Shane G. Poplawski

Memory acquisition and retrieval impact different epigenetic processes that regulate gene expression.

BackgroundA fundamental question in neuroscience is how memories are stored and retrieved in the brain. Long-term memory formation requires transcription, translation and epigenetic processes that control gene expression. Thus, characterizing genome-wide the transcriptional changes that occur after memory acquisition and retrieval is of broad interest and importance. Genome-wide technologies are commonly used to interrogate transcriptional changes in discovery-based approaches. Their ability to increase scientific insight beyond traditional candidate gene approaches, however, is usually hindered by batch effects and other sources of unwanted variation, which are particularly hard to control in the study of brain and behavior.ResultsWe examined genome-wide gene expression after contextual conditioning in the mouse hippocampus, a brain region essential for learning and memory, at all the time-points in which inhibiting transcription has been shown to impair memory formation. We show that most of the variance in gene expression is not due to conditioning and that by removing unwanted variance through additional normalization we are able provide novel biological insights. In particular, we show that genes downregulated by memory acquisition and retrieval impact different functions: chromatin assembly and RNA processing, respectively. Levels of histone 2A variant H2AB are reduced only following acquisition, a finding we confirmed using quantitative proteomics. On the other hand, splicing factor Rbfox1 and NMDA receptor-dependent microRNA miR-219 are only downregulated after retrieval, accompanied by an increase in protein levels of miR-219 target CAMKIIγ.ConclusionsWe provide a thorough characterization of coding and non-coding gene expression during long-term memory formation. We demonstrate that unwanted variance dominates the signal in transcriptional studies of learning and memory and introduce the removal of unwanted variance through normalization as a necessary step for the analysis of genome-wide transcriptional studies in the context of brain and behavior. We show for the first time that histone variants are downregulated after memory acquisition, and splicing factors and microRNAs after memory retrieval. Our results provide mechanistic insights into the molecular basis of cognition by highlighting the differential involvement of epigenetic mechanisms, such as histone variants and post-transcriptional RNA regulation, after acquisition and retrieval of memory.

Gadd45b knockout mice exhibit selective deficits in hippocampus-dependent long-term memory

Growth arrest and DNA damage-inducible β (Gadd45b) has been shown to be involved in DNA demethylation and may be important for cognitive processes. Gadd45b is abnormally expressed in subjects with autism and psychosis, two disorders associated with cognitive deficits. Furthermore, several high-throughput screens have identified Gadd45b as a candidate plasticity-related gene. However, a direct demonstration of a link between Gadd45b and memory has not been established. The current studies first determined whether expression of the Gadd45 family of genes was affected by contextual fear conditioning. Gadd45b, and to a lesser extent Gadd45g, were up-regulated in the hippocampus following contextual fear conditioning, whereas Gadd45a was not. Next, Gadd45b knockout mice were tested for contextual and cued fear conditioning. Gadd45b knockout mice exhibited a significant deficit in long-term contextual fear conditioning; however, they displayed normal levels of short-term contextual fear conditioning. No differences between Gadd45b knockout and wild-type mice were observed in cued fear conditioning. Because cued fear conditioning is hippocampus independent, while contextual fear conditioning is hippocampus dependent, the current studies suggest that Gadd45b may be important for long-term hippocampus-dependent memory storage. Therefore, Gadd45b may be a novel therapeutic target for the cognitive deficits associated with many neurodevelopmental, neurological, and psychiatric disorders.

Sleep deprivation impairs memory by attenuating mTORC1-dependent protein synthesis

Sleep deprivation impairs memory formation by suppressing protein synthesis. How sleep deprivation impairs memory Sleep deprivation impairs learning and memory. Tudor et al. (see also the Focus by Sweatt and Hawkins) found that sleep deprivation in mice suppressed activation of the kinase complex mTORC1 and consequently protein synthesis in hippocampal neurons, which impaired their memory. Restoring protein synthesis by increasing the amount of phosphorylated 4EBP2 protein in the hippocampus—a function normally performed by mTORC1—protected mice from the memory impairment caused by sleep deprivation. The findings reveal a molecular mechanism by which sleep loss impairs memory. Sleep deprivation is a public health epidemic that causes wide-ranging deleterious consequences, including impaired memory and cognition. Protein synthesis in hippocampal neurons promotes memory and cognition. The kinase complex mammalian target of rapamycin complex 1 (mTORC1) stimulates protein synthesis by phosphorylating and inhibiting the eukaryotic translation initiation factor 4E–binding protein 2 (4EBP2). We investigated the involvement of the mTORC1-4EBP2 axis in the molecular mechanisms mediating the cognitive deficits caused by sleep deprivation in mice. Using an in vivo protein translation assay, we found that loss of sleep impaired protein synthesis in the hippocampus. Five hours of sleep loss attenuated both mTORC1-mediated phosphorylation of 4EBP2 and the interaction between eukaryotic initiation factor 4E (eIF4E) and eIF4G in the hippocampi of sleep-deprived mice. Increasing the abundance of 4EBP2 in hippocampal excitatory neurons before sleep deprivation increased the abundance of phosphorylated 4EBP2, restored the amount of eIF4E-eIF4G interaction and hippocampal protein synthesis to that seen in mice that were not sleep-deprived, and prevented the hippocampus-dependent memory deficits associated with sleep loss. These findings collectively demonstrate that 4EBP2-regulated protein synthesis is a critical mediator of the memory deficits caused by sleep deprivation.

