Developmental maturation of millimeter-scale functional networks across brain areas

Nathaniel J. Powell, Bettina Hein, Deyue Kong, Jonas Elpelt, Haleigh N Mulholland, Rayne Holland, Matthias Kaschube, Gordon B. Smith

TL;DR

This study reveals that millimeter-scale functional networks in the ferret brain, characterized by modular and correlated activity, are a conserved feature across diverse cortical areas during development. While these networks generally mature by becoming less modular and more complex, visual cortex (V1) shows unique developmental trajectories, suggesting an interplay between a common developmental program and area-specific factors.

What Problem Does This Paper Address?

The brain's cortex is organized into specialized areas that process different types of information, from basic sensory input to complex cognitive functions. A fundamental question in neuroscience is how this functional diversity arises during development. While it's known that early in development, various cortical areas share a common, modular functional organization, it's unclear whether this organization follows a similar developmental path across all areas or if it diverges. This paper investigates whether the modular and correlated functional networks observed in early development persist and mature similarly across primary sensory areas (visual, auditory, somatosensory) and higher-order association areas (prefrontal, posterior parietal) as the brain transitions from internally generated activity to externally driven sensory processing.

Methods

The researchers used in vivo wide-field calcium imaging in ferrets to observe spontaneous neural activity. This technique allows for the visualization of neural activity across large areas of the cortex. They expressed the genetically encoded calcium indicator GCaMP6s in targeted cortical regions via viral injection. Spontaneous activity was recorded at two key developmental stages: around eye-opening (postnatal days P27-32) and a few weeks later (P39-43), a period when visual cortex responses approach mature levels.

To analyze the data, they employed several computational methods:

  • Event detection: Identifying periods of significant spontaneous neural activity.
  • Modularity estimation: Quantifying the degree of organized, patchy activity patterns within these events. This was done by analyzing the spatial autocorrelation of the activity.
  • Wavelength estimation: Determining the typical spatial scale of these modular patterns.
  • Module amplitude calculation: Assessing the selectivity of modular activation.
  • Correlation pattern analysis: Measuring how activity in different brain regions or locations within a region is related over distance. This involved calculating the variance of pixel-wise correlations at a specific distance from a "seed point."
  • Dimensionality estimation: Quantifying the complexity or richness of the neural activity patterns by estimating the effective number of dimensions needed to describe them.

They also used shuffled control ensembles to statistically assess the significance of their findings, ensuring that observed patterns were not due to random chance.

Datasets

The study utilized data collected from ferret kits. The specific developmental ages examined were postnatal days P27-32 and P39-43. Data from an earlier developmental stage (P21-24) from a previous study by Powell et al. (2024) was re-analyzed and included for comparison, providing a developmental timeline from before eye-opening to well after. No new benchmarks were introduced; rather, the study builds upon existing methodologies and datasets.

Key Insights and Ideas

  • Ubiquitous Modular Networks: Millimeter-scale modular functional networks are not confined to primary sensory areas but are a common feature across diverse cortical regions, including association areas, even after eye-opening and the onset of externally driven sensory input.
  • Shared Developmental Trajectory: While area-specific differences exist, cortical areas generally follow a similar developmental path, characterized by a decrease in modularity and correlation strength, and an increase in the dimensionality of activity.
  • V1's Unique Path: The visual cortex (V1) stands out with a distinct developmental trajectory. Its correlation strength remains high with age, and its activity remains relatively low-dimensional compared to other areas. This suggests that V1's specialized processing might be supported by unique network maturation processes.
  • Interplay of Common and Specific Development: The findings highlight a dynamic interplay between a conserved developmental program that maintains modular networks and area-specific factors that lead to functional specialization.
  • Functional Significance of Modularity: The persistence of modular organization suggests it plays a crucial role in functional representations across various brain areas, potentially by facilitating efficient local processing and communication.

Key Results

  • Modularity Decline: Modularity of spontaneous events significantly declined with age across all cortical areas examined (P21-24 to P39-43). However, events remained significantly modular compared to controls at all ages.
  • Wavelength Changes: The spatial wavelength of active modules showed a slight but significant decrease with age, indicating a general shift towards higher spatial frequencies. There were also significant differences in wavelength between specific cortical areas (e.g., V1 vs. PFC).
  • Module Amplitude Reduction: The amplitude of active modules, reflecting selectivity, significantly declined with age across all areas, with V1 showing distinct differences compared to other areas.
  • Correlation Strength Decrease: Long-range correlations (estimated at 1.8-2.2 mm) showed a significant decline with age in all areas except V1, which maintained high correlation strength.
  • Dimensionality Increase: The dimensionality of spontaneous activity significantly increased with age in most cortical areas (PFC, PPC, A1, S1), indicating a richer and more complex activity space. V1 showed minimal change in dimensionality with age.
  • Area-Specific Interactions: Statistical analyses revealed significant interactions between age and area for module amplitude, correlation strength, and dimensionality, underscoring the area-specific developmental trajectories.

Limitations and Future Work

The authors acknowledge that their experiments were conducted under light isoflurane anesthesia. While they cite previous work suggesting this doesn't alter the spatial structure of activity in V1, they note that direct comparisons with awake states in other cortical areas were not performed. Future work could directly test this by comparing awake and anesthetized states in PFC, PPC, A1, and S1.

They also suggest that future research could investigate the temporal structure of activity and explore how it relates to spatial patterns, particularly in the context of Up/Down states observed during sleep and anesthesia. The use of head-mounted miniature microscopes is proposed for future studies to examine spatial and temporal activity across ages and states (awake, sleep) in more detail.

Furthermore, the paper raises questions about the circuit mechanisms underlying modular organization and its developmental changes. Future studies could directly measure or manipulate lateral interactions (excitation/inhibition) to understand their role in shaping modular networks and their maturation. Optogenetic stimulation experiments are also suggested to disentangle the contributions of input versus recurrent interactions. Finally, the potential role of long-range horizontal connections in stabilizing large-scale correlated networks and contributing to area-specific differences warrants further investigation.

Summary updated

Abstract

Processing sensory information, generating perceptions, and shaping behavior engages neural networks in brain areas with highly varied representations, ranging from unimodal sensory cortices to higher-order association areas. In early development, these areas share a common distributed and modular functional organization, but it is not known whether this undergoes a common developmental trajectory, or whether such organization persists only in some brain areas. Here, we examine the development of network organization across diverse cortical regions in ferrets using in vivo wide field calcium imaging of spontaneous activity. In both primary sensory (visual, auditory, and somatosensory) and higher order association (prefrontal and posterior parietal) areas, spontaneous activity remained significantly modular with pronounced millimeter-scale correlations over a 3-wk period spanning eye opening and the transition to externally-driven sensory activity. Over this period, cortical areas exhibited a roughly similar set of developmental changes, along with area-specific differences. Modularity and long-range correlation strength generally decreased with age, along with increases in the dimensionality of activity, although these effects were not uniform across all brain areas. These results indicate an interplay of area-specific factors with a conserved developmental program that maintains modular functional networks, suggesting modular organization may be involved in functional representations in diverse brain areas.
BibTeX
@inproceedings{powell2025developmental,
  author = {Nathaniel J. Powell and Bettina Hein and Deyue Kong and Jonas Elpelt and Haleigh N Mulholland and Rayne Holland and Matthias Kaschube and Gordon B. Smith},
  title = {{Developmental maturation of millimeter-scale functional networks across brain areas}},
  booktitle = {Cerebral},
  year = {2025},
  doi = {10.1093/cercor/bhaf007},
}
Figure