We recorded local field potentials (LFPs) from cervical spinal cord (C5–C8) in monkeys performing a precision grip task and examined their coherence with electromyographic (EMG) activities (spinomuscular coherence) recorded from hand and arm muscles. Among 164 LFP-EMG pairs, significant coherence was found in 34 pairs (21%). We classified the coherence into two groups based on its frequency range, narrowband coherence, and broadband coherence. The narrowband coherence was restricted to discrete frequencies in the range of 14–55 Hz and was widespread throughout the superficial and deep gray matter. In contrast, the broadband coherence distributed between 10 and 95 Hz and was found only in the ventral half of the spinal cord. The narrowband coherence suggests that oscillations, which have been described in many motor control areas of the brain, could also pass though spinal interneurons to affect motor output and sensorimotor integration. On the other hand, the broadband coherence could be a unique feature of spinal motoneuron-muscle physiology.
Rhythmic oscillatory activity has commonly been observed in the mammalian CNS's and elucidating its significance in regulating behavior is an important question in neuroscience (MacKay 1997). This oscillatory activity may have a role in controlling volitional movement because it has been found in various key motor control areas including subthalamic nucleus (Marsden et al. 2001), motor thalamus (Marsden et al. 2000), supplementary motor area (Ohara et al. 2001), and cerebeller cortex (Aumann and Fetz 2004; Pellerin and Lamarre 1997). Oscillatory activity in the primary motor cortex (M1) seems to be involved in controlling finger movements. For example, M1 local field potentials (LFPs) exhibit oscillations around 15–30 Hz (beta-band) during precision grip performance in monkeys (Baker et al. 1997) as originally found in a magnetoencephalographic study with human subjects performing isometric finger contractions (Conway et al. 1995), which are dynamically coherent with electromyographic (EMG) activities (corticomuscular coherence) (Baker et al. 1997; Kilner et al. 2000; Riddle and Baker 2006). Furthermore, the phase-locked activity of M1 pyramidal tract neurons with oscillating LFP (Baker et al. 2003), as well as the resetting or modulation of cortical oscillation (Jackson et al. 2002) and corticomuscular coherence (Hansen and Nielsen 2004) by pyramidal tract stimulation, all suggest that the corticospinal pathway is involved in the generation of corticomuscular coherence.
Sensory input from peripheral afferent neurons is also known to affect the beta-band oscillations in the motor system. Digital nerve anesthesia (Fisher et al. 2002) or deafferentation (Kilner et al. 2004) reduces the coherence between different agonist muscle EMGs during precision grip. Coherence between sensorimotor electroencephalography and EMG of hand and arm muscles is also affected by arm cooling suggesting peripheral afferent and efferent pathways contribute to the generation of beta-band coherence (Riddle and Baker 2005). Finally, the direct recording of single-unit activity from the C8/T1 dorsal root ganglia (DRGs) in behaving monkeys showed that afferent neurons and muscles exhibit coherent oscillations during wrist movements (Baker et al. 2006). These data strongly implicate afferent input to the CNS in the genesis of coherent beta-band oscillations in motor system during voluntary movement.
The spinal cord is the first relay of both afferent and descending pathways, and it is well known that spinal interneurons receive highly convergent input from a number of motor centers as well as peripheral afferents (Baldissela et al. 1981). It is reasonable to postulate therefore that oscillatory activity in the descending and afferent pathways is relayed through spinal interneurons, and that spinal interneurons can affect and can be affected by this oscillatory activity. Therefore it is important to know if coherent oscillations exist in spinal interneurons during movement. However, no previous reports have examined this question in awake, behaving animals.
The aims of this study were to determine if the oscillations are present in areas of spinal cord containing spinal interneurons and, if so, are they coherent with similar activity in forelimb muscles during voluntary movement (spinomuscular coherence). To this end, we have recorded cervical spinal cord LFPs and forelimb muscle EMGs in monkeys performing a precision grip task. We provide the first evidence indicating the oscillations of spinal LFPs and the spinomuscular coherence during voluntary movement. Preliminary results have been presented previously (Takei and Seki 2007).
The data presented here were taken from two male Macaca fuscata monkeys (monkey A: 6.8 kg and monkey U: 8.5 kg). The experiments were performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and were approved by the Animal Research Committee at National Institute for Physiological Sciences.
A monkey was trained to sit in a chair with its right and left elbow restrained while it performed a precision grip task using a custom-made manipulandum (modified from Lemon et al. 1986) (Fig. 1, A and B). The monkey inserted his thumb and index finger through separate holes in a horizontal plate to access the levers of the manipulandum. Fingers 3–5 were inserted through another hole. The manipulandum comprised two spring-loaded levers (10 cm length), each with a potentiometer (Model 357, Vishay Spectrol) fitted to its pivot point to report lever position and a touch sensor (D5C-1DA0, Omron, 30–100 pF) to measure thumb and index finger contact with the levers. The binary state of both touch sensors was constantly monitored throughout the experiments to determine when the signal from potentiometer represented thumb and index finger positions (i.e., when thumb and index finger were both in contact with the levers). A trial would not begin until both touch sensors were activated and would abort immediately on loss of contact with either lever. Strain gauge was attached to each lever for measuring the force exerted by both fingers.
Thumb and index finger positions were continuously presented to the monkey via the positions of two rectangle cursors displayed on a screen in front of monkey (Fig. 1C). Two target boxes were also displayed and the monkey was required to keep each cursor inside its target box during a trial.
Each trial began with the presentation of two target boxes positioned so that the thumb and index finger were 3.0 cm apart (trial start, Fig. 1C). After 1.0–2.0 s, the “out” targets disappeared and two “center” targets appeared simultaneously, signaling monkey to flex its thumb and index finger (go) to bring the cursors into the center targets. The required displacement of both fingers was 0.7–1.3 cm (monkey A) and 0.3–0.8 cm (monkey U), which corresponded to a force of 0.9–1.1 and 0.4–0.6 N, respectively. The monkey was required to maintain the lever positions within the center targets for 1.0–2.0 s then release them back into the out targets once the center targets disappeared (go2). Successful completion of a trial was rewarded with a drop of applesauce. On average, ca 1,500 successful trials/recording session with the success rate of ca 80% were performed by both monkeys.
After behavioral training was complete, three separate surgeries were performed to implant a head restraint, a recording chamber over the cervical spinal cord, and EMG wire electrodes into multiple forelimb muscles. All surgeries were performed on different days using isoflurane (1.0–2.0% in 2:1 O2:N2O) or sevoflurane anesthesia (1.5–3.0% in 2:1 O2:N2O) and aseptic conditions. During the head-restraint and spinal-recording-chamber surgeries, the monkey was immobilized with intravenous pancuronium bromide (Mioblock, Organon, 0.05 mg/kg every hour) and artificially respired. Respiration rate was adjusted to keep EtCO2 within 20–25 mmHg. Two head restraints (plastic tubes) were fixed to the skull with titanium screws and dental acrylic. An oval-shaped spinal recording chamber (Perlmutter et al. 1998) was implanted over the lower cervical spinal cord. C3–T2 vertebrae were exposed bilaterally, and titanium screws were inserted into the lateral mass of each vertebra. After performing a unilateral laminectomy of C3–C7 vertebrae, the recording chamber was positioned over the laminectomy and cemented in place with dental acrylic. For recording EMGs, pairs of stainless steel wires (AS632, Cooner Wire) were implanted subcutaneously in 19 muscles of monkey A, including four hand muscles [first dorsal interosseous (FDI), adductor pollicis (ADP), abductor pollicis brevis (AbPB), abductor digiti minimi (AbDM)], 14 forearm muscles [flexor digitorum superficialis (FDS), radial and ulnar parts of flexor digitorum profundus (FDPr and FDPu), flexor carpi radialis (FCR), flexor carpi ulnaris (FCU), palmaris longus (PL), extensor digitorum-2,3 (ED23), extensor digitorum communis (EDC), extensor digitorum-4,5 (ED45), extensor carpi radialis longus and brevis (ECRl and ECRb), extensor carpi ulnaris (ECU), brachioradialis (BRD), pronator teres (PT)], and one upper arm muscle [biceps brachii (B)]. For monkey U, EMG wires were acutely inserted percutaneously in two muscles (FDI and AbDM) on each experimental day. The location of each EMG implant was confirmed by evoking joint- and muscle movement using low-intensity electrical stimulations applied through the wire electrodes during and after surgery.
During each recording session, the monkey's head was fixed to the chair with plastic rods and a glass-insulated tungsten microelectrode (impedance: 1–2 MΩ at 1 kHz) was inserted into cervical spinal cord. The position of the electrode was controlled by using a hydraulic microdrive (MO-951, Narishige Scientific Instrument) and a custom-made X–Y stage, both of which were mounted on the recording chamber. The signal from the microelectrode was amplified (×1,000) and filtered (0.1 Hz to 10 kHz) using a differential amplifier (Model 1800, A–M Systems) for LFPs. A silver-ball electrode was also inserted into spinal chamber as a reference electrode and placed on the scar tissue overlying the cord surface. In addition, output from the amplifier was high-pass filtered (0.3–10 kHz) for monitoring action potentials from single spinal neurons.
Due to difficulties in locating recording sites using conventional histological methods following chronic spinal recordings, we used the depth of the electrode tip below the point at which single-unit activity was first recorded in each penetration. After unit activity was first encountered, the electrode was advanced and LFPs were recorded where single- or multiunit activities were simultaneously recorded. This ensured that the electrode tip was located in spinal gray matter during the recording of LFPs. LFPs were digitized at 20 kHz. EMGs were amplified and filtered using a multichannel differential amplifier (SS-6110, Nihon Kohden, ×3,000–25,000, 5 Hz to 3 kHz) and digitized at 5 kHz. Signals from potentiometers, strain gauges, and touch sensors were digitized at 1 kHz. All subsequent analyses were performed off-line using MATLAB (Mathworks).
First, LFPs were band-pass filtered (4th-order Butterworth filter between 3 and 100 Hz) and down-sampled to 250-Hz sampling rate. EMGs were high-pass filtered at 30 Hz (4th-order Butterworth filter), rectified, and down-sampled to 250 Hz. Subsequent analyses were restricted to the continuous recordings from 0 to 1.024 s (256 points) after the termination of dynamic movement (end grip) of successful trials (Fig. 1D). The end grip was determined as the time when the rate of change of the summed forces exerted by index finger and thumb returned to <1 mN/s after monkey acquired the center targets.
For spectral analysis, the 256-point recordings were divided into 128-point non-overlapping segments for Fourier transform. This allowed investigation of spectral measurements with a frequency resolution of 1.95 Hz. Only LFP-EMG pairs that had >99 segments of data (each 128-point long) were included in the analysis, and LFPs from the intraspinal sites <150 μm apart were pooled to avoid redundancy. Then the one-sided power spectra for LFP and EMG signal were calculated. Denoting the Fourier transform of the ith section of LFPs and EMGs as F1,i(f) and F2,i(f), respectively, the power spectrum of each signal (j = 1,2) was calculated as (1) where L is the number of data segments available and * denotes the complex conjugate (Witham and Baker 2007). Using this normalization, P(f) has units of μV2. Then the coherence between LFP and EMG was calculated as (2) A significance threshold level S was calculated according to Rosenberg et al. (1989) as (3) where α is a significance level, which was set to 0.05 in this study.
In the preliminary analyses, we sometimes observed that significant coherence at very low frequencies (<5 Hz) and/or around 60 Hz. The former could be affected by signal co-variation produced by task performance, and the latter was caused by main noise. For these reasons, the frequency ranges of <5 Hz and 55–65 Hz were excluded from the further analyses. The signal filter of LFPs had a flat frequency response (>−1 dB) in the frequency range between 5 and 95 Hz. Significant coherence peaks were determined as the frequency range where three of five consecutive bins exceeded threshold S. Width of the coherence peak was measured with the first and the last frequency bin in each coherence peak. When a coherence spectrum showed more than one coherence peak, a coherence peak with widest frequency band was analyzed.
The coherence phase for the significant bins was calculated as (4) 95% confidence limits on the phase estimates, (θ ± Δθ), were determined according to the formulae given by Rosenberg et al. (1989) (5) In situations where two signals are correlated with a fixed time delay, the phase difference between two signals will be a linear function of frequency (Rosenberg et al. 1989). To determine if there was evidence of fixed-delay coupling, a regression line was fitted to the phase-frequency relationship. If the slope was significantly different from zero (P < 0.05, t-test on regression coefficient), the constant delay (τ ms) was estimated from the line's slope as (6) where A is the line's slope (rad/Hz). The negative slope (positive delay) indicates that LFP precedes EMG with the constant delay and vice versa.
Time-domain analysis was also applied to measure the correlation between LFP and EMG signals as a supplement to the frequency-domain analysis. This involved calculating the cumulant density function, which is defined as the inverse Fourier transform of the cross-spectrum (Halliday et al. 1995). To estimate the cumulant density functions, LFPs and EMGs were down-sampled to 5 kHz without rectification, and data in the time window (0–1.024 s after end grip) and frequency range (5–95 Hz except for 55–65 Hz) were adopted. The cumulant density functions were estimated in a range of time lag from −100 to 100 ms, and positive lag indicated LFP preceding EMG.
Electrical cross-talk among EMGs
Electrical cross-talk between EMGs might cause redundant coherences between LFP-EMG pairs because cross-talk signal from a muscle that is functionally coupled with the spinal cord may give rise to false positive coherence between an uncoupled LFP-EMG pair. To eliminate such redundant coherence, the extent of electrical cross-talk among EMGs recorded from different muscles was quantified using the method developed by Kilner et al. (2002). First, EMGs were down-sampled to 1 kHz and third-order differentiated without rectification. Cross-correlation functions between each EMG pair were then calculated as (7) where f1 and f2 are two differentiated EMGs, f̄1 and f̄2 are their mean values and σ1 and σ2 are their SDs. r was calculated with 25-ms lags and a maximum of a modulus of r, |r|max, was used as an index of the extent of cross-talk. In each experimental day, |r|max was calculated for every possible muscle pair using a 1-min epoch of EMGs. In this study, if |r|max was >0.25 in each muscle pair (cf. Kilner et al. 2002), one muscle (selected randomly) was added and the other was eliminated from the data pool.
LFPs were recorded from 22 sites in the cervical spinal cord (C5–C8, 15 from monkey A and 7 from monkey U). From 299 LFP-EMG pairs recorded, 135 pairs were eliminated because of significant EMG-EMG cross-talk. Consequently, further analysis was performed on 164 LFP-EMG pairs (150 and 14 pairs from monkeys A and U, respectively).
Figure 2, A and B, shows LFP power spectra recorded from two separate intraspinal sites (monkeys U and A, respectively). In these spectra, there were peaks at around 10 Hz (A) or 25 Hz (B). Among the LFP power spectra recorded from 22 intraspinal sites, 14 (64%) LFPs showed similar peaks between 10 and 30 Hz. Figure 2, C and D, shows the corresponding power spectra of rectified EMGs recorded from the FDI muscle. Broad peaks around 30 Hz were dominant in both spectra. From the LFP-EMG pairs (A–C and B–D) the coherence spectra between LFP and EMG were calculated (E and F, respectively). Although the LFP and EMG power spectra showed similar characteristics in both cases, the coherence spectra were clearly different. The first example (E) shows significant coherence only within a very restricted frequency range between 13.7 and 21.5 Hz (peak width = 7.8 Hz, see shaded area in E). In contrast, the second example (F) shows significant coherence over the entire frequency range examined (13.7–95.7 Hz, peak width = 82.0 Hz, see the shaded area in F). Figure 2, G and H, shows the coherence phase plotted versus frequency. Both examples show a linearly decreasing phase-frequency relationship. Such a negative linear relationship indicates that LFP preceded EMG by a constant time delay. A regression line fitted to this relationship had a slope significantly different from zero (P < 0.05); the slope corresponded to a delay of 30.2 ms (G) and 8.7 ms (H).
Among the 164 LFP-EMG pairs, 34 pairs (21%) showed at least one significant peak in their coherence spectrum (30/150 and 4/14 from monkeys A and U, respectively). In relation to the LFP oscillation, significant coherence peak was observed in 28 of 95 pairs (29.5%) with oscillated LFP and 6 of 69 pairs (8.7%) with nonoscillated LFP. Quantitative comparison among these coherence spectra were made by measuring the width of the significant peak in each spectrum (see Fig. 2, E and F). Figure 3A shows the distribution of the peak width in the 34 coherent LFP-EMG pairs plotted from the pooled data of two monkeys. The width of coherence peak was broadly distributed from ∼4 to 86 Hz. For subsequent analysis, LFP-EMG pairs were categorized into two groups: the pairs with a “narrow” coherence peak [≤25 Hz, n = 25 (21 and 4 from monkeys A and U, respectively)] and the pairs with a “broad” coherence peak [>25 Hz, n = 9 (9 and 0 from monkeys A and U, respectively)]. Absence of broad coherence peak in monkey U could be ascribed to the fewer number of LFP-EMG pairs recorded in this monkey. Figure 3, B and C, shows the percentage of LFP-EMG pairs with significant coherence peak at a given frequency in the narrow group (B) and the broad group (C). The (dashed lines) indicate the significance limit based on binomial test (P < 0.05). The bins that exceeded this significance limit were mainly distributed to 14–55 in the narrow group. On the other hand, these significant bins were distributed almost all frequency range in the broad group. From these results, we have concluded that the spinomuscular coherence occurred in two different modes, narrowband coherence (NB) and broadband coherence (BB).
Figure 3, D and E, shows a histogram of the estimated time delay between LFP and EMG in each group (D: NB, E: BB). In NB, 4 of 25 LFP-EMG pairs (16%) showed a linear phase-frequency relationship which was significantly different from zero, and the time delay was widely distributed from −66.2 to 38.9 ms [3.4 ± 47.9 (SD) ms]. More importantly, a majority of NB (21/25) did not show such a significant linear relationship, indicating that the coherence was not produced by a fixed-delay transmission between two signals. On the other hand, all BB pairs (9/9) showed a significant negative linear relationship, and the estimated delays fell in a range of 5.6–12.3 ms (8.5 ± 1.8 ms). This indicates that the LFP led the EMG with the constant delay in the BB pairs.
To further characterize these two modes of coherence, they were analyzed with respect to depth of the recording sites in the spinal cord (Fig. 3F). The proportion of LFP-EMG pairs with significant coherence (NB or BB) was not uniformly distributed with depth in the spinal cord, rather it was unimodal with a peak at 2,500 μm. Interestingly, the NB (▪) was present at every depth while the BB (▪) was restricted to depths between 2,000–3,000 μm. This difference indicates that the BB is restricted to the ventral portion of spinal cord whereas the NB can be found throughout dorsal and ventral spinal cord.
Time domain analysis provided additional features of NB and BB. Shown as insets in Fig. 4, A and B, are the cumulant density functions between the LFP and the EMG, the coherence of which is shown in Fig. 2, E and F, respectively. The cumulant density functions show significant negative peaks (▵) at 32.2 ms (A) and 7.4 ms (B), indicating the EMGs lag behind the LFPs by these delays, and these time lags are consistent with the time delays estimated from the phase-frequency plots (30.2 ms and 8.7 ms; see Fig. 2, G and H). A significant correlation (R2 = 0.98, P < 0.001) was found between the time delay estimated from phase-frequency plot (frequency-domain analysis) and the time-lag in the cumulant density function (time-domain analysis) in the LFP-EMG pairs, which showed a significant linear phase-frequency relationship (n = 13). This result confirms the validity of our estimation of the time delay between LFP and EMG signals in the frequency domain analysis. In addition, the time domain analysis showed the difference in duration of the cumulant peak between NB and BB. For example, the cumulant density function in Fig. 4A, inset, showed a broader peak than that of Fig. 4B, and the peak width at half-maximum (PWHM, ▴) was 24.0 ms (NB) and 7.8 ms (BB). The main part of Fig. 4, A and B, illustrates the histogram of PWHM of each NB and BB pair. BB showed a shorter PWHM (7.0–9.8 ms, median: 8.0 ms) than NB (6.2–41.4 ms, median: 10.0 ms; Wilcoxon rank sum test, P < 0.05), indicating that the cumulant of BB had a narrower peak than that of NB.
In this study we describe oscillation in LFPs recorded from spinal gray matter in monkeys performing a precision grip task as well as its coherence with the activity of forelimb muscles. We describe two different types of spinomuscular coherence, one operating in a restricted frequency range between 14 and 55 Hz (NB), another with a much broader range of coherent frequencies (BB). Furthermore, we show that the NB exists at dorsal and ventral levels of the cervical spinal gray matter, whereas the BB is spatially restricted to the ventral portion.
Oscillation of spinal LFP
As shown in Fig. 2, A and B, a majority (63%) of spinal LFPs exhibited a peak between 10 and 30 Hz in their power spectrum. These results suggest that spinal interneurons could exhibit oscillatory activity around beta-band frequencies, a phenomenon that has been widely reported in many other motor control areas (Aumann and Fetz 2004; Baker et al. 1997, 2003; Marsden et al. 2000, 2001; Murthy and Fetz 1992, 1996; Ohara et al. 2001; Pellerin and Lamarre 1997). The size and prevalence of spinal LFP oscillations described in the present study were not remarkable in comparison with those in the cerebral cortex (Baker et al. 1997, 2003) and DRG (Baker et al. 2006). Prut and Perlmutter (2003) found less synchronous activity occurred between pairs of spinal interneurons that receive common input and proposed a specific decorrelation mechanism that reduces synchronization among spinal neurons. The relatively small oscillatory peak found in the present study, therefore might be due to decorrelation of the oscillatory input from afferent and descending systems. In addition, the cortical pyramidal cells possess strongly polarized dendrites (Jones 1986). In contrast, some of spinal interneuron possess more radially arranged dendrites (Willis and Coggeshall 2004). These differences in the geometric configuration may explain weaker LFP signals in the spinal cord. Nevertheless, to our knowledge, this is the first report to show that spinal LFPs exhibit oscillation around beta-band frequency in awake, behaving animal. Because the significant LFP-EMG coherence has been more frequently observed in the oscillated LFPs (29.5% versus 8.7%), the oscillation of spinal LFP might be involved in spinomuscular coherence.
Coherence between spinal LFP and EMG
Twenty-one percent of LFP-EMG pairs showed significant coherence peaks, and this spinomuscular coherence was of two different types, NB and BB.
NB was distributed widely throughout the spinal gray matter (Fig. 3F), and two potential mechanisms for the genesis of NB would be suggested from this result, oscillatory output from and oscillatory input to the spinal interneurons.
Some of spinal segmental interneurons are known to have direct and indirect projections to motoneurons, and their locations are distributed widely throughout the dorsoventral extent of the spinal gray matter (Kitazawa et al. 1993; Perlmutter et al. 1998). If these premotor spinal interneurons show the oscillatory activity as shown in Fig. 2, A and B, their oscillatory outputs may drive oscillations in target motoneurons and muscles during voluntary movement. Similar mechanism has already been proposed in the M1 and corticomuscular coherence. In M1, the oscillatory activity of corticospinal cells, which have direct and indirect connections to motoneurons, is proposed to be a source of corticomuscular coherence during voluntary movement (Baker et al. 1997, 2003; Hansen and Nielsen 2004; Jackson et al. 2002; Mima et al. 2000). On the other hand, abundant afferent fibers are known to terminate within the spinal gray matter, and the location of these terminals also distributed widely throughout the dorsoventral extent of the spinal gray matter (Brown 1981; Ishizuka et al. 1979; Willis and Coggeshall 2004). Therefore if the oscillatory muscle activity (Piper 1912) induces oscillatory afferent activity, it could also drive oscillation in spinal neurons and LFPs. In fact, group Ia primary afferents, which are known to have highly divergent intraspinal projections (Brown 1981; Ishizuka et al. 1979; Willis and Coggeshall 2004), show oscillatory firing and coherence with EMGs during voluntary wrist movement (Baker et al. 2006). We applied the coherence phase analysis to dissociate these two mechanisms, and only a minority of LFP-EMG pairs showed significant slopes in their phase-frequency relationship. This result may suggest that each NB coherence has been originated from both mechanisms.
As shown in Fig. 3B, we found the NB in two specific frequency ranges, 14–27 and 39–55 Hz, and each frequency range is corresponding to the beta (15–30 Hz) and gamma (30–80 Hz) band. Most previous studies have reported the beta-band corticomuscular coherence occurred during weak and steady force output (Baker et al. 1997; Kilner et al. 2000; Riddle and Baker 2006), and they may have a function to improve the performance of steady-state motor output (Baker et al. 1999). Thus the spinomuscular coherence at beta-band during the hold period of precision grip task could represent the fact that spinal neurons may be involved in this cortico-muscular coupling. On the other hand, the gamma-band corticomuscular coherence has been rarely reported during weak and steady force output (cf. Brown et al. 1998; Omlor et al. 2007; Schoffelen et al. 2005). Significance of the gamma-band spinomuscular coherence for controlling precision grip is currently unclear, but it could help integrating multimodal afferent inputs required to produce precise force output (Omlor et al. 2007).
Another type of spinomuscular coherence described in the present study exhibited a much wider range of coherent frequencies (10–95 Hz), suggesting synchronous activity of LFPs and EMGs in a wider frequency band. In contrast to the NB, the BB was recorded only from the ventral portion of the spinal cord (Fig. 3F). The most likely source of this BB coherence is therefore the activity of motoneurons projecting muscles active during precision grip. This conclusion was also confirmed by the facts that all BB pairs (9/9) showed a significant negative linear relationship with constant time delay (8.5 ± 1.8 ms) and that BB pairs had narrower peaks in the cumulant density functions than that of NB. Because it can be assumed that a narrower correlation peak reflects a more direct coupling between the neuronal activities (Vaughan and Kirkwood 1997), the result suggested BB was generated by relatively direct connections from motoneurons to muscles, whereas NB can be generated the more indirect connections between spinal interneurons and muscles. It is well known that the neuromuscular junction has a high safety margin that minimizes transmission failure from motoneurons to muscle units (Trontelj et al. 2002), so motoneuronal activity could be directly represented in muscle activity. In fact, the firing pattern of every recorded motoneuron was closely correlated to the EMG profiles of target muscles in walking cats (Hoffer et al. 1987). In addition, motoneurons are known to respond well to the oscillatory excitatory input with wide frequency range (Matthews 1997). Therefore it is likely that spinal LFPs with BB reflect the activity of motoneurons that can drive motor units at a wide range of different frequencies.
In conclusion, we found the oscillatory LFPs in the spinal gray matter of monkeys performing a precision grip task that was coherent with forelimb muscle activity in either a restricted or broad frequency range. Spinomuscular coherence could reflect oscillatory input from descending and/or peripheral afferents. Elucidating mechanisms underlying spinomuscular coherence should facilitate our understanding the functional roles of motor system oscillatory activity in controlling voluntary movements.
This study was supported by Japan Society for the Promotion of Science KAKENHI (1802928) and Grant-in-Aid for Scientific Research on Priority Areas [Mobilligence] and [System study on higher-order brain function] from Ministry of Education, Culture, Sports, Science and Technology (18020030, 18047027).
The authors thank Drs. Stuart Baker, Tim Aumann, and Tadashi Isa for helpful comments and discussions, Dr. Paul Cheney for the advice of EMG implant to hand muscles, N. Takahashi for expert technical assistance, L. Shupe for programming assistance, and B. Brown for surgical advice. The authors also thank Dr. Tadashi Isa for his generous support in setting up our laboratory.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2008 by the American Physiological Society