Role of the supplementary motor area during reproduction of supra-second time intervals: An intracerebral EEG study

Abstract

The supplementary motor area (SMA) has been shown to be involved in interval timing but its precise role remains a matter of debate. The present study was aimed at examining, by means of intracerebral EEG recordings, the time course of the activity in this structure, as well as in other functionally connected cortical (frontal, cingulate, insular and temporal) areas, during a visual time reproduction task. Four patients undergoing stereo-electroencephalography (SEEG) for presurgical investigation of refractory focal epilepsy were enrolled. They were selected on the presence of depth electrodes implanted within the SMA. They were instructed to encode, keep in memory and then reproduce the duration (3, 5 and 7 s) of emotionally-neutral or negative pictures. Emotional stimuli were used with the aim of examining neural correlates of temporal distortions induced by emotion. Event-related potentials (ERPs) were analyzed during three periods: During and at the extinction of the target interval (TI) and at the beginning of the reproduction interval (RI). Electrophysiological data revealed an ERP time-locked to TI-offset whose amplitude varied monotonically with TI-duration. This effect was observed in three out of the four patients, especially within the SMA and the insula. It also involved the middle and anterior cingulate cortex, the superior, middle and inferior frontal gyri and the paracentral lobule. These effects were modulated by the prior TI-duration and predicted variations in temporal reproduction accuracy. In contrast, modulations of ERPs with TI-duration, emotion or temporal performance during the target or the reproduction interval were modest and less consistent across patients. These results demonstrate that, during reproduction of supra-second time intervals, the SMA, in concert with a fronto-insular network, is involved at the end of the target interval, and suggest a role in the duration categorization and decision making operations or alternatively in the preparedness of the timing of the future movement that will be executed during the reproduction phase.

Introduction

The ability to perceive the passage of time is essential for individuals to adapt to a changing environment. Numerous studies aiming at elucidating the neurobiological basis of time perception in the milliseconds to seconds range (i.e., interval timing) presuppose the existence of a dedicated clock in the brain and are based on the “pacemaker-accumulator” model (Gibbon, 1977). According to this information-processing model, pulses are emitted by an oscillatory pacemaker with a certain frequency and gathered into an “accumulator” via a “switch” (controlled by attention) that closes and opens at the onset and the offset of the to-be-timed event, respectively. The content of the accumulator indexing the perceived event duration is stored transiently in a working memory system and/or permanently in a reference memory system. Finally, a decision mechanism compares the current duration values with those in reference memory to perform an adequate behavior.

Neuroimaging and neuropsychological studies have suggested that interval timing mechanisms are distributed in the brain and could rely on a large network including the cerebellum, the basal ganglia as well as the frontal, parietal and temporal cortices. It has been proposed that the engagement of each specific brain region could depend on the duration range (seconds or sub-seconds), the sensory modality (visual, auditory, tactile or multimodal) or the nature (perceptual or sensori-motor) of the task at hand (Ivry and Spencer, 2004; Lewis and Miall, 2003; Merchant et al., 2013). Interestingly, a review of functional magnetic resonance imaging (fMRI) studies by Wiener et al. (2010) using a conjunction analysis pointed the supplementary motor area (SMA) and the right inferior frontal gyrus (IFG) as the only structures consistently activated irrespective of a specific timing context.

Great attention has been given to the SMA which, in connection with the striatum, might provide the neural substrate of the accumulation process commonly postulated in models of time processing (Pouthas et al., 2005). This putative role is supported by a high number of fMRI studies showing that SMA activity during timing perceptive task increases with physical duration (Pouthas et al., 2005; Wencil et al., 2010), perceived duration (Tipples et al., 2013), or the amount of attention allocated to the duration of stimulus (Coull et al., 2004; Henry et al., 2015; Herrmann et al., 2014). Furthermore, Coull et al. (2015) suggested that SMA selectively codes the accumulation of temporal magnitude by showing that its activity increased incrementally as a function of both physical and perceived duration, but not as function of distance. This role of SMA in temporal accumulation is also in line with single-cell recordings in monkeys showing increasing SMA activity as a function of duration during temporal prediction (Akkal et al., 2004), or reproduction of single (Mita et al., 2009) or multiple intervals (Merchant et al., 2013). Similarly, scalp EEG studies in humans have shown that the amplitude of the contingent negative variation (CNV), a slow negative potential usually recorded over medial frontal electrodes and supposed to originate from the SMA (Gómez et al., 2003; Kononowicz et al., 2015; Mento et al., 2013), varies in line with the perceived duration (Macar and Vidal, 2002). When physically identical target durations are compared to a previously memorized standard, a larger negativity is associated to a longer duration judgment (Bendixen et al., 2005; Macar et al., 1999). This pattern of results suggests that the level of activation for the subjectively longer durations is relatively high as compared to the pre-defined level stored in reference memory and provides a strong argument that the CNV indexes the accumulation process (Macar et al., 1999).

