Systematic annotation of a complete adult male Drosophila nerve cord connectome reveals principles of functional organisation

Elizabeth C. Marin, Billy J Morris, Tomke Stürner, Andrew S Champion, Dominik Krzemiński, Griffin Badalamente, Marina Gkantia, Christopher R Dunne, Katharina Eichler, Shin-ya Takemura, Imaan Fm Tamimi, Siqi Fang, Sung Soo Moon, Han Sj Cheong, Feng Li, Philipp Schlegel, Sebastian E. Ahnert, Stuart Berg, Gwyneth M Card, Marta Costa, David Shepherd, Gregory S.x.e. Jefferis2024 · DOI

Abstract

The acquisition of a complete connectome is a milestone in neuroscience, yet a raw wiring diagram requires systematic annotation to be biologically interpretable. This study provides a comprehensive atlas of the adult male Drosophila ventral nerve cord (VNC), categorizing nearly 16,000 neurons. By integrating developmental lineage data, symmetry, and connectivity-based clustering, we identify principles of functional organization, including the distinction between early-born "primary" neurons and later-born "secondary" neurons. We further map sensory modalities and identify serial homologues across thoracic segments, providing a foundational framework for understanding the neural basis of insect locomotion and behavior.

Introduction

A connectome, in its rawest form, is a massive graph of nodes and edges representing neurons and synapses. While the reconstruction of the adult male Drosophila nerve cord (MANC) provides the structural "wiring diagram" (Takemura et al., 2023), the biological meaning of these connections remains opaque without systematic labeling. The Ventral Nerve Cord (VNC) serves as the insect's equivalent to the vertebrate spinal cord, housing the circuitry required for walking, flying, and grooming. This paper details the effort to transform the MANC dataset into a searchable, categorized, and biologically grounded atlas.

The Challenge of the Unlabeled Connectome

The VNC contains approximately 16,000 neurons, the majority of which were previously unnamed and functionally uncharacterized. To bridge the gap between high-resolution electron microscopy data and decades of "light-level" biological research, a systematic nomenclature was required. This nomenclature needed to account for the VNC's segmental nature, its developmental origins, and the diverse sensory inputs it processes.

Methodological Framework for Annotation

The researchers employed a multi-layered approach to categorize 15,762 neurons:

  • Developmental Mapping: Neurons were assigned to "hemilineages"—groups of cells derived from the same neural stem cell (neuroblast). Hemilineage identity is a strong predictor of a neuron’s neurotransmitter profile and general morphology.
  • Symmetry and Serial Homology: The team identified left-right pairs and "serial homologues"—neurons that repeat across the three thoracic segments (T1, T2, and T3) responsible for controlling the front, middle, and hind legs.
  • Connectivity-Based Clustering: Using cosine similarity of synaptic connections, neurons were grouped into "types." Neurons sharing similar input and output partners were classified together.
  • Machine Learning and NBLAST: Support Vector Machines (SVM) predicted the "birthtime" of neurons (embryonic vs. larval) based on morphological features like fiber diameter. NBLAST was used to compare 3D EM reconstructions with existing light-microscopy libraries.

Organizational Principles of the VNC

The annotation process revealed several key insights into the functional logic of the nervous system:

The "Business End" of the Nervous System

While the brain is larger, the VNC exhibits a higher density of sensory and motor neurons. The ratio of intrinsic (local) neurons to motor neurons in the VNC is 9:1, a stark contrast to the 55:1 ratio found in the brain, highlighting the VNC's role in direct motor control.

Developmental Birth Order

The study found that "primary" neurons (born early in the embryo) form a "rich club" within the nervous system. These neurons are centrally located, highly connected, and act as global hubs. In contrast, "secondary" neurons (born later during the larval stage) are more specialized and insular.

Leg vs. Flight Circuitry

The organizational logic differs significantly between locomotor modalities. Leg-related circuits are primarily local and ipsilateral (restricted to one side). Conversely, flight-related circuits in the "tectulum" exhibit complex bilateral and diagonal coordination, reflecting the millisecond-level synchronization required for wing control.

Key Results and Data Integration

The systematic effort achieved 88% coverage in assigning developmental hemilineages to VNC neurons. Furthermore, over 5,000 sensory neurons were categorized by modality, including chemosensory (taste/smell), tactile (touch), and proprioceptive (body position) inputs. By identifying serial homologues for over one-third of all neurons, the researchers uncovered recurring "circuit motifs" that are reused across segments for locomotor control.

To ensure community utility, all annotations have been integrated into the neuPrint and Clio web applications, providing a programmatically accessible resource for global research.

Limitations and Future Directions

Despite the comprehensiveness of this atlas, challenges remain. Truncated neurons at the volume edges and staining artifacts can lead to missing synaptic data. Additionally, the current automated methods focus on chemical synapses, likely underrepresenting electrical "gap junctions." The abdominal segments (ANm) also remain more difficult to categorize than the thoracic segments due to a lack of clear segmental repeats.

The ultimate goal is "whole-body" connectomics. By linking this VNC atlas to the complete brain connectome, researchers will be able to trace the entire path of a neural command—from the decision-making centers in the head to the motor execution in the legs and wings.

Summary updated

Abstract

Summary Nervous systems function as ensembles of neurons communicating via synaptic connections, and a functional understanding of nervous systems requires extensive knowledge of their connectomes. In a companion paper (Takemura et al., 2023), we describe the acquisition of a complete fruit fly nerve cord connectome, the first for an animal that can walk or fly. Here, to efficiently navigate and to appreciate the biological significance of this connectome, we categorise and name nearly all neurons systematically and link them to the experimental literature. We employ a system of hierarchical coarse annotations and group similar neurons across the midline and across segments, then define systematic cell types for sensory neurons, intrinsic neurons, ascending neurons, and non-motor efferent neurons. Stereotyped arrays of neuroblasts generate related neuron populations called hemilineages that repeat across the segments of the nerve cord. We confirm that larval-born neurons from a given hemilineage generally express the same neurotransmitter but find that earlier born neurons often express a different one. We match over 35% of intrinsic, ascending, and non-motor efferent neurons across segments, defining serial sets which were crucial for systematic typing of motor neurons and sensory neurons. We assign a sensory modality to over 5000 sensory neurons, cluster them by connectivity, and identify serially homologous cell types and a layered organisation likely corresponding to peripheral topography. Finally, we present selected examples of sensory circuits predicated on programmatic analysis of a complete VNC connectome. Our annotations are critical for analysing the structure of descending input to the nerve cord and of motor output, both described in a third companion paper (Cheong et al., 2023). These annotations are being released as part of the neuprint.janelia.org and clio.janelia.org web applications and also serve as the basis for programmatic analysis of the connectome through dedicated tools that we describe in this paper.
BibTeX
@misc{marin2024systematic,
  author = {Elizabeth C. Marin and Billy J Morris and Tomke Stürner and Andrew S Champion and Dominik Krzemiński and Griffin Badalamente and Marina Gkantia and Christopher R Dunne and Katharina Eichler and Shin-ya Takemura and Imaan FM Tamimi and Siqi Fang and Sung Soo Moon and Han SJ Cheong and Feng Li and Philipp Schlegel and Sebastian E. Ahnert and Stuart Berg and Gwyneth M Card and Marta Costa and David Shepherd and Gregory S.X.E. Jefferis and Gregory SXE Jefferis},
  title = {{Systematic annotation of a complete adult male Drosophila nerve cord connectome reveals principles of functional organisation}},
  year = {2024},
  doi = {10.7554/elife.97766},
}

Loading related content...

Figure