During the past decade, our understanding of axon-pathfinding mechanisms has been advanced through the use of invertebrate and vertebrate model systems. Two well-known model systems have provided many insights into the processes of axon pathfinding and we will confine our discussion to these examples: the vertebrate retinal projection and the midline of the developing
CNS. Importantly, many of the guidance cues described in the previous section play integral roles in mediating axon guidance during development of these systems. Moreover, these two systems provide an introduction to the general processes that underlie axon guidance in other neuronal populations.
The retinal projection is the axonal pathway linking retinal ganglion cells (RGCs) of the eye to their primary target in the brain, the optic tectum (in lower vertebrates) or superior colliculus (in mammals) (Figures 6a-6d). The eye is a highly complex light-capturing organ which generates neuronal signals corresponding to external visual information and conveys them via the retinal projection to visual centers in the brain. To ensure that visual information is correctly relayed to the CNS, axons of embryonic RGCs must find their way from the eye to their CNS targets. The number of RGC axons that must navigate correctly from the eye is impressive. In mice, for example, more than 50000 RGC axons exit from each developing eye, while in humans, over a million RGC axons must navigate accurately during embryonic development.
In this section, we will describe the different decisions that RGC axons make at various choice points during pathfinding to their targets in the brain. Although much remains unknown, it is clear that attractive and repulsive cues in the developing brain act in concert to guide RGC axons correctly. An important concept, shared by navigating axonal populations across a wide evolutionary distance, is that the retinal pathway is broken into smaller segments with distinct molecular characteristics. The subdivision into shorter distinct segments is used to simplify navigation, as well as allowing multiple regulatory check points.
220.127.116.11.1 General organization of the retinal pathway The first step in the journey of a RGC axon is to extend across the (vitreal) surface of the retina toward the optic nerve head (ONH), or future optic disk, where they form a bundle of axons that will exit the eye. Axons extend through the ONH toward the back of the eye (pigmented epithelium) where they emerge as a tightly fasciculated bundle to form the optic nerve. The optic nerve enters the brain at the ventral diencephalon and axons grow toward the optic chiasm at the midline. The chiasm is a key choice point as axons may cross to the other
Figure 6 Xenopus retinotectal projection. In the Xenopus retinotectal projection, RGC axons navigate from the retina to the optic tectum in the brain. Guidance molecules are expressed at key choice points to ensure fidelity of pathfinding. Some of the major pathfinding decisions are made at: a, the optic nerve head; b, the optic chiasm; c, the optic tract; and d, the tectum itself (see text for details).
side of the brain or remain ipsilateral. After passing through the optic chiasm, RGC axons form the optic tract which courses dorsally along the lateral surface of the diencephalon. The final step in the retinal pathway involves the departure of RGC axons from the optic tract and entry into the main synaptic target, the optic tectum in lower vertebrates, the superior colliculus in mammals, in the midbrain. Within the target area, retinal axons terminate next to their retinal neighbors, thus forming a precise topographic map of visual space (Figures 6a-6d).
18.104.22.168.2 Pathfinding across and out of the retina In the first step of the retinal pathway, axons converge on the ONH. Remarkably, axons arising from all points of origin across the retinal surface orient immediately toward the ONH, suggesting that they receive directional information early in axonogenesis (reviewed in Oster et al., 2004). The mechanisms responsible for this directed intraretinal growth, although not well understood, are beginning to be uncovered (reviewed in Mann et al., 2004). For example, experimental studies that perturb the function of cell adhesion molecules (CAMs), such as L1, neural CAM (NCAM), and neurolin, that are expressed on both RGC axons and their substrate neuroepithelial cells, show that these molecules play a role in intraretinal growth (Brittis et al., 1992; Ott et al., 1998). In addition, chondroitin sulfate proteoglycans (CSPGs) have been proposed to play a role in directing axons toward the ONH. These extracellular matrix proteins, known to inhibit axon growth in vitro (Snow et al., 1991), are expressed in a receding ring in the developing rat retina just peripheral to differentiating RGCs. Enzymatic removal of CSPGs causes defects in directional guidance with some axons erroneously heading peripherally (Brittis et al., 1992), suggesting that the ring of CSPG may serve to push growing axons centrally. However, a CSPG ring has only been observed in the rat retina, indicating that it may not represent a common guidance mechanism. Another class of molecule implicated in intraretinal guidance is the B-type family of Eph tyrosine kinases. Some members, such as EphB2, are expressed in a ventral-to-dorsal gradient in the retina and mutant mice lacking EphB2 and EphB3 display pathfinding errors within the retina (Birgbauer et al., 2000).
When axons reach the ONH they make a sharp change in direction (approximately 90°) to enter the ONH, marking this as a critical decision point. This behavior is mediated by the chemoattractant netrin-1, which is localized to the ONH (de la Torre et al., 1997; Deiner et al., 1997). In netrin-1 and DCC-deficient mice, axon growth across the retinal surface to the ONH is unaffected, but many axons fail to enter the ONH, resulting in optic nerve hypopla-sia (Deiner et al., 1997). In this context, netrin-1 appears to be acting as a short-range attractant, guiding RGC axons that have arrived at the ONH into the optic nerve itself. This contrasts with its well-characterized long-range role in guiding com-missural axons to the floorplate of the ventral spinal cord (Tessier-Lavigne et al., 1988). Within the optic nerve, axons fasciculate into bundles through the action of homophilic binding of axonally expressed adhesion molecules. Sema5A, which is known to repel retinal axons in vitro, helps to confine the axons to the nerve bundle, as it is expressed as an inhibitory sheath around the nerve, preventing them from wandering off the defined path (Goldberg et al., 2004).
22.214.171.124.3 Axon divergence at the chiasm From the optic nerve, RGC axons enter the brain and approach the midline, a key intermediate target. At this point, axons have to make the critical decision of whether to cross the CNS midline at the optic chiasm. At a functional level, the decision for RGC axons to cross or not cross at the chiasm correlates with the degree of binocular overlap. In species with eyes placed laterally there is no binocular overlap and the retinal projections are completely crossed (i.e., contralateral). Thus, the visual information from the left and the right eyes is processed independently. In species with forward-facing eyes and, hence, binocular vision, the visual fields of both eyes overlap to some extent. For example, a subset of RGCs in the left eye will receive stimuli that overlap with those in the right eye. To process such shared visual information, some RGC
axons from the region of binocular overlap project ipsilaterally while others project contralaterally. This arrangement enables information about the same point in visual space from the two eyes to be brought together in the brain. Thus, in vertebrates with binocular vision, the ability to integrate shared visual information relies on the decision of RGC axons to cross or not to cross at the chiasm.
The first insights into the molecular regulation of crossing came from studies in the amphibian, Xenopus. In the filter-feeding tadpoles the eyes are placed laterally, so there is no binocular overlap as all the visual projections are crossed. During metamorphosis, the eyes migrate dorsally to the top of the head, creating binocular vision essential for the frog's new prey-catching lifestyle. Beginning at metamorphosis, axons arising from RGCs in the ventrotemporal binocular part of the retina extend ipsilaterally. What makes these axons alter their behavior? It turns out that ephrin-B expression is switched on during metamorphosis at the midline at the optic chiasm (Nakagawa et al., 2000). Ephrin-B is the ligand for EphB receptors (Figure 5) and the ventrotemporal axons express high levels of EphB receptors. Since ephrin-B acts repulsively, the current working model is that ventrotemporal axons are deflected into the ipsilateral optic tract by the ephrin-B signal encountered at the midline chiasm in metamorphosing and postmetamorphic frogs. The same mechanism seems to play a role in other vertebrate species such as mice (reviewed in Mann et al., 2004). Significantly, animals without binocular vision, such as zebra fish and chick, lack ephrin-B expression at the chiasm, suggesting that this mechanism has arisen during evolution to accommodate the development of binocular vision in vertebrates.
The determinant of regionalized EphB expression in the retina appears to be the transcription factor Zic-2, which is exclusively expressed in the ventro-temporal retina after the onset of metamorphosis in Xenopus and in the temporal retina in embryonic mice (Herrera et al., 2003). As with ephrin-B expression at the chiasm, Zic-2 is not expressed in the retina of organisms without visual field overlap, such as the chick (Herrera et al., 2003). In addition, the transcription factor Isl-2 is expressed only by RGCs that cross the chiasm and has been shown to repress the expression of Zic-2, preventing them projecting ipsilaterally (Pak et al., 2004). Therefore, Zic-2 and Isl-2 transcription factors are expressed in mutually exclusive areas of the retina and are key regulators of an RGC's sensitivity to directional signals at the midline.
