Abnormal motor unit synchronization of antagonist muscles underlies pathological co-contraction in upper limb dystonia

Transcription

1 Brain (1998), 121, Abnormal motor unit synchronization of antagonist muscles underlies pathological co-contraction in upper limb dystonia S. F. Farmer, 1,2 G. L. Sheean, 1,2 M. J. Mayston, 3 J. C. Rothwell, 1 C. D. Marsden, 1 B. A. Conway, 4 D. M. Halliday, 5 J. R. Rosenberg 5 and J. A. Stephens 3 1 The National Hospital for Neurology and Neurosurgery, Correspondence to: Dr S. F. Farmer, Department of 2 Department of Neurology, St Mary s Hospital, Neurology, The National Hospital for Neurology and 3 Department of Physiology, University College, London, Neurosurgery, Queen Square, London WC1N 3BG, UK and 4 Bioengineering Unit, University of Strathclyde, and 5 Institute of Biomedical and Life Sciences, University of Glasgow, Glasgow, UK Summary The aim of this study was to examine the pathophysiological mechanisms underlying co-contraction in patients with dystonia (n 6) and writer s cramp (n 5). Multi-unit needle and surface EMGs were recorded from extensor carpi radialis (ECR) and flexor carpi radialis (FCR) muscles during motor tasks that elicited dystonia or writer s cramp. The EMGs from ECR and FCR were recorded simultaneously and analysed using cross-correlation analysis. Similar recordings were obtained from healthy age- and sex-matched control subjects (n 8). Despite co-contraction of the muscles, cross-correlograms from the healthy subjects did not reveal evidence of motor unit synchronization. Crosscorrelograms from the dystonic subjects revealed a central peak with a median duration of 37 ms, indicating broadpeak motor unit synchronization. Cross-correlograms from patients with writer s cramp were either flat or modulated by a Hz tremor. Frequency-domain analysis of ECR and FCR EMGs demonstrated significant coherence in the patients with dystonia and writer s cramp. These results indicate that co-contraction in dystonia is neurophysiologically distinct from voluntary co-contraction and is produced by abnormal synchronization of presynaptic inputs to antagonist motor neuron pools. ECR and FCR co-contraction in writer s cramp may be a compensatory process under voluntary control. Keywords: dystonia; writer s cramp; co-contraction; motor unit synchronization; tremor Abbreviations: ECR extensor carpi radialis; FCR flexor carpi radialis Introduction Dystonia may be defined as a disorder of muscle activation in which there is involuntary, excessive and inappropriate muscle activity resulting in muscle cramps, spasms and abnormal posture. Dystonias may be generalized, segmental or focal; some dystonias are task specific. The role of central motor pathways in producing dystonia is strongly supported by studies of symptomatic dystonias which may be associated with lesions in the putamen, caudate nucleus or thalamus (Marsden et al., 1985; Pettigrew and Jankovic, 1985; Lee and Marsden, 1994). However, while it is generally agreed that the dystonias result from disordered function of the central motor system, their pathophysiology remains obscure. Investigations of the dystonias, using conditioning type paradigms, have revealed subtle neurophysiological abnormalities. Hoffman-reflex tests reveal a reduction in Oxford University Press 1998 reciprocal inhibition in generalized dystonia and writer s cramp (Rothwell et al., 1983; Nakashima et al., 1989; Panizza et al., 1989). The R1 and R2 components and the R2 recovery cycle of the blink reflex are enhanced in blepharospasm and oromandibular dystonia (Berardelli et al., 1985). Longlatency stretch reflexes are altered in their muscle localization, duration and size, possibly reflecting altered excitability of cortical premotor areas (Rothwell, et al., 1983; Tatton et al., 1984; Nauman and Reiners, 1997). The N30 component of the somatosensory evoked potential may be increased in amplitude (Reilly et al., 1992). Magnetic brain stimulation has demonstrated abnormally enhanced excitability of the primary motor cortex (Ridding et al., 1995; Ikoma et al., 1996). The above studies rely on external stimuli in order to study

2 802 S. F. Farmer et al. the excitability of reflex, efferent and afferent pathways. There have been relatively few studies of the spatiotemporal pattern of ongoing EMG in different muscles during dystonic contraction. The results of such studies in focal dystonias, such as writer s cramp, show that there are abnormal cocontracting bursts of EMG activity in antagonist muscles (Hughes and Mcllellan, 1985; Cohen and Hallett, 1988). During rapid elbow flexion movements, patients with dystonia exhibit slow and variable movements with prolongation of the first agonist burst, along with abnormal co-contraction and spread of muscle activation to involve muscles not normally involved in the movement (van der Kamp et al., 1989). Thus clinical and EMG observation, evoked response and conditioning studies point to a lack of focusing of the central motor command. This may result from altered excitability at the cortical level due to a disruption of the basal ganglia circuit s ability to excite and inhibit corticomotor neurons selectively. In addition, there may be alterations in subcortical motor pathways and the spinal motor apparatus. The focusing of the motor command and the basic neurophysiological mechanisms that underlie muscle co-contraction may be studied by using cross-correlation analysis to detect the presence or absence of motor unit synchronization. The temporal structure of common presynaptic input to motor neurons may be further explored through use of coherence analysis. In this paper we examine the hypothesis that abnormalities of the spatiotemporal pattern of EMG activity underlie dystonic co-contraction of muscles and that these are produced by abnormal synchronization of presynaptic inputs to motor neurons. Cross-correlation analysis of EMG signals in mammals allows the spatiotemporal organization of presynaptic inputs to motor neurons to be defined. A peak of duration 6 ms centred near time zero in the cross-correlogram constructed between the discharges of pairs of single motor units indicates short-term motor unit synchronization, and thus the presence of common presynaptic