Modeling the Spread and Effect of β-Amyloid on a Circuit of Memory Neurons

Modeling the Spread and Effect of β-Amyloid on a Circuit of Memory Neurons Results Introduction Discussion/ Conclusions Alzheimer’s disease(AD) is an irreversible, progressive neurodegenerative disease in which patients experience impaired memory, cognition, and inability to carry out daily tasks. Importance of our research: Our model can be modified easily and be used in future research to gain a deeper understanding of plaques spread and memory degradation as Alzheimer’s progresses. The Amyloid Hypothesis states that the accumulation of β-Amyloid plaques, which result from an abnormal production of the amyloid precursor protein, is the primary influence driving the progression of AD. Neurons affected by β-Amyloid plaques experience decreased ion voltage and channel activities, eventually degenerating and losing the ability to depolarize and connect with other neurons. The spread of β-Amyloid in the brain via plasma exosomes clustering around the plaques is one of the major causes of memory loss and symptoms of dementia due to increased level of synaptic dysfunction. https://news.berkeley.edu/2012/01/23/engaged-brain-a myloid-alzheimers/ Above is a graph of model accuracy during the response task at selected days. Note the stable function of the network until the rapid decrease near the end. Less than 100% accuracy near the beginning is due to model’s misperception of the stimulus (10% probability) Below is a heatmap of connection weights at selected days. Yellow pixels indicate a stronger connection weight, while purple regions represent regions with weights near zero. Note how bright yellow regions exist even in later stages, possibly due to these synapses attempting to compensate for synapstic loss in other neurons. Day 0 Day 2200 Day 2600 Day 3000 We heavily modified a pre-existing working memory model (Duggins et. al 2017) to accommodate for the spread of β-Amyloid and find the extent to which it affects the brain and observe the effect of gradual increase in synapse loss over time. The model of working memory used performs a two choice paradigm of the Delayed Response Task. The project aims to model the spread of β-Amyloid in Alzheimer’s patients to better understand the effects plaques have on a circuit of memory cells, as well as the progression of the disease. Our project also investigates relatively unexplored topics such as the specific effect of amyloid plaques on storing and recalling memory, which potentially can be used to create therapeutics for the disease. Potential Applications Shows that Deep Brain Stimulation may be a viable option to improve memory performance (though perhaps as a more palliative care instead), corroborating the findings of Xia et al. It also suggests that memory improvements might be due to stimulation of network-level circuits, rather than cellular level processes. More work needs to be done in refining the model’s many parameters to achieve biological realism. The two choice DRT paradigm presented is not particularly representative of higher level cognition and reasoning, so more work may be done to improve on this aspect. Future research in the field might focus on refining the synaptic degradation algorithm to allow for more parameters to be considered (rather than just probability over time) Day 3400 Day 3600 Day 4000 Day 4200 Methods ● We used a spiking neural network and the Nengo framework to simulate the memory network’s performance on a delayed task response lasting 15 seconds. The model represents a circuit of 2000 neurons capable of processing an input (250 neurons), using working memory to temporarily store input information (1500 neurons), and returning an output based on the original input after a given time period (250 neurons). ● We then measured the accuracy of the memory circuit at various time by checking if the output produced by the model was consistent with the input. We also monitored the “cue” encoded within the memory by summing the spiking activity of the neurons within the memory circuit at each time step. ● We modified the strength of connection weights between neurons within the network in order to simulate the spread of β-Amyloid. The synapses were sorted into seven levels of degradation ranging from healthy to non-functional — where each level represents a different synaptic transmission level. A neuron is considered non-functional when it reaches a synaptic transmission level of 1/e (~37%). ● At specific time intervals, we run the model to quantify and interpret effects of amyloid spread on working memory. The spread of affected synapses within the network itself was modeled by two Poisson distributions centered about a “ground zero” synapse. One distribution represented the infected neurons, and another wider distribution represented the affected connection. ● We adapted an exponential decay function from Kashyap et al. 2019 to model the synaptic transmission of an infected synapse over time. ● We implemented a function to simulate Deep Brain Stimulation on the diseased network. A 90 microsecond pulse was delivered 130 times as a second to simulate the rectangular function used by Xia et al. 2020 to improve function in mice with Alzheimer’s in mice.