A molecular and neuronal basis for amino acid sensing in the Drosophila larva

Amino acids are important nutrients for animals, reflected in conserved internal pathways in vertebrates and invertebrates for monitoring cellular levels of these compounds. In mammals, sensory cells and metabotropic glutamate receptor-related taste receptors that detect environmental sources of amino acids in food are also well-characterised. By contrast, it is unclear how insects perceive this class of molecules through peripheral chemosensory mechanisms. Here we investigate amino acid sensing in Drosophila melanogaster larvae, which feed ravenously to support their rapid growth. We show that larvae display diverse behaviours (attraction, aversion, neutral) towards different amino acids, which depend upon stimulus concentration. Some of these behaviours require IR76b, a member of the variant ionotropic glutamate receptor repertoire of invertebrate chemoreceptors. IR76b is broadly expressed in larval taste neurons, suggesting a role as a co-receptor. We identify a subpopulation of these neurons that displays physiological activation by some, but not all, amino acids, and which mediate suppression of feeding by high concentrations of at least a subset of these compounds. Our data reveal the first elements of a sophisticated neuronal and molecular substrate by which these animals detect and behave towards external sources of amino acids.

Scientific RepoRts | 6:34871 | DOI: 10.1038/srep34871 a key mechanism for environmental amino acid sensing, the observed heterogeneity in perception suggests that these stimuli act through additional pathways.
Dietary amino acids are also critical for insects, notably to support a high rate of egg production in females 13 . Although there has been some investigation of behavioural responses of insects towards individual amino acids 14 , the sensory mechanisms allowing them to perceive these chemicals in the environment are largely unknown. Electrophysiological responses of chemosensory neurons in adult feeding organs to individual (or mixes of) amino acids have been described in several species [15][16][17][18] , but the behavioural roles of these neurons are unexplored. In the genetic model Drosophila melanogaster, where substantial progress has been made in revealing the molecular and neuronal basis of sweet and bitter tastants 19 , amino acid sensing mechanisms are surprisingly poorly understood. This might reflect, in part, the observation that adult Drosophila display robust behavioural responses towards amino acids only when previously deprived of these nutrients 20 , similar to locusts 18 . It remains unclear whether such plasticity in behaviour reflects the function of specific peripheral chemoreceptive pathways or internal metabolic amino acid sensors.
Larval Drosophila represent an appealing system for investigating sensory detection of amino acids, because these animals display a remarkably persistent appetite to support a 250-fold increase in mass from hatching to  pupation. In this study, we have investigated the behavioural responses of larvae to amino acids and delineated some of the relevant chemosensory receptors and neurons.

Results
Drosophila larvae display innate, stimulus-specific responses to amino acids. We first asked whether larvae show innate behavioural responses to amino acids using a modified version of a paradigm for measuring taste preferences 21 . In brief, experimentally-naïve animals were allowed to wander freely on the surface of a soft agarose substrate in a four-quadrant circular arena, in which two opposing quadrants also contained a specific amino acid (Fig. 1A). The distribution of animals on quadrants containing or lacking the amino acid was quantified after 20 minutes and used to generate a Preference Index for the stimulus (Fig. 1A). Larvae display diverse behavioural responses to the 20 canonical L-amino acids (Fig. 1B). Seven amino acids promoted varying degrees of attraction (threonine, arginine, asparagine, phenylalanine, cysteine, glutamic acid and aspartic acid), one (leucine) was aversive, and the remainder were neutral. Essential amino acids are not confined to any particular category of response (Fig. 1B, blue labelling). Furthermore, we did not observe any clear relationship between physico-chemical properties of different amino acids and the behavioural responses they evoke; although 6 out of 7 appetitive amino acids are polar, six other polar amino acids are neutral (Fig. 1C). These results suggest the existence of multiple mechanisms that mediate different behavioural responses to distinct amino acid stimuli.
