Introduction

Life is a rich multisensory experience for most animals. As a result, nervous systems have evolved to use multisensory representations of objects, scenes and events to most effectively guide behaviour¹. It is widely appreciated that multisensory learning improves memory performance, from children in the classroom to rodents and insects in controlled laboratory experiments^2–5. Moreover, an apparently universal and unexplained feature of multisensory learning is that it improves subsequent memory performance even for the separate unisensory components^3,5. Studies in humans and mammals have suggested that multisensory learning benefits from interactions between modality specific cortices that were co-active during training, and that individual senses can reactivate both areas at testing^3,6–9. In addition, cells in different brain regions respond to multiple sensory cues and the proportions/numbers change after multisensory learning^1,10–12. However, we currently lack detailed mechanistic understanding of how multisensory learning converts neurons from being modality selective to multimodal, and how enhanced multisensory and unisensory memory performance could be supported by such a process.

In Drosophila, unique populations of mushroom body (MB) Kenyon Cells (KCs) receive predominant and anatomically segregated dendritic input from olfactory or visual projection neurons (as well as local visual interneurons) and their axons project as parallel streams into the MB lobes. There, successive compartments of each KC’s axonal arbor are intersected by the presynapses of dopaminergic neurons (DANs) that convey the reinforcing effects of appetitive or aversive stimuli^13,14. Reinforcing dopamine depresses synapses between active KCs and the compartment-restricted dendrites of downstream mushroom body output neurons (MBONs) to code valence-specific memories^15–17.

Multisensory learning improves memory

To study multisensory learning in Drosophila we adapted the olfactory T-maze¹⁸ so that colours and odours can be presented together (Figure. 1a). Food-deprived flies were trained by presenting them with a colour and/or odour (conditioned stimulus minus, CS−), followed by another colour and/or odour (conditioned stimulus plus, CS+) paired with sugar reward (Figure. 1a-b). When trained and tested with only colours [Visual learning, V], flies did not show significant learned preference for the previously sugar paired colour (Figure. 1b-d; Figure 1 - figure supplement 1a, 24h memory). However, combining colours with odours [Congruent protocol, C] produced robust and long-lasting memory, which was significantly enhanced over that formed by training with only odours [Olfactory learning, O] (Figure 1b-d, and Figure 1 - figure supplement 1a). If the colour and odour pairs were reversed between training and testing [Incongruent protocol, I] (Figure 1b-d; Figure 1 - figure supplement 1a,b), no memory enhancement was observed. Furthermore, memory enhancement was not apparent if the same colour was presented with CS− and CS+ odours during training and testing (Figure 1 figure supplement 1c, C v Csc). The memory enhancing effect of multisensory learning therefore requires a learned relationship between specific colour and odour combinations.

Figure 1.

To further investigate multimodal memory enhancement, we restricted presentation of multisensory cues to either training or testing. Multisensory training improved memory retrieval even when testing used unisensory cues [Olfactory Retrieval, OR; Visual Retrieval, VR] (Figure 1e,f). In contrast, presenting multisensory stimuli during testing did not improve performance in olfactory trained flies [Multisensory Retrieval, MSR] (Figure 1 - figure supplement 1d). Moreover, the greatest improvement in performance was observed when multisensory stimuli were used during training and testing (Figure 1 - figure supplement 1e). Therefore, multisensory training enhances memory performance for the individual colour and odour memory components, and congruence of colour and odour information between training and testing further improves performance. Although our experiments and elsewhere ^5,19 imply that flies distinguish green and blue colours, we do not discount a contribution of hue and luminance.

