Figure 3 Anti-agrin mAbs stain stacks of flattened cisternae in motor neurons characteristic of the Golgi apparatus. Cytoplasm of a motor neuron in the ventral horn of a frog spinal cord treated with a mixture of mAbs 3B5 and 5B1. Scale bar, 1 μm. in figure 5, these extracts caused the aggregation of AChRs on cultured chick myotubes. The extract from the electric lobe also caused AChE aggregation (data not shown).

To determine whether the AChR- and AChE-aggregating activities in the electric lobe extracts were antigenically related to agrin, we assayed the ability of anti-agrin antibodies to immunoprecipitate the activities. For these experiments, we used five mAbs, each of which was directed against a different epitope on agrin (Reist et al. 1987). Each of the mAbs immunoprecipitated nearly all of the AChR-aggregating activity from the extracts of electric lobe of Torpedo brain, as they did agrin from electric organ extracts (Reist et al. 1987). Two of these mAbs (the only ones tested) also immunoprecipitated electric lobe AChE-aggregating activity, as expected (data not shown). Most of the AChR-aggregating activity in extracts of the spinal cords of Torpedos, frogs, and chicks was also immunoprecipitated by anti-agrin antibodies. Thus, the AChR- and AChE-aggregating molecules in extracts of the electric lobe and spinal cord are antigenically similar to agrin.

Agrin-Like AChR-Aggregating Activity in Extracts of
Motor Neurons

To learn whether any of the AChR-aggregating activity detected in spinal cord extracts is derived from motor neurons, we separated motor neurons

Figure 4 Selective staining of motor neurons and nonneural structures by anti-agrin mAbs. Cross section of the lumbosacral region of a spinal cord from a 10 d chick embryo incubated with mAb 5B1. Motor neurons and the pial surface of the spinal cord are intensely stained. Capillaries (arrowhead) are lightly stained; compare with intense staining of capillaries at a later stage of development in figure 2. Glial cells and other neurons are not stained. The intensely stained structures outside of the spinal cord are ventral roots; much of the stain is probably in the Schwann cell basal lamina, which is known to stain intensely in the adult (Reist et al. 1987). The lightly stained region (arrow) of the spinal cord extending from the motor neurons to the ventral root was observed at higher magnification to be composed of narrow cell processes having a nearly uniform diameter, probably motor axons. Scale bar, 200 μm.

from other cellular components in the spinal cord of 6 d chick embryos according to the method described by Dohrmann et al. (1986). The separation procedure results in two cellular fractions, one in which more than 95 percent of the cells are motor neurons and the other with no or relatively few motor neurons (Dohrmann et al. 1986). When extracts of each fraction were made and tested for AChR-aggregating activity, we found that the specific activity of the motor neuron-containing fraction was sevenfold greater than that of the nonmotor neuron-containing fraction. Moreover, more than 60 percent of this activity was immunoprecipitated by anti-agrin antibodies. Thus, motor

neurons contain AChR-aggregating molecules antigenically related to agrin, and they are present at the time neuromuscular synapses are beginning to form.

Discussion

We find that the cell bodies of motor neurons stain with anti-agrin mAbs as does the synaptic basal lamina at neuromuscular junctions, and that the staining is concentrated in the Golgi apparatus, which processes proteins for secretion. We also provide evidence that motor neurons contain agrin-like AChR/AChE-aggregating molecules. Taken together, these findings support our hypothesis, as proposed by Nitkin et al. (1987), that motor neurons synthesize agrin or agrin-like AChR/AChE-aggregating molecules and release them at their axon terminals to become incorporated into the basal lamina of the synaptic cleft, and that these molecules account for the synaptic basal lamina's ability to induce and maintain AChR aggregates on muscle fibers. Our evidence that such molecules are present in the cell bodies of motor neurons in embryos and normal adults suggests further that they also account for the motor neuron's ability to cause the formation of postsynaptic specializations on mature muscle fibers. Our hypothesis does not rule out a role for other neuron-derived factors in the formation of the postsynaptic apparatus, such as regulating the levels of AChRs. In this regard, electromechanical activity, ARIA, and CGRP are discussed by Nitkin et al. (1987).

An alternative interpretation of our observation that the cell bodies of motor neurons contain agrin-like molecules is that such molecules are produced and secreted by muscle fibers and/or Schwann cells and are taken up by motor neurons. However, there is no evidence that proteins secreted by Schwann cells, muscle fibers, or other target cells and internalized by

Figure 5 (see facing page) Extracts of motor neuron-containing regions of the CNS of Torpedo, frog, and chick contain AChR-aggregating activity, which is immunoprecipitated by anti-agrin mAbs. Each of several different mAbs immunoprecipitated AChR-aggregating activity from the Torpedo electric lobe. mAb 6D4, which stains neuromuscular junctions and motor neurons in Torpedo but not in frog and chick, immunoprecipitates AChR-aggregating activity from the electric lobe and spinal cord of Torpedo, but not from the spinal cords of the other two species. On the other hand, mAbs 5B1 and 3B5, which do stain neuromuscular junctions in frog and chick, also immunoprecipitated the activity in the extracts from these two species. Such findings would be expected if the active molecules were identical or closely related to the molecules that stain in motor neurons and at the neuromuscular junction. Data expressed as mean ± S.E.M.; the number of observations is given in parentheses above the bars. Our methods of measuring activity and of immunoprecipitation are given in Godfrey et al. 1984 and Reist et al. 1987.

neurons appear in the neuron's Golgi apparatus, although certain plant lectins that bind tightly to membrane glycoproteins have been shown to appear in the Golgi apparatus of neurons after internalization (Gonatas et al. 1975).

The hypothesis that agrin, an extracellular matrix molecule, mediates the nerve-induced formation and maintenance of AChR and AChE aggregates on muscle fibers raises many questions. For example, does agrin trigger the formation of the entire postsynaptic apparatus? The postsynaptic apparatus consists of several components in addition to the high concentration of AChRs and AChE. They include high concentrations of heparan sulfate proteoglycan and butyrylcholinesterase on the cell surface, of a 43 kD protein associated with the cytoplasmic domain of the AChR, and of cytoskeletal elements and infoldings of the plasma membrane (junctional folds). Studies are underway to determine whether agrin-induced patches of AChR and AChE have each of these components. Already, the patches have been found to contain a high concentration of heparan sulfate proteoglycan, butyrylcholinesterase, and the 43 kD protein (Smith et al. 1987; Wallace 1987).

Are there agrin or agrin-like molecules at neuron-to-neuron synapses? Indeed, in unpublished experiments we have detected AChR-aggregating activity in extracts from regions of the Torpedo and frog brain that contain few motor neurons. The specific activity of such molecules from these regions of Torpedo brain was slightly above background, but the activity from frog brain was nearly as great as from the spinal cord, apparently too great to be accounted for by the presence of the relatively few motor neurons. The active molecules in the frog brain extracts were immunoprecipitated with anti-agrin mAbs, indicating that they are antigenically similar to agrin. We have not observed anti-agrin mAb staining in nonmotor neurons or at neuron-toneuron synapses. It may well be that agrin-like molecules are produced by many types of neurons to both induce and maintain the formation of postsynaptic specializations at their synapses but are too low in concentration to be detected by our staining techniques.

Do agrin-like molecules play a role in the aggregation of cell surface proteins at sites other than synapses? We have previously reported that, in frog, anti-agrin mAbs stain the external surface of axonal membranes at nodes of Ranvier (Reist et al. 1987), sites where sodium channels are aggregated. We also have reported that the basal lamina between capillary endothelium and astrocyte endfoot processes in the CNS stains (Magill-Solc and McMahan, submitted); the astrocyte endfoot processes have a high concentration of potassium channels (Newman 1986). Clearly, the discovery of agrin may have several interesting consequences.

Acknowledgments

These studies were supported by National Institutes of Health grant NS14506, grants from the Wills Foundation, the Weingart Foundation, Mr.