Formins at a glance

Formins are conserved actin polymerization machines that have instrumental roles in controlling rearrangements of the actin cytoskeleton and have recently been shown to directly regulate microtubule dynamics. Here, and on the accompanying poster, we aim to organize a rapidly expanding body of literature on this diverse protein family, summarizing the common properties that apply to most formins, and highlighting recent advances in understanding formin structure, mechanism, activity and regulation at the molecular and cellular levels.

Formins were first identified in flies, mice and yeast as genes that, when mutated, cause severe defects in cytokinesis, polarity, and cell and tissue morphogenesis (Mass et al., 1990; Jackson-Grusby et al., 1992; Castrillon and Wasserman, 1994; Kohno et al., 1996). Subsequent studies in budding yeast Saccharomyces cerevisiae revealed that formins directly nucleate the assembly of actin filaments, and that these activities are essential in vivo for the assembly of actin cables that direct polarized cell growth (Evangelista et al., 2002; Pruyne et al., 2002; Sagot et al., 2002a; Sagot et al., 2002b). Since then, members of this large protein family (encoded by 15 genes in mammals, two in S. cerevisiae, three in Saccharomyces pombe, and six in Drosophila melangastor) have been demonstrated to have crucial roles in an increasingly wide range of cytoskeleton-based processes (as illustrated in the Poster).

Cellular functions of formins

In yeast, formins have an essential role in the assembly of cytokinetic rings and cables that direct intracellular transport (Chang, 1999; Feierbach and Chang, 2001; Evangelista et al., 2002; Dong et al., 2003; Cheung and Wu, 2004; Ingouff et al., 2005; Cheung et al., 2010). In animal cells, formins are required for the assembly of filopodia (Pellegrin and Mellor, 2005; Schirenbeck et al., 2005; Yang et al., 2007; Block et al., 2008; Matusek et al., 2008; Harris et al., 2010), lamellipodia (Yang et al., 2007; Sarmiento et al., 2008; Block et al., 2012), stress fibers (Ishizaki et al., 2001; Satoh and Tominaga, 2001; Gasteier et al., 2003; Peng et al., 2003; Hotulainen and Lappalainen, 2006; Sato et al., 2006; Takeya et al., 2008), cytoplasmic actin networks used for long-range vesicle transport (Leader et al., 2002; Azoury et al., 2008; Li et al., 2008; Pfender et al., 2011; Schuh, 2011), cytokinetic actin rings (Severson et al., 2002; Watanabe et al., 2008) and phagocytic cups (Brandt et al., 2007). Formins also have essential roles in physiological processes ranging from cell motility in the immune system (Yayoshi-Yamamoto et al., 2000; Eisenmann et al., 2007; Shi et al., 2009), to gastrulation and neural tube closure (Habas et al., 2001; Sato et al., 2006; Lai et al., 2008), heart morphogenesis (Iskratsch et al., 2010; Li et al., 2011), kidney morphogenesis (Brown et al., 2010; Boyer et al., 2011a; Boyer et al., 2011b), and dendritic spine formation in neurons. Some of these functions depend on the actin nucleation and elongation activities of formins (Bartolini et al., 2008; Andrés-Delgado et al., 2010; Madrid et al., 2010; Andrés-Delgado et al., 2012; Ramabhadran et al., 2012; Stastna et al., 2012), which have been demonstrated by investigating point mutations that impair in these activities (Xu et al., 2004; Lu et al., 2007; Ramabhadran et al., 2012). However, for many of the other in vivo functions of formins, it has not yet been determined which of their activities are required. Moreover, some formins exhibit activities beyond actin assembly, e.g. actin bundling, severing, depolymerization and microtubule (MT) binding (see below). Thus, an important future challenge is to develop new mutants as tools that are capable of separately disrupting each formin activity.

Localization and activation of formins

Formins are recruited and activated at different sites in cells, where they perform their diverse roles in cytoskeletal reorganization. Of the 15 vertebrate formins, the largest subset are Diaphanous-related formins (DRFs), which have an N-terminal GTPase-binding domain (GBD), an adjacent DID (Diaphanous inhibitory domain) and a C-terminal DAD (Diaphanous autoregulatory domain) (see Poster) (Alberts, 2001; Li and Higgs, 2003; Otomo et al., 2005a; Rose et al., 2005; Otomo et al., 2010). DRFs are autoinhibited through DID–DAD interactions. Recent crystallographic and single-particle electron microscopy studies show that, in the autoinhibited conformation, the N-terminus physically obstructs the ability of the C-terminus to polymerize actin (Nezami et al., 2010; Otomo et al., 2010; Maiti et al., 2012). Binding of active Rho-GTPases to GBD activates the formin by releasing these DID–DAD interactions. However, activation is noticeably incomplete (Li and Higgs, 2003; Maiti et al., 2012), suggesting that other factors are required to fully activate formins.

