| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J Neurophysiol (December 1, 2002). 10.1152/jn.00070.2002
Submitted on 4 February 2002
Accepted on 6 August 2002
1Department of Physiology, Hirosaki University School of Medicine, Hirosaki 036-8562; and 2Department of Physiology, Tohoku University School of Medicine, Sendai 980-8575 Japan
 |
ABSTRACT |
|---|
 TOP
 ABSTRACT
 INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Kurata, Kiyoshi and Eiji Hoshi. Movement-Related Neuronal Activity Reflecting the Transformation of Coordinates in the Ventral Premotor Cortex of Monkeys. J. Neurophysiol. 88: 3118-3132, 2002. We examined how the transformation of coordinates from visual to motor space is reflected by neuronal activity in the ventral premotor cortex (PMv) of monkeys. Three monkeys were trained to reach with their right hand for a target that appeared on a screen. While performing the task, the monkeys wore prisms that shifted the image of the target 10°, left or right, or wore no prisms, for a block of 200 trials. The nine targets were located in the same positions in visual space regardless of whether the prisms were present. Wearing the prisms required the monkeys to initiate a movement in a direction that was different from the apparent target location. Thus using the prisms, we could dissociate visual space from motor space. While the monkey performed the behavioral task, we recorded neuronal activity in the left PMv and primary motor cortex (MI), and various kinds of task-related neuronal activity were found in the motor areas. These included neurons that changed their activity during a reaction time (RT) period (the period between target presentation and movement onset), which were called "movement-related neurons" and selected for analysis. In these neurons, activity during a movement time (MT) period was also compared. Using general linear models for our statistical analysis, the neurons were then classified into four types: those whose activity was consistently dependent on location of targets in the visual coordinates regardless of whether the prisms were present or absent (V type); those that were consistently dependent on target location in the motor coordinates only; those that had different activity for both of the motor and visual coordinates; and those that had nondifferential activity for the two types of coordinates. The proportion of the four types of the neurons differed significantly between the PMv and MI. Most remarkably, neurons with V-type activity were almost exclusively recorded in the PMv and were almost exclusively found during the RT period. Such activity was never observed in an electromyogram of the working forelimb. Based on these observations, we postulate that the V and other types may represent the various intermediate stages of the transformation of coordinates and that the PMv plays a crucial role in transforming coordinates from visual to motor space.
| |
INTRODUCTION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
In visually guided goal-directed
movements, the CNS should first map the location of movement targets in
a retinocentric visual-reference frame and then transform the
visual-reference frame into an intrinsic motor-reference frame directly
or indirectly through several intermediate representations in the
head- or body-part-centered coordinates (Alexander and Crutcher
1990
; Andersen et al. 1993; Atkeson
1989; Kawato et al. 1988). Such sensorimotor
integration is known as transformation of coordinates and has been the
subject of continuous interest for researchers in the field.
Although the transformation in the visually guided movements can
potentially occur in a number of sites in the brain, recent neurophysiological and anatomical studies have provided much evidence that the ventral premotor cortex (PMv) is one of the best candidates. First, the PMv receives abundant corticocortical projections from the
parietal cortex (Cavada and Goldman-Rakic 1989;
Godschalk et al. 1984; Kurata 1991),
which contains cells with a visual receptive field (Colby et al.
1993). The PMv is also heavily interconnected with the primary
motor cortex (MI) (Leichnetz 1986; Muakkassa and
Strick 1979) and projects to the spinal cord (Dum and
Strick 1991; He et al. 1993). There is a sensory
aspect of the transformation in the PMv: PMv neurons respond to visual
and somatosensory stimuli conveying spatial information in
retinocentric, head-centered, or body-part-centered coordinates
(Boussaoud et al. 1993; Fogassi et al.
1996; Godschalk et al. 1985; Graziano et
al. 1994, 1997b; Rizzolatti et al. 1981). On the
other hand, the activity of neurons in the PMv also changes before
forelimb movements (Caminiti et al. 1991;
Godschalk et al. 1985; Kurata 1989, 1993;
Kurata and Hoffman 1994; Kurata and Tanji
1986; Murata et al. 1997; Weinrich et al.
1984). Many of these movement-related neurons are selectively active when visual stimuli trigger a movement (Kurata and Wise 1988; Murata et al. 1997; Mushiake et al.
1991; Okano and Tanji 1987; Rizzolatti et
al. 1987; Romo and Schultz 1987). Recently, motor aspects of the transformation in the PMv have been reported: the
PMv contains neurons encoding the direction of reaching movement in
retinocentric coordinates (Mushiake et al. 1997)
and those encoding the direction of wrist movements in extrinsic space
independent of forearm posture (Kakei et al. 2001).
Furthermore, removal or inactivation of the PMv results in visuomotor
deficits (Kurata and Hoshi 1999; Schieber
2000) or visual neglects (Rizzolatti et al.
1983).
However, little is yet known whether movement-related activity in the PMv reflects a head-centered visual-reference frame as an intermediate representation in a series of the transformations. To directly assess this possibility, we dissociated visual and motor-reference frames by applying shift prisms to the monkeys that were allowed to move their eyes freely during a reaching task. This protocol enabled us to examine which reference frame, visual, motor, or intermediate, was represented by neuronal activity in the PMv. Thus we focused on movement-related neurons whose activity changed during the reaction time (RT) period when visuomotor transformation for goal-directed reaching was expected to occur and compared their activity with that during a movement time (MT) period when the reaching was being executed. We also recorded movement-related activity in the MI to determine whether the PMv and MI have differential roles in the transformation.
| |
METHODS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Subjects and apparatus
Three male Japanese monkeys (Macaca fuscata)
weighing 5.1-6.2 kg were trained to perform a visually guided task
that involved reaching toward a target with the right arm. All the
experiments were conducted in accordance with the standards of the
Guide for the Humane Care and Use of Animals published by
the American Physiological Society (http://www.the-aps.org/pub
affairs/humane/pa_aps_guiding.htm). The apparatus and two of
the three monkeys (monkeys 1 and 2) were the same
as used in our previous report (Kurata and Hoshi 1999). The monkeys sat comfortably in a primate chair facing a 14-in CRT
screen covered with a transparent touch panel that monitored the
position of the monkey's hand on the screen by detecting local pressure. The positions on the touch screen were sampled at 500 Hz
through an 8-channel, 12-bit A/D converter and stored in a laboratory
computer. The screen was placed 30 cm away from the monkey's eyes, and
the vertical centers of the screen and the monkey's head and body were
aligned. An apparatus with two pairs of 4 × 4-cm wedge prisms
(10° to the left or right) was placed immediately in front of the
monkey's eyes. The apparatus was 46 cm wide, 17 cm high, and 8 cm deep
and had a 4 × 8-cm hole in which a pair of the prisms was placed.
Within the apparatus, each pair of the prisms was aligned horizontally
on a mounting frame, and the prisms were separated by 5 cm, which
matched the monkey's interocular distance. The mounting frame also had
a pair of 4 × 4-cm holes without prisms. The monkey's visual
field through the prisms apparatus matched the size of the CRT screen.
Due to the size and placement of the prism apparatus, the monkey was not able to see its hand until the hand entered the visual field seen
through the hole of the prism apparatus. A switch made of a 5 × 10-cm acrylic plate was placed at the end of the right armrest to serve
as a hold key. The monkey's left arm was immobilized on the arm rest
by Velcro straps.
Behavioral task
In each trial, the monkeys were required to reach for a 5 × 8-mm blue rectangular target that randomly appeared in one of nine
locations (Fig. 1, A-C), with
or without the prisms. The targets were separated by 6.0 cm or 10° of
visual angle horizontally and vertically. Throughout the sessions, the
location of the targets remained the same in terms of the visual
coordinates, regardless of whether the 10° wedge prisms were present
(Figure 1A-D) (see also Fig. 2 in Kurata and
Hoshi 1999). The central target was always located at the
intersection of the vertical and horizontal meridians of the visual
coordinates (V08 in Fig. 1D), regardless of the prism
conditions. When no prism was applied (Fig. 1B), the visual
and motor coordinates of the targets matched. When the left 10° wedge
prisms were applied, the motor coordinates were shifted 10° to the
right of the visual coordinates (Fig. 1A). Conversely, when
the right 10° wedge prisms were applied, the motor coordinates were
shifted 10° to the left of the visual coordinates (Fig.
1C). The conditions in Fig. 1, A-C, were called the left prism, no prism, and right prism conditions, respectively (Kurata and Hoshi 1999). There were 15 motor targets
(M01-15) in the motor coordinates (Fig. 1E). The target V08
corresponded to M07, M08, M09 in the right, no, and left prism
conditions, respectively.
|
To start each trial, the monkey was required to press the hold key with
its right hand. When it pressed the key continuously for a period
varying between 1.5 and 3.0 s, a target was selected randomly from
the nine possible locations and presented on the screen (Fig. 1). If
the monkey released the hold key (movement onset) within 500 ms of the
target's appearance, reached the touch screen within 500 ms of the
movement onset, and hit the correct target, then 0.1 ml of orange juice
was delivered as a reward. The RT was defined as the period between the
appearance of the target and movement onset. The MT was defined as the
period between movement onset and contact with the screen. We required
the monkeys to initiate and execute the movement quickly so that they
transformed the coordinates during the RT period with minimal sensory
feedback during the MT period (Kurata and Hoshi 1999).
