Sensorimotor adaptation ensures movement accuracy despite continuously changing environment and body. Adaptation of saccadic eye movements is a classical model of sensorimotor adaptation. Beside the well-established role of the brainstem–cerebellum in the adaptation of reactive saccades (RSs), the cerebral cortex has been suggested to be involved in the adaptation of voluntary saccades (VSs). Here, we provide direct evidence for a causal involvement of the parietal cortex in saccadic adaptation. First, the posterior intraparietal sulcus (pIPS) was identified in each subject using functional magnetic resonance imaging (fMRI). Then, a saccadic adaptation paradigm was used to progressively reduce the amplitude of RSs and VSs, while single-pulse transcranial magnetic stimulation (spTMS) was applied over the right pIPS. The perturbations of pIPS resulted in impairment for the adaptation of VSs, selectively when spTMS was applied 60 ms after saccade onset. In contrast, the adaptation of RSs was facilitated by spTMS applied 90 ms after saccade initiation. The differential effect of spTMS relative to saccade types suggests a direct interference with pIPS activity for the VS adaptation and a remote interference with brainstem–cerebellum activity for the RS adaptation. These results support the hypothesis that the adaptation of VSs and RSs involves different neuronal substrates.
Learning new motor skills, or optimizing existing ones, require the ability to detect and correct movement errors. Enduring errors can be progressively corrected thanks to adaptation mechanisms. In laboratory conditions, motor errors eliciting sensorimotor adaptation are induced by perturbations of the movement, or its goal. For visuo-manual adaptation, the perceived trajectory or the actual trajectory of the limb is perturbed respectively through optical distortion, or force fields. Previous studies indicate that visuo-manual adaptation involves neural plasticity mechanisms in the cerebellum (Weiner et al. 1983; Martin et al. 1996; Pisella et al. 2005; Luaute et al. 2009; Taylor et al. 2010; Werner et al. 2010; Donchin et al. 2012) and in parieto-frontal areas of the cerebral cortex (Clower et al. 1996; Della-Maggiore et al. 2004; Pisella et al. 2004; Hadipour-Niktarash et al. 2007; Luaute et al. 2009). In contrast, only the brainstem–cerebellar centers have so far been shown to contribute to the adaptation of eye movements (smooth pursuit: Chou and Lisberger 2004, vestibulo-ocular-reflex: Anzai et al. 2010, vergence: Takagi et al. 2001, saccades: See reviews in Hopp and Fuchs 2004; Tian et al. 2009; Iwamoto and Kaku 2010; Pelisson et al. 2010).
The saccadic system provides one of the most well-established models of sensorimotor adaptation, because saccades are too fast to permit online corrections of the ongoing eye trajectory and because the saccadic system has a major contribution to normal visual function (for reviews see Hopp and Fuchs 2004; Tian et al. 2009; Iwamoto and Kaku 2010; Pelisson et al. 2010). Reactive saccades (RSs) elicited by a sudden presentation of a visual target can be adapted by the classical saccadic adaptation paradigm (the “double-step target” paradigm: McLaughlin 1967). In this paradigm, a displacement of the saccadic target during the eye movements elicits an error signal. Because of saccadic suppression (Bridgeman et al. 1994), this intrasaccadic displacement is usually not consciously perceived by the subjects. Repeating this error signal over tens of trials leads to a progressive restoration of saccade accuracy, through recalibration of oculomotor commands. In daily conditions, most saccades are generated during the self-paced viewing of a visual scene and are based on target selection and motor decision processes driven by endogenous cues. Such scanning saccades belong to the category of voluntary saccades (VSs; for a discussion about the convenient and common use of the terms “reactive (or reflexive)” and “voluntary”, see Walker and McSorley 2006). The mechanisms of saccadic adaptation, mostly studied with RSs, have long been thought to involve exclusively the cerebellum and associated brainstem structures (for reviews see Hopp and Fuchs 2004; Tian et al. 2009; Iwamoto and Kaku 2010; Pelisson et al. 2010; Prsa and Thier 2011). However, mounting evidence suggest that the adaptation of RSs and VSs involves different mechanisms, and partially separate neural substrates (Erkelens and Hulleman 1993; Deubel 1995; Fujita et al. 2002; Hopp and Fuchs 2002; Gaveau et al. 2005; Collins and Dore-Mazars 2006; Alahyane et al. 2007; Cotti et al. 2007; Alahyane, Devauchelle et al. 2008, Alahyane, Fonteille et al. 2008; Cotti et al. 2009; Zimmermann and Lappe 2009, 2011; Hopp and Fuchs 2010; Panouilleres et al. 2011). By assessing the transfer of adaptation to saccadic, visuo-manual, or perceptual responses, these behavioral studies support the hypothesis that VS adaptation is acting on both sensory and motor processes and that RS adaptation is acting preferentially on motor processes. Two relatively recent lines of evidence even suggest that the parietal cortex may play a key role in the adaptation of VSs. First, 2 of the behavioral studies quoted above (Cotti et al. 2007, 2009) proposed that the early stages of the sensorimotor transformation affected by VS adaptation should rely on the frontal and/or the parietal cortex. Secondly, a recent functional magnetic resonance imaging (fMRI) study of saccadic adaptation highlighted the involvement of the posterior intraparietal sulcus (pIPS) in the adaptation of VSs, but not in the adaptation of RSs (Gerardin et al. 2012). To date, however, there is no direct evidence showing that the parietal cortex is actually necessary for saccadic adaptation.
The first aim of the present study was to assess the role of the pIPS in saccadic adaptation. We decided to target this area, which is thought to be the parietal eye fields (for review see: Muri 2006), because of the fMRI results reported by Gerardin et al. (2012). Further, we focussed on the pIPS of the right hemisphere because in the latter study, adaptation of leftward saccades was associated with activation in the contralateral (right), but not in the ipsilateral (left) hemisphere. The second aim of the study was to test whether the pIPS is similarly involved in the adaptation of VSs and RSs. To these aims, we used fMRI-guided single-pulse transcranial magnetic stimulation (spTMS) to interfere with the normal processing of the right pIPS during the adaptation of RSs and VSs.
Materials and Methods
Eighteen healthy subjects participated in the study. Of these, 6 participated only in the first experiment (VS experiment); 6 participated only in the second experiment (RS experiment); and 6 participated in both experiments. Thus, each experiment involved 12 subjects. The 2 groups of 12 subjects were matched in age (VS experiment: 29 ± 9.5; RS experiment: 29 ± 9.6 years) and gender (6 females and 6 males in each group). Thirteen participants were naïve to the goals of the study. All subjects had normal or corrected-to-normal vision, and no history of neurological or psychological disorder. Subjects gave their informed written consent and were financially compensated for their participation. All procedures complied with the Ethical Principles of the World Medical Association (Declaration of Helsinki) and were approved by the local ethics committee (CCPPRB Lyon B).
Each experiment involved 2 steps. In the first step, fMRI was used to localize the oculomotor region of the posterior part of the right intraparietal sulcus (pIPS) in each subject (see for detailed procedures: Gerardin et al. 2012). In the second step, spTMS was used to assess the role of the pIPS in saccadic adaptation.
fMRI was performed with a 1.5-T Siemens Sonata MRI scanner at CERMEP (Bron, France). Following a T1-weighted anatomical scan (1 × 1 × 1 mm), T2*-weighted functional scans (EPI, echo-planar imaging) were acquired from 29 axial slices covering the entire head (repetition time: 3000 ms, echo time: 50 ms, rotation angle: 80°, resolution: 3 × 3 × 3 mm). These anatomical and functional scans were performed using an 8-channel SENSE™ head coil.
The stimuli used during the functional scans were produced using light-emitting diodes (LEDs, 3 mm diameter) located at the rear of the magnet and visible through a mirror attached to the head coil. An infrared eye tracker (ISCAN, Woburn, MA, United States of America) located below the LEDs recorded, at a frequency of 240 Hz, the vertical and horizontal positions of the left eye via the same mirror. The eye tracker was calibrated by asking subjects to fixate a series of 9 LEDs covering the field used during scanning. The triggering of the scans, the monitoring of eye movements, and the control of the visual stimuli were controlled in real time by customary software.
Functional saccade localizer scans were obtained by alternating pro- and antisaccade blocks with central fixation blocks. During prosaccade blocks, a green fixation point indicated to the subjects that they had to make a saccade toward the target. During antisaccade blocks, a red fixation point indicated to the subjects that they had to make a saccade in the opposite direction, toward the mirror position of the target. Targets were presented for 1500 ms at a random location of 4°, 10°, 16°, or 20° in the left or right visual hemifield.
The fMRI data were analyzed with the Brain Voyager QX software (Brain Innovations, Maastricht, The Netherlands). Preprocessing of functional images involved correcting for slice scan time and head movements, temporal high-pass filtering (3 cycles), and linear trend removal. No spatial smoothing was used. The functional images were aligned with the anatomical data (transformed into Talairach space). For each subject, the oculomotor part of the right pIPS was identified by contrasting all blocks of saccades (pro- and anti-) with fixation blocks, and the locus of highest activity in the pIPS was used for the spTMS sessions.
spTMS sessions were run in a dark room. Subjects were sitting 57 cm away from a 140-Hz computer screen (30° × 40°), with their head stabilized by a chin rest, cheekbone rests, frontal support, and a band behind the head. The presentation of visual stimuli (6-mm diameter black circles on a gray background) on the screen was controlled by a Visual Stimuli Generation system (CRS, Cambridge, United Kingdom).
Binocular eye movements were recorded using an infrared tracker (Eyelink 1000; SR Research, Canada) with a frequency of 500 Hz and a spatial resolution of 0.05°. At the beginning of each session, the eye tracker was calibrated by presenting a sequence of 9 fixation points forming an array on the computer screen and asking subjects to fixate each point in turn. Laboratory-made software allowed online monitoring of eye movements, triggering of the visual stimulation, and triggering of the spTMS pulses relative to the timing of primary saccade detections. Eye-movement data were stored for offline analysis.
spTMS was delivered via a figure-of-eight coil (90 mm) coupled to a Magstim Rapid. Prior to each session, transcranial magnetic stimulation (TMS) was applied to the right motor cortex to determine the motor threshold, which was defined as the lowest stimulation intensity inducing a visible twitch in the resting contralateral hand in 5 of 10 trials (Schutter and van Honk 2006). The intensity of spTMS applied over the right pIPS and the vertex during the sessions was set to 120% of the motor threshold, corresponding to an average intensity of 67% (range: 52–84%) of the maximum output intensity (2T). The coil was precisely placed over the right pIPS by using a neuronavigation system (SofTaxicOptic, EMS s.r.l., Bologna, Italy), which was fed with the individual Talairach coordinates of fMRI identified pIPS. The coil was maintained in place with a hydrostatic arm (Manfrotto, Feltre, Italy) for the entire duration of the session.
