Neural activity profiles reveal overlapping, intermingled subpopulations spanning area borders in mouse sensorimotor cortex

By performing cellular-resolution calcium imaging across five regions of mouse sensorimotor cortex during a reach-to-grasp task, the study reveals that while neuronal response properties shift abruptly at anatomical borders, behaviorally relevant activity is organized into distinct, intermingled subpopulations that span multiple areas rather than being strictly segregated by cortical region.

The Gist

Imagine your brain's sensorimotor cortex (the part that controls your movements and feels your touch) as a massive, bustling city. For a long time, scientists thought this city was divided into strict, walled-off neighborhoods: a "Motor District" for planning moves, a "Sensory District" for feeling things, and specific zones for your arm, leg, and face. The assumption was that if you lived in the Motor District, you only did motor things, and if you lived in the Sensory District, you only did sensory things.

This new study, however, is like sending a drone with a high-definition camera over that city to see what the people are actually doing. The researchers didn't just look at the street signs (anatomy); they watched the behavior of over 39,000 individual neurons (the citizens) while mice performed a complex task: reaching for water from 15 different spouts.

Here is what they found, explained simply:

1. The City is Messier Than the Map

When the researchers looked at how these neurons reacted, they found that the "neighborhoods" aren't as rigid as the map suggests.

• The Old View: Think of it like a school where the Math class is in Room 101 and the Art class is in Room 102. You expect all Math students to be in Room 101.
• The New Reality: It's more like a giant open-plan office. While there are general "zones" (like the front of the office being more focused on planning and the back being more focused on reacting), the people doing specific jobs are mixed together. A person doing "Math" might be sitting right next to someone doing "Art," and they are both working on the same project.

2. The "Salt-and-Pepper" Mix

The most surprising discovery is that these different types of neurons are intermingled.

• The Analogy: Imagine a bowl of salt and pepper. You might expect the salt to be in one corner and the pepper in another. But in this brain, the salt and pepper are mixed so thoroughly that if you took a tiny pinch from anywhere in the bowl, you'd get a little bit of both.
• What this means: Even within a specific anatomical area (like the "Forearm Zone"), you find neurons that act like "planners" and neurons that act like "reactors" all sitting right next to each other. They aren't segregated into neat little groups.

3. The Four "Super-Teams"

Instead of organizing by where the neuron lives, the researchers found that neurons organize by what they do. They identified four distinct "super-teams" of neurons that span across the whole city:

• The "Anterior" Team (The Planners): These are the strategists. They are mostly found in the front of the motor area. They are very precise, focused on where the target is, and they keep that plan in mind for a long time. They are like the project manager who keeps the big picture in mind.
• The "Forelimb Motor" Team (The Executors): These span the motor area and the forearm sensory area. They are also focused on the target but are more about the actual movement execution. They are like the construction crew getting the job done.
• The "Forelimb Somatomotor" Team (The Sensory-Motor Mix): These are found in the sensory area and parts of the motor area. They are quick, brief responders. They are like the security guards who react instantly to a specific event (like the paw lifting) but don't hold a long-term plan.
• The "Hindlimb Somatomotor" Team (The Stabilizers): These are found in the leg and trunk areas. They are a bit more chaotic, firing at different times depending on the specific target. They seem to be helping the animal brace itself and stay balanced while reaching.

4. The "Ghost" Overlap

Here is the kicker: These four teams overlap.

• The Analogy: Imagine four different radio stations broadcasting over the same city. In the "Motor District," you might hear mostly the "Planner" station, but you can also hear the "Executor" station. In the "Sensory District," you hear mostly the "Stabilizer" station, but the "Executor" station is also playing loudly.
• The Result: The reason the "Motor District" looks different from the "Sensory District" isn't because they have different types of people. It's because they have different mixes of the same four teams. The "Motor District" just has a higher volume of Planners and Executors, while the "Sensory District" has a higher volume of Stabilizers and Sensory-Mixers.

Why Does This Matter?

This changes how we think about the brain.

• Old Idea: The brain is a collection of specialized rooms.
• New Idea: The brain is a dynamic, overlapping network. The "specialty" of a brain area isn't just about where it is, but about the mixture of teams working there.

In a nutshell: The brain doesn't sort its workers into separate buildings. Instead, it has four types of workers (Planners, Executors, Reactors, Stabilizers) who are mixed together like salt and pepper across the whole city. The "Motor City" just happens to have more Planners, while the "Sensory City" has more Stabilizers, but they are all working together in the same messy, beautiful, intermingled space to help the mouse grab that water.

Technical Summary

1. Problem Statement

While anatomical atlases and projection tracing have established that the mammalian cortex is divided into distinct areas (e.g., M1, M2, S1) with specific connectivity, the functional organization of neural activity at the single-cell resolution during complex behavior remains poorly understood. Previous large-scale recordings suggest motor signals are widespread, but the specific spatial distribution of how individual neurons encode movement parameters (tuning, timing, linearity) across the sensorimotor sheet is unclear. The central question is: How are the response properties of individual cortical cells organized across sensorimotor cortex during behavior, and do these functional organizations align with anatomical borders or somatotopic maps?

2. Methodology

The authors employed a high-throughput, cellular-resolution approach to map neural activity across multiple cortical areas simultaneously.

