Structure and activity in the same piece of brain
The MICrONS consortium released a landmark functional-connectomics dataset in Nature on 9 April 2025. It combines calcium-imaging recordings from roughly 75,000 neurons with an electron-microscope reconstruction containing more than 200,000 cells and about half a billion synapses. Both views cover the same approximately cubic millimetre of visual cortex from one mouse.
A wiring diagram alone shows which neurons may communicate, but not how they respond during behaviour. Activity recordings alone reveal response patterns without resolving every physical connection. MICrONS joins those levels at an unprecedented scale, allowing researchers to ask how a neuron's visual preferences relate to the cells that send it signals and the cells it contacts.
Building the dataset
While the mouse was awake and viewing natural and synthetic visual stimuli, researchers used two-photon calcium imaging across primary and higher visual areas. Calcium-sensitive fluorescence acts as an indirect indicator of neural activity. It is not a perfect measurement of every electrical spike, but it makes simultaneous recording from large populations possible.
After the functional experiments, the team preserved the same tissue, cut it into extremely thin serial sections and imaged the sections with transmission electron microscopy. Computational systems segmented cells, traced axons and dendrites, and proposed synaptic contacts in three dimensions. Human and automated proofreading then corrected important reconstructions. Finally, the researchers co-registered the activity data with the anatomical volume so responses could be assigned to corresponding cells.
The released resource includes interfaces and software for querying synapses, cell bodies and reconstructions. Making the data open is central to its value: many groups can test competing circuit hypotheses without reproducing a petabyte-scale imaging campaign.
Early results reveal statistical wiring rules
Companion analyses used the dataset to examine excitatory and inhibitory cell types, dendritic form and communication across cortical layers and visual areas. One prominent result is a broad like-to-like rule: excitatory neurons with similar response properties are more likely to connect, including across layers and between areas. Researchers also found higher-order organisation that could not be reduced entirely to the preferences of isolated neuron pairs.
These are probabilistic rules, not a fixed code in which every similarly tuned neuron connects. Physical proximity strongly affects whether an axon can meet a dendrite, and cell type, cortical layer and many unmeasured biological factors also matter. The large volume lets researchers model these influences together rather than treating a few reconstructed cells as representative.
A cubic millimetre is both huge and tiny
At synaptic resolution, one cubic millimetre is a massive technical achievement. Relative to a whole mouse brain, it remains a small sample, and it covers visual cortex rather than every brain system. The most detailed structure-function alignment comes primarily from one animal. Individual brains differ, and tissue preparation captures anatomy at one endpoint after the earlier activity sessions.
The resource therefore is not a complete mouse connectome and says even less directly about the human brain. Calcium signals are slower than electrical spikes, reconstruction errors remain, and many synapses' physiological strength is unknown. A visible anatomical contact does not alone reveal whether its influence is strong, weak or changing with experience.
Why the map changes what can be tested
Its importance lies in linking scales. Scientists can start with a neuron's observed response, follow its local and long-range anatomy, compare it with gene-informed cell classes and test network models against measured connectivity. Digital models trained to predict visual responses can be checked against real synapses instead of only against output activity.
Future volumes from more animals, brain regions and behavioural conditions will show which rules generalise. Faster proofreading and new microscopy methods may reduce the extraordinary time and computing required. MICrONS does not finish the problem of understanding the brain; it creates a shared empirical object detailed enough for theories of cortical computation to fail in precise, informative ways.
