Abstract
Cocaine is an addictive psychostimulant, and the risk of developing cocaine use disorder (CUD) is highly heritable. Little is known about the specific genes and mechanisms that lead to the development of CUD, and there are currently no FDA-approved pharmacotherapies that can treat it. Drosophila has proven an effective model organism to identify genes and mechanisms underlying addiction, especially alcohol use disorder. While flies exposed to cocaine display features of acute intoxication like those observed in mammals, including hyperactivity and reduced sleep, to date, there is no model of preferential cocaine self-administration in flies. Here, we assayed cocaine consumption in Drosophila males, as well as preference in a two-choice paradigm. We also investigated mechanisms involved in cocaine taste sensing using genetic and imaging tools. We show that cocaine is innately aversive to flies and that this avoidance depends on bitter sensing. Gustatory sensory neurons expressing the Gr66a bitter receptor are activated upon exposure to cocaine. Silencing of these bitter-sensing neurons or mutation of Gr66a reduces cocaine avoidance. In a longitudinal choice assay, these flies develop preference for cocaine-containing solutions within 12–18 h, whereas control flies do not. Our findings show that bitter sensation protects flies from developing cocaine self-administration preference. Conversely, silencing bitter perception enables us to use Drosophila as a model for experience-dependent cocaine self-administration preference. This opens the door to testing human variants associated with CUD for their causative role in cocaine self-administration in this highly tractable model organism.
Significance Statement
Cocaine use disorder (CUD) is a highly heritable condition for which there are no effective treatments. Testing the many human genetic variants linked to CUD requires a cost-effective, genetically tractable model. Here, we show that bitter-sensing neurons prevent cocaine self-administration in Drosophila. Furthermore, we demonstrate that disrupting Drosophila bitter perception enables a model for experience-dependent cocaine preference. Our findings underscore the potential of Drosophila as a crucial tool for identifying the genetic mechanisms underlying CUD, aiding in the discovery of new therapeutic targets, and contributing to the development of effective treatments for this highly heritable disease.
Introduction
Cocaine is a psychostimulant with a high addictive potential and a greater risk of producing dependence than any other psychostimulant drug (Nutt et al., 2010), and it is known to increase dopamine in areas of the human brain associated with reward (Volkow et al., 1999). Cocaine is used by >20 million individuals globally, and it is estimated that >24,000 people in the USA died from an overdose death involving cocaine in 2021 (CDC WONDER Database). However, currently, there are no FDA-approved pharmacotherapies to treat cocaine use disorder (CUD). Family studies show that CUD has an estimated genetic heritability of ∼72% (Goldman et al., 2005), with genetics influencing both initiation of cocaine use (Rhee et al., 2003) and addiction risk (Kendler et al., 2000). Despite this strong genetic component, little is known about the specific genes and molecular mechanisms that lead to the development of CUD. Recent genome- and transcriptome-wide association studies have identified hundreds of putative risk genes for the development of CUD (Cabana-Domínguez et al., 2019; Huggett and Stallings, 2020a,b), and they suggest that addictive disorders are largely polygenic. However, most of the associated genes have not been investigated regarding their causality for CUD-relevant phenotypes.
In contrast to research in humans, animal models of addiction provide the possibility to test the role of specific genes and genetic manipulations in the development of substance use disorder. In rodent models of addiction, escalation of drug-taking behavior and development of drug preference are typically studied in experiments that involve self-administration (Deroche et al., 1999). The behavioral response to cocaine in rodent assays of choice includes features of CUD observed in humans, where cocaine consumption and preference increase over time (Garcia et al., 2020). Given the polygenic nature of CUD as well as the generation time of rodents, efficient investigation of multiple candidate genes and the ability to probe interactions between such genes requires a model organism with face-valid responses to cocaine and a good economy of scale. Drosophila is an excellent genetic and molecular experimental model organism to study behavior, with ∼75% of disease-related human genes having functional orthologs in flies (Reiter et al., 2001). Drosophila has proven an effective model for studying complex addictive disorders such as alcoholism (Chvilicek et al., 2020), and flies express a range of alcohol-induced behaviors similar to those observed in intoxicated humans (Singh and Heberlein, 2000). Many genes identified based on their involvement in Drosophila alcohol responses have helped uncover corresponding genes involved in alcohol use disorder in humans (Grotewiel and Bettinger, 2015; Ojelade et al., 2015; Jia et al., 2016; Gonzalez et al., 2018; Evangelou et al., 2019; Lathen et al., 2020).
Research with other drugs of abuse, including psychostimulants, has been lagging, one reason being the lack of a reliable self-administration assay that would allow investigations into the development of self-administration preference. Recent papers have shown preferential self-administration of methamphetamine (Kanno et al., 2021; Rigo et al., 2021) and amphetamine (Belovich et al., 2019) in flies, and -omics approaches have identified many conserved genes in psychostimulant responses and self-administration (Highfill et al., 2019; Baker et al., 2021a,b). Flies exposed to cocaine display features of intoxication like those observed in mammals, including hyperactivity and reduced sleep (McClung and Hirsh, 1998; Lebestky et al., 2009). To date, however, there is no model of preferential cocaine self-administration in flies. Here, we show that cocaine is innately aversive to flies and that its avoidance depends on bitter sensing. Gustatory sensory neurons expressing the Gr66a bitter receptor are activated upon exposure to cocaine. Furthermore, flies with these bitter-sensing neurons silenced, as well as Gr66a mutant flies, show significantly reduced naive cocaine avoidance and develop experience-dependent preference for the drug, unlike controls. Thus, bitter sensation protects flies from developing cocaine preference, while silencing bitter perception enables us to use Drosophila as a model organism to investigate the genetics of preferential cocaine self-administration.
