54: The Growth and Guidance of Axons

• Differences in the Molecular Properties of Axons and Dendrites Emerge Early in Development

  • Neuronal Polarity Is Established Through Rearrangements of the Cytoskeleton

  • Dendrites Are Patterned by Intrinsic and Extrinsic Factors

• The Growth Cone Is a Sensory Transducer and a Motor Structure

• Molecular Cues Guide Axons to Their Targets

• The Growth of Retinal Ganglion Axons Is Oriented in a Series of Discrete Steps

  • Growth Cones Diverge at the Optic Chiasm

  • Ephrins Provide Gradients of Inhibitory Signals in the Brain

• Axons from Some Spinal Neurons Cross the Midline

  • Netrins Direct Developing Commissural Axons Across the Midline

  • Chemoattractant and Chemorepellent Factors Pattern the Midline

• An Overall View

In the two preceding chapters we saw how neurons are generated in appropriate numbers, at correct times, and in the right places. These early developmental steps set the stage for later events that direct neurons to form functional connections with target cells. To form connections neurons have to extend long processes—axons and dendrites—which permit connectivity with postsynaptic cells and synaptic input from other neurons.

The growing axon may have to travel a long distance—up to several meters in a giraffe—and ignore many inappropriate neuronal partners before terminating in just the right region and recognizing its correct synaptic targets. In this chapter we examine how neurons elaborate axons and dendrites, and how axons are guided to their targets. In subsequent chapters we consider how neurons form synapses and how patterns of connections are shaped by activity.

We begin this chapter by discussing how certain neuronal processes become axons and others dendrites. We then consider the challenges that face an axon as it projects along tortuous pathways to its target. Finally, we illustrate general features of axonal guidance by describing the development of two well-studied axonal pathways: one that conveys visual information from the retina to the brain and another that conveys cutaneous sensory information from the spinal cord to the brain.

The processes of neurons vary enormously in their length, thickness, branching pattern, and molecular architecture. Nonetheless, most neuronal processes fit two functional categories: axons and dendrites. More than a century ago Santiago Ramón y Cajal hypothesized that this distinction underlies the ability of neurons to transmit information in a particular direction, an idea he formalized as the law of dynamic polarization. Cajal wrote that "the transmission of the nerve impulse is always from the dendritic branches and the cell body to the axon." In the decades before electrophysiological methods were up to the task, this law provided a means of analyzing neural circuits histologically. Although exceptions have been found, Ramón y Cajal's law remains a basic principle that relates structure and function in the nervous system and highlights the importance of knowing how neurons acquire their polarized form.

Neuronal Polarity Is Established Through Rearrangements of the Cytoskeleton

Much of our knowledge about neuronal polarization comes from studies of neurons taken from the rodent brain and grown in tissue culture. Hippocampal neurons grown in isolation develop processes reminiscent of those seen in vivo: a single, long, cylindrical axon and several shorter, tapered dendrites (Figure 54–1A). Axons and dendrites soon acquire molecular distinctions, as cytoskeletal and synaptic proteins are targeted to these components. For example, a particular form of the Tau protein is localized in axons and the MAP2 protein in dendrites (Figure 54–1B)

Cultured neurons are especially useful for developmental studies because they initially show no obvious sign of polarization and acquire their specialized features gradually in a stereotyped sequence of cellular steps. This sequence begins with extension of several short processes, each equivalent to the others. Soon thereafter, one process is established as an axon and the remaining processes acquire dendritic features (Figure 54–1A).

How does this occur? Cytoskeletal proteins that maintain elongated processes and drive growth are central to this process. If the actin filaments in an early neurite are destabilized, the ensuing cytoskeletal re-arrangements commit the neurite to becoming the axon, while the remaining neurites become dendrites. If the nascent axon is removed, one of the remaining neurites quickly assumes an axonal character. This sequence suggests that axonal specification is a key event in neuronal polarization and that signals from newly formed axons both suppress the generation of additional axons and promote dendrite formation.

The nature of the axonally derived signal that represses other axons is not known. But some insight into signals that control cytoskeletal arrangements has come from the study of a group of proteins encoded by the Par complex genes. As first shown in the nematode worm Caenorhabditis elegans, Par proteins are involved in diverse aspects of cytoskeletal reorganization, including the polarization of neuronal processes. Mammalian forebrain neurons lacking Par3, Par4, Par6, or relatives of Par1 grow multiple processes that are intermediate in length between axons and dendrites and bear markers of both processes (Figure 54–1B).

Although neurons grown in culture are similar to those in the brain, they are deprived of key extrinsic cues and signals. Cultured neurons become randomly arranged with respect to each other, whereas in many regions of the developing brain neurons line up in rows, with their dendrites pointing in the same direction (Figure 54–2A). This difference in vivo and in vitro implies that extrinsic signals regulate the polarization machinery. In the developing brain the local release of semaphorins, and other axonal guidance factors that we discuss later in the chapter, may help to orient dendrites (Figure 54–2C). The job of the Par protein complex is to link these extracellular signals to the cellular machinery that rearranges the cytoskeleton, a process achieved in part through the regulation of proteins that modify actin or tubulin function. In fact both the Tau protein in axons and the MAP2 protein in dendrites associate with and affect microtubules.

If local signals are needed to polarize neurons in the brain, how is polarity established in the uniform environment of a tissue culture? One possible explanation is that minor variations in the intensity of signaling within a neuron, or in signals from its immediate environment, will activate Par proteins in one small domain of the neuron, triggering the nearest process to become an axon. If, by happenstance, one process grows slightly faster than its neighbors, its chances of becoming an axon increase markedly (Figure 54–2B). Presumably, this proto-axonal process emits signals that decrease the chance of other processes following suit, forcing them to become dendrites.

Dendrites Are Patterned by Intrinsic and Extrinsic Factors

Although processes destined to become axons and dendrites initially are indistinguishable, they soon acquire distinct features. Nascent dendrites grow at acute angles and their branches are generally more numerous and closer to the cell body than those of axons. In addition, small protrusions called spines extend from the distal branches of dendrites. Finally, some dendritic branches are retracted or "pruned" to give the arbor its final and definitive shape (Figure 54–3).

Although the core features of dendrite formation are common to many neurons, there is a striking variation in dendrite number, shape, and branching pattern among neuronal types. Indeed, the shape and form of its dendritic arbor is one of the main ways in which neurons can be classified. Cerebellar Purkinje cells can be distinguished from granule cells, spinal motor neurons, and hippocampal pyramidal neurons simply by looking at the pattern of their dendrites.

How is dendritic pattern established? Neurons must have intrinsic information about their shape because the patterns in tissue culture are strikingly reminiscent of those in vivo (Figure 54–4). The transcriptional programs that specify neuronal subtype (see Chapter 52) presumably also encode information about neuronal shape. A second mechanism for establishing the pattern of dendritic arbors is the recognition of one dendrite by others of the same cell. In the Drosophila nervous system repellent signaling between dendrites of the same neuron, mediated by the different isoforms of a recognition protein known as DS-CAM, helps to ensure that dendrites expand over a broad territory rather than bunching together.

The dendrites of neighboring neurons also provide cues. In many cases the dendrites of a particular class of neuron cover a surface with minimal overlap, a spacing pattern called tiling. The tiling of dendrites appears to result from specific inhibitory interactions among the dendrites of a particular class of neuron. Tiling allows each class of neuron to receive information from the entire surface or area it innervates. Tiling of a region by the dendrites of one class of neuron also avoids the confusion that could arise if the dendrites of many different neurons occupied the same area.

Thus interactions between dendrites establish an overall arborization pattern through a mixture of intrinsic and extracellular mechanisms. Once a cell process is committed to an axonal or dendritic character, it becomes sensitive to a variety of intrinsic and extrinsic signals that determine its trajectory. For dendrites these patterning signals determine neuronal morphology. For axons, which we consider next, the signals guide the axons to their targets.

Once an axon forms it begins to grow toward its synaptic target. The key neuronal element responsible for axonal growth is a specialized structure at the tip of the axon called the growth cone. Both axons and dendrites use growth cones for elongation, but those linked to axons have been studied more intensively.

Ramón y Cajal discovered the growth cone and had the key insight that it was responsible for axonal pathfinding. With static images alone for inspiration (Figure 54–5A), he envisioned the growth cone to be "endowed with exquisite chemical sensitivity, rapid ameboid movements and a certain motive force, thanks to which it is able to proceed forward and overcome obstacles met in its way … until it reaches its destination."

Many studies over the past century have confirmed Ramón y Cajal's intuition. We now know that the growth cone is both a sensory structure that receives directional cues from the environment and a motor structure whose activity drives axon elongation. Ramón y Cajal also pondered "what mysterious forces precede the appearance of these processes … promote their growth and ramification … and finally establish those protoplasmic kisses … which seem to constitute the final ecstasy of an epic love story." In more modern and prosaic terms we now know that the growth cone guides the axon by transducing positive and negative cues into signals that regulate the cytoskeleton, thereby determining the course and rate of axonal outgrowth.

Growth cones have three main compartments. Their central core is rich in microtubules, mitochondria, and other organelles. Long slender extensions called filopodia project from the body of the growth cone. Between the filopodia lie lamellipodia, which are also motile and give the growth cone its characteristic ruffled appearance (Figure 54–5C).

Growth cones sense environmental signals through their filopodia: rod-like, actin-rich, membrane-limited structures that are highly motile. Their surface membranes bear receptors for the molecules that serve as directional cues for the axon. Their length—tens of micrometers in some cases—permits the filopodia to sample environments far in advance of the central core of the growth core. Their rapid movements permit them to compile a detailed inventory of the environment, and their flexibility permits them to navigate around cells and other obstacles.

When filopodia encounter signals in the environment, the growth cone is stimulated to advance, retract, or turn. Several motors power these orienting behaviors. One source of power is the movement of actin along myosin, an interaction similar to the one that powers the contraction of skeletal muscle fibers, although the actin and myosin of neurons are different from those in muscle. The assembly of actin monomers into polymeric filaments also contributes a propulsive force for filopodial extension. Acting in parallel, actin filaments constantly depolymerize at the base of filopodia. Depolymerization slows during periods of growth cone advance, leading to greater net forward motion. The movement of membranes along the substrate provides yet another source of forward motion.

The contribution of each type of molecular motor to the advance of the growth cone is likely to vary from one situation to another. With all of these motors the final step involves the flow of microtubules from the central core of the growth cone into the newly extended tip, thus moving the growth cone ahead and leaving in its wake a new segment of axon. New lamellipodia and filopodia form in the advancing growth cone and the cycle repeats (Figure 54–6).

Accurate pathfinding can occur only if the growth cone's motor action is linked to its sensory function. It is crucial therefore that the recognition proteins on the filopodia are signal-inducing receptors and not merely binding moieties that mediate adhesion. The binding of a ligand to its receptor affects growth in diverse ways. It stimulates the formation, accumulation, and even the breakdown of soluble intracellular molecules that function as second messengers. These second messengers affect the organization of the cytoskeleton, and in this way regulate the direction and rate of movement of the growth cone.

One important second messenger is calcium. The calcium concentration in growth cones is regulated by the activation of receptors on filopodia and this affects the organization of the cytoskeleton, which in turn modulates motility. Growth cone motility is optimal within a narrow range of calcium concentrations, called a set point. Activation of filopodia on one side of the growth cone leads to a concentration gradient of calcium across the growth cone, providing a possible basis for changes in the direction of growth.

Other second messengers that link receptors and motor molecules include cyclic nucleotides. These nucleotide second messengers modulate the activity of enzymes such as protein kinases, protein phosphatases, and rho-family guanosine triphosphatases (GTPases). In turn these messengers and enzymes regulate the activity of proteins that regulate the polymerization and depolymerization of actin filaments, thereby promoting or inhibiting axonal extension.

The critical role of intracellular signals in growth cone motility and orientation can be demonstrated using embryonic neurons grown in culture. Application of growth factors to one side of a growth cone activates receptors locally and leads to extension and turning of the growth cone toward the source of the signal. In essence, the factor attracts the growth cone. Yet when cyclic adenosine monophosphate (cAMP), levels in the neuron are decreased, the same stimulus acts as a repellent and the growth cone turns away from the signal (Figure 54–7). Other repulsive factors can become attractive when levels of the second messenger cyclic guanosine 3′-5′-monophosphate (cGMP) are raised.

Another strategy for coupling the sensory and motor functions of the growth cone is to engage the cytoskeleton directly, through the intracellular domain of receptors. Integrin receptors couple to actin in growth cones when they bind molecules associated with the surface of adjoining cells or the extracellular matrix, thereby influencing motility. Recently, it has been found that growth cones contain messenger RNAs as well as the machinery for protein synthesis. The activation of receptors can lead to local synthesis of new motor proteins precisely when and where they are needed. Thus the growth cone has many strategies and mechanisms for integrating molecular signals to direct the axon in specific directions.

For much of the 20th century a debate raged between advocates of two very different views of how growth cones navigate embryonic terrains to reach their targets. A molecular view of axonal guidance was first articulated at the turn of the 20th century by the physiologist J. N. Langley. But by the 1930s many eminent biologists, including Paul Weiss, believed that axonal outgrowth was essentially random and that appropriate connections persisted largely because of productive, matching patterns of electrical activity in the axon and its target cell.

In our molecular age Weiss's ideas may seem simplistic but they were not unreasonable at the time. In tissue culture axons grow preferentially along mechanical discontinuities (scratches and bumps on a cover slip), and embryonic nerve trunks often align themselves with solid supports (blood vessels or cartilage). It seemed logical to Weiss that mechanical guidance, called stereotropism, could account for axonal patterning. We are quite comfortable today with the idea that electrical signals can be used to change the way current flows in a computer without the need to resolder connections. Likewise, patterns of activity and experience can strengthen or weaken neural connections without requiring the formation of new axonal pathways. Then why not consider that congruent activity, called resonance by Weiss, is involved in establishing appropriate connections?

Today few scientists believe that stereotaxis or resonance are crucial forces in neuronal connectivity. The tipping point that shifted opinion in favor of the molecular view was an experiment performed with frogs and other amphibia in the 1940s by Roger Sperry (ironically, a student of Weiss). Sperry manipulated the information carried from the eye to the brain by the axons of retinal ganglion cells. These axons terminate in their target areas—the lateral geniculate body in the thalamus and the superior colliculus (or optic tectum) in the midbrain—in such a way that an orderly retinotropic map of the visual field is created.

Because of the optics of the eye, the visual image on the retina is an inversion of the visual field. The retinal ganglion cells reinvert the image by the manner in which their axons terminate in the optic tectum, the main visual center in the brain of frogs (Figure 54–8A). If the optic nerve is cut the animal is blinded. In lower vertebrates cut retinal axons can reestablish projections to the tectum, whereupon vision is restored. This is not the case in mammals, as we will discuss in Chapter 57.

Sperry performed a simple yet profound experiment that demonstrated the organization of the retinotectal projection. He severed the optic nerve in a frog and then rotated the eye in its socket by 180 degrees before regeneration of the nerve. Remarkably, the frog exhibited orderly responses to visual input, but the behavior was wrong. When the frog was presented with a fly on the ground it jumped up, and when offered a fly above its head it struck downward (Figure 54–8B). Importantly, the animal never learned to correct its mistakes. Sperry suggested—and later verified with anatomical and physiological methods—that the retinal axons had reinnervated their original tectal targets, even though these connections provided the brain with erroneous spatial information that led to aberrant behavior. The inference of these experiments was that recognition between axons and their targets relied on molecular matching rather than functional validation and refinement of random connections.

But Weiss's ideas are by no means obsolete. Indeed, we now recognize that the activity of neural circuits can play a crucial role in shaping connectivity. The current view is that molecular matching predominates during embryonic development, and that activity and experience modify circuits after they have formed. In this chapter and the next we describe the molecular cues that guide the formation of neural connections, and then in Chapter 56 we examine the role of activity and experience in the fine-tuning of synaptic connections.

Sperry's conjecture, often called the chemospecificity hypothesis, prompted developmental neurobiologists to search for axonal and synaptic "recognition molecules." Success was limited for the first few decades, in part because these molecules are present in small amounts and on discrete subsets of neurons and there were no effective methods for isolating rare molecules from complex tissues. As with the search for neural inducers, advances in biochemical and molecular biological methods gradually made this task more feasible, and many proteins involved in the guidance of axons to their targets have now been discovered. These proteins typically consist of pairs of ligands and receptors: The ligands are presented by cells along the pathway an axon follows and the receptors by the growth cone itself.

