Extracellular discharges from single neurons in the internal segment of the globus pallidus (GPi) were recorded and analyzed for rate changes associated with visually guided forearm rotations to four different targets. We sought to examine how GPi neurons contribute to movement preparation and execution. Unit discharge from 108 GPi neurons recorded in 35 electrode penetrations was aligned to the time of various behavioral events, including the onset of cued and return movements. In total, 39 of 108 GPi neurons (36%) were task-modulated, demonstrating statistically significant changes in discharge rate at various times between the presentation of visual cues and movement generation. Most often, strong modulation in discharge rate occurred selectively during either the cued (n = 32) or return (n = 2) phases of the task, although a few neurons (n = 5) were well-modulated during both movement phases. Of the 34 neurons that were modulated exclusively during cued or return movements, 50% (n = 17) were modulated similarly in association with movements to any target. The remaining 17 neurons exhibited considerable diversity in their discharge properties associated with movements to each target. Cued phases of behavior were always rewarded if executed correctly, whereas return phases were never rewarded. Overall, these data reveal that many GPi neurons discharged in a context-dependent manner, being modulated during cued, rewarded movements, but not during similar self-paced, unrewarded movements. When considered in the light of other observations, the context-dependence we have observed seems likely to be influenced by the animal's expectation of reward.
Differing theories of globus pallidus (GP) function have emerged from previous studies of behaviorally associated GP neuronal activity. Some studies have indicated that these neurons carry information related to movement kinematics (Georgopoulos et al. 1983; Mitchell et al. 1987; Turner and Anderson 1997). Others have provided evidence that the GP (internal and external segments) may mediate cognitive aspects of movement such as multiple-segment movement sequencing (Mushiake and Strick 1995) or movement preparation (Brotchie et al. 1991a; Georgopoulos et al. 1983; Jaeger et al. 1993). All of these studies provide evidence for the great diversity of discharge patterns exhibited by GP neurons in different behavioral conditions.
The variety in task-related discharge properties is consistent with evidence supporting the presence of anatomical projections from multiple cortical areas (Alexander et al. 1986;Strick et al. 1995) that carry information through the basal ganglia. The degree of convergence of cortical inputs within the striatum has not yet been definitively demonstrated. Strick's findings support a theory in which discrete cortical areas project to topographically distinct regions of the striatum and GP (Strick et al. 1995). Yet, other anatomical studies have provided evidence for convergence of functionally related cortical inputs within the striatum (Flaherty and Graybiel 1993;McFarland and Haber 2000; Parthasarathy et al. 1992; Yeterian and Van Hoesen 1978). Nonetheless, there is general agreement that neighboring striatopallidal efferent populations project to functionally specific populations of pallidal neurons with adjacent terminal arborizations that do not intermix (Parent et al. 1997). Likewise, globus pallidus internal segment (GPi) efferents project to thalamic neuron populations that, in turn, project to single cortical neuron populations, as demonstrated by thalamic labeling following cortical injection of transneuronal virus (Middleton and Strick 1997).
In addition to the task-related properties that may be a consequence of their cortical inputs, GP neuronal responses may potentially be influenced by another variable. Substantia nigra pars compacta (SNpc) neurons release dopamine onto frontal cortical neurons and striatal medium spiny neurons which has been reported to modulate the activity of striatal spiny neurons (Akaike et al. 1987;Kiyatkin and Rebec 1996). These SNpc neurons have been demonstrated to provide signals during behavior that may be related to the prediction of future events, progress through a behavioral task, or the likelihood of reward (Hollerman et al. 1998;Kawagoe et al. 1998; Schultz 1998).
Owing to the fact that they receive cortical and dopaminergic afferents, striatal neurons may convey information to the pallidum that reflects the processing of both types of influences. Consequently, the activity of a given pallidal output neuron may be related to motor or associative aspects of behavior, but could also convey information about progression through a task or potential for a reward. This reward relationship has been ascribed to striatal neurons having firing patterns sensitive both to eye or to limb movements andreward schedule. For instance, some caudate neuron responses have been described that appeared to be contingent on behavioral cues that were predictive of the prospect of reward if targeted saccadic eye movements were executed correctly (Kawagoe et al. 1998). In another experiment, visual signals were presented within a predictable trial sequence providing contextual cues that could be used to determine the appropriate motor response in a go, no-go arm movement task (Schultz et al. 1998). Putamen neuron discharge was enhanced in both go and no-goconditions if contextual cues were predictive of reward. Movement-associated modulation was completely absent if there was no possibility of reward, suggesting that it was not exclusively the movement, but also the behavioral context that determined the neuronal discharge pattern.
