Patterned activation of unpaired median neurons during fictive crawling in Manduca sexta larvae

Successful execution of behavioral movements requires flexibility in patterned motor activity and muscle function. Sensory feedback and neuromodulation are two mechanisms that permit flexibility in central circuits and peripheral motor function in the face of changing conditions in the internal or external environment (Burrows, 1996; Harris-Warrick and Marder, 1991; Sombati and Hoyle, 1984; Ramirez and Pearson, 1991a,b). The recruitment of neuromodulatory neurons during patterned motor activity may be essential both for adapting muscle function to meet particular demands and for modulation of the motor pattern itself. One neuromodulator with multiple functions within the insect nervous system is the biogenic amine octopamine (Evans, 1980, 1985).

Biochemical (Evans and O’Shea, 1978) and immunohistochemical (Konings et al., 1988; Stevenson et al., 1992) methods reveal a group of peripherally projecting modulatory neurons in the locust, the efferent unpaired median (UM) neurons, which contain octopamine. These neurons have also been identified in many other insect species (Plotnikova, 1969; Crossman et al., 1971, 1972; Hoyle, 1978; Stevenson and Spörhase-Eichmann, 1995; Brookes, 1988; Brookes and Weevers, 1988) and, from the dorsal or ventral position of their somata, they have been named dorsal unpaired median (DUM) (Hoyle, 1975, 1978) or ventral unpaired median (VUM) (Kondoh and Obara, 1982; Watson, 1984) neurons. These cells have unique bifurcating primary neurites, which give rise to two axons that innervate the same targets on both sides of a segment. Neurons with similar characteristics have been identified in the hawkmoth Manduca sexta (Pflüger et al., 1993). Larval efferent UM neurons in M. sexta undergo dendritic remodeling, but persist through metamorphosis to innervate targets in the adult (Pflüger et al., 1993).

Efferent UM neuron activity in locusts and other insects modulates the effectiveness of neuromuscular systems. The twitch tension and relaxation rate of skeletal muscle are enhanced (O’Shea and Evans, 1979), whereas myogenic contractions of visceral muscles are inhibited (Lange and Orchard, 1984a,b; Orchard and Lange, 1986; Kalogianni and Pflüger, 1992; Kalogianni and Theophilidis, 1993). Modulatory effects of octopamine on peripheral sense organs have also been reported in locusts (Ramirez and Orchard, 1990; Matheson, 1997). Octopamine is known to affect muscle metabolism (Candy et al., 1997), so UM neurons may also exert metabolic effects on target tissues.

Unpaired median neurons are also suspected to be the mediators of octopaminergic effects within central neuropils. For example, flight-like motor patterns are induced by ionophoretic or pressure injection of octopamine into specific locations within locust thoracic ganglia (Sombati and Hoyle, 1984; Stevenson and Kutsch, 1988). This may be due to the induction of plateau potentials and bursting in certain interneurons of the flight circuit (Ramirez and Pearson, 1991a,b). The possibility that these effects are ordinarily mediated by the release of octopamine from the central processes of efferent UM neurons remains open to question (Pflüger and Watson, 1995; Parker, 1996), but central release from the processes of local or intersegmental octopaminergic cells that lack peripheral axons may occur within the locust pterothoracic ganglia (Stevenson et al., 1994), and good evidence for the release of octopamine by UM neurons within the brain comes from studies of olfactory learning in honeybees (Hammer, 1993).

Efferent UM neurons are activated during the flight motor pattern of locusts (Ramirez and Orchard, 1990), during walking in crickets (Gras et al., 1990) and by ventral giant interneurons in the cockroach (Pollack et al., 1988). Recent studies in the locust provide evidence that specific patterns of recruitment occur during different motor tasks. For example, only three of the nineteen efferent UM neurons in the metathoracic ganglion are strongly activated in parallel with motor circuits during hindleg kicking (Burrows and Pflüger, 1995), whereas another distinct sub-population is inhibited during the same movement. The same sub-population of UM neurons is also inhibited during locust flight (Duch et al., 1997). In another example, efferent UM neurons that project to leg muscles generate bursts of action potentials with each cycle of activity during the pilocarpine-induced rhythmic leg motoneuron pattern in locusts (Baudoux et al., 1998). The sources of UM neuron drive include sensory neurons and unidentified intrasegmental and intersegmental interneurons (Duch and Pflüger, 1995; 1996; Duch et al., 1997). One source of efferent UM neuron drive may be located within the subesophageal ganglion. In the absence of patterned motor activity, synaptic inputs that occur in common in all efferent UM neurons were observed in M. sexta, and their source was located to within the subesophageal ganglion (Pflüger et al., 1993). Similar synaptic inputs have been described in the silkmoth (Brookes and Weevers, 1988) and locust (Watson, 1984; Pflüger and Watson, 1995), but their role in activating UM neurons during movement remains unclear.

