The supplementary motor area (SMA) has long been thought to have a special role in the internal generation of complex movements. Yet, a number of recent functional imaging studies indicate that the SMA is activated during the execution of simple movements guided by sensory cues. The extent of participation of the cingulate motor areas in visually guided movements also is unclear. To explore these issues we used the 2-deoxyglucose (2DG) technique to measure functional activation in the motor areas on the medial wall of the hemisphere in monkeys trained to perform visually guided reaching movements to randomly presented targets. This approach enabled us to make precise comparisons between sites of activation and the location of specific premotor areas on the medial wall of the hemisphere. We found that the SMA was strongly activated during reaching to different visual targets. Indeed, its activation was comparable to that of the primary motor cortex (M1). In contrast, none of the cingulate motor areas displayed significantly increased activation specifically related to arm movements. Our results provide further support for the involvement of the SMA in visually guided movements. Furthermore, our observations suggest that during externally guided reaching, SMA activation is tightly coupled to that of M1, but dissociated from that of the cingulate motor areas.
Concepts about the function of the supplementary motor area (SMA) have been strongly influenced by the results from early functional imaging studies that emphasized its involvement in ‘higher-order’ aspects of motor behavior (Goldberg, 1985; Picard and Strick, 1996a; Tanji, 1994, 2001). For example, the SMA was reported to be the only region of the brain activated when subjects imagined that they were performing a complex sequence of finger movements (Roland et al., 1980). The primary motor cortex (M1) was activated when the subjects actually performed the sequence. In contrast, during simple movements, activation was present in M1, but not the SMA. These and other observations lead to the suggestion that the SMA was a supramotor area specifically involved in the internal generation of complex movements (Orgogozo and Larsen, 1979; Roland et al., 1980; Goldberg, 1985).
In recent years it has become clear that the anatomical organization of the medial wall of the hemisphere is more complicated than previously thought (Hutchins et al., 1988; Dum and Strick, 1991; Luppino et al., 1991; Shima et al., 1991; He et al., 1995; Picard and Strick, 1996a). In addition to the SMA, the medial wall contains three separate premotor areas which are buried in the cingulate sulcus. These cingulate motor areas (CMA) project directly to M1 and include: the CMAd on the dorsal bank, the CMAv on the ventral bank and the CMAr which is located more rostrally on the dorsal and ventral banks of the cingulate sulcus (Hutchins et al., 1988; Dum and Strick, 1991; Wang et al., 2001). Like M1, the SMA and all of the CMA project directly to the spinal cord (Dum and Strick, 1991, 1996; He et al., 1993, 1995).
The medial wall of the hemisphere also contains another cortical field in the rostral portion of area 6 which has come to be termed the PreSMA (Luppino et al., 1991; Matelli et al., 1991; Matsuzaka et al., 1992). Although at one time the PreSMA was included within the SMA, it is now recognized as a separate cortical field. Unlike the premotor areas, the PreSMA does not have substantial connections with M1 and it does not project to the spinal cord (Dum and Strick, 1991; He et al., 1995; Wang et al., 2001). Instead, the PreSMA is heavily interconnected with regions of prefrontal cortex (Bates and Goldman-Rakic, 1993; Luppino et al., 1993; Lu et al., 1994; Wang et al., 2001). Based on this and other evidence, we have argued that the PreSMA is more properly considered a region of prefrontal cortex than a premotor area (Picard and Strick, 2001).
Advances in our understanding of the anatomical subdivisions of the medial wall have led us to be concerned that activity which was formerly attributed to the SMA was actually located in nearby cortical fields. Similarly, the results of recent imaging studies in humans have led us to question whether the SMA is exclusively involved in movements that are internally generated (Picard and Strick, 1996a, 2001; Petit et al., 1998; Deiber et al., 1999; Jenkins et al., 2000; Thickbroom et al., 2000; Crosson et al., 2001; Weeks et al., 2001; Cunnington et al., 2002). However, accurate localization of activations in human studies presents a special problem in part because the location of the motor areas is not well defined in the human and has a large variability. In addition, only relative levels of activations between tasks are frequently reported. For these reasons, it is difficult to get a clear sense of the basic patterns of activation on the medial wall of the hemisphere during movement and of how these patterns relate to the motor areas that are known to exist in primates.
In neuron recording studies in monkeys, the functions of the SMA have often been probed using tasks that emphasize higher order aspects of motor control such as temporal sequence planning (Tanji, 2001). Nonetheless, a number of recent studies have found that the majority of SMA neurons show activity related to movement kinematics and dynamics (Clower and Alexander, 1998; Padoa-Schioppa et al., 2002; Russo et al., 2002). These results suggest that a substantial amount of activity in the SMA may be devoted to fundamental aspects of motor control.
