Classical reward conditioning in Drosophila melanogaster


  • Y-C. Kim,

    1. Neuroscience and Genetics Graduate Program, The Huck Institute, and §Department of Biology, 208 Mueller Laboratory, Pennsylvania State University, University Park, PA, USA
    Search for more papers by this author
  • H-G. Lee,

    1. Neuroscience and Genetics Graduate Program, The Huck Institute, and §Department of Biology, 208 Mueller Laboratory, Pennsylvania State University, University Park, PA, USA
    Search for more papers by this author
  • and ,, , K-A. Han

    Corresponding author
    1. Neuroscience and Genetics Graduate Program, The Huck Institute, and §Department of Biology, 208 Mueller Laboratory, Pennsylvania State University, University Park, PA, USA
      *K-A. Han, 208 Mueller Laboratory, Department of Biology, Pennsylvania State University, University Park, PA, USA. E-mail:
    Search for more papers by this author

*K-A. Han, 208 Mueller Laboratory, Department of Biology, Pennsylvania State University, University Park, PA, USA. E-mail:


Negatively reinforced olfactory conditioning has been widely employed to identify learning and memory genes, signal transduction pathways and neural circuitry in Drosophila. To delineate the molecular and cellular processes underlying reward-mediated learning and memory, we developed a novel assay system for positively reinforced olfactory conditioning. In this assay, flies were involuntarily exposed to the appetitive unconditioned stimulus sucrose along with a conditioned stimulus odour during training and their preference for the odour previously associated with sucrose was measured to assess learning and memory capacities. After one training session, wild-type Canton S flies displayed reliable performance, which was enhanced after two training cycles with 1-min or 15-min inter-training intervals. Higher performance scores were also obtained with increasing sucrose concentration. Memory in Canton S flies decayed slowly when measured at 30 min, 1 h and 3 h after training; whereas, it had declined significantly at 6 h and 12 h post-training. When learning mutant t βh flies, which are deficient in octopamine, were challenged, they exhibited poor performance, validating the utility of this assay. As the Drosophila model offers vast genetic and transgenic resources, the new appetitive conditioning described here provides a useful tool with which to elucidate the molecular and cellular underpinnings of reward learning and memory. Similar to negatively reinforced conditioning, this reward conditioning represents classical olfactory conditioning. Thus, comparative analyses of learning and memory mutants in two assays may help identify the molecular and cellular components that are specific to the unconditioned stimulus information used in conditioning.


The capacity to learn and remember distinct environmental cues or actions associated with harm or food reward is fundamental to survival in all animals. To investigate the neurobiological basis of associative learning and memory, simple conditioning paradigms have been employed in various animal models and they are largely classified into classical or Pavlovian conditioning (learning the relationship between two external stimuli) and operant or instrumental conditioning (learning the association between an action and its consequence) (Brembs 2003; Domjan 2005). In both learning situations the reinforcing stimulus can be either aversive or rewarding. Regarding mechanisms of different types of learning, some of the key questions are whether classical vs. operant or reward vs. aversive conditioning involve distinct or overlapping neural structures and cellular processes. These questions are as yet unanswered.

The fruit fly, Drosophila melanogaster, has been tremendously helpful to the analysis of molecular components and cellular pathways that mediate associative learning and memory because it provides an extensive resource for genetic and transgenic mutants, tools and information. Moreover, its relatively short life cycle (∼12 days) makes it quite feasible to investigate a large number of genetically homogeneous animals, avoiding differences caused by an individual’s genetic background and thus correlating behavioural phenotypes with function of a gene or manipulation under test. One of the well-established paradigms in Drosophila is classical olfactory conditioning, which tests the flies’ capacity to associate olfactory [odour, conditioned stimulus (CS)] and aversive mechanosensory [electric shock, unconditioned stimulus (US)] inputs. Typically, a group of flies housed in a plastic tube with an electrifiable grid inside are exposed to a first odour (CS+) accompanied by an electric shock (US) followed by a second odour (CS−) without a shock (Tully & Quinn 1985). Subsequently, flies are tested in a T-maze with two odours presented and their avoidance of the CS+ odour is measured to represent learning or memory retention. This assay has been extensively used to identify numerous molecules (neurotransmitters, ion channel components of various signal transduction pathways and transcription factors) and specific neural substrates required for the acquisition or memory retention of negatively reinforced olfactory conditioning in Drosophila (Belvin & Yin 1997; Davis 2005; Gerber et al. 2004; Margulies et al. 2005; Schwaerzel et al. 2003; Sokolowski 2001; Waddell & Quinn 2001).

Olfactory conditioning utilizing positive reinforcement is also described in Drosophila although it is rather underrepresented. The paradigm first developed by Tempel et al. (Tempel et al. 1983) employs a counter-current distribution apparatus in which flies are allowed to run into a tube containing a copper grid coated with CS+ odour and sucrose, and then are transferred to a tube coated with CS− odour without sucrose. The test is carried out in a T-maze and measures preference to the CS+ for learning or memory retention. This paradigm represents operant conditioning because flies that walk freely into the tube experience CS and US inputs and learn their consequence; whereas, the electric shock-mediated olfactory conditioning described above is classical conditioning, because flies do not have control over the occurrence of CS and US inputs. To uncover molecular, cellular and neural substrates mediating negative vs. positive reinforcement for olfactory learning and memory, it would therefore be ideal to employ an apparatus that represents reward classical conditioning. A recent study (Schwaerzel et al. 2003) describes such a paradigm and utilizes a training tube similar to that used for electric-shock conditioning. For training, flies are passively transferred to the tube, of which the internal surface is covered with filter paper containing sucrose, and exposed to CS+ odour. Subsequently, flies are transferred to another tube containing a filter paper soaked with water and exposed to CS− odour. Wild-type Canton S flies, when tested immediately after training, display stable, but relatively poor performance (learning score 17.7 ± 2.6 out of a possible 100). Thus in this assay, it may not be feasible to identify subtle phenotypes of certain learning or memory mutants. Here, we report on a new training apparatus, named the Sliding box, for reward olfactory learning that represents classical conditioning. This assay is not only easy to set up but it also involves minimal physical stress to flies during training and thus yields consistent behavioural scores that are higher than those reported by the previous study (Schwaerzel et al. 2003). Thus, this new paradigm provides a useful means to delineate the neurobiological basis underlying distinct types of associative learning and memory.

Materials and methods

Drosophila strains and culture

All flies were reared on standard cornmeal/agar medium at 25 °C and 50–60% relative humidity under a 12 h light/12 h dark cycle. A wild-type strain used in this study was Canton S. Isogenic w1118 was used as a control for the homozygous t βhnM18 mutant, which is deficient in tyramine β-hydroxylase (Monastirioti et al. 1996). The t βhnM18 was kindly provided by Dr Monastirioti (Institute of Molecular Biology and Biotechnology, Greece). Flies (4–6 days old) were collected and housed in fresh food vials for at least 1 day before use.

Behavioural apparatus and assays

Flies were starved in the vials containing Kimwipes soaked with water for 20 h before training. Dehydration of flies significantly affected viability and behavioural scores; thus, it was critical for flies to have direct access to water during starvation and Kimwipes, compared to filter papers, were highly effective in providing sufficient water. Odorants, isoamyl acetate (IAA) and ethylacetate (EA), were diluted in mineral oil and sucrose solution was made in deionized water. All chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA).

The new training apparatus, the Sliding box (Fig. 1), represents a prototype and consists of three parts, a main chamber (Fig. 1b) and two identical plates (Fig. 1a), made of Plexiglas. The main chamber (outer dimension, 50 mm W × 70 mm L × 30 mm H) has two open slots on each of two long sides for supporting the two plates together. One of the two short sides of the chamber has a hole to accommodate an adaptor for introducing flies and airflow with or without odour. The other short side has five small holes for ventilation. Four corners inside the chamber at the side with an adaptor have triangle-shaped blockades to prevent flies escaping from odour exposure. One side of each plate (65 mm W × 340 mm L × 5 mm H) has six indented rectangular spaces (48 mm × 38 mm × 1 mm), which hold filter papers (Whatman Inc, Florham Park, NJ, USA). The plates are inserted into the main chamber in such a way that the filter papers face each other inside the chamber, yielding a compartment (inner dimension, 40 mm W × 60 mm L × 10 mm H) where the flies are housed and trained. The chamber can be slid into six different positions along the parallel plates and each position is assigned by parallel filter papers, which serve as a ceiling and a floor of the assembled compartment (Fig. 1d, schematic presentation). Five sets of filter papers (each set consisting of a top and a bottom filter paper) are sequentially soaked with water, sucrose, water, water, and water for training and the sixth set serves as a backup (Fig. 1c). The assembled unit is stably secured on the flat Plexiglas plate with four sets of holders (Fig. 1c).

