Overview of MANC connectome.

A. Schematic of the Drosophila melanogaster central nervous system, showing brain and ventral nerve cord (VNC) as well as neuron classes which constitute VNC motor circuits: descending neurons (DNs), intrinsic neurons (INs), motor neurons (MNs), and ascending neurons (ANs). Numbers in parentheses indicate the total number of that neuron class found in MANC. Approximate areas indicated by dashed lines correspond to the neuropils controlling several of the fly’s primary motor systems (legs, wings, halteres). The three thoracic segments are indicated with T1, T2 and T3, and the abdominal neuropil is abbreviated ANm. Pie chart shows the neuron class composition of postsynaptic partners of all DNs, the majority of which are intrinsic neurons. B-C. Cross section of the D. melanogaster neck connective at the location of the black dashed line in A. DN profiles are color coded by tract membership (B, see Figure 2) and predicted neurotransmitter (C, see Figure 4). Side bar charts show distribution of DNs per tract or predicted neurotransmitter. Predicted neurotransmitter type is ‘unknown’ if prediction probability was <0.7). D. DNs and MNs were identified by matching MANC EM reconstructed neurons to published light microscopic (LM) level descriptions of DNs (mainly Namiki et al., 2018, for more references see Supplementary file 1) and MNs either manually by eye or aided by neuronbridge queries of color depth maximum intensity projections (MIP) (Clements et al., 2022; Otsuna et al., 2018) of driver lines against the EM dataset. Left, DN identification example: A neuronbridge query for SS02384, which labels DNa10 (Namiki et al., 2018), revealed group 10506 to be a good match. 29.1% of MANC DNs matched a previous DN description in the literature. Center, MN identification example for the dorsal neuropils. Matching to SS02623 (which contains the tergotrochanteral muscle motor neuron, TTMn) revealed group 10068. Right, MN identification example for the leg neuropils. A recent study identified all leg MNs in the T1 compartment (FANC, Azevedo et al., 2020). T1 leg MNs in this dataset (MANC) were matched to FANC using NBLAST. Cross-matching of leg MNs across FANC and MANC aided in confirming LM identification of leg MNs in both volumes. Furthermore, serially homologous neurons of T1 leg MNs in T2 and T3 were identified in MANC (see Materials & Methods). The bar chart shows the status of LM identification of MNs. 75% of dorsal neuropil MNs are matched to previous descriptions. 35.9% of leg MNs are matched to FANC. By identifying serial homologous neurons of the T1 leg MNs, the muscle targets of an additional 50% of leg MNs were characterized. We did not match abdominal MNs to light level descriptions in this study. E. Left, glossary of terms and acronyms used in this study. Right, systematic naming of DNs and MNs. For more information on other neuron classes and terms, see accompanying paper (Marin et al., 2023).

Overview of Identified DNs.

Morphology of DN types described by (Namiki et al., 2018) identified in the EM volume. Nomenclature according to Namiki et al., 2018. Three DN types could not be identified in the EM dataset (indicated by an empty VNC outline and “not found”). See Supplementary file 1 for more information on identification confidence and other synonyms of DNs in the literature.

Tract-based analysis of descending neurons.

A. Tract analysis of all left side DNs (right side DNs are mirror symmetric). DNs run through different tracts in the VNC (Boerner and Duch, 2010; Court et al., 2020; Namiki et al., 2018). Tract names are based on previous publications: DLT, dorsal lateral tract of dorsal cervical fasciculus; MDA, median dorsal abdominal tract; MTD, median tract of dorsal cervical fasciculus; DMT, dorsal medial tract; ITD, intermediate tract of dorsal cervical fasciculus; VLT, ventral lateral tract; DLV, dorsal lateral tract of ventral cervical fasciculus; VTV, ventral median tract of ventral cervical fasciculus. We identified two previously unknown tracts: a third type of MTD and CVL, curved ventral lateral tract. DNs with a short primary neurite were not assigned to a tract in agreement with (Namiki et al., 2018). B. Tract meshes created from skeletons of DNs assigned to each tract respectively in dorsal, ventral and lateral view. Colors in B correspond to A. C. Cross section (frontal) view of all DN primary neurites color coded by their tract membership (DNs without tract membership were omitted for clarity). Position (1, 2, 3) of section in the VNC indicated by lines on the VNC ventral view. D. Number of DNs for each tract separated by cervical connective side. E. DNs were grouped based on morphology and connectivity. Groups can consist of one (pair) or more (population) DNs per side. DNs for which we could not identify a group were labeled ‘ungrouped’. Shown are the compositions of DN grouping status separated by tract membership. F. Correlation of soma location and tract membership for identified DN types (based on light microscopy data, Namiki et al., 2018).

Tract-based analysis of descending neurons.

Tract based analysis of right side descending neurons and tract meshes. A. Tract analysis of all DNs. DNs run through different tracts in the VNC (Boerner and Duch, 2010; Court et al., 2020; Namiki et al., 2018). Tract names are based on previous publications: DLT, dorsal lateral tract of dorsal cervical fasciculus; MDA, median dorsal abdominal tract; MTD, median tract of dorsal cervical fasciculus; DMT, dorsal medial tract; ITD, intermediate tract of dorsal cervical fasciculus; VLT, ventral lateral tract; DLV, dorsal lateral tract of ventral cervical fasciculus; VTV, ventral median tract of ventral cervical fasciculus. We identified two previously unknown tracts: a third type of MTD and CVL, curved ventral lateral tract. DNs with short primary neurite were not assigned to a tract in agreement with (Namiki et al., 2018). Shown are DNs of the right side in a ventral view (see Figure 2 for left side DNs).

DN subclasses.

Descending neurons were analyzed based on neuropil innervation. A. VNC neuropils. B. X-axis indicates target neuropil designations (the ROI neuropils that together receive more than 80% of the DNs output, see Materials & Methods) and bar color indicates DN grouping status (as in Figure 2). Target neuropil groups that were used to subsequently define DN subclasses are in black, all grey target neuropil groups received the subclass DNxn (see Figure 3–Supplement 1). C. The relationship between neuropil location of DN presynaptic sites and tract membership. Total synapse number normalized by the number of DNs in the tract is shown. D. DNs are assigned a subclass based on their target neuropil: DNnt, Neck Tectulum; DNwt, Wing Tectulum; DNht, Haltere Tectulum; DNut, upper tectulum, if they target a combination of neck, wing, and haltere neuropils; DNfl, DNml, DNhl, innervating single leg neuropils front leg, middle leg, and hind leg respectively; DNxl, if they target a combination of several leg neuropils; DNit, Intermediate Tectulum; DNlt, Lower Tectulum; DNad, Abdominal Neuromeres. DNs that innervate more are referred to as DNxn, for multiple neuropils. E. Heatmap representation of DN neuropil innervation, by percent of each DN’s synaptic output. Every row is a single DN output. Bar along the y-axis shows the subclass assigned to different DNs, grey are those in xn category (see Figure 3–Supplement 1). F. The number of DNs in a given subclass assigned to either left or right side. There is a single pair of DNs for the subclasses DNwt, DNml and DNhl (see images in C). G. DN postsynaptic partner composition by neuron class and interneuron subclass.