To examine the involvement of the SMA, CMA and PreSMA in basic aspects of motor behavior, we analyzed the pattern of radiolabeled 2-deoxyglucose (2DG) uptake in the medial wall of the hemisphere in monkeys during the performance of reaching movements to randomly presented visual targets. Monkeys are ideally suited for these studies because the anatomical organization of the medial wall motor areas is well characterized in this primate. We used the 2DG technique because it provides spatial information at a resolution that is high enough to differentiate sites of activation in closely adjoining areas. We compared the patterns of activation in monkeys who performed a control task (LICK task) or one of two visually guided reaching tasks (TRACK task and RANDOM task). This allowed us to distinguish activations specifically related to the performance of visually guided reaching movements from those activations associated with other task components such as reward consumption. The reaching tasks in this study also allowed us to compare patterns of activation to those found when monkeys performed similar movements generated from memory (Picard and Strick, 1997). Some of these results have been presented in abstract form (Picard and Strick, 1996b).
Materials and Methods
Six monkeys (Macaca nemestrina or mulatta, weight: 3.0–7.8 kg) were used. The monkeys were seated in a primate chair with the left arm and one leg comfortably strapped down for immobilization. A stainless steel tube mounted on the primate chair at mouth level delivered controlled amounts of liquid rewards. The care of the animals and the experimental protocols used adhered to the NIH Guide for the Care and Use of Laboratory Animals and the US Public Health Service Policy on Humane Care and Use of Laboratory Animals. All procedures used followed institutional guidelines and were approved by the relevant committees.
We examined patterns of activation on the medial wall during three tasks: a control task (LICK) and two reaching tasks (TRACK and RANDOM) (see below for details). The essential features of the two reaching tasks were similar. In both, the monkeys performed reaching movements to one of five possible targets. A visual cue told the monkey which target to contact. Movement to the target was required immediately after presentation of this cue. Neither task included an instructed delay period and neither allowed for target selection by the monkey. The reaching movements were performed one after another without pause or return to a ‘home’ position. The main difference between the two reaching tasks is that in the TRACK task the monkeys performed sequences of three individually cued movements followed by an inter-trial interval, whereas in the RANDOM task the reaching movements were performed continuously. A detailed description of each task follows.
Two monkeys were trained for 16–19 months to perform sequences of three visually guided reaching movements (Mushiake and Strick, 1993, 1995). Each monkey faced a custom-built panel with five touch-sensitive targets positioned in a horizontal row. Five red light-emitting diodes (LEDs) located directly above each target served as ‘target cues’. The monkey initiated a trial by placing its right hand on the hold key which was located at the monkey’s waist level. A 200 Hz auditory signal was generated for the duration of contact with the hold key. The monkey was required to maintain contact with the hold key for a variable time (0.6–1.5 s). At the end of the hold period, one target cue was lit and a brief auditory ‘Go’ signal (500 Hz) was presented. The monkey was trained to release the hold key and quickly (<1 s) contact the cued target. Correct contact of the target immediately triggered the illumination of a second target cue over another target. The monkey had to quickly move from the first to the second target. When the monkey contacted the second target, a third target cue was illuminated and the monkey was required to move quickly to the third target. A brief 1 kHz tone signaled the correct completion of the sequence of three movements. The monkey was rewarded with a drop of fruit juice 100 ms later. If the monkey made an error, the trial was aborted and the same set of target cues was repeated on the next trial. Trials were separated by 1.1 or 1.5 s inter-trial intervals. Then, a new trial was initiated with a different sequence of target cues. The reaction times and eye movements in response to the target cues indicate that both monkeys relied on the cues to visually instruct each movement. At the time of the experiments, the two monkeys performed correctly on 97% and 98% of the trials.
Two other monkeys were trained for 12–17 months to perform continuous reaching movements to visual targets (RANDOM task). The RANDOM task was devised to maximize the frequency of visually guided movements during the period of 2DG uptake. Each monkey faced a touch-sensitive monitor (Elo Touch Systems Inc., Oak Ridge, TN) that displayed the outlines of five targets arranged in a horizontal row. On each trial, one of the targets was filled with yellow. The monkey was required to quickly (<800 ms) touch the filled target with its right hand. Contact outside of the target area was signaled with a 50 Hz tone. Trials were repeated at the occurrence of an incorrect response. Correct contact within the filled target was indicated by a brief 1 kHz tone. A new target was filled 100 ms after a correct response and the monkey was again required to quickly touch the indicated target. The short delay between response and new target presentation prevented the monkey from guessing the location of the next target and prompted the monkey to move its hand directly from one target to the next without pause. Thus, movements were performed continuously during the RANDOM task. The monkey was rewarded with a drop of water after every fourth or fifth correct response. The targets were randomly selected on each trial with the only constraint that a target not be presented twice in succession. Thus, the movement that the monkey made on each trial of the RANDOM task was visually cued. At the time of the experiments, the two monkeys performed correctly on 93% and 98% of the trials.
To control for activations that were not directly related to the performance of the reaching movements, another two monkeys were trained for 5–6 months to perform a licking task (LICK task). As in the TRACK task, each monkey faced a panel with five targets positioned in a horizontal row. The monkey was presented with visual and auditory signals that were comparable to those of the TRACK task. However, these signals were meaningless during the LICK task and the monkey did not respond to them. The animal received drops of fruit juice that were delivered at variable intervals which approximated the rate of reward delivery during the TRACK task (3.2–4.8 s). By comparison, reward delivery occurred on average every 2.3 s during the RANDOM task.
We followed conventional procedures for qualitative analysis of 2DG uptake (Juliano et al., 1981). On the day of the experiment, an i.v. catheter was inserted transcutaneously into the long saphenous vein and was secured for 2DG and drug injections. The task was then initiated. After the monkey achieved stable task performance (5–10 min), it was given a pulse i.v. injection of [14C]2DG in sterile saline (100 μCi/kg, 55 mCi/mmol, American Radiolabeled Chemicals, St Louis, MO). The catheter was immediately flushed with a solution of sterile saline and heparin (1.0–1.3%). All solutions were warmed to 38°C prior to injection. The monkey continued to perform the task for 35–45 min. The monkey was then deeply anesthetized (sodium pentobarbital, 40 mg/kg, i.v.) and perfused through the heart with phosphate buffer (pH 7.4) followed by 4% paraformaldehyde, and then a solution of 4% paraformaldehyde and glycerin (10%). The brain was extracted, blocked and immediately frozen by immersion into isopentane chilled to −60°C. The rostro-caudal extent of frontal lobe blocks was 24–39 mm. The brain blocks were stored at −70°C until processing.
Tissue Processing and Mapping Procedures
The brain blocks were cut in the coronal plane in 30 μm thick serial sections (25 μm for case N1) in a cryostat at −18 to −22°C. Every section was picked-up on glass slides except for monkey N1 (LICK task) in which every fourth section was saved. The brain sections used for 2DG analysis were quickly dried on a hot plate at 65°C. Autoradiographs of the sections were made with SB-5 or Biomax MR-1 film (Kodak: Rochester, NY). Images of the autoradiographs were captured using a Dage-MTI (Michigan City, IN) Series 68 camera and digitized at a pixel resolution of 55–77 μm by 46–62 μm. We obtained estimates of 2DG uptake by converting gray values from the digitized autoradiographs to 14C tissue concentration equivalents based on 14C standards (Amersham, Arlington Heights, IL) applied to each film. We measured 14C values in the middle cortical layers. To do this, we drew a line through the cortex approximately midway between the cortical surface and the white matter (Fig. 1A). The 14C values in four pixels (~200–300 μm), two on each side of this line, were averaged. These averaged values were then used as the local measurements of activation in our flattened maps of the medial wall. These maps were constructed using sections spaced 90 μm apart (100 μm for N1) (Fig. 1B). The methods for unfolding the medial wall and generating a flattened map have been described in detail elsewhere (Dum and Strick, 1991; He et al., 1993, 1995). Serial sections spaced 300 μm apart (N1, 400 μm) were Nissl-stained for estimation of cytoarchitectonic boundaries.
Normalization of Activation
To compare patterns of activation between animals, we transformed 14C concentration values to a normalized activation scale (% range) (Picard and Strick, 1997). To accomplish this, the range of activation was defined for each animal as the difference between peak 14C tissue concentration and background concentration (Fig. 1C). Peak concentration was measured in the portion of the face representation of the primary somatosensory cortex (S1) that is located on the precentral gyrus, caudal to the inferior precentral sulcus (Pritchard et al., 1986; Cipolloni and Pandya, 1999). We measured background cortical concentration in a portion of the leg representation of area 4 (M1) that is on the precentral gyrus close to the midline. The values assigned to peak and background were the median 14C concentration in a 2 mm2 area in each region. Then, normalized values of activation were calculated for each pixel:
The validity of this normalization procedure relies on the presence of stable metabolic features across animals. This is in fact what we observed. 2DG uptake was low in the leg representation of M1 (background) in every monkey because they made few if any leg movements during the tasks. Subtraction of this measure of non-specific 2DG uptake in cortex minimized the effect of subject-dependent variables, such as cerebral blood flow or blood glucose level, on our measure of activation. Likewise, intense activation in the oral representation of S1 (peak) was present in all monkeys because they all consumed liquid rewards during the tasks. This measure of peak activation set an upper bound on 2DG uptake that was related to the number of rewards each monkey earned (and the trials each monkey performed) during the experimental period. Further evidence for the validity of this normalization procedure is the observation that the background/peak ratio was similar across experimental tasks [F(2,3) = 1.08, P = 0.44] (Fig. 2A).
We would like to emphasize that we took our measurement of background from a region of the leg representation in M1 where 2DG uptake was low in our trained monkeys (Fig. 1C). Other regions of hindlimb sensorimotor cortex display higher levels of uptake. For example, bilateral activation is present in a region of area 3a which is located medially in the anterior bank of the central sulcus (not illustrated).
Based on data from prior anatomical and physiological studies (Muakkassa and Strick, 1979; Mitz and Wise, 1987; Dum and Strick, 1991; Luppino et al., 1991; Matelli et al., 1991; Matsuzaka et al., 1992; Tanji, 1994; He et al., 1995; Morecraft et al., 1996; Picard and Strick, 1996a; Tokuno et al., 1997), we outlined the location of six regions on the medial wall for analysis: the arm and face representations of the SMA, the arm representations of the CMAd and CMAv and the CMAr and PreSMA (Figs 1B and 4). For comparison, activation was also measured in the arm representation of M1 (area 4) in the anterior bank of the central sulcus and on the precentral gyrus (Dum and Strick, 1991; He et al., 1995). We searched for the peak activation in each defined area by measuring the median values of 2DG labeling in 2 mm2 areas of interest. Significant activation in an area of interest was defined as peak 2DG uptake greater than three standard deviations (SD) above background. Depending upon the animal, this threshold represents normalized activation greater than the 18–24% range (median = 20%). Task-related differences in the degree of activation were assessed by a two-way (Task × Area) analysis of variance using normalized activation measures. When appropriate, planned comparisons of peak activation in the selected regions of the medial wall and M1 during the tasks were made. Considering the small number of subjects in each group (n = 2), the significance threshold for these contrasts was set at P < 0.05. However, all significant task differences in the motor areas had probabilities <0.005. Only activations in the left hemisphere (contralateral to the moving arm in the reaching tasks) are examined in this report.
Validation of the Procedure for the Normalization of 2DG Uptake
In addition to the ‘peak’ and ‘background’ regions used for normalization, several cortical areas that are not directly involved in motor control of the arm also displayed consistent patterns of 2DG uptake across monkeys. The auditory cortex in the lateral sulcus (Jones et al., 1995; Hackett et al., 1998) and the gustatory area on the precentral gyrus (Pritchard et al., 1986; Ogawa, 1994) contained high concentrations of 2DG in all monkeys. In contrast, 2DG uptake was consistently below our measure of background in the portion of the anterior cingulate gyrus located just above the genu of the corpus callosum. Peak activation in these areas was measured and compared as described above. Activation in each of these areas was task-independent [main effect of task F(2,9) = 1.9, P > 0.2] (Fig. 2B). For example, normalized activation in the auditory cortex varied by less than 11% across the tasks and the differences observed were not statistically significant (P > 0.7).
On the other hand, there was a notable decrease of activation of the gustatory area in the monkeys who performed the RANDOM task compared with the monkeys who performed the LICK and the TRACK tasks. The reduction of activation in the RANDOM task was explained by the fact that monkeys in this group received water for reward, whereas fruit juice was used for the LICK and TRACK tasks (t-test, t = 2.988, d.f. = 4, P = 0.04). This result is not surprising since water is a less potent stimulus than fruit juice for neurons in the gustatory cortex of macaques (Ito and Ogawa, 1994; Scott and Plata-Salamán, 1999). No other area examined showed a significant modulation of activation with the type of reward used. The patterns of 2DG uptake in the auditory cortex, anterior cingulate gyrus and in the gustatory area that we observed for the subjects of this study were confirmed in a separate analysis that included data from four additional monkeys used in other experiments (n = 10). Taken together these observations indicate that our normalization procedure adequately controlled for global metabolic variations between animals. As a consequence, we are able to express the relative patterns of 2DG uptake present in the cortex of different animals on a common scale.
The premotor areas on the medial wall of the hemisphere and the PreSMA displayed consistent patterns of 2DG uptake associated with the LICK, TRACK and RANDOM tasks. These patterns will be described in the sections below. Other sites on the medial wall also displayed 2DG uptake. For example, activation was present in caudal regions of the cingulate sulcus in one of the LICK animals and in all of the TRACK and RANDOM animals (Fig. 3). In addition, a narrow band of 2DG uptake was present in portions of areas 23a, 30 and 29 just above the corpus callosum (Vogt et al., 1987) in all animals (not illustrated). The rostral extent of this activation varied considerably from animal to animal. Because these regions of activation lie outside of the premotor areas and the PreSMA, they will not be considered further.
Activation Related to Reward Consumption
Significant activation was present in two foci within the motor areas on the medial wall of the hemisphere during the LICK task (Fig. 3, top). A large region of the superior frontal gyrus, close to the dorsal surface of the hemisphere, had the most intense activation. This focus spanned ~5 mm in the rostro-caudal axis, with its center located near the level of the genu of the arcuate sulcus, i.e. between the PreSMA and the core of the SMA. Thus, the rostral portion of the activated region overlaps the face representation of the SMA (SMA-face) (Muakkassa and Strick, 1979; Mitz and Wise, 1987; Luppino et al., 1991; Morecraft et al., 1996; Tokuno et al., 1997). However, the activation extends caudally, behind the level of the genu of the arcuate sulcus, into a region of the SMA that projects to upper cervical segments of the spinal cord (Dum and Strick, 1991; He et al., 1995). Intracortical stimulation in this region evokes face, neck and upper trunk movements (Mitz and Wise, 1987; Luppino et al., 1991). Thus, it is likely that the entire region of 2DG activation includes portions of the face, neck and upper trunk representations of the SMA. However, for brevity, we will refer to this as ‘SMA-face’ activation.
The second site of activation associated with the LICK task was located more rostrally within the banks of the cingulate sulcus. It is likely that this region includes the face representation of the CMAr (CMAr-face) (Muakkassa and Strick, 1979; Morecraft et al., 1996; Tokuno et al., 1997). An isolated area of 2DG uptake was present in one of the LICK animals on the superior frontal gyrus at levels rostral to the genu of the corpus callosum (Fig. 3). This activation lies rostral to the arm representation of the PreSMA defined in other studies (Luppino et al., 1991; Akkal et al., 2002; Fujii et al., 2002; Matsuzaka et al., 1992). Activation at this site was not present in the other LICK animal or in the TRACK or RANDOM animals (Fig. 3) and therefore, may not be task-related.
The other premotor areas on the medial wall and the PreSMA displayed little or no activation associated with the LICK task (Figs 3 and 4; Table 1). For example, less than 6% of the SMA-arm and less than 1% of the CMAd and CMAv contained activated pixels associated with the LICK task. These results indicate that the activity of orofacial and other muscles associated with reward consumption did not engage the core of the arm representations in the SMA, CMAd, CMAv, CMAr and the PreSMA.
Because the TRACK and RANDOM monkeys performed the trained movements for liquid rewards, they displayed activations at the same sites as monkeys who performed the LICK task (Figs 3 and 4, Table 1). For example, intense activation was present in the SMA-face in all animals. In fact, the peak intensity of activation in SMA-face and in the CMAr did not significantly differ between the tasks (SMA-face, P > 0.31; CMAr, P > 0.18). To distinguish 2DG uptake evoked by different behaviors in our figures we display activations associated with reward consumption in grayscale and those associated with reaching movements in color (Figs 3–5).
The TRACK and RANDOM tasks evoked activation in a dorsal region of the superior frontal gyrus that was located rostral to the PreSMA. This labeling was part of a larger area of activation in the supplementary eye field which is located near the medial wall, but largely on adjacent portions of the lateral surface of the hemisphere (Schlag and Schlag-Rey, 1987; Fujii et al., 2002). Because activation in this region is unlikely to be specifically related to arm movements, we also display it in grayscale (Fig. 3).
Activation Related to Visually Guided Reaching
In addition to the activation related to reward consumption, a unique set of activations was present in the monkeys who performed the TRACK and RANDOM tasks (Table 1). The most intense of these activations was found at two sites on the superior frontal gyrus. One site was located within the arm representation of the SMA and the other was located within the PreSMA (Figs 3–5).
The uptake of 2DG in the SMA-arm and in the PreSMA during the reaching tasks was remarkable for its intensity, extent and consistency. Indeed, the intensity of activation in both cortical areas (54–64% range; Table 1) was comparable to that observed in the arm area of M1 in the same animals (56% range, P = 0.45; Table 1, Fig. 5). Significant activation nearly filled the regions considered to be the SMA-arm and PreSMA. Our prior anatomical studies suggest that there is a gradient of arm representation within the SMA with the representation of the distal arm located more caudally and ventrally in the SMA than the representation of the proximal arm (Dum and Strick, 1991; He et al., 1995). Based on this gradient, intense 2DG uptake was present in both the proximal and distal arm representations of the SMA; however, more of the proximal arm representation was strongly activated (Figs 3 and 4). Overall, we did not see substantial differences between the TRACK and RANDOM tasks in either the patterns or the peaks of activation in the SMA-arm (P > 0.95) and PreSMA (P > 0.46) (Figs 3 and 4, Table 1).
In contrast to the intense activation seen in the SMA-arm and PreSMA, the arm representations of the CMAd and CMAv displayed little or no 2DG uptake in either the TRACK or the RANDOM task. Although weakly activated pixels were seen in the CMAd and CMAv of some animals, the number of these pixels was not large enough for either area to reach our threshold for statistical significance in any animal (see Materials and Methods and Table 1). In fact, the activations of the CMAd and CMAv were not significantly different during the LICK, TRACK and RANDOM tasks (CMAd, P > 0.4; CMAv, P > 0.32). The modest activation seen in the CMAd and CMAv of the TRACK animal displayed in Figure 3 (middle) was the largest activation seen in any animal that performed visually guided reaching movements. Indeed, the extent of the activation seen in the CMAd and CMAv of the RANDOM animal displayed in Figure 3 (bottom) was more typical of that seen in the other animals in this experimental series.
