Rac1-knock-out CGNs show defects in migration

Neuronal migration is one early polarized event during neuronal development. To test whether neuronal migration is impaired in the absence of Racs, we focused on CGNs that express only Rac1. We stained cerebellar sections with antibodies against the CGN markers NeuN and Pax-6 (Yamasaki et al., 2001; Weyer and Schilling, 2003). In the P18 wild-type cerebellum, NeuN-positive CGNs were almost exclusively present in the internal granule layer (IGL) (Fig. 1B). By contrast, in the Rac1-null cerebellum, we found a fivefold increase in the number of NeuN-positive cells in the ML compared with control (1.1 ± 0.2 in Rac1-null vs 0.2 ± 0.1 cells in wild type; average ± SEM/micrometers of ML from three animals; n = 3 sections/cerebellum; p < 0.05) (Fig. 1B, arrowheads). Similarly, Pax-6-positive cells were still present in the EGL and ML of the P18 Rac1^−/− cerebellum (data not shown).

In contrast to the aberrant localization of CGNs, Purkinje neurons, which express Rac3 (Bolis et al., 2003), formed their typical monolayer in the P18 Rac1-null cerebellum (Fig. 1C). We also examined the in vivo organization of Bergmann glia by immunocytochemistry using the GFAP, but found no obvious differences between wild-type and Rac1-knock-out cerebella (Fig. 1D). The majority of Rac1^−/− CGNs were in close proximity to GFAP-positive glia fibers (Fig. 1D, arrowheads).

To directly analyze inward migration of CGNs, we labeled dividing granule cell precursors by injecting BrdU in P7–P8 pups (Borghesani et al., 2002). At 24 h after injection, BrdU-labeled cells resided in the EGL in both the wild-type and Rac1^−/− cerebellum (Fig. 2A, arrowheads). At 96 h after injection, wild-type BrdU-labeled cells had reached the IGL, whereas many Rac1^−/− BrdU-labeled cells resided within the EGL and ML (Fig. 2A, arrowheads). These results indicate a defect in migration of CGNs in the absence of Rac1.

Open in a separate window

Figure 2.

Abnormal in vivo and in vitro migration of Rac1^−/− CGNs. A, Dividing cerebellar granule cell precursors were labeled by systemic injection of BrdU into P7 wild-type and Rac1-deficient mice. At 24 h after BrdU injection, the majority of BrdU-labeled granule cells are in the EGL in both the wild-type and Rac1^−/− cerebellum (arrowheads). At 96 h after injection, the majority of BrdU-labeled granule cells in the wild-type cerebellum have reached the IGL (arrowhead). By contrast, in the Rac1^−/− cerebellum, more BrdU-labeled granule cells (arrowheads) stayed in the EGL and ML. Scale bar, 50 μm. B, EGL explants from P5 wild-type and Rac1^−/− mice were stained with DAPI after 2 DIV. Rac1^−/− neurons migrate from the explant less efficiently than wild-type neurons. Scale bar, 200 μm.

Although we found Rac1-knock-out CGNs in close proximity with Bergmann glia fibers, we could not exclude the possibility that the loss of Rac1 interferes with the ability of neurons to associate and/or dissociate with the glia fibers, resulting in an impaired neuronal migration. To test whether the migration defect in the Rac1-null CGNs is cell autonomous, we studied their migration in cultured EGL microexplants (Fig. 2B). As reported previously (Nagata and Nakatsuji, 1990, 1994), we confirmed that cells migrating out of the explants are CGNs (data not shown). Their migration in microexplants occurs independently of Bergmann glia, making this system well suited to study cell-autonomous migratory cues (Nagata and Nakatsuji, 1990, 1994). After 2 DIV, we found more than a twofold decrease in the number of Rac1-knock-out granule cells migrating out of the explant compared with wild-type (913.8 ± 140 in Rac1-knock-out vs 2231 ± 322.1 cells in wild type; average ± SEM/square micrometers of explant surface from five independent cultures; n = 100 explants; p < 0.01) (Fig. 2B). Furthermore, Rac1-knock-out granule neurons migrated far shorter distances and distributed over smaller occupancy area compared with migrating wild-type neurons (0.2 ± 0.1 in Rac1-knock-out vs 0.5 ± 0.1 mm² in wild type; average ± SEM from five independent cultures; n = 100 explants; p < 0.001). Together, our data demonstrate that Rac1 is necessary for the migration of CGNs in a cell-autonomous fashion.

