Abp1 utilizes the Arp2/3 complex activator Scar/WAVE in bristle development

Shaping of the plasma membrane is crucial for cell morphology changes. The forces required for such processes are brought about by cortical actin dynamics. Nucleation is rate-limiting in actin filament formation (Pollard, 2007). One of the most prominent nucleators is the Arp2/3 complex (Pollard, 2007). It is controlled by members of the Scar/WAVE and Wiskott-Aldrich syndrome protein (WASP) family (Stradal and Scita, 2006; Takenawa and Suetsugu, 2007).

In contrast to vertebrate genomes, Drosophila melanogaster only carries a single gene of each family, wasp and scar, therefore, genetic analyses in Drosophila were very valuable and led to the assignment of the WASP and the Scar pathways to specific functions (Ben-Yaacov et al., 2001; Zallen et al., 2002; Tal et al., 2002; Coyle et al., 2004). Notably, studies in flies also suggested some cross-talk between Scar and WASP (Bogdan and Klämbt, 2003; Bogdan et al., 2004; Bogdan et al., 2005; Fricke et al., 2009).

Cell cortex targeting is a crucial aspect in the control of Arp2/3 functions. It is thought to be mediated by the Arp2/3 complex binding to the sides of existing cortical actin filaments and by lipid interactions of Arp2/3 activators (Pollard, 2007). Phosphatidyl-inositol-(4,5)-bisphosphate (PIP₂) binds to and controls the activity of N-WASP (Rohatgi et al., 2001). Phosphatidyl-inositol-(3,4,5)-trisphosphate (PIP₃) was found to be important for Scar/WAVE activity (Oikawa et al., 2004). Dictyostelium WASp binds both PIP₂ and PIP₃ (Myers et al., 2005). In general, such lipid signals are combined with further cues. WASP and N-WASP are kept inactive by autoinhibition, which can be released by different means. Control of stability and activity of Scar is mediated by the Scar/WAVE complex including Nap1/Kette, Sra-1/PIR121, Abi and HSCP300, among which Sra-1/PIR121 associates with Rac1 (Stradal and Scita, 2006; Takenawa and Suetsugu, 2007). However, relatively little is known about how these multiple layers of regulation work and are coordinated in vivo.

Mammalian Abp1, an SH3 domain-containing F-actin-binding protein (Kessels et al., 2000), is a crucial N-WASP-binding and -activating protein that cooperates with Cdc42 in controlling N-WASP functions in early neuromorphogenesis (Pinyol et al., 2007; Dharmalingam et al., 2009) and mediates dendritic spine enlargement in an N-WASP- and Arp2/3-dependent manner (Haeckel et al., 2008). These studies firmly established a physiological role of Abp1/N-WASP complexes.

In this study, we show that Abp1 control of Arp2/3 complex functions is not restricted to N-WASP autoinhibition release. Genetic analyses in Drosophila and biochemical experiments reveal that Abp1 uses the Scar pathway to control Arp2/3-mediated, cortical actin nucleation during bristle development. Abp1 is hereby functionally distinguished from the established Scar complex components. Abp1 coordinates Scar- and Arp2/3-mediated actin nucleation with phosphoinositide signals of the plasma membrane without affecting the Scar protein stability. The fact that knockout of abp1 copies the effects of Scar RNAi therefore reflects a lack of Scar's biological functions in abp1^KO flies and does not merely reflect a reduction of Scar protein levels upon loss of the Scar-associating component Abp1. Thus, control of Scar functions at the cell cortex is far more complex than previously thought and critically involves Scar-associating proteins beyond the established Scar complex components that are only beginning to emerge.

To gain novel insights into the regulation of Arp2/3 complex-mediated actin nucleation during physiological processes, we set out to identify an ortholog of Abp1, which physically and functionally interfaces with the Arp2/3 complex activator N-WASP in mammals (Pinyol et al., 2007; Haeckel et al., 2008) and, in yeast, activates the Arp2/3 complex directly (Goode et al., 2001), in the multicellular and genetically well-accessible model organism Drosophila melanogaster. Database analyses revealed that gene locus CG10083 encodes the only putative Abp1 protein in Drosophila. Its domain structure is consistent with that of mammalian Abp1 containing a relatively well-conserved N-terminal part composed of an ADF-H domain and a helical domain and a C-terminal part comprising an uncharacterized linker area and a well-conserved, Src Homology 3 (SH3) domain (Fig. 1A; supplementary material Fig. S1). Drosophila Abp1 lacks the acidic motifs binding the Arp2/3 complex in yeast (Goode et al., 2001) and thus rather resembles mammalian Abp1 proteins.

To characterize Drosophila Abp1 biochemically, we cloned full-length fly Abp1 and then tested a GST fusion protein thereof for F-actin binding (Fig. 1B). Similar to mammalian Abp1 (Kessels et al., 2000), Drosophila Abp1 directly associated with F-actin in cosedimentation assays (Fig. 1B; pellet fraction). Deletion analyses demonstrated that C-terminal parts of Abp1 were not able to bind to F-actin. Only proteins containing the N-terminal ADF-H and helical domains showed F-actin-binding in vitro (Fig. 1B).