Some 2DG uptake associated with the TRACK and RANDOM tasks was found in caudal regions of the medial wall that lie outside of the arm representations of the SMA, CMAd and CMAv (He et al., 1995) (Fig. 3). A few sparse foci of activation were located in parts of the leg/lower body representation of the SMA and M1 which lie caudal to the main focus of activation in the SMA-arm. Comparable activation is seen in this region in monkeys trained to perform reaching movements that were ‘internally guided’ (Picard and Strick, 1997) (unpublished observations). Although monkeys did not make overt leg movements during the reaching tasks, the animals commonly gripped the bars of the primate chair with their feet, possibly to help stabilize their posture. Thus, it is likely that these activations are incidental to the performance of reaching movements rather than being specifically related to them. Likewise, we saw some separate clusters of activation on the dorsal bank of the cingulate sulcus caudal to the arm representation of the CMAd. This region does not have a specific designation, but is known to contain cells that project to multiple segments of the spinal cord (He et al., 1995).
One of the major results of the present study is that both the SMA and the PreSMA display substantial uptake of 2DG in association with visually guided reaching movements. Indeed, the second surprising result is that the intensity of activation in the SMA and PreSMA is comparable to that observed in the arm area of M1 in the same animals. Taken together, these results are clearly at odds with proposals that the SMA and PreSMA are exclusively involved in higher-order aspects of motor programming and in the generation of internally guided movements.
For the monkey, there are well-defined maps of the motor areas on the medial wall of the hemisphere (Dum and Strick, 1991; Luppino et al., 1991; He et al., 1995). These maps provide the critical framework for interpreting sites of activation in this topographically complex region. Based on this analysis, we were able to draw unequivocal associations between sites of activation and cytoarchitectonic areas of cortex. A comparable analysis is problematic for human studies where it is not feasible to define the boundaries of cortical areas with certainty and where large variations in gyral and sulcal patterns are commonplace (Ono et al., 1990; Paus et al., 1996a,b). The comparison between monkey and human imaging results provides a means to bridge this data gap.
Our results on the SMA clearly dissociate it from the three cingulate motor areas. All of these areas have one important aspect of their anatomy in common. They all are interconnected with M1 (Dum and Strick, 1991; Shima et al., 1991; Luppino et al., 1993; Tokuno and Tanji, 1993; Hatanaka et al., 2001; Wang et al., 2001). However, only the SMA displayed activation of the same magnitude as M1 during the visually guided tasks we used. In fact, neither the CMAd, CMAv or the arm representation of the CMAr displayed significant activation during the TRACK and RANDOM tasks. Thus, the patterns of activation for the motor areas on the medial wall cannot be explained solely in reference to their connections with M1.
The functional dissociation that we observed is particularly notable for the CMAd, which is located adjacent to the SMA in a subfield of area 6 [area 6c of (Dum and Strick, 1991; He et al., 1995)]. Despite the proximity of the SMA and CMAd, our knowledge of the anatomy in this region of the medial wall coupled with the high resolution of the 2DG technique has enabled us to see the differential patterns of activation in these two areas. In monkeys, neurons in both the SMA and in the CMAd are active during relatively simple tasks (Shima et al., 1991; Cadoret and Smith, 1997; Russo et al., 2002). However, our results suggest that the CMAd is not as involved as the SMA in visually guided reaching, at least in our tasks. This observation mirrors our previous results of differential activation in the CMAd and SMA associated with the performance of internally generated sequences of movements. In this case, the CMAd was the most intensely activated motor area on the medial wall of the hemisphere (Picard and Strick, 1997). Together, our results suggest that information processing in the two areas is in some respects very different.
On the other hand, although there are major anatomical differences in the connectivity of the SMA and PreSMA (Bates and Goldman-Rakic, 1993; Luppino et al., 1993; Lu et al., 1994; He et al., 1995; Picard and Strick, 1996a, 2001), we did not observe major differences in their patterns of activation during the TRACK and RANDOM tasks. Neuron recording studies do observe differences between SMA and PreSMA responses during various behavioral conditions, although few of these differences are absolute (Matsuzaka et al., 1992; Matsuzaka and Tanji, 1996; Shima et al., 1996; Clower and Alexander, 1998; Nakamura et al., 1998; Shima and Tanji, 2000). In other circumstances, the SMA and PreSMA appear to be similar (Shima and Tanji, 1998; Hernández et al., 2002). Apparently, our tasks did not include behaviors that were sufficient to cause differential activation of the SMA and PreSMA.
As noted in the Introduction, the SMA and the PreSMA were generally considered to be concerned with internally guided movements and the generation of movement sequences. This view was first prompted by reports from imaging studies that SMA activation is present during the preparation and performance of complex movement sequences, but not during the performance of simple movements (Orgogozo and Larsen, 1979; Roland et al., 1980). This view was further supported by the observation that some single neurons in the SMA of trained monkeys display enhanced or exclusive activity during the performance of sequential movements that were internally guided (Mushiake et al., 1990, 1991; Halsband et al., 1994; Shima and Tanji, 2000). On the other hand, the results from neuron recordings also provide evidence for SMA involvement in visually guided movements. When Mushiake et al. (Mushiake et al., 1991) compared the activity of SMA neurons during visually guided and internally generated tasks, they found that 61% of the task-related neurons displayed enhanced or preferential activity before or during the internally generated movements. However ~90% of the task-related neurons changed their activity before or during the visually guided movements. Furthermore, ~20% of the task-related neurons displayed enhanced or preferential activity during the visual task. Thus, although the SMA may be involved in aspects of internally guided movements, the present results emphasize that the contribution of this cortical area to the generation and control of visually guided movements should not be neglected.
