FHOD1 is a combined actin filament capping and bundling factor that selectively associates with actin arcs and stress fibers

Cell movements and dynamic changes in cell morphology play key roles during tissue regeneration, immune responses, embryonic development, and wound healing in eukaryotic organisms. Remodeling of the actin cytoskeleton regulates many of the associated cellular functions such as protrusion, adhesion, and retraction. Although a wide array of actin filament assemblies exists in multicellular organisms, three main categories can be distinguished that play fundamental roles in cell migration and cellular morphogenesis: (1) the lamellipodium, a veil-like membrane protrusion at the leading edge of a migrating cell containing a branched actin meshwork, (2) filopodia and microvilli, finger-like protrusions of the plasma membrane, stabilized by an F-actin bundle of varying thickness, and (3) actin stress fibers that can form at least three different assembly categories, such as dorsal stress fibers, actin arcs and ventral stress fibers (Hotulainen and Lappalainen, 2006; Ridley, 2011).

A rich variety of regulators assembles, maintains and destroys actin cytoskeletal structures. Among these, formins constitute the largest and most diverse family of actin polymerization promoting factors (Chesarone et al., 2010). Fifteen human formin variants are known, many of which contain additional splicing isoforms that cluster into eight different families (Schönichen and Geyer, 2010). The defining element to all formins is the presence of an FH2 domain that renders the protein dimeric. For many formins (e.g. mDia1), it has been shown that the FH2 domain promotes barbed end elongation of straight actin filaments, most likely supported in combination with the proline-rich FH1 domain, which concentrates actin–profilin units at the site of polymerization. Diaphanous related formins (DRFs) contain an additional array of regulating domains at their N-termini that constitute a GTPase-binding domain (GBD) for activation, a canonical FH3 domain composed of armadillo repeats and additional helical elements. DRFs are negatively regulated by an intra-molecular interaction of a C-terminal Diaphanous autoregulation domain (DAD) with the N-terminal FH3 domain (Alberts 2001; Watanabe et al., 1999), whose release is thought to be mediated by competition displacement with an activated Rho-GTPase (Chesarone et al., 2010; Schönichen and Geyer, 2010). In contrast, activation of the human FH1/FH2 domain-containing protein 1 (FHOD1) was shown to require phosphorylation of serine and threonine residues within the DAD by the Rho-associated kinase ROCK (Takeya et al., 2008).

De novo actin filament formation requires an initial nucleation process, i.e. the stabilization of an actin dimer or trimer that acts as nucleating seed followed by an elongation process, during which new actin subunits are preferentially added to the fast growing plus or barbed ends. Actin dimer or trimer formation is kinetically unfavorable, but strongly supported by dimeric formin FH2 domains. In addition, formins support subsequent elongation leading to rapid polymerization of straight, unbranched actin filaments (Chesarone et al., 2010). Based on our current understanding, formins stay bound to the barbed end of the growing filament as leaky cappers, where they promote filament elongation. In cells, this behavior effectively leads to transport of the formin through the cytoplasm or into protrusions, by transmitting forces from actin polymerization to the plasma membrane (Romero et al., 2004; Zigmond et al., 2003). To date, the effect of full length formins on actin polymerization has not been analyzed in vitro, mainly due to technical issues to purify or produce the stable, dimeric ∼125 kDa multidomain protein. Instead, only deletion constructs containing FH1 and FH2 domains of mammalian and ascomycota formins were investigated.

In cells, FHOD1 has been shown to induce stress fiber formation in conjunction with either activated RhoA or Rac1 GTPases (Gasteier et al., 2003; Takeya and Sumimoto, 2003; Westendorf, 2001). Stress fibers are contractile actomyosin bundles, which play important roles in embryonic development, wound healing and cell migration (Hotulainen and Lappalainen, 2006). They are composed of numerous, short actin filaments with alternating polarity (Cramer et al., 1997). These filaments are cross-linked by α-actinin and possibly also by other actin-bundling proteins. α-Actinin and myosin display a periodic distribution along stress fibers typical also for other types of contractile structures, such as myofibrils of muscle cells. In the present study, we investigated the function of FHOD1 by performing in vitro actin polymerization and bundling assays and followed its distribution and movements on actin-rich structures in living cells with high temporal and spatial resolution. Based on the functional characterization of FHOD1 domains, a model is presented, which proposes that FHOD1 stabilizes F-actin bundles by protecting barbed ends and connecting them to adjacent actin filaments, a function that is enforced in the presence of tropomyosin.