We next raised and affinity-purified rabbit antibodies against fly Abp1. Western blot analyses of recombinant proteins and fly extracts as well as immunofluorescence analysis of COS-7 cells transfected with myc-tagged Abp1 show that the antibodies specifically detect Abp1 (supplementary material Fig. S2).

In line with Abp1's ability to associate with F-actin in vitro, the subcellular distribution of endogenous Abp1 in spreading Drosophila S2 cells largely overlapped with F-actin at the cellular cortex (Fig. 1C; for merged image see supplementary material Fig. S3). Anti-Abp1 and phalloidin signal intensity profiles across the cells showed good overlap of Abp1 with F-actin especially at the cell periphery (Fig. 1D). Cortical localization of Abp1 was also evident in embryonic and larval epithelia, such as the imaginal wing discs of third instar larvae (Fig. 1E, inset in 1F).

In flies, Abp1 was present at the F-actin-rich cortex of S2 cells and epithelial cells (Fig. 1) but certain F-actin-rich areas, such as adherens junctions, were not marked by an equal enrichment for Abp1 (not shown). We thus reasoned that Abp1 may not only associate with F-actin but also with the plasma membrane directly. In fractionations of both larval and S2 cell extracts, a subpool of Abp1 indeed floated to sucrose gradient fractions 4/5 representing the plasma membrane fractions, as shown by a membrane-anchored form of Abp1 (myrAbp1#12) floating to the same fractions (Fig. 2A).

In vitro reconstitutions of Abp1/lipid interactions using liposomes and purified Abp1 proved that the associations are direct. GST-Abp1 and deletion mutants lacking the C-terminal domains (Abp1 ΔSH3 and Abp1 N-term) floated together with liposomes generated from Folch-fractions, i.e. to density gradient fraction 2 (Fig. 2B). Liposome floatation to fraction 2 was visualized by fluorescent lipids (not shown). In contrast, neither GST nor GST-fusion proteins containing the C-terminal Abp1 domains floated (Fig. 2B).

Interestingly, whereas Abp1 failed to bind to pure cholesterol-based liposomes supplemented with 20% phosphatidylcholine (PC), addition of 20% PIP₂ to cholesterol-based liposomes led to robust association. Likewise, addition of 20% PIP₃ or phosphatidylserine (PS) resulted in Abp1 association. Combinations of PS with PIP₂ or PIP₃ resulted in enhanced liposome binding of Abp1 (Fig. 2C). Abp1 thus exhibits preferences for certain lipids, e.g. PS, PIP₂ and PIP₃, all of which are enriched in the plasma membrane.

A comparison of the subcellular distribution of Abp1 and PIP₃ showed that both components exhibit a clear colocalization at the cell cortex of spreading cells (Fig. 2D, for merged image see supplementary material Fig. S4). Plotting Abp1 and PIP₃ signal intensities across the cells confirmed the good spatial overlap of Abp1 with PIP₃ especially at the cell cortex (Fig. 2E). These data suggest that the interaction of Abp1 with PIP₃, which we unravelled in our in vitro reconstitutions with purified components (Fig. 2C), also is of relevance in vivo.

Association with lipids enriched in the plasma membrane may represent an important aspect in Abp1's function. How can cell biological functions be revealed that specifically correspond to the membrane-associated subpool of Abp1? To address such functions, we designed a constitutive gain-of-function situation by generating Gal4-inducible transgenes encoding for GFP-tagged Abp1 proteins carrying a myristoylation motif (myrAbp1; see supplementary material Table S1 for an overview of fly strains used in this study). Density gradient analyses of protein extracts from transgenic larvae showed that myrAbp1-GFP indeed effectively accumulated in the plasma membrane fractions. About half of the total protein floated up to fraction 4/5 (Fig. 2A).

Ubiquitous expression of myrAbp1 was lethal (data not shown). These observations suggested that Abp1 is a potent cytoskeletal effector. Therefore, to genetically address the putative cytoskeletal functions of Abp1, a physiological context was required, which is not necessary for survival but is firmly established to depend on actin filament organization and dynamics. Since we observed a strong anti-Abp1 immunolabeling of imaginal discs (Fig. 1F), we analyzed the bristles developing from sensory organ precursor cells of these tissues. Bristle structures including the microchaetes and the much larger macrochaetes are actin-based and have proven extremely useful as read-out for cytoskeletal defects. Bristles contain non-Arp2/3-dependent actin bundles and dynamic actin structures, which depend on Arp2/3-mediated actin nucleation. Both actin structures together form a highly ordered array (Tilney and DeRosier, 2005). Distortions of the overall organization of the cortical actin cytoskeleton directly lead to improper bristle morphologies (Hudson and Cooley, 2002; Bogdan and Klämbt, 2003; Frank et al., 2006).

