- Top of page
- Materials and Methods
- Conflict of Interest
Learning can be defined as a change in behaviour driven by the memory of previous experience (Davis 2005) and may help an animal to adapt its behaviour in response to changing environmental circumstances (Dukas 2004). While many vertebrates have been shown to rely on learning under natural conditions (Dukas 2004), evidence for ecological relevance of learning in insects is mostly limited to bees (e.g., Menzel and Muller 1996; Hill et al. 1997; Menzel et al. 2005; Raine and Chittka 2008; Ings et al. 2012), parasitoids (e.g., van Nouhuys and Kaartinen 2008; Hoedjes et al. 2011; Froissart et al. 2012; Thiel et al. 2013) and macrolepidoptera (e.g., Rausher 1978; Stanton 1984; Cunningham et al. 2004; Snell-Rood and Papaj 2009). Several of those studies provide ecological underpinning for specific laboratory assays of learning in those insect groups (e.g., Menzel and Muller 1996; Raine and Chittka 2008; Hoedjes et al. 2011; Thiel et al. 2013). However, not all laboratory learning assays extrapolate to nature. For example, although honey bees remember flowers associated with perceived danger in the laboratory, they are apparently unable to learn to avoid flowers with predatory crab spiders in the field (Dukas et al. 2005). Thus, even though many more insect species have been shown to learn in laboratory conditioning assays, extrapolating from such assays to nature may be problematic, especially when the assays do not have an obvious connection with the animal's ecology.
A case in point is the fruit fly Drosophila melanogaster, a favourite model species for studying the genetics, neural mechanisms and evolution of associative learning (Fig. 1; Davis 2005; Kawecki 2010). Several experimental paradigms for quantifying associative learning in Drosophila in the laboratory have been developed, involving associations of odours or visual cues with shock, heat, bitter taste or sugar reward (e.g., Tempel et al. 1983; Tully and Quinn 1985; Scherer et al. 2003; Foucaud et al. 2010). The use of these paradigms has greatly advanced our understanding of the mechanisms of learning, but their relevance to what Drosophila may learn in nature is unclear. In these assays the flies are either immobilized or confined to a very small space; the experimental stimuli are strong and the flies cannot avoid perceiving them. A few somewhat more ecologically relevant laboratory assays have demonstrated that experience can change flies' mating and oviposition behaviour (Wolf and Heisenberg 1991; Mery and Kawecki 2002; Dukas 2005b). Still, even those assays confine flies to highly spatially restricted and very simple environments. It is not clear to what extent those learned responses would scale up to natural environments, where the spatial scale is orders of magnitude larger and a multitude of stimuli compete for the fly's attention. For example, a male fruit fly constrained with an unreceptive female in a small space subsequently refrains from courting even receptive females for several hours (“courtship conditioning”; Siegel and Hall 1979); however, this does not occur under less constrained conditions more akin to flies' natural environment (Dukas 2005a). Similarly poor correspondence between laboratory and field have been found for an innate behavioural pattern – circadian activity rhythm (Vanin et al. 2012).
Not having a nest or brood care, fruit flies may not need the cognitive abilities required for homing, while their short lifespan under natural conditions (Rosewell and Shorrocks 1987) would limit their chances to benefit from past experience. It could be argued that the learning observed in the laboratory indirectly supports the relevance of learning ability to fitness in nature: if learning were not beneficial, this costly trait (Mery and Kawecki 2003) would be eliminated by natural selection. Yet, it is also possible that some basic level of learning ability is conserved as a by-product of general neuronal plasticity (important in nervous system development), even if learning is ecologically irrelevant or costly (Dukas 2009). So the fact that fruit flies learn in the laboratory does not necessarily imply that learning is ecologically relevant for this animal in nature.
Knowing if and what fruit flies learn in nature would throw light on the evolutionary forces maintaining learning ability and provide ecological underpinning for the neuroscience-oriented research on learning in this model species. Furthermore, learning has been proposed to buffer populations against environmental fluctuations (Stephens 1991), affect predator-prey population dynamics (Ishii and Shimada 2012), facilitate expansion into novel habitats (Sutter and Kawecki 2009), modulate evolutionary change (Paenke et al. 2007) and initiate speciation (Thorpe and Jones 1937; Dukas 2005b). Thus, showing that even a small, short lived, non-social insect makes use of learning in nature would greatly extend the potential taxonomic relevance of those hypotheses.
