. Author manuscript; available in PMC: 2026 Oct 23.

Published in final edited form as: Curr Biol. 2026 Sep 23;36(20):5285–5299.e7. doi: 10.1016/j.cub.2026.08.070

A peripheral feedforward circuit amplifies taste signals, ensuring behavioral reliability

Bo-Mi Song ¹, Bong-Hyun Kim ², Mark Stopfer ^1,^*

PMCID: PMC13688462 NIHMSID: NIHMS2209771 PMID: 42777715

Summary

Mechanisms ensuring reliable behavioral control amidst complex, noisy sensory environments are often attributed to the central brain, yet computations occurring at the extreme sensory periphery remain poorly understood. Recent evidence hints at complex computations at the earliest stages of taste detection, but whether peripheral circuits of gustatory receptor neurons (GRNs) exist to perform them remains unknown. Here, using transsynaptic mapping and paired electrophysiological recordings in Drosophila, we reveal a feedforward circuit of monosynaptically connected GRNs. When driven by tastants, these GRNs significantly increase spiking in follower GRNs, integrating, amplifying, and stabilizing taste signals before they reach the central nervous system. Connectivity among these GRNs is selective and directional rather than all-to-all, and is mediated by the nicotinic acetylcholine alpha 5 receptor subunit. Genetic ablation of this subunit significantly impairs functional interactions between GRNs and significantly decreases the reliability of taste-driven behaviors, phenotypes reversed by cell-specific rescue of receptor expression. These results demonstrate that GRNs are organized into functional circuits that stabilize taste signals and gustatory behaviors.

eTOC Blurb

Song et al. identify a feedforward circuit of synaptically connected peripheral taste-sensing neurons in Drosophila. Connectivity is selective, mediated by the nAChR α5 receptor, and strengthens and stabilizes taste signals before they reach the brain. Cell-specific ablation of the receptor decreases the reliability of synaptic function and taste-driven behavior.

Graphical Abstract

graphic file with name nihms-2209771-f0007.webp

Introduction

A basic challenge of nervous systems is ensuring reliable control over behavior. Complexity in the sensory environment and noise affecting every stage of neural processing can limit the reliability of responses of individual neurons^1,2. Thus, a fundamental question in neuroscience is how populations of neurons collectively process sensory information to ensure reliable behaviors^3–5.

Gustation is essential for survival and requires reliable control of behaviors that regulate food acceptance and avoidance. Traditionally, gustatory processing was viewed as fairly simple: distinct receptor cell populations detect basic taste categories, including bitter, sweet, salty, sour, umami, and convey this information through separate pathways that drive specific behaviors^6–9. However, recent studies suggest greater complexity: gustatory receptor neurons (GRNs, neurons expressing gustatory receptors) and their targets have been shown to employ temporal coding^10,11 and are regulated by support cells, feedback from the brain, and interactions with other GRNs^9,12–19. These findings suggest complex neural computations may arise in the earliest stages of gustation. In the central brain, computations like these are achieved by neural circuitry. But whether peripheral circuits comprised of GRNs exist and how they might contribute to shaping behavior remain unknown.

Because the stimulus sensitivities, temporal response dynamics, and other properties of taste receptor cells vary substantially, even within a receptor class^20–23, understanding the first stages of gustatory processing requires neuron-level descriptions of GRNs and their structural and functional organization. Specifically, it is necessary to determine which cells are connected, whether and how they functionally interact, and whether their interactions exert a meaningful influence on downstream behavioral outputs.

Owing to its relatively simple and accessible anatomical structure and its abundance of genetic tools, Drosophila melanogaster provides a useful model for exploring peripheral taste circuits at the level of individual cells. The fly’s feeding organ, the labellum, located on the distal tip of the proboscis, has received special scrutiny. Each hemisphere of the labellum contains about 30 sensilla, hair-like structures that are in some ways analogous to taste buds, but individually identifiable by morphology and location²⁰. Each sensillum houses two or four GRNs, nearly exclusively expressing one of five receptor proteins^24,25. Comprehensive spatial maps of GRNs expressing these type-specific molecular markers are available²⁴ so that, by combining anatomical and genetic information, individual GRNs can be located and identified across individual flies. The taste-elicited responses of GRNs can be assessed with electrophysiological recordings made with electrodes placed over the sensillum or at its base²⁶, revealing the properties of individual GRNs^11,20–22. Together, these features make the Drosophila labellum well suited for dissecting peripheral gustatory circuits at the resolution of identified individual neurons.

A recent connectomic analysis of labellar gustatory axonal projections, derived from a complete adult Drosophila brain volume reconstructed via serial-section electron microscopy, revealed synaptic connectivity among GRNs²⁵. However, because the brain volume does not include the proboscis, the identities of connected GRNs are unknown. Further, it is unknown whether such connections support functional interactions or whether or how the circuitry influences gustatory behavior.

Here, by combining comprehensive genetically-driven transsynaptic mapping with paired electrophysiological recordings, we demonstrate that five identifiable GRNs, all with similar tuning for bitter compounds, are monosynaptically connected. Notably, the functional interactions between these GRNs are selective and heterogeneous in strength and direction, forming an activity-promoting feedforward circuit that integrates, amplifies, and stabilizes spiking responses to tastes. We also identify the nicotinic acetylcholine receptor subunit mediating most inter-GRN interactions, and, using behavioral experiments in which this subunit was selectively blocked and rescued, we demonstrate that these synaptic interactions enhance the reliability of behavioral regulation. Together, our results establish circuit-level computation among peripheral gustatory receptor neurons as a previously unrecognized mechanism for reliable behavioral control.

RESULTS

Labellar GRNs form two distinct patterns of direct synaptic connectivity.

To identify synaptic connections between labellar GRNs, we performed systematic genetically-driven anterograde transsynaptic mapping²⁷. GRNs are housed within sensilla, which are classified by size as S, I, or L (small, intermediate, or large²⁰). Each sensillum contains two or four GRNs, each of which nearly exclusively expresses one of five receptor proteins – Gr66a, Gr64f, Ppk23, Ppk28 and Ir94e^24,25 (see Figure 1A for a spatial map). To map synaptically connected GRNs, we used trans-Tango with five Gal4 drivers that together label more than 93% of labellar GRNs²⁴. This tracing approach labeled presynaptic GRNs with membrane-bound GFP and their postsynaptic GRNs with tdTomato (Figure 1B).

trans-Tango tracing revealed two distinct connectivity patterns (Figure 1C-G). First, Gr66a- and Ppk23-Gal4-expressing GRNs exhibited tdTomato only in GFP-positive cells (Gr66a-Gal4>trans-Tango: n=13, N=7; Ppk23-Gal4>trans-Tango: n=7, N=4; n=number of samples, N=number of independent experiments) (Figure 1C-D), indicating that these GRNs connect only to other GRNs expressing the same gustatory receptors. Second, Gr64f or Ppk28-Gal4-expressing GRNs showed tdTomato in both GFP-positive and GFP-negative GRNs (Gr64f-Gal4>trans-Tango: n=11, N=7; Ppk28-Gal4>trans-Tango: n=8, N=4) (Figure 1E-F), revealing connections to other GRNs expressing either the same or different gustatory receptors. Ir94e-Gal4 did not trigger tdTomato expression in GRNs innervating labellar sensilla (Ir94e-Gal4>trans-Tango: n=9, N=4) (Figure 1G), suggesting they do not form direct synaptic connections with other GRNs. Ppk23-Gal4 and Gr66a-Gal4 overlap in a small subset of GRNs²⁴ and thus, synaptically connected GRNs identified in Ppk23-Gal4>trans-Tango and Gr66a-Gal4>trans-Tango flies may coincide. Gene expression driven by Ir94e-Gal4 was reliable but weak (Figure S1); thus, we cannot exclude the possibility that, in this case, Tango-ligand expression was insufficient for trans-synaptic labeling. Together, these results demonstrate that GRNs form selective monosynaptic connections with two distinct connectivity patterns defined by gustatory receptor type.

Similarly-tuned Gr66a-expressing GRNs form synaptic connections.

The availability of several-Gal4 drivers targeting different subsets of Gr66a-expressing GRNs made systematic analysis tractable, leading us to focus our study on Gr66a-expressing neurons (Figure 2A)²⁰. Except for sensilla lacking Gr66a expression, each I- and S-sensillum is innervated by a single Gr66a-expressing GRN (Figure 2A). These neurons form synaptic connections exclusively with one another. Thus, we could identify synaptically connected Gr66a-expressing GRNs by locating sensilla innervated by GFP- or tdTomato-expressing GRNs (Figure 1B-C).

Figure 2. — (A) Spatial distribution of Gr66a-expressing GRNs (orange dots) and genetic coverage by subtype-specific Gal4 drivers in S and I sensilla. Gr47a-Gal4 targets I8–I10; Gr59c-Gal4 targets I0–I6 and (S0), S1, S2, S6, S7, and S10; Gr93a-Gal4 targets (S0), S1, S2, S6, S7, and S10; Gr22f-Gal4 targets S3, S5, and S9. (B-E) Representative confocal images of labella expressing trans-Tango under subtype-specific Gal4 drivers targeting subsets of Gr66a-expressing GRNs. GFP (left), tdTomato (middle), and composite (right) channels are shown. Schematics at left summarize connectivity patterns. Dotted lines in some panels highlight neural fibers. (B) Gr47a-Gal4 shows no tdTomato signal, indicating no detectable synaptic connectivity. (C) Gr59c-Gal4 shows tdTomato colocalization with GFP in both S and I sensilla. (D) Gr22f shows no tdTomato signal, indicating no detectable synaptic connectivity. (E) Gr93a-Gal4 shows tdTomato colocalization with GFP in S sensilla. GFP-positive Gr66a-expressing GRNs in S sensilla labeled in cyan text. GRNs consistently identified as postsynaptic partners are labeled in magenta text. GRNs expressing both GFP and tdTomato are labeled in white text. N-values indicate labella (n) and independent experiments (N): Gr47a, n=7, N=5; Gr59c, n=8, N=7; Gr22f, n=9, N=4; Gr93a, n=11, N=5. Scale bars, 30 μm. Related to Figures S2-S3.

Three-Gal4 drivers, Gr47a, Gr59c, and Gr22f, target different subsets of Gr66a-expressing GRNs and together cover the entire population of Gr66a-expressing sensilla in the labellum (Figure 2A)²⁰. Of these, only Gr59c-Gal4 promoted tdTomato expression in GRNs (Gr47a-Gal4>trans-Tango: n=7, N=5; Gr59c-Gal4>trans-Tango: n=8, N=7; Gr22f-Gal4>trans-Tango: n=9, N=4), and all tdTomato-expressing GRNs in Gr59c-Gal4>trans-Tango flies also expressed GFP (Figure 2B-D, Figure S2A; Gr59c-Gal4>trans-Tango: S0-S10, n=3, N=3; S1-S10, n=3, N=3). Gr59c-Gal4 drives expression in Gr66a-expressing GRNs in sensilla I0-I6 and S1, S2, S6, S7, and S10 (and S0 when present) (Figure 2A)²⁰, suggesting that Gr66a-expressing GRNs innervating these sensilla are synaptically interconnected.

