Download Primary motor cortex neurons classified in a postural task predict

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Human leg wikipedia , lookup

Undulatory locomotion wikipedia , lookup

Transcript
Articles in PresS. J Neurophysiol (February 3, 2016). doi:10.1152/jn.00971.2015
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Title: Primary motor cortex neurons classified in a postural task predict muscle activation patterns in a
reaching task
Running Head: M1 prediction of movement from postural classification
ETHAN A HEMING1, TIMOTHY P. LILLICRAP2, MOHSEN OMRANI1, TROY M HERTER3, J.
ANDREW PRUSZYNSKI4,5,6, STEPHEN H. SCOTT1,7,8
1. Centre for Neuroscience Studies, Queen’s University, Kingston, ON, Canada
2. Google DeepMind, London, Kings Cross, UK
3. Department of Exercise Science, University of South Carolina, Columbia, SC, USA
4. Department of Physiology and Pharmacology, Western University, London, ON, Canada
5. Robarts Research Institute, Western University, London, ON, Canada
6. Brain and Mind Institute, Western University, London, ON, Canada
7. Department of Biomedical and Molecular Sciences, Queen’s University, Kingston, ON, Canada
8. Department of Medicine, Queen’s University, Kingston, ON, Canada
17
18
Corresponding Authors:
19
Ethan Heming: [email protected]
20
Stephen Scott: [email protected]
21
22
1
Copyright © 2016 by the American Physiological Society.
23
24
Abstract:
25
Primary motor cortex (M1) activity correlates with many motor variables making it difficult to
26
demonstrate how it participates in motor control. We developed a two-stage process to separate the
27
process of classifying the motor field of M1 neurons from the process of predicting the spatiotemporal
28
patterns of its motor field during reaching. We tested our approach with a neural network model that
29
controlled a two-joint arm to show the statistical relationship between network connectivity and neural
30
activity across different motor tasks. In rhesus monkeys, M1 neurons classified by this method showed
31
preferred reaching directions similar to their associated muscle groups. Importantly, the neural
32
population signals predicted the spatiotemporal dynamics of their associated muscle groups, although a
33
sub-group of atypical neurons reversed their directional preference suggesting a selective role in
34
antagonist control. These results highlight that M1 provides important details on the spatiotemporal
35
patterns of muscle activity during motor skills such as reaching.
36
37
38
Keywords:
39
•
primary motor cortex
40
•
motor fields
41
•
neural network model
42
•
electromyography
43
2
44
INTRODUCTION
The motor system precisely controls the activity of muscles to generate smooth and accurate
45
46
motor actions. For example, goal-directed reaching involves agonist muscle activity to propel the hand
47
toward the goal and antagonistic muscle activity to decelerate and stop at the goal (Marsden et al., 1983;
48
Wierzbicka et al., 1986; Flanders et al., 1994). The selection, onset time and magnitude of muscle
49
activity during reaching depends on many factors such as target and initial limb position, arm geometry,
50
and external loads (Caminiti et al., 1990; Karst and Hasan, 1991; Hong et al., 1994; Scott, 1997).
Primary motor cortex (M1) plays an important role in voluntary motor functions, such as
51
52
reaching, but its specific role remains debated. The classic dichotomy is whether its activity reflects
53
muscles or movements (Phillips, 1975). The latter suggests that M1’s contribution is to define
54
behavioural goals (e.g. direction and extent of movements) leaving subcortical and spinal structures as
55
the primary source for generating spatiotemporal patterns of muscle activity (Georgopoulos, 1995;
56
Raphael et al., 2010). Indeed, many studies have shown that M1 activity correlates with many high level
57
parameters such as direction of hand motion (Georgopoulos et al., 1982; Toxopeus et al., 2011; Philip et
58
al., 2013), hand movement speed (Schwartz, 1993), and target direction and speed (Johnson et al.,
59
1999).
60
The other end of the spectrum is the view that primary motor cortex directly generates
61
spatiotemporal patterns of muscle activity for goal-directed movements (Bennett and Lemon, 1994;
62
Scott, 1997, 2003; Cherian et al., 2013). Although the exact patterns of muscle activity for limb
63
movement are only specified at the spinal level, as spinal afferent feedback will influence motor output,
64
the idea is that basic features such as the selection, timing and magnitude of muscle activity are specified
65
by neurons in M1. Indeed, several studies have quantified how M1 activity correlates with the activity of
66
hand and wrist muscles (Evarts, 1968; Humphrey, 1972; Bennett and Lemon, 1996; Kakei et al., 1999;
3
67
Oby et al., 2013) or proximal arm muscles (Murphy et al., 1985; Scott, 1997; Todorov, 2000; Sergio and
68
Kalaska, 2003; Cherian et al., 2013).
69
The lack of strong causal evidence for one of these options indicates that identifying a simple
70
correlation is not sufficient to identify the role of M1 in voluntary motor control. There are many
71
different patterns of muscle and neural activity, and finding arbitrary correlations is relatively easy
72
(Humphrey, 1972). One way to circumvent this problem is to dissociate different movement parameters,
73
such as by making movements with different torques or arm configurations (Fromm and Evarts, 1977;
74
Scott and Kalaska, 1997; Kakei et al., 1999; Sergio and Kalaska, 2003; Cherian et al., 2013). Inevitably,
75
these studies identify that some neuronal activity in M1 can reflect many different features of motor
76
actions. However, the presence of some high level information in M1 does not preclude its role in
77
specifying spatiotemporal patterns of muscle activity.
78
Better support for a role in generating patterns of muscle activity is to separate the process for
79
identifying the portion of the motor periphery associated with a given neuron – its motor field – from the
80
process of relating its spatiotemporal activity to muscles in that motor field. A similar concept is the
81
sensory field of a neuron, which is portion of the periphery to which applying stimuli elicits firing.
82
Identifying a sensory field is relatively straightforward as it requires one to simply apply or present a
83
controlled stimuli. It is less clear how to define motor fields.
84
The synaptic connectivity of neurons that project to proximal limb motor neuron pools can be
85
roughly identified by spike-triggered averaging (Fetz and Cheney, 1978; Cheney and Fetz, 1980, 1985;
86
Kasser and Cheney, 1985; McKiernan et al., 1998). However the variability in STA timing, and the
87
inability to record from all muscles at once, makes it very difficult to describe a complete connectivity
88
mapping from STA connectivity. If the scope of investigation is limited to monosynaptically connected
89
corticomotorneurons (CM cells), the proximal limb has reduced CM cell representation compared to the
4
90
distal limb (Buys et al., 1986; Palmer and Ashby, 1992) and CM cell distribution is limited to the caudal
91
(new) section of M1 (Rathelot and Strick, 2009). Intra-cortical micro-stimulation can be used to identify
92
muscle connectivity by repeated stimulation and stimulus-triggered-averaging (Cheney and Fetz, 1985;
93
Buys et al., 1986), however this technique does not guarantee single-neuron activation. Cheney and Fetz
94
(1985) and Lemon et al. (1987) show increased muscle activity for increased stimulation current and
95
conclude that stimulation likely activates multiple neurons in proximity.
96
The present study used two different behavioural tasks to separate the process for identifying a
97
neuron’s motor field from the process of comparing spatiotemporal patterns of activity between neurons
98
and muscles. First, the motor field of M1 neurons and limb muscles were identified based on their load
99
preference using a posture task consisting of combinations of shoulder and elbow flexion or extension
100
torques (Figure 1A). A neural network model that we used to drive our predictions (Lillicrap and Scott,
101
2013) highlighted that the torque preference of the model’s ‘cortical’ units correlated with the torque
102
preference of their synaptically associated muscle groups. The model predicted that neurons with similar
103
motor fields – defined by the posture task – would show correlated directional preferences in a reaching
104
task. We tested this prediction by examining the directional preferences of monkey M1 neurons during
105
center-out reaching to a range of peripheral targets (Figure 1B). Consistent with our model, we found
106
that the activity of each neuronal population predicted the spatiotemporal patterns of their respective
107
muscle groups. Interestingly, a small subset of the neurons in each group possessed directional
108
preferences during reaching that were opposite to that observed for its associated muscle group. These
109
atypical neurons may be indicative of selective control when the muscle acts as an antagonist during
110
motor skills.
