Sexually-dimorphic neurons in the Drosophila whole-brain connectome

Abstract

Sexual dimorphisms are present across brains. Male and female brains contain sets of cell types with differences in cell number, morphology, or synaptic connectivity between the two sexes. These differences are driven by differentially-expressed transcription factors, which set the stage for disparate sexual and social behaviors observed between males and females, such as courtship, aggression, receptivity, and mating. In the Drosophila brain, sexual dimorphisms result from differential expression of two transcription factors, Fruitless (Fru) and Doublesex (Dsx), and genetic reagents driven by enhancers for Fru and Dsx label sexually-dimorphic neurons in both male and female brains. The recent release of the first whole-brain connectome for Drosophila provides a unique opportunity to study the connectivity between these neurons as well as their integration into the larger brain network. Here, we identify 91 putative Fru or Dsx cell types, comprising ~1400 neurons, within the whole-brain connectome, using morphological similarity between electron microscopic (EM) reconstructions and light microscopic (LM) images of known Fru and Dsx neurons. We discover that while Fru and Dsx neurons are highly interconnected, each cell type typically receives more inputs from and sends more outputs to non-Fru/Dsx neurons. We characterize the connectivity in the Fru/Dsx networks to predict the function of cell types not previously characterized, we measure distances to the sensory periphery and uncover multisensory interactions, and we map connections to descending neurons that drive behavior. All Fru and Dsx labels reported here are shared within FlyWire Codex (codex.flywire.ai; gene==Fruitless or Doublesex); this work is a critical first step towards deciphering the neural basis of sexually-dimorphic behaviors and for making comparisons with future connectomes of the male brain.

Figure 1.. Identifying putative Fruitless and Doublesex (Fru/Dsx) neurons in the FlyWire brain connectome

(A) FlyWire (EM) candidates of Fru/Dsx types were found by direct comparison to light microscopy (LM) stacks of Fru/Dsx neurons (single neuron ‘clones’ or sparse neuron labeling with genetic lines (Cachero et al., 2010; Yu et al., 2010; Chiang et al., 2011; Liu, 2012; Wu et al., 2016; McKellar et al., 2019; Wang et al., 2020 b, 2021; Nojima et al., 2021; Sterne et al., 2021; Court et al., 2023)) or via matching to previously identified Fru/Dsx candidates in the hemibrain dataset (Scheffer et al., 2020; Schlegel et al., 2024). In both cases, FlyWire candidates were compared with all available sources (see Table 1). Additional Fru/Dsx candidate neurons in FlyWire were identified based on existing annotations of FlyWire neurons in Codex (Matsliah et al., 2024; Schlegel et al., 2024); see Methods). For a given LM source there may by a single Fru/Dsx type (e.g., aDT10; top row), or multiple types that both overlap with the LM morphology, but are not similar in morphology (e.g., Fru/Dsx aDT5a and aDT5b; middle row). Since Fru/Dsx types are defined only by morphology, a given Fru/Dsx may comprise multiple FlyWire subtypes that differ by their connectivity (e.g., pMP6; bottom row). There are a total of 91 Fru/Dsx types and 236 subtypes in the dataset. (B) We identified 1407 putative Fru/Dsx cells in total, here sorted by ‘superclass’ (Dorkenwald et al., 2024) and rendered to a brain template. Our dataset excludes y Fru/Dsx neurons in the mushroom body (see Methods), and in the following superclasses: sensory and visual centrifugal (see codex.flywire.ai for more information on superclasses). (C) Number of Fru/Dsx cells in each superclass and hemisphere. (D) Number of cells per type in the left (green) and right (purple) hemispheres. Type names are colored by superclass (as in B-C). (E) Soma clusters of putative Fruitless types in the posterior (left) and anterior (right) brain, colored by known or predicted (Eckstein et al., 2024) neurotransmitter expression. When classifier prediction did not agree on at least 60% of individual neurons, the type was colored back. (F) same as in (E), but for all Doublesex types, in posterior view (all Dsx cells in the brain are posterior, except aDN). The Dsx ascending SAG neurons are marked outside the brain (not at their physical location) with an upward arrow. Fru pMP3 and Dsx pCd1 share morphology, as do Fru pMP5 and Dsx pCd2. Because we are unable to distinguish these types in the connectome, they are referred to as pMP3/pCd1 and pMP5/pCd2. (G) All posterior (left) and anterior (right) putative Dsx neurons rendered to a brain template and colored by type. aDN neurons (2 per hemisphere) are shown only in the left hemisphere.