One strategy for characterizing patterns of modulatory neuron recruitment is to record from identified UM neurons during rhythmic fictive motor patterns that are produced by the perfusion of the muscarinic agonist pilocarpine (Baudoux et al., 1998; Büschges et al., 1995; Chrachri and Clarac, 1990; Ryckebusch and Laurent, 1993, 1994). Pilocarpine induces a pronounced inter-and intrasegmentally coordinated motor pattern, called fictive crawling, in isolated nerve cords from M. sexta larvae (Johnston and Levine, 1996). The motor pattern is qualitatively and quantitatively similar to that associated with crawling (Johnston and Levine, 1996). During each cycle of this fictive crawling, the motoneurons that innervate the body wall muscles produce bursts of activity that proceed in a posterior-to-anterior sequence through the abdominal segments. The posterior-to-anterior sequence continues as motoneurons innervating the levator and depressor muscles of the three pairs of thoracic legs sequentially produce alternating bursts of activity during each cycle. The activity is synchronous in motoneurons that innervate the right and left legs or sides of the body. Here, we take advantage of this system in M. sexta to characterize the pattern of efferent UM neuron recruitment during fictive crawling. A preliminary account of this work has been presented (Johnston et al., 1997).

Manduca sexta larvae, obtained from a colony at the University of Arizona, were reared on an artificial diet (modified from Bell and Joachim, 1976) and maintained on a 17 h:7 h light:dark photoperiod at 26°C and 50–60 % relative humidity. Larvae were used for experiments on the day following the molt into the fifth (final) larval instar (mass approximately 5–6 g).

The techniques employed in this study were similar to those described previously (Johnston and Levine, 1996). Animals were anesthetized on ice for 30–50 min before dissection. All dissections were carried out under chilled saline consisting of (in mmol l⁻¹) 140 NaCl, 5 KCl, 4 CaCl₂, 28 glucose, and 5 Hepes; pH 7.4 (Trimmer and Weeks, 1989). Anesthetized larvae were pinned dorsal side up in a dish lined with silicone elastomer (Sylgard; Dow Corning). A dorsal incision was made from the head to the most posterior segment, and the gut was removed. Specific nerve branches of each ganglion were dissected free from the surrounding tissue. The nerve cord from the subesophageal ganglion to the terminal abdominal ganglion was dissected from the body and pinned ventral side up in a second Sylgard-lined dish filled with chilled saline. The ventral surface of each ganglion was treated briefly with 3 % (w/v) collagenase–dispase (Sigma) in saline and desheathed using fine forceps. The isolated nerve cord was perfused continuously with saline or pilocarpine-HCl (10⁻³ mol l⁻¹; Sigma) dissolved in saline, both at room temperature (21–23°C) (flow rate approximately 15 ml min⁻¹). During some experiments, the subesophageal ganglion was removed by cutting the interganglionic connectives anterior to the prothoracic ganglion. To accomplish this without disturbing ongoing intracellular recordings in more posterior ganglia, the prothoracic ganglion was pinned to the recording chamber but the subesophageal ganglion remained loose, thereby allowing the connectives to be cut with minimal movement and without loss of intracellular recordings.

