Formin 2 regulates the stabilization of filopodial tip adhesions in growth cones and affects neuronal outgrowth and pathfinding in vivo

Establishment of accurate neuronal connectivity during development is central to nervous system function. Growth cones are dynamic chemosensory structures at the tips of growing axons and dendrites that guide these processes to their synaptic targets. Rod-like, actin-based filopodia on growth cones serve as antennae for guidance cues, which trigger local remodeling of the cytoskeleton to achieve directional motility (Bentley and Toroian-Raymond, 1986; Gomez et al., 2001; Heckman and Plummer, 2013). Additionally, filopodial contractility generates the traction forces necessary for growth cone translocation (Chan and Odde, 2008; Dennerll et al., 1989). Persistent elongation of the filopodia and their ability to generate traction depend on the formation of short-lived, integrin-based point contacts along their length and at the tip. These adhesion sites act as a molecular clutch, transiently anchoring the rearward actin flow to the integrin receptors engaged with the extracellular matrix (ECM) (Bornschlögl, 2013; Mallavarapu and Mitchison, 1999; Mitchison and Kirschner, 1988). Clutch engagement slows down the retrograde flow, allowing cytoskeletal forces to be transduced onto the substrate as traction. In addition, unbalanced actin polymerization at the tip increases membrane protrusion. Recently, a second mechanism of filopodial force generation has been described involving helical buckling of filopodial actin bundles pulling on the substrate engaged to the filopodial tip (Leijnse et al., 2015). These studies highlight the importance of filopodial tip adhesions, although less is known about the mechanisms underlying adhesion dynamics and force transduction through the filopodial shaft.

In neuronal growth cones, integrins have been shown to cluster at filopodial tips upon induction with NGF (Grabham and Goldberg, 1997). Integrin-dependent signaling, analogous to that seen in non-neuronal cells, has been reported in a few studies (Gomez and Letourneau, 1994; Gomez et al., 1996; Myers and Gomez, 2011) and mechanochemical signaling through focal adhesion kinase (FAK) has been shown to stabilize the early adhesions (Moore et al., 2009, 2012; Robles and Gomez, 2006).

Multiple actin remodeling activities regulating filopodial initiation and elongation have been described (Gupton and Gertler, 2007; Lundquist, 2009; Mattila and Lappalainen, 2008). Several members of the formin family have been implicated in filopodia formation, including the Diaphanous-related formins (Block et al., 2008; Goh and Ahmed, 2012; Schirenbeck et al., 2005), DAAM (Jaiswal et al., 2013) and FMNL2 (Block et al., 2012). In the growth cone, the role of Dia2 (Diaph2) and DAAM has been highlighted (Dent et al., 2007).

Formin 2 (Fmn2) was originally described as enriched in neuronal tissue (Leader and Leder, 2000), but little is known about its function in the nervous system. Recently, mutations in FMN2 have been associated with human intellectual and cognitive disabilities, and correlated with reduced dendritic spines and synaptic density in mice (Law et al., 2014; Mozhui et al., 2008). Mammalian Fmn2 mediates spindle positioning and meiosis during oocyte development by generating contractile actin networks (Leader et al., 2002; Schuh and Ellenberg, 2008). This activity is conserved in the Drosophila ortholog of Fmn2 (Rosales-Nieves et al., 2006).

In this study, we find in chick that Fmn2 regulates growth cone translocation by mediating filopodial mechanotransduction and tip adhesion stabilization. In mouse fibroblasts, Fmn2 maintains focal adhesion stability via an analogous function in the ventral stress fibers. We show that Fmn2 is necessary for pathfinding by spinal commissural neurons, underscoring the importance of Fmn2 in the development of neural circuits.

We evaluated the differential expression of Fmn2 mRNA using quantitative real-time PCR (qRT-PCR) in the developing chick embryo. Fmn2 transcript was enriched in the central nervous system, with the highest expression in the developing spinal cord (Fig. 1A). Monitoring Fmn2 transcript levels in whole embryos revealed persistent expression through developmental stages HH14 to HH22 (Fig. 1B). This period coincides with the phase of outgrowth and floor plate crossing by the spinal commissural neurons (Yaginuma et al., 1994).

