Non-negative Matrix Factorization as a Tool to Distinguish Between Synaptic Vesicles in Different Functional States

Abstract

Synaptic vesicles (SVs) undergo multiple steps of functional maturation (priming) before being fusion competent. We present an analysis technique, which decomposes the time course of quantal release during repetitive stimulation as a sum of contributions of SVs, which existed in distinct functional states prior to stimulation. Such states may represent different degrees of maturation in priming or relate to different molecular composition of the release apparatus. We apply the method to rat calyx of Held synapses. These synapses display a high degree of variability, both with respect to synaptic strength and short-term plasticity during high-frequency stimulus trains. The method successfully describes time courses of quantal release at individual synapses as linear combinations of three components, representing contributions from functionally distinct SV subpools, with variability among synapses largely covered by differences in subpool sizes. Assuming that SVs transit in sequence through at least two priming steps before being released by an action potential (AP) we interpret the components as representing SVs which had been ‘fully primed’, ‘incompletely primed’ or undocked prior to stimulation. Given these assumptions, the analysis reports an initial release probability of 0.43 for SVs that were fully primed prior to stimulation. Release probability of that component was found to increase during high-frequency stimulation, leading to rapid depletion of that subpool. SVs that were incompletely primed at rest rapidly obtain fusion-competence during repetitive stimulation and contribute the majority of release after 3–5 stimuli.