A Sensorimotor Circuit in Mouse Cortex for Visual Flow Predictions

Sensory- and Motor-Related Responses of Layer 1 Neurons in the Mouse Visual Cortex

Cortical layer 1 (L1) contains a sparse and molecularly distinct population of inhibitory interneurons. Their location makes them ideally suited for affecting computations involving long-range corticocortical and subcortical inputs, yet their response properties remain largely unexplored. Here we attempt to characterize some of the functional properties of these neurons in the primary visual cortex of awake mice. We find that the strongest driver of L1 neuron activity is locomotion, with at least half of L1 neurons displaying locomotion-related activity. Visual responses are present in a similar fraction of neurons, but these responses are weaker and frequently suppressive. We also find that ∼43% of L1 neurons respond to noise stimuli and at least 14% respond to whisker touch, with these two populations being statistically independent. Finally, we find that 45% of L1 neurons have generally weak responses correlated with whisking activity. Overall, the spatial distributions of modality-specific responses were more or less random. Our work helps to establish the basic sensory- and motor-related responses of L1 interneurons, revealing several previously unreported characteristics.SIGNIFICANCE STATEMENT Cortical processing even in primary sensory areas is strongly influenced by nonlocal corticocortical and neuromodulatory inputs. Many of these inputs are known to converge onto layer 1 where they target not only distal dendrites of pyramidal neurons but also a sparse population of inhibitory neurons. Previous studies have suggested that layer 1 neurons may play a crucial role in mediating the effects of these long-range projections, but the different types of inputs have mostly been studied in isolation. Here, we take a closer look at the response properties of layer 1 neurons in mouse visual cortex, examining both their visual properties, likely caused by direct thalamocortical inputs, and other sensory and motor properties, likely reflecting corticocortical and neuromodulatory inputs.

Distinct Waking States for Strong Evoked Responses in Primary Visual Cortex and Optimal Visual Detection Performance

Variability in cortical neuronal responses to sensory stimuli and in perceptual decision making performance is substantial. Moment-to-moment fluctuations in waking state or arousal can account for much of this variability. Yet, this variability is rarely characterized across the full spectrum of waking states, leaving the characteristics of the optimal state for sensory processing unresolved. Using pupillometry in concert with extracellular multiunit and intracellular whole-cell recordings, we found that the magnitude and reliability of visually evoked responses in primary visual cortex (V1) of awake, passively behaving male mice increase as a function of arousal and are largest during sustained locomotion periods. During these high-arousal, sustained locomotion periods, cortical neuronal membrane potential was at its most depolarized and least variable. Contrastingly, behavioral performance of mice on two distinct visual detection tasks was generally best at a range of intermediate arousal levels, but worst during high arousal with locomotion. These results suggest that large, reliable responses to visual stimuli in V1 occur at a distinct arousal level from that associated with optimal visual detection performance. Our results clarify the relation between neuronal responsiveness and the continuum of waking states, and suggest new complexities in the relation between primary sensory cortical activity and behavior.SIGNIFICANCE STATEMENT Cortical sensory processing strongly depends on arousal. In the mouse visual system, locomotion (associated with high arousal) has previously been shown to enhance the sensory responses of neurons in primary visual cortex (V1). Yet, arousal fluctuates on a moment-to-moment basis, even during quiescent periods. The characteristics of V1 sensory processing across the continuum of arousal are unclear. Furthermore, the arousal level corresponding to optimal visual detection performance is unknown. We show that the magnitude and reliability of sensory-evoked V1 responses are monotonic increasing functions of arousal, and largest during locomotion. Visual detection behavior, however, is suboptimal during high arousal with locomotion, and usually best during intermediate arousal. Our study provides a more complete picture of the dependence of V1 sensory processing on arousal.

Top-Down Suppression of Sensory Cortex in an NMDAR Hypofunction Model of Psychosis

Conceptual and computational models have been advanced that propose that perceptual disturbances in psychosis, such as hallucinations, may arise due to a disruption in the balance between bottom-up (ie sensory) and top-down (ie from higher brain areas) information streams in sensory cortex. However, the neural activity underlying this hypothesized alteration remains largely unexplored. Pharmacological N-methyl-d-aspartate receptor (NMDAR) antagonism presents an attractive model to examine potential changes as it acutely recapitulates many of the symptoms of schizophrenia including hallucinations, and NMDAR hypofunction is strongly implicated in the pathogenesis of schizophrenia as evidenced by large-scale genetic studies. Here we use in vivo 2-photon imaging to measure frontal top-down signals from the anterior cingulate cortex (ACC) and their influence on activity of the primary visual cortex (V1) in mice during pharmacologically induced NMDAR hypofunction. We find that global NMDAR hypofunction causes a significant increase in activation of top-down ACC axons, and that surprisingly this is associated with an ACC-dependent net suppression of spontaneous activity in V1 as well as a reduction in V1 sensory-evoked activity. These findings are consistent with a model in which perceptual disturbances in psychosis are caused in part by aberrant top-down frontal cortex activity that suppresses the transmission of sensory signals through early sensory areas.

