Journal of Neurophysiology

Modulating Neural Networks With Transcranial Magnetic Stimulation Applied Over the Dorsal Premotor and Primary Motor Cortices

Philippe A. Chouinard, Ysbrand D. Van Der Werf, Gabriel Leonard, Tomás Paus

Abstract

Our study uses the combined transcranial magnetic stimulation/positron emission tomography (TMS/PET) method for elucidating neural connectivity of the human motor system. We first altered motor excitability by applying low-frequency repetitive TMS over two cortical motor regions in separate experiments: the dorsal premotor and primary motor cortices. We then assessed the consequences of modulating motor excitability by applying single-pulse TMS over the primary motor cortex and measuring: 1) muscle responses with electromyography and 2) cerebral blood flow with PET. Low-frequency repetitive stimulation reduced muscle responses to a similar degree in both experiments. To map networks of brain regions in which activity changes reflected modulation of motor excitability, we generated t-statistical maps of correlations between reductions in muscle response and differences in cerebral blood flow. Low-frequency repetitive stimulation altered neural activity differently in both experiments. Neural modulation occurred in multiple brain regions after dorsal premotor cortex stimulation; these included motor regions in the frontal cortex as well as more associational regions in the parietal and prefrontal cortices. In contrast, neural modulation occurred in a smaller number of brain regions after primary motor cortex stimulation, many of these confined to the motor system. These findings are consistent with the known differences between the dorsal premotor and primary motor cortices in the extent of cortico-cortical anatomical connectivity in the monkey.

INTRODUCTION

The cortical motor system can be separated into the primary motor and the nonprimary motor areas. The nonprimary motor areas are defined as all regions in the frontal lobe that have the potential to influence motor output at the level of both the primary motor cortex and the spinal cord (Dum and Strick 1991); these include the premotor, supplementary motor, and cingulate motor areas. Transcranial magnetic stimulation (TMS) applied in trains of pulses can modulate the motor system in a temporary fashion, lasting beyond the duration of stimulation. Studies that have examined these effects generally applied repetitive stimulation over the primary motor cortex and measured the modulation of motor-evoked potentials (MEPs) recorded in the contralateral hand muscles. Typically, low-stimulation frequencies of 1 to 2 Hz induce inhibitory effects (e.g., Chen et al. 1997; Gerschlager et al. 2001; Maeda et al. 2000; Muellbacher et al. 2000) and high-stimulation frequencies between 5 and 20 Hz induce facilitory effects (e.g., Maeda et al. 2000; Pascual-Leone et al. 1994; Peinemann et al. 2000; Romeo et al. 2000). Cortical mechanisms are believed to mediate both inhibitory (Touge et al. 2001) and facilitory (Baradelli et al. 1998) effects.

Recent studies demonstrate that low-frequency repetitive TMS applied over the premotor cortex can also induce changes in motor excitability as reflected by: 1) decreases in the amplitude of MEPs elicited by single-pulse stimuli (Gerschlager et al. 2001); 2) increases in intracortical facilitation to paired-pulse stimuli (Münchau et al. 2002); and 3) reductions in the duration of the silent period (Münchau et al. 2002). These results suggest that repetitive stimulation over the premotor cortex can also modulate the output of the motor system, mediated perhaps by direct cortico-cortical connections between the premotor and primary motor cortices.

The question we address here is whether repetitive TMS applied over the dorsal premotor cortex, and over the primary motor cortex in a separate experiment, can alter neural activity at distal sites connected synaptically. Previous studies (Bohning et al. 1999; Fox et al. 1997; Paus et al. 1997, 1998; Siebner et al. 1998, 2000) have established the combination of functional brain imaging and TMS as an effective method to measure changes in neural activity induced by repetitive stimulation (reviewed in Paus 2002). Previous positron emission tomography (PET) studies have already described changes in blood flow and glucose metabolism during repetitive stimulation applied over the primary motor cortex (Fox et al. 1997; Paus et al. 1998; Siebner et al. 2001). These studies reveal focal changes in the primary motor cortex as well as in distant regions known to be connected synaptically in the monkey, including the premotor and supplementary motor areas. These results suggest that for certain neural networks, connectivity patterns identified in monkeys are similar in humans.

By applying repetitive TMS over two subdivisions of the cortical motor system in the same group of subjects, we can potentially map two networks and the manner in which each is modulated. Because repetitive TMS applied over the dorsal premotor cortex or the primary motor cortex can reduce MEP amplitudes, we used this change in MEP as an index of effectiveness for altering neural activity by repetitive stimulation. To map networks of brain regions in which activity changes reflected modulation of motor excitability, we generated t-statistical maps of correlations between reductions in muscle response and differences in cerebral blood flow (CBF).

