Central relay of bitter taste to the protocerebrum by peptidergic interneurons in the Drosophila brain

Bitter is a taste modality associated with toxic substances evoking aversive behaviour in most animals, and the valence of different taste modalities is conserved between mammals and Drosophila. Despite knowledge gathered in the past on the peripheral perception of taste, little is known about the identity of taste interneurons in the brain. Here we show that hugin neuropeptide-containing neurons in the Drosophila larval brain are necessary for avoidance behaviour to caffeine, and when activated, result in cessation of feeding and mediates a bitter taste signal within the brain. Hugin neuropeptide-containing neurons project to the neurosecretory region of the protocerebrum and functional imaging demonstrates that these neurons are activated by bitter stimuli and by activation of bitter sensory receptor neurons. We propose that hugin neurons projecting to the protocerebrum act as gustatory interneurons relaying bitter taste information to higher brain centres in Drosophila larvae.

D etailed knowledge exists on the anatomical distribution and function of gustatory receptors in mammals and Drosophila [1][2][3][4][5] . In Drosophila 60 gustatory receptor genes encode 68 gustatory receptors [6][7][8] , with the majority of these receptors detecting bitter compounds 9 . Although gustatory receptors in Drosophila share no homology to mammalian taste receptors, the strategy used in both to detect a taste molecule, process its information and the valence of aversive bitter and appetitive sweet stimuli share similarities 4 . In contrast to the extensive knowledge on the peripheral coding of taste in flies and mammals, much less is known about the central pathways that relay and translate these signals into meaningful behaviour. Although broad regions in different parts of the brain have been shown to respond to various taste cues, there is little information on the molecular identity of specific neurons that convey different taste modalities to the higher brain centres 10,11 . Recently, secondary neurons that relay sweet taste from subesophageal zone (SEZ) to the antennal mechanosensory motor centre of adult Drosophila were characterized 12 . Analogous secondary neurons for other taste modalities have not yet been identified.
A candidate for conveying bitter taste from the SEZ to higher brain centres are neurons that express the hugin neuropeptide 13,14 (referred to as hugin neurons), whose arborizations in Drosophila larvae overlap with that of bitter gustatory receptor neurons (GRNs) expressing the caffeine receptor GR66a [15][16][17] . In adult Drosophila, GR66a was shown to represent a bitter receptor for detection of caffeine 9,16,17 and inactivation of GR66a positive neurons leads to impairment of caffeine aversion 17   larvae, artificial activation of GR66a positive neurons leads to aversive behaviour 18 . Thus, hugin neurons were good candidates for acting as a central relay for bitter information from sensory neurons. Using classical two-choice behavioural experiments, electrophysiological measurements as well as calcium imaging analysis, we now show that hugin neurons relay caffeine as well as other bitter taste signals from sensory neurons to the protocerebrum, acting as bitter taste interneurons in the Drosophila larval brain.

Results
Hugin neurons are required for caffeine avoidance response. We first asked whether the hugin neurons make contacts with caffeine responsive GR66a neurons. Using the GFP reconstitution across synaptic partners (GRASP) approach 19 , we could indeed observe a GRASP signal in the SEZ, indicating that caffeine receptor neurons and hugin neurons are in close proximity to each other (Fig. 1a,b).
