FHOD1 regulates stress fiber organization by controlling the dynamics of transverse arcs and dorsal fibers

ABSTRACT The formin FHOD1 (formin homology 2 domain containing protein 1) can act as a capping and bundling protein in vitro. In cells, active FHOD1 stimulates the formation of ventral stress fibers. However, the cellular mechanisms by which this phenotype is produced and the physiological relevance of FHOD1 function are not currently understood. Here, we first show that FHOD1 controls the formation of two distinct stress fiber precursors differentially. On the one hand, it inhibits dorsal fiber growth, which requires the polymerization of parallel bundles of long actin filaments. On the other hand, it stimulates transverse arcs that are formed by the fusion of short antiparallel actin filaments. This combined action is crucial for the maturation of stress fibers and their spatio-temporal organization, and a lack of FHOD1 function perturbs dynamic cell behavior during cell migration. Furthermore, we show that the GTPase-binding and formin homology 3 domains (GBD and FH3) are responsible for stress fiber association and colocalization with myosin. Surprisingly, a version of FHOD1 that lacks these domains nevertheless retains its full capacity to stimulate arc and ventral stress fiber formation. Based on our findings, we propose a mechanism in which FHOD1 promotes the formation of short actin filaments and transiently associates with transverse arcs, thus providing tight temporal and spatial control of the formation and turnover of transverse arcs into mature ventral stress fibers during dynamic cell behavior.


INTRODUCTION
Diaphanous-related formins (DRFs) form an evolutionarily conserved family of actin regulators, which are involved in numerous cellular processes, such as morphogenesis, cell division and cell polarity. In general, formins are well known for their ability to act as processive actin nucleators or 'leaky cappers' for actin filaments -they catalyze the nucleation of new actin filaments and stay attached to the barbed end of the growing filament (Kovar and Pollard, 2004;Higashida et al., 2004). However, the efficiency of actin nucleation and of subsequent processive monomer addition varies greatly among individual family members (Paul and Pollard, 2009). In addition to their nucleation activity, several formins have been described to have other modulatory roles, such as actin filament bundling or actin severing and de-polymerization (Harris et al., 2006;Harris et al., 2010), suggesting that they might also play additional roles in cells (Harris et al., 2004;Chhabra and Higgs, 2006;Chesarone et al., 2010).
For multiple family members, the highly conserved formin homology 2 (FH2) domain was shown to be sufficient to catalyze actin nucleation in vitro. The FH1 domain is thought to promote processive filament elongation by recruiting profilin-bound monomeric actin (Sagot et al., 2002;Paul and Pollard, 2008;Romero et al., 2004). Two additional modules shared among DRFs -the C-terminal diaphanous autoregulatory domain (DAD) and the N-terminal diaphanous inhibitory domain (DID) regioninteract with one other, keeping the protein in the inactive conformation (Alberts, 2001;Faix and Grosse, 2006;Schönichen and Geyer, 2010). In mDia1 (DIAPH1), this auto-inhibitory intramolecular interaction was shown to be released by the binding of the Rho GTPase RhoA to the N-terminal GTPasebinding domain (GBD), thus stimulating the actin-nucleating activity of this formin and promoting actin polymerization into stress fibers from focal adhesions Hotulainen and Lappalainen, 2006). ROCK, another prominent effector of RhoA, phosphorylates myosin II and increases the motor activity of myosin to promote the bundling and contractility of existing actin filaments . Thus, RhoA controls stress fiber formation through the concerted activation of at least two downstream effectors, Rho kinase (ROCK) and mDia1 (Watanabe et al., 1999).
The formin FHOD1 (formin homology 2 domain containing protein 1) is also thought to play a role in stress fiber formation (Koka et al., 2003). FHOD1 shares a similar domain organization with that of other DRFs and is also activated by a mechanism involving the release of auto-inhibition (Schönichen et al., 2006). The truncation of the C-terminal amino acids 1012-1164, which include the DAD region and a short sequence of unknown function (called 'X'), leads to constitutive activation of this formin and results in the formation of prominent stress fibers throughout the cell (Koka et al., 2003;Gasteier et al., 2003). The RhoA effector ROCK phosphorylates FHOD1 at three residues within this C-terminal region (S1131, S1137, T1141), disrupting the inhibitory interaction between the N-terminal FH3 domain (DID in mDia1) and the DAD region (Takeya et al., 2008). Similar to the C-terminal truncation mutant, the overexpression of the corresponding triple phosphomimetic mutant induces the formation of thick linear stress fibers, suggesting that ROCKmediated phosphorylation is involved in the activation of FHOD1 (Takeya et al., 2008). In addition, it is suggested that the Rho GTPase Rac1 releases the auto-inhibition of FHOD1 by binding to the GBD region in a manner similar to the interaction between RhoA and mDia1 (Westendorf, 2001). However, there is little experimental evidence to support this claim and it is currently unclear how Rac1 activity affects FHOD1 in cells. Active FHOD1 has been shown to localize to stress fibers, and its overexpression in cells induces the formation of thick rigid actin fibers (Koka et al., 2003;Gasteier et al., 2003). The truncation of a region encompassing the GBD, FH3 domain and a part of the coiled-coil domain (amino acids 1-573) prevents the binding of FHOD1 to stress fibers (Takeya and Sumimoto, 2003;Schönichen et al., 2013); however, the functional role of the association of FHOD1 with stress fibers was not explored in cells.
