Site-specific regulation of the GEF Cdc24p by the scaffold protein Far1p during yeast mating

Far1p is required to recruit Cdc24p to the polarization site signaled by the mating partner

To test whether Far1p is required to recruit Cdc24p to the site of polarization during mating, we compared the localization of Cdc24p-GFP in wild-type (Figure 1A) and orientation-deficient far1-c (Figure 1, panel B; Chenevert et al, 1994) cells by time-lapse microscopy. We filmed several isolated pairs of MATa and MATα cells (labeled with ConA-TRITC) that landed very close to each on the slide; approximately 75% of wild-type cells (determined for 15 cell pairs in nine independent time-lapse studies) mated under these conditions, as determined by the formation of a zygote (Figure 1A). Importantly, Cdc24p-GFP was recruited to the site closest to the mating partner (arrow), implying that the cells were able to orient correctly in this natural pheromone gradient. In contrast, in all analyzed far1-c cells (23 far1-c cell pairs in 11 independent movies), Cdc24p-GFP was recruited to a random position on the cell cortex with respect to the mating partner (arrow). Only one of these cell pairs formed a zygote (4%), but instead resumed budding after a short cell cycle delay. Taken together, these results demonstrate that Far1p is required to translate a natural pheromone gradient into directed cell polarity during mating.

Far1p is able to ectopically recruit Cdc24p to the plasma membrane

To determine whether Far1p is able to recruit Cdc24p to the site of polarization, we fused the myristoylation signal of Gpa1p (MGCTVSTQTIGDESD) to the amino-terminus of Far1p⁵¹⁻⁸³⁰ (myr-Far1p⁵¹⁻⁸³⁰), which lacks its major nuclear localization signals (Figure 2A; Blondel et al, 1999). For control, we fused a mutated myristoylation signal, where the myristoylated glycine residue has been changed to a non-myristoylatable alanine residue (myrG2A-Far1p⁵¹⁻⁸³⁰). As expected, myr-Far1p⁵¹⁻⁸³⁰-GFP, but not myrG2A-Far1p⁵¹⁻⁸³⁰-GFP, was uniformly localized at the plasma membrane (Figure 2B), and a fraction was also detectable at internal membranes. Importantly, myr-Far1p⁵¹⁻⁸³⁰ was sufficient to recruit Cdc24p-GFP to the plasma membrane (Figure 2, panel C). To avoid any interference with the cell cycle stage, we expressed myr-Far1p⁵¹⁻⁸³⁰ or for control myrG2A-Far1p⁵¹⁻⁸³⁰ in YMG258 cells depleted for the G1 cyclins, which uniformly arrest in G1 prior to bud emergence (Gulli et al, 2000). Expression of myr-Far1p⁵¹⁻⁷³⁴-H7, which harbors a short carboxy-terminal deletion of the Cdc24p-binding site on Far1p (Figure 2, panel A; Valtz et al, 1995), failed to recruit Cdc24p-GFP. Taken together, these results demonstrate that Far1p is able to ectopically recruit Cdc24p to the plasma membrane.

