The peripheral chemosensory system of the larvae is subdivided into bilaterally symmetrical, ‘external’ and ‘internal’ taste organs (Stocker, 2008). The three external organs reside near the tip of the head and are organized as paired ganglia, the dorsal, terminal, and ventral organ ganglia (DOG, TOG, and VOG) that house the GRN cell bodies with dendritic extensions in respective sensory organs (DO, TO, and VO) at the head surface, while carrying information via their axons to the subesophageal zone (SEZ) in the brain (Figure 1A). The DOG harbors 21 olfactory neurons and 18 presumptive GRNs. For clarity, GRN numbers refer to neurons in one of the two, bilaterally symmetrical taste organs. The GRNs located in the DOG fall into two distinct groups, based on their dendritic extensions: 11 presumptive GRNs, 4 of which were shown to express Gr genes (Apostolopoulou et al., 2015), send dendrites to the base of the DO, while 7 neurons extend dendrites to the TO (the dorsolateral group, Figure 1A; Kwon et al., 2011). The TOG contains 30 neurons, with dendrites located in the TO (the distal group). The internal taste structures, referred to as the dorsal/ventral pharyngeal sense organ (DPS/VPS, 33 neurons) and the posterior pharyngeal sense organs (PPS, 6 neurons) are located along the pharynx and sense chemicals as they are moved toward the digestive system. We note that not all these neurons are confirmed GRNs, either by function or expression of Gr or IR genes, albeit based on their location and anatomy, most are thought to be GRNs (Rist and Thum, 2017; Sánchez-Alcañiz et al., 2018; Stewart et al., 2015).

Our initial expression analysis of the Gr28 genes revealed that four of the six Gr28 genes (Gr28a, Gr28b.a, Gr28b.c, and Gr28b.e) were expressed in larval taste organs, in addition to cells in the gut, the brain, and non-chemosensory cells of the larvae (Mishra et al., 2018; Thorne and Amrein, 2008). This and all previous Gr expression studies were performed using bimodal expression systems, mostly GAL4/UAS, whereby Gr promotors driving GAL4 are assumed to faithfully reproduce expression of the respective Gr genes. Importantly, we analyzed two to four Gr28-GAL4 insertion lines for each transgene, and at least two generated the same expression pattern (Mishra et al., 2018; Thorne and Amrein, 2008), providing evidence that the drivers reflect a fairly accurate expression profile of respective endogenous genes. To further delineate the putative chemosensory roles of these genes, we performed a more detailed co-expression analysis between the Gr28 genes and the bitter taste receptor gene Gr66a by combining the GAL4/UAS (labeling Gr28 expressing neurons) with the LexA/lexAop (marking Gr66a neurons) system (Figure 1C and D). In the TOG, we found expression of all four Gr28 genes, along with that of Gr66a, which was expressed in three or four neurons (this number is slightly smaller than the six neurons previously reported by Kwon et al., 2011). GAL4 drivers for Gr28b.a, Gr28b.c, and Gr28b.e are co-expressed in a single neuron with Gr66a-LexA (Figure 1C). In contrast, Gr28a-GAL4 is expressed in a distinct TOG neuron than Gr66a-LexA (Figure 1C) and the Gr28b genes, an observation we independently confirmed using a Gr28b.c-LexA driver (Figure 1—figure supplement 1). In the DOG, we find only a single Gr28a-GAL4 neuron, while none of the Gr28b genes, or Gr66a, is expressed there (Figure 1C). None of the Gr28-GAL4 drivers was expressed in the VOG. In the internal sensory organs, Gr66a-LexA is found in one GRN in the DPS/VPS, where it is co-expressed with Gr28b.c-GAL4 as well as Gr28a-GAL4 (Figure 1D). In the PPS, none of the Gr28b genes is expressed, but Gr28a-GAL4 is found in two neurons, one of which also expresses Gr66a-LexA (Figure 1D). In summary, the internal sensory organs can be subdivided into three distinct groups (Figure 1E): Gr66a/Gr28b.c/Gr28a-positive neurons, Gr66a/Gr28a-positive neurons, and Gr28a^only neurons, which contrasts the external sensory organs where Gr28a and the Gr28b genes are expressed in a mutually exclusive fashion.