Learning, memory and consolidation mechanisms for behavioral control in hierarchically organized cortico-basal ganglia systems

This article aims to provide a synthesis on the question how brain structures cooperate to accomplish hierarchically organized behaviors, characterized by low‐level, habitual routines nested in larger sequences of planned, goal‐directed behavior. The functioning of a connected set of brain structures—prefrontal cortex, hippocampus, striatum, and dopaminergic mesencephalon—is reviewed in relation to two important distinctions: (a) goal‐directed as opposed to habitual behavior and (b) model‐based and model‐free learning. Recent evidence indicates that the orbitomedial prefrontal cortices not only subserve goal‐directed behavior and model‐based learning, but also code the “landscape” (task space) of behaviorally relevant variables. While the hippocampus stands out for its role in coding and memorizing world state representations, it is argued to function in model‐based learning but is not required for coding of action–outcome contingencies, illustrating that goal‐directed behavior is not congruent with model‐based learning. While the dorsolateral and dorsomedial striatum largely conform to the dichotomy between habitual versus goal‐directed behavior, ventral striatal functions go beyond this distinction. Next, we contextualize findings on coding of reward‐prediction errors by ventral tegmental dopamine neurons to suggest a broader role of mesencephalic dopamine cells, viz. in behavioral reactivity and signaling unexpected sensory changes. We hypothesize that goal‐directed behavior is hierarchically organized in interconnected cortico‐basal ganglia loops, where a limbic‐affective prefrontal‐ventral striatal loop controls action selection in a dorsomedial prefrontal–striatal loop, which in turn regulates activity in sensorimotor‐dorsolateral striatal circuits. This structure for behavioral organization requires alignment with mechanisms for memory formation and consolidation. We propose that frontal corticothalamic circuits form a high‐level loop for memory processing that initiates and temporally organizes nested activities in lower‐level loops, including the hippocampus and the ripple‐associated replay it generates. The evidence on hierarchically organized behavior converges with that on consolidation mechanisms in suggesting a frontal‐to‐caudal directionality in processing control.

V1 microcircuits underlying mouse visual behavior

Visual behavior is based on the concerted activity of neurons in visual areas, where sensory signals are integrated with top-down information. In the past decade, the advent of new tools, such as functional imaging of populations of identified single neurons, high-density electrophysiology, virus-assisted circuit mapping, and precisely timed, cell-type specific manipulations, has advanced our understanding of the neuronal microcircuits underlying visual behavior. Studies in head-fixed mice, where such tools can routinely be applied, begin to provide new insights into the neural code of primary visual cortex (V1) underlying visual perception, and the micro-circuits of attention, predictive processing, and learning.

Structural transition in the collective behavior of cognitive agents

Living organisms process information to interact and adapt to their surroundings with the goal of finding food, mating, or averting hazards. The structure of their environment has profound repercussions through both selecting their internal architecture and also inducing adaptive responses to environmental cues and stimuli. Adaptive collective behavior underpinned by specialized optimization strategies is ubiquitous in the natural world. We develop a minimal model of agents that explore their environment by means of sampling trajectories. The spatial information stored in the sampling trajectories is our minimal definition of a cognitive map. We find that, as cognitive agents build and update their internal, cognitive representation of the causal structure of their environment, complex patterns emerge in the system, where the onset of pattern formation relates to the spatial overlap of cognitive maps. Exchange of information among the agents leads to an order-disorder transition. As a result of the spontaneous breaking of translational symmetry, a Goldstone mode emerges, which points at a collective mechanism of information transfer among cognitive organisms. These findings may be generally applicable to the design of decentralized, artificial-intelligence swarm systems.

Towards a Unified View on Pathways and Functions of Neural Recurrent Processing

There are three neural feedback pathways to the primary visual cortex (V1): corticocortical, pulvinocortical, and cholinergic. What are the respective functions of these three projections? Possible functions range from contextual modulation of stimulus processing and feedback of high-level information to predictive processing (PP). How are these functions subserved by different pathways and can they be integrated into an overarching theoretical framework? We propose that corticocortical and pulvinocortical connections are involved in all three functions, whereas the role of cholinergic projections is limited by their slow response to stimuli. PP provides a broad explanatory framework under which stimulus-context modulation and high-level processing are subsumed, involving multiple feedback pathways that provide mechanisms for inferring and interpreting what sensory inputs are about.

A Head-Mounted Camera System Integrates Detailed Behavioral Monitoring with Multichannel Electrophysiology in Freely Moving Mice

Breakthroughs in understanding the neural basis of natural behavior require neural recording and intervention to be paired with high-fidelity multimodal behavioral monitoring. An extensive genetic toolkit for neural circuit dissection, and well-developed neural recording technology, make the mouse a powerful model organism for systems neuroscience. However, most methods for high-bandwidth acquisition of behavioral data in mice rely upon fixed-position cameras and other off-animal devices, complicating the monitoring of animals freely engaged in natural behaviors. Here, we report the development of a lightweight head-mounted camera system combined with head-movement sensors to simultaneously monitor eye position, pupil dilation, whisking, and pinna movements along with head motion in unrestrained, freely behaving mice. The power of the combined technology is demonstrated by observations linking eye position to head orientation; whisking to non-tactile stimulation; and, in electrophysiological experiments, visual cortical activity to volitional head movements.

•

Eyes, whiskers, head, and neural activity monitored in freely moving mice

•

System generates stable video output and leaves mouse behavior largely unchanged

•

Close link between eye and head movements at both slow and fast timescales

•

Active head movements in the dark strongly modulate primary visual cortex activity

Internally Generated Predictions Enhance Neural and Behavioral Detection of Sensory Stimuli in an Electric Fish

Studies of cerebellum-like circuits in fish have demonstrated that synaptic plasticity shapes the motor corollary discharge responses of granule cells into highly-specific predictions of self-generated sensory input. However, the functional significance of such predictions, known as negative images, has not been directly tested. Here we provide evidence for improvements in neural coding and behavioral detection of prey-like stimuli due to negative images. In addition, we find that manipulating synaptic plasticity leads to specific changes in circuit output that disrupt neural coding and detection of prey-like stimuli. These results link synaptic plasticity, neural coding, and behavior and also provide a circuit-level account of how combining external sensory input with internally generated predictions enhances sensory processing.

