Formin-dependent actin assembly is regulated by distinct modes of Rho signaling in yeast

The determination of cell shape and organization requires coordination between signal transduction pathways, the cytoskeleton, and membrane traffic. In the budding yeast Saccharomyces cerevisiae, the actin cytoskeleton, without participation of microtubules, polarizes cell growth (Bretscher, 2003). During bud growth, actin filaments assemble into cortical patches that cluster in the expanding bud, and into cables that extend along the axis of growth (Adams and Pringle, 1984). Cortical patches are implicated in endocytic internalization (Munn, 2001), and their assembly depends upon the actin-nucleating Arp2/3 complex (Winter et al., 1999). Actin cables function as tracks that guide myosin-V–driven secretory vesicle, organelle, and mRNA transport and nonmyosin-driven mitochondrial transport into the growing bud (Schott et al., 2002). The assembly of actin cables is dependent on the functionally redundant formins Bni1p and Bnr1p (Evangelista et al., 2002; Sagot et al., 2002a). In contrast to the assembly of patches, cable assembly is Arp2/3 independent (Evangelista et al., 2002).

Like many formins from animals and fungi, yeast Bni1p and Bnr1p contain an NH₂-terminal Rho-binding domain (RBD),^* a central proline-rich formin homology (FH) domain 1 that binds profilin, and a COOH-terminal FH2 domain (Kohno et al., 1996; Evangelista et al., 1997; Imamura et al., 1997) (Fig. 1, a and d). The isolated FH2 domain of Bni1p can serve as a nucleator for actin assembly in vitro (Pruyne et al., 2002; Sagot et al., 2002b). Because the FH2 domain has the novel capacity to remain bound to the barbed end of the assembling filament, Bni1p can potentially serve as a nucleator and filament anchor during actin cable assembly at the cell cortex (Pruyne et al., 2002).

Based on studies of the mammalian formins mDia1 and mDia2, the RBD is proposed to negatively regulate formin activity by binding a sequence COOH terminal to the FH2 domain termed the Dia autoregulatory domain (DAD) (Alberts, 2001; Palazzo et al., 2001). The association of GTP-bound Rho to the formin RBD relieves this interaction, thereby activating the formin. Consistent with this model, related DAD sequences are present in Bni1p and Bnr1p (Fig. 1), and Bni1p constructs lacking the RBD or DAD stimulate excessive filament assembly in vivo compared with full-length Bni1p (Evangelista et al., 2002; Sagot et al., 2002a).

Yeast contains six Rho-family GTPases (Cdc42p and Rho1–5p) that participate in multiple aspects of cell polarity, including septin organization (Holly and Blumer, 1999; Weiss et al., 2000; Gladfelter et al., 2002), the regulation of secretion (Adamo et al., 1999, 2001; Robinson et al., 1999; Guo et al., 2001; Zhang et al., 2001), and the stimulation of cell wall synthesis (Cabib et al., 1998). All six Rho proteins also have roles in regulating actin polarity. Cdc42p is required for actin polarization at bud emergence, with multiple effectors implicated in this process (Adams et al., 1990; Brown et al., 1997; Chen et al., 1997; Evangelista et al., 1997; Holly and Blumer, 1999; Bi et al., 2000; Lamson et al., 2002). Rho3p and Rho4p share a function in polarizing actin, though so far no physiologically relevant effectors have been defined (Matsui and Toh-e, 1992b; Imai et al., 1996). Rho1p functions through Pkc1p to regulate actin in a complex manner, driving MAPK-independent depolarization of the actin cytoskeleton in response to cell wall injuries and then stimulating a MAPK-dependent repolarization (Delley and Hall, 1999). Overexpressed Rho2p can substitute for the actin-polarizing activity of Rho1p (Marcoux et al., 2000), and Rho5p negatively regulates cell wall stress-induced actin depolarization (Schmitz et al., 2002), but effectors for these proteins are unknown.

The Bni1p and Bnr1p RBDs show two-hybrid interactions with multiple Rho proteins (Kohno et al., 1996; Evangelista et al., 1997; Imamura et al., 1997), but the contribution of each Rho protein to regulating formin-dependent actin cable assembly was not known. We find that multiple Rho-dependent pathways converge upon Bni1p and Bnr1p to regulate their activity, including the semiredundant Rho3p and Rho4p and the highly conserved homologues of RhoA (Rho1p) and CDC42 (Cdc42p). Thus, the formins are key targets for integrating signaling pathways in controlling actin polarity.

