Abstract
Feeding decisions are fundamental to survival, and decision making is often disrupted in disease1,2, yet the neuronal and molecular mechanisms of adaptive decision making are not well understood. Here we show that the neural activity in a small population of neurons projecting to the fan-shaped body in the central brain of Drosophila represents food choice during sensory conflict. We found that hungry flies made tradeoffs between appetitive and aversive values of food in a decision making task to choose unpalatable bittersweet food with high sucrose concentration over sucrose-only food with less sucrose. Using cell-specific optogenetics and receptor RNAi knockdown during the decision task, we identified an upstream neuropeptidergic and dopaminergic network that likely relays internal state and other decision relevant information, like valence and previous experience, to the fan-shaped body. Importantly, calcium imaging revealed that these fan-shaped body neurons were strongly inhibited by rejected food choice, suggesting that this neural activity is a representation of behavioral choice. FB response to food choice is modulated by taste, previous experience, and hunger state, which the fan-shaped body neurons likely integrate to encode choice before relaying decision information to downstream motor circuits for behavioral implementation. Our results uncover a neural substrate for choice encoding in a genetically tractable model to enable mechanistic dissection of decision making at neuronal, cellular, and molecular levels.
Main
Animals integrate food-related sensory information from their external environment with their internal state in order to make adaptive decisions. Often food-related sensory information is conflicting in valence. For example, Drosophila flies forage on decomposing fruits and, when hungry, must balance obtaining essential nutrition with avoiding toxins, pathogens, etc. As flies forage, sweet and bitter taste receptors on their legs and wings signal the presence of sweet nutritive food and bitter potential toxins3. Flies must adaptively weigh and integrate this conflicting information before consumption to enhance reproductive success. We investigated how value-based decisions are made in the brain of a hungry fly using an experimental paradigm in which freely walking flies sample and choose between different sweet-only and bittersweet foods (Fig. 1a). We quantified food choice and manipulated subsets of neurons while flies engaged in this decision task with conflicting taste information (Fig. 1a).
Hungry flies make tradeoffs when faced with conflicting sensory information
We tested wild-type flies deprived of food for different durations over a range of increasing concentration of sweet-only (sucrose) option against a constant bittersweet (sucrose + quinine) option. When choosing between a low sucrose concentration sweet-only option and a high sucrose concentration bittersweet option, flies prefer higher sucrose bittersweet (Fig. 1b). As sucrose concentration of the sweet-only choice increased, flies increasingly preferred it over bittersweet. This dose-dependent change in preference suggests that at higher sucrose concentrations of sweet-only option the caloric advantage in choosing a less palatable bittersweet food was lost (Fig. 1b). In the absence of bitter, flies always chose the sweeter option (Extended Data Fig. 1a). Flies equally preferred sweet-only and bittersweet option at 10-fold sucrose concentration difference (Fig. 1b, 50 mM vs. 500 mM sucrose+1 mM quinine). This equal-preference point was identical at all of the tested food deprivation durations (Fig. 1b). The equal-preference point depends on the sucrose concentration ratio between the two options and not absolute concentration (Extended Data Fig. 1b), indicating that there was no saturation of taste sensation at the concentrations used. These results indicate that hungry flies tradeoff the appetitive (sweet) and aversive (bitter) values of food in making feeding decisions.
To further understand decision making behavior, we also recorded location of flies at the end of decision task, and as expected, position preference mirrored ingested food preference (Fig. 1c). Social interaction between animals can have effects on decision making. To keep the task similar to fly’s natural social environment, we used random proportions of males and females per trial. There was no effect of male-to-female ratio on ingested food preference (Fig. 1d). Previous studies have shown that group size can affect Drosophila behavior4,5. At equal-preference condition (21h deprivation, 50 mM sucrose), food preference and group size (Fig. 1e), food preference and percent of flies that ate (Fig. 1f), and percent of flies that ate and group size (Fig. 1g) were not correlated. There was also no significant prediction of preference index by group size or percent of flies that ate in a multiple regression model (Fig. 1h, Extended Table 1), indicating no interaction between these variables in the decision task.