The Role of Neuronal Oscillations in Visual Active Sensing

Visual perception is most often studied as a "passive" process in which an observer fixates steadily at point in space so that stimuli can be delivered to the system with spatial precision. Analysis of neuronal signals related to vision is generally keyed to stimulus onset, stimulus movement, etc.; i.e., events external to the observer. In natural "active" vision, however, information is systematically acquired by using eye movements including rapid (saccadic) eye movements, as well as smooth ocular pursuit of moving objects and slower drifts. Here we consider the use of alternating saccades and fixations to gather information from a visual scene. The underlying motor sampling plan contains highly reliable information regarding "where" and "when" the eyes will land, this information can be used predictively to modify firing properties of neurons precisely at the time when this "contextual" information is most useful - when a volley of retinal input enters the system at the onset of each fixation. Analyses focusing on neural events leading to and resulting from shifts in fixation, as well as visual events external to the observer, can provide a more complete and mechanistic understanding of visual information processing. Studies thus far suggest that active vision may be a fundamentally different from that process we usually study with more traditional passive viewing paradigms. In this Perspective we note that active saccadic sampling behavior imposes robust temporal patterning on the activity of neuron ensembles and large-scale neural dynamics throughout the brain's visual pathways whose mechanistic effects on information processing are not yet fully understood. The spatio-temporal sequence of eye movements elicits a succession of temporally predictable quasi-rhythmic sensory inputs, whose encoding is enhanced by entrainment of low frequency oscillations to the rate of eye movements. Review of the pertinent findings underscores the fact that temporal coordination between motor and visual cortices is critical for understanding neural dynamics of active vision and posits that phase entrainment of neuronal oscillations plays a mechanistic role in this process.

From Hiring to Firing: Activation of Inhibitory Neurons and Their Recruitment in Behavior

The investigation of GABAergic inhibitory circuits has substantially expanded over the past few years. The development of new tools and technology has allowed investigators to classify many diverse groups of inhibitory neurons by several delineating factors: these include their connectivity motifs, expression of specific molecular markers, receptor diversity, and ultimately their role in brain function. Despite this progress, however, there is still limited understanding of how GABAergic neurons are recruited by their input and how their activity is modulated by behavioral states. This limitation is primarily due to the fact that studies of GABAergic inhibition are mainly geared toward determining how, once activated, inhibitory circuits regulate the activity of excitatory neurons. In this review article, we will outline recent work investigating the anatomical and physiological properties of inputs that activate cortical GABAergic neurons, and discuss how these inhibitory cells are differentially recruited during behavior.

Fitting predictive coding to the neurophysiological data

Recent neurophysiological data showing the effects of locomotion on neural activity in mouse primary visual cortex has been interpreted as providing strong support for the predictive coding account of cortical function. Specifically, this work has been interpreted as providing direct evidence that prediction-error, a distinguishing property of predictive coding, is encoded in cortex. This article evaluates these claims and highlights some of the discrepancies between the proposed predictive coding model and the neuro-biology. Furthermore, it is shown that the model can be modified so as to fit the empirical data more successfully.

Layer-specific integration of locomotion and sensory information in mouse barrel cortex

During navigation, rodents continually sample the environment with their whiskers. How locomotion modulates neuronal activity in somatosensory cortex, and how it is integrated with whisker-touch remains unclear. Here, we compared neuronal activity in layer 2/3 (L2/3) and L5 of barrel cortex using calcium imaging in mice running in a tactile virtual reality. Both layers increase their activity during running and concomitant whisking, in the absence of touch. Fewer neurons are modulated by whisking alone. Whereas L5 neurons respond transiently to wall-touch during running, L2/3 neurons show sustained activity. Consistently, neurons encoding running-with-touch are more abundant in L2/3 and they encode the run-speed better during touch. Few neurons across layers were also sensitive to abrupt perturbations of tactile flow during running. In summary, locomotion significantly enhances barrel cortex activity across layers with L5 neurons mainly reporting changes in touch conditions and L2/3 neurons continually integrating tactile stimuli with running.

The influence of locomotion on somatosensory processing in barrel cortex is not well understood. Here the authors report distinct layer-specific responses, with L5 primarily reporting changes in touch condition while L2/3 neurons integrating touch and locomotion continuously.

Timing Is of the Essence: Improvement in Perception During Active Sensing

Active sensing refers to the concept of animals perceiving their environment while involving self-initiated motor acts. As a consequence of these motor acts, this activity produces direct and timely changes in the sensory surface. Is the brain able to take advantage of the precise time-locking that occurs during active sensing? Is the intrinsic predictability present during active sensing, impacting the sensory processes? We conjecture that if stimuli presentation is evoked by a self-initiated motor act, sensory discrimination and timing accuracy would improve. We studied this phenomenon when rats had to locate the position of a brief light stimulus, either when it was elicited by a warning light [passive condition (PC)] or when it was generated by a lever press [active condition (AC)]. We found that during the PC, rats had 66% of correct responses, vs. a significantly higher 77% of correct responses in AC. Furthermore, reaction times reduced from 1,181 ms during AC to 816 ms during PC For the latter condition, the probability of detecting the side of the light stimulus was negatively correlated with the time lag between the motor act and the evoked light and with a 38% reduction on performance per second of delay. These experiment shows that the mechanism that underlies sensory improvement during active behaviors have a constrained time dynamic, where the peak performances occur during the motor act, decreasing proportionally to the lag between the motor act and the stimulus presentation. This result is consistent with the evidence already found in humans, of a precise time dynamic of the improvement of sensory acuity after a motor act and reveals an equivalent process in rodents. Our results support the idea that perception and action are precisely coordinated in the brain.

In Vivo Submillisecond Two-Photon Optogenetics with Temporally Focused Patterned Light

To better examine circuit mechanisms underlying perception and behavior, researchers need tools to enable temporally precise control of action-potential generation of individual cells from neuronal ensembles. Here we demonstrate that such precision can be achieved with two-photon (2P) temporally focused computer-generated holography to control neuronal excitability at the supragranular layers of anesthetized and awake visual cortex in both male and female mice. Using 2P-guided whole-cell or cell-attached recordings in positive neurons expressing any of the three opsins ReaChR, CoChR, or ChrimsonR, we investigated the dependence of spiking activity on the opsin's channel kinetics. We found that in all cases the use of brief illumination (≤10 ms) induces spikes of millisecond temporal resolution and submillisecond precision, which were preserved upon repetitive illuminations up to tens of hertz. To reach high temporal precision, we used a large illumination spot covering the entire cell body and an amplified laser at high peak power and low excitation intensity (on average ≤0.2 mW/μm2), thus minimizing the risk for nonlinear photodamage effects. Finally, by combining 2P holographic excitation with electrophysiological recordings and calcium imaging using GCaMP6s, we investigated the factors, including illumination shape and intensity, opsin distribution in the target cell, and cell morphology, which affect the spatial selectivity of single-cell and multicell holographic activation. Parallel optical control of neuronal activity with cellular resolution and millisecond temporal precision should make it easier to investigate neuronal connections and find further links between connectivity, microcircuit dynamics, and brain functions.SIGNIFICANCE STATEMENT Recent developments in the field of optogenetics has enabled researchers to probe the neuronal microcircuit with light by optically actuating genetically encoded light-sensitive opsins expressed in the target cells. Here, we applied holographic light shaping and temporal focusing to simultaneously deliver axially confined holographic patterns to opsin-positive cells in the living mouse cortex. Parallel illumination efficiently induced action potentials with high temporal resolution and precision for three opsins of different kinetics. We extended the parallel optogenetic activation at low intensity to multiple neurons and concurrently monitored their calcium dynamics. These results demonstrate fast and temporally precise in vivo control of a neuronal subpopulation, opening new opportunities for revealing circuit mechanisms underlying brain functions.

Locomotion-dependent remapping of distributed cortical networks

The interactions between neocortical areas are fluid and state-dependent, but how individual neurons couple to cortex-wide network dynamics remains poorly understood. We correlated the spiking of neurons in primary visual (V1) and retrosplenial (RSP) cortex to activity across dorsal cortex, recorded simultaneously by widefield calcium imaging. Neurons were correlated with distinct and reproducible patterns of activity across the cortical surface; while some fired predominantly with their local area, others coupled to activity in distal areas. The extent of distal coupling was predicted by how strongly neurons correlated with the local network. Changes in brain state triggered by locomotion strengthened affiliations of V1 neurons with higher visual and motor areas, while strengthening distal affiliations of RSP neurons with sensory cortices. Thus, the diverse coupling of individual neurons to cortex-wide activity patterns is restructured by running in an area-specific manner, resulting in a shift in the mode of cortical processing during locomotion.

Locations in the Neocortex: A Theory of Sensorimotor Object Recognition Using Cortical Grid Cells

The neocortex is capable of anticipating the sensory results of movement but the neural mechanisms are poorly understood. In the entorhinal cortex, grid cells represent the location of an animal in its environment, and this location is updated through movement and path integration. In this paper, we propose that sensory neocortex incorporates movement using grid cell-like neurons that represent the location of sensors on an object. We describe a two-layer neural network model that uses cortical grid cells and path integration to robustly learn and recognize objects through movement and predict sensory stimuli after movement. A layer of cells consisting of several grid cell-like modules represents a location in the reference frame of a specific object. Another layer of cells which processes sensory input receives this location input as context and uses it to encode the sensory input in the object's reference frame. Sensory input causes the network to invoke previously learned locations that are consistent with the input, and motor input causes the network to update those locations. Simulations show that the model can learn hundreds of objects even when object features alone are insufficient for disambiguation. We discuss the relationship of the model to cortical circuitry and suggest that the reciprocal connections between layers 4 and 6 fit the requirements of the model. We propose that the subgranular layers of cortical columns employ grid cell-like mechanisms to represent object specific locations that are updated through movement.

The power of predictions: An emerging paradigm for psychological research

The last two decades of neuroscience research has produced a growing number of studies that suggest the various psychological phenomena are produced by predictive processes in the brain. When considered together, these studies form a coherent, neurobiologically-inspired research program for guiding psychological research about the mind and behavior. In this paper, we briefly consider the common assumptions and hypotheses that unify an emerging framework and discuss its ramifications, both for improving the replicability and robustness of psychological research and for innovating psychological theory by suggesting an alternative ontology of the human mind.

