henhaobio
Sign in

bioRxivNeurosciencePreprint

A Connectome of the Male Drosophila Ventral Nerve Cord

Shin-ya TakemuraiD, Kenneth J. HayworthiD, Gary B. HuangiD, Michał JanuszewskiiD, Zhiyuan LuiD, Elizabeth C. MariniD, Stephan PreibischiD, C. Shan XuiD, John BogoviciD, Andrew S ChampioniD, Han SJ CheongiD, Marta CostaiD, Katharina EichleriD, William T. KatziD, Christopher J. KnechtiD, Feng LiiD, Billy J MorrisiD, Christopher OrdishiD, Patricia K. RivliniD, Philipp SchlegeliD, Kazunori ShinomiyaiD, Tomke StürneriD, Ting ZhaoiD, Griffin S. BadalamenteiD, Dennis BaileyiD, , Stuart BergiD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Animal behavior is principally expressed through neural control of muscles. Therefore understanding how the brain controls behavior requires mapping neuronal circuits all the way to motor neurons. We have previously established technology to collect large-volume electron microscopy data sets of neural tissue and fully reconstruct the morphology of the neurons and their chemical synaptic connections throughout the volume. Using these tools we generated a dense wiring diagram, or connectome, for a large portion of the Drosophila central brain. However, in most animals, including the fly, the majority of motor neurons are located outside the brain in a neural center closer to the body, i.e. the mammalian spinal cord or insect ventral nerve cord (VNC). In this paper, we extend our effort to map full neural circuits for behavior by generating a connectome of the VNC of a male fly.

In-depth summary

Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 1 (6 Jun 2023) · CC BY 4.0

Roughly 23 thousand traced neurons and 10 million TBars of the male Drosophila nerve cord, reconstructed from electron microscopy, typed, and queryable in neuPrint down to motor neurons.

Dense fly connectomes so far (optic lobe columns, the female hemibrain) omit premotor circuits and motor neurons, which lie mostly in the ventral nerve cord (VNC), so central computations could not be followed to muscles. This is presented as the first densely reconstructed VNC connectome, taken from a male to include male-specific courtship and mating motor programs and to complement a partially proofread female VNC.

Key results

  1. ShownMANC, a dense connectome of an adult male Drosophila VNC, is publicly queryable in neuPrint, with Python, R and Cypher APIs and a dataset-specific malevnc R package.

    About 23 thousand traced neurons, 10 million TBars and 74 million PSDs (postsynaptic densities), plus 44m of neuronal cable. Per-neuropil completeness (the share of synapses with both partners in identified neurons) is given in Table 1, not in the prose. · 1 Introduction and Motivation; 2.8 Proofreading; Tables

  2. ShownCutting the >500 µm-long VNC into 25 µm hot-knife slabs let it be imaged by FIB-SEM in parallel and then stitched back together computationally.

    26 slabs were cut and 25 were imaged on seven customized FIB-SEM systems over six months, at 8-nm XY pixels with 8-nm milling steps. A Zeiss Capella FIB column replaced the FEI Magnum used for the hemibrain. Slabs were aligned with a render-based pipeline, surface finding and distributed deformable registration across hot-knife seams. · 2.2 Hot knife cutting; 2.3 Imaging; 2.4 Alignment

  3. SuggestedMulti-scale flood-filling networks, with lower-resolution models trained on hemibrain neurites, segmented the VNC directly. Added merge constraints limited bodies containing several merged cells.

    The 32 and 16 nm models, trained on 111 proofread hemibrain neurites, worked on CLAHE-normalized VNC images without CycleGAN normalization. SECGAN output replaced only voxels within 5 voxels of slab interfaces. Merges were forbidden between annotated somas, neck fibers and nerve bundles, and between objects above 10M voxels. The proofreading time saved is not quantified. · 2.5 Segmentation

  4. ClaimedEach presynaptic site carries predicted probabilities for acetylcholine, GABA, glutamate or none. Every neuron has a coarse class, a systematic type and a type name, plus hemilineage, left-right group and serial-set annotations where these apply.

