Proteins involved in sterol synthesis interact with Ste20 and regulate cell polarity

The small Rho GTPase Cdc42 plays a central role in establishing and maintaining cell polarity in eukaryotes (Etienne-Manneville, 2004; Jaffe and Hall, 2005). In the yeast Saccharomyces cerevisiae, Cdc42 regulates different types of polarized growth at several stages of the life cycle. During budding, growth is restricted to a zone adjacent to the previous budding site in haploid cells or at either pole of the cell in diploids. First, a hierarchy of positional signals defines the bud site (Casamayor and Snyder, 2002). Haploid cells use an axial budding pattern in which cells bud adjacent to the preceding site of cytokinesis. Diploid cells exhibit a more complex bipolar budding pattern: daughter cells bud opposite to their previous birth pole and mother cells bud at either pole (Casamayor and Snyder, 2002). Then, growth is focused at that location and is restricted to the new bud by a diffusion barrier made of septins that is placed as a collar at the neck between mother and daughter cell (Douglas et al., 2005; Versele and Thorner, 2005). S. cerevisiae also exhibits polarized growth during mating. Binding of a pheromone secreted by cells of the opposite mating type to specific membrane receptors results in the stimulation of a mitogen-activated protein (MAP) kinase signaling cascade and in polarized growth towards the mating partner (Dohlman and Thorner, 2001). Cell polarization is also required for the transformation from cellular yeast to a filamentous form. In diploid yeast, the lack of nitrogen causes the transition to pseudohyphal growth, in which cells are elongated and bud in a unipolar fashion (Palecek et al., 2002). In haploid cells, nutrient limitation results in a similar developmental switch that allows cells to penetrate the surface of an agar medium in a process called invasive growth (Palecek et al., 2002).

Cdc42 promotes polarized growth through multiple pathways, including polarization of the actin cytoskeleton and directed vesicle trafficking. Among the Cdc42 effectors that trigger these pathways are Iqg1, Gic1, Gic2 (Brown et al., 1997; Chen et al., 1997; Epp and Chant, 1997; Osman and Cerione, 1998) and members of the p21-activated kinase (PAK) family. S. cerevisiae encodes three of these highly conserved serine/threonine protein kinases: Ste20, Cla4 and Skm1. Whereas little is known about the function of Skm1 (Martin et al., 1997), Cla4 has been shown to be involved in septin ring assembly and actin polarization (Cvrckova et al., 1995; Holly and Blumer, 1999; Longtine et al., 2000; Dobbelaere et al., 2003; Schmidt et al., 2003; Versele and Thorner, 2004). In addition, Cla4 has an important role in mitotic entry and exit (Höfken and Schiebel, 2002; Sakchaisri et al., 2004). The founding member of the PAK family, Ste20, promotes actin polarization via the type I myosins Myo3 and Myo5 (Wu et al., 1996; Evangelista et al., 2000; Geli et al., 2000; Lechler et al., 2000), and possibly by activation of the formin homolog Bni1 (Goehring et al., 2003). Ste20 also regulates mating, filamentous growth and osmotic stress response by phosphorylation, and activation of the MAP kinase kinase kinase Ste11 (Roberts and Fink, 1994; Wu et al., 1995; van Drogen et al., 2000). Furthermore, Ste20 has an undefined role in mitotic exit (Höfken and Schiebel, 2002) and phosphorylates histone H2B during hydrogen peroxide-induced apoptosis (Ahn et al., 2005).

To gain further insight into the role of Ste20 in cell polarity, we performed a screen for interactors of Ste20 using the split-ubiquitin system. Three proteins involved in sterol biosynthesis – Erg4, Cbr1 and Ncp1 – were identified in this screen. The phenotypes of the corresponding deletion strains indicate that these proteins share functions with Ste20 in cell polarization. Erg4, Ncp1 and Cbr1 play a role in bud site selection, apical growth, bud morphology, cell wall assembly, mating and haploid invasive growth. Thus, our results imply that sterol synthesis is crucial for cell polarization.

