Developmental Biology 205, 22–32 (1999) Article ID dbio.1998.9104, available online at http://www.idealibrary.com on

View metadata, citation and similar papers at core.ac.uk brought to you by CORE Overexpression of Isoforms in Xenopus provided by Elsevier - Publisher Connector Embryos Alters the Distribution of Synaptic Acetylcholine Receptors during Development of the

Earl W. Godfrey,* Jeremy Roe,* and R. David Heathcote† *Department of Cell Biology, Neurobiology, and Anatomy, Medical College of Wisconsin, Milwaukee, Wisconsin 53226; and †Department of Biological Sciences, University of Wisconsin—Milwaukee, Milwaukee, Wisconsin 53201

Synapse formation involves a large number of macromolecules found in both presynaptic nerve terminals and postsynaptic cells. Many of the molecules involved in of the neuromuscular junction have been discovered through morphological localization to the and functional cell culture assays, but their role in embryonic development has been more difficult to study. One of the best understood of these molecules is agrin, a synaptic secreted by both motor neurons and muscle cells, that organizes the postsynaptic apparatus, including high-density aggregates of acetylcholine receptors (AChRs), at the neuromuscular junction. We tested the specific hypothesis that different agrin isoforms made by neurons and muscle cells contribute to agrin’s synapse organizing activity in the embryo. Agrin isoforms were overexpressed by injecting synthetic RNA into Xenopus laevis embryos at the one- or two-cell stage. To mark cells containing agrin RNA, green fluorescent protein (GFP) RNA was coinjected. The relative area of muscle AChR aggregates was measured by confocal microscopy and image analysis in GFP-positive segments of injected embryos. Innervated regions of myotomal muscles were compared in animals injected with a mixture of agrin and GFP RNAs or with GFP RNA alone. Overexpression of COOH-terminal 95-kDa fragments of a rat agrin isoform made only by neurons (4,8) and the major isoform (0,0) made by muscle cells both increased AChR cluster area by 100–200%. Rat agrin protein was colocalized with AChR aggregates in innervated regions of muscles in injected embryos. These results show that agrin derived from both the nerve terminal and the muscle cell could contribute to synaptic differentiation at the embryonic neuromuscular junction. They further demonstrate the usefulness of overexpression by RNA injection as an assay for molecular function in embryonic synapse formation. © 1999 Academic Press Key Words: agrin; synaptogenesis; neuromuscular junction; acetylcholine receptors; Xenopus embryos.

INTRODUCTION splicing of its mRNA (Ferns et al., 1992; Tsim et al., 1992). Among these, isoforms with inserts at both of two positions Several molecules play a role in the formation and differ- near the C terminus of the protein (“A” and “B,” Ruegg et entiation of neuromuscular . Agrin is a widely al., 1992; “y” and “z;” Hoch et al. 1993) are exclusively distributed basal lamina protein that is concentrated at the neuronal (Smith and O’Dowd, 1994). Compared to isoforms neuromuscular junction (Reist et al., 1987; Godfrey et al., (e.g., 0,0 and 4,0) made by nonneural tissues, including 1988a,b), where it is localized to the synaptic basal lamina muscle (Ruegg et al., 1992; Hoch et al., 1993), the neuronal (Reist et al., 1987). Agrin is made and secreted by both isoforms (e.g., 4,8) have much greater AChR aggregating motor neurons and muscle cells (Magill-Solc and McMa- activity on cultured muscle cells, when they are added to han, 1988; Cohen and Godfrey, 1992; Ruegg et al., 1992; the culture medium as soluble (Hoch et al., 1994; Lieth and Fallon, 1993; Ma et al., 1994, 1995). Agrin directs Gesemann et al., 1995). However, when the 0,0 and 4,0 the formation of the postsynaptic apparatus, including a isoforms are expressed on the surface of transfected cells, high-density aggregation of acetylcholine receptors (AChRs), their ability to aggregate AChRs upon contact with cocul- and consists of several isoforms derived from alternative tured myotubes is comparable to that of neuronal isoforms

0012-1606/99 $30.00 Copyright © 1999 by Academic Press 22 All rights of reproduction in any form reserved. Agrin Isoforms Alter Embryonic Synapse Formation 23

(Ferns et al., 1992, 1993). Here we show that embryonic agrin protein were used as templates for RNA synthesis after overexpression of agrin isoforms normally secreted by both subcloning into pCS2ϩ as ClaI/XbaI fragments. A cDNA encoding neurons (4,8) and muscle cells (0,0) increases the differen- a red-shifted variant (F64L S65T) of the jellyfish green fluorescent tiation of postsynaptic structures, in contrast to some protein (GFP-LT; gift of Dr. Richard Dorsky, University of Califor- ϩ results from cell culture assays and transfection of adult nia, San Diego) in pCS2 was used to prepare GFP RNA. muscles. These results suggest that both neuronal and muscle agrin isoforms play a role in postsynaptic differen- RNA Synthesis tiation of the embryonic neuromuscular junction. The pCS2ϩ constructs were linearized with XbaI (for rat ) The function of proteins in synapse formation has largely or NotI (for GFP). RNAs were synthesized with SP6 RNA polymer- been extrapolated from results of cell culture assays. For ase using the mMessage mMachine kit (Ambion) and analyzed and example, motor neuron–muscle cell cultures have been quantitated after electrophoresis on agarose gels. Agrin RNAs were very useful for studying early stages of synapse formation about 3.6 kb and GFP RNA was 1.2 kb, as expected. RNA (e.g., Cohen and Godfrey, 1992), but later stages of synap- concentrations of 0.5–4 ␮g/␮l (agrin) and 0.1 ␮g/␮l (GFP) were used togenesis have not been well characterized in these cul- for injection. tures. Another valuable tool is the elimination of candidate molecules by altering their genes in transgenic “knockout” RNA Injection mice. Use of knockout mice has shown that agrin (Gautam Embryos of X. laevis were obtained by in vitro fertilization et al., 1996), muscle-specific kinase (DeChiara et al., 1996), (Moon and Christian, 1987) and injected with synthetic RNAs at and rapsyn (Gautam et al., 1995) are essential for the the one- or two-cell stages using a Nanoject (Drummond) or formation of embryonic neuromuscular junctions. How- Picospritzer II (General Valve) through glass pipets with a tip ever, since mice lacking expression of any of these proteins diameter of 10–20 ␮m. Embryos were injected (Moon and Chris- die at birth, the study of synaptic development is difficult in tian, 1987) with 5 nl per blastomere containing GFP RNA (0.5 ng) these animals. We have developed a novel approach to test or agrin (2.5–20 ng) plus GFP (0.5 ng) RNA. the role of proteins in the formation of the neuromuscular synapse in the developing embryo. The test involves inject- Screening, Fixation, and Labeling Acetylcholine ing RNA into early blastomeres to overexpress candidate Receptors molecules during development of the frog, Xenopus laevis. Injection of RNAs into early Xenopus embryos is fre- Following injection of RNAs, embryos were transferred into ϫ quently used to study the role of proteins in early develop- 0.1 modified Barth’s saline (Gurdon and Wickens, 1983) and allowed to develop at room temperature. The next day, they were ment (Vize et al., 1991). Some studies have also utilized this screened for normal development and incubated at 10–18°C over- method to evaluate the role of molecules involved in night. Embryos were staged according to Nieuwkoop and Faber organogenesis. For example, Gove et al. (1997) used overex- (1994). After neuromuscular synapse formation (usually stage 31), pression of the transcription factor GATA-6 to examine its embryos were screened for GFP fluorescence. Normal embryos effect on heart development in Xenopus, which they as- with fluorescence in the trunk myotomes and/or neural tube were sayed in embryos at stages 26–35. In the present study we fixed in 4% paraformaldehyde for1hat4°Candrinsed in Ringer, used RNA injection to overexpress both neuronal and and the skin was removed with fine forceps. Embryos were incu- nonneuronal isoforms of agrin. We assayed the effects of bated in rhodamine–␣-bungarotoxin (0.5–2 ␮g/ml, Molecular overexpression at stages 28–33/34 (1.3–1.8 days of develop- Probes) in 1% goat serum in Ringer overnight at 4°C, washed, ment), during the period when synapses form in myotomal incubated1hinmounting medium (Valnes and Brandtzaeg, 1985), and mounted on slides. muscles. We show that RNA injection is a reliable method for overexpression of proteins in the developing neuromus- cular system. We used confocal microscopy to image seg- Confocal Microscopy mental groups of neuromuscular synapses labeled with AChR aggregates were imaged by confocal microscopy. Images fluorescent probes, thus enabling us to assay the effect of were taken from the fifth through the seventh myotomes in any protein on the development of the embryonic neuro- embryos with GFP fluorescence in this region. These myotomes muscular junction. were chosen because the synaptic bands of AChR aggregates were more regular and continuous than in more rostral or caudal myotomes. To compare the extent of AChR clustering in synaptic METHODS regions of control and experimental animals, six different Z-series of four optical sections (at 1-␮m intervals) were acquired from four cDNAs and Subcloning to six animals in each group. Microscope settings remained con- stant during imaging of all groups of animals within an experiment To synthesize mRNAs that could be translated readily in Xeno- to ensure comparability of the images. pus embryos, cDNAs were subcloned into the multiple cloning site ϩ of pCS2 (Rupp et al., 1994; Turner and Weintraub, 1994). Rat Image Analysis agrin cDNAs [0,0 “muscle” and 4,8 “neuronal” isoforms at posi- tions y and z (Hoch et al., 1993); gift of Dr. Michael Ferns, Montreal The area of AChR aggregates in myotomal muscles was mea- General Hospital] encoding the COOH-terminal 95-kDa of the sured with Metamorph image analysis software (Universal Imag-

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved. 24 Godfrey, Roe, and Heathcote

ing). Threshold was set to mark the aggregates in each series of confocal images, and aggregate area was converted to a percentage of the total area of the entire series of images.

Statistics Mean values and SE were calculated for each group of images. To create a more normal distribution of the data (percentages of area containing AChR aggregates), a one-tailed t test was performed after an arcsin conversion (Zar, 1974), comparing values of mean percentage area from each injected group of embryos with values from the GFP control group.

Immunofluorescence Rat agrin protein was detected in RNA-injected embryos with a rabbit anti-rat agrin antiserum (1:1000) made against the COOH- terminal 95-kDa portion of isoform 4,8 (gift from Drs. Janice Sugiyama and Zach Hall, National Institute of Mental Health). This antibody did not cross-react with endogenous Xenopus agrin. Antibody-binding sites were revealed with FITC-labeled goat anti- rabbit antibodies (Cappel) and imaged by confocal microscopy.

FIG. 1. Innervation of myotomal muscles in the Xenopus embryo, Reverse Transcription–Polymerase Chain Reaction stage 31. (a) Diagram of stage 31 embryo showing myotomal (RT-PCR) muscles (shaded). (b) Innervation of both ends of muscle cells by To determine how long injected RNA persisted in the embryo, motor axons is indicated by arrows. Bands of synapses from RT-PCR was performed. Groups of five embryos, uninjected or adjacent myotomes are closer together than indicated and cannot injected with 10 ng of rat agrin (4,8) RNA, were frozen at stages 2, always be distinguished by light microscopy. 9, 32, and 35/36. Total RNA was isolated using the acid guanidinium/phenol method of Chomzcynski and Sacchi (1987). cDNA synthesis and PCR amplification were performed using RT-PCR beads (Pharmacia) with 350 ng total RNA. Rat agrin tively small variability resulted in SE of 10–20% of the Ј Ј forward and reverse primers were 5 -AAG GTC AAC ACC AAC-3 mean in uninjected and GFP RNA-injected embryos (e.g., Ј (bases 5722–5736) and 5 -GGT CGA TAA TAC GAC TCA CTA Table 1). The organization of Xenopus embryo neuromus- TAG GGA GGG AGT GGG GCA GGG-3Ј (T7 promoter sequence cular junctions allowed us to image large segmental en- and bases 6007–6021; Rupp et al., 1991). Primers for Xenopus ornithine decarboxylase (ODC; Bassez et al., 1990) were used as a sembles of synapses by confocal microscopy (e.g., Fig. 5), control for RNA purification. Amplification was performed in a thus averaging variation between synapses in a myotome. PTC-100 thermal cycler (M.J. Research) using 23 cycles of 95, 55, Development of these synapses occurs rapidly, and the and 72°C (each 1 min). A control reaction without template RNA free-living embryos are readily accessible. The first func- was included. The relative intensity of DNA bands was determined tional neuromuscular synapses form 1 day after fertiliza- by laser scanning of ethidium bromide-stained agarose gels using a tion, as evidenced by the onset of muscle movement FluoroImager (Molecular Dynamics) and ImageQuant software. (Nieuwkoop and Faber, 1994), the appearance of synaptic potentials and nerve–muscle contacts (Kullberg et al., 1977), and the aggregation of AChRs at these early synapses RESULTS (Chow and Cohen, 1983; Cohen and Godfrey, 1992). Since some injected RNAs persist past this stage (see below; Fig. Innervation of Myotomal Muscles 4), altering expression of proteins during neuromuscular We chose the Xenopus embryo neuromuscular system synaptogenesis by RNA injection into one- or two-cell because it allowed us to perform a convenient, rapid, and embryos was possible. quantitative assay of synapse formation in the embryo. Synapses in the myotomal muscles of Xenopus tadpoles GFP Is a Marker for RNA Injection and form in regular rows at both ends of each myotome, Translation resulting in synaptic “stripes” bordering the septa between myotomes (arrows, Fig. 1b). The regular nature of the To visualize the successful injection, diffusion, and transla- myotomes and the synapses within them resulted in small tion of synthetic mRNA in the embryo, we injected a mixture enough variability in the relative area of synaptic AChR of RNAs encoding agrin and GFP. Embryos injected with GFP aggregates that differences between control and experimen- RNA alone served as a control. In Fig. 2, an embryo is shown tal groups could readily be detected. Typically, this rela- in which one blastomere had been injected with GFP RNA at

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved. Agrin Isoforms Alter Embryonic Synapse Formation 25

FIG. 2. GFP is a marker for successful RNA injection and indicates the extent of diffusion and translation of injected RNA and the persistence of the protein product. One blastomere of a two-cell stage embryo was injected with GFP RNA. GFP fluorescence is seen in most cells on the right (upper) half of this tadpole (stage 35/36, 2.1 days of development). FIG. 3. Rat agrin is colocalized with AChR aggregates in confocal images of the innervated region of myotomal muscles. Embryos were injected at the one-cell stage with rat agrin RNA (neuronal isoform 4,8) in the absence of GFP RNA. Whole mounts of fixed stage 31 (1.6 day) embryos were stained as described under Methods. (A) Rat agrin immunofluorescence (FITC-labeled antibody). (B) AChR aggregates labeled with rhodamine–␣-bungarotoxin. (C) Superimposed images from A and B showing colocalized rat agrin and AChR aggregates (yellow). Arrows indicate synaptic regions adjacent to the myotomal septum. Bar, 20 ␮m.

the two-cell stage. At 2 days of development (stage 35/36), deposited at neuromuscular junctions in the injected embryos bright GFP fluorescence was seen in most cells derived from and that it was biologically active. Superimposing these two the injected blastomere, indicating that GFP RNA diffused images shows considerable colocalization of rat agrin with widely and evenly within this blastomere before it divided. AChR aggregates (yellow, Fig. 3C), particularly in the inner- Thus, GFP is a reliable marker that enables us to screen large vated region. In other embryos, punctate deposits of rat agrin numbers of living embryos for successful injection and trans- immunoreactivity were also seen in other parts of myotomes lation of RNA and to identify regions of injected embryos that (data not shown). The identity and AChR-aggregating activity are also likely to express the experimental gene product. of rat agrin encoded by the synthetic RNA were confirmed by in vitro translation, followed by agrin bioassay (AChR aggre- gation in primary chick muscle cultures) and immunoblotting Overexpressed Rat Agrin Is Colocalized with of the translation product with the species-specific antibody AChR Aggregates in Embryonic Muscles (data not shown). Overexpressed rat agrin (neuronal isoform 4,8; Fig. 3A) was colocalized with AChR aggregates (Fig. 3B) by labeling agrin Rat Agrin RNA Persists during Neuromuscular RNA-injected embryos with a species-specific anti-rat agrin Synaptogenesis antibody. Rat agrin protein was found in a punctate pattern in the innervated region of 1.6-day (stage 31) embryonic muscles Although the capped synthetic RNA that we injected into (myotomal septum; arrows, Fig. 3), suggesting that it was embryos should be stable, even stable RNAs are eventually

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved. 26 Godfrey, Roe, and Heathcote

FIG. 4. Injected agrin RNA persists throughout embryogenesis. Rat agrin RNA was detected in RNA-injected embryos by RT-PCR. (a) Rat agrin sequence. (b) Ornithine decarboxylase sequence from Xenopus as an internal control. Lanes: 1, no RNA control; 2–5, embryos injected with agrin RNA (10 ng) at the one-cell stage and analyzed at stages 2, 9, 32, and 35/36 (2 h–2.1 days); 6 and 7, uninjected control embryos (stages 9 and 32).

degraded. To determine whether rat agrin RNA persisted through the initial period of synaptogenesis, we used RT- PCR and RNase protection assays. Rat agrin RNA was detected in injected tadpoles as late as 1.7 days (stage 32) by RNase protection (data not shown) and 2.1 days (stage 35/36) by RT-PCR (Fig. 4). The amount of rat agrin cDNA amplified from RNA in hatched tadpoles (2.1 days, stage 35/36) was 25% of that amplified from embryos Ϸ1 h after injection (stage 2), as determined by laser scanning and quantitation of fluorescence intensity of the bands shown photographically in Fig. 4. Thus, significant amounts of injected rat agrin RNA (Fig. 4) and protein (Fig. 3) persisted well after neuromuscular synaptogenesis began.

Overexpression of Neuronal and Muscle Agrin Isoforms Increases AChR Aggregation in Synaptic Regions of Embryonic Muscle To assay the effects of overexpression of agrin isoforms on synapse formation, we compared AChR aggregation at the neuromuscular junctions of animals injected with rat agrin RNAs plus GFP marker RNA, those injected with GFP RNA alone, and uninjected animals. Recombinant rat agrin aggregated AChRs on cultured Xenopus muscle cells (M. W. Cohen, personal communication). Overexpression of both neuronal (4, 8) and muscle (0, 0) isoforms of rat agrin

FIG. 5. Overexpression of neuronal and muscle agrin isoforms increases AChR aggregate area in innervated regions of myotomal muscles. Xenopus embryos were injected at the one-cell stage with RNAs encoding GFP (A), rat neuronal agrin (4,8) and GFP (B), or rat muscle agrin (0,0) and GFP (C). AChRs were labeled with rhodamine–␣-bungarotoxin and images were obtained by confocal microscopy at stage 31 (1.6 days). Overlaid series of four images, taken at 1-mm intervals, are shown. Bands of AChR aggregates (bright clusters) running horizontally across the images represent innervated regions at the ends of two adjacent myotomes; muscle cells run vertically, parallel to the anterior–posterior axis of the embryo. The images shown were taken from the same experiment; AChR aggregate areas were 0.61% (A), 2.94% (B), and 2.16% (C) of area imaged. Bar, 20 ␮m. Agrin Isoforms Alter Embryonic Synapse Formation 27

TABLE 1 Overexpression of Neuronal and Muscle Agrin Isoforms in Xenopus Embryos Increases Acetylcholine Receptor Aggregation in Myotomal Muscles

AChR aggregate area P value Expt RNA injected (ng) Stage fixed (%) Ϯ SE Area ratio agrin:GFP (vs GFP)

1 None 33/34 0.32 Ϯ 0.03 — 1.00 GFP (0.5) 33/34 0.26 Ϯ 0.03 — — Agrin 4,8 (10) ϩ GFP (0.5) 33/34 0.58 Ϯ 0.07 2.23 0.001 Agrin 0,0 (10) ϩ GFP (0.5) 33/34 0.78 Ϯ 0.14 3.00 0.006 2 GFP (0.5) 31 0.28 Ϯ 0.07 — — Agrin 4,8 (2.5) ϩ GFP (0.5) 31 0.64 Ϯ 0.17 2.29 0.049 Agrin 4,8 (5) ϩ GFP (0.5) 31 0.74 Ϯ 0.03a 2.64 0.003 Agrin 4,8 (10) ϩ GFP (0.5) 31 1.05 Ϯ 0.30 3.75 0.007

Note. Agrin was overexpressed in Xenopus embryos by injection of RNA encoding either 4,8 (neuronal) or 0,0 (muscle) isoforms. Area ratio refers to the ratio of AChR aggregate area in agrin RNA-injected groups compared to corresponding GFP RNA-injected control groups. a Average of data from three microscopic fields. All other groups consists of data from six different fields from four to six embryos.

also resulted in a significant increase in AChR aggregation shown), the relative increase due to agrin overexpression in embryonic muscle cells (Figs. 5B and 5C). Most of the was similar (Table 2). Injection of increasing amounts of increase in aggregation was in normally innervated regions neuronal agrin RNA (4,8 isoform) led to increasing AChR of muscle (horizontal stripes of clusters, Fig. 5), but in some aggregation in the embryos (Experiment 2, Table 1). experiments there was also an increase in extrajunctional Due to conflicting reports of AChR-aggregating activity of aggregation (above and below stripes of clusters). AChR agrin isoforms in culture, it was unclear whether the 0,0 aggregation was strikingly different in muscles of control isoform, which is the major form made by muscle cells, would (GFP RNA alone, Fig. 5A; uninjected, not shown) and display activity in the embryo. However, overexpression of experimental (agrin plus GFP RNA, Figs. 5B and 5C) em- both 4,8 and 0,0 isoforms caused significant increases in bryos. There were more and larger AChR aggregates in AChR aggregation in embryonic muscles (Table 1). No obvi- muscles of embryos overexpressing agrin than in those ous difference in the pattern of AChR aggregates induced by expressing GFP alone. Overexpression of either neuronal or the two isoforms was noticed. The frequency of experiments muscle agrin isoforms resulted in AChR aggregate areas showing significant increases in embryonic AChR aggregation two- to fourfold those of control embryos (Table 1). The (Table 2) was similar for groups of embryos injected with RNA increase in AChR aggregation was similar in regions of encoding the 4,8 neuronal isoform (65%) and the 0,0 muscle injected embryos in which GFP was expressed in the neural isoform (50%). The potency of these isoforms in the embryo tube and regions with GFP in myotomes (data not shown). was also similar. In groups of embryos showing a significant Although the AChR aggregate area in both control and increase, AChR aggregate area averaged 2.33- (4,8) and 2.68- experimental animals varied between experiments (not (0,0) fold that of controls, respectively. Thus, agrin isoforms

TABLE 2 Summary of Agrin RNA Injection Experiments

% of groups Significantly increased Area ratio agrin:GFP No significant difference significantly RNA injected groups* (avg Ϯ SE) groups* increased

Neuronal agrin 4,8 11 2.33 Ϯ 0.22 6 65 (Ն10 ng) Muscle agrin 0,0 3 2.68 Ϯ 0.30 3 50 (Ն10 ng)

Note. Each group consisted of four to six imaged embryos which had been injected at the one- or two-cell stage with the same RNA (type and concentration) in the same experiment, and which contained GFP in the areas analyzed. * “Significantly increased groups” had an average AChR aggregate area significantly (P Ͻ 0.05) greater than that of control animals injected with GFP RNA alone, while “no significant difference groups” did not (P Ͼ 0.05). Area ratios shown were calculated from groups of embryos showing significant increases over control.

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved. 28 Godfrey, Roe, and Heathcote produced by both the nerve terminal and the muscle cell may the amount of RNA remaining at this stage was signifi- contribute to the differentiation of the postsynaptic apparatus cantly lower than originally injected (25% or less of the in the embryo. original amount), the protein expressed was abundant enough that immunofluorescence revealed rat agrin colo- calized at AChR aggregates in muscle. While these results DISCUSSION show that protein translated from RNAs injected into embryos at the one-cell stage could be detected as late as Overexpression of Synaptic Proteins during stage 31, we found that injection of at least 10 ng of agrin Embryonic Synapse Formation RNA per embryo was required to stimulate increased AChR While many synaptic macromolecules have been dis- aggregation consistently. Thus, the RNA injection method covered, knowledge of molecular mechanisms by which is suitable for overexpression of synaptic molecules and synapses are organized remains sketchy. Considerable observation of the effects on neuromuscular synapse forma- progress has been made, however, in understanding some tion. of the molecular interactions involved in the organiza- tion of the postsynaptic apparatus by agrin, different Overexpression of Both Neuronal and Muscle isoforms of which are externalized at the neuromuscular Agrin Isoforms Increased AChR Aggregation in junction by motor nerve terminals and muscle cells Embryonic Muscles (Ruegg et al., 1992; Ferns et al., 1993; Ma et al., 1994). Many of these interactions have been established by Overexpression of agrin in the Xenopus embryo produced examining protein–protein interactions in cell-free sys- 100–200% increases in AChR aggregate area in myotomal tems (e.g., Campanelli et al., 1996; Gesemann et al., muscles. The extent of the increase in AChR aggregation 1996; Hopf and Hoch, 1996; O’Toole et al., 1996) or in depended on the amount of agrin RNA injected, consistent transfected cells (Apel et al., 1995, 1997; Denzer et al., with the dose dependence of AChR aggregation in agrin- 1997; Fuhrer et al., 1997). Corresponding studies of treated muscle cell cultures (Godfrey et al., 1984). Thus, embryonic synapse formation have recently been focused agrin levels can be manipulated in the developing embryo, on null mutant knockout mice, which have confirmed a resulting in an appropriate biological response. It appears requirement for agrin, MuSK, and rapsyn for the forma- that agrin receptors in these embryonic muscles were not tion of the postsynaptic specialization of the neuromus- saturated by endogenous agrin and that the area of AChR cular junction in vivo (Gautam et al., 1995; DeChiara et aggregates reflected the extent of agrin receptor activation al., 1996; Gautam et al., 1996). Because the developmen- in the embryo. An excess of agrin receptors at the neuro- tal role of molecules in the living embryo can only be muscular junction might facilitate the growth of the syn- established by in vivo studies, additional embryonic apse during development and its remodeling in the adult. assay systems are required to gain a more complete The results were variable in two ways. First, the relative understanding of mechanisms of embryonic synaptogen- area of AChR aggregates in muscle varied between experi- esis. To address this need, we have utilized overexpres- ments in both controls and agrin RNA-injected embryos sion of synaptic molecules by RNA injection in Xenopus (Table 1 and data not shown). Second, some of the groups of embryos to study their function in the formation of agrin RNA-injected embryos did not show a significant neuromuscular synapses. increase in AChR aggregate area when compared to controls During development of myotomal muscles in the Xeno- (Table 2). The variability in average area between groups of pus embryo, innervation of muscle cells by motor axons embryos in different experiments could reflect biological from spinal cord begins at about 1 day of development (stage variability between groups of embryos with different par- 22; Kullberg et al., 1977). One of the first events seen at ents, variability in labeling of AChRs, and differences nascent neuromuscular synapses is the aggregation of between experiments in settings on the confocal micro- muscle cell AChRs at nerve–muscle contacts (Chow and scope and resulting intensity of aggregates in images. The Cohen, 1983; Cohen and Godfrey, 1992). Appearance of lack of a significant increase in AChR aggregate area in AChR aggregates can therefore reflect the establishment of some groups of agrin RNA-injected embryos could be new synapses in the embryonic muscles. We have measured caused by variability in the stability or translation of agrin the area of AChR aggregates as a reflection of the ability of RNA, differences in the extent of diffusion of injected RNA the overexpressed molecules to affect the differentiation of (which would affect its concentration), and/or variability of the postsynaptic membrane. AChR aggregate area within a group of embryos. All imaged regions of agrin RNA-injected embryos contained visible GFP; however, agrin RNA or protein may be degraded more Injected RNA and Encoded Protein Persist during rapidly than GFP. The doses of agrin RNA we used could be Synaptogenesis in the Embryo close to the minimum needed to consistently cause the Injected agrin RNA and protein translated from it were observed effect. Injecting smaller amounts of RNA resulted detectable in injected embryos up to the stage when they in a larger proportion of experiments with no significant were fixed for analysis of synaptic differentiation. Although increase in area. However, we did not increase the amount

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved. Agrin Isoforms Alter Embryonic Synapse Formation 29 of RNA injected to more than 10–15 ng/embryo because of heart (Mjaatvedt et al., 1987) and in chick and mouse increased toxicity seen after injecting larger amounts of embryo premuscle masses in the limb bud (Godfrey et al., RNA. 1988b; Godfrey and Gradall, 1998). Perhaps most of the Both an agrin isoform found only in neurons (4,8) and one agrin overexpressed by embryonic muscle cells in our found in many cell types that is also the major isoform in experiments was more diffusely distributed on the surface muscle cells (0,0) produced increased clustering in our of the cells, where it caused less AChR aggregation than experiments. The frequency and extent of the increase were agrin which was overexpressed by neurons and concen- similar for the two isoforms, indicating that their potency trated in the synaptic region by nerve terminals. in embryonic muscle was similar. Some investigators have Our results showing the AChR-aggregating activity of a referred to the 0,0 isoform as “inactive,” since little or no muscle agrin isoform in the embryo raise the question of AChR-aggregating activity is seen when recombinant agrin the role of agrin secreted by muscle cells at the normal 0,0 is added to the culture medium of muscle cells (Ruegg et embryonic synapse. Some neurons in the ciliary ganglion of al., 1992; Ferns et al., 1993). However, even when assayed chick embryos express the agrin 4,0 isoform, which is also in this way, agrin extracted from adult skeletal muscle had expressed by muscle cells (Smith and O’Dowd, 1994; Hoch a small but measurable activity (Godfrey, 1991). When agrin et al., 1993), but it is not known if this isoform is also cDNA constructs were used to transfect adult rat muscle expressed by motor neurons. However, the 0,0 isoform is fibers, neuronal isoforms of agrin induced complete the predominant one expressed in embryonic chick and rat postsynaptic specializations extrajunctionally (Cohen et muscles (Ruegg et al., 1992; Hoch et al., 1993; Ma et al., al., 1997; Jones et al., 1997), while other isoforms did not 1994). Unfortunately, the normal expression of agrin in (Cohen et al., 1997). In contrast to these results, other muscle cells of Xenopus embryos has not been character- studies have shown that both sets of isoforms can display ized, since the available anti-agrin antibodies do not label AChR-aggregating activity. When the 0,0 and 4,0 isoforms these cells (Cohen and Godfrey, 1992), and Xenopus agrin were expressed on the surface of transfected cells which cDNA has not yet been cloned. Muscle-derived agrin is were cocultured with muscle cells (Ferns et al., 1993), these concentrated at AChR clusters induced by contact with isoforms were about 80% as active as the neuronal iso- motor axons (Lieth and Fallon, 1993), but agrin immuno- forms. In another study, NG108-15 neuroblastoma–glioma fluorescence surrounding embryonic muscle fibers is much hybrid cells induced AChR aggregation on muscle cells, but less intense than at the neuromuscular junction (Godfrey et expressed only the 0,0 isoform of agrin (Pun and Tsim, al., 1988b; Hoch et al., 1993). Thus, muscle agrin is likely to 1997). Taken together, these results suggest that the agrin be more concentrated at the synapse than extrasynaptically, isoforms found in muscle cells have the ability to organize which could explain why isoforms expressed in muscle the postsynaptic apparatus, but this activity may not be cells do not normally induce extrasynaptic AChR aggrega- apparent unless agrin is presented to the muscle cell on the tion. Since muscle agrin is colocalized with AChR aggre- surface of a cell or in the matrix. Such solid-phase presen- gates induced on muscle cells by neuronal agrin (Lieth et tation of agrin would allow agrin isoforms (such as 0,0), al., 1992), it is likely that muscle agrin isoforms also which may have a low affinity for the agrin receptors when become concentrated at the embryonic synapse, where they presented in solution, to bind to these receptors with an could contribute to postsynaptic differentiation. effectively higher affinity, stimulating AChR aggregation. Thus, the method of assaying agrin’s synapse-organizing Distribution of AChR Aggregates Induced by activity is critical. Assays using cell-attached agrin may Overexpressed Agrin more closely resemble the situation at the developing neuromuscular junction than assays in which agrin is added In most of the embryos overexpressing agrin, AChR to the medium of cultured muscle cells. aggregation increased primarily in the regions of innerva- In Xenopus embryos, neuronal agrin is likely to be tion at the ends of myotomes. This may have resulted from presented to muscle cells on the surface of motor axons greater agrin overexpression by neurons than by muscle rather than being released into the extracellular space. This cells or it could reflect the cellular geometry of the synapse, hypothesis is suggested by previous results from Xenopus at which nerve terminals concentrate neuronal agrin. Con- spinal cord–muscle cell cocultures (Cohen and Godfrey, centration of agrin receptors at the ends of the muscle cells 1992), in which most AChR aggregates at nerve–muscle could also contribute to preferential aggregation of AChRs contacts were colocalized with punctate aggregates of neu- at those sites. In our experiments, agrin RNA was injected ronal agrin on the surface of the axons. The punctate at the one-cell stage, and regions of embryos expressing GFP distribution of rat agrin in injected embryos (Fig. 3) is in myotomes or in the spinal cord were selected for analy- consistent with the findings in culture; it was seen in both sis. No apparent differences in the pattern or area of AChR innervated regions at the ends of myotomes (e.g., Fig. 3) and aggregates were seen when GFP was expressed primarily in also scattered throughout the myotome in some embryos spinal cord or in muscle. Since injected agrin RNA was (not shown). Puncta containing agrin could result from the present in muscle cells and in the spinal cord, the synaptic assembly of extracellular matrix molecules into particulate localization of induced AChR aggregates may reflect greater complexes, similar to those seen in early embryonic chick stability or translation efficiency of rat agrin RNA in

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved. 30 Godfrey, Roe, and Heathcote neurons than in muscle cells, and/or targeting or clustering valuable reagents, technical suggestions, and encouragement. of muscle agrin in synaptic regions of these cells, as We thank Drs. Monroe Cohen and Wesley Thompson for care- discussed above. However, these parameters were not ad- fully and critically reading the manuscript, Carolyn Snyder for dressed in our experiments. preparing Fig. 1, and Deb Generotzky for photography. This In some experiments, overexpression of agrin resulted in work was funded by the National Institute of Mental Health, NIH (1R01MH57545, E.W.G. and R.D.H.) and the Muscular an increase in AChR aggregation away from the ends of the Dystrophy Association (E.W.G.). myotomes, in regions of muscle cells that are not normally innervated. We assume that these ectopic aggregates re- sulted from increased secretion of agrin by muscle cells to levels normally found only at the synapse. While we saw rat REFERENCES agrin immunofluorescence concentrated in the innervated septal regions of myotomes in embryos overexpressing Alder, J., Kanki, H., Valtorta, F., Greengard, P., and Poo, M. (1995). agrin (Fig. 3), we also observed punctate deposits of rat agrin Overexpression of enhances se- throughout the myotomes of some injected embryos. How- cretion at Xenopus neuromuscular synapses. J. Neurosci. 15, ever, we did not determine whether this “extrasynaptic” 511–519. agrin was colocalized with AChR aggregates. While we Apel, E. D., Roberds, S. L., Campbell, K. P., and Merlie, J. P. (1995). cannot rule out the possibility that agrin overexpression Rapsyn may function as a link between the acetylcholine recep- stimulated the growth of motor axons into the myotomes, tor and the agrin-binding dystrophin-associated glycoprotein. and that the extrasynaptic aggregates were induced by agrin Neuron 15, 115–126. presented by ectopic nerve terminals, we consider this Apel, E. D., Glass, D. J., Moscoso, L. M., Yancopolous, G. D., and possibility unlikely. Since eliminating agrin expression Sanes, J. R. (1997). Rapsyn is required for MuSK signaling and recruits synaptic components to a MuSK-containing scaffold. stimulates axonal overgrowth in muscle (Gautam et al., Neuron 18, 623–635. 1996), we would expect agrin overexpression to inhibit the Bassez, T., Paris, J., Omilli, F., Dorel, C., and Osborne, H. B. (1990). growth of motor axons, consistent with agrin’s inhibitory Post-transcriptional regulation of ornithine decarboxylase in effect on neurite outgrowth in culture (Chang et al., 1997). Xenopus laevis oocytes. Development 110, 955–962. Ben Aziz-Aloya, R., Seidman, S., Timberg, R., Sternfeld, M., Zakut, General Applicability of RNA Injection for H., and Soreq, H. (1993). Expression of a human acetylcholines- Overexpression of Synaptic Proteins terase promoter–reporter construct in developing neuromuscular junctions of Xenopus embryos. Proc. Natl. Acad. Sci. USA 90, Overexpression in Xenopus embryos has been used to test 2471–2475. the role of specific proteins in function and development of Campanelli, J. T., Gayer, G. G., and Scheller, R. H. (1996). Alter- cellular mechanisms involved in synaptic transmission at native RNA splicing that determines agrin activity regulates ␣ the neuromuscular junction. Synaptic proteins which were binding to heparin and -. Development 122, 1663– overexpressed by injection of either DNA or RNA into early 1672. Chang, D., Woo, J. S., Campanelli, J., Scheller, R. H., and Ignatius, embryos include acetylcholinesterase (AChE; Ben Aziz- M. J. (1997). Agrin inhibits neurite outgrowth but promotes Aloya et al., 1993; Seidman et al., 1994; Shapira et al., attachment of embryonic motor and sensory neurons. Dev. Biol. 1994), acetylcholine receptors (Shapira et al., 1994), synap- 181, 21–35. sin I (Lu et al., 1992), IIa (Schaeffer et al., 1994), Chomczynski, P., and Sacchi, N. (1987). Single-step method of and synaptophysin (Alder et al., 1995). These studies sug- RNA isolation by acid guanidinium thiocyanate–phenol– gested that some of these proteins may be rate-limiting in chloroform extraction. Anal. Biochem. 162, 156–159. the formation of functional embryonic synapses. Overex- Chow, I., and Cohen, M. W. (1983). Developmental changes in the pression has not previously been used in Xenopus embryos distribution of acetylcholine receptors in the myotomes of Xe- to evaluate molecules believed to be involved in directing nopus laevis. J. Physiol. London 339, 553–571. the differentiation of the neuromuscular junction. In this Cohen, I., Rimer, M., Lomo, T., and McMahan, U. J. (1997). Agrin-induced postsynaptic-like apparatus in skeletal muscle study, we have overexpressed agrin, a protein essential for fibers in vivo. Mol. Cell. Neurosci. 9, 237–253. organization of this synapse. The approach we have out- Cohen, M. W., and Godfrey, E. W. (1992). Early appearance of and lined can be applied to testing additional hypotheses about neuronal contribution to agrin-like molecules at embryonic frog the role of specific proteins, their isoforms, and domains in nerve–muscle synapses formed in culture. J. Neurosci. 12, 2982– embryonic synaptogenesis. 2992. DeChiara, T. M., Bowen, D. C., Valenzuela, D. M., Simmons, M. V., Poueymirou, W. T., Thomas, S., Kinetz, E., Compton, ACKNOWLEDGMENTS D. L., Rojas, E., Park, J. S., Smith, C., DiStefano, P. S., Glass, D. J., Burden, S. J., and Yancopoulos, G. D. (1996). The receptor We thank Haig Tcheurekdjian, Daniel Byrne, Peter Schmidt, tyrosine kinase MuSK is required for neuromuscular junction Sheila Baker, Jon Ekman, and Sarah Rickert for their assistance formation in vivo. Cell 85, 501–512. in developing the methods and performing experiments in this Denzer, A. J., Brandenburger, R., Gesemann, M., Chiquet, M., and study. Drs. Richard Dorsky, Janice Sugiyama, Paul Krieg, Mon- Ruegg, M. A. (1997). Agrin binds to the nerve–muscle basal roe Cohen, and Linda Gont and Mr. Brett Schroeder provided lamina via . J. Cell Biol. 137, 671–683.

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved. Agrin Isoforms Alter Embryonic Synapse Formation 31

Ferns, M., Hoch, W., Campanelli, J. T., Rupp, R., Hall, Z. W., and clustering are overlapping but not identical with the Scheller, R. H. (1992). RNA splicing regulates agrin-mediated ␣-dystroglycan-binding region. J. Biol. Chem. 271, 5231–5236. acetylcholine receptor clustering activity on cultured myotubes. Jones, G., Meier, T., Lichtsteiner, M., Witzemann, V., Sakmann, B., Neuron 8, 1079–1086. and Brenner, H. R. (1997). Induction by agrin of ectopic and Ferns, M. J., Campanelli, J. T., Hoch, W., Scheller, R. H., and Hall, functional postsynaptic-like membrane in innervated muscle. Z. (1993). The ability of agrin to cluster AChRs depends on Proc. Natl. Acad. Sci. USA 94, 2654–2659. and on cell surface proteoglycans. Neuron 11, Kullberg, R. W., Lentz, T. L., and Cohen, M. W. (1977). Develop- 491–502. ment of the myotomal neuromuscular junction in Xenopus Fuhrer, C., Sugiyama, J. E., Taylor, R. G., and Hall, Z. W. (1997). laevis: An electrophysiological and fine-structural study. Dev. Association of muscle-specific kinase MuSK with the acetylcho- Biol. 60, 101–129. line receptor in mammalian muscle. EMBO J. 16, 4951–4960. Lieth, E., Cardasis, C., and Fallon, J. R. (1992). Muscle-derived Gautam, M., Noakes, P. G., Mudd, J., Nichol, M., Chu, G. C., agrin in cultured myotubes: Expression in the basal lamina Sanes, J. R., and Merlie, J. P. (1995). Failure of postsynaptic and at induced acetylcholine receptor clusters. Dev. Biol. 149, specialization to develop at neuromuscular junctions of rapsyn- 41–52. deficient mice. Nature 377, 232–236. Lieth, E., and Fallon, J. R. (1993). Muscle agrin: Neural regulation Gautam, M., Noakes, P. G., Moscoso, L., Rupp, F., Scheller, and localization at nerve-induced acetylcholine-receptor clus- R. H., Merlie, J. P., and Sanes, J. R. (1996). Defective neuro- ters. J. Neurosci. 13, 2509–2514. muscular synaptogenesis in agrin-deficient mutant mice. Cell Lu, B., Greengard, P., and Poo, M.-M. (1992). Exogenous 85, 525–535. promotes functional maturation of developing neuromuscular Gesemann, M., Denzer, A. J., and Ruegg, M. A. (1995). Acetylcho- synapses. Neuron 8, 521–529. line receptor-aggregating activity of agrin isoforms and mapping Ma, E., Morgan, R. M., and Godfrey, E. W. (1994). Distribution of of the active site. J. Cell Biol. 128, 625–636. agrin mRNAs in the chick embryo nervous system. J. Neurosci. Gesemann, M., Cavalli, V., Denzer, A. J., Brancaccio, A., Schuma- 14, 2943–2952. cher, B., and Ruegg, M. A. (1996). Alternative splicing of agrin Ma, E., Morgan, R. M., and Godfrey, E. W. (1995). Agrin mRNA variants are differentially regulated in developing chick embryo alters its binding to heparin, dystroglycan, and the putative agrin spinal cord and sensory ganglia. J. Neurobiol. 26, 585–597. receptor. Neuron 16, 755–767. Magill-Solc, C., and McMahan, U. J. (1988). Motor neurons contain Godfrey, E. W. (1991). Comparison of agrin-like proteins from agrin-like molecules. J. Cell Biol. 107, 1825–1833. the extracellular matrix of chicken kidney and muscle with Mjaatvedt, C. H., Lepera, R. C., and Markwald, R. R. (1987). neural agrin, a synapse organizing protein. Exp. Cell Res. 195, Myocardial specificity for initiating endothelial–mesenchymal 99–109. cell transition in embryonic chick heart correlates with a par- Godfrey, E. W., and Gradall, K. S. (1998). Basal lamina components ticulate distribution of fibronectin. Dev. Biol. 119, 59–67. are concentrated in myogenic regions of the mouse limb bud. Moon, R. T., and Christian, J. L. (1989). Microinjection and expres- Anat. Embryol. 198, 481–486. sion of synthetic mRNAs in Xenopus embryos. Technique 1, Godfrey, E. W., Nitkin, R. M., Wallace, B. G., Rubin, L. L., and 76–89. McMahan, U. J. (1984). Components of Torpedo electric organ Nieuwkoop, P. D., and Faber, J. (1994). “Normal Table of Xenopus and muscle that cause aggregation of acetylcholine receptors on laevis (Daudin)” (originally published in 1967 [2nd ed.], North cultured muscle cells. J. Cell Biol. 99, 615–627. Holland, Amsterdam). Garland, New York and London. Godfrey, E. W., Dietz, M. E., Morstad, A. L., Wallskog, P. A., and O’Toole, J. J., Deyst, K. A., Bowe, M. A., Nastuk, M. A., McKech- Yorde, D. E. (1988a). Acetylcholine receptor-aggregating proteins nie, B. A., and Fallon, J. R. (1996). Alternative splicing of agrin are associated with the extracellular matrix of many tissues in regulates its binding to heparin, ␣-dystroglycan, and the cell Torpedo. J. Cell Biol. 106, 1263–1272. surface. Proc. Natl. Acad. Sci. USA 93, 7369–7374. Godfrey, E. W., Siebenlist, R. E., Wallskog, P. A., Walters, L. M., Pun, S., and Tsim, K. W. K. (1997). Antisense agrin cDNA trans- Bolender, D. L., and Yorde, D. E. (1988b). Basal lamina compo- fection blocks neuroblastoma cell-induced acetylcholine recep- nents are concentrated in premuscle masses and at early acetyl- tor aggregation when co-cultured with myotubes. Mol. Cell. choline receptor clusters in chick embryo hindlimb muscles. Neurosci. 10, 87–99. Dev. Biol. 130, 471–486. Reist, N. E., Magill, C., and McMahan, U. J. (1987). Agrin-like Gove, C., Walmsley, M., Nijjar, S., Bertwistle, D., Guille, M., molecules at synaptic sites in normal, denervated and damaged Partington, G., Bomford, A., and Patient, R. (1997). Over- skeletal muscles. J. Cell Biol. 105, 2457–2469. expression of GATA-6 in Xenopus embryos blocks differentia- Ruegg, M. A., Tsim, K. W. K., Horton, S. E., Kroeger, S., Escher, G., tion of heart precursors. EMBO J. 16, 355–368. Gensch, E. M., and McMahan, U. J. (1992). The agrin gene codes Gurdon, J. B., and Wickens, M. P. (1983). The use of Xenopus for a family of basal lamina proteins that differ in function and oocytes for the expression of cloned genes. Methods Enzymol. distribution. Neuron 8, 1–20. 101, 370–386. Rupp, F., Payan, D. G., Magill-Solc, C., Cowan, D. G., and Scheller, Hoch, W., Ferns, M., Campanelli, J. T., Hall, Z. W., and Scheller, R. H. (1991). Structure and expression of a rat agrin. Neuron 6, R. H. (1993). Developmental regulation of highly active alterna- 811–823. tively spliced forms of agrin. Neuron 11, 479–490. Rupp, R. A. W., Sinder, L., and Weintraub, H. (1994). Xenopus Hoch, W., Campanelli, J. T., Harrison, S., and Scheller, R. H. (1994). embryos regulate the nuclear localization of XmyoD. Genes Dev. Structural domains of agrin required for clustering of nicotinic 8, 1311–1323. acetylcholine receptors. EMBO J. 13, 2814–2821. Schaeffer, E., Alder, J., Greengard, P., and Poo, M. M. (1994). Hopf, C., and Hoch, W. (1996). Agrin-binding to ␣-dystroglycan: Synapsin IIa accelerates functional development of neuromuscu- Domains of agrin necessary to induce acetylcholine receptor- lar synapses. Proc. Nat. Acad. Sci. USA 91, 3882–3886.

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved. 32 Godfrey, Roe, and Heathcote

Seidman, S., Ben Aziz-Aloya, R., Timberg, R., Loewenstein, Y., Turner, D. L., and Weintraub, H. (1994). Expression of achaete- Velan, B., Shafferman, A., Liao, J., Norgaard-Pedersen, B., Brod- scute homolog 3 in Xenopus embryos converts ectodermal cells beck, U., and Soreq, H. (1994). Overexpressed monomeric human to a neural fate. Genes Dev. 8, 1434–1447. acetylcholinesterase induces subtle ultrastructural modifica- Valnes, K., and Brandtzaeg, P. (1985). Retardation of immunofluo- tions in developing neuromuscular junctions of Xenopus laevis rescence fading during microscopy. J. Histochem. Cytochem. 33, 755–761. embryos. J. Neurochem. 62, 1670–1681. Vize, P. D., Melton, D. A., Hemmati-Brivanlou, A., and Harland, Shapira, M., Seidman, S., Sternfeld, M., Timberg, R., Kaufer, D., R. M. (1991). Assays for gene function in developing Xenopus Patrick, J., and Soreq, H. (1994). Transgenic engineering of embryos. In “Methods in Cell Biology” (B. K. Kay and H. B. neuromuscular junctions in Xenopus laevis embryos transiently Peng, Eds.), Vol. 36, pp. 367–387. Academic Press, San overexpressing key cholinergic proteins. Proc. Natl. Acad. Sci. Diego. USA 91, 9072–9076. Zar, J. H. (1974). “Biostatistical Analysis.” Prentice-Hall, Engle- Smith, M. A., and O’Dowd, D. K. (1994). Cell-specific regulation of wood Cliffs, NJ. agrin RNA splicing in the chick ciliary ganglion. Neuron 12, 795–804. Received for publication August 5, 1998 Tsim, K. W. K., Ruegg, M. A., Escher, G., Kro¨ger, S., and McMahan, Revised October 5, 1998 U. J. (1992). cDNA that encodes active agrin. Neuron 8, 677–689. Accepted October 6, 1998

Copyright © 1999 by Academic Press. All rights of reproduction in any form reserved.