Cross-modal involvement of the primary somatosensory cortex in visual working memory: A repetitive TMS study

Abstract

Recent literature suggests that the primary somatosensory cortex (S1), once thought to be a low-level area only modality-specific, is also involved in higher-level, cross-modal, cognitive functions. In particular, electrophysiological studies have highlighted that the cross-modal activation of this area may also extend to visual Working Memory (WM), being part of a mnemonic network specific for the temporary storage and manipulation of visual information concerning bodies and body-related actions. However, the causal recruitment of S1 in the WM network remains speculation.

In the present study, by taking advantage of repetitive Transcranial Magnetic Stimulation (rTMS), we look for causal evidence that S1 is implicated in the retention of visual stimuli that are salient for this cortical area. To this purpose, in a first experiment, high-frequency (10 Hz) rTMS was delivered over S1 of the right hemisphere, and over two control sites, the right lateral occipital cortex (LOC) and the right dorsolateral prefrontal cortex (dlPFC), during the maintenance phase of a high-load delayed match-to-sample task in which body-related visual stimuli (non-symbolic hand gestures) have to be retained. In a second experiment, the specificity of S1 recruitment was deepened by using a version of the delayed match-to-sample task in which visual stimuli depict geometrical shapes (non-body related stimuli).

Results show that rTMS perturbation of S1 activity leads to an enhancement of participants’ performance that is selective for body-related visual stimuli; instead, the stimulation of the right LOC and dlPFC does not affect the temporary storage of body-related visual stimuli. These findings suggest that S1 may be recruited in visual WM when information to store (and recall) is salient for this area, corroborating models which suggest the existence of a dedicated mnemonic system for body-related information in which also somatosensory cortices play a key role, likely thanks to their cross-modal (visuo-tactile) properties.

Introduction

In the last two decades, the role of the primary somatosensory cortex (S1) in high-order cognitive functions has been widely investigated. Studies using different techniques, like functional magnetic resonance imaging (fMRI), electroencephalography (EEG) or non-invasive brain stimulation, showed that S1 is activated not only during the elaboration of afferent somatosensory stimuli and vicarious somatosensation by touch observation, but it is also involved in higher-order cognitive functions such as emotion recognition, motor learning by observation and body representation (e.g., Avenanti et al., 2007, Blakemore et al., 2005, Bolognini et al., 2014, Lametti and Watkins, 2016, McGregor et al., 2016, Pisoni et al., 2018, Pitcher et al., 2008, Rossetti et al., 2012, Sel et al., 2014, Valchev et al., 2017, Zazio et al., 2019). The activation of S1 seems to extend also to memory and attention of touch (e.g., Bauer et al., 2006, Forster et al., 2016, Gallace and Spence, 2009, Ku et al., 2015, Papagno et al., 2016, Papagno et al., 2017, Preuschhof et al., 2006, Taylor-Clarke et al., 2002, Zhao and Ku, 2018).

In particular, considering the domain of memory, Christophel and Haynes (2014), in an fMRI experiment, showed that the retention of complex visuo-spatial pattern stimuli (i.e., moving dots), besides activating lower-level visual areas and the posterior parietal cortex, also recruited S1. This result would suggest that visual features may be encoded through a cross-modal (visual-tactile) mapping, a process by which S1 takes part into the storage of visual information in working memory (WM) (Christophel & Haynes, 2014). Other clues on the potential cross-modal recruitment of somatosensory cortical areas in visual WM are provided by the studies of Ku and colleagues (2015, 2017, 2018): these authors showed that participants’ performance in a tactile-visual delayed match-to-sample task was impaired when single-pulse transcranial magnetic stimulation (TMS) is delivered over S1 300 ms after the onset of the tactile stimulus. Instead, high-order cortical areas, like the posterior parietal cortex and the dorsolateral prefrontal cortex (dlPFC), were recruited during this cross-modal task at different (longer) timepoints of the mnemonic process. However, in such experiments, the tasks were not purely visual; indeed, the authors adopted cross-modal (tactile-to-visual) WM paradigms requiring to exchange information between touch and vision (Ku et al., 2015, Zhao and Ku, 2018, Zhao et al., 2017).

