Correlation Between Muscle Structure and Filter Characteristics of the Muscle–Joint System in Three Orthopteran Insect Species

Investigations in the field of neuroethology have led to a detailed insight into the neuronal mechanisms underlying behaviour and behavioural specializations (e.g. Burrows, 1984; Breidbach and Kutsch, 1995; Schildberger and Elsner, 1994). Besides analysing the neuronal basis of specific behaviours in individual species, numerous investigations have also tried to compare neuronal networks in related species exhibiting (i) the same behaviours (e.g. Hennig, 1994; Stumpner et al. 1995; von Helversen and von Helversen, 1994; Schildberger, 1994; Elsner, 1994; Michelsen, 1994) or (ii) different behaviours due to different ethological requirements (Ebner and Bässler, 1978; Büschges and Wolf, 1995). Most of these investigations have dealt with the generation of the motor output pattern (e.g. Robertson and Olberg, 1988; Katz and Tazaki, 1992). However, the motor output is not the same as the movement itself. Therefore, we here attempt to correlate features of the neuromuscular system with details of movements for which the ethological significance is known.

The femur–tibia (FT) control system in orthopterans is well suited for such a comparative investigation for three reasons: (i) it is one of the best-known joint control loops (see Bässler, 1993, for summaries and comparisons with other joint control loops). (ii) The FT control loop in phasmids and locusts apparently consists of the same neuronal elements and connections (e.g. Burrows, 1989, 1992; Büschges and Wolf, 1995). (iii) There are strong quantitative differences in the performance of the reflex loop between different species that enable behavioural specializations (Bässler, 1983a, and below). These can, at least in part, be attributed to different ecological niches.

Phasmids exhibit catalepsy as one of the behavioural components of twig mimesis (see Bässler, 1983a), while locusts show no catalepsy. To be able to generate catalepsy, the joint control loops of stick insects must have a very low position sensitivity combined with a high non-linear velocity-dependency and a high gain (see Bässler, 1993). In contrast, locust joint control loops are more linear, have a considerably smaller velocity-sensitivity and a rather low gain (Ebner and Bässler, 1978; Büschges and Wolf, 1995; Field and Burrows, 1982). These differences are most obvious in the ‘open-loop configuration’ in ‘inactive’ animals (for a definition of inactive, see Bässler, 1983a). Open-loop configuration in the case of the FT joint means that the apodeme of the femoral chordotonal organ (fCO), the sensor of the FT control system, is cut. The fCO is then stimulated mechanically and the movement of the tibia generated by the evoked resistance reflex is monitored. From the movement responses to sinusoidal stimulation of the fCO, amplitude–frequency plots can be calculated which describe the dependency of the amplitude of the induced joint movement on stimulus frequency (see Fig. 1B). In such a plot, two characteristics are especially important, the maximum response amplitude and the stimulus frequency above which the response amplitude declines.

The response amplitude can be described by the dimensionless term ‘gain’ (output amplitude divided by input amplitude) since the mechanical stimulus to the fCO can be converted into the joint angle variation it signals (for details, see Bässler, 1983a). Because gain is independent of body size, it can be used in comparisons between different species.

To characterise the stimulus frequency above which there is a decline in response amplitude, the ‘upper corner frequency’ is normally used. This value is a measure of the fastest stimulus frequency that results in a movement by the joint control network. In linear systems, it is defined as the specific frequency at which a regression line through the declining part of the amplitude–frequency plot meets a horizontal line drawn through the maximal response amplitude. Since biological systems are not completely linear, only approximations of their upper corner frequencies can be obtained and they are usually taken as the frequency at which the joint control system produces a movement with 70 % of its maximal response amplitude.

The open-loop FT control system of phasmids has a high maximal gain value (up to nearly 10; Bässler, 1983a; Weiland et al. 1986). The upper corner frequency differs among species. For Carausius morosus, it is approximately 0.5–1.0 Hz (Bässler, 1972; Bässler et al. 1974; Kittmann, 1984, 1991) but for Cuniculina impigra it is in the range 0.1–0.2 Hz (Bässler and Foth, 1982). In contrast, the FT control system of the locust middle leg has a maximal gain of less than 1 and an upper corner frequency of less than 0.05 Hz (Ebner and Bässler, 1978). The gain and upper corner frequency of a control system affect each other; for example, control systems with high gain tend to become unstable under closed-loop conditions if the upper corner frequency is too high (for a detailed discussion, see Bässler, 1983a). Therefore, the relatively low upper corner frequency of the phasmid FT control loops protects them from instability and can be considered to be a consequence of the high gain that is necessary for the generation of catalepsy.