Finally, gene knockout studies in mice have shown that Slit1 and Slit2 are involved in the correct development of the chiasm. Mice deficient for both genes exhibit multiple retinal axon-pathfinding errors, including the formation of an ectopic chiasm (Plump et al., 2002). Slit1 and Slit2 are expressed in complementary domains surrounding the path of the growing retinal axons, creating a repulsive barrier around the chiasm, which acts to channel the axons into a narrow corridor across the ventral diencephalon. Slits, therefore, unlike ephrin-B, are not involved in the decision to cross the midline but rather in the anterior-posterior positioning of the chiasm. Together these two complementary molecular systems help to determine the exact path of axons across the midline.
126.96.36.199.4 Pathfinding in the optic tract In the last segment of their journey, RGC axons have to elongate along the optic tract in the diencephalon and perform a 45° turn posteriorly in the mid-dien-cephalon to reach the anterior border of the tectum. Again, a variety of different cues work together to guide axons in the optic tract, and the expression of repulsive molecules stops them from leaving the optic tract and innervating inappropriate territories.
In vitro studies performed in Xenopus have highlighted a role for the netrin-1 and Sema3A cues in RGC axon pathfinding in the optic tract. In addition to its function at the ONH, netrin-1 is also expressed in the dorsal diencephalon in an area that is nonoverlapping with, but adjacent to, growing retinal axons, suggesting that it might act as a repellent cue to prevent RGC axons from leaving the optic tract (Shewan et al., 2002). Sema3A is expressed along the boundary of a segment of the optic tract and may act as a repulsive cue forcing RGC axons to turn posteriorly (Campbell et al., 2001).
A zebra fish mutant, called astray, corresponding to a mutation in the Robo2 gene has provided evidence that an interaction between Slit and Robo is essential for axon pathfinding in the optic tract. The astray mutant phenotype exhibits severe retinal axon-pathfinding defects, including defasciculation of axons in the optic tract and widespread invasion of inappropriate regions of the brain (Fricke et al., 2001; Hutson and Chien, 2002). This suggests that Slit/Robo signaling may be required to prevent axons from leaving the optic tract. These results are also consistent with the finding that Slit2 controls RGC axon pathfinding and targeting in vivo within the rat diencephalon (Ringstedt et al., 2000).
Slit/Robo signaling also seems to be implicated in the topographic sorting of axons in the optic tract.
In this process, axons from the dorsal part of the retina are sorted into the ventral brachium of the optic tract, and axons from the ventral part of the retina axons are sorted into the dorsal brachium of the optic tract in fish (Scholes, 1979; Stuermer, 1988). In a screen for zebra fish mutants with retinal axon-pathfinding defects, two mutants were isolated, exhibiting optic tract-sorting defects: boxer (box), dackel (dak). Boxer and dackel were identified to encode for genes involved in heparan sulfate proteoglycan (HSPG) biosynthesis and therefore in both mutants the level of HSPG is dramatically reduced (Lee et al., 2004). Remarkably, the dak box double mutant exhibits a severe phenotype similar to the astray zebra fish mutant, indicating that HSPG biosynthesis regulates Slit function. Indeed, there is accumulating evidence that heparan sulfate is essential for the function of Slit (Liang et al., 1999; Hu, 2001; Ronca et al., 2001). However, exactly how the HSPGs interact with Slit and regulate its function remains unknown.
Further evidence supports an essential role for HSPGs and CSPGs in axon navigation within the optic tract. Both are highly enriched in the Xenopus optic tract and enzymatic removal of HSPG during development results in abnormally short retinal projections, suggesting that HSPG plays a role in promoting growth. Addition of an HSPG-binding growth factor, FGF2, to such HSPG-free brains enables axons to continue growing but they then follow extremely aberrant trajectories, suggesting that HSPG is dually involved in guidance and growth (Walz et al., 1997). Exogenous HSPG added in vivo to the embryonic diencephalon causes a highly penetrant mistargeting phenotype in which retinal axons avoid entering the tectum, suggesting that HSPG plays an important role in target recognition (Walz et al., 1997). Exogenously applied CSPGs lead to defasciculation of the optic tract and axons invade inappropriate territories (Walz et al., 2002). In addition, it has been found recently that both HSPGs and CSPGs interact with the guidance cue Sema5A and differentially modulate its action on axons in the developing rat brain (Kantor et al., 2004).
At the end of their journey, RGC axons enter their target area, and are topographically sorted to ensure correct mapping of the retinal image. This topographic mapping mechanism involves the A- and B-type of ephrin guidance cues and their Eph receptors. Since mapping is not strictly part of the axon-pathfinding process, it is not included in this article (but is reviewed in McLaughlin et al., 2003).