To identify regulators and targets of Ste20, we screened for interactors of Ste20 by employing the split-ubiquitin technique (Johnsson and Varshavsky, 1994). The split-ubiquitin method is based on the ability of the N- and C-terminal halves of ubiquitin to assemble into a quasi-native ubiquitin. UBPs, ubiquitin-specific proteases, which are present in all eukaryotic cells, recognize the reconstituted ubiquitin, but not its halves, and cleave the ubiquitin moiety off a reporter protein that had been linked to the C-terminus of ubiquitin (Fig. 1A). A modified version of the enzyme Ura3 that carries an additional arginine at the extreme N-terminus (RUra3) was used as the reporter. Once cleaved off from the C-terminal domain of ubiquitin, RUra3 was rapidly degraded by enzymes of the N- end rule pathway of protein degradation. Therefore, close proximity between two proteins fused to the N- and C-terminal half of ubiquitin, respectively, results in non-growth on medium lacking uracil. Conversely, growth on 5-FOA indicated protein interaction as well, because 5-FOA was converted by Ura3 into 5-fluorouracil, which is toxic for the cell. Using this approach, 22 Ste20-interacting proteins were identified. Among these proteins was Bem1 (Fig. 1B), which acts as a scaffold that links Ste20 with Cdc42 and other factors (Leeuw et al., 1995; Butty et al., 1998; Drees et al., 2001). In this study we focus on Erg4, Cbr1 and Ncp1, because all three proteins are involved in sterol biosynthesis (Reddy et al., 1977; Aoyama et al., 1981; Yoshida, 1988; Aoyama et al., 1989; Kelly et al., 1995; Lamb et al., 1999; Zweytick et al., 2000) (Fig. 8) and have not previously been reported to interact with Ste20. The remaining genes will be described elsewhere. As explained above, the interactions between Ste20 and Erg4, Cbr1 as well as Ncp1 were demonstrated by non-growth on medium lacking uracil (Fig. 1B).

To confirm these interactions using an independent approach, recombinant GST and GST-Ste20 purified from Escherichia coli were bound to glutathione-sepharose beads, which were then incubated with a yeast extract of ERG4-9MYC cells. Erg4-9myc interacted with GST-Ste20 but not with GST (Fig. 1C). In a similar experiment using a yeast extract of 3HA-CBR1 cells we demonstrated a specific binding between Ste20 and Cbr1 (Fig. 1D). However, we did not observe these interactions by co-immunoprecipitation, possibly owing to the transient or weak nature of these interactions.

In order to find out whether the identified genes share a common function with STE20, we examined their genetic interactions. STE20 deletion is lethal in a cla4Δ background because of functional redundancy of these PAK-family kinases (Cvrckova et al., 1995) (Fig. 2A). We reasoned that important downstream effectors of Ste20 might show the same negative interaction with cla4Δ. To test this hypothesis we individually deleted ERG4, NCP1 and CBR1 in a cla4Δ strain. In fact, cells deleted for CLA4 and ERG4 were no longer viable, whereas single deletions of each gene resulted in no difference in growth compared to wild-type cells (Fig. 2A). Deletion of NCP1 alone resulted in slower growth, but in a cla4Δ background NCP1 was essential for viability (Fig. 2A). These data indicate that Erg4 and Ncp1 share functions with Cla4, possibly as downstream effectors of Ste20. The deletion of CBR1 in a cla4-null mutant had no growth defect (Fig. 2A).

We also observed reduced growth for the ste20Δ ncp1Δ double mutant compared with the single deletion of NCP1 (Fig. 2B). This effect was weak but reproducible and indicates that the situation might be more complex. Therefore, we tested whether single deletions of ERG4, CBR1 and NCP1 are synthetically lethal with other deletions that show negative genetic interactions with ste20Δ. It has been reported that STE20 becomes essential in the absence of OCH1, LTE1 or SWE1 (Lee and Elion, 1999; Höfken and Schiebel, 2002; Goehring et al., 2003). Interestingly, deletion of either CLA4 or NCP1 also resulted in lethality in an och1Δ background, whereas deletion of CBR1 in these mutants had no effect on growth (Fig. 2C). Unexpectedly, deletion of ERG4 suppressed the slight growth defect of och1Δ (Fig. 2C). Because OCH1 encodes an α-1,6-mannosyltransferase with a role in cell wall biosynthesis (Nakayama et al., 1992), STE20, CLA4, NCP1 and ERG4 might also contribute to this process. In contrast to a previous report (Goehring et al., 2003), the ste20Δ swe1Δ double mutant grew normally in our hands and no other genetic interactions were identified with SWE1 (see supplementary material Fig. S1B). In a similar way, no additional phenotypes were observed for the lte1 double mutants (see supplementary material Fig. S1A).

The synthetic lethality between cla4Δ and erg4Δ, as well as between cla4Δ and ncp1Δ, implies that ERG4 and NCP1, like STE20, have overlapping functions with CLA4 during the cell cycle. Therefore, we tested whether the deletion of ERG4 or NCP1 in addition to that of CBR1 results in phenotypes similar to those described for either ste20Δ or cla4Δ strains. Ste20 has previously been described to participate in bipolar bud site selection (Sheu et al., 2000) and, interestingly, ERG4 was identified in a genome-wide screen for genes involved in bipolar budding (Ni and Snyder, 2001). We confirmed the requirement of ERG4 and STE20 for bud site selection in diploids (see supplementary material Fig. S2A). However, whereas the STE20-deletion strain mostly exhibited a unipolar budding pattern, the bud site was chosen randomly in the majority of erg4Δ/erg4Δ cells (Ni and Snyder, 2001) (see supplementary material Fig. S2A). In contrast to the diploid strains, the erg4Δ and ste20Δ mutants, as well as cbr1Δ and ncp1Δ cells, displayed a normal axial budding pattern in haploid cells (see supplementary material Fig. S2B).

Next, the role of Erg4, Ncp1 and Cbr1 in apical bud growth was tested. During budding, the cyclin-dependent kinase Cdc28 promotes polarized apical growth when coupled to the G₁ cyclins and promotes isotropic growth when associated with mitotic cyclins (Lew and Reed, 1995). The apical growth phase can be prolonged by G₁ cyclin overexpression, resulting in hyperelongated buds (Lew and Reed, 1995) (Fig. 3A,B). Cells deleted for genes encoding cell polarity proteins such as Ste20 form no or fewer hyperpolarized buds in response to overexpression of the G₁ cyclin CLN1 (Sheu et al., 2000). To test whether Erg4 and Cbr1 are involved in apical bud growth, we overexpressed CLN1 in the corresponding deletion strains and scored for the presence of hyperelongated buds. The wild-type strain and cells deleted for CBR1 showed more than 60% of buds were hyperpolarized after 4 hours induction (Fig. 3B). By contrast, in erg4Δ and ste20Δ cells, the number of hyperelongated buds was reduced to only about 20% (Fig. 3B). Immunoblot analysis revealed that the mutant and wild-type cells expressed comparable levels of galactose-induced CLN1 (Fig. 3C). Unfortunately, this assay could not be performed with ncp1Δ cells because of the slow growth of this strain (Fig. 2A). Taken together, the data demonstrate an involvement of ERG4 in bud site selection and apical bud growth.

Rho GTPases and their downstream effectors play an important role in cell wall assembly. They are required for the polarization of synthetic proteins such as glucan and chitin synthase (Levin, 2005). Various mutant strains, which are defective in cell wall integrity, rapidly lyse in distilled water when grown at 37°C (Levin, 2005). We observed such a temperature-sensitive cell lysis phenotype for cla4Δ, erg4Δ and ncp1Δ cells, but not for the CBR1- and STE20-deletion strains (data not shown). Cell wall integrity mutants are often also sensitive to compounds like Calcofluor White, which interferes with cell wall assembly (Ram et al., 1994). Therefore, we tested whether the deletion of the genes identified in the screen results in an increased Calcofluor White sensitivity. In fact, erg4Δ cells grew much slower on Calcofluor White (Fig. 4A). Next we analyzed the role of Erg4, Ncp1 and Cbr1 in cell wall formation using a more reliable and quantitative assay. Here, cells were incubated overnight with various amounts of zymolyase, an enzyme that degrades the principal cell wall component β-1,3-glucan (Levin, 2005), and, subsequently, the survival rate was determined. Hypersensitivity to this zymolyase treatment was observed for erg4Δ and ncp1Δ cells, and this hypersensitivity was even more pronounced for the cla4Δ strain (Fig. 4B). The deletion of CBR1 or STE20 had no effect (Fig. 4B). Together with the genetic interactions of och1Δ (Fig. 2C), these results suggest a function for Cla4, Erg4 and Ncp1 in cell wall biosynthesis.

