|Home | About | Journals | Submit | Contact Us | Français|
The Saccharomyces cerevisiae PAK (p21-activated kinase) family kinase Ste20 functions in several signal transduction pathways, including pheromone response, filamentous growth, and hyperosmotic resistance. The GTPase Cdc42 localizes and activates Ste20 by binding to an autoinhibitory motif within Ste20 called the CRIB domain. Another factor that functions with Ste20 and Cdc42 is the protein Bem1. Bem1 has two SH3 domains, but target ligands for these domains have not been described. Here we identify an evolutionarily conserved binding site for Bem1 between the CRIB and kinase domains of Ste20. Mutation of tandem proline-rich (PxxP) motifs in this region disrupts Bem1 binding, suggesting that it serves as a ligand for a Bem1 SH3 domain. These PxxP motif mutations affect signaling additively with CRIB domain mutations, indicating that Bem1 and Cdc42 make separable contributions to Ste20 function, which cooperate to promote optimal signaling. This PxxP region also binds another SH3 domain protein, Nbp2, but analysis of bem1Δ versus nbp2Δ strains shows that the signaling defects of PxxP mutants result from impaired binding to Bem1 rather than from impaired binding to Nbp2. Finally, the PxxP mutations also reduce signaling by constitutively active Ste20, suggesting that postactivation functions of PAKs can be promoted by SH3 domain proteins, possibly by colocalizing PAKs with their substrates. The overall results also illustrate how the final signaling function of a protein can be governed by combinatorial addition of multiple, independent protein-protein interaction modules.
Signal transduction pathways in eukaryotic cells use a variety of modular protein-protein interaction domains to provide functional linkages between individual signaling components (7, 51). Genetic exchange of sequences encoding these protein-protein interaction modules is thought to have facilitated the evolution of signaling proteins with new properties, by creating new combinations of protein contacts that in sum would govern the protein's function (51). An example of a ubiquitous interaction module is the SH3 domain (30, 45), which binds target proteins by recognizing proline-rich ligands that usually conform to the general consensus sequence PxxP (where P stands for proline and x stands for any residue). In the budding yeast Saccharomyces cerevisiae, genomic techniques have identified all proteins with SH3 domains, and peptide library screening methods have been used to define the optimum ligands for most of them (67). In theory, the existence of well-characterized signal transduction pathways combined with the ease of genetic manipulations makes the yeast system an ideal test bed for studying the role of SH3 domain-mediated interactions in signaling. Surprisingly, however, there are few instances in which an SH3-mediated interaction is known to participate in a well-defined signaling pathway in yeast. The best-characterized example is the hyperosmotic response (HOG) pathway, which requires interaction between an SH3 domain from the transmembrane sensor protein Sho1 and a proline-rich motif in the kinase/scaffold protein Pbs2 (40, 43, 59, 76). Another well-studied SH3 domain protein in yeast is Bem1, which functions in the control of polarized growth during both normal cell division and the mating response (4, 8, 9, 12, 24, 29) and which also has a role in facilitating signal transduction in response to mating pheromones (39, 47). Bem1 has no known enzymatic activity but has a PX (phox homology) domain that binds phosphoinositides, as well as several protein-protein interaction domains including two SH3 domains (SH3-1 and SH3-2) and a PB1 (phox and Bem1) domain. Its ability to interact with multiple proteins involved in polarized growth has led to the suggestion that Bem1 may serve as an organizing scaffold to help promote interactions between its binding partners (8, 9, 24, 29, 35). Despite considerable interest in the function of Bem1, no functionally significant target binding sites have been identified for either of its two SH3 domains (67). This study reports the identification of such a site in the protein kinase Ste20.
The budding yeast kinase Ste20 is a founding member of the p21-activated kinase (PAK) family of protein kinases (16). Originally identified as a component in the yeast mating pathway (33, 58), Ste20 also functions in signaling pathways that regulate filamentous growth and the osmotic stress response (37, 49, 57, 61). In addition, Ste20 is involved in the control of actin organization, polarized growth, and cell cycle transitions (15, 19, 23, 27, 28, 63, 73). In the mating pathway, Ste20 activates a mitogen-activated protein (MAP) kinase cascade in response to extracellular mating pheromones (reviewed in references 17, 20, and 25). Binding of these pheromones to G protein-coupled receptors triggers the release of Gβγ dimers, which bind Ste20 (36) and also recruit to the plasma membrane the kinase cascade scaffold protein Ste5 (22, 56, 69, 71). These Gβγ-Ste20 and Gβγ-Ste5 interactions result in the Ste5-associated kinase, Ste11 (a MAP kinase kinase kinase [MAPKKK]), being brought into close proximity with Ste20, which consequently phosphorylates and activates Ste11 (56, 68, 74). This activating event is then propagated through successive kinases in the mating pathway MAP kinase cascade.