A decision making neuronal ensemble converges on the fan-shaped body
During foraging, animals compute value estimates of internal hunger state and external sensory environment such valence of available foods, location of food, etc. Various neuromodulators regulate hunger dependent food intake6-12, reward13-16 or punishment17, as well as memory14,16,18. The mushroom body is an insect central brain region involved in gustatory learning and memory19,20 and valence encoding21, and is thought to be a major center controlling higher-order behaviors22-24. The insect central complex is an evolutionarily conserved central brain region whose ellipsoid body and protocerebral bridge sub-regions have been implicated in navigation25-32 and sleep33-35. The central complex fan-shaped body, a laminar neural sub-region, has been implicated in sleep36-39 and ethanol preference40,41. The fan-shaped body was particularly interesting to us because several neuromodulators42, their receptors37,43,44, as well as dopaminergic inputs45,46 co-localize in its layers. We hypothesized that value estimates of internal state and external environment from modulatory neurons will be required for integration by higher brain regions for decision making. To test this, we manipulated genetically targeted cell-specific neural expression using GAL4-UAS binary expression system47. We acutely optogenetically activated subsets of neurons using CsChrimson channelrhodopsin48 while flies actively sampled and consumed food at the equal-preference condition (Fig. 1a, 1b, 21h food deprivation, 50 mM sucrose vs 500mM sucrose+1mM quinine). It is not only activation of neurons but also inhibition that can modulate behavior. Therefore, for the next part of the screen, we optogenetically inhibited select genotypes from activation screen, using the anion-conducting channelrhodopsin GtACR149. Genotypes were selected for inhibition screen based on the following pre-defined rules: a genotype with preference index lower than -0.3, or higher than 0.3, or a genotype with change in feeding during activation. This optogenetic interrogation of modulatory neurons and higher order brain regions revealed neuropeptidergic neurons (Leucokinin, Allatostatin A, NPF, DH44), subsets of dopaminergic neurons, and a narrow subset of fan-shaped body layer 6 neurons (FBl6) whose activation or inhibition significantly shifted food choice in the equal-preference condition (Fig. 2a).
Activation of Leucokinin (Lk) neurons suppressed feeding in food deprived flies (Fig. 2a, 2b left panel, Extended Fig. 2a-b), suggesting that Lk may relay metabolic state information. To confirm that Lk secreted by these neurons was the molecular basis of this feeding suppression, we simultaneously knocked down Lk expression with genetically encoded RNAi while optogenetically activating Lk neurons. The majority of flies consumed food during simultaneous Lk RNAi and activation, while almost no flies consumed when Lk neurons were activated without Lk RNAi. This indicates that Lk secretion mediates feeding suppression by Lk neurons (Fig. 2b left panel, Extended Fig. 2a-b). Optogenetic silencing of Lk neurons shifted preference towards bittersweet (Fig. 2a, 2b left panel, Extended Fig. 2a-b). Feeding suppression on Lk neuron activation implies a decrease in perceived hunger level of food deprived flies, which is consistent with implied increased perceived hunger level on Lk neuron inhibition leading to increased preference for high sucrose bittersweet food. Activation of Allatostatin A (AstA) neurons shifted the preference towards sweet, while inhibition shifted the preference towards bittersweet (Fig. 2a, 2c left panel). We confirmed that AstA was the molecular basis of this shift in preference by simultaneous AstA RNAi knockdown and activation of AstA neurons (Fig. 2c left panel). Flies preferred sweet on activation of NPF neurons, and this shift was abolished by simultaneous activation and NPF RNAi knockdown (Fig. 2a, 2d left panel). Activation of DH44 neurons had no significant effect, but inhibition shifted the preference towards bittersweet food (Fig. 2a, 2e left). Dopaminergic subsets involved in aversive memory (Fig. 2a PPL1 γ2α’1)50, taste conditioning (Fig. 2a PPL1 α3)18, and long-term memory (Fig. 2a PAM α1)14 also affected food choice. Activation of these dopaminergic subsets shifted the preference toward bittersweet (Fig. 2a). Activation of neurons from different mushroom body lobes, a brain region controlling higher-order behaviors, had no effect on preference. However, inhibition of a specific subset of fan-shaped body neurons, FBl6, shifted preference toward bittersweet (Fig. 2a, 3a left panel). Value estimates of internal state and external sensory environment, which are likely computed by modulatory neurons, are crucial for decision making. Fan-shaped body has co-localization of several neuromodulators and their receptors37,42-46 and likely integrates the value estimates it receives from modulatory neurons.
To determine whether the neurons we identified in this optogenetic screen are connected in a behaviorally relevant ensemble, we employed a chemoconnectomics approach51 exploring cell-specific genetically encoded RNAi knockdown of neuropeptide and dopamine receptors. Knockdown of neuropeptide or dopamine receptors in Lk neurons did not shift preference (Fig. 2b right panel), implying that Lk neurons receive food preference and hunger related information from other neurons. Dopaminergic Dop1R1 receptor RNAi in AstA neurons shifted preference towards sweet (Fig. 2c right panel), suggesting that AstA neurons receive food preference related dopaminergic inputs. Lkr and Dop1R1 receptor RNAi in NPF neurons shifted preference toward sweet (Fig. 2d right panel), suggesting that NPF neurons receive food preference relevant Lk and dopaminergic inputs. DopEcR receptor RNAi in DH44 neurons shifted preference toward bittersweet (Fig. 2e right panel), suggesting that DH44 neurons receive food preference relevant dopaminergic inputs.
Importantly, RNAi knockdown of Lkr, AstA-R1, or DH44-R1 receptors in FBl6 neurons shifted the preference toward bittersweet (Fig. 3a right panel), indicating that FBl6 neurons are modulated by these three neuropeptides to affect food choice. Furthermore, change in food preference on receptor RNAi in FBl6 mirrors change in food preference on respective neuropeptide neuron inhibition. For example, AstA-R1 receptor RNAi in FBl6 neurons should inhibit AstA input to FBl6, that is, have the effect that is equivalent of inhibiting AstA neurons. Consistently, both AstA-R1 receptor RNAi in FBl6 neurons (Fig. 3a right panel), and AstA neuron inhibition (Fig. 2c left panel) shifted food preference towards bittersweet. Similarly, both DH44-R1 receptor RNAi in FBl6 (Fig. 3a right panel) and DH44 neuron inhibition (Fig. 2d right panel) shifted preference towards bittersweet. Both Lkr receptor RNAi in FBl6 (Fig. 3a right panel), and Lk neuron inhibition (Fig. 2b right panel) also shifted preference towards bittersweet. RNAi of dopamine receptors in FBl6 had no effect (Fig. 3a right panel). This matrixed strategy mapped the neuromodulatory connections between nodes in the ensemble to control choice, and uncovered a previously unknown convergence node (FBl6) that is well positioned to integrate sensory, metabolic, and experiential information for decision making.
Fan-shaped body neurons encode choice
Value estimates of internal state like degree of hunger, and external environment like appetitive or aversive value of food (valence) and past experience, are integrated and transformed into choice. This raises the question of whether FBl6 neurons compute value estimates or integrate these estimates to encode choice. If FBl6 neurons estimated value of or encoded metabolic parameters such as hunger or satiety, manipulating their activity would be expected to influence feeding behavior. During FBl6 neural manipulation, majority of food deprived flies consumed food while majority of fed flies did not (Fig. 3b, Extended Table 1), demonstrating that hunger state is not affected by FBl6 neural activity. There was no significant difference in total amount of food consumed by flies during FBl6 neural manipulation compared to control flies (Fig. 3c, Extended Table 1). There was also no significant difference in the amount of sweet versus bittersweet food consumed per fly during FBl6 neural manipulation compared to controls (Fig. 3d, Extended Table 1). The shift in food preference during FBl6 inhibition (Fig. 3a left panel) was due to larger number of flies preferring to consume bittersweet over sweet food rather than each fly consuming larger quantity bittersweet food. Taken together, these results demonstrate that FBl6 does not encode or affect metabolic signals of hunger or satiety.
Next, we asked if activity of FBl6 neurons was inherently rewarding or aversive, that is, had inherent valence, which could shift food preference. To test this, we quantified place preference for illuminated versus dark parts of fly arena without food, during optogenetic manipulation of FBl6 neurons. FBl6 neural manipulation had no effect on preference for illuminated versus dark parts (Fig. 3e), demonstrating that FBl6 activation or inhibition is neither inherently rewarding nor aversive.
Animals accumulate past experience to inform future decisions. We hypothesized that FBl6 integrates hunger and food-related value estimates with experiential information for decision making. To understand how past experience affects FBl6 activity, we recorded FBl6 neural activity in flies that had different food-related experiences. Flies were presented taste stimuli from the equal-preference condition (Fig. 4a) while ratiometric Ca2+ activity in FBl6 was measured using GCaMP6f52 and tdTomato (Fig. 4a, b). First, we tested the effect of hunger on FBl6 neural activity in naïve flies, i.e., flies that had not experienced the decision task at all. FBl6 neurons of naïve food-deprived flies were strongly inhibited by the bittersweet stimulus, but not sweet (Fig. 4d-e, naïve deprived). Flies often find bittersweet food aversive20 and inhibition of naïve fly FBl6 neural activity in response to bittersweet stimulus may be a representation of rejected choice. Consistently, if FBl6 activity inhibition represents rejected food choice then neural activity in naïve fed flies should be strongly inhibited by both sweet and bittersweet stimuli because fed flies reject both foods in decision task (Fig. 3b). Indeed, FBl6 neurons of naïve fed flies showed strong inhibitory response to both bittersweet and sweet stimuli (Fig. 4d-e, naïve fed).
Next, we asked, if similar to naïve flies, FBl6 neural activity also represents behavioral choice in flies that experienced the decision task and made different food choices. FBl6 neurons of flies that chose sweet food were strongly inhibited by rejected bittersweet stimulus but not by chosen sweet (Fig. 4d-e, chose sweet). Correspondingly, FBl6 neurons of flies that chose bittersweet food were strongly inhibited by rejected sweet stimulus but not by chosen bittersweet (Fig. 4d-e, chose bittersweet). FBl6 neurons of flies that chose neither food, i.e. rejected both, were strongly inhibited by both bittersweet and sweet stimuli (Fig. 4d-e, chose neither). Overall, FBl6 neural activity is always strongly inhibited by food that a fly rejects, demonstrating that suppression of FBl6 activity is the neural representation of behavioral food choice. This neural representation is modulated by taste (sweet vs. bittersweet), previous experience (naïve vs. experience with two-choice conflict), as well as hunger state (naïve food deprived vs. fed) (Fig. 4d-e). FBl6 neurons likely receive these different types of information directly through AstA, DH44, and Lk receptor signaling, and indirectly through NPF and dopamine pathways of the decision ensemble, for integrating them to form a representation of choice before sending information to downstream motor circuits for decision implementation (Fig. 4f).
Discussion
Animals make decisions about which foods to consume by integrating their internal physiological state with external sensory cues. Here we delineated a neuronal ensemble in Drosophila that underlies food-related decision making during sensory conflict between sweet and bittersweet food choices (Fig. 4f). Activating or silencing particular nodes in this ensemble shifts the decision balance between sweet and bittersweet food (Fig. 2b-e, 3a). This ensemble convergences on to FBl6 and FBl6 neurons likely integrate information from the upstream modulatory network to transform it into the neural representation of food choice (Fig. 4f).
Organisms must assess and assign value estimates to their external environment and internal state before integrating these estimates for adaptive decision making. Neuromodulatory subsets in the decision ensemble that we have identified have roles in hunger dependent food intake behavior, reward, valence, and memory. These modulatory neurons are well positioned to estimate value of the sensory environment and internal hunger state. For example, AstA neuron activation shifts food preference from carbohydrates to protein6, while DH44 neurons sense sugars53 and amino acids11. AstA and DH44 neurons may, therefore, convey food quality information to FBl6. NPF neuron activation is inherently rewarding54 and may convey food valence information. Lk neurons have been implicated in nutrient sensing9 and their activation suppresses feeding in food deprived flies (Fig. 2a-b), suggesting that internal metabolic state information may reach FBl6 through Lk/Lkr signaling. Dopaminergic subsets involved in aversive memory (PPL1 γ2α’1)50, taste conditioning (PPL1 α3)18, and memory (PAM α1)14 also affected food choice (Fig. 2a) and may provide an error signal for predicting and updating value estimates similar to primate dopaminergic ventral tegmental area55. FBl6 neurons have axonal projections in the fan-shaped body45,46, dense dendritic projections in the superior medial protocerebrum (SMP), and sparse dendritic projections in superior intermediate protocerebrum (SIP) and superior lateral protocerebrum (SLP)45,46 (Fig. 4g). In these higher brain regions, FBl6 receives synaptic inputs from dopaminergic neurons38,39 that regulate sleep38,39 and ethanol preference41. Interestingly, direct dopaminergic input to FBl6 through dopamine receptors did not influence food choice (Fig. 3a). Instead indirect dopaminergic inputs conveyed by neuromodulatory neurons regulated food choice (Fig. 2b-e). Mammalian studies provide converging evidence on multiple interconnected networks in frontal cortex and basal ganglia that compute and store value estimates of sensory environment and motor events in that environment required for decision making55,56. The neural ensemble described in this study has a similar framework of interconnected networks that potentially store, compute, and update value estimates for decision making.
A value integrator for food-related decision making requires estimates of taste identity, previous experience, and hunger state. FBl6 neuron activity is modulated by these parameters but it is not yet clear how the information brought to FBl6 from upstream network is integrated and transformed into the representation of behavioral choice before it is sent to downstream motor neurons for decision implementation (Fig. 4f). Decision making theories in mammals have traditionally focused on how values are represented in the brain55,56, but how the brain integrates value information to make decisions when competing alternatives are present is still unclear57. Future investigations are required to identify and test proposed hypotheses of specific role of each node in the ensemble, how inputs from different nodes are integrated in FBl6, how this integration is transformed into the representation of choice, and which downstream motor circuits are involved in decision implementation.
Author contributions
P.S. conceptualized the study, designed and performed experiments, and analyzed data; L.Y.M. contributed to experimental design and fly dissection; P.S. and M.N.N. interpreted data; P.S. and M.N.N. prepared the manuscript with inputs from all authors.
Methods
Fly husbandry
Flies were cultured on standard cornmeal medium on 12:12 light:dark cycle at 25°C. w1118 lab stock was used as wild type. ss00208 and ss00225 unpublished split-GAL4 lines were a generous gift from Gerry Rubin. All genotypes and their sources are described in Extended Table 2. 2-5 day old flies were wet starved for 2-48 h (based on experiment design) on a wet Kimwipe with 1.5 ml distilled water. For optogenetic experiments, flies were food deprived for 21 h before testing on 0.4 mM all-trans Retinal (Cayman Chemicals) in 1% agar. Flies for RNAi knockdown and their controls were moved to 28°C for 21 h the day before testing, i.e., during food deprivation, to induce strong RNAi. RNAi control was created for each GAL4 line by crossing the respective GAL4 to UAS-Valium (see Extended Table 2). All RNAi lines that we used were from Harvard TRiP project1,2 and have been validated by independent groups (see Extended Table 2). Flies for simultaneous optogenetic and RNAi experiments were created using the genotypes mentioned in Extended Table 2. All experiments were conducted at Zeitgeber Time 3-6.
Two-choice assay and optogenetics
Sweet foods were made with different concentrations of sucrose and bittersweet foods with 500mM sucrose (Sigma) and 1mM quinine (Alfa Aesar or Beantown Chemicals, CAS#207671-44-1) dissolved in 1% agarose (AmericanBio) made in distilled water. 0.04% w/v red dye (Sulforhodamine B, MP Biochemicals, CAS# 3520-42-1) and 0.02% w/v blue dye (Erioglaucine A, Alfa Aesar, CAS# 3844-45-9) were used for food coloring. Dye colors were alternated between sweet and bittersweet foods for each condition and there was no preference for one dye over the other at the concentrations used. Fly arenas were prepared by pouring agarose based foods in two-compartment petri-dishes (90-100 mm diameter) from Kord Valmark, EMS, or Fisher Scientific. Because of a thin physical barrier between the two compartments in the arena there was no diffusion between the two foods. Groups of 20-35 flies were aspirated and introduced into the arena 5-10 sec before the start of the experiment. All experiments were conducted in dark so that there was no effect of food color on preference. Arenas were placed on a platform with IR backlight for video recording using a Flea Pointgrey camera (FL3-U3-20E4C/M) at 15 fps. For optogenetics, we used high-power LEDs (Luxeon) placed adjacent to backlight IR LEDs (based on Janelia ID&F design) of 627 nm (for CsChrimson) and 520 nm (for GtACR1) that were controlled using Arduino Uno. For optogenetic screen, both red and green lights were pulsed at 100% max intensity, 50Hz, 25% duty cycle. For follow up experiments, CsChrimson experiments were conducted at 25% max intensity; GtACR1 follow up was done at screen condition. Light was pulsed for the entire duration of the experiment. At the end of the experiment, flies were anaesthetized using CO2 and their belly color was recorded under a dissection microscope. Preference index (PI) was calculated as (no. of sweet food flies + 0.5 no. of both food flies) - (no. of bittersweet food flies + 0.5 no. of both food flies) / no. of total flies that ate, where negative PI would mean that more number of flies ate bittersweet food.
Food intake quantification
Food intake was quantified using spectrophotometry as previously described3,4. After recording belly color flies were frozen in 1.5 ml Eppendorf tubes at -20°C until intake quantification (1-2 days). Flies from each trial were separately homogenized in distilled water (5 µl /fly) using a motorized pestle (BT Labsystems, BT703) for 1.5 min and centrifuged at 13000 rpm for 5 min. Absorbance of the debris-cleared 2 µl supernatant was measured on NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific) at 565 nm (for red dye) and 630nm (for blue dye). Flies that ate uncolored 1% agarose with 50mM sucrose were used as blank for baseline control. Red and blue dye concentrations were interpolated using their respective standard curves (GraphPad Prism) acquired from serial dilutions of single dyes in distilled water. Since we knew the number of flies that ate each color per trial, we could calculate per fly blue and red concentrations in the same solution.
Calcium imaging and data analysis
3-5 day old flies (naïve or after two-choice assay) were aspirated and positioned in a custom made fly holder in which they were glued using two-part transparent epoxy (Devcon). Only the top of fly head (for imaging) and the forelegs (for taste delivery) were outside the holder, while the rest of the fly, including proboscis were restrained in the fly holder. No anesthesia was used. A small piece of head cuticle was dissected and air sacs removed using a 30-gauge syringe needle and fine forceps, immediately followed by sealing the head capsule with a translucent surgical silicone adhesive (Kwil-Sil, WPI). Dissected fly was then placed in a humidified chamber for 15 min recovery before imaging.
Calcium imaging was performed on a Zeiss Axio Examiner upright microscope with 20x air objective and a Colibri module for LED control. tdTomato was excited at 555 nm (80% intensity) and GCaMP6 at 470 nm (100% intensity). An image splitter (Photometrics DV-2) was used to split red and green channels and acquire simultaneous images for tdTomato and GCaMP6 using a Hamamatsu ORCA-R2 C10600 camera. Images were acquired at 5 fps with variable baseline (3 to 10 s required for stable tastant delivery) followed by tastant application to the forelegs, using a syringe, for 3 s and 4 s of no tastant. Excess tastant was wicked from the forelegs using absorbent tissue paper between each application. 10 s inter-trial interval was used during which all lights were off. Water was always applied first, followed by either sweet or bittersweet tastant. Sequence of sweet and bittersweet was alternated between flies. Sweet: 50 mM sucrose in distilled water; bittersweet: 500 mM sucrose + 1 mM quinine in distilled water. For flies that chose sweet in the two-choice behavior assay only trials with sweet as the first tastant were averaged and for flies that chose bittersweet, only trials with bittersweet as the first tastant were averaged.
Pixel intensities were extracted in Fiji followed by data analysis in MATLAB, both using custom written code. After background subtraction using Fiji’s rolling-ball method (20 px), ROIs were manually drawn and saved on the tdTomato image, and superimposed on the GCaMP image (both reporters were expressed in the same neurons using the same driver) for mean ROI pixel intensity extraction. The saved intensity signals were then analyzed in MATLAB. tdTomato and GCaMP traces were individually corrected for photobleaching by fitting a single exponential function. Corrected GCaMP trace was then divided by the corrected tdTomato trace to obtain the ratiometric fluorescence trace (R). For relative fluorescence fold change (ΔR/R0) determination, baseline fluorescence (R0) was calculated by averaging R over 2 s preceding tastant application. Peak ΔR/R0 was calculated during 4 s following tastant application.
EM reconstruction
Electron microscopy images were reconstructed from publically available Janelia FlyEM hemibrain data using neuPRINT5. Neuron identities were confirmed in NeuronBridge6 by cross-referencing EM traced FBl6 neurons matched with light microscopy images of 84C10-GAL4 from FlyLight7. FBl6 mesh, whole FB mesh, and SMP, SIP and SLP brain region meshes were used to depict brain regions with neural projection areas.
Statistics
All data were plotted in either Python or MATLAB using custom written code. Statistics were carried out in GraphPad Prism. If all data passed Kolmogorov-Smirnov normality test, ANOVA was conducted, otherwise Kruskal-Wallis test was conducted, both followed by appropriate post-hoc tests. Details of statistics for each figure are provided in Extended Table 1. Sample sizes are reported in parentheses next to dataset name in Extended Table 1.
Extended figure legends
Acknowledgements
We would like to thank Peter Niesman for help with calcium imaging data collection and Jason Braco for helpful discussions. Yichen Luo from John Carlson’s lab provided useful information about taste delivery imaging rig. Tanya Wolff generously provided insights into fan-shaped body neuroanatomy. We would also like to thank Gerry Rubin and Tanya Wolff for sharing unpublished fly lines ss00208 and ss00225. These studies were supported in part by the National Institute of Neurological Disorder and Stroke, National Institutes of Health (NIH) (RO1NS091070) and the National Institute of General Medical Sciences, NIH (RO1GM098932).