Some labella exhibited anatomical variations in sensilla number and placement²⁸. In these atypical labella, we found the number and distribution of Gr66a-expressing GRNs also varied, but, notably, connectivity patterns were preserved with respect to receptor type (Figure S3). We selected animals with stereotypical S-sensilla anatomy for further analysis, as anatomical variability was less frequent in S- than I-sensilla, and Gr59c-Gal4-expressing GRNs in S-sensilla were the only cells consistently labeled by tdTomato (Figure 2C, Figure S2A).

While the tuning properties of Gr66a-expressing GRNs are variable, those innervating sensilla S0, S1, S2, S6, S7, and S10 share similar tuning profiles and form a distinct functional subclass, S-a²⁰. In Gr59c-Gal4 > trans-Tango flies, we identified neurons in these five sensilla by their co-expression of GFP and tdTomato. We hypothesized that GRNs sharing similar tunings are preferentially interconnected. To test this, we used Gr93a-Gal4, which selectively and comprehensively targets Gr66a-expressing GRNs innervating these five S-sensilla²⁰ (Figure 2E, Figure S2B). Consistent with our hypothesis, all Gr93a-Gal4-expressing GRNs co-expressed GFP and tdTomato (Gr93a-Gal4>trans-Tango: S0-S10, n=5, N=3; S1-S10, n=4, N=3; Figure 2E, Figure S2B). This result confirms that similarly tuned Gr66a-expressing GRNs innervating these five S-sensilla form synaptic connections.

Similarly-tuned Gr66a-expressing GRNs show functional feedforward excitatory interactions.

Our trans-Tango results provided anatomical evidence that five similarly tuned, Gr66a-expressing GRNs are directly connected to each other by synapses. To examine whether these GRN pairs functionally interact while responding to taste stimuli, we designed a two-channel sensilla recording system to independently manipulate candidate paired pre- and postsynaptic GRNs while simultaneously monitoring their responses through taste-filled recording electrodes. We used this system to test whether activating a GRN identified by transsynaptic mapping alters the activity of its putative postsynaptic partner GRN. Our transsynaptic mapping had revealed five interconnected Gr66a-expressing GRNs in S1, S2, S6, S7, and S10, potentially forming 20 directional GRN pairs. Excluding pairs containing the S1 sensillum (which is difficult to access with electrodes), we tested the remaining 12 candidate GRN pairs in flies with stereotypical labella (see STAR Methods for details).

Gr66a-expressing GRNs are responsive to bitter chemicals, and the five identified GRNs innervating S-sensilla are activated by 1 mM lobeline^20,21. Therefore, we prepared electrodes filled with 1 mM lobeline diluted in water, plus 30 mM tricholine citrate to suppress responses of GRNs sensitive to water²⁹. We first elicited spiking in a candidate postsynaptic GRN by placing the first electrode over its sensillum, then waited until spiking in the GRN, initially rapid, reached a slower steady-state (Figure 3A). We elicited spiking in the postsynaptic GRNs for two reasons: to create a baseline of excitation against which to detect possible inhibition, and to characterize the spike waveforms generated by bitter-responsive postsynaptic GRNs. Next, we elicited spiking in the candidate presynaptic GRN by placing the second electrode over its sensillum (Figure 3A), thus allowing us to measure whether spiking in the candidate presynaptic GRN altered the spiking of the candidate postsynaptic GRN.

We tested each candidate GRN pair with four trials per experiment. Each neuron within a sensillum generates a unique spike waveform, allowing us to isolate individual GRNs using a spike-sorting algorithm³⁰. Throughout each experiment, we held the postsynaptic recording electrode in a stable position to ensure consistent recordings (Figure 3B). After sorting the spikes recorded from the postsynaptic GRN, we calculated its firing rate in the 4 seconds before and after the first spike in the presynaptic GRN in each trial (Figure 3C), and then compared the mean firing rates from each experiment. We found that activating the presynaptic GRN significantly increased the firing rate of the postsynaptic GRN in seven of the 12 GRN pairs (Figures 3D-E; Table S1), establishing their functional connectivity.

Synaptic strength was consistent within each GRN pair across animals but differed significantly between GRN pairs (Figure 3E; Table S1). While some GRN pairs displayed asymmetrical bidirectional interactions, others interacted unidirectionally or showed no functional connectivity (Figure 3E and 3F). From the 12 GRN pairs tested, we selected the four with the strongest functional interactions for further analysis (Figure 3E, left of dotted vertical line). We repeated the electrophysiology experiment using a different bitter tastant, 10 mM caffeine, another known activator of Gr66a-expressing GRNs in S1, S2, S6, S7, and S10^20,21,31. Like lobeline, caffeine elicited consistent and significant increases in postsynaptic GRN activity following presynaptic activation in all four GRN pairs (Figure S4A; Table S1).

Each sensillum houses a mechanosensory neuron along with GRNs³², raising the possibility that placing an electrode over the sensillum could drive mechanosensory responses that confound taste responses. We ruled this out by repeating the experiment described above but omitting the taste chemical from the electrode contacting the presynaptic sensillum. This treatment elicited no significant changes in the firing rates of the postsynaptic GRNs in any pair except one, S2-S7 (Figure S4B; Table S1). In this test, as above, the electrode contained 30 mM tricholine citrate. A previous report showed that 30 mM tricholine citrate moderately activates Gr66a-expressing GRNs in some S sensilla²². To further investigate the S2-S7 GRN pair, we filled the electrode targeting the presynaptic S2 sensillum with distilled water, omitting tricholine citrate. In this condition, the firing rate of the postsynaptic GRN in S7 was not affected (Figure S4B; Table S1). This explains the response we had observed in the S2-S7 GRN pair: 30 mM tricholine citrate had activated the presynaptic Gr66a-expressing GRN in S2, thereby increasing firing in the postsynaptic GRN in S7. Thus, these results suggest that presynaptic GRNs affect postsynaptic GRNs through taste-driven, rather than mechanical, activation. Consistent with this conclusion, in the three other GRN pairs, postsynaptic firing increased reliably and significantly when the presynaptic GRN was stimulated with 1 mM lobeline, but not when it received the control stimulus (Table S1). Our analysis of functional interactions among GRN pairs revealed not all-to-all interconnectivity, but rather a circuit whose strongest connections provide mainly feedforward excitation driven by bitter tastants (Figure 3F).

Because each S-sensillum contains four types of GRNs, it was not immediately clear which cell type had generated which spike waveform, potentially confounding our ability to know which GRN had responded to presynaptic activation. We based our identification of the Gr66a-expressing GRNs on prior reports that Gr66a-expressing GRNs are activated by bitter chemicals^20,21. However, we needed additional experiments to exclude the possibility that other types of GRNs within the sensillum also responded to 1 mM lobeline or 10 mM caffeine.

We used two approaches to identify the synaptically connected neurons. First, we used a genetic manipulation to specifically block synaptic transmission from Gr66a-expressing GRNs by expressing tetanus toxin light chain in them³³. We found that, compared to control, this blockade significantly reduced inter-GRN interactions, confirming that synaptic output from Gr66a-expressing GRNs drives postsynaptic activity (Figure 4A; Tables S2-S3). Second, with an optogenetic manipulation, we tested whether the postsynaptic GRNs participating in synaptic interactions also express Gr66a. For this, we repeated the paired recording experiments described above with electrodes filled with 1 mM lobeline, followed by optogenetic stimulation of each type of GRN housed in S-sensilla. Stimulation was achieved by performing experiments in flies expressing CsChrimson³⁴ driven by different GRN type-specific Gal4 drivers (Figure 4B and Figure S5). After sorting the spikes recorded from the postsynaptic GRN channel, we identified the waveforms of spikes that increased in frequency upon taste-driven activation of the presynaptic GRN, and asked which type of GRN, when activated through optogenetic stimulation, led to an increase in that waveform's frequency. In all GRN pairs, the spike waveform that increased during taste-driven synaptic interaction also increased in frequency when Gr66a-expressing GRNs were optogenetically stimulated (Figure 4B; n=3, N=3 for all four tested pairs). Conversely, when non-Gr66a-expressing types of GRNs—either Gr64f- or glutamatergic Ppk23-expressing²⁴—were optogenetically stimulated, spikes matching that waveform did not increase in frequency. Instead, the matching postsynaptic spiking appeared to be substantially or completely suppressed, possibly through ephaptic interactions among GRNs within each sensillum^18,19 (Figure S5, n≥3, N=3 for all four tested pairs). We did not test Ppk28-expressing GRNs because, in these experiments, they are suppressed by 30 mM tricholine citrate and are thus unlikely to contribute to the functional interactions we observed between GRNs. Our results confirm that functionally interacting presynaptic and postsynaptic neurons are Gr66a-expressing GRNs innervating the five sensilla.

Figure 4. — (A) Pre- (cyan) and postsynaptic (magenta) GRNs were stimulated with 1mM lobeline in 30mM TCC. (A) Synaptic transmission from Gr66a-expressing GRNs is essential for reliable inter-GRN interactions. Mean firing rates of postsynaptic GRNs increased after presynaptic GRNs were activated in Gal4 control flies, but not in flies expressing tetanus toxin light chain (indicated with Xs) in Gr66a-expressing GRNs. (B) The same postsynaptic GRN spike waveforms increased by presynaptic GRN activation increased further upon optogenetic stimulation of Gr66a-expressing GRNs; thus, postsynaptic GRNs express Gr66a. Representative raster plots of four functionally interacting GRN pairs, corresponding activity histograms, and the consistent postsynaptic GRN spike waveform (spikes recorded across full trials are superimposed) are shown. Optogenetic stimulation was delivered 25–29 sec after the onset of recording. n.s.: not significant. See Table S2-S3 for experimental design and statistical analysis. Related to Figure S5.

Together, these electrophysiology experiments demonstrate that seven pairs of Gr66a-expressing GRNs innervating the S sensilla functionally interact through synapses. The only connectivity model consistent with all our observations is a feedforward, activity-promoting circuit (Figure 3F).

nAChR α5 mediates inter-GRN excitatory synaptic interactions.

To assess the circuiťs function, we sought to identify and disrupt postsynaptic receptors mediating the interactions between GRNs. Since all Gr66a-expressing GRNs in the labellum are cholinergic²⁴, we extracted RNA from dissected labella, prepared complementary DNA, and performed a PCR-based screen for all 12 acetylcholine receptor subunit genes known to be expressed in D. melanogaster. Of these, seven receptor genes yielded DNA bands of expected size in two independent PCR amplifications (Figure S6).