The question arises as to whether these differences between species mainly correlate with properties of the neuronal networks controlling the activity of the motoneurones in the FT joint or whether they are also correlated with species-specific properties of the muscles and joints involved. For both phasmid species, the dependence of the firing frequency of the extensor tibiae motoneurones upon fCO stimulus frequency is known (see summary in Bässler, 1983a). Amplitude– frequency characteristics of the responses of the slow extensor tibiae motor neurone (SETi) show that SETi firing frequency increases steadily up to a stimulus frequency of 5 Hz. Because (i) SETi is the only neurone determining extensor force under these conditions and (ii) the force of the flexor tibiae has similar characteristics (Storrer and Cruse, 1977), the decline in movement amplitude with higher fCO stimulus frequencies in phasmids must originate from properties of the tibial muscles and/or the leg joint itself. In the following, these properties are summarised by the term ‘muscle–joint system’. The muscle–joint systems of both phasmid species used in this study apparently form low-pass filters that transform only slow modulations of SETi firing into appropriate movements.

To obtain an understanding of the factors affecting the upper corner frequencies of phasmid muscle–joint systems, we used the following arguments. (i) If the structure or the arrangement of the muscle fibres is involved in the generation of the upper corner frequency of the muscle–joint system, there should be a correlation between appropriate structural variables and the upper corner frequencies of Carausius morosus and Cuniculina impigra. (ii) The locust middle leg, as an unspecialised walking leg, should not possess specializations that are directly or indirectly necessary for catalepsy. Thus, the amplitude–frequency characteristics of the locust muscle–joint system should be compared with those of the phasmid species.

We focused on the extensor tibiae muscle for studying putative correlations between muscle structure and joint control properties because the extensor tibiae muscle in all three species has a much simpler innervation than the flexor muscle (Burns and Usherwood, 1978, 1979; Debrodt and Bässler, 1989; Theophilidis and Burns, 1983). We investigated and compared (i) the arrangement of the muscle fibres, (ii) the distribution of the innervation patterns of the muscle fibres and (iii) the distribution of muscle fibre types by histochemical staining for myofibrillar ATPase (mATPase) according to established procedures (Müller et al. 1992). In the companion paper (Bässler and Stein, 1996), we then analyse possible causes of the reported correlation and their consequences for other properties of this system.

Adult females of two stick insect species, Carausius morosus Brunner and Cuniculina impigra Redtenbacher (syn. Baculum impigrum Brunner), and adults of both sexes of the locust Locusta migratoria L. were obtained from colonies at Kaiserslautern University, Germany.

The experimental animal was fixed dorsal side up on a foam platform and its right middle leg was fixed, using dental cement (Protemp, ESPE), perpendicular to the thorax on a foam platform. Preparation of the femur and cutting and mounting of the receptor apodeme of the fCO in a stimulation clamp were performed according to previously published procedures (e.g. stick insect: Bässler, 1983a; locust: Büschges and Wolf, 1995). The fCO was stimulated (Fig. 1) conventionally using a pen-motor (Hellige) and a sine-wave generator (Tektronix), as described in detail in Bässler (1986). In Cuniculina impigra, a movement of the fCO apodeme by 400 μm corresponds to a change in joint angle of 40 °, while in Carausius morosus 200 μm corresponds to 40 ° (Weiland et al. 1986). In Locusta migratoria, the fCO apodeme had to be moved by approximately 600 μm for the same joint movement of 40 ° (Field and Pflüger, 1989).

The fCO was stimulated using sinusoidal stimuli (Fig. 1B) of 40 ° amplitude around a centre position that corresponded to the 90 ° position of the FT joint. Stimulation frequency ranged from 0.01 to 5 Hz. Elongation of the fCO mimics imposed flexion of the FT joint and relaxation of the fCO mimics imposed extension. SETi activity was recorded (see below). As a measure of SETi response amplitude, the number of spikes that occurred during the second half of stretching was counted because the spike frequency is maximal at this time (see inset in Fig. 1B). The spike frequency could be used as a measure of response amplitude because the minimum spike frequency during fCO relaxation reached zero in all cases. As the FT control system is known to show marked habituation (Kittmann, 1991), the animal was touched with a brush between each stimulation sequence at a given frequency in order to dishabituate the experimental animal. Only the first five full cycles were considered for each stimulus frequency. The values obtained (Fig. 1B) are expressed as spikes s⁻¹ and they give the average activity of SETi for the first five cycles.

Extracellular recordings were made conventionally either by pressing the nerve against a trachea using myogram electrodes with the insulation removed from the tips (for details, see Pflüger, 1977) or by using a paraffin-oil hook electrode (Schmitz et al. 1988) to record en passant the activity of the extensor nerve.

