Patterns of fusimotor activity during locomotion in the decerebrate cat deduced from recordings from hindlimb muscle spindles

Abstract

1. Recordings have been made from multiple single muscle spindle afferents from medial gastrocnemius (MG) and tibialis anterior (TA) muscles of one hindlimb in decerebrate cats, together with ankle rotation and EMG signals, during treadmill locomotion. Whilst the other three limbs walked freely, the experimental limb was denervated except for the nerves to MG and TA and secured so that it could rotate only at the ankle joint, without any external load. Each afferent was characterised by succinylcholine testing with regard to its intrafusal fibre contacts. Active movements were recorded and then replayed through a servo mechanism to reproduce the muscle length changes passively after using a barbiturate to suppress γ-motor firing.

2. The difference in secondary afferent firing obtained by subtracting the discharge during passive movements from that during active movements was taken to represent the profile of static fusimotor activity. This indicated an increase before the onset of movement followed by a strongly modulated discharge in parallel with muscle shortening during locomotion. The pattern of static firing matched the pattern of unloaded muscle shortening very closely in the case of TA and with some phase advance in the case of MG. The same effects were observed in primary afferents.

3. Primary afferents with bag₁ (b₁) contacts in addition showed higher firing frequencies during muscle lengthening in active than in passive movements. This indicated increased dynamic fusimotor firing during active locomotion. There was no evidence as to whether this fluctuated during the movement cycles.

4. When the mean active minus passive difference profile of firing in bag₂-chain (b₂c) type primary afferents was subtracted from that for b₁b₂c afferents, the difference was dominated by a peak centred on the moment of maximum lengthening velocity (v).

5. The component of the active minus passive difference firing due to b₁ fibre contacts could be modelled by f(t) =av (where a is a constant) during lengthening and by f(t) = 0.2av during shortening. The remainder of the difference signal matched the predictions of the static fusimotor signal derived from secondary afferents.

6. The findings are discussed in relation to the concept that the modulated static fusimotor pattern may represent a ‘temporal template’ of the expected movement, though the relationship of the results to locomotion in the intact animal will require further investigation. The analysis of the data indicates that the combined action of muscle length changes and static and dynamic fusimotor activity to determine primary afferent firing can be understood in terms of the interaction between the b₁ and b₂c impulse initiation sites.

The part played by muscle spindles in the control of natural movements must depend on how the static and dynamic gamma (γ_s and γ_d) fusimotor systems are activated. However, their patterns of activity have been difficult to elucidate, because technical problems severely limit the possibilities for directly recording from γ-motoneurones. The alternative approach of deducing γ-patterns from spindle afferent records has been used in a variety of reduced preparations (Eldred et al. 1953; Taylor & Davey, 1968; Perret & Busser, 1972; Perret & Berthoz, 1973) and with particular success in locomotor movements (Prochazka et al. 1976, 1979; Loeb & Duysens, 1979; Loeb et al. 1985) and jaw movements (Taylor & Cody, 1974; Cody et al. 1975) in intact, normally active animals. The interpretation of the spindle afferent recordings, however, is difficult because of the complexity of the interaction of muscle length changes and γ_s and γ_d discharge in determining the firing frequency of spindle primary and secondary sensory endings. One method which has been developed to deal with this is to attempt to match the spindle afferent patterns recorded in chronic experiments with recordings from similar spindle afferents in separate acute experiments, in which the recorded natural muscle length changes are imposed passively. Various patterns of γ_s and γ_d stimulation are then tried iteratively until the match is optimised (Hulliger & Prochazka, 1983). The limitations of this method are that only one afferent can be studied at a time and that the afferent used for simulation is different from the one recorded under natural conditions. Even so, the results that have been obtained do tend to indicate a generally increased activity in γ_s and γ_d systems during locomotion, with the possibility of some phasically modulated γ_s activity in parallel with α-motor firing (Prochazka, 1996).

The earlier work on locomotion in decorticate cats using spindle recordings (Perret & Busser, 1972; Perret & Berthoz, 1973) also strongly suggested that γ_s activity was phasically modulated. Later studies of the decerebrate cat were extended to include direct recordings from γ-axons (Murphy et al. 1984). These were found to belong to one or other of two functional groups. One group was described as ‘deeply modulated with each step (phasically modulated)‘, the other type ‘was not highly modulated with each step (tonically modulated)‘. It was concluded that the former were dynamic and the latter static γ-motoneurones. This seemed to conflict with the previous results quoted above for spindle recordings in decorticates in which strong modulation appeared to be a feature of static fusimotor activity. In the experiments of Murphy et al. (1984) identification of γ-efferents as static or dynamic depended on relating the observed resting firing frequencies of γ-axons recorded in-continuity in ventral root filaments to the effects of stimulating these on spindle primary responses to sinusoidal stretch. High resting frequencies were associated with dynamic effects and low resting frequencies with static effects. Support for these conclusions was later provided by results using small amplitude sinusoidal muscle stretch to look for γ_s and γ_d effects on spindle afferents during locomotion (Taylor et al. 1985) and further data based on the same approach has been presented for hindlimb extensors (Murphy & Hammond, 1990) and for intercostal muscles (Greer & Stein, 1989, 1990). However, some uncertainty has persisted because the technically difficult procedure on which the identification was based (in-continuity recording from and stimulation of single axons in ventral root filaments) could only be completed in a small number of cases and has not been the subject of independent confirmation. Also, the use of sinusoidal stretches for diagnosing static and dynamic effects on spindles has been found to be potentially unreliable in some situations (Rodgers et al. 1994). Added to this is the observation that in ankle flexor muscles in the decerebrate cat it appears to be the γ_s neurones which are modulated in locomotion (Murphy & Hammond, 1993). Studies in the jaw-closing muscles, from both spindle afferent and γ-motor recordings, have been consistently in favour of modulated patterns of firing in γ_s neurones (e.g. Taylor & Appenteng, 1981; Taylor et al. 1997). Although jaw muscles serve very different functions from hindlimb extensor muscles, it would be surprising if there were to be such a basic difference in the way in which the fusimotor system is used in the two cases.

It is evident that more data on spindle behaviour in locomotion are needed and that it would be a great advantage to be able to compare the behaviour of spindles during active movements with the behaviour of the same spindles under otherwise identical conditions, when the movements are exactly reproduced passively. Also, because primary and secondary afferents exhibit such a wide range of properties and patterns of connection to the three different intrafusal muscle fibre types (Taylor et al. 1992b), it is desirable to record from several simultaneously and to characterise their individual intrafusal contacts. We therefore report in this paper experiments on locomoting decerebrate cats in which a number of spindle afferents from flexor and extensor muscles are fully characterised in terms of their intrafusal fibre contacts and recorded from in both active and passively reproduced movements. Preliminary accounts of this work have appeared in abstract form (Taylor et al. 1999a,b).

METHODS

Experiments were performed on 15 adult female cats in the weight range 2.5-4.0 kg. Anaesthesia was induced with 3 % halothane in oxygen delivered into a box at 5 l min⁻¹ and continued with 1.5-2 % halothane in 50 % air and 50 % oxygen, delivered first through a mask and subsequently through a tracheal cannula. A double lumen cannula was inserted into the right external jugular vein, so that succinylcholine (SCh) or anaesthetic agents could be administered later in the experiment and kept separate. A loop of thread was placed around the right carotid artery, to allow it to be tied off later, whilst the left carotid artery was tied and cannulated for monitoring blood pressure and heart rate. Respiratory rate, end-tidal CO₂ and rectal temperature were monitored throughout and kept within normal limits. The maintenance of full surgical anaesthesia was confirmed by steady blood pressure (100/60 to 110/70 mmHg), pulse rate (70–80 beats min⁻¹) and end-tidal P_CO2 (40–45 mmHg), and muscle relaxation. Antibiotic was given as a single intramuscular dose (flucloxacillin, 200 mg). The bladder was emptied by suprapubic puncture or via the urethra by passing a catheter.

