Mouse ACF7 and Drosophila Short stop modulate filopodia formation and microtubule organisation during neuronal growth

Spectraplakins are multi-talented scaffolding molecules that are highly conserved across the animal kingdom. The family comprises mammalian Bpag1/dystonin and ACF7/MACF1, Drosophila Short stop (Shot), and Caenorhabditis elegans Vab-10 (Jefferson et al., 2004; Röper et al., 2002). Spectraplakins play crucial roles in a broad spectrum of cellular contexts, ranging from highly dynamic roles in development or wound healing to more structural functions in cell or tissue maintenance (Sonnenberg and Liem, 2007).

Mammalian BPAG1 and ACF7 are both strongly expressed in the CNS with partially distinct expression patterns (Bernier et al., 2000; Leung et al., 2001). BPAG1 has been shown to play an important role in neuronal maintenance (Sonnenberg and Liem, 2007; Young and Kothary, 2007). However, we have no knowledge at all of the neural function of ACF7. We hypothesised that ACF7 is more likely to be involved in developmental aspects in the nervous system, based on insights into the role of ACF7 in non-neuronal cells. Thus, ACF7 is known to localise Axin in the context of Wnt signalling during formation of the primitive streak, node and mesoderm (Chen et al., 2006), arrange polarity factors at the leading edge of migrating cells, and act as an actin-microtubule linker to coordinate microtubule dynamics in cell migration (Kodama et al., 2003). Whether such functions of ACF7 are likewise applicable to its potential role in the nervous system remains to be elucidated.

Drosophila Short stop (Shot) is an excellent model for the analysis of spectraplakin function. Domain compositions and existing isoforms of Shot are very similar to both ACF7 and BPAG1, and all three display a high degree of sequence similarity (Jefferson et al., 2004; Röper et al., 2002). Shot is the only spectraplakin in Drosophila and would therefore be expected to integrate potential homologous functions of both ACF7 and BPAG1. Accordingly, the range of its reported phenotypes is wide, and Shot plays structural roles and/or mediates signalling events in these contexts. Reported shot mutant phenotypes comprise structural and adhesion defects in the epidermis (Prokop et al., 1998; Röper and Brown, 2003), developmental aberrations of foregut and tracheal tubes (Fuss et al., 2004; Lee et al., 2003), and failed association of microtubules with the fusome during oogenesis (Röper and Brown, 2004).

Particularly intriguing are further roles of Shot during nervous system development (Lee et al., 2000; Lee et al., 2007; Prokop et al., 1998), which could represent a paradigm for the neural function of ACF7 in mammals. Thus, Shot promotes the growth of nerves, dendrites and synaptic terminals, and its function is required in sensory, motor and interneurons alike (Gao et al., 1999; Lee et al., 2000; Prokop et al., 1998; Reuter et al., 2003). Structure-function analyses of Shot carried out in the context of motor axon growth have revealed a requirement for actin-binding Calponin domains, the microtubule-binding Gas2 domain, and for the EF-hand motifs (putative Ca²⁺-binding and shown to bind the translational regulator Kra/eIF5C) (Lee et al., 2000; Lee et al., 2007). However, we have no insights into the subcellular processes regulated by Shot during neuronal growth.

To advance our understanding of spectraplakin function in the nervous system, we focussed on the role of ACF7 and Shot. To this end, we successfully established knockdown of ACF7 in neuronal mouse cultures, and capitalised on Drosophila primary cultures, which have only recently become available for the study of subcellular neuronal phenotypes (Matusek et al., 2008; Sánchez-Soriano et al., 2007) (and see supplementary material Fig. S1). We demonstrated that ACF7 and Shot display similar requirements during axonal extension and pinpointed two subcellular roles for ACF7 and Shot: first, a role in organising neuronal microtubules, which is essential for axon extension; and second, a role in regulating filopodia formation. Capitalising on the tractability of the Drosophila system, we were able to refine our insights into these subcellular roles. We demonstrated that both functions have very distinct domain requirements. Furthermore, we have provided the first mechanistic links by demonstrating that the role of Shot in filopodia formation occurs in genetic interaction with the putative translational regulator Krasavietz/eIF5C. The data presented here not only highlight the first known function for ACF7 in the nervous system, but also highlight the subcellular roles of both ACF7 and Shot in neuronal growth.