Intracellular recordings from muscle fibres were conventionally obtained using electrodes filled with 3 mol l⁻¹ KCl for the phasmids (in order to be able to compare the data with previous recordings; Bässler and Storrer, 1980) or 2 mol l⁻¹ potassium acetate for the locust, and having a resistance of 20–50 MΩ. We could not find differences in the parameters measured between electrolytes (e.g. resting membrane potential). For identification of muscle fibres, the electrodes were filled with 4 % Lucifer Yellow (tip) and 1 mol l⁻¹ LiCl (shaft). The experimental animals were fixed either dorsal-side-up (Carausius morosus, Cuniculina impigra) or ventral-side-up (all three species) on a foam platform. The femur of the middle or hindleg was fixed horizontally with dental cement (Protemp, ESPE) on a foam platform at an angle of approximately 90 ° to the body’s longitudinal axis. The parts of the femur that were dissected were surrounded by a small basin filled with saline. In the dorsal-side-up preparations of the phasmids, a small window was cut into the dorsal cuticle of the femur to expose the muscle fibres. Care was taken to ensure that only a small number of muscle fibre origins had to be removed. In the ventral-side-up preparations, the flexor tibiae muscle, the crural nerve and the retractor unguis apodeme were removed, together with most of the ventral cuticle in that region of the leg. The large leg trachea (supplying the flexor tibiae muscle) was cut distally and the cut end was raised above the surface of the saline so that saline could not be sucked into it. The diaphragm between the flexor and extensor haemolymph space was carefully removed, leaving the smaller leg trachea (supplying the extensor tibiae muscle) intact.

Autotomised or cut middle and hindlegs were opened on the ventral side over the whole length in locust or Carausius saline (Bässler, 1997). The tibiae were cut off distal to the FT joint, the tendon of the flexor tibiae muscle was cut at mid-femur; hence, the diaphragm separating the extensor tibiae and flexor tibiae muscles was destroyed at this point. The leg was dried on cellulose tissue, embedded in Tissue Tec medium and shock-frozen in an aluminium mould in liquid nitrogen. Preparations were mounted with Tissue Tec on a cold chuck of a microtome (Reichert and Jung) at −15 to −20 °C. Serial cross sections (30 μm) of the whole leg (including flexor tibiae and extensor tibiae muscles) were made. Individual cross sections were then placed alternately onto two different cold, glycerine–albumin-coated microscope slides (both middle and hindlegs). They were quickly thawed and air-dried on a heating plate at 40 °C.

Staining for mATPase activity was performed according to Padykula and Hermann (1955) with modifications adopted from Müller et al. (1992). The reaction medium was adjusted to pH 8.4 (incubation time 10–15 min). Possible isoforms of mATPase were detected on the basis of their different pH stability. Therefore, the sections on one microscope slide were preincubated for 5 min at pH 4.7 and on the other slide for 10 min at pH 10.2 before they were transferred to another reaction medium at pH 9.4 (incubation time 30–45 min; Brooke and Kaiser, 1970; Guth and Samaha, 1970).

Mean values of samples with N<30 were compared using a modified t-test according to Dixon and Massey (1969). Binomial distributions were compared using Fisher’s test. Means and samples were regarded throughout the investigation as significantly different when P<0.01.

In order to estimate the filter characteristics of the muscle–joint system in the FT joint, the amplitude–frequency characteristics of the activity of the extensor motoneurone during sinusoidal stimulation of the fCO must be known. For the two phasmid species, Carausius morosus and Cuniculina impigra, the amplitude–frequency characteristics of the slow extensor tibiae motoneurone (SETi) have been reported previously. For both species, the SETi response increases up to stimulus frequencies of approximately 5 Hz and decreases at higher stimulus frequencies (Carausius morosus: Bässler, 1983a; Cuniculina impigra: Bässler et al. 1982; Bässler, 1983b; summarized in Bässler, 1983a). However, these data were collected in different ways; for example, by measuring the maximum spike frequency, the number of spikes per cycle or the modulation in soma membrane potential. In addition, for locusts, only the amplitude–frequency characteristics of tibial movements have been published (Ebner and Bässler, 1978) while the frequency-dependence of the motoneuronal response is not known. Thus, to obtain the best comparison possible, the amplitude–frequency characteristics were re-investigated in all three species and the data were evaluated in the same way (see Materials and methods section).