• Subjects & Behavior: Six head-fixed mice were trained on a 15-target delayed reach-to-grasp-to-drink task. The task involved reaching with the right paw to one of 15 randomly selected spout positions arranged in a concave grid. Mice achieved a 97.1% success rate.
• Imaging: The researchers used two-photon calcium imaging (GCaMP8s) to record from Layer 2/3 pyramidal neurons.
  • Scale: They imaged 39,398 neurons across 98 fields of view (FOVs).
  • Coverage: The imaging covered five key regions: Secondary Motor Cortex (M2), Primary Motor Cortex (M1), and the primary somatosensory areas for the forelimb (S1-fl), hindlimb (S1-hl), and trunk (S1-tr).
  • Alignment: FOVs were aligned to the Allen Mouse Brain Common Coordinate Framework (CCF v3) using widefield imaging and paw vibration mapping.
• Feature Engineering: Instead of relying solely on unsupervised dimensionality reduction (like PCA), the authors defined five interpretable metrics to characterize each neuron's trial-averaged Peri-Event Time Histogram (PETH):
  1. Response Duration: Autocorrelation width (sustained vs. transient).
  2. Peak Time Variation: Standard deviation of peak firing times across different targets.
  3. Tuning Sharpness: Selectivity for specific targets (broad vs. sharp).
  4. Target Tuning Linearity: How linearly firing rates scaled with the physical distance of the target.
  5. Tuning Persistence: Consistency of the tuning profile across different time points during the movement.
• Statistical Analysis:
  • Gradient Analysis: Computed spatial gradients of feature maps to identify sharp transitions (boundaries).
  • Dimensionality Reduction: Used t-SNE to visualize the distribution of feature vectors.
  • Generative Modeling: Fit Gaussian Mixture Models (GMMs) to the feature distributions within each anatomical area to identify multimodal subpopulations.
  • Cross-Area Likelihood: Calculated the likelihood of neurons from one area belonging to the distribution of another to test for shared subpopulations.

3. Key Contributions

• Dense Functional Atlas: Created the first high-resolution, single-cell resolution map of sensorimotor cortex activity spanning five distinct areas during a complex, multi-target reaching task.
• Feature-Based Organization: Demonstrated that functional organization is defined not just by what is encoded (e.g., target location), but by how it is encoded (e.g., linearity, persistence, duration).
• Discovery of Area-Spanning Subpopulations: Identified that the cortex is not organized into strictly segregated functional zones. Instead, distinct neural subpopulations with characteristic activity profiles exist that span multiple anatomical areas and are salt-and-pepper intermingled within those areas.
• Methodological Framework: Provided a robust pipeline for combining interpretable feature engineering with generative modeling to uncover latent structure in large-scale neural datasets.

4. Key Results

A. Heterogeneity and Spatial Structure

• Modulation: Task-modulated neurons were found in all areas, but the fraction varied (M2: 14%, M1: 23%, S1-fl: 30%). Modulation was highest at the M1/S1-fl border.
• Temporal Onsets: Activation onset times followed a somatotopic gradient, with S1-trunk/hindlimb activating earliest (likely for stabilization), followed by M2 and M1.
• Feature Boundaries: Spatial gradients of the five metrics revealed that response properties change most abruptly at anatomical borders (e.g., M1/M2) and somatotopic borders (e.g., forelimb vs. hindlimb).
  • Motor areas (M1/M2) showed sharper tuning, more linear target encoding, and higher tuning persistence.
  • Somatosensory areas (S1) showed more heterogeneous, transient, and non-linear responses.

B. Multimodality and Subpopulations

• Multimodal Distributions: Within each anatomical area, the distribution of response features was multimodal (not unimodal), suggesting distinct functional "types" of neurons coexist within a single area.
• Four Distinct Subpopulations: GMM analysis across all areas revealed four distinct subpopulations that spanned anatomical boundaries:
  1. Anterior Subpopulation: Unique to M2. Characterized by high tuning persistence, sharp tuning, and high linearity (abstract target representation).
  2. Forelimb Motor Subpopulation: Spanned M2 and the forelimb region of M1. Highly linear tuning but less sharp and less persistent than the Anterior type.
  3. Forelimb Somatomotor Subpopulation: Spanned M1 (forelimb/orofacial) and anterior S1-fl. Broadly tuned, non-linear, low persistence, and very brief responses.
  4. Hindlimb Somatomotor Subpopulation: Centered on S1-hl, extending to S1-tr and posterior M1/M2. Similar to Forelimb Somatomotor but with high peak time variability (firing at different times for different targets).

C. Spatial Intermingling

• Overlapping Footprints: The spatial likelihood maps of these subpopulations overlapped significantly. For example, the Anterior and Forelimb Motor subpopulations overlapped in M2.
• Salt-and-Pepper Intermingling: In overlap zones, neurons from different subpopulations were physically intermingled at the single-cell level.
• Homogeneity: Neurons belonging to the same subpopulation exhibited similar response profiles regardless of their specific anatomical location (e.g., a Forelimb Somatomotor neuron in M1 looked like one in S1-fl).

5. Significance

• Redefining Cortical Organization: The study challenges the classical view of the cortex as a set of functionally distinct, non-overlapping areas. Instead, it proposes a model where functional subpopulations are distributed across the cortical sheet, with anatomical areas acting as mixtures of these subpopulations in varying proportions.
• Circuit Implications: The existence of area-spanning subpopulations suggests that specific functional circuits may be distributed across multiple areas (recurrent circuits) or that activity is broadcast from a source area to others, creating shared functional motifs.
• Complement to Anatomy: These activity-based maps complement existing anatomical and genetic atlases. They reveal that while anatomical borders exist, the functional "code" of the cortex is organized by overlapping, intermingled populations that transcend these borders.
• Future Directions: The findings suggest that understanding motor control requires analyzing the specific mixture of subpopulations in a given region rather than treating the region as a monolithic functional unit. This framework can be applied to other behaviors and cortical layers to further dissect the neural basis of movement.