To assess the intracellular distribution of Fmn2 in neurons, we raised a polyclonal antibody against chick Fmn2 (anti-gFmn2). The specificity of the antibody was confirmed by western blot on lysates from control and two independent Fmn2-specific translation-blocking morpholino-transfected spinal cord tissue (Fig. 1C, Fig. S1A,A′). In immunoblots of spinal lysates, anti-gFmn2 antibody detected a single band at the expected size, which was significantly reduced in lysates from embryos electroporated with either of the two Fmn2-specific morpholinos.

As Fmn2 mRNA showed sustained expression in the developing spinal cord, we analyzed the distribution of endogenous Fmn2 protein by immunofluorescence in spinal neurons from HH26/27 chick embryos cultured on fibronectin-coated cover glass. Fmn2 was seen to preferentially localize to the actin bundles, which extended into the growth cone lamellipodia (Fig. 1D-D″, Fig. S1B-C″). Occasionally, Fmn2 was detected at the filopodial tip, perhaps indicative of a role in the formation or elongation of filopodia (Fig. 1D-D″). Growth cones transfected with Fmn2-specific morpholinos showed an overall reduction in endogenous Fmn2 protein levels (Fig. 1E-E″, Fig. S1D). Additional evidence for antibody specificity came from pre-adsorption of the anti-gFmn2 antibody with the epitope peptide prior to immunofluorescence studies. Peptide pre-adsorption resulted in a strong reduction in the overall staining with no signal detected on filopodial actin bundles (Fig. S1D).

Further, we evaluated the localization of transfected GFP-tagged mouse Fmn2 (mFmn2-GFP) in spinal neurons. To reduce the cytosolic signal resulting from overexpression, growth cones were permeabilized prior to fixation. Similar to the distribution of endogenous chick Fmn2, mFmn2-GFP was detected along actin bundles in filopodia and their extended roots. mFmn2-GFP also localized to the tips of some filopodia (Fig. 1F-F″). By contrast, overexpression of soluble GFP showed a diffuse distribution within the central region of the growth cone and was not enriched in filopodial actin bundles (Fig. S2).

Thus, using two complementary approaches – antibody-based detection of endogenous chick Fmn2 and expression of the mouse ortholog of Fmn2 – we show that Fmn2 preferentially localizes to actin bundles in growth cone filopodia.

We analyzed Fmn2 morphant spinal growth cones from HH26/27 chick embryos to evaluate the effect of Fmn2 depletion on growth cone morphology (Fig. 2A,B). The overall growth cone area was reduced in the Fmn2 morphants compared with the control treatment (P=0.0019; Fig. 2C).

Formins are known regulators of actin filament nucleation, elongation and bundling (Bor et al., 2012; Chesarone and Goode, 2009; Rosales-Nieves et al., 2006; Schuh, 2011). The Drosophila ortholog of Fmn2, Cappuccino, is also reported to have actin nucleating and bundling activities (Vizcarra et al., 2014). In order to assess actin organization directly in the chick growth cone, we evaluated actin compaction using homoFRET between phalloidin-Alexa 488 molecules labeling F-actin. homoFRET is measured by analyzing the extent of depolarization of the fluorescence emission after excitation with polarized light. Fluorescence anisotropy is used as a measure of fluorescence depolarization (Chan et al., 2011; Ganguly et al., 2011; Iyer et al., 2012; Tramier and Coppey-Moisan, 2008). Anisotropy measurements were performed by collecting the parallel and perpendicular emissions from the phalloidin-stained growth cones. The anisotropy in the system was calculated using standard equations (see Materials and Methods). A compact arrangement of fluorophores results in higher homoFRET and, consequently, reduced anisotropy, whereas higher anisotropy indicates low homoFRET resulting from a less compact organization of fluorophores. Knockdown of Fmn2 resulted in an increase in fluorescence anisotropy compared with controls (P=0.014), indicating reduced homoFRET and therefore decreased actin compaction in these growth cones (Fig. 2D-F).