Activating the hugin neurons causes the larvae to stop feeding and move out of a strongly appetitive food source (yeast) 20 , which could be due to activation of an aversive bitter taste pathway. We therefore tested the behavioural response to caffeine in a two choice assay. When hugin neurons are activated with the temperature sensitive cation channel dTrpA1, the animals showed impaired aversion to caffeine stimuli (Fig. 1c). We next asked how the animals would behave if hugin neuronal activity was suppressed, by ablating the hugin neurons. These animals showed significantly less avoidance to caffeine (Fig. 1d,e). To exclude potential developmental effects, we also expressed the temperature sensitive mutant form of dynamin (shibire TS ) in the hugin neurons, which leads to a block of synaptic release in a  temperature dependent manner. The loss of proper bitter aversion still persisted (Fig. 1d,f). We next asked if hugin neurons were involved in other taste modalities. We first tested high salt (2 M NaCl), which is very aversive for larvae 21 , in two-choice assays. As with caffeine, activation of the hugin neurons led to animals showing impaired aversion to high salt (Fig. 2a). However, no difference to control was observed when hugin neurons were ablated, indicating that hugin neurons are not required for the behavioural response to high salt (Fig. 2b). High caffeine and salt levels are both aversive gustatory stimuli. To determine how manipulation of hugin neuronal activity affected the response to an appetitive cue, we performed two-choice assays with 1 M fructose. In this case, activation led to impaired attraction to fructose, whereas ablation had no effect relative to control (Fig. 2c,d; Supplementary Fig. 1). Thus, when hugin neurons are activated, the larvae are no longer able to respond properly to different taste modalities, such as yeast protein 20 , fructose, high salt and caffeine. This chemosensory response was specific for taste, as olfactory behaviour was not affected (Fig. 3), showing that activation of hugin neurons selectively disrupts gustatory chemosensory processing. However, inhibiting the hugin neurons results in the inability to respond appropriately to caffeine, indicating that hugin neurons process bitter taste information. In sum, when hugin neurons are activated, larvae perceive all tested substrates (water, caffeine, salt and fructose) as bitter tasting. In contrast, when hugin neurons are inactivated, only the bitter substrate choice is affected since proper detection of bitter compounds no longer occurs.
Distinct hugin neurons modulate taste and feeding behaviour.
As the hugin neuronal cluster is composed of different classes with different projection targets 13,14 , we next wanted to determine which class was responsible for taste processing and feeding regulation. Through promoter deletion analysis, we generated a hugin promoter-Gal4 line which showed target gene expression exclusively in the eight hugin cells (four per hemisphere) that project to the protocerebrum (HugPC-Gal4, Fig. 4a). To see whether this subset of hugin neurons (huginPC neurons) is necessary for proper bitter taste processing, we ablated these neurons (Fig. 4b) and tested larvae in the caffeine two-choice experiment. Ablation of the huginPC neurons resulted in the inability to appropriately avoid caffeine as compared with control animals (Fig. 4c). Consistent with the ablation of all hugin neurons ( Fig. 2), ablating just the huginPC neurons had no effect compared with controls on high salt or fructose (Fig. 4d,e, controls are re-plotted from Fig. 2b,d). We note that, in the case of 2 M NaCl, there was a significant difference to one of the controls (UAS-rpr;;hid control larvae). Therefore, we additionally tested the animals on a lower, but still aversive salt concentration (500 mM), which showed that huginPC neurons are not necessary for proper high salt aversion ( Supplementary Fig. 2, for descriptive statistics see Supplementary Table 7). Activating just the huginPC neurons was sufficient to suppress food intake, induce wandering-like behaviour ( Fig. 5a; Supplementary Fig. 3), as well as decrease the motor pattern of pharyngeal muscle contractions (Fig. 5b). These results indicated that the hugin neurons responsible for feeding and caffeine taste mediated behaviours are those that project to the protocerebrum.
Activity of huginPC neurons upon bitter stimulation. To further investigate the connection between bitter taste and hugin neurons, we asked if the huginPC neurons could be activated by caffeine using CaMPARI (Calcium Modulated Photoactivatable Ratiometric Integrator), that allows monitoring of calcium activity in intact animals. Calcium activity and simultaneous presence of ultraviolet-light (405 nm) lead to an irreversible conversion from green to red fluorescence of the neurons of interest 22 . When we placed intact larvae in solutions containing water and water mixed with caffeine, fructose, high NaCl or yeast (Fig. 6a), huginPC neurons were strongly activated by caffeine (Fig. 6b,c). They also showed a concentration dependent increase in calcium activity with increasing caffeine concentrations (Fig. 6d). Other bitter tastants, such as quinine and denatonium, also activated huginPC neurons (Fig. 6e). We tested the behavioural relevance for this rise in calcium activity for denatonium in two-choice experiments, since denatonium led to robust avoidance behaviour in control larvae at 32°C (experimental temperature). The calcium activity in huginPC neurons was behaviourally relevant since ablation of these neurons resulted in larvae with impaired aversion to the taste of denatonium (Fig. 6f) (no difference when tested against chance levels, see Supplementary Table 5). Unexpectedly, we observed decrease in huginPC calcium activity in larvae placed in high salt, fructose and yeast relative to water alone (Fig. 6c). Although the functional significance of this repression is not clear, it has been shown that bitter taste pathways can inhibit sweet pathways in adult 23 and larval 24 Drosophila, reflecting an interaction between pathways involving different taste modalities. This crossregulation might be a general strategy to efficiently detect distinct taste modalities, since in case of the huginPC neurons, a sweet taste stimulus would inhibit the activity of bitter interneurons.
