Memory loss reversed in mouse models of early Alzheimer’s Disease

According to the Alzheimer’s Association (, one in every nine Americans over 65 years old is living with Alzheimer’s Disease (AD), the devastating neurodegenerative disorder characterized by progressive memory loss, broad cognitive dysfunction and dementia.  As the “Baby Boomer” generation ages, the number of afflicted individuals is expected to rise nearly 3-fold by 2050.

Episodic memory loss is one of the earliest neurological manifestations observed in AD patients, but studies have not revealed definitively whether the pathway affected is in memory encoding and storage or memory retrieval. In a March 2016 study published in Nature (Roy et al. 2016), Dr. Susumu Tonegawa’s group at the Massachusetts Institute of Technology (MIT) used a sophisticated opotogenetic approach to directly stimulate memory “engram” neurons in the hippocampus of early AD mouse models.  Their treatment restored learned behaviors in the AD mice that could no longer be induced by contextual cues.  Their data reveal that the cognitive decline observed in early AD is due to defects in memory retrieval, and have implications which could impact clinical therapies for improving memory recall in patients with AD.

Optogenetic stimulation of neurons in the hippocampus restores long-term memory in an AD mouse model

Nine-month-old AD mice (B6C3-Tg(APPswe,PSEN1dE9)85Dbo/Mmjax (004462/034829-JAX)), which overexpress both a chimeric mouse/human b-amyloid precursor protein with the Swedish mutant (APPswe) and the delta exon 9 variant of human presenillin 1 (PSEN1dE9), show severe AD-associated b-amyloid plaque deposits, whereas 7 month-old mice do not.  Therefore, the Tonegawa group chose the latter as their model for early AD.  Because the hippocampus has a crucial role in episodic memory storage, the Tonegawa team focused their study on neurons in this region, specifically in the dentate gyrus (DG), and neurons that innervate them in the entorhinal cortex.  As their behavioral paradigm to study short- and long-term memory, they used conditional fear conditioning (CFC).  In this model, mice are given an aversive stimulus (such as a foot shock) in the context of a specific environment.  After training the mice typically “freeze” when placed in the same environment even in the absence of the stimulus.  To test short-term and long-term memory integration, the mice are placed in the aversive context 1 hour and 24 hours after training, respectively.  After CFC training, the 9 month-old mice showed memory deficits (reduced freezing incidence) in both short- and long-term memory, whereas the 7 month-old mice only showed defects in long-term memory.  Granule cell numbers in the DG were the same in the 7 month-old AD mice as in wild-type controls, but fewer neurons were activated during long-term memory trials as revealed by c-Fos+ immunohistochemical staining (a marker for neuron activation).

To study whether the failure for the 7 month-old AD mice to retain long term memory was due a defect in memory storage or in memory recall, the Tonegawa team injected mice stereotaxically with adeno-associated viral (AAV) viral vectors that permitted simultaneous doxycycline-regulated expression of the optogenetic protein channelrhodopsin (ChR2) and yellow fluorescent protein (YFP) labelling of neurons activated during CFC training. The mice were also fitted with an optic fibre headpiece to allow laser-stimulated activation of the ChR2 protein. Delivery of blue light to ChR2-expressing neurons induces them to fire and activate their downstream targets.  ChR2-YFP expression had no affect the animals’ long-term memory after CFC training – the mice still failed to freeze 24 hours.  When the CFC-trained, ChR2-YFP-expressing mice we placed in a control environment that normal does not induce freezing, activation of the ChR2-expressing neurons with blue light induced the animals to freeze with a similar incidence as in CFC-trained, ChR2-YFP-expressing wild-type controls.  These data demonstrated that re-activation of neurons that were activated during the CFC training period could restore the contextual memory in the 7 month-old AD mice, indicating that the long-term memory deficit was in memory recall.   The corrected memory recall was not permanent, however:  when the mice were retested 24 hours after the light-induced memory correction, they showed the same memory defect as before.  Using similar optogentic techniques, the MIT team corrected long-term memory defect in to two other AD mouse models with long-term memory defects, including 3xTg-AD mice (B6;129-Psen1tm1Mpm Tg(APPSwe,tauP301L)1Lfa/Mmjax (004807/034830-JAX)).

Increasing dendritic spine density improves long-term memory in AD mice

The density of dendritic spines – the sites on neurons where the make contact with their upstream partners – has been implicated in AD memory deficits, and previous studies by the Tonegawa group had found that long-term memory defects in early AD mice correlate with decreases in dendritic spine density on DG memory engram neurons.  These observations suggested that increasing dendritic spine numbers might improve the memory deficits in the early AD mice.  To test this, the MIT researchers used their optogenetic system to induce long-term potentiation (LTP) - a phenomenon in which synaptic connections are strengthened by repeated stimulation by upstream neurons - in memory engram neurons in vivo by using light to activate channelrhodopsin-expressing neurons in the entorhinal cortex (EC) that innervate them. The investigators had previously demonstrated that LTP rapidly induces spine formation in downstream neuronal targets.  Following CFC training, optogenetically-induced LTP increased dendritic spine numbers in the DG memory engram neurons, and restored long-term memory, which persisted for at least six days after training.  Specifically ablating the DG memory engram cells with a diphtheria toxin-based protocol following training and LTP-induced memory retrieval reversed the effect, demonstrating that increases in the dendritic spines on the memory engram cells and not on other neurons, was responsible for the improved memory recall. The optogenetic-LTP protocol also improved memory retrieval in other two other cognitive, behavioral tests:  inhibitory avoidance tests – a test in which animals are trained to avoid a specific area where the previously received an aversive stimulus- and in novel object recognition tests, which evaluates recognition memory. 

Together, the data published by the MIT group suggest that episodic memory degradation associated with early AD is a problem in memory retrieval.  Their data further suggests that therapies aimed at increasing dendritic spine density in neurons encoding these memories or strategies to stimulate these neurons directly may help to delay or reverse cognitive deficits as AD progresses.  Interestingly, applying the optogenetic-LTP protocol to a broad field of EC neurons – not just those that project to the DG memory engram cells – did not rescue long-term memory in the AD mice.  These data suggest that direct targeting of neuronal subsets that encode specific memories may be required for their effective retrieval.


Roy DS et al. 2016.  Memory retrieval by activating engram cells in mouse models of early Alzheimer’s disease.  Nature. 531: 508-512,  PMID: 26982728