Experimental Procedures of spTMS Sessions
The involvement of the pIPS in saccadic adaptation was assessed separately for VSs (VS experiment) and for RSs (RS experiment). Each experiment involved 4 spTMS sessions separated by at least 5 days. spTMS was applied to the right pIPS 30 ms after the detection of the primary horizontal saccade for one session, 60 ms after saccade detection for another session, and 90 ms after saccade detection in another session (Fig. 1). In the control session, spTMS was applied over the vertex. During this control stimulation condition, the timing of spTMS (relative to the detection of the primary horizontal saccade) was set at 30 ms for 4 subjects, 60 ms for 4 other subjects, and 90 ms for the remaining 4 subjects. A 4-way analysis of variance (ANOVA; with phase: pre × adaptation blocks × post, saccade direction, saccade type, and stimulus timing as factors) showed that, for this control session, there was no effect of timing on saccadic gain change (F2,18 < 2.34; P > 0.12). Accordingly, the data collected during the control session were pooled across the 3 timing conditions, separately for VS experiment and for RS experiment. The order of the 4 sessions was counterbalanced across subjects.
Each spTMS session involved 3 phases: a preadaptation phase; an adaptation phase; and a postadaptation phase. During the adaptation phase, adaptation was elicited using a classical double-step target procedure for RSs (McLaughlin 1967; Alahyane et al. 2007) or for VSs (Alahyane et al. 2007). The differences between adaptation protocols were maximally reduced, given the objective to elicit RSs and VSs. For both saccades types, 3 points were presented in each trial (Fig. 2), and subjects had to discriminate a letter located inside one of them. Trials always started by the production of a vertical saccade from the upper to the central point, followed by a horizontal (reactive or voluntary) saccade from the central to the lateral point. Therefore, the main differences between protocols reside on the mode of saccade triggering and on the number of targets present when the second saccade started, as detailed in the following. In both cases, the visual scene was shifted at the onset of the horizontal saccade and in the direction opposite to that of the saccade. This intrasaccadic “backward step” of the visual scene induced an error between the saccade endpoint and the target position. For the VS experiment (Fig. 2a), subjects had to explore a display containing 3 targets. Each adaptation trial started with the presentation of a fixation point, 4° above the center of the screen. After 1600 ms, a circle surrounding the fixation point and 2 targets appeared. One target (central target) was located immediately below the fixation point, and another target (lateral target) was located in the left or the right hemifield, at ±8°. The extinction of the circle 500 ms later indicated to the subjects that they had to make, first, a vertical saccade toward the central target, and then, a horizontal saccade toward the lateral target. The second saccade was initiated voluntarily, at the subject's own pace. Once this horizontal VS was detected (velocity threshold: 80–90°/s), the fixation point and the 2 targets were displaced in the direction opposite to that of the saccade. For the RS experiment (Fig. 2b), each trial started with the presentation of the fixation point, 4° above the center of the screen. Then, after 1700 ms, the fixation point suddenly jumped toward the center of the screen. Subjects were required to make a vertical saccade toward this central target. Then, after a random delay (800–1200 ms) following the end of this saccade, the central target was switched off and, simultaneously, a lateral target appeared in the left or the right hemifield, at ±8°. Subjects had to move their eyes toward this new target as soon as it appeared. Once this horizontal RS was detected (velocity threshold: 80–90°/s), the target was displaced backward. In both RS and VS experiments, the intrasaccadic step corresponded to 25% of the initial target eccentricity for adaptation blocks 1 and 2 (48 trials each), and to 40% of the initial target eccentricity for adaptation blocks 3 and 4 (48 trials each). spTMS was applied for all trials of the adaptation phase. The visual display disappeared 50 ms after the end of the horizontal saccade (Fig. 1). The 50-ms duration was based on 2 considerations. First, given that the effects of spTMS are usually short-lived, the use of short visual stimulus durations would restrict the temporal window over which the stepped target is visible and could thus increase the likelihood of interfering with adaptation. Secondly, it has been previously shown that a 50-ms target duration is sufficient to induce an optimal adaptation of both RSs and VSs (Panouilleres et al. 2011).
Both preadaptation and postadaptation phases consisted of a block of 24 trials without spTMS and a block of 24 trials with spTMS; in the latter block, the same spTMS timing as in the corresponding adaptation phase was used. In the preadaptation phase, the no spTMS block was performed before the spTMS block, whereas in the postadaptation phase, the sequence was reversed. Each 24-trial block comprised 12 rightward saccade trials and 12 leftward saccade trials, in random order. The design of the pre- and postadaptation trials was the same as for the adaptation phase, except that the visual display was switched off as soon as the horizontal saccade was detected.