IR76b is required for behavioural attraction to amino acids. Although   compounds) 24,25 , while individual GRs and IRs are expressed in small numbers of neurons in one or more of these chemosensory organs [26][27][28][29][30] . Remarkably few relationships have been defined between specific gustatory ligands and particular sensory neurons in the larva [31][32][33] . We were particularly interested to test the contribution of IRs to amino acid detection, because these receptors are distantly related to ionotropic glutamate receptors (iGluRs) 27,34,35 . iGluRs are best known for their role in mediating synaptic transmission, but homologues in prokaryotes and plants have been suggested to function as more broadly-tuned amino acid sensors 36,37 . In the adult olfactory and gustatory systems, most IRs are expressed in discrete, functionally specialised neurons 27,34,38,39 , but three (IR8a, IR25a, IR76b) are expressed in multiple neuron classes. These IRs appear to function as co-receptors together with "tuning" IR subunits that define ligand specificity 40 . We reasoned that study of these co-receptors might provide an entry-point to determine the contribution of this repertoire to amino acid sensing.
IR8a expression was not detected in the larva, using available antibodies or genetic reporters 40 (data not shown). By contrast, consistent with previous analyses 26-28 , we observed expression of IR25a in multiple sensory neurons in all of the larval chemosensory organs ( Fig. 2B and data not shown). An IR76b reporter (IR76b-Gal4 38 ) is expressed in a similarly broad pattern, except the Dorsal Organ, with substantial, but not complete, overlap with the IR25a-positive cells ( Fig. 2B and data not shown).
To test the contribution of IR25a and IR76b to amino acid sensing, we examined the responses of mutant larvae lacking these receptor genes to the stimuli that evoked the strongest attraction (aspartic acid, glutamic acid, cysteine and phenylalanine). IR25a mutant larvae exhibited statistically indistinguishable responses when compared to control genotypes, indicating that this co-receptor is dispensable for amino acid sensing (Fig. 2C). However, IR76b mutants displayed substantially decreased or abolished behavioural responses to aspartic acid, glutamic acid and phenylalanine (Fig. 2D). These defects could be rescued by expression of an IR76b cDNA under the control of IR76b-Gal4, confirming these neurons as the site of action of this receptor (Fig. 2D). Interestingly,

IR56c-Gal4
IR94e-Gal4 IR94h Glutamine 200 mM Fluorescence intensity (a.u.)  IR76b mutants still preferred cysteine (Fig. 2D); we speculate that this response might reflect, in part, a contribution of the olfactory system to detection of this odoriferous amino acid 41 .

IR60c-Gal4 > UAS-GCaMP3
IR76b-Gal4 neurons respond physiologically to amino acids. We next asked whether neurons labelled by IR76b-Gal4 respond physiologically to amino acids. We developed a preparation that allowed us to measure changes in calcium levels (using the genetically-encoded reporter GCaMP3 42 ) in the axonal projections of these neurons in the brain upon application of taste stimuli to the larval head (Fig. 3A). Consistent with the large number of gustatory neurons labelled by IR76b-Gal4 (Fig. 2B), the corresponding axon termini innervate a broad region of the subesophageal zone (SEZ), their primary target area in the larval brain (Fig. 3B). Application of arginine, an amino acid that elicits preference behaviour in larvae (Fig. 1B), evoked robust increases in calcium levels (Fig. 3C). These calcium changes were spatially restricted to subregions of the axonal innervations, generally within the posterior half of the SEZ, suggesting that only a subset of IR76b-Gal4 neurons are sensitive to amino acids (Fig. 3C, left). Analysis of the temporal dynamics revealed a fast onset of the response (Fig. 3C, right), although it was not possible to calculate precisely the response latency due to the method of stimulus presentation.
We extended this analysis by measuring responses to 18 other amino acids (tyrosine was excluded because of its low solubility in water). Increases in calcium levels were observed upon presentation of nine amino acids; this subset did not correspond precisely with the stimuli that trigger preference behaviour (Fig. 3D). This discordance might reflect the existence of neurons outside the IR76b-Gal4 population that contribute to behaviours towards at least some of the amino acids and/or that we are unable to detect physiological responses to certain ligands with our calcium imaging approach.