We started our investigation by establishing siRNA knockdown of ACF7 protein to study its function. To this end, we transfected siRNA directed against ACF7 into Neuro2A cells, a well-established neuronal cell line (Olmsted et al., 1970). Quantitative western blot analyses revealed a >75% reduction of protein content of a ∼600 kDa band detected by anti-ACF7 antibody (Fig. 1A). This band corresponds in size to the long Shot-LA isoform implicated in axon extension in Drosophila (Lee and Kolodziej, 2002). To measure potential effects of ACF7 knockdown on axon length, sham- and siRNA-transfected Neuro2A cells were serum-starved and treated with retinoic acid to induce neurite growth. ACF7 knockdown caused a highly significant ∼35% reduction in axonal length, when compared to control cells (Fig. 1B). To ensure results obtained were not due to off-target effects, we used a second independent ACF7 siRNA in parallel experiments, with a similar reduction in axonal length observed (Fig. 1B). To strengthen these findings, we extended our studies to mouse primary cortical neurons. Cortical neurons analysed at 5 days in vitro and three days after transfection with either ACF7 siRNA, displayed ∼25% shorter axons than control-transfected neurons (Fig. 1C).

We next tested whether a similar axon extension phenotype would be seen upon loss of Shot function in our recently established Drosophila primary neuronal cultures (Matusek et al., 2008). To this end, neurons were extracted from embryos that were either wild type or mutant for shot, carrying either the shot^sf20 or shot³ mutant alleles believed to abolish all Shot function (Lee et al., 2000; Prokop et al., 1998). In such cultures, both shot mutant alleles cause a significant reduction in axonal length (∼50%), as compared to wild-type neurons (Fig. 1D; illustrated for shot³ in Fig. 1F). Even after 3 days in culture [a stage when neurons have long undergone synaptic differentiation (Küppers-Munther et al., 2004)], a similar reduction in axonal length was observed (data not shown), indicating that the defect represents a true premature stall.

Taken together, both ACF7 and Shot are essential for proper axonal elongation in cultured neurons, providing a first indication that the function of spectraplakins in axon extension might be evolutionarily conserved.

In order to understand the role of ACF7 and Shot in axonal extension, we investigated their functions at the subcellular level. Both proteins are known to interact with F-actin and microtubules (Lee and Kolodziej, 2002; Leung et al., 1999), and we focussed our analyses on the organisation of these two cytoskeletal components.

We initiated these studies by analysing microtubule organisation in vertebrate neurons. In controls, a larger fraction of neurons displays microtubules that tend to be tightly bundled along the axon and appear relatively straight when splaying into the growth cone periphery. Knockdown of ACF7 in Neuro2A cells increased the percentage of axons with unbundled and looped microtubules by 22% (Fig. 2A-C, Table 1). The effect of ACF7 knockdown on microtubules was even stronger in mouse cortical neurons (40% increase, Table 1), where the siRNA-treated neurons had a strong tendency to display unbundled and heavily looped microtubules (Fig. 2D-F). Comparable results were found for shot mutant Drosophila primary neurons, which showed a 25-29% increase of neurons with unbundled microtubules (Table 1). Microtubules of such shot mutant neurons displayed irregular trajectories and frequent loops in axons and growth cones (Fig. 1F; Fig. 2H′).

Therefore, our data suggest that Shot and ACF7 both play essential roles in organising neuronal microtubules. All three culture systems used here show a clear correlation between microtubule organisation and axon length, suggesting that microtubular defects are the cause of impaired axon extension.

To pave the way towards a mechanistic understanding of the roles of ACF7 and Shot in microtubule organisation, we attempted to identify the protein domains required in this context. To this end, we carried out a structure-function study for Shot, capitalising on the genetic and experimental tractability of the Drosophila model system, which is likewise applicable to the primary Drosophila cell culture system used here.