While many GPi studies have examined neuronal responses from trial onset through movement completion and reward, few have contrasted neuronal discharge during movements executed in different behavioral contexts. Therefore the present study compared cued, rewarded movements, and similar self-paced, unrewarded movements to determine whether the associated discharge was context dependent.
Single unit extracellular recordings were obtained in the GPi of an awake, behaving adult female macaque monkey in accordance with guidelines established by the NIH and the Northwestern Institutional Animal Care and Use Committee. Tungsten microelectrodes (FHC, Bowdoinham, Maine) were inserted into a recording chamber oriented 45° from vertical in the coronal plane. Recording location was determined using published coordinates from other laboratories (DeLong and Georgopoulos 1979; DeLong et al. 1985; Horak and Anderson 1984) that were adjusted for this monkey using fiducial markings in preoperative MR images. For each penetration, single units were discriminated and identified as putamen or external and internal pallidal neurons (GPe and GPi, respectively) based on standard criteria [electrode depth, firing rate, incidence of pausing (Anderson 1977;DeLong 1971)].
The monkey faced a white screen containing five light emitting diodes (LEDs) positioned at 35° increments about an arc (shown schematically in Fig. 1, top) and grasped a handle that rotated ±100° from vertical during forearm supination and pronation. A handle-mounted laser projected onto the screen, providing continuous visual feedback about handle position, both during cued and self-paced movements. Trial onset was designated by briefly lighting all five LEDs, after which only the 0° target remained lit. Once the animal held the handle within ±8° of the lit 0° target for a fixed 500 ms duration, one of the other LEDs was randomly selected and lit. After a variable hold period (0–1000 ms), a long duration (440 ms) tone triggered the monkey to rotate the handle to the instructed LED. The animal was required to move to within ±8° of the target within 2 s of the movement trigger. On target acquisition, both LEDs were turned off. At the completion of a fixed 500 ms hold period in the instructed target window, a liquid reward was delivered. Following reward delivery, the animal usually returned the handle to the center target with a self-paced movement prior to the lighting of five LEDs at the onset of the next trial. Comparison of the two behavioral contexts in which movements were generated shows that cued movements were triggered by an auditory tone and directed to lit LED targets, while return movements were not triggered by a tone and were directed to unlit targets (Fig. 1, top). The cued movements always started from a single, central location, while return movements were initiated from any of the four off-center targets. Furthermore, correct movements to the instructed target elicited a liquid reward, while return movements to the fixation target were unrewarded. Data were collected continuously over 10–12 min intervals.
For each neuron, data from successful trials were grouped according to the instructed target. Rasters and cumulative histograms were generated by aligning neuronal discharge with respect to the time of a variety of behavioral events (trial onset, target on, tone on, target off, and reward) and to the onset of both cued and return movements. Movement onset was defined as the point at which the derivative of the handle position signal (velocity) exceeded a fixed threshold of three times the standard deviation of the noise floor of the velocity trace. The maximum rate change (burst or pause, or both in the case of biphasic responses) within 500 ms of the trigger event was determined for each histogram. The mean discharge was determined in a 250-ms window around these time points. Modulation in discharge rate during the bursts or pauses associated with the cued movement was compared to the mean rate in the period 1000–750 ms before cued movement onset. Modulation in discharge rate during the return movements was compared to the mean rate in the period 750–1000 ms after return movement onset. Neurons were included in subsequent analyses if the event-related discharge differed significantly from baseline (small-sample test statistict for the difference between two means, P < 0.05). In some cases (8/39), the change in discharge occurred after the presentation of the visual cue, but before movement onset, and could not be definitively attributed to either behavioral event. Neurons were subsequently classified according to the magnitude of phasic modulation in conjunction with cued and return movements. If phasic modulation with similar statistical significance were detected during both the cued and return movements, the neuron was classified as “context-independent.” Alternatively, strong phasic modulation associated exclusively with either cued or return movements led us to categorize the neuron as “context-dependent.”