For at least 2 wk before data collection, the monkeys were trained to
perform the task with and without the prisms. In any prism condition,
they were allowed to hit the screen first (not necessarily the target) and then make a corrective movement of their hand to the target. The
corrective movements were highly variable from trial to trial. In some
trials, the monkeys hit the targets without any corrective movements,
whereas, in other trials, they continued to make the movements until
they got rewards. We analyzed neuronal and EMG data only during the
period up to the first contact with the screen. The monkeys adapted to
the prisms, and after adaptation, the first contact point on the screen
was generally close to the target, irrespective of the prism condition
(see Fig. 3 of Kurata and Hoshi 1999). Eye movements
were not monitored in monkeys 1 and 2 but were
monitored in monkey 3 using an infrared oculometer system
(R21CA, RMS, Hirosaki, Japan). All the monkeys were free to move their
eyes at any time during a trial.
EMG recording
Electromyographic (EMG) activity was monitored bilaterally with surface and wire electrodes from the following muscles: the biceps and triceps brachii, anterior and lateral deltoid, extensor carpi radialis, flexor carpi ulnaris, trapezius, supraspinatus, infraspinatus, pectoralis major, rhomboid, thoracic and lumbar paravertebral muscles, gluteus maximus, quadriceps, and tibialis anterior. EMG activity from each muscle was stably recorded twice, before chamber implantation and near the completion of data collection from each monkey. The EMG data recorded at the two stages were separately analyzed (see following text). The EMG data were sampled at 100 Hz through the A/D converter and stored with the position data in the laboratory computer.
Neuronal recording method
After completion of the behavioral training, a stainless steel recording chamber (27 × 27 mm) and head-fixation bolts were implanted on the skull under aseptic conditions. The monkeys were anesthetized with pentobarbital sodium (30 mg/kg im) after induction with ketamine hydrochloride (8 mg/kg im) and atropine sulfate. During surgery, additional ketamine hydrochloride was given as necessary. Antibiotics and analgesics were used to prevent postsurgical infection and pain.
After complete recovery from the surgery (more than 7 days), neuronal
activity was recorded from the PMv and MI during task performance. The
explored areas were selected based on the central and arcuate sulci and
the arcuate spur that were observed during surgery. It was confirmed by
histological reconstruction that the areas covered the proximal
forelimb representations of the PMv and MI (Gentilucci et al.
1988; Kurata and Tanji 1986). We used
glass-insulated Elgiloy microelectrodes (1.5-2.0 M
at 333 Hz) for
single-unit recording. The microelectrodes were inserted through the
dura mater with a hydraulic microdrive (Narishige, MO95). The same
microelectrodes were used for intracortical microstimulation (ICMS).
Each ICMS consisted of a train of 11 cathodal pulses of less than 50 µA and 0.2-ms duration at 333 Hz. ICMS was used to identify the MI
physiologically, and neuronal activity within the MI was recorded in
the proximal forelimb representation areas but not in the distal
forelimb or orofacial areas. We isolated neurons with an increased
discharge rate during the RT period and then examined their response to
visual stimuli. Two types of visual stimuli were used: the same stimuli
that served as targets were presented for 100 ms and an experimenter's
hand moving toward the monkey's eyes (Fogassi et al. 1992,
1996). When a neuron showed increased activity during the RT
period and did not respond to these visual stimuli, its activity was
recorded in the three prism conditions. Blocks of trials were recorded
in the no prism condition first, then in either the right or left prism
condition, followed by the no prism condition a second time, and
finally in the remaining prism condition. The order of the right and
left prism conditions was randomized in a daily session. The second no
prism condition was used to check the excitability of the cell. In each
block, neuronal activity was recorded for approximately 200 trials,
i.e., more than 20 trials for each of the nine targets. If stable
neuronal activity was recorded, an optional recording was done in the
no prism condition using two sets of targets shifted to the locations in Fig. 1, A and C.
Data analysis
For each of the nine targets in the three prism
conditions, a raster display of recorded neuronal activity was aligned
with the movement onset, and a peri-event histogram with a 20-ms
binwidth was created to show neuronal activity for each target in one
of the prism conditions (Figs. 5, 7, and 9). For each display and quantitative analysis, the data after the first 10 trials during adaptation to the prisms (Kurata and Hoshi 1999) were
taken because, in this study, it was essential to obtain the data best
corresponding to the motor coordinates in each prism condition.
