We examined the use of social information in fruitfly larvae, which represent an ideal model system owing to their robust learning abilities, small number of neurons and well-studied neurogenetics. Focal larvae showed attraction to the distinct odour emanating from food occupied by other larvae. In controlled learning experiments, focal larvae preferred novel odours previously paired with food occupied by other larvae over novel odours previously paired with unoccupied food. When we gave groups of larvae a choice between food patches differing in quality, more larvae aggregated on the higher-quality food, suggesting that attraction to and learning about cues associated with other larvae can be beneficial. Furthermore, larvae were more likely to find the best available food patch in trials when that food patch was occupied by other larvae than in trials when that food patch was unoccupied. Our data suggest, however, that the benefits from joining others may be at least partially offset by the fitness costs of increased competition, because larvae reared in isolation did as well as or better than larvae reared in groups on three key fitness parameters: developmental rate, survival rate and adult dry body mass. Our work establishes fruitfly larvae as a highly tractable model species for further research on the mechanisms that modulate behaviour and learning in a social context.
There has recently been increased interest in establishing simple, tractable model systems for research on the evolution of and neurogenetic mechanisms underlying social behaviour [1–3]. In addition to basic interest in social behaviour , such research may help us form the foundation for treatments of social disorders in humans [5–8]. A key feature of social animals is the ability to engage in social learning, defined as the acquisition of novel information from other individuals. We still do not know how prevalent social learning is among animal species. However, it has had remarkable effects on some species, most notably on humans, in which it has generated a rich culture . While there has been intensive research on the evolution of social behaviour, empirical work on the evolution of social learning is rather limited. Furthermore, until recently, most research on social learning has focused on vertebrates and eusocial insects [10–12].
As a part of a series of experiments on the evolution of social learning in insects , we examined social behaviour and social information use in fruitfly (Drosophila melanogaster) larvae. Adult fruitflies are moderately social. Most notably, the pheromone cis-vaccenyl acetate (cVA), produced by males and transferred to females during copulation, serves as a long-distance attractant promoting adult aggregation [14–16]. Both cVA and an individual's cuticular hydrocarbons modulate aggression between males [17,18]. Social experience also influences fruitflies’ circadian rhythms and the expression of cuticular hydrocarbons [19–21]. Finally, adult fruitflies show social learning in the contexts of egg laying and mate choice [22–25]. Because adult female fruitflies tend to aggregate and lay eggs at a single site, many larvae typically share a food substrate, and thus social behaviour may occur at the larval stage as well. Identifying social interactions among larvae opens opportunities for analysing social behaviour and the use of social information in a simple and tractable model system with well-studied learning abilities [26–28] and neurobiology [29–31]. We began by examining social attraction in the larvae. We then tested whether larvae learn to prefer cues associated with other larvae. Finally, having found both social attraction to and learning from social cues, we assessed some of the benefits and costs larvae incur from joining other larvae.
2. Material and methods
We maintained three population cages each containing several hundred D. melanogaster Canton-S on abundant standard food at 25°C and 60% relative humidity, and on a 12 L : 12 D cycle with lights on at 01.00. This irregular light cycle placed peak egg-laying at midday, so that we could collect experimental eggs within a very short time window of about 1 h. We collected eggs on 85 mm (diameter) Petri dishes filled with 10 ml of standard food and covered with 0.7 ml of live-yeast suspension (30 g dry live yeast l−1 of warm water) to stimulate egg laying . Immediately following egg laying, we transferred these dishes to an incubation chamber maintained at 25°C, high humidity and total darkness. We conducted all further manipulations and tests under far red light, which fruitflies cannot see , in order to minimize disturbance and phototaxis. Data are archived in Dryad (doi:10.5061/dryad.qq304).
(b) Food preparation
In several experiments, we created social and non-social food discs. We placed discs of food (ranging from 1.15 to 2.5 ml, depending on the experiment) in 85 mm Petri dishes containing a thin layer of agar. To social discs, we added groups of 20–30 randomly selected larvae, which fed on the discs for 18–42 h prior to testing, depending on the experiment. After such feeding, we considered food to be used, as opposed to unused fresh food, which was identical in quality and age, but had not been occupied by larvae. Used food has a notably different texture, smell and (presumably) taste than fresh food. Because the larvae on social stimuli may have provided social cues to the focal larvae, we refer to them throughout as ‘models’.
