Agrin is an extracellular matrix (ECM) protein with a calculated relative molecular mass of more than 200 kD that induces the aggregation of acetylcholine receptors (AChRs) at the neuromuscular junction. This activity has been mapped to its COOH terminus. In an attempt to identify the functions of the NH2-terminal end, we have now characterized full-length chick agrin. We show that chick agrin encoded by a previously described cDNA is not secreted from transfected cells. Secretion is achieved with a construct that includes an additional 350 bp derived from the 5' end of chick agrin mRNA. Recombinant agrin is a heparan sulfate proteoglycan (HSPG) of more than 400 kD with glycosaminoglycan side chains attached only to the NH2-terminal half. Endogenous agrin in tissue homogenates also has an apparent molecular mass of > 400 kD. While the amino acid sequence encoded by the 350-bp extension has no homology to published rat agrin, it includes a stretch of 15 amino acids that is 80% identical to a previously identified bovine HSPG. The extension is required for binding of agrin to ECM. AChR aggregates induced by recombinant agrin that includes the extension are considerably smaller than those induced by agrin fragments, suggesting that binding of agrin to ECM modulates the size of receptor clusters. In addition, we found a site encoding seven amino acids at the NH2-terminal end of agrin that is alternatively spliced. While motor neurons express the splice variant with the seven amino acid long insert, muscle cells mainly synthesize isoforms that lack this insert. In conclusion, the cDNAs described here code for chick agrin that has all the characteristics previously allocated to endogenous agrin.

Agrin is a basal lamina protein that induces aggregation of acetylcholine receptors (AChRs) and other molecules at the developing neuromuscular junction. Alternative splicing of chick agrin mRNA at two sites, A and B, gives rise to eight possible isoforms of which five are expressed in vivo. Motor neurons express high levels of isoforms with inserts at sites A and B, muscle cells synthesize isoforms that lack amino acids at the B-site. To obtain further insights into the mechanism of agrin-induced AChR aggregation, we have determined the EC50 (effective concentration to induce half-maximal AChR clustering) of each agrin isoform and of truncation mutants. On chick myotubes, EC50 of the COOH-terminal, 95-kD fragment of agrinA4B8 was approximately 35 pM, of agrinA4B19 approximately 110 pM and of agrinA4B11 approximately 5 nM. While some AChR clusters were observed with 64 nM of agrinA4B0, no activity was detected for agrinA0B0. Recombinant full-length chick agrin and a 100-kD fragment of ray agrin showed similar EC50 values. A 45-kD, COOH-terminal fragment of agrinA4B8 retained high activity (EC50 approximately equal to 130 pM) and a 21-kD fragment was still active, but required higher concentrations (EC50 approximately equal to 13 nM). Unlike the 45-kD fragment, the 21-kD fragment neither bound to heparin nor did heparin inhibit its capability to induce AChR aggregation. These data show quantitatively that agrinA4B8 and agrinA4B19, expressed in motor neurons, are most active, while no activity is detected in agrinA0B0, the dominant isoform synthesized by muscle cells. Furthermore, our results show that a fragment comprising site B8 and the most COOH-terminal G-like domain is sufficient for this activity, and that agrin domains required for binding to heparin and those for AChR aggregation are distinct from each other.