To keep subjects attentive during sessions, we required them to perform a discrimination task. After every trial, they had to indicate (by pressing a key) whether the letter “E” (not shown in Fig. 2), which was displayed on the fixation point (for reactive trials) or the central target (for voluntary trials), was complete or truncated.
To evaluate if our 2 protocols were triggering the 2 different types of saccades, a separate, complementary experiment, ran without spTMS, was conducted on age-matched neurologically healthy participants (N = 5). In this experiment, we assessed the proportion in which adaptation transferred from RSs to VSs and vice versa. Results revealed that the transfers of adaptation between saccades elicited in the 2 protocols were similar to those reported in previous studies both for RSs and for VSs, hence validating our protocols as eliciting these different saccade categories (Supplementary Fig. 1).
TMS Data Analysis
The eye-movement data were analyzed using a custom program developed in Matlab v.7.1 (MathWorks Inc., Natick, MA, United States of America). Data from the left and right eyes were averaged, and all subsequent analyses were performed on the horizontal and vertical components of the cyclopean eye position. The start and end times of all saccades were identified based on a velocity threshold of 50°/s. Trials for which a primary saccade was not correctly detected online, or was contaminated by eye blinks, were eliminated; 4.8 ± 3.7% of all trials were eliminated in this way.
Each primary horizontal saccade was described by its latency, amplitude, and gain. Saccade latency was defined as the delay between horizontal saccade onset and the time of lateral target presentation for RSs, and between horizontal saccade onset and the time of vertical saccade completion (i.e. central target fixation time) for VSs. Saccade amplitude was computed as the difference between the initial and final positions of the eye. Saccade gain was obtained by dividing horizontal saccade amplitude by retinal error, the latter being defined as the difference between the target position and the starting position of the saccade. In each spTMS session, the mean saccadic gain was computed separately for rightward and leftward saccades and for the 8 blocks of trials (2 preadaptation blocks, 4 adaptation blocks, and 2 postadaptation blocks). For each block, saccades with a gain which differed from the mean by at least 3 standard deviations were excluded from analysis. The gain change of each saccade of the spTMS blocks following the preadaptation phase (4 adaptation blocks and 1 postadaptation block) was calculated as follows:
Negative (positive) values of gain changes corresponded to a decrease (increase) of saccade amplitude relative to preadaptation.
During the adaptation phase, secondary corrective saccades were identified as saccades directed toward the stepped target position and initiated within 500 ms after the offset of the primary saccade. The gain of these corrective saccades was calculated in the same way as for primary saccades. Their latency was computed as the duration of fixation since the preceding primary saccade. The proportion of corrective saccades (out of the total number of adaptive trials) was also measured.
Statistical analyses were performed with Statistica 9 (Statsoft Inc., Tulsa, OK, United States of America). The latency and the gain of saccades measured during the preadaptation phase were submitted to a 4-way ANOVA with 3 within-subject factors (TMS: on or off; saccade direction: leftward or rightward; and TMS type: pIPS 30 ms, pIPS 60 ms, pIPS 90 ms, or vertex) and 1 between-subject factor (experiment: VS or RS). Gain changes during the adaptation phase, which were computed relative to the gain measured during the preadaptation phase with TMS, were analyzed using a 4-way ANOVA with 3 within-subjects factors (block of trials: 1–4; saccade direction; and TMS type) and 1 between-subject factor (experiment). The same 4-way ANOVA was also used to analyze the latency, the proportion, and the gain of corrective saccades. Gain changes in postadaptation with and without TMS, which were computed relative to the gain measured during the preadaptation (with and without TMS, respectively), were submitted to a 4-way ANOVA with 3 within-subject factors (TMS, saccade direction, and TMS type) and 1 between-subject factor (experiment). ANOVAs yielding significant results were followed by post hoc Tukey's honestly significant difference (HSD) tests. The significance level was set at P < 0.05.
This study was aimed at evaluating the involvement of the parietal cortex in saccadic adaptation. First, the oculomotor region of the pIPS was localized by fMRI. Then, in 2 separate experiments, spTMS was applied in each trial of an adaptation paradigm, to assess the implication of pIPS in the adaptation of VSs (VS experiment) and of RSs (RS experiment).
The fMRI saccade localizer scans revealed a cluster of activated voxels in the pIPS in each subject. A representative individual example is illustrated in Figure 3. Over all subjects, the average Talairach coordinates corresponding to the location of the peak of activity in pIPS were: x = 13 ± 5, y = −63 ± 6, and z = 52 ± 5.
In the following, we present the results of the 2 experiments. The results of the preadaptation phase will be presented first, followed by the results of the adaptation and postadaptation phases.