Ste20 is normally enriched at the cell periphery, at regions undergoing polarized cell growth such as the tips of buds and mating projections (34, 53, 72). This polarized localization depends on interaction with the membrane-bound GTPase Cdc42 (1, 31, 34, 47, 53). Cdc42 itself is similarly localized and associates directly with the plasma membrane via posttranslational modification of its carboxyl terminus with a lipophilic prenyl moiety (77). In addition to localizing Ste20, Cdc42 activates Ste20 kinase activity, because the domain that binds Cdc42 (the CRIB domain) overlaps an autoinhibitory domain that regulates Ste20 kinase activity (31, 65). Point mutations in the Ste20 CRIB domain that precisely disrupt Cdc42 binding without disrupting the autoinhibitory function cause Ste20 to become refractory to activation by Cdc42 and cause a strong loss of signaling activity in vivo (1, 31). In contrast, mutations that disrupt the autoinhibitory function of the CRIB domain, or that remove the CRIB domain altogether, confer constitutive (Cdc42-independent) activity on Ste20 (31, 47). Thus, the localization and activity of Ste20 are normally intimately connected due to their dependence on Cdc42, though this association can be uncoupled by deleting the CRIB domain, which yields Ste20 that is active yet delocalized (1, 31, 34, 47, 53). Remarkably, although signaling by this CRIB deletion form of Ste20 is detectably reduced (e.g., in the mating response), it is not reduced as drastically as might be expected from its strong delocalization, given that signaling is thought to be initiated at the plasma membrane (31, 56, 69). One possible explanation for this relatively mild defect is that Ste20 might contain residual localization ability by binding to other polarized proteins.
Previous work (18, 35) has found that Ste20 can interact with Bem1, which shows a cortical localization similar to that of Cdc42 and Ste20 (2, 29). Interactions between PAKs and SH3 domain proteins are common in other systems (6, 11, 26, 38, 41). Some evidence suggests that Bem1 may serve a parallel or overlapping role with Cdc42, since overexpression of Bem1 can partially compensate for defects in Ste20 localization and signaling in certain cdc42 mutants (47). Nevertheless, the manner in which Bem1 affects pheromone signaling remains poorly defined, in part because Bem1 also binds other signaling proteins in the pheromone response pathway such as Ste5 and Far1 (35, 39). In this study, we investigate the binding between Ste20 and Bem1 and the contribution of this interaction to Ste20 function. We demonstrate that the regulatory (nonkinase) half of Ste20, which contains the Cdc42-binding domain, also contains a separate proline-rich motif that binds to Bem1. We then report on how mutations that perturb Bem1-Ste20 binding affect Ste20 function in several signaling pathways, both in the presence and in the absence of functional Cdc42-Ste20 interaction.
Yeast strains are listed in Table Table1.1. New strains were constructed as follows. PPY1249 is a segregant from a cross between the unpublished strain PPY650 (MATα ste20Δ) and YEL285 (MATa ADE2 ste20Δ285-582) (34). PPY1287 (ssk1Δ ste20Δ) was obtained via a cross between SO621 (MATa ssk1::HIS3 ste50::LEU2), a gift from S. O'Rourke, and the unpublished strain PPY1281 (MATα ste20-1::TRP1). PPY1341 and PPY1343 are segregants from a cross between strains PPY1329 and PPY1331, which were constructed as follows: PPY1329 contains a HIS2-marked FUS1-lacZ reporter integrated at the FUS1 locus, introduced by transformation with plasmid pFL-HIS2 (47) into strain MOSY0151 (MATα ste20::TRP1), a gift from D. Lew; PPY1331 (MATa bem1::ura3::LEU2) was derived by transformation of strain JMY1128 (bem1::URA3), a gift from D. Lew, with the marker swap construct pUC4-ura3::LEU2 (46). PPY1356 harbors an NBP2::myc13-kanR fusion, inserted at the genomic NBP2 locus by transformation of strain DLY1 with KpnI-digested pPP1446, which contains a 477-bp C-terminal fragment of NBP2 lacking the native stop codon in the BamHI site of the myc13-kanR plasmid pFA6a-13Myc-kanMX6 (70). PPY1415 and PPY1456 were derived from PPY1343 and PPY913, respectively, by replacing the complete NBP2 open reading frame with an nbp2::kanR fragment, obtained by PCR of the kanR cassette from pFA6a-kanMX6 (70) with primers whose 5′ ends contain 40 bp of homology to sequences immediately upstream and downstream of the NBP2 open reading frame. PPY1646 (ssk2Δ ssk22Δ ste20Δ) was created by deleting STE20 by use of plasmid pDH104 (74) in strain TM252 (40). PPY1691 was derived from the unpublished strain PPY860 (ste20Δ ste5Δ) by transformation with the hog1::hisG::URA3::hisG fragment from pBC99 (49), selection for uracil prototrophy, and then selection for 5-fluoroorotic acid (5-FOA) resistance.
Plasmids used in this study are listed in Table Table2.2. Signaling properties of Ste20 mutants were studied in two plasmid contexts, pRL116 and pPP1001. pRL116 (34) is a CEN URA3 plasmid expressing a GFP(S65T)-STE20 fusion gene from the native STE20 promoter. pPP1001 (31) is a CEN URA3 plasmid expressing the native, untagged STE20 gene from its own promoter. The degree to which these plasmids complement ste20Δ phenotypes has been discussed previously (31).
Point mutations in STE20 were first generated in pRL116, by Pfu polymerase-mediated extension of complementary mutagenic oligonucleotides, according to the instructions in the QuikChange site-directed mutagenesis kit (Stratagene). To ensure the absence of spurious mutations, restriction fragments containing the desired mutations were transferred back into unmutagenized plasmids, and the transferred fragments were sequenced. The P475G P477A (PP-GA) mutation was transferred on a SalI-SgrAI fragment into the green fluorescent protein (GFP)-STE20 plasmids pRL116, pPP964, and pPP1010 to create pPP1211, pPP1212, and pPP1213, respectively, and into the non-GFP fusion constructs pPP1001, pPP1002, and pPP1011 to create pPP1214, pPP1215, and pPP1216, respectively. In addition, the P475G P477A mutation was introduced directly by site-directed mutagenesis of pB20N2 to create pPP1180. Point mutations are marked by the gain or loss of restriction sites: a loss of EcoRV for the S338A mutation and a gain of BlpI for the P475G P477A mutation.