As a prerequisite for our studies, we assessed the ability of GFP-tagged Shot (Shot::GFP) to rescue shot mutant phenotypes of primary neurons. We used one of the long isoforms, Shot-LA, which has been reported to rescue shot mutant axon extension phenotypes in vivo (Lee and Kolodziej, 2002). Using the Gal4-UAS system of targeted gene expression (Duffy, 2002), Shot::GFP was expressed in shot mutant embryos using a scabrous-Gal4 driver, which drives gene expression throughout the nervous system (UAS-shot::GFP/+ or Y; shot³/Df(2R)MK1, scabrous-Gal4). These embryos were used to generate neuronal cultures. In these cultures, almost all neurons displayed strong Shot::GFP expression, which localised along the entire length of axons at modest levels, but was strongly enriched at growth cones. Therefore, Shot::GFP was found in strategic positions from where it could directly bind to microtubules and influence their organisation (Fig. 3A,C,D). To consolidate this finding, we carried out staining against endogenous Shot in culture and visualised Shot in vivo. All these experiments revealed a similar distribution of Shot within growing neurons (details in Fig. 3). Importantly, Shot::GFP fully restored the microtubule unbundling and axonal length phenotypes (Fig. 4B; Table 1), an essential prerequisite for its use in structure-functional analyses.

To carry out structure-function analyses, we targeted derivatives of Shot::GFP to shot mutant embryos. These derivatives carried specific deletions of first Calponin, Gas2 or EF-hand domains, respectively (Fig. 4A). These domains were chosen as candidates to mediate Shot function in microtubule organisation because of their demonstrated role in nerve extension in vivo (Lee and Kolodziej, 2002). ΔGas2 (lacking the microtubule-binding Gas2 domain) failed to rescue the microtubule and axon length mutant phenotypes of shot, confirming that microtubule-association through the Gas2 domain is essential in this context (Fig. 4B; Table 1). In addition, ΔCalp1 (lacks F-actin-binding first Calponin domain) failed to rescue both phenotypes (Fig. 4B; Table 1). Because this result was less expected, we carried out additional analyses with primary neurons derived from shot^kakP2 mutant embryos, which fail to express the endogenous Shot isoforms containing the first Calponin domain (Bottenberg et al., 2009; Lee et al., 2000; Röper and Brown, 2003). Confirming our results with ΔCalp1, shot^kakP2 mutant neurons showed increased microtubule unbundling and shorter axons (Fig. 4B; Table 1). We therefore conclude that F-actin interaction of Shot is required for its function in this context. By contrast, ΔEF (lacks both EF-hand domains) could rescue both phenotypes (Fig. 4B; Table 1).

These results suggest that Shot acts as an actin-microtubule linker in organising neuronal microtubules, a role that is crucial for regulation of axon length. However, these results leave unexplained the requirement of EF-hand domains for the axonal extension in vivo that was reported previously (Lee and Kolodziej, 2002).

ACF7 and Shot harbour N-terminal F-actin-binding Calponin domains, and we have shown for Shot that these domains are crucial for its role in microtubule organisation. Vice versa, F-actin structures might be regulated through ACF7 and Shot function. In accordance with this notion, we find endogenous Shot and Shot::GFP to localise along filopodia of Drosophila growth cones in culture (Fig. 3A′,D,E′). As shown recently, Drosophila primary neurons are well suited for quantitative analyses of filopodia (Matusek et al., 2008), and we capitalised on this read-out for our studies of Shot. Such analyses demonstrated that growth cones of primary neurons obtained from the two independent loss-of-function mutant alleles of shot (shot^sf20 and shot³) showed a 40-45% reduction in filopodia number (Fig. 5C,D,F; see also Fig. 2H″). This phenotype was a direct consequence of loss of Shot function, because it was fully rescued by expression of Shot::GFP in shot mutant neurons (Fig. 5F).