Materials and Methods
Drosophila stocks
Behavioral experiments were performed with male flies raised in 12 h light/dark conditions at 25°C with 70% humidity on standard cornmeal/molasses food. Flies used in behavioral experiments were 3–6 d of age at the start of the experiments. Male w* Berlin flies were chosen as a genetic wild-type strain for all preliminary experiments of consumption and preference. For assays requiring food deprivation, flies were kept in vials containing 0.7% water agar. Apart from the wild-type w* Berlin (Rothenfluh et al., 2006) flies, the rest of the fly stocks were obtained from Bloomington Drosophila Stock Center: Gr66a-Gal4 (RRID:BDSC_28801), UAS-Kir2.1 (RRID:BDSC_6596), Gr66a mutant (RRID:BDSC_28804), Gr66a Rescue (RRID:BDSC_35528), UAS-EGFP (RRID:BDSC_5431), and UAS-GCaMP6s (RRID:BDSC_42746).
Blue dye assay
Male flies, 3–6 d old, were collected under CO2 anesthesia and allowed to recover for 24 h in vials with standard cornmeal food. After recovery, flies were wet starved for the indicated period in vials with 0.7% water agar (Fisher Bioreagents, catalog #BP2641-500 CAS:9002-18-0). Following starvation, flies were transferred to feeding vials with filter paper strips (Kimtech, catalog #34155; 7 × 1.75 cm) coated in 350 μl of the indicated feeding solution. All feeding solutions contained 0.3% (v/v) blue food dye (FD&C Blue#1 Spectrum, catalog #FD110; CAS: 3844-45-9) and the indicated amounts of sucrose (Fisher Bioreagents, catalog #BP220-1 CAS:57-50-1), cocaine (NIDA, catalog #9041-001 CAS:53-21-4), lobeline (Alfa Aesar, catalog #A17326 CAS:134-64-5), ʟ-canavanine (Sigma, catalog #C9758 CAS:2219-31-0), or denatonium (Aldrich, catalog #D5765-5G CAS:3734-33-6). After a 4 min feeding assay, flies were transferred to empty vials and frozen at −80°C. Flies that ate were identified under a stereomicroscope based on the presence of blue dye and transferred to Eppendorf tubes in groups of five. Then, 50 μl of water was added to each tube, and flies were homogenized with a small battery-powered pestle. Homogenates were spun down for 5 min at 14,000 rpm. Centrifugation results in a tissue pellet, leaving a supernatant of blue-dyed solution under a thin lipid layer. A NanoDrop 2000 spectrophotometer (Thermo Scientific) was used to measure absorbance on 2 μl samples of the dyed supernatant, with special care being taken to avoid the transfer of the topmost lipid layer. Absorbance was measured at 630 and 700 nm to quantify consumption. The volume (nl) consumed by each fly was calculated using the formula (OD 630 nm − 1.1 * OD 700 nm) * CF, where CF is an empirically determined conversion factor specific to the 3% Blue#1 stock solution.
Fluorometric reading assay of preference (FRAP)
Naive cocaine preference was tested in a 30 min two-choice fluorometric plate-reading assay described previously (Peru y Colón de Portugal et al., 2014), with some modifications. Male flies were collected in groups of 35 under CO2 anesthesia and allowed to recover for 24 h in vials with standard cornmeal food. After recovery, flies were transferred to starvation vials containing 0.7% water agar and food deprived for the indicated amount of time. Next, flies were gently transferred to 60-well plates using a mouth pipette. After loading the flies, the holes in the lid of each plate were covered with tape, the plates were turned upside down, and the flies were allowed to feed for 30 min. The 60-well plates (MilliporeSigma, catalog #M0815-100EA) were prepared to offer a choice between two equivalent sucrose solutions (340 or 100 mM), where one solution was supplemented with cocaine, lobeline, or ʟ-canavanine. The feeding solutions in each plate were labeled with distinct fluorescent dyes, 0.005% rhodamine B (Acros Organics, catalog #132311000 CAS:81-88-9) and 0.003% fluorescein (Sigma-Aldrich, catalog #F6377-100G CAS: 518-47-8) to determine relative consumption. To control for color bias, we ran two plates for each pair of feeding solutions, where the solutions in each plate were reciprocally dyed. Then, 100 μl of 0.005% rhodamine B sucrose solution and 100 μl of 0.003% fluorescein sucrose solution were mixed to obtain a standard 50/50 dye solution and added to 20 wells of a 60-well plate (10 μl per well). For all other plates, 100 μl of 0.005% rhodamine B liquid sucrose solution and 100 μl of 0.003% fluorescein liquid sucrose solution were distributed between 20 wells in a staggered, offset pattern. After the assay, the plates were transferred to the −80°C freezer to flash freeze flies. We used a 384-well Greiner plate (Greiner Bio-One, catalog #781201), suspending three flies in each well with 60 μl of water (8 wells per plate). Next, we loaded the plates into a Fluoroskan Ascent microplate fluorometer. We measured the relative emission intensity of each dye by scanning with filter pairs at 485/527 and 542/591, which correspond to fluorescein and rhodamine B, respectively. These data were used to generate a ratiometric index of consumption representing a relative preference for each solution.