Live imaging revealed impaired growth cone motility in Fmn2 morphants compared with controls (Fig. 2G,G′, Movie 1). Fmn2 knockdown reduced growth cone motility speeds (P=0.0098) and displacements (P=0.0011) in comparison to the controls (Fig. 2H,I). In addition, Fmn2 morphant growth cones covered significantly less distance over time (P=0.013; Fig. S1F). The ratio of the displacement to the distance covered by the growth cone was used as a measure of the persistent directionality for a given growth cone. Compared with the controls, Fmn2 morphants showed impaired persistent directional translocation (P=0.0006; Fig. 2J).

Taken together, our analysis reveals a function for Fmn2 in organizing the growth cone actin cytoskeleton and in the regulation of growth cone structure and motility.

As Fmn2 localized to actin bundles in the filopodia, we assessed the effects of Fmn2 depletion on filopodia. Both filopodia numbers per growth cone (P=0.0022) and filopodial lengths (P=0.0011) decreased upon Fmn2 depletion (Fig. 3A,B). Live, time-lapse imaging showed that both the number of filopodia initiation events per growth cone and the elongation rates were unaltered upon Fmn2 knockdown (Fig. 3C,D, Fig. S4). Filopodial persistence was evaluated by measuring filopodial lifetimes. Depletion of Fmn2 resulted in reduced filopodial lifetimes (P=0.02) and an increase in the number of short-lived filopodia (Fig. 3E,F).

To validate a role for Fmn2 at the growth cone, we introduced the morpholino-resistant mouse ortholog of Fmn2 in the morphant chick neurons. This successfully rescued the growth cone area, filopodial numbers and filopodial lengths in the Fmn2 morphant growth cones (Fig. 3G-I).

These experiments implicate Fmn2 function in growth cone cytoarchitecture and in filopodia dynamics by influencing filopodial stability.

The formation of substrate attachments is central to filopodial stabilization. We tested the ability of growth cones to develop point contacts along the filopodia using autophosphorylation of focal adhesion kinase at Y397 (pFAK) as a readout of integrin-dependent development of substrate attachments. Knockdown of Fmn2 resulted in an overall reduction of the pFAK signal at the growth cone (P=0.0031; Fig. S3F,G). The distribution of substrate adhesions was analyzed by pFAK fluorescence intensity line scans along the filopodia. Control morpholino-treated filopodia showed multiple point contacts along the length, with a distinct pFAK intensity peak at the distal tip (Fig. 4A′ inset, C). However, Fmn2 morphant filopodia lacked this tip accumulation (Fig. 4B′ inset, C). In order to directly compare pFAK distribution across control and Fmn2 knockdown filopodia, which have intrinsic length differences, we normalized the pFAK signal in the distal segment of the filopodia to the total pFAK signal in the filopodia (distal and proximal segments). This analysis revealed that Fmn2 depletion resulted in a considerable decrease in the pFAK signal from the distal segment of the filopodia (P=0.02; Fig. 4D). These data reveal Fmn2 function in influencing filopodial dynamics via stabilization of tip adhesions.

The importance of tip adhesions has been highlighted in recent studies on filopodial traction forces (Bornschlögl et al., 2013; Chan and Odde, 2008; Hoffmann and Schäfer, 2010; Romero et al., 2013). Our observation of reduced pFAK signal in Fmn2 morphant filopodial tips prompted us to investigate filopodial traction. Growth cones were cultured on compliant gels embedded with fluorescent beads (Fig. 5A, Movie 2), and the displacement of individual beads at the filopodial tips (Fig. 5A′-A‴) was used as an indirect measure of forces exerted by the filopodia on the substrate (Chan and Odde, 2008). Fmn2 knockdown resulted in a reduced ability to generate traction on the underlying substrate, as seen by the smaller displacement of the beads in comparison to the control filopodia (P<0.001; Fig. 5C). Although the frequency of pulling by individual filopodia did not differ significantly (Fig. 5B), the velocity of the pulling events was attenuated upon Fmn2 depletion (P<0.001; Fig. 5D). These data implicate filopodial Fmn2 activity as being necessary for point contact stabilization and the generation of traction on the substrate.

Our data from chick neuronal filopodia prompted us to investigate whether Fmn2 has a conserved function in stabilizing focal adhesions (FAs) in fibroblasts. qRT-PCR analysis suggested that siRNA treatment could deplete the endogenous Fmn2 transcript in mouse NIH3T3 fibroblasts (Fig. S5A). Consistent with our observation in chick growth cones, depletion of Fmn2 reduced the pFAK (Y397) intensity at the FAs (P=0.0047; Fig. S5B).

