| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, Kansas
Submitted 13 August 2007; accepted in final form 23 December 2007
 |
ABSTRACT |
|---|
|
0.8) to target muscle EMG activity. Our results provide evidence in support of the notion that corticomotoneuronal output from primary motor cortex encodes movement in a framework of muscle-based parameters, specifically muscle-activation patterns as reflected in EMG activity. |
|
INTRODUCTION |
|---|
|
; Fetz and Cheney 1980; Schieber and Rivlis 2005). Similarly, postspike suppression (PSpS) identifies the presence of underlying inhibitory synaptic linkages (Kasser and Cheney 1985). Muscles with PSpF or PSpS are defined as the cell's target muscles. PSpF effects with durations at half magnitude of 
9 ms can be attributed to underlying monosynaptic connections (Baker and Lemon 1998; Schieber and Rivlis 2005). Accordingly, cells producing these PSpF effects can be more confidently categorized as corticomotoneuronal (CM) cells.
Given that the presence of PSpF is evidence of an underlying synaptic linkage and that neurons producing PSpF represent the output signal from motor cortex to spinal motoneurons, a fundamental issue concerns the extent to which the activity of these cortical cells predicts or even encodes target muscle EMG activity (Hamed et al. 2007; Schieber and Rivlis 2007; Towsend et al. 2006). There is an underlying assumption that if postspike effects on muscle activity are functionally meaningful, then the cells producing the effects and their target muscles should show some level of covarying activity during task performance. Our previous work (McKiernan et al. 2000), using long-duration cross-correlations of continuous data (Houk et al. 1987), suggests this is true for identified CM cells. Taking this analysis a step further, one might also expect that the temporal pattern of activity of an individual CM cell might closely resemble the temporal pattern of target muscle EMG activity. However, these expectations must be tempered by the fact that muscle activation reflects the summation of converging excitatory postsynaptic potentials from many cells terminating within the motoneuron pool. A single cell will make only a small contribution to overall motor unit activation, so its relationship to the pattern of target muscle activity may be weak and variable. In view of this, one minimal expectation might be that the activity of the majority of CM cells and their facilitated muscles should at least show coactivation during the same segment of a movement task and that their peaks of activity should exhibit overlap. Although individual CM cells might have temporal patterns of activation that closely match the pattern of target muscle EMG activity, this is not essential. However, it is true that the ensemble activity of an identified population of CM cells sharing a common target muscle should have a temporal pattern of activity during movement that closely resembles the pattern of EMG activity, assuming that CM input to the motoneuron pool is a major factor driving motoneuron depolarization underlying an EMG peak and assuming a relatively linear transformation of cortical spike trains into EMG activity.
To further investigate the extent of covariation between CM cells and their target muscles, we have identified where peaks in their activity occur during a forelimb reach-to-grasp task and have quantified the extent of overlap between them. The results show that 71% of CM cell peaks match a target muscle peak in the same task segment. CM cell peaks show an average of 74% overlap with peaks in their target muscles. We also report significantly improved correlations between the ensemble activity of a population of CM cells influencing the same target muscle and that muscle's EMG activity compared with the individual CM cell correlations.
|
|
METHODS |
|---|
|
Two male rhesus macaques (Macaca mulatta) were trained to perform a reach-to-grasp task as described previously (McKiernan et al. 1998). Inside a sound-attenuating chamber, the monkey was seated in a custom-built primate chair facing a computer monitor providing audiovisual cues. The monkey's left arm was comfortably restrained and the task was performed with the right arm. The task was initiated when the monkey placed its right hand, palm down, on a pressure-detecting plate (home plate) at waist level in front on the right side. Holding the plate down for a preprogrammed length of time (1–2 s) triggered the release of a food reward into a cylindrical well at arm's length from the monkey. The monkey then grasped and brought the food reward to its mouth. This task provided a robust paradigm in which to test relationships between CM cell and target muscle activity. The task broadly coactivated both proximal and distal forelimb muscles while at the same time yielding a relatively high level of fractionation in terms of the detailed structure of the EMG pattern in different muscles.
Surgical procedures
After training, a 22-mm-diameter stainless steel chamber was centered over the hand area of M1 of the left hemisphere of each monkey and anchored to the skull with 12 vitallium screws and dental acrylic. Threaded stainless steel nuts were also attached over the occipital aspect of the skull using 12 additional vitallium screws and dental acrylic. These nuts provided a point of attachment for a flexible head-restraint system during recording (McKiernan et al. 1998, 2000).
EMG activity was recorded with pairs of insulated, multistranded stainless steel wires inserted transcutaneously into each of the target muscles (McKiernan et al. 1998, 2000). Electrode locations were confirmed by stimulation through the electrode pair and observation of appropriate muscle twitches. Electrode wires and connector terminals were affixed using medical adhesive tape. Following surgery, the monkeys wore a Kevlar vest and sleeve to protect the implant. EMG activity was recorded simultaneously from 22–24 forelimb muscles (Table 1).
|
Cortical recording
Single cells in primary motor cortex (M1) were recorded using glass-and-Mylar–insulated platinum–iridium electrodes with typical impedances between 0.7 and 1.5 M
. A recording electrode was positioned within the chamber using an X–Y coordinate manipulator and was advanced into the cortex with a manual hydraulic microdrive (FHC). Electrode orientation was at a right angle to the cortical surface.
Spike-triggered averages
Cortical cell activity, EMG activity, and position signals were recorded on analog tape using a 28-channel TEAC instrumentation recorder. SpTAs and response averages were compiled off-line using a custom software package (Windows Neural Averager, L. Shupe, University of Washington, Seattle, WA). The action potentials of single cells in M1 served as the triggers for computing SpTAs. Single-unit spikes were isolated using an Alpha Omega MSD spike discriminator. EMG activity was routinely filtered from 30 Hz to 1 kHz, digitized at 4 kHz, and full-wave rectified. Averages were compiled using an epoch of 60 ms, extending from 20 ms before to 40 ms after the unit spike.
Segments of EMG activity associated with each spike were evaluated by the software and accepted for averaging only if the average of all data points over the entire epoch was 5% of full-scale input. This prevented averaging EMG segments where activity was minimal or absent (McKiernan et al. 1998).
Categorization and quantification of postspike effects and cell firing frequency
The CM cells analyzed here were used in previous studies of postspike effects in forelimb muscles (McKiernan et al. 1998, 2000). For the present analysis, the postspike effects of many of the cells were recomputed from tape playback and enhanced by increasing the number of trigger events.
