Interaction between visuospatial attention and semantic memory in large-scale cortical networks

Introduction: Spatial attention and semantic memory are two typical examples of cognitive processes that are considered to be independent of each other and involve separate, although anatomically contiguous, brain networks (Corbetta & Shulman, 2002; Wirth et al., 2011). A series of EEG-TMS studies from our group showed that stimulating the left intraparietal sulcus and the left temporoparietal junction selectively disrupts reaction times in a visuospatial attention and a semantic discrimination task, respectively (Capotosto et al., 2017; 2023; Croce et al., 2018). However, we often need attention and memory together in everyday life, so the question arises of how different large-scale brain networks interact when a task demands the integration of independent cognitive processes. Here, we present a novel task requiring the combination of visuospatial attention and semantic memory retrieval, and results from a task-related fMRI study to offer insight into their interaction. Methods: Twenty-six right-handed young, healthy volunteers underwent a scanning session in a 3T Siemens Prisma MR scanner, including four functional runs (TR=0.8s, TE=0.03s, 2.4 mm isotropic voxels, 583 volumes, 60 slices). A 2x2 factorial block design with visuospatial attention (VSA) and Semantic memory (SEM) as factors gave rise to 4 conditions. The two SEM conditions required to judge whether a visually presented word referred to a living entity, while the non-semantic conditions required to judge whether a visually presented non-word contained the letter "A." In the two VSA conditions, the target word appeared at a peripheral location preceded by a central spatial cue (80% valid), while in the non-attentional conditions, the target word appeared centrally preceded by a non-spatial cue. Thus, in the crucial VSA+SEM combination, participants performed a semantic task on spatially cued target words (Fig. 1). Images were preprocessed with fMRIprep and analyzed with SPM12 with a conventional univariate random-effects analysis (p < 0.05 cluster FDR, cluster-forming threshold p < 0.001 uncorrected). This project was funded by Next Generation EU through the Italian Ministry of University and Research (PRIN2022 20225PJ2JM, CUP B53D2301432 0006). Results: The main effect of VSA (Fig. 2a) yielded significant differential activity in a bilateral parieto-frontal network comprising the superior and intra-parietal regions (SPL, IPS), the intraoccipital sulcus, and the frontal eye fields (FEF). The main effect of SEM (Fig. 2b) yielded activation in the left inferior frontal gyrus (IFG), the bilateral pre-supplementary motor area (pre-SMA), the right anterior insula (aINS), the left ventral temporal lobe (vTL), the left middle temporal sulcus (MTS), and the left orbito-frontal cortex (OFC). Crucially, the right anterior insula activity (Fig. 2c) was significantly higher in the VSA+SEM condition than in both the VSA-only and SEM-only conditions. Conclusions: As expected, the dorsal attentional network was recruited as the main effect of VSA. On the other hand, SEM recruited left language areas and the ventral temporal lobe, which is typically associated with semantic memory. Interestingly, the right anterior insula response was selectively higher in the combined task. This is consistent with the general view of the anterior insula as a hub mediating dynamic interactions between other large-scale brain networks (Menon & Uddin, 2010). We propose that aINS not only "switches" between large-scale cortical networks involved in externally and internally oriented cognition but also contributes to their integrated work when required by the task demands as in the present case.