Sleep deprivation causes memory deficits by negatively impacting neuronal connectivity in hippocampal area CA1

Brief periods of sleep loss have long-lasting consequences such as impaired memory consolidation. Structural changes in synaptic connectivity have been proposed as a substrate of memory storage. Here, we examine the impact of brief periods of sleep deprivation on dendritic structure. In mice, we find that five hours of sleep deprivation decreases dendritic spine numbers selectively in hippocampal area CA1 and increased activity of the filamentous actin severing protein cofilin. Recovery sleep normalizes these structural alterations. Suppression of cofilin function prevents spine loss, deficits in hippocampal synaptic plasticity, and impairments in long-term memory caused by sleep deprivation. The elevated cofilin activity is caused by cAMP-degrading phosphodiesterase-4A5 (PDE4A5), which hampers cAMP-PKA-LIMK signaling. Attenuating PDE4A5 function prevents changes in cAMP-PKA-LIMK-cofilin signaling and cognitive deficits associated with sleep deprivation. Our work demonstrates the necessity of an intact cAMP-PDE4-PKA-LIMK-cofilin activation-signaling pathway for sleep deprivation-induced memory disruption and reduction in hippocampal spine density. DOI: http://dx.doi.org/10.7554/eLife.13424.001

How data analysis affects power, reproducibility and biological insight of RNA-seq studies in complex datasets

The sequencing of the full transcriptome (RNA-seq) has become the preferred choice for the measurement of genome-wide gene expression. Despite its widespread use, challenges remain in RNA-seq data analysis. One often-overlooked aspect is normalization. Despite the fact that a variety of factors or ‘batch effects’ can contribute unwanted variation to the data, commonly used RNA-seq normalization methods only correct for sequencing depth. The study of gene expression is particularly problematic when it is influenced simultaneously by a variety of biological factors in addition to the one of interest. Using examples from experimental neuroscience, we show that batch effects can dominate the signal of interest; and that the choice of normalization method affects the power and reproducibility of the results. While commonly used global normalization methods are not able to adequately normalize the data, more recently developed RNA-seq normalization can. We focus on one particular method, RUVSeq and show that it is able to increase power and biological insight of the results. Finally, we provide a tutorial outlining the implementation of RUVSeq normalization that is applicable to a broad range of studies as well as meta-analysis of publicly available data.

Object-location training elicits an overlapping but temporally distinct transcriptional profile from contextual fear conditioning.

Hippocampus-dependent learning is known to induce changes in gene expression, but information on gene expression differences between different learning paradigms that require the hippocampus is limited. The bulk of studies investigating RNA expression after learning use the contextual fear conditioning task, which couples a novel environment with a footshock. Although contextual fear conditioning has been useful in discovering gene targets, gene expression after spatial memory tasks has received less attention. In this study, we used the object-location memory task and studied gene expression at two time points after learning in a high-throughput manner using a microfluidic qPCR approach. We found that expression of the classic immediate-early genes changes after object-location training in a fashion similar to that observed after contextual fear conditioning. However, the temporal dynamics of gene expression are different between the two tasks, with object-location memory producing gene expression changes that last at least 2 hours. Our findings indicate that different training paradigms may give rise to distinct temporal dynamics of gene expression after learning.

Learning induces the translin/trax RNase complex to express activin receptors for persistent memory

Long-lasting forms of synaptic plasticity and memory require de novo protein synthesis. Yet, how learning triggers this process to form memory is unclear. Translin/trax is a candidate to drive this learning-induced memory mechanism by suppressing microRNA-mediated translational silencing at activated synapses. We find that mice lacking translin/trax display defects in synaptic tagging, which requires protein synthesis at activated synapses, and long-term memory. Hippocampal samples harvested from these mice following learning show increases in several disease-related microRNAs targeting the activin A receptor type 1C (ACVR1C), a component of the transforming growth factor-β receptor superfamily. Furthermore, the absence of translin/trax abolishes synaptic upregulation of ACVR1C protein after learning. Finally, synaptic tagging and long-term memory deficits in mice lacking translin/trax are mimicked by ACVR1C inhibition. Thus, we define a new memory mechanism by which learning reverses microRNA-mediated silencing of the novel plasticity protein ACVR1C via translin/trax.

Combination therapy prevents amyloid-dependent and -independent structural changes

Neuropathological features of Alzheimers disease (AD) are recapitulated in transgenic mice expressing familial AD-causing mutations, but ectopic transgene overexpression makes it difficult to relate these abnormalities to disease pathogenesis. Alternatively, the APP/PS-1 double knock-in (DKI) mouse produces mutant amyloid precursor protein (APP) and presenilin-1 (PS-1) with normal levels and regulatory controls. Here, we investigated effects of amyloid on brain structure and neuroplasticity by vaccinating DKI mice with amyloid-β starting at 8 months of age. At 14 months, vaccination blocked cerebral amyloid deposition and its attendant microglial activation. Neuropil abnormalities were pronounced only within plaques, and included circumscribed loss and dysmorphology of axons, dendrites, terminals and spines. Blockade of amyloid deposition restored neuropil integrity. Amyloid removal did not rescue reductions in the hippocampal neural progenitor and neuroblast populations, but adding 1 month of voluntary exercise to amyloid-β vaccination markedly stimulated hippocampal neurogenesis. These results identify amyloid-dependent and -independent structural changes in the DKI mouse model of AD. Combining exercise with amyloid-directed immunotherapy produces greater restoration of brain structure and neuroplasticity than is achieved with either maneuver alone.

Contextual fear conditioning induces differential alternative splicing.

The process of memory consolidation requires transcription and translation to form long-term memories. Significant effort has been dedicated to understanding changes in hippocampal gene expression after contextual fear conditioning. However, alternative splicing by differential transcript regulation during this time period has received less attention. Here, we use RNA-seq to determine exon-level changes in expression after contextual fear conditioning and retrieval. Our work reveals that a short variant of Homer1, Ania-3, is regulated by contextual fear conditioning. The ribosome biogenesis regulator Las1l, small nucleolar RNA Snord14e, and the RNA-binding protein Rbm3 also change specific transcript usage after fear conditioning. The changes in Ania-3 and Las1l are specific to either the new context or the context-shock association, while the changes in Rbm3 occur after context or shock only. Our analysis revealed novel transcript regulation of previously undetected changes after learning, revealing the importance of high throughput sequencing approaches in the study of gene expression changes after learning.

Learning-dependent chromatin remodeling highlights noncoding regulatory regions linked to autism

High-throughput sequencing analysis of learning-induced epigenetic changes in the mouse hippocampus reveals regulatory regions relevant to autism. Deciphering the chromatin in learning and autism Both development and learning are shaped through epigenetics—modifications to the DNA or chromatin that alter gene expression without changing the underlying DNA sequence. The intellectual disorder ASD (autism spectrum disorder) is idiopathic, but it is associated with changes in gene expression that alter the development of neuronal circuitry in the brain and impair some forms of learning. Using a new technique called DEScan, Koberstein et al. explored learning-induced changes to the epigenetic landscape of the hippocampus (a region critical for memory) in mice. They found that learning was mediated through changes to regulatory regions, particularly the activation of alternative promoters, of many genes associated with ASD. These findings identify an epigenetic source of one gene alteration associated with ASD and, more broadly, demonstrate a new technique for exploring learning disorders in animal models. Autism spectrum disorder (ASD) is a prevalent neurodevelopmental disorder that is associated with genetic risk factors. Most human disease-associated single-nucleotide polymorphisms (SNPs) are not located in genes but rather are in regulatory regions that control gene expression. The function of regulatory regions is determined through epigenetic mechanisms. Parallels between the cellular basis of development and the formation of long-term memory have long been recognized, particularly the role of epigenetic mechanisms in both processes. We analyzed how learning alters chromatin accessibility in the mouse hippocampus using a new high-throughput sequencing bioinformatics strategy we call DEScan (differential enrichment scan). DEScan, which enabled the analysis of data from epigenomic experiments containing multiple replicates, revealed changes in chromatin accessibility at 2365 regulatory regions—most of which were promoters. Learning-regulated promoters were active during forebrain development in mice and were enriched in epigenetic modifications indicative of bivalent promoters. These promoters were disproportionally intronic, showed a complex relationship with gene expression and alternative splicing during memory consolidation and retrieval, and were enriched in the data set relative to known ASD risk genes. Genotyping in a clinical cohort within one of these promoters (SHANK3 promoter 6) revealed that the SNP rs6010065 was associated with ASD. Our data support the idea that learning recapitulates development at the epigenetic level and demonstrate that behaviorally induced epigenetic changes in mice can highlight regulatory regions relevant to brain disorders in patients.