Surface electromyograms (EMGs) were collected during some of the final GPi recording sessions from triceps, biceps, brachioradialis, pronator teres, and flexor carpi radialis muscles. Skin surfaces were shaved and cleaned with alcohol prior to application of bipolar, surface EMG electrodes whose contacts were lightly coated with conductive gel. EMG signals were amplified and high-pass filtered at 75 Hz. During most experiments, behavioral signals were all sampled at 200 Hz; however, when EMGs were collected, all signals were sampled at 1000 Hz. EMG signals were subsequently rectified and smoothed.
During the final recording session, histological lesions (50 μA for 20–30 s) were placed in several locations within the recording chamber using stainless steel microelectrodes (FHC). After lesioning, the monkey received a lethal dose of sodium pentobarbital and was perfused with physiological saline followed by 4% paraformaldehyde and 1% potassium ferrocyanide. The brain was removed, blocked, cryoprotected, and sectioned at 50 μm. Sections were mounted and stained with thionin and neutral red. Electrode penetrations were reconstructed and the distribution of neurons in anterior-posterior, medial-lateral, and dorsal-ventral dimensions was plotted. To summarize this distribution, the extent of the nucleus along each dimension was divided in half, and the cell was assigned to one half or the other.
In total, 108 GPi neurons were isolated during 35 electrode penetrations, of which 35% (39/108) were task-modulated, exhibiting statistically significant changes in discharge rate at any time between the presentation of visual cues and end of movement. Representative responses from three well-modulated GPi neurons during the execution of the task are shown in Fig. 1. Neuron B181 (Fig.1 A) paused phasically in association with cued movements to all four targets (all P < 0.0001, small sample test statistic t for difference between two means). The depths of modulation of each of the pauses were similar despite differences in movement amplitude and direction, although the latency of the pauses relative to movement onset differed with movement direction. These observations suggest that discharges from this neuron were not strongly associated with movement kinematics per se. Modulation, which was clearly evident during cued movements in either direction, was notably absent during both clockwise and counterclockwisereturn movements (all P > 0.05). Behavioral context affected the discharge pattern of neuron B251 in a similar fashion (Fig. 1 B). In contrast to the pausing ofneuron B181, however, this unit burst phasically in association with cued movements (all P < 0.01), but did not burst during the return movements (all P > 0.05). Like many of the neurons in this study (n = 17),neurons B181 and B251 were modulated similarly during cued movements to all four targets (n = 17). In contrast, neuron B121 is more representative of the task-modulated neurons that exhibited considerable diversity in their discharge properties associated with movements to different targets (n = 17). Unit B121 (Fig. 1 C) burst prior to the cued −70° (P < 0.0001) and 35° (P < 0.010) movements and paused during the execution of the −70° movement (P < 0.020). It was unmodulated with all other cued and return movements (P> 0.05).
Most of the task-related neurons were also modulated in a manner that depended on the context, i.e., strong, phasic modulation occurred only in conjunction with rewarded, cued movements (32/39) or in a few units, only with unrewarded, return movements (2/39). A small fraction of GPi neurons (5/39) were strongly modulated during both cued and return portions of the task and were considered to have a context-independent discharge pattern. Half of the context-dependent neurons (17/34) were similarly modulated in association with movements to all four targets as in Fig. 1, A and B. The remaining context-dependent neurons (n = 17) were preferentially modulated in association with movements to specific targets as in Fig.1 C.
The context-dependent discharge shown in Fig. 1 was derived from comparison of cued and return movements that were in opposite directions. To eliminate this factor as an explanation for the disparate neuronal discharge observed in the two conditions, the discharge associated with cued and return movements of thesame amplitude and direction was examined as shown in Fig.2. This figure shows the discharge patterns associated with cued and return movements to the −70° target (A, B) and +70° target (C,D) for four neurons. Unit discharge during cued (left column in A, B, C, and D) and return (middle column) movements is presented as in Fig. 1. In addition, this figure includes return movements from a different target that were of the same amplitude and direction as the cued movement (right column, kinematically similar). Although all neurons shown were well-modulated during the cued movements (P< 0.002), none were strongly modulated with any of the return movements (P > 0.20), except D(P = 0.044) or kinematically similar return movements (P > 0.30), except B (P = 0.010). Although these neurons were modulated in association with both cued and return movements to some targets (P < 0.05), the differences in the magnitude of the significance during these movements led us to characterize them as context-dependent. One would expect that the cued and kinematically similar movements might have elicited similar responses; however, the lack of modulation during either the actual return, or the kinematically similar return movements clearly excludes directional preference as an explanation for the differences in modulation for any given neuron with context-dependent discharge properties. Thus even when kinematically similar movements were compared, most (34/39) task-modulated neurons responded selectively during the cued or return movement but not both. Consequently, these data support the hypothesis that most GPi neurons are selectively modulated within specific contextual settings.