To explore FHOD1's actin assembly properties in vitro, several FHOD1 constructs were generated (Fig. 1A). Besides the 127 kDa full length protein (fl), we generated a truncation mutant lacking the C-terminal DAD (ΔC), an N-terminal deletion mutant composed of the canonical FH1–FH2–DAD domains (ΔN), a mutant containing only the N-terminal regulation domains (N), and the central part connecting the canonical FH3 and FH1–FH2 domains (M). The FH2 domain of mDia1 served as control protein in all actin assembly assays. Although the mDia1 FH2 domain could be expressed in E. coli, expression of all recombinant FHOD1 constructs containing the FH2 domain required the baculovirus expression system, indicating that FHOD1's FH2 domain potentially contains unknown structural elements that require eukaryotic chaperones or posttranslational modifications for proper maturation. However, all proteins could be purified to homogeneity (Fig. 1B). Single particle electron micrographs confirmed the integrity of full length FHOD1 by the characteristic torus-shaped conformation (Fig. 1C).

We examined actin filament nucleation and elongation activities of FHOD1 proteins by fluorescence measurements of pyrenyl-actin (Fig. 1D). At a concentration of 2 µM, actin spontaneously polymerized at a basal level, while the FH2 domain of mDia1 effectively stimulated actin elongation, similarly as described before (Li and Higgs, 2003; Romero et al., 2004). In contrast, addition of 0.3 µM FHOD1 either as full length protein, or ΔC or ΔN variants, almost completely inhibited actin polymerization, suggesting capping activity of this formin (Fig. 1D). The inhibitory effect of FHOD1 ΔC and fl on actin polymerization could be confirmed by total internal reflection fluorescence (TIRF) microscopy, where no indication of filament formation was seen (Fig. 1E). The N-terminal FHOD1 domains 1–573 lacking the FH2 domain instead showed no effect on actin polymerization in the pyrene-actin assay (supplementary material Fig. S1).

In cells, expression of FHOD1 ΔC leads to a robust increase in number and thickness of actin stress fibers, which suggests an active role of this formin in cellular actin organization (Gasteier et al., 2003; Schönichen et al., 2006). Thus, our finding that FHOD1 might be a negative regulator of actin polymerization was surprising and prompted us to investigate the molecular role of FHOD1 on actin polymerization in more detail. In time resolved actin polymerization assays, addition of the proposed constitutively active FHOD1 ΔC variant in increasing concentrations from 1 to 100 nM led to dose-dependent inhibition of actin polymerization (Fig. 2A). A similar effect was observed when full length FHOD1 was applied (Fig. 2B), indicating its ability to interact with actin despite presence of its proposed autoinhibitory regulation. Addition of the DAD at 10-fold molar excess over FHOD1 fl to release a potential autoinhibitory state did not stimulate polymerization, but again inhibited actin polymerization. Phosphorylation of FHOD1 fl by incubation with recombinant ROCK1, which had been shown to activate the protein in cells by phosphorylation of the DAD (Hannemann et al., 2008; Takeya et al., 2008), did not increase polymerization, but suggested enhanced inhibition (Fig. 2B). In presence of 3 µM profilin, addition of the FHOD1 FH1–FH2–DAD domain construct (ΔN) led again to significantly reduced polymerization activity (Fig. 2C).

Actin filament elongation experiments were performed at reduced concentrations of 0.5 µM G-actin, which allowed distinguishing between actin nucleation and elongation capabilities, as they require the presence of actin seeds for free barbed end elongation activity (Fig. 2D,E). Increasing concentrations of FHOD1 ΔC led to dose-dependent reduction of the elongation activity both in the absence or presence of 3 µM profilin. The inhibitory effect of FHOD1 ΔC was reduced in presence of profilin, suggesting that FHOD1 is able to elongate actin filaments by incorporation of actin–profilin complexes (Fig. 2E). This effect was quantified, indicating comparable exponential reduction of filament elongation by increasing amounts of FHOD1 ΔC in the presence or absence of profilin albeit to different base-line activities (Fig. 2F). In conclusion, we could not detect actin nucleation activity of FHOD1, neither as full length protein nor in the truncated form lacking the autoregulation domain, but found weak elongation activity in presence of profilin.

As FHOD1 was found to inhibit actin polymerization and elongation, we wondered if FHOD1 is an actin filament capping protein. We thus examined FHOD1 fl and ΔC in F-actin dilution assays and found that both proteins inhibited F-actin depolymerization (Fig. 2G,H). FHOD1 fl showed weaker inhibition at 300 nM concentration than FHOD1 ΔC suggesting that autoinhibition regulates association with F-actin barbed ends (Fig. 2G). A continuous decrease in the F-actin depolymerization activity was observed upon increasing concentrations of FHOD1 ΔC (Fig. 2H) as seen as well from the quantified depolymerization rates (Fig. 2I).

Activation of FHOD1 in cells induces stress fibers (Gasteier et al., 2003; Schönichen et al., 2006; Takeya and Sumimoto, 2003). Because stress fibers contain F-actin bundles, we tested how FHOD1 binds to F-actin and if it is sufficient to induce F-actin bundles in vitro. High speed centrifugation assays revealed that all tested FHOD1 constructs (fl, ΔC, ΔN, N and M), as well as mDia1 FH2 appeared in the pellet fraction, and thus bind to F-actin (Fig. 3A). These results confirmed previous studies that suggested F-actin binding activities of an N-terminal FHOD1 construct lacking the FH2 domain (Takeya and Sumimoto, 2003). As FHOD1 N and ΔN do not overlap significantly in primary sequence, this observation suggests that two independent F-actin binding sides are present in FHOD1. Interestingly, while FHOD1 fl and ΔC bound F-actin very efficiently, only a small fraction of constructs FHOD1 M, ΔN and N and mDia1 FH2 was present in the pellet fractions. In case of FHOD1 ΔN and the control protein mDia1, this finding can be explained by binding exclusively to barbed ends via the FH2 domain shared in these constructs. Because the number of barbed ends is significantly smaller than the number of potential binding sites along actin filaments, the amount of FH2 domain binding in the pellet fraction is expected to be smaller. However, the independent F-actin binding site in the FHOD1 N-terminus might be able to bind to the sides of filaments. Additional barbed-end binding or dimerization via the FH2 domain could increase overall actin binding efficiency. Indeed, compared to FHOD1 N and M, significantly more FHOD1 fl and ΔC was pulled down by F-actin, indicating that two regions of FHOD1 can act synergistically to increase F-actin binding efficiency.

The F-actin bundling activity of FHOD1 constructs was next analyzed using low speed centrifugation. The data obtained showed that FHOD1 fl and particularly FHOD1 ΔC induced sedimentation of F-actin (Fig. 3B). Low-speed sedimentation of actin in the presence of constructs M and ΔN (at the limited amounts available) was comparable to negative controls, suggesting that those N-terminal domains themselves are not able to bundle F-actin (Fig. 3B). These data suggest that two independent actin-binding sites at the N- and C-terminal domains of FHOD1 cooperate for actin filament bundling. Whereas the N-terminus is likely to have F-actin side-binding activity, barbed-end binding and/or protein dimerization via the FH2 domain is required for the formation of stable higher-order structures composed of FHOD1 and actin bundles.

To examine the capability of FHOD1 for actin bundling we applied TIRF microscopy using polymerizing actin filaments. Two FHOD1 proteins ΔC and fl were added at similar concentrations to polymerizing actin filaments using actin alone as control. Formation of thick F-actin bundles and a rigid F-actin network was indeed observed upon FHOD1 addition as actin filaments stopped fluctuating (Fig. 3C; supplementary material Movie 1). This effect was much stronger for FHOD1 ΔC compared to fl, indicating that release of the DAD from the N-terminal FH3 recognition domain is required for full bundling activity.

In addition we tested F-actin bundling by FHOD1 using electron microscopy (EM) after negative staining. FHOD1 ΔC and fl were incubated with freshly polymerized F-actin at molar ratios of 1∶1, 1∶10, 1∶100 and 1∶1,000 (FHOD1 versus actin monomer concentration) for varying periods of time. F-actin was polymerized in the presence of the barbed end capper gelsolin at 100∶1 ratio to obtain filaments of similar lengths. The actin concentration was set to 0.1 mg/ml. Furthermore, we employed F-actin decorated with skeletal muscle tropomyosin at a molar ratio of 7∶1 in some experiments. In presence of FHOD1 fl we detected normal appearing actin filaments and occasionally bundles comprising 2 to 3 actin filaments and in between noisy sprinkles and torus-shaped structures, most probably indicating free FHOD1 (Fig. 3D, lower panels). Otherwise no indication was obtained for an alteration of the filament assembly and shape. In contrast, FHOD1 ΔC induced F-actin bundles containing about 2 to 5 parallel actin filaments at ratios of 1∶10 and 1∶100 after overnight incubation at room temperature (Fig. 3D, middle panels). Upon addition of gelsolin, the actin filaments appeared shortened, potentially due to the increasing number of barbed end cappers, while the bundle thickness remained similar. FHOD1 also bundled F-actin decorated with tropomyosin as indicated by an increase in the length and thickness of F-actin bundles compared to the treatment with gelsolin only (Fig. 3D, right panels). This suggests that the F-actin side-binding capacity of FHOD1 is maintained upon filament stabilization with tropomyosin.

Our in vitro data suggest that FHOD1 does not nucleate and assemble actin filaments de novo as other formins but rather bundles and stabilizes existing actin filaments. We next sought to investigate the biological significance of these in vitro activities. Previous studies applied speckle microscopy for detection of individual molecular complexes of related actin regulators, such as mDia and the Arp2/3 complex (Higashida et al., 2004; Miyoshi et al., 2006) as individual ‘speckles’ (Danuser and Waterman-Storer, 2006; Watanabe et al., 1999). We used this method to detect potential enrichment of several FHOD1 constructs in subcellular structures and to study its dynamic localization in living cells, and focused our analysis on the actin-rich cell cortex by using TIRF microscopy. We therefore expressed FHOD1 constructs at low levels in two mammalian cell lines: (1) Neuro-2a neuroblastoma cells expressing dominant positive Rac1, for robust formation of large, dynamic lamellipodia and filopodia, along with less dynamic actin arcs, and (2) COS7 fibroblast-like kidney cells, which form smaller, dynamic lamellipodia along with less dynamic ventral stress fibers terminating in focal adhesions.

