[page 76↓]

6.  Bilaterally synchronous oscillatory EMG-EMG activity evoked by the acoustic startle the healthy human

A major impetus to the study of motor control in both health and disease has been the realisation that the cortico-spinal system in the human tends to drive synchronisation of motor units over the 15-30 Hz band, to the extent that the coupling between EMG signals at this frequency may be taken as a surrogate marker of cortico-spinal activity (Farmer et al., 1993; Conway et al., 1995; Salenius et al., 1997; Baker et al., 1997; Mima et al., 1998b; Brown et al., 1998; Brown et al., 1999; Kilner et al., 1999; Halliday et al., 1998; Gross et al., 2000; Marsden et al., 2001; Forss et al., 2002; Grosse et al., 2003). In contrast, know­ledge of reticulospinal function has been limited by the inaccessibility of this system, both in terms of recording and stimulating, and by the absence of a known surrogate measure in EMG, with the exception of the drive to motor units at above 60 Hz during breathing (Kirkwood et al, 1982; Carr et al, 1994). Here, we use the acoustic startle response (ASR) to demonstrate that some forms of reticulospinal activity in the human are associated with a characteristic pattern of bilaterally synchronous oscillations with a frequency of 10-20 Hz between motor units.

As in other animals there are several lines of evidence that the ASR in humans is re­layed in the reticular formation of the lower brainstem and uses reticulospinal efferents (Hammond, 1973; Leitner et al, 1980; Davis et al, 1982; Lingenhöhl and Friauf, 1992; Lingenhöhl et al, 1992; Koch and Schnitzler, 1997; Koch, 1998; Yeomans and Frankland, 1996; Yeomans et al., 2002). Firstly, the startle reflex exists in anencephalic infants (Edin­ger and Fisher, 1913). Secondly, the caudo-rostral pattern of recruitment of cranial nerve innervated muscles suggests a generator in the caudal brainstem in the startle reflex (Brown et al., 1991a; Valldeoriola et al., 1997; Matsumoto et al., 1992). Thirdly, symptomatic ca­ses of exaggerated startle involve brainstem pathology and sometimes responses at a laten­cy only compatible with a brainstem relay (Brown et al., 1991b; Matsumoto et al., 1992). Finally, the startle reflex is diminished in the Steele-Richardson-Olszewski syndrome, in [page 77↓]which there are widespread pathological changes in the brainstem including degeneration of the pontine reticular formation with severe neuronal loss (Vidailhet et al., 1992).

We recorded EMG activity in proximal and distal upper extremity muscles during the physiological startle reflex in order to define any common drive to motoneurones from the reticulospinal system in the human. Given that the reticulospinal system projects bilaterally and preferentially innervates motoneurones of proximal muscles (Kuypers, 1981), we pre­dicted that a reticulospinal drive would be evident as significant EMG-EMG coherence between homologous proximal muscle pairs, with less coupling between hand muscles on the two sides of the body.

6.1. Methods

6.1.1. Subjects and recording procedure

Healthy subjects gave their informed consent to the study which was approved by the by the Joint Research Ethics Committee of the National Hospital for Neurology and Neurosur­gery and the Institute of Neurology. 28 subjects were recorded but only 15 had at least two auditory startle reflexes upon testing. Only the results in these subjects (13 female, 2 male; mean age: 29 yrs; range: 20-59 yrs) were therefore analysed. EMG was recorded from del­toid, biceps, finger flexor, and first dorsal interosseous (1DI) using surface electrodes (Ag-Ag, 9mm diameter) placed 3 cm apart on the muscle belly, with the exception of 1DI where the reference electrode was placed over the proximal metocarpo-phalangeal joint of the in­dex finger. EMG was also recorded from sternocleidomastoid and the onset of activity in this muscle was used to trigger the selection of post-stimulation startle blocks (see later). Facial muscles, which are usually activated during the startle response such as orbicularis oculi, masseter or mentalis (Brown et al., 1991) were not recorded as significant cross-talk between these muscles was to be expected. Coherence between right and left sternocleido­mastoid muscles was not evaluated for similar reasons.

[page 78↓]

Subjects sat on a chair and were asked to provide a gentle background contraction of deltoid, biceps, finger flexors and 1DI, bilaterally, while [1] unexpected acoustic stimuli (1 kHz tone of 50 ms duration at 98 dB) were delivered pseudorandomly and binaurally through headphones once every 5 minutes or so, or [2] they voluntarily mimicked a startle response at a rate of once every 10 s after an initial learning session. The voluntary startles served to show that any drive identified in the ASR was not related to background contraction, some non-specific feature of phasic movements or of the analytical approach utilised. However, given that the physiological cortico-spinal drive to motor units is attenuated during movement (Brown et al., 1998; Kilner et al., 1999), voluntary startles did not allow us to contrast the pattern of the corticospinal drive to muscles with that evident during reflex startles. We therefore [3] also asked the same subjects to tonically contract their deltoid, biceps and 1DI muscles bilaterally at a level under 50% maximal voluntary contraction for a period of about 60s.

EMG was band pass filtered between 53 and 1000 Hz and signals were amplified and digitised with 12-bit resolution by a CED 1401 analogue-to-digital converter. The sampling rate was 2 kHz. Signals were displayed and stored on a PC by a software package (CED Spike 2, version 4).

6.1.2. Analysis

Frequency analysis was performed upon the ASRs recorded in the 15 subjects. To this end the 0.92 s following the onset of each ASR (defined in sternocleidomastoid) was extracted from the recording. Only ASRs with a mean rectified EMG level in each block greater than thrice pre-stimulation background (in sternocleidomastoid, deltoid and biceps muscles) were analysed. In this way habituated startles were avoided. All the extracted ASRs were then concatenated to give a total record of 100 s, which was downsampled by averaging successive pairs of data points after digitally low pass filtering at 500Hz to avoid aliasing. Sham startle responses and submaximal voluntary contraction were processed in a [page 79↓]similar fashion, concatenating the same data lengths as used in the ASR from each subject. Table 1 shows how many ASRs individual subjects contributed to the whole sample.

Table 6.1: Number of startle blocks in each subject contributing to the entire ASR sample (block size: 0.92s)

number of startles

no. of individuals



















Coherence and cumulant density estimates were estimated from rectified EMG using methods outlined in chapter II. The discrete Fourier transform and parameters derived from it were estimated by dividing the concatenated records into a number of disjoint sections of equal duration (512 data points), and estimating spectra by averaging across these discrete sections (Halliday et al. , 1995). The frequency resolution of all spectra was 2 Hz.

6.1.3. Statistics

The power in each bin of autospectra was expressed as the relative percentage of the total power of each autospectrum to facilitate comparison between muscles and subjects. The variance of the coherence was normalised as outlined in chapter II using the Fisher transform. To test normalised power and transformed EMG-EMG coherences for statistical significance a repeated measures general linear model was performed using the three con­traction conditions and frequency band as the main effects. Separate models were per­formed for deltoid and 1DI and for deltoid- biceps and finger flexor-1DI, respectively. Where results were non-spherical, a Greenhouse-Geisser correction was used and when differences were significant a pair-wise Students-t-test was carried out.

[page 80↓]

6.2.  Results

Fig.6.1 compares a typical ASR with a voluntary sham startle in one of the subjects. There is evidence of phasic discharges repeating every 70-80 ms in deltoid during the ASR but not the voluntary movement from the same subject.

Fig. 6.1. EMG record of a typical reflex startle (A) and voluntary sham startle (B) in the same heal­thy subject. Note phasic dischar­ges repeating every 70-80 ms in deltoid during the ASR.

Fig. 6.2 demonstrates the averaged spectra of the percentage total EMG power for deltoid, biceps and 1DI, with the pooled data from homologous muscles on the two sides of [page 81↓]the body. The results from ASR, voluntary sham startles and tonic voluntary contractions are illustrated. Deltoid EMG has a peak centred around 12-14 Hz during the ASR. A similar, albeit less distinct feature, is seen in the ASR spectrum from biceps. This feature is absent during voluntary sham startles and tonic voluntary contraction.

Fig. 6.2. Averaged spectra of the percentage total EMG po­wer in ASR, voluntary sham startles and tonic voluntary con­traction in deltoid (A), biceps (B) and 1DI (C). Homologous muscles from the two sides of the body have been pooled in 15 subjects to give 390 data blocks. Note the peak centred around 14 Hz during reflex star­tles in deltoid (arrowed).

[page 82↓]The coherence spectra between right and left deltoid, biceps and 1DI during ASR, voluntary sham startles and tonic voluntary contractions are given in Fig 6.3A, C and E. The deltoid-deltoid and biceps-biceps EMG coherence during the ASR was above the 95%-significance level between 10 and 20 Hz and showed a discrete peak around 12-14 Hz. The peak was biggest in deltoid (Fig 6.3 A), where about 20 % of the activity at 12 Hz was synchronised between the two sides of the body. Conversely, in the voluntary sham startle and tonic voluntary contraction there was only minor coherence above 10 Hz. Note that 1DI-1DI coherence (Fig 6.3 E) was little different in the ASR, sham startle and tonic voluntary contraction. To check, whether volume conduction could account for the cohe­rence between bilateral muscles we levelled the surface recorded analogue EMG signals and then performed frequency analysis on the two resulting point processes. The result for deltoid, the muscle with the shortest distance between itself and its homologue, is shown in Fig 6.3A (inset). There remains a clear peak at around 14 Hz in the point process coherence pooled across subjects. Note that, in line with the lower information content of the point process, the coherence was lower than between the analogue signals (Fig 6.3A).

Fig 6.3B, D and F are the cumulant density estimates for the ASR. The cumulant density estimate for deltoid has a broad central peak with side-lobes every 70-80 ms (Fig 6.3B). Side-lobes are much less distinct in biceps (Fig 6.3B) and absent in 1DI (Fig 6.3B). They were also absent during voluntary sham startles and tonic voluntary contractions (not shown) in all of the muscles. The cumulant density function was estimated from blocked and hanning windowed data. Note that the cross-correlograms between homologous mus­cles were almost identical to the cumulant density estimates (Fig 6.3B, D and F), so that the periodicity evident in the cumulant density estimates for homologous deltoid and biceps muscle pairs were not epiphenomena of the way in which data were blocked.

[page 83↓]

Fig. 6.3. Coherence spectra between right and left deltoid (A), biceps (C) and 1DI (E) during ASR, voluntary sham startles and tonic voluntary contraction and cumulant density estimates for the same muscles during the ASR (B, D and F). Only the spectra from the ASR in deltoid and biceps have a discrete peak in coherence around 14 Hz (arrows). The coherence spectrum for levelled deltoid-deltoid EMG pooled over 15 subjects is shown in the inset to (A). Note the peak at around 14 Hz (arrow) in the point process coherence in the ASR but not sham startles or voluntary contraction. There is also considerable coherence <10 Hz. This was dimi­nished by detrending the data (not shown), although the latter did not affect the coherence in the 10-20 Hz band. The cumulant density estimate (black line) for deltoid (B) has a broad central peak with side-lobes eve­ry 70 ms during the ASR. Side-lobes are less distinct in biceps (D) and absent in 1DI (F). Cross-correlograms (grey lines in B, D and F) match the cumulant density estimates in deltoid and biceps (black lines in B, D, F). Peak-to-peak r-value in deltoid is 0.16; same scaling for biceps and 1DI. Dottes lines indicate 95% confi­dence limit of the cumlant density estimate. 100 s of data drawn from 15 subjects.

