Regulation of the Formin for3p by cdc42p and bud6p

Abstract

Formins are conserved actin nucleators responsible for the assembly of diverse actin structures. Many formins are controlled through an autoinhibitory mechanism involving the interaction of a C-terminal DAD sequence with an N-terminal DID sequence. Here, we show that the fission yeast formin for3p, which mediates actin cable assembly and polarized cell growth, is regulated by a similar autoinhibitory mechanism in vivo. Multiple sites govern for3p localization to cell tips. The localization and activity of for3p are inhibited by an intramolecular interaction of divergent DAD and DID-like sequences. A for3p DAD mutant expressed at endogenous levels produces more robust actin cables, which appear to have normal organization and dynamics. We identify cdc42p as the primary Rho GTPase involved in actin cable assembly and for3p regulation. Both cdc42p, which binds at the N terminus of for3p, and bud6p, which binds near the C-terminal DAD-like sequence, are needed for for3p localization and full activity, but a mutation in the for3p DAD restores for3p localization and other phenotypes of cdc42 and bud6 mutants. In particular, the for3p DAD mutation suppresses the bipolar growth (NETO) defect of bud6Δ cells. These findings suggest that cdc42p and bud6p activate for3p by relieving autoinhibition.

INTRODUCTION

Formins are key regulators of the actin cytoskeleton and form a large family conserved in all eukaryotes (Wallar and Alberts, 2003; Faix and Grosse, 2006). These proteins are necessary for the formation of numerous actin structures, including stress fibers, filopodia, cytokinetic actin rings, junctional actin structures, or actin cables. The proper regulation of formins is likely to be critical for cellular processes such as cell migration, cytokinesis, cell adhesion, and cell polarity.

A well characterized biochemical activity of formins is to nucleate and elongate linear actin filaments (Pruyne et al., 2002; Sagot et al., 2002b; Kovar et al., 2003; Li and Higgs, 2003; Moseley et al., 2004). This activity occurs through the formin-homology (FH) 2 domain, which dimerizes to form a doughnut-shaped structure containing multiple actin-binding sites in its core (Xu et al., 2004; Otomo et al., 2005b). This dimer is thought to stabilize otherwise unstable intermediates in the assembly of new actin filaments, and it binds processively to the fast-elongating barbed end of existing actin filaments (Pring et al., 2003; Kovar and Pollard, 2004). The adjacent FH1 domain binds profilin-actin and helps accelerate the elongation of actin filaments (Chang et al., 1997; Evangelista et al., 1997; Watanabe et al., 1997; Romero et al., 2004; Kovar et al., 2006). The budding yeast formin Bni1p also binds the actin monomer-binding protein Bud6p, which, similar to profilin, stimulates the activity of the FH2 domain (Moseley and Goode, 2005). In addition to their activity in actin filament nucleation and elongation, some formins have been suggested to also function in actin bundling and severing, further contributing to the remodeling of actin structures (Harris et al., 2004, 2006; Moseley and Goode, 2005; Harris et al., 2006).

The potent activity of formins needs to be carefully regulated in vivo to generate the correct actin structures at the right place and time. Autoinhibition has been proposed to be an important mechanism by which formin activity is controlled. Initially discovered in diaphanous-related formins, this mechanism relies on the interaction of a conserved C-terminal motif, named Diaphanous Autoregulatory Domain (DAD), with the Diaphanous Inhibitory Domain (DID) in the N-terminal region (Watanabe et al., 1999; Alberts, 2001; Li and Higgs, 2003, 2004; Wallar et al., 2006). Support for this model largely came from the observation that exogenous expression of truncated formin constructs, lacking either N or C terminus, or harboring a mutated DAD region, led to an overabundance of actin structures, such as filopodia or stress fibers (Watanabe et al., 1999; Tominaga et al., 2000; Koka et al., 2003; Schonichen et al., 2006; Wallar et al., 2006). It is not clear whether all formins use an autoinhibitory mechanism for regulation, because many formins do not contain an obvious DAD-like sequence at the primary sequence level (Higgs and Peterson, 2004).

The autoinhibition of many formins is regulated by small Rho GTPases. Different formins may be regulated by different Rho GTPases; for example, Cdc42 is thought to regulate the formins mDia2 and FRLα in mammalian cells, whereas Rho3p and Rho4p regulate Bni1p in the budding yeast (Dong et al., 2003; Peng et al., 2003; Seth et al., 2006). Generally, specific Rho(s) bind to formins at an N-terminal G protein binding-domain partially overlapping with the DID domain. This binding activates formins by relieving the autoinhibitory interaction between the DAD and DID, because Rho competes with the DAD for DID binding (Dong et al., 2003; Lammers et al., 2005; Otomo et al., 2005a; Rose et al., 2005; Nezami et al., 2006). In vitro experiments have shown that the activity of the FH1–FH2 domain is indeed inhibited by binding of the N terminus to the DAD (Li and Higgs, 2003). Binding of active Rho not only regulates the biochemical activity of formins but also may control their ability to localize to the cell cortex (Seth et al., 2006). N-terminal truncations or overexpression of the DAD domain of the formin Bni1p have also been shown to suppress growth defects in rho mutants in budding yeast, suggesting that activation of formins is a major function of these rho proteins (Dong et al., 2003). Genetic and biochemical studies are beginning to indicate the existence of additional modes of formin regulation, such as phosphorylation and through interactions with other proteins (Dong et al., 2003; Li and Higgs, 2003; Matheos et al., 2004; Moseley et al., 2004; Martin et al., 2005; Eisenmann et al., 2007).

In this study, we have focused on for3p, one of the three formins in the fission yeast Schizosaccharomyces pombe. Fission yeast cells are rod-shaped and grow exclusively at cell tips. The formin for3p localizes to cell tips, and it is responsible for the assembly of a polarized array of actin cables (Feierbach and Chang, 2001; Nakano et al., 2002). These cables are bundles of short linear actin filaments largely oriented with their barbed ends facing the cell tip (Kamasaki et al., 2005). Actin cables contribute to polarized cell growth, in part, by acting as tracks for the delivery of myosin V-driven vesicles to cell tips (Schott et al., 2002; Salas-Pino and Chang, unpublished observations). Analyses of for3p movements suggest that the activity and localization of for3p is highly dynamic on the time scale of seconds (Martin and Chang, 2006). At the cell tip, for3p may be active at cell tips for actin assembly for only a few seconds, before it is released onto assembling actin cables, and travels away from cell tips with the retrograde flow of actin in cables. Because only 20–30% of the total for3p protein is present at cell tips at any given time, much of for3p protein is probably inactive. How for3p is localized and regulated at the cell tip is not well understood. Based upon sequence comparisons, for3p was categorized as a DAD-less formin (Higgs and Peterson, 2004). It has been shown to bind to activated rho3p and cdc42p but not other rhos, in two-hybrid assays (Nakano et al., 2002). However, the physiological significance of these interactions is not clear, especially because the function of these rhos in actin cable assembly has not yet been elucidated.

The formin for3p plays an important role in regulating the growth pattern of fission yeast cells. After cell division, cells initially grow in a monopolar manner, and then they initiate polarized growth at the second cell end in the G2 phase of the cell cycle (Martin and Chang, 2005). This transition to bipolar growth, termed new end take-off (NETO), relies on the localization of for3p to the new end. In this process, the regulation of for3p is dependent on tea1p, tea4p, and bud6p, all of which are necessary for NETO and reside in large complexes with for3p at cell tips (Verde et al., 1995; Feierbach and Chang, 2001; Glynn et al., 2001; Martin et al., 2005). Similar to its Saccharomyces cerevisiae orthologue, bud6p binds to the C terminus of for3p and is required for proper for3p localization and robust actin cable assembly (Feierbach and Chang, 2001; Ozaki-Kuroda et al., 2001). However, the precise functions of bud6p are unknown.

Here, we show that for3p is regulated by autoinhibition through an intramolecular interaction. We demonstrate that cdc42p is the primary Rho-GTPase for actin cable assembly and that it functions to relieve for3p autoinhibition. We further show that bud6p binds for3p at a site overlapping with the DAD, and that all bud6Δ phenotypes can be rescued by preventing for3p autoinhibition, suggesting that bud6p also regulates for3p autoinhibition. The proper localization of the formin depends on relief of autoinhibition and the binding of multiple sites on the formin to cortical docking factors. Finally, these studies address how regulation of for3p contributes to controlling bipolar cell growth.

MATERIALS AND METHODS

Yeast Strains, Media, and Genetic Methods

The S. pombe strains used in this study are listed in Supplemental Table 1. Standard methods for S. pombe media and genetic manipulations were used throughout. All plasmids were sequenced. Tagged strains were constructed using a polymerase chain reaction (PCR)-based approach and confirmed by analytical PCR.

The for3DAD* and for3ΔFH3 mutations were created by PCR stitching in two rounds of PCR to generate pGBD-for3(1261-1461)LLT-AAA and pREP42-EGFP-for3NΔ(353-406), respectively. All plasmids used in this study are listed in Supplemental Table 2. To generate the for3DAD* and for3ΔFH3 alleles in vivo, the genomic for3 DAD and FH3 regions were replaced with ura4⁺ in a wild-type strain, to yield strain YSM816 and YSM327, respectively. These strains were transformed with a linearized for3DAD* or for3ΔFH3 fragment. ura4⁻ colonies were selected on 5-fluoroorotic acid, and integration was confirmed by PCR and sequenced.

To isolate cdc42 mutants, we introduced random mutations into the entire region of cdc42 by the error-prone PCR method (Cadwell and Joyce, 1992). PCR was carried out with AmpliTaq DNA polymerase (PerkinElmer-Cetus, Boston, MA) in manufacturer-supplied reaction buffer mixtures containing 1 mM dCTP/dTTP and 0.2 mM dGTP/dATP, 2.5 mM MgCl₂, with 100 ng/100 μl reaction mixture of purified pSKcdc42 as a template. Oligonucleotide primers 5′-TATATACATATGCCCACCATTA AGTGTGTC-3′ and 5′-TATA- GGATCCTTACAGTACCAAACACTTTGAC-3′ were designed to amplify the entire cdc42 open reading frame (ORF) flanked by NdeI–BamHI (underlined). The collection of cdc42 ORFs was cloned into a Bluescript plasmid, in which 500 base pairs of the cdc42 5′ region and 500 base pairs of the cdc42 3′ region had previously been cloned on either side of the kanamycin resistance gene (KanMX). The ORFs were cloned adjacent to the 5′ region by using the NdeI–BamHI sites. A second PCR amplification was used to generate a DNA fragment, including 5′ region-cdc42-KanMX-3′region, with free ends homologous to the flanking regions of cdc42⁺. This fragment was used to transform an h+ leu1-32 ura4D-18 (PPG103) strain. Transformant clones were selected in YES+G418 plates at 25°C and screened for thermosensitivity at 36.5°C. Colonies that exhibited temperature-sensitive (ts) growth were selected and the cdc42 allele carried by one of them was named cdc42-1625 and analyzed further.

To generate hemagglutinin (HA)-cdc42 and HA-cdc42-1625 strains, a genomic version of cdc42⁺ or cdc42-1625 with the HA epitope coding sequence fused at the beginning of the ORF was generated by inserting in a Bluescript plasmid 500 base pairs of the 5′ cdc42 flanking sequences, the HA sequence, the cdc42⁺ or cdc42-1625 ORF, the ura4⁺ gene, and 500 base pairs of the 3′ cdc42 flanking sequence. The insert was transformed into an h+leu1-32 ura4D-18 (PPG103) strain, and stable transformants were selected and screened by PCR for the appropriate gene replacement

We found that cdc42-1625 bud6 double mutants were strictly dependent on the presence of a pREP41-HA-cdc42 plasmid for growth; although this plasmid was lost from >20% of cdc42-1625 single mutant cells grown in nonselective media for >10 generations, it was not lost from cdc42-1625 bud6Δ cells, demonstrating that this double mutant is not viable.

Microscopy

Microscopy was performed using either a widefield fluorescence microscope or a spinning disk confocal microscope essentially as described previously (Martin and Chang, 2006), and images were acquired, processed, and analyzed with the OpenLab software (Improvision, Coventry, United Kingdom).

To prepare cells for live imaging of green fluorescent protein (GFP) or red fluorescent protein (RFP)-tagged proteins, strains were grown in Edinburgh minimal media supplemented with appropriate amino acids and 5 times the amount of adenine overnight at 30°C to log phase (or 25°C for cdc42-1625 strains and corresponding wild-type controls) and observed at room temperature (22–24°C), unless otherwise noted. To image live actin cables, GFP–CH domain expression was induced for 16 h in medium lacking thiamine. To image GFP-for3N, expression was induced for 14–16 h in medium lacking thiamine. For all other imaging, cells were grown in YE5S medium at 30 or 25°C, unless noted otherwise. Actin staining was performed as described using AlexaFluor 488-phalloidin (Invitrogen, Carlsbad, CA) with a fixation time of 40 min (Pelham and Chang, 2001). For actin staining of GFP-tagged strains, AlexaFluor 488-phalloidin was also used, because the GFP signal was not resistant to the fixation and staining procedure, and thus it did not interfere with imaging of the actin cytoskeleton (data not shown). We prepared 100X stock solutions of latrunculin A (LatA) (provided by Phil Cruz, University of California, Santa Cruz) in dimethyl sulfoxide.

