Stimulus-triggered averaging (StTA) of electromyographic (EMG) activity from 24 simultaneously recorded forelimb muscles was used to investigate properties of primary motor cortex (M1) output in the macaque monkey. Two monkeys were trained to perform a reach-to-grasp task requiring multijoint coordination of the forelimb. EMG activity was recorded from 24 forelimb muscles including 5 shoulder, 7 elbow, 5 wrist, 5 digit, and 2 intrinsic hand muscles. Microstimulation (15 μA at 15 Hz) was delivered throughout the movement task. From 297 stimulation sites in M1, a total of 2,079 poststimulus effects (PStE) were obtained including 1,398 poststimulus facilitation (PStF) effects and 681 poststimulus suppression (PStS) effects. Of the PStF effects, 60% were in distal and 40% in proximal muscles; 43% were of extensors and 47% flexors. For PStS, the corresponding numbers were 55 and 45% and 36 and 55%, respectively. M1 output effects showed extensive cofacilitation of proximal and distal muscles (96 sites, 42%) including 47 sites that facilitated at least one shoulder, elbow, and distal muscle, 45 sites that facilitated an elbow muscle and a distal muscle, and 22 sites that facilitated at least one muscle at all joints. The muscle synergies represented by outputs from these sites may serve an important role in the production of coordinated, multijoint movements. M1 output effects showed many similarities with red nucleus output although red nucleus effects were generally weaker and showed a strong bias toward facilitation of extensor muscles and a greater tendency to facilitate synergies involving muscles at noncontiguous joints.
Linkages between primary motor cortex (M1) and spinal motoneurons in primates have been investigated for many years using a variety of anatomical and electrophysiological methods (Asanuma et al. 1978; Cheney et al. 1991; Donoghue et al. 1992; Futami et al. 1979; Hepp-Raymond 1988; Lemon 1988; Phillips and Porter 1977; Porter and Lemon 1993; Preston et al. 1967; Shinoda et al. 1979). Studies based on stimulation of the cortical surface have demonstrated monosynaptic excitatory postsynaptic potentials (EPSPs) in motoneurons of both distal and proximal muscles of the monkey forelimb and the hindlimb with consistently more frequent and stronger projections to motoneurons of distal muscles (Clough et al. 1968; Jankowska et al. 1975a; Phillips and Porter 1964; Preston et al. 1967). More recently, Fritz et al. (1985) and Lemon (1990) confirmed these observations in a study of 358 motoneurons supplying elbow, forearm, and intrinsic hand muscles. EPSPs were evoked by stimulating axons in the medullary pyramid in an effort to activate the entire corticospinal input to motoneurons.
Results from studies using spike-triggered averaging of electromyographic (EMG) activity from forelimb muscles are also relevant to the issue of cortical control of distal and proximal muscles. Based on a sample of 112 corticomotoneuronal (CM) cells recorded from the forelimb representation of M1, McKiernan et al. (1998) showed that the magnitude of postspike facilitation (PSpF) was continuously graded along the proximal-distal axis with the weakest effects in proximal muscles and the strongest effects in intrinsic hand muscles. The magnitude of PSpF for intrinsic hand muscles was about three times that of shoulder muscles. It is also noteworthy that 45% of 112 CM cells tested in the study by McKiernan et al. (1998) produced postspike effects (PSpE) in combinations of proximal and distal forelimb muscles.
Recent transcranial magnetic stimulation studies in humans have also demonstrated projections from primary M1 to motoneurons of both proximal and distal arm muscles (Colebatch et al. 1990; Lemon et al. 1995; Palmer and Ashby 1992). In subjects performing voluntary reaching movements during stimulation, Colebatch et al. (1990) reported that the CM projection to proximal muscles is as strong as that to distal muscles. In contrast, Palmer and Ashby (1992) found that TMS produced stronger net facilitation of distal muscle motoneurons than proximal motoneurons.
Whereas numerous different approaches have provided important insights concerning the organization of corticospinal input to muscles at different joints, some limitations are evident. Intracellular recording experiments have been highly informative but have been limited by the number of motoneurons that can be tested and by difficulty in identifying the specific target muscles of motoneurons. Studies with transcranial magnetic stimulation (TMS) and transcranial electrical stimulation (TES) in humans have provided data paralleling those from monkeys and with largely consistent conclusions. However, the number of simultaneously recorded muscles has been quite limited, and the extent of cortical activation with TMS is also unclear, although it is certainly large compared with low-intensity intracortical microstimulation (ICMS).
In the present study, we have attempted to overcome these limitations by recording the activity of a relatively large number (24) of forelimb muscles simultaneously, including shoulder, elbow, wrist, digit, and intrinsic hand muscles, in monkeys performing a stereotyped reach-to-grasp movement. Moreover, we have systematically explored the entire forelimb representation of M1 using stimulus-triggered averaging of EMG activity at 15 μA to generate output maps of these muscle groups that were published in a previous paper (Park et al. 2001). We have now used this same data for a systematic and detailed analysis of the sign, latency, magnitude, and distribution of output from M1 cortex. Data for this analysis were limited to cortical stimulation sites close to or in layer V. We believe the results from this approach provide an unbiased depiction of the distribution and magnitude of M1 cortical output to different forelimb muscle groups in the primate. The results also provide a comparison of output from three zones in forelimb M1 (Park et al. 2001), one influencing only distal muscles, one only proximal muscles, and the third influencing combinations of proximal and distal muscles. Finally, the results provide a database for comparison with data from red nucleus obtained under similar conditions (Belhaj-Saïf et al.1998).
