Multiple time courses of somatosensory responses in human cortex

Abstract

Here we show how anatomical and functional data recorded from patients undergoing stereo-EEG can be used to decompose the cortical processing following nerve stimulation in different stages characterized by specific topography and time course. Tibial, median and trigeminal nerves were stimulated in 96 patients, and the increase in gamma power was evaluated over 11878 cortical sites. All three nerve datasets exhibited similar clusters of time courses: phasic, delayed/prolonged and tonic, which differed in topography, temporal organization and degree of spatial overlap. Strong phasic responses of the three nerves followed the classical somatotopic organization of SI, with no overlap in either time or space. Delayed responses presented overlaps between pairs of body parts in both time and space, and were confined to the dorsal motor cortices. Finally, tonic responses occurred in the perisylvian region including posterior insular cortex and were evoked by the stimulation of all three nerves, lacking any spatial and temporal specificity. These data indicate that the somatosensory processing following nerve stimulation is a multi-stage hierarchical process common to all three nerves, with the different stages likely subserving different functions. While phasic responses represent the neural basis of tactile perception, multi-nerve tonic responses may represent the neural signature of processes sustaining the capacity to become aware of tactile stimuli.

Highlights

• •StereoEEG shows different time courses in responses to peripheral nerve stimulations.
• •Strong phasic responses likely reflect thalamic input into primary somatosensory cortex.
• •Delayed responses subserve sensorimotor synergies in motor and premotor cortex.
• •Tonic multi-nerve responses in perisylvian region may be markers of tactile awareness.

Introduction

The organization of human somatosensory cortex has been intensively studied in these last years with functional MRI and with electro-/magneto-encephalography. A frequently used procedure in these investigations was the description of the neural activity evoked by the electrical stimulation of peripheral nerves. While neuroimaging studies have localized the most important nodes contributing to somatosensory processing, namely contralateral primary somatosensory areas (SI), bilateral secondary somatosensory area (SII) and insula (Ferretti et al., 2007), they did not provide a description of the temporal dynamics characterizing this network. This information was so far largely investigated in EEG (Cruccu et al., 2008) and MEG studies (see Hari and Forss, 1999), which identified several sequential components of the somatosensory processing time course.

ERP studies reported the presence of three families of ERP components: earliest components, likely reflecting SI activity, were elicited contralaterally by the stimulation of median nerve at about 20 ms (N20), while the peak following tibial nerve stimulation occurs at about 40 ms. Middle-latency peaks (N30 – N60) are typically recorded more anteriorly for median nerve stimulation, but their generators are still matter of debate and far from being clarified (Barba et al., 2001, Barba et al., 2002a, Barba et al., 2002b, Barba et al., 2005). Activation of parietal opercular cortex was found to be bilateral, typically beginning at 60–70ms, and continuing up to 200 ms for both nerves. Further active regions were located in posterior parietal cortex and in the premotor cortices. The techniques used so far suffer from limitations in either spatial (MEG) or temporal (fMRI) resolution that hinder the obtainment of detailed four-dimensional maps of brain activity. In fact, functional MRI technique has some important limitations (Logothetis, 2008) as the neuronal activity is only indirectly derived from a haemodynamic signal, and this can lead to erroneous localization, as BOLD signals are biased towards the larger veins (Disbrow et al., 2000). In addition, the poor temporal resolution of fMRI data (typically one sample each 2–3 s) hinders the determination of temporal sequences in stimulus processing, and questions its sensitivity to weak but consistent short-lasting neural activities. On the contrary, MEG data are limited to the cortical surface plus the superficial sectors of the sulci including the inter-haemispheric fissure and the cortical generators contributing to the somatosensory evoked fields can be only coarsely defined given the poor EEG (Burle et al., 2015) and MEG localization power (1–1.5 cm, Hedrich et al., 2017).

These limitations can be overcome by using the stereo-EEG technique, which, when sufficient number of subjects are tested, allows the combination of spatially precise anatomical data with temporally highly resolved electrophysiological findings. In a previous paper (Avanzini et al., 2016), we used stereo-EEG to depict the somatosensory processing after electrical stimulation of median nerve in a large group of epileptic patients. The main results were 1) the large extent of the cortical surface activated by hand somatosensory stimulation, extending far beyond SI and SII, and 2) the presence of several temporal patterns of activation, localized in different cortical areas. In particular, a fast phasic component characterized the primary somatosensory complex (SI), a delayed activity was observed in the dorsal sector of motor region and, finally, a tonic activity lasting more than 200 ms was present in the perisylvian region including posterior insular cortex. Overall, this study indicated that the somatosensory processing of hand stimuli is a complex and multi-stage hierarchical process. However, the stimulation of a single body effector prevented the discussion of the possible functional roles of each of these stages.

To overcome this limitation, in the present study we mapped the cortical processing following the stimulation of tibial, median and trigeminal nerve in 96 patients. The first aim of the present study was to demonstrate that phasic, delayed and tonic activities are not peculiar for hand somatosensory stimulation (Avanzini et al., 2016), but rather characterize the processing of somatosensory stimuli delivered to any body part. In addition, by evaluating the spatio-temporal arrangement of each temporal pattern and relating them to previous physiological and clinical studies, we provided evidence that the different temporal profiles represent the neural signatures of different somatosensory processing functions.

Materials and methods

Participants

Stereo-EEG data were collected in 96 patients (51 males, 45 females) suffering from drug-resistant focal epilepsy. Only patients presenting no (n = 76) anatomical alterations or little abnormalities outside the sensorimotor areas (n = 20), as evident on MR, were included. The latter group included 14 patients with positive MR showing alterations of the temporal lobe (4 hippocampal sclerosis, 3 minimal periventricular nodular heterotopia, 2 cavernomas of temporal pole, 2 focal cortical dysplasia - FCDII, and 3 post ischemic injuries), 4 patients presenting alteration of the posterior parietal lobe without overlap with active ROIs reported in the study (2 FCDII, one cavernoma and one post ischemic injury), one patient presenting a FCDII of the occipital lobe, and one patient with an anterior periventricular nodular heterotopia.

All patients were stereotactically implanted with intracerebral electrodes as part of their pre-surgical evaluation, at the “Claudio Munari” Center for Epilepsy Surgery, Ospedale Niguarda-Ca’ Granda, Milan, Italy. Implantation sites were selected purely on the basis of seizure semiology, scalp-EEG recordings and neuroimaging examinations. Patients were fully informed regarding the electrode implantation and sEEG recordings. The present study received the approval of the Ethics Committee of Niguarda Hospital (ID 939–2.12.2013). Intracerebral recordings were performed according to sEEG methodology to identify the cerebral epileptogenic zones as well as the structures involved in the irritative propagation of seizure activity (Munari et al., 1994, Cossu et al., 2005, Cardinale et al., 2013, Cardinale et al., 2016). No seizure occurred, no alterations in the sleep/wake cycle were observed, and the anti-epileptic pharmacological treatment was not modified during the 24 h prior to the experimental recording. Neurological examination was unremarkable in all cases; in particular, no motor or sensory deficit was found in any patient.