Rac1-knock-out CGNs show impaired axon growth

Axon growth and migration occur simultaneously during the development of CGNs (Hatten and Heintz, 1995; Yamasaki et al., 2001). In addition, defects in the axon formation often correlate with aberrant migration of CGNs (Tomoda et al., 1999; Yamasaki et al., 2001). During radial migration, CGNs extend two axons, the parallel fibers, that reside in the ML (Hatten et al., 1997) and can be visualized with an antibody against the cell adhesion molecule L1 (Fig. 3A) (Sakurai et al., 2001). In the Rac1^−/− cerebellum, only few L1-positive axons were observed (Fig. 3A, arrows). Moreover, axonal fibers in the ML visualized by neurofilament immunostaining were short and misoriented (Fig. 3B, arrows). To further analyze axonal tracts in vivo, we injected DiI in the wild-type and Rac1-knock-out cerebella (Fig. 3C). At 2 d after DiI injection, labeled axons of CGNs were visible in the wild-type (Fig. 3C, asterisks), but not in the Rac1-knock-out cerebellum. At 4 d after DiI injection, typical bipolar and parallel axonal tracts extended over distances >400 μm from the injection site in the wild-type cerebellum (Fig. 3C, asterisks). In contrast, no axonal labeling was observed in the Rac1-knock-out cerebellum. This suggests that Rac1 signaling is critically required during parallel fiber formation. Of note, Rac1-knock-out Purkinje cells, which express Rac3, do not show defects in extending their axon (Fig. 1C, arrowheads).

Open in a separate window

Figure 3.

Rac1^−/− neurons show deficient axon extension in vivo. A, P15 coronal sections from the middle part of the cerebellum immunostained with an antibody against L1 that visualizes parallel fibers in the ML. Images of boxed regions of the ML are depicted at higher magnification. L1 immunostaining reveals less densely packed parallel fibers in the ML of the Rac1-null cerebellum compared with wild type (arrows). Scale bars, 30 μm. B, P18 midsagittal sections immunostained with an antibody against neurofilament to visualize axons. Images of boxed regions of the ML are depicted at higher magnification. Long axonal fibers can be observed in the wild-type cerebellum and short and misrouted axons in the Rac1^−/− cerebellum (arrows). Scale bars, 30 μm. C, DiI-labeled parallel fibers in P15 coronal sections of wild-type and Rac1^−/− cerebella. At 2 d after DiI injection, parallel fiber labeling within the ML is visible in the wild-type cerebellum (asterisks), but not in the Rac1^−/− cerebellum. At 4 d after injection, long parallel fibers project from both sides of the DiI injection place in the wild-type cerebellum, but this labeling is absent in the Rac1^−/− cerebellum. The pial surface of the cerebellum is indicated by arrowheads. Note that the intensity of DiI labeling at the injection site in wild-type and Rac1^−/− cerebellum is comparable (2 d after DiI injection). Scale bar, 100 μm.

To assess whether Rac1 affects axon growth in a cell-autonomous manner, when separated from spatial cues, we examined axon formation in dissociated CGNs, an established model of neuronal polarity (Hatten and Heintz, 1995; Powell et al., 1997; Zmuda and Rivas, 1998). Cultured CGNs undergo a set of well described morphological changes that resemble those occurring in vivo (Powell et al., 1997). Long neurites that develop after 2 DIV correspond to axons and can be morphologically distinguished from dendrites, whose development initiates at later stages (5–6 DIV). We found a threefold decrease in the number of Rac1^−/− neurons that extended axons >100 μm after 2 DIV (Fig. 4A,B). Additionally, we assessed the ability of Rac1-knock-out CGNs to form axons. To this end, we used the Tau-1 antibody that recognizes a dephosphorylated Tau epitope that is axon specific (Mandell and Banker, 1996). We found a 50% reduced ability of Rac1-knock-out CGNs to form axons compared with wild type (supplemental Fig. S3, available at www.jneurosci.org as supplemental material). These data suggest a role of Rac1 in axon growth. Interestingly, Rac1^−/− neurons showed an increase in the average number of neurites per cell. Whereas >80% of wild-type neurons had one to two neurites per cell, the majority of Rac1^−/− neurons contained three or more neurites (Fig. 4A,C). Nevertheless, the average length of all neurites at 2 DIV decreased from 147.8 ± 5.1 μm in wild-type neurons to 113.4 ± 1.4 μm in Rac1-null CGNs (average ± SEM from three independent cultures; n = 100 cells/culture; p < 0,05). Time-lapse videomicroscopy analysis revealed that Rac1-null neurons initiated neurite formation at the same time as wild-type neurons (supplemental Fig. S4A,B, Videos 1, 2, available at www.jneurosci.org as supplemental material). We observed a constant growth rate of both wild-type and Rac1-null neurons during the first 24 h. Wild-type neurons exhibited a growth rate of 1.5 μm/h. Instead, Rac1^−/− neurons elongated their longest neurite slowly, with a growth rate of 0.8 μm/h. After 24 h, wild-type neurons continued extending their axons further (supplemental Fig. S4B, available at www.jneurosci.org as supplemental material), whereas Rac1^−/− neurons formed additional neurites (supplemental Fig. S4C, available at www.jneurosci.org as supplemental material).

Open in a separate window

Figure 4.

Reduced length and an increased number of neurites in cultured Rac1^−/− CGNs can be reversed by overexpressing Rac1. A, Dissociated CGNs were immunostained after 2 DIV with an antibody against Tuj1. Rac1^−/− CGNs display several shorter Tuj1-positive neurites, whereas wild-type CGNs typically exhibit one or two longer Tuj1-positive neurites. Scale bar, 25 μm. B, Quantification of the length of the longest Tuj1-stained neurite. The values represent averages ± SEM from three independent cultures (n = 100 cells/culture; ***p < 0.001; **p < 0.01). C, Quantification of the neurite number. The values represent averages ± SEM from three independent cultures (n = 100 cells/culture; *p < 0.05). D, Quantification of the length of the longest neurite in CGNs cotransfected with pEYFP-C1 and pcDNA3.1 (ctrl) or with pEYFP-C1 and pRac1 (pRac1) at 3 DIV. The values represent averages ± SEM from four independent cultures (n = 100 cells/culture; **p < 0.01). E, Quantification of the neurite number in CGNs cotransfected with pEYFP-C1 and pcDNA3.1 (ctrl) or with pEYFP-C1 and pRac1 (pRac1) at 3 DIV. The values represent averages ± SEM from three independent cultures (n = 100 cells/culture; *p < 0.05).

To examine whether the defect in axon outgrowth was specifically caused by Rac1 ablation, we coexpressed wild-type Rac1 and YFP in Rac1^−/− CGNs. The length of the longest process increased by >50% on overexpression of Rac1 in Rac1^−/− neurons compared with Rac1^−/− neurons transfected with the control plasmid (Fig. 4D). On Rac1 overexpression, we could no longer observe a significant reduction of neurite length between wild-type and Rac1^−/− CGNs (Fig. 4D). Additionally, overexpression of Rac1 in Rac1-deficient neurons decreased the average number of neurites per cell (Fig. 4E). The majority of Rac1^−/− neurons transfected with Rac1 contained one to two neurites. Overexpression of Rac1 in wild-type neurons affected neither their neurite length nor their number of neurites (Fig. 4D,E). Similar to Rac1, we found that overexpressing Rac3 in Rac1-null neurons significantly increased the length of the longest process (supplemental Fig. S5C, available at www.jneurosci.org as supplemental material) and reduced neurite number (data not shown), whereas it had no effect in wild-type neurons. This suggests that Rac3 and Rac1 have overlapping functions in axon growth.

Rac1^−/− neurons lack lamellipodia

To investigate whether the defect in axon formation is caused by aberrant regulation of the actin cytoskeleton, we stained cultured CGNs from wild-type and Rac1-knock-out animals with fluorescently labeled phalloidin. Lamellipodia were visualized using an antibody against cortactin (Wu and Parsons, 1993). Wild-type growth cones contained both long and thin actin protrusions, filopodia (Fig. 5A, arrowhead), and fanlike lamellipodia (Fig. 5A, arrow). By contrast, Rac1-null neurons were typically devoid of lamellipodia (Fig. 5A). Quantification of the proportion of cells bearing lamellipodia showed a 70% reduction in the Rac1-null CGNs compared with wild type (21.4 ± 4.1 in Rac1-null vs 71 ± 1.7% in wild type; average ± SEM from three independent cultures; n = 100 cells/culture; p < 0.01). Consistent with the absence of lamellipodia, we found that the growth cone surface of the Rac1-deficient CGNs reduced to 44.3 ± 1.6% of the wild-type growth cone surface (average ± SEM from three independent cultures; n = 20 cells/culture; p < 0.001). Additionally, Rac1-null neurites typically showed actin-rich filopodia at their ends (Fig. 5A, arrowheads). Examination of cultured CGNs by videomicroscopy revealed dynamic filopodia in wild-type and Rac1-knock-out neurons (Fig. 5B; supplemental Videos 3, 4, available at www.jneurosci.org as supplemental material), but the specific absence of lamellipodia dynamics in Rac1-null CGNs (supplemental Video 4, available at www.jneurosci.org as supplemental material). We also observed that the relative amount of F-actin in the Rac1-knock-out growth cones reduced to 49 ± 1% of the wild type (average ± SEM from four independent cultures; n = 100 cells/culture; p < 0.01). Thus, our data demonstrate that Rac1 is essential for lamellipodia formation and growth cone dynamics of CGNs.

Open in a separate window

Figure 5.

Rac1^−/− CGNs lack lamellipodia and show reduced actin dynamics. A, Dissociated CGNs at 1.5 DIV were costained with rhodamine–phalloidin to visualize actin filaments and with an antibody against cortactin to visualize lamellipodia. Images from wild-type and two different Rac1^−/− neurons are shown. Cortactin and phalloidin staining of a wild-type neuron reveal a lamellipodium (arrow) and a filopodium (arrowhead), whereas Rac1^−/− neurons only have filopodia at their neurite tips (arrowheads) but lack lamellipodia. Scale bar, 10 μm. B, Time-lapse videomicroscopy of representative wild-type and Rac1^−/− CGNs at 1.5 DIV. Three frames at 1 min intervals are shown. A dynamic lamellipodium of a wild-type growth cone is indicated by an arrow. A filopodium of a Rac1^−/− growth cone is indicated by an arrowhead. Scale bar, 10 μm.

Modulating the actin cytoskeleton increases the ability of Rac1-knock-out neurons to form axons

Actin dynamics orchestrate neuronal polarization and axon growth. In fact, depolymerization of the actin cytoskeleton induces formation of multiple axons in dissociated hippocampal neurons (Bradke and Dotti, 1999, 2000; Kunda et al., 2001; Schwamborn and Püschel, 2004). We wondered whether the observed changes in the actin cytoskeleton of Rac1-null neurons are underlying defects in neuronal polarization. We therefore pharmacologically depolymerized the actin cytoskeleton and assessed axon growth in Rac1^−/− CGNs (Fig. 6A–C). Cytochalasin D treatment of Rac1^−/− CGNs increased by 2.4-fold the length of the longest neurite (Fig. 6A,C) and by 2-fold the ability to form Tau-1-positive axons (Fig. 6A,B). Cytochalasin D treatment of Rac1^−/− CGNs also decreased their neurite number (Fig. 6D). In contrast to hippocampal neurons, we did not observe the formation of multiple axons on treatment of CGNs with cytochalasin D as previously reported (Zmuda and Rivas, 2000). These differences may stem from different intrinsic developmental properties of these two distinct neuronal populations. Together, our work reveals that depolymerization of the actin cytoskeleton by cytochalasin D abrogated the difference in axon growth and neurite number between Rac1-knock-out and wild-type CGNs.

Open in a separate window

Figure 6.

Actin depolymerization reverses Rac1-mediated defects in axon growth. A, Dissociated CGNs were treated with cytochalasin D (CytoD) or control treated with DMSO 4 h after plating. Cells were fixed at 2 DIV and immunostained for the axonal marker Tau-1. CytoD-treated Rac1^−/− neurons extend longer, Tau-1-positive axons bearing a typical proximodistal staining gradient. DMSO-treated Rac1^−/− neurons extend several shorter, Tau-1-negative neurites. Cytochalasin D treatment abrogates the difference in axon formation between Rac1^−/− and wild-type CGNs. Scale bar, 30 μm. B, Quantification of axon formation. Cells, bearing at least one Tau-1-positive axon (proximodistal staining gradient), were scored as positive. The values represent averages ± SEM from three independent cultures (n = 100 cells/culture; **p < 0.01). C, D, Dissociated CGNs were treated with CytoD or DMSO 4 h after plating. Cells were fixed at 1 DIV and visualized by immunostaining for Tuj1. Quantification of the length of the longest Tuj1-stained neurite (C) or neurite number (D). The values represent averages ± SEM from three independent cultures (n = 100 cells/culture; *p < 0.05; ***p < 0.001).

Modest stabilization of microtubules also induces the growth of multiple axons in dissociated hippocampal neurons (Gomis-Rüth et al., 2008; Witte et al., 2008) and enhances neurite outgrowth of CGNs when plated on the inhibitory substrate myelin (Ertürk et al., 2007). However, treating Rac1^−/− CGNs with the microtubule-stabilizing drug Taxol did not affect their neurite length (data not shown). Therefore, our experiments suggest that Rac1 predominantly regulates neurite outgrowth via modulating the organization of the actin cytoskeleton.

Rac1-knock-out cerebella show a decrease in PAK1 phosphorylation

Since expression of constitutively active Rac in fibroblasts causes a decrease in Rho activity (Sander et al., 1999), we examined whether there was a cross talk between Rac1 and RhoA in the cerebellum. However, neither total RhoA expression levels (data not shown) nor the levels of active RhoA in Rac1^−/− cerebellar extracts were altered (Fig. 7A). We also did not find any change in Cdc42 activation in Rac1-null compared with wild-type cerebellar extracts (Fig. 7A). This indicates that defects in the Rac1^−/− cerebellum are not mediated through RhoA or Cdc42 signaling.

Open in a separate window

Figure 7.

Rac1 deletion results in a reduced activity of PAK but does not change the activity of cofilin. A, ELISA-based assay for RhoA and Cdc42 activation. The results are expressed as ratios between A₄₉₀ in wild-type and Rac1^−/− extracts. The values represent averages ± SEM from three animals. There is no significant difference between the activity of RhoA or Cdc42 in Rac1^−/− and wild-type cerebella. Cortical extracts were used as control. B, Analysis of phospho-PAK1 (p-PAK1) and PAK1 protein levels from P9 cortices and cerebella. Rac1^−/− cerebellar extracts show reduced levels of p-PAK1 compared with wild-type extracts. C, Dissociated CGNs at 2 DIV were immunostained with an antibody against p-PAK1/2. A decreased level of p-PAK1/2 can be observed in neurites of Rac1^−/− CGNs. By contrast, higher levels of p-PAK1/2 accumulate in a neurite of a wild-type CGN. Scale bar, 15 μm. D, Quantification of the length of the longest neurite in CGNs cotransfected with pEYFP-C1 and pcDNA3.1/myc-His A (ctrl) or pEYFP-C1 and pPAKT423E (pPAKT423E) at 3 DIV. The values represent averages ± SEM from three independent cultures (n = 100 cells/culture; *p < 0.05; **p < 0.01). Overexpression of constitutively active PAK1 (PAKT423E) in Rac1^−/− CGNs does not increase neurite length. E, Analysis of phosphorylated-cofilin (p-cofilin) and cofilin protein levels from P9 cortices and cerebella. Wild-type and Rac1^−/− extracts show comparable levels of p-cofilin.

We assessed potential effectors of Rac1 that regulate actin dynamics. Previous work showed that PAK1 regulates axon formation in hippocampal neurons (Jacobs et al., 2007). We observed a decrease in the phosphorylation of PAK1 at Ser^198/203 in Rac1^−/− cerebellar extracts (Fig. 7B), indicating a reduced catalytic activity of PAK1 (Shamah et al., 2001). The total PAK1 protein levels remained constant (Fig. 7B). Decrease in the phosphorylation of PAK1 was mainly attributable to decreased levels of phosphorylated PAK in neurites of Rac1^−/− CGNs (Fig. 7C).