In both cell types, wild-type FHOD1 co-localized strongly with actin structures known to form anti-parallel bundles: actin arcs and ventral stress fibers (Fig. 4A,B). This localization pattern is even more pronounced for the dominant active ΔC fragment. In contrast, FHOD1 ΔN, which contains the proposed barbed-end binding FH1–FH2 domains, did not show any particular enrichment at the actin-rich cell cortex. The N-terminal fragment 1–573, which lacks the actin barbed-end binding FH1–FH2 domains, also strongly localized to actin arcs and ventral stress fibers. Taken together, our data suggested that the N-terminal region is sufficient to confer selective binding of FHOD1 to actin-arcs and stress fibers.

Constitutively active mutants of the related formin mDia1 are highly mobile in living cells due to their processive barbed-end capping activity. Because FHOD1 elongates actin filaments only poorly in vitro, we examined the dynamic behavior of individual FHOD1 ΔC speckles to test, whether this constitutively active mutant shows typical motility of a processive cap in living cells. It should be noted here that the FHOD1 ΔN mutant is not expected to act as a constitutively active mutant, as it is lacking the essential N-terminal F-actin binding site. mDia1 lacks such N-terminal F-actin binding site, and a ΔN mutant is widely accepted to be constitutively active.

Kymograph analyses in lamellipodia showed that FHOD1 ΔC speckles were distributed sparsely throughout the lamellipodium, where they were predominantly shifted laterally at the speed and in the direction of retrograde flow (Fig. 5A,B, left; supplementary material Movies 2 and 3, upper panel), suggesting that FHOD1 ΔC interacts with preformed actin filaments either by binding at the side of filaments, by capping filament ends or both. FHOD1 ΔC speckles localized to actin arcs at the base of the lamellipodium, which was characterized by much slower retrograde movements (supplementary material Fig. S2; Movie 4). In all actin-rich regions, most FHOD1 ΔC speckles were only transiently detected for one single 400 ms frame (supplementary material Movie 3) suggesting that cortex-associated FHOD1 ΔC exchanges rapidly with a cytosolic pool.

In contrast, the dynamic behavior of mDia1 ΔN speckles was highly divergent from FHOD1 ΔC speckles. mDia1 ΔN did not translocate with the retrograde flow towards the cell interior, but rather in the opposite direction towards the leading edge (anterograde) at much faster speeds (Fig. 5A,B, right; supplementary material Movies 2 and 3, lower panel). As an apparent consequence of actin filament elongation and its processive capping function, mDia1 ΔN accumulated at the very leading edge of the cell. This is in contrast to FHOD1 ΔC, which accumulated at the base of the lamellipodium. Thus, our observation, that FHOD1 does not preferentially bind to cellular locations enriched in barbed ends, but rather interacts transiently with preformed actin structures, suggested that FHOD1 does not primarily function as a processive actin filament capper.

While we never observed directional, anterograde translocation of FHOD1 ΔC speckles, a small subpopulation of FHOD1 ΔC speckles accumulated transiently at the leading edge of migrating cells. In a few cases, we even observed a strong accumulation of FHOD1 ΔC speckles at this cellular region (supplementary material Fig. S2; Movie 5). However, the dynamics of the majority of FHOD1 ΔC speckles was in agreement with a role for FHOD1 as an actin filament side-binding and capping protein, which is particularly associated with actin filament bundles in actin arcs and stress fibers.

In the present study we show that human FHOD1 does not de novo polymerize actin filaments, but instead binds dynamically to and bundles actin filaments throughout the cell. The canonical FH2 domain exhibits a capping activity on the actin filament barbed end, whereas the central helical region in between the FH3 and FH1 domains interacts with sides of actin filaments.

Given that FHOD1 ΔC has the strongest effect on F-actin bundling suggests that this activity is regulated by the interaction of the DAD with the FH3 domain. In cells, activation of FHOD1 has been shown to require phosphorylation by ROCK1 (Hannemann et al., 2008; Takeya et al., 2008). In contrast, an activation mechanism via binding to a lipidated Rho-GTPase, as e.g. described for mDia1 (Watanabe et al., 1999), that would lead to retention of the formin at the plasma membrane, is not observed for FHOD1 (Schulte et al., 2008). Instead, activated FHOD1 transiently associates with various actin structures and ultimately accumulates in actin arcs and stress fibers, where it could protect the filament strain from depolymerization while simultaneously binding to neighboring actin filaments. FHOD1 thus represents a new class of actin regulators by combining capping activity at the barbed end with F-actin side-binding activity to effectively protect actin filaments in cytoskeletal structures.