Ste20 also functions in a signal transduction pathway that governs filamentous growth (Roberts and Fink, 1994). As a consequence, cells lacking STE20 fail to invade agar (Roberts and Fink, 1994) (Fig. 5A). To analyze the role of ERG4, NCP1 and CBR1 in this pathway, we individually deleted these genes in cells of the Σ1278b strain background, which is necessary for tests of filamentous growth. The CBR1 deletion strain was undistinguishable from wild type (Fig. 5A). ncp1Δ cells were unable to invade agar, whereas the deletion of ERG4 resulted in a hyperinvasive phenotype (Fig. 5A). These strains were further examined for filamentous morphology. To this end, cells were spread onto synthetic medium lacking glucose. Under these conditions, wild-type cells undergo few divisions and form small colonies of branched and elongated cells (Cullen and Sprague, 2000) (Fig. 5B). By contrast, colonies of the ncp1Δ mutant exhibited a reduction in branching and elongation (Fig. 5B). Cells isolated from these colonies were mainly spherical without a bud or with only a small round bud (Fig. 5C). Buds of the few large-budded cells were round rather than elongated (Fig. 5C). Consistent with this result, actin was evenly distributed within the cell and not polarized to the bud, as occurs in wild-type cells (Fig. 5D). Interestingly, we did not observe any morphological difference between wild-type and erg4Δ microcolonies (Fig. 5B). Therefore, the mechanisms of hyperinvasive growth in this strain remain unclear.

Ste20 was originally identified as a component in the yeast mating pathway (Leberer et al., 1992; Ramer and Davis, 1993). Upon pheromone stimulation, Ste20 triggers a MAP kinase cascade, which leads to transcriptional activation of mating-specific genes such as FUS1 and cell cycle arrest in G₁ (Dohlman and Thorner, 2001). Furthermore, the actin cytoskeleton and other factors polarize, resulting in the formation of a mating projection (Dohlman and Thorner, 2001). Therefore, we tested whether the proteins identified in the screen are involved in mating. Almost all wild-type cells (96%) treated with pheromone formed a mating projection after 150 minutes and had the actin cytoskeleton polarized towards the tip of that projection (Fig. 6A). Longer incubation of cells with a high concentration of mating pheromone leads to the initiation and termination of projection growth in regular cycles (Bucking-Throm et al., 1973). After 6 hours, 90% of the cells had two or more mating projections (Fig. 6A). Cells lacking CBR1 behaved like wild type (Fig. 6A). Again, this assay could not be performed with ncp1Δ cells because of their slow growth (Fig. 2A). In contrast to wild type, the erg4Δ mutant completely arrested in G₁ phase but more than 90% of the cells failed to form a mating projection (Fig. 6A). The actin cytoskeleton in these unpolarized cells was evenly distributed. Even after 6 hours, multiple mating projections were never observed in the erg4Δ strain. As a consequence of that lack of polarization, erg4Δ cells also had a reduced mating efficiency (Fig. 6B). We further examined other polarity markers, such as the mating-specific protein Fus1, which is restricted to the tip of the mating projection (Fig. 6C). Interestingly, Fus1 was also enriched at a small membrane region in most cells deleted for ERG4. By sharp contrast, Ste20, which also localizes at the tip of the mating projection, was completely mislocalized in erg4Δ cells (Fig. 6D). However, ERG4 was not required for the polarized localization of Ste20 at bud tips during vegetative growth (see supplementary material Fig. S3A).

The observed polarization defect of erg4Δ cells could be attributed to the defect in ergosterol synthesis or might be caused by another activity of the Erg4 protein. To distinguish these two possibilities, the erg4 deletion mutant was grown anaerobically in the presence of α-factor and ergosterol. Under these conditions cells take up sterol from the medium (Raychaudhuri and Prinz, 2006) and both wild type and the erg4Δ strain polarized to the same extent (see supplementary material Fig. S3A). Thus, efficient pheromone-induced polarization requires Erg4 enzyme activity.

