QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Govek, E.-E. Articles by Van Aelst, L. Search for Related Content PubMed PubMed Citation Articles by Govek, E.-E. Articles by Van Aelst, L. Related Content Neurobiology Signal Transduction Social Bookmarking                   What's this?

REVIEW

The role of the Rho GTPases in neuronal development

¹ Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, 11724, USA; ² Molecular and Cellular Biology Program, State University of New York at Stony Brook, Stony Brook, New York, 11794, USA

	Abstract

Top Abstract The Rho GTPases Neurite initiation and outgrowth Neuronal polarization, axon... Axon pathfinding and guidance The Rho GTPases and... The Rho GTPases and... Rho signaling and nervous... Nonsyndromic or nonspecific X... Syndromic X-linked MR (MRXS) Autosomal syndromic MR Concluding remarks Acknowledgments References

 
Our brain serves as a center for cognitive function and neurons within the brain relay and store information about our surroundings and experiences. Modulation of this complex neuronal circuitry allows us to process that information and respond appropriately. Proper development of neurons is therefore vital to the mental health of an individual, and perturbations in their signaling or morphology are likely to result in cognitive impairment. The development of a neuron requires a series of steps that begins with migration from its birth place and initiation of process outgrowth, and ultimately leads to differentiation and the formation of connections that allow it to communicate with appropriate targets. Over the past several years, it has become clear that the Rho family of GTPases and related molecules play an important role in various aspects of neuronal development, including neurite outgrowth and differentiation, axon pathfinding, and dendritic spine formation and maintenance. Given the importance of these molecules in these processes, it is therefore not surprising that mutations in genes encoding a number of regulators and effectors of the Rho GTPases have been associated with human neurological diseases. This review will focus on the role of the Rho GTPases and their associated signaling molecules throughout neuronal development and discuss how perturbations in Rho GTPase signaling may lead to cognitive disorders.

[Keywords: Rho GTPases; neurite formation; axon formation, guidance, and regeneration; dendrite formation; spine morphogenesis; nervous system disorders]

	The Rho GTPases

Top Abstract The Rho GTPases Neurite initiation and outgrowth Neuronal polarization, axon... Axon pathfinding and guidance The Rho GTPases and... The Rho GTPases and... Rho signaling and nervous... Nonsyndromic or nonspecific X... Syndromic X-linked MR (MRXS) Autosomal syndromic MR Concluding remarks Acknowledgments References

Of the Rho GTPase family members, RhoA, Rac1, and Cdc42 have been characterized most extensively. These Rho GTPases are best known for their effects on the actin cytoskeleton. In classic fibroblast studies, activation of RhoA, Rac1, and Cdc42 leads to reorganization of the actin cytoskeleton into distinct structures: stress fibers and focal adhesions, veil-like lamellipodia, and filopodial microspikes, respectively. In addition to their regulation of the actin cytoskeleton, the Rho GTPases have been shown to play a role in transcriptional activation, membrane trafficking, and microtubule dynamics. These cellular processes contribute to the effects of the Rho GTPases on cell growth control, cytokinesis, cell motility, cell–cell and cell–extracellular matrix adhesion, cell transformation and invasion, and recently, neuronal development. Given the dependency of neuronal development on regulation of the actin and microtubule cytoskeletons by the Rho GTPases, a brief description of key effectors of Rac1, Cdc42 and RhoA, and their general mechanisms of action on the cytoskeleton is presented here and schematized in Figure 1. Their involvement in neuronal development is subsequently discussed in the appropriate sections below. For more comprehensive reviews on Rho effectors, the reader is referred to the following references: Van Aelst and D'Souza-Schorey (1997); Bishop and Hall (2000); Gundersen (2002); Luo (2002); Bokoch (2003); Miki and Takenawa (2003); Riento and Ridley (2003); Burridge and Wennerberg (2004); Millard et al. (2004).

View larger version (40K): [in this window] [in a new window]   Figure 1. Rho GTPase effectors implicated in actin and microtubule dynamics. See main text for explanation. (Toca-1) Transducer of Cdc42-dependent actin assembly; (WIP) WASP-interacting protein; (WASP) Wiskott-Aldrich-syndrome protein; (Arp2/3) actin-related proteins 2 and 3; (PAK) p21-activated kinases; (LIMK) Lin-11, Isl-1, and Mec-3 kinase; (Cdk5) cyclin-dependent kinase 5; (IRSp53) insulin receptor substrate of 53 kDa; (Mena) mammalian Ena (Enabled); (WAVE) WASP family Verprolin-homologous protein; (CYFIP) cytoplasmic FMR1-interacting protein; (PIR121) a p53-inducible mRNA; (Nap125) Nck-associated protein; (Abi2) Abl interactor 2; (HSPC) heat-shock protein C; (MLCK) myosin light chain kinase; (MLC) myosin light chain; (MLCP) myosin light chain phosphatase; (Dia) Diaphanous-related formins.

Accumulating evidence indicates that another key mechanism by which Rac and Cdc42 relay signals to the actin cytoskeleton involves the Wiskott-Aldrich-syndrome family of scaffolding proteins (Machesky and Insall 1999; Suetsugu et al. 2002; Miki and Takenawa 2003; Snapper and Rosen 2003; Millard et al. 2004; Smith and Li 2004). The Wiskott-Aldrich-syndrome protein (WASP) and its closest cousin neuronal WASP (N-WASP) are regulated by Cdc42 (Rohatgi et al. 1999). Three other members of this family, WAVE1–3 (also known as Scar proteins) mediate actin-based processes triggered by Rac (Miki et al. 1998; Machesky et al. 1999; Suetsugu et al. 1999; Yamazaki et al. 2003; Yan et al. 2003). Both the WASP and WAVE family members are linked to the actin cytoskeleton through their interaction with the Arp2/3 complex (see reviews above). In the case of WASP/N-WASP, these proteins have been shown to directly bind to the activated form of Cdc42. This induces a conformational change that releases the WASP VCA domain from auto-inhibition, allowing it to activate the Arp2/3 complex to nucleate the formation of new actin filaments in vitro (Machesky et al. 1999; Rohatgi et al. 1999, 2000; Kim et al. 2000). Recent studies, however, have revealed that an additional component, Toca-1 (transducer of Cdc42-dependent actin assembly), is required to mediate Cdc42-induced actin polymerization in a physiological context. Toca-1 binds both N-WASP and Cdc42 and mediates actin nucleation by either directly activating N-WASP and/or inhibiting WIP, which is a negative regulator of N-WASP (Ho et al. 2004).

The WAVE proteins mediate actin cytoskeletal changes downstream of Rac without directly binding to Rac. In an effort to find a molecular connection between Rac and WAVE, a recent study succeeded in isolating an inhibitory WAVE complex that is responsive to Rac signaling (Eden et al. 2002). In its inactive state, this complex includes Nap125 (Nck-associated protein), PIR121 (a p53-inducible mRNA), Abi2 (Abl interactor 2) and the heat-shock protein, HSPC300. Notably, Nap125 (also called Nap1, NCKAP1, and Hem2) has been linked to Rac1 via p140 protein (Kitamura et al. 1997; Yamamoto et al. 2001). The p140 protein corresponds to Sra-1, which is an isogene of PIR121, and directly interacts with both Rac1 and Nap125 (Kobayashi et al. 1998). The PIR121 protein was previously identified as a profilin-interacting protein called POP (Witke et al. 1998), and interestingly both p140/Sra-1 and PIR121 proteins have recently been identified as fragile X mental retardation protein (FMRP)-interacting proteins and are referred to as CYFIP1 and CYFIP2, respectively (Schenck et al. 2001). When active Rac (Rac-GTP) is added, the complex dissociates, freeing WAVE and HSPC300, which then activates the Arp2/3 complex to induce actin polymerization. Subsequent studies in Drosophila and Caenorhabditis elegans have shown that the orthologs of Sra-1/PIR121 and Nap125 act in a common pathway linked to Rac-mediated actin-based protrusions and cell migration (Soto et al. 2002; Kunda et al. 2003; Schenck et al. 2003). Moreover, recent studies in mammalian cells also revealed that Abi1, Sra-1/PIR121, and Nap1 are essential intermediates of a signaling pathway from Rac activation to WAVE2-based nucleation of lamellipodial actin filaments. These studies, however, show that dissociation of the Abi1/Nap1/PIR121 complex from WAVE2 does not occur in vivo and that Rac activation recruits this complex to lamellipodia to cause site-directed nucleation of actin filaments (Innocenti et al. 2004; Steffen et al. 2004).

Another molecule, the insulin receptor substrate of 53 kDa (IRSp53), has been shown to bind both Rac and Cdc42, and links these Rho GTPases to WAVE2 and Mena (mammalian Ena), respectively (Miki et al. 2000; Krugmann et al. 2001; Miki and Takenawa 2002). The N-terminal portion of IRSp53 also binds a partial CRIB (Cdc42/Rac1 interactive binding) domain within itself, producing an intramolecular autoinhibitory interaction. This intramolecular interaction appears to be relieved by the binding of a Rho GTPase to the partial CRIB motif, or by the binding of another effector to the SH3 domain of IRSp53, which allows for the initiation of actin filament assembly (Krugmann et al. 2001). Overexpression of IRSp53 in fibroblasts causes the formation of filopodia, and IRSp53 synergizes with Mena to do so (Krugmann et al. 2001). Interestingly, a more recent study shows that IRSp53 localizes to both lamellipodia and filopodia, as does WAVE2, even in the absence of Ena/Vasp family members, including Mena (Nakagawa et al. 2003). In addition to the above interactions, IRSp53 binds the post-synaptic scaffolding molecules Shank1 (Bockmann et al. 2002; Soltau et al. 2002) and post-synaptic density (PSD)-95 (Soltau et al. 2004). Activation of IRSp53 by Cdc42 allows docking of a preassembled IRSp53/PSD-95 complex to the proline-rich region of Shank1 (Soltau et al. 2004). Given that Shank and PSD-95 organize a complex array of signaling molecules at the synapse (Ehlers 1999; Tu et al. 1999; Sheng 2001; Boeckers et al. 2002) and are important for the structure and function of dendritic spines (Ehlers 1999; Sala et al. 2001; Sheng 2001; Boeckers et al. 2002), IRSp53 via its association with these molecules is likely to be important for post-synaptic function as well. Interestingly, -PIX, a GEF for Rac and Cdc42, also binds the PDZ domain of Shank (Park et al. 2003) and plays a role in spine morphogenesis (Zhang et al. 2003). This GEF/scaffolding molecule complex may help to confer specificity on Rho GTPase signaling by linking a particular Rho GTPase with a particular effector, such as IRSp53. Lastly, mDia binds IRSp53 in a Rho-dependent manner (Fujiwara et al. 2000), further implicating IRSp53 in cytoskeletal reorganization.

Two major downstream effectors of RhoA that mediate this GTPase's effects on the cytoskeleton are members of the Rho-kinase (also called ROK/ROCK) family and the Diaphanous formin subfamily (Dia). Rho-kinase was identified as ROK (Leung et al. 1995) and ROCK2 (Matsui et al. 1996; Nakagawa et al. 1996), with ROK (Leung et al. 1996)/ROCK1 (Ishizaki et al. 1996) constituting an isoform of Rho-kinase. Rho-kinases are serine/threonine kinases that play several roles in RhoA-induced actin reorganization (Riento and Ridley 2003). They control actin filament bundling by directly phosphorylating and activating MLC, or by phosphorylating and inactivating MLC phosphatase, thereby indirectly increasing MLC phosphorylation and activation (Amano et al. 1996; Kimura et al. 1996). Furthermore, Rho-kinases may promote F-actin accumulation by phosphorylating and activating LIMK, which in turn phosphorylates and inactivates the actin depolymerization factor (ADF) cofilin (Maekawa et al. 1999; Sumi et al. 1999, 2001; Ohashi et al. 2000a,b; Amano et al. 2001).