drive from last order branched-axon presynaptic inputs (Sears and Stagg, 1976; Kirkwood and Sears, 1978). In animal studies, motor unit synchronization detects the presence of common presynaptic input to different muscles, activation of these common presynaptic inputs, will promote synergistic muscle activation (Kirkwood et al., 1982b). The presence of broad cross-correlogram peaks, of duration 40 ms (broad-peak synchronization) indicates common drive to motor neurons from presynaptic inputs the discharges of which are themselves synchronized. This has been termed presynaptic synchronization (Kirkwood et al., 1982a, 1984). Cross-correlation of single- and multi-unit EMG in man may reveal both short-term and broad-peak motor unit synchronization. Short-term synchronization is detected through cross correlation of EMGs produced during voluntary isometric contraction. Short-duration peaks are observed when the discharges of pairs of single motor units recorded from within the same muscle are cross-correlated (Datta and Stephens, 1990), or when the cross-correlation is performed between single or multi-unit EMGs of synergistic muscles, the actions of which are closely related (Bremner et al., 1991a, b). Broad-peak synchronization has been observed in individual or pairs of muscles similar to those in which shortterm synchronization is detected, but only when certain pathological processes such as stroke or spinal cord disease have disrupted function in the corticospinal tract (Datta et al., 1991; Farmer et al., 1993a). In general, motor unit synchronization is detected within muscles or between muscles that act as synergists. There are several reports of short-term synchronization of antagonist intrinsic hand muscles and antagonist leg muscles during co-contraction (Bremner et al., 1991b; Nielsen and Kagamihara, 1994). However, the effects are very weak and infrequently detected. Synchronization has never been reported between EMGs of antagonist muscle pairs acting about the wrist and it has been suggested that reciprocal inhibition between upper and lower limb antagonists during co-contraction may be revealed as a small decrease in joint motor neuron firing probability at time zero (Gibbs et al., 1994; Nielsen and Kagamihara, 1994). The detection of synchronization of the EMGs of muscles with antagonistic actions in dystonic subjects would be strong evidence for the presence of abnormal common presynaptic input to antagonist motor neuron pools. The presence of significant coherence between EMGs in man may reveal the temporal structure of common presynaptic inputs. In humans, EMGs from within and between upper limb muscles display coherence in the frequency ranges 1 12 Hz and Hz (Farmer et al., 1993b; Conway et al., 1995). Coherence has not been detected between antagonist muscle EMGs (S.F.F., personal observation). In this study we have obtained multi-unit needle and surface EMG from patients with primary and symptomatic upper limb dystonias. Similar recordings were obtained from patients with writer s cramp. Control recordings were made from non-dystonic subjects attempting to reproduce the types of muscle co-contraction observed in dystonia. In patients with hemidystonia control recordings have been obtained from the non-dystonic limb. The study describes the results of cross-correlation and coherence analysis of the EMG signals. These results have been previously published in abstract form (Farmer et al., 1995). Methods Subjects Six patients with upper limb dystonia were studied (four female, two male; median age 42 years, range years). The median duration of symptoms was 30 years, range 4 60 years. In all subjects the preferred hand either at the time of study or at the onset of the dystonia was the right. In three subjects the dystonia was symptomatic and confined to the limb contralateral to a known CNS lesion. In three patients the dystonia was primary. In one of these patients it

3 Mechanisms of co-contraction in dystonia 803 Table 1 Dystonia and writer s cramp patient details Patient (Sex) Age (years) Diagnosis Duration (years) Botox Drug treatment Lesion on scan Dystonia patients 1 (F) 17 PGTD 7 No BHZ/TBZ No 2 (M) 52 PSD 11 Yes CLON/BHZ* No 3 (M) 58 PSD 30 No None No 4 (F) 33 SLHD 11 Yes None Yes (lentiform infarct) 5 (F) 60 SLHD ~58 Yes BHZ No scan 6 (F) 31 SLHD ~29 Yes None Yes (porencephaly) Writer s cramp patients 7 (M) 55 DWC 14 No None No 8 (F) 30 DWC 2 Yes None No 9 (M) 31 SWC 2 Yes None No 10 (M) 17 SWC 1 No None No 11 (M) 32 SWC 8 No None No PGTD primary generalized torsion dystonia; PSD primary segmental dystonia; SLHD symptomatic left hemidystonia; DWC dystonic writer s cramp; SWC simple writer s cramp; BHZ benzhexol; TBZ tetrabenazine; CLON clonazepam. *No treatment for 2 years. was generalized. In the patients with primary dystonia there was a varying degree of involvement of both upper limbs. Five patients with writer s cramp were studied (one female, four male; median age 31 years, range years). All were right-handed. The median duration of symptoms was 2 years, range 1 14 years. Three of the these subjects had simple writer s cramp and two dystonic writer s cramp according to the definition of Marsden and Sheehy (1990). The three subjects with simple writer s cramp all displayed a mild tremor during writing (frequency ~11 Hz). All three patients with hemidystonia had received EMG guided treatment with botulinum toxin into the forearm muscles between 1 and 3 months prior to the study. One patient with primary dystonia had been treated with botulinum toxin into the forearm muscles 3 months prior to the study. One patient with simple writer s cramp had been treated with forearm muscle botulinum toxin 3 months prior to the study. At the time of the study two patients with dystonia were taking drug treatment for the movement disorder. The details of patients with dystonia and writer s cramp are summarized in Table 1. Eight healthy right-handed volunteers (four female, four male; median age 34 years, range years) acted as control subjects. None of these subjects had any history of neurological disease, and clinical examination did not reveal any evidence of tremor or dystonia. All subjects gave written, informed consent and the study was carried out with approval from the National Hospital for Neurology and Neurosurgery ethical committee. Experimental protocol Needle and surface EMG recordings were obtained from extensor and flexor carpi radialis (ECR and FCR) during cocontraction of the muscles. Recordings were made from both right and left upper limbs. One monopolar concentric needle electrode (core area mm 2 ) was inserted into the belly of each of the two muscles under study. Surface electrodes (diameter 20 mm) were placed over the muscles and adjusted in order to record EMG exclusively from either ECR or FCR. The subjects were asked to grip an object using all the fingers of the hand and to slightly extend the wrist such that EMG activity could be produced in both channels as steadily as possible at ~20 50% the maximum voluntary contraction. Some patients with dystonia experienced considerable difficulty in maintaining a steady force of contraction. The patients with writer s cramp were, in addition, asked to grasp a pen using a conventional grip between the thumb, index and middle fingers. In addition to these activation tasks, the healthy subjects were also asked to voluntarily produce simultaneous bursts of EMG in ECR and FCR as rapidly as possible. Typically, each recording epoch lasted 2 4 min. Of the dystonic patients, five underwent multi-unit needle EMG recordings, one underwent surface and needle EMG recordings. In all of the patients with writer s cramp, needle EMG recordings were obtained. Of the healthy control subjects, five underwent needle and surface EMG recordings, and three underwent surface EMG recordings only. The results of data analysis were not affected by the recording modality used. The EMG was amplified and filtered ( 3 db at 2 khz and 16 khz for needle recordings; 3 db at 20 Hz and 5 khz for surface recordings). The amplified filtered EMG record was stored on magnetic tape for off-line analysis. Figure 1 illustrates raw EMG from a healthy subject and a patient with primary upper limb dystonia. Data collection and analysis The needle and surface multi-unit EMGs were treated as stochastic point processes. TTL (transistor transistor logic) pulses were generated by passing the amplified EMG through a level detection circuit. The timing of each motor unit discharge was indicated by the pulse generated every time the leading edge of the motor unit action potential crossed

4 804 S. F. Farmer et al. Fig. 1 EMG activity recorded from extensor carpi radialis (ECR) and flexor carpi radialis (FCR) muscles during co-contraction. (A) EMGs from a healthy subject performing steady co-contraction. (B) EMGs from a healthy subject voluntarily producing cocontracting bursts of EMG in ECR and FCR as rapidly as possible. (C) Longer section of EMG, same recording as B, but with a contracted time-scale. (D) EMGs from a subject with primary segmental upper limb dystonia attempting to perform the same steady cocontraction task as the healthy subject. (E) Longer section of EMG same recording as D, but with a contracted time-scale. the threshold of the level detection circuit. It was ensured that only one trigger pulse was produced per action potential. The times of occurrence of the motor unit spikes were determined from the train of TTL pulses using a laboratory interface (CED 1401; Cambridge Electronic Design, Cambridge, UK) driven by a microcomputer. The times of occurrence of the motor units were used to perform time-domain analysis. Cross-correlation histograms were calculated using CED 1401 software. The analysis was performed using at least 2000 reference spikes. The EMG from the ECR was always used as the reference for crosscorrelation analysis. The cross-correlation histogram may be calculated for single as well as multi-unit data, the analysis may be applied to surface as well as needle multi-unit EMGs (Sears and Stagg, 1976; Harrison et al., 1991). In the case of multi-unit analysis, the resulting cross-correlation histogram represents the linear sum of all pair-wise comparisons between the discriminated motor units in the two EMG channels. In the case where the times of occurrence of the motor unit action potentials are independent, the histogram does not deviate significantly, at any time lag, from a mean level which is simply dependent on the total number of discriminated spikes in the two EMG channels and the histogram bin width. The cross-correlation histograms were calculated using a 1-ms bin width (the effective filtering for such analysis is thus DC to 500 Hz). The statistical significance of any peaks detected in the cross-correlation histogram was assessed according to the criteria of Sears and Stagg (1976). The magnitude of any peak discovered was expressed as a normalized index k (the number of counts in the bin at the histogram peak divided by the mean level of counts in the histogram; see Sears and Stagg, 1976). The position of the peak with respect to time zero was determined from the time of the maximum bin count. The duration of the peak was determined visually. Frequency-domain analysis was performed on the times of occurrence of the motor unit action potentials extracted from the EMG data. The same data files used for crosscorrelation analysis were used to calculate the coherence between the ECR and FCR EMGs. The times of occurrence of the motor unit spikes were transformed into the frequency domain through application of the finite Fourier transform to L disjoint samples of record length T (T 1024 ms, where the total record duration (R) is given by R LT (for details, see Rosenberg et al., 1989). The coherence may be defined as the correlation between the fourier transforms of two spike trains. This can be interpreted as the covariance between the two processes at frequency λ, normalized by the product of the variance of each process alone (Rosenberg et al., 1989). The coherence between two EMGs N1 and N2 is denoted by R 12 (λ) 2, and is written as R 12 (λ) 2 f 12 (λ) 2 /[f 11 (λ) f 22 (λ)], where λ is the frequency in Hertz, f 12 (λ) represents the cross spectrum and f 11 (λ) and f 22 (λ) represent the autospectra of the two component processes. The phase relationship, φ 12 (λ), between the two processes f 11 and f 22 is given by argument of the cross spectrum f 12 (λ), i.e. φ 12 (λ) arg{f 12 (λ)}.

5 Mechanisms of co-contraction in dystonia 805 Calculation of the coherence at any given frequency provides a bounded measure of the association between the two processes. The coherence at any given frequency necessarily takes on values between zero and one, with zero occurring in the case of linear independence of the two processes. The coherence spectra and corresponding confidence limits were calculated using programs written by D.M.H. and J.R.R. (for details, see Rosenberg et al., 1989; Farmer et al., 1993b). Results Figure 1A shows EMGs recorded from the ECR and FCR in a healthy subject who activated these muscles through constant, sustained isometric co-contraction required to hold a round object using all the fingers of the right hand. Figure 1 (B and C) shows EMGs from the ECR and FCR in a healthy subject attempting to voluntarily co-modulate the EMGs as rapidly as possible. Figure 1 (D and E) shows the ECR and FCR EMGs from a dystonic subject attempting to perform a constant co-contraction task similar to that attempted by the healthy subject in Fig. 1A. It can be seen that the raw EMG of the dystonic subject contains bursts of synchronous EMG activity lasting ms. A similar pattern of synchronous bursty EMG activity may be mimicked voluntarily by healthy subjects (see Fig. 1B and C). Figure 2 illustrates the results of cross-correlating the entire record length (2 min) of extensor and flexor EMGs, samples of which are shown in Fig. 1. During steady cocontraction, cross-correlation of ECR and FCR EMGs in the healthy subject revealed a flat histogram, thus there is no evidence of motor unit synchronization. In contrast, crosscorrelation analysis of ECR and FCR EMGs from the dystonic subject revealed a peak at time zero (duration 35 ms), indicating synchronization over this time scale between the antagonist muscle EMGs (see Fig. 2D). Figure 2B shows the results of cross-correlating ECR and FCR EMGs in a healthy subject attempting to mimic dystonia by producing rapid cocontracting EMG bursts. It can be seen that the synchronous bursts of EMG produced by the subject (see also Fig. 1B and C), produce a modulation pattern in the cross-correlogram characterized by a peak of duration 200 ms, flanked by further peaks of similar duration displaced from the central peak by an interval that approximately corresponds to the inter-burst interval of the EMGs (500 ms). This is further emphasized by Fig. 1C in which a 2-s segment of EMG has been plotted. The corresponding cross-correlogram for the data has been plotted for total lags of 100 and 1000 ms (Fig. 2B and C). These cross-correlograms clearly show the co-modulation pattern and may be contrasted with those plotted at similar lags for the dystonic subject in which the synchronization occurs over a 35-ms time span and is restricted to time zero (Fig. 2D and E). During steady co-contraction the cross-correlograms of all eight healthy subjects were flat. Correlograms, very similar to those in Fig. 2B and C, could be voluntarily produced by all of the healthy subjects by co-contracting ECR and FCR as rapidly as possible. No healthy subject produced a crosscorrelogram peak of duration 200 ms. The crosscorrelogram of this type of EMG pattern is, thus, quite distinct from that seen in the patients with dystonia. Figure 3 contains data from a patient with symptomatic left hemidystonia in whom there was an infarct in the right lentiform nucleus. Figure 3A shows a cross-correlogram constructed between the ECR and FCR EMGs recorded from the unaffected right arm during the steady activation task which involved voluntary muscle co-contraction. Figure 3B shows a cross-correlogram constructed between ECR and FCR EMGs recorded from the affected left arm during performance of the same muscle activation task which, in this case, evoked dystonic muscle co-contraction. The data from the unaffected arm does not show any features in the cross-correlogram, and is indistinguishable from that produced by healthy subjects during steady co-contraction (cf. Figs 2A and 3A). In contrast, the data from the affected side shows a peak at time zero, of duration 45 ms (Fig. 3B). For comparison, the ECR FCR cross-correlograms are illustrated for both right and left arms in a patient with primary bi-brachial dystonia, in whom attempted steady cocontraction of either arm evoked dystonia (Fig. 3C and D). In this case, cross-correlograms constructed for the right and left arms both contain peaks at time zero, with durations 39 and 40 ms, respectively. Figure 4 shows examples of the cross-correlation peaks between ECR and FCR EMGs observed in the six dystonic patients. The left-hand panels contain data from the three patients with primary dystonia (Fig. 4A C). The right-hand panels display data from the affected limb of three patients with symptomatic hemidystonia (Fig. 4D F). The time course of the central peaks is similar in both groups of patients (mean duration 38.3 and 34.8 ms, respectively). Five subjects with writer s cramp were studied. Although two subjects typically developed cramping pain in the forearms during performance of other tasks, suggesting a diagnosis of dystonic writer s cramp, the subjects were significantly impaired only during writing. During writing, all of these patients adopted an abnormal posture with a tendency for the thumb and index finger to flex, along with hyperpronation of the wrist in association with discomfort of the hand and forearm. In three subjects a rapid tremor developed during gripping of the pen. Cross-correlation of ECR and FCR EMGs did not reveal evidence of abnormal motor unit synchronization (see Fig. 5A). In this example there is a suggestion of a dip at time zero, indicating a small decrease (below chance level) in the probability of joint motor neuron firing. In the three subjects who developed tremor, an alternating tremor pattern was observed in the cross-correlogram. This effect is shown in two crosscorrelograms between ECR and FCR with a total lag of 100 and 1000 ms (see Fig. 5B and C). The results of cross-correlation analysis are summarized in Table 2. The size of the cross-correlogram peak is

6 806 S. F. Farmer et al. Fig. 2 Cross-correlograms constructed between the entire record lengths of ECR and FCR EMG (~2 min) using ECR as the reference. Raw samples of the data are illustrated in Fig. 1. The cross-correlogram constructed from the healthy subject during steady cocontraction (A) is flat. The cross-correlogram constructed from the healthy subject during voluntary bursts of co-contraction (B) contains a peak of duration 200 ms, centred at time zero. The same data as B, plotted with a contracted time scale (C), demonstrates that the cross-correlogram mirrors the EMG modulation shown in Fig. 1C. The cross-correlogram constructed from the dystonic patient during steady co-contraction (D) contains a peak of duration 35 ms, centred at time zero, indicating abnormal motor unit synchronization. (E) The same data as D plotted with a contracted time scale. Cross-correlogram bin width is 1 and 10 ms. The number of reference spikes used for A was 4519; for B and C it was 6554, and for D and E, expressed as the index k, and its duration in milliseconds. The displacement of the peak from time zero is expressed in milliseconds, with negative values (to the left) reflecting discharges in the FCR channel which precede those in the ECR channel. Although the centre of the peak was distributed either side of time zero (range 0 8 ms; see Table 2) these variations in lag are no greater than the 8 ms detected by Bremner et al. (1991a) in their study of short-term synchronization in hand and arm muscles. The delays are consistent with there being variable peripheral conduction times between the motor neurons and the EMG recording electrodes. Frequency-domain analysis of ECR and FCR EMGs in the healthy subjects and the unaffected limb of patients with symptomatic hemidystonias did not reveal significant coherence at any frequency during voluntary co-contraction. Frequency-domain analysis of ECR and FCR EMGs from the affected limbs of patients with idiopathic and symptomatic dystonias revealed significant coherence in the frequency ranges 1 12 and Hz; peak coherence in the higher range was generally close to 20 Hz. Examples of coherence are shown in Fig. 6. Figure 6A shows data from a healthy subject performing voluntary co-contraction; the spectra are essentially flat with no more crossings of the 95% confidence limit than would be expected by chance alone. Figure 6B shows ECR FCR coherence spectra from a subject with primary segmental dystonia which contains peaks in the ranges 1 3 and Hz, the main effect being at 21 Hz. Figure 6C illustrates ECR FCR coherence from the affected limb of a patient with symptomatic left hemidystonia, significant coherence is present at 1 3 Hz and Hz, the main effect being at 15 Hz; coherence spectra from the unaffected limb of this patient were flat. Figure 6D illustrates ECR FCR coherence from a patient with writer s cramp and tremor. In contrast to the dystonic patients, a highly discrete peak is present at 12 Hz, corresponding to the frequency of the tremor. The corresponding phase plots for Fig. 6B and C demonstrated that the coherence was present at zero phase lag; the phase for Fig. 6D indicated that the coupling occurred at 100 ms from time zero (see also Fig. 5). The results of coherence analysis are summarized in Table 3. It can be seen that, in general, patients with dystonia showed coherence in low and high frequency ranges. Within these broad ranges coherence was detectable; however, the magnitude of the effect is expressed as the maximum coherence detected within the range. In all cases the frequency-domain coupling

7 Mechanisms of co-contraction in dystonia 807 Fig. 3 Cross-correlograms constructed between ECR and FCR EMGs in a patient with symptomatic hemidystonia and a patient with primary segmental dystonia during co-contraction. (A) The cross-correlogram from the unaffected arm of the patient with hemidystonia is flat. (B) The cross-correlogram from the affected limb of the patient with hemidystonia contains a peak of duration 44 ms, centred about time zero. Cross-correlograms from right (C) and left (D) arms in the patient with primary segmental dystonia affecting both limbs contain peaks of duration 39 ms and 40 ms, respectively, slightly offset to the right of time zero. Cross-correlogram bin width is 1 ms. The numbers of reference spikes used for A, B, C and D were 5973, 3899, 2018 and 5007, respectively. occurred at zero phase. The subjects with writer s cramp alone did not display ECR FCR coherence; those with writer s cramp and tremor showed discrete coherence corresponding to the frequency of the tremor that was always out of phase. Discussion Differences between voluntary modulation, writer s cramp and dystonia This study describes, for the first time, abnormal motor unit synchronization of antagonist arm muscles, and indicates the presence of abnormal overflow of the motor command in patients with dystonia. The finding of a central peak in the cross-correlogram indicates that two motor neuron pools share part of their drive from a common presynaptic input. The data from the patients with dystonia and writer s cramp, and the healthy subjects producing co-contracting bursts, are clearly less stationary than those of healthy subjects performing steady co-contraction. However, as pointed out by Sears and Stagg (1976) in their analysis of respiratory rate modulation of EMG, the cross-correlation histogram actually expresses the non-stationarity as well as the synchronization which occurs on a millisecond time scale (see their Fig. 6 and Fig. 2 in the present paper). Crosscorrelation analysis thus reveals the fine structure of the cocontracting bursts of EMG in dystonic patients and shows that it differs from that produced by steady voluntary cocontraction and voluntary bursts of co-contraction in healthy subjects. There was no evidence of motor unit synchronization during steady isometric co-contraction of ECR and FCR of the unaffected limb of patients with hemidystonia, or in healthy subjects. Likewise, in two of the patients with writer s cramp, there was no evidence of synchronization during cocontraction, even when the posture required to grip a pen was clearly abnormal. The three patients in whom gripping a pen induced an alternating 11- or 12-Hz tremor exhibited an alternating cross-correlogram with no evidence of a central peak. When healthy subjects were asked to produce cocontracting bursts of EMG, the cross-correlogram simply reflected the time course of the voluntary co-contraction. The ability to detect different patterns of low frequency modulation of motor unit activity is well recognized both

8 808 S. F. Farmer et al. Fig. 4 Representative cross-correlograms constructed between ECR and FCR EMGs during co-contraction in the six patients with dystonia. Cross-correlograms from a subject with primary generalized torsion dystonia (A) and two subjects with primary segmental dystonia (B and C). Each contains a peak centred at or near to time zero. The peak durations are 45, 31 and 35 ms, respectively. (D F) Cross-correlograms from the affected limbs of three subjects with symptomatic hemidystonia. Each contains a peak centred at time zero. The peak durations are 45, 38 and 36 ms, respectively. Cross-correlogram bin width is 1 ms. The numbers of reference spikes used for A, B, C, D, E and F were 2960, 4258, 6554, 8572, 4161 and 2922, respectively. in animal and human experiments. For example, crosscorrelograms of cat thoracic motor units contain a pattern corresponding to the respiratory rate as well as short-term synchronization (Sears and Stagg, 1976; Kirkwood and Sears, 1978). In man, voluntary co-modulation of motor unit firing rate induces patterns corresponding to the frequency of the modulation (Bremner et al., 1991a; Farmer et al., 1993b). The important point, however, is that voluntary co-contraction is limited by the firing rate of the motor units and thus may, if necessary, be decoupled. In contrast, broad-peak synchronization reflects a process that occurs on a time scale of ms, i.e. within the interspike interval of motor units and thus reflects a process that cannot be easily uncoupled. The failure to find abnormal motor unit synchronization in patients with writer s cramp may indicate either that writer s cramp is not associated with abnormal common input to agonist and antagonist motor neurons, suggesting different pathophysiology compared with dystonia, or that the muscles sampled, i.e. ECR and FCR, are not those in which an abnormality of synchronization is present. It is a common clinical observation that, to compensate for the difficulties experienced during writing, these patients attempt to fix the wrist through excessive co-contraction of forearm muscles. The results of this study would suggest that this form of cocontraction is similar to that employed by healthy subjects, and thus may be a voluntary compensatory strategy rather than part of the dystonia. A study focusing on EMGs from different agonist

9 Mechanisms of co-contraction in dystonia 809 Fig. 5 Cross-correlograms constructed between ECR and FCR EMGs during muscle co-contraction induced through holding a pen in two patients with writer s cramp. (A) The cross-correlogram from a patient with writer s cramp, but no tremor, does not reveal evidence of abnormal motor unit synchronization. (B) The cross-correlogram from a patient with writer s cramp who developed tremor during the cocontraction task shows an alternating tremor pattern with period 100 ms. (C) The same data as B plotted on a longer time scale. Crosscorrelogram bin width is 1 and 10 ms. The numbers of reference spikes used for A and B were 3342 and 2863, respectively. Table 2 Characteristics of central cross-correlogram peaks Patient Hand (n*) Normalized size: median k (range) Duration (ms): median (range) Offset (ms): median (range) 1 R (2) R (8) 1.54 ( ) 36.5 ( ) 5.0 ( 4.0 to 8.0) L (7) 1.62 ( ) 34.0 ( ) 2.0 ( 7.0 to 8.0) 3 R (1) L (2) R (2) No peak No peak NA L (2) 1.29 ( ) 34.5 ( ) 0.5 ( 1.0 to 0.0) 5 R (4) 1.29 ( ) 38.0 ( ) 0.0 L (2) No peak No peak NA 6 R (4) No peak No peak NA L (3) 1.69 ( ) 36.0 ( ) 0.0 ( 1.0 to 5.0) R right; L left. NA not applicable. *n number of separate recording epochs for each hand. antagonist pairs of muscles acting on the fingers may yet detect evidence of abnormal motor unit synchronization in patients with writer s cramp. In addition, because of the task dependency of the cramps experienced by these patients, further work is also needed to study the task dependency of any motor unit synchronization detected. The neurophysiological interpretation of abnormal motor unit synchronization The duration of the central cross-correlogram peaks observed in this study is in the range termed broad-peak synchronization. The time-course of broad-peak synchronization is too great to be matched by the equations of Kirkwood and Sears (1978), thus indicating that it results from the increase in joint motor neuron firing probability that results from polysynaptic EPSPs (excitatory postsynaptic potentials). Kirkwood et al. (1982a) demonstrated that broadpeak synchronization of thoracic motor units in the cat was enhanced by reduced levels of anaesthesia and through increasing the central respiratory drive to the motor neurons by raising levels of the inspired carbon dioxide. Thus in the cat respiratory motor system broad-peak motor unit synchronization is produced by conditions which favour the development of increased excitability of spinal cord interneurons such that their discharges become synchronized (Kirkwood et al., 1982a, 1984). Furthermore, it has been demonstrated that broad-peak motor unit synchronization can be produced by spinal cord lesions and that, once established, it is shared by motor units from anatomically distant spinal segments. Kirkwood et al. (1984) postulated that spinal lesions release the activity of spinal cord interneurons through interruption of inhibitory control normally exerted by the descending bulbospinal tract. Findings analogous to those of Kirkwood et al. (1984) have been described in man following CNS lesions (Datta et al., 1991; Farmer et al., 1993a). In these studies broadpeak synchronization was only observed in recordings from pairs of motor units recorded within hand or leg muscles, or between muscles normally expected to show short-term synchronization, e.g. first and second dorsal interosseous muscles. In experiments on stroke patients in whom the lesion affected the motor cortex and or descending corticospinal tract pathways, abnormal synchronization of antagonist muscles of the type described in the present study has not been detected (S.F.F., personal observation). The demonstration of broad-peak synchronization of antagonist muscles leads us to conclude that, in dystonia,

10 810 S. F. Farmer et al. Fig. 6 Coherence spectra constructed between ECR and FCR EMGs during muscle co-contraction. (A) The coherence from a healthy subject does not show any significant spectral components. (B) The coherence from an affected limb of a patient with primary segmental dystonia shows peaks at 1 Hz and 20 Hz. (C) The coherence from the affected limb of a patient with symptomatic hemidystonia shows peaks at 1 Hz and 15 Hz. (D) The coherence from the affected limb of a patient with writer s cramp and tremor shows a highly discrete peak at 12 Hz. Bin width (A D) is1hz. antagonist motor neurons receive an abnormal common drive as a result of abnormal synchronization of presynaptic motor neuron inputs. In dystonia, common polysynaptic EPSPs are delivered to antagonist muscle motor neuron pools, resulting in an increase in joint motor neuron firing probability lasting for the duration of the central cross-correlogram peak (i.e ms). On the basis of the size of the central peaks detected (k-range ) it may be estimated that during dystonic co-contraction 25 50% of the total synaptic drive to the ECR and FCR motor neurons is shared (see Tuck, 1977). Thus, the magnitude of the obligatory common input and its time-course would be expected to have a profound effect on the ability of these patients to control antagonist muscles independently. Neurophysiological mechanisms of dystonia Spinal mechanisms The finding that agonist antagonist co-contraction in dystonia results from abnormal synchronization of presynaptic inputs to motor neurons accords with earlier studies in which a reduction of reciprocal inhibition between forearm antagonist muscles has been identified (Rothwell et al., 1983; Nakashima et al., 1989). These results were obtained from recordings of the inhibition of the Hoffman reflex in forearm flexor muscles produced by conditioning radial nerve stimuli. In upper limb muscles there was no difference between healthy control subjects and dystonic subjects in early reciprocal inhibition (conditioning test interval 0 1 ms) attributable to the Ia inhibitory interneuron. However, compared with normal subjects, dystonic subjects show a striking reduction in the later phase of reciprocal inhibition (conditioning-test intervals of 5 30 ms). This inhibition is thought to represent presynaptic inhibition of Ia afferent fibres. There are also abnormalities of an even later phase of reciprocal inhibition (conditioning-test interval ms), which is thought to be mediated by as yet unknown polysynaptic pathways. The studies of reciprocal inhibition have several implications for the interpretation of agonist antagonist synchronization. First, the present results indicate that the common drive to ECR and FCR is generated through synchronization of polysynaptic inputs to motor neurons. This finding strongly argues against any abnormality of the Ia inhibitory interneuron in dystonia, as this would involve

11 Mechanisms of co-contraction in dystonia 811 Table 3 Coherence between ECR and FCR multi-unit EMGs Patient Hand (n*) Coherence frequency (Hz) Peak size (median coherence) Low range High range Low range High range 1 R (2) (at 9 Hz) 0.10 (at 20 Hz) 2 R (8) (at 9 Hz) 0.10 (at 21 Hz) L (7) (at 8 Hz) 0.20 (at 20 Hz) 3 R (1) (at 1 Hz) 0.06 (at 21 Hz) L (2) (at 1 Hz) 0.19 (at 18 Hz) 4 R (2) None None NA NA L (2) None NA 0.05 (at 15 Hz) 5 R (4) (at 12 Hz) 0.05 (at 25 Hz) L (4) None None NA NA 6 R (4) None None NA NA L (3) (at 1 Hz) 0.06 (at 17 Hz) 7 R (4) None None NA NA 8 R (3) None None NA NA 9 R (4) 12 Hz 12 Hz 0.50 (at 12 Hz) NA 10 R (3) 11 Hz 11 Hz 0.40 (at 11 Hz) NA 11 R (6) 11 Hz 11 Hz 0.30 (at 11 Hz) NA R right; L left; ECR extensor carpi radialis; FCR flexor carpi radialis; *n number of separate recording epochs for each hand. Discrete peak present. changes in disynaptic pathways and the time course of any central cross-correlogram peak would be expected to be considerably narrower than that detected in this study. During normal voluntary co-contraction, evidence from primate and human experiments suggests that the cortical drive to the Ia inhibitory interneuron is reduced (Fetz et al., 1989; Nielsen and Kagamihara, 1992). This disfacilitation of the Ia inhibitory interneuron may promote muscle co-contraction; it does not, however, allow synchronization of forearm antagonist muscle EMGs to occur. Thus, in normal cocontraction the ECR and FCR motor neuron pools are still controlled independently. The study of Nakashima et al. (1989) found abnormal late-phase reciprocal inhibition in writer s cramp and hemiparesis as well as in dystonia. Whilst these authors suggest that the abnormality of reciprocal inhibition is greater for dystonia than for writer s cramp, there is nevertheless a clear reduction in reciprocal inhibition in subjects with writer s cramp. We did not examine reciprocal inhibition in our patients with writer s cramp. However, our patients did not differ clinically from those described by Nakashima et al. (1989) and, thus, might be expected to possess reduced levels of late reciprocal inhibition between ECR and FCR. The absence of broad-peak synchronization of ECR and FCR activity in our patients with writer s cramp suggests that abnormalities of reciprocal inhibition at the spinal level may not be sufficient to cause the abnormal motor unit synchronization. The minor differences in the amount of reduced inhibition described by Nakashima et al. (1989) would not seem to be sufficient to explain the presence of ECR FCR synchronization in dystonia and its absence in writer s cramp. The above considerations suggest that, in dystonia, abnormal reciprocal inhibition between antagonists may favour the development of abnormal agonist antagonist synchronization but synchronization is not an inevitable consequence of changes in reciprocal inhibition at the spinal level. In addition, loss of inhibitory control over nonreciprocal spinal excitatory pathways may also contribute to these findings. In the following section we consider the arguments in favour of dystonic co-contraction arising as the result of abnormal synchronization of corticomotor neurons. Cortical mechanisms Primate studies demonstrate that abnormal muscle cocontraction may be produced by both cortical and sub-cortical lesions. Inactivation of the globus pallidus with muscimol promotes muscle co-contraction (Mink and Thatch, 1991). The GABA (gamma-aminobutyric acid) antagonist bicuculline, when injected into the primary motor cortex, produces motor instability, characterized by excessive arm muscle co-contraction. This suggests that GABAergic inhibition at the cortical level is important for reciprocal organization of limb movements (Matsumara et al., 1991). In man, PET studies in dystonias suggest a variety of alterations in cortical activation. However, there are apparent and, as yet, unexplained differences between acquired and idiopathic dystonias at the level of primary motor cortex activation. Ceballos-Baumann et al. (1995a) reported underactivation of caudal supplementary motor area and underactivation of the primary motor cortex in idiopathic (primary) torsion dystonia, with over-activation of premotor cortex, dorsolateral prefrontal cortex, rostral supplementary motor area, anterior cingulate and insular. In contrast, the same researchers reported over-activation of primary motor cortex in acquired hemidystonia (Ceballos-Baumann et al., 1995b). These results suggest that the primary motor cortex behaves differently in idiopathic dystonia and symptomatic dystonia.

