HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) All Versions of this Article: 92/6/3482    most recent 00547.2004v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (27) Citing Articles via Google Scholar Google Scholar Articles by Hoshi, E. Articles by Tanji, J. Search for Related Content PubMed PubMed Citation Articles by Hoshi, E. Articles by Tanji, J.

Differential Roles of Neuronal Activity in the Supplementary and Presupplementary Motor Areas: From Information Retrieval to Motor Planning and Execution

¹Department of Physiology, Tohoku University School of Medicine, Sendai 980-8575; and ²The Core Research for Evolutional Science and Technology Program, Japan Science and Technology Agency, Kawaguchi 332-0012, Japan

Submitted 26 May 2004; accepted in final form 20 July 2004

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
We explored functional differences between the supplementary and presupplementary motor areas (SMA and pre-SMA, respectively) systematically with respect to multiple behavioral factors, ranging from the retrieval and processing of associative visual signals to the planning and execution of target-reaching movement. We analyzed neuronal activity while monkeys performed a behavioral task in which two visual instruction cues were given successively with a delay: one cue instructed the location of the reach target, and the other instructed arm use (right or left). After a second delay, the monkey received a motor-set cue to be prepared to make the reaching movement as instructed. Finally, after a GO signal, it reached for the instructed target with the instructed arm. We found the following apparent differences in activity: 1) neuronal activity preceding the appearance of visual cues was more frequent in the pre-SMA; 2) a majority of pre-SMA neurons, but many fewer SMA neurons, responded to the first or second cue, reflecting what was shown or instructed; 3) in addition, pre-SMA neurons often reflected information combining the instructions in the first and second cues; 4) during the motor-set period, pre-SMA neurons preferentially reflected the location of the target, while SMA neurons mainly reflected which arm to use; and 5) when executing the movement, a majority of SMA neurons increased their activity and were largely selective for the use of either the ipsilateral or contralateral arm. In contrast, the activity of pre-SMA neurons tended to be suppressed. These findings point to the functional specialization of the two areas, with respect to receiving associative cues, information processing, motor behavior planning, and movement execution.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
The supplementary (SMA) and presupplementary (pre-SMA) motor areas, which are located in area 6 on the medial wall of the frontal cortex of primates, are separate motor areas with distinct structural and functional properties (Picard and Strick 1996; Tanji 1996). Anatomical studies revealed differences in the connectivity of the two areas. Only the SMA has direct connections with the primary motor cortex and descending output to the spinal cord (Dum and Strick 1991; Macpherson et al. 1982; Maier et al. 2002), whereas the pre-SMA receives afferents from the dorsolateral prefrontal cortex (Lu et al. 1994; Luppino et al. 1993). Thalamic projections to the two areas come from largely separate areas (Matelli and Luppino 1996). The SMA is organized somatotopically in the caudal to rostral direction (Matsuzaka and Tanji 1996; Tanji 1994). Intracortical microstimulation evokes body movements, allowing construction of a somatotopic motor map (Luppino et al. 1991; Mitz and Wise 1987). In contrast, for the pre-SMA, microstimulation of this area had complex effects involving the forelimb (Luppino et al. 1991). Ablation of the SMA produces specific deficits in bimanual coordination (Brinkman 1984) or internally guided or instructed movements (Chen et al. 1995; Kazennikov et al. 1998; Kermadi et al. 1997; Thaler et al. 1995), whereas chemical inactivation causes the failure of the sequential execution of multiple movements (Shima and Tanji 1998). Clinical studies reported comparable impairments with SMA lesions: the failure of sequential motor performance (Laplane et al. 1977) or a reduction in spontaneous movements (Krainik et al. 2001). On the other hand, lesions of the pre-SMA caused deficits in updating sequential movements (Shima and Tanji 1998) and the acquisition of sequential procedures (Nakamura et al. 1999). Brain imaging studies in humans have implicated the pre-SMA in motor-task learning (Friston et al. 1992; Hikosaka et al. 1996) and in performing motor tasks with higher cognitive demands (Deiber et al. 1991; Sergent et al. 1992; Zatorre et al. 1994).

Studies examining single-cell activity have pointed to the different roles these two areas play in controlling motor behavior. First, it was found that SMA neurons exhibited marked activity during an arm movement, whereas pre-SMA neurons were more active in response to visual signals or during a preparatory period (Matsuzaka et al. 1992). Second, with a motor task requiring subjects to switch the direction of forthcoming reaching movement, neurons that were selectively active when shifting the direction of action occurred more often in the pre-SMA than in the SMA (Matsuzaka and Tanji 1996). Third, neuron activity selectively associated with capturing a spatial target with either a saccade or an arm reach, i.e., neurons involved in effecter-independent reaching, was more frequent in the pre-SMA (Fujii et al. 2002). Finally, with sequential movement tasks, it was found that 1) SMA neurons were more active when memory guided a sequence of movements (Mushiake et al. 1991; Tanji and Shima 1994), whereas pre-SMA neurons were more active when visual signals guided the sequence (Halsband et al. 1994); 2) neuron activity during the linking of two different movements was found more often in the SMA (Shima and Tanji 2000; Tanji 2001), whereas neurons selectively active when updating sequential order were more frequent in the pre-SMA (Shima et al. 1996); 3) neurons selective for the numerical order were more frequent in the pre-SMA (Clower and Alexander 1998; Shima and Tanji 2000); and 4) neuronal activity during the acquisition of sequential movements were more abundant in the pre-SMA (Nakamura et al. 1998; however, see Lee and Quessy 2003).