Identification of a single neuronal subset responsive to amino acids. We next sought to characterise in more detail the amino acid-sensitive subset(s) of IR76b-Gal4 neurons. We reasoned that the fast neuronal responses to amino acids in the imaging experiments would be consistent with a peripheral, rather than internal location of the sensory endings 33 and therefore focussed our analysis on neurons located in the Terminal Organ. Previous studies identified Gal4 driver lines built from putative regulatory elements of two IR genes (IR7a and IR47a) that are expressed in subsets of Terminal Organ neurons 27,28 . In the course of characterisation of a complete set of IR-Gal4 drivers (V.C. and R.B., unpublished data), we found an additional eight lines (containing regulatory regions for IR7b, IR7d, IR7e, IR7g, IR56c, IR60c, IR94e or IR94h) that are expressed in 1-4 neurons in the Terminal Organ, and whose axons project to the SEZ (Fig. 4A).
To test whether any of these IR-Gal4-expressing neuron populations are sensitive to amino acids, we screened them for calcium responses to glutamine, which evoked one of the strongest responses in IR76b-Gal4 neurons (Fig. 3D). (We avoided glycine, which produced the highest responses, as this amino acid might also be detected by sweet sensing neurons 43 ). For two lines, IR7b-Gal4 and IR7d-Gal4, basal GCaMP3 fluorescence was too weak for reliable measurement of responses. Of those that we were able to analyse, glutamine-evoked calcium increases were detected in only a single population of neurons, labelled by IR60c-Gal4 (Fig. 4B). These responses displayed a similarly fast onset to those of IR76b-Gal4 neurons (Fig. 4C).
We expanded our physiological profiling of IR60c-Gal4 neurons and found that they exhibited strong responses to precisely the same subset of amino acids as IR76b-Gal4 neurons (Fig. 5A), as well as lysine (potentially reflecting the higher stimulus concentration used in these experiments). Responses were observed for L-isomers but not D-isomers (Fig. 5B), consistent with the neurons only detecting amino acids used in protein synthesis. Moreover, no substantial calcium changes were observed in response to the water solvent or any tested sweet or bitter compound (Fig. 5C).
IR60c-Gal4 neurons are not required for amino acid preference. As IR60c-Gal4 neurons were the only identified subpopulation of cells in the Terminal Organ that was activated by amino acids, we tested whether these neurons mediate appetitive behavioural responses of larvae to these compounds. We silenced IR60c-Gal4 neurons (using the inwardly rectifying potassium channel Kir2.1 44 ) and tested the consequence for larval preference for four amino acids that evoke appetitive behaviour (glutamic acid, cysteine, phenylalanine and arginine). Surprisingly, these animals displayed indistinguishable responses compared to genetic controls (Fig. 6), indicating that IR60c-Gal4 neurons are not required for behavioural attraction to amino acids.
IR60c-Gal4 neurons and IR76b mediate feeding suppression by amino acids. Given the lack of requirement of IR60c-Gal4 neurons for attraction to amino acids, we surmised that they control other types of behaviour. We therefore developed an experimental paradigm in which larvae are allowed to feed from a sucrose solution containing an edible dye to allow spectrophotometric quantification of food intake (see Methods) (Fig. 7A). The role of IR60c-Gal4 neurons in modulating such feeding was assessed by artificially activating these cells with VR1, a mammalian ion channel gated by capsaicin (which is a behaviourally neutral compound for form: UAS-GCaMP3;IRxx-Gal4) to presentation of glutamine (200 mM; preliminary observations prompted our use of higher concentrations than in Fig. 3, to ensure detection of responses in the sparse innervations of individual populations of IR-Gal4 neurons). IR7b-Gal4 and IR7d-Gal4 were excluded because the basal GCaMP3 fluorescence with these drivers was too low to perform imaging experiments. (Mean response ± SEM is shown; N ≥ 5 animals per stimulus). (C) Top: False-coloured representation of GCaMP3 fluorescence intensity in one hemisphere of the SEZ of a UAS-GCaMP3;IR60c-Gal4 animal, before and after presentation of 200 mM glutamine. Scale bar: 20 μ m. Bottom: example time course of the calcium response of IR60c-Gal4 neurons to glutamine presentation (arrowhead) from the region of interest indicated by a dashed circle on the upper panels.