Figure 3 shows the discharge of a GPi neuron during cued and kinematically similar counterclockwise 70° movements along with position, velocity, and EMG signals which were aligned with respect to the onset of movement and ensemble averaged. This neuron was strongly modulated in association with the cued counterclockwise movements of both 35° and 70° amplitudes (P < 0.0001); however, only the 70° movement is shown for simplicity. Similar to the neurons in Fig. 2, this neuron was modulated at a level that exceeds the statistical threshold to be considered task-modulated (P < 0.05). However, the magnitude of the modulation during the return movements (P > 0.03) was considerably smaller than that during the cued movements. Comparison of the cued and return movements from the ±70° targets revealed some small differences in muscle activation patterns that likely reflect differences in agonist/antagonist muscle activation during execution of these two opposing movements (data not shown). However, comparison of EMG activity during the cued movements and kinematically similar return movements (e.g., −70° cued movement compared with the return from the +70° target as shown in Fig. 3) illustrates that muscle activation patterns were quite similar.
Reconstruction of lesions and recording positions indicated that neurons were sampled from the full extent of GPi, with the exception of the ventralposteriomedial region of the nucleus. The distribution of neurons that were modulated and those that were not modulated is shown in Table 1. While most task-modulated neurons were encountered in the anterior half of the nucleus (31/39), they were evenly distributed along the dorsal-ventral (18 dorsal, 21 ventral) and medial-lateral (19 medial, 20 lateral) axes. There was no apparent clustering of neurons along any axis with respect to context-dependence. However, the context-dependent neurons that exhibited modulation primarily in association with presentation of visual cues (8/39) tended to be clustered in the medial and anterior portion of the nucleus.
In this study, neuronal discharge during cued andreturn phases of a movement task was compared. Most task-modulated GPi neurons exhibited strong, phasic changes in discharge rate in association with cued, but not return movements. Cued movements were instructed by lit targets, triggered by a tone, and always rewarded if executed correctly. Return movements were usually executed while targets were off, were self-paced, and were never rewarded. Thus, many GPi neurons seemed to be selectively modulated during movements generated in a specific behavioral context. While small differences in the EMG patterns during cued and return movements were observed as shown in Fig. 3, these were unlikely to provide an explanation for such dramatic differences in neuronal activity during the two phases of movement. Although we were unable to exclude the influence of the other contextual cues on neuronal discharge using this behavioral paradigm, we speculate that the monkey identified the context for each movement, and that reward expectation was the most important variable influencing GPi neuronal activity. Previous reports of GPi neurons have described a variety of response patterns that appeared to be influenced by variables of target dependence (Mushiake and Strick 1995), movement duration (Anderson and Turner 1991), movement direction (Georgopoulos et al. 1983; Turner and Anderson 1997), movement amplitude, or velocity preference (Anderson and Turner 1991; Georgopoulos et al. 1983). Consistent with these previous reports, the task-modulated neurons in this study also exhibited great diversity in their response patterns that may reflect the influence of some of these variables. However, the neurons studied in this report appeared to be influenced by an additional contextual variable that contributed to the discharge patterns by determining whether the unit will be strongly modulated in association with cued movements, return movements, or both.
