Examining a punishment-related brain circuit with miniature fluorescence microscopes and deep learning.

In humans experiencing substance use disorder (SUD), abstinence from drug use is often motivated by a desire to avoid some undesirable consequence of further use: health effects, legal ramifications, etc. This process can be experimentally modeled in rodents by training and subsequently punishing an operant response in a context-induced reinstatement procedure. Understanding the biobehavioral mechanisms underlying punishment learning is critical to understanding both abstinence and relapse in individuals with SUD. To date, most investigations into the neural mechanisms of context-induced reinstatement following punishment have utilized discrete loss-of-function manipulations that do not capture ongoing changes in neural circuitry related to punishment-induced behavior change. Here, we describe a two-pronged approach to analyzing the biobehavioral mechanisms of punishment learning using miniature fluorescence microscopes and deep learning algorithms. We review recent advancements in both techniques and consider a target neural circuit.

A modular, cost-effective, versatile, open-source operant box solution for long-term miniscope imaging, 3D tracking, and deep learning behavioral analysis.

In this procedure we have included an open-source method for a customized operant chamber optimized for long-term miniature microscope (miniscope) recordings. •The miniscope box is designed to function with custom or typical med-associates style accessories (e.g., houselights, levers, etc.).•The majority of parts can be directly purchased which minimizes the need for skilled and time-consuming labor.•We include designs and estimated pricing for a single box but it is recommended to build these in larger batches to efficiently utilize bulk ordering of certain components.

Jump-GRS: a multi-phase approach to structured pruning of neural networks for neural decoding.

Objective.Neural decoding, an important area of neural engineering, helps to link neural activity to behavior. Deep neural networks (DNNs), which are becoming increasingly popular in many application fields of machine learning, show promising performance in neural decoding compared to traditional neural decoding methods. Various neural decoding applications, such as brain computer interface applications, require both high decoding accuracy and real-time decoding speed. Pruning methods are used to produce compact DNN models for faster computational speed. Greedy inter-layer order with Random Selection (GRS) is a recently-designed structured pruning method that derives compact DNN models for calcium-imaging-based neural decoding. Although GRS has advantages in terms of detailed structure analysis and consideration of both learned information and model structure during the pruning process, the method is very computationally intensive, and is not feasible when large-scale DNN models need to be pruned within typical constraints on time and computational resources. Large-scale DNN models arise in neural decoding when large numbers of neurons are involved. In this paper, we build on GRS to develop a new structured pruning algorithm called jump GRS (JGRS) that is designed to efficiently compress large-scale DNN models.Approach.On top of GRS, JGRS implements a 'jump mechanism', which bypasses retraining intermediate models when model accuracy is relatively less sensitive to pruning operations. Design of the jump mechanism is motivated by identifying different phases of the structured pruning process, where retraining can be done infrequently in earlier phases without sacrificing accuracy. The jump mechanism helps to significantly speed up execution of the pruning process and greatly enhance its scalability. We compare the pruning performance and speed of JGRS and GRS with extensive experiments in the context of neural decoding.Main results.Our results demonstrate that JGRS provides significantly faster pruning speed compared to GRS, and at the same time, JGRS provides pruned models that are similarly compact as those generated by GRS.Significance.In our experiments, we demonstrate that JGRS achieves on average 9%-20% more compressed models compared to GRS with 2-8 times faster speed (less time required for pruning) across four different initial models on a relevant dataset for neural data analysis.

Neural ensembles in the murine medial prefrontal cortex process distinct information during visual perceptual learning.

BACKGROUND: Perceptual learning refers to an augmentation of an organism's ability to respond to external stimuli, which has been described in most sensory modalities. Visual perceptual learning (VPL) is a manifestation of plasticity in visual information processing that occurs in the adult brain, and can be used to ameliorate the ability of patients with visual defects mainly based on an improvement of detection or discrimination of features in visual tasks. While some brain regions such as the primary visual cortex have been described to participate in VPL, the way more general high-level cognitive brain areas are involved in this process remains unclear. Here, we showed that the medial prefrontal cortex (mPFC) was essential for both the training and maintenance processes of VPL in mouse models. RESULTS: We built a new VPL model in a custom-designed training chamber to enable the utilization of miniScopes when mice freely executed the VPL task. We found that pyramidal neurons in the mPFC participate in both the training process and maintenance of VPL. By recording the calcium activity of mPFC pyramidal neurons while mice freely executed the task, distinct ON and OFF neural ensembles tuned to different behaviors were identified, which might encode different cognitive information. Decoding analysis showed that mouse behaviors could be well predicted using the activity of each ON ensemble. Furthermore, VPL recruited more reward-related components in the mPFC. CONCLUSION: We revealed the neural mechanism underlying vision improvement following VPL and identify distinct ON and OFF neural ensembles in the mPFC that tuned to different information during visual perceptual training. These results uncover an important role of the mPFC in VPL, with more reward-related components being also involved, and pave the way for future clarification of the reward signal coding rules in VPL.