Whereas loss of either ERG4 or NCP1 resulted in various polarity defects, no phenotypes were observed for cbr1Δ cells in the assays described above. We reasoned that this lack of mutant phenotype in the cbr1Δ strain might be due to redundancy. In fact, Ncp1 and Cbr1 have overlapping functions in sterol synthesis. They both transfer electrons to proteins of the sterol biosynthetic pathway, such as cytochromes P450 (see Discussion for details). Therefore, we deleted both genes and characterized the cbr1Δ ncp1Δ double mutant. As described above, the deletion of the NCP1 gene alone resulted in slower growth, whereas cbr1Δ cells behaved like the wild-type strain (Fig. 2A, Fig. 7A). Interestingly, cbr1Δ ncp1Δ cells were unable to grow. To examine the terminal phenotype of this double mutant, we generated a conditional lethal strain using a temperature-sensitive degron (td). The degron is added to the N-terminus of a protein, causing it to be degraded by a specific ubiquitin-mediated pathway when cells are shifted from 23°C to 37°C (Labib et al., 2000). As expected, cbr1Δ ncp1-td cells could not be distinguished from wild type at the permissive temperature but were inviable when grown at 37°C (Fig. 7B). We were now able to study the phenotypic consequences of the simultaneous loss of CBR1 and NCP1 function. At 3 hours after Ncp1 depletion was induced, the large majority of cells (around 70%) showed abnormal bud morphology (Fig. 7C,D). Most of these cells had hyperelongated buds, some of them with a constriction, but pointed or bent buds were observed as well. By contrast, only a negligible percentage (2-3%) of wild type, cbr1Δ and ncp1-td had similar defects (Fig. 7D). This clearly indicates that Ncp1 and Cbr1 have overlapping functions in bud morphology. Because these morphological defects are often due to an aberrant localization or organization of septins, we examined a GFP-tagged version of the septin Cdc12 in cbr1Δ ncp1-td cells. However, Cdc12-GFP was normally localized and organized in these cells (see supplementary material Fig. S4A). Many Cdc42 effectors play a role in exit from mitosis and the corresponding mutants arrest as large-budded cells with an elongated mitotic spindle and separated nuclei (Höfken and Schiebel, 2002; Jensen et al., 2002; Seshan et al., 2002; Höfken and Schiebel, 2004). Therefore, we examined whether the large-budded cbr1Δ ncp1-td cells have a defect in mitotic exit as well. However, we did not observe a significant increase in cells with two separated DAPI-staining regions in these cells (see supplementary material Fig. S4B). This indicates that the cbr1Δ ncp1-td strain has a morphological and not a mitotic exit defect.

We also tested whether the cbr1Δ ncp1-td mutant had other defects. These cells formed a normal mating projection, as occurred in wild-type cells and in the corresponding single mutants (see supplementary material Fig. S5A). The actin cytoskeleton and Ste20-GFP were highly polarized in all strains (see supplementary material Fig. S5A). Interestingly, when cells were stained with Calcofluor White to examine the budding pattern, we observed a diffused chitin staining in the cbr1Δ ncp1-td double mutant (see supplementary material Fig. S5B). Notably, such a phenotype has also been described for other cell polarity mutants (Novick and Botstein, 1985).

The Cdc42 effector Ste20 promotes the establishment and maintenance of cell polarity during vegetative growth, mating and filamentous growth. Although some downstream targets of Ste20 have been described, the underlying molecular mechanisms of cell polarization are poorly understood. Here, we describe the identification and characterization of three novel interactors: Erg4, Cbr1 and Ncp1. The fact that Bem1, a well-characterized interactor of Ste20, was found in this screen as well, demonstrates that the split-ubiquitin technique used in this study is suitable for the identification of proteins that bind to Ste20 under physiological conditions. Notably, we can confirm the interaction between Ste20 and Erg4, as well as between Ste20 and Cbr1, by pull-down assays. We further show negative genetic interactions, which are consistent with the observed protein interactions. The deletion of either ERG4 or NCP1 is lethal in a cla4Δ background. Such a negative interaction with CLA4 has been described for the STE20 deletion previously (Cvrckova et al., 1995). Therefore, Erg4 and Ncp1 might act in the same pathway as Ste20 – either upstream or downstream. The synthetic lethality also indicates that Erg4 and Ncp1 have overlapping functions with Cla4 during vegetative growth. In fact, in line with previous reports, we demonstrated that the deletion of either ERG4 or NCP1 results in defects similar to those described for either ste20Δ or cla4Δ strains (Ni and Snyder, 2001; Keniry et al., 2004).

The first polarized process in the cell cycle is the selection of a new bud site. As previously reported (Sheu et al., 2000; Ni and Snyder, 2001) and shown by us here, both Erg4 and Ste20 contribute to bud site selection in diploid but not haploid cells. Thus, Ste20 and Erg4 might act together to regulate bud site selection in diploids.

Once a bud site has been selected, the actin cytoskeleton orients towards this site and guides secretory vesicles to the cell surface, where they fuse. As a bud forms, this apical growth continues. During the G₂-M transition, cells switch to isotropic growth, which results in the formation of an ellipsoid-shaped bud. Our data demonstrate that Erg4 has a role in apical growth, as previously described for Ste20 (Sheu et al., 2000). Therefore, both proteins might cooperate in this process.