FLIC
Fly Liquid Food Interaction Counter (FLIC) assays were performed as previously described (Ro et al., 2014), with some modifications. Male flies, 3–6 d old, were collected with CO2 anesthesia at least 24 h before each experiment. Flies were food deprived for the indicated time in vials with 0.7% agar to prevent dehydration. After starvation, we used a mouth pipette to transfer 10 flies to each arena in the FLIC for a 30 min assay of preference. Flies in each arena had free access to two wells, allowing a choice between two different feeding solutions. The number and duration of leg events (LE) and proboscis events (PE) for each well in the Drosophila feeding monitor were recorded and analyzed as described previously using the FLIC (Ro et al., 2014). PE and LE were defined as having a peak signal amplitude greater than or less than 100 AU, respectively (Ro et al., 2014; Mishra et al., 2021). Raw data from each experiment were processed to analyze proboscis preference, event preference, and percent of proboscis events. As proboscis interactions indicate feeding events, the cumulative length of proboscis events at each feeding well provides an indirect measure of feeding activity. The proboscis preference index represents the relative interaction time between two wells and indirectly measures feeding preference. Proboscis preference was calculated as the difference in interaction time between two wells divided by the total interaction time. The results of this analysis give a value between +1 and −1, with +1 representing absolute preference, −1 representing total avoidance, and 0 indicating indifference. Event preference was calculated using the same formula but is based on the count of events rather than the length of events. The proportion of proboscis events to total events was used to determine acceptance for each well. Flies were not food deprived before FLIC experiments lasting longer than 30 min (Fig. 4), and a liquid reservoir (Corning Flask reservoir, MilliporeSigma, catalog #CLS430639) was attached to the FLIC platform to ensure the solutions did not evaporate during the assay.
Calcium imaging gustatory neuron responses in Drosophila legs
Calcium imaging taste response in Drosophila legs was performed as previously described (Miyamoto et al., 2013), with some modifications. Flies were anesthetized with CO2, and forceps were used to remove the forelegs by cutting between the femur and tibia. Double-sided tape was used to fix the tibia to a slide, and another piece of tape was used to secure the leg, leaving the distal fourth and fifth tarsal segments exposed over the glass. The exposed tarsal segments were covered with water, and the slide was transferred to the microscope for imaging. The legs were imaged for 30–60 s to establish baseline fluorescence and then treated with the indicated solutions. Imaging was performed with an Olympus CKX53 inverted microscope using a 40× objective and an X-Cite 120q fluorescence illumination light source. The preparation was excited with a mercury arc lamp using the CKX3-B excitation fluorescence mirror unit with an excitation filter (BP460-490C), dichroic mirror (LP500), and emission filter (520IF). Image acquisition was performed using an Olympus DP72 microscope digital camera with DP Manager software (cellSens). Acquisition settings for representative images were optimized to enhance structural aspects of tarsal segments and support visualization of Gr66a-positive neurons. For calcium imaging, images were acquired at 2 fps with an exposure time of 500 ms and at the minimum setting for light sensitivity (ISO 200).
Image processing and analysis
Image sequences for each leg (240 frames) were processed using ImageJ. Images were cropped to include the fourth and fifth tarsal segments, and X–Y drift was corrected with TurboReg based on an average intensity Z-projection. ROIs were identified by analyzing particles in a standard deviation Z-projection and applying the triangle autothreshold method. Background fluorescence was measured within ellipses adjacent to taste cell ROIs. Intensity measurements for each frame were used to plot changes in intensity over time. Twenty frames (10 s) were analyzed to establish a baseline ΔF/F. Solutions were applied between frames 20–30 (15–20 s), and responses were analyzed across the last 200 frames (100 s). The change in fluorescence intensity (ΔF) relative to baseline fluorescence intensity (F) was used to plot ΔF/F for each acquisition series. Untreated cells were imaged to measure photobleaching. Results were fit to an exponential decay function and applied to correct signal intensity across each treatment condition.
Data analysis and statistics
Statistical analyses were performed using GraphPad Prism 9. Data from most feeding experiments were not normally distributed; therefore, preference and consumption experiments were analyzed using nonparametric statistics. The two-tailed Mann–Whitney U test was used to determine the significance of pairwise comparisons in the blue dye feeding assay of consumption, the FRAP, and FLIC. The Wilcoxon signed rank test with a two-tailed p value was used in the FRAP and FLIC to compare the median preference score of each group to a hypothetical median of zero. Box plots in all consumption and preference experiments represent median values and interquartile range, with whiskers depicting 10th and 90th percentiles.
Results
Cocaine causes reduced food ingestion and avoidance
While it has been known for some time that acute cocaine exposure has hyperactivating effects on Drosophila (McClung and Hirsh, 1998; Bainton et al., 2000), few studies have investigated cocaine self-administration. To determine whether Drosophila would consume cocaine, flies were (mildly) food deprived for 6 h to encourage eating, and consumption of 340 mM sucrose with blue dye and varying concentrations of cocaine was measured (Fig. 1A). We started with mM cocaine concentrations, based on Drosophila literature with cocaine (McClung and Hirsh, 1998; Andretic et al., 1999; Bainton et al., 2000; Chang et al., 2006; Filošević et al., 2018) and also amphetamine (Vanderwerf et al., 2015; Rigo et al., 2021).
