Origins of the insect enteric nervous system: differentiation of the enteric ganglia from a neurogenic epithelium

The assembly of the nervous system involves both the proliferation of undifferentiated cells and their commitment to express particular morphological and biochemical characteristics. In many invertebrates, these two aspects of neurogenesis are closely linked. During the formation of the insect CNS, for example, most embryonic neurons are derived from identified stem cell neuroblasts, and the phenotypes that they subsequently express may be determined by the mitotic sequence in which they arise (Bate, 1976; Thomas et al. 1984; Taghert and Goodman, 1984; Doe et al. 1988). Moreover, the locations of most neurons within the developing ganglia are determined primarily by the spatial arrangement of the stem cells from which they arise, with only minor repositioning of neighboring cells as development proceeds (Goodman et al. 1984; Campos-Ortega and Hartenstein, 1985). Quite a different sequence of events occurs during retinal formation in Drosophila, however, in which the proliferative phase of neurogenesis produces a field of uncommitted precursor cells, while neuronal differentiation is subsequently directed by a series of position-specific inductive interactions (Ready et al. 1976; Basler and Hafen, 1989; Baker et al. 1990).

A third neurogenic program has recently been characterized during the formation of the enteric nervous system (ENS) of the moth, Manduca sexta (Copenhaver and Taghert, 1990). As in other insects, the ENS of Manduca spans the length of the alimentary tract and provides innervation to the visceral musculature and several related structures. Within the ENS of Manduca, two distinct domains can be identified: anteriorly, the enteric neurons are organized into several small ganglia on the foregut surface, while more posteriorly, a distinct set of cells is distributed throughout a branching nerve plexus (the enteric plexus) that spans the foregut-midgut boundary. This latter population has been named the EP cell group (Copenhaver and Taghert, 1989a; see Fig. 1). In a previous report (Copenhaver and Taghert, 1990), we showed that the EP cells arise en masse from a neurogenic placode within the foregut epithelium, but delay their differentiation until after they have migrated across the gut surface and reached their mature positions (Copenhaver and Taghert, 1989b). Moreover, at least some aspects of their differentiation (including neurotransmitter expression) are delayed until after migration is complete, and then are expressed in a position-specific manner (Copenhaver and Taghert, 1989a). In this regard, differentiation of the EP cell population resembles patterns of neurogenesis seen in developing vertebrate systems, in which both central and peripheral neurons undergo extensive migratory displacements and only later express phenotypes that are correlated with specific regions of the nervous system (Rakic and Sidman, 1973; Le Douarin, 1982; Sanes, 1989; Austin and Cepko, 1990).

To complete our characterization of neurogenesis in the embryonic ENS in Manduca, we have examined the developmental events by which the second major group of enteric neurons, those constituting the frontal and hypocerebral ganglia, arise and become committed to express their mature phenotypes. We now report that the neurons of the enteric ganglia arise from three neurogenic zones in the foregut epithelium that, surprisingly, are not mitotically active. Rather, these three zones produce chains of undifferentiated cells that delaminate from the epithelium onto the foregut surface and only subsequently undergo a limited phase of mitotic activity prior to their differentiation. Unlike the large neuroblasts of the insect central nervous system, each of these zone-derived precursors gives rise to 2-4 presumptive neurons of approximately equal size, a pattern of mitosis that is reminiscent of the midline precursor (MP) class of neural progenitors in the insect CNS (Bate and Grunewald, 1981; Goodman et al. 1981). Subsequently, the zone-derived neurons undergo extensive reorganization and delay their differentiation as the primordial ganglia are formed, similar to the delayed sequence of differentiation seen in the enteric plexus. Once the ganglia are assembled, however, a number of the neurons can be recognized on the basis of cell-specific phenotypes, while many other cells can be categorized as members of a particular subtype. Thus in the ENS of Manduca, a novel blend of neurogenic and differentiative events that have been previously associated with other regions of the nervous system gives rise to the enteric ganglia.

Animal rearing and embryonic staging was performed as previously described (Copenhaver and Taghert, 1989a, 1990) and by reference to published schedules of external and internal markers (Dorn et al. 1987; Broadie et al. 1991). The developing ENS was visualized by dissecting embryos in the following medium (vol/vol): 50% Schneider’s Drosophila medium, 40% Eagle’s basic salts, 9.9% heat-inactivated fetal calf serum, 0.09% penicillin-streptomycin (Sigma #P-0906), 0.01 % insulin, supplemented with Manduca hemolymph (Copenhaver and Taghert, 1990; after Chen and Levi-Montalcini, 1969; and Seecof et al. 1971). For cell counts, the enteric ganglia of embryos and postembryonic larvae were stained in whole-mount with toluidine blue (after the method of Fahrbach and Truman, 1987). Structural features of individual neurons were revealed by iontophoresis of the dye lucifer yellow CH (5% in 2 M LiCl; Sigma), followed by the application of an anti-lucifer yellow antiserum (1:400; Taghert et al. 1982).