Olfactory retrieval requires visual KCs

Dendrites of the numerically larger populations of olfactory KCs within the αβ, α′β′ and γ lobes (ie. γ main, γm KCs) occupy the main calyces of the MB, whereas the relatively small populations of αβ posterior (αβp) and γ dorsal (γd) KCs receive predominantly visual information via dendrites in the accessory calyces¹⁴. γd KCs were previously implicated in colour learning^19,20. We tested roles for visual γd and αβp KCs in multisensory learning and memory, using cell-specific expression of a UAS-Shibire^ts1 (Shi^ts1) transgene which encodes a dominant temperature-sensitive dynamin²¹. At temperatures >30°C, Shi^ts1 blocks membrane recycling and thus impairs synaptic transmission. Function can be restored by returning flies to <29°C. Blocking output from γd and αβp KCs during testing at 6h abolished visual enhancement of performance in the congruent protocol, and removed the interference of incongruent color cues. In both instances memory performance was similar to that of flies tested with odours alone (Figure 2a-d, Figure 2 - figure supplement 1a-f). These results suggest activity in γd and αβp KCs represents the visual component of multisensory memory (see also Figure 2 - figure supplement 1g-h). Surprisingly, blocking γd KC (but not αβp KCs) output during testing also impaired performance for odour-only memory retrieval in multisensory trained flies (Figure 2e-f), despite having no effect on memory retrieval in flies trained with only odours²² (Figure 2 - figure supplement 1a,d). This unexpected result led us to hypothesize that multisensory learning might expand the representation of odours to include ‘visual’ γd KCs.

Figure 2.

DPM neurons are positioned to bridge KC sensory streams

Anatomy of the MB network suggests two possible ways to confer odour responsiveness to γd KC axons: via KC-KC synapses or neurons positioned to bridge the different KC streams. We queried the anatomical feasibility of these routes using the complete MB connectome of a single adult female fly ‘hemibrain’ electron microscope (EM) volume^32,33. Although most (562 of 590) γm KCs make synapses with γd KCs, number and placement of these connections does not support every γd KC to receive γm input in every γ lobe compartment. In addition, KC-KC connections were reported to suppress activity in neighboring KCs³⁴. We therefore moved to identify neurons that could bridge different KC sensory streams, and studied the fine anatomy of the γ lobe innervation of the potentially excitatory serotonergic Dorsal Paired Medial (DPM) neuron in the hemibrain EM volume. DPM neurons send separate branches that densely innervate the vertical and horizontal lobes and distal peduncle of the MB, where they are both pre- and post-synaptic to KCs^35–37. Ultrastructure of DPM neuron projections in the γ lobe revealed two branches within γ1 and other ventral and dorsal branches passing through the γ2 - γ5 compartments (Figure 3a). Positions of DPM neuron synapses onto γd KCs follow the γd KC axon bundle as it winds around the γ lobe from ventral in γ1 to dorsal in γ5 compartment (Figure 3a).

Figure 3.

Annotating a dendrogram of DPM neurites (Figure 3b) with γ lobe compartment boundaries (based on DAN connectivity), synapses from γm KCs and those to γd KCs, showed that unique branches of the DPM neuron can provide compartment-specific microcircuit bridges between γm and γd KCs. DPM neurons can also bridge γd to γm KCs (Figure 3c). The large GABAergic anterior paired lateral (APL)³⁸ neuron was found to make synapses along DPM branches in the γ lobe (Figure 3b,c), suggesting DPM bridging can be regulated by local inhibition. Importantly, the APL neuron receives many DAN inputs within each compartment and can therefore also be regulated with region specificity³⁹ (Figure 3d), to potentially release specific DPM branches from APL inhibition.

APL and DPM neurons bridge sensory streams in multisensory memory

We challenged a putative microcircuit bridge model by independently manipulating APL and DPM neurons. Expression in APL neurons of the Dop2R dopamine receptor has been linked to aversive learning⁴⁰ and transciptional profiling suggests APL neurons also express DopEcR receptor⁴¹. Both of these dopamine receptors are inhibitory^42,43. We therefore used tubP-GAL80^{ts 44} to temporally-restrict transgenic RNAi expression in APL neurons to test a role of these receptors in multisensory learning. Knocking down Dop2R in adult APL neurons abolished multisensory enhancement of olfactory retrieval performance (Figure 4a, Figure 4 - figure supplement 1a). A mild defect was also observed for odour memory following olfactory appetitive conditioning, however the difference was only significant to one control (Figure 4b, Figure 4 - supplement 1b). In contrast, DopEcR RNAi had no effect in either experiment (Figure 4 -figure supplement 1c,d). These results are consistent with inhibition of APL neurons by reinforcing dopamine, to allow recruitment of γd KCs into the olfactory memory engram during multisensory learning.