We monitored paxillin-labeled FAs using live total internal reflection fluorescence (TIRF) microscopy to evaluate the effect of Fmn2 knockdown on FA dynamics in NIH3T3 cells (Fig. 6A,B). FA parameters were extracted using an automated image analysis platform (Berginski et al., 2011). Compared with the control cells, the Fmn2 siRNA-treated cells showed reduced FA size (P<0.0001), although the number of FAs per cell was not altered significantly (Fig. 6C,D). Further analysis revealed that although Fmn2-deficient cells showed comparable FA assembly rates, they disassembled more quickly (P=0.0002; Fig. 6E,G). In addition, the duration of the disassembly phase was increased (P<0.0001; Fig. 6H) and the assembly phase was reduced following Fmn2 knockdown (P=0.0071; Fig. 6F).

Taken together, these data show that FAs in Fmn2-depleted cells show increased disassembly and are consequently unstable and smaller.

Expression of mFmn2-GFP in NIH3T3 cells showed a heterogeneous distribution. mFmn2-GFP was localized to irregularly shaped, often fibrillar structures closely apposed to the ventral surface (Fig. 7A,B). Typically, these structures were subnuclear and perinuclear and were largely devoid of actin. Interestingly, Fmn2 was also found to localize along ventral actin stress fibers that span two FAs (Fig. 7D-F′). The Fmn2-labeled ventral stress fibers were not only perinuclear, forming a cage-like network at the base of the nucleus, but also extended towards the periphery. Fmn2 was unevenly distributed along the length of these ventral stress fibers (Fig. 7D-H). Fmn2 on stress fibers was restricted to the juxta-FA actin bundle but was largely undetected in the central region of the paxillin-labeled FAs (Fig. 7C).

To test if Fmn2 activity influences actin dynamics at the ventral stress fibers, we employed fluorescence recovery after photobleaching (FRAP) analysis on NIH3T3 cells co-transfected with siRNAs and mGFP-Actin (Fig. 7I). Analysis of the FRAP curves revealed longer recovery times (P<0.0001) in Fmn2-deficient cells (fast t_1/2=55.84±8.49 s; slow t_1/2=4.25±3.27 s) compared with the control cells (fast t_1/2=80.68±26.21 s; slow t_1/2=8.47±5.04 s). This suggests that Fmn2 activity is required for efficient actin monomer exchange in the ventral stress fibers. Mechanotransduction along stress fibers has been reported to influence FA stability (Burridge and Wittchen, 2013; Oakes et al., 2012; Walcott and Sun, 2010) and the defects in FA dynamics observed upon Fmn2 knockdown are likely to be a consequence of impaired ventral stress fiber assembly.

As our in vitro data implicated Fmn2 function in growth cone dynamics of spinal neurons, we investigated its in vivo relevance by depleting Fmn2 from developing chick spinal cords. Midline crossing of spinal commissural interneurons was evaluated following in ovo electroporation of control and Fmn2 translation-blocking morpholinos.

Outgrowth and floor plate crossing of spinal commissural neurons occurs between HH19 and HH26 (Yaginuma et al., 1994). Initial experiments involved unilateral in ovo electroporation of morpholinos at HH14 followed by immunostaining of HH26 cryosections with axonin 1 antibodies. Axonin 1 is expressed on the axons of pre-crossing spinal commissurals but its expression disappears following floor plate crossing (Baeriswyl and Stoeckli, 2008; Fitzli et al., 2000; Stoeckli and Landmesser, 1995). At HH26, no axonin 1 signal was detected in the spinal cords of control morpholino-treated embryos. However, all stage-matched Fmn2 morphant embryos showed sustained axonin 1 staining only in the electroporated half of the spinal cord (Fig. S6A-D′; three embryos in each group).

Encouraged by these results we investigated the role of Fmn2 in spinal circuit development by evaluating the Atoh1-expressing dI1 subpopulation of spinal commissural interneurons. Morpholinos were unilaterally co-electroporated along with Atoh1-tau-mCherry at HH14 and stage-matched spinal cords were analyzed in open-book preparations at HH26 to visualize the dI1 commissural neurons. In control embryos, the majority of the commissural neurons had crossed the floor plate and taken a rostral turn along a stereotyped sigmoidal trajectory (Fig. 8A-A‴). However, the Fmn2 morphant embryos showed compromised midline crossing, with many dI1 tracts stalling at the floor plate (Fig. 8B-B‴). This analysis showed that depletion of Fmn2 reduces the ability of the commissural neurons to efficiently cross the floor plate as compared with the control morpholino-treated embryos (Fig. 8D,E). We further analyzed the contralateral sigmoidal trajectories. Compared with control embryos, Fmn2 morphants showed a reduction in the proportion of post-crossing axonal tracts displaying the stereotyped sigmoidal trajectory (Fig. 8F). To test for a cell-autonomous requirement for Fmn2 in dI1 neuron pathfinding, we expressed the mouse ortholog of Fmn2 specifically in these commissural neurons using the Atoh1 enhancer (Atoh1-mFmn2). The chick-specific morpholino used to deplete endogenous Fmn2 had no binding sites on the mouse Fmn2 cDNA. Co-electroporation of Atoh1-mFmn2 and control morpholino had no discernible effect on the dI1 neurons. However, expression of Atoh1-mFmn2 in dI1 neurons rescued both the midline crossing and contralateral trajectory phenotypes of the Fmn2 morphants (Fig. 8C-F).

Taken together, our analysis highlights the cell-autonomous requirement of Fmn2 in the development of dI1 spinal commissural neuron circuits and emphasizes the evolutionary conservation of Fmn2 function in avian and mammalian lineages.

Filopodia are central to the establishment of neuronal circuits during development. A number of proteins have been identified in recent years, including members of the formin family, as being necessary for the initiation and elongation of filopodia. However, less is known about activities that regulate the stability and maturation of the filopodial tip adhesions and, in turn, the generation of traction forces. We identify the Rho-independent formin Fmn2 as a crucial regulator of filopodial adhesion dynamics and growth cone motility. Fmn2 function in vivo was tested in chick dI1 spinal commissural neurons and was found to be necessary for outgrowth and pathfinding.

Fmn2 localized to filopodial actin bundles in spinal neurons. Knockdown of Fmn2 reduced the number and length of filopodia. Unlike other formins (Block et al., 2008; Goh and Ahmed, 2012; Mellor, 2010), Fmn2 depletion did not affect filopodial elongation rates or the number of initiation events. However, filopodial lifetimes were severely reduced, suggesting compromised filopodial stability. Although it is possible that Fmn2 null neurons may reveal defects in filopodia initiation and elongation not observed in transient knockdowns, our results support a dominant and novel function for Fmn2 in filopodial stabilization via tip adhesions.

The persistence of protrusive structures and their function in motility depend on the formation of stable substrate attachments. FAK autophosphorylation at Y397 is central to assembly and mechanotransduction at integrin-based adhesion sites. In growth cones, FAK activity mediates filopodial dynamics, development of adhesions and traction forces and pathfinding (Chacón et al., 2012; Moore et al., 2012; Robles and Gomez, 2006). Fmn2 depletion reduced pFAK (Y397) at the filopodia tips, suggesting impaired adhesive stability, in concurrence with the decreased filopodial lifetimes observed.

Filopodial tip adhesions regulate filopodial contractility and the generation of traction forces (Moore et al., 2009, 2012; Romero et al., 2013). We find that Fmn2-depleted filopodia are severely compromised in their ability to generate traction forces, possibly owing to impaired engagement of the cytoskeleton with the ECM. Collectively, these data agree with the growth cone translocation defects observed in Fmn2 morphant neurons.

Forces originating from F-actin retrograde flow are transduced onto the substrate by dynamic molecular clutches that link the actin to the extracellular substrate (Bornschlögl et al., 2013; Chan and Odde, 2008; Mallavarapu and Mitchison, 1999). Additionally, helical buckling-induced shortening of the filopodial actin shaft can also contribute to filopodial traction (Leijnse et al., 2015). We propose that Fmn2 knockdown results in disorganized actin bundles in filopodia, impairing its ability to transduce forces to the tip adhesions (Fig. 8G). Reduced actin compaction in Fmn2-depleted growth cones provides early evidence for such a function. Several formins, including Fmn2 and its fly ortholog, have been suggested to have actin bundling functions (Bohnert et al., 2013; Esue et al., 2008; Harris et al., 2006; Machaidze et al., 2010; Schönichen et al., 2013; Vizcarra et al., 2014). Thus, it seems feasible that Fmn2 maintains the structural integrity of filopodial actin bundles, in turn influencing the efficiency of clutch engagement and force transduction.