The left hindlimb and hip muscles were denervated except for the nerves to medial gastrocnemius (MG) and tibialis anterior (TA) muscles. The nerves to these two muscles were isolated and enclosed in plastic recording cuffs. The cuffs were 8 mm long, cut from nylon catheter (1.65 mm external diameter; No. 5 FG) and contained three silver wires spaced symmetrically 2 mm apart. The neurogram was recorded differentially between the central wire and the outer two joined so as to discriminate against muscle interference. Pairs of silver wires were inserted into triceps brachii, MG and TA on both sides for electromyogram (EMG) recording (bandpass 80–2.5 kHz) and displayed on monitor oscilloscopes. The six amplified signals were also full-wave rectified, low-pass filtered (-3 dB at 12 Hz) and sampled by computer at 50 Hz.

The animal was supported with its head in a stereotaxic apparatus and with a clamp on the L2 vertebra. The pelvis was supported by a rod passing through the iliac bones, just behind their crests, in order to avoid the noxious pressure which can result from the more usual pins (K. G. Pearson, personal communication). Local anaesthetic was infiltrated around the lumbar and pelvic supports. Body temperature was regulated by radiant heat lamps. Below the animal was mounted the belt of a servo-controlled treadmill, which could be raised to come in contact with the cat's paws at the optimum height. Whilst the forelimbs and right hindlimb rested on the treadmill belt, the left hindlimb was supported above the belt by two clamps, one on the femur, the other at the lower end of the tibia. The hip was extended to bring the femur 30 deg back from the vertical and the knee joint was usually set at a right angle. The ankle joint took up its natural angle dependent upon the balance of tone in the MG and TA muscles and the weight of the foot. This ranged between a right angle and an additional 20 deg of extension (plantarflexion). A light metal strip was bound firmly along the medial side of the foot and pivoted on an instrument grade potentiometer, mounted co-axial with the ankle joint, so that rotation about this joint could be recorded with very little loading. Also mounted on the spindle of the potentiometer, but normally free to rotate independently, was a cylindrical pulley. A braided nylon thread was wound around this and tensioned to a yoke attached to a servo-controlled electromagnetic puller (Type V409; Ling Dynamic Systems Ltd, Royston, Herts, UK). Thus, when the pulley was locked to the spindle, the motion of the servo system was converted linearly into rotation of the ankle joint. This method of applying muscle length changes had the distinct advantages of avoiding dissection of the tendons and of keeping the muscles in good condition and working at their normal lengths and directions of pull. The terms ‘ankle extension and flexion’ are used in the text instead of ‘plantarflexion and dorsiflexion’, respectively.

A left-sided laminectomy from L4 to L7 was performed and the dura opened to expose the left dorsal roots. A few rootlets from L7 and S1 were cut and divided to isolate between six and eight single afferent units sensitive to ankle flexion or extension. Because the lumbar spine moved significantly during locomotion, it was necessary to mount the array of eight silver hook electrodes on a plastic frame (30 mm × 15 mm) sutured to the spinous processes. This frame also carried a small black glass plate on which to dissect the filaments. The differential inputs to the filament amplifiers were connected in common to a silver reference electrode on which were placed some discarded inactive filaments. Filaments were dissected to provide large single unit spikes with very high signal-to-noise ratio. Because of this the recordings remained reliable despite the movements during locomotion and the EMG from back muscles. The single units were identified as spindle afferents from MG or TA by their response to muscle stretches and to muscle twitches elicited by stimulating through the nerve cuff electrodes. Any possible remaining doubt regarding the distinction between Golgi tendon organs and spindles was resolved by the use of succinylcholine (described below) and the observation of fusimotor-induced fluctuations in firing frequency in response to skin stroking and brainstem stimulation and their abolition by deep anaesthesia. Conduction velocities were measured from permanent records by backward spike-triggered averaging from the dorsal rootlet single unit spikes to the neurogram recorded via the appropriate nerve cuff (Kirkwood & Sears, 1980). Conduction times could be resolved to better than 0.1 ms. Conduction distances measured at the end of experiments by laying a thread along the exposed nerve were reliable to ±5 mm in approximately 150 mm for MG and 200 mm for TA. Conduction velocities were therefore thought to be reliable to about ±7 % for fibres conducting at 70 m s⁻¹. Level discriminators for each spike channel supplied TTL pulses to a computer interface (CED 1401 plus, Cambridge Electronic Design Ltd, Cambridge, UK) and a 66 MHz PC 486 computer. Using programs based on Spike 2 software (CED), the afferent activity was displayed as instantaneous frequency, together with the ankle angle sampled at 100 Hz and the smoothed, rectified EMG signals. Further data processing was performed using the Spike2 analysis package and Kaleidagraph software (Synergy Software Inc., Reading, PA, USA).

Once the array of afferents was set up, the right carotid artery was tied and pre-mammillary decerebration carried out, with complete removal of the forebrain. The anterior part of the tentorium was removed to allow a vertical stimulating electrode access to the midbrain locomotor region (MLR), taken to coincide with the cuneiform nucleus at co-ordinates of 3.5 mm lateral and 2.7 mm caudal to ear bar zero (Berman, 1968). The most effective depth was usually 6 mm below the midbrain surface (Mori et al. 1977). Stimuli were negative 0.5 ms pulses, 20 Hz at 0.5-1.5 V. Halothane anaesthesia was discontinued and locomotor movements commonly occurred spontaneously 0.5-1 h later and were then facilitated by the moving treadmill. Otherwise, repetitive stimulation in the MLR (on the right or on both sides), combined with treadmill movement, generated prolonged periods of vigorous and apparently well co-ordinated stepping with the three limbs on the belt and regular rhythmic movements at the left ankle joint. Data were gathered by the computer during these active movements and at the same time ankle movements were recorded on magnetic tape (TEAC RD 120). This tape recording could then be replayed through the servo mechanism described above, set up to reproduce the active movements exactly, whilst continuing to record from the same spindle afferents. This was done after administering a barbiturate to suppress fusimotor activity. Sodium pentobarbitone (3 successive doses of 12 mg in 1 ml i.v. at 1 min intervals) was effective for this purpose, but since this did not allow recovery for further locomotion, thiopentone (12.5 mg ml⁻¹) and brietal (5 mg ml⁻¹) were also tried. However, even with these short-acting agents, regular locomotion was never restored well enough to permit recording from a second batch of afferents. The use of halothane or enflurane to suppress γ-activity was precluded because of their property of enhancing spindle primary dynamic sensitivity (Andrew, 1966). Following the replay of ankle movements under deep anaesthesia, the spindle afferent responses to ramp and hold ankle rotations were studied before and for 5 min after giving SCh (200 μg kg⁻¹i.v.). The injection was preceded by a period of 30 s stimulation at 10 Hz of both MG and TA via their nerve cuffs, in order to enhance the muscle blood flow and so to ensure reliable delivery of the SCh to the muscles. The effects on the afferent discharge were interpreted according to the scheme detailed previously (Taylor et al. 1992a) in terms of the contacts of each afferent on the bag₁ (b₁) and bag₂ (b₂) intrafusal muscle fibres. This permitted each afferent to be classified as b₁c, b₁b₂c, b₂c or chain (c) type. The chief importance of this is so as to distinguish those primary afferents which have no contacts on b₁ fibres. It is also found that some afferents in the secondary conduction velocity range do have some b₁ fibre contacts (Banks et al. 1982). Testing with SCh is also important for distinguishing primary and secondary afferents in the conduction velocity range of 60 to 70 m s⁻¹. Any such afferents with a clear bag₁ influence are taken as primaries and those without as secondaries.