To monitor motor activity in the isolated nerve cord, extracellular glass-tipped suction electrodes were used to record from selected nerve branches of specific thoracic and abdominal ganglia. In the thorax, we monitored the motor activity from the two main branches of the leg nerve. The 1b branch carries the axons that innervate the femoral levator and the tibial extensor. The axons in the 2a branch innervate femoral depressors and the flexors of the more distal leg segments (Johnston and Levine, 1966; Kent and Levine, 1988). In the abdomen, we focused on the posterior branch of the dorsal nerve (DNp), which carries the axons of motoneurons that innervate the lateral, ventral and oblique intersegmental muscles of the body wall (Levine and Truman, 1985). The extracellular signals were amplified with differential a.c. amplifiers (A-M Systems), filtered, and stored on a Vetter PCM eight-channel recorder for subsequent analysis. For intracellular recordings, motoneuron somata were impaled with glass micropipette electrodes (resistance 30–60 MΩ) filled with 2 mol l⁻¹ potassium acetate. Intracellular recordings were amplified (Axoclamp 2A; Axon Instruments) and stored on video tape. Both intracellular and extracellular signals were played back onto a chart recorder and subsequently transferred to a computer (acquisition rate 13 kHz per channel) for analysis using Data-Pac II software (Run Technologies).

To reveal the dendritic anatomy of abdominal and thoracic UM neurons, and to confirm their identification by visualizing axonal projection pathways, UM neuron somata were impaled with glass microelectrodes filled with 3 % biocytin (resistance 30–50 MΩ). Depolarizing current pulses of 5–10 nA amplitude and 500 ms duration were injected into the cells at a rate of 1 Hz for 20–60 min. To allow staining of the UM cell axons in the periphery, the preparations were left first under constant saline flow for 60 min and then kept at 7°C for another 6–8 h. Preparations were processed as described previously (Consoulas et al., 1996), then photographed with a compound microscope or drawn using a camera lucida attachment.

The efferent UM neurons within the segmental ganglia of M. sexta larvae have prominent cell bodies located on the posterior midline and close to the ventral surface of the ganglion. Three UM neurons are located in every thoracic ganglion. Two of these project in the dorsal segmental nerve to body wall muscles, while the third enters the thoracic legs within nerve 2a (C. Consoulas, R. M. Johnston, H.-J. Pflüger and R. B. Levine, unpublished obervations). Two UM neurons are found within each of the unfused abdominal ganglia. Although they have similar dendritic structures, one projects into the lateral and posterior branches of the dorsal segmental nerve, whereas the other projects into the anterior branch exclusively; however, both innervate body wall muscles (Pflüger et al., 1993). Intracellular recordings from the soma of any of these efferent UM neurons reveal broad overshooting action potentials, a characteristic hyperpolarizing undershoot (see Fig. 1A,B) and prominent excitatory postsynaptic potentials (EPSPs) (see Figs 3–5).

In the presence of the muscarinic agonist pilocarpine, isolated ventral nerve cords from M. sexta larvae display a robust motor pattern that is qualitatively and quantitatively similar to that associated with crawling (Johnston and Levine, 1996). The fictive crawling motor pattern is distinguished by bursts of activity in body wall motoneurons that proceed in a posterior-to-anterior sequence through the abdominal segments. Subsequently, during each cycle, the motoneurons innervating the levator and depressor muscles of the three pairs of thoracic legs sequentially produce alternating bursts of activity, which are synchronous in nerves that innervate the right and left legs or sides of the body. Of the 45 isolated nerve cord preparations used for this study, 28 displayed patterned activity in response to pilocarpine that could be identified as fictive crawling using these features.

In 15 of the preparations in which the isolated nerve cords produced a clear fictive crawling pattern, intracellular recordings from several different identified efferent UM neurons revealed patterned activity, with depolarizations and action potential bursts accompanying each cycle of motor activity in the extracellular nerves (Fig. 1). In the remaining 13 preparations, there were rhythmic bouts of depolarizing synaptic inputs to the UM neurons, but these had a minimal effect on the patterning of action potential activity. Since all the UM neurons monitored within a preparation behaved in a similar manner, the difference was dependent on the preparation rather than on the specific identity of the UM neuron. In the former group of preparations, each depolarization of the membrane potential led to 1–3 action potentials in the UM neurons (Fig. 1A,B). Hyperpolarizing the UM neuron membrane potential with constant-current injection prevented action potential production, but the rhythmic depolarizations increased in amplitude, suggesting that they were dependent on synaptic drive (Fig. 1C).