The Diaphanous-related formins are defined by their ability to interact with activated Rho GTPases, and murine members include mDia1 (which binds RhoA-C), mDia2 (which binds RhoA and Cdc42), and mDia3 (which binds Cdc42, RhoA, Rac1, and RhoD) (Olson 2003; Wallar and Alberts 2003; Yasuda et al. 2004). As seen for WASP and PAK proteins, binding of an activated Rho GTPase seems to relieve an autoinhibitory interaction between the NH₂- and COOH-terminal regions of Dia proteins, which have been shown to regulate the actin and microtubule cytoskeletons (Alberts 2001). Yeast formins and mDia1 bind the barbed ends of actin filaments while still allowing elongation (Pring et al. 2003; Zigmond et al. 2003), and they enhance actin-nucleation via their Formin homology FH1-FH2 or FH2 domains (Tominaga et al. 2000; Copeland and Treisman 2002; Pruyne et al. 2002; Sagot et al. 2002; Kovar et al. 2003; Li and Higgs 2003; Pring et al. 2003; Shimada et al. 2004). In mammalian cells, a large proportion of actin monomer is bound to Profilin, preventing spontaneous nucleation and the addition of actin monomer to the pointed end of the filament, but not to the barbed end (Pollard et al. 2000). The binding of Profilin to the FH1 domain of mDia1 may allow mDia1 to use Profilin-bound actin monomers for nucleation (Li and Higgs 2003; Higashida et al. 2004). Another study demonstrates that mDia cooperates with Rho-kinase to induce actin fiber formation (Maekawa et al. 1999). However, this action of mDia seems to be exerted through its effect on microtubule alignment (Ishizaki et al. 2001), and recent evidence shows that Dia proteins function as downstream effectors of RhoA to regulate the formation and orientation of stable microtubules (Ishizaki et al. 2001; Palazzo et al. 2001).

Taken together, these studies show that the Rho GTP-ases exert their effects on the cytoskeleton through a large number of effectors. The involvement and potential functions of these Rho effectors in neuritogenesis, axon formation and guidance, dendritic development, and dendritic spine formation are discussed below.

	Neurite initiation and outgrowth

Top Abstract The Rho GTPases Neurite initiation and outgrowth Neuronal polarization, axon... Axon pathfinding and guidance The Rho GTPases and... The Rho GTPases and... Rho signaling and nervous... Nonsyndromic or nonspecific X... Syndromic X-linked MR (MRXS) Autosomal syndromic MR Concluding remarks Acknowledgments References

 
Immediately after neuronal commitment, extracellular cues activate membrane receptors to induce neuritogenesis, the formation of cylindrical extensions off of the neuronal cell body that serve as precursors to axons and dendrites. Each neuronal population forms neurites in a manner that is specific to its program of differentiation. However, regardless of the type of neuron, the original round shape of the cell must be broken to form a bud that develops into a neurite. Neurite initiation and outgrowth involves coordinated changes between the actin cytoskeleton, which provides a means of generating force within the cell, and the microtubule network, which stabilizes and maintains the neurites (da Silva and Dotti 2002). As a regulator of both actin cytoskeletal reorganization and microtubule orientation and stabilization, the Rho GTPases have a profound effect on neuritogenesis.

Rho GTPase studies in neuronal cell lines established their involvement in neurite formation and suggested that they act antagonistically toward each other to determine neuronal morphology. Treatment of rat pheochromocytoma PC12 cells with nerve growth factor (NGF), or serum starvation of mouse N1E-115 neuroblastoma cells, results in the formation of neurites, which depends on Rac and Cdc42 activity, since DN mutant forms of these GTPases inhibit neurite outgrowth (Sarner et al. 2000; Aoki et al. 2004). Interestingly, a recent study using FRET-based probes has shown that localized activation of Rac1 and Cdc42 is required for neurite outgrowth. Immediately after the addition of NGF to PC12 cells, Rac1 and Cdc42 are transiently activated in broad areas of the cell periphery, with a subsequent localized cycling of activity and inactivity of these GTPases at the mobile tips of protrusions. High Rac1 activity is observed in the distal half of neurite tips, while strong Cdc42 activity is concentrated in microspikes projecting from the tips (Aoki et al. 2004). Furthermore, NGF-induced recruitment of Rac1 to cell surface sites to form filamentous actin-rich protrusions is associated with a concomitant decrease in RhoA activity (Yamaguchi et al. 2001). Conversely, Rho activation is generally associated with inhibition of neurite initiation and retraction in PC12 and N1E-115 cells. RhoA signaling induces the formation of a thick ring-like structure of cortical actin filaments at the cell periphery and inhibits NGF-induced recruitment of Rac1 to protrusions (Yamaguchi et al. 2001). Furthermore, CA RhoA is sufficient to prevent neurite initiation and induce neurite retraction (Kranenburg et al. 1997; Amano et al. 1998; Hirose et al. 1998; Katoh et al. 1998a,b; Sebok et al. 1999), while inhibition of Rho using C3 exoenzyme from Clostridium botulinum or DN RhoA mimics actin-based lamellipodia and filopodia structures induced by Rac1 and Cdc42 and induces neurite formation (Tigyi et al. 1996a; Kozma et al. 1997; Kranenburg et al. 1997; Sebok et al. 1999; Brouns et al. 2001; Fujita et al. 2001). Taken together, these studies show that Rac and Cdc42 activation promote the formation of lamellipodia and filopodia, and play a role in neurite formation, while Rho activity prevents neurite initiation and induces neurite retraction.

The generalization that Rac and Cdc42 activation promote neurite formation, while Rho activation antagonizes it in neuronal cell lines, holds true in several primary cell systems. In agreement with the PC12 and N1E-115 experiments above, neurite outgrowth is stimulated by the chick-specific Rac1 subtype cRac1B in chick primary retinal neurons and inhibited by an inactive form of cRac1B (Albertinazzi et al. 1998), and CA Rac1 increases and DN Rac1 decreases neurite extension in dissociated rat hippocampal neurons (Schwamborn and Puschel 2004). CA Rho prevents neurite outgrowth of cultured hippocampal neurons, including that promoted by neurotrophin 3 (NT-3), brain-derived neurotrophic factor (BDNF), NGF, and coating on a laminin substrate (Da Silva et al. 2003; Schwamborn and Puschel 2004), and neurite outgrowth is promoted by Rho inhibition using C3 exoenzyme in chick dorsal root ganglia (DRG) and cultured rat hippocampal neurons (Jin and Strittmatter 1997; Sarner et al. 2000; Da Silva et al. 2003; Fournier et al. 2003). C3 exoenzyme also rescues the arrest of neurite formation and elongation that results from the plating of cultured rat hippocampal neurons on a collagen substrate (Da Silva et al. 2003).