In Carausius morosus and Cuniculina impigra, the SETi firing rate increased steadily with increasing stimulus frequency (Fig. 1B), thus confirming previous reports (Bässler, 1983a,b; Bässler et al. 1982; Kittmann, 1984). In contrast, SETi firing rate in Locusta migratoria dropped significantly between stimulus frequencies of 0.01 and 0.3 Hz. SETi activity showed a significant increase compared with the value at 0.01 Hz only at higher stimulus frequencies (>1 Hz).

Comparisons between the amplitude–frequency characteristics of SETi (Fig. 1B) and values for gain of tibial movement taken from the literature (Fig. 1C) confirm that in both phasmid species the upper corner frequency in the movement response occurs some way down the slope of the increase in SETi activity with stimulus frequency. Thus, the upper corner frequency in these phasmids must be a result of the filter characteristics of the muscle–joint system. These filter characteristics differ for the two species investigated.

In locusts, the upper corner frequency (<0.05 Hz; Ebner and Bässler, 1978) appears, at least in part, to be determined by neuronal mechanisms that are also responsible for the decrease in SETi activity with increasing stimulus frequencies between 0.01 and 0.3 Hz. This decrease in SETi activity corresponds to the much smaller velocity-dependence of the locust FT control loop found at the motoneuronal as well as the interneuronal level (for details, see Büschges and Wolf, 1995). The comparison also shows that the higher gain in the movement response (Fig. 1C) of the phasmids is mirrored by a higher SETi spike frequency (Fig. 1B), i.e. the high gain of the joint control system in phasmids is, at least in part, a neuronal attribute.

The extensor tibiae is a pinnate muscle which extends over almost the entire length of the femur (length of femur; Cuniculina impigra, middle leg 2.0–2.5 cm, hindleg 2.4–3.0 cm; Carausius morosus, middle leg 1.0–1.3 cm; hindleg 1.2–1.6 cm; Locusta migratoria, middle leg 0.8–1.0 cm). In locusts, accessory extensor and flexor muscles are known in the tibiae of the middle and hindleg (e.g. Theophilidis and Burns, 1979); however, these were not investigated in the present study.

In both phasmid species, the extensor muscle fibre bundles originate only from the dorsal roof of the femur, except in the most distal region (Fig. 2; Bässler and Storrer, 1980). To aid comparison among the three species in this study, the extensor muscle was divided into four regions, which were defined with a FT joint angle of approximately 90 ° (Fig. 1A). Region a covered the most proximal third of the muscle, ending in the region of the beginning of the extensor muscle apodeme. Region b covered the middle portion of the muscle from approximately 30 % to approximately 85 % of its length. The most distal portion, where we identified in both phasmids a previously undescribed row of muscle fibres originating on the posterior cuticle, was divided into two regions (Fig. 2). The posterior row of muscle fibres started at approximately 85 % of the muscle length, and region c was defined as the anterior part of the muscle and region d as the posterior part. This distal posterior row (region d) is larger in Cuniculina impigra than in Carausius morosus, but even in Cuniculina impigra it only occupies approximately 10–15 % of the muscle length (Fig. 2). The fibres of this row are separated from the rest of the extensor muscle by the small leg trachea, which runs dorsal to these fibres. In both middle and hindlegs of the phasmids, the extensor muscle has the same anatomy.

In Locusta migratoria middle legs, extensor muscle fibre bundles originate on the cuticle in two rows, an anterior–dorsal row and a posterior row (see also Fig. 6). The most proximal fibre bundle inserts on the dorsal roof of the femur (Burns and Usherwood, 1979). The most distal bundle in the posterior row has a considerably broader insertion base on the cuticle than the other bundles and its fibres appear to be shorter, thus showing some structural similarities with phasmids.

The innervation pattern of the extensor muscle was investigated using intracellular recordings from muscle fibres and simultaneous extracellular recordings from the extensor nerve (F2 in phasmids, N5B2 in the locust). As previously described for insects (cockroach, Atwood et al. 1969; locust, Burns and Usherwood, 1979; stick insect, Bässler and Storrer, 1980), we found three types of muscle fibre innervation: (1) only FETi (singly innervated; not shown), (2) FETi, SETi and CI₁ (triply innervated; Fig. 3A), (3) SETi and CI₁ (dually innervated; Fig. 3B). The frequency of these three fibre types changed as a function of location within the muscle (locust, Burns and Usherwood, 1979; stick insect, Bässler and Storrer, 1980).

