It has proven difficult to separate functional areas in the prefrontal cortex (PFC), an area implicated in attention, memory, and distraction handling. Here, we assessed in healthy human subjects whether PFC subareas have different roles in top-down regulation of sensory functions by determining how the neural links between the PFC and the primary somatosensory cortex (S1) modulate tactile perceptions. Anatomical connections between the S1 representation area of the cutaneous test site and the PFC were determined using probabilistic tractography. Single-pulse navigated transcranial magnetic stimulation of the middle frontal gyrus–S1 link, but not that of the superior frontal gyrus–S1 link, impaired the ability to discriminate between single and twin tactile pulses. The impairment occurred within a restricted time window and skin area. The spatially and temporally organized top-down control of tactile discrimination through a segregated PFC–S1 pathway suggests functional specialization of PFC subareas in fine-tuned regulation of information processing.
Magnetic resonance imaging (MRI)-guided navigated transcranial magnetic stimulation (nTMS) with single monopolar pulses provides an anatomically and temporally accurate method for cortical stimulation and perturbation of cortical function (Hannula et al. 2005). In the study of the somatosensory system, nTMS has been used to verify the well-established concept that somatotopic representation of afferent inputs in the primary somatosensory cortex (S1) has an important role in mediating signals evoking tactile sensations (Werner and Whitsel 1973; Romo and Salinas 2001). This is indicated by impairment of tactile perceptions (Hannula et al. 2005) and tactile temporal discrimination ability (Hannula et al. 2008) by single-pulse nTMS of the S1 representation area of the cutaneous test site (S1hotspot[hs]). While the contribution of the prefrontal cortex (PFC) to the perception of tactile signals is not well known, there is earlier evidence suggesting that, in addition to its many other roles (Goldman-Rakic 1987; Fuster 2008), the PFC is associated with the processing or gating of tactile information (e.g., Yamaguchi and Knight 1990; Carlson et al. 1997; Chao and Knight 1998; Romo et al. 1999; Romo and Salinas 2003; Stoeckel et al. 2003; Numminen et al. 2004; Pastor et al. 2004; Postle 2005; Pleger et al. 2006; Preuschhof et al. 2006; Kostopoulos et al. 2007). Earlier tract-tracing studies in nonhuman primates indicate that the PFC has cortico-cortical connections with the S1 (Preuss and Goldman-Rakic 1989), which provide potential pathways for top-down modulation of the S1 by the PFC. A combination of nTMS with diffusion-weighted MRI and tractography allows selective stimulation of segregated cortico-cortical or corticofugal fiber tracts and thereby, a noninvasive study of their functional roles in the living human brain (Hannula et al. 2010). The hypothesis that the PFC–S1 link, indeed, is involved in the modulation of somatosensory signals is supported by a recent experimental finding in healthy subjects, showing that nTMS of the PFC attenuated the distractive effect of task-irrelevant signals during tactile working memory (WM) maintenance and thereby, facilitated WM performance (Hannula et al. 2010). Importantly, memory facilitation occurred only when nTMS was applied during tactile WM maintenance to a PFC site which according to tractography was connected to the S1hs, implying that the improved memory performance was due to top-down suppression of distractive tactile signals.
Here, we proposed a hypothesis that a segregated PFC–S1 link has a specific role in top-down control of tactile perceptions. We tested this hypothesis by studying how nTMS of the PFC–S1 link influences tactile discrimination ability in a task in which the subject attempts to discriminate between single and twin tactile pulses applied to the skin of the hand. Moreover, we determined whether the PFC–S1 link-mediated top-down effect has a temporal and spatial organization, and whether the role of the PFC–S1 link depends on its origin within the PFC.
Materials and Methods
Eight healthy volunteers participated in the experiments (3 females and 5 males, age range from 23 to 31 years). All subjects were right handed. Five of the subjects participated in all 3 main experiments, while 3 subjects participated only in 2 of them (Supplementary Table 1). All subjects signed an informed consent before participating in the experiment. The experimental protocol took into account the code of ethics as defined in the World Medical Association's Declaration of Helsinki, and it was approved by the Ethical Committee of the Helsinki University Central Hospital.