Figure 2.. LM-to-EM matches for each Fru/Dsx type

For each one of the 91 Fru/Dsx types, a light microscopy (LM) image (top) and electron microscopy (EM) reconstruction from FlyWire (bottom) is shown - for EM images, we chose a single example cell. The LM source for each type and a FlyWire link corresponding to the displayed neuron are provided in Table 1. In two cases a single LM image was used for multiple Fru/Dsx types: pC2l (pC2la, pC2lb, pC2lc, pC2ld) and LC14a (LC14a1, LC14a2). Several Fru/Dsx types have additional subtypes (based on annotations in Codex) - see Table 1 and Fig. 3. Fru/Dsx types are ordered as in Table 1 (first Fru, then Fru/Dsx, and then Dsx). Cell type name is colored as elsewhere: Fru in blue, Dsx in red, and Fru-Dsx (Fru/Dsx types pMP3-pCd1 and pMP5-pCd2; for those types there is shared morphology between a Fru and a Dsx type) in gray. For cell type a SP30, the source image is not from LM, but rather a hemibrain segment that was previously identified as aSP30 (ID 800579317). This image is marked with *.

Figure 3 -. Connectivity in the Fru/Dsx network

(A) All 1407 putative Fru/Dsx neurons (black), and a set of 1407 neurons from one example ‘matched network’ (purple). ‘Matched networks’ (see Methods) have the same superclass and spatial distribution as the Fru/Dsx neurons. (B) (left) Fru/Dsx neurons (black) make more connections within-network (versus out-of-network), as compared with 100 matched networks. A 5 synapse minimum threshold for a connection was applied. (right) Fru/Dsx neurons (black) have more input and output synapses within-network, as compared with 100 matched networks (purple). Synapses within a primary type were excluded for both Fru/Dsx and matched networks, to avoid possible bias for within network connectivity due to more cells per type in the Fru/Dsx network compared to a matched network. (C) Synaptic Input (blue) and output (red) bias distributions, measuring the bias of each Fru/Dsx type to form more (positive bias) or less (negative bias) synapses with Fru/Dsx partners compared to the number of synapses with any of its partners (see Methods). The input and output biases were calculated separately for each individual Fru/Dsx cell, and then averaged for all the cells of a given type. For example, pC1a has a strong bias for connections with Fru/Dsx partners for both inputs and outputs, while for pC1d, the strong bias is only for the output connections. (D) The percentage of synapses each Fru/Dsx subtype has with any Fru/Dsx partner (including within type), for inputs (synapses with presynaptic partners; blue) and outputs (postsynaptic partners; red). When a cell type has multiple subtypes, * or ** appear near the subtype name. ** indicates that for a given type, this subtype has the most connections with other Fru/Dsx neurons. For example, pMP3/pCd1-CB3301 has 79% of input synapses from other Fru/Dsx neurons and 52% output synapses, more than the input+output % for any other subtype of pMP3/pCd1. When a given Fru/Dsx type has multiple subtypes but some cells do not have a defined primary type in Codex, the Fru/Dsx type is used also as the subtype name (e.g., ‘pSP9-pSP9’). This is in contrast to cases where there is only one primary type per Fru/Dsx type (e.g., pC1a or aSP30) that are not marked with an asterisk. The total number of synapses per type and the number of Fru/Dsx synapses per type both exclude within-subtype connections. (E) Same as (D), but for FlyWire cell types not in the Fru/Dsx list, that have a large proportion of connections with Fru/Dsx neurons (50% or more). These cell types are candidates for being Fruitless or Doublesex positive.

Figure 4 -. Connectivity-based clustering of Fru/Dsx neurons

Fru/Dsx neurons were clustered into 26 clusters based on their connectivity (input and output) with all cell types in the whole brain conenctome (see Methods). Cluster colors are arbitrary, except that neighbouring clusters are assigned different colors. Fru cell types are indicated in blue, Dsx cell types in dark red, and Fru/Dsx cell types in gray. Consistent with previous clustering into subtypes (Matsliah et al., 2023; Schlegel et al., 2024), cells from the same primary type (or Fru/Dsx subtype) share the same cluster. A black filled circle indicates a descending type.

Figure 5 -. Network of significant connections between Fru/Dsx subtypes

Network diagram for Fru/Dsx types. Edge from type A to type B is shown only if at least 3% of A’s output synapses are to B cells and at least 3% of B’s input synapses are from A cells (‘symmetric threshold’, see Methods). Edge widths further represent the absolute count of synapses connecting the types, and edge tips represent excitation/inhibition (arrow=excitation, round=inhibition) based on neurotransmitter prediction (Cholinergic cells are considered excitatory, while GABAergic and Glutamatergic cells are considered inhibitory). Node colors represent superclasses, and their shapes represent genes (ellipse=Dsx, octagon=Fru, rectangle=Dsx+Fru, hexagon=other). 119 out of 236 Fru/Dsx subtypes are present in the network.

Figure 6 -. Predicting Function from Connectivity

(A) Strongest direct and indirect connections of pMN1-DNp13 and PMN2-vpoDN types with Fru/Dsx types. First, the subgroup of Fru/Dsx types with a strong (1% symmetric threshold; see Methods) directly connected to pMN1,2 were added. Then, the Fru/Dsx types connected to this subgroup (using a 3% symmetric threshold) were added. Line width reflects connection strength (S), using three categories for the strength. Types that are known to be tuned to ‘Pulse song’ (one of the two major courtship song types in D. melanogaster) are marked with a red box. None of the types in this diagram is known to be tuned to ‘Sine song’. Excitatory and inhibitory connections are represented by pointed and flat arrows, respectively. Connections are considered inhibitory if predicted in FlyWire as GABAergic or glutamatergic, and as excitatory if predicted as cholinergic, with the following caveat: SAG neurons are predicted to be Serotonergic in FlyWire but are shown to be cholinergic (Wang et al., 2020b). ‘Fru’, ‘Dsx’ and ‘Fru or Dsx’ are colored in blue, red and gray as done elsewhere. (B) Three pC2lb subtypes that appear in (A) are rendered to brain template in FlyWire: pC2la-CL313 (red), pC2lb-AVLP567 (green) and pC2lb-AVLP569 (blue). Arrows point to processes that exist in pC2lb but not in pC2la. Dashed arrows - small posterior-lateral processes that exist in pC2lb-AVLP567 but not in pC2lb-AVLP569. Note that for simplicity, only direct connections with LC31b are shown, and not connections between the types that are connected to LC31b, even if they are strongly connected with each other. (C) Direct connections between Fru LC31b and all Fru/Dsx types. A symmetric threshold of 0.05% was used, lower than the threshold used in (A). LC31b is a visual projecting neuron that is directly connected upstream with Fru+ and Dsx+ auditory neurons that are tuned to Pulse-song. This suggests multisensory integration in the rejection-controlling circuit. Note that for simplicity, only direct connections with LC31b are shown, and not connections between the types that are connected to LC31b, even if they are strongly connected with each other. (D) Innervation of the anterior and posterior lobula by example LC10a and LC31b cells. (E) The descending types Dsx pMN1-DNp13 and Fru pIP9 are rendered in FlyWire and colored in green (Dsx+) or purple (Fru+). Note that as those are descending neurons, the ventral part, in the VNC, is missing in our images. Each type has exactly 1 cell per hemisphere. pMN1-DNp13 has shared morphology (as well as shared connectivity, see Fig 4, Fig 5 and Fig. 6A) with pIP9, possibly also sharing functional roles. (F)-(G) Direct connections between aSP8 or aSP30 in panel F and G, respectively and all Fru/Dsx types. A symmetric threshold of 0.05% was used as in (C). (F) aSP8 inhibits both pMN1-DNp13 and pMN2-vpoDN and cells that are connected directly to those two types, therefore possibly driving ‘global inhibition’ in the rejection and receptivity. (G) aSP30 receives excitatory inputs from pMN2-vpoDN (which controls a receptivity female behavior, vaginal plate opening) and from other cells types which are strongly connected to pMN2-vpoDN, and sends inhibitory connections to pMN1-DNp13 (which controls a rejecting female behavior, ovipositor extrusion) and to cells that are strongly connected to pMN1-DNp13, therefore driving ‘cross inhibition’. aSP8-vpoIN¹ = AVLP008; aSP8-vpoIN² = LHAV4c2; aSP8³ = aSP8; aSP8b¹⁻³ = AVLP010,011,013; aIPgb¹ = CB2131; aIPgc¹ = CB1127; aIP1¹ = CB2986; aIP1² = CB2409; pC2la¹ = CL313; pC2la² = CB2375, pC2lb¹ = AVLP567; pC2lb² = AVLP569; pIP14¹ = AOTU062; pSP7¹ - SLPpm3_H01; pMP5-pCd2¹ = SMP286.

Figure 7 -. Information flow from sensory modalities to Fru/Dsx neurons

(A) Top, UMAP analysis of the matrix of normalized traversal distances (Schlegel et al., 2021; Dorkenwald et al., 2024), resulting in a 2D representation of each neuron in a sensory space (see Methods). Each dot represents a single neuron. Fru, Dsx, Fru-Dsx neurons are colored in blue, red and gray, respectively. The approximated average location of the cells belonging to some types are shown. The Bottom, we coloured neurons in the UMAP plot by the rank order in which they are reached from 4 seed groups (Mech. -JO: mechanosensory-JO; see (Dorkenwald et al., 2024)). Purple neurons are reached earlier than green neurons. For example, mAL neurons (see arrow in top panel) are positioned around cells with early rank from the Gustatory sensory neurons (bottom panel ‘Gustatory’), consistent with their role in processing gustatory cues (Clowney et al., 2015), while aDT3 has small rank from the olfactory seed (Jefferis et al., 2007; Datta et al., 2008; Yu et al., 2010), while aSP18 has small rank from the Mech. -JO seed, consistent with its low rank from Mechanosensory-JO (see 7(C); (Liu, 2012)). (B) Fru/Dsx neurons with different ranks from different sensory seeds (mechanosensory-JO, olfactory, gustatory, or visual projection; see Methods). No mechanosensory-JO, olfactory and gustatory sensory neurons (rank 1) exist in our Fru/Dsx network, while some visual projection neurons do. (C) Fru/Dsx subtypes are sorted by normalized rank from Mechanosensory-JO. Known auditory types (Deutsch et al., 2019; Baker et al., 2022) are colored according to their tuning to syllables of courtship song, ‘pulse’ or ‘sine’. Intermediate turning implies no strong preference to Pulse over Sine song as previously defined (Baker et al., 2022). (D-F) same as (C) for Gustatory (D), Olfactory (E), and Visual projection (F) seeds. Fru/Dsx subtypes are colored by superclass. Normalized ranks from sensory modalities were calculated as before (see Methods), by converting absolute ranks to percentile as previously done (Dorkenwald et al., 2024), and averaged over all the cells in a given subtype. For example, aSP18-CB1125 has a normalized rank of 1% from mechanosensory-JO, implying that 99% of FlyWire cells have a higher absolute rank than an average cell in the aSP18-CB1125 group. (G–I) Fru/Dsx subtypes with normalized rank < 10% with pairs of modalities: mechanosensory-JO and olfactory (G), mechanosensory-JO and gustatory (H), gustatory and olfactory (I). Insets show rendered neurons for some multisensory types. Only around half of the aSP9 neurons (28/54) are multisensory (according to their ranks). Compared to all 17/17 (all subtypes) for aIP1 and 45/48 cells (11/12 subtypes) for pSP9. Only one aSP8 subtype (LHAV4c2; 3 cells) is multisensory. Pairs with visual-projection are not shown: no type has normalized rank <10% from both visual proj. and gustatory seeds. Types pSP5-CB3776 and aIP3b-CB1688 have norm. rank<10% with vis. Proj. and norm. rank <10% with olfactory or mech.-JO seeds, respectively.

Figure 8 -. Fru/Dsx neuron connectivity to descending neurons

(A) Strong direct synaptic connections (1% or larger, symmetric threshold; see Methods) between Fru/Dsx neurons and DNs (including DNs that are Fru/Dsx, such as pMN1-DNp13 and pMN2-vpoDN and DNs that are not in the Fru/Dsx list). This map reveals DNs that may be important for driving sexually dimorphic behaviors. Some of these have already been characterized (for example, Fru oviDNs or Dsx vpoDN (Wang et al., 2020b, 2021; Vijayan et al., 2023)), but most have not. Fru, Dsx, Fru-Dsx types are colored blue, brown and gray as elsewhere. Descending neurons not in the Fru/Dsx list are marked in black. Possibly some of these types are Fruitless or Doublesex positive. All Descending neurons are marked with an underline. Note that some descending neurons are upstream of other descending neurons, and are therefore not at the bottom row of the map. (B) Examples for connections between pairs of descending cells: fru+pIP9 is upstream of both pMN12-DNp60 (left), and pMP12-DNp67 (middle). Both pMP12-DNp60 and pMP12-DNp67 are directly upstream of DNge144, a pair of descending neurons (one per hemisphere) that are not in the Fru/Dsx list. Synapses are shown as red circles. (C) Three Descending types have strong direct connections upstream the descending cell type DNg19 (which is not in the Fru/Dsx list): fru+aSP22 (left), dsx+pMN2-vpoDN (middle) and fru+Bluebell-DNg60. Synapses are marked in yellow.