In the most general terms, guidance cues can be presented on cell surfaces, in the extracellular matrix, or in soluble form. They act by either promoting or inhibiting outgrowth of the axon. Most receptors are embedded in the growth cone membrane; they have an extracellular domain that selectively binds the cognate ligand and an intracellular domain that couples to the cytoskeleton, either directly or through intermediates such as second messengers. Activation of these receptors can promote or inhibit neurite outgrowth. A ligand presented to one side of the growth cone can result in turning. In this way the local distribution of environmental cues can steer the growth cone.

As a result of these recent discoveries, axon guidance—a process that appeared mysterious years ago—can now be viewed as the orderly consequence of protein-protein interactions that instruct the growth cone to grow, turn, and stop (Figures 54–9 and 54–10). This limited set of instructions is sufficient, when presented with spatial precision, to choreograph growth cone behaviors with exquisite subtlety. Axonal guidance can therefore be explained by describing how and where ligands are presented and how the growth cone integrates this information to generate an orderly response. In the rest of the chapter we illustrate lessons learned by describing the journeys of two types of axons: the axons of retinal ganglion neurons and those of a particular class of sensory relay neurons in the spinal cord.

Sperry's experiment implied the existence of axon guidance cues but did not reveal where they were or how they worked. For a time, one prominent view was that recognition occurred mostly at or near the target and that mechanical forces or long-range chemotactic factors sufficed to get axons to the vicinity of the target.

We now know that axons reach distant targets in a series of discrete steps, making frequent decisions at closely spaced intervals along their route. To illustrate this point we shall trace in greater detail the path that Sperry was trying to understand, that of a retinal axon growing to the optic tectum.

Growth Cones Diverge at the Optic Chiasm

The first task of the axon of a retinal ganglion cell is to leave the retina. As it enters the optic fiber layer it extends along the basal lamina and glial end-feet positioned at the retina's edge. The growth of the axon is oriented from the outset, indicating that it can read directional cues in the environment. As it approaches the center of the retina it comes under the influence of attractants emanating from the optic nerve head (the junction of the optic nerve with the retina proper), which guide it into the optic stalk. It then follows the optic nerve toward the brain (Figure 54–11).

The first axons to travel this route follow the cells of the optic stalk, the rudiment of the neural tube that connects the retina to the diencephalon from which it arose. These "pioneer" axons then serve as scaffolds for later-arriving axons, which are able to extend accurately simply by following their predecessors (see strategy 3 in Figure 54–9). Once they reach the optic chiasm, however, the retinal axons must make a choice. Axons that arise from neurons in the nasal hemiretina of each eye cross the chiasm and proceed to the opposite side of the brain, whereas those from the temporal half are deflected as they reach the chiasm and so stay on the same side of the brain (Figure 54–12A).

This divergence in trajectory reflects the differential responses of axons from the nasal and temporal hemiretinas to guidance cues presented by midline chiasm cells. Some retinal axons contact and traverse chiasm cells, whereas others are inhibited by these cells and deflected away, thus remaining in the ipsilateral side. One of the key molecules presented by chiasm cells is a membrane-bound repellent of the ephrin-B family (Figure 54–12B), which also figures in later steps of retinal ganglion cell axon guidance.

The fraction of temporal retinal axons that project ipsilaterally varies among species: a few in lower vertebrates, some in rodents, and many in humans. These differences reflect placement of the eyes. In many animals the eyes point to the sides and monitor different parts of the visual world, so that information from the two eyes need not be combined. In humans both eyes look forward and sample largely overlapping regions of the visual world, so coordination of visual input is essential.

After crossing the optic chiasm, retinal axons assemble in the optic tract along the ventral surface of the diencephalon. Axons then leave the tract at different points. In most vertebrate species the tectum of the midbrain (called the superior colliculus in mammals) is the major target of retinal axons, but a small number of axons project to the lateral geniculate nucleus of the thalamus. In humans, however, most axons project to the lateral geniculate, a sizable number reach the colliculus, and small numbers project to the pulvinar, superchiasmatic nucleus, and pretectal nuclei. Within these targets different retinal axons project to different regions. As Sperry showed, the retinal axons form a precise retinotopic map on the tectal surface. Similar maps form in other areas innervated by retinal axons.

Having reached an appropriate position within the tectum, retinal axons need to find an appropriate synaptic partner. To achieve this last leg of their journey, retinal axons turn and dive into the tectal neuropil (Figure 54–11), descending along the surface of radial glial cells, which provide a scaffold for radial axonal growth. Although radial glial cells span the entire extent of the neuroepithelium, each retinal axon confines its synaptic terminals to a single layer. The dendrites of many postsynaptic cells extend through multiple layers and form synapses along their whole length, but retinal inputs are restricted to a small fraction of the target neuron's dendritic tree. These organizational features imply that layer-specific cues arrest axonal elongation and trigger arborization.

Axons therefore solve the problem of long-distance navigation by dividing their journey into short segments, and by recognizing and responding to intermediate targets along the path to their final targets. Some intermediate targets, such as the optic chiasm, are "decision" regions where axons need to diverge.

Reliance on intermediate targets is an effective solution to the problem of long-distance axonal navigation but is not the only one. In some cases the first axons reach their targets when the embryo is small and the distance to be covered is short. These "pioneer" axons respond to molecular cues embedded in cells or the extracellular matrix along their way. The first axons to exit the retina fall within this class. Axons that appear later, when distances are longer and obstacles more numerous, can reach their targets by following the pioneers. Yet another guidance mechanism is a molecular gradient. Indeed, as we will see, gradients of cell-surface molecules in the tectum inform axons about their proper termination zone.

Ephrins Provide Gradients of Inhibitory Signals in the Brain

So far we have seen how retinal axons reach the tectum by responding to a series of discrete directional cues. However, these choices during growth do not account for the smoothly graded connections implied by Sperry's analysis of the retinotopic map in the tectum. The quest for the hypothetical "graded map molecules" became a major focus for developmental neurobiologists, and so we describe it in some detail.

A key breakthrough in the quest for these molecules came with the development of bioassays in which explants from defined portions of the retina were laid on substrates of tectal membrane fragments. The membrane fragments were taken from defined anteroposterior portions of the tectum and arranged in alternating stripes. Axons from the temporal (posterior) hemiretina were found to grow preferentially on membranes from anterior tectum, a preference similar to that exhibited in vivo (Figure 54–13). This preference was found to result from the presence of inhibitory factors in posterior membranes rather than from attractive or adhesive substances in anterior membranes. This observation was one of the first to emphasize the role of inhibitory or repellent substances in axon guidance.

This stripe assay permitted the characterization of an inhibitory cue, present in membranes from the posterior but not the anterior tectum. Independently, molecular biologists identified a family of receptor tyrosine kinases, the Eph kinases, and a large family of membrane-associated ligands, the ephrins. Both receptors and ligands are divided into A and B subfamilies. The ephrin-A proteins bind and activate EphA kinases; conversely, ephrin-B proteins bind and activate EphB kinases (Figure 54–14).

The two lines of research converged when the tectal inhibitory cue was identified as ephrin-A5. We now know that the Eph kinases and ephrins serve many functions in neural and nonneural tissues and that each class of proteins can serve as ligands or receptors, depending on cellular context. In the developing nervous system these proteins comprise a major group of repellent signals.

Ephrin-Eph interactions account for formation of the retinotopic map in the tectum. In the tectum ephrin-A2 and ephrin-A5 levels are graded along the anteroposterior axis, and in the retina the levels of the Eph receptors are also graded along the anteroposterior axis. These gradients run in opposite directions in the tectum: ephrin-A grades from anterior-low to posterior-high and in the retina Eph-A kinase from posterior-high to anterior-low (Figure 54–14A). Such counter-gradients account, at least in part, for topographic mapping. Axons from posterior retinal ganglion cells with high levels of receptors are repelled most strongly by the high level of ephrin-A in the posterior tectum and thus are confined to the anterior tectum. The less sensitive axons from the anterior retina are able to penetrate further into the posterior domain of the tectum. Ephrin-A2 and -A5 are therefore strong candidates for chemospecificity factors of the type postulated by Sperry.

The crucial role of ephrins and Eph kinases in the formation of retinotopic maps has been confirmed in vivo. Overexpression of ephrin-A2 in the developing optic tectum of chick embryos generates small patches of cells in the rostral tectum that are abnormally rich in ephrin-A2. Temporal retinal axons, which normally avoid the ephrin-rich caudal tectum, also avoid these patches in the rostral tectum, and they terminate in abnormal positions. In contrast, nasal retinal axons, which normally grow toward the caudal tectum, are not perturbed by encounters with excess ephrin.

Conversely, in mice with targeted mutations in the ephrin-A2 and ephrin-A5 genes some posterior retinal axons terminate in inappropriately posterior tectal regions (Figure 54–14B). Anterior retinal axons, which naturally express low levels of EphA proteins, project normally in these mutants. In mice lacking both ephrin-A proteins these deficits are more severe than with either single mutant. Thus the interaction of ephrin-A with EphA receptors is crucial for the targeting of retinal axons in the tectum. These ephrin/EphA pairs possess the properties of the recognition molecules that Sperry predicted were necessary to direct topographic mapping along the anteroposterior axis of the tectum.

But the retinal map is two-dimensional, so what establishes order along the dorsoventral axis? The ephrin/EphB pairs are involved in establishing this axis. Just as ephrin-A and EphA are graded along the anteroposterior axis, ephrin-B and EphB are graded along the dorsoventral axis, and genetic manipulation of ephrin-B and EphB levels affects dorsoventral mapping (Figure 54–14C). Thus the retinotopic map is arranged along rectangular coordinates with ephrin/EphA and ephrin/EphB labeling the anteroposterior and dorsoventral axes, respectively.

This simple view is satisfying, but the reality is more complex. First, Eph kinases are expressed in the tectum as well as in the retina, and ephrins are expressed in the retina as well as in the tectum. Thus, so-called "cis" interactions (Eph and ephrin on the same cell) as well as "trans" interactions (Eph on growth cone, ephrin on target cell) may be involved. Second, both ligands and receptors are present at multiple points along the optic pathway and play multiple roles. As we have seen, ephrin/EphB interactions affect not only dorsoventral mapping but also the decision of an axon to cross to the contralateral side at the optic chiasm. Finally, in visual circuits precise axonal mapping is regulated by patterns of neural activity, as discussed in the next two chapters. Nonetheless, we now have the initial outline of a molecular strategy for the formation of specific topographic projections from the eye to the brain.

One of the fundamental features of the central nervous system is the need to coordinate activity on both sides of the body. To accomplish this task certain axons need to project to the opposite side.

We have seen one example of axonal crossing in the optic chiasm. Another example that has been studied in detail is the axonal crossing of commissural neurons that convey sensory information from the spinal cord to the brain at the ventral midline of the spinal cord across the floor plate. After crossing, axons turn abruptly and grow up toward the brain. This simple trajectory raises several questions. How do these axons reach the ventral midline? On arrival how do they cross the midline? After crossing how do they ignore cues that axons on the other side are using to get to the midline? In other words, why do they turn toward the brain instead of crossing back?

Netrins Direct Developing Commissural Axons Across the Midline

Many of the neurons that send axons across the ventral midline are generated in the dorsal half of the spinal cord. The first task for these axons is to reach the ventral midline. Ramón y Cajal considered the possibility that chemotactic factors emitted by targets could attract axons, but this idea lay dormant for nearly a century. We now know that such chemotropic factors do exist and one of them, the protein netrin-1, mediates the chemoattractant activity of the floor plate. When presented in culture, netrin attracts commissural axons; when mice are deprived of netrin-1 function, axons fail to reach the floor plate (Figure 54–15).

