Drosophila twinfilin is required for cell migration and synaptic endocytosis

The dynamic actin cytoskeleton is essential for a wide range of cellular processes, including endocytosis, cytokinesis, axonal growth and motility. The dynamic nature of the actin structure derives from the fact that monomeric globular (G) actin subunits assemble at the barbed ends of filamentous (F) actin, while actin monomers dissociate from the pointed ends of F-actin (Pantaloni et al., 2001; Paavilainen et al., 2004). Under steady-state conditions, actin filaments undergo ‘treadmilling’, a process in which actin monomers depolymerized from one end of the filament are recycled and polymerized onto the other end, allowing directional growth of actin filaments without changing their length. This treadmilling process occurs much faster, allowing sustained actin filament growth, in such processes as lamellipodial motility during axonal growth and cell migration, and vesicle internalization during endocytosis. This dynamic treadmilling is regulated by a number of molecules, including highly conserved actin-monomer-binding proteins: profilin, actin-depolymerizing factor (ADF, also known as cofilin), cyclase-associated protein (CAP) and twinfilin. Profilin promotes actin polymerization, whereas the latter three proteins promote actin disassembly, inhibit actin polymerization or both (Palmgren et al., 2002; Paavilainen et al., 2004; Winder and Ayscough, 2005).

Twinfilin is a small protein that sequesters actin monomers. It harbors two ADF-H (ADF-homologous) domains. Like ADF/cofilin, twinfilin shows preferential binding to ADP–G-actin and forms a 1:1 complex with G-actin that prevents its assembly into actin filaments in vitro (Goode et al., 1998; Palmgren et al., 2001; Ojala et al., 2002). Twinfilin is highly abundant in yeast and mammals (Palmgren et al., 2001; Vartiainen et al., 2003); it localizes to subcellular regions of rapid actin turnover, such as cortical actin patches in yeast (Goode et al., 1998; Palmgren et al., 2001), lamellipodia in mammalian cells (Vartiainen et al., 2000; Vartiainen et al., 2003) and the ends of actin bundles in developing Drosophila bristles (Wahlström et al., 2001). The localization of yeast twinfilin to actin patches depends on both its actin-monomer-binding ability and the presence of barbed-end-capping proteins (Palmgren et al., 2001; Falck et al., 2004).

Recently, it has been reported that yeast twinfilin also interacts with actin filaments and promotes filament severing at low pH in vitro; its severing activity is inhibited by capping protein and phosphatidylinositol (4,5)-bisphosphate [PtdIns(4,5)P₂] (Moseley et al., 2006). Thus, twinfilin might coordinate filament severing and monomer sequestration at sites of rapid actin turnover (Moseley et al., 2006). In addition to F-actin severing, twinfilins from vertebrates, but not from lower organisms, were found to cap the barbed ends of F-actin, with preferential affinity for ADP-bound ends, thus inhibiting the elongation of actin filaments (Helfer et al., 2006; Paavilainen et al., 2007).

Deletion of twf in yeast results in mildly perturbed cortical actin patches with increased F-actin staining, defects in the bipolar bud-site selection pattern, and synthetic lethality with certain ADF/cofilin and profilin mutations (Goode et al., 1998; Wolven et al., 2000; Falck et al., 2004). By contrast, overexpression of twf in yeast leads to complete depolarization of the actin cytoskeleton and accumulation of cytoplasmic actin bars (Goode et al., 1998). Similarly, overexpression of mouse twinfilin gives rise to a decrease in stress fibers and formation of abnormal actin filament structures in cultured cells (Vartiainen et al., 2000). In summary, alterations of twinfilin expression result in abnormal F-actin structures. However, the physiological function of twinfilin in a multicellular organism remains unclear.

Wahlström et al. showed that the Drosophila genome encodes a single homologue of yeast and mammalian twinfilins (Wahlström et al., 2001). Phenotypic analysis of a twf hypomorphic allele revealed that twf is involved in actin-based bristle development (Wahlström et al., 2001). However, there are no reports of studies on twf-null mutants in metazoans. Here, we present the first systematic analysis of a twf-null mutant in a multicellular organism. Drosophila twf nulls are larval lethal with few adult escapers, indicating that it is critical for viability in flies. We provide evidence demonstrating that twinfilin negatively regulates F-actin formation in multiple cellular contexts. More importantly, we report for the first time that twinfilin is required for different cellular processes in neuronal and non-neuronal systems, including axonal growth in the brain, border cell (BC) migration in the ovary and presynaptic endocytosis at neuromuscular junction (NMJ) synapses. We suggest that the various phenotypes observed in disparate cell types or cellular contexts of twf mutants are caused by a shared underlying mechanism of disrupted actin dynamics.

EP(3)3701 inserts in the first intron of twf and is a strong hypomorphic allele of twf (Fig. 1) (Wahlström et al., 2001). This allele is the only twf mutation characterized to date in Drosophila. To further understand the in vivo functions of twf, we generated null twf mutants using P-element-mediated excision mutagenesis. We recovered 305 independent excision lines. The breakpoints were first identified by PCR and then pinpointed by DNA sequencing. Here, we report two imprecise deletions, 110 and 102, in which most of the twf coding region is deleted (Fig. 1A). In addition to twf, the gene encoding dAbi (the Drosophila Abelson interacting protein) lies next to the EP(3)3701 insertion (Fig. 1A). Using northern analysis, Wahlström et al. showed that the EP3701 insertion does not affect dAbi expression at the transcriptional level (Wahlström et al., 2001). To examine whether the twf deletions also affect dAbi expression, we performed western blotting to detect dAbi protein levels. Our results showed that dAbi protein levels remained unchanged in all twf mutant lines (Fig. 1B). Thus, the EP insertion, subsequently referred to as twf³⁷⁰¹, and the two imprecise excisions (twf¹¹⁰ and twf¹⁰²) do not affect the adjacent dAbi gene and are twf-specific mutant alleles. Ubiquitous expression of twf small interfering RNA (siRNA) and upstream activating sequence (UAS)-twf by act-Gal4 caused dramatically reduced and increased expression of twinfilin, respectively (Fig. 1B). More importantly, western blotting and immunostaining showed that twinfilin was not detected in the 110 and 102 deletion lines (Fig. 1B,I,J) indicating that these deletions are null alleles of twf.