Finally we asked if there is a functional connection, in addition to the anatomical connection (Fig. 1a,b), between the caffeine sensing GR66a neurons and hugin neurons. To this end, we activated the GR66a neurons and then monitored the activity of hugin neurons using the calcium indicator GCaMP6s (Fig. 7a,b). Activation of GR66a by dTrpA1 and simultaneous calcium imaging of huginPC projections resulted in the induction of rhythmically occurring calcium peaks, demonstrating a functional connection between GR66a and huginPC neurons (Fig. 7b).

Discussion
The bitter taste rejection response is important for all animals that encounter toxic or harmful food in their environment. Here we showed that the hugin neurons in the Drosophila larval brain function as a relay between bitter sensory neurons and higher brain centres (Fig. 8). Strikingly, activation of the hugin neurons made the animals significantly more insensitive to substrates with negative valence like bitter (caffeine) and salt (high NaCl), as well as positive valence like sweet (fructose). In other words, when the hugin neurons are active these animals 'think' they are tasting bitter and therefore become insensitive to other gustatory cues. This is in line with observations made in mice, in which optogenetically activating bitter cortex neurons caused animals to avoid an empty chamber illuminated with blue light. In this situation, although mice do not actually taste something bitter, they avoid the empty chamber since the bitter perception has been optogenetically induced in the central nervous system (CNS) and the mice 'think' they are tasting a bitter substance 25 . In our previous work, activation of all hugin neurons led to behavioural and physiological phenotypes such as decreased feeding, decrease in neural activity of the antennal nerve (AN), and induction of a wandering-like behaviour 20 . We have now pinpointed the neurons responsible specifically to those that project to the protocerebrum. These neurons not only respond to bitter stimuli, but also show a concentration dependent increase in calcium activity in response to caffeine. Dose dependent coding of bitter taste stimuli was previously shown to occur in peripheral bitter sensory neurons, where bitter sensilla exhibit dose dependent responses to various bitter compounds 9 . Larvae in which the huginPC neurons have been ablated still showed some  avoidance to caffeine. Whether this implies the existence of other interneurons being involved in caffeine taste processing remains to be determined. Interestingly, the huginPC neurons are inhibited when larvae taste other modalities like salt (NaCl), sugar (fructose) or protein (yeast). This may indicate that taste pathways in the brain are segregated, but influence each other, as previously suggested 10 .
Bitter compounds may be able to inhibit the sweet-sensing response to ensure that bitter taste cannot be masked by sweet tasting food. This provides an efficient strategy for the detection of potentially harmful or toxic substances in food 26,27 . For appetitive tastes like fructose and yeast, bitter interneurons neurons like the huginPC neurons in the CNS may become inhibited to ensure appropriate behaviour to pleasant food. Salt is a bivalent taste modality, that is, low doses of salt drive appetitive behaviour, whereas high doses of salt are aversive to larval 21,28 and adult 29,30 Drosophila. Inhibition of huginPC neurons when larvae are tasting salt might be due to a different processing circuit for different concentrations of salt and the decision to either take up low doses or reject high doses.
Taken together, we propose that hugin neuropeptide neurons projecting to the protocerebrum represent a hub between bitter gustatory receptor neurons and higher brain centres that integrate bitter sensory information in the brain, and through its activity, influences the decision of the animal to avoid a bitter food source. The identification of second order gustatory neurons for bitter taste will not only provide valuable insights into bitter taste pathways in Drosophila, but may also help in assigning a potentially novel role of its mammalian homologue, Neuromedin U, in taste processing.  Generation of transgenic flies. For HugPC-Gal4 line, a 544 bp Hugin fragment 155 bp upstream of the ATG was amplified with primer1 (AAG GGT TTG GTT TAA TTT ATT TAT GTC ATA) and primer2 (GAG CCT GAT TAG GTC CCT GAT GTT TAA ACT T) and cloned into pCaSpeR-AUG-Gal4-X vector (Addgene plasmid 8,378. The construct was injected into w [1118] ).

Methods
Two-choice gustatory assay. To measure the preference index of larvae towards given appetitive or aversive substrates, 9 mm diameter petri dishes were filled with 20 ml warm water agar (2.125% Agar-Agar, Kobe). After 20 min of air drying half of the agar was cut away and discarded. Compounds (2 M NaCl, 1 M Fructose, 200 mM Caffeine, 10 mM Denatonium) were diluted in warm agar in the given concentration until the agar fluid was clear, and filled in the other half of the petri dish (10 ml). After air drying again for 20 min, petri dishes were prewarmed in the 32°C incubator 1.5 h prior to the experiment. All two-choice experiments were performed at 32°C for comparability between all genotypes. For each experiment 30 larvae were taken out of standard fly food and washed with tap water. Larvae were then placed on the water side of the two-choice dishes and videotaped for 20 min. Videos were processed with FIJI (ImageJ) and analysed using a custom written script for FIJI. Analysis of PI values started 60 s after the start of the experiment to ensure proper tracking of larvae due larval accumulation at the beginning by placing them on the pure agar side. PI gustatory was calculated as (#larvae substrate À #larvae water )/#larvae total .
One-choice olfactory assay. For testing response to an attractive odour (apple vinegar), a one-choice assay was performed. Agar plates were used (as described earlier for two-choice petri dishes) and placed into the incubator at 32°C for 1.5 h. Larvae were videotaped at 32°C for 5 min, and then an Eppendorf cup (1.5 ml) with filter paper soaked with apple vinegar was placed on one side of the water agar plate above the larvae such that they were not able to reach the odorant source. Movement of larvae was analysed using the custom made FIJI macro.  Table 4.
Food intake assay. Apple juice agar plates were prepared with a spot of red yeast paste in the middle of the plate. Plates were then placed in incubators precooled to 18°C or prewarmed to 32°C for 2 h. After 30 min starvation, five larvae were transferred on top of the yeast paste of a plate and videotaped for 20 min. After 20 min of videotaping, larvae were transferred into a small cell strainer and washed with 60°C hot water. Larvae were then transferred to glass slides for photo documentation and analysed with the open source software FIJI (ImageJ) and a custom written analysis macro, which calculated the percentage of the red stained surface of the body compared with the whole body of the larva. To calculate the fold change of food intake from 18 to 32°C, the value of all 32°C values was divided by the mean value of all 18°C values.
Electrophysiology. Third instar larvae were dissected in 35 mm petri dishes coated with 5 ml two-component silicone (Elastosil RT). Larvae were pinned down dorsal side up at the anterior and posterior end using 77 mm thick sharp etched tungsten needles. The larva was cut open longitudinally along the dorsal midline and the cuticle was pinned aside with 40 mm tungsten needles. Interior organs like fat body, intestine or salivary glands were removed except for the cephalopharyngeal skeleton and CNS with attached nerves of interest. Eye and leg imaginal discs were also removed. A transversal cut of the cuticle was performed beneath the CNS to reveal the AN. Nerves not needed for the respective recording were cut. A piece of thinned Parafilm was placed beneath the nerve of interest. This nerve was isolated from the surrounding solution with two adjacent jelly pools. Motor output of the AN was measured using custom made silver wire electrodes connected to a preamplifier (Model MA103, Ansgar Büschges group electronics lab). The preamplifier was connected to a four-channel amplifier/signal conditioner (Model MA 102, Ansgar Büschges group electronics lab). All recorded signals were amplified (amplification factor: 5,000) and filtered (bandpass:  ARTICLE controller (Thorlabs, LEDDB1 driven with 1000 mA) and positioned 18 cm above the solution with the larva. PCR plate of 96-well was filled with 50 ml containing the given taste stimuli in tap water. Larvae were placed into the solution for 2 min and then ultraviolet light was applied for 30 s. Afterwards brains were dissected in PBS and mounted onto a Poly-L-lysine (Sigma, Lot # SLBG4596V) coated cover slide with a drop of PBS. All z-stacks of the HugPC neurons were aquired using a ZEISS LSM 780 Laser scanning microscope with LCI Plan-Neofluar 25 Â /0.8 Imm Korr DIC M27. For quantification of green to red photoconversion maximum intensity projections of the acquired z-stacks were used and a portion of the cytoplasmatic region of each cell was analysed to obtain data for green and red fluorescence intensity. Red fluorescence intensity was divided by green fluorescence intensity to get F red /F green ratio. A 'no ultraviolet light' control was included to show that scanning of the CNS without being exposed to ultraviolet light does not convert green to red fluorescence.
Calcium imaging with GCaMP. Freshly dissected CNS of feeding third instar larvae of the genotypes UAS-dTrpA1/Hug1.2lexA; GR66a-Gal4/lexAop-GCamP6s-p10 (experiment) and Hug1.2lexA/CyO; lexAop-GCamP6s/TM3, Sb (control) were placed with the SEZ region up on a Poly-L-lysine coated coverslide in a drop of saline. Coverslide was attached to a custom built heating device consisting of a 1.5 cm 2 Peltier element for shifting the temperature from 20 to 30°C by applying specific voltage values. Images were acquired with a Zeiss LSM 780 laser scanning microscope as time series with a speed of 781.96 ms (B1.28 Hz) using a Zeiss LCI 'Plan-Neofluar' 25 Â per 0.8 Imm Korr DIC M27 objective dipped in the saline solution. Region of interest covered the complete HugPC neuronal 'sprinkler-like' arborization pattern in the protocerebrum.
Statistics. For comparison of two groups in the two choice assays Mann-Whitney-Rank-Sum-Test was used. Statistical data were acquired as cumulative PI values of the last 5 min of the 20 min experiment and displayed as boxplots. For food intake analysis 32°C values (% of red yeast in gut relative to whole body) were divided by the mean of all 18°C values to gather the fold change of food intake. Fold changes were then compared with the Mann-Whitney-Rank-sum-Test. For electrophysiological data cycle frequencies were analysed at 18°C for 60 s and 32°C for 60 s. The fold change was calculated between 18 and 32°C. Per larva a maximum of four temperature steps could be applied to the CNS during recordings. All fold changes of one genotype were then compared with the Mann-Whitney-Rank-Sum-Test with the other genotypes. The relative fold change is displayed as subtraction of the mean of OrgR Â dTrpA1 control values from the fold change values of all genotypes (OrgR Â dTrpA1 set to 0). CaMPARI data were analysed by calculating the mean F red /F green value of all eight HuginPC neurons of one larva to determine one mean F red /F green value per larva. Values were then compared with the Mann-Whitney-Rank-Sum-Test to water control.
Data availability. The authors declare that the data supporting the findings of this study are available within the article (and its Supplementary Information files), or available from the authors upon request. Bitter taste, detected in peripheral taste organs, activates huginPC neurons. Activity of huginPC neurons projecting to higher brain centres leads to a decrease in food intake and in the cycle frequency of feeding related motor patterns in the antennal nerve (AN). Activation of huginPC neurons also leads to aversion to different gustatory substrates, including yeast, thereby triggering aversive behaviour. Thus, huginPC neurons act as relay of bitter taste to the protocerebrum in the Drosophila larval brain.