In contrast to the above findings, there are several instances in which CA and DN mutant forms of Rac and Cdc42 have been found to induce opposite phenotypes. Both CA Rac1 and CA Cdc42 expressing Drosophila giant fiber system neurons exhibit a lack of neurite outgrowth (Allen et al. 2000), CA Rac1 decreases the length of the longest neurite in cultured rat cortical neurons (Kubo et al. 2002), and DN Rac promotes neurite outgrowth in chick DRG (Fournier et al. 2003). These differential effects of the Rho GTPases on neurite formation may be due to the diversity of species, cell types, and age of cells or organisms used in primary cell studies, or reflect the need for the Rho GTPase to cycle between an active GTP-bound and inactive GDP-bound state in order to properly regulate neuritogenesis. This last possibility is emphasized in studies in which CA and DN mutants of a Rho GTPase produce the same effect. For example, both CA and DN Rac1 mutants retard growth cone advance, neurite outgrowth, and differentiation in primary chick embryo motor neurons (Kuhn et al. 1998). The important information gleaned from these studies is that the Rho GTPases play an important role in neurite formation in primary neurons, and that tight regulation and cycling of the Rho GTPases are required for normal neurite initiation and outgrowth.

Rac and Cdc42 signaling pathways that promote neurite formation

Given the effects of the Rho GTPases on neurite formation discussed above, a recurrent theme is the requirement of concurrent Rac and Cdc42 activation and Rho inactivation to promote neurite formation. Signaling cues that positively affect Rac and Cdc42 activity, while negatively impacting Rho activity, include growth factors, receptors, and Rho GTPase regulatory molecules. NGF is well known for its ability to stimulate the formation of neurites in neuronal cell lines (Negishi and Katoh 2002), and the NGF-elicited signaling pathway that leads to neurite formation through modulation of Rho GTPase activity has been largely elucidated. The Ras-linked tyrosine kinase receptor TrkA mediates NGF-activation of Rac1 and inactivation of RhoA (Nusser et al. 2002), and PI 3-kinase is required for NGF activation of both Rac1 and Cdc42 (Nusser et al. 2002; Aoki et al. 2004), NGF-induced inactivation of RhoA (Nusser et al. 2002), and NGF and Ras-induced process formation (Kobayashi et al. 1997; Kita et al. 1998; Sarner et al. 2000). Furthermore, activated Ras and PI 3-kinase-induced neurite outgrowth requires both Rac1 and Cdc42 activity, while neurite outgrowth induced by activated Cdc42 is Rac1 dependent (Sarner et al. 2000). Conversely, CA RhoA is able to antagonize neurite outgrowth induced by activated Ras (Sarner et al. 2000), and NGF treatment of PC12 cells causes a decrease in active Rho and its effector Rho-kinase downstream of Rac activation (Nusser et al. 2002). Thus, the following signaling cascade has been proposed for NGF-induced neurite outgrowth: NGF, TrkA, Ras, PI 3-kinase, Cdc42, and Rac, with a concomitant decrease in Rho and Rho-kinase activities downstream of Rac. NGF-induced Rac and Cdc42 activation and neurite formation may be further modulated by a specific splice variant of the RhoGAP Nadrin, Nadrin-116. Nadrin-116 inhibits NGF-induced neurite outgrowth in PC12 cells, which is dependent on Nadrin's GAP activity, suggesting a Rac- or Cdc42-directed preference for this molecule in relation to neurite formation (Furuta et al. 2002).

In addition to acting downstream of Ras signaling to regulate neuritogenesis, Rac and Cdc42 have been found to mediate the neurite-promoting effects of another Ras-like small GTPase, Rin. Rin is expressed predominantly in adult neurons and binds calmodulin (CaM) (Lee et al. 1996). Rin expression in PC12 cells increases Rac and Cdc42 activity and induces neurite formation in a MAP-kinase pathway independent manner. Rin-induced neurite outgrowth can be suppressed by DN Rac, DN Cdc42, or CaM inhibitor, and a Rin mutant unable to associate with CaM fails to induce neurite outgrowth. Furthermore, interfering with Rin function inhibits potentiation of neurite formation induced by a combination of forskolin, an adenylyl cyclase activator, and KCl, which induces extracellular calcium entry through voltage-dependent calcium channels. Rin-induced neurite outgrowth therefore requires Rac and Cdc42 activation and Rin association with CaM, and may involve Rin/calcium/CaM-mediated neuronal signaling. It should be noted that Rin expression also increases Rho activity, and interfering with RhoA levels using RNA interference (RNAi) induces the formation of more neurite branch points in Rin-induced neurites (Hoshino and Nakamura 2003). However, given that Rin-induced Rho activation does not appear to antagonize neurite formation as expected, how Rin balances Rac, Cdc42, and Rho activity to promote neurite formation remains to be elucidated.

Several Rac-specific GEFs have also been shown to play a role in neurite formation, including Tiam1 (the invasion-inducing T-lymphoma and metastasis 1 protein), STEF (SIF and Tiam1-like exchange factor), and FIR (FERM domain including RhoGEF). Tiam1 is a Rac-specific GEF that is highly expressed in the developing nervous system and may play a role in the migration of granule cells to their final destination (Habets et al. 1994; Ehler et al. 1997). Overexpression of Tiam1 in N1E-115 cells promotes cell spreading and neurite outgrowth on the extracellular matrix protein laminin through the recruitment of the 61 integrin and activation of Rac1. Neurite formation induced by Tiam1 can be enhanced by inactivation of RhoA and overcome by coexpression of CA RhoA (Leeuwen et al. 1997), once again highlighting the importance of Rac activation and Rho inactivation for neurite formation. Tiam1 also plays a role in neurite outgrowth promoted by ephrin-B1/EphB2-mediated reverse signaling and ephrin-A1/EphA2-mediated forward signaling (Tanaka et al. 2004). Ephrins are transmembrane or glycosylphosphatidylinositol-anchored molecules that are ligands of Eph receptor tyrosine kinases (Wilkinson 2001; Cutforth and Harrison 2002; Guan and Rao 2003; Huber et al. 2003). Tiam1 interacts with both ephrin-B1 and EphA2, and Rac1 is activated by extracellular stimulation of clustered soluble EphB2 receptors in cells coexpressing Tiam1 and ephrin-B1 or by soluble ephrin-A1 in cells coexpressing Tiam1 and EphA2. Importantly, primary cortical neurons from mouse embryos and neuroblastoma cells significantly extend neurites on surfaces coated with the extracellular domain of EphB2 or ephrin-A1, an effect that is negated by expression of DN ephrin-B1, DN EphA2, or DN Tiam1 (Tanaka et al. 2004). Another Rac1-specific GEF, STEF, is predominantly expressed in the brain, and CA STEF induces neurite-like processes that can be inhibited by DN Rac1 in N1E-115 cells (Matsuo et al. 2002), suggesting that STEF, like Tiam, activates Rac to promote neurite formation. In contrast to these findings, the Rac-specific GEF FIR appears to hinder neurite growth in cortical neurons. Ectopic expression of FIR results in shortened neurites and excessive growth cones in a Rac-dependent manner, as does CA Rac1 in this system (Kubo et al. 2002).

