Architecture of Cerebral Neurosecretory Cell Systems in the Silkworm Bombyx Mori

Insect nervous systems have many specific cells that can be visualized by the characteristic chromophilic materials present in their cytoplasm. Most such ‘neurosecretory cells’ (NSCs) make a specific terminal structure in specialized neurohaemal regions in order to release their products into the haemolymph (Rowell, 1976; Raabe, 1983). Many neurohormones identified in insect nervous systems seem to be derived from these NSCs (Holman et al. 1990).

Morphological and physiological characterization of individual NSCs as well as identification of their hormonal products are vital when attempting to elucidate related neuroendocrine functions. Retrograde cobalt fills combined with silver intensification have revealed the dendritic fields and axonal pathways of NSCs in various species of insects (Pipa, 1978; Koontz and Edwards, 1980; Buys and Gibbs, 1981; Copenhaver and Truman, 1986b). Intracellular injection of a marker dye from a micropipette into a single cell can reveal its complete structure: this is important for identifying a particular cell (Copenhaver and Truman, 1986a), for classifying a cell (Carrow et al. 1984; Copenhaver and Truman, 1986b) and for the possible identification of homologous cells in different species. In addition, intracellular micropipettes can pick up electrical activity that provides clues about the neurosecretory nature of the impaled cell, since the time course of action potentials differs between NSCs and non-NSCs in both vertebrates and invertebrates (Yagi and Iwasaki, 1977; Miyazaki, 1980; Carrow et al. 1984).

The silkworm, Bombyx mori, is one insect in which (neuro)endocrine mechanisms have been extensively studied. The brain has many types of putative NSCs distinguishable by the size and position of their somata, by the characteristic appearance of their cytoplasmic chromophilic inclusions (Panov and Kind, 1963; Bassurmanova and Panov, 1967) and by retrograde cobalt fills (Morohoshi et al. 1977). Recently, three types of NSCs have been identified immunohistochemically (Mizoguchi et al. 1987, 1990; Kono et al. 1990). In the present study, the central and terminal structures of all putative NSCs in the silkworm brain were individually examined using intracellular injections of the fluorescent dye Lucifer Yellow after recording electrical potentials.

The fifth-instar larvae of Bombyx mori L. used throughout this study were Fl hybrids of the races ‘Kinshu’ and ‘Showa’. The larvae were reared at 26°C on a commercially available artificial diet (Kyodo Shiryo Co., Japan).

The gross anatomical features of the cerebral neuroendocrine system and its associated targets were examined by dissecting anaesthetized animals in a physiological saline and fixing the tissues with a 70% ethanol solution.

For electrophysiological studies, the head of a larva was usually isolated and mounted in an experimental chamber using beeswax. Sometimes the head of an intact larva was fixed to the chamber after plugging the mouthparts with a quickdrying glue. The brain was exposed by making a window in the front of the head. The chamber was filled with physiological saline designed to mimic extracellular conditions within the brain (Miyazaki, 1980). A pair of insect pins was placed under the brain to provide a supporting platform. After incubation for 20–30 s in saline containing 1% pronase, the sheath of the brain was partially removed using a fine needle.

Putative NSC somata in the viable brain can be identified under the dissecting microscope by their reflective opalescence. Under visual guidance, an opalescent soma was penetrated with a bevelled glass microelectrode filled with Imoll⁻¹potassium acetate or 5% Lucifer Yellow CH (Sigma) dissolved in 0.15 mol l⁻¹LiCl. An indifferent electrode was placed in the chamber. The microelectrode was connected to a preamplifier (MEZ-8201, Nihon Kohden), which allowed simultaneous recording and current injection via a bridge circuit. Intracellular potentials and signals for injected currents were displayed and stored on a memory oscilloscope (VC-10, Nihon Kohden). Stored signals were replayed and recorded on a chart recorder.

Lucifer Yellow (LY) was ionophoretically injected by passing 300 ms hyperpolarizing pulses of current (10 nA) at 1Hz for 5℃10 min. Subsequently the preparation was maintained at room temperature (19–23 °C) for 1–2 h (isolated head) or at 4°C for 1 day (whole animal). The tissues containing the dye were dissected out together with parts of associated structures (e.g. aorta and oesophagus) and fixed with 4% formaldehyde in 0.15 mol l⁻¹ sodium phosphate buffer (pH7.4) for 1h. The tissues were then washed for 30min in the buffer, dehydrated, cleared with methyl salicylate and gently compressed under the weight of a coverslip or by a finger. The preparations were viewed and photographed immediately at various focal planes with a fluorescence photomicroscope (Nikon). The structure of LY-filled cells was reconstructed from photographic films, using a photographic enlarging projector.

