Inhibitors of neuronal regeneration: mediators and signaling mechanisms

Transcription

1 Neurochemistry International 42 (2003) Review Inhibitors of neuronal regeneration: mediators and signaling mechanisms Bor Luen Tang a,b, a NCA Laboratory, Institute of Molecular and Cell Biology, 30 Medical Drive, Singapore , Singapore b Department of Biochemistry, National University of Singapore, 10 Kent Ridge Crescent, Singapore , Singapore Received 28 January 2002 Abstract Neuritogenesis and its inhibition are opposite and balancing processes during development as well as pathological states of adult neuron. In particular, the inability of adult central nervous system (CNS) neurons to regenerate upon injury has been attributed to both a lack of neuritogenic ability and the presence of neuronal growth inhibitors in the CNS environment. I review here recent progress in our understanding of neuritogenic inhibitors, with particular emphasis on those with a role in the inhibition of neuronal regeneration in the CNS, their signaling cascades and signal mediators. Neurotrophines acting through the tropomyosin-related kinase (Trk) family and p75 receptors promote neuritogenesis, which appears to require sustained activation of the mitogen activated protein (MAP) kinase pathway, and/or the activation of phosphotidylinositol 3-kinase (PI3 kinase). During development, a plethora of guidance factors and their receptors navigate the growing axon. However, much remained to be learned about the signaling receptors and pathways that mediate the activity of inhibitors of CNS regeneration. There is growing evidence that neuronal guidance molecules, particularly semaphorins, may also have a role as inhibitors of CNS regeneration. Although direct links have not yet been established in many cases, signals from these agents may ultimately converge upon the modulators and effectors of the Rho-family GTPases. Rho-family GTPases and their effectors modulate the activities of actin modifying molecules such as cofilin and profilin, resulting in cytoskeletal changes associated with growth cone extension or retraction Elsevier Science Ltd. All rights reserved. Keywords: Actin; Neurite; Neurotrophin; Nogo; Rac; Rho; Semaphorin 1. Introduction The sprouting of neurites, the growth of an axon and the extension of a dendritic tree are key morphological features characterizing neuronal development. The opposite processes of growth cone retardation and neurite retraction, are no less important in mediating neuronal path finding, in the eventual formation of synaptic connections and the subsequent establishment of new connections. Neurite outgrowth and retraction are also important for neuronal plasticity as well as neuronal regeneration (or failure to regenerate) from injuries or neuropathological conditions. The in vitro differentiation to a neuronal-like phenotype by PC12 cells, a phaeochromocytoma line, has been a popular cell culture model for neuritogenesis (or its inhibition) in vitro. Upon treatment of a PC12 culture with nerve growth factor (NGF), a morphologically differentiated phenotype characterized by the outgrowth of neurites extending for several cell bodies in length is obvious after 3 4 days. The Tel.: ; fax: address: mcbtbl@imcb.nus.edu.sg (B.L. Tang). model is limited by the inability of these outgrowths to eventually form functional synaptic connections. A more physiological cell culture model for neuronal growth is a rodent hippocampal culture. Careful morphological studies reveal several distinct phases in neuronal growth upon plating of hippocampal cells (Bradke and Dotti, 2000). The above mentioned and similar cell culture models have been used extensively to study the process of neuritogenesis. Neurite outgrowth in vitro can usually be initiated by the binding of a neurotrophic ligand to its receptor, resulting in cascades of signal transduction events that coordinate changes in cellular cytoskeleton, which would in turn mediate the growth process. However, not all growth factors capable of initiating a mitogenic signaling cascade in neurons are capable of inducing neurite outgrowth. On the other hand, there exist a large repertoire of both soluble and membrane bound molecules that can bring neurite extension to a halt. The neuronal guidance molecules represent attractive or repulsive cues acting through established cellular receptors. These molecules and their receptors mediate neuronal pathfinding during development and their signaling pathways had come to light in the past few /03/$ see front matter 2003 Elsevier Science Ltd. All rights reserved. PII: S (02)

2 190 B.L. Tang / Neurochemistry International 42 (2003) years. There exist also a diverse class of neuritogenic inhibitors associated with myelin and glial scars that prevail in central nervous system (CNS) lesions and are implicated in the prevention of regeneration of injured CNS neurons. The molecular identity of some of these inhibitors, such as myelin-associated glycoprotein (MAG), Nogo-A and chondroitin sulfate proteoglycans (CSPGs) are now known. However, their transducing receptors and the signaling cascades that they initiate that would lead to growth retardation have yet to be explored extensively. In spite of their diversity, the signaling of all inducers and inhibitors of neuritogenesis must ultimately act by imposing changes in the cellular cytoskeletal components. In cellular terms, this probably means engaging the Rho-family GTPases via their regulatory guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs) and impinging on their effectors such as Rho-associated kinases (ROCKs) and p21-activated kinases (PAKs) at some point. Before looking at neuritogenesis inhibitors in greater detail, let us first have a glimpse of the fascinating and yet incompletely understood process of neuritogenesis. 