Formins were identified as de novo actin filament nucleators and are often referred to as actin nucleating factors in the current literature. However, we could not confirm a ‘formin-typical’ activity for FHOD1, but found that FHOD1 inhibits de novo actin nucleation in vitro. FHOD1 does not seem to be an outlier in this respect as some formins inhibit rather than increase actin polymerization in vitro, including FHOD3 (Taniguchi et al., 2009), and FMNL2 (Block et al., 2012). Hence, nucleation of actin filaments in vitro does not appear to be a common property of formins. In fact, it is emerging that actin nucleation activity of formins may involve additional cellular co-factors: The yeast formin Bni1 interacts via the FH2 domain with the polarity factor Bud6 (Moseley and Goode, 2005; Tu et al., 2012), the drosophila formin Cappuccino and Spire cooperate for actin nucleation (Quinlan et al., 2007), and mDia1 synergizes with the APC (adenomatous polyposis coli) protein for actin nucleation (Breitsprecher et al., 2012; Okada et al., 2010). Thus, we cannot rule out that FHOD1 may associate with an additional factor for actin nucleation in cells. FHOD1 has been shown not only to be activated by but also to interact with ROCK1 via the N-terminal half of its FH2 domain (Hannemann et al., 2008), and thus ROCK1 may be a candidate to potentially cooperate in actin filament nucleation.

It has been reported that the FH2 domain of certain formins, such as mDia2, is able to bundle actin filaments in vitro (Harris et al., 2006). Although mDia2 indeed stabilizes cortical actin and free actin filament barbed ends in focal adhesions (Gupton et al., 2007), it remains unclear whether actin filament bundling activity by the FH2 domain is relevant in cells. However, for the FHOD1 FH2 domain we neither observe actin filament bundling activity in vitro nor does it localize to F-actin in cells. In contrast, N-terminal binding regions of FHOD1 are required for in vitro bundling activity, which is also required for F-actin localization. Hence, our data suggest that the mechanisms for actin filaments bundling of FHOD1 and other formins differ strongly.

Live-cell imaging using EGFP–FHOD1 and mCherry–actin revealed that FHOD1 localized near the plasma membrane to associate with cortical F-actin. Interestingly, for a close homologue to FHOD, ForC from Dictyostelium discoideum, the N-terminal putative GTPase binding domain was shown to interact with phosphoinositides in vitro and was required for ForC targeting to cell–cell contacts and early phagocytic cups in cells (Dames et al., 2011). Similarly to FHOD1, this domain exhibits the ubiquitin superfold, and its appearance in domain architecture and structure is reminiscent to the F0 domain in Kindlin and Talin proteins that is known to target FERM proteins to focal adhesions (Legate et al., 2011; Perera et al., 2011). Deletion of the N-terminal region 1–115 in FHOD1 abrogates its capability to stabilize actin fibers (Schulte et al., 2008) indicating that association to the plasma membrane might be required for FHOD1 function. Of note, ForC does not contain a proline-rich FH1 domain that could recruit profilin–actin complexes for actin polymerization, indicating a loss of actin polymerization activity in ForC, and supporting properties of FHOD1 described in the present paper.

The F-actin bundling mechanism identified in the present study fits well to the known biological properties of FHOD1. Similar to other diaphanous formins, activated forms of FHOD1 potently induce the formation of F-actin stress fibers in mammalian cells (Gasteier et al., 2003; Schulte et al., 2008; Westendorf, 2001). As a notable difference to other DRFs, these stress fibers are significantly thicker and FHOD1 remains associated with these filament bundles to decorate the stress fibers. This phenomenon can now be attributed to FHOD1's unique properties such as sustained barbed end association and filament bundling. Moreover, the mobility of filament-associated FHOD1 away from the plasma membrane observed by live cell imaging explains why FHOD1 specifically facilitates the formation of cytoplasmic F-actin structures such as stress fibers and actin arcs. Another common feature of formins consists of their ability to induce the nuclear transcription serum response factor (SRF) to modulate the expression of specific target genes (Tominaga et al., 2000). This effect mirrors formin-mediated regulation of actin dynamics since enhanced actin polymerization liberates myocardin-related transcription factor (MRTF) cofactors from G-actin to facilitate their import into the nucleus and subsequent activation of transcription (Jurmeister et al., 2012; Staus et al., 2011). As activated FHOD1 acts as potent inducer of SRF-mediated transcription (Gasteier et al., 2003; Westendorf, 2001) without inducing de novo actin polymerization, our results indicate that reducing G-actin pools, e.g. by stabilization and/or bundling of F-actin structures, can suffice for the induction of transcriptional responses.

ROCK1 activation leads to increased actin stress fiber formation via myosin light chain phosphorylation (Wójciak-Stothard et al., 2001). However, whether FHOD1 and myosin light chain phosphorylation act cooperatively or independently to increase actin stress fibers and whether they regulate different architectures of bundled actin filaments is unknown. FHOD1 may stabilize and bundle assembled stress fibers, while myosin motors generate force for contraction. Alternatively, FHOD1 could elongate barbed ends with its FH2 domain while the N-terminal domains stay bound to adjacent actin filaments for anchoring the formin. As FH2 domains can generate piconewton forces, this could be an alternative mechanism for contraction (Kovar and Pollard, 2004).

