Low-noise encoding of active touch by layer 4 in the somatosensory cortex

Samuel Andrew Hires, Diego A. Gutnisky, Jianing Yu, Daniel H. O’Connor, Karel SvobodaeLife 2015 · DOI

TL;DR

This study reveals that layer 4 neurons in the mouse somatosensory cortex encode tactile information with remarkable temporal precision and surprisingly low noise, challenging previous assumptions about cortical "noisiness." This suggests that the brain might leverage precise spike timing for efficient sensory processing, rather than relying solely on population-averaged firing rates.

What Problem Does This Paper Address?

A long-standing debate in neuroscience concerns the "noisiness" of cortical spike trains. While spike trains often appear irregular and variable across repeated stimuli, it's unclear whether this variability is an inherent, irreducible property of neural circuits (internal noise) or due to uncontrolled factors like behavioral state or sensory input fluctuations (external noise). Estimating true internal noise is challenging because it requires precise control and tracking of behavior and brain state. This paper investigates this question in the context of active touch, where mice use their whiskers to explore objects, a scenario that allows for high-precision behavioral control and sensory input measurement.

Methods

The researchers recorded from excitatory neurons in layer 4 (L4) of the mouse barrel cortex while mice performed an active tactile object localization task. Key methodological aspects include:

  • Active Tactile Task: Mice learned to discriminate object locations using a single whisker. This active engagement allowed for precise control over sensory input.
  • High-Precision Behavioral Tracking: Whisker movements (position, phase, curvature) and touch events were tracked with millisecond temporal resolution.
  • Loose-Seal Juxtacellular Recordings: This technique was used to record from individual neurons, minimizing disruption and allowing for accurate spike detection.
  • Fano Factor (FF) Analysis: This standard measure was used to quantify spike count variability. The authors employed sophisticated analysis to calculate FF under various conditions, including aligning spikes to specific behavioral events and sorting touches by their characteristics.
  • Decoding Analysis: Simulated populations of L4 neurons were used to assess their ability to decode touch timing and whisking phase.

Datasets

The study utilized recordings from 21 mice, with 52 loose-seal cell-attached recordings from L4 neurons. These recordings were collected during the described active tactile task. No new benchmark datasets were introduced; rather, the study generated novel data from a controlled experimental paradigm.

Key Insights and Ideas

  • Touch Dominates L4 Activity: Over 70% of spikes in L4 neurons occurred within millisecond-timescale epochs immediately following touch onset. This highlights the critical role of touch as a primary driver of L4 activity.
  • Temporal Precision is Key: The timing of spikes, not just their rate, appears to carry significant information. Spikes were precisely coupled to touch onset, often occurring within 8 ms.
  • Low Spike Count Variability: The variance of touch responses was often close to the theoretical minimum (binomial limit), suggesting minimal irreducible noise. This is a significant departure from previous observations of higher variability in other cortical areas or under passive stimulation.
  • Active Sensing Reduces Noise: The active nature of the task, where mice control their sensory input, seems to be a crucial factor in achieving this low-noise encoding, potentially by recruiting circuit mechanisms that enhance precision.
  • Touch Decoding is Highly Efficient: A small population of L4 neurons (as few as 95) could reliably detect touch events with high temporal precision (10 ms integration time).
  • Whisking Phase Encoding is Coarse: While L4 neurons are modulated by whisking phase, the encoding is less precise and requires a much larger population for decoding, suggesting it plays a secondary role in this specific task.

Key Results

  • Spike Timing: Over 70% of spikes were temporally coupled to touch onset, with short latencies (onset ~8 ms).
  • Fano Factor: For touch responses, the FF was often close to the binomial limit (as low as 0.54 ± 0.27 when accounting for similar touches), indicating minimal spike count variability.
  • Touch Decoding: A population of just 95 L4 neurons could detect touch with 95% accuracy. Decoding elapsed time from touch was also highly precise, with a minimum uncertainty of 0.55 ms.
  • Whisking Phase Decoding: A large population (1000 neurons) was required for even a rough estimate of whisking phase (76% correct discrimination), indicating a coarser representation.

Limitations and Future Work

  • Generalizability: The study focused on L4 of the barrel cortex during a specific tactile task. It remains to be seen if these findings generalize to other cortical areas, sensory modalities, or more complex behaviors.
  • Internal vs. External Noise: While the study strongly suggests low irreducible noise, definitively separating internal from external noise is inherently difficult. The authors controlled for many external factors, but subtle influences might remain.
  • Role of Inhibition: The authors hypothesize that strong, rapid feedforward inhibition might contribute to the observed binary-like responses and low variability. Further investigation into the precise role of inhibitory circuits would be valuable.
  • Brain State: While active behavior was controlled, the precise internal brain state (e.g., attention levels beyond task engagement) was not explicitly measured, which could still influence variability.
  • Future Directions: The authors suggest exploring how the brain interprets and utilizes this precise, low-noise touch information for object identification and decision-making. Understanding the downstream processing of this information in higher cortical areas is a key next step.

Summary updated

Abstract

Cortical spike trains often appear noisy, with the timing and number of spikes varying across repetitions of stimuli. Spiking variability can arise from internal (behavioral state, unreliable neurons, or chaotic dynamics in neural circuits) and external (uncontrolled behavior or sensory stimuli) sources. The amount of irreducible internal noise in spike trains, an important constraint on models of cortical networks, has been difficult to estimate, since behavior and brain state must be precisely controlled or tracked. We recorded from excitatory barrel cortex neurons in layer 4 during active behavior, where mice control tactile input through learned whisker movements. Touch was the dominant sensorimotor feature, with >70% spikes occurring in millisecond timescale epochs after touch onset. The variance of touch responses was smaller than expected from Poisson processes, often reaching the theoretical minimum. Layer 4 spike trains thus reflect the millisecond-timescale structure of tactile input with little noise.
BibTeX
@inproceedings{hires2015low,
  author = {Samuel Andrew Hires and Diego A. Gutnisky and Jianing Yu and Daniel H. O’Connor and Karel Svoboda},
  title = {{Low-noise encoding of active touch by layer 4 in the somatosensory cortex}},
  booktitle = {eLife},
  year = {2015},
  doi = {10.7554/elife.06619},
}
Figure