The Eph Ligand Family and Neuronal Topographic Mapping

Hwai-Jung Cheng

One of the central questions in neurobiology is how the neurons in the nervous system are connected. To process and respond to information from the outside world properly, the connections from the peripheral nervous system to higher hierarchies in the central nervous system have to be arranged such that the order of the presynaptic neurons is faithfully reflected by the spatial order of their connections with the postsynaptic neurons, with nearest-neighbor relationships preserved. This type of spatial relationship is known as a topographic map. Topographic maps have been demonstrated in sensory inputs, motor outputs, and various neuronal connections within the brain (1, 2). With regard to the development of topographic order, the visual projection from the retina to the tectum is the best studied. The widely accepted hypothesis to explain the formation of the retinotectal topographic map is the chemoaffinity theory, first proposed more than 50 years ago by R. W. Sperry. This theory postulates the existence of graded chemical labels across both projecting (retina) and target (tectum) fields. These labels could give each point on the retina and tectum a unique positional identity and would allow each retinal axon to recognize its proper tectal termination site. The postulated chemical labels are often referred to as Sperry molecules, but the identification of these molecules has been elusive over the last several decades (3,4).

Receptor tyrosine kinases (RTKs) and their cognate ligands are important molecules implicated in the process of cellular interactions during embryonic development (5). The RTKs can be divided into families based on structural homology, and the Eph family is by far the largest. Expression pattern studies of Eph receptors have suggested roles in the development of the nervous system. However, all members of this family were identified as orphan receptors without known ligands. In order to investigate the biological function of the Eph family, I set out to clone the ligand(s) for two closely related Eph receptors, Mek4 (6) and Sek (7).

To identify the ligand(s) for Mek4 and Sek receptors, I tested a new approach to determine the spatial distribution of li-gands in embryos (8). We term this procedure "AP in situ", for alkaline phosphatase or affinity probe in situ. The receptor extracellular domain was fused to a human placental alkaline phosphatase (AP) tag to make a soluble receptor fusion protein. The Mek4-AP and Sek-AP fusion proteins detected high ligand binding activity in the midbrain (also known as tectum in lower vertebrates) of embryos. Based on this spatial information, a cDNA expression library was prepared and ELF-1 (Eph ligand family–1) was cloned from this library by screening with the Mek4-AP and Sek-AP probes. The expression of ELF-1 in the embryonic tectum is surprisingly strong, suggesting its role in the development of the retinotectal topographic map.

AP in situ provides a rapid, efficient, and generally applicable way to identify and clone new ligands. Moreover, the same approach can be generalized from ligands to receptors or other interaction partners. For example, a receptor for leptin, the murine obesity gene product, was recently cloned by this approach (9). In addition to identifying and cloning new molecules, AP in situ is a general method for testing the distribution of binding activity of molecules and can provide important biological information on the distribution of cognate binding sites—information not available from existing methods such as RNA hybridization or immunohistochemistry. The utility of this approach is illustrated by our experiments on the development of the retinotectal topographic map.

The strong expression of ELF-1 in the midbrain led us to speculate that ELF-1 and its receptor might be Sperry molecules acting in retinotectal development. We therefore characterized RNA expression patterns of ELF-1 and Mek4 in the retinotectal system of the chicken, a species with a prominent and well-characterized retinotectal projection (10). We showed that the expression of ELF-1 is in a gradient in the tectum, and its receptor Mek4 is in a matching gradient in the retina. Moreover, both ELF-1 and Mek4 are expressed within the time-window when the development of the retinotectal map occurs. To directly test the binding activity of Mek4 and ELF-1, AP in situ with Mek4-AP and ELF1-AP as probes were used. In Sperry's chemoaffinity theory, complementary matching gradients are essential. That is, the binding sites of a molecule with graded distribution on one topographic field are supposed to be in a matching gradient on the reciprocal field. Since AP fusion protein can act as an affinity probe to directly assay the binding activity of the fused molecule, AP in situ is a useful tool to test for molecular complementarity of positional gradient molecules in topographic projection maps. We showed that the binding activities of Mek4-AP on the tectum and ELF1-AP on the retinal axon fibers are in matching gradients, providing direct evidence for molecular complementarity of gradients in reciprocal fields of the retinotectal system.

To directly test for the effect of ELF-1 on the development of visual topographic map in vivo, we retrovirally overexpressed ELF-1 in the tectum. Our data have shown that ELF-1 can discriminate between axons from different positions with repellent effects on temporal but not nasal axons. These results show that ELF-1 has topographically specific effects on retinal axon mapping (11). Since ELF-1 was cloned, more than six Eph ligands have been identified (12). RAGS, another Eph ligand, is also expressed in a tectal gradient and has retinal axon repellent activity in vitro (13). These experiments strongly suggest a role for Eph family ligands and receptors in the establishment of visual topographic map. In addition, several members of the Eph ligands and receptors are known to show localized expression in other regions of the nervous system (12). It seems plausible that they might function in several developmental fields and may act as molecules responsible for developmental topographic map formation of the nervous system in general.

Hwai-Jung Cheng is at 2F-3, No. 42, Sec 5, Min-Chuan East Road, Taipei 105, Taiwan. E-mail: hjcheng{at}


  1. S. B. Udin and J. W. Fawcett, Annu. Rev. Neurosci.11, 289 (1988).
  2. C. S. Goodman and C. J. Shatz, Cell 72 (Suppl.), 77 (1993).
  3. C. E. Holt and W. A. Harris, J. Neurobiol.24, 1400 (1993).
  4. P. A. Garrity and S. L. Zipursky, Cell83, 177 (1995).
  5. P. van der Geer, T. Hunter, R. A. Lindberg, Annu. Rev. Cell Biol.10, 251 (1994).
  6. F. G. Sajjadi, E. B. Pasquale, S. Subramani, New Biol.3, 769 (1991).
  7. P. Gilardi-Hebenstreit et al., Oncogene7, 2499 (1992).
  8. H. J. Cheng and J. G. Flanagan, Cell79, 157 (1994).
  9. L. A. Tartaglia et al., ibid. 83, 1263 (1995).
  10. H. J. Cheng, M. Nakamoto, A. D. Bergemann, J. G. Flanagan, ibid.82, 371 (1995).
  11. M. Nakamoto, H. J. Cheng, C. Yoon, J. G. Flanagan, unpublished data.
  12. M. Tessier-Lavigne, Cell82, 345 (1995).
  13. U. Drescher et al., ibid.359, 370 (1995).