Here, we also show that Erg4, Ncp1 and Cla4 are involved in cell wall assembly. The corresponding deletion mutants not only interact with OCH1, which encodes a α-1,6-mannosyltransferase and contributes to cell wall biosynthesis, but also exhibit a zymolyase hypersensitivity. Furthermore, we demonstrate that chitin deposition is not restricted to bud scars, but is evenly distributed in the cell wall of cells lacking CBR1 and NCP1. Proper cell wall biogenesis requires polarization of synthetic enzymes such as chitin and glucan synthase. Rho1 is considered a master regulator in this process (Levin, 2005). By contrast, little is known about the role of Cdc42 and its downstream targets in cell wall maintenance. Various genes involved in the synthesis of the cell wall components chitin and β-1,3-glucan become essential in the absence of CLA4 (Goehring et al., 2003; Lesage et al., 2005). Together with our data, this suggests that Cla4, like Erg4 and Ncp1, regulates cell wall assembly by a hitherto unknown mechanism.

Ste20 and other Cdc42 effectors trigger exit from mitosis (Höfken and Schiebel, 2002; Jensen et al., 2002; Seshan et al., 2002; Höfken and Schiebel, 2004). In contrast to Cla4 and Gic1, the molecular mechanisms of Ste20 action in this process remain unclear. Recently, Ncp1 has also been described to contribute to mitotic exit (Keniry et al., 2004). Because our data show that Ncp1 binds to Ste20, it is tempting to speculate that Ncp1 and Ste20 act together in mitotic exit.

Like Ste20, Erg4 and Ncp1 not only have various functions during vegetative growth but also play a role in mating and filamentous growth. Ste20 is an essential component of the yeast mating pathway (Leberer et al., 1992; Ramer and Davis, 1993). Upon pheromone stimulation, Ste20 activates a MAP kinase cascade, which results in a cell cycle arrest in G₁ phase. Via actin polarization, these cells then form a mating projection and fuse with a cell of the opposite mating type (Dohlman and Thorner, 2001). Here, we show that cells in which ERG4 is deleted undergo a cell cycle arrest but fail to polarize the actin cytoskeleton. Consequently, a reduced mating efficiency was observed for these cells. These overlapping phenotypes, together with the protein interaction data, suggest that Ste20 and Erg4 cooperate in cell polarization during mating.

Finally, we demonstrate that Erg4 and Ncp1 contribute to haploid invasive growth, although with opposing functions. The ERG4 deletion strain displayed hyperinvasive growth. Thus, Erg4 is a negative regulator of filamentous growth. By contrast, loss of NCP1 resulted in a complete loss of invasive growth due to a polarization defect. This is in accordance with the phenotype described for ste20Δ cells.

Due to redundancy (see below), the CBR1 deletion alone has no phenotypic effects. By contrast, the inactivation of CBR1 and NCP1 is lethal and results in bud morphology defects, which demonstrates that Cbr1, like Erg4 and Ncp1, plays a role in cell polarity

Erg4, Cbr1 and Ncp1 are all involved in the biosynthesis of ergosterol (Fig. 8), the predominant sterol in budding yeast. The biosynthesis of sterols in yeast can be divided into two stages. The early sterol biosynthetic pathway starts with acetyl CoA and provides essential cellular constituents, such as isopentenyl diphosphate, quinones and dolichol (Daum et al., 1998). It also provides squalene, the first specific intermediate of the ergosterol biosynthetic pathway. The proteins of the later sterol pathway, including Erg4, Ncp1 and Cbr1, catalyze the transformation of squalene to ergosterol (Daum et al., 1998) (Fig. 8). Whereas Erg4 catalyzes the final step in this pathway (Zweytick et al., 2000), Ncp1 and Cbr1 have a more indirect role. Ncp1 transfers electrons from NADPH to two cytochrome P450 enzymes involved in sterol synthesis, Erg5 (sterol 22-desaturase; CYP61) and Erg11 (sterol 14α-demethylase; CYP51) (Aoyama et al., 1989; Kelly et al., 1995) (Figs 8, 9). Erg1 (squalene epoxidase/squalene monooxygenase), another enzyme in the sterol biosynthetic pathway, also requires Ncp1 (Yoshida, 1988). Cbr1 (NADH cytochrome b₅ reductase) acts in a very similar way to Ncp1 and has overlapping electron acceptors. It transports electrons from NADH via Cyb5 (cytochrome b₅) to Erg11, and very probably to Erg1 and Erg5 as well (Lamb et al., 1999) (Figs 8, 9). Additionally, the Cbr1-Cyb5 electron transport system participates in other sterol biosynthetic reactions – such as C5-desaturation, catalyzed by Erg3, and C4-demethylation, catalyzed by Erg25, Erg26 and Erg27 (Reddy et al., 1977; Aoyama et al., 1981). The redundant functions of Cbr1 and Ncp1 explains why no phenotypes were observed for the CBR1-deletion strain. Furthermore, the lethality of the double mutant cbr1Δ ncp1Δ is in accordance with the biochemical data. These cells presumably die because the ergosterol biosynthesis pathway is perturbed at several steps.