Whole-mount immunohistochemical staining was performed as previously described (Copenhaver and Taghert, 1989a), using 2 % paraformaldehyde or a modified Zamboni’s fixative (4 % paraformaldehyde, 15 % saturated picric acid, in sodium phosphate buffer, pH7.2). Ascites fluid containing monoclonal antibody TN-1, which recognizes a cell-surface molecule that may be related to fasciclin II (Nardi, 1990), was used to visualize the developing ENS throughout the period of ganglion formation (at dilutions of 1:20000 to 1:40000; Copenhaver and Taghert, 1989b). Rabbit antisera made against commercially synthesized FMRFamide (Phe-Met-Arg-Phe-NH₂; Sigma) were used at concentrations of 1:4000 to 1:8000. Controls for this antiserum included the application of pre-immune serum in place of the immune serum and preabsorption of the antisera with the conjugated antigen, both of which resulted in an inhibition of positive staining.

Antisera to a number of other candidate transmitter substances were used at the following concentrations: anti-adipokinetic hormone (AKH; gift of Dr H. Schooneveld) sera #241 and #433 at 1:400 and 1:1000, respectively; anti-proctolin (gift of Dr M. O’Shea) at 1:500; anti-Substance P (serum R5; gift of Dr J. Krause) at 1:5000; anti-choline acetyltransferase (CHAT; gift of Dr M. Forte) at 1:400; anti-g-aminobutyric acid (GABA; gift of Dr J. Hildebrand) and anti-[ Drosophila]-dopa decarboxylase (DDC; gift of Dr R. Hodgetts) were both used at 1:1000 to 1:2000. Commercially available antisera to serotonin (Immunonuclear Inc) was used at 1:400 to 1:1000. Monoclonal antibodies to gastrin/CCK (gift of Drs A. Strack and A. Lowey) and to the molluscan small cardioactive peptide (SCP_B; gift of Drs B. Masinovsky and A. O. D. Willows) were used at dilutions of 1:200 and 1:20, respectively. A monoclonal antibody against [Dros-ophila]-choline acetyl transferase (gift of Drs M. Forte and W. Wolfgang) was used at 1:50 concentration. For most of the immunohistochemical reactions described above, we did not perform all of the appropriate controls needed to characterize the transmitter-related epitopes; rather, the various antisera were used as a means of distinguishing individual phenotypes within the enteric ganglia, as a prelude to more extensive analyses of specific cell lineages in the developing embryo.

Staged embryos were selected at specific times throughout the period of ganglion formation (24–40% of development; 1 % of development equals ∼1 h of real time) and labelled with the thymidine analogue, 5-bromo-2’-deoxyuridine (BrdU; Sigma) using previously described methods (Truman and Bate, 1988; Bodmer et al. 1989; Copenhaver and Taghert, 1990). After 2h of incubation with BrdU (50μgml^-1), the preparations were fixed and stained with an antibody to BrdU at concentrations of 1:30 to 1:50 (Becton-Dickinson; see Gratzner, 1982). Cell lineage relationships within the developing enteric ganglia were examined using intracellular injections of fluorochrome-coupled dextran amines (Copenhaver and Taghert, 1990; after Gimlich and Braun, 1985, and Wetts and Fraser, 1988). Lysinated tetramethylrhodamine dextran amine (LRD; 10×10³M_r) or lysinated fluorescein dextran amine (FRD; 10×10³M_r from Molecular Probes, Inc.) was injected into individual zone-derived cells at times throughout this same period of development, and the number and distributions of labelled progeny were examined after 24 h in embryonic culture. Preparations were also routinely counterstained with TN-1 to visualize the overall morphology of the ENS.

As in other insect species, the ENS of Manduca spans the length of the alimentary tract (Fig. 1A), supplying innervation to the visceral musculature of the foregut, midgut and hindgut, and to several adjacent structures. The gut also receives innervation from the brain and terminal abdominal ganglia, but most of its innervation is supplied by discrete populations of enteric neurons that are localized within specific regions of the ENS on the gut surface. Previously, we have described the differentiation of the EP cell group, which occupies a branching nerve network (the enteric plexus) that spans the foregut-midgut boundary (Fig. IB; Copenhaver and Taghert, 1989b, 1990). In this report, we have focussed on the developmental origins of the second major group of enteric neurons that form the anterior enteric ganglia and associated neural structures.

Most of the neurons in the anterior ENS are organized into two contiguous ganglia, the frontal and hypocerebral ganglia (FG and HG) that lie superficially on the dorsal surface of the pharyngeal region of the foregut (Fig. IB). Typical of all insects, the frontal ganglion in Manduca is the largest structure of the ENS. It lies just rostral to the brain and is connected to the tritocerebral brain lobes via the bilaterally paired frontal ganglion connectives (FGC). The general morphology of the frontal ganglion is analogous to the organization of ganglia of the insect CNS: its ∼70 neurons are segregated in the form of a cortical dome (Fig. 2A), 2-4 cell layers thick, that surrounds a ventral neuropil. Immediately caudal to the frontal ganglion is the hypocerebral ganglion (Fig. IB), so named because it sits directly below the brain on the gut surface. This elongate ganglion contains ∼40 neurons that are arrayed in loose columns along the length of the ganglion.

Posteriorly, the hypocerebral ganglion is continuous with the recurrent nerve, which in Manduca also contains a small population of enteric neurons that are variably dispersed along the length of the nerve. The recurrent nerve is also connected to the major neurohemal organs of the brain (the corpora cardiaca-corpora allata complex, or CC-CA), via bilaterally paired nerves (Fig. IB). These nerves have been given a variety of names (Fraser and Pipa, 1977; Copenhaver and Truman, 1986), but will be described in this paper as the ‘nervi cardiostomatogastrici’ (NCS; after Penzlin, 1985). Near the foregut-midgut boundary, the recurrent nerve is continuous with the apex of the enteric plexus, from which the plexus nerves bifurcate and extend onto adjacent regions of the visceral musculature.

Within the enteric ganglia of Manduca, cell-specific phenotypes could be distinguished both by anatomical and biochemical criteria. The relative positioning of the neuronal somata varied somewhat from animal to animal and was often asymmetric across the midline of the ganglia (e.g. Fig. 2A, 2B). Nevertheless, a number of the cells could be routinely identified in postembryonic larvae on the basis of morphology, transmitter phenotype, and approximate soma position. For example, two of the largest neurons (∼40μm) in the rostral region of the frontal ganglion stained with an antiserum against the molluscan neuropeptide FMRFa-mide (Fig. 2B). The axons of these cells could be routinely traced along the recurrent nerve and onto specific muscle fibers near the foregut-midgut boundary (Fig. 2F). A second set of four neurons in the frontal ganglion reacted positively to an antiserum against the C-terminus of the insect neuropeptide, AKH (Fig. 2C), and these cells possessed axons that extended through the enteric plexus and onto the muscle bands of the midgut (not shown).

Many other neurons in the frontal ganglion could not be uniquely identified but could be categorized into one of several distinct subtypes: for example, several smaller cells (2-4) were stained with anti-serotonin antibodies and a variable number of cells (10-14) were stained weakly with anti-dopa decarboxylase (not shown). Consistent patterns of transmitter expression were also observed in the hypocerebral ganglion and recurrent nerve, including one pair of bipolar neurons that were immunoreactive to anti-FMRFamide (Fig. 2E), and an additional subset of cells that stained more weakly with anti-DDC (not shown). Several antisera that label structures in the ENS in other studies (e.g. Homberg et al. 1987; Davis et al. 1989) did not stain any of the enteric neurons in Manduca, including antisera to proctolin and GABA, although immunoreactive processes projecting to the frontal ganglia from the brain were detected with antisera to GABA and CHAT (not shown). In addition, FMRFamide-immu-noreactive processes could be detected within the NCS nerves that join the recurrent nerve with the CC-CA (Fig. 2D). These fibers originate from a set of cerebral neurosecretory cells within the protocerebrum of the brain (a subset of the ‘type Ila’ cells; Copenhaver and Truman, 1986; see also Homberg et al. 1991). Antisera against AKH also stained the intrinsic neurons of the corpora cardiaca, some of which sent immunoreactive processes into the NCS (not shown). As discussed below, the anterior ENS and the corpora cardiaca are developmentally related; these observations suggest that the two structures may share overlapping functions, as well.

The morphological phenotypes of cells within the enteric ganglia were examined in more detail by systematically injecting individual cells in different regions of the ganglia of mature embryos with lucifer yellow. The results of 72 such preparations revealed a number of representative cell types (each identified in 5⩾3 fills) that could be grouped by their primary projection patterns, as shown in Fig. 3. In both the frontal and hypocerebral ganglia, several morphological subtypes were found to possess unilateral (or asymmetric) projection patterns (e.g. Fig. 3A, D, I), but the vast majority of dye-injected cells had bilaterally symmetrical processes. Within the frontal ganglion, neurons could be classified on the basis of whether their processes were confined to the ENS (Fig. 3A-C) or whether they also projected to the CNS (Fig. 3D-E). In the latter case, they usually terminated in the tritocerebral lobes of the brain, although a few cells projected into the protocerebrum or through the brain connectives and into the subesophageal ganglion (not shown). An additional subset of cells sent processes both to the CNS and to the gut surface, providing innervation either to the nearby pharyngeal musculature (Fig. 3F) or to other foregut musculature (Fig. 3G). Similarly within the hypocerebral ganglion, one subset of cells was found that projected only to peripheral regions of the ENS and gut surface (Fig. 3H-J), while another subset sent bilateral projections to the CNS, as well (Fig. 3K-L). These morphological classes are similar to the subtypes of enteric neurons that have been described in orthopteran species (Gundel and Penzlin, 1978; Kirby et al. 1984).