Data were collected from the left primary motor cortex of two male rhesus monkeys (Macaca mulatta; ∼9 kg, 6 yr old). The stimulus-triggered averaging (StTA) data on which the analysis in this paper is based are the same data used in a previous publication focusing on mapping the distribution of effects (Park et al. 2001). The monkeys were trained to perform a reach-to-grasp task requiring coactivation of multiple proximal and distal forelimb muscles in natural, functional synergies. Training procedures and the behavioral task have been described in detail previously (Belhaj-Saïf et al. 1998; McKiernan et al. 1998). During each data collection session, the monkey was seated in a custom primate chair and placed inside a sound-attenuating chamber. The left forelimb of the monkey was restrained during task performance, whereas the right forelimb had freedom of movement.
The reach-to-grasp task consisted of four different phases. Performance was guided by audio and video cues provided by an IBM-compatible computer. The task began with the monkey's right hand resting on a home plate device at waist height and his elbow flexed at ∼90°. Having the hand on the plate for a preprogrammed length of time triggered the release of a food reward and a go signal. In the second phase of the task, the monkey reached into the target cylinder to grasp the food pellet with its fingers using a precision grip. The target cylinder was located at shoulder level, a little less than one arm length away, and oriented ∼20° from vertical. During this phase, the arm was fully extended. In the third phase, the monkey flexed its elbow and wrist to bring the pellet to its mouth. Finally, in the last phase of the task, the monkey returned its hand to the home-plate starting position.
On completion of training, each monkey was implanted with a cortical recording chamber and EMG electrodes. For all implant surgeries, the monkeys were tranquilized initially with ketamine (10 mg/kg), administered atropine, and subsequently anesthetized with isoflurane gas. Both monkeys received prophylactic antibiotic (Penicillin G, Benzathaine/Procaine combination, 40,000 IU/kg every 3 days) before and after surgery and analgesic medication (Buprenorphine 0.5 mg/kg every 12 h for 3–4 days) postoperatively. All surgeries were performed in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care using full sterile procedures. All procedures conformed to the Guide for the Care and Use of Laboratory Animals, published by the United States Department of Health and Human Services and the National Institutes of Health.
A magnetic resonance imaging (MRI)-compatible plastic chamber allowing exploration of a 30-mm-diam cortical area was stereotaxically implanted over the forelimb area on the left hemisphere of each monkey using procedures fully described previously (Kasser and Cheney 1985; McKiernan et al. 1998). The chambers were centered stereotaxically at anterior 21.6 mm, lateral 11.4 mm (monkey M) and at anterior 16.0 mm, lateral 7.4 mm (monkey D) with a 30° angle to the midsagittal plane. For MRI compatibility, titanium screws (Bioplate, Los Angeles, CA) and titanium restraining nuts (McMaster-Carr, Chicago, IL) were used. In addition, a titanium screw (Synthes, Monument, CO) in contact with the dura served as a reference ground for electrophysiology.
EMG activity from 24 muscles of the forelimb was recorded using pairs of multi-stranded stainless steel wires (Cooner Wire, Chatsworth, CA) implanted during a sterile surgical operation. One monkey was implanted using a modular subcutaneous implant technique, and the other was implanted using a cranial subcutaneous implant technique. These procedures are described in detail elsewhere (Park et al. 2000). Briefly, for both techniques, pairs of wires for each muscle were tunneled subcutaneously to their target muscles. The modular subcutaneous implant technique used four connector modules (ITT Cannon), two placed above and two below the elbow. The cranial subcutaneous implant technique used one circular connector (Wire Pro, Salem, NJ) module anchored to the skull near the cortical recording chamber. The wire insertion points for specific muscles were identified on the basis of external landmarks and palpation of muscle bellies. The wires of each pair were bared of insulation for ∼2 mm at the tip and inserted into the muscle with a separation of ∼5 mm. Proper placement was tested by stimulating through the wires with short trains or single pulses while observing the evoked movements. The wires were removed and reinserted if necessary.
EMGs were recorded from five shoulder muscles: pectoralis major (PEC), anterior deltoid (ADE), posterior deltoid (PDE), teres major (TMAJ), and latissimus dorsi (LAT); seven elbow muscles: biceps short head (BIS), biceps long head (BIL), brachialis (BRA), brachioradialis (BR), triceps long head (TLON), triceps lateral head (TLAT), and dorso-epitrochlearis (DE); five wrist muscles: extensor carpi radialis (ECR), extensor carpi ulnaris (ECU), flexor carpi radialis (FCR), flexor carpi ulnaris (FCU), and palmaris longus (PL); five digit muscles: extensor digitorum communis (EDC), extensor digitorum 2 and 3 (ED23), extensor digitorum 4 and 5 (ED45), flexor digitorum superficialis (FDS), and flexor digitorum profundus (FDP); and two intrinsic hand muscles: abductor pollicis brevis (APB) and first dorsal interosseus (FDI). At regular intervals, the monkeys were tranquilized with ketamine and the implants were tested to confirm electrode location.
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 for all other muscles. Cross-talk peaks in nontrigger muscles were expressed as a percent of the peak in the trigger muscle. In addition to common recording of the same motor unit potentials through different sets of electrodes (electrical cross-talk), physiological synchrony between motor units can contribute to cross-talk peaks. To adjust for this fact, we allowed cross-talk up to and including 15% of the trigger muscle peak. Any muscle showing cross-talk >15% was eliminated from the data base (Fetz and Cheney 1980). In this study, no significant poststimulus effects had to be eliminated due to cross-talk.
Cortical electrical activity and EMG activity were simultaneously monitored along with task-related signals. For cortical recording and stimulation, glass and Mylar-insulated platinum-iridium electrodes with impedances between 0.7 and 1.5 MΩ (Frederick Haer, Bowdoinham, ME) were used. Electrode penetrations were made systematically in precentral and postcentral cortex in a 1-mm grid interval. In some areas, electrode tracks were placed in the center of the 1-mm square formed by four adjacent tracks to achieve greater spatial resolution. The electrode was advanced with a manual hydraulic microdrive and stimulation was performed at 0.5-mm intervals, starting from the first cortical electrical activity encountered. The stimulus sites that yielded forelimb data used in this study are shown in Fig. 1, C and D. These maps are reproduced from Park et al. (2001) and show the precentral cortex unfolded with a grid of black dots indicating cortical stimulation sites corresponding to layer V. The forelimb representation shows a characteristic and reproducible intra-areal somatotopy with a core distal muscle representation (dark blue), a horseshoe shaped peripheral proximal muscle zone (red), and a zone producing effects in both proximal and distal muscles (PDC zone, purple). Additional tracks around the perimeter of the forelimb representation either did not yield effects or produced muscle twitches in the face region (light blue) or the hindlimb region (green).
While the monkey performed the reach-to-grasp task, stimuli (15 μA at 15 Hz) were applied through the electrode and served as triggers for computing StTAs. Individual stimuli were symmetrical biphasic pulses—a 0.2-ms negative pulse followed by a 0.2-ms positive pulse. EMGs were filtered from 30 Hz to 1 kHz, digitized at a rate of 4 kHz, and full-wave rectified. Averages were compiled using a 60-ms epoch, including 20 ms before the trigger to 40 ms after the trigger. Stimuli were applied throughout all phases of the reach-to-grasp task and the assessment of effects was based on StTAs of ≥500 trigger events. Segments of EMG activity associated with each stimulus were evaluated and accepted for averaging only when the average of all EMG data points over the entire 60-ms epoch was equal to or >5% of the full-scale signal. This prevented averaging segments in which EMG activity was minimal or absent (McKiernan et al. 1998).
At some stimulation sites, averages were computed at 30 μA if no poststimulus effects (PStEs) were obtained at 15 μA. When no PStEs were detected with 30 μA, repetitive ICMS (R-ICMS) was performed to determine the motor output representation, if any, from that site. R-ICMS consisted of a train of 10 symmetrical biphasic stimulus pulses (negative–positive with total duration of 0.4 ms) at a frequency of 330 Hz (Asanuma and Rosén 1972), and an intensity of 15 and/or 30 μA. Evoked movements and muscle contractions detected with palpation were noted.
MRI of cortex and analysis
MRI was used for confirmation of electrode track locations. MRI studies were performed ∼5 mo after the cortical recording chamber implant but before the EMG implant. The MRI protocol is fully described in Park et al. (2001). The monkeys were tranquilized with ketamine and atropine and subsequently anesthetized with isoflurane gas. To give the MR images a reference framework, a custom-designed chamber cap filled with MR opaque marker (liquid vitamin E) was used to identify the anterior–posterior (A-P) and medial–lateral (M-L) axes of the cortical recording chamber. Image reconstruction and analysis was performed using Omniview 2D and 3D visualization software (3D Biomedical Imaging, Shawnee Mission, KS), the details of which have been described previously (Park et al. 2001). Oblique parasagittal images of the brain were obtained at a 30° angle to the midsagittal plane. These images were in register with the chamber coordinate system. For example, an oblique parasagittal image at lateral 4 would represent a slice through the cortex showing all electrode tracks for which the M-L chamber coordinate was lateral 4. These images were then traced to highlight gray matter, white matter, and central sulcus of the M1, and used to estimate electrode track and stimulation site placement (Fig. 1B).
At each stimulation site, averages were obtained from all 24 muscles. PStF and poststimulus suppression (PStS) effects were computer measured as described in detail by Mewes and Cheney (1991). Mean baseline activity and SD were measured for each average in the pretrigger period (−20 to −7.5 ms pretrigger, 12.5 ms total). Nonstationary, ramping baseline activity was routinely subtracted from the average using custom data-analysis software. Baseline ramp subtraction avoids what would otherwise be an artifact in the measurement of poststimulus effect magnitude produced by a rising or falling baseline. StTAs were identified as having a significant poststimulus effect if the peak or trough of the effect exceeded ±2 SD of baseline for a period ≥0.75 ms (3 points). It should be emphasized that this was the statistical criterion applied to identify effects for analysis and should not be equated to the duration of poststimulus effects. The onset latency of PStF and PStS was measured as the point where the envelope of the effect intersected the line representing two SDs away from the baseline. The magnitude of PStF and PStS was expressed as the percent increase or decrease in EMG activity above (facilitation) or below (suppression) baseline (Cheney and Fetz 1985; Cheney et al. 1991; Kasser and Cheney 1985). Peak values were measured as the highest point in the peak of facilitation or lowest point in the trough of suppression.
Subsequently, stimulation sites corresponding most closely to cortical layer V in the anterior bank of the central sulcus were selected from the resultant three-dimensional matrix of stimulation site data. First, all electrode tracks were grouped according to their M-L coordinate. Within each group, the tracks were then ordered according to their A-P coordinate. On the basis of electrophysiological data and observations, a parasagittal diagram was constructed to represent the cortex that was explored and stimulated (Fig. 1A). White matter was identified by a sharp decrease or loss of background cell activity. Sensory cortex was identified by the presence of distinctive spike activity and characteristic receptive fields (Widener and Cheney 1997). For each electrode track, sites corresponding to cortical layer V were identified using a combination of electrode depth, strength of PStF effects, and reconstruction of precentral geometry in relation to MRI sections (Fig. 1B). Electrode penetrations on the convexity of the gyrus traversed cortical layers perpendicularly, and in these cases, it was relatively easy to identify the stimulation site closest to layer V. For electrode penetrations traversing the depth of the precentral gyrus and extending roughly parallel to the cortical layers, it was more difficult to identify layer V sites. In these cases, output effects from sites at the same depth from different electrode tracks along the A-P axis were compared with aid in identifying sites most likely corresponding to layer V.
Table 1 summarizes the data collected from the left M1 in two rhesus monkeys (M. mulatta). Stimulus-triggered averages (15 μA) of rectified EMG activity were collected from 24 forelimb muscles of the shoulder, elbow, wrist, digits, and intrinsic hand. R-ICMS (10 pulses at 330 Hz) was performed at sites where no PStEs were observed with StTA at intensities ≤30 μA. This was done to identify representations for muscles other than those with EMG electrodes, for example, the trunk, hindlimb, and face. Data were collected from a total of 2,836 M1 sites (StTA data from 2,477 and R-ICMS data from 359). On the basis of the criteria described earlier, stimulation sites corresponding to cortical layer V were selected, and only their PStEs were used for analysis. This yielded a total of 297 sites and 7,128 individual StTA records from which 2,079 individual PStEs were obtained including 1,398 (67%) PStF effects and 681 (33%) PStS effects. Seven hundred thirty seven (53%) of these PStEs were biphasic, consisting of facilitation followed by suppression. This suppression is probably associated with underlying inhibitory postsynaptic potentials (Lemon et al. 1987), although a contribution from motoneuron refractoriness and postexcitatory depression cannot be ruled out. Nevertheless, because onset latency and magnitude could have been obscured or altered by overlap with the earlier facilitation, the magnitudes of biphasic inhibitory effects were not measured.
Three classes of PStE were defined based on peak magnitude: weak (2–3 times the SD of the baseline points), moderate (3–6 times the SD of the baseline points), and strong (>6 times the SD of the baseline points) (Belhaj-Saïf et al. 1998; Park et al. 2001). Of the 1,398 PStF effects obtained, 9% were weak, 29% were moderate, and 62% were strong; of the 681 PStS effects obtained, 18% were weak, 54% were moderate, and 28% were strong.
Latency and magnitude
Only moderate and strong PStEs were used to calculate latency data because of the greater uncertainty in identifying the onsets and peaks of weak effects. The average PStF onset and peak latencies were 9.2 ± 1.6 and 11.6 ± 2.0 ms, respectively, compared with PStS onset and peak latencies of 14.5 ± 3.6 and 18.8 ± 5.6 ms, respectively. PStS onset and peak latencies were 5.3 and 7.2 ms longer than those for PStF. Table 2 summarizes the average onset and peak latencies for moderate and strong PStF for muscles at different joints. Statistical comparison of mean PStF onset latency at different joints were all significant except digit versus wrist (P ≤ 0.0001, 1-way ANOVA).
Each stimulation site was categorized based on distribution of PStF in different muscles. The stimulation site was considered proximal (P) if the PStF effects were only in proximal muscles (Fig. 2A). Similarly, the site was considered distal (D) if the PStF effects were only in distal muscles (Fig. 2C). If the stimulation site facilitated at least one proximal and one distal muscle, it was categorized as a proximal-distal cofacilitation (PDC) site (Fig. 2B). Table 3 shows the mean onset latencies of PStF effects for each of these categories. The onset latencies from P sites were greater than those in the same muscles from PDC sites, although only the comparison of elbow muscles reached statistical significance (P ≤ 0.05, t-test). No differences in onset latency were found between D sites and effects in the corresponding muscles elicited from PDC sites.
Figure 3 shows the distribution of PStF onset latencies for muscles acting at different joints of the forelimb. The distributions for proximal muscles were broader as reflected in the greater SDs and width of the distributions at half-maximum amplitude (3.5, 2.5, 1.5, 2.0, and 2.0 ms, respectively, for shoulder, elbow, wrist, digit, and intrinsic hand muscles). The greater variability in onset latency for proximal muscles may reflect a greater diversity of synaptic linkages and/or corticospinal neuron conduction velocities.
The average PStF magnitude expressed in peak percent increase (ppi) above baseline at 15 μA was 57.6 ± 70.7 compared with −20.4 ± 5.7 ms for PStS. The magnitude of PStF showed a progressive increase for each muscle group in going from the most proximal muscles to the most distal muscles (Table 2). Statistical comparison of mean PStF magnitude at different joints were all significant except digit versus intrinsic (P ≤ 0.0001, 1-way ANOVA).
Figure 4 shows the distribution of PStF magnitude for muscles acting at different joints of the forelimb. The distributions for distal muscles are broader than those for proximal muscles with substantial numbers of effects greater than a magnitude of 100 ppi. Only a few elbow muscle effects were >100 ppi, and none of the shoulder muscle effects were >100 ppi. The maximum effect obtained (830 ppi) was in a forearm digit muscle. Because of the skewed distributions of PStF magnitude, means and medians were substantially different for the wrist, digit, and intrinsic hand muscles (Table 2). Nevertheless, both the means and medians show the same trends. Distal muscle PStF is stronger than proximal muscle PStF, and there is a consistent increase in magnitude moving distally from shoulder muscles to elbow muscles, to wrist and digit muscles, and finally to intrinsic hand muscles.
Figure 5 shows that there was tendency for stronger PStF to be associated with shorter latencies. This tendency was most pronounced for elbow, wrist, and digit PStF (P < 0.001, t-test) but was also statistically significant for intrinsic hand muscle PStF (P < 0.01, t-test). The inverse relationship between magnitude and latency suggests that stronger PStF effects are mediated by a faster conducting pathway or a more direct synaptic linkage or both.
Distribution of PStEs
Figure 6, A and B, shows the distributions of moderate and strong PStEs in shoulder, elbow, wrist, digit, and intrinsic hand muscles. Of 1,398 PStF effects, 60% were in distal muscles including 25% in wrist, 25% in digit, and 10% in intrinsic hand muscles. Forty percent of PStF effects were in proximal muscles including 11% in shoulder and 29% in elbow muscles. Inhibitory effects showed a similar distribution. Of 681 PStS effects examined, 55% were in distal muscles including 24% in wrist, 22% in digit and 9% in intrinsic hand muscles. Forty-five percent of PStS effects were in proximal muscles including 16% in shoulder and 29% in elbow muscles. These distributions for PStF or PStS were not substantially altered by excluding weak effects.
The number of muscles recorded at each joint could have influenced the distribution of effects in muscles at different joints. For the shoulder, wrist, and digits, five muscles were sampled; for the elbow, seven were sampled; and only 2 intrinsic hand muscles were sampled. After normalizing for the number of muscles sampled, the incidence of effects in intrinsic hand muscles was more prominent (Fig. 6, C and D) matching that for the wrist and digit muscles. The incidence of elbow muscle effects was lower and shoulder muscles showed the lowest incidence of effects.
Extensor/flexor distribution of M1 output
Overall, PStF was similarly distributed in extensors and flexors of the forelimb—43% of all PStF effects were in extensor muscles (PDE, TMAJ, LAT, TLON, TLAT, DE, ECR, ECU, EDC, ED23, or ED45), 47% in flexor muscles (PEC, ADE, BIS, BIL BRA, BR, FCR, FCU, PL, FDP, or FDS). FDI and APB each accounted for an additional 5% of PStF effects. The distribution remained relatively unchanged when only strong and moderate effects were considered (42% for extensor muscles, 48% for the flexor muscles, 5% each for FDI and APB). Because there were equal numbers of extensors and flexors, normalizing for the number of recorded muscles was not necessary.
Nearly equal numbers of cortical sites produced effects in extensor and flexor muscles. Of 272 stimulation sites producing PStEs, 68% showed PStF in at least one extensor muscle and 62% showed PStF in at least one flexor muscle. Twenty-five percent of sites produced effects in FDI and 25% in APB.
PStS was more heavily weighted toward flexor muscles. Fifty-five percent of PStS effects were in flexor muscles; 36% were in extensor muscles. Many PStS effects were in antagonists of facilitated muscles at the same joint. FDI and APB accounted for an additional 4 and 5% of PStS effects respectively. Of 272 stimulation sites yielding PStEs, 50% produced PStS in at least one extensor muscle, 58% in at least one flexor muscle. An additional 12 and 10% of sites produced PStS in APB and FDI, respectively. Considering only strong and moderate PStS effects, the distribution remained relatively unchanged (36% for extensors, 54% for the flexors, 4% for FDI, and 6% for APB).
Figure 7 shows the distribution of all PStF and PStS effects in flexors and extensors of the shoulder, elbow, wrist, digits, and intrinsic hand muscles (FDI and APB). PStF was more common in extensors than flexors at the shoulder and digit joints and more common in flexors than extensors at the elbow and wrist joints (Fig. 7A). PStS was also more common in flexors than extensors at the elbow and wrist joints (Fig. 7B). After adjusting for the number of recorded extensor and flexor muscles at each joint (3 extensors and 2 flexors at the shoulder and digit joints, 2 extensors and 3 flexors at the wrist, and 3 extensors and 4 flexors at elbow), some differences became less prominent (Figs. 7C). PStF was more common in extensors at the shoulder, wrist, and forearm digits while PStF remained more common in flexor muscles of the elbow. PStS was more common in flexor muscles at all joints except the intrinsic hand (Fig. 7D).
Figure 8 shows the number of the PStF and PStS effects observed in each of 24 forelimb muscles sampled. The muscles most frequently facilitated by M1 were PDE for the shoulder, BIL and BRA for the elbow, ECU and ECR for the wrist, and EDC, ED23, and ED45 for the digits. The most frequently suppressed muscles were PDE and ADE for the shoulder, BIS, BRA, BR, and TLAT for the elbow, FCR and FCU for the wrist, and FDS for the digits. APB and FDI showed similar numbers of effects.
It should be emphasized that results very similar to those presented in the preceding text were obtained when the data were recalculated using a uniform 1-mm spacing of recording sites throughout the cortex. This issue is presented more fully in the discussion.
StTA muscle field
The term “muscle field” refers to the set of muscles with significant facilitation from single cells in spike-triggered averages of EMG activity (Cheney et al. 1991; Fetz and Cheney 1978, 1979). Muscle fields can also be characterized for sites within motor cortex activated by microstimuli, reflecting the output effects of a collection of neurons activated by the stimulus. For the purposes of this study, muscle fieldst was defined as the number of muscles showing PStF and/or PStS from sites of stimulation within M1. A superscript “st” designation will distinguish this use of the term “muscle field” from its more conventional use to describe the set of target muscles associated with single neurons as revealed by spike-triggered averaging of EMG activity. Also, in many places throughout the text, muscle field will be linked with the prefixes StTA, PStF, or PStS to further emphasize that these are muscle fields associated with stimulus-evoked output effects. The mean muscle fieldst including both PStF and PStS effects was 8.7 ± 5.0 for all sites of stimulation with 6.2 ± 4.6 muscles showing PStF and 3.5 ± 2.4 showing PStS (Fig. 9). Excluding weak effects reduced the PStF-muscle fieldst size only slightly from 6.2 to 5.8. The largest number of muscles showing PStEs from a single M1 site was 17. At this site, none of the effects were weak. Seventy-one percent of sites (193/226) yielding PStF, facilitated two or more muscles at the same joint or different joints. At 15% of sites (33/226), PStF appeared in just one muscle, and in most cases (28/33 sites), the effects were either moderate or strong. Interestingly, 30 (91%) of the 33 sites producing effects in a single muscle involved proximal muscles; only 3 (9%) involved distal muscles (FCR, ED23 and FDI).
Analysis of muscle fieldst size for different categories of M1 sites (Table 4) revealed the following significant differences (P < 0.05, t-test). The muscle fieldst size was larger for distal muscles than proximal muscles for all categories of cortical sites (P, D, and PDC) and for both PStF and PStS. The average muscle fieldst size for PStF in distal muscles from PDC sites was smaller than for pure distal sites (D), whereas the muscle fieldst size in proximal muscles was greater than for pure proximal sites. PStS-muscle fieldst sizes for proximal and distal muscles were similar at P, D, and PDC sites.
Divergence of output effects to muscles at multiple joints
Cofacilitation involving combinations of distal and proximal muscles was a very common pattern of M1 output accounting for 42% (96/226) of sites producing PStF (Fig. 10A). Forty percent of sites produced effects in only proximal muscles and 18% in only distal muscles. Including both PStF and pure PStS effects, 65% (147/226) of sites influenced both proximal and distal muscles. Of the 96 sites that cofacilitated proximal and distal muscles, 23% facilitated at least one muscle at all joints (shoulder, elbow, wrist, digit, and intrinsic hand), 49% facilitated at least one shoulder, elbow, and distal muscle, 47% facilitated an elbow muscle together with a distal muscle, and only 4% facilitated some combination of shoulder and distal muscles (Fig. 10B). Excluding weak effects did not substantially change these synergy patterns.
Table 5 summarizes the frequency with which different combinations of facilitated muscles at different joints were observed. The table is organized in ascending order of the number of muscle groups facilitated. All 31 possible combinations are shown, although not all were represented in M1 output. Table 5 also includes the observed probability of occurrence of different muscle group synergies together with a predicted probability of occurrence. To calculate the predicted probability of occurrence, each combination was initially assigned a probability of 0.0323 (1/number of possible combinations) in an effort to simulate an underlying random process guiding formation of corticomotoneuronal synaptic connections. This probability was then adjusted for the number of muscles associated with different joints and the number of recorded muscles. Joints with more muscles would have a greater probability of showing an effect, that is, a synaptic connection. Similarly, the fewer muscles actually recorded at a joint, the lower the probability of detecting an effect. The column on the far right (Table 5) gives the deviation of observed probability from expected probability. A plus sign represents a higher than expected rate of occurrence; a minus sign is a lower rate; and zero is essentially equal to the predicted rate of occurrence. Each plus and minus sign represents a difference of 1 SD. In general, the muscle synergies that occur at a higher rate than predicted (≥2 SD) are ones in which contiguous groups of muscles were facilitated, for example, facilitation of shoulder and elbow muscle groups or, in the most distributed case, facilitation of muscles at all five locations—shoulder, elbow, wrist, digits, and intrinsic hand. The one exception to this rule was high incidence, compared with predicted, of facilitation limited to elbow muscle. All other single muscle group synergies occurred at a rate equal to or lower than expected. All other combinations that were less common than predicted involved muscles at noncontiguous joints, for example, shoulder with digit muscles.
In absolute terms (number of sites column), the most frequently observed pattern was facilitation of only elbow muscles. The next most common pattern was cofacilitation of both shoulder and elbow muscles. Additional highly represented patterns included shoulder muscles alone, combinations of shoulder and elbow muscles with various distal muscles and combinations of distal muscles (wrist + digit + intrinsic hand and wrist + digit). The least common synergies involved combinations of muscles at noncontiguous joints. This included six synergies that were not represented in M1 output.
This paper presents a detailed analysis of the magnitude, latency, and distribution of PStF and PStS from M1 cortex to 24 muscles of the forelimb in rhesus macaques. The results are significant in several respects. First, the data build on previous studies using various approaches demonstrating stronger effects from M1 cortex in more distal muscles. Our results show that the magnitude of M1 effects on muscle activity increases along the proximal-distal axis from shoulder to intrinsic hand muscles. Second, we show that the vast majority of M1 sites (73%) facilitate muscle synergies involving two or more forelimb joints and that an overwhelming number of these sites (90%) involve muscles at contiguous joints. Finally, the data are important in providing a comprehensive depiction of M1 output to forelimb muscles in the macaque monkey under normal conditions. This data can then serve as the basis for functional comparisons with other descending systems, other cortical motor areas, and M1 cortex after various experimental manipulations.
Latency and magnitude of poststimulus effects
Latencies of effects on proximal and distal muscles reflect a combination of conduction distance and synaptic linkage. Rather than showing a progressive increase in latency in moving from the most proximal to the most distal muscles, the onset latencies we have reported were actually lowest for wrist and forearm digit muscles. Shoulder and elbow muscles showed significantly greater latencies, and we speculate that this may be due to a less direct synaptic linkage, although we cannot rule out slower conduction times. Intrinsic hand muscles had the longest mean onset latency, and this can easily be accounted for based on the additional conduction distance. Onset latencies for PStF are very similar to those previously reported by Cheney and Fetz (1985) in a study of forearm wrist and digit muscles. In that study, the onset latency of PStF evoked from the sites of corticomotoneuronal cells was 8.3 ms for 15-μA stimuli. The magnitude of PStF increased continuously the more distal the muscle group. This preference toward activation of distal muscles was also evident in the number of PStF and PStS effects obtained, particularly after normalizing for the number of muscles recorded. The number of effects obtained was very similar for wrist, digit, and intrinsic hand muscles, but the numbers for elbow and shoulder muscles were clearly lower.
Issues related to mapping method
The spacing of electrode tracks was generally 1 mm. However, for tracks running down the bank of the precentral gyrus, data were collected at 0.5-mm intervals. To what extent might this disparity in mapping density have influenced the results? A mapping density of 0.5 mm was not used throughout because of the large number of electrode tracks that would have been required. Regardless of the sampling density, the data on which means for latency, magnitude and muscle fieldst were derived came from samples (sites) covering the entire forelimb representation, and there is every reason to believe that these samples constitute a valid and representative depiction of the entire area from which they were obtained. However, data reporting output characteristics in terms of the number of sites or number of PStF or PStS effects could have been biased by differences in sampling density at different locations in the map. To address this issue, stimulation sites were removed from some tracks to yield maps with a uniform sampling density of 1 mm. Figures 6–8 and 10 and Table 5 were then regenerated using this data. While this exercise produced small changes in some of the numbers in these graphs, the fundamental trends as described in the results were still clear and unchanged.
Another issue concerns the basis for the decision to use a 15-μA stimulus for mapping. Cheney and Fetz (1985) calculated the extent of physical current spread from different intensities of microstimulation. Current spread is given by the expression where r0 is the radius of the cortical volume containing directly activated cells, i is the stimulus current, and k is the proportionality constant. Various values for k have been proposed; however, Cheney and Fetz (1985) found that a minimal value of k (250 μA/mm2) yielded the best fit to their results. Using this value for k, a stimulus of 15 μA would produce current spread capable of exciting neurons over a radius of ∼250 μm. As long as the sites of stimulation were close to or in layer 5, spacing the stimulus sites at 0.5 mm would mean that nearly all corticospinal neurons would have been activated. However, spacing the stimuli at 1-mm intervals might have left as much as 75% of the layer 5 cortical output territory unactivated by the stimulus. Spacing electrode tracks at 1-mm intervals but stimulating at 0.5-mm intervals over the depth of the electrode penetration (which we did) might have left 60% of M1 output unactivated, taking into account that more of the representation was in the bank of the precentral gyrus and was sampled more densely. In either case, it is clear that the output effects reported here would only constitute a sample of the complete M1 output. However, Sawyer et al. (1979) found that PStF in specific muscles at 10 μA remained at or above half-maximum amplitude for a mean distance of 850 μm along the bank of the precentral gyrus. Also the profile of PStF across muscles remained very similar for distances of ≥1 mm. This suggests that our data based on 15-μA stimuli and a track spacing of 1 mm with stimulation at 0.5-mm intervals may have activated a much larger fraction of the complete M1 output than suggested by the effective stimulus distance calculation.
Cheney and Fetz (1985) showed that the profile of PStF in wrist and digit forearm muscles obtained with stimulus triggered averaging of EMG activity closely matched the profile of postspike facilitation obtained from the same sites except that PStF was much stronger than PSpF (e.g., 6.3 times at 5 μA). Based on these results, they postulated that corticospinal output neurons are organized in clusters or at least grouped such that neighboring neurons have the same target muscles, that is, the same muscle fields. Stimulus intensities of 5, 10, and, in some cases, 20 μA provided a good match between PStF to PSpF from the same site. Higher intensities generally produced increases in PStF magnitude with recruitment of new muscles, particularly at ≥20 μA. Therefore we felt that 15 μA was a good compromise between producing clear output effects but also ones that might closely match the muscle fields of individual cells at that site.
Distribution of PStF in proximal versus distal muscles
Findings presented in this paper show that both proximal and distal muscles are common targets of CM cells and that the effects are generally more prominent in distal muscles. The findings support the conclusion of McKiernan et al. (1998) based on spike-triggered averaging that CM cells make more frequent and potent terminations in motoneuron pools of distal compared with proximal muscles. Unlike StTA, SpTA makes it possible to study the synaptic output linkage of single cells in the M1. Therefore the findings of StTA, which systematically explored the entire forelimb area of M1, in conjunction with previous findings using SpTA, demonstrate that the CM cells target both proximal and distal muscles and the strength of effects increases the more distal the muscle.
Data presented in this paper also show that the vast majority of M1 sites (Table 5) facilitate muscle synergies involving two or more forelimb joints. Of these sites, 90% produced effects in muscles at contiguous joints. For example, there were more sites facilitating shoulder, elbow, wrist, and digit muscles in combination than those facilitating shoulder, elbow, and digit muscles (Table 5). Only 16 sites produced effects in muscles at noncontiguous joints. Moreover, seven synergy patterns involving muscles at noncontiguous joints were not represented at all in the cortical output of either monkey. This finding is more significant in view of the systematic approach to mapping we have applied. Sampling was extended in all directions until no effects were obtained. At these sites, repetitive ICMS was then used to evoke movements and establish that the site contained the representation of a body part other than the forelimb. Also worth noting is the fact that stimulus-triggered averaging of EMG activity during movement is a sensitive method capable of revealing subthreshold effects.
We suggest that sites producing cofacilitation of muscles at multiple joints may be particularly important for activation of synergies needed in executing coordinated multijoint arm and hand movements. The fact that the synergy pattern often involves both proximal and distal muscles is consistent with the observation that both proximal and distal muscles become broadly coactive during reaching movements (McKiernan et al. 1998). Activation of proximal muscles not only has the role of prime mover in bringing the hand to a target but also serves as a stabilizing factor for postural support of the hand and to counteract interaction torques.
Distribution of PStF in flexors versus extensors
Based on spike-triggered averaging data, Fetz and Cheney (1980) reported that wrist extensor muscles as a group were more strongly and frequently facilitated by CM cells than flexor muscles. McKiernan et al. (1998) also reported results from spike-triggered averaging consistent with this conclusion. This present StTA study also shows that wrist extensors were more commonly facilitated than flexors and flexors were more commonly inhibited. This general pattern was also true for forearm digit muscles but was reversed at the elbow where flexors were more commonly facilitated and suppressed than extensors.
Comparison of M1 and red nucleus output properties
Table 6 is a summary comparing the key features of M1 and magnocellular red nucleus (RNm) output based on StTA data collected under similar conditions (Belhaj-Saïf et al. 1998). To summarize the most noteworthy comparisons: RNm output is weaker than M1 to all muscle groups except the shoulder, overall RNm muscle fieldst size is smaller, the distribution of effects from RNm and M1 to different muscle groups is broadly similar for both PStF and PStS with the possible exception of the distal muscles and a greater emphasis on proximal-distal synergies, RNm output shows an overwhelming preference toward facilitation of extensor muscles whereas M1 facilitation is equally distributed to flexors and extensors, and both M1 and RNm show a bias toward suppression of flexor muscles, although this bias is stronger for RNm.
Cofacilitation of proximal and distal muscles was observed at 42% of M1 sites tested. In comparison, Belhaj-Saïf et al. (1998) reported that 61% of sites facilitated a combination of proximal and distal muscles. The most common synergies represented by M1 output involved muscles at contiguous joints. Common synergies included cofacilitation of both shoulder and elbow muscles, facilitation of only elbow muscles, facilitation of only shoulder muscles, facilitation of combinations of shoulder and elbow muscles with various distal muscles, and facilitation of various distal muscle combinations (wrist + digit + intrinsic hand and wrist + digit). Similar synergies were observed for RNm output. For example, the most common synergy involved a combination of shoulder, elbow, wrist, and digit muscles. RNm output showed a greater tendency than M1 for representation of muscles at noncontiguous joints. For instance, facilitation of shoulder, wrist, and digit muscles as a synergy was the fourth most common pattern in RNm output but was only found at one M1 site. Facilitation of shoulder and digit muscles was also a common RNm synergy but was not represented in M1 output.
Mechanism of proximal-distal muscle cofacilitation
The present StTA data demonstrate a clear pattern of cofacilitation of muscles at both proximal and distal joints with 42% of sites stimulated showing facilitation of both proximal and distal muscles. This supports the findings of McKiernan et al. (1998) showing that 45% of individual CM cells recorded produced PSpEs in both proximal and distal muscles.
Several different mechanisms can be proposed to explain cofacilitation of proximal and distal muscles in stimulus-triggered averages (Belhaj-Saïf et al. 1998). Perhaps the simplest explanation and the one we favor is the existence of individual cortical cells that make spinal terminations with motoneurons of both proximal and distal muscles. This mechanism is supported by the fact that a large number of individual cortical cells (45%) were found to produce cofacilitation of proximal and distal muscles in SpTAs of EMG activity (McKiernan et al. 1998). The sites producing cofacilitation of proximal and distal muscles form a zone in M1 separating a core distal representation and a peripheral band of proximal representation (Park et al. 2001). It should be noted that in previous work we have shown that the profile of PStF across forearm muscles (and the muscle fieldst) at ≤15 μA closely matches that of PSpF from individual neurons recorded at the site of stimulation so the results in this study are likely to reflect output patterns of individual corticospinal neurons (Cheney and Fetz 1985; Cheney et al. 1991).
One alternative possibility is that the cofacilitation may be due to the activation of two populations of neurons that individually facilitate either proximal or distal motoneurons but not both. These neurons may overlap in their spatial distribution in the cortex. However, it seems unlikely that this mechanism was a major contributor to the cofacilitation observed in our study because the sites producing cofacilitation form a distinct and continuous zone in M1 separating a core distal region from a peripheral horseshoe-shaped zone of proximal muscle representation (Park et al. 2001). The dimensions and boundaries of the proximal-distal cofacilitation zone cannot be accounted for based on simple stimulus spread to the distal and proximal zones. Also this mechanism does not easily explain the existence of large numbers of individual cortical cells that produce cofacilitation of proximal and distal muscles. Another possibility is the existence of excitatory collaterals between two populations of neurons the terminal connections of which are individually confined to motoneurons of either proximal or distal muscles. However, in addition to the arguments raised in the preceding text, the effects mediated through additional synapses would be expected to be weaker and have somewhat longer latency than the direct effects, which is not consistent with our results. A final possibility is activation of afferent input fibers supplying separate populations of CM cells that individually facilitate either proximal or distal motoneurons but not both. This possibility deserves particular attention given the fact that ICMS is likely to activate cortical output neurons indirectly through synaptic inputs (Jankowska et al. 1975b). However, explaining the existence of a large, distinct zone producing cofacilitation of proximal and distal muscles, as described by Park et al. (2001), is tenuous using this as the sole mechanism.
This work was supported by National Institutes of Health Grants NS-39023 and HD-02528.
We thank Dr. Matt Mayo for expert help with the statistical analysis.
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 © 2004 by the American Physiological Society