In addition to our results, other groups have reported the involvement of ergosterol-synthesizing proteins in homotypic vacuole fusion and endocytosis (Kato and Wickner, 2001; Seeley et al., 2002; Pichler and Riezman, 2004). Cdc42 and its effectors also play a role in these processes. Ste20 and Cla4 promote actin patch assembly and, as a consequence of that, endocytosis via type I myosin phosphorylation (Wu et al., 1996; Evangelista et al., 2000; Geli et al., 2000; Lechler et al., 2000). By contrast, the role of Cdc42 and other polarity proteins in vacuole maintenance is poorly understood (Wickner, 2002). Recently, homologs of oxysterol-binding proteins, a family of proteins that regulate the synthesis and transport of sterols, were found to participate in Cdc42-dependant polarity (Kozminski et al., 2006). All these observations, together with our data, demonstrate the requirement of sterol synthesis for cell polarization and possibly for other functions involving Cdc42. It is important to notice that the major regulatory steps in the post-squalene biosynthetic pathway, catalyzed by Erg1 and Erg11 (Veen et al., 2003) (Fig. 8), both depend on electron transfer from Ncp1 and Cbr1. Therefore, the whole pathway might be regulated through Ste20. This is of particular interest because homologs of Ste20, Ncp1, Cbr1 and Cyb5 have been found in mammals, and four of the enzyme activities depending on Ncp1 and/or Cbr1, including the major regulatory steps of the pathway, are conserved throughout evolution (Fig. 8). Thus, this proposed regulatory module (Fig. 9) might be conserved from yeast to humans. Interestingly, emb-8, the NCP1 homolog in Caenorhabditis elegans, has a role in polarization during early embryonic development (Rappleye et al., 2003). Furthermore, it has recently been reported that, in Arabidopsis thaliana, squalene epoxidase (Erg1 in budding yeast), a major enzyme of the sterol biosynthetic pathway that depends on Ncp1 and Cbr1 (Fig. 8), regulates the polarized growth of roots and hypocotyl (Rasbery et al., 2007).

Although these observations support a conserved regulatory link between Ste20 and ergosterol pathway genes such as ERG4, CBR1 and NCP1, neither these findings nor our results indicate whether Erg4, Cbr1 and Ncp1 lie upstream or downstream of Ste20. Although the delocalization of Ste20-GFP in mating projections of erg4Δ cells supports the ergosterol pathway lying upstream of Ste20, we find it tempting to also speculate that Ste20 might activate these proteins. This important question will be addressed in the future.

Whereas Ste20 is enriched at sites of polarized growth at the plasma membrane (Peter et al., 1996), Erg4, Cbr1 and Ncp1 associate with the endoplasmic reticulum (ER) (Kreibich et al., 1983; Zweytick et al., 2000). This raises the issue of where and how these proteins interact. Although Ste20 is activated by Cdc42 at the plasma membrane, it is also present in the cytosol (Peter et al., 1996). Importantly, Cbr1 and Ncp1 face the cytoplasmic side of the ER (Kreibich et al., 1983) and could, therefore, be easily accessible for cytosolic Ste20. Alternatively, plasma membrane-bound Ste20 could interact with Erg4, Cbr1 and Ncp1 in the ER. It has been reported that a subfraction of the ER closely associates with the plasma membrane (Pichler et al., 2001). Interestingly, enzymes of sterol synthesis are enriched at these contact sites (Pichler et al., 2001).

Our data suggest that sterol composition of membranes is crucial for the establishment of cell polarity. It has been suggested that membrane microdomains rich in sterol and sphingolipids known as lipid rafts compartmentalize the plasma membrane and might play an important role in cell polarization (Rajendran and Simons, 2005). In budding yeast, sterol-rich membranes have been observed to localize in a polarized manner towards the tip of the mating projection (Bagnat and Simons, 2002). Thus, there they might contribute to the polarized localization of proteins involved in the mating process. However, it is not clear whether ergosterol molecules at the mating projection really form lipid rafts. Another open question is whether asymmetric lipid distribution is sufficient for polarized protein localization. Furthermore, the existence of lipid rafts is controversial and is still being debated (Munro, 2003). In addition, it has been questioned whether sterols are polarized in budding yeast (Valdez-Taubas and Pelham, 2003). Because endocytosis requires sterol (Pichler and Riezman, 2004) and the endocytic pathway is necessary and sufficient for cell polarization (Valdez-Taubas and Pelham, 2003), sterols could establish polarity via endocytosis. Further work will determine the role of sterols in cell polarity and whether Ste20 regulates sterol synthesis.

All yeast strains are listed in Table 1. The strains used in this study were in the YPH499 background with the exception of strains used for invasive growth and mating assays. For invasive growth experiments, strains in the Σ1278b background were used (MBY15, MBY16, MBY22, PPY966 and PPY1209). Strains for quantitative mating assays (ESM356, ESM357, THY580, THY581, THY595 and THY597) are derivatives of FY1679. All constructs used in this work are listed in Table 2.

For the split-ubiquitin screen, ste20Δ cells carrying pMET25-STE20-CUBI-RURA3 were transformed with a split-ubiquitin library and grown on 5-FOA plates for 3 days at 30°C. The interaction assays were performed as described previously (Johnsson and Varshavsky, 1994).

Genes were subcloned after PCR amplification to confirm the interaction.

GST and GST-Ste20 were expressed in E. coli BL21 (DE3) and purified using glutathione-sepharose (Amersham Biosciences). The immobilized GST and GST-Ste20 proteins were presented to a yeast lysate of 3HA-CBR1 and ERG4-9myc, respectively, for 9 minutes at 4°C in IP buffer (20 mM Tris pH 7.5, 100 mM NaCl, 10 mM EDTA, 1 mM EGTA, 5% glycerol, 1% NP-40, 1% BSA). After five washes with IP buffer, the associated proteins were eluted with sample buffer and analyzed by immunoblotting. Monoclonal mouse anti-HA (12CA5) was obtained from Roche. Mouse monoclonal anti-Myc (9E10) and rabbit polyclonal anti-Cdc11 were from Santa Cruz Biotechnology. Secondary antibodies were from Jackson Research Laboratories.

Cells carrying the plasmid pMT485 (GAL1-CLN1-3HA) were grown overnight in selective medium. Exponentially growing cells were induced with galactose for 4 hours, fixed for microscopic examination and prepared for immunoblot analysis.

Cells were inoculated to an OD₆₀₀ of 0.025 in YPD supplemented with various amounts of zymolyase 20T (Seikagaku) and grown overnight at 30°C. Various dilutions of the cell suspensions were then plated on YPD medium and grown for 2 days at 30°C to determine the number of colony forming units. Zymolyase sensitivity was expressed as percentage of growth compared with growth in YPD without zymolyase.

For agar invasion assays, 10⁵ cells of an overnight culture were spotted on YPD and grown for 2 days at 30°C. Plates were photographed before and after being rinsed under a gentle stream of deionized water. For analysis of filamentous colony morphology, cells were grown to stationary phase in synthetic complete medium, washed twice with water and spread onto synthetic complete medium lacking glucose at a concentration of 10⁵ cells/plate. Plates were incubated at 30°C for 18 hours and analyzed microscopically. Single cells of these microcolonies were analyzed by scraping off the colonies from the plates using a cell scraper and subsequent fixation with formaldehyde for morphological examination and actin staining.

Logarithmically growing cells were incubated with 1 μg/ml α-factor for 150 and 360 minutes. These cells were fixed with formaldehyde for morphological examination, analysis of GFP-tagged proteins and actin staining. For quantitative mating assays 3×10⁶ cells of each mating type were mixed and collected on nitrocellulose filters. The filters were placed on YPD plates for 4 hours. Filters were then suspended in water and serial dilutions were plated on double-selective SC plates to determine the number of diploids. Mating efficiency was calculated as the percentage of input cells that formed diploids.

F-actin of yeast cells was stained with rhodamine-phalloidin as described previously (Höfken and Schiebel, 2002). Cells were examined with a Zeiss Axiovert 200M fluorescence microscope equipped with a 100× Plan oil-immersion objective and images were captured using a Zeiss AxioCam MRm CCD camera.

We are very grateful to Nils Johnsson for providing us with the split-ubiquitin library and plasmids. We thank C. David Allis, Kathryn Ayscough, Peter Pryciak, Elmar Schiebel, Kazuma Tanaka and Mike Tyers for strains and plasmids. We thank Melanie Boß for excellent technical support. This work is part of the doctoral thesis of C.T. and the diploma thesis of D.H. The project was supported by the Deutsche Forschungsgemeinschaft (HO 2098/3-1).