DeepVID v2: self-supervised denoising with decoupled spatiotemporal enhancement for low-photon voltage imaging.

Significance: Voltage imaging is a powerful tool for studying the dynamics of neuronal activities in the brain. However, voltage imaging data are fundamentally corrupted by severe Poisson noise in the low-photon regime, which hinders the accurate extraction of neuronal activities. Self-supervised deep learning denoising methods have shown great potential in addressing the challenges in low-photon voltage imaging without the need for ground-truth but usually suffer from the trade-off between spatial and temporal performances. Aim: We present DeepVID v2, a self-supervised denoising framework with decoupled spatial and temporal enhancement capability to significantly augment low-photon voltage imaging. Approach: DeepVID v2 is built on our original DeepVID framework, which performs frame-based denoising by utilizing a sequence of frames around the central frame targeted for denoising to leverage temporal information and ensure consistency. Similar to DeepVID, the network further integrates multiple blind pixels in the central frame to enrich the learning of local spatial information. In addition, DeepVID v2 introduces a new spatial prior extraction branch to capture fine structural details to learn high spatial resolution information. Two variants of DeepVID v2 are introduced to meet specific denoising needs: an online version tailored for real-time inference with a limited number of frames and an offline version designed to leverage the full dataset, achieving optimal temporal and spatial performances. Results: We demonstrate that DeepVID v2 is able to overcome the trade-off between spatial and temporal performances and achieve superior denoising capability in resolving both high-resolution spatial structures and rapid temporal neuronal activities. We further show that DeepVID v2 can generalize to different imaging conditions, including time-series measurements with various signal-to-noise ratios and extreme low-photon conditions. Conclusions: Our results underscore DeepVID v2 as a promising tool for enhancing voltage imaging. This framework has the potential to generalize to other low-photon imaging modalities and greatly facilitate the study of neuronal activities in the brain.

Perirhinal cortex learns a predictive map of the task environment.

Goal-directed tasks involve acquiring an internal model, known as a predictive map, of relevant stimuli and associated outcomes to guide behavior. Here, we identified neural signatures of a predictive map of task behavior in perirhinal cortex (Prh). Mice learned to perform a tactile working memory task by classifying sequential whisker stimuli over multiple training stages. Chronic two-photon calcium imaging, population analysis, and computational modeling revealed that Prh encodes stimulus features as sensory prediction errors. Prh forms stable stimulus-outcome associations that can progressively be decoded earlier in the trial as training advances and that generalize as animals learn new contingencies. Stimulus-outcome associations are linked to prospective network activity encoding possible expected outcomes. This link is mediated by cholinergic signaling to guide task performance, demonstrated by acetylcholine imaging and systemic pharmacological perturbation. We propose that Prh combines error-driven and map-like properties to acquire a predictive map of learned task behavior.

DeepVID v2: Self-Supervised Denoising with Decoupled Spatiotemporal Enhancement for Low-Photon Voltage Imaging.

Significance: Voltage imaging is a powerful tool for studying the dynamics of neuronal activities in the brain. However, voltage imaging data are fundamentally corrupted by severe Poisson noise in the low-photon regime, which hinders the accurate extraction of neuronal activities. Self-supervised deep learning denoising methods have shown great potential in addressing the challenges in low-photon voltage imaging without the need for ground truth, but usually suffer from the tradeoff between spatial and temporal performance. Aim: We present DeepVID v2, a novel self-supervised denoising framework with decoupled spatial and temporal enhancement capability to significantly augment low-photon voltage imaging. Approach: DeepVID v2 is built on our original DeepVID framework,1,2 which performs frame-based denoising by utilizing a sequence of frames around the central frame targeted for denoising to leverage temporal information and ensure consistency. The network further integrates multiple blind pixels in the central frame to enrich the learning of local spatial information. Additionally, DeepVID v2 introduces a new edge extraction branch to capture fine structural details in order to learn high spatial resolution information. Results: We demonstrate that DeepVID v2 is able to overcome the tradeoff between spatial and temporal performance, and achieve superior denoising capability in resolving both high-resolution spatial structures and rapid temporal neuronal activities. We further show that DeepVID v2 is able to generalize to different imaging conditions, including time-series measurements with various signal-to-noise ratios (SNRs) and in extreme low-photon conditions. Conclusions: Our results underscore DeepVID v2 as a promising tool for enhancing voltage imaging. This framework has the potential to generalize to other low-photon imaging modalities and greatly facilitate the study of neuronal activities in the brain.