2. Neurotrophic factors and signal transduction events in neurite outgrowth The mammalian neurotrophines NGF, brain-derived growth factor (BDNF), neurotrophin-3 (NT-3) and neurotrophin-4 (NT-4) bind to and activate two classes of receptors: the tropomyosin-related kinase (Trk) family receptors and the p75 neurotrophin receptor (which is a member of the tumor necrosis factor (TNF) receptor superfamily). Neurotrophins and their receptors appear to be unique for higher organisms and are not found in invertebrates. These ligands and receptors interact with each other in a complex manner to mediate the survival, growth, death and differentiation process of neurons, and are the subjects of several excellent recent reviews (Kaplan and Miller, 2000; Sofroniew et al., 2001; Patapoutian and Reichardt, 2001; Lee et al., 2001). We shall focus our attention here on neurotrophin signaling pathways that lead towards neurite outgrowth in in vitro culture models. Several growth factors that are mitogenic for fibroblasts, such the epidermal growth factor (EGF) are also mitogenic for PC12 or neuroblastoma cells in culture. The point to note is that these are not quite capable of inducing a full differentiation program in PC12 cells as does NGF. For that matter, only some of these, such as PDGF- and insulin are able to induce some kind of limited neurite formation. One of the distinct features of NGF-induced signal transduction cascade in PC12 cells believed to result in the initiation and maintenance of the full differentiation program is the need for sustained activation of a mitogen activated protein (MAP) kinase activation pathway distinct from the classical Raf Ras pathway (Patapoutian and Reichardt, 2001). This pathway is probably initiated by the binding of the adaptor protein, fibroblast growth factor receptor substrate (FRS)-2, to tyrosine-phosphorylated Trk. In a signaling cascade involving molecules such as Crk and C3G, MAP kinase is activated via a C3G/Rap1/B-Raf dependent pathway. Prolonged and sustained MAP kinase activity is essential for differentiation, as transient MAP kinase activity induced by EGF is mitogenic for PC12 cells but does not lead to differentiation. However, sustained MAP kinase activity alone is insufficient for differentiation, which appears to need also the activation of the phosphatidylinositol-3 kinase (PI3 kinase) via the Ras pathway. PI3 kinase is known to affect the activities of GEFs of the Rho-family GTPases. Therefore, as far as neurite outgrowth is concerned, the involvement of PI3 kinase activation by NGF is, in a sense, expected. Early observations that wortmannin (a PI3 kinase inhibitor) inhibits neurite outgrowth appeared to support this notion (Kimura et al., 1994). On the other hand, the fact that a factor such as EGF, which activates PI3 kinase but could not initiate neurite outgrowth, suggests that the signaling process leading to neuritogenesis is not so simple (Raffioni and Bradshaw, 1992). Furthermore, PDGF- mutants that do not bind and activate PI3 kinase could nonetheless promote neurite outgrowth in PC12 cells (Vetter and Bishop, 1995). Gain-of-function experiments involving inducible expression of constitutively active PI3 kinase using a Cre/loxP system (Kobayashi et al., 1997) or microinjection of the active kinase in PC12 cells (Kita et al., 1998), provided evidence that activated PI3 kinase induced the growth of processes that are actually incomplete neurites. These processes lack the accumulation of F-actin and GAP-43 at the growth cones, and possibly resulted from the reorganization of microtubules without concommitent changes in the actin cytoskeleton. One possible explanation for some of the discrepancies discussed above is that inhibition of PI3 kinase may not just act to inhibit the signal that promote neurite outgrowth, but rather generates a signal that actively promotes neurite retraction or growth cone collapse (Sanchez et al., 2001). On the other hand, recent data also suggested an explanation as to why NGF could induce neurite outgrowth whereas EGF could not. There appears to be subsequent downstream differences in terms of activation and the localization of a Rho family member, Rac (Yasui et al., 2001). Therefore, although the link between neurotrophin-induced differentiation and neurite outgrowth is not yet clear in molecular terms, it is likely to ultimately involve the Rho-family GTPases. 3. Neuritogenic inhibitors in lesioned central nervous system It is a well-known fact that unlike neurons of the peripheral nervous system (PNS), CNS neurons almost invariably fail to regenerate after injury (Ramon y Cajal, 1928; Goldberg and Barres, 2000). This could be due, in part, to some intrinsic loss of plasticity in adult CNS neurons, evidenced by the fact that embryonic CNS neurons have a much

3 B.L. Tang / Neurochemistry International 42 (2003) better regenerative capacity (Davies et al., 1993). However, a major reason why CNS neurons do not regenerate has its roots in the non-permissive neuritogenesis environment of the adult CNS, especially after episodes of tissue damage. This has been demonstrated by the early work of Aguayo and colleagues, who showed that injured rat CNS neurons could in fact grow into peripheral nerve grafts (David and Aguayo, 1981). Further experiments showed that a more permissive growth environment for injured CNS neurons could also be provided by PNS Schwann cell implants (Li and Raisman, 1994) and by myelin-free spinal cord (Savio and Schwab, 1990). In response to injury, the CNS undergoes an injury response known as reactive gliosis or glial scarring (Fawcett and Asher, 1999; Fournier and Strittmatter, 2001). The morphological and biochemical changes associated with glial scar formation pose a physical barrier and contribute to a non-permissive environment for neuronal re-growth. However, glial scars can take quite some time, up to weeks, to be fully formed. The delay in immediate re-growth of freshly damaged neurons may therefore have something to do with CNS myelin itself. It follows that the CNS myelin made by oligodendrocytes contains substances that inhibit neuronal regeneration. Elegant experiments from the laboratory of Martin Schwab provided evidence that dissociated optic nerves could extend neurites over astrocytes and immature oligodendrocytes but not over differentiated oligodendrocytes (Schwab and Caroni, 1988). This and subsequent work from Schwab and others clearly demonstrated the inhibitory property of CNS myelin. Furthermore, inhibitory substances in the CNS myelin seem to account for a major part of the CNS regeneration inhibitory activity. This is evidenced by the observations that animals immunized with myelin could extensively regenerate a large number of nerve fibers and demonstrate enhanced functional recovery (Huang et al., 1999), and treatment of CNS lesioned rats with the monoclonal antibody IN-1 (raised against a myelin inhibitor Nogo, as describe below) results in regenerative nerve fiber growth (Bregman et al., 1995; Karim et al., 2001). To date, three classes of CNS-myelin-associated inhibitors have been identified, namely, myelin-associated glycoprotein (MAG), Nogo and CSPGs. On the other hand, there is also substantial experimental evidence to suggest that myelin-containing adult CNS white matter can support regeneration of adult neurons, provided that neurons are experimental grafted in a way that does not induce glial scarring (Davies et al., 