The principal form of data processing was the construction of cycle histograms using Spike 2 software. Periods of locomotion with cycles of regular shape and amplitude were chosen and markers placed on the movement record at the peaks of ankle extension. Averages were then computed of afferent spike channels, movement and EMG about these marks for a minimum of 10 step cycles. Cycle averages of afferent impulses were computed by the probability density method. Cycle histograms were computed of the numbers of spikes in each of 50 bins per cycle divided by the number of cycles. These numbers divided by the bin widths gave units of frequency (see for example Matthews & Stein, 1969). In order to make illustrations clearer, the procedure was adopted in most cases of truncating the averages to exactly one cycle length and replotting part of this.

Animals were killed at the end of the experiments with a barbiturate overdose. In five cases they were then perfused via the thoracic aorta with heparinised saline followed by 10 % formol saline. Dissection of the fixed brains showed that the transections passed from just in front of the superior colliculi to just in front of the mammillary bodies. Vibratome sections of the brainstem were cut transversely at 50 μm and stained with Cresyl Violet, to locate the stimulating electrode tracks. The most successful stimulating sites were found to have been 6 mm below the surface at the centre of the inferior colliculus, a region corresponding to the cuneiform nucleus.

RESULTS

The data reported are derived from recordings from 67 spindle afferents (40 from MG and 27 from TA). An example of the type of recording obtained is shown in Fig. 1. In this case MLR stimulation combined with treadmill movement led to continuous regular cyclical ankle movements of 20 deg peak-to-peak amplitude at 1.05 Hz, with alternating TA and MG EMG activity. The afferent recordings from three MG primary units and three TA secondary units revealed varying amplitudes of modulation up to 300 impulses s⁻¹. The most strongly modulated were two of the TA secondary units, whilst the third showed the least variation. In all cases there was a progressive increase in mean firing rate in the period following the start of MLR stimulation and before the treadmill started. Since this was seen to affect the b₂c type secondary afferents, as well as the primary afferents, it must indicate some build up of static fusimotor activity to both flexors and extensors before the onset of locomotion. The variability of afferent firing is also consistent with the presence of natural static fusimotor activity because the unfused contraction of chain fibres is known to produce such variability (Taylor et al. 1998). In Fig. 1 the initial increase in afferent firing frequency is much greater in TA than in MG, but the relative effects on the two muscles varied from one experiment to another.

Figure 1. Example of recordings made in a decerebrate cat during treadmill locomotion.

Three legs walked on the treadmill, while the left hindleg was fixed and allowed to rotate only at the ankle. The horizontal continuous line at the top indicates the period of MLR stimulation and the dashed line indicates when the treadmill was in motion. The two records below these are smoothed rectified EMG signals from left TA and MG. The next record below these is of ankle rotation. Successively below this are instantaneous frequency records (in impulses (imp) s⁻¹) from three MG and three TA spindle afferents. In each case the afferent type in terms of intrafusal fibre contacts is indicated together with afferent conduction velocity.

The behaviour of two of these afferents is shown in more detail in the upper five traces in Fig. 2, taken from the period in Fig. 1 showing established regular walking. Their firing patterns are quite consistent throughout the 10 cycles of movement shown, but they are clearly not simple reflections of the ankle angles. Though the receptors lie in antagonist muscles, their firing patterns do not vary in a simple inverse fashion. When γ-motor activity was suppressed with sodium pentobarbitone and exactly the same ankle movements imposed passively, discharge patterns were changed as seen in the lowest four traces. There is no sign of EMG activity in either muscle and the firing frequencies of the spindle afferents are much reduced. Their patterns are simpler and more obviously reciprocal. Nevertheless, whilst the MG unit fires only towards the peak of muscle lengthening, the TA unit fires throughout the cycle. The TA unit is clearly a secondary afferent by virtue of its conduction velocity and its b₂c type, as revealed by SCh testing. Its passive stretch response also accords with this. The MG unit has a conduction velocity in the borderline region between primary and secondary ranges, but should probably be regarded as primary, because of its b₁b₂c type. Its weak response to passive length changes is appropriate for a primary afferent in a slack muscle. The large excess of firing in the active relative to the passive movements indicates the occurrence of considerable fusimotor activity during locomotion. Further presentation of the results is best made systematically, taking in turn secondary, b₂c primary and b₁b₂c type primary afferents, first from TA then from MG.

Figure 2. Comparison of responses recorded during active walking movements and during the same ankle movements imposed passively.

Period of established walking taken from the data of Fig. 1. The upper five records taken during active movements show respectively left TA and MG smoothed rectified EMG, instantaneous frequency of a MG primary and a TA secondary afferent and ankle rotation. The four records below this resulted from imposing the same movements passively. Note the absence of EMG activity in the next two records and the greatly reduced spindle afferent firing in the lowest two records. Dashed and continuous vertical lines indicate points of maximum ankle flexion and extension, respectively.

Secondary afferents in TA

The behaviour of secondary afferents is particularly important for deducing fusimotor activity because they are almost exclusively influenced by the static γ-system. In Fig. 3 an experiment is illustrated in which two TA secondary afferents were recorded simultaneously during regular, large amplitude movements (Fig. 3A and B, filled circles). The data have been converted into cycle histograms of afferent firing and cycle averages of the other variables (see Methods). The two afferents showed strongly modulated discharge reaching 190–210 impulses s⁻¹. However, this was not a simple response to muscle lengthening, since a large part of the increase occurred during the muscle shortening phase (i.e. before the vertical datum lines drawn on these records). When the movements were repeated passively (Fig. 3A and B, open circles), with γ-firing suppressed by deep anaesthesia, there was virtually no afferent firing during shortening of TA and the response to lengthening reached only 60–70 impulses s⁻¹. The mean firing frequency record obtained during passive movements was subtracted point by point from the record obtained during active movements. The resulting mean difference records, shown in Fig. 3C and D rose to peaks of 125 and 170 impulses s⁻¹ very close to the moment of transition from shortening to lengthening (vertical lines), with a rapid fall thereafter. The difference records actually went negative by about 50 impulses s⁻¹ at the points of maximum lengthening, indicating less firing at this point in the active than in the passive movements. The excess firing during active shortening can only be explained by a rising profile of γ-static firing during this period. It seems likely that the rapid fall during the lengthening would have occurred as a result of a sudden fall in γ_s firing rate. The EMG records in Fig. 3C and D (dashed lines) show quite close parallels with the difference records, which implies that in this case the γ_s pattern may be similar to that of the α-motor firing. Even more striking, however, is the parallel between the difference records and the angular displacement as shown in Fig. 4A. Since these two secondary afferents behaved similarly, the ensemble mean of their difference plots has been computed and is shown as filled circles. The angular displacement record has been inverted, so that shortening of TA (ankle flexion) is upward to match the increase in afferent firing frequency difference (active minus passive). The displacement record is also scaled arbitrarily to have the same peak-to-peak amplitude as the mean frequency difference record. The waveforms are evidently very similar and this is emphasised by the plot of difference frequency against angle in Fig. 4B. By contrast, the plot of difference frequency and mean TA EMG in Fig. 4C shows a less exact parallel. The plot of difference frequency against EMG in Fig. 4D emphasises that the relationship is more irregular and with evidence of considerable phase shift. A total of nine TA secondary afferents were studied, of which eight showed essentially the same patterns as described above, but in three cases with some phase advance of difference frequency relative to shortening. In one case the firing during active movements was raised tonically, without significant modulation. It is known that, at constant length, γ_s stimulation causes biasing of muscle spindle secondary firing proportional to the stimulation frequency and that this bias sums quite linearly with frequency increases caused by stretch (Andersson et al. 1968; Ellaway et al. 1996). It therefore seems reasonable to propose that the difference frequency plots computed for TA secondary afferents in the present study generally reflect the underlying (suitably scaled) firing frequency profiles of γ_s neurones. Thus, the evidence from TA secondaries supports the view that γ_s discharge fluctuates, rising to a maximum towards the end of shortening and then falling rapidly.