Based on a recent study by Krainer and colleagues (Krainer et al., 2013), FHOD1 expression is among the highest of all 15 human formins tested in a large variety of cell lines and tissues, suggesting that it is of broad physiological importance in cytoskeletal organization. Despite this finding, it is currently not known how FHOD1 is involved in the regulation of the complex actin meshwork and its coordinated reorganization during cellular morphogenesis or directional cell migration. In adherent cells, several distinct types of stress fibers exist, each of which could be regulated in different ways by endogenous FHOD1. Contractile transverse arcs are curved structures formed by the anti-parallel assembly of short actin filaments (Zhang et al., 2003). After their initial formation in the lamellipodium, they undergo retrograde flow towards the cell center (Heath, 1983;Small et al., 1998;Hotulainen and Lappalainen, 2006). Myosin II motors and the filament-bundling protein a-actinin are incorporated into arcs in a periodically alternating pattern (Hotulainen and Lappalainen, 2006). During their retrograde flow, arcs are not linked to focal adhesions but associate with dorsal stress fibers, which are also generated in the leading edge of motile cells. The distal ends of dorsal stress fibers associate with focal adhesions, whereas their proximal ends are linked to the retrograde flow of the transverse arc meshwork (Small et al., 1998;Hotulainen and Lappalainen, 2006). In contrast to arcs, dorsal stress fibers are formed by the polymerization of long parallel actin filaments at focal adhesions. Lastly, ventral stress fibers are coupled with focal adhesions at both ends, and they are thought to be formed either by the fusion of transverse arcs with two flanking dorsal stress fibers or by the end-to-end fusion of two dorsal stress fibers (Small et al., 1998;Hotulainen and Lappalainen, 2006). The stress fibers that are induced by active FHOD1 mutants might be related to such ventral stress fibers. However, the precise formation mechanism and spatial organization of these thick actin fibers are still poorly understood. Based on its homology to other DRFs, FHOD1 would be expected to play a role in those processes by nucleating and stimulating actin polymerization in stress fibers. However, a recent study by Schönichen and colleagues (Schönichen et al., 2013) showed that FHOD1 does not act as an actin filament nucleator but instead can cap and bundle actin filaments in vitro. It is currently unclear what role these molecular functions of FHOD1 play in cells and whether they are involved in the enhanced formation of ventral stress fibers by the activated formin.
Here we use a combination of live-cell microscopy, expression of truncation mutants and acute pharmacological perturbation to investigate FHOD1 function in cells. Our studies reveal that this formin distinctly regulates the formation and growth of different stress fiber types. RNAi-mediated silencing of FHOD1 expression results in the collapse of the cellular stress fiber meshwork and causes defects in cell migration and cell spreading. Together, our work supports a model in which FHOD1 plays a central role in the spatial and temporal coordination of cellular stress fiber dynamics.

RESULTS
FHOD1 depletion leads to the collapse of the stress fiber meshwork in U2OS cells To study the cellular mechanism by which FHOD1 induces stress fibers in cells, we used RNAi to deplete the endogenous formin. The efficient knock-down of the protein was confirmed by western blotting (supplementary material Fig. S1A). First, transverse arcs and dorsal stress fibers were quantified in RNAi-treated cells. We distinguished between these distinct stress fiber types based on their pattern of association with focal adhesions (Hotulainen and Lappalainen, 2006): actin arcs are not linked to focal adhesions but instead are linked to dorsal stress fibers, whereas the distal ends of dorsal stress fibers are associated with focal adhesions, and their proximal ends are associated with actin arcs. After FHOD1 depletion, both wellorganized actin arcs and dorsal stress fibers were observed less frequently (Fig. 1A,B). Instead, cells preferentially formed thick and less dynamic peripheral actin bundles (Fig. 1A, yellow arrow, 1B). These findings were confirmed with four individual siRNA oligonucleotides to exclude off-target effects (supplementary material Fig. S1B). As reported previously (Small et al., 1998;Hotulainen and Lappalainen, 2006), transverse arcs were located on the dorsal cell surface (Fig. 1C). We next investigated whether FHOD1 plays a role in the spatial organization of arcs within individual cells. We restricted our morphometric measurements to those cells that were still able to form arcs, and we found that FHOD1 depletion decreased the arccovered cell area ( Fig. 1D; see supplementary material Fig. S1C for definition of area quantification). Furthermore, the average actin fluorescence intensity was reduced in this area, suggesting that the density of arcs is reduced as well (Fig. 1E). Interestingly, we observed a substantial increase in the number of cells that contained prominent stellate accumulations of actin fibers ( Fig. 1F-H). The exclusive presence of the focal adhesion marker paxillin at the distal ends of the individual fibers and the strong myosin II staining in the center suggested that these unusual structures might be generated by the collapse of the meshwork of transverse arcs and dorsal stress fibers (Fig. 1G,H). Live-cell imaging revealed that these structures are only formed transiently (Fig. 1I), leaving behind the less dynamic peripheral actin bundles described above (Fig. 1A, yellow arrow). In addition to the clear long-term effects of FHOD1 depletion on actin network organization, cytochalasin D washout experiments revealed impaired stress fiber recovery in cells lacking FHOD1, suggesting that this formin is not only crucial for long-term network organization but also for the dynamic de novo formation of stress fibers (supplementary material Fig. S2). These striking defects in actin stress fiber organization were accompanied by a significant decrease in migration efficiency (supplementary material Fig.  S3A-D).
