A critical period plasticity framework for the sensorimotor-association axis of cortical neurodevelopment

Abstract

To understand human brain development it is necessary to describe not only the spatiotemporal patterns of neurodevelopment but also the neurobiological mechanisms that underlie them. Human neuroimaging studies have provided evidence for a hierarchical sensorimotor-to-association (S-A) axis of cortical neurodevelopment. Understanding the biological mechanisms that underlie this program of development using traditional neuroimaging approaches has been challenging. Animal models have been used to identify periods of enhanced experience-dependent plasticity - 'critical periods' - that progress along cortical hierarchies and are governed by a conserved set of neurobiological mechanisms that promote and then restrict plasticity. In this review we hypothesize that the S-A axis of cortical development in humans is partly driven by the cascading maturation of critical period plasticity mechanisms. We then describe how recent advances in in vivo neuroimaging approaches provide a promising path toward testing this hypothesis by linking signals derived from non-invasive imaging to critical period mechanisms.

Keywords: adolescence; chemogenetic fMRI; excitation; inhibition; neuroplasticity; pharmacological fMRI.

Figure 1.. A critical period model for development along the S-A axis.

A) Left: The sensorimotor-association (S-A) axis is a large-scale axis of human brain organization that spans continuously from functionally-specific primary sensory and motor regions (sensorimotor pole; yellow), to modality-specific, multimodal, and then integrative transmodal association cortices (association pole; purple). Right: The S-A axis captures spatial variation in circuit properties thought to influence the hierarchical cascading of developmental critical periods, including anatomical hierarchy position, intracortical myelin density, and excitatory (E) and inhibitory (I) circuit properties. B) The S-A axis captures spatial and temporal variability in the development of functional and structural cortical properties. Functional Connectivity: The magnitude and direction of age-related changes in between-network functional connectivity are largely explained by a network’s position in the cortical functional hierarchy, a functional MRI-based homologue of the S-A axis. Each dot represents a single functional network. Positive and negative age effects indicate that a network’s average connectivity with other cortical areas increased with age (functional integration) or decreased with age (functional segregation), respectively. Cortical Myelin: The age of maximal myelin growth progressively increases along the S-A axis, with regions nearer the association pole showing a peak growth rate at later developmental stages. Each dot represents an individual cortical region. Cortical myelin content was proxied by the T1w/T2w ratio. Intrinsic Activity Amplitude: The principal axis of intrinsic activity amplitude development is strongly related to the S-A axis, revealing convergent spatial embedding of developmental and organizational axes. The principal developmental axis was obtained from a principal component analysis conducted on regional trajectories of fMRI fluctuation amplitude development; each dot represents an individual region. This developmental axis captured 87% of variance in regional maturational profiles. C) A critical period plasticity model for the S-A axis of neurodevelopment provides mechanistic insight into hierarchically organized windows of neurodevelopmental plasticity. Left: Windows of critical period plasticity are theorized to occur earliest, in infancy, at the sensorimotor pole of the S-A axis and latest, in adolescence, at the association pole of the axis, with a gradient of developmental timing in between. Right: Critical period mechanisms are hypothesized to underlie windows of developmental plasticity across the human cortex. Critical periods are initiated by opening mechanisms, including the initial strengthening of parvalbumin (PV) interneuron cell signaling and declines in a circuit’s excitation/inhibition (E/I) ratio. Critical periods are terminated by closing mechanisms, including the formation of cortical myelin and perineural nets, which often form around PV cells. Panels A and B (right) are adapted with permission from[5]. Panel B (left) is adapted from [29] under license https://creativecommons.org/licenses/by/4.0/. Panel B (center) is adapted from [31].

Figure 2.. Approaches to investigating changes in the excitation/inhibition ratio using fMRI.

A) Pharmacological functional MRI: Drugs can be used to manipulate the E/I ratio while fMRI data are recorded (drug manipulation) to measure the effect on the fMRI signal. Recent work used a benzodiazepine manipulation to train a machine learning classifier to distinguish patterns of fMRI connectivity generated from drug (i.e., reduced E/I) from placebo scans (classifier biomarker). The trained and validated E/I biomarker was then used to discover evidence of declining E/I in the association cortex during youth (developmental effect). B) Chemogenetic functional MRI: The E/I ratio can be precisely manipulated using Designer Receptors Exclusively Activated by Designer Drugs (DREADD) animal fMRI studies. Recent work used DREADDs to chemogenetically manipulate the E/I ratio in somatosensory cortex (SSp) or prefrontal cortex (PFC) while recording fMRI data (left). Chemogenetically decreasing E/I by increasing inhibition results in increased fMRI connectivity to anatomically connected regions (e.g., retrosplenial cortex, center). Conversely, chemogenetically increasing E/I by either increasing excitation or reducing inhibition results in reduced fMRI connectivity to homotopic cortex (right). C) Biophysical models of functional MRI: Biophysical models place a local circuit model in each region of cortex that has biologically plausible parameters that correspond to excitatory and inhibitory neuron populations. Communication between circuits is scaled by pairwise structural connectivity (left, top). A whole-brain dynamical model produces simulated activity that can be transformed to simulated fMRI BOLD signal. The whole-brain model can be further informed by spatial heterogeneity in the neurobiological properties of each cortical area (left, bottom). Simulated fMRI data can be used to construct a simulated fMRI connectivity matrix that can be compared to empirical fMRI connectivity matrices (middle). Recent work demonstrated that a biophysical model informed by a spatial gradient of E/I gene expression captures true patterns of fMRI connectivity better than chance or competing gradients of neurobiology. Biophysical models could now be used to investigate how E/I model parameters vary across development. Top panel (A) is adapted from [67]. Middle panel (B) is adapted from [72] under http://creativecommons.org/licenses/by/4.0/ and from [73]. Bottom panel (C) is adapted from [80].