Associative learning in larval and adult Drosophila is impaired by the dopamine-synthesis inhibitor 3-Iodo-L-tyrosine

Dopamine signaling serves multiple functions, including movement initiation, sleep regulation, motivation, learning, memory extinction and forgetting (Berke, 2018; Meder et al., 2019; Oishi and Lazarus, 2017; Schultz, 2007; Yamamoto and Seto, 2014). In particular, it is crucial for conferring reinforcement signals that teach animals about the causal structure of the world (Ryvkin et al., 2018; Schultz, 2015; Waddell, 2013; Yamamoto and Vernier, 2011). This role of dopamine is found across the animal kingdom, including the fruit fly Drosophila melanogaster. For this model organism, a rich genetic toolbox is available to study the functions of the dopaminergic system. Here, we employ a complementary approach using pharmacological intervention.

Since the 1970s, both adult and larval D. melanogaster have been established as powerful model organisms to investigate Pavlovian conditioning, using odors as the conditioned stimulus (CS) and various types of rewarding and punishing unconditioned stimuli (US) (adults: Busto et al., 2010; McGuire et al., 2005; Perisse et al., 2013; Quinn et al., 1974; larvae: Diegelmann et al., 2013; Gerber and Stocker, 2007; Scherer et al., 2003; Thum and Gerber, 2019; Widmann et al., 2018). The genetic tools available for D. melanogaster have allowed the neurogenetic mechanisms of learning and memory to be investigated, and revealed many striking similarities between the dopaminergic systems of flies and mammals, including humans (reviewed in Yamamoto and Seto, 2014). To mention but a few, flies and mammals share the majority of genes involved in dopamine synthesis, secretion and signaling (Clark et al., 1978; Karam et al., 2020; Riemensperger et al., 2011; Yamamoto and Seto, 2014), as well as the crucial role of dopaminergic neurons in reinforcement signaling (Burke et al., 2012; Liu et al., 2012; Schroll et al., 2006; Schwaerzel et al., 2003; Selcho et al., 2009; reviewed in Scaplen and Kaun, 2016). Of note, in D. melanogaster different sets of dopaminergic neurons signal appetitive or aversive reinforcement, respectively, to distinct compartments of the insects’ memory center, the mushroom body, which harbors a sparse and specific representation of the olfactory environment (Diegelmann et al., 2013; Guven-Ozkan and Davis, 2014; Heisenberg, 2003; Owald and Waddell, 2015; Thum and Gerber, 2019). A similar dichotomy of appetitive and aversive reinforcement signals carried by different sets of dopaminergic neurons may also be emerging in vertebrates (Groessl et al., 2018; Lammel et al., 2012; Menegas et al., 2018). Due to the seductive power, ease and elegance of the available genetic tools in D. melanogaster, however, other useful techniques are used less often in the field. For example, feeding or injecting drugs, although lacking the neuronal specificity of many transgenic tools, is a convenient way of exerting acute systemic effects. Furthermore, these approaches can be combined with genetic methods like cell-specific optogenetic manipulations, allowing greater flexibility in manipulating the animals’ nervous system.

Many drugs affecting the dopamine system in mammals are also effective in flies (Nichols, 2006; Pandey and Nichols, 2011). For example, drugs that target mammalian D1 and D2 receptors have already been used pharmacologically to activate and inhibit their Drosophila homologs in vivo (Chang et al., 2006; Srivastava et al., 2005; Yellman et al., 1997). Also, drugs that induce dopamine deficiency have been found to influence various brain functions. For example, 3-Iodo-L-tyrosine (3IY; other abbreviations sometimes used are 3-IY and 3-IT) interferes with dopamine synthesis by inhibiting the tyrosine hydroxylase enzyme (TH) that catalyzes the conversion of L-tyrosine to L-3,4-dihydroxyphenylalanine (L-DOPA), a precursor of dopamine. As a result, 3IY reduces dopamine levels (Bainton et al., 2000; Fernandez et al., 2017; Neckameyer, 1996) (Fig. S1A). Feeding 3IY to flies decreases activity/locomotion and increases sleep (Andretic et al., 2005; Cichewicz et al., 2017; Tomita et al., 2015; Ueno and Kume, 2014), increases ethanol preference (Ojelade et al., 2019 preprint), and alters courtship behavior (Monier et al., 2019; Neckameyer, 1998; Wicker-Thomas and Hamann, 2008). Regarding learning and memory, 3IY feeding impairs visual and olfactory learning, as well as long-term appetitive ethanol memory in adult flies (Kaun et al., 2011; Seugnet et al., 2008; Zhang et al., 2008). Importantly, these effects of 3IY-induced dopamine deficiency can be substantially rescued by additionally feeding L-DOPA to the flies (Cichewicz et al., 2017; Monier et al., 2019; Riemensperger et al., 2011; Zhang et al., 2008).

