Neuronal activity in sensory cortex predicts the specificity of learning

Katherine C. Wood, Christopher F. Angeloni, Karmi Oxman, Claudia Clopath, Maria N. GeffenbioRxiv 2020 · DOI

TL;DR

This study reveals that the pre-learning activity patterns in the auditory cortex predict how specifically an individual mouse will learn to fear a dangerous stimulus. Interestingly, the post-learning activity patterns, even though reorganized, do not correlate with this learned specificity, suggesting the auditory cortex's role is primarily in the acquisition, not the storage, of fear memory.

What Problem Does This Paper Address?

Fear learning is crucial for survival, but individuals vary greatly in how they learn to associate a neutral cue with danger. Some learn to fear only the specific dangerous cue (high specificity), while others generalize their fear to similar, safe cues (low specificity). This generalization is particularly problematic in disorders like PTSD. While we know that the auditory cortex (AC) is involved in fear learning, the specific neuronal mechanisms that explain why some individuals learn more specifically than others remain unclear. This paper aims to uncover the neuronal basis for this inter-subject variability in fear specificity.

Methods

The researchers used two-photon calcium imaging to record neuronal activity in the auditory cortex of head-fixed mice. They employed differential fear conditioning (DFC), where a specific tone (CS+) was paired with a foot-shock, while another tone (CS-) was not. They tracked the same neurons over multiple imaging sessions, both before and after DFC. To quantify neuronal activity and its predictive power, they used:

  • Z-score difference (Zdiff): A measure of how well individual neurons could discriminate between the CS+ and CS- tones before conditioning.
  • Support Vector Machine (SVM): A machine learning classifier trained on population-level neuronal responses to discriminate between CS+ and CS- before conditioning.
  • Freezing behavior: A common measure of fear in rodents, used to assess learning specificity after conditioning.

Datasets

Not explicitly detailed as separate datasets; the study utilized data collected from mice undergoing the described experimental paradigm.

Key Insights and Ideas

  • Predictive power of pre-learning AC activity: The most striking finding is that the state of the auditory cortex before any fear conditioning occurs is a strong predictor of how specifically an animal will learn. This suggests that individual differences in sensory processing might pre-dispose animals to more specialized or generalized fear learning.
  • Auditory cortex's role in acquisition, not storage: While the AC's activity reorganizes after fear conditioning, this reorganization doesn't predict learning specificity. This strongly implies that the AC is critical for learning the association and specificity, but not for retaining the fear memory itself.
  • Population coding matters: Both individual neuron discriminability (Zdiff) and population-level decoding (SVM) before conditioning predicted learning specificity, highlighting the importance of collective neuronal activity.

Key Results

  • Pre-DFC neuronal discriminability predicts learning specificity: Mean Zdiff scores and SVM performance from pre-DFC imaging sessions significantly correlated with learning specificity measured 24 hours post-DFC.
  • Post-DFC neuronal activity does not predict specificity: Correlations between post-DFC neuronal activity and learning specificity were not significant.
  • Auditory cortex activity changes after DFC: Normalized responses to the CS+ tone increased in conditioned mice after DFC, suggesting some plasticity, but this change didn't correlate with learning specificity.
  • Model supports findings: A computational model incorporating auditory nuclei and the amygdala demonstrated that AC tuning width (representing discriminability) could explain variations in fear specificity and generalization.

Limitations and Future Work

  • Pure tone stimuli: The study used pure tones, which are simpler than natural sounds. Future work could explore if these findings extend to more complex auditory stimuli.
  • Auditory cortex layers: Recordings were limited to layers 2 and 3 of the AC. Investigating other layers might reveal more specific changes.
  • Neuronal subtypes: The study did not distinguish between different neuronal subtypes, which might have differential roles in fear conditioning.
  • MGB and BLA roles: The authors suggest future studies should explore the role of the medial geniculate body (MGB) and basolateral amygdala (BLA) and their projections in fear learning and memory.
  • Random drift: The paper suggests that random drift in neuronal tuning might explain the lack of correlation between neuronal tuning and learning specificity after conditioning.

Summary updated

Abstract

Abstract Learning to avoid dangerous signals while preserving normal responses to safe stimuli is essential for everyday behavior and survival. Fear learning has a high level of inter-subject variability. Following identical experiences, subjects exhibit fear specificities ranging from high (specializing fear to only the dangerous stimulus) to low (generalizing fear to safe stimuli). Pathological fear generalization underlies emotional disorders, such as post-traumatic stress disorder. The neuronal basis of fear specificity remains unknown. Here, we identified the neuronal code that underlies inter-subject variability in fear specificity using longitudinal imaging of neuronal activity before and after differential fear conditioning in the auditory cortex of mice. Neuronal activity prior to, but not after learning predicted the level of specificity following fear conditioning across subjects. Stimulus representation in auditory cortex was reorganized following conditioning. However, the reorganized neuronal activity did not relate to the specificity of learning. These results present a novel neuronal code that determines individual patterns in learning.
BibTeX
@misc{wood2020neuronal,
  author = {Katherine C. Wood and Christopher F. Angeloni and Karmi Oxman and Claudia Clopath and Maria N. Geffen},
  title = {{Neuronal activity in sensory cortex predicts the specificity of learning}},
  year = {2020},
  doi = {10.1101/2020.06.02.128702},
  howpublished = {bioRxiv},
}
Figure