The p21-activated Protein Kinase-related Kinase Cla4 Is a Coincidence Detector of Signaling by Cdc42 and Phosphatidylinositol 4-Phosphate*

Signal transduction pathways that co-regulate a given biological process often are organized into networks by molecules that act as coincidence detectors. Phosphoinositides and the Rho-type GTPase Cdc42 regulate overlapping processes in all eukaryotic cells. However, the coincidence detectors that link these pathways into networks remain unknown. Here we show that the p21-activated protein kinase-related kinase Cla4 of yeast integrates signaling by Cdc42 and phosphatidylinositol 4-phosphate (PI4P). We found that the Cla4 pleckstrin homology (PH) domain binds in vitro to several phosphoinositide species. To determine which phosphoinositides regulate Cla4 in vivo , we analyzed phosphatidylinositol kinase mutants ( stt4, mss4, and pik1 ). This indicated that the plasma membrane pool of PI4P, but not phosphatidylinositol 4,5-bisphosphate or the Golgi pool of PI4P, is required for localization of Cla4 to sites of polarized growth. A combination of the Cdc42-bind-ing and PH domains of Cla4 was necessary and sufficient for localization to sites of polarized growth. Point mutations affecting either domain impaired the ability of Cla4 to regulate cell morphogenesis and the mitotic exit network (localization of Lte1). Therefore, Cla4 must retain the ability to NheI sites was inserted, resulting in pRS314Cla4-GFP. Prior studies have shown that this C- terminal fusion is fully functional (39). To generate GFP-tagged Cla4 truncation mutants, we used site-directed mutagenesis to introduce NheI sites at desired locations in the CLA4 gene on pRS314Cla4. The GFP coding region flanked by NheI sites then was inserted into the NheI site of the engineered CLA4 gene cassette. In so doing we created the following pRS314Cla4-GFP derivatives: Cla4-(1–246)-GFP (GFP inserted after amino acid 246) and Cla4-(1–527)-GFP (GFP inserted after amino acid 527). In-frame internal deletion mutations in CLA4 were generated by using site-directed mutagenesis to introduce two NheI sites into pRS314Cla4 at locations flanking the regions of Cla4 to be deleted. Following NheI digestion, purification, and ligation of the cleaved plasmid, we generated mutants lacking sequences encoding the PH domain or Cdc42-binding domain (PBD). To generate GFP-tagged forms of these internal Cla4 deletion mutants, we replaced most of the CLA4 coding region of pRS314Cla4-GFP (contained on a XhoI fragment) with XhoI fragments from the PH or PBD deletion mutant plasmids. Amino acid substitutions within the phosphoinositide- and Cdc42-binding surfaces of the PH domain and PBD of Cla4 were intro- duced by site-directed mutagenesis (K71N, K96N, and K99N in the PH domain or H192L and H195L in the PBD). The entire coding and promoter regions of all cloning and mutagenesis products were the PH constructed

from mitosis (mitotic exit network (MEN)) 1 (11)(12)(13), and stimulates docking and fusion of the lysosome-like vacuole (14 -16). Phosphoinositide signaling has several functions in yeast. The type III PI 4-kinase Stt4 and the PI4P 5-kinase Mss4 are required for polarization of the cortical actin cytoskeleton and regulation of a protein kinase C-activated MAP kinase cascade (17)(18)(19)(20). The Golgi-localized type III PI 4-kinase Pik1 regulates trafficking through the Golgi apparatus (19,21). Yeast cells also express a type II PI 4-kinase (Lsb6) that localizes to the plasma membrane and vacuole (22,23); however, the function of this enzyme is unknown. The PI 3-kinase encoded by VPS34 is required for protein sorting in the endocytic pathway (24).
Because Cdc42 and phosphoinositide signaling pathways coregulate overlapping sets of biological processes, cells must integrate these pathways into networks. Signal integration could occur directly via multidomain proteins that are activated simultaneously by two or more signals (25,26), thereby functioning as coincidence detectors. An example is N-WASp, which stimulates the nucleation of actin filament assembly by the Arp2/3 complex in mammalian cells (27). N-WASp is activated synergistically upon binding the Rho-type GTPase Cdc42 and phosphatidylinositol 4,5-bisphosphate (PI4,5P2 (27)).
The mechanisms that integrate Cdc42 and phosphoinositide signaling in yeast are unknown. However, several proteins have the potential to act directly as coincidence detectors of these two signaling pathways. One such protein is the PAKrelated protein kinase Cla4. Cla4 possesses in its non-catalytic N-terminal region a Cdc42-binding domain (PBD) and a plekstrin homology (PH) domain (28), which potentially binds phosphoinositides. Other PH domain-containing proteins that interact with Cdc42 include Cdc24 (the guanine nucleotide exchange factor for Cdc42 (29)) and the related proteins Boi1/2 (unknown function (30)). Whether the PH domain of Cdc24 binds phosphoinositides is unknown. However, the PH domain of Boi1 does bind PI4,5P2 (31). Furthermore, Boi1 PH domain point mutants impaired in PI4,5P2 binding fail to function or localize to the bud cortex (31). The Cdc42 scaffold protein Bem1 possesses a phox homology domain that binds phosphatidylinositol 3-phosphate (PI3P (32)), which has a prominent role in protein trafficking to the vacuole (24).