Targeted expression of myrAbp1 in sensory organ precursor cells by sca-Gal4 caused a striking bristle phenotype: microchaetes were split (Fig. 3). Detailed scanning electron microscopy (SEM) analyses showed that the phenotype ranged from short microchaetes with multiple splits (≥4 visible tips) to microchaetes with only one or two splits either occurring laterally or within the upper third of the microchaete (Fig. 3A,B). Transmission electron microscopy (TEM) analyses of sections close to the base of split microchaetes demonstrated that these split structures clearly had a common base with one nerve cell extension in the center (central area, not preserved) arguing against the possibility that the phenotype reflects bristle duplications due to impaired cell fate determination (Fig. 3C,D).

Quantitative analyses of two UAS-myrAbp1-GFP effector lines with crosses raised at 29°C to maximize expression levels showed that about 80% of all microchaete-type bristles on the notum of adult flies were short and split in line#12 (myrAbp1#12), whereas the more modest expressing line#1 (myrAbp1#1) (for western blot analysis see supplementary material Fig. S5) had 32% of the microchaetes split (Fig. 1E).

Microchaete splitting was accompanied with length reductions. Whereas unsplit myrAbp1#1 microchaetes were unchanged in length and unsplit line#12 microchaetes only showed a modest reduction in length, the split microchaetes of both lines showed a strong reduction in length. MyrAbp1#12, which exhibited the stronger split-bristle phenotype, also showed a stronger reduction in split microchaete length when compared to myrAbp1#1 (Fig. 3F).

In contrast to the length, the diameter of split bristles (measured at a height of 2 µm, i.e. at the base) did not show such drastic differences in myrAbp1-expressing flies when compared to wt bristles (Fig. 3G). The volumes of myr-Abp1 expressing microchaetes also did not differ significantly from wt (Fig. 3H). It thus seems that splitting of microchaetes occurs at the expense of their length.

A split-bristle phenotype has also been observed upon expression of myristoylated Kette (myrKette), the Drosophila ortholog of Nap1, a cytoskeletal effector working via the WASP pathway in split-bristle induction (Bogdan and Klämbt, 2003). Mammalian Abp1 associates with N-WASP (Pinyol et al., 2007; Haeckel et al., 2008). To genetically dissect the actin nucleation pathway Abp1 uses for split-bristle induction, as a prerequisite, we first analyzed heterozygosity for wasp¹, wasp³, scar^k03107 and scar^Δ37/+ as well as Arp2 RNAi in control experiments. None of these flies exhibited split bristles, i.e. all of them were suitable for genetic dissection of the molecular requirements of the myrAbp1-induced phenotype (supplementary material Fig. S6A,B).

Surprisingly, neither heterozygosity for wasp¹ nor for wasp³ suppressed the myrAbp1-induced phenotype (Fig. 4A,B). These data strongly suggest that, in contrast to Abp1 and N-WASP's tight cooperation in early mammalian neuromorphogenesis (Pinyol et al. 2007; Dharmalingam et al., 2009), Abp1 and WASP do not cooperate in bristle formation in flies.

To clarify, whether split-bristle formation by Abp1 reflects Arp2/3 complex functions at all, we next knocked down Arp2 by Gal4-induced expression of Arp2-specific dsRNA. Upon Arp2 RNAi, the frequency of split microchaetes induced by plasma membrane-targeted Abp1 was strongly suppressed (Fig. 4C). Quantitative analyses showed that the suppression was about 60% (Fig. 4D). Coexpression of UAS-GFP had no such suppressive effect (Fig. 4D). This strongly argued that the reduction of the myrAbp1 phenotype by Arp2 RNAi was not caused by a putative reduction of Gal4 availability (Fig. 4C,D). Instead, these results strongly suggested that myrAbp1 did employ cortical Arp2/3 complex functions for split-bristle formation but apparently used a pathway of Arp2/3 activation different from WASP.

We next assessed the possibility that Arp2/3 complex-mediated actin nucleation pathways distinct from WASP account for the observed bristle phenotype. As fly Abp1 has no recognizable interface for direct Arp2/3 complex association (Fig. 1A), we addressed Arp2/3 complex activation via Scar. We crossed scar^k03107 and scar^Δ37 flies with myrAbp1 flies. MyrAbp1-induced split microchaete formation was suppressed by both heterozygosity for scar^k03107 and scar^Δ37, respectively (Fig. 4C,D). Altogether, these data show that Scar and Arp2/3 complex functions are crucial for the myrAbp1-induced phenotype and suggest that Abp1 uses Scar as down-stream effector for Arp2/3 complex-mediated actin nucleation.

To map which Abp1 domains are required for its effects on bristle morphology, we generated transgenic flies expressing membrane-targeted deletion mutants of Abp1 and asked which of these would still give rise to the split-bristle phenotype. Attaching a combination of the lipid- and the F-actin-binding N-terminal domains of Abp1, i.e. the ADF-H and the helical domain, to the membrane did not lead to any split bristles despite expression levels being similar to those of myrAbp1 (Fig. 5A,B; supplementary material Fig. S7). Similarly, deleting just the SH3 domain (ΔSH3) also did not lead to any split bristles (Fig. 5A,B). This showed that the C-terminal part of Abp1 containing the SH3 domain was critical for phenotype induction (Fig. 5A,B).

We next asked whether the F-actin binding property of Abp1 is important for split-bristle induction or whether membrane attachment of the C-terminal domains of Abp1 alone would be sufficient for split-bristle formation. The myristoylated Abp1 C-terminus (myrAbp1 C-term) induced a phenotype highly analogous to the full-length protein (Fig. 5A). Quantitative examinations showed that similarly, a very high penetrance of the phenotype was obtained (Fig. 5B). More detailed analyses showed that the strong reduction of microchaete length accompanying microchaete splitting by the full-length protein was likewise observed upon plasma membrane targeting of the SH3-domain-containing Abp1 C-terminus. As for the full-length protein, dramatic effects on diameter and volume were not observed. The phenotype caused by membrane anchoring of the SH3 domain thus very well mirrored the full-length phenotype (Fig. 5A–E).

In contrast, expression of F-actin-binding myrAbp1 mutants lacking the SH3 domain failed to cause a split-bristle phenotype. Membrane targeting of the F-actin and lipid-binding N-terminal part of Abp1 only slightly increased microchaete length (Fig. 5A,C). Thus, membrane-targeting and C-terminal interactions of Abp1 are critical and sufficient for the induction of split microchaetes.

How might Abp1 employ Scar as effector leading to Arp2/3 complex activation? It seemed likely that a combination of Abp1's membrane association and an interaction with Scar mediated by the Abp1 C-terminal half may constitute the molecular mechanism. We thus tested the Abp1 SH3 domain for interaction with Scar. Affinity purifications with immobilized recombinant Abp1 SH3 domain demonstrated that endogenous Scar was efficiently and specifically coprecipitated from fly head extracts. Since a P523A mutant of the SH3 domain did not interact with Scar, the interaction seemed to be based on a classical SH3 domain and PxxP motif association (Fig. 5F).

Further analyses showed that binding is mediated by the proline-rich domain (PRD) of Scar. GFP-Scar PRD was specifically precipitated by the Abp1 SH3 domain (Fig. 5G). We next conducted endogenous coimmunoprecipitation experiments to evaluate whether Abp1 indeed associates with Scar in vivo. Our anti-Abp1 antibodies successfully immunoprecipitated endogenous Abp1 from Drosophila S2 cell lysates (Fig. 5H, upper panel). Further analyses of immuno-isolated protein complexes showed that endogenous Scar was specifically coimmunoprecipitated (Fig. 5H, lower panel). Additionally, immunofluorescence microscopy showed that both proteins colocalized at the leading edge of spreading S2 cells (Fig. 5I). Thus, protein complexes containing both Abp1 and Scar are formed both in vitro and in vivo.

Expression of myristoylated Kette led to split microchaetes by using WASP as effector (Bogdan and Klämbt, 2003). If myrAbp1 employs Scar and myrKette uses WASP to induce split microchaetes, together they can be expected to synergize. To test this prediction, we coexpressed myrAbp1 (line#1) and myrKette at 25°C. This temperature was low enough to yield only a very mild split-bristle phenotype for each individual line, so that a putative phenotype enhancement would be easily detectable (Fig. 5J,K). Strikingly, coexpression of myrAbp1 and myrKette led to a dramatic phenotype enhancement. Split-bristle frequency increased more than tenfold (Fig. 5J,K). This highlights the strong synergistic activity of Abp1 and Kette and their downstream effectors Scar and WASP.

To study abp1 loss-of-function phenotypes in Drosophila sensory organ formation, we analyzed the Abp1 expression in flies with PBac transposon insertions in the 5′-UTR and ∼600 bp upstream of the Abp1-encoding gene (locus CG10083; Fig. 6A), respectively. Whereas Abp1 expression in RB e01789 flies was not affected, the insertion WH f05024 diminished Abp1 to undetectable levels in immunoblots of larval body wall extracts (Fig. 6B). In head extracts, however, in which Abp1 expression was higher, modest levels of Abp1 were still detectable (not shown).

To obtain a null allele, we therefore deleted the presumptive promoter region and part of the 5′-UTR via FRT-mediated trans-chromosomal recombination between the two PBac insertions (Fig. 6A). The resulting allele, abp1^KO, lacked Abp1 in head extracts (data not shown) and body wall extracts (Fig. 6B). To minimize putative second-site mutational effects, a small deficiency, Df(3L)Exel6119, which covers the abp1 gene entirely (Fig. 6A), was employed for heteroallelic combinations. The resulting flies [abp1^KO/Df(3L)] also lacked Abp1, as demonstrated by both Western blot analysis and by a lack of anti-Abp1 immunostaining in imaginal wing discs (Fig. 6B,C).

Abp1 knockout had no striking effect on the viability or fertility of flies (data not shown) suggesting that, like in yeast and mammals, Abp1 functions are partially secured by functional redundancies (Drubin et al., 1988; Lila and Drubin, 1997; Connert et al., 2006).

We next analyzed abp1^KO/Df(3L) flies for defects in sensory organ formation using both light microscopy and EM (Figs. 6, 7; supplementary material Fig. S8A). Abp1 mutant flies showed kinked microchaetes. High-power magnifications of SEM images clearly resolved the longitudinal ridges and furrows of the microchaetes and showed that kinks in almost all cases occurred at the positions of joining ridges [abp1^KO/Df(3L); Fig. 6D]. Importantly, precise excision of the PBac insertion WH f05024 [abp1^WH-rev/Df(3L)] did not only restore Abp1 expression to wt levels (Fig. 6B) but also led to wt microchaetes (Fig. 6D; supplementary material Fig. S8A). The phenotypes of abp1^KO/Df(3L) flies that we observed are therefore specifically caused by abp1 disruption. Thus, Abp1 is required for proper formation of microchaetes.

Strikingly, a virtually identical low-penetrance kinked-bristle phenotype was described upon degradation of Scar due to loss of sra-1 or HSPC300 (Bogdan et al., 2004; Qurashi et al., 2007). Analyses of scar^k03107/+ and scar^Δ37/+ flies confirmed that reduced Scar functions lead to kinked microchaetes (supplementary material Fig. S8B). Consistently, also Scar RNAi flies showed a similar phenotype (Fig. 6D). In contrast, wasp¹/Df3450 did not show any distortions of microchaetes formed by these flies, i.e. did not phenocopy the abp1-knockout phenotype (Fig. 6D). Further control experiments showed that wasp and scar heterozygous backgrounds, respectively, did not lead to any obvious alterations of the abp1^KO/Df(3L)-induced kinked-microchaete phenotype (supplementary material Fig. S6C).

To ensure that the microchaete distortions observed in abp1- and scar-deficient flies indeed represented impairments of the functions of the actin nucleator Arp2/3 complex, we finally conducted comparative SEM analyses with Arp2 RNAi flies. Importantly, Arp2-knockdown flies also exhibited kinked microchaetes and thereby mirrored the abp1 and scar loss-of-function phenotypes (Fig. 6D).

Bristles are thought to be composed of a pile of segments of laterally aligned filament bundles and both the segments and also the filament bundles are joined and intercalated by fine actin structures generated by Arp2/3-mediated actin nucleation (Tilney and DeRosier, 2005; Hudson and Cooley, 2002; Frank et al., 2006). Macrochaete-type bristles are large enough to allow for phenotypical analyses of the organization of bristle substructures. SEM analysis of abp1^KO/Df(3L) dorso-central macrochaetes of the thorax provided further evidence for Abp1's specific control of Arp2/3-mediated actin nucleation. Whereas the generation of the actin bundles, which underlie the ridges of the bristles, was not affected by loss of Abp1, abp1 deficiency seemed to affect the intercalating structures because abp1-knockout bristles exhibited distorted arrangements of ridges resulting in a plethora of irregularly joined and fused ridges (Fig. 7A, arrows) and led to an increased number of ridges (Fig. 7A). Quantitative analyses showed that this increase in F-actin-bundle-containing ridges in the SEM images of the dorso-central macrochaetes was highly significantly different from wt macrochaetes (P<0.001) and corresponded to an increase in the density of ridges per macrochaete perimeter (Fig. 7B,C).

This phenotype was specifically due to loss of abp1 because the phenotype was not observable in revertants [abp1^WH-rev/Df(3L)]. Furthermore, the abp1-knockout phenotype was suppressed by about half by re-expression of Abp1 under sca-Gal4 control in the abp1-knockout background [abp1^KO/Df(3L)+Abp1] (Fig. 7A–C).

Importantly, a significant increase of the number of ridges as in abp1^KO/Df(3L) flies was also observable in high-power SEM analyses of dorso-central macrochaetes of both Scar RNAi and Arp2 RNAi flies (Fig. 7D–F). As for the kinked-microchaetes phenotype (supplementary material Fig. S8C), heterozygosity for scar^k03107 and wasp¹ did not modulate the Abp1 loss-of-function phenotype (supplementary material Fig. S9).

Together, our analyses of the loss-of-function phenotypes of Abp1, Scar and Arp2 further support the idea that Scar and Abp1 act in the same pathway and reveal that the Arp2/3 complex, Scar and Abp1 are crucial factors for the formation of intercalating actin structures indispensable for the overall organization of the actin bundles giving rise to the ridges of bristles and the proper formation of these sensory organs.

How is Abp1 linked to Scar functionally? Scar/WAVE protein stability is controlled by components of the Scar complex. Loss of any of these components leads to diminished levels of Scar and therefore gives rise to scar mutant phenotypes (Stradal and Scita, 2006; Takenawa and Suetsugu, 2007; Qurashi et al., 2007).

It seemed likely that the same was true for the newly identified Scar interaction partner Abp1. The reduction of Scar levels in larval body wall extracts from scar^k03107/+ and T80-Gal4+ScarRNAi flies, respectively, was readily detectable (Fig. 8A). However, no reduction of Scar levels in abp1^KO/Df(3L) flies was observed (Fig. 8A). Quantitative western blot analyses confirmed the observation that knockout of abp1 did not affect the stability of Scar. Scar protein expression levels in abp1^KO/Df(3L) flies were similar to wt (Fig. 8B). These results show that Abp1's role is not to stabilize Scar and that abp1-knockout phenotypes are not simply caused by a concomitant lack of Scar. Instead, our data strongly suggest that Abp1 is a crucial, lipid-binding coordinator and promoter of the physiological functions of Scar and the Arp2/3 complex at the cell cortex.

Membrane targeting is a crucial aspect in the spatial and temporal control of Arp2/3-mediated actin nucleation in various cellular and developmental processes. Whereas previous studies suggested that membrane association is mediated by Scar/WAVE and WASP proteins, our data show that phosphoinositides do not exclusively act on Arp2/3-activators. We reveal that Abp1 is physically and functionally interfacing with Scar and that a subpool of Abp1 is membrane-associated by interactions with PS, PIP₂ and PIP_3. This creates a layer of Abp1-mediated Arp2/3 regulation specifically at the cell cortex.

Our in vitro reconstitutions proved that Abp1's lipid interactions are direct, specific and strong enough to withstand floatation through sucrose. The in vivo relevance of this novel Abp1 function is supported by subcellular fractionations and by colocalization of Abp1 with PIP₃ especially at the plasma membrane of S2 cells.

The need for tight control of the membrane-associated pool of Abp1 is evident from the dramatic effects observed upon constitutive Abp1 presentation at the plasma membrane, such as reduced viability of myrAbp1-expressing larvae (unpublished observations) and split bristles. Our studies of the cytoskeletal functions required for microchaete formation clearly demonstrate a role of membrane-associated Abp1 in Arp2/3-mediated actin nucleation, as strong induction of actin nucleation specifically beneath the plasma membrane is thought to be the driving force behind bristle splitting (Bogdan and Klämbt, 2003).

Surprisingly, genetic dissection of the Abp1-induced split-bristle phenotype revealed that Abp1-induced cortical actin nucleation critically relies on membrane anchoring of the Abp1 SH3 domain and on the Arp2/3 complex but not on the Abp1 SH3 domain binding partner WASP. This finding was unexpected, because mammalian Abp1 was established to be an important positive regulator of N-WASP- and Arp2/3-mediated actin nucleation in vitro and in the morphogenesis of hippocampal neurons (Pinyol et al., 2007; Dharmalingam et al., 2009) as well as in formation and maturation of postsynapses (Haeckel et al., 2008). Our observations are, however, in line with the fact that myrKette induced a phenotype similar to constitutively membrane-anchored Abp1 and that myrKette strongly synergized with myrAbp1. These results suggested that Abp1 does not compete with Kette for WASP but uses an alternative activator of the Arp2/3 complex.

Our analyses revealed that Abp1 interfaces with the Scar pathway in proper bristle formation. Seven lines of evidence support our finding that Abp1 and Scar are allied in a close physical and functional interaction. First, genetic analyses suggest that Abp1 acts in an Arp2/3 complex-dependent pathway independent from WASP. Scar is the most prominent alternative activator of the Arp2/3 complex. Second, we observed that Abp1 was recruited to the leading edge of spreading S2 cells, where Scar controls actin nucleation (Rogers et al., 2003). Third, both Scar (Oikawa et al., 2004) and Abp1 associate with PIP₃. Fourth, coprecipitation of endogenous Scar with the Abp1 SH3 domain, coprecipitation of GFP-Scar PRD and coimmunoprecipitations of endogenous Abp1 with Scar from fly head extracts revealed that both proteins are associated in vitro and also in vivo and unravelled the Abp1 and Scar protein domains involved in the interaction. Fifth, genetic dissection of the Abp1 domain requirements for split-microchaete formation demonstrated that the Scar-binding Abp1 SH3 domain is crucial and sufficient for phenotype induction, whereas an excess of the F-actin-binding modules of Abp1 at the plasma membrane merely increased the length of microchaetes. The latter effect is in line with elongations of postsynaptic spines upon overexpression of Abp1 mutants containing the N-terminal F-actin binding domains (Haeckel et al., 2008). Sixth, the Abp1-induced split-bristle phenotype critically depended on Scar, as it was strongly suppressed by heterozygosity for scar^k03107. And finally, seventh, abp1^KO/Df(3L) flies show phenotypes, which are likewise observed upon reduction of Scar by three different means but not upon loss of WASP function. Abp1 knockout led to kinked microchaetes. Abp1 deficiency furthermore led to increased numbers and densities of macrochaete ridges. Again, this phenotype was also observed upon Scar RNAi. Importantly, both the kinked-microchaete phenotype and the macrochaete phenotypes observed upon knockout of abp1 were completely absent in revertant fly mutants [abp1^WH-rev/Df(3L)]. The phenotypes of abp1 knockout observed in bristle formation are thus specifically caused by abp1 disruption.

Strikingly, and in sharp contrast to the chaperoning components of the Scar-complex, the abp1 knockout phenotypes were not simply caused by loss of Abp1 leading to a concomitant reduction of Scar levels. Thus, Abp1 does not merely affect Scar stability. Instead, Abp1 seems to trigger and steer Scar functions at the cell cortex. Our analyses show that this function of Abp1 is crucial for the Scar- and Arp2/3-mediated actin nucleation pathway. The fact that the Abp1-Scar interaction we discovered indeed is part of Arp2/3-dependent cytoskeletal functions is firmly established by our genetic analyses. The split-bristle formation induced by myrAbp1 was not only dependent on Scar but was also dependent on Arp2, as shown by Arp2 RNAi. Furthermore, Arp2 RNAi phenocopied abp1 knockout in proper bristle development by leading to both kinked microchaetes as well as to an excess of macrochaete ridges.

The observation that the actin filaments of the bundles forming the ridges are not negatively affected in abp1 knockout flies is in line with these structures being generated by Arp2/3-independent actin nucleation, as shown by analyses of arp3^EP3640/+ macrochaetes (Hudson and Cooley, 2002; Frank et al., 2006) and by our comparative Arp2 RNAi studies. Abp1, Scar and Arp2/3-mediated actin filaments thus seem to play an important role in joining Arp2/3-independent F-actin bundles and in controlling their abundance.

Together, our finding that the F-actin-binding protein Abp1 interfaces with lipids enriched in the plasma membrane and employs the Arp2/3 complex activator Scar adds yet another layer of complexity to the scrupulous control of Arp2/3 complex mediated actin nucleation at the cell cortex. Our study demonstrates that this layer of control represented by Abp1 is crucial for controlling Scar- and Arp2/3-mediated actin nucleation in time and space and that this function of Abp1 is critical for proper sensory organ formation in flies.

Fly strains used are summarized in supplementary material Table S1. All crosses were performed at 25°C unless indicated otherwise. Canton S and w¹¹¹⁸ served as a wt control. Transgenic flies were generated by P-element mediated germline transformation of w¹¹¹⁸ flies (Spradling, 1986). Abp1^KO flies were generated by FRT-mediated recombination between piggyBac alleles WH^f05024 and RB^e01789 through heatshock-induced expression of Flp recombinase in heteroallelic flies (Parks et al., 2004). Precise excision of the piggyBac insertion WH^f05024 leading to abp1^WH-rev flies with restored Abp1 expression was achieved by crossing to a piggyBac-transposase-expressing stock (Bloomington#8285). Abp1 knockout flies and revertant mutants were evaluated by single-fly PCRs and anti-Abp1 western blotting.

Drosophila Abp1 cDNA was assembled from fragments amplified from a cDNA library and of EST-clone SD28142 using an internal XhoI site. For RNAi, a tail-to-tail tandem of bp 625–1593 and bp 1425–625 was cloned into pUAST (Brand and Perrimon, 1993). To generate pUAST-myr, a fragment encoding the N-myristoylation site of Scr64B was inserted by oligonucleotide annealing. Abp1 SH3^P523A was generated by PCR using the primers ggactgtttctggcaaactatg and catagtttgccagaaacagtcc. Variants of Abp1 and Scar-PRD (aa 457–614) were generated by PCR and cloned into pEGFP-N3 (Clontech), pUAST, pUAST-myr, pCMV-Tag3B (Stratagene), pGEX-5X-1 (Promega) and/or pMal-C2 (New England Biolabs). A Scar cDNA-plasmid was a generous gift of E. Schejter. Bacterial fusion proteins were expressed and purified according to published methods (Kessels and Qualmann, 2006).

Affinity-purified GST-Abp1 ΔSH3 (aa 1–475) was used to immunize rabbits and guinea pigs (Pineda Antikörper Service). Antisera were purified with MBP-Abp1 ΔH3, as described (Kessels et al., 2000). Other primary antibodies used include anti-Scar (Bogdan et al., 2005), monoclonal anti-actin (Sigma), anti-Dlg (Hybridoma Bank, University of Iowa) and anti-PIP₃ antibodies (MoBiTec; Chen et al., 2002; Kwon et al., 2010). Primary antibody incubations were performed overnight at 4°C or for 2 hours at room temperature. Secondary antibodies and phalloidin conjugated to Alexa 488 and Alexa 568, respectively, were from Molecular Probes. Cy3 and Cy5 conjugates were from Jackson Laboratories. DyLight800- and DyLight680-conjugated goat anti-rabbit, anti-mouse, anti-guineapig antibodies were from Pierce. Imaging was performed on a AxioObserver/ApoTome (Zeiss).

HEK293 and COS-7 cells were maintained, transfected and subjected to immunofluorescence analyses as described (Kessels and Qualmann, 2002). To visualize spreading HEK293 cells, cells were trypsinized, re-seeded and incubated for 90 minutes before fixation with 4% paraformaldehyde (PFA) in PBS. Drosophila S2 cells were propagated in Schneider's Drosophila medium supplemented with 10% FBS and antibiotics at 27°C. Cells were plated, pre-treated with 0.5 mg/ml concanavalinA for 30 minutes, fixed and processed for immunofluorescence as above.

Immunostaining of whole mount embryos was done according to Goldstein and Fyrberg (Goldstein and Fyrberg, 1994). Imaginal discs were dissected from 3rd instar larvae and fixed for 20 minutes in 4% PFA.

For protein extraction, 12 dissected larvae per genotype were homogenized in 60 µl buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% NP40, 0.5% deoxycholate, 1% SDS and Complete protease inhibitors. Homogenates were cleared from cuticle debris by centrifugation (5 minutes, 3000 g) and analyzed by SDS-PAGE, Coomassie staining and immunoblotting.

For fractionation, 30 body walls of 3rd instar larvae per genotype were lyzed for 20 minutes in 525 µl HEPES buffer (10 mM HEPES, pH 7.4, 1 mM EGTA, 0.1 mM MgCl₂, 150 mM NaCl, protease inhibitors) containing 1 mM DTT at 4°C. Density gradient centrifugation and sampling was performed as described (Dharmalingam et al., 2009; Schwintzer et al., 2011).

Quantitative immunoblot analyses were performed with fluorescent antibody conjugates using a LI-COR Odyssey detection system.

S2 cells were lyzed by repeated passage through 22-gauge needles. Fly heads were harvested by agitating approximately 50 shock-frozen flies per sample in a sieve. The material was homogenized in 0.5 ml HEPES buffer containing 1% Triton X-100 and incubated for 45 minutes at 4°C. After centrifugation (10 minutes, 16,000 g), supernatants were used for coimmunoprecipitation and pull-down assays, respectively, as described previously (Qualmann et al., 2004).

For coimmunoprecipitations, rabbit anti-Abp1 antibodies were immobilized on protein-A agarose (Santa Cruz). After incubation with S2 cell extracts, co-immunoprecipitated Scar was detected with guinea pig anti-Scar antibodies (Bogdan et al., 2005). Non-immune rabbit IgGs (Santa Cruz) served as a control.

F-actin cosedimentation assays were performed according to published methods (Kessels et al., 2000) with 2 µM GST-Abp1 fusion proteins and 8 µM actin.

Liposomes were either made from Folch fraction I or from purified lipids as described (Reeves and Dowben, 1969; Dharmalingam et al., 2009). For lipid binding assays, 1 mg/ml methanol solutions of lipids were dried and resuspended in 100 µl 20 mM HEPES, 100 mM KCl, 1 mM EDTA, pH 7.4. Thereafter, 0.3 M sucrose per 1 mM lipid was added and the mixture was incubated over night at 45°C. After brief vortexing, 1 mM liposomes were incubated with 1 µM protein (10 minutes, RT) and then mixed with 150 µl 75% sucrose. The mixture was overlaid with 200 µl 35% sucrose and 200 µl buffer and centrifuged at 200,000 g for 30 minutes at 4°C. Six gradient fractions of 100 µl were collected from top to bottom and analyzed by SDS-PAGE and immunoblotting.

Adult flies were dehydrated in an ethanol series (30%, 40%, 50%, 60%, 70%, 80%, 90%, 100%; 30–60 minutes each and 100% ethanol overnight), air dried and subjected to sputter coating with a gold film. SEM was performed using a LEO-1450 VP (Zeiss).

ScaGal4-driven myrAbp1#12 pupae and adult wt flies were fixed with glutaraldehyde (2% in 0.2 M phosphate buffer, pH 6.8, 1 hour, RT) and osmium tetroxide (1% in 0.2 M phosphate buffer, pH 6.8, 1 hour, RT). Sections from dorsal thoraxes were dehydrated in rising ethanol concentrations and infiltrated with araldite resin. Resin curing was performed for 48 hours at 60°C. The embedded samples were sectioned with a LKB 8800A Ultratome III (LKB Produkter AB) equipped with a diamond knife. Ultrathin sections were stained with uranyl acetate and lead citrate and examined in an EM 900 (Zeiss).

Microchaetes were quantitatively examined by measuring their length and their diameter at the base. Volumes in good approximation cone-shaped (unsplit) and cylindrical (split) microchaetes, respectively, were calculated using the width and length determinations.

Dorso-central macrochaete diameters were measured at the base (at a height of 2 µm), all ridges visible in SEM images (i.e. half perimeter) were counted and their density was calculated per 10 µm of the perimeter.

We thank R. Kaiser and C. Kämnitz for their help with SEM and TEM and I. Linke and M. Öhler for excellent technical support. We are grateful to A. Zellhof, C. Klämbt, S. Bogdan, E. Schejter and the VDRC and Bloomington stock centers for providing reagents.