Hypothetically, GPi receives this contextual information via striatal inputs and conveys it to thalamic neurons via specific channels through the basal ganglia, thereby encoding information about motivation or the potential for reward. Since most task-modulated striatal neurons seem to be influenced by the expectation of reward (Hollerman et al. 1998), one might expect many GPi neurons to be associated with reward-related context detection as well. This may be especially true for GPi neurons receiving input from neurons in the ventral striatum that were shown to carry signals related to progression through a task, or proximity to reward (Shidara et al. 1998). The fact that the majority of task-modulated neurons responded to both movement-related variables and behavioral context is one of the most striking observations in the present study. This may reflect convergence of movement and motivation-related signals at the striatal level, since neurons that appear to carry information about movement preparation or initiation also appear to encode information about behavioral context. Alternatively, it may reflect the prominence of striatopallidal projections that are related to goal attainment, or even the importance of reward in the areas of prefrontal cortex that provide inputs to striatal and GPi neurons.
Other laboratories have observed context-dependent firing behavior in pallidal neurons. GPi neurons have been described that were modulated differently in association with sequential wrist flexion and extension movements (Brotchie et al. 1991b) depending on whether or not the next movement was predictable. These findings suggest that GPi neurons may indeed participate in the process of behavioral context encoding. Since the animal may have been more likely to prepare for and successfully complete a predictable movement than an unpredictable one, this may further reflect reward expectation. Caudate and pallidal neurons were studied during flexion/extension hand movements made either to an instructed target, or as corrective movements following torque step perturbations (Aldridge et al. 1980). Analogous to the findings of Brotchie et al., neuronal responses for similar movements differed as a function of the behavioral context in which the movements were generated, suggesting that both striatal and pallidal neurons participate in this process of context encoding.
Context-dependent discharge patterns in GPi could possibly be associated with other aspects of behavior. Some channels through the basal ganglia, for example those receiving input from prefrontal cortical neurons, are likely to participate in the sequencing of multiple-segment movements. Prefrontal cortical neurons have been described (Barone and Joseph 1989; Tanji et al. 1996) that were selectively modulated during specific segments of a movement sequence. These neuronal responses were sometimes specific to the order in which a movement segment occurred within a sequence. Similarly, Mushiake and Strick (1995)described pallidal neurons that responded during specific segments of a three-segment arm movement task, most frequently when generated from memory. The authors suggest that this may be evidence for GPi participation in the encoding of “spatiotemporal characteristics of a sequential movement” (Mushiake and Strick 1995). These pallidal neurons were presumed to be located outside of the primary motor area of the pallidum, potentially in an area receiving prefrontal or supplementary motor cortical inputs. Hence, an alternative explanation for the context-dependent modulation that we observed may be that the neurons were encoding either the initial or return movement segment within the complete trial sequence. However, the paucity of neurons responding exclusively during return movements (2/39) suggests that this is not the case.
Observations suggest that neurons in the pars reticulata of the substantia nigra (SNpr), the other output nucleus of the basal ganglia, participate in the encoding of behavioral cues used for task completion. Hikosaka and Wurtz (1989) described saccade-related neurons with responses that varied as a function of the behavioral context in which the saccade was generated. Recently, SNpr discharge patterns associated with saccadic eye movements were shown to vary systematically with the reinforcement associated with the behavioral contexts in which the saccades were generated (Handel and Glimcher 2000). Owing to these findings in the nuclear homologue of GPi, the discharge patterns that we observed during visually guided forelimb movement are not unexpected and may reflect progression through a behavior or an anticipation of reward. This SNpr finding is consistent with recent reports of caudate neuron modulation during presentation of targets to which correctly executed eye movements would be rewarded (Kawagoe et al. 1998). While some investigators have suggested that the basal ganglia are mainly involved in encoding movement kinematics, the current study provides evidence that they are additionally involved in context encoding. GPi neuronal discharge may be modulated by the context in which movements are generated, in particular, information about movement sequence or reward prediction. Experiments are in progress to further characterize the salient aspects of the behavior which influence the firing patterns of GPi neurons in specific behavioral contexts.
The authors thank A. Gruber for valuable discussions of this work and K. Novak for assistance with behavioral software development.
This work was supported by the Ruggles Fellowship in Movement Disorders, Department of Neurology, Evanston Hospital, Evanston, IL, and by a National Institute of Mental Health Center Grant (MH-48185-09), J. C. Houk, Center Director.
Address for reprint requests: L. E. Miller, Northwestern University Medical School, Dept. of Physiology, 303 E. Chicago Ave., Chicago, IL 60611 (E-mail:).
- Copyright © 2001 The American Physiological Society