Here we have determined whether the PAK-related kinase Cla4 functions as a coincidence detector of signaling by Cdc42 and phosphoinositides. By analyzing phosphoinositide binding to the Cla4 PH domain in vitro, Cla4 localization in phosphatidylinositol (PI) kinase mutants, and Cla4 function in vivo, we provide evidence indicating that Cla4 must bind both Cdc42 and PI4P, as expected for a coincidence detector of these two signaling pathways.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Enzymes used for recombinant DNA manipulations were purchased from commercial sources and used as recommended by the manufacturers. The complete CLA4 gene, including the promoter and 3Ј-untranslated region, was amplified using primers (forward 5Ј-CCACACCCCGGGGAGAAATCACCACTGGAAAC-3Ј and reverse 5Ј-CCACACGGGCCCTAGAAGCTGAAGCATGGACG-3Ј), digested with ApaI and SmaI (sites underlined) to generate appropriate ends, and cloned into pRS314 or pRS315 (38) to generate pRS314Cla4 and pRS315Cla4. To tag the Cla4 protein at its C terminus with green fluorescent protein (GFP), we used site-directed mutagenesis to generate a unique NheI site just prior to the CLA4 stop codon of pRS314Cla4 into which the GFP coding region flanked by NheI sites was inserted, resulting in pRS314Cla4-GFP. Prior studies have shown that this Cterminal fusion is fully functional (39). To generate GFP-tagged Cla4 truncation mutants, we used site-directed mutagenesis to introduce NheI sites at desired locations in the CLA4 gene on pRS314Cla4. The GFP coding region flanked by NheI sites then was inserted into the NheI site of the engineered CLA4 gene cassette. In so doing we created the following pRS314Cla4-GFP derivatives: Cla4-(1-246)-GFP (GFP inserted after amino acid 246) and Cla4-(1-527)-GFP (GFP inserted after amino acid 527). In-frame internal deletion mutations in CLA4 were generated by using site-directed mutagenesis to introduce two NheI sites into pRS314Cla4 at locations flanking the regions of Cla4 to be deleted. Following NheI digestion, purification, and ligation of the cleaved plasmid, we generated mutants lacking sequences encoding the PH domain or Cdc42-binding domain (PBD). To generate GFP-tagged forms of these internal Cla4 deletion mutants, we replaced most of the CLA4 coding region of pRS314Cla4-GFP (contained on a XhoI fragment) with XhoI fragments from the PH or PBD deletion mutant plasmids. Amino acid substitutions within the phosphoinositide-and Cdc42-binding surfaces of the PH domain and PBD of Cla4 were introduced by site-directed mutagenesis (K71N, K96N, and K99N in the PH domain or H192L and H195L in the PBD). The entire coding and promoter regions of all cloning and mutagenesis products were sequenced.
GFP fusions to the PH domain or PBD of Cla4 were constructed as follows. The parent plasmid was pUG36, a single copy vector in which the MET25 promoter drives expression of EGFP. Sequences encoding the Cla4 PH domain (amino acids 62-181) were amplified as a fragment flanked by HindIII and XhoI sites and cloned into pUG36 cleaved with HindIII and XhoI. A second Cla4 PH domain (flanked with HindIII sites and lacking a C-terminal stop codon) was then inserted into pUG36-Cla4PH at the HindIII site. The stop codon between the two PH do-mains was eliminated by site-directed mutagenesis so that the resultant plasmid pUG36-GFP-2XCla4PH expressed a GFP fusion linked to two tandem copies of the Cla4 PH domain. A similar procedure was used to generate a plasmid (pUG36 -2XPLC␦PH) that expressed GFP fused to two tandem copies of the PH domain of human PLC␦ (residues 1-175 (40)) in which the two PH domain-encoding sequences were flanked by HindIII/XhoI and HindIII sites. To generate a TRP1-marked plasmid expressing the GFP-PLC-PH domain fusion, we subcloned a SacI/KpnI fragment from pUG36 -2XPLC␦PH into pRS314. The Cla4 PBD (amino acid residues 184 -242) was cloned into pUG36 at the ClaI site. Similar methods were used to generate a construct containing two PBDs fused to GFP. To create a plasmid expressing GFP fused to both the PH and PBD of Cla4, we amplified sequences encoding residues 62-242 of Cla4 flanked by two ClaI sites and inserted this fragment in the ClaI site of pUG36, yielding pUG36-Cla4PHPBD. Site-directed mutagenesis of the above-noted plasmids was used to generate mutant forms of the Cla4 PH domain defective in phosphoinositide binding (K71N, K96N, and K99N).
GST fusion proteins containing the Cla4 or PLC␦ PH domain were generated using the pGEX-3X (Amersham Biosciences) vector. The Cla4 PH domain (amino acids 62-181) and the human PLC␦ PH domain (amino acids 1-175) were amplified flanked by appropriate restriction sites and inserted into pGEX-3X at the EcoRI site or BamHI/EcoRI sites, respectively. For expression of the untagged Cla4 PH domain, the Cla4 PH domain-encoding sequence was amplified with NdeI/BglII ends and cloned into the pET11a vector (Stratagene) digested with NdeI/BamHI.
Escherichia coli and Yeast Strains-E. coli strains (Stratagene) used were DH5␣, XL1-Blue for mutagenesis experiments, and BL21DE3 cells for GST fusion protein expression. All Saccharomyces cerevisiae strains used in this study are listed in Table I. The yeast strains generated in this study were derivatives of W303. A 3X-GFP-TRP1 cassette was integrated in W303 at the LTE1 locus using pBJ1368 linearized with EcoRV as described (41). Clones selected on media lacking tryptophan were screened by microscopy for cortical GFP fluorescence. The genes encoding CLA4 and SWE1 were disrupted with PCR-generated kanamycin and hygromycin resistance gene cassettes, respectively. Disruptions were confirmed by PCR of yeast genomic DNA and by evaluation of bud morphology.
Protein Purification, Protein-Lipid Overlays, and Surface Plasmon Resonance Experiments-Recombinant GST fusion proteins were purified from E. coli. Briefly, fresh BL21DE3 transformants expressing GST, GST-Cla4PH, or GST-PLC␦PH were grown overnight in a 2-ml LB-amp culture. One liter of LB-amp was inoculated and grown to an A 600 of 0.6 -0.8. Isopropyl-␤-D-thiogalactoside was added to a final concentration of 0.2 mM to induce fusion protein expression, and the cultures were grown for 16 h at 25°C. Cells were collected by centrifugation for 20 min at 1000 ϫ g. Pellets were suspended in 25 ml of ice-cold Buffer A (50 mM Hepes, 0.5 mM EDTA, 300 mM NaCl, 1 mM dithiothreitol, pH 7.4) containing protease inhibitors. Cells were lysed by one round of freeze-thawing in the presence of 2 mM dithiothreitol, treated with lysozyme and Triton X-100 (0.5% final concentration), and incubated on ice for 2-4 h. Samples were clarified by centrifugation at 20,000 ϫ g for 1 h at 4°C. The resultant supernatant fraction was filtered before overnight incubation with 2 ml of glutathione-Sepharose 4B (Amersham Biosciences). The supernatant/glutathione-agarose mixture was poured into a column and washed with Buffer B (20 mM Hepes, 25 mM NaCl, 2% glycerol, 1 mM dithiothreitol, pH 7.4) with and then without Triton X-100. GST fusion proteins were eluted with buffer (50 mM Tris-HCl, 150 mM NaCl, 25 mM glutathione, pH 8), and fractions were collected. Relevant fractions were identified by measuring protein concentrations, which were pooled and dialyzed overnight. GST fusion proteins were frozen in liquid nitrogen and stored at Ϫ80°C in glycerolcontaining buffer. Protein-lipid overlays were performed as described previously (42). Briefly, nitrocellulose membranes spotted with 100 pmol of various purified phospholipids (Echelon, Inc.) were incubated in blocking buffer (TBST (50 mM Tris HCl, pH 7.5, 150 mM NaCl, 0.1% Tween containing 3% fatty acid-free bovine serum albumin (Sigma)) for at least 1 h prior to addition of GST fusion proteins. Recombinant GST, GST-Cla4PH, or GST-PLC␦PH (0.5 g/ml) was added to the blocking buffer, and blots were incubated overnight at 4°C. Blots were washed with TBST (6 times over 30 min) and incubated with rabbit polyclonal antibody to GST for 2 h at room temperature (1:10,000 dilution in TBST). After blots were washed, they were incubated with goat anti-rabbit horseradish peroxidase-conjugated secondary antibody (1:2000 dilution in TBST; ICN Biomedicals, Inc.). Protein binding was detected by enhanced chemiluminescence (Amersham Biosciences).
Surface plasmon resonance experiments were performed as described (32). Dioleoylphosphatidylcholine vesicles lacking or containing 3% (mol/mol) PI3P, PI4P, PI4,5P2, or PI3,5P2 were prepared and immobilized onto an L1 sensor chip using a Biacore X instrument. For purified GST-PH fusion proteins tested, protein concentration was determined by comparison with bovine serum albumin standards on SDS-PAGE, and for the binding analysis of purified untagged PH domains, protein concentration was determined by absorbance at 280 nm.
Microscopy-To localize GFP-tagged proteins expressed in yeast, we transformed relevant yeast strains with plasmids expressing GFPtagged proteins of interest or fused GFP to the protein of interest expressed from the chromosome. Plasmid-containing cells were grown overnight in 2 ml of selective media, diluted the next day, and grown to early log phase. For experiments using temperature-sensitive PI kinase mutants, a portion of the culture was shifted from permissive (26°C) to non-permissive (37°C) temperature. After incubation for 1-2 h, cells were collected by brief centrifugation and visualized by fluorescence microscopy with a temperature-controlled microscope as described below. For studies involving yeast strains expressing mutant forms of the Stt4 or Mss4 PI kinases, we grew wild type and mutant strains in media containing 1 M sorbitol to provide osmotic support and prevent cell lysis as described previously (19,43). Fluorescence images of live cells were collected using a cooled CCD camera (DAGE) mounted on a temperature-controlled Olympus IX 70 inverted microscope equipped with a UPlanApo 100ϫ objective. Images were acquired and analyzed with NIH Image 1.62.
Immunoblotting-Yeast cell extracts were prepared by lysing cells with NaOH (44). Proteins were resolved by SDS-PAGE and transferred to nitrocellulose. Blots were blocked, probed with a rabbit polyclonal antibody against GFP (1:2000 dilution (45)) and goat anti-rabbit horseradish peroxidase-conjugated secondary antibody (ICN Biomedicals, Inc.), and visualized by enhanced chemiluminescence detection (Amersham Biosciences).

RESULTS
To determine how Cdc42 and phosphoinositide signaling in yeast is integrated into networks that co-regulate specific cell biological processes, we focused on Cla4, a PAK-related protein kinase that is a direct effector of Cdc42 (Fig. 1A). The rationale for doing so was severalfold. In contrast to other proteins such as Boi1 that bind Cdc42 by poorly understood mechanisms, Cla4 uses its conserved PBD to bind Cdc42 (46). Furthermore, Cla4 probably binds Cdc42 by a mechanism highly similar to that used by mammalian PAK1 because the residues in PAK1 that bind Cdc42 are conserved in Cla4 (Fig. 1B) (47). Likewise, structural studies of PH domain-containing proteins have identified sequence features directly involved in binding the phosphoinositide head group (48). Several of these residues are conserved in the Cla4 PH domain (Fig. 1C), although it is unknown whether the Cla4 PH domain binds phosphoinositides. Finally, phosphoinositide kinase mutants are available to determine whether Cla4 is regulated by phosphoinositide production (17)(18)(19)(20)(21).
The PH Domain of Cla4 Binds Phosphoinositides in Vitro-We determined initially whether the Cla4 PH domain binds phosphoinositides in vitro by performing protein-lipid overlay assays and surface plasmon resonance experiments. For protein-lipid overlay assays, we incubated nitrocellulose membranes containing spots of pure phosphoinositides or other lipids with purified GST fusion proteins containing the PH domain of Cla4 or the PH domain of PLC␦ as a positive control. Membranes were washed and probed with an anti-GST antibody to detect binding of GST-tagged PH domains. The results indicated that the GST-Cla4 PH fusion protein binds most strongly to the monophosphorylated phosphoinositides PI3P, PI4P, and PI5P, and to PI3,5P2 ( Fig. 2A). The Cla4 PH domain bound less strongly to several other lipids including PI4,5P2. At an equivalent concentration, a GST fusion containing the PLC␦ PH domain bound strongly to PI4,5P2 and weakly to other lipids ( Fig. 2A), confirming the specificity of this PH domain (49). GST without a PH domain did not bind (data not shown). Therefore, the PH domain of Cla4 is similar to many PH domains that bind in vitro to several phosphoinositide species (50).
To complement qualitative measures of phosphoinositide-PH domain interaction determined by lipid overlay assays, we performed quantitative surface plasmon resonance (SPR) experiments to determine the affinity of the Cla4 PH domain for various phosphoinositides. These experiments used dioleoylphosphatidylcholine vesicles containing 3% (mol/mol) PI3P, PI4P, PI3,5P2, or PI4,5P2 immobilized on sensor chips. Responses of sensor chips were recorded following exposure to  various concentrations of the purified His-tagged Cla4 PH domain. The resultant dose-response curves indicated that the Cla4 PH domain binds with highest affinity to PI4,5P2 (K d ϭ 20 M) and PI3,5P2 (K d ϭ 21 M) (Fig. 2B). By comparison, the PLC␦ PH domain was found to bind PI4,5P2 with a K d of 1.7 M (data not shown). We also found that the Cla4 PH domain binds PI3P or PI4P with lower affinity (K d Ͼ 100 M; Fig. 2B). The relative affinity of the Cla4 PH domain for various phosphoinositides suggested by the results of SPR experiments apparently differed from that suggested by protein-lipid overlay assays. This may have occurred because protein-lipid overlay assays use pure phosphoinositides adsorbed to blots, which may result in lipid head group packing that differs significantly from what occurs in biological membranes or lipid vesicles used for SPR experiments. Regardless of the explanation, the results of both approaches indicated that the PH domain of Cla4 is like most PH domains, which bind phosphoinositides in vitro with relatively low affinity and selectivity (5). Therefore, the in vitro binding data were insufficient to indicate which phosphoinositide species are likely to regulate Cla4 in vivo. To address this question, it was necessary to determine whether Cla4 localization or function was altered when the levels of specific phosphoinositides were manipulated in vivo.
Localization of Cla4-GFP Requires the Plasma Membrane Pool of PI4P Produced by the PI 4-Kinase Stt4 -Phosphoinositide levels in yeast can be altered either by inactivating temperaturesensitive forms of specific PI kinases or by overexpressing phos-phoinositide phosphatases such as the SacI or Inp families from inducible promoters. We employed the former approach because we found that the latter required long induction times (Ͼ4 h), which could affect Cla4 localization indirectly.
To determine which phosphoinositide species regulate Cla4 in vivo, we analyzed cells expressing temperature-sensitive forms of the essential PI 4-kinases Stt4 or Pik1 or the essential PI4P 5-kinase Mss4. Stt4 and Mss4 synthesize, respectively, PI4P and PI4,5P2 on the plasma membrane and are required for polarization of the actin cytoskeleton (17)(18)(19)(20). Pik1 synthesizes the Golgi pool of PI4P and promotes secretion but apparently is uninvolved in cell or actin polarization (19,21). Accordingly, we assessed Cla4 function in wild type cells, stt4, mss4, or pik1 temperature-sensitive mutants at permissive (25°C) and non-permissive (38°C) temperatures by examining the localization of a functional Cla4-GFP fusion protein to sites of polarized growth, where Cdc42 is concentrated. Because Cla4-GFP localization is independent of the actin cytoskeleton (39), any effects of inactivating PI kinases on Cla4-GFP localization could not be caused indirectly by depolarization of the actin cytoskeleton that occurs in stt4 and mss4 mutants.
The results indicated that polarized localization of Cla4-GFP requires the PI 4-kinase Stt4 but not the PI4P 5-kinase Mss4 or the Golgi-localized PI 4-kinase Pik1. In the stt4 temperaturesensitive mutant incubated at permissive temperatures, fluorescence microscopy of living cells indicated that Cla4-GFP was polarized normally to the bud (88% of the cells; Fig. 3, A and B).
In contrast, at non-permissive temperatures we found that Cla4-GFP was poorly polarized or diffusely localized in 62% of budded stt4 mutant cells (Fig. 3, A and B). Partial mislocalization of Cla4-GFP might occur because Stt4 is not fully inactivated, as suggested by observations indicating that total PI4P levels are reduced 2.5-fold in stt4 mutants (20). In contrast, Cla4-GFP localization was polarized normally (polarized localization in 84 -91% of budded cells) in wild type cells, mss4, or pik1 temperature-sensitive mutants incubated at permissive or non-permissive temperatures (Fig. 3, C and D, and data not shown). Inactivation of Mss4 at non-permissive temperature was confirmed by the loss of plasma membrane localization of a PI4,5P2-dependent marker (GFP fused to two copies of the PH domain of PLC␦; GFP-PLC␦PH 2 in Fig. 3E). Likewise, inactivation of Pik1 at non-permissive temperature was confirmed by the appearance of fragmented vacuoles (data not shown (20)). Therefore, localization of Cla4-GFP requires the plasma membrane pool of PI4P generated by the PI 4-kinase Stt4, but not the Golgi pool of PI4P generated by the PI 4-kinase Pik1, or the plasma membrane pool of PI4,5P2 synthesized by Mss4.