To first investigate whether any of the nonessential Rho proteins are required to stimulate actin cable assembly, the actin cytoskeleton was examined in yeast lacking the Rho isoforms Rho2p, Rho3p, Rho4p, and Rho5p. The deletion strains rho2Δ, rho4Δ, and rho5Δ all grew as well as wild type, whereas the rho3Δ strain grew more slowly (unpublished data), and each deletion strain displayed normal actin cables when stained for either actin (Act1p) or tropomyosin (Tpm1p) (Fig. 2 a). Thus, none of these Rho proteins alone is required to stimulate both Bni1p and Bnr1p.

We have previously shown that overexpression of full-length Bni1p causes the loss of extended actin cables in the mother cell and the appearance of excessive cable-like filaments in the bud. This shift from cables to cable-like filaments is thought to reflect excessive nucleation of filaments that depletes the cable-stabilizing protein tropomyosin (Evangelista et al., 2002). To evaluate whether Bni1p activity depends upon any of the nonessential Rho proteins, we expressed full-length myc-tagged Bni1p in the rho2Δ, rho3Δ, rho4Δ, and rho5Δ strains. Cable-like filaments accumulated in the buds of wild-type, rho2Δ, rho4Δ, and rho5Δ cells but not in rho3Δ yeast (Fig. 2, b and c). Western blotting showed that rho3Δ yeast produce Bni1p–myc to the same extent as wild-type cells (Fig. 2 d), and reintroduction of RHO3 into the rho3Δ strain restored cable-like filaments with overexpression of Bni1p (Fig. 2, b and c), showing that overexpressed Bni1p requires Rho3p to assemble filaments in the bud.

As Bni1p and Bnr1p perform an essential function in the assembly of actin cables (Evangelista et al., 2002; Sagot et al., 2002a), we generated a strain lacking both RHO3 and BNR1 to examine whether Bni1p expressed at endogenous levels also requires RHO3. Consistent with a strong dependence of actin cable assembly by Bni1p upon Rho3p, the rho3Δbnr1Δ cells grew very poorly and contained few actin cables (Fig. 3, a and b) but accumulated actin bars, a form of aggregated monomeric actin common to cytoskeletal mutants.

To explore whether the requirement of Rho3p is to activate full-length Bni1p or if it plays a different role in filament assembly, constitutively active Bni1p lacking the regulatory RBD (Bni1pΔRBD; Fig. 1 b) was overexpressed in wild-type and rho deletion strains. In all strains, including rho3Δ yeast, Bni1pΔRBD stimulated massive assembly of cable-like filaments (Fig. 2, e and f). Furthermore, Bni1pΔRBD expressed from the BNI1 promoter was able to rescue the synthetic sick phenotype and restore normal actin cables to bnr1Δrho3Δ yeast (Fig. 3, a and b). Thus, the full-length formin Bni1p requires Rho3p to stimulate filament assembly, but this requirement can be bypassed by a formin lacking the Rho-binding regulatory region.

The viability of the rho3Δbnr1Δ strain suggested that Bni1p has some residual ability to assemble cables in the absence of Rho3p. One possible explanation is that other Rho proteins play a role in this process. One likely candidate is Rho4p. Simultaneous deletion of RHO3 and RHO4 is lethal, and overexpression of Rho4p can suppress growth defects of rho3Δ cells (Matsui and Toh-e, 1992a), indicating that the two share some redundant function. Double conditional rho3 rho4 mutants display severe cytoskeletal defects (Matsui and Toh-e, 1992b; Imai et al., 1996), yet the effectors for these Rho isoforms in organizing the cytoskeleton are unknown. If they share an essential role in formin activation, we predicted that expression of the constitutively active Bni1pΔRBD might bypass the requirement for these two Rho proteins. Indeed, expression of Bni1pΔRBD from the BNI1 promoter rescued the viability of the double rho3Δrho4Δ mutant (Fig. 3 c). An attractive explanation is that the normal function of Rho3p and Rho4p is to bind the Bni1p RBD to relieve the inhibitory interaction between the RBD and the DAD.