It should be emphasized that while Fig. 3 shows representative examples of the different cell classes in the enteric ganglia, these drawings do not indicate all of the anatomical variations that we have observed. For example, both the FMRFamide-positive and AKH-positive cell types shown in Fig. 2 have morphologies that are generally similar to the cell class illustrated in Fig. 3C, although both sets of peptidergic neurons gave rise to much more extensive peripheral terminations, as already described. It was also not possible to examine every neuron unambiguously, given the variability in soma positions that occurred in both the frontal and hypocerebral ganglia. Nevertheless, these observations indicate that the anterior enteric ganglia of Manduca are organized in a manner that is similar to the insect CNS: within the enteric ganglia, cells form discrete cortical layers that are segregated from the underlying neuropilar regions. In addition, at least some of the neurons are individually identifiable by anatomical and biochemical criteria, while many of the smaller neurons may be classified on the basis of their transmitter phenotypes or shared morphological characteristics. As described below, the variability that we observed in the relative positions of particular neurons could be traced to the initial assembly of the enteric ganglia during embryogenesis.

To characterize the formation of the enteric ganglia in Manduca, we visualized the embryonic ENS at progressive stages of development using the monoclonal antibody TN-1 (Taghert et al. 1983; Carr and Taghert, 1988), which selectively labels the neuronal components of the ENS as they differentiate on the gut surface (Copenhaver and Taghert, 19896). The earliest indication of neurogenesis in the ENS was detected at about 24% of embryogenesis, with the appearance of three neurogenic ‘zones’ in the mid-dorsal epithelium of the stomodeum (Fig. 4, 5A). These three zones, which we designated Z_1; Z2 and Z₃, differentiated at about the same time, each consisting of a small number of epithelial cells (6-10 cells) that were weakly TN-1 positive and that had begun to delaminate from the epithelial layer. When viewed dorsally (Fig. 5E), the zones initially appeared as rounded structures that could be distinguished from the surrounding epithelium; this image is reminiscent of the ‘clear areas’ reported by Baden (1936) in the foregut of the embryonic grasshopper. A sagittal view of this process (Fig. 5A, B) showed that the appearance of the zones resulted from a marked change in the shape of the zone cells, the basal ends of each cell becoming increasingly enlarged with a corresponding narrowing of the apical portions. As development progressed, the clustering of these cells gave each zone a distinctive cone-shaped appearance, which tapered at its apical margin and expanded onto the basal epithelial surface (Fig. 5B, C).

Over the next 15 % of development (25–40 %), these three zones gave rise to columns of cells that extended along the foregut surface (Fig. 4). Individual cells emerged from the zones, gradually losing continuity with the epithelial layer and assuming a rounded morphology. As additional cells were generated from the zones, those cells that had already emerged became displaced anteriorly, so that three distinct chains of cells could soon be distinguished on the foregut surface (Fig. 5B, C). These three chains overlapped, the cells from zone 3 extending over the cells of zone 2 and then over zone 1 (Fig. 5C). As the stomodeum continued to elongate, cells from all three zones became increasingly intermingled, forming a loose cellular ridge that subsequently became reorganized into all of the major structures of the anterior ENS. Those cells nearest the anterior margin of the foregut aggregated into a bulbous cluster (Fig. 4, 30-39%), which gradually condensed into the incipient frontal ganglion (Fig. 4, 5D). At the same time, adjacent cells coalesced to form the hypocerebral ganglion, while more posteriorly, the residual cell columns became progressively thinner and elongated to form the recurrent nerve (Fig. 5F). Other cells from the anterior cluster migrated laterally and rostrally, establishing the major nerve roots associated with the enteric ganglia (Fig. 4, 36-42%).