The observed disruption in actin network organization was also paralleled by alterations in the size of focal adhesions. We quantified this effect by measuring the average number and size of focal adhesions in confocal micrographs of cells expressing paxillin fused to the fluorescent protein mKate ( Fig. 2A). We found that the average size of focal adhesions in FHOD1-depleted cells was  S3E). We found that the early phase of U2OS cell spreading on collagen type I was significantly impaired, as reflected by a smaller average adhesion area compared with control cells (Fig. 2D,E, 30 min). Interestingly, this difference was lost at a later time-point (Fig. 2D,E, 60 min), suggesting compensation by other, possibly Rac1-mediated, mechanisms (Price et al., 1998;del Pozo et al., 2000).

FHOD1 promotes transverse arc formation and turnover into ventral stress fibers
In order to study the cellular mechanism by which FHOD1 affects the cellular stress fiber organization, we performed cytochalasin D washout experiments using cells expressing FHOD1 mutants. Compared with control EGFP-transfected cells, overexpression of wild-type FHOD1 did not significantly alter stress fiber recovery dynamics (Fig. 3A,B). To study the effect of activated FHOD1, we used a point mutation (FHOD1 V228E) that induces an open protein conformation and strongly stimulates stress fiber formation in cells (Schulte et al., 2008). In contrast to Cterminally truncated constitutively active FHOD1 (FHOD1 1-1011), this mutant retains an intact C-terminus. After cytochalasin D washout, FHOD1 V228E induced an enhanced generation of dorsal transverse arcs as compared with the EGFP control and wild-type FHOD1 (Fig. 3A,B). This observation is further supported by an analysis of confocal F-actin image stacks, which showed that FHOD1 V228E induced enhanced F-actin staining on the dorsal cell surface (Fig. 3C).
In addition, live-cell imaging during cytochalasin D washout revealed that the enhanced arc formation was paralleled by an accelerated transformation of arcs into straight stress fibers (   the known long-term effect of enhanced stress fiber formation after the overexpression of constitutively active FHOD1 mutants (Koka et al., 2003;Gasteier et al., 2003;Takeya and Sumimoto, 2003;Schulte et al., 2008). Our data now show that FHOD1 stimulates the generation of transverse arcs and that constitutively active FHOD1 causes these to mature more rapidly into linear stress fiber bundles ( Fig. 3D-F).
We next examined the effect of FHOD1 activation or inactivation on dorsal stress fibers. In cells lacking this formin, we found that the average growth rate of dorsal stress fibers and the average maximal fiber length were significantly increased ( Fig. 4A,B). Furthermore, in FHOD1-depleted cells, the majority of mature dorsal stress fibers bent or buckled instead of growing straight and perpendicular to the cell edge as in control cells (Fig. 4C,D; supplementary material Movie 2). This suggests that the formation of dorsal stress fibers is stimulated in the absence of FHOD1 and that their regulation is perturbed, leading to aberrant turnover dynamics. Conversely, dorsal stress fibers were observed less frequently, and their average length was decreased after cytochalasin D washout, if activated FHOD1 was expressed ( Fig. 4E-G). In contrast to other formins, such as mDia1, which was proposed to stimulate actin polymerization by processive actin filament elongation (Higashida et al., 2004), this inhibitory effect of FHOD1 might be due to an alternative mechanism involving filament end capping.

FHOD1 localizes to contractile anti-parallel actin arrays at sites of myosin association
To gain more insight into the mechanism of FHOD1-mediated actin arc formation we next studied its intracellular localization by using confocal scanning and total internal reflection fluorescence (TIRF) microscopy at low expression levels. Wildtype FHOD1 is largely cytosolic; however, because of the increased signal-to-background ratio in TIRF microscopy, we found that wild-type FHOD1 was distributed in an irregular punctate pattern, with some preference for ventral stress fibers. These structures are known to form anti-parallel actin-filament arrays [i.e. arrays of mixed actin filament polarity (Naumanen et al., 2008)] (Fig. 5A, yellow arrow). Actin arcs, which also contain anti-parallel actin arrays, undergo rapid retrograde flow along the dorsal side of the cell and are therefore hardly detected using TIRF microscopy. The constitutively active mutant FHOD1 1-1011 was strongly associated with both ventral stress fibers and arcs and therefore was easily detected in the complete cell volume by using confocal scanning microscopy (Fig. 5B, yellow and green arrows, respectively). Interestingly, active FHOD1 did not bind to the majority of dorsal stress fiber structures, which are thought to be composed of parallel actin arrays (i.e. unidirectional actin filament arrays) (Fig. 5B, orange arrows). Thus, the preferential binding of FHOD1 to ventral stress fibers and actin arcs compared with dorsal fibers suggests that FHOD1 selectively binds to antiparallel actin arrays. However, FHOD1 did bind to the distal ends of dorsal fibers that overlapped with focal adhesions (Fig. 5B, red arrows), as did myosin (Fig. 5C, red and white arrowheads; supplementary material Fig. S4A), suggesting coordinated functionalities of the two proteins in these areas.