The GAL1pr-GST-STE20 constructs encoding wild-type Ste20 (pRDSTE20ATG) and Ste20ΔN (pRDSTE20-RI), as well as the vector control (pRD56), have been described previously (19, 53, 55). Note that pRDSTE20ATG was also called CY409 in reference 53 and was mistakenly given the name of the empty vector, pRD56, in reference 19, as a consequence of which the GAL1pr-GST-ste20-K649M construct was designated pRD56-K649M (19). Variants pPP1254 (PP-GA) and pPP1327 (L369G) were derived from pRDSTE20ATG by transfer of the SalI-SgrAI fragment from pPP1211 or the BsrGI-SgrAI fragment from pPP1117.
The two-hybrid plasmids pB20N2 and pPP1062 have been described previously (31). Additional two-hybrid plasmids, pB20N, pB20BA, pB20BB, and pB20BC, were constructed by PCR amplification of the STE20 codons indicated in Table Table22 (followed by a stop codon introduced by the downstream primer) and insertion into pBTM116 (3). Sequences in pB20N were inserted as an EcoRI-PstI fragment; those in pB20BA, pB20BB, pB20BC, and pPP1309 were inserted as BamHI-PstI fragments. In pB20N2, pB20BA, and pB20BC, a silent mutation from the downstream PCR primer disrupts the native EcoRI site near STE20 codon 499. The PP-GA mutation in pPP1180 was introduced by directed mutagenesis of pB20N2 as described above. Thirteen additional derivatives of pB20N2 containing point mutations at single codons from Lys469 to Ser481 (K469A through S481A) were made by first introducing a silent XhoI site (lowercase letters) overlapping STE20 codons 463 to 465 (TCc tcg agt) and then replacing the XhoI-PstI fragment with XhoI/PstI-cut PCR fragments in which mutations were introduced by the upstream (forward) primer.
The AD-Nbp2 plasmid pPP1355 was made by PCR amplification of full-length NBP2 and insertion into pGADXP as an EcoRI-SalI fragment. The activation domain (AD) fusions to Bem11-551 (pRL5.2) and Bem1157-551 (pRL51.1) have been described previously (35). Plasmid pXP-BEM1 was constructed by PCR amplification of full-length BEM1 and ligation into pGADXP as an EcoRI-BglII fragment. Coligation of two PCR fragments was used to generate pXP-BEM1-W108A (EcoRI-HindIII and HindIII-BglII) and pXP-BEM1-W192A (EcoRI-SalI and SalI-BglII); the W108A mutation introduces an MluI site, and the W192A mutation disrupts a BlpI site.
Halo assays of growth arrest, patch mating assays, and FUS1-lacZ transcription assays in response to α-factor or galactose-inducible constructs were performed as described previously (31). For cross talk assays, transformants were grown overnight in selective glucose medium to mid-exponential phase; these cultures were split in two, diluted with an equal volume of growth medium with or without 2 M sorbitol (final concentration, 0 or 1 M), and then incubated for 3 h at 30°C before the β-galactosidase assay. Liquid β-galactosidase assays (31) and filter β-galactosidase assays (3) were performed as described previously.
For coprecipitation of Bem1-myc or Nbp2-myc with glutathione S-transferase (GST)-Ste20, transformants were first grown overnight at 30°C to an optical density at 660 nm of 0.6 to 1.0 in uracil-deficient medium (−Ura)-raffinose containing 0.1% glucose; then they were induced with galactose (2%) for 3 h. Equivalent numbers of cells (usually 1 × 108 to 4 × 108) were harvested, and clarified cell lysates were prepared by glass bead lysis as described elsewhere (31), except that additional Triton X-100 was not added after the glass bead lysis step. GST-Ste20 was precipitated from the clarified lysates by addition of 40 μl of a 50% slurry (in buffer B) of glutathione-Sepharose beads (Amersham Pharmacia Biotech), followed by mixing on a nutator for 1 h at 4°C. Beads were pelleted at 2,400 × g, washed four times with 200 μl of cold buffer B, resuspended in 25 μl of 1.25× sodium dodecyl sulfate-polyacrylamide gel electrophoresis sample buffer, boiled 2 min, and centrifuged 5 min. Eight microliters was run on duplicate sodium dodecyl sulfate-7% polyacrylamide gels, followed by transfer and immunodetection as described elsewhere (31). GST-Ste20 was detected by using a mouse anti-GST antibody at a 1:1,000 dilution (B-14; Santa Cruz Biotechnology). Bem1-myc and Nbp2-myc were detected by using a rabbit anti-myc antibody at 1:2,000 (a gift of the R. Vallee lab, Columbia University). Some Bem1-myc was always detectable in the precipitates, even from control lysates of cells expressing GST alone.
Agar invasion assays were performed as described previously (13). Briefly, cells picked from freshly grown (2- to 4-day-old) colonies on −Ura plates were patched onto yeast extract-peptone-dextrose (YPD) plates, incubated for 30 h at 30°C, and then photographed before and after rinsing under a gentle stream of deionized water. To analyze filamentous morphology, cells were plated onto a low-glucose −Ura medium (14) and incubated at 30°C for 24 h prior to photomicroscopy.