In larvae, 3IY feeding has been used to study the developmental effects of dopamine (Neckameyer, 1996, reviewed in Verlinden, 2018) as well as the characterization of dopamine synthesis, reuptake and release (Pyakurel et al., 2018; Xiao and Venton, 2015). Furthermore, 3IY has been found to attenuate the increase in sugar feeding elicited by food odors, an effect that likewise was reversed by additional L-DOPA feeding (Wang et al., 2013).

Here, we provide the first investigation of the effects of feeding 3IY and/or L-DOPA on Pavlovian conditioning in larval D. melanogaster, and report detailed protocols of drug application and behavioral controls. Furthermore, we also feed 3IY and/or L-DOPA to adult flies. We study the drugs' impact on learning about optogenetic activation of an identified dopminergic neuron to examplify the potential of combining genetic and pharmacological approaches, as the drugs' effects on wild-type behavior has previously been shown.

We first investigated the effects of 3IY feeding on D. melanogaster larvae. In an approach modified from Neckameyer (1996), cohorts of 4-day-old larvae were placed on a PET mesh soaked with a yeast solution mixed with 3IY at the indicated concentrations, or without 3IY. After 24 h, the larvae underwent a single-trial Pavlovian training with an odor and a fructose reward, following established protocols (Michels et al., 2017; Saumweber et al., 2011; Scherer et al., 2003; Weiglein et al., 2019): one cohort of larvae was trained by a paired presentation of odor and reward, and a second cohort was trained reciprocally, i.e. with separated, unpaired presentations of odor and reward. In control larvae that were kept on a yeast solution without 3IY, an appetitive associative memory was revealed by higher odor preferences after paired than after unpaired training in a subsequent test (Fig. S1B), indicated by positive performance index (PI) scores (Fig. 1B, left-most box plot). When we performed the same learning experiment with larvae fed with various concentrations of 3IY, we observed decreased memory scores with increased 3IY concentrations. Significantly reduced scores were found for a concentration of 5 mg/ml (Fig. 1B; Fig. S1B), a result we replicated in an independent experiment (Fig. 1C; Fig. S1C). However, we noticed that many larvae had died due to the treatment, and the cuticle of many of the surviving animals was darkened (not shown). We therefore wondered whether the treatment may generally impair behavioral faculties. Indeed, innate odor preference was found to be impaired in 3IY-fed larvae (Fig. 1D). This prompted us to test their basic locomotion on an empty, tasteless Petri dish without odor or sugar, and to analyze their behavior using custom-made analysis software (Paisios et al., 2017). Typically, larvae move by relatively straight runs, interrupted by turning maneuvers indicated by lateral head movements called head casts (HC) (Fig. S1D) (Gershow et al., 2012; Gomez-Marin and Louis, 2014; Gomez-Marin et al., 2011; Paisios et al., 2017; Thane et al., 2019). Analysis of these parameters of locomotion revealed that the animals’ run speed was unchanged by 3IY feeding (Fig. 1E). However, the larvae fed with 3IY systematically performed fewer and larger HCs than control animals (Fig. 1F,G; Fig. S1E-H).

Thus, feeding the larvae with 5 mg/ml 3IY for 24 h seemed to impair their basic behavioral faculties, suggesting that the reduced memory scores that we observed after the treatment might be secondary to such general impairment. Therefore, we next sought to reduce the ‘side effects’ of 3IY feeding.