Figure 3. Behaviour of two TA secondary afferents in active and passive movements.

In all four panels the ankle movement is shown as a continuous line with an arrowhead indicating direction of muscle lengthening. A and B show for each of the two units, cycle histograms of firing frequency in active (•) and in passive (○) movement. In C and D the mean difference records are shown, active minus passive (•) and the cycle means of smoothed rectified EMG in TA (dashed line). Vertical lines indicate transition from muscle shortening to lengthening. Records are means of 25 cycles from the data of Fig. 1. Both afferents were b₂c type. Their conduction velocities were 62.1 and 56.8 m s⁻¹.

Figure 4. Ensemble means of TA secondary difference records related to movement and EMG.

In A and C the ensemble mean of the difference records from Fig. 3C and D is plotted (•) superimposed on the ankle rotation record (A) and the mean smoothed rectified EMG record (C). Note in A that the direction of muscle lengthening is downward. The difference signal is also plotted at 20 ms intervals against ankle angular rotation in B and against mean EMG in D. In B and D, the straight lines represent the best fit linear regressions, the parameters of which are shown.

Primary afferents in TA lacking bag₁ fibre contacts

If the above interpretations of secondary behaviour in terms of γ_s activity are correct, then it should be possible to find similar behaviour in afferents from b₂c primary endings. In various muscles only 8–22 % of primaries are reported to be b₂c (Eldred et al. 1974; Banks et al. 1982; Taylor et al. 1992a), but in one experiment two of six TA primary afferents were shown by SCh testing to be of this type. Figure 5A shows the difference records for these two units with the movement record (continuous line) inverted and shifted on the y axis to overlie them. With the arbitrarily chosen scale the profiles of the difference signals are fitted reasonably well by the movement record, but there are more irregular discrepancies and phase advance than in the case of the secondary data in Fig. 4. It is possible that this is because the response of the primary afferents to stretch (even in the absence of bag₁ fibre contacts) is more velocity dependent than is the case for secondaries. Accordingly, we have assumed a response function of the form f(t) =adx/dt+bx+c (where f is afferent frequency and x is angular displacement) and have computed the optimum values for the constants a, b and c for the mean of these two units by multivariate regression between the displacement record and the passive afferent response (see Prochazka & Gorassini, 1998a). The linear transfer function resulting from this was then inverted and used to operate on the mean difference record (by means of Matlab Simulink software) and so to derive an estimate of the γ_s-induced intrafusal fibre length change. The assumption made was that the intrafusal length changes caused by static fusimotor action summed linearly with those caused by muscle stretch. The result of this process is shown in Fig. 5B, with the original displacement record (continuous line) and the predicted muscle length change (dashed line) scaled to the same peak-to-peak values. The close similarity of the plots strongly suggests that, when allowance is made for their different dynamic properties, the b₂c primary afferents, like the secondary afferents, also reveal that the underlying profile of γ_s-induced intrafusal length change resembles the unloaded movement. The actual profile of γ_s firing must be slightly phase advanced relative to this, because of the time taken for contraction of intrafusal muscle fibres (Bessou & Pagès, 1975; Hulliger, 1979). However, it is the timing of the afferent signal relative to the movement which is important for the purposes of control.

Figure 5. Active minus passive difference records for ensemble of two b₂c primary afferents in TA.

A and B show ensemble mean difference records (•) for two TA b₂c primary afferents (conduction velocities 78.9 and 96.2 m s⁻¹) from a period of spontaneous walking (31 cycles). The continuous lines in A and B are the mean ankle angle records with arrowheads (in A) to show direction of muscle lengthening. In B the dashed line shows the movement record predicted from the difference record by compensating for the transfer characteristic of the b₂c afferents (see text). Note the close agreement with the actual angle record, particularly during the fast components of active flexion and extension.

Firing patterns in TA primary afferents with bag₁ fibre contacts

In order to deduce γ_d firing patterns we must analyse records from those primary afferents, which have been shown by SCh testing to have endings on b₁ fibres. Figure 6 shows data from the three TA primary b₁b₂c type afferents recorded at the same time as the b₂c afferents described in Fig. 5. Figure 6A, B and C shows the averaged movement waveform and the mean discharge for the three units in active and passive movements. If we consider the afferent behaviour during the muscle lengthening phase, it is clear that the averaged firing frequency reached in active movements (filled circles) is very much greater than that in the passive state with fusimotor activity suppressed (open circles). The effects of fusimotor activity are emphasised in Fig. 6D, E and F, in which the filled circles represent the mean differences in firing frequency (active minus passive). The difference value reached during the lengthening phase ranged from 110 to 220 impulses s⁻¹, and this suggests the presence of considerable dynamic fusimotor firing during active locomotion, acting to sensitise the b₁ fibre to stretch. Of the 15 b₁b₂c type primary afferents studied, 12 showed evidence of this kind for varying degrees of dynamic activation, whilst three showed no such effect. Two TA secondary afferents were found with SCh testing to have significant b₁ fibre connections and one of these showed a very marked dynamic increase during locomotion. One b₁c type TA unit was moderately influenced.

Figure 6. Three examples of TA b₁b₂c primary afferent responses during spontaneous locomotion.

A–C show cycle histograms from 31 cycles of afferent firing in three b₁b₂c primary afferents recorded at the same time as the records of Fig. 5 during active (•) and passive (○) movements. Unit conduction velocities 130.6, 136.7 and 112.0 m s⁻¹, respectively. In all panels ankle rotation is shown as continuous line with arrowheads to indicate direction of muscle lengthening. D–F show the corresponding mean difference frequency records (•).

Estimating the contribution of dynamic fusimotor activity

If it can be assumed that the passive length transducing properties of the primary endings are the same irrespective of whether they do or do not terminate on a b₁ fibre (Gioux et al. 1991) then it should be possible to estimate the contribution of b₁ contraction to the difference signal. This has been attempted as shown in Fig. 7. In Fig. 7A the ensemble mean difference signals are shown for the three b₁b₂c type afferents from Fig. 6 (Fig. 7A, filled circles) and the two b₂c type afferents from Fig. 5 (Fig. 7A, open circles) recorded simultaneously. The records are conspicuously similar throughout the shortening period, but diverge widely during lengthening. The difference between these records (b₁b₂c – b₂c) is shown in Fig. 7B and this reveals a brief period of extra firing centred at the moment of maximum lengthening velocity of TA. This is precisely what would have been expected from the presence of a dynamic effect on the b₁b₂c endings. It is entirely consistent with the previous conclusion that enhanced γ_d activity accompanies the locomotor movements. Under conditions of rapid stretch in natural locomotion, the firing of Ia afferents has been found to be dominated by the velocity of stretch (Prochazka & Gorassini, 1998a). With this in mind, we have derived velocity of lengthening (adx/dt) for TA in Fig. 7 and find (with a suitable choice of the value of a) that its positive part predicts the most prominent part of the difference frequency attributed to b₁ fibre activation (dashed line in Fig. 7B). We found no advantage in using velocity raised to the power 0.6, as has previously been suggested (Prochazka & Gorassini, 1998a).