The deletion mutants were larval lethal with very few adult escapers (Table 1). Lethality was rescued to different extents by the expression of twf in muscles by C57-Gal4, in neurons by pan-neuronal elav-Gal4 or ubiquitously by actin-Gal4 using the UAS-Gal4 system (Table 1), indicating that twinfilin is required for the normal development of different tissues, particularly the nervous system. Drosophila twinfilin is ubiquitously expressed in embryos and adult ovaries with high enrichment in BCs, and at various developmental stages (Wahlström et al., 2001). However, its expression in the nervous system has not been documented. Here, we report that twinfilin is enriched in the neuronal soma and processes in the larval and adult brain (supplementary material Fig. S1), as well as at peripheral NMJ synapses. Triple labeling of NMJ synapses with anti-twinfilin, the presynaptic membrane marker anti-horseradish peroxidase (anti-HRP) and the postsynaptic marker anti-Dlg showed that twinfilin mainly colocalized with postsynaptic Dlg (Fig. 1C-H). Twinfilin staining was hardly detectable at the postsynaptic site in twf-null mutants (Fig. 1I,J).

Precise regulation of actin dynamics is crucial for axonal growth. It has been reported that the actin-monomer-binding proteins ADF/cofilin and Ciboulot (encoded by cib), a homologue of β-thymosins with the biochemical properties of profilin, affect axonal growth in the Drosophila brain; both cib and ADF/cofilin-encoding tsr mutants show axonal growth defects (Boquet et al., 2000; Ng and Luo, 2004). To examine a possible role for twinfilin in axonal growth, we examined the neuronal morphology of the mushroom body (MB), for which neuronal development is well documented. Anti-Fas II strongly labels the α and β axonal lobes of the MB (Fig. 2). In wild-type flies, the β lobes terminate near the midline, but rarely grow across it (Fig. 2A). By contrast, 31.4% of the β lobes (n=51) in the twf³⁷⁰¹ hypomorph crossed the midline and, in severe cases (23.5%), the two β lobes were completely fused (Fig. 2B,C,G). twf¹⁰² null mutants showed a more severe midline-crossing phenotype (17.5% mild and 35.7% severe; n=40; Fig. 2D,G). The twf¹¹⁰ null allele showed phenotypes similar to those of twf³⁷⁰¹and twf¹⁰² (Fig. 2G). Thus, about 50% of the homozygous mutants of the three different alleles had apparent MB midline-crossing defects. The hemizygotes had a weaker but similar phenotype to the homozygotes, suggesting the presence of a background mutation on the mutated chromosomes or a suppressor on the deficiency chromosome (Fig. 2G). However, the midline-crossing phenotype was rescued, at least partially, by the expression of twf driven by MB-specific OK107-Gal4 in the twf³⁷⁰¹ background (Fig. 2G). Together, these data demonstrate that the axonal midline-crossing phenotype is caused specifically by twf mutations. The midline-crossing phenotype in twf mutants is similar to that reported for cib-overexpressing animals (Boquet et al., 2000).

In addition to the midline-crossing defect, other axonal abnormalities of MB were also observed. In the wild type, individual α/β neurons extend a medial β lobe and a dorsal α lobe from the peduncle (Fig. 2A). By contrast, twf¹⁰²-null mutants had a missing α lobe, accompanied by thickening of the β lobe (Fig. 2E,F). Either uni- or bi-lateral loss of α lobes occurred with a higher frequency in twf¹⁰² and twf¹¹⁰ null mutants than in hypomorphic twf³⁷⁰¹ (Fig. 2H). In twf¹⁰² mutants, 20% of the MBs (n=40) showed unilateral α-lobe loss, whereas 25% of the MBs showed bilateral α-lobe loss. Interestingly, the β lobes were always bifurcated (Fig. 2E) or were much thicker (Fig. 2F) when the ipsilateral α lobe was missing. One possibility for the bifurcated or thickened β lobes is that the α lobe projects parallel to its ipsilateral β lobe, or that the α and β lobes fail to branch at the base of the peduncle, thus forming thickened β lobes instead. These data together show that twf regulates axonal growth and guidance.

As shown above, twinfilin is enriched at the postsynaptic site (Fig. 1C-J). It has been established that F-actin is also enriched at the postsynaptic site. Postsynaptic F-actin can be best visualized using fluorescently conjugated phalloidin by means of thin (0.4 μm) confocal sectioning (Coyle et al., 2004; Chen et al., 2005). To test whether twinfilin regulates the formation of F-actin, we examined the level of F-actin at the postsynaptic site in the wild type and twf mutants (Fig. 3A,B). In wild-type larvae, postsynaptic F-actin surrounded presynaptic boutons, whereas in twf¹¹⁰ null mutants, a markedly increased level of F-actin was observed, using anti-HRP staining as an internal control (compare Fig. 3B with 3A), even though the NMJ morphology was normal. Statistical analyses showed that twf¹¹⁰ mutants had significantly increased phalloidin-staining intensity compared with the wild-type controls (Fig. 3D; 135.6±3.2 for mutants versus 100±1.8 for wild type; n=8; P<0.001). Conversely, when twf was overexpressed postsynaptically by muscle-specific C57-Gal4, the F-actin level showed a significant 17% reduction in staining intensity compared with the wild type (P<0.001; Fig. 3D).

