Artificial sweeteners differentially activate sweet and bitter gustatory neurons in Drosophila

Abstract

Artificial sweeteners are highly sweet, non-nutritive compounds that have become increasingly popular over recent decades despite research suggesting that their consumption has unintended consequences. Specifically, there is evidence suggesting that some of these chemicals interact with bitter taste receptors, implying that sweeteners likely generate complex chemosensory signals. Here, we report the basic sensory characteristics of sweeteners in Drosophila, a common model system used to study the impacts of diet, and find that all noncaloric sweeteners inhibited appetitive feeding responses at higher concentrations. At a cellular level, we found that sucralose and rebaudioside A co-activated sweet and bitter gustatory receptor neurons (GRNs), two populations that reciprocally impact feeding behavior, while aspartame only activated bitter cells. We assessed the behavioral impacts of sweet and bitter co-activation and found that low concentrations of sucralose signal appetitive feeding while high concentrations signal feeding aversion. Finally, silencing bitter GRNs reduced the aversive signal elicited by high concentrations of sucralose and significantly increased sucralose feeding behaviors. Together, we conclude that artificial sweeteners generate a gustatory signal that is more complex than “sweetness” alone, and this bitter co-activation has behaviorally relevant effects on feeding that may help flies flexibly respond to these unique compounds.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-08467-4.

Introduction

Artificial sweeteners are intensely sweet, non-nutritive compounds that have emerged as popular way to reduce caloric intake while maintaining palatability¹. Sweeteners have become an increasingly common feature of modern diets²a trend that is expected to persist³. The FDA and other regulatory agencies have currently approved several artificial sweeteners for general consumption, including sucralose (Splenda^®), saccharin (Sweet’n Low^®), aspartame (Equal^®, NutraSweet^®), neotame(Newtame^®), advantame (NutraSweet^®), acesulfame potassium (Sunett^®, Sweet One^®), and sugar alcohols^4,5. Beyond these synthetic chemicals, naturally derived compounds like stevia (Truvia^®) and its extracts, stevioside and rebaudioside A (Reb A), have also grown in popularity as low-calorie sweet additives⁶. While these sweeteners are considered biologically inert and safe for consumption⁷their increased prevalence has led to apprehension about their long-term safety⁸. The World Health Organization (WHO) has recently advised against using sweeteners for weight loss or dieting, citing potentially harmful physiological effects⁹. Both animal and clinical investigations have suggested gastrointestinal^10–13metabolic^14–16and cardiovascular^5,17,18 risks associated with sweetener consumption. While many of these health concerns have centered on metabolic outcomes and the implications of experiencing sweetness that is dissociated from energy intake¹⁴artificial sweeteners are considered to be hundreds to thousands of times sweeter than sucrose for humans¹⁹. These unique chemosensory properties have raised questions about how the gustatory system encodes and processes these compounds. Previous work has demonstrated that sweeteners are detected by human sweet taste receptors (T1R2/T1R3)^20–22, but in vitro evidence suggests that sucralose also binds to bitter receptors (T2Rs)^23,24including those expressed extra-orally²⁵. These findings align with behavioral evidence indicating that sucralose and other sweeteners possess a bitter “aftertaste” quality for both humans and rodents^26–28. However, the exact cellular mechanisms and behavioral consequences of sweetener chemosensation remain unclear, particularly regarding potential off-target bitter signaling.

Drosophila melanogaster has been an advantageous model organism for studying both gustatory processing²⁹ and how artificial sweeteners impact health-related phenotypes^30–41including feeding. Initial investigations into sweetener feeding behavior demonstrated that flies preferentially consume sucralose, aspartame, and saccharin, implying a conserved attraction to these compounds⁴². Later research examined the long-term effects of sweetener exposure and proposed that dietary sucralose leads to enhanced sweet taste sensitivity and increased food intake via a neuronal starvation response³⁹. In contrast, another study found that sucralose deters feeding⁴¹leading to further controversy about the effects of dietary sweeteners⁴⁰. However, this debate did not consider how sucralose is detected by flies at a sensory level.

In Drosophila, taste processing is initiated by peripheral gustatory receptor neurons (GRNs) distributed throughout several structures, but most prominently within taste sensilla located on the fly’s primary mouthpart - the labellum^43,44. Labellar GRNs detect chemical stimuli and directly transmit taste signals to the subesophageal zone (SEZ) of the fly brain for higher-order processing⁴⁵. Among the five recognized GRN classes, sugar-sensing “sweet” cells are the primary appetitive GRN population and express sugar-specific gustatory receptors, such as Gr64a-f^46–48. In contrast, bitter GRNs represent the main aversive GRN population and detect unpalatable compounds via several types of taste receptors, including Gr66a^45,49. Accordingly, activation of sweet or bitter GRNs has strong but opposite effects on feeding^49,50. Taking advantage of the abundant genetic and neurobiological tools available in Drosophila^51,52previous investigation of these taste cell populations has revealed complex and combinatorial mechanisms of gustatory processing for several classes of tastants^48,53–57.

While prior research has indicated mild attraction to sweeteners at a behavioral level⁴²they only looked at the effects of stimulating taste cells in the legs (tarsal GRNs). How flies respond to sweetener stimulation on the labellum, where the majority of GRNs are located⁵⁸remains unclear. Moreover, the cellular responses elicited by these compounds in Drosophila have not been described. Here, we recorded behavioral responses to labellar sweetener stimulation, showing that flies exhibit concentration-dependent taste responses to different compounds. We then characterized the neuronal responses to sucralose and found that it co-activates sweet and bitter GRNs to modulate feeding behavior. We also included aspartame and Reb A in these analyses to demonstrate that bitter cellular responses extend to sweeteners with different chemical structures. Together, our findings shed light on the basic chemosensory underpinnings of artificial sweeteners and specifically highlight bitter taste cell co-activation as an important consideration for future sweetener research.

