|Home | About | Journals | Submit | Contact Us | Français|
The ability of neurons to form a single axon and multiple dendrites underlies the directional flow of information transfer in the central nervous system. Dendrites and axons are molecularly and functionally distinct domains. Dendrites integrate synaptic inputs, triggering the generation of action potentials at the level of the soma. Action potentials then propagate along the axon, which makes presynaptic contacts onto target cells. This article reviews what is known about the cellular and molecular mechanisms underlying the ability of neurons to initiate and extend a single axon during development. Remarkably, neurons can polarize to form a single axon, multiple dendrites, and later establish functional synaptic contacts in reductionist in vitro conditions. This approach became, and remains, the dominant model to study axon initiation and growth and has yielded the identification of many molecules that regulate axon formation in vitro ( Dotti et al. 1988). At present, only a few of the genes identified using in vitro approaches have been shown to be required for axon initiation and outgrowth in vivo. In vitro, axon initiation and elongation are largely intrinsic properties of neurons that are established in the absence of relevant extracellular cues. However, the importance of extracellular cues to axon initiation and outgrowth in vivo is emerging as a major theme in neural development ( Barnes and Polleux 2009). In this article, we focus our attention on the extracellular cues and signaling pathways required in vivo for axon initiation and axon extension.
Historically, the advent of in vitro dissociated neuronal cultures provided an experimental template for improving our understanding of the cell biology of neuronal polarity, including the specification of the molecular identity of axon and dendrites (Goslin and Banker 1989; Craig and Banker 1994). Careful analysis of these cultures led to the observation that cultured hippocampal neurons transition through several stages: from freshly plated stage-1 cells bearing immature neurites to stage-5 cells that show mature axons, dendrites, dendritic spines, and functional synapses (Dotti et al. 1988; Craig and Banker 1994) (Fig. 1A). It should be noted that in the classical E18 rat hippocampal cultures, most plated cells are polarized postmitotic neurons before dissociation. It is therefore important to keep in mind that, in this in vitro model, molecular manipulations act on previously polarized neurons that may retain some aspects of their initial polarization. This can be critical for interpreting these experimental results. Recent advances in techniques such as in utero or ex utero cortical electroporation (Saito and Nakatsuji 2001; Tabata and Nakajima 2001; Hatanaka and Murakami 2002; Hand et al. 2005), provide a paradigm for (a) manipulating gene expression in progenitors (i.e., before neuronal polarization occurs on cell-cycle exit), and for (b) visualizing the earliest stages of neuronal polarization in a contextual cellular and molecular environment (i.e., in organotypic slices or intact embryonic brain) (Hand et al. 2005; Barnes et al. 2007; Calderon de Anda et al. 2008).
Neuronal polarization can be divided into several specific steps in vivo. On cell-cycle exit, mammalian neurons usually migrate over a long distance before reaching their final destination. In vivo, most neurons undergo axon–dendrite polarization during migration. Although migrating, neocortical pyramidal neurons form a leading process and a trailing process, each becoming the axon or the dendrite (Fig. 1B). Careful examination of the morphological transition between neural progenitors and postmitotic neurons reveals that neurons can inherit their axon and dendrite polarity directly from the apico-basal polarity of their progenitors. This is the case for retinal ganglion cells and bipolar cells in the developing vertebrate retina (Hinds and Hinds 1978; Morgan et al. 2006; Zolessi et al. 2006; reviewed in Barnes and Polleux 2009). In other neuronal subpopulations undergoing long-range migration, neuronal morphogenesis undergoes extensive stereotypical changes, leading to polarized outgrowth of their axon and dendrites. This is the case for cerebellar granule neurons (CGN) as well as cortical and hippocampal pyramidal neurons (PN), two of the best-studied models of neuronal polarization (Rakic 1971; Rakic 1972; Shoukimas and Hinds 1978; Gao and Hatten 1993; Komuro et al. 2001; Hatanaka and Murakami 2002; Noctor et al. 2004). Both CGN and PN acquire their axon–dendrite polarity from the polarized emergence during migration of their trailing and leading processes, respectively (Fig. 1B). Therefore, in these two neuronal cell types, an important functional relationship exists between the molecular mechanisms underlying polarized migration and the final axon–dendrite polarity.
As we discuss later, an emerging concept from recent work primarily in Caenorhabditis elegans suggests that in vivo, the “symmetry-breaking” events that lead to the emergence of the dendrite and the axon require the ability of postmitotic neurons to sense gradients of extracellular cues leading to the asymmetric activation of signaling pathways underlying the emergence of the axon. In this article, we provide a synthetic model of the molecular mechanisms underlying axon specification and axon growth in vitro and in vivo.
As mentioned previously, most studies published over the past two decades in this field have been performed using in vitro approaches. An important paradigm for confirming the regulatory role of a gene in neuronal polarity is to show that down-regulation of its expression, using shRNA technology or gene knockout technology, is required for axon formation. These experiments are typically performed using axon-specific markers and measurement of neurite length, because the axon usually grows five to 10 times faster than do neurites becoming dendrites. However, this type of evidence may not be sufficient to distinguish unambiguously an effector of axon specification from a molecule that is simply required for axon growth (Jiang et al. 2005). Conversely, showing that overexpression or overactivation of a candidate molecule leads to the emergence of multiple neurites displaying the molecular identity of an axon is generally interpreted to suggest that this molecule is sufficient to confer axon identity. However, this approach is limited because it relies on overexpression, which can be complicated by abnormal activation of a pathway normally not involved in axon specification or neuronal polarity. Given these technical issues, a more biologically relevant validation should include the test of the requirement of a candidate gene for neuronal polarity in vivo, or ex vivo, using gene knockout or shRNA-mediated knockdown technologies (Barnes et al. 2007).