DN subclass.

A. DNs of subclass xn, which consisted of large groups (>15 single DNs) of DNs with common target distribution in the VNC. Number of single DNs within each xn subclass and names of combined target neuropils shown below each group. ‘Multi’ DNxns had 80% of the DN output in more than two neuropils. B. Heatmap representation of DN neuropil innervation by number of neurons. Every row is a single DN. Bar along the y-axis shows the neuropil targets (with >15 DNs) assigned to subclass DNxn, coloured only the ones shown in A. C. Number of DNs by tract innervating neuropils. Innervation in a neuropil is counted if the DN has more than 5% of its output to that neuropil. D. DN output to a neuron class in percent for every single DN, plotted on the x-axis. Red arrows point to two small groups of DNs, one with a high proportion of output to efferent neurons and the other with a high proportion of output to local sensory neurons.

Neuropil innervation for descending neuron types described in Namiki et al., 2018.

A. Neuropil innervation of DN types. Filled pixels indicate that at least 5% of a DN group’s presynaptic sites are located in a given ROI. X-axis order determined by (Namiki et al., 2018), Figure 7. B-C. Autocorrelation matrix of DN innervation pattern in the VNC. The Pearson’s correlation coefficient was calculated for each pair of ROIs between DN innervation of identified canonical Namiki types (B) and all types in the MANC EM volume (C).

Neurotransmitter prediction for DNs.

A-D. Morphology of example DNs with high probability neurotransmitter predictions for acetylcholine (A), GABA (B), glutamate (C) and very low probability neurotransmitter prediction (D). Four presynaptic sites are shown for each DN with yellow arrows pointing to the presynaptic density (T-bar) and synaptic cleft. D. DNp20 has few vesicles and may be electrically coupled (indicated by larger size diameter of the axon). E. Counts of pre- and postsynaptic sites of DNs and neurotransmitter predictions shown as shapes (acetylcholine, circle; GABA, triangle; glutamate, square). The prediction probability is color coded. Red circle indicates DNs shown in D, black rectangle indicate DNs shown in F and gray rectangle DNs shown in G. F. Morphology of two DN pairs with high synaptic input in the VNC. For both the predicted neurotransmitter is GABA. G. Morphology of three DN pairs with high synaptic output in the VNC of which the neurotransmitter of two was predicted GABA and one acetylcholine. The type of each DN is indicated in the bottom left. Black scale bars are 50 µm, red scale bars are 500 nm. Scale bars in EM synapse panels are the same if not indicated otherwise. H. Neurotransmitter identity for DNs from FISH technique, see Figure 4–Supplement 1 for images and DN identities. I. Example images from the FISH analysis, DNb01 (SS02383) and DNa01 (SS00731). Dotted yellow line indicates the DN soma location (see Material and Methods for details).

DN FISH for neurotransmitter markers.

To survey neurotransmitter usage in DNs, adult Drosophila brains expressing marker genes in individual DN types were probed using either EASI-FISH (Eddison, 2022; Wang et al., 2021) or a standardized FISH technique (Meissner et al., 2019). Example results for nine DNs are shown here. A-F. For EASI-FISH, brains from DN split-Gal4 lines expressing UAS-CsChrimson-mVenus were probed for the indicated combinations of two neurotransmitter markers, as well as immunostained for GFP. Labeled DN types and their split-Gal4 lines are as follows: A. DNa01 (SS00731), B. DNg34 (SS58292), C. systematic type DNxl046 (SS93763, synonym: oval), D. DNp27 (SS00923), E. DNb01 (SS02383), F. DNp20 (SS01078). G-I. For standard FISH, brains from DN split-Gal4 lines expressing UAS-7xHaloTag::CAAX were probed for ChAT, Gad1 and VGlut, and stained with fluorescent conjugated HaloTag ligand. Labeled DN types and their split-Gal4 lines are as follows: G. DNa08 (SS02393), H. DNb02 (SS01060), I. DNp24 (SS00732). Soma of DNs of interest are outlined in all FISH channels. Maximum intensity projections in all images are from a Z-range in each stack that closely flanks the soma of interest. J. Summary of EM neurotransmitter prediction and FISH-determined neurotransmitter usage for all DN types tested (y-axis). Multiple colors per row for FISH results indicates positive marker expression for multiple neurotransmitters. (*) indicates uncertain FISH results. Red DN labels indicate systematic types (and synonyms if known) matched against new split-Gal4 lines. Their type names will be revised in future work (Eichler et al., in prep).

Motor neurons of the Drosophila male VNC.

A. VNC nerves from which motor neurons (MNs) exit. Nerve abbreviations are as follows: cervical nerve (CvN), dorsal prothoracic nerve (DProN), ventral prothoracic nerve (VProN), prothoracic accessory nerve (ProAN), prothoracic leg nerve (ProLN), anterior dorsal mesothoracic nerve (ADMN), posterior dorsal mesothoracic nerve (PDMN), mesothoracic accessory nerve (MesoAN), mesothoracic leg nerve (MesoLN), dorsal metathoracic nerve (DMetaN), metathoracic leg nerve (MetaLN), first abdominal nerve (AbN1), second abdominal nerve (AbN2), third abdominal nerve (AbN3), fourth abdominal nerve (AbN4), and abdominal nerve trunk (AbNT). B. All MNs found in MANC, classified by exit nerve. Number in parentheses indicates cell count. C. MN synaptic input by neuropil. D. MN subclass classification by anatomical muscle category. E. MN count per subclass by exit nerve side: left (L), right (R) or not determined (ND). Exit nerve side was not determined only for a subset of MNs exiting near the midline through the AbNT.

Reconstruction state of motor neurons in MANC.

A. Motor neurons (MNs) per nerve and side, classified by whether they have less than 70% (0.7) of the predicted postsynaptic site counts compared to their groupwise maximum, i.e. the postsynaptic count of the most well-traced member of each MN’s morphological group in MANC. This classification flags MNs with likely reconstruction issues. Note that MN groups where all members of the group have reconstruction issues may not be flagged. B-D. Example MN groups with one well- and one less-well-reconstructed member: B. Sternal rotator anterior MN (group 13300), C. Ti extensor MN (group 10347), D. MNhm43 (group 17216). E. Seven examples (out of 14) of ungrouped MNs where their reconstruction state or that of their morphological match on the opposite side precluded group assignment.

Upper tectular motor neuron identification in MANC.

