Detection of sweet tastants by a conserved group of insect gustatory receptors

An Ectopic Expression System to Decode Taste Receptors.

We selected the CO₂-sensing ab1C olfactory neuron for ectopic expression of Grs based on our previous study (12). The ab1C neuron, housed with three other neurons in ab1 large basoconic sensilla in the fly antenna, is unique in that it expresses Gr21a and Gr63a but no members of the Or or Ir gene families (20, 21, 26, 27). Exogenous receptors can be expressed in the ab1C neuron via the GAL4/Upstream Activating Sequence (UAS) system by using either Gr21a or Gr63a promoter sequences (Fig. 1A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

An in vivo ectopic expression system for analysis of individual Grs. (A) Schematic of ectopic expression in the ab1C neuron with trace from wild-type ab1 sensillum depicting activities of the four ab1 neurons. Glass micropipettes for tastant recordings contain sensillum lymph ringer (SLR) control (gray) or stimulus in SLR (blue). (B) Sample ab1 recordings in flies expressing Gr5a in ab1C neurons (ab1C:Gr5a-2x). Black dots indicate ab1C spikes. (C) Mean responses of ab1C:Gr5a-1x and ab1C:Gr5a-2x neurons. Baseline activity to SLR is not subtracted from stimulus-evoked activity. Sugars were tested at a concentration of 100 mM. Letters indicate statistical significance (P < 0.001; one-way ANOVA with Tukey’s post hoc test; n = 6–12). (D) Dose-dependent response to trehalose (n = 6–12). (E) Mean responses of ab1C:Gr5a neurons generated with Gr21a– or Gr63a–GAL4 as indicated to 100 mM sugars (n = 6). (F) Mean responses of ab1C:Gr5a-2x neurons in wild-type (+Gr63a) or ΔGr63a (–Gr63a) flies to 100 mM sugars (n = 10–14). All genotypes in E and F were compared with each other by using two-way ANOVA with Tukey’s post hoc test. Only genotypes with one copy of UAS–Gr5a (E) are significantly different from genotypes with two copies of UAS–Gr5a (F) (P < 0.05).

One consideration is that tastants would have to be delivered by contact with the lymph of the ab1 sensillum. Previous studies have shown that nonvolatile agonists or antagonists of the Or coreceptor, Orco, are able to act on olfactory neurons simply by their inclusion in the recording micropipette (28, 29). Others have reported injected delivery of inhibitors into the sensillum lymph that mimics the effects of in vitro studies (30⇓–32). We therefore dissolved tastants in electrolyte solution and delivered them by piercing the sensillum. As a control, we used this technique to deliver 10⁻⁴ ethyl acetate [solubility in water ∼8% (vol/vol) at 25 °C] and compared ensuing activity with responses obtained via an odor delivery system. In both cases, we observed robust responses to ethyl acetate (Fig. S1). Although neuronal activity was lower when the stimulus was delivered by micropipette, these results show that the dissolved stimulus is sufficient for olfactory neuron activity.

We next systematically characterized the suitability of the ab1C ectopic expression system for decoding taste receptors. We first expressed a trehalose receptor, Gr5a (19, 33), and asked whether its expression was sufficient to confer trehalose response in the ab1C neuron. Each ab1 sensillum was pierced twice for recordings, first to measure baseline activity with electrolyte alone and a second time to measure ligand-evoked activity by using electrolyte solution containing either trehalose or sucrose. Application of sucrose did not alter the firing rate of ab1C neurons expressing Gr5a, designated ab1C:Gr5a (Fig. 1 B and C). By contrast, trehalose evoked a significant increase in ab1C:Gr5a activity. Subsequent recordings with electrolyte solution showed a return of ab1C activity to baseline levels (Fig. S1). Notably, the response was dependent on the gene dosage of Gr5a (Fig. 1C) as well as the concentration of trehalose (Fig. 1D). Regardless of the number of copies of UAS–Gr5a, response to sucrose was indistinguishable from that of electrolyte alone (Fig. 1 B and C), suggesting that Gr5a-mediated response was specific for trehalose.

Flies lacking Gr5a have reduced taste neuron responses not only to trehalose, but also to three other tested sugars: glucose, methyl-α-glucopyranoside (m-glucoside), and melezitose (9). One possible explanation is that Gr5a is directly involved in detection of all four sugar tastants. Alternatively, Gr5a may be involved only indirectly, perhaps as a coreceptor, for detection of glucose, m-glucoside, and melezitose. To distinguish between these possibilities, we examined responses of the ab1C:Gr5a neuron to the four sugars. We also took this opportunity to compare response profiles of ab1C:Gr5a neurons generated independently using either Gr21a–GAL4 or Gr63a–GAL4.

Our recordings showed that the ab1C:Gr5a neuron responded to all Gr5a-dependent sugars (Fig. 1E), indicating a direct role for Gr5a in detecting these sweet tastants. Moreover, similar levels of mean responses to each sugar were obtained irrespective of the GAL4 driver that was used (Fig. 1E). Therefore, in subsequent experiments, Gr21a–GAL4 and Gr63a–GAL4 were treated as equivalent, and receptors were expressed singly in the ab1C neuron via two copies each of GAL4 and UAS transgenes.

It is curious that Gr5a was functional in the ab1C neuron when other attempts at heterologous expression of Gr5a in vivo have failed. We wondered whether the endogenous Gr21a/Gr63a receptor was a factor in its suitability. Although a Gr21a mutant is not available, we examined Gr5a-mediated responses in ab1C neurons of ΔGr63a flies (20). Responses to the four sugars were not significantly altered by the absence of Gr63a (Fig. 1F), indicating that an intact Gr21a/Gr63a receptor is not necessary for functional expression of an exogenous Gr gene, Gr5a. Importantly, no new responses were observed in ab1C:Gr5a neurons of Gr63a mutants (sucrose, 0.17 ± 0.21; maltose, 0.08 ± 0.49; maltotriose, 0.08 ± 0.45; fructose, −0.42 ± 0.52; glycerol, 1.00 ± 0.68; n = 6 for each stimulus), thereby ruling out the possibility that Gr63a may interfere with Gr5a function. Based on these results, subsequent ectopic expression experiments were carried out in Gr63a⁺ flies.

All Sweet Grs Confer Responses to Sweet Tastants.

Having established the ab1C neuron as a suitable host for Gr proteins, we performed a systematic analysis of tastant response profiles of the entire sweet Gr clade. Grs were individually expressed in the ab1C neuron, and ab1C:GrX responses were measured against a panel of sweet tastants. Based on our previous observations that only ∼15 compounds from a large stimulus panel strongly activated sweet taste neurons of L-type labellar sensilla, and that all responses were dependent on either Gr5a or Gr64a (9), we selected a diagnostic subset of nine tastants to represent the structural diversity of ligands and the breadth of responses in taste neurons. Specifically, we included four sugars that depend on Gr5a for responses in taste neurons, four others that depend on Gr64a, and a known ligand of Gr64e, glycerol (8, 9, 12). Importantly, the selected tastant panel represents sugars and sugar alcohols that are behaviorally very attractive to flies (34). Sugars were tested at a concentration of 100 mM and glycerol at 10% (vol/vol), concentrations that typically evoke robust responses in taste neurons. One exception was maltotriose, which was tested at a concentration of 250 mM based on initial analysis of dose-dependent responses of ab1C:Gr64a (Fig. S2). Each Gr–tastant combination was tested in at least six sensilla derived from at least two different flies.

ab1C neurons expressing either Gr5a or Gr64a were responsive to complementary subsets of tastants. The ab1C:Gr5a neuron responded to the four Gr5a-dependent sugars (as shown in Fig. 1F), but not to the four Gr64a-dependent sugars or to glycerol (Fig. 2A). Likewise, the ab1C:Gr64a neuron was selectively activated by Gr64a-dependent sugars and glycerol, but not by Gr5a-dependent sugars. Thus, Gr5a and Gr64a, which are broadly required for sugar sensing (9), are directly involved in recognition of nonoverlapping subsets of sweet tastants.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Tastant response profiles of sweet taste receptors. (A) Mean electrophysiological responses of ab1C:GrX neurons. All sugars were tested at a concentration of 100 mM, except maltotriose at 250 mM and glycerol at 10% (vol/vol). For each data point, n = 6–14. *P < 0.05; **P < 0.001 [vs. control ab1C flies (w¹¹¹⁸)]. (B) Mean electrophysiological responses of Gr5a expressed alone (–) or with the indicated receptor in ab1C neurons. For each data point, n = 6–7. *P < 0.05; **P < 0.001 [vs. Gr5a alone (–)]. (C) Phylogenetic tree of sweet Grs (Left) and heat map of mean neuronal responses of ab1C:GrX neurons to indicated sweet tastants. Data are the same as in A. Heat map was made with PAST (http://folk.uio.no/ohammer/past).

Of the eight Grs that we expressed in the ab1C neuron, every one conferred a significant response to at least one of the tested compounds (Fig. 2A), demonstrating that all receptors of the sweet clade participate in detection of sweet tastants. With the exception of sucrose, each tastant evoked a response from more than one sweet taste receptor. Conversely, each receptor was activated by more than one sweet tastant.

Previous genetic analyses have implicated possible heteromeric complexes in sugar detection. For example, Gr64f is required in combination with Gr5a for response to trehalose (13). We therefore asked whether coexpressed pairs of receptors conferred synergistically higher responses compared with either receptor alone. Specifically, we tested Gr5a in pairwise combinations with the other sweet receptors and measured responses to the Gr5a-dependent sugars. We observed that the presence of another Gr did not result in significant increases in mean sugar responses (Fig. 2B). In fact, coexpression of Gr64a generally depressed responses to Gr5a-dependent sugars, which is consistent with their nonoverlapping functions in taste neurons (9). Thus, at least in the context of the ab1C neuron, expression of an individual sweet taste receptor suffices to confer responses to sweet tastants.

A heat map of ab1C:GrX responses ordered by the sequence relationship between receptor proteins is shown in Fig. 2C. The distribution of responses across the entire set of receptors revealed extensive overlap of Gr–ligand interactions, showing that individual tastants are detected by multiple sweet Grs. Interestingly, even with the limited panel of tastants that we analyzed, two loosely divided subgroups of receptors emerged from the response matrix: one group sharing functional overlap with Gr5a and a second with Gr64a. These results provide the framework to understand how largely distinct subsets of the sweet receptor clade may function with Gr5a or Gr64a to encode sweet tastants.

Ectopic Responses Are Validated by Mutant Analysis.

We wanted to validate the observed Gr response profiles in ab1C neurons with physiological analysis of neurons in which these receptors are expressed. We therefore compared the responses of ab1C:GrX neurons to those of sweet taste neurons in the corresponding Gr mutants. Single Gr mutants have been reported for four of these receptors, Gr5a, Gr61a, Gr64a, and Gr64e (8, 9, 12, 33); in addition, a Minos element insertion allele became available for Gr64f (35). Importantly, all five of these receptors are present in sweet taste neurons of L-type sensilla in the labellum as demonstrated by expression or functional analysis (5, 9). We measured responses of sweet taste neurons in L-type sensilla (Fig. S3A) in each of these mutants using our panel of nine sweet tastants, which were tested at concentrations of 100 mM (Fig. 3A and Fig. S3B) and 1 M (Fig. S3C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Sweet taste responses in Gr mutants. (A) Mean responses of sweet taste neurons in L-type sensilla to indicated tastants. Indicated genotypes were: w¹¹¹⁸ (wild-type), ΔEP(X)-5 (ΔGr5a), Gr64f^MB12243 (ΔGr64f), Gr64e^MB03533 (ΔGr64e), Gr61a¹ (ΔGr61a), and Gr64a¹ (ΔGr64a). All stimuli were tested at a concentration of 100 mM, except glycerol (10%). *P < 0.05; **P < 0.001 (one-way ANOVA with one-tailed Dunnett’s t test vs. wild-type; n = 6–22). (B) Heat maps of ab1C:GrX responses (Upper) and percent reduction in taste neuron responses in corresponding GrX mutants (Lower); the latter only includes data points significantly different from wild-type in C. Percent loss of response was calculated by using [(wild type – mutant)/wild type] x 100. Heat maps were made by using JMP 10 (www.jmp.com). (C) Scatter plot of percent loss of response in Gr mutant and ab1C:GrX response (gain) for each GrX–ligand combination. Filled circles indicate taste neuron responses that are significantly reduced in mutant flies (ΔGrX); open circles indicate those that are not. Shaded area indicates ab1C:GrX responses that are not statistically significant.

We observed complete overlap between Gr5a– and Gr64a–ligand interactions obtained in ab1C:GrX neurons and by loss of sensitivity in Gr5a and Gr64a mutants (Fig. 3 A and B). Although not as complete, there was also substantial overlap between the results of gain-of-function and loss-of-function analyses for Gr61a, Gr64e, and Gr64f (Fig. 3B and Fig. S4A). Of most importance is that every single one of the observed ab1C:GrX responses was validated by a significant reduction in sweet taste neuron response in the corresponding Gr mutant. We visualized the relationship between gain and loss values in a scatter plot (Fig. 3C) and ran a Spearman’s correlation, which showed a strong positive correlation (r_s = 0.759, n = 45, P < 0.001). Together, these results authenticate the functional analysis of singly expressed Grs in the ab1C neuron. Moreover, the lack of unsupported interactions in the ab1C neuron makes it unlikely that Gr21a/Gr63a alter the response specificities of ectopically expressed receptors.

The mutant analysis also revealed eight instances in which a Gr (specifically, Gr61a, Gr64e, and Gr64f) was necessary for the full extent of a response in sweet taste neurons but was not sufficient to confer sensitivity to the corresponding tastant in the ab1C:GrX neuron (Fig. 3B). Moreover, the eight responses belonging to this category were not restricted to either Gr5a- or Gr64a-dependent tastants in that Gr61a, Gr64e, and Gr64f mutants all showed significant reductions in response to one or more tastants of both subsets. Our results are consistent with a model in which the tuning breadth of sweet taste neurons is collectively determined by each of the many Grs that are expressed in them and support extensive functional and overlapping interactions among individual receptors of the sweet clade.

Ectopic Analysis Is Validated by a Bitter Receptor.

To further investigate the ab1C neuron as a reliable system for reporting GrX activity, we tested a bitter receptor, Gr59c, which has been implicated in detection of specific aversive compounds (5). Gr59c is derived from the I-a class of labellar bitter taste neurons, which respond to lobeline, denatonium, and escin. Misexpression of Gr59c in other classes of labellar bitter neurons increases responses to I-a ligands (5). Consistent with these results, expression of Gr59c in the ab1C neuron conferred responses to lobeline, denatonium, and escin, but not to caffeine, theophylline, or sucrose, which do not activate I-a bitter neurons (Fig. 4 A and B). The Gr59c response profile was not altered by the absence of Gr63a (Fig. 4C). Thus, similar to the sweet receptor Gr5a, the specificity of the Gr59c bitter receptor appears to be preserved in the ab1C neuron.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Functional analysis of other Drosophila and mosquito Grs. (A) Mean response of ab1C neurons in control w¹¹¹⁸ flies (ab1C) and flies expressing Gr59c (ab1C:Gr59c). *P < 0.05 (vs. control; n = 6–12). Sucrose was tested at a concentration of 100 mM and bitter compounds at 10 mM. (B) Dose-dependent response of ab1C:Gr59c (n = 6–12). (C) Sample recordings and mean responses of ab1C:Gr59c in wild-type (+Gr63a) and ΔGr63a (–Gr63a) flies to indicated stimuli (10 mM). Genotypes are not significantly different (P > 0.05; n = 6). Concentrations were as in A. (D) Dose-dependent response of ab1C:DmGr43a (n = 6–11). (E) Mean responses of ab1C neurons in control w¹¹¹⁸ flies (ab1C) and flies expressing Drosophila Gr43a (ab1C:DmGr43a) or its mosquito ortholog (ab1C:AgGr25) to indicated stimuli tested at a concentration of 100 mM, except maltotriose (250 mM) and glycerol (10%). *P < 0.05; **P < 0.001 (vs. ab1C; n = 6–12).

Functional Expression of an Orthologous Pair of Fly and Mosquito Grs.

Recent studies showed that Gr43a functions as an internal fructose-sensing receptor in vivo (36) and confers fructose response when heterologously expressed in COS-7 cells (22). We therefore tested whether expression of Gr43a was sufficient to confer fructose sensitivity in ab1C neurons. Recordings with a range of concentrations revealed a dose-dependent response to fructose in ab1C:Gr43a neurons (Fig. 4C). The concentration range over which Gr43a was active in the ab1C neuron was comparable to that observed by imaging Ca²⁺ activity in Gr43a-labeled neurons in the legs and the lateral protocerebrum region of the brain (36).

Given the compatibility of the ab1C neuron with Drosophila sweet and bitter taste receptors, we wanted to test whether it could be adopted for functional analysis of taste receptors from other insects such as A. gambiae. Gr genes of D. melanogaster and A. gambiae are highly divergent with few one-to-one orthologs, which include the Gr43a and AgGr25 pair (Fig. S5). We expressed AgGr25 in the ab1C neuron using a UAS–AgGr25 transgene constructed from A. gambiae genomic DNA. We assayed responses of ab1C neurons expressing either DmGr43a or AgGr25 to the nine selected sweet tastants. Both DmGr43a and AgGr25 conferred robust responses to fructose and some other sugars, including glucose (Fig. 4D), which is present in the hemolymph along with fructose and trehalose (37, 38) (Fig. 4E). In fact, there was a large overlap in excitatory responses of DmGr43a and AgGr25, with maltotriose being the only exception that evoked a response exclusively from the fly ortholog. Thus, both DmGr43a and AgGr25 are responsive to sweet tastants. Moreover, the observation that AgGr25 can function in the absence of any other mosquito proteins suggests compatibility between the ab1C neuron and Grs of A. gambiae and potentially other insects as well.