Given the reported role of dopamine and the TH enzyme in development and cuticle formation (Friggi-Grelin et al., 2003; Hsouna et al., 2007; Neckameyer, 1996; Neckameyer and White, 1993; reviewed in Verlinden, 2018), the timing of 3IY feeding is likely to have an impact. In order to minimize developmental effects, it seems desirable to apply 3IY as late as possible in the larval life cycle (and yet early enough to be able to finish the experiment before the larvae start to pupate). We therefore reduced the duration of 3IY feeding to 4 h, which allowed for the feeding of 3IY to 5-day-old animals. After this shortened feeding protocol too, memory scores were reduced compared to controls (Fig. 2A; Fig. S2A). Critically, the animals’ basic behavioral faculties turned out to be intact: no impairment in innate odor preference (Fig. 2B) or sugar preference (Fig. 2C) was detectable. Thus, the shortened feeding of 3IY specifically impaired associative memory without impairing task-relevant behavioral faculties (nor did we observe any dead or darkened larvae; not shown). This conclusion was also supported by a more detailed analysis of locomotion that revealed only very mild differences to controls (Fig. 2D–F, for more details, see Fig. S2B–E). However, we cannot rule out the possibility of impairments in locomotion or other basic behavioral faculties after the animals underwent the training procedure, caused e.g. by fatigue or adaptation to the stimuli used. Given that we used a very short one-trial training paradigm (about 6 min in total), such effects seem not too likely. Notably, we detected a small increase in the HC rate after 4 h of 3IY feeding (Fig. 2E). This effect seems to be contradictory to the decrease in the HC rate after 24 h of 3IY feeding (Fig. 1F). A closer look revealed that after 4 h feeding the HC rate is increased only for large HC (Fig. S2B,C). After 24 h feeding, the same effect is observed, but additionally the rate of small HC is reduced (Fig. S1E,F), resulting in a total decrease of the HC rate. How these effects of 3IY feeding exactly come about remains unclear.

We next tried to rescue the effect of 3IY on the TH enzyme by additionally feeding the animals with L-DOPA (Fig. S1A). To this end, we fed animals either with plain yeast solution (control), or 5 mg/ml 3IY, or with both 5 mg/ml 3IY and 10 mg/ml L-DOPA. The memory scores were impaired in larvae fed with 3IY alone (Fig. 3A; Fig. S3A), replicating the results from Fig. 2A. These reduced memory scores were restored to control levels by additionally feeding L-DOPA to the larvae (Fig. 3A; Fig. S3A). Innate odor and sugar preferences were not affected by either 3IY or combined 3IY and L-DOPA feeding, confirming that both effects were specific for associative learning (Fig. 3B,C). Importantly, while a repetition of the experiment from Fig. 3A replicated the finding that L-DOPA feeding can restore memory scores upon 3IY treatment, we also showed that the feeding of L-DOPA alone did not increase memory scores (Fig. 3D; Fig. S3B).

After demonstrating the effect of 3IY feeding on associative learning about natural sugar rewards in larvae, we sought to combine 3IY feeding with genetic manipulations of the dopaminergic system, and at the same time to study how broadly applicable the 3IY approach might be. Therefore, we applied it to a different learning paradigm, by using (i) adult flies instead of larvae; (ii) a two-odor differential paradigm instead of a one-odor, ‘absolute’ paradigm; and (iii) an optogenetic punishment instead of a natural taste reward (Fig. 4). Specifically, we expressed the blue-light-gated cation channel channelrhodopsin-2-XXL as the optogenetic effector (ChR2-XXL; Dawydow et al., 2014) in a single dopaminergic neuron per brain hemisphere, called PPL1-γ1pedc (alternative nomenclatures PPL1-01 and MB-MP1), as covered by the Split-GAL4 driver strain MB320C (Aso et al., 2014). This neuron, when optogenetically activated, carries an internal punishment signal sufficient to establish an aversive associative memory when paired with an odor (Aso and Rubin, 2016; Hige et al., 2015; König et al., 2018) (Fig. 4B, left-most box plot). Upon feeding 3IY for 48 h before training, memory scores were decreased, an effect that was restored by L-DOPA feeding (Fig. 4B; Fig. S4A). The effect of 3IY in reducing memory scores increased with increasing 3IY concentrations (Fig. 4C; Fig. S4B), and was equally observed in female and male flies (Fig. 4D; Fig. S5). Critically, 3IY feeding left innate odor preference to either odor unaffected (Fig. 4E,F), which also implies that the animals’ locomotor abilities were intact to an extent that allowed normal odor preferences. We therefore did not perform detailed locomotion analyses. Furthermore, feeding L-DOPA alone did not increase memory scores (Fig. 4G; Fig. S4C). Thus, feeding 3IY specifically impaired associative learning via PPL1-γ1pedc activation in adult flies, but kept their task-relevant behavioral capacities intact.