Spatial regulation of protein expression by sel ective degradation has been shown in several contexts during neuronal development, including: axonal pruning (Watts et al. 2004); various aspects of axon guidance (Campbell and Holt 2001; DiAntonio et al. 2001; Bloom et al. 2007; Lewcock et al. 2007); synapse formation (DiAntonio et al. 2001; Nakata et al. 2005); synapse maintenance (DiAntonio et al. 2001; Aravamudan and Broadie 2003; Ehlers 2003; Speese et al. 2003); and synapse elimination (Ding et al. 2007) (reviewed in DiAntonio and Hicke 2004). Acute treatment with the proteosome inhibitor lactacystin blocks axogenesis in dorsal root ganglion cells (Klimaschewski et al. 2006). The protein kinase AKT is critical for neuronal polarity (see the following) and appears to undergo selective degradation (Yan et al. 2006), which selectively targets the inactive pool of AKT in neurites, resulting in a net enrichment of activated AKT in a single process becoming the axon. This phenomenon is consistent with the negative feedback signal model proposed by Kaibuchi and colleagues (Arimura and Kaibuchi 2007) to explain specification of a single axon during neuronal polarization. Schwamborn et al. recently showed that the small GTPase Rap1b is regulated by a similar scheme because the active form of Rap1b is spared from degradation and ultimately enriched in the axon (Schwamborn et al. 2007b). Additional work has shown that an interaction between Smurf2 and the polarity scaffold PAR3 must exist for proper neuron polarization (Schwamborn et al. 2007a). More complete understanding of this mechanism will require identification of both upstream regulatory signals responsible for its spatially limited activity and the cohort of proteins targeted for degradation in the nascent axon and dendrites.
Another potential mechanism that could locally regulate proteins involved in axon specification or axon growth is local translation. The presence of mRNA and the translation machinery in developing axon and dendrites has been well-established (Job and Eberwine 2001), and their role in local protein synthesis is starting to emerge as an important molecular mechanism underlying synaptic plasticity (Sutton and Schuman 2006) and axon guidance (Lin and Holt 2008). However, the functional importance of local protein synthesis in axon specification and axon growth is still debated (Hengst and Jaffrey 2007). A recent study, using compartmentalized application of axon growth-stimulating cues only to axons, shows that intra-axonal protein translation is required for stimulated, but not basal, axon outgrowth (Hengst et al. 2009). Interestingly, the authors found that local translation of Par3 mRNA is triggered following both NGF or Netrin-1 application (Hengst et al. 2009). The definition of the repertoire of mRNA undergoing local translation, and the characterization of the functional importance in vitro and in vivo of these locally translated mRNAs, will undoubtedly improve our understanding of the molecular mechanisms underlying axon elongation.
Appropriate regulation of the actin and microtubule cytoskeleton is critical for neuronal polarization and has been the focus of numerous studies. Experiments using the actin-destabilizing agents lactrunculin B and cytochalasin D indicate that remodeling of the actin-based cytoskeleton is an important regulatory step in axon formation (Bradke and Dotti 1999). Specifically, actin-depolymerization localized to a single neurite in unpolarized stage 2 hippocampal neurons is sufficient to confer axonal identity. For more details, we refer to Gertler (2010).
APC (Adenomatous polyposis coli) is a well-established effector of GSK-3 that is enriched in the neurite that will become the axon early in neuronal polarization (Shi et al. 2004). Most cell biological evidence suggests that APC enhances microtubule stability and that APC can bind microtubule plus ends via its EB1 binding domain (Aoki and Taketo 2007). Phosphorylation of APC by GSK-3β reduces its ability to bind the plus ends of microtubules, and inhibition of GSK-3β leads to an accumulation of APC in multiple neurites (Shi et al. 2004). Expression of truncated forms of APC is sufficient to inhibit axon formation and elongation (Shi et al. 2004; Zhou et al. 2004; Purro et al. 2008). Recent work suggests that the APC/GSK-3 dyad regulates targeting of another polarity protein PAR3 (Shi et al. 2004; Zhou et al. 2004). Further, growth factor-triggered inactivation of GSK-3β by PI3K signaling acts through APC to control axon elongation (Zhou et al. 2004). Two other GSK-3 targets, the microtubule associated proteins (MAPs) MAP1b (Gonzalez-Billault et al. 2004) and Tau (Sperber et al. 1995), display reduced microtubule binding when phosphorylated by GSK-3. These results emphasize that the microtubule cytoskeleton is a major endpoint for polarity regulators. Interestingly, PTEN was also recently identified as a GSK-3β substrate (Maccario et al. 2007), and this may define a negative feedback loop for AKT signaling following activation via stabilization of PTEN.
A pioneering genetic screen performed by Kemphues and colleagues in the late 1980s identified six Par genes encoding distinct protein families that play critical roles in epithelial cell polarity during development and metastasis (Kemphues et al. 1988; Goldstein and Macara 2007). Par4 (also known as the tumor-suppressor LKB1 or STK11) translocates from the nucleus and is activated by hetero-dimerization with one of two related pseudokinases known as Stradα and β (Dorfman and Macara 2008). In addition to binding Strad, LKB1 function in neuronal polarity requires its phosphorylation at S431, a target of both protein kinase A and the p90RSK kinase (Collins et al. 2000; Sapkota et al. 2001). This phosphorylation event can be triggered by extracellular cues such as BDNF (brain derived neurotrophic factor) (Shelly et al. 2007) (Fig. 2). Future investigations will identify the relevant extracellular cues and the corresponding signaling pathways that trigger phosphorylation of LKB1 in position S431, thereby specifying the axon in developing cortical pyramidal neurons in vivo.
Once LKB1 is activated by binding to its necessary coactivator Strad, and S431-phosphorylation occurs (only in the neurite becoming the axon), LKB1 phosphorylates SAD-A/B kinases. In addition, LBK1 probably also phosphorylates microtubule affinity-regulated kinases, MARK1-4 (Matenia and Mandelkow 2009), which are required in part for axon specification because they phosphorylate microtubule-associated proteins such as Tau. On the basis of SAD kinase function in presynaptic vesicular clustering observed in C. elegans (Crump et al. 2001), we hypothesize that SAD-A/B kinases specify axon identity by directing presynaptic vesicular trafficking in the neurite becoming the axon. Most importantly, genetic deletion of LKB1 in cortical pyramidal neurons prevents axon formation, whereas overexpression of LKB1 and its coactivator Strad in neural progenitors or LKB1 alone in postmitotic cells is sufficient to lead to the formation of multiple axons (Asada et al. 2007; Barnes et al. 2007; Shelly et al. 2007).
LKB1 also phosphorylates and activates a family of 13 protein kinases related to the C. elegans PAR1 protein (Lizcano et al. 2004). To date, three of these have been implicated in regulating axon formation: SAD-A and SAD-B kinases, and also MARK-2 (microtubule affinity regulating kinase-2; also called Par1b). RNAi knockdown of SAD kinases partially abrogates the ability of LKB1 overactivation to induce the formation of multiple axons in cortical neurons, indicating that LKB1's function in promoting axogenesis largely derives from activation of SAD-A/B kinases (Barnes et al. 2007). Double knockout mice for SAD-A and SAD-B results in neurons that cannot form axons in vivo (Kishi et al. 2005). SAD and MARK kinases target several MAPs, including MAP2, MAP4, and Tau, by phosphorylating three K-X-G-S motifs within each protein, reducing their microtubule binding affinity and thus destabilizes microtubules (Illenberger et al. 1996; Drewes et al. 1997; reviewed in Matenia and Mandelkow 2009). Little is known about SAD kinase regulation. However, a recent study implicates the tuberous sclerosis complex (TSC) genes TSC1/2 in regulating SAD protein abundance (Choi et al. 2008). The microtubule regulatory scheme is the same for the four members of MARK kinase family, but to date only MARK2 has been implicated in the regulation of neuronal polarity (Biernat et al. 2002; Chen et al. 2006).
The core components of the other major proteins in the polarity complex identified in C. elegans are the scaffolding proteins PAR-3 and PAR-6. Many binding partners for this protein complex have been implicated in regulating the polarity of epithelial cells. Neuroepithelial radial glia target the PAR3/6 complex in their apical domain along the ventricular wall (Manabe et al. 2002; Costa et al. 2008). Proteins reported to exist in a complex with PAR3/6 include atypical forms of protein kinase C (aPKC: PKCλ and ζ) (Joberty et al. 2000; Lin et al. 2000; Qiu et al. 2000), the small GTPase cdc42 (Joberty et al. 2000; Lin et al. 2000; Qiu et al. 2000), the kinesin motor protein KIF3A (Nishimura et al. 2004), the guanine exchange factor Tiam1/STEF (Chen and Macara 2005; Nishimura et al. 2005), the lipid and protein phosphatase PTEN (von Stein et al. 2005; Feng et al. 2008), the GTPase activating protein (GAP) p190RhoGAP (Zhang and Macara 2008), the tumor suppressor lethal giant larvae (lgl) (Plant et al. 2003), the scaffold protein inscuteable (Schober et al. 1999), the ubiquitin ligases Smurf1 (Ozdamar et al. 2005) and Smurf2 (Schwamborn et al. 2007b), and the transforming growth factor receptor 1 (TGFβR1) (Ozdamar et al. 2005). Each of these proteins has also been implicated in controlling polarity in nonneuronal cells as part of the PAR3/6 complex.