3. Social attraction
We began our investigations by testing for simple social information use: attraction to a substrate frequented by others. We placed a social and a non-social food disc on opposite sides of a Petri dish containing a thin layer of agar (figure 1a). We tested each focal third-instar larva individually by placing it through a 1 cm hole in the lid at the centre of the Petri dish, equidistant to either disc, and recording its choice, defined as making contact with a disc within 5 min (see electronic supplementary material).
First, in experiment 1A, we gave focal larvae a choice between a social and a non-social disc. We conducted tests in 60 mm agar Petri dishes with the food discs placed on opposite sides. Focals were placed 7 mm from either disc. In this experiment, however, we reared focals with others for the first 2 days of life and so they may have learned to prefer the familiar cues associated with others. To eliminate this possibility, in experiment 1B, we reared each focal larva individually by placing each egg into its own 60 mm Petri dish containing 0.3 ml of standard food, which is abundant for a single larva. These isolated larvae experienced no other larvae prior to testing.
Next, in experiment 1C, we tested whether the social attraction observed in the first two experiments was a general phenomenon by testing larvae from a wild-caught population feeding on and tested with ripe banana. In experiment 1D, we tested which cue served as the attractant for the focal larvae: the distinct odour emanating from food previously occupied by larvae for about a day or some cue directly originating from the model larvae. We compared larval attraction with social foods in two tests each involving a choice between foods: (i) used food without models versus fresh food, and (ii) fresh food with models versus fresh food. As a control, we also included our baseline test involving used food with models versus fresh food. Used food consisted of a food disc consumed by 20 early-third-instar model larvae for 24 h. Depending on the treatment, we left the models, removed models from used food discs or added models to fresh food discs immediately before testing.
Finally, in experiment 1E, we assessed whether the attractive cue associated with used food was due to the presence of larvae, and not due merely to an increased salience of the food (e.g. because of increased surface area). We tested larval attraction to used food without models versus artificially ‘used’ fresh food, which we had made to resemble used food by artificially simulating larval foraging with a needle, generating small grooves and scratches in the surface and underside of the food disc.
Focal larvae reared both in groups and in isolation showed significant attraction to the social food discs of larvae and used food (groups: 66.7%, n = 126, generalized linear model (GLM) intercept: χ21 = 13.8, p < 0.001; figure 1b, experiment 1A; isolation: 71.6%, n = 67, χ21 = 11.7, p = 0.001; figure 1b, experiment 1B). Similarly, focal larvae from a wild population showed significant attraction to the banana slice that had been used by larvae overnight (68.4%, n = 76, χ21 = 10.0, p = 0.002; figure 1b, experiment 1C). In the test of the nature of the attractive cue, focal larvae showed significant attraction to the used food without model larvae but not to fresh food containing model larvae (respectively: 63.0%, n = 81, χ21 = 5.5, p = 0.019; and 48.4%, n = 64, χ21 = 0.03, p = 0.874; figure 1c, experiment 1D). As before, focal larvae showed significant attraction to used food occupied by models (72.0%, n = 82, χ21 = 14.6, p < 0.001; figure 1c, experiment 1D). Larval attraction to the social food was similar in the tests consisting of used food with models and used food without models (p = 0.245). Attraction to fresh food with models was significantly lower than attraction to used food with models (p = 0.004). Attraction to used food persisted even when the alternative food was similarly textured artificially ‘used’ food (68.0%, n = 50, χ21 = 6.2, p = 0.013; figure 1c, experiment 1E).
4. Learning from social cues
Next, we asked whether larvae learn to prefer novel cues associated with other larvae. All experiments consisted of pairing one novel odour with a social food and another novel odour with a non-social food, and then testing the subsequent odour preference (figure 2a). Training and preference test (see electronic supplementary material) were adapted from previous larval learning assays [27,28,33].
In experiment 2A, one odour was paired with a 1.25 ml social food disc occupied by 30 early-third-instar models, which had been feeding on that disc for 18 h, and the other odour was paired with a non-social food disc, consisting of fresh food without models. Next, in experiment 2B, we tested which component of the social experience was critical for the learned odour preference: used food or the model larvae per se. We had two treatments in which we trained larvae with one odour paired with non-social food (fresh food without models) and the other odour paired with either (i) used food without models or (ii) fresh food with models. As a control, we also included our baseline test, which paired one odour with used food with models and the other with fresh food. Additionally, we removed a circle of 1 cm diameter (0.1 ml) from the centre of each disc to ensure that focal larvae could more easily contact model larvae, which often crawl beneath the food discs. For used food without models and fresh food with models, we removed or added models, respectively, immediately prior to training.