Bitter taste perception contributes to the avoidance of cocaine consumption in Drosophila. A, A schematic of the experimental timeline for the blue dye feeding assay. Flies were food deprived for 6 h and then offered sucrose (340 mM) or sucrose (340 mM) with cocaine in a 4 min feeding. Relative consumption was calculated for flies that ate by expressing intake (nl) in each condition as a proportion of the average consumption by flies offered plain sucrose solution in parallel. B, Cocaine significantly reduces consumption (H(5) = 73.35, p = 2.05 × 10−14 Kruskal–Wallis test) for control flies. Multiple comparison-adjusted p values shown versus 0 mM cocaine (Dunn's correction). Here, and in the following Figures *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns, not significant. Data are shown as median with quartile boxes and Min-Max whiskers. C, Experimental timeline for the FRAP in D. Flies were food deprived for 6 h and then given access to sucrose (340 mM) and sucrose (340 mM) with cocaine or a bitter control compound (10 mM ʟ-canavanine) during a 30 min assay of choice. D, Cocaine dose has a significant effect on preference in control flies (H(3) = 34.12, p = 1.869 × 10−7 Kruskal–Wallis test). The significance of comparisons between measured PI (preference index) and indifference (a hypothetical PI = 0) are indicated below each column (Wilcoxon signed rank test with Bonferroni adjustment). E, Experimental timeline for F. Preference toward different cocaine concentrations in 340 mM sucrose measured in the FRAP in flies with bitter sensory neurons silenced (Gr66a > Kir2.1) and controls (UAS-Kir2.1). A significant difference in PI between genotypes is indicated (Mann–Whitney test). ʟ-Canavanine (ʟ-can) was used as an known bitter control. G, A timeline for a FRAP experiment (H) using flies with a deletion of Gr66a (Gr66a Deletion) at the endogenous locus and matched controls that express functional Gr66a (Gr66a Rescue). H, Lobeline was used as a known bitter control, and significant differences in PI between genotypes are indicated. All exact p values and detailed statistics are described in Extended Data Table 1.
Flies showed a dose-dependent decrease in food consumption, with virtually no consumption of 15 mM cocaine-supplemented sucrose (Fig. 1B). The flies that ingested 10 mM cocaine showed apparent incapacitation, even seizures, while 3 mM cocaine often led to pronounced grooming behavior, loss of righting, and loss of negative geotaxis. These effects of intoxication are similar to behaviors associated with cocaine overdose or chronic cocaine abuse in humans, which include hyperkinesia, seizures, and loss of coordination (Cardoso and Jankovic, 1993; Lipton et al., 2000; Narula et al., 2017). The effects of acute high-dose cocaine exposure have been already reported (McClung and Hirsh, 1998; Bainton et al., 2000). We were interested in establishing a model that mimics the initial stages of substance use disorder, in which the use of the drug has positive valence and therefore causes an escalation in consumption. Ingestion of lower concentrations (see below) did not cause overt behavioral phenotypes.
Because a decrease in consumption was already apparent when sucrose was supplemented with 0.5 mM cocaine (Fig. 1B), we decided to test consumption preference at these lower cocaine concentrations in a 30 min 2-choice assay (FRAP, for fluorescence reading assay of preference; Fig. 1C; Peru y Colón de Portugal et al., 2014). In the FRAP, 6 h food-deprived flies showed significant dose-dependent avoidance of cocaine as low as 0.1 mM (Fig. 1D).
Flies avoid consuming cocaine via bitter sensation
Cocaine is an alkaloid phytotoxin produced by the coca plant. Since flies perceive most alkaloids as bitter and, therefore, avoid them (Dweck and Carlson, 2020), we asked whether bitter sensing was involved in cocaine avoidance. We first reduced bitter taste sensation in flies using the Gal4/UAS system (Brand and Perrimon, 1993) by expressing the inwardly rectifying potassium channel Kir2.1 (Baines et al., 2001) to electrically silence neurons using Gr66a-Gal4, a driver that is specifically expressed in bitter-sensing gustatory neurons (Weiss et al., 2011). We used the FRAP to determine preference between sucrose alone versus sucrose supplemented with different cocaine concentrations, or a known bitter compound, ʟ-canavanine, as a positive control (Fig. 1E; Lee et al., 2012). Silencing bitter gustatory neurons (Gr66a > Kir2.1) significantly decreased (but did not abrogate) avoidance of ʟ-canavanine, as well as of cocaine (Fig. 1F), suggesting that flies sense cocaine as a bitter tastant.
To determine whether the avoidance driven by bitter-sensing Gr66a-positive neurons was mediated by the G-protein coupled Gr66a receptor complex itself, as opposed to other receptors, e.g., ionotropic receptors (Moon et al., 2006), we examined cocaine preference in flies containing a deletion of Gr66a and its genetic rescue control (Moon et al., 2006). Because avoidance of 1.5 mM cocaine was still strong in controls, we further diluted the cocaine in our feeding solutions, and wet-starved flies for 18 h, which increases the flies’ motivation to eat and decreases their bitter perception (LeDue et al., 2016; Mishra et al., 2021; Fig. 1G). To ensure flies still sampled both solutions and did not gorge themselves on the first solution encountered, we also lowered the sucrose concentration to 100 mM. We observed a significant reduction in avoidance of cocaine in Gr66a Deletion flies versus Gr66a Rescue controls (Fig. 1H). Avoidance of lobeline, a bitter phytotoxin detected by Gr66a (Sung et al., 2017) and positive control, was also reduced in Gr66a Deletion flies. These results show that G-coupled protein receptor complexes that include Gr66a contribute to the detection of cocaine as an aversive bitter tastant in Drosophila.