The netrin protein is structurally related to the protein product of unc-6, a gene shown to regulate axon guidance in the nematode C. elegans. Two other C. elegans genes, unc-5 and unc-40, encode receptors for the unc-6 protein. Vertebrate netrin receptors are related to the unc-5 and unc-40 receptors: The unc-5H proteins are homologs of unc-5, and DCC (deleted in colorectal cancer) and neurogenin are related to unc-40 (see Figure 54–10G). These receptors are members of the immunoglobulin superfamily, and their functions have been remarkably conserved throughout animal evolution (Figure 54–16). This conservation supports the use of simple and genetically accessible invertebrates to unravel developmental complexities. In no area has this approach been more fruitful than in the analysis of axon guidance. Dozens of genes that affect this process were first identified and cloned in Drosophila and C. elegans and then shown to play important and related roles in mammals.

Other signaling systems work with netrins to guide commissural axons on this initial phase of their journey to the brain. For example, bone morphogenetic proteins secreted by the roof plate act as repellants and begin to direct commissural axons ventrally.

Chemoattractant and Chemorepellent Factors Pattern the Midline

Once commissural axons reach the midline, they find themselves exposed to the highest available level of netrin-1. Yet this netrin-rich environment does not keep the axons at the midline indefinitely. Instead they cross to the other side of the spinal cord, while at the same time their contralateral counterparts are navigating up the netrin chemoattractant gradient.

This puzzling behavior is explained by the fact that growth cones change their responsiveness to attractive and repellent signals as a consequence of exposure to floor plate signals. This switch illustrates an important property of intermediate targets involved in axon guidance. Factors presented by intermediate targets not only influence the growth of axons but also change the sensitivity of the growth cone, preparing it for the next leg of its journey.

Once axons arrive at the floor plate, they become sensitive to the chemorepellent signal slit, which is secreted by floor plate cells (Figure 54–17). Before commissural axons reach the floor plate the robo proteins that serve as slit receptors are kept inactive by expression of a related protein, rig-1. As axons reach the floor plate rig-1 is lost, unleashing robo activity and causing axons to respond to the repellant actions of slit. This repellant action propels growth cones down the slit gradient into the contralateral side of the spinal cord. In addition, activated robo forms a complex with DCC, which renders these netrin receptors incapable of responding to their ligand. The decreased sensitivity of growth cones to the attractive properties of the floor plate helps to account for the transient nature of the floor plate's influence on axons.

Finally, once axons have left the floor plate they turn rostrally toward their eventual synaptic targets in the brain. A rostrocaudal gradient of Wnt proteins expressed by floor plate cells appears to direct axon growth rostrally at the ventral midline (Figure 54–17D). Thus different cues guide commissural axons during distinct phases of their overall trajectory. This same process is presumably played out for hundreds and even thousands of classes of neurons to establish the mature pattern of brain wiring.

Many developmental processes shape precise patterns of connections in the nervous system, but none is more important than the guidance of axons from their origin to appropriate targets. The specificity required to achieve the correct wiring is extraordinary. Axons need to grow long distances, bypassing numerous targets along the way, before reaching and forming connections with appropriate synaptic partners.

The nervous system has devised elaborate mechanisms to achieve its basic wiring plan. At the cellular level one basic strategy is to break the long journey into manageable legs. Thus axons do not set off like explorers of old, crossing completely uncharted territory in search of distant goals. Instead, they receive guidance at intervals along the way. Some axons that grow early in the embryo, when distances are short, serve as scaffolds for later-growing axons. Other axons grow along epithelial surfaces or extracellular matrices. Often so-called guidepost cells or intermediate targets mark sites at which axons need to make divergent choices.

The axon receives and responds to these guidance clues with a specialized terminal apparatus, the growth cone. The growth cone is both a sensory and a motor structure. It bears numerous receptors that bind environmental cues as well as cytoskeletal proteins and actin-based motors that propel it forward. And it contains signal transduction systems that convert ligand-receptor binding into instructions that steer the growth cone.

In the past several years rapid progress has been made in the identification of molecules that guide axons. Guidance cues include soluble, membrane-bound, and extracellular matrix molecules. Many of the soluble molecules are members of gene families, such as the ephrins, netrins, slits, semaphorins, and laminins. Families of membrane-bound receptors on the growth cones include adhesion molecules, Eph kinases, neuropilins, plexins, robos, and integrins. Mutations in genes that encode these ligands and receptors can result in developmental neurological disorders.