To further understand the role of twinfilin in regulating F-actin formation, we examined the level of F-actin in the cluster of 6-10 migratory BCs where twinfilin is highly expressed (Fig. 4A) (Wahlström et al., 2001). Twinfilin is mainly cytoplasmic, whereas F-actin is localized to the cortical region of the BCs and is highly enriched in the periphery of the BC cluster (Fig. 4A). Statistical analysis showed that the twf³⁷⁰¹ mutation resulted in a twofold increase in F-actin staining intensity in the BCs (203.5±11.1 in twf mutants versus 100±6.2 in wild type; n=10 for both; P<0.001; Fig. 4D). Conversely, overexpressing twf in BCs by slbo-Gal4 resulted in a significant 35% decrease in the F-actin level (P<0.001; Fig. 4D).

In addition to the mutant analysis, we also performed mosaic analysis to better examine the role of twinfilin in F-actin formation. As twf is highly expressed in the large, single layer of follicle cells of a developing egg chamber (Wahlström et al., 2001), where the effect of actin regulators such as profilin and cofilin on F-actin formation has been well studied (Baum and Perrimon, 2001), we generated mitotic clones mutant for twf¹¹⁰ in the follicle cells. Cells from a planar section in twf¹¹⁰ mutant clones exhibited a significant increase in the F-actin level compared with the adjacent wild-type cells (Fig. 4E-H). To better reveal F-actin misregulation in twf mutants, we analyzed F-actin staining in a cross-section of the epithelial follicle cells (Fig. 4I-L). An obvious increase in F-actin was largely limited to the apical surface and lateral junctions, but not to the basal regions of the twf¹¹⁰ mutant follicle cells (Fig. 4I-L), consistent with the increased F-actin level in the planar view (Fig. 4F). Increased F-actin was also observed in twf¹¹⁰ mutant clones in eye disc epithelial cells (supplementary material Fig. S2). Taken together, both mutant and clonal analyses demonstrate that Drosophila twinfilin negatively regulates F-actin formation in multiple cellular contexts.

We have previously shown that ADF/cofilin suppresses F-actin formation and is required for BC migration (Chen et al., 2001). Because both twinfilin and ADF/cofilin suppress F-actin formation, we then tested whether twinfilin also plays a similar role in BC migration. The BCs in stage 9 egg chambers migrate as a cluster between the nurse cells and eventually reach the border between the nurse cells and the oocyte by early stage 10. The stereotypic migration of BCs can be easily quantified; thus, it serves as an excellent model for evaluating the role of a gene in cell migration. It was previously reported that hypomorphic twf³⁷⁰¹ displays no obvious defects in BC migration at 25°C (Wahlström et al., 2001), which we have confirmed. However, we detected a moderate but significant delay in BC migration in both twf³⁷⁰¹ homozygous and twf³⁷⁰¹/twf¹¹⁰ heteroallelic mutants when they were cultured at 29°C, a suboptimal temperature for fruit flies (Fig. 5H). Targeted overexpression of twf in BCs by BC-specific slbo-Gal4 at 29°C showed a mild migration delay (Fig. 5H), indicating that an apparent reduction in F-actin levels in the BC cluster does not have an obvious effect on BC migration. However, twf RNAi knockdown under identical conditions resulted in a significant migration delay: 48% of the BC clusters examined failed to reach the border, whereas only 4% of the BC clusters in GFP control animals had yet to reach their destination (Fig. 5H). In the overexpression and RNAi knockdown assays, transgenic flies were cultured at 29°C for a stronger effect due to higher expression of Gal4 at that temperature. Similarly, twf¹¹⁰ mitotic clones in the BC clusters also resulted in a migration delay at 25°C: 52% of the BC clusters containing twf¹¹⁰ mutant clones displayed a migration delay, whereas 96% of the wild-type BC clusters reached their final destination (Fig. 5I). These results together demonstrate an essential role for twf in BC migration.

It has been demonstrated that the actin cytoskeleton plays a role in the postsynaptic membrane localization of glutamate receptors (GluRs). This modulates the postsynaptic response to neurotransmitters (Shirao and Sekino, 2001; Coleman et al., 2003; Chen et al., 2005). Two subtypes of GluRs are characterized by the presence of either the GluRIIA or GluRIIB subunit, together with shared GluRIIC, GluRIID and GluRIIE subunits (Chen et al., 2005). The A- and B-type receptors have functional differences and are spatially segregated as homotypic clusters (DiAntonio et al., 1999; Marrus et al., 2004). In mammals, extensive studies have shown that GluRs cluster at postsynaptic sites through interactions with the actin cytoskeleton (Rosenmund and Westbrook, 1993; Shirao and Sekino, 2001; Coleman et al., 2003).

We investigated whether actin-mediated GluR localization at the postsynaptic site was disrupted in twf mutants. As shown in Fig. 6, GluRIIA levels were mildly but significantly decreased in twf-null mutants (77.8±2.0; n=12; P<0.001) compared with the wild type (100±3.1; n=12). By contrast, GluRIIB levels remained the same in twf mutants (96.9±2.7; n=12; P>0.1) as in the wild type (100±2.6; n=12; Fig. 6B,D). To ascertain whether the decreased levels of GluRIIA in twf mutants were specifically due to the absence of twinfilin, we examined the expression of GluRs when twf was expressed by muscle-specific C57-Gal4 in the twf mutant background. Expression of twf postsynaptically rescued the decreased GluRIIA level in the twf mutants (96.2±2.4; n=6; P>0.1), although the level of GluRIIB was mildly but significantly reduced (88.9±3; P<0.05; Fig. 6C,D). These data demonstrate that twf regulates the postsynaptic expression of glutamate receptor GluRIIA in a subtype-specific manner.

GluRIIA was reported to be anchored to F-actin at NMJ synapses through Coracle, the Drosophila homologue of actin-interacting 4.1 protein (Chen et al., 2005). To test whether Coracle expression at NMJ synapses was altered in twf mutants, we stained NMJ synapses with guinea-pig anti-Coracle. The intensity of anti-Coracle staining was significantly reduced by 30% in twf¹¹⁰ mutants compared with the wild type (n=10; P<0.001; Fig. 6E,F), suggesting that the reduced GluRIIA in twf mutants might be caused by decreased Coracle.

In yeast, twf mutants have no defects in fluid-phase marker uptake (Goode et al., 1998). However, in cultured mammalian cells, the subcellular distribution of endocytic organelles is disrupted when twf is knocked down (Pelkmans et al., 2005); expression of either the N- or the C-terminal ADF-H domain of mouse twinfilin leads to decreased efficiency of transferrin uptake (Helfer et al., 2006). These data indicate that twinfilin might be involved in endocytosis and in actin-driven internalization or mobility of endocytic vesicles in metazoans. However, a possible role for twinfilin in endocytosis at the multicellular organism level has not yet been tested.

To examine whether endocytosis at NMJ synapses is affected in twf mutants, we performed fluorescent dye FM1-43 uptake experiments. FM1-43 can be taken up in synaptic vesicles by endocytosis and released by exocytosis. Compared to the wild-type controls, which displayed intense labeling of FM1-43 at synaptic boutons (the revertant 138 also showed synaptic loading of FM1-43 comparable to that of the wild type, data not shown), twf-null mutants showed a significantly reduced amount of loaded dye (Fig. 7). The relative fluorescent intensity of twf mutants was ~60% of that of the wild-type controls (n=8; P<0.001; Fig. 7A,B,G). Correspondingly, the amount of remaining dye in the mutants after unloading triggered by high K⁺ stimulation was also reduced compared with that in the controls (P<0.01; Fig. 7C,D,G). However, the ratio of unloaded to loaded FM1-43 showed no difference between twf mutants and wild-type controls, indicating that exocytosis is not affected (Fig. 7H). The decrease in the amount of loaded and unloaded FM1-43 dye in twf mutants was largely rescued by presynaptic expression of twf by elav-Gal4 (Fig. 7G). These results indicate that twf is involved specifically in endocytosis but not in exocytosis.

As shown above, twf mutants show increased F-actin formation (Figs 3 and 4; supplementary material Fig. S2). To further understand whether the endocytosis defect in twf mutants is caused by the increased F-actin level, we examined endocytosis after pharmacological disruption of the actin cytoskeleton. We treated NMJ synapses before FM1-43 dye uptake with a low concentration (0.1 μM) of latrunculin A (LatA), which promotes actin depolymerization by sequestering G-actin. The relative amount of loaded FM1-43 in wild-type synapses was significantly reduced compared with the untreated controls (Fig. 7E,I; 21.5±1.3 in twf mutants versus 100±4.1 in the wild type; P<0.001). By contrast, LatA treatment of the twf mutants showed no significant effect on endocytosis (Fig. 7F,I; 61.1±6.6 for LatA treated versus 61.7±6.6 for untreated; P>0.05). However, when treated with a high concentration of LatA (10 μM), endocytosis was completely blocked in both wild-type and mutant animals (data not shown). The pharmacological analysis showed that endocytosis in twf mutants is more resistant to LatA treatment at a low concentration of 0.1 μM, consistent with our finding presented above that the levels of F-actin are increased in the twf mutants.

To further ascertain whether twf is required for endocytosis, we examined synaptic neurotransmission in twf mutants using physiological assays. We first recorded the postsynaptic excitatory junctional potentials (EJPs) triggered by nerve stimulation and spontaneous miniature excitatory junctional potentials (mEJPs) without nerve stimulation at NMJ synapses of late third instar larvae. Evoked EJP amplitudes upon low-frequency stimulation of 0.3 Hz in twf¹¹⁰ and control animals were similar (Fig. 8A-C; P>0.05). In addition, we evaluated spontaneous neurotransmitter release by analyzing mEJPs. The mEJP amplitudes were not significantly altered in twf mutants compared with the wild type (Fig. 8D). The mEJP frequency in homozygous twf¹¹⁰ mutants showed a significant decrease (P<0.05), but the decrease in mEJP frequency did not present in twf¹¹⁰ hemizygous mutants (P>0.05), suggesting that the phenotype is not caused by the twf mutation (Fig. 8E). The quantal content, calculated by dividing the corrected EJP amplitude by the mEJP amplitude, was not altered either (Fig. 8F; P>0.05). Thus, basal transmission upon low-frequency stimulation is not affected in the twf mutants.

The twf mutations impair endocytosis but not exocytosis (Fig. 7), and endocytosis mutants such as endophilin and dap160/intersectin mutants cannot maintain neurotransmission during intense stimulation because of insufficient replenishment of synaptic vesicles (Verstreken et al., 2002; Koh et al., 2004). Therefore, we then examined the EJPs when intense stimulation at 10 Hz was delivered to the nerve for 600 seconds. In control animals, the EJP amplitudes gradually declined to about 60% of the original level within the 600 second stimulation period. By contrast, under the same stimulation regime, the EJP amplitudes detected in the twf¹¹⁰-null mutants dropped sharply in the first 240 second period (Fig. 8G) and reached about 20% of the original level. The hemizygous mutants (twf¹¹⁰ over-deficiency) showed a similar phenotype as the homozygous twf¹¹⁰ mutants. Furthermore, presynaptic expression of twinfilin driven by elav-Gal4 completely rescued the enhanced decrease in EJP amplitudes in twf¹¹⁰ mutants (Fig. 8G). Together, these data demonstrate that the dramatic reduction in EJPs under high-frequency stimulation, consistent with the defective endocytosis detected by the FM1-43 uptake assay, is caused specifically by twf mutations.

Yeast twf-null mutants are viable, but synthetically lethal in combination with profilin or ADF/cofilin mutants (Goode et al., 1998; Wolven et al., 2000; Falck et al., 2004). In Drosophila, transheterozygotes of twf and tsr, which encodes ADF/cofilin, show a subtle bristle phenotype reminiscent of that seen in homozygous twf mutants (Wahlström et al., 2001), consistent with the fact that both twf and tsr suppress F-actin formation (Figs 3 and 4) (Chen et al., 2001). To further understand the role of twinfilin in regulating the actin cytoskeleton, we examined the genetic interaction of twinfilin with profilin, which promotes actin polymerization and is encoded by chic. Wild-type bristles have smooth, parallel ridges, whereas the bristles of the twf-null adult escapers show forked tips and severely twisted ridges (Fig. 9A,B). Bristles of tsr^N96A/+;twf¹¹⁰/+ transheterozygotes showed mildly disorganized ridges and hooked tips, confirming the previous report (Fig. 9C) (Wahlström et al., 2001). Because profilin promotes F-actin assembly, whereas twinfilin inhibits F-actin formation, we would expect a rescue effect if the amount of profilin is reduced in the twf mutant background. This is indeed the case. Homozygous chic²²¹ is embryonically lethal. Double mutants of twf and chic²²¹ are embryonically lethal, precluding the examination of bristles, whereas reducing profilin by half partially rescued the bristle phenotypes of twf¹¹⁰ homozygous mutants (compare Fig. 9D with Fig. 9B). As a control, chic²²¹/+;twf¹¹⁰/+ transheterozygotes showed wild-type bristles (data not shown).

In addition, the bristle length of twf¹¹⁰ animals is significantly shorter than that of the wild type (Fig. 9E). Statistical analysis showed that the reduced bristle length in twf mutants was also partially rescued by reducing chic expression by half (177.4±6.7 μm in twf mutants and 199.1±8.2 μm in chic²²¹/+;twf¹¹⁰; n>19; P<0.05; Fig. 9E). The interaction between twf and chic is similar to that reported for the interaction between capping protein and chic (Hopmann and Miller, 2003). Capping protein limits F-actin formation by binding to the F-actin barbed ends.

To further confirm the genetic interaction between twf and chic, we examined FM1-43 uptake at NMJ synapses and found that the reduced FM dye uptake in twf¹¹⁰ mutants was rescued substantially in the double mutants (P<0.01; supplementary material Fig. S3), consolidating the antagonistic effect between the two. The genetic interaction assays indicate that twf and chic function antagonistically in regulating actin dynamics during bristle development and synaptic endocytosis.

Twinfilin is well established as a protein that sequesters actin monomers (Goode et al., 1998; Wahlström et al., 2001; Helfer et al., 2006). However, its in vivo function remains largely unclear. Our work provides multiple lines of evidence from both mutant and mosaic clone analyses demonstrating that Drosophila twf suppresses F-actin formation in subcellular regions of high actin dynamics in multiple cellular contexts, that is, NMJ synapses, migrating BCs and epithelial follicle cells. In twf mutants and mutant clones, there is excessive F-actin, whereas overexpression of twf leads to reduction in the F-actin level (Figs 3 and 4; supplementary material Fig. S1). In yeast, twf mutation also results in enlargement of cortical actin patches (Goode et al., 1998; Moseley et al., 2006). The increase in the F-actin level of twf mutant yeast and Drosophila is consistent with in vitro studies demonstrating that twinfilin inhibits nucleotide exchange on actin monomers and prevents assembly of the monomer into filaments (Palmgren et al., 2002). It is noteworthy that loss of ADF/cofilin function also results in excessive F-actin in yeast and in Drosophila (Lappalainen and Drubin, 1997; Chen et al., 2001). Moreover, genetic interaction assays in Drosophila showed that twinfilin acts in the same process as ADF/cofilin, but antagonistically with profilin in actin-based bristle development (Fig. 9) (Wahlström et al., 2001). Thus, our work provides multiple lines of in vivo evidence implicating twinfilin in regulating F-actin formation and actin dynamics.

Twinfilin promotes actin-based motility in an in vitro biomimetic motility assay (Helfer et al., 2006). In S2 cells, RNAi knockdown of twinfilin expands the lamellipodia at the expense of the lamellum (Iwasa and Mullins, 2007). However, its requirement in cellular processes such as axonal growth and cell migration has not been tested. We found that twf mutants show overgrown axons in the brain, in contrast to tsr mutants (Ng and Luo, 2002), but similar to that reported for dfmr1 mutants (Michel et al., 2004) and for cib-overexpressing animals (Boquet et al., 2000). Cib is an actin-binding protein with a function similar to profilin (Boquet et al., 2000). BC migration in the egg chamber of Drosophila is a well-studied cell behavior that is genetically tractable (Naora and Montell, 2005). Many proteins, including the actin-monomer-binding proteins profilin and ADF/cofilin, have been identified to be required for this process (Chen et al., 2001; Naora and Montell, 2005). Our work from mutant and clonal analyses, as well as RNAi knockdown experiments, provides evidence demonstrating, for the first time, that twinfilin is required for BC migration in the developing ovary (Fig. 5).

The mechanism by which neurotransmitter receptors are anchored to the postsynaptic membrane is largely unknown. Our immunohistochemical analysis revealed that the post-synaptic expression of GluRIIA, but not GluRIIB, is specifically reduced in twf mutants, demonstrating that twf regulates GluR localization at synapses in a receptor-subtype-specific manner. Coracle, the fly homologue of mammalian actin-interacting protein 4.1 that binds directly and specifically to GluRIIA (Chen et al., 2005), was also reduced in twf mutants (Fig. 6E,F). These results agree well with previous findings that Coracle is required for the proper localization of GluRIIA, but not GluRIIB (Chen et al., 2005). Pharmacological disruption of F-actin phenocopies the coracle mutants with reduced GluRIIA expression (Chen et al., 2005). However, in twf mutants, there was more F-actin than in the wild type (Fig. 3), but GluRIIA was specifically downregulated, suggesting that the mere presence of F-actin is not sufficient to localize GluRIIA correctly. As twinfilin is conserved throughout evolution and actin regulates the postsynaptic localization of mammalian GluRs (Coleman et al., 2003; Cingolani and Goda, 2008), we predict a similar role for twinfilin in mammals.

Although yeast twf-null mutants have no defects in endocytosis (Goode et al., 1998), mammalian twinfilin is involved in clathrin-mediated endocytosis and redistribution of endocytic organelles in cultured cells (Pelkmans et al., 2005). Overexpression of either the N- or the C-terminal ADF-H domain of mammalian twinfilin in cultured cells also results in defects in receptor-mediated endocytosis (Helfer et al., 2006). Consistent with the studies in cultured cells (Pelkmans et al., 2005; Helfer et al., 2006), our phenotypic analyses of Drosophila mutants demonstrate, for the first time, that twinfilin is required for presynaptic endocytosis in vivo (Figs 7, 8), even though it is enriched at the postsynaptic site (Fig. 1). First, twf mutants had endocytic defects at NMJ synapses, as shown by the FM1-43 uptake assay (Fig. 7A-D,G). Second, pharmacological disruption of F-actin by LatA at a low concentration of 0.1 μM in the wild type caused endocytic defects, but had little effect on endocytosis in twf mutants (Fig. 7E-F,I), consistent with the presence of increased F-actin levels in twf mutants (Figs 3, 4). Third, the EJP amplitude dropped precipitously in the twf mutant synapses under intense stimulation (Fig. 8G), whereas the basal transmission is normal (Fig. 8A-F), indicating an inability to retrieve synaptic vesicles. Last, endocytic protein AP180, but not other synaptic proteins such as α-adaptin, Eps15 and EndoA, was mildly but significantly reduced at NMJ synapses in twf mutants by immunostaining (supplementary material Fig. S4). Collectively, our data demonstrate that Drosophila twinfilin is required for endocytosis. It has been well established that a highly dynamic actin cytoskeleton is essential for endocytosis (Kaksonen et al., 2006; Smythe and Ayscough, 2006). However, the mechanism by which twf regulates endocytosis is currently unknown. One simple explanation is that the increased level of F-actin in the twf mutants hinders the actin-driven internalization or mobility of endocytic vesicles. Alternatively, the endocytosis defects in twf mutants could be caused by misregulation of the endocytic protein AP180.

Interestingly, mutants of endocytic genes such as endophilin, synaptojanin, dynamin and AP180 all show dramatic increases in the number of satellite boutons at NMJ synapses (Dickman et al., 2006). Similarly, mutations in other actin regulators, such as CYFIP (cytoplasmic FMRP interacting protein), WASP (Wiskott-Aldrich syndrome protein) and Nwk (nervous wreck), also exhibit prominent defects in synaptic growth (Schenck et al., 2003; Coyle et al., 2004). As shown in this study, twinfilin is involved in endocytosis and negatively regulates F-actin formation. However, the NMJ synapses in twinfilin mutants examined by immunostaining displayed no obvious morphological abnormalities (data not shown), demonstrating that, in contrast to endocytic proteins and the other actin regulators, twinfilin is not essential for synapse growth, but rather is involved in specific synaptic functions, that is, postsynaptic localization of glutamate receptors and presynaptic endocytosis. The mechanism by which twf functions at synapses, as related to endocytic proteins such as AP180 and the other actin regulators, remains to be elucidated.

Like twinfilin, other actin-monomer-binding proteins, such as ADF/cofilin and profilin, are also essential for endocytosis in both yeast and mammalian cells. ADF/cofilin, which promotes F-actin depolymerization, is required for endocytosis and membrane trafficking to the vacuoles in yeast (Lappalainen and Drubin, 1997; Idrissi et al., 2002; Okreglak and Drubin, 2007). Profilin, which promotes F-actin polymerization, has the opposite effect to ADF/cofilin on endocytosis in cultured cells (Gareus et al., 2006). Loss of mouse brain-specific profilin 2 results in increased endocytosis, whereas its overexpression leads to compromised endocytosis in cultured neurons (Gareus et al., 2006). It is apparent that, in addition to twinfilin, ADF/cofilin and profilin, the two crucial players in actin filament turnover, are also involved in endocytosis.