Despite these reports, more studies of neuronal activity relevant to other aspects of behavioral processes are still necessary. Recent brain imaging studies in humans suggest that the pre-SMA participates in another aspect of behavioral control. Two studies reported that the pre-SMA was active in conditional motor behavior in which auditory or visual signals were associated with the selection of finger movements (Kurata et al. 2000; Sakai et al. 2000). These reports suggest that the pre-SMA, but not the SMA, is involved in mapping sensory signals to movements (Picard and Strick 2001), inviting studies to analyze how sensory signals are received, processed, and transformed into information useful for motor selection at the single-cell level. On the other hand, the nature of preparatory activity preponderant in both the SMA and pre-SMA remains to be clarified (Matsuzaka et al. 1992; Tanji 1996).

To address these issues, we devised a behavioral task that separated each step in the information processing of the visuomotor transformation required to associate visual signals with movements. In the initial part of the task, two instructions indicating arm use and target location were given successively with a delay between them, requiring the subjects to detect sensory signals and to extract necessary information. Subsequently, information had to be integrated to generate information to plan forthcoming actions. Then, a motor-set period was introduced to sort the preparatory processes before initiating a reaching movement in response to a trigger signal. We will show that neurons in the SMA and pre-SMA exhibit differential properties with respect to receiving associative cues, retrieving and integrating information, planning motor behavior, and executing movements.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Animals and apparatus

We studied two male monkeys (Macaca fuscata, 8 kg) that were cared for in accordance with the National Institutes of Health guidelines and the Guidelines for Animal Care and Use published by our institute. The two monkeys were also used in previous studies (Hoshi and Tanji 2000, 2002, 2004). During the experimental sessions, each monkey sat in a chair with its head restrained. We installed two touch-pads (17 cm apart) in front of the chair, and a color monitor equipped with a touch-sensitive screen was placed in front of the monkey (30 cm from its eyes). Eye positions were monitored with an infrared eye-camera system (R-21C-AS, RMS, Hirosaki, Japan). Neuronal activity was recorded with glass-insulated Elgiloy-alloy microelectrodes (1 2 M at 333 Hz), which were inserted through the dura mater using a hydraulic microdrive (MO-81, Narishige, Tokyo, Japan). Single-unit potentials were amplified with a multichannel processor and sorted using a multispike detector (MCP plus 8, MSD, Alpha Omega Engineering, Nazareth, Israel). EMG activity was recorded with silver wire electrodes. The EMG activity was amplified and digitized with an A/D converter, and the digital values were stored in a laboratory computer. The TEMPO/Win system (Reflective Computing, St. Louis, MO) controlled the behavioral task and saved data for off-line analysis.

Behavioral task

The monkeys were trained to perform a target-reach task by following two sets of instructions, one indicating the target location and the other indicating which arm to use to reach for the target (Fig. 1A). The task commenced when the monkey placed a hand on each touch-pad, after an inter-trial interval of 3 s, and gazed at a fixation point (FP) 1.2° in diameter that appeared at the center of the touch-sensitive screen. If fixation was maintained for 1,200 ms, the monkey was given the first instruction (1st cue, 400-ms duration), which contained information about either the target location or which arm to use. A small, colored cue that was superimposed on the central FP indicated the type of instruction, i.e., whether the instruction was related to the target location or the arm to use. For monkey 1, a green circle or red square indicated the instruction for arm use, whereas a blue circle or red cross indicated the instruction for target location. For monkey 2, a green square and blue cross indicated the instruction for arm use and target location, respectively. A white square (8 x 8°) that appeared to the left or right of the FP, appearing at the same time as the colored cue, indicated the laterality of arm use (for the arm instruction) or target location (for the target instruction). If fixation was maintained for 1,200 ms during the subsequent delay period (1st delay), the second instruction (2nd cue, 400 ms) was given to complete the information for the subsequent action. Thereafter, if fixation was maintained for 1,200 ms during the second delay, squares appeared on each side of the FP (set cue, 1,000 ms), telling the monkey to get ready to reach for the target when the FP disappeared (the "GO" signal). If the monkey subsequently reached for the target with a reaction time of <1 s, it was rewarded with fruit juice. Before the GO signal appeared, monkey 1 was required to fixate on the FP for 800 1,200 ms. The order of appearance of the target and arm instructions was alternated in a block of 20 trials, and laterality was randomized within each block. A series of five 250-Hz tones after a reward signaled reversal of the order of instructions.