1997, 1999). In such experimental systems, growing axons still stopped or turned away from areas with high levels of CSPGs, as discussed below Myelin-associated glycoprotein MAG is a transmembrane protein belonging to a subgroup of sialic acid dependent adhesion molecule of the immunoglobulin superfamily. It can either promote or inhibit neuronal outgrowth depending on the age of the neuron tested, but is inhibitory for most mature neurons (McKerracher et al., 1994; Mukhopadhyay et al., 1994). Although MAG knockout mice display little difference in terms of the regenerative ability of damaged spinal axons compared to wild-type mice (Bartsch et al., 1995), there is another evidence that suggests a role for MAG in neuronal regeneration in vivo. It has been reported that a soluble proteolytic fragment corresponding to the extracellular domain of MAG exist in vivo (termed dmag). This proteolytic fragment is a potent growth inhibitor when released from purified myelin (Tang et al., 1997). In fact, it was very recently shown that dmag might account for the majority of axonal regeneration activity released from damaged white matter (Tang et al., 2001). There is very noteworthy recent progress in the understanding of signal transduction pathways responsible for the neuronal growth inhibitory effect of MAG. It was found early on that the function of MAG could be modulated by a mechanism that is camp-dependent. camp levels are dramatically higher in young neurons in which axonal growth is actually promoted by MAG, than in the same types of neurons that, when older, are inhibited by MAG. Inhibition of protein kinase A, a downstream effector of camp, prevents MAG promotion of neurite growth from young neurons, and elevating camp blocks MAG inhibition of neurite outgrowth in older neurons (Cai et al., 2001). Most recently, it was shown that the neurotrophin receptor p75 is a signal-transducing element for the MAG-mediated neuronal growth inhibition (Yamashita et al., 2002). Adult DRGs or postnatal cerebellar neurons from mice carrying a targeted disruption of the third exon of p75 are insensitive to MAG-induced neurite outgrowth inhibition. Mechanistically, MAG activates RhoA in the presence of p75 in wild-type cells, leading to retarded outgrowth. Although p75 appears not to be directly interacting with MAG, it does interact specifically with the ganglioside GT1b, which is one of the binding partners of MAG. The p75 and GT1b may thus form a receptor complex for MAG to transmit its growth inhibitory signals Nogo By performing size fractionation of myelin proteins and careful analysis of the inhibitory properties of each fraction, Schwab and colleagues first identified two inhibitory fractions with the molecular masses of 35 and 250 kda (which came to be known as the neurite growth inhibitors, or NI-35/250). A blocking monoclonal antibody, IN-1, raised against the 250 kda fraction was able to block the latter s inhibitory activity (Caroni and Schwab, 1988a,b). IN-1 turned out to be able to promote neuronal regeneration in various CNS injury models and in some cases lead to some degree of functional recovery (Karim et al., 2001). Importantly, Schwab s laboratory went on to purify the bovine equivalent of rat NI-250, and its microsequencing revealed six

4 192 B.L. Tang / Neurochemistry International 42 (2003) partial peptide sequences (Spillmann et al., 1998). These sequences led three groups to simultaneously report the molecular cloning of Nogo (Chen et al., 2000; GrandPre et al., 2000; Prinjha et al., 2000). Nogo has three isoforms: Nogo-A (1192 residues), Nogo-B (373 residues) and Nogo-C (199 residues). Nogo-A is the longest and corresponds to NI-250, and Nogo-B is a splice isoform that is likely to be NI-35. Nogo-C has the same C-terminus as Nogo-A and Nogo-B but has a different N-terminus generated by alternative promoter usage. All three isoforms of Nogo share the same C-terminus region that bears significant homology to the reticulon family proteins (Van de Velde et al., 1994; Roebroek et al., 1996), thus making Nogo the fourth member of the family. Analysis of the expression of the Nogo isoforms revealed that Nogo-A is CNS restricted, and is found mainly in oligodendrocytes and a subset of neurons (Huber et al., 2002). Nogo-B is not enriched in brain but can be found in the optic nerve, kidney and lung. Nogo-C is most abundant in skeletal muscle but can also be found in the CNS. Like other members of the reticulon family, Nogos have a dilysine ER retention motif at its C-terminus and is indeed found to be largely on intracellular membranes. However, there is also a significant amount of surface expression of Nogo in oligodendrocytes (Chen et al., 2000; GrandPre et al., 2000; Prinjha et al., 2000). The membrane topology of Nogo is still controversial, and the model best supported by available experimental evidence is that shown in Fig. 1. In this model, Nogo has two transmembrane domains and a major portion of the molecule is cytoplasmically oriented. A 66 amino acid loop (known as Nogo-66) between the two putative transmembrane domains is believed to be oriented extracellularly. Extracts from cells expressing Nogo-A inhibits neurite extension as well as spreading of fibroblast cells in vitro. Like MAG, its inhibitory effect is much more pronounced on mature neurons. More detailed analysis revealed that while the N-terminal portion of Nogo-A appears to have a general growth inhibitory activity, the Nogo-66 s growth inhibitory effect is confined to neuronal cells (GrandPre et al., 2000). Based on this, a brain enriched, high affinity receptor molecule for Nogo-66 was cloned (Fournier et al., 2001). The Nogo-66 receptor (NgR) is a 473 amino acid GPI-linked protein with eight leucine-rich repeats. Early embryonic chick retinal neurons that are normally unresponsive to Nogo, became responsive upon expression of exogenous NgR a clear functional evidence that NgR mediates the neuronal growth inhibitory activity of Nogo-66. The exact signaling mechanism by which the inhibitory activity is executed has not yet been elucidated, but is likely to involve immediate alterations of growth cone cytoskeleton via Rho-family GTPases and perhaps also delayed changes in the expression of growth cone proliferation enhancers such as GAP-43 (Ng and Tang, 2002). Since NgR does not span the plasma membrane, it is likely to engage some other molecules that could transmit the Nogo signal. It is therefore likely to be interacting, constitutively or upon Nogo binding, with a signal receptor complex that has a membrane-spanning component. The identity of this receptor is still unknown, but