    The classifier was trained on 187 neurons identified from light-level data (67 acetylcholine, 55 GABA, 65 glutamate) with an 80%/20% train/validation split; no accuracy is reported in the text. Typing covered 2065 descending and motor neurons and 21683 sensory and intrinsic neurons, detailed in companion papers. · 2.10 Neurotransmitter prediction; 2.11 Hemilineage annotation; 2.13 Group and serial set annotation; 2.14 Cell typing and nomenclature

  5. ClaimedMANC is described as the first complete nerve cord connectome, the first covering a bilaterally complete region of an adult central nervous system, and the connectome capturing the most synaptic connectivity released to date.

    Stated in the concluding remarks. The text gives no quantitative comparison with other connectomes. · 2.15 Concluding Remarks and Future Research

Methods, data and reproducibility

The VNC of a five-day-old male was cut into 25 µm hot-knife slabs, imaged by focused ion beam scanning electron microscopy (FIB-SEM) at 8 nm, stitched, and segmented with multi-scale flood-filling networks plus a synapse detector. Neurons were proofread using expected left-right partners, registered to the JRC2018M template, and annotated with predicted neurotransmitters, hemilineages and cell types.

  • FIB-SEM volume of the ventral nerve cord of one five-day-old male (Canton S G1 x w^1118), in 25 slabs at 8 nm · Drosophila melanogaster · n = roughly 23 thousand traced neurons · neuPrint dataset MANC
  • Manually proofread hemibrain neurites used to train the lower-resolution flood-filling network models · Drosophila melanogaster (female central brain) · n = 111 neurites
  • Ground-truth neurotransmitter labels from morphological identification with light-level data, provided by the Dickson lab · Drosophila melanogaster · n = 187 neurons (67 acetylcholine, 55 GABA, 65 glutamate)
  • JRC2018M male ventral nerve cord template used for registering EM to light microscopy · Drosophila melanogaster
Methods & tools
Zeiss Versa 520 x-ray CT, FIB-SEM (Zeiss Capella FIB column), render web services, hot-knife (surface finding and stitching), SIFT, multi-scale flood-filling networks, SECGAN, elastix, BigWarp, NBLAST / CBLAST / NBLAST-DASH, neuPrint, malevnc R package
Experimental conditions
Imaging used seven customized FIB-SEM systems running in parallel for six months. No computational hardware or run times are given for alignment, segmentation or synapse detection.
Reproducibility gaps
neuPrint access requires a Google account. Detailed alignment methods are deferred to a paper in preparation (Preibisch, Trautman and Saalfeld, 2023). No code links are given for segmentation, synapse detection or the neurotransmitter models.
Limitations (5)
  • Nerves hold far fewer synapses and are much less completely reconstructed than neuropils, largely because of truncation damage during sample preparation. · stated by the authors
  • Neuropils were defined from synapse point clouds, with boundaries drawn by hand; some structures in the standard VNC nomenclature could not be identified this way. · stated by the authors
  • Neurotransmitter predictions cover only acetylcholine, GABA and glutamate (plus an unknown class). The per-synapse values are relative probabilities, and other transmitters may be present. · stated by the authors
  • Single specimen, so wiring variability between animals cannot be assessed here; the authors defer this to a matched sample in a forthcoming male central nervous system dataset. · noted in this summary
  • Most quality metrics are outside the prose: synapse precision-recall appears only in Fig. 4 and completeness only in Table 1. Neurotransmitter prediction accuracy is not reported, and validation of cell typing is deferred to the companion papers. · noted in this summary

Read this if you plan to query the MANC connectome in neuPrint or malevnc and need to know how the male fly nerve cord was imaged, segmented, proofread and annotated.

Citations
74

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISAzevedo '24Cheong '23Stürner '24Bates '26Marin '23Chen '18
87% semantic matchNeuroscience
Anthony W. Azevedo, Ellen Lesser, Jasper S. Phelps+32 more·Nature·Jun 26, 2024· 119 citations