Dynamic functional connectivity in mouse widefield calcium activity relates to future learning rates
Poster No:
1541
Submission Type:
Abstract Submission
Authors:
Giuseppe de Alteriis1, Matthew Harvey1, Adil Khan1, Dafnis Batalle2
Institutions:
1King's College London, London, Greater London, 2King's College London, London, N/A
First Author:
Co-Author(s):
Introduction:
A key property of brain-wide networks is the dynamic nature of the interactions between their nodes (Hutchison et al., 2013). While functional connectivity has been studied using correlation-based approaches, the dynamical properties of functional connectivity are poorly understood and their relation to behaviour and cognition are largely unknown. Here we extend methods used for the analysis of dynamic functional connectivity (dFC) in fMRI (Allen et al., 2014; Cabral et al., 2017), to widefield calcium imaging of mouse cortex. This allows 1) to seek a more mechanistic understanding of dFC patterns in widefield calcium imaging, given its higher temporal resolution (Ts=36 ms) 2) to apply dFC to investigate the properties of brain dynamics during different behavioural states (stationary vs. locomotion), and their association with learning rates in a visual discrimination task.
Methods:
We performed widefield calcium imaging of mouse dorsal cortex expressing GCaMP7f. We characterised dFC in n=12 mice (6 wildtype mice and 6 mice with a knockout of the Neurexin 1-alpha gene, which is associated with cognitive impairments and autism).
We selected a 1s-long sliding window and obtained time-varying sliding correlation matrices (dFC matrices). We reduced the large dimensionality of pixel-wise dFC matrices by approximating each dFC matrix by its leading eigenvector (i.e., performing Leading Eigenvector Dynamics Analysis -LEiDA, Fig 1A). The LEiDA signal indicates how much a brain area engages in interactions with others. Using k-means clustering of the LEiDA signal we identified 7 connectivity patterns (transient brain states) that all mice exhibit (Fig 1B). For each state we computed Fractional Occupancy (FO, percentage of time in a particular state), Dwelling Time (DT, average duration of a state bout), and metastability (the variability of dFC, i.e. standard deviation of LEiDA for each state), a metric of brain dynamic flexibility. Mice were considered in a locomotion condition if the speed of the wheel was higher than 5 cm/s, otherwise, they were considered in a stationary condition.
We selected a 1s-long sliding window and obtained time-varying sliding correlation matrices (dFC matrices). We reduced the large dimensionality of pixel-wise dFC matrices by approximating each dFC matrix by its leading eigenvector (i.e., performing Leading Eigenvector Dynamics Analysis -LEiDA, Fig 1A). The LEiDA signal indicates how much a brain area engages in interactions with others. Using k-means clustering of the LEiDA signal we identified 7 connectivity patterns (transient brain states) that all mice exhibit (Fig 1B). For each state we computed Fractional Occupancy (FO, percentage of time in a particular state), Dwelling Time (DT, average duration of a state bout), and metastability (the variability of dFC, i.e. standard deviation of LEiDA for each state), a metric of brain dynamic flexibility. Mice were considered in a locomotion condition if the speed of the wheel was higher than 5 cm/s, otherwise, they were considered in a stationary condition.
Results:
The 7 brain states that we identified involve brain areas that are homologous to those found in human fMRI (Fig 1B). We found distinct dFC patterns between stationary and locomotion conditions, with multiple DT and FO state features (but not metastability) correlated with the animals' running speed. Locomotion was characterized by increased FO and DT of the olfactory and somatosensory states and decreased in motor states (Fig 2A).
Furthermore, the landscape of dFC was atypical in the Neurexin 1-alpha KO mice when compared with wildtype mice. We found decreased metastability during stationary conditions in the knockout mice in the global, olfactory and motor states (Fig 2B). Multiple distinct dFC patterns were also identified when comparing wildtype and knockout mice (Fig 2B).
Finally, after the recordings, mice were trained on a visual discrimination task. Metastability in the stationary condition before behavioural training was correlated with the learning rate in the visual discrimination task (Fig 2C).
Furthermore, the landscape of dFC was atypical in the Neurexin 1-alpha KO mice when compared with wildtype mice. We found decreased metastability during stationary conditions in the knockout mice in the global, olfactory and motor states (Fig 2B). Multiple distinct dFC patterns were also identified when comparing wildtype and knockout mice (Fig 2B).
Finally, after the recordings, mice were trained on a visual discrimination task. Metastability in the stationary condition before behavioural training was correlated with the learning rate in the visual discrimination task (Fig 2C).
·Figure 1: pipeline of the method, brain states
·Figure 2: Leida encodes behaviour, Leida predicts learning rate
Conclusions:
The differences in locomotion versus stationary conditions validate the ability of our method to distinguish the underlying patterns of brain-wide interactions supporting different behavioural states. This also suggests a behavioural relevance of the LEiDA patterns, which up to now have been mainly studied in resting-state fMRI. Our results also support the role of metastability as a promising neuromechanistic biomarker of psychiatric disorders (Hancock et al., 2023).
Additionally, our results suggest that metastability is linked to cognitive abilities, being predictive of learning performance, and highlight the central role of brain-wide dynamics in cognition and flexible behaviour.
Additionally, our results suggest that metastability is linked to cognitive abilities, being predictive of learning performance, and highlight the central role of brain-wide dynamics in cognition and flexible behaviour.
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural) 1
Methods Development 2
Novel Imaging Acquisition Methods:
Imaging Methods Other
Keywords:
Autism
Computational Neuroscience
Learning
Machine Learning
1|2Indicates the priority used for review
Provide references using author date format
Cabral, J., Vidaurre, D., Marques, P., Magalhães, R., Silva Moreira, P., Miguel Soares, J., Deco, G., Sousa, N., & Kringelbach, M. L. (2017). Cognitive performance in healthy older adults relates to spontaneous switching between states of functional connectivity during rest. Scientific Reports, 7(1), Articolo 1. https://doi.org/10.1038/s41598-017-05425-7
Hancock, F., Rosas, F. E., McCutcheon, R. A., Cabral, J., Dipasquale, O., & Turkheimer, F. E. (2023). Metastability as a candidate neuromechanistic biomarker of schizophrenia pathology. PLOS ONE, 18(3), e0282707. https://doi.org/10.1371/journal.pone.0282707
Hutchison, R. M., Womelsdorf, T., Allen, E. A., Bandettini, P. A., Calhoun, V. D., Corbetta, M., Della Penna, S., Duyn, J. H., Glover, G. H., Gonzalez-Castillo, J., Handwerker, D. A., Keilholz, S., Kiviniemi, V., Leopold, D. A., de Pasquale, F., Sporns, O., Walter, M., & Chang, C. (2013). Dynamic functional connectivity: Promise, issues, and interpretations. NeuroImage, 80, 360–378. https://doi.org/10.1016/j.neuroimage.2013.05.079