12 812 S. F. Farmer et al. The present study found the time course of agonist antagonist broad-peak synchronization was similar in primary and symptomatic dystonias, suggesting that the pathophysiological mechanisms underlying co-contraction are the same in the two conditions. The finding of under-activation of the primary motor cortex in PET experiments is difficult to reconcile with neurophysiological studies of dystonia in which enhanced motor cortex excitability has been demonstrated. The results of magnetic brain stimulation experiments suggest that there is abnormal excitability and reduced inhibition at the level of the primary motor cortex in focal task-specific dystonias (Ridding et al., 1995; Ikoma et al., 1996). The study of Ridding et al. (1995) recorded EMGs from the first dorsal interosseous muscle, therefore it is not known whether patients with focal task-specific dystonias show muscle or task-specific differences in cortical inhibition of ECR and FCR. Ikoma et al. (1996) demonstrated increased MEP (motor evoked potential) responses in the FCR in focal dystonia; they did not study the ECR. The findings of the present study provide indirect evidence for the motor cortex being the source of an abnormal common drive to the ECR and FCR in dystonia but not in writer s cramp. The contribution of premotor cortical areas, if any, to short-term motor unit synchronization in man is as yet unknown. There is ample evidence, however, that activity in the primary motor cortex and corticospinal tract is closely involved in the production of short-term synchronization (Farmer et al., 1990, 1991, 1993a; Datta et al., 1991). Activity in the corticospinal tract also contributes to common modulation of human motor unit activity in the Hz frequency range (Farmer et al., 1993b). The origin of common Hz modulation of motor neuron activity is highly likely to be Hz periodic synchronization of neurons of the primary sensorimotor cortex (Conway et al., 1995; Murthy and Fetz, 1996a, b; Baker et al., 1997). Thus corticospinal inputs to upper limb motor neurons may be responsible for short-term synchronization; in addition, rhythmic synchronization of the corticomotor neurons is detectable as coherence between motor unit discharges. Coherence between ECR and FCR EMGs reveals frequency components in the range 1 12 Hz and Hz at near-zero phase lag, indicating that in dystonia these frequencies are common to agonist and antagonist motor neurons. In contrast, the patients with writer s cramp and tremor exhibit discrete coherence at 11 or 12 Hz with a phase lag of 100 ms. Common frequency components in the range 1 3 Hz reflect common firing rate modulation of motor unit discharge, which in part may reflect voluntary changes in the central motor command (Farmer et al., 1993b). Rhythmic modulation of motor neuron discharge in the frequency range Hz may potentially arise as the result of activity in a variety of motor neuron presynaptic inputs. However, the similarity of these frequencies to those known to arise from periodic synchronization of neurons in the primary motor cortex at least raises the possibility that part of the abnormal common drive to antagonist muscles in dystonia arises as a result of abnormal periodic synchronization of corticomotor neurons. To summarize, changes in late reciprocal inhibition at the spinal level may, through enhanced activation of polysynaptic pathways, encourage presynaptic synchronization of agonist and antagonist presynaptic inputs, but abnormalities of late reciprocal inhibition are present between ECR and FCR muscles in a group of patients in whom we did not observe abnormal synchronization. Abnormal cortical inhibition is present in patients with focal task-specific dystonias. It is possible that, in dystonia, loss of inhibition at the cortical level results in enhanced levels of synchronization of cortical neurons, resulting in abnormal synchronization and coherence between ECR and FCR EMGs. One would have to postulate, however, that this process is muscle- and task-specific, which is why it was not detected in the ECR and FCR in writer s cramp patients. We next intend to study the task-specificity of motor unit synchronization and coherence in antagonist intrinsic hand muscles in writer s cramp. Conclusions The results presented in this paper indicate that the abnormal co-contractions in dystonia result from abnormal synchronization of presynaptic inputs to forearm antagonist muscles. These abnormalities may result from alterations in the behaviour of motor neuron inputs at various levels including the corticospinal tract. With suitable refinements, the approach outlined in this paper may provide a simple test to distinguish voluntary from dystonic muscle cocontraction and thus prove a useful diagnostic test for dystonia. Acknowledgements S. F. Farmer, D. M. Halliday, B. A. Conway and J. R. Rosenberg were supported by the Wellcome Trust. The authors also gratefully acknowledge the support of the Medical Research Council of Great Britain. References Baker SN, Olivier E, Lemon RN. Coherent oscillations in monkey motor cortex and hand muscle EMG show task-dependent modulation. J Physiol (Lond) 1997; 501: Berardelli A, Rothwell JC, Day BL, Marsden CD. Pathophysiology of blepharospasm and oromandibular dystonia. Brain 1985; 108: Bremner FD, Baker JR, Stephens JA. Correlation between the discharges of motor units recorded from the same and from different finger muscles in man. J Physiol (Lond) 1991a; 432: Bremner FD, Baker JR, Stephens JA. Variation in the degree of synchronization exhibited by motor units lying in different finger muscles in man. J Physiol (Lond) 1991b; 432: Ceballos-Baumann AO, Passingham RE, Warner T, Playford ED,