Five healthy males (aged 27 to 36, right handed) with no history of neurological or psychiatric illness participated. All were screened for MRI and TMS compatibility and gave written informed consent in accord with local ethics. The study was approved by the joint ethics committee of the National Hospital for Neurology and Neurosurgery (University College London Hospitals National Heath Service Trust) and Institute of Neurology (University College London). Our stimulation protocol conformed to published TMS guidelines (Wassermann, 1998 (link)). The scalp position for placing the TMS coil over right parietal cortex was first determined outside the scanner, using the same approach as by Seyal and colleagues (Seyal et al., 1995 (link)). For this purpose, we identified the motor hotspot for the left thenar muscles (motor threshold: mean 63.6 % ± 4.2 S.E.M. of TMS stimulator output) and then moved the TMS coil backwards from the motor hotspot by 2-4 cm to the first point at which there was no longer any hand contractions induced when stimulating at the high TMS intensity of 110% of the resting motor threshold. We used an intensity of 50% of the motor threshold for our low intensity TMS during the main combined TMS-fMRI experiment (see below). This motor threshold was determined inside the scanner room with the same equipment as used during scanning. While some studies suggest that cross-modal interaction can have an impact on TMS thresholds (Ramos-Estebanez et al., 2007 (link)), our main focus for the concurrent TMS-fMRI experiment here was on any interaction between the effects of high versus low TMS, during the presence or absence of concurrent tactile input to the other hemisphere (see Introduction).As described below, we found robust effects of TMS during scanning on somatosensory responses in the other hemisphere, thus confirming that our TMS protocol was effective.
A pair of surface-adhesive electrodes was positioned on the right wrist of the subject for median-nerve stimulation. Constant current pulses (square wave, 200 μs duration) were applied to this site using a neurostimulator (DS7A, Digitimer, Hertfordshire, UK) located within a shielded box (to preclude MR artefacts) inside the scanner room. Stimulation of the right-median nerve in particular was confirmed by subjects’ verbal report of sensation in the first three fingers. We ensured that the median-nerve stimulation intensities used did not induce any twitching. For each subject, sensory threshold (mean 2.4 ± 0.16 mA) was determined by the method of limits (single pulses lasting 200 μs), and stimulation intensity for the experiment was then set to three times that sensory threshold (7.2 ± 0.5 mA), which is clearly detectable but does not induce any muscle effects. We used such suprathreshold somatosensory stimulation to ensure that increased activation of contralateral left SI by right median-nerve stimulation could be reliably detected in fMRI (Arthurs et al., 2000 (link);Blankenburg et al., 2003 (link)).
Our scanning experiment had a fully randomized 2 × 2 factorial design with the orthogonal factors of right parietal TMS (high versus low intensity) and right-wrist median-nerve stimulation (present versus absent). Each trial consisted of three successive ‘mini blocks’ of stimulation, each lasting 500 ms. On a random half of such trials, we applied right-wrist median-nerve stimulation in three trains of 5 pulses (each at 10 Hz). The other half of trials had no somatosensory stimulation. Orthogonally to this, TMS bursts (also 5 pulses at 10 Hz, and thus similar to the combined TMS-fMRI protocols in prior studies from our lab, see Bestmann et al., 2008b (link);Ruff et al., 2006 (link);Ruff et al., 2008 (link)) were applied on each trial with either high or low TMS intensity (random half of trials each). On those 50% of trials in the 2 × 2 factorial design that had both TMS application and median-nerve stimulation on the same trial, the trains of TMS or median-nerve stimulation (both 5 pulses at 10Hz) were temporally interleaved with a 180-degree difference in phase. Thus, each such trial started with a TMS pulse, followed 50 ms later by the first somatosensory stimulation, followed 50 ms later by the next TMS pulse and so on. After each trial, a rest period without any stimulation (neither median nerve nor TMS) was included, lasting four image volumes.
Functional data were acquired on a 1.5T whole-body scanner (Magnetom Sonata, Siemens Medical System, Erlangen, Germany), operating with the standard CP receive head and body transmit coil. We used a multi-slice gradient echo EPI sequence (39 slices, 64 × 96 matrix (readout × phase-encoding), in-plane resolution: 3 × 3 mm, 2 mm slice thickness, 50% spatial gap between adjacent slices, TE=50ms, TR=2880 ms, 2298 Hz/pixel bandwidth, echo spacing 500μs). In addition, oversampling (50%) was used in the phase-encoding direction to shift any possible ghost artifact induced by mere presence of the TMS coil outside of the volume of interest. The last seven slices (33-39, lasting 630 ms) were recorded without an MR excitation high frequency pulse. This enabled us to apply the TMS-pulses and the somatosensory stimulation always within this period, hence without any potential corruption of functional image volumes. In addition, this ensured a constant auditory input from the scanner, as gradients during these slices remained turned on. The acquisition time for one volume was 3.51 sec. Two sessions were acquired for each subject (each session comprised 320 volumes, including 5 dummy scans to allow T1 saturation). We chose to compare high-intensity (effective) TMS versus low-intensity (less effective or ineffective) TMS at a single right-parietal site, similar to the site used by Seyal et al. (1995) (link), rather than comparing different TMS sites, due to the technical problems that relocating the TMS probe within an fMRI session would inevitably cause. Moreover, the appropriateness of high versus low TMS comparisons at a given site during concurrent fMRI has recently been established by other studies from our group using similar concurrent TMS-fMRI protocols (Bestmann et al., 2008b (link);Ruff et al., 2006 (link)).
Each session included 44 randomly intermingled trials, 11 per condition in the 2 × 2 design. To preclude visual changes (e.g., from blinks), subject kept their eyes closed throughout scanning. In addition, to ensure that TMS could not lead to any changes in performance that might otherwise have complicated interpretation of the physiological fMRI data, subjects had no behavioural task during scanning (see also Ruff et al., 2006 (link)). Thus, as per the Introduction, our a-priori aim was to test for a physiological interaction of right parietal TMS with right-wrist somatosensory input, in terms of the BOLD response of left SI (and possibly the thalamus as well: see below). Nevertheless, we also ran a behavioral follow-up experiment outside the scanner (see below), which confirmed that our particular right-parietal TMS protocol, using bursts at 10Hz, could indeed produce the same behavioral effect originally documented by Seyal et al, (1995) (link), namely enhancement of somatosensory detection on the ipsilateral right hand.
TMS during scanning was applied using a Magstim Super Rapid stimulator and MR-compatible non-ferrous figure-of-eight coil with a small-diameter (30mm inner diameter, 70mm outer diameter, 15 turns each winding, wire size 5×1.5mm, 22.9μH inductance, 4.7kVA predicted maximal current at 100%) from the MAGSTIM Company, Dyfed, UK. The coil was positioned over the stimulation site tangentially to the scalp, at approximately 45° from the midline, inducing a biphasic current with an initial anteroposterior direction. We ensured that TMS did not induce any muscle twitches in the experiment (see above for TMS-site selection). The coil was held fixed by a non-ferromagnetic custom-built coil holder, and the participant’s head was fixed with vacuum cushions. To avoid any radio-frequency interference of TMS with image acquisition, the TMS-stimulator was placed in a shielded metal cabinet in the scanner room, and the TMS cable was passed through a custom filter box (the Magstim Company, Dyfed, UK) and further ferrite sleeves (Wuerth Elektronik, Waldenburg, Germany); see also (Bestmann et al., 2008b (link);Ruff et al., 2006 (link);Ruff et al., 2008 (link)). Furthermore, the TMS coil was connected to the stimulator in parallel to a high voltage relay (Magstim ES9486, The Magstim Company). During volume acquisition, this relay was closed, shorting any potential leaking-current. Thus, any current flow through the stimulation coil originating from the stimulator was eliminated while it waited to release a pulse. The relay was opened 50 ms prior to a TMS train, and closed 11 ms after the last TMS pulse of a train.
All stimuli were controlled using the MATLAB (The Mathworks, Natick, Massachusetts, USA) toolbox Cogent 2000 (http://www.vislab.ucl.ac.uk/Cogent/ ), running on a conventional PC. Image processing and analysis was performed with SPM2 (htt://www.fil.ion.ucl.ac.uk/spm ). Functional images were reconstructed offline, and the first five images of each run discarded to avoid T1 equilibration effects. In accord with the standard SPM approach, the remaining functional images were realigned to the first of the series, corrected for movement-induced image distortions, normalized to the MNI anatomical standard space and spatially smoothed with a 9 mm FWHM Gaussian kernel in accord with the standard SPM approach. In addition, the fMRI data were temporally band-pass filtered (lower/upper cutoff-frequency at 7 and 128 sec, respectively).
Statistical parametric maps were calculated by multiple regressions of the data onto a model of the hemodynamic response (Friston et al., 1995 (link)). This model contained regressors for the onsets of every ‘mini-block’ for each of the four conditions in the 2×2 design, convolved with the canonical hemodynamic response function in SPM2. An autocorrelation model was used in order to account for scan-to-scan dependencies in the error term. Statistical inference used a fixed effect model, in accord with the limited number of subjects (n=5) available for this demanding combined TMS-fMRI protocol. However, we also inspected individual data to ensure that the critical fMRI pattern was observed for all subjects (see below). For unrestricted whole-brain analyses we used a threshold of p<0.05, family-wise error (FWE) corrected for the entire image volume. For analyses of activity in brain areas for which we had clear a priori hypotheses (e.g., left somatosensory cortex and thalamus, see Introduction), critical effects were inspected in volumes-of-interest (VOIs) derived either by anatomical criteria (e.g., the thalamus was defined by means of a computerized cytoarchitectonic atlas, seehttp://www.loni.ucla.edu/ICBM/Downloads/Downloads_Atlases.shtml ), or functionally by inclusive masking with an orthogonal contrast used to define specific brain areas of interest (e.g., activation for presence minus absence of right-hand median-nerve stimulation was used to confirm functional localization of contralateral left somatosensory cortex). The results of these hypothesis-driven analyses are all reported at a threshold of p<0.05, FWE-corrected for the VOI (Worsley et al., 1996 (link)).
Here we used bursts of TMS at 10Hz during scanning in order to drive reliable BOLD responses by the TMS (Bestmann et al., 2008b (link);Ruff et al., 2006 (link);Ruff et al., 2008 (link)). Our particular TMS protocol thus differed from the original Seyal et al. (1995) (link) study, which had used single-pulse TMS. Accordingly, we also conducted a new psychophysical study outside the scanner that sought to replicate the classical behavioral findings of Seyal et al (1995) (link), but now using the identical 10Hz-burst TMS protocol as in our concurrent TMS-fMRI study. This follow-up behavioral experiment was conducted in 4 additional subjects, who were again screened for TMS compatibility and gave written informed consent in accord with local ethics. Briefly, we applied 5 pulses of 10 Hz rTMS at the outset of each trial, either at 110% or 50% of motor threshold, as during scanning. The TMS coil was again localized over right parietal lobe, using the identical procedure as for the main fMRI experiment. On a random half of these trials, peri-threshold right median nerve stimulation (of the same duration and timing relative to TMS burst as for the fMRI experiment) was applied during TMS. On the other half of trials, TMS was applied in the absence of median nerve stimulation. After each trial, subjects were asked to respond by button-press whether right-hand tactile stimulation was present or not. Subjects each completed 4 blocks of 60 trials. Tactile stimulation intensity was determined separately for each block, with the aim of keeping the intensities peri-threshold. Some blocks were removed from analysis because of greater than 90% accuracy (3 blocks), a bias towards responding ‘absent’ on more than 90% of trials (2 blocks), or a technical malfunction (2 blocks). Nine blocks remained, yielding a total of 540 trials. Sensitivity (d’) and response bias (criterion) were calculated for each retained block, and paired t-tests were performed to determine the effect of TMS intensity on d’ across blocks.
A pair of surface-adhesive electrodes was positioned on the right wrist of the subject for median-nerve stimulation. Constant current pulses (square wave, 200 μs duration) were applied to this site using a neurostimulator (DS7A, Digitimer, Hertfordshire, UK) located within a shielded box (to preclude MR artefacts) inside the scanner room. Stimulation of the right-median nerve in particular was confirmed by subjects’ verbal report of sensation in the first three fingers. We ensured that the median-nerve stimulation intensities used did not induce any twitching. For each subject, sensory threshold (mean 2.4 ± 0.16 mA) was determined by the method of limits (single pulses lasting 200 μs), and stimulation intensity for the experiment was then set to three times that sensory threshold (7.2 ± 0.5 mA), which is clearly detectable but does not induce any muscle effects. We used such suprathreshold somatosensory stimulation to ensure that increased activation of contralateral left SI by right median-nerve stimulation could be reliably detected in fMRI (Arthurs et al., 2000 (link);Blankenburg et al., 2003 (link)).
Our scanning experiment had a fully randomized 2 × 2 factorial design with the orthogonal factors of right parietal TMS (high versus low intensity) and right-wrist median-nerve stimulation (present versus absent). Each trial consisted of three successive ‘mini blocks’ of stimulation, each lasting 500 ms. On a random half of such trials, we applied right-wrist median-nerve stimulation in three trains of 5 pulses (each at 10 Hz). The other half of trials had no somatosensory stimulation. Orthogonally to this, TMS bursts (also 5 pulses at 10 Hz, and thus similar to the combined TMS-fMRI protocols in prior studies from our lab, see Bestmann et al., 2008b (link);Ruff et al., 2006 (link);Ruff et al., 2008 (link)) were applied on each trial with either high or low TMS intensity (random half of trials each). On those 50% of trials in the 2 × 2 factorial design that had both TMS application and median-nerve stimulation on the same trial, the trains of TMS or median-nerve stimulation (both 5 pulses at 10Hz) were temporally interleaved with a 180-degree difference in phase. Thus, each such trial started with a TMS pulse, followed 50 ms later by the first somatosensory stimulation, followed 50 ms later by the next TMS pulse and so on. After each trial, a rest period without any stimulation (neither median nerve nor TMS) was included, lasting four image volumes.
Functional data were acquired on a 1.5T whole-body scanner (Magnetom Sonata, Siemens Medical System, Erlangen, Germany), operating with the standard CP receive head and body transmit coil. We used a multi-slice gradient echo EPI sequence (39 slices, 64 × 96 matrix (readout × phase-encoding), in-plane resolution: 3 × 3 mm, 2 mm slice thickness, 50% spatial gap between adjacent slices, TE=50ms, TR=2880 ms, 2298 Hz/pixel bandwidth, echo spacing 500μs). In addition, oversampling (50%) was used in the phase-encoding direction to shift any possible ghost artifact induced by mere presence of the TMS coil outside of the volume of interest. The last seven slices (33-39, lasting 630 ms) were recorded without an MR excitation high frequency pulse. This enabled us to apply the TMS-pulses and the somatosensory stimulation always within this period, hence without any potential corruption of functional image volumes. In addition, this ensured a constant auditory input from the scanner, as gradients during these slices remained turned on. The acquisition time for one volume was 3.51 sec. Two sessions were acquired for each subject (each session comprised 320 volumes, including 5 dummy scans to allow T1 saturation). We chose to compare high-intensity (effective) TMS versus low-intensity (less effective or ineffective) TMS at a single right-parietal site, similar to the site used by Seyal et al. (1995) (link), rather than comparing different TMS sites, due to the technical problems that relocating the TMS probe within an fMRI session would inevitably cause. Moreover, the appropriateness of high versus low TMS comparisons at a given site during concurrent fMRI has recently been established by other studies from our group using similar concurrent TMS-fMRI protocols (Bestmann et al., 2008b (link);Ruff et al., 2006 (link)).
Each session included 44 randomly intermingled trials, 11 per condition in the 2 × 2 design. To preclude visual changes (e.g., from blinks), subject kept their eyes closed throughout scanning. In addition, to ensure that TMS could not lead to any changes in performance that might otherwise have complicated interpretation of the physiological fMRI data, subjects had no behavioural task during scanning (see also Ruff et al., 2006 (link)). Thus, as per the Introduction, our a-priori aim was to test for a physiological interaction of right parietal TMS with right-wrist somatosensory input, in terms of the BOLD response of left SI (and possibly the thalamus as well: see below). Nevertheless, we also ran a behavioral follow-up experiment outside the scanner (see below), which confirmed that our particular right-parietal TMS protocol, using bursts at 10Hz, could indeed produce the same behavioral effect originally documented by Seyal et al, (1995) (link), namely enhancement of somatosensory detection on the ipsilateral right hand.
TMS during scanning was applied using a Magstim Super Rapid stimulator and MR-compatible non-ferrous figure-of-eight coil with a small-diameter (30mm inner diameter, 70mm outer diameter, 15 turns each winding, wire size 5×1.5mm, 22.9μH inductance, 4.7kVA predicted maximal current at 100%) from the MAGSTIM Company, Dyfed, UK. The coil was positioned over the stimulation site tangentially to the scalp, at approximately 45° from the midline, inducing a biphasic current with an initial anteroposterior direction. We ensured that TMS did not induce any muscle twitches in the experiment (see above for TMS-site selection). The coil was held fixed by a non-ferromagnetic custom-built coil holder, and the participant’s head was fixed with vacuum cushions. To avoid any radio-frequency interference of TMS with image acquisition, the TMS-stimulator was placed in a shielded metal cabinet in the scanner room, and the TMS cable was passed through a custom filter box (the Magstim Company, Dyfed, UK) and further ferrite sleeves (Wuerth Elektronik, Waldenburg, Germany); see also (Bestmann et al., 2008b (link);Ruff et al., 2006 (link);Ruff et al., 2008 (link)). Furthermore, the TMS coil was connected to the stimulator in parallel to a high voltage relay (Magstim ES9486, The Magstim Company). During volume acquisition, this relay was closed, shorting any potential leaking-current. Thus, any current flow through the stimulation coil originating from the stimulator was eliminated while it waited to release a pulse. The relay was opened 50 ms prior to a TMS train, and closed 11 ms after the last TMS pulse of a train.
All stimuli were controlled using the MATLAB (The Mathworks, Natick, Massachusetts, USA) toolbox Cogent 2000 (
Statistical parametric maps were calculated by multiple regressions of the data onto a model of the hemodynamic response (Friston et al., 1995 (link)). This model contained regressors for the onsets of every ‘mini-block’ for each of the four conditions in the 2×2 design, convolved with the canonical hemodynamic response function in SPM2. An autocorrelation model was used in order to account for scan-to-scan dependencies in the error term. Statistical inference used a fixed effect model, in accord with the limited number of subjects (n=5) available for this demanding combined TMS-fMRI protocol. However, we also inspected individual data to ensure that the critical fMRI pattern was observed for all subjects (see below). For unrestricted whole-brain analyses we used a threshold of p<0.05, family-wise error (FWE) corrected for the entire image volume. For analyses of activity in brain areas for which we had clear a priori hypotheses (e.g., left somatosensory cortex and thalamus, see Introduction), critical effects were inspected in volumes-of-interest (VOIs) derived either by anatomical criteria (e.g., the thalamus was defined by means of a computerized cytoarchitectonic atlas, see
Here we used bursts of TMS at 10Hz during scanning in order to drive reliable BOLD responses by the TMS (Bestmann et al., 2008b (link);Ruff et al., 2006 (link);Ruff et al., 2008 (link)). Our particular TMS protocol thus differed from the original Seyal et al. (1995) (link) study, which had used single-pulse TMS. Accordingly, we also conducted a new psychophysical study outside the scanner that sought to replicate the classical behavioral findings of Seyal et al (1995) (link), but now using the identical 10Hz-burst TMS protocol as in our concurrent TMS-fMRI study. This follow-up behavioral experiment was conducted in 4 additional subjects, who were again screened for TMS compatibility and gave written informed consent in accord with local ethics. Briefly, we applied 5 pulses of 10 Hz rTMS at the outset of each trial, either at 110% or 50% of motor threshold, as during scanning. The TMS coil was again localized over right parietal lobe, using the identical procedure as for the main fMRI experiment. On a random half of these trials, peri-threshold right median nerve stimulation (of the same duration and timing relative to TMS burst as for the fMRI experiment) was applied during TMS. On the other half of trials, TMS was applied in the absence of median nerve stimulation. After each trial, subjects were asked to respond by button-press whether right-hand tactile stimulation was present or not. Subjects each completed 4 blocks of 60 trials. Tactile stimulation intensity was determined separately for each block, with the aim of keeping the intensities peri-threshold. Some blocks were removed from analysis because of greater than 90% accuracy (3 blocks), a bias towards responding ‘absent’ on more than 90% of trials (2 blocks), or a technical malfunction (2 blocks). Nine blocks remained, yielding a total of 540 trials. Sensitivity (d’) and response bias (criterion) were calculated for each retained block, and paired t-tests were performed to determine the effect of TMS intensity on d’ across blocks.