The loss of Cla4 localization in stt4 mutants might have been caused by impairment of Cdc42 activation. This possibility existed because Cdc24, the exchange factor for Cdc42 (29), possesses a PH domain that potentially binds phosphoinositides. However, this possibility was excluded because localization of GFP-tagged Gic1, a PBD domain-containing protein whose localization and function depends on its ability to bind Cdc42 (34,51), was nearly normal in stt4 mutants incubated at non-permissive temperature (Fig. 3, F and G).
In contrast to Cla4, plasma membrane localization of Rom2, the only previously described direct target of the PI4P 3 PI4,5P2 pathway, requires the sequential action of Stt4 and Mss4 to generate PI4,5P2 (20). Synthesis of PI4P may be sufficient for Cla4 localization because the PH domain of this PAK homolog may have higher affinity for PI4P than does the PH domain of Rom2. Alternatively, the PH and Cdc42 binding domains of Cla4 may function synergistically to localize the protein to sites of polarized growth. According to this model, a combination of the Cdc42 and PH domains, but neither domain alone, is sufficient for polarized localization. This hypothesis was tested as described below.
The PH Domain and PBD of Cla4 Are Necessary and Sufficient for Localization to Sites of Polarized Growth-To identify domains of Cla4 that are necessary and sufficient for polarized localization of this PAK family kinase, we constructed a series of GFP-tagged Cla4 truncation and deletion mutants expressed from the CLA4 promoter on single copy plasmids. Results of fluorescence microscopy of cells expressing these GFP fusions indicated that localization determinants in Cla4 were present within its N-terminal non-catalytic domain containing the PH and Cdc42 binding (PBD) domains. Polarized localization was observed with a Cla4 mutant lacking the kinase domain (Cla4-(1-527)-GFP; Fig. 4A) or with a Cla4 mutant containing the N-terminal domain plus the PH domain and PBD (Cla4-(1-246)-GFP; Fig. 4A). In contrast, internal deletion mutants lacking either the PH or PBD of Cla4 (⌬54 -181 or ⌬181-247, respectively) were expressed well but localized to the cytoplasm (Fig. 4, A and C).
We subsequently analyzed GFP-tagged full-length Cla4 bearing amino acid substitutions in the PH domain or PBD that disrupt phosphoinositide or Cdc42 binding, respectively. These experiments used structural information from the G, quantitation of GFP-Gic1 localization in wild type (STT4) and stt4-4 mutants at permissive (25°C) and non-permissive (37°C) temperature. Cells (n Ͼ 150/group) were scored for proper localization of GFP-Gic1 to sites of polarized growth (tips and sides of small-and medium-sized buds) or improper localization (cytoplasmic or partial localization); the difference between stt4-4 cells incubated at 25 and 38°C was not statistically significant.
PBD and PH domains of other proteins. We introduced three lysine 3 asparagine substitutions (K71N, K96N, and K99N) in or near the putative phosphoinositide binding pocket of the PH domain in full-length Cla4. Basic residues in the corresponding sites of other PH domains bind the negatively charged head groups of phosphoinositides (48). Similar substitutions of basic residues in other PH domains disrupt phosphoinositide binding without affecting PH domain folding (reviewed in Refs. 5 and 52). We confirmed that these three substitutions disrupted the ability of the Cla4 PH domain to bind phosphoinositides by performing surface plasmon resonance experiments with sensor chips adsorbed with vesicles containing 3% (mol/mol) PI4,5P2 (Fig. 2B). When this mutant PH domain was present in full-length Cla4-GFP, the protein was expressed at normal levels but localized to the cytoplasm (Fig. 4, B and C).
We next investigated the role of the Cla4 PBD by introducing two substitutions (H192L and H195L) within the Cdc42-binding surface of full-length Cla4-GFP. As indicated by sequence alignments (Fig. 1 and data not shown), equivalents of these histidine residues in Cla4 are present in the PBDs of many Cdc42 targets. Furthermore, mutagenesis studies of vertebrate PAK1, ACK, and WASP and the NMR structure of Cdc42bound PAK1 have demonstrated that equivalent substitutions of these histidine residues greatly impair Cdc42 binding (47,(53)(54)(55). Indeed, we found that this mutant form of Cla4-GFP (H192L and H195L) mislocalized to the cytoplasm even though it was well expressed (Fig. 4, B and C). Therefore, the results of these point mutagenesis studies indicated that Cla4 localization requires residues implicated directly in binding phosphoinositides and Cdc42, supporting the hypothesis that Cla4 is a coincidence detector of these two signaling pathways.
