Characterization of Pak2p, a Pleckstrin Homology Domain-containing, p21-activated Protein Kinase from Fission Yeast*

p21-activated kinases (PAKs) bind to and are activated by Rho family GTPases such as Cdc42 and Rac. Since these GTPases play key roles in regulating cell polarity, stress responses, and cell cycle progression, the ability of PAK to affect these processes has been examined. We previously showed that fission yeastpak1 + encodes an essential protein that affects mating and cell polarity. Here, we characterize a secondpak gene (pak2 +) fromSchizosaccharomyces pombe. Like the Saccharomyces cerevisiae proteins Cla4p and Skm1p, fission yeast Pak2p contains an N-terminal pleckstrin homology domain in addition to a p21-binding domain and a protein kinase domain that are common to other members of the PAK family. Unlike pak1 +,pak2 + is not essential for vegetative growth or for mating in S. pombe. Overexpression of the wild-typepak2 + allele suppresses the lethal growth defect associated with deletion of pak1 +, and this suppression requires both the pleckstrin homology- and the p21-binding domains of Pak2p, as well as kinase activity. A substantial fraction of Pak2p is associated with membranous components, an association mediated both by the pleckstrin homology- and by the p21-binding domains. These results show that S. pombeencodes at least two pak genes with distinct functions and suggest that the membrane localization of Pak2p, directed by its interactions with membrane lipids and Cdc42p, is critical to its biological activity.

Signal transduction by small GTPases of the Rho family has become the focus of intensive investigation in recent years. Rho GTPases (in mammals, Cdc42, Rac, and Rho) regulate a variety of signaling pathways in higher eukaryotes, including those that affect the organization of cortical actin, stress responses, gene transcription, and cell cycle progression (1). In budding and fission yeasts, Cdc42p has been implicated in the generation and maintenance of cell polarity. However, the mechanisms by which Cdc42p and other Rho family GTPases regulate these processes are poorly understood.
Several candidate effectors for Rho GTPases have been identified recently, including a variety of protein kinases, GTPase activating proteins, and other proteins that lack obvious enzymatic function. Among these candidate effectors, the p21-activated serine/threonine kinases (PAKs) 1 are currently the best characterized (2). PAKs bind to activated (GTP-bound) forms of Cdc42 and Rac, but not Rho, and are activated as a result of this binding. Like Cdc42 and Rac, PAKs stimulate stress-activated protein kinases (3)(4)(5)(6). Moreover, dominant interfering forms of PAK block stress-activated protein kinase activation by Cdc42 and Rac1, suggesting that PAKs may be effectors for these GTPases in stress signaling pathways. The role of PAKs in mediating other Cdc42/Rac functions, such as cytoskeletal rearrangements and G 1 progression, is less clear. In mammalian fibroblasts, expression of activated Pak1 induces changes in focal complex formation and F actin distribution similar but not identical to those induced by Rac1 and Cdc42 (7,8). Therefore, PAKs may activate multiple pathways, including those that affect gene transcription and those that affect actin dynamics.
To better define the targets and functions of PAK, we and others have examined PAKs in yeast. Three PAK homologs have been characterized in budding yeast. The PAK homolog Ste20p is required for pheromone-induced G 1 arrest and activation of the transcription factor Ste12p as well as for certain specialized morphogenic processes, such as formation of mating projections and filamentous growth (9 -11). A second PAK-like kinase, Cla4p, has been shown to regulate formation of the bud neck and may also have other functions that are partially redundant with Ste20p (12). A third kinase of this class, Skm1p, is dispensable for normal growth and mating but yields a multibudded phenotype when overexpressed, suggesting a role in cell morphogenesis (13). Unlike Ste20p or mammalian PAKs, both Cla4p and Skm1p contain a pleckstrin homology (PH) domain near the N terminus. The functional significance of this extra domain is not known, although it seems likely that it serves to target these kinases to appropriate locations within membrane structures.
We recently characterized a PAK-like protein in S. pombe. This kinase, known as Pak1p or Shk1p, is essential for viability and mating and may play important roles in the generation and/or maintenance of cell polarity (14,15). Expression of pak1 ϩ in S. cerevisiae cells deleted for STE20 relieves the mating defect associated with this mutation, indicating that fission yeast pak1 ϩ is to some degree functionally homologous to budding yeast STE20. During the isolation of pak1 ϩ clones, we uncovered a second PAK homolog (pak2 ϩ ). The protein encoded by this gene, like budding yeast Cla4p and Skm1p, contains an N-terminal PH domain. In this study, we report the biochemical and functional characterization of Pak2p.
Isolation of cDNA Clones-The primers 5Ј-CGG-GAT-CCG-TNG-CNA-T(A/C/T)A-A(A/G)C-A(A/G)A-TGA-A-3Ј and 5Ј-CGG-AAT-TCN-GGN-GG(C/T)-TCN-CC(C/T)-TC(A/G/T)-ATC-AT-3Ј were used to amplify an ϳ500-base pair pak-like gene fragment from an S. pombe cDNA library, and this PCR product was used to probe the library as described previously (14). Most positive clones contained partial or full-length pak1 ϩ , but one contained a related sequence that we termed pak2 ϩ . The sequence of this clone appeared to represent a partial cDNA, since the predicted protein product lacked a recognizable N-terminal p21-binding domain, which is present in Pak1p and other PAKs. While we attempted to isolate a full-length cDNA, the complete pak2 ϩ gene was uncovered during the course of a comprehensive international S. pombe genome sequencing effort. We obtained this clone (ICRFc1F5) from the Imperial Cancer Research Fund Reference Library (21). The genomic pak2 ϩ clone was identical to our cDNA clone, but it encoded in addition a potential N-terminal PH domain and a p21-binding domain.
Plasmids-The pak2 ϩ coding sequence (ϳ1.8 kilobases (kb)) was amplified by PCR using Deep Vent polymerase (New England Biolabs) from a cosmid (ICRFc1F5) containing the genomic pak2 ϩ clone. The PCR product was subcloned as a SmaI/Acc65I fragment into the hemagglutinin (HA) epitope-tagging vector pJ3H (22). This vector was then cut with Acc65I, blunt-ended, and then digested with SalI and subcloned into SalI-, SmaI-cut pREP3X (a conditional S. pombe expression vector) (23). Epitope-tagged Pak1p constructs have been described previously (14). Pak2p point mutants (Pak2p H136L,H140L and Pak2p K343A) were created using the Altered States mutagenesis system (Promega) and confirmed by sequence analysis. An N-terminal truncation mutant that deletes the PH domain (⌬PH-Pak2p) was created by digesting pJ3H-pak2 with NheI and XbaI and then religating the linearized 4.8-kb fragment. The pak2 insert from the resulting plasmid, which encodes HA-tagged Pak2p lacking the N-terminal PH domain, was then subcloned into pREP3X as described above.
Southern Blot Analyses-Genomic S. pombe DNA was isolated using the protocol of Hoffman and Winston (25). Five g of DNA was digested with EcoRV, electrophoresed, and transferred to a nylon membrane and cross-linked using UV irradiation. Filters were hybridized overnight at 65°C in 0.5 M NaPO 4 , pH 7.2, 7% SDS, 1% bovine serum albumin with full-length (1.6-kb) [ 32 P]dCTP-labeled pak2 ϩ cDNA and washed twice with 0.2ϫ SSC, 0.2% SDS at 65°C. Autoradiograms were obtained by exposing the blots to Kodak XAR film with intensifying screens at Ϫ70°C for 24 h.
Protein Kinase Assays-S. pombe cells were transformed with pREP3X-HA-Pak2 and grown overnight in leucine-deficient, thiamine-deficient liquid medium. Cells were lysed in Nonidet P-40 lysis buffer (50 mM Tris-HCl, pH 8.0, 137 mM NaCl, 10% glycerol, 1% Nonidet P-40, 50 mM NaF, 10 mM ␤-glycerol-phosphate, containing 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride, and 10 g/ml aprotinin) using glass beads, and HA-tagged Pak2p was immunoprecipitated using the monoclonal anti-HA antibody 12CA5 (Babco). The immunoprecipitates were washed twice with lysis buffer and then once with kinase buffer (50 mM Hepes, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol). The immunoprecipitates were then resuspended in 25 l of kinase buffer containing 20 M unlabeled ATP and 5 Ci of [␥ 32 P]ATP, 5 g of myelin basic protein (Sigma), with or without 10 g of GTP-loaded Gst-Cdc42 (14), and incubated at 30°C for 30 min. Reactions were terminated by adding 5 l of 6ϫ SDS-polyacrylamide gel electrophoresis sample buffer and then boiled for 5 min. Samples were separated by 10% SDSpolyacrylamide gel electrophoresis. The gels were either transferred to a polyvinylidine difluoride membrane and immunoblotted with 12CA5 or dried on Whatman 3 MM paper and autoradiographed using Kodak XAR film.
Interaction Trap Binding Assays-Full-length wild-type and mutant pak2 ϩ cDNAs were subcloned into the bait vector pEG202 (26). Wildtype and mutant forms of S. pombe cdc42 cDNA were subcloned into the activation domain plasmid pJG4 -5 (26). All cdc42 alleles encode a mutation (C189S) that eliminates the C-terminal prenylation site to promote nuclear entry and reduce toxicity (14). Bait vectors, activation domain vectors, and a lacZ reporter were co-transformed into EGY48, and transformants were selected on dextrose-containing media lacking uracil, leucine, and histidine. Three independent colonies for each transfection were analyzed for reporter activation. The colonies were replica-plated to galactose-containing media to induce production of the bait protein, and the colonies were assayed for lacZ activity (19) and for growth on media lacking leucine.
Cell Fractionation-10 ml of CHP428 cells transformed with pREP3X-Pak2 expression vectors were grown at 30°C in leucine-deficient EMM, with or without 5 M thiamine, to an OD of 1.0. The cells were collected, washed with water, and resuspended in 0.2 ml of lysis buffer (0.8 M sorbitol, 1 mM EDTA, 10 mM MOPS, pH 7.0) with protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin). The cells were lysed on ice by vortexing with 400 -500-m acid-washed glass beads. Cell lysates were spun at 500 ϫ g for 4 min at 4°C, and the pellets were resuspended in the same volume of lysis buffer as the supernatants. The 500 ϫ g supernatants were then spun at 10,000 ϫ g for 10 min at 4°C, and the pellets were resuspended in the same volume of lysis buffer as the supernatants. Equal volumes of each fraction were loaded onto a 10% SDS-polyacrylamide gel for immunoblot analysis.
S. pombe Suppression Assays-The pak1 ϩ gene was deleted in a CHP428/429 diploid as described previously (14), except that a his7 ϩ marker was used in place of ura4 ϩ . Diploids were then transformed with pREP3X encoding various forms of Pak1p or Pak2p as described under "Results." The diploids were sporulated, and at least 10 tetrads were dissected for each construct. Spores isolated in this manner were scored for growth on YES media and on EMM media lacking histidine. Random spores, generated by glusulase treatment of tetrads, were also analyzed for growth on EMM media lacking histidine in the presence or absence of thiamine. pak1::his7 cells supported by pREP3X-based plasmids were tested for mating competence by plating to malt extract media with an h ϩ (CHP428) or h Ϫ strain (CHP429) of S. pombe.
Overexpression Phenotypes in S. pombe-pak1 ϩ , pak2 ϩ , or CLA4, under the control of the thiamine-repressible nmt1 ϩ promoter, were introduced into S. pombe CH428 cells. To observe the growth of cells expressing these alleles, transformants were successively subcultured twice to selective plates minus thiamine. To observe the overexpression phenotype, 5-ml cultures were grown in EMM plus adenine in the presence of 15 M thiamine to mid-log phase (OD Ϸ 1) and washed twice with 5 ml of water. The cells were then resuspended in 5 ml at OD ϭ 0.5 and grown to mid-log to late log phase (OD ϭ 1-2). These cultures were diluted to OD ϭ 0.2 and grown overnight two successive times. Final cultures ranged from early to late log phase (OD ϭ 0.2-2).
Microscopic Analyses-Cells were examined and photographed using an Olympus BH-2 (S. pombe) or a Leica Aristoplan (S. cerevisiae) microscope equipped with differential interference optics. Pictures were taken using Kodak TMAX 400 ASA film.

Isolation of a Second PAK Homolog from Fission Yeast-
During the course of cloning pak1 ϩ from a S. pombe cDNA library (14), we isolated a second, pak-like gene fragment. We were unable to obtain a full-length clone of this gene from this library, but the entire gene was subsequently sequenced as part of a comprehensive fission yeast genome sequencing effort. The sequence of the complete gene clearly indicates that it is a member of the PAK family; thus we have termed it pak2 ϩ (Fig. 1A).
The pak2 ϩ gene potentially encodes a protein of 589 amino acids. The domain structure of the predicted Pak2p protein closely resembles that of two S. cerevisiae proteins, Cla4p and Skm1p (2). Like these molecules, the predicted Pak2p protein contains a PH domain at the N terminus, followed by a p21binding domain, and a serine/threonine protein kinase domain that occupies the C-terminal half of the protein (Fig. 1B).
Unlike many other PAK homologs, S. pombe Pak2p does not contain an acid-rich region C-terminal to the p21-binding domain or notable proline-rich regions. Outside of the conserved PH, p21-binding, and kinase domains, Pak2p is not similar to any other members of the PAK family.
Pak2p Kinase Activity Is Stimulated by Cdc42p-To confirm that Pak2p is, in fact, a p21-activated kinase, we immunoprecipitated HA-tagged Pak2p from S. pombe that overexpress this protein. Pak2p immunoprecipitates were then assayed for protein kinase activity in the presence of GTP-loaded, wildtype Cdc42p; a constitutively active (V12) form; a dominantnegative (N17) form; or a control protein. Both wild-type and constitutively activate Cdc42p markedly stimulated Pak2p ki- nase activity, as reflected in increased phosphorylation of myelin basic protein in an in vitro kinase assay and in the generation of "upshifted" Pak2p bands on anti-HA immunoblot (Fig.  2). In contrast, dominant negative Cdc42p had no effect on basal kinase activity. These data confirm that Pak2p, like Pak1p and other members of the PAK family of protein kinases, can be activated by Cdc42p.