Electrode implantation

Most implantations were unilateral, as clinical evidence generally indicated the hemisphere generating the seizures. Only 8 of the 96 patients were implanted bilaterally, resulting in a total of 104 implanted hemispheres. A number of depth electrodes (range: 9–19: average 13) were implanted into different regions of the hemispheres using stereotactic coordinates. Each cylindrical electrode had a diameter of 0.8 mm and consisted of 8–18 2 mm-long contacts (leads), spaced 1.5 mm apart (DIXI^®, Besancon, France).

Immediately after the implantation, cone-beam computed tomography (CBCT) was obtained with the O-arm scanner (Medtronic, Minneapolis, Minnesota) and registered to pre-implantation 3D-T1-weighted MR images. Subsequently, multimodal scenes were built with the 3D Slicer software package (Fedorov et al., 2012), and the exact position of each lead was determined, at the single patient level, looking at multiplanar reconstructions (Dale et al., 1999). Following clinical conventions, all leads were labeled by a letter corresponding to the electrode shaft, followed by a number sequentially increasing starting from the tip of the electrode.

Nerve stimulation

The day after the implantation, patients were admitted to the neurology ward, to undergo clinical and neuropsychological tests to functionally characterize the explored cortical regions. To map leads processing somatosensory information, three separate blocks of peripheral electrical stimulation were administered to the patients by means of a Nihon Kohden - Neuropack M1 MEB-9200 stimulator. Stimulations are delivered to three peripheral nerves, i.e. median for the hand-, tibial for the foot- and mandibular branch of trigeminal nerve for the mouth-somatosensory processing. The present study received the approval of the Ethical Committee of Niguarda Hospital (ID 939, date 4.7.2014).

During the stimulations, patients are lying in bed with eyes closed. Stimulation was delivered contralaterally to the recorded hemisphere, using 100 constant-current pulses (0.2 ms duration) at 1 Hz. For patients with bilateral implantations, two separate sets of stimulations were administered. The intensity and exact site of stimulation were varied until an observable twitch (of thumb, foot and jaw respectively) was obtained. The motor threshold in our sample ranged from 6.3 to 13.8 mA for tibial nerve dataset, from 3.2 to 5.8 mA for median nerve stimulations and from 2.4 to 4.7 mA for trigeminal nerve ones. For each block and patient, the stimulation intensity was set at 10% above the motor threshold. As a control, most of the patients (n = 70) were also tested with a lower stimulation intensity, i.e. 20% below the motor threshold.

Anatomical reconstruction of electrodes

The aim of the reconstruction was to locate the recording leads precisely in the individual cortical surfaces using the multimodal explorations performed in each patient, and import these locations into a common template. Each patient underwent a structural MRI (Achieva 1.5T, Philips Medical Systems) before electrodes implantation. The T1 images were segmented using FreeSurfer software (Reuter et al., 2012) to identify the pial and the white matter surfaces of the native mesh of each patient. The quality of segmentation was verified by visual inspection of the resulting surfaces. The mid-thickness surface, i.e. the average surface in between the pial and white matter surfaces, was extracted. Moreover, the sulcal pattern was extracted from the pial surface using a procedure evaluating the normalized geometric depth of each point (folding index, Van Essen and Drury, 1997). The more negative this parameter, the deeper the point is buried in a sulcus. After the implantation, each patient underwent a volumetric brain CT, to locate precisely the recording leads using the artifacts generated on CT images. The MRI and CT datasets were co-registered (FLIRT, 6 degrees of freedom, mutual information, see Jenkinson and Smith, 2001) to get the anatomical brain data and the implanted electrodes into the same coordinate space. Thresholding of the CT signal intensity was used to segregate the recording leads, and to reconstruct their position in the CT volume. During this process, the knowledge of the number of implanted electrodes, the number of leads on each electrode, and their entry and target points were used as constraints to ensure a reliable reconstruction.

The space in between the pial and the white matter surfaces constitute a ribbon corresponding to the grey matter thickness. Since the spatial resolution of CT images exceeded that of the MRI imaging and the CT images were reduced to binary information (1: presence of a recording lead; 0: absence), the cortical ribbon was oversampled at a resolution of 0.4 mm in all directions. The intersection of CT images with the ribbon detected the contact points located in the cortical grey matter. Thus at this stage a lead located in the grey matter is represented by the number of voxels it shares with the grey matter. For those leads in the grey matter, the centroid of the artifact was computed and projected onto the nearest node of the mid-thickness native surface, thus obtaining a single surface node representing the projection of a lead located in grey matter.

Finally, the individual mid-thickness surface was resampled to match the number of nodes (163.842) of the template (Fs-LR-average) and co-registered with this template using Freesurfer_to_fs_LR pipeline (http://brainvis.wustl.edu/wiki/index.php/Caret:Download). In this step single lead representations were enhanced by arbitrarily adding 6 nodes (on the hexagonal mesh) surrounding the original single nodes on the native surfaces, to ensure that all leads in the individual surfaces were maintained on the template surface. To quantify the goodness of fit between the native meshes and the template, we computed, according to Abdollahi et al., (2014), two local indices (i.e. deformation and distortion, evaluated for each node of the mesh), indexing the shrinkage and translation of each node during the transformation.

In conclusion, the recording leads located in the grey matter are represented in three different formats in the reconstruction procedure: as the number of 0.064 mm³ native voxels, and as a fixed number (7) of nodes in the native and in the template meshes. While the first format provides the precise location in the cortical depth, the latter two provide precise information about location in the cortical surface (at mid-thickness).

The usage of a brain template allowed us not only to directly overlap results from different patients, but also to compare functional SEEG results with complementary information such as the location of cytoarchitectonical areas. Using CARET software (Van Essen, 2012), the position of each single lead related to cytoarchitectonic areas of sensorimotor regions (Geyer et al., 1999, Geyer et al., 2000, Grefkes et al., 2001), superior and inferior parietal cortex (Scheperjans et al., 2008a, Scheperjans et al., 2008b, Caspers et al., 2006, Caspers et al., 2008, Choi et al., 2006), parietal operculum (Eickhoff et al., 2006), and areas 44 and 45 (Amunts et al., 1999) The subdivision between dorsal and ventral premotor areas was made according to Tomassini et al. (2007). All these anatomical subdivisions are shown in Fig. S1 in both inflated and flattened views of the right hemisphere.