Furthermore, the data were added to the database only when neuronal
activity was stably recorded for more than 10 trials for each target in a prism condition and the neuron showed a consistent activity pattern
throughout the recorded trials. The mean discharge rate and its SD
during the 0.5- to 1.5-s interval before target presentation (premovement period) were calculated first. If the neuronal activity exceeded 2.56 SD (P < 0.01) during the RT in at least
two consecutive bins of at least three of the nine histograms in the no
prism condition, it was defined as movement-related activity. Then the time when the activity first exceeded the threshold value was defined
as the onset of neuronal activity in each display. After the earliest
onset of neuronal activity was obtained from the nine displays, the
mean discharge rate during the period from the earliest onset of
neuronal activity to movement onset was calculated in each trial,
regardless of whether the neuron showed a statistically significant
change during the period, in each of the nine displays. The mean
discharge rate during the period in each trial was used for subsequent
quantitative and statistical analyses. For each of the movement-related
neurons, a mean discharge rate and SD during MT in each trial were also
calculated for quantitative and statistical analysis.
To classify neuronal activity, the mean discharge rates during
the sampling times (either RT or MT) were compared statistically using
the general linear model (GLM) of SYSTAT for Windows (ver. 8.0.2, Chicago, IL). First, we selected neurons with activity dependent on
target location in either the visual or motor coordinates. Without the
variation, it cannot be judged whether the activity reflects visual or
motor space. Second, it was essential to judge whether the activity
reflected visual or motor space, and/or prism effects. Thus we compared
neuronal activity during the RT or MT at the visual and motor target
locations (Fig. 1, D and E) in the three prism
conditions, using linear regression models
|
(1) |
|
(2) |
If factor analysis for prism was not statistically significant (P > 0.01) in Eq. 1, but was statistically significant (P < 0.01) in Eq. 2 and if that for targetv was statistically significant (P < 0.01) in Eq. 1, then the activity was judged as consistently dependent on target location in the visual coordinates only, regardless of whether the prisms were present or absent (termed V type). This means that when a target was located at the same position in visual coordinates, the neuronal activity was always constant regardless of the prisms, and, dependent on the prisms, the activity was variable even when a target was constantly located at a given position in the motor coordinates. Similarly, if factor analysis for prism was not statistically significant (P > 0.01) in Eq. 2 but was statistically significant (P < 0.01) in Eq. 1 and if that for targetm was statistically significant (P < 0.01) in Eq. 2, then the activity was judged as consistently dependent on target location in the motor coordinates (termed M type). This indicates that when a target was located at the same position in the motor coordinates, the neuronal activity was always constant regardless of the prisms, and, dependent on the prisms, the activity was variable even when a target was constantly located at a given position in the visual coordinates. If factor analysis for prism was statistically significant (P < 0.01) in both Eqs. 1 and 2, then the activity was judged as differently active in both of the visual and motor coordinates (termed B type). In other words, dependent on the prisms, the neuronal activity was variable even when a target was located at the same position in the visual or motor coordinates. Finally, if factor analysis for prism was not statistically significant (P > 0.01) in Eqs. 1 nor 2, then the activity was judged as indifferently active in the two coordinates (termed N type). That is, N-type activity showed no prism effect, even when the neurons changed their activity depending on target locations in the visual or motor coordinates. The discharge rates of the movement-related neurons during MT and quantified EMG data were similarly analyzed and were classified into the four types (V, M, B, and N). Equations 1 and 2 were also used to examine whether differences in reaction and movement times and accuracy of reaching between visual and motor targets and prism conditions were statistically significant. In these analyses, we set 0.01 as the significance level, because each unit activity was dually compared with both Eqs. 1 and 2, so that type-1 error could be lower than 0.05 usually adopted for GLM. When an optional recording was obtained in the no prism condition at the target locations in Fig. 1, A and C, the data were included in the statistical analysis. As a result, the two equations were more balanced, and enabled us to check the validity of Eq. 2 without the data collected in the prism conditions.
After the neuronal activities were classified into the four types, the
correlation coefficients of each neuron to the visual and motor
coordinates were computed to obtain to what extent the neuron modulated
its activity dependent on the coordinates, using the average firing
rates during RT and MT. The formula used was
|
(3) |
, is called the Pearson product moment correlation
coefficient and has been used for a similar purpose to that used in
this study (Batista et al. 1999). To compute the
correlation in visual space coordinates (rvisual),
xi is the average firing rate for a reach
to a given visual target i under the no prism condition, and
yi is a reach to the same target under
either the left or right prism conditions, 
is the
average of the xi; 
is
the average of the xi; Histology
After collecting single-unit data, muscimol was injected into
the PMv of monkeys 1 and 2 (Kurata and
Hoshi 1999). When all the experiments were completed,
electrolytic marking lesions were produced by passing 20 µA of
cathodal DC through the microelectrodes for 15 s. Nine to 10 days
later, the monkeys were deeply anesthetized with pentobarbital (50 mg/kg im) and were perfused through the heart with saline followed by a
fixative containing 3.7% formaldehyde in 0.1 M phosphate buffer at pH
7.4, followed by 10 and 20% sucrose solutions in 0.1 M phosphate
buffer at pH 7.4.
After marking the location of the recording chamber with five pins at
known electrode coordinates, the brain was removed from the skull and
photographed. Later it was sectioned serially at 50-µm thickness in
the frontal plane using a freezing microtome. The PMv was defined as
the area within the dysgranular frontal cortex rostral to the primary
motor cortex (MI) and lateral to the arcuate spur, where ICMS at an
intensity less than 50 µA does not evoke muscle activity
(Barbas and Pandya 1987; Kurata 1993, 1994; Kurata and Hoffman 1994; Kurata and
Hoshi 1999; Matelli et al. 1985).
| |
RESULTS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
RT and MT
The mean RTs and MTs times for movements toward the nine targets in the three different prism conditions by the three monkeys are shown in Fig. 2 and listed in Table 1. In all the monkeys, the mean RTs were shorter than 300 ms, regardless of target location or prism condition, and there were no statistically significant differences in RT between any visual or motor target locations, or between any prism conditions (GLM, P > 0.01). MT to a given target did not vary with prism condition. Difference between prism conditions (prism in Eq. 2) was statistically not significant (GLM, P > 0.01). On the other hand, MT depended on the motor target location. Difference between the motor target locations (targetm in Eq. 2) was statistically significant (GLM, P < 0.01). Because the distances from the hold key (starting point) to the former visual targets were longer than the distances to the latter targets, these observations suggest that the MT depended on the movement distance, and not on presence of the prisms.
|
|
Reaching movement accuracy
Table 2 shows the average movement
accuracy in the three prism conditions. To measure movement accuracy,
the first 10 trials after changing the prisms, during which adaptation
was taking place, were excluded (see Fig. 3 in Kurata and Hoshi
1999). Differences in the movement accuracy of monkey
1 under the three prism conditions were not statistically
significant (GLM, P > 0.01), whereas monkeys 2 and 3 showed statistically significant difference in
movement accuracy between the prism conditions (GLM, P < 0.01). In the three monkeys, means of the first contact points were
located to the left of the target. This is partly due to mechanical
properties of the touch screen. Because the monkeys tended to contact
the screen with their index, middle, and ring fingers, the contact point was detected as a weighted average of the multiple contact points, which were usually left of the target. Compared with the inter-target distance (60 mm), the average distance between the center
of each target and the first contact point was relatively small,
regardless of the presence or absence of the shift prisms.
|
EMG activity during task performance
Of the muscles recorded, the anterior deltoid was found to be a prime mover for the reaching movements in all three monkeys. The muscle was active immediately before and during reaching movements. Figures 3 and 4 show representative EMG activity from the right anterior deltoid of monkey 1 and its quantified data during RT (between EMG onset and movement onset) and MT, respectively. In the motor coordinates, differences in the EMG activity during both RT and MT between prism conditions (prism in Eq. 2) were not statistically significant (GLM, P > 0.01), while difference between the motor targets (targetm in Eq. 2) was statistically significant (GLM, P < 0.01; Fig. 4, right). In the visual coordinates, however, differences in the EMG activity during RT and MT between prism conditions (prism in Eq. 1) were statistically significant (GLM, P < 0.01; Fig. 4, left). The EMG activities during both RT and MT were classified as M type. Thus the right anterior deltoid muscle changed its activity in relation to the targets which the movements were directed in the motor, but not visual, coordinates.
|
|
The other right shoulder and upper arm muscles, such as supra- and infraspinatus, rhomboid, trapezius, and teres major, were active during the task, but the activity was similar for each of the nine targets. The right proximal and distal forelimb muscles, such as the biceps and triceps brachii, extensor carpi radialis, flexor carpi ulnaris, changed activity after onset of reaching movements. None of the left shoulder or forelimb muscles changed activity during reaching movements. No muscle was specifically active during the holding period before a target was presented. Table 3 summarizes activity patterns of the recorded muscles in the three monkeys, classified as M, N, and B types. No V-type activity was found in any muscle.
|
Movement-related activity reflecting visual space
We found V-type activity almost exclusively during the RT period, and the activity was almost exclusively recorded in the PMv (Table 4). Figure 5 shows an example of V-type activity recorded in the PMv. The neuron changed its activity during RT, and no significant change of activity was observed during MT. The neuron showed activity dependent on target location during RT. Most remarkably, activities toward the same visual targets were almost identical in the three prism conditions. In contrast, the activities toward the same motor targets differed in the prism conditions. Figure 6 shows a quantitative analysis of the data shown in Fig. 5 in the visual and motor coordinates. Difference in the activity during RT between the visual targets (targetv in Eq. 1) was statistically significant (GLM, P < 0.01). When activities in the visual coordinates were compared, difference between prism conditions (prism in Eq. 1) was not statistically significant (GLM, P > 0.01). In contrast, when activities in the motor coordinates were compared, prism in Eq. 2 was statistically significant (GLM, P < 0.01). Thus the PMv activity during RT was classified as V type.
|
|
|
Movement-related activity reflecting motor space
We found M-type activity predominantly during the RT period, and the activity constituted 19 and 37% of the classified activities (V, M, B, or N) during RT in the PMv and MI, respectively (Table 4). Figure 7 shows a representative M-type activity during RT recorded in the PMv. Activities toward the same motor targets were almost identical in the three prism conditions. In contrast, the activities toward the same visual targets differed in the prism conditions. Figure 8 shows quantified data of the neuron in Fig. 7 that were aligned in the visual and motor coordinates. When activities during RT were compared in the motor coordinates, the difference between prism conditions (prism in Eq. 2) was not statistically significant (GLM, P > 0.01). In contrast, when the data were compared in the visual coordinates, difference between prism conditions (prism in Eq. 1) was statistically significant (GLM, P < 0.01). During MT, activity of the same neuron was classified as B type.
|
|
Movement-related activities classified as B and N types
B-type activities were found during RT and MT, and was found about equally in the PMv and MI (Table 4). Figure 9 shows a representative B-type activity during RT recorded in the PMv, and its activity during MT was also classified as B type. Figure 10 shows quantified data of the neuron in Fig. 9, which were aligned in the visual and motor coordinates. During RT and MT, activities toward the same motor and visual targets differed in the three prism conditions. When activities during RT and MT were compared in the visual and motor coordinates, differences between prism conditions (prism in Eqs. 1 and 2) were statistically significant (GLM, P < 0.01). Thus the activities were classified as B type. We also found activities classified as N type during RT and MT in the PMv and MI (data not shown).
|
|
Number of neurons with each classified type during RT and MT
Tables 4 and 5 summarize the number of neurons with classified types in the PMv and MI of the three monkeys. During RT, the most remarkable distinction between the PMv and MI was the proportion of neurons with V-type activity. The neurons with V-type activity constituted 21% in the PMv but only 1% in the MI (1 neuron). In contrast, M-type neurons were more numerous in MI the (37%) than in the PMv (19%). Proportions of B and N-type neurons were similar in the two areas (35 vs. 36% for B type and 26 vs. 26% for N type).
|
Of the 117 PMv neurons and 76 MI neurons whose activity were classified as the four types during RT, 107 (91.4%) and 69 (90.8%) neurons, respectively, were also active during MT. However, the proportion of the classification types changed greatly from RT to MT. During MT, none of the PMv neurons with V-type activity during RT were classified as V type but changed to the other types. Some neurons with M-type activity during RT was also classified as M type during MT. Others with M-type activity changed to either B (Figs. 7 and 8) or N types, but not to V type, during MT. Table 5 indicates that there was no systematic change from one type of activity during RT to the other type during MT. As a whole, neurons with B-type activity were most numerous both in the PMv (56%) and MI (66%) during MT. No activity type was consistently related to eye movements throughout RT and MT.
For 42 of the statistically analyzed 285 neurons (24 of 119 PMv neurons and 18 of 76 the MI neurons), an optional recording was obtained in the no prism condition at the target locations in Fig. 1, A and C (see METHODS). When the data with optional recording were compared with those without the recording, a majority of the neurons during RT [22 of 24 PMv neurons (91.7%) and 15 of 18 MI neurons (83.3%)] and MT [19 of 21 PMv neurons (90.5%) and 13 of 17 MI neurons (76.5%)] were classified as the same type.
Correlation of the classified activities in the PMv and MI to the visual and motor coordinates
Tables 6 and 7 indicate how closely each type of activity in the PMv and MI was correlated to the visual or motor coordinates, by showing means ± SE of the correlation coefficients (rvisual vs. rmotor calculated from Eq. 3). During RT (Table 6), V-type activities frequently had rvisual values close to 1.0. In the PMv, mean rvisual values of the V-type activities were 0.710-0.865, whereas their mean rmotor values were much less (0.352-0.577). By contrast, M-type activities frequently had rmotor values close to 1.0. Mean values of their rmotor were 0.858-0.767 in the MI and 0.915-0.828 in the PMv, and the values were higher than those of their rvisual (0.552-0.614 in the MI and 0.303-0.780 in the PMv). Common to the PMv and MI, B-type activities frequently had relatively high rvisual and rmotor values (most frequently > 0.5), whereas N neurons had lower rvisual and rmotor values (most frequently < 0.5). Difference in the correlation coefficient (rvisual or rmotor) among the four types of activities was statistically significant particularly in the PMv. These trends were similar during MT (Table 7), although very few V-type activities were recorded even in the PMv.