An additional protein linked to Rac and Cdc42-promoted neurite outgrowth is CD47 (also known as integrin-associated protein), a transmembrane protein that belongs to the immunoglobulin (Ig) superfamily. CD47 has been shown to activate Rac and Cdc42 and induce the formation of neurites and filopodia through Rac and Cdc42, respectively, in N1E-115 neuroblastoma cells. CD47-induced neurite and filopodia formation are also likely triggered by the CD47 ligand SHPS-1 (BIT/SIRP) and involve integrins that contain the 3 subunit (Miyashita et al. 2004).

Once activated, Rac and Cdc42 relay neurite-promoting signals by binding and activating downstream effector molecules. PAK kinases are prominent effectors of both Rac and Cdc42 shown to play a role in neurite formation. PAK1 induces neurite outgrowth independent of its kinase activity when targeted to membrane in PC12 cells (Daniels et al. 1998) and is also indirectly linked to neurite outgrowth in cortical neurons through its association with both Rac and p35 (CDK5 kinase) (Nikolic et al. 1998). p35 is a neuron-specific regulator for CDK5 that colocalizes with Rac in neuronal growth cones, associates directly with Rac in a GTP-dependent manner (Nikolic et al. 1998), and regulates neurite outgrowth in cortical neurons in culture (Nikolic et al. 1996). Interestingly, PAK1 is present in a Rac/p35/Cdk5 complex and active p35 causes hyperphosphorylation of PAK1 in a Rac-dependent manner, resulting in down-regulation of PAK1 activity. Thus Rac not only activates PAK1, but also appears to regulate the duration of its activity via a p35/CDK5 complex (Nikolic et al. 1998), perhaps to modulate neurite formation. PAK5 is another member of the PAK family that has been implicated in neurite formation. PAK5 is highly expressed in mammalian brain and induces the formation of filopodia and neurite-like processes in N1E-115 cells. However, unlike PAK1, the action of PAK5 on neurite outgrowth is dependent upon its kinase activity. Furthermore, an activated PAK5 inhibits Rho activity and its effect on neurite outgrowth can be abolished by an activated RhoA mutant (Dan et al. 2002), suggesting that PAK5 acts downstream of Rac/Cdc42 to antagonize Rho signaling.

Other Cdc42 effectors involved in neurite outgrowth include myotonic dystrophy kinase-related Cdc42-binding kinase (MRCK), N-WASP, and Cdc42Hs-associated kinase-1 (ACK-1). Mutants of MRCK, including a kinase-dead MRCK, and N-WASP, including N-WASP mutants unable to bind Cdc42 or activate Arp2/3, are able to block NGF-induced neurite outgrowth in PC12 cells (Chen et al. 1999; Banzai et al. 2000), and the N-WASP mutant unable to activate Arp2/3 inhibits neurite formation in cultured rat hippocampal neurons (Banzai et al. 2000). Cdc42 and its effector ACK-1 also regulate neurite formation downstream of muscarinic cholinergic receptor (mAChR) activation in human SH-SY5Y neuroblastoma cells expressing M₃-type muscarinic receptors. The mechanism by which this occurs involves phosphorylation of ACK-1 via a Fyn tyrosine kinase signaling pathway, which is dependent upon Rho-kinase signaling (Linseman et al. 2001). In addition to the above Cdc42 effectors, an adaptor protein that binds Cdc42, the 58-kDa substrate of the insulin receptor tyrosine kinase (IRS-58) has been shown to colocalize with F-actin and induce the formation of filopodia in fibroblasts and promote highly complex neurite outgrowth in N1E-115 cells, while an IRS-58 mutant unable to bind Cdc42 antagonizes these effects (Govind et al. 2001). Notably, IRS-58 is identical to IRSp53, except for differences in post-translational modifications (Abbott et al. 1999). Taken together, these findings not only implicate these Cdc42 effectors in neurite formation, but also underscore the importance of actin cytoskeletal changes for this process.

The role of Rho signaling in neurite retraction

While Rac and Cdc42 activation, coupled with Rho inactivation, promote neurite formation, Rho activation antagonizes this effect and causes neurite retraction, a process that may play an important role in the remodeling of neuronal protrusions important for guidance and synaptic plasticity. Rho activation and neurite retraction have been linked to lysophosphatidic acid (LPA) activation of G-protein coupled receptors, thrombin receptor activating peptide (TRP) activation of thrombin receptor, activation of prostaglandin E receptor EP3 subtype, activation of sphingosine-1-phosphate receptor, and the addition of serum to cells (Jalink et al. 1994; Katoh et al. 1996, 1998a; Postma et al. 1996). It was originally noted that the addition of LPA or TRP causes rapid growth cone collapse, neurite retraction, and cell rounding, and that C. botulinum C3 exoenzyme is able to inhibit this (Jalink et al. 1994; Tigyi et al. 1996a,b). C3 exoenzyme ADP ribosylates and specifically inhibits Rho activity and is able to cause neurite outgrowth (Nishiki et al. 1990; Kamata et al. 1994; Tigyi et al. 1996a; Kozma et al. 1997; Kranenburg et al. 1997; Sebok et al. 1999; Brouns et al. 2001; Fujita et al. 2001). Later studies suggested that LPA activates Rho through the heterotrimeric G-proteins (Tigyi et al. 1996a), including G12, G13, and Gq, though they are believed to induce Rho-dependent neurite retraction through different signaling pathways. Gq-induced neurite retraction can be blocked by inhibiting protein kinase C or extracellular calcium, and both G13 and Gq-elicited neurite retraction, as well as LPA-induced RhoA activation, can be blocked by tyrosine kinase inhibitors (Katoh et al. 1998b; Kranenburg et al. 1999). In nonneuronal cells, the Rho-specific GEF PDZ–RhoGEF mediates both G12 and G13-induced Rho activation (Fukuhara et al. 1999), and p115 RhoGEF stimulates G12 and G13 activity and mediates G13-induced Rho activation, which is negatively regulated by G12 (Hart et al. 1998; Kozasa et al. 1998).

Rho regulatory molecules implicated in neurite retraction include a more recently identified Rho-specific GEF, KIAA0380, and a number of Rho GAPs. The RhoGEF KIAA0380 is related to p115 RhoGEF, is highly expressed in brain, activates Rho-kinase, and induces stress fiber formation in nonneuronal cells. In Neuro2a cells, KIAA0380 is present in the tips of neurites, and expression of an N-terminal fragment inhibits LPA-induced neurite retraction, suggesting that KIAA0380 plays a role in neurite retraction through Rho signaling (Togashi et al. 2000). Rho GAPs involved in neurite retraction include p190 RhoGAP and Grit. p190 RhoGAP is a Rho-specific GAP (Billuart et al. 2001) which is the principal Src (Src/Fyn) substrate in developing and mature nervous system. Overexpression of p190 RhoGAP causes extensive neurite outgrowth on laminin in N1E-115 cells and extensive outgrowth of long and highly branched neurites in neuroblastoma N2A cells. This growth is dependent on the RhoGAP domain and can be blocked by RhoA (Brouns et al. 2001). Grit is a novel GAP for the Rho family of GTPases that is highly expressed in neuronal cells. Grit directly interacts with the TrkA NGF receptor, and the GAP domain alone of Grit exhibits GAP activity mainly toward RhoA and Cdc42, with only moderate activity toward Rac1. Overexpression of the TrkA-binding region of Grit inhibits NGF-induced neurite elongation in PC12 cells, while full-length Grit promotes neurite formation upon NGF stimulation, suggesting Rho-directed activity in these cells (Nakamura et al. 2002). Lastly, a novel RhoGAP family member, p200RhoGAP, acts preferentially toward RhoA and Rac1, and costains with cortical actin in naive N1E-115 cells and the actin-rich ends of neurites in differentiated cells. Expression of either full-length p200RhoGAP or the RhoGAP domain alone induces differentiation (Moon et al. 2003), a phenotype reminiscent of decreased Rho activity.

Additional Rho proteins implicated in neurite formation are Dvl, Wnt, and -catenin. Dvl mediates Wnt signaling in the -catenin and planar cell polarity pathways and has recently been shown to play a role in Wnt-3a-dependent neurite retraction. Dvl-1 is capable of activating Rho and Rho-kinase in PC12 cells, and expression of Dvl-1 inhibits NGF-induced neurite outgrowth in PC12 cells and serum-starvation-dependent neurite outgrowth in N1E-115 cells, which can both be prevented by the Rho-kinase inhibitor Y-27632. Furthermore, a Dvl-1 mutant incapable of activating Rho-kinase fails to induce neurite retraction. With regard to Wnt signaling, NGF-induced neurite formation in PC12 cells is suppressed by Wnt-1 or Wnt-3, and neurite formation under these conditions is enhanced by Y-27632. Wnt-3a protein also stimulates -catenin and Rho-kinase activity, and the Dvl-1 mutant unable to activate Rho-kinase prevents Wnt-3a-induced neurite retraction (Kishida et al. 2004). Together, these data indicate that Dvl acts downstream of Wnt to regulate Rho and Rho-kinase activities necessary for neurite retraction. Another potential Rho protein implicated in neurite formation is -catenin, which is a neuronal protein that binds to the juxtamembrane segment of classical cadherins, and interacts with cortactin. -Catenin is phosphorylated by Src family kinases, which inhibits binding of -catenin to cortactin and causes cells to extend unbranched primary processes upon NGF treatment in PC12 cells. Inhibition of Rho using C3 exoenzyme changes the effect of -catenin from primary process extension to branch formation, and both DN RhoA and Rho-kinase inhibitor inhibit -catenin process formation. An activated RhoA mutant also decreases the length of protrusions compared with -catenin alone. Interestingly, growth induced by -catenin is more robust and is restricted to dendrites in cultured hippocampal neurons (Martinez et al. 2003). Thus these findings suggest that -catenin regulates process extension through Rho signaling and plays a role in dendrite formation.

A major downstream target of RhoA, its effector Rho-kinase, has been shown to mediate Rho-driven neurite retraction. Both wild-type and CA Rho-kinase arrest cells in a round state or induce neurite retraction (Amano et al. 1998; Hirose et al. 1998; Katoh et al. 1998a; Da Silva et al. 2003), including that promoted by NT-3, BDNF, NGF, and a laminin-coated substrate in cultured rat hippocampal neurons (Da Silva et al. 2003). Conversely, DN Rho-kinase or the Rho-kinase-specific inhibitor Y-27632 promotes the formation of neurites (Hirose et al. 1998; Fujita et al. 2001; Da Silva et al. 2003), DN Rho-kinase mutants inhibit both LPA-elicited (Amano et al. 1998; Hirose et al. 1998) and EP3 prostaglandin E receptor-induced neurite retraction (Katoh et al. 1998a), and Rho-kinase inhibition via Y-27632 rescues collagen-induced arrest of neurite sprouting and elongation in cultured rat hippocampal neurons (Da Silva et al. 2003).

The effects of activated RhoA and Rho-kinase on neurite retraction are likely brought about by increased actomyosin contractility. LPA activation of RhoA causes rapid F-actin assembly, along with neurite retraction and cell rounding, suggesting that contraction of the cortical actin cytoskeleton is important for neurite retraction (Kranenburg et al. 1997). Rho-kinase has been shown to directly phosphorylate and activate MLC (Amano et al. 1996, 1998) and is required for LPA-induced MLC phosphorylation in N1E-115 cells (Hirose et al. 1998). The use of a mutant MLC with an activated myosin ATPase promotes neurite retraction (Amano et al. 1998), while a MLCK inhibitor blocks LPA and TRP-induced neurite retraction (Jalink et al. 1994). Conversely, NGF and C3-induced neurite outgrowth can be counteracted by a phosphatase inhibitor, and both NGF and C3 exoenzyme treatment of PC12 cells and of C3 exoenzyme have been shown to decrease MLC phosphorylation. These findings suggest that NGF induces neurite outgrowth in part through inhibition of the RhoA/Rho-kinase signaling pathway that results in MLCP activation, causing a transient decrease in phosphorylated MLC (Fujita et al. 2001). The finding that myosin IIA antisense oligonucleotides inhibit LPA or thrombin-induced neurite retraction in Neuro-2A neuroblastoma cells, as does Y-27632, shows that the myosin isoform IIA is the myosin motor that drives neurite retraction (Wylie and Chantler 2003).

Additional Rho-kinase substrates linked to regulation of the actin cytoskeleton are the LIM kinases. As mentioned in the introduction, LIM kinases are serine/threonine kinases that phosphorylate and thus inactivate the actin depolymerizing protein cofilin. LPA treatment of N1E-115 cells causes phosphorylation of cofilin in a Y-27632-sensitive manner. This increase in cofilin phosphorylation is achieved by Rho-kinase's direct phosphorylation and activation of LIM-kinase, which then phosphorylates cofilin (Maekawa et al. 1999). N-terminal LIM domains also inhibit PC12 cell differentiation after stimulation with NGF and Y-27632, while the PDZ domain of LIM-kinase only reduces neurite outgrowth induced by Y-27632 (Birkenfeld et al. 2001).

The actin monomer binding protein Profilin has also been suggested to regulate neuritogenesis downstream of Rho-kinase by modulating actin stability. Different Profilin isoforms include Profilin I (PI), which is ubiquitously expressed, Profilin II (PIIa) and Profilin IIb (PIIb), which are largely restricted to the brain (Witke et al. 1998), and a third Profilin that is restricted to kidney and testes (E. Hu et al. 2001). Rho-kinase has been shown to associate with and phosphorylate PIIa (Witke et al. 2001; Da Silva et al. 2003), though this phosphorylation is not necessary for the binding of these two proteins (Da Silva et al. 2003). Interestingly, PIIa deficiency or overexpression causes phenotypes similar to impaired RhoA/Rho-kinase function and RhoA/Rho-kinase activation, respectively, in cultured rat hippocampal neurons. Neurons from PII-deficient mice or those subjected to antisense directed against PIIa exhibit an increase in neurite number, length, and branching, and a decrease in F-actin content. Overexpression of PIIa has the opposite effect. PIIa overexpression is capable of rescuing the effects of Rho and Rho-kinase inhibition on neurite outgrowth and F-actin content, while a reduction in PIIa ameliorates the effects of CA RhoA and Rho-kinase. Similarly, PIIa expression antagonizes increased neurite number and length and decreased F-actin caused by the actin depolymerizing factor cytocholasin D, suggesting that PIIa increases actin stability by increasing F-actin content. PIIa expression also antagonizes the induction of neurite formation by NT-3, BDNF, NGF, and a laminin-coated substrate, as do CA RhoA and Rho-kinase, and reduction of PIIa, Rho, or Rho-kinase activity rescues collagen-induced arrest of neurite sprouting and elongation. Thus, PIIa acts downstream of RhoA/Rho-kinase signaling to regulate actin stability during neuritogenesis in mammalian hippocampal neurons (Da Silva et al. 2003).

A recently discovered Rho-kinase-interacting protein potentially involved in neurite outgrowth is p21^Cip1/WAF1, which binds to and inhibits both cyclin/Cdk kinases and proliferating cell nuclear antigen. It is induced in the cytoplasm during differentiation of chick retinal precursor cells and N1E-115 cells. Without its nuclear localization signal, it affects the formation of actin structures similar to inactivation of Rho and forms a complex with Rho-kinase that inhibits its activity in vitro and in vivo. Neurite outgrowth and branching from hippocampal neurons are promoted if p21^Cip1/WAF1 is expressed abundantly in the cytoplasm (Tanaka et al. 2002). Together, the studies above demonstrate an important link between Rho-kinase and proteins associated with regulation of the actin cytoskeleton, underscoring the importance of actin reorganization for neurite retraction.

Rho-kinase has also been implicated in the reorganization of microtubules and intermediate filaments necessary for neurite retraction. An early study showed that DN Rho-kinase interferes with cytoskeletal collapse of microtubules and the intermediate filament peripherin caused by the presence of serum in N1E-115 cells (Hirose et al. 1998). Interestingly, tubulin disappears in retracting neurites, whereas vimentin and actin remain colocalized. Vimentin is one of the intermediate filaments and a major cytoskeletal component in developing neurons, and is phosphorylated by Rho-kinase, which causes disassembly of these filaments. Interfering with Rho-kinase activity, using a DN mutant or inhibitor, abolishes Rho-kinase-induced phosphorylation of vimentin and results in irregular neurite outgrowth (Nakamura et al. 2000). Thus the Rho/Rho-kinase signaling pathway not only regulates actin organization necessary for growth cone collapse, but also the microtubule and intermediate filament cytoskeletons vital to cell structure.

The role of other Rho GTPases in neurite formation

Additional Rho GTPases that play a role in neurite formation and retraction include RhoG, Rnd1/2, TC10, and RhoT. RhoG has been implicated in several signaling pathways and complexes that contribute to its ability to promote neurite formation. Downstream of Ras, RhoG activates Rac and Cdc42 to promote neurite outgrowth. Wild-type RhoG is capable of inducing neurite outgrowth in PC12 cells independent of NGF stimulation and enhances outgrowth in its presence in a Rac1- and Cdc42-dependent manner. In support of this, DN mutants of Rac1 and Cdc42 interfere with RhoG-induced neurite outgrowth, CA RhoG elevates endogenous Rac1 and Cdc42 activities, and DN RhoG experiments show that interfering with RhoG function suppresses outgrowth, including that induced by activated Ras (Katoh et al. 2000). RhoG activity is further regulated by the RhoGEF Trio, which contains two RhoGEF domains; GEFD1, which activates Rac through RhoG; and GEFD2, which activates RhoA. Human Trio induces neurite outgrowth in PC12 cells in a GEFD1-dependent manner through RhoG (Estrach et al. 2002), indicating that neurite outgrowth involves Trio/RhoG/Rac1-Cdc42 signaling. A recent study suggests that RhoG also interacts with Elmo in a GTP-dependent manner and forms a ternary complex with Dock180 to activate Rac1. Coexpression of CA RhoG with Elmo and Dock180 promotes relocalization of Elmo and Dock180 from the cytosol to the plasma membrane, and increases Dock180 and Elmo-mediated activation of Rac1. In addition, DN RhoG, and Dock180 and Elmo mutants, are all able to prevent NGF-induced neurite outgrowth in PC12 cells (Katoh and Negishi 2003). Another RhoG-interacting protein, kinectin, is a regulator of microtubule-dependent kinesin activity. Kinectin selectively binds GTP-bound RhoG, inhibition of RhoG activity leads to relocalization of both kinectin and RhoG from the cell periphery to a perinuclear distribution, and changes in the activity of both proteins influence microtubule transport of lysosomes in nonneuronal cells (Vignal et al. 2001). Thus kinectin's ability to link RhoG to microtubule-dependent transport in cells may prove important for neuronal differentiation. Together, these studies show that RhoG acts downstream of Ras and Trio, and may act in concert with Elmo and Dock180, to activate Rac and Cdc42 in order to promote neurite formation. Furthermore, RhoG could potentially regulate microtubule-dependent transport necessary for the growth of neuronal processes.