Results

Flies exhibit dose-dependent taste responses to multiple artificial sweeteners

We first sought to describe the taste responses to artificial sweeteners at a behavioral level. To accomplish this, we assembled a panel of common sweeteners and tested how flies respond to each compound using the proboscis extension response (PER) assay⁵⁹specifically stimulating a fly’s labellum with a test solution and recording if they extend their proboscis to initiate feeding (Fig. 1A). The proportion of flies that extend to each test stimulation can be quantified and used as a measure of appetitive taste responsiveness. We used a concentration series for each of the seven sweeteners in our panel: maltitol, saccharin, cyclamate, aspartame, neotame, rebaudioside A (Reb A), and sucralose. Concentrations were chosen based on previous work in Drosophila⁴²and include lower concentrations in the range found in sweetened drinks⁶⁰ as well as some higher concentrations since we did not know if the sweeteners would be similarly potent ligands for fly taste receptors. Across our panel, we observed dose-dependent taste responses to the different sweeteners, with most compounds eliciting the highest level of PER at low/moderate concentrations (Fig. 1).

Fig. 1.

Flies exhibit diverse and concentration-dependent taste responses to a panel of artificial sweeteners. (A) Labellar proboscis extension response (PER) assay schematic showing stimulation with a test solution and a resulting extension. (B-H) Labellar PER with a panel of seven sweeteners presented to flies in a concentration series. %PER refers to the fraction of flies that extended to each respective concentration. n = 46–99. (I) Two-Choice FLIC feeding assay with a sucralose concentration series. Flies were allowed to choose between two food sources: H₂O and one of the four sucralose concentrations. Preference index over three hours is depicted. n = 15–18 flies per genotype. All data from mated female Canton-S flies. Data plotted as mean ± SEM. ns = no significance, **p < 0.01, ***p < 0.001, ****p < 0.0001 by ordinary one-way ANOVA with Dunnett’s multiple comparisons test. Graphics were generated with BioRender.com.

Unlike the other non-nutritive compounds in our panel, maltitol is a semi-caloric sugar alcohol with a similar level of sweetness as sucrose in humans⁶¹. Maltitol exhibited the strongest responses of any sweetener we tested and PER rates increased at higher concentrations (Fig. 1B), mimicking sucrose PER^62,63 and implying that flies are more strongly attracted to even partially nutritive sweeteners. Stimulating with cyclamate, a relatively mild sweetener used widely in Europe⁶⁴also resulted in higher PER rates compared to the other sweeteners (Fig. 1D). However, cyclamate is the sodium salt of cyclamic acid, and its PER curve resembles that of sodium^48,65suggesting that flies likely detect and respond to the salt taste in addition to any of cyclamate’s sweetness. Responses to saccharin, aspartame, neotame, and Reb A all shared a similar pattern - a small peak at 0.001–0.01 mM followed by very minimal responses at higher concentrations (Fig. 1C, E, F and G).

In comparison, flies were more attracted to sucralose, with responses peaking around 10 mM and remaining consistent at higher concentrations (Fig. 1H). Because of this unique response pattern and its prominence in previous fly sweetener research^37–42we aimed to more thoroughly characterize how flies interact with sucralose across different concentrations. We used the Fly Liquid Interaction Counter (FLIC)⁶⁶ to quantify nuanced feeding metrics over a longer duration. Flies were individually loaded into chambers containing two food options: water and sucralose (2-choice FLIC). Over a three-hour period, preference index, interactions, feeding events, and mean event duration were recorded for each fly. We found that flies exhibit a strong preference for low concentrations of sucralose (10–50 mM), but this preference was neutralized at high concentrations (500 mM) (Fig. 1I, Fig. S1A). Additionally, we analyzed how the feeding preference for these sucralose concentrations changed over time during the FLIC assay. There appeared to be a subtle trend towards a decrease in 100 mM preference index at the final time period, but we found that the preferences were not statistically different from the first time period for any concentration (Fig. S1B), implying that their initial sensory experience largely shapes their ongoing feeding behavior. Combined with mild responses in our PER experiment, a lack of feeding preference for high concentration sucralose suggests that the taste system may be integrating additional gustatory signals beyond just saccharinity.

Sucralose activates sweet taste cells to promote feeding attraction

To more thoroughly characterize these taste responses, we explored how artificial sweeteners are encoded by the fly gustatory system at a cellular level. As sucralose is considered 600 times sweeter than sucrose for mammals⁶⁷we began by analyzing the neuronal responses to sucralose in sweet GRNs (Gr64f+), the primary appetitive taste cell population in Drosophila^46,47. This functional analysis was accomplished using in vivo calcium imaging of labellar GRNs^{48,53,54,68–72}. We found that stimulation with 10 mM elicited a clear response in at least one fly, but 100 and 500 mM sucralose elicited a significant increase in sweet GRN calcium activity compared to water, a negative control (Fig. 2A), indicating that these appetitive taste cells respond to sucralose. Interestingly, the magnitude of these responses was surprisingly moderate, particularly when compared to the 1 M sucrose positive control (Fig. 2A).

Fig. 2.