Figure 1.

Sliding box conditioning apparatus and training regimen. (a) Two plates with six indented surfaces for holding filter papers. (b) Top (left), long-side (centre) and short-side (right) views of the main chamber. (c) Training set-up consisting of Sliding box assembly, odour bottles, tubing, and electric valves for air/odour delivery. (d) Schematic presentation of training in Sliding box. (e) Standard training schedule.

For training, 50 to 60 flies were transferred to the main chamber positioned at the first water-soaked filter papers and were exposed to air for 30 seconds (Fig. 1e). Subsequently, the chamber carrying flies was gently slid to the next position with sugar-containing filter papers and the flies were exposed to the first odour CS+ for 1 min. The chamber was then slid to serial compartments where flies received air for 45 seconds, the second odour (CS−) for 1 min, and air for 30 seconds consecutively (Fig. 1d,e). A second set of flies was trained with the odours presented in a reversed order to counterbalance any odour bias. The trained flies were transferred to a T-maze with two odours presented and tested for their preference of the CS+ odour immediately after or at various time intervals after training to measure memory retention. Odours were delivered from glass vials containing diluted odorants to the main chamber via airflow (flow speed, 3.5 l/min) in silicon tubing. Switching of air and odour was controlled by four three-way electrical solenoid valves (Omega, Stamford, CT, USA) (Fig. 1c). The performance index (PI) was calculated by subtracting the percentage of flies that chose CS− (i.e. the wrong choice) from the percentage of flies that chose CS+ (the right choice) and an average PI of two counter-balanced trainings was used to represent one data point. If all of the trained flies preferred CS+, the PI would be 100, representing perfect learning or memory retention. If the trained flies chose CS− and CS+ equally, PI would be 0 (no learning or no memory).

Data analysis

All data are reported as mean ± standard error of the mean. Statistical analyses were performed using MINITAB 14 (Minitab Inc., State College, PA, USA). Student’s t-test was used for comparing the means of two groups. Analysis of variance (anova) with post hoc Tukey–Kramer tests was used to measure differences among the means of more than two groups.


Sliding box paradigm

The sucrose-mediated olfactory conditioning depicted here calls for an association of olfactory (CS) and appetitive gustatory (US) inputs and measures preference behaviour as an indicator of learning and memory. To establish a classical conditioning paradigm of this behavioural plasticity, we developed a new apparatus named Sliding box. In the Sliding box, flies were housed in a movable rectangular chamber of which the internal top and bottom surfaces were provided by two long parallel plates containing five sets of filter papers (Fig. 1). Upon sliding the chamber to different areas of the plates, flies were forced to contact water or sucrose on filter papers. During training, flies were exposed to CS+ odour in a compartment with sugar surfaces, thus involuntarily receiving an appetitive sugar US through the legs and the proboscis; they were then slid to other compartments with water surfaces to be exposed to CS− odour or air (Fig. 1d, e). The leg tarsal segments, which directly contact the sucrose, contain taste neurons whose axons project to the sub- oesophageal ganglion in the brain (Stocker 1994; Wang et al. 2004). They respond to appetitive stimulus as sucrose directly applied to the tarsi stimulates the proboscis extension reflex and feeding behaviour (DeJianne et al. 1985; Stocker 1977). In this assay therefore, flies did not have control over the appetitive US occurrence in a similar manner to those flies receiving aversive electric shocks on the legs in negatively reinforced olfactory conditioning; thus, both paradigms represent classical, as opposed to operant, conditioning. To establish optimal training conditions for this new assay, we explored several parameters including multiple training sessions, inter-training intervals, stimulus saliencies and training procedures.