We have previously summarized the growing evidence from human studies that the SMA is activated during ‘simple’ voluntary movements (Picard and Strick, 1996a, 2001; Petit et al., 1998; Deiber et al., 1999; Jenkins et al., 2000; Thickbroom et al., 2000; Crosson et al., 2001; Krainik et al., 2001; Weeks et al., 2001; Cunnington et al., 2002). In addition, single neuron recording studies in awake trained primates found that SMA neurons are active in relation to the performance of simple movements, such as a ‘key press’ (Brinkman and Porter, 1979; Tanji and Kurata, 1979, 1982; Okano and Tanji, 1987; Tanji et al., 1988; Thaler et al., 1988; Mushiake et al., 1991). Our observations confirm the results of these prior studies and add that SMA involvement in simple voluntary movements may be quite substantial. Intense activation was present throughout the arm area of the SMA in both of the visually guided reaching tasks we trained monkeys to perform. This implies that neuronal activity was widespread in the SMA during the performance of these relatively simple, externally guided movements.
The extensive activation in the SMA during visually guided reaching stands in sharp contrast with the more modest activation observed in monkeys who performed internally generated movements that were otherwise comparable to those of the TRACK task (Picard and Strick, 1997). For internally generated movements, several small and scattered foci of activation were located within the distal and proximal arm representation of the SMA [see Fig. 2A of Picard and Strick (Picard and Strick, 1997)]. Although in some pixels the intensity of activation was comparable to that found during visually guided reaching, the overall measure of activation in the SMA was about half as much as for the visually guided tasks. Activations in face representations on the medial wall (CMAr, SMA-face) and in the three control areas (see Methods) were not significantly different in the REM task and the tasks used here (Picard and Strick, 1997) (this study and unpublished observations). These observations suggest that, globally, more neuronal activity was present in the SMA for the control of visually guided reaching than for performance based on memorized sequence information. While our previous results confirmed that the SMA participates in aspects of internally generated movements, the present results clearly show that overall, the SMA is more activated during simple, externally guided movements like our tasks.
Our results raise a critical question: what does the presence of intense 2DG uptake in the SMA (and for that matter the PreSMA) mean? The precise relationship between measures of activation and neuronal activity is unclear. 2DG uptake is thought to reflect mostly presynaptic activity at both excitatory and inhibitory synapses (Jueptner and Weiller, 1995). In fact, there are circumstances where intense 2DG labeling is predominantly related to the activity of inhibitory synapses with the net effect being a reduction of neuronal activity at the site of uptake (Nudo and Masterson, 1986; Jueptner and Weiller, 1995). 2DG is also taken up by the cell bodies of excitatory and inhibitory neurons (McCasland and Hibbard, 1997). However, despite the uncertainty about the precise relationship between 2DG uptake and cellular processes, the location of 2DG uptake in the cerebral cortex has been shown to correspond closely to the location of active neurons (Juliano and Whitsel, 1987; Tootell et al., 1988; Gilbert and Weisel, 1989). Moreover, there is evidence that the relative change in cerebral metabolism at an activated site is directly related to the relative change in firing frequency of cortical neurons at that site (Smith et al., 2002). Thus, it is not unreasonable to propose that the 2DG uptake in the SMA (and PreSMA) during the visually guided tasks reflects or is closely related to increased activity of cortical neurons in these areas. Consequently, for the remainder of this discussion we will assume that increased 2DG uptake reflects increased activity of single neurons at the activation site.
The SMA, like M1, has efferents that project directly to the spinal cord (Macpherson et al., 1982a; Hutchins et al., 1988; Dum and Strick, 1991; He et al., 1995). Some SMA efferents make monosynaptic connections with motoneurons (Dum and Strick, 1996; Maier et al., 2002). It is possible to evoke limb movements by stimulation of the SMA with relatively low currents (Macpherson et al., 1982b; Mitz and Wise, 1987; Luppino et al., 1991; Matsuzaka et al., 1992). Some SMA neurons discharge before movement onset and before the muscle activity associated with reaching movements (Brinkman and Porter, 1979; Mushiake et al., 1991). In fact, a substantial proportion of SMA neurons have ‘M1-like’ properties in their pattern and timing of discharge during reaching movements to visual targets (Chen et al., 1991). From this perspective, the SMA activation we observed may lead to descending commands that assist in the generation of motor output directly at the level of the spinal cord. Of course this suggestion does not preclude SMA activation from influencing motor output at other levels of the neuraxis.
If the SMA makes an essential contribution to the generation of visually guided movements, then one would predict that performance of these movements would be disrupted by SMA inactivation or lesion. Studies on this issue have given inconsistent results. In many cases inactivation or lesion of the SMA in monkeys did not impair the performance of simple externally guided movements (Brinkman, 1984; Chen et al., 1995; Thaler et al., 1995; Kermadi et al., 1997; Shima and Tanji, 1998; Tanji, 2001). On the other hand, cooling in the region of the SMA/ PreSMA grossly altered parameters of performance (reaction time, force, errors) of a simple key press task triggered by sensory signals (Tanji et al., 1985). Similarly, prolonged reaction time and movement time were the main effects of SMA inactivation during performance of a sequential button push task (Nakamura et al., 1999). In humans, surgical resection of the SMA region that was specifically activated during simple movements was highly correlated with the occurrence of post-surgical motor deficits including hemiplegia (Krainik et al., 2001).
At this point, it is unclear why an SMA lesion produced a marked motor deficit in some cases and little deficit in others. Perhaps the SMA contribution to motor behavior is highly task dependent. The types of movement that tend to be spared by SMA lesion/inactivation in monkey studies were commonly intermittent movements to single targets performed either spontaneously (self-paced) or triggered by infrequent sensory stimuli (Chen et al., 1995; Thaler et al., 1995; Kermadi et al., 1997; Shima and Tanji, 1998). In contrast, our tasks were more challenging in terms of motor performance and required monkeys to make rapid, accurate responses to multiple spatial locations in the frontal plane. In this respect, it is noteworthy that SMA neurons exhibit a high degree of selectivity for spatial variables (Clower and Alexander, 1998; Russo et al., 2002). Thus, a critical test of SMA function would be to examine the consequences of an SMA lesion or inactivation on performance of a task like ours where activation of the cortical area is known to be high. This test would help to determine whether the SMA activation that we and others have seen is truly a reflection of its contribution to the generation of visually guided movements or a reflection of its involvement in processes that are associated with reaching, but not critical to the actual generation of motor output (Rushworth et al., 2002).
In summary, our results show that the SMA is strongly activated during the performance of relatively simple visually guided movements. The precise nature of the contribution of the SMA to this type of movements is unclear. However, the intensity and extent of its functional activation, as well as the coupling with activation in M1, suggest that the SMA contributes substantially to some fundamental aspects of motor behavior. Although historically disregarded, our results stress that these aspects of SMA function need to be further explored.
This material is based upon work supported in part by the Office of Research and Development, Medical Research Service and Rehabilitation R&D Service, Department of Veterans Affairs, and by USPHS NS24328 (PLS). We are grateful to Mike Page for the development of computer programs for data acquisition and analysis and to Michele O’Malley for expert technical assistance.
Send reprint requests to Nathalie Picard, Department of Neurobiology, University of Pittsburgh School of Medicine, W1640 Biomedical Science Tower, 200 Lothrop Street, Pittsburgh, PA 15261, USA. Email: firstname.lastname@example.org.
Address correspondence to Peter L. Strick, Department of Neurobiology, University of Pittsburgh School of Medicine, W1640 Biomedical Science Tower, 200 Lothrop Street, Pittsburgh, PA 15261, USA. Email: email@example.com.
|*Significantly greater than background.|
|PreSMA (arm)||12.5 ± 5.6||54.4 ± 1.3*||64.0 ± 11.1*|
|CMAr (face)||38.9 ± 0.7*||52.0 ± 11.4*||50.4 ± 14.2*|
|SMA (face)||50.8 ± 13.1*||56.5 ± 1.8*||64.0 ± 9.2*|
|SMA (arm)||6.8 ± 4.4||60.6 ± 8.3*||61.4 ± 8.0*|
|CMAd (arm)||3.3 ± 1.8||14.2 ± 17.1||11.3 ± 7.5|
|CMAv (arm)||8.5 ± 17.8||1.5 ± 37.7||14.3 ± 2.9|
|M1 (arm)||2.9 ± 11.6||56.8 ± 15.0*||56.3 ± 10.4*|
|*Significantly greater than background.|
|PreSMA (arm)||12.5 ± 5.6||54.4 ± 1.3*||64.0 ± 11.1*|
|CMAr (face)||38.9 ± 0.7*||52.0 ± 11.4*||50.4 ± 14.2*|
|SMA (face)||50.8 ± 13.1*||56.5 ± 1.8*||64.0 ± 9.2*|
|SMA (arm)||6.8 ± 4.4||60.6 ± 8.3*||61.4 ± 8.0*|
|CMAd (arm)||3.3 ± 1.8||14.2 ± 17.1||11.3 ± 7.5|
|CMAv (arm)||8.5 ± 17.8||1.5 ± 37.7||14.3 ± 2.9|
|M1 (arm)||2.9 ± 11.6||56.8 ± 15.0*||56.3 ± 10.4*|