Arcs are formed in the lamellipodium by the bundling and fusion of short a-actinin-bound actin filaments that undergo rapid retrograde flow (Hotulainen and Lappalainen, 2006;Burnette et al., 2011;Tojkander et al., 2012). Myosin II is incorporated into these actin filament bundles, generating an alternating pattern with a-actinin. Because FHOD1 associates with stress fibers, it might play a role in coordinating those processes. Previously, it was unclear whether FHOD1 cooperates with a-actinin-mediated filament assembly or myosin-mediated contractility. Here, we found that activated FHOD1 (EGFP-FHOD1 V228E) substantially colocalized with myosin and was excluded from a-actinin-enriched regions ( Fig. 5D; supplementary material Fig. S4B). Thus, our results suggest that FHOD1 might cooperate with myosin to regulate the formation of contractile stress fibers.

Recruitment of FHOD1 to stress fibers and myosin colocalization requires an N-terminal targeting region
The stress fiber localization of FHOD1 is mediated by sequences within its N-terminal region (1-573) that consists of multiple parts, including the GBD, FH3 domain, linker and helical domain (Takeya and Sumimoto, 2003;Schönichen et al., 2013). The FH3 domain adjacent to the GBD bears the capacity to interact with the Cterminal DAD to mediate auto-inhibition (Takeya and Sumimoto, 2003;Schönichen et al., 2006;Schulte et al., 2008). The helical domain was suggested to be responsible for FHOD1 side-binding to actin filaments in vitro and might therefore mediate filament bundling (Schönichen et al., 2013). However, the precise Nterminal region that is responsible for the stress fiber localization of the protein had not been determined. To address this, we generated various FHOD1-truncation mutants and analyzed their subcellular localization (Fig. 6A,B; supplementary material Fig. S5B).
As expected, full-length activated FHOD1 V228E robustly associated with stress fibers, similar to the previously described constitutively active mutant FHOD1 1-1011 (Schulte et al., 2008;Koka et al., 2003;Gasteier et al., 2003;Takeya and Sumimoto, 2003) (Fig. 6A,B). In agreement with previous studies, the entire N-terminus (1-573) was found along stress fibers (Fig. 6B). Here, we show that GBD alone (1-115) was predominantly cytosolic, whereas GBD-FH3 (1-339) displayed a distinct association with stress fibers (Fig. 6B) and colocalization with myosin II (Fig. 6C). By contrast, a construct lacking these domains (340-1164) was again mostly cytosolic, as judged by confocal microscopy (data not shown), and did not localize along stress fibers. However, using TIRF microscopy, we found a weak but reproducible accumulation of FHOD1 340-1164 along the entire length of focal adhesions ( Fig. 6B; supplementary material Fig.  S5C). Interestingly, when stress fiber formation was stimulated acutely by cytochalasin D washout, this N-terminal truncation mutant strongly accumulated in the lamellipodium (Fig. 6D). Lamellipodia were rarely observed after long-term expression of FHOD1 340-1164; however, if they were present, we also observed enrichment of this mutant in this cell region in steadystate conditions (data not shown). Together, these observations show that the GBD-FH3 domains can mediate the stress fiber association of FHOD1 and its colocalization with myosin II independent of the two domains that were previously proposed to mediate F-actin interaction -the FH2 domain (617-1011) and the helical domain (396-573) (Pruyne et al., 2002;Schönichen and Geyer, 2010;Schönichen et al., 2013). In agreement with this, a construct containing only the linker region and the helical domain (340-573) localized mostly diffusely in the cells (supplementary material Fig. S5B).
As shown recently, truncation of the entire N-terminus (574-1164) effectively ablated the stress fiber association of FHOD1 (Fig. 6B) (Schönichen et al., 2013). Here, the majority of this truncated FHOD1 was cytosolic, but it was also found in dot-like accumulations throughout the cell body, with a preference for focal adhesions, suggesting that this construct has a tendency to form aggregates but nevertheless retains some subcellulartargeting functionality (Fig. 6B).
We next studied the functional role of the FHOD1 N-terminal stress fiber localization domain and the helical domain. Surprisingly, in long-term expression as well as cytochalasin D washout experiments, FHOD1 340-1164, which lacks the stressfiber-targeting GBD-FH3 domains (Fig. 6B), nevertheless stimulated the formation of thick linear stress fibers to a similar extent as the constitutively active FHOD1 V228E (Fig. 7A-C). FHOD1 340-1164 still encompasses the helical domain (396-573) that was suggested to bind to actin filaments in vitro through side-binding, and therefore it might mediate filament bundling (Schönichen et al., 2013). Thus, the helical domain might play a functional role in stress fiber formation. Indeed, the truncated mutant that lacks the helical domain (FHOD1 574-1164) failed to stimulate stress fiber formation (Fig. 7A,B).
Thus, FHOD1 can induce stress fibers independently of the Nterminal GBD-FH3 domains (1-339), which are responsible for stress fiber targeting. Our results therefore suggest that the helical domain (396-573) plays a key role in the enhanced stress fiber formation shown by the FHOD1 340-1164 mutant, potentially by mediating the accumulation of the protein in the lamellipodium (Fig. 6D), which might lead to the formation of short actin filament bundles through its proposed role in actin bundling (Schönichen et al., 2013).
We next characterized the spatial organization of the thick actin stress fiber bundles that are induced by activated FHOD1 V228E in more detail by using confocal F-actin image stacks. We found that the majority of those bundles were localized on the ventral side of cells, where they associated with focal adhesions on both ends. Thus, those bundles were mainly composed of ventral stress fibers (Fig. 7D, left). Interestingly, we also observed a minor fraction of thick actin bundles that formed an arch-like structure, being dorsally localized only in their central region (Fig. 7D, right, white arrow) and ventrally localized in the flanking fiber ends (Fig. 7D, right, white arrowhead), which were associated with focal adhesions. Thus, while sharing partial dorsal localization with actin arcs, these arch-like structures also showed properties of ventral stress fibers.

The C-terminal FHOD1 domain is required for the efficient formation of ventral stress fibers
The C-terminal DAD region of FHOD1 mediates its autoinhibitory conformation by interacting with the N-terminal FH3 domain (Takeya and Sumimoto, 2003;Schönichen et al., 2006).
The C-terminally truncated FHOD1 construct (FHOD1 1-1011) was therefore used in previous studies as a model for constitutively active FHOD1 with an open conformation. To test whether the FHOD1 C-terminus might play additional functional roles, we compared the stress fiber formation of FHOD1 1-1011 with that of the FHOD1 point mutant (V228E), which contains the entire C-terminus. Similar to FHOD1 1-1011, FHOD1 V228E is in the active conformation because this mutation blocks the intramolecular auto-inhibitory interaction (Schulte et al., 2008). We analyzed the effect of these FHOD1 constructs on actin organization by quantifying the number of cells that contain an increased number of mature ventral stress fibers (Fig. 8A). As expected, wild-type FHOD1 only weakly enhanced the formation of ventral stress fibers, whereas the FHOD1 1-1011 mutant significantly stimulated their formation almost threefold as compared with EGFP (Fig. 8A). Interestingly, the FHOD1 V228E mutant had an even more pronounced effect on the formation of ventral stress fibers. This was surprising, as confocal imaging and cross-correlation analysis of FHOD1 fluorescence with the corresponding actin signals revealed that the FHOD1 V228E mutant was only weakly associated with stress fibers as compared with FHOD1 1-1011 (Fig. 8B,C).
Although FHOD1 1-1011 lacks the three C-terminal residues that can be phosphorylated by the upstream regulator ROCK, these sites are still present in FHOD1 V228E. To study the effects of ROCK inhibition on the localization of these mutants, we combined acute pharmacological inhibition of the kinase with live-cell TIRF microscopy. We found that wild-type FHOD1 rapidly dissociated from actin fibers after the addition of the ROCK inhibitor Y-27632 (50 mM for 5 min). This suggests that the phosphorylation of FHOD1 by ROCK antagonizes constitutive de-phosphorylation, thereby allowing dynamic control of the association of FHOD1 with stress fibers (Fig. 8D,E). As expected, FHOD1 1-1011 remained mostly at stress fibers after ROCK inhibition. However, FHOD1 V228E dissociated substantially from stress fibers after Y-27632 treatment, similar to wild-type FHOD1 (Fig. 8D, arrows, 8E). This finding suggests that, even in the FHOD1 V228E mutant, phosphorylation of the C-terminus still controls the actin-binding function that is mediated by the N-terminal GBD-FH3 domains. Taken together, our observations suggest that the C-terminus plays not only an important role in FHOD1 auto-inhibition, but also a more direct functional role in the stimulation of stress fiber formation by FHOD1.

DISCUSSION
Here, we show that FHOD1 controls the coordinated maturation of stress fibers by its distinct effects on two dynamic precursor stress fiber types: it stimulates the formation of transverse arcs and inhibits the growth of dorsal fibers. These combined activities of FHOD1 lead to the efficient formation of less-dynamic ventral stress fibers. Our detailed analysis of this process offers novel mechanistic insight into the role of FHOD1 in the dynamic processes that build contractile actin structures in cells.