We then tested whether similar phenotypes would also occur upon ACF7 knockdown in mouse neuronal cells. Unfortunately, the culturing conditions for primary cortical neurons (high cell density, low transfection rates and the long culturing period required to achieve knockdown of ACF7) did not allow us to quantify filopodia numbers. However, such counts were achievable in retinoic-acid-induced Neuro2A cells. When comparing growth cones of siRNA-treated and control Neuro2A cells, the number of filopodia was reduced, although this effect was far milder (22–25%) (Fig. 5A,B,E) than observed in shot mutant neurons.

Hence, we uncovered a second subcellular function of Shot and potentially ACF7 in the regulation of filopodia formation. Given the importance of F-actin in growth cones (Dent and Gertler, 2003), this role of ACF7 and Shot very probably relates to axonal growth regulation.

Capitalising on the genetic tractability of the Drosophila culture system, we also investigated the domain requirements of Shot in this subcellular context (Fig. 5F). To our surprise, ΔCalp1 fully rescued the filopodia phenotype, and this result was confirmed by cultured neurons derived from shot^kakP2 mutant neurons. Similarly, we could discard the requirement for the microtubule-binding domain Gas2 in filopodia formation, because rescue could also be achieved with ΔGas2. We also obtained full rescue of filopodia with a C-terminal construct of Shot (EGG), containing only EF-hand motifs, the Gas2 domain and the very C-terminus of Shot (Fig. 4A), thus excluding a requirement for the second Calponin domain, the Plakin domain and the Spectrin-repeat rod for filopodia formation. Interestingly, only a Shot construct lacking the EF-hand motifs failed to produce a full rescue of filopodia formation (Fig. 5F).

Taken together, our data obtained from analyses of Shot suggest that the novel function of ACF7 and Shot in filopodia formation does not depend on their actin-microtubule linker function. Instead, the EF-hand domains are required in this context, which could also explain their importance for axon extension observed in vivo (see Discussion).

In the context of axonal midline crossing in the Drosophila embryonic nervous system, the EF-hand motifs of Shot were shown to mediate genetic interaction with the putative translational regulator Krasavietz/eIF5C (Kra), and biochemical assays have suggested that this interaction is direct (Lee et al., 2007). Because we found the EF-hand motifs to be important for Shot-mediated filopodia formation, we tested whether such function also depends on an interaction with Kra.

To this end, we first investigated the effect of Kra deficiency on growth cones. Analyses of two independent loss-of-function mutant alleles of kra displayed normal appearance with respect to microtubule organisation (unlike shot), but a reduction in filopodia number comparable to shot³ and shot^sf20 (Fig. 6A,B,G). Immunocytochemistry of Drosophila primary neurons revealed that Kra is strongly expressed in cell bodies, and displays a punctate staining along axons and throughout growth cones and filopodia, and that this staining is strongly reduced in kra mutant neurons (Fig. 6C,E). Kra::HA targeted to cultured neurons displays a very similar pattern to that of endogenous Kra and localises to similar areas as Shot. Occasional dots of Kra::HA colocalise with Shot in double-stained neurons, supporting the hypothesis that both proteins might interact physically (Fig. 6F). To gain potential genetic support for this notion, we raised primary cultures from shot^+/–;kra^+/– trans-heterozygous mutant embryos (carrying one mutant copy of shot and kra, respectively). These neurons displayed a strong reduction in filopodia number, whereas control neurons heterozygous for only one of these genes displayed normal filopodia numbers (Fig. 6G).

Taken together, our localisation studies and genetic analyses together with existing biochemical data (Lee et al., 2007) strongly suggest that Shot interacts with Kra via its EF-hand domains in order to regulate filopodia formation. Given the fact that ACF7 also contains EF-hand motifs (Jefferson et al., 2004; Röper et al., 2002) and that there is a mammalian homologue for Kra (Lee et al., 2007), the same kind of regulation might also occur in mammalian growth cones.

We have uncovered the conserved subcellular roles of two spectraplakin homologues, murine ACF7 and Drosophila Shot, during neuronal growth. A neural role for ACF7 was anticipated by its embryonic expression pattern (Bernier et al., 2000; Leung et al., 2001), but our results identify the first functional role for ACF7 in the nervous system. For Shot, a role in axonal growth had been demonstrated previously (Lee et al., 2000; Lee and Luo, 1999; Prokop et al., 1998), but we had no insights into the subcellular processes in which it might function. We pinpoint two distinct subcellular roles conserved between ACF7 and Shot in the context of axonal growth. First, both regulate the organisation of neuronal microtubules, a phenomenon we linked to the ability of axons to extend at normal rates. Second, both regulate filopodia formation. These two roles relate to distinct molecular mechanisms, as indicated by the different domain requirements demonstrated using Drosophila Shot as a paradigm. Continuing on this path, we were able to pinpoint a mechanistic link for the role in filopodia formation by demonstrating a genetic interaction with the translational regulator Krasavietz/eIF5C.

The ability of ACF7 and Shot to regulate microtubules and filopodia in the context of neuronal growth emphasises their role as cytoskeletal regulators. Our insights now make it possible to integrate ACF7 and Shot into current models of neuronal growth and to compare their neuronal and non-neuronal functions. Current models of axonal growth suggest that the extension of axons is executed by the microtubule-related machinery, whereas the actin-related machinery is predominantly required to regulate the directionality of this growth (Dent and Gertler, 2003). Our data indicate that ACF7 and Shot modulate both these machineries (Fig. 7), as discussed below.

ACF7 and Shot promote axonal growth, and this property depends on their roles in organising neuronal microtubules. This causal relationship between microtubule organisation and axon length is demonstrated by taxol-mediated rescue of axon extension phenotypes in all three culture systems used (N.S.-S. and M.T., unpublished data). It is further supported by the fact that the same combination of domains (Calponin and Gas2) is required for Shot function in both microtubule coalescence and axon extension. We propose therefore that ACF7 and Shot are important mediators of the microtubule-related pushing force underlying axonal extension. Similar correlations between unbundling of microtubules and impaired axon extension have been suggested from work on retraction bulbs during regeneration and on factors such as Lis1, Dynein, Doublecortin and APC during development (Ahmad et al., 2006; Bielas et al., 2007; Ertürk et al., 2007; Grabham et al., 2007; Purro et al., 2008). Those studies do not only support our conclusion, but might provide some of the factors that cooperate with Shot and ACF7 in axon extension.

One potential model for the role of ACF7 and Shot in microtubule organisation is based on their microtubule-bundling and microtubule-stabilising activities. Thus, when C-terminal constructs of ACF7/MACF1 or of Shot were transfected into non-neuronal cells, they were shown to bundle microtubules and render them resistant to the microtubule-destabilising drug nocodazole, and the Gas2 domain was shown to be essential in this context (Lee and Kolodziej, 2002; Sun et al., 2001) (and our own observations). We could substantiate these findings through nocodazole treatment of wild-type and shot mutant primary Drosophila neurons, which demonstrated that endogenous Shot indeed mediates microtubule stability (N.S.-S., unpublished data). Vulnerability to nocodazol in shot mutant neurons could be rescued through targeted expression of ΔCalp1 but not ΔGas2 (N.S.-S., unpublished data). Therefore, in contrast to microtubule organisation and neurite length, which both depend on actin-microtubule linker activity of Shot, microtubule stabilisation essentially requires only the Gas2 domain (in agreement with the above-mentioned findings in non-neuronal cells).

Considering these arguments and facts, we believe that the Gas2-mediated bundling activity of Shot or ACF7 might contribute to their microtubule-organising activity, but this is an incomplete explanation for their roles in axon extension. Instead, we favour a model in which Shot primarily regulates microtubule dynamics through three possible mechanisms: (1) Microtubules in ACF7- or Shot-deficient neurons fail to grow on a straight path in axons and growth cones; (2) anterograde axonal transport of microtubules is impaired; and (3) microtubule polymerisation processes are less persistent. Impairment of any of these three mechanisms (in addition to the microtubule-stabilising functions of Shot downstream of these processes) could explain a decrease in the microtubule-dependent forces required for axon extension. The ∼600 kDa large isoforms of ACF7 and Shot required for axon extension would be long enough [several hundred nanometers (Röper et al., 2002)] to link via their N-termini to F-actin of the submembraneous axonal cortex, whilst simultaneously attaching to microtubules with their C-termini. Therefore, the actin cytoskeleton would serve as a track for elongating microtubules in mechanism 1 [as suggested for ACF7 in other cellular contexts (Kodama et al., 2003; Kodama et al., 2004)], it would provide the essential substrate for dynein-mediated anterograde transport of microtubule fragments in mechanism 2 (Myers et al., 2006), or it would trigger Shot and ACF7 to influence the composition of plus-end tracking proteins (+TIPs) to favour microtubule-polymerising activity in mechanism 3. Attractive in the context of the tracking and polymerisation models is the general assumption that ACF7 and Shot can interact with microtubule plus-end-binding proteins, such as EB1 (Lansbergen and Akhmanova, 2006; Subramanian et al., 2003).

We demonstrate a novel function for ACF7 and Shot in filopodia formation. Surprisingly, our analyses of Shot in this context show that this function does not involve the F-actin- or microtubule-binding domains but, instead, requires its EF-hand motifs. EF-hand domains are generally believed to bind calcium (Bhattacharya et al., 2004), and intracellular calcium levels are potent regulators of growth cone dynamics (Henley and Poo, 2004). Experiments addressing this possibility for Shot have so far been inconclusive (data not shown), and we do not rule out that Shot activity might potentially be regulated through the calcium-binding capabilities of its EF-hand motifs. However, there is clear biochemical proof that the EF-hand motifs of Shot can physically interact with the potential translational regulator Kra/eIF5C (Lee et al., 2007). Here we show that Kra indeed localises at the growth cone, i.e. in a position where it could cooperate with Shot in regulating the actin cytoskeleton. We propose that this interaction could lead to the regulation of local translation events to influence the subcellular concentration levels of actin regulators or actin itself, thus tipping the balance for or against the induction of new filopodia. Such a model would be consistent with the fact that growth cone guidance can be regulated through local translation events (Lin and Holt, 2007), and with the fact that filopodia formation is induced locally (Mattila and Lappalainen, 2008). Because EF-hand motifs are present in ACF7 (Röper et al., 2002) and a mammalian homologue exists for Kra (Lee et al., 2007), the function of Shot in the regulation of filopodia formation might be conserved.

The EF-hand-dependent promotion of filopodia through Shot is dispensable for growth cone advance and axon extension in culture. This contrasts with previous observations that the EF-hand motifs are required for the extension of motor axons in vivo (Lee and Kolodziej, 2002; Lee et al., 2007). We made similar observations for neurons lacking Profilin or treated with F-actin-destabilising drugs, two conditions that likewise affect filopodia formation (C.G.-P., N.S.-S. and A.P., unpublished data). In both cases, axons grow to normal length or even longer in culture, whereas in vivo they cause severe stall phenotypes (Kaufmann et al., 1998; Wills et al., 1999). These data, together with the findings for Shot, strongly suggest that defects in F-actin regulation do not affect the machinery of axon extension per se, but rather axonal pathfinding abilities. We hypothesise that an aberrant pathfinding machinery inhibits the axon extension machinery in the presence of certain extracellular signals (as is the case in vivo), whereas it permits advance at default rate in the absence of such signals (as is probably the case in our primary neuronal cultures: cells are treated with proteases, cultured on glass with no coating, and with a very low abundance of support cells). Assigning the Kra-related function of Shot in filopodia formation to a role in pathfinding, is in agreement with the fact that this interaction is required for correct axonal navigation at the CNS midline in vivo (Lee et al., 2007).

Taken together, the different requirements of ACF7 and Shot for the regulation of both microtubules and filopodia highlight their relevance for neuronal growth and make them important factors to be considered in future models of growth cone advance.