In conclusion, we have seen that a combination of molecules act as attractive and/or repulsive cues to guide RGC axons during their journey to innervate their appropriate target in the brain. However, only a small part of this complex process has been elucidated so far, and we are just beginning to understand the mechanisms and the molecular determinants involved in each step of the retinal pathway.
188.8.131.52 The Midline Choice Point as a Model System
second model system that has provided many important insights into the mechanisms underlying axon guidance has been the ventral nerve cord of Drosophila. Many neurons from this population are interneurons, that is, neurons that connect with other neurons. As with many other regions in the developing CNS of bilaterally symmetrical organisms, most (but not all) of these interneurons extend axons toward the midline. A brief synopsis of the development of the Drosophila ventral nerve cord is outlined in Figure 7a. In each segment of the ventral nerve cord, interneuron axons cross the midline at either of two defined points, known as the anterior and posterior commissures, and then project parallel to the midline in distinct longitudinal fascicles. Importantly, those axons that have crossed the mid-line never do so again in the wild-type situation.
Thus, the ventral midline represents an important choice point for interneuron growth cones. Decisions that need to be correctly made include: (1) whether to remain ipsilateral or to cross the midline and project contralaterally; (2) which commissure to cross in; and (3) which longitudinal fascicle to extend in postcrossing. How is this accomplished? A number of genetic and biochemical studies have led to the identification of many factors responsible for these decisions. Firstly, crossing axons are attracted to the midline by netrin-1-A and netrin-1-B expressed by midline cells (Harris et al., 1996; Mitchell et al., 1996). These axons must then leave the attractive midline to continue toward their targets.
Large-scale screens for Drosophila mutants that display guidance defects at the midline have isolated three key players involved in expelling crossing axons from the midline: Commissureless (Comm), Robo, and Slit (reviewed in Kaprielian et al., 2001). Comm mutants lack all commissural tracts. Robo mutants exhibit a thickened midline, a result of ipsilateral axons crossing the midline and contralateral axons recrossing the midline several times. Slit mutants show a thickened midline, resulting from all spinal axons collapsing onto the midline. These observations led to the following model for axon guidance at the Drosophila ventral midline: Slit secreted by the midline glial cells acts as a short-range repellent. Axons that extend contralaterally repress the surface expression of Robo until past the midline, after which it is upregulated at the growth cone surface to ensure exit from the midline and to prevent recrossing. Axons that remain ipsilateral express Robo from the outset and so are repelled from the midline by Slit. A combination of Robo and the other two Slit receptors, Robo2 and Robo3, is then used to specify lateral positions within the longitudinal axon bundles (Rajagopalan et al., 2000; Simpson et al., 2000; Figure 7).
Comm regulates the surface expression of Robo, therefore being the switch mechanism between the contra- and ipsilateral decision (Keleman et al., 2002). A recent study has resolved how this occurs (Keleman et al., 2005). In commissural axons expressing Comm, Robo is sorted to the degradation machinery, so that no Robo reaches the growth cone surface and these axons can cross the midline. In postcrossing axons and ipsilateral neurons, which lack Comm, Robo is targeted to the growth cone membrane; therefore, these axons are repelled by the midline repellent Slit.
184.108.40.206.2 The vertebrate spinal cord The vertebrate spinal cord is another key model system for axon guidance and a comprehensive review of this area can be found elsewhere (reviewed in Kaprielian et al., 2001). Importantly, the role of many of the molecules that regulate midline crossing has been strongly conserved during evolution. In the developing vertebrate CNS, commissural interneurons in the dorsal regions of the spinal cord extend axons ventrally toward the spinal cord floorplate (Figure 7b). This specialized area, considered to be functionally equivalent to the midline of the Drosophila ventral nerve cord, is where commis-sural axons cross the midline, and constitutes an important intermediate choice point for axon guidance. The way commissural axons are initially attracted to the floorplate is analogous to the situation in Drosophila. Both in vitro and in vivo studies have shown that netrin-1, expressed by the floor-plate cells, attracts these axons to the midline (Kennedy et al., 1994; Serafini et al., 1994). The way in which these axons are compelled to leave the floorplate is also similar. Midline cells also express homologues of the Slit gene, namely Slit1, Slit2, and Slit3. After crossing, commissural axons gain sensitivity to these repellents, and exit the mid-line. Removal of all three Slit genes results in guidance errors indicative of axons no longer being repelled from the floorplate postcrossing (Long et al., 2004).
Segment boundary - Anterior commissure
Posterior commissure Segment boundary
Ventral midline t-*
Netrin Slit midline
Was this article helpful?