We attempted to rescue the growth defect of Rac1-knock-out neurons by manipulating PAK1. Transfection of a phosphomimetic PAK1 mutant (PAKT423E) caused no increase in neurite length (Fig. 7D) or number (data not shown) in Rac1-knock-out neurons, suggesting that activating PAK1 alone is not sufficient to rescue axon formation. It has been shown that membrane-targeting and kinase activity of PAK1 are required for axon formation in hippocampal neurons (Jacobs et al., 2007). However, overexpression of the membrane-targeted PAK1 construct (Pak1-CAAX) also did not rescue neurite length or axon formation in Rac1-knock-out neurons (supplemental Fig. S5D, available at www.jneurosci.org as supplemental material) (data not shown). In another study, it has been shown that overexpression of the kinase-dead, membrane-targeted form of PAK1 (PakR299-CAAX) induces neurite outgrowth in PC12 cells, suggesting that the effect of PAK1 on neurite outgrowth is independent of its kinase activity but requires membrane localization (Daniels et al., 1998). We therefore also overexpressed a PakR299-CAAX in Rac1^−/− CGNs, but failed to detect an increase in neurite length or axon formation (supplemental Fig. S5E, available at www.jneurosci.org as supplemental material) (data not shown).

We also analyzed the activity of the PAK effector cofilin, an F-actin depolymerizing and severing factor that affects axon formation in hippocampal neurons (Garvalov et al., 2007; Jacobs et al., 2007) but found no detectable differences in phosphorylation of cofilin between wild-type and Rac1-knock-out cerebellar extracts (Fig. 7E). Together, our data suggest that Rac1 regulates axon growth through pathways other than PAK-cofilin.

Rac1-knock-out neurons mislocalize the WAVE complex from the growth cone plasma membrane

Overexpression studies in fibroblasts and melanoma cells have shown that the WAVE complex regulates lamellipodia formation and that Rac1 is necessary for its recruitment to the plasma membrane (Steffen et al., 2004). The WAVE complex, comprising WAVE, Sra-1, Nck-associated protein 1 (Nap1), Abelson-interacting protein 1 (Abi1), and HSPC300, has been implicated in axon development in vertebrate and invertebrate neurons (Hummel et al., 2000; Bogdan et al., 2004; Kawano et al., 2005; Yokota et al., 2007). We therefore assessed whether WAVE signaling is disturbed in Rac1-knock-out neurons. First, we investigated protein levels of the WAVE complex components Sra-1, Nap1, and WAVE, but could not detect any differences between wild-type and Rac1-knock-out brains (Fig. 8D). As WAVE recruitment to the plasma membrane is crucial for its signaling (Steffen et al., 2004; Stradal et al., 2004), we analyzed the intracellular localization of WAVE protein using a pan-WAVE antibody. In wild-type neurons, WAVE localized to the plasma membrane of lamellipodial protrusions (Fig. 8A,B). By contrast, in Rac1-knock-out neurons, WAVE was absent from the plasma membrane (Fig. 8A,B). As the vast majority of Rac1^−/− neurons contained no lamellipodia (Fig. 5B), we selected for the rare Rac1^−/− neurons that still contained some lamellipodia-like protrusions. Even in those neurons, WAVE, as well as other components of the WAVE complex, including Abi1, Sra-1 (Fig. 8A–C), and Nap1 (data not shown), failed to localize to the plasma membrane of the growth cone. Only 20–30% of the Rac1-knock-out neurons containing some lamellipodia remnants showed Abi1, Sra-1, and WAVE localization at the plasma membrane of the growth cone (Fig. 8C). These results suggest that Rac1 regulates the WAVE complex activity by localizing it to the plasma membrane of neuronal growth cones.

Open in a separate window

Figure 8.

Rac1^−/− neurons mislocalize WAVE from the growth cone plasma membrane. A, Dissociated CGNs at 1 DIV were costained with rhodamine–phalloidin and an antibody against WAVE (pan-WAVE). Rac1^−/− neurons that still contained some lamellipodia-like protrusions were examined for WAVE localization. Scale bar, 10 μm. B, Merged images of boxed regions in A are depicted at higher magnification. WAVE is absent from the edges of Rac1^−/− growth cones, whereas in wild-type neurons WAVE predominantly localizes to membranes of lamellipodial protrusions (arrowhead). Scale bar, 5 μm. C, Quantification of plasma membrane localization of Abi1, Sra-1, and WAVE. CGNs, showing plasma membrane localization of the respective protein in at least one growth cone, were scored as positive. The values represent averages ± SEM from three independent cultures (n = 100 cells/culture; *p < 0.05; **p < 0.01; ***p < 0.001). D, Mislocalization of the WAVE complex in the Rac1^−/− neurons is not caused by protein degradation. Western blot analysis of P9 cortical and cerebellar extracts reveals no difference in the protein amounts of the respective WAVE complex components.

Neuronal migration requires Rac1 function

Rac1 was proposed to regulate cell migration (Ridley, 2001a). However, this role has been challenged in recent in vivo studies (Bokoch, 2005; Vidali et al., 2006; Wheeler et al., 2006; Kassai et al., 2008). For example, migration of cortical neurons in cortex-restricted Rac1-knock-out mice is not impaired (Kassai et al., 2008). Here, we showed that Rac1 signaling is essential for efficient migration of one of the largest population of neurons in the brain, CGNs. Although one interpretation could be that Rac1 regulates migration in a neuron-specific manner, we favor the hypothesis that Rac3 functions in a redundant fashion with Rac1. Hence, the role of Rac1 in the cortex may be masked because of the presence of Rac3. In line with this idea, we found that Rac1-deficient Purkinje cells, which express Rac3 (Bolis et al., 2003), were able to form their typical monolayer.

Apart from regulating neuronal migration, we found that Rac1 was important for neuronal survival. Thus, our in vivo data support previous cell culture overexpression studies (Linseman et al., 2001; Le et al., 2005) and reveal antiapoptotic effects of Rac1 signaling in the cerebellum. This particular role of Rac1 may be more widespread in the nervous system because Rac1 deficiency in the forebrain also results in an increased rate of apoptosis (Chen et al., 2009). However, this study revealed a positive role of Rac1 in proliferation, whereas our study shows that Rac1 is dispensable for proliferation during cerebellar development. Together, this suggests that regulation of cell proliferation by Rac1 may be more cell type specific.

Rac1 ablation from the nervous system impairs axon growth

Previous in vitro studies suggested Rac1 as both a positive and a negative regulator of axonogenesis (Govek et al., 2005). These discrepancies may stem from differences in the cellular systems analyzed or may be attributable to the fact that most of the previous studies used dominant-negative and constitutively active mutants, which can exhibit nonspecific or nonphysiological effects resulting from perturbation of multiple Rho GTPases (Wang and Zheng, 2007).

Analysis in mice have shown that Rac1-deficient cortical neurons exhibit axon guidance defects and fail to cross the midline. However, axon growth defects could not be observed (Chen et al., 2007; Kassai et al., 2008). One potential reason for the normal polarization of the Rac1-ablated cortical neurons is that Rac3 may compensate for some Rac1-dependent functions (Bolis et al., 2003; Corbetta et al., 2009). Consistent with this view, knocking out Rac3 alone does not lead to defects in axon development (Corbetta et al., 2005; Gualdoni et al., 2007). However, downregulation of Rac1, in the Rac3-knock-out neurons, revealed changes in dendrite development and spine formation (Gualdoni et al., 2007; Corbetta et al., 2009). It is therefore possible that, in contrast to CGNs, axon formation in hippocampal neurons occurs independently of Rac1. Alternatively, some residual Rac1 activity could have still been present during axonal formation that occurs early during development (Bradke and Dotti, 2000).

Our study clearly identifies Rac1 as a positive physiological regulator of axon growth in the mammalian nervous system. In addition to the in vivo defect in axon formation in the Rac1-knock-out cerebellum, cultured Rac1^−/− CGNs are disrupted in their polarization and show markedly reduced axon growth, indicating that this phenotype is cell intrinsic.

Rac1 regulates axon formation through the actin cytoskeleton

Axon outgrowth relies on coordinated regulation of the actin cytoskeleton and microtubule dynamics (da Silva and Dotti, 2002; Dent and Gertler, 2003; Witte and Bradke, 2008). Previous studies using mutant Rho-GTPases indicated that Rac1 may regulate both actin and microtubule dynamics (Kunda et al., 2001; Ng and Luo, 2004; Watabe-Uchida et al., 2006a,b). However, although we did not study microtubule dynamics in detail, staining of the microtubules did not reveal any obvious defects in the microtubule organization, and microtubule stabilization did not rescue axon formation in the Rac1-knock-out CGNs (data not shown). Instead, we observed a strong defect in the organization and dynamics of the actin cytoskeleton in the growth cone. These Rac1-induced changes of the actin cytoskeleton did not prevent neurite formation, but selectively axon growth. Furthermore, actin depolymerization reversed the defect in axon formation. This indicates that Rac1 regulates axon formation by acting predominantly through the actin cytoskeleton.