The cortical microcircuitry of contextual processing in mice, monkeys, and humans
Poster No:
2559
Submission Type:
Abstract Submission
Authors:
Lars Muckli1, Christiaan Levelt2, Tyler Morgan3, Paolo Papale4, Pieter Roelfsema4, Koen Seignette2, Matthew Self2, Michele Svanera5, Lucy Petro6
Institutions:
1University of Glasgow, Glasgow, Scotland, 2Netherlands Institute for Neuroscience, Amsterdam, Netherlands, 3National Institute of Mental Health, Bethesda, MD, 4Netherlands Institute for Neuroscience (KNAW), Amsterdam, Netherlands, 5University of Glasgow, Glasgow, United Kingdom, 6School of Psycholog and Neuroscience, University of Glasgow, Glasgow, Scotland
First Author:
Co-Author(s):
Introduction:
The architecture of the cerebral cortex supports the integration of sensory inputs with current demands, expectations and prior knowledge, in distinct processing streams flowing in bottom up and top down directions, embedded in laminar microcircuits (e.g. Larkum, 2013). However, the content of internal representations and the functional roles of top-down processing remain often theoretical or lack mechanistic detail. We present across-species studies using a visual occlusion paradigm designed to isolate the influence of top-down information. This approach masks a portion of an image (the "occluded" region). This region of the visual field in the primary visual cortex of humans, monkeys and mice does not receive bottom-up receptive field stimulation, and neuronal recordings therefore measure functional and structural principles of top-down processing.
Methods:
(Humans) We recorded 3T (n=23) and laminar 7T (n = 8) fMRI data in human V1 while participants viewed partially occluded natural scene images. MRI acquisition was as follows: (3T: 18 slices; voxel size: 3 mm, isotropic; 0.3 mm inter-slice gap; TR = 1000 ms; TE = 30 ms; matrix size = 70×64; FOV = 210×192 mm, and 7T: 38 slices; isotropic 0.8 mm3; TE = 17 ms; TR= 2000 ms; FOV=128 x 128 mm2, matrix: 160 x 160).
(Mice) We recorded chronic two-photon and wide-field calcium imaging in awake mice in L2/3 and L5. We mapped single cell receptive fields before presenting partially occluded natural scenes that were either novel or had been presented during a training phase. Animals trained to detect fullscreen natural scenes were subsequently presented with fullscreen or occluded natural scenes.
(Monkeys) We recorded spiking activity in V1 neurons in two monkeys that were presented with an occluder over the receptive field. Partially occluded scenes were presented for 700ms. (Model) We built a generative encoder-decoder architecture to model human fMRI responses measuring cortical feedback processing (Svanera et al., 2021).
(Mice) We recorded chronic two-photon and wide-field calcium imaging in awake mice in L2/3 and L5. We mapped single cell receptive fields before presenting partially occluded natural scenes that were either novel or had been presented during a training phase. Animals trained to detect fullscreen natural scenes were subsequently presented with fullscreen or occluded natural scenes.
(Monkeys) We recorded spiking activity in V1 neurons in two monkeys that were presented with an occluder over the receptive field. Partially occluded scenes were presented for 700ms. (Model) We built a generative encoder-decoder architecture to model human fMRI responses measuring cortical feedback processing (Svanera et al., 2021).
Results:
In human, monkey and mouse V1, scene identity can be decoded using multivariate pattern analyses of data taken from a region processing only the occluded image patch. Laminar imaging localises this top-down contextual information to superficial layers of V1 (Muckli et al., 2015, Figure 1A). This contextual information in the non-stimulated retinotopic region of V1 correlates with orientation information found in internal model predictions of scenes, sampled in humans as behavioural line drawings of missing scene features (Morgan et al., 2019, Figure 1A). The dynamics of this contextual feedback information, studied using electrophysiology in monkeys, shows that V1 neurons in the occluded region (that don't receive feed-forward information) process scene-specific contextual information less than 20 ms later than neurons whose receptive fields are presented with the scene stimulus. These contextual representations were correlated between humans and monkeys, suggesting cross-species similarity in the content of perceptual internal models supplied by higher visual areas to V1 (Papale at al., 2023, Figure 1B). In mice, distinct populations of L2/3 neurons code for either sensory or feedback information, and contextual responses to occluded scenes are stronger in trained animals, suggesting that these responses can be explained in a predictive processing framework (Figure 1C, Seignette et al., 2023).
·Multispecies contextual feedback in non-stimulated V1
Conclusions:
Assessing cognitive functions and neuronal substrates of top-down microcircuits can be achieved with the occlusion paradigm. We showed evidence in humans that top-down signals include predictive world models that are updated by information from our senses. The occlusion paradigm can also be implemented in a multispecies approach, with surprisingly consistent results in human and nonhuman primates and mice, opening the door for studying the involvement of finer-grained (cellular) neural mechanisms in feedback processing (Muckli et al., 2023).
Modeling and Analysis Methods:
Methods Development
Multivariate Approaches 2
Neuroinformatics and Data Sharing:
Databasing and Data Sharing
Novel Imaging Acquisition Methods:
Imaging Methods Other
Perception, Attention and Motor Behavior:
Perception: Visual 1
Keywords:
ANIMAL STUDIES
Cortical Layers
Cross-Species Homologues
Meta- Analysis
Multivariate
Optical Imaging Systems (OIS)
Vision
1|2Indicates the priority used for review
Provide references using author date format
Muckli, L., et al. (2015). Contextual feedback to superficial layers of V1. Current Biology, 25(20), 2690-2695.
Muckli,L et al. (2023). The cortical microcircuitry of predictions and context – a multi-scale perspective (v_0.1). Zenodo. https://doi.org/10.5281/zenodo.8380094
Papale, P., et al. (2023). The representation of occluded image regions in area V1 of monkeys and humans. Current Biology.
Seignette, K. et al. (2023). Layer-specific plasticity of feedforward and contextual neuronal population in mouse visual cortex. Society for Neuroscience 25.08.