Categorization of effects in SpTAs was based on the latency and width of effects. We estimated the minimum reasonable latency for PSpF of muscles at different joints to be: 3.4 ms for shoulder muscles, 4.2 ms for elbow muscles, and 6.0 ms for intrinsic hand muscles (McKiernan et al. 1998, 2000). Effects with shorter latencies were presumed to have synchrony components. Schieber and Rivlis (2005) evaluated PSpF effects using a criterion developed by Baker and Lemon (1998) derived from a spike-triggered averaging simulation model. This model suggests that pure PSpF effects arising from underlying monosynaptic connections with motoneurons can be identified based on the peak width of PSpF at half-magnitude (PWHM). A PWHM of 9 ms was suggested as an effective criterion for identifying PSpF effects that are most likely due to underlying monosynaptic PSpF (Baker and Lemon 1998). A PWHM of 9 ms was the criterion applied by Schieber and Rivlis (2005) and we have also adopted this criterion. Taking into account these latency and width factors, in this study we categorized PSpF effects as 1) pure PSpF if this was the only effect present and its PWHM was 9 ms; 2) PSpF on synchrony (PSpF+Sync) if a primary PSpF could be identified based on a discontinuity in the slope of the rising phase of an underlying synchrony facilitation and the primary PSpF effect possessed a latency consistent with a minimum cortex to muscle pathway (Flament et al. 1992); 3) late widening PSpF (Schieber and Rivlis 2005) if only a primary effect was present but the PWHM was >9 ms and its latency could be explained without requiring the presence of synchrony; and 4) pure synchrony facilitation (SyncF) if the effect was broad with an onset latency inconsistent with a realistic minimum cortex to muscle pathway and no primary PSpF could be identified as a sharp peak riding on a broad synchrony peak. A similar categorization was used for suppression effects. Although synchrony effects are of interest and may contain a component mediated by a synaptic output linkage between the cortical cell and motoneurons, for the purposes of this study, we have excluded synchrony effects from the analysis. All effects included in this study were either pure PSpF or late widening PSpF effects. For convenience, we will refer to cortical cells producing these effects as CM cells.
All identified postspike effects were assigned a ranking of weak, moderate, or strong based on the magnitude of the effect (Fig. 1). Nonstationary, ramping baseline activity was routinely subtracted from SpTAs using our analysis software. The EMG values from a range of bins in the pretrigger period were averaged to arrive at a baseline mean and SD. The baseline typically was determined by averaging a 10-ms segment of each record during the pretrigger period. The onset and offset of each peak were determined as the points where the record crossed a level equivalent to +2SD above the mean of the baseline EMG (see McKiernan et al. 1998; Fig. 4A). Peaks <2SD of baseline and peaks that remained >2SD for less than a 0.75-ms period were considered insignificant and the average was categorized as having no effect (Fig. 1). The color coding of effects based on magnitude used in Fig. 1 is maintained throughout all figures presented herein.
|
|
). Ten thousand was approximately equal to the median of the number of trigger events for all PSpFs analyzed (Park et al. 2004). Expressions for these measures are as follows
 |
 |
 |
After normalizing the P/N ratio, the magnitude of PSpF effects were categorized as follows (Fig. 1). Weak PSpF effects had peaks >2SD of mean baseline activity but <4SD; moderate effects had peaks 4SD of mean baseline activity but <7SD; and strong PSpF effects had peaks of 7SD.
The depth of modulation (DOM) in CM cell firing rate (in Hertz) was measured for all peaks using response-average records referenced to different parts of the movement sequence. CM cell activity peaks were identified in segments of the record >2SD of the baseline points. Baseline was determined from activity while the monkey's hand was on home plate (segment 1 in Fig. 2) and EMG activity was largely absent. DOM was then calculated by subtracting the cell's lowest firing rate during baseline activity from its highest firing rate during the peak of activity. Peaks in CM cell activity were then ranked by magnitude as primary (highest peak value), secondary (second highest peak value), tertiary (third highest), and quaternary (fourth highest).
|
Quantification of cell–muscle covariation
For each response average, peaks in CM cell and EMG activity were assigned to one of ten segments of the reach-to-grasp task as illustrated in Fig. 2. The details of timing for defining the boundaries of each task segment are given in the legend for Fig. 2. These segments were then used to constitute the criterion for determining whether peaks in CM cell activity were associated with peaks in target muscle EMG activity. Peaks in CM cell and target muscle EMG activity were considered "matching" if they both fell within the same segment of the task. The durations of segment 1 (on home plate) and segment 5 (in the food cylinder) were considerably longer than those of other segments and potentially could allow nonoverlapping peaks in CM cell and muscle activity to be called "matching." However, the mean peak time difference between CM cell and target muscle EMG activity was not significantly greater for these movement segments compared with other movement segments.
Our goal in segmenting the reach-to-grasp task was not only to identify the location of CM cell firing rate peaks relative to functionally distinct task segments, but also to establish a sufficient number of segments to provide reasonable temporal resolution. The onset and duration of segment 8 (at the mouth) were estimated based on the fact that the monkey's hand reached its mouth about halfway between exiting the food well and depressing home plate.
As noted earlier, one objective of this approach was to document what phase of the reach-to-grasp task engaged the activity of each CM cell and its target muscles. This approach also provided a measure of the extent to which peaks in CM cell and target muscle activity occurred during the same functionally distinct task segment. Given the fact that a single CM cell is just one of hundreds of cells contributing to the activity of motoneurons belonging to the target muscle, it is unreasonable to expect that the cell and muscle peaks should necessarily be completely overlapping and coincident. However, if the cell is part of a larger neural network causally involved in generating muscle activity, it is reasonable to expect that the peaks of activity should at least be partially overlapping and would occur during the same functional task segment. To quantify the temporal coupling between CM cell and target muscle EMG activity we measured the time difference between their matching peaks—that is, peaks falling within the same segment of the task—and the extent of overlap between peaks.
Covariation was visualized and quantified by plotting CM cell firing rate in response averages against target muscle EMG point for point as a scatterplot (Griffin et al. 2004; Schieber and Rivlis 2007). Pearson correlation coefficients (r) were then calculated for these scatterplots. Four response averages were generated for each CM cell as described earlier. The analysis period was sufficiently long to contain the entire movement cycle within each average. However, the average producing the highest peak in cell activity revealed the aspect of movement the cell was best related to and this average was used to calculate the correlation coefficient. For example, in Fig. 4, all four averages show a peak in CM cell activity corresponding to exiting the food well. However, the peak was most sharply defined in the average triggered from exiting the food well so that the average was selected for performing the correlation analysis. However, the correlation coefficients were very similar for all four sets of averages belonging to a particular cell.
