henhao
Sign in
NeurosciencebioRxiv
Preprint

Transforming descending input into motor output: An analysis of the Drosophila Male Adult Nerve Cord connectome

Han SJ CheongiD, Katharina EichleriD, Tomke StürneriD, Samuel K. AsinofiD, Andrew S ChampioniD, Elizabeth C. MariniD, Tess Baker OramiD, Marissa G. SumathipalaiD, Lalanti VenkatasubramanianiD, Shigehiro NamikiiD, Igor SiwanowicziD, Marta CostaiD, Stuart BergiD, Janelia FlyEM Project Team, Gregory S.X.E. JefferisiD, Gwyneth M CardiD

Abstract

In most animals, a relatively small number of descending neurons (DNs) connect higher brain centers in the animals head to circuits and motor neurons (MNs) in the nerve cord of the animals body that effect movement of the limbs. To understand how brain signals generate behavior, it is critical to understand how these descending pathways are organized onto the body MNs. In the fly, Drosophila melanogaster, MNs controlling muscles in the leg, wing, and other motor systems reside in a ventral nerve cord (VNC), analogous to the mammalian spinal cord. In companion papers, we introduced a densely-reconstructed connectome of the Drosophila Male Adult Nerve Cord (MANC, (Takemura et al., 2024)), including cell type and developmental lineage annotation (Marin et al., 2024), which provides complete VNC connectivity at synaptic resolution. Here, we present a first look at the organization of the VNC networks connecting DNs to MNs based on this new connectome information. We proofread and curated all DNs and MNs to ensure accuracy and reliability, then systematically matched DN axon terminals and MN dendrites with light microscopy data to link their VNC morphology with their brain inputs or muscle targets. We report both broad organizational patterns of the entire network and fine-scale analysis of selected circuits of interest. We discover that direct DN-MN connections are infrequent and identify communities of intrinsic neurons linked to control of different motor systems, including putative ventral circuits for walking, dorsal circuits for flight steering and power generation, and intermediate circuits in the lower tectulum for coordinated action of wings and legs. Our analysis generates hypotheses for future functional experiments and, together with the MANC connectome, empowers others to investigate these and other circuits of the Drosophila ventral nerve cord in richer mechanistic detail.

PDFDOICodeData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
54
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (3 Jun 2025) · CC BY 4.0

1328 descending and 733 motor neurons of the male Drosophila nerve cord connectome were curated; direct links between them are infrequent, and intrinsic-neuron communities route control per motor system.

In the fly ventral nerve cord (VNC), only limited circuits linking brain-derived descending neurons (DNs) to motor neurons (MNs) were understood, and concurrent female nerve cord (FANC) work examined MNs and their immediate upstream partners. Every DN and MN in the densely reconstructed Male Adult Nerve Cord (MANC) connectome is proofread and typed, and matched to light-level data where possible, enabling VNC-wide analysis of DN-to-MN pathways.

Key results

  1. ShownMonosynaptic DN input is a small share of most MNs' input, so control runs mostly through VNC intrinsic circuits; neck MNs and a subset of abdominal MNs are the main exceptions.

    A groupwise mean of 7% of MN input comes from DNs, yet about 80% of MN groups receive at least 1% from some DN group; most neck MNs, some abdominal MNs and several wing and haltere MNs get >20% (up to ~60%). Only 13/92 lower-tectulum-targeting DNs give at least 1% input to any MN. Bayesian graph traversal places haltere-tectulum DNs about half as far from neck, wing and haltere MNs as from leg MNs. · 2.4.1 DN to MN connectivity across the VNC; 2.7.1 Organizational logic of lower tectulum…; 3 Discussion

  2. ShownPremotor intrinsic neurons form communities tied to motor systems, plus intermediate ones with little MN output feeding others; cyclic triads are less frequent in upper tectulum and leg neuropils (like Hemibrain ventrolateral neuropils) but more frequent among intermediate and lower tectulum-restricted cells.

    Infomap yielded 37 communities, 18 with 50 or more cells: one per leg neuropil per side, five upper tectulum, three abdominal, three intermediate. Community 7 (~1800 cells) is the top non-self input to most others. Cyclic triads in the whole VNC, upper tectulum and leg neuropils are rarer than in Hemibrain mushroom body and central complex; lower/intermediate tectulum-restricted cells reach similar levels. · 2.4.2 Community structure of VNC networks

  3. ShownAll DNs and MNs were proofread, typed and given systematic names. The DN count exceeds earlier estimates, two previously undescribed DN tracts were found, and many DNs plus most wing, haltere and leg MNs were linked to light-level types or muscles.

    1328 DNs versus earlier estimates of 700 to 1100; 733 MNs (510 in the nerves FANC traced, where 507 were reported). 29% of DNs matched to published types (45.9% with companion split-Gal4 work); 85% placed in tracts, including new MTD-III and CVL. 170 of 319 MN groups identified: all but one of 26 wing MNs, 142 of 142 T1 leg MNs and, putatively, 198 of 252 T2/T3 leg MNs. · 2.1 Connectome generation and cell typing; 2.2 Identification of descending neurons; 2.3 Identification of motor neurons; 3 Discussion

  4. SuggestedWing premotor circuits separate power from steering. Power MNs share diffuse, bilateral excitatory input from a population of upper tectulum interneurons, while steering MNs get lateralized DN drive, partly through wing contralateral haltere interneurons (w-cHINs) that appear electrically coupled.

    About 21 pairs of tectular INs (Tect INs) together supply 7-23% of each power MN's input, at a mean of 22 synapses per Tect IN per power MN. Five w-cHIN groups (14 cell pairs) receive substantial DN input; membrane contact areas suggest putative electrical links to contralateral b1, b2 or iii3 steering MNs. Roughly 30% of wing/haltere-tectulum DNs give at least 1% direct input to a wing or haltere MN. · 2.5.1 Descending neurons and networks underlying…; 2.5.2 Laterized descending control of steering…; 2.5.3 Wing power output is likely driven by…

  5. SuggestedEach leg is controlled mainly by a serially repeated local premotor circuit dominated by inhibitory inputs. The smaller inter-leg network is mostly cholinergic, unlike earlier models with inhibitory coupling between per-leg pattern generators, and most leg DNs target the local circuits.

    Over 75% of leg MN presynaptic partners are INs restricted to one leg neuropil; 52% of these (about 1200 per neuropil) are serially matched, and Infomap recovered six similar ~1200-neuron leg communities. Direct inhibitory input comes from over 57% GABAergic and glutamatergic serial leg neurons. Strong left-right links exist for T1 and T3 leg INs but not T2. · 2.6.1 The leg premotor circuit is serially…; 2.6.2 Inter-leg coordination is not purely…; 2.6.3 Descending neurons provide input to both…; 3 Discussion

Methods, data and reproducibility

Every DN and MN in the MANC electron-microscopy connectome was proofread, typed by NBLAST and matched to light-level driver lines, FANC or serial homologs; networks were analyzed by input fractions, Bayesian graph traversal, Infomap communities, triad motifs and indirect connectivity. DN transmitter predictions were checked by fluorescent in situ hybridization (FISH, EASI-FISH).

  • Male Adult Nerve Cord (MANC) EM connectome v1.2.3, densely reconstructed and annotated · Drosophila melanogaster (male) · n = ~23,500 neurons (1328 DNs, 733 MNs) · neuPrint dataset manc:v1.2.3
  • Female Adult Nerve Cord (FANC) EM volume; T1 leg MNs transformed to MANC space for NBLAST matching · Drosophila melanogaster (female)
  • Hemibrain connectome regions (mushroom body, central complex, ventrolateral neuropils) used as recurrence comparison · Drosophila melanogaster
  • Split-Gal4 lines for DN types probed by FISH or EASI-FISH for neurotransmitter markers · Drosophila melanogaster · n = 53 DN types (15 also probed for minor transmitters)
  • Light-microscopy images of Gal4/split-Gal4 lines (Janelia collection via NeuronBridge, Virtual Fly Brain, literature) for DN and MN matching · Drosophila melanogaster
  • VNC longitudinal tract meshes from Virtual Fly Brain (JRC2018VNC_U space) · Drosophila melanogaster · VFB_00104642, VFB_00104635, VFB_00104636, VFB_00104637, VFB_00104634, VFB_00104638, VFB_00104633
Methods & tools
neuPrint, natverse, neuprintr, malevnc, NBLAST, navis, Infomap, Cytoscape 3.9.1 (RCy3, igraph), Blender, NeuronBridge, EASI-FISH, alphashape3d
Code
https://natverse.github.io/malevnchttps://natverse.org/neuprintrhttps://zenodo.org/records/8191725
Data & accessions
https://neuprint.janelia.org/https://neuprint.janelia.org/?dataset=manc%3Av1.2.3&qt=findneuronshttps://www.virtualflybrain.org/https://github.com/clio-janelia/clio_website
Software & environment
MANC v1.2.3 queried via neuPrint API; R natverse (neuprintr, malevnc); RCy3 2.17.1, igraph 1.3.0, Cytoscape 3.9.1; Blender 2.82 (tract meshes) and 3.4 (contact areas); navis Python library for graph traversal.
Reproducibility gaps
Analysis and figure code is available only on request from the corresponding authors; DN/MN annotations and light-level matches are provided as supplementary files and in neuPrint.
Limitations (6)
  • DN transmitter predictions have uncertain fidelity (DNs were absent from classifier training): FISH backed 23 of 28 cholinergic and 9 of 10 glutamatergic calls but only 3 of 9 GABAergic ones; some circuit readings also rest on sub-0.7 predictions and on glutamate being inhibitory. · stated by the authors
  • Electrical synapses cannot be resolved in this EM data, so w-cHIN couplings and other putative electrical neurons rest on membrane-contact measurements and await experimental confirmation. · stated by the authors
  • Dissection-related axonal damage caused degeneration in roughly 49% of leg MNs (and 5% of the remaining MNs), leaving their reconstructed morphology and connectivity sparser; neck and abdominal MNs remain largely unidentified. · stated by the authors
  • Infomap is one of many community detection algorithms and the functional meaning of its communities is unverified; analyses also omit non-synaptic and neuromodulatory release and treat synapse counts as synapse weights. Functional data are lacking for most VNC cells. · stated by the authors
  • All connectivity comes from one male nerve cord volume; apart from MN counts and T1 leg MN matching against FANC, consistency across animals or sexes is not tested. · noted in this summary
  • Many cutoffs (1% groupwise input, 10% output with 100 synapses, 20% traversal cap, 40-synapse leg threshold) shape the reported pathways, but no sensitivity analysis of these choices is reported. · noted in this summary

Who should read:Read this if you study how descending commands drive walking, flight steering or escape takeoff in Drosophila, or plan to mine MANC using curated DN and MN identities.

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
THISTakemura '23Namiki '18Azevedo '24Stürner '24Ceballos '26Marin '23
86% semantic matchNeuroscience
Shin-ya Takemura, Kenneth J. Hayworth, Gary B. Huang+81 more·bioRxiv·Jun 5, 2023· 74 citations