SEEG data recording and processing

The sEEG trace was recorded with a Neurofax EEG-1100 (Nihon Kohden System, Japan) at 1 kHz sampling rate. For each implanted patient, the recording procedure started with the selection of an intracranial reference, which was chosen by the neurologists using both anatomical and functional criteria. The reference was computed as the average of two adjacent leads both exploring white matter. These leads are selected by clinicians, ad-hoc for each patient, as leads not presenting any response to standard clinical stimulations, including somatosensory, visual and acoustical stimulations. Nor did their electrical stimulation evoke any sensory and/or motor behavior.

The neurologists visually inspected recordings for the presence of ictal epileptic discharges (IEDs) during the stimulation protocol. In 86 patients, the regions of interest were devoid of any IED. In the remaining 10 patients, sparse IEDs were recorded during the stimulation protocol. Even in these cases false positive responses were considered unlikely since none of the patients included showed reflex IEDs, and thus IEDs were not synchronized to the stimulus. This was confirmed by visual inspection of the quality of the data averaged.

Recordings from all leads in the grey matter were filtered (band-pass: 0.015–500 Hz, notch: 50 Hz) to avoid aliasing effects, and analyzed by means of complex Morlet's wavelet decomposition. Following previous intracranial studies (Vidal et al., 2010, Caruana et al., 2014a, Caruana et al., 2014b), gamma power was estimated for 10 adjacent non-overlapping 10 Hz frequency bands between 50 and 150 Hz, integrating EEGLAB routines into an in-house pipeline. The window used for the computation was frequency dependent and set to 3 cycles using the newtimef of EEGLAB. This gamma power was computed at 60 time points equally distributed between [-100; 500] ms relative to the electrical stimulation delivery. This sampling of gamma power each 10 ms provides an appropriate description of the post stimulation time course of neural activity. Note that power for each frequency was transformed into z-scores relative to the pre-stimulus interval to normalize data across patients and leads, and subsequently the overall gamma reactivity was computed as the average z-score of the 10 frequency bands covering the range 50–150 Hz.

Leads responsive to each nerve stimulation were identified by applying a paired t-test, where power for each post-stimulus time bin was contrasted against the average pre-stimulus period. Significance threshold was adjusted according to Bonferroni correction (N = 50, p = .001) to account for multiple comparisons. In consideration of the lack of complete independence across time bins, and to further decrease the false positive ratio, only leads with significant gamma increase in at least 3 consecutive time bins of which the peak amplitude exceeded 0.25 z-scores were designated as responsive.

Within each dataset, a k-mean clustering was applied to the gamma power time course of the responsive leads, independently of their location. Nineteen clustering procedures were performed imposing increasing numbers of clusters (2−20), and computing silhouette values (Rousseeuw, 1987) to evaluate clustering validity. The optimal clustering was defined by the maximal average silhouette value. Leads with negative silhouette values were included in responsiveness analyses, but discarded from any subsequent cluster-specific computation. A repeated-measurement ANOVA with time and cluster as factors was conducted for each stimulation dataset to validate the functional difference across the identified clusters. T-tests (p < .001 to account for 50 time bins) were used to evaluate post-hoc comparisons.

Chi-squared tests were conducted to test the homogeneity of cluster distribution across the three datasets (see Table 1), across the three datasets but separately for left and right hemispheres (see Table S1), and finally, across anatomical ROIs, separately for each nerve and hemisphere (see Table 2). The ROIs considered for this analysis were SI complex (3a, 3b, 1 and 2), primary motor area, premotor regions, anterior parietal cortices (PFs plus IPs), parietal operculum (OPs) and insular cortex. In all these tests, the absolute number of leads responding following a cluster-specific pattern served as variable. In addition, since all three nerve stimulations were tested on all leads, we defined the number of non-responsive leads, leads responding specifically to only one nerve stimulation, bi-nerve leads responsive to two out of the three stimulations and, finally, unspecific tri-nerve leads activated by all three stimulations (see Table S2&S3). A chi-squared test was used to evaluate whether the proportion of multi-nerve leads varied across the two hemispheres.

Table 1.

Characteristics of the time course clusters in datasets relative to the 3 nerve stimulation; raw data are on the left, including the number of leads per cluster, its percentage over the total number of responsive leads and the number of leads with negative silhouette; patients, average and SD silhouette, peak amplitude and timing refer only to leads with positive silhouette.

TIBIAL	Number of Leads	Percentage	Negative Silhouette	Patients	Average Silhouette	Standard Deviation Silhouette	Gamma power peak	Peak Timing
Early Strong-phasic	32	8,4%	2	11	0,429	0209	3,29	30-40 ms
Late Strong-phasic	26	6,8%	1	12	0,459	0224	4,23	40-50 ms
Middle-phasic	47	12,3%	5	16	0,357	0189	2,17	40-50 ms
Weak-phasic	97	25,5%	8	34	0,398	0157	0,96	40-50 ms
Delayed	22	5,8%	7	9	0,212	0135	1,31	60-70 ms
Tonic	157	41,2%	0	62	0,605	0150	0,22	110-120 ms
	381	100%	23
MEDIAN	Number of Leads	Percentage	Negative Silhouette	Patients	Average Silhouette	Standard Deviation Silhouette	Gamma power peak	Peak Timing
Strong-phasic	154	14,2%	7	35	0,557	0209	4,42	20-30 ms
Middle-phasic	228	21,0%	8	44	0,479	0192	2,42	20-30 ms
Weak-phasic	343	31,5%	0	64	0,460	0166	1,03	20-30 ms
Prolonged	78	7,2%	29	16	0,264	0155	2,31	20-30 ms
Tonic	285	26,1%	12	74	0,472	0182	0,31	50-60 ms
	1088	100%	56
TRIGEMINAL	Number of Leads	Percentage	Negative Silhouette	Patients	Average Silhouette	Standard Deviation Silhouette	Gamma power peak	Peak Timing
Strong-phasic	10	3,9%	1	6	0,543	0107	3,07	20-30 ms
Middle-phasic	31	12,1%	7	12	0,335	0149	1,64	20-30 ms
Weak-phasic	79	30,7%	0	32	0,447	0162	0,70	20-30 ms
Delayed	35	13,6%	3	12	0,344	0177	0,96	50-60 ms
Tonic	102	39,7%	1	42	0,487	0178	0,22	70-80 ms
	257	100%	12

Table 2.

Distribution of the time course clusters of the 3 nerves in anatomical regions. Each cell contains the numbers of leads and, in brackets, the individual chi[HYPHEN]values against the overall distribution within each hemisphere. Chi[HYPHEN]values are hatched in red if significantly larger than expected (chi > 5) and blue if significantly smaller than expected (chi < 5)

When region of interest analysis was performed to compare responsive patterns across nerves, before averaging leads from a given region, we normalized their gamma power time course, by setting the peak of each single lead equal to 1. In this way, the temporal pattern across datasets was highlighted, discarding the mere amplitude differences.