For Locusta migratoria, in agreement with previous investigations (e.g. Burns and Usherwood, 1979), we found mostly singly innervated fibres in the most proximal fibre bundle (region a, Fig. 1A; Table 1). Towards the middle of the femur (region b, Fig. 1A), there was a decrease in the percentage of singly innervated fibres, an increase in the number of triply innervated fibres and a small, but increasing, number of dually innervated fibres (Table 1). In the distal region (regions c and d, Fig. 1A), only triply and dually innervated fibres were present (Table 1).

In contrast, in both phasmid species, we found a different innervation pattern from that previously reported (Bässler and Storrer, 1980). In the proximal region (region a), as in the locust, mostly singly innervated fibres were found (Table 1). Triply innervated fibres were present throughout the muscle, with a lower percentage at both ends (Table 1). However, small numbers of dually innervated fibres were detected over the whole length of the muscle, especially in the dorsal-side-up preparation. In the distal region, their number showed a steep and marked increase (see below). The percentage of dually innervated fibres found in the distal region in Cuniculina impigra was significantly higher than that in Carausius morosus. These results thus differ markedly from previous descriptions of fibre type distribution in the extensor tibiae of phasmids (Bässler and Storrer, 1980). These differences are likely to be due to the fact that in the present investigation muscle fibre recordings were made from the dorsal as well as the ventral side of the muscle, whereas in previous investigations on phasmids recordings were made only from the anterior side.

In the proximal (a) and middle (b) regions of the extensor muscle, there were no significant differences in the percentage of fibre types between the three species. It was only in the distal region of the extensor muscle that we detected significant differences in innervation pattern between the three species investigated. Between Locusta migratoria and Carausius morosus, we found no difference in fibre type distribution for regions c and d (Table 1). In contrast fibre type composition in regions c and d in Cuniculina impigra did differ significantly (Table 1). In region c (dorsally originating distal fibres), triply innervated and dually innervated fibres were detected in approximately equal numbers (Table 1). In region d, however, most of the fibres were dually innervated. Only a few triply innervated fibres were found and all were located at the proximal end of this region.

From the relative sizes of the excitatory junctional potentials (EJPs), there appeared to be gradual transitions between the three fibre types [e.g. for triply innervated muscle fibres (Fig. 4A–C) and dually innervated fibres (Fig. 4D) in a Cuniculina impigra hindleg]. Similar results were obtained for Cuniculina impigra middle legs and for the middle legs of Carausius morosus and Locusta migratoria.

In triply and dually innervated muscle fibres in the proximal and middle region of the extensor muscle, CI₁ usually elicited relatively small IJPs (Fig. 4; see also Fig. 8A), while they were always large in dually innervated fibres of the distal region (Fig. 3B). Thus, in many cases, IJPs could only be detected unequivocally in dually innervated fibres of the proximal and middle regions and in triply innervated fibres when they occurred at relatively depolarised membrane potentials, e.g. during spontaneous bursts of motoneurone activity (Fig. 3A). This was true for all three species investigated. In addition, in the companion paper (Bässler and Stein, 1996), it will be shown that CI₁ activity decreases the force of contraction of dually innervated muscle fibres in the distal portion of the extensor muscle much more strongly than in the other portions of the muscle, indicating that there might be different types of dually innervated fibres.

As a consequence, a distinction between dually and triply innervated fibres was clearest in the distal portion of the extensor muscle, i.e. in regions c and d (Fig. 2). However, for comparison with previous studies, we also give fibre type percentage values for the two more proximal regions in Table 1. In contrast to the basic similarities in the innervation pattern of the extensor muscle between the three species, we observed a phenomenon in Locusta migratoria muscle fibres that was not found in Carausius morosus or Cuniculina impigra muscle fibres. In 10 out of 25 dually innervated fibres and 16 out of 46 triply innervated fibres of Locusta migratoria, the frequency of EJPs induced by SETi activity was lower than the frequency of SETi action potentials recorded from the extensor nerve. This phenomenon was only present at high SETi discharge rates, exceeding 100 spikes s⁻¹ (Fig. 5). EJPs in the muscle fibre followed SETi discharge rates reliably with a 1:1 correlation up to approximately 100 spikes s⁻¹. Between 100 and 140 spikes s⁻¹, SETi spikes sometimes did not induce EJPs in the muscle fibre, leading to an EJP frequency lower than the motoneuronal discharge rate. Above 140 spikes s⁻¹, the EJP frequency in the muscle fibre was always lower than the neuronal discharge rate. Any influence of the location of the extracellular extensor nerve recording electrode (N5B2) on this phenomenon could be excluded since the recording electrode was sometimes located proximal, and sometimes far distal, to the recorded muscle fibre. Only one out of 28 singly innervated fibres showed this phenomenon for FETi. EJP failure was also observed with orthodromic stimulation of motoneurones via nerve N5 for SETi and for FETi via nerve N3B (results not shown). In three triply innervated muscle fibres, motoneurone discharge rates above 110 spikes s⁻¹ were found to be correlated with a failure of EJP occurrence. This phenomenon could, in part, be due to propagation failures of motoneurone action potentials at branching points of the extensor nerve (Theophilidis, 1988; see Discussion).