We next tested whether natural amino acid stimuli could also evoke feeding suppression through IR60c-Gal4 neurons. Indeed, adding a high concentration of arginine to the sucrose solution almost completely inhibited feeding in genetic controls (Fig. 7C,D). This inhibition was lifted when IR60c-Gal4 neurons were silenced (with Shibire ts1 47 ) (Fig. 7C), with these animals feeding at similar levels to assays where no amino acid was  present (Fig. 7E). This phenotype did not appear to be due to enhanced perception of sucrose, because blocking IR60c-Gal4 neuron function had no effect on feeding on sucrose alone (Fig. 7E).
We extended this analysis to other amino acid ligands of IR60c-Gal4 neurons: alanine, lysine and glycine. As observed with arginine, high concentrations of alanine suppressed feeding, and this depended upon functional IR60c-Gal4 neurons (Fig. 7F-G). By contrast, addition of lysine or glycine revealed two different effects. Lysine suppressed feeding even when IR60c-Gal4 neurons were blocked (Fig. 7H), suggesting that additional sensory pathways mediate feeding suppression by this amino acid. Glycine did not produce any feeding inhibition, even in animals with functional IR60c-Gal4 neurons (Fig. 7I). Given that glycine can activate IR60c-Gal4 neurons (Fig. 5A), the latter observation suggests that this amino acid is also sensed by a distinct attraction-mediating pathway that overrides IR60c-Gal4 neuron-dependent aversion.
We confirmed a role for IR60c-Gal4 neurons in mediating aversion to high concentrations of amino acids in a gustatory choice assay for substrates containing sucrose alone versus sucrose supplemented with a high concentration of alanine (which gave the most robust phenotype in the feeding assay) (Fig. 7F). While control larvae avoided sucrose medium containing this amino acid, animals with silenced IR60c-Gal4 neurons did not (Fig. 7J).
Finally, we tested whether the putative co-receptor IR76b is also involved in feeding suppression by high concentrations of amino acids. While wild-type and heterozygous IR76b control strains barely ate sucrose containing amino acids, animals lacking IR76b robustly consumed this solution (Fig. 7K). This phenotype is similar to that observed when inhibiting IR60c-Gal4 neurons. These observations indicate that IR76b is not only required for appetitive behaviour towards amino acids in palatable concentrations (Fig. 2D), but also for feeding suppression by high concentrations of these compounds.

Distinct pathways for larval aversion.
Larval behavioural aversion to bitter compounds, such as caffeine, has been reported upon activation of GR66a-expressing Terminal Organ neurons 45 , prompting us to investigate the role of IR60c-Gal4 neurons in caffeine-related behaviour. Similar to the effect of high concentrations of amino acids, caffeine potently suppressed larval feeding on sucrose (Fig. 8A). However, this suppression is unaffected by inhibition of IR60c-Gal4 neurons (Fig. 8A). This result is consistent with our demonstration that caffeine does not activate these neurons (Fig. 5C), and suggests the existence of different pathways for feeding suppression by amino acids and bitter substances. Indeed, anatomical comparison using transgenic reporters for IR60c and GR66a revealed that there is no overlap between these populations of neurons in the Terminal Organ (Fig. 8B). However, the axonal projections of these neurons partially overlap in the SEZ (Fig. 8B), raising the possibility that they converge on at least some common second-order neurons.

Discussion
This study is, to our knowledge, the first to investigate the molecular and neuronal mechanisms underlying amino acid detection in the Drosophila larva. Despite the numerical simplicity of this animal's chemosensory system 22,23 , our data reveal an unexpected complexity in how it responds to this class of compounds. We find that different amino acids can produce behavioural responses of opposing valence. Moreover an amino acid stimulus that produces appetitive behaviour at low concentration can evoke aversion at higher concentrations. Behavioural valence does not correlate with the essential/non-essential dietary requirements for amino acids, nor with their physico-chemical properties (as observed in an independent study (T. Tanimura, personal communication)). These observations suggest that the assignment of value to amino acids is a complex process, potentially related to the diverse properties and biological functions of these chemicals 1 . It is important to note that the presumed "innate" behaviours we have measured in third instar larvae are not necessarily fixed across life stages and strains of different genetic background (T. Tanimura, personal communication), as previously noted in adult Drosophila 48 .
