Reinforcement Learning Recruits Somata and Apical Dendrites across Layers of Primary Sensory Cortex

Goal-Directed Learning is Multidimensional and Accompanied by Diverse and Widespread Changes in Neocortical Signaling

New tasks are often learned in stages with each stage reflecting a different learning challenge. Accordingly, each learning stage is likely mediated by distinct neuronal processes. And yet, most rodent studies of the neuronal correlates of goal-directed learning focus on individual outcome measures and individual brain regions. Here, we longitudinally studied mice from naive to expert performance in a head-fixed, operant conditioning whisker discrimination task. In addition to tracking the primary behavioral outcome of stimulus discrimination, we tracked and compared an array of object-based and temporal-based behavioral measures. These behavioral analyses identify multiple, partially overlapping learning stages in this task, consistent with initial response implementation, early stimulus-response generalization, and late response inhibition. To begin to understand the neuronal foundations of these learning processes, we performed widefield Ca2+ imaging of dorsal neocortex throughout learning and correlated behavioral measures with neuronal activity. We found distinct and widespread correlations between neocortical activation patterns and various behavioral measures. For example, improvements in sensory discrimination correlated with target stimulus evoked activations of licking-related cortices along with distractor stimulus evoked global cortical suppression. Our study reveals multidimensional learning for a simple goal-directed learning task and generates hypotheses for the neuronal modulations underlying these various learning processes.

Rule-based modulation of a sensorimotor transformation across cortical areas

Flexible responses to sensory stimuli based on changing rules are critical for adapting to a dynamic environment. However, it remains unclear how the brain encodes rule information and uses this information to guide behavioral responses to sensory stimuli. Here, we made single-unit recordings while head-fixed mice performed a cross-modal sensory selection task in which they switched between two rules in different blocks of trials: licking in response to tactile stimuli applied to a whisker while rejecting visual stimuli, or licking to visual stimuli while rejecting the tactile stimuli. Along a cortical sensorimotor processing stream including the primary (S1) and secondary (S2) somatosensory areas, and the medial (MM) and anterolateral (ALM) motor areas, the single-trial activity of individual neurons distinguished between the two rules both prior to and in response to the tactile stimulus. Variable rule-dependent responses to identical stimuli could in principle occur via appropriate configuration of pre-stimulus preparatory states of a neural population, which would shape the subsequent response. We hypothesized that neural populations in S1, S2, MM and ALM would show preparatory activity states that were set in a rule-dependent manner to cause processing of sensory information according to the current rule. This hypothesis was supported for the motor cortical areas by findings that (1) the current task rule could be decoded from pre-stimulus population activity in ALM and MM; (2) neural subspaces containing the population activity differed between the two rules; and (3) optogenetic disruption of pre-stimulus states within ALM and MM impaired task performance. Our findings indicate that flexible selection of an appropriate action in response to a sensory input can occur via configuration of preparatory states in the motor cortex.

Auditory Corticofugal Neurons Transmit Auditory and Non-auditory Information During Behavior

Layer 5 pyramidal neurons of sensory cortices project "corticofugal" axons to myriad sub-cortical targets, thereby broadcasting high-level signals important for perception and learning. Recent studies suggest dendritic Ca2+ spikes as key biophysical mechanisms supporting corticofugal neuron function: these long-lasting events drive burst firing, thereby initiating uniquely powerful signals to modulate sub-cortical representations and trigger learning-related plasticity. However, the behavioral relevance of corticofugal dendritic spikes is poorly understood. We shed light on this issue using 2-photon Ca2+ imaging of auditory corticofugal dendrites as mice of either sex engage in a GO/NO-GO sound-discrimination task. Unexpectedly, only a minority of dendritic spikes were triggered by behaviorally relevant sounds under our conditions. Task related dendritic activity instead mostly followed sound cue termination and co-occurred with mice's instrumental licking during the answer period of behavioral trials, irrespective of reward consumption. Temporally selective, optogenetic silencing of corticofugal neurons during the trial answer period impaired auditory discrimination learning. Thus, auditory corticofugal systems' contribution to learning and plasticity may be partially nonsensory in nature.

Distributed and specific encoding of sensory, motor, and decision information in the mouse neocortex during goal-directed behavior

Goal-directed behaviors involve coordinated activity in many cortical areas, but whether the encoding of task variables is distributed across areas or is more specifically represented in distinct areas remains unclear. Here, we compared representations of sensory, motor, and decision information in the whisker primary somatosensory cortex, medial prefrontal cortex, and tongue-jaw primary motor cortex in mice trained to lick in response to a whisker stimulus with mice that were not taught this association. Irrespective of learning, properties of the sensory stimulus were best encoded in the sensory cortex, whereas fine movement kinematics were best represented in the motor cortex. However, movement initiation and the decision to lick in response to the whisker stimulus were represented in all three areas, with decision neurons in the medial prefrontal cortex being more selective, showing minimal sensory responses in miss trials and motor responses during spontaneous licks. Our results reconcile previous studies indicating highly specific vs. highly distributed sensorimotor processing.

