Adult neurogenesis acts as a neural regularizer.

New neurons are continuously generated in the subgranular zone of the dentate gyrus throughout adulthood. These new neurons gradually integrate into hippocampal circuits, forming new naive synapses. Viewed from this perspective, these new neurons may represent a significant source of "wiring" noise in hippocampal networks. In machine learning, such noise injection is commonly used as a regularization technique. Regularization techniques help prevent overfitting training data and allow models to generalize learning to new, unseen data. Using a computational modeling approach, here we ask whether a neurogenesis-like process similarly acts as a regularizer, facilitating generalization in a category learning task. In a convolutional neural network (CNN) trained on the CIFAR-10 object recognition dataset, we modeled neurogenesis as a replacement/turnover mechanism, where weights for a randomly chosen small subset of hidden layer neurons were reinitialized to new values as the model learned to categorize 10 different classes of objects. We found that neurogenesis enhanced generalization on unseen test data compared to networks with no neurogenesis. Moreover, neurogenic networks either outperformed or performed similarly to networks with conventional noise injection (i.e., dropout, weight decay, and neural noise). These results suggest that neurogenesis can enhance generalization in hippocampal learning through noise injection, expanding on the roles that neurogenesis may have in cognition.

Excitability mediates allocation of pre-configured ensembles to a hippocampal engram supporting contextual conditioned threat in mice.

Little is understood about how engrams, sparse groups of neurons that store memories, are formed endogenously. Here, we combined calcium imaging, activity tagging, and optogenetics to examine the role of neuronal excitability and pre-existing functional connectivity on the allocation of mouse cornu ammonis area 1 (CA1) hippocampal neurons to an engram ensemble supporting a contextual threat memory. Engram neurons (high activity during recall or TRAP2-tagged during training) were more active than non-engram neurons 3 h (but not 24 h to 5 days) before training. Consistent with this, optogenetically inhibiting scFLARE2-tagged neurons active in homecage 3 h, but not 24 h, before conditioning disrupted memory retrieval, indicating that neurons with higher pre-training excitability were allocated to the engram. We also observed stable pre-configured functionally connected sub-ensembles of neurons whose activity cycled over days. Sub-ensembles that were more active before training were allocated to the engram, and their functional connectivity increased at training. Therefore, both neuronal excitability and pre-configured functional connectivity mediate allocation to an engram ensemble.

A shift in the mechanisms controlling hippocampal engram formation during brain maturation.

The ability to form precise, episodic memories develops with age, with young children only able to form gist-like memories that lack precision. The cellular and molecular events in the developing hippocampus that underlie the emergence of precise, episodic-like memory are unclear. In mice, the absence of a competitive neuronal engram allocation process in the immature hippocampus precluded the formation of sparse engrams and precise memories until the fourth postnatal week, when inhibitory circuits in the hippocampus mature. This age-dependent shift in precision of episodic-like memories involved the functional maturation of parvalbumin-expressing interneurons in subfield CA1 through assembly of extracellular perineuronal nets, which is necessary and sufficient for the onset of competitive neuronal allocation, sparse engram formation, and memory precision.

Restoration of hippocampal neural precursor function by ablation of senescent cells in the aging stem cell niche.

Senescent cells are responsible, in part, for tissue decline during aging. Here, we focused on CNS neural precursor cells (NPCs) to ask if this is because senescent cells in stem cell niches impair precursor-mediated tissue maintenance. We demonstrate an aging-dependent accumulation of senescent cells, largely senescent NPCs, within the hippocampal stem cell niche coincident with declining adult neurogenesis. Pharmacological ablation of senescent cells via acute systemic administration of the senolytic drug ABT-263 (Navitoclax) caused a rapid increase in NPC proliferation and neurogenesis. Genetic ablation of senescent cells similarly activated hippocampal NPCs. This acute burst of neurogenesis had long-term effects in middle-aged mice. One month post-ABT-263, adult-born hippocampal neuron numbers increased and hippocampus-dependent spatial memory was enhanced. These data support a model where senescent niche cells negatively influence neighboring non-senescent NPCs during aging, and ablation of these senescent cells partially restores neurogenesis and hippocampus-dependent cognition.

An inhibitory hippocampal-thalamic pathway modulates remote memory retrieval.

Memories are supported by distributed hippocampal-thalamic-cortical networks, but the brain regions that contribute to network activity may vary with memory age. This process of reorganization is referred to as systems consolidation, and previous studies have examined the relationship between the activation of different hippocampal, thalamic, and cortical brain regions and memory age at the time of recall. While the activation of some brain regions increases with memory age, other regions become less active. In mice, here we show that the active disengagement of one such brain region, the anterodorsal thalamic nucleus, is necessary for recall at remote time-points and, in addition, which projection(s) mediate such inhibition. Specifically, we identified a sparse inhibitory projection from CA3 to the anterodorsal thalamic nucleus that becomes more active during systems consolidation, such that it is necessary for contextual fear memory retrieval at remote, but not recent, time-points post-learning.

Automated Curation of CNMF-E-Extracted ROI Spatial Footprints and Calcium Traces Using Open-Source AutoML Tools.