Cocaine reduces feeding initiation mediated by peripheral leg sensing
Flies integrate sensory signals from multiple taste organs when deciding whether to initiate feeding (Freeman and Dahanukar, 2015). Gr66a-positive neurons are found in the pharynx, labial palp—part of the consummatory organ, the proboscis—and in the flies’ two forelegs (Dunipace et al., 2001). In the fly liquid-food interaction counter (FLIC; Ro et al., 2014), interactions with liquid food via the leg or proboscis close an electrical circuit, allowing quantification of the duration of the interaction (Fig. 2A). The signal's amplitude differentiates between leg interactions, which are tasting events, and proboscis interactions, which are consummatory events with a larger amplitude. The proboscis interaction duration is proportional to the amount of food consumed (Itskov et al., 2014; Ro et al., 2014). The fraction of time spent interacting with cocaine + sucrose versus plain sucrose reflects consumption preference, and flies avoided cocaine in the FLIC, even at concentrations as low as 50 µM (Fig. 2B). Gr66a Deletion flies showed reduced cocaine avoidance in the FLIC compared with Gr66a Rescue flies (Fig. 2C), similar to our result in the FRAP fluorescence ingestion assay (Fig. 1H).
Cocaine reduces feeding initiation mediated by peripheral leg sensing. A, Schematic of the experimental timeline with a representation of a single FLIC arena containing ten flies with access to two feeding wells. One well in each arena contains plain sucrose (well a), while the opposite contains sucrose with cocaine (or bitter control, well b). The magnified section shows two of the 10 flies interacting with the solution in well a. The fly on the left side tastes the sucrose with their leg, while the fly on the right feeds with their proboscis, events that can be differentiated and that are recorded in their duration. B, C, Proboscis event duration preference index scores (proboscis interaction, PI) are shown for w* Berlin controls (B) and Gr66a Deletion and Rescue flies (C). D–F, The fraction of proboscis events (compared with total proboscis + leg events) at each well—a measure of food palatability—are used to estimate variability in attraction between pairs of wells in each condition. ʟ-Canavanine is used as a known bitter control, and significant differences in the fraction of proboscis events between pairs of wells are indicated (Mann–Whitney test) for w* Berlin controls (D) and Gr66a Deletion (E) and Rescue flies (F).
A tasting event with the leg precedes each consummatory proboscis event. Some leg events do not lead to consumption but to food rejection. The fraction of proboscis events to total (proboscis + leg) events thus represents a measure of feeding initiation, or food palatability. A fraction of 0 indicates complete rejection of the food. When we food-deprived flies for 18 h and then offered them a choice between 100 mM sucrose supplemented with different cocaine concentrations and plain sucrose for 30 min in the FLIC, cocaine caused a significant reduction in food palatability, i.e., a decrease in the fraction of proboscis interaction events (Fig. 2D). This suggests that leg-mediated bitter sensing of cocaine reduced the flies’ willingness to consume the cocaine-supplemented sucrose. Indeed, unlike Gr66a Rescue flies (Fig. 2E), Gr66a Deletion flies did not show reduced feeding initiation on cocaine-supplemented sucrose compared with plain sucrose (Fig. 2F). In fact, the Gr66a Deletion flies showed increased initiation on 50 µM cocaine-supplemented sucrose, compared with plain sucrose. Our data suggest that Gr66a receptors in bitter-sensing gustatory leg neurons participate in mediating cocaine aversion.
Gr66a-positive leg neurons are activated by cocaine
Because our data suggested that Gr66a-positive neurons in the forelegs sense cocaine, we performed functional imaging of those neurons. We used enhanced green fluorescent protein (EGFP; Goentoro et al., 2006) to confirm that the Gr66a-Gal4 driver is expressed in two bilateral pairs of neurons in both the forelegs’ fourth and fifth tarsal segments, as described before (Dunipace et al., 2001). We observed the same foreleg neurons when Gr66a-Gal4 was used to express the genetically encoded calcium indicator GCaMP6s (Chen et al., 2013; Fig. 3A) and imaged their activity in response to sucrose, denatonium (a known bitter substance; Weiss et al., 2011), and cocaine (100 or 1,000 μM). While our behavioral data showed changes in preference at doses lower than 100 μM, we did not perform calcium imaging at those lower doses due to the limitations of our ex vivo experimental approach. In our preparation, we cannot dissect the neuronal tissue from the highly autofluorescent cuticle; therefore, the GFP response signal needs to be strong enough to pass that additional layer. While lower concentrations could increase the activity of those neurons, we would not be able to observe it in our preparation. As expected, no change in baseline fluorescence intensity was observed after sucrose application (since it is not a bitter substance detected by bitter gustatory sensing neurons), while fluorescence intensity increased when the forelegs were exposed to denatonium (Fig. 3B–E). The responses we recorded for denatonium are comparable with those published in a similar ex vivo preparation measuring activation of GCamP6s in taste neurons of the Drosophila pharynx (Chen et al., 2019). Importantly, we also observed a dose-dependent increase in GCaMP6s signal intensity in forelegs exposed to cocaine (Fig. 3B–E). This indicates that cocaine is directly sensed by peripheral gustatory neurons in Drosophila forelegs.
