Netrin-1 has been known to be an important axon guidance cue that plays a key role in neuronal wiring. We have determined crystal structure of netrin-1 in complex with one of its receptor, DCC (Deleted in Colorectal Cancer). Netrin-1 is in fact a bi-functional ligand. When it binds to DCC, which is constitutively expressed on the surface of axonal growth cone, it will trigger axon attraction toward the source of netrin-1. When another receptor UNC5 co-exists on the surface of growth cone, netrin-1 will bring DCC and UNC5 together, resulting in axon repulsion. DCC receptor is composed of 4 Ig-like domains at N-terminal, followed by 6 fibronectin (FN) domains. Since netrin-1 is known to bind membrane-proximal FN domains, we used FN5-FN6 for co-crystallization. Our structure demonstrates that netrin-1 simultaneously binds two DCC. The binding site 1 is DCC-specific whereas the site 2 is more generic. We propose that the more generic binding site can be replaced by other receptor, like UNC5. Functional data confirm that UNC5 is indeed able to out-compete DCC at the site 2 for repulsion outcome.
We have determined structures of N-terminal 4 Ig-like domains of a cell adhesion molecule, Dscam from Drosophila. The four domains fold into a horseshoe configuration. We unveil that the homophilic dimerization of this horseshoe is physiological. Drosophila Dscam has astonishing 40 thousands isoforms! Each neuron only expresses about 25 unique isoforms. During neuronal wiring, a self-avoidance mechanism ensures that the extension from the same neuron will not connect, only extension from different neuron is possibly connected. We show structurally that this is mediated by homophilic dimerization of horseshoe unit between the same Dscam isoform.
We have determined structure of DCC N-terminal four Ig-like domains, which also fold into horseshoe configuration. We have compared 4 published structures of Ig superfamily members that have a similar horseshoe unit, and identified structural elements that determine the horseshoe conformation. We further predict the existence of 23 horseshoe-like receptor structures in the whole human genome and have determined that these receptors overwhelmingly serve as neuroreceptors.