Using deep learning to study emotional behavior in rodent models.

Quantifying emotional aspects of animal behavior (e.g., anxiety, social interactions, reward, and stress responses) is a major focus of neuroscience research. Because manual scoring of emotion-related behaviors is time-consuming and subjective, classical methods rely on easily quantified measures such as lever pressing or time spent in different zones of an apparatus (e.g., open vs. closed arms of an elevated plus maze). Recent advancements have made it easier to extract pose information from videos, and multiple approaches for extracting nuanced information about behavioral states from pose estimation data have been proposed. These include supervised, unsupervised, and self-supervised approaches, employing a variety of different model types. Representations of behavioral states derived from these methods can be correlated with recordings of neural activity to increase the scope of connections that can be drawn between the brain and behavior. In this mini review, we will discuss how deep learning techniques can be used in behavioral experiments and how different model architectures and training paradigms influence the type of representation that can be obtained.

GRIN lens applications for studying neurobiology of substance use disorder.

Substance use disorder (SUD) is associated with severe health and social consequences. Continued drug use results in alterations of circuits within the mesolimbic dopamine system. It is critical to observe longitudinal impacts of SUD on neural activity in vivo to identify SUD predispositions, develop pharmaceuticals to prevent overdose, and help people suffering from SUD. However, implicated SUD associated areas are buried in deep brain which makes in vivo observation of neural activity challenging. The gradient index (GRIN) lens can probe these regions in mice and rats. In this short communications review, we will discuss how the GRIN lens can be coupled with other technologies such as miniaturized microscopes, fiberscopes, fMRI, and optogenetics to fully explore in vivo SUD research. Particularly, GRIN lens allows in vivo observation of deep brain regions implicated in SUD, differentiation of genetically distinct neurons, examination of individual cells longitudinally, correlation of neuronal patters with SUD behavior, and manipulation of neural circuits.

A miniature fluorescence microscope for multi-plane imaging.

Miniature fluorescence microscopes are becoming an increasingly established tool to investigate neural circuits in freely moving animals. In this work we present a lightweight one-photon microscope capable of imaging at different focal depths. The focal plane can be changed dynamically by modulating the pulse width of the control signal to a variable focus liquid lens, which is synchronized to the image sensor to enable changing focal plane between frames. The system was tested by imaging GCaMP7f expressing neurons in the mouse medial prefrontal cortex (mPFC) in vivo during open field test. Results showed that with the proposed design it is possible to image neurons across an axial scan of ~ 60 µm, resulting in a ~ 40% increase of total neurons imaged compared to single plane imaging.

Dysbindin-1, BDNF, and GABAergic Transmission in Schizophrenia.

Schizophrenia is a psychiatric disorder characterized by hallucinations, anhedonia, disordered thinking, and cognitive impairments. Both genetic and environmental factors contribute to schizophrenia. Dysbindin-1 (DTNBP1) and brain-derived neurotrophic factor (BDNF) are both genetic factors associated with schizophrenia. Mice lacking Dtnbp1 showed behavioral deficits similar to human patients suffering from schizophrenia. DTNBP1 plays important functions in synapse formation and maintenance, receptor trafficking, and neurotransmitter release. DTNBP1 is co-assembled with 7 other proteins into a large protein complex, known as the biogenesis of lysosome-related organelles complex-1 (BLOC-1). Large dense-core vesicles (LDCVs) are involved in the secretion of hormones and neuropeptides, including BDNF. BDNF plays important roles in neuronal development, survival, and synaptic plasticity. BDNF is also critical in maintaining GABAergic inhibitory transmission in the brain. Two studies independently showed that DTNBP1 mediated activity-dependent BDNF secretion to maintain inhibitory transmission. Imbalance of excitatory and inhibitory neural activities is thought to contribute to schizophrenia. In this mini-review, we will discuss a potential pathogenetic mechanism for schizophrenia involving DTNBP1, BDNF, and inhibitory transmission. We will also discuss how these processes are interrelated and associated with a higher risk of schizophrenia development.

