System level analyses of motor-related neural activities in larval Drosophila

Youngteak Yoon, Jeonghyuk Park, Atsushi Taniguchi, Hiroshi Kohsaka, Ken Nakae, Shigenori Nonaka, Shin Ishii, Akinao NosebioRxiv 2018 · DOI

TL;DR

This paper presents a novel computational pipeline for analyzing neural activity in the larval Drosophila central nervous system (CNS) during fictive locomotion. By combining advanced light-sheet microscopy with machine learning, they can identify specific neurons that are preferentially active during forward or backward movements, offering a systems-level view of motor control.

What Problem Does This Paper Address?

Understanding how the central nervous system (CNS) generates complex behaviors like locomotion is a fundamental challenge in neuroscience. While muscle activity patterns can be observed, directly linking them to the underlying neural activity in a comprehensive way has been difficult, especially in systems with thousands of neurons. Previous studies were often limited to smaller nervous systems or focused on individual neurons or small populations. This paper aims to bridge this gap by providing a method to analyze the activity of a large population of neurons in the Drosophila larval CNS during specific motor behaviors (forward and backward locomotion) to uncover the neural circuits involved.

Methods

The researchers employed a multi-pronged approach:

  • Advanced Imaging: They used a digital scanned light-sheet microscopy (ezDSLM) system to capture neural activity in isolated Drosophila larval CNS. This allowed for high temporal resolution (0.6s/volume) to record dynamic calcium signals (using GCaMP6f) and high spatial resolution to identify individual neurons (using a nuclear marker, mCherry.nls).
  • Automated Cell Detection and Signal Extraction: A key innovation is their image processing pipeline. It automatically detects neuron locations based on nuclear staining and then extracts the GCaMP fluorescence signals from these identified cells. This overcomes the manual labor and potential bias of manual cell identification.
  • Machine Learning for Motor State Classification: They adapted a previously developed machine learning method (CNN and hierarchical agglomerative clustering) to classify the recorded neural activity into distinct motor states: forward wave (FW), backward wave (BW), anterior burst (AT), and quiescence state (QS). This allows for objective assignment of time windows corresponding to specific behaviors.
  • Identification of Biased Neurons: By combining the motor state classification with the extracted neural activity, they identified neurons whose activity patterns were significantly correlated with either forward or backward locomotion. They defined specific criteria (e.g., maximum fluorescence change during BW vs. FW) to categorize these "biased" neurons.

Datasets

Not discussed in the paper. The study uses Drosophila larvae as the biological system and generates its own imaging data.

Key Insights and Ideas

  • Systems-Level View of Motor Control: The study demonstrates the feasibility of analyzing neural activity across a large population of neurons in the Drosophila larval CNS to understand motor control, moving beyond single-neuron or small-circuit analyses.
  • Machine Learning as a Powerful Tool: The successful application of their machine learning pipeline for both motor state classification and identifying biased neurons highlights the power of these methods for dissecting complex neural dynamics.
  • Identification of Novel Candidate Neurons: They identified specific groups of neurons, particularly near the SEZ-thoracic boundary, that are preferentially active during the early phase of backward locomotion. This suggests these neurons play a crucial role in initiating or regulating backward movement.
  • Scattered Nature of Motor Circuits: The finding that biased neurons are distributed across the VNC suggests that the circuits controlling forward and backward locomotion are not localized to a single region but are rather spread throughout the ventral nerve cord.

Key Results

  • The automated cell detection pipeline achieved an 86.2% hit rate compared to human annotation.
  • They successfully classified distinct motor states (FW, BW, AT, QS) from the neural activity data.
  • Approximately 150 BW-biased and 150 FW-biased cells were identified for each motor pattern.
  • A specific group of BW-biased neurons was found to be activated early in backward locomotion but not in forward locomotion, suggesting a role in initiating backward movement.

Limitations and Future Work

  • Neuron Identity: A significant limitation, acknowledged by the authors, is the difficulty in identifying the specific identity of the neurons showing characteristic activity patterns due to the pan-neuronal expression of GCaMP. Future work will involve combining this technique with sparse Gal4 lines to identify specific neuron types.
  • Brain Exclusion: For temporal resolution, parts of the brain were excluded from imaging. While previous studies suggest the VNC can autonomously generate locomotion, a complete understanding might require including brain activity.
  • Functional Characterization: The identified biased neurons are candidates. Further experiments using optogenetics and genetic manipulations are needed to confirm their specific roles in motor control.
  • Broader Behaviors: The study focused on forward and backward locomotion. Future work could extend to other behaviors like turning or more complex motor sequences.

Summary updated

Abstract

ABSTRACT The way in which the central nervous system (CNS) governs animal movement is complex and difficult to solve solely by the analyses of muscle movement patterns. We tackle this problem by observing the activity of a large population of neurons in the CNS of larval Drosophila . We focused on two major behaviors of the larvae, forward and backward locomotion, and analyzed the neuronal activity related to these behaviors during fictive locomotion that spontaneously occurs in the isolated CNS. We expressed genetically-encoded calcium indicator, GCaMP, and a nuclear marker in all neurons and used digital scanned light-sheet microscopy to record neural activities in the entire ventral nerve cord at a fast frame rate. We developed image processing tools that automatically detect the cell position based on the nuclear staining and allocate the activity signals to each detected cell. We also applied a machine learning-based method that we developed recently to assign motor status in each time frame. Based on these methods, we find cells whose activity is biased to forward versus backward locomotion and vice versa. In particular, we identified a group of neurons near the boundary of subesophageal zone (SEZ) and thoracic neuromeres, which are strongly active during an early phase of backward but not forward fictive locomotion. Our experimental procedure and computational pipeline enable systematic identification of neurons to show characteristic motor activities in larval Drosophila .
BibTeX
@misc{yoon2018system,
  author = {Youngteak Yoon and Jeonghyuk Park and Atsushi Taniguchi and Hiroshi Kohsaka and Ken Nakae and Shigenori Nonaka and Shin Ishii and Akinao Nose},
  title = {{System level analyses of motor-related neural activities in larval <i>Drosophila</i>}},
  year = {2018},
  doi = {10.1101/292136},
  howpublished = {bioRxiv},
}
Figure