High-speed low-light in vivo two-photon voltage imaging of large neuronal populations.

Monitoring spiking activity across large neuronal populations at behaviorally relevant timescales is critical for understanding neural circuit function. Unlike calcium imaging, voltage imaging requires kilohertz sampling rates that reduce fluorescence detection to near shot-noise levels. High-photon flux excitation can overcome photon-limited shot noise, but photobleaching and photodamage restrict the number and duration of simultaneously imaged neurons. We investigated an alternative approach aimed at low two-photon flux, which is voltage imaging below the shot-noise limit. This framework involved developing positive-going voltage indicators with improved spike detection (SpikeyGi and SpikeyGi2); a two-photon microscope ('SMURF') for kilohertz frame rate imaging across a 0.4 mm × 0.4 mm field of view; and a self-supervised denoising algorithm (DeepVID) for inferring fluorescence from shot-noise-limited signals. Through these combined advances, we achieved simultaneous high-speed deep-tissue imaging of more than 100 densely labeled neurons over 1 hour in awake behaving mice. This demonstrates a scalable approach for voltage imaging across increasing neuronal populations.

PERIRHINAL CORTEX LEARNS A PREDICTIVE MAP (INTERNAL MODEL) OF THE TASK ENVIRONMENT.

Goal-directed tasks involve acquiring an internal model, known as a predictive map, of relevant stimuli and associated outcomes to guide behavior. Here, we identified neural signatures of a predictive map of task behavior in perirhinal cortex (Prh). Mice learned to perform a tactile working memory task by classifying sequential whisker stimuli over multiple training stages. Chemogenetic inactivation demonstrated that Prh is involved in task learning. Chronic two-photon calcium imaging, population analysis, and computational modeling revealed that Prh encodes stimulus features as sensory prediction errors. Prh forms stable stimulus-outcome associations that expand in a retrospective manner and generalize as animals learn new contingencies. Stimulus-outcome associations are linked to prospective network activity encoding possible expected outcomes. This link is mediated by cholinergic signaling to guide task performance, demonstrated by acetylcholine imaging and perturbation. We propose that Prh combines error-driven and map-like properties to acquire a predictive map of learned task behavior.

Dense functional and molecular readout of a circuit hub in sensory cortex.

Although single-cell transcriptomics of the neocortex has uncovered more than 300 putative cell types, whether this molecular classification predicts distinct functional roles is unclear. We combined two-photon calcium imaging with spatial transcriptomics to functionally and molecularly investigate cortical circuits. We characterized behavior-related responses across major neuronal subclasses in layers 2 or 3 of the primary somatosensory cortex as mice performed a tactile working memory task. We identified an excitatory intratelencephalic cell type, Baz1a, that exhibits high tactile feature selectivity. Baz1a neurons homeostatically maintain stimulus responsiveness during altered experience and show persistent enrichment of subsets of immediately early genes. Functional and anatomical connectivity reveals that Baz1a neurons residing in upper portions of layers 2 or 3 preferentially innervate somatostatin-expressing inhibitory neurons. This motif defines a circuit hub that orchestrates local sensory processing in superficial layers of the neocortex.

Flexible simultaneous mesoscale two-photon imaging of neural activity at high speeds.

Understanding brain function requires monitoring local and global brain dynamics. Two-photon imaging of the brain across mesoscopic scales has presented trade-offs between imaging area and acquisition speed. We describe a flexible cellular resolution two-photon microscope capable of simultaneous video rate acquisition of four independently targetable brain regions spanning an approximate five-millimeter field of view. With this system, we demonstrate the ability to measure calcium activity across mouse sensorimotor cortex at behaviorally relevant timescales.

jYCaMP: an optimized calcium indicator for two-photon imaging at fiber laser wavelengths.

Femtosecond lasers at fixed wavelengths above 1,000 nm are powerful, stable and inexpensive, making them promising sources for two-photon microscopy. Biosensors optimized for these wavelengths are needed for both next-generation microscopes and affordable turn-key systems. Here we report jYCaMP1, a yellow variant of the calcium indicator jGCaMP7 that outperforms its parent in mice and flies at excitation wavelengths above 1,000 nm and enables improved two-color calcium imaging with red fluorescent protein-based indicators.

Context-Dependent Sensory Processing across Primary and Secondary Somatosensory Cortex.

To interpret the environment, our brain must evaluate external stimuli against internal representations from past experiences. How primary (S1) and secondary (S2) somatosensory cortices process stimuli depending on recent experiences is unclear. Using simultaneous multi-area population imaging of projection neurons and focal optogenetic inactivation, we studied mice performing a whisker-based working memory task. We find that activity reflecting a current stimulus, the recollection of a previous stimulus (cued recall), and the stimulus category are distributed across S1 and S2. Despite this overlapping representation, S2 is important for processing cued recall responses and transmitting these responses to S1. S2 network properties differ from S1, wherein S2 persistently encodes cued recall and the stimulus category under passive conditions. Although both areas encode the stimulus category, only information in S1 is important for task performance through pathways that do not necessarily include S2. These findings reveal both distributed and segregated roles for S1 and S2 in context-dependent sensory processing.

CELLULAR RESOLUTION IMAGING OF NEURONAL ACTIVITY ACROSS SPACE AND TIME IN THE MAMMALIAN BRAIN.

While the action potential has long been understood to be the fundamental bit of information in brain, how these spikes encode representations of stimuli and drive behavior remains unclear. Large-scale neuronal recordings with cellular and spike-time resolution spanning multiple brain regions are needed to capture relevant network dynamics that can be sparse and distributed across the population. This review focuses on recent advancements in optical methods that have pushed the boundaries for simultaneous population recordings at increasing volumes, distances, depths, and speeds. The integration of these technologies will be critical for overcoming fundamental limits in the pursuit of whole brain imaging in mammalian species.

Neocortical dynamics during whisker-based sensory discrimination in head-restrained mice.

A fundamental task frequently encountered by brains is to rapidly and reliably discriminate between sensory stimuli of the same modality, be it distinct auditory sounds, odors, visual patterns, or tactile textures. A key mammalian brain structure involved in discrimination behavior is the neocortex. Sensory processing not only involves the respective primary sensory area, which is crucial for perceptual detection, but additionally relies on cortico-cortical communication among several regions including higher-order sensory areas as well as frontal cortical areas. It remains elusive how these regions exchange information to process neural representations of distinct stimuli to bring about a decision and initiate appropriate behavioral responses. Likewise, it is poorly understood how these neural computations are conjured during task learning. In this review, we discuss recent studies investigating cortical dynamics during discrimination behaviors that utilize head-fixed behavioral tasks in combination with in vivo electrophysiology, two-photon calcium imaging, and cell-type-specific targeting. We particularly focus on information flow in distinct cortico-cortical pathways when mice use their whiskers to discriminate between different objects or different locations. Within the primary and secondary somatosensory cortices (S1 and S2, respectively) as well as vibrissae motor cortex (M1), intermingled functional representations of touch, whisking, and licking were found, which partially re-organized during discrimination learning. These findings provide first glimpses of cortico-cortical communication but emphasize that for understanding the complete process of discrimination it will be crucial to elucidate the details of how neural processing is coordinated across brain-wide neuronal networks including the S1-S2-M1 triangle and cortical areas beyond.

Long-range cortical dynamics: a perspective from the mouse sensorimotor whisker system.

In the mammalian neocortex, the capacity to dynamically route and coordinate the exchange of information between areas is a critical feature of cognitive function, enabling processes such as higher-level sensory processing and sensorimotor integration. Despite the importance attributed to long-range connections between cortical areas, their exact operations and role in cortical function remain an open question. In recent years, progress has been made in understanding long-range cortical circuits through work focused on the mouse sensorimotor whisker system. In this review, we examine recent studies dissecting long-range circuits involved in whisker sensorimotor processing as an entry point for understanding the rules that govern long-range cortical circuit function.

An R-CaMP1.07 reporter mouse for cell-type-specific expression of a sensitive red fluorescent calcium indicator.

Genetically encoded calcium indicators (GECIs) enable imaging of in vivo brain cell activity with high sensitivity and specificity. In contrast to viral infection or in utero electroporation, indicator expression in transgenic reporter lines is induced noninvasively, reliably, and homogenously. Recently, Cre/tTA-dependent reporter mice were introduced, which provide high-level expression of green fluorescent GECIs in a cell-type-specific and inducible manner when crossed with Cre and tTA driver mice. Here, we generated and characterized the first red-shifted GECI reporter line of this type using R-CaMP1.07, a red fluorescent indicator that is efficiently two-photon excited above 1000 nm. By crossing the new R-CaMP1.07 reporter line to Cre lines driving layer-specific expression in neocortex we demonstrate its high fidelity for reporting action potential firing in vivo, long-term stability over months, and versatile use for functional imaging of excitatory neurons across all cortical layers, especially in the previously difficult to access layers 4 and 6.

Long-range population dynamics of anatomically defined neocortical networks.

The coordination of activity across neocortical areas is essential for mammalian brain function. Understanding this process requires simultaneous functional measurements across the cortex. In order to dissociate direct cortico-cortical interactions from other sources of neuronal correlations, it is furthermore desirable to target cross-areal recordings to neuronal subpopulations that anatomically project between areas. Here, we combined anatomical tracers with a novel multi-area two-photon microscope to perform simultaneous calcium imaging across mouse primary (S1) and secondary (S2) somatosensory whisker cortex during texture discrimination behavior, specifically identifying feedforward and feedback neurons. We find that coordination of S1-S2 activity increases during motor behaviors such as goal-directed whisking and licking. This effect was not specific to identified feedforward and feedback neurons. However, these mutually projecting neurons especially participated in inter-areal coordination when motor behavior was paired with whisker-texture touches, suggesting that direct S1-S2 interactions are sensory-dependent. Our results demonstrate specific functional coordination of anatomically-identified projection neurons across sensory cortices.

Pathway-specific reorganization of projection neurons in somatosensory cortex during learning.

In the mammalian brain, sensory cortices exhibit plasticity during task learning, but how this alters information transferred between connected cortical areas remains unknown. We found that divergent subpopulations of cortico-cortical neurons in mouse whisker primary somatosensory cortex (S1) undergo functional changes reflecting learned behavior. We chronically imaged activity of S1 neurons projecting to secondary somatosensory (S2) or primary motor (M1) cortex in mice learning a texture discrimination task. Mice adopted an active whisking strategy that enhanced texture-related whisker kinematics, correlating with task performance. M1-projecting neurons reliably encoded basic kinematics features, and an additional subset of touch-related neurons was recruited that persisted past training. The number of S2-projecting touch neurons remained constant, but improved their discrimination of trial types through reorganization while developing activity patterns capable of discriminating the animal's decision. We propose that learning-related changes in S1 enhance sensory representations in a pathway-specific manner, providing downstream areas with task-relevant information for behavior.

Imaging neuronal populations in behaving rodents: paradigms for studying neural circuits underlying behavior in the mammalian cortex.

Understanding the neural correlates of behavior in the mammalian cortex requires measurements of activity in awake, behaving animals. Rodents have emerged as a powerful model for dissecting the cortical circuits underlying behavior attributable to the convergence of several methods. Genetically encoded calcium indicators combined with viral-mediated or transgenic tools enable chronic monitoring of calcium signals in neuronal populations and subcellular structures of identified cell types. Stable one- and two-photon imaging of neuronal activity in awake, behaving animals is now possible using new behavioral paradigms in head-fixed animals, or using novel miniature head-mounted microscopes in freely moving animals. This mini-symposium will highlight recent applications of these methods for studying sensorimotor integration, decision making, learning, and memory in cortical and subcortical brain areas. We will outline future prospects and challenges for identifying the neural underpinnings of task-dependent behavior using cellular imaging in rodents.

Online correction of licking-induced brain motion during two-photon imaging with a tunable lens.

Two-photon calcium imaging in awake, head-fixed animals enables the measurement of neuronal activity during behaviour. Often, licking for the retrieval of water reward is used as a measurable report of the animal's decision during reward-driven behaviour. However, licking behaviour can induce severe motion artifacts that interfere with two-photon imaging of cellular activity. Here, we describe a simple method for the online correction of licking-induced focus shifts for two-photon calcium imaging of neocortical neurons in the head-fixed mouse. We found that licking causes a stereotyped drop of neocortical tissue, shifting neurons up to 20 µm out of focus. Based on the measurement of licking with a piezo film sensor, we developed a feedback model, which provides a corrective signal for fast optical focus adjustments with an electrically tunable lens. Using online correction with this feedback model, we demonstrate a reduction of licking-related focus changes below 3 µm, minimizing motion artifact contamination of cellular calcium signals. Focus correction with a tunable lens is a simple and effective method to improve the ability to monitor neuronal activity during reward-based behaviour.

Behaviour-dependent recruitment of long-range projection neurons in somatosensory cortex.

In the mammalian neocortex, segregated processing streams are thought to be important for forming sensory representations of the environment, but how local information in primary sensory cortex is transmitted to other distant cortical areas during behaviour is unclear. Here we show task-dependent activation of distinct, largely non-overlapping long-range projection neurons in the whisker region of primary somatosensory cortex (S1) in awake, behaving mice. Using two-photon calcium imaging, we monitored neuronal activity in anatomically identified S1 neurons projecting to secondary somatosensory (S2) or primary motor (M1) cortex in mice using their whiskers to perform a texture-discrimination task or a task that required them to detect the presence of an object at a certain location. Whisking-related cells were found among S2-projecting (S2P) but not M1-projecting (M1P) neurons. A higher fraction of S2P than M1P neurons showed touch-related responses during texture discrimination, whereas a higher fraction of M1P than S2P neurons showed touch-related responses during the detection task. In both tasks, S2P and M1P neurons could discriminate similarly between trials producing different behavioural decisions. However, in trials producing the same decision, S2P neurons performed better at discriminating texture, whereas M1P neurons were better at discriminating location. Sensory stimulus features alone were not sufficient to elicit these differences, suggesting that selective transmission of S1 information to S2 and M1 is driven by behaviour.

Highly specific structural plasticity of inhibitory circuits in the adult neocortex.

Inhibitory neurons are known to play a vital role in defining the window for critical period plasticity during development, and it is increasingly apparent that they continue to exert powerful control over experience-dependent cortical plasticity in adulthood. Recent in vivo imaging studies demonstrate that long-term plasticity of inhibitory circuits is manifested at an anatomical level. Changes in sensory experience drive structural remodeling in inhibitory interneurons in a cell-type and circuit-specific manner. Inhibitory synapse formation and elimination can occur with a great deal of spatial and temporal precision and are locally coordinated with excitatory synaptic changes on the same neuron. We suggest that the specificity of inhibitory synapse dynamics may serve to differentially modulate activity across the dendritic arbor, to selectively tune parts of a local circuit, or potentially discriminate between activities in distinct local circuits. We further review evidence suggesting that inhibitory circuit structural changes instruct excitatory/inhibitory balance while enabling functional reorganization to occur through Hebbian forms of plasticity.