Aberrant neural activity in prefrontal pyramidal neurons lacking TDP-43 precedes neuron loss.

Mislocalization of TAR DNA binding protein 43 kDa (TARDBP, or TDP-43) is a principal pathological hallmark identified in cases of neurodegenerative disorders such as amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). As an RNA binding protein, TDP-43 serves in the nuclear compartment to repress non-conserved cryptic exons to ensure the normal transcriptome. Multiple lines of evidence from animal models and human studies support the view that loss of TDP-43 leads to neuron loss, independent of its cytosolic aggregation. However, the underlying pathogenic pathways driven by the loss-of-function mechanism are still poorly defined. We employed a genetic approach to determine the impact of TDP-43 loss in pyramidal neurons of the prefrontal cortex (PFC). Using a custom-built miniscope imaging system, we performed repetitive in vivo calcium imaging from freely behaving mice for up to 7 months. By comparing calcium activity in PFC pyramidal neurons between TDP-43 depleted and TDP-43 intact mice, we demonstrated remarkably increased numbers of pyramidal neurons exhibiting hyperactive calcium activity after short-term TDP-43 depletion, followed by rapid activity declines prior to neuron loss. Our results suggest aberrant neural activity driven by loss of TDP-43 as the pathogenic pathway at early stage in ALS and FTD.

Striatal direct pathway neurons play leading roles in accelerating rotarod motor skill learning.

Dorsal striatum is important for movement control and motor skill learning. However, it remains unclear how the spatially and temporally distributed striatal medium spiny neuron (MSN) activity in the direct and indirect pathways (D1 and D2 MSNs, respectively) encodes motor skill learning. Combining miniature fluorescence microscopy with an accelerating rotarod procedure, we identified two distinct MSN subpopulations involved in accelerating rotarod learning. In both D1 and D2 MSNs, we observed neurons that displayed activity tuned to acceleration during early stages of trials, as well as movement speed during late stages of trials. We found a distinct evolution trajectory for early-stage neurons during motor skill learning, with the evolution of D1 MSNs correlating strongly with performance improvement. Importantly, optogenetic inhibition of the early-stage neural activity in D1 MSNs, but not D2 MSNs, impaired accelerating rotarod learning. Together, this study provides insight into striatal D1 and D2 MSNs encoding motor skill learning.

Characterization of operant social interaction in rats: effects of access duration, effort, peer familiarity, housing conditions, and choice between social interaction vs. food or remifentanil.

RATIONALE AND OBJECTIVE: Social factors play a critical role in drug addiction. We recently showed that rats will abstain from methamphetamine, cocaine, heroin, and remifentanil self-administration when given a choice between the addictive drug and operant social interaction. Here, we further characterized operant social interaction by determining the effects of access duration, effort, peer familiarity, and housing conditions. We also determined choice between social interaction vs. palatable food or remifentanil. METHODS: We first trained single-housed male and female rats to lever-press for social interaction with a sex- and age-matched peer. Next, we determined effects of access duration (3.75 to 240 s), effort (increasing fixed-ratio schedule requirements or progressive ratio schedule), peer familiarity (familiar vs. unfamiliar), and housing conditions (single vs. paired housing) on social self-administration. We also determined choice between social interaction vs. palatable food pellets or intravenous remifentanil (0, 1, 10 µg/kg/infusion). RESULTS: Increasing access duration to a peer decreased social self-administration under fixed ratio but not progressive ratio schedule; the rats showed similar preference for short vs. long access duration. Social self-administration under different fixed ratio requirements was higher in single-housed than in paired-housed rats and higher for a familiar vs. unfamiliar partner in single-housed but not paired-housed rats. Response rates of food-sated rats under increasing fixed-ratio requirements were higher for palatable food than for social interaction. The rats strongly preferred palatable food over social interaction and showed dose-dependent preference for social interaction vs. remifentanil. CONCLUSIONS: We identified parameters influencing the reinforcing effects of operant social interaction and introduce a choice procedure sensitive to remifentanil self-administration dose.

Detailed mapping of behavior reveals the formation of prelimbic neural ensembles across operant learning.