Layer 1 neocortex: Gating and integrating multidimensional signals

Layer 1 (L1) of the neocortex acts as a nexus for the collection and processing of widespread information. By integrating ascending inputs with extensive top-down activity, this layer likely provides critical information regulating how the perception of sensory inputs is reconciled with expectation. This is accomplished by sorting, directing, and integrating the complex network of excitatory inputs that converge onto L1. These signals are combined with neuromodulatory afferents and gated by the wealth of inhibitory interneurons that either are embedded within L1 or send axons from other cortical layers. Together, these interactions dynamically calibrate information flow throughout the neocortex. This review will primarily focus on L1 within the primary sensory cortex and will use these insights to understand L1 in other cortical areas.

History information emerges in the cortex during learning

We learn from our experience but the underlying neuronal mechanisms incorporating past information to facilitate learning is relatively unknown. Specifically, which cortical areas encode history-related information and how is this information modulated across learning? To study the relationship between history and learning, we continuously imaged cortex-wide calcium dynamics as mice learn to use their whiskers to discriminate between two different textures. We mainly focused on comparing the same trial type with different trial history, that is, a different preceding trial. We found trial history information in barrel cortex (BC) during stimulus presentation. Importantly, trial history in BC emerged only as the mouse learned the task. Next, we also found learning-dependent trial history information in rostrolateral (RL) association cortex that emerges before stimulus presentation, preceding activity in BC. Trial history was also encoded in other cortical areas and was not related to differences in body movements. Interestingly, a binary classifier could discriminate trial history at the single trial level just as well as current information both in BC and RL. These findings suggest that past experience emerges in the cortex around the time of learning, starting from higher-order association area RL and propagating down (i.e., top-down projection) to lower-order BC where it can be integrated with incoming sensory information. This integration between the past and present may facilitate learning.

Convolutional neural networks for vision neuroscience: significance, developments, and outstanding issues

Convolutional Neural Networks (CNN) are a class of machine learning models predominately used in computer vision tasks and can achieve human-like performance through learning from experience. Their striking similarities to the structural and functional principles of the primate visual system allow for comparisons between these artificial networks and their biological counterparts, enabling exploration of how visual functions and neural representations may emerge in the real brain from a limited set of computational principles. After considering the basic features of CNNs, we discuss the opportunities and challenges of endorsing CNNs as in silico models of the primate visual system. Specifically, we highlight several emerging notions about the anatomical and physiological properties of the visual system that still need to be systematically integrated into current CNN models. These tenets include the implementation of parallel processing pathways from the early stages of retinal input and the reconsideration of several assumptions concerning the serial progression of information flow. We suggest design choices and architectural constraints that could facilitate a closer alignment with biology provide causal evidence of the predictive link between the artificial and biological visual systems. Adopting this principled perspective could potentially lead to new research questions and applications of CNNs beyond modeling object recognition.

A transthalamic pathway crucial for perception

Perception arises from activity between cortical areas, first primary cortex and then higher order cortices. This communication is served in part by transthalamic (cortico-thalamo-cortical) pathways, which ubiquitously parallel direct corticocortical pathways, but their role in sensory processing has largely remained unexplored. Here, we show that the transthalamic pathway linking somatosensory cortices propagates task-relevant information required for correct sensory decisions. Using optogenetics, we specifically inhibited the pathway at its synapse in higher order somatosensory thalamus of mice performing a texture-based discrimination task. We concurrently monitored the cellular effects of inhibition in primary or secondary cortex using two-photon calcium imaging. Inhibition severely impaired performance despite intact direct corticocortical projections, thus challenging the purely corticocentric map of perception. Interestingly, the inhibition did not reduce overall cell responsiveness to texture stimulation in somatosensory cortex, but rather disrupted the texture selectivity of cells, a discriminability that develops over task learning. This discriminability was more disrupted in the secondary than primary somatosensory cortex, emphasizing the feedforward influence of the transthalamic route. Transthalamic pathways thus appear critical in delivering performance-relevant information to higher order cortex and are critical hierarchical pathways in perceptual decision-making.

An Open-Source Virtual Reality System for the Measurement of Spatial Learning in Head-Restrained Mice

Head-restrained behavioral experiments in mice allow neuroscientists to observe neural circuit activity with high-resolution electrophysiological and optical imaging tools while delivering precise sensory stimuli to a behaving animal. Recently, human and rodent studies using virtual reality (VR) environments have shown VR to be an important tool for uncovering the neural mechanisms underlying spatial learning in the hippocampus and cortex, due to the extremely precise control over parameters such as spatial and contextual cues. Setting up virtual environments for rodent spatial behaviors can, however, be costly and require an extensive background in engineering and computer programming. Here, we present a simple yet powerful system based upon inexpensive, modular, open-source hardware and software that enables researchers to study spatial learning in head-restrained mice using a VR environment. This system uses coupled microcontrollers to measure locomotion and deliver behavioral stimuli while head-restrained mice run on a wheel in concert with a virtual linear track environment rendered by a graphical software package running on a single-board computer. The emphasis on distributed processing allows researchers to design flexible, modular systems to elicit and measure complex spatial behaviors in mice in order to determine the connection between neural circuit activity and spatial learning in the mammalian brain.

