1998), but because these process are not within the realm of the growth cone, they are not part of this review.

Transcription

1 The Multiple Decisions Made by Growth Cones of RGCs as They Navigate from the Retina to the Tectum in Xenopus Embryos Kevin S. Dingwell, Christine E. Holt, William A. Harris Department of Anatomy, University of Cambridge, Downing Street, Cambridge, United Kingdom, CB2 3DY Received 20 April 2000; accepted 25 April 2000 ABSTRACT: Retinal ganglion cells (RGCs) of Xenopus laevis send axons along a stereospecific pathway from the retina to their target the optic tectum. Viewed from the point of the growth cone, this journey is reflected by discrete processes of axon initiation, axon outgrowth, navigation, target recognition, and innervation. These processes are characterised by distinct signalling mechanisms that trigger dynamic changes in growth cone morphology and behavior. Here we review work primarily from our laboratory, examining these events from a cellular and molecular perspective, focusing on the roles of FGFs, netrins, receptors, and intracellular effectors John Wiley & Sons, Inc. J Neurobiol 44: , 2000 Keywords: Xenopus; RGC; CAM; netrin; FGF; ephrin The growth of retinal ganglion cell (RGC) axons to their tectal targets is perhaps one of the best understood cases of axon navigation over an entire pathway. Decades of work from many laboratories have yielded insights into both the cellular and molecular mechanisms that regulate growth cone guidance in the retinotectal system, from the initiation of axonogenesis, to axon outgrowth and pathfinding, to target recognition, and finally to topographical map formation (Hynes and Lander, 1992; Holt and Harris, 1993; Chien and Harris, 1994; Tessier-Lavigne and Goodman, 1996). We discuss this journey from the viewpoint of the RGC growth cone: what it senses and how it responds. In the retinotectal system we have additional information about synapse formation and activity-dependent plasticity of connections in the tectum (see review by Holt and Harris, 1993; Zhang et al., Corresponding author: W. A. Harris (harris@mole.cam.ac.uk). Contract grant sponsor: Medical Research Council (CEH, WAH). Contract grant sponsor: Wellcome Trust (CEH, WAH). KSD is a Hitchings-Elion fellow of the Burroughs Wellcome Fund John Wiley & Sons, Inc. 1998), but because these process are not within the realm of the growth cone, they are not part of this review. AXONOGENESIS AND GROWTH CONE EXTENSION RGCs in the dorsocentral Xenopus retina are the first to extend axons beginning around stage 28, and axonogenesis continues in a dorsal ventral gradient (Holt, 1984, 1989). Axonogenesis is first evident as a polarized thickening of the plasma membrane close to or at the vitreal surface of RGCs [Holt, 1989; Fig. 1(A)]. In some cases axons appear from other parts of the cell or from a primary dendrite shaft. Once axons reach the vitreal surface, they immediately orient and extend along the ganglion cell fiber layer toward the optic nerve head [Fig. 1(B)], where they leave the eye and join the optic nerve [Fig. 1(C)]. In vitro, components of the extracellular matrix (e.g., laminin, fibronectin) and a variety of cell adhesion molecules (CAMs: IgG superfamily, cadherin families) and substrate adhesion molecules (SAMs; 246

2 RGC Growth Cone Dynamics 247 Figure 1 Development of the Xenopus visual system. (A) Transverse sections show the position and growth cone dynamics of pioneering axons from the right eye. RGCs from the dorsocentral retina are the first to extend axons beginning at stage 28. (B) These pioneering axons grow ventrally close to the vitreal surface of the eye toward the ONH. (C) At the ONH, growth cones become highly complex as they dive down into the ON. (D) Growth cones appear torpedo shaped as they grow along the ON. (E) Growth cones then adopt a complex morphology when they reach the ventral diencephalon, at the brain entry point. At stage 32, growth cones cross over to the contralateral side of the brain at the optic chiasm (F) and then navigate in the neuroepithelium close to the pial surface. Axons first reach the mid optic tract by stage 35/36, where they make a caudal turn that reorients them with the tectum (G). The first growth cones reach the anterior tectum by stage 39 and undergo dramatic morphological changes (H). Within the tectum, terminal arbors develop that form a topographical map (I). ONH, optic nerve head; ON, optic nerve; vs, vitreal surface; ps, pial surface.