For HA–Cdc42p localization, cells were fixed with methanol-free 16% formaldehyde (Polysciences, Warrington, PA) for 40 min as described (Hagan and Hyams, 1988). Cells were incubated for 16 h at 4°C with a 1:200 dilution of anti-HA antibody in PEMBAL. Then, 3 × 1 h washes with PEMBAL [100 mM piperazine-N, N′-bis(2-ethanesulfonic acid) pH 6.9, 1 mM EGTA, 1 mM MgSO₄, 1% bovine serum albumin, 100 mM L-lysine hydrochloride, 0.1% NaN₃] at room temperature were performed. Cells were incubated for 2 h at room temperature with a 1:1000 dilution of anti-mouse AlexaFluor 594-conjugated antibody; Invitrogen). Then, we preformed 3 × 1 h washes with PEMBAL at room temperature before cells were analyzed under the microscope.

Data Analysis

For3p movements were analyzed with kymographs, which were constructed using the “Volume Slicing” tool of the OpenLab software, as described previously (Martin and Chang, 2006). Fluorescence activation after photobleaching (FRAP) analysis was performed as described previously (Martin and Chang, 2006).

Quantification of the fluorescence intensity of actin cables was performed using ImageJ software (National Institutes of Health, Bethesda, MD). We drew a line roughly perpendicular to the long axis of the cell across actin cables in the middle third of the cell (making sure the line did not cross actin patches), and we measured the peaks of the fluorescence profile along this line with the “plot profile” tool. From each maximum value, we subtracted the background fluorescence (level of the fluorescence profile outside the cell). Note that the fluorescence value is arbitrary and varies from one experiment to the other depending on imaging conditions, but that under identical staining and imaging conditions, in five of five experiments for3DAD* cells show stronger actin cables than wild-type cells. Length and width of cells was measured on calcofluor-stained septated cells in the middle of the short and long axis, respectively.

Two-Hybrid Analysis, In Vitro Binding Assays, and Western Blotting

Two-hybrid interactions were tested on medium lacking histidine in strain AH109 (Clontech, Mountain View, CA). Maltose binding protein (MBP) and MBP-for3N(1-702) were expressed in E. coli, affinity purified on an amylose resin (Clontech), and stored at 4°C on the resin in the presence of sodium azide. 6His-tagged for3C(630-1461) and for3CDAD*(630-1461) were expressed in E. coli, purified on nickel columns, eluted in 250 mM imidazole, and stored at −80°C. In vitro binding assays were performed in microbio-spin columns (Bio-Rad, Hercules, CA) by adding equivalent amounts of His-tagged protein diluted eight times in B buffer (20 mM Tris, pH 8.0, 20 mM KCl, 130 mM NaCl, 1 mM MgCl₂, and 2 mM EDTA) to MBP fusion-coupled amylose resin and incubating for 2 h at 4°C. The resin was subsequently washed three times with B buffer and one time with B buffer + 0.05% NP-40 and two times with B buffer. MBP-fusions and associated proteins were eluted in 50 μl of 10 mM maltose. After addition of sample buffer, samples were analyzed by SDS-polyacrylamide gel electrophoresis, Coomassie staining, and Western blotting.

Antibodies used were as follows: monoclonal mouse anti-HA (HA.11, Covance; or 12CA5, Roche Diagnostics, Indianapolis, IN), monoclonal mouse anti-polyhistine (HIS-1; Sigma-Aldrich, St. Louis, MO), monoclonal mouse anti-GFP (clones 7.1 and 13.1; Roche Diagnostics), anti-actin antibody (MP Biomedicals, Irvine, CA), and monoclonal anti-α-tubulin (TAT).

RESULTS

Cdc42p Is Necessary for Cell Morphology, Actin Cables, and for3p Localization

Small Rho GTPases bind and regulate formins in many organisms. In the fission yeast, activated alleles of rho3p and cdc42p have been shown to bind the N-terminal region of for3p (Nakano et al., 2002); however the functional significance of these interactions has not been explored. For3p localization and actin cable organization seemed normal in rho3Δ cells (data not shown), suggesting that rho3p may not be the major regulator of for3p. Therefore, we focused our attention to cdc42p.

cdc42⁺ is an essential gene (Miller and Johnson, 1994), and so we generated point mutant alleles (see Materials and Methods). In this study, we characterized one of these alleles, cdc42-1625. Strains bearing this allele were temperature sensitive for growth at 36°C (Figure 1, A and B). This growth defect was rescued by wild-type cdc42⁺ and the constitutively active allele cdc42G12V expressed from plasmids, but not by the inactive allele cdc42T17N (Figure 1B), indicating that cdc42-1625 is a hypomorphic allele.

Figure 1.

Cdc42p is necessary for cell morphology, actin cable organization and for3p localization. (A) Growth curve of wild-type and cdc42-1625 cells incubated at 28 and 36°C. (B) Rescue of the temperature-sensitive growth defect of cdc42-1625 by different cdc42 alleles. cdc42-1625 cells were transformed with the expression plasmid pREP41-HA carrying either no insert, wild-type cdc42⁺, constitutively active cdc42G12V, or dominant-negative cdc42T17N alleles and grown on Edinburgh minimal media plates for 2 d at 25 and 36°C. (C) Immunoblot of HA-cdc42p and HA-cdc42p-1625 expressed from the endogenous promoter. Twenty-five micrograms of total yeast extracts from exponential cultures at 25 and 36°C were loaded in each lane. Levels were monitored with the anti-actin antibody. Although wild-type cdc42p shows increased levels at 36°C, this increase failed to happen in the cdc42-1625 mutant. (D) Immunofluorescence of HA-cdc42p and HA-cdc42p-1625 with anti-HA antibody. Both wild-type and mutant cdc42p localize to growing cell tips. Differential interference contrast (DIC) of the permeabilized cells is shown at the bottom. (E) Morphology of wild-type and cdc42-1625 cells grown to log phase at 28 and 36°C for 5 h. DIC and calcofluor-staining images are shown. (F) Rescue of cell morphology (DIC; top) and actin cables (AlexaFluor-phalloidin; bottom) by wild-type cdc42⁺ in cdc42-1625 cells at 36°C. (G) Projection images of spinning disk confocal stacks of AlexaFluor 488-phalloidin–stained cdc42-1625 (top) and wild-type (bottom) cells grown at 25°C (left) or 36°C for 1 h (right). Note that actin cables are extremely weak, but they were still present in cdc42-1625 cells (arrowheads). (H) Single focal plane widefield fluorescence images of myo52p-tomato in cdc42-1625 (top) and wild-type (bottom) cells grown at 25°C. (I) Single focal plane widefield fluorescence images of for3p-3GFP in cdc42-1625 (top) and wild-type (bottom) cells grown at 25°C. Note that for3p largely fails to localize to cell tips in the mutant cells. All bars, 2 μm.