The present study demonstrates that both in larval and adult D. melanogaster, and in two very different kinds of tasks, feeding 3IY can specifically impair associative learning while innate task-relevant behavior remains intact. In either case, the observed memory impairment was rescued by feeding L-DOPA, suggesting that the 3IY-impairment was indeed caused by an inhibition of the TH enzyme that catalyzes the synthesis of L-DOPA. Regarding adult flies, these results are in line with previous studies that showed that 3IY feeding impairs associative learning about ethanol, quinine or electric shock (Kaun et al., 2011; Seugnet et al., 2008; Zhang et al., 2008). Here, we find a similar impairment of learning about optogenetic PPL1-γ1pedc activation. Previously, a constitutive RNA-interference knockdown of TH in PPL1-γ1pedc revealed that punishment learning by PPL1-γ1pedc activation is dependent on dopamine synthesis in this same neuron (König et al., 2018). Using the more acute, albeit systemic approach of feeding 3IY, we provided an independent confirmation of these results (Fig. 4). Regarding larvae, genetic approaches have uncovered an important role of dopamine for odor-taste associative learning (Rohwedder et al., 2016; Selcho et al., 2009). This is further supported here by an independent pharmacological approach (Figs 2 and 3). Although not unexpected, these results are interesting in themselves by demonstrating for the first time that an acute inhibition of TH impairs associative learning in larvae. This is critical to disentangle acute effects from potential developmental impairments or their compensation.

Indeed, our experiments demonstrate why drug feeding offers a valuable additional approach to manipulate the dopaminergic system of D. melanogaster. It is easy to apply, quick, comparably cheap, and it allows inducing the desired effect shortly before the experiment. The approach also does not require generating new fly strains, but can be easily combined with the use of already available genetic tools. As an example, for the experiments shown in Fig. 4 we optogenetically activated a specific dopaminergic neuron, while inhibiting the TH enzyme in a both systemic and inducible manner. In order to perform the same type of manipulation by genetic means alone, one would have to combine at least five genetic constructs for driving expression of channelrhodopsin-2-XXL in the neuron of interest, as well as of an RNAi against TH in the whole body, plus e.g. a Gal80^ts construct to make the expression of the RNAi inducible. Although that is certainly possible, feeding 3IY is the quicker and easier option. Also, the effects of the drugs can be titrated relatively conveniently by adjusting the concentration and the duration of feeding (Figs 1 and 2). This makes it possible to find a trade-off between maximizing the intended effect on learning and memory and minimizing developmental side effects, or effects on locomotion or sensory function. Furthermore, drugs with comparable effects in different organisms allow for elegant translational research across different species.

The obvious drawback of drug feeding in comparison to present genetic tools is the lack of spatial specificity. However, in some situations, this may actually be advantageous, for example when asking whether a newly discovered process is dependent on synthesis of dopamine at all. In this case, drugs can be used as a first screening, followed up by spatially specific genetic approaches (see also Ojelade et al., 2019 preprint). To give an example, using the genetic driver strain TH-Gal4, which then was believed to cover all dopaminergic neurons, Schwaerzel et al. (2003) suggested that dopaminergic neurons were responsible only for punishment, but not reward signaling (see also Schroll et al., 2006, regarding larvae). This was reconsidered about 10 years later, when refined genetic reagents became available showing that TH-Gal4 largely missed a cluster of dopaminergic neurons that do indeed signal reward (Burke et al., 2012; Liu et al., 2012; larvae: Rohwedder et al., 2016). A systemic pharmacological approach could have made the discovery that dopaminergic neurons carry punishment as well as reward signals possible right away.

Taken together, pharmacological approaches like the one used here enrich the neurogenetic toolbox available for Drosophila and should be considered by the community when investigating the principles of dopaminergic system function.

Drosophila melanogaster were raised in mass culture on standard cornmeal-molasses food and maintained at 25°C, 60–70% relative humidity, and a 12:12 h light/dark cycle.

For larval behavior experiments, we used third instar, feeding-stage wild-type Canton Special larvae of either sex, aged 4 or 5 days after egg laying, as mentioned along with the results. For adult behavior experiments, the split-GAL4 driver strain MB320C (detailed information can be found in the relevant database http://splitgal4.janelia.org/cgi-bin/splitgal4.cgi as well as in Aso et al., 2014), covering the PPL1-γ1pedc neurons (alternative nomenclatures: PPL1-01 and MB-MP1), was crossed to UAS-ChR2-XXL (Bloomington, stock number: 58374, Dawydow et al., 2014) as the effector and kept in darkness throughout to avoid optogenetic activation by room light. Flies of either sex, aged 1 to 4 days after hatching, were used.