3 248 Dingwell et al. integrins) have been shown to promote neurite initiation [Hynes and Lander, 1992; Fig. 2(A,B)]. To address whether CAMs and SAMs play significant roles in the initiation of axons in vivo, various dominant negative forms of CAMs and SAMs have been expressed in RGCs using an in vivo transfection technique (Holt et al., 1990). Expression constructs are targeted to the eye primordia of late neural plate stage embryos. The expression of mutant proteins in some retinal precursors establishes a mosaic in which some RGCs express the transgenes in an otherwise genetically wild-type background. This technique serves as a powerful approach to specifically target expression of foreign genes to RGCs in order to determine their roles during visual system development. Integrin Signaling The integrin complex, a heterodimer of a and b subunits, is a structural component of focal adhesions and acts as a link between the plasma membrane and the extracellular matrix (ECM; Hynes, 1992; Aota and Yamada, 1997). Expression of a variety of chimeric integrins, which are either impaired in b 1 subunit binding to the RGD site in fibronectin or fail to localize to focal adhesions, severely compromise the initiation of RGC axons and their extension within the retina (Lilienbaum et al., 1995). These data suggest that cell adhesion to the ECM via b 1 integrins is required for normal axon outgrowth. However, although substrate-adhesion is essential for generating the forces required for axon outgrowth, simple cell cell or cell ECM adhesion may not be sufficient for promoting axonogenesis. Adhesion is poorly correlated with stimulating neurite growth in culture (Gunderson, 1987), and numerous CAMs can promote adhesion but do not stimulate growth (Gunderson, 1987; Hall et al., 1987; Lemmon et al., 1992). Together, these results suggest that there are additional signaling events involved in CAM-mediated axonogenesis. Focal adhesions are not only sites that couple the plasma membrane with the cytoskeleton, but are also macromolecular signaling complexes (Giancotti and Ruoslahti, 1999; Fig. 3). Composed of both structural (e.g., cytoskeleton-binding) and signaling proteins, focal adhesions play important roles in a variety of cellular processes (e.g., cell cycle, cell migration, and anoikis; Aota and Yamada, 1997; Giancotti and Ruoslahti, 1999). Key components of the focal adhesion signaling complex are the cytoplasmic tyrosine kinases (TKs), pp60 c-src and pp125fak (focal adhesion kinase). Both TKs play important roles in regulating the array of extracellular signals elicited by integrins, neuropeptides, and growth factor receptors. They are therefore prime candidates for mediating either integrin- or CAM-dependent axonogenesis. FAK is strongly expressed in RGC neurites, growth cones, and filopodia in a punctate pattern that colocalizes with phosphotyrosine immunolabelling (Worley and Holt, 1996). Herbimycin A (HA), a TK inhibitor, reversibly inhibits axon extension from RGC in culture and in vivo (Worley and Holt, 1996a,b). In addition, HA treatment results in the redistribution of FAK from filopodia into a more diffuse pattern within the center of the growth cone (Worley and Holt, 1996b). Likewise, src is highly expressed in RGCs and within the optic nerve, and its function is correlated with neurite outgrowth (Maness et al., 1988, 1990; Cox and Maness, 1991; Bixby and Jhabvala, 1993). A constitutively active form of src induces neurite outgrowth when expressed in PC12 cells (Cox and Maness, 1991). In cerebellar neurons derived from either src 2/2 or fyn 2/2 mice, neurite outgrowth is severely impaired when plated on certain CAMs, suggesting that src acts downstream of CAM binding (Beggs et al., 1994; Ignelzi et al., 1994). In the Xenopus visual system, expression of constitutively active (ca) forms of either c-src or its neuronal splice variant n-src severely impairs axonogenesis and subsequent axon outgrowth (Worley et al., 1997). For ca c-src, up to 85% of the RGC axons terminate at the optic nerve head (ONH). A direct link between src and FAK function in mediating CAMdependent axon outgrowth has yet to be shown, although recent evidence suggests that src-dependent phosphorylation of cellular substrates depends on FAK activity (Schaller et al., 1999). Therefore, FAK activation following integrin binding in RGCs may recruit src TKs to focal adhesions, and thereby regulate phosphorylation of cytoskeletal proteins required for initiating and maintaining axon outgrowth (Fig. 3). CAM Signaling Cadherin-dependent axon outgrowth appears to involve a signaling mechanism in addition to cell adhesion. Neural or N-cadherin is a 130-kD, type 1 membrane glycoprotein that mediates Ca 12 -dependent cell cell adhesion (reviewed in Shapiro and Colman, 1998) and promotes neurite outgrowth from chick ciliary ganglion neurons and Xenopus RGCs (Tomaselli et al., 1988; Riehl et al., 1996). The cytoplasmic domain of N-cadherin is coupled to the actin cytoskeleton via interactions with catenins. However, the outgrowth-promoting property of N-cadherin in vivo does not require the catenin binding domain, suggesting that a separate region of the c-terminus regulates axon outgrowth (Riehl et al., 1996). Expression of the juxtamembrane region of N-cadherin, which lacks the

4 RGC Growth Cone Dynamics 249 Figure 2 Some of the molecules that regulate RGC growth cones dynamics during the development of the Xenopus visual system. Axonogenesis occurs at the vitreal surface of RGCs. Outgrowth of axons is then stimulated by a variety of molecules within the retina and optic tract. At key points along the visual pathway, growth cones make specific navigational choices leading to a change in their direction of growth. Once growth cones have reached their target, they undergo a series of dynamic morphological changes including the loss of the primary growth cone, and the development of backbranches along the axon shaft. Within the target, terminal arbor formation begins in which appropriate topographical choices are made.

5 250 Dingwell et al. catenin binding domain, is a potent inhibitor of axon outgrowth, yet has little or no effect on epithelial cell cell adhesion (Riehl et al., 1996). This indicates that there may be protein protein interactions with the juxtamembrane region of N-cadherin and other cytoplasmic proteins. The juxtamembrane region of N- cadherin binds to the cytosolic protein p120 ctn, a newly discovered member of the catenin family. The function of p120 ctn remains to be elucidated although in epithelial cells it may play a role in regulating the strength of cadherin-based adhesion (reviewed in Provost and Rimm, 1999). Overexpression of p120 ctn in fibroblasts results in a remarkable phenotype characterized by the induction of long filopodial-like processes (Reynolds et al., 1996). Therefore, it is tempting to speculate that p120 ctn (originally identified as a substrate for c-src) may be an important component of the N-cadherin mediated axon-outgrowth signal (Fig. 3). Figure 3 Model for signal transduction cascades in Xenopus growth cones during the development of the visual system. Axonogenesis is regulated by at least three signals propagated by N-cadherin, b1-integrins, and the FGFR. N-cadherin dependent axonogenesis requires the juxtamembrane region of the cadherin molecule and may involve the catenin-like protein p120. pp60-src and focal adhesion kinase (FAK), likely act downstream of integrin binding, whereas FGFR signaling is triggered by non-fgf ligand(s). All signals likely converge on the activation of the small GTPase Rac. Axon outgrowth is stimulated in the retina and optic tract through the PLCg pathway via FGF and N-cadherin dependent activation of the FGFR. b1- integrin signaling, possibly through src/fak, is necessary for promoting outgrowth only in the retina. Target recognition and innervation requires an FGFR signal, possibly through SH2/SH3 domain containing adapter/effector proteins that activate the GTPases Cdc42 and RhoA. FGFR, fibroblast growth factor receptor; JM, juxtamembrane; FAK, focal adhesion kinase; SH2/SH3, src homology region 2/3. Growth Factors In addition to CAMs and ECM, axon initiation and outgrowth in the visual system depends on growth factor receptors, such as the fibroblast growth factor receptor [FGFR; Fig. 2(A,B)]. RGC growth cones express FGFRs, and axon outgrowth is strongly stimulated by FGF2 (McFarlane et al., 1995). FGFs are also abundantly expressed along the length of the optic pathway (McFarlane et al., 1995), suggesting that FGF and the FGFR are involved in stimulating axon outgrowth. When a dominant negative form of the FGFR, XFD, is expressed in RGCs, axon outgrowth is significantly impaired, such that axons advance at only 60% of the normal speed (McFarlane et al., 1996). XFD, however, has no effect on axonogenesis. In contrast, expression of the FGFR mutant HAVf, which is unable to bind FGF and therefore cannot act as a dominant negative (Amaya et al., 1993), inhibits axonogenesis in vivo, suggesting that a non-fgf ligand is involved. The identity of the non- FGF ligand(s) remains to be identified. CAMs may represent a family of non-fgf FGFR ligands involved in promoting axon outgrowth [Fig. 2(B)]. In culture, the CAMs L1, N-CAM, and N- cadherin, exert potent axon outgrowth-promoting activity in an FGFR-dependent manner (Doherty and Walsh, 1996; Lom et al., 1998; Fig. 3). Although a physical association of the FGFR and CAMs has yet to be shown, the CAM homology domains (CHD) may be involved in heterophilic protein protein interactions between the FGFR and CAMs. Peptides corresponding the CHD domains inhibit FGFR-dependent CAM-stimulated axon outgrowth (Williams et al., 1994). Thus, at least three FGFR-dependent signals are elicited in the RGC growth cone. First, non-fgf ligand activation is required for axonogenesis, which appears to be distinct from the second CAM-mediated, FGFR-dependent stimulation of axon outgrowth. Finally, there is third, FGF-dependent FGFR activation of axon extension. How can three FGFR-dependent signals elicit several different growth cone responses? First, cellular responses to FGF signaling vary, depending on the spatial and temporal expression of both ligand and receptor (Klint and Claesson- Welsh, 1999). Three different FGFRs are expressed in the developing retina at a time when axonogenesis is occurring (McFarlane et al., 1998) and therefore could

6 RGC Growth Cone Dynamics 251 account for some of the differences in the FGFRdependent signaling events occurring during axonogenesis versus axon outgrowth. Second, FGFR ligands exhibit different binding affinities for their receptors, resulting in differences in the strength, nature, and duration of the signaling cascade (Marshall, 1995). Finally, the cellular response following activation of different FGFRs can differ even when the receptors are activated by a common ligand (Raffioni et al., 1999). The identification of the both the FGF and non-fgfr ligands expressed in the retina await identification. Activation of the FGFR leads to the phosphorylation of specific tyrosine residues with the FGFR s cytoplasmic domain (reviewed in Klint and Claesson- Welsh, 1999). These tyrosine residues act as binding sites for a large number of adapter and effector molecules that are responsible for transducing the array of FGF signals. Thus, our understanding of the varied roles of FGFR-dependent signaling during axonogenesis and axon outgrowth depends on determining which second-messenger cascade is turned on, when is it turned on, and by which ligand and receptor (Fig. 3). Rho Family of GTPases Stimulation of axonogenesis and growth cone advance, whether through CAMs, the ECM, or growth factor receptors, must at some point act through cytoplasmic effectors in order to modify the actin cytoskeleton (Figs. 2 and 3). The prime candidates are the Rho family of small GTPases. The Rho-dependent control of the actin cytoskeleton has been well characterized in fibroblasts and is now being investigated in neurons (Luo et al., 1997). Axonogenesis in the Xenopus visual system is particularly sensitive to the unregulated activity of the Rho GTPases. Expression of constitutively active forms of RhoA, Rac1, and Cdc42 essentially eliminates RGC axons (Ruchhoeft et al., 1999). However, in RGCs only Rac1 function appears to be necessary for axonogenesis, because only expression of a dominant negative Rac1 is able to inhibit axon initiation (Ruchhoeft et al., 1999). This is similar to observations made in Drosophila (Luo et al., 1994). A possible effector of Rac1 in regulating axonogenesis is the complex of cyclin-dependent kinase 5 (cdk5) and its regulatory subunit p35 (Nikolic et al., 1998). Cdk5, originally isolated through its structural homology to human Cdc2, is abundantly expressed in the nervous system, with a staining pattern similar to actin filaments (Nikolic et al., 1996). In addition, the cdk5/p35 complex colocalizes with Rac1 in growth cones (Nikolic et al., 1996). The dn cdk5 is able to partially rescue axonogenesis in ca Rac1 expressing RGCs (Ruchhoeft et al., 1999). Because growth cones appear to be sensitive to perturbations of Rho-family GTPase activity, the dn Cdk5 likely compensates for the ca Rac1, thus bringing Rac1 activity to a level amenable to the formation and extension of axons. Rac, and possibly Cdc42, antagonise Rho activity by promoting the phosphorylation of myosin-ii heavy chain (MHC) (van Leeuwen et al., 1999). Non muscle myosin-ii is found in an actinomyosin cytoskeletal complex that is rapidly remodeled during growth cone advance and retraction. Activation of myosin-ii by Rho-depending kinases leads to actinomyosin contraction and growth cone retraction (Burridge and Chrzanowska-Wodnicka, 1996). Rac-mediated phosphorylation of MHC causes the disassembly of myosin-ii from the actinomyosin complexes and cell spreading (van Leeuwen et al., 1999). In contrast, Rho promotes myosin-ii activity by stimulating the phosphorylation of myosin-ii regulatory light chain (MLC; Amano et al., 1996; Kimura et al., 1996). In Xenopus axons, disruption of myosin actin interactions or inhibition of MLC kinase (MLCK), causes growth cone collapse and stalling, suggesting that myosin-ii activity is required for growth cone motility (Ruchhoeft and Harris, 1997). Therefore, actinomyosin dynamics may be finely regulated by a series of phosphorylation events of MLC and MHC by the Rho GTPases. Overexpression of Rho GTPases in RGC axons also induces a series of dramatic changes in growth cone morphologies (Ruchhoeft et al., 1999). Expression of and RhoA induces the formation of abnormal, thickened filopodia with a balled appearance, and reduces the area of the growth cone. Consistent with its role in antagonizing Rho activity (reviewed in Lim et al., 1996), expression of wt Cdc42 increases the size and complexity of growth cones and induces the formation of backbranches along the axon shaft. There is also a decrease in the rate of axon extension. In contrast, expression of a dn form of Ccd42 reduces the overall complexity resulting in growth cones with smaller areas and decreases in the number of filopodia, suggesting that RhoA activity may be increased in these growth cones. GROWTH CONE DYNAMICS AT CHOICE POINTS Prototypic growth cones as observed in vitro, are flattened structures characterized by the extension of several filopodia that radiate from the junction of lamellipodia. In vivo, however, RGC growth cones display a wide range of morphologies varying from simple club shapes to highly complex filiform types

7 252 Dingwell et al. (Holt, 1989; Fig. 1). These morphologies appear to vary, depending on where in the optic pathway the growth cone is observed (Holt, 1989; Fig. 1). When single growth cones, filled with lucifer yellow or with the lipophilic dye DiI are visualized at different points along the optic pathway, they tend to be more complex at points where important pathway decisions or choice points are made (Harris et al., 1987; Holt, 1989) [Fig. 1(C,E,F)]. For example, complex growth cones having more filopodia are observed at the optic nerve head and optic chiasm of Xenopus, two points in the pathway where growth cones may make a dramatic changes in their direction of growth [Holt, 1989; Fig. 1(C,F)]. Less complex club-shaped growth cones or ones extending a single filopodium are seen on the retinal surface and in the optic nerve [Fig. 1(B,D)]. These observations are consistent with a similar study made in the mouse visual system (Bovolenta and Mason, 1987). Together, these analyses of growth cones in vivo demonstrate that the dynamics of growth cone morphology varies in a position-specific manner during the development of the visual projection. Growth cones tend to be simple during outgrowth and then acquire a more complex morphology as they reach and turn at key decision points, supporting the idea that these changes reflect responses to cues within their microenvironment. The ability of growth cones to detect and respond to their environment is an integral trait of growth cones themselves. In Xenopus, RGC axons severed from their soma are fully capable of growing and responding appropriately to cues along the optic tract (Harris et al., 1987). HRP-labeled axons isolated from their soma by removing the retina continue to grow and navigate appropriately for up to 4 h. Severed axons do not stray from their normal trajectory, and if growth cones are close enough to their target, they undergo the characteristic morphological and behavioral changes associated with target recognition, including a decrease in speed and the initiation of backbranches [Fig. 1(H); see below]. Together these data demonstrate that isolated growth cones possess the means necessary to detect and respond to pathway and target derived cues. Role of Filopodia Filopodia are thought to play numerous roles during growth cone guidance ranging from motor, steering, and sensory functions (Kater and Rehder, 1995). Therefore, the increased complexity of RGC growth cones at choice points may reflect the importance of filopodia in growth cone guidance. To address this question, the ability of RGC growth cones lacking filopodia to make correct pathway decisions has been tested (Chien et al., 1993). Using an exposed brain preparation, it is possible to treat the developing optic tract with cytochalsin B (CB) at a time when growth cones have crossed the chiasm and have entered the contralateral ventral optic tract. DiI-labeled growth cones treated with CB lack filopodia, yet have active lamellipodia. CB-treated axons have a decrease in the rate of growth. In addition, there is a general disorganization of the optic tract in CB-exposed brains. Normally, when growth cones reach the mid-diencephalon, they make a 50 turn posteriorly, which realigns them with the optic tectum [Harris et al., 1987; Fig. 1(G)]. The CB-treated growth cones, however, fail to reorient and instead continue to grow dorsally toward the pineal. The failure of the CB-treated growth cones to turn at the choice point may reflect an inability either to sense the guidance cue or to respond by turning toward the cue. Evidence from other systems would suggest that filopodia are required for both functions. Because of a high surface area-to-volume ratio, filopodia are able to sample environmental cues effectively. They can also integrate information from multiple signals at one time (Kater and Rehder, 1995). In addition, isolated filopodia can function independently and will respond to an external signal in a contractile manner (Davenport et al., 1993). Recent experiments examining the response of Xenopus spinal neurons toward a chemotropic gradient of glutamate strongly suggest a dual function for filopodia during growth cone turning (Zheng et al., 1996). Preceding the turning response of the growth cone toward a glutamate gradient, an asymmetric distribution of filopodia develops on the side of the growth cone facing the glutamate source. Growth cones that lack filopodia fail to turn toward the gradient yet continue to advance normally. Together, these data demonstrate the importance of filopodia to sense extracellular cues and to translate this information into a mechanical reorientation of the growth cone. Exiting the Eye One of the most dramatic changes in the morphology of RGC growth cones occurs when axons leave the ganglion cell fiber layer and enter the optic nerve head [ONH; Fig. 1(C)], to join the optic nerve. A prime candidate for promoting this turn toward the ONH is the soluble chemoattractant netrin-1 (de la Torre et al., 1997; Deiner et al., 1997; Hopker et al., 1999). Netrin-1 is expressed in the optic disk of a variety of vertebrates, including fish, rodent, and chick (Kennedy et al., 1994; Deiner et al., 1997; Lauderdale et al., 1997; Livesey and Hunt, 1997; Strahle et al., 1997). Likewise, in Xenopus, RNA in situ hybridiza-

8 RGC Growth Cone Dynamics 253 a small glass pipette. This rapidly establishes a gradient in the order of a 1000-fold between the pipette tip and the growth cone. When subjected to a netrin-1 gradient, RGC growth cones specifically orient in a DCC-dependent manner toward the source of netrin-1 (de la Torre et al., 1997). In addition, application of netrin-1 to cultures produces a dramatic change in growth cone morphology, reminiscent of the complex growth cones observed in vivo at the ONH [de la Torre et al., 1997; Holt, 1989; Fig. 1(C)]. The turning response by the growth cone toward a netrin-1 gradient is converted to a repulsive response by the depletion of intracellular camp or the inhibition of protein kinase A (PKA; Ming et al., 1997). These findings place netrin-1 within the group I of soluble chemotropic agents (Song and Poo, 1999). Netrin-1, brainderived neurotropic factor (BDNF), acetylcholine (ACh), and myelin-associated glycoprotein (MAG) are included in group I because the direction of their induced turns depends on the level of cytosolic camp, the activity of PKA, and the presence of extracellular Ca 21. Growth cone responses to group II members, which include semaphorin III (SemaIII) and NT-3, on the other hand, are independent of Ca 21 and are regulated by cytosolic cgmp and PKG. Figure 4 Model for growth cone navigation at the optic nerve head. RGC axons grow along the vitreal surface of the retina toward the ONH in close contact with the outgrowth promoting ECM component laminin. Netrin-1 is abundantly expressed throughout the ONH. Netrin signaling on the ONH side of the growth cone increases the concentration of camp. In contrast, on the vitreal side of the growth cone, laminin-1 antagonizes the netrin signal and lowers the camp levels, thus creating a camp gradient across the growth cone. The high camp levels triggers actin polymerization and growth cone turning toward the ONH, whereas low levels of camp leads to actin depolymerization and growth cone repulsion away from the vitreal surface. RGC, retinal ganglion cell; ONH, optic nerve head; ECM, extracellular matrix; DCC, deleted in colorectal cancer. tion experiments show that netrin-1 is expressed in the ONH and optic nerve (ON; de la Torre et al., 1997; Fig. 4). The receptor for netrin-1, Deleted in colorectal cancer (DCC), is also expressed in RGC axons (de la Torre et al., 1997), consistent with a role for netrin-1 in mediating the exit of growth cones from the eye. To address whether netrin-1 alone could cause growth cone turning, the effect of soluble netrin-1 gradients on RGC growth cone behavior have been examined in culture using the growth cone turning assay (Lohof et al., 1992). In this assay, growth cones are subjected to a stable gradient of soluble chemoattractant produced following its pulsatile release from ECM Modulation of Netrin Response The conversion of growth cone attraction to repulsion may play a significant role during the exit of growth cones through the ONH [Hopker et al., 1999; Figs. 1(C) and 4]. As described above, growth cones in culture turn toward a gradient of netrin-1; however, if the same growth cones are cultured on the ECM substrate laminin, the turning response is converted to repulsion (Hopker et al., 1999). Repulsion to a netrin-1 gradient can be inhibited if growth cones are treated with function-blocking antibodies against b 1 integrins, a component of the laminin receptor a 6 b 1 integrin complex. This strongly suggests that b 1 integrins are involved in changing the growth cone response to netrin gradients. A soluble peptide fragment of the laminin-1 B1 chain YIGSR, which mediates cell attachment but has no neurite outgrowth promoting activity, mimics the laminin-induced switch from attraction to repulsion. Surprisingly, however, a 19- amino acid peptide containing the IKVAV sequence from the laminin-1 A1 chain, which promotes neuritic formation and outgrowth (Sephel et al., 1989), has no effect on the turning response. Function-blocking antibodies to the 67-kD laminin receptor, which binds to the YIGSR peptide and may function to stabilize laminin binding to integrin receptors, are able to block both laminin and YIGSR effects on growth cone turning. Application of Sp-cAMPS (a camp analogue

9 254 Dingwell et al. that activates PKA) abolishes the YIGSR effect on growth cone turning. Likewise, YIGSR-treated growth cones show a net decrease in camp levels compared to growth cones treated with netrin alone. Therefore, the B1 chain of laminin through the YIGSR peptide may be responsible for mediating the conversion in growth cone responses. To address whether laminin and specifically the YIGSR peptide could affect growth cone dynamics in vivo, developing retinae have been treated with YIGSR peptides (Hopker et al., 1999) at the time when RGC are first extending axons across the vitreal surface. In contrast to control-treated eyes in which axons converge at the ONH into a single large fascicle that exits the eye [Fig. 1(C)], RGC axons in YIGSRtreated eyes form disorganized fascicles that in many cases fail to exit the eye. Together, these data suggest that laminin and its receptors play a critical role in regulating the orderly arrangement and navigation of RGC axons at the ONH. A question that has remained unanswered for some time is how are axons able to turn away from the vitreal surface and enter the ONH (Fig. 1). These results suggest that growth cone responses to a netrin gradient may be involved in mediating this event. As discussed previously, the response of RGC axons to a netrin gradient can be either attractive or repulsive, depending on the level of intracellular camp. In the case of group I chemotropic agents such as netrin, high camp levels lead to actin polymerization, whereas low levels of camp trigger actin depolymerization. By manipulating the levels of camp across a growth cone, it is then possible to control the level of actin polymerization/depolymerization and cause turning toward or away from a guidance cue. In fact, an increase in the concentration of intracellular camp of only 10% across a growth cone is sufficient to induce attractive turning (Lohof et al., 1992). At the ONH, laminin-1 is localized to the vitreous surface (Hopker et al., 1999), whereas netrin-1 is expressed abundantly throughout the ONH (de la Torre et al., 1997; Fig. 4). Thus, RGC growth cones reaching the ONH are confronted with two opposing signals that could establish an intracellular camp gradient. Low levels of camp could be established in the growth cone at the vitreal surface by netrin/laminin responses, whereas high concentrations of camp could develop on the ONH side in the presence of netrin only. Together, the difference in camp concentrations across the growth cone could cause an increase in actin polymerization on the ONH side and depolymerization close to the vitreal surface, thus promoting a turn away from the fiber layer and growth into the ONH (Fig. 4). Ephrin-B Dependent Turning at the Optic Chiasm In the premetamorphic tadpole, all RGC axons cross at the optic chiasm and project to the contralateral side of the brain [Fig. 1(F)]. However, during metamorphosis in which the tadpole undergoes extensive external changes, an ipsilateral retinothalamic projection arises from a subpopulation of newly born cells in the temporal and ventral retina (Hoskins and Grobstein, 1985a,b). This ipsilateral projection does not arise if development is stalled at a premetamorphic stage (Hoskins and Grobstein, 1984, 1985c). Moreover, transplants of postmetamorphic ventral retina, which would normally project ipsilaterally, into retinas of young embryos do not develop an ipsilateral projection (Nakagawa et al., 2000). Together, these data suggest that early tadpoles do not express the appropriate ipsilateral guidance cue(s) at the chiasm. The Eph receptors and their ligands, the ephrins, are required for a variety of developmental processes including the appropriate topographic mapping of RGC axons to the optic tectum (Cheng et al., 1995; Drescher et al., 1995). Because Eph receptors are expressed at high levels in the ventrotemporal retina (Cheng et al., 1995; Drescher et al., 1995), Eph ephrin interactions at the chiasm could account for the ipsilateral projection. Indeed, there is a stage-dependent expression of ephrin-b at the chiasm. Ephrin-B expression rises sharply following metamorphosis (stage 60) and is not detected in the chiasm of premetamorphic tadpoles (stage 54; Nakagawa et al., 2000). Targetted expression of ephrin-b2 in the chiasm of premetamorphic tadpoles is sufficient to induce an ectopic ipsilateral retinal projection that is restricted to the EphB-expressing population of RGCs growth cones from the ventral retina (Nakagawa et al., 2000). Thus, the stage-dependent EphB and ephrin-b interaction at the chiasm can generate an ipsilateral projection that is essential for establishing binocular vision in the adult frog. Growth Cone Turning in the Optic Tract The development of the visual projection depends on an orderly advance RGC growth cones. This is achieved by an integrated response of the growth cone to a variety environmental cues acting either from a distance through diffusion gradients or by local shortrange effects that include cell cell interactions [Tessier-Lavigne and Goodman, 1996; Fig. 2(C)]. In the Xenopus visual system, both types of cues appear to be important for growth cone turning at choice points. The turn at the mid optic tract, where growth cones turn caudally to reorient themselves with their

10 RGC Growth Cone Dynamics 255 target, involves local positional cues tightly associated with the neuroepithelium. Experiments in which ectopic third eyes are transplanted to various regions of the developing brain suggest that a soluble cue emanating from the target might be responsible for attracting growth cones to the tectum (Harris, 1986). In the majority of cases, growth cones from the transplanted eyes navigate directly toward the optic tectum, even when they enter the brain far from the optic tract. However, when the tectum is removed, RGC axons still retain their ability to navigate appropriately along the optic tract right up to the border where their target would normally have been found (Taylor, 1990). Coculture experiments fail to provide any evidence that the tectum secrets long-range, diffusible signals that attract RGC axons. Growth cones extending from retinal explants cocultured near tectal explants in a collagen matrix do not grow preferentially toward the tectal explant; instead, they grow equally toward or away from the piece of tectum (Harris et al., 1985). One must therefore consider the possibility of a global guidance system in which positional information is contained in the neuroepithelium. Consistent with this idea is the result that growth cones that enter a rotated piece of neuroepithelium are deflected in the same direction as the tissue rotation (Harris, 1989; Taylor, 1991). These positional cues appear early during development, appearing as early as stage 24 (Harris, 1989; Taylor, 1991; Cornel and Holt, 1992). TARGET SELECTION Changes in Growth Cone Dynamics During axon growth along the optic tract, growth cones advance at a constant rate of approximately 50 mm h 21 (Harris et al., 1987). However, when they reach the border of their target, the optic tectum, growth cones undergo a rapid and remarkable change in behavior [Harris et al., 1987; Fig. 1(H,I)]. Their growth rate decreases significantly to about 15 mm h 21, and they lose their characteristic morphology, becoming highly complex and elongated with lamellipodia and filopodia extending in all directions, essentially adopting a hairy appearance (Harris et al., 1987; Holt, 1989). Growth cones also begin to meander through the tectum with sharp turns, in marked contrast to the smooth trajectories observed while axons are growing across the forebrain (Harris et al., 1987). Target innervation also triggers the development of small side (terminal) branches along the axon shaft, leading to the development of a complex network of terminal arbors [Sakaguchi and Murphy, 1985; Harris et al., 1987; Holt, 1989; Fig. 1(I)]. The primary axon always advances by way of a single growth cone, although growth cones are rarely observed on the secondary branches (Harris et al., 1987). However, once terminal branching begins, the appearance of the primary growth cone changes to resemble a side branch (Harris et al., 1987). FGF What triggers these dramatic changes in behavior? One possibility is a switch in the extracellular environment to one that does not favor axon growth. Laminin, for example, is expressed only along the optic tract during the time that RGCs are extending toward the tectum (Cohen et al., 1987). At later stages, when cultured in vitro, these axons lose their response to laminin in a target-dependent manner (Cohen et al., 1986, 1989). Thus, a change in laminin expression and/or growth cone response to laminin at the tectal border may trigger a change in axon growth rate. In the Xenopus visual system, the growth factor FGF2 together with FGF2-binding heparan sulfate proteoglycan sidechains (HS) are potent stimulators of RGC axon outgrowth [McFarlane et al., 1995; Walz et al., 1997; Fig. 2(B)]. Both molecules are highly expressed along the optic tract, but their levels drop off significantly within the anterior tectum (Mc- Farlane et al., 1995; Walz et al., 1997). Exogenous application of either FGF2 or HS to the developing optic tract causes a severe mistargeting of retinal growth cones in which axons grow dorsally or ventrally around their target (McFarlane et al., 1995; Walz et al., 1997). These data are consistent with a model in which a change in the rate of axon outgrowth at the target border is critical for target recognition and innervation. Because FGF2 is a potent stimulator of axon outgrowth, the continued stimulation by FGF2 may make the growth cones insensitive to cues within their target. However, experiments in which RGC axons express a dominant negative FGFR, XFD, argues against this hypothesis. XFD-expressing axons are significantly impaired in their ability to grow, yet they are nevertheless able to navigate appropriately along the optic pathway (McFarlane et al., 1996). However, when reaching the tectal border, many of these axons veer around rather than entering their target. In contrast, growth cones expressing nonfunctional FGFRs (i.e., that cannot act as dominant negatives nor stimulate FGFR activity), enter the tectum normally (McFarlane et al., 1996). This clearly shows that FGFR signaling is required for target recognition, yet the evidence appears paradoxical, because either sustained increases or decreases in FGFR signaling

11 256 Dingwell et al. result in similar targeting defects. One simple model to explain these observations is that during target recognition, growth cones sense a change in FGFR signaling, likely from high to low, which then triggers their morphological and behavioral changes. Such growth cones switch from active growth to arborization [Figs. 1(H) and 3]. FGFR Signalling FGFRs signal through a multitude of secondary pathways (e.g., src, PLCg, ras/raf/mapk; reviewed in Klint and Claesson-Welsh, 1999). During growth cone extension, FGFR signaling whether stimulated via FGF2 or CAMs appears to be mediated by the activation of the phospholipase C-g (PLCg) pathway [Doherty and Walsh, 1996; Lom et al., 1998; Figs. 2(B,D) and 3]. Pharmacological inhibition of the PLCg pathway significantly reduces FGFR-dependent neurite outgrowth in culture (Saffell et al., 1997). Likewise, inhibition of PLCg activity in Xenopus embryos using the exposed brain preparation reversibly impairs RGC axon outgrowth; yet, significantly, there is no effect on target recognition (Lom et al., 1998). Therefore, target recognition and innervation probably involve an FGFR-dependent signaling cascade that does not include PLCg. Activated FGFRs in Xenopus associate with a number of adapter/effector molecules (SH2/SH3 domain proteins) that link activated receptors to downstream second-messenger cascades (Anderson et al., 1990; Klint and Claesson- Welsh, 1999: Fig. 3). Adapters that are known to associate with the FGFR in Xenopus include Grb2 and Nck (Ryan and Gillespie, 1994). Nck may be a candidate component of the FGFR-dependent target recognition signal. In Drosophila, DOCK (the homologue to vertebrate Nck) mutants are characterized by visual system pathfinding errors in which R cell axons exhibit pronounced targeting defects (Garrity et al., 1996). SemaIIIA and cgmp The group II guidance protein, semaiiia, induces an age-dependent response in Xenopus RGC growth cones (D. Campbell, J. Lopez, T. Odagiri, personal communication). At early stages (before stage 32), growth cones are not repulsed by a gradient of semaiiia nor collapse when semaiiia is added to the culture medium. However, axons from retinal explants older than stage 32 respond robustly to semaiiia, and these responses are converted by the activation of PKG. In the Xenopus visual system the semaiiia receptor neuropilin-1 is not expressed until after stage 32 (Fujisawa et al., 1995), which would explain the age-dependent responses. In addition, semaiiia in Xenopus is predominantly expressed within the caudal third of the tectum (Regan and Tannahill, personal communication). Therefore, semaiiia may play a role in visual system development following target innervation, such as preventing axons from overshooting their target and entering the hindbrain. Alternatively, semaiiia may act to promote branching and arborization of axons within their target [Fig. 1(H,I)]. In culture, Xenopus growth cones recovering from semaiii-induced collapse are more likely to form branches than control growth cones (J. Lopez, D. Campbell, personal communication). TOPOGRAPHICAL MAP FORMATION Once within their target, growth cones form a topographical map corresponding to the position of their cell bodies within the retina [reviewed in Holt and Harris, 1993; Figs. 1(I) and 2(E)]. Axons from the anterior retina project to the posterior tectum, posterior retina to the anterior tectum, dorsal retina to the ventral tectum, and ventral retina to the dorsal tectum. Considerable effort has been made in identifying and characterizing what Sperry (1963) referred to as cytochemical tags. It is now thought that Eph receptor tyrosine kinases and their ligands, the ephrins, expressed in gradients within the retina and tectum, respectively, are partially responsible for appropriate topographical mapping (Cheng et al., 1995; Drescher et al., 1995). We will not review the evidence for this here, because this work has largely been done in other systems, and it is reviewed elsewhere (Drescher et al., 1997 and references therein). In Xenopus, Eph receptors and ephrins are expressed in appropriate gradients in the retina and tectum (S. Nakagawa and F. Mann, personal communication) and are likely to play a similar role in this system. CONCLUSION The retinotectal system in Xenopus is an excellent model system for studying the mechanisms that control growth cone dynamics along an entire neural pathway. This review describes a wide breadth of research from our laboratories aimed at understanding the molecular and cellular mechanisms involved in this process. And, although our efforts may have addressed many questions, there are still many more to be answered. The main challenges ahead will be to find the molecular mechanisms that integrate the various guidance cues and then drive growth cone dynamics during particular pathfinding events.

12 RGC Growth Cone Dynamics 257 The authors thank the many past and present members of the Harris and Holt laboratories who contributed to the majority of work cited in this review. They also thank D. Campbell, J. Lopez, F. Mann, S. Nakagawa, T. Odagiri, A. Regan, and D. Tannahill for allowing citation of their unpublished work. REFERENCES Amano M, Ito M, Kimura K, Fukata Y, Chihara K, Nakano T, Matsuura Y, Kaibuchi K Phosphorylation and activation of myosin by Rho-associated kinase (Rhokinase). J Biol Chem 271: Amaya E, Stein PA, Musci TJ, Kirschner MW FGF signaling in the early specification of mesoderm in Xenopus. Development 118: Anderson D, Koch CA, Grey L, Ellis C, Moran ME, Pawson T Binding of SH2 domains of phospholipase C- gamma-1, GAP and Src to activated growth factor receptors. Nature 250: Aota S, Yamada KM Integrin functions and signal transduction. Adv Exp Med Biol 400B: Beggs HE, Soriano P, Maness PF NCAM-dependent neurite outgrowth is inhibited in neurons from Fyn-minus mice. J Cell Biol 127: Bixby JL, Jhabvala P Tyrosine phosphorylation in early embryonic growth cones. J Neurosci 13: Bovolenta P, Mason C Growth cone morphology varies with position in the developing mouse visual pathway from retina to first targets. J Neurosci 7: Burridge K, Chrzanowska-Wodnicka M Focal adhesions, contractility, and signalling. Annu Rev Cell Dev Biol 12: Cheng HJ, Nakamoto M, Bergemann AD, Flanagan JG Complementary gradients in expression and binding of ELF-1 and Mek4 in development of the topographic retinotectal projection map. Cell 82: Chien CB, Harris WA Axonal guidance from retina to tectum in embryonic Xenopus. Curr Top Dev Biol 29: Chien CB, Rosenthal DE, Harris WA, and Holt CE Navigational errors made by growth cones without filopodia in the embryonic Xenopus brain. Neuron 11: Cohen J, Burne JF, McKinlay C, Winter J The role of laminin and the laminin/fibronectin receptor complex in the outgrowth of retinal ganglion cell axons. Dev Biol 122: Cohen J, Burne JF, Winter J, Bartlett P Retinal ganglion cells lose response to laminin with maturation. Nature 322: Cohen J, Nurcombe V, Jeffrey P, Edgar D Developmental loss of functional laminin receptors on retinal ganglion cells is regulated by their target tissue, the optic tectum. Development 107: Cornel E, Holt C Precocious pathfinding: retinal axons can navigate in an axonless brain. Neuron 9: Cox ME, Maness PF Neurite extension and protein tyrosine phosphorylation elicited by inducible expression of the v-src oncogene in a PC12 cell line. Exp Cell Res 195: Davenport RW, Dou P, Rehder V, Kater SB A sensory role for neuronal growth cone filopodia. Nature 361: de la Torre JR, Hopker VH, Ming GL, Poo MM, Tessier- Lavigne M, Hemmati-Brivanlou A, Holt CE Turning of retinal growth cones in a netrin-1 gradient mediated by the netrin receptor DCC. Neuron 19: Deiner MS, Kennedy TE, Fazeli A, Serafini T, Skarnes WC, Tessier-Lavigne M, Sretevan DW Netrin-1 and DCC mediate axon guidance locally at the optic disc: loss of function leads to optic nerve hypoplasia. Neuron 19: Doherty P, Walsh FS CAM-FGF receptor interactions: a model for axonal growth. Mol Cell Neurosci 8: Drescher U, Bonhoeffer F, Muller BK The Eph family in retinal axon guidance. Curr Opin Neurobiol 7: Drescher U, Kremoser C, Handwerker C, Loschinger J, Noda M, Bonhoeffer F In vitro guidance of retinal ganglion cell axons by RAGS, a 25 kda tectal protein related to ligands for Eph receptor tyrosine kinases. Cell 82: Fujisawa H, Takagi S, Hirata T Growth-associated expression of a membrane protein, neuropilin, in Xenopus optic nerve fibers. Dev Neurosci 17: Garrity PA, Rao Y, Salecker I, McGlade J, Pawson T, Zipursky SL Drosophila photoreceptor axon guidance and targeting requires the dreadlocks SH2/SH3 adapter protein. Cell 85: Giancotti FG, Ruoslahti E Integrin signaling. Science 285: Gunderson RW Response of sensory neurites and growth cones to patterned substrata of laminin and fibronectin in vitro. Dev Biol 121: Hall DE, Neugebauer KM, Reichardt LF Embryonic neural retinal cell response to extracellular matrix proteins: developmental changes and effects of the cell substratum attachment antibody (CSAT). J Cell Biol 104: Harris WA Homing behaviour of axons in the embryonic vertebrate brain. Nature 320: Harris WA Local positional cues in the neuroepithelium guide retinal axons in embryonic Xenopus brain. Nature 339: Harris WA, Holt CE, Bonhoeffer F Retinal axons with and without their somata, growing to and arborizing in the tectum of Xenopus embryos: a time-lapse video study of single fibres in vivo. Development 101: Harris WA, Holt CE, Smith TA, Gallenson N Growth cones of developing retinal cells in vivo, on culture surfaces, and in collagen matrices. J Neurosci Res 13: Holt CE Does timing of axon outgrowth influence initial retinotectal topography in Xenopus? J Neurosci 4: