Functional specialization in dorsal and ventral premotor areas

Abstract

The premotor cortex (PM) in the bilateral lateral hemisphere of nonhuman primates and the human has been implicated in the sensorial guidance of movements. This is in contrast to more medial motor areas that are involved more in the temporal structuring of movements based on memorized information. The PM is further subdivided into dorsal (PMd) and ventral (PMv) parts. In this chapter, we describe our attempts to find differences in the use of these two areas in a nonhuman primate for programming future motor actions based on visual signals. We show that neurons in the PMv are involved primarily in receiving visuospatial signals and in specifying the spatial location of the target to be reached. In contrast, neurons in the PMd are involved more in integrating information about which arm to use and the target to be reached. Thus, PMd neurons are more implicated than those of he PMv in the preparation for a future motor action.

Introduction

A large body of evidence has accumulated to show that the primate brain has multiple motor areas that individually are further subdivided into subareas (Matelli et al., 1985, Barbas and Pandya, 1987, Dum and Strick, 1996). The supplementary motor area (SMA) and rostrally adjacent presupplementary motor area (pre-SMA) and supplementary eye field (SEF) are located in the medial part of the frontal cortex, rostral to the primary motor cortex (MI). The cingulate motor areas (CMA), with their rostral and caudal subdivisions, are in the banks of the cingulate sulcus. Functional specializations in all these areas have been thoroughly reviewed (Passingham,1993, Tanji, 1994, Tanji, 1996, Rizzolatti et al., 1998, Tanji, 2001; Rouiller et al., this volume). In addition, at least two limb-motor areas are known to exist rostral to MI in the premotor cortex (PM), which is located in the lateral surface of the frontal cortex. Each area, termed the dorsal and ventral premotor cortex (PMd and PMv), has been suggested to be further divided into subareas (Gentilucci et al., 1988, Rizzolatti et al., 1988, Kurata, 1994, Fujii et al., 2000). The role played by these two areas has been discussed previously (Wise, 1985, Wise et al., 1997, Rizzolatti et al., 1998). Their characteristics are far from understood, however. In this report, we briefly review previous lesion and unit-recording studies, which compared the motor function of these two areas. Finally, we present our view on the selective use of the PMd and PMv in the planning and execution of visually guided limb movements (see also Weber and He, Chapter 45 of this volume).

In his pioneering lesion study, Passingham (1985) showed that the PM is essential for motor selection based on visual cues. Subsequent lesion studies have revealed functional deficits selective to lesions restricted to the PMd versus PMv. PMv lesions have revealed attentional deficits in the peripersonal space within a reaching distance (Rizzolatti et al., 1983), and deficits in shift-prism adaptation (Kurata and Hoshi, 1999) and hand-preshaping for grasping objects (Fogassi et al., 2001). A tendency to select an object ipsilateral to the lesioned PMv has also been reported (Schieber, 2000). On the other hand, lesions in the PMd were shown to lead to an inability in planning adequate wrist movements based on visual conditional cues (Kurata and Hoffman, 1994). Importantly, the above deficits were observed in the absence of difficulties in perceiving visual information, or in executing limb movements per se (Passingham, 1985, Kurata and Hoshi, 1999).

Neurons in the PMd and PMv respond to the appearance of visual signals instructing future movements and they exhibit sustained activity during the subsequent motor-set period. Their discharge is related to different aspects of forthcoming motor behavior, however. Activity of PMd neurons reflects forthcoming movement parameters like direction and amplitude (Fu et al., 1993, Kurata, 1993, Crammond and Kalaska, 2000), movement trajectory (Hocherman and Wise, 1990), and a combination of target location and movement trajectory (Shen and Alexander, 1997). On the other hand, activity of PMv neurons reflects target location (Gentilucci et al., 1988, Mushiake et al., 1997), the three-dimensional shape of motor targets (Murata et al., 1997) and peripersonal space (Graziano et al., 1997). Boussaoud and Wise, 1993a, Boussaoud and Wise, 1993b found differences in the firing patterns PMd and PMv neurons in monkeys performing a behavioral task. Most PMd neurons were more active when the direction of a future movement was instructed, whereas most PMv neurons were more active when a locus of attention was indicated. Furthermore, PMd neuronal activity reflected the direction of movements rather than the characteristics of the visual cues.

In summary, both lesion and unit-recording studies have suggested differences in the use of PMd and PMv for the visual guidance of movements and for instructing future actions by use of visual information. We describe in the next section, the extension of these findings by defining more precisely how each of the two areas is involved in retrieving information out of visual signals and in formulating plans for future actions.

When intending to initiate an action, an object in the outside world is selected and a decision made concerning which part (effector) of the body is to be used for the action. For example, to plan an arm-reach movement, different sets of information are required for the selection (choice) of both the target to reach and the arm to use. These sets must then be integrated in order to specify an appropriate motor program for the arm-reach. If the source of information is visual signals, then the most plausible site of information processing for the motor planning is neuronal circuitry within the PM. With this construct in mind, we designed a series of experiments to investigate the role of PMd versus PMv in retrieving, processing and integrating information to plan arm-reaching movements.

We trained two monkeys (Macaca fuscata) to plan and prepare future movements following two instruction cues (Hoshi and Tanji, 2000). One cue indicated the arm-choice (right vs. left) for the reaching movement, and the other indicated the target-choice (again right vs. left). Each instruction consisted of a color cue and a white square. The color cue indicated whether the instruction was for arm-choice or for target-choice, and the location of the white square indicated the side (right vs. left) for arm-choice or target-choice. Since the two instructions were presented separately and sequentially in a randomized order, the monkeys were required to (1) retrieve information for arm-choice versus target-choice indicated by the first cue, (2) retain this information, (3) retrieve information indicated by the second cue, and (4) integrate the first and second set of information in order to plan and prepare future movements.

Neuronal activity was recorded in the PMd and PMv while the monkey was performing the task. After the appearance of the first cue, different groups of PMd neurons extracted and retained specific information about arm-choice and target-choice. For example, Fig. 1 shows that the activity of Cell 1 (in the PMd) was most active when the first cue instructed right-arm choice. On the other hand, activity of Cell 2 (also in the PMd) was prominent when the first cue instructed reaching to the correct target. The existence of these two types of activity suggested that PMd neurons selectively collect arm-choice and target-choice information, and retain this information before the second information (instruction) is provided.

In contrast to PMd cells, those tested in PMv were mainly selective during the first cue for the physical location of the white square, with far fewer encountered that selectively represented arm-choice or target-choice information. The second-cue responses were also different for PMd versus PMv cells. Many neurons in PMd, but not PMv, integrated the two sets of information and specified future action. For example, Fig. 1 shows that Cell 3 (another PMd cell) was active in the second delay if the two cues indicated right-arm choice and left-target choice, regardless of the order of instruction presentation. This result suggested that the PMd is more involved than PMv in integrating the arm-choice and target-choice information. Furthermore, during the motor-set period, while the monkey prepared a future movement just before the appearance of the GO signal, the two factors were differently represented in PMd versus PMv. Selectivity for the target was more represented in the PMv. In contrast, selectivity for the arm-choice was more represented in the PMd. Interestingly, the number of PMd neurons encountered that were selective for arm-choice versus target-choice was quite similar, thereby possibly further supporting the concept that the integration of arm-choice and target-choice occurs in PMd. In contrast, PMv neurons mainly represented target-choice information rather than arm-choice.

The above results indicate that neurons in PMd have a capacity to selectively represent information about arm-choice versus target-choice. Once the two sets of information become available in PMd, they are quickly integrated. During the motor-set period, PMd neurons also achieve a representation of future action (i.e., a hybrid of arm-choice and target-choice information). In contrast, neurons in PMv mainly represented right versus left aspects of the two instruction cues, and, during the motor-set period, they tend to represent location of the target, regardless of the effector (arm) to be used.

As reviewed in the previous section, many lines of evidence now point to functional specificity within the PMd and PMv. Based on these results, we propose some functional differences between PMd and PMv neurons and their circuitry. PMd retrieves and holds both target information and body-part information, and integrate the two sets of information to plan an action. PMv mainly represents the nature of targets, regardless of the specific effectors required to reach them.

How does the above concept relate to the broad array of previous findings on these two structures? The loss of sensorimotor integration in planning movements after a PMd lesion can be viewed as a deficit in retrieving information and/or in integrating multiple sets of information, such as the choice of an effector to reach a specific target. Many aspects of neuronal activity found in the PMd can be viewed as collecting and integrating component information of movements, and planning and preparing future movements. On the other hand, a PMv-lesion-evoked deficit in locating a target or in specifying a three-dimensional shape can be explained as a failure in representing properties of targets that are relevant to an intended action. It has been shown that neurons in PMv represent the nature of targets for action, such as the location of motor targets or their three-dimensional shape (Rizzolatti et al., 1998, Hoshi and Tanji, 2001). These functional properties, which are unique to the PMd and PMv, seem to be provided by the wealth of anatomical connections with other areas in the CNS, such as the parietal and prefrontal cortex. In future research, it will be necessary to clarify how the PMd and PMv operate as part of global networks for the control of movement.