Diffusion-Weighted Magnetic Resonance Imaging Data Acquisition
MRI scans were obtained with a 3.0-T scanner (Signa VH/I Excite II; GE Healthcare, Chalfont St. Giles, UK) equipped with an 8-channel high-resolution brain array head coil (GE Signa Excite, GE Healthcare) at the Advanced Magnetic Imaging Centre (AMI Centre, Aalto University, Espoo, Finland).
The diffusion imaging scheme consisted of acquiring a set of 60 diffusion-weighted images with noncollinear diffusion gradients (1000 s/mm2, Δ = 30 ms, δ = 24 ms) and 4 nondiffusion-weighted images employing a single-shot diffusion-weighted echo-planar imaging sequence. The orientation of diffusion gradients was chosen to minimize the directional bias in the measurements (Jones et al. 1999). The configuration of diffusion gradients was obtained from a 60-electron repulsion scheme computed on the surface of a unit sphere. A set of 54 slices was acquired using a matrix of 128 × 128 and voxel dimensions of 1.9 × 1.9 × 3.0 mm, resulting in a rectangular field of view (FOV) of 240 mm. The echo time (TE) was 79 ms, and the repetition time (TR) was 10 000 ms. A manually adjusted high-order shim procedure was performed prior to acquisition of diffusion-weighted images. The imaging was performed twice, resulting in a data set consisting of 128 volumes.
A T1-weighted 3-dimensional (3D) anatomical volume was acquired with a rectangular FOV of 240 mm, matrix of 256 × 256, voxel dimensions of 0.9 × 0.9 × 1.0 mm, TE = 1.9 ms, TR = 9.1 ms, IT = 300 ms, flip angle = 15°, and NEX = 2. A total of 162 axial slices were collected.
Postprocessing of Diffusion-Weighted Images
The diffusion-weighted data set was corrected for subject motion and eddy currents with the FMRIB's diffusion toolbox using the nondiffusion-weighted volume as a reference. The data were prepared for probabilistic tractography by performing a Bayesian estimation of diffusion parameters with the BEDPOSTX tool in the FSL software package (www.fmrib.ox.ac.uk/analysis/; Behrens, Johansen-Berg, et al. 2003; Behrens, Woolrich, et al. 2003). The anatomical T1-weighted volumes where skull stripped (BET, FMRIB, Oxford, UK; Smith 2002) and coregistered (FLIRT tool in the FSL package; Jenkinson et al. 2002).
Navigated Transcranial Magnetic Stimulation
The eXimia NBS system locates the TMS coil with an optical tracking system that can recognize the TMS tracking tools with a precision of <1 mm (Hannula et al. 2005). The eXimia NBS system takes into account individual head shape and size as the stimulation coil is modeled and the calculation of the intracranial electric field is based on the spherical model (Sarvas 1987; Ruohonen and Ilmoniemi 2002; Tarkiainen et al. 2003) matched to the individual MRIs.
Mapping of the M1 cortex determined the optimal coil position to produce a motor response of the relaxed abductor pollicis brevis muscle. Coil orientation was such that the induced electric field was aimed at the motor cortex, anterior to the central sulcus. First, the area that produced the highest motor response was located, and the coil was rotated to determine the optimal coil orientation. This coil location was selected as target for determining the resting motor threshold. Motor evoked potentials were measured with continuous on-line surface electromyography. The lowest TMS intensity at which ≥5 of 10 pulses to the optimal motor area resulted in a motor evoked potential of 50 μV (peak-to-peak) or greater was considered the motor threshold. In tactile discrimination experiments, TMS was applied at 120% of the motor threshold, which corresponded to 65 ± 12.4% of the maximal output of the stimulator.
Determination of the Thenar Representation on S1
When determining functionally the thenar representation site on S1 in a blocking experiment (Hannula et al. 2005), electrical test stimuli were delivered at threshold intensity to the thenar skin of the dominant right hand using a constant current stimulator. TMS was applied at 120% of the motor threshold, and with a time delay of 20 ms from the cutaneous stimulus. TMS pulses and tactile stimuli were programmed with the Presentation software (Neurobehavioral Systems, Albany, CA, USA). Subjects were instructed to attend to the thenar test region and after each TMS pulse to answer “yes” if they felt a tactile stimulus, “no” if not, and “maybe” if they were uncertain. S1 was probed with TMS until a location, S1hs, was found in which TMS succeeded to block at least 75% of the cutaneous test pulses. Our previous study demonstrated that when mapping the S1 representation of a stimulus site in the hand thenar using the same protocol and methodology as in the present study, the functional resolution of monopolar nTMS is 8–13 mm or better (Hannula et al. 2005).