In mammalian cells, expression of the mDia DAD has been shown to also bypass the requirement for Rho activation of a formin by competing with the DAD of the endogenous full-length formin for binding the RBD (Alberts, 2001). To more specifically determine whether Rho3p and Rho4p play a role in regulating the interaction of the Bni1p RBD and DAD, we generated a construct containing the DAD of Bni1p (Fig. 1 c). Expression of this construct was able to rescue growth and permit actin cable assembly in rho3Δrho4Δ yeast (Fig. 3, c and d). As expected, rescue by the Bni1p DAD was dependent upon full-length formins, because Bni1pDADCOOH was unable to rescue the lethality of bni1Δbnr1Δ (unpublished data). These results suggest that Rho3p and, to a lesser extent, Rho4p share a role in regulating the interaction between the RBD and DAD of the yeast formin Bni1p.

Rho4p is the only Rho protein that binds the Bnr1p RBD, as analyzed by two-hybrid or in vitro binding assays (Imamura et al., 1997). To ascertain whether Rho4p is the preferred Rho activator for Bnr1p, the growth rates of wild-type, rho4Δ, bni1Δ, and rho4Δbni1Δ strains were compared. No obvious growth defects were seen in any of these strains, and the actin cables of rho4Δbni1Δ were normal, indicating that Bnr1p can function in the absence of Rho4p (Fig. 3 d).

The strong dependence of Bni1p upon Rho3p suggests that the actin cables present in rho3Δ yeast are dependent upon Bnr1p. Consistent with this, the growth rates of rho3Δ and rho3Δbni1Δ strains were similar (unpublished data), and the actin cables of rho3Δbni1Δ were normal (Fig. 3 d), suggesting that Bnr1p can also function in the absence of Rho3p. However, the lethality of rho3Δrho4Δ yeast can be rescued by an activated version of Bnr1p (Bnr1pΔRBD) (Fig. 3 c), suggesting that yeast lacking Rho3p and Rho4p are deficient in all formin function. Also, the deletion of RHO3 results in slow growth, but this can be rescued by the activated Bni1ΔRBD (unpublished data), suggesting that the slow growth of our rho3Δ strain results from diminished formin activity. Notably, loss of BNI1 in our strain background does not show a slow growth phenotype, suggesting that the loss of RHO3 also diminishes activation of Bnr1p. Thus, although Bnr1p-mediated filament assembly can occur in the presence of Rho3p or Rho4p, activation by Rho3p appears to be more important for normal growth. Our combined results indicate that the shared essential function of Rho3p and Rho4p is to activate the formins, specifically through disrupting the RBD–DAD interaction, and that Rho3p plays a more important role in this process.

Loss of Cdc42p function causes multiple defects in polarized growth (Kozminski et al., 2000). However, many conditional alleles arrest as unbudded cells that grow isotropically (Adams et al., 1990; Miller and Johnson, 1997; Kozminski et al., 2000), similar to the terminal phenotype of double conditional bni1bnr1 mutants (Evangelista et al., 2002). To determine whether Cdc42p regulates Bni1p or Bnr1p activity, we examined whether actin cables are present in a temperature-sensitive cdc42-1 strain. When cdc42-1 yeast were shifted to 35°C, polarized actin cables were lost from unbudded cells but, to a large extent, were retained in budded cells (Fig. 4, a and b). The defect in the unbudded cdc42-1 cells did not appear to be a loss of cables but an inability to properly organize them. The unbudded cdc42-1 cells retained actin cables to a similar extent as wild-type cells but in a disorganized distribution.

To determine whether these phenotypes were specific to the cdc42-1 allele or were a general phenotype for loss of Cdc42p function, five conditional alleles (cdc42-101, cdc42-118, cdc42-123, cdc42-124, and cdc42-129) isolated from a collection of 37 site-directed cdc42 mutants (Kozminski et al., 2000) were examined for actin cable defects under restrictive conditions. Similar to cdc42-1, all five alleles showed a profound loss in polarized actin cables in unbudded cells yet retained a polarized network of cables in budded cells (Fig. 4 c). Also consistent with cdc42-1, all the cdc42 mutants retained actin cables in unbudded cells in a disorganized distribution (unpublished data). Because Bni1p and Bnr1p are required for actin cable assembly throughout the cell cycle (Evangelista et al., 2002; Sagot et al., 2002a), we conclude that Cdc42p is not essential for formin activation. Consistent with this, the expression of the activated formin construct Bni1pΔRBD was unable to rescue any of the defects of the cdc42-1 mutant; actin cables still became disorganized in unbudded cells at the restrictive temperature, and the mutants arrested as large and unbudded cells (unpublished data). Thus, Cdc42p appears to be important for the proper organization of cables during bud emergence but not for the bulk assembly of cables that occurs in unbudded cells.