For GFP-Ste20 observations, transformants were grown at 30°C in −Ura-glucose or −Ura-His-glucose medium, and cells were examined without fixation by using a Nikon E600 epifluorescence microscope equipped with a 50× Plan oil immersion objective. Cell morphology was photographed by using a Nikon Labophot-2 microscope equipped with a 30× extra-long working distance objective and Hoffman modulation contrast optics.
As described in the introduction, when active Ste20 is delocalized by removal of the Cdc42-binding (CRIB) domain, a measurable but mild signaling defect results. To explain why this defect was relatively small, we considered the possibility that interaction between Ste20 and Bem1 could provide some residual localization information. To determine whether Bem1 and Cdc42 could in fact contribute independent, additive effects on Ste20 function, we first mapped the binding site for Bem1 on Ste20. Using a two-hybrid assay and fusions to different fragments of Ste20, we found that residues 434 to 499, located between the CRIB and kinase domains of Ste20, were both necessary and sufficient for binding to Bem1 and were dispensable for binding to Cdc42 (Fig. (Fig.1A).1A). Within this region lies a proline-rich segment (Fig. (Fig.1B)1B) containing two adjacent matches to a core binding motif (PxxP) for SH3 domains (45), raising the possibility that these are recognized by a Bem1 SH3 domain. To test this notion, we made point mutations that simultaneously disrupt one proline in each motif (Pro475 to Gly and Pro477 to Ala) (Fig. (Fig.1B).1B). Indeed, this double mutation (designated PP-GA) eliminated Bem1 binding by the two-hybrid assay as strongly as did deletion of residues 434 to 499 but had no effect on Cdc42 binding (Fig. 1A and C). By contrast, two other PxxP sites elsewhere in the Ste20 N terminus (Fig. (Fig.1A)1A) were neither necessary nor sufficient for Bem1 binding (Fig. 1A and C).
To further substantiate these findings, we used a coprecipitation assay involving full-length GST-Ste20 and Bem1-myc fusion proteins. We found that the PP-GA mutation strongly decreased the amount of Bem1-myc that coprecipitated with GST-Ste20 (Fig. (Fig.1D).1D). The residual interaction was eliminated by complete deletion of the 495 N-terminal residues of Ste20 (GST-Ste20ΔN). These coprecipitation results are consistent with those from the two-hybrid assays and indicate that the PP-GA mutation disrupts a predominant, though perhaps not exclusive, site of interaction between the full-length Ste20 and Bem1 proteins.
Additional evidence suggested that this tandem PxxP motif binds the second SH3 domain (SH3-2) in Bem1. First, the pattern of two-hybrid binding was the same regardless of whether the Bem1 constructs contained both SH3 domains (residues 1 to 551) or only the second SH3 domain (residues 157 to 551) (see Fig. Fig.1A1A legend). Second, we made point mutations within full-length Bem1 in each SH3 domain (W108A in SH3-1 and W192A in SH3-2) at highly conserved tryptophans that are involved in binding proline-rich ligands (32). Only the mutation in the second SH3 domain (W192A) disrupted binding to Ste20; the mutation in the first SH3 domain (W108A) did not (Table (Table3).3). This was true when either of two Ste20 fragments (comprising residues 1 to 499 or 434 to 499) was used, while full-length Ste20 (residues 1 to 939) interacted negligibly with Bem1 by this assay, as observed previously (35). The W192A mutation also reduced the binding of Bem1 to Cdc24 (which served here as a control), but this was a much milder effect than the complete elimination of Ste20 binding. Together, therefore, these results show that the SH3-1 domain is dispensable, but an intact SH3-2 domain in Bem1 is required, for binding to the proline-rich motif in Ste20.
To study the effects of Bem1 binding on Ste20 signaling, the PP-GA mutation was introduced into vectors expressing full-length STE20 from its own promoter in either GFP-tagged or untagged (native) forms. Previous work showed that the GFP-tagged construct supplies a wild-type level of Ste20 function, whereas the untagged, native construct functions at a slightly reduced level (likely due to the lack of 3′ untranslated sequences ), which can mildly enhance phenotypic severity (31). When introduced into ste20Δ cells and tested for effects on pheromone response (Fig. (Fig.2A),2A), the PP-GA mutant showed a mild defect in FUS1-lacZ induction (60% reduced from that of the wild type), though not strong enough to cause a clear phenotype in assays of growth arrest or mating (Fig. (Fig.2A).2A). To address the relationship between Bem1 binding and Cdc42 binding, the PP-GA mutation was combined with two previously characterized mutations in the CRIB domain (31), which disrupt Cdc42 binding either mildly (S338A) or completely (Δ334-369). The partial effect of the PP-GA mutation became more pronounced when coupled with the mildest Cdc42-binding mutation, S338A, such that the combined mutant was strongly defective in FUS1-lacZ, arrest, and mating assays (Fig. (Fig.2A).2A). The PP-GA mutation also reduced signaling when combined with the complete CRIB domain deletion, Δ334-369, suggesting that Bem1 binding can affect Ste20 signaling in the absence of Ste20-Cdc42 interaction. Altogether, these observations show that the Bem1-binding domain has a positive effect on Ste20 signaling, that Cdc42 and Bem1 make separable contributions to Ste20 function, and that these two interactions are not redundant but in fact collaborate to promote optimal signaling by Ste20.