On average, 5 (range 2–8) sets of 8 stimulus presentations were needed to find the S1hs (as indicated by successful blocking of cutaneous sensations by nTMS of S1). In the sham condition, the stimulus coil was kept approximately 5 cm above the head. In the successful blocking sessions, subjects answered no 6.9 ± 0.7, yes 0.7 ± 0.76, and maybe 0.3 ± 0.5 times out of 8 (mean ± SD). In the sham sessions, subjects answered no 2.6 ± 1.1, yes 3.9 ± 1.1, and maybe 1.6 ± 1.3 times out of 8. Subjects did not receive any feedback of their performance in the blocking sessions. When the blocking was successful, the cutaneous stimulus or TMS evoked no tactile sensations in the cutaneous test site.
Connections from the S1hs to the middle frontal gyrus (MFG) and superior frontal gyrus (SFG) were probed with probabilistic tractography using the FSL 4.0 software (FMRIB). After functional determination of the S1hs, coordinates of the maximal estimated electric field (E-field) produced by the stimulating coil at the S1hs were recorded. The coordinates were then transferred to the anatomical T1 image and registered to the diffusion space. A region corresponding to the S1hs, with a diameter of approximately 10 mm, was selected around the maximum E-field coordinate as a seed mask for tractography. Tracts were generated starting from the seed mask (step length = 0.5; number of steps = 2000; number of particles = 5000; curvature threshold = 0.2), which resulted in each voxel attaining a connectivity value, corresponding to the number of streamlines passing through the voxel and the seed region. The threshold was set at a connectivity value of >100 (Ciccarelli et al. 2006). The same threshold is previously described and chosen on the basis that lower values give unspecific connections to voxels that are widespread across the brain (Ciccarelli et al. 2006). The connectivity value had to be reduced in one subject to 48 to find a solid tract to the SFG, and in one subject to 30 to find a solid tract to the MFG. In these 2 subjects, unspecific connections were found only with connectivity values less than the reported ones.
In this report, MFGhs and SFGhs refer to the most anterior PFC sites that had a tractography-informed connection with the S1 representation site of the thenar skin site (S1hs) in which we attempted to modulate tactile discrimination by nTMS of the PFC (Supplementary Fig. 1 and Table 2).
Tactile Discrimination Task
Two Ag–AgCl skin electrodes (Ambu, Ballerup, Denmark) were fixed on the thenar or hypothenar area of the dominant hand, and electrical stimuli were delivered via a Grass PSIU6 constant current stimulator (Grass Instruments, Quincy, MA, USA) with a pulse duration of 0.2 ms. Tactile threshold for single pulses applied at intervals varying between 1 and 2 s was determined by the method of limits, and it was defined as the stimulus intensity that was felt by the subject in at least 90% of the trials.
It should be noted that while the delayed matching to sample paradigm that inherently involves a memory or decision component has been commonly used in assessing tactile discrimination ability, in the present study we used a strictly perceptual task with no involvement of WM. In the tactile discrimination task of the present study, cutaneous test stimuli were applied to the thenar or hypothenar area of the dominant hand at an intensity that was twice as high as the perception threshold for a single test pulse. The single nTMS pulse was applied at the intensity of 120% of the motor threshold to the MFGhs and SFGhs, or at the intensity of 120% of the motor threshold approximately 5 cm above the head (sham TMS). Delivery of the cortical stimulus was accompanied by that of the cutaneous single or twin pulse. Each cortical stimulus condition (MFGhs, SFGhs, and sham) was studied in a separate session. The interstimulus interval (ISI) of the twin pulse was chosen individually based on a preliminary study in which we determined the ISI at which the subjects could discriminate a single pulse from a twin pulse in about 90% of stimulus presentations. In general, this ISI was 90 ms. The delay from the delivery of the cutaneous test pulse to that of the cortical stimulus was 0 ms, except for the first main experiment of the present study in which 3 different delays were used: 0, 20, or 50 ms. With twin cutaneous test pulses, the delay from the cutaneous stimulus to the nTMS delivery was calculated from the second pulse of the cutaneous twin pulses. In each experimental condition (Supplementary Table 1), both the cutaneous single and twin pulses were presented 8 times. To allow counterbalancing not only between but also within subjects, each condition was split into 2 separate sessions. Each psychophysical session lasted 2–3 min. The number of daily sessions was 6–10, and the subjects had a possibility to have a rest and refreshments between sessions. Tactile threshold was also determined between the sessions to exclude the possibility that a change in the threshold contributes to the results. Moreover, counterbalancing of all stimulus conditions was considered to minimize a potential bias caused by changes occurring as a function of elapsed time or repetitive testing.