Responsiveness maps

By spatially filtering using the notion of geodesic distance (Knutsen et al., 2010), we generated for each nerve the overall responsiveness map and relative responsiveness maps (Avanzini et al., 2016). To this aim, for each cortical node we defined the nodes within a disk of 1 cm of geodesic radius from the original node, and weighted the contribution of each node by logistic function with unitary amplitude, a steepness of 2 and a midpoint at 7.5 mm. As a result, each node of the cortical mesh was associated with a collection of surrounding nodes, with all nodes within 5 mm from the origin maximally weighted, while those between 5 and 10 mm were gradually reduced in weight, to avoid edge effects. A threshold of 10% was used for the overall responsiveness, while for relative responsiveness maps the threshold was set at the chance level for a given cluster to appear (1/n, e.g. with 5 clusters, 1/5 = 20%).

We computed two relationships among the three-nerves datasets. Within each dataset, we compared the overall responsiveness between the two hemispheres via a chi-squared test in order to determine lateralization effects of cortical somatosensory processing. As this test was computed once per each node exceeding the 10% of overall responsiveness in at least one hemisphere, a false discovery rate (FDR, p < .05) approach was used to correct for multiple comparisons. In consideration of the fact that different populations are tested for left and right hemispheres, we restricted results by applying three constraints: 1) we considered results only from regions sufficiently covered in both subsamples (more than 3 leads in the disk), 2) we considered nodes whose absolute difference between left and right overall responsiveness was higher than 25% and, finally, 3) we applied a spatial clustering (minimum size: 300 nodes) so as to include only spatially reliable results.

Subsequently, to compare results across nerves we superimposed on the same map the results of all three nerve stimulations, using the three components of the RGB color map. Tibial responses were quantitatively rendered by the intensity of the green palette, median responses were reported in red and trigeminal ones were indicated in blue. Thanks to this coding, regions presenting an overlap of hand and foot representation tended towards yellow, joint hand and mouth representations to purple tones and foot-mouth responsiveness to cyan. Finally, any cortical area that is responsive to all three stimulations will be colored in white. Color labeling was restricted to regions whose overall responsiveness was higher than 10%.

Results

Recordings were obtained from 16599 leads in 96 patients, of which 11878 (5632 in left hemisphere, 6246 in the right) were localized in the cortical grey matter according to the anatomical reconstruction procedure (see Avanzini et al., 2016). The patient population tested is the same as that reported in the previous paper concerning the activation by median nerve stimulation, except for three patients who underwent only a partial stimulation protocol. Hence, the sampling density maps computed for the two hemispheres (Fig. S2) are virtually the same as those in Avanzini et al. (2016). They show the extensive, bilateral coverage of the cortical sheet, with sampling density peaks (yellow colored areas) distributed bilaterally over mesial temporal areas, middle and superior temporal gyri, parietal and frontal opercular regions, inferior frontal areas and anterior mesial regions. In Fig. S2 a threshold of 3 leads per disk (corresponding roughly to 1 lead/cm²) was applied to visualize only the sufficiently sampled cortical regions. Only the frontal and occipital poles of the two hemispheres, as well as sectors of the cortical crown, were under-sampled (see regions lacking color in Fig. S2), due to the surgical procedure and to the anatomical and vascular surgical constraints. These regions were excluded from subsequent analyses, so as to avoid biasing results by small numbers of leads tested, and they will be indicated in the images when relevant. In the next paragraphs, we will first report separately the results obtained by stimulating each of the three nerves.

Tibial nerve stimulation

Statistical analysis revealed that 381 of the leads exploring grey matter presented a significant increase in broadband gamma power (50–150 Hz) in response to tibial nerve stimulation (178 in the left hemisphere, 203 in the right). Fig. 1 shows the overall (and timeless) responsiveness (OR, responsive leads in percentage of explored leads) maps for the two hemispheres. As expected, OR was mostly located in the dorsal part of the sensorimotor cortex, extending onto its mesial surface. On the left side, the activation peaks are located in lateral and dorsal areas 3a and 4 (Fig. 1A, OR around 80%). Large OR values were also found in their mesial counterparts (Fig. 1C). Other reliable activations were present in areas 3b, 1, and in the upper part of dorsal premotor cortex (all reaching over 50% OR). Of note, the large responsiveness of mesial cortices, encompassing mesial area 5, primary somatosensory regions and supplementary motor areas. In addition, a strongly responsive zone located in opercular and insular cortices was also found, including parietal operculum and posterior insular long gyrus (OR peaks at 40%). The map obtained for right hemisphere (Fig. 1B and D) was largely similar to that in the left hemisphere (Fig. 1A, C).

Fig. 1.

Overall Responsiveness maps for tibial nerve stimulation. Overall responsiveness (responsive leads as a percent of total explored leads per disc) maps for left (A) and right (B) hemispheres. Only surface nodes with values exceeding 10% are shown. Panels C and D show the same datasets, respectively, but shown from a medial perspective onto the inflated hemispheres, allowing to appreciate the continuity of maps in the mesial regions without the cuts inherent to the flat maps. White borders refer to cytoarchitectonic areas of sensorimotor regions (Geyer et al., 1999, Geyer et al., 2000, Grefkes et al., 2001), superior and inferior parietal cortex (Scheperjans et al., 2008a, Scheperjans et al., 2008b, Caspers et al., 2006, Caspers et al., 2008, Choi et al., 2006), parietal operculum (Eickhoff et al., 2006), and areas 44 and 45 (Amunts et al., 1999). The subdivision between dorsal and ventral premotor areas was defined according to Tomassini et al. (2007). The separation between mesial and dorsal premotor cortices (PMm and PMd) follows the crown of the hemisphere. In addition, long and short gyri of insula were anatomically defined using the gyral pattern. Black hatchings indicate under-sampled regions.

To rule out that the responses of motor and premotor areas were due to the intensity of stimulation, just exceeding motor threshold, we computed the overall responsiveness for a second set of recordings, obtained in 70 patients, additionally tested with a lower intensity, specifically set at 20% below the motor threshold. The proportions of active leads in the primary somatosensory complex (areas 3a, 3b, 1 and 2) and motor cortex (area 4 and premotor regions) areas did not differ between the supra- and sub-threshold data (see Fig. S3, panel A, B) in either hemisphere (χ² = 2.76 for left hemisphere, χ² = 3.02 for right, both ns).