Interestingly, FHOD1 was preferentially associated with actin structures that have anti-parallel orientation, such as ventral stress fibers and transverse arcs. By contrast, FHOD1 was largely absent from dorsal fibers, which mainly consist of (C) Acute stress fiber formation upon cytochalasin D washout. Cells expressing EGFP alone, EGFP-FHOD1 wild-type (FHOD1 WT) and active EGFP-FHOD1 mutants were treated with cytochalasin D (2 mM for 90 min), followed by extensive washout to stimulate stress fiber recovery. Cells were fixed and stained with Rhodamine-phalloidin (F-actin) at 30 min after washout. The relative area of the cells that was covered with stress fibers was quantified. Data show the mean6s.e.m. of n556-74 cells for each condition from three experiments. Welch's t-test was performed in C. n.s. not significant; ***P,0.001. (D) 3D organization of stress fiber bundles stimulated by FHOD1 V228E. Depicted is a single confocal plane on the ventral cell side (left) and maximum projections of confocal Z-stacks (F-actin, red; paxillin, green). Z coordinates are indicated by the color bar (right). The arrow points to the dorsally localized region of a linear stress fiber and the arrowhead shows the ventral end of a linear fiber. Scale bar: 10 mm.
parallel actin-bundles (Cramer et al., 1997;Svitkina et al., 1997;Pellegrin and Mellor, 2007). This preference of FHOD1 for anti-parallel actin fibers is shared by myosin II (Verkhovsky and Borisy, 1993), which has to bind such anti-parallel fibers to generate contractile forces (Clark et al., 2007). The only region close to dorsal stress fibers where there was association of FHOD1 and myosin II overlapped substantially with focal adhesions, suggesting that a subpopulation of anti-parallel-oriented actin filaments, albeit with currently unknown function, might be present in that region. It is currently unclear how FHOD1 might be able to associate selectively with anti-parallel actin bundles. One possibility would be by the binding of FHOD1 to myosin II, either directly or indirectly through adapter proteins. This idea is supported by our observation that FHOD1 is colocalized with myosin-II-rich stripes and not with a-actinin within stress fibers. The N-terminal GBD-FH3 region is required for the colocalization of FHOD1with myosin II and might thus mediate such coordinated recruitment.
Both myosin II and FHOD1 are activated by phosphorylation that is mediated by the serine/threonine kinase ROCK Kimura et al., 1996;Hannemann et al., 2008;Takeya et al., 2008). The upstream activator of ROCK, the small GTPase RhoA, is known to form spatial activity gradients that correlate with cell motility, cell shape and polarity (Pertz et al., 2006;Nalbant et al., 2009). Thus, in these cellular processes, the RhoA-ROCK pathway is a likely candidate to coordinate both the spatio-temporal organization of stress fiber formation through FHOD1 and the contractility of these stress fibers through myosin II.
Although the regulation of FHOD1 by ROCK-mediated phosphorylation was proposed previously (Hannemann et al., 2008;Takeya et al., 2008), the functional role of this modification was not fully explored. On the one hand, the effect of ROCK on the subcellular localization of FHOD1 and the timescale at which ROCK mediates FHOD1 regulation were unknown. On the other hand, experiments that use C-terminal truncation mutants or phosphomimetic mutations are difficult to interpret because they might affect other functions of the C-terminus that are not directly related to auto-inhibition. Indeed, the C-terminal DAD region of the formin mDia1 was recently suggested to play an active role in filament nucleation (Gould et al., 2011). Here, we show that, in comparison to the C-terminal truncation (FHOD1 1-1011), the point mutation V228E is an even stronger activator of stress fiber formation, although it shows significantly less colocalization with Fig. 8. The C-terminal regulatory domain is required for FHOD1 effector function to stimulate efficient stress fiber maturation. (A) The FHOD1 V228E mutant is a more potent inducer of ventral stress fibers compared with the Cterminally truncated FHOD1 1-1011. Maximum projections of confocal Z-stacks (F-actin) of cells transfected with EGFP-FHOD1 constructs and EGFP control were generated. Cells covered entirely with linear stress fibers (linear stress fiber phenotype) were quantified. Data show the means6s.d. from three experiments (n546-62 cells). (B) Stress fiber association of constitutively active FHOD1 mutants. Upper panel, representative maximum projections of confocal Zstacks showing EGFP-FHOD1 and F-actin. Lower panel, line scans along the red and green arrows shown in the upper panels were used to analyze the correlation of FHOD1 (green) and F-actin (red) distribution (ImageJ, plot profile function). Lines were drawn perpendicular to stress fibers from the leading edge to the center. The normalized intensity profiles are shown. Scale bars: 10 mm. (C) Pearson's correlation coefficient (GraphPad Prism) was used to quantify the correlation of FHOD1 and F-actin intensities obtained from line scans (n526 for FHOD1 1-1011 and n524 FHOD1 V228E cells from three experiments). Red lines indicate the mean. (D) TIRF images of a representative cell expressing EGFP-FHOD1 wild-type and mCherry-actin prior to and 5 min post ROCK inhibition by Y-27632 (50 mM). Yellow arrows show a representative stress fiber. Scale bar: 20 mm. (E) Quantification of stress fiber dissociation of EGFP-FHOD1 wild-type and the two constitutively activated constructs EGFP-FHOD1 V228E and EGFP-FHOD1 1-1011. The decrease in the average fluorescence intensity along individual stress fibers in each EGFP-FHOD1 TIRF image was measured using ImageJ software and normalized to the change of the corresponding mCherry-actin signal (n514-16 cells from three to four experiments for each construct; red line marks the mean). ***P,0.001, **P,0.01, *P,0.05, n.s., not significant. stress fibers. This suggests that the C-terminus of FHOD1 indeed plays additional roles in FHOD1 function, potentially by augmenting actin monomer recruitment, as reported for the formins mDia1 and FMNL2 (Gould et al., 2011;Block et al., 2012). Alternatively, the sensitivity of the FHOD1 association with stress fibers to ROCK inhibition suggests that dynamic binding and unbinding of FHOD1 to stress fibers, mediated by cycles of phosphorylation and dephosphorylation, might be necessary to activate FHOD1 function fully.