We have identified a single class of broadly (but not universally) tuned amino acid sensing neurons, marked by IR60c-Gal4, in the main gustatory organ of the larva. Unexpectedly, the physiological profile of these neurons does not show concordance with the observed behavioural sensitivity or valence for these compounds. This observation implies the existence of additional sensory pathways for amino acids not revealed in the present study, or which exhibit responses below the detection threshold of our experimental assays.
Our demonstration that at least some of the behavioural responses to amino acids require the IR76b co-receptor provides the first evidence for the molecular mechanism of this taste modality. It seems very unlikely, however, that IR76b functions as an amino acid sensor itself. This receptor is broadly expressed in multiple populations of chemosensory neurons, and, in adults, has been implicated in sensory transduction of diverse odours in the olfactory system 40,49 , and in polyamine and salt sensing in distinct subsets of neurons in the gustatory system 49,50 . IR76b has been suggested to function independently as a salt sensor (potentially by acting as sodium leak on 50 mM sucrose containing 500 mM alanine and 0.4% Brilliant Blue for 30 minutes at the restrictive (F) or permissive (G) temperature. (Kruskal-Wallis, ***p < 0.001; N ≥ 12 groups of 30 larvae per genotype). (H-I) Absorbance of extracts from homogenised larvae of the same genotypes as in (C) after animals have fed on 50 mM sucrose containing 500 mM lysine (H) or 500 mM glycine (I) and 0.4% Brilliant Blue for 30 minutes at the restrictive temperature. (Kruskal-Wallis, ***p < 0.001; N ≥ 12 groups of 30 larvae per genotype). (J) Preference Indices of larvae in which IR60c-Gal4 neurons are silenced with Kir2.1 (UAS-Kir2.1:GFP/+ ;IR60c-Gal4/+ ), together with controls (UAS-Kir2.1:GFP/+ and IR60c-Gal4/+ ; "+ " represents Canton S-derived chromosomes) for 50 mM sucrose versus 50 mM sucrose + 500 mM alanine. (Wilcoxon Signed Rank Test, **p < 0.01, ***p < 0.001; N ≥ 69 groups of ~15 animals per genotype). (K) Absorbance of extracts from homogenised wild-type (w 1118 ), homozygous IR76b 1 mutant and heterozygous IR76b 1 /+ ("+ " represents w 1118 derived chromosomes) larvae after animals have fed on 50 mM sucrose containing 500 mM alanine and 0.4% Brilliant Blue for 30 minutes. (Kruskal-Wallis, ***p < 0.001; N = 16 groups of 30 larvae per genotype). channel) 50 , but in other contexts it appears to function as a co-receptor with other more selectively-expressed IRs that determine ligand specificity 40,49 . We propose that IR76b must also act with one, or more, different tuning IRs to mediate responses to amino acids. Although we identified the IR60c-Gal4 expressing neurons as amino acid sensors, IR60c does not itself appear to be required for physiological responses of these cells (V.C., J.R.A. and R.B., unpublished data), suggesting that other receptors act with IR76b in these neurons. Regardless, the implication of a variant iGluR in peripheral amino acid sensing in insects provides an interesting molecular analogy to the role of the mGluR-related T1R1/T1R3 umami sensors in mammals 11,12 , and might be reflective of a broader role of the iGluR superfamily in amino acid sensing 36,37 . IR-Gal4 transgenes. Primers were designed to amplify putative regulatory regions (covering the entire 5′ intergenic region) from Oregon R genomic DNA with appropriate flanking restriction sites (Supplementary Table 1). PCR products were cloned in pGEM-T Easy (Promega), end-sequenced, and sub-cloned into the pGal4 attB vector 27 . All constructs were integrated into the attP2 landing site at cytological location 68A 55 , by phiC31-mediated transgenesis (Genetic Services, Inc.).