Trigeminal nerve stimulation

Statistical analysis revealed that 257 of the leads exploring grey matter presented a significant gamma power increase in response to the stimulation of the mandibular branch of the trigeminal nerve (82 in the left hemisphere, 175 in the right). The overall responsiveness map of the left hemisphere (Fig. 2A) showed the involvement of ventral parts of primary somatosensory areas 3b (OR: 42%), 3a (OR: 35%) and, at a lesser degree, area 2 (OR: 29%). No activity was found in area 4, nor in dorsal (PMd) and ventral premotor areas (PMv). As for the tibial nerve, perisylvian regions were responsive, including sectors of the parietal operculum (particularly OP1 and PR).

Fig. 2.

Overall Responsiveness maps for trigeminal nerve stimulation. Overall responsiveness (responsive leads as a percent of total explored leads per disc) maps for left (A) and right (B) hemispheres. Only surface nodes with values exceeding 10% are shown. Same conventions as in Fig. 1. LgI: long insular gyri. White hatchings indicate under-sampled regions.

The overall responsiveness pattern of the right hemisphere (Fig. 2B) was very similar for the primary somatosensory areas and perisylvian regions. However, the responsiveness of motor and premotor regions was strikingly different. While these areas were silent in the left hemisphere, two sectors were responsive in area 4 as well as large regions in both right PMd and PMv.

The proportions of active leads in the primary somatosensory complex (3a, 3b, 1 and 2) and motor cortex (4 and premotor regions) areas did not differ between the supra- and sub-threshold data in the right hemisphere (χ² = 2.99, ns). This analysis was not performed for the left hemisphere given the lack of responsiveness in the supra-threshold set recordings of motor cortex (see Fig. S3, panel E, F). Thus the response to electrical stimulation of the trigeminal nerve, as that of the tibial nerve, reflects primarily sensory processing.

Median nerve stimulation

The neural response following median nerve stimulation obtained using sEEG were reported in a previous paper (Avanzini et al., 2016). Here, results are based on 96 out of the 99 patients included in that publication. In the present sample, 1088 leads (489 left, 599 right) were responsive to median nerve stimulation, resulting in an overall responsiveness pattern virtually identical to that previously reported (Fig. S4).

Overall pattern of activation following stimulation of the 3 nerves

Once identified the topographical distribution of cortical regions responsive to each of the three nerves, we merged the overall responsiveness maps into a single somatotopic map for the two hemispheres (Fig. 3). These maps, based on 632 and 753 responsive leads in left and right hemisphere respectively, show both the territories belonging to single nerve representations (green, blue and red as before) and the regions in which two nerves are represented (yellow for tibial-median and pink for median-trigeminal overlap), with white corresponding to response to all three stimulated nerves.

Fig. 3.

Overall somatotopic organization. Overall responsiveness to the three nerve stimulations are superimposed on the same map for left (panel A) and right (panel B) hemisphere by using the three components of the RGB color map. Tibial responses are rendered by the intensity of the green palette, median responses in red and trigeminal in blue. Yellow regions correspond to overlap of hand and foot representations, purple regions indicate joint hand and mouth representations. Finally, cortical areas responsive to all three stimulations are colored white. Only surface nodes with values exceeding 10% are shown. Same conventions as in Fig. 1. Refer to Fig. S1 for the full description of cytoarchitectonic boundaries.

In both hemispheres (Fig. 3) a classical somatotopic organization with the foot represented dorsally, the hand in a middle position and the mouth ventrally, was evident in primary somatosensory cortex, with overlapping regions in the transition between median and trigeminal nerve territories. These transition regions generally correspond to the presence of bi-nerve leads responsive to the two overlapping nerves (Fig. S5). It is worth noting that the foot representation extends onto the medial surface, beyond the four-cytoarchitectonic areas corresponding to the primary somatosensory complex as defined in Geyer et al. (2000). This observation is in line with previous MEG (Hari and Forss, 1999) and intracranial studies (Allison et al., 1996), suggesting this area as the continuation of the lower-limb primary somatosensory cortex.

Rostrally, in the motor and premotor cortices, the three nerve stimulations elicited activations, which followed the classical homonculus pattern, with the exception of the absence of trigeminal responses in the left motor and premotor areas. Compared to SI, larger regions of overlap were found in the motor region, in particular in the right hemisphere (see Fig. 3B). Interesting to note the rich responsiveness of anterior intraparietal sulcus and surrounding cortices, which appear to be exclusively devoted to hand representations.

With respect to the perisylvian regions, three nerves representations overlapped in opercular and insular cortices (see also white dots in Fig. S5), occurring mainly in the posterior parietal operculum and the dorsal part of the second long gyrus of the insula. A site in anterior parietal operculum, corresponding roughly to area labeled PR by Disbrow et al. (2000), was recruited by both hand and mouth stimulations. Finally, when balancing the levels of overall responsiveness for the three nerves, no clear somatotopic arrangement emerged in SII. However, when increasing the threshold for the responsiveness to median nerve stimulation, a weak somatotopic arrangement emerged in the whole parietal operculum, with the foot represented ventrally (mainly OP2), the hand mainly centered on the posterior part of OP1 and, finally, the mouth occupying a more anterior sector of OP1 (Fig. S6 top). Of note, little activity was found in OP4, and indeed only 7/55 and 14/66 leads were responsive in left and right OP4 respectively, compared to 33/56 and 33/63 in left and right OP1 (Fig. S6 bottom).

These results are largely in agreement with previous results, thus indicating that sEEG technique provides valuable results about the physiological activity of brain processing. The major advantage of the technique, however, is the detailed temporal resolution it yields for recordings throughout the cortex. The subsequent sections exploit this novel advantage.

Prototypical time courses of responses to nerve stimulation

To characterize the temporal pattern of the responses to each nerve stimulation, we performed a clustering procedure over the time course of gamma power for all responsive leads. Of note, this step was performed independently of the location of individual leads. Fig. 4 shows the average time course for each cluster of leads in the three different nerves (tibial in panel A, median in panel B and trigeminal in panel C), along with the respective standard deviation intervals (see Fig. S6 for corresponding ERPs). Table 1 reports for each nerve and cluster, the quality of the clustering (less than 5% negative silhouettes, ie misclassified leads), the number of leads and patients contributing to each specific cluster, as well as their temporal characteristics in terms of peak amplitude and latency, underscoring the robustness of the clustering. Since responsive leads were grouped according to the similarity in their time course, ANOVA not surprisingly returned significant main effects of TIME and CLUSTER, as well as significant TIME*CLUSTER interactions for all three nerves. The significance of these interactions (tibial: F_5,295 = 193.8, p < .0001; median: F_4,236 = 709.7, p < .0001; trigeminal: F_4,236 = 92.1, p < .0001) confirm that the clusters differed in time courses.

Fig. 4.

Clustering of time courses. Results of clustering on gamma power time course are shown for all the three datasets (tibial in panel A, median in B, trigeminal in C). All curves (color code, see inset) represent the cluster centroid ± standard deviation. The time window is limited to [-20; 300] ms relative to the stimulus delivery.

In all datasets, three prototypical response families consistently appear. The first is composed by early and short-lasting phasic responses, further subdivided into three clusters according to their peak amplitude. These classes are labeled strong, middle and weak phasic (blue, yellow and red curves, respectively). As expected from literature, the latency of phasic responses after trigeminal stimulation was the shortest (peaks at 10–20 or 20–30ms), and tibial the longest (peaks at 30–40 or 40–50ms). Note that in tibial nerve dataset there were two strong phasic clusters rather than the only one seen in the other nerves. This was most likely due either to the variable height of the enrolled patients (see Baumgärtner et al., 1998) and/or to the difference in fiber conduction velocities that becomes more evident the longer is the pathway.

The second prototype is represented by phasic responses, whose time course was prolonged in the case of median nerve (with response ending at 70 ms), or delayed in the case of tibial and trigeminal nerves, with similar peak latencies around 60–70ms (green curves). The third family is represented by long-lasting tonic responses (black curves), which did not exhibit a clear peak, but rather a sustained activation with shallow ascending and descending phases, lasting for more than 200 ms.

A chi-square test was performed to examine the distribution of different clusters across the three datasets. The distributions differed significantly across nerves (Table S1, χ² = 80.19, p < .0001). In particular, the individual chi values (see Table 2), indicated that trigeminal nerve stimulation evoked a high proportion of delayed responses (trigeminal: 13% compared to median: 7.2% and tibial 5.8%), while median nerve stimulation elicited a lower proportion of tonic responses relative to the other two nerves (median: 26.2% compared to tibial: 41.2% and trigeminal 39.7%).

Integrating time and space: maps of time-course clusters

To evaluate whether the different clusters were prevalent in specific brain regions, we conducted a chi-square test for each dataset and hemisphere on the number of leads attributed to each cluster and exploring the SI complex (3a, 3b, 1 and 2), primary motor area, premotor cortices, anterior parietal regions (IPs plus PFs), opercular areas and, finally, insular cortex. All tests indicated significantly different distributions across those regions (see Table 2). For median nerve, strong phasic responses were mainly present in SI (33 and 39 leads for left and right hemisphere), primary motor area housed mostly middle or weak phasic patterns, premotor cortex showed a high proportion of delayed responses and, finally, tonic responses were prevalent in opercular and insular cortices. The same class of responses was vastly under-represented in suprasylvian sensorimotor regions.

Despite the lower number of active leads for tibial and trigeminal stimulations, the chi-square test on these datasets confirmed the differences in time-course distribution in the different anatomical regions, notably the specificity of tonic pattern for peri-sylvian regions and anterior parietal areas. Less consistent were the findings about the strong-phasic responses in SI for the tibial nerve. Indeed, only 5 leads (2 left, 3 right) covering SI responded in a strong phasic fashion to tibial nerve stimulation, while 17 (10 left, 7 right) were identified in primary motor area. As one can see from Fig. 4, the low number of strong phasic responses in SI for tibial nerve stimulation is due to the restriction of SI cytoarchitectonical boundaries to the lateral cortical surface (Geyer et al., 1999, Geyer et al., 2000), with no extension in the mesial surface, where many leads exhibited a strong-phasic (SP) pattern. Considering the presence of strong-phasic responses on this part of mesial surface, one may suggest that this is a prolongation of the SI foot representation.

In order to explore the topographical cluster-specific distributions without a-priori constraints, we computed the relative responsiveness maps, i.e. the percentage of responsive leads belonging to a cluster, thresholding the results to show only regions exceeding the chance level (see Methods). Such maps, which are informative of which regions house prevalently a given response cluster, are shown in Fig. 5, Fig. 6, Fig. 7.

Fig. 5.

Strong phasic cluster maps. A & C: relative responsiveness (leads belonging to one cluster as a percentage of total number of responsive leads per disc) maps of left (A) and right (C) hemispheres for strong phasic cluster. Color code as in Fig. 3. Only nodes with values exceeding the chance level (20% with 5 clusters, 16.6% with 6 clusters) are shown. Same anatomical conventions as in Fig. 1. B: cluster centroids for the three datasets, with all curves in grey except the strong phasic ones, colored code as in the maps. Black hatchings indicate under-sampled regions.

Fig. 6.

Delayed/prolonged cluster maps. A & C: relative responsiveness (leads belonging to one cluster as a percentage of total number of responsive leads per disc) maps of left (A) and right (C) hemispheres for delayed/prolonged cluster. Color code as in Fig. 3. Only nodes with values exceeding the chance level (20% with 5 clusters, 16.6% with 6 clusters) are shown. Same anatomical conventions as in Fig. 1. B: cluster centroids for the three datasets, with all curves in grey except the delayed/prolonged ones, colored code as in the maps. PMd: dorsal premotor cortex.

Fig. 7.

Tonic cluster. A & C: relative responsiveness (leads belonging to one cluster as a percentage of total number of responsive leads per disc) maps of left (A) and right (C) hemispheres for tonic cluster. Color code as in Fig. 3. Only nodes with values exceeding the chance level (20% with 5 clusters, 16.6% with 6 clusters) are shown. Same anatomical conventions as in Fig. 1. B: cluster centroids for the three datasets, with all curves in grey except the tonic ones, colored code as in the maps. SgI and LgI: short and long insular gyri, respectively.

Fig. 5 shows the relative responsiveness map for strong-phasic cluster. The tibial nerve representation in the left hemisphere (Fig. 5A) occupied the dorsal most part of primary somatosensory complex, extending to its mesial counterpart. The same time-course following median nerve stimulation mapped onto the middle part of the primary somatosensory complex extending to anterior intraparietal sulcus. Finally, trigeminal nerve stimulation produced strong phasic responses only in a focal ventral sector of areas 3a and 3b. The strong phasic map of the right hemisphere (Fig. 5C) followed a distribution virtually identical to that in the left hemisphere, with a slightly larger involvement of the parietal lobe and of motor cortex for the median nerve. Of note, the three nerve representations showed no overlap in the strong-phasic maps, contrary to the overall somatotopy (Fig. 3).

The prolonged/delayed activity map was sparse in the left hemisphere (Fig. 6A). In contrast, this cluster was more extensive in the right hemisphere, covering various sectors of the motor system (panel 6C). It is worth noting the presence in this map of regions with overlapping hand and mouth responses in the primary hand motor area, and overlap between hand and foot responses at the junction between area 4 and PMd. Isolate sites in response to trigeminal and tibial nerve stimulations were found in ventral primary motor area and dorsal premotor cortex, respectively.

To clarify the interaction between phasic and delayed responses in the two primary motor sites responsive to trigeminal and median nerve stimulation, we built two ROIs corresponding to these two sites (labeled 4d1 and 4v, Fig. S7A). For the leads located in these ROIs (12 leads from 8 patients in 4d1, 14 leads from 6 patients in 4v), we computed and averaged the normalized time course of gamma power. In 4d1 the responsiveness to median nerve (red) is prolonged and that to trigeminal nerve (blue) is delayed to create a 50 ms period during which both representations are strongly active (Fig. S7B). In 4v, on the contrary, a phasic activity is visible in response to median nerve stimulation, while the temporal pattern in response to trigeminal nerve is a mixture of phasic and delayed responses, resulting in a double-peak average profile (Fig. S7C). Next to the 4d1 ROI there is a small region of overlapping delayed/prolonged activity for median and tibial (4d2), creating a short 30 ms period of simultaneous strong activity in these two representations.

The tonic relative responsiveness map of left hemisphere (Fig. 7A) showed a very focal pattern. The activation was almost entirely housed in the perisylvian region, including OP1, OP2, PFcm, and the posterior insular long gyrus, with an additional site adjacent to ventral premotor cortex and area 44. Notably, operculo-insular regions showed a strong overlap among the three nerve representations (white in Fig. 7). The tonic map for right hemisphere (Fig. 7C) resembled the left one, but with the addition of the dorsal part of the first insular long gyrus responsive to all three nerves. Very little somatotopic arrangement was discernible in any part of these tonic maps. Given the extremely large overlap among all tested nerves, we evaluated also the number of single leads responsive to all three stimulations (tri-nerve leads). Virtually all tri-nerve leads were consistently tonic (red dots in Fig. S8) and were located in the perisylvian regions, predominantly on the right side. The association of multi-nerve responsiveness and right hemisphere (Table S2) was very significant (χ² = 22.94, p < .0001), throughout the parietal operculum (Table S3).

The ERPs obtained in exemplar leads with strong phasic, delayed/prolonged and tonic behavior for the three nerve stimulations are reported in Fig. S9.

Discussion

Our results show that electrical stimulation of peripheral sensory nerves activates a wide region of cortex including SI, SII, M1, dorsal and ventral PM, SMA, posterior insula, and rostral parts of PPC and of the parietal operculum. Viewing these maps through the glasses of time course categorization identified for the first time different patterns of temporal responsiveness across these cortical areas.

Timeless topography of areas responsive to the three nerve stimulation

Before discussing the main findings of our study, i.e. the decomposition in time of somatosensory processing, it is worth to mention the novel results revealed by the timeless responsiveness maps. While the activation in SI, SII and the insula is largely documented by previous literature, two findings are of particular interest. The first is the consistent activation of motor cortices, including both M1 and premotor areas, regardless of the presence of muscle contraction. This is in agreement with electrophysiological studies in the monkey, reporting the presence both in M1 and in the premotor area of a large population of neurons which are activated by proprioceptive or tactile stimulations with clear and well-defined receptive fields (Lemon and Porter, 1976, Rizzolatti et al., 1981, Gentilucci et al., 1988). Surprisingly, fMRI experiments failed to show a somatosensory activation of motor areas following electrical stimulation of peripheral nerves. The only exception is a recent paper by Sanchez-Panchuelo et al. (2016), who reported in a 7T fMRI study a BOLD activation of motor and premotor cortices following intraneural stimulation of the median nerve. The inconsistency between 3T fMRI on one side, and 7T fMRI and single neuron recordings on the other is discussed below.

The other interesting finding was the engagement of inferior parietal cortex, and in particular of the anterior intraparietal sulcus (phAIP), following hand stimulation, while no activity could be detected either for the mouth or for the leg. This is in agreement with neurophysiological data that indicate that both in the monkey (Murata et al., 2000) and in humans (Dijkerman and de Haan, 2007) AIP is involved in transforming object affordances into appropriate hand shapes. It is however somehow surprising that virtually all the inferior parietal lobe appears to be devoted to hand representation, lacking any representation of the mouth and legs. This is consistent with the notion that the development of the hand activity was fundamental in the evolution of primates for manipulative actions specialization and tool creation.

The phasic somatotopic maps

The strong-phasic maps are more selective than the overall responsiveness maps (compare Fig. 3, Fig. 5) and match the classical representation of body parts as demonstrated by single neuron recordings in macaque SI (Kaas et al., 1979, Iwamura and Tanaka, 1978). In agreement with these findings, we also found a smaller representation of the mouth in areas 1 and 2, which are relatively new in the primate evolution (Krubitzer and Disbrow, 2008). These strong-phasic maps most likely reflect the direct input from human ventral posterior (VPL and VPM) thalamic nuclei to SI. Indeed, in macaque VP projects strongly to areas 3a, 3b, 1, 2 and weakly to area 5 (Padberg et al., 2009).

Weak/middle-phasic maps include fields mainly outside those showing strong-phasic responses. Besides some territories in SI, weak/middle phasic responses covered motor and premotor areas, intraparietal sulcus, as well as small regions in the parietal operculum. It is likely that these maps reflect cortico-cortical connections rather than direct thalamic input from VPL and VPM. The direct route from nucleus VL to motor cortex is the other main candidate, but peripheral inputs to its neurons are few and poorly characterized (Asanuma et al., 1974).

Concerning somatosensory regions, two different kinds of connections may account for this weak/middle phasic responses. One could be represented by the lateral intra-areal connections, which have been documented in the macaque monkey (Reed et al., 2008). Such lateral connections may generate the bi-nerve responses typical for the transitions between hand and mouth territories (see for example Fornia et al., 2016) or hand and foot territories and may play a key role in the reorganization following nerve lesions (Kaas et al., 1983). Alternatively, weak/middle-phasic responses might also reflect connections between neighboring areas by short subcortical association fibers (U fibers, Nieuwenhuys et al., 2007).

The comparison of the present study with previous fMRI investigations (see Ferretti et al., 2007) reveals that most of the areas found to be active in the present study and lacking in fMRI experiments are endowed with the weak/middle phasic temporal profile, suggesting that its detection requires sensitive recording techniques. In fact, in order to yield consistent fMRI haemodynamic variations, the underlying neural activity should be either strong (as in the case of strong-phasic responses), or prolonged in time (as in the case of tonic responses). If these conditions are not met, the area under the curve of the activation, i.e. the integral of locally generated energy, may be too low to result in a statistically significant BOLD variation at 3T. This interpretation could explain the discrepancy between electrophysiological and neuroimaging studies described above. As a further argument in favor of this interpretation, higher sensitivity 7T fMRI as that used by Sanchez-Panchuelo et al. (2016) allowed them to observe activations in motor and premotor areas.

