Recording Chronically From the Same Neurons in Awake, Behaving Primates

Andreas S. Tolias, Alexander S. Ecker, Athanassios G. Siapas, A Hoenselaar, Georgios A. Keliris, Nikos K. LogothetisJournal of Neurophysiology 2007 · DOI

TL;DR

This paper introduces a novel statistical framework and chronic recording techniques to reliably track individual neurons in awake, behaving primates over extended periods (days to weeks). This breakthrough enables the study of how neural circuits change during learning and memory formation, a critical step towards understanding complex brain functions.

What Problem Does This Paper Address?

Understanding how learning and memory are encoded and stored in the brain requires observing how individual neurons and their interactions change over time. Traditional electrophysiological recordings are often "non-chronic," meaning they capture neural activity for short durations, making it impossible to track the same neurons across learning experiences. This limitation hinders the study of crucial questions, such as how memories are gradually formed across distributed neural circuits or how the brain generalizes learned information to novel situations. The authors aim to bridge this gap by developing a method for stable, long-term recording of individual neurons in primates.

Methods

The core of this paper lies in a novel statistical framework designed to identify and track individual neurons across multiple recording sessions. This framework leverages spike waveform information captured by tetrodes (bundles of four electrodes).

  • Spike Sorting: A mixture of Gaussians model is fitted to spike waveform data in a 12-dimensional feature space (derived from principal components of the waveforms). This is achieved using the Split and Merge Expectation Maximization (SMEM) algorithm.
  • Single Unit Isolation Quality: A contamination measure is used to quantify the quality of isolated units, ensuring that only well-separated neurons are included in the analysis.
  • Long-Term Stability Evaluation: A key innovation is the development of a method to quantitatively assess the stability of recorded neurons over days and weeks. This involves:
  • Distance Measures: Two distance measures are computed between average waveforms of neurons recorded on consecutive days: one capturing shape differences (normalized Euclidean distance after scaling) and another capturing amplitude differences.
  • Null Distribution: A "null distribution" of pairwise distances is empirically determined from data collected during periods when tetrode stability was intentionally compromised (e.g., through daily adjustments). This represents the expected distance between waveforms of different neurons.
  • Statistical Hypothesis Testing: By comparing the experimental distribution of distances (from stable recording periods) to the null distribution, the authors can statistically determine if two recorded units are likely the same neuron. A significant deviation towards smaller distances in the experimental distribution indicates recording stability.
  • Orientation Tuning Function Analysis: As a validation, the stability of orientation tuning functions of neurons identified as stable by the waveform analysis is also examined.

Datasets

The study utilized electrophysiological recordings from awake, behaving macaques. The data was collected over extended periods, with some recordings spanning up to one year after implantation. Specific details about the number of recording sessions and the duration of stable recording periods are provided in the text and figures (e.g., Period A: 22 recording sessions over 23 days; Period B: 9 recording sessions over 9 days).

Key Insights and Ideas

  • Extracellular Signatures for Neuron Identification: The paper demonstrates that spike waveform characteristics, when analyzed with a sophisticated statistical framework, provide a reliable "extracellular signature" for identifying individual neurons over long periods. This moves beyond simply assuming waveform similarity implies the same neuron.
  • Quantitative Stability Assessment: The development of a statistically rigorous method to assess single-unit stability, grounded in an empirically derived null distribution, is a significant contribution. This allows for objective evaluation of recording reliability.
  • Linking Waveform Stability to Functional Stability: The study shows a strong correlation between the stability of spike waveforms and the stability of neuronal functional properties (orientation tuning). This provides a powerful validation of their stability metric.
  • Opening Doors for Learning Research: The ability to reliably track neurons over days and weeks is a critical enabler for studying the dynamic changes in neural circuits that underlie learning and memory.

Key Results

  • The developed statistical framework can quantitatively assess the accuracy of identifying stable neurons across successive recording sessions, demonstrating the ability to record from the same neurons for multiple consecutive days and weeks.
  • In recording sessions where multiple neurons were isolated from a tetrode, the method successfully identified and characterized these units.
  • Analysis of orientation tuning functions revealed that neurons identified as stable by waveform analysis also exhibited highly consistent tuning functions across days.
  • The study quantified the probability of a neuron being recorded on subsequent days, showing that a significant percentage of neurons could be tracked for several days (e.g., 61% for at least 2 days in Period A, 41% in Period B).
  • The paper also reported the absolute number of neurons recorded for extended periods, with some neurons being tracked for over a week.

Limitations and Future Work

  • Manual Intervention for Complex Cases: While the model fitting is largely automated, the authors acknowledge that in rare cases of overestimation of model complexity (leading to multiple mixtures for a single neuron), manual merging of these mixtures is required. Additionally, grouping components for complex spiking neurons may require user intervention.
  • Tetrode Geometry and Source Localization: A major improvement suggested for future work is to incorporate exact tetrode geometry. Knowing the precise positions of the four tetrode tips would allow for triangulation of neuronal sources using amplitude information, potentially leading to better separability of clusters.
  • Potential Bias in Null Distribution Estimation: While the authors argue their null distribution estimation is relatively unbiased, they acknowledge that tetrode properties could change over time, or different neuron types might be sampled during adjustment periods, potentially biasing the estimate.
  • Generalization of Findings: The study focused on area V1. Future work could explore the applicability of these methods to other brain areas and different behavioral tasks.

Summary updated

Abstract

Understanding the mechanisms of learning requires characterizing how the response properties of individual neurons and interactions across populations of neurons change over time. To study learning in vivo, we need the ability to track an electrophysiological signature that uniquely identifies each recorded neuron for extended periods of time. We have identified such an extracellular signature using a statistical framework that allows quantification of the accuracy by which stable neurons can be identified across successive recording sessions. Our statistical framework uses spike waveform information recorded on a tetrode's four channels to define a measure of similarity between neurons recorded across time. We use this framework to quantitatively demonstrate for the first time the ability to record from the same neurons across multiple consecutive days and weeks. The chronic recording techniques and methods of analyses we report can be used to characterize the changes in brain circuits due to learning.
BibTeX
@article{tolias2007recording,
  author = {Andreas S. Tolias and Alexander S. Ecker and Athanassios G. Siapas and A Hoenselaar and Georgios A. Keliris and Nikos K. Logothetis},
  title = {{Recording Chronically From the Same Neurons in Awake, Behaving Primates}},
  journal = {Journal of Neurophysiology},
  year = {2007},
  doi = {10.1152/jn.00260.2007},
}
Figure