Longitudinal consensus clustering reveals the functional architecture of the developing rat brain

Abstract

Resting-state fMRI-based mapping of functional networks in developing human populations holds promise for enhancing the diagnosis and treatment of psychiatric disorders, which often emerge during adolescence. Preclinical models offer experimentally tractable lifespans to investigate these trajectories longitudinally, while also enabling direct experimental manipulations that can reveal causal influences on brain and behavioural outcomes. However, unlike in the human neuroimaging field, preclinical neuroimaging often lacks standardisation, making comparison across studies and translation of findings more difficult. In particular, we lack developmentally informed functional brain atlases. Working with a longitudinal sex-balanced resting fMRI dataset collected using a standardised protocol, we aimed to identify robust functional networks in the rat brain and investigate how they change over development (P28, P35, P49, P70 and P91). Employing a consensus clustering approach informed by cluster quality metrics, we mapped functional network development of the rat brain across multiple spatial scales (k=3, k=5 and k=7) from juvenile (pre-puberty) to early adulthood. Force-directed spring embeddings and graph metrics showed global shifts in brain organisation; with increased global system segregation over time, while global network integration (participation coefficient) decreased. At the network-level, the rat brain was characterised by functionally heterogenous network maturation that recapitulated sensorimotor - higher cognitive gradients of development observed in humans. Longitudinal mapping over the adolescent period revealed both linear and non-linear developmental trajectories, and shows that maturation of the rat brain is characterised by network fractionation and delayed cortical specialisation, with the brain transitioning from globally interconnected juvenile networks into functionally segregated adult networks. The functional architecture we describe offers a generalisable structure and nomenclature for the preclinical imaging community, analogous to the Yeo 7-network parcellation in humans. By openly sharing all data, analytic resources, and outputs with the community, this work provides a developmentally informed functional atlas developed under a standardised protocol for anaesthesia and data acquisition and offers resources that will enable preclinical researchers to further advance our understanding of brain development during this critical period.