To determine whether the PH and/or PBD of Cla4 is sufficient for polarized localization, we fused each domain or a combination of both to GFP. We found that a GFP fusion containing the PH domain and PBD (GFP-PH-PBD) was polarized (Fig. 5A); some cytoplasmic and vacuolar membrane localization was also observed (Fig. 5A). In contrast, the same fusion protein carrying substitutions that impair phosphoinositide binding to the PH domain (K71N, K96N, and K99N; indicated as GFP-mPH-PBD in Fig. 5, A and B) was well expressed but localized to the cytoplasm and nucleus (Fig. 5, A and B). A GFP fusion protein containing either one or two copies of the PBD was well expressed but failed to polarize or localize to the plasma membrane (Fig. 5, A and B). GFP fused to one copy of the Cla4 PH domain localized weakly to the plasma membrane (Fig. 5A). However, a GFP fusion containing two copies of the PH domain associated efficiently with the plasma membrane, although in a non-polarized distribution (Fig. 5A). Association with other intracellular compartments such as the Golgi appa-  (K71N, K96N, and K99N) and a wild type PBD; PBD (residues 184 -242); PBD 2 (two copies of the PBD); PH (residues 62-184); PH 2 (two copies of the PH domain). B, expression of GFP-tagged proteins detected by anti-GFP immunoblotting of crude yeast lysates.  (K71N, K96N, and K99N) and PBD (H192L and H195L). C, expression of mutant proteins detected by anti-GFP immunoblotting of crude yeast lysates. ratus or vacuole was not evident (Fig. 5A). Results obtained with these Cla4 fragments therefore suggested that the PH domain of Cla4 provides affinity for the plasma membrane and that Cdc42 binding refines Cla4 localization to sites of polarized growth where Cdc42 is concentrated. Other domains such as the proline-rich region adjacent to the PBD probably also refine Cla4 localization to sites of polarized growth, as indicated by the more highly polarized distribution of Cla4-(1-527)-GFP (Fig. 4) relative to GFP-PH-PBD (Fig. 5). This hypothesis is consistent with findings indicating that the proline-rich domain of Cla4 binds an SH3 domain of Bem1 (56), a polarity scaffold in the Cdc42 pathway.
Cla4 Function Requires Binding to Cdc42 and Phosphoinositides-To determine whether Cla4 functions as a coincidence detector for phosphoinositide and Cdc42 signaling, we analyzed the ability of Cla4 point mutants defective in phosphoinositide or Cdc42 binding to rescue phenotypes of cla4⌬ mutants. This approach was used instead of examining PI kinase mutants, which have pleiotropic phenotypes that could indirectly affect processes regulated by Cla4. The hallmark phenotype of cells lacking Cla4 is the exaggerated apical growth of the bud, resulting in highly elongated cells (28). This phenotype is thought to occur in part because septin assembly at the incipient bud site is defective (28), which activates a morphogenesis checkpoint that inhibits cyclin-cdk (Clb-Cdc28) complexes required to drive G 2 /M progression (57). A second phenotype of cla4⌬ cells is a delay in mitotic exit (11)(12)(13). Mitotic exit is impaired because recent studies have indicated that Cla4 is required for phosphorylation of Lte1 (11,12), a guanine-nucleotide exchange factor for the Tem1 GTPase in the MEN (58). Cla4 is also required for Lte1 function as assessed by restricted localization of this exchange factor to the cortex of the bud (11). Indeed, in cla4⌬ cells Lte1 localizes to the cortex of mother and bud, resulting in a delay in mitotic exit (11,12).
By using rescue of these cla4⌬ mutant phenotypes as assays, we determined whether expression of mutant Cla4-GFP or untagged mutant Cla4 defective in Cdc42 or phosphoinositide binding was functional. As a control, we showed that cla4⌬ cells expressing wild type Cla4-GFP from a single copy plasmid and the CLA4 promoter had normal bud morphology in 95% of the cells (Fig. 6A). In contrast, expression of mutant Cla4-GFP in which the PH domain or PBD was deleted failed to rescue the elongated bud phenotype (only 40% of cells had normal buds, similar to empty vector controls; Fig. 6A), as reported previously (46). These loss of function phenotypes were not due to complete loss of Cla4 kinase activity, because the PH⌬ mutant has been shown previously to have detectable (albeit reduced) kinase activity, whereas the PBD⌬ mutant has elevated kinase activity relative to wild type Cla4 (46).
We also determined whether expression of Cla4 point mutants defective in Cdc42 or phosphoinositide binding could rescue the bud morphology phenotype of cla4⌬ cells. Expression of full-length Cla4-GFP bearing substitutions in the PH domain (K71N, K96N, and K99N) that strongly impaired phosphoinositide binding failed to rescue the elongated bud phenotype of cla4⌬ cells (only 50% of cells had normal buds; Fig. 6A). Similarly, expression of full-length Cla4-GFP bearing substitutions within the Cdc42-binding surface of the PBD (H182L and H185L) only partially rescued the elongated bud phenotype of cla4⌬ cells (75% of cells had normal buds; Fig. 6A). The partial function of this protein may be due to high constitutive kinase activity, as suggested by the effects of identical mutations in human PAK1 (53). Nevertheless, the fact that this mutant form of Cla4 is mislocalized and fails to rescue fully the bud morphology phenotype of cla4⌬ cells demonstrates that Cdc42 binding is required for full function of Cla4 in vivo.
Similar results were obtained when Lte1 localization was used as an indicator of Cla4 function in the mitotic exit network. These experiments used Lte1 fused to three tandem copies of GFP (Lte1-3XGFP) as an indicator of Lte1 function and MEN activity, as substantiated by previous studies (11). As expected, Lte1-3XGFP localized exclusively to the cortex of the bud in wild type cells (Fig. 6B). To examine Lte1-3XGFP localization in cells lacking Cla4, it was necessary to prevent the formation of elongated buds. This was accomplished by deleting SWE1, which disables the morphogenesis checkpoint pathway (57). As shown previously (11), cla4⌬ swe1⌬ double mutants exhibited normal bud morphology but mislocalized Lte1-3XGFP to the cortex of both the mother and bud (Fig. 6B). Lte1-3XGFP localization was rescued when wild type Cla4 was expressed from a plasmid (Fig. 6, A and B). In contrast, Lte1-FIG. 6. Functional analysis of Cla4 mutants defective in Cdc42 or phosphoinositide binding. A, quantitation of bud morphology and mitotic exit phenotypes of cla4⌬ cells expressing the indicated cla4 alleles from the CLA4 promoter on single copy plasmids. Bud morphology (wild type ovoid buds versus elongated buds characteristic of cla4⌬ mutants) were scored visually (n Ͼ 400/ group; error bars indicate S.E.). Mitotic exit phenotypes were scored by analyzing the localization of a functional Lte1-3XGFP fusion protein in cla4⌬ swe1⌬ cells expressing the indicated CLA4 alleles from the CLA4 promoter on a single copy plasmid (n ϭ 100 -200 cells/group; error bars indicate S.E.). The SWE1 gene was deleted to disable the morphogenesis checkpoint that otherwise elicits an elongated bud morphology in cla4⌬ cells. Wild type CLA4 function in mitotic exit was indicated by localization of Lte1-3XGFP restricted to the bud cortex, whereas loss of CLA4 function in mitotic exit was indicated by localization of Lte-3XGFP on the cortex of mother and bud. B, representative fluorescence images of Lte1-3XGFP localization in wild type cells or in cla4⌬ swe1⌬ cells expressing the indicated CLA4 alleles from single copy plasmids.