Methods
Behavioural analyses. Larval gustatory choice assays. All chemicals were from Sigma-Aldrich or Applichem (Supplementary Table 2 "agarose only" quadrants, and two "agarose + tastant" quadrants. Media were allowed to cool to room temperature and polymerise for > 1 hour. Plates were prepared freshly on the same day as behavioural assays. Third instar larvae were collected, rinsed with tap water and kept for ~1 hour on 0.8% agarose. Groups of 30 animals were placed at the centre of each Petri dish under red light. For Figs 6 and 7J: split Petri dishes (60 mm diameter, VWR) were used and the agarose concentration was 1%. Groups of 15 animals were collected and kept for 1 hour in a droplet of tap water. For the assays in Fig. 7J, larvae were used immediately after collection. Animals were placed in the centre of the Petri dish under white light. For all assays, the number of animals on each substrate was counted after 20 minutes and the Preference Index was calculated as shown in Fig. 1A. Statistics and plots were made with R (R-Development-Core-Team; www.R-project.org).
Larval feeding assays. Groups of 30 early third instar larvae were collected and kept for 1 hour in 500 μ l PBS in 1.5 ml microcentrifuge tubes placed on their sides. Tubes were incubated at the desired permissive (23 °C) or restrictive (32 °C) temperature for an additional 20 minutes. 500 μ l of 2X tastant solution and Brilliant Blue FCF (Spectrum Chemical; final concentration 0.4%) in PBS was added to the tubes and larvae were allowed to feed for 30 minutes at the selected temperature. Tubes were transferred to ice for 15 minutes to anesthetise animals and terminate feeding. Larvae were washed 4x with PBS, homogenised in 100 μ l PBS and centrifuged for 12 minutes at 13,000 rpm. 50 μ l of the supernatant was pipetted into 50 μ l PBS and the absorbance of this solution at 625 nm was measured using a NanoDrop. A solution from wild-type larvae that had not been fed with dye was used as a blank for the measurement. Plots and analyses were performed with Prism 6 (GraphPad Software; www.graphpad.com).
Calcium imaging. A head of a third instar larva expressing UAS-GCaMP3 under the control of the desired IR-Gal4 driver was dissected so that the brain remained attached to the mouthparts. Imaginal discs and salivary glands were removed. The head was placed on a 50 μ l drop of 1.5% low melting point agarose (Promega) in Drosophila sugar-free adult haemolymph-like (AHL) Ringer's saline (108 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 8.2 mM MgCl 2 , 4 mM NaHCO 3 , 1 mM NaH 2 PO4, 5 mM HEPES pH 7.5) at 50 °C on a 25 × 60 mm glass coverslip and gently pulled away from the drop so that the brain directly lay on the coverslip. After polymerisation a second 50 μ l drop of 1.5% agarose at 50 °C was added on top of sample. Once polymerised, agarose was removed from the tip of the larval head with a sharp blade to uncover peripheral sensory organs, and replaced with 5 μ l sugar-free AHL Ringer's saline, to which an equal volume of tastant solution was added at the appropriate time to stimulate sensory neurons. The subesophageal zone (SEZ) was visualised with a Zeiss LSM 510 Meta inverted microscope using an EC Plan-NEOFLUAR 40x/1,30 Oil DIC objective. Images were acquired with an AxioCam MRm camera using LSM 3.5 Software. Typically, recordings were made for 87.5 seconds with 4 images per second (350 frames) and tastants were added after 25 seconds (frame 100). As tastants could not be washed from the preparation, only one stimulus was used per animal, except where a tastant did not elicit a response; in these cases, a positive control stimulus (i.e., an established amino acid agonist) was subsequently applied to verify viability and responsiveness of the preparation.
Lateral movement on collected image stacks was corrected using the StackReg plugin (bigwww.epfl.ch/ thevenaz/stackreg) in NIH ImageJ. Image stacks were further processed using an adapted custom MATLAB (MathWorks) script 38 . The relative percentage change in fluorescence (Δ F/F) was measured as: (F i − F 0 )/F 0 × 100, where F 0 is the mean fluorescence value of frames 20-95 (before stimulus presentation), and F i is the fluorescence value for the i th frame. Both sides of the SEZ were imaged and for each sample, the Region of Interest (ROI) was a 10 μ m diameter circle around an area displaying high fluorescence changes (determined by visual inspection), which was invariably within the posterior half of one of the two sides of the SEZ projection, where Terminal Organ neuron innervations are located. For each frame, the average Δ F/F was calculated from this ROI. The peak value recorded within the 25 seconds (100 frames) following stimulation was defined as "Δ F/F max". Plots were produced with custom scripts in R and ImageJ.