We previously showed that at least one of the six Gr28 genes is necessary for feeding attraction to RNA, ribonucleosides, and ribose using a well-established two-choice feeding preference assay (Mishra et al., 2018; Figure 2A). Specifically, larvae homozygous mutant for Gr28 (ΔGr28, a deletion of the entire Gr28a gene and more than half of the coding region of all Gr28b genes, Figure 1B) lose their ability to sense these compounds, a phenotype that is restored when single UAS-Gr28 reporter transgenes are expressed in Gr28a-GAL4 neurons. The largely non-overlapping expression of Gr28a and the Gr28b genes suggests that respective neurons represent functionally distinct entities. To investigate this possibility, we took advantage of the mammalian VR1 protein, a TRP channel that is activated by capsaicin (Caterina et al., 1997). Drosophila have no VR1 like-gene in their genome and do not respond to capsaicin behaviorally, but flies are attracted to this chemical when a modified VR1 gene (VR1E600K) is expressed in sweet taste neurons (Marella et al., 2006). Thus, we expressed VR1E600K (henceforth referred to UAS-VR1) under the control of the four Gr28-GAL4 drivers (Gr28a-GAL4, Gr28b.a-GAL4, Gr28b.c-GAL4, and Gr28b.e-GAL4) in larvae and tested their response to capsaicin using the two-choice preference assay (Figure 2A). Additionally, we expressed UAS-VR1 in bitter neurons as well as appetitive fructose sensing neurons using respective GAL4 drivers (Gr66a-GAL4 and Gr43a^GAL4) (Mishra et al., 2013; Scott et al., 2001), which served as control larvae (note that fructose sensing Gr43a^GAL4 neurons are distinct from Gr28a-GAL4 neurons; Mishra et al., 2018). Just like adult flies, w¹¹¹⁸ control larvae lacking either a UAS-VR1 reporter, a Gr-GAL4 driver, or both, were unresponsive to 0.1 mM capsaicin, while the positive control larvae expressing UAS-VR1 in either fructose sensing GRNs or bitter taste GRNs showed robust attraction to or avoidance of capsaicin (Figure 2B). Consistent with previous findings, which identified Gr28a GRNs as appetitive neurons (Mishra et al., 2018), larvae expressing UAS-VR1 under the control of Gr28a-GAL4 showed strong appetitive responses to capsaicin (Figure 2B). In contrast, larvae expressing UAS-VR1 under the control of Gr28b.c-GAL4 showed robust avoidance behavior to capsaicin, while expression in the single TOG GRN using either Gr28b.a-GAL4 or Gr28b.e-GAL4 (also expressing Gr28b.c; Figure 1C) caused neither attraction to nor avoidance of capsaicin.

Overlap between Gr28a and Gr66a in internal GRNs raises the question about their contribution to appetitive and/or avoidance behavior. We deemed it unlikely that these neurons were critical for appetitive behavior since the one pair located in the DPS/VPS (expressing also Gr28b.c) is necessary for capsaicin avoidance (see above). Indeed, when VR1 was suppressed in all Gr66a/Gr28b.c neurons by means of the GAL4 suppressor GAL80, these larvae, expressing VR1 in Gr28a^only GRNs, remained strongly attracted to capsaicin (Figure 2B, last panel; for effective suppression, see Figure 2—figure supplement 1). Taken together, this analysis indicates that Gr28a-GAL4 is expressed in appetitive-inducing neurons, while Gr28b.c-GAL4 neurons mediate avoidance behavior.

The capsaicin experiments above suggest that the two larval Gr28b.c-GAL4 GRNs mediate negative valence. Previous reports have shown that larvae avoid many bitter compounds (Apostolopoulou et al., 2016; Apostolopoulou et al., 2015; Choi et al., 2020; Choi et al., 2016; van Giesen et al., 2016), and thus, we expected that eliminating activity of Gr28b.c-GAL4 GRNs would result in loss of avoidance behavior to at least some bitter chemicals. Indeed, when Gr28b.c-GAL4 GRN activity was blocked using the inward-rectifying potassium channel Kir2.1 (Baines et al., 2001; Paradis et al., 2001), larvae no longer avoided the four tested bitter compounds denatonium, quinine, lobeline, and caffeine (Figure 3). A compound-specific avoidance phenotype to lobeline and caffeine was observed when the Gr28b.e-GAL4 GRNs in the TOG were inactivated. These data, together with the capsaicin experiments, suggest that two pairs of GRNs, one in the TOG and one in the DPS/VPS, are necessary and sufficient for avoidance of these four bitter tasting chemicals.