DNA sequencing of this allele revealed a single mutation that substitutes the alanine residue at position 159 for valine. This residue, close to the C terminus, is conserved in Cdc42 from different species, including mammals, and maps to one of the four Cdc42 domains implicated in GTP binding and hydrolysis (Johnson, 1999). Immunoblots showed that the cdc42p-1625 mutant protein was expressed but exhibited faster mobility and decreased levels at 36°C compared with wild-type cdc42p (Figure 1C). In contrast, wild-type cdc42p levels were higher at 36°C than at 25°C. The cdc42p-1625 protein was still properly localized to growing cell tips and septa, suggesting that this mutation does not affect membrane association (Figure 1D). cdc42-1625 mutant phenotypes described below were apparent at both 25 and 36°C, suggesting that function was lost even at 25°C. The growth defect at 36°C may be due to a higher requirement of cdc42p at elevated temperatures. Subsequent experiments described in this article were thus conducted at 25°C, except where noted.

We next tested the effect of cdc42p on cell morphology and actin cable formation. cdc42-1625 cells were misshapen and sometimes displayed misplaced septa (Figure 1E). Cells showed more pronounced morphological defects and increased cell lysis at 36°C (Figure 1E). Significantly, cdc42-1625 cells had very few short, weakly staining actin cables at both 25 and 36°C, as shown by AlexaFluor-phalloidin staining (Figure 1G). Actin patches were depolarized from cell tips in >90% of cells (n > 100), a possible consequence of the cable defect. To probe the function of actin cables, we also examined the distribution of a barbed end-directed type V myosin, which is normally targeted to cell tips by moving on actin cables (Motegi et al., 2001). Using a myo52p-tomato fusion, we found that myo52p dots failed to accumulate at cell tips in >90% of cdc42-1625 cells (n = 200; Figure 1H), consistent with defects in actin cable organization. Expression of cdc42⁺ in cdc42-1625 cells rescued the morphological and actin cable defects (Figure 1F).

To examine for3p distribution, we used a for3p fusion construct tagged with three tandem copies of GFP; this fusion is expressed from the chromosomal for3 locus from the endogenous promoter and is functional (Martin and Chang, 2006). In cdc42-1625 strains, for3p-3GFP failed to accumulate at cell tips (Figure 6I). Thus, similar to what has been seen with many diaphanous-related formins (Watanabe et al., 1997; Gasman et al., 2003; Gasteier et al., 2003; Peng et al., 2003; Seth et al., 2006), cdc42p is required for the efficient localization of full-length for3p to cell tips. Thus, cdc42p is the primary Rho protein in fission yeast responsible for regulation of actin cable assembly and formin for3p localization.

Figure 6.

Cdc42p relieves for3p autoinhibition to regulate its localization. (A) Single focal plane widefield fluorescent image of for3p-3GFP (left) and for3pDAD*-2GFP (right) in cdc42-1625 cells. Note that the DAD* mutation is sufficient to restore for3p localization to cell tips in cdc42-1625 mutant cells (arrowheads). (B) DIC images of for3-3GFP cdc42-1625 (left) and for3DAD*-2GFP cdc42-1625 (right) cells. The DAD* mutation improves the morphology of cdc42-1625 mutant cells. (C) Projection images of spinning disk confocal stacks of AlexaFluor 488-phalloidin–stained for3-3GFP cdc42-1625 (left) and for3DAD*-2GFP cdc42-1625 (right) cells. Arrowheads point at weak actin cables. Note that the DAD* mutation fails to restore wild-type actin cables in cdc42-1625 mutant cells.

For3p Contains Three Localization Domains

To study the mechanisms responsible for for3p regulation, we next used a mutational approach to define sites on for3p that affect its localization and activity. We and others previously showed that the N terminus of for3p (amino acids [aa] 1-702; for3N) is sufficient to localize to cell tips (Nakano et al., 2002; Martin et al., 2005) (Figure 2B). This N-terminal region also contains a Rho-binding site (Nakano et al., 2002). Residues in this region can be aligned, albeit poorly, to the DID and dimerization domains of mDia (Supplemental Figure 1). The predicted FH3 region (aa 306–507, as defined by SMART) corresponds to the C-terminal part of this DID and dimerization-like domains.

Figure 2.

For3p contains three independent localization domains. (A) Scheme of for3p fragments that were fused to GFP to assay localization. All constructs were expressed from a plasmid under control of the medial-strength nmt1 promoter. (B) Projection images of spinning disk confocal stacks of GFP-for3N (left), GFP-for3(137-515) (middle), and GFP-for3NΔ(353-406) (right). Yellow arrowheads indicate cell tips and septum localization. GFP-for3N is also detected at the spindle pole body. (C) Projection images of spinning disk confocal stacks of GFP-for3C (left), GFP-for3C treated with 200 μM LatA (middle), and GFP-for3C-I930A (right). Blue arrowhead shows actin cable localization. Yellow arrowheads indicate cell tips and septum localization.

Further truncation analyses showed that the DID and dimerization-like domains (aa 137–515) were sufficient for cell tip localization (Figure 2B). In contrast, a 54 amino acid in-frame deletion within the DID (for3NΔ353-406) abrogated cortical association of the for3N fragment (Figure 2B). Because this region contains a potential dimerization domain (Otomo et al., 2005a; Rose et al., 2005), we verified that for3N was not localizing to the cortex just by binding to the endogenous full-length for3p: GFP-for3N also localized normally in a for3Δ mutant (data not shown). Thus, the N-terminal portion of for3p contains a localization domain in a region of for3p that overlaps with the DID.

Additional sites in the C-terminal portion of the protein also seem to contribute to for3p localization. We expressed the C-terminal half of for3p (aa 630-1461; for3C), including the FH1 and FH2 domains, as a GFP fusion from a plasmid. GFP-for3C localized to two distinct cellular structures: actin cables and cell tips (Figure 2C). Both localizations were present even when the fragment was expressed in for3Δ mutants (data not shown). The cable staining was distributed evenly throughout the cable, in a different pattern from the discrete dots seen with full-length protein. GFP-for3C also localized to actin rings, but not actin patches. This cable localization was mediated by actin binding in the FH2 region, because a point mutation in the actin-binding interface within the FH2 domain (I930A) abolished cable localization but not the cell tip localization of for3C (Figure 2C). The cell tip distribution was clearer with this mutant, or when actin structures were disrupted with LatA (a drug that sequesters actin monomers) (Figure 2C). Together, these results using for3p fragments reveal three distinct localization domains: two cortical targeting regions that localize it to the cortex and the central FH2 domain that mediates actin cable association.

For3p Exhibits an Intramolecular Interaction Mediated by a DAD-like Domain

In diaphanous-like formins, an intramolecular DID–DAD interaction controls formin activity. We asked whether a similar mode of regulation operates in for3p. We found that for3p N- and C-terminal halves interact with each other, by using in a two-hybrid assay and also with in vitro binding assays with purified protein fragments (Figure 3). Although initial sequence searches did not reveal obvious homology to other DAD sequences (Higgs and Peterson, 2004), mapping the interaction region in the C terminus defined a 25-amino acid region necessary for binding to the N terminus of for3p (Figure 3A). This “DAD-like” region did not have strong homology with canonical DAD domains, but it did contain two leucines and a few basic residues that could be aligned with other DAD domains (Figure 3B). To test whether this putative DAD-like domain was responsible for this interaction, we mutated these residues to alanines—(1396)LLT(1398), or R(1409), K(1411), and K(1413). In the two-hybrid assay, these mutations abolished binding to a fragment of the for3p N terminus. These mutations specifically affected binding to the N terminus of for3p, because they did not affect binding to bud6p or localization of for3C to cell tips (Figure 3, A–C; data not shown) (Feierbach et al., 2004).