METHODS

We acquired a total of six 60-s 15O-H2O PET scans in each of two sessions. In both sessions, we scanned subjects during the following conditions: 1) no TMS before repetitive stimulation (Base); 2) single-pulse TMS before repetitive stimulation (Pre); 3) single-pulse TMS shortly after repetitive stimulation (Post1); 4) single-pulse TMS about 10 min after repetitive stimulation (Post2); 5) single-pulse TMS about 20 min after repetitive stimulation (Post3); and, 6) single-pulse TMS about 30 min after repetitive stimulation (Post4). We counterbalanced the order of the Base and Pre conditions across subjects. During 5 of the 6 PET scans, we applied 12 suprathreshold single pulses of TMS over the left primary motor cortex while recording MEPs in the right first dorsal interosseous muscle. Between the second and third PET scans, we applied a 15-min train of 1-Hz subthreshold repetitive TMS over the left dorsal premotor cortex in one experiment and over the left primary motor cortex in the other experiment. These experiments were conducted on separate days and in a counterbalanced order; we refer to these as the dorsal premotor and primary motor experiments.

We calculated the amount of MEP reduction induced by repetitive stimulation for every single-pulse condition and used this measure as an index of effectiveness in modulating neural activity. To reveal brain regions modulated by the repetitive stimulation, we performed correlations between reductions in MEP and differences in CBF. In the single-pulse conditions, we applied 20 single pulses of TMS on average every 5 s (range: 4 to 6 s to minimize anticipation), 12 of which occurred during PET scanning.

Subjects

Four female and three male right-handed subjects (19 to 27 yr of age, mean ± SD, 23±3) participated in the study after giving informed written consent. The Research Ethics Board of the Montreal Neurological Institute and Hospital approved all experimental procedures. We preselected subjects for their low resting motor thresholds (rMT) to prevent overheating of the stimulating coil. We determined thresholds for the relaxed right first dorsal interosseus muscle before both experiments by first determining the optimal position for activating the muscle and then by reducing the stimulation intensity (in 1% steps) from an initial suprathreshold level until we found the lowest stimulus intensity sufficient to induce 5 MEPs of ≥50 μV in a series of 10 stimuli applied at least every approximately 5 s. We also preselected for right-handed subjects as determined by a handedness questionnaire (Crovitz and Zener 1965).

Transcranial magnetic stimulation

We carried out TMS using a Cadwell (Kennewick, WA) high-speed magnetic stimulator and a Cadwell figure-of-eight stimulating coil (Corticoil, 2 tear-shaped coils of approximately 5-cm diameter each). We chose this coil because it produces a magnetic-field maximum of sufficiently small width to allow stimulation of the dorsal premotor cortex without encroaching on the primary motor cortex. In the scanner, a mechanical arm held the coil over the optimal position for eliciting a muscle twitch in the right index finger. We used a suprathreshold intensity of 115% rMT for single-pulse TMS and a subthreshold intensity of 90% rMT for repetitive TMS. Subthreshold intensities allow for more focal stimulation by narrowing the magnetic field produced by the coil, thus enabling better spatial resolution for examining changes between the location of stimulation and more distant cortical structures (Gerschlager et al. 2001; Münchau et al. 2002; Pascual-Leone et al. 1993).

Targeting the stimulation locations

We used a four-step procedure to place the stimulating coil over our stimulation locations. This procedure, developed in our first TMS/PET study (Paus 1999; Paus et al. 1997), takes advantage of standardized stereotaxic space (Talairach and Tournoux 1988). First, we acquired magnetic resonance (MR) images (170 contiguous 1-mm-thick sagittal slices) of the subject's brain using a Siemens Vision 1.5-T system and transformed these images into standardized stereotaxic space using an automatic feature-matching algorithm (Collins et al. 1994). Second, we derived locations for the primary motor and dorsal premotor cortices using information gained in previous brain imaging studies. We derived a probabilistic location for the primary motor cortex (X = –31, Y = –22, Z = 52; Paus et al. 1998) by averaging the coordinates reported in eight previous studies examining blood-flow activation when subjects moved the fingers of their right hand (Colebatch et al. 1991; Dettmers et al. 1995; Grafton et al. 1993; Jahanshahi et al. 1995; Jenkins et al. 1994; Matelli et al. 1993: Paus et al. 1993; Schlaug et al. 1994). This location served as an estimate as to where we should place the TMS coil relative to the subject's head in the scanner; subsequent adjustments in coil positioning were made (see following text). We defined a location for the dorsal premotor cortex (X = –21, Y = –2, Z = 52) as being 10 mm medial and 20 mm anterior to the probabilistic location of the primary motor cortex. This location was estimated by a PET study carried out by Fink et al. (1997) and was used in a previous TMS study of the premotor cortex (Schulter et al. 1998). Third, we transformed these two locations to the subject's brain coordinate space using an inverse version of the native-to-standardized transformation matrix.

The final step required us to position the coil over these locations, now marked on the MR images, which we achieved using frameless stereotaxy. With the subject lying on the couch of the scanner, we first registered the subject's head with the MR images and then placed the coil over the target locations by tracking the position and three-dimensional orientation of the coil with an infrared optical-tracking system (Polaris System, Northern Digital, Waterloo, Ontario, Canada, and Brainsight software, Rogue Research, Montreal, Quebec, Canada). We then locked the coil in place after finding these locations. In the case of the primary motor cortex, we made further adjustments in coil positioning to where stimulation resulted in the maximum MEP amplitude. To ensure that we used the same position for subsequent coil placements over the primary motor cortex, we first defined its position in the subject's brain coordinate space and then marked its position on the subject's MR images. We held the coil in different orientations when stimulating the primary motor and dorsal premotor cortices. For the primary motor cortex, we oriented the coil tangentially to the scalp with the short axis of the figure-of-eight coil angled at 45° relative to the interhemispheric fissure and approximately perpendicular to the central sulcus. For the dorsal premotor cortex, we oriented the coil tangentially to the scalp with the short axis of the figure-of-eight coil perpendicular to the interhemispheric fissure. For primary motor and dorsal premotor stimulation, the resulting induced electric current in the brain flowed in posterior-to-anterior and lateral-to-medial directions, respectively.

Verifying final coil positions over the primary motor cortex

Interpretation of results acquired with TMS/PET depends critically on the accuracy of coil positioning. In the present study, this applies specifically to the dorsal premotor experiment where we moved the coil from the primary motor cortex (Pre scan) to the dorsal premotor cortex (between scans 2 and 3) and then back to the primary motor cortex (Post scans). Using a procedure described in detail elsewhere (Paus and Wolforth 1998), we used 10-min transmission scans to verify coil positions relative to the acquired PET and MR images. We acquired transmission scans at the beginning of the dorsal premotor and primary motor experiments, and an additional transmission scan at the end of the dorsal premotor experiment. These transmission images showed us the coil's position relative to the subject's head. We then registered an X-ray image of the coil to these images and projected a straight rod orthogonal to the plane of the coil from the coil center. After PET-to-PET, PET-to-MR, and MR-to-standardized space transformations, we superimposed the locations of the rod on an average anatomical MR image of all subjects. This indicates the projected center of the coil in the brain; the figure-of-eight coil used in this study stimulates an estimated volume of 20 × 20 × 10 mm (Cohen et al. 1990; Maccabee et al. 1990; Wassermann et al. 1996).

Positron emission tomography

We instructed subjects to relax and keep their eyes closed during PET scanning. Subjects used a bite-bar to maintain a constant head position during the experiments. We measured CBF with a CTI/Siemens HR+ 63-slice tomograph scanner operated in three-dimensional (3D) acquisition mode during 60-s scans using the 15O-labeled H2O bolus method (Raichle et al. 1983). In each scan, we injected 10 mCi of 15O-labeled H2O into the left antecubital vein. Acquired CBF images were reconstructed with a 14-mm Hanning filter, normalized for differences in global CBF (normalized CBF), coregistered with the individual MR images (Woods et al. 1993), and transformed into standardized stereotaxic space (Talairach and Tournoux 1988) by means of an automated feature-matching algorithm (Collins et al. 1994). We placed four 0.5-mm-thick sheets of well-grounded mumetal to protect the photomultipliers inside the PET scanner from the effects of the coil-generated magnetic field. The mu-metal, however, can attenuate gamma rays and in turn decrease the number of detected coincidence counts (Paus 2002). The transmission data acquired at the beginning of the experiments were also used to correct for the attenuation of gamma rays caused by all objects in the scanner, including the coil, the coil mount, and the metal sheets.

Analyses of muscle-evoked potentials

We recorded MEPs from the right first dorsal interosseus muscle using Ag/AgCl surface electrodes fixed on the skin with a bellytendon montage. We sampled the electromyographic (EMG) signal using an EMG channel of a 60-channel TMS-compatible electroencephalography system (Virtanen et al. 1999) with the amplifier's bandwidth set at 0.1–500 Hz and the sampling rate set at 1.45 kHz. We measured the peak-to-peak amplitudes for each MEP off-line. For practical reasons, we began to deliver single pulses of TMS at the time the radioactive tracer was injected. Acquisition after injection varies from one person to another and there is no way of knowing exactly which of the MEPs occurred during scanning. We therefore calculated the muscle response for a given scan as a percentage of the mean MEP amplitude during the Pre scan based on the 20 trials. We evaluated the effects of repetitive TMS on motor excitability by analysis of variance (ANOVA) using a model of repeated measures with Time as a within-subject factor. We used Tukey's HSD tests, which correct for multiple comparisons, for all post hoc pairwise comparisons. We considered values statistically significant at P < 0.05. We also performed a Wilcoxon signed ranks test to determine whether rMT values were significantly different between the two experiments.

Analyses of cerebral blood flow

We used a two-step process to generate t-statistical maps. We first subtracted CBF acquired before repetitive TMS from CBF acquired after repetitive TMS. We performed this initial subtraction to obtain CBF differences contrasting scans obtained before and after repetitive TMS, and to remove confounding intersubject variability. We then correlated these subtractions with the relative amount of MEP reduction, which we calculated in the same way as in a previous TMS/PET study (Strafella and Paus 2001) that examined the effects of double-pulse stimulation on CBF: {[1 – (MEP amplitude at a given post-rTMS condition/MEP amplitude at the pre-rTMS condition)] × 100}. We carried out calculations for the t-statistical maps for each of the 3D volume elements (voxels) constituting the entire scanned volume, which tested whether at a given voxel the slope of the regression was significantly different from zero.

After generating our t-statistical maps, we evaluated the presence of a significant peak by a method based on a 3D Gaussian random-field theory, with correction for the multiple comparisons involved in searching the entire volume (Worsley et al. 1992). Using this method, we performed both an exploratory search of the entire brain and a directed search in specific brain regions. For an exploratory search, we considered values equal to or exceeding a criterion of t = 4.5 as significant (P < 0.000003, 2-tailed, uncorrected), yielding a false positive rate of 0.04 (corrected) in 400 resolution elements (each of which has dimensions 14 × 14 × 14 mm) for a brain volume of 1,100 cm3. For a directed search, we considered values equal to or exceeding a criterion of t = 3.5 as significant (P < 0.0002, 2-tailed, uncorrected), yielding a false positive rate of 0.01 (corrected) in two resolution elements (each of which has dimensions of 14 × 14 × 14 mm) for a volume of 5 cm3. We performed our directed search in the dorsal premotor cortex, in the primary motor cortex, and in brain regions known to be connected with these regions in nonhuman primates (Fig. 1). To perform this search in the human brain, we relied on previous functional brain imaging studies that mapped their putative homologues; we describe these later in the discussion. We determined anatomical locations of all significant t-statistic peaks by examining the merged image of our t-statistical maps with the transformed averaged MR image of all subjects in standardized stereotaxic space (Talairach and Tournoux 1988). We performed additional subtractions of the CBF data to examine primarily the local effects at the stimulation sites. The first subtraction examined the possible, but unlikely, local effects of single-pulse TMS and consisted of subtracting CBF in the Base scan from CBF in the Pre scan. The second subtraction examined the presence of local effects of repetitive TMS and consisted of subtracting CBF in the Pre scan from the average CBF of all Post scans. All brain regions that showed significant CBF changes are reported.

fig. 1.

Overview of possible connections in human cerebral cortex derived from anatomical studies performed by others in monkey. A: predicted brain regions connected with dorsal premotor cortex. B: predicted brain regions connected with primary motor cortex (reviewed in Matelli and Luppino 2000 for parieto-frontal circuits; Matelli and Luppino 1997 for functional anatomy of human and nonhuman primate motor cortical areas; and Parent and Hazrati 1995 for cortical–subcortical connections). No distinction is made between left and right hemispheres. Dorsal premotor cortex in this schematic consists of two anatomical areas [i.e., caudal (F2) and rostral (F7)], each with distinct interconnections with other brain areas. PE, PEc-PEip, and PGm are cytoarchitectonically defined areas of parietal cortex. F1 to F7 are subdivisions of cortical motor system in frontal lobe. Abbreviations: PMd, dorsal premotor area; M1, primary motor area; CMA, cingulate motor area; MIP, medial intraparietal area; and SMA, supplementary motor area.

We also provide additional analyses that examine similarities and differences between the effects of repetitive TMS over the dorsal premotor and primary motor cortices. To examine similarities, we carried out a conjunction analysis (Price and Friston 1997). This analysis tests for the presence of correlations in both experiments by revealing the maximum peaks in the two contrasts. We considered values equal to or exceeding a criterion of t = 3.5 as significant (Worsley and Friston 2000). To examine differences, we directly tested for differences in the CBF difference/MEP reduction relationship between the dorsal premotor and primary motor experiments. We first extracted CBF values from volumes of interest (VOIs) centered at the X, Y, and Z coordinates of our correlation peaks (Tables 2 and 3) and then, for each brain region, we used ANOVA to test for differences in the slope of their correlations between the two experiments. We used Bonferroni corrections to take into account multiple comparisons and considered values statistically significant at P (corrected) < 0.05.

View this table:
table 2.

Effects of repetitive stimulation over the dorsal premotor cortex on cerebral blood flow

View this table:
table 3.

Effects of repetitive stimulation over the primary motor cortex on cerebral blood flow

RESULTS

All subjects tolerated the study well without noticeable adverse effects related to TMS and/or the scanning procedures. We excluded data from two subjects in the dorsal premotor experiment from our analyses because of head movement. Figure 2 illustrates the end result of all coil placements over the primary motor cortex for both experiments.

fig. 2.

Verification of coil positioning over primary motor cortex. Superimposed are virtual rods derived from transmission scans that indicate end result of all coil placements over primary motor cortex. Two spheres represent probabilistic locations for dorsal premotor (PMd) and primary motor (MI) cortices.

Effects of repetitive TMS on MEP amplitudes

Repeated-measures ANOVA revealed a significant effect of Time on the mean MEP amplitude in the dorsal premotor experiment [F(4,16) = 3.48, P < 0.05] and in the primary motor experiment [F(4,26) = 3.11, P < 0.05]. These results indicate that MEP amplitudes changed during the course of the two experiments. Using Tukey's HSD pairwise comparison tests, we further examined the pattern of MEP changes (Fig. 3). The second Post scan showed significantly smaller MEP amplitudes compared with baseline in both experiments (both P < 0.05). No other pairwise comparisons differed significantly. A Wilcoxon signed ranks test revealed that there was no significant difference between rMT values in the two experiments [W(4) = –1.83, P = 0.07; subjects 6 and 7 excluded). Although this was not significant, rMT values tended to be lower in the dorsal premotor experiment compared with the primary motor experiment (Table 1).

fig. 3.

Effects of repetitive stimulation on motor-evoked potentials. Mean (± SE) percentage motor-evoked potential (MEP) amplitude change at Post conditions compared with Pre conditions in both dorsal premotor and primary motor experiments. Asterisks denote significant differences compared with Pre conditions (*P < 0.05).

View this table:
table 1.

rMTs for each subject

Effects of repetitive TMS over the dorsal premotor cortex on CBF

Figure 4A and Table 2 summarize the findings in the dorsal premotor experiment and show all brain regions that presented significant positive and negative correlations between Post-Pre CBF differences and the amount of MEP reduction. Figure 5, A and B provides plots of CBF differences versus MEP reduction for two of these brain regions: the right anterior parietal and ventral premotor cortices. Motor-related regions with positive correlations include: the left and right ventral premotor areas in the precentral region of the operculi, the left and right cingulate motor areas in the cingulate gyri/sulci, the right premotor area in the precentral sulcus, the right supplementary motor area in the medial frontal gyrus, and the right putamen. Motor-related regions with negative correlations include: the left dorsal premotor area in the precentral gyrus/sulcus (30 mm lateral and 9 mm caudal to the targeted site of repetitive TMS and unlikely to indicate a local effect of stimulation) and the right sensorimotor area in the paracentral lobule.

fig. 4.

Effects of repetitive stimulation on cerebral blood flow. A, top half: horizontal slices of brain regions with positive correlations (t > 3.5) obtained in PMd experiment. B, bottom half: horizontal slices of brain regions with positive correlations (t < 3.5) obtained in M1 experiment.

fig. 5.

Cerebral blood flow (CBF) differences plotted vs. the amount of reduction in MEPs. Figure shows extracted CBF values using VOIs centered at X, Y, and Z coordinates of three correlation peaks. A and B: extracted CBF values with VOIs centered at right anterior intraparietal and ventral premotor cortices in dorsal premotor experiment. C: extracted CBF values with VOI centered at right primary motor cortex in primary motor experiment. Abbreviations: AIP, putative anterior intraparietal area; PMv, ventral premotor area; and M1, primary motor area.

Parietal brain regions with positive correlations include: the right posterior portion of the superior parietal lobule/intraparietal sulcus (putative medial intraparietal area), the anterior portion of the right inferior parietal lobule/intraparietal sulcus (putative anterior intraparietal area), and the right inferior parietal lobule/postcentral sulcus. Prefrontal brain regions with positive correlations include: the left and right inferior frontal gyrus/sulcus (ventrolateral prefrontal cortex) and the right middle frontal gyrus/sulcus (dorsolateral prefrontal cortex). One medial temporal-lobe region with a positive correlation was found in the right hippocampus. Negative correlations were mostly confined to several areas in the primary and associational visual cortices.

No significant correlations occurred either at the local site of repetitive TMS (i.e., left dorsal premotor cortex) or at the site of single-pulse TMS (i.e., left primary motor cortex). Further examination using direct subtraction analyses did not reveal significant CBF changes at either of the two sites of stimulation, which equally suggests that no local effects of TMS occurred in the left dorsal premotor cortex or in the left primary motor cortex. A direct subtraction of the Base scan from the Pre scan revealed CBF increases in the left presupplementary area on the medial frontal gyrus (X = –5, Y = 15, Z = 51; t = 3.8) and CBF decreases in the right superior parietal lobule/intraparietal sulcus (putative medial intraparietal area; X = 31, Y = –64, Z = 54; t = –4.1). A direct subtraction of the Pre scan from the average of all Post scans revealed no significant CBF differences anywhere in the brain.

Effects of repetitive TMS over the primary motor cortex on CBF

Figure 4B and Table 3 summarize the findings in the primary motor experiment and show all brain regions that presented significant positive and negative correlations between Post-Pre CBF differences and the amount of MEP reduction. Figure 5C provides a plot of CBF differences versus MEP reduction for one of these brain regions, the right primary motor cortex. Motor-related regions with positive correlations include: the left cingulate motor area in the cingulate gyrus/sulcus, the left putamen, the right primary motor area in the precentral gyrus/central sulcus, the right ventral-lateral thalamic nucleus, and the left cerebellum. Negative correlations were mostly confined to several areas in the primary and associational visual cortices.

No significant correlation occurred at the location of single-pulse TMS and repetitive TMS (i.e., left primary motor cortex). A direct subtraction of the Base scan from the Pre scan did not reveal any local changes in CBF. The same subtraction revealed CBF increases in the left primary visual cortex in the calcarine sulcus (X = –4, Y = –86, Z = 10; t = 5.0) and the right primary visual cortex in the calcarine sulcus (X = 7, Y = –71, Z = 14; t = 4.8). A direct subtraction of the Pre scan from the average of all Post scans, however, revealed a near significant increase of CBF at the stimulated region (X = –35, Y = –26, Z = 51, t = 3.2). The same subtraction also revealed CBF increases in the right cingulate motor area in the cingulate gyrus/sulcus (X = 1, Y = 18, Z = 45; t = 3.7) and in the left dorsal premotor cortex in the superior frontal sulcus (X = –33, Y = 6, Z = 52; t = 3.6), as well as CBF decreases in the left primary visual cortex in the calcarine sulcus (X = –5, Y = –85, Z = 12; t = 4.9). These two subtraction-based results suggest local effects of repetitive TMS but not of single-pulse TMS in the left primary motor cortex.

Conjunction analysis

Table 4 summarizes the findings of our conjunction analysis and lists all brain regions that presented significant correlations between Post-Pre CBF differences and the amount of MEP reduction in both experiments. Brain regions with significant positive correlations include: the right hippocampus and the right mesencephalon, both of which were approximately in the same horizontal plane (Z between –12 and –16). Except for one location in the right cerebellum, brain regions with significant negative correlations were all confined to the primary and associational visual cortices.

View this table:
table 4.

Similarities between the effects of repetitive stimulation over the dorsal premotor and primary motor cortices on cerebral blood flow

Contrast analysis

Table 5 summarizes the findings from our ANOVA that tested for differences in the CBF difference MEP reduction relationship between the dorsal premotor and primary motor experiments. We also present in this table Pearson's correlation coefficients between CBF differences and the amount of MEP reduction. Overall, our analysis confirms minimal overlap in the effects of repetitive TMS applied over the dorsal premotor and primary motor cortices on possible fronto-parietal circuits. Similar to the results in the conjunction analysis, the right mesencephalon showed relatively large Pearson's correlation coefficients for both experiments, suggesting strong positive relationships between CBF differences and the amount of MEP reduction; although these still showed significantly different relationships. Contrary to the results in the conjunction analysis, the right hippocampal formation showed a small Pearson's correlation coefficient for the primary motor experiment. This is likely because we extracted VOI at its correlation peak in the dorsal premotor experiment, which was about 5 mm more medial and 5 mm more dorsal than its activation peak reported in the conjunction analysis.

View this table:
table 5.

Differences between the effects of repetitive stimulation over the dorsal premotor and primary motor cortices on cerebral blood flow

DISCUSSION

Our results demonstrate that low-frequency repetitive TMS applied over the dorsal premotor and primary motor cortices produced similar inhibitory effects on MEPs but influenced cerebral activity differently. Repetitive stimulation over the dorsal premotor cortex resulted in the modulation of a network encompassing a number of brain regions; these include several regions in the parietal and prefrontal cortices. In contrast, repetitive stimulation over the primary motor cortex resulted in the modulation of a network encompassing a smaller number of brain regions; many of these were confined to the cortical and subcortical motor system. In the ensuing discussion we first address methodological issues and then discuss our findings in the light of studies performed by others in the monkey.

Methodological issues

Although we showed significant reductions in motor excitability after both applications of repetitive stimulation over the dorsal premotor and primary motor cortices, we also noted considerable interindividual differences. Some subjects showed greater reductions in MEP amplitude (1, 3, 5, and 6) compared with others (2 and 7), and one subject (4) showed increases in MEP amplitude. This is consistent with previous findings suggesting that it might be necessary to individualize parameters of repetitive TMS to achieve a consistent change in motor excitability across all subjects (Maeda et al. 2000). It is unlikely that this variability resulted from changes in coil positioning. Verifications of final coil positioning showed that we placed the coil consistently over the primary motor cortex. Most subjects showed minimal head movements as evident from their blood-flow images; we excluded two subjects who had head movements, in the dorsal premotor experiment, from the analyses.

Similar to Gerschlager et al. (2001) and Münchau et al. (2002), we demonstrated changes in motor excitability after applying repetitive TMS over the dorsal premotor cortex. Unlike the aforementioned studies, we held the coil in different orientations when stimulating the dorsal premotor and primary motor cortices. We chose different coil orientations to reduce the likelihood that stimulation of the dorsal premotor cortex would encroach on the primary motor cortex. Also, unlike the aforementioned studies, we demonstrated a reduction of motor excitability after repetitive TMS over the primary motor cortex. We might have had better access to the primary motor cortex by stimulating at a higher intensity (90% rMT as opposed to 80–90% active MT); as suggested by Gerschlager et al. (2001), it might be easier to stimulate the premotor cortex than relatively deeper structures like the primary motor cortex, located in the anterior bank of the central sulcus. The small figure-of-eight coil used in this study (diameter = 5 cm) delivers higher intensities while maintaining focality and stimulates an estimated volume of 20 × 20 × 10 mm (Cohen et al. 1990; Maccabee et al. 1990; Wassermann et al. 1996). It is unlikely, therefore, that the spread of current to premotor areas induced the effects obtained in the primary motor experiment.

MEPs obtained in the dorsal premotor and primary motor experiments are associated with changes in the size of muscle twitches and, presumably, differential sensory feedback from the hand muscles to the brain. This feedback could conceivably confound the blood-flow response. Two important features of our data argue against this possibility. First of all, we observed no significant blood-flow changes in the contralateral sensory cortices or contralateral sensory thalamus in either experiment, which suggests that the possible effects of the 12 muscle twitches on blood-flow response were negligible. Second, our results show that the depression of MEP amplitudes followed a similar time course in both experiments (Fig. 3). If our correlations resulted from changes in sensory feedback, we would have obtained more equivalent changes in blood flow from stimulating the dorsal premotor and primary motor cortices; this was not the case (Table 5).

The lack of blood-flow changes to single-pulse stimulation applied during the scans is not surprising in light of the low number of pulses (12 pulses/scan). On the other hand, we would expect blood-flow changes after the 15-min train of 1-Hz repetitive stimulation. We found a significant increase in local blood flow in scans acquired after repetitive stimulation over the primary motor cortex, but no such effects after repetitive stimulation over dorsal premotor cortex. Assuming tight coupling between excitatory synaptic activity and blood flow (Logothetis et al. 2001; Mathiesen et al. 1998; for review, see Paus 2002), we hypothesize that the local effects of low-frequency repetitive stimulation on inhibitory and excitatory neurotransmission canceled out while the distal effects remained. The latter might be related to the fact that the majority of cortico-cortical and cortico-subcortical projections are glutamatergic and, hence, their activation is more likely to influence blood flow in their target regions. As for a lack of a distal effect in the left primary motor cortex after dorsal premotor stimulation, this finding raises the possibility that the observed changes in MEP amplitudes are mediated by cortico-spinal projections originating in the dorsal premotor cortex rather than cortico-cortical connections between the dorsal premotor and primary motor cortices. We also hypothesize that the lateral-to-medial orientation of the short axis of the stimulating coil (virtual anode-cathode), as used in the dorsal premotor experiment, influenced preferentially inter-hemispheric rather than intra-hemispheric cortico-cortical projections. This could explain the general lack of distal effects in the left hemisphere compared with the right hemisphere after repetitive stimulation over the dorsal premotor cortex.

Before proceeding to the interpretation of the results, we should mention some important aspects related to our correlations. A positive correlation reflects an increase in blood-flow response with the amount of MEP reduction and a negative correlation reflects a decrease in blood-flow response with the amount of MEP reduction. Brain regions that show these correlations were modulated in parallel with MEP reduction but, in some cases, modulation could have resulted from nonspecific effects of TMS. Our correlation analyses showed that most negative correlations were located in the primary and associational visual areas. Our conjunction analysis further showed that the majority of these were present in both the dorsal premotor and primary motor experiments. Together, these results suggest that our negative correlations were largely the result of nonspecific effects of TMS, one possibility of which is the result of changes in arousal levels. Several attention studies observed blood-flow fluctuations in similar brain regions and attribute these changes to differences in arousal levels and/or to cross-modal suppression (reviewed in Paus 2000). The rest of this discussion therefore concentrates on our positive correlations, which were confined predominantly to motor areas and putative fronto-parietal circuits.

Dorsal premotor experiment

Several anatomically distinct areas constitute the premotor cortex, each with a potentially different specialization. In our study, repetitive stimulation likely affected two distinct dorsal premotor areas (reviewed in Piccard and Strick 2000), that is, those identified in the monkey as the caudal premotor area F2, which has substantial connections with the primary motor area (Barbas and Pandya 1987; Dum and Strick 1991), and the rostral dorsal premotor area F7, which has substantial connections with the prefrontal cortex (Barbas and Pandya 1987; Lu et al. 1994). Repetitive stimulation over the dorsal premotor cortex might also have affected the frontal eye field; our stimulation site was in close proximity to the probabilistic location of this area as established by Paus (1996) in a metaanalysis of oculomotor neuroimaging studies.

The dorsal premotor cortex plays a prominent role in coupling arbitrary sensory cues to motor acts (for review, see Freund 1996). Studies in the monkey reveal that lesions to the dorsal premotor cortex disrupt the animal's ability to use such cues to make or withhold particular movements (Halsband and Passingham 1982, 1985; Petrides 1985); the same is true for patients with damage to the dorsal premotor cortex (Halsband and Freund 1990). The parietal cortex receives somatosensory and visual inputs, and encompasses several subdivisions that have reciprocal connections with motor areas in the frontal cortex, each with a specific target with which it is most densely connected. These circuits provide an anatomical basis for the transformation of sensory information into motor actions (Matelli and Luppino 2000; Rizzolatti et al. 1998). Anatomical studies in the monkey reveal circuits that include the dorsal premotor area (F2/F7) as their frontal component; one of these is the MIP-F2 circuit (Matelli et al. 1998). A combination of somatosensory and visual information used for the visual guidance of arm movement trajectories is thought to reach F2 from MIP (Colby and Duhamel 1991; Galletti et al. 1996; Matelli and Luppino 2000).

In view of these data, we postulate that our findings may show the human homologue of the MIP-F2 circuit. The circuit follows from correlations observed in the right premotor cortex in the precentral sulcus and in the right medial intraparietal cortex along the posterior superior parietal lobule. Stimulation of the left dorsal premotor cortex might have modulated the right premotor cortex through commissural connections (Marconi et al. 2002; Pandya and Vignolo 1971). Our MIP coordinates (X = 36, Y = –64, Z = 54) are similar to those (X = –33, Y = –60, Z = 54) established in a previous functional MR imaging study of response switching, which required subjects to switch between two different visuomotor-related intentions (Rushworth et al. 2001). Other functional brain imaging studies show comparable metabolic changes in both the posterior parietal and premotor cortices as subjects selected motor acts based on visual stimuli (Dieber et al. 1997; Grafton et al. 1998; Paus et al. 1993).

Our results also suggest an additional parieto-frontal circuit that connects the right ventral premotor area (PMv) in the precentral operculum with the right putative anterior intraparietal area (AIP) in the lateral bank of the intraparietal sulcus along the anterior inferior parietal lobule. Stimulation of the left dorsal premotor cortex might have modulated the right ventral premotor cortex through commissural connections (Marconi et al. 2002). Marconi et al. (2002) recently demonstrated in the monkey that callosal connections exist between the dorsal premotor cortex in one hemisphere and the ventral premotor cortex in the opposite hemisphere. Connections also exist in the monkey between PMv (F5) and the more anterior part of the intraparietal cortex (Luppino et al. 1999). Both F5 and AIP neurons code for selective hand manipulations, grasping movements, and various visual characteristics of 3D objects (Murata et al. 1997; Rizzolatti et al. 1988). In view of these findings, Jeannerod et al. (1995) suggested that the F5-AIP circuit plays a role in transforming the properties of a 3D object into the appropriate hand movements required to grasp it. Previous PET data indicate that similar activations occur in the human PMv during the presentation of 3D objects (Grafton et al. 1997) with coordinates (X = –48, Y = –2, Z = 29) that are slightly more dorsal than our PMv coordinates (X = 52, Y = –6, Z = 12 and X = –43, Y = –6, Z = 14).

The prefrontal cortex plays a prominent role in executive functions (reviewed in Fuster 1993; Petrides 2000). To select relevant information for action, the prefrontal cortex has access, through its connections with other brain structures, to sensory and spatial aspects of the environment, mnemonic information acquired through experience, and motor control (reviewed in Barbas 2000). These motor output–related connections mainly arise from the premotor cortices (Barbas and Pandya 1987; Lu et al. 1994) and might explain our additional correlations in the prefrontal cortices. Anatomical data in the monkey show reciprocal connections of the prefrontal cortex and the premotor cortex in an orderly pattern along dorsal and ventral axes; interconnections between the two axes are sparse (Barbas 1988, 1992; Barbas and Pandya 1989). One would therefore predict that the blood-flow changes we observed in the ventrolateral prefrontal cortices arose from connections with the ventral premotor cortices.

Primary motor experiment

A conjunction analysis performed on our data revealed little overlap in the positive correlations obtained between the dorsal premotor and primary motor experiments. Similarly, ANOVA revealed that most of the brain regions with correlations showed significant differences between the two experiments. These findings suggest that we mapped two separate networks and lend support to the notion that the dorsal premotor and primary motor cortices differ in their functional properties. Unlike the dorsal premotor cortex, the primary motor cortex plays a role mainly in the execution of voluntary movements. Studies in the monkey reveal that lesions to the primary motor cortex disrupt the execution of skilled movements to a greater extent than lesions to nonprimary motor cortices (Passingham 1985; Passingham et al. 1983; Petrides 1985). Of all the cortical motor areas, the primary motor cortex contains the highest percentage (31%) of large corticospinal neurons (Dum and Strick 1991), which directly generate movement of the limbs (for review, see Evarts 1981).

The primary motor cortex connects predominantly with nonprimary motor and nonprimary somatosensory cortices; connections between the primary motor cortex and other cortical structures are sparse (Fig. 1). Visual and/or auditory information that influence movements must first be processed by associational and/or higher-order sensory cortices, and then be communicated to the nonprimary motor cortices (for review, see Ghez et al. 1991). The nonprimary motor cortices can in turn use this information to coordinate motor output at the level of both the primary motor cortex and the spinal cord (Dum and Strick 1991). We propose that our data from the primary motor experiment reflect this pattern of connections: the network mapped in the primary motor experiment encompasses correlations confined mainly to nonprimary motor cortices and subcortical motor structures.

Brain regions with significant correlations in the primary motor experiment include the right cingulate motor area, the left putamen, the right primary motor area, and the right ventral-lateral thalamic nucleus/internal global pallidus. Correlations in this experiment reflect both direct and indirect connections with the stimulation site (i.e., the left primary motor cortex). The cingulate motor area represents most likely the human homologue of CMAr, or the rostral cingulate zone, which is located anterior to the anterior commisure (Paus et al. 1993; Piccard and Strick 1996). The correlation in the left ipsilateral putamen suggests cortico-striatal projections from the primary motor area to the lateral putamen (Takada et al. 1998). The presence of a blood-flow response in the contralateral primary motor area suggests commissural connectivity from the stimulated hemisphere to the unstimulated hemisphere (Jenny 1979; Rouiller et al. 1994). The right ventrallateral thalamus and the right cingulate motor area might reflect indirect connections with the site of stimulation, the left primary motor cortex, mediated perhaps through the right primary motor cortex. Both the ventral-lateral thalamus and the internal globus pallidus are components of cortico-basal gangliathalamo-cortical loops related to the control of movement (Parent and Hazrati 1995).

In conclusion, the data presented here suggest that we mapped two separate motor-related networks and provide complementary insights into the function of the dorsal premotor and primary motor cortices.

DISCLOSURES

This work was supported by the Canadian Institute of Health Research (MT-14505) and by the Canadian Foundation for Innovation.

Acknowledgments

We thank Dr. K. E. Watkins for statistical advice.

Footnotes

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

References

View Abstract