We next examined whether any of the Gr28b proteins is part of a taste receptor complex detecting any of these bitter chemicals (Figure 4). Surprisingly, only avoidance of denatonium was affected in larvae lacking the Gr28 gene cluster (ΔGr28; Figure 1B), while avoidance to quinine was somewhat reduced, albeit not significantly, and avoidance to lobeline or caffeine remained robust. In fact, avoidance to caffeine increased modestly, but significantly (Figure 4A). We then examined whether any of the Gr28 proteins was sufficient to restore denatonium avoidance by expressing individual Gr28 genes under the control of the Gr28b.c-GAL4 driver. Indeed, either Gr28b.a or Gr28b.c expression led to a full recovery of denatonium avoidance, while expression of any other Gr28b gene, or Gr28a, failed to do so (Figure 4B). This observation suggests that despite the high level of similarity between these receptors, recognition of denatonium is dependent on specific structural features present in Gr28b.a and Gr28b.c, but not in any of the other Gr28 proteins.

Since Gr66a-LexA is co-expressed in all Gr28b-GAL4-expressing GRNs, we wondered whether Gr66a is a component of the denatonium receptor. Previous work had established that Gr66a is required for caffeine avoidance in both larvae and adult flies (Apostolopoulou et al., 2016; Lee et al., 2009; Moon et al., 2006), which we confirmed (Figure 4—figure supplement 1). Surprisingly, avoidance of denatonium and quinine was not diminished, but increased significantly (Figure 4—figure supplement 1). Given the multimeric nature of bitter taste receptors (Sung et al., 2017), one possibility is that the absence of a Gr subunit not required for the detection of denatonium (Gr66a) could favor formation of multimeric complexes containing Gr subunits that recognize this compound (such as Gr28b.a and/or Gr28b.c).

Finally, we investigated neuronal responses in larvae expressing the Ca²⁺ indicator GCaMP6m in Gr28b.c-GAL4 GRNs (Figure 5). We developed a whole animal imaging preparation, whereby larvae were placed in an ‘imaging chamber’ to minimize head movements (Figure 5A), and visualized neural activity of the Gr28b.c-GAL4 GRN in the TOG in real time (Chen et al., 2013) upon exposure to the four bitter compounds, as well as sucrose, ribose, and fructose (Figure 5B–D). All bitter compounds elicited rapid Ca²⁺ increases in Gr28b.c-GAL4 GRNs, while none of the sugars did (Figure 5C and D). When neural activity was recorded in Gr28b.c-GAL4 GRNs of ΔGr28 homozygous mutant larvae (Figure 5E), Ca²⁺ responses to denatonium and quinine were severely reduced, while responses to both caffeine and lobeline were not affected. Re-expression of either Gr28b.a or Gr28b.c, but not Gr28b.b, Gr28b.d, Gr28b.e, or Gr28a rescued Ca²⁺ response to denatonium, but not to quinine (Figure 5F and G). Together, these experiments identified Gr28b.c and Gr28b.a as redundant subunits of a denatonium receptor complex, a complex that does not require Gr66a or any of the other Gr28b subunits.

Our expression analysis of the bitter taste receptor gene Gr66a and the Gr28 genes in larvae is consistent with earlier studies, despite some small variation in neuron number (Choi et al., 2016; Kwon et al., 2011), which is likely due to the use of different GAL4 driver lines and/or variability in expression levels. Importantly, all Gr28b genes are co-expressed with the bitter taste receptor gene Gr66a (Figure 1) and probably several other putative bitter Gr genes (Kwon et al., 2011; Rist and Thum, 2017), while Gr28a is found in a largely, but not entirely, distinct set of GRNs. Whether and what kind of Gr genes might be co-expressed with Gr28a in Gr28a^only GRNs will require more in-depth expression studies and might shed light on other receptors involved in appetitive behaviors of larvae.