We consider the potential function of some of these PAR3/6 interacting proteins (Lgl, Inscuteable, and Smurf1/2) later in this review in the context of progenitor polarity and protein stability, respectively. PAR3/6 are enriched in the nascent axon in stage 3 hippocampal neurons, and overexpression of wild-type and truncated forms of either PAR3 or PAR6 perturb the formation of a single axonal process in hippocampal neurons (Shi et al. 2003). However, in Drosophila, orthologs of PAR3 (bazooka), PAR6, or aPKC do not appear to be required for proper axon–dendrite specification (Rolls and Doe 2004). This could mean that PAR3/6 have acquired a function in neuronal polarity late in the evolution of the vertebrate radiation. Further, there is so far no genetic loss-of-function evidence in vertebrates (especially in mammals) demonstrating that Par3 and Par6 are required for axon specification. This evidence is clearly more challenging to obtain in mammals than in Drosophila because of potential genetic redundancy: There are four Par6 genes and two Par3-like genes in mammalian genomes (Goldstein and Macara 2007; Barnes et al. 2008).
Small GTPases are critical regulators of cytoskeletal and membrane dynamics underlying cell motility, cell polarity, and cell growth. Not surprisingly, both Ras- and Rho-family small GTPases have been shown to be involved in axon specification and axon growth. We refer the reader to Hall and Lalli (2010) for a comprehensive review of the roles played by small GTPases in the regulation of neuronal morphology.
The phosphatidylinositol-3 kinase (PI3K) family regulates diverse biological functions. These include cell polarity, cell motility, and chemotaxis, as well as neuronal migration and polarization; however, these observations are drawn almost exclusively from the use of pharmacological inhibitors such as Wortmanin or LY294002 (Polleux et al. 2002; Shi et al. 2003; Jossin and Goffinet 2007; Zhou et al. 2007). The best characterized class Ia PI3-kinase (PI3KcIa) is involved in the formation of phosphatidylinositol (3,4,5)-triphosphate (PIP3), and it lies downstream of Ras and upstream of protein kinase B (or AKT) during signal transduction. Work from several groups implicates PI3K in axon specification based on the fact that pharmacologic inhibition of PI3K activity using LY294002 or Wortmannin prevents axon formation (Shi et al. 2003; Menager et al. 2004; Jiang et al. 2005; Yoshimura et al. 2006). However, these data have to be interpreted carefully because these inhibitors are not class-specific. They inhibit the three main classes of PI3K (Stack and Emr 1994), including phosphatidylinositol-3 kinase class III (PI3Kc3, also called Vps34) that produces exclusively phosphatidylinositol-3 monophosphate (PI(3)P) and regulates endocytosis, vesicular trafficking, trimeric G-protein signaling, and the mTOR (mammalian target of rapamycin) pathway. Using the pleckstrin homology (PH) domain of AKT fused to GFP (PHAKT-GFP) as a biosensor for PIP3 formation, Menager et al. (2004) have shown that PIP3 accumulates selectively within a single neurite following local application of laminin to a single neurite of stage 2 hippocampal neurons. Future work using genetic loss-of-function approaches will address when and where PI3K activation occurs in vivo during neuronal polarization, and also which class of PI3K is required (class 1, 2, or 3).
PTEN (Phosphatase and Tensin homolog deleted on chromosome 10) is a lipid and protein phosphatase that acts in direct opposition to PI3K activity because PTEN dephosphorylates PIP3, generating PIP2 and thus limiting PIP3 signaling both spatially and temporally. Increasing levels of PTEN expression leads to a loss of axon formation (Shi et al. 2003; Jiang et al. 2005), whereas reduction of PTEN expression via RNAi-mediated knockdown leads to a multiple axon phenotype (Jiang et al. 2005). This effect is consistent with the gain-of-function mutation of PI3K described previously and highlights the critical need for maintaining the delicate balance of phospholipid composition at the membrane to ensure proper neuronal polarization and axon formation. PTEN, remarkably, via regulation of the mTOR pathway, has a strong function in suppressing injury-induced axon growth in CNS neurons (Park et al. 2008).
Several proteins are recruited via their PIP3-specific PH domains to membrane sites created by PI3KcIa activity. The protein kinase AKT, also called protein kinase B (PKB), undergoes such a translocation to the membrane via its PH domain. This step is required for AKT's activation by dual phosphorylation on T308 and S473 by the membrane-targeted protein kinases PKD1 and PKD2, respectively. This activated form of AKT is enriched in growth cones of polarized neurons (Shi et al. 2003). When a myristoylation site is added to recombinant AKT (myr-AKT), it is constitutively targeted to the membrane, independent of PI3K, and therefore is constitutively active. When overexpressed in neurons, this form of AKT is sufficient for multiple axon formation (Yoshimura et al. 2006), consistent with a unified pathway in which AKT acts downstream of PI3K in regulating axon formation.
GSK-3 (glycogen synthase kinase 3) is a well-studied serine/threonine protein kinase that functions in the regulation of multiple intracellular processes including pathways downstream of receptor tyrosine kinases and Wnt/Frizzled signaling. Two genes encoding GSK-3 (α and β) in mammals perform essentially redundant functions. GSK-3 has the unusual property of being constitutively active, a state that is reversed following phosphorylation at Ser9 in GSK-3β, or Ser21 in GSK-3α, by multiple kinases including AKT, ILK, and atypical protein kinase C (aPKC) (Etienne-Manneville and Hall 2003). Recent in vitro work implicates GSK3β as a critical regulator of neuronal polarity. Experiments using several types of GSK-3 inhibitors indicate that GSK-3α/β act as negative regulators of axon formation because they lead to formation of multiple axons (Jiang et al. 2005; Yoshimura et al. 2005).
A recent study suggests that the situation may be somewhat more complex in vivo. Using double knockin mice bearing single point mutations in GSK-3βS9A and GSK-3αS21A, Gartner et al. reported no obvious deficits in neuronal morphogenesis in vivo, or in vitro (Gartner et al. 2006). However, using inhibitors of GSK-3α/β such as lithium chloride or, more specifically, SB-415286, SB-216763, and AR-A014418, Gartner et al. (2006) were able to replicate the multiple axon phenotype obtained by others (Jiang et al. 2005; Garrido et al. 2007). These results indicate that although the exact role of Ser9/Ser21 phosphorylation in GSK-3 inactivation remains to be understood, or may involve an alternate site (Thornton et al. 2008), it is clear that the catalytic activity of GSK-3 is a critical regulator of neuronal polarity in these in vitro paradigms.
Several downstream targets of GSK-3 are potential effectors of neuronal polarity, and many involve regulation of the cytoskeleton. Collapsin-response mediator protein-2 (CRMP-2) is one such microtubule-binding protein that is enriched in tip of the nascent axon and is regulated by GSK-3β, such that phosphorylated CRMP-2 displays a decreased binding affinity for tubulin heterodimers (Inagaki et al. 2001; Yoshimura et al. 2005; reviewed in Arimura et al. 2004). As seen for other polarity regulators, overexpression of CRMP-2 is sufficient to induce the formation of multiple axons, and truncated forms of CRMP-2 can impair axon formation (Inagaki et al. 2001).
A key issue for efficient axon elongation is regulation of microtubules polymerization. Although this question is only beginning to be addressed in primary neurons, there are some indications about kinases that may orchestrate this process. Abundant in vitro evidence suggests a potentially important role for GSK-3. GSK-3 has multiple substrates that have the potential to regulate tubulin polymerization and microtubule stability. These include the microtubule plus-end binding proteins APC and CLASP2, CRMP-2, which is localized to axon tips and may play a role in cargo delivery and tubulin polymerization, and the microtubule associated proteins MAP1B and tau (reviewed in Zhou and Snider 2006). Details of how these proteins function and how they are regulated by GSK-3 phosphorylation remain obscure. However, abundant in vitro evidence implicates GSK-3 as having an important role in axon elongation.
Virtually all studies indicate that low doses of GSK-3 inhibitors induce axon branching. This may be a counterpart to the elaboration of multiple axons in the hippocampal paradigm outline above. Normally, GSK-3 inhibition and APC plus-end binging are localized to the axon growth cone downstream of PI3K signaling (Eickholt et al. 2002; Zhou et al. 2004; Cosker et al. 2008). More general inhibition of GSK-3 may induce multiple foci along the axon capable of supporting efficient tubulin polymerization, thus resulting in branching. In DRG neurons, cGMP is an important mediator of axon branching, and a key cGMP dependent protein kinase (PrkG1) directly binds to and regulates GSK-3 activity (Zhao et al. 2009). There is a clear in vivo correlate to these observations, because elimination of PrkG1 strongly reduces the normal bifurcation of the central process of DRG neurons. Importantly, there appear to be other in vivo correlates of this regulation. Wnt3 can induce branching of the central process of proprioceptive DRG neurons, and it seems reasonable that GSK-3 inhibition mediated via canonical Wnt signaling is a critical component of this branch induction (Krylova et al. 2002).
As noted previously, two GSK-3 family members, GSK-3α and GSK-3β, have largely redundant functions. Elimination of either family member alone has little effect on axon growth, either in vitro or in vivo (Kim et al. 2006; Kim et al. 2009). However, complete inhibition of GSK-3 activity via inhibitors or shRNA directed at both family members appears to block axon growth altogether in vitro (Shi et al. 2004; Kim et al. 2006; Garrido et al. 2007). Presumably, some baseline level of GSK-3 activity is required for efficient microtubule assembly. Floxed alleles for both GSK-3α and GSK-3β have recently been generated (Kim et al. 2009), and therefore GSK-3 functions in axon growth in vivo should be defined shortly. GSK-3 regulation of axon growth may have important clinical significance because GSK-3 inhibition has produced modest, but statistically significant, improvement in rodent spinal cord injury models (Dill et al. 2008).
It is logical to suppose that many of the principles associated with axon initiation are also operative in the setting of axon growth and extension toward target fields. In the case of RAS and its downstream effector, the RAF/MEK/ERK cascade, abundant in vitro evidence supports an important function in axon growth. Thus, many studies with the major pharmacological inhibitors PD98059 and U0126 show striking inhibition of axon growth induced by neurotrophins and other factors acting via receptor tyrosine kinases (see Zhou and Snider 2006, for a review). Further, both loss and gain of function studies in vitro suggested strong axon growth promoting functions for Ras, Raf, and ERK/MAPK (Markus et al. 2002). Indeed, ERK/MAPK appear well-positioned to regulate axon growth via its well-characterized regulation of a broad range of transcription factors, including SRF, and local regulation of the axonal cytoskeleton via phosphorylation of several microtubule-associated proteins. Perhaps surprisingly, evidence to date indicates a role for ERK/MAPK signaling in target field arborization but not in long-range axon extension. Thus, early ablation of the ERK/MAPK mediator SRF in DRG neurons using a Cre recombinase active in neural crest cells showed a striking ability of DRG neurons to extend long distances into the hindlimb during embryonic development (Wickramasinghe et al. 2008). However, axonal arborization into cutaneous target fields was abolished. These findings indicate that initial axon extension and also long-range growth can occur in the absence of SRF, but when axons encounter NGF ERK/MAPK SRF signaling, SRF is required for growth responses. More recent studies eliminating both ERK1 and ERK2 also show that long-range axon growth in both the PNS and CNS appears unimpeded in the absence of ERK in many classes of neurons (Samuels et al. 2008; Newbern and Snider, unpubl. observations).
Recent findings have revealed a critical role for c-Jun amino-terminal kinase (JNK) signaling, classically considered a degenerative signal and a stress-response signaling pathway, in the regulation of axonal outgrowth. Analysis of the effects of JNK deletion on axonal elongation in vivo has been complicated by the presence of three partially redundant JNK isoforms, and also by JNK's key role in cell death (Kuan et al. 1999; Sabapathy et al. 1999; Brecht et al. 2005). Nonetheless, deletion of a single isoform, JNK1, disrupts axon tract maintenance in vivo (Chang et al. 2003; Bjorkblom et al. 2005). The role of JNKs in axonal polarity and initiation is further supported by many cell culture studies showing that JNK inhibition attenuates neuritogenesis (Leppa et al. 1998; Waetzig and Herdegen 2003; Eom et al. 2005; Oliva et al. 2006; Newbern et al. 2007). Numerous lines of evidence suggest this pathway is also important for axonal extension (Leppa et al. 1998; Rosso et al. 2005; Xiao et al. 2006; Ciani and Salinas 2007; Eminel et al. 2008). JNKs show constitutively high activity within neurons and phosphorylate various cytoskeletal proteins involved in axon extension, including MAP1B, MAP2, tau, and SCG10, among others (Kyriakis and Avruch 1990; Otto et al. 2000; Neidhart et al. 2001; Tararuk et al. 2006; Yamauchi et al. 2006; Ciani and Salinas 2007). Axonal transport is modulated by JNK, and it has been proposed that JNK triggers the release of cargoes, such as tubulin, from kinesin complexes (Stagi et al. 2006; Horiuchi et al. 2007). It is clear that the repertoire of JNK substrates is well-suited to mediate many aspects of axonal development.
Transcription is a key aspect of axonal elongation, targeting, and regeneration after injury (Polleux et al. 2007). Surprisingly, despite intensive transcriptional profiling efforts, few transcriptional regulators of axon extension have been identified. One elegant mechanism has been defined involving the anaphase promoting complex (APC), which functions as an E3 ubiquitin ligase in the regulation of cyclins and other mediators of cell-cycle control. In postmitotic neurons, APC and its activator protein, Cdh1, stimulate degradation of the transcription factor SnoN (Konishi et al. 2004; Stegmuller et al. 2006). SnoN levels are also regulated by TGF-β/SMAD-2 signaling (Stegmuller et al. 2008). Knockdown experiments in vivo show that reduction of Cdh1 enhances axon growth of cerebellar granule cells and, conversely, reduction of SnoN expression reduces granule cell axon growth. One important target of SnoN is the scaffolding protein Cdc1, which controls JNK activity (Ikeuchi et al. 2009). Cell biological work in vitro, and also shRNA experiments in vivo, strongly point to SnoN as a key transcriptional regulator of axon growth during nervous system development. However, genetic loss-of-function mouse models will be needed to fully appreciate the requirement of this signaling pathway for axon elongation in these and other neuronal subtypes in vivo.
Other transcriptional regulators that have been documented to control axon extension include CREB and NFAT transcription complexes (Lonze et al. 2002; Graef et al. 2003). The role of the NFATc family is particularly striking in vivo. In mice lacking 3 NFATc family members, peripheral axon extension fails almost completely at early developmental stages. Further, longitudinal sensory axon tracts do not form in the spinal cord, and commissural neuron axons do not cross the midline near the floor plate. In vitro experiments reveal that both NGF and Netrin1-mediated responses are abolished (Graef et al. 2003). Importantly, the loss of NFATc family members can be mimicked by inhibiting Calcineurin, and it is important to note that NFATc subunits are controlled by both ERK/MAPK and GSK-3 phosphorylation (Crabtree and Olson 2002). To date, the role of NFATc family members has been studied only at the earliest stages of axon growth because of the lack of viability of the null mutants beyond mid-embryonic stages.