In addition, there are convergent findings that our brain is equipped with a WM sub-system that is specific for the temporary storage and manipulation of visual information concerning bodies and body-related actions (Galvez-Pol et al., 2020, Rumiati and Tessari, 2002, Shen et al., 2014, Smyth et al., 1988, Smyth and Pendleton, 1989, Wood, 2007). The neural underpinnings of this visual ‘body-related’ memory system comprise a complex network of cerebral areas including specialized visual regions, such as the fusiform gyrus and the extrastriate body area (Peelen and Downing, 2005, Urgesi et al., 2007), premotor and tempo-parietal regions related to the action observation network (Cai et al., 2018, Liu et al., 2019, Lu et al., 2016), but also ‘low-level’ somatosensory and motor areas (Arslanova et al., 2019, Galvez-Pol et al., 2018, Galvez-Pol et al., 2018). Overall, this memory network strongly overlaps the ones implicated in the processing of our own body and motor representations (Longo et al., 2010, Molenberghs et al., 2012). The role of S1 in this WM sub-system has been recently highlighted in an EEG study showing that contralateral delayed activity (CDA), a neurophysiological marker of WM (for a review, see: Luria, Balaban, Awh, & Vogel, 2016), increased in left and right somatosensory cortices selectively during the maintenance of visual body-related stimuli (i.e., non-symbolic hand gestures); moreover, the magnitude of this increase was correlated with the number of stimuli to memorize (Galvez-Pol, Calvo-Merino, et al., 2018).

Despite these recent findings, no study has investigated the causal involvement of S1 in visual WM so far. To this aim, we have run a TMS study to obtain the first causal evidence of the actual recruitment of S1 during a visual body-related WM task. Indeed, as repetitive TMS (rTMS) allows to prove the causal relationship between the stimulated cortical area and its involvement in a cognitive function (Bolognini & Ro, 2010), we applied this technique ‘online’ to interfere with S1 functioning while participants performed a WM task, i.e., delayed match-to-sample task (e.g., Galvez-Pol et al., 2018, Vogel and Machizawa, 2004, Vogel et al., 2005) where the visual stimuli to memorize depicted body-related information (hand gestures – Experiment 1). If S1 is recruited for processing bodily stimuli during visual WM tasks, then the perturbation of its activity should affect participants’ performance when rTMS is delivered at a timepoint corresponding to the maintenance phase of the mnemonic process. Following the evidence that higher memory loads lead to a greater CDA in the electrodes overlapping somatosensory cortices (Galvez-Pol, Calvo-Merino, et al., 2018), we decided to use an experimental task with a high number of visual body-related stimuli to memorize (Luck & Vogel, 2013). We chose S1 of the right hemisphere, following Baddeley’s model (2000) proposing that visual WM is mainly represented in the right hemisphere (Baddeley, 2000), as well as evidence of a right hemisphere dominance for touch observation (e.g., Blakemore et al., 2005, Bolognini et al., 2013) and body representation (e.g., Longo et al., 2010, Tsakiris et al., 2007).

Two cortical areas of the right hemisphere were also targeted with TMS: the lateral occipital cortex (LOC) and dlPFC. In our starting hypothesis, the right LOC should act as a control site; this extrastriate area is implicated in the feature-based analysis of visual information (e.g., Amedi et al., 2001, Grill-Spector et al., 2001, Kourtzi and Kanwisher, 2001, Malach et al., 1995) but its role in the processing and the possible storing of body-related visual stimuli still needs to be determined (Gayet et al., 2018, Heuer et al., 2016, Pitcher et al., 2009, Xu, 2018, Xu and Chun, 2006). Conversely, the dlPFC was chosen according to a recent proposal that visual WM storage does not rely on sensory processing areas, but rather on specialized frontal (and parietal) areas not involved in low-level sensory processing per se (such as S1) (Xu, 2017). The right dlPFC is considered a key node of the WM network; indeed, recent studies suggest that its activation is related to the manipulation of stored information rather than to their maintenance (e.g., Barbey, Koenigs, & Grafman, 2013; C. Kim et al., 2015, Postle et al., 2006, Rowe et al., 2002, Veltman et al., 2003). However, previous TMS studies, where both the right and left dlPFC were stimulated, not always were successful in modulating participants’ performance during either visual and non-visual WM tasks (e.g., Bagherzadeh et al., 2016, Beynel et al., 2019, Chung et al., 2018, Hamidi et al., 2009). Hence, by stimulating LOC and dlPFC of the right hemisphere, we aimed at verifying the causal involvement of S1 in WM, assessing its main role in processing visual body-related stimuli (at variance with LOC) and in their storage (at variance with dlPFC); this would offer a better comprehension of its role in visual WM as compared to other areas of the same network. Indeed, by targeting two other functionally relevant areas, our study may provide stronger clues on the functional selectivity of S1 in visual WM tasks, rather than simply controlling for sensory and placebo rTMS confounds, as it would have happened if we had used a sham stimulation or a functionally ‘neutral’ area (such as the vertex) (Duecker & Sack, 2015).

In the second experiment (Experiment 2), we further explored the selective recruitment of the right S1 for body-related visual stimuli by assessing the effect of S1 rTMS on a delayed match-to-sample task depicting geometric shapes; we expected no effect of rTMS over S1 if visual stimuli are not body-related (e.g., Galvez-Pol, Calvo-Merino, et al., 2018).