Because Ste20Δ334-369 is a constitutively active kinase (31), the fact that its mating pathway signaling is reduced by the PP-GA mutation suggests that Bem1 can influence Ste20 at a postactivation step, possibly by promoting colocalization with substrates. To rule out the alternative possibility that the PP-GA mutation affects signaling by disrupting the binding of Ste20 to Gβγ or Ste5, each of which can interact with Ste20 (35, 36), we activated signaling in ways that bypass these factors. The requirement for pheromone and Gβγ was bypassed by expression of membrane-targeted Ste5 (Ste5-CTM ) or mutationally activated Ste5 (Ste5-P44L-GST ). Here, the PP-GA mutation mildly affected signaling by Ste20 with an intact CRIB domain and clearly decreased signaling by the Ste20Δ334-369 derivative, which lacks the CRIB domain (Fig. (Fig.2B).2B). To bypass the requirement for Ste5 (as well as pheromone and Gβγ), we took advantage of Ste5-independent cross talk activation of the mating pathway, which occurs when hog1Δ cells are subjected to hypertonic stress (49). To ensure that Ste5 made no contribution, these experiments were performed in hog1Δ ste20Δ ste5Δ cells. As with pheromone signaling, cross talk signaling was moderately reduced by single mutations in either the CRIB or the Bem1-binding domain and was strongly reduced when mutations in the two domains were combined (Fig. (Fig.2C).2C). In total, these results indicate that the PP-GA mutation affects the ability of Ste20 to activate downstream signaling rather than affecting its ability to be regulated by pheromone, Gβγ, Ste5, or Cdc42.
For several reasons, we do not believe that the PP-GA phenotypes are due to reduced Ste20 protein levels: (i) the mutation did not affect GST-Ste20 levels when expressed from a strong promoter (see Fig. Fig.1D);1D); (ii) we saw no consistent effect on levels of natively expressed GFP-Ste20 fusions, though these were only faintly detectable by immunoblotting even after immunoprecipitation with anti-GFP antibodies from large (50- to 100-ml) cultures (data not shown); and (iii) further experiments below show that the PP-GA mutation confers no defect in a specific genetic context (bem1Δ cells), making trivial explanations for the signaling phenotypes unlikely.
Ste20 is ordinarily not essential for viability, but ste20Δ cla4Δ cells are nonviable; therefore, Ste20 can perform an essential function that is redundant with that of its related PAK family kinase, Cla4, and that requires the Ste20 CRIB domain (31, 34, 53). This essential function of Ste20 appears to be less dependent on interaction with Bem1 than on interaction with Cdc42, as evidenced by the fact that the PP-GA mutant was better able to support growth upon Cla4 inactivation than was the S338A mutant (Fig. (Fig.3A,3A, top). Nevertheless, quantitative measurement of growth rates (Fig. (Fig.3A,3A, bottom) showed that the PP-GA mutant was slightly impaired, though not as strongly as the CRIB domain mutants (S338A and Δ334-369), which themselves retained partial function in this assay. The morphology of these cells was somewhat heterogeneous (Fig. (Fig.3B)3B) but showed general trends suggesting a unique phenotype for the PP-GA mutant that is neither wild type nor null. The cla4 cells harboring the Ste20 PP-GA mutant had relatively normal bud necks, compared to those of the vector controls, but they tended to form rounder, less elongated buds than cells with wild-type Ste20, which showed elongated buds characteristic of cla4 STE20 cells (15). By comparison, the CRIB domain mutants showed more examples than the PP-GA mutant of both elongated buds and wide bud necks, plus additional morphologies (Fig. (Fig.3B)3B) (data not shown), suggesting that the PP-GA morphology is not simply attributable to partial loss of function. While the PP-GA morphology resembles that of wild-type cells, it is not the phenotype of wild-type Ste20 in a cla4 background, and this suggests that the persistent apical bud growth of cla4 STE20 cells (15) depends on Ste20-Bem1 interaction, perhaps causing Ste20 to act apically rather than isotropically. This view is consistent with proposed roles for Bem1 in bud emergence, apical polarization, and symmetry breaking (8, 9, 29), as well as with the requirement for both Bem1 and Ste20 in the persistent apical growth of cdc34 mutants (24, 63).
Cortical localization of GFP-Ste20 was only mildly decreased by the PP-GA mutation (data not shown); this decrease was not as severe as that of even the mildest Cdc42-binding mutant, Ste20S338A (31), suggesting that Bem1 may assist in localizing Ste20 to the cell cortex (47) but makes a less critical contribution than does Cdc42. The fact that the CRIB domain mutations (S338A and Δ334-369) on their own cause strong delocalization (31, 34, 53) made it difficult to assess whether combination with the PP-GA mutation would confer a more severe defect in localization, as it does for signaling. To circumvent this difficulty, we examined GFP-Ste20 localization in cells overexpressing Bem1; Bem1 overexpression had been shown previously to restore GFP-Ste20 cortical localization in cdc42 mutant cells (47). Under these conditions, the Δ334-369 derivative of GFP-Ste20 was detected at sites of polarized growth in some cells, whereas the combined Δ334-369 PP-GA derivative was not (Fig. (Fig.3C).3C). When the CRIB domain was left intact, the PP-GA mutant showed polarized localization similar to that of wild-type Ste20, but in general it was less intensely enriched at these locations. These results support the general notion that Bem1 can help localize Ste20, though Cdc42 is the predominant localizing factor in otherwise-normal cells.