In vivo 1-photon (1p) calcium imaging is an increasingly prevalent method in behavioral neuroscience. Numerous analysis pipelines have been developed to improve the reliability and scalability of pre-processing and ROI extraction for these large calcium imaging datasets. Despite these advancements in pre-processing methods, manual curation of the extracted spatial footprints and calcium traces of neurons remains important for quality control. Here, we propose an additional semi-automated curation step for sorting spatial footprints and calcium traces from putative neurons extracted using the popular constrained non-negative matrixfactorization for microendoscopic data (CNMF-E) algorithm. We used the automated machine learning (AutoML) tools TPOT and AutoSklearn to generate classifiers to curate the extracted ROIs trained on a subset of human-labeled data. AutoSklearn produced the best performing classifier, achieving an F1 score >92% on the ground truth test dataset. This automated approach is a useful strategy for filtering ROIs with relatively few labeled data points and can be easily added to pre-existing pipelines currently using CNMF-E for ROI extraction.

Forgetting at biologically realistic levels of neurogenesis in a large-scale hippocampal model.

Neurogenesis persists throughout life in the dentate gyrus region of the mammalian hippocampus. Computational models have established that the addition of neurons degrades existing memories (i.e., produces forgetting). These predictions are supported by empirical observations in rodents, where post-training increases in neurogenesis also promote forgetting of hippocampus-dependent memories. However, in these computational models which use 10-1,000 neurons to represent the dentate gyrus, forgetting is only observed at rates of new neuron addition that greatly exceed adult neurogenesis rates observed in vivo. In order to address this, here we generated an artificial neural network which incorporated more realistic features of the hippocampus - including increased network size (with up to 20,000 dentate gyrus neurons), sparse activity, and sparse connectivity - features that were not present in earlier models. In addition, we explored how properties of new neurons - their connectivity, excitability, and plasticity - impact forgetting using a pattern categorization task. Our results revealed that neurogenic networks forget previously learned input-output pattern associations. This forgetting predicted a performance enhancement in subsequent conflictual learning, compared to static networks (with no added neurons). These effects were especially sensitive to changes in increased output connectivity and excitability of new neurons. Crucially, forgetting was observed at much lower rates of neurogenesis in larger networks, with the addition of as little as 0.2% of the total DG population sufficient to induce forgetting.

Memory formation in the absence of experience.

Memory is coded by patterns of neural activity in distinct circuits. Therefore, it should be possible to reverse engineer a memory by artificially creating these patterns of activity in the absence of a sensory experience. In olfactory conditioning, an odor conditioned stimulus (CS) is paired with an unconditioned stimulus (US; for example, a footshock), and the resulting CS-US association guides future behavior. Here we replaced the odor CS with optogenetic stimulation of a specific olfactory glomerulus and the US with optogenetic stimulation of distinct inputs into the ventral tegmental area that mediate either aversion or reward. In doing so, we created a fully artificial memory in mice. Similarly to a natural memory, this artificial memory depended on CS-US contingency during training, and the conditioned response was specific to the CS and reflected the US valence. Moreover, both real and implanted memories engaged overlapping brain circuits and depended on basolateral amygdala activity for expression.

A Compact Head-Mounted Endoscope for In Vivo Calcium Imaging in Freely Behaving Mice.

Miniaturized fluorescence microscopes for imaging calcium transients are a promising tool for investigating the relationship between behavior and population-level neuronal activity in rodents. However, commercially available miniature microscopes may be costly and, because they are closed source, may not be easily modified based on particular experimental requirements. Here, we describe how to build and use a low-cost compact head-mounted endoscope (CHEndoscope) system for in vivo calcium imaging. The CHEndoscope uses an implanted gradient index lens along with the genetically encoded calcium indicator GCaMP6 to image calcium transients from hundreds of neurons simultaneously in awake behaving mice. This system is affordable, open source, and flexible, permitting modification depending on the particular experiment. This article describes in detail the assembly, surgical implantation, data collection, and processing of calcium signals using the CHEndoscope system. The aim of this open framework is to provide an accessible set of miniaturized calcium imaging tools for the neuroscience research community. © 2018 by John Wiley & Sons, Inc.

Chemogenetic Interrogation of a Brain-wide Fear Memory Network in Mice.

Behavior depends on coordinated activity across multiple brain regions. Within such networks, highly connected hub regions are assumed to disproportionately influence behavioral output, although this hypothesis has not been systematically evaluated. Previously, by mapping brain-wide expression of the activity-regulated gene c-fos, we identified a network of brain regions co-activated by fear memory. To test the hypothesis that hub regions are more important for network function, here, we simulated node deletion in silico in this behaviorally defined functional network. Removal of high degree nodes produced the greatest network disruption (e.g., reduction in global efficiency). To test these predictions in vivo, we examined the impact of post-training chemogenetic silencing of different network nodes on fear memory consolidation. In a series of independent experiments encompassing 25% of network nodes (i.e., 21/84 brain regions), we found that node degree accurately predicted observed deficits in memory consolidation, with silencing of highly connected hubs producing the largest impairments.