Opposing Influence of Sensory and Motor Cortical Input on Striatal Circuitry and Choice Behavior

The striatum is the main input nucleus of the basal ganglia and is a key site of sensorimotor integration. While the striatum receives extensive excitatory afferents from the cerebral cortex, the influence of different cortical areas on striatal circuitry and behavior is unknown. Here, we find that corticostriatal inputs from whisker-related primary somatosensory (S1) and motor (M1) cortex differentially innervate projection neurons and interneurons in the dorsal striatum and exert opposing effects on sensory-guided behavior. Optogenetic stimulation of S1-corticostriatal afferents in ex vivo recordings produced larger postsynaptic potentials in striatal parvalbumin (PV)-expressing interneurons than D1- or D2-expressing spiny projection neurons (SPNs), an effect not observed for M1-corticostriatal afferents. Critically, in vivo optogenetic stimulation of S1-corticostriatal afferents produced task-specific behavioral inhibition, which was bidirectionally modulated by striatal PV interneurons. Optogenetic stimulation of M1 afferents produced the opposite behavioral effect. Thus, our results suggest opposing roles for sensory and motor cortex in behavioral choice via distinct influences on striatal circuitry.

Theories of Error Back-Propagation in the Brain

This review article summarises recently proposed theories on how neural circuits in the brain could approximate the error back-propagation algorithm used by artificial neural networks. Computational models implementing these theories achieve learning as efficient as artificial neural networks, but they use simple synaptic plasticity rules based on activity of presynaptic and postsynaptic neurons. The models have similarities, such as including both feedforward and feedback connections, allowing information about error to propagate throughout the network. Furthermore, they incorporate experimental evidence on neural connectivity, responses, and plasticity. These models provide insights on how brain networks might be organised such that modification of synaptic weights on multiple levels of cortical hierarchy leads to improved performance on tasks.

A Framework for Intelligence and Cortical Function Based on Grid Cells in the Neocortex

How the neocortex works is a mystery. In this paper we propose a novel framework for understanding its function. Grid cells are neurons in the entorhinal cortex that represent the location of an animal in its environment. Recent evidence suggests that grid cell-like neurons may also be present in the neocortex. We propose that grid cells exist throughout the neocortex, in every region and in every cortical column. They define a location-based framework for how the neocortex functions. Whereas grid cells in the entorhinal cortex represent the location of one thing, the body relative to its environment, we propose that cortical grid cells simultaneously represent the location of many things. Cortical columns in somatosensory cortex track the location of tactile features relative to the object being touched and cortical columns in visual cortex track the location of visual features relative to the object being viewed. We propose that mechanisms in the entorhinal cortex and hippocampus that evolved for learning the structure of environments are now used by the neocortex to learn the structure of objects. Having a representation of location in each cortical column suggests mechanisms for how the neocortex represents object compositionality and object behaviors. It leads to the hypothesis that every part of the neocortex learns complete models of objects and that there are many models of each object distributed throughout the neocortex. The similarity of circuitry observed in all cortical regions is strong evidence that even high-level cognitive tasks are learned and represented in a location-based framework.

More than Just a "Motor": Recent Surprises from the Frontal Cortex

Motor and premotor cortices are crucial for the control of movements. However, we still know little about how these areas contribute to higher-order motor control, such as deciding which movements to make and when to make them. Here we focus on rodent studies and review recent findings, which suggest that-in addition to motor control-neurons in motor cortices play a role in sensory integration, behavioral strategizing, working memory, and decision-making. We suggest that these seemingly disparate functions may subserve an evolutionarily conserved role in sensorimotor cognition and that further study of rodent motor cortices could make a major contribution to our understanding of the evolution and function of the mammalian frontal cortex.

Reconstruction of Intratelencephalic Neurons in the Mouse Secondary Motor Cortex Reveals the Diverse Projection Patterns of Single Neurons

The secondary motor cortex (MOs) plays crucial roles in cognitive and executive processes and has reciprocal connections with numerous cortices in rodents. However, descriptions of the neuronal morphologies and projection patterns of the MOs at the level of a single neuron are lacking, severely hindering the comprehensive understanding of the wiring diagram of the MOs. Herein, we used a Cre-dependent adeno-associated virus (AAV) to fluorescently label ~80 pyramidal neurons nearby or in the MOs and acquired an uninterrupted whole-brain 3D dataset at a voxel resolution of 0.2 × 0.2 × 1 μm with a whole-brain fluorescence imaging system (fMOST). Based on our 3D dataset, we reconstructed the complete morphologies of 36 individual intratelencephalic (IT) neurons nearby or in the MOs and analyzed the projection patterns and projection strengths of these neurons at a single-neuron level based on several parameters, including the projection areas, the total number of branches, the fiber length, and the total number of terminal tips. We obtained a neuron with an axonal length of 318.43 mm, which is by far the longest reported axonal length. Our results show that all individual neurons in the MOs, regardless of whether they are located in layer 2/3 or layer 5, display diverse projection patterns and projection strengths, implying that these neurons might be involved in different brain circuits at different intensities. The results lay a solid foundation for exploring the relationship between neuronal morphologies and behavioral functions of the MOs at the level of a single neuron.

Predictive Processing: A Canonical Cortical Computation

This perspective describes predictive processing as a computational framework for understanding cortical function in the context of emerging evidence, with a focus on sensory processing. We discuss how the predictive processing framework may be implemented at the level of cortical circuits and how its implementation could be falsified experimentally. Lastly, we summarize the general implications of predictive processing on cortical function in healthy and diseased states.

Sensory Processing Across Conscious and Nonconscious Brain States: From Single Neurons to Distributed Networks for Inferential Representation

Neuronal activity is markedly different across brain states: it varies from desynchronized activity during wakefulness to the synchronous alternation between active and silent states characteristic of deep sleep. Surprisingly, limited attention has been paid to investigating how brain states affect sensory processing. While it was long assumed that the brain was mostly disconnected from external stimuli during sleep, an increasing number of studies indicates that sensory stimuli continue to be processed across all brain states-albeit differently. In this review article, we first discuss what constitutes a brain state. We argue that-next to global, behavioral states such as wakefulness and sleep-there is a concomitant need to distinguish bouts of oscillatory dynamics with specific global/local activity patterns and lasting for a few hundreds of milliseconds, as these can lead to the same sensory stimulus being either perceived or not. We define these short-lasting bouts as micro-states. We proceed to characterize how sensory-evoked neural responses vary between conscious and nonconscious states. We focus on two complementary aspects: neuronal ensembles and inter-areal communication. First, we review which features of ensemble activity are conducive to perception, and how these features vary across brain states. Properties such as heterogeneity, sparsity and synchronicity in neuronal ensembles will especially be considered as essential correlates of conscious processing. Second, we discuss how inter-areal communication varies across brain states and how this may affect brain operations and sensory processing. Finally, we discuss predictive coding (PC) and the concept of multi-level representations as a key framework for understanding conscious sensory processing. In this framework the brain implements conscious representations as inferences about world states across multiple representational levels. In this representational hierarchy, low-level inference may be carried out nonconsciously, whereas high levels integrate across different sensory modalities and larger spatial scales, correlating with conscious processing. This inferential framework is used to interpret several cellular and population-level findings in the context of brain states, and we briefly compare its implications to two other theories of consciousness. In conclusion, this review article, provides foundations to guide future studies aiming to uncover the mechanisms of sensory processing and perception across brain states.

A cortical filter that learns to suppress the acoustic consequences of movement

Sounds can arise from the environment and also predictably from many of our own movements, such as vocalizing, walking, or playing music. The capacity to anticipate these movement-related (reafferent) sounds and distinguish them from environmental sounds is essential for normal hearing^1,2, but the neural circuits that learn to anticipate the often arbitrary and changeable sounds that result from our movements remain largely unknown. Here we developed an acoustic virtual reality (aVR) system in which a mouse learned to associate a novel sound with its locomotor movements, allowing us to identify the neural circuit mechanisms that learn to suppress reafferent sounds and to probe the behavioural consequences of this predictable sensorimotor experience. We found that aVR experience gradually and selectively suppressed auditory cortical responses to the reafferent frequency, in part by strengthening motor cortical activation of auditory cortical inhibitory neurons that respond to the reafferent tone. This plasticity is behaviourally adaptive, as aVR-experienced mice showed an enhanced ability to detect non-reafferent tones during movement. Together, these findings describe a dynamic sensory filter that involves motor cortical inputs to the auditory cortex that can be shaped by experience to selectively suppress the predictable acoustic consequences of movement.

Sensorimotor integration in Caenorhabditis elegans: a reappraisal towards dynamic and distributed computations

The nematode Caenorhabditis elegans is a tractable model system to study locomotion, sensory navigation and decision-making. In its natural habitat, it is thought to navigate complex multisensory environments in order to find food and mating partners, while avoiding threats like predators or toxic environments. While research in past decades has shed much light on the functions and mechanisms of selected sensory neurons, we are just at the brink of understanding how sensory information is integrated by interneuron circuits for action selection in the worm. Recent technological advances have enabled whole-brain Ca2+ imaging and Ca2+ imaging of neuronal activity in freely moving worms. A common principle emerging across multiple studies is that most interneuron activities are tightly coupled to the worm's instantaneous behaviour; notably, these observations encompass neurons receiving direct sensory neuron inputs. The new findings suggest that in the C. elegans brain, sensory and motor representations are integrated already at the uppermost sensory processing layers. Moreover, these results challenge a perhaps more intuitive view of sequential feed-forward sensory pathways that converge onto premotor interneurons and motor neurons. We propose that sensorimotor integration occurs rather in a distributed dynamical fashion. In this perspective article, we will explore this view, discuss the challenges and implications of these discoveries on the interpretation and design of neural activity experiments, and discuss possible functions. Furthermore, we will discuss the broader context of similar findings in fruit flies and rodents, which suggest generalizable principles that can be learnt from this amenable nematode model organism.This article is part of a discussion meeting issue 'Connectome to behaviour: modelling C. elegans at cellular resolution'.

Dendritic solutions to the credit assignment problem

Guaranteeing that synaptic plasticity leads to effective learning requires a means for assigning credit to each neuron for its contribution to behavior. The 'credit assignment problem' refers to the fact that credit assignment is non-trivial in hierarchical networks with multiple stages of processing. One difficulty is that if credit signals are integrated with other inputs, then it is hard for synaptic plasticity rules to distinguish credit-related activity from non-credit-related activity. A potential solution is to use the spatial layout and non-linear properties of dendrites to distinguish credit signals from other inputs. In cortical pyramidal neurons, evidence hints that top-down feedback signals are integrated in the distal apical dendrites and have a distinct impact on spike-firing and synaptic plasticity. This suggests that the distal apical dendrites of pyramidal neurons help the brain to solve the credit assignment problem.