It is known that the enzymatic status of muscle fibres, i.e. their contractile protein composition, is correlated with their contraction properties (e.g. Stokes, 1987). For the locust coxa rotator muscle (M92, Snodgrass, 1929), it was shown that singly innervated fast-contracting muscle fibres have a high content of the enzyme myofibrillar actomyosin ATPase (mATPase), an mATPase isoform with a high reaction velocity and that inactivates at alkaline pH. In contrast, slow-contracting, dually innervated muscle fibres, contain an mATPase isoform with a low reaction velocity and insensitive to preincubation at alkaline pH (Müller et al. 1992). In order to characterise the muscle fibres with respect to their contraction properties, we measured mATPase activity histochemically in serial cross sections of extensor muscles. Using this technique, differences in the strength of staining correlate with differences in the reaction velocity of non-inactivated mATPase. Therefore, one can distinguish between fast- and slow-contracting fibres. The different isoforms of mATPase can be distinguished not only on the basis of staining intensity, but also on the basis of differences in their pH-stability (Morgan et al. 1980; Padykula and Hermann, 1955). We optimised the procedure for the extensor muscle of the species investigated. In all three species, the fast isoform of mATPase shows a high reaction velocity without preincubation (pH 8.4) and inactivates after alkaline preincubation (pH 10.2). The slow isoform of mATPase has a lower reaction velocity at pH 8.4 and remains active after alkaline preincubation. Therefore, so-called fast fibres stained darkly without preincubation but were unstained after alkaline preincubation and so-called slow fibres stained darkly after alkaline preincubation. As previously shown for the locust anterior coxa rotator muscle (Müller et al. 1992), some fibres (termed intermediate fibres) were lightly to strongly stained after both procedures.

At pH 8.4, the differences between the reaction velocity of the fast and the slow mATPase isoforms were smaller in the phasmid species than in the locust. Interestingly, preincubation at pH 4.7, which was reported to stain the fast isoform of mATPase and to inactivate the slow isoform in cockroach mesocoxal muscles (Stokes et al. 1979; Morgan et al. 1980), gave reasonable results in stick insects, but not in locusts (see also Müller et al. 1992). At pH 4.7 in locusts, both isoforms inactivated at the same time, precluding any distinction between them. This suggests that differences in the mATPase isoforms exist between locusts and stick insects.

Because of differences in staining quality without alkaline preincubation and after acidic preincubation, we quantitatively evaluated mATPase activity only after alkaline preincubation at pH 10.2. Therefore, any intermediate fibres would be counted as slow fibres (alkali-stable fibres).

Fig. 6 shows the distribution of slow fibres in middle leg cross sections typical for the three species. Cross sections were taken from the proximal and middle regions a and b (see Fig. 2) and from the distal region (i.e. regions c and d in Fig. 2) after alkaline preincubation. Slow fibres are clearly detectable by their dark staining, while the fast fibres remained unstained. There was an increase in the number of slow fibres from proximal to distal in all three species. The reverse was the case for the fast fibres. These results are in accordance with the gross fibre innervation pattern of the extensor muscle (Burns and Usherwood, 1979; Bässler and Storrer, 1980; this study). We evaluated the number of fibres present of each type. In each cross section, the total number of fibres and the numbers of slow and fast fibres were counted. Three or four middle legs from each species were evaluated and the results are shown in Fig. 7. In the proximal and distal region of the extensor muscle of Cuniculina impigra, there were often tiny muscle fibres that looked as if a single fibre had been split into several pieces (see Fig. 6). Therefore, the total number of fibres in a cross section increased towards both ends in this species (Fig. 7A).

In both Cuniculina impigra and Carausius morosus, there was a small and nearly constant percentage of slow fibres (0–10 %) over more than 75 % of the muscle length (Fig. 7B,D). These fibres were found in the most dorsal region of the muscle (Fig. 6) and appeared to be shorter than the other fibres, because they had a larger insertion angle at the cuticle when consecutive cross sections were compared. Only in the distal quarter of the muscle was there a marked increase in the percentage of slow fibres. In Cuniculina impigra, the percentage of slow fibres approached 100 % at the most distal position, while in Carausius morosus this percentage reached a value of approximately 60–70 %. In Cuniculina impigra, the fibres of the posterior distal row of muscle fibres (region c in Fig. 2) were all slow fibres except for a few rare fast fibres at its proximal end. The percentage of fast fibres in these species changed correspondingly (Fig. 7B,D). The same pattern was found for the hindleg extensor muscles of Carausius morosus and Cuniculina impigra. In Locusta migratoria the situation was different. There was a gradual increase in the percentage of slow fibres from the most proximal region (0 % slow fibres) to the most distal region (approximately 40 % slow fibres) and a corresponding decrease in the percentage of fast fibres (Fig. 7F).

In summary, it appears that there are similarities in fibre composition between all three species. Similarities between Carausius morosus and Cuniculina impigra are clear in the dependence of percentage fibre type on muscle location (Fig. 7B,D). However, the absolute values for percentage fibre type were similar for Carausius morosus and Locusta impigra (Fig. 7D,F).

Comparison of the histochemical fibre-type characterisation (Fig. 7) with the innervation pattern (Table 1) suggests that the singly (FETi) innervated fibres are fast fibres and the dually (SETi, CI₁) innervated fibres are slow fibres. The triply (FETi, SETi, CI₁) innervated fibres could not be clearly attributed to either type and they were not likely all to be intermediate fibres as they occurred more frequently than the latter. The same conclusion was drawn for the locust M92 muscle (Müller et al. 1992). To investigate further, we recorded from triply innervated (FETi, SETi, CI₁) fibres (Fig. 8A) and, after physiological characterisation, filled them with a fluorescent dye (Lucifer Yellow). Negative current pulses of −10 nA amplitude were then injected at intervals of 1 s for 60–90 min. The dye was then given approximately 60 min for diffusion inside the muscle fibre. In unstained serial sections of the leg, the labelled fibres could be identified by their fluorescence in at least some consecutive sections close to the location of the recording electrode (Fig. 8B). In serial sections, mATPase activity was afterwards investigated with and without alkaline preincubation in order to characterise the fibre types (Fig. 8C,D). The example shown in Fig. 8 is a fibre identified as a fast fibre.

We treated a total of 10 triply innervated muscle fibres from the middle and distal regions of the locust extensor muscle in this way. Six were identified as fast fibres. Four fibres exhibited the staining properties of intermediate fibres. From these results, it is clear that triply innervated fibres are not a uniform group as they express not only fast but also intermediate properties.

The present investigation attempts to elucidate the correlation between neuromuscular structure and properties and the characteristics of joint control systems in three orthopteran insect species in order to detect specializations correlated with specific behavioural repertoires. We found that differences in the neuronal systems and in the muscle–joint systems are correlated with differences in the properties of the joint control systems in the three species investigated.

Previous measurements of tibial movements and muscle forces have shown that the femur–tibia control system in orthopteran insects shows marked low-pass filter properties, although the tuning may differ as a result of different behavioural specializations (see Introduction). The response of the FT control system in phasmids and locusts, measured as movement of the tibia, can be formally described as two parallel components (for the extensor and flexor tibiae muscles). The extensor component consists of a sequence incorporating a high-pass filter (situated in the fCO), a rectifier (formed by FETi and SETi) and a low-pass filter (formed by the muscles; Cruse and Storrer, 1977; Bässler, 1983a,b). We have investigated the motoneuronal output for sinusoidal stimuli. For both phasmids investigated, Carausius morosus and Cuniculina impigra, we found, in accordance with previous investigations, that SETi activity increased over the range of fCO stimulus frequencies tested (0.01–5 Hz; Fig. 1B, Bässler, 1983a). In these two species, the low-pass filter is not located at the motoneuronal level; thus, the low-pass filter characteristics obviously originate from the characteristics of the muscle–joint system.

In Locusta migratoria, in contrast, there was a continuous decrease in SETi activity in the stimulus frequency range from 0.01 to 0.3 Hz. This decrease approximately parallels the decrease in movement amplitude for the complete system over this frequency range and suggests that neuronal properties might contribute to the low-pass filter. SETi activity was found to increase with increasing stimulus frequency only above a stimulus frequency of 1 Hz. Thus, for faster stimuli, the decrease in movement amplitude cannot be attributed to neuronal properties and is obviously also derived from the properties of the muscle–joint system. In contrast to the situation in phasmids, an additional mechanism in locusts affects low-pass filter characteristics. We found that not all locust extensor muscle fibres are able to respond to high motoneuronal discharge rates, i.e. they do not show an EJP for every motoneuronal spike. However, this phenomenon only occurs at relatively high SETi frequencies. It is not known at present whether this phenomenon originates from the presynaptic or postsynaptic properties of neuromuscular transmission at the muscle fibres. A partial explanation for this phenomenon may be the conduction failures at branching points of the extensor nerve reported by Theophilidis (1984, 1988) for another orthopteran insect, Decticus albifrons.

In all three species, there is a similar innervation pattern of the extensor muscle (this study; Bässler and Storrer, 1980; Burns and Usherwood, 1979). The percentage of singly innervated fibres decreases from approximately 80–90 % at the proximal end of the extensor to zero at distal locations, while the percentage of dually and triply innervated fibres increases. A similar pattern is found from the histochemical properties of the muscle fibres: the numbers of slow fibres proportionally increase while the numbers of fast fibres decrease more distally along the extensor muscle. However, marked differences between the three species were detected. In Cuniculina impigra, which has a low upper corner frequency of the muscle–joint system, there is a larger percentage of slow fibres in the distal portion of the extensor muscle compared with that of Carausius morosus and Locusta migratoria. Indeed, the percentage of slow fibres increases to almost 100 % in Cuniculina impigra in this region largely due to the characteristics of the muscle fibres in the distal posterior row. These slow muscle fibres were found to be almost exclusively dually innervated (SETi, CI₁). It is conceivable that these reported differences influence the relatively low upper corner frequency in Cuniculina impigra compared with that of Carausius morosus: the existence of a high percentage of slow innervated fibres will result in an increase in tonic muscle force with increasing stimulus frequency, leading to a relative stiffening of the leg joint at high stimulus frequencies. The companion paper (Bässler and Stein, 1996) will verify this suggestion with an analysis of the specific role of these fibres in the generation of the joint control properties of Cuniculina impigra.

In invertebrate muscles, muscle fibre heterogeneity has been described frequently (e.g. Govind and Atwood, 1982; Rathmayer and Maier, 1987). Nevertheless, for insect muscles innervated by only three distinct motoneurones, three classic types of muscle fibre innervation have been described (cockroach, Atwood et al. 1969; locust, Burns and Usherwood, 1979; Hoyle, 1978; Müller et al. 1992; stick insect, Bässler and Storrer, 1980). There are singly innervated muscle fibres, i.e. fibres innervated by one fast motoneurone; dually innervated muscle fibres, i.e. fibres innervated by one slow motoneurone and a common inhibitor motoneurone; and triply innervated muscle fibres, i.e. fibres innervated by all three types of motoneurone. For one insect leg muscle, the coxa rotator muscle (M92; Snodgrass, 1929) of the locust, there is evidence for a close correlation between fibre innervation type and histochemical and mechanical properties. For this muscle, three different classes of muscle fibres were distinguished and named: type I fibres are innervated by the slow excitatory motor neurone and CI₁, stain for mATPase as slow fibres and show a slow contraction velocity; type IIa fibres are triply innervated, stain for mATPase as fast or slow fibres or both and have a fast contraction velocity; type IIb fibres are innervated by the fast excitatory motoneurone only, stain for mATPase as fast fibres and exhibit a fast contraction velocity. In their study, Müller et al. (1992) grouped the histochemical, electrophysiological and mechanical properties of the fibres investigated into these three different classes on the basis of the fibre location within the muscle; however, they did not identify the different parameters for individual muscle fibres. Our findings on the innervation pattern of extensor muscle fibres indicate that there may be no clear distinctions in the fibre innervation pattern, i.e. they support the previous results of Hoyle (1978; see his Table 1). In addition, we have shown that triply innervated muscle fibres can exhibit differing histochemical properties. In summary, our data indicate (see also Hoyle, 1978) that the usual classification of fibres may not be applicable to triply innervated muscles of insects. These results, therefore, question the concept of clearly delimited classic fibre types.

It is highly likely that these relationships between muscle structure and filter characteristics are not the only factors affecting the properties of the joint control system. However, it is clear from our investigation that, in addition to characteristics of the neuronal network, specific properties of the muscle–joint system are strongly correlated with the properties of joint control systems. In the following paper (Bässler and Stein, 1996) it will be shown that, particularly in the case of the extensor tibiae muscle in Cuniculina impigra, this contribution can result in a functional partitioning of the muscle.

We thank Professor Dr W. Rathmayer, Dr H. Wolf and Dr A. E. Sauer for helpful discussions on our data. We also thank Dr A. E. Sauer and Dr J. Schmidt, as well as two anonymous referees, for their comments on the manuscript. In addition, we thank Tove Heller, Ilse Winkler and Sybille Watt for technical assistance. This work was supported by DFG grants to U.B. (Ba578) and A.B. (Bu857).