Cortical glutamatergic projection neuron types contribute to distinct functional subnetworks

The cellular basis of cerebral cortex functional architecture remains not well understood. A major challenge is to monitor and decipher neural network dynamics across broad cortical areas yet with projection-neuron-type resolution in real time during behavior. Combining genetic targeting and wide-field imaging, we monitored activity dynamics of subcortical-projecting (PTFezf2) and intratelencephalic-projecting (ITPlxnD1) types across dorsal cortex of mice during different brain states and behaviors. ITPlxnD1 and PTFezf2 neurons showed distinct activation patterns during wakeful resting, during spontaneous movements and upon sensory stimulation. Distinct ITPlxnD1 and PTFezf2 subnetworks were dynamically tuned to different sensorimotor components of a naturalistic feeding behavior, and optogenetic inhibition of ITsPlxnD1 and PTsFezf2 in subnetwork nodes disrupted distinct components of this behavior. Lastly, ITPlxnD1 and PTFezf2 projection patterns are consistent with their subnetwork activation patterns. Our results show that, in addition to the concept of columnar organization, dynamic areal and projection-neuron-type specific subnetworks are a key feature of cortical functional architecture linking microcircuit components with global brain networks.

Cocaine regulates sensory filtering in cortical pyramidal neurons

Exposure to cocaine leads to robust changes in the structure and function of neurons within the mesocorticolimbic pathway. However, little is known about how cocaine influences the processing of information within the sensory cortex. We address this by using patch-clamp and juxtacellular voltage recordings and two-photon Ca2+ imaging in vivo to investigate the influence of acute cocaine exposure on layer 2/3 (L2/3) pyramidal neurons within the primary somatosensory cortex (S1). Here, cocaine dampens membrane potential state transitions and decreases spontaneous somatic action potentials and Ca2+ transients. In contrast to the uniform decrease in background spontaneous activity, cocaine has a heterogeneous influence on sensory encoding, increasing tactile-evoked responses in dendrites that do not typically encode sensory information and decreasing responses in those dendrites that are more reliable sensory encoders. Combined, these findings suggest that cocaine acts as a filter that suppresses background noise to selectively modulate incoming sensory information.

Quantitative Fluorescence Analysis Reveals Dendrite-Specific Thalamocortical Plasticity in L5 Pyramidal Neurons during Learning

High-throughput anatomic data can stimulate and constrain new hypotheses about how neural circuits change in response to experience. Here, we use fluorescence-based reagents for presynaptic and postsynaptic labeling to monitor changes in thalamocortical synapses onto different compartments of layer 5 (L5) pyramidal (Pyr) neurons in somatosensory (barrel) cortex from mixed-sex mice during whisker-dependent learning (Audette et al., 2019). Using axonal fills and molecular-genetic tags for synapse identification in fixed tissue from Rbp4-Cre transgenic mice, we found that thalamocortical synapses from the higher-order posterior medial thalamic nucleus showed rapid morphologic changes in both presynaptic and postsynaptic structures at the earliest stages of sensory association training. Detected increases in thalamocortical synaptic size were compartment specific, occurring selectively in the proximal dendrites onto L5 Pyr and not at inputs onto their apical tufts in L1. Both axonal and dendritic changes were transient, normalizing back to baseline as animals became expert in the task. Anatomical measurements were corroborated by electrophysiological recordings at different stages of training. Thus, fluorescence-based analysis of input- and target-specific synapses can reveal compartment-specific changes in synapse properties during learning.SIGNIFICANCE STATEMENT Synaptic changes underlie the cellular basis of learning, experience, and neurologic diseases. Neuroanatomical methods to assess synaptic plasticity can provide critical spatial information necessary for building models of neuronal computations during learning and experience but are technically and fiscally intensive. Here, we describe a confocal fluorescence microscopy-based analytical method to assess input, cell type, and dendritic location-specific synaptic plasticity in a sensory learning assay. Our method not only confirms prior electrophysiological measurements but allows us to predict functional strength of synapses in a pathway-specific manner. Our findings also indicate that changes in primary sensory cortices are transient, occurring during early learning. Fluorescence-based synapse identification can be an efficient and easily adopted approach to study synaptic changes in a variety of experimental paradigms.

