Disentangling Sub-Millisecond Processes within an Auditory Transduction Chain

Tim Gollisch, Andreas M. V HerzPLoS 2004 · DOI

Abstract

Every sensation begins with the conversion of a sensory stimulus into the response of a receptor neuron. Typically, this involves a sequence of multiple biophysical processes that cannot all be monitored directly. In this work, we present an approach based on analyzing "iso-response" stimuli—different inputs that produce the same final output probability. Applied to insect auditory receptor cells, this technique reveals sub-millisecond dynamics of eardrum vibration and electrical potential, yielding a quantitative four-step cascade model that explains high temporal resolution under natural stimulation. Owing to its simplicity and generality, the presented method is readily applicable to other nonlinear cascades and a large variety of signal-processing systems.

Introduction

Sensory transduction is a lightning-fast relay: air pressure vibrates a membrane, opening ion channels, changing cell voltage, and finally triggering a neural spike. Studying this chain is difficult because researchers typically only observe the final spike. Intermediate steps are often too fast to measure directly, and the inherent timing jitter of neural firing (approximately 1 ms) masks sub-millisecond upstream dynamics. Furthermore, the fragile nature of the auditory periphery makes direct mechanical measurement invasive. This study introduces a non-invasive method to "peek inside" this black box using only output spikes.

The Iso-Response Method and LNLN Model

The researchers developed a method based on iso-response stimuli. Instead of measuring how output changes with input, they kept the output constant (e.g., a 70% firing probability) and varied the input parameters.

The four-step auditory transduction chain model: from sound wave stimulus to neural response.

The system is modeled as a four-step "Linear-Nonlinear-Linear-Nonlinear" (LNLN) cascade:

  1. L1 (Mechanical Filter): Linear eardrum vibration.
  2. N1 (Squaring Nonlinearity): Conversion of vibration energy into current via mechanosensors.
  3. L2 (Electrical Filter): Charge accumulation and ionic current integration on the cell membrane.
  4. N2 (Spike Generation): Nonlinear thresholding resulting in an action potential.

By using a "two-click" stimulation protocol—varying the amplitude and timing interval between two sound pulses—the authors could algebraically separate the time constants of mechanical vibration from electrical integration.

Key Insights: Bypassing the Jitter Barrier

While the timing of a single spike is imprecise, the probability of a spike is a stable metric. By focusing on probability, the researchers achieved microsecond-level resolution, far exceeding the 1 ms spike jitter.

A critical observation was the "geometry of response." For short click intervals, iso-response curves were straight lines (summing pressure). For longer intervals, they became circles (summing energy/squared pressure). This shift confirms a squaring nonlinearity between two filtering steps, likely representing the opening of mechanosensory ion channels where current is proportional to mechanical oscillation energy.

Results and Predictive Power

The method successfully extracted eardrum resonance frequencies (5–15 kHz) and damping constants purely from spike data. It also measured the membrane's integration time constant, a value previously inaccessible without damaging the tiny receptor cells.

The resulting model accurately predicted responses to complex three-click patterns and reconstructed traditional frequency-tuning curves without additional parameter tuning. This demonstrates that the high temporal resolution of these neurons is a direct consequence of the specific filters in the transduction chain.

Discussion and Future Directions

The model assumes a feedforward structure, which is appropriate for the locust auditory system but would require modification for systems with complex feedback, like the mammalian cochlea. While the study focused on sub-millisecond scales, slower processes like spike-frequency adaptation remain for future exploration.

Ultimately, this "iso-response" framework is a general tool. It can be applied to disentangle hidden steps in various signal-processing chains, from gene regulatory networks to other sensory modalities like vision or touch.

Summary updated

Abstract

Every sensation begins with the conversion of a sensory stimulus into the response of a receptor neuron. Typically, this involves a sequence of multiple biophysical processes that cannot all be monitored directly. In this work, we present an approach that is based on analyzing different stimuli that cause the same final output, here defined as the probability of the receptor neuron to fire a single action potential. Comparing such iso-response stimuli within the framework of nonlinear cascade models allows us to extract the characteristics of individual signal-processing steps with a temporal resolution much finer than the trial-to-trial variability of the measured output spike times. Applied to insect auditory receptor cells, the technique reveals the sub-millisecond dynamics of the eardrum vibration and of the electrical potential and yields a quantitative four-step cascade model. The model accounts for the tuning properties of this class of neurons and explains their high temporal resolution under natural stimulation. Owing to its simplicity and generality, the presented method is readily applicable to other nonlinear cascades and a large variety of signal-processing systems.
BibTeX
@inproceedings{gollisch2004disentangling,
  author = {Tim Gollisch and Andreas M. V Herz},
  title = {{Disentangling Sub-Millisecond Processes within an Auditory Transduction Chain}},
  booktitle = {PLoS},
  year = {2004},
  doi = {10.1371/journal.pbio.0030008},
}
Figure