Of the 24 SH3 domain proteins encoded in the S. cerevisiae genome, only Bem1 has previously been implicated in pheromone signaling, though others (e.g., Cdc25, Boi1, Boi2, Myo3, and Myo5) can serve as Ste20 substrates or binding partners or can affect Ste20-related functions (18, 52, 73). Therefore, it was conceivable that the phenotypic effects of the Ste20 PP-GA mutation reflect a defect in binding to an SH3 domain protein other than Bem1. A recent phage display screen (67) identified consensus ligands for 20 yeast SH3 domains (not including the Bem1 SH3-2 domain). The SH3 domain of Nbp2, a potential adaptor protein with roles in stress response pathways (42, 48), preferentially bound the consensus sequence PxRPaPxxP, which matches the Bem1-binding region in Ste20 (Fig. (Fig.1B).1B). Hence, this Ste20 sequence might bind more than one SH3 domain. Indeed, we found that Nbp2 displayed a two-hybrid interaction with the isolated Bem1-binding fragment of Ste20 (residues 434 to 499), though it interacted negligibly with a larger Ste20 fragment (residues 1 to 499) that can bind Bem1 (Fig. (Fig.4A).4A). Furthermore, Nbp2 coprecipitated with a GST fusion to full-length Ste20 in a manner that, like Bem1, required the Ste20 N terminus and was disrupted by the PP-GA mutation (Fig. (Fig.4B).4B). In contrast, mutations causing kinase hyperactivity (L369G ) or inactivity (K649M [19, 34, 53, 55, 74]) did not disrupt Ste20 interaction with either protein (Fig. (Fig.4B),4B), suggesting that the other mutations (PP-GA and ΔN) disrupt binding by perturbing the binding site rather than by indirect effects on Ste20 activity. Together, these results show that both Bem1 and Nbp2 can associate with Ste20 and that they have parallel binding requirements. Thus, without further analysis, conclusions regarding the role of the Bem1-binding site in Ste20 signaling become somewhat ambiguous given the intracellular presence of Nbp2 and multiple other SH3 domains that could have ligand-binding preferences overlapping those of Bem1.
To address whether the pheromone response defects resulting from the Ste20 PP-GA mutation relate to Bem1, we compared the effects of this mutation in BEM1 versus bem1Δ cells (Fig. (Fig.5A).5A). In support of the relevance of the PxxP motifs to Bem1 binding, deletion of BEM1 caused signaling defects that were similar in magnitude to those caused by the PP-GA mutation, regardless of the status of the CRIB domain. Most importantly, the PP-GA mutation conferred no defect on bem1Δ cells (Fig. (Fig.5A,5A, right), indicating that the intact (PP) domain contributes its positive signaling role only when Bem1 is present. These results also show that the PP-GA phenotypes do not result from trivial defects such as misfolding or instability, as the mutation has no consequence when Bem1 is absent. Note that mutations in the CRIB domain (S338A and Δ334-369) reduced signaling both in the presence and in the absence of Bem1 (Fig. (Fig.5A),5A), demonstrating that bem1Δ cells are not generally insensitive to perturbations in Ste20 function but instead are specifically insensitive to the PP-GA mutation. In contrast to the results for bem1Δ cells, the PP-GA mutation still caused a pheromone response defect in nbp2Δ cells, and did so in every CRIB domain context tested (Fig. (Fig.5B),5B), indicating that the Bem1-binding domain can perform its signaling role irrespective of the presence or absence of Nbp2.
Altogether, these results indicate that the signaling defect resulting from the PP-GA mutation can be attributed to a disruption in Ste20-Bem1 interaction and that this interaction contributes to signaling in a manner that is clearly separable from interactions of Ste20 with other binding partners such as Cdc42 and Nbp2. It is also clear that in the absence of Bem1, the ability of the CRIB domain to bind Cdc42 is still critical (Fig. (Fig.5A,5A, right), because the S338A mutation caused an exceptionally strong defect in bem1Δ cells and this defect was suppressed somewhat by deletion of the entire CRIB domain, which (like Cdc42 binding) releases Ste20 from autoinhibition. This behavior suggests that binding and activation by Cdc42 still regulate Ste20 signaling even in the absence of Bem1 or the Bem1-binding domain, and therefore Bem1 is not required as a cofactor in the activation of Ste20 by Cdc42. Instead, interaction with Bem1 makes Ste20 signaling more efficient regardless of the status of Cdc42-Ste20 interaction.
In addition to pheromone response, Ste20 functions in the filamentous growth and HOG pathways (37, 49, 57, 61). In each case, as with pheromone response, Ste20 functions downstream of Cdc42 and stimulates a MAP kinase cascade by activation of the MAPKKK Ste11 (31, 46, 57, 68). We tested the Ste20 mutants for function in the filamentous growth pathway by an agar invasion test (Fig. (Fig.6A)6A) and by examining the morphology of cells grown on low-glucose medium (Fig. (Fig.6B).6B). Previous work indicated a role for Cdc42-Ste20 interaction, as agar invasion was strongly defective when Cdc42 binding alone was disrupted (by an S338A H345G mutation) but was largely restored when the autoinhibitory effect of the CRIB domain was also eliminated (31). Here, the PP-GA mutation had no evident effect on its own, but it eliminated the residual function of the Δ334-369 derivative, as revealed by comparison of the Δ334-369 and Δ334-369 PP-GA mutants (Fig. (Fig.6A).6A). The S338A mutant was more defective than the Δ334-369 mutant, but a faint residual haze of adherence was reproducibly observed; this haze was eliminated in the S338A PP-GA combined mutant. A similar hierarchy of phenotypes was apparent when these cells were examined for filamentous morphology (Fig. (Fig.6B):6B): (i) the PP-GA mutation alone caused at most a mild reduction in the extent of elongated, branched morphology; (ii) each of the two mutations in the CRIB domain, S338A and Δ334-369, caused a stronger reduction in elongation and branching but with some residual filamentous characteristics still apparent (e.g., slightly elongated cells, and uneven colony edges with cells occasionally protruding from the colony mass); (iii) these residual features were eliminated when the PP-GA mutation was combined with either of the CRIB domain mutations.