Overexpression of Cdc42p is reported to bypass the requirement for Rho3p and Rho4p (Matsui and Toh-e, 1992b). To examine the ability of the various Rho isoforms to replace Rho3p and Rho4p function, we introduced CDC42, RHO1, RHO2, or RHO5 on high-copy plasmids into rho3Δrho4Δ yeast grown in the presence of pRS316-RHO3 and then tested each strain for its ability to lose the RHO3 plasmid. Consistent with Matsui and Toh-e (1992b), CDC42 was able to rescue the lethality of rho3Δrho4Δ, and the rescued cells displayed normal actin cables (Fig. 5, a and b), indicating that at least one yeast formin was active under these conditions. However, none of the other three Rho proteins was able to rescue the lethality of rho3Δrho4Δ (unpublished data). Cdc42p was also able to bypass the synthetic sickness and actin cable defects of rho3Δbnr1Δ yeast (unpublished data), indicating that when overexpressed, Cdc42p can relieve the dependence of Bni1p activation upon Rho3p.

The essential functions of Rho1p in cell wall synthesis and in actin organization have been dissected using conditional mutations specifically defective for one function or the other (Saka et al., 2001). One mutation, rho1-2, was shown to have a depolarized actin cytoskeleton after 5 h at the restrictive temperature (Helliwell et al., 1998). We found that when shifted to 37°C, rho1-2, but not RHO1, cells lost actin cables after 15 min, suggesting that Rho1p might also be important for Bni1p- and Bnr1p-stimulated filament assembly (Fig. 6, a and b). In support of this, when full-length Bni1p was overexpressed in rho1-2 cells, cable-like filaments accumulated in the bud at the permissive temperature, but after 15 min at 37°C, these cable-like filaments were no longer present (Fig. 6, c and d). To determine if Rho1p is required for the activation of Bni1p, we expressed Bni1pΔRBD from the BNI1 promoter in rho1-2 cells. When shifted to 37°C, actin cables now remained in the rho1-2 cells (Fig. 7, a and b), although viability was not rescued, suggesting that Rho1p is required for the activation of Bni1p and Bnr1p, but failure in formin activation is not the only defect of rho1-2 cells.

Rho1p is able to bind several putative effectors, but the growth and polarity defects of rho1-2 mutants are specifically suppressed by an activated form of the Rho1p effector protein kinase C (Pkc1p*) (Helliwell et al., 1998). Expression of Pkc1p* from the PKC1 promoter restored normal actin cables to rho1-2 cells at 37°C (Fig. 8 a), and when Pkc1p* was coexpressed with full-length Bni1p in rho1-2 cells, the activated kinase permitted cable-like filament assembly at 37°C (Fig. 8, b and c). Thus, activated Pkc1p can provide the Rho1p-dependent signal for activation of full-length Bni1p.

Signaling from Rho1p to the actin cytoskeleton bifurcates at the level of Pkc1p (Delley and Hall, 1999). One signaling pathway functions through the cell wall stress response MAPK cascade composed of Bck1p (MAPK kinase kinase), Mkk1/2p (MAPK kinases), and Slt2p (MAPK). Pkc1p targets of the second pathway are not known. Depending upon strain background, deletions of the Slt2p MAPK cascade can have either severe or mild effects upon the cell, resulting either in slow growth, cell lysis, and temperature sensitivity, or in normal growth but temperature sensitivity (Mazzoni et al., 1993; Delley and Hall, 1999). To determine whether the MAPK cascade is involved in formin activation, we examined mutants lacking components of the cascade from both types of backgrounds. In both a bck1Δ strain with severe growth defects and a slt2Δ strain with mild defects, actin cables were present at room temperature and throughout a 1-h shift to 37°C (unpublished data). Expression of Slt2p from a 2μ high-copy plasmid can rescue the polarity of cortical actin patches in rho1-2 yeast but cannot rescue viability (Helliwell et al., 1998). However, overexpression of Slt2p was unable to restore actin cables in rho1-2 yeast at 37°C and was also unable to allow overexpressed Bni1p to generate cable-like filaments in rho1-2 yeast (unpublished data), suggesting that the Slt2p MAPK cascade is not involved in Rho1p-stimulated activation of the formins.

Although PCK1 is essential, a deletion can be rescued by a dominant active allele of the MAPK kinase kinase BCK1 (BCK1-20) (Lee and Levin, 1992; Levin and Bartlett-Heubusch, 1992). To determine whether Pkc1p is required for Rho1p-dependent activation of the formins, pkc1Δ BCK1-20 cells were examined at several temperatures. At room temperature, actin cables were present (unpublished data), but when shifted to 37°C, actin cables disassembled in the pck1Δ BCK1-20 cells within 15 min, just like the rho1-2 mutants (Fig. 8, d and e). A temperature-sensitive PKC1 allele (pkc1-2ts) (Lee and Levin, 1992) yielded similar results (unpublished data), whereas wild-type PKC1 or PKC1 BCK1-20 controls showed no loss of actin cables. Thus, Rho1p works through its effector Pkc1p to activate the formins in a MAPK-independent manner, though this signaling appears to be required only at elevated temperatures.

To determine whether Pkc1p signaling can bypass the requirement for Rho3p-dependent activation of Bni1p, Pkc1p* was expressed in rho3Δbnr1Δ yeast. The presence of the activated kinase was unable to rescue the slow growth of this strain (Fig. 8 f) or restore actin cables and eliminate actin bars (Fig. 8 g), suggesting that Pkc1p cannot bypass the requirement for Rho3p to activate Bni1p. To determine whether Rho3p can bypass the need for Rho1p/Pkc1p signaling in activating Bni1p, we coexpressed full-length Bni1p with an activated GTPase-deficient allele of RHO3 (RHO3-V25) (Adamo et al., 1999) from a CEN plasmid in rho1-2 yeast. The presence of activated Rho3p was unable to bypass the requirement of Bni1p-stimulated filament assembly for Rho1p (unpublished data), indicating that both Rho1p/Pkc1p and Rho3p are required for efficient activation of Bni1p at 37°C. Similarly, overexpression of CDC42 was unable to suppress the rho1-2 mutant at 37°, either with respect to growth or the actin cable defect. Based on the loss of all cables in rho1-2 yeast, Rho1p/Pkc1p is likely to also be required for the activation of Bnr1p.

The formins provide a crucial target for regulating cell polarity in yeast. As nucleators of actin filaments, the formins define the spatial organization of the actin cable array, which in turn guides polarized secretion, organelle segregation, and mitotic spindle orientation (Evangelista et al., 2002; Pruyne et al., 2002; Sagot et al., 2002a,b). We have found that four of yeast's six Rho proteins are involved in regulating the formins Bni1p and Bnr1p in several distinct ways.

Rho3p and Rho4p were previously known to share an essential role in yeast growth (Matsui and Toh-e, 1992a), and we find here that this function appears to be regulation of the inhibitory interaction between the RBD and DAD of the yeast formins. Thus, these two Rho proteins become dispensable in cells expressing Bni1p or Bnr1p from which the RBD has been deleted, or when the RBD–DAD interaction is disrupted by overexpression of exogenous DAD sequence. By several criteria, Rho3p appears to be the more important of the two GTPases. Loss of Rho3p almost completely eliminates Bni1p-dependent filament assembly. Furthermore, rho3Δ yeast grow slowly, but this slow growth can be bypassed by an activated Bni1p, suggesting that the growth defect is related to a loss of signaling to the formins. As bni1Δ does not cause a similar decrease in growth rate in our strain background, we suggest that the loss of Rho3p eliminates a significant component of Bnr1p activation.

An attractive model would have Rho3p and Rho4p activate the formins directly. GTP-bound forms of multiple Rho proteins show two-hybrid interactions with the RBD of the formins, although the reported interactions do not completely coincide with our in vivo analysis; Bni1p RBD interacts with GTP-Cdc42p, -Rho1p, -Rho3p, and -Rho4p (Kohno et al., 1996; Evangelista et al., 1997), but Bnr1p interacts only with GTP-Rho4p (Imamura et al., 1997). Our results suggest that the most physiologically important interaction for Bni1p is with Rho3p. However, with no reported interaction between Bnr1p and Rho3p, it is possible that signaling between these two molecules is indirect.

In addition to formins, Rho3p and Rho4p can bind to several proteins involved in polarized secretion, including the exocyst proteins Sec3p and Exo70p and (for Rho3p) the myosin-V Myo2p (Adamo et al., 1999; Robinson et al., 1999; Guo et al., 2001). Furthermore, mutants for RHO3 show defects in exocytosis (Adamo et al., 1999), but the contributions of the Rho proteins to secretion and to actin organization seem to be distinct. Many suppressors of rho3Δ growth defects encode proteins involved in the secretory pathway, but these are not able to rescue rho3Δrho4Δ lethality (Matsui and Toh-e, 1992b; Imai et al., 1996; Kagami et al., 1997). Also, defects in the function of Sec3p or Myo2p do not directly affect the organization of actin cables (Haarer et al., 1996; Schott et al., 1999). Finally, rho3 alleles specifically defective in secretion retain a normal actin cytoskeleton (Adamo et al., 1999), suggesting that they are not defective in activating the formins. Despite the importance of Rho3p for exocytosis, the growth defects of our rho3Δ mutants appear to be limited to formin activation, as they can be rescued by an activated allele of Bni1p. This discrepancy might reflect a difference in strain background or a difference in growth conditions, such that the requirement for Rho3p in exocytosis does not impact the growth rate under the conditions used in this study.

A previous screen identified two high-copy suppressors of rho3Δrho4Δ lethality, the highly conserved Rho family member Cdc42p and the Cdc42p-binding scaffold protein Bem1p (Matsui and Toh-e, 1992b). We found that overexpressed Cdc42p could activate the formins to generate actin cables in rho3Δrho4Δ yeast and could specifically activate Bni1p in rho3Δbnr1Δ yeast. The ability of Cdc42p to associate with the Bni1p RBD (Evangelista et al., 1997; Imamura et al., 1997) supports the possibility that Cdc42p might directly activate Bni1p.

The need to overexpress Cdc42p to rescue the rho3Δrho4Δ lethality suggests that the normal role of Cdc42p is more restricted. Previous work demonstrated that Cdc42p plays an important role in organizing actin during bud emergence (Adams et al., 1990), and we found that the loss of Cdc42p function in six conditional cdc42 alleles specifically compromised actin cable organization in unbudded cells, but cables appeared to be unperturbed in budded cells. Notably, the levels of actin cables did not appear to be diminished in the unbudded cells, suggesting that the formins were still activated, presumably by Rho3p and Rho4p. This suggests that Cdc42p plays a role in properly recruiting the active formins to the nascent bud site. In fact, previous results have demonstrated that Bni1p is unable to localize in the absence of Cdc42p function (Jaquenoud and Peter, 2000; Ozaki-Kuroda et al., 2001). The inability of cdc42-ts strains to organize actin cables in unbudded cells can account for two other phenotypes of the arrested cells, isotropic growth and depolarized cortical patches, because both of these phenotypes can also arise as a secondary consequence of loss of transport along actin cables (Pruyne et al., 1998; Schott et al., 1999; Evangelista et al., 2002; Sagot et al., 2002a). The ability of Cdc42p to replace Rho3p and Rho4p when overexpressed may reflect a subtle role for Cdc42p in regulating formin activity for cable polarization under normal conditions.

The conserved RhoA homologue, Rho1p, makes an additional, independent contribution to the activation of the formins. Loss of Rho1p function at 37°C in rho1-2 yeast eliminates both actin cable assembly by endogenous formins and cable-like filament accumulation stimulated by exogenous Bni1p. The expression of the activated Bni1pΔRBD construct restored cables to rho1-2 yeast at the restrictive temperature, though the viability of the rho1-2 yeast was not rescued. Despite the ability of Rho1p to bind the Bni1p RBD in two-hybrid assays (Kohno et al., 1996), our evidence suggests that Rho1p does not directly regulate the formin RBD–DAD interaction but acts indirectly through the Rho1p effector, Pkc1p. Yeast without functional Pkc1p also lost actin cables at 37°C, whereas expression of an activated kinase Pkc1p* restored cables to rho1-2 yeast and permitted overexpressed Bni1p to generate ectopic filaments in rho1-2 yeast. The Pkc1p/Rho1p dependence of formin activation was only observable at elevated temperatures, suggesting that some other change in the cell status under these conditions impinges on the formins such that a Pkc1p-dependent signal is required to maintain their activity.

The Pkc1p-dependent signaling did not appear to act through the downstream MAPK cascade, as cells lacking MAPK components retained actin cables even at 37°C, and overexpression of the MAPK Slt2p was unable to restore cables in rho1-2 yeast or restore Bni1p function in rho1-2 yeast. As further confirmation that formin function and MAPK signaling are distinct events, defects in formin function (e.g., bni1Δ) and loss of MAPK signaling (e.g., slt2Δ) have additive deleterious effects (Fujiwara et al., 1998). While this paper was in preparation, it was reported that Rho1p and formins are necessary for the assembly of the contractile ring at elevated temperatures (Tolliday et al., 2002). These findings are consistent with our results and suggest that the role of Rho1p in that process might also be indirect through Pkc1p.

Thus, we find that three distinct Rho-dependent signals regulate the proper function of the formins. Rho3p and Rho4p share a critical role in activating the formins. This activity cannot be replaced by Rho1p/Pkc1p signaling, either through Rho1p overexpression or through activation of Pkc1p. Cdc42p can replace Rho3p/Rho4p when overexpressed, but its role appears to normally be restricted to organizing the formins for bud emergence. Similarly, the requirements for Rho1p and Pkc1p appear to be restricted to conditions that trigger a cell wall stress response (i.e., 37°C). Thus, regulation of the formins may require multiple events, for example, recruitment of the formin, a specific phosphorylation of the RBD, and the binding of a Rho protein to the RBD. A similar phenomenon is seen with Pkc1p, where binding of Rho1 and phosphorylation by the redundant kinases Pkh1/2p contribute to Pkc1p activation (Inagaki et al., 1999). It will be interesting to determine whether the animal Rho1p homologue, RhoA, regulates formins in a similarly indirect manner through its Pkc1p-related kinase effectors, the PRKs (Amano et al., 1996; Watanabe et al., 1996; Vincent and Settleman, 1997).

The multiple inputs to Bni1p and Bnr1p activation link formin-mediated actin assembly into multiple essential regulatory pathways. Undoubtedly, other inputs contribute to the regulation of formin activity. For example, osmotic stabilization can also rescue viability of rho3Δrho4Δ (Matsui and Toh-e, 1992b), suggesting that high osmolarity activates an alternative signaling pathway to bypass the requirement for Rho3p/Rho4p. The conservation of formins and their roles in regulating cytoskeletal organization suggests that these principles of multiple activating inputs are likely to be conserved across the eukaryotes.

All yeast strains used in this study are described in Table I. Single RHO deletion strains were obtained from American Type Culture Collection except the rho3Δ strains. The rho3Δ strains were generated by transformation of the diploid strain ABY500 with a PCR-derived kanamycin resistance knockout construct generated from template pUG6 (Guldener et al., 1996) using primers 5′-CTTTATAACAACAAGTTTTTCATTAAAATCCATAGTAGCACAAGGTAAAGGGAAACAAACCAGCTGAAGCTTCGT-3′ and 5′-TTTGTTTAGTCTCTCTTTTTGGGTTCTATTCTTTTGTGTGTGCGTCCGTGTCTTATTATAGCATAGGCCACTAGT-39, followed by sporulation and isolation of kanamycin-resistant segregants. The deletion was confirmed by PCR amplification of the RHO3 locus using primers outside the deleted region.