A variety of Rho-GTPases recruit DRFs to different locations in the cell for localized actin assembly (Evangelista et al., 1997; Watanabe et al., 1997; Ishizaki et al., 2001; Nakano et al., 2002; Tolliday et al., 2002; Pellegrin and Mellor, 2005; Seth et al., 2006; Martin et al., 2007; Block et al., 2012) (see Poster), yet additional factors can also regulate formins. For example, the S. cerevisiae formin Bnr1 and S. pombe formin Cdc12 each harbor at least two separate localization sequences that independently target the formin in vivo (Gao et al., 2010), suggesting that a combination of cues and binding partners control formin recruitment. Indeed, one of the sequences in Bnr1 was shown to interact with a septin-associated kinase that controls Bnr1 function (Buttery et al., 2012). In addition, the S. cerevisiae DRF Bni1 is phosphorylated at its N- and C-termini by Prk1 kinase, which facilitates its release from autoinhibition (Wang et al., 2009). In mammals, the formin protein diaphanous homolog 2 (DIAPH2, hereafter referred to as mDia3) is phosphorylated by Aurora B kinase (Cheng et al., 2011), the formin homology domain proteins 1 and 3 (FHOD1 and FHOD3, respectively) are phosphorylated by cGMP-dependent protein kinase 1 (PRKG1) and casein kinase 2 subunit α (CSNK2A1, also known as CK2) (Hannemann et al., 2008; Iskratsch et al., 2010; Iskratsch et al., 2012), and mDia2 and FHOD1 are activated through phosphorylation by Rho-associated protein kinase (ROCK) (Takeya et al., 2008; Staus et al., 2011). Inverted formin-2 (INF2) and formin-like protein 2 (FMNL2, also known as FRL3) are also farnesylated and myristoylated, respectively, which promotes their membrane targeting (Chhabra et al., 2009; Block et al., 2012). Moreover, mDia1 and mDia2, as well as the plant formins AFH1, formin1 and class II formin, directly bind phospholipid membranes (Cheung et al., 2010; Ramalingam et al., 2010; Gorelik et al., 2011; Martinière et al., 2011; van Gisbergen et al., 2012). These observations show that localization and activation of formins depend on their diverse interactions and that localization and activation, in some cases, serve as convergent inputs from multiple signalling pathways.

Biochemical activities of formins

Formins are large, dimeric multi-domain proteins with a modular design (see poster). Their signature features are the C-terminal formin homology 1 and 2 domains (FH1 and FH2, respectively) (Chesarone et al., 2010; Schönichen and Geyer, 2010). The only known exception is Dictyostelium ForC, which lacks an FH1 domain (Rivero et al., 2005). Most formins also have C-terminal tail regions that, sometimes, include DAD domains and/or Wiskott-Aldrich syndrome homology region 2 (WH2)-like domains. In comparison, the N-terminal halves of formins are more variable and have main roles in directing their localization.

Most purified formins exhibit the following three activities: (1) nucleation of actin assembly, (2) processive movement on growing barbed ends of actin filaments while protecting them from capping proteins and (3) profilin-dependent acceleration of actin filament elongation. Additional activities in different subsets of formins include MT binding (see below) and actin filament bundling, severing and/or depolymerization (Harris et al., 2004; Michelot et al., 2005; Moseley and Goode, 2005; Chhabra and Higgs, 2006; Harris and Higgs, 2006; Esue et al., 2008; Barkó et al., 2010; Harris et al., 2010; Machaidze et al., 2010; Scott et al., 2011; Skillman et al., 2012). All known interactions of formins with actin and MTs are mediated by the FH1 and FH2 domains and/or the tail regions. The FH1 domain is predicted to be extended and unstructured, and contains multiple proline-rich motifs that recruit complexes that consist of actin monomers and profilin, a small abundant protein that binds the majority of ATP–actin monomers in cells (Chang et al., 1997; Imamura et al., 1997; Sagot et al., 2002b; Kovar et al., 2003; Kovar et al., 2005). The FH2 domain forms a ring-shaped anti-parallel dimer, in which the two halves are held together by interactions of ‘lasso’ and ‘post’ segments, generating a flexibly tethered dimer (Xu et al., 2004). FH2 dimers bind with high affinity to the barbed ends of actin filaments (Pruyne et al., 2002; Kovar et al., 2003; Zigmond, 2004), and each functional half of the dimer contains two different actin-binding sites; one marked by a conserved isoleucine residue (I1431 in Bni1) and the other by a conserved lysine residue (K1601 in Bni1) (Xu et al., 2004; Otomo et al., 2005b; Lu et al., 2007). Mutation of these residues (e.g. in Bni1, mDia1, mDia2, FRL2, Daam1, INF2) strongly reduces actin nucleation and elongation activities of the respective formins and compromises actin assembly in vivo (Shimada et al., 2004; Xu et al., 2004; Lu et al., 2007; Bartolini et al., 2008; Harris et al., 2010; Ramabhadran et al., 2012).

Mechanism of actin nucleation

The precise mechanism by which formins nucleate the assembly of actin filaments is still being worked out. Initially, it was proposed that formins nucleate actin filaments by capturing and stabilizing spontaneously formed actin dimers and trimers (Pring et al., 2003) (see Poster), on the basis that the FH2 domain alone is sufficient for nucleation in vitro but lacks detectable binding affinity for actin monomers. Subsequently, however, it was shown that FH2-domain-mediated nucleation is very inefficient when using profilin-bound actin monomers, the predominant substrate that is available for actin polymerization in cells (Chesarone et al., 2010). More recent studies have shown that the C-terminal tail regions of formins bind actin monomers and enhance nucleation in the presence of profilin, which might explain how formins nucleate actin assembly in vivo (Gould et al., 2011; Heimsath and Higgs, 2012). In addition, the tail regions of formins can interact with other factors that bind actin monomers with high affinity and promote nucleation (see below). Thus, the interplay between formins and nucleation co-factors is a powerful means by which both, tight control and robust stimulation of actin nucleation, can be achieved in vivo (Blanchoin and Michelot, 2012). Moreover, interactions between FH1 and the profilin-actin complex might contribute to nucleation (Paul and Pollard, 2008). Thus, nucleation triggered through formins in the presence of profilin may involve contributions by their FH1 and FH2 domains, and the tail regions.

Although most formins promote actin nucleation and elongation, the strength of these activities can vary drastically. For example, S. pombe Cdc12 is a potent nucleator with a nucleation efficiency of over 50% (every second formin dimer nucleates a filament) (Neidt et al., 2008), whereas other formins, such as Daam1, FMNL3 (also known as FRL2) and FMNL2</emph> have nucleation efficiencies of below 1% (Vaillant et al., 2008; Block et al., 2012). These differences may reflect diverse in vivo requirements, e.g. the need for a slower and more controlled nucleation in some instances, or the dedication of a formin to actin filament elongation rather than nucleation. Indeed, one study has proposed that the primary role of FMNL2 at the leading edge of the cell is to capture free actin filament barbed ends that are nucleated by the Arp2/3 complex and elongate those filaments to drive filopodial and lamellipodial extension (Block et al., 2012). The weaker nucleation by some formins that has been observed in vitro might also reflect their stronger dependence on co-factors or nucleation promoting factors (NPFs) in vivo.

Formin-interacting NPFs are believed to help formins overcome the barrier that profilin constitutes to actin nucleation. Profilin suppresses the self-association of actin into dimers and trimers, and reduces the efficiency of formin nucleation (Neidt et al., 2008; Paul and Pollard, 2008; Scott et al., 2011). However, NPFs effectively compete with profilin for actin monomer binding, and organize mutiple actin monomers in proximity to the formin FH2 domain. Such NPF-formin pairs include Bud6–Bni1, Spire–FMN (Spir and Cappuccino in Drosophila) and adenomateous polyposis coli protein (APC)–mDia1 (Moseley et al., 2004; Quinlan et al., 2007; Webb et al., 2009; Okada et al., 2010; Graziano et al., 2011; Tu et al., 2012). In vivo, Bni1–Bud6 and APC–mDia1 function together to assemble actin cables and pseudocleavage furrows, respectively, and have also been shown to directly interact in vitro to assemble actin in the presence of profilin and/or capping protein (Moseley et al., 2004; Okada et al., 2010; Graziano et al., 2011; Breitsprecher et al., 2012). Spire and FMN co-function in vivo to assemble cytoplasmic actin meshworks (Schumacher et al., 2004; Pfender et al., 2011; Schuh, 2011) but, perplexingly, Spire inhibits rather than enhances the nucleation activity of FMN in vitro (Quinlan et al., 2007; Vizcarra et al., 2011; Zeth et al., 2011), suggesting that additional factors are required to activate collaborative actin assembly through Spire–FMN. More recently, we have begun to address the question of how NPF–formin pairs co-assemble actin filaments by using triple-color total internal reflection fluorescence (TIRF) microscopy at the single-molecule level (Breitsprecher et al., 2012). This study revealed that, during the early phases of nucleation, mDia1 and APC molecules associate, with APC being mainly responsible for recruiting actin monomers. Subsequently, upon actin polymerization, APC and mDia1 separate, with APC remaining at the nucleation site and mDia1 moving processively along the growing barbed end, where it protects the filament from capping protein (Breitsprecher et al., 2012). An important next goal will be to determine whether other NPF–formin pairs use similar or distinct mechanisms.