The peak membrane depolarization and action potential activity in several different types of UM neuron, including those innervating the abdominal and thoracic body-wall muscles (Fig. 1A,C) or the thoracic leg muscles (Fig. 1B), occurred during the phase of the crawling cycle when there were bursts of motoneuron activity in the thoracic leg levator nerves (Figs 1A–C, 2). Furthermore, there was substantial overlap in the timing of membrane depolarization and action potential bursts in simultaneous recordings from several different pairs of UM neurons (Figs 1A, 2, 3B), even when they were separated by several segments. Thus, the rhythmic depolarization of efferent UM neurons during fictive crawling did not necessarily occur in phase with activity in the excitatory motoneurons that project to the same target muscles, although there was usually some overlap. For example, a phase plot constructed from 39 cycles of activity (3.9 min) of the preparation shown in Fig. 1A revealed overlapping peaks of action potential activity in UM neurons in A1 and T3 during the phase of the cycle when leg levator motoneurons were active (Fig. 2). Nevertheless, owing to the long duration of abdominal motoneuron bursts, there was overlap between activity in the UM neuron in A1 and the activity of excitatory motoneurons that project to the same muscles.

In the isolated ventral nerve cord with the subesophageal ganglion attached, prominent EPSPs occurred in common, but with an anterior-to-posterior delay, in all the efferent UM neurons (Pflüger et al., 1993). In the absence of pilocarpine, these EPSPs occurred continuously, with no apparent patterning, as each UM neuron reached its action potential threshold at different times (Figs 3A, 4A). After the addition of pilocarpine and the onset of fictive crawling, the prominent EPSPs increased in frequency and became clustered into bursts that occurred with each cycle of activity in the motor roots (Figs 3B, 4Bi,ii, 5C,D). These caused compound depolarizations that usually led to action potentials in the UM neurons. As was true of the EPSPs that occurred in the absence of pilocarpine (see also Pflüger et al., 1993), there was an anterior-to-posterior delay between the appearance of the EPSPs in efferent UM neurons of different segments (Fig. 4Bi,ii), but this delay was brief relative to the cycle period. The delay of approximately 40 ms for EPSPs in A2 and A6 yields a conduction velocity of approximately 3 m s⁻¹, a rate similar to that calculated in a previous study (Pflüger et al., 1993).

The prominent EPSPs that occur in the absence of pilocarpine are not dependent upon the brain, but are lost following removal of the subesophageal ganglion (Pflüger et al., 1993). Along with the anterior-to-posterior delay between their occurrence in different segments, this suggests that the EPSPs are generated by neurons that descend from the subesophageal ganglion. Similarly, during pilocarpine-induced fictive crawling, the rhythmic bursts of prominent EPSPs in the efferent UM neurons were dependent upon the presence of the subesophageal ganglion. Following removal of the subesophageal ganglion in preparations that were displaying fictive crawling, the bursts of prominent EPSPs in the UM neurons ceased (N=13; Figs 3C, 4C, 6B,C). In some cases, the UM neurons continued to fire irregularly following removal of the subesophageal ganglion (Fig. 3C), whereas in other cases weak membrane potential oscillations and small EPSPs, perhaps reflecting local segmental inputs, were apparent (Fig. 6B,C).

The patterned motoneuron activity that occurs in the segmental nerves during fictive crawling does not depend upon the subesophageal ganglion, nor does the subesophageal ganglion have to be exposed to pilocarpine for fictive crawling to be evoked (Johnston and Levine, 1996). The prominent EPSPs that caused rhythmic depolarization of the UM neurons during fictive crawling, however, depended on the subesophageal ganglion. This raises the possibility that pilocarpine acts on the subesophageal ganglion to increase the descending drive of the efferent UM neurons.

