Neural function relies on cellular energy supplies meeting the episodic demands of synaptic activity, but little is known about the extent to which power demands (energy demands per unit time) fluctuate, or the mechanisms that match supply with demand. Here, in individually-identified glutamatergic motor neuron terminals of Drosophila larvae, we leveraged prior macroscopic estimates of power demand to generate profiles of power demand from one action potential to the next. These profiles show that signaling demands can exceed non-signaling demands by 17-fold within milliseconds, and terminals with the greatest fluctuation (volatility) in power demand have the greatest mitochondrial volume and packing density. We elaborated on this quantitative approach to simulate adenosine triphosphate (ATP) levels during activity and drove ATP production as a function of the reciprocal of the energy state, but this canonical feedback mechanism appeared to be unable to prevent ATP depletion during locomotion. Muscle cells possess a phosphagen system to buffer ATP levels but phosphagen systems have not been described for motor nerve terminals. We examined these terminals for evidence of a phosphagen system and found the mitochondria to be heavily decorated with an arginine kinase, the key element of invertebrate phosphagen systems. Similarly, an examination of mouse cholinergic motor nerve terminals found mitochondrial creatine kinases, the vertebrate analogues of arginine kinases. Knock down of arginine kinase in Drosophila resulted in rapid depletion of presynaptic ATP during activity, indicating that, in motor nerve terminals, as in muscle, phosphagen systems play a critical role in matching power supply with demand.

Calcium ions are key signaling molecules in dendritic spines, the small neurotransmitter-receiving protrusions along dendrites. Their dynamics have been shown to regulate many downstream phenomena including synaptic plasticity and learning. Previously we and others have shown that the subcellular morphology of spines can affect signaling dynamics of Ca 2þ and cAMP. In this work we construct a reaction-diffusion Partial Differential Equation (PDE) model of the dynamics of calcium in response to varied electrical stimuli. On realistic mesh geometries of dendritic spines generated from 3D electron micrographs via our open-source workflow including softwares IMOD, GAMer 2, and Blender, the PDEs are solved using FEniCS, an open-source finite element solver. The resulting simulations across many dendritic spines are analyzed for correlations between calcium signals and spine geometries. We posit that these robust ultrastructure-signaling relationships represent possible mechanisms of how dendritic spines can learn and process information.