Sexual dimorphism in the complete connectome of the Drosophila male central nervous system

Abstract

Sex differences in behaviour exist across all animals, typically under strong genetic regulation. In Drosophila, fruitless/doublesex transcription factors can identify dimorphic neurons but their organisation into functional circuits remains unclear. We present the connectome of the entire Drosophila male central nervous system. This contains 166,691 neurons spanning the brain and nerve cord, fully proofread and annotated including fruitless/doublesex expression and 11,691 types. We provide the first comprehensive comparison between male and female brain connectomes to synaptic resolution, finding 7,205 isomorphic, 114 dimorphic, 262 male-specific and 69 female-specific types. This resource enables analysis of full sensory-to-motor circuits underlying complex behaviours and the impact of dimorphic elements. Sex-specific/dimorphic neurons are concentrated in higher brain centres while the sensory and motor periphery are largely isomorphic. Within higher centres, male-specific connections are organised into hotspots defined by male-specific neurons or arbours. Numerous circuit switches reroute sensory information to form antagonistic circuits controlling opposing behaviours.

Figure 1 |. A densely annotated and cross-matched male CNS connectome

a Outlines of the aligned male CNS volume, coloured by CNS region (left). Sagittal slice through the EM image data (top right). Insets show (from left): image data, neuron segmentation, and synapse detection. b 3D rendering of all neurons in the connectome. Frontal view; VNC tilted down for visualisation. Numbers represent neuron counts per brain region, including sensory neurons. c Synaptic completion rates for pre- and postsynapses per brain region. d Percentage of connections where both pre- and postsynaptic partners are proofread broken down by neuropil region. Bar width corresponds to size of neuropil. e Percentages of neurons for main superclasses across brain regions (left). Comparison of neuron counts to existing connectomes: female brain (FAFB/FlyWire, top right) and male nerve cord (MANC, bottom right). f Number of neurons per annotation field. The number of unique annotations for each field is shown at the top of the bar. side combines fields somaSide and rootSide; hemilineage combines fields itoleeHl and trumanHl. g Example of the hierarchical annotation of neurons, from superclass through hemilineage, supertype and type. The annotation and the number of neurons in that category is shown in grey. h Percentage of neurons matched to an existing dataset in the whole male CNS and for each CNS region. i Neurons were matched to existing connectomes using a combination of spatial transforms + NBLAST and connectivity co-clustering. Example shown here is DNde003, a descending neuron type, with a 1:1 match in FlyWire and MANC, and matched to two types in hemibrain. j Full view of DNde003 in male CNS, FlyWire and MANC. The number of neurons in each dataset is shown in brackets. EM: electron microscopy; OL: optic lobes; CB: central brain; MANC: male adult nerve cord connectome; VNC: ventral nerve cord.

Figure 2 |. Information flow from sensory input to motor output organizes circuits spanning brain and nerve cord

a Schematic of sensory groups (by class) and motor group (by subclass). Bars show number of neurons per group, lines the peripheral origin or target. Asterisk mark examples in (b). b Sample detailed annotation for sensory and motor neurons. Top: morphology of neck motor neuron CvN7. Stacked bar for exit nerve and hemilineage (itoleeHl). Bottom: wing campaniform sensilla sensory type SNpp32 in the VNC. Stacked bar for subclass and entry nerve. Grey shades indicate annotation values besides those of the sample neurons. c Directed flow diagram of connections from sensory to motor (and endocrine) through the entire CNS. Colours indicate neurotransmitter identity, dashed lines show feedback connections. Edge width is proportional to the number of synapses between groups, node width is proportional to synaptic connections within each group. Edges with fewer than 20k synapses omitted. d Schematic of maxflow analysis. Left: normalized edge weights establish potential information flow capacities between nodes. Right: maxflow assignment of flow value across each edge: flow is constrained by the minimum capacity along a path. e Example of maxflow routes from olfactory sensory inputs to front leg motor outputs. Left: neurons (flow value > 0.25) colored according to their pseudo-layer ordering. Right: strong flow partners (by type) upstream and downstream of example neurons in type DNp44. Arrow to (f) indicates where the flow for this sensory-motor pairing fills in DNp44’s overall sensory-to-motor flow matrix. f Example of mean maxflow through neurons in type DNp44 (left) and sensory modality preference averaged from flow (right). g Flow for neck connective neurons. Proportion of flow utilization of different superclasses (left), max flow preference (middle), and flow specificity for one sensory group, motor group or a single sensory to motor group (right). h UMAP embedding of DN and AN types by sensorimotor flow, coloured by superclass and sized by presynaptic sites, outlined by numbered clusters (top). Types shown in (m) are annotated. Bottom left: coloured by preferred sensory modality, sized by maximum sensory preference value. Bottom right: as above but for motor preference. i Composition of superclass, preferred sensory preference and motor preference for clusters in (h). Arrows indicate clusters containing types in (m). j Behavioral categorization of clusters. Top: histogram of cluster membership of neck connective types described in literature by behavior category. Bottom: compatibility of clusters with behavior categories as predicted based on maxflow analysis. Frequency of descriptions from top panel are overlaid in green. Receiver operating characteristic area under curve 0.84 when compatibility is treated as a multilabel classifier of described behavioral category. k Connection pairs of neck connective types by superclass with mean weight above 20. Plotted by strength of connection in the brain and VNC. Arrows highlight examples of strong connections. Numbers above show type-type connections for each motif. l Axon-dendrite split of DNs and ANs shows the type of connections made by pairs between these superclasses. Only the three most prominent types of connections are shown: axon to dendrite, dendrite to dendrite, axon to axon. m Example of a strong DN to DN connection that is 91% axo-axonic. DNb07 and DNp63 have similar axonal arbours in the VNC, but form distinct dendritic arbours in the brain. The glutamatergic DNb07 synapses onto the cholinergic DNp63 in both the brain and VNC, an inhibitory motif. 30% of their downstream partners are shared, yet the only common target in the brain is AN27X015. AN27X015 projects back from the VNC axonal domain of DNb07 and DNp63 to their brain dendrites, but is preferential for mechanosensory (tactile & proprioceptive) modalities and proboscis and abdominal motor domains.

Figure 3 |. Sexual dimorphism in the fly brain.

a Schematic illustrating differences between iso- dimorphic and sex-specific neurons in morphology or connectivity space. See also Table 1. b Number of dimorphic & sex-specific neurons broken down by superclass. c Example for isomorphic cell type. d Classification of dimorphic types based on morphological differences in the brain. e Example for dimorphic cell type with extra branches. Arrowheads highlight ventro-lateral axons only present in male neurons of this type. f Fraction of synapses on male-specific branches versus total synapse count for each outgoing connection made by AOTU008. g Examples for distributed connections and connections highly localised to male-specific branches. h Example for sex-specific cell type: GNG700m (green). The closest isomorphic cell type (AVLP613, magenta) is shown for comparison. Arrowheads point out differences in axonal projections. i Spatial distribution of dimorphic and sex-specific neurons’ somata. Central brain somas are coloured in green and magenta for male and female, respectively. Pie charts show the proportion of central brain neurons that are dimorphic/sex-specific. j Left: example hemilineage in male and female with sex-specific neurons highlighted in red. Right: neuron count per hemilineage in male versus female. Discounting sex-specific neurons tends to align numbers. The labelled hemilineages collectively produce >50% of all dimorphic & sex-specific neurons in the male. k Distribution of pre- (outputs) and postsynapses (inputs) of dimorphic & sex-specific neurons in male and female connectome. Arrowheads highlight differences between sexes. l Layer assignments relative to sensory inputs for all neurons in the male CNS split by dimorphism. Data points are coloured by superclass. Upper half shows the kernel density estimate of the underlying distribution. Mean layers for antennal lobe projection neurons (ALPNs), Kenyon Cells and central complex neurons are shown as landmarks. Boxplots shows median (vertical line), mean (circle), 1st-3rd quantile (box) and 1.5IQR). m Light-microscopy (LM) image of a fruitless (fru+) clone (top row) and matching neurons found in the EM (middle row). Arrowhead points at variable branches in female aSP-g clones. n Fraction of neurons in the central brain labeled as fruitless- and doublesex-expressing (dsx+) based on LM-EM matches. o Left: pC1 (includes P1 neurons) as an example for a synonym assigned to a group of dimorphic neurons. Right: fraction of dimorphic neurons matched to prior literature.

Figure 4 |. Sexual dimorphism in the visual system reveals a functional “love-spot”

a Diagram of the optic lobe showing the different neuropils and neuronal superclasses. LA: lamina; ME: medulla; LO: lobula; LP: lobula plate; AME: accessory medulla; OLSNs: optic lobe sensory neurons; OLINs: optic lobe intrinsic neurons; VPNs: visual projection neurons; VCNs: visual centrifugal neurons. b Number of types (left) and neurons (right) per superclass across several categories. A type was classified as male-specific only if it was symmetrical; otherwise, it was annotated as unmatched outlier. c Neuron count per type between male and female. Female counts were normalised (x1.134) to account for the ~100 additional ommatidia in the male eye. Only types matched one-to-one between male and female are included. Photoreceptors and Lai are excluded, as many are missing in the male CNS. d The four male-specific visual types. Scale bar, 100 μm. e-f Examples male-specific (e) and dimorphic (f) types that converge onto LoVP92 axons. Inset: close-up and shifted view of the boxed area in e. Arrowheads indicate the location of LoVP92 axons. Stacked bar charts show the input-normalized proportion of dimorphism status of connection partners onto DNpe002. g Dimorphic synapse distribution around LoVP92 axons (black). Green dots show all synapses between dimorphic or sex-specific types. LoVP92 output synapses are shown in magenta. Right: heatmap showing the number of cell types per 10 μm³ voxel in the right hemisphere of the central brain. h Spatial map showing Dm3a,b,c inputs to TmY21 in the third layer of the ME. i Output-normalized percentage of VPN outputs to dimorphic or sex-specific types in the female (left) and the male (right). Dotted lines indicate the average percentage output from all VPNs onto sex-specific or dimorphic partners. Types highlighted in grey are frontally biased, as shown in panel j. j Examples of frontally-biased VPNs. Spatial coverage heatmaps show input synapse distributions mapped onto a Mollweide projection of the right compound eye’s visual field. Color scale bars show input synapse count. k Heatmap showing the strongest inputs from optic lobe intrinsic neurons to frontally-biased VPNs. l The dimorphic type TmY21 in the male (top) and female (bottom). Arrowheads indicate differences in arborisation between the sexes. m Fraction of input connections for each pair of connections between optic lobe intrinsic neurons in the male and female. Only edges with more than 500 synapses are shown. Lines indicate changes in relative weight by factors of 5 and 10. n Network diagram showing male-specific connections between frontal VPNs and DNs. Only DNs reached within two hops were included, using a 2% input threshold – except for DNa02 and DNae002, where a 0.5% threshold was used. AOTU008 did not meet the DN threshold but was included because its top two input partners are VES200m and VES202m. o Diagram of a frontal VPN to DN circuit in the fly brain and nerve cord (left), alongside a simplified circuit diagram (right). VES200m links visual detection – via LoVP92 inputs from the lobula – to the steering descending neurons DNg13. Unilateral activation or inhibition of DNg13 has been shown to cause ipsilateral, or contralateral, turning, respectively. LoVP92 activates VES200m bilaterally, and VES200m inhibits DNg13 ipsilaterally. As a result, activation of LoVP92 leads to inhibition of both DNg13 neurons, preventing turning.

Figure 5 |. A recurrent feedback loop controls courtship song detection and action in the auditory system

a Overview of the auditory pathway and recurrent feedback circuit. Song induced vibrations move the arista to activate Johnston’s organ B neurons, activating a sensorimotor loop that provides feedback from song generating networks back to the central brain. b Circuit schematics for male (left) and female (right) from JO-B to higher order processing centres. Boxes denote cell types, edge thickness scales with synaptic weight, and edge colors indicate transmitter identity. Node colors mark sex classes: isomorphic (white), sexually dimorphic (gold), sex specific (blue). Line widths denote input-normalized connection strengths. c Top 10 output partners of vPN1 (top), vpoEN (middle) and vpoIN (bottom) in males and females. Bars show the fraction of total synaptic outputs (right) (output-normalized) to each identified type across sexes. d Quantification of partner identity for vPN1, vpoEN and vpoIN. Top: fraction of inputs from sex-specific/dimorphic neurons in the female (left) and male (right). Bottom: fraction of outputs to sex-specific/dimorphic neurons in female (left) and male (right). e Connectivity between central brain song detection and ventral nerve cord song production circuits. The established pulse and sine song network in the wing neuropil is reproduced and connected to identified descending neurons and to male specific ascending neurons that project back to central brain nodes. Line widths denote input-normalized connection strengths. Scale bars as indicated.

Figure 6 |. Dimorphism emerges in the third-order of the olfactory system

a Illustration of the olfactory system describing flow of information from sensory neurons to the antennal lobe (cb_sensory) and from there to the lateral horn and mushroom body (cb_intrinsic) via antennal lobe projection neurons (class: ALPN). The left hand side shows example neurons, the right hand side shows the diagrammatic flow of information. AL: antennal lobe, LH: lateral horn, MB: mushroom body, PED: peduncle, CA: calyx. b Rendering of pheromone responsive ORNs: DA1, DL3, VA1d, VA1v. 10 neurons are shown for each type. Below, the percentage of pheromone RNs relative to all RNs. c Comparison of the number of RNs per type in the male (green) and female (FAFB/FlyWire: magenta) brains. The thermo/hygrosensitive glomeruli (VP1–5) are shown on the right of the plot. d Comparison of the percentage input to uniglomerular ALPNs in the male and female brains (FAFB/FlyWire) for pheromone (pink, distinguishing excitatory from inhibitory) and non-pheromone (grey) types. e Number of pheromone and non-pheromone ALPNs per side in the male and female brains. An ALPN is defined as being either from class ALPN or ALON and receiving at least 5% of its input from pheromone ORNs and their uniglomerular PNs. f One of two sexually dimorphic ALPN, MZ_lv2PN, in the male (green) and FAFB/FlyWire (pink). Below, a heatmap of its top RN input (over 0.1% per type) in the male and female brains. g Left: ALPNs that provide at least 10% of their output to sex-specific or dimorphic neurons, in male or female brains. The plot shows percentage output in the male and female (FAFB/FlyWire) brains to dimorphic (yellow) and sex-specific (blue) neurons. Right: ALPN postsynapses of cell types shown in the plot, coloured by type of dimorphism. h Correlation analysis of the percentage of pheromone input for ALPNs (direct and two-hop) versus the percentage output to dimorphic or sex specific neurons. Pheromone ALPNs in pink. On the right: Bootstrapped comparisons of sex-specific/dimorphic targeting across ALPN groups defined by dominant RN input valence. i Network diagram showing strong and selected downstream targets of the uniglomerular pheromone ALPNs (up to 2 hops).

Figure 7 |. Dimorphic second-order gustatory neurons separate contact chemosensation from taste

a Schematic representation (left) and 3D representation of taste sensory subclasses. Taste peg, labellar bristle, and pharyngeal sensilla GRNs (top inset) enter the nervous system via nerves targeting the brain; leg and wing bristle GRNs enter via the VNC. Bottom inset: schematic representation of taste bristle external structure. b Output-normalized proportion (%) of GRNs output to types annotated as sex-specific (blue) or sexually dimorphic (yellow) according to subclass and anatomical location: brain (top), ascending (middle), VNC (bottom). c For each subclass, the main GRN types targeting sex-specific and sexually dimorphic partners in females (left) and males (right).Types with >1% dimorphic outputs in either sex are shown. See Fig S7a for subclasses. d Visualisation of the top ten sex-specific and sexually dimorphic (green) versus isomorphic (black) downstream partner cell types of taste sensory subclasses. Sex-specific and sexually dimorphic neurons extend distinct projections to AVLP, SMP, SLP, SIP, FLA, LegNpT1 (black arrows and Fig S7c). e Input-normalized proportion of putatively pheromone-responsive versus non-pheromone responsive neurons in leg and wing GRN subclasses according to receptor-type mapping. f Left: Downstream partners of putatively pheromone-responsive and non-pheromone responsive local leg and wing GRNs, coloured by superclass. Percentages are output-normalized. Within-type connections are removed; Right: AN types selective for pheromone-responsive GRN input include a single pair of ANXXX093 neurons and 3 serially homologous cell types, AN05B023, AN05B102, AN09B017, each containing multiple subtypes recorded in neuprint.

Figure 8 |. Sexual dimorphism in brain connectivity

a Dimorphic up- and downstream synaptic partners as a fraction of total synaptic partners for all cross-matched male cell types. Arrows on the y-axis indicate means for iso-, dimorphic and male-specific types. b Illustration of neuron-level dimorphism and expected resulting edge dimorphisms. c t-statistic to find significantly different edges between male and female. The denominator is a pooled standard deviation combining the separate standard deviations for male and female edge weights (see Methods for details). d Male (scaled) versus female type-to-type edge weights coloured by false discovery rate (FDR)-corrected p-value. Dotted envelope demarcates a 30% difference in connection weights. e Fraction of dimorphic, isomorphic and noisy connections (male + female) broken down by edge type. False-positives and -negatives are labeled based on expectation as laid out in a. f Proportion of edge types by total number (left) and by synapse count (right) in males (top) and females (bottom). False-positives and -negatives (g) were re-assigned to isomorphic and dimorphic edges, respectively. g Fraction of dimorphic edges/synapses not counting noisy connections. h Male-female differences in spatial distribution of synapses in dimorphic edges. i Example of an isomorphic cell type (vpoEN) with a large fraction of dimorphic outputs (by synapse count) in both male and female. j Fraction of iso- and dimorphic types with at least X% dimorphic in- or outputs (by synapse count) in male. k Fraction of dimorphic in- versus outputs (by synapse count) per cell type in the male. Arrows correspond to labeled thresholds in panel k. l Examples for dimorphic cell types that were split into isomorphic and male-specific branches. m,n For each split type, the fraction of total synapses from isomorphic and dimorphic connections found on male-specific branches (m), and the fraction of dimorphic synapses found on male-specific vs isomorphic branches (n). Arrows indicate the means along the respective axes.

Figure 9 |. Sexual dimorphism in brain networks

a Hierarchical community detection partitions the cell types of the male central brain into blocks of similar connectivity. Clusters at 6 different levels of the hierarchy (left) are derived from the type-to-type connectivity matrix (right). The IDs of male-enriched clusters are shown in red. b Zoom-in on inset highlighted in panel a showing the hierarchical partitions in the adjacency matrix, with clusters enriched with non-isomorphic types in red. Rows are coloured by superclass. c Proportion of within-community synapse counts across the hierarchical levels. d Breakdown of cell type identity in the 13 enriched clusters at level 6. e Network graph of the neighbourhood (≥1,000 synapse threshold) around enriched cluster 158. f Rendering of enriched cluster 158, the most statistically overrepresented for male-specific or dimorphic cell types. g Fruitless and doublesex expression among non-isomorphic cell types in enriched versus non-enriched clusters.