Far1p may activate the GEF activity of Cdc24p in vivo

Overexpression of myr-Far1p⁵¹⁻⁸³⁰ was toxic in wild-type cells, while myrG2A-Far1p⁵¹⁻⁸³⁰ or untagged Far1p⁵¹⁻⁸³⁰ had no effect (Figure 3A and data not shown). This was surprising because uniform recruitment of Cdc24p to the plasma membrane is not sufficient for its activation (see accompanying paper by Shimada et al). However, the cells expressing myr-Far1p⁵¹⁻⁸³⁰ predominantly arrested with a large unbudded morphology and an unpolarized actin cytoskeleton (Figure 3B), characteristic for uniform activation of Cdc24p (Shimada et al, 2004). These results imply that Far1p may not only recruit Cdc24p to the plasma membrane but may also directly or indirectly increase its GEF activity. Supporting this hypothesis, overexpression of myr-Far1p⁵¹⁻⁸³⁰ in cdc24-m1 cells (Nern and Arkowitz, 1999), which harbor a point mutation in the Far1p-interaction site (Butty et al, 1998; Nern and Arkowitz, 1998), had no effect (Figure 3, panel C, upper half of the plate). Likewise, cells overexpressing myr-Far1p⁵¹⁻⁸³⁰-H7 were viable (panel A). Together, these results demonstrate that binding of Cdc24p to Far1p is required for the toxicity induced by myr-Far1p⁵¹⁻⁸³⁰. The presence of Cdc42p-GTP can be indirectly measured by the phosphorylation state of Cdc24p (Gulli et al, 2000; Bose et al, 2001) and membrane recruitment of the Cdc42p-effector Gic2p (Jaquenoud and Peter, 2000). Indeed, expression of myr-Far1p⁵¹⁻⁸³⁰ in wild-type cells resulted in a uniform accumulation of Gic2p-GFP at the plasma membrane in 72% of all unbudded cells (Figure 3, panel D), and induced hyperphosphorylation of Cdc24p similar to expression of Cdc42p-G12V (Figure 3, panel E). Finally, GST-CRIB pull-down assays confirmed that cells expressing myr-Far1p⁵¹⁻⁸³⁰ arrested with increased levels of activated Cdc42p (Figure 3, panel F). Based on these results, we conclude that Far1p not only recruits Cdc24p to the site of polarized growth, but may also be involved in its activation in vivo.

Activation of Cdc24p in vivo by a carboxy-terminal domain of Far1p may require dimerization

To identify the Far1p domains involved in membrane targeting and activation of Cdc24p, we used an assay described in the accompanying paper by Shimada et al (2004). Briefly, Cdc24p was uniformly targeted to the plasma membrane by fusing its amino-terminus to a myristoylation signal (myr-Cdc24p; Figure 4A), or for control the non-myristoylatable G2A version (myrG2A-Cdc24p). However, membrane-bound Cdc24p was not sufficient to induce toxicity, most likely because its GEF activity needs to be activated by binding to the small GTPase Rsr1p/Bud1p (Figure 4B; see accompanying paper). Importantly, expression of a carboxy-terminal fragment of Far1p fused to GST was toxic in combination with myr-Cdc24p (Figure 4B), and the cells arrested as large unbudded cells (data not shown). Toxicity induced by myr-Cdc24p required its specific interaction with Far1p, as co-expression of the same fragment with myr-Cdc24p-S189P had no effect in this in vivo assay. Conversely, myr-Cdc24p-G168D, which is specifically defective for binding to Bud1p/Rsr1p (Shimada et al, 2004), was activated by the carboxy-terminal Far1p-fragment, supporting the notion that specific binding of Far1p to Cdc24p is required for toxicity.

As GST is known to induce dimerization of proteins (Lim et al, 1994), we wanted to test whether dimerization of the carboxy-terminal fragment by GST is important for its ability to activate Cdc24p in vivo. Indeed, replacement of the GST moiety by an HA epitope abolished the toxic effect induced by the Far1p fragment (Figure 4C), although the fragment was expressed and interacted efficiently with Cdc24p. Further support for a role of dimerization in vivo also stems from the following experiment (Figure 4D). Surprisingly, co-expression of the GST-tagged carboxy-terminal fragment of Far1p and an inactive myr-Cdc24p-ΔDH mutant in wild-type cells was still partially toxic (Figure 4, panel D). However, we found that this toxicity depended on endogenous Cdc24p, because expression of the two proteins in cdc24-S189P cells (Nern and Arkowitz, 1998) had no effect (Figure 4, panel D). These results show that (1) functional Cdc24p is required for the toxic effect induced by the carboxy-terminal fragment of Far1p and (2) GST-induced dimerization of two carboxy-terminal fragments is required to recruit and activate endogenous Cdc24p, as schematically shown in Figure 4, panel D (right panel). In analogy with Ste5p (Elion, 2001), we speculate that the ring-finger domain of full-length Far1p may be involved in dimerization, and that predominantly the dimerized carboxy-terminal domain of Far1p may function as a potent activator of Cdc24p. However, we have so far been unable to detect Far1p dimers in vivo (P Wiget and M Peter, unpublished results).

Identification of Far1p mutations unable to activate Cdc24p in vivo

By deletion analysis, we mapped the activation domain of Far1p between amino acids 546 and 830 (Figure 5A). This minimal fragment was able to both interact (Figure 5D) and activate Cdc24p, as observed in our in vivo assay (Figure 5, panel B, data not shown). We next searched for point mutants within this minimal Far1p fragment that were no longer toxic when co-expressed with myr-Cdc24p (Figure 5, panel C). Interestingly, we found that mutations in residues 638 (MM6; G to E change) and 650 (MM18; G to R change) strongly reduced the Cdc24p-activation activity of this Far1p fragment in vivo. Both residues are conserved among Far1p homologues from other yeast species, but are absent in Ste5p (Supplementary Figure S1). Two-hybrid assays further showed that both residues were required for efficient binding of the minimal Far1p fragment to Cdc24p (Figure 5D). Taken together, these results identify the Far1p–Cdc24p-binding domain, and suggest that this domain may also be involved in the activation of Cdc24p by Far1p in vivo.

To confirm that the Far1p–Cdc24p interaction is biologically relevant, we analyzed far1∷LEU2 cells expressing the MM6 and MM18 mutations in the full-length Far1p context from the endogenous promoter. As expected, the cells were able to arrest the cell cycle in response to pheromones (Figure 5, panel E), but exhibited a strong bilateral mating defect (Figure 5, panel F). These results strongly suggest that the ability of Far1p to bind and activate Cdc24p is important for the establishment of cell polarity during mating.

The PH-like domain of Far1p may be required for efficient membrane localization in vivo

Both Far1p and Ste5p contain a conserved domain within their carboxy-terminus, which resembles a PH domain (amino acids 418–545 of Far1p; Supplementary Figure S2; Kai Hoffmann, personal communication). To test whether this conserved PH-like domain is functionally important, we constructed Far1p mutants deleted for this motif (Far1p-ΔPH-like) and expressed the mutant protein from its endogenous promoter in both wild-type and far1Δ cells. Like wild-type Far1p, Far1p-ΔPH-like was efficiently induced and phosphorylated in response to α factor (Figure 6B), and was able to complement the cell cycle arrest defect of far1Δ cells (Figure 6, panel C). However, far1Δ cells expressing Far1p-ΔPH-like exhibit a bilateral mating defect (Figure 6, panel D), although they are able to form shmoos with normal efficiency (data not shown). As published earlier (Nern and Arkowitz, 1999), wild-type Far1p-GFP was localized at shmoo tips after exposing cells to α-factor for 30 min. In contrast, little Far1p-ΔPH-like-GFP was detected at shmoo tips under the same conditions (Figure 6, panel E), although Far1p-ΔPH-like was able to efficiently interact with Gβγ and Cdc24p (Butty et al, 1998 and data not shown). To test whether artificial membrane localization of Far1p-ΔPH-like may bypass the need for its PH-like domain, we fused a myristoylation signal to the amino-terminus of Far1p-ΔPH-like (myr-Far1p-ΔPH-like). As shown in Figure 6F, artificially membrane-targeted myr-Far1p-ΔPH-like was toxic when overexpressed in wild-type cells (Figure 6, panel F), suggesting that it is able to efficiently recruit and activate Cdc24p in vivo. Taken together, these results suggest that the PH-like domain contributes to membrane recruitment and/or stabilization of Far1p at sites of polarized growth. At present, we do not know whether the PH-like domain is able to interact with lipids or proteins.

Binding of Gβγ is needed to activate Far1p

Surprisingly, we found that the toxicity of myr-Far1p was dependent on the Gβ subunit STE4 (Figure 7A). In contrast, myr-Far1p⁵¹⁻⁸³⁰ was toxic in cells deleted for STE11 or STE5, implying that an intact mating MAP-kinase signal transduction cascade is not necessary for the activation of Cdc24p. Gβγ has been shown to interact with the ring-finger domain of Far1p and Ste5p (Elion, 2001), and recent experiments indicate that this interaction may be important to activate Ste5p in vivo (Sette et al, 2000). Interestingly, wild-type cells expressing myristoylated Far1p-B4, which harbors a mutation in the ring-finger domain and as a result shows reduced binding to Ste4p (Valtz et al, 1995; Butty et al, 1998), were able to form colonies, although the growth rate was reduced compared to vector controls. Taken together, these results suggest that binding of Gβγ to Far1p is not only involved in membrane recruitment of Far1p but is also needed for its ability to activate Cdc24p in vivo. In analogy with Ste5p (Elion, 2001), we hypothesized that binding of Gβγ to Far1p may induce a conformational change necessary to expose the carboxy-terminal activation domain to Cdc24p. Indeed, in contrast to full-length cytoplasmic Far1p-nls1/2 (Blondel et al, 1999), expression of GST-Far1p-390-830 was toxic in ste4Δ cells expressing myr-Cdc24p (Figure 7C). Residues downstream of the ring-finger domain have been implicated in the activation of Ste5p by Gβγ (Sette et al, 2000), and these residues are conserved in Far1p (Figure 8A). Interestingly, myr-Far1p-P246A was toxic in both wild-type and ste4Δ cells (Figure 8B), implying that Gβγ was no longer required for its activation. Mutations in residues 248 and 251 gave similar results, although their toxicity in ste4Δ cells was somewhat weaker (data not shown). As shown in Figure 8C and D, far1Δ cells expressing Far1p-P246A from its endogenous promoter without myristoylation tag were able to mate efficiently with wild-type or orientation-deficient far1-c-mating testers (Figure 8, panels C and D). Taken together, these results suggest that Gβγ fulfills two distinct roles during mating polarization: first, it targets Far1p to the correct site at the plasma membrane, and second, it triggers a conformational change in Far1p that allows the activation of Cdc24p in vivo.

Far1p is required to orient cell polarity in a morphogenetic gradient

Using time-lapse microcopy of mating cells, we were able to directly visualize the orientation defect of far1-c cells in a natural pheromone gradient. Although the far1-c cells responded efficiently to pheromones as judged by induction of a transient cell cycle delay, Cdc24p-GFP accumulated at random sites with respect to the position of the mating partner. In several cases, we were able to confirm that the chosen site indeed corresponds to the incipient bud site (Valtz et al, 1995), which is located next to the previous site of cell division (Roemer et al, 1996). Thus, in the absence of functional Far1p, Cdc24p is ‘free' to interact with Rsr1p/Bud1p at the genetically determined bud site (Park et al, 2002; Shimada et al, 2004). However, mating movies with GFP-tagged markers for the axial bud site are needed to settle this point. As we could never detect Cdc24p-GFP at correct sites even early in the response, we believe that Cdc24p is not able to interact with Gβγ in the absence of functional Far1p. With the single-cell system at hand, we should now be able to follow the behavior of several polarity markers as well signaling proteins in real time in a physiologically relevant situation.

Far1p may function as a site-specific activator of Cdc24p during mating

Our results suggest that Far1p may play a dual role in the site-specific activation of Cdc24p during mating. First, Far1p may recruit Cdc24p to the site of receptor activation marked by Gβγ, and second, binding of Far1p to Cdc24p at the cell cortex may be involved in the activation of Cdc24p in vivo. Indeed, artificially membrane-targeted Far1p is able to ectopically localize Cdc24p uniformly to the plasma membrane. This localization requires interaction of Far1p with Cdc24p, as mutations in either Cdc24p or Far1p that interfere with their binding prevent Cdc24p recruitment. Interestingly, wild-type cells overexpressing myr-Far1p⁵¹⁻⁸³⁰, as well as cells co-expressing myr-Cdc24p with wild-type Far1p, failed to divide. In both situations, the cells accumulated predominantly as large unbudded cells, most likely because they uniformly activate Cdc24p all over their surface. Supporting this notion, Cdc42p-GTP levels were increased in these cells. Moreover, myr-Far1p was not toxic when expressed in cdc24-m1 cells (Nern and Arkowitz, 1998), which express a Cdc24p mutant defective for Far1p binding (Butty et al, 1998; Nern and Arkowitz, 1999). Thus, in contrast to direct membrane recruitment of Cdc24p via a myristoylation signal (Figure 4A and B; Shimada et al, 2004), membrane recruitment of Cdc24p by Far1p was sufficient to activate Cdc42p in vivo. These results imply that Far1p not only recruits Cdc24p to the correct location, but it may also function directly or indirectly as an activator of its GEF activity. Far1p may thus function during mating in an analogous manner as Rsr1p/Bud1p at bud emergence (Shimada et al, 2004).

Using a deletion approach, we identified two novel functional domains in Far1p. First, a conserved PH-like domain is required for efficient membrane localization of Far1p. This domain is separate from its ring-finger motif, which is necessary and sufficient for the interaction with Gβγ (Butty et al, 1998), and thought to be involved in the recruitment of Far1p to the site of polarization during mating. Second, we mapped the minimal region involved in binding and activation of Cdc24p to the carboxy-terminus of Far1p encompassing amino acids 546–830. We found two single point mutations within this domain that disrupt binding of Cdc24p to Far1p. The mutations affect two closely spaced glycine residues, which are conserved among Far1p homologues, but are not present in Ste5p. However, we were so far unable to identify mutations within this domain, which separate binding and activation of Cdc24p.

Dual role of Gβγ in the regulation of Far1p

Our results imply that Gβγ not only recruits Far1p to the site of receptor activation, but binding of Gβγ to Far1p may also be required for its ability to activate Cdc24p in vivo. Several lines of evidence suggest that the carboxy-terminal activation domain is not accessible to Cdc24p in the absence of Gβγ (Figure 9). First, in contrast to expression of the carboxy-terminal fragment of Far1p, activation of Cdc24p by full-length Far1p requires Ste4p. Second, binding of Far1p and Gβγ is necessary to render full-length myristoylated Far1p toxic. Finally, point mutations next to the ring-finger domain of Far1p alleviate the requirement for Ste4p binding, implying that this region may be involved in an auto-inhibition mechanism. We speculate that binding of Gβγ to the ring-finger domain of Far1p triggers a conformational change, which allows its carboxy-terminal activation domain to activate Cdc24p (Figure 9). In analogy to the Wiscott–Aldrich syndrome protein (WASP) (Caron, 2002), it is possible that binding of Gβγ releases an auto-inhibitory domain (Pufall and Graves, 2002). Alternatively, Gβγ may promote dimerization of Far1p. Interestingly, artificial dimerization of the carboxy-terminal domain of Far1p by GST was necessary for its ability to activate Cdc24p in vivo. As the ring-finger domain of Ste5p has been shown to induce dimerization (Yablonski et al, 1996; Inouye et al, 1997), we speculate that the ring-finger domain of full-length Far1p may be involved in dimerizing its carboxy-terminal Cdc24p-activation domain upon pheromone signaling. However, we have so far not been able to detect Far1p dimers in vivo.

Role of Far1p in orienting polarized growth in a morphogenetic pheromone gradient

A detailed understanding of the molecular mechanisms of how yeast cells orient their cell surface growth towards their mating partner in a morphogenetic gradient is beginning to emerge. The uniformly distributed pheromone receptors are asymmetrically activated by binding to the pheromones, secreted by the nearest mating partner (Dohlman and Thorner, 2001). The activated pheromone receptors trigger the local dissociation of Gα from Gβγ. In turn, Gβγ activates the MAP-kinase signaling cascade by binding to Ste5p (Elion, 2001), while it recruits and activates the GEF Cdc24p via Far1p. Based on the experiments discussed here and elsewhere, we propose that binding of Gβγ recruits Far1p to the site of receptor activation, where it is stabilized via its PH-like domain. Binding of Gβγ also triggers a conformational change in Far1p, which may allow its carboxy-terminal domain to activate the GEF Cdc24p by a mechanism that involves dimerization (Figure 9). Locally activated Cdc42p promotes actin nucleation predominantly via the formin Bni1p, thereby assembling actin cables which serve as tracks to deliver vesicles containing new membrane material to the site of polarization (Pruyne and Bretscher, 2000a). Recent evidence has shown that Cdc42p is able to self-organize by a mechanism that requires actin and proteins involved in endocytosis and secretion (Wedlich-Soldner et al, 2003), implying that vesicle and membrane trafficking may play a role in the amplification of polarity signals. Moreover, specific membrane domains at sites of polarized growth may contribute to the stabilization and maintenance of critical polarized markers (Bagnat and Simons, 2002), perhaps by clustering the signaling receptor. Further work will be required to understand how the local activation of Cdc24p by Far1p is amplified into a robust polarization signal.

Strains constructions and genetic manipulations

Yeast strains are described in Table I of the supplementary data. The genotypes of the yeast strains are: W303 (ade2-1, trp1-1, can1-100, leu2-3,112, his3-11,15, ura3, ssd1-d2) or EG123 (ade2-101, trp1-Δ99, can1-100, leu2-Δ1, his4-519), unless noted otherwise. Standard yeast growth conditions and genetic manipulations were used as described (Sprague and Thorner, 1992). CDC24-GFP constructs were linearized with Bsu36 I for integration at the TRP1 locus. The strains YYS185 andYYS187 were obtained by backcrossing the strains cdc24-4 (YEF313) or RAY916 (cdc24∷HISG cdc24-m1) four times to K700 (W303) in order to improve Gal expression. Mating and halo assays were carried out as described previously (Valtz and Peter, 1997).

DNA manipulations and two-hybrid assays

Plasmids are described in Table II of the supplementary data. Standard procedures were used for recombinant DNA manipulations (Ausubel et al, 1991). PCR reactions were performed with the Expand polymerase kit as recommended by the manufacturer (Roche). Site-directed mutagenesis was performed by PCR and confirmed by sequencing. Oligonucleotides were synthesized by Genset (Paris, France). Details of plasmid constructions and oligo sequences are available upon request. The myristoylation sequence of Gpa1p (MGCTVSTQTIGDESD) was used to artificially target Far1p. Far1p-MM6 and MM18 were isolated by error-prone PCR of wild-type Far1p and identified by screening for specific mating defects when expressed in far1Δ cells (YMP1054). The mutations were identified by sequencing both strands. Two-hybrid assays were performed in YACB165 (Butty et al, 1998) containing the LacZ-reporter plasmid pSH18.34 (Gyuris et al, 1993). Miller units are the average of at least three independent experiments with standard deviations (Brown et al, 1997).

Microscopy

Proteins tagged with S65T variant of green fluorescent protein (GFP) were visualized on a Zeiss Axiovert 200M fluorescence microscope using a Zeiss GFP filterset no.10, an Orca-ER CCD camera (Hamamatsu, Japan) and Openlab software (Improvision, UK). Proteins expressed from the GAL promoter were induced by the addition of 2% galactose for 3 h. For quantitation, at least 200 cells were analyzed. Spinning-disc confocal images were acquired using the same microscope setup described above, but additionally equipped with a spinning-disc head (Visitech-International, UK).

For time-lapse microscopy, agarose pads and cell mounting on microscope slides was carried out as described previously (Hoepfner et al, 2000). Briefly, 1 ml of log-phase MATα EG123 cells grown in synthetic medium containing 4% glucose (SD) were pelleted and resuspended in PBS. MATα cells were stained by the addition of 5 μl 1 mg/ml ConcanavalinA-TRITC (ConA-TRITC, Sigma) for 5 min. Cells were then pelleted and washed twice in PBS. A volume of 1 ml of log-phase MATa WT or far1-c cells expressing CDC24 from the CYC1 promotor was mixed with the stained MATα cells, pelleted and resuspended in 50 μl SD containing all the required amino acids. Cells were imaged by acquiring three GFP focal planes 1 μm apart from each other every 150 s for up to 8 h.

Cell cycle synchronization

Exponentially growing cln1,2,3Δ METCLN2 (YMG258) cells were arrested in G1 by repressing CLN2 for 3 h in selective medium containing 2 mM methionine (Gulli et al, 2000). Efficient cell cycle arrest was controlled microscopically (Jaquenoud and Peter, 2000). The expression of myr-Far1p from the GAL1 promoter was induced for 3 h by adding 2% galactose. Where indicated, 30 μg/ml α factor (LIPAL-Biochemicals, Zurich) was added for 2 h.

GST-CRIB pull-down experiments, antibodies and immunoblotting

Standard conditions were used for yeast cell extracts and immunoblotting (Harlow and Lane, 1988). GST-CRIB pull-down experiments were carried out as described (Shimada et al, 2004). Polyclonal anti-Far1p, anti-Cdc24p and anti-Cdc42p antibodies have been described previously (Leeuw et al, 1995; Butty et al, 1998; Tjandra et al, 1998). 9E10 and monoclonal anti-GFP antibodies were obtained from the ISREC antibody facility. HA11 and anti-actin antibodies were purchased from Babco (Berkeley) and Boehringer Mannheim, respectively.