Function in the HOG pathway was assessed by growth on high-osmolarity media (Fig. (Fig.6C).6C). To make growth dependent on the Ste20 branch of the HOG pathway, ssk1Δ or ssk2Δ ssk22Δ deletion mutations were used to inactivate a parallel, redundant branch (49, 57). None of the single Ste20 mutations caused an observable defect, but the Δ334-369 PP-GA combined mutation caused a strong reduction in growth on high-osmolarity media (Fig. (Fig.6C),6C), with partial function evident upon extended incubation (6 or 7 days). Although a previous study reported a HOG signaling defect for the Δ334-369 allele (57), our tests did not uncover this defect in either of two strain backgrounds. The reasons for this difference are not clear (possibilities include differences in growth media and plasmid backbones) but may ultimately represent differences in detection—i.e., while the earlier study detected the partial defect in the Δ334-369 allele, our conditions may have been more permissive, allowing detection of the stronger Δ334-369 PP-GA defect. In this view, each result is informative, and it would be inaccurate to classify these alleles as simply functional or nonfunctional. Overall, however, the results in both the filamentation and HOG pathways were largely reminiscent of those in the mating pathway, in that single mutations in the Bem1-binding and Cdc42-binding domains conferred only partial defects, whereas double mutations combined to produce stronger phenotypes.
Previous attempts to determine the recognition sequence for the Bem1 SH3-2 domain were unsuccessful, due to poor expression in Escherichia coli (67, 76). To gain some insight into this issue, we examined evolutionary conservation of the Bem1-binding site in Ste20 and probed the conserved sequence with additional mutations. Comparison of this site among fungal orthologs of Ste20 (Fig. (Fig.7B)7B) shows that the strongest conservation occurs within a block of 11 residues (FIPSRPAPKPP; residues 470 to 480 in Ste20), which includes both of the tandem PxxP motifs targeted by the PP-GA mutation. Functional PxxP motifs often exist within more-constrained 7-residue sites: class I (+x@Px@P) and class II (@Px@Px+) (where + and @ stand for positively charged and aliphatic residues, respectively) sites (45). Both tandem motifs in Ste20 have positively charged residues at the −3 position with respect to PxxP (i.e., KFIPSRP and RPAPKPP), characteristic of class I sites (45), though only the second motif is a perfect match with the class I consensus (Fig. (Fig.7B).7B). Conservation of residues in both motifs could imply that they function redundantly or additively, or simply that one binding motif consists of more than the prototypical seven residues. We introduced point mutations across a 13-residue region (KFIPSRPAPKPPS) that includes the core 11-residue sequence plus 1 residue to either side (Fig. (Fig.7C).7C). Consistent with the conservation, Bem1 binding was minimally affected by mutations at the periphery (K469A and S481A) but was strongly disrupted by mutation at most of the 11 core residues, including each single mutation present in the original PP-GA double mutant (P475G and P477A). Mutations in the core sequence with the least impact (S473A and K478A) affect analogous positions within each of the tandem class I-like motifs, and indeed SH3 domains rarely show preferences at this position (45, 67). Overall, these results show that the tandem motifs are not redundant, and the possibility that they function additively also seems unlikely, because removal of the positive charge at the −3 position was tolerated in the first motif (K469A in KFIPSRP) but not in the second motif (R474G in RPAPKPP). Nevertheless, the second motif is not sufficient; mutations at flanking residues (e.g., F470, I471, P472) were strongly disruptive. Therefore, while not exhaustively defining the optimal ligand, these observations suggest that the Bem1 SH3-2 domain recognizes sequence features spread over a broad region of 11 residues, including a 7-residue class I motif plus four N-terminally flanking residues. The involvement of these flanking residues is reminiscent of some other SH3 domains (60, 66, 67), and it supports the functional significance of their evolutionary conservation (see Discussion).
In this study we identify and functionally characterize a proline-rich motif within Ste20, a PAK family kinase, that provides a binding site for the SH3 domain protein Bem1. Mutations in this motif (PP-GA) reduce pheromone-responsive signaling in a way that mimics a bem1Δ mutation, and the intact Bem1-binding domain enhances signaling only in cells that express Bem1, clearly demonstrating the physiological significance of the Bem1-Ste20 interaction. Bem1 and Cdc42 bind to distinct domains in Ste20, and combined disruption of Bem1-binding and Cdc42-binding domains impairs Ste20 signaling more strongly than either disruption alone, indicating that these interactions contribute in an additive fashion to optimal function. The additive effect of Bem1 and Cdc42 interactions and their physical independence offer what is perhaps the most broadly applicable aspect of our observations: they provide a concrete example of how changes in function can be conferred incrementally on a protein by the acquisition (e.g., during evolution) of new protein-protein interaction modules (51).