Fly strains used: shot³ (Lee et al., 2000), shot^SF20 (Prokop et al., 1998), shot^kakP2 (Gregory and Brown, 1998; Lee et al., 2000), Df(2R)MK1 (Strumpf and Volk, 1998), scabrous-Gal4 (Mlodzik et al., 1990), kra¹, kra², Uas-kra-HA (Lee et al., 2007), UAS-shot-LA-GFP (Shot::GFP), UAS-shot-LC-GFP (ΔCalp1), UAS-shot-LA-ΔGas2-GFP (ΔGas2), UAS-shot-LA-ΔEF-GFP (ΔEF) (Lee and Kolodziej, 2002) and UAS-EGG (Subramanian et al., 2003). All mutant alleles of shot were analysed in hemizygosis over the deficiency Df(2R)MK1, i.e. the mentioning of shot^XY throughout the text always refers to shot^XY/Df(2R)MK1.

Drosophila primary cell cultures were generated as described previously (Sánchez-Soriano et al., 2005). In brief, cells were removed with micromanipulator-attached capillaries from stage 11 embryos (6–7 hours after egg lay at 25°C) and transferred into dispersion medium [200 ml contained 30 ml HBSS (Gibco), 3 ml penicillin-streptomycin-solution (Gibco), 0.01 g phenylthio urea (Sigma), 170 ml distilled water, 0.5 mg/ml collagenase (Worthington, Cellsystems) and 2 mg dispase (Roche)]. Digestion of cells was stopped after 3 minutes through addition of cell culture medium; cells were centrifuged at 120 × g for 4 minutes, the supernatant removed and the final volume of cell culture medium added. Aliquots of 30-40 μl were transferred to flat-bottom wells in glass slides, which were sealed air-tight with cover slips. Cells were subsequently kept in standard Schneider's medium. Cells were grown directly on glass or on a coating containing 0.5 mg/ml concanavalin A (Sigma). For drug treatment, solutions of 20 μM nocodazole or 1 nM taxol (both Sigma) in Schneider's medium were prepared from stock solution in DMSO. For controls, equivalent concentrations of DMSO were diluted in Schneider's medium. Taxol was applied for 4 hours and nocodazol for 30 minutes before fixation.

Cultured Drosophila neurons were analysed 6 hours after plating. They were fixed (30 minutes in 4% paraformaldehyde in 0.05 M phosphate buffer, pH 7-7.2), then washed in PBS containing 0.1% Triton X-100 (PBT). Incubation with antibodies was performed in PBT without blocking reagents using anti-Shot [1:200, guinea pig (Strumpf and Volk, 1998)], anti-Fasciclin2 (1:20, mouse; DSHB), anti-tubulin (1:1000, mouse; Sigma), anti-GFP (1:10000, goat), anti-Kra [1:500, guinea pig (Lee et al., 2007)], anti-HA (1:100, rat; Roche), Cy5-conjucated anti-HRP (1:200; Jackson ImmunoResearch), and FITC-, Cy3- or Cy5-conjugated secondary antibodies (1:200, donkey; Jackson ImmunoResearch). Filamentous actin was stained with TRITC- and FITC-conjugated phalloidin (Sigma). Stained Drosophila neurons or embryos were mounted in Vecta-shield (Vector Labs).

The mouse neuroblastoma cell line Neuro2A (CCL-131, American Cell Type Culture Collection) was cultured in DMEM (Dulbecco's modified Eagle's medium) containing 10% fetal calf serum (FCS), 1% penicillin/streptomycin and 10 mM HEPES. Cells were differentiated by addition of 20 nM retinoic acid (Sigma) plus 0.1% FCS in DMEM over a 24-hour period prior to fixation or lysis (Ebneth et al., 1998).

Primary mouse cortical neurons were prepared as previously described (Dajas-Bailador et al., 2008). Briefly, cortices from E17 C57/b6 mice were dissected, gently triturated and plated onto poly-L-ornithine-coated culture plates in Neurobasal medium (Invitrogen) containing 5 mM glutamine and 2% B27 supplement (Invitrogen).

