Circuit-level mechanisms of memory consolidation

Understanding memory is a central goal of neuroscience, with potentially far-reaching consequences for treating Alzheimer's disease and other dementias. Memory formation requires 'offline consolidation', whereby the neuronal traces representing newly-acquired experiences are selectively stabilised during sleep. Following their retrieval, consolidated memories undergo an additional process of reconsolidation that further stabilizes them for long-term expression.

Damage to particular brain regions results in selective behavioural impairments. The hippocampus, for example, encodes new memories about specific events and places (episodic and spatial memories). We lack a comprehensive understanding of how new memories gain long-term expression but two distinctive patterns of hippocampal electrical activity could promote consolidation processes: sharp-wave ripples (SWRs) and 'dentate spikes'. SWRs have received much scientific attention, and disrupting SWRs impairs memory. In contrast, little is known about dentate spikes, a misleading term that refers not to the action potentials of individual dentate granule cells (DGCs) but to a large population event that recruits many DGCs. No one has ever silenced dentate spikes to determine their role in memory. Our proposed experiments will address this important gap in our knowledge, using cutting-edge technology to detect and silence dentate spikes in the mouse brain in post-learning sleep, thereby revealing their contribution to memory consolidation.

Our approach will utilise sophisticated genetic approaches that can deliver light-sensitive proteins into particular neurons (e.g. DGCs). When stimulated by light, these proteins will silence or activate those neurons. We will use real-time detection of dentate spikes to trigger light-stimulation, giving us precise control over neuronal activity during sleep.

We will assess the contribution of dentate spikes to two distinct forms of memory, both of which require the hippocampus. First we will investigate the effects of dentate spike silencing on associative memory, e.g. learning that cue A predicts outcome X. Ordinarily, learning simple associations does not require the hippocampus, but if the relationship is made ambiguous (e.g. such that outcome X only follows cue A on a subset of trials), the hippocampus then becomes necessary. Second, we will investigate the effects of silencing dentate spikes on non-associative memories, e.g. using the relative novelty or familiarity of mnemonic cues to guide behavioural choices. Consolidation is thought to be critical for associative but not non-associative memories. If silencing dentate spikes also affects non-associative memory, this would suggest a more general role in memory, rather than in consolidation per se. Moreover, in control conditions we will silence DGCs during sleep but NOT during dentate spikes to see whether this also affects memory consolidation.

Next, we will determine the neuronal inputs driving dentate spikes. We will use a recently developed technique to selectively target neurons in the neocortex that project directly to DGCs and determine how activating or silencing those neocortical cells alters dentate spikes.

Finally, we will test how inhibiting an already consolidated memory during 'reconsolidation' affects its long-term expression. We will use a special genetically-modified mouse line that can drive the expression of a light-sensitive neuronal inhibitor selectively in cells that were active during a particular learning episode. At a later time we will reactivate this memory, which places the memory in a labile state, and then determine how inhibiting dentate spikes following this reactivation affects the reconsolidation of this memory.

Collectively, our experiments will make a major contribution to a comprehensive understanding of the circuit-level mechanisms underlying the long-lasting expression of memory.

First, in mice implanted with tetrodes and optic fibres in the dentate gyrus (DG) we will use a closed-loop brain-machine interface to detect dentate spikes and silence dentate granule cells (DGCs), virally transfected with the light-sensitive rapid neuronal silencer ArchT. We will use a transgenic mouse line (expressing Cre recombinase under the metabotropic glutamate receptor-2 promoter) which shows exquisite specificity for DGCs, with no observable expression in CA1-3 or the hilus. We will silence DGCs during sleep after learning events to determine the contribution of dentate spikes to the neuronal and behavioural correlates of memory.

Next we will investigate the neuronal inputs to DGCs that drive dentate spikes. We will use two transgenic mouse lines (Ai32 and Ai40D) that express the light-driven neuronal activator channelrhodopsin-2 (ChR2) and silencer ArchT, respectively, only in Cre-expressing cells. We will inject a retrograde adeno-associated viral vector into the DG to transduce DGC-projecting medial entorhinal cortex (MEC) cells with Cre recombinase. The presence of Cre in these MEC cells will drive ChR2 in Ai32 mice and ArchT in Ai40D mice. We will then test how stimulating DGC-projecting MEC cells during sleep alters the probability or amplitude of dentate spikes.

Finally, we will use 'TetTag' mice in which ArchT expression is driven by the immediate-early gene cFos. ArchT expression is controlled by an experimenter-defined window because it is suppressed by dietary doxycycline (dox). When dox is removed from the diet, neuronal activation induces cFos which then drives ArchT expression in those neurons. Thus we can 'tag' DGCs active during a specific memory episode. We will not interfere with the initial memory consolidation but will later 'reactivate' this memory by presenting retrieval cues which renders the memory labile. We will test how inhibiting the tagged DGCs during dentate spikes following this reactivation affects subsequent memory.

This proposal describes fundamental neuroscience research and therefore its main economic and societal impacts are likely to be increasing knowledge about brain function and effective information processing. As such, we do not anticipate any immediate commercially exploitable outputs or treatments for disease. Nevertheless, the potential consequences of better understanding the mechanisms underlying memory formation could have a significant impact on society at a later time. This proposal will also inform current research efforts in artificial intelligence and deep-learning networks.

Who will benefit?
In addition to the academic community (see Academic Beneficiaries), the main potential beneficiaries of our work will be the pharmaceutical industry, clinicians, and ultimately patient populations from within the general public.

How will they benefit?
Memory impairment is a major aspect of aging, which has considerable impact on the quality of life of individuals. As people live longer, Age-Associated Memory Impairment (AAMI) will become even more of an issue. The hippocampus is a brain region that is intimately associated with memory, and it exhibits important structural, physiological and neurochemical changes with aging. Hippocampus-dependent memories (e.g. episodic and spatial memories) appear to be particularly vulnerable to decline with aging, consistent with the neurobiological changes that occur in the hippocampus as individuals get older. Indeed, age-related cognitive decline has been strongly linked to impairments in hippocampal dependent forms of spatial and episodic memory. In addition, hippocampal dysfunction is a key feature of various other psychiatric and neurological disorders including anxiety, depression, schizophrenia and ischaemic brain injury. Thus, understanding how the hippocampus subserves memory function is likely to be of great importance to both pre-clinical and clinical researchers, the pharmaceutical industry, clinicians and ultimately to the patient population.

Our research is not designed to explicitly identify targets for drug development but, ultimately, a better understanding of the physiological processes that underlie memory formation could increase the likelihood of new treatments for dementia (not only in specific patients groups with conditions like Alzheimer's Disease, but also in the aging population more generally). Indeed, treatments that might restore hippocampal function could also be relevant for a variety of psychiatric disorders (e.g. schizophrenia, anxiety and depression). The development of novel treatment strategies and therapies will produce both economic and societal benefits, with the ultimate endpoint of improving human health.

The knowledge obtained from this proposal will also make an important contribution to the increased and timely interest on brain-machine interfaces, closed-loop systems and artificial intelligence. Indeed, offline consolidation processes have not yet been considered in these research areas while they are likely to greatly optimise information processing.

Finally, via our public-engagement activities, the general public will benefit from an increased knowledge and understanding of the physiological processes that underlie memory consolidation. Specifically, we will organise a public-focused event to discuss the concepts and technologies described in this proposal, and how they might lead to better treatments for age-associated memory impairments in humans (see pathways to impact).