My Career – Chapter 2: Non-Synaptic Plasticity

Integrating Behavior and Neuronal Populations at the Single-Cell Level to Decode Sensory Processing and Memory Engrams

Context and Significance

This chapter marks a pivotal phase in my research career at the University of Chicago, where I actively expanded my expertise beyond biology, immersing myself in engineering and computational neuroscience. Building on the foundation of my earlier work (see Chapter 1), I set out to study behavior and neural activity synchronously, breaking away from traditional approaches that examined these aspects in isolation.

To achieve this, I independently designed and built custom tools capable of imaging single-cell-level data from large neuronal populations – up to 400 neurons (video here) – while simultaneously capturing multimodal behavioral datasets. This required mastering CAD design, 3D printing, system engineering, and programming (MatLab, Python, and C++). These tools integrated neuronal recording with two-photon microscopy and behavioral analysis, facilitating precise studies of neural plasticity in awake, active animals.

My research focused on understanding how sensory experiences shape neural circuits and memory engrams, with an emphasis on intrinsic plasticity (IP) in sensory coding and ensemble formation in the barrel cortex. This work challenged the traditional view that synaptic plasticity alone drives memory and learning, instead revealing the critical role of changes in intrinsic excitability in integrating neurons into memory ensembles.

"Dr Lambot developed a novel behavioral paradigm to study natural touch in rodents. This involved engineering novel instruments to evoke natural touch and the application of advance machine learning algorithms to quantify natural touch. Of course, Dr Lambot is a neuroscientist so she combined this new set of instruments with brain recordings. [...] The breadth of her work is highly unusual and demonstrates a tremendous intellect."

Jason N. MacLean, PhD Professor of Neurobiology and Director of Undergraduate Studies in Neuroscience at the University of Chicago

The Interplay of Plasticity Mechanisms

Memory engrams—groups of interconnected neurons activated during learning—require both synaptic and intrinsic plasticity. While LTP strengthens connections, IP enhances a neuron’s likelihood of firing, ensuring its participation in ensemble activity.

Key findings include:

Intrinsic Plasticity’s Functional Role
Changes in excitability enable neurons previously unresponsive to sensory input to fire action potentials, allowing them to integrate into memory ensembles.
Mechanistic Insights
This process depends on SK2 channels, which regulate potassium conductance and excitability thresholds, linking membrane potential dynamics to ensemble coding.
Innovation in Experimental Design
By combining advanced imaging, behavioral setups, and computational tools, I captured how sensory stimulation reshapes neural activity at both single-cell and population levels.

Behavioral Integration and Imaging

This project involved designing and implementing a comprehensive setup to measure neuronal activity and mouse behavior simultaneously, a significant leap from my earlier work (see Chapter 1). It required years of iterative design, problem-solving, and interdisciplinary expertise.

Experimental Setup
I developed a custom platform combining two-photon calcium imaging with precise behavioral tracking, tailored for awake, head-fixed mice. The system recorded activity from hundreds of layer 2/3 neurons in the barrel cortex during voluntary whisker exploration—a behavior critical for sensory processing. By synchronizing multiple data streams with high temporal precision, I captured both neuronal and behavioral dynamics, enabling naturalistic paradigms where mice explored textured surfaces actively or passively, akin to human tactile exploration.

Each component of the setup was carefully engineered for accuracy and reliability:

Whisker Kinematics: Individual whiskers were illuminated with infrared light and recorded at high frequencies (200–500 Hz) using two synchronized cameras. These recordings were processed through deep learning algorithms (DeepLabCut) to generate precise 3D pose estimations, capturing detailed dynamics of active and passive whisker movements. Check this video for an example of whisker dynamics with an overlay of detections generated through deep learning.
Tactile Stimulus Control: A rotating cylinder, controlled by a servo motor via Arduino, moved in and out of reach of the whiskers to simulate tactile interactions. I recorded both the digital write signals from the Arduino (indicating the intended motor commands) and the actual movement of the cylinder using a secondary rotary encoder. This dual recording approach ensured a precise and reliable log of the tactile stimulus position and timing.
Pupilometry: A unique feature of the setup leveraged the scattered infrared light from the two-photon excitation used for neuronal imaging. This scattered light illuminated the pupils, creating high-contrast images where the pupils appeared as bright white spots. This incidental illumination proved invaluable, enabling me to capture high-quality pupil dynamics without additional hardware. The data was compressed using FFmpeg and processed with DeepLabCut to extract detailed measurements of pupil diameter and its changes over time, offering insights into arousal and attention states. Check this video for an example of pupil dynamics with an overlay of detections generated through deep learning.
Locomotion Dynamics: A rotary encoder attached to a treadmill measured running speed and movement patterns, correlating locomotion with neural responses to sensory stimuli.

All components, including the two-photon imaging system for recording neuronal activity, were synchronized via a master trigger. This ensured perfect temporal alignment across modalities, allowing me to investigate how sensory inputs, behavioral states, and neural activity interacted in real-time. Look at the following video for a detailed explanation of all components and an example recording:

"Dr. Lambot’s groundbreaking research has demonstrated that memories can be artificially implanted without natural stimuli, leading to significant implications for the study of memory and its associated disorders. Her work holds great promise for developing novel treatments for memory-related conditions such as dementia or Alzheimer’s disease."

Christian Hansel, PhD Professor of Neurobiology at the University of Chicago

In some instances, we used controlled visual stimuli to locate and study memory formation in the primary visual cortex (V1). The stimuli, similar to the examples shown in this video, were generated using MATLAB’s Psychtoolbox, enabling precise and reproducible experiments. These foundational studies provided insights into V1’s role in memory and sensory processing.

Functional Networks and Ensemble Analysis

Plasticity and Ensemble Dynamics

The ultimate goal of my experiments was to understand how memory formation reshapes neural circuits. By inducing artificial tactile memories in mice, I observed two pivotal outcomes:

Refined Coupling in Engrams
Neuronal coupling within the engram strengthened, enhancing coordination among ensemble members and reflecting increased network efficiency.
Engram Expansion
New neurons were recruited into the engram, primarily drawn from previously unreliable populations. These neurons transitioned from dormant to active participants in memory encoding, contributing to the reinforcement of the implanted memory.

To uncover these dynamics, I employed encoding models to predict neuronal activity based on behavioral and neural coupling data. Directed, weighted networks mapped interactions within ensembles, revealing how sensory inputs and intrinsic network dynamics shaped ensemble activity. By analyzing shifts in behavioral predictors—such as whisker velocity and pupil dynamics—I demonstrated that memory induction not only strengthened existing connections but also restructured the engram to integrate new contributors.

Key Findings:

Engram Formation and Stability: Sensory stimulation increased the number of reliably responsive neurons in the barrel cortex, an effect that persisted across days.
Intrinsic Plasticity and Spike Firing: Patch-clamp recordings showed that intrinsic plasticity (IP) amplified spike output, even when synaptic input remained unchanged.
Behavioral Relevance: Genetically modified mice lacking IP (SK2-KO) exhibited impaired tactile learning, underscoring the functional importance of intrinsic plasticity.

By synthesizing computational methods, machine learning, and biological insights, this work provided a comprehensive understanding of sensory processing and memory plasticity at both the population and single-neuron levels. These findings offer a framework for understanding how memories are formed, stabilized, and reinforced through the recruitment and refinement of neural ensembles.

Description

Behavioral Imaging Integration
Developed a cutting-edge tool to simultaneously study behavior and neuronal activity, bridging the gap between engineering, computational neuroscience, and biology.