No spTMS Effect on Preadaptation Saccade Parameters
To assess whether spTMS could modify baseline saccadic parameters, we compared the latency and gain of saccades measured during preadaptation between spTMS and no spTMS blocks. A repeated-measure ANOVA (see Materials and Methods) revealed that, as expected, saccade latency differed depending on the saccade category (effect of experiment: VS vs. RS—F1,22 = 43.7, P = 1.2 × 10−6), but also on the presence of TMS (on vs. off—F1,22 = 34.0, P = 7.2 × 10−6). The significant interaction between these 2 factors (F1,22 = 19.5, P = 2.2 × 10−4) indicated that spTMS had a stronger effect on the latency of VSs (430 ± 14.3 ms without spTMS vs. 378 ± 13.5 ms with spTMS) than on the latency of RSs (196 ± 2.6 ms without spTMS vs. 189 ± 2.8 ms with spTMS). Note that this facilitatory effect of spTMS on saccade initiation was unspecific, as it occurred independently of the stimulation site (no effect of TMS type: F3,66 < 1.43, P > 0.24). Note further that this latency decrease was larger for VSs (characterized by their long latency) than for RSs, which is compatible with a nonspecific alerting effect of the TMS pulses. Alternatively, since the preadaptation with spTMS was always performed after the preadaptation without spTMS, the decrease of saccade latency might also originate from a nonspecific training effect. Concerning the saccadic gain, no effect of spTMS was found, either for RSs or for VSs, irrespective of the stimulation site (pIPS or vertex) and of the timing (30, 60 or90 ms—no effect of TMS type: F3,66 < 2.47, P > 0.07, Fig. 4a,b). Thus, any specific effect of spTMS over pIPS on saccade latency and gain can be ruled out.
Effect of spTMS Over the Right pIPS on Saccadic Adaptation
The gain of RSs and VSs decreased progressively during the adaptation phase in all spTMS sessions (Fig. 5). This saccade shortening is consistent with the direction of the intrasaccadic target step, opposite to the primary saccade. To test whether spTMS over the right pIPS influenced the adaptation, the mean saccadic gain change calculated relative to the mean gain in preadaptation with spTMS was submitted to a 4-way ANOVA with the following factors: Block (1–4), saccade direction (left vs. right), experiment (VS vs. RS), TMS type (pIPS 30, 60, 90 ms and vertex). A significant effect of block on saccadic gain change was observed (F3,66 = 195.9, P < 0.001), consistent with a gradual adaptation of saccades. As expected (Alahyane et al. 2007; Panouilleres, Salemme et al. 2012), no difference in gain changes was observed between the 2 experiments in the vertex session (post hoc Tukey's HSD tests, P > 0.73). Interestingly, gain changes were affected differently depending on spTMS timing and on experiments (experiment × TMS type interaction: F3,66 = 3.0, P = 0.038) and also depending on the adaptation blocks (block × experiment × TMS type interaction: F9,198 = 2.0, P = 0.038). Post hoc tests revealed that, in VS experiment, the application of spTMS over the right pIPS at 60 ms reduced the gain change of VSs, when compared with the vertex session. Specifically, for leftward VSs (Fig. 3a, left panel), a significant decrease in adaptation compared with the vertex session was observed for the last 3 adaptation blocks (post hoc Tukey's HSD tests, P < 0.019). For rightward VSs (Fig. 5a, right panel), a significant decrease of adaptation was observed for the last adaptation block (post hoc Tukey's HSD test, P = 0.004). In contrast, in RS experiment (Fig. 5b), a significant difference of gain changes between the right pIPS and the vertex sessions was observed only for the 90-ms spTMS timing and for the rightward saccades during the last adaptation block (post hoc Tukey's HSD test, P = 0.0004). In this case, the adaptation of RSs was larger for the pIPS session than for the vertex session. For reactive leftward saccades, there was also a trend toward larger gain changes when spTMS is applied over the pIPS at 90 ms than when applied over the vertex (Fig. 5b, left panel). Notwithstanding, this trend failed to reach significance (post hoc Tukey's HSD test, P > 0.07). No difference of gain change was observed between the session where spTMS was applied over pIPS at 30 ms and the vertex session, regardless of experiments (P > 0.26).
In sum, the application of spTMS over the right pIPS 60 ms after saccade detection resulted in less adaptation of VSs in both directions (VS experiment). In contrast, the application of spTMS over the right pIPS at 90 ms enhanced the gain change of rightward RSs at the end of the adaptation phase (RS experiment).
No Effect of spTMS Over the Right pIPS on the Production of Corrective Saccades
Progressive changes in primary saccade gain induced by repeated perturbations, such as intrasaccadic target steps, reflect the operation of an adaptive mechanism. The error signals induced by these perturbations also trigger secondary corrective saccades. Since the application of spTMS over the right pIPS was found to affect saccadic adaptation, we thought it important to assess whether it interfered also with corrective saccades. In principle, spTMS could influence the initiation of corrective saccades, thus affecting their latency. It could also interfere with the detection of the postsaccadic visual feedback, and as a result, affect the rate of the occurrence of corrective saccades. Lastly, spTMS could affect the processing of the size of the visual error and thus modify the gain of corrective saccades. All of these parameters of corrective saccades (latency, rate, and gain; Table 1) were analyzed separately using 4-way ANOVA with the same factors as above (block, saccade direction, experiment, and TMS type). No significant effect of TMS type factor was observed (F3526 < 1.8, P > 0.14), indicating that the production of corrective saccades remained unchanged regardless of the site (pIPS and vertex) and timing (30, 60, or 90 ms) of the spTMS.
|TMS type||pIPS (ms)||Vertex|
|Latency (ms)||282 ± 7.9||284 ± 7.8||295 ± 7.3||270 ± 7.6|
|Rate (%)||27.5 ± 2.0||29.3 ± 2.1||31.9 ± 2.2||32.9 ± 2.1|
|Gain||1.15 ± 0.06||1.38 ± 0.13||1.28 ± 0.09||1.17 ± 0.10|
|TMS type||pIPS (ms)||Vertex|
|Latency (ms)||282 ± 7.9||284 ± 7.8||295 ± 7.3||270 ± 7.6|
|Rate (%)||27.5 ± 2.0||29.3 ± 2.1||31.9 ± 2.2||32.9 ± 2.1|
|Gain||1.15 ± 0.06||1.38 ± 0.13||1.28 ± 0.09||1.17 ± 0.10|
Note: The mean parameter values (±SEMs) are depicted separately for the 4 TMS types. No significant effect of the TMS type factor was observed (4-way ANOVA, F3,526 < 1.8, P > 0.14).