Gr66a-positive leg neurons are activated by cocaine. A, A representative image of GCaMP6s fluorescence from Gr66a-positive neurons in the forelegs of adult Drosophila males. B, GCaMP6s fluorescence signal recorded after exposing forelegs to 20 µM denatonium (red trace), 1,000 µM cocaine (teal trace), 100 µM cocaine (purple trace), and 100 mM sucrose (gray trace). The calibration bar to the left of the Y-axis represents the signal intensity range observed during acquisition. C, Heat map kymographs showing the change in signal intensity over time in recordings from individual neurons (individual lines within a treatment group). D, Average peak intensity after treatment (20–120 s). Mean and SEM are shown for neurons in each group. There was a significant effect of treatment (H(3) = 37.01, p = 4.578 × 10−8, Kruskal–Wallis, one-way ANOVA) and significant differences from sucrose. E, Representative images of the maximum change in fluorescence signal recorded in each treatment condition. Top row, Background fluorescence signal from GCamP6s-positive neurons. Two images from each treatment condition are displayed to show cell responses in the fourth and fifth tarsal segments. Middle row, Projections of maximum signal change from baseline before treatment. Bottom row, Projections of maximum signal change from baseline during treatment.
Reducing flies’ bitter perception of cocaine leads to experience-dependent cocaine consumption preference
All of our data above suggested that flies avoid cocaine, consistent with other published results showing cocaine avoidance (or indifference at lower concentrations) after 24 h in a 2-choice feeding assay (Kanno et al., 2021). We have previously shown that flies can develop ethanol consumption preference within 24 h in a 2-choice assay, starting from naive avoidance of 15% ethanol (Butts et al., 2019). We therefore performed longitudinal FLIC assays for 24 h. Since even at 50 µM, flies showed a strong aversion to cocaine (Fig. 2D), we tested longitudinal preference in Gr66a Deletion flies at 50, 20, and 5 μM, but these flies showed strong avoidance (preference < −0.5 after 24 h; data not shown). Consequently, we reduced the concentration of the drug in our feeding solutions even further. In control UAS-Kir2.1 flies, there were no significant differences in preference index from indifference, i.e., PI = 0, in the last 6 h of the assay (to allow time for preference to develop) at concentrations from 0.25 to 1 µM cocaine in 100 mM sucrose (Fig. 4A). In contrast, experimental Gr66a > Kir2.1 flies, with silenced bitter perception, showed significant preference for 0.5 µM cocaine (median preference = 0.56; Fig. 4B). We then examined the 24 h data in 6 h intervals. UAS-Kir2.1 control flies showed no effect of time and only significant preference at 12–18 h (Fig. 4C), while bitter-silenced Gr66a > Kir2.1 flies showed an effect of time and significant preference at 6–12, 12–18, and 18–24 h in the FLIC (Fig. 4D). As we did before (Butts et al., 2019), we examined the linear regression for the data (which was a better fit for the majority of the data than an asymptotic growth curve) and found that experimental, but not control, flies showed a significantly positive slope over time to 0.5 µM cocaine (Fig. 4E). Thus, increasing preference developed—sooner and more sustained—in flies lacking bitter perception of cocaine. To corroborate our finding of experience-dependent preference for cocaine self-administration, we used Gr66a Deletion and Gr66a Rescue flies at the same cocaine concentration. Neither genotype showed an effect of time (Rescue p = 0.47, while Deletion p = 0.07). However, unlike Gr66a Rescue flies, Gr66a Deletion flies did show significant preference at the end of the assay (18–24 h; Fig. 4F,G). Furthermore, the slope of the preference index was nominally significantly >0 for Gr66a Deletion flies, which, albeit did not carry over after Bonferroni’s correction (Fig. 4H). Together, these data suggest that bitter perception protects Drosophila from developing cocaine preference and self-administration.
Reducing flies’ bitter perception of cocaine leads to experience-dependent cocaine consumption preference. A, B, Dose–curve response of a 24 h FLIC assay showing preference indices in the last 6 h interval of the assay. Here, and in the following panels, stars at the bottom indicate whether PI is significantly different from 0 (Bonferroni adjusted). Control UAS-Kir2.1 (A) and experimental Gr66a > Kir2.1 on the right (B). C, D, Preference indices for every 6 h time interval in the 24 h FLIC assay. No significant effect of time in control flies (C), Friedman test, while the time effect was significant in experimental flies (D). E, slopes of linear regressions from C and D with 95% confidence interval. Bonferroni-adjusted significance level of the slope not being 0 at the bottom. F, G, Preference indices for every 6 h time interval in the 24 h FLIC assay for Gr66a Rescue and Deletion flies. No significant effect of time p = 0.47 in Rescue, and p = 0.07 in Deletion flies. No significant effect of time in Rescue flies (F), Friedman test, while the time effect was significant in Deletion flies (G). H, slopes of linear regressions from (F, G) with nominal p value shown for Deletion data.