View larger version (25K): [in this window] [in a new window]   FIG. 1. Behavioral task and recording site. A: temporal sequence of the behavioral events. Top: trial in which 2 instructions were given, namely, which arm to use ("arm") and which target to reach ("target"), in that order. Bottom: trial in which the 2 instructions were given in reverse order. B: top view of the surface of the frontal cortex showing the recording sites. Neuronal data were recorded in the presupplementary motor areas (pre-SMA; red) and arm area of the supplementary motor areas (SMA; blue) in the right hemisphere. C: coronal section of the pre-SMA along the dotted line labeled C in B. D: coronal section of the arm region in the SMA along the dotted line labeled D in B. PS, principal sulcus; AS, arcuate sulcus; SPS, superior precentral sulcus; CeS, central sulcus; CiS, cingulate sulcus. Scale bars, 5 mm.

Definition of Task-related Neurons and 10 Task Periods.    We sampled all neurons that were monitored during at least four blocks of trials (i.e., 80 trials). For the purpose of defining neuronal activity as task-related, we initially divided the behavioral task into the following six periods: 1) control, 200 700 ms after attaining fixation; 2) prefirst cue, the 500-ms period before the first cue appeared; 3) first cue and delay, from 100 ms after the first cue onset until onset of the second cue; 4) second cue and delay, from 100 ms after onset of the second cue until onset of the set cue; 5) set cue, from onset of the set cue until the GO signal appeared; and 6) movement, the 500-ms period around the time when movement started. We classified a neuron as "task-related" if the distribution of the discharge rate (spikes/s) was significantly different in at least one of eight trial types (ANOVA, P < 0.05, repeated over 8 trial types with 8 sequences of the 1st and 2nd cues). For the purpose of statistical analysis and display, data were aligned separately for the five task events (onsets of the 1st and 2nd cues, the set cue, GO, and the time of screen touch). These data were analyzed separately before being merged at the midpoint of the first and second delays and of the set-cue period (i.e., 600 ms after cue offset and 600 ms before the onset of the 2nd cue or the set cue, and 600 ms after the set-cue onset and 600 ms before the GO signal).

Subsequently, for the purpose of examining properties of neuronal activity with statistical analyses, we divided the total task phases into one "control period" (200–700 ms after attaining fixation) and 10 "task periods," as follows: 1) precue, 500-ms period before the onset of the first cue; 2) first cue, 100 500 ms after the onset of the first cue; 3) early first delay, 500 1,000 ms after the onset of the first cue; 4) late first delay, last 500 ms before the onset of the second cue; 5) second cue, 100 500 ms after the onset of the second cue; 6) early second delay, 500 1,000 ms after the onset of the second cue; 7) late second delay, last 500 ms before the onset of the set cue; 8) early set-cue, 500-ms period after the onset of the set cue; 9) late set-cue, 500 ms before GO appeared; and 10) movement, 500-ms period before the screen was touched.

Statistical Analysis Using Interspike Intervals.    To analyze neuronal activity with high temporal resolution, we first calculated the instantaneous firing rate as the inverse of the interspike interval (inverse-ISI, 1-ms resolution). Since the rate of neuronal discharge tended to follow a Poisson distribution, the inverse-ISI data were square-root-transformed to stabilize the variance (Zar 1999).

To estimate how neuronal activity reflected information contained in the first, second, or both cues, we used a one-way ANOVA. We examined how well each of the following formulas expressed neuronal activity

(1)

(2)

(3)

In these formulas, the firing rate index is for the transformed inverse-ISI data that were sampled every 10 ms, ₀ is the intercept, and a, b, and c are coefficients. The categorical factors for the first and second CUE are the four instructions provided in the cues (right arm, right target, left arm, and left target). The categorical factors for COMBINATION are the four possible combinations of arm use and target location given by the first and second cues. First, we calculated the probability (P value) that the coefficient of each formula equaled zero. We calculated P values for each 10-ms time-point (i.e., bin) using a custom-made algorithm that was executed with commercially available software (MATLAB 6.5, MathWorks, Natick, MA). We took P < 0.01 to be statistically significant. Then, we calculated the sum of squares (SS) between groups and divided this value by the total SS to obtain the SS ratio. These SS values were obtained from ANOVA tables using a custom-made algorithm that was executed with commercial software (MATLAB 6.5, MathWorks). The SS ratio was analyzed for each 10-ms bin of data. The larger the SS ratio, the better the firing rate index formula represented neuronal activity. Based on the analysis of probability and the SS ratio, we classified neurons into four categories, according to whether the instantaneous activity was best and significantly represented by 1) the first cue, 2) the second cue, 3) the combination of arm target information, or 4) none of the regression coefficients were significantly different from zero. This classification was carried out for every 10-ms bin.

Linear Model Analysis After the Appearance of the Second Cue.    For activity after the second cue appeared, we quantified the extent to which neuronal activity represented selectivity for the second cue or the combination of the first and second cues. We used the following linear model to execute an ANOVA analysis

(4)

In this formula, the firing rate index is the square-root transformed firing rate during the second delay period, ₀ is the intercept, and ₁ and ₂ are coefficients. The categorical factors for the second cue (CUE2) are the four instructions given by the second cue (right arm, right target, left arm, and left target). The categorical factors for the combination of both the first and second cues (COMBINATION) are the four possible combinations of the two instructions indicating arm use and target location. We classified the neurons into four groups by looking at the probability that coefficients ₁ and ₂ were zero: 1) combination-only-selective group (P_{1 = 0} 0.01 and P_{2 = 0} < 0.01), 2) second-cue-only-selective group (P_{1 = 0} < 0.01 and P_{2 = 0} 0.01), 3) selective-for-both group (P_{1 = 0} < 0.01 and P_{2 = 0} < 0.01), and 4) nonselective group (P_{1 = 0} 0.01 and P_{2 = 0} 0.01).

Quantification of Selectivity for Arm Use and Target Location.    To examine the extent to which individual neurons exhibited selectivity for 1) the location of the target or 2) arm use during the set-cue and movement periods, we applied a multiple regression analysis using the following model formula

(5)

The default values of the variables for right and left were 0 and 1, respectively. We calculated the values of slopes ₁ and ₂ (by dividing the difference in activity in spikes/s by the dimensionless initial variable values 1 and 0) to assess target location and arm use selectivity, respectively. If ₁ > 0, it meant greater selectivity for the left target. If ₂ > 0, the selectivity was greater for the left arm. A merit of this analysis was that we could measure the selectivity with the dimension of firing rate (spikes/s).

Quantification of Muscle Activity.    To quantify the activity of muscles during movement execution, we calculated two indexes, the arm index and target index, based on the rectified EMG averaged over 20 trials for each movement. The indexes are defined as follows

(6)

(7)

In the formulas, V is the integrated value of the rectified EMG during the movement period, categorized using the appropriate subscript (RA, right arm; LA, left arm; RT, right target; LT, left target). Therefore V_RA-RT means muscle activity while reaching to the right target with the right arm. The indexes, which range from –1 to +1, include information on laterality (positive for "left" preference).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Sampling neuronal activity in the pre-SMA and SMA

Before we started collecting neuronal activity, we mapped the somatotopic organization of the SMA in the medial wall of the superior frontal gyrus using intracortical microstimulation (ICMS; 11–44 pulses; 200-µs width at 333 Hz; current, 5–50 µA) and by observing neuronal responses to the somatosensory stimuli applied by the experimenters. From caudal to rostral in the SMA, we found somatotopic representation arranged in the order leg, hip, trunk, arm, and face, as reported previously (Luppino et al. 1991; Matsuzaka et al. 1992; Mitz and Wise 1987). In the area rostral to the face area of the SMA, the ICMS effects required longer pulse trains, and neuronal responses to the somatosensory stimuli were weaker and less frequent. Instead, we observed ample visual responses to moving objects. We defined this area as the pre-SMA (Matsuzaka and Tanji 1996; Matsuzaka et al. 1992). The recording sites were reconstructed histologically using iron deposition produced by passing a positive DC current through the tips of the microelectrodes.

We recorded neuronal activity in the two areas (Fig. 1, B and C) in the right hemisphere. We monitored activity of every neuron we encountered by making peri-event rastergrams aligned at several task events. If the activity was judged to be modified by the appearance of any events, we continued to record the activity for off-line analysis. We analyzed 329 SMA neurons (148 in monkey 1 and 181 in monkey 2) and 349 pre-SMA neurons (107 in monkey 1 and 242 in monkey 2) that were found to be task-related (see METHODS). One-third of pre-SMA neurons we encountered and monitored on-line were found task-related with the off-line analysis. During recording, the success rate for the behavioral task was >96% for both monkeys.

Frequency of a change in activity during each task period

To study the overall changes in neuronal activity, we analyzed how many neurons in the pre-SMA and SMA showed increased or decreased activity during each of the 10 task periods (see METHODS). For each task period, we applied a paired t-test for each of the eight sequences of the first and second cues (paired t-test, = 0.05, corrected for 8 repeated analyses) compared with the control period. The results are summarized in Fig. 2. The fractions of neurons showing increased or decreased activity are depicted with thin solid and dotted lines, respectively. The bold line indicates the fraction of neurons with either increased or decreased activity (a neuron can show an increase in one trial type and a decrease in another, thus the sum of the fractions can exceed the total number of neurons). In the pre-SMA (Fig. 2A), >30% of the neurons exhibited changes in activity in the 10 task periods, most markedly in the second cue period, when 231 neurons (66%) showed changes (Fig. 2A, ). In contrast, in the SMA (Fig. 2B), the fraction of activity-modulated neurons remained low before the onset of the set cue and rose sharply toward the initiation of the reaching movement; 227 neurons (68%) changed their activity (Fig. 2B, ). The ² test revealed that a change in activity was more frequent among pre-SMA neurons from the precue to the late set-cue periods (P < 0.001 for each comparison), whereas activity changes during the movement period were more frequent in the SMA (P < 0.0001).

View larger version (30K): [in this window] [in a new window]   FIG. 2. Time courses of the change in neuronal activity in the pre-SMA (A) and SMA (B) over all the task periods. Thick lines indicate the fraction of neurons whose activity increased or decreased significantly compared with activity during the control period. Thin solid and dotted lines represent the fractions of neurons whose activity increased or decreased compared with activity during the control period, respectively. Gray areas indicate task periods during which visual cues were presented (from left to right: 1st cue, 2nd cue, and set cue). ***P < 0.0001 or **P < 0.001: task phase during which the fraction of neurons was larger in the pre-SMA or SMA. Numbers at the top of each panel are the actual numbers of neurons that changed activity (i.e., increase or decrease).

Before the first cue appeared (i.e., the precue period), 107 (30%) neurons in the pre-SMA showed significant modification of their activity compared with activity during the control period (paired t-test, = 0.05). This modification resembled that reported in the dorsal premotor cortex as activity anticipating the appearance of the cue (Mauritz and Wise 1986). In contrast, only 20 (6%) neurons in the SMA showed this modification. The frequencies of occurrence of the anticipatory activity differed significantly between the two areas (Fig. 2; Pearson's ² test with Yates' continuity correction, ² = 65.609, df = 1, P < 0.0001). Precue anticipatory activity was also found in 39 (11%) pre-SMA and 7 (2%) SMA neurons before the second cue. Furthermore, we found that 79 (22%) pre-SMA and 22 (6%) SMA neurons exhibited precue anticipatory activity before the set cue. The anticipatory activity before the second- and set-cue was counted 1) if the activity during the late delay period differed from that in the early delay period (paired t-test, = 0.01) and 2) if the activity did not reflect specific information given by the cues (P > 0.01 in every factor analysis). An example of pre-SMA neurons showing anticipatory activity is shown in Fig. 3. This neuron showed build-up activity before the appearance of the first, second, and set cues.

View larger version (79K): [in this window] [in a new window]   FIG. 3. Anticipatory activity of a pre-SMA neuron. Rasters and spike density functions indicate activity in trials sorted according to 8 task schedules with different instruction-cue orders. In the raster displays, each row represents a trial, and each dot represents a discharge from the neuron. Spike density functions (SDFs; Gaussian kernel, = 20 ms; mean ± SE) are below the raster display. Ordinate represents the instantaneous firing rate (spikes/s). Rasters and SDFs are aligned to the onset of the 1st cue, 2nd cue, set cue, and GO, which were merged at the midpoint of the 1st delay, 2nd delay, and set-cue period. From left to right, gray areas indicate when the 1st, 2nd, and set cues were presented. The instruction given by the cue is indicated above each panel (RA, right arm; LA, left arm; RT, right target; LT, left target). To the right of each panel, an additional display shows activity during the movement period (aligned to the screen touch, ). Tic marks on the horizontal axis are placed at 400-ms intervals. This neuron showed anticipatory activity before the 1st, 2nd, and set cues appeared.

Neuronal activity in the pre-SMA and SMA after the first cue

After the first cue was presented, >40% of pre-SMA neurons exhibited significant changes in activity compared with activity during the control period [n = 173 (49%) during the first-cue period, n = 153 (43%) during the early delay, and n = 168 (48%) for the late delay]. In contrast, fewer SMA neurons responded to the first cue [n = 50 (15%), 46 (13%), and 54 (16%) during the cue, early delay, and late delay periods, respectively]. The distributions of neurons with activity changes in the three periods differed significantly (Fig. 2; ² test, P < 0.0001).

We found that the activity properties of pre-SMA neurons could be grouped into three different types. The first type reflected the position of the white square in the first visual cue. An example of preferential activity for the position of the white square is shown in Fig. 4. The neuron was distinctly more active when the first cue was for either the left target or left arm than it was when the cue was for the right target or right arm. The common factor in the signals that led to an increase in neuronal activity was the appearance of the white square on the left. The second type of activity reflected the fact that the first cue contained an instruction for the location of the target. An example of a neuron of this type is shown in Fig. 5, where the first delay activity was selective for the right-side target. The third type of activity reflected the fact that the first cue had indicated which arm to use. In the example shown in Fig. 6, the pre-SMA neuron responded phasically to the cue that instructed the use of the right "arm." However, its activity was suppressed after the right "target" instruction. During the first delay period, for this particular neuron, the activity was more vigorous when the cue instructed the use of the left "arm" than left "target." Statistical tests supported these points. During the cue period, activity was significantly modified by two factors: type of instruction and position of the white square (2-way ANOVA, P < 0.0001 for main factors of type of instruction and the position of the white square). Furthermore, activity after the "right arm" instruction was significantly larger than activity after the other three instructions (P < 0.0001, Bonferroni pairwise comparisons). Similarly, during the early and late delay periods, activity was significantly modified by the two factors, and the activity was more vigorous when the cue instructed the "left arm use" than when instructed other three instructions (P < 0.004, Bonferroni pairwise comparisons).

View larger version (47K): [in this window] [in a new window]   FIG. 4. Cue-position selective activity of a pre-SMA neuron. This neuron was similarly active during the 1st delay period if the cue instructed either the left target or use of the left arm. Display format for this figure and Figs. 5, 6, 9, 12, 13, and 17 is the same as for Fig. 3.

View larger version (49K): [in this window] [in a new window]   FIG. 5. Target-instruction selective activity of a pre-SMA neuron. This neuron was active during the 1st delay period if the 1st cue instructed reach to the RT.

View larger version (76K): [in this window] [in a new window]   FIG. 6. Arm-instruction selective activity of a pre-SMA neuron. This neuron responded to 1st cues instructing the use of the RA. In addition, activity during the 1st delay period was greater if the 1st cue instructed the use of the LA.

View larger version (28K): [in this window] [in a new window]   FIG. 7. Cue-position and type-of-instruction selectivity in pre-SMA and SMA. Pie charts summarize proportion of neurons classified into 4 categories by 2-way ANOVA analysis (2 factors: cue POSITION and type of INSTRUCTION). Top: data for the pre-SMA. Bottom: data for the SMA. Actual number of neurons in each category is shown next to the corresponding portion of each pie graph. Position-only neurons were significantly selective (P < 0.01) only for the main factor POSITION. Instruction-only neurons were significantly selective (P < 0.01) only for the main factor INSTRUCTION. Neurons classified as "position and instruction" were significantly selective (P < 0.01) for both main factors or for the interaction between them.

View larger version (24K): [in this window] [in a new window]   FIG. 8. Time course of the selectivity for neuronal activity during the 1st cue and delay periods: bin-by-bin plots of the fraction of neurons exhibiting selectivity for each category. Top: data for the pre-SMA. Bottom: data for the SMA. Left column: position selectivity for the white square. Gray areas in each panel indicate when the cues appeared. Solid lines represent the fraction of neurons that was position-selective, calculated successively for each 10-ms bin. Dotted lines represent the fraction of neurons that was position-selective only (not instruction-selective). Tic marks on the horizontal axis are placed at 400-ms intervals. Right column: time course of instruction selectivity. Solid lines represent the fraction of neurons that exhibited arm or target instruction selectivity. Dotted lines represent the fraction of neurons that exhibited selectivity for the arm instruction.

In summary, for activity during the first cue and delay periods, neurons that were selective for either the position of the cue or the type of instruction were much more numerous in the pre-SMA than in the SMA. Further, the position selectivity in the pre-SMA appeared promptly after the first cue appeared, which was followed by the instruction selectivity 130 ms later.

Neuronal activity in the pre-SMA and SMA after the second cue

After the second cue appeared, the number of pre-SMA neurons showing changes in activity reached a peak (as many as 66% of all task-related neurons, see Fig. 2A). In contrast, only 23% of SMA neurons changed their activity in response to the second cue (Fig. 2B). The distributions of neurons with activity changes during the cue, early delay, and late delay periods differed significantly between the two areas (² test, = 0.01).

We found that the properties of responses to the second cue differed from those to the first cue in both areas. Although some neurons exhibited responses reflecting what the second cue indicated or instructed, these neurons representing the second cue were in the minority. In the majority of cases, we found that the selectivity was not merely for the reflection of the second cue itself. Rather, activity reflected information provided by the combination of two instructions (arm use and target location) given by the first and second cues. In the example shown in Fig. 9, the pre-SMA neuron was active most intensely when the combination of two cues was for the left arm and left target, irrespective of the order of presentation. Statistical tests revealed that activity of the neuron was significantly modified by the two factors of CUE2 and COMBINATION during the early delay period (P < 0.0001 for the 2 factors, see Eq. 4 in METHODS) and only by COMBINATION during the late delay period (P < 0.0001 for COMBINATION and P > 0.75 for CUE2).

View larger version (52K): [in this window] [in a new window]   FIG. 9. Activity of a pre-SMA neuron reflecting the combination of 2 instructions, as well as the 2nd cue. This neuron was active most intensely if the combination of the 2 instructions was LA and LT, regardless of the order of the 2 instructions (bottom).

View larger version (22K): [in this window] [in a new window]   FIG. 10. Second-cue and 2-cue-combination selectivity in the pre-SMA and SMA. Each pie graph summarizes the proportion of neurons classified into 4 categories in the linear model analysis (2 factors: 2nd CUE and COMBINATION of 2 cues). Top: data for the pre-SMA. Bottom: data for the SMA. Actual number of neurons in each category is shown next to each subcategory, whose identity is shown below. The 2nd-cue-only neurons were significant (P < 0.01) only for the main factor 2nd CUE. The combination-only neurons were significant (P < 0.01) only for the main factor COMBINATION. Neurons classified as both 2nd-cue and combination were significant (P < 0.01) for the 2 main factors.

View larger version (44K): [in this window] [in a new window]   FIG. 11. Time course of neuronal activity selectively representing the 1st and 2nd cues and the combination of both cues. A and B: bin-by-bin plot of selective activity expressed as the fraction of neurons that were best and significantly selective for the 1st cue (black trace), 2nd cue (blue), or both (combination selectivity, red). A: data for pre-SMA neurons. B: data for SMA neurons.

Neuronal activity in the pre-SMA and SMA during the set-cue period

The appearance of the set cue told the subjects to get ready to reach for the target in response to the GO signal. Throughout the delay period, >60% of the pre-SMA neurons exhibited changes in activity (Fig. 2A). In the SMA, there was a smaller proportion of set-cue-responsive neurons during the early part of the set-cue period (Fig. 2B; ² test, P < 0.003). Subsequently, the fraction of set-cue-modulated neurons increased sharply. The results of the quantitative analysis are shown in Fig. 11, based on calculations using Eqs. 1–3. The results revealed that during the set-cue period, most of the activity changes reflected the combination of the two cues (red traces in Fig. 11, A and B). In the pre-SMA, combination selectivity was reflected in 15–20% of the neurons, whereas in the SMA, the combination-selective neurons increased sharply toward the end of the delay, surpassing neurons in the pre-SMA.

A remarkable finding concerning the delay-period activity was that pre-SMA neurons preferentially represented the location of the reach target rather than arm use. A typical example of such target representation is shown in Fig. 12. This shows a pre-SMA neuron in which set-cue period activity was apparent when the target was on the left, regardless of the arm used. In contrast, SMA neurons were more selective for arm use, particularly during the late set-cue period preceding the GO signal. Such activity is exemplified in Fig. 13, where activity was observed selectively when the subject was prepared to use its left arm.

View larger version (58K): [in this window] [in a new window]   FIG. 12. Set-related activity of a pre-SMA neuron. This neuron was intensely active if the future target was on the left, regardless of arm use.

View larger version (64K): [in this window] [in a new window]   FIG. 13. Set-related activity of a SMA neuron. This neuron was more active if the subject was prepared to reach with the left arm.

View larger version (31K): [in this window] [in a new window]   FIG. 14. Arm use and target location selectivity of set- and movement-related activity in the pre-SMA and SMA. In each panel, pie charts summarize the proportion of neurons classified into 4 categories by the 3-way ANOVA analysis (3 factors: arm use, target location, and order of the 2 instructions). Top: data for the pre-SMA. Bottom: data for the SMA. Actual number of neurons in each category is shown next to each segment, and these are identified below the segments. The arm use–only neurons were significant (P < 0.01) only for the main factor ARM. The target location–only neurons were significant (P < 0.01) only for the main factor TARGET. Both arm use and target location neurons were significant (P < 0.01) for both of the main factors or for the interaction between ARMx TARGET.

View larger version (34K): [in this window] [in a new window]   FIG. 15. Quantitative analysis of arm use and target location representation for individual pre-SMA neurons classified as set- and movement-related. A: scatterplot of the regression slope of arm use (horizontal axis) vs. the regression slope of target location (vertical axis) for 222 pre-SMA neurons during the late set-cue period. B: cumulative fraction of absolute values of the regression slopes of arm use and target location for the 222 pre-SMA neurons. C: scatterplot of the regression slope of arm use (horizontal axis) vs. the regression slope of target location for 196 pre-SMA neurons during the movement period. D: cumulative fraction of absolute values of regression slopes of arm use and target location for the 196 pre-SMA neurons during the movement period.

View larger version (35K): [in this window] [in a new window]   FIG. 16. Quantitative analysis of arm use and target location representation for individual SMA neurons. A: scatterplot of the regression slope of arm use vs. the regression slope of target location for 170 SMA neurons during the late set-cue period. B: cumulative fraction of absolute values of regression slopes of arm use and target location for the 170 SMA neurons. C: scatterplot of the regression slope of arm use vs. the regression slope of target location for 227 SMA neurons during the movement period. D: cumulative fraction of absolute values of the regression slopes of arm use and target location for the 227 SMA neurons during the movement period.

Neuronal activity during movement execution

In the SMA, 68% of task-related neurons changed their activity during the movement-execution period (Fig. 2B). Of these 227 neurons, 181 (79%) exhibited increased activity. In the pre-SMA, 56% of the task-related neurons showed changes in activity during the movement period (Fig. 2A). Of note, a majority (n = 132, 67%) of the 196 movement-related neurons showed decreased activity. In both areas, neuronal activity was selective for information reflecting the combination of Cue 1 and Cue 2, rather than reflecting what Cue 1 or Cue 2 showed or instructed individually (Fig. 11, A and B). Greater selectivity was observed for SMA neurons than for pre-SMA neurons.

We found that a majority of the selectivity exhibited by movement-related SMA neurons reflected arm use, as apparent in the SMA activity shown in Fig. 17. In that example, the activity was intense preceding the reach movement using the left arm. To systematically analyze how selectivity for location and arm use were represented in SMA and pre-SMA neurons, we applied three-way ANOVA to movement-related neurons for the factors TARGET location, ARM use, and ORDER of instructions. In the SMA (Fig. 14, bottom), selectivity for arm use was observed in 126 (38%) neurons (P < 0.01 for ARM or P < 0.01 for ARM x TARGET), and selectivity for target location was observed in 65 (19%) neurons (P < 0.01 for TARGETor P < 0.01 for ARM x TARGET). In the pre-SMA (Fig. 14, top), 58 neurons (16%) showed selectivity for arm use (P < 0.01 for ARM or P < 0.01 for ARM x TARGET), and 56 neurons (16%) showed selectivity for target location (P < 0.01 for TARGETor P < 0.01 for ARM x TARGET).

View larger version (47K): [in this window] [in a new window]   FIG. 17. Movement-related activity of an SMA neuron. Intense activity was observed exclusively while reaching toward targets with the left arm.

Time course of arm use and reach-target representation from cue 2 reception to movement

Figure 11 shows that the representation of information given with either Cue 1 or Cue 2 alone faded after the delay periods. In contrast, the combination of information given by both cues was represented throughout the Cue-2 delay, set-cue, and movement periods. We were interested in the time course of arm use and reach-target representation because both are vital for planning and executing the reach movement. This included the information represented in the period from the reception of Cue 2 until movement execution. Therefore we analyzed the activity of all neurons that were defined as best and significantly selective for the combination of two instructions (Fig. 11, red traces). We applied three-way ANOVA (with the factors target location, arm use, and order of instructions) to a series of inverse-ISI spike data in every 10-ms bin. For each 10-ms bin, we classified the activity into four categories: 1) selective for arm use (ARM < 0.01 or ARM x TARGET < 0.01), 2) selective for target location (TARGET < 0.01 or ARM x TARGET < 0.01), 3) selective for both arm use and target location (ARM < 0.01 and TARGET < 0.01, or ARM x TARGET < 0.01), and 4) nonselective. The results of this analysis are shown in Fig. 18 (Fig. 18A for the pre-SMA and Fig. 18B for the SMA), in which we plot, bin by bin, the fraction of neurons whose activity is assigned to categories 1–3. Blue, green, and black traces represent the fractions of neurons classified as selective for target location, arm use, and both, respectively. Figure 18A shows that the appearance of Cue 2 provides both target and arm information to pre-SMA neurons. Throughout the subsequent preparatory period, target information is maintained, while arm use information decreases. Toward movement initiation, target information decays, rather than increases, in the pre-SMA. In contrast, the level of target representation stays low in the SMA (Fig. 18B). Instead, arm use representation predominates and grows rapidly from the middle of the set-cue period to movement initiation.

View larger version (42K): [in this window] [in a new window]   FIG. 18. Time course of neuronal activity representing arm use and target location among neurons defined as instruction-combination selective. A and B: bin-by-bin plot of selective activity expressed as the fraction of neurons that were selective for target location (blue), arm use (green), and both target location and arm use (black). A: data for pre-SMA neurons. B: data for SMA neurons.

In addition to neuronal recordings, we monitored the following muscles bilaterally during task performance: the biceps and triceps brachii, deltoid (anterior, lateral, and posterior heads), trapezius, flexor and extensor carpi radialis, supraspinatus, infraspinatus, pectoralis major, rhomboid, and neck and paravertebral muscles. We found that the 12 forearm muscles, but not the neck or paravertebral muscles, increased their activity in association with movement execution. Despite their movement-associated activity, they did not show consistent changes in activity before actual execution of the movements. To quantify the activity of the 12 muscles during the movement, we calculated two indexes, the arm index and target index (Eqs. 6 and 7 in METHODS), based on the rectified EMG averaged over 20 trials for each movement.

We examined the distribution of the arm and target indexes. Examples of data recorded from the 12 muscles in the left arm of each monkey (12 x 2 = 24 muscles) are shown in the scatterplot in Fig. 19A. The target index was distributed around the horizontal line, indicating values close to zero. The arm index was distributed widely in the positive range, indicating left-arm preference. To compare arm and target representations statistically, we applied the Kolmogorov-Smirnov test to the absolute indexes. As shown with black cumulative plots in Fig. 19B, the arm index was much greater than the target index (KS = 0.9583, P < 0.0001). We performed the same analysis repeatedly on data obtained in 12 sessions, continuing to reach the same conclusion. These results indicate that muscle activity mainly reflected arm use and only represented target location to a small degree. This conclusion was reasonable because the horizontal distance between the two targets was small in our experimental system (55 mm or 10.5°).

View larger version (17K): [in this window] [in a new window]   FIG. 19. Muscle activity and movement-related SMA neuronal activity during reaching movements. A: scatterplot of the arm index (horizontal axis) vs. target index (vertical axis) for 24 muscles on the left side of the body (12 muscles from each monkey: *, muscle activity of monkey 1; , muscle activity of monkey 2) that showed movement-related activity. B: black lines denote cumulative fraction of absolute values of arm index (solid line) and target index (dotted line) of muscles shown in A. Gray lines denote cumulative fraction of absolute values of arm index (solid line) and target index (dotted line) of movement-related SMA neurons (n = 227).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
In this study, we observed striking differences in the activity of SMA and pre-SMA neurons with respect to five behavioral aspects. First, neuronal activity preceding the appearance of visual cues was more frequent in the pre-SMA. Second, in response to the appearance of the first instructional cue, a sizeable number of pre-SMA neurons, but few SMA neurons, were active during the cue and delay periods. Third, neurons responding to the second instructional cue were much more frequent in the pre-SMA. In addition, pre-SMA neurons often reflected information combining the instructions in the first and second cues. Fourth, in response to the set cue's prompting preparation for a forthcoming movement, pre-SMA neurons preferentially reflected the location of the target. In contrast, SMA neurons mainly reflected which arm to use. Fifth, during the execution of the reaching movement, the majority of SMA neurons increased their activity, which was largely selective for use of either the ipsilateral or contralateral arm. In contrast, the activity of the majority of pre-SMA neurons tended to be suppressed. Based on these findings, we discuss the implications of the differential properties of SMA and pre-SMA neurons, which suggest functional specialization of the two areas, with respect to receiving associative cues, processing information, planning motor behavior, and executing movement.

Pre–cue activity in the pre-SMA

As many as 30% of the task-related neurons in the pre-SMA changed activity before the onset of the instruction cues. This activity is akin to the precue anticipatory activity reported in the dorsal premotor cortex (Mauritz and Wise 1986) and might reflect the anticipation of