[page 84↓]However, pooled coherence spectra, such as those shown in Fig 6.2 and Fig 6.3B, D and F can be relatively dominated by a few individuals with very high EMG-EMG coherence and confidence levels established across the whole spectrum (Halliday et al., 1995) do not necessarily take this into account. To corroborate the consistency of our findings across the subject group we therefore randomly divided the sample of 100 s of EMG from each condition into five segments consisting of 20 s. The percentage total power in each of the five segments was entered into a general linear model with conditions (3 levels: ASR, sham startle, voluntary contraction) and frequency (2 levels: 10-20 Hz, 20-30 Hz) as main effects. There was a significant interaction between condition and frequency in deltoid (F[2;8] = 50.843, p <0.001) but not for 1DI. Post hoc analysis revealed a significant diffe­rence between the ASR and both the sham startles (p=0.01) and voluntary contraction (p=0.017) in the 10-20 Hz frequency band. Fig. 6.4 A shows the averaged normalised power across the five segments of 20 s.

Fig. 6.4. Five blocks of 20 s of data have been analysed, the power norma­lised, the coherence transformed and averaged for the 10-20 and 20-30 Hz bands in deltoid and 1DI. (A) Norma­lised power. (B) Transformed cohe­rences. Bars indicate standard error of the means. Asterixes indicate statisti­cally significant differences between conditions (p<0.05).

[page 85↓]

Similarly, transformed coherences from the 20s segments were entered into a Gene­ral Linear Model which also showed a significant main effect for frequency and condition only for deltoid (F[2;8] = 27.948, p =0.01). Here, differences were significant between the ASR and both the sham startles (p=0.006) and voluntary contractions (p<0.01) as well as between sham startles and voluntary contractions (p=0.03) in the 10-20 Hz band. Averaged transformed coherence from the five 20 s segments are illustrated in Fig. 6.4B. Note that power and coherence in the 10-20 Hz band were both higher in deltoid in the ASR than in sham startles or tonic voluntary contraction, so that changes in coherence were not due to modulations in non-linearly related frequency components (Florian et al., 1998).

In addition, the pattern of pooled EMG-EMG coherence detailed in Fig 6.3 was repre­sented individually among those subjects with more than 10 blocks of EMG during reflex startles (i.e. sufficient to estimate coherence). Figure 6.5 contrasts power and coherence spectra in reflex and voluntary startles in two such subjects.

Fig. 6.5. Power and cohe­rence spectra (insets) in re­flex and voluntary sham startles in two subjects. A. 15 s concatenated data. B. 13s concatenated data. Note the spectral peaks at about 14 Hz (arrowed) in proximal muscle pairs in the ASR.

[page 86↓]Finally it should be considered whether the coherence between homologous muscles at 10-20 Hz during the ASR could reflect the square wave nature of the acoustic stimulus. Could the stimulus have elicited a pulse of bilateral reflex EMG activity of similar dura­tion, which then appeared in coherence spectra as a peak in coupling? Ordinarily we would expect a 50 ms acoustic stimulus to lead to an EMG burst of 50 ms duration and contami­nate coherence spectra with a peak at 20 Hz, rather higher than seen here. Even if we were to consider a delay in the offset of the EMG pulse, to give a period of 70-80 ms appropriate for the frequencies detected in this study, there are several reasons for believing this to be an unlikely explanation. First, reflex EMG activity was elicited in several muscles but only coherence between deltoid and, to a lesser extent biceps, demonstrated a peak at 10-20 Hz. Second, both the raw EMG records (Fig 6.1A), the cumulant density estimates and the cross-correlograms of deltoid and biceps muscles (Fig 6.3) indicated that reflex EMG bursts at 70-80 ms were repetitive rather than single.

6.3. Discussion

It could be demonstrated that the normal acoustic startle reflex is associated with a ten­dency for motor units in homologous muscles on the two sides of the body to synchronous­ly discharge in the 10-20 Hz band. The synchronising influence was strongest for the more proximal muscles of the upper limb, and was not an artefact of the analysis technique or of volume conduction. Thus, voluntary sham startles analysed in an identical fashion failed to demonstrate this feature and bilateral coherence was evident between homologous proxi­mal muscles regardless of whether analogue or levelled EMG was analysed. Given the strong evidence that the startle reflex is mediated by the reticulospinal system this common oscillatory drive to the two sides of the body in the ASR can be ascribed to reticulospinal activity. This is supported by the finding of synchronised discharges in this frequency range between reticular neurones in the lower brainstem of dogs (Schulz et al., 1985). Interesting­ly, the tetanic fusion frequency of upper limb muscles is around 15 Hz, suggesting that the [page 87↓]reticulospinal drive at this frequency may have mechanically important effects (Nathan and Tavi, 1990).