A. Thoracic organization of wing muscles shown from a lateral (top panel) or medial (bottom panel) view. Muscle categories and abbreviations are as follows: power muscles (dorsal longitudinal muscles, DLM; dorsal ventral muscles, DVM1-3), steering muscles (basalar, b1-3; first axillary, i1, i2; third axillary, iii1, iii3, iii4; fourth axillary, hg1-4), and indirect control muscles (tergopleural, tp; pleurosternal, ps1, ps2; tergotrochanteral, TTM). The hg2 muscle in the lateral view, and the power muscles and TTM in the medial view are omitted for clarity. B. Identification of wing muscle MNs from light-level data (further see Supplementary file 3). The iii4 MN match (MNwm35) is putative, and the muscle target of MNwm36 is unknown. C. Organization of haltere muscles. Muscle categories and abbreviations are as follows: power muscle (haltere dorsal ventral muscle, hDVM), and steering muscles (haltere basalar, hb1, hb2; first axillary, hi1, hi2; third axillary, hiii1-3). D. Identification of haltere muscle MNs from light-level data. Two candidates for the hi2 MN with similar morphology and connectivity are present in MANC. MNhm43 and MNhm42 innervate the hb1 and hb2 muscles, however the exact target of each MN is not yet known. While additional putative haltere MNs are not easily determined in EM as both haltere MNs and abdominal MNs exit the AbN1 nerve, three potential haltere MNs are subsetted by their arborization in the haltere neuropil and exit through AbN1, and two potential haltere MNs by exit through the haltere nerve, DMetaN).

Wing MN drivers and their muscle targets determined in this paper.

A. VNC expression pattern (green) of the split-Gal4 line SS98650 (VT058382-AD; VT042734-DBD) compared to its MN match in EM (B). C. b3 muscle innervation (green) in a longitudinal section of the thorax in SS98650. D. VNC expression pattern of SS38113 (Sterne et al., 2021) compared to its MN match (systematic type) in EM (E). Expression in the wing MN is stochastic in this line, and its MN morphology and muscle innervation was corroborated with another stochastic line, OL0070B (not shown; Wu et al., 2016). F. hg4 muscle innervation a longitudinal section of the thorax in SS38113. G. VNC expression pattern of SS98638 (VT026026-AD; VT064565-DBD) compared to its MN match in EM (H). I. ps2 muscle innervation in a longitudinal section of the thorax in SS98650. All VNCs are counterstained for mAb nc82 (magenta), while all muscles are counterstained with phalloidin (magenta).

Serial leg motor neurons.

A. Diagram of a prothoracic leg showing the position of muscle groups innervated by motor neurons (MNs), some of which attach across the thorax, and others of which are intrinsic to the coxa, trochanter, femur, or tibia leg segments. Adapted from Brierley et al., 2012. On the right the number of MNs that could be assigned a muscle target split into soma neuromere (T1, T2, T3). B. Example of two serial MN sets that innervate the long tendon muscles (LTM) and two serial MN sets that innervate the tibia extensor muscles, one for slow and one for fast tibia extension. C. All MNs that are in a serial set listed by their muscle target. In brackets the number of serial sets found / number of neurons in T1 pairs. Colors by the muscle they innervate as shown in the cartoon in A. D. For three of the muscle targets we were not able to find corresponding neurons in T2 and T3 leg neuropils. Thus only the T1 leg neuropils have muscle target assignments. E. Cosine similarity of all leg MNs by their connections to serial local neurons in T1, T2 and T3 leg neuropils. MNs are organized by muscle targets and leg compartment, shown in corresponding coloured bars. All MN targets in T1 were assigned by careful matching to the FANC dataset (Azevedo et al., 2022), while all MN targets in T2 and T3 were assigned by serial sets (see Materials & Methods).

Unidentified leg MNs in MANC.

A. Cosine similarity of all leg MN groups by their connections to serial local neurons in T1, T2 and T3 leg neuropils. MNs are organized by muscle targets and leg compartment, shown in corresponding coloured bars. All MN targets in T1 were assigned by careful matching to the FANC dataset (Azevedo et al., 2022, see Materials & Methods). B. Four serial MN sets that could not be assigned to the same muscle target across segments. The first two examples each include a T2 pair that has been matched to the literature type STTMm (marked with *). By connectivity and morphology they have a serial pair in T3 (and in T1 a MN assigned to the Tergotr. MN). C. All MNs by group in T2 that we were unable to assign a serial set. D. All MNs by group in T3 that we were unable to assign a serial set.

DN to MN connectivity across the VNC.

A. Pair vs. population classification for all DN and MN types, as determined by the number of cells per neuron morphological group. DN and MN subclasses are defined in Figures 3 and 5. The DN subclasses DNwt, DNht, DNml and DNhl, and the MN subclass MNxm are not shown as they consist of few cells (3 groups or less). B. Hemilineage composition of DN downstream and MN upstream partners, compared to hemilineage motor activation phenotype as summarized from Harris et al., 2015. Motor phenotypes are marked “+” if ≥50% of flies display it upon hemilineage activation in Harris et al., 2015. C. All VNC intrinsic premotor groups (intrinsic, ascending, efferent and ascending efferent neurons) by their DN input count and local output count (Count of groups at a ≥1% groupwise input threshold). Intrinsic neuron groups are categorized by their primary input neuropil (≥50% postsynapses). D. Neuron groups with high DN input and local output (≥90th percentile for both) as defined in C. E. Direct DN to MN connectivity, shown by input fraction per MN group contributed by DNs (left), and thresholded DN input count (≥1% groupwise input) per MN group (right). F. MNs with high DN input (≥20% total MN group input) from E. G. Bayesian graph traversal model for assessing distance from individual DN groups (source layer) to all VNC neurons (target layer). All neurons are assigned a layer based on their synaptic connectivity with neurons in prior layers, where the probability of traversal is scaled by the input fraction contributed by neurons of prior layers. Thus, layer assignments are a nonlinear measure of distance combining path length and synapse connectivity strength. H. MN groupwise layer means for all DN groups by Bayesian graph traversal runs starting with individual DN groups as the source layer. I. Example per-layer cell class composition of Bayesian graph traversal for DNa02 and DNa08. J. Example layer-to-layer connectivity of Bayesian graph traversal for DNa02 and DNa08, up to 10 layers. Layers without neurons are omitted.

Bayesian graph traversal from DNs to MNs.

Heatmap of all-to-all mean DN group to MN group layer distance for Bayesian graph traversal runs starting with individual DN groups as the source layer.

Structure of VNC networks.

A. Community structure of VNC intrinsic premotor networks (all neurons excluding descending, sensory and motor) as partitioned by the Infomap algorithm (Rosvall et al., 2009; Rosvall and Bergstrom, 2011, 2008; Smiljanić et al., 2021). Communities are numbered arbitrarily, and communities of at least 50 cells are shown on graph and colored by their constituent neurons’ primary input and output neuropils (see C). Edges between communities are thresholded at ≥25000 synapses. Dotted circles and their labels indicate putative broad motor function based on neuropil arborization and connectivity with MN subclasses (see C & D). B. Neuron count of Infomap communities. Communities on the x-axis are rank-ordered by decreasing neuron count. C. Infomap community-to-community connectivity. D. Community neuropil input and output, calculated by the % neuropil input and output of all neurons in the community. E. Community DN input and MN output by DN/MN subclass. Connectivity in graphs are thresholded at ≥5000 synapses. Light shaded area in graphs correspond to Infomap communities as shown in A. The DN subclasses DNwt, DNml and DNhl consist of only a single DN group each and are not shown. F. Heatmaps of community DN input and MN output. G. Recurrence in VNC networks as evaluated by triad census, compared to brain neuropils with high recurrence (mushroom body, MB, and central complex, CX) and low recurrence (right ventrolateral neuropils, VLNP(R)). Triad motifs are classified by their isomorphism class into feedforward, feedback and recurrent categories. Synaptic connectivity between neurons are thresholded at ≥0.1% input. Neurons per neuropil are subsetted by having ≥50% input and output in the target neuropil, and further excluding DNs, sensory and MNs. H. Recurrence in neuropil networks as evaluated by all-to-all neuron shortest cycle length at a synapse threshold of ≥0.1% input. Neurons per neuropil are subsetted by having ≥50% input and output in the target neuropil, and further excluding DNs, sensory and MNs.

VNC Infomap communities and within-community recurrence.

A. Community neuropil input and output for all Infomap communities (excluding communities with 1 neuron, or that have zero pre- or postsynaptic sites), calculated by the % neuropil input and output of all neurons in the community. B. All Infomap community-to-community connectivity. C. Recurrence in communities with ≥50 cells as evaluated by triad census. Synaptic connectivity between neurons are thresholded at ≥0.1% input. D. Recurrence in communities with ≥50 cells as evaluated by all-to-all neuron shortest cycle length at a synapse threshold of ≥0.1% input.

Dipteran steering muscle function

. Activity of steering muscles in Drosophila and other dipterans during turning in flight on the ipsilateral (inner side of turn) vs contralateral (outer side of turn) sides. Evidence compiled from Egelhaaf, 1989; Heide, 1975; Heide and Götz, 1996; Lehmann and Götz, 1996; Lindsay et al., 2017; Nachtigall and Wilson, 1967. The iii2 muscle, which is not present in Drosophila, is omitted.

Overview of wing DNs and wing networks.

A. DN groups with ≥10% output (groupwise mean) and ≥100 presynaptic sites (groupwise sum) in the wing tectulum (WTct) and haltere tectulum (HTct). DNs are categorized and colored by their subclass as defined in Figure 3. B. WTct/HTct DNs of interest that may be implicated in flight motor control. C. Neuropil-wise and cell class-wise efferent connectivity of WTct/HTct DNs (left) and their direct downstream targets that receive ≥1% groupwise input from WTct/HTct DNs (right). In this and subsequent heatmaps, DNs of interest from A are shown with arrows (DN pairs) or arrowheads (DN populations) with the same color as DN labels in A. D. DN input count (number of WTct/HTct DN groups providing ≥1% groupwise input) of WTct/HTct DN downstream intrinsic premotor neuron groups (IN/AN/EN/EA), classified by top input/output neuropil of their VNC Infomap communities (see Figure 9). Groups in the “multi com.” category have individual neurons assigned to different communities. E. Morphologies of top WTct/HTct DN targets from C that have thresholded input from 10 or more DN groups. F. Community structure of wing-related Infomap communities and their connectivity with WTct DNs and wing/haltere MNs. Subcommunities were defined by a second round of Infomap community detection on cells of each community. DN grouping is defined by hierarchical clustering of individual WTct/HTct DN groups’ output to subcommunities (right heatmap). Edges on graph are thresholded at ≥2000 synapses. G. Direct connectivity from WTct/HTct DNs to wing and haltere MNs. DN groups shown have ≥1% input to at least 1 MN group. DNs and MNs are named by type, with group appended if multiple groups share a type. H. Indirect WTct/HTct DN-MN connectivity by normalized indirect connectivity strength (see Materials & Methods). DN groups shown have ≥5 normalized indirect connectivity strength values to a MN group. Normalized indirect connectivity strength is then rescaled to a maximum of 1 row-wise (DN groupwise).

Connectivity of wing network Infomap subcommunities to MNs.

A. Synaptic input fractions contributed by individual subcommunities to wing and haltere MNs of a group. B. Putative courtship-related communities as indicated by LM-EM matching to fruitless-positive neuron types (Lillvis et al., 2023).

Putative flight steering circuits through DNa04/05 and w-cHINs.

A. Modulation of fly wingbeat amplitude (WBA) and steering muscle (SM) activity during turning in flight. SM activity is simplified from Lindsay et al., 2017 (also see Table 1). B. Pathways of DNa04, DNa05 and other DNs with similar VNC morphology to steering MNs and haltere MNs. Node labels are neuron type, appended with a five- or 6-digit number for neuron group if a subset of neurons of a type are shown, while number in parentheses indicates cell count per node. Synapse weight (neuron-to-neuron mean) is thresholded at ≥20, except for connectivity of wing contralateral haltere interneurons (w-cHINs) to MNs which may have electrical synapses (see C). C. Mean surface contact area of single w-cHINs of a group to all MNs of a group, suggestive of electrical synaptic connectivity. Contact area is calculated using 3D neuron meshes derived from EM segmentation; for each MN/w-cHIN pair, meshes are expanded by 0.1 μm, and the intersection surface area is calculated. D. Examples of w-cHIN contact with MNs, with intersection area (yellow) shown through the neuron meshes. E. Lateralized pathway from DNa04 to steering MNs through w-cHINs and associated cells as a candidate steering microcircuit, and predicted effect on steering MN lateralized control. For clarity, the right side DNa04, w-cHIN group 10147, IN18B041, IN06A002 and IN06A044 group 19157, are omitted. While neurotransmitter prediction for IN18B041, N06A002 and IN19A026 group 10764 are below probability threshold (<0.7; round edge arrowheads), their hemilineages suggest that IN06A002 (6A) and IN19A026 group 10764 (19A) are GABAergic, while IN18B041 (18B) is cholinergic (Lacin et al., 2019). Side annotation is neck side for DNs, output nerve side for MNs, and soma side for others. Synapse weight (neuron-to-neuron mean) is thresholded at ≥20, except for connectivity from w-cHINs to MNs. F. Convergence of DN and sensory input onto w-cHINs. All DNs contributing ≥1% groupwise input to a w-cHIN group are shown as a single node, while sensory neurons grouped by anatomical origin contributing ≥1% input to w-cHIN groups are shown. G. Morphology of w-cHINs (top, colored by group), pathway of DNa04 to contralateral b1 and b3 MNs through w-cHIN group 10073, IN18B041 and IN19A026 group 10764 (middle), and convergence of SNs and DNs onto w-cHINs (bottom). H. Direct lateralized DN input to steering MNs, haltere MNs (from A) and w-cHINs, for DNs providing groupwise input of ≥1% to any. Ipsilateral-contralateral index is calculated by ipsilateral minus contralateral connectivity, normalized by their sum. w-cHIN side is considered to be their output side, contralateral to dendrites. DN labels colored red are not within the WTct/HTct DN set defined in Figure 10.

Putative electrical interneurons in the Upper Tectulum.

A. Top 50 intrinsic neuron groups in MANC rank-ordered by low presynaptic site count per neuron volume (groupwise mean), which is suggestive of non-chemical synaptic transmission such as electrical synaptic connectivity. The mean of all VNC intrinsic neurons is also shown. Within the top 50, groups with dense core vesicles (which suggests extrasynaptic neurotransmitter release) or significant reconstruction issues were manually evaluated and excluded (see Supplementary file 5). Neurons are named by type, with their group appended if multiple groups share a type. B. Morphology of top 10 rank-ordered putative electrical interneuron groups besides the w-cHINs shown in Figure 11. C. Direct DN input count (at ≥1% groupwise input), and neuropil input and output of the top 50 putative electrical interneuron groups. D. Mean surface contact area between putative electrical interneurons and all wing and haltere MNs of a group. Contact area was determined by expanding each interneuron and MN by 0.1 μm and calculating the surface area of intersection meshes between every interneuron-MN pair. E. Example interneuron-MN pairs with high contact area as determined in D. Intersection area is shown in yellow on top of the neurons. All views shown are ventral.

Additional connectivity of DNs controlling steering MNs.

A-B. Lateralized indirect connectivity of DNs to steering MNs, haltere MNs and w-cHINs, for path length of 2 (B) and 3 (C). Un-normalized indirect connectivity strength is used, as a comparison between path lengths is not carried out. Ipsilateral-contralateral index is calculated by ipsilateral minus contralateral connectivity strengths, divided by their sum. For this calculation, w-cHIN side is considered to be their output side, contralateral to dendrites. Only DNs above an arbitrary cutoff value for indirect connectivity strength with MNs/w-cHINs are shown. C. Highly bilateral pathway of DNa10 to steering MNs. Synapse weight (group sum) is thresholded at ≥100. D. Lateralized path of DNp03 to steering MNs through w-cHINs and other cells. For clarity, only the left DNp03 and INs (by dendrites side) are shown, except for IN12A059 group 19537, a bilateral population cell type. Synapse weight (group sum) is thresholded at ≥20. Side annotations in node names are neck input side for DNs, output nerve side for MNs, and soma side for others.

Sensory input to w-cHINs and their target MNs.

Adjacency matrix of synapse weights of individual sensory neurons to individual w-cHINs or w-cHIN target MNs. Only sensory neurons which contribute 2 or more synapses to individual w-cHINs or MNs are shown. Side annotations are input nerve side for sensory neurons, output nerve side for MNs, and soma side for w-cHINs (same side as their dendrites).

Putative power muscle control circuit in flight.

A. Power generation by the power muscles (PM) in flight drives wingbeats. Drosophila wingbeats are over an order of magnitude faster than firing of power MNs. The power MNs within the same motor unit fire in a loose, staggered pattern, likely to smooth out PM calcium changes and wing power output. Wingbeat and power MN traces artistically rendered from prior descriptions (Harcombe and Wyman, 1978, 1977; Tanouye and Wyman, 1981; Wyman, 1966). B. Upstream connectivity of the power MNs (dorsal lateral, DLMns and dorsal ventral, DVMns). The bulk of upstream connectivity of the power MNs is contributed by 42 tectular intrinsic neurons (Tect INs), with many DNs (here subsetted by ≥1% groupwise connectivity with Tect IN groups) and other neuron types with connectivity to Tect INs. Upstream connectivity to the power MNs are also associated with inputs to the ps1 and MNwm36 wing MNs, and the mesVUM-MJ which targets the dorsal lateral muscles. Edges are thresholded at ≥100 synapses. The Tect INs comprise neuron groups 15113, 13060, 14502, 18519, 21307, 15337, 22521, 21381, 22289, 15788, 14625. The Tect IN-like neurons comprise groups 13874, 18143, 22530, 18638. C. Morphology of the Tect IN population, shown together with a single DLMn. D. Connectivity of the Tect INs (individual neurons on x-axis) with upstream DNs and other major inputs, and downstream MNs/efferents (combined by group). Laterality index is the mean of the difference between side-separated connectivity, normalized by per-side cell count and further normalized by sum of side-separated connectivity. The Tect INs show variation in the combinations and strength of their upstream and downstream connectivity, which is partially reflected in their morphological variation (see E). Neuron groups shown on y-axis are a selection from the top input or output neuron types/groups connected to Tect INs, and named by type, with group (5- or 6-digit number) appended if more than 1 group share a type. Dendrogram shows hierarchical clustering of Tect INs by the subset of upstream and downstream connectivity shown here. E. Morphological variation in Tect INs. F. Lateralized connectivity of DNa08 and DNp31 with Tect INs, IN06B066 group 16779 and power MNs, forming a motif where the same DNs appear to excite and inhibit power MNs. Side annotation is neck side for DNs, nerve side for MNs, and soma side for others. G. Connectivity of three example power MN-associated DNs (DNp31, DNg02 and DNp03) with indirect control MNs. H. Schematic of putative power MN control circuit. In flight, DN input controls Tect IN activity, which then excites power MNs through ‘diffuse’ connectivity (individually small synaptic connections, but collectively strong at the population level). DNa08 and DNp31 further excites IN06B066 group 16779, which may form an inhibition-stabilized with Tect INs to limit runaway excitation. Power MNs sum ‘diffuse’ excitation from Tect INs to individually reach their spiking threshold, and weak electrical connectivity between power MNs further enforce an inhibition-like connectivity between them, such that power MNs spike in a slow, staggered manner to smooth out power muscle calcium changes.

Additional connectivity from DNs to power motor neurons.

A. Tect IN connectivity within the population. Tect INs provide weak recurrent input to each other, with a subset that contributes more input to others. Order of neurons and dendrogram on both axes is the same as for x-axis in Figure 12D. B. Pathways of DNa10 connectivity to power and indirect control MNs. IN00A057 group 16145 and IN12A059 group 19537 also have connectivity to steering MNs, see Figure 11–Supplement 1. Tect IN connectivity to power MNs not shown for clarity. C. Pathways of the courtship DNs pIP10 and pMP2 to power MNs. IN06B066 group 16779 has connectivity with Tect INs (see Figure 12F). Identification of the TN1a, dPR1, vPR9 and vMS11 cell types in MANC by Lillvis et al., 2023. D. Lateralized wing and proximal wing campaniform sensilla sensory neuron connectivity to Tect INs and IN06B066 group 22716. Node labels for all graphs are neuron type, appended with five- or 6-digit numbers for neuron groups (if a subset of groups within a type are shown), while number in parentheses indicates cell count per node. Labels for sensory neurons are derived from their synonyms instead. Synapse weights of all graphs are thresholded at ≥100.

Standard Leg Connectome.

Leg premotor circuit local to all three leg neuropils. A. Input to leg MNs in percent. Colour indicates the class of the input neurons, and the intrinsic neurons are additionally split up along the x-axis by their subclass (prefix). B. Over 75% of input onto leg MNs comes from intrinsic neurons restricted to the leg neuropils. Schematic illustrates the distinction between the three types of restricted neurons: IR, CR, BR. Next to it an example of a IR serial set (10652, type IN08A006) colored by leg neuropil. On the right the number of neurons in the VNC with those prefixes. Color indicates if the neuron is sorted into a serial set. C. Premotor circuit of all Leg MNs that are in serial sets. All connections between serial leg restricted neurons and MNs are shown (weight > 40), collapsed by serial set (see Figure 13—Supplement 1 for differences between leg circuits). Color of intrinsic neurons indicates the neurotransmitter prediction. Squares are single serial sets, hexagons are grouped serial sets (See Supplementary file 6 for more information). D. Examples taken from C to show a GABA, Glutamate and GABA/Cholinergic mechanism of controlling the leg MNs. E. Interconnection between the MNs in the Thorax with the Fe reductor. F. Major serial sets controlling the trochanter MNs. MN nodes in D, E and F are colored by the color of the leg muscle they control in the inset leg schematics.

Differences in the Standard Leg Connectome across segments.

Difference between the mean connection of the serial set and T1, T2 or T3 specific connections in A, B, C accordingly. Weight difference is color-coded, with red indicating stronger connections and blue meaning weaker connections compared to the mean connection of the serial set. Intrinsic serial sets are in light blue, motor neuron serial sets by muscle target in purple. Squares are single serial sets, hexagons are grouped sets.

Coordination of leg premotor circuits.

Grouped intrinsic neurons upstream of the standard leg premotor circuit with effective connectivity (see Materials & Methods). A. Serially repeating intrinsic neurons that ascend or descend a segment with their axonic projections (sequential). Grouped by hemilineage or neurotransmitter, for ungrouped version see Figure 14– Supplement 1. B. Ipsilaterally connecting intrinsic neurons, targeting several leg premotor circuits (all_L or all_R). Often projecting from other neuropils, see example IN21B002. C. Intrinsic neurons connecting T1 leg or T3 legs bilaterally (bil_T1 and bil_T3), example shown below. D. Intrinsic neurons projecting from T1 or T2 on one side to several leg premotor circuits on the other side (proj_T1 or proj_T2). E. Intrinsic neurons grouped by neurotransmitter that innervate all leg neuropils on both sides (bil_all). For details and examples, see Supplementary file 1. F. Summary of the groups and their connection to the standard leg premotor circuits of the six legs. Numbers indicate the number of DN types that connect to these groups with weight >40. See Supplementary file 7 for neuron types and details.

Additional upstream connectivity to leg circuit.

A. Sequential neurons (see Figure 14A) split by type and hemisegment. B. Neurons that project to a single leg neuropil, only the left hand side shown. Most are neurons restricted to the same leg neuropil and the standard leg group (grouped into one node, restricted). A few neurons project from another neuropil to a single leg neuropil. C. Neurons that innervate all leg neuropils (see Figure 14E) split by type. D. Example images from ventral and side view for neuron types with red stroke around the node in A, B or C. The first example shows a neuron projecting from the ovoid to the T3 leg neuropil, involved in wing grooming (Zhang and Simpson, 2022).

Descending neurons innervating the leg neuropils.

DNs that innervate the different leg neuropils with at least 10% of their presynaptic budget and 100 presynaptic sites (group total for both). Their morphology is shown in a ventral and lateral view color coded by their subclass. The number of groups fulfilling the criteria is indicated in brackets and pie charts on the right show subclass (legend at the bottom) composition of the DNs for each innervation type. A. DNs innervating LegNpT1 indicated by red area mesh in VNC cartoon on the left. B. DNs innervating LegNpT2 indicated by orange area mesh in VNC cartoon on the left. C. DNs innervating LegNpT3 indicated by orange area mesh in VNC cartoon on the left. For A, B and C, DNs that also dedicated more than 10% of their output budget to the respective other LegNps are excluded. D. DNs that innervate all three thoracic leg neuropils fulfilling the above criteria for each LegNp.

Effective connectivity of DNfl subclass neurons.

A. The effective connectivity of DN types to the leg MN muscle targets ipsilateral and contralateral to the root side of the DN. The MNs are separated by the three segments T1-T3. The shade of magenta encodes for the layer in which the DN targets the MN (1-5), while the size of the square reflects the connectivity score, which reflects the strength of the connection (see Materials & Methods for details). B. Examples indicated with an arrow in A. The top row shows the morphology of the DN. The pink arrow indicates the root side of the DN. Underneath are schematics of the 6 leg muscles with the normalized connectivity score to the different muscles highlighted in shades of grey. The bottom row shows the percent direct output to neuron classes and subclasses in the form of prefixes (Marin et al., 2023).

Effective connectivity of DNxl subclass neurons.

A. The effective connectivity of DN types to the leg MN muscle targets ipsilateral and contralateral to the root side of the DN. The MNs are separated by the three segments T1-T3. The shade of magenta encodes for the layer in which the DN targets the MN (1-5), while the size of the square reflects the connectivity score, which reflects the strength of the connection (see Materials & Methods for details).

Examples of DNxl subclass neurons.

Examples indicated with an arrow in Figure 15—Supplement 2. A. Morphology of the DNxl neurons. The pink arrow indicates the root side of the DN. B. Schematics of the six leg muscles with the normalized connectivity score to the different muscles highlighted in shades of grey. C. The percent direct output to neuron classes and subclasses in the form of prefixes (Marin et al., 2023).

Effective connectivity of DNxn subclass neurons.

A. The effective connectivity of DN types to the leg MN muscle targets ipsilateral and contralateral to the root side of the DN. Only those DNxn neurons were chosen that had the highest effective connectivity scores to leg MNs. The MNs are separated by the three segments T1-T3. The shade of magenta encodes for the layer in which the DN targets the MN (1-5), while the size of the square reflects the connectivity score, which reflects the strength of the connection (see Materials & Methods for details). B. Examples indicated with an arrow in A. The top row shows the morphology of the DN. The pink arrow indicates the root side of the DN. Underneath are schematics of the 6 leg muscles with the normalized connectivity score to the different muscles highlighted in shades of grey. C. DNxn neurons innervate several neuropils. The effective connectivity to other MNs targets are shown, separated by ipsilateral and contralateral.

Connectivity of DNa02 and DNg13.

A. Light microscopy images of DNa02 and DNg13 from Namiki et al., 2018 with their MANC axonic match. B. Connection of DNa02 and DNg13 in MANC into the leg coordination circuit groups from Figure 14. C. Morphology of selected neurons downstream of DNg13 or DNa02. D. Leg Muscles targeted by the two DNs, coloured by putative inhibition, activation or disinhibition. E. Illustration showing the connectivity across segments and sides for selected downstream circuitry with the predicted putative inhibition, disinhibition and activation of MNs marked on the ipsilateral and contralateral side in respect to the brain morphology and soma location. Note that the connections of only one DN (left side dendrite and soma DN) of the pair is shown, the right side DN has the same mirrored connectivity, but is omitted for clarity. Furthermore, we omitted circuitry of these DNs to MNs other than the swing/stance MNs for clarity. Neurons that are common downstream partners of both DNa02 and DNg13 are marked by a black stroke around the node.

Effective connectivity of DNa02 and DNg13.

Connectivity to ipsi- and contralateral leg MNs for the three neuromeres separately. A. DNa02 and B. DNg13 connectivity. Connectivity score represents the strongest connection in the network (either direct or via interneurons, four layers at maximum) based on the normalized matrix product of percent of input to the receiving neuron (top row, see Methods). The layer in which the highest connectivity score was present is shown in the bottom row plots. First layer represents direct DN to MN connectivity, second layer is DN-> IN -> MN, third layer includes two interneurons and so on. Leg muscle cartoons show a normalized score averaged across all three legs for each muscle. Darker color indicates stronger effective connectivity from the DN to MNs controlling the muscle.

Connectivity of DNa01, MDN and DNxl023, DNxl024.

A. Axon morphology of identified DNa01, MDN and two unknown DNs, DNxl023 and DNxl024 that target the three leg neuropils. B. Connection of DNa01, MDN and DNxl023/DNxl024 in MANC into the leg coordination circuit groups from Figure 14. C. Morphology of selected MANC neurons downstream, including LBL40 (Feng et al., 2020; Rayshubskiy et al., 2020). D. Predicted modulation of the leg muscles across the 6 legs, coloured by putative inhibition, activation or disinhibition. E. Illustration showing the connectivity across segments and sides for selected downstream circuitry of the DNs. Predicted putative inhibition, disinhibition and activation of MNs marked on the ipsi and contralateral side in respect to the brain morphology and soma location of only one DN (left side dendrite and soma DN for DNa01 and MDN, left root side for DNxl023/DNxl024). For clarity and comparison the circuitry concentrates on connections from the neurons shown in C. Neurons that are common downstream partners of all 4 DN types are marked by a black stroke around the node.

Effective connectivity of DNa01 and MDN.

Connectivity to ipsi- and contralateral leg MNs for the three neuromeres separately. A. DNa01 and B. MDN connectivity. Connectivity score represents the strongest connection in the network (either direct or via interneurons, four layers at maximum) based on the normalized matrix product of percent of input to the receiving neuron (top row, see Methods). The layer in which the highest connectivity score was present is shown in the bottom row plots. First layer represents direct DN to MN connectivity, second layer is DN-> IN -> MN, third layer includes two interneurons and so on. Leg muscle cartoons show a normalized score averaged across all three legs for each muscle. Darker color indicates stronger effective connectivity from the DN to MNs controlling the muscle.

Effective connectivity of DNxl023 and DNxl024.

Connectivity to ipsi- and contralateral leg MNs for the three neuromeres separately. A. DNxl023 and DNxl024 joined connectivity. Connectivity score represents the strongest connection in the network (either direct or via interneurons, four at maximum) based on the normalized matrix product of percent of input to the receiving neuron (top row, see Methods). The layer in which the highest connectivity score was present is shown in the bottom row plots. First layer represents direct DN to MN connectivity, second layer is DN-> IN -> MN, third layer includes two interneurons and so on. Leg muscle cartoons show a normalized score averaged across all three legs for each muscle. Darker color indicates stronger effective connectivity from the DN to MNs controlling the muscle.

Overview of lower tectulum circuits.

A. Left: morphology of the LTct-DNs (DN groups with ≥100 synapses and ≥10% of their synaptic output in the lower tectulum). Color corresponds to DN subclasses (e.g. xl, xn, lt) as in Figure 3. Right: morphologies of LTct-DNs with known behavioral functions. B. Left: Efferent connectivity of LTct-DNs. Left sidebar, subclass of LTct-DNs, with colors as in A (left panel). Colored arrows on the left of the sidebar indicate rows that correspond to examples in A. Left heatmap, proportion of each LTct-DN’s synapse-wise output in each region of the VNC. Right heatmap, proportion of LTct-DN’s synapse-wise output, onto each downstream cell class (counting only neurons receiving ≥1% of their synaptic input, minimum 50 synapses, from an LTct-DN). Right: as on the left side, but each row corresponds to one AN, IN, or EN downstream of an LTct-DN (“LTct-DN downstream neurons”, same connectivity criteria as above). C. Heatmap depicting all direct connections between LTct-DNs and MNs (groupwise). Any DNs that do not provide ≥1% synaptic input to a MN and any MNs that do not receive ≥1% synaptic input from an LTct-DN were excluded. Groups on the x-axis are named according to MN type; if types do not refer to a unique group then group number is appended. D. Heatmap depicting normalized indirect synaptic connectivity (as in Figure 6G) from each LTct-DN onto all MNs within five downstream layers. Any DNs that do not have an indirect connectivity score ≥5 onto any MN and any MNs that do not receive indirect connectivity ≥5 from any LTct-DN were excluded. For visualization, all heatmap values are re-normalized based on row-wise maxima. Colored arrows indicate rows corresponding to examples in A. E. Diagram depicting connectivity between neurons directly downstream of LTct-DNs (same subset as in right panels of B), binned according to the Infomap community they were assigned to (see Figure 9). Minimum connectivity to display an edge is 5000 synapses, and communities with fewer than 60 LTct-DN downstream members are not displayed. F. Communities 7 and 14 receive high levels of LTct-DN input. Number of LTct-DN input partners (neurons providing ≥1% of synaptic input) per downstream neuron, segregated by community assignment on the y-axis. G. Examples of neurons from communities 7 and 14 that receive high levels of LTct-DN input (≥1% of input from ≥5 LTct-DNs) and project widely throughout the VNC. Regions in diagrams are colored according to the amount of output they receive from the depicted neuron (normalized to the region with the highest output). Text indicates the community membership of each neuron, the number of inputs from LTct-DNs to each example neuron, and the number of outputs from that neuron to all neurons in the VNC (≥1% threshold).

Upstream and downstream circuits of the Giant Fiber in the VNC.

A. Schematic of differences between short-mode and long-mode takeoffs, including the characteristic patterns of motor activity, the durations of the behaviors, and the neurons known or hypothesized to be involved in each. B. Diagram depicting connectivity of neuron types that provide at least 1% of the Giant Fiber’s VNC-intrinsic input, as well as a selection of neurons that have strong input to those cells. C. Heatmap showing synaptic connectivity between the 17 groups that provide at least 1% of the Giant Fiber’s VNC-intrinsic synaptic input as well as IN05B072, a group of important second-order inputs. Connectivity is expressed as the fraction of the postsynaptic cell’s total synaptic input (groupwise). D. Morphology of GF-targeting ANs and INs. Top row shows morphology from a ventral perspective and bottom row shows morphology as viewed from the left side of VNC. The axons of the GF are shown in indigo in each image. Symbols next to neuron names indicate putative neurotransmitter identity: circle corresponds to a cholinergic prediction, triangle to a GABAergic prediction, and square to a Glutamatergic prediction. Grey symbols indicate that neurotransmitter prediction is low confidence (prediction probability <0.7). Left, morphology of the axon-targeting group AN08B069. Middle, six types of terminal-targeting INs divided into three sets based on morphological criteria. Right, morphology of IN05B072. E. Major groupwise synaptic outputs of the GF. Synaptic output is plotted as the % of the downstream group’s synaptic input that is provided by the GF. Only neurons that provide at least 1% of input are shown (min. 30 synapses). Neurons with known electrical connectivity to the GF are shown in dark boxes. F. Left, network diagram depicting direct and indirect connectivity between GF and two downstream partners that trigger takeoff behaviors: the TTMns, which trigger jumping, and DLMns, which initiate wing tuck or downstroke, via different subsets of GFC2 neurons. Right, morphologies of the GFC2_direct and GFC2_indirect groups. The GFC2_direct set contains the following groups: 13127 and 13645. The GFC2_indirect contains the following groups: 14527, 13479, and 15505. G. Left, network diagram depicting connectivity between GF and downstream partners that putatively control post-takeoff positioning and leg flexion. MNs are included in this plot so long as they receive at least 50 synapses (groupwise) from GFC1, GFC3, and GFC34. All three INs contact tibia flexor MNs and trochanter flexor MNs, although they vary in their other targets and in which neuromeres they project to. Right, morphologies of the GFC1, GFC3, and GFC4 groups.

Circuits mediating escape takeoff.

A. Connectivity diagram depicting strong descending connectivity onto MNs critical for long-mode takeoffs. For clarity, several groups are combined into larger categories based on shared connectivity: the “GFC2_indirect” node contains groups 14527, 13479, and 15505, the “GFC2_direct” node contains groups 13127 and 13645, the “Uni. DVMn input INs” node contains groups 17159 and 16088, and the Tect INs node contains the Wing Upper Tectular neurons described in Figure 12. MNs are grouped according to their target muscle. All other nodes contain bodyids with a shared type. B. Connectivity between SNs and escape-relevant INs and MNs. Left, heatmap depicting connectivity from SNs onto neurons directly presynaptic to TTMn. Connectivity is determined by a 1% input and 25 synapse threshold (groupwise). Rows are grouped by synonyms, with related synonyms combined based on the body parts in which they originate (e.g. wing CS SNs and wing margin bristle SNs are both included in the “Wing SN” category). Columns from neurons of interest are labeled with arrows. Right, as above but for neurons one or two hops upstream of the PSI (neurons on the right panel provide at least 1% of synaptic input to neurons on the left panel). C. Diagram showing strong synaptic connectivity between SNs, neurons of interest (including a subset labeled in C), and the TTMn (top) or PSI (bottom). E. Connectivity diagram of DNs upstream of AN05B006 10603 and AN05B006 10221 and their strongest postsynaptic targets. In order to be included on this diagram, presynaptic neurons must be DNs that provide at least 1% of synaptic input to either AN05B006 10603 or AN05B006 10221. Postsynaptic neurons must receive at least 1% of their input and at least 150 synapses (groupwise) from either AN05B006 10603 or AN05B006 10221.

Circuits mediating pre-takeoff postural adjustments.

A. Venn diagram depicting shared first-order downstream connectivity of DNp02 and DNp11 in the VNC. Numbers refer to the number of groups that receive more than 75 synapses and more than 0.5% of their input from either or both DNs. B. Attributes of neurons downstream of DNp02 and DNp11. Top heatmap (“Inp.”): number of synapses onto each group from DNp02 and DNp11. Top bar refers to the group’s assignment in A. All rows are sorted according to the number of synaptic inputs from DNp02 (for the neurons that receive input from DNp02) or from DNp11 (for the neurons that only receive DNp11 input). Second heatmap (“Lat”): Ipsi-contra lateralization index (as in Figure 11H), indicating how bilaterally distributed inputs from DNp02 and DNp11 are. Index was not computed for groups that did not meet the threshold for strong synaptic connectivity. Third heatmap (“Output”): fraction of each group’s output in the indicated VNC regions. Bottom heatmaps (“Attributes”): cell class and neurotransmitter prediction for each group, colored as in Figure 1. C. Connectivity between DNp02, all DNp02-exclusive downstream neurons from panels A/B that synapse onto a MN type in LegNpT2, and those downstream MNs. In order to be included in the diagram MNs must be in LegNpT2 and receive at least 20 synapses from all DNp11-downstream neurons. Downstream neurons must provide at least 5 synapses onto all MNs in this set. Right inset, a diagram of the putative cumulative effects of activating DNp02 on T2 leg muscles based on the connectivity of neurons described in this panel. D. Connectivity between DNp11, DNp11-exclusive downstream neurons, and LegNpT2 MNs as in C. E. Diagram summarizing the putative functional impact of activating synaptic partners downstream of DNp02, DNp11, or both DNs.

Summary of wing, leg and LTct motor control.

A. Wing motor control summary. On the left, wing control circuits are segregated into power and steering circuits, which each also contribute to the activity of indirect control MNs. On the right, neurotransmitter predictions and electrical transmission assignments for neurons of VNC origin (except MNs) that are predicted to play a role in wing motor control (Infomap communities 8, 9, 10, 16). Neurotransmitter predictions are thresholded at ≥0.7 probability and assigned to the unknown (unk.) category if below threshold. Known or putative electrical transmission (from the ‘transmission’ neuprint field) are shown for neurons with unknown neurotransmitter prediction. B. Leg coordination summary. On the left, a schematic adapted from (DeAngelis et al., 2019) showing a potential minimal model to coordinate the 6 legs to allow tripod and tetrapod movement. In the center, a summary of the actual connections seen upstream of the leg premotor and motor circuits, summarized from Figure 14. Line is thicker if there are more than 5 pairs of neurons in that group of the US connectivity (see Figure 14). Glutamatergic neurotransmitter predictions are considered to be inhibitory for this schematic. On the right, comparison of neurotransmitter predictions for the restricted leg premotor neurons and the leg coordination neurons. C. Summary of LTct circuits that likely contribute to escape-related behavior. On the left, distinct DN types converge onto a set of motor effectors that each control overlapping facets of an escape behavior. In the middle, LTct-DN targets ascending neurons that likely play a dual role in actuating motor output and relaying motor information back to the brain (i.e. efference copy). On the right, a subset of DNs unrelated to the control of takeoffs activate inhibitory neurons that suppress takeoff-related actions, thus mediating descending action selection.

VNC tracts according to Court et al., 2020 and Namiki et al., 2018

Drosophila lines used in this study.

Imagery for newly generated split-Gal4 combinations may be found at https://splitgal4.janelia.org/precomputed/Cheong,%20Eichler,%20Stuerner%202023.html