Is there any in vivo evidence for the role of extracellular cues in the specification of neuronal polarity? Important progress toward our understanding of the molecular and cellular mechanisms specifying axon initiation during neuronal polarization has been made in C. elegans (Quinn and Wadsworth 2008). This work has markedly enhanced our understanding of how extracellular cues instruct axon initiation in vivo. The neurons of the nematode have a stereotyped morphology, e.g., specific projections along the dorso–ventral and anterior–posterior body axes. Elegant experiments have identified an extracellular cue, UNC-6 (netrin), along with its receptor UNC-40 (DCC), as critical for orchestrating axon initiation in vivo (Adler et al. 2006). This work also identified downstream proteins in this pathway, including (mammalian orthologs are shown in parenthesis where established): AGE-1 (PI3K), DAF-18 (PTEN), UNC-34 (Enabled), CED-10 (Rac), UNC-115/AbLIM, and MIG-10/Lamellipodin. The current model for the relationship of these genes to UNC-6 signaling involves DAF-18's limitation of AGE-1 activity following UNC-40 stimulation and the asymmetric recruitment of MIG-10 to the plasma membrane. This recruitment requires activated CED-10 (Rac) directly binding to MIG-10 (Quinn et al. 2008) and also the involvement of the PAK-like kinase Pak-1 (Adler et al. 2006). Another regulator thought to act in concert with MIG-10 to drive directed filopodial formation is the Enabled homolog, UNC-34 (Chang et al. 2006). SLT-1 (Slit) is another extracellular cue that likely acts through MIG-10 recruitment to control neuronal polarization (Chang et al. 2006).
Two other studies have identified the diffusible signal Wnt and its receptor as critical regulators of axon specification and neuronal polarity (Hilliard and Bargmann 2006; Prasad and Clark 2006). In addition to identifying loss-of-function for Lin-44 (Wnt) and its receptor Lin-17 (Frizzled) as being required for these processes, one of the screens also identified VPS-35, a component of retromer complex that regulates vesicular traffic and is required for proper Wnt secretion (Prasad and Clark 2006; Pan et al. 2008).
Careful live imaging experiments of Xenopus retinal ganglion cell polarization revealed that polarized axon outgrowth requires some unidentified extracellular cues present in the basal lamina (Zolessi et al. 2006). This work strongly suggests that: (a) the basal lamina contains important extracellular cues that play a role in the polarized emergence of the axon of RGC neurons and that (b) RGC neurons inherit the intrinsic apico-basal polarity of their progenitor, at least with regard to the Par3/aPKC components of the polarity complex.
Several lines of evidence suggest that extracellular cues can direct the polarized emergence of the axon and dendrites both in vitro and in vivo. One paradigm involves dissociated cortical or hippocampal pyramidal neurons plated on striped substrates coated with two different cell adhesion molecules (Laminin and NgCAM, for example) (reviewed in Barnes and Polleux 2009). The first immature neurite of E18 hippocampal neurons that contacts the boundary between two stripes reproducibly becomes the axon (Esch et al. 1999). Using a similar approach, Shelly and colleagues showed that neurites of immature hippocampal neurons growing on a patterned substrate can detect the presence of BDNF, which plays an instructive role in axon specification because the first neurite contacting a BDNF stripe reproducibly becomes the axon (Shelly et al. 2007). The effect of BDNF on axon specification requires cAMP-dependent protein kinase (PKA) activation and phosphorylation of LKB1 in position 431 by PKA (Shelly et al. 2007), suggesting that LKB1 phosphorylation on S431 acts as a detector of a break in neuronal symmetry induced by extracellular cues such as BDNF in this in vitro context.
Using a slice overlay assay, Polleux et al. have shown that only a couple of hours after plating, the vast majority of cortical neurons displayed a single short axon directed ventrally toward the ventricle (Polleux et al. 1998), as is found in vivo (see Fig. 1). These authors went on to show that the class 3 secreted semaphorin, Sema3A, which is enriched in the most superficial part of the cortical wall (the top of the cortical plate) (see also Chen et al. 2008), plays a role in repulsing axon initiation ventrally toward the ventricle (Polleux et al. 1998). This work suggests that the polarized emergence of a single axon is controlled at least in part by extracellular cues expressed in a graded manner along the neuron's migratory path.
A central unresolved issue is whether or not neuronal activity is required for proper axon elongation in vivo. Careful in vivo analyses provide strong evidence suggesting that for specific neuronal subtypes, such as retinal ganglion cells (RGC), axon growth stimulated by various neurotrophic is normally remarkably slow, but profoundly potentiated by physiological levels of electrical activity, either spontaneously generated within explants or mimicked by multielectrode silicon chip stimulation (Goldberg et al. 2002). In vivo, elegant imaging studies have provided direct evidence for a negative relationship between the frequency of spontaneous calcium transients and the rate of axon outgrowth (Gomez et al. 1995; Gomez and Spitzer 1999). The relevance of these findings to other neuronal populations in various organisms will need to be explored before these findings can be generalized; however, modulation of neuronal activity is an attractive mechanism for regulating axon growth and targeting in vivo by tuning the response to extrinsic axon growth-promoting cues and axon guidance cues.
Once axon specification occurs, the axonal process begins to grow at a rate approximately 10 times faster than the other processes. Although there is some intrinsic capability of a process to extend at a rapid rate, most experimental paradigms that have employed highly reduced conditions indicate that extracellular factors are required for the gene transcription and cytoskeletal assembly that underlies rapid axon extension (Lentz et al. 1999; Goldberg et al. 2002; Ozdinler and Macklis 2006).