To determine which of these receptors are expressed in the functionally interacting Gr66a-expressing GRNs in S sensilla, we used double-immunolabeling to identify cells co-expressing Gr66a and each candidate acetylcholine receptor subunit (Figure S7). To identify GRNs expressing specific receptors, we labeled them with tdTomato using genetic drivers that reflect endogenous expression patterns^35,36. Because a driver for nicotinic receptor α4 was not available, we excluded its gene from the analysis. Of the six tested receptor candidates, only nicotinic acetylcholine receptor subunit α5 (nAChR α5) was found in the functionally interacting Gr66a-expressing GRNs in S sensilla (Figures 5A and S7). Notably, a previous study showed that nAChR α5 subunits can form a homomeric functional acetylcholine-gated ion channel³⁷.

Figure 5. — (A) Representative confocal image of labellum showing overall expression of nAChR α5 (cyan) and Gr66a (magenta). (A’-A”) Higher-magnification views of sensilla S1, S2, S10 (A’) and S6, S7 (A”) showing composite channels (left), nAChR α5 (middle), and Gr66a (right). Dotted circles highlight individual GRNs showing colocalization, indicating positive detections in functionally interacting GRNs. Scale bars, 30 μm (A); 10 μm (A’-A”). (B) nAChR α5 null mutation abolished taste-driven functional interactions in most GRN pairs tested; only S2→S7 GRNs retained inter-GRN interactions, revealing heterogeneous synaptic mechanisms across interacting GRNs. Pre- (cyan) and postsynaptic (magenta) GRNs were both stimulated with lobeline in TCC. nAChR α5 null mutants with restored nAChR α5 expression selectively in Gr93a-Gal4-expressing GRNs (E), but not control genotypes (C-D), show inter-GRN interactions in all tested pairs. Lines connect paired measurements. n.s.: not significant. See Table S3-S4 for experimental design and statistical analysis. Related to Figures S6-S8.

If nAChR α5 mediates functional interactions between these GRNs, disrupting the receptor gene should abolish the ability of presynaptic GRNs to affect spiking in postsynaptic GRNs. To test this, we used CRISPR/Cas9 to generate nAChR α5 null mutants (Figure S8). Indeed, in mutant flies, activating presynaptic GRNs did not significantly change postsynaptic firing in three of the four GRN pairs tested (Figure 5B; Tables S3-S4); synaptic interactions in the S2→S7 pair remained detectable (Figure 5B; Table S4), suggesting it may use different acetylcholine receptors or neurotransmitters. In mutant flies, using Gr93a-Gal4 to restore nAChR α5 expression specifically in these GRNs rescued their synaptic function (Figure 5C-E; Tables S3-S4).

We next asked whether the loss of detectable synaptic interactions at individual GRN pairs impaired the reliability of signal amplification across the inter-GRN circuitry. In wild-type flies and in mutants with GRN-specific nAChR α5 rescue, nearly every recording showed an increase in postsynaptic firing following presynaptic activation; in nAChR α5 mutants, this fraction was markedly reduced in all GRN pairs except S2→S7. Across the circuit, presynaptic activation drove a significantly less consistent postsynaptic firing increase in mutants than in wild-type or rescue flies (STAR Methods; Table S3). Together, these results indicate that nAChR α5 subunits form the homomeric acetylcholine-gated ion channel mediating, in a cell-autonomous manner, excitatory synaptic interactions in most pairs of the inter-GRN circuitry, and further suggest that nAChR α5 promotes taste signals by increasing the reliability of inter-GRN interactions.

The feedforward amplification circuit enhances behavioral reliability.

Next, we asked whether and how the feedforward circuit of functionally interacting GRNs contributes to gustatory behavior. Activating Gr66a-expressing GRNs is known to suppress feeding and other taste-driven behaviors in several ways^20,38,39, leading us to hypothesize that the GRNs comprising the feedforward circuit contribute to this suppression. Because the circuiťs feedforward interactions appear to integrate the responses of multiple GRNs to tastants, we further hypothesized that the circuit helps amplify and stabilize this suppressive effect. To test this, we used an automated Fly Liquid Interaction Counter (FLIC) system^40–42 to examine how wild-type and genetically manipulated, individual freely-moving flies interact with a food source when the feedforward circuit is active.

To interrogate the feedforward circuit, we used Gr93a-Gal4 to express the light-gated cation channel CsChrimson in functionally interacting GRNs³⁴. All-trans-retinal (ATR)-fed flies were monitored through three phases: an initial feeding period (“Before”), an “On” period featuring closed-loop illumination triggered by gustation (Figure S9), and a final recovery period (“After”) (Figure 6A). Because bitter-responsive GRNs suppress feeding, we predicted that gustation bouts in “On” periods would be shorter than in “Before” and “After” periods in ATR-fed flies expressing CsChrimson in the group of functionally interacting GRNs (“Gr93a>CsChrimson +ATR”), but not in the control groups. As hypothesized, only the “Gr93a>CsChrimson +ATR group” exhibited a significant reduction in median gustation bout duration during illumination. This effect was specific to the “On” period and was absent in all control groups (Figure 6B and C; Table S5).

Figure 6. — (A) Representative examples of gustation bouts (dots) in single animals of indicated genotype measured by FLIC (Fly Liquid-food Interaction Counter) before, during (“On”), and after closed-loop optogenetic activation. Closed-loop illuminations triggered by initiation of gustation bouts were engaged only during the On period, suppressing gustation in ATR-fed flies expressing CsChrimson in Gr93a-Gal4-expressing GRNs. (B-C) Group results: (B) median gustation bout duration for individual flies of genotype indicated in panel A; optogenetic activation reduced feeding bout duration. Lines connect measurements from the same fly. (C) Median gustation bout duration compared by genotype for each condition. (D-E) Optogenetic illumination reduced the number of gustation bouts (mean number of gustation bouts per illumination shown in D) and the duration of initial gustation bouts (median duration of first gustation bouts per illumination shown in E). (F) Circuit models of interacting Gr93a-Gal4-expressing GRNs of the indicated genotypes during optogenetic stimulation. Arrow thickness indicates functional synapse strength. Dark pink: GRNs responding to both optogenetic activation and feedforward synaptic input; light pink: GRNs whose feedforward synaptic input is partly or completely blocked by receptor knockout; gray: nonresponsive GRNs. (G) Comparison of variances in the median gustation bout duration among the animals of indicated genotypes. Dotted line at 0.8 s (B-C, E, G) indicates the shortest median duration observed across all control groups during the On period. Test groups exhibiting median durations of gustation bouts equal to or exceeding 0.8 seconds show less reliable behavioral control of feeding. n.s.: not significant. See Table S5, Data S4 and S5 for experimental design and statistical analysis. Raw data are available in Data S1, S2 and S3. Related to Figures S8-S11.

To better understand how the functionally interacting GRNs suppress gustation, we tested two possibilities: that these GRNs reduce the initiation of gustation bouts, and that they induce premature termination of ongoing bouts. We tested the first possibility by asking whether there were fewer individual gustation bouts in each illumination in the “Gr93a>CsChrimson +ATR” group than control groups. Indeed, this was the case (Figure 6D; Table S5). We then tested the second possibility by asking whether the median duration of first gustation bouts in each illumination was shorter in the “Gr93a>CsChrimson +ATR” group than in control groups. (First gustation bouts begin before the onset of illumination and thus would not be affected by optogenetic activation of the functionally interacting GRNs at the initiation step; Figure S9). We found bout duration was indeed shorter in the “Gr93a>CsChrimson +ATR” group than control groups, indicating that activating these GRNs induced premature termination of gustation bouts (Figure 6E; Table S5). These results demonstrate that activating the functionally interacting GRNs suppresses gustation by inhibiting its initiation and inducing its premature termination.

To determine whether disrupting the feedforward GRN circuitry alters the suppression of gustatory behaviors, we tested the nAChR α5 null mutant, which blocks synaptic interactions among most of the functionally interacting GRNs, in the FLIC apparatus (see nAChR α5 KO, Gr93a-Gal4>CsChrimson in Figure 6F). We found the feedforward circuit does not drive gustatory suppression per se; instead, the circuit enhances its reliability. Specifically, the nAChR α5 null mutation did not block gustatory suppression, nor did it affect the median duration of gustatory behavior in the “On” period (Figure S10; Data S4.xlsx). However, gustatory behavior in flies with blocked feedforward circuitry was substantially and significantly less reliable than in control flies, with increased variance in bout duration only during the “On” period. Gr93a-Gal4-driven expression of the receptor in these GRNs in mutant flies rescued response reliability to the control level (Figure 6G, Table S5), indicating that the greater behavioral variance was caused by blocking excitatory synaptic interactions among the functionally interacting GRNs. Notably, disrupting the feedforward circuit substantially increased the proportion of animals showing prolonged responses: 21% of the mutants (nAChR α5 KO, Gr93a-Gal4>CsChrimson; 10/48) exhibited a median response duration equal to or exceeding 0.8 seconds, which is the shortest median duration observed across all control groups during the “On” period. In contrast, only 7–8% of animals expressing the endogenous receptor (Gr93a-Gal4>CsChrimson +ATR; 2/25) or the rescue construct (nAChR α5 KO, Gr93a-Gal4>CsChrimson, cDNA rescue; 2/41) showed durations at or above this threshold (Figure 6G). Thus, mutant flies were three times more likely to show unreliable feeding suppression than control and rescue groups.

In the labellum, the Gr93a-Gal4 driver we used for behavioral testing is expressed exclusively in functionally interacting GRNs. However, we observed additional expression in a small number of non-gustatory neurons in the second and third antennal segments (Figure S11A-A’). These antennal neurons project to the antennal mechanosensory and motor center (AMMC) and the antennal lobe (AL), respectively (Figure S11A”).

To determine whether these non-labellar neurons contribute to the feeding suppression observed during Gr93a-Gal4-driven CsChrimson stimulation, we repeated the FLIC assay using antenna-ablated flies and mock-operated controls. Both groups exhibited robust suppression of gustation (Figure S11B). Further, antenna-ablated flies showed no increase in median gustation duration or variance relative to controls, ruling out a role for antennal neurons in this behavioral response (Figure S11C; Data S5.xlsx). While antenna-ablated flies displayed shorter median gustation durations overall, it is unclear whether this reflects a functional role for the antennae in promoting feeding or a nonspecific effect of tissue injury.

Together, our results establish that excitatory synaptic interactions among functionally interacting GRNs enhance the reliability of gustatory behavior suppression.

Discussion

A peripheral feedforward amplification circuit for behavioral reliability

Our study reveals a previously unknown gustatory circuit consisting of monosynaptically connected, similarly-tuned bitter-sensing GRNs. These connections, mostly mediated by nAChR α5 receptors, vary in functional strength and direction between cell pairs. Our study did not investigate the specific mechanism mediating this synaptic interaction, but we speculate the spikes we detected in postsynaptic GRNs reflect backpropagating action potentials: axo-axonic synaptic inputs from presynaptic GRNs depolarize the axon and trigger action potentials that propagate retrogradely to the dendrite, where they are detected by tip recording. Together, the interacting GRNs constitute a feedforward network at the extreme sensory periphery that promotes the reliability of neural activity. These findings challenge traditional models of gustation by revealing that peripheral receptor neurons interact directly in circuit-level computations to integrate and stabilize sensory signals, ensuring reliable gustatory behaviors.

Notably, this inter-GRN circuitry is composed of a single subclass of bitter-sensing neurons with similar properties, all responding to a broad range of aversive taste chemicals by firing at relatively low rates²⁰. In some ways, the circuitry seems reminiscent of olfactory glomeruli, in which the synaptic outputs of multiple, similarly-tuned olfactory sensory neurons are integrated to enhance signal strength, add structural redundancy, and reduce noise from local odorant fluctuations⁴³.

Our behavioral experiments demonstrate that activating Gr93a-expressing GRNs suppresses gustatory behavior by both inhibiting its initiation and inducing premature termination. Our cell-specific rescue experiments establish that inter-GRN synaptic interactions contribute to stabilizing rather than simply amplifying signals that suppress feeding. In this experiment, all Gr93a-expressing GRNs were optogenetically activated, including GRNs within the nAChR α5 knockout mutant, in which only the synaptic interactions between GRNs were selectively disrupted. Notably, in mutants, behavioral response reliability was substantially and significantly impaired. We observed no changes in gustatory bout duration, likely because activating the GRNs provided sufficient drive to suppress gustatory behavior in most trials, leaving any change in bout duration undetectable.

How is this outcome achieved? Our results show that the mechanism ensuring behavioral reliability begins in the periphery but likely depends on the combined action of the inter-GRN feedforward circuit and downstream follower neurons whose role we have not directly tested. Our analysis of presynaptic-driven postsynaptic GRN activation showed that nAChR α5 is essential for reliable synaptic interactions within this circuit. We speculate that these stabilized interactions allow the feedforward circuit to amplify and integrate individually weak GRN signals into a stronger and more reliable collective output, providing the excitatory drive needed to cross a response threshold in follower neurons, and thereby robustly engage the pharyngeal motor neurons that suppress gustatory behavior. In the nAChR α5 mutant, the loss of this stabilization may cause excitatory drive to fall short of this threshold, resulting in inconsistent motor neuron engagement. There is precedent for this kind of integrative circuitry: in the mouse midbrain, for example, feedforward networks linking individually weak and unreliable synapses between the medial superior colliculus (mSC) and the dorsal periaqueductal gray (dPAG) are essential for surpassing synaptic thresholds required to initiate escape responses⁴⁴. Our results suggest feedforward integration may be a general motif for converting weak or variable signals into robust outputs.

Studies of feeding in Drosophila point to a practical approach for testing these ideas. Drosophila feeding behaviors proceed through a stereotyped, well-characterized sequence: flies approach the food, suppress ongoing locomotion, adjust their posture to establish contact, initiate a proboscis extension reflex (PER), and, finally, retract the proboscis^45,46. Several of these actions are driven by motor neurons in the subesophageal zone (SEZ) that innervate pharyngeal musculature⁴⁶, including motor neurons downstream from Gr66a-expressing GRNs that facilitate proboscis retraction and suppress sucrose-evoked extension⁴⁷.Testing our model directly at this level of the circuit will clarify how integration at the sensory periphery translates into reliable behavioral control.

Combining multiple approaches to assess function

Our transsynaptic mapping focused exclusively on homotypic GRN pairs, leaving for future investigation the identities of the heterotypic connections we observed. Although the focus of our work precludes a full comparison with recent connectome-based analyses^25,48, a preliminary comparison between those datasets and our genetic tracings revealed several notable discrepancies. We did not observe the synaptic connections between Ppk23-expressing GRNs and other GRN types predicted by the connectome analysis, nor did we observe predicted synapses between Ir94e-expressing GRNs. Despite the sparse synaptic connections reported in connectome datasets, we provide electrophysiological evidence for significant functional communication between Gr66a-expressing GRNs. Discrepancies like these could arise from incomplete connectome coverage of distal GRN connections or the different expression strengths of genetic drivers used for our transsynaptic mapping. Yet these discrepancies also highlight the complementary value of functional approaches for revealing interactions that structural analyses alone cannot predict.

Our findings reveal unexpected diversity among pairs of GRNs expressing the same receptor type, including differences in wiring patterns, signaling properties, and synaptic strength, opening new avenues for understanding the circuit organization, mechanisms, and functional logic of inter-GRN interactions. To better understand this selectivity and heterogeneity, a more complete description of the circuitry, including its embedding within networks receiving input from other GRNs and feedback from higher gustatory areas⁴⁸, will be needed. We found synaptic connections between GRNs sharing broad tuning for bitter tastants. It will now be interesting to determine whether synaptic connectivity among other Gr66a-expressing GRNs correlates with tuning properties. Our study also revealed anatomical variation among individual flies, including differences in the number and distribution of Gr66a-expressing GRNs (in atypical labella). These observations suggest that some types of structural variability are common in the peripheral gustatory system, though the functional consequences for circuit computation remain to be determined. Further work combining multiple types of analyses will deepen our understanding of peripheral gustatory circuits and more broadly illuminate the principles by which sensory systems establish reliable behavioral control at the earliest stages of neural processing.

Resource Availability

Lead Contact:

Requests for resources or further information should be directed to and will be fulfilled by the lead contact, Mark Stopfer (stopferm@mail.nih.gov)

Materials Availability

There are no restrictions on material availability of any reagent produced in this work. All plasmids generated in this study will be made available upon request. All fly stocks generated in this study (Key resources table) will be made available through the Bloomington Drosophila Stock Center (https://bdsc.indiana.edu/) upon publication. Fly stocks, primer sequences used in this study are summarized in the Key resources table.

