Spatial and single-cell transcriptomics reveal neuron-astrocyte interplay in long-term memory

Background

How we learn and remember things is an integral part of who we are. When we learn something, connections between neurons in our brain start to form, a process referred to as encoding. What we call a “memory” is the reactivation of a specific group of neurons involved in making those connections in the first place. Memory is further divided into short-term memory, information processed in a short period of time, and long-term memory which makes it possible to store information for longer periods of time¹. Astonishingly, the adult human brain has around 86 billion neurons that can each connect to 10,000 other neurons — our brain’s capacity to learn and store information seems nearly endless².

It has been widely accepted that memory consolidation, encoding a memory, and reconsolidation, recalling a memory, is a biological process. In this preprint, Wenfei Sun and his colleagues combined single-cell spatial transcriptomics and single-cell RNA-sequencing (scRNA-seq) to elucidate the role of the basolateral amygdala (BLA) region of the brain in long-term fear memory. Although extensive research has shown that the BLA is implicated in short- and long-term memory formation, the implications of specific BLA cell subsets and transcriptional programs in long-term memory formation remains largely unexplored.

Key findings

BLA engram cell heterogeneity 

The study utilized TRAP2 mice crossed with Ai14 tdTomato reporter mice that express tdT upon stimulation and tamoxifen exposure. At day 0, mice underwent fear conditioning involving the use of an electrical shock. To trigger long-term memory recall, mice received tamoxifen injections at day 16. This experimental set-up allowed the researchers to mark engram cells (tdT+ cells), the cells specifically activated during recall. The BLA of mice were then analyzed by single-cell spatial transcriptomics and scRNA-seq nine days after recall.

Single-cell spatial transcriptomic results revealed that the BLA is highly heterogenous, comprised of astrocytes, microglia, oligodendrocytes, OPCs, endothelial cells, pericytes, as well as excitatory and inhibitory neuronal subtypes. A differential gene expression (DEG) analysis between tdT+ neurons in the No Fear (NF) and Fear training and Recall (FR) conditions revealed that Sv2c (synaptic vesicle) and Penk (neuropeptide) were upregulated in FR inhibitory engram neurons while Tac2 (neuropeptide) was downregulated. In terms of FR excitatory engram neurons, Penk was also upregulated, as well as Dusp1 and Fos: immediate early-genes transcribed rapidly after stimulation.

Wenfei Sun and his colleagues used full-length deep scRNA-seq to obtain an in-depth view of the transcriptome of engram cells within the BLA. Interestingly, across all conditions the relative abundance of cell types was conserved leading the authors to suggest that long-term fear memory formation is not implicated in cellular architecture remodeling of the BLA. Subclustering the previously identified neuron cluster resulted in the identification of seven neuron subsets, three of which were glutamatergic, while the remaining four were GABAergic. The authors demonstrated that two GABAergic inhibitory neurons (BLA.Int.Gpr88 and BLA.Int.Crhbp) are highly implicated in memory consolidation. Interestingly, Tac1 was downregulated over 6-fold and Penk was upregulated over 4-fold in the BLA.Int.Gpr88  neurons of the FR versus NF conditions. Moreover, genes involved in mitogen-activated kinase pathways and brain-derived neurotrophic factor (BDNF) signaling, both of which have been shown to be implicated in fear-memory, were enriched in the FR condition.

Long-term memory and the neuro-astrocyte spatial niche

Astrocytes, a specialized glial cell, interact with neurons and respond to neuronal activity. Wenfei Sun and his colleagues wanted to determine if, like neurons, astrocytes also exhibit changes in gene expression upon memory consolidation. They found that astrocytes were, in fact, the only non-neuronal cells exhibiting transcriptional changes. Subclustering the astrocyte cluster revealed five astrocyte subsets, of which Astro_1, Astro_4 and Astro_5 showed signs of activation. Interestingly, Astro_1 is more active in the NF condition, while Astro_5 is more active in the FR condition. As such, memory consolidation may induce and shift astrocyte activation. In accordance with scRNA-seq results, a subcluster of activated astrocytes was identified in the spatial transcriptomic dataset belonging to the FR condition.

By spatially resolving the BLA microenvironment surrounding peri-engram neurons – neurons surrounded by a specialized structure called perineuronal nets – the authors identified proximal peri-engram astrocytes that exhibited an enriched expression of Igfbp2. IGFBP2 excretion by astrocytes has been shown to play a role in synaptic transmission and neuron excitability. Moreover, Wenfei Sun and his colleagues demonstrated that upon inhibiting astrocyte activation in the BLA, fear conditioning was impaired which further supports the functional role of astrocytes in memory formation.

Acting together: the prefrontal cortex (PFC) and BLA

An integrated DEG analysis of the BLA scRNA-seq dataset generated in this study and a PFC scRNA-seq dataset generated in a previous study showed that 95% of differentially expressed genes in neurons were co-regulated in the same way. As such, the authors provide evidence that engram neurons across multiple brain regions are associated with conserved transcriptional changes upon memory consolidation. One of the upregulated genes in both regions was Plk2, a transcriptional target of Npas4, suggesting that engram neurons across multiple brain regions may even share transcriptional mechanisms during memory consolidation.

Why this work is important

The scRNA-seq data presented in this preprint reveals the cellular complexity of the BLA and PFC regions, as well as the changes in transcriptional programs across conditions. The use of MERSCOPE, a single-cell spatial transcriptomic technology, has made it possible for Wenfei Sun and his team to reveal neuron-astrocyte interactions in situ during memory consolidation. Neurons communicate to consolidate and recall memories; this preprint gives us (1) valuable insight into the transcriptional changes involved in neuronal communication in long-term memory and (2) highlights that the neuron-astrocyte colocalization within the brain plays an equally important role in long-term memory.

Questions for the authors

1. A study by Nguyen, Phi T., et al. from the University of California, San Francisco demonstrated that microglia interact with neurons to promote synapse plasticity in the hippocampus³. Have you checked for potential neuron-microglia colocalization in the BLA? Do you believe that microglia could also play an important role in the BLA in long-term memory?

2. Have you considered using CellChat or CellPhoneDB to infer inter-cellular communication within your scRNA-seq dataset?

Sources

[1] Camina, Eduardo, and Francisco Güell. “The neuroanatomical, neurophysiological and psychological basis of memory: Current models and their origins.” Frontiers in pharmacology 8 (2017): 438.

[2] Azevedo, Frederico AC, et al. “Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled‐up primate brain.” Journal of Comparative Neurology 513.5 (2009): 532-541.

[3] Nguyen, Phi T., et al. “Microglial remodeling of the extracellular matrix promotes synapse plasticity.” Cell 182.2 (2020): 388-403.

 

doi: https://doi.org/10.1242/prelights.34504

Read preprint (No Ratings Yet)