Effects of multiple trainings and inter-training intervals

In electric shock-mediated olfactory conditioning, flies exhibit better performance with more training and the degree of enhancement depends on the inter-training intervals (Beck et al. 2000; Tully & Quinn 1985). To examine whether performance in sucrose-mediated olfactory conditioning was influenced by multiple training and inter-training intervals, wild-type Canton S flies were subjected to appetitive conditioning using 2 m sucrose, which is close to the saturation concentration (2.62 M at 25 °C) (Mathlouthi & Reiser 1995), as the US in the Sliding box. In the previous study employing the sugar tube (Schwaerzel et al. 2003), the odorants IAA and EA were used as CS and they were therefore used in this study also. After surveying several different concentrations, 2% IAA and 2% EA were chosen because Canton S flies showed similar levels of avoidance for both odours against air (avoidance score for 2% IAA, 63.8 ± 5.1, = 6; avoidance score for 2% EA, 65.4 ± 1.9, = 6) and naive flies without training were equally distributed between 2% IAA and 2% EA (Fig. 2a). After one training session, flies displayed a substantial level of performance, which was enhanced after an additional training session with 1-min inter-training interval (Fig. 2a). Four cycles of training also improved performance but the difference in the performance scores obtained after two and four trainings was not statistically significant. Previous studies of operant counter-current and classical sugar-tube conditioning show performance scores of Canton S flies in the range 30–36 and 17.7 ± 2.6, respectively, after two cycles of training (Schwaerzel et al. 2003; Tempel et al. 1983). Thus, the performance scores in the range 30–40 observed in this study (Fig. 1–4) were comparable to or higher than the previously reported scores.

Figure 2.

Effects of multiple training sessions and inter-training interval. (a) Canton S flies were tested for their preference to the sugar-associated CS+ odour right after one (1T), two (2T), or four (4T) cycles of training or without training (0T). One-factor anova revealed a significant effect of training (F3,44 = 14.09, < 0.0001, = 6–18 per group). Post hoc Tukey comparisons showed the significant difference between performance scores for non-trained (0T) and trained (1T, 2T or 4T) flies. The performance score with 1T was significantly different from that of 2T or 4T (two-tailed Student’s t-tests, < 0.05; marked by an asterisk); however, there was no difference between 2T and 4T performance scores (> 0.05). (b) Flies were tested immediately after two (2T) or four (4T) training sessions with 1-, 15-, or 30-min inter-training intervals. One- factor anova revealed a small yet significant effect of inter-training interval in 2T (F2,45 = 3.31, < 0.05, = 12–18) but no significant effect in 4T (F2,15 = 0.73, > 0.05, = 6). The performance score for the 30-min inter-training interval in 2T was significantly different from those of the 1-min or 15-min inter-training intervals (two-tailed Student’s t-test, < 0.05; marked by an asterisk) but there was no difference between the 1- and 15-min inter-training intervals (> 0.05).

Figure 3.

Effects of sucrose concentration and CS order. Flies were trained with different concentrations (0.1, 1, or 2 m) of sucrose solution (a) or with the standard (CS+/CS−) or reversed (CS−/CS+) order of CS (b), and were tested immediately after training. One-factor anova revealed a significant effect of sucrose concentration (F2,15 = 4.28, < 0.05, = 6 per group). Post hoc Tukey comparisons showed a significant difference between 0.1 and 2 m sucrose conditions (marked by an asterisk). There was no significant difference between different orders of CS (two-tailed Student’s t-test, > 0.05, = 6).

Figure 4.

Memory retention. Canton S flies displayed gradual memory decay when tested at various times after training (F5,30 = 5.6, < 0.001, = 6 per time-point). Post hoc Tukey comparisons revealed that the performance scores at 6 and 12 h post-training were significantly different from those at other time-points (marked by an asterisk).