The Impact of Visual Cues, Reward, and Motor Feedback on the Representation of Behaviorally Relevant Spatial Locations in Primary Visual Cortex

The integration of visual stimuli and motor feedback is critical for successful visually guided navigation. These signals have been shown to shape neuronal activity in the primary visual cortex (V1), in an experience-dependent manner. Here, we examined whether visual, reward, and self-motion-related inputs are integrated in order to encode behaviorally relevant locations in V1 neurons. Using a behavioral task in a virtual environment, we monitored layer 2/3 neuronal activity as mice learned to locate a reward along a linear corridor. With learning, a subset of neurons became responsive to the expected reward location. Without a visual cue to the reward location, both behavioral and neuronal responses relied on self-motion-derived estimations. However, when visual cues were available, both neuronal and behavioral responses were driven by visual information. Therefore, a population of V1 neurons encode behaviorally relevant spatial locations, based on either visual cues or on self-motion feedback when visual cues are absent.

Cholinergic modulation of spatial learning, memory and navigation

Spatial learning, including encoding and retrieval of spatial memories as well as holding spatial information in working memory generally serving navigation under a broad range of circumstances, relies on a network of structures. While central to this network are medial temporal lobe structures with a widely appreciated crucial function of the hippocampus, neocortical areas such as the posterior parietal cortex and the retrosplenial cortex also play essential roles. Since the hippocampus receives its main subcortical input from the medial septum of the basal forebrain (BF) cholinergic system, it is not surprising that the potential role of the septo-hippocampal pathway in spatial navigation has been investigated in many studies. Much less is known of the involvement in spatial cognition of the parallel projection system linking the posterior BF with neocortical areas. Here we review the current state of the art of the division of labour within this complex 'navigation system', with special focus on how subcortical cholinergic inputs may regulate various aspects of spatial learning, memory and navigation.

Top-down control: A unified principle of cortical learning

Cognitive control of the brain flexibly maps incoming sensory information onto execution of actions appropriate for the current goal. Learning is a process that enables the brain to estimate current states of the world by extracting its spatiotemporal structure and generate goal-directed motor outputs through selective association of events or movement refinement. Accumulating evidence suggests that top-down control from higher-order brain areas modulates downstream neural activity and changes local computations that are critical for the execution of learned behavior. Recent technological advances in multi-site recordings and optogenetic approaches are beginning to reveal more direct evidence of top-down cognitive control by monitoring and perturbing activity of top-down inputs and observing its causal consequences on behavior and downstream neural dynamics. Here I highlight that learning-related changes in neural circuits in distinct domains of learning converge onto a unified principle; namely recruitment of top-down control whether it involves sensory, motor or offline learning. Recruitment of top-down control may reflect experience-dependent adaptation and integration of internal models for refined state estimation and goal-directed optimal behavior.

How Movement Modulates Hearing

Hearing is often viewed as a passive process: Sound enters the ear, triggers a cascade of activity through the auditory system, and culminates in an auditory percept. In contrast to a passive process, motor-related signals strongly modulate the auditory system from the eardrum to the cortex. The motor modulation of auditory activity is most well documented during speech and other vocalizations but also can be detected during a wide variety of other sound-generating behaviors. An influential idea is that these motor-related signals suppress neural responses to predictable movement-generated sounds, thereby enhancing sensitivity to environmental sounds during movement while helping to detect errors in learned acoustic behaviors, including speech and musicianship. Findings in humans, monkeys, songbirds, and mice provide new insights into the circuits that convey motor-related signals to the auditory system, while lending support to the idea that these signals function predictively to facilitate hearing and vocal learning.

Sparse bursts optimize information transmission in a multiplexed neural code

Many cortical neurons combine the information ascending and descending the cortical hierarchy. In the classical view, this information is combined nonlinearly to give rise to a single firing-rate output, which collapses all input streams into one. We analyze the extent to which neurons can simultaneously represent multiple input streams by using a code that distinguishes spike timing patterns at the level of a neural ensemble. Using computational simulations constrained by experimental data, we show that cortical neurons are well suited to generate such multiplexing. Interestingly, this neural code maximizes information for short and sparse bursts, a regime consistent with in vivo recordings. Neurons can also demultiplex this information, using specific connectivity patterns. The anatomy of the adult mammalian cortex suggests that these connectivity patterns are used by the nervous system to maintain sparse bursting and optimal multiplexing. Contrary to firing-rate coding, our findings indicate that the physiology and anatomy of the cortex may be interpreted as optimizing the transmission of multiple independent signals to different targets.

Contextual signals in visual cortex

Vision is an active process. What we perceive strongly depends on our actions, intentions and expectations. During visual processing, these internal signals therefore need to be integrated with the visual information from the retina. The mechanisms of how this is achieved by the visual system are still poorly understood. Advances in recording and manipulating neuronal activity in specific cell types and axonal projections together with tools for circuit tracing are beginning to shed light on the neuronal circuit mechanisms of how internal, contextual signals shape sensory representations. Here we review recent work, primarily in mice, that has advanced our understanding of these processes, focusing on contextual signals related to locomotion, behavioural relevance and predictions.

Sensory representation of an auditory cued tactile stimulus in the posterior parietal cortex of the mouse

Sensory association cortices receive diverse inputs with their role in representing and integrating multi-sensory content remaining unclear. Here we examined the neuronal correlates of an auditory-tactile stimulus sequence in the posterior parietal cortex (PPC) using 2-photon calcium imaging in awake mice. We find that neuronal subpopulations in layer 2/3 of PPC reliably represent texture-touch events, in addition to auditory cues that presage the incoming tactile stimulus. Notably, altering the flow of sensory events through omission of the cued texture touch elicited large responses in a subset of neurons hardly responsive to or even inhibited by the tactile stimuli. Hence, PPC neurons were able to discriminate not only tactile stimulus features (i.e., texture graininess) but also between the presence and omission of the texture stimulus. Whereas some of the neurons responsive to texture omission were driven by looming-like auditory sounds others became recruited only with tactile sensory experience. These findings indicate that layer 2/3 neuronal populations in PPC potentially encode correlates of expectancy in addition to auditory and tactile stimuli.

Action and learning shape the activity of neuronal circuits in the visual cortex

•

Arousal and locomotion modulate neuronal activity in primary visual cortex.

•

Neurons in primary visual cortex respond to visuomotor mismatch.

•

Experience shapes neuronal responses to familiar stimuli, reward and object location.

•

Neuronal representations of visual stimuli are modulated according to the behavioural relevance of the stimuli.

•

Neuromodulatory, top-down and thalamocortical inputs convey arousal-related and motor-related signals to primary visual cortex.

Nonsensory variables strongly influence neuronal activity in the adult mouse primary visual cortex. Neuronal responses to visual stimuli are modulated by behavioural state, such as arousal and motor activity, and are shaped by experience. This dynamic process leads to neural representations in the visual cortex that reflect stimulus familiarity, expectations of reward and object location, and mismatch between self-motion and visual-flow. The recent development of genetic tools and recording techniques in awake behaving mice has enabled the investigation of the circuit mechanisms underlying state-dependent and experience-dependent neuronal representations in primary visual cortex. These neuronal circuits involve neuromodulatory, top-down cortico-cortical and thalamocortical pathways. The functions of nonsensory signals at this early stage of visual information processing are now beginning to be unravelled.

A Circuit for Integration of Head- and Visual-Motion Signals in Layer 6 of Mouse Primary Visual Cortex

To interpret visual-motion events, the underlying computation must involve internal reference to the motion status of the observer's head. We show here that layer 6 (L6) principal neurons in mouse primary visual cortex (V1) receive a diffuse, vestibular-mediated synaptic input that signals the angular velocity of horizontal rotation. Behavioral and theoretical experiments indicate that these inputs, distributed over a network of 100 L6 neurons, provide both a reliable estimate and, therefore, physiological separation of head-velocity signals. During head rotation in the presence of visual stimuli, L6 neurons exhibit postsynaptic responses that approximate the arithmetic sum of the vestibular and visual-motion response. Functional input mapping reveals that these internal motion signals arrive into L6 via a direct projection from the retrosplenial cortex. We therefore propose that visual-motion processing in V1 L6 is multisensory and contextually dependent on the motion status of the animal's head.

The influence of locomotion on sensory processing and its underlying neuronal circuits

Processing of sensory information can be modulated in both cortex and thalamus by behavioral context, such as locomotion. During active behaviors, coding of sensory stimuli and perception are improved, in particular during physical activity of moderate intensity. These locomotion-related modulations seem to arise from a combination of mechanisms, including neuromodulation, the recruitment of inhibitory interneurons, and specific top-down or motor-related inputs. The application of new experimental methods in mice during walking under head-fixation on treadmills made it possible to study the circuit and cellular basis underlying modulations by behavioral context with unprecedented detail. This article reviews the current state of these studies and highlights some important open questions.

Towards deep learning with segregated dendrites

Deep learning has led to significant advances in artificial intelligence, in part, by adopting strategies motivated by neurophysiology. However, it is unclear whether deep learning could occur in the real brain. Here, we show that a deep learning algorithm that utilizes multi-compartment neurons might help us to understand how the neocortex optimizes cost functions. Like neocortical pyramidal neurons, neurons in our model receive sensory information and higher-order feedback in electrotonically segregated compartments. Thanks to this segregation, neurons in different layers of the network can coordinate synaptic weight updates. As a result, the network learns to categorize images better than a single layer network. Furthermore, we show that our algorithm takes advantage of multilayer architectures to identify useful higher-order representations—the hallmark of deep learning. This work demonstrates that deep learning can be achieved using segregated dendritic compartments, which may help to explain the morphology of neocortical pyramidal neurons.

Artificial intelligence has made major progress in recent years thanks to a technique known as deep learning, which works by mimicking the human brain. When computers employ deep learning, they learn by using networks made up of many layers of simulated neurons. Deep learning has opened the door to computers with human – or even super-human – levels of skill in recognizing images, processing speech and controlling vehicles. But many neuroscientists are skeptical about whether the brain itself performs deep learning.

The patterns of activity that occur in computer networks during deep learning resemble those seen in human brains. But some features of deep learning seem incompatible with how the brain works. Moreover, neurons in artificial networks are much simpler than our own neurons. For instance, in the region of the brain responsible for thinking and planning, most neurons have complex tree-like shapes. Each cell has ‘roots’ deep inside the brain and ‘branches’ close to the surface. By contrast, simulated neurons have a uniform structure.

To find out whether networks made up of more realistic simulated neurons could be used to make deep learning more biologically realistic, Guerguiev et al. designed artificial neurons with two compartments, similar to the ‘roots’ and ‘branches’. The network learned to recognize hand-written digits more easily when it had many layers than when it had only a few. This shows that artificial neurons more like those in the brain can enable deep learning. It even suggests that our own neurons may have evolved their shape to support this process.

If confirmed, the link between neuronal shape and deep learning could help us develop better brain-computer interfaces. These allow people to use their brain activity to control devices such as artificial limbs. Despite advances in computing, we are still superior to computers when it comes to learning. Understanding how our own brains show deep learning could thus help us develop better, more human-like artificial intelligence in the future.

Cortical Processing: How Mice Predict the Visual Effects of Locomotion

New research identifies a frontal area in the mouse neocortex that sends predictions of locomotion-coupled visual flow to visual cortex. The findings support predictive coding theories of cortical processing.