Anti-parallel acto-myosin stress fibers transduce mechanical forces between the cytoskeleton network and FAs in non-neuronal cells. We found that Fmn2 regulates ventral stress fiber turnover and FA disassembly. Stress fiber dynamics is known to influence FA stability (Burridge and Wittchen, 2013; Chang and Kumar, 2013; Colombelli et al., 2009; Oakes et al., 2012) and these observations are therefore analogous and supportive of our conclusions regarding Fmn2 function in growth cone filopodia. Fmn2 is detected in the myosin-dependent FA proteome (Kuo et al., 2011), and its juxta-FA localization might suggest a function in coupling actin bundles to the adhesion complex, as proposed for INF2 (Skau et al., 2015). Additionally, the cage-like organization of Fmn2 stress fibers around the nucleus might suggest a role in nuclear positioning.

Fmn2 was found to be necessary for commissural neuron outgrowth and pathfinding. This function appears to be conserved across mammalian and avian lineages. Stalling of dI1 axons at the floor plate and aberrant axonin 1 expression upon Fmn2 depletion are consistent with growth cone translocation deficits identified in this study and may represent a delay in axonal outgrowth. The defects in the contralateral side could be an indirect consequence of the delayed arrival of growth cones at the relevant choice points. An additional, independent function of Fmn2 in guidance cue-dependent directional motility is also possible, but elucidation of this role will require an understanding of the upstream regulation of Fmn2 activity.

Our study investigated Fmn2 function in a specific spinal subpopulation, and it remains to be tested whether Fmn2 is a regulator of growth cone dynamics in actively extending neurons in general. In fact, our observation of Fmn2 function in ventral stress fibers and associated FAs in fibroblasts might suggest a broader role in cellular morphogenesis and motility that is not limited to neurons.

Recently, mutations in Fmn2 have been associated with human intellectual disabilities and with defective fear conditioning and reduced dendritic spine density in mice (Law et al., 2014; Mozhui et al., 2008; Peleg et al., 2010). Fmn2 mutations are also associated with increased metastatic potential (Charfi et al., 2011; Lynch et al., 2013). The actin organizing and adhesion stabilization functions of Fmn2 identified in our study are likely to offer mechanistic insights to these observations.

This study is the first to characterize Fmn2 function in neurons and identifies an unexpected role in stabilizing adhesive structures via actin bundle assembly, disruption of which leads to impaired development of spinal circuits. Together, these findings highlight the conserved role of Fmn2 in neuronal development and provide a novel mechanistic framework for formin function in cell motility.

Freshly fertilized, White Leghorn eggs were procured from Dr BV Rao Institute of Poultry Management and Technology, Pune, India. All animal protocols were in accordance with the procedures approved by the Institutional Animal Ethics Committee, IISER Pune, India. Embryos at stage HH14 were used for electroporation. Spinal cord electroporations were performed similarly to Baeriswyl et al. (2008). See the supplementary Materials and Methods for electroporation details.

Two translation-blocking morpholino sequences were tested (5′-3′): Fmn2 MO#1, CACGGTTTTTATTGCCTGCATTGCA; and Fmn2 MO#2, CCATCTTGATTCCCCATGATTTTTC. Standard control morpholino CCTCTTACCTCAGTTACAATTTATA was used as a negative control. Atoh1-tau-mCherry was a generous gift from Prof. Jane Johnson (UT Southwestern Medical Center). The original clone for mouse Fmn2 was obtained from Prof. Philip Leder (Harvard Medical School). See the supplementary Materials and Methods for construction of Atoh1-mFmn2 and pCAG-mFmn2-GFP.

Tissues were pooled from three chick embryos and a standard total RNA isolation procedure was followed using Trizol (Invitrogen). cDNA was prepared using 2 μg total RNA, oligo-dT (18mer) and M-MLV reverse transcriptase (Promega) according to the manufacturer's recommendations. For whole-embryo qRT-PCR analysis, RNA from three chick embryos was pooled. Each sample had three technical replicates and each experiment was repeated three times. The pooled data were used for statistical analysis and graphical representations. Fmn2 mRNA was normalized to β-actin and represented using brain expression as the reference.

Primers used (forward and reverse): Fmn2, TCAGCAGCGGATTCTGAGGCTAAA and ATGCAAGTCCTCTTGACTGGCTGA; β-actin, AGACATCAGGGTGTGATGGTTGGT and TCCCAGTTGGTGACAATACCGTGT. For qRT-PCR analysis of NIH3T3 cells, including sequences of mouse Fmn2 and β-actin primers, see the supplementary Materials and Methods.

A unique peptide sequence for chick Fmn2 (CRQKKGKSLYNIRPK) was selected for antibody generation in rabbits. The peptide was synthesized and injected into rabbit for antibody generation. Antisera (anti-gFmn2) were collected periodically after booster doses and the antibody titer was estimated by ELISA. See the supplementary Materials and Methods for antibody validation.

Spinal cords from HH26/27 chick embryos were dissected in PBS and trypsinized using 0.05% trypsin, 0.48 mM EDTA (Gibco). The cell suspension was transferred to an electroporation cuvette with appropriate plasmid and morpholino combinations; 10 μg plasmid and 100 μM morpholino were used for transfections. A NEPA21 type II square wave electroporator (Nepagene) was used. Typical transfection efficiency in these experiments was in the range 25-30%. Post-electroporation immunostaining, imaging and quantification of signal, including the analysis of pFAK along filopodia, were undertaken using standard procedures as detailed in the supplementary Materials and Methods.

Twenty-four hours after electroporation and culturing, the culture medium was replaced with fresh medium. The samples were imaged using a PlanApoN 60×/1.49 objective on an Olympus IX81 system equipped with a Hamamatsu ORCA-R2 CCD camera. Images were captured every 10 s for up to 30 min. The system was equipped with a focus drift correction mechanism and the entire imaging setup was maintained at 37°C without CO₂ during the entire imaging procedure. See the supplementary Materials and Methods for details.

Filopodial traction force assays were modified from an earlier protocol (Chan and Odde, 2008), as detailed in the supplementary Materials and Methods.

NIH3T3 fibroblasts were obtained from Dr N. Balasubramanian (IISER Pune) and were tested for contamination prior to use. The cells were maintained using standard culturing procedures in DMEM plus 10% heat-inactivated FBS, 1× Pen Strep (Gibco) and 1× L-glutamine (Gibco, 2 mM). Approximately 2×10⁴ cells were seeded in 50 ng/ml fibronectin-coated 4-well cover glass-bottom Nunc Lab-Tek chambers and allowed to attach overnight. siRNA (lacZ sense strand, 5′-CGUCGACGGAAUACUUCGAUU-3′; Fmn2 sense strand, 5′-UGGUUAGACUUGUGGGUAAUU-3′), paxillin-EGFP (a gift from Rick Horwitz, University of Virginia; Addgene plasmid #15233), paxillin-mCherry (a gift from Martin Schwartz, Yale University), Actin-mRFP (a gift from Bidisha Sinha, IISER Kolkata) and mGFP-Actin (a gift from Ryohei Yasuda, Max Planck Florida Institute of Neuroscience; Addgene plasmid #21948) were transfected using Lipofectamine 2000 according to the manufacturer's protocol (Invitrogen). The cells were incubated for 24 h before imaging.

Twenty-four hours post siRNA plus paxillin-EGFP transfections, the culture medium was replaced with cell imaging medium (L-15 plus 10% heat-inactivated FBS, 1× Pen Strep, 2 mM L-glutamine). The cells were imaged in TIRF mode using an ApoN/TIRF 60×/1.49 objective on an Olympus IX81 system equipped with a Hamamatsu ORCA-R2 CCD camera and the Olympus Cell TIRF module. Images were acquired every 10 s for 30 min. The system was equipped with focus drift correction and the entire imaging setup was maintained at 37°C without CO₂ during the entire imaging procedure. See the supplementary Materials and Methods for details. FAs from ten cells in each treatment representing three biological replicates were used for graphical representation and statistical analysis.