To examine the effect of inter-training intervals in multiple training sessions on performance in the Sliding box, Canton S flies were subjected to two or four training cycles with 1, 15 or 30-min inter-training intervals. After two training sessions, performance with inter-training intervals of 1 and 15 min was comparable; however, the performance score with a 30-min inter-training interval was significantly lower than that with 1- or 15-min inter-training intervals. Performance was not significantly affected by different inter-training intervals after four training sessions, although there was a tendency for a longer inter-training interval yielding a lower performance score (Fig. 2b). Overall, the highest performance score was obtained with a 1-min inter-training interval; thus, all experiments described below were carried out with two training sessions and a 1-min inter-training interval.

Effects of sucrose concentration and training procedure

The associative strength of US and CS for learning depends on the saliency of the stimuli (Beck et al. 2000; Tully & Quinn 1985) and the context of their presentation (Tully & Quinn 1985). Thus, we investigated the effectiveness of three different concentrations of sucrose solution, 0.1, 1 and 2 m. Flies conditioned with higher concentrations of sucrose (possibly stronger US saliency) displayed better performance (Fig. 3a): while the performance score of flies trained with 2 m sucrose was not significantly different from that of flies trained with 1 m sucrose, it was significantly higher than that for 0.1 m sucrose. Since flies trained with 2 m sucrose yielded the highest performance score, 2 m sucrose solution was used as the US in all the conditioning experiments described here. In electric shock-mediated conditioning, flies trained with CS+ followed by CS− learn better than those trained in the reverse order of CS (Tully & Quinn 1985). When the effect of CS order during training was explored in the Sliding box, Canton S flies exhibited similar performance with either CS order (Fig. 3b); thus, training in the order of CS+ and CS− was adopted as a standard procedure.

Memory retention

Most, if not all, memories gradually decay over time; however, the decay kinetics is variable depending on conditioning types (Tempel et al. 1983). To establish a memory retention curve, Canton S flies were subjected to two training cycles with a 1-min inter-training interval in the Sliding box and were tested right after (<3 min) or at 0.5, 1, 3, 6 and 12 h after training. Performance at 30 min or 1 h after training was comparable to that immediately after training (Fig. 4), indicating negligible memory decay for at least 1 h. On the other hand, performance was compromised at 3 h post-training and significantly worsened at 6 h. Nonetheless, the level of performance observed at 6 h was maintained at 12 h post-training, suggesting that memory formed in appetitive conditioning lasted for a relatively long time. A significant percentage of flies died at later time-points, presumably as the result of starvation, thus it was not feasible to test memory retention for longer than 12 h.

Analysis of tβh

To ascertain the utility of a new conditioning assay, it is critical to demonstrate its sensitivity to detect phenotypes associated with certain genetic mutations or epigenetic effects. Tyramine β-hydroxylase is an essential enzyme for the biosynthesis of octopamine, a major monoamine in invertebrates (Monastirioti et al. 1996). Flies with a mutation in the tyramine β hydroxylase gene, t βhnM18, have been previously shown to have an undetectable level of octopamine and to exhibit an impaired acquisition of appetitive conditioning (Schwaerzel et al. 2003). When homozygous t βhnM18 flies were challenged in the Sliding box, they displayed poorer performance immediately after training compared to the control w1118 fly, which has the same genetic background as t βhnM18 (Fig. 5), indicating that octopamine is crucial for appetitive learning in the Sliding box as well. This suggests that the new positively reinforced classical olfactory conditioning is able to distinguish impaired performance resulting from mutations in learning or memory genes, or potential factors influencing learning and memory, such as aging or drug treatments.

Figure 5.

Impaired learning of the octopamine-less t βh mutant. The control w1118 and t βhnM18 flies were tested immediately after training. Two-tailed Student’s t-test showed the significant difference between their performance scores (< 0.01, = 6 per group; marked by an asterisk).


In this report, we described a novel conditioning system, the Sliding box, for testing flies’ capacity to learn and remember the smell associated with sugar reward and establishment of its optimal conditions. This paradigm offers several advantageous features. First, the mechanical stress received by flies during training (via gentle sliding) was probably milder than that caused by passive transfer of flies from one tube to another in sugar-tube conditioning (Schwaerzel et al. 2003). This may enhance the flies’ attention to the relevant cues for learning and thus may account for their better performance (PI immediately after training, 30–40) in the Sliding box than in the sugar-tube conditioning (PI immediately after training, 17.7) (Schwaerzel et al. 2003). Second, while flies in the counter-current and the sugar-tube apparatus are housed in the same compartment for air exposure during training, flies in the Sliding box were placed in different compartments before, between, and after CS (Fig. 1e). This would probably reduce cross-contamination of CS and US, improving learning scores. Third, the Sliding box paradigm represents classical conditioning similar to the commonly utilized electric shock-mediated conditioning. Moreover, the same CS odours and testing apparatus (T-maze) were used in Sliding box and electric shock-mediated conditioning. Consequently, differences in the flies’ performance in these classical conditioning assays could be correlated only with the nature of the US information that flies receive during training. Thus, comparative analysis of various mutants in these assays may provide valuable insights into common vs. distinct molecular, cellular and neural components for processing punishment vs. reward information.

Comparing the performance of Canton S flies in Sliding box, counter-current, and sugar-tube appetitive conditioning, we noted several differences and similarities. As described above, performance after two cycles of training with a 1-min inter-training interval in the Sliding box was similar to that in the operant counter-current paradigm but was better than that in the classical sugar-tube paradigm. No information is available regarding whether more than two cycles of training enhance performance in the counter-current assay, whereas four cycles of training with 1-min inter-training interval did not improve performance in Sliding box and sugar-tube conditioning. This suggests that the physiological processes mediating two classical conditioning types, Sliding box and the sugar tube, may be comparable. It is conceivable that increased training of more than two cycles in the Sliding box may affect long-term memory but not learning or short-term memory, similar to electric shock-mediated conditioning (Tully et al. 1994). Performance scores for memory retention at 30 min, 1 h and 3 h after training in Sliding box remained between 20 and 30 but were dramatically reduced at 6 and 12 h post-training (Fig. 4). Although no information is available on memory retention in sugar-tube conditioning, memories formed in operant counter-current conditioning have declined slowly because only a small decay is noticeable at 6 and 12 h post-training (Tempel et al. 1983). It is conceivable that different memory decay kinetics in the Sliding box and counter-current conditioning may be associated with distinct types of learning (passive learning in classical Sliding box conditioning vs. active learning in operant counter-current conditioning). Nonetheless, substantial memory retention up to 3 h after training in the Sliding box should be significant to identify and characterize molecules important for learning and short-term memory in reward conditioning, as shown with the tβh mutant (Fig. 5).

Notably, Schwaerzel et al. elegantly demonstrated common as well as distinct components that were essential for electric shock- and sucrose-mediated conditioning (Schwaerzel et al. 2003). While both conditioning types require rutabaga-adenylyl cyclase function in the mushroom body intrinsic neurons, US inputs for aversive electric shock and appetitive sucrose are delivered by dopamine and octopamine neurons, respectively, to the mushroom bodies. Consistently, two dopamine and one octopamine receptors that activate cAMP increases are shown to be highly enriched in the mushroom bodies (Han et al. 1998; Han et al. 1996; Kim et al. 2003; Lee et al. 2003). The newly implemented conditioning described here provides an additional tool to address specific roles of these receptors and to uncover new players, which will ultimately help delineate the neurobiological underpinnings of punishment vs. reward olfactory learning and memory. In addition, it should also be feasible to establish other types of conditioning by adopting different US or CS in the Sliding box. For example, olfactory CS may be replaced with visual CS by applying distinct colours or visual patterns to filter papers. This would help to identify molecular and cellular components involved in olfactory vs. visual learning and memory.


We greatly appreciate Dr Maria Monastirioti for providing t βhnM18 flies, Drs Kyungok Cho (Baylor College of Medicine, Houston, TX, USA) and C. Nelson Hayes for valuable comments on the manuscript, and an undergraduate research participant Sue Hae Chae for technical assistance. This work was supported by National Institute of Child Heath and Human Development Grant HD048766.