The task of the subjects was to assess after each test stimulus delivery whether a single or a twin pulse was presented. Therefore, they were asked to press the left (single stimulus) or right (twin stimulus) button of a computer mouse with their left middle or index finger, respectively, after each stimulus delivery. Response time from the test stimulus presentation to the button press was sampled in all conditions. After pressing the button of a computer mouse, the perception induced by each cutaneous test stimulus (single or double pulse) was verified by a verbal report of the subject. In all experimental conditions and in all subjects, the perceptual responses sampled with the computer mouse and by a verbal report proved to be identical. For assessment of the subject's capacity to discriminate between single and twin pulses, receiver operating characteristic (ROC) curve analysis was performed (Swets 1973) using the MedCalc (Mariakerke, Belgium) software. When entering the data, reporting a single pulse when the stimulus was a single pulse was considered a correct rejection, whereas reporting a twin pulse when the stimulus was a single pulse was considered a false alarm. Reporting a twin pulse when the stimulus was a twin pulse was considered a hit, whereas reporting a single pulse when the stimulus was a twin pulse was considered a miss. In each condition, the ROC curve analysis was made by comparing responses with single and twin pulses. The area under the ROC curve (AUC) was used as an index of the subject's discriminative capacity. The AUC value varies from 0.5 to 1.0: the value of 0.5 indicates that the subject cannot discriminate a single stimulus from a twin stimulus, whereas 1.0 indicates that the subject can perfectly discriminate the twin stimulus from a single stimulus.
In addition to reporting whether the subject felt 1 or 2 test pulses in the skin, the subject was asked to give verbally the confidence rating for each of his response using the scale of 1–3 (1 = low level of confidence/a guess; 2 = medium level of confidence; and 3 = high level of confidence).
Course of the Study
For each subject, the experiment started with MRI scanning. Next, the subjects participated in a session, in which the motor threshold and S1hs were determined. Then, tractography was performed to determine stimulation targets in the PFC. The 3 main experiments (Supplementary Table 1) were performed on separate days: 1) Temporal properties of top-down modulation, 2) spatial properties of top-down modulation, 3) PFC subregion in top-down modulation. In each of the main experiments, each cortical stimulus condition was tested in a separate session, and the order of testing different stimulus conditions was varied between and within the subjects. Moreover, the order of presenting single and twin cutaneous pulses within sessions was semi-random. The duration of testing during 1 day varied from 1 to 2h. Testing sessions had a maximum duration of 10 min, and with intervals for rest and refreshments between the sessions. In all subjects, nTMS of the MFGhs or SFGhs was applied in the left hemisphere (i.e., contralateral to their dominant hand).
Statistical analysis of the data was performed using 1- or 2-way repeated-measures analysis of variance followed by Tukey's test. A P-value of <0.05 was considered to represent a significant difference.
The experiments were performed in 8 healthy subjects, who had participated in a diffusion-weighted MRI scanning and in whom the location of the S1hs (the S1 representation area of the hand thenar) was established by determining the cortical site on the postcentral gyrus, where the sensation of the cutaneous test stimulus could be blocked by nTMS (Supplementary Table 1). PFC sites with neural connections with the S1hs were located using probabilistic tractography (Fig. 1, Supplementary Fig. 1 and Table 2). When assessing top-down control of tactile temporal discrimination by the PFC, the subjects discriminated between electric single and twin pulses applied to the hand thenar or hypothenar skin while a single monophasic nTMS pulse was delivered into the PFC sites linked with S1hs. A sham TMS stimulus was delivered in control trials.