Key Resources Table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-GFP Chicken polyclonal Antibody	aves labs	GFP-1020
Anti-RFP Rabbit polyclonal Antibody Pre-adsorbed	Rockland	600-401-379
Anti-nc82 Mouse Antibody	Developmental Studies Hybridoma Bank	nc82 Supernatant
Goat anti-Chicken IgY (H+L) Secondary Antibody, Alexa Fluor^™ 488	Thermo Fisher Scientific	A-11039
Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 568	Thermo Fisher Scientific	A-11011
Goat anti Mouse IgG (H+L) Secondary Antibody, Alexa Fluor 633, Invitrogen	Thermo Fisher Scientific	A-21052
Chemicals, peptides, and recombinant proteins
all-trans-retinal (ATR)	Sigma-Aldrich	R2500
Lobeline hydrochloride	Sigma-Aldrich	141879
Caffeine	Sigma-Aldrich	W222402
Sucrose	Sigma-Aldrich	S0389
Dimethylsulfoxide (DMSO)	Sigma-Aldrich	34869
Tricholine Citrate	Sigma-Aldrich	T0252
Critical commercial assays
Direct-zol RNA Microprep kit	Zymo Research	R2061
Superscript IV one-step RT-PCR system	Thermo Fisher	12594025
FirstChoice RLM-RACE kit	Thermo Fisher	AM1700
GeneArt Gibson Assembly HiFi Master Mix	Thermo Fisher	A46627
Deposited data
Raw and analyzed data	This study	Supplementary Table 1-7, Zenodo DOI: https://doi.org/10.5281/zenodo.18887927
Experimental models: Organisms/strains
y[1] w[*] P{y[+t7.7] w[+mC]=UAS-myrGFP.QUAS-mtdTomato-3xHA}su(Hw)attP8; P{y[+t7.7] w[+mC]=trans-Tango}attP40	Bloomington Stock Center	RRID: BDSC 77124
w[*]; wg[Sp-1]/CyO; P{w[+mC]=Gr66a-GAL4.1.8}1/TM3, Sb[1]	Bloomington Stock Center	RRID: BDSC 57670
w[1118]; P{w[+mC]=ppk23-Gal4.2.695}2; TM2/TM6B	Bloomington Stock Center	RRID: BDSC 93026
[*]; P{w[+mC]=UAS-mCD8::GFP.L}LL5; P{w[+mC]=Gr64f-GAL4.9.7}1/TM3, Sb[1]	Bloomington Stock Center	RRID: BDSC 57668
[1118]; Bl[1]/CyO; P{w[+mC]=ppk28-GAL4.C}3	Bloomington Stock Center	RRID: BDSC 93020
w[*]; P{w[+mC]=UAS-mCD8::GFP.L}LL5/CyO; P{y[+t7.7] w[+mC]=Ir94e-GAL4.K}attP2	Bloomington Stock Center	RRID: BDSC 60725
w[*]; P{w[+mC]=Gr47a-GAL4.7.321}2/CyO; Dr[1]/TM3, Sb[1]	Bloomington Stock Center	RRID: BDSC 57638
[*]; wg[Sp-1]/CyO; P{w[+mC]=Gr59c-GAL4.5.5}6	Bloomington Stock Center	RRID: BDSC 57650
w[*]; wg[Sp-1]/CyO; P{w[+mC]=Gr93a-GAL4.5}2DD	Bloomington Stock Center	RRID: BDSC 67679
w[*]; wg[Sp-1]/CyO; P{w[+mC]=Gr22f-GAL4.2.5}3	Bloomington Stock Center	RRID: BDSC 57609
Canton-S	Quan Yuan lab
P{w[+mC]=UAS-TeTxLC.tnt}C1, w[*]	Bloomington Stock Center	RRID: BDSC 28996
w[1118] P{y[+t7.7] w[+mC]=20XUAS-IVS-CsChrimson.mVenus}attP18	Bloomington Stock Center	RRID: BDSC 55134
y[1] w[*]; Mi{Trojan-GAL4.2}nAChRalpha5[MI13859-TG4.2]/SM6a	Bloomington Stock Center	RRID: BDSC 76755
w[*]; Bl[1]/CyO; P{w[+mC]=LexAop-CD2::HRP}3/TM6B, Tb[1]	Bloomington Stock Center	RRID: BDSC 56522
[1118], P{w[+mC]=Gr66a-lexA.S}1; Bl[1]/CyO; TM2/TM6B	Bloomington Stock Center	RRID: BDSC 93023
y[1] w[*] P{y[+t7.7] w[+mC]=10XUAS-IVS-myr:: tdTomato}su(Hw)attP8	Bloomington Stock Center	RRID: BDSC 32223
y[1] sc[*] v[1] sev[21]; P{y[+t7.7] v[+t1.8]=WKO.1-E4}attP40	Bloomington Stock Center	RRID: BDSC 82478
w[*]; PBac{y[+mDint2] RFP[DsRed.OpIE2]=nanos-Cas9.T2A.GFP}VK00033	Bloomington Stock Center	RRID: BDSC 79004
nAchR alpha5 receptor null mutant	This study
w[1118]; P{w[+mC]=nAchR alpha5 cDNA}/TM6B	This study
w[1118]	Bloomington Stock Center	RRID: BDSC 5905
y[1] w[*]; Mi{Trojan-3XGAL80.0}ChAT[MI04508-T3XG80.0] CG7715[MI04508-T3XG80.0-X]/TM3, Sb[1]	Bloomington Stock Center	RRID: BDSC 84649
TI{2A-GAL4}mAChR-A[2A-GAL4]/CyO	Bloomington Stock Center	RRID: BDSC 84649
TI{2A-GAL4}mAChR-B[2A-GAL4]/TM3, Sb[1]	Bloomington Stock Center	RRID: BDSC 84650
w[*] TI{2A-GAL4}nAChRalpha3[2AGAL4]	Bloomington Stock Center	RRID: BDSC 84664
y[1] w[*] Mi{Trojan-GAL4.1}nAChRalpha7[MI12545-TG4.1]	Bloomington Stock Center	RRID: BDSC 77828
TI{2A-GAL4}nAChRbeta2[2A-GAL4]	Bloomington Stock Center	84666
Oligonucleotides for PCR screen
ATGAAGTGGTTTCAAGTGACCATAGA		CG2302_F1
TGGTCATTATCTGGTTCTTCAGATTG		CG2302_R1
GTGACCATAGATATGATTCTTGTGCT		CG2302_F2
CATCGATCAGTTGCGATAGCTTC		CG2302_R2
ATGGACTCCCCGCTGCCA		CG4128_F1
GAGAAGCTGATTCTTCTCGTCC		CG4128_R1
GTCCTGTTGATCTTTCTGGCGAT		CG4128_F2
GATGATCTGCTGCAGCGTCAG		CG4128_R2
TCACATTCAAGGCATATGAGAGAGTA		CG4356_F1
TGCGTACTGTGTCACAGACACTA		CG4356_R1
TACATACAGTGTAAGCAAACGACATCC		CG4356_F2
GACACTACAAATGTCGCTTTGATTTCG		CG4356_R2
CGGCTGGCTGGTGGAATC		CG5610_F1
ACCAACGAAACGGATGACTCGT		CG5610_R1
GTGGAATCTAATCATCAG		CG5610_F2
ACAATAACCTTTACAGCA		CG5610_R2
ACTCTGATGACGAAAGCCACTC		CG6798_F1
TTCCGTGAGATCGATGCCAACTT		CG6798_R1
CTGAAGTATACGGGAGAAGTCTTTTGG		CG6798_F2
CTTGCACTAGGTTACTGCCAGG		CG6798_R2
AGGACGAGTTCAATGCGGAAGA		CG6844_F1
CTACGTTCTCATGAATTTAATTCTTCTTCT		CG6844_R1
AGTTCAATGCGGAAGATCAGGAC		CG6844_F2
CGGTTAGATTGTAGATTTGCTTGGC		CG6844_R2
ATGAAAAGCAGACATATGAAGGGG		CG7918_F1
TCGTCGAAAAAGTTTGAACACAGTTTATA		CG7918_R1
GGTAGGCCATAAGAAGACCAAAGAAAAG		CG7918_F2
GCAGGATTGACATCGAACTGCTTAAC		CG7918_R2
TACTTTCGATCCCATGCTATTACATTTT		CG8109_F1
TGTACTTCCATCCTTGCCTCG		CG8109_R1
CGGCAGCAATTCATTCATTATTCACTATC		CG8109_F2
CTTTGTACTGCATATTGGTTTTATCAAGTTTTAC C		CG8109_R2
GAAGAGCGCTTGGTGCGTG		CG11348_F1
CAGACGCAACACCCCGATG		CG11348_R1
ATGGAGTCTTCCTGCAAATCCTG		CG11348_F2
GAAAAGGCCACAAGCACCAG		CG11348_R2
TAATCTTCTACAGCACCAGCCTG		CG11822_F1
ACTCCTGCTCGTCCGCCT		CG11822_R1
ATGACGACGACTCCCAAGATAAAG		CG11822_F2
CAACTCGTCGATGTCTATGTAGGTGAC		CG11822_R2
GGTCACACTATCTATCTCCATTCTTAT		CG12414_F1
GCTGGCAAAAAGTCAATAAAGACG		CG12414_R1
GTCTTCACGTGTTCTTTCTTTTGGTG		CG12414_F2
CTTAACCCAGGGTGCCATTTTATG		CG12414_R2
CCAACAACCAACAAACATAAGACTGTG		CG32975_F1
GGATACTCTATCATGTCTCGATATCGGAATAC		CG32975_R1
CGACGAAAAAGAAAACCAGCAACC		CG32975_F2
CTCGATATCGGAATACTGAACTCATTGTC		CG32975_R2
Recombinant DNA
PJFRC7-20XUAS-IVS-mCD8::GFP	Addgene	RRID: 26220
UAS-nAChR α5	This study
Software and algorithms
Zen	Zeiss
ImageJ/Fiji	NIH	RRID: SCR_003070
Imaris	Bitplane	RRID: SCR_007370
Matlab	Mathworks	RRID: SCR_001622
Labview	National Instrument	RRID: SCR_014325
R studio	Posit Software	RRID: SCR_000432
Jupyter Notebook	Python.org	RRID: SCR_008394
Prism 9	Graphpad	RRID: SCR_002798
Other
Normal Goat Serum	Thermo Fisher Scientific	01-6201
Poly-L-lysine hydrobromide	Sigma-Aldrich	P1524
Vectashield plus	Vector Laboratories	H1900
Kwik-sil	WPI
Trizol	Thermo Fisher Scientific	15596026
Optolid	Sable Systems International	FLIC-TOP-Opto12-R

Open in a new tab

Data and Code Availability

Data: raw data for figures have been deposited on Zenodo (DOI: https://doi.org/10.5281/zenodo.18887927) and are publicly available as of the date of publication.
Code: Python code for FLIC analysis has been deposited on Github (https://github.com/joshuabhk/flic-analysis). Matlab code for spike sorting has been deposited on bitbucket (https://bitbucket.org/zaldworth/stopferlabspikesorter/src/main/). Both codes are publicly available.
Additional information: Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Drosophila stocks

Flies were grown on Jazz mix culture medium (Fisher Scientific).

Flies not used in behavioral experiments were raised at 25°C on 12hr light: 12hr dark cycle and 70% humidity. Fly lines provided from the Bloomington Drosophila Stock Center and created in this study were used; RRIDs are listed in the Key resources table. nAChR α5 mutant lines created in this study (Figure S8) were backcrossed five times against strain w1118.

Age and cultivation:

Flies for trans-Tango were collected within four days of eclosion and were then kept for 15 days at 21°C before they were dissected. For other expression studies, flies collected within four days of eclosion were then kept for seven days at 25°C before they were dissected. For all electrophysiology experiments, flies were collected within four days of eclosion, transferred to fresh food, and kept for four days at 25°C prior to the experiments. Unless indicated as “-ATR,” flies for behavioral experiments were grown on Jazz mix culture medium containing 1.5mM all-trans-retinal (ATR) and kept in constant dark: culture vials were wrapped in foil to protect ATR from light-induced degradation. Flies were collected within four days of eclosion and kept for five days in foil-wrapped vials containing the culture medium with ATR at 25°C. “–ATR” flies were treated the same as ‘+ATR’ flies, except they were raised on ATR-free medium. For some experiments, antennae were removed as follows: individual flies were anesthetized on ice, secured in a custom-made collar, and both antennae were removed at the base with forceps immediately before the behavior test. Mock-treated control flies underwent the same procedure except antennae were left intact.

Sex reporting:

To avoid possible sexual dimorphisms, only mated females were used in all our experiments. Accordingly, the generalizability of findings to males was not assessed.

METHOD DETAILS

Electrophysiology

Recording setup:

Two-channel over-the-sensilla recordings were made using a chlorided silver wire electrode (A-M systems, 782500, 0.01” silver wire), glass electrodes, a preamplifier (Grass, P55 AC preamplifier) and an amplifier (Precision, Model 440), and were digitized at 10KHz and recorded by a custom program (Labview). Glass electrodes were used for electrical reference (tip diameter ~ 30um) and for sensilla recordings (tip diameter = ~15um). Flies were prepared and single sensilla recording was performed as described in²⁶, with some modifications.

Fly preparation:

Briefly, a fly was immobilized by inserting a sharp glass reference electrode filled with Drosophila Ringer’s solution through the posterior end of the thorax to the tip of labellum. The reference electrode was mounted at such an angle that all S-sensilla in the right hemisphere of the labellum were in roughly the same focal plane. Recordings were made only from S-sensilla on the right hemisphere. Although we found with trans-Tango experiments that the presence of S0 sensilla in some flies affected neither the spatial distribution nor synaptic connectivity of Gr66a-expressing GRNs innervating S sensilla (Figure S2), we restricted our analysis to flies whose labella have only S1-S10 sensilla to avoid any potential variations in wiring and to exclude GRN pairs difficult to access with electrodes.

Experimental procedure:

All recordings were performed at room temperature. Before each experiment, we evaluated the health of the target GRNs by briefly contacting a cognate S-sensilla with a tastant-filled electrode; only S sensilla that responded to the tastant by spiking were used for experiments. To test whether spiking in one GRN could affect the spiking of another, two glass electrodes filled with tastant solutions were used to stimulate and record from putative pre- and postsynaptic GRNs simultaneously. We activated the specific GRN pairs identified in our trans-Tango transsynaptic mapping using 1 mM lobeline or 10 mM caffeine, both of which are established activators of Gr66a-expressing cells²⁰. An electrode was first placed over a sensillum housing a putative postsynaptic GRN. Because this electrode included a tastant solution, the GRN typically began to spike. Recording trials began when the spiking, initially rapid, reached a slower steady state, typically after 1–2 mins. After 4 sec, a second recording electrode was placed over a different sensillum housing the putative presynaptic GRN and was held in place until the trial ended (total length of each recording trial was 15–20 secs except for trials shown in Figure 4B and Figure S5 that were 35secs). The first electrode, contacting the putative postsynaptic GRN, was held in place across all trials. To monitor the stability of the recording, spike waveforms were inspected throughout the experiment. To perform multiple trials, the second electrode was retracted and replaced onto the same putative presynaptic sensillum three times when testing with a 10mM caffeine solution, or four times with a 1mM lobeline solution, with an interstimulus interval of at least 60sec. Experiments in which a fly twitched its proboscis were excluded from analysis. We recorded from several GRN pairs in each animal, rinsing the labellum with distilled water between recordings from each pair. Per condition, six to nine animals were tested.

