We have used recombinant retroviruses as lineage markers to study the genealogy of motoneurons (MNs) in the chick spinal cord. We infected individual progenitors by injecting virions into the neural tube at stages 11-18, a few cell divisions before MNs are born. The descendants of infected cells were subsequently detected with a histochemical stain for beta-galactosidase (lacZ), the product of the retrovirally introduced gene. Clonally related, lacZ-positive cells formed clusters that were usually radial or planar in shape. The cells that comprised these clones were classified by morphology, size, and location. About 15% of the clones in the spinal cord contained MNs, and these were studied further. Multicellular clones that contained only MNs were infrequent. Instead, close relatives of MNs included a variety of other neurons, as well as glia and ependymal cells. Most non-MNs in these clones were found in the ventral and intermediate parts of the spinal cord. Neurons included interneurons and autonomic preganglionic neurons in the column of Terni. Labeled glia were found in both the gray and white matter and included astrocytes and cells tentatively identified as oligodendrocytes. Thus, even shortly before MNs are born, their progenitors are multipotential. Clonally related MNs were not restricted to a single motor pool. Some clones contained MNs in both the medial and lateral parts of the lateral motor column, which are known to innervate distinct groups of limb muscles. Furthermore, some clones contained MNs in the medial motor column (which innervate axial muscles) as well as in the lateral motor column. In contrast, the dispersal of clonally related MNs (and other neurons) was restricted in the rostrocaudal axis; most clones were less than one-quarter segment in length. Thus, MNs derived from a single progenitor are more likely to share rostrocaudal position than synaptic targets. To investigate the fate of clonally related MNs, we counted the number of MNs per clone at times before, during, and after the major period of MN death. The number of MNs per clone declined in precise parallel with the total number of MNs during this period, suggesting that neurons are eliminated without regard to their clone of origin. This result implies that the decision to live or die occurs on a cell-by-cell rather than a clone-by-clone basis.

Developing motor axons induce synaptic specializations in muscle fibers, including preferential transcription of acetylcholine receptor (AChR) subunit genes by subsynaptic nuclei. One candidate nerve-derived signaling molecule is AChR-inducing activity (ARIA)/heregulin, a ligand of the erbB family of receptor tyrosine kinases. Here, we asked whether ARIA and erbB kinases are expressed in patterns compatible with their proposed signaling roles. In developing muscle, ARIA was present not only at synaptic sites, but also in extrasynaptic regions of the muscle fiber. ARIA was synthesized, rather than merely taken up, by muscle cells, as indicated by the presence of ARIA mRNA in muscle and of ARIA protein in a clonal muscle cell line. ARIA-responsive myotubes expressed both erbB2 and erbB3, but little EGFR/erbB1 or erbB4. In adults, erbB2 and erbB3 were localized to the postsynaptic membrane. ErbB3 was restricted to the postsynaptic membrane perinatally, at a time when ARIA was still broadly distributed. Thus, our data are consistent with a model in which ARIA interacts with erbB kinases on the muscle cell surface to provide a local signal that induces synaptic expression of AChR genes. However, much of the ARIA is produced by muscle, not nerve, and the spatially restricted response may result from the localization of erbB kinases as well as of ARIA. Finally, we show that erbB3 is not concentrated at synaptic sites in mutant mice that lack rapsyn, a cytoskeletal protein required for AChR clustering, suggesting that pathways for synaptic AChR expression and clustering interact.

We have studied the segmental innervation of 2 rat skeletal muscles, the diaphragm and the serratus anterior. Both muscles are thin, flat, and composed of several sectors that form a clear rostrocaudal progression. Each is innervated through a single nerve, which is in turn supplied by motor neurons from several cervical spinal segments. Using intracellular recording, we found that in both cases, the rostrocaudal axis of the motor pool is systematically mapped onto the rostrocaudal axis of the muscle's surface. For the diaphragm, electrophysiological results were confirmed by immunohistochemical identification of denervated fibers following section of single ventral roots and by retrograde labeling of motoneurons following localized application of fluorescent dyes. In addition, an immunohistochemical method was used to study the arrangement of motor axons in the phrenic nerve, which supplies the diaphragm, and to show that contributions from individual ventral roots are compartmentalized within this nerve. We suggest that segmental ordering of axons in the nerve, axonal guidance at branch points in the nerve, and positional labels within the muscle may all contribute to the rostrocaudal mapping of motor pools onto muscle.

To study the migration of chick spinal cord neurons, we labeled individual cells in the ventricular zone with recombinant retroviruses, then identified their progeny histochemically. First, we analyzed cell mixing in the ventricular zone. Some clones labeled at early neural tube stages spread widely along both the dorsoventral and rostrocaudal axes. However, clones labeled later were confined to narrow domains along both axes. These results imply that displacement of cells within the ventricular zone becomes progressively restricted. Second, we studied the migration of cells out of the ventricular zone by infecting embryos at a fixed stage and varying the time of analysis. At first, most clones consisted of radial arrays of cells, suggesting that the initial migration is predominantly radial. In many clones, however, neurons turned orthogonally from parental radial arrays and migrated along the path of circumferentially oriented axons. By hatching, clonally related cells in the gray matter were usually distributed in narrow transverse slabs, but some white matter glial cells had migrated longitudinally for up to several segments. We conclude that the dispersal of clonally related cells results from (1) early mixing of progenitors within the neural tube; (2) radial stacking of progeny in the ventricular zone; (3) migration of progeny from the ventricular zone in spoke-like routes; (4) circumferential migration of some neurons along axons; (5) short-distance dispersal of differentiating neurons; and (6) a late, longitudinal migration of glia through white matter tracts. Finally, we show that floor plate cells differ from other spinal cord cells in both their

We recently generated and characterized transgenic mice in which regulatory sequences from a myosin light chain gene (MLC1f/3f) are linked to the chloramphenicol acetyltransferase (CAT) gene. Transgene expression in these mice is specific to skeletal muscle and graded along the rostrocaudal axis: adult muscles derived from successively more caudal somites express successively higher levels of CAT. To investigate the cellular basis of these patterns of expression, we developed and used a histochemical stain that allows detection of CAT in individual cells. Our main results are as follows: (a) Within muscles, CAT is detected only in muscle fibers and not in associated connective tissue, blood vessels, or nerves. Thus, the tissue specificity of transgene expression observed by biochemical assay reflects a cell-type specificity demonstrable histochemically. (b) Within individual muscles, CAT levels vary with fiber type. Like the endogenous MLC1f/3f gene, the transgene is expressed at higher levels in fast-twitch (type II) than in slow-twitch (type I) muscle fibers. In addition, CAT levels vary among type II fiber subtypes, in the order IIB greater than IIX greater than IIA. (c) Among muscles that are similar in fiber type composition, the average level of CAT per fiber varies with rostrocaudal position. This position-dependent variation in CAT level is apparent even when fibers of a single type are compared. From these results, we conclude that fiber type and position affect CAT expression independently. We therefore infer the existence of separate fiber type-specific and positionally graded transcriptional regulators that act together to determine levels of transgene expression.