Some LY-filled cells were examined for paraldehyde-fuchsin-positive materials. After observing the fluorescence, the brains were rehydrated and refixed in aqueous Bouin’s solution for 24 h in the dark. The tissues were then stained with paraldehyde-fuchsin according to Dogra and Tandan (1964) for whole-mount preparations.

The morphological organization of the retrocerebral complex and its associated structures in Bombyx larvae is illustrated in Fig. 1. It is similar to that of Manduca sexta larvae (Nijhout, 1975; Copenhaver and Truman, 1986b). A pair of corpora cardiaca (CC) lies posterior to the brain and each CC is connected to the posterior face of the brain by a fused nerve, the nervus corporis cardiaci 1 and 2 (NCC1+2). The CC receives contacts from the nervus corporis cardiaca 3 (NCC3), which arises from the lateral face of the brain as part of the tegumentary nerve and projects to the stomodeal dilator muscles via the CC. The corpus allatum (CA) lies posterior to the CC. The allatal nerve, nervus corporis allati (NCA), connecting the CC to the CA, is very short, being little more than a constriction between the two organs. Each CC gives rise to two additional nerves, medially the nervus corporis cardiaci-nervus recurrens (NCC−RN) and ventrally the nervus corporis cardiaci ventralis (NCC−V). The two NCC-RNs from the bilaterally paired CC join beneath the aorta and run for several hundred micrometres on the surface of the recurrent nerve (RN) before merging into it. The NCC−V runs through the underlying hypopharyngeal musculature and joins a branch of the maxillary nerve of the suboesophageal ganglion (SG). The SG is connected to three thoracic ganglia and eight abdominal ganglia.

The putative NSCs in the brain occur in two clusters, medial and lateral (Fig. 2). Most of the NSCs in the viable brain were clearly visible because of the way they diffracted light. Three distinct groups of cells were usually recognizable in the medial cluster, on the basis of the size, shape and position of their somata; the three groups are here termed Ml, M2 and M3 cells following the nomenclature of Panov and Kind (1963). The relative locations of the somata within the Ml group varied considerably from animal to animal. However, four cells with large somata (40–50 μm in diameter) were readily distinguishable from other cells with small somata (20–30 μm in diameter) in each hemisphere of the brain. Two pairs of M2 cells were usually distinguishable by their ellipsoidal somata located close to the midline of the brain. One or two M3 cells with relatively large and round somata (30–35 μm in diameter) were sometimes faintly discernible in the anteroventral area of both brain hemispheres.

The lateral cluster was always made up of several cells. No individual cell in this cluster was accurately identifiable on the basis of the appearance of its soma in the viable brain, though histological studies revealed three types of cell in this cluster (Panov and Kind, 1963; see Fig. 2).

Retrograde LY filling via the NCC3 in a preliminary experiment revealed three cells with somata in the lateral surface of the tritocerebrum (Fig. 2) and thick dendrites in a ventral region of the tritocerebral neuropile. The fill was carried out because these tritocerebral cells had been visualized previously by retrograde cobalt fills via the NCC3 and were considered to be members of an NSC group (Morohoshi et al. 1977). It was very difficult to identify the tritocerebral cells as they lacked the opalescent appearance of NSCs and they did not stain with paraldehyde-fuchsin.

One hundred and eight cells in 102 animals were filled with Lucifer Yellow. The distribution and overall characteristic profiles of the cells are summarized in Table 1. Examples of the dye-filled cells are shown in Figs 3–5. Fig. 3 shows central and terminal arborizations of an Ml cell with a large soma. A single neurite extending from the soma runs ventrally for a short distance and bifurcates into an axonal process and a large dendritic process that extends branches to the dorsal region of the ipsilateral protocerebrum. The axonal process produces a few thin branches before and after crossing the midline of the brain. A contralateral branch extends to the dorsal protocerebrum. The axon runs along the ventral surface of the contralateral protocerebral neuropile, enters the NCC1+2 and passes through the CC without branching to terminate in the CA. The terminal neurohaemal arborization was confined to the superficial region of the CA (Fig. 3B). Fig. 4 shows an Ml cell with a small soma. The cell has a relatively narrow dendritic field in the medial region of the protocerebrum in both hemispheres of the brain. The axon of the cell projects to the contralateral CC via NCC1+2, where the terminal branches form a complex network of numerous varicosities (Fig. 4B). A few terminal branches leave the CC towards the recurrent nerve via the NCC−RN (Fig. 4C) and the NCC−V (Fig. 4D). The terminal branches do not invade the CA. Fig. 5 shows a lateral cell that projects an axon to the contralateral CA. The neurite originating from the soma travels ventromedially towards the midline and crosses into the contralateral hemisphere of the brain. The neurite is densely fasciculated. The axon, after taking a characteristic pathway in the contralateral hemisphere (for details, see Fig. 10), leaves the brain and terminates in the CA without branching in the CC (Fig. 5B). The axon usually bifurcates in the NCA or CA and spreads varicose terminal processes on the surface of the CA. The terminal of the lateral cell was easily distinguishable from that of the medial cell shown in Fig. 3 by its relatively thin and sparse terminal branches.

This group of cells could be divided into at least five subgroups according to the pattern of the dendritic branches (Fig. 6). The cells of subgroup Mia are characterized by thick ipsilateral and thin contralateral dendrites extending to the dorsal protocerebrum as well by large somata (Fig. 6A). Another cell of the same subgroup is illustrated in Fig. 3. Mia cells always terminated in the CA without branching in the CC. The branching pattern of their dendrites, including other short processes in the medial area of the brain in different animals, showed little variability, suggesting that the Mia cells may represent a functionally homogeneous group. The structural profile of the Mia cells is identical to that of bombyxin-producing cells (Mizoguchi et al. 1987).

Mlb cells usually have a dendritic field confined to the medial region of the protocerebrum (Fig. 6B, see also Fig. 4). A few cells extended a slender process to the ventral area of the ipsilateral protocerebrum. The axon terminals always formed a complex network of branches and varicosities in the CC and there were a few branches extending out of the CC to the NCC−RN and NCC−V (Fig. 4B–D). The branches extending to the NCC-RN usually terminated after running for several hundred micrometres on the surface of the RN, although in a few cells these branches terminated before reaching the RN or after running the entire length of the contralateral NCC−RN. Some varicose processes were also found in the NCA, but they did not invade the CA. These terminal profiles of Mlb cells in the CC complex (the CC and accessory nerves of the CC) were common to the other three subgroups of cells with small somata, as described below.

The Mlc cells gave rise to dorsal dendrites in both hemispheres, whereas the Mid cells lacked the contralateral dendrite (Fig. 6C,D). Cells of both subgroups showed variability in the length of the additional minor branches.

Cells of subgroup M1e are characterized by the large ventral dendrite that arises from a thickening of the neurite in the contralateral hemisphere (Fig. 6E,F). The axon also originates from the thickening and projects to the CC complex. The cells of this subgroup showed considerable variations in additional dendritic processes. Most cells had a thin process extending dorsally (and then turning laterally) in the contralateral hemisphere (Fig. 6F). Some cells extended a (slender) ipsilateral process to the ventral (Fig. 6F) or dorsal area of the protocerebrum (Fig. 6E). This relatively large variability in distribution of the minor processes may suggest a further division of this subgroup of cells, although in this report I put them into the same group.

These cells spread dendritic branches in the anteromedial region of the ipsi- and contralateral hemispheres of the brain (Fig. 7A). The axons run along the ventral surface of the contralateral protocerebral neuropile, send out several fine processes to the tritocerebrum, and travel down the circumoesophageal connective to the suboesophageal ganglion (SG) where they have a few short processes (inset to Fig. 7A). Although the axons were observed to run down the ventral nerve cord connecting the three thoracic ganglia, they could not be traced to their terminals because the dye was only faintly fluorescent in the axons and was partially obscured by the autofluorescence of the thoracic and abdominal ganglia.

Electrical stimulation of the somata of the M2 cells induced a short-duration action potential apparently differing from the typical NSC’s action potential (see Fig. 12). This result could have been obtained because the electrode was introduced into a non-neurosecretory cell located in the close vicinity of the M2 cell rather than in the M2 cell itself. However, the impaled cell was shown to be the M2 cell following staining with paraldehyde-fuchsin after LY injection (Fig. 8). The location, shape and staining properties of the cell soma were all typical of M2 cells.

This group of cells also has dendritic fields in the anteromedial region of the brain (Fig. 7B). However, the axons do not cross the midline of the brain, but traverse the ventral surface of the ipsilateral protocerebral neuropile and enter the SG where they give off few branches (inset to Fig. 7B). Although the axons could be traced only to the SG and to the adjacent ganglion using the dye-filling method described in this study, immunohistochemical techniques, using a monoclonal antibody against the eclosion hormone, have revealed that the axons of the M3 cells extend the full length of the ventral nerve cord and project into the proctodeal nerve (T. Ichikawa, in preparation).

Cells located in the lateral cluster were divided into three subgroups, La−Lc, according to their central and terminal structures.

Fig. 9 shows the central pattern of branching of a cell from the first subgroup. The neurite arising from the soma first runs ventromedially towards the midline of the brain and then dorsolaterally for some distance after crossing the midline, finally turning ventrally to enter the NCC1+2. The densely fasciculated neurite, the axonal pathway and the terminal profile in the CA are identical to those of the prothoracicotropic hormone (PTTH)-producing cells identified immunohisto-chemically (Mizoguchi et al. 1990).

Fig. 10 shows four cells belonging to the second (Lb) and third subgroups (Lc) of the lateral cells. They send two large dendritic collaterals into the medial and ventral regions of the protocerebrum and axons to the ipsilateral NCC1+2. The Lb cells can be differentiated from the Lc cells on the basis of their terminal structures; the former have terminal branches in the CA (Fig. 10A, bottom), whereas the latter, like the Mlb cells (Fig. 4), send terminal branches to the CC complex (Fig. 10B, bottom), though some Lc cells have a few terminal processes running onto the surface of the CA (Fig. 10B). Although the dendritic arborizations of the two subgroups of cells appear complicated and similar, the two can be distinguished on the basis of the branching point of the medial dendrites (arrowheads in Fig. 10). The dendrites of Lc cells originate from the axons, whereas those of Lb cells arise from neurites connecting the somata and the ventral dendritic branches (arrows in Fig. 10).

The dye-filled electrode was advanced into the ventral region of the neuropile, where the tritocerebral cells extended thick dendrites, until it encountered a cell to fill with LY. About 40% of the filled cells sent axons into the NCC3 and had a dense dendritic arborization in the neuropile (Fig. 11). The axons running down the NCC3 passed the CC without branching and extended several long terminal branches among the stomodeal dilator muscles (data not shown).

The duration of action potentials recorded from B. mori NSCs has been reported to be significantly longer than that of non-NSCs (Miyazaki, 1980). Action potentials were recorded from impaled cells, before injecting the dye. They were induced by current injection because many cells showed no spontaneous activity. All impaled cells in the medial and lateral clusters, except the M2 cells, showed large, overshooting and prolonged action potentials (Table 1). Fig. 12A shows such action potentials obtained by injecting current pulses (InA, 300 ms in duration) into a cell of subgroup Mia. The cell produced three action potentials with a duration of 20–30 ms at half-amplitude. Fig. 12B shows a typical action potential with a short duration recorded from the soma of an M2 cell. The duration of the action potential of M2 cells, 2.5 ms, was about one-tenth of that of other medial and lateral cells (see Table 1). It also differed significantly from the duration of action potentials (1ms) recorded from a neurite (dendrite) of the tritocerebral cells (Fig. 12C), which innervate the stomodeal dilator muscles (Fig. 11).

Histological studies using specific ‘neurosecretory’ stains have shown that in the brain of B. mori the medial (group Ml) cell cluster contained at least four and the lateral cell cluster at least three types of cells distinguishable by the size and position of their somata and by the staining properties of their cytoplasmic inclusions (Panov and Kind, 1963; Bassurmanova and Panov, 1967). The five Ml and three lateral cells identified in the present study (Figs 6, 9, 10) illustrate the multiplicity of morphological cell types present in the insect brain. These morphologically and histologically different cells may regulate different physiological and behavioural processes by producing different neurosecretory signals in the form of neurohormones. Various peptide hormones have been extracted from insect nervous systems and their amino acid sequences have been analyzed (Holman et al. 1990). In B. mori, three species of cerebral neuropeptides have been characterized and immunohistochemical studies using monoclonal antibodies have revealed the localization (and structural characteristics) of the cells responsible for the production of peptides. Bombyxin is an insulin-like peptide that activates the prothoracic glands of the saturniid moth Samia cynthia ricini (though inactive on Bombyx mori) and is produced by four pairs of medial cells with large somata (Mizoguchi et al. 1987). The bombyxin-producing cells were positively identified as Mia cells in the present study (Figs 3 and 6A) because of their uniquely large somata and from the distribution of their axon terminals in the CA. Prothoracicotropic hormone (PITH) is a peptide that activates the prothoracic glands of B. mori and is synthesized by two pairs of lateral cells that project to the contralateral CA (Mizoguchi et al. 1990). The immunohistochemical profiles of their characteristic axonal pathway and axon terminals in the CA clearly demonstrate that the PllH-producing cells correspond to the La cells (Figs 5 and 9). Eclosion hormone (EH) is a peptide that triggers pupal-adult ecdysis (Fugo et al. 1984). Although immunohistochemistry using the anti-EH antibody did not reveal their dendritic structure and axonal pathway, the ventromedial localization of the cells reactive to the antibody suggests that they correspond to the M3 cells (Kono et al. 1990). In an immunohistochemical experiment, I observed that the M3 cells filled with Lucifer Yellow were immunoreactive to the anti-EH antibody produced by Kono et al. (1990). In addition, the experiment revealed that these cells had axonal pathways leading to the proctodeal nerve (T. Ichikawa, in preparation).

The axonal pathway of the M3 cells is interesting because it is thought that the EH synthesized by the medial cells may be transported to the CC-CA complex to be liberated from there into the haemolymph. This is suggested by the strong EH activity in extracts from the medial part of the brain and CC-CA complex and in the haemolymph (Fugo and Iwata, 1983; Fugo et al. 1984; Sakakibara and Fugo, 1990). However, the M3 cells do not send any processes to the CC-CA complex (Fig. 7B). In Manduca sexta there are two sources of EH. The EH that triggers the larval-larval and larval-pupal ecdyses is produced by two pairs of ventromedial cells (Truman and Copenhaver, 1989), whereas the EH responsible for the pupal-adult ecdysis is produced by five pairs of lateral cells that project to the ipsilateral CC-CA complex (Copenhaver and Truman, 1986a). The same situation may occur in B. mori because (1) the axonal pathways of the M3 cells of B. mori and of the ventromedial EH cells of M. sexta are the same, suggesting that they are homologous, and (2) an extract from the lateral part of the B. mori brain showed weak, but significant, EH activity (Fugo and Iwata, 1983). If this proposal proves to be correct, then the Lb cells are the most likely candidates for the lateral source of EH, because these cells send axons to the ipsilateral CA (Fig. 10). Confirmatory evidence could probably be acquired by assaying bioactive material released from the cells by intracellular stimulation, an approach used with success in M. sexta (Copenhaver and Truman, 1986a).

Intensive intracellular studies on the structure of single NSCs in Manduca sexta revealed at least three morphological classes of monopolar cells in both the medial (corresponding to the Ml group according to the terminology used in the present study) and the lateral clusters in pupal and adult brain, though an additional class of NSCs appeared in the medial cluster during adult development (Carrow et al. 1984; Copenhaver and Truman, 1986b). Intracellular fills in the present study on the larval brain of Bombyx mori revealed five morphological classes of NSCs in the medial (Ml) and three in the lateral clusters. Despite the species difference and the different stages of development examined, the close structural similarities between some cells of M. sexta and B. mori has facilitated identification of homologous cells in the two species. In B. mori the NSCs projecting to the retrocerebral complex have terminal branches in either the CC or the CA, but not both (Figs 3–5). In M. sexta, a similar restriction of the terminal branches has been observed in the pupal cells (Carrow et al. 1984). In contrast, many cells in the adult brain share two retrocerebral organs for their terminals, and some cells terminate on the surface of the aorta (Copenhaver and Truman, 1986b). Thus, it is easier to draw comparisons between the larval cells of B. mori and the pupal cells of M. sexta. The Mia cell of B. mori (Fig. 6A) apparently corresponds to the medial la cell of M. sexta, which is characterized by its large soma, ipsilateral major dendrites and a terminal in the CA (Figs 4A, 5A in Carrow et al. 1984). The profile of the remaining two classes of M. sexta medial cells whose dendritic fields are limited to the ipsilateral hemisphere does not fit that of any medial cells of B. mori, though these cells commonly terminate in the CC. In the case of lateral cells, it is obvious that the La cell of B. mori (Fig. 9) corresponds to the Ila cell of M. sexta, since both have a densely fasciculated neurite in the ipsilateral protocerebrum, a characteristic axonal pathway and terminate in the contralateral CA (Fig. 4C in Carrow et al. 1984). Immunohistochemical studies have shown that the Ila and La cells both produce the same hormone, PTTH (O’Brien et al. 1988; Mizoguchi et al. 1990). The Lb and Lc cells (Fig. 10) may also be related to the Ilb(CA) and Ilb(CC) cells that terminate in the ipsilateral CA and CC, respectively (Fig. 7 in Carrow et al. 1984).