following from the recent lead with regards to the mechanism of MAG signaling (Yamashita et al., 2002), it may be worth to focus on known signal transducers of inhibitory cues. It is interesting to note that Nogo-B (which is not particularly enriched in the brain) could induce apoptosis in cancer cells (Li et al., 2001). The apoptosis inducing activity is associated with the C-terminus of the molecule (including the Nogo-66 portion), but it is not clear if it acts through the NgR. Earlier observations showed that treatment of rat dorsal root ganglion neurons with NI-35 (which is likely to be Nogo-B) resulted in a rapid and large increase in intracellular Ca 2+ before growth cone collapse takes place (Bandtlow et al., 1993; Bandtlow and Loschinger, 1997). The death inducing properties of Nogo-B in cancer cells is interesting and has potentially useful in therapeutic terms Chondroitin sulfate proteoglycans The glial reaction to injury results in the recruitment of a variety of glial cells, including microglia, oligodendrocyte precursors, meningeal cells, astrocytes and stem cells to the lesion site. Oligodendrocyte precursors and reactive astrocytes produce CSPGs, and there is considerable increase in CSPG expression around CNS lesions (Levine, 1994; Stichel et al., 1995; Fawcett and Asher, 1999; Asher et al., 2000). These CSPGs (phosphacan, neurocan, brevican, versican and NG2) have been shown to be potent inhibitors of axon growth in vitro. One of the major evidence for the involvement of CSPGs as neuronal regeneration inhibitors in vivo is the recent demonstration that treatment with chondroitinase ABC enhanced regenerative sprouting of nigrostriatal fibers of adult rat (Moon et al., 2001). While there is little doubt that the molecules MAG and Nogo discussed above are physiological inhibitors of CNS neuronal regeneration, elevated levels of CSPGs may prove to be the ultimate stumbling block for any neuronal growth. By using a microtransplantation technique that minimizes scarring, minute volumes of dissociated DRG neurons injected directly into adult rat CNS pathways can grow axons for long distances within the CNS white matter (Davies et al., 1997). Furthermore, chronically or acutely degenerating white matter could also support regeneration (Davies et al., 1999). However, abortive regeneration and axonal growth arrest with growth cone dystrophy occurs at proteoglycan-rich boundaries or at the molecular barrier of lesion sites. Neither the surface receptors that interacts with nor the signaling cascades induced by CSPGs leading to inhibition of neuronal regeneration are known in any detail. However, it was recently shown that the CSPG neurocan interacts with a cell surface glycosyltransferase (Li et al., 2000). This glycosyltransferase is a GPI-linked molecule that exerts an influence on two other important cell adhesion molecules that have critical roles in neurite outgrowth, namely 1-integrin

5 B.L. Tang / Neurochemistry International 42 (2003) Fig. 1. Known myelin and glial scar associated inhibitors. These include the two splice isoforms of Nogo, Nogo-A and Nogo-C that are brain enriched, the myelin-associated glycoprotein (MAG), the extracellular portion (dmag) generated by proteolytic cleavage and the myriad of chondroitin sulfate proteoglycans (CSPGs). The respective transmembrane mediators for the large class of CSPGs and their signaling intermediates are not known with certainty (represented by box with a? ). The NgR is a GPI-linked protein, and the actual signal transduction must involve a yet unidentified co-receptor or co-receptor complex (? box with dotted outline). For MAG, its signal transducer p75 appears to function as a receptor complex with the ganglioside GT1b. The signaling pathways from all these inhibitors may ultimately converge on the Rho GTPases, which could in turn regulate the reorganization of the actin cytoskeleton (green wavy lines) and microtubules (black lines). The plasma membrane indicated is that of a glial cell. and N-cadherin (Bixby et al., 1988). Neurocan binding coordinately inhibits both 1-integrin and N-cadherin-mediated cell adhesion and neurite outgrowth. CSPGs are generally known to interact with a myriad of molecules on the cell surface and the extracellular matrix, and are likely to elicit varied and complex effects, thus compounding the difficulty in dissecting their signaling pathways and mechanisms. 4. Guidance molecules and guidance receptors in the inhibition of neuronal regeneration The struggle by a severed CNS axon to regenerate from injury may appear to be quite different from the ordered and progressive axonal outgrowth that occurs during development. The latter is guided by spatially and temporarily changing patterns of attractant and repellent activities in the

6 194 B.L. Tang / Neurochemistry International 42 (2003) vicinity of the growth cone. The process usually involves massive morphological changes over a good length of time. However, there are now substantial evidence to implicate the influence of chemorepulsive molecules and their receptors in CNS regeneration. One of which is the sustained expression of some of these repellent proteins well into adulthood, as well as their injury-induced re-expression. While many neuronal guidance molecules exhibit bi-functionality, being either attractive or repulsive depending on the internal state of the growth cone in question, the semaphorins, ephrins, netrins and slits are largely known to be axonal repellents. The following discussions focus on the semaphorins, as convincing evidence exists for their role in neuronal regeneration, and include recent findings on their signaling mechanism Semaphorins and neuronal regeneration after injury Semaphorins are a family of soluble and membrane-bound axonal guidance factors (for extended reviews on semaphorins and axonal guidance, see recent excellent articles of Mueller (1999), Liu and Strittmatter (2001) and Song and Poo (2001)). This large group of molecules has at least 30 members and can be divided into eight classes based on their species of origin and structural homologies (Semaphorin nomenclature committee, 1999). The prototype of these, semaphorin 3A (Sema3A), or known previously as collapsin-1, is the first member of the better characterized class 3 semaphorins to be isolated (via biochemical fractionation as a potent inducer of growth cone collapse in vitro) (Luo et al., 1993). Evidence for a role for semaphorins in neuronal regeneration came largely from studies of their levels and expression in nerve injury models (Pasterkamp and Verhaagen, 2001). In adult rats subjected to a unilateral olfactory