Measurement of EMG cross talk
Cross talk between EMG electrodes was evaluated by constructing EMG triggered averages. This procedure involved using the motor unit potentials from one muscle as triggers for compiling averages of rectified EMG activity of all other muscles. The criterion established by Buys et al. (1986) was used to eliminate effects with cross talk. To be accepted as a valid postspike effect, the ratio of PSpF between test and trigger muscle needed to exceed the ratio of their cross-talk peaks by a factor of two or more. One muscle of any muscle pair that did not meet this criterion was eliminated from the data base. Based on this criterion, we eliminated at total of 11 effects from both monkeys over the course of four EMG implants.
Cortical maps
The procedure used for producing a two-dimensional rendering of the location of cortical sites was described previously (Park et al. 2001). Briefly, the cortex was unfolded and the locations of cells were mapped onto a two-dimensional cortical sheet based on the cell's X–Y coordinate, known architectural landmarks, and observations noted during the cortical implant surgeries (Fig. 3).
|
|
|
RESULTS |
|---|
|
, 2000). Spike-triggered averaging of EMG activity from 22–24 forelimb muscles yielded 187 postspike and synchrony effects as follows: 135 pure PSpF or late-widening PSpF effects, 14 SyncF, 7 PSpF+S, and 31 pure PSpS. Totals number of pure or late-widening PSpF effects obtained by joint were 18 shoulder, 28 elbow, 27 wrist, 23 intrinsic hand, and 39 digit. Of the total, 13% were strong effects (>7SDs of the baseline points), 41% were moderate effects (4- to 7SDs of the baseline points), and 46% were weak effects (2- to 4SDs of the baseline points). Eighty percent of the CM cells facilitated more than one muscle; 61% facilitated three or more muscles. CM cell–target muscle modulation during reach-to-grasp task
Response averages referenced to leaving home plate, entering the food well, leaving the food well, and returning to home plate were generated for each CM cell (Figs. 2 and 4). The maximum DOM observed among the 44 CM cells during the reach-to-grasp task ranged from 186 Hz for the 1° peak to 12 Hz for the 4° peak. The overall mean DOM for primary peaks was 80 and 56 Hz across all peaks. Figure 4 shows an example of a complete set of four response averages compiled for one CM cell. As noted in METHODS, the analysis period was sufficiently long that all segments of the task are present in each response average. The peak in activity for the cell in Fig. 4 was strongest in the response average triggered from exiting the target food well, although the peak of activity actually occurred about midway through segment 5 of the task (digits in the food well). The discharge peaked about 300 ms before leaving the food well with a DOM of 97 Hz. All four of the cell's facilitated target muscles (green and blue records corresponding to moderate and weak PSpF effects, respectively) show a peak in EMG activity within the same segment of the task (gray shading), defined as a "matching" peak. Several nontarget muscles also showed matching peaks of activity including extensor carpi ulnaris (ECU); extensor digitorum 2 and 3 (ED2,3); extensor digitorum 4 and 5 (ED4,5); extensor digitorum communis (EDC); extensor carpi radialis (ECR); flexor carpi ulnaris (FCU); and triceps long head (TLON). The peaks in activity of the cell's facilitated muscles lagged the CM cell's peak by 30–140 ms, but they all [except first dorsal interosseus (FDI)] began to rise in advance of the CM cell's peak. All of the target muscle peaks were present in the same segment of the task and overlapped substantially with the cell's peak.
For one of three CM cells with single peaks of activity, the primary EMG peaks in all facilitated target muscles occurred in the same segment of the task (100% matching); the match was 25% (i.e., one of four target muscles) for another cell and for the third cell, the primary target muscle EMG peaks were in different segments of the task. However, most CM cells had multiple peaks of modulation during the task. Total numbers of activity peaks were as follows: 3 had one peak, 6 had two peaks, 14 had three peaks, and 21 had four peaks.
The peaks in CM cell activity were distributed over the entire movement cycle. Figure 5 shows a histogram of all the CM cell peaks associated with each segment of the task coded for whether it was the cell's primary (strongest) peak or a weaker peak. The majority of firing rate peaks (44%) occurred in segments 4 (entering target food well), 5 (in target food well), and 6 (exiting target food well) of the movement cycle. Substantial numbers (28%) were also associated with segments 8 (at the mouth) and 9 (in transit back to home plate). It is noteworthy that these are all phases of the task that most heavily rely on skilled use of the distal muscles and correlates with the fact that a majority of cells (52%) facilitated distal muscles exclusively or most strongly. For example, the cell in Fig. 4 facilitated digit and wrist flexor muscles and showed a single strong peak in segment 5 of the task, undoubtedly associated with flexion of the wrist and digits related to grasp of the food pellet. The concentration of peaks in activity associated with activity in the food well and at the mouth reflects the importance of CM cells in controlling distal muscles associated with shaping the hand, grasping the reward, and release of the food pellet into the mouth. Relatively few CM cell peaks (3.5%) were associated with segments 10 (depression of home plate) and 1 (hold on home plate). The EMG levels during these segments of the task were also relatively low.
|
Peaks in CM cell firing rate and target muscle EMG activity were compared to determine the extent to which they occurred in the same segment of the reach-to-grasp movement task (defined as matching peaks). This approach is based on the rationale that although the timing and duration of peaks in individual CM cells and target muscles would not be expected to correlate perfectly, they should at least be associated with the same functional segment of the task and show some overlap. Figure 6 shows the results obtained using criteria that varied in the level of rigor needed to conclude that the cell's peaks matched the target muscle's peaks, with Fig. 6A being the most rigorous and Fig. 6C the least rigorous. In Fig. 6A, we determined the number of cells whose 1° peak was in the same segment of the task as the 1° EMG peaks of the target muscles. Since most CM cells had multiple target muscles, the percentage given for each cell reflects the fraction of target muscles that met the criterion. Secondary firing rate peaks were ignored. For 64% of CM cells (numbers 1–28) none of the target muscle primary peaks matched the cell's primary peak. For this strictest criterion, the mean CM cell–target muscle peak match was 20%, including the cells with zero matches. Some CM cells (7.0%) showed a 100% match, that is, all the cell's target muscles had their primary peak in the same segment of the task as the CM cell. The outcome did not correlate with either the number of facilitated muscles or the number of muscle peaks. We then relaxed the criterion and determined for each CM cell whether its 1° peak was in the same segment of the task as any peak in the target muscle EMG (Fig. 6B). Once again the percentage given for each cell reflects the number of target muscles that met this criterion. This yielded a mean CM cell–target muscle peak match of 45%; that is, 45% of target muscles had a peak of some magnitude in the same segment of the task as the primary peak of the CM cell. In Fig. 6C, we determined for each CM cell the percentage of target muscles that had a peak of any magnitude that matched a CM cell peak of any magnitude. This yielded an average match of 85%. Overall, 71% of CM cell firing rate peaks had a matching target muscle EMG peak. Nearly all CM cells (95%) had at least one peak that matched a peak in a target muscle.
|
Figure 7 shows two examples of identifying matching peaks in CM cell firing rate and EMG activity (peaks occurring in the same segment of the task) and how these data were used to construct the plots in Fig. 6. A subset of task segments are color coded and labeled 4–9 at the bottom of the figure. CM cell 105N6, represented by the black bars in Fig. 6, and CM cell 65N6, represented by the dark gray bars in Fig. 6, both show four peaks of activity. Both cells have a primary peak (highest firing rate) associated with segment 6 (exiting the target food well) of the reach-to-grasp task. Only 105N6 has a primary peak that matches a primary peak of EMG activity in one of its facilitated muscles [abductor pollicis brevis (APB)]. Since 105N6 had seven target muscles, 14% of all target muscles had primary peaks that matched the cell's primary peak. Similarly, 65N6 showed a 0% match (0/3) to primary peaks in its target muscles. However, 105N6's primary peak matches three of the nonprimary peaks in its muscles [the tertiary peak of triceps lateral head (TLAT) and the secondary peak of both brachialis (BRA) and brachioradialis (BR)]. This yields a 57% match between the cell's primary peak and any target muscle EMG peak (Fig. 6B). 65N6 shows only one target muscle peak match with its primary peak (secondary peak of ED4,5) yielding a 33% match based on the criterion of Fig. 6B. Taking this analysis further, all seven of 105N6's target muscles show at least one peak that matched one of the cell's peaks, yielding a value of 100% in Fig. 6C. By this same criterion, 65N6 had two target muscles with peaks that matched one of the cell's peaks yielding a 67% match in Fig. 6C. Since 105N6 has 4 peaks of activity and facilitates 7 muscles, the total possible matches would be 28. However, only 14 peaks in EMG activity actually match peaks in CM cell activity—a 50% match in Fig. 6D. 65N6 has 4 peaks of activity and facilitates 3 muscles, yielding 12 total possible matches. However, only 2 actual matches were observed for this cell and its target muscles—a 17% match in Fig. 6D.
|
To provide detailed information on timing, we measured the time lag between matching peaks in CM cell and target muscle EMG activity. Figure 8, A and B shows the distribution of time lags plotted according to the strength of synaptic connection (magnitude of PSpF; Fig. 1) and cell firing rate modulation (DOM). The cell–muscle peak time difference was determined using the time corresponding to the highest point in the peak for both unit and EMG activity. Fifty-six percent of peaks were within ±100 ms of each other. Based on analysis of 190 matching activity peaks, the CM cell peak led the target muscle EMG peak by an average of 23 ± 150 ms (Table 2). The median CM cell to EMG peak time differences were not statistically significant for the different distributions based on the magnitude of PSpF (P = 0.68, Kruskal–Wallis) or DOM (P = 0.18, Mann–Whitney). The peaks in the timing distributions for different strengths of PSpF were similar. However, the tightest coupling (smallest range) between peak time in CM cell activity and target muscle EMG activity occurred for cell–muscle pairs exhibiting strong and moderate PSpF (P < 0.01, Levene median test). A similar result was obtained for DOM. The distribution of timing between CM cell and target muscle EMG peaks was narrower (less variability) for CM cells with high DOM, >75 Hz, compared with those with DOM, <75 Hz (P < 0.05, Levene median test). The same result was obtained with a DOM cutoff of 50 Hz. Note that DOM and strength of PSpF were not significantly correlated (r = 0.01, P = 0.95).
|
|
50% overlap: 81% of CM cell peaks showed 50% overlap by one or more individual target muscles. Correlations between CM cell activity and target muscle EMG activity
Cell–target muscle covariation during the reach-to-grasp task was quantified by plotting the average CM cell firing rate during reach-to-grasp against target muscle EMG (Fig. 9). Scatterplots were generated from this procedure and subjected to correlation analysis. Pearson's correlation coefficient (r) was used to quantify the covariation of CM cell and target muscle activity during the reach-to-grasp movement cycle. Firing rate was plotted against EMG activity with no time shift, based on the rationale that the time delay between the firing of a CM cell and its effect on muscle activity should be roughly equal to the conduction time through the CM pathway to muscles and should be approximated by the peak latency of PSpF (see DISCUSSION). This latency is in the range of 8–14 ms (Park et al. 2004), depending on the muscle, and can be ignored for this analysis because it is close to our sampling rate, that is, one sample point. Firing rates and corresponding target muscle EMGs that have the same temporal profile with no phase shift should have correlations close to one. In Fig. 9, the bulk of points in the >60-Hz firing rate range are from a part of the response average record containing the cell's primary activity peak. During much of this time, ECR's EMG activity was relatively flat. This generates a group of points that are relatively constant on the EMG axis but vary over the range from 60 to 100 Hz on the CM cell firing rate axis. These points tend to diminish the overall correlation since throughout most of the remainder of the record, CM cell and muscle activity covary more closely. The broader, slower trends in firing rate and EMG activity contribute significantly to the overall strength of the correlation.
|
|
The magnitude of pure PSpF (PPI and normalized P/N ratio) was plotted against the Pearson correlation coefficient (r) for all 135 cell–target muscle pairs that showed PSpF and had 2,000 or more sweeps in the SpTA. Although the correlations were weak, PSpF magnitude measured as P/N showed a significant positive relationship with CM cell–target muscle covariation (r = 0.25, P < 0.01), although this weakened to only a trend toward significance when P/N was normalized (r = 0.13, P = 0.14). Using PPI as a measure of PSpF yielded no significance or trend (r = 0.03, P = 0.72). It is worth noting that differences in baseline magnitude can potentially distort the true strength of PSpF based on PPI measurements.
DOM relationships
Depths of modulation of individual CM cell firing rate peaks were plotted against Pearson's correlation of the covariation between the cell firing rate and target muscle EMG activity. Of 115 cell–muscle pairs at least one had a "matching" peak of activity. In the case of multiple "matching" peaks, the values used were based on the response average with the highest DOM peak. There was no statistically significant tendency for r to be higher for greater DOMs. There was no relationship between DOM and any measure of PSpF magnitude.
Covariation and PSpS
For 31 cell–target muscle pairs that exhibited PSpS, 29% (9/31) had a negative r (compared with 16% of cell–target muscle pairs producing PSpF effects). The magnitudes of pure PSpS effects (PPI and normalized P/N) were plotted against r for all 31 cell–target muscle pairs that showed PSpS and had 2,000 or more sweeps in the SpTA. PSpS magnitude did not show a statistically significant relationship with r nor did the relationship change when the analysis was limited to moderate and strong effects.
Covariation and synchrony effects
The analysis thus far was limited to PSpF or PSpS without evidence of early onset synchrony. However, we did identify synchrony effects and test their relationship to the strength of covariation. The magnitude of synchrony effects expressed as PPI or normalized P/N was plotted against r for all 21 cell–target muscle pairs that showed either SyncF (n = 14) or PSpF+Sync (n = 7) and had 2,000 or more sweeps in the SpTA. The strength of covariation between CM cell and target muscle activity based on r was not significantly correlated with SyncF magnitude. This was also true for effects rated as moderate or strong.
Correlations with a CM cell's full muscle field
One factor that might contribute to weak covariation between CM cells and their target muscles is the fact that the output from most CM cells is not limited to one muscle but rather diverges to influence multiple muscles. This raises the possibility that the activity of a CM cell might covary more closely with the summed activity of all of its target muscles rather than with any one muscle. To test this hypothesis, the response averages of all target muscles for 12 representative CM cells were summed together after weighting by the magnitude of PSpF for each muscle. CM cells were selected using the criterion that the PSpF in at least one muscle had to be strong or moderate. Scatterplots were generated by plotting each CM cell's firing rate record against the summed EMG activity of all its facilitated target muscles. The resulting correlation coefficient was then evaluated for improvement compared with that of the individual cell–target muscle pairs. Only 3 of 12 CM cells showed stronger correlations with the summed target muscle EMG record compared with the best correlation with an individual muscle. All 3 of these were CM cells with a distal only or proximal only muscle field. Also, the mean of correlations between the CM cell's firing rate record and the summed EMG of all its target muscles was not significantly different from the corresponding mean of all the individual CM cell–target muscle EMG correlations (P = 0.41).
Populations of CM cells converging on a common target muscle
A major contributor to disparities evident earlier between CM cell and target muscle covariation is undoubtedly the fact that the activation of muscles is the result of synaptic input from many CM cells (and other cells), not just the recorded cell. Clearly, the input from one cell alone will have only a weak effect on the firing of motoneurons and, based on that, it is perhaps unrealistic to expect that the activity of one CM cell should correlate closely with the activity of a particular muscle, even though the cell directly facilitates that muscle. However, it is reasonable to expect that a population of CM cells influencing the same muscle should be a much better predictor of the pattern of EMG activity (Fetz et al. 1989; Griffin et al. 2004; Schieber and Rivlis 2007). To test this hypothesis, we identified populations of CM cells influencing the same target muscle and correlated the summed population activity to the muscle's EMG activity. Of course, the optimal way to perform this experiment would be to simultaneously record from many CM cells that all have at least one target muscle in common. However, lacking this type of data, which would undoubtedly be very difficult to obtain, we have tried to take advantage of our existing data from sequentially recorded individual CM cells. To simulate the conditions that would exist with simultaneously recorded CM cells, we have only selected cells for which the temporal pattern of EMG activity in the muscle of interest was very similar. For example, Fig. 11B shows the EMG records for ED2,3 recorded with three different CM cells (Fig. 11A) aligned on entering the food well (Fig. 11E). The EMG records for ED2,3 in Fig. 11B have the same number of peaks with similar timing and width, thereby meeting the criterion for creating a population from the corresponding CM cells. The average firing rate records of these CM cells were then summed together as were their associated EMG records (Table 3). Our data base contained many more individual CM cells that facilitated each of these muscles, but we excluded them based on the dissimilarity in their EMG pattern during the reach-to-grasp task.
|
|
We were able to apply this analysis to seven muscles in total (Table 3). For four of these muscles, the criterion that the temporal pattern of EMG activity for the muscle of interest had to be similar for each individual CM cell required splitting the CM cells for these muscles into multiple populations. Our final data set consisted of 10 CM cell populations ranging in size from three to five cells. For all but two (Table 3, ED2,3 population 2, ECR) of these CM cell populations, the population correlation coefficient was either equal to or greater than the correlation coefficient of any individual cell–target muscle pair in the population. However, all but one of the 10 population correlation coefficients were greater than the corresponding means of the individual cell–target muscle correlations. Additionally, the overall mean of the 10 population correlation coefficients was significantly greater than the overall mean of the individual correlation coefficients for each population (r = 0.75 vs. r = 0.58, P = 0.02).
We went to great lengths to select CM cells that had a very similar pattern of EMG activity for the muscle in question. The lowest value of the correlation coefficients between each muscle in a set and the summed EMG for that set ranged from 0.79 to 0.96 (Table 3, column F). Nevertheless, to further test the possibility that improvement in the correlation between the population CM cell firing rate and summed EMG records could have been due to some nonspecific smoothing effect of summing records together, we compared the set of r values obtained from correlating the population CM cell activity and summed EMG (Table 3, column E) to the mean r values obtained from correlations of the population CM cell activity to the individual EMG records (Table 3, column G). This comparison was not significant (P = 0.43, Mann–Whitney test), supporting the contention of statistical equivalence between the summed and individual EMG records. Finally, we also compared the mean r from the individual CM cell–muscle pair correlations (Table 3, column D) against the mean r derived from correlating the population CM cell activity with each individual muscle EMG (Table 3, column G). The population CM cell activity yielded a stronger correlation, but fell slightly below the 0.05 level of statistical significance (r = 0.55 vs. r = 0.44, P = 0.08). Nevertheless, it seems reasonable to conclude that using the population CM cell activity—rather than summing together the EMG records—was a major factor contributing to the improvement in CM cell–muscle EMG correlations.
|
|
DISCUSSION |
|---|
|
The interpretation of data presented herein is subject to two points of view. On the one hand it could be argued that for neurons constituting a descending system, which is supposedly driving muscle activity, the level of mismatch between the firing rate peaks of individual CM cells and their target muscles seems rather astounding. For example, on average, only 20% of CM cells had their primary peaks in the same segment of the task as the primary peaks in their target muscles. Relaxing the criterion to include any magnitude EMG peak occurring in the same segment of the task as the cell's primary peak resulted in a match of 45%—better, but still surprisingly low.
Alternatively, the similarities in activity between the temporal pattern of activity in CM cells and their target muscles could be emphasized. For example, nearly all CM cells (95%) had a least one firing rate peak that matched (occurred in the same task segment) an EMG peak in at least one of its target muscles. CM cell firing rate peaks also showed substantial overlap (mean = 74%), with peaks in individual target muscle EMG records and the amount of overlap increased to 90% for cell–muscle pairs producing strong PSpF. In this case, it should be noted that even though CM cell and target muscle peaks overlap, the activity of the two may have actually been negatively correlated during part of the overlap period, with one signal increasing and another decreasing. Nevertheless, these results are evidence in support of the general conclusion that CM cells exhibit relatively strong and consistent coactivation with their target muscles and this is particularly true of CM cell–target muscle pairs exhibiting strong PSpF.
This conclusion supports the findings of McKiernan and colleagues (2000) whose study used long-duration cross-correlation analysis. Although the long-duration cross-correlation method has its strengths—for example, it yields a coefficient that describes the correlation and a measure of time lag—it lacks information about where the activity for both the cell and muscle occurs relative to the task. In the present study, we have been able to describe where CM cell activity peaks occur in relation to their target muscles in a linear nonshifted correlation. This study also demonstrates that the highest peak of cell activity is often not associated with the activity peak of the cell's target muscle and therefore shifting the EMG signal to match the highest cell activity peak may be imposing an arbitrary association. Also, since the long-duration cross-correlation method uses single continuous trial records, it is subject to trial-by-trial variability. The analysis method of this study uses averages of multiple trials that remove the trial-by-trial variability. We have made the argument that if CM cells are linearly encoding EMG activity, using the present analysis methods, one would expect to see correlation coefficients approaching one without shifting the signals relative to one another.
Correlation studies are an approach to quantifying the extent of linear covariation between CM cells and EMG activity. We plotted the average firing rate records of CM cells against the corresponding target muscle EMG records and subjected the resulting scatterplots to correlation analysis (Fig. 9). The correlation coefficients ranged from –0.69 to 0.91 for individual cell–muscle pairs with PSpF. The median r value was 0.46 with a peak between 0.5 and 0.6 (Fig. 10). Overall, the correlations for individual cell–muscle pairs would have to be judged as relatively weak, a result that is consistent with the findings of other studies on cortical cells and their facilitated muscles (Schieber and Rivlis 2007). One might expect our results to show even weaker correlation coefficients than those reported by Schieber and Rivlis (2007) since we have used a highly complex multijoint reaching task that broadly activates forelimb muscles while at the same time fractionating peaks of activity into unique synergies and ultimately providing a robust paradigm with which to test relationships between CM cell and target muscle activity.
What factors might contribute to the existence of major disparities in the location of movement-related activity peaks in CM cells compared with their target muscles and to associated weak correlation coefficients? Certainly a major issue is the fact that the depolarization of motoneurons underlying muscle EMG activity results not from the action of just one CM cell but from many CM cells converging on a particular motoneuron pool. In addition, there are numerous additional sources of input to the motoneuron pool that can influence motoneuron activity independent of corticospinal input. At any given time during movement, a single motoneuron is receiving modulated input from hundreds, if not thousands, of afferent neurons. Another factor that might degrade the fidelity of covariation between a CM cell and its target muscles is the fact that most CM cells do not influence just one muscle; rather they influence multiple muscles as a synergy. We tested the possibility that correlations might be stronger if a CM cell's complete muscle field were taken into account. Each muscle of a CM cell's muscle field was weighted according to the magnitude of PSpF and the resulting EMG records were then summed together. The summed record was correlated with the cell's firing rate record. However, in most cases, the summed record did not result in significantly stronger correlations than the individual muscle EMG records. Using a similar approach, Schieber and Rivlis (2007) also reported that summing the EMG records of all the target muscles failed to substantially improve the correlations. However, in an interesting modification of this type of analysis, Townsend et al. (2006) recently showed that the EMG activity of all a cell's target muscles could be used to accurately predict CM cell activity and that the prediction accuracy increased with the size of the muscle field.
In view of the potential sources of disparity, it is only reasonable to predict major dissimilarities in the pattern of activity of any single CM cell and its target muscles. In fact, it might be considered remarkable that the timing of firing rate peaks between single CM cells and target muscle EMG activity are as close as they are and that the correlation coefficients are as strong as they are.
Predicting EMG from population CM cell activity
Assume that corticospinal input to motoneurons is the principal driving force under at least some conditions, essentially eliminating multiple sources of synaptic input as a factor contributing to degradation in the strength and quality of covariation between CM cell and EMG activity. In this case, motoneurons would be depolarized by the actions of multiple CM cells and other corticospinal neurons. The ensemble firing rate record of a sufficiently large population of CM cells synaptically coupled to motoneurons of the same muscle might then approach a perfect correlation with the muscle's EMG activity. To the extent that this was possible within our data set, we attempted to test this possibility. We found that in many cases (6 of 10), the temporal pattern of the ensemble firing rate record for the CM population closely resembled (r 0.8) the EMG activity of the common target muscle (Fig. 11). Perhaps most noteworthy is the fact that in all cases except one, the population correlation coefficient was greater than the corresponding mean of the individual cell–target muscle correlations (Table 3). The one exception was ECR, where the population and individual r values were essentially the same. Moreover, the mean of the population correlation coefficients for all 10 muscles tested was significantly greater (P = 0.02) than the mean of all the individual cell–target muscle correlations. Finally, in all cases except two (Table 3), the population r was essentially equal to or greater than the highest individual cell–muscle correlation.
Some individual cell–target muscle pairs had very strong correlations as Schieber and Rivlis (2007) also reported. However, the key issues are whether the population correlation is better than the individual cell–target muscle correlations and whether the final population correlation achieves a level consistent with concluding that the cells as a population could potentially account for a large part of the time-varying pattern of EMG activity during movement. We believe our data are consistent with this interpretation and add further support to the notion that CM cell output encodes muscle activation (EMG) and should be viewed within the context of a muscle-based coordinate system (Hamed et al. 2007; Holdefer and Miller 2002; Morrow and Miller 2003; Morrow et al. 2007; Mussa-Ivaldi 1988; Todorov 2000; Townsend et al. 2006).
Due to the small size of our populations, we could not analyze, in any meaningful way, changes in the population r with addition of new cells and increase in the size of the population. However, Schieber and Rivlis (2007) were able to do this, using larger populations of CM cells recorded in relation to finger movements. They showed that the pattern of improvement or decline with cell number depended on the order in which cells were added into the population. Using an order that was essentially random, the population r value fluctuated over a large range with small numbers of cells but then converged toward the final population r. However, despite larger populations of CM cells, the correlations reported by Schieber and Rivlis (2007) were weaker overall than those we have reported in this study. Their strongest population r value was 0.657 (R2 value of 0.431). In contrast, 60% (6 of 10) of our CM cell populations had greater correlations than this and the overall mean r value was 0.75. The reason for this difference is unclear. The muscles that form our CM cell populations are entirely distal muscles, mostly digit muscles (Table 3). Although our behavioral task was an unconstrained "free-form" task that might have provided a greater opportunity for yielding a higher level of sculpting of individual muscle EMG activity than the digit flexion/extension task used by Schieber and Rivlis (2007), the fact that their correlations included 12 separate movement conditions potentially added a much greater opportunity for disparities to occur between cell and muscle activity and this may have contributed to the differences in the strength of correlations between our two studies.
Our results also suggest that cortical input to the motoneuron pool dominates the activity of the motoneurons during the reach-to-grasp movement. If not, other excitatory inputs to the motoneuron pool must show temporal modulation closely matching that of the CM cell input. A significant contributor to the strength of correlations observed in our data involves the broader periods of coactivation. We agree with the interpretation of Towsend et al. (2006) that this broad coactivation "accounts for the general correlation between the envelopes of cell and muscle activity." Superimposed on this broad coactivation are peaks and valleys of activity. Our analysis of these peaks in activity showed a relatively poor correlation between the existence of CM cell primary activity peaks and primary peaks or lesser peaks in the target muscle EMG activity. However, it was true that for 73% of the CM cells, at least one peak in each of the cell's target muscles had a matching peak of some size in CM cell activity. Moreover, the timing of the peaks was relatively tight (25-ms mean with EMG peak lagging; Table 2).
CM cell effects on motor unit firing: timing issues
What timing should be expected between peaks in CM cell activity and the effect of that activity on motor unit firing rate? Many studies going back to the original work of Evarts (1968) demonstrated that cells in motor cortex show a wide range of timing relationships relative to movement onset with some neurons beginning to fire before the onset of movement and others following the onset of movement. However, nearly all these studies have shown that the mean onset time of the cortical cell population ranges from 60 to 150 ms before the onset of movement (Porter and Lemon 1993). Extending this analysis to CM cells, Fetz and Cheney (1980) showed that the mean onset of activity relative to the onset of target muscle EMG activity for a simple alternating wrist flexion/extension task was 71 ms (phasic–tonic CM cells). Despite these findings, we agree with Schieber and Rivlis (2007) that logical analysis would suggest that the timing should equal the conduction time through the pathway from cortical cell discharge to motor unit discharge (Morrow and Miller 2003; Towsend et al. 2006). This time can be estimated from the onset latency of PSpF. However, the cell's peak effect on motor unit firing would more likely correspond to the peak latency of PSpF. It is reasonable to conclude that the timing difference between a CM cell's firing rate peak and its maximum effect on motor units should also be the peak latency of PSpF. Peak PSpF latencies range from 9 to 13 ms depending on the muscle (McKiernan et al. 1998). Our sampling rate for response averages was 100 Hz or 10 ms for both unit activity and EMG channels. This means that the time shift expected between a CM cell's firing rate and its effect on motoneurons is about equal to one sample point—in other words, negligible for our purposes. Accordingly, in plotting CM cell firing rate against EMG activity and performing the Pearson correlation analysis, we did not time shift the records in an effort to achieve stronger correlations. Time shifting records might have provided stronger correlations in some cases, but we believe that such time shifting does not match the reality of timing that should exist between peaks in CM cell activity and when that activity should exert its maximum excitatory influence over motoneurons (Morrow and Miller 2003; Schieber and Rivlis 2007; Towsend et al. 2006).
Our data provide some support for this view of the timing between CM cell activity and target muscle EMG. Of the 190 cell–target muscle activity peaks occurring during the same segment of the reach-to-grasp task, the peak of CM cell activity led the peak in target muscle EMG by an average of 25 ± 150 ms (Table 2). This number is very close to the estimated time of 9–13 ms based on the peak latency of PSpF. Restricting this analysis to peaks occurring during the same segment of the task is justified because other peaks would be unlikely to be causally related. It is also noteworthy that the tightest coupling (smallest range) between peak time in CM cell activity and target muscle EMG activity occurred for cell–muscle pairs exhibiting strong PSpF effects.
The mean EMG peak time lag is notably shorter than the 71 ms reported in a previous study of the timing between CM cell (phasic–tonic cells) and muscle activity (Cheney and Fetz 1980). This difference may be due to differences in the behavioral tasks. The step-tracking task used by Cheney and Fetz (1980) in which wrist movement alternated between flexion and extension position zones engaged the activity of wrist and digit muscles in a heavily reciprocal pattern. While in one zone, the antagonist muscles were generally inactive and their motoneurons were hyperpolarized. Movement toward the opposite target zone then involved activation of the CM cells for that direction. However, before the appearance of agonist muscle EMG for that direction, the CM cells need to depolarize motoneurons from their hyperpolarized level to firing threshold. The amount of time needed for motoneurons to reach threshold and start firing would contribute to the time delay between the onset of CM cell firing and the onset of target muscle EMG activity. The reach-to-grasp task we have used in the present study differs fundamentally from the reciprocal wrist movement task in that it requires a "free-form," coordinated, multijoint reaching movement to a visual target where a food morsel is grasped and carried to the mouth and then the hand is returned to the starting point. EMG activity during this task shows broad coactivation throughout most of the task, with specific sculpting of EMG peaks and valleys evident for individual muscles. What is significant about this task is that EMG activity is always present (except on home plate), so peaks in CM cell firing rate should be translated immediately into firing rate changes of the motoneuron without the need to first depolarize the motoneuron to threshold. This fact could have significantly reduced the time difference observed in this study between CM cell firing rate peaks and corresponding target muscle EMG peaks.
Schieber and Rivlis (2007) tested the effect of time shifting the population activity of cortical cells with respect to the cell's target muscle and found that, in one monkey, the maximum correlation was obtained with the EMG delayed 40–60 ms from the cell activity. The effect of time shifting was not nearly as dramatic in another monkey. How might this time shift be reconciled with expectations based on conduction time in the corticospinal pathway? It is tempting to suggest that in the finger flexion/extension task of Schieber and Rivlis (2007), the possible lack of background EMG and need to raise motoneurons to firing threshold might also apply. However, as pointed out by Morrow and Miller (2003), it is difficult to explain the results of correlation studies involving activity over the whole movement cycle, if the delay of 40–60 ms is present only at the onset of movement. They further raise the possibility that persistent inward currents in motoneurons (Lee and Heckman 1998) essentially act as a low-pass filtered amplifier to produce currents that are substantially delayed from and greater than the synaptic currents. Although the correct explanation of these timing disparities remains unknown, the findings we have reported here suggest that the disparity may not be as large as previously thought.
Overall summary and conclusions
In this paper we report the results of a study of the functional activity patterns of 44 identified CM cells and their target muscles in relation to a free-form reach-to-grasp task. The peaks in activity of individual CM cells were about evenly distributed throughout the movement task, except for the starting position where EMG activity was minimal or absent. CM cell peaks occurred during segments of the task that in general correlated with the occurrence of peaks in target muscle EMG activity. Although many examples of strong correlations between the activity of individual CM cells and their facilitated target muscles were found, overall, the correlations were relatively weak. However, this should not be surprising given the large number of synaptic inputs driving motoneurons and the relatively small contribution made by any single input neuron. Although individual cell–target muscle correlations were relatively weak, the ensemble firing rate records of populations of CM cells sharing a common target muscle produced significantly stronger correlations than the mean of the individual cell–target muscle correlations. The results provide further evidence in support of the notion that cortical output encodes muscle-based parameters, specifically muscle activation as reflected in EMG activity. Morrow and Miller (2003) demonstrated that the ensemble activity of a relatively small number of unidentified cortical cells, time shifted according to the phase differences observed in analog cross-correlations, very closely matched the EMG activity of agonist muscles. Our data extend this to identified CM cells and shows that without any time shifting, the ensemble activity of small populations of CM cells produces a relatively good match (r 0.8) to target muscle EMG activity.
|
|
GRANTS |
|---|
|
|
|
ACKNOWLEDGMENTS |
|---|
|
Present addresses: A. Belhaj-Saïf, University of Fribourg, Institute of Physiology, Rue du Musee 5, CH-1700 Fribourg, Switzerland; B. McKiernan, Rockhurst University, Department of Physical Therapy, 123 Van Ackeren, Kansas City, MO 64110.
|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: P. D. Cheney, Department of Molecular and Integrative Physiology, University of Kansas Medical Center, Kansas City, KS 66160-7336 (E-mail: pcheney{at}kumc.edu)
|
|
REFERENCES |
|---|
|
Belhaj-Saïf A, Hill Karrer J, Cheney PD. Distribution and characteristics of poststimulus effects in proximal and distal forelimb muscles from red nucleus in the monkey. J Neurophysiol 79: 1777–1789, 1998.
Buys EJ, Lemon RN, Mantel GW, Muir RB. Selective facilitation of different hand muscles by single corticospinal neurones in the conscious monkey. J Physiol 381: 529–549, 1986.
Evarts EV. Relation of pyramidal tract activity to force exerted during voluntary movement. J Neurophysiol 31: 14–27, 1968.
Fetz EE, Cheney PD. Postspike facilitation of forelimb muscle activity by primate corticomotoneuronal cells. J Neurophysiol 44: 751–772, 1980.
Fetz EE, Cheney PD, Mewes K, Palmer S. Control of forelimb muscle activity by populations of corticomotoneuronal and rubromotoneuronal cells. Prog Brain Res 80: 437–449, 1989.[Web of Science][Medline]
Flament D, Fortier PA, Fetz EE. Response patterns and postspike effects of peripheral afferents in dorsal root ganglia of behaving monkeys. J Neurophysiol 67: 875–888, 1992.
Griffin DM, Hudson HM, Boudrias MH, Belhaj-Saïf A, McKiernan BJ, Cheney PD. Patterns of activation of primary motor cortex neurons (M1) and their target muscles during a reach-to-grasp task. Program No. 655.9. 2004 Abstract Viewer and Itinerary Planner. Washington, DC: Society for Neuroscience, 2004. Online.
Hamed SB, Schieber MH, Pouget A. Decoding M1 neurons during multiple finger movements. J Neurophysiol 98: 327–333, 2007.
Holdefer RN, Miller LE. Primary motor cortical neurons encode functional muscle synergies. Exp Brain Res 146: 233–243, 2002.[CrossRef][Web of Science][Medline]
Houk JC, Dessem DA, Miller LE, Sybirska FH. Correlation and spectral analysis of relations between single unit discharge and muscle activities. J Neurosci Methods 21: 201–224, 1987.[CrossRef][Web of Science][Medline]
Kasser RJ, Cheney PD. Characteristics of corticomotoneuronal postspike facilitation and reciprocal suppression of EMG activity in the monkey. J Neurophysiol 53: 959–978, 1985.
Lee RH, Heckman CJ. Bistability in spinal motoneurons in vivo: systematic variation in persistent inward currents. J Neurophysiol 80: 583–593, 1998.
McKiernan BJ, Marcario JK, Hill Karrer J, Cheney PD. Corticomotoneuronal postspike effects in shoulder, elbow, wrist, digit, and intrinsic hand muscles during a reach and prehension task. J Neurophysiol 80: 1961–1980, 1998.
McKiernan BJ, Marcario JK, Hill Karrer J, Cheney PD. Correlations between corticomotoneuronal (CM) cell postspike effects and cell–target muscle covariation. J Neurophysiol 83: 99–115, 2000.
Morrow MM, Jordan LR, Miller LE. Direct comparison of the task-dependent discharge of M1 in hand space and muscle space. J Neurophysiol 97: 1786–1798, 2007.
Morrow MM, Miller LE. Prediction of muscle activity by populations of sequentially recorded primary motor cortex neurons. J Neurophysiol 89: 2279–2288, 2003.
Mussa-Ivaldi FA. Do neurons in motor cortex encode movement direction? An alternative hypothesis. Neurosci Lett 91: 106–111, 1988.[CrossRef][Web of Science][Medline]
Park MC, Belhaj-Saïf A, Cheney PD. Properties of primary motor cortex output to forelimb muscles in rhesus macaques. J Neurophysiol 92: 2968–2984, 2004.
Park MC, Belhaj-Saïf A, Gordon M, Cheney PD. Consistent features in the forelimb representation of primary motor cortex in rhesus macaques. J Neurosci 21: 2784–2792, 2001.
Porter R, Lemon RN. Corticospinal Function and Voluntary Movement. Oxford, UK: Clarendon Press, 1993.
Schieber MH, Rivlis G. A spectrum from pure postspike effects to synchrony effects in spike-triggered averages of electromyographic activity during skilled finger movements. J Neurophysiol 94: 3325–3341, 2005.
Schieber MH, Rivlis G. Partial reconstruction of muscle activity from a pruned network of diverse motor cortex cells. J Neurophysiol 97: 70–82, 2007.
Todorov E. Direct cortical control of muscle activation in voluntary arm movements: a model. Nat Neurosci 3: 391–398, 2000.[CrossRef][Web of Science][Medline]
Townsend BR, Paninski L, Lemon RN. Linear encoding of muscle activity in primate motor cortex and cerebellum. J Neurophysiol 96: 2578–2592, 2006.
This article has been cited by other articles:
|
|
 |
|
  W. S. Smith and E. E. Fetz Synaptic Linkages Between Corticomotoneuronal Cells Affecting Forelimb Muscles in Behaving Primates J Neurophysiol, August 1, 2009; 102(2): 1040 - 1048. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Post, R. Bakels, and I. Zijdewind Inadvertent Contralateral Activity during a Sustained Unilateral Contraction Reflects the Direction of Target Movement J. Neurosci., May 13, 2009; 29(19): 6353 - 6357. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
TOP
ABSTRACT
INTRODUCTION