Figure 4.

We tested a role of the serotonergic DPM neurons in expanding the olfactory engram to include visually responsive KCs using expression of UAS-Shi^ts1. Temporally restricting transmission from DPM neurons either during acquisition (Figure 4c, Figure 4 - supplement 1e) or retrieval (Figure 4d; Figure 4 - supplement 1e) significantly impaired the olfactory retrieval of multisensory memory. Blocking DPM neuron output also impaired retrieval of visual memory after multisensory learning (Figure 4 - supplement 1f,g). These same manipulations had no effect on odour memory after unisensory olfactory learning (Figure 4 - supplement 1h), consistent with prior work⁴⁵. These data are consistent with DPM neuron output being required during learning to bind together simultaneously active KC streams, whereas DPM neuron output during memory retrieval provides the connection for odour-driven γm KCs to activate the relevant γd KCs.

If multisensory learning establishes DPM neuron microcircuit bridges between olfactory γm and visual γd KCs, output from the γm KCs should be required during multisensory learning, like that of DPM neurons. We therefore used UAS-Shi^ts1 to test whether γm was required during learning for enhancement of olfactory and visual memory retrieval after multisensory training. Whereas blocking γ KCs during multisensory learning significantly impaired both olfactory and visual retrieval (Figure 4e, f, Figure 4 - figure supplement 1i, j), it did not alter olfactory learning (Figure 4g), as in prior studies^46–48. A requirement for γm KC output for multisensory memory is therefore consistent with the DPM bridge model, and suggests that multisensory learning involves different plasticity rules to that of unisensory olfactory memory^16,49.

For DPM neuron-released serotonin (5-hydroxytryptamine, 5-HT) to mediate a bridge from γm to γd KCs, there should be a requirement for KC expressed 5-HT receptors. Since 5-HT can exert excitatory effects through 5-HT2A and 5-HT7 type receptors⁵¹, we used RNAi to knockdown these receptors in γd KCs. Reducing 5-HT2A but not 5-HT7 or 5-HT2B receptor expression impaired olfactory memory performance after multisensory training (Figure 4h) but not olfactory training (Figure 4 - figure supplement 1k). Taken together the anatomical and behavioural genetic data lead us to propose that reinforcer-evoked compartment specific dopamine releases APL-mediated inhibition. This could facilitate the same reinforcing dopamine to induce KC-DPM and DPM-KC plasticity that forms excitatory serotonergic DPM microcircuit bridges between the relevant γm and γd KCs activated by specific odour-colour combinations (Figure 4i).

Engram expansion benefits new learning

An expansion of the CS+ odour representation into a particular segment of the γd axons after multisensory training might be expected to facilitate subsequent learning with the same odour, if the next DAN teaching signal intersects the expanded KC representation. We reasoned that if DANs direct the recruitment of γd axons to become odour activated, an aversive learning event which requires dopamine release from the PPL1-γ1pedc DANs that innervate the most proximal γ1 lobe compartment³¹ should confer odour-responsiveness onto all downstream γd axon segments from γ1 to γ5 (Figure 5a). In contrast, reward learning which requires dopamine from PAM-γ4 and PAM-γ5 DANs, might only make these terminal segments of the γd KCs odour responsive (Figure 5b). We first confirmed that multisensory colour and odour aversive (electric shock) training produced memory enhancement for both combined and individual cues, similar to that observed by others 5, and after appetitive training (Figure 5 - figure supplement 1a-f), and that γd KCs were also required for odour memory enhancement following aversive multisensory training (Figure 5 - figure supplement 1g-l).

Figure 5.

We then tested the axonal recruitment model by sequentially training flies with either an aversive (dopamine in γ1 and γ2) or appetitive (dopamine in γ4 and γ5) multisensory protocol followed by unisensory odour-reward or odour-punishment learning (Figure 5c,d). Prior multisensory aversive training significantly enhanced subsequent odour-reward learning (Figure 5c). However, no enhancement was apparent if aversive odour learning followed multisensory appetitive learning (Figure 5d). These data are consistent with a multisensory training-dependent expansion of the CS+ odour representation being included in the next CS+ odour memory engram when appropriate γd axon segments have become CS+ odour-activated. Larger colour and odour memory engrams provide a mechanism for how odour and colour memory performance is enhanced following multisensory training (Figure 1e,f), and explains why odour memory retrieval in this context acquires a requirement for γd KC output (Figure 2e).

Discussion

Our study describes a neural mechanism in Drosophila through which multisensory learning improves subsequent memory performance, even for individual sensory cues. A single training trial with visual cues could only generate robust memory performance if they were combined with odours during training, similar to visual rhythm perception learning in humans, which requires accompanying auditory information⁵². We propose that multisensory learning binds together information from temporally contingent odours and colours within axons of mushroom body γ KCs, via serotonergic DPM neurons, whose activity also defines the coincidence time window⁵³. For this to occur, we reason that learning-driven binding converts axons of visually (presumably colour) selective KCs to also become responsive to the temporally contingent trained odour. We assume the reverse may also apply, that axons of olfactory selective KCs become activated by the temporally contingent trained colour. While predominant dendritic input defines γm KCs as being olfactory and γd as being visual, the anatomy of DPM neurons, the positions of reinforcing DANs, and sequential learning experiments suggest that segments of their axons may become multimodal after multisensory learning. This result suggests that γ KCs are a likely substrate where other temporally contingent sensory information can be integrated with that of explicit sensory cues^54,55.

Although our experiments mostly focused on odour activated γm KCs recruiting colour γd KCs via DPM microcircuits, the observed behavioural enhancement of visual memory following multisensory learning and the reciprocal connectivity of DPM neurons suggests that DPM neurons could mediate a reverse bridge, and even cross modal interactions between other sensory streams. In so doing, multisensory learning uses DPM neurons to link KCs that are responsive to each temporally contingent sensory cue and expands representations of each cue into that of the other. Such a cross-modal expansion would allow multisensory experience to be efficiently retrieved by combined cues and by each individually. As a result, trained flies can evoke a memory of a visual experience with the learned odour, and memory of an odour with the learned colour. These findings provide a neural mechanism through which the fly achieves a conceptual equivalent of hippocampus-dependent pattern completion in mammals, where partial cues can retrieve a more complete memory representation⁵⁶. Interestingly, human patients with schizophrenia and autism exhibit deficits in multisensory integration⁵⁷ and these conditions have been linked to serotonergic dysfunction and 5-HT2A receptors⁵⁸. Our work here suggests that inappropriate routing of multisensory percepts may contribute to these conditions. Moreover, the excitatory 5-HT2A receptors that mediate multisensory binding, are the major targets of hallucinogenic drugs⁵⁹.

Methods

Fly strains

All Drosophila melanogaster strains were reared at 25 °C and 40-50% humidity, except where noted, on standard cornmeal-agar food (100 g l^-1 anhydrous d-glucose, 47.27 g l^-1 organic maize flour, 25 g l^-1 autolyzed yeast, 7.18 g l^-1 agar, 12.18 g Tegosept dissolved in 8.36 ml absolute ethanol, per liter of fly food) in 12:12 h light:dark cycle. Canton-S flies were used as wild-type (WT) and originated from William Quinn’s laboratory (Massachusetts Institute of Technology, Cambridge, MA, USA). The following GAL4 lines were used in the behaviour experiments: MB607B-GAL4^60,61, MB009B-GAL4^60,61, c708a-GAL4⁶², VT43924- GAL4.2⁶³, VT64246-GAL4⁶⁴. Temperature controlled blocking of neuronal output was achieved by expressing the UAS-Shi^{ts1 65} transgene under the control of the MB607B-GAL4^60,61, MB009B-GAL4^60,61, c708a-GAL4⁶² and VT64246-GAL4⁶⁴ drivers. For RNAi knockdown experiments involving APL, tubP-GAL80^{ts 66}, VT43924-GAL4.2⁶³ flies were crossed with UAS-Dop2R RNAi⁶⁷ and UAS-DopEcR RNAi (VDRC ID 103494) flies. The same driver line was crossed to wild-type flies and the RNAi background strain (VDRC ID 60100), as controls. For RNAi knockdown experiments involving γd KCs, MB607B-GAL4^60,61 flies were crossed with UAS-5-HT2a RNAi (BDSC 31882), UAS-5-HT2b RNAi (BDSC 60488) and UAS-5-HT7 RNAi (BDSC 27273) flies. The same driver line was crossed to the RNAi background strain (BDSC 36304), as controls. We used both male and female flies for all experiments.

Behavioural experiments

Male flies from the GAL4 lines were crossed to UAS-Shi^ts1 virgin females, except for experiments involving c708a-GAL4, where UAS-Shi^ts1 males were crossed to c708a-GAL4 virgin females. For heterozygous controls, GAL4 or UAS-Shi^ts1 flies were crossed to WT flies. In RNAi experiments GAL4 or RNAi flies were crossed to the appropriate RNAi background strains or WT flies, respectively. All flies were raised at 25 °C, except where noted below for manipulation of RNAi expression. Populations of 2-8 day-old flies were used in all experiments.

For appetitive conditioning experiments, 80-100 flies were placed in a 25 ml vial containing 1% agar (as a water source) and a 20 × 60 mm piece of filter paper for 19–22 h before training and were kept starved for the entire experiment, except when assaying 24 h memory where flies were fed for 30 min after training then returned to starvation vials until testing. For aversive conditioning experiments, 80-100 flies were placed in a vial containing standard food and a piece of filter paper for 14–22 h before behavioural experiments.

For experiments involving neuronal blocking with UAS-Shi^ts1, a schematic of the timeline of temperature shifting is provided in each figure. For Shi^ts1 experiments, flies were transferred to restrictive 33 °C 30 min prior to training and/or testing. For RNAi experiments involving tub P-GAL80^ts; VT43924-GAL4.2, flies were raised at 18 °C and shifted to 29 °C after eclosion to induce RNAi expression for 3 d before the behavioural experiments. The flies remained at 29 °C for the duration of the experiments.

All behavioural experiments were conducted using a standard T-maze that was modified to allow simultaneous delivery of colour and odour stimuli. The T-maze which is made from translucent plastic was covered in opaque blackout film to minimise interference between the visual stimuli when they were used in parallel. Odours were 4-methylcyclohexanol (MCH) and 3-octanol (OCT) diluted in mineral oil (at ≈1:10^-3 dilution). Colours were provided by lightemitting diodes (LEDs); green LEDs with wavelength 530 +/- 10nm (ProLight Opto, PM2E-3LGE-SD) and blue LEDs with wavelength 465 +/- 10nm (ProLight Opto, PM2B-3LDE-SD). Four LEDs were assembled in a circuit built onto a heat sink and were mounted securely on top of the odour delivery tubes. The intensities of the LEDs were adjusted so that naïve flies showed no phototactic preference between the illuminated T-maze arms. Visual stimuli were presented in the same manner and same intensity for both training and testing. For appetitive experiments the testing tubes were lined with filter paper and for aversive experiments the testing tubes were lined with non-electrified shock grids. Experiments were performed in an environmental chamber set to the desired temperature and 55-65% relative humidity. Flies were handled prior to training and testing under overhead red light.

Appetitive conditioning was performed essentially as previously described⁶⁸. Briefly, flies were exposed for 2 min to stimuli Y (Y_Colour and/or Y_Odour) without reinforcement in a tube with dry filter paper (the Conditioned Stimulus minus, CS−), 30 s of clean air, then 2 min with stimuli X (X_Colour and/or X_Odour) presented with 5.8 M sucrose dried on a filter paper (the Conditioned Stimulus plus, CS+). For aversive olfactory conditioning^69,70, flies recieved 1 min exposure to stimuli X (X_Colour and/or X_Odour) paired with twelve 90 V electric shocks at 5 s intervals (CS+), 45 s of clean air, followed by 1 min exposure to stimuli Y (Y_Colour and/or Y_Odour) without reinforcement (CS−). Electric shocks were delivered using a Grass S48 Square Pulse Stimulator (Grass Technology). Shock grids are those previously described⁷¹ and consist of interleaved copper rows printed on transparent Mylar film which allows coloured light to pass through.

Memory performance was assessed by testing flies for their preference between the CS− and the CS+ colours and/or odours for 2 min. Odour testing was performed in darkness. The flies in each arm were collected and transferred to polystyrene tubes (Falcon 14mL Round Bottom Polypropylene Test Tube with Cap). Tubes with flies were frozen at −20°C and flies were then removed and manually counted.

Performance Indices were calculated as the number of flies in the CS+ arm minus the number in the CS− arm, divided by the total number of flies. For all behavioural experiments, a single sample, or n, represents the average Performance Index from two independent groups of flies trained with the reciprocal colour/odour combinations as CS+ and CS−. The total n for each experiment was acquired over 3 different training sessions on different days.

Six behavioural protocols were used:

1. Visual (V) Learning: colours (X_Colour and Y_Colour) were used as CS+ and CS−.

2. Olfactory (O) Learning: odours (X_Odour and Y_Odour) were used as CS+ and CS−.

3. Congruent (C) protocol: colours and odours were combined (X_Colour + X_Odour and Y_Colour + Y_Odour) as CS+ and CS−. The same colours and odours combinations were used during training and testing.

4. Incongruent (I) protocol: colour and odours stimulus contingencies were switched between training (X_Colour + Y_Odour and Y_Colour + X_Odour) and testing (X_Colour + X_Odour and Y_Colour + Y_Odour). The V and O protocols are unisensory, whereas C and I are multisensory.

5. Olfactory retrieval (OR): flies were trained as in the Congruent (C) protocol , but only odours (XOdour and YOdour) were presented as the choice at test.

6. Visual retrieval (VR): flies were trained as in the Congruent (C) protocol , but only colours (X_Colour and Y_Colour) were presented as the choice at test.

   The sequential learning experiments depicted in Figure 5c and 5d used aversive or appetitive Congruent (C) multisensory training followed by unisensory appetitive or aversive Olfactory (O) learning, then testing using olfactory retrieval (OR).

Statistical Analysis

Statistical analyses were performed in GraphPad Prism. All behavioural data were analyzed with an unpaired two-sided t-test, Mann-Whitney U-test, one-way ANOVA or Kruskal-Wallis H-test followed by a posthoc Tukey’s, Dunnet’s or Dunn’s multiple comparisons test. No statistical methods were used to predetermine sample size. Sample sizes were similar to other publications in the field.

Blinding, Randomization

The experiments were randomized with appropriate controls present in each independent experiment. All genotypes tested and analyzed were self-blinded to the experimenter.

Neuroanatomy, Connectivity, Dendrograms

Neuromorphological calculations and connectivity analyses were performed, and dendrograms calculated & plotted, with scripts based on NAVis 1.2.1 library functions in Python 3.8.8 (https://pypi.org/project/navis/; https://github.com/navis-org/navis)⁸⁶ and data from the Drosophila hemibrain (v.1.2.1) (https://neuprint.janelia.org)^87,88. All neuronal skeletons were healed (navis.heal_skeleton (method = “ALL”, max_dist = “100 nanometer”, min_size = 10)), rerooted (navis.reroot_skeleton (x.soma)) and strongly down sampled with conserved connectors (navis.downsample_neuron(downsampling_factor = 1000, preserve_nodes = ‘connectors’)). 3D representations of neurons shaded by Strahler order were generated with navis.plot2d (method=’3d’, shade_by=’strahler_index’), after pruning twigs with Strahler order ≤1 (navis.prune_by_strahler()). Where applicable only branches in specific volumes were considered (navis.in_volume()). Volumes were obtained from neuprint (v.1.2.1) with fetch_roi(). Connectivity was analyzed using unpruned neurons and with compartment specificity (navis.in_volume()).

Custom scripts based on navis.plot_flat() were used to generate dendrograms of DPM and APL neurons with twigs of Strahler order ≤1 pruned. MB compartment boundaries were defined by connectivity to DANs of the respective compartments. Branches outside the γ-lobe were downsized manually to increase visibility of γ-lobe compartments. Synapses are filtered by in_volume() and displayed on branches with Strahler order >1. Connectivity statistics are based on unpruned neurons and synapses between neurons were obtained with R based natverse:: neuprint_get_synapses() (https://natverse.org)⁸⁶ scripts and processed with custom scripts in Python.

Figure supplements

Figure 1 – figure supplement 1.

Figure 2 – figure supplement 1.

Figure 4 – figure supplement 1.

Figure 5 – figure supplement 1.

Data availability

Supplementary Table 1 and Supplementary Table 2 and data supporting the findings of this study are available at https://github.com/CNCBWaddellTmaze/multisensory-memory-Drosophila. Dataset of Drosophila hemibrain used for Methods section “Neuroanatomy, Connectivity, Dendrograms” is publicly available at https://neuprint.janelia.org. Code used for Methods section “Neuroanatomy, Connectivity, Dendrograms” is publicly available at https://pypi.org/project/navis/, https://github.com/navis-org/navis and https://natverse.org.

Acknowledgements

We thank R. Brain, K. delos Santos, R. Busby, M. Moreno Gasulla, J-P. Moszynski, F. Woods for technical assistance. P. Jacob contributed to behavioural experiments. We thank G. Rubin, FlyLight, B. Dickson, A. Lin and the Vienna Drosophila Resource Center and the Bloomington Stock Center for flies. Z.O and A.A.C were funded by EMBO Long-term postdoctoral fellowships (ALTF 311-2017 and ALTF 1300-2024). P.V-G. was funded by the Sloane Robinson/Clarendon Scholarship given by the University’s Clarendon Fund scheme and Keble College and from the Consejo Nacional de Ciencia y Tecnología (CONACYT). S.W. was funded by a Wellcome Principal Research Fellowship (200846), a Wellcome Discovery Award (225192), an ERC Advanced Grant (789274), and Wellcome Collaborative Awards (203261 and 209235).

Additional information

Author contributions

Designed research ZO, CT, AAC, AM, MK, SW, Performed research ZO, NO, AAC, AM, MK, CS, KD, PVG, Analyzed data ZO, NO, CS, KD, AAC, AM, MK, SW Resources SW Writing SW, ZO, NO, AAC, Supervision SW, Funding Acquisition SW, ZO, AAC.

Funding

Wellcome Trust (WT)

https://doi.org/10.35802/200846

• Scott Waddell

Wellcome Trust (WT)

https://doi.org/10.35802/225192

• Scott Waddell

Wellcome Trust (WT) (203261)

• Scott Waddell

Wellcome Trust (WT) (209235)

• Scott Waddell

EC | European Research Council (ERC)

https://doi.org/10.3030/789274

• Scott Waddell

European Molecular Biology Organization (EMBO) (ALTF 311-2017)

• Zeynep Okray

European Molecular Biology Organization (EMBO) (ALTF 1300-2024)

• Anna A Cook

Consejo Nacional de Ciencia y Tecnología, Paraguay (CONACYT)

• Paola Vargas-Gutierrez

Additional files

Supplementary Table 1.

Supplementary Table 2.

Supplementary Data.