Numerous in vitro studies, mostly of the gain-of-function variety, show dramatic effects of several neuronal growth factors on neuronal classes bearing receptors for these molecules (reviewed in Chao 2003; Reichardt 2006). An early complication in interpreting such studies was the separation of survival versus growth-promoting effects. However, recently, there has been broad use of the manipulation of the Bcl-2 apoptosis pathway to address this issue. Neuronal survival can be maintained indefinitely in the absence of exogenous factors using overexpression of Bcl-2, Bcl-X, or elimination of Bax (reviewed in Akhtar et al. 2004; Youle and Strasser 2008). Using this approach, multiple studies have now established robust axon-growth promoting effects of growth factors acting through receptor tyrosine kinases and cytokines (Lentz et al. 1999; Goldberg et al. 2002).
Despite these robust effects, the in vitro studies showing the ability of several factors to trigger growth have to be interpreted with caution. First, factors that show strong effects under reduced conditions in vitro may not be required in the rich in vivo environment. Axons are exposed to a myriad of diffusible and surface anchored growth-promoting molecules in vivo that may compensate for the loss of a specific factor. Second, the ability to trigger axon growth in a dish usually does not establish when and where these factors regulate axon growth along its trajectory in vivo. Robust growth promoting effects in vitro may be irrelevant to axon extension but may indicate, instead, that a factor is required for induction of collateral branching (Wang et al. 1999; Brose and Tessier-Lavigne 2000) or for induction of terminal branching near target cells (Hall et al. 2000; Krylova et al. 2002; Salinas and Zou 2008). Finally, if a rich mixture of surface molecules (e.g., Matrigel) is presented to the neuron, growth promoting effects of any specific factor may be difficult to detect, even when such molecules are known to be important for axon growth or targeting in vivo (Graef et al. 2003; Wickramasinghe et al. 2008).
Mouse genetic studies have provided accumulating evidence for in vivo roles for neurotrophins and GDNF family members. Interestingly, the earliest phases of PNS axon growth are clearly neurotrophin independent. Thus, BDNF and NT-3–sensitive trigeminal neurons extend toward their target fields in the periphery and are sensitive to both TrkB and TrkC deletion (O'Connor and Tessier-Lavigne 1999). GDNF family members may have a role in early phases of extension of sympathetic ganglion cell axons, but whether this generalizes to other PNS neurons is not yet certain (Honma et al. 2002; Kuruvilla et al. 2004; Luo et al. 2007). Later stages of axon growth along vessels appear to require NT-3 acting locally (Kuruvilla et al. 2004). Similarly, within the CNS, in vivo evidence does not support a role for neurotrophins or GDNF family members in the earliest stages of axon extension (Chao 2003). Gain of function studies show strong effects of neurotrophins on axon branching, but in only one case has a known role for a neurotrophin related to an axon trajectory been established (Ma et al. 2002).
Although the early phases of axon growth may be neurotrophin-independent, later stages of axon growth associated with growth of limbs also clearly require NT signaling (Patel et al. 2000; Tucker et al. 2001; Patel et al. 2003). Furthermore, it is now clearly established that NGF and NT-3 regulate target field innervation. In mice lacking NGF or TrkA, where neuronal survival is preserved by absence of BAX, NGF-responsive axons do not invade cutaneous target fields. Similarly, proprioceptive axons do not innervate muscle spindles in the absence of NT-3 or TrkC (Patel et al. 2003). Finally, in the setting where the ERK/MAPK effectors RAF kinase and SRF are absent, terminal axonal arborization, but not long range extension, is impaired at embryonic stages (Zhong et al. 2007; Wickramasinghe et al. 2008).
Within the CNS, studies of mice harboring null and conditional mutants for members of the neurotrophin or GDNF families have not shown significant effects on axon extension and targeting. An exception is the central projections of DRG neurons, where elimination of NT-3 in concert with elimination of BAX abrogates central branching of proprioceptive axons via regulation of the transcription factor ER81 (Patel et al. 2003). There are numerous other examples in the CNS where elimination of transcription factors affects axon guidance and targeting (Polleux et al. 2007), but whether these effects are controlled by growth factor signaling is still poorly understood. Finally, neutralizing IGF1 receptor signaling with direct antibody injections has shown failure of axon elongation and targeting of CST axons in the spinal cord (Ozdinler and Macklis 2006). This suggests that RTK signaling and pathways considered above may be important for growth and targeting of CNS axon tracts. However, for many CNS neurons, precise factors that regulate axon extension remain to be identified.
Our knowledge of the intrinsic molecular mechanisms underlying the ability of neurons to polarize by extending a single axon and multiple dendrites has clearly improved over the past two decades (Craig and Banker, 1994; Arimura and Kaibuchi 2007). In contrast, our understanding of how these intrinsic signaling pathways are regulated by the action of extracellular cues during neuronal polarization in vivo is still fragmented (Barnes and Polleux 2009). Future work will clearly identify: (1) the extracellular cues and cellular mechanisms underlying axon and dendrite specification in mammalian CNS neurons in vivo; (2) how these extracellular cues regulate the spatial pattern of activation of specific signaling pathways during neuronal polarization; and (3) how axon growth is regulated in vivo during normal development and following injury.
Editors: Marc Tessier-Lavigne and Alex L. Kolodkin
Additional Perspectives on Neuronal Guidance available at www.cshperspectives.org