Pak2p Interacts with Activated Cdc42p-To establish that the p21-binding domain in Pak2p is functional, we carried out interaction-trap assays (27). Co-expression of LexA-Pak1p or LexA-Pak2p with activation domain-Cdc42p resulted in activation of the lacZ reporter (Table I) as well as LEU2 (not shown). These reporters were not activated when the yeast were grown on dextrose-containing media, which prevents induction of Cdc42p expression from the activation domain vector (not shown), or when a nonspecific bait (RBP7) was used and the yeast were grown on galactose-containing media ( Table I). The interaction among Pak1p, Pak2p, and Cdc42p was greatly increased when a constitutively activated (V12) form of Cdc42p was used and abolished when an inactive (N17) (not shown) or an effector domain (A35) mutant was used. These results indicate that, like Pak1p, Pak2p can interact with Cdc42p within cells. Furthermore, the Pak2p/Cdc42p interaction is enhanced by mutations that activate Cdc42p and requires an intact effector domain. As with mammalian Pak1 (7), replacing two conserved histidine residues within the p21-binding domain of Pak2p ablates binding to Cdc42p. Neither a mutation that abolishes kinase activity nor deletion of the PH domain, however, measurably affects binding of Pak2p to Cdc42p in the interaction trap assay. In all cases, expression of both baits and interactors was confirmed by immunoblot (Fig. 3).
pak2 ϩ Is Not an Essential Gene-We and others have previously established that pak1 ϩ is essential for viability and is likely to play a role in mating as well (14,15). To address the functions of the related pak2 ϩ gene, we used the ura4 ϩ marker to disrupt the coding sequence of one copy of this gene in diploid cells. After confirming the presence of the disrupted allele by Southern blot (not shown), the appropriate diploids were sporulated and tetrad dissections were carried out. Of 14 tetrads dissected, 12 gave rise to four viable colonies, while the remaining two yielded three colonies. Those tetrads that yielded four colonies displayed 2:2 segregation for uracil auxotrophy. These results indicate that pak2 ϩ is not required for viability. Haploids containing the disrupted pak2 ϩ allele were isolated and tested for growth rate at 20, 30, and 36°C; viability under stress (e.g. post-heat shock, osmotic shock, nutrient-deprivation); morphology; and mating competence (not shown). None of these tests revealed a phenotype for pak2 disruption.
pak2 ϩ Overexpression Can Partially Replace pak1 ϩ Func-tion-To assess whether pak2 ϩ and pak1 ϩ have overlapping functions, we first asked if high level expression of pak2 ϩ could allow cells to remain viable in the absence of pak1 ϩ . pak1 ϩ /pak1::his7 diploids were transformed with a thiamineregulated expression plasmid (pREP3X) bearing either no insert or various forms of fission yeast pak1 ϩ or pak2 ϩ , or budding yeast CLA4. The cells were made to sporulate by plating on maltose, and the resulting tetrads were then dissected and plated to complete medium. Diploid cells transformed with the control vector gave rise to two viable and two dead spores (12 tetrads examined), as expected for disruption of an essential gene such as pak1 ϩ . Expression of pak1 ϩ from the pREP3X vector rescued pak1:his7 ϩ haploids; three or four spores germinated from most of the tetrads derived from pak1 ϩ /pak1::his7 ϩ (pREP3X-pak1 ϩ ) diploids, one or two of which bore the pak1::his7 ϩ allele. That some spores failed to germinate can probably be attributed to the low copy number of the episomal expression vector pREP3X, which may not be present in sufficient numbers to distribute to all four spores. ⌬N-pak1, which lacks the p21-binding domain and which we have previously shown to encode an unstable form of Pak1p (14), also supported the growth of pak1::his7 ϩ haploids but only when expressed at high levels (i.e. on medium lacking thiamine) (Fig. 4B).
We found that overexpression of pak2 ϩ can also suppress the   growth defect associated with loss of pak1 ϩ function (Fig. 4B). Suppression was not evident when thiamine was included in the growth medium (which reduces expression from the pREP3X vector), indicating that viability in pak1⌬ cells requires high levels of pak2 ϩ expression (data not shown). This conclusion is also supported by the finding that cells transformed with pREP81-pak2 ϩ , which has a mutated thiamineresponsive promoter that yields expression levels that are only about 1% of pREP3X-pak1 ϩ (23), did not suppress pak1 disruptants (data not shown). Interestingly, suppression of pak1⌬ lethality was not apparent in cells expressing pak2 mutants lacking either functional PH or p21-binding domains or in cells expressing kinase-dead pak2 (Fig. 4B), despite adequate expression of these mutant proteins (Fig. 4C). These results indicate that the PH and p21-binding domains, as well as intact kinase function, are required for pak2 ϩ to suppress the growth defect associated with loss of pak1 ϩ . Like pak2 ϩ , CLA4 from S. cerevisiae can also suppress the pak1 ϩ deletion. We have not tested if the PH and p21-binding domains of Cla4p are required for this suppression.
We further assessed the ability of pak2 ϩ to replace pak1 ϩ function using mating and sporulation as end points. Both cdc42 ϩ and pak1 ϩ have been implicated in the mating process in fission yeast; cells overexpressing defective forms of these genes fail to mate normally (14,15). When grown on thiaminedeficient media, pak1⌬ cells supported by Pak1p, ⌬N-Pak1p, Pak2p, or Cla4p mate normally with wild-type partners, and sporulation of diploids is not notably affected (Table II). Support by plasmids encoding ⌬N-Pak1p, Pak2p, and Cla4p required high levels of expression, since mating function was poor in cells grown in the presence of thiamine. In contrast, in crosses involving two mutant cells (i.e. crosses in which both partners bear the pak1::his7 allele, supported by a pREP3Xbased plasmid), only Pak1p and ⌬N-Pak1p, but not Pak2p or Cla4p, efficiently supported mating. Therefore, high levels of Pak2p can fully compensate for Pak1p's vegetative functions but only partially compensate for its sexual functions. These results suggest that Pak1p and -2p can share certain substrates that are required for viability, but not for mating.
Subcellular Localization-Because we found that the PH and p21-binding domains of Pak2p are required for its ability to suppress the growth defect of pak1⌬ cells, we assessed the subcellular distribution of wild-type Pak2p and of Pak2p mutants lacking a functional PH or p21-binding domain. Exponentially growing cells expressing these proteins were fractionated into low speed (500 ϫ g) and high speed (10,000 ϫ g) pellets and supernatants, which were analyzed by immunoblot for the presence of Pak2p (Fig. 5). The low speed pellet (P1) largely consists of unbroken cells, nuclei, and large organelles, while the low speed supernatant (S1) contains both cytosol and membranous elements (28). The high speed pellet (P2) is comprised largely of plasma membrane, and the high speed supernatant contains cytosol and secretory granules. Wild-type Pak2p is found almost equally in all four fractions, indicating that both membrane-bound and cytosolic pools of this protein are present in the cell. Surprisingly, both the ⌬PH and the p21-binding domain Pak2p mutants had similar subcellular distributions, indistinguishable from that of the wild-type enzyme. However, a mutant lacking function of both domains (⌬PH-Pak2p H136L,H140L), was primarily cytosolic (Ͼ80%, as assessed by densitometry of immunoblots). These data indicate that the PH and p21-binding domains both contribute to the association of

Pak2p with membranes.
Trans-species Complementation-As Pak2p is structurally homologous to Cla4p, we asked whether it is also functionally homologous. Like pak2 ϩ , overexpression of CLA4 was able able to suppress the loss of viability associated with pak1 ϩ deletion in S. pombe (Fig. 4), as well as restore mating function (Table  II). To more fully explore this issue, we did the reciprocal experiments in cla4⌬ or ste20⌬ S. cerevisiae cells. In ste20⌬ cells, expression of transgenic STE20 or pak1 ϩ restores mating, as shown previously (Fig. 6 and Ref. 14). Interestingly, we found that multicopy expression of CLA4 also restores mating competence, suggesting that high levels of Cla4p can share overlapping functions in a morphogenic pathway (12). Although CLA4 and STE20 have previously been shown to share overlapping functions in a morphogenic pathway (12), to our knowledge CLA4 has not previously been shown to substitute for STE20 in the mating pathway. Unlike CLA4, expression of S. pombe pak2 ϩ did not restore mating to ste20⌬ cells. Protein expression for all transgenes was verified by immunoblot (not shown); thus, Pak2p failed to complement Ste20p and Cla4p despite adequate expression.
In budding yeast, simultaneous deletion of STE20 and CLA4 results in a vegetative growth defect (12). To analyze whether pak1 ϩ or pak2 ϩ is able to complement this defect, the diploid strain YEL252, heterozygous for ste20⌬::URA3/STE20 cla4⌬::TRP1/CLA4, was transformed with the multicopy plasmid pYES2 carrying either pak1 ϩ or pak2 ϩ , respectively, or, as positive controls, with the multicopy plasmids pVTU-STE20 (9) and pRL21 (10) carrying either STE20 or CLA4, respectively. The diploid transformants were then sporulated and dissected for tetrad analysis. We dissected more than 30 asci for each transformant. We found no viable double mutant spores in tetrads transformed with either pak1 ϩ or pak2 ϩ but found 16 double mutant spores transformed with the STE20 plasmid and 13 double mutant spores transformed with the CLA4 plasmid. Putative pak1 ϩ or pak2 ϩ transformants predicted to be double ste20⌬ cla4⌬ mutants germinated and produced aberrantly shaped buds but finally lysed after prolonged incubation.
To further examine whether pak1 ϩ or pak2 ϩ is able to complement the growth defect of ste20 cla4 double mutant cells, we transformed strain M96007-1D, which is deleted for CLA4 and carries a temperature-sensitive mutation in STE20. Cells were transformed at the permissive temperature (room temperature) and then analyzed for growth at the restrictive temperature (37°C). We found that mutant cells transformed with the STE20 or CLA4 plasmids were able to grow at 37°C, whereas transformants with pak1 ϩ or pak2 ϩ plasmids stopped growing and underwent lysis at this temperature (not shown).
Taken together, these data indicate that while pak1 ϩ com- FIG. 5. Immunoblot analysis of fractionated S. pombe cell lysates. 10 ml of CHP428 cells transformed with the indicated pREP3X-Pak2 expression vectors were grown at 30°C, in the presence (not shown) or absence of thiamine, in leucine-deficient media to mid/late log phase. The cells were then lysed with glass beads, and protein extracts were spun at 500 ϫ g to produce pellet (P) and supernatant (S) fractions; the supernatant fractions were subsequently spun at 10,000 ϫ g. Equal volumes of protein extract were fractionated by 10% SDS-polyacrylamide gel electrophoresis, transferred to a polyvinylidine difluoride membrane, and immunoblotted with anti-HA antibodies. This result is indicative of two independent experiments. 6. Complementation of the mating defect of ste20⌬ mutant  cells. a, the ste20⌬ Mata strain YEL206 was transformed with the indicated plasmids. Patches of independent transformants were grown on glucose (2%) medium under selective conditions (left). Mating was tested by replica-plating the patches onto a lawn of the MAT␣ tester strain DC17 on permissive YEP-galactose (4%) medium. Mating was allowed to proceed overnight at 30°C, and the formation of diploids was then analyzed by replica-plating the cells onto selective glucose (2%) medium (right). b, YEL206 cells, transformed with YES2-pak1 ϩ or -pak2 ϩ , were grown in galactose-containing, uracil-deficient medium for 24 h. Lysates from these cells were immunoprecipitated with polyclonal anti-HA antibodies (Santa Cruz Biotechnology), assayed for (auto)protein kinase activity, and immunoblotted with monoclonal anti-HA (12CA5) antibodies.

TABLE II
Suppression of the mating defect in pak1⌬ cells pak1 ϩ /pak1::his7 diploids were transformed with pREP3X vectors bearing the indicated genes (shown in parentheses). Following selection of transformed cells on leucine-deficient media, the diploids were sporulated on malt extract plates for 2 days at 29°C. Asci were treated with glusulase to release free spores, which were then germinated on histidine-, leucine-deficient media. Several independent clones, of mating type h ϩ , from among the resulting colonies (which bear the disrupted pak1::his7 allele, supported by a pREP3X-based plasmid) were then tested for mating with a wild-type, h Ϫ (CHP429) strain and also for mating with the indicated h Ϫ mutant strains on S. pombe minimal medium (without nitrogen) plates with appropriate auxotrophic supplements in the presence or absence of thiamine. The number of zygotes, asci, and unmated cells were determined microscopically. The mating values represent averages from three independent determinations. plements the mating defect associated with loss of STE20, neither pak1 ϩ nor pak2 ϩ complements the cytokinesis function of CLA4 or the essential functions that STE20 and CLA4 share during vegetative growth.
Overexpression Phenotype-We have previously shown that overexpression of wild-type or kinase-dead pak1 ϩ in S. pombe results in aberrant morphology, characterized by maldistribution of cortical actin (14). To determine the effects of pak2 ϩ , we transformed wild-type cells with either an empty expression vector or vectors bearing pak1 ϩ , pak2 ϩ , or CLA4. Cells transformed with empty vector display normal morphology (Fig. 8A). pak2 ϩ overexpression is characterized by the frequent (Ͼ60%) appearence of irregular, dysmorphic cells (Fig. 8C). The morphology of these cells is similar, but distinguishable from, that of cells overexpressing pak1 ϩ (Fig. 8, compare B and C). Low level expression of pak2 ϩ does not notably affect morphology (Fig. 8D) Overexpression of CLA4 also alters the morphology of S. pombe, giving rise to small cells, which occasionally bulge at the septum (Fig. 8E). These morphological defects are only apparent when the cells are grown in thiamine-deficient media, indicating that these phenotypes are due to overexpression of the indicated genes (not shown). Therefore, overexpression of pak2 ϩ , like pak1 ϩ , affects morphogenesis in fission yeast, consistent with the proposed role of Ste20-like kinases in actin polarization pathways in this organism and in other eukaryotes (2). DISCUSSION In this report, we characterize a second Pak-like protein in S. pombe. Like Pak1p, Pak2p binds to activated Cdc42p and affects cell morphology when overexpressed. However, unlike pak1 ϩ , pak2 ϩ is not essential for cell viability. pak2-null cells exhibit normal vegetative growth rates, mating capability, and survival under stress. High level expression of pak2 ϩ suppresses lethality associated with loss of pak1 ϩ , indicating a degree of functional overlap between these proteins. The PH and p21-binding domains, as well as kinase function, are required for the growth of cells lacking the pak1 ϩ gene. Transspecies complementation tests in S. cerevisiae show that pak2 ϩ suppresses neither the mating nor the morphologic defects in STE20-and CLA4-null cells, respectively, nor the growth defect in cells deleted for both STE20 and CLA4. Thus, Pak2p, while similar in structure to budding yeast Cla4p, appears to be functionally distinct. Pak2p may instead represent a homolog of the recently described S. cerevisiae protein Skm1p (13). Like pak2 ϩ in S. pombe, disruption of SKM1 in S. cerevisiae evokes no obvious phenotype, and overexpression leads to aberrant morphology. These results indicate that Pak2p may regulate distinct signaling pathways from Cdc42p in fission yeast.
In budding yeast, the p21-binding domain of Ste20p is dispensable for mating but is required for the vegetative functions of this protein and for filamentous growth (10,29). Mutant Ste20p, lacking the p21-binding domain, is diffusely localized throughout the cell, whereas wild-type Ste20p is concentrated at emerging bud tips during vegetative growth and at shmoo tips in cells arrested with ␣-factor. The kinase activity of the wild-type and mutant forms of Ste20p is comparable, indicat- ing that binding to Cdc42p is not absolutely required for activation of this kinase in vitro. These findings suggest that binding to Cdc42 is important for proper localization of Ste20p and that proper localization is required for the biological function of Ste20p during vegetative growth but not for mating.
In contrast to Ste20p in budding yeast, the p21-binding domain of Pak1p is dispensable both for vegetative growth and for mating in S. pombe. Does this indicate that the consequences of the Cdc42p/Pak1p interaction in fission yeast differ from that of Cdc42p/Ste20p in S. cerevisiae? Unlike wild-type Pak1, ⌬N-Pak1 (which lacks the p21-binding domain) is only capable of supporting the growth of such cells when expressed at high level (i.e. on thiamine-deficient media). This behavior can be interpreted in several ways. For example, at high levels of expression, small amounts of ⌬N-Pak1p may localize to the required site(s) of action, i.e. to areas occupied by Cdc42p, although no targeting sequence is present. This scenario is especially plausible if deletion of the N terminus activates Pak1p, since N-terminal truncations of S. cerevisiae Ste20p activate that kinase (9,30). Second, Pak1p may contain a second, weak, and hitherto unrecognized, binding site for Cdc42p. This site may contribute to physiologically significant amounts of Cdc42p binding under conditions of overexpression. However, we have been unable to identify such a cryptic binding site either by in vitro or by interaction trap binding assays. Finally, unlike Ste20p in S. cerevisiae, Pak1p localization may not be required for its vegetative functions in S. pombe. In this interpretation, the requirement for high level expression may merely reflect the fact that the ⌬N-Pak1p protein is unstable (14).
Overexpression of pak2 ϩ partially suppresses the lethal growth defect associated with loss of pak1 ϩ . Unlike Pak1p, the N terminus of Pak2p, containing the PH and the p21-binding domains, are required for this suppression. PH domains are often found in proteins that affect cytoskeletal function and are thought to direct proteins to certain membrane compartments through their ability to bind phospholipids (31)(32)(33). The removal of this domain from Pak2 may cause this protein to mislocalize from its normal site(s) of action, perhaps diminishing or precluding interaction with activators, such as Cdc42p, or key downstream targets for its kinase activity. It is interesting to note that the PH domain of Cla4p has also recently been shown to be required for its biological function (34). While ⌬PH-Pak2p can bind Cdc42p in an interaction trap assay (Table I), it is possible that the association of these two proteins is diminished in vivo in S. pombe cells. This interpretation implies that, unlike Pak1p, localization of Pak2p may be crucial for at least some of its biological functions. Disabling the p21binding domain of Pak2p has a similar effect; the Cdc42-binding minus mutant also fails to suppress the growth defect of pak1-null cells. This finding supports the notion that proper subcellular localization is important to the function of Pak2p. Perhaps Pak2p is inactive in the absence of association with Cdc42p, whereas Pak1p has a higher basal kinase activity. In this scenario, the presence of the PH and p21-binding domains are necessary because together they direct interaction with Cdc42p, which is required for Pak2p activation.
cdc42 is an essential gene in fission yeast, involved in the regulation of cell polarity as well as the mating response. The effectors used by Cdc42p to regulate these processes are unknown. Because Pak1p and Pak2p bind to and are activated by Cdc42p, they represent attractive candidates in mediating one of more of Cdc42p's functions. Although the terminal phenotype of pak1-null cells resembles those seen in cells bearing cdc42 disruption, overexpression of pak1 ϩ does not suppress the lethal growth defect associated with loss of cdc42 (14). The failure of pak1 ϩ to suppress loss of cdc42 could mean that Cdc42p has other effectors that are required for viability or that Pak1p must be activated by Cdc42p to function effectively. We believe the first of these scenarios to be more likely than the second, since a pak1 mutant, encoding a constitutively activated form of Pak1p (Pak1p T539E ), also fails to suppress loss of cdc42 ϩ . 3 Could Pak2p represent such a second effector? Like pak1 ϩ , overexpression of alleles encoding either wild type or an activated form of Pak2p fail to suppress loss of cdc42 ϩ . 3 Although we have not tested whether Pak1p and Pak2p together mediate those functions of Cdc42p that are required for viability, this seems unlikely, since both proteins would be expected to compete for common binding sites on Cdc42p and thus would not be activated simultaneously. Therefore, we expect that S. pombe Cdc42p has other effectors that are vital to its vegetative functions. A similar conclusion has been drawn regarding Cdc42p function in S. cerevisiae. In this organism, ste20 cla4 mutant cells still nucleate and polarize actin (12). In addition, schmoo morphogenesis and actin repolarization occur normally in cells expressing a Ste20p mutant unable to bind Cdc42p (10,29). Mammalian Cdc42 and the related protein Rac are also thought to have multiple effectors, including PAKs and other p21-binding proteins (2), Rho kinase (35,36), and proteins of unknown function, such as POR1 (37). Consistent with this notion, expression of activated forms of human or rat Pak1 in murine fibroblasts recapitulate some, but not all, of the functions of Cdc42 and Rac (7,8). In addition, mutant forms of Cdc42 and Rac1 that are unable to bind PAK nevertheless retain the ability to reorganize actin and to drive cell cycle progression (35,36). Thus, the weight of evidence from these diverse organisms is that Cdc42 and Rac, like the related protein Ras, recruit an array of effector molecules that mediate their biological functions.
Among the known functions of fission yeast Cdc42p are activation of the Byr2p/Byr1p/Spk1p mating cascade and maintenance of cell polarity. Does either Pak1p or Pak2p have a role in mediating these signaling pathways? In budding yeast, the Pak1p homolog Ste20p activates a protein kinase cascade, apparently through a direct phosphorylation of Ste11p (18). In mammalian cells, PAKs 1 and 3 have also been shown to stimulate stress-activated kinase cascades, although direct activation of Ste11p homolog (i.e. a Mekk) has not been demonstrated (4 -6, 38). The existing evidence for a connection between PAKs and the mating protein kinase cascade in fission yeast is indirect, based on the ability of inactivated forms of these kinases to interfere with mating. Although we have not yet been able to document an interaction between either Pak1p or Pak2p and the Mekk homolog Byr2p, 3 we consider it likely that Pak1p and/or Pak2p mediates activation of the mating kinase cascade by Cdc42p, given the high degree of similarity in the design of the mating molecular machinery in S. cerevisiae and S. pombe. Whether fission yeast PAKs also mediate other functions of Cdc42p is less clear, but, in the case of Pak1p, this kinase clearly has important targets other than Byr2p, since disruption of pak1 results in lethality, not just sterility. While these vital Pak1p targets are unknown, some are likely to be shared in common with Pak2p, since overexpression of the latter kinase suppresses pak1⌬ lethality. Since overexpression of Pak1p or Pak2p in S. pombe yields a characteristic morphologic defect, it is possible that these kinases play a role in Cdc42p-mediated actin reorganization.
The relationship between fission yeast Pak1p and -2p is reminiscent of, but not identical to, that of budding yeast Ste20p and Cla4p. Both pairs of kinases share certain functions in mating and cell morphogenesis but also play unique roles in the cell. Given the inability of Pak2p to complement Cla4p function, it is possible that Pak2p is more closely related to the recently characterized, PH-domain containing kinase, Skm1p (13). Clearly, the multiplicity and conservation of Pak-like kinases among diverse eukaryotes as well as their effects on regulating such basic processes as mating and actin polarization warrant further investigation to define the precise functions and targets of these proteins.