Axon guidance

A Roundabout way of avoiding the midline

Article metrics

Bilaterally symmetrical animals must be able to integrate sensory inputs and coordinate motor control on both sides of the body. Thus, many neurons in the central nervous system (CNS) project their axons to the opposite side of the body, whereas others project axons that remain on the same side. In the latest issues of Cell1,2 and Neuron3, the groups of Corey Goodman, Guy Tear, Marc Tessier-Lavigne and Cori Bargmann report that, from worms and flies to rats and humans, a common mechanism determines which axons cross the midline and which do not.

In insects and vertebrates, the two symmetrical halves of the CNS are separated by a specialized group of cells. Located at the ventral midline of the CNS, these are called midline cells in insects, and they form the floor plate in vertebrates. Most CNS axons initially grow towards these midline cells and then turn longitudinally — either on their own side (ipsilateral), or by first crossing the midline and then turning (contralateral). As they project longitudinally, these axons skirt along the edge of the midline, but never cross it again. Those axons that cross the midline form the commissures that allow sensory information and motor instructions to pass from one side of the animal to the other.

A few years ago, there was much excitement when midline cells in worms, flies and vertebrates were all found to secrete the same type of chemoattractants, netrins, to guide axons along the initial part of their trajectories towards the midline4,8. But evidence has been accumulating from studies in the fly9, grasshopper10, fish11 and chick12 suggesting that the midline is not only the source of a long-range attractive signal, but also of a short-range repulsive signal. Kidd et al.1 and Zallen et al.2 now identify a family of putative receptors for such a repulsive signal — the Roundabout (Robo) family. So far, one robo gene has been identified in worms, two in flies, two in the rat and two in humans, suggesting that the short-range repulsive signal has been just as highly conserved during the course of evolution as the long-range attractive signal.

Mutations in the robo gene were originally recovered in a screen for abnormal axon-projection patterns in the Drosophila CNS9. In robo mutant embryos, axons meander back and forth across the midline. Mutations in sax3 — a closely related Caenorhabditis elegans gene — result in a similar phenotype2. The worm ventral nerve cord is asymmetrical but, nevertheless, it is divided into discrete left and right axon bundles. Axons mainly cross from left to right only at the anterior and posterior ends of the nerve cord. In sax3 mutants, however, axons cross the midline in either direction along the length of the nerve cord. Kidd et al. and Zallen et al. have also found that the robo and sax3 genes encode immunoglobulin-type transmembrane receptors, which are expressed on neuronal growth cones as they encounter the midline1,2. One of the rat robo genes is also expressed by spinal-cord neurons when their axons are responding to guidance cues from the floor plate. So rat robo, like the invertebrate robo genes, may be reading a signal that prevents these axons from repeatedly crossing the midline1.

A combination of a long-range attractant and short-range repellent expressed by the same midline cells would elegantly explain why most axons first project towards the midline, and, on reaching it, turn to follow a parallel pathway, never straying across the midline or back towards the periphery. But why do most axons cross the midline before making this turn? And why do they cross only once? Conceivably, crossing and non-crossing axons might differ in their responses to the attractive or the repulsive cue.

It is not yet clear whether netrins are also involved in the crossing decision, because it is experimentally difficult to separate guidance towards and across the midline. For example, the partial loss of commissures in netrin mutant fly embryos7,8 may occur because many commissural axons fail even to make contact with midline cells. Still, netrins are not the only attractive signals provided by the midline and floor plate, and commissural and longitudinal axons may differ in their sensitivity to some of these other cues.

Differential sensitivity to the midline repellent, at least in flies, is dramatically shown by Kidd et al.1. The Robo protein is expressed on the growth cones of the longitudinal axons, yet it is almost completely absent from commissural axons — until they have crossed the midline. This explains why commissural (but not longitudinal) axons can cross the midline, and also why, having crossed once, commissural axons never turn back and cross again.

In a companion paper, Kidd et al.3 show that Robo is downregulated by a transmembrane protein encoded by the commissureless (comm) gene. As the name implies, axons cannot cross the midline in comm mutants9. The Comm protein is expressed by a group of midline glial cells, positioned exactly where the commissures are pioneered13. The protein is transferred to the commissural axons that contact these cells, inhibiting Robo and facilitating passage of these axons across the midline3,13 (Fig. 1). Because both commissural and longitudinal axons contact these cells, they must differ in their ability to respond to Comm. Kidd et al.3 suggest that commissural axons may initially express lower levels of Robo, so they will be more sensitive to downregulation by Comm. Or, alternatively, perhaps only commissural axons express a Comm receptor.

Figure 1: Model for regulation of midline crossing by axons in the central nervous system of Drosophila embryos.

The midline is the source of a diffusible attractant (netrins; green) and, as described by Kidd et al.1 and Zallen et al.2, a membrane-bound repellent (the Roundabout or Robo ligand; brown). Axons projecting ipsilaterally (on the side from which they originated) express high levels of Robo (blue), whereas contralaterally projecting axons (on the opposite side) express high levels of Robo only after crossing the midline. The commissureless (Comm) protein (yellow) is transferred from midline glia to commissural axons13, where Kidd et al.3 have found that it downregulates Robo and thus allows these axons to cross the midline.

The high structural and functional conservation in the netrin and robo families reveals a common underlying logic in the control of axonal traffic by specialized midline structures in the nervous systems of worms, flies, mice and humans — a long-range attractant guides axons to the midline, and a short-range repellent prevents them from crossing it. Commissural axons are granted temporary immunity against this repellent, allowing them just a single pass across the midline.

These studies1,3 raise many questions. For example, what is the nature of the Robo ligand, the putative midline repellent? How does Comm downregulate Robo, and how is Robo upregulated after the midline crossing? It will also be interesting to find out whether other species use Comm homologues to allow transit across a repulsive midline. Although no vertebrate homologue has yet been reported, a similar mechanism probably conducts commissural axons across the vertebrate floor plate. But as sequencing of the C. elegans genome nears completion, it seems that worms lack a comm gene, just as their nerve cord lacks commissures.

Evolution seems to have enjoyed a relatively free rein in designing the CNS, coming up with different cellular architectures and developmental strategies in nematodes, arthropods and chordates. One common feature is the organization of axon projections into an orthogonal grid of circumferential and longitudinal nerve fibres. Highly conserved guidance mechanisms employing members of the netrin and robo families can account for much of this pattern.


  1. 1

    Kidd, al. Cell 92, 205–215 (1998).

  2. 2

    Zallen, J. A., Yi, B. A. & Bargmann, C. Cell 92, 217–227 (1998).

  3. 3

    Kidd, T., Russell, C., Goodman, C. S. & Tear, G. Neuron 20, 25–33 (1998).

  4. 4

    Ishii, N., Wadsworth, W. G., Stern, B. D., Culotti, J. G. & Hedgecock, E. M. Neuron 9, 873–881 (1992).

  5. 5

    Serafini, al. Cell 78, 409–424 (1994).

  6. 6

    Kennedy, T. E., Serafini, T., de la Torre, J. R. & Tessier-Lavigne, M. Cell 78, 425–435 (1994).

  7. 7

    Mitchell, K. al. Neuron 17, 203–215 (1996).

  8. 8

    Harris, R., Sabatelli, L. M. & Seeger, M. A. Neuron 17, 217–228 (1996).

  9. 9

    Seeger, M., Tear, G., Ferres-Marco, D. & Goodman, C. S. Neuron 10, 409–426 (1993).

  10. 10

    Myers, P. Z. & Bastiani, M. J. J. Neurosci. 13, 127–143 (1993).

  11. 11

    Bernhardt, R. R., Nguyen, N. & Kuwada, J. Y. Neuron 8, 869–882 (1992).

  12. 12

    Stoeckli, E., Sonderegger, P., Pollerberg, G. E. & Landmesser, L. T. Neuron 18, 209–221 (1997).

  13. 13

    Tear, al. Neuron 16, 501–514 (1996).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.