To test this possibility, isolated nerve cords were placed in chambers containing petroleum jelly barriers that allowed subsets of ganglia to be exposed selectively to pilocarpine (N=2; Fig. 5A). Fictive crawling activity was still evoked in the motor roots following exposure to pilocarpine of only those segmental ganglia posterior to A1 (Fig. 5B,C). Moreover, the increased frequency of prominent EPSPs (Fig. 5B), and their grouping into bursts, still occurred in the efferent UM neurons whether or not they had been exposed directly to pilocarpine (Fig. 5C,D). Following removal of the subesophageal ganglion in this preparation (Fig. 6A), the fictive crawling pattern continued in the motor roots, but became more irregular (Fig. 6B). The prominent EPSPs disappeared, although small EPSPs and membrane potential depolarizations of the UM neuron still occurred and were often correlated with bursts of activity in the motor roots (Fig. 6B,C). Although the UM neuron gradually depolarized by approximately 10 mV following removal of the subesophageal ganglion in this preparation, this did not account for the absence of the prominent EPSPs. As with the preparation shown in Figs 3 and 4, the prominent EPSPs disappeared immediately after removal of the subesophageal ganglion and were not revealed when the membrane potential was hyperpolarized with current injection. Thus, although a significant source of rhythmic synaptic drive to efferent UM neurons is derived from the subesophageal ganglion during fictive crawling, the activation of this drive does not require the direct exposure of the subesophageal ganglion to pilocarpine.

Efferent UM neurons are activated rhythmically during fictive crawling so that, in most preparations, depolarizations and action potentials occur with each cycle of activity in the body-wall and thoracic leg motoneurons. This activity is relatively synchronous in all the segmental efferent UM neurons regardless of the muscles innervated by their peripheral axons. Thus, the two efferent UM neurons within each abdominal ganglion are active together with those in the thoracic ganglia, including the one that projects to the muscles of the thoracic legs. The efferent UM neurons were depolarized and produced action potentials during the portion of the crawling cycle when the thoracic leg levator motoneurons were active. Similarly, in the presence of pilocarpine, UM neurons that innervate leg muscles in the locust were driven rhythmically to produce bursts of action potentials during bursts of activity in the levator motoneurons (Baudoux et al., 1998).

The functional significance of this phasing is unknown at present, but the efferent UM neurons were not necessarily activated together with the excitatory motoneurons that project to the same muscles (Johnston and Levine, 1996). Although the thoracic leg UM neuron projects in nerve 2a (C. Consoulas, R. M. Johnston, H. J. Pflüger and R. B. Levine, unpublished obervations), which innervates the depressor muscles of the larval leg, it was depolarized during the phase of the fictive crawling cycle when all leg levator motoneurons were active. Nevertheless, octopamine that is released peripherally into the vicinity of the target muscles as a result of UM neuron activity may have relatively long-term effects on both the contraction amplitude and relaxation rate of the muscle (Evans and O’Shea, 1978), although these effects have yet to be demonstrated in M. sexta. Thus, precise coupling of UM neuron activity to that of the muscles that they innervate may be unnecessary, as long as octopamine release is ensured during periods of crawling. Alternatively, the coupling observed in the isolated central nervous system (CNS) preparation may be different from that in an intact animal during crawling. For example, segmental sensory input may force efferent UM neuron activity into a clear phase relationship with the motoneurons innervating specific muscles. Another possibility is that UM neuron activity during crawling is important for the release of octopamine into the CNS, which may augment rhythmic motor activity. Although few central output sites were identified in an ultrastructural analysis of efferent UM neurons in the locust (Pflüger and Watson, 1995), central effects on motoneurons in this preparation have been described (Parker, 1996). However, current injection into UM neurons did not alter the rhythmic motor activity induced in locust leg motoneurons by pilocarpine (Baudoux et al., 1998).

The rhythmic depolarizations during fictive crawling in M. sexta are evoked primarily by prominent EPSPs that occur in common in abdominal and thoracic efferent UM neurons during pairwise recordings. These may be generated by the same unidentified presynaptic neurons that are the source of the EPSPs that occur in common during pairwise UM neuron recordings in the absence of fictive crawling (Pflüger et al., 1993). In both cases, the EPSPs are dependent upon the presence of the subesophageal ganglion and occur with an anterior-to-posterior delay. In the presence of pilocarpine, these prominent EPSPs increase in frequency and become grouped into bursts that cause membrane potential depolarization and spiking of the efferent UM neurons. However, the rhythmic activity of UM neurons during fictive crawling does not depend exclusively on drive from the subesophageal ganglion. Following removal of the subesophageal ganglion, weak depolarizations of the UM neurons, which may reflect segmental inputs, occurred with bursts of motoneuron activity. Although the prominent EPSPs in efferent UM neurons are dependent upon the subesophageal ganglion, their increase in frequency and grouping into bursts during fictive crawling do not depend upon exposure of the subesophageal ganglion to pilocarpine. Both occur upon exposure of more posterior ganglia, which also evokes fictive crawling. One hypothesis that is consistent with this observation is that descending interneurons within the subesophageal ganglion, which evoke EPSPs in all efferent UM neurons, receive phasic excitatory or inhibitory drive from the crawling central pattern generator. The anterior-to-posterior delay between the onset of EPSPs in UM neurons in two different segments is constant and is consistent with that expected as a result of action potential conduction between segments (Pflüger et al., 1993). Nevertheless, there may be multiple descending interneurons from the subesophageal ganglion that mediate these prominent EPSPs, and they need not synapse directly onto the UM neurons within the segmental ganglia.