Preference development is driven by learning cocaine’s location
To determine which assay parameter was driving the increasing proboscis interaction preference for sucrose + cocaine over time in experimental bitter perception-deficient flies, we further scrutinized the FLIC data from Figure 4. First, we analyzed the fraction of proboscis events over time in each genotype as a readout for the palatability of sucrose with or without cocaine (Fig. 5A), essentially asking whether flies acquired a literal taste for cocaine and would be more likely to transition from tasting with their legs to consumption with their proboscis. Three of four genotypes showed an effect of time and both bitter-blunted genotypes showed an interaction of time × solution. We therefore calculated the initiation preference, i.e., on which solution flies would rather initiate feeding over time. None of these initiation preferences showed an increase over time, and only few isolated preference points were significantly >0 (Fig. 5B). Thus, a (relative) increase in the palatability of sucrose + cocaine (over sucrose) did not drive the overall increasing consumption preference (Fig. 4).
Flies do not acquire taste preference for cocaine. A, Feeding initiation, reflecting food palatability/tastiness as measured by the fraction of proboscis events, on sucrose (left) and sucrose + cocaine (right). Significance summary for time effect (T), solution effect (S), and TxS interaction (I) is shown for each genotype (2-way ANOVA). Here, and in following panels and figures, the four data categories on the X-axis represent four successive 6 h intervals in the FLIC assay (data from Fig. 4C–H). B, Initiation preference (sucrose + cocaine top, sucrose only on bottom half of each graph). LR indicates the significance level of a linear regression and the Bonferroni-adjusted significance summary at the bottom indicates any deviations from indifference, where PI = 0. Initiation preference does not increase over time for any of the genotypes.
We next focused on individual proboscis events and first asked whether the number of proboscis interactions would specifically increase on sucrose + cocaine (over sucrose). While the total number of proboscis events showed an effect of time for every genotype, we observed no time × solution interaction (Fig. 6A), and the overall decrease in proboscis events may reflect ongoing satiety during the assay time. When we examined the average duration of individual proboscis events over time, we found significant effects of time (three of four genotypes; Fig. 6B), with increasing event duration over time. Linear regression slopes were positive for three of four genotypes on sucrose + cocaine and one genotype on sucrose-only (Fig. 6C). Pooling the data from all genotypes revealed a significant difference between the slopes on the two solutions (F = 4.453, DFn = 1, DFd = 349, p = 0.0355), but that difference was no longer significant when we analyzed bitter-blunted flies (F = 2.940, DFn = 1, DFd = 182, p = 0.0881) and controls (F = 2.612, DFn = 1, DFd = 187, p = 0.1078) separately. Thus, the duration of proboscis events on sucrose + cocaine increased over time, and while that seemed exacerbated in bitter-blunted flies (Fig. 6C, largest slopes), it was not significantly different from control flies. Therefore, cocaine feeding bout duration increased but did not drive the difference in overall consumption preference between bitter-blunted and control flies.
Duration of individual proboscis feeding events increase over time on sucrose + cocaine. A, Number of proboscis events over time. B, C, Average duration of individual proboscis events over time (B) with the linear regression slopes for both solutions (C). Significance summary below indicate deviation from no correlation, where slope = 0. The duration of proboscis events increases over time on sucrose + cocaine, but not with significant bitter-blunted specificity, thus not driving overall preference.
Lastly, we examined the total number of leg + proboscis events and found consistently significant effects of time (all four genotypes) and of solution (in three of four genotypes; Fig. 7A). In Gr66 > Kir2.1 we also found a highly significant interaction of time × solution (Fig. 7A). We therefore determined the total leg + proboscis event preference index, i.e., which well the flies would rather visit, and found significant increases over time in the preference to visit the sucrose + cocaine well in the two experimental bitter-blunted genotypes, but not their controls (Fig. 7B). This suggests that the main driving force for overall increased consumption preference was flies’ increasing preference to visit the well containing sucrose + cocaine, over the sucrose-only well. Taken together with unchanged feeding initiation (Fig. 5B), and a trend for increasing feeding bout duration on sucrose + cocaine (Fig. 6), more sucrose + cocaine visits are the main driving force to result in increasing self-administration preference for cocaine.
Increasing preference to visit the sucrose + cocaine well drives consumption preference. A, Total number of (leg + proboscis) events on each solution. B, Event preference, i.e., which well flies prefer to visit over time. Increasing preference to visit the cocaine + sucrose well is observed in bitter-blunted flies.
Discussion
Cocaine elevates dopamine and mood in humans (Volkow et al., 2009) and has strong addictive potential (Nestler, 2005), but there are currently no FDA-approved drugs to treat CUD. Given its large heritability and the polygenicity of its phenotypes, it is crucial to develop genetically tractable models of CUD to pave the path toward novel therapeutics and test candidate genes from human genome-wide association studies. Model organisms, like Drosophila, have proven relevant in understanding the genetics of other substance use disorders, notably alcohol use disorder (Lathen et al., 2020). While flies have been successfully used to identify genes that influence acute cocaine consumption (Highfill et al., 2019; Kanno et al., 2021; Baker et al., 2021b), there is currently no Drosophila model for preferential self-administration of cocaine. Our results explain why that is the case: cocaine activates bitter-sensing peripheral gustatory neurons, causing preingestive avoidance, and therefore, flies do not consume enough cocaine to develop preference. Our results show that flies directly sense cocaine by the legs, in neurons expressing the Gr66a coreceptor. Flies with a Gr66a mutation, or in which Gr66-expressing neurons are silenced, show a significant reduction in cocaine avoidance, suggesting that Gr66a-containing bitter receptor complexes detect cocaine. At this point, it remains to be determined which other G-coupled receptors might be part of the Gr66a-containing heterotrimeric complex. While cocaine also tastes bitter to humans, it is not known whether changes in taste receptors alter the risk of developing a CUD. Bitter perception does, however, modulate alcohol consumption in humans, and polymorphisms in genes encoding bitter receptor proteins have been associated with alcohol consumption (Duffy et al., 2004; Lanier et al., 2005; Hayes et al., 2011; Dotson et al., 2012), binge-like drinking behavior (Wang et al., 2007), and risk of developing alcohol use disorder (Edenberg and Foroud, 2006; Hinrichs et al., 2006). It is also noteworthy that two genes associated with cocaine use or dependence, respectively, encode putative olfactory receptors (Gelernter et al., 2014; Wu et al., 2019), suggesting that peripheral sensation of cocaine might influence the risk of cocaine taking in humans as well.