The prelimbic cortex (PrL) is involved in the organization of operant behaviors, but the relationship between longitudinal PrL neural activity and operant learning and performance is unknown. Here, we developed deep behavior mapping (DBM) to identify behavioral microstates in video recordings. We combined DBM with longitudinal calcium imaging to quantify behavioral tuning in PrL neurons as mice learned an operant task. We found that a subset of PrL neurons were strongly tuned to highly specific behavioral microstates, both task and non-task related. Overlapping neural ensembles were tiled across consecutive microstates in the response-reinforcer sequence, forming a continuous map. As mice learned the operant task, weakly tuned neurons were recruited into new ensembles, with a bias toward behaviors similar to their initial tuning. In summary, our data suggest that the PrL contains neural ensembles that jointly encode a map of behavioral states that is fine grained, is continuous, and grows during operant learning.

LEARNING COMPACT DNN MODELS FOR BEHAVIOR PREDICTION FROM NEURAL ACTIVITY OF CALCIUM IMAGING.

In this paper, we develop methods for efficient and accurate information extraction from calcium-imaging-based neural signals. The particular form of information extraction we investigate involves predicting behavior variables linked to animals from which the calcium imaging signals are acquired. More specifically, we develop algorithms to systematically generate compact deep neural network (DNN) models for accurate and efficient calcium-imaging-based predictive modeling. We also develop a software tool, called NeuroGRS, to apply the proposed methods for compact DNN derivation with a high degree of automation. GRS stands for Greedy inter-layer order with Random Selection of intra-layer units, which describes the central algorithm developed in this work for deriving compact DNN structures. Through extensive experiments using NeuroGRS and calcium imaging data, we demonstrate that our methods enable highly streamlined information extraction from calcium images of the brain with minimal loss in accuracy compared to much more computationally expensive approaches.

Anterograde transneuronal tracing and genetic control with engineered yellow fever vaccine YFV-17D.

Transneuronal viruses are powerful tools for tracing neuronal circuits or delivering genes to specific neurons in the brain. While there are multiple retrograde viruses, few anterograde viruses are available. Further, available anterograde viruses often have limitations such as retrograde transport, high neuronal toxicity or weak signals. We developed an anterograde viral system based on a live attenuated vaccine for yellow fever-YFV-17D. Replication- or packaging-deficient mutants of YFV-17D can be reconstituted in the brain, leading to efficient synapse-specific and anterograde-only transneuronal spreading, which can be controlled to achieve either monosynaptic or polysynaptic tracing. Moreover, inducible transient replication of YFV-17D mutant is sufficient to induce permanent transneuronal genetic modifications without causing neuronal toxicity. The engineered YFV-17D systems can be used to express fluorescent markers, sensors or effectors in downstream neurons, thus providing versatile tools for mapping and functionally controlling neuronal circuits.

Circuit Investigation of Social Interaction and Substance Use Disorder Using Miniscopes.

Substance use disorder (SUD) is comorbid with devastating health issues, social withdrawal, and isolation. Successful clinical treatments for SUD have used social interventions. Neurons can encode drug cues, and drug cues can trigger relapse. It is important to study how the activity in circuits and embedded cell types that encode drug cues develop in SUD. Exploring shared neurobiology between social interaction (SI) and SUD may explain why humans with access to social treatments still experience relapse. However, circuitry remains poorly characterized due to technical challenges in studying the complicated nature of SI and SUD. To understand the neural correlates of SI and SUD, it is important to: (1) identify cell types and circuits associated with SI and SUD, (2) record and manipulate neural activity encoding drug and social rewards over time, (3) monitor unrestrained animal behavior that allows reliable drug self-administration (SA) and SI. Miniaturized fluorescence microscopes (miniscopes) are ideally suited to meet these requirements. They can be used with gradient index (GRIN) lenses to image from deep brain structures implicated in SUD. Miniscopes can be combined with genetically encoded reporters to extract cell-type specific information. In this mini-review, we explore how miniscopes can be leveraged to uncover neural components of SI and SUD and advance potential therapeutic interventions.

Neural Decoding on Imbalanced Calcium Imaging Data with a Network of Support Vector Machines.

We present a novel neural decoding system for calcium imaging data. Miniature calcium imaging is of great utility for examining population neural activity of animals. Our neural decoding system is developed using a carefully-designed support vector machine subsystem together with dataflow-based techniques for system design, which capture the high-level structure of the application and enable powerful system-level analysis and optimization. Also, we introduce a framework for handling imbalanced data. This addresses a problem of imbalanced datasets, which arises commonly in neural decoding applications, as well as in a wide variety of other applications in biomedical engineering and advanced robotics. We developed an ensemble learning based method to tackle this problem. The proposed framework systemically incorporates two heterogeneous model characteristics into a combined model. Through extensive experiments, we evaluate the proposed system using calcium imaging datasets in which neural activities of D1 medium spiny neurons in the dorsal striatum were recorded. The results show that the F 1 score of the proposed system is significantly better than those of previously developed neural decoding systems for calcium imaging.

An open source motorized swivel for in vivo neural and behavioral recordings.

In this work we propose an open source, cost-effective motorized swivel for behavioral and neural recordings in small rodents, offering a flexible solution for managing cable twisting and tangling in a variety of experimental settings with minimal human supervision.•The device operates independently of the data acquisition system, and it can be controlled through any popular platform such as Arduino or Raspberry Pi.•All mechanical parts are 3D-printed, allowing to customize the design to fit specific experimental needs, and electromechanical components can be sourced from all major distributors, keeping the cost for the entire system under $500.•The proposed commutator is compatible with commercial or custom data acquisition systems supporting up to 10 data lines (2 for LVDS signals) and 2 power lines.

An open-source capacitive touch sensing device for three chamber social behavior test.

A common feature of many neuropsychiatric disorders is deficit in social behavior. In order to study mouse models for such disorders, several behavioral tests involving social interaction with other mice have been developed. While a precise annotation of rodent behavioral state is necessary for these types of experiments, manual annotation of rodent social behavior is time-consuming and subjective. Therefore, an automated system that can instantly and independently quantify the animal's social exploration is desirable. We developed a capacitive touch device for automated detection of direct social-exploration in a modified three-chamber social behavior test. In this device, capacitive sensors can readily detect nose-pokes and other direct physical touches from the rodent under investigation. In addition, a conductive barrier makes mouse behavioral output immediately available for real-time use, by sending data to a host computer via a custom Field-Programmable Gate Array (FPGA) platform. Our capacitive touch sensing device produced similar results to the manually annotated data, demonstrating the ability to instantly and independently analyze direct social-exploration of animals in a social behavior test. Compared to the manual annotation method, this capacitive touch sensing system can be used to instantaneously quantify direct social-exploration, saving significant amount of time of post-hoc video scoring. Furthermore, this low-cost method enhances the objectivity of data by reducing experimenter involvement in analysis.

Real-Time Neuron Detection and Neural Signal Extraction Platform for Miniature Calcium Imaging.

Real-time neuron detection and neural activity extraction are critical components of real-time neural decoding. In this paper, we propose a novel real-time neuron detection and activity extraction system using a dataflow framework to provide real-time performance and adaptability to new algorithms and hardware platforms. The proposed system was evaluated on simulated calcium imaging data, calcium imaging data with manual annotation, and calcium imaging data of the anterior lateral motor cortex. We found that the proposed system accurately detected neurons and extracted neural activities in real time without any requirement for expensive, cumbersome, or special-purpose computing hardware. We expect that the system will enable cost-effective, real-time calcium imaging-based neural decoding, leading to precise neuromodulation.

Circuit Mechanisms of Neurodegenerative Diseases: A New Frontier With Miniature Fluorescence Microscopy.

Neurodegenerative diseases (NDDs), such as Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease (HD), amyotrophic lateral sclerosis (ALS), and frontotemporal dementia (FTD), are devastating age-associated brain disorders. Significant efforts have been made to uncover the molecular and cellular pathogenic mechanisms that underlie NDDs. However, our understanding of the neural circuit mechanisms that mediate NDDs and associated symptomatic features have been hindered by technological limitations. Our inability to identify and track individual neurons longitudinally in subcortical brain regions that are preferentially targeted in NDDs has left gaping holes in our knowledge of NDDs. Recent development and advancement of the miniature fluorescence microscope (miniscope) has opened up new avenues for examining spatially and temporally coordinated activity from hundreds of cells in deep brain structures in freely moving rodents. In the present mini-review, we examine the capabilities of current and future miniscope tools and discuss the innovative applications of miniscope imaging techniques that can push the boundaries of our understanding of neural circuit mechanisms of NDDs into new territories.