Prior to behavioral experiments, animals were fed with solutions of 3-Iodo-L-tyrosine (3IY; stored at −20°C; CAS: 70-78-0, Sigma-Aldrich, Steinheim, Germany) and/or 3,4-dihydroxyphenylalanine (L-DOPA; CAS: 59-92-7, Sigma-Aldrich) at concentrations of 5 mg/ml and 10 mg/ml, respectively, as explained in more detail below. To facilitate reproducibility, we measured the absorption of the solutions in the UV-visible spectrum, using a NanoDrop 2000c spectrometer (ThermoFisher Scientific, Dreiich, Germany). For 5 mg/ml 3IY in distilled water, we found the wavelength of maximal absorption to be 280 nm, and the average absorption at this wavelength to be 4.46. For 10 mg/ml L-DOPA in distilled water, we determined a wavelength of maximal absorption of 280 nm, and an absorption at this wavelength of 7.66.

A 0.5 mg/ml yeast solution was prepared from fresh baker's yeast (common supermarket brands) diluted in tap water and stored for up to 5 days at 4°C in a closed bottle. Samples of 2 ml yeast solution were filled into a 15 ml Falcon tube and kept for a few minutes in a warm water bath. 3-Iodo-L-tyrosine (3IY; stored at −20°C; CAS: 70-78-0, Sigma-Aldrich) was added at a concentration of 5 mg/ml to the respective sample, if not mentioned otherwise. Notably, in contrast to earlier studies using 10 mg/ml or more (Neckameyer, 1996; Wang et al., 2013), we were not able to dissolve concentrations higher than 5 mg/ml. In some experiments, 3,4-dihydroxyphenylalanine (L-DOPA; CAS: 59-92-7, Sigma-Aldrich) was added at a concentration of 10 mg/ml, either to pure yeast solution, or to yeast solution with 5 mg/ml 3IY.

The solutions were thoroughly mixed by attaching the Falcon tubes to a shaker at high speed for approximately 60 min. Empty vials of 5 cm diameter were equipped with two layers of mesh (PET, 500 µm mesh size). Samples of the mixed yeast solution with or without additional substances were distributed onto the mesh of one vial. Larvae of the third instar feeding stage were collected from the fly food by adding 15% sucrose solution (D-Sucrose; CAS: 57-50-1, Roth, Karlsruhe, Germany; in dH₂O) so that the larvae floated up and could be transferred to a Petri dish filled with tap water using a tip-cut plastic pipette. After being rinsed in water, the larvae were loaded onto a filter (pluriStrainer 70 µm, pluriSelect Life Science, Leipzig, Germany) to separate them from water and small food particles, and transferred with a brush to one of the prepared vials. For yeast solutions containing different drugs and/or concentrations, different brushes were used. The larvae were left to feed on the respective yeast solution for 24 or 4 h at 25°C and 60–70% relative humidity. The desired number of larvae were collected with a brush, briefly rinsed in water, and afterwards used in the respective experiment.

Experiments for appetitive odor-fructose associative memory (Saumweber et al., 2011; Scherer et al., 2003) were performed using a one-odor, single-training-trial protocol described in Weiglein et al. (2019) (Fig. 1A, left). For example, two custom-made Teflon containers of 5 mm diameter were filled with 10 µl of odor substance (n-amylacetate, AM; CAS: 628-63-7, Merck, Darmstadt, Germany; diluted 1:20 in paraffin oil; CAS: 8042-47-5, AppliChem, Darmstadt, Germany) and closed with lids perforated with 5–10 holes, each of approximately 0.5 mm diameter. These odor containers were located on opposite sides of a Petri dish (9 cm inner diameter; Nr. 82.1472 Sarstedt, Nümbrecht, Germany) filled with 1% agarose solution (electrophoresis grade; CAS: 9012-36-6, Roth, Karlsruhe, Germany) and additionally containing fructose (FRU; 2 M; purity 99%; CAS: 57-48-7 Roth, Karlsruhe, Germany) as a taste reward (+). Cohorts of approximately 30 larvae were placed at the center of the Petri dish and allowed to move about the Petri dish for 2.5 min. Subsequently, they were transferred with a brush to a fresh Petri dish that was filled with plain, tasteless agarose and equipped with two empty Teflon containers (EM). For each cohort trained in such a paired way (paired training; AM+/EM), a second cohort of larvae received the odor unpaired from the fructose reward (unpaired training; EM+/AM). In half of the cases the sequence of events was reversed (EM/AM+, AM/EM+, respectively).

