Pyramidal cell types drive functionally distinct cortical activity patterns during decision-making

Understanding how the neocortex generates complex behaviors requires a deep understanding of its constituent cell types and their interactions. While the neocortex exhibits a conserved microcircuit motif across layers, each layer comprises distinct cell types categorized by genetic markers, morphology, projections, or lineage. The functional roles of inhibitory interneurons are relatively well-understood, thanks to cell-type-specific mouse lines which have revealed their roles in network synchronization and state-dependent sensory processing. However, the roles of glutamatergic pyramidal neurons (PyNs), constituting ~80% of cortical neurons and forming nearly all long-range projections, remain less clear. PyNs are far more diverse than interneurons, with RNA sequencing suggesting at least 100 subtypes. They are broadly categorized into two main types based on long-range projections: intratelencephalic (IT) neurons projecting to other cortical areas and the striatum, and pyramidal tract (PT) neurons projecting to subcortical structures like the pons and thalamus. PT and IT neurons also differ electrophysiologically, in dendritic arborization, local connectivity, and sensory tuning. Notably, only PT neurons in sensory cortex are required for perception of certain stimuli, suggesting distinct information streams. Previous studies have focused on single cortical areas, leaving open the question of whether PyN-specific subcircuits are a general principle of cortical organization. This study aimed to address this gap using widefield calcium imaging to measure cortex-wide activity with cell-type specificity.

Studies on cortical interneurons using cell-type-specific mouse lines have significantly advanced our understanding of inhibitory circuit motifs and their roles in network synchronization and state-dependent sensory processing. However, the functional roles of glutamatergic pyramidal neurons (PyNs), which comprise the majority of cortical neurons, remain less well-defined. While often considered a homogeneous group, PyNs exhibit significant diversity in their gene expression, morphology, long-range projections, and electrophysiological properties. Previous research has identified at least 100 putative PyN subtypes within cortical layers. Major PyN subtypes have been categorized based on their projection targets: intratelencephalic (IT) neurons projecting within the telencephalon and pyramidal tract (PT) neurons projecting to subcortical structures. Studies suggest that these subtypes process separate streams of information, with PT neurons particularly important for sensory perception in sensory cortices. The functional divergence of PyN types suggests that PT and IT neurons form parallel subnetworks that independently process different information streams. However, previous research has largely focused on single cortical areas, leaving it unclear whether PyN-type-specific subcircuits are a general feature across the cortex.

The researchers employed several key methodologies in this study: 1. **Genetic and Retrograde Labeling:** They used CreER lines (Fezf2-CreER for PT neurons and PlexinD1-CreER for IT neurons) crossed with Ai162 mice to achieve PyN-type-specific expression of the calcium indicator GCaMP6s. Retrograde labeling was also used to target corticostriatal (CStr) projection neurons by injecting CAV-2-Cre in reporter mice. This enabled cell-type-specific measurement of neural activity. 2. **Widefield Calcium Imaging:** This technique allowed for simultaneous measurement of neural activity across the dorsal cortex with high temporal resolution. Retinotopic mapping was used to identify visual areas. The imaging data were analyzed using dimensionality reduction techniques such as semi-nonnegative matrix factorization (sNMF) and uniform manifold approximation and projection (UMAP) to identify unique activity patterns associated with each PyN type. 3. **Auditory Decision-Making Task:** Mice were trained to perform an auditory discrimination task where they had to discriminate between left and right auditory stimuli and lick a corresponding spout to receive water. This task provided a behavioral context for assessing the functional roles of different PyN types. 4. **Optogenetic Inactivation:** Using the inhibitory opsin stGtACR2, the researchers performed PyN-type-specific optogenetic inactivation in auditory, parietal, and frontal cortices. This allowed them to causally test the functional role of each PyN type in decision-making. 5. **Data Analysis:** The study utilized various data analysis techniques including semi-nonnegative matrix factorization (sNMF), uniform manifold approximation and projection (UMAP), linear encoding models, and logistic regression decoders to analyze the imaging data and uncover task-related activity patterns and causal relationships between PyN types and behavior. 6. **Two-photon Calcium Imaging:** In addition to widefield imaging, two-photon calcium imaging was used at cellular resolution to confirm findings in frontal cortex, particularly focusing on the choice selectivity of CStr neurons. This provided complementary data to support and refine the conclusions from widefield imaging.

The study revealed several key findings: 1. **PyN-type-Specific Cortex-Wide Activity Patterns:** Dimensionality reduction and clustering analyses revealed that each PyN type (PT, IT, and CStr) exhibited unique cortex-wide activity patterns. These patterns were consistent across individual mice and sessions, suggesting the existence of specialized subcircuits for each PyN type. Analysis using localized sNMF (LocaNMF) indicated that these PyN-type specific differences were not simply due to differences in overall activity levels in specific cortical areas, but reflected differences in the coordinated activation of multi-area cortical networks. Interestingly, PyN-predictive LocaNMF components were significantly smaller than nonspecific components, suggesting that different PyNs might be most active in distinct subregions within traditionally defined cortical areas. 2. **Distinct Task-Related Activity:** During the auditory decision-making task, each PyN type showed distinct task-related activity. Encoding and decoding models revealed that parietal cortex primarily showed stimulus-related activity, with PT neurons having the strongest causal role in sensory processing. Frontal cortex showed choice-related activity, with all PyN types being required for accurate choices but displaying distinct choice tuning. Surprisingly, IT neurons showed a mild ipsilateral choice preference, a pattern not observed in PT neurons. 3. **CStr Neurons as a Functionally Divergent IT Subclass:** Further analysis revealed that CStr neurons form a functionally distinct subclass of IT neurons, contributing to the ipsilateral choice preference observed in IT neurons. Two-photon imaging confirmed that CStr neurons, particularly in superficial layers of the anterior lateral motor cortex (ALM), showed a stronger preference for ipsilateral choices than unlabeled PyNs. 4. **PyN-Type-Specific Causal Contributions:** Optogenetic inactivation experiments confirmed the distinct functional roles of different PyN types. Inactivation of PT neurons in parietal cortex strongly impaired sensory processing, while inactivation of all PyN types in frontal cortex impaired choice formation and retention. The effects of inactivation varied across different task periods, suggesting specific roles for each PyN type at different stages of the decision-making process.

This study provides compelling evidence for the functional diversity of pyramidal neuron types in the cortex and their distinct contributions to perceptual decision-making. The findings challenge the traditional view of PyNs as a monolithic group and highlight the importance of considering cell-type-specific dynamics when investigating cortical function. The observation of unique cortex-wide activity patterns for each PyN type suggests that different PyN subtypes form specialized subcircuits that independently process information. The distinct roles of PyN types in sensory processing (parietal cortex) and choice formation (frontal cortex) support a model of parallel processing, where different neuronal populations contribute to different aspects of behavior. The strong causal role of PT neurons in parietal cortex for sensory processing is consistent with their role in relaying sensory information to subcortical structures. The finding that all PyN types in frontal cortex are needed for accurate choices, despite their distinct choice tuning, points to an intricate interplay between these populations for decision formation. The discovery of ipsilateral choice preference in a subset of IT neurons, specifically those projecting to the striatum, suggests a complex mechanism involving striatal circuits for decision execution. These findings have significant implications for understanding cortical computation and how neuronal diversity underpins complex behavior.

This study demonstrates that different pyramidal neuron types exhibit functionally distinct, cortex-wide neural dynamics with separate roles during perceptual decision-making. The unique activity patterns and causal contributions of each PyN type highlight the importance of considering neuronal heterogeneity when studying cortical circuits. Future research should focus on further dissecting the functional roles of specific PyN subtypes and understanding the interactions between these subpopulations to generate complex behavior. The development of advanced genetic tools and imaging techniques will be crucial in achieving this goal.

While this study provides valuable insights into the functional diversity of pyramidal neurons, some limitations should be considered. The study primarily focused on a specific auditory decision-making task, and the generalizability of the findings to other tasks or sensory modalities needs further investigation. The optogenetic inactivation experiments targeted relatively large cortical areas, and future studies with more precise targeting could provide a more refined understanding of the functional roles of specific subregions. The study primarily utilized widefield imaging, which provides a mesoscale view of neural activity. Combining widefield imaging with cellular resolution techniques like two-photon imaging could provide a more complete picture of the underlying neural dynamics.