Rnd proteins, which include Rnd1, Rnd2, and Rnd3 (also known as RhoE), are relatively new members of the Rho family of GTPases that have low intrinsic GTPase activity and are thought to be CA (Foster et al. 1996; Guasch et al. 1998; Nobes et al. 1998). In general, Rnd1 and Rnd3 prevent stress fiber and focal adhesion formation (Guasch et al. 1998; Nobes et al. 1998), suggesting that they function in part by antagonizing the RhoA signaling pathway. In line with this, Rnd1 causes the formation of numerous neurites that contain microtubules but little filamentous actin and neurofilaments in PC12 cells, and the formation of these processes by Rnd1 is inhibited by DN Rac1, suggesting that Rnd1 induces Rac-dependent neurite formation by disrupting cortical actin filaments (Aoki et al. 2000). More recent studies provide direct evidence for Rnd inhibition of RhoA signaling in nonneuronal cells, where Rnd proteins regulate reorganization of the actin cytoskeleton by reducing cellular levels of GTP-bound RhoA through a p190 RhoGAP mechanism and by binding and sequestering Rho-kinase to prevent it from phosphorylating MLCP (Riento et al. 2003; Wennerberg et al. 2003). Rnd2 has also been implicated in neurite formation. Rnd2 interacts with the protein Rapostlin in a GTP-dependent manner, and Rapostlin binds directly to microtubules and regulates reorganization of both actin filaments and microtubules. Rapostlin induces neurite branching in response to Rnd2 in PC12 cells and primary hippocampal neurons, an effect that is dependent upon the microtubule-binding region of Rnd2 (Fujita et al. 2002). Together, the Rnd GTPases appear to promote neurite formation and branching by antagonizing RhoA signaling and coordinating actin and microtubule cytoskeletal changes.

TC10 and RhoT have also been implicated in neurite outgrowth. TC10 was identified as a gene whose expression is up-regulated in response to nerve injury, while RhoA, Rac1, and Cdc42 only show slight up-regulation (Tanabe et al. 2000). This GTPase can associate with some of the same downstream effectors as Cdc42 and Rac, including PAK, N-WASP, and MRCK (Neudauer et al. 1998; Abe et al. 2003). TC10 is highly expressed in muscular tissue and brain and is induced during differentiation of C2 skeletal muscle cells and neuronal differentiation in N1E-115 and PC12 cells. RhoT is expressed in heart and uterus and during neuronal differentiation of N1E-115 cells. Both TC10 and RhoT have been found to induce neurite outgrowth in PC12 and N1E-115 cells (Abe et al. 2003), and TC10 induces neurite extension in cultured rat DRG neurons (Tanabe et al. 2000). DN mutants of TC10 and RhoT are capable of preventing neuronal differentiation induced by dibutyryl cyclic AMP in PC12 cells and by serum starvation in N1E-115 cells (Abe et al. 2003), but DN TC10 is unable to inhibit NGF-induced neurite outgrowth in PC12 cells (Murphy et al. 2001). In addition, neurite formation induced by TC10 and RhoT is mediated by N-WASP, given that DN N-WASP mutants, including one deficient for GTPase binding and one deficient for ARP2/3 binding, can inhibit TC10 and RhoT-induced neurite outgrowth (Abe et al. 2003).

	Neuronal polarization, axon growth, and regeneration

Top Abstract The Rho GTPases Neurite initiation and outgrowth Neuronal polarization, axon... Axon pathfinding and guidance The Rho GTPases and... The Rho GTPases and... Rho signaling and nervous... Nonsyndromic or nonspecific X... Syndromic X-linked MR (MRXS) Autosomal syndromic MR Concluding remarks Acknowledgments References

 
How does a neuron develop the unique and intricate architecture that allows it to receive and store information vital for cognitive function? Neurons possess a highly polarized structure that provides for the vectorial flow of information. They typically have several dendrites that receive inputs from presynaptic neurons and one axon that relays information to post-synaptic neurons. The development of these processes has been studied extensively in rat primary hippocampal cultures and consists of five stages (see Fig. 2). Upon plating, round neurons attach to the substratum and form a lamellipodium (Stage 1), which is breached by the sprouting of several minor neurites (Stage 2). These neurites are cylindrical extensions with a growth cone at the distal tip, and while they continue to extend and retract, there is no net elongation of these processes during this time. Within 24 h, one of these neurites will elongate rapidly without retraction to form the axon (Stage 3), and several days later the remaining neurites will grow at a slower rate to form the dendrites (Stage 4). Lastly, the axon develops further, dendrites continue to branch, and dendritic spines are formed, allowing for the formation of synaptic contacts and spontaneous electrical activity throughout the neuronal network (Stage 5) (Dotti et al. 1988; Higgins et al. 1997; da Silva and Dotti 2002; Arimura et al. 2004).

View larger version (14K): [in this window] [in a new window]   Figure 2. Establishment of polarity and stages of neuronal development in hippocampal neurons. Cultured hippocampal neurons form a lamellipodium when they attach to the substratum (Stage 1), and shortly after neurites begin to sprout (Stage 2). The neurites are cylindrical protrusions that contain a growth cone at their distal tip and they lack molecular and structural characteristics of mature axonal or dendritic processes. These neurites extend and retract without net elongation. The following day, one of the neurites with an enlarged growth cone at its tip (see darkened arrow) elongates rapidly without retraction to form the axon (Stage 3; see open arrow). Several days later, the remaining neurites continue to grow and branch to form the dendrites (Stage 4), and in a final step of maturation, the axon and dendrites develop further and dendritic protrusions, or spines, appear (Stage 5). See text for further explanation and references.