The filament stabilizing function ascribed to FHOD1 in the present study is in line with recent analyses of FHOD1 (Mi-Mi et al., 2012) and FHOD3 (Iskratsch et al., 2010; Taniguchi et al., 2009) on the regulation of actin assembly and sarcomere organization in muscle cells. A specific splice variant of FHOD3 was shown to be located mainly in the Z disc of the mature heart muscle and may thus assist actin turnover or reduced barbed end depolymerization (Iskratsch et al., 2010). In mature myofibrils, both pointed and barbed ends of the thin filaments are capped and stabilized by tropomodulin and CapZ, respectively. While the mechanism of thin filament assembly and maintenance are still under debate (Sparrow and Schöck, 2009), FHOD3 as a barbed end capping protein could adopt a function similar to CapZ in muscle cells. Notably, FHOD3 is about 250 amino acids longer in size than FHOD1. Whereas the GBD–FH3 and FH1–FH2 elements align overall well between both homologues, it is particularly the connecting region identified in the present study as F-actin side binding element (construct M, Fig. 1A) that varies substantially between both proteins. This suggests that the mode of bundling to neighboring filaments by this region in FHOD1 could adopt an alternative function in the assembly mechanism mediated by its homologue FHOD3, e.g. in cardiomyocytes and striated muscle cells.

On the basis of these results we propose a model, in which FHOD1 selectively stabilizes actin filaments by associating to the growing actin filament at the side of filament polymerization and travelling with the filament into cell shape maintaining cytoskeletal substructures (Fig. 5C). FHOD1 could sustain actin arc formations at nascent focal adhesions that were recently proposed to mediate cellular protrusions (Burnette et al., 2011), or associate to stress fibers, which are highly decorated with myosin II and tropomyosin (Tojkander et al., 2011). The combination of capping activity at the filament barbed end with F-actin side-binding activity appears complementary to other bundling factors as α-actinin or capping factors as gelsolin, whose expression levels are however much higher in eukaryotic cells. These distinct cross-linkers could associate transiently to the actin filaments to orchestrate the reorganization of stress fibers. The diverging regulatory pathways that control these cross-linkers are integrated into the filament reorganization process and may allow for the cytoskeletal dynamics and plasticity. With these features, the mammalian formin FHOD1 represents a new class of actin regulators to effectively protect and bundle actin filaments.

Detailed protocols of plasmid cloning, protein expression and purification of recombinant FHOD1 are provided in the supplementary material. Briefly, plasmids encoding human FHOD1 1–573, 1–377 and 396–573 as well as mouse mDia1 748–1175 were transformed into T1 phage resistant BL21(DE3) cells, expressed at 20°C for 14 h and purified following standard protocols. FHOD1 constructs containing the FH2 domain (FHOD1 fl, ΔC and ΔN) were expressed in Sf21 cells using a baculovirus expression system and purified as N-terminally His-tagged fusion proteins.

Plasmids encoding human FHOD1 1–573, 1–377 and 396–573 as well as mouse mDia1 748–1175 were transformed into T1 phage resistant BL21(DE3) cells and expressed at 20°C for 14 h after induction with isopropyl-β-D-1-thiogalactopyranoside. Cells were harvested by a 20 min centrifugation step at 4,500 g, resuspended in buffer A (50 mM Hepes pH 7.6, 150 mM NaCl, 5 mM EDTA, 1 mM DTE, 1 mM PMSF) and lysed using a microfluidizer. The lysate was cleared by centrifugation at 10,000 g for 45 min. GST fusion proteins were purified using GSH Sepharose FastFlow (GE Healthcare) using Äkta Prime FPLC systems. The resin was washed with buffer A and B (50 mM NaP_i pH 8, 1 M NaCl, 5 mM EDTA, 1 mM DTE, 1 mM PMSF) until A₂₈₀ reached zero. GST fusion proteins were eluted with buffer A containing 10 mM GSH. Fractions containing the fusion protein were pooled and treated with TEV protease (home-made) in a molar ratio of 1∶200 at 4°C overnight.

FHOD1 constructs containing the FH2 domain (FHOD1 fl, ΔC and ΔN) were expressed in Sf21 cells (Invitrogen) using the pFastBac baculovirus expression system (Invitrogen). Sf21 cells were transfected using Lipofectamine (Invitrogen) with plasmids encoding N-terminally His-tagged fusion proteins. After virus amplification, Sf21 cultures were infected with concentrated virus stock and proteins were expressed for 72 h at 27°C. Cells were harvested by centrifugation at 2,500 g for 20 min at 4°C, and lysed in buffer C (50 mM Hepes pH 8, 500 mM NaCl, 20 mM imidazole, 5 mM β-mercaptoethanol, 1 mM PMSF) by sonication. The cleared lysate was applied to Ni-NTA resin (Qiagen). The resin was washed with buffer C until the OD₂₈₀ reached zero. Proteins were eluted by running a gradient from 20 to 250 mM imidazole in buffer C. Peak fractions were analyzed by SDS-PAGE and fractions containing the proteins of interest were pooled, concentrated and treated with TEV protease.

After protease cleavage, the protein solution was concentrated at 3,500 g using Amicon Ultra-15 centrifugation devices with the appropriate pore size (Millipore). 2 ml protein solutions were applied to a Superdex S75 16/60 column (GE Healthcare), equilibrated in buffer D (20 mM Hepes pH 7.5, 50 KCl, 1 mM MgCl₂, 1 mM EGTA, 1 mM DTT) with a GSH or Ni-NTA column attached in series. FHOD1 proteins eluted at its expected monomer or dimer (constructs encompassing FH2 domains) weight. Homogeneity and identity of FHOD1 proteins were determined by SDS-PAGE analysis and eventually electrospray mass spectrometry, respectively. Proteins containing the FH2 domain were immediately studied in pyrene assay measurements or stored at 4°C for less than one week. All other proteins were stored at −80°C. Profilin was expressed as His-tagged protein, purified by standard Ni-NTA chromatography and stored at −80°C. For speckle microscopy, FHOD1 constructs were subcloned into a pEGFP-based vector (Clontech), which was modified to contain a minimal CMV promoter (ΔCMV) (Watanabe and Mitchison, 2002).

Actin nucleation and F-actin elongation assays were performed as described (Moseley et al., 2004). All proteins investigated were carefully dialyzed or gel filtrated in buffer B to avoid artifacts due to buffer differences. Pyrene-labeled actin was purchased from Cytoskeleton Inc., and dissolved to a concentration of 20 µg/µl. 5 µl aliquots were stored at −80°C. After thawing, one aliquot was diluted in 225 µl G buffer and actin as depolymerized by incubation at 23°C for 1 h. This aliquot was then centrifuged for 1 h at 80,000 g using a Beckman Optima TL-100 and a TLA-100 rotor at 4°C to get rid of actin nuclei. 200 µl of the supernatant containing 10 µM monomeric actin were taken carefully for actin assembly assays.

In polymerization experiments, monomeric pyrene-actin was used at a final concentration of 2 µM. After a 3 min incubation step with ME buffer (1 mM MgCl₂, 1 mM EGTA) to exchange actin bound Ca²⁺ against Mg²⁺, proteins of interest in buffer B were added and reactions were started with 10× KMEI. 100 µl of these mixtures were quickly applied to an UV cuvette and pyrene fluorescence (excited at 365 nm, emitted at 407 nm) was recorded.

For actin filament elongation assays, G-Mg buffer (G buffer with MgCl₂ instead of CaCl₂), F-actin, buffer B with or without protein, and 10× KMEI were mixed. The solution was sheared by passaging five times through a 27-gauge needle in order to generate actin seeds. An invariable volume of this solution was mixed with pre-incubated pyrenyl-actin and ME. 100 µl of this mixture were transferred into a fluorescence cuvette. Pyrene fluorescence was monitored immediately. Final concentrations of actin and actin seeds used in these assays were 0.5 µM and 333 nM, respectively.

In F-actin dilution assays, 10 µM actin (comprised with 5% pyrenyl-actin) was polymerized by addition of 1× KMEI for 1 h at room temperature. F-actin was 1∶10 diluted into G buffer with 1× KMEI in absence and presence of FHOD1 fl and FHOD1 ΔC. The decay in fluorescence was monitored for 10 min.

For TIRFM, coverslips were gently cleaned with ethanol, Milli-Q water, Hellmanex II (Fluka, NO. 61257), and rinsed in Milli-Q water. Clean coverslips were subsequently incubated with PEG-silane (MPEG-SIL-5000, 0.1 mg/ml in 96% Ethanol) for 12 hours at room temperature. PEGylated coverslips were then rinsed in Milli-Q water, air dried and stored at 4°C until use. Polymerization experiments were carried out in a Leica microscope (DMI 6000B) using 10 to 15% labeled G-actin (A488 rabbit muscle actin, Invitrogen). Images were taken every 5 seconds using an EMCCD camera (Hamamatsu C9100-02), with the excitation laser being shuttered between images to avoid photo-bleaching. Polymerization was started by the addition of 2× TIRF polymerization buffer (20 mM imidazole, 100 mM KCl, 2 mM MgCl₂, 2 mM EGTA, 0.4 mM ATP, 100 mM DTT, 2% methylcellulose, 10 µg/ml glucose oxidase, 50 µg/ml catalase, 250 µg/ml glucose, pH 7.2–7.4). After the addition of the TIRF polymerization buffer the sample was incubated 2 minutes on ice before it was loaded onto the coverslip.

Actin was isolated from acetone-dried powder prepared from rabbit skeletal muscle as described (Mannherz et al., 2007). Actin was first depolymerized at 5 µM on ice for 1 h in G buffer (5 mM Tris/HCl, pH 8.0 and 0.2 mM CaCl₂, 0.2 mM ATP, 0.5 mM DTT). After a centrifugation step (14,000 rpm, 30 min) actin was polymerized by addition of 10× KMEI (500 mM KCl, 20 mM MgCl₂, 10 mM EGTA, 100 mM imidazole) for 2 h at 23°C. 5 µM phalloidin (Sigma) was added to the assay in order to stabilize F-actin. F-actin was added to the solutions containing proteins of interest using cut 200 µl tips and incubated 30 min at 23°C. After that, reaction mixtures were centrifuged at 80,000 g at 4°C for 20 min in a Beckman Optima TL-100 bench top ultracentrifuge using a TLA-100 rotor. 80 µl of the supernatants were taken and dried in a SpeedVac. The remaining supernatant was removed and pellets were washed briefly with 100 µl polymerization buffer. To both supernatants and pellets, 20 µl of 1× SDS sample buffer was added. 10 µl of the samples were applied to and analyzed by Coomassie-stained SDS-PAGE.

F-actin bundling assays were performed as described in the actin binding section. The only difference affected the centrifugation step after mixing of actin and probed proteins. Here, samples were applied to 10,000 g for 20 min at 4°C using a bench top centrifuge. In addition, F-actin bundling was probed by chemical cross-linking using p-NN′-phenylene-bis-maleimide (PBM) as described (Knight and Offer, 1978) and confirmed by SDS-PAGE analysis performed as given (Hesterkamp et al., 1993). The PBM-generated ‘lower dimer’ obtained by cross-linking F-actin bundles obtained in the presence of 50 mM MgCl₂ was purified by gel filtration as described (Silván et al., 2012).

For EM analysis actin was polymerized by addition of 2 mM MgCl₂ and diluted to 0.1 mg/ml (2.4 µM) in polymerization buffer. Depending on the experimental conditions gelsolin and tropomyosin were added at molar ratios of 1∶100 and 1∶7, respectively and incubated for 15 min at RT. Additionally, either FHOD1 ΔC or full length FHOD1 at a concentration of 240 nM or 24 nM were added.

After 5 h or overnight incubation on ice, samples were prepared for electron microscopy similarly as described (Ohi et al., 2004). Briefly, a 3 µl drop of sample solution was adsorbed for 45 sec to a freshly glow-discharged copper grid (AGAR Scientific) covered by a thin, continuous carbon film. Excess sample was blotted using filter paper (Whatman No.4), washed twice with sample buffer and stained for 30 sec with 0.75% uranyl formate. Specimens were imaged at room temperature using a JEM1400 electron microscope (JEOL) equipped with a LaB₆ electron source. The microscope was operated at an acceleration voltage of 120 kV. Images were taken at a calibrated magnification of either 6,660× or 53,800× and recorded on a 4k×4k TemCam-F416 CMOS camera (TVIPS).

COS7 cells (ATCC, Teddington, UK) and Neuro-2a cells (DMSZ, Heidelberg) were cultured using standard conditions. For microscopy, 50,000 cells were plated on individual glass bottom dishes (MatTek Corp., Ashland) and transfected using Fugene-6 (Roche, Mannheim) on the following day according to the manufacturer's protocol. The following day, cells were imaged via TIRF microscopy in CO₂-independent growth medium containing serum and HEPES (Pan Biotech, Aidenbach) at 35°C in a home-built incubation chamber. For TIRF microscopy, an Olympus IX-81 microscope, equipped with a Plan APO 60× oil immersion TIRF microscopy objective (NA = 1.45) and 1.6 optovar, was used. A triple bandpass dichroic mirror (U-M3TIR405/488/561, Olympus, Hamburg) was combined with Semrock Brightline emission filters (HC 520/35 and HC 629/53, AHF Analysentechnik, Tübingen) and CellR diode lasers 488 (20 mW), 561 (25 mW). For detection, an EMCCD camera (C9100-13, Hamamatsu, Herrsching am Ammersee) was used at maximal gain. All microscope components were controlled by the CellR software (Olympus, Hamburg). For speckle microscopy, cells that express very low amounts of the FHOD1 constructs were selected. In these conditions, exposure times of at least 800 ms were required to visualize individual speckles. Image processing was performed using ImageJ (NIH, Bethesda) and Photoshop (Adobe, München). Image manipulations were limited to cropping, scaling, rotation, adjustment of levels, and addition of clearly identifiable labels or symbols. To correct for lateral drift in the stage position, an image stabilizer plugin for ImageJ was used (K. Li, ‘The image stabilizer plugin for ImageJ’, http://www.cs.cmu.edu/~kangli/code/Image_Stabilizer.html, February 2008). The intensity levels in all image panels in time series are adjusted identically.

We thank Karin Vogel-Bachmayr and Diana Ludwig for expert technical assistance and Sascha Gentz for synthesis of the DAD peptide. In addition, we thank Naoki Watanabe (Tohoku University) for supplying the ΔCMV expression plasmid.