Introduction
Even in total silence, auditory nerve (AN) fibers exhibit "spontaneous" firing. This activity is not merely random noise; it follows specific statistical patterns that serve as a functional fingerprint of the auditory system. Two key metrics describe this: the Inter-Spike-Interval (ISI) distribution, which measures the time between consecutive spikes, and the Fano factor, which quantifies firing correlations over time.
Traditional models, such as the widely used Meddis (2006) model, rely on post-synaptic "relative refractoriness" to explain these patterns. However, these models often fail to replicate the precise ISI shapes and long-term correlations observed in empirical data. This paper investigates whether these phenomena can instead be explained by biophysical processes occurring pre-synaptically within the inner hair cell (IHC).
Methods: Refining the IHC-AN Complex Model
Researchers modified the standard Meddis (2006) IHC-AN complex model by introducing three primary pre-synaptic mechanisms:
- Short-Term Depression (STDv): A mechanism where the probability of vesicle release drops immediately after an event and recovers as the "immediate store" of vesicles is replenished. This is analogous to models of cortical synaptic depression applied to the IHC ribbon synapse.
- Biophysical Calcium Channel Gating: Deterministic calcium equations were replaced with a stochastic four-state model to simulate the random opening and closing of individual calcium channels.
- Phenomenological Channel Noise: As an alternative to discrete gating, an Ornstein-Uhlenbeck (OU) process—a mean-reverting stochastic process—was added to the calcium dynamics to represent continuous channel noise.
The model was validated against high-quality empirical datasets, including cat auditory nerve data from Heil et al. (2007) and mammalian data from Teich and Lowen (1994).
Key Insights and Results
Pre-synaptic vs. Post-synaptic Drivers
The study suggests that what was previously attributed to post-synaptic "refractoriness" (the neuron being "tired") may actually be driven by pre-synaptic "short-term depression" (the synapse running low on ready-to-release vesicles). The STD model produced an ISI probability density function that closely matched the mixture of exponential and gamma distributions observed in biological recordings.
The Role of Noise and Correlation
Surprisingly, simple discrete biophysical gating of calcium channels was insufficient to generate long-term correlations. The "aha moment" occurred when combining the STD model with phenomenological OU noise. This combination was the only one that qualitatively matched both the short-term dip and the long-term rise in Fano factor curves, which indicate strong positive correlations over several seconds.
Biologically Plausible Time Constants
The model achieved its best fit for ISI data when the time constant for short-term depression was set to approximately 2.5 ms. This aligns closely with observed recovery periods in experimental synapse studies, providing further evidence for the pre-synaptic origin of these statistics.
Discussion and Future Work
While the phenomenological OU noise provides an excellent fit, it remains a mathematical abstraction. The authors acknowledge the need for more detailed biophysical models—perhaps where a single calcium channel controls a specific vesicle docking site—to bridge this gap.
Refining these stochastic models is not just a theoretical exercise; it has practical implications for the design of cochlear implants. By better replicating the natural stochasticity of healthy hearing, researchers may eventually develop stimulation patterns that more effectively restore auditory function.
Summary updated