Sucralose activates sweet GRNs to signal feeding attraction at low concentrations. (A) In vivo calcium imaging of sweet GRNs (Gr64f > GCaMP6f) during labellar sucralose stimulation. Calcium responses measured as ΔF/F (Z-score) over time at each concentration (left) and peak ΔF/F (right). Blue lines under each curve indicate when the stimulus was on the labellum. n = 12 flies. (B) Prolonged taste modulation PER paradigm consisting of three consecutive stimulations: an initial water stimulation A, a 50 mM sucralose stimulation, and a delayed water stimulation B. Modulation assessed after 20 s (right) or 5 min (left). n = 39–40 flies. Data from Canton-S flies. (C) Optogenetic silencing of sweet GRNs (Gr64f > GtACR1) during labellar sucralose PER. Flies were stimulated with 50 mM sucralose while exposed to green light. %PER to this stimulation was compared between flies pre-fed retinal (+ Retinal) and control flies (-Retinal). n = 35–38 flies per condition. (D) Constitutive silencing of sweet GRNs (Gr64f > Kir2.1) during the 2-choice FLIC feeding assay. Flies were allowed to choose between two food options, H₂O and 100 mM sucralose, for three hours. Preference index calculated based off number of interactions with each food option. n = 20–22 flies per genotype. All experiments used mated females. Data plotted as mean ± SEM. ns = no significance, *p < 0.05, ***p < 0.001, ****p < 0.0001 by repeated-measures ANOVA with Dunnett’s multiple comparisons test (A), Wilcoxon matched-pairs signed rank test (B), Fisher’s exact test (C), or ordinary one-way ANOVA with Dunnett’s multiple comparisons test (D). Graphics were generated with BioRender.com.

We then asked if this level of sweet activation represents a signal strong enough to impact feeding. Previous research on sweet GRNs revealed that both sucrose stimulation and optogenetic activation of these cells were sufficient to enhance subsequent PER to neutral stimuli, even after a prolonged delay⁵⁰. This finding suggests that activation of sweet GRNs primes the fly to respond appetitively to upcoming gustatory stimuli, thereby encouraging feeding. To determine if sucralose also impacts feeding in this manner, we utilized a similar prolonged taste modulation PER paradigm. Flies were presented with three consecutive stimulations: a baseline neutral water stimulation, a sucralose stimulation, and a second, delayed water stimulation (Fig. 2B). We found that stimulating flies with 50 mM sucralose, an attractive concentration (Fig. 1H and I), significantly enhanced PER to the second water stimulus both acutely (20 s) and after a prolonged delay (5 min) (Fig. 2B). Presenting two water stimuli alone did not enhance the second water response, confirming this is a consequence of the appetitive stimulus and not due to a change in water appetite (Fig. S2A). Next, we tested whether exposure to sucralose would also enhance the response to a second 50 mM sucralose stimulation, but there was no change (Fig. S2B). This implies that the appetitive signal elicited by 50 mM sucralose can enhance neutral stimuli but does not enhance a subsequent response to sucralose.

To verify that appetitive responses to sucralose were driven by sweet GRN activation, we silenced these neurons in two behavioral assays. First, we optogenetically silenced sweet GRNs via selective expression of GtACR1, an anion channel that hyperpolarizes neurons upon green light stimulation⁷³. This approach requires pre-feeding with all-trans-retinal, a chromophore necessary for proper GtACR1 function^68,74,75. Gr64f > GtACR1 flies without retinal pre-feeding served as experimental controls. We implemented this manipulation into our labellar PER assay by acutely exposing flies to green light alongside 50 mM sucralose. Consistent with our prediction, sucralose responses were significantly attenuated compared to inactive controls (Fig. 2C). Second, we used the 2-choice FLIC assay to test how sucralose feeding behavior would change when sweet GRNs were genetically inactivated via Kir2.1, an inward-rectifying potassium channel that constitutively silences neurons⁷⁶. Aligning with our optogenetic silencing results, Gr64f > Kir2.1 flies showed reduced preferences for both low and moderate concentrations of sucralose compared to genetic controls (Fig. 2D, S2C). The preference for 10 mM sucralose was significantly reduced with sweet GRN silencing, indicating that this low concentration must activate sweet GRNs to some extent, even if it was not detectable in the calcium imaging experiment (Fig. 2A). The preference for 100 mM sucralose started to flip to a negative preference, suggesting there may be an aversive pathway influencing behavior at moderate concentrations. Together, these findings suggest that sucralose activates sweet GRNs and that lower concentrations signal prolonged feeding attraction.

Sucralose activates bitter taste cells to limit feeding

While higher concentrations of sucralose produced increasing calcium signals in sweet GRNs (Fig. 2A), sucralose PER did not increase at higher concentrations (Fig. 1H), and our 2-choice FLIC revealed that flies do not prefer high concentrations. We reasoned that this discrepancy could be explained by additional GRN populations responding to sucralose and forming a more complex gustatory signal. Therefore, we hypothesized that aversive GRNs are also activated upon sucralose stimulation and that the resulting signal has biologically relevant effects on feeding. Using the same in vivo calcium imaging approach, we assessed how bitter GRNs (Gr66a+), the primary aversive GRN population, respond to sucralose detection. We found that stimulation with 100 and 500 mM sucralose significantly activated bitter GRNs (Fig. 3A). Unlike the mild calcium signals seen in sweet GRNs (Fig. 2A), the levels of bitter GRN activation by sucralose were much higher and comparable in magnitude to those elicited by the caffeine positive control stimulation (Fig. 3A).

Fig. 3.

Sucralose activates bitter GRNs to signal feeding aversion at high concentrations. (A) In vivo calcium imaging of bitter GRNs (Gr66a > GCaMP6f) during labellar sucralose stimulation. Calcium responses measured as ΔF/F (Z-score) over time at each concentration (left) and peak ΔF/F (right). Blue lines under each curve indicate when the stimulus was on the labellum. n = 11 flies. (B) Prolonged taste modulation PER paradigm consisting of two repeated 500 mM sucralose stimulations separated by a 20-second or 5-minute delay. n = 66–67 flies. Data from Canton-S flies. (C) Optogenetic silencing of bitter GRNs (Gr66a > GtACR1) during prolonged taste modulation PER assay. Flies received two consecutive 500 mM sucralose stimulations separated by 20 s while exposed to green light. %PER to the initial and delayed stimulations were compared between flies pre-fed retinal (+ Retinal) and control flies (-Retinal). n = 32 flies per condition. (D) Constitutive silencing of bitter GRNs (Gr66a > Kir2.1) during the 2-choice FLIC feeding assay. Flies were allowed to choose between two food options, H₂O and 100 mM sucralose, for three hours. Preference index calculated based off number of interactions with each food option. n = 18–26 flies per genotype. All experiments used mated females. Data plotted as mean ± SEM. ns = no significance, *p < 0.05, ***p < 0.001, ****p < 0.0001 by repeated-measures ANOVA with Dunnett’s multiple comparisons test (A), Wilcoxon matched-pairs signed rank test (B), Fisher’s exact test (C), or ordinary one-way ANOVA with Dunnett’s multiple comparisons test (D). Graphics were generated with BioRender.com.

Next, we tested if bitter GRN activation by sucralose represented a salient aversive feeding signal. Since 500 mM sucralose exhibited the strongest bitter GRN activation (Fig. 3A), we used this concentration as the representative bitter stimulus in our prolonged taste modulation PER assay. It was previously established that this paradigm also detects the ability of bitter GRN activation to suppress future taste responses⁵⁰. We tested if initial exposure to 500 mM sucralose could produce an aversive signal strong enough to attenuate subsequent taste responses to the same 500 mM stimulation. Flies responded strongly to the initial sucralose stimulation, but PER to the second stimulation was significantly weaker after either a 20-second or 5-minute delay (Fig. 3B). The reduced responses to the second stimulation suggest the initial detection produced an aversive “aftertaste” signal that diminished sucralose attraction. To determine if this aversive signal was able to suppress another appetitive stimulus, we compared the response to 20 mM sucrose before and after 500 mM sucralose exposure (Fig. S3A). We did not see any significant reduction in the sucrose response after 20 s–5 min, suggesting that sucralose exposure is not sufficient to reduce the response to other appetitive stimuli.

To validate that this sucralose-induced aversive signal is driven by bitter GRN activation, we predicted that silencing these cells would dampen the aversive component of sucralose and enhance taste responses as a result. We optogenetically silenced bitter GRNs in our aversive prolonged taste modulation PER assay by exposing flies to green light during two consecutive 500 mM sucralose stimulations separated by a 20-second delay. Taste responses to the initial stimulation were similarly strong for both groups, but the flies pre-fed with retinal exhibited a significantly higher level of PER with the second stimulation compared to inactive controls (Fig. 3C). These findings suggest that silencing bitter GRNs reduces the aversive signal elicited by 500 mM sucralose, confirming that the activation of these neurons drives the unpalatable aspect of sucralose detection. We also constitutively silenced bitter GRNs using Kir2.1 during the 2-choice FLIC assay to further assess the impact of bitter GRN activity on sucralose feeding. Flies with silenced bitter GRNs showed increased sucralose preference, interactions, and events compared to genetic controls (Fig. 3D, S3B). To confirm these results with acute silencing, we repeated our 2-choice FLIC experiment in flies with bitter GRNs optogenetically silenced by green light exposure throughout the assay⁷⁷. Compared to controls, flies pre-fed with retinal exhibited a significantly stronger sucralose preference (Fig. S3C). Collectively, our results demonstrate that sucralose co-activates sweet and bitter GRNs to form a concentration-sensitive gustatory code comprised of opposing attractive and aversive signals that reciprocally modulate feeding.

Aspartame and Reb A minimally activate GRNs to impact feeding behaviors

To broaden our investigation of sweetener chemosensation, we assessed the cellular and behavioral responses for two additional chemicals: aspartame and Reb A. We chose these two compounds based on their popularity^78,79 and to ensure our analysis included sweeteners with diverse chemical structures. In our initial PER experiments, both aspartame and Reb A produced minimal taste responses, particularly at high concentrations (Fig. 1E and G). If the fly gustatory system encodes these chemicals like sucralose, this response pattern could indicate sweet and bitter GRN co-activation. Surprisingly, we did not observe any cellular responses to aspartame in sweet GRNs (Fig. 4A, S4A). However, aspartame stimulation induced significant calcium activity in bitter cells (Figs. 4B, S4B), including an OFF peak with stimulus removal. Despite clear bitter activation, optogenetic silencing of bitter GRNs during low or high concentration aspartame stimulation had no effect on taste responses in our prolonged taste modulation PER assay (Fig. S4C). Moreover, constitutive bitter GRN silencing did not alter aspartame feeding behavior (Fig. 4C, S4D), implying either that this signal is not salient for feeding or that without a simultaneous appetitive signal, aspartame will never be preferred.

Fig. 4.

Aspartame and Reb A exhibit limited GRN responses that minimally impact feeding. (A, B) In vivo calcium imaging of sweet (A) and bitter (B) GRNs during labellar aspartame stimulation. Calcium responses measured as peak ΔF/F (Z-score). n = 12 flies per experiment. (C) Constitutive silencing of bitter GRNs (Gr66a > Kir2.1) during the 2-choice FLIC feeding assay. Flies were allowed to choose between two food sources, H₂O and 10 mM aspartame, for three hours. Preference index calculated based off number of interactions with each food option. n = 26–27 flies per genotype. (D, E) In vivo calcium imaging of sweet (D) and bitter (E) GRNs during labellar Reb A stimulation. Calcium responses measured as peak ΔF/F (Z-score). n = 13 (D) and 11 (E). (F) Constitutive silencing of bitter GRNs (Gr66a > Kir2.1) during the 2-choice FLIC feeding assay. Flies were allowed to choose between two food options, H₂O and 10 mM Reb A, for three hours. Preference index calculated based off number of interactions with each food option. n = 21–28 flies per genotype. (G) Proposed model of how sweet and bitter GRNs encode several artificial sweeteners and how these signals impact feeding. Red-green gradients depict how low concentrations of sucralose and Reb A predominantly activate sweet GRNs while higher concentrations co-activate bitter GRNs. Aspartame only activates bitter GRNs. Solid lines represent signals that affect feeding while dotted lines represent signals that do not significantly impact feeding. All experiments used mated females. Data plotted as mean ± SEM. ns  no significance, *p < 0.05, **p < 0.01, ****p < 0.0001 by repeated-measures ANOVA with Dunnett’s multiple comparisons test (A, B, D, E) or ordinary one-way ANOVA with Dunnett’s multiple comparisons test (C, F). Graphics were generated with BioRender.com.

In contrast, Reb A mildly activated both sweet and bitter GRNs. Sweet GRNs exhibited increased calcium activity when stimulated with Reb A (Fig. 4D, S5A), similar to sucralose (Fig. 2A). Additionally, the magnitude of Reb A’s bitter responses was relatively low compared to the other sweeteners (Figs. 4E, S5B). Silencing bitter GRNs had no impact on Reb A taste response modulation (Fig. S5C) or Reb A feeding, which did show a mild positive preference (Fig. 4F, S5D). The lack of behavioral phenotypes following bitter silencing implies that the aversive signals are insufficient to override the appetitive signals. Overall, we propose that multiple artificial sweeteners are sensed by both appetitive and aversive GRN populations (Fig. 4G). Sweet GRNs mediate appetitive responses to lower concentrations of sucralose and Reb A, while bitter GRNs respond to aspartame and higher concentrations of sucralose and Reb A. Sweet and bitter co-activation by sucralose generates a balanced gustatory signal that modulates feeding in a concentration-dependent manner. In contrast, the distinct GRN responses to aspartame and Reb A have minimal effects on feeding behavior (Fig. 4G).

Discussion

Despite evidence indicating that bitter aftertastes associated with sweeteners are potentially due to bitter receptor (T2R) binding^23,24characterizing the impacts of sweetener sensation on consumption requires in vivo analyses that are currently challenging to achieve with mammals. These limitations have helped Drosophila become a common model organism for studying how dietary sweetener exposure impacts lifespan, feeding, and other phenotypes^30–42. However, these studies may have limited relevance for mammals and humans if the sensory experience associated with these compounds differ between species. In the current study, we described the cellular basis of Drosophila sweetener gustation in vivo and explored how bitter taste cell co-activation impacts sweetener feeding behavior.

Artificial sweeteners induce responses in both sweet and bitter gustatory neurons

Our labellar PER experiments showed only mild sweetener attraction – implying that sweeteners either do not strongly activate sweet GRNs, or they co-activate sweet and aversive GRNs. To confirm, we recorded calcium responses from sweet and bitter GRNs during in vivo sweetener stimulation. This technique has successfully been used to describe mechanisms of peripheral taste processing in living, awake flies^{48,53,54,68–71}an approach that is currently not feasible with mammalian model systems. Sweet GRNs responded to both sucralose and Reb A, but surprisingly, not to aspartame, while bitter GRNs responded strongly to sucralose and moderately to aspartame and Reb A. These results indicate that sucralose may represent a well-conserved sweet ligand for mammalian and Drosophila taste receptors, potentially due to its structural similarities with sucrose. However, sucralose appears to be a much more potent activator of T1R2/T1R3 compared to fly taste receptors^80,81. This difference could be driven by the significant molecular differences between mammalian T1Rs and Drosophila sugar GRs which are structurally unrelated, and fly GRs likely function as ionotropic cation channels instead of GPCRs⁸².

One recent study in flies suggests that the sugar receptor, Gr64a, mediates a dietary sucralose-induced increase in feeding⁸³. This result indicates that this canonical sugar receptor is involved in sucralose detection, in line with the sweetener’s structural similarities with sucrose. Interestingly, the same study also demonstrated that mutation of both Gr64a and Gr5a, a trehalose receptor, abolished sucralose-induced hyperphagia and instead reducing food consumption⁸³. This finding suggests that disrupting sweet signaling at the molecular level unveils an aversive effect of dietary sucralose, which could support our model of bitter co-activation. For the molecular basis of bitter sweetener signaling, in vitro studies with mammalian cells have helped reveal that two bitter receptors expressed in human taste papillae, TAS2R43 and TAS2R44, specifically detect saccharin and acesulfame K²⁴. In flies, it remains unclear which gustatory receptors mediate the robust neuronal responses we observed in bitter GRNs. Future research can explore whether sucralose and other chemicals bind to canonical bitter GRs or alternative classes of taste receptors, including ionotropic receptors (IRs), rhodopsins, TRP channels, or even bacterial peptidoglycan receptors expressed in bitter GRNs^{49,57,84–89}.