One of the most intriguing observations in this study was that FHOD1 has at least two distinct cellular functions: (1) stimulating stress fibers that contain anti-parallel filaments (arcs and ventral stress fibers) and (2) inhibiting the growth of stress fibers that contain long parallel filaments (dorsal stress fibers). Our studies give important novel insight into the molecular basis of how FHOD1 could perform those functions: (1) We were able to narrow down the region of FHOD1 that mediates the association of the protein with stress fibers and its colocalization with myosin II to amino acids 1-339.
(2) We found that a truncation mutant lacking the GBD-FH3 domains accumulated in the lamellipodium and along the entire length of focal adhesions. (3) We found that persistent association of FHOD1 with stress fibers is not necessary for the stimulation of stress fiber formation. (4) Further truncation of a helical domain (396-573) that was recently suggested to mediate actin filament bundling in vitro (Schönichen et al., 2013) prevented the FHOD1-mediated stimulation of stress fibers. Together, these observations suggest that the helical, FH1-, FH2-and C-terminal domains are crucial for both the stimulation of arcs by FHOD1 and for their turn-over into ventral stress fibers. Previous studies proposed that the short bundled actin filaments that form transverse arcs originate from the lamellipodium (Hotulainen and Lappalainen, 2006;Burnette et al., 2011;Tojkander et al., 2012). Therefore, the lamellipodial localization of the highly active truncation mutant 340-1164 during acute arc stimulation suggests that FHOD1 might mediate the bundling of preformed short actin filaments at the leading edge of the cell through its helical domain, and that these filaments are then efficiently assembled into transverse arcs at the interface with the lamellum.
Classically, formins are thought to act as processive actincapping proteins that can stimulate actin polymerization (Zigmond et al., 2003;Shemesh et al., 2005b;Otomo et al., 2005). However, recent biochemical studies show that purified active FHOD1 inhibits actin polymerization in addition to filament bundling (Schönichen et al., 2013). In cells, dorsal stress fibers require rapid filament elongation at the associated focal adhesion, which is mediated by the efficient processive capping protein mDia1 (Hotulainen and Lappalainen, 2006;Oakes et al., 2012). At these focal adhesions, FHOD1 might act as a capping protein, which would inhibit the growth of associated dorsal stress fibers by competing with mDia1. Our observation that a FHOD1 mutant lacking the stress-fibertargeting GBD-FH3 domains associates more prominently with focal adhesions in steady-state conditions further supports this idea. In the lamellipodium, linear filaments are generated by mDia2 or FMNL2 (Yang et al., 2007;Tojkander et al., 2012;Block et al., 2012). Here, FHOD1 might enrich the pool of short actin filaments either by permanent or slow processive capping. Concomitantly, FHOD1 might promote transverse arc assembly by bundling those preformed short filaments. Thus, it is intriguing that FHOD1 by the same molecular activity might control opposing processes at dorsal stress fibers and arcs. Finally, whereas the central and C-terminal domains of FHOD1 are responsible for robust filament formation, the N-terminal stressfiber-targeting region (1-339) might facilitate the precise spatial targeting of FHOD1 to 'fine tune' its cellular function. These processes in the generation of mature contractile stress fibers are summarized in a working model that integrates the two proposed molecular functions of FHOD1 -filament capping and bundlinginto a dynamic process that is tightly controlled in space and time (supplementary material Fig. S6).
In summary, our study establishes a key role for the formin FHOD1 in the spatio-temporal control of the dynamics of the actin filament meshwork in adherent cells. FHOD1 promotes the efficient formation of transverse actin arcs in the leading edge and, by restricting the length of dorsal stress fibers, it promotes the coordinated turn-over of actin arcs into mature contractile stress fibers.
Cells were treated with the mycotoxin cytochalasin D (2 mM for 90 min) (C8273, Sigma Aldrich) to allow complete depolymerization of F-actin. Treatment started 72 h after the transfection of cells with siRNAs or 16 h after the transfection of cells with EGFP or EGFP-FHOD1 constructs. The recovery of stress fibers was stimulated by five washing steps with growth medium. The selective ROCK inhibitor Y-27632 (Y0503) was purchased from Sigma-Aldrich.

Microscopy
Epifluorescence and phase-contrast imaging were performed with an Eclipse Ti inverted microscope (Nikon) equipped with a motorized stage, a built-in Perfect Focus System and a CoolSNAP HQ2 camera (Photometrics). Images were acquired using a 620/0.45NA air or a 660/1.40NA oil-immersion objective. Acquisition was controlled by NIS-Elements Imaging Software (Nikon). Random positions were chosen using the multi-position tool.
Time-lapse TIRF microscopy was performed on an Eclipse Ti-E (Nikon) inverted microscope with a motorized TIRF Illuminator Unit (Nikon), an Andor AOTF Laser Combiner (Andor Technology) and a Clara Interline CCD camera (Andor Technology). The laser lines used for the excitation of EGFP and mCherry were 488 nm and 561 nm, respectively. Images were acquired by using an Apo TIRF 6100/1.49NA oil-immersion objective (Nikon). Acquisition was controlled by Andor IQ Software (Andor Technology).
Wound healing assay IBIDI culture inserts (IBIDI) were used to study wound-healing efficiency. Cells (8610 3 ) were seeded in each compartment of the culture inserts and were directly transfected with 10 nM siRNA using HiPerFect (Qiagen). The inserts were removed 72 h after siRNA transfection and the growth medium was replaced with growth medium supplemented with 10 mM HEPES. Cell migration was monitored by time-lapse phase-contrast microscopy (with a frame rate of 30 s). For each condition, ten different positions along the wound were recorded in each experiment. The migration efficiency was quantified by measuring the migrated distance (ImageJ). Lines were drawn along the wound edge at the start and end of the assay and the average movement was calculated from all positions.

Cell adhesion and spreading
For adhesion and spreading assays, cells were detached from culture dishes using Trypsin-EDTA and were plated onto collagen-type-I-coated glass coverslips in growth medium 72 h after siRNA transfection. After 15, 30 or 60 min, the coverslips were rinsed once with PBS, fixed with 4% formaldehyde and stained with DAPI (nuclei) and Rhodaminephalloidin (F-actin). Epifluorescence images were acquired using a 620/ 0.45NA air objective. Adhered cells were quantified from 25 random regions for each condition and time-point. To quantify cell spreading, 20 random positions were acquired using a 660/1.40NA oil-immersion objective for each condition, and the cell area was measured using ImageJ software. The focal adhesion size and number per cell were quantified from confocal images of cells expressing mKate-paxillin using ImageJ. Thresholds were set manually to isolate focal adhesions properly. The average area and number of focal adhesions per cell were quantified from binary images using the analyze particle tool in ImageJ. Particles smaller than 0.25 mm 2 and larger than 10 mm 2 were excluded to avoid nonspecific background signal and neighboring adhesions that were not properly separated.

RNA isolation, reverse transcription and quantitative real-time PCR analysis
Total RNA was extracted using the RNeasy Mini kit (Qiagen). Reverse transcription was performed with SuperscriptH II Reverse Transcriptase (Life Technologies) according to the manufacturer's protocols using 500 ng of total RNA. Quantitative real-time PCR was performed with a StepOnePlus TM Real-Time PCR system (Life Technologies) using the Fast SYBR H Green Master Mix (Life Technologies) and human FHOD1specific primer pairs (forward, 59-TACACGGTCACCCTCATCAA-39; reverse, 59-AGTGCATCCGTCACATCGTA-39). GAPDH, ACTB and RRN18S were used as housekeeping genes (QuantiTect Primer Assays, Qiagen). Relative quantification was performed using the efficiencycorrected relative quantification method (Pfaffl, 2001). Primer efficiencies were calculated for each primer pair by performing dilution series experiments.

Western blot analysis
Cells were washed once with ice-cold PBS and were lysed in ice-cold radioimmunoprecipitation assay buffer [50 mM Tris pH 7.5, 150 mM NaCl, 1% NP-40, 0.25% sodium deoxycholate, 1 mM EDTA, 16protease inhibitor cocktail and 16PhosSTOP phosphatase inhibitor cocktail (Roche)]. Insoluble cell debris were removed by centrifugation at 13,000 g for 10 min at 4˚C. The protein concentration of the supernatants was determined by the Bradford method (BioRad). Equal amounts of total protein were mixed with 56Laemmli sample buffer, boiled at 95˚C for 10 min and separated by SDS-PAGE. After electrophoresis, proteins were transferred to a polyvinylidene fluoride (PVDF) membrane (Thermo Scientific) using a Biometra fastblot B34 blotting device (Biometra). Blots were blocked for 1 h at room temperature with 5% nonfat dry milk in TBS-T (20 mM Tris pH 7.6, 137 mM NaCl and 0.1% Tween-20) and incubated overnight at 4˚C with the primary antibodies anti-FHOD1 (1:200, FM3521, ECM Bioscience) or anti-a-tubulin (1:20,000, clone B-5-1-2, Sigma-Aldrich). Membranes were washed three times with TBS-T and were incubated with HRP-conjugated antimouse secondary antibody (1:20,000; Santa Cruz Biotechnology) for 1 h at room temperature. After additional washing steps with TBS-T and TBS (20 mM Tris pH 7.6 and 137 mM NaCl), protein bands were visualized using ECL western blotting substrate (Pierce). Images were captured with a Fusion Fx7 system and were quantified with Bio-1D software (Peqlab).

Statistical analysis
Independent two-tailed Student's t-tests were used unless otherwise stated. In cases where there were unequal variances, Welch's t-test was used. The resulting P-values are indicated in the figures and legends.