The results from experiment 2B indicated that focal larvae learned to prefer novel odours associated with both used food with no larvae and models on fresh food (figure 2b). In experiment 2C, we directly tested which factor was more important to the larvae: used food or other larvae. We tested whether focal larvae preferred an odour previously paired with (i) used food without models or an odour previously paired with (ii) fresh food with models. As a control, we simultaneously replicated experiment 2A. If larvae do not learn from their direct interactions with others, then they should prefer an odour paired with used food over an odour paired with others on fresh food. If, however, direct interactions with others improve the perceived quality of a food, then they should prefer the odour paired with used food less strongly than controls. Additionally, we observed a subset (71%) of the fresh food with larvae training dishes to quantify social interactions between focals and models. We recorded the proportion of time (out of the total nine possible minutes) that each focal larva spent within 2 mm (approx. 1 body length) of a model larva. Typically, focal larvae crawled beside and remained in contact with other larvae. Once a focal was near models, it usually stayed close to them for the remainder of the training session.
In experiment 2A, focal larvae chose the odour previously paired with social food (used food with models) more frequently than the odour paired with non-social food (77.5%, n = 71, GLM: χ21 = 15.4, p < 0.001; figure 2b, experiment 2A). In experiment 2B, focals chose the odour previously paired with the social food more frequently in all treatments: when the social food was used food with models (65.6%, n = 61; GLM: χ21 = 5.5, p = 0.019; figure 2b, experiment 2B), used food without models (66.2%, n = 65, χ21 = 6.4, p = 0.012) and fresh food with models (61.3%, n = 62; χ21 = 3.8, p = 0.050). There was no overall difference in the frequency of social choices between the three tests (GLM: χ22 = 0.429, p = 0.807), and pairwise comparisons revealed no significant differences between the three tests (all p > 0.528; figure 2b, experiment 2B). In experiment 2C, focals did not differ in preference for odours previously paired with used food without models or fresh food with models (50%, n = 48, GLM: χ21 = 0.01, p = 0.937; figure 2c), and the presence of model larvae on the fresh food significantly reduced preference for the odour paired with used food in test trials compared with controls (GLM: χ21 = 4.1, p = 0.044; figure 2c). We replicated our previous results from experiment 2A, with larvae choosing an odour previously paired with used food with models significantly more often than an odour previously paired with fresh food alone (68.6%, n = 51, GLM: χ21 = 7.1, p = 0.008). Our quantification of social interactions revealed that focal larvae spent 52.4 ± 3.8% (n = 41) of their time within 2 mm of model larvae.
5. Benefits and costs of joining others
In our final three experiments, we addressed the ultimate evolutionary question of why focal larvae prefer to join others. First, we asked whether an aggregation of larvae can be a valuable source of foraging information to other larvae. If groups of larvae tend to aggregate at the best sites in their environment, then individuals can rely on the cues of foraging conspecifics to quickly locate high-quality sites. In experiment 3A, we tested whether groups of larvae are more likely to aggregate on the best available food in their environment. We randomly selected 30 larvae and placed them at the edge of an 85 mm agar dish, 3 cm from two 2.5 ml discs of food (2.3 cm diameter, 6 mm thick). Dishes contained either (i) one disc of standard food (100%) and one disc of 50% food, or (ii) one disc of 50% food and one disc of 25% food. Additionally, the food discs were presented in one of two possible configurations: touching or separated by 1 cm.
In experiment 3B, we tested whether individual larvae were better at locating the best locally available food patch when that patch was occupied by other larvae than when it was unoccupied. We allowed focal larvae to choose between a low- and high-quality food in one of two conditions. In the models-absent condition, individual focal larvae could choose between the two food patches based on food-derived cues only. In the models-present condition, we placed 30 larvae on the higher-quality food disc 18 h prior to testing. In short, we gave larvae a choice between (i) low-quality food and (ii) either social or non-social high-quality food. We analysed the frequency of choices with a generalized linear model with a binomial distribution and logit link function, including factors for the presence/absence of model larvae, foods available, side of food discs and relevant interactions.
Finally, in experiment 3C, we assessed the developmental effects of group foraging. We measured key parameters related to fitness as a function of larval group size. We transferred 1, 3, 10 or 50 eggs to dishes with 2.5 ml of standard food immediately after egg laying. As a reference, fruitfly laboratories typically rear a few dozen flies per vial containing 5 ml of standard food [34,35]. We recorded larval developmental rate, egg-to-adult survival and adult body mass. See electronic supplementary material for further details. If foraging aggregations improve fitness in this context, then we would expect moderately sized groups of larvae to develop faster, larger and with lower mortality rates than either larvae reared alone or in large groups with increased competition.
In experiment 3A (group choice), larvae showed significant preference for aggregating on the higher-quality food for all food combinations and configurations (all t49 > 3.7, all p < 0.001; figure 3a). For the 100% versus 50% nutrition food tests, the proportion of larvae choosing the 100% food was 0.808 ± 0.024 when the foods were touching and 0.617 ± 0.031 when 1 cm apart. For the 50% versus 25% food tests, the proportion choosing the 50% food was 0.917 ± 0.011 when touching and 0.708 ± 0.031 when 1 cm apart (figure 3a). When the discs were touching, a significantly greater proportion of larvae chose the higher-quality food than when the discs were 1 cm apart (F1,192 = 56.3, p < 0.001). When the available foods were 50% and 25%, a greater proportion of larvae chose the higher-quality food than when the two foods were 100% and 50% (F1,192 = 18.6, p < 0.001). There was no significant effect of side (F1,192 = 1.0, p = 0.331), and no significant interactions (all p > 0.365).
In experiment 3B (individual choice), focal larvae chose the higher-quality food more often in the presence than the absence of model larvae (GLM: χ21 = 6.7, p = 0.009). In the presence of models, focal larvae chose the higher-nutrition food significantly more frequently in both the 100% versus 50% and the 50% versus 25% food conditions (respectively, 76.9%, n = 39, GLM: χ21 = 10.0, p = 0.002; and 66.7%, n = 39, χ21 = 4.1, p = 0.042; figure 3b). Without model larvae on the higher-quality food, focals did not differ from chance (respectively, 48.6%, n = 37, χ21 = 0.4, p = 0.842; and 54.1%, n = 37, χ21 = 0.2, p = 0.619; figure 3b). There was no significant effect of food types available, side of food disc presentation or the interaction between available foods and the presence of model larvae (all p ≥ 0.288). The presence of model larvae did not affect choice latency (58.7 ± 6.3 versus 61.0 ± 8.0 s, with and without larvae, respectively; t150 = 0.2, p = 0.815).
In experiment 3C, larval density negatively affected developmental rate, survival and adult body mass (figure 4). Larval density decreased developmental rate (Kaplan–Meier survival analysis with Mantel–Cox log rank chi-square: χ23 = 29.6, p < 0.001; figure 4a). Post hoc comparisons revealed that larval development was significantly slower in the density of 50 larvae than all others (all p < 0.001), and that 1 and 3 versus 10 approached significance (respectively, p = 0.090 and p = 0.059). Density negatively affected egg-to-adult survivorship (F3,36 = 6.4, p = 0.001; figure 4b). Post hoc comparisons showed that the density of 10 larvae had the lowest survivorship, significantly lower than densities of 1 and 3 larvae (Tukey HSD, respectively, p = 0.025 and p = 0.001). A planned contrast of low density (1 and 3) versus high density (10 and 50) revealed a significantly lower survivorship in the higher-density than the low-density treatments (t36 = 3.7, p < 0.001). Increasing density also significantly reduced adult body mass in both males (F3,98 = 118.5, p < 0.001) and females (F3,79 = 69.3, p < 0.001; figure 4c). See electronic supplementary material for further details.
Our main findings were that (i) fruitfly larvae are attracted to odours emanating from food used by other larvae; (ii) larvae prefer novel odours previously associated with other larvae over novel odours previously associated with non-social alternatives; (iii) for a foraging larva, other larvae can be a useful source of social information about high-quality food; and (iv) when larvae join others, they may incur costs owing to competition. We discuss each of these results in turn.
(a) Social attraction
In our first series of experiments, we found that focal larvae showed significant attraction to food patches occupied by other larvae, and this was consistent whether or not we reared focal larvae in a group or in isolation (figure 1b). This indicates that focal larvae did not merely show attraction to an already-familiar group setting. Furthermore, we replicated the social attraction results using larvae from a recently collected wild population reared on natural fruit (figure 1b). Larvae far away from food rely on cues that lead them back to food, and cues of other feeding larvae are especially relevant because they indicate that others have found a site with sufficiently high-quality food. Moreover, food patches that have been occupied by larvae for several hours develop a distinct odour. Experiment 1E suggests that larvae are attracted to this odour (figure 1c) and not to the direct presence of larvae at a food site. Finally, experiment 1E indicates that the attractive odour is associated with feeding larvae rather than with mere mechanical disturbance of the food. The tendency of animals to join others and form aggregations has been studied for a long time [36–38]. Our experimental work on fruitfly larvae allows us to link work on social attraction to simple cases of social information use in a leading model system highly amenable to experimental manipulation in both evolutionary ecological and neurogenetic arenas.
One could argue that the larvae in our experiments (figure 1) did not actually show social attraction in the strict sense because they were not attracted directly to others, but instead to the volatiles in food consumed by others. However, social attraction should always be based on the most relevant and salient cues available, and the ultimate cause of all social attraction is some fitness benefit such as the opportunity to locate and feed on higher-quality food [37,38].
(b) Learning from social cues
To assess the magnitude of social information use by larvae, we asked whether larvae assigned higher values to novel odours associated with relevant social settings. In agreement with the data for social attraction, we found that the larvae preferred novel odours previously associated with either used food occupied by larvae or used food from which we had removed the larvae (figure 2b). Interestingly, larvae also preferred odours paired with fresh food occupied by larvae over odours paired with fresh, unoccupied food (figure 2b), and larvae did not prefer odours paired with used food over odours paired with fresh food containing models (figure 2c), which suggests that experiencing direct interactions with other larvae on a food increases the perceived quality of that food.
(c) Benefits and costs of joining others
Our model system is somewhat unique because it allows us to quantify potential benefits and costs of social information use. We found that, given a choice between foods of different quality, groups of larvae were more likely to settle on the better option (figure 3a). Importantly, the distance between the high- and low-quality food patches had strong effects on larval choice, with fewer larvae settling on the high-quality food when the inter-patch distance was greater (figure 3a, white versus black bars), suggesting that limited mobility and perception may prevent larvae from readily locating the best available food patches. Given such limitations, it may be highly beneficial for larvae to be attracted to odour cues associated with others and to learn about novel cues associated with others. Indeed, we found that focal larvae were significantly better at locating a higher-quality food when that food was occupied by larvae than when it was unoccupied (figure 3b).
While the information gleaned by seeking others has obvious benefits, we also documented some costs. Isolated larvae had the heaviest adult dry body mass (figure 4c). This can translate into higher fitness, because males prefer larger females, which are more fecund [39,40], and larger males have a mating advantage owing to both superior fighting ability and female preference for larger males [41–43]. Moreover, isolated larvae did as well as or better than a modest group of 10 larvae in terms of developmental rate and survival from egg to adult (figure 4a,b). Costs associated with aggregation are well known from a large variety of species [36,38], and our results are consistent with those showing such costs among D. melanogaster in both laboratory and natural settings [44,45].
One can imagine some benefits from being in a small group, including suppressing mould, enhancing the growth of preferred species of yeast and bacteria, and improved ability to dig into the substrate [45–49]. Such benefits, however, may not be important in our laboratory settings, where we provide larvae with a diet containing yeast and a mould inhibitor. We cannot yet provide an estimate of the net benefit larvae may gain from joining others in natural settings. Overall, though, our results are in agreement with previous work highlighting the trade-offs involved in joining others: individuals searching for the best available site may rely on the inadvertent social information of others who have already found such a site; by joining others, however, an individual increases the level of competition at that site [36,38].
(d) Conclusions and prospects
We have established fruitfly larvae as a simple, highly tractable model system for studying social behaviour and socially influenced learning. This is especially exciting given that larvae have only about 3000 functional neurons and that there are powerful tools available for studying their neurogenetics [30,50,51]. The most logically consistent explanation for our results is that focal larvae use cues of others as a guide to superior feeding sites. Learning about novel cues associated with others and then preferring such cues over alternatives constitutes social learning, defined as the acquisition of new information by an individual (observer) through interaction with either another individual (model) or cues left by that individual . While one can question whether such simple social learning can inform us about elaborate cases of social learning among vertebrates, experience clearly indicates that simple, tractable behaviours and brain functions identified in fruitflies have been instrumental for furthering our understanding of behaviour and cognition in more complex animals, including humans [52,53]. Further work on fruitfly larvae can elucidate the social cues or signals they rely on, and the neurobiological pathways that modulate behaviour and learning in a social context.
We thank A. Luby, S. Lalani, M. Lee, K. Wong, S. Greenaway and A. Adamiec for assistance, S. Balshine, A. Campos and L. Dukas and two anonymous referees for comments on the manuscript, and the Natural Sciences and Engineering Research Council of Canada, Canada Foundation for Innovation, and Ontario Innovation Trust (R.D.) and McMaster School of Graduate Studies (Z.D.) for financial support.
- Received May 31, 2013.
- Accepted July 9, 2013.
- © 2013 The Author(s) Published by the Royal Society. All rights reserved.