Both Gr66a mutants, as well as fly with Gr66a-expressing neurons silenced, showed a reduction in naive cocaine avoidance. However, both still avoided cocaine. One reason for this might be that there may be other gustatory receptors able to sense cocaine, especially at higher concentrations, that do not require Gr66a to form a functional complex. Furthermore, we detected weak activation of Gr66a-expressing leg neurons by 100 µM cocaine (Fig. 3), yet flies showed cocaine consumption avoidance even at 5 µM (data not shown). This would suggest sensitivity limitations of our assay, and/or the presence of additional labial or pharyngeal bitter sensing neurons with greater sensitivity than the leg ones (Chen et al., 2019; Devineni et al., 2021). An additional explanation for the avoidance of low-concentration cocaine could be that it causes an acute adverse physiological response when ingested that causes consummatory aversion. In humans, cocaine consumption does not only lead to euphoria and mental alertness, but it also causes nausea, high blood pressure, irregular heartbeat, and even anxiety and paranoia (Goldstein et al., 2009; Riezzo et al., 2012; Fonseca and Ferro, 2013). Adverse effects in flies, even if slow in onset, could be similarly detrimental to long-term health and could also explain why even bitter-silenced flies do not develop preference for 1 µM cocaine (Fig. 5A,B). What exactly these adverse responses might be, and whether they depend on cocaine-mediated inhibition of the dopamine transporter (Chen et al., 2005), remains to be determined. In flies, distinct dopamine circuits are involved in appetitive and aversive memory conditioning (Kaun and Rothenfluh, 2017). Acute, as well as learned, aversion could, therefore, be contributing reasons for the lack of preference development at higher concentrations.
Importantly, we did find that when flies’ bitter sensing is blunted, they are able to develop experience-dependent preference for cocaine. Flies with either a mutation in the Gr66a bitter receptor, or with silenced Gr66a-containing neurons, did not show naive preference or aversiveness toward cocaine at very low doses. However, within 12–18 h, those flies developed consumption preference toward cocaine. Our data suggested that flies do not “develop a taste” for cocaine, since the cocaine-containing solution does not gain increasing palatability (feeding initiation) over the sucrose-only solution with experience. Instead, flies learned to visit the cocaine-containing well more often than the sucrose-only well (Fig. 5F,G). As cocaine inhibits the dopamine transporter both in flies and mammals (Heberlein et al., 2009), a possible explanation for the behaviors we observe is that cocaine acts on the Drosophila dopaminergic PAM neurons, which are known to be required for sugar-mediated behavioral reinforcement learning (Burke et al., 2012; Liu et al., 2012). While the final dose that led to increasing consumption preference was significantly lower than the concentrations we started out with, based on Figure 1, the concentration that would lead to a dose of cocaine ingestion equivalent to a human recreational dose of ∼2.5 mg/kg (Smith et al., 2001) would be 20 µM. While this is still considerably higher than the 0.5 µM we found to be reinforcing (Fig. 4), it is also much lower than the mM concentrations of psychostimulants commonly used in Drosophila. This suggests that experiments with lower concentrations of cocaine may be more effective in modeling human recreational doses rather than mimicking acute high-dose cocaine consumption, which would be more akin to overdoses. Our work adds cocaine to the list of drugs of abuse that produce experience-dependent voluntary consumption preference and self-administration in flies, thus opening the door to utilize this highly amenable model organism for elucidating the molecular mechanisms of genes associated with CUD in humans.
Inclusion and Diversity
Two of the authors of this paper come from a disadvantaged background. One author of this paper self-identifies as living with a disability.
Footnotes
We thank the members of the Rodan and Rothenfluh labs for continued discussion and suggestions. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. This work was supported by the Huntsman Mental Health Institute, the University of Utah Molecular Medicine Program, and NIH: the National Institute of Diabetes and Digestive and Kidney Diseases (R01DK110358 to A.R.R.), the National Institute on Drug Abuse (grants K01DA058919 to I.T. and R21DA049635, R21DA040439 to A.R.), and the National Institute on Alcohol Abuse and Alcoholism (grants R01AA026818, R01AA019536-S1, and R01AA030881 to A.R.).
↵*I.T. and P.N.C-B. contributed equally to this work.
The authors declare no competing financial interests.
This paper contains supplemental material available at: https://doi.org/10.1523/JNEUROSCI.1040-24.2025
- Correspondence should be addressed to Adrian Rothenfluh at Adrian.Rothenfluh{at}hsc.utah.edu.