While the initial formation of the three neurogenic zones occurred more or less simultaneously (around 24 % ), the contributions of the different zones varied both in duration and in the number of cells that they elaborated onto the foregut surface. Zone 3, the most posterior zone (Fig. 4, 5A), persisted for the shortest period. As with the other zones, zone 3 initially gave rise to a continuous chain of cells that extended onto the foregut surface (Fig. 5A). Starting around 27%, however, this zone underwent a transformation, during which all of the residual cells within the zone gradually emerged onto the foregut surface (Fig. 4). As a result, by 30% of development, the zone was no longer apparent in the gut epithelium (Fig. 5B). This same sequence of events subsequently occurred in zones 1 and 2, as well. The residual cells of zone 2 gradually emerged from the epithelium between 36 and 38% (Fig. 4, 5C), while the cells of zone 1 maintained a persistent connection with the epithelial layer until 38-40 % (see Fig. 5D, F). By 42 %, all three zones had disappeared, leaving only the mid-dorsal ridge of cells on the foregut for the continued differentiation of the ENS.

Concurrent with this sequence of zone-derived neurogenesis, a second major program of neurogenesis commenced and gave rise to the EP cell population of the enteric plexus. As previously described (Copenhaver and Taghert, 1990), the EP cells are derived from an epithelial placode that invaginates into the body cavity from the protruding lip of the dorsal foregut (invaginating from the apical to the basal surface of the epithelium). In the present study, we observed that the neurogenic zones of the anterior ganglia and this neurogenic placode were spatially related but temporally distinct. As shown in Figs 4 and 5, the EP cell placode first differentiated at about 30% of development, encompassing the narrow strip of epithelium that had initially given rise to the zone 3 cells. However, the neurogenic placode appeared only after zone 3 had been obliterated. During the subsequent invagination of the placode, the last of the zone 3-derived cells remained closely apposed to the emerging cell packet; these residual zone 3 cells were clearly distinguishable as a small cluster of rounded cells at the apex of the EP cell group (Fig. 5D,F, arrowhead). In this manner, the more anterior domains of the ENS established continuity with the components of enteric plexus from the earliest stages of their differentiation, a relationship that was maintained throughout all subsequent phases of development.

To characterize the mitotic patterns that were associated with the three neurogenic zones, we next mapped the distributions of cells that were undergoing active DNA synthesis throughout the period of enteric ganglion formation, using the BrdU-antiBrdU technique (see methods). Surprisingly, during the initial phases of neurogenesis (24-30%), we could detect no labelled cells either within the three zones themselves or in the delaminating cell groups (Fig. 6), although there was substantial labelling in the surrounding epithelium. Only after individual cells had completely left the epithelial layer could labelled nuclei be detected within the zone-derived groups. In any particular developmental stage, only a small number of zone-derived cells were found to be immunoreactive for BrdU (typically between 4-10 cells per preparation), and these were scattered along the length of the foregut surface (Fig. 6, 32-40%). At no time could we detect labelled nuclei within the neurogenic zones themselves; rather, immunoreactive cells were often found just outside of the zones, presumably having recently emerged from the epithelial layer. Scattered examples of labelled cells could also be found in more rostral regions of the developing ENS, as well.

In general, we saw no evidence for organized patterns of mitosis during the subsequent coalescence of the zone-derived cells, although synthetically active nuclei occasionally were seen within the cell groups of both the frontal and hypocerebral ganglia. In rare preparations, a coordinated pattern of labelled nuclei was detected in the vicinity of individual zones just prior to their obliteration (Fig. 6, 36 % ) : in these cases, many of the residual cells that were completing the process of delamination became immunoreactive for BrdU simultaneously, forming a ‘halo’ of labelled nuclei around the remnants of the zone. The fact that this pattern was only seen in a few preparations (3 out of 50) suggests either that this coordinated mitotic pattern occurred within an extremely narrow developmental window, or that this pattern was simply due to the chance synchronization of cells in the same phase of their differentiative cycle.

During the subsequent period of enteric ganglion formation, a scattered number of zone-derived cells were invariably labelled (38–40%), suggesting that at least some of the zone-derived cells underwent more than one round of mitosis. However, by 42 –43 %, the generation of neurons in the enteric ganglia was essentially complete: numerical analyses of toluidine-blue-stained preparations at these stages indicated that the mature complement of frontal ganglion cells had already been attained (68.6±7.8 cells, n=16), although the ganglion continued to expand in size throughout the remainder of embryonic and postembryonic development. Also during subsequent development, specific subsets of cells derived from zones 2 and 3 did continue to be mitotically active, producing a large number of glial-like cells that populated the branches of the enteric plexus as well as the recurrent nerve (to be described in a subsequent report).

The results of the foregoing experiments indicated that while the neurogenic zones gave rise to the progenitors of the enteric ganglia, the zones themselves were not sites of active cell proliferation. To gain insight into the contributions of individual progenitors to the developing ENS, we next examined the clonal relationships of cells within the enteric ganglia using intracellular injections of fluorochrome-coupled dextran amines in embryonic culture. While individual cells within a particular zone could be labelled by this technique, variability in cellular positions within the zones precluded us from determining whether specific precursor cells from a particular zone could be linked with unique cell lineages. Nevertheless, these experiments revealed the replicative patterns involved in the formation of the enteric ganglia and provided insight into the patterns and distribution of progeny that were derived from the three different zones.

When individual zone cells were injected during the initial phases of neurogenesis (25-28 %), a small subset of labelled neurons were typically labelled within the coalescing ganglia (Figs 7,8). The progeny of individual cells were usually clustered in either the frontal or hypocerebral ganglion (e.g. Fig. 8C,F), although labelled neurons were occasionally distributed across both regions (Fig. 7A). The number of labelled neurons in these preparations was consistently low (Fig. 9), with a maximum of four cells being labelled following the injection of a progenitor cell near one of the zones (Figs 7A, 8C,F). More often, we subsequently detected only one or two labelled neurons within the developing ganglia (Fig. 7B), regardless of the position occupied by the progenitor cell at the time of injection. In many cases, labelled neurons could be seen extending neurites in patterns reminiscent of the phenotypes that we observed in the mature ganglia (compare Figs 3 and 7). When more than one labelled cell was detected, the cells were typically similar in size and extent of outgrowth. Not all of the neurons derived from an individual zone cell acquired identical morphological characteristics, however, indicating that individual progenitor cells could give rise to more than one neuronal subtype within the enteric ganglia.

During subsequent phases of development (29-36%), the average number of cells that were labelled in this manner gradually declined (Fig. 9): injections resulting in as many as four labelled neurons became relatively rare, while more frequently only one or two fluorescent neurons were detected (Figs 7, 8B,E). By the time the enteric ganglia had begun to coalesce (37-40%), most of the injected cells had already begun to differentiate (as indicated by the onset of neurite outgrowth), and typically these cells did not undergo any additional rounds of mitosis (Figs 7, 8A,D). At all developmental ages examined, a few preparations resulted in no labelled neurons (indicated in Fig. 9 as ‘0’ labelled progeny), indicating that the injected cell may have died. We have not yet detected any evidence for significant patterns of stage-specific or systematic cell death, however.

We also used these experiments to examine the relative contributions of the three zones to the different structures of the ENS. During the initial phases of neurogenesis (25-28%), cells derived from all three zones were found to contribute progeny to either the frontal or hypocerebral ganglion and occasionally to both, as already mentioned. These results are in agreement with our observations of TN-1 stained preparations (Figs 3, 4), in which we observed the chains of cells from all three zones streaming anteriorly in an intermingled fashion prior to the formation of the enteric ganglia. With continued development, however, the positions of labelled progeny became increasingly restricted in a predictable manner: cells derived from the most anterior zone (zone 1) were most likely to be incorporated into the coalescing frontal ganglion, while cells from the more posterior zones typically produced cells that occupied the hypocerebral ganglion or recurrent nerve. Thus there was a gradual restriction in the contribution of the neurogenic zones to the different regions of the ENS.

While most of the experiments in this paper focussed on the neurogenic phase of enteric ganglion formation (25 –40% of development), the differentiation of the anterior ENS occupied a substantially longer period of embryogenesis. As previously mentioned, some of the zone-derived cells associated with the enteric ganglia also grew out to form nerve roots (Fig. 4). In addition, a number of other cell groups were detected with TN-1 staining that were not zone-derived but that delaminated from more lateral regions of the foregut epithelium during the formation of the ganglia (Figs 10, 11). These additional cell groups subsequently participated in the differentiation of the major nerves of the frontal ganglion, hypocerebral ganglion and corpora cardiaca.

The initial differentiation of these lateral groups commenced soon after the neurogenic zones had begun to elaborate neuronal precursors: as early as 30% of development, bilaterally paired sets of epithelial cells on either side of the neurogenic zones had begun to delaminate from the foregut epithelium. The earliest and most prominent of these groups subsequently aggregated on either side of the frontal ganglion to help form the frontal ganglion connectives (Fig. 10). Other cell groups differentiated next to the hypocerebral ganglion and gradually formed nerves projecting to the buccal and pharyngeal musculature (Figs 10,11). At the same time, paired groups of cells also emerged from ventrolateral regions of the foregut, posterior to the level of the hypocerebral ganglion (Fig. 10, 35%; see curved arrows). Cells from these groups gradually shifted both medially towards the recurrent nerve and laterally towards the paired corpora allata (structures that invaginate from the ectodermal layer of the body wall), forming the NCS nerves. In addition, a small number (5 –8) of cells became incorporated into the distal portions of the NCS nerves adjacent to the corpora allata, forming the intrinsic neurons of the corpora cardiaca. Thus while all of the ganglionic neurons of the ENS are derived from the three neurogenic zones, an independent population of epithelial cells contributes substantially to the differentiation of ancillary portions of the ENS in Manduca, including the intrinsic neurons of the neurohemal complex of the brain.

These experiments show that the enteric ganglia of Manduca arise via a neurogenic sequence that is distinct from the patterns of differentiation seen in other regions of the developing ENS (Copenhaver and Taghert, 1990). As summarized in Fig. 12, three neurogenic zones appear simultaneously within the columnar epithelium of the foregut (at about 24 % of development; panel A) and give rise to a progression of neuronal precursor cells. Without undergoing mitosis, epithelial cells in proximity to these three zones are apparently recruited into the zones and undergo an epithelial-to-mesenchymal transformation, during which the cells delaminate onto the foregut surface (Fig. 12, panel B). Then, only after they have emerged from the epithelial layer do the cells enter a limited phase of mitotic activity. Both the distribution of mitotically active cells (revealed by the BrdU technique) and the lineage relationships among the enteric neurons (as indicated by clonal analysis) have shown that each precursor produces a maximum of 2 –4 progeny of approximately equal size and similar developmental potentials, the majority being destined to assume neuronal phenotypes.

As precursor cells continue to be generated from the three zones, the cells and their progeny shift forward, intermingling as they aggregate into the rostral clusters that will form the enteric ganglia (Fig. 12, panel C). The neurogenic zones remain active for only a limited phase of development: starting with the most posterior zone (Z₃), each zone gradually disappears from the epithelial layer as all of its residual cells emerge onto the foregut surface (Fig. 12, panels C and D). At the same time, the neurogenic placode that will produce the EP cells of the enteric plexus (Copenhaver and Taghert, 1990) begins to form, incorporating the epithelial region from which the zone 3 cells initially emerged (Fig. 12, panel D). As the ganglia coalesce, some variability occurs in the relative positioning of individual neurons within the ganglia (as manifested by the variable positions of specific cells observed in post-embryonic animals). Nevertheless, at least some of the neurons subsequently express individually identifiable characteristics (Fig. 12, panel E), including specific transmitter phenotypes and morphological features, while many other ganglionic neurons can be classified by subtype. Zone cells also coalesce to form the recurrent nerve that serves as a persistent fink between the enteric ganglia and more posterior domains of the ENS, including the developing enteric plexus. By 40% of development, essentially all of the neurons that will be incorporated into the enteric ganglia have been generated, and the proliferative phase of neurogenesis is complete.

In several respects, the origins of the enteric ganglia in Manduca resembles patterns of neurogenesis that have been described for the insect CNS. For example, the generation of most central neurons in insects begins with the differentiation of uniquely identifiable stem cells (neuroblasts) from the neuroepithelium, a process that is regulated by inhibitory interactions among equipotent ectodermal cells (Doe and Goodman, 1986). The initial specification of the neurogenic zones in the insect stomodeum may be regulated in an analogous manner. Each zone differentiates at a specific (middorsal) position from within a field of apparently uniform cells, in the absence of any obvious prespecification of the epithelial layer. As with neuroblasts in the insect CNS, individual zone cells then lose their continuity with neighboring cells and round out onto the epithelial surface. However, while most neuroblasts subsequently undergo an extended sequence of asymmetric divisions and produce many neuronal progeny (Bate, 1976; Goodman et al. 1984; Taghert and Goodman, 1984), the zone-derived precursors are mitotically active for only a short period of time. At most, individual precursors were found to produce four progeny and more typically gave rise to only one or two differentiated neurons. In this regard, the zone-derived precursors more closely resemble the midline precursor (MP) class of neural progenitors in the insect CNS, each of which undergoes a single, symmetric division and gives rise to a pair of identified neurons (Bate and Grunwald, 1981; Goodman et al. 1981). Thus one view of the neurogenic zones in the ENS is that they each provide a series of MP-like precursors, whose progeny assume a variety of mature phenotypic characteristics only after they have migrated into the developing enteric ganglia.

An important aspect of zone neurogenesis is the process by which the neurons of the enteric ganglia become committed to express cell-specific phenotypes. Individual neurons may differentiate according to their mitotic ancestry or they may be regulated by positional information encountered during subsequent phases of their development. For example, in each neuromere of the developing insect CNS, ∼7 MP cells emerge from distinct regions of the neurectoderm and give rise to unique sets of neurons whose phenotypes are correlated with the initial precursor positions (Bate and Grunwald, 1981; Goodman et al. 1981). In contrast, the progenitors of the enteric ganglia generate a greater variety of cell types (Figs 2 and 3) but themselves arise from just one of three positions (zones Z₁-Z₃) within the stomodeal epithelium. If a lineage-based mechanism regulates the differentiation of these neurons, then the properties of the zones must change with time so that the precursor cells that emerge from each zone can be assigned distinct fates. Alternatively, a lineageindependent model of neuronal differentiation in the ENS would involve regulatory interactions among the zone-derived progeny during their incorporation into the developing enteric ganglia. In this scenario, heterogeneity of neuronal phenotype would be regulated by a stepwise series of cellular interactions, as has been documented during the formation of the insect retina (Tomlinson and Ready, 1987; Baker et al. 1990). A more thorough examination of cell lineage relationships within the ENS (in terms of specific neuronal phenotypes) and an analysis of the developmental potential of individual precursor cells will be needed to discriminate between these possibilities.

It should be noted that our characterization of transmitter and morphological phenotypes within the enteric ganglia did not provide an exhaustive index of the different cell types found in the ENS but was intended to demonstrate the nature of cellular identities within the ganglia, and to provide markers for future investigations into the regulation of neurogenesis. A number of the observations reported in this paper are similar to features of the insect ENS that have been reported elsewhere, including cells (or processes) that are immunoreactive for GABA (Homberg et al. 1987), serotonin (Radwan et al. 1989), AKH (Schooneveld et al. 1985; Homberg et al. 1991), and FMRFamide (White et al. 1986; Carroll et al. 1986; Myers and Evans, 1987; Homberg et al. 1990).

The neurogenic sequence that we have described for the anterior ENS differs markedly from the sequence that gives rise to the other major group of enteric neurons in Manduca, namely the EP cells of the enteric plexus (Copenhaver and Taghert, 1989a,b). Whereas the zone-derived precursors delaminate sequentially from the foregut epithelium and then become mitotically active, the EP cells arise en masse from an epithelial placode and become mitotically inactive as they leave the columnar layer (Copenhaver and Taghert, 1990). While at least some of the zone-derived neurons assume cell-specific phenotypic traits, the EP cells do not assume uniquely identifiable characteristics but constitute several distinct subtypes that are intermingled within the enteric plexus (Copenhaver and Taghert, 1989a). In addition, while both the zone-derived and plexus-derived cell groups undergo substantial rearrangements prior to their terminal differentiation, the outcome of migration in the two regions of the ENS is strikingly different: anteriorly, the zone cells coalesce into discrete ganglia, while posteriorly, the EP cells become progressively dispersed throughout the nerves of the plexus.

Despite these differences between the two domains of the ENS, the neurogenic zones and the neurogenic placode are developmentally related: the EP cell placode appeared immediately after zone 3 (the most posterior zone) had disappeared, and its boundaries were larger, encompassing the region of epithelium previously defined as zone 3 cells (shown schematically in Fig. 12). This relationship is maintained during subsequent stages of embryogenesis, as well, with the residual zone 3-cells eventually forming the neural pathway through which neurons from both the ganglia and the enteric plexus send their axonal processes (Copenhaver and Taghert, 1989a and unpublished observations), llhus the zones and the EP placode arise from overlapping regions of the foregut but are temporally distinct, suggesting that similar mechanisms may control the initial commitment of the gut epithelium into these two programs of neuronal differentiation.

In comparing the differentiation of the ENS in different insects, at least two (and usually three) ‘clear areas’ or invaginations have been detected in the developing foregut of other species, which have been ascribed a neurogenic function (Heider, 1889; Poulson, 1965; Anderson, 1972; Kobayashi and Ando, 1983). Undoubtedly, these invaginations are equivalent to the three neurogenic zones that we have characterized in Manduca. In Drosophila, few (if any) mitoses have been detected within the invaginations themselves, but sporadic divisions have been reported within the delaminated cell groups (Campos-Ortega and Hartenstein, 1985). These observations corroborate our own analysis of the neurogenic sequence that gives rise to the insect enteric ganglia. With respect to the later stages of development, we have shown that a number of cell groups besides the three neurogenic zones delaminate from the foregut epithelium and contribute to the formation of the ENS. The observation that the corpora cardiaca arise in this fashion is novel, in that the derivation of these structures in other insects has been been attributed to one or more of the neurogenic zones (Schoeller, 1964; Kobayashi and Ando, 1983; Campos-Ortega and Hartenstein, 1985). Our observations suggest that in Manduca, the neurogenic potential of the foregut epithelium may be more widespread than was previously appreciated.

We wish to thank Drs H. Schooneveld, M. O’Shea, J. Krause, J. Hildebrand, R. Hodgetts, A. Strack, A. Lowey, B. Masinovsky, A. O. D. Willows, M. Forte, and W. Wolfgang for gifts of the antisera used in this work. We also thank Ms Marisa LaGrange for excellent technical assistance, and Drs J. C. Weeks and S. Matsumoto for critical readings of this manuscript. This research was supported by NIH grant #NS-21749 to P.H.T., NIH fellowship #F32NS07957 to P.F.C., and a research initiation grant from the Medical Research Foundation of Oregon (to P.F.C.).