Cortex-wide response mode of VIP-expressing inhibitory neurons by reward and punishment

Neocortex is classically divided into distinct areas, each specializing in different function, but all could benefit from reinforcement feedback to inform and update local processing. Yet it remains elusive how global signals like reward and punishment are represented in local cortical computations. Previously, we identified a cortical neuron type, vasoactive intestinal polypeptide (VIP)-expressing interneurons, in auditory cortex that is recruited by behavioral reinforcers and mediates disinhibitory control by inhibiting other inhibitory neurons. As the same disinhibitory cortical circuit is present virtually throughout cortex, we wondered whether VIP neurons are likewise recruited by reinforcers throughout cortex. We monitored VIP neural activity in dozens of cortical regions using three-dimensional random access two-photon microscopy and fiber photometry while mice learned an auditory discrimination task. We found that reward and punishment during initial learning produce rapid, cortex-wide activation of most VIP interneurons. This global recruitment mode showed variations in temporal dynamics in individual neurons and across areas. Neither the weak sensory tuning of VIP interneurons in visual cortex nor their arousal state modulation was fully predictive of reinforcer responses. We suggest that the global response mode of cortical VIP interneurons supports a cell-type-specific circuit mechanism by which organism-level information about reinforcers regulates local circuit processing and plasticity.

Learning enhances encoding of time and temporal surprise in mouse primary sensory cortex

Primary sensory cortex has long been believed to play a straightforward role in the initial processing of sensory information. Yet, the superficial layers of cortex overall are sparsely active, even during sensory stimulation; additionally, cortical activity is influenced by other modalities, task context, reward, and behavioral state. Our study demonstrates that reinforcement learning dramatically alters representations among longitudinally imaged neurons in superficial layers of mouse primary somatosensory cortex. Learning an object detection task recruits previously unresponsive neurons, enlarging the neuronal population sensitive to touch and behavioral choice. Cortical responses decrease upon repeated stimulus presentation outside of the behavioral task. Moreover, training improves population encoding of the passage of time, and unexpected deviations in trial timing elicit even stronger responses than touches do. In conclusion, the superficial layers of sensory cortex exhibit a high degree of learning-dependent plasticity and are strongly modulated by non-sensory but behaviorally-relevant features, such as timing and surprise.

Activity in the superficial layers of the sensory cortex is believed to be largely driven by incoming sensory stimuli. Here the authors demonstrate how learning changes neural responses to sensations according to both behavioral relevance and timing, suggesting a high degree of non-sensory modulation.

Two-photon calcium imaging of neuronal activity

In vivo two-photon calcium imaging (2PCI) is a technique used for recording neuronal activity in the intact brain. It is based on the principle that, when neurons fire action potentials, intracellular calcium levels rise, which can be detected using fluorescent molecules that bind to calcium. This Primer is designed for scientists who are considering embarking on experiments with 2PCI. We provide the reader with a background on the basic concepts behind calcium imaging and on the reasons why 2PCI is an increasingly powerful and versatile technique in neuroscience. The Primer explains the different steps involved in experiments with 2PCI, provides examples of what ideal preparations should look like and explains how data are analysed. We also discuss some of the current limitations of the technique, and the types of solutions to circumvent them. Finally, we conclude by anticipating what the future of 2PCI might look like, emphasizing some of the analysis pipelines that are being developed and international efforts for data sharing.

Falling Short: The Contribution of Central Insulin Receptors to Gait Dysregulation in Brain Aging

Insulin resistance, which manifests as a reduction of insulin receptor signaling, is known to correlate with pathological changes in peripheral tissues as well as in the brain. Central insulin resistance has been associated with impaired cognitive performance, decreased neuronal health, and reduced brain metabolism; however, the mechanisms underlying central insulin resistance and its impact on brain regions outside of those associated with cognition remain unclear. Falls are a leading cause of both fatal and non-fatal injuries in the older population. Despite this, there is a paucity of work focused on age-dependent alterations in brain regions associated with ambulatory control or potential therapeutic approaches to target these processes. Here, we discuss age-dependent alterations in central modalities that may contribute to gait dysregulation, summarize current data supporting the role of insulin signaling in the brain, and highlight key findings that suggest insulin receptor sensitivity may be preserved in the aged brain. Finally, we present novel results showing that administration of insulin to the somatosensory cortex of aged animals can alter neuronal communication, cerebral blood flow, and the motivation to ambulate, emphasizing the need for further investigations of intranasal insulin as a clinical management strategy in the older population.

Sensitivity of the S1 neuronal calcium network to insulin and Bay-K 8644 in vivo: Relationship to gait, motivation, and aging processes