bulbectomy, high levels of Sema3A were detected in regenerating neurons a week post-operation (Williams-Hogarth et al., 2000). Another report showed that bulbectomy induced the proliferation and migration of non-neuronal cells expressing high levels of Sema3A mrna that occupy the core of the glial scar (Pasterkamp et al., 1998). These cells are likely to be fibroblasts of meningeal origin based on their morphology and immunohistochemical profiles (Pasterkamp et al., 1999). Axotomy of the olfactory nerve, on the other hand, induced a local and transient up-regulation of Sema3A expression at the lesion site. Olfactory axons can grow back to their original targets by circumventing Sema3A expressing areas in the lesion zone. Thus, it appears that Sema3A expression can block regeneration, and that regeneration to the extent of re-innervation is possible in a less severe injury paradigm where a regenerating axon can avoid areas with high Sema3A expression. Several lines of evidence support a role for semaphorins and their receptors in scar formation after CNS injury. NP-1, one of the semaphorin receptor component, is present in CNS scar tissue (Pasterkamp et al., 1999). Several semaphorin members, including Sema3A, Sema3B, Sema3C, Sema3E and Sema3F are present in lesionassociated fibroblasts. Based on their role in other tissues, these semaphorins may modulate-lesion-associated signaling events and immune modulation. Sema3A can in fact act as a neuronal cell death factor in vitro, inducing apoptosis (Gagliardini and Fankhauser, 1999; Shirvan et al., 1999). A particularly interesting piece of circumstantial evidence for an inhibitory role for semaphorins in CNS regeneration is that Sema3A is not present at the injury site in a peripheral nerve (which of course had no problem regenerating). Moreover, peripheral motor neurons appear to downregulate Sema3A mrna expression in response to injury (Pasterkamp et al., 1999) Semaphorin and semaphorin receptor signal transduction Semaphorins function through receptor complexes consisting of the neuropilin-plexin family proteins (Takahashi et al., 1999; Tamagnone and Comoglio, 2000). Most of the transmembrane semaphorins interact directly with plexins, which appear to be sufficient for their signaling (Tamagnone et al., 1999). The class 3 semaphorins, however, require the neuropilins as co-receptors, where neuropilin provides the membrane binding capacity while the plexin serve as the signal transducer. Interestingly, plexins share a homologous N-terminal domain (known as the Sema domain) with its ligands. Sema3A signaling requires the clustering of plexin-1a with the Sema3A/neuropilin-1 complex. Signaling from plexin-1 is auto-inhibited by its Sema domain, and the auto-inhibition is released by Sema3A/neuropilin-1 binding (Takahashi and Strittmatter, 2001). The cytoplasmic domain of plexin can be tyrosyl-phosphorylated, suggesting that plexins can signal via association with a tyrosine kinase (Tamagnone et al., 1999). The cytoplasmic domain of neuropilin-1 has been shown to interact with a PSD-95/Dlg/ZO-1 (PDZ) domain-containing protein that may also be a signal transducer (Cai and Reed, 1999). In terms of collapsin-1/sema3a s ability to induce growth cone collapse, several cytosolic components appear to mediate semaphorin signaling downstream of plexins. The most prominent of these are the collapsin response mediator protein (CRMP) and the Rho-family GTPases. The C. elegans unc33 homologue CRMP was first identified in Strittmatter s laboratory in an expression screen for mediators of Sema3A signaling (Goshima et al., 1995). Five members of the CRMP family are now known. These are differentially expressed in the developing and adult brain. The unc33 mutation in C. elegans results in abnormal axonal outgrowth, and the signaling pathways of different guidance cues may converge on the CRMP family. A direct involvement of CRMP in Sema3A-induced growth cone collapse is demonstrated by the inhibition of the process by an anti-crmp antibody (Wong and Strittmatter, 1997). Although it is clear that CRMP is a major functional link

7 B.L. Tang / Neurochemistry International 42 (2003) between plexin activation and growth cone collapse, the intermediate steps upstream and downstream of CRMP are not known with certainty (Liu and Strittmatter, 2001). CRMP2 can be phosphorylated by a serine-threonine kinase of unknown identity (Gu et al., 2000) while CRMP3 associates with a brain specific unc-33 family protein known as the CRMP-associated molecule (CRAM) as well as a tyrosine kinase (Inatome et al., 2000). These associations have yet to be linked to plexin or the function of CRMP in growth cone collapse. There is strong evidence suggesting that semaphorininduced growth cone collapse involves the direct activation of Rho-family GTPases. Introduction of a dominant-negative form of Rac1 into DRG neurons attenuates their response to Sema3A (Jin and Strittmatter, 1997). Likewise, adenoviral-mediated expression of constitutively activated Rac1 (and interestingly, constitutively activated Cdc42 as well) negated Sema3A-induced growth cone collapse and promoted neurite outgrowth on a Sema3A substrate (Kuhn et al., 1999). A peptide corresponding to amino acids of Rac1 that binds directly to the established Rac1-interacting molecules competes with activated Rac1 for target binding, and inhibits Sema3A-induced growth-cone collapse (Vastrik et al., 1999). In the case of Sema3A signaling, there is no absolute evidence indicating that plexin activation directly activates Rac1 (although plexin-a1 does interact with another Rho-family GTPase, Rnd1 (see below)). However, there is a particularly interesting piece of evidence from Strittmatter s laboratory implicating the direct regulation of Rac1 by semaphorin. Sema3A Fig. 2. The signaling pathway of Sema3A in inducing growth cone collapse. Sema3A (or collapsin-1) is a secreted soluble factor that binds to its receptor complex consisting of neuropilin-1 (NP-1) and plexin-1a, probably in the form of a dimer. The binding of Sema3A-NP-1 to plexin-1a activates it, and activated plexin-1a in turn engages Rac and consequently its downstream effectors, notably LIM-kinase and cofilin. This process appears to be dependent on the collapsin response mediator protein (CRMP). Direct engagement of a Rho-family GTPase by a surface receptor is not the norm. For another guidance molecule ephrin-a (EphA), stimulated EphA receptors interact with ephexin, which is a GEF for RhoA. Activation of RhoA by ephexin can potentially engage its downstream effector ROCK, which could in turn activate LIM-kinase or further engage CRMP.

