Synaptic strength instability from stochastic gene expression in neurons

Memories can last a lifetime. For many decades, neuroscientists assumed that the stability of our long-term memories was due to stability in the underlying brain circuitry where the information is stored. This picture has changed dramatically over the past 10-15 years, when experimental neuroscientists used cutting edge microscopy and molecular biology methods to track the sizes of synapses - the connections between neurons where information is thought to be stored - longitudinally in the brains of animals over days and weeks. It turned out that individual synapses continually fluctuate in size on a timescale of hours-days, much faster than the years timescale of long-term memories. This contradiction poses a deep challenge for our understanding of memory, and although theoretical neuroscientists have some ideas for how it may be resolved, it basically remains a mystery.

A second source of confusion is the origin of the fluctuations. Beyond a general belief that "biology is noisy", it is not clear what is driving these spontaneous changes in synapse size. An understanding of their sources will be important, for three reasons: 1) it will tell us which cellular properties regulate the fluctuations and which don't; 2) it will give constraints for higher level theoretical models of memory; 3) it may give clues into potential beneficial roles for the fluctuations. In this project we will use mathematical modelling, computer simulations, and data analysis to build and test a new theory of synaptic fluctuations, based on the hypothesis that they arise from gene expression noise.

All cells turn genes on and off to implement cellular functions, and neurons are no exception. When a cell turns a gene 'on', it triggers a biochemical signalling cascade that results in more of its corresponding protein being manufactured. However because this process happens at the level of single molecules diffusing randomly, it is somewhat unreliable, and the amount of protein that gets manufactured varies from moment to moment in a partially uncontrolled way. This 'gene expression noise' is heavily studied in simple cells like bacteria, yeast and some mammalian neurons. Here we will ask, for the first time, if it can plausibly explain synapse size fluctuations in neurons. We will adapt existing mathematical models of stochastic gene expression developed for simple cells, and adapt them for neurons. Neurons are unlike most cells because of their extended tree-shapes, which complicates the mathematical analysis somewhat but can introduce some counter-intuitive effects. First we will analyse these 'simple' mathematical models to get an insight into the key components controlling fluctuation size and timescale. Then we will run detailed computer simulations of more complicated versions of the models with more biological details added. These will produce hard quantitative predictions. Finally, we will test the model's predictions against three previously recorded datasets, provided by our international collaborators.

If successful, this interdisciplinary project will open up new avenues of research on synaptic fluctuations, and give clues for solving the puzzle of how brains can store stable long-term memories despite their unstable components.

Classic neuroscientific theories of long-term memory posit that information is stored in the brain via the strengths of synaptic connections between neurons. However, accumulating evidence has shown that individual synaptic strengths in mammalian brains continually fluctuate in size, typically over a 2 or 3 fold range, on a timescale of hours-days. Since we know from everyday experience that memories can persist for months to years, these findings pose a fundamental challenge to the field's core theories. To solve this paradox, it will be critical to understand the source of these fluctuations, and characterise their properties. Currently this is poorly understood. In this project we will test the new hypothesis that these synaptic strength fluctuations are driven by stochastic protein expression mechanisms in neurons: noisy transcription, translation, degradation, diffusion and trafficking of mRNA and protein.

For preliminary work we adapted existing models of stochastic gene expression in bacteria and similar simple-shaped cells from the systems biology literature to neurons - the key extension being the addition of a spatial tree geometry. When inserting parameters for a key synaptic protein CaMKII, we found that indeed synaptic protein content fluctuations appear with magnitude similar to that reported empirically. This project will develop and test the theory in three phases: 1) use methods from the mathematics of stochastic processes to derive equations for synaptic protein fluctuations in a toy model; 2) use numerical simulations of detailed neuron models to make quantitative predictions; 3) test the predictions against existing experimental data from collaborators. In doing so we aim to build the first theory for stochastic gene expression in neurons, and attempt to link it to synaptic strength fluctuations. We believe this inter-disciplinary project will lead to a step-change in our understanding of synaptic plasticity and long-term memory.