111
112
113
Experimental Procedures
5
114
115
Subjects and apparatus
Five male rhesus monkeys (Macaca mulatta, 6–12 kg, monkeys A-E) were trained to perform
116
reaching and posture tasks while attached to a robotic upper-limb exoskeleton (KINARM; BKIN
117
Technologies, Kingston, Ontario, Canada). This exoskeleton maintained the arm in a horizontal plane at
118
shoulder height permitting motion at the shoulder and elbow, and allowing mechanical torques to be
119
applied at either joint (Scott, 1999). Targets and a dot representing hand feedback location were
120
presented in the plane of the arm via reflection in semi-transparent glass. These experiments were
121
conducted in accordance with the Queen's University Animal Care Committee.
122
123
124
Behavioural Tasks
Posture task. This task has been previously described (Cabel et al., 2001; Herter et al., 2007). In
125
each trial, a constant torque was applied to the elbow and/or shoulder. The monkey then stabilized the
126
hand within a central 0.8cm-wide stationary target for at least 3 seconds. Nine constant torques were
127
used, consisting of elbow flexion (EF) or extension (EE), shoulder flexion (SF) or extension (SE), four
128
multi-joint torques (SF+EF, SF+EE, SE+EF, SE+EE), and an unloaded condition (Figure 1A). Torques
129
of magnitude 0.12Nm were used for monkeys A-C,E and 0.32 Nm for Monkey D. Five blocks were
130
presented, each containing the 9 load conditions in random order, for a total of 45 trials.
131
Reaching task. This task has been previously described (Kurtzer et al., 2006b). Monkeys began
132
each trial by maintaining their index finger (white dot) within a central start target (8mm radius). This
133
start target was positioned such that the shoulder and elbow were at approximately 30⁰ and 90⁰,
134
respectively. After a random time period (1.5-2.0s), a peripheral target (12mm radius) then illuminated
135
6cm from the central target. The monkey then moved between the start and peripheral target in 220 to
136
350ms, generating total reach times of ~500 to 600ms when including intra-target acceleration and
6
137
deceleration. Eight such peripheral targets were located around the start target. For monkeys A and B,
138
these were distributed such that they were roughly distributed uniformly in torque space in two
139
arrangements (Figure 1B i,ii). For Monkeys C-E, the targets were uniformly distributed in Cartesian
140
space in two arrangements (Figure 1B iii,iv). In monkey D, some trials were also performed with 3cm
141
reaches (target pattern not shown). Five blocks were presented, each containing the 8 reach directions in
142
random order, for a total of 40 trials.
143
144
145
Data collection
Neuronal recordings were obtained from the elbow/shoulder region of primary motor cortex
146
(M1), contralateral to the arm used to perform the behavioural tasks, using standard recording
147
techniques (Scott and Kalaska, 1997; Scott et al., 2001). Activity of these neurons was recorded during
148
both the reaching and posture tasks.
149
We examined the activity of elbow and shoulder flexor and extensor muscles using
150
electromyography during the posture and reaching tasks. In most cases, we recorded through a pair of
151
percutaneous Teflon-coated 50-μm stainless steel wires using standard recording techniques (Scott and
152
Kalaska, 1997). In monkeys A and C, we recorded some muscle activity from chronically implanted
153
bipolar multi-strand electrodes (Scott and Kalaska, 1997; Kurtzer et al., 2006b). Chronic electrode
154
recordings were only selected if they occurred more than a week apart, to minimize redundant data.
155
Recordings were taken from biceps (12 percutaneous, 13 chronic), brachioradialis (7, 11), brachialis (6,
156
6), long head triceps (5, 19), posterior deltoid (8, 10), lateral triceps (8, 3), middle triceps (2, 0), anterior
157
deltoid (3, 11), and pectoralis major (6, 6). To ensure that the percutaneous recordings were not being
158
skewed by the chronic recordings, we compared the mean preferred torque direction (PTD, described in
159
Data analysis) between percutaneous and chronic recordings across muscles, and found no difference
7
160
(paired t-test, t=0.02, p=0.98), and while the average variance in PTD was lower for chronic recordings
161
(3.4º) than percutaneous (7.0º).
162
Joint kinematic data and EMG were recorded at 1 kHz (Monkeys A–C) or 4 kHz (Monkeys D,E).
163
Neuronal firing was binned into 5ms bins, and EMG and kinematic data were down-sampled to 200Hz
164
for data storage.
165
166
167
Data analysis
Preferred torque direction. Combinations of shoulder and elbow torques were described in
168
torque space where shoulder torque was represented along the x-axis and elbow torque was represented
169
along the y-axis. Positive torque, in each axis, was defined as flexor torque to oppose joint-extending
170
applied torques, so that shoulder flexor torque was at 0⁰, elbow extensor at 90⁰, shoulder extensor at
171
180⁰, and elbow extensor at 270⁰. A plane was fit to EMG activity or cell firing rate associated with the
172
elbow/shoulder torque in this space.. We used only the EMG activity or cell firing in the last two
173
seconds of hold time to ensure that recordings were from a period of stationary posture. If the plane had
174
a statistically significant slope, the angle of maximal slope was defined as the torque which elicited
175
either a muscle’s maximal activation or neuron’s maximal firing rate. This preferred torque direction
176
(PTD) was measured counter-clockwise from shoulder flexion.
177
Preferred reaching direction. A given reach movement was assigned a reach direction in
178
Cartesian space, increasing counter-clockwise from the positive x-axis. It was calculated by the angle
179
from the origin to the hand position at maximal tangential velocity. EMG activity and neuronal firing
180
rate were integrated around (-50ms to 150ms) movement onset, which was defined as the time when the
181
hand first attained 5% of maximum hand speed. This activity was fit to a plane based on the angle of
182
each reach and the activity during each reach. If the plane had significant slope, the preferred reaching
8
183
direction (PRD) was defined as the angle which had maximal slope on the plane. It described the
184
maximum spatial modulation of either a muscle’s activation or neuron’s firing rate.
185
Spatiotemporal dynamics of reaching. Activities in each trial were aligned temporally on the
186
calculated reach onset time. Baseline activity was removed by calculating the mean activity of each
187
neuron or muscle recording during the initial hold period over all trials and subtracted from all trials for
188
that recording. Activity for each muscle or neuron was then divided by the maximum activity across
189
reaching directions for that recording. Data were smoothed using a Gaussian kernel with a 5ms standard
190
deviation.
191
192
193
Modelling
To predict the responses of a neural system with defined motor fields, we implemented a static
194
neural network model similar to that developed by Lillicrap & Scott (2013). . The model consisted of a
195
vector of ‘cortical’ units, z, which controlled 6 muscle groups, u, controlling a two-joint arm in the
196
horizontal plane via linear output weights, w (Figure 2A). Muscle activity was kept positive via the
197
standard sigmoid - that is, u = σ(w · z). Modelled muscle activity generated torques and hand velocities
198
via a function that approximated the biomechanics of a 2-joint revolute arm constrained to move in the
199
horizontal plane. This function was computed by linearizing a dynamic model that included limb
200
geometry, intersegmental dynamics, and mono- and bi-articular muscles with force generation
201
dependent on length and velocity. Muscle tension forces, t, are obtained by element-wise multiplication
202
of muscle activity with linearized F-L/V scaling factors appropriate for the movement direction, i.e., t =
203
H · u. Joint torques are computed via: τ = M t. And hand velocity is determined by the linear
204
transformation, y = G F τ, where F and G are local linear approximations to limb dynamics and the
205
geometric mapping between joint and hand velocity, respectively. This static model was derived as a
9
206
simplified version of a dynamic model which executed reaching movements over a sequence of time
207
steps and in which the network model was connected in closed loop with the arm. One of the findings
208
of this previous work was that a static model based on a linearization of the dynamic version captured
209
the most salient features of the population neural activity (Lillicrap & Scott 2013). The static version
210
also has the benefit of being easier to optimize, analyze, and understand. Parameters for the limb
211
biomechanics were derived from published work on monkey limb and muscle characteristics (Cheng
212
and Scott, 2000; Singh et al., 2002; Graham and Scott, 2003).
We optimized z to solve analogues of the posture and reach tasks while keeping the square of the
213
214
neural and muscle activities small. For the reaching task, the model captures movement initiation. In
215
this case z* is found by minimizing the difference between target and actual hand velocity for a given
216
movement,
217
−
=
− , while keeping unit and muscle activity small; that is,
∗
=
=
+ ‖ ‖ + ‖ ‖. For the posture task, the model captures the steady state condition
218
during which the joint torques are countered and the arm is at zero velocity. In this case z* is found by
219
minimizing the difference between applied and actual joint torques,
220
and muscle activity small; that is,
221
and β are set to 1e-6. Importantly, for a given simulation, the elements of the matrix w were drawn
222
randomly from a normal distribution (σ = 0.05) and were unaltered during optimization. This meant that
223
a given unit maintained the same relative contribution to each muscle at the periphery across tasks. Unit
224
activity was optimized to generate 16 target torques (posture task) and 16 target velocities (reach task) -
225
both equally distributed about the unit circle. In practice, we used a preconditioned conjugate gradient
226
descent algorithm with back-tracking line searches to find an optimal vector of activity for a given trial.
227
In line with the analysis of biological data, the activity of z and u across all 16 torques and reach
228
directions were plane fit to determine preferred reach and torque directions.
∗
=
=
10
−
=
− , while keeping unit
+ ‖ ‖ + ‖ ‖. In both cases, α
229
230
231
232
233
RESULTS
234
235
236
Neural network model: Relation of torque preference and motor field
We used a static neural network model similar to that developed by Lillicrap & Scott (2013) to
237
examine the relationship between neural connectivity, torque preferences during posture, and directional
238
tuning during reaching. The activation of model cortical units, z, were optimized to generate
239
combinations of elbow and shoulder torque for the posture task and different hand velocities for the
240
reaching task. Though we used a PCG algorithm for the results reported here, the same essential results
241
can be obtained with virtually any gradient based optimization routine. In particular, we find the same
242
results using L-BFGS and stochastic gradient descent (SGD), though SGD takes significantly longer to
243
converge. Given that the optimization we perform is non-linear and high dimensional, we are not able
244
to find a global minimum. Our results are thus based on local minima - but they are robust minima in the
245
following sense: we repeated the simulation 10 times from random initializations of the synaptic weight
246
matrix and found the same characteristic pattern of PTD/PRD distributions in each case. Thus, there
247
appears to be a large family of such minima - all of which produce similar behavioral performance and
248
PTD/PRD distributions.
249
We compared the torque preference of units in the network to its connectivity to evaluate how
250
well its torque preference estimated its motor field. To quantify torque preference, we fit a plane to the
251
activation of each unit across torque combinations in the two-dimensional elbow/shoulder torque space.
11
252
The preferred torque direction (PTD) of a unit was the angle in torque space to which the unit was
253
maximally active. Across simulations, we found that 99.4% of units had a significant (p<0.01) PTD in
254
the posture task. The distribution of the significant PTDs for one training session can be seen in Figure
255
3A. The resulting distribution was bimodal (r=0.31, p<0.001), aligned in much the same manner as the
256
previously reported distribution of M1 neurons (Herter et al., 2007; Pruszynski et al., 2014), with the
257
majority of units related to whole-limb flexion or extension.
258
In such a straightforward model, one might assume that a given unit would always show an
259
identical relationship between its torque preference and its anatomical connectivity given that the unit
260
can only produce torque in a given direction when activated alone. We calculated the motor field
261
preferred torque (MFPTD) direction of a given unit, zi, by multiplying its output weights (motor field),
262
wij, with the vector of preferred torque direction of each output unit in the posture task, uj, and then
263
vector summating. It is important to note that the preferred torque direction of each output unit is not the
264
simple direction of force production for that unit. Due to the redundant force profiles of the biarticular
265
muscles and the dynamics of the limb the preferred direction of an output unit rotates away from its
266
simple force production direction (Lillicrap and Scott, 2013). We found a circular correlation (r=0.86,
267
p<0.001) between the unit’s MFPTD and its preferred torque direction (PTD) in the posture task (Figure
268
2B), however the relationship was not perfect. To see how different connectivity patterns might be
269
influencing this relationship, we looked at the degree to which a unit co-activated opposite muscle
270
groups together (co-contraction). We normalized each set of weights from a unit to its muscle groups,
271
multiplied these by their respective MFPTD vectors, and vector averaged them. The length of the
272
resulting vector for each unit was used as an indicator of co-contraction, with a short vector indicating
273
more co-contraction. The average difference between PTD and MFPTD was 22⁰. Units with low co-
274
contraction (MFPTD vector length > 0.4, 9.1% of units) had an average difference of 7⁰ (r=0.99),
12
275
whereas units with high co-contraction (MFPTD vector length < 0.1, 11.7% of units) had an average
276
difference of 62⁰ (r=0.40). This suggests that the statistical dispersion in the relationship between torque
277
preference and anatomical connectivity is caused by those units with stronger synaptic connections to
278
antagonist muscles.
279
280
281
Neural Network Model: Relation of torque and reaching direction
We used our model to predict the relationship between preferred torque direction and reaching
282
activity for units in the neural network model. Reaching activity was described by a preferred reaching
283
direction (PRD) – the direction in Cartesian space toward which a reach movement would elicit
284
maximal unit activity. Across all simulations, we found that 98.7% of units had a significant PRD (plane
285
fit p<0.01). The distribution of significant PRDs for one training session of units can be seen in Figure
286
3B. The bimodality (r=0.52, p<0.001) towards the top-left and bottom-right that emerges is stable across
287
simulations. The distribution was consistent for static and dynamic networks (Lillicrap and Scott, 2013)
288
and previous observations on M1 neurons (Scott et al., 2001).
289
Novel in this study, we compared PTD and PRD for each model unit which had both a
290
significant PTD in the posture task and a significant PRD in the reaching task (98.1%). As can be seen
291
in Figure 4A, there is a consistent relationship between the PTDs and PRDs, where two clusters emerge
292
due to the interaction of the bimodalities noted in each distribution. Figure 4B displays a histogram of
293
the difference of PTD and PRD angles for all units, with the presence of a mean systematic shift of 152⁰
294
(circular correlation, r=0.84, p<0.001). This angle roughly corresponds to the shift in coordinate frames
295
between torque and hand space, used to define unit responses in the posture and reaching tasks,
296
respectively. This rotational shift was stable across all ten simulation sets (150⁰-154⁰, circular SD=1.2⁰),
297
indicating that the result is robust. There was some dispersion in the relationship between a unit’s PTD
13
298
and PRD (circular SD=27⁰). Even though individual units had different PRD and PTD tuning across
299
simulations, both the population distribution, and the relation between a given unit’s PTD and PRD
300
remained statistically the same.
301
302
303
Non-Human Primate Recordings: Muscle recordings
We recorded and analysed the EMG activity of 9 muscles spanning the shoulder and/or elbow
304
(151 suitable recordings) in 5 macaque monkeys in posture and reaching tasks. In the posture task, a
305
monkey’s arm was maintained in the horizontal plane while combinations of flexor and/or extensor step
306
torques were applied to the shoulder and/or elbow (Figure 1A). Recordings were made while the
307
monkey held their hand stationary at a central position and countered these torques. Figure 3C shows an
308
example of pectoralis major activity during the posture task. As previously reported (Kurtzer et al.,
309
2006a), this shoulder flexor, modulated its activity for torques applied at the elbow in addition to loads
310
at the shoulder and, as a result, the preferred torque direction (PTD) for this muscle did not lie at
311
shoulder flexion (i.e. 0⁰) line, as one would expect if the muscle activity reflected its anatomical action.
312
Shifts across all muscles (Figure 3E) showed stereotypical clustering of preferred torque directions - one
313
in the elbow-extension/shoulder-flexion quadrant and the other in the elbow-flexion/shoulder-extension
314
quadrant (Kurtzer et al., 2006a).
315
In the reaching task, NHPs made fast center-out reaches to peripheral targets arranged around the
316
starting position (Figure 1B). Figure 3D shows an example of pectoralis major muscle activity during
317
reaching. Maximal activity in the example was observed towards the top left target reflecting the need
318
for a large shoulder flexor torque to accelerate the arm towards that target (Graham et al., 2003). In the
319
bottom right direction, an activity peak can be seen about 200 ms later related to an antagonist burst to
320
decelerate the arm. EMG activity was integrated over a 200 ms period starting 150 ms before movement
14
321
onset (grey bar), to capture the activity associated with initial acceleration. Across all muscles, PRDs
322
(128 significant muscle samples) were skewed towards one of two directions (Figure 3F), one centred
323
around 110⁰ and one around 280⁰ (Kurtzer et al., 2006a).
324
As with the novel model analysis, we compared PTD and PRD for each muscle which had both a
325
significant PTD in the posture task and a significant PRD in the reaching task (N=122). A strong
326
consistent relationship between the PTDs and PRDs can be seen in Figure 4C where two clusters
327
emerge: one (bottom left cluster) that includes the shoulder extensors, the elbow flexors and all bi-
328
articulars and the other (top right) that includes the elbow extensors and shoulder flexors. Figure 4D
329
displays a histogram of the difference of PTD and PRD angles for all muscles. There is a systematic
330
shift of 152⁰ (circular r=0.87, p<0.001), very similar to that obtained for the network model.
331
332
333
Monkey M1 recordings
We recorded and analysed the activity of 540 M1 neurons in 5 monkeys in the posture and
334
reaching tasks, with the same epochs as for muscle activity. Of these, 373 showed a significant PTD in
335
the posture task (example neuron in Figure 3G) and 424 M1 neurons that showed a significant PRD
336
during the reaching task (example neuron in Figure 3H). As seen in Figure 3I and 3J, the distributions of
337
PTDs and PRDs for all M1 neurons showed stereotypical bimodalities as have been previously reported
338
(Scott et al., 2001; Kurtzer et al., 2006a).
339
As with muscles, we compared PTD and PRD for each neuron which had both a significant PTD
340
and PRD (N=314). Figure 4E shows two distinct clusters when comparing the PTD and PRD of M1
341
neurons: one cluster with a PRD at 301⁰ and a PTD at 140⁰ (bottom left cluster), and another cluster
342
with a PRD at 123⁰ and a PTD at 325⁰. The relation between the two preferred directions of M1 neurons
343
is clearly shown in Figure 4F as a 151⁰ shift (circular r=0.43, p<0.001) between PRD and PTD. Cluster
15
344
locations were maxima of cell preferred directions convolved with a 2D windowed Gaussian distribution
345
(σ = 10⁰). This shift and clustering is very similar to both the units of the network model, and the muscle
346
recordings. Unlike the muscles and network model, a notable portion of M1 neurons show PTD-PRD
347
differences not located near the diagonal of the plot. The result is a relatively low PTD-PRD correlation
348
(r=0.43 cells, 0.84 model, 0.87 muscles) and large variance in their shift values (circular SD=61⁰ cells,
349
27⁰ model, 24⁰ muscles). We estimated whether these variances were significantly different by sampled
350
the PTD-PRD difference values with replacement to create distributions of variances. The distribution of
351
PTD-PRD variance for neurons had no overlap with either the muscle or model distributions. The
352
muscle and model distributions of variance shared a 30% overlap.
353
354
355
Prediction of patterns of muscle activity from M1 population signals
Our first step to predict muscle activity during reaching was to classify the M1 neurons based on
356
their similarity to muscle groups during a postural load task. Figure 5 highlights the distribution of PTDs
357
of muscle for shoulder and elbow muscles. We defined four ranges in torque space representing flexors
358
and extensors at each joint: elbow flexors (90⁰ to 135⁰), shoulder extensors (135⁰ to 180⁰), elbow
359
extensors (270⁰ to 315⁰), and shoulder flexors (315⁰ to 360⁰). The four torque groups captured 75% of
360
motor cortical neurons: 55 neurons were classified as elbow flexor neurons, 71 as shoulder extensor
361
neurons, 54 as elbow extensor neurons, and 68 shoulder flexor neurons. When applied to our network
362
model, these groupings captured 60.8% of units, with 14.5% to 15.6% of units in each group.
363
Figure 6 displays the distribution of PRDs for each muscle group, the associated M1 neurons,
364
and associated network units. Note that these distributions represent vertical slices, 45⁰ thick, of the
365
PRD-PTD relationship shown in Figure 4E. The preferred reaching direction distributions for each
366
group of muscles were unimodal, with shoulder flexors preferring 134⁰ (r=0.85, p<0.001), elbow flexors
16
367
preferring 261⁰ (r=0.83, p<0.001), shoulder extensors preferring 317⁰ (r=0.99, p<0.001), and elbow
368
extensors preferring 98⁰ (r=0.63, p=0.03). Dispersion was relatively small for each muscle group
369
illustrating that muscle within a group had similar reaching preferences.
370
Each group of M1 neurons showed a significant unimodal distribution of PRDs, nearly identical
371
to that of their comparative muscle group (Figure 6). Shoulder flexor neurons showed a preferred
372
reaching direction of 128⁰ (r=0.66, p<0.001), elbow flexor neurons showed a preferred reaching
373
direction of 293⁰ (r=0.32, p<0.001), shoulder extensor neurons showed a preferred reaching direction of
374
313⁰ (r=0.48, p<0.001), and elbow extensor neurons showed a preferred reaching direction of 95⁰
375
(r=0.35, p=0.002). M1 neurons associated with each muscle group showed a much greater range of
376
preferred reach directions. Watson-Williams tests between muscles and cells showed no difference in
377
means, except a small (33.8 degrees), but significant difference for elbow flexors (SF: F1,72=0.02
378
p=0.65, EF: F1,102=5.12 p=0.03, SE: F1,101=0.13 p=0.71, SF: F1,55=0.01 p=0.92).
379
The torque groupings were also used to classify neural network units in a similar fashion (Figure
380
6). Shoulder flexor units showed a preferred reaching direction of 119⁰ (r=0.92, p<0.001), elbow flexor
381
units showed a preferred reaching direction of 269⁰ (r=0.88, p<0.001), shoulder extensor units showed a
382
preferred reaching direction of 298⁰ (r=0.91, p<0.001), and elbow extensor units showed a preferred
383
reaching direction of 89⁰ (r=0.89, p<0.001). As the neural network units followed the same pattern as
384
both muscles and M1 neurons in distribution, it was not surprising that each torque-based group also had
385
close overlap with the PRD of both M1 neurons and muscles.
386
387
388
389
Muscle-like spatiotemporal dynamics
While predicting the agonist burst activity of the spatiotemporal activity was a first step, we
further examined whether M1 neurons could predict the full temporal pattern of shoulder and elbow
17
390
muscle activity across directions during reaching. We evaluated the population activity of each group of
391
M1 neurons over time and compared this to the activity of each group of muscles.
392
Illustrations of normalized population activity over time and across reach direction for each
393
group of M1 neurons can be seen in Figure 7A. Clear maximums of activity emerge prior to reach onset
394
in a primary direction for each group. Given that PRD was determined by the epoch of time around
395
reach onset, it is not a surprise that these maximums align with the PRD of each torque-classified group.
396
Qualitatively, the overall patterns of activity are similar between M1 and their corresponding muscle
397
groups. Given that each group seemed to have a similar profile of activity across time but for different
398
reach directions, we computed 2-dimensional correlations between each M1 group and all muscle
399
groups (Figure 7B) and found high degrees of correlation (shoulder flexor r=0.82, elbow flexor r=0.74,
400
shoulder extensor r=0.74, elbow extensor r=0.75). Each muscle group’s spatiotemporal activity was
401
predicted best by its respective M1 neuronal group, with latencies of maximal correlation that show M1
402
activity precedes muscle activity (shoulder flexor -30ms, elbow flexor -70ms, shoulder extensor -30ms,
403
elbow extensor -65ms). Figure 7C displays the aggregate of all four groups highlighting how specific
404
shifts in the timing and magnitude of agonist muscle activity before movement is preceded (average
405
latency -35ms) by similar spatiotemporal patterns in M1 (Scott, 1997).
406
We also investigated the spatiotemporal output of the subset of M1 neurons that appeared to have
407
an atypical, opposing PRD-PTD relationship. We decomposed the aggregate M1 activity by selecting
408
neurons that were within 30⁰ of the expected relationship (around the 151⁰ PRD-PTD), and those that
409
were opposite to these. As can be seen in Figure 8A and 8B respectively, both sets of neurons seemed to
410
have bursts of activity; however, the atypical M1 neurons had later temporal characteristics for their
411
burst (Figure 8C), which occurred in the opposite reaching direction. A cross-correlation of the
412
maximum aggregate activity showed that the atypical activity burst occurred 50ms later than the typical
18
413
agonist burst. Interestingly, the atypical neurons’ agonist-direction activity (solid red line) appeared to
414
be larger with a longer duration than the regular population’s antagonist-direction activity (dashed black
415
line), lasting well after the reaching movement was completed. Thus across the population, the preferred
416
direction of these atypical cells was aligned with its motor field after movement when maintaining the
417
hand at the peripheral target.
418
19
419
420
421
DISCUSSION
Several studies have shown correlations between M1 activity and EMG in monkeys (Evarts,
422
1968; Humphrey, 1972; Fetz et al., 1989; Bennett and Lemon, 1996) and cats (Drew, 1993; Drew et al.,
423
2002). However, given the diversity of EMG activity across muscles during a motor action, one can
424
likely find that a neuron’s activity will correlate with at least one of them. Here we show these
425
correlations remain when classification and prediction are separated. We classified M1 neurons into one
426
of four distinct motor fields (elbow flexor, elbow extensor, shoulder flexor, and shoulder extensor) using
427
a static load task. We then generated a population signal from each group and predicted the
428
spatiotemporal activity of the corresponding muscle during a dynamic reaching task. The preferred
429
direction of most neurons matched the directional preference of muscles with the same motor field.
430
Interestingly, a small proportion of neurons had preferences in the opposite direction.
431
A receptive field of a sensory neuron can be defined as the part of the body to which a stimulus
432
elicits activity by this neuron. These can be easily defined by repeated application of stimuli and finding
433
the parts of the body to which firing of a neuron correlates. It is implicitly assumed from this that a
434
neuron that fires in relation to a stimulus is caused by that stimulus. A related concept is a motor field,
435
which can be defined as the portion of the body associated with a given motor-related neuron. While the
436
target innervation, or muscle field, of some neurons, such as corticomotoneurons, can be identified by
437
spike-triggered averaging of EMG (Cheney and Fetz, 1980, 1985), this does not necessarily define the
438
portions of the body to which the neuron fires. Even when limiting to CM cells which have
439
monosynaptic connections to motoneuron pools, Griffin et al. (2015) showed how directional tuning, or
440
motor field, of some CM cells does not necessarily align with the muscle field, or connectivity. As well,
441
CM cells represent only a small fraction of corticospinal neurons that predominantly target the distal
20
442
musculature in primates (Buys et al., 1986; Palmer and Ashby, 1992), and are localized mainly in the
443
caudal portion of M1 or “new M1” (Rathelot and Strick, 2009). More directly, a motor field can be
444
defined based on whether a neuron is active when the animal performs a motor action with a given body
445
part, as we have done in the present study. However, unlike sensory fields, it is not necessarily true that
446
there is a causal link between the neuron’s activity (motor field) and motor output of that body part.
447
We examined this relationship using a relatively complex artificial neural network with 1000
448
neural units to observe the relationship between a unit’s connectivity and activity relative to the motor
449
periphery. The model displayed a strong similarity between the unit’s torque preference (PTD) and the
450
activity preference of its motor field (MFPTD) during the postural load task. However, the directional
451
preferences were not identical as there was a mean difference of 22 degrees. Thus, in general, the
452
activity of the unit during a task was a good indicator of its synaptic connectivity to the motor periphery.
453
This was more accurate for the units that had strong unidirectional connectivity without strong synaptic
454
connections to antagonist muscles.
455
The presence of PTD-MFPTD variability in the artificial neural network parallels previous
456
experimental work demonstrating a statistical relationship between the muscle field of CM cells (defined
457
by STA, Fetz and Cheney, 1978) and the associated patterns of activity of forearm and hand muscles
458
during a precision grip task (Bennett and Lemon, 1994). In that study, the observed correlations were
459
relatively modest, the highest of which was 0.69. Some of this variability may be explained by task-
460
dependant variation in the connectivity using STA (Buys et al., 1986). However, our network model
461
showed that a more complex pattern of synaptic connectivity may contribute to a lower overall
462
correlation value. Ultimately, our model suggests that a perfect correlation will not exist between muscle
463
fields and motor fields as the neuronal activity reflects not only the activity of target muscles, but also
464
the activity of the rest of the network (Lillicrap and Scott, 2013).
21
465
Our model also demonstrated that the preferred pattern of activity relative to the motor periphery
466
remains relatively constant across motor behaviours. In other words, if a unit was maximally active
467
when the elbow flexors were maximally active for postural loads, it will tend to be active when the
468
elbow flexors are maximally active during reaching. A fixed motor field is consistent with the
469
observation that neural responses can be predicted when loads are combined at the shoulder and elbow
470
(Gribble and Scott, 2002) and are similar for transient and sustained loads (Herter et al., 2009;
471
Pruszynski et al., 2014). It is also consistent with several studies that highlight how the reaching
472
directional preference of neurons and muscles both tend to rotate similarly with applied curl fields
473
(Cherian et al., 2013), changes in arm geometry (Scott and Kalaska, 1997) or start position (Caminiti et
474
al., 1991; Sergio and Kalaska, 1997). Again, there is some dispersion in the directional preference across
475
tasks or conditions, likely reflecting how unit activity is influenced by factors beyond its motor field.
476
Similar methodology to the present study was used by Sergio et al. (2005) to examine M1 activity
477
between isometric force production and a dynamic reaching paradigm. Their results show similar
478
patterns of correlation between isometric force and reaching preferred direction which support the
479
notion of M1 motor fields that persist across tasks. Here we show this specific correlation between the
480
activity of M1 neurons and its motor field.
481
It is important to recognize that a fixed motor field does not mean that a neuron is always active
482
when its associated muscles are active during voluntary motor actions. It has been shown that neurons
483
strongly active during precision grip were much less active for power grip that required more muscle
484
force (Muir and Lemon, 1983). As well, load representations can change between posture and
485
movement, but their directional preference remains relatively constant (Kurtzer et al., 2005) and can be
486
only active when the muscle is used for a specific phase of movement (Griffin et al., 2015). Even when
487
neural activity reflects features such as the direction of the spatial goal, these neurons likely remain
22
488
associated with a specific body part such as the wrist (Kakei et al., 1999) or elbow/shoulder (Sergio and
489
Kalaska, 2003).
490
While the essential PRD and PTD patterns are quite similar between network model and real
491
neurons, the variability in the difference between the PRDs and PTDs is notably greater in the neurons
492
than in the model. There are many possible sources for this difference in observed variability. First,
493
there is noise in the empirical estimates of the PRDs and PTDs, whereas in the model these quantities
494
can be measured perfectly. As well, the only source of noise in the model, given a fixed and
495
deterministic optimization procedure, is the random initialization of the weights, w. Second, our model
496
assumes that all of the model cortical units directly contributed to driving muscle activity by way of
497
monosynaptic connections. In reality, only a subset of M1 neurons project to the spinal cord, and most
498
of these synapse predominantly on spinal interneurons and can be influenced by spinal processing. As
499
well, we may expect that neurons that do not project to the spinal cord – which may connect to other
500
brain areas (Turner and DeLong, 2000), or recurrently within M1 – to show activity dissociated from the
501
muscles. As well, motor cortical neurons reflect aspects of motor function beyond patterns of motor
502
output, such as correlates of movement kinematics (Kalaska et al., 1989) and behavioural goals (Kakei
503
et al., 1999). Neural activity is also context (Hepp-Reymond et al., 1999), phase (Griffin et al., 2015)
504
and task-dependent (Buys et al., 1986; Kurtzer et al., 2005). These complexities will necessarily
505
influence the relationship between motor cortical activity across postural and movement tasks and likely
506
increase variability in the difference between their PRD and PTD.
507
At the same time, given the complexity of neural processing in M1 it is surprising its activity can
508
predict the patterns of muscle activity. The patterns are not exact however particularly related to activity
509
patterns when the muscles are antagonists. The general similarity between M1 and muscle activity
510
suggests that spinal processing is not dramatically altering descending commands (Lillicrap and Scott,
23
511
2013). While adding an additional network layer to our model to simulate the influence of a spinal cord
512
or other systems may have given us a higher network variability, adding complexity was not the aim of
513
this study. We were not interested in exactly modeling the descending stream from primary motor
514
cortex, but rather to producing testable hypotheses that stem from a system with an unchanging
515
relationship with the periphery.
516
Our analysis focussed on exploring the relationship between M1 neurons and flexor and extensor
517
muscle groups at the shoulder and elbow. However, this does not mean that M1 neurons are associated
518
with all muscles within a given group, nor that they are only associated with a single muscle group.
519
Most certainly M1 neurons synapse onto many motor neurons (directly or indirectly) including those in
520
different muscle groups and muscles that span multiple joints as demonstrated with STA (Cheney and
521
Fetz, 1985; McKiernan et al., 1998; Smith and Fetz, 2009). We find several neurons with load
522
preferences related to combined flexor or extensor loads at the shoulder and elbow, a pattern of activity
523
not observed for any muscles (Kurtzer et al., 2006b). These neurons likely have muscle fields that span
524
both joints. Even those neurons with load preferences associated with a single muscle group, likely have
525
a motor field beyond that group.
526
The idea of a motor field is closely related to the idea of muscle synergies. While the basic
527
definition of a muscle synergy is simply the cooperative action of two or more muscles, some theorize
528
that the motor system simplifies control by only varying a small number of muscle synergies (d’Avella
529
et al., 2006). One prediction from this theory is that there would be a fixed number of motor fields
530
represented in a region like M1. In our postural load task, these synergies should appear as clusters of
531
preferred loads in our M1 sample. In fact, the distribution of load preferences do appear to be clustered
532
into two groups one related to whole-limb flexion and one related to whole-limb extension (Kurtzer et
533
al., 2006a; Herter et al., 2009). However, this bias appears to be due to the presence of bi-articular
24
534
muscles (Lillicrap and Scott, 2013) – in a model with no bi-articular muscles, the distribution of load
535
preferences is relatively uniform with no obvious clustering. The units spanned all possible
536
combinations of control, rather than showing a reduced set of controls. The spanned distribution was
537
mirrored by the distributions of load and reach preferences observed for neurons in M1. As well, studies
538
using STA also highlight diverse connectivity in their cell samples (Smith and Fetz, 2009; Alstermark
539
and Isa, 2012). Thus, there are likely many more motor fields (and muscle synergies) represented in M1
540
neurons than muscles in the body.
541
It was surprising to find atypical cells where the directional preference of the neuron during
542
reaching was opposite to that of its muscle group identified from the postural load task. This sub-
543
population of neurons could not be identified in Scott (1997). In that study, neurons were first divided
544
into elbow and shoulder groupings by their response to passive movement of the arm. Separation into
545
flexor or extensor sub-groups was defined by their preferred direction during reaching, and thus, they
546
would have simply been classified with the antagonist muscle group at that joint.
547
The onset time of this atypical population of neurons was delayed as compared to neurons with
548
directional preferences aligned with the directional preference of their motor field. One possible
549
explanation is that these neurons are preferentially involved in controlling muscle activity when it is an
550
antagonist to decelerate the limb. Drew and colleagues (Drew, 1993; Drew et al., 2002) found that some
551
M1 neurons in cats during locomotion were only active when a specific modification of the muscle
552
activity was required, suggesting that the activity of some M1 neurons could be task-dependent. Task-
553
dependent processing in M1 has also been observed in previous reaching and posture comparisons
554
(Kurtzer et al., 2005). While the latency between atypical and typical activity was much shorter than the
555
latency between the muscular agonist and antagonist, a recent study by Griffin et al. (2015) also showed
556
that cells which were functionally tuned for antagonist activity show activity only shortly after those
25
557
tuned for agonist activity. The antagonist activity was relatively weak in the present study due to the
558
speed of reaching and contribution of friction from the exoskeleton assisting the deceleration of the
559
limb. The potential contribution of these atypical neurons for controlling antagonist muscle activity
560
could be examined by modifying the speed of the movement or inertia of the limb.
561
562
26
563
564
565
566
Acknowledgements
567
568
We would like to acknowledge Kim Moore, Simone Appaqaq, Justin Peterson, and Helen Bretzke for
their laboratory and technical assistance.
569
570
Grants
571
572
573
574
This work was supported by the Canadian Institutes of Health Research (CIHR). E. Heming received a
Doctoral Award from NSERC. M. Omrani received a Vanier Doctoral Award from CIHR. J. A.
Pruszynski received salary awards from CIHR and the Human Frontier Science Program. S. H. Scott is
supported by a GSK-CIHR Chair in Neuroscience.
575
576
Disclosures
577
578
S. H. Scott is associated with BKIN Technologies, which commercializes the KINARM device used in
this study.
27
579
References
580
Alstermark B, Isa T. Circuits for Skilled Reaching and Grasping. Annu Rev Neurosci 35: 559–578, 2012.
581
582
d’Avella A, Portone A, Fernandez L, Lacquaniti F. Control of Fast-Reaching Movements by Muscle Synergy
Combinations. J Neurosci 26: 7791–7810, 2006.
583
584
Bennett KM, Lemon RN. The influence of single monkey cortico-motoneuronal cells at different levels of
activity in target muscles. J Physiol 477: 291–307, 1994.
585
586
Bennett KM, Lemon RN. Corticomotoneuronal contribution to the fractionation of muscle activity during
precision grip in the monkey. J Neurophysiol 75: 1826–1842, 1996.
587
588
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.
589
590
Cabel DW, Cisek P, Scott SH. Neural Activity in Primary Motor Cortex Related to Mechanical Loads Applied
to the Shoulder and Elbow During a Postural Task. J Neurophysiol 86: 2102–2108, 2001.
591
592
593
Caminiti R, Johnson PB, Galli C, Ferraina S, Burnod Y. Making arm movements within different parts of
space: the premotor and motor cortical representation of a coordinate system for reaching to visual targets. J
Neurosci 11: 1182–1197, 1991.
594
595
Caminiti R, Johnson PB, Urbano A. Making arm movements within different parts of space: dynamic aspects in
the primate motor cortex. J Neurosci 10: 2039–2058, 1990.
596
597
Cheney PD, Fetz EE. Functional classes of primate corticomotoneuronal cells and their relation to active force. J
Neurophysiol 44: 773–791, 1980.
598
599
600
Cheney PD, Fetz EE. Comparable patterns of muscle facilitation evoked by individual corticomotoneuronal
(CM) cells and by single intracortical microstimuli in primates: evidence for functional groups of CM cells. J
Neurophysiol 53: 786–804, 1985.
601
602
Cheng EJ, Scott SH. Morphometry of Macaca mulatta forelimb. I. Shoulder and elbow muscles and segment
inertial parameters. J Morphol 245: 206–224, 2000.
603
604
Cherian A, Fernandes HL, Miller LE. Primary motor cortical discharge during force field adaptation reflects
muscle-like dynamics. J Neurophysiol 110: 768–783, 2013.
605
606
Drew T. Motor cortical activity during voluntary gait modifications in the cat. I. Cells related to the forelimbs. J
Neurophysiol 70: 179–199, 1993.
607
608
Drew T, Jiang W, Widajewicz W. Contributions of the motor cortex to the control of the hindlimbs during
locomotion in the cat. Brain Res Rev 40: 178–191, 2002.
609
610
Evarts EV. Relation of pyramidal tract activity to force exerted during voluntary movement. J Neurophysiol 31:
14–27, 1968.
611
Fetz EE, Cheney PD. Muscle fields of primate corticomotoneuronal cells. J Physiol (Paris) 74: 239–245, 1978.
28
612
613
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; discussion 427–430, 1989.
614
615
Flanders M, Pellegrini JJ, Soechting JF. Spatial/temporal characteristics of a motor pattern for reaching. J
Neurophysiol 71: 811–813, 1994.
616
617
Fromm C, Evarts EV. Relation of motor cortex neurons to precisely controlled and ballistic movements.
Neurosci Lett 5: 259–265, 1977.
618
Georgopoulos AP. Current issues in directional motor control. Trends Neurosci 18: 506–510, 1995.
619
620
Georgopoulos AP, Kalaska JF, Caminiti R, Massey JT. On the relations between the direction of twodimensional arm movements and cell discharge in primate motor cortex. J Neurosci 2: 1527–1537, 1982.
621
622
Graham KM, Moore KD, Cabel DW, Gribble PL, Cisek P, Scott SH. Kinematics and Kinetics of Multijoint
Reaching in Nonhuman Primates. J Neurophysiol 89: 2667–2677, 2003.
623
624
Graham KM, Scott SH. Morphometry of macaca mulatta forelimb. III. moment arm of shoulder and elbow
muscles. J Morphol 255: 301–314, 2003.
625
626
Gribble PL, Scott SH. Overlap of internal models in motor cortex for mechanical loads during reaching. Nature
417: 938–941, 2002.
627
628
Griffin DM, Hoffman DS, Strick PL. Corticomotoneuronal cells are “functionally tuned.” Science 350: 667–
670, 2015.
629
630
Hepp-Reymond M-C, Kirkpatrick-Tanner M, Gabernet L, Qi H-X, Weber B. Context-dependent force
coding in motor and premotor cortical areas. Exp Brain Res 128: 123–133, 1999.
631
632
Herter TM, Korbel T, Scott SH. Comparison of Neural Responses in Primary Motor Cortex to Transient and
Continuous Loads During Posture. J Neurophysiol 101: 150–163, 2009.
633
634
Herter TM, Kurtzer I, Cabel DW, Haunts KA, Scott SH. Characterization of Torque-Related Activity in
Primary Motor Cortex During a Multijoint Postural Task. J Neurophysiol 97: 2887–2899, 2007.
635
636
Hong DA, Corcos DM, Gottlieb GL. Task dependent patterns of muscle activation at the shoulder and elbow for
unconstrained arm movements. J Neurophysiol 71: 1261–1265, 1994.
637
Humphrey DR. Relating motor cortex spike trains to measures of motor performance. Brain Res 40: 7–18, 1972.
638
639
Johnson MTV, Coltz JD, Ebner TJ. Encoding of target direction and speed during visual instruction and arm
tracking in dorsal premotor and primary motor cortical neurons. Eur J Neurosci 11: 4433–4445, 1999.
640
641
Kakei S, Hoffman DS, Strick PL. Muscle and Movement Representations in the Primary Motor Cortex. Science
285: 2136–2139, 1999.
642
643
644
Kalaska JF, Cohen DA, Hyde ML, Prud’homme M. A comparison of movement direction-related versus load
direction-related activity in primate motor cortex, using a two-dimensional reaching task. J Neurosci Off J Soc
Neurosci 9: 2080–2102, 1989.
29
645
646
Karst GM, Hasan Z. Timing and magnitude of electromyographic activity for two-joint arm movements in
different directions. J Neurophysiol 66: 1594–1604, 1991.
647
648
Kasser RJ, Cheney PD. Characteristics of corticomotoneuronal postspike facilitation and reciprocal suppression
of EMG activity in the monkey. J Neurophysiol 53: 959–978, 1985.
649
650
Kurtzer I, Herter TM, Scott SH. Random change in cortical load representation suggests distinct control of
posture and movement. Nat Neurosci 8: 498–504, 2005.
651
652
Kurtzer I, Herter TM, Scott SH. Nonuniform Distribution of Reach-Related and Torque-Related Activity in
Upper Arm Muscles and Neurons of Primary Motor Cortex. J Neurophysiol 96: 3220–3230, 2006a.
653
654
Kurtzer I, Pruszynski JA, Herter TM, Scott SH. Primate Upper Limb Muscles Exhibit Activity Patterns That
Differ From Their Anatomical Action During a Postural Task. J Neurophysiol 95: 493–504, 2006b.
655
656
Lemon RN, Muir RB, Mantel GWH. The effects upon the activity of hand and forearm muscles of intracortical
stimulation in the vicinity of corticomotor neurones in the conscious monkey. Exp Brain Res 66: 621–637, 1987.
657
658
Lillicrap TP, Scott SH. Preference Distributions of Primary Motor Cortex Neurons Reflect Control Solutions
Optimized for Limb Biomechanics. Neuron 77: 168–179, 2013.
659
660
Marsden CD, Obeso JA, Rothwell JC. The function of the antagonist muscle during fast limb movements in
man. J Physiol 335: 1–13, 1983.
661
662
663
McKiernan BJ, Marcario JK, Karrer JH, 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.
664
Muir RB, Lemon RN. Corticospinal neurons with a special role in precision grip. Brain Res 261: 312–316, 1983.
665
666
Murphy JT, Wong YC, Kwan HC. Sequential activation of neurons in primate motor cortex during unrestrained
forelimb movement. J Neurophysiol 53: 435–445, 1985.
667
668
Oby ER, Ethier C, Miller LE. Movement representation in the primary motor cortex and its contribution to
generalizable EMG predictions. J Neurophysiol 109: 666–678, 2013.
669
670
Palmer E, Ashby P. Corticospinal projections to upper limb motoneurones in humans. J Physiol 448: 397–412,
1992.
671
672
Philip BA, Rao N, Donoghue JP. Simultaneous reconstruction of continuous hand movements from primary
motor and posterior parietal cortex. Exp Brain Res 225: 361–375, 2013.
673
674
Phillips CG. Laying the ghost of “muscles versus movements.” Can J Neurol Sci J Can Sci Neurol 2: 209–218,
1975.
675
676
Pruszynski JA, Omrani M, Scott SH. Goal-Dependent Modulation of Fast Feedback Responses in Primary
Motor Cortex. J Neurosci 34: 4608–4617, 2014.
30
677
678
Raphael G, Tsianos GA, Loeb GE. Spinal-Like Regulator Facilitates Control of a Two-Degree-of-Freedom
Wrist. J Neurosci 30: 9431–9444, 2010.
679
680
Rathelot J-A, Strick PL. Subdivisions of primary motor cortex based on cortico-motoneuronal cells. Proc Natl
Acad Sci 106: 918–923, 2009.
681
682
Schwartz AB. Motor cortical activity during drawing movements: population representation during sinusoid
tracing. J Neurophysiol 70: 28–36, 1993.
683
684
Scott SH. Comparison of Onset Time and Magnitude of Activity for Proximal Arm Muscles and Motor Cortical
Cells Before Reaching Movements. J Neurophysiol 77: 1016–1022, 1997.
685
686
Scott SH. Apparatus for measuring and perturbing shoulder and elbow joint positions and torques during
reaching. J Neurosci Methods 89: 119–127, 1999.
687
688
Scott SH. The role of primary motor cortex in goal-directed movements: insights from neurophysiological studies
on non-human primates. Curr Opin Neurobiol 13: 671–677, 2003.
689
690
Scott SH, Gribble PL, Graham KM, Cabel DW. Dissociation between hand motion and population vectors
from neural activity in motor cortex. Nature 413: 161–165, 2001.
691
692
Scott SH, Kalaska JF. Reaching Movements With Similar Hand Paths But Different Arm Orientations. I.
Activity of Individual Cells in Motor Cortex. J Neurophysiol 77: 826–852, 1997.
693
694
Sergio LE, Hamel-Pâquet C, Kalaska JF. Motor Cortex Neural Correlates of Output Kinematics and Kinetics
During Isometric-Force and Arm-Reaching Tasks. J Neurophysiol 94: 2353–2378, 2005.
695
696
697
Sergio LE, Kalaska JF. Systematic Changes in Directional Tuning of Motor Cortex Cell Activity With Hand
Location in the Workspace During Generation of Static Isometric Forces in Constant Spatial Directions. J
Neurophysiol 78: 1170–1174, 1997.
698
699
Sergio LE, Kalaska JF. Systematic Changes in Motor Cortex Cell Activity With Arm Posture During
Directional Isometric Force Generation. J Neurophysiol 89: 212–228, 2003.
700
701
Singh K, Melis EH, Richmond FJR, Scott SH. Morphometry of Macaca mulatta forelimb. II. Fiber-type
composition in shoulder and elbow muscles. J Morphol 251: 323–332, 2002.
702
703
Smith WS, Fetz EE. Synaptic Linkages Between Corticomotoneuronal Cells Affecting Forelimb Muscles in
Behaving Primates. J Neurophysiol 102: 1040–1048, 2009.
704
705
Todorov E. Direct cortical control of muscle activation in voluntary arm movements: a model. Nat Neurosci 3:
391–398, 2000.
706
707
Toxopeus CM, de Jong BM, Valsan G, Conway BA, Leenders KL, Maurits NM. Direction of Movement Is
Encoded in the Human Primary Motor Cortex. PLoS ONE 6: e27838, 2011.
708
709
Turner RS, DeLong MR. Corticostriatal Activity in Primary Motor Cortex of the Macaque. J Neurosci 20:
7096–7108, 2000.
31
710
711
Wierzbicka MM, Wiegner AW, Shahani BT. Role of agonist and antagonist muscles in fast arm movements in
man. Exp Brain Res 63: 331–340, 1986.
712
713
32
714
Figure 1. Experimental posture and reaching tasks. A. In the posture task, the monkey was required to
715
maintain its hand at a central target while flexion and/or extension torques were applied at the elbow and
716
shoulder. The right panel illustrates the eight combinations of torques and an unloaded condition. B. In
717
the reaching task, relatively fast reaches were made from a central origin to eight peripheral targets. The
718
right panel displays the four different combinations of targets used in the datasets.
719
720
Figure 2. Predictions from the static neural network model. A. A representative schematic of one of the
721
models. Two network units, z, are shown connected to muscle outputs, u, via random weightings shown
722
by the thickness of the lines. The output variables, u, drove mono- and bi-articular muscles on a two-
723
joint arm, as shown. The values of the units in z were optimized to solve analogues of the reaching and
724
posture tasks. B. The relationship between each unit’s preferred torque direction in the posture task and
725
the calculated preferred torque direction of its motor field (defined by synaptic connectivity) in one
726
network of 1000 units. The size of the circle for each unit corresponds to its strength of tuning. Note that
727
the muscles, plotted here for reference, must lie on the diagonal as they correspond with their own motor
728
field.
729
730
Figure 3. Preferred torque and reaching directions. A. Polar histogram of each network units’ preferred
731
torque direction (PTD) in one network. Angles are in torque space (shoulder torque x-axis, elbow torque
732
y-axis). B. Polar histogram of each network units’ preferred reaching direction (PRD) in one network. C.
733
Exemplar activity of pectoralis major in eight shoulder/elbow loaded conditions (unloaded condition not
734
shown) during the last three seconds of in the posture task. The arrow in the centre denotes the muscle’s
735
preferred torque direction (351 degrees). D. Exemplar activity of pectoralis major around reach onset
736
while reaching to the eight targets (set iii in Fig. 1B). The arrow in the centre denotes the muscle’s PRD
33
737
(140 degrees). E-F. Averages of the PTD and PRD of all muscle samples are shown with 95%
738
confidence intervals as shaded areas around each line. G. Activity of an exemplar M1 neuron in the
739
eight shoulder/elbow loaded conditions (unloaded condition not shown) during the last three seconds of
740
in the posture task. The arrow in the centre denotes the neuron’s preferred torque direction (155
741
degrees). H. Activity of an exemplar M1 neuron around reach onset while reaching to the eight targets
742
(Set iii in Fig. 1B). The arrow in the centre denotes the neuron’s PRD (119 degrees). I-J. Polar
743
histograms of all significant PTDs and PRDs across the population of recorded M1 neurons.
744
745
Figure 4. E. Relationship between PTD and PRD (directional preferences must be significant in both
746
tasks) for network units, muscles, and M1 neurons. A. Scatter relationship for network units. B.
747
Histogram of the difference between PTD and PRD for network units, inset shows as polar plot. C.
748
Scatter relationship for muscle recordings. D. Histogram of the difference between PTD and PRD for
749
muscle recordings. E. Scatter relationship for M1 neuron recordings. F. Histogram of the difference
750
between PTD and PRD for M1 neuron recordings.
751
752
Figure 5. Distribution of load preferences for M1 neurons and individual muscles. The number of
753
recordings for each muscle is indicated in brackets beside the muscle name. The four assigned motor
754
fields are shown as vertical shaded stripes with their titles marked at the top.
755
756
Figure 6. Distribution of preferred reaching direction for each muscle group, and assigned M1 neurons
757
and network units. Each distribution is normalized as a percentage of the total number of significant
758
PRDs. The x-axis starts at rightward reach (0⁰) and increases counter clockwise.
759
34
760
Figure 7. Spatiotemporal dynamics of muscle groups and associated M1 neuron populations during
761
reaching. A. Colormaps of the aggregate activity over time (x-axis) and across reach direction (y-axis)
762
of each muscle group and associated population of M1 neurons. Data are smoothed in the y-axis for
763
display purposes. B. Two-dimensional correlation coefficients between the spatiotemporal patterns of
764
muscle and neuron groups. Each graph is one muscle group and the bars show the coefficients between
765
that muscle’s spatiotemporal pattern and the pattern of the four populations of M1 neurons. C.
766
Aggregate activity of all neurons and all muscles were generated by rotating the y-axis of each group so
767
that their directions of maximal agonist activity were aligned with the Agonist direction, and then
768
averaged.
769
770
Figure 8. Aggregate spatiotemporal colormaps of M1 neurons that have similar and opposing PRDs
771
compared to their associated muscle group. A. Colormap of population activity of all neurons whose
772
PRD was within 30⁰ of its assigned muscle group (typical neurons). Aggregate of neural activity was
773
rotated as in Figure 7. B Colormap of population activity of all neurons whose PRD was greater than
774
150⁰ away from that of its assigned muscle group (atypical neurons). C. Spatiotemporal pattern of
775
activation for typical (black lines) and atypical (red lines) neurons when the muscle group acts as an
776
agonist (solid line) and an antagonist (dashed line).
777
778
35
A.
Postural contraction
against loads
B.
Monkey reaches to
peripheral targets
i
ii
iii
iv
A.
w
z
1
u
Elb Flex
2
Elb Ext
3
Sho Flex
4
5
Sho Ext
6
Bi Flex
Bi Ext
N
B.
Motor Field PTD (°)
360
0
0
Posture PTD (°)
360
A. Model Posture PTD C.
EMG
in Posture Task
E. EMG Posture PTD
G.
M1 cell
in Posture Task
I.
N=373
Model Reach PRD
D.
EMG
in Reaching Task
F.
EMG Reach PRD
N=125
1au
B.
1s
20 sp/s
BR
Bi
Br
DA
DP
PM
Tlat
Tlong
Tmid
1s
H.
M1 cell
in Reaching Task
J.
M1 Reach PRD
N=424
20 sp/s
1au
N=133
1s
M1 Posture PTD
1s
B.
A.
Neurons (%)
Reach PRD (°)
20
0
0
Posture PTD (°)
10
0
360
C.
0
D.
Recordings (%)
0
90
180
270
360
270
360
270
360
Difference (°)
PRD-PTD
30
179
Reach PRD (°)
152°
15
5
180
25
N=122
152°
20
15
10
5
180
0
Posture PTD (°)
0
360
E.
F.
179
0
12
10
Neurons (%)
Reach PRD (°)
PRD-PTD
25
179
0
180
90
180
Difference (°)
PRD-PTD
N=314
151°
8
6
4
2
0
Posture PTD (°)
360
0
0
90
180
Difference (°)
Elbow Shoulder
flexor extensor
Elbow Shoulder
extensor flexor
Model units
M1 neurons (373)
Biceps (25)
Brachioradialis (18)
Brachialis (12)
Long head triceps (24)
Posterior deltoid (18)
Lateral triceps (8)
Middle triceps (2)
Anterior deltoid (14)
Pectoralis major (12)
0
45
90
135
180
225
270
315
360
A.
shoulder flexor
B.
elbow flexor
D.
elbow extensor
30%
Muscles
M1 neurons
Network
20%
10%
0%
C.
shoulder extensor
30%
20%
10%
0%
0°
360° 0°
360°
Preferred reaching direction Preferred reaching direction
EMG behavior
B.
C.
1
Aggregate M1 prediction
Antagonist
Correlation
Shoulder
flexor
related
M1 prediction
Reach direction
A.
0
Agonist
Reach direction
Elbow
flexor
related
Correlation
1
0
Aggregate EMG behavior
Antagonist
Correlation
Shoulder
extensor
related
Reach direction
1
0
Agonist
Reach direction
Correlation
1
Elbow
extensor
related
0
-1
-1
1
Time around reach onset (s)
1
SF EF SE EE
neuron group
-1
0
Time around reach onset (s)
1
Agonist
Antagonist
A.
Agonist
Antagonist
B.
-1000
100
Percent max activity
C.
1000
0
Time around reach onset (ms)
80
60
40
20
0
-500
0
500
Time around reach onset (ms)