Post-transfection, the cells were fixed and stained with 20 units of phalloidin (Invitrogen) or live imaged using standard protocols. Twenty-four hours post siRNA plus mGFP-Actin transfection, ventral stress fibers of NIH3T3 cells were subjected to FRAP analysis on a Zeiss 710 inverted confocal microscope. Cells from three independent biological replicates were analyzed and the pooled data used for graphical representation and subsequent statistical analysis. See the supplementary Materials and Methods for details.

For both fixed and live imaged growth cones, finger-like projections from the growth cone periphery that were ≥2 μm in length were considered as filopodia (Fig. S4). All analysis was carried out using ImageJ (NIH). For details of the image analysis see the supplementary Materials and Methods. For fluorescence quantification, all imaging conditions were kept constant between the experimental sets being compared.

Following polarized excitation, the change in the fluorescence anisotropy measures the homoFRET efficiency. Under these conditions, increased homoFRET indicates higher order compaction as reported by a reduction in fluorescence anisotropy (Chan et al., 2011; Ganguly et al., 2011; Iyer et al., 2012; Tramier and Coppey-Moisan, 2008).

Morpholino-treated growth cones were fixed and stained with 20 units of phalloidin-Alexa Fluor 488 (Invitrogen) as above. The parallel and perpendicular emission components were collected using a PlanApo 63×/1.4 on a Zeiss LSM 710 confocal system equipped with an anisotropy module. Images were processed by applying the standard anisotropy equation: D=[(S1−G×S2)/(S1+2G×S2)]×255, where D is the anisotropy value, S1 and S2 are the parallel and perpendicular components of emission, respectively, and G is the grating factor for the imaging system which is used to correct any bias towards a particular emission component. To scale the output image to an 8-bit image a factor of 255 is incorporated in the equation. The grating factor was calculated using 1 mM FITC solution in water and was found to be 0.98. All the mathematical transformations were performed using 64-bit Zen 2012 Black edition (Zeiss). The anisotropy image so obtained was imported into ImageJ and a 2×2 median filter was applied to the entire image. Using the original phalloidin image, an outline of the growth cone was drawn using the polygon tool in ImageJ. This region of interest (ROI) was overlaid on the anisotropy image and the anisotropy values enclosed within this ROI were used for comparison across different treatments. Data were analyzed from three independent biological replicates representing 20 growth cones for each treatment. The ImageJ heatmap plugin was used to represent the output images.

Following morpholino electroporation (at HH14) and subsequent incubation, the embryos were harvested at HH26 and dissected as described previously (Avraham et al., 2010). The region of the spinal cord between the developing fore and hind limbs was used for analysis. The tissue was fixed in 4% formaldehyde for 30 min and mounted in 1× PBS for imaging. The entire depth of the sample was imaged as a z-stack using a PlanApo 20×/0.8 water-immersion objective on a Zeiss LSM 780 confocal system in the tiling mode. Analysis of spinal commissural axon tracts is described in the supplementary Materials and Methods. Each embryo was considered an independent biological replicate.

Raw images were processed as described in the supplementary Materials and Methods. All the raw measurements were imported to GraphPad Prism 5 for graphical representation. The box and whisker plots are represented using the Tukey method. The bottom and top of the box are the first and third quartiles, the whiskers span the lowest datum within the 1.5 interquartile range of the lower quartile and the highest datum within the 1.5 interquartile range of the upper quartile; the outliers are represented separately beyond the whiskers. The number of data points used for graphical representation and subsequent statistical analysis are represented below each treatment in the graphs. These numbers represent the pooled data from three or more independent biological replicates (exact numbers are indicated in the relevant subsections). All statistical comparisons for box plots were made using the Mann–Whitney test or Kruskal–Wallis test followed by Dunn's post-test in Prism 5. Bar graphs show mean±s.e.m. The multiple comparisons were made using one-way ANOVA followed by Dunnett's post-test.

We acknowledge the help of Dhriti Nagar in preparing the schematics and N. K. Subhedar for critical reading and helpful comments.