A few studies have addressed the effect of experience on behaviour of Drosophila under field conditions, with mixed results. Jaenike (1985, 1986)) reports increased attraction of D. melanogaster (and Drosophila tripunctata) in field releases to food on which the flies had been previously kept in the laboratory for a week. Similarly, D. melanogaster which emerged from pupa in immediate vicinity of a particularly flavoured food were subsequently more attracted to that flavour under field conditions (Jaenike 1988). However, other studies with similar design (Hoffmann and Turelli 1985; Hoffmann 1988; Turelli and Hoffmann 1988), found no or inconsistent evidence for effects of prior food exposure on its subsequent attractiveness to flies released in the field. All those studies involved non-differential learning; i.e., the flies only acquired experience of one food type rather than comparing food sources of different palatability. Thus, where effects were detected, they could reflect imprinting or sensitization (or habituation in cases where reduced attraction to previously experienced food was found; Hoffmann and Turelli 1985; Turelli and Hoffmann 1988), irrespective of food quality. Furthermore, while flies possess consolidated memory and can remember an association between shock and odour for over at least 24 h in laboratory assays (Dubnau and Tully 1998; Isabel et al. 2004; Mery et al. 2007), it is not clear if consolidated memory is relevant to food choice in nature.
In this paper we address the ecological relevance of learning in fruit flies by testing if they learn about food quality, retain this information overnight, and use it to choose between food types in a greenhouse setting. This setting emulates the natural environment in that expanse of space the flies can explore is vast compared to typical laboratory learning assays, and the food sources must be located from afar and approached in flight. Furthermore, plants, pots, soil and greenhouse construction elements provide olfactory and visual complexity and heterogeneity. In our assays, flies were first given the opportunity to learn which of two food substrates (apple- or orange-flavoured) tastes good and which one is less palatable as a result of being laced with quinine (“learning phase”). Subsequently, their attraction to the two food sources was assayed in the greenhouse by setting out traps baited with apple and orange (“test phase”). If their food choices were affected by past experience, the flies should shift their food preference in the test phase towards the previously palatable flavour. We report three experiments which show that fruit flies modify their food preference under the greenhouse conditions based on what they previously learned in the laboratory (experiment 1), that they do so even after remaining in the greenhouse overnight, indicating involvement of consolidated memory (experiment 2), and that they can acquire new learned information while free-flying in the greenhouse (experiment 3).
- Top of page
- Materials and Methods
- Conflict of Interest
In this study we demonstrated that the relative attraction of fruit flies flying in a greenhouse environment to alternative odours is affected by the quality of food with which these odours were previously associated. Flies that had experienced palatable apple-flavoured food and unpalatable orange-flavoured food were more likely to be attracted to the odour of apple than flies with the opposite experience. This occurred both when the experience had been acquired under free-flying conditions and under more artificial and confined conditions in the laboratory. The degree to which flies' preference for apple versus orange odour was modified after the flies could learn while free flying in the greenhouse (Experiment 3) was similar to that observed after the flies were first subject to a learning treatment in the laboratory and subsequently tested in the greenhouse (Experiment 1). Furthermore, while the effect of learning in our experiments may seem small, it was similar in magnitude to that reported in a laboratory oviposition learning assay conducted in small boxes preventing flies from flying and only containing the two substrates (Mery and Kawecki 2002). It thus appears that the greenhouse environment, which is richer and more “noisy” in terms of stimuli and which requires flight for exploration, neither poses an obstacle to learning new information, nor significantly impairs flies' ability to make foraging choices based on previously learned information.
The effect of experience on food preference is overlaid on the pre-existing innate preference, and the limited magnitude of this effect means that learning in our study did not lead to reversal of innate preferences. In particular, in experiment 1 flies showed overall greater attraction to the orange than to the apple odour, as confirmed by the release of unconditioned flies. While flies that experienced orange as unpalatable significantly shifted the relative preference away from orange and towards apple, they still chose apple at most half of the time. Conversely, in later experiments (2 and 3) flies were in general considerably more attracted to apple (confirmed by the second release of unconditioned flies), so that even those that experience orange as palatable and apple as bitter tended to choose apple traps as often or more often than orange traps. However, similar observations have been made in laboratory learning assays (Mery and Kawecki 2004). Thus the inability of learning to reverse pre-existing innate biases points to limits on Drosophila learning, but it does not negate the fact that learning, defined as a changed of food odour choice based on past experience, occurred in our experiments. Furthermore, if an animal shows a strong but maladaptive innate preference for a particular food, even a small reduction of this preference may have large fitness consequences. In an extreme case when the innately preferred food turns out to be lethally toxic, reducing the likelihood of choosing it from 90% to 80% would double fitness.
While the greenhouse obviously still differs from a natural environment, its spatial scale and relative complexity are presumably more relevant to the natural ecology of Drosophila than the confines of laboratory learning assays (see 'Introduction'). A few previous studies (Jaenike 1985, 1986, 1988) demonstrated an effect of adult experience acquired at the time of emergence from pupa on food preference in free-flying flies outside of the laboratory (although other studies failed to find it; see the 'Introduction'). We extend those results in three ways.
First, in previous studies of the effect of experience on food preference in free-flying flies, flies were exposed to the focal odour or flavour upon emergence. These experiments were thus specifically meant to address the “chemical legacy hypothesis” (Corbet 1985), according to which chemical signals encountered by an insect upon emergence may influence its subsequent feeding and oviposition behaviour. In contrast, our flies emerged in standard culture vials and first fed on a standard cornmeal/sugar/yeast medium; they were only exposed to the focal food substrates several (3–10) days after emergence. Thus, our results, in particular experiment 3, indicate that the ability of flies to learn under free-flying semi-natural conditions extends beyond being conditioned to the first food encountered in adult life.
Second, in contrast to those previous field studies, in our study flies were exposed to two substrates, both having an attractive odour but one being highly palatable and the other less so due to presence of quinine. Even though the flies likely spent more time on the more palatable food during the learning phase, the proximity of the two substrates meant that they were exposed to both odours. This excludes simple odour exposure as the cause of the change in subsequent odour preference. Rather, our results are most parsimoniously explained by flies having learnt the association between food odour and its palatability, in analogy to associative learning about food quality observed under laboratory conditions (Mery and Kawecki 2002).
Third, we have shown that memory persists overnight under the free-flying greenhouse setting in the absence of the focal odours, indicating that a consolidated form of memory is involved. Two forms of consolidated memory − long-term memory (LTM) and anaesthesia-resistant memory − have been described in laboratory olfactory classical conditioning in Drosophila; they are the only memory forms that persist beyond several hours (Isabel et al. 2004; Davis 2005). In laboratory studies of consolidated memory flies are maintained between conditioning and test under conditions that minimize exposure to odours and other stimuli, which may be particularly favourable to consolidation and retention of memory (Dacher and Smith 2008; Lagasse et al. 2009; Burns et al. 2011). Hoffmann and Turelli (1985) reported apparent retention of memory about food in flies maintained over 24 h under such minimal-stimulus laboratory conditions (vials with no food) and subsequently released and tested in the field. However, this result was inconsistent between experiments and may have been confounded by the degree of food fermentation (Hoffmann 1985); another study reported that the effect of experience disappears in the field within hours (Jaenike 1986). Our results thus provide the strongest support yet for the importance of consolidated memory in Drosophila foraging. In our Experiment 2 the flies still showed increased preference for the food they previously experienced as more palatable, even though in the meantime they spent 13–16 h in the greenhouse, in the absence of the focal food odours, but exposed to other odours such as those emanating from potted plants and soil. While part of the overall time spent in the greenhouse was spent roosting on plants as soil (as we observed), the flies are also likely to have actively sought moisture and food, in particular during the natural peaks of activity at dusk and dawn. The stimuli encountered during that time might have interfered with the pre-existing memory (Dacher and Smith 2008; Lagasse et al. 2009). The fact that their memories of food-associated odours from the previous day persisted under these conditions is a strong indication that such memories may also persist in nature. Thus, while we cannot discern whether flies in our assays relied on long-term or anesthesia-resistant memory, our study indicates that consolidated memory may be ecologically relevant even for small, short lived insects.
Overall, our results provide strong support for the notion that fruit flies modify their food choice in a natural context based on previous experience. Learning detected in laboratory conditioning paradigms is thus not merely an artefact of strong stimuli implemented in such assays, small spatial scale and absence of other stimuli. Learning about food and oviposition sites is likely to be important for D. melanogaster in nature. Fruit flies feed on decaying fruits, the quality and palatability of which vary in space and time, depending on the stage of decomposition, the microorganisms involved, and the degree of desiccation. Volatile compounds through which flies are attracted to fermenting substrates (Markow and O'Grady 2008) may not reflect the quality of the substrate accurately. If so, it will often be beneficial for flies to be able to adjust their attraction to volatiles based on their local, recent experience. While it has been suggested that fruit flies spend extended periods of time on or near the food sites (Spieth and Heed 1972), they may be driven away temporarily, for example, by the heat of mid-day (Feder 1997), as well as roost away from the feeding sites overnight. If that is the case, retaining learned information over prolonged periods might be useful in relocating temporarily abandoned, but still appropriate food sites, or locating new ones with similar properties. While this remains to be directly demonstrated, our results support the plausibility of such a scenario. They also provide an ecological underpinning for laboratory learning assays involving associations between food and odours and used to study the mechanisms or evolution of learning in this model species (Mery and Kawecki 2002; Chabaud et al. 2006). Finally, fruit flies probably do not possess extraordinary learning abilities compared to other insect groups (Dukas 2008). Hence, our results suggest that the ecological significance of associative learning in insects extends beyond the special cases of social Hymenoptera, parasitoids and butterflies reviewed in the 'Introduction'.