Why should neurones in the primate motor cortex tend to synchronise at 20-30 Hz (Murthy and Fetz, 1996; Baker et al., 1997; Baker et al., 1999; Baker et al., 2001 ) whereas those giving rise to the reticulospinal drive following acoustic stimulation tend to synchro­nise at 10-20 Hz? On the one hand it seems possible that this reflects differences in the central network properties of the respective sites. On the other hand, differences in periphe­ral feedback delays are unlikely to account for the different synchronisation patterns. If anything the shorter peripheral conduction time to and from proximal muscles would fa­vour a higher frequency drive to proximal than distal muscles, and yet the reticulospinal drive to proximal muscles causes synchronisation at lower frequency than the corticospinal drive preferentially distributed to distal upper limb muscles.

The reticulospinal drive demonstrated here can be contrasted in character with the corti­cospinal drive to muscle. The latter does not lead to bilateral synchronisation, preferentially involves distal limb muscles and results in EMG-EMG coherence at generally higher frequencies (Farmer et al., 1990; Carr et al., 1994; Farmer et al., 1993; Marsden et al., 1999). In voluntary tonic contractions of weak to moderate intensity cortical drive leads to contralateral EMG-EMG coherence over the 15-30 Hz band (Kilner et al., 1999), whereas during strong contractions or movement cortical drive tends to synchronise motor units in the Piper range of 30-60 Hz (Brown, 2000). Recently, corticomuscular coherence has been reported in the frequency range of physiological tremor (8-12 Hz). However, this coupling is generally weak and it is unclear whether it is afferent or efferent in origin (Marsden et al, 2001; Raethjen et al., 2000). In any case, physiological postural, action and force tremors are not bilaterally synchronous (Marsden et al., 1969; Vallbo and Wessberg, 1993), al­though, exceptionally, pathological tremors may exhibit synchronisation across the muscles of the two sides of the body (Lauk et al., 1999; Raethjen et al., 2000; O’Sullivan et al., 2002).

In particular, the bilaterally coherent muscle activity documented here is reminiscent of that seen in the pathological condition of primary orthostatic tremor. The latter is associa­ [page 88↓] ted with strong synchronisation of muscle activity within and between limbs at around 13-18 Hz. Synchronisation is most evident upon standing, and characteristically abates during the swing phase of gait (Heilman, 1984; Britton et al., 1992; McManis and Sharbrough, 1993). The tremor frequency overlaps with the frequency of synchronisation seen in the ASR and it is interesting to note that reticulospinal neurones in the cat medullary and cau­dal pontine regions exhibit phasic modulation that is correlated to locomotor activity (Drew et al., 1986; Orlovsky, 1970; Perreault et al., 1993; Shimamura and Kogure, 1983; Shima­mura et al., 1982). A similar phenomenon may be seen in humans where the 8-20 Hz drive to tibialis anterior motor units is suppressed during the mid-swing phase of walking (Halli­day et al., 2003). Recently, Sharott et al (2002) showed that healthy subjects could develop a similar bilateral synchronisation at around 13-18 Hz in leg muscles when particularly un­steady. The possibility arises that the upper limb drive in the normal auditory startle reflex, pathological primary orthostatic tremor and the orthostatic tremor of posturally challenged healthy subjects involve a similar reticulospinal generator.

In summary, a specific pattern of EMG-EMG coherence that is associated with non-res­piratory reticulospinal activity in the human. The challenge now is to define when this drive is manifest in health and how it may be deranged in disease.

© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 3.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML generated: