Statistically unbiased prediction enables accurate denoising of voltage imaging data.

Here we report SUPPORT (statistically unbiased prediction utilizing spatiotemporal information in imaging data), a self-supervised learning method for removing Poisson-Gaussian noise in voltage imaging data. SUPPORT is based on the insight that a pixel value in voltage imaging data is highly dependent on its spatiotemporal neighboring pixels, even when its temporally adjacent frames alone do not provide useful information for statistical prediction. Such dependency is captured and used by a convolutional neural network with a spatiotemporal blind spot to accurately denoise voltage imaging data in which the existence of the action potential in a time frame cannot be inferred by the information in other frames. Through simulations and experiments, we show that SUPPORT enables precise denoising of voltage imaging data and other types of microscopy image while preserving the underlying dynamics within the scene.

Population imaging of neural activity in awake behaving mice.

A longstanding goal in neuroscience has been to image membrane voltage across a population of individual neurons in an awake, behaving mammal. Here we describe a genetically encoded fluorescent voltage indicator, SomArchon, which exhibits millisecond response times and is compatible with optogenetic control, and which increases the sensitivity, signal-to-noise ratio, and number of neurons observable several-fold over previously published fully genetically encoded reagents1-8. Under conventional one-photon microscopy, SomArchon enables the routine population analysis of around 13 neurons at once, in multiple brain regions (cortex, hippocampus, and striatum) of head-fixed, awake, behaving mice. Using SomArchon, we detected both positive and negative responses of striatal neurons during movement, as previously reported by electrophysiology but not easily detected using modern calcium imaging techniques9-11, highlighting the power of voltage imaging to reveal bidirectional modulation. We also examined how spikes relate to the subthreshold theta oscillations of individual hippocampal neurons, with SomArchon showing that the spikes of individual neurons are more phase-locked to their own subthreshold theta oscillations than to local field potential theta oscillations. Thus, SomArchon reports both spikes and subthreshold voltage dynamics in awake, behaving mice.

CaMKIIα-Positive Interneurons Identified via a microRNA-Based Viral Gene Targeting Strategy.

Single-cell analysis is revealing increasing diversity in gene expression profiles among brain cells. Traditional promotor-based viral gene expression techniques, however, cannot capture the growing variety among single cells. We demonstrate a novel viral gene expression strategy to target cells with specific miRNA expression using miRNA-guided neuron tags (mAGNET). We designed mAGNET viral vectors containing a CaMKIIα promoter and microRNA-128 (miR-128) binding sites, and labeled CaMKIIα+ cells with naturally low expression of miR-128 (Lm128C cells) in male and female mice. Although CaMKIIα has traditionally been considered as an excitatory neuron marker, our single-cell sequencing results reveal that Lm128C cells are CaMKIIα+ inhibitory neurons of parvalbumin or somatostatin subtypes. Further evaluation of the physiological properties of Lm128C cell in brain slices showed that Lm128C cells exhibit elevated membrane excitability, with biophysical properties closely resembling those of fast-spiking interneurons, consistent with previous transcriptomic findings of miR-128 in regulating gene networks that govern membrane excitability. To further demonstrate the utility of this new viral expression strategy, we expressed GCaMP6f in Lm128C cells in the superficial layers of the motor cortex and performed in vivo calcium imaging in mice during locomotion. We found that Lm128C cells exhibit elevated calcium event rates and greater intrapopulation correlation than the overall CaMKIIα+ cells during movement. In summary, the miRNA-based viral gene targeting strategy described here allows us to label a sparse population of CaMKIIα+ interneurons for functional studies, providing new capabilities to investigate the relationship between gene expression and physiological properties in the brain.SIGNIFICANCE STATEMENT We report the discovery of a class of CaMKIIα+ cortical interneurons, labeled via a novel miRNA-based viral gene targeting strategy, combinatorial to traditional promoter-based strategies. The fact that we found a small, yet distinct, population of cortical inhibitory neurons that express CaMKIIα demonstrates that CaMKIIα is not as specific for excitatory neurons as commonly believed. As single-cell sequencing tools are providing increasing insights into the gene expression diversity of neurons, including miRNA profile data, we expect that the miRNA-based gene targeting strategy presented here can help delineate many neuron populations whose physiological properties can be readily related to the miRNA gene regulatory networks.