Mechanisms of actin filament elongation

Once an actin filament is nucleated, the dimeric FH2 domain processively tracks the growing barbed end, which permits the addition of tens of thousands of actin subunits before it finally dissociates. The tracking mechanism is thought to involve transient, alternating contacts of the two halves of the FH2 domain dimer with the two terminal actin subunits of the filament. During these movements, the FH2 dimer is believed to switch between an ‘open state’ that allows actin monomer addition and a capped or ‘closed state’ that does not (see Poster) (Zigmond et al., 2003; Kozlov and Bershadsky, 2004; Moseley et al., 2004; Romero et al., 2004; Xu et al., 2004; Otomo et al., 2005b; Kovar et al., 2006; Vavylonis et al., 2006; Paul and Pollard, 2009a). The fraction of time a formin spends in the open versus the closed state is its ‘gating factor’ (Paul and Pollard, 2009a) and ranges from almost 0 (capped) for S. pombe Cdc12 to nearly 1 (uninhibited elongation) for mDia1 and Caenorhabditis elegans CYK-1 (Kovar et al., 2006; Neidt et al., 2008). Owing to differences in gating factors, different formins slow filament elongation and depolymerization to variable degrees in the absence of profilin (Kovar et al., 2003; Harris et al., 2004; Kovar et al., 2006; Neidt et al., 2008; Neidt et al., 2009; Block et al., 2012).

In the presence of profilin, however, the FH1 domains recruit profilin–actin complexes and accelerate the addition of actin monomers to the FH2-capped barbed end of up to tenfold over the rate of elongation at free barbed ends (Kovar et al., 2003; Romero et al., 2004; Kovar et al., 2006; Paul and Pollard, 2008; Paul and Pollard, 2009b; Vidali et al., 2009) (see Poster). How actin subunits are delivered from the FH1 to the FH2 domains is not fully understood, but might involve direct interactions between FH2– and FH1–profilin–actin (Ezezika et al., 2009; Neidt et al., 2009). Furthermore, the profilin-binding sites in the FH1 domain are arranged in a specific order, with low-affinity sites nearest to the FH2 domain and high-affinity sites more distal from it (Courtemanche and Pollard, 2012), possibly to optimize actin transfer to the barbed end. Importantly, it has been shown that formin processivity depends neither on profilin nor on the hydrolysis of ATP on newly added actin subunits (Kovar et al., 2006; Michelot et al., 2007; Paul and Pollard, 2009b; Ramalingam et al., 2010; Mizuno et al., 2011; Breitsprecher et al., 2012). Thus, the release of free energy, which accompanies the addition of actin subunits at the barbed end of the filament seems to be sufficient to drive formin FH2 processive movement.

Owing to the helical pitch of the actin filament, each step taken by an untethered formin on the barbed end is accompanied by a rotation of 14 degrees. At in vivo rates of actin polymerization, this would cause a formin to ‘spin’ at over 1000 rpm, which may be difficult to achieve for membrane-bound, immobilized formins. This problem has been referred to as the ‘rotation paradox’, and was initially addressed using single-filament TIRF microscopy and passively immobilized formins, which led to the conclusion that formins do not faithfully rotate during filament elongation, suggesting that they, instead, slip around the barbed end to release torsional stress (Shemesh et al., 2005). However, this model has recently been challenged in a different study, which employed single-molecule fluorescence polarization and in which helical rotations of filaments were observed that are elongated by tightly immobilized formins (Mizuno et al., 2011). Thus, further experimental studies are needed to determine whether formins are capable of either rotating or slipping depending on the forces or constraints that are applied.

Regulation of formin processivity

In vitro, formins remain processively attached to growing ends of filaments for minutes without dissociating, and produce actin polymers that are over 50 µm long (Kovar et al., 2006; Neidt et al., 2008; Breitsprecher et al., 2012). These properties suggest that, in vivo, there is a need for regulatory mechanisms to limit the duration and/or rate of formin processivity because actin filaments in cells are typically no longer than 1 µm (Pollard and Borisy, 2003). To date, two such mechanisms have been described for S. cerevisiae Bnr1: (1) Bud14 binds to its FH2 domain, which catalyzes Bnr1 displacement from filament ends (Chesarone et al., 2009); (2) the myosin-passenger protein Smy1, which also binds to the Bnr1 FH2 domain, interferes with actin filament elongation, slowing it down to half the normal speed (Chesarone-Cataldo et al., 2011) (see Poster). In vivo, both Bud14 and Smy1 are required to prevent the overgrowth of actin cables; loss of both proteins at the same time results in even more-severe defects of cable organization and cell growth. In addition, capping protein itself might competitively displace formins from filament ends, as suggested by its genetic interactions with Bud14 in S. cerevisiae (Chesarone et al., 2009) and its antagonistic interactions with mDia1 in mammalian cells (Bartolini et al., 2012). Another potential group of formin regulators are FH1-domain ligands, for example the Slit–Robo GTPase-activating proteins (srGAPs) (Mason et al., 2011), which might interfere with the binding of profiling–actin complexes to slow or arrest elongation (see Poster). Moreover, the Arp2/3-activating WAVE complex has been shown to directly inhibit mDia2, possibly as part of a regulatory circuit that balances the activity of the WAVE–Arp2/3 complex in lamellipodia formation with that of mDia2 in extending filopodia (Beli et al., 2008).

Regulation of microtubule dynamics

In vivo observations have long suggested that formins also regulate MT organization and dynamics (Kikyo et al., 1999; Lee et al., 1999; Chang, 2000; Ishizaki et al., 2001; Kato et al., 2001; Leader et al., 2002; Deeks et al., 2010; Li et al., 2010). Expression of constitutively active mDia constructs results in co-alignment of MTs and actin fibers (Ishizaki et al., 2001), promotes MT targeting to the cell periphery (Yamana et al., 2006; Goulimari et al., 2008; Zaoui et al., 2010) and stabilizes MTs (Palazzo et al., 2001; Wen et al., 2004; Eng et al., 2006; Bartolini et al., 2008). mDia mutants that fail to polymerize actin still bind to and stabilize MTs in vitro and in vivo (Bartolini et al., 2008). Furthermore, mDia3 mediates MT attachment to kinetochores independent of actin polymerization (Yasuda et al., 2004; Cheng et al., 2011), and INF2 promotes MT stabilization to reorient centrosomes (Andrés-Delgado et al., 2012). Furthermore, inverted formin 1 (INF1) colocalizes with MTs and promotes their bundling and acetylation in vivo, consistent with a role in stabilizing MT arrays (Young et al., 2008). More recently, it was shown that MT acetylation in cells is induced upon overexpression of each of ten different mammalian formins (Thurston et al., 2012). Thus, formins appear to stabilize MTs both through their direct binding (Bartolini et al., 2008; Gaillard et al., 2011) and/or by altering the post-translational state of MTs (Wen et al., 2004; Bartolini et al., 2008; Thurston et al., 2012).

Biochemical studies have begun to define the molecular interactions of the formins mDia1, mDia2 and INF2 with MTs (Bartolini et al., 2008; Gaillard et al., 2011). These studies have revealed surprising differences in how formin domains interact with MTs – as well as differences in their binding affinities and stoichiometries – suggesting that these formins have distinct roles in regulating MTs. In addition, these studies have shown that the presence of MTs inhibits the actin assembly activities of mDia1 and mDia2 but not of INF2 (Gaillard et al., 2011), suggesting that some formins can coordinate their effects on MTs and actin filaments. This view is supported by the interaction of formins with the MT plus-end tracking proteins (+TIPs) APC and EB1, and the cytoplasmic linker protein 170 (CLIP-170; tea1p in yeast) (Feierbach et al., 2004; Wen et al., 2004; Bartolini et al., 2008; Lewkowicz et al., 2008; Cheng et al., 2011), thereby representing an area ripe for future investigation at both biochemical and cellular levels.

Conclusions and perspectives

Ten years of genetic and biochemical characterization has demonstrated that formins in different species directly catalyze actin filament nucleation and elongation, but that the strength of their activities vary greatly, as do their in vivo binding partners, localization and regulation. Furthermore, many formins have additional activities (e.g. actin filament depolymerization, severing and bundling, and MT stabilization and bundling) that are still not well understood and deserve more attention (see Poster). For these reasons, it is crucial to characterize the properties of each formin individually before drawing conclusions about its cellular roles. Moreover, it is important to consider formin activities in the context of their ligands, which can transform their activities in a variety of ways.