Robust single-shot 3D fluorescence imaging in scattering media with a simulator-trained neural network.

Imaging through scattering is a pervasive and difficult problem in many biological applications. The high background and the exponentially attenuated target signals due to scattering fundamentally limits the imaging depth of fluorescence microscopy. Light-field systems are favorable for high-speed volumetric imaging, but the 2D-to-3D reconstruction is fundamentally ill-posed, and scattering exacerbates the condition of the inverse problem. Here, we develop a scattering simulator that models low-contrast target signals buried in heterogeneous strong background. We then train a deep neural network solely on synthetic data to descatter and reconstruct a 3D volume from a single-shot light-field measurement with low signal-to-background ratio (SBR). We apply this network to our previously developed computational miniature mesoscope and demonstrate the robustness of our deep learning algorithm on scattering phantoms with different scattering conditions. The network can robustly reconstruct emitters in 3D with a 2D measurement of SBR as low as 1.05 and as deep as a scattering length. We analyze fundamental tradeoffs based on network design factors and out-of-distribution data that affect the deep learning model's generalizability to real experimental data. Broadly, we believe that our simulator-based deep learning approach can be applied to a wide range of imaging through scattering techniques where experimental paired training data is lacking.

Targeted micro-fiber arrays for measuring and manipulating localized multi-scale neural dynamics over large, deep brain volumes during behavior.

Neural population dynamics relevant to behavior vary over multiple spatial and temporal scales across three-dimensional volumes. Current optical approaches lack the spatial coverage and resolution necessary to measure and manipulate naturally occurring patterns of large-scale, distributed dynamics within and across deep brain regions such as the striatum. We designed a new micro-fiber array approach capable of chronically measuring and optogenetically manipulating local dynamics across over 100 targeted locations simultaneously in head-fixed and freely moving mice, enabling the investigation of cell-type- and neurotransmitter-specific signals over arbitrary 3D volumes at a spatial resolution and coverage previously inaccessible. We applied this method to resolve rapid dopamine release dynamics across the striatum, revealing distinct, modality-specific spatiotemporal patterns in response to salient sensory stimuli extending over millimeters of tissue. Targeted optogenetics enabled flexible control of neural signaling on multiple spatial scales, better matching endogenous signaling patterns, and the spatial localization of behavioral function across large circuits.

Targeted micro-fiber arrays for measuring and manipulating localized multi-scale neural dynamics over large, deep brain volumes during behavior.

Neural population dynamics relevant for behavior vary over multiple spatial and temporal scales across 3-dimensional volumes. Current optical approaches lack the spatial coverage and resolution necessary to measure and manipulate naturally occurring patterns of large-scale, distributed dynamics within and across deep brain regions such as the striatum. We designed a new micro-fiber array and imaging approach capable of chronically measuring and optogenetically manipulating local dynamics across over 100 targeted locations simultaneously in head-fixed and freely moving mice. We developed a semi-automated micro-CT based strategy to precisely localize positions of each optical fiber. This highly-customizable approach enables investigation of multi-scale spatial and temporal patterns of cell-type and neurotransmitter specific signals over arbitrary 3-D volumes at a spatial resolution and coverage previously inaccessible. We applied this method to resolve rapid dopamine release dynamics across the striatum volume which revealed distinct, modality specific spatiotemporal patterns in response to salient sensory stimuli extending over millimeters of tissue. Targeted optogenetics through our fiber arrays enabled flexible control of neural signaling on multiple spatial scales, better matching endogenous signaling patterns, and spatial localization of behavioral function across large circuits.

Robust single-shot 3D fluorescence imaging in scattering media with a simulator-trained neural network.

Imaging through scattering is a pervasive and difficult problem in many biological applications. The high background and the exponentially attenuated target signals due to scattering fundamentally limits the imaging depth of fluorescence microscopy. Light-field systems are favorable for high-speed volumetric imaging, but the 2D-to-3D reconstruction is fundamentally ill-posed, and scattering exacerbates the condition of the inverse problem. Here, we develop a scattering simulator that models low-contrast target signals buried in heterogeneous strong background. We then train a deep neural network solely on synthetic data to descatter and reconstruct a 3D volume from a single-shot light-field measurement with low signal-to-background ratio (SBR). We apply this network to our previously developed Computational Miniature Mesoscope and demonstrate the robustness of our deep learning algorithm on scattering phantoms with different scattering conditions. The network can robustly reconstruct emitters in 3D with a 2D measurement of SBR as low as 1.05 and as deep as a scattering length. We analyze fundamental tradeoffs based on network design factors and out-of-distribution data that affect the deep learning model's generalizability to real experimental data. Broadly, we believe that our simulator-based deep learning approach can be applied to a wide range of imaging through scattering techniques where experimental paired training data is lacking.

Visual Cues Predictive of Behaviorally Neutral Outcomes Evoke Persistent but Not Interval Timing Activity in V1, Whereas Aversive Conditioning Suppresses This Activity.

Cue-evoked persistent activity is neural activity that persists beyond stimulation of a sensory cue and has been described in many regions of the brain, including primary sensory areas. Nonetheless, the functional role that persistent activity plays in primary sensory areas is enigmatic. However, one form of persistent activity in a primary sensory area is the representation of time between a visual stimulus and a water reward. This "reward timing activity"-observed within the primary visual cortex-has been implicated in informing the timing of visually cued, reward-seeking actions. Although rewarding outcomes are sufficient to engender interval timing activity within V1, it is unclear to what extent cue-evoked persistent activity exists outside of reward conditioning, and whether temporal relationships to other outcomes (such as behaviorally neutral or aversive outcomes) are able to engender timing activity. Here we describe the existence of cue-evoked persistent activity in mouse V1 following three conditioning strategies: pseudo-conditioning (where unpaired, monocular visual stimuli are repeatedly presented to an animal), neutral conditioning (where monocular visual stimuli are paired with a binocular visual stimulus, at a delay), and aversive conditioning (where monocular visual stimuli are paired with a tail shock, at a delay). We find that these conditioning strategies exhibit persistent activity that takes one of three forms, a sustained increase of activity; a sustained decrease of activity; or a delayed, transient peak of activity, as previously observed following conditioning with delayed reward. However, these conditioning strategies do not result in visually cued interval timing activity, as observed following appetitive conditioning. Moreover, we find that neutral conditioning increases the magnitude of cue-evoked responses whereas aversive conditioning strongly diminished both the response magnitude and the prevalence of cue-evoked persistent activity. These results demonstrate that cue-evoked persistent activity within V1 can exist outside of conditioning visual stimuli with delayed outcomes and that this persistent activity can be uniquely modulated across different conditioning strategies using unconditioned stimuli of varying behavioral relevance. Together, these data extend our understanding of cue-evoked persistent activity within a primary sensory cortical network and its ability to be modulated by salient outcomes.

Reward Timing and Its Expression by Inhibitory Interneurons in the Mouse Primary Visual Cortex.

The primary sensory cortex has historically been studied as a low-level feature detector, but has more recently been implicated in many higher-level cognitive functions. For instance, after an animal learns that a light predicts water at a fixed delay, neurons in the primary visual cortex (V1) can produce "reward timing activity" (i.e., spike modulation of various forms that relate the interval between the visual stimulus and expected reward). Local manipulations to V1 implicate it as a site of learning reward timing activity (as opposed to simply reporting timing information from another region via feedback input). However, the manner by which V1 then produces these representations is unknown. Here, we combine behavior, in vivo electrophysiology, and optogenetics to investigate the characteristics of and circuit mechanisms underlying V1 reward timing in the head-fixed mouse. We find that reward timing activity is present in mouse V1, that inhibitory interneurons participate in reward timing, and that these representations are consistent with a theorized network architecture. Together, these results deepen our understanding of V1 reward timing and the manner by which it is produced.

Are You There, Cortex? It's Me, Acetylcholine.

It is not well understood how associations between two temporally removed stimuli form. In this issue of Neuron, Guo et al. (2019) implicate basal forebrain cholinergic neurons as providing a link between auditory cues and the aversive outcomes they predict.

Visually cued action timing in the primary visual cortex.

Most behaviors are generated in three steps: sensing the external world, processing that information to instruct decision-making, and producing a motor action. Sensory areas, especially primary sensory cortices, have long been held to be involved only in the first step of this sequence. Here, we develop a visually cued interval timing task that requires rats to decide when to perform an action following a brief visual stimulus. Using single-unit recordings and optogenetics in this task, we show that activity generated by the primary visual cortex (V1) embodies the target interval and may instruct the decision to time the action on a trial-by-trial basis. A spiking neuronal model of local recurrent connections in V1 produces neural responses that predict and drive the timing of future actions, rationalizing our observations. Our data demonstrate that the primary visual cortex may contribute to the instruction of visually cued timed actions.

Licking microstructure reveals rapid attenuation of neophobia.

Many animals hesitate when initially consuming a novel food and increase their consumption of that food between the first and second sessions of access-a process termed attenuation of neophobia (AN). AN has received attention as a model of learning and memory; it has been suggested that plasticity resulting from an association of the novel tastant with "safe outcome" results in a change in the neural response to the tastant during the second session, such that consumption increases. Most studies have reported that AN emerges only an hour or more after the end of the first exposure to the tastant, consistent with what is known of learning-related plasticity. But these studies have typically measured consumption, rather than real-time behavior, and thus the possibility exists that a more rapidly developing AN remains to be discovered. Here, we tested this possibility, examining both consumption and individual lick times in a novel variant of a brief-access task (BAT). When quantified in terms of consumption, data from the BAT accorded well with the results of a classic one-bottle task-both revealed neophobia/AN specific to higher concentrations (for instance, 28mM) of saccharin. An analysis of licking microstructure, however, additionally revealed a real-time correlate of neophobia-an explicit tendency, similarly specific for 28-mM saccharin, to cut short the initial bout of licks in a single trial (compared with water). This relative hesitancy (i.e., the shortness of the first lick bout to 28-mM saccharin compared with water) that constitutes neophobia not only disappeared between sessions but also gradually declined in magnitude across session 1. These data demonstrate that the BAT accurately measures AN, and that aspects of AN-and the processes underlying familiarization-begin within minutes of the very first taste.

Lateral hypothalamus contains two types of palatability-related taste responses with distinct dynamics.

The taste of foods, in particular the palatability of these tastes, exerts a powerful influence on our feeding choices. Although the lateral hypothalamus (LH) has long been known to regulate feeding behavior, taste processing in LH remains relatively understudied. Here, we examined single-unit LH responses in rats subjected to a battery of taste stimuli that differed in both chemical composition and palatability. Like neurons in cortex and amygdala, LH neurons produced a brief epoch of nonspecific responses followed by a protracted period of taste-specific firing. Unlike in cortex, however, where palatability-related information only appears 500 ms after the onset of taste-specific firing, taste specificity in LH was dominated by palatability-related firing, consistent with LH's role as a feeding center. Upon closer inspection, taste-specific LH neurons fell reliably into one of two subtypes: the first type showed a reliable affinity for palatable tastes, low spontaneous firing rates, phasic responses, and relatively narrow tuning; the second type showed strongest modulation to aversive tastes, high spontaneous firing rates, protracted responses, and broader tuning. Although neurons producing both types of responses were found within the same regions of LH, cross-correlation analyses suggest that they may participate in distinct functional networks. Our data shed light on the implementation of palatability processing both within LH and throughout the taste circuit, and may ultimately have implications for LH's role in the formation and maintenance of taste preferences and aversions.