Neuronal hippocampal Ca²⁺ dysregulation is a critical component of cognitive decline in brain aging and Alzheimer's disease and is suggested to impact communication and excitability through the activation of a larger after hyperpolarization. However, few studies have tested for the presence of Ca²⁺ dysregulation in vivo, how it manifests, and whether it impacts network function across hundreds of neurons. Here, we tested for neuronal Ca²⁺ network dysregulation in vivo in the primary somatosensory cortex (S1) of anesthetized young and aged male Fisher 344 rats using single‐cell resolution techniques. Because S1 is involved in sensory discrimination and proprioception, we tested for alterations in ambulatory performance in the aged animal and investigated two potential pathways underlying these central aging‐ and Ca²⁺‐dependent changes. Compared to young, aged animals displayed increased overall activity and connectivity of the network as well as decreased ambulatory speed. In aged animals, intranasal insulin (INI) increased network synchronicity and ambulatory speed. Importantly, in young animals, delivery of the L‐type voltage‐gated Ca²⁺ channel modifier Bay‐K 8644 altered network properties, replicating some of the changes seen in the older animal. These results suggest that hippocampal Ca²⁺ dysregulation may be generalizable to other areas, such as S1, and might engage modalities that are associated with locomotor stability and motivation to ambulate. Further, given the safety profile of INI in the clinic and the evidence presented here showing that this central dysregulation is sensitive to insulin, we suggest that these processes can be targeted to potentially increase motivation and coordination while also reducing fall frequency with age.

Dynamic compartmental computations in tuft dendrites of layer 5 neurons during motor behavior

Tuft dendrites of layer 5 pyramidal neurons form specialized compartments important for motor learning and performance, yet their computational capabilities remain unclear. Structural-functional mapping of the tuft tree from the motor cortex during motor tasks revealed two morphologically distinct populations of layer 5 pyramidal tract neurons (PTNs) that exhibit specific tuft computational properties. Early bifurcating and large nexus PTNs showed marked tuft functional compartmentalization, representing different motor variable combinations within and between their two tuft hemi-trees. By contrast, late bifurcating and smaller nexus PTNs showed synchronous tuft activation. Dendritic structure and dynamic recruitment of the N-methyl-d-aspartate (NMDA)-spiking mechanism explained the differential compartmentalization patterns. Our findings support a morphologically dependent framework for motor computations, in which independent amplification units can be combinatorically recruited to represent different motor sequences within the same tree.

Acoustic Bioprinting of Patient-Derived Organoids for Predicting Cancer Therapy Responses

Cancer models, which are biologically representative of patient tumors, can predict the treatment responses and help determine the most appropriate cancer treatment for individual patients. Here, a point-of-care testing system called acoustically bioprinted patient-derived microtissues (PDMs) that can model cancer invasion and predict treatment response in individual patients with colorectal cancer (CRC), is reported. The PDMs are composed of patient-derived colorectal tumors and healthy organoids which can be precisely arranged by acoustic bioprinting approach for recapulating primary tissue's architecture. Particularly, these tumor organoids can be efficiently generated and can apprehend histological, genomic, and phenotypical characteristics of primary tumors. Consequently, these PDMs allow physiologically relevant in vitro drug (5-fluorouracil) screens, thus predicting the paired patient's responses to chemotherapy. A correlation between organoid invasion speed and normalized spreading speed of the paired patients is further established. It provides a quantitative indicator to help doctors make better decisions on ultimate anus-preserving operation for extremely low CRC patients. Thus, by combing acoustic bioprinting and organoid cultures, this method may open an avenue to establish complex 3D tissue models for precision and personalized medicine.

The role of higher order thalamus during learning and correct performance in goal-directed behavior

The thalamus is a gateway to the cortex. Cortical encoding of complex behavior can therefore only be understood by considering the thalamic processing of sensory and internally generated information. Here, we use two-photon Ca²⁺ imaging and optogenetics to investigate the role of axonal projections from the posteromedial nucleus of the thalamus (POm) to the forepaw area of the mouse primary somatosensory cortex (forepaw S1). By recording the activity of POm axonal projections within forepaw S1 during expert and chance performance in two tactile goal-directed tasks, we demonstrate that POm axons increase activity in the response and, to a lesser extent, reward epochs specifically during correct HIT performance. When performing at chance level during learning of a new behavior, POm axonal activity was decreased to naive rates and did not correlate with task performance. However, once evoked, the Ca²⁺ transients were larger than during expert performance, suggesting POm input to S1 differentially encodes chance and expert performance. Furthermore, the POm influences goal-directed behavior, as photoinactivation of archaerhodopsin-expressing neurons in the POm decreased the learning rate and overall success in the behavioral task. Taken together, these findings expand the known roles of the higher-thalamic nuclei, illustrating the POm encodes and influences correct action during learning and performance in a sensory-based goal-directed behavior.

High-frequency burst spiking in layer 5 thick-tufted pyramids of rat primary somatosensory cortex encodes exploratory touch

Diversity of cell-types that collectively shape the cortical microcircuit ensures the necessary computational richness to orchestrate a wide variety of behaviors. The information content embedded in spiking activity of identified cell-types remain unclear to a large extent. Here, we recorded spike responses upon whisker touch of anatomically identified excitatory cell-types in primary somatosensory cortex in naive, untrained rats. We find major differences across layers and cell-types. The temporal structure of spontaneous spiking contains high-frequency bursts (≥100 Hz) in all morphological cell-types but a significant increase upon whisker touch is restricted to layer L5 thick-tufted pyramids (L5tts) and thus provides a distinct neurophysiological signature. We find that whisker touch can also be decoded from L5tt bursting, but not from other cell-types. We observed high-frequency bursts in L5tts projecting to different subcortical regions, including thalamus, midbrain and brainstem. We conclude that bursts in L5tts allow accurate coding and decoding of exploratory whisker touch.

