Overlapping connectivity gradients in the anterior temporal lobe underlie semantic cognition

Abstract

Convergent evidence from neuroimaging, computational, and clinical research has shown that the anterior temporal lobe (ATL) is critically involved in two key aspects of semantic cognition: the representation of semantic knowledge, and the executive regulation of this knowledge. Both are necessary for integrating features to understand concepts, and to integrate concepts to understand discourse. Here, we tested the hypothesis that these differential aspects of integration map onto different patterns of ATL connectivity. Specifically, we hypothesized that there are two overlapping modes of functional connectivity of the ATL that each predict distinct aspects of semantic cognition on an individual level. We used a novel analytical approach (connectopic mapping) to identify the first two dominant modes connection topographies (i.e. maps of spatially varying connectivity) in the ATL in 766 participants (Human Connectome Project), and summarized these into 16 parameters that reflect inter-individual differences in their functional organization. If these connection topographies reflect the ATL's functional multiplicity, then we would expect to find a dissociation where one mode (but not the other) correlates with cross-modal matching of verbal and visual information (picture vocabulary naming), and the other (but not the former) correlates with how quickly and accurately relevant semantic information is retrieved (story comprehension). Our analysis revealed a gradient of spatially varying connectivity along the inferior-superior axis, and secondly, an anterior to posterior gradient. Multiple regression analyses revealed a double dissociation such that individual differences in the inferior-superior gradient are predictive of differences in story comprehension, whereas the anterior-posterior gradient maps onto differences in picture vocabulary naming, but not vice versa. These findings indicate that overlapping gradients of functional connectivity in the ATL are related to differential behaviors, which is important for understanding how its functional organization underlies its multiple functions.

Correspondence of functional connectivity gradients across human isocortex, cerebellum, and hippocampus

“…Accumulating evidence suggests that the organization of the isocortex 8,22,23,25,[61][62][63][64][65] , the cerebellum 38,39 , and the hippocampus 7,40-42,66 can be described with multiple gradients of functional features that map onto structural features in the human brain. In the present study, analyses of two large datasets (N > 2000) replicated two gradients identified by published research describing cerebellar and hippocampal functional connectivity to the isocortex 38,42 , and additionally revealed the topography of the third gradient in each structure that, to our knowledge, has not previously reported in the literature.…”

Functional connectivity gradients as a common neural architecture for predictive processing in the human brain

“…Accumulating evidence reveals that the organization of the cerebral cortex 30,[32][33][34][35][36][64][65][66] , the cerebellum 38,59 , and the hippocampus 39-41,67,68 can be described with multiple gradients of structural and functional features in humans. In the cerebral cortex, converging evidence from network-, circuit-, and cytoarchitectural-levels of analysis suggest that such gradients can be interpreted as the organizing principles underlying predictive processing 36,55,56,69,70 , guiding the flow of prediction signals, prediction error signals, and precision signals.…”

A dynamic gradient architecture generates brain activity states

“…Recent work indicates that these dimensions reflect gradients of continuous spatial variation in regional functional connectivity ( Haak et al, 2018 ; Margulies et al, 2016 ; Zhang et al, 2019 ) and cytoarchitecture ( Burt et al, 2018 ; Paquola et al, 2019 ; Wang, 2020 ). These gradients capture both the major sensorimotor-to-association axis of cortical organization ( Sydnor et al, 2021 ) and finer-grained delineations between subregions supporting distinct cognitive processes like in the anterior temporal lobe ( Faber et al, 2020 ). While gradients appear to reflect regional heterogeneity in circuit properties that sculpt brain activity dynamics ( Kong et al, 2021 ), it remains unclear how spatially fixed functional gradients can dynamically coordinate to generate distinct brain activity states.…”