In summary, we have shown that twf mutants exhibit increased F-actin formation in disparate cellular contexts and defects in different cellular functions, from axonal growth and cell migration to synaptic endocytosis. A common feature of these processes is that they all depend on a highly dynamic actin cytoskeleton. When the function of twinfilin is abolished in these processes, dynamic actin would be turned into a more ‘stagnant’ state, which is reflected in the increased F-actin formation observed in different cellular contexts. The disruption of actin dynamics in twf mutants could compromise endocytosis in NMJ synapses and lamellipodial extension in both growing axons and migrating cells. Thus, our study reveals a crucial in vivo role for twinfilin in promoting actin dynamics, which is essential for a range of cellular processes in both neuronal and non-neuronal systems.

Fly cultures were maintained at 25°C unless specified. w¹¹¹⁸ flies were used as the wild-type control in all experiments. twf³⁷⁰¹ is a hypomorphic allele of twf as previously described (Wahlström et al., 2001). Two twf deletion alleles, twf¹¹⁰ and twf¹⁰², were generated by means of P-element-mediated excision from EP(3)3701 (Fig. 1). Df(3R)suHw⁷ (deleting 88A9-B2 carrying twf), tsr^N96A and chic²²¹ were from the Bloomington Stock Center. The twf RNAi transgenic stock was obtained from the Vienna Drosophila RNAi Center. The UAS-twf strain was generated by inserting the full-length ORF of twf from the pGAT2 expression vector (Wahlström et al., 2001) into the pUAST vector, followed by germline transformation of the construct into w¹¹¹⁸ as the host. For targeted knockdown, overexpression or rescue experiments, the tissue-specific Gal4 drivers elav-Gal4 (pan-neuronal), OK107-Gal4 (mushroom body), C57-Gal4 (muscular), slbo-Gal4 (BCs) and act-Gal4 (ubiquitous) were used. To obtain twf mutant clones in the ovary, y w; FRT82B twf¹¹⁰/TM6B was crossed with y w hs-Flp; FRT82B ubi-GFP/TM3. All progeny were heat shocked at 37°C, 2 hours a day for 3 days beginning at the second-instar larval stage. Upon eclosion, flies were transferred to fresh food and dissected 3 days later.

Immunostaining of various tissues was performed essentially as previously described (Zhang et al., 2001; Jin et al., 2009). Specifically, late third-instar larvae were dissected in Ca²⁺-free standard saline and then fixed in fresh 4% paraformaldehyde for 30 minutes for all NMJ preparations. For GluRIIA and GluRIIB staining, larvae were dissected in saline with 2 mM L-glutamate, followed by ice-cold methanol fixation for 5 minutes. Adult brains were dissected in PBS and then fixed in fresh 4% paraformaldehyde for 40 minutes. Ovaries were dissected in PBS and then fixed in a mixture of devitellinizing buffer (7% formaldehyde) and heptane (Sigma) (1:6) for 10 minutes. The following primary antibodies or reagents were used: FITC-conjugated goat anti-HRP (Jackson ImmunoResearch, 1:100), Texas-Red phalloidin (Molecular Probes, 1:6 for NMJ staining), TRITC phalloidin (Sigma, 1:100 for ovary staining), rabbit anti-GluRIIB (from Aaron DiAntonio, Washington University, WA, 1:2500), rabbit anti-twinfilin (1:2000) (Wahlström et al., 2001) and guinea-pig anti-Coracle (from Richard Fehon, University of Chicago, IL, 1:250). Monoclonal mouse antibodies anti-Dlg (4F3; 1:1000), anti-Fas II (1D4; 1:50) and anti-GluRIIA (1:50) were from the Developmental Studies Hybridoma Bank at the University of Iowa. AlexaFluor-488, -568 or -635-conjugated anti-mouse or anti-rabbit secondary antibodies (Invitrogen; 1:1000) were used. All images were collected using a Leica SP5 laser scanning confocal microscope or an Olympus BX51 fluorescent microscope, and processed with Adobe Photoshop 8.0.

For western blotting, 20 adult heads from each genotype were homogenized in 120 μl common lysis buffer containing 2% SDS on ice. The immunoblots were incubated overnight at 4°C with the following primary antibodies: anti-twinfilin (1:20,000), anti-dAbi (from Sven Bagdan, Wilhelms-University Münster, Germany; 1:1000), anti-α-tubulin (mAb B-5-1-2; Sigma; 1:70,000) and anti-actin (mAb1501; Chemicon; 1:35,000). The secondary antibody HRP-labeled anti-mouse IgG was diluted at 1:20,000 in PBS-Tween containing 5% milk. Protein bands were visualized with HRP chromogenic reagents from Millipore.

Scanning electron microscopy

Scanning electron microscopy (SEM) of bristles was carried out as previously published (Hopmann and Miller, 2003). Briefly, after anaesthetizing with CO₂, adult flies with wings removed and abdomens punctured were transferred to a solution of 25% glycerol in 95% ethanol for at least 2 days. The fixed flies were dehydrated through a graded series of ethanol solutions: 70%, 95% and three times of 100% ethanol for 15 minutes each. The ethanol was then replaced with hexamethyldisilazane with four changes to prevent distortion, followed by drying overnight and sputter coating with gold before phototaking with a Hitachi S-3000N scanning electron microscope. Statistical analysis of bristle length was performed (following Hopmann and Miller, 2003).