Mechanisms of generation of membrane potential resonance in a neuron with multiple resonant ionic currents.

Neuronal membrane potential resonance (MPR) is associated with subthreshold and network oscillations. A number of voltage-gated ionic currents can contribute to the generation or amplification of MPR, but how the interaction of these currents with linear currents contributes to MPR is not well understood. We explored this in the pacemaker PD neurons of the crab pyloric network. The PD neuron MPR is sensitive to blockers of H- (IH) and calcium-currents (ICa). We used the impedance profile of the biological PD neuron, measured in voltage clamp, to constrain parameter values of a conductance-based model using a genetic algorithm and obtained many optimal parameter combinations. Unlike most cases of MPR, in these optimal models, the values of resonant- (fres) and phasonant- (fϕ = 0) frequencies were almost identical. Taking advantage of this fact, we linked the peak phase of ionic currents to their amplitude, in order to provide a mechanistic explanation the dependence of MPR on the ICa gating variable time constants. Additionally, we found that distinct pairwise correlations between ICa parameters contributed to the maintenance of fres and resonance power (QZ). Measurements of the PD neuron MPR at more hyperpolarized voltages resulted in a reduction of fres but no change in QZ. Constraining the optimal models using these data unmasked a positive correlation between the maximal conductances of IH and ICa. Thus, although IH is not necessary for MPR in this neuron type, it contributes indirectly by constraining the parameters of ICa.

The frequency preference of neurons and synapses in a recurrent oscillatory network.

A variety of neurons and synapses shows a maximal response at a preferred frequency, generally considered to be important in shaping network activity. We are interested in whether all neurons and synapses in a recurrent oscillatory network can have preferred frequencies and, if so, whether these frequencies are the same or correlated, and whether they influence the network activity. We address this question using identified neurons in the pyloric network of the crab Cancer borealis. Previous work has shown that the pyloric pacemaker neurons exhibit membrane potential resonance whose resonance frequency is correlated with the network frequency. The follower lateral pyloric (LP) neuron makes reciprocally inhibitory synapses with the pacemakers. We find that LP shows resonance at a higher frequency than the pacemakers and the network frequency falls between the two. We also find that the reciprocal synapses between the pacemakers and LP have preferred frequencies but at significantly lower values. The preferred frequency of the LP to pacemaker synapse is correlated with the presynaptic preferred frequency, which is most pronounced when the peak voltage of the LP waveform is within the dynamic range of the synaptic activation curve and a shift in the activation curve by the modulatory neuropeptide proctolin shifts the frequency preference. Proctolin also changes the power of the LP neuron resonance without significantly changing the resonance frequency. These results indicate that different neuron types and synapses in a network may have distinct preferred frequencies, which are subject to neuromodulation and may interact to shape network oscillations.

Ultrasound pulse repetition frequency preferentially activates different neuron populations independent of cell type.

Transcranial ultrasound activates mechanosensitive cellular signaling and modulates neural dynamics. Given that intrinsic neuronal activity is limited to a couple hundred hertz and often exhibits frequency preference, we examined whether pulsing ultrasound at physiologic pulse repetition frequencies (PRFs) could selectively influence neuronal activity in the mammalian brain. We performed calcium imaging of individual motor cortex neurons, while delivering 0.35 MHz ultrasound at PRFs of 10, 40, and 140 Hz in awake mice. We found that most neurons were preferentially activated by only one of the three PRFs, highlighting unique cellular effects of physiologic PRFs. Further, ultrasound evoked responses were similar between excitatory neurons and parvalbumin positive interneurons regardless of PRFs, indicating that individual cell sensitivity dominates ultrasound-evoked effects, consistent with the heterogeneous mechanosensitive channel expression we found across single neurons in mice and humans. These results highlight the feasibility of tuning ultrasound neuromodulation effects through varying PRFs.