Thus, the effects of spTMS over the right pIPS on saccadic gain changes reported in the previous section are not accompanied by spTMS effects on corrective saccades.
Effect of spTMS Over the Right pIPS on Adaptation After-Effects
During the postadaptation phase, subjects performed saccades toward targets that disappeared at saccade onset. This allowed us to assess the extent to which saccadic gain changes persisted after the adaptation phase, when tested in the absence of any error signal. The gain difference between these saccades and those generated during the preadaptation phase is called “adaptation after-effect,” a typical measure of the retention of adaptation. This adaptation after-effect was computed separately for the postadaptation spTMS block (Fig. 6a,b, left panels) and for the postadaptation no spTMS block (Fig. 6a,b, right panels). After-effects were analyzed using a 4-way ANOVA with 3 within-subject factors (TMS: on vs. off, saccade direction, and TMS type) and 1 between-subject factor (experiment). As revealed by this ANOVA (TMS type × experiment interaction, F3,66 = 3.0, P = 0.038), the adaptation after-effects were differently affected by the TMS type and the experiment factors. For VSs (VS experiment, Fig. 6a, left panel), the after-effects measured in the postadaptation spTMS block were significantly smaller in the session with spTMS over the right pIPS at 60 ms than in the vertex session (post hoc Tukey's HSD test, P < 0.048). This indicates that the inhibitory effect of right pIPS-TMS on the adaptation of VSs, which was reported above, persisted in the postadaptation spTMS block. This inhibitory effect of TMS on adaptation was still observed in the postadaptation no spTMS block (Fig. 6a, right panel). Indeed, the after-effect measured in the session with spTMS over pIPS at 60 ms was significantly reduced relative to the vertex session for rightward saccades (post hoc Tukey's HSD test, P = 0.03), and this reduction approached significance for leftward saccades (post hoc Tukey's HSD test, P = 0.057).
By contrast, for RSs, no significant effect of spTMS on after-effects was observed, whether measured in postadaptation with spTMS (Fig. 6b, left panel, post hoc Tukey's HSD tests, P > 0.91), or without spTMS (Fig. 6b, right panel, post hoc Tukey's HSD test, P > 0.67). Despite a trend for larger after-effects at this timing than for the vertex session, this negative outcome indicates that the facilitatory effect of pIPS-TMS with a 90-ms delay, which was observed during the adaptation phase, was not maintained in the postadaptation phase.
To summarize, for VSs, the adaptation after-effects were significantly smaller in the session with spTMS applied over the pIPS at 60 ms than in the vertex session (VS experiment). In contrast, for RSs, spTMS never affected the after-effect (RS experiment).
Single-pulse TMS applied over the right pIPS, 60 ms after the onset of leftward and rightward VSs, largely reduced the amount of adaptation and the size of the after-effect, when compared with the application of TMS over the vertex. A markedly different pattern of results was found for RSs: Right pIPS-TMS applied 90 ms after the onset of rightward saccades led to larger adaptation at the end of the adaptation phase, with no persistent change in the after-effect. Given the lack of specific modifications of baseline saccade parameters, this pattern of TMS effects clearly results from the modification of different adaptation or error processing mechanisms engaged by the RS and VS protocols.
To the best of our knowledge, the results of this study provide the first direct evidence that parietal regions are causally involved in saccadic plasticity. Consistent with this finding, the results of 2 recent fMRI studies suggest a role of the cerebral cortex in saccadic adaptation. The first study showed an involvement of the supplementary eye fields and of the temporo-insular cortex in the adaptation of RSs (Blurton et al. 2012). The second study found that a network of cerebellar and cerebral areas is implicated in the adaptive processes of RSs and VSs. In particular, the right pIPS was specifically involved in the adaptation of VSs (Gerardin et al. 2012). Here, by using fMRI-guided spTMS over this area, we could perturb selectively the adaptation of VSs, whereas the adaptation of RSs was not impaired, but rather enhanced. Thus, the present study shows that the pIPS is necessary for the optimal adaptation of VSs and may mediate the strengthening of the adaptation of RSs.
An important methodological aspect of the present study is that the intraparietal area targeted by spTMS was functionally localized on a single-subject basis using fMRI. The localizer task was designed to allow precise determination of the oculomotor region of the pIPS. This region is currently thought to correspond to the parietal eye fields, which are known to play a crucial role in the production of saccades (for review see: Muri 2006). This area is also believed to be the human homolog of the lateral intraparietal (LIP) area in the monkey—consistent with the results of a recent fMRI study in humans (Galati et al. 2011). It is commonly accepted that the parietal cortex contributes more to the generation of RSs than VSs. This view is supported by data showing that permanent or reversible lesions of the parietal cortex specifically delay the initiation of RSs (Pierrot-Deseilligny et al. 1991, 2004; Rivaud et al. 1994; Muri and Nyffeler 2008). From this perspective, the finding that the pIPS is necessary for the adaptation of VSs might seem unexpected. However, the contradiction is only apparent, as the notion that the parietal cortex is more involved in RSs than VSs has been challenged and amended by recent studies. First, data from nonhuman primate models indicate that saccade deficits induced by inactivation of the LIP are related specifically to selection and/or attentional mechanisms, and not to saccade execution processes per se (Wardak et al. 2002). Similarly, saccade deficits of neurological patients suffering from lesions of the intraparietal region have been linked to impairments of visuomotor, or saccade programming processes, rather than to saccade initiation processes (Rafal 2006). Secondly, fMRI studies have shown that parietal areas are more activated for VS generation than for RS generation (for review see: McDowell et al. 2008). Event-related fMRI studies suggest that this stronger activation may be associated with saccade preparation processes (Brown et al. 2004, 2008). Altogether, these studies indicate that the parietal cortex is involved both in the initiation processes of RSs and in the visuo-motor programming processes of VSs. Consistent with the latter notion, the presently demonstrated effect of spTMS over pIPS on VSs adds sensorimotor adaptation to the multiple functions of the parietal cortex in saccadic control.
A major finding of the present study is the opposite effects of spTMS over the pIPS for the adaptation of the 2 categories of saccades. Indeed, spTMS at the 60 ms timing strongly impaired the adaptation of VSs, whereas spTMS at the 90 ms timing improved the adaptation of RSs. A recent spTMS study showed that TMS applied over the lateral cerebellum can also have opposite effects on RS adaptation, but this time in relation to the direction of adaptation (Panouilleres, Neggers et al. 2012). Indeed, irrespective of the timing of application, TMS over the cerebellar lobule Crus I potentiated the adaptive lengthening and depressed the adaptive shortening of saccades. The authors of this study proposed that different neuronal populations located in the lobule Crus I may be specifically involved in the 2 adaptive processes. Another study reported the existence of inhibitory and excitatory effects depending on the timing of spTMS and on the behavioral task (Nyffeler et al. 2004). The authors found that spTMS applied over the frontal eye field (FEF) at target onset, decreased the latency of saccades in a gap task (RSs), but not in an overlap task (VSs). In contrast, spTMS applied over the FEF after target disappearance, significantly increased saccade latency in both tasks. The authors proposed that a direct spTMS interference with the FEF was responsible for the inhibitory effect on saccade initiation and that an indirect TMS interference with the superior colliculus was responsible for the facilitatory effect. In the next paragraph, we will consider whether direct/local versus indirect/remote TMS effects could explain the present pattern of findings.
Irrespective of its nature (inhibitory of facilitatory), the mere presence of an effect of spTMS over pIPS on RS adaptation could not be predicted based on previous works. There is a large consensus that the adaptation of RSs involves motor stages of the sensorimotor transformation at the brainstem–cerebellar level (for reviews see: Hopp and Fuchs 2004; Tian et al. 2009; Pelisson et al. 2010; Prsa and Thier 2011). Comparatively, suggestions that saccadic adaptation may involve sensory stages in the cerebral cortex are based on sparse and indirect evidence, and concern VSs (Deubel 1995; Gancarz and Grossberg 1999; Cotti et al. 2007, 2009) or the adaptive lengthening of RSs (Semmlow et al. 1989; Hernandez et al. 2008). To our knowledge, only 2 behavioral studies implied cerebral structures in the adaptive shortening of RSs, one reporting adaptation deficits in 2 patients with thalamic lesion (Gaymard et al. 2001), the other one showing a visual mislocalization in healthy subjects when adaptation was induced in a new protocol (Zimmermann and Lappe 2010), which differs from the classical double-step protocol used here. Therefore, we consider that a note of caution should be taken in interpreting our findings on RS adaptation as evidence that the parietal cortex is the site where adaptive changes develop. Rather, the parietal cortex could be involved in the processing of error signals that lead to saccadic adaptation. However, the fact that the processing of error signals occurs earlier for RSs than for VSs (Panouilleres et al. 2011) is not compatible with the relatively late timing in which spTMS was effective for RSs (90 ms) compared with VSs (60 ms). Further, the effect of pIPS-TMS applied with a delay of 90 ms on RS gain changes was not maintained in the postadaptation phase. Finally, spTMS facilitated, rather than impaired, the adaptation of RSs. Thus, contrary to the case of VSs, TMS might not disrupt an ongoing neural activity specifically related to saccadic adaptation, but could remotely alter neural activity in the cerebellum (e.g. Nyffeler et al. 2004), via existing parietal–cerebellar connections (for reviews, see Glickstein 2003; Ramnani 2012), leading to a facilitation of RS adaptation. Indeed, effects of TMS on remote but connected areas have been largely documented using positron emission topography (Paus et al. 1997), fMRI (Ruff et al. 2006), electroencephalography (Fuggetta et al. 2005; Taylor et al. 2007), or a second “test” TMS pulse (Ugawa et al. 1995; Pascual-Leone and Walsh 2001; Silvanto et al. 2006; Ruff et al. 2008).
The application of spTMS over the right pIPS modified the adaptation of VSs in both directions (left and right). Previous studies have shown that the right parietal cortex is dominant for both visual hemifields in visual, attentional, and/or motor processes (Corbetta et al. 1993; Mangun et al. 1994; van Koningsbruggen et al. 2009). The bilateral effect of spTMS on saccade adaptation could then be due to this dominance of the right parietal cortex. For the VSs, a significant impairment on the development and the retention of adaptation was found when the spTMS was applied 60 ms after saccade onset. In this case, the pulse occurred ∼20 ms after saccade completion (mean saccade duration: 40 ms). This indicates that the adaptation of VSs critically depends on functions taking place in the right pIPS almost immediately after saccade termination.
What could be the nature of these postsaccadic functions? One possibility is that spTMS may have interfered with postsaccadic visual information processing in the pIPS. Visual responses organized according to the retinotopic maps in IPS have indeed been described in several reports (Medendorp et al. 2005; Schluppeck et al. 2005; Swisher et al. 2007). In the light of the strong dissociation between the mechanisms leading to adaptation and those leading to corrective saccades (e.g. Wallman and Fuchs 1998; Bahcall and Kowler 2000; Noto and Robinson 2001; Panouilleres et al. 2011), our study may help understand whether these parietal responses encode “early” visual information, driving both adaptation and corrective saccades, or encode visuo-motor signals specialized for each type of correction. If the effect of spTMS was due to an interference with early visual processes, then it should not depend on the type of saccade (reactive or voluntary), and it should also be observed for corrective saccades. The present results invalidate both of these predictions. Thus, the most parsimonious conclusion is that spTMS did not act at such early sensory level, but more likely interfered with visuo-motor processes that specifically lead to saccadic adaptation. Both error signals processing and enduring changes of oculomotor commands take place during the adaptation phases. Although conceptually distinct, these 2 components of adaptation cannot be dissociated in our study, like in most previous studies (including the fMRI study of Gerardin et al. 2012). The question of whether spTMS on pIPS has altered the “error signal” and/or the “oculomotor commands” component of adaptation can nevertheless be discussed in relation to the available literature. The error signal hypothesis is coherent with the proposed contribution of IPS in error processing for the adaptation of arm reaching movements to a velocity-dependent force field (Della-Maggiore et al. 2004), or to optical prisms (Luaute et al. 2009). Concerning saccadic adaptation, it is believed that error signals result from a comparison between the expected postsaccadic error, which is predicted based on a copy of the motor command (efference copy), and the actual error, which is sampled after saccade completion (Bahcall and Kowler 2000; Wong and Shelhamer 2011; Collins and Wallman 2012). Interestingly, the parietal cortex is known to be involved in remapping the representation of visuo-spatial information through eye movements (Duhamel et al. 1992; Medendorp et al. 2003; Merriam et al. 2003), a mechanism that also requires the use of efference copy. Along this line, TMS studies have further indicated that the right posterior parietal cortex is causally involved in the spatial remapping of visual targets for perception (Chang and Ro 2007; Prime et al. 2008; van Koningsbruggen et al. 2009), or sequential saccades programming (Van Donkelaar and Muri 2002; Morris et al. 2007). The present findings suggest that the role of the parietal cortex in spatial remapping could also serve an error signals monitoring function in the context of VS adaptation. Note that the different effect of TMS on the right pIPS in the adaptation of VSs and RSs is consistent with the dissociation in the processing of error signals between these saccade types (Panouilleres et al. 2011). In contrast, according to the oculomotor commands hypothesis, plastic changes underlying the adaptation of VSs could take place in pIPS, as suggested in previous behavioral studies (Cotti et al. 2007, 2009). Thus, the application of TMS to the right pIPS 60 ms after saccade onset may have interfered with these plastic changes, resulting in a slower development and a smaller after-effect of adaptation. To conclude, the critical involvement of the right pIPS in voluntary saccadic adaptation depends either on error-signal processing, or on plastic changes related to adaptation, and further studies are necessary to tease apart these 2 possibilities.
In conclusion, we demonstrate that the right pIPS is causally involved in saccadic adaptation. Interestingly, we found that spTMS affected adaptation differently depending on the type of saccades. Both adaptation and retention of VSs were impaired with spTMS applied over the pIPS 60 ms after saccade onset. On the contrary, the adaptation of RSs was facilitated when spTMS was applied 90 ms after saccade onset. We propose that these differential effects may respectively emerge from a direct effect of the TMS over the pIPS and from an indirect effect over the cerebellum. Thus, this study provides direct evidence consistent with the hypothesis that the adaptation of VSs and RSs involves different neural mechanisms.
This work was supported by Agence Nationale de la Recherche (ANR-06-NEURO-001) to DP and by Institut National de la Santé et de la Recherche Médicale (INSERM) U864.
We thank all subjects for their participation. We also thank Prof. C. Tilikete for seeing all participants before their inclusion, Dr V. Gaveau for helpful comments and F. Volland for building the fMRI-compatible set-up. The staff of Centre d'Etude et de Recherche Multimodal et Pluridisciplinaire en Imagerie du vivant (CERMEP) is also acknowledged for the fMRI exams. Conflict of Interest: None declared.