Optogenetic Validation:

Sensilla contain multiple GRNs, each producing spikes with distinct waveforms. To further validate the identity of postsynaptic GRNs, we isolated the spike waveform that increased its firing rate when the presynaptic GRN was activated and used an optogenetic approach to test which postsynaptic GRN subtype generated the waveform. For this, we made paired recordings using electrodes filled with 1 mM lobeline on the four GRN pairs that exhibited the strongest functional interactions in flies expressing CsChrimson under the control of type-specific Gal4 drivers—Gr66a, Gr64f, and Ppk23^vglut. Ppk23^vglut, which targets only glutamatergic Ppk23-expressing GRNs, was chosen for CsChrimson expression because cholinergic GRNs of this type also express Gr66a²⁴. In each trial, a 4-sec illumination (631± 33nm LED (Zeiss Colibri 5 system) at ~7mWatt/cm²) was delivered 25 seconds after recording began to evoke spiking from the CsChrimson-expressing GRNs³⁴. The timing and intensity of illumination were controlled by custom Labview (National Instruments) code driving the Colibri 5 system. Per condition, three or more animals were tested.

Spike sorting and Data analysis

General:

To assign spikes recorded from sensilla to individual GRNs, we used a semiautomated, model-based spike sorting method described previously³⁰. This method compares the whole waveforms of spike signals for each trial of an experiment to a waveform template built from one trial. After sorting extracellular spike waveforms, we made quantitative estimates of cluster quality throughout each experiment. We included only experiments in which the sorted waveforms had physiological shapes, were well isolated in principal component space, and estimated Type 1 and Type 2 errors caused by overlap between clusters or the occurrence of nonphysiological interspike intervals were less than 5% of the dataset. To ensure unbiased selection of units, spike sorting and unit selection were performed before any further analysis. The 4-sec windows labeled 'Before' and 'After' (Figure 3C) were aligned to the timing of the first spike from the presynaptic GRN. When stimulation with 30 mM Tricholine citrate or distilled water failed to elicit a spike, the time of contact between the solution in the electrode and the sensillum was used as the reference point. Trials in which a taste chemical used to stimulate the presynaptic GRN elicited firing of <1Hz were excluded from the analysis. The Matlab code for spike sorting is available at: https://bitbucket.org/zaldworth/stopferlabspikesorter/src/main/

Tastant stimuli

For electrophysiological recordings, 30mM tricholine citrate, an electrolyte that also reduces responses to water, was used alone as a control, or added to tastant solutions used for electrophysiology. 1mM lobeline and 10mM caffeine were chosen to activate Gr66a-expressing GRNs in cognate sensilla²⁰. Consistent with a prior report²², we found that 30mM tricholine citrate sometimes activated the GRNs, especially the one in S2. To prevent caffeine from crystallizing at the tip of the recording electrodes, DMSO (final concentration 5%) was included. Recording electrodes were replaced if solutions crystallized. For the FLIC assay, 30mM sucrose solution was used as food. Detailed information about chemicals used in this study is provided in the Key resources table.

Molecular biology

PCR screen for AChR expression:

Total RNA was extracted using phenol-chloroform (Allen, E. (2016) RNA Extraction from Drosophila Tissues using TRIzol Reagent V.2, protocols.io. DOI: dx.doi.org/10.17504/protocols.io.fgtbjwn) from 250 labella dissected from female adult flies. Using Superscript IV one-step RT-PCR system, extracted RNA was used to screen for receptor candidates expressed in the labella. For each target receptor gene, two separate sets of primers were chosen to yield two different amplicons (see Key resources table).

Generation of nAChR α5 null mutant:

To generate a specific deletion of the nAChR α5 gene by CRISPR/CAS9, gRNAs targeting all four isoforms encoded by the gene and CAS9 protein were expressed during early embryonic development by crossing transgenic flies that expressed either gene. Progeny expressing both gRNAs and CAS9 were crossed to balancer flies to establish fly lines carrying individual mutations. Subsequently, 2–3 individual flies from each line were used for sequencing the target site in their genomes. Lines carrying nonsense mutations were identified with an online translation tool (https://web.expasy.org/translate/) to search for mutations that result in premature termination of all four isoforms.

Cloning of nAChR α5 cDNA and generation of a rescue strain:

Total RNA was extracted from 100 dissected labella from female adult flies using Trizol and the Direct-zol RNA Microprep kit. Complete cDNA that encodes nAChR α5 was cloned using the FirstChoice RLM-RACE kit. In brief, after ligating adaptors at either the 5’ or 3’ end of extracted mRNA and reverse-transcribing the RNA into cDNA, we performed 5’- and 3’-RACE (Rapid Amplification of cDNA Ends) on cDNAs encoding nAChR α5. Gene specific primers were designed so that, with an overlapping sequence for PCR stitching, the two resulting amplicons from the 5’- and 3’-RACE reconstituted the full coding region of nAChR α5 cDNA that was identical across all 4 isoforms. The cDNA followed by T2A sequence was cloned upstream of mCD8::GFP sequence in an expression vector using GeneArt Gibson Assembly HiFi Master Mix. The resulting expression vector was injected with a helper plasmid expressing phi31 to an nAChR α5 null mutant (#9) carrying an attp2 docking site (constructed from RRID: BDSC_8622).

Immunohistochemistry, Microscopy and Image Processing

Immunohistochemistry experiments were conducted mainly as described²⁴. Anesthetized flies were briefly submerged in 70% ethanol and transferred to cold PBS. After dissection, proboscises, antennae, or brains with ventral nerve cords were incubated in fixative (4% paraformaldehyde dissolved in PBS) for 30 min at room temperature on a rocker platform. After four washes in PBS/1% Triton X-100, tissues were incubated in a blocking solution (10% goat serum in PBS/1% Triton X-100) for 30 min at room temperature on a rocker platform. Subsequently, samples were incubated with primary antibodies (anti-GFP chicken (1:1000), anti-RFP rabbit (1:500), anti-nc82 mouse (1:50)) for 3–5 days, then washed and incubated with secondary antibodies (anti-chicken 488, anti-rabbit 568, anti-mouse 633, 1:200) overnight at 4°C on a rocker platform. After clearing proboscises in 100% glycerol, the prepared tissues were mounted on a glass slide using Kwik-sil, set under a cover glass, and immediately infused with 100% glycerol. After sealing the cover glass with nail polish, confocal serial optical sections at 1um intervals were taken by a Zeiss 710/780/900 laser scanning microscope under a LCI Plan-Apochromat 25x objective (NA 0.8). To enhance the visibility of sensilla, a maximum filter of radius 2 was applied to the DIC (Differential Interference Contrast) channel. A Gaussian filter with sigma 2 was also applied to each fluorescence channel to reduce noise. After processing in ImageJ, images were imported into Imaris where color was assigned to each channel and the intensities of fluorescence signals were adjusted. Individual S-sensilla were identified with reference to S6, which is longer and more lateral than the others. Antennae and brains were cleared with Vectashield Plus and mounted on poly-L-lysine coated slides (Description: https://www.janelia.org/sites/default/files/FL%20Recipe%20-%20Poly-L-Lysine%20.pdf).

Behavior

FLIC assay setup:

Automated FLIC (Fly Liquid Food Interaction Counter) assays were performed as described in⁴⁰ with modifications. The FLIC system consists of food reservoirs, a Master Control Unit (MCU), optolids for delivering illumination yoked to behaviors and sets of behavior test chambers containing single feeding wells (DFMs, Drosophila Feeding Monitors). Briefly, wells in the DFMs were filled to the brim with 30 mM sucrose. Mated adult female flies (5–9 days post-eclosion) were anesthetized on ice and loaded singly into each test chamber. Every 200 milliseconds, the MCU recorded voltage signals reflecting gustation bouts from each chamber and, during the “On” period, triggered illumination when the signal exceeded the programmed threshold (FLIC signal = 20). Recorded values from all chambers in all DFMs were imported and analyzed using custom Python code adapted from R code (https://github.com/PletcherLab/FLIC_R_Code) provided with the FLIC apparatus. The Python code is available at https://github.com/joshuabhk/flic-analysis.

Experimental procedure:

To enable temporally precise manipulation of the feedforward GRN circuit during gustatory bouts in individual flies, we expressed CsChrimson—a light-activated cation channel that depolarizes target neurons in the presence of all-trans-retinal—in Gr93a-Gal4-expressing GRNs. The optolid-FLIC system triggered illumination (627 nm at ~0.7 mW/cm²) immediately after the first gustatory bout started, and, depending on bout duration, lasted 1–2.5 seconds (see Figure S9). Flies did not show consistent gustatory behaviors during the first 24 hours in the test chambers, so we began our analysis after this time, dividing the following 60 hrs into “Before,” “On,” and “After” periods (Figure 6A). During the Before and After periods, flies fed normally as their behaviors were recorded. During the On period, closed-loop illumination triggered by gustatory bouts was enabled in each test chamber. Following published procedures⁴⁰, gustatory bouts were defined as sequences of voltage signals that continuously remained above a lower threshold, with at least one value exceeding an upper threshold (thresholds were set to 10 and 20, respectively; see Figure S9). Python code was used to identify individual gustation bouts and confirm the accuracy of illumination times (examples are shown in Figure S9). Flies showing very low gustatory activity (fewer than 30 gustation bouts per 12 hrs, observed in approximately one in ten flies) were considered unhealthy and were excluded from further analysis.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistics were analyzed using Prism 9 or Python’s built-in libraries.

Sample size estimation:

Sample sizes for all electrophysiology experiments were determined by an a priori power analysis using Cohen's d estimated from pilot experiments. The sample size calculated for wild-type GRN pairs stimulated by lobeline was used for all other experimental groups. Statistical power for Levene’s test was estimated via Monte Carlo simulation (1000 iterations per condition) using scipy.stats in python, with power defined as the proportion of iterations rejecting the null hypothesis at α=0.05. These analyses showed our sampling was sufficient to support all our conclusions..

Normality test:

Each dataset was examined for normality using an appropriate test (D'Agostino-Pearson omnibus normality test, Anderson-Darling test, Shapiro-Wilk normality test and Kolmogorov-Smirnov normality test with Dallal-Wilkinson-Lilliefors correction); because none followed a normal distribution, subsequent comparisons were made using nonparametric tests.

Statistical test for electrophysiology:

We chose a statistical approach designed to evaluate the reliability of firing rate changes under different experimental conditions. To assess whether activating presynaptic GRNs consistently altered the firing rates of postsynaptic GRNs, we used paired nonparametric Wilcoxon signed-rank tests to compare postsynaptic firing rates before and after activating the presynaptic GRN (Figures 3–5). To quantify and compare this consistency while accounting for the paired, non-parametric nature of the data and differences in baseline firing rates, effect sizes and sample sizes across GRN pairs, we calculated rank-biserial correlations^49,50 for each GRN pair as an independent stratum⁵¹, and combined effects across strata using unweighted averaging. The resulting combined statistic was evaluated using a permutation test for equality of correlation-type statistics^52,53.