Figure 7. Prediction of the component of primary afferent firing due to the bag₁ fibre.

In A the ensemble mean active minus passive difference signals are shown for the three b₁b₂c primary afferents (•) from Fig. 6 and from the two b₂c primary afferents (○) from Fig. 5 recorded at the same time (31 cycles). Note the close similarity of the two difference records during the muscle shortening phase and their divergence with the onset of lengthening. In B the symbols (•) plot the difference of the b₁b₂c mean minus the b₂c mean. The ankle angular velocity (dashed line) has been obtained by differentiating the ankle angle record (continuous line in A).

It may be questioned whether it is valid simply to subtract the b₂c difference record from the b₁b₂c difference record, because it is known that b₁ generated firing does not sum simply with bag₂-chain generated firing (Banks et al. 1997). There are two separate impulse initiation sites connected to the main Ia axon by a variable number of myelinated segments. This means that the site which fires at the higher frequency may occlude the other or they may sum together to a variable extent, depending on the electrotonic interaction between them. However, in Fig. 7 there has probably been very little interaction or occlusion between the two sites because Fig. 7A indicates that the b₂c contribution was falling rapidly to zero at the beginning of the muscle lengthening phase. It was only in this phase that the b₁ fibre could have made a contribution. On those occasions when there is evidence for overlap of activity in the two impulse initiation sites, it may be possible to estimate the b₁ contribution from the velocity record and to subtract this (suitably scaled) from the difference signal, to leave the b₂c component. This has been tried in Fig. 8 for data from an ensemble of five b₁b₂c primary units in TA. In Fig. 8A the difference signal (filled circles) is superimposed on the ankle angle record (arrow indicates TA stretch during ankle extension). The stretch velocity has been derived (open circles) and its positive values scaled to match the difference frequency during the rapid stretch phase. This is supposed to represent the contribution to firing from the b₁ site, on the expectation that it occludes the b₂c site firing at this time. The negative part of the velocity has been scaled down to 0.2 of this, on the basis that during shortening the b₂c site dominates, but can be affected by some electrotonic spread from the b₁ site. The value of 0.2 is chosen with regard to the observations of Banks et al. (1997), which suggest that this is the average strength of the electrotonic effect. The complete scaled velocity signal was then subtracted from the ensemble difference signal to give an estimate of the b₂c component (Fig. 8B, filled circles). In this case, the ankle angle signal has been inverted, to show that the static component, so estimated from the b₁b₂c primary afferents, is remarkably well matched to the record of muscle shortening. In this it is consistent with the prediction of the γ_s firing derived from the secondary afferents. This analysis was tried with data from two ensembles of three units and one with five and all behaved similarly.

Figure 8. Prediction of the component of firing in b₁b₂c TA primary afferents due to static fusimotor firing.

In A the mean active minus passive difference record (•) is shown for an ensemble of five b₁b₂c primary afferents from TA recorded simultaneously during 35 cycles of locomotion induced by MLR stimulation. The continuous line indicates ankle rotation with muscle lengthening direction indicated by the arrow. The open circles (○) show the result of differentiating the angular displacement trace and scaling it to match the difference trace during extension and further scaling by a factor of 0.2 during the flexion phase (see text). In B the result (•) of subtracting the scaled velocity record from the b₁b₂c record from A is shown. The ankle angle record (continuous line) has been inverted to show muscle shortening upward. Afferent conduction velocities: 94.7, 67.3, 87.8, 90.5 and 103.2 m s⁻¹. In this figure the average has not been truncated to one cycle and repeated.

Behaviour of MG afferents

Of the 40 MG afferents recorded in these experiments 4 were secondary and 36 primary. The majority (27) were b₁b₂c primaries and it therefore seems best to concentrate on these in the first place. Their general behaviour indicated, as in the case of TA afferents, that there was an increase in discharge of both static and dynamic fusimotor systems, with additional phasic modulation of γ_s during the active muscle shortening. The basis for this conclusion is illustrated in Fig. 9, which shows the ensemble mean of the activity of four b₁b₂c primaries in panels A and C and the activity of a secondary afferent in panels B and D, all recorded in the same period of locomotion. The most prominent feature of the active discharge record for the ensemble of primaries (Fig. 9A, filled circles) is the brief period of afferent firing centred at the point of maximum velocity of muscle stretch (arrow 1), which reaches 80 impulses s⁻¹ There is no equivalent discharge in the passive movement record (open circles) and consequently this increase in afferent discharge must have been due to the presence of fusimotor activity in the active movements, which was not present in the passive movements. Being centred on the period of maximum stretch velocity, the extra response is typical of the dynamic responsiveness of primary afferents in the presence of dynamic fusimotor activity (Prochazka & Gorassini, 1998a). The absence of any such response in the passive movement situation is consistent with the evidence that the bag₁ intrafustretch in the absence of dynamic fusimotor activity (Dutia sal fibre contributes very little to the spindle response to & Price, 1990; Proske et al. 1991; Taylor et al. 1992a). Confirmation of the identification of this enhanced firing during stretch as due to b₁ fibre activity is the fact that the equivalent effect in the secondary afferent (Fig. 9B and D) is very small.

Figure 9. Patterns of firing of MG primary and secondary afferents during locomotion.

A and C refer to an ensemble of four MG b₁b₂c primary afferents (conduction velocities: 104.2, 96.2, 100.0 and 106.4 m s⁻¹) recorded during 35 cycles of locomotion during MLR stimulation. B and D refer to a single b₂c secondary afferent (56.8 m s⁻¹) recorded at the same time. In A and B the cycle histograms of afferent firing are shown for active (•) and for passive (○) movements. In C and D the difference signals (•) are shown (active minus passive). In all cases the mean ankle angle record is the continuous line with the arrowhead on the trace showing direction of muscle lengthening. In A, open arrow 1 shows the peak of afferent firing during rapid lengthening and open arrow 2 shows the peak of firing at the end of lengthening (see text for details).