Aspartame and Reb A are 200 and 400 times sweeter than sucrose in humans, respectively^6,90. However, both sweeteners possess complex chemical structures that may be less compatible with fly taste receptors, leading to weaker activation compared to sucralose. Sweet and bitter GRN co-activation by Reb A is consistent with previous work demonstrating attraction to the sweetener in mice⁹¹but also molecular evidence of interacting with sweet and bitter human taste receptors^92,93. Interestingly, Reb A is considered less bitter for humans than the other popular stevia derivative, stevioside⁹⁴and future studies can compare the mechanisms of Reb A processing with other steviol glycosides. Conversely, aspartame only activated bitter GRNs and included a peak of activity upon stimulus removal. These bitter OFF-responses have been observed with other aversive tastants^95–97 and may indicate that more complex cellular kinetics contribute to aspartame processing. Aspartame only activating bitter cells aligns with a study in mice that reported minimal attraction to this sweetener⁹⁸suggesting the “sweetness” associated with this compound is not conserved in either model organism. There is also evidence indicating that aspartame interacts with TRPV1 receptors that are expressed in taste cells⁹⁹however the mechanisms underlying chemosensation of this particular sweetener, and their apparent divergence in humans, remain unclear.

Combining our cellular results with our behavioral findings, we propose a combinatorial encoding mechanism for sucralose that is concentration dependent. Sweet GRNs respond to lower concentrations to signal feeding attraction, whereas high concentrations co-activate bitter GRNs and signal feeding avoidance (Fig. 4G). This encoding pattern represents a common theme for the Drosophila peripheral gustatory system. Similar to sucralose, salts (NaCl)⁴⁸fatty acids (hexanoic acid)⁵⁵carboxylic acids (acetic acid)⁵⁴and amino acids (arginine)⁵⁶ activate sweet GRNs at low concentrations to promote feeding while higher concentrations activate bitter GRNs to deter feeding. This appetitive-aversive balance likely enables flies to flexibly respond to chemicals that are beneficial to some extent, but harmful in high quantities. Of note, not all synthetic molecules will show a similar pattern of concentration-dependent attraction. For example, the pesticide Flonicamid is only detected by aversive pathways¹⁰⁰. One caveat of our in vivo calcium imaging approach is that it does not directly measure neuronal activity, however previous investigation has confirmed that GCaMP signals are highly correlated with action potentials^101,102. Future work using electrophysiological tip recordings can confirm if sweeteners similarly induce action potentials in sweet and bitter GRNs. While our analysis specifically focuses on sweet and bitter GRNs, many tastants exhibit combinatorial encoding mechanisms that involve the three other labellar GRN classes (“Water”, “Salty”, and “IR94e”)^{48,57,103,104}. Currently, it remains unknown if these additional populations contribute to sweetener detection and processing.

Notably, our results show that while high concentrations of sucralose strongly activate bitter GRNs, sweet GRN calcium responses remain relatively modest. This discrepancy may reflect distinct sucralose binding affinities for the receptors expressed in each GRN population. Alternatively, weak sweet responses could indicate active dampening of sweet GRN output through acute habituation¹⁰⁵ or by co-activated bitter GRNs. This suppression could arise from direct lateral connections between GRNs^106–108bitter-induced GABAergic feedback on sweet GRNs¹⁰⁹or even ephaptic coupling between neighboring GRNs that tunes sensitivity through field-based interactions¹¹⁰. Additionally, the odorant-binding protein, OBP49a, has been shown to bind bitter compounds and inhibit sweet GRN activity¹¹¹suggesting that indirect molecular mechanisms may also modulate sweet GRNs during concurrent bitter activation. Potential suppressive crosstalk between bitter and sweet GRNs is also consistent with our behavioral observation that high concentration sucralose reduces subsequent feeding responses. We found that the aversive signal elicited by high concentration sucralose modulates responses to the same, repeated stimulus (Fig. 3B), but not to a separate appetitive stimuli like sucrose (Fig. S3A). This nuance indicates that the aversive signal elicited by sucralose specifically regulates feeding upon recognition of the same tastant. Together, these results imply that bitter GRN co-activation not only shapes immediate sweetener encoding in the periphery, but also generates lasting influences on feeding behavior.

Artificial sweeteners differ in their behavioral valence and impact of bitter signaling

Many of the previous Drosophila investigations of sweeteners have focused on how chronic intake affects several phenotypes, including feeding^{32,33,38–41}. However, these studies often do not quantify baseline interest in the sweeteners or assume that they elicit only attractive/sweet sensory signals. Additionally, sensory-induced physiological or metabolic changes also impact how dietary manipulations affect the animal¹¹²but these interactions are often not addressed in this current area of research. Here, we characterized and compared the behavioral valencies of several artificial sweeteners. The limited attraction we observed towards many of the chemicals in our labellar PER experiments is consistent with previous tarsal PER data suggesting similarly mild response levels⁴². We found positive preferences for sucralose and Reb A, which also agree with this previous investigation, but our results differ with aspartame. They observed a significant preference for 2 mM, but since aspartame in this range did not activate sweet GRNs on the labellum, it is currently unclear what would mediate this attraction. Moreover, the lack of aspartame feeding preference we observed agrees with our calcium imaging results and previous mouse data⁹⁸further indicating that the palatability of this sweetener may be specific to humans.

Although bitter GRNs respond to sucralose, aspartame, and Reb A, our silencing experiments indicate that bitter cell activation only significantly modulates sucralose feeding. This result provides further evidence that sucralose evokes a stronger behavioral valence compared to the other sweeteners, and that bitter co-activation significantly modulates sucralose feeding. Additionally, our behavioral evidence indicates that flies respond appetitively to high-concentration sucralose upon first encounter, but the accompanying bitter co-activation generates an aversive aftertaste signal that suppresses responses to subsequent stimuli, consistent with behavioral studies in mammals^26–28. Potential mechanisms underlying this phenomenon at the level of primary GRNs are described above, but it is also worth emphasizing that we do not know how these opposing gustatory signals are processed downstream to ultimately modify feeding output. Evidence from human neuroimaging studies demonstrate that sucralose and sucrose activate similar primary taste pathways, but sucralose produces significantly weaker responses in the insula and regions of the midbrain associated with feeding behavior¹¹³. Future analysis is required to determine how sucralose interacts with both sweet and bitter downstream circuitry.

In conclusion, our results indicate that flies exhibit diverse behavioral and cellular responses to various classes of artificial sweeteners. Future studies utilizing these sweeteners in Drosophila should anticipate sensory responses that are not exclusively sweet or appetitive. Moreover, future work with sucralose should expect concentration-dependent effects on feeding behavior. The chemosensory experience associated with sweetener consumption should be considered when evaluating the impact of diets containing artificial sweeteners on various phenotypes to better understand the unintended consequences of these chemicals.

Methods

Flies

Experimental flies were kept on regular cornmeal food at 25 °C in 60% relative humidity. Mated female flies between 2 and 10 days old were used in all experiments. The Canton-Special (CS) strain was used as the wild-type Drosophila melanogaster strain, and the following stocks were used to generate the genotypes indicated in each figure: Gr64f-Gal4¹¹⁴, Gr66a-Gal4⁸⁴, UAS-Kir2.1⁷⁶, UAS-GtACR1 (BDSC_92983), UAS-GCaMP6f (BDSC_52869).

Chemicals

All-trans-retinal (ATR) (Sigma R2500) was made up in 100% EtOH (Pharmco 111000200), kept at -20 °C, and diluted to a final concentration of 1 mM with EtOH of the same dilution given as a vehicle. Other test solution stocks were made as follows: sucrose (Sigma S7903): 1 M, sucralose (Sigma 69293): 0.1, 1, 10,100, 500 mM; Saccharin (Sigma 26047): 0.001, 0.01, 0.1, 1, 5, 10 mM; Cyclamate (Sigma C9131): 0, 1, 10, 50, 100, 200 mM; Aspartame (Sigma W700655): 0.001, 0.01, 0.1, 1, 5, 10 mM; Neotame (Interchim DW615): 0.001, 0.01, 0.1, 1, 5, 10 mM; Rebaudioside A (Sigma 01432): 0.001, 0.01, 0.1, 1, 5, 10 mM, Maltitol (Sigma M8892): 1, 10, 50, 100, 200 mM; Caffeine (Sigma C0750): 100mM.

Proboscis extension response (PER)

For the labellar PER experiments, flies were placed inside a pipette tip cut to size so that only the head was exposed. Flies were then sealed into the tube with tape and then adhered to a glass slide with double-sided tape. Flies were allowed 1 to 2 h to recover before testing began. Flies were stimulated with water on their labellum and allowed to drink until satiated. Each fly was then stimulated with increasing concentration of tastant and provided with water between each stimulus. For all prolonged taste modulation PER experiments, flies were stimulated with water on their labellum and allowed to drink until satiated. For the water control experiment (Fig. S2A), flies were stimulated water first, then stimulated with water a second time after 20 s–5 min. For all other experiments, flies were given two stimuli as indicated and then given a third stimulus after 20 s of 5 min, similar to previously published experiments⁵⁰. For all PER experiments, proboscis extension to each tastant was recorded as a positive (100) or negative response (0). For all experiments, flies received a final stimulation of 1 M sucrose as a positive control. Any flies that failed to respond to the 1 M sucrose were excluded.

Optogenetic PER

Flies were collected and put on standard cornmeal food supplemented with 10 µL of either all-trans-retinal (ATR) or vehicle for two days. Flies were then transferred to food deprivation vials containing 1% agar (Sigma A1296) with either ATR or vehicle for one day prior to experiment. All vials were kept at 25 °C and covered with foil to reduce light exposure. Flies were mounted in the same manner as labellar PER, but in a low luminosity environment. Flies were stimulated with water and allowed to drink until satiated. For all experiments, stimulations were combined with a green light LED (3.5 mW) held just above the labellum. For the sweet GRN silencing, flies were first given water (negative control), followed by 50 mM sucralose + green light. For the bitter GRN silencing prolonged taste modulation PER experiments, the green light was turned on prior to the initial stimulation of 500 mM sucralose, and 20 s later, flies were provided a second 500 mM sucralose stimulation (Fig. 3C). The same protocol was used with 0.1 mM and 10 mM aspartame (Fig. S4C); and with 0.1 mM and 10 mM Reb A (Fig. S5C). For all experiments, flies received a final stimulation of 1 M sucrose as a positive control. Any flies that responded to the water negative control or failed to respond to the 1 M sucrose were excluded.

Calcium imaging

In vivo imaging of labellar GRN axon terminals was performed as previously described⁴⁸. First, mated female flies were CO₂ anesthetized and mounted into a specialized chamber in which their heads are secured with nail polish. Once secured, the fly’s labellum was manually extended and waxed into this position to ensure that the proboscis was unobstructed. Flies recovered in a humidity chamber for one hour. After recovery, cuticle above the sub-esophageal zone (the imaging area) was dissected to expose the brain, and Adult Hemolymph-Like (AHL) solution (108 mM NaCl, 5 mM KCl, 4 mM NaHCO₃, 1 mM NaH₂PO₄, 5 mM HEPES, 15 mM ribose, 2 mM Ca²⁺, 8.2 mM Mg²⁺, pH 7.5) was continuously applied to the brain. The imaging area was cleared of obstructions by cutting and removing the respiratory tissues and part of the esophagus. Flies were imaged with a 3i Spinning disc Confocal station (Zeiss upright microscope, 2Kx2K 40 fps sCMOS camera, CSU-W1 T1 50 mm spinning disc). During image acquisition with a 40x water immersion objective, AHL was used as the immersive solution. Once positioned for imaging, baseline fluorescence was recorded for 5 s before the fly’s labellum was stimulated with the tastant for 5 s, and the recording continued for a total of 17 s. The tastant was directly applied to the labellum via a micromanipulator and a capillary tube which fit directly over the fly’s labellum. Each fly was exposed to water (negative control), increasing concentrations of indicated sweeteners, and a positive control (1 M sucrose for Gr64f GRNs, 100 mM caffeine for Gr66a GRNs). The camera exposure was set to 100 ms and 170 frames were captured at a rate of 9.5 Hz in Slidebook (SCR_014300). Videos were opened in FIJI (SCR_002285, SCR_003070) where a ROI was drawn around the projections and the fluorescence signal over time extracted. Ten consecutive points were chosen as a baseline and ΔF/F was calculated as a z-score ((F – mean baseline F)/standard deviation baseline F) for each timepoint. The three consecutive highest points during the 5 s stimulation were averaged as the peak ON response. Flies with no visible response to the positive controls were excluded.

Fly liquid interaction counter (FLIC)

FLIC experiments were performed as previously described^66,68. For optogenetic FLIC assays, flies were collected and put on standard cornmeal food supplemented with either all-trans-retinal or vehicle for two days. Flies were then transferred to food deprivation vials containing 1% agar with either all-trans-retinal or vehicle for one day prior to experiment. All vials were kept at 25 °C and covered with foil to reduce light exposure. Flies were then individually transferred into behavioral feeding chambers via mouth pipette. Within each chamber, flies had access to two food sources (2-choice FLIC assay). Both options were connected to distinct capacitance sensors on Drosophila Feeding Monitors (DFMs) (Sable Systems) that recorded the number of interactions between the fly and each respective food source. Interaction with one side activated a green-light LED within that chamber (FLIC opto-lid, Sable systems), while the other side did not produce any light activation. For chronic silencing assays, flies were transferred to 1% agar food deprivation vials for 1 day prior to experiment and kept at 25 °C. Flies were then loaded into similar 2-choice feeding chambers with standard FLIC lids (Sable Systems) that lacked LED activation.

For both assays, interactions between a fly and each food source were recorded for 3 h. A custom R script (SCR_000432), based off the Pletcher Lab, was used to analyze this raw output (full code on GitHub http://github.com/MStanleyLab/FLIC_code). Any interactions detected during the loading process were removed to avoid potential artifacts caused by loading flies into the chambers. Data were analyzed similarly to previous publications to quantify preference index, feeding interactions, feeding events, and feeding event duration^66,68. Measurements were taken every 200 ms, minimum signal threshold was set to 10, and feeding threshold was set to 20. For the optogenetic assays, green-light LED activation threshold was also set to 20 and the green light had an intensity of 1.5 mW. An interaction was recorded if the signal exceeded the feeding threshold. Preference index was calculated for each fly using: ((interactions on side A – interactions on side B) / total interactions). A feeding event was defined as 10 consecutive interactions above the minimum threshold, as long as one of these readings reached the feeding threshold. Sequential periods of continuous interactions were combined into one single event if the gaps of inactivity between them lasted less than 1 s. Feeding event duration was measured in seconds. To account for potential baseline differences between feeding chambers, flies of a particular condition or genotype were varied by position on the DFMs. For our 2-choice sucralose concentration series experiments (Fig. 1I, S1A), concentrations were run in parallel. Data from chambers that exhibited disrupted signal detection, including 0 values or exceedingly high values (> 1000 discrepancy between raw value and filtered value), were removed. Flies that failed to significantly interact with the food (< 15 interactions) were also excluded. Lastly, we applied a ROUT outlier test for each feeding metric (preference index, interactions, events, and event duration) to remove any other significant outliers.

Quantification and statistical analysis

All statistical tests were performed in GraphPad Prism 10 software (SCR_002798). Specific tests are indicated in the figure legends along with the number of replicates, which were generally chosen based on variance and effect sizes in accordance with previously published literature using the same assays. Experimental or genotype controls were always run in parallel. Behavioral assays were repeated over multiple days and genetic crosses. Data are plotted as mean +/- SEM in bar graphs and line graphs. Asterisks indicate *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Pierre-Yves Musso and Molly Stanley conceptualized, supervised, and funded the project. Christian Arntsen wrote the original manuscript, generated all figures, and performed statistical analyses. Christian Arntsen, Pierre-Yves Musso, Isabelle Chauvel, Stéphane Fraichard, Stéphane Dupas, and Jérôme Cortot performed the PER experiments. Christian Arntsen, Jake Grenon, and Kayla Audette performed the FLIC experiments. Molly Stanley and Kayla Audette performed the calcium imaging experiments. Jake Grenon, Isabelle Chauvel, Stéphane Fraichard, Stéphane Dupas, Jérôme Cortot, Kayla Audette, Pierre-Yves Musso and Molly Stanley contributed to reviewing and revising the original manuscript.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary files. Data files can also be found in this Mendeley Dataset: doi: 10.17632/vr53j5n9tf.1.

Declarations

Competing interests

The authors declare no competing interests.