3XGFP localization was not rescued when the phosphoinositide binding-defective point mutant (K71N, K96N, and K99N) was expressed (Fig. 6, A and B). Expression of the PBD point mutant (H191L and H192L) partially rescued Lte1 localization (Fig. 6A), indicating that Cdc42 binding is required for full Cla4 function in the MEN pathway. Again, the residual ability of this Cla4 mutant to localize Lte1 may be due to its elevated kinase activity, as suggested by biochemical studies of the identical substitutions in mammalian PAK1 (53). Therefore, we concluded that Cla4 must bind both Cdc42 and phosphoinositides to localize to sites of polarized growth and function as a regulator of cell morphogenesis and mitotic exit. Accordingly, this PAK-family kinase appears to be a coincidence detector that directly integrates signaling by Cdc42 and phosphatidylinositol 4-phosphate. These findings also provide the first evidence indicating that mitotic exit requires phosphoinositide signaling.

DISCUSSION
Previous studies have shown that Cla4, one of three PAKrelated protein kinases of the yeast S. cerevisiae, is a direct effector of Cdc42 in pathways that control cell polarity and morphogenesis, cytoskeletal organization, mitotic exit, and docking and fusion of the lysosome-like vacuole (7, 11-13, 16, 28, 39, 46, 56, 59). It also has been established that Cla4 deleted of its Cdc42-binding domain (PBD) or PH domain is non-functional (46). However, the function of the Cla4 PH domain has remained unknown.
Results described herein demonstrate that several phosphoinositides can bind the Cla4 PH domain and that Cdc42 and phosphoinositide binding is required for Cla4 localization and function. As discussed below, these and other findings indicate the following: 1) Cla4 is an effector of the plasma membrane pool of PI4P; 2) Cla4 functions as a coincidence detector of signaling Cdc42 and PI4P; and 3) PI4P synthesis promotes mitotic exit.
Cla4 Is an Effector of the Plasma Membrane Pool of PI4P-We have found that the Cla4 PH domain binds in vitro to several phosphoinositide species with low to moderate affinity (ϳ10 -100 M), similar to many PH domains (50). The ability of most PH domains to bind relatively non-selectively to several phosphoinositide species generally is of unknown biological significance. However, our findings demonstrate that the ability of Cla4 to bind phosphoinositides is critical for the localization of this protein kinase to sites of polarized growth and for its function as a regulator of cell morphogenesis and mitotic exit. One function of the Cla4 PH domain may be to provide affinity for the plasma membrane because GFP fused to two copies of this domain associates efficiently and specifically with the plasma membrane. The preference of the Cla4 PH domain for the plasma membrane is striking because the membranes of other organelles such as the Golgi apparatus and vacuole contain various phosphoinositide species (60). Association of the Cla4 PH domain with the plasma membrane is not due its deposition via bulk membrane flow. This possibility is excluded by previous observations showing that other PH domains associate specifically with other membranes or intracellular organelles in yeast such as the Golgi apparatus (43).
Our analysis of PI kinase mutants indicates that Cla4 localization requires the plasma membrane pool of PI4P, but not the plasma membrane pool of PI4,5P2 or the Golgi pool of PI4P. This conclusion is supported by the observation that Cla4-GFP is mislocalized when the plasma membrane-associated PI 4-kinase Stt4 is inactivated, but not when the Golgi-localized PI 4-kinase Pik1 or the plasma membrane-bound PI4P 5-kinase Mss4 is inactivated. PI4P rather than PI4,5P2 appears to be the lipid that regulates Cla4 localization in vivo, even though the Cla4 PH domain binds PI4,5P2 more tightly than PI4P in vitro. We suggest this because Cla4 localization is impaired in stt4 but not mss4 mutants even though PI4,5P2 levels are reduced more dramatically in mss4 than in stt4 mutants incubated at non-permissive temperatures (19,43). The results of in vivo studies may differ from those obtained from in vitro studies of phosphoinositide binding because factors present in cells may increase the specificity or affinity of the Cla4 PH domain for PI4P relative to PI4,5P2, or because PI4P is more available than PI4,5P2 where Cla4 localizes.
Plasma membrane-selective localization of the Cla4 PH domain is not determined strictly by affinity for PI4P because the specificity of a given PH domain for various phosphoinositides in vitro does not predict where the PH domain localizes in cells. This point is illustrated by the observation that the PH domains of Cla4 and FAPP1 bind in vitro with some preference for PI4,5P2, yet the Cla4 PH domain associates with the plasma membrane and the FAPP1 PH domain localizes to the Golgi (43). Furthermore, whereas Cla4 localization to the plasma membrane requires PI4P synthesized by the plasma membrane-bound PI 4-kinase Stt4, the PH domain of FAPP1 requires PI4P synthesized by the Golgi-associated PI 4-kinase Pik1. Indeed, the mechanisms determining the membrane or organelle specificity of various PH domains is an interesting problem remaining to be solved. Nevertheless, the results described herein are sufficient to indicate that Cla4 is the first known effector of the plasma membrane pool of PI4P generated by Stt4. In contrast, Rom2, a guanine nucleotide exchange factor for Rho1 that regulates yeast cell wall biosynthesis and integrity (61), is an effector of PI4,5P2 (20).
Cla4 Is a Coincidence Detector of Cdc42 and PI4P-In confirmation of previous studies (46), we have found that Cla4 localization and function requires Cdc42 binding. An internal deletion that removes the Cla4 PBD or amino acid substitutions in the Cla4 PBD analogous to those in mammalian PAK that strongly impair Cdc42 binding results in cytoplasmic localization of Cla4 and significant loss of function. In contrast to Cla4, localization of PAKs in mammalian cells is not strictly dependent upon the ability to bind Cdc42 (53). Instead, PAK recruitment to focal contacts apparently involves binding to other proteins such as the Cdc42/Rac exchange factor PIX and the paxillin kinase linker PKL (62).
The role of the Cla4 PBD appears to differ from that of the PH domain. Although the PH domain provides affinity for the plasma membrane, the PBD does not because GFP fused to one or two copies of this domain fails to localize to the plasma membrane or sites of polarized growth. These observations suggest that Cla4 uses its PH domain to associate with the plasma membrane such that subsequent binding to Cdc42 restricts Cla4 localization to sites of polarized growth and stimulates Cla4 kinase activity. Thus, an inability to associate with the plasma membrane may explain why Cla4 mutants lacking a PH domain have reduced kinase activity (46). Taken together, our findings and those of previous investigations indicate that Cla4 requires input from both the phosphoinositide and Cdc42 signaling pathways, the hallmark of a coincidence detector (25).
Because two of the three PAK homologs (Cla4 and Skm1) in budding yeast and the Cla4 homolog required for virulence of Candida albicans (63) possess Cdc42-binding and PH domains, these protein kinases have the potential to function as coincidence detectors of Cdc42 and phosphoinositide signaling. In contrast, Ste20 (the third PAK homolog of budding yeast) lacks a PH domain, suggesting that phosphoinositides do not bind to and regulate this kinase directly. Assuming this is true, the distinct biological functions of Ste20 and Cla4 might reflect, in part, differences in the input pathways that regulate these kinases. For example, Ste20 regulates mitotic exit independently of Lte1 (the exchange factor for the Tem1 GTPase that triggers the mitotic exit network (11)), whereas Cla4 promotes mitotic exit by regulating Lte1 phosphorylation and localization (11,12). However, the specificity of the kinase domains Ste20 and Cla4 is known to determine many of the distinct biological functions of these proteins (64), although this conclusion remains to be investigated in the context of mitotic exit regulation. Thus, the biological specificity of Ste20 and Cla4 may also be determined in some cases by differences in the input mechanisms that localize and/or activate these kinases at sites of polarized growth.
Regulation of Mitosis by Phosphatidylinositol 4-Phosphate Synthesis-PI4P synthesis has been implicated in a yeast cell cycle checkpoint that delays cell division (65). The checkpoint delays cell division until the mitotic spindle is correctly positioned along the mother-bud axis. The checkpoint is essential for the viability of mutant cells lacking components of the dynein-dynactin complex, which have defects in spindle positioning. Therefore, inactivation of genes encoding components of this checkpoint is lethal in cells lacking dynein (65). Indeed, a partial loss of function mutation in STT4 is lethal in dynactin complex mutants (65), suggesting that the PI 4-kinase encoded by STT4 is required for checkpoint function. However, an alternative explanation is that a spindle-positioning defect is lethal in combination with a defect in mitotic exit. We suggest this because our results show that Stt4 regulates Cla4 localization and that Cla4 promotes mitotic exit by mechanisms that require the ability to bind phosphoinositides. Further studies will be necessary to evaluate these models by determining whether Stt4 is a component of the spindle positioning checkpoint and/or the mitotic exit network.

Mammalian Protein Kinases That Potentially Function as Coincidence Detectors of Phosphoinositide and Cdc42
Signaling-PAK family kinases in animal cells lack PH domains and have not been shown to bind phosphoinositides (66). However, PAKs in mammalian cells function indirectly as integrators of Cdc42 and phosphoinositide signaling pathways because the phosphoinositide-activated kinase PDK1 phosphorylates the activation loop of PAK (67). PAK is also activated by sphingosine (66,68), indicating that it potentially functions as a coincidence detector of Cdc42 and sphingosine signaling pathways.
In contrast, the myotonic related Cdc42-binding kinase (a protein kinase that regulates filopodia formation (69)) may function as a coincidence detector of Cdc42 and phosphoinositide signaling in mammalian cells. Like Cla4, myotonic dystrophy-related Cdc42-binding kinase has PH and PBD domains in its N-terminal regulatory domain. Therefore, it would be interesting to determine whether the PH domain of myotonic dystrophy-related Cdc42-binding kinase binds phosphoinositides and whether this interaction is required for this kinase to regulate filopodia assembly.
In conclusion, Cla4 appears to regulate a host of cell biological processes in yeast by functioning as a coincidence detector that directly integrates Cdc42 and PI4P signaling pathways into a network. Further studies will indicate whether other coincidence detectors for these pathways exist in yeast and mammalian cells. Disrupting Cdc42 and phosphoinositide signaling networks at nodal integration points may provide a novel means of treating human immunodeficiency virus-positive patients afflicted with opportunistic fungal pathogens like Candida albicans whose virulence requires a Cla4-like kinase (63).