8 196 B.L. Tang / Neurochemistry International 42 (2003) treatment results in ligand dependent F-actin distribution and re-localization of Rac1 with the ligand-aggregated semaphorin receptors (Fournier et al., 2000). Evidence for a direct physical association between plexin and Rac1 comes from studies of plexin-b1 in non-neuronal cells. Data from three different laboratories indicate that plexin-b1 directly interacts with activated Rac1 and Rnd1 (Vikis et al., 2000; Rohm et al., 2000; Driessens et al., 2001). The interaction appears to be rather specific, as Rac1 does not interact with plexin-a3 or plexin-c1, neither does RhoA or Cdc42 interact with plexin-b1 (Vikis et al., 2000). Curiously, the binding of GTP and the integrity of the Rac effector domain are required for the interaction with plexin-b1, which effectively places Rac1 upstream of plexin-b signaling. It is therefore not yet clear how activation of plexin-b1 causes the activation of Rac. Indeed, plexin-b1 clustering in fibroblasts does not cause the formation of lamellipodia, but instead results in the assembly of actin-myosin filaments and cell contraction, which indicates RhoA activation and not Rac1 (Driessens et al., 2001). Although the role of RhoA in semaphorin signaling is even less clear than that of Rac1, the above results would suggest that RhoA is downstream of Rac1 in response to plexin-b1 activation. What acts downstream of the Rho-family GTPases in the case of Sema3A? The actin binding protein cofilin, or actin depolymerizing factor (ADF) promotes F-actin turnover by causing actin depolymerization. Over-expression of cofilin in primary neurons enhances neurite outgrowth (Meberg and Bamburg, 2000), whereas antisense oligonucleotides against cofilin inhibit outgrowth (Kuhn et al., 2000). The activity of cofilin is mediated by the serine-threonine kinase, LIM-kinase, which inactivates cofilin by phosphorylating it (Sumi et al., 1999). LIM-kinase is, in turn regulated by the effectors of the Rho-family GTPases such as ROCK (Maekawa et al., 2000) and PAK (Edwards et al., 1999). A recent report provided a molecular link between Sema3A-induced actin rearrangements and growth cone collapse. Sema3A treatment of growth cones resulted in sequential phosphorylation and dephosphorylation of cofilin. The collapsing activity of Sema3A is suppressed by a peptide containing a cofilin phosphorylation site, which inhibited LIM-kinase activity in vitro and in vivo. A dominant-negative form of LIM-kinase, which could not be activated by ROCK or PAK, suppressed the collapsing activity of Sema3A (Aizawa et al., 2001). The signaling cascade of Sema3A is represented diagrammatically in Fig. 2, together with that of ephrin-a (which appears to utilize a more conventional route of signaling that involves the Rho GTPases, via a GEF (see below)). 5. The Rho-family GTPases and their function in neurite outgrowth and retardation Small GTPases of the Rho-family are important regulators of actin cytoskeleton in all eukaryotic cells (for recent extensive reviews, see Hall (1998), Kaibuchi et al. (1999), Bishop and Hall (2000), Bar-Sagi and Hall (2000)). Most findings are based on RhoA, Rac1 and Cdc42, which are the most widely expressed members. In fibroblasts, RhoA regulates cell contractility and the assembly of actin stress fibers, Rac1 stimulates the formation of lamellipodia while Cdc42 regulates filopodia formation (Nobes and Hall, 1995). All three GTPases and their modulators and effectors have been implicated in neurite outgrowth and/or axonal guidance (Sarner et al., 2000; for a recent extended review, see Luo (2001)). As discussed above, Sema3A-induced growth cone collapse can be inhibited by dominant-negative forms of Rac1 or Cdc42. On the contrary, the growth inhibitory effect of myelin extract is inhibited by constitutively activated Rac1 or RhoA (Kuhn et al., 1999). That opposite mutants of Rac1 are effective against different inhibitory signals argues against a common signaling pathway underlying the action of neurite growth inhibition and growth cone collapse. However, all agents that either promote neurite growth or retard neurite growth would appear to function through the Rho family of GTPases, with Rac1 promoting neurite extension in general while RhoA induces retraction. The relationship and sequence of action between the different members of Rho GTPases in neurons are complicated and are distinctly different from that in fibroblasts Novel Rho family members implicated in neurite outgrowth Several recent reports have implicated some of the less known members of the Rho-family GTPases in regulating neurite outgrowth. These include RhoG, Rnd1 and TC10. Expression of wild-type RhoG in PC12 cells induced neurite outgrowth in the absence of NGF (Katoh et al., 2000). Neurite outgrowth induced by NGF was enhanced by expression of wild-type RhoG and suppressed by expression of dominant-negative RhoG. Likewise, constitutively active Ras-induced neurite outgrowth was also suppressed by dominant-negative RhoG. Cells transfected with constitutively active RhoG have elevated Rac1 and Cdc42 activities and extended short neurites, but they developed large lamellipodial or filopodial structures at the tips of neurites. Co-expression with dominant-negative Rac1 or Cdc42 inhibited RhoG-induced neurite outgrowth. The above results suggest that RhoG acts upstream of Rac and Cdc42, but there is no evidence yet to suggest that RhoG acts downstream of guidance receptors or inhibitory signals. TC10 mrna is most abundant in heart and skeletal muscle, not brain (Neudauer et al., 1998). In fact, the expression of TC10 is normally lower in developing and mature motor neurons compared with other Rho family members such as RhoA, Rac1, and Cdc42. However, Tanabe et al. (2000) demonstrated that TC10 mrna expression in motor neurons was dramatically induced by axotomy. Cultured dorsal root ganglia exhibited dramatic neurite extension upon adenovirus-mediated expression of TC10. The authors

9 B.L. Tang / Neurochemistry International 42 (2003) concluded that TC10 expression is induced by nerve injury to play a crucial role in nerve regeneration, particularly neurite elongation, perhaps via cooperation with other Rho family members. Rnd1 was the first Rho-family GTPase shown to interact with plexin-a1 (Rohm et al., 2000). Expression of Rnd1 in PC12 cells caused the formation of neuritic processes with disruption of the cortical actin filaments (Aoki et al., 2000). Neurite formation induced by Rnd1 was inhibited by dominant-negative Rac1, which suggests that Rac1 acts downstream of Rnd1. On the other hand, a very recent report showed that recruitment of active Rnd1 is sufficient to trigger plexin-a1-mediated growth cone collapse even in the absence of Sema3A. Interestingly, the effect of Rnd1 is block by RhoD (Zanata et al., 2002). Further studies are needed to fully define the physiological relevance of the apparent signaling pathway from Sema3A-plexin-1A through Rnd Modulators of Rho-family GTPases activity in neurite outgrowth Modulation of Rho-family GTPases directly by receptors, as in the case of plexins above, is probably the exception rather than the rule. Most guidance cues act through one or more of the three classes of proteins that regulate the guanine nucleotide binding status of Rho-family GTPases. These are the GEFs, the guanine nucleotide dissociation inhibitors (GDIs) and the GAPs. The GEFs activate Rho-family GTPases by promoting the exchange of GDP for GTP, whereas the GAPs inactivate them by enhancing their intrinsic GTPase activity. The GDIs participate in the regulation of both the GDP/GTP cycle and the membrane association/dissociation cycle. GEFs for Rho-family GTPases are critical modulators of neuritogenesis and growth cone guidance. Many of these are enriched in the brain, and have been localized to growth cones and dendrites (Togashi et al., 2000, Kunda et al., 2001). The invasion-inducing T-lymphoma invasion and metastasis 1 (TIAM1) protein functions as a GEF for Rac1. TIAM1 expression in the developing brain is differentiation dependent. Early studies showed that over-expression of TIAM1 induces cell spreading and affects neurite outgrowth in N1E-115 neuroblastoma cells (Van Leeuwen et al., 1997). These effects are Rac-dependent. Cells over-expressing TIAM1 no longer respond to the Rho-mediated lysophosphatidic acid-induced neurite retraction and cell rounding, suggesting that TIAM1-induced activation of Rac antagonizes Rho signaling. Neurite formation induced by TIAM1 or Rac1 is further promoted by inactivating Rho. These results provide the first demonstration that Racand Rho-mediated pathways oppose each other, and that a balance between these pathways determines neuronal morphology. Recent studies further implicate TIAM1 in the establishment of neuronal polarity (Kunda et al., 2001). In cultured primary neurons, TIAM1 localizes to the neurite and its growth cone, where it associates with microtubules. Neurons over-expressing TIAM1 extend several axon-like neurites, whereas suppression of TIAM1 prevents axon formation. A splice isoform of kalirin, kalirin-7, is a Rac specific GEF, the first GEF localized to the post-synaptic density (Penzes et al., 2000), and it may function to regulate the morphology of the dendritic spine. Kalirin-7 is a large multi-domain protein (containing nine spectrin-like repeats, a Dbl homology domain, and a pleckstrin homology domain) and could be linked to many signaling and adaptor molecules. It shares significant homology to Trio, another GEF whose function in photoreceptor axon pathfinding in Drosophila has recently been investigated in detail (Newsome et al., 2000; Lin and Greenberg, 2000). Perhaps not surprisingly, another GEF of Rac1, the still life and TIAM1-like exchange factor (STEF), which is predominantly expressed in the brain during development, has very recently been included in the above list. Ectopic expression of STEF in N1E-115 neuroblastoma cells induced neurite-like processes, while the expression of a fragment corresponding to its membrane association domain, PHnTSS, resulted in inhibition of neurite outgrowth (Matsuo et al., 2002). The participation of GAPs in neuritogenesis and axonal guidance has also come to light recently. The p190 RhoGAP is shown to be the principal substrate for the neuronally enriched Src kinase, which has been implicated in several aspects of neural development and nervous system function. Mice lacking functional p190 RhoGAP exhibit defects in axon guidance and fasciculation. The p190 RhoGAP is co-enriched with F-actin in the distal tips of axons, and over-expression of p190 RhoGAP in neuroblastoma cells promotes extensive neurite outgrowth (Brouns et al., 2001). The p190 RhoGAP, like TIAM1 (Van Leeuwen et al., 1997), transduces signals downstream of cell-surface adhesion molecules, as p190-rhogap-mediated neurite outgrowth is promoted by the extracellular matrix protein laminin. p190 RhoGAP might therefore mediate a Src-dependent adhesion signal for neuritogenesis to the actin cytoskeleton through the Rho GTPase. As mentioned above, as far as growth cone collapsing activity of Sema3A is concerned, there is no clear evidence for the involvement of GEFs or GAPs. An interesting point to note is that plexin family members have a weak sequence homology to Ras GAPs (not Rho GAPs) (Takahashi et al., 1999; Rohm et al., 2000). This, together with the in vitro interaction demonstrated between plexins and members of the Rho-family, led to the speculation that semaphorins directly engage the Rho GTPases and by pass the need for GEFs or GAPs. This is clearly not the case for other inhibitory cues such as the ephrins and the Eph receptors, which is the largest family of receptor tyrosine kinases (for a recent extended review on ephrins and Eph receptors, see Mueller (1999), Schmucker and Zipursky (2001) and Wilkinson (2001)). The cloning and characterization of ephexin, a novel Eph receptor-interacting protein that is a member of the Dbl

10 198 B.L. Tang / Neurochemistry International 42 (2003) family of Rho GEFs, has recently been reported (Shamah et al., 2001). The ephrin-a stimulation of EphA receptors modulates the activity of ephexin leading to RhoA activation, Cdc42 and Rac1 inhibition, and cell morphology changes. A mutant form of ephexin interferes with ephrin-a-induced growth cone collapse in primary neurons. The association of ephexin with Eph receptors constitutes a molecular link between Eph receptors and the actin cytoskeleton. This is shown in Figs. 2 and 3 alongside the signaling pathway of Sema3A. Fig. 3. The simplified general scheme of signaling cascades that regulate neurite outgrowth and retraction. The various ligands that would either promote neurite outgrowth or retraction interact with their respective plasma membrane receptor or receptor complexes. In the case of the CNS regeneration inhibitors MAG and CSPG, the respective membrane proteins that mediate their effects have not been identified with certainty (???). Neurotrophins receptor engages adaptor molecules (Grb2 and Nck for example) for their downstream signaling, whereas semaphorin functions through cytosolic components such as the collapsin response mediator protein (CRMP). The signaling pathways converge on the Rho-family GTPases with or without going through a modulator (GEFs, GDIs or GAPs). The activated Rho-family GTPases interact with their effectors, which regulate the functions of the actin-modifying proteins via other intermediates such as LIM-kinase.

11 B.L. Tang / Neurochemistry International 42 (2003) Effectors of Rho-family GTPases in neurite outgrowth Many downstream targets of Rho GTPases are known and some of these have clear roles in neurite extension or axonal guidance. The neural Wiskott Aldrich syndrome protein (N-WASP) is an actin-regulating protein that induces filopodia formation downstream of Cdc42. It has a verprolin, cofilin homology and acidic (VCA) domain, known to be required for the activation of the Arp2/3 complex that induces actin polymerization. Expression of a VCA domain mutant form of N-WASP with diminished Arp2/3 activating ability, or a mutant that is unable to bind Cdc42, suppressed neurite extension of PC12 cells and severely inhibited the neurite extension of primary hippocampal neurons (Banzai et al., 2000). The p21-activated kinase 1 (PAK1) is a serine-threonine kinase acting downstream of Rac and Cdc42. It was shown that targeting of PAK1 to the plasma membrane via a C-terminal isoprenylation sequence induced PC12 cells to extend neurites similar to those stimulated by NGF (Daniels et al., 1998). This effect was independent of PAK1 kinase activity but was dependent on structural domains within both the N- and C-terminal portions of the molecule. Neurite outgrowth in PC12 cells stimulated by NGF were effectively inhibited by these regions. Curiously, another report suggests that PAK promotes morphological changes in PC12 by acting upstream of Rac (Obermeier et al., 1998), which is in better agreement with the fact that the kinase activity of PAK is not involved in the changes observed. The p160rock, a downstream effector of RhoA, has also been implicated in inhibition of neuritogenesis. Microinjection of the catalytic domain of p160rock into NGF-differentiated PC12 cells induced neurite retraction, in a similar manner to that induced by the microinjection of a constitutively active Rho (Katoh et al., 1997). In cultured cerebellar granule neurons. The p160rock inhibition triggered immediate outgrowth of membrane ruffles and filopodia, followed by the generation of initial growth cone-like membrane domains from which axonal processes arose (Bito et al., 2000). Neurite outgrowth appears to involve two components, the formation of actin-rich filopodia and lamellipodia at the growth cone in the direction of growth as well as traction force generated by the growth cone that pulls on the neurite. The latter is thought to involve actomyosin-generated force acting in combination with substratum adhesion. Two isoforms of non-muscle myosin II heavy chain (A and B) are present in neuronal tissues. Studies using drugs and antisense knockdown approaches in neurons or neuroblastoma cells have suggested a role for myosin IIB in neurite outgrowth or growth cone motility (Ruchhoeft and Harris, 1997; Wylie et al., 1998). On the other hand, incubation of neuroblastoma cells with isoform specific antisense oligonucleotides for myosin IIA resulted in a loss of cell adhesion. The activity of myosin in neurite extension or collapse is also regulated by the Rho GTPases. Rac has been shown to antagonize Rho by regulating the phosphorylation of the myosin II heavy chain (Van Leeuwen et al., 1999). Furthermore, the myosin light chain kinase (MLCK) is subjected to regulation by PAK1 phosphorylation, which decreases its activity (Sanders et al., 1999). 6. Future perspectives Great strides have been made in the past 2 3 years in our understanding of the molecules that are involved in regulating neurite outgrowth and retraction as well as axonal guidance during development. These appear to be fundamentally related processes that must ultimately involve modulation of cellular adhesion and cytoskeletal components. Much is already known about the modulation of actin by the Rho GTPases. However, the process of neuritogenesis and axonal outgrowth involves integrated changes not just in actin microfilaments alone, but also microtubules and myosin. Furthermore, all these changes must be operatively coupled to membrane transport processes (Tang, 2001). Much more remains to be learnt about the latter processes. Based on current knowledge about the signaling pathways of guidance receptors, one can be optimistic that the specific signal transducers and signaling pathways of the CNS inhibitors of regeneration MAG, Nogo and the CSPGs will be revealed and properly delineated in good time by contemporary approaches. The undertaking to unravel the signaling intermediates would undoubtedly be aided by the human genome data that is now fully in place, as well as the myriad of genomics and proteomics approaches that are now available. It would be surprising if these pathways turn out to be completely alien and unrelated to what is already known. Comparatively, identifying the missing links in the signaling cascades and further distinguishing the true physiological intermediates of a pathway from a plethora of related or permissive factors is a task of a different magnitude. Expression of dominant-negative mutants by DNA transfection, viral transduction or protein microinjection can provide the first clues, but the perturbations generated are all grossly non-physiological in nature. As a result, the effects elicited can also be exaggerated and non-specific. In order to do the appropriate loss and gain of function experiments in more physiological settings, one would have to create conditional knockout and knock in animals that would in turn provide cells and tissues samples where the expression of the genes in question can be modulated at ease. Furthermore, it would seem impossible to perform adequate quantitative or even semi-quantitative analysis of the input output relationship of the various components of the interwoven signaling network without the aid of computational simulations (Song and Poo, 2001; Jordan et al., 2000). A multi-disciplinary approach that constantly taps on technological advances would be the way to go.