Comparative connectomics of Drosophila descending and ascending neurons

Abstract

In most complex nervous systems there is a clear anatomical separation between the nerve cord, which contains most of the final motor outputs necessary for behaviour, and the brain. In insects, the neck connective is both a physical and an information bottleneck connecting the brain and the ventral nerve cord (an analogue of the spinal cord) and comprises diverse populations of descending neurons (DNs), ascending neurons (ANs) and sensory ascending neurons, which are crucial for sensorimotor signalling and control. Here, by integrating three separate electron microscopy (EM) datasets^1-4, we provide a complete connectomic description of the ANs and DNs of the Drosophila female nervous system and compare them with neurons of the male nerve cord. Proofread neuronal reconstructions are matched across hemispheres, datasets and sexes. Crucially, we also match 51% of DN cell types to light-level data⁵ defining specific driver lines, as well as classifying all ascending populations. We use these results to reveal the anatomical and circuit logic of neck connective neurons. We observe connected chains of DNs and ANs spanning the neck, which may subserve motor sequences. We provide a complete description of sexually dimorphic DN and AN populations, with detailed analyses of selected circuits for reproductive behaviours, including male courtship⁶ (DNa12; also known as aSP22) and song production⁷ (AN neurons from hemilineage 08B) and female ovipositor extrusion⁸ (DNp13). Our work provides EM-level circuit analyses that span the entire central nervous system of an adult animal.

Fig. 1. Reconstruction and identification of three neuronal classes across three datasets.

a, Schematic of the CNS with the three neuronal classes that pass through the neck connective: DNs, ANs and SAs. FANC neurons are shown in MANC space here and in all following figures. A, anterior; D, dorsal; P, posterior; V, ventral. b, Number of neurons in each class and dataset. c, Transects through the neck of the three datasets: female adult fly brain (FAFB), male adult nerve cord (MANC) and female adult nerve cord (FANC). These neck connective transects were used as seedplanes to find and reconstruct the three classes of neurons shown in different colours. White arrows mark SA bundles. Scale bars, 5 μm. d, Number of DNs and ANs that have been left–right matched into pairs or groups in the three datasets. e, Number of DNs and ANs that have a match across the two VNC datasets. f, SAs were assigned modalities by matching to LM images. Left, example of a LM image of a femoral chordotonal organ club; white arrows point at the neuron of interest. Right, the EM reconstructions that were matched to the image. g, DNs were identified in all three EM datasets by matching the EM reconstructions to LM-level descriptions (see Supplementary File 4). Left, example of a LM image of DNa01 in the brain and VNC and next to it the FAFB, FANC and MANC EM reconstructions that were matched to those images. Right, quantification of DNs identified in all three datasets. h, DNs annotated by their soma location, brain and VNC neuropil innervation and the longitudinal tract they take in the VNC (DNa02 is used as an example). DNa, DNs with anterior dorsal soma; PS_LAL, posterior slope and lateral accessory lobe; xl, multiple leg neuropils; fl, front leg; ml, mid-leg; hl, hind leg; ITD, intermediate tract of dorsal cervical fasciculus. See Extended Data Figs. 3–6 for images of DNs across the neck connective, coloured by these four annotations.

Fig. 2. DN matching to new genetic driver lines.

Morphology of identified DNs across all three datasets with nomenclature as described previously. Four types could be found only in the brain (DNp52, DNp66, DNg58 and DNg101), and DNp61 is marked as not found because it is a duplicate. See Extended Data Fig. 7 for matching to DNs previously characterized at light level. See Supplementary File 4 for details on the DN identification. DN morphologies from the female datasets (FAFB and FANC) are in black; the male dataset (MANC) is in red. DNs can be queried and viewed in interactive 3D at https://tinyurl.com/NeckConnective.

Fig. 3. Sensory ranking of DNs.

a, Pie charts showing the neuron class composition of input and output partners of all DNs in the brain (corresponds to FlyWire super_class). Total synapse numbers are shown at the bottom. b, Clustering of FAFB DNs by their sensory input rank (all apart from sensory DNs: DNx01, DNx02 and LN-DNs). The ranks, ranging from 1 to 12, taken from a previous study, are defined as the traversal distances from a given sensory modality to each DN and then averaged by type. Low rank indicates a more direct connection from sensory modality to DN type. A cut height of 5 (dotted line in dendrogram) produces 16 clusters. JO, Johnston’s organ. Asterisks denote clusters with morphologies shown in d. c, Clusters shown in b by the brain neuropil assigned to DN types as a percentage of all DNs in that cluster (for abbreviations, see Extended Data Fig. 4). d, DN morphologies of clusters that are close in rank to several sensory modalities in the brain (clusters marked with asterisks in b). e–i, DN morphologies of clusters that are close in rank to one particular sensory modality (e, gustatory; f, mechanosensory: eye bristles and head bristles; g, mechanosensory: JO and auditory; h, visual projection and ocellar; i, olfactory: thermosensory and hygrosensory). Plots on the left show the average rank of the clusters defined in b for the different sensory modalities. The average rank is plotted for all 353 DN types in FAFB. Centre lines denote median; two hinges denote first and third quartiles; whiskers extend at most 1.5 × interquartile range (IQR) from hinge; and outlying points are shown. Arrows point to specific DN types that stand out. DN morphologies are plotted in their brain neuropil colours.

Fig. 4. Direct DN and AN connections.

Connectivity of ANs and SAs (ANs/SAs) to DNs and vice versa. a, Direct connectivity of ANs/SAs onto DNs in the brain. Connections between DNs and ANs/SAs are averaged by type and plotted by mean weight in per cent to mean weight. Arrows point to the two strongest connections in weight from ANs/SAs onto DNs. b, Direct connectivity of DNs in the VNC onto ANs/SAs. Connections are averaged by type, as in a. Arrows point to the one connection that stands out in both MANC and FANC. The weight in a,b is the number of presynapses. Dots are coloured by brain or VNC neuropil (see Extended Data Figs. 4 and 5 for abbreviations). c, Effective connectivity to MN targets ipsilateral and contralateral to the root side of DNx02 (n = 4). Centre lines denote median; two hinges denote first and third quartiles; whiskers extend at most 1.5 × interquartile range (IQR) from hinge; and outlying points are shown. d, Morphology of DNx02 and AN06B025 in the brain and VNC. The EM morphology from female datasets (FAFB and FANC) is in black; the male dataset (MANC) is in red. e, DNx02 circuit in the brain (FAFB-FlyWire) and in the VNC (MANC). Connections in both datasets are averaged by type and shown in the per cent input to the receiving neuron. AN_SPS_GNG_1 corresponds to AN06B057 in the VNC and targets the neck neuropil (see Supplementary File 12).

Fig. 5. Comparisons across VNC datasets.

a–d, DNa02 as an example of a stereotyped circuit in the VNC. This analysis is provided at https://github.com/flyconnectome/2023neckconnective/blob/main/code/3_DNa02_downstream.Rmd. a, DNa02 output partners in MANC and FANC compared across sides of the same dataset. Best-fit slope is 1.02 for MANC and 0.45 for FANC. b, DNa02 output partner types that receive more than 1% of DNa02 output (all but three matched between MANC and FANC). Left, raw synapse numbers (best-fit slope = 0.42); right, the same neurons shown by the input per cent (best-fit slope = 0.69) to the receiving neuron. The Pearson correlation coefficient (Cor) is shown for a,b. c, Top partner types targeted by DNa02 in MANC and FANC: three sets of serial leg-restricted neurons (IN08A006, IN19A003 and IN13B001), the w-cHIN and a bilaterally projecting neuron (IN07B006). Arrow thickness corresponds to the per cent input to the receiving neuron, and only values higher than 1% are shown. All nodes are single neurons apart from the w-cHIN, number in brackets. d, EM morphologies of the neurons shown in the connectivity graphs in c. Reconstructions from FANC are in black; those from MANC are in red. e, MANC leg premotor circuit neurons (by number of neurons and type) published in a previous study, matched to FANC. All leg-restricted serial sets were found, although some are missing on one or the other side in FANC. All apart from three types of leg coordination neurons were matched to FANC. EM morphologies of those three unmatched types are shown on the right as potentially male-specific neurons.

Fig. 6. Sex-specific or sexually dimorphic neurons.

a, Morphology of DNs in three datasets and ANs in the two VNC datasets that are sexually dimorphic (sex. dimorphic) or sexually specific (sex-specific) as described in the literature or predicted by the matching. The EM morphology from female datasets (FAFB and FANC) is in black; the male dataset (MANC) is in red. The number of sexually dimorphic and sex-specific DNs and ANs in the VNC can be found on the right. b, Density of pre- or postsynapses of the DNs and ANs shown in a compared with previously published images of enlarged regions in the female and male CNS. c, Downstream or upstream partners of sexually dimorphic or sex-specific DNs or ANs in FANC and MANC. Arrows point to the strongest partners by number of synaptic connections and number of neurons connecting onto them. d, Reconstructions of partner neurons in MANC (c, top row) or in FANC that were matched to MANC neurons of that type.

Fig. 7. Sexually dimorphic and sex-specific DNs.

a, Proportion of DNs that are sex-specific or sexually dimorphic by dataset and primary input neuropil. b, Morphology of DNs in the three datasets belonging to the oviDN hemilineage. Two of the six types are uniquely identifiable but we cannot match them across the neck (unmatched types, oviDNd and oviDNv). c, EM morphologies identified within the LM images from a female and a male of the oviDN-SS2 line. d, EM morphologies of previously LM-characterized male-specific DNs and new potentially male-specific DNs. e, EM morphology of the female-specific DN vpoDN (DNp37). f, EM morphologies of LM-characterized sexually dimorphic DNs and new potentially sexually dimorphic DNs. g,h, Connectivity downstream of the sexually dimorphic DNp13 in MANC (g) and FANC (h). There is just one downstream partner in common between the two sexes, IN12A002 (marked with an asterisk). All other partners are sex-specific (coloured black or red) or dimorphic in their connections (pink), or exist in both datasets but are not downstream of DNp13 in the other dataset (coloured grey). The female-specific INXXX998 is also a downstream target of vpoDN. Tergotr., Tergotrochanter; Acc. ti flexor, accessory tibia flexor. i, EM morphology of some of the top VNC targets. j,k, Connectivity downstream of the sexually dimorphic DNa12 (aSP22) in MANC (j) and FANC (k). There are eight downstream neurons in common. Only T1 leg MNs have been systematically identified between the two datasets, so other FANC leg MNs are not shown. Asterisks indicate common downstream partners of dimorphic DN pairs. Tr, Trochanter; Ti, Tibia; Fe, Femur; Ta, Tarsus. l, EM morphology of some of the top VNC targets. The EM morphology from female datasets (FAFB and FANC) is in black; the male dataset (MANC) is in red.

Fig. 8. Sexually dimorphic and sex-specific ANs.

a, Proportion of ANs that are potentially sex-specific or potentially sexually dimorphic by hemilineage, soma neuromere and primary input neuropil. b,c, Morphology of ANs that are potentially sex-specific in males (b, MANC neurons) and females (c, FANC neurons) by hemilineage. FANC neurons in c were assigned hemilineages and soma neuromere if possible and given new type names. d,e, Input circuit in the VNC to potentially sex-specific AN types of hemilineage 08B with soma location in T1 (black box in b,c). d, Male-specific ANs. e, Female-specific ANs. AN08B074 is the previously described male-specific AN vPR1, which as previously hypothesized receives input from DNxn046 (pMP2). Morphology of AN types underneath. All input neurons with more than 2% input onto the receiving AN are shown. FANC neurons in e were matched to MANC neuron types by morphology and connectivity and given the MANC names with an addition of f for female.

Extended Data Fig. 1. Cross section (frontal) of the neck connective in the three datasets.

a, All neurons in the neck connective colour-coded by their longitudinal tract, soma location or predicted neurotransmitter. b, Number of neurons by longitudinal tract, soma location or neurotransmitter in the three datasets. Neurotransmitter predictions are not yet available in FANC. c, Colour legend with details to the abbreviations used in a,b.

Extended Data Fig. 2. Sensory ascending neurons.

a, Morphology of SAs identified in the three EM volumes. In black the EM morphology of DNs from female datasets (FAFB, FANC), in red from the male dataset (MANC). Next to them the LM images that allowed a grouping into sensory subclasses. b, Tract-based analysis of SAs in MANC. None of the SAs project along the MTD, or VLT tract. c, Number of SAs in each tract. d, Number of SA grouped into pairs or populations. e, Correlation of entry nerve to tract membership for MANC SAs.

Extended Data Fig. 3. Morphology matched across the neck—soma location.

Morphology of LM matched DNs across the three datasets colour-coded by cell body location according to a previous study. a, DNa neurons have an anterior dorsal soma; b, DNb an anterior ventral soma; c, DNc a soma in the pars intercerebralis; d, DNd a soma in an anterior outside cell cluster; e, DNp are on the posterior surface; f, DNg are located in the GNG; g, DNx are outside the brain and h, all identified DNs coloured by soma location. In each panel the top images show reconstruction in FAFB in anterior and lateral view; the two bottom left images show MANC and two bottom right FANC in ventral and lateral view, respectively. The bar charts represent the distribution of the VNC longitudinal tract characteristics and VNC neuropil innervation for the neurons in each category - see colour legend.

Extended Data Fig. 4. Morphology matched across the neck—brain neuropil.

Morphology of LM matched DNs across the three datasets colour-coded by their brain neuropil innervation. DNs with input neuropil. a, Superior medial protocerebrum and superior lateral protocerebrum (SMP_SLP); b, vest (VES); c, ocellar ganglion and lobular (OC_LO, vision related); d, posterior lateral protocerebrum (PLP); e, prow and flange (PRW_FLA); f, posterior slope and lateral accessory lobe (IPS_SPS_LAL); g, antennal mechanosensory and motor centre and wedge (AMMC_WED, auditory related); h, gnathal ganglia (GNG); i, multiple innervations of neuropils across the brain (multi) and j, all identified DNs coloured by brain neuropil groups. In each panel the top images show reconstruction in FAFB in anterior and lateral view; the two bottom left images show MANC and two bottom right FANC in ventral and lateral view, respectively. The bar charts represent the distribution of the VNC characteristics longitudinal tract and VNC neuropil innervation for the neurons in each category - see colour legend.

Extended Data Fig. 5. Morphology matched across the neck—VNC neuropil.

Morphology of LM matched DNs across the three datasets colour-coded by VNC neuropil innervation. DNs with output neuropil. a, front leg (fl); b, hind leg (hl); c, multiple innervation in leg compartments (xl); d, neck tectulum (nt); e, wing tectulum (wt); f, multiple innervation into upper tectulum neuropils of the neck, wings and halteres (ut); g, intermediate tectulum (it); h, lower tectulum (lt); i, abdomen (ad); j, multiple innervations of neuropils across the VNC (xn) and k, all identified DNs coloured by VNC neuropil group. In each panel the top images show reconstruction in FAFB in anterior and lateral view; the two bottom left images show MANC and two bottom right FANC in ventral and lateral view, respectively. The bar charts represent the distribution of the brain characteristics soma location and brain neuropil innervation for the neurons in each category - see colour legend.

Extended Data Fig. 6. Morphology matched across the neck—tract.

Morphology of LM matched DNs across the three datasets colour-coded by longitudinal tract membership in the VNC. a, Overview of longitudinal tract in MANC as presented in another study. b, All identified DNs coloured by the tract they take in the VNC. DNs in the tract. c, dorsal lateral tract (DLT); d, median dorsal abdominal tract (MDA); e, ventral route of the mediate tract of dorsal cervical fasciculus (MTD-I); f, dorsal route of the mediate tract of dorsal cervical fasciculus (MTD-II); g, lateral route of the mediate tract of dorsal cervical fasciculus (MTD-III); h, dorsal median tract (DMT); i, intermediate tract of dorsal cervical fasciculus (ITD); j, ventral lateral tract (VLT); k, dorsal lateral tract of ventral cervical fasciculus (DLV); l, ventral median tract of ventral cervical fasciculus (VTV); m, curved ventral lateral tract (CVL) and n, no tract membership (none). In each panel the top images show reconstruction in FAFB in anterior and lateral view; the two bottom left images show MANC and two bottom right FANC in ventral and lateral view, respectively. The bar charts represent the distribution of brain neuropil innervation for the neurons in each category - see colour legend.

Extended Data Fig. 7. DN matching to genetic driver lines.

Morphology of identified DNs across all three datasets with nomenclature as described in a previous study. Two DN types could not be identified (DNd01, DNg25) in any of the three EM datasets and one DN type (DNg28) is only identifiable in the brain. See Supplementary File 4 for slide codes and for DN synonyms from the literature. In black the morphology of DNs from the female datasets (FAFB, FANC) in red from the male dataset (MANC). This figure is also provided in high resolution and DNs can be viewed in 3D at https://tinyurl.com/NeckConnective.

Extended Data Fig. 8. Brain neuropil groups.

Morphology of DNs and ANs by primary brain neuropil. a, All DNs in FAFB-FlyWire were assigned one or two input neuropils. DNs that receive input from more than two neuropils are referred to as multi. b, All ANs in FAFB-FlyWire were assigned one or two output neuropils. ANs that output to more than two neuropils are referred to as multi. Morphologies are coloured by broader neuropil groups: auditory related neuropils (AVLP_AMMC_WED), higher-order multimodal sensory integration 1 (SMP_SLP), vision related (OC_LO), multimodal sensory integration 2 (PVLP, PLP), sensory modalities from the GNG (GNG, SAD), multimodal sensory integration and steering (PS_LAL), vest related (VES) and flange/prow (PRW_FLA).

Extended Data Fig. 9. Neuropil-based analysis of DNs in FANC.

a, Number of DNs/ANs assigned a VNC output/input neuropil in MANC and FANC. Green triangle indicates differences between datasets. A small number of DN types classified as upper tectulum DNs (DNut) in MANC fall below our 80% synaptic output threshold and thus are assigned to multiple neuropil innervating (xn). This is due to slight differences in neuropil meshes between the two datasets. b, Primary neuropil assignment of DNs in the FANC dataset to compare to previously published ones in the MANC dataset. *the DN type targeting specifically the middle leg neuropil in MANC (DNml) has a considerable amount (>5%) of its synapses in the front leg neuropil in FANC, and is therefore in the neuropil category xl (for multiple leg neuropil innervating). c, Synaptic output in % by VNC neuropil of matched DNs in MANC and FANC. Each row represents one DN type, order is conserved between the two datasets. Left bar indicates the previously assigned neuropil-based subclasses from the MANC dataset.

Extended Data Fig. 10. Tract-based analysis of DNs in FANC and ANs in MANC.

a, Tract assignment of all left side DNs in the FANC dataset to compare to previously published tract assignment in the MANC dataset. b, Number of DNs for each tract in comparison to MANC DNs (dotted line). c, DNs grouped into pairs or populations comparing FANC to MANC. d, Correlation of soma location and tract membership for identified FANC DN types based on LM images from genetic driver lines. e, Tract assignment of all left side ANs in the MANC dataset. None of the ANs project along the MTD-II or MTD-III tract. A small additional tract was observed for ANs, referred to as AN-specific dorsal medial tract (ADM). f, Number of ANs in each tract. g, ANs grouped into pairs or populations comparing MANC to FANC. h, Correlation of hemilineage and neuromere to tract membership for MANC ANs. i, Number of DNs/ANs assigned a VNC long tract in MANC and FANC.

Extended Data Fig. 11. Potentially sexually dimorphic or sex-specific ANs in the VNC.

a, Morphology of all the potentially male-specific ANs by type. b, Morphology of the potentially female-specific ANs by newly assigned types. c, Morphology of the potentially sexually dimorphic ANs by MANC type names. In black the EM morphology from the female dataset (FANC) in red from the male dataset (MANC). Stars indicate ANs with missing soma in FANC due to missing EM image data.