Development of Segment- and Target-Related Neuronal Identity in the Medicinal Leech

Neurons adopt several strategies to establish their connectivity patterns, ranging from relatively simple choices of defined substrata (Bastiani and Goodman, 1984; Berlot and Goodman, 1984; Ho and Goodman, 1982; Jellies and Kristan, 1988b) to exuberant outgrowth that is modified secondarily as the circuits become functional (Wolszon and Macagno, 1992). Growth cones appear to be specified to respond in particular ways to different cues in the environment, giving rise to stereotyped nerve pathways and patterns of synaptic connectivity. Vertebrate motor neuron growth cones, for example, respond to different environmental features in highly specific ways. There are permissive and general pathways that can be followed by many axonal growth cones, while subsets of growth cones respond to more specific local cues to reach appropriate targets even when the targets are displaced (Lance-Jones and Landmesser, 1981; Landmesser, 1984; Tosney and Landmesser, 1984, 1985). Similar examples of pathfinding and growth cone migration have been found elsewhere in vertebrate development, including other motor systems (Eisen and Pike, 1992; Eisen et al. 1989; Myers et al. 1986; Westerfield et al. 1986), sensory systems (Bastmeyer et al. 1990; Godement et al. 1990; O’Rourke and Fraser, 1990; Scott, 1988; Stuermer, 1988) and central nervous system (CNS) neurons (Easter et al. 1992; Kuwada and Bernhardt, 1990; Kuwada et al. 1990; Wilson and Easter, 1991).

One common developmental strategy involves the initial projection of multiple processes into several target fields, including the correct one, with secondary retraction of the inappropriate projections and/or stabilization of the appropriate ones (Cowan et al. 1984). The initial projection of many axons has been interpreted as a means of ensuring adequate innervation of the available target (Purves, 1988). However, the mechanisms driving this developmental plasticity are not well understood. Despite advances in cellular development, there are few systems in which neurons of known behavioral relevance and synaptic connectivity can be examined with both anatomical and physiological techniques during normal embryogenesis.

The leech has proved useful for studies on cellular development in general (Baptista et al. 1990; Blair et al. 1990; Blair and Weisblat, 1982, 1984; Gao and Macagno, 1988; Jellies, 1990; Jellies and Kristan, 1988a,b; Jellies et al. 1987; Johansen et al. 1984, 1985; Loer et al. 1987; Loer and Kristan, 1989a,b; Macagno et al. 1983, 1990; Macagno and Stewart, 1987; Shankland, 1984, 1987, 1991; Shankland and Martindale, 1989b; Stuart et al. 1989; Torrence et al. 1989; Torrence and Stuart, 1986). In large part this success is related to the ease with which individual neurons can be identified and their role in behavior assessed (Muller et al. 1981). One particularly tractable system for analysis is the neurogenic heart.

The leech has a closed circulatory system that uses bilaterally paired muscular heart tubes to propel the blood (Stent et al. 1979). The heart excitor motor neurons (HEs) are readily identifiable, occurring as a bilaterally symmetrical pair with their somata on the ventral surface of the ganglion and with a novel axonal projection for neurons in this location; they send an axon ipsilaterally out of the anterior root to innervate the lateral heart tubes (Maranto and Calabrese, 1984,a; Thompson and Stent, 1976). HEs have a stereotyped central morphology and are segmentally iterated in midbody segments 3-18 (Shafer and Calabrese, 1981; Thompson and Stent, 1976), but have not been identified in the first two or the last three midbody segments. Although the heart tubes have an intrinsic myogenic rhythm, their functional rhythmic contractions are superimposed on them by a central pattern generator (CPG), which modulates their tonic activity by regular periodic inhibition from identified heart interneurons (HNs) residing in segments 1-7 (Maranto and Calabrese, 1984a,b; Thompson and Stent, 1976; Tolbert and Calabrese, 1985). Furthermore, the HEs are cholinergic (Calabrese and Maranto, 1986) and contain and use FMRFamide (or something virtually indistinguishable from it) as a modulator (Calabrese and Norris, 1989; Evans and Calabrese, 1989; Kuhlman et al. 1985a,b).

The present study represents our initial efforts to examine the normal development of HEs using intracellular recording and dye-filling. We present evidence that HEs transiently grow into both the appropriate target region (heart tube) and other body wall territories and that they subsequently retract their incorrect projections as their correct peripheral arborization expands. Furthermore, we show that this anatomical remodelling occurs well before the emergence of the centrally driven pattern of rhythmic activity. Additionally, we present evidence that there appear to be segmental homologs of HEs in the first two and last three midbody segments and that these homologs are initially similar to definitive HEs in their peripheral projections, but their peripheral arborizations neither retract during embryogenesis nor become rhythmically driven by central input. Some of these results have appeared previously in abstract form (Jellies and Kopp, 1991).

The leech is an annelid composed of 32 segments. The tubular body wall of the adult is composed of four discrete layers: an outer epidermal layer, circumferential (circular) muscles, oblique muscles and an inner layer of longitudinal muscles (Sawyer, 1986). The central nervous system is composed of a ventral chain of ganglia, one ganglion per segment, each containing about 200 different neurons (Macagno, 1980). All the ganglia are similar, and many of the neurons have been uniquely identified (Muller, 1979; Ort et al. 1974; Stuart, 1976). Leeches, Hirudo medicinalis L., were obtained from a breeding colony at the University of Alabama at Birmingham or from a commercial supplier (Leeches USA, New York). Breeding, maintenance and staging were carried out as previously described (Fernández and Stent, 1982; Jellies et al. 1987) at 22-24°C, except that embryos were maintained in non-sterile stock solutions of 0. 0005% sea salt (w/w) (Tropic Marin, Dr Biener GMBH Aquarientechnik, Wartenberg, Germany) in 0.001 mol l⁻¹ Hepes buffer (pH 7.4) to which was added sufficient penicillin/streptomycin (Sigma) for a final dilution of 1000 i.u. mP¹ (penicillin) and 1000 mg ml⁻¹ (streptomycin) just before use.

Cells in dissected germinal plates or ganglia were intracellularly filled with Lucifer Yellow (LY) in a 3% aqueous solution or horseradish peroxidase (HRP) as described previously (Jellies and Kristan, 1991; Jellies et al. 1987). Dye-filled cells were drawn using a camera lucida.

Intracellular recordings were made using LY-filled (Stewart, 1978) micropipettes (approximately 150 MQ) that had been pulled on a Flaming-Brown puller. Signals were amplified using a preamplifier (Getting Instrument Company) and displayed on a storage oscilloscope (Tektronix). Traces were photographed directly from the oscilloscope screen using polaroid film. A comparable semiintact preparation retaining the body wall from the head to midbody segment 5 or 6 was employed for embryos, juveniles and adults. After cutting the dorsal midline in animals anesthetized in 8% ethanol/saline (embryos) or by cold (older animals) and removing gut contents, the leech was pinned skin-side upwards and the body wall and blood sinus opened at the ventral midline to expose the CNS. For mechanical stability (the embryo undergoes constant muscular contractions) it proved necessary to draw the cut edges of the ventral body wall back towards the midline with a set of fine wire pins. Leaving uncut body wall between ganglia in embryos resulted in mechanically unstable preparations. Preparations were pinned in the recording chamber (a Sylgard-coated slide for embryos) in normal saline. Although body wall contractions resumed within seconds, preparations were allowed to recover for at least 10 min before intracellular recordings were made. After recording, the cells were filled with LYas above. 15 embryonic and 25 mature motor neurons were successfully identified by intracellular recording and subsequent dye filling.

For immunostaining, leech embryos or the dissected ganglia (with or without LY-filled neurons) were fixed overnight at 4°C in 4% paraformaldehyde in 0.1 mol l⁻¹ phosphate buffer (pH 7.4). Rinsed germinal plates were incubated with a primary monoclonal antibody (mAb) directed against human fibronectin (Sigma) at 1:1000 in phosphate-buffered saline (PBS) with 5% goat serum, 0.5% Triton X-100 and 0.05% sodium azide (all reagents from Sigma) overnight at 4°C with constant agitation. After rinsing in PBS, tissues were incubated in secondary antiserum overnight at 4°C (goat anti-mouse IgG, Boehringer Mannheim) conjugated to rhodamine and diluted (1:100-1:250) in the same buffer as the primary mAb. Preparations were then rinsed with PBS, dehydrated in ethanol, cleared in methyl salicylate and mounted in acrylic dissolved in toluene. Whole mounts were viewed using a Leitz compound microscope with epifluorescence and appropriate filter sets and photographed on Ektachrome color slide film (ASA200, Kodak).

HE motor neurons from juvenile and mature leeches (90 days to >1 year) were readily identifiable by their position in the ganglion, rhythmic membrane potential oscillations and morphology (Fig. 1). Embryonic HEs were found close to their adult positions. In E10-E16 (E=embryonic day) embryos they were among the largest somata on the ventral surface of the ganglion (Fig. 1), roughly comparable in size to the Retzius neurons. The only other neuron in the anterior packet that has an ipsilateral projection is cell 157 (Nusbaum, 1984); this cell was encountered twice and was easily distinguished from the HE by its position, size (it is smaller than the HE in embryos) and peripheral projection out of the posterior, rather than the anterior, root (not shown).

Early in development, each HE had an elaborate arborization within the ventral body wall (Fig. 2) in addition to the expected one in the vicinity of the nascent heart tube. Rather than being a random over-exuberance, this body wall arborization always projected roughly parallel to the heart tube arborization. The major difference in these arborizations was position, one being located more medially. This medial terminal field has not been described in mature animals by workers discussing HE anatomy in the periphery (Maranto and Calabrese, 1984a), and our intracellular preparations of juvenile and adult HEs consistently failed to reveal it even when dye had moved to the first bifurcation in the vascular nerve. Thus, branches of the mature HE in ventral body wall are rare and of small caliber if they are present at all, but the extent of peripheral branching of the mature HE remains to be examined in detail.

We examined staged embryos using intracellular dye-filling from E10 to E12 as well as E14 and juveniles, concentrating on anterior segments (1-4 and 7-9). The extraaxonal arborization medial to the nephridiopores was apparent at the earliest time examined, when the HE had also projected to the appropriate target area (Fig. 3). The extra projection was a robust arborization in each case, but it decreased in extent as the heart tube arborization expanded. However, once begun, this change was fairly rapid. Both arborizations appeared to grow in absolute size for several days with a dramatic change between E11 and E12 (Fig. 3). Each of the 17 E12 dye-filled HEs showed a prominent coverage of the heart tube with branches ramifying extensively in three dimensions over the heart tube, while the more ventral arborization seemed to have regressed.

These qualitative changes were confirmed by counting total branch points in each of the two fields at E10, E11 and E12. Branch points residing in a given field (medial or lateral defined by a line connecting nephridiopore centers) were counted regardless of their origin. Because we were interested in how the heart tube (lateral) innervation changed with respect to that of the body wall (medial), and in an effort to compensate for different qualities of dye-fills (HRP was used exclusively for this analysis and some dye-fills were more complete than others), branch points were expressed as the number of lateral ones divided by the number of medial ones (Fig. 3). At E10 both peripheral arborizations were comparable, with a slight bias towards more branch points in the medial field. Changes in this measure could come about by changes in either lateral or medial fields independent of one another. Indeed, by E11 both fields had expanded in absolute size (Fig. 3), and while there was a slight increase in branch points in the lateral relative to the medial field (ordinate value approximately 2), this difference was not statistically significant (Student’s t-test, P>0.01). There was a dramatic change between E11 and E12 because of continued expansion by the lateral field as well as retraction of most branches in the medial field. The increase at E12 was statistically different from the measurements at earlier stages (Student’s t-test, P≤0.00l).

Although older (and larger) cells were indeed more poorly dye-filled in our studies, it seems likely that this progression represents a true retraction of processes rather than incomplete dye-filling since the more medial processes, if present, should have been more completely revealed by our injections into the soma. The loss of branches in the medial field was confirmed by both HRP and LY injections into E14 embryos and juveniles where the HE axon could be followed laterally to the heart tube without branching (with the exception of a single E14 HE that had two relatively small processes in the medial field). Since relatively short survival times were used in these studies (about 1 h for HRP), dye-filling was not extensive and these older preparations were, therefore, not included in the quantification of branch points. Furthermore, as the number of medial branch points approaches zero the relative measure used here becomes inadequate (undefined).

Mature HEs have a characteristic central arborization that rarely expands extensively across the midline of the ganglion but does project some contralateral dendrites (Thompson and Stent, 1976; Tolbert and Calabrese, 1985) and contains a prominent tuft of posterior-projecting dendrites. In contrast to the peripheral arborizations, these central features seemed to develop in a straightforward fashion, with embryonic HEs having small arborizations that mirrored the adult form (Figs 3, 4, 5). These central projections were not quantified in the present study.

Another indication of plasticity in HE peripheral arborizations (and the only obvious central process variation) was the presence of a small-caliber axon projecting out of the posterior root in addition to the axon projecting out of the anterior root (Fig. 4). This extra axon arose from one of the 2-3 major posterior-projecting neurites near the bend in the cell that directs the major axon out the anterior root. We found that about 34% (12 out of 35) of the cells had these projections (considering only those for E11 and E12 used for the quantification in Fig. 3). This is in contrast to such axons which are found on adult HEs much more rarely, in only about 4% of the neurons (Shafer and Calabrese, 1981). In embryos, these extra axons conformed to one of two configurations in the periphery. They either established a highly branched arborization in the posterior/ventral territory (five cells) (Fig. 4A) or projected unbranched within the dorsal-posterior nerve beyond the level of the heart tube (seven cells) (Fig. 4B). Furthermore, in each of the five cases where the axon branched, there was no major posterior-projecting branch from the anterior root in the ventral territory.

The membrane potential of embryonic HEs was typically about –35 mV. Since the young cells were easily damaged by impalement (most impalements resulted in a burst of action potentials followed by a slow depolarization) such voltage measurements can only be considered approximate. HEs that depolarized to about –20 mV were considered to be badly damaged and not examined further. At E23 we obtained electrical recordings from four HEs in segments 3 and 4 (from three different animals) that were stable for 10-15 min. E23 HEs all exhibited spontaneous rhythmicity resembling that of the adult (Fig. 5) with 3-5 s periods of prominent inhibitory postsynaptic potentials (IPSPs) throughout, but concentrated in the episodes of reduced activity.

We also recorded spontaneous electrical activity from younger embryos (Fig. 5). In this case, recordings were obtained from six HEs in segments 3 and 4 (from six different animals) at E16. In two cases (from the same group of siblings), HE activity was predominantly tonic and there were occasional prominent IPSPs (not shown). Periods of activity were broken by irregular silent periods of 5-15 s during which there were still isolated IPSPs but no barrage sufficient to alter the resting potential overtly. Activity in the remaining E16 HEs appeared to be more intermediate between tonic and rhythmic (Fig. 5). The activity in these cells was robust, they exhibited prominent IPSPs and also silent periods associated with bursts of IPSPs that served to hyperpolarize the membrane potential (Fig. 5). The activity in these cells could not be classified as rhythmic in the sense that, over the relatively short duration of the recordings in this study, there was no apparent regularity in periodicity. Interestingly, this latter group of four HEs was from the same sibling group, which was different from that of the previous two E16 HEs.

In adult Hirudo medicinalis, there are no HE motor neurons in the extreme anterior or posterior midbody segments (Shafer and Calabrese, 1981; Thompson and Stent, 1976). This is based upon the functional definition of these motor neurons, in that there are no cells in these ganglia driven by HNs to be rhythmically active. We discovered that there might be HE homologs in these segments. In the absence of direct lineage studies in Hirudo, however, this determination cannot be made unambiguously. With this caveat, these cells will be tentatively called HE homologs. These HE homologs share a common (fibronectin) epitope with the definitive HEs in segments 3-18 (Fig. 6). Double-labels of late E12 embryos using the mAb and LY injection into HEs demonstrated that the HEs and HE homologs are the only neuronal somata that present this epitope at a level that can be visualized above background. A high rhodamine background was, however, a persistent problem and so we cannot exclude the possibility that other neurons also express the epitope at levels below background. No claim about the identity of the epitope or functional consequence is intended. These homologs were found in midbody segments 1 and 2 as well as 19, 20 and 21; we have concentrated on the anterior ones. The mAb also labelled all muscle cells in all layers, including those within the CNS (Tulsi and Coggeshall, 1971), the body wall and the developing heart tubes (not shown), and persisted on HE somata throughout embryogenesis and into the juvenile stage when HEs could be positively identified by their rhythmic activity. Because of high background fluorescence, we have not yet determined unambiguously whether the fibronectin epitope persists on older (adult) HEs.

We succeeded in identifying these homologs in mature animals and juveniles and, as predicted, they showed no spontaneous rhythmic activity (Fig. 7). The activity of these homologs was tonic with occasional prominent IPSPs. We do not yet know the source of this input, but there was no evidence of rhythmicity, either in action potential activity or in the IPSPs, in the relatively short recordings made in this study. Anatomically, the homologs were located in the same position as definitive HEs, had a relatively large soma and an ipsilateral projection out into the anterior root (Fig. 7).

During the early stages of embryogenesis the HE homologs could not easily be distinguished from definitive HEs. They project peripheral arborizations to the developing heart tube primordium and also elaborate a more medial arborization roughly parallel to the lateral one (Fig. 8). These homologs expressed a very different growth strategy from that of the definitive HEs, elaborating both lateral and medial arborizations equally over the same time that definitive HEs were pursuing an overgrowth/retraction strategy (Fig. 8). This was quantified in a similar way as before, this time extrapolating the line through nephridiopores into segment 1 (well-defined nephridiopores begin in segment 2). Both the lateral and medial arborizations maintained the same number of branch points throughout the period examined; there was no statistically significant difference between any of the three ages (Student’s t-test, P>0.01). This means that each of the 19 cells filled with HRP in segments 1 and 2 at E10, E11 and E12 had roughly equivalent lateral and medial peripheral arborizations. This could be achieved by having both arborizations regress, both remain stable or both expand. We found that both arborizations appeared to expand (Fig. 8). At E10 and E11 there was no statistically significant difference between the branch point measure for definitive HEs and that for HE homologs (Student’s t-test, P>0.01), whereas the difference at E12 was statistically significant (Student’s t-test, P≤0.000l). We have not yet determined whether the homologs ever innervate the heart tube or whether they retain both extensive peripheral arborizations in the adult. However, in contrast to the definitive HEs, a limited number of dye-filled HE homologs at E14 revealed both medial and lateral arborizations present at that time (not shown).

The neuropilar arborizations of HE homologs were qualitatively indistinguishable from those of definitive HEs, having a prominent dendritic field projecting both medially and posteriorly with little extension across the midline (Figs 7 and 8). No effort was made to quantify these central arborizations in this study and it is not yet known whether there are more subtle differences in dendritic distribution. In contrast to the definitive HEs, we found no examples of homologs that projected an extra axon out of the posterior root.

We obtained stable recordings from three HE homologs in segment 2 at E23 (three different animals). Each showed an activity pattern comparable to that in the adult, having robust tonic activity with occasional hyperpolarizing IPSPs (Fig. 9). Stable impalements of two HE homologs at E16 revealed the same pattern of activity and no evidence of regular periodicity (Fig. 9).

The present studies relate to several areas of investigation that are usually considered independently: the projection of axons via directed migration of growth cones, the selection of appropriate targets and subsequent maturation of synaptic connections, and the emergence of coordinated behavior. Presently unresolved is the issue of how interdependent these events are. For example, navigation of growth cones to a target region is, a priori, a necessary condition for generating a synapse between a particular neuron and its target, yet this could be done in a number of ways ranging from intrinsic programming of branch points, axonal lengths and trajectories to completely random projections with correct ones being selected/directed by a neuron-target interaction. Likewise, without the development of specific neuronal connectivity patterns there would be no overt behavior, but perhaps some features of the initial behavioral patterns provide feedback to influence the development of the circuitry or effectors.

These studies have established that HEs produce extra peripheral arborizations and subsequently retract them. Furthermore, the extra axonal arborizations are not random projections but are oriented longitudinally and resemble the early heart tube arborization except in lateral-medial position. At the stages considered here, there are many alternative nerve pathways (Jellies and Kristan, 1988b; Jellies et al. 1987; Loer et al. 1987) that are not utilized by the HE axons. We are currently examining these developing arborizations at the ultrastructural level to determine whether they are in comparable locations in three dimensions. Similarly, in insects the early projection of some motor neurons may involve more substantial examples of initial extra projections than previously suspected (Myers et al. 1990). In leech there is an early patterning by muscle cells that establishes the orthogonal axes of the germinal plate (Torrence and Stuart, 1986) and which is known to exert some control over patterns of neuronal migration (Stuart et al. 1989; Torrence et al. 1989). Our results suggest the working hypothesis that the early HE projections are responding to just such a general set of migration cues to establish the position of longitudinally oriented arborizations, only one of which is in the location of the developing heart tube. Future studies will examine the development of the heart tube primordium to determine whether it might not be a competent target for HEs at the time they reach it, but that shortly thereafter it becomes recognizable by the motor neurons. Our present results lead to the prediction that this proposed recognition event (whether trophic, differential adhesivity or some other) would coincide with the rapid transition in peripheral arborization morphology such that they expand over the surface of the forming heart tube at the expense of the more medial body wall projections. This is clearly a speculative notion that remains to be examined further.

Our results also make it unlikely that the centrally patterned rhythmic activity is directly involved in determining the peripheral innervation fields of HEs since the rhythmic activity in HEs did not appear until some time after E16, at least 4 days after the dramatic change in peripheral arborizations had begun. However, our results also imply that the HN interneurons quite early become competent to drive HEs with inhibitory input, whether rhythmic or not, and so could possibly influence HE development via that interaction. Future experiments must examine the HNs as well as the HEs during development to address this possibility. An alternative explanation for our observations of spontaneous activity might be that the CPG is functional earlier than we have proposed but that, because of the more delicate nature of the preparation, our techniques have compromised the CPG. This is clearly possible but seems unlikely because when stable recordings were obtained the activity was robust, IPSPs were consistently present and were not diminished in amplitude, and the motor neuron action potentials were often 3-4 times as large as those recorded from adult HEs. Perhaps these small cells have a higher input impedance and/or the geometry is such that the spike-initiating zone is electrically closer to the recording site. Whatever the underlying cause, these features do not seem to indicate a compromised preparation. This will be addressed in future work by video analysis of the emerging behavior patterns as well as extra-and intracellular recording of underlying electrical activity. Our results do, however, leave open the possibility that peripheral target interactions could influence the formation of appropriate central connections. This remains to be examined with respect to the neurogenic heart.

In those cases where the emergence of motor control has been examined, there is some evidence that the basic patterns of coordination can develop in the relative absence of overt feedback. It is interesting to note that the development of the neurogenic rhythm in the present study (Fig. 5) might follow a roughly similar time course (emerging relatively late in embryogenesis) to the development of swimming behavior (Reynolds and Kristan, 1989), which is another prominent rhythmic activity pattern in these animals. The development of leech swimming was examined (Glover and Kramer, 1982) by ablating the serotonergic cells early in development, well before the swimming pattern generator had developed (as monitored behaviorally). Serotonergic neurons potentiate swimming behavior (Nusbaum and Kristan, 1986; Willard, 1981). After a period of maturation, such animals were incapable of swimming upon appropriate stimulation, yet when serotonin was provided these very same animals could then be induced to swim. The most reasonable broad interpretation of this work seems to be that the swimming circuitry had developed appropriately in the absence of overt swimming movements. Likewise in metamorphosing moths, while there are tantalizing correlations that implicate causal roles for neuron-target interactions in specifying particular phenotypes (Levine and Weeks, 1989), motor neurons and muscles appear to respond to hormonal cues independently rather than influencing each other directly (Levine and Weeks, 1989) and dendritic growth can be stimulated by direct hormonal manipulations (Truman and Reiss, 1988). This relative indepen-dence of motor development is also seen in vertebrates (Oppenheim and Haverkamp, 1986). However, it is important to bear in mind that, while overt motor patterns can develop independently, it may be the case that considerable interactions at the cellular level had led to the stereotypy in motor output.

Indeed, there is considerable precedence in leech development for early interactions that lead to selective cell death or selective retraction of axons (Gao and Macagno, 1988; Loer et al. 1987; Loer and Kristan, 1989b,c; Macagno et al. 1990; Macagno and Stewart, 1987; Shankland, 1991; Shankland and Martindale, 1989a; Wolszon and Macagno, 1992). In one case it has been shown that peripheral target interactions could determine functional synaptic connectivity within the CNS (Loer and Kristan, 1989a). Thus, perhaps both retraction of inappropriate HE arborizations as well as central connectivity to the appropriate CPG can be influenced by target contact. Future experimental manipulations will be needed to address this possibility.

The discovery of HE homologs is interesting in this light as well. Since all HEs initially resemble each other and innervation of the heart tube is secondary, perhaps the phenotype of the HE homologs is a default condition and they are prevented from innervating heart tubes either by competition from definitive HEs in caudal segments or by their lineal descent. Perhaps eliminating the HEs in segments 3 (and/or 4) at an early stage might permit the homologs to innervate the heart tubes and develop the more specialized phenotype. It is interesting in this regard that previously unknown homologs of another segment-specific set of neurons involved in modulation of the heart, the heart accessory (HA) neurons (Calabrese, 1984), have been tentatively identified (Stewart et al. 1991). In these studies it was found that both the definitive HAs and the HA-like neurons received segmentally variable inhibitory input from the HNs. In contrast to the HEs and HE homologs of the present study, the segmental differences in HA and HA-like neurons developed very gradually over the course of embryogenesis and they eventually acquired different antigenic determinants.

In summary, the present study shows that HEs transiently extend similar peripheral arborizations into two different territories, both to the appropriate target region (heart tube) and into other body wall territories, and that they subsequently retract their incorrect projections over the same time course as their correct peripheral arborization expands. This transition is extremely rapid considering the time course of leech development. We have proposed that the early growth may be in response to general cues for migration and patterning analogous to the more global pathways suggested for motor neuron growth in vertebrate limb development (Tosney, 1991) and that HE-heart tube interactions represent a specific local interaction. Furthermore, we have shown that this anatomical remodelling occurs well before the emergence of the centrally driven pattern of rhythmic activity and have suggested that specific target interactions might play a role in both the anatomical patterning and the development of central connectivity. Additionally, we have presented evidence, based upon cell position, central and peripheral anatomy and antigenic similarity, that there are probably segmental homologs of HEs in the first two and last three midbody segments. These homologs are initially similar to definitive HEs in their peripheral projections, but their peripheral arborizations do not regress during embryogen-esis and they do not become driven rhythmically by central input. Thus, the development of the neurogenic heart of the medicinal leech should prove to be a particularly tractable system in which to examine the mechanisms that link the embryogenesis of particular neurons and their synaptic connections with the emergence of refined patterns of overt behavior.

We thank M. P. Nusbaum for helpful comments on an early draft of this manuscript and D. McCarthy for technical assistance. Supported by PHS NS 28603 (J.J.). J.J. is an Alfred P. Sloan Research Fellow.