The delayed/prolonged maps

These regions were predominantly localized in area 4 and PMd of the right hemisphere. While some of them were activated by only one nerve stimulation, two areas along the precentral gyrus showed overlapping delayed responses following stimulations of two nerves, namely median with either tibial or trigeminal nerves (areas 4d1 and 4d2 in Fig. S7). Integration of somatosensory responses from different body parts was described in the premotor cortex of the monkey (Rizzolatti et al., 1981). “Tactile” premotor neurons (70%) had their receptive field formed by one or two spatially separated responding areas, generally the hand and the mouth (Rizzolatti et al., 1981). Authors interpreted these findings as evidence in favor of a region in the agranular cortex organizing hand to mouth movements. Desmurget and coworkers (2015) recently described the presence, along the precentral gyrus, of both motor and sensori-motor hand-mouth synergies by delivering peripheral and intracortical electrical stimulation to intra-operative patients, extending previous observations to humans.

Our spatio-temporal data provide evidence of a possible mechanism for the integration of hand and mouth sensory responses. Two patches in the precentral gyrus responded in a delayed fashion to trigeminal nerve stimulation. The lower one was located in the mouth motor field, while the upper one was located in the hand motor field. This latter patch exhibits a delayed responsiveness pattern also following median nerve stimulation, with a time course prolonged up to 70 ms. The delayed time course for trigeminal nerve peaked exactly at the same latency. This temporal relationship between mouth- and hand-related responses, with the hand starting first and the mouth later, can be accounted for by the master role that the hand plays in typical hand-mouth interactions. Consistent with this interpretation, the tibial nerve stimulation evoked a cluster of delayed responses highly similar to the trigeminal one, reinforcing the notion of the hand playing generally the master role in the organization of multiple body-parts movements.

The prevalence of the right hemisphere for hand-mouth delayed responses is likely due to the lateralization of masticatory behavior to the right hemisphere, possibly related to the opposite lateralization of speech processing. The lack of responsiveness to trigeminal stimulation in the left motor cortex, observed here, supports this hypothesis, as well as fMRI data, showing consistently stronger activation of the mouth sector in the right than the left precentral gyrus during the observation of ingestive mouth movements (Buccino et al., 2004, Jastorff et al., 2010).

The tonic maps

One of the most intriguing findings of the present study is that the stimulation of all peripheral nerves elicited a different pattern of activity, i.e. tonic responses, almost exclusively restricted to perisylvian regions, notably parietal operculum and the posterior insula. No clear somatotopic arrangement was found for tonic responses, and this notion was reinforced by the fact that most of the tonic leads responded to two or all three-nerve stimulation. Given the lack of both spatial and temporal selectivity of tonic responses, these regions are unlikely to complement SI in encoding precise locations on body surface.

Previous studies have proposed that a network including SI, posterior parietal cortices (PPC) and the insula is involved in bodily awareness and tactile perception (Melzack, 1990, Berlucchi and Aglioti, 1997, Dijkerman and de Haan, 2007), where each area would play a different role. Indeed, while lesions of SI result in tactile and proprioceptive impairments but without higher-order body-awareness deficits, lesions to the PPC can result in body awareness disturbances like anosognosia for hemianesthesia (Spinazzola et al., 2008, Romano et al., 2014), or body ownership alterations like somatoparaphrenia (Gandola et al., 2012). In addition to the inferior parietal cortex, clear evidence indicate that also the insula plays a role in corporeal awareness. Patients with insular seizures may experience somatic hallucinations (Roper et al., 1993) and lesion of the same structure may determine somatoparaphrenia (Cereda et al., 2002). These conclusions were reinforced in a review by Karnath and Baier (2010) based on modern lesional analysis and neuroimaging data, demonstrating a prominent role of the right insula for our sense of limb ownership, and further strengthened by Tsakiris et al. (2007), who showed how a network composed by right posterior insula and right frontal operculum sustains our sense of limb ownership.

Overall, this picture returns a functional dissociation between SI and PPC/insula during somatosensory processing, which is extremely well fitting with the functional results obtained for phasic and tonic responses in our study. Indeed, while phasic responses may mainly subserve tactile processing and discrimination, the tonic multi-nerve activity recorded from perisylvian region (mostly for the right hemisphere) might represent the electrophysiological correlate of neural processes underlying tactile awareness and limb ownership (see Auksztulewicz et al., 2012 for further links between long latency activity and somatosensory awareness). The link between tonic activity and tactile awareness may explain also previous ERP studies (Schubert et al., 2006), which reported that early components are independent of stimulus perception, while for consciously perceived stimuli amplitude enhancements are observed for the component at 100 and 140 ms.

Finally, the association between awareness and tonic activity may not be restricted to somatosensory processing. Indeed, Fisch et al. (2009) demonstrated that visual awareness requires sustained activity in human inferior temporal cortex. Such notion would be of great impact not only for the clarification of neurophysiological mechanism leading to tactile awareness, but also for providing clinicians with an indication about a neural signature of perceptual awareness to be searched in presurgical evaluations.

The possible dual role of secondary somatosensory cortex

Despite the prevalence of tonic activity, SII showed also a non-negligible number of weak/middle phasic responses, in particular for the hand (11 leads with weak phasic response out of 66 responsive). We cannot be definite about the anatomical bases of phasic and tonic responses in SII. In macaques, SII receives direct somatic input from VPL and VPM (Jones and Powell, 1970) and VPi (Disbrow et al., 2003) as well as cutaneous input from areas 3b and 1 (Disbrow et al., 2003). Earlier, we proposed that the weak phasic activity in SII might reflect this serial cortico-cortical connection (Avanzini et al., 2016). However, a contribution of thalamic inputs cannot be excluded. Regardless of their origin, we propose that these phasic signals are the neural substrate for a complete perception of touch, which is partially impaired (hypoaesthesia) following lesions of SII (Preusser et al., 2015). This finding is not in contrast with the proposed role of right perisylvian areas in tactile awareness, because most patients of Preusser et al. (2015) had left hemisphere damage, and patients showing neglect were excluded from the study. Phasic signals in SII, given their predominance in the median nerve sample, might be at the basis of the interaction between SII and PMv (Matelli et al., 1986) and AIP (Borra et al., 2008) for the tactile control of object grasping, and manipulation in general.

We propose that the co-existence of two extremely different patterns of responsiveness within SII may indicate that this area plays a dual functional role in somatosensation. On one side, tonic activity might subserve tactile awareness. On the other side, phasic activity may reflect a role in touch processing for a variety of higher order tactile functions like texture discrimination and haptic processing (Sathian et al., 2011; see Sathian, 2016).

Appendix A. Supplementary data

The following is the supplementary data related to this article: