Functional clustering of dendritic activity during decision-making

Aaron Kerlin, Boaz Mohar, Daniel Flickinger, Bryan MacLennan, Matthew B Dean, Courtney Davis, Nelson Spruston, Karel SvobodaeLife 2019 · DOI

TL;DR

This paper uses a novel high-resolution, large-scale imaging system to investigate dendritic activity in motor cortex neurons during a decision-making task. They found that while most dendritic activity is globally driven, task-specific signals are spatially clustered within branches, suggesting a mechanism for individual neurons to process complex information and support learning.

What Problem Does This Paper Address?

Dendrites, the branched extensions of neurons, are known to have "active properties" that can perform local computations. However, there's been a long-standing debate about how prevalent and selective these local operations are in vivo, especially during complex behaviors like decision-making. Previous studies often had limitations in imaging resolution, temporal speed, or the extent of the dendritic tree they could observe, leading to conflicting views on whether local dendritic activity is rare and globally driven, or if it plays a significant role in processing specific information. This paper aims to resolve this by providing a more comprehensive view of dendritic activity during a behavioral task.

Methods

The authors developed a custom two-photon laser-scanning microscope capable of high-speed (15 Hz), high-resolution imaging of dendrites and individual spines across large sections (up to 300 µm) of the dendritic tree. This system allowed them to simultaneously record calcium signals from the soma and extensive dendritic arbors. They also developed novel analysis methods for motion correction and for estimating the spatial scales of activity within dendrites and spines. A key aspect of their analysis involved a sophisticated method to subtract the contribution of back-propagating action potentials (bAPs) from dendritic signals, allowing them to better isolate task-related activity.

Datasets

The study used calcium imaging data from GCaMP6f-expressing pyramidal neurons in the anterior lateral motor (ALM) cortex of mice performing a tactile decision-making task. The task involved mice learning to lick left or right based on the location of a presented pole. The data collected included calcium transients from the soma, dendrites, and individual spines across different behavioral epochs (sampling, delay, response).

Key Insights and Ideas

  • Global vs. Local Activity: The study found that the majority of dendritic calcium transients were coincident with global events (likely driven by action potentials reaching the dendrite). However, they also identified a significant population of task-associated calcium signals that were not globally driven.
  • Spatial Clustering: These task-associated signals were not randomly distributed. Instead, they were compartmentalized by dendritic branching and clustered within branches over approximately 10 µm. This suggests that specific dendritic branches might be specialized for processing certain aspects of the task.
  • Functional Specialization within Neurons: The diverse behavior-related signals were intermingled throughout the dendritic arbor, implying that a single neuron can represent multiple, distinct pieces of information. This spatial organization could significantly enhance a neuron's computational capacity and its ability to support learning.
  • Novel Imaging and Analysis: The development of a high-resolution, large-scale imaging system and advanced analysis techniques (especially for bAP subtraction) were crucial for overcoming previous limitations and providing a more accurate picture of dendritic activity in vivo.

Key Results

  • The custom microscope allowed imaging of up to 300 µm of dendrite and hundreds of spines at 15 Hz.
  • While most dendritic calcium transients were global, a subset of task-associated signals showed compartmentalization.
  • These task-associated signals clustered within branches over approximately 10 µm.
  • Signal correlations between nearby dendritic segments and spines decayed exponentially with a length constant of ~10 µm, indicating functional similarity over short distances.
  • Dendritic branching points acted as barriers, compartmentalizing task-related calcium signals, with correlations being lower across branches than within them.

Limitations and Future Work

The authors acknowledge that their method for subtracting bAP components, while advanced, still has limitations and that their simulations showed that inferred correlations between input and output could be inaccurate. They suggest that future work could incorporate more sophisticated nonlinear inference methods and a priori information about spike trains to more accurately disambiguate pre- and post-synaptic activity. They also note that the observed long-lasting dendritic events in L5 neurons could be part of the cellular mechanism for short-term memory, suggesting future investigations into this role. Finally, they propose that exploring the relationship between anatomical changes in dendrites and clustered task-related activity during learning could be a fruitful avenue for future research.

Summary updated

Abstract

The active properties of dendrites can support local nonlinear operations, but previous imaging and electrophysiological measurements have produced conflicting views regarding the prevalence and selectivity of local nonlinearities in vivo. We imaged calcium signals in pyramidal cell dendrites in the motor cortex of mice performing a tactile decision task. A custom microscope allowed us to image the soma and up to 300 μm of contiguous dendrite at 15 Hz, while resolving individual spines. New analysis methods were used to estimate the frequency and spatial scales of activity in dendritic branches and spines. The majority of dendritic calcium transients were coincident with global events. However, task-associated calcium signals in dendrites and spines were compartmentalized by dendritic branching and clustered within branches over approximately 10 μm. Diverse behavior-related signals were intermingled and distributed throughout the dendritic arbor, potentially supporting a large learning capacity in individual neurons.
BibTeX
@inproceedings{kerlin2019functional,
  author = {Aaron Kerlin and Boaz Mohar and Daniel Flickinger and Bryan MacLennan and Matthew B Dean and Courtney Davis and Nelson Spruston and Karel Svoboda},
  title = {{Functional clustering of dendritic activity during decision-making}},
  booktitle = {eLife},
  year = {2019},
  doi = {10.7554/elife.46966},
}
Figure