PERSISTENCE AND VOLATILITY IN SPINE MORPHODYNAMICS

Synaptic strength is continuously shaped by activity-dependent plasticity as well as ongoing structural turnover of synaptic components. While long-term memory is thought to rely on stable synaptic changes, in vivo imaging has revealed substantial volatility of dendritic spines, even for apparently persistent synapses. How long-term memory can coexist with such fluctuations remains poorly understood.
Here, we analyze dendritic spine morphodynamics across multiple timescales using in vivo STED nanoscopy data, quantifying temporal fluctuations of spine head size, neck length, and neck width as well as their mutual statistical covariation over intervals ranging from hours to weeks. Cross-covariances reveal robust interrelations, indicating that spine feature fluctuations are not driven by independent noise but by partially coordinated processes.
To explain these observations, we develop an event-based stochastic model in which spine morphology evolves through a mixture of coordinated and feature-specific events acting on top of a quenched baseline. Model comparison shows that neither purely activity-independent stochastic dynamics nor a single common STDP-like driver can account for the measured covariances. Instead, spine morphodynamics require the coexistence of coordinated events affecting multiple features and independent events targeting individual features. Finally, we introduce a data-driven generative model that reconstructs realistic spine trajectories from sparse measurements while preserving both volatility and persistence.
Together, our results provide a mechanistic framework for how synaptic structures can remain stable over long timescales while retaining the flexibility required for ongoing plasticity.