|
|
Lead times of the movement-related activity before movement onset
We analyzed the lead-time from the onset of the V-, M-, B-, and N-type activities to movement onset. Table 8 shows the mean lead-times of each classified types during RT in the PMv and MI of the three monkeys. In monkeys 1 and 2, the lead-times of the movement-related activities in the PMv were significantly shorter than those in the MI (GLM, P < 0.05). In both the PMv and MI, however, difference in the lead-time between the V, M, B, and N types was not statistically significant (GLM, P > 0.05).
|
Location of neurons
Figure 11 shows the distribution of movement-related neurons, classified by activity during RT, in the PMv and MI of monkey 2. In the MI, the neurons were located mainly in the proximal forelimb and trunk representation areas. In the PMv, on the other hand, the majority of these neurons were located in the caudal part of the PMv, close to the border between the PMv and MI. As shown in Fig. 11, the V-, M-, B, and N-type neurons were similarly distributed in the PMv and MI, and from surface views we could not discern any tendency toward clustering.
|
Figure 12 shows a histological reconstruction of the location of the movement-related neurons in the PMv. Most were located on the dorsal surface of the PMv and within the ventral bank of the arcuate spur. Only a few neurons were recorded in the caudal bank of the arcuate sulcus in the three monkeys (data not shown). The V-, M-, B-, and N-type neurons were recorded at various depths in the same track. In some tracks, the M-type neurons were located deeper than the V-type neurons. In other tracks, however, they were intermingled, and no constant tendency for the location of V-, M-, B-, and N-type neurons was observed.
|
| |
DISCUSSION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Classification of movement-related activity by statistical models
We found several types of movement-related activity in the PMv during RT and MT: activity closely related to head-centered visual space (V type), activity closely related to motor space (M type), and activity reflecting both visual and motor space (B type), and activity reflecting neither of them (N type). Of these four types of movement-related activity, the presence of neurons with V-type activity during RT clearly distinguished the PMv from the MI. The MI had virtually no V type in RT and MT periods. The anterior deltoid muscle, the prime mover of reaching, showed M-type activity during RT. Other muscles showed M, B, and N types during RT and MT, but we did not observe V-type activity in any muscle.
Classification by the statistical models (Eqs. 1 and 2) seems appropriate to distinguish the activity types. In the analysis, the motor coordinates of the targets were used to judge whether the activity was dependent on where the reaching movements were directed. Ideally, the first contact point on the screen should be used to examine this relationship; however, it is impossible to examine such data statistically because the first contact points were not exactly on the targets, but the monkeys tended to hit slightly to the left of the targets (Table 3). Thus it is highly possible that we underestimated the number of M-type activities and overestimated the number of B- and N-type activities. On the other hand, we could statistically verify whether the activities reflected visual space because the targets had the identical location in visual space regardless of the application of prisms.
When the data were obtained optionally at the motor target locations in Fig. 1, A and C, in the no prism condition, Eqs. 1 and 2 in METHODS were more balanced. When the movement-related activities with and without the optional data were compared using the two equations, a vast majority of them were classified as the same type (see RESULTS). Thus we confirmed the validity of Eq. 2, even when the optional data were not recorded. Furthermore, using Eq. 3 in METHODS, V and M types were found highly correlated to the visual and motor coordinates of target location, respectively (Tables 6 and 7). This suggests that V-type activity reflects visual space, whereas motor space is represented in M-type activity. Accordingly, we will first discuss our interpretation of the functional roles of these neurons in reaching motor behavior.
Movement-related activity closely linked to visual space
First, it should be stressed that the V-type activity in the PMv
in this study was not passive response to visuospatial inputs (Boussaoud et al. 1993; Fogassi et al.
1996; Godschalk et al. 1985; Graziano et
al. 1994; Rizzolatti et al. 1981;
Weinrich and Wise 1982). Instead, they were active only
when reaching movements were initiated after visual targets were
presented. Thus we regarded the V-type activity as
"movement-related" (Kurata and Hoffman 1994;
Kurata and Tanji 1986). Similar observations in the PMv were recently reported (Mushiake et al. 1997), but those
activities are considered different from the V type in this study. They
required the monkeys to make visual fixation at one of two points
throughout a trial (Mushiake et al. 1997). During the
trial, a target peripheral to the fixation point was presented. In
their study, they found that the movement-related activity of some PMv
neurons was dependent on the retinocentric location of the targets. In
our study, on the other hand, the monkeys were free to make eye
movements during any phase of a trial. Accordingly, we found
movement-related neurons in the PMv whose activity was dependent on the
head-centered visual location of the targets, and not on their
retinocentric location or the direction of reaching.
It is reported that the PMv contains neurons whose visual and
somatosensory receptive fields remained in the same location even when
the monkey was not fixating and the eyes were moving (Graziano
and Gross 1998). Furthermore, a subset of the neurons continued
to respond in the dark as if the object were still present and visible.
Such cells exhibit "object permanence," encoding the presence of an
object that is no longer visible. It has been suggested that those
cells may underlie the ability to reach toward or avoid objects that
are no longer directly visible (Graziano et al. 1997a).
We propose that the V-type activity observed in this study reflects
motor commands directly converted from either retinocentric or
head-centered visuospatial information, represented by the neuronal
responses to visual stimuli in the PMv (Boussaoud et al.
1993; Fogassi et al. 1996; Graziano et
al. 1994; Rizzolatti et al. 1981) and
intraparietal cortex (Colby et al. 1993, 1995) that
project to the PMv (Kurata 1991; see also a review by
Wise et al. 1997). This should be clarified in future
experiments. Because few V-type activities were recorded in the MI, it
is also suggested that direct spatial conversion from visual input to motor output occurs in the PMv but much less in the MI.
Because visual and motor space were dissociated in our experiment
design, it can be interpreted that the V-type activity represents an
extrinsic frame, related to the direction of movement in visual space,
and that the M-type activity reflects an intrinsic frame, related to
the direction of movement in motor space. Our results are comparable to
the reports by Kakei et al. (1999, 2001) who dissociated
different coordinate frames related to wrist movements: extrinsic
(related to the direction of movement in space) and intrinsic (related
to the activity of individual or groups of muscles and related to the
angle of the wrist joint) frames. They found that nearly all of the
movement-related neurons in the PMv were "extrinsic-like"
(Kakei et al. 2001), whereas the MI contains both
"muscle-like" and extrinsic-like activities (Kakei et al. 1999). Although the proportion of extrinsic and "intrinsic"
activities in the PMv and MI seems different from our results, it can
be explained by difference in experiment designs. It is more important to notice that the PMv contains neurons representing an extrinsic frame, thus contributing to transformation of coordinates.
Movement-related activity possibly representing final motor commands
In contrast to V type, M-type activities were frequently recorded
in both the PMv and MI, and their activity changed in association with
target location or the direction of reaching toward them but not with
the visuospatial target location. These observations suggest that the
activity of the M-type activity reflects final motor commands to lower
motor centers such as the spinal cord, to which both the PMv and MI
project directly (Dum and Strick 1991; He et al.
1993). We did not find any distinction between the M-type
activity in the PMv and MI. However, it is possible that output from
the neurons with M-type activity in the PMv may be sent to the MI
(Leichnetz 1986; Muakkassa and Strick
1979), where it is further processed to make muscle activation
patterns within the MI.
Movement-related activity other than V and M types
Besides the V and M types, we recorded a number of movement-related neurons in the PMv and MI that did not show a close relationship to visual or motor coordinates (B and N types). The presence of the neurons can be interpreted in four ways. First, they could convey information on motor space or final motor commands because similar activities were found in EMG of the upper arm and shoulder (Table 3), i.e., in activity during MT. Second, the neurons may represent an intermediate stage of processing that is necessary for coordinate transformation, and the PMv may modify the relationship between the V and M types by changing the activity of the B and N types. Third, they might represent noisy elements of neuronal activity and play no specific role in the transformation of coordinates or in generating the final motor commands necessary for directing the arm toward the target. Fourth, the neurons might contribute to other aspects of motor control rather than those required in our task. Because the design of our study did not enable us to identify which interpretation is most appropriate, other studies will be necessary to specify the functional role of the B- and N-type activities.
| |
ACKNOWLEDGMENTS |
|---|
We thank Prof. J. Tanji for continuous support and advice on this study. Dr. S. P. Wise provided critical advice on an early version of this manuscript.
This work was supported by the Ministry of Education, Science, and Culture of Japan (Grants 08279204, 08680878, 09268204, 10164205, 11145203, and 12210025), the "Research for the Future" Program (Grant 96L00206) of the Japan Society for the Promotion of Science, and the Cooperation Research Program of Primate Research Institute, Kyoto University.
Present address of E. Hoshi: University of Pittsburgh, Department of Neurobiology, W1640 Biomedical Science Tower, 200 Lothrop Street, Pittsburgh, PA 15261.
| |
FOOTNOTES |
|---|
Address for reprint requests: K. Kurata, Department of Physiology, Hirosaki University School of Medicine, Hirosaki 036-8562 Japan (E-mail: kuratak{at}cc.hirosaki-u.ac.jp).
| |
REFERENCES |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
) on the CRT screen (the
large rectangles) in the left (A), no (B), and
right (C) prism conditions. Note that the target locations
are the same in the visual coordinates, regardless of the prism
condition (D). The large rectangles in A-C, E
indicate the edge of the screen and touch panel. The central target in
the visual coordinates (V08) was aligned at the intersection of the
vertical and horizontal meridians. There were 15 motor targets
(M01-15) in the motor coordinates (E, see text).
motor, in
Eq. 
