Neurons extend axons, which must grow along specific pathways during development. The discovery of particular patterns of expression of so-called Robo proteins provides clues to how axons find their way.
During development, neurons send out projections called axons, which must grow along an appropriate route — selecting from a multitude of possibilities — to reach their target. In organisms that are bilaterally symmetrical, many axons must also cross the central line (the midline) of the central nervous system, forming connections called commissures, which join the two sides of the body. Other axons, meanwhile, remain on one side. Four papers in Cell and Neuron1,2,3,4 now reveal the surprising result that, in the fruitfly Drosophila melanogaster, a protein called Slit, expressed at the midline, has a say both in controlling midline crossing and in directing an axon's subsequent choice of route. These activities require the previously known Roundabout (Robo) protein, found on the surface of neurons, and two newly discovered members of the same family.
In the central nervous system (CNS) of insects and vertebrates, neurons initially project axons towards specialized cells at the midline. The axons then turn into a longitudinal pathway, either before or after crossing the midline. The specialized midline cells secrete an axonal repellent — Slit — which binds to Robo5. Certain axons are attracted to midline cells and express Robo on their surface only after crossing the midline. This stops them from crossing back over. Meanwhile, axons that remain on one side of the CNS express Robo continuously. In other words, by signalling through Robo, Slit hangs up a 'do not enter' sign, stopping axons from coming into the midline.
In the absence of Robo, many axons freely cross and recross the midline6 (Fig. 1). But when Slit function is experimentally eliminated, all axons remain at the midline. This suggests that Slit may normally signal through other receptors in addition to Robo, providing a 'get out' signal. As now shown by Simpson et al.1 and Rajagopalan et al.2, one of the new Robo proteins, Robo2, is required for this signal. Robo2 is first expressed on all longitudinal axons in a pattern comparable to that of Robo. Eliminating the function of both Robo and Robo2 has the same effect as eliminating Slit — axons are unable to leave the midline (Fig. 1).
Having made its initial decision about crossing the midline, each axon turns into one of about 20 different longitudinal pathways, which are positioned at various distances either side of the midline. Axons in the various longitudinal pathways express certain surface markers that identify them. For example, the cell-adhesion molecule Fasciclin II is expressed in three major axon pathways either side of the midline.
The existence of such restricted expression patterns gave rise to the 'labelled pathways' hypothesis, which states that axons use these adhesive surface cues to decide which pathway to enter7. What has not been clear, however, is how the axons are directed the appropriate distance from the midline before they recognize the surface cues. Simpson et al.3 and Rajagopalan et al.4 now suggest that the two new Robo proteins — Robo2 and Robo3 — provide a combinatorial code that sets the sensitivity of an axon to the midline signal Slit, and hence the distance the axon travels from the midline (Fig. 2).
At the time that axons choose their longitudinal route, Robo2 is expressed only on those axons that extend within the most lateral third of the longitudinal pathways — that is, those pathways furthest from the midline. The expression of Robo3 likewise begins at this time, and is restricted to axons that extend within the lateral two-thirds of the longitudinal pathways. Robo, by contrast, is expressed by all axons in the longitudinal pathways.
The final, overlapping pattern of expression of Robo, Robo2 and Robo3 sets up a series of zones within the longitudinal pathways (Fig. 2). Each zone includes one of the Fasciclin-II-expressing pathways. The medial pathway — that nearest to the midline — is defined by the expression of Robo, the intermediate pathway by Robo and Robo3, and the lateral pathway by all three Robo proteins. Experimentally changing the combination of Robo molecules expressed by the axons drives them to extend within different longitudinal domains3,4. For example, eliminating Robo3 causes axons to extend into a more medial domain, resulting in a fusion of the intermediate Fasciclin-II-expressing pathway with the medial pathway. Similarly, the forced expression of Robo2 in neurons expressing only Robo causes their axons to extend more laterally.
In other words, the specific combination of Robo proteins expressed by an axon — its 'Robo code' — determines the approximate longitudinal region it reaches. The final choice of route is then mediated by local attractive interactions between other surface markers. This combination of long-range repulsion (in this case mediated by Slit, produced at the midline, and the Robo proteins, found on axons) and local cues in the target area has also been suggested as a mechanism for the topographical mapping of axons in the brain's tectum8,9.
This model suggests that, although Slit is expressed by midline cells, the protein must be distributed across the CNS. Although a gradient of Slit in the CNS has not yet been detected, it surely exists. Cells must be able to detect Slit at some distance from the midline, as the migration of muscle precursor cells at the lateral edge of the CNS is known also to be regulated by Slit5. The Robo molecules bind Slit, albeit with differing efficiencies, and switches in pathway choice based on changes in the Robo code produce the predicted discrete axonal movements towards or away from the midline, supporting the idea of a graded signal.
Do all the Robo proteins use the same intracellular signalling pathway to control axon extension in response to Slit? Robo2 and Robo3 lack two of Robo's four conserved intracellular domains, which bind to the intracellular signalling molecules Enabled and Abl10. So the newly identified Robo molecules are likely to transmit a qualitatively different intracellular signal, although what this is remains unknown. Moreover, Simpson et al.1 show that the Robo molecules can form homodimers and heterodimers, which is likely to affect the signalling output.
How is the graded Slit signal generated? How do the Robo molecules transmit their signals into the axon? And is this type of combinatorial code also used in vertebrates to direct and position neurons? The number of species in which Slit-to-Robo signalling has been found is increasing rapidly11,12,13, so it probably won't be long before we find the answers to these questions.