A key finding of the work presented here is the observation that Gr28a^only and Gr28b.c neurons dictate distinct behavioral programs, the former representing an ensemble of neurons that instruct larvae to ‘go toward’ a chemical source and consume it, while the latter do the opposite (Figure 2). This observation is reminiscent of a seminal study by Troemel and colleagues in the Caenorhabditis elegans chemosensory system, who reported that the valence of a chemical compound is dependent on the identity of a neuron, and not the identity of the molecular receptor the neuron expresses (Troemel et al., 1997). The number of ‘go-away’ GRNs in Drosophila larvae co-expressing Gr28b.c and Gr66a is remarkably small, consisting of only two pairs, one in the TO and the other in the DPS/VPS. It seems likely that this is the smallest, minimal subset of neurons sufficient to induce avoidance behavior as expression of VR1 in only the TO pair (under the control of either Gr28b.a-GAL4 or Gr28b.e-GAL4) has no behavioral effect when challenged with capsaicin. The ‘go-to’ neurons are characterized by expression of Gr28a and represent a slightly larger set of four GRN pairs (Gr28a^only GRNs) (Figure 7) . Thus, the minimal requirement to induce ‘go-to’ and ‘go-away’ behavior is defined by distinct sets of GRNs, and each appears to be composed of neurons located in both external and the internal taste organs. Co-expression of the RNA taste receptor Gr28a in the DPS/VPS GRN essential for bitter taste (Figures 2B and 3) raises interesting questions about additional functions for Gr28a in bitter taste. We note that the sweet taste receptor Gr5a, a subunit of a multimeric trehalose receptor, is also expressed in non-sweet neurons of unknown function (Fujii et al., 2015).

Both behavioral analyses and Ca²⁺ imaging experiments implicate at least one Gr28b protein as an essential component of a denatonium receptor complex as ΔGr28 larvae exhibit total loss of avoidance (Figure 4) and respective GRNs fail to elicit a response upon exposure to this chemical (Figure 3). What the precise composition of that complex is remains to be determined, but recovery of denatonium responses by expressing either Gr28b.c and/or Gr28b.a indicates that either one of these (or possibly both) is an essential subunit, in addition to other Grs expressed in this GRNs, such as Gr22a and Gr59c (Choi et al., 2020; Rist and Thum, 2017), while Gr66a is unlikely to be part of such a complex based on our behavioral analysis (Figure 4—figure supplement 1).

Since only Gr28b.a and Gr28b.c can rescue denatonium responses in Gr28b.c GRNs of ΔGr28 mutant larvae, sequence comparison between the unique N-terminal halves of the Gr28 proteins comprising the first four transmembrane domains and the extracellular loops 1 and 2 might provide insights as to possible residues important for ligand recognition. When interrogating these regions, only seven residues are identical between Gr28b.a and Gr28b.c (Table 1 and Figure 6). Reducing the stringency requirement by allowing one of the remaining receptors to share the same residue, nine additional sites are identified. One experimental avenue to validate these residues as important sites contributing to denatonium binding might involve introduction of point mutations that converts respective amino acids of other Gr28b proteins into those found in Gr28b.a/Gr28b.c.

The role of Gr28b proteins in quinine detection is less clear, and the different phenotypes observed in behavioral experiments and Ca²⁺ imaging suggest that at least two molecular types of quinine receptors exist in larvae. Ca²⁺ imaging experiments implicate a role for multiple Gr28b subunits in a quinine receptor complex in the TOG neuron since single Gr28b genes cannot restore the loss of quinine response in ΔGr28 larvae (Figure 5E and G). However, because ΔGr28 mutant larvae still avoid quinine (Figure 4), at least one Gr28b-independent receptor must exist in one or several other GRNs, one of which is likely the 28b.c GRN in the DPS/VPS, since expression of Kir2.1 in that neuron, along with the one in the TOG, totally abolishes quinine avoidance (Figure 3). We note that functional redundancy is also observed in sweet taste receptors, where we found that different combinations of sugar Gr genes can restore responses to the same sugar when expressed in tarsal sweet GRNs of sugar blind flies (Fujii et al. unpublished).

Flies were maintained on standard corn meal food in plastic vials under a 12 hr light/dark cycle at 25°C. The w¹¹¹⁸ strain (Bloomington Drosophila Stock Center, number 3605) was used as a wild-type control. Fly strains used: Gr28a-GAL4(SF36S) for Figures 1 and 2B, and SF36E1 for Figure 2B and Figure 2—figure supplement 1, Gr28b.a-GAL4(NT42aC51a), Gr28b.c-GAL4(NT21B1) (Thorne and Amrein, 2008); Gr28b.e-GAL4(Gr28a3AII) and Gr66a-GAL4 (Scott et al., 2001); ΔGr28(54B3) (Mishra et al., 2018); UAS-Gr28a, UAS-Gr28b.a, UAS-Gr28b.b, UAS-Gr28b.c, UAS-Gr28b.d, and UAS-Gr28b.e (Ni et al., 2013); Gr43a^GAL4 (Miyamoto et al., 2012); Gr66a-LexA and lexAop-GAL80 (Thistle et al., 2012); UAS-VR1E600K (Marella et al., 2006); UAS-Kir2.1-GFP (Baines et al., 2001; Paradis et al., 2001); lexAop-rCD2:GFP (Lai and Lee, 2006), UAS-GCaMP6m, UAS-mCD8:RFP and Gr66a^ex83 (Bloomington Drosophila Stock Center, numbers 42748, 32220, and 35528); Gr28b.c-LexA(#8).

Caffeine (Cat# C0750), capsaicin (Cat# M2028), denatonium benzoate (Cat# D5765), lobeline hydrochloride (Cat# 141879), D-(-)-ribose (Cat# R7500), and quinine hydrochloride dihydrate (Cat# Q1125) were purchased from MilliporeSigma, with a purity of >95%. Fructose (Cat# F1092) and agarose (Cat# 20-102) were purchased from Spectrum chemical and Apexbio, respectively. Sucrose (mfr. no. 8360-06) and charcoal (Cat# 1560-01) were purchased from Macron Fine Chemicals and J.T. Baker, respectively. A stock solution for capsaicin (20 mM) was prepared in 70% ethanol and stored at 4°C protected from light for up to 1 y. Stock solutions for bitter chemicals were prepared in Millipore Q water and stored at –20°C. Stock solutions for sugars were prepared in Millipore Q water and stored at 4°C for up to 1 mo. A stock solution for ribose was treated with charcoal (10% of the weight of ribose used for stock solution) overnight at 4°C and sterile-filtrated (0.45 μm) to remove unrelated odor. Stock solutions were diluted to the final concentration using Millipore Q water prior to each experiment.

Immunofluorescence of larval heads was performed based on the protocol described in Croset and colleagues (Croset et al., 2016) with minor modification. Heads of third-instar larvae were dissected using microscissors in phosphate-buffered saline (PBS) and immediately fixed in PBS with 4% paraformaldehyde for 1 hr at 4°C. They were washed six times in washing buffer (PBS with 0.1% Triton X-100) for 20 min and blocked for 1 hr in washing buffer containing 5% heat-inactivated goat serum (SouthernBiotech, Cat# 0060-01), followed by incubation with the primary antibodies (rabbit anti-GFP, 1:1000 dilution; rat anti-mCD8, 1:200 dilution, Thermo Fisher Scientific) at 4°C overnight. The next day, heads were washed six times for 20 min in washing buffer and blocked in washing buffer containing 5% heat-inactivated goat serum for 1 hr, followed by incubation with the secondary antibodies (goat anti-rabbit Alexa 488, 1:500 dilution, Thermo Fisher Scientific; goat anti-rat Cy3, 1:300 dilution, Jackson ImmunoResearch Laboratories Inc) at 4°C overnight. Finally, heads were washed six times in washing buffer for 20 min each at room temperature under gentle agitation. Heads were then mounted with VectaShield (Vector Lab, Cat# H-1200) on a microscope slide and images were obtained using a Nikon A1R confocal microscope system. Adobe Photoshop 2022 was used further to process images.

Two-choice preference assay of larvae was conducted as described in Mishra et al., 2013 with minor modifications. Flies were placed on standard corn meal food in plastic vials and allowed to lay eggs for 24 hr under a 12 hr light/dark cycle at 25°C. Flies were removed from food vials and feeding-stage third-instar larvae were collected. Agarose food dishes for two-choice preference assay were prepared just prior each experiment as follows: Petri dishes (60 × 15 mm, Falcon, Cat# REF353004) with two halves marked on the bottom were filled with melted plain 1% agarose or 1% agarose containing 1.75% ethanol (for capsaicin preference). After the agarose solidified, one half was removed and replaced with 1% agarose solution containing taste ligands (capsaicin or bitter compound). For each experiment, 15 larvae from food vials were briefly rinsed twice with Millipore Q water and placed along the middle separating pure and ligand containing agarose. After 16 min, images were taken for record keeping and used to calculate larval preference indices. Larvae that crawled onto the wall of a dish or dug in the agarose were excluded. The preference index (P.I.) was calculated as follow: PI = (N_tastant – N_plain)/N_Total, whereby N is the number of larvae in the tastant sector, the plain agarose sector, and the total number, respectively. Positive values indicate a preference for capsaicin or bitter compound while negative values indicate repulsion (avoidance).

Calcium imaging was performed in Gr28b.c GRNs expressed in the terminal organ of feeding-stage, third-instar larvae, reared as described for the larval two-choice preference assay. For each experiment, larvae from food vials were briefly rinsed twice with Millipore Q water and were mounted dorsally on a large microscope cover glass (24 × 50 mm, VWR, Cat# 16004-098) using double-sided scotch tape and covered with a small microscope cover glass (12CIR-1, Thermo Fisher Scientific, Cat# 1254580). Millipore Q water (40 µl) was applied to the tip of the larval head, and the preparation was placed on the stage of a Nikon eclipse Ti inverted microscope. Images were obtained every 500 ms, starting 15 s before application and ending 105 s after ligand application. Each recording was initiated by applying water (40 μl) to set a baseline. The first ligand solution (40 μl of bitter chemical or sugar) was applied thereafter, followed by five washes with carrier (100 μl of water). After a 3 min pause to allow the preparation to recalibrate, a second ligand solution (40 μl bitter chemical or sugar) was applied. To assure validity in experiments with Gr28 mutants and rescues, each recording was concluded with application of caffeine, and recordings were included only if caffeine generated a positive response. Baseline fluorescence, which was determined from the average of five frame measurements from a region next to the cell immediately before ligand application, was subtracted from the actual measurements. ∆F/F (%) = (fluorescence light intensity of the cell body – baseline/baseline) × 100. ∆F/F (max %) is the maximum value within 40 s after ligand application.

To generate the Gr28b.c-LexA driver, a 1.3 kb DNA fragment immediately upstream of the Gr28b.c start codon was amplified from w¹¹¹⁸ flies using a forward (5′-AATCTAGGTACCCCGGCTGCTCGTCTCCCTGGATGT-3′) and a reverse (5′- CGTCAAACTAGTGACCGCTTCGTTTGAGCTTCAACC-3′) primer. Acc65I and SpeI sites included in the primer sequence (underlined) were incorporated such that the amplified fragment was amenable to directional cloning into the LexA vector CMC105 (Larsson et al., 2004). The clone chosen was confirmed by DNA sequence analysis. Transgenic flies were generated by standard P-element transformation of w¹¹¹⁸ embryos (Rainbow Transgenic Flies Inc, Camarillo, CA).

Statistical analyses were conducted using Prism software 9.5.1 (GraphPad Software). Larval two-choice preference assay and Ca²⁺ imaging data were analyzed for normal distribution using D’Agostino–Pearson omnibus and Shapiro–Wilk normality tests. When groups did not meet the assumption for normal distribution, nonparametric statistics was used. For comparison between multiple groups, one-way ANOVA or Kruskal–Wallis test by ranks (nonparametric one-way ANOVA) was performed to test for difference of mean or rank distribution. As a post hoc test, Bonferroni’s or Dunn’s (nonparametric) multiple comparison tests were employed to compare two specific groups. One-way ANOVA with Bonferroni’s multiple comparison tests were used in Figure 2B. Kruskal–Wallis test by ranks with Dunn’s multiple comparison tests were used in Figures 3—5F and G. For comparison between two groups, Mann–Whitney U test (nonparametric t -test, Figures 4A,, 5D and E, Figure 2—figure supplement 1B, and Figure 4—figure supplement 1) with two-tailed P- value were used. The sample size for larval two-choice preference assays and Ca²⁺ imaging experiments werewas based on Mishra et al., 2018.