Figure 3.

Identification of a DAD-like region in for3p necessary for interaction with the N terminus. (A) Mapping of the DAD domain by two-hybrid assay. Interaction of C-terminal for3p fragments cloned in the pGBD vector with pGAD-for3N(1-702) was assayed by growth on −His plates. (B) Sequence alignment of for3p DAD with defined DAD domains from S. cerevisiae, Drosophila, and mouse formins. Stretches of basic residues are shown in blue. Mutation of the residues indicated in red to alanines abolished interaction with for3p N terminus. (C) Two-hybrid assay on SC-His plate of pGAD-for3N(1-702), pGAD-bud6C(581-1385) and empty pGAD with wild-type, LLT-AAA, and RKK-AAA pGBD-for3C(1261-1461). Numbers refer to constructs indicated in A. Mutations in the DAD region specifically abolish interaction with for3N but not with bud6C. (D) 6His-for3C binds directly and specifically to MBP-for3N. Binding of bacterially expressed proteins was assayed by affinity column. 6His-for3C (aa 630-1461) binds to MBP-for3N(1-702), but not MBP alone. This interaction is compromised by the LLT-AAA mutation (DAD*). 6His-tagged protein fragments were detected by Western blotting with an anti-6His antibody. Coomassie staining of MBP-fusions indicates equal loading.

Phenotype of a for3DAD* Mutant

We next tested the function of this intramolecular interaction in vivo. Because previous studies on formin autoinhibition commonly carry caveats of overexpression, we expressed a for3p DAD mutant protein at endogenous levels. The wild-type for3⁺ gene on the chromosome was replaced with a for3DAD* mutation that carries one of the DAD mutations described above [(1396)LLT(1398) to alanines mutation] in the full-length for3 gene. This for3 allele was thus expressed from the endogenous for3 promoter as the sole for3 gene in the cell. Western blot analysis confirmed that for3pDAD* is expressed at the same level as wild-type for3p (Figure 4A).

Figure 4.

Phenotype of the for3DAD* mutant. (A) Immnunoblot of for3p-HA and for3pDAD*-HA expressed from the endogenous promoter. Twelve micrograms of total yeast extract were loaded in each lane. Levels were monitored with the TAT anti-α-tubulin antibody. (B) FRAP analysis of for3p-3GFP and for3pDAD*-2GFP. Using a laser scanning microscope, the entire cell tip of cells expressing either for3p-3GFP or for3pDAD*-2GFP from the endogenous promoter was photobleached and imaged every 4 s thereafter to monitor fluorescence recovery. Each trace represents the average value for the indicated number of experiments. Error bars represent the SE. (C) Single focal plane widefield fluorescence images of for3p-3GFP and for3pDAD*-2GFP expressed from the endogenous promoter. Note that for3pDAD* localizes to cell tips, if not even more tightly than wild type for3p. (D) Projection images of spinning disk confocal stacks of AlexaFluor 488-phalloidin stained wild-type (top) and for3DAD* (bottom) cells. Note that actin cables in the for3DAD* mutant stain more brightly than in wild-type cells. (E) Quantification of the fluorescence intensity of actin cables in wild-type and for3DAD* cells. We measured the peaks of the fluorescence profile along a line drawn across actin cables and subtracted background value. A histogram of these values is shown. Note that the fluorescence value is arbitrary and varies from one experiment to the other, but that under identical staining and imaging conditions, for3DAD* cells always show stronger actin cable staining than wild-type cells.

This for3pDAD* mutant protein was localized properly. To assess the localization of for3pDAD*, we tagged it with two tandem copies of GFP. For3pDAD*-2GFP localized efficiently to cell tips and septa. Time lapse showed dynamic behavior indistinguishable from wild-type for3p. Small dots of for3pDAD*-2GFP displayed linear movements away from cell tips at the same rate as wild-type for3p-3GFP (rate of 0.29 ± 0.09 μm/s, n = 63 vs. 0.30 ± 0.12 μm/s for wild-type under the same conditions, n = 42), and FRAP analysis showed turnover rates at cell tips similar to wild-type for3p (t_1/2 = 10–12 s) (Figure 4B and Supplemental Movies 1 and 2). The only slight difference we noticed was that for3pDAD*-2GFP seemed to display a somewhat tighter distribution at cell tips than wild-type for3p-3GFP (Figure 4C). Thus, the dynamic behavior of for3p is not primarily controlled by DAD-dependent regulation. In addition, because the rate of for3p dot movement corresponds to that of actin cable assembly (Martin and Chang, 2006), this suggested that the rate of actin polymerization was also not affected by the DAD mutation.

for3DAD* cells had consistently more robust actin cables than wild-type, as revealed by phalloidin staining (Figure 4D). These cables were arranged in a longitudinal manner as in wild type, but quantification of their fluorescence intensity showed that they were more brightly stained than their wild-type counterpart. This result was confirmed and quantified in five independent experiments comparing side-by-side staining of nontagged and GFP-tagged for3 and for3DAD* cells. One representative quantification is shown in Figure 4E. We probed the dynamic properties of these cables by inhibiting actin polymerization with LatA. The cables in for3DAD* cells depolymerized in the same time period as those in wild-type cells, suggesting that the increased amounts of actin polymer in these cables was not due to overstabilization of actin filaments (data not shown). Thus, these cells may have more actin filaments in the cables; this result may be explained by the slightly increased amounts of for3DAD* at cell tips.

The DAD* mutation had no noticeable effect on the viability and growth of cells. for3DAD* cells were also morphologically normal, retaining their characteristic rod shape and growth patterns (Figure 8D). However, we noticed that for3DAD* cells were slightly, but reproducibly, longer than wild-type cells, suggesting a very modest hyperpolarization (14.48 ± 0.89 μm in wild-type vs. 14.99 ± 0.94 μm in for3DAD* mutant cells; Student's t test, p < 0.00001; n = 140).

Figure 8.

Mutation of the for3p DAD domain restores wild-type actin cables and bipolar growth in bud6Δ mutants. (A) Projection images of spinning disk confocal stacks of AlexaFluor 488-phalloidin–stained for3-3GFP (wild-type), bud6Δ for3-3GFP (bud6Δ), and bud6Δ for3DAD*-2GFP (bud6Δ for3DAD*) cells. Note that the intensity of actin cables is lower in bud6Δ cells compared with wild-type cells, but that this phenotype is suppressed by the DAD* mutation. (B) Quantification of the fluorescence intensity of actin cables in for3-3GFP, for3DAD*-2GFP, bud6Δ for3-3GFP, and bud6Δ for3DAD*-2GFP cells, as in Figure 4E. (C) Calcofluor staining of bud6Δ cells expressing either for3-3GFP (wt for3) or for3DAD*-2GFP (for3DAD*), showing monopolar and bipolar growth, respectively. (D) Quantification of new end growth in strains of the indicated phenotypes. The length between the inner side of the birth scar (black line in the calcofluor staining) and the tip of the cell was measured, as shown in C. This length is of up to 1.5 μm in cells that fail to initiate growth at the new end.

In summary, these data suggest that interaction between the N and C termini of for3p does have an inhibitory effect on actin cable assembly in vivo. However, this DAD-dependent autoinhibition has only a relatively mild effect, and it is not responsible for many parameters involved in actin cable assembly, such as the rate of actin filament assembly, actin stabilization, or formin release from the cortex.

Function of the DID Domain

We also tested the function of the N-terminal DID domain. Because of the poor conservation of the region with the DID domains of other formins (Supplemental Figure 1), we were unable to identify specific residues that might mediate cdc42p or DAD binding. Thus, we generated a small in-frame deletion in the region [for3Δ(353-406); Figure 5A]. This mutation was introduced into the context of the full-length for3⁺ gene and expressed at the endogenous locus as sole for3 copy. Based upon our truncation analysis described above, we predicted that this allele may be defective in aspects of cell tip localization, autoinhibition, and interaction with cdc42p.

Figure 5.

Deletion within the for3p DID leads to disorganized actin cables. (A) Scheme of the Δ353-406 deletion within full-length for3p. (B) DIC image of for3Δ353-406-myc cells. (C) Projection images of spinning disk confocal stacks of AlexaFluor 488-phalloidin–stained wild-type and for3Δ(353-406)-myc cells. Note the disorganized appearance of actin cables in for3Δ(353-406)-myc cells. Yellow arrowheads indicate locations from which some actin cables seem to radiate. (D) Single focal plane widefield fluorescence images of myo52p-GFP in wild-type (left) and for3Δ(353-406)-myc (right) cells. (E) Single focal plane widefield fluorescence images of for3p-3GFP and for3pΔ(353-406)-2GFP (right). (F) Single focal plane widefield fluorescence images of for3p-3GFP (left) and for3pΔ(353-406)-2GFP (right) in cells treated with 200 μM LatA. Red arrowheads highlight cell tip localization of for3pΔ(353-406)-2GFP.

for3Δ(353-406) cells had a more severe phenotype than the DAD* mutant. The cells were aberrantly shaped, displaying round, lemon or pear shapes, similar to for3Δ cells (Figure 5B). for3Δ(353-406) cells displayed an impressive network of actin cables, which were often disorganized or radiating from a focal point away from the cell tip (Figure 5C). Live imaging of actin cables confirmed these results and showed how cables were dynamic and often looped and organized transversally in the cell (Supplemental Movies 3 and 4). Actin patches were delocalized from cell tips in 73% of cells (n = 104). In for3Δ(353-406) cells, myo52p dots failed to accumulate to cell tips, but they still moved in linear paths in the cell, suggesting that the organization of actin cables was disrupted. (Figure 5D and Supplemental Movie 5). Thus, the DID domain is required for the correct organization of actin cables and for proper cell polarization.

We examined the localization of for3pΔ(353-406) by tagging it with two GFPs. For3pΔ(353-406) failed to localize efficiently to cell tips, but it still formed motile dots that moved in a linear manner in the cell (Figure 5E; data not shown). Curiously, treatment with LatA led to accumulation of this protein to cell tips. This effect suggested that in the absence of LatA, the mutant for3p may be preferentially bound to actin cables; however, after the actin cables are depolymerized in LatA treatment, for3p may be targeted to cell tips through the C-terminal cortical localization domain (Figure 5F), similar to what was seen in the for3C fragment.

Cdc42p Regulates for3p Localization by Relief of Autoinhibition

We next tested how cdc42p may affect for3p localization and/or activity. As shown above, for3p-3GFP largely failed to accumulate at cell tips in cdc42-1625 strains. We considered two ways by which cdc42p could affect for3p localization: first, cdc42p binding at the N terminus could directly tether for3p to cell tips; second, cdc42p may be needed to alter the conformation of for3p by antagonizing for3p autoinhibition (Seth et al., 2006).

If cdc42p primarily regulates for3p autoinhibition, we predicted that the DAD* mutation would bypass the need for cdc42p-mediated activation. Consistent with this model, we found that for3pDAD*-2GFP efficiently localized to cell tips in cdc42-1625 cells (Figure 6A). Thus, the localization defect seen in cdc42 mutants is not due to loss of a cortical docking site, but rather to an inability to activate for3p.

for3DAD*-2GFP also restored a more cylindrical cell morphology to the cdc42 mutants, making them longer and thinner (Figure 6B). Cell length and width were 14.23 ± 1.08 and 4.85 ± 0.44 μm in cdc42-1625 for3DAD*-2GFP (n = 70) compared with 12.88 ± 1.07 and 5.53 ± 0.54 μm in cdc42-1625 for3-3GFP (n = 97), respectively (Student's t-test, p < 1 × 10⁻¹³). The localization of markers for cell polarity such as actin patches and type V myosin were also improved by the for3DAD* mutation: myo52p-tomato dots and actin patches accumulated at the cell tip in 44 and 37% of cells, respectively (compared with <10% of cdc42-1625 cells, n >150). However, this for3 allele suppressed only some the defects associated with the cdc42-1625 mutant. In particular, actin cable staining was not significantly restored (Figure 6C). This suggested that cdc42p has other effectors involved in actin cable assembly or stability.

The for3 DAD Mutation Suppresses a bud6 Mutant

Another potential for3p activator is bud6p. In both budding and fission yeasts, bud6p directly binds the C terminus of formins (Bni1p and for3p, respectively), and it is necessary for their efficient cortical localization and for efficient actin cable assembly (Ozaki-Kuroda et al., 2001; Feierbach et al., 2004; Moseley et al., 2004). For3p–3GFP localization to cell tips was partly compromised in bud6Δ cells, similar to what was shown previously. This localization defect was due to defects in cortical association and not to excessive actin cable binding, because it was not improved by LatA treatment (Figure 7B).

Figure 7.

Bud6p targets for3p to cell tips through both anchoring and relief of autoinhibition. (A) Mapping of the BBS in for3p by two-hybrid analysis. Interaction of C-terminal for3p fragments cloned in the pGBD vector with pGAD-bud6C(581-1385) was assayed by growth on −His plates. Note how the minimum interaction domain overlaps with the DAD region. (B) Single focal plane widefield fluorescent images of for3p-3GFP in bud6Δ cells. Cells on the right were treated with 200 μM LatA. (C) Single focal plane widefield fluorescent images of for3pDAD*-2GFP bud6Δ cells. Cells on the right were treated with 200 μM LatA. Note that the DAD* mutation is sufficient to restore efficient for3p localization to cell tips in bud6Δ mutant cells (blue arrowheads). (D) Projection images of spinning disk confocal stacks of GFP-for3N (left) and GFP-for3C (middle), and GFP-for3C-I930A (right) expressed from the medial-strength nmt1 promoter in bud6Δ mutant cells. GFP-for3C fails to localize to cell tips. (E) Single focal plane widefield fluorescent images of cdc42p-GFP in wild-type and bud6Δ cells. (F) Single focal plane widefield fluorescent images of bud6p-GFP in wild-type and cdc42-1625 cells.

We defined the minimal bud6p binding site (BBS) of for3p by using the two-hybrid system (Figure 7A). Interestingly, we found that the minimal BBS overlaps significantly with the DAD in for3p. The S. cerevisiae BBS on Bni1p has also been mapped to a site adjacent to the DAD (Moseley et al., 2004). However, key amino acids mediating S. cerevisiae Bud6p binding are not conserved in the S. pombe BBS, suggesting that this site is not well conserved, at least on the level of primary sequence. We note that residues that affect DAD activity were separable from those mediating bud6p association, because the specific for3DAD* mutation, which disrupted intramolecular binding, did not affect bud6p binding (Figure 3C).

Bud6p binding in the vicinity of the DAD raised the question of whether bud6p may affect autoinhibition. As with cdc42, we found, importantly, that the DAD* mutation restored for3p localization to cell tips in bud6Δ cells (Figure 7C). Thus, autoinhibition, not a lack of a cortical tether, may prevent full-length for3p from localizing properly in bud6Δ mutants.

for3 DAD* Rescues Actin Cable Assembly and Bipolar Growth Defects in bud6Δ Cells

The for3DAD* mutation also rescued other phenotypes associated with bud6Δ cells. Actin cables are generally less robust in bud6Δ cells (Feierbach et al., 2004). The for3DAD* mutation was sufficient to restore actin cables in these cells, as measured by fluorescence intensity of AlexaFluor-phalloidin staining (Figure 8, A and B).

Bud6p has also been shown to be required for the efficient initiation of a second site of polarized growth at NETO (Glynn et al., 2001). It is not known whether this function depends on for3p or represents an independent activity of bud6p. Remarkably, the for3DAD* allele was sufficient to restore bipolar growth in bud6Δ cells: whereas only ∼40% of bud6Δ cells initiate a second site of growth, >70% of bud6Δ for3DAD* cells undergo NETO and grow in a bipolar manner (Figure 8, C and D). Similar results were obtained using for3⁺ and for3DAD* alleles tagged with HA instead of GFP (data not shown). Thus, preventing for3p autoinhibition also bypasses the need for bud6p in NETO. In contrast, the for3DAD* mutation did not restore bipolar growth in other NETO mutants, such as tea1Δ or tea4Δ, and it did not seem to cause premature bipolar growth in interphase wild-type cells (Figure 8D; data not shown). Thus, this effect of for3DAD* was specific toward bud6Δ cells. In summary, all known phenotypes of bud6Δ cells—defective for3p localization, weak actin cables, and NETO failure—can be rescued by the for3DAD* mutation. The simplest explanation for these effects is that bud6p has a primary role in relieving autoinhibition of for3p (see Discussion).

Bud6 Also Mediates Cortical Localization

Although bud6p was not required for for3pDAD* localization in the context of the whole protein, it was needed for the proper cell tip localization of for3C, the C-terminal fragment of for3p. Because this fragment cannot be autoinhibited, these effects of bud6p could not be explained solely by its effects on autoinhibition. Introduction of the DAD* mutation in for3C did not restore its localization to cell tips in bud6Δ cells, showing that the DAD* mutation itself did not somehow produce a novel cortical targeting signal (data not shown).

In bud6 mutant cells, GFP-for3C failed to localize to cell tips, but it still decorated actin cables (Figure 7D). When we prevented actin cable association, either by LatA treatment or by using the FH2 I930A mutant, GFP-for3C in bud6Δ cells did not accumulate at cell tips, and it was instead completely diffuse (Figure 7D; data not shown). In contrast, GFP-for3N localized normally to both cell tips in bud6Δ cells. These results suggest that bud6p has two functions: one in regulating autoinhibition and the second as a physical tether that contributes toward its localization at cell tips. We note that this tethering function of bud6p is usually masked in the context of the full-length for3p protein because of a functionally redundant localization site in the N terminus.

Relationship of bud6p and cdc42p

Because bud6p and cdc42p seem to share functions in regulating for3p, we investigated the possible relationship between these proteins. Several lines of evidence suggest that these factors function in independent ways. First, cdc42p and bud6p bind different for3p domains. Second, cdc42p localized well to growing cell tips in bud6Δ cells (Figure 7E). Also, bud6p–GFP localization to cell tips was unaffected in cdc42-1625 cells (Figure 7F). Third, cdc42-1625 and bud6Δ mutants were synthetic lethal (at 25°C). Thus, cdc42p and bud6p may not act solely in a common pathway but rather represent parallel pathways that both regulate for3p, as well as possibly other targets in cell polarization.

DISCUSSION

Here, we have dissected mechanisms of regulation and localization of the formin for3p. We show that for3p possesses at least two domains that mediate cell tip localization—in the N and C termini. A third domain, the FH2 domain, mediates its association with actin cables. Similar to diaphanous-related formins, the ability of for3p to localize and drive actin assembly depends on relieving the autoinhibitory binding of DAD and DID-like sequences at its N and C termini. Our results suggest that for3p in a closed, autoinhibited state cannot bind to cortical tethers and thus cannot localize. We provide evidence that for3p is converted to an open active conformation by cdc42p and bud6p. The active for3p can now bind to the cell tip and assemble actin filaments for construction of the actin cable.

Multiple Localization Domains Target for3p to Cell Tips

The anchoring of for3p at the cortex depends on at least two cortical tethers. Our data suggest that two sites on for3p mediate its localization to the cell tips, one site at the N terminus (in the vicinity of the DID domain) and one site near the C terminus. Bud6p, which binds in the vicinity of the DAD domain, is a likely cortical anchor for the C terminus. The corresponding cortical anchor(s) for the N terminus are not yet known. Tea4p is necessary for tethering for3N (as well as full-length for3p) at nongrowing cell ends, but because for3p still localizes to one cell tip in tea4Δ cells, additional factor(s) must contribute to tethering the formin to growing cell ends (Martin et al., 2005). It is not clear yet whether cdc42p could be one such factor. Although for3pDAD* and a for3N fragment can still bind to cell tips in a cdc42 mutant, this cdc42 mutant does not represent a complete null.

For3p Autoinhibition

Like diaphanous-related formins, for3p is regulated by an intramolecular interaction between functional DAD and DID domains. The poor conservation of these domains with other formins raises the possibility that this mode of regulation may be more widely used even in “non-diaphanous-related” formins.

Our study of the for3DAD* mutant revealed that autoinhibition has a surprisingly minor effect on the normal dynamic regulation of the formin. When the for3pDAD* mutant protein is expressed at endogenous levels, it has only mild effects on the organization of the actin cytoskeleton or cell growth, at least under the standard laboratory conditions tested. The amount of actin staining in each actin cable is increased, suggesting that there are more actin filaments within each bundle, or that individual filaments in the cable are longer. However, the rate of actin assembly, measured by the movement of for3p dots, was not increased, and the general organization of the actin cytoskeleton was normal. The DAD* did not obviously perturb the dynamic localization patterns of for3p, suggesting that this autoinhibitory regulation is not responsible for events such as its release from the cortex.

This mild phenotype is in contrast to studies in other cell types in which more dramatic effects are seen upon overexpression of truncated formin fragments (Watanabe et al., 1999; Tominaga et al., 2000; Evangelista et al., 2002; Sagot et al., 2002a; Dong et al., 2003; Koka et al., 2003; Schonichen et al., 2006). With these other studies, the more extreme effects may have been caused in part by formin overexpression and in part by the fact that formin fragments were likely delocalized from their intracellular location. It is possible that in fission yeast, most for3p molecules in the cell may normally already exist in the open conformation. In addition, for3p activity may be controlled by additional signals, such as through interactions with other proteins at the cell tips, or posttranslational modifications.

Cdc42p and Formin Regulation

Different Rho-formin pairs may regulate different actin structures in various cell types (Faix and Grosse, 2006). Our studies identify cdc42p as the yeast Rho family member needed for efficient actin cable assembly and for3p regulation in fission yeast. The five other Rho-GTPases have other functions in cell morphogenesis, and some are implicated in regulating cell wall synthesis; however, we note that our studies do not rule out minor or redundant contributions of these other Rhos on for3p regulation. Interestingly, none of them to date have been implicated in contractile ring assembly and cdc12p formin regulation in cytokinesis.

The suppression of cdc42 by a for3DAD* mutation provides strong evidence that cdc42p normally activates for3p by relieving its autoinhibition; in a cdc42 mutant, for3p may be largely autoinhibited. These studies together with biochemical and structural data on many formins lead to a model in which cdc42p binding at the DID at the N terminus of for3p leads to relief of autoinhibition (Dong et al., 2003; Li and Higgs, 2003; Lammers et al., 2005; Otomo et al., 2005a; Rose et al., 2005; Nezami et al., 2006). However, because the for3DAD* mutation does not suppress all of cdc42 defects, it is likely that cdc42p has other effectors involved in cell polarization.

Bud6p and Multiple Aspects of Formin Regulation

Bud6p may regulate formin activity in multiple ways: through regulation of autoinhibition, as a cortical tether, and possibly as a cofactor in actin assembly. Our finding that the for3DAD* mutant restores all the phenotypes ascribed to bud6Δ cells provides a strong genetic argument that bud6p functions primarily in vivo to antagonize autoinhibition. The binding of bud6p at a site at or near the DAD suggests an attractive model in which bud6p could block the DAD from binding to the DID. This is by far the simplest model to explain our data, and alternative models are considerably more complicated. For example, bud6p could activate for3p through some other mechanism, but the increase of activity of the for3pDAD* mutant may compensate for the decrease in activity in the bud6Δ mutant. However, this model would not account for how the DAD* mutation restores for3p localization in bud6Δ cells. Future studies will be needed to test more directly the function of bud6p on autoinhibition.

Bud6p also seems to have a function as one of the cortical tethers that localize for3p at the cell tip. This function is redundant with tethers that bind around the DID domain. A third potential function is in regulation of actin assembly by the formin FH2 domain. In S. cerevisiae, in vitro experiments have shown that Bud6p binds to G-actin, accelerates nucleotide exchange on actin, and enhances the actin assembly activity of the Bni1p FH2 domain (Moseley et al., 2004). Bud6p thus may act like profilin to increase the local concentration of G-actin at the actin assembly site at the formin core. Because for3p activity has not been reconstituted in vitro yet, we have not been able to test the role of bud6p on this activity in the fission yeast proteins directly. However, the robust actin cables seen in a bud6Δ for3DAD* mutant suggest that bud6p is not essential for the activity of a fully uninhibited formin in vivo.

The presence of multiple activating factors poses the question whether each modulates for3p independently, for example, in different cellular contexts, or whether they act in a concerted manner for full activation. For example, bud6p may prime the formin in a conformation to facilitate binding of cdc42p, or it may help to maintain for3p in the open conformation. Although obvious homologues of bud6p are not found outside of fungi, it will be interesting to investigate whether other formins are regulated at the C-terminal regions by analogous regulators. There is indeed some evidence from in vitro experiments that RhoA is not sufficient for full relief of mDia1 autoinhibition (Li and Higgs, 2003).

For3p and Regulation of Bipolar Growth

This work provides further evidence that the activation of for3p is a key step in establishment of bipolar growth at NETO. Two proteins required for NETO, tea4p and bud6p, have been identified to interact with and regulate for3p (Glynn et al., 2001; Feierbach et al., 2004; Martin et al., 2005). The ability of the for3DAD* allele to restore NETO in bud6 mutants suggests that the role of bud6p in NETO is to regulate for3p.

A current model for NETO is that, during polarity establishment in the G2 phase, for3p is recruited and activated at the new cell end by the concerted actions of at least three proteins: bud6p, cdc42p, and tea4p. Because expression of the for3DAD* allele is not sufficient by itself for promoting bipolar growth in all cells, relief of autoinhibition is not the rate-limiting step in formin activation at NETO. However, relief of for3p autoinhibition by cdc42p and/or bud6p may be a necessary step to allow for3p to bind to cortical docking proteins present at the new end, such as tea4p. Bud6p, which is also directly or indirectly recruited to the new end by tea1p and tea4p (Feierbach et al., 2004; Martin, unpublished observations), may also contribute to the tethering of for3p. At the new cell end, activation of actin assembly by the FH1–FH2 domains may be mediated by relief of autoinhibition and by other modes of activation, such as via tea4p (Martin et al., 2005). Localized formin-mediated assembly of actin cables then leads to initiation of polarized cell growth.

Other organisms may use similar autoinhibitory regulation of formins to modulate patterns of cell polarization: DAD-dependent regulation of a formin in Ashbya gossypii also affect the hyphal growth pattern of this filamentous fungus (Schmitz et al., 2006). Further investigation into these and other modes of formin regulation will reveal whether multiple activators also control formin autoinhibition in other species and provide a better understanding of the precise regulatory mechanisms underlying cell polarization and regulation of actin assembly.

Abbreviations used:

DAD

Diaphanous Autoregulatory Domain

DID

Diaphanous Inhibitory Domain

FH

Formin Homology

LatA

Latrunculin A

NETO

New End Take Off.