In this context, it is noteworthy that the subesophageal ganglion has been regarded as an important integrative center for locomotory control, although the experimental basis for this idea is rather limited (see Burrows, 1996). Decerebrated stick insects with both circumesophageal connectives cut exhibit long bouts of walking activity, suggesting an inhibitory influence of the brain onto locomotory centers in the subesophageal ganglion. In contrast, stick insects with the neck connectives (i.e. just posterior to the subesophageal ganglion) cut do not express any walking activity (B ässler, 1983). From electrical stimulation of fiber populations in the circumesophageal and neck connectives, Kien (1983) postulated effects on many different aspects of walking such as initiation, speed, step length and coordination. In a subsequent paper, a few descending interneurons of the subesophageal ganglion were filled with dye and their structure was revealed (Kien and Altman, 1984). Descending interneurons of the subesophageal ganglion are also involved in the control of locust flight (Ramirez, 1988), of locust (Ramirez and Pearson, 1989) and cricket (Otto and Januszewski, 1989) ventilation and of stridulation in the cricket (Otto and Hennig, 1993) and grasshopper (Hedwig, 1986). Our results suggest that the subesophageal ganglion may also function to control the excitability of neuromodulatory neurons.

The existence of EPSPs that occur in common in all efferent UM neurons in larval M. sexta, and their role in providing common drive during fictive crawling, is in contrast to results from the adult locust, where the total population of UM neurons is divided into sub-populations that are only active during specific patterns of motor behavior (Burrows and Pflüger, 1995; Duch and Pflüger, 1995, 1996; Baudoux et al., 1998) and where sub-populations exist that are specifically inhibited during ongoing motor behavior such as flight (Duch et al., 1997). In the locust, these sub-populations are defined by the presence (or absence) of common EPSPs, which come either from local sources within the same ganglion or from the subesophageal ganglion (Duch and Pflüger, 1995, 1996). In contrast to M. sexta, sub-populations of UM neurons exist in the locust that do not rely on intersegmental inputs from the subesophageal ganglion for patterned activation. For example, in a completely isolated metathoracic ganglion, only UM neurons innervating the leg muscles were coupled to the rhythmic leg motoneuron activity induced by pilocarpine (Baudoux et al., 1998).

There are several possible reasons for the more specific recruitment strategies in the locust and for the greater importance of local segmental drive of efferent UM neurons during pilocarpine-induced rhythmic leg motor activity. First, the metathoracic ganglion of the adult locust is involved in several distinct types of behavior, which may require unique patterns of UM neuron activity or specific sets of UM neurons. Second, fictive crawling in M. sexta shows well-structured intersegmental coordination that, unlike adult walking, includes the abdominal segments. Finally, the pilocarpine-induced pattern in the locust leg motoneurons involves bilateral alternation (Ryckebusch and Laurent, 1993), rather than synchronous activation across the ganglion as in fictive crawling in M. sexta (Johnston and Levine, 1996). Locust DUM neurons, which project bilaterally to innervate both legs, couple to the rhythmic leg motoneuron pattern in both hemiganglia (Baudoux et al., 1998). Although common drive to all efferent UM neurons from the subesophageal ganglion remains in M. sexta adults (Pflüger et al., 1993), it will be interesting to determine whether other sources of synaptic input cause the UM neurons that project to different targets to display unique activity patterns during adult walking and flight behavior.