The Bem1-Ste20 interaction is pertinent to other cases where SH3 domain proteins are thought to promote localization and/or activation of PAKs (6, 11, 26, 38, 41). Our findings suggest that SH3-PAK interactions can contribute to signal transduction even after activation by the GTPase, as disruption of the Bem1-Ste20 interaction or deletion of BEM1 reduces signaling by the Cdc42-independent hyperactive mutant, Ste20Δ334-369 (Fig. (Fig.22 and and5A)5A) (47). This Bem1 dependence also implies that although Ste20Δ334-369 appears to be largely delocalized (31, 34, 53), it may still signal predominantly at the cell periphery, because that is where Bem1 is enriched (2, 29). Therefore, the simplest explanation for the effects of Bem1-Ste20 binding is that Bem1 helps colocalize activated Ste20 with its substrates, for instance, by maintaining a pool of active Ste20 at the cell periphery. This can explain why Bem1 binding enhances pheromone signaling, which is initiated at the plasma membrane (56, 69), but what about other signaling, such as that by Ste5P44L-GST (Fig. (Fig.2B)2B) or hyperosmotic cross talk (Fig. (Fig.2C)?2C)? In fact, in each of these alternative settings there is reason to suspect that signaling is still localized. For example, signaling by Ste5P44L-GST is reduced when Ste20 is delocalized by mutations that disrupt Cdc42 binding, irrespective of Ste20 activity levels (e.g., Fig. Fig.2B)2B) (31); hence, Ste5P44L-GST might still signal preferentially at the cell periphery, despite not being obviously enriched there (62). Also, hyperosmotic cross talk is initiated by the polarized transmembrane proteins Sho1 and Msb2 (49, 50), and hence the contribution of Ste20 may also occur primarily at these polarized sites.
Our mutational analysis of the Ste20 Bem1-binding site suggests that the Bem1 SH3-2 domain recognizes a rather broad motif, with important sequence features spread over 11 residues (Fig. (Fig.7C),7C), including a 7-residue class I consensus SH3-binding sequence (45) plus N-terminally flanking sequence. This 11-residue motif is strongly conserved among Ste20 orthologs (Fig. (Fig.7B),7B), suggesting that it confers a function that is under selective pressure. The disruption of Bem1 binding by mutations in Ste20 residues F470 and I471, at positions −7 and −6 (with respect to PxxP), is reminiscent of SH3 domains from Src family kinases, which show a preference for hydrophobic residues at these positions (60, 66). Hydrophobic residues are also present at the −7 position of PxxP motifs in other proteins that bind the Bem1 SH3-2 domain, such as Boi1, Boi2, and Mps1 (5, 44), though the functional significance of those interactions remains unclear. Whether other yeast SH3 domains show biases at these positions would have been missed in previous work due to the use of peptide libraries with only nine residues (67), but biases were found at positions −5 and −4 in several cases, and three SH3 domains showed preferences for tandem PxxP motifs similar to the Ste20 sequence (i.e., PxxPxPxxP). Altogether, these observations suggest that while SH3 domains can recognize short sequences such as the 7-residue class I or class II motifs (45), residues at the flanking 4 to 5 positions can also influence binding (60, 66).
Similarities to the Bem1-binding domain are also present in more-distant Ste20 orthologs and in the other two S. cerevisiae PAKs, Cla4 and Skm1 (Fig. (Fig.7D).7D). Bem1 does bind Cla4 (8, 24), likely mediated by the Cla4 sequence FQPQRTAPKPP (residues 451 to 461) (Fig. (Fig.7D),7D), which resembles the Ste20 motif both in sequence and in its positioning between the CRIB and kinase domains. The Schizosaccharomyces pombe ortholog of Ste20, Pak1/Shk1, contains two strong matches to the Bem1-binding motif (Fig. (Fig.7D):7D): one (#1) upstream of the CRIB domain and another (#2) between the CRIB and kinase domains. Indeed, fragments that include either motif can bind Scd2, the S. pombe homolog of Bem1 (11). Each of these other sequences contains some of the flanking features present in Ste20 (Fig. (Fig.7D).7D). Interestingly, Bem1 and Scd2 themselves each contain a class I motif within their PX domains, which are proposed to bind in cis to their own SH3-2 domains, possibly generating an autoinhibited conformation (21). While each sequence matches the class I motif consensus, they lack the additional conserved features of the broader Bem1-binding motif in Ste20 (Fig. (Fig.7D).7D). This lack could be due to other functional constraints on the PX domain, but a more tantalizing possibility is that it reflects negative selection against an optimal binding motif, which if present as an intramolecular ligand (in cis) might bind so tightly as to always outcompete trans ligands.
Finally, our observations show that two different SH3 domains, the Nbp2 SH3 domain and the Bem1 SH3-2 domain, can bind the same target site in vivo. At present, there is little to suggest that Ste20 function is affected by Nbp2, which so far is implicated only in stress response pathways (42, 48), but this remains a possibility given that the homolog of Nbp2 in S. pombe, Skb5, can activate the kinase activity of the S. pombe Ste20 homolog, Pak1/Shk1 (75). Recent work on the SH3-mediated interaction between Sho1 and Pbs2 suggests that other yeast SH3 domains have been under negative selection to reduce binding to the Pbs2 ligand, ensuring highly specific binding of Pbs2 to only the Sho1 SH3 domain (76). Our findings suggest the converse possibility that single SH3 ligands may interact with multiple different SH3 domains in the same cell, which may also be implied by the multitude of SH3 binding partners for the yeast protein Las17 (67). It remains to be determined whether this serves a functional purpose in the case of Ste20 interactions with Bem1 versus Nbp2, or in other situations in which multiple SH3 domains share ligand preferences.
We are grateful to D. McCollum, A. Neiman, D. Lew, S. O'Rourke, H. Saito, R. Vallee, and M. Whiteway for plasmids, strains, and antibodies and to R. Lamson for additional technical assistance.
This work was supported by a grant from the NIH (GM57769) to P.M.P.