Cohorts of approximately 20–30 experimentally naïve larvae were collected, briefly washed in tap water, and placed onto a Petri dish with an AM container on one side and an EM container on the other side (Fig. 1A, second from left). After 3 min, the odor preference was determined as detailed in Eqn 1.

Cohorts of approximately 20 larvae were placed on an empty, plain-agarose-filled Petri dish without odor or reward (Fig. 1A, right). For 3 min, they were video-recorded while they freely moved in the dish. The videos were analyzed offline using custom-made tracking software described in Paisios et al. (2017). In brief, larvae alternately perform relatively straight forward-locomotion, called runs, and lateral head movements, called head casts (HC) that are often followed by changes in direction. This leads to a typical zig-zagging pattern of locomotion (Gershow et al., 2012; Gomez-Marin and Louis, 2014; Gomez-Marin et al., 2011). As described in detail by Paisios et al. (2017), an HC was detected whenever the angular velocity of a vector through the animal's head exceeded a threshold of 35°/s and ended as soon as that angular velocity dropped below the threshold again. The time during which an animal was not head-casting was regarded as a run, deducting 1.5 s before and after an HC to exclude the decelerating and accelerating phases that usually happen before and after an HC, respectively. Three aspects of behavior were analyzed: the run speed was determined as the average speed (mm/s) of the larval midpoint during runs; the rate of HCs was determined as the number of HCs per second (HC/s); and the size of HCs was determined by the HC angle. Accordingly, the animal's bending angle as the angle between vectors through the head and tail was determined before and after an HC. Then, the HC angle was calculated as the difference between the animal's bending angle after an HC and the bending angle before an HC. For a detailed description, see Paisios et al. (2017).

To analyze the HC behavior in more detail, we determined the HC rate and HC angle separately for small and large HCs. The discriminatory threshold for large HCs of an HC angle >20° was based on previous studies (Paisios et al., 2017; Schleyer et al., 2015; Thane et al., 2019).

For 3IY feeding in adult flies, a 5% sucrose solution (CAS: 57-50-1, Hartenstein, Würzburg, Germany) was prepared. This solution was either used pure, or mixed with 5 mg/ml 3IY, or with 10 mg/ml L-DOPA, or with both, in an analogous manner to that described above for the larval case. Hatched adults of the genotype MB320C;ChR2-XXL were collected in fresh food vials and kept under the normal culture conditions mentioned above, at least overnight and at most until 4 days after hatching. Flies were transferred to new vials containing a tissue (Fripa, Düren, Germany) soaked with 1.8 ml of sucrose solution that either did or did not contain 3IY and/or L-DOPA, as mentioned in the results section. After 40–48 h under otherwise normal culture conditions, the flies were trained and/or tested en masse.

Two-tailed, non-parametric statistics were used throughout to analyze the behavioral data. For comparisons of a group's scores with chance levels (zero), one-sample sign tests (OSS) were applied. To compare across multiple independent groups, Kruskal–Wallis tests (KW) with subsequent pair-wise Mann–Whitney U-tests (MWU) were used (Statistica 13, StatSoft Inc, Tulsa, USA). To ensure a within-experiment error rate below 5%, a Bonferroni–Holm (BH) correction for multiple comparisons was employed (Holm, 1979). Sample sizes (biological replications) were estimated based on previous studies with small to medium effect sizes (König et al., 2018; Weiglein et al., 2019). None of the specific experiments reported here had previously been performed in our laboratory, although the basic behavioral paradigms are regularly used. Experimenters were blind to treatment condition during the experiments with larvae, and during the fly counting for the experiments with adults. Data are presented as box plots showing the median as the middle line, the 25 and 75% quantiles as box boundaries, and the 10 and 90% quantiles as whiskers. All data from behavioral experiments are documented in Table S1.

Discussions with B. Gerber and P. Stevenson, as well as technical assistance by A. Ciuraszkiewicz-Wojciech, B. Kracht, T. Niewalda and M. Thane, are gratefully acknowledged. We thank R.D.V. Glasgow (Zaragoza, Spain) for language editing. This study received institutional support from the Otto von Guericke Universität Magdeburg (OVGU), the Wissenschaftsgemeinschaft Gottfried Wilhelm Leibniz (WGL) and the Leibniz Institute for Neurobiology (LIN).