Differences in layered activation of the insula during interoceptive vs. exteroceptive attention

2343 

Abstract Submission 

Matthias Müller-Schrader¹, Frederike Petzschner², Jakob Heinzle¹, Lars Kasper³, Katharina Wellstein¹, Johanna Bayer⁴, Maria Engel⁵, Lilian Weber⁶, Klaas Prüssmann⁵, Klaas Stephan¹

¹Translational Neuromodeling Unit (TNU), University of Zurich & ETH Zurich, Zurich, Zurich, ²Brown University, Providence, RI, ³University of Toronto, Toronto, Ontario, ⁴Donders Institute for Brain, Cognition and Behaviour, Nijmegen, Gelderland, ⁵Institute for Biomedical Engineering, ETH Zürich & University of Zurich, Zurich, Zurich, ⁶Department of Psychiatry, University of Oxford, Oxford, Oxfordshire

Matthias Müller-Schrader  
Translational Neuromodeling Unit (TNU), University of Zurich & ETH Zurich
Zurich, Zurich

Frederike Petzschner  
Brown University
Providence, RI

Jakob Heinzle  
Translational Neuromodeling Unit (TNU), University of Zurich & ETH Zurich
Zurich, Zurich

Lars Kasper  
University of Toronto
Toronto, Ontario

Katharina Wellstein  
Translational Neuromodeling Unit (TNU), University of Zurich & ETH Zurich
Zurich, Zurich

Johanna Bayer  
Donders Institute for Brain, Cognition and Behaviour
Nijmegen, Gelderland

Maria Engel  
Institute for Biomedical Engineering, ETH Zürich & University of Zurich
Zurich, Zurich

Lilian Weber  
Department of Psychiatry, University of Oxford
Oxford, Oxfordshire

Klaas Prüssmann  
Institute for Biomedical Engineering, ETH Zürich & University of Zurich
Zurich, Zurich

Klaas Stephan  
Translational Neuromodeling Unit (TNU), University of Zurich & ETH Zurich
Zurich, Zurich

Attention modulates the processing of exteroceptive stimuli, such as visual inputs (Moore et al. 2017). Potential mechanisms include top-down cortical projections to lower sensory regions and neuromodulatory transmitter release via projections from brainstem nuclei, both of which occur in layer-specific manners. By contrast, the impact of attention on the perception of bodily states (interoception) remains less explored. Here, we present a preregistered analysis of layered 7T fMRI data, focusing on the insula, a central area for interoception. We investigated whether insular subregions show different BOLD activity between attention to one's own heartbeat and attention to auditory sounds, and whether this difference varied between upper vs. lower cortical layers.

40 participants performed a heartbeat attention task (Petzschner et al. 2019). In each block (duration 20s, 16 per run), participants were instructed to either focus on their heartbeat (interoceptive condition) or on continuously present auditory white noise (exteroceptive condition). The study was approved by the cantonal ethics committee of Zurich (2016-01297).
FMRI images were acquired on a Philips Achieva 7T MR system with quadrature transmit and 32-channel head receive coils, using a high-resolution single-shot spiral-out trajectory with 0.9 mm in-plane resolution (R=4, TE=15ms, 50ms readout duration), 36 slices (0.9mm/0.1mm slice-gap/TR=3.128s) and fat suppression module (SPIR). Images were reconstructed by inversion of an advanced signal model, taking sensitivity encoding and static and dynamic field inhomogeneities into account (Kasper et al. 2022).
Functional images were slice-timed, realigned and co-registered to an anatomical reference image. The first level GLM included two main regressors (attention to heart/sound), as well as regressors accounting for physiological noise, head motion, and motion censoring.
Layered masks (upper/lower) were created for each individual by combining anatomical masks of insular regions (Fan et al. 2016) with 25%/75%-equivolume-cortical depth masks (Waehnert et al. 2014) obtained by segmentation of T1-weighted images [0.8mm isotropic resolution] with Freesurfer) and a functional mask for the contrast attention to heart vs. sound (from all other subjects, p_cluster<0.05 with CDT p<0.001), see Figure 1.
Data processing and analyses followed a preregistered analysis plan (Müller-Schrader et al. 2023). A 2x2 repeated measures ANOVA was performed for each insular region (factors layer [upper/lower] and condition [attention to heart/sound]). We tested the layer-by-condition interaction and corrected for multiple comparisons across insular regions.

Complete fMRI data were available for 25 out of 40 participants. 19 datasets could be reconstructed with sufficient quality for layered analyses. Within the functional mask, three insular regions had notable activations (>10 activated voxels). Attention to heartbeats activated the regions more strongly than attention to sound, and this difference was greater in upper vs. lower layers (Figure 2). This layer-by-condition interaction was significant in both left (p=0.00036, F = 19.55) and right (p=0.002, F=13.233) dorsal dysgranular insula, and exhibited a continuous trend across cortical depth (Figure 2).

Previous layered fMRI have examined exteroceptive (visual) attention effects, with diverse results (Mourik et al, 2023, Liu et al, 2021, Muckli et al, 2015). This study is the first to report layer-specific attention effects for interoceptive stimuli. Our preregistered analysis found that activation of the insula by attending to heartbeats vs. sounds was modulated by cortical depth, revealing a layer-by-condition interaction, with larger differences in the upper layer. Notably, it is possible that these layer differences may have been influenced by venous blood inflow effects (from lower to upper layers); we will investigate this in further analyses.

Activation (eg. BOLD task-fMRI)

BOLD fMRI ¹

Attention: Auditory/Tactile/Motor

Perception: Multisensory and Crossmodal

Perception and Attention Other ²

FUNCTIONAL MRI

HIGH FIELD MR

Perception

Pre-registration

Other - Interoception; laminar fMRI; attention; Insula