Specifically, for each animal and each GRN pair, we calculated the change in firing rate, defined as d_i = after_i − before_i, where i indexes the individual recordings within a given genotype and GRN pair, after activating the presynaptic GRN. We measured how consistently these changes pointed in one direction, within a genotype, using the matched-pairs rank-biserial correlation:

r = (W + - W -) / (W + + W -) = W / [n (n + 1) / 2]

Here, $W +$ and $W -$ are the sums of the ranks of $|d_{-} i|$ for the positive and negative differences, respectively (tied $|d_{-} i|$ values are given the average of their ranks, and pairs where $d_{-} i = 0$ are dropped; $n$ is the resulting number of non-zero paired differences). $W = W + - W -$ is the Wilcoxon signed-rank statistic. $r$ ranges from −1 to +1; biologically, r = +1 indicates that all animals showed an increase in postsynaptic firing, regardless of the magnitude of that increase; r = 0 indicates that increases and decreases were roughly equally mixed, even if some individual changes were large; r = −1 indicates that all animals showed a decrease in postsynaptic firing, regardless of magnitude.

To compare responses across two genotypes, we calculated $r$ separately for both, within each of the $k$ GRN pairs (treating each pair as a stratum). We then took the difference between the two genotypes' $r$ values within each stratum, $Δ r_j = r_j (1) - r_j (2)$ . To compare the effect of genotype at the circuit level, we averaged this difference across all strata. $j$ indexes the $k$ GRN pairs:

Δ \bar{r}_obs = (1 / k) Σ_j Δ r_j

To determine whether observed differences were larger than expected by chance, we built a null distribution using permutation⁵⁴ calculated separately within each stratum: within each GRN pair, we pooled the animals from both genotypes together, then randomly reassigned them back into two groups of the original sizes. We recalculated $r$ for each newly-shuffled group (re-ranking the data within each shuffled group), giving a permuted per-pair difference $Δ r_j^{*}$ . We then averaged these across the k pairs to obtain a single permuted, averaged statistic,

△ \bar{r} * = (1 / k) Σ_j △ r_j *

We repeated this process 200,000 times to build up a distribution of $△ \bar{r} *$ values that could occur under random reassignment. The one-sided p-value was then calculated as:

p = (# \{△ \bar{r} * \geq △ \bar{r}_obs\} + 1) / (N + 1)

⁵⁵

where $# {\cdot}$ counts how many of the permuted, averaged $△ \bar{r} *$ values were greater than or equal to the observed $△ \bar{r}_obs$ . We used one-tailed tests because we predicted in advance that functional genotypes would show more consistent effects. Because shuffling was done separately within each GRN pair, the pairing between GRN circuits and genotype was preserved during permutation. We applied this procedure to the following comparisons: increased postsynaptic GRN firing rate in wild-type flies following tastant vs control stimulation (Figures 3D vs. S4B), Gr66a GRN silenced vs control flies (Figure 4A vs. 4B), wild-type versus mutant flies (Figures 3D vs. 5B), and rescue versus non-rescued control flies (Figures 5E vs. 5C-D).

Statistical test for behavior test:

Median bout durations during the Before, On, and After behavior test periods were compared across genotypes using two-tailed, paired nonparametric Friedman tests (Figure 6B and Figure S10A; for raw data, see Data S1.xlsx). Differences in median bout duration, gustatory bout number per illumination, and median duration of first bouts across genotypes were assessed using two-tailed, nonparametric Kruskal-Wallis tests followed by Dunn’s post hoc test (Figure 6C-E and Figure S10B; see Data S1-S3.xlsx for raw data). We assessed equality of variance in median bout duration across genotypes using a one-tailed, nonparametric version of Levene’s test, testing the hypothesis that the receptor knockout mutant exhibits greater variance than either the wild-type or the receptor-restored mutant (in which receptor expression was specifically restored in Gr93a-Gal4-expressing GRNs). Detailed statistical results are provided in Table S1-S5, Data S4 and S5.xlsx.

Supplementary Material

Data S1. Number and log2-transformed median duration of detected gustatory bouts across genotypes. Related to Figure 6.

NIHMS2209771-supplement-1.xlsx^{(26.3KB, xlsx)}

Data S2. log2 Median Duration of 1st bout during illumination across different genotypes. Related to Figure 6E.

NIHMS2209771-supplement-2.xlsx^{(16.5KB, xlsx)}

Data S3. Fraction of animals showing different numbers of bouts per illumination and calculated average number of bouts per illumination. Related to Figure 6D.

NIHMS2209771-supplement-3.xlsx^{(32.6KB, xlsx)}

Data S4. Statistical analysis to compare mean of log2-transformed median duration of gustatory bouts across genotypes. Related to Figures 6 and S10.

NIHMS2209771-supplement-4.xlsx^{(11.3KB, xlsx)}

Data S5. Statistical analysis to compare mean and variance of log2-transformed median duration of gustatory bouts between antenna-ablated and control animals. Related to Figures 6 and S11.

NIHMS2209771-supplement-5.xlsx^{(12.4KB, xlsx)}

Document S1. Figures S1-S11 and Tables S1-S5.

NIHMS2209771-supplement-6.pdf^{(24.6MB, pdf)}

Highlights.

Functional synapses connect bitter-sensing gustatory receptor neurons (GRNs).
GRN connectivity is selective and directional, not all-to-all.
The nAChR α5 receptor mediates this inter-GRN excitation.
Receptor blockade impairs synaptic function and taste-driven behavior.

Acknowledgements

We thank B. Pritchard and G. Dold (Section of Instrumentation, NIMH) for assistance in configuring two-channel sensilla recordings. We also thank Z. Aldworth (NICHD), V. Schram (Microscopy and Imaging Center, NICHD), and S. Pletcher (University of Michigan) for assistance with spike sorting, image processing, and the FLIC assay, respectively. We thank the Bloomington Drosophila Stock Center for fly stocks and Addgene for plasmids. We thank members of the Stopfer laboratory, E. Ginger (NINDS), B. White (NIMH), Q. Yuan (NINDS) and the NIH Drosophila Neurobiology Consortium for helpful discussions and feedback on the manuscript. This research was supported by the Intramural Research Program of the National Institutes of Health (NIH). The contributions of the NIH authors are considered Works of the United States Government. The findings and conclusions presented in this paper are those of the author and do not necessarily reflect the views of the NIH or the U.S. Department of Health and Human Services. An intramural grant from NIH-NICHD to M. Stopfer (1ZIAHD008760-20) A career development award from NIH-NICHD to B. Song

Footnotes

Declaration of interests: The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process During the preparation of this work, the authors used ChatGPT and Claude Sonnet 5 to help revise some sentences. The authors reviewed and edited the output as needed and take full responsibility for the content of the published article.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

1.Marder E, and Goaillard J-M (2006). Variability, compensation and homeostasis in neuron and network function. Nat. Rev. Neurosci. 7, 563–574. [DOI] [PubMed] [Google Scholar]
2.Faisal AA, Selen LPJ, and Wolpert DM (2008). Noise in the nervous system. Nat. Rev. Neurosci. 9, 292–303. [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Carandini M, and Heeger DJ (2011). Normalization as a canonical neural computation. Nat. Rev. Neurosci. 13, 51–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Panzeri S, Moroni M, Safaai H, and Harvey CD (2022). The structures and functions of correlations in neural population codes. Nat. Rev. Neurosci. 23, 551–567. [DOI] [PubMed] [Google Scholar]
5.Turrigiano G (2011). Too many cooks? Intrinsic and synaptic homeostatic mechanisms in cortical circuit refinement. Annu. Rev. Neurosci. 34, 89–103. [DOI] [PubMed] [Google Scholar]
6.Hoon MA, Adler E, Lindemeier J, Battey JF, Ryba NJP, and Zuker CS (1999). Putative mammalian taste receptors: a class of taste-specific GPCRs with distinct topographic selectivity. Cell 96, 541–551. [DOI] [PubMed] [Google Scholar]
7.Adler E, Hoon MA, Mueller KL, Chandrashekar J, Ryba NJ, and Zuker CS (2000). A novel family of mammalian taste receptors. Cell 100, 693–702. [DOI] [PubMed] [Google Scholar]
8.Nelson G, Hoon MA, Chandrashekar J, Zhang Y, Ryba NJ, and Zuker CS (2001). Mammalian sweet taste receptors. Cell 106, 381–390. [DOI] [PubMed] [Google Scholar]
9.Roper SD (2022). Encoding taste: from receptors to perception. Handb. Exp. Pharmacol. 275, 53–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Devineni AV, Deere JU, Sun B, and Axel R (2021). Individual bitter-sensing neurons in Drosophila exhibit both ON and OFF responses that influence synaptic plasticity. Curr. Biol. 31, 5533–5546. [DOI] [PubMed] [Google Scholar]
11.Dweck HKM, and Carlson JR (2023). Diverse mechanisms of taste coding in Drosophila. Sci. Adv. 9, eadj7032. [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Jeong YT, Shim J, Oh SR, Yoon HI, Kim CH, Moon SJ, and Montell C (2013). An odorant-binding protein required for suppression of sweet taste by bitter chemicals. Neuron 79, 725–737. [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Youn H, Kirkhart C, Chia J, and Scott K (2018). A subset of octopaminergic neurons that promotes feeding initiation in Drosophila melanogaster. PLoS One 13, e0198362. [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Li Q, DeBeaubien NA, Sokabe T, and Montell C (2020). Temperature and sweet taste integration in Drosophila. Curr. Biol. 30, 2051–2067. [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Roper SD (2021). Chemical and electrical synaptic interactions among taste bud cells. Curr. Opin. Physiol. 20, 118–125. [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Chu B, Chui V, Mann K, and Gordon MD (2014). Presynaptic gain control drives sweet and bitter taste integration in Drosophila. Curr. Biol. 24, 1978–1984. [DOI] [PubMed] [Google Scholar]
17.Bohra AA, Kallman BR, Reichert H, and VijayRaghavan K (2018). Identification of a single pair of interneurons for bitter taste processing in the Drosophila brain. Curr. Biol. 28, 847–858. [DOI] [PubMed] [Google Scholar]
18.Lee M, Park SH, Joo KM, Kwon JY, and Lee K-H, and Kang K (2024). Drosophila HCN mediates gustatory homeostasis by preserving sensillar transepithelial potential in sweet environments. eLife 13, RP96602. [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Lee M, Kim SY, Park T, Yoon S-E, Kim Y-J, Joo KM, Kwon JY, Kim K, and Kang K (2025). An evolutionarily conserved cation channel tunes the sensitivity of gustatory neurons to ephaptic inhibition in Drosophila. Proc. Natl. Acad. Sci. USA 122, e2413134122. [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Weiss LA, Dahanukar A, Kwon JY, Banerjee D, and Carlson JR (2011). The molecular and cellular basis of bitter taste in Drosophila. Neuron 69, 258–272. [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Dweck HKM, and Carlson JR (2020). Molecular logic and evolution of bitter taste in Drosophila. Curr. Biol. 30, 17–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Delventhal R, Menuz K, Joseph R, Park J, Sun JS, and Carlson JR (2017). The taste response to ammonia in Drosophila. Sci. Rep. 7, 43754. [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Tomchik SM, Berg S, Kim JW, Chaudhari N, and Roper SD (2007). Breadth of tuning and taste coding in mammalian taste buds. J. Neurosci. 27, 10840–10848. [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Jaeger AH, Stanley M, Weiss ZF, Musso P-Y, Chan RCW, Zhang H, Feldman-Kiss D, and Gordon MD (2018). A complex peripheral code for salt taste in Drosophila. eLife 7, e37167. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Engert S, Sterne GR, Bock DD, and Scott K (2022). Drosophila gustatory projections are segregated by taste modality and connectivity. eLife 11, e78110. [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Delventhal R, Kiely A, and Carlson JR (2014). Electrophysiological recording from Drosophila labellar taste sensilla. J. Vis. Exp. 84, e51355. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Talay M, Richman EB, Snell NJ, Hartmann GG, Fisher JD, Sorkaç A, Santoyo JF, Chou-Freed C, Nair N, Johnson M, et al. (2017). Transsynaptic mapping of second-order taste neurons in flies by trans-Tango. Neuron 96, 783–795. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Hiroi M, Marion-Poll F, and Tanimura T (2002). Differentiated response to sugars among labellar chemosensilla in Drosophila. Zoolog. Sci. 19, 1009–1018. [DOI] [PubMed] [Google Scholar]
29.Wieczorek H, and Wolff G (1989). The labellar sugar receptor of Drosophila. J. Comp. Physiol. A 164, 825–834. [Google Scholar]
30.Reiter S, Campillo Rodriguez C, Sun K, and Stopfer M (2015). Spatiotemporal coding of individual chemicals by the gustatory system. J. Neurosci. 35, 12309–12321. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Lee Y, Moon SJ, and Montell C (2009). Multiple gustatory receptors required for the caffeine response in Drosophila. Proc. Natl. Acad. Sci. USA 106, 4495–4500. [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Jeong YT, Oh SM, Shim J, Seo JT, Kwon JY, and Moon SJ (2016). Mechanosensory neurons control sweet sensing in Drosophila. Nat. Commun. 7, 12872. [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Niemann H, Blasi J, and Jahn R (1994). Clostridial neurotoxins: new tools for dissecting exocytosis. Trends Cell Biol. 4, 179–185. [DOI] [PubMed] [Google Scholar]
34.Klapoetke NC, Murata Y, Kim SS, Pulver SR, Birdsey-Benson A, Cho YK, Morimoto TK, Chuong AS, Carpenter EJ, Tian Z, et al. (2014). Independent optical excitation of distinct neural populations. Nat. Methods 11, 338–346. [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Deng B, Li Q, Liu X, Cao Y, Li B, Qian Y, Xu R, Mao R, Zhou E, Zhang W, et al. (2019). Chemoconnectomics: mapping chemical transmission in Drosophila. Neuron 101, 876–893. [DOI] [PubMed] [Google Scholar]
36.Lee P-T, Zirin J, Kanca O, Lin W-W, Schulze KL, Li-Kroeger D, Tao R, Devereaux C, Hu Y, Chung V, et al. (2018). A gene-specific T2A-GAL4 library for Drosophila. eLife 7, e35574. [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Lansdell SJ, Collins T, Goodchild J, and Millar NS (2012). The Drosophila nicotinic acetylcholine receptor subunits Dα5 and Dα7 form functional homomeric and heteromeric ion channels. BMC Neurosci. 13, 73. [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Rimal S, Sang J, Poudel S, Thakur D, Montell C, and Lee Y (2019). Mechanism of acetic acid gustatory repulsion in Drosophila. Cell Rep. 26, 1432–1442. [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Deere JU, Sarkissian AA, Yang M, Uttley HA, Martinez Santana N, Nguyen L, Ravi K, and Devineni AV (2023). Selective integration of diverse taste inputs within a single taste modality. eLife 12, e84856. [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Ro J, Harvanek ZM, and Pletcher SD (2014). FLIC: high-throughput, continuous analysis of feeding behaviors in Drosophila. PLoS One 9, e101107. [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Weaver KJ, Raju S, Rucker RA, Chakraborty T, Holt RA, and Pletcher SD (2023). Behavioral dissection of hunger states in Drosophila. eLife 12, RP84537. [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Guillemin J, Li J, Li V, McDowell SAT, Audette K, Davis G, Jelen M, Slamani S, Kelliher L, Gordon MD, et al. (2024). Taste cells expressing Ionotropic Receptor 94e reciprocally impact feeding and egg laying in Drosophila. Cell Rep. 43, 114625. [DOI] [PubMed] [Google Scholar]
43.Kay LM, and Stopfer M (2006). Information processing in the olfactory systems of insects and vertebrates. Semin. Cell Dev. Biol. 17, 433–442. [DOI] [PubMed] [Google Scholar]
44.Evans DA, Stempel AV, Vale R, Ruehle S, Lefler Y, and Branco T (2018). A synaptic threshold mechanism for computing escape decisions. Nature 558, 590–594. [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Thoma V, Kobayashi K, and Tanimoto H (2017). The role of the gustatory system in the coordination of feeding. eNeuro 4, ENEURO.0324–17.2017. [DOI] [PMC free article] [PubMed] [Google Scholar]
46.McKellar CE, Siwanowicz I, Dickson BJ, and Simpson JH (2020). Controlling motor neurons of every muscle for fly proboscis reaching. eLife 9, e54978. [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Zhou Y, Cao L-H, Sui X-W, Guo X-Q, and Luo D-G (2019). Mechanosensory circuits coordinate two opposing motor actions in Drosophila feeding. Sci. Adv. 5, eaaw5141. [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Walker SR, Peña-Garcia M, and Devineni AV (2025). Connectomic analysis of taste circuits in Drosophila. Sci. Rep. 15, 5278. [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Cureton EE (1956). Rank-biserial correlation. Psychometrika 3, 287–290. [Google Scholar]
50.Glass GV (1966). Note on rank biserial correlation. Educ. Psychol. Meas. 26, 623–631. [Google Scholar]
51.Hodges JL Jr., and Lehmann EL (1962). Rank methods for combination of independent experiments in analysis of variance. Ann. Math. Stat. 33, 482–497. [Google Scholar]
52.Sakaori F (2002). A nonparametric test for the equality of dependent correlation coefficients under normality. Commun. Stat. Theory Methods 31, 2379–2389. [Google Scholar]
53.Omelka M, and Pauly M (2012). Testing equality of correlation coefficients in two populations via permutation methods. J. Stat. Plan. Inference 142, 1396–1406. [Google Scholar]
54.Ernst MD (2024). Permutation methods: a basis for exact inference. Stat. Sci. 19, 676–685. [Google Scholar]
55.Phipson B, and Smyth GK (2010). Permutation p-values should never be zero: calculating exact p-values when permutations are randomly drawn. Stat. Appl. Genet. Mol. Biol. 9, Article39. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1. Number and log2-transformed median duration of detected gustatory bouts across genotypes. Related to Figure 6.

NIHMS2209771-supplement-1.xlsx^{(26.3KB, xlsx)}

Data S2. log2 Median Duration of 1st bout during illumination across different genotypes. Related to Figure 6E.

NIHMS2209771-supplement-2.xlsx^{(16.5KB, xlsx)}

Data S3. Fraction of animals showing different numbers of bouts per illumination and calculated average number of bouts per illumination. Related to Figure 6D.

NIHMS2209771-supplement-3.xlsx^{(32.6KB, xlsx)}

Data S4. Statistical analysis to compare mean of log2-transformed median duration of gustatory bouts across genotypes. Related to Figures 6 and S10.

NIHMS2209771-supplement-4.xlsx^{(11.3KB, xlsx)}

Data S5. Statistical analysis to compare mean and variance of log2-transformed median duration of gustatory bouts between antenna-ablated and control animals. Related to Figures 6 and S11.

NIHMS2209771-supplement-5.xlsx^{(12.4KB, xlsx)}

Document S1. Figures S1-S11 and Tables S1-S5.

NIHMS2209771-supplement-6.pdf^{(24.6MB, pdf)}

Data Availability Statement

Data: raw data for figures have been deposited on Zenodo (DOI: https://doi.org/10.5281/zenodo.18887927) and are publicly available as of the date of publication.
Code: Python code for FLIC analysis has been deposited on Github (https://github.com/joshuabhk/flic-analysis). Matlab code for spike sorting has been deposited on bitbucket (https://bitbucket.org/zaldworth/stopferlabspikesorter/src/main/). Both codes are publicly available.
Additional information: Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

PERMALINK

A peripheral feedforward circuit amplifies taste signals, ensuring behavioral reliability

Bo-Mi Song

Bong-Hyun Kim

Mark Stopfer

Roles

Summary

eTOC Blurb

Graphical Abstract

Introduction

RESULTS

Labellar GRNs form two distinct patterns of direct synaptic connectivity.

Figure 1. Systematic anterograde transsynaptic mapping reveals two distinct patterns of synaptic connectivity among gustatory receptor neurons on the labellum, the taste apparatus around the fly’s mouth.

Similarly-tuned Gr66a-expressing GRNs form synaptic connections.

Figure 2. Synapses connect genetically-defined subsets of Gr66a-expressing GRNs in the labellum.

Similarly-tuned Gr66a-expressing GRNs show functional feedforward excitatory interactions.

Figure 3. Tastants trigger functional, excitatory synaptic interactions between GRNs in S sensilla.

Figure 4. Interacting GRNs all express Gr66a.

nAChR α5 mediates inter-GRN excitatory synaptic interactions.

Figure 5. The nAChR α5 subunit is required for reliable communication between interacting, Gr66a-expressing GRNs.

The feedforward amplification circuit enhances behavioral reliability.

Figure 6. Optogenetic activation of Gr93a-Gal4-expressing GRNs suppresses gustatory behavior by inhibiting its initiation and inducing its termination. nAChR α5-mediated inter-GRN interactions among these neurons stabilize suppression of gustatory behavior.

Discussion

A peripheral feedforward amplification circuit for behavioral reliability

Combining multiple approaches to assess function

Resource Availability

Lead Contact:

Materials Availability

Key Resources Table.

Data and Code Availability

STAR Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Drosophila stocks

Flies were grown on Jazz mix culture medium (Fisher Scientific).

Age and cultivation:

Sex reporting:

METHOD DETAILS

Electrophysiology

Recording setup:

Fly preparation:

Experimental procedure:

Optogenetic Validation:

Spike sorting and Data analysis

General:

Tastant stimuli

Molecular biology

PCR screen for AChR expression:

Generation of nAChR α5 null mutant:

Cloning of nAChR α5 cDNA and generation of a rescue strain:

Immunohistochemistry, Microscopy and Image Processing

Behavior

FLIC assay setup:

Experimental procedure:

QUANTIFICATION AND STATISTICAL ANALYSIS

Sample size estimation:

Normality test:

Statistical test for electrophysiology:

Statistical test for behavior test:

Supplementary Material

Highlights.

Acknowledgements

Footnotes

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases