Abstract
Neuronal responses to external stimuli vary from trial to trial partly because they depend on continuous spontaneous variations of the state of neural circuits, reflected in variations of ongoing activity prior to stimulus presentation. Understanding how post-stimulus responses relate to the pre-stimulus spontaneous activity is thus important to understand how state dependence affects information processing and neural coding, and how state variations can be discounted to better decode single-trial neural responses. Here we exploited high-resolution CMOS electrode arrays to record simultaneously from thousands of electrodes in in-vitro cultures stimulated at specific sites. We used information-theoretic analyses to study how ongoing activity affects the information that neuronal responses carry about the location of the stimuli. We found that responses exhibited state dependence on the time between the last spontaneous burst and the stimulus presentation and that the dependence could be described with a linear model. Importantly, we found that a small number of selected neurons carry most of the stimulus information and contribute to the state-dependent information gain. This suggests that a major value of large-scale recording is that it individuates the small subset of neurons that carry most information and that benefit the most from knowledge of its state dependence.
Introduction
A central challenge in neuroscience is understanding the trial-to-trial variability of neural responses to identical stimuli. Often dismissed as stochastic noise, this variability may instead reflect the internal state of the neural circuit at the moment of stimulation. This study investigates the relationship between pre-stimulus spontaneous activity and post-stimulus responses in large-scale neural networks. By identifying key state variables, researchers aim to improve stimulus decoding and understand how information is distributed across the network.
Methods
The researchers utilized high-density CMOS micro-electrode arrays (MEAs) to record from 4,096 electrodes simultaneously in primary hippocampal neuronal cultures from rats ($n=5$). Cultures were grown for 24 days in-vitro. Electrical pulses were delivered from eight distinct stimulation sites.
The analysis focused on:
- State Variable Identification: Specifically, the time elapsed since the last spontaneous network burst ($T_B$).
- Information Theory: Mutual Information was used to quantify the stimulus information in neural responses and the gain provided by accounting for the pre-stimulus state.
- Modeling: Testing additive versus multiplicative gain models to describe the state-response relationship.
- Pharmacological Modulation: Application of norepinephrine to alter network synchrony.
Results
The Role of Network State ($T_B$)
The study identified $T_B$ as a primary driver of response variability. Immediately following a spontaneous burst, the network enters a "refractory" state where responses to stimuli are significantly attenuated. As $T_B$ increases, the network recovers, and the number of evoked spikes grows.
Multiplicative Gain and Information Sparsity
The relationship between the pre-stimulus state and the response was found to be multiplicative rather than additive. This suggests that the internal state acts as a gain control mechanism for the entire network.
Furthermore, the study revealed that stimulus information is highly concentrated. A small subset of neurons (approximately 1.5–2% of the total population) was found to carry 95% of the stimulus-related information. These "hub" neurons are also the ones that benefit most from accounting for state-dependence, which increased stimulus information by an average of 51%.
Effects of Neuromodulation
Norepinephrine treatment decreased network synchrony. While this led to an increase in raw stimulus information, it simultaneously decreased the information gain derived from state-dependence, indicating that less synchronized states are more reliable but less influenced by immediate history.
Discussion
These findings highlight that neural "noise" is often structured information about the network's internal state. The use of high-density recordings is crucial not because every neuron is equally informative, but because it allows for the identification of the sparse subset of neurons that dominate the functional code. Future research is needed to determine if downstream biological decoders utilize this state-dependent information in real-time and to validate these findings in in-vivo systems.
Summary updated