Two independent ACF7 siRNA oligos were designed using the siDESIGN Center (Dharmacon); ACF7 siRNA1 (nucleotides 10626–10644) 5′-CCTCAGAACTTTAGAACAA-3′, ACF7 siRNA 2 (nucleotides 14691-14709) 5′-GAACATAGACAGAGTTAAA-3′ (nucleotides numbered using GenBank AF150755 sequence). Control siRNA was purchased from Dharmacon.

Neuro2A cells were co-transfected with ACF7 siRNA (25 nM) and RISC-free siGLO red transfection reagent (25 nM; Dharmacon) using Lipofecatamine 2000 (Invitrogen). At 4 hours post-transfection, cells were sorted by flow cytometry for siGLO-red-positive cells and plated in six-well culture plates. For some microcsopy experiments, cells were not sorted but, instead, co-transfected with 25 nM ACF7 siRNA plus pmaxGFP (AMAXA, Cologne, Germany), and GFP expression was used as a marker of transfection. For microscopy, cells were transferred to poly-D-lysine-coated coverslips (5 K/cm²) 24 hours after transfection, induced with retinoic acid at 48 hours and fixed at 72 hours.

Primary cortical neurons were cultured for 24 hours before co-transfection with pmaxGFP (AMAXA) and ACF7 siRNA (50 nM) using Lipofecatmine 2000, as described previously (Dajas-Bailador et al., 2008). GFP expression in neurons was used as a marker of transfection. Fixation occurred 72 hours after transfection.

For taxol treatments, both cell types were kept in medium containing 1 nM taxol for 48 hours before fixation.

For western blot analysis, N2A cells in six-well plates were induced with retinoic acid 48 hours post-transfection and lysed at 72 hours. Cell lysates were prepared using 1% Triton, 1 mM sodium vanadate, 2 mM EDTA, protease inhibitor cocktail (Sigma), resolved by SDS-PAGE (3-8% Tris-acetate gradient gel; Invitrogen) and proteins transferred to PVDF membrane using the XCell blot system (Invitrogen). Membranes were incubated in blocking buffer (Sigma), followed by incubation with primary antibody (anti-ACF7 clone CU119, rabbit, 1:2000; kind gift from Ronald Liem) and anti-talin, (goat, 1:1000; Santa Cruz Biotechnology). Membranes were then incubated with secondary antibodies (1:5000) labelled with either Alexa-Flour 680 (Invitrogen) or IRDye-800 (Rockland) and membranes analysed using the Odyssey Infrared Imaging System (LI-COR).

At 72 hours post-transfection with siRNA, Neuro2A cells or primary mouse cortical neurons were fixed (using 4% paraformaldehyde, 0.25% glutaraldehyde, 0.1% Triton, 10 μM taxol in 60 mM PIPES, 25 mM HEPES, 10 mM EGTA, and 2 mM MgCl₂, pH 6.9), blocked using normal donkey serum (Jackson ImmunoResearch) and incubated with primary antibody directed against tubulin (1:500, Millipore) or GFP (1:500, Abcam) diluted in PBS containing 0.3% Triton. Cells were then incubated with FITC-, Cy3- or Cy5-conjugated secondary antibodies (1:200, donkey; Jackson ImmunoResearch). Coverslips were mounted onto slides using Prolong Gold anti-fade reagent (Invitrogen).

Most images were taken using an AxioCam camera mounted on an Olympus BX50WI microscope. Fig. 6F-F′″ were obtained on an Olympus IX71 microscope controlled by a Deltavision system (Applied Precision) and processed using the Enhanced Ratio deconvolution algorithm (10 cycles, medium noise filter). All axonal measurements were performed using ImageJ software. Axonal measurements were performed using ImageJ software. Statistics were carried out in Sigma Stat using a t-test or, if data failed the normality test, Mann-Whitney rank sum test. All data representations are as mean ± s.e.m.