However, other empirical findings raise questions on the exact process that is reflected by the CNV recorded during the to-be-timed interval (Kononowicz et al., 2017; van Rijn et al., 2011; for review). Using a time production procedure, the seminal work by Macar et al. (1999) has shown that larger CNV amplitudes are observed when the target duration is over-produced, which seems to be at odds with the construct of temporal accumulation. Indeed, if a higher level of negativity reflects an accelerated accumulation of pulses, one can predict that the threshold value corresponding to the target duration is reached earlier (van Rijn et al., 2011). High-amplitude CNV trials are thus supposed to be associated with a response that occurred before the target duration and not afterwards. Interestingly, in reproduction tasks for which a number of different target intervals were used, the amplitude of the CNV recorded during the reproduction phase has been found to be negatively correlated with the duration of the target interval, a finding which appears to be inconsistent with an accumulation account (Elbert et al., 1991; Kononowicz et al., 2015). In addition, larger amplitude of both the early (initial CNV: iCNV) and late (late CNV: lCNV) components of the CNV during the reproduction phase has also been associated with underproduction of intervals and this effect was more pronounced for the shorter reproduction interval (Kononowicz et al., 2015). From these results, Kononowicz et al. (2015) proposed that the negative relationship between temporal estimates and iCNV amplitude is likely to reflect preparatory and anticipatory processes initiated right after the beginning of the reproduced interval and suggested that the preparation level is larger for the shorter target interval (Kononowicz et al., 2015).

The challenge in establishing the factors on which the link between CNV and time processing depends is emphasized by recent electrophysiogical studies that have failed to find any performance-related effects on the CNV amplitude in similar motor timing tasks (Kononowicz and van Rijn, 2011; Tamm et al., 2014; Kononowicz et al., 2017; for review). Moreover, some studies demonstrated that the subjective timing of supra-second time intervals was better reflected by EEG activity occurring at the beginning of the interval (Kononowicz et al., 2015; Kononowicz and van Rijn, 2015) or by post-interval ERPs that follow the CNV resolution (i.e. amplitude of N1-P2 complex; Kononowicz and van Rijn, 2014a). In the context of a time production task, in which a motor response is generated, higher beta power measured at the onset of the to-be-timed interval appears to be associated with overproduced durations (Kononowicz and van Rijn, 2015). Among the post-duration ERPs, the N1-P2 complex varies according to the magnitude of difference between target and comparison intervals (Kononowicz and van Rijn, 2014a) and the late positive component of timing (LPCt) is modulated with the difficulty of temporal decision (i.e. when spacing between comparison intervals is reduced; Paul et al., 2011) or with physical duration (Gontier et al., 2009; Lindbergh and Kieffaber, 2013). Collectively, these electrophysiological studies indicated that at least some timing mechanisms continue after the offset of the to-be-timed stimulus and questioned the role of climbing neural activity in the subjective timing, as indexed by the CNV (Kononowicz et al., 2017; for review).

Finally, fMRI studies found an activity resembling to an accumulation process in the middle (Bueti and Macaluso, 2011) and posterior (Wittmann et al., 2010) parts of the insula but not in the SMA during the encoding phase of temporal reproduction tasks. This activity would be followed by the activation of a frontal network encompassing the pre-SMA and the anterior insular cortex/inferior frontal gyrus (AIC/IFG) during the reproduction phase, suggesting that SMA is not involved in the accumulation process. Preparatory processes to an upcoming stimulus (Kononowicz and van Rijn, 2014a), motor inhibition (Coull et al., 2016; Kononowicz and van Rijn, 2015), keeping a representation of the target interval during the reproduction phase (Wittmann et al., 2010), or general cognitive effort irrespective of whether the task requires to process temporal information or not (Livesey et al., 2007) have been considered as alternative proposals to describe the role of the SMA in interval timing.

Stereo-electroencephalographic (SEEG) recordings offer a unique opportunity to explore the time course of activity in targeted cortical brain regions thus making possible precise measures of where and when activity occurs in the brain. To date, there has been no attempt to examine the neural correlates of time perception by means of this technique. Given the putative role of SMA in the build-up and in the keeping of a representation of a time interval, the aim of the present study was to investigate the temporal dynamics of intracerebral EEG activity in this brain region during the encoding, the retention and the reproduction of time intervals. Considering the hypothesis of a “temporal hub” with a distributed network of brain structures (Merchant et al., 2013, for review), activity was also examined in the other implanted brain regions. The task was similar to that used by Wittmann et al. (2010) or by Kononowicz et al. (2015). Participants were presented a target interval (TI: 3, 5 and 7 s) they had to encode, to keep in memory, and then to reproduce. In order to bias the subjective perception of time, emotionally-negative pictures known to induce distortion of temporal judgment (Droit-Volet and Meck, 2007, for review) and neutral pictures were presented during the encoding phase. This task was presented to four subjects with drug-refractory epilepsy who had intracerebral electrodes implanted in the SMA. Because a putative neural accumulator located in the SMA is supposed to be involved during the coding and retrieval of a target interval (Coull et al., 2008), we predicted that the amplitude of the CNV-like activity recorded in the SMA during the encoding and reproduction phases correlates with the subjective duration (i.e. with trial-to-trial fluctuations of temporal performance), and also with temporal distortions induced by emotion. Within this conceptual framework, we expected that the amplitude of the CNV-like activity evoked by the target interval (TI) and the reproduction interval (RI) should increase and decrease, respectively, when participants over-produced temporal intervals. Alternatively, if SMA activity locked to the onset of the RI reflects motor preparation and expectation, higher amplitude of the CNV-like activity should be associated with both the short TI-duration (i.e. 3s interval; Elbert et al., 1991; Gibbons and Rammsayer, 2004; Kononowicz et al., 2015) and under-produced intervals (Kononowicz et al., 2015). Under the assumption that temporal processing continues after TI-offset, we predicted that the brain responses recorded during the retention phase would be modulated according to the length of TI-duration and temporal performance.