In order to investigate the information encoded by spiking activity in different neuronal cell types in the primary somatosensory cortex, de Kock et al performed electrophysiological recordings in untrained rats. They demonstrated that an increase in high-frequency burst spiking in thick tufted pyramids in layer 5 of the cortex allow accurate encoding of exploratory whisker touch.

Rethinking Single Neuron Electrical Compartmentalization: Dendritic Contributions to Network Computation In Vivo

Decades of experimental and theoretical work support a now well-established theory that active dendritic processing contributes to the computational power of individual neurons. This theory is based on the high degree of electrical compartmentalization observed in the dendrites of single neurons in ex vivo preparations. Compartmentalization allows dendrites to conduct semi-independent operations on their inputs before final integration and output at the axon, producing a "network-in-a-neuron." However, recent in vivo functional imaging experiments in mouse cortex have reported surprisingly little evidence for strong dendritic compartmentalization. In this review, we contextualize these new findings and discuss their impact on the future of the field. Specifically, we consider how highly coordinated, and thus less compartmentalized, activity in soma and dendrites can contribute to cortical computations including nonlinear mixed selectivity, prediction/expectation, multiplexing, and credit assignment.

Neocortical Layer 1: An Elegant Solution to Top-Down and Bottom-Up Integration

Many of our daily activities, such as riding a bike to work or reading a book in a noisy cafe, and highly skilled activities, such as a professional playing a tennis match or a violin concerto, depend upon the ability of the brain to quickly make moment-to-moment adjustments to our behavior in response to the results of our actions. Particularly, they depend upon the ability of the neocortex to integrate the information provided by the sensory organs (bottom-up information) with internally generated signals such as expectations or attentional signals (top-down information). This integration occurs in pyramidal cells (PCs) and their long apical dendrite, which branches extensively into a dendritic tuft in layer 1 (L1). The outermost layer of the neocortex, L1 is highly conserved across cortical areas and species. Importantly, L1 is the predominant input layer for top-down information, relayed by a rich, dense mesh of long-range projections that provide signals to the tuft branches of the PCs. Here, we discuss recent progress in our understanding of the composition of L1 and review evidence that L1 processing contributes to functions such as sensory perception, cross-modal integration, controlling states of consciousness, attention, and learning.

Prolonged anesthesia alters brain synaptic architecture

Prolonged medically induced coma (pMIC) is carried out routinely in intensive care medicine. pMIC leads to cognitive impairment, yet the underlying neuromorphological correlates are still unknown, as no direct studies of MIC exceeding ∼6 h on neural circuits exist. Here, we establish pMIC (up to 24 h) in adolescent and mature mice, and combine longitudinal two-photon imaging of cortical synapses with repeated behavioral object recognition assessments. We find that pMIC affects object recognition, and that it is associated with enhanced synaptic turnover, generated by enhanced synapse formation during pMIC, while the postanesthetic period is dominated by synaptic loss. Our results demonstrate major side effects of prolonged anesthesia on neural circuit structure.

Sequence Learning Induces Selectivity to Multiple Task Parameters in Mouse Somatosensory Cortex

Sequential temporal ordering and patterning are key features of natural signals, used by the brain to decode stimuli and perceive them as sensory objects. To explore how cortical neuronal activity underpins sequence discrimination, we developed a task in which mice distinguished between tactile "word" sequences constructed from distinct vibrations delivered to the whiskers, assembled in different orders. Animals licked to report the presence of the target sequence. Mice could respond to the earliest possible cues allowing discrimination, effectively solving the task as a "detection of change" problem, but enhanced their performance when responding later. Optogenetic inactivation showed that the somatosensory cortex was necessary for sequence discrimination. Two-photon imaging in layer 2/3 of the primary somatosensory "barrel" cortex (S1bf) revealed that, in well-trained animals, neurons had heterogeneous selectivity to multiple task variables including not just sensory input but also the animal's action decision and the trial outcome (presence or absence of the predicted reward). Many neurons were activated preceding goal-directed licking, thus reflecting the animal's learned action in response to the target sequence; these neurons were found as soon as mice learned to associate the rewarded sequence with licking. In contrast, learning evoked smaller changes in sensory response tuning: neurons responding to stimulus features were found in naive mice, and training did not generate neurons with enhanced temporal integration or categorical responses. Therefore, in S1bf, sequence learning results in neurons whose activity reflects the learned association between target sequence and licking rather than a refined representation of sensory features.