The other prominent feature of the primary ensemble records is the peak of firing seen at the end of muscle lengthening and the beginning of shortening (arrow 2). In Fig. 9 the mean firing frequency during active locomotion shown in A (filled circles) rises towards the end of the muscle stretch phase to 35 impulses s⁻¹, but then increases suddenly to 65 impulses s⁻¹. There follows a rapid fall to 10 impulses s⁻¹ with the onset of muscle shortening. The mean response to the movement pattern repeated passively under anaesthesia (open circles) shows much less firing under these conditions. The difference record (Fig. 9C) shows the excess firing during active movements to reach 45 impulses s⁻¹. At first sight it might be thought that this could be explained as a consequence of increased stretch sensitivity due to dynamic fusimotor activity. However, the secondary afferent records in Fig. 9B and D reveal the importance of the γ_s contribution. In Fig. 9B the active firing pattern of the secondary at the end of lengthening and the beginning of rapid shortening is similar to that in the primary ensemble. The active minus passive difference record (Fig. 9D) reaches a peak of 95 impulses s⁻¹ after the onset of rapid shortening and does not fall below 60 impulses s⁻¹ throughout the shortening period. This indicates the presence of a strongly modulated γ_s discharge with approximately this profile, starting in the latter part of lengthening and continuing through the shortening phase. Returning to the primary afferent ensemble difference record (Fig. 9C), it is now clear that the latter part of the γ_s burst is much less effective in maintaining afferent firing in the primary afferents during rapid shortening than in the secondary. Indeed, without the secondary afferent record the presence of the static burst might have been overlooked. Although it happened that only four MG secondary afferents were recorded in these experiments, they all showed similar behaviour. Their active minus passive movement difference signals were consistent in indicating tonically raised levels of γ_s discharge with modulation phase advanced with respect to the MG muscle shortening movement.

Separation of the contribution of static and dynamic fusimotor activity to MG primary afferent firing

In presenting the data for TA afferents above it appeared that the contribution of the b₁ fibre to the primary afferent response could be predicted from stretch velocity using the expression f(t) =adx/dt, during lengthening and f(t) = 0.2adx/dt, during shortening. The same procedure has been tried with an ensemble of four MG b₁b₂c primaries in Fig. 10. In Fig. 10A the difference signal (filled circles) is seen to have two peaks, one just before the beginning of shortening and the other centred on the first and fastest part of lengthening. The open circles show the velocity signal scaled as suggested above. In Fig. 10B the scaled velocity has been subtracted from the ensemble difference record and is shown as filled squares. This has removed most of the response to stretch and reveals that the γ_s part of the response starts just before shortening and stops abruptly at the end of this phase. The effectiveness of the process of subtracting the estimated dynamic component is shown in Fig. 10B by the superimposed plot of the secondary difference signal from Fig. 9D (Fig. 10B, open squares). Although the frequency scales are different, the timing and general shape of the frequency changes during active muscle shortening are similar. The fact that the computation does not necessarily predict the mean firing frequency of the difference signal correctly is probably not of great significance, considering the differences between spindles. It is the prediction of the time course of the varying static fusimotor firing which is impressive.

Figure 10. Prediction of the component of firing in b₁b₂c MG primary afferents due to static fusimotor firing.

In A the mean active minus passive difference record (•) is shown for the ensemble of four b₁b₂c primary afferents from MG shown in Fig. 9. The continuous line indicates ankle rotation with muscle lengthening direction indicated by the arrowhead. This signal has been differentiated and scaled (○) to overlie the b₁b₂c record during the lengthening phase. During the shortening phase the differentiated record has been further scaled by a factor of 0.2 (see text). In B the result (▪) of subtracting the scaled velocity record from the b₁b₂c record from A (right hand frequency scale) is shown. Superimposed on this is the difference record (□) for the secondary afferent of Fig. 9D (left hand frequency scale).

Comparison of MG and TA

As shown above, both MG and TA spindle records show signs of similar patterns of static and dynamic fusimotor activity during locomotion. However, there were some detailed differences in the patterns of spindle activity. Comparison of Fig. 4A with Fig. 9D shows that, though the difference signal for secondary afferents (representing static fusimotor firing) in both cases matched the muscle shortening record, its onset was relatively phase advanced in the case of MG. In this case the advance amounted to 166 ms in cycles lasting 704 ms. Figure 10B shows that the static pattern estimated from MG primaries was also advanced by 158 ms in the same cycles. This feature was observed regularly for single units and for ensembles. The other point of difference between TA and MG spindle records is seen in b₁b₂c primary afferents. While in the case of TA primaries, there is usually one clear peak of firing per cycle centred on the maximum velocity of lengthening (Fig. 6), in MG there are two peaks per cycle (Fig. 9A). One of these occurs as expected at the point of maximum stretch velocity, but there is another just before the onset of shortening. This second peak is clearly seen from the secondary records (Fig. 9B and D) to be due to the underlying static fusimotor burst. Its excitatory effect on the primaries is interrupted by the rapid muscle shortening occurring in the MG, which we have seen exerts a negative velocity effect. The primaries in TA (Fig. 6) do not show two separate peaks of firing because the static fusimotor firing profile coincides precisely with the muscle shortening, which is slower than in MG. The TA afferents therefore continue to increase their frequency to the end of shortening at which point the sudden additional contribution due to muscle stretch produces the conspicuous single peak. Thus, although the underlying pattern of fusimotor activity is similar in TA and MG, the phase advance of the static burst in MG combined with the different relative speeds of muscle shortening yield distinctly different net afferent discharge patterns. The data suggest that there may be a higher level of tonic γ_s firing directed towards MG than towards TA, since the troughs of the difference records for TA afferents commonly fell to zero, whereas they were generally above zero in the case of MG.

DISCUSSION

The methods

In this report, recordings have been made from MG and TA muscle spindle afferents during locomotor movements in pre-collicular decerebrate cats, in order to deduce the underlying patterns of activity in static and dynamic γ-motoneurones. Although such an approach has been used before in animals which were lightly anaesthetised (Taylor & Appenteng, 1981), decerebrate or decorticate (Perret & Busser, 1972; Perret & Berthoz, 1973; Cabelguen, 1981; Taylor et al. 1985) and normal (Prochazka et al. 1977; Loeb et al. 1985), there has not been complete agreement on the interpretation of the results in terms of patterns of static and dynamic fusimotor activity. In the present experiments a number of new features have been introduced to facilitate the interpretation. Firstly, the afferents have been characterised by SCh testing, which reveals the presence or absence of contacts of each afferent on the intrafusal b₁ and b₂ fibres, which cannot otherwise be predicted with confidence from conduction velocity. Secondly, afferent discharge patterns in active movements are compared here with those obtained by imposing the same movements passively, in the presence of barbiturate anaesthetics to abolish α- and γ-motor output. Thirdly, the reliability of the data has also been enhanced by using cycle averages for single afferents and ensemble averages for several afferents of the same type. Recent new observations on the jaw muscles in lightly anaesthetised cats have shown the value of these methods in interpreting spindle recordings (Taylor et al. 1997). Fourthly, the availability of simultaneous recordings from several afferents in both MG and TA has clearly revealed the range of differences that exists amongst afferents. Most of these differences can now be explained in terms of recently expanded knowledge of the details of sensory and motor innervation of the individual intrafusal muscle fibres (Taylor et al. 1992a, 1998; Banks, 1994; Prochazka, 1996). Finally, relative to the other various studies on decerebrate preparations quoted above, the present experiments keep closer to the normal by allowing two muscles to exert their proper actions about a joint and by preserving some of their normal afferent innervation.

With regard to the preparation used, there are likely to be some differences in details of control of locomotion in the decerebrate cat from the normal cat. Nevertheless, locomotion in the decerebrate cat can be regular, vigorous and give every appearance of normal rhythm and posture of the limbs. The locomotor pattern also adapts readily to the speed of the treadmill belt. Thus such abnormalities as may exist in the decerebrate cat are not likely to have affected the main features of fusimotor control. One important aspect of the preparation is the extensive denervation and immobilisation of the hindleg under study. Technically, this has been essential to allow for isolation and characterisation of multiple single spindle afferents and the exact reproduction of the active movements of the ankle by the servo mechanism, but it is very likely to have distorted any reflex influences which may normally be exerted on α- and γ-motor activity. Another difference from normal conditions was the fact that the ankle movements had no external loading. As a consequence the speed of ankle extension in particular was greater than normal, since there was no ground contact of the foot during the stance phase. Because TA normally contracts with little external loading, the conditions for that muscle in the present work may not be very abnormal. The results obtained can be expected to provide a useful insight into the patterns of static and dynamic fusimotor activity generated centrally, though these may be modified reflexly in normal walking.

The results

The following basic observations were made. Firstly, afferent firing increased in the few seconds preceding the onset of locomotor movements and since this affected secondary afferents as much as primary afferents, it clearly indicates a build up of static fusimotor activity in preparation for the movements. This may be part of the normal phenomenon of ‘fusimotor set’, described by Prochazka et al. (1987). Secondly, during active muscle shortening, afferents fired at much higher frequencies than in the same phase of passively imposed movements. This effect was very prominent in secondary afferents and in b₂c primaries and so must indicate increased firing in those efferents supplying b₂ and chain fibres, namely γ_s motoneurones. The static fusimotor effect was so strong that secondary afferents often increased their firing frequencies during the active muscle shortening phase. The evidence for static increase during shortening was also present in b₁b₂c and b₂c primaries, but was less striking than in secondaries. This is probably because the dynamic sensitivity of primaries causes a negative velocity effect during muscle shortening. The third main observation was that b₁b₂c primaries, but not b₂c primaries or secondaries, often showed an increased firing during muscle lengthening in active movements over that seen in passive movements. This is most readily explained as due to the presence of increased γ_d firing during locomotion. In summary, locomotion in the decerebrate preparation appears to be accompanied by static fusimotor firing which is both raised and strongly phasically increased during muscle shortening. There is often in addition some increase in dynamic fusimotor firing, possibly another aspect of fusimotor set. With regard to the dynamic activity, we do not have any direct evidence as to whether or not any modulation of the γ_d firing occurs. In some previous work on decorticate or decerebrate cats evidence has been found for stronger tonic and α-linked static fusimotor activity in flexors than in extensors (Perret & Busser, 1972; Cabelguen, 1981). In the present observations the fluctuations in static firing have tended to be rather more marked for TA than for MG, and there was some evidence for a stronger tonic discharge in MG.

Interpretations

These qualitative conclusions appear to be dependable, as they are based on simple observations and deductions using well-established features of fusimotor effects on spindle firing. Confidence in the findings is also enhanced by their repeatability and by the use of ensembles of units of similar type. However, more quantitative analysis of the data has also been attempted and this deserves further examination. First, there is the process of subtraction of mean afferent discharge patterns in passive stretch conditions from that recorded during active movements. The difference signal is regarded as due to fusimotor firing which occurs during the active movement. Strictly, the validity of this approach requires that the length changes which occur in the spindles during active movements are exactly reproduced in the passive state by reproducing the ankle movements. It has been pointed out that this may not be true if the active muscle contraction causes appreciable stretch of the elastic components of the muscle and its tendon (Hoffer et al. 1989). We do not believe this to have been a significant problem in the present experiments, because the active movements were essentially unloaded so that the muscle forces generated would have been small. Also, any yielding of the tendon would have occurred during active muscle shortening, the time when we observe an excess of firing in the active movement relative to the passive movement. This excess, which is attributed to increased γ_s firing, would actually have been greater if allowance could have been made for the effects of tendon compliance.

It has generally been thought that the interaction of dynamic and static fusimotor firing with length to generate afferent activity is too complex to allow a quantitative assessment of γ_s and γ_d patterns from muscle length and spindle records. It is for this reason that an iterative simulation method was developed by Hulliger & Prochazka (1983). This involved studying the behaviour of single spindle afferents in acute experiments while movements, previously recorded in chronic experiments, were applied passively. Static and dynamic γ-axons were stimulated in various patterns until the best match of spindle firing was obtained. The limitations of this method as applied so far are that different spindles have had to be studied in the two parts of the experiment and, although they were selected to be as similar as possible regarding their passive stretch responses, they were not known to be the same with respect to their terminations on b₁ and b₂ intrafusal fibres, nor were they usually from the same muscle. The present work suggests that in some realistic situations the static and dynamic fusimotor contributions to afferent firing can indeed be deduced directly from the active minus passive difference signal, without resorting to a simulation procedure. The crucial step is first to examine the behaviour of secondary afferents, in which the absence of any b₁ effect means that the difference signal depends purely on static fusimotor activity. The contribution of β-innervation is probably small, because α-motoneurones fire at frequencies which are so much lower than the natural γ_s range. The evidence is that in secondary afferents there is a simple algebraic summation of static γ-firing (suitably scaled) with muscle length to determine afferent frequency (Andersson et al. 1968; Ellaway et al. 1996). Consequently, the difference signal should represent the mean profile of γ_s firing. Confidence in this conclusion is enhanced by the observation that the difference signal for b₂c primary afferents is similar to that for secondaries.

When the same principle was applied to b₁b₂c primary afferent firing, the difference signal during muscle shortening was found to be less than that for secondaries. This may be explained as a consequence of the b₁ fibre contributing a negative velocity component. In the shortening phase the b₁ fibre impulse initiation site will become inactive, but hyperpolarisation due to negative velocity will be expected to spread electrotonically to the b₂c impulse initiation site. The best evidence indicates that on average one site can affect the other electrotonically to the extent of about 20 % (Banks et al. 1997). The results shown in the present study confirm that when allowance is made for this velocity effect on primaries during muscle shortening, the difference signal does come to resemble that for secondaries.

During the muscle lengthening phase the difference signal falls to a low value in the case of secondary afferents, but in b₁b₂c primaries there is a rapid increase centred on the moment of maximum velocity of stretch. If it is true that the patterned γ_s firing falls to a low value at this time, then the firing at the b₂c impulse initiation site may be occluded by the firing of the b₁ site. The firing of the primary afferent will then be expected to reflect the characteristics of the stretched active b₁ fibre, namely to give a frequency proportional to the velocity (see Prochazka & Gorassini, 1998a), as observed here.

Relation to previous studies

The literature on this subject has become extensive and somewhat contentious and can only be briefly reviewed here. The earliest observations of hindlimb spindle activity in decerebrate locomotion (Severin et al. 1967) indicated the presence of fusimotor activity increasing during muscle contraction. Subsequently, in decorticate cats recordings from spindle afferents identified as primary or secondary strongly suggested that this fluctuating activity was predominantly static, but could say little regarding dynamic outflow (Perret & Busser, 1972). The direct γ-motor recordings of Murphy et al. (1984) have been reviewed in the Introduction above. These original studies concerned triceps surae only and their conclusion that γ_s firing was tonically raised and rather modestly modulated was surprising in relation to the earlier findings in decorticates. Such a pattern of activity was also different from that deduced for jaw movements in normal and in lightly anaesthetised cats (Taylor & Appenteng, 1981; Taylor et al. 1997). The possibility that the identification of static and dynamic axons might have been unreliable has been discussed and it should be emphasised that the resting firing frequency of γ-motoneurones in decerebrates is so variable that it cannot be a dependable basis for classification. However, some recordings of firing of triceps γ-motoneurones identified as static did actually show quite significant modulation in time with the locomotor movements (Murphy et al. 1984) and this modulation may have been thought to have been relatively modest because of the much more striking frequency fluctuations in those γ-motoneurones identified as dynamic. The present experiments, while indicating increased dynamic sensitivity of primary spindle afferents during muscle stretch do not exclude the possibility that dynamic fusimotor firing might fluctuate during the stepping cycle. The full resolution of the controversy regarding patterns of fusimotor activity will require simultaneous spindle and γ-motor recordings. There is perhaps less basis for disagreement between the different groups since the report (Murphy & Hammond, 1993) that in TA it is indeed the γ_s axons which show a well modulated discharge pattern. The finding in this paper of less tonic static activity in TA than in MG may also be consistent with the present data.

A considerable body of data on hindlimb spindle firing in natural locomotion has been accumulated (reviewed in Hulliger et al. 1989) and attempts have recently been made to derive general models to describe it (Prochazka & Gorassini, 1998a). Most interesting for the present case is the analysis of triceps surae spindle records (Prochazka & Gorassini, 1998b), from which it was concluded that some phasic, EMG-linked static fusimotor activity was needed to make simple passive models of primary and secondary afferents fit the natural records reasonably well. It was not tested whether the fit might have been improved further by taking the muscle shortening record as the profile for the γ_s activity and varying the phase relations. A feature of the present observations is the obvious importance of having records of spindle firing in response to passively reproduced movements in the absence of fusimotor activity. If such records had been available for the analyses of walking in the conscious cats, it would have been more evident how much fusimotor activity would have been needed to have produced the observed firing frequencies.

The records obtained by Loeb & Hoffer (1985), from a Ia afferent in vastus medialis during natural walking (their Fig. 1), did show two main bursts of firing per cycle. The larger one, reaching 175 impulses s⁻¹, lasted throughout the active period of the muscle during the stance phase, despite the presence of slow shortening. The other, briefly reaching 130 impulses s⁻¹, was centred on the moment of maximum velocity of stretch of the inactive muscle, during the swing phase. This is basically the same result as seen for MG spindles in the present work. Moreover, the new insights presented here enable us to interpret the effects which Loeb & Hoffer (1985) then observed of applying dilute lidocaine to the femoral nerve. The first effect was to increase the discharge during active stance and greatly to reduce response during swing. This can now be explained by supposing that in this particular case there was an initial reduction in γ_d impulses reaching the spindle. This would have had the effect of reducing the sensitivity to stretch. It would also have reduced the negative velocity effect conducted electrotonically from the bag₁ fibre ending to the bag₂-chain ending during the period of active shortening. In this period the afferent firing is dominated by the bag₂-chain ending, then under the influence of the burst of γ_s firing. As the lidocaine block spread to involve all the γ-axons the afferent firing during muscle shortening was also reduced, as expected from the view that this was normally supported by elevated γ-motor activity. Other examples are more difficult to interpret, but in all cases lidocaine reduced and simplified the afferent discharge.

Functional implications

Now that it is evident from the recordings in decerebrate cats how strongly the static fusimotor firing may be modulated, it must be questioned whether simple spindle models with constant parameters can be justified. This is especially the case because of the evidence that spindle primary afferent firing may be dominated by the b₁ or by the b₂-chain impulse initiation sites according to the conditions and that the two sites can interact electrotonically. To be useful in future work, models of spindle primary afferent behaviour will need to contain two separate subsystems representing the two impulse initiation sites, with appropriate provision for their interaction. The question as to whether there may be some degree of separate innervation of b₂ and chain fibres, so providing for two types of static fusimotor action, cannot yet be regarded as settled. Some recent evidence is in favour (Taylor et al. 1998), but other work is not (Celichowski et al. 1994). One relevant consideration is that when γ_s axons are stimulated with a stimulus train, frequency-modulated sinusoidally at 1 Hz, those axons supplying b₂ and chain fibres jointly cause strong afferent modulation with little phase shift (Ellaway et al. 1996). Those static axons supplying b₂ fibres alone produce very little modulation. It is possible, therefore, that to produce a tonic bias of afferent discharge, activity might be concentrated in those γ_s motoneurones that predominantly supply b₂ fibres. To produce the rapidly varying effect needed to oppose spindle silencing during muscle shortening, those γ_s motoneurones operating chain or b₂ and chain fibres jointly might be used. Further progress in this connection will require systematic direct recordings of the behaviour of γ_s motoneurones in locomotion.

The findings of this study potentially simplify the prediction of the firing of primary and secondary afferents. To a first approximation, secondary firing is a simple scaled version of the sum of muscle stretch and static fusimotor firing frequency. Primary firing during rapid muscle stretch in the presence of γ_d firing is dominated by the b₁ fibre, which causes afferent discharge to be proportional to stretch velocity. During shortening, the b₂-chain combination takes over, so that afferent frequency is determined as in the secondary afferent, but with a scaled negative component from the b₁ fibre. It may be that in previous attempts to obtain descriptions of the effects of γ_s and γ_d stimulation, too much emphasis has been placed on finding general descriptions for changes in stretch response parameters. In seeking a place for muscle spindles in movement control, investigators have commonly tried to visualise the spindle as a length transducer capable of description by a transfer function in which parameters are set by the fusimotor signals. This attitude is understandable in that it potentially allows one to draw on the techniques of engineering systems analysis, but may be inappropriate when trying to understand a system which has arisen through a long process of cumulative selective evolution, rather than by prospective design. The problem is that the models used may be too simple. Specifically, it is important to recognise the way in which the three intrafusal fibre types contribute to the afferent firing under different conditions. One rational approach is to study spindles under those conditions in which they normally operate (namely, natural active movements), to determine the patterns of fusimotor activity which actually occur and then to consider what functional (and hence survival) advantages might arise as a consequence. With this in mind, we can see from the present study that the initial increase in static fusimotor activity will have two potentially useful effects. First, it will increase the barrage of excitatory afferent activity reaching the α-motoneurones and so prepare them to respond to the output of the central pattern generator. Secondly, it will make the spindles able to signal shortening as well as lengthening, which most spindle afferents cannot do in the absence of static fusimotor bias. The observed modulation of static fusimotor activity superimposed on this bias acts to reduce the tendency of the spindle afferents to go silent during active muscle shortening and thereby extends their useful range of length transduction. We have previously suggested that the static fusimotor drive may function as a ‘temporal template of the intended movement’ in studies of jaw muscles (Taylor & Appenteng, 1981; Taylor et al. 1997) and hindlimb muscles (Taylor et al. 1999b), without being very precise as to the meaning of ‘intended’ in this context. It is perhaps better to use the phrase ‘expected movement’ in future and to define this as the movement which would take place through the action of the central pattern generator in the absence of external loading. The present finding that the estimate of the γ_s profile resembles the unloaded shortening better than the EMG signal is consistent with this idea. The implication of this is that any departure of the actual movement from this template would lead to a spindle output that could be considered as an error signal to be used for reflex correction or for plastic adaptation of the central pattern output. In this context, the phase advance of the static fusimotor output relative to shortening in MG may be an adaptation to help resist lengthening of extensors at the onset of stance. Increased γ_d output increases primary afferent response to the stretch phase of movement, but this cannot be useful as a correcting signal in the muscle of origin through the stretch reflex, because during the lengthening, which occurs as the muscle relaxes, the α-motoneurones are actively hyperpolarised (Shefchyk & Jordan, 1985). The burst of firing in primaries during muscle relaxation actually mirrors the stretch velocity and so peaks at the moment of maximum rate of stretch. The usefulness of this would probably be in providing a timing signal to keep the central pattern generator entrained to the natural movement speed of the limb (Whelan et al. 1995) and so to ensure smooth, energy-efficient locomotion.