Thalamus exhibits less sensory variability quenching than cortex

Eva Poland, Tobias H. Donner, Kai-Markus Müller, David A. Leopold, Melanie WilkeScientific 2019 · DOI

TL;DR

Cortical neurons show a significant decrease in trial-to-trial variability when a stimulus appears, a phenomenon known as "variability quenching." This study reveals that the thalamus, particularly the pulvinar, exhibits much lower baseline variability than the cortex and largely fails to show this stimulus-induced quenching effect. This suggests fundamental differences in how these brain regions process sensory information.

What Problem Does This Paper Address?

Neural activity, even when responding to the same stimulus repeatedly, is inherently noisy and variable. While this variability is well-studied in the cortex, where it's known to decrease significantly upon stimulus onset (variability quenching), its presence and dynamics in the thalamus, a crucial relay and processing center for sensory information, remain largely unknown. Specifically, the authors investigate whether higher-order thalamic nuclei like the pulvinar, which have extensive reciprocal connections with the cortex, also exhibit this stimulus-driven reduction in variability. Understanding this could shed light on how sensory information is transformed and processed as it ascends from the sensory periphery to higher cortical areas.

Methods

The researchers recorded electrophysiological data (multi-unit and single-unit activity) from two adult rhesus macaques performing a visual detection task. They simultaneously recorded from several brain regions: the visual cortex (area V4), the lateral geniculate nucleus (LGN - a first-order thalamic nucleus), and two parts of the pulvinar (ventral and dorsal - higher-order thalamic nuclei).

The core of their analysis focused on trial-to-trial variability, quantified using the Fano factor (the ratio of spike count variance to mean spike count). They compared Fano factors in two periods: a stable fixation period before stimulus onset and a period immediately following stimulus onset. They also performed control analyses to rule out confounding factors like differences in response strength, mean firing rates, and even motor responses associated with the task. To further investigate the role of firing rate changes, they employed a distribution-matching procedure to equalize firing rates across regions before comparing variability. They also used a random dot motion (RDM) stimulus that elicited fewer responses to assess variability quenching in non-responsive neurons.

Datasets

The study utilized electrophysiological recordings from two adult rhesus macaques performing a visual detection task. The data was collected during 113 recording sessions. No new benchmarks or datasets were introduced; the focus was on analyzing existing electrophysiological recordings.

Key Insights and Ideas

  • Thalamus is inherently less "noisy" than cortex: Even before any stimulus appears, the pulvinar and LGN show significantly lower trial-to-trial spiking variability compared to area V4. This suggests a more stable baseline processing in these thalamic structures.
  • Variability quenching is a cortical phenomenon: The robust decrease in variability observed in V4 upon stimulus onset is largely absent in the thalamic nuclei. This indicates that the mechanisms responsible for stabilizing neural activity in response to sensory input might be primarily cortical.
  • Pulvinar subregions differ: While neither the ventral nor dorsal pulvinar showed significant quenching, the dorsal pulvinar exhibited a small, marginal decrease, which was linked to the motor response required by the task. The ventral pulvinar, despite its strong connections with V4, showed no quenching at all.
  • Quenching isn't just about firing rate changes: Even when firing rates were matched across regions, V4 still showed variability quenching, while the thalamus did not. This suggests that the underlying mechanisms are not simply a consequence of increased or decreased firing.
  • Low baseline variability limits quenching potential: The authors propose that because the thalamus already operates with low variability, there's less room for further reduction upon stimulus onset.

Key Results

  • Baseline Variability: Fano factors prior to stimulus onset were significantly lower in vPul (1.13 ± 0.25), dPul (1.26 ± 0.59), and LGN (1.28 ± 0.45) compared to V4 (1.80 ± 0.98).
  • Variability Quenching:
    • Area V4 showed a substantial decrease in Fano factor (mean change -16%, p=6.91e-11).
  • Ventral pulvinar showed no significant change (mean change 0%, p=0.29).
  • LGN showed a slight, non-significant increase (mean change 2%, p=0.55).
  • Dorsal pulvinar showed a small but statistically significant decrease (mean change -4%, p=1.72e-05).
  • Response Strength: The quenching effect in V4 was negatively correlated with response strength, suggesting that stronger responses might contribute to greater variability reduction. However, this correlation was absent or weaker in the thalamic nuclei.
  • Non-responsive sites: Quenching was observed in V4 even in sites that did not show significant responses to the stimulus, further supporting that it's not solely driven by direct sensory input.

Limitations and Future Work

  • Limited number of animals: The study was conducted on only two macaques, which might limit the generalizability of the findings.
  • No direct manipulation of attention: The study did not directly manipulate attentional states, which are known to influence both cortical and thalamic activity and variability. The authors suggest that future work could explore how attention modulates these differences.
  • Motor response confound: The small quenching effect in the dorsal pulvinar was linked to the motor response required by the task. Future studies could use paradigms that decouple sensory processing from motor preparation to isolate the sensory quenching effect more cleanly.
  • Network architecture: The authors speculate that differences in local network architecture between the thalamus and cortex might explain the distinct variability dynamics. Further investigation into these microcircuitry differences could be fruitful.

The study highlights a fundamental divergence in how the thalamus and cortex handle neural variability, suggesting that the thalamus might provide a more stable substrate for sensory processing, while the cortex actively refines this information through stimulus-driven stabilization.

Summary updated

Abstract

Spiking activity exhibits a large degree of variability across identical trials, which has been shown to be significantly reduced by stimulus onset in a wide range of cortical areas. Whether similar dynamics apply to the thalamus and in particular to the pulvinar is largely unknown. Here, we examined electrophysiological recordings from two adult rhesus macaques performing a perceptual task and comparatively investigated trial-to-trial variability in higher-order thalamus (ventral and dorsal pulvinar), the lateral geniculate nucleus (LGN) and visual cortex (area V4) prior to and following the presentation of a visual stimulus. We found spiking variability during stable fixation prior to stimulus onset to be considerably lower in both pulvinar and the LGN as compared to area V4. In contrast to the prominent variability reduction in V4 upon stimulus onset, variability in the thalamic nuclei was largely unaffected by visual stimulation. There was a small but significant variability decrease in the dorsal pulvinar, but not in the ventral portion of the pulvinar, which is closely connected to visual cortices and would thus have been expected to reflect cortical response properties. This dissociation did not stem from differences in response strength or mean firing rates and indicates fundamental differences in variability quenching between thalamus and cortex.
BibTeX
@inproceedings{poland2019thalamus,
  author = {Eva Poland and Tobias H. Donner and Kai-Markus Müller and David A. Leopold and Melanie Wilke},
  title = {{Thalamus exhibits less sensory variability quenching than cortex}},
  booktitle = {Scientific},
  year = {2019},
  doi = {10.1038/s41598-019-43934-9},
}
Figure