For the FM1-43 uptake assay, we followed previously published protocols (Kuromi and Kidokoro, 1998; Dickman et al., 2005). Late third-instar larvae were dissected in a low Ca²⁺ saline and washed with the normal medium (Jan and Jan, 1976). To load the fluorescent FM1-43 dye, membrane depolarization was achieved by incubating the preparations in a high K⁺ (90 mM) solution containing 10 μM FM1-43 (Molecular Probes) for 5 minutes. The preparations were vigorously washed three times with Ca²⁺-free saline for 5 minutes. The loaded synapses were imaged on a Leica SP5 confocal microscope with a 40× water-immersion lens. After imaging, the preparations were rested for 5 minutes and high K⁺ saline without FM1-43 was applied again for 5 minutes to elicit exocytosis. To investigate the effect of actin drugs on synaptic vesicle endocytosis, the dissected larval preparations were incubated for 5 minutes with 0.1 or 10 μM LatA, an actin-monomer-sequestering drug, before loading FM1-43. The composition of the normal medium (Jan and Jan, 1976) was as follows: 128 mM NaCl, 35.5 mM sucrose, 2 mM KCl, 4 mM MgCl₂, 1.8 mM CaCl₂ and 5 mM (pH 7.3) HEPES. For low Ca²⁺ saline, 0.2 mM instead of 1.8 mM Ca²⁺ was added to the normal medium, whereas for Ca²⁺-free saline, CaCl₂ in the normal medium was replaced by 0.5 mM EGTA. High K⁺ saline contained 90 mM KCl, whereas NaCl was reduced by an equivalent amount.

Image J was used to quantify fluorescence intensities. Antibody staining and the FM1-43 dye uptake assay were repeated at least three times and the different genotypes were processed in the same tube for better comparison. For each channel, an arbitrary threshold was set and used for all relevant images. The sum of pixel intensities above the threshold was recorded for each channel. The ratio of phalloidin:anti-HRP, anti-GluRIIA:anti-HRP or anti-GluRIIB:anti-HRP staining intensity was calculated for each genotype in Figs 3 and 6. Three individual boutons were analyzed for phalloidin staining quantification per animal (Fig. 3), whereas two or three NMJs in the A2 or A3 segment of each animal were analyzed for GluR staining (Fig. 6) and FM1-43 live imaging (Fig. 7). For comparison of the F-actin level in BCs (Fig. 4), ovaries expressing UAS-twf or twf siRNA were mixed with the wild-type controls. They were then processed in the same tube and photographed using the same settings. The genotypes were unambiguously identified by the presence or absence of GFP and anti-twinfilin staining. Statistical analysis of fluorescence intensities was performed using the Student's t-test.

Only stage 10 egg chambers were selected and examined for quantification of BC migration. Flies examined for BC migration were raised at 29°C for twf overexpression, RNAi knockdown and twf mutants or at 25°C for clonal analysis of twf mutants. The extent of BC migration is measured as 0% (no migration), <50%, >50% and 100% (reaching the border) for quantitative analyses. Wild-type BCs would have migrated to the 100% position in a stage 10 egg chamber, whereas a migration delay means that the BC cluster fails to migrate to the final 100% position.

Intracellular recordings were carried out at 20°C, following a conventional procedure (Jan and Jan, 1976; Jin et al., 2009). For the basal transmission assay, wandering third-instar larvae were dissected in Ca²⁺-free HL3.1 saline (Feng et al., 2004) and recorded in HL3.1 saline containing 0.25 mM Ca²⁺. Intracellular microelectrodes with a resistance of 8-20 MΩ filled with 3M KCl were used for the assay. Recordings were performed using an Axoclamp 2B amplifier (Axon Instruments) in Bridge mode. Data were filtered at 1 kHz, digitized using Digitizer 1322A (Axon Instruments) and collected with Clampex 9.1 software. EJPs were evoked at 0.3 Hz by a suction electrode with a depolarizing pulse delivered by a Grass S48 stimulator (Astro-Grass Inc.). EJPs were recorded from muscle 6 of abdominal segment A2 or A3, followed by miniature EJP (mEJP) recording for 120 seconds. EJP and mEJP recordings were processed with Clampfit 10.2 software. Quantal content was calculated by dividing the corrected EJP amplitude by the mEJP amplitude according to a classical protocol (Martin, 1955). The EJP correction for nonlinear summation was done using a reversal potential of 10 mV. At least eight animals were recorded for each genotype. For examining synaptic transmission under tetanic stimulation, synapses were stimulated at 10 Hz for 10 minutes and recorded in a modified HL3.1 saline with 10 mM extracellular Ca²⁺ (Dickman et al., 2005).

We thank Vivian Budnik, Sven Bogdan, Richard Fehon, Aaron DiAntonio, the Bloomington Stock Center, the Vienna Drosophila RNAi Center and the Developmental Studies Hybridoma Bank, University of Iowa for fly stocks and antibodies. We thank Pekka Lappalainen for comments on the manuscript and Zhihua Liu for advice on electrophysiology. We are grateful to Xing Ma for assistance in the identification of twf mutants and analysis of mushroom-body phenotypes. Discussions with David Featherstone and members of our laboratories are acknowledged. This work is supported by grants from the National Science Foundation of China (30430250 and 30525015 to Y.Q.Z.; 30570910 and 90608018 to J.C.), the Ministry of Science and Technology of China (2007CB947200 to Y.Q Z.; 2007CB947101 and 2006CB943503 to J.C.) and the Chinese Academy of Sciences (KSCX1-YW-R-69 to Y.Q.Z.).