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

Reviewer #1 (Public review):

Summary:

Here, the authors address the organization of reach-related activity in layer 2/3 across a broad swath of anterodorsal neocortex that included large subregions of M1, M2, and S1. In mice performing a novel variant water-reaching task, the authors measured activity using two-photon fluorescence imaging of a GECI expressed in excitatory projection neurons. The authors found a substantial diversity of response patterns using a number of metrics they developed for characterizing the PETHs of neurons across reach conditions (target locations). By mapping single-neuron properties across the cortex, the authors found substantial spatial variation, only some of which aligned with traditional boundaries between cortical regions. Using Gaussian mixture models, the authors found evidence of distinct response types in each region, with several types prominent in multiple cortical regions. Aggregating across regions, four primary subpopulations were apparent, each distinct in its average response properties. Strikingly, each subpopulation was observed in multiple regions, but subpopulation members from different regions exhibited largely similar response properties.

Strengths:

The work addresses a fundamental question in the field that has not previously been addressed at cellular resolution across such a broad cortical extent. I see this as truly foundational work that will support future investigation of how the rodent brain drives and controls reaching.

The quantification is thoughtful and rigorous. It is great that the authors provide an explanation for and intuition behind their response metrics, rather than burying everything in the Methods.

The Discussion and general contextualization of the results are thorough, thoughtful, and strong. It is great that the authors avoid the common over-interpretation of classical observations regarding cortical organization that are endemic in the field.

All things considered, this is the best paper regarding spatial structure in the motor system I have ever read. The breadth of cellular resolution activity measurement, the rigor of the quantification, and the clear and open-minded interrogation of the data collectively have produced a very special piece of work.

Weaknesses:

The behavioral task is very impressive and an important contribution to the field in its own right. However, given that it appears substantially different from the one used in the previous paper, the characterization of the behavior provided in the Results is too brief. More illustration of the behavior would be helpful. For example, it is rather deep into the paper when the authors reveal that the mice can whisk to help localize the target location. That should be expressed at the outset when the behavior is first described. Other suggestions for elaborating the behavior description are included below.

Statistical support for key claims is lacking. For example, "The five areas of interest varied in the fraction of neurons that were modulated: M2 had 14%, M1 had 23%, S1-fl had 30%, S1-hl had 25%, and S1-tr had 27%" - I cannot locate the statistical tests showing that these values are actually different. Another example is Figure 7, where a key observation is that distributions of PETH features are distinct across regions. It is clear that at least some distributions are not overlapping, but a clearer statistical basis for this key claim should be provided.

I understand that the authors are planning a follow-up study that addresses the relation between activity patterns and kinematics. One question about interpreting the results here though, is how much the activity variation across target locations may relate to the kinematic differences across these different conditions, as opposed to true higher-order movement features like reach direction.

Reviewer #2 (Public review):

Summary:

The functional parcellation of cortical areas is a critical question in neuroscience. This is particularly true in frontal areas in mice. While sensory areas are relatively well characterized by their tuning to sensory stimuli, the situation is much less clear for motor areas. This has become even more ambiguous since recent studies using large-scale neuronal recordings consistently report mixed sensory and motor-related activity throughout the brain, and motor mapping studies have shown that movements evoked by cortical stimulation are by no means limited to motor areas alone. Here, the authors use a correlation approach combining large-scale functional imaging at cellular resolution with movement-tracking in mice executing a reaching task. Across multiple recording sessions in the same animals, the authors have imaged a large portion of the sensorimotor cortex at cellular resolution in mice performing a reaching task, recording the activity of nearly 40,000 neurons. By aligning the calcium signal of each neuron to three task events-the Go cue triggering the reach, the onset of paw lift, and the contact between the paw and the target-for different target positions, the authors identified different response patterns distributed differently across cortical areas. They defined a set of features that describe the neurons' response pattern, representing the temporal dynamics and tuning properties for the different target positions. These features were used to construct cortical maps, and the authors show that, interestingly, gradient maps obtained from the first derivative of the feature maps reveal sharp discontinuities at the boundaries between anatomically defined cortical areas. Using dimensionality reduction of the neuronal response features, the authors found that, despite clear differences in their average response properties, individual neurons from the same cortical areas do not form distinct clusters in the reduced-dimensional space. In fact, most areas contain heterogeneous neuronal populations, and most neuronal populations are present in multiple areas, albeit in different proportions. Interestingly, the authors identified four neuronal subpopulations based on the distance between the components of the Gaussian mixture model used to model the distribution of neurons within each area. One of these subpopulations is almost exclusively represented in the anterior M2 cortex, while another is broadly distributed across the different areas.

Strengths:

This article is based on an impressive dataset of nearly 40,000 neurons covering a large portion of the sensorimotor cortex and on innovative analytical approaches. This study is likely the first to clearly demonstrate boundaries between cortical areas defined based on the responses of individual neurons. This innovative approach to functional mapping of cortical areas potentially opens up new perspectives for higher-resolution mapping of frontal cortical areas, using a broader repertoire of sensory and motor evoked responses.

Weaknesses:

The second part of the article, which presents multimodal responses in the cortical areas, seems to be a perhaps overly complicated way of showing what has already been demonstrated in numerous recent publications, but these new analyses expand upon these previous observations by revealing an interesting functional organization of the sensorimotor cortex, highlighting interesting similarities and differences between certain areas.