Striatal cholinergic interneuron membrane voltage tracks locomotor rhythms in mice.

Rhythmic neural network activity has been broadly linked to behavior. However, it is unclear how membrane potentials of individual neurons track behavioral rhythms, even though many neurons exhibit pace-making properties in isolated brain circuits. To examine whether single-cell voltage rhythmicity is coupled to behavioral rhythms, we focused on delta-frequencies (1-4 Hz) that are known to occur at both the neural network and behavioral levels. We performed membrane voltage imaging of individual striatal neurons simultaneously with network-level local field potential recordings in mice during voluntary movement. We report sustained delta oscillations in the membrane potentials of many striatal neurons, particularly cholinergic interneurons, which organize spikes and network oscillations at beta-frequencies (20-40 Hz) associated with locomotion. Furthermore, the delta-frequency patterned cellular dynamics are coupled to animals' stepping cycles. Thus, delta-rhythmic cellular dynamics in cholinergic interneurons, known for their autonomous pace-making capabilities, play an important role in regulating network rhythmicity and movement patterning.

The autism spectrum disorder risk gene NEXMIF over-synchronizes hippocampal CA1 network and alters neuronal coding.

Mutations in autism spectrum disorder (ASD) risk genes disrupt neural network dynamics that ultimately lead to abnormal behavior. To understand how ASD-risk genes influence neural circuit computation during behavior, we analyzed the hippocampal network by performing large-scale cellular calcium imaging from hundreds of individual CA1 neurons simultaneously in transgenic mice with total knockout of the X-linked ASD-risk gene NEXMIF (neurite extension and migration factor). As NEXMIF knockout in mice led to profound learning and memory deficits, we examined the CA1 network during voluntary locomotion, a fundamental component of spatial memory. We found that NEXMIF knockout does not alter the overall excitability of individual neurons but exaggerates movement-related neuronal responses. To quantify network functional connectivity changes, we applied closeness centrality analysis from graph theory to our large-scale calcium imaging datasets, in addition to using the conventional pairwise correlation analysis. Closeness centrality analysis considers both the number of connections and the connection strength between neurons within a network. We found that in wild-type mice the CA1 network desynchronizes during locomotion, consistent with increased network information coding during active behavior. Upon NEXMIF knockout, CA1 network is over-synchronized regardless of behavioral state and fails to desynchronize during locomotion, highlighting how perturbations in ASD-implicated genes create abnormal network synchronization that could contribute to ASD-related behaviors.

Theta and gamma rhythmic coding through two spike output modes in the hippocampus during spatial navigation.

Hippocampal CA1 neurons generate single spikes and stereotyped bursts of spikes. However, it is unclear how individual neurons dynamically switch between these output modes and whether these two spiking outputs relay distinct information. We performed extracellular recordings in spatially navigating rats and cellular voltage imaging and optogenetics in awake mice. We found that spike bursts are preferentially linked to cellular and network theta rhythms (3-12 Hz) and encode an animal's position via theta phase precession, particularly as animals are entering a place field. In contrast, single spikes exhibit additional coupling to gamma rhythms (30-100 Hz), particularly as animals leave a place field. Biophysical modeling suggests that intracellular properties alone are sufficient to explain the observed input frequency-dependent spike coding. Thus, hippocampal neurons regulate the generation of bursts and single spikes according to frequency-specific network and intracellular dynamics, suggesting that these spiking modes perform distinct computations to support spatial behavior.

The membrane potential waveform of bursting pacemaker neurons is a predictor of their preferred frequency and the network cycle frequency.

Many oscillatory networks involve neurons that exhibit intrinsic rhythmicity but possess a large variety of voltage-gated currents that interact in a complex fashion, making it difficult to determine which factors control frequency. Yet these neurons often have preferred (resonance) frequencies that can be close to the network frequency. Because the preferred frequency results from the dynamics of ionic currents, it can be assumed to depend on parameters that determine the neuron's oscillatory waveform shape. The pyloric network frequency in the crab Cancer borealis is correlated with the preferred frequency of its bursting pacemaker neurons anterior burster and pyloric dilator (PD). We measured the preferred frequency of the PD neuron in voltage clamp, which allows control of the oscillation voltage range and waveforms (sine waves and realistic oscillation waveforms), and showed that (1) the preferred frequency depends on the voltage range of the oscillating voltage waveform; (2) the slope of the waveform near its peak has a strongly negative correlation with the preferred frequency; and (3) correlations between parameters of the PD neuron oscillation waveform and its preferred frequency can be used to predict shifts in the network frequency. As predicted by these results, dynamic clamp shifts of the upper or lower voltage limits of the PD neuron waveform during ongoing oscillations changed the network frequency, consistent with the predictions from the preferred frequency. These results show that the voltage waveform of oscillatory neurons can be predictive of their preferred frequency and thus the network oscillation frequency.

Distinct Spiking Patterns of Excitatory and Inhibitory Neurons and LFP Oscillations in Prefrontal Cortex During Sensory Discrimination.

Prefrontal cortex (PFC) are broadly linked to various aspects of behavior. During sensory discrimination, PFC neurons can encode a range of task related information, including the identity of sensory stimuli and related behavioral outcome. However, it remains largely unclear how different neuron subtypes and local field potential (LFP) oscillation features in the mouse PFC are modulated during sensory discrimination. To understand how excitatory and inhibitory PFC neurons are selectively engaged during sensory discrimination and how their activity relates to LFP oscillations, we used tetrode recordings to probe well-isolated individual neurons, and LFP oscillations, in mice performing a three-choice auditory discrimination task. We found that a majority of PFC neurons, 78% of the 711 recorded individual neurons, exhibited sensory discrimination related responses that are context and task dependent. Using spike waveforms, we classified these responsive neurons into putative excitatory neurons with broad waveforms or putative inhibitory neurons with narrow waveforms, and found that both neuron subtypes were transiently modulated, with individual neurons' responses peaking throughout the entire duration of the trial. While the number of responsive excitatory neurons remain largely constant throughout the trial, an increasing fraction of inhibitory neurons were gradually recruited as the trial progressed. Further examination of the coherence between individual neurons and LFPs revealed that inhibitory neurons exhibit higher spike-field coherence with LFP oscillations than excitatory neurons during all aspects of the trial and across multiple frequency bands. Together, our results demonstrate that PFC excitatory neurons are continuously engaged during sensory discrimination, whereas PFC inhibitory neurons are increasingly recruited as the trial progresses and preferentially coordinated with LFP oscillations. These results demonstrate increasing involvement of inhibitory neurons in shaping the overall PFC dynamics toward the completion of the sensory discrimination task.

Region-specific effects of ultrasound on individual neurons in the awake mammalian brain.

Ultrasound modulates brain activity. However, it remains unclear how ultrasound affects individual neurons in the brain, where neural circuit architecture is intact and different brain regions exhibit distinct tissue properties. Using a high-resolution calcium imaging technique, we characterized the effect of ultrasound stimulation on thousands of individual neurons in the hippocampus and the motor cortex of awake mice. We found that brief 100-ms-long ultrasound pulses increase intracellular calcium in a large fraction of individual neurons in both brain regions. Ultrasound-evoked calcium response in hippocampal neurons exhibits a rapid onset with a latency shorter than 50 ms. The evoked response in the hippocampus is shorter in duration and smaller in magnitude than that in the motor cortex. These results demonstrate that noninvasive ultrasound stimulation transiently increases intracellular calcium in individual neurons in awake mice, and the evoked response profiles are brain region specific.

Large-scale voltage imaging in behaving mice using targeted illumination.

Recent improvements in genetically encoded voltage indicators enabled optical imaging of action potentials and subthreshold transmembrane voltage in vivo. To perform high-speed voltage imaging of many neurons simultaneously over a large anatomical area, widefield microscopy remains an essential tool. However, the lack of optical sectioning makes widefield microscopy prone to background cross-contamination. We implemented a digital-micromirror-device-based targeted illumination strategy to restrict illumination to the cells of interest and quantified the resulting improvement both theoretically and experimentally with SomArchon expressing neurons. We found that targeted illumination increased SomArchon signal contrast, decreased photobleaching, and reduced background cross-contamination. With the use of a high-speed, large-area sCMOS camera, we routinely imaged tens of spiking neurons simultaneously over minutes in behaving mice. Thus, the targeted illumination strategy described here offers a simple solution for widefield voltage imaging of many neurons over a large field of view in behaving animals.

High-contrast multifocus microscopy with a single camera and z-splitter prism.

Optical microscopy has been an indispensable tool for studying complex biological systems, but is often hampered by problems of speed and complexity when performing 3D volumetric imaging. Here, we present a multifocus imaging strategy based on the use of a simple z-splitter prism that can be assembled from off-the-shelf components. Our technique enables a widefield image stack to be distributed onto a single camera and recorded simultaneously. We exploit the volumetric nature of our image acquisition by further introducing a novel extended-volume 3D deconvolution strategy to suppress far-out-of-focus fluorescence background to significantly improve the contrast of our recorded images, conferring to our system a capacity for quasi-optical sectioning. By swapping in different z-splitter configurations, we can prioritize high speed or large 3D field-of-view imaging depending on the application of interest. Moreover, our system can be readily applied to a variety of imaging modalities in addition to fluorescence, such as phase-contrast and darkfield imaging. Because of its simplicity, versatility, and performance, we believe our system will be a useful tool for general biological or biomedical imaging applications.

Voltage Imaging of Cardiac Cells and Tissue Using the Genetically Encoded Voltage Sensor Archon1.

Precise measurement of action potentials (APs) is needed to observe electrical activity and cellular communication within cardiac tissue. Voltage-sensitive dyes (VSDs) are traditionally used to measure cardiac APs; however, they require acute chemical addition that prevents chronic imaging. Genetically encoded voltage indicators (GEVIs) enable long-term studies of APs without the need of chemical additions, but current GEVIs used in cardiac tissue exhibit poor kinetics and/or low signal to noise (SNR). Here, we demonstrate the use of Archon1, a recently developed GEVI, in hiPSC-derived cardiomyocytes (CMs). When expressed in CMs, Archon1 demonstrated fast kinetics comparable with patch-clamp electrophysiology and high SNR significantly greater than the VSD Di-8-ANEPPS. Additionally, Archon1 enabled monitoring of APs across multiple cells simultaneously in 3D cardiac tissues. These results highlight Archon1's capability to investigate the electrical activity of CMs in a variety of applications and its potential to probe functionally complex in vitro models, as well as in vivo systems.

Optoacoustic brain stimulation at submillimeter spatial precision.

Low-intensity ultrasound is an emerging modality for neuromodulation. Yet, transcranial neuromodulation using low-frequency piezo-based transducers offers poor spatial confinement of excitation volume, often bigger than a few millimeters in diameter. In addition, the bulky size limits their implementation in a wearable setting and prevents integration with other experimental modalities. Here, we report spatially confined optoacoustic neural stimulation through a miniaturized Fiber-Optoacoustic Converter (FOC). The FOC has a diameter of 600 µm and generates omnidirectional ultrasound wave locally at the fiber tip through the optoacoustic effect. We show that the acoustic wave generated by FOC can directly activate individual cultured neurons and generate intracellular Ca2+ transients. The FOC activates neurons within a radius of 500 µm around the fiber tip, delivering superior spatial resolution over conventional piezo-based low-frequency transducers. Finally, we demonstrate direct and spatially confined neural stimulation of mouse brain and modulation of motor activity in vivo.

Precision Calcium Imaging of Dense Neural Populations via a Cell-Body-Targeted Calcium Indicator.

Methods for one-photon fluorescent imaging of calcium dynamics can capture the activity of hundreds of neurons across large fields of view at a low equipment complexity and cost. In contrast to two-photon methods, however, one-photon methods suffer from higher levels of crosstalk from neuropil, resulting in a decreased signal-to-noise ratio and artifactual correlations of neural activity. We address this problem by engineering cell-body-targeted variants of the fluorescent calcium indicators GCaMP6f and GCaMP7f. We screened fusions of GCaMP to natural, as well as artificial, peptides and identified fusions that localized GCaMP to within 50 µm of the cell body of neurons in mice and larval zebrafish. One-photon imaging of soma-targeted GCaMP in dense neural circuits reported fewer artifactual spikes from neuropil, an increased signal-to-noise ratio, and decreased artifactual correlation across neurons. Thus, soma-targeting of fluorescent calcium indicators facilitates usage of simple, powerful, one-photon methods for imaging neural calcium dynamics.

Video-rate volumetric neuronal imaging using 3D targeted illumination.

Fast volumetric microscopy is required to monitor large-scale neural ensembles with high spatio-temporal resolution. Widefield fluorescence microscopy can image large 2D fields of view at high resolution and speed while remaining simple and costeffective. A focal sweep add-on can further extend the capacity of widefield microscopy by enabling extended-depth-of-field (EDOF) imaging, but suffers from an inability to reject out-of-focus fluorescence background. Here, by using a digital micromirror device to target only in-focus sample features, we perform EDOF imaging with greatly enhanced contrast and signal-to-noise ratio, while reducing the light dosage delivered to the sample. Image quality is further improved by the application of a robust deconvolution algorithm. We demonstrate the advantages of our technique for in vivo calcium imaging in the mouse brain.

Automatic Cell Segmentation by Adaptive Thresholding (ACSAT) for Large-Scale Calcium Imaging Datasets.

Advances in calcium imaging have made it possible to record from an increasingly larger number of neurons simultaneously. Neuroscientists can now routinely image hundreds to thousands of individual neurons. An emerging technical challenge that parallels the advancement in imaging a large number of individual neurons is the processing of correspondingly large datasets. One important step is the identification of individual neurons. Traditional methods rely mainly on manual or semimanual inspection, which cannot be scaled for processing large datasets. To address this challenge, we focused on developing an automated segmentation method, which we refer to as automated cell segmentation by adaptive thresholding (ACSAT). ACSAT works with a time-collapsed image and includes an iterative procedure that automatically calculates global and local threshold values during successive iterations based on the distribution of image pixel intensities. Thus, the algorithm is capable of handling variations in morphological details and in fluorescence intensities in different calcium imaging datasets. In this paper, we demonstrate the utility of ACSAT by testing it on 500 simulated datasets, two wide-field hippocampus datasets, a wide-field striatum dataset, a wide-field cell culture dataset, and a two-photon hippocampus dataset. For the simulated datasets with truth, ACSAT achieved >80% recall and precision when the signal-to-noise ratio was no less than ∼24 dB.

A MicroRNA-Based Gene-Targeting Tool for Virally Labeling Interneurons in the Rodent Cortex.

More specific and broadly applicable viral gene-targeting tools for labeling neuron subtypes are needed to advance neuroscience research, especially in rodent transgenic disease models and genetically intractable species. Here, we develop a viral vector that restricts transgene expression to GABAergic interneurons in the rodent neocortex by exploiting endogenous microRNA regulation. Our interneuron-targeting, microRNA-guided neuron tag, "GABA mAGNET," achieves >95% interneuron selective labeling in the mouse cortex, including in a murine model of autism and also some preferential labeling of interneurons in the rat brain. We demonstrate an application of our GABA mAGNET by performing simultaneous, in vivo optogenetic control of two distinct neuron subtypes. This interneuron labeling tool highlights the potential of microRNA-based viral gene targeting to specific neuron subtypes.