A Dendrite-Focused Framework for Understanding the Actions of Ketamine and Psychedelics

Pilot studies have hinted that serotonergic psychedelics such as psilocybin may relieve depression, and could possibly do so by promoting neural plasticity. Intriguingly, another psychotomimetic compound, ketamine, is a fast-acting antidepressant and induces synapse formation. The similarities in behavioral and neural effects have been puzzling because the compounds target distinct molecular receptors in the brain. In this opinion article, we develop a conceptual framework that suggests the actions of ketamine and serotonergic psychedelics may converge at the dendrites, to both enhance and suppress membrane excitability. We speculate that mismatches in the opposing actions on dendritic excitability may relate to these compounds' cell-type and region selectivity, their moderate range of effects and toxicity, and their plasticity-promoting capacities.

The connectome of the adult Drosophila mushroom body provides insights into function

Making inferences about the computations performed by neuronal circuits from synapse-level connectivity maps is an emerging opportunity in neuroscience. The mushroom body (MB) is well positioned for developing and testing such an approach due to its conserved neuronal architecture, recently completed dense connectome, and extensive prior experimental studies of its roles in learning, memory, and activity regulation. Here, we identify new components of the MB circuit in Drosophila, including extensive visual input and MB output neurons (MBONs) with direct connections to descending neurons. We find unexpected structure in sensory inputs, in the transfer of information about different sensory modalities to MBONs, and in the modulation of that transfer by dopaminergic neurons (DANs). We provide insights into the circuitry used to integrate MB outputs, connectivity between the MB and the central complex and inputs to DANs, including feedback from MBONs. Our results provide a foundation for further theoretical and experimental work.

Being superficial: a developmental viewpoint on cortical layer 1 wiring

Functioning of the neocortex relies on a complex architecture of circuits, as illustrated by the causal link between neocortical excitation/inhibition imbalance and the etiology of several neurodevelopmental disorders. An important entry point to cortical circuits is located in the superficial layer 1 (L1), which contains mostly local and long-range inputs and sparse inhibitory interneurons that collectively regulate cerebral functions. While increasing evidence indicates that L1 has important physiological roles, our understanding of how it wires up during development remains limited. Here, we provide an integrated overview of L1 anatomy, function and development, with a focus on transient early born Cajal-Retzius neurons, and highlight open questions key for progressing our understanding of this essential yet understudied layer of the cerebral cortex.

Dynamic perceptual feature selectivity in primary somatosensory cortex upon reversal learning

Neurons in primary sensory cortex encode a variety of stimulus features upon perceptual learning. However, it is unclear whether the acquired stimulus selectivity remains stable when the same input is perceived in a different context. Here, we monitor the activity of individual neurons in the mouse primary somatosensory cortex during reward-based texture discrimination. We track their stimulus selectivity before and after changing reward contingencies, which allows us to identify various classes of neurons. We find neurons that stably represented a texture or the upcoming behavioral choice, but the majority is dynamic. Among those, a subpopulation of neurons regains texture selectivity contingent on the associated reward value. These value-sensitive neurons forecast the onset of learning by displaying a distinct and transient increase in activity, depending on past behavioral experience. Thus, stimulus selectivity of excitatory neurons during perceptual learning is dynamic and largely relies on behavioral contingencies, even in primary sensory cortex.

Do cortical neurons stably represent stimulus features in different contexts? Here, using calcium imaging, the authors show that texture selectivity of individual neurons is dynamic during reversal learning. For a subclass this is contingent on the associated reward and forecasts the onset of learning.

Cell-Type-Specific Outcome Representation in the Primary Motor Cortex

Adaptive movements are critical for animal survival. To guide future actions, the brain monitors various outcomes, including achievement of movement and appetitive goals. The nature of these outcome signals and their neuronal and network realization in the motor cortex (M1), which directs skilled movements, is largely unknown. Using a dexterity task, calcium imaging, optogenetic perturbations, and behavioral manipulations, we studied outcome signals in the murine forelimb M1. We found two populations of layer 2-3 neurons, termed success- and failure-related neurons, that develop with training, and report end results of trials. In these neurons, prolonged responses were recorded after success or failure trials independent of reward and kinematics. In addition, the initial state of layer 5 pyramidal tract neurons contained a memory trace of the previous trial's outcome. Intertrial cortical activity was needed to learn new task requirements. These M1 layer-specific performance outcome signals may support reinforcement motor learning of skilled behavior.

Voltage Imaging of Cortical Oscillations in Layer 1 with Two-Photon Microscopy

Membrane voltage oscillations in layer 1 (L1) of primary sensory cortices might be important indicators of cortical gain control, attentional focusing, and signal integration. However, electric field recordings are hampered by the low seal resistance of electrodes close to the brain surface. To study L1 membrane voltage oscillations, we synthesized a new voltage-sensitive dye, di1-ANNINE (anellated hemicyanine)-6plus, that can diffuse into tissue. We applied it with a new surgery, leaving the dura intact but allowing injection of large quantities of staining solution, and imaged cortical membrane potential oscillations with two-photon microscopy depth-resolved (25-100 μm below dura) in anesthetized and awake mice. We found delta (0.5-4 Hz), theta (4-10 Hz), low beta (10-20 Hz), and low gamma (30-40 Hz) oscillations. All oscillations were stronger in awake animals. While the power of delta, theta, and low beta oscillations increased with depth, the power of low gamma was more constant throughout L1. These findings identify L1 as an important coordination hub for the dynamic binding process of neurons mediated by oscillations.

EZcalcium: Open-Source Toolbox for Analysis of Calcium Imaging Data

Fluorescence calcium imaging using a range of microscopy approaches, such as two-photon excitation or head-mounted “miniscopes,” is one of the preferred methods to record neuronal activity and glial signals in various experimental settings, including acute brain slices, brain organoids, and behaving animals. Because changes in the fluorescence intensity of genetically encoded or chemical calcium indicators correlate with action potential firing in neurons, data analysis is based on inferring such spiking from changes in pixel intensity values across time within different regions of interest. However, the algorithms necessary to extract biologically relevant information from these fluorescent signals are complex and require significant expertise in programming to develop robust analysis pipelines. For decades, the only way to perform these analyses was for individual laboratories to write their custom code. These routines were typically not well annotated and lacked intuitive graphical user interfaces (GUIs), which made it difficult for scientists in other laboratories to adopt them. Although the panorama is changing with recent tools like CaImAn, Suite2P, and others, there is still a barrier for many laboratories to adopt these packages, especially for potential users without sophisticated programming skills. As two-photon microscopes are becoming increasingly affordable, the bottleneck is no longer the hardware, but the software used to analyze the calcium data optimally and consistently across different groups. We addressed this unmet need by incorporating recent software solutions, namely NoRMCorre and CaImAn, for motion correction, segmentation, signal extraction, and deconvolution of calcium imaging data into an open-source, easy to use, GUI-based, intuitive and automated data analysis software package, which we named EZcalcium.

Investigating learning-related neural circuitry with chronic in vivo optical imaging

Fundamental aspects of brain function, including development, plasticity, learning, and memory, can take place over time scales of days to years. Chronic in vivo imaging of neural activity with cellular resolution is a powerful method for tracking the long-term activity of neural circuits. We review recent advances in our understanding of neural circuit function from diverse brain regions that have been enabled by chronic in vivo cellular imaging. Insight into the neural basis of learning and decision-making, in particular, benefit from the ability to acquire longitudinal data from genetically identified neuronal populations, deep brain areas, and subcellular structures. We propose that combining chronic imaging with further experimental and computational innovations will advance our understanding of the neural circuit mechanisms of brain function.

Activity-dependent death of transient Cajal-Retzius neurons is required for functional cortical wiring

Programmed cell death and early activity contribute to the emergence of functional cortical circuits. While most neuronal populations are scaled-down by death, some subpopulations are entirely eliminated, raising the question of the importance of such demise for cortical wiring. Here, we addressed this issue by focusing on Cajal-Retzius neurons (CRs), key players in cortical development that are eliminated in postnatal mice in part via Bax-dependent apoptosis. Using Bax-conditional mutants and CR hyperpolarization, we show that the survival of electrically active subsets of CRs triggers an increase in both dendrite complexity and spine density of upper layer pyramidal neurons, leading to an excitation/inhibition imbalance. The survival of these CRs is induced by hyperpolarization, highlighting an interplay between early activity and neuronal elimination. Taken together, our study reveals a novel activity-dependent programmed cell death process required for the removal of transient immature neurons and the proper wiring of functional cortical circuits.

Continual Learning in a Multi-Layer Network of an Electric Fish

Distributing learning across multiple layers has proven extremely powerful in artificial neural networks. However, little is known about how multi-layer learning is implemented in the brain. Here, we provide an account of learning across multiple processing layers in the electrosensory lobe (ELL) of mormyrid fish and report how it solves problems well known from machine learning. Because the ELL operates and learns continuously, it must reconcile learning and signaling functions without switching its mode of operation. We show that this is accomplished through a functional compartmentalization within intermediate layer neurons in which inputs driving learning differentially affect dendritic and axonal spikes. We also find that connectivity based on learning rather than sensory response selectivity assures that plasticity at synapses onto intermediate-layer neurons is matched to the requirements of output neurons. The mechanisms we uncover have relevance to learning in the cerebellum, hippocampus, and cerebral cortex, as well as in artificial systems.

A deep learning framework for neuroscience

Systems neuroscience seeks explanations for how the brain implements a wide variety of perceptual, cognitive and motor tasks. Conversely, artificial intelligence attempts to design computational systems based on the tasks they will have to solve. In artificial neural networks, the three components specified by design are the objective functions, the learning rules and the architectures. With the growing success of deep learning, which utilizes brain-inspired architectures, these three designed components have increasingly become central to how we model, engineer and optimize complex artificial learning systems. Here we argue that a greater focus on these components would also benefit systems neuroscience. We give examples of how this optimization-based framework can drive theoretical and experimental progress in neuroscience. We contend that this principled perspective on systems neuroscience will help to generate more rapid progress.