Temporal Properties of the PFC-Induced Modulation of Tactile Perceptions
In the PFC, the MFG through a neural link to S1hs contributed to the control of tactile discrimination ability within a restricted time window as indicated by the following findings. When a single nTMS pulse was applied to the MFGhs (the MFG site connected with the S1hs), tactile discrimination ability in the hand thenar was suppressed (main effect of nTMS: F1,6 = 9.94, P = 0.02; Fig. 2a). The impairment of discrimination ability was most robust when the nTMS to the MFGhs was applied at the same time as the cutaneous test stimulation (0- ms delay) and reduced when the delay from the cutaneous stimulation to the nTMS was prolonged from 0 to 50 ms (main effect of delay: F2,12 = 5.89, P = 0.017). Two additional and independent parameters assessed in the same sessions, the subject's response time and response confidence, corroborated the findings of the top-down control of discrimination ability. The nTMS of the MFGhs produced a close to significant prolongation of the response time (F1,6= 4.42, P = 0.080; Supplementary Fig. 2a) and reduced the subject's confidence rating of the response (F1,6 = 13.15, P = 0.011; Supplementary Fig. 3a).
Spatial Properties of the PFC-Induced Modulation of Tactile Perceptions
Top-down control of tactile discrimination through the MFG–S1 link was spatially organized. This was indicated by the finding that the suppression of tactile discrimination by nTMS of the MFGhs (for thenar) was significantly stronger in the hand thenar than hypothenar (main effect of cutaneous test site: F1,7 = 9.14, P = 0.019; Fig. 2b). Moreover, the suppression of tactile discrimination (main effect of nTMS: F1,7 = 13.1, P = 0.0085) varied with the cortical stimulus condition (interaction of cutaneous test site with the cortical stimulation condition: F1,7 = 5.83, P = 0.047). In line with this, post hoc testing indicated that nTMS of the MFGhs (for thenar) produced a significant impairment of tactile discrimination ability only in the thenar skin. The effect by nTMS of the MFGhs was short of significance on the response time or confidence rating in the experiment in which spatial properties of top-down modulation were assessed (Supplementary Figs 2b and 3b).
While in the main experiments reported here, the cutaneous test stimulus was applied to the dominant hand contralateral to the cortical stimulation site, in an additional experiment with 2 subjects the cutaneous test stimulation was applied to the hand thenar ipsilateral to the cortical stimulation site. In these 2 subjects, nTMS of the MFGhs failed to reduce tactile temporal discrimination ability in the ipsilateral hand thenar (the AUC curve was >0.9 in both subjects with and without nTMS of the ipsilateral MFGhs).
Influence of the PFC Subarea on the PFC-Induced Perceptual Effect
In the current experiments, the top-down control of tactile temporal discrimination was a specific property of the MFGhs–S1 link rather than a general property of the neural tracts connecting the PFC with S1. This is indicated by the finding that while within the PFC the SFG and the MFG were connected with the S1hs, the suppression of tactile temporal discrimination ability varied significantly with the cortical stimulus condition (sham, MFGhs, and SFGhs: F2,10 = 6.52, P = 0.015; Fig. 2c). Post hoc testing indicated that nTMS of the MFGhs produced a significantly stronger suppression of tactile temporal discrimination than that of the SFGhs. Response time (F2,10 = 4.46, P = 0.041; Supplementary Fig. 2c) varied with the cortical stimulus condition, whereas the cortical stimulus condition failed to have a significant effect on confidence ratings (F2,10 = 0.38; Supplementary Fig. 3c). Post hoc testing indicated that nTMS of the MFGhs produced a significantly longer response time than that of the SFGhs (Supplementary Fig. 2c).
Our results demonstrate reduction of tactile discrimination by nTMS of the PFC. The association of reduced tactile discrimination with the nTMS-induced activation of a top-down pathway from PFC to S1 is in line with the hypothesis that one of the functions of the PFC is sensory gating that facilitates, through suppression of distractive signals, attention to behaviorally relevant signals and memory maintenance (Chao and Knight 1998; Postle 2005; Hannula et al. 2010). Impairment of tactile discrimination by nTMS of the MFGhs, but not by nTMS of the SFGhs, indicates a selective involvement of the MFG–S1 link in the gating of tactile signals by the PFC. Since the present study focused on assessing discrimination of temporal properties of tactile test stimuli (discrimination of single from twin pulses) and since spatial stimulus properties have proved effective in activating SFG in earlier brain imaging studies, at least in WM tasks (Carlson et al. 1998; Courtney et al. 1998), it remains to be studied whether nTMS of the SFGhs might influence tactile spatial perceptions. Interestingly, the MFG–S1 link-mediated control of tactile discrimination in the present study was temporally and spatially organized, indicating that, through this link, the PFC may induce subtle modifications of tactile perceptions which may be of importance when performing tactile discrimination tasks.
The dependence of the top-down control effect both from the stimulated PFC–S1 pathway and the cutaneous test site also suggests that the segregated top-down links from the PFC to the sensory cortical areas as well as neurons within the PFC subregions (Artchakov et al. 2009) have quite specific rather than more general modulatory roles in top-down control of cognitive functions. In line with this, our recent study indicated that the MFG–S1 link reduces the influence of tactile rather than visual distraction during tactile WM maintenance (Savolainen et al. 2011). Moreover, nTMS of the MFGhs, but not that of an adjacent MFG area, suppressed somatosensory evoked potentials in the S1 cortex (Hannula et al. 2010). This finding also gives physiological evidence for functional specialization within the PFC.
While the present findings are most likely explained by/related to sensory gating mechanisms (Chao and Knight 1998; Postle 2005), the PFC is also involved in multiple other functions, one which is decision-making. For example, the left dorsolateral PFC is considered to be among structures associated with perceptual decision-making in various sensory tasks (Heekeren et al. 2008). Of particular interest for the present study is an earlier human brain imaging result, showing that activity changes in the left dorsolateral PFC encode stimulus representations underlying decisions in a tactile temporal discrimination task (Pleger et al. 2006). In the present study, nTMS was also applied to the left PFC. Therefore, it would be tempting to speculate that the PFC–S1 link that induced modulation of tactile temporal discrimination is involved in the PFC circuitry associated with perceptual decision-making. However, the earlier human brain imaging study demonstrating the association of the left dorsolateral PFC activation with perceptual decision-making used a delayed matching to sample tactile temporal discrimination task (Pleger et al. 2006). This involved a significant contribution of WM when the subjects compared the perceptions induced by 2 subsequent stimulus presentations (Pleger et al. 2006). In contrast, the present study with presentations of cutaneous single or twin pulses that subjects classified without a comparison stimulus was strictly a perceptual task, without the contribution of WM. Parallel assessment of perceptions using 2 different types of responses (pressing a mouse button with fingers, and giving a verbal report) and the analysis of the results using methods based on the signal detection theory (Swets 1973) further emphasize that the present results reflect PFC-induced effects on perceptions rather than motor or other cognitive–motivational behaviors. For these reasons, it is likely that the earlier results on perceptual decision-making in a tactile temporal discrimination task (Pleger et al. 2006) deal with mechanisms that are at least partly different from those studied in the present strictly perceptual task. Moreover, based on some earlier findings (Heekeren et al. 2008), it might be expected that the neural circuitry underlying decision-making does not have as spatially restricted effects on tactile temporal discrimination as nTMS of the MFGhs had in the present study. Taken together, these observations are in line with the hypothesis that the MFG–S1 link plays a role in fine-tuning of tactile perceptions, probably as part of the sensory gating mechanisms (Chao and Knight 2005; Postle 2005), although its contributions to other functions still need to be studied.
The demonstration of a spatially and temporally organized top-down control of tactile temporal discrimination by the MFG–S1 link in the present study indicates that the tractography-guided single-pulse nTMS allows anatomically and temporally accurate perturbation of cognitive functions and thereby, assessment of causality between a cognitive function and a segregated fiber tract in the living human brain.
The MFG–S1 link, but not the SFG–S1 link, controls tactile temporal discrimination in a spatially and temporally organized fashion. This finding suggests a functional specialization of PFC subareas in fine-tuned regulation of information processing.
This work was supported by the aivoAALTO project of the Aalto University, Espoo, Finland, the Sigrid Jusélius Foundation, Helsinki, Finland, the Päivikki and Sakari Sohlberg Foundation, Helsinki, Finland, and the Academy of Finland, Helsinki, Finland.
Conflict of Interest: One of the authors (P.S.) is an employee of the company (Nexstim Ltd.) that manufactured the TMS device used in this study. Other authors declare no conflicts of interest.