p2529 (CEN URA3 GAL1-BNI1-13myc) and p2528 (CEN URA3 GAL1-BNI1ΔRBD-13myc) were as previously described (Evangelista et al., 2002). pPKC1* (YCp50-PKC1^R398P) was a gift from M. Hall (University of Basel, Basel, Switzerland) (Helliwell et al., 1998). pRS316-RHO3V25 was a gift from P. Brennwald (University of North Carolina at Chapel Hill, Chapel Hill, NC) (Adamo et al., 1999). pRS313-RHO3 (CEN HIS3 RHO3) was constructed by introduction of a 1.19-kb PCR-amplified fragment, including the RHO3 open reading frame and 290-bp upstream and 200-bp downstream sequences, inserted between XbaI and SacI sites of pRS313 (Sikorski and Hieter, 1989). PCR was performed with primers 5′-GCTCTAGACATCCACTCCCAGG-3′ and 5′-TCGAGCTCACATGCTGGAGG-3′. To generate pYD518 (CEN TRP1 GAL1-BNI1-13myc), the GAL1-10 promoter from p2529 was ligated between the EcoRI and BamHI sites of pRS314 (Sikorski and Hieter, 1989), and the 6.4-kb BNI1-13myc fragment from p2529 between the BamHI and NotI sites. To generate plasmids pYD125 (CEN HIS3 BNI1ΔRBD-myc) and pYD126 (CEN LEU2 BNI1ΔRBD-myc) for expression of BNI1ΔRBD behind the BNI1 promoter, a 450-bp fragment of the BNI1 promoter region was amplified by PCR with primers and cloned into the XhoI and BamHI sites of pRS313 (for pYD125) or pRS315 (for pYD126). Subsequently, a 5.1-kb BNI1ΔRBD-13myc fragment from p2528 was ligated between the BamHI and NotI sites. To generate pYD127 (CEN LEU2 BNI1-DAD-C-terminus-myc) for expression of the BNI1 DAD and COOH terminus from the BNI1 promoter, a PCR fragment encompassing the 480 3′-terminal bases of BNI1 plus ∼500 bp of additional myc and downstream sequences was amplified from p2529 using primers 5′-CGGGATCCATGGATAGGCGCGCTGTTAT-3′ and T7 and then cloned between the BamHI and NotI sites of pYD126, replacing BNI1ΔRBD. To generate pYD128 (CEN LEU2 BNR1ΔRBD) for expression of BNR1ΔRBD behind the BNI1 promoter, a 2.9-kb BNR1ΔRBD fragment with EcoRI and XbaI sites was amplified by PCR with primers 5′-CGGAATTCATGAAGGCAAAAGATAGTCCCG-3′ and 5′-GCTCTAGAATCTGTCCATCTCCAAATC-3′, and the BNI1 promoter fragment with XhoI and EcoRI sites was amplified by PCR with primers 5′-CCCTCGAGTTTTTGACGTGGAAGAG-3′ and 5′-CGGAATTCTTCCTTTCCTTCTCTTCC-3′, and then a three-piece ligation was performed to clone the BNI1 promoter and BNR1ΔRBD between the XhoI and XbaI sites of pRS315. To construct pRS425-CDC42 (2μ LEU2 CDC42), a 1.18-kb fragment, including the CDC42 open reading frame and 380-bp upstream and 220-bp downstream sequences, was amplified by PCR with primers 5′-TCGAGCTCGCATTAAAGATGTCTTC-3′ and 5′-GCTCTAGAGCGTTGAAGAACATGG-3′ and ligated between the SacI and XbaI sites of pRS425 (Christianson et al., 1992). All PCR-derived constructs were verified by sequencing.

Cells were prepared, fixed, and stained using antibodies to actin, tropomyosin, and the myc epitope as previously described (Pruyne et al., 1998; Evangelista et al., 2002). Cells were categorized as unbudded, small budded (bud lengths <1/3 the length of the mother cell), medium budded (bud lengths <2/3 the length of the mother cell but ≥1/3 the length of the mother cell), or large budded (bud length ≥2/3 the length of the mother cell). For each assay, 100 cells of the indicated categories from asynchronous cultures were scored. When scoring for the presence of actin cables, cells of all categories were scored positive if cables were visible by eye, and negative if not. When scoring for the presence of polarized actin cables, cells of all categories were scored as positive if the majority of cables present in a cell were aligned along the axis of the cell or emanated from a nascent bud site. Cells were scored as negative for polarized actin cables if no cables were detectable or if half or more of the cables present in a cell were not aligned along the growth axis or associated with a bud site. When scoring for the accumulation of cable-like filaments in the bud, small- and medium-budded cells were scored positive if the stain in the bud showed a clearly visible increase in fluorescence beyond that seen in wild-type cells. Note that for the images shown in this paper, the intensity has been digitally reduced to allow dimmer portions of the displayed cells to be visible. The increased intensity of stain due to formin-stimulated accumulation of filaments is clearly discernible by eye when compared with controls. For the galactose induction experiments, midlog phase cultures grown in defined raffinose medium were induced by the addition of 2% galactose for 2 h.

Cells were grown up to midlog phase in defined raffinose medium and induced by the addition of 2% galactose for 2 h. Samples were equalized based on OD₆₀₀, and extracts were isolated as previously described (Horvath and Riezman, 1994) and resolved by SDS-PAGE. Blots were probed with 9E10 (anti-myc) or B28 (anti–yeast actin).

We would like to thank Marie Evangelista, Michael Hall, Ruth Collins, Tim Huffaker, Clarence Chan, Susan Henry, David Levin, Keith Kozminski, and Patrick Brennwald for providing plasmids and yeast strains and Charles Boone and Daniel Schott for critical reading of the manuscript.

This work was supported by the National Institutes of Health (GM39066) to A. Bretscher.