Axon terminals in the CC always contained a few processes extending to the NCC−RN (and RN) and the NCC−V (Fig. 4). These processes have numerous varicosities along their entire length and do not leave from nerves to reach specific target tissues. Axon terminals in the CA spread their varicose processes on the surface of the CA, but few occur inside the CA (Figs 3, 5). These findings suggest that the retrocerebral complex, including the peripheral nerves of the CC, serves as a release site for the neurosecretory products of the cerebral NSCs into the haemolymph. There was no consistent picture to suggest a direct delivery of the secretory product to a specific target tissue such as the CA, though a few lateral cells did extend a small number of varicose processes to the CA (Fig. 10B). Such a direct delivery mechanism has been proposed in the cerebral NSC system of M. sexta (Carrow et al. 1984; Copenhaver and Truman, 1986a) and other insects (Raabe, 1983).

Intracellular potentials have been recorded from NSCs in several insect species (Gosbee et al. 1968; Wilkins and Mote, 1970; Cook and Milligan, 1972; Norman, 1973; Carrow et al. 1984; Copenhaver and Truman, 1986a,b), including Bombyx mori (Miyazaki, 1980). The NSC soma has a typical electrically excitable membrane and produces large, overshooting action potentials of relatively long duration. The prolonged action potentials recorded from the Ml, M3 and lateral cells in the present study coincide with the general features of electrical activity that have been described for NSCs. In contrast, the M2 cells generate an action potential with a short duration that is practically identical to the action potentials of non-NSCs (Fig. 12). The duration of the action potential in M2 cells was twice that of the tritocerebral cells (Table 1). However, this may simply be due to different recording sites, the soma in the case of the M2 cells and dendrites for the tritocerebral cells. Most insect neurones have electrically non-excitable somata. In these cells, the action potentials recorded in the soma can become much smaller and slower than those recorded in the axon (or the dendrite), depending on the passive cable properties of the cells. The relatively small and slow action potentials of the M2 cells suggest that they may be attenuated potentials that passively invade the electrically non-excitable soma of the cells (Fig. 12B).

The M2 cells contain some paraldehyde-fuchsin (PF)-positive material (Fig. 8). Chromophilic cytoplasmic inclusions, however, do not always indicate a neurosecretory function for cells, since certain pigments and cytoplasmic organelles can also react to PF and to other ‘neurosecretory stains’ (Rowell, 1976; Panov, 1980; Raabe, 1983). The origin of the PF-positive material and the siting of the cell terminals remain to be examined. The electrophysiological and histological characteristics of the M2 cells in B. mori are interesting, because the cells of group M2 (and of group M3) appear to be present not only in Lepidoptera (Panov and Kind, 1963) but also in other holometabolous insects (Panov, 1979). The cells in Chironomus plumosus appear to terminate in the last abdominal ganglion (Panov, 1979).

The tritocerebral cells project axons via the NCC3 to the stomodeal dilator muscles, with no sign of any interaction in the CC (Fig. 11). The action potential produced by the tritocerebral cells was characteristic of an ordinary non-neurosecretory neurone (Fig. 12C). There is no histological evidence that they possess chromophilic neurosecretory material. Copenhaver and Truman (1986b) demonstrated that electrical stimulation of the NCC3 in M. sexta induced contractions in the various muscle groups. All these observations taken together suggest that B. mori tritocerebral cells are motoneurones.

M. Ohara provided comments on the manuscript.