We have established a new simple behavioral paradigm in Drosophila melanogaster to determine how genes and the environment influence the behavior of flies within a social group. Specifically, we measure social space as the distance between two flies. The majority of Canton-s flies, regardless of their gender, are within two body lengths from each other. Their social experience affects this behavior, with social isolation reducing and mating enhancing social space respectively, in both males and females. Unlike several other social behaviors in the fly, including the formation of social groups themselves (a well-described behavior), social space does not require the perception of the previously identified aggregation pheromone cis-vaccenyl acetate. Conversely, performance of the assay in darkness or mutations in the eye pigmentation gene white increased social space. Our results establish a new assay for the genetic dissection of a fundamental mode of social interaction.
Mogilner et al. (2003) defined ‘social space’, as the measure of the distance between two individuals. These authors used this measure to mathematically model the spacing of individuals in a social aggregate (such as swarm, flock, school, or herd of birds, fish, locust or groups of bacteria –Mogilner et al. 2003). Body length is an important metric to determine social space in a variety of species, including gulls, separated by one body length in flocks (Emlen 1952), or tufted ducks spaced two to three body lengths apart (Conder 1949). Social space probably results from an equilibrium between multiple attractive and repulsive cues (reviewed by Mogilner et al. 2003), in addition to environmental factors. For example, schools of anchovies increase their social space at night relative to their daytime behavior (Aoki & Inagaki 1988). Similar to many other social behaviors, social space is likely to be subject to a variety of other, complex gene–environment interactions.
Drosophila is emerging as a powerful model for the study of the genetic basis of social behavior (Sokolowski 2010). Flies show a tendency to be attracted to one another and to congregate in all types of test chambers (Bolduc et al. 2010; Lefranc et al. 2001; Navarro & del Solar 1975). Importantly, increased proximity to conspecifics is required prior to other, more complex behaviors (Chen et al. 2002; Connolly & Tully 1998). Thus, social aggregation precedes mating or aggression. However, the mechanisms that determine the behavior of flies within a stable social group are not known.
Here, we describe a simple assay to study a fundamental form of social behavior in the fruit fly: social space within a social group. The dynamics of this behavior differs from other well-studied Drosophila behaviors including the formation of groups (Bartelt et al. 1985; Lof et al. 2009; Rohlfs & Hoffmeister 2004) and the exploratory behavior of groups (Tinette et al. 2004). Furthermore, unlike the formation and exploration of groups, we find social space does not depend on classical odorant perception and therefore may employ different genetic pathways.
Materials and methods
Drosophila stocks and husbandry
Drosophila stocks were raised in standard food cornmeal/molasses/ agar bottles or vials at 25°C with a relative humidity of 20–40% in a 12-h dark–light cycle. Canton-s (Cs) and w1118Cs10 (w1118 outcrossed 10 times to Cs) were from our laboratory stocks (Simon et al. 2003). The following lines were placed in Cs background through outcrosses: Or83b1 and Or83b2 (Larsson et al. 2004), generously provided by Dr John Carlson from Yale University, were outcrossed five times to w1118Cs10. The mutant paraslb1 (Lilly et al. 1994) in a Cs background was generously provided by Dr Joel Levine (University of Toronto).
All behavioral experiments were performed in a genotype-balanced manner (see Drosophila stocks and husbandry section). To minimize the disruption of standard environmental conditions, flies were reared in bottles and thus socially enriched unless noted otherwise in the text, kept as mixed genders to allow mating, and kept with standard food at all times prior to testing. Flies were separated by gender the day prior to each experiment.
As described by (Connolly & Tully 1998) and (Simon et al. 2006), all experiments used flies naive to the test performed. However, for most experiments, they were raised in bottles and therefore socially enriched. Unless otherwise noted in the text, the flies were collected from the bottles when ∼3–4 days old and sexed the day before the experiment, under cold anesthesia, and experiments were performed under ambient light. The flies were allowed to habituate to the testing room for 2 h before each experiment. The experiments were performed at the same time of the day, in a range of 3–4 h in the afternoon (between ZT5 and ZT9), to reduce variation in performance linked to circadian rhythm. All behavioral assays were carried out in the same dedicated room at ∼25°C.
Social space assay
The vertical triangle test chamber was constructed using two square glass plates (18 × 18 cm), separated by 0.47 cm using acrylic spacers to confine the flies to a two-dimensional space. Four spacers were used: two right triangles for each side and two rectangular at the bottom (Fig. 1b), thereby defining an isosceles triangular internal chamber (base: 15.2 cm, height: 15.2 cm). The two bottom spacers were transiently separated to allow the introduction of flies to the chamber.
The horizontal circular chamber shown in Fig. 1a and analyzed in Figs. 1c and 2f–i was created using the plastic culture dish (9 cm of internal diameter and 0.9 cm deep), covered with a glass plate. The glass cover was briefly removed from chamber to allow the introduction of flies.
Flies were raised in bottles, collected under cold anesthesia 3–4 days old after eclosion and placed in vials (40/vial) 1 day prior to the experiment. Two h prior to each experiment, the vials were habituated to the test room (25°C). Flies were collected from vials, using gentle suction with a mouth aspirator (Ejima & Griffith 2010) and introduced into the chamber through the opening between the bottom spacers. After the entrance was closed, the bottom of the chamber was banged on a laboratory bench three times, to ensure that all flies were at the same starting point. Digital images were collected after the flies reached a stable position (up to 20 min).
Digital images were imported in PowerPoint (Microsoft PowerPoint® 2004 for Mac). Each fly in the image was represented by a small dot (0.05 cm in diameter) drawn at the center of the thorax. An automated measure of the nearest neighbor to each dot (fly) was determined using Lispix, a public domain image analysis program (written by David S. Bright, Microanalysis Research Group, NIST; web page: http://www.nist.gov/lispix/LxDoc/home.html). Of note, although there are multiple programs to follow moving flies, to our knowledge there is no dedicated program to analyze still flies.
This data set was imported into Excel (Microsoft Excel® 2004 for Mac), and Prism 4 (GraphPad Software, San Diego, CA, USA) used for most of the statistical analysis. Comparison of histograms was performed using an online Kolmogorov–Smirnov test because these data do not follow Gaussians distribution (KS-test – web page: http://www.physics.csbsju.edu/stats/KS-test.html). All experiments were run at least in triplicate.
To measure the effect of group size on social space, flies were reared in bottles and housed at a density of 40 flies/vial for 12–18 h prior to the assay. A few minutes prior to testing, different number of flies to were introduced into the test chamber using gentle suction with a mouth aspirator: either the whole vial (40 flies), or 30, 20 and 10 flies, were counted as they were collected.
Social space index
The calculation of the social space index (SSI) is based on the values obtained in the histogram representations of the social distance. SSI equals the percentage of flies in the first bin minus the percentage of flies in the second bin (SSI = first bin - second bin). As described in Results section, an SSI =< 0 suggests a lack of social interactions. Kolmogorov–Smirnov tests were used to statistically analyze histograms, and t-tests used for the SSIs.
Generation of a random distribution
Social space under experimental conditions was compared with Monte Carlo in-silico simulations of random distributions of flies. A custom script in the C programming language was created to simulate 20 spatially random realizations of a population of 40 flies in a triangle-shaped or circle-shaped enclosure identical to those used for behavioral experiments. For each simulation, we calculated a histogram of the distances to the nearest neighbor. A representative random distribution is shown in Fig. 2a,f, and the mean and the standard error of each histogram bin from the 20 simulations are shown in Fig. 2d,h.
To study the regulation of local enhancement, we measured the distance between individual flies and their closest neighbor: their ‘social space’ (Mogilner et al. 2003). As described by others, we found that flies form groups in several different types of chambers (Fig. 1a, also see Bolduc et al. 2010; Lefranc et al. 2001; Navarro & del Solar 1975). However, locomotion in horizontally oriented chambers is dominated by exploration and dispersal rather than local enhancement, making it difficult to obtain stable measurements of social space (Fig. 1c, also see Lefranc et al. 2001; Simon & Dickinson 2010; Tinette et al. 2004). In our hands, horizontal group of flies do not assume a static position even after spending 1 h in the chamber. In contrast, in vertically oriented chambers, we find that the flies stop moving and assume a stable position within ∼20 min. The shape of the container also affected the flies' position and behavior, and we found that a triangular shape was most useful. Using the vertically oriented, triangular chamber shown in Fig. 1b, flies show a consistent sequence of behaviors amenable to acquisition of data on social space. For the first few minutes after placing the flies in the chamber, they display an escape response, manifested as negative geotaxis. Negative geotactic behavior ends as the flies crowd into the upper tip of the triangular assay chamber. They then move away from each other and locomote for several minutes (∼5–10 min for males, and up to 20 min for females) in an apparent attempt to reduce local crowding and to explore their new environment. At the end of the exploration phase, the flies remain essentially in one location and engage in sporadic grooming behavior for up to 45–55 min, and their social distance is stable throughout that time (Fig. 1f, cf. 15 vs. 45 min for males, 25 vs. 55 min for females). The relative stability of the distance between flies at this stage allows the easy acquisition of digital images for subsequent quantitation of the flies' location. Although it may also be possible to analyze social space using moving animals, the analysis would be complicated by variations in locomotor speed, and generally more technically demanding than the analysis of still pictures. To confine the image analysis of the final position to only two dimensions, the depth of the chamber was restricted to 0.47 cm (3/16′′). This arrangement prevented two flies from occupying the same x, y coordinate (Materials and methods).
Quantitation of the flies' position in the test chamber showed a surprisingly consistent distribution of distances between each fly and its closest neighbor. We find that 56% ± 3 (mean ± SEM) of the flies lie within 0–0.5 cm from their nearest neighbor (∼two body lengths in magnitude and comparable to social distances observed in other species –Mogilner et al. 2003), and 18% ± 2 are within 0.5–1 cm. The remaining 20% of the population are further apart as shown in Fig. 1d. The pattern of the fly's social space is the same in both genders, consistent with previous observations (Navarro & del Solar 1975 and Fig. 1d–f).
To confirm that the distance between the flies in our assay was governed by social interactions rather than the result of a random distribution, we used a computer simulation to map the location of 40 randomly placed dots (Fig. 2a). In addition, to account for the effects of centrophobism and geotaxis in individual flies lacking all social interactions, we assayed single flies 40 times and merged the data into a single combined image (Fig. 2b). We compared these distributions with that of 40 flies assayed together as described above (Fig. 2c). The flies assayed individually were not only attracted to the sides of the chamber (a well-described behavior –Simon & Dickinson 2010; Valente & Mitra 2007) but also moved away from the top. In contrast, we observed that the flies assayed in-group displayed less attraction to the sides and form a more robust aggregate at the top of the chamber. The flies that do not aggregate at the top and migrate to the sides of the chamber also form closely situated pairs (Fig. 2c).
For each condition, the data representing the distance between each fly (or random dot) was binned and used to generate a histogram, in which we represent the percentage of flies (y-axis) for every 0.5 cm increment (bins on the x− axis –Fig. 2d). The patterns of the three histograms differed significantly (Kolmogorov–Smirnov test, P < 0.00001) showing that interactions between the individual flies influence their distribution in the test chamber.
The most obvious difference between the histograms was the relative size of the first and second bins. For flies tested together under ‘social conditions' (i.e. with other flies), the first bin was consistently larger than the second (Fig. 2d). In other words, there was a higher percentage of flies within two body lengths of each other as compared with those that were >2–4 body lengths from the nearest fly. In contrast, for flies tested separately, the first and second bins were roughly the same size. We therefore generated a simple ‘social space index’ (SSI) by subtracting the percentage of flies in the second bin from the percentage of flies in the first bin (Fig. 2e). Using this metric, we obtain a value of 0 or less when flies behave similarly to those tested individually (with the second bin containing a similar or higher percentage of flies than the first one), and therefore assign 0 as the baseline SSI representing social space in the absence of social interactions (such as in Fig. 2b). We obtain an SSI of 100 when all of the flies are in bin 1 (or 100% of the flies 0–0.5 cm from each other, two body lengths apart or less), which we define as a state of maximal social interaction. For 40 Cs flies of 3–4 days old, on average, we obtain an SSI of ∼40 (38.4 ± 4.6, n = 21 –Fig. 2e).
As explained above, we chose to use a vertical triangular chamber for most tests. However, we also confirmed that non-random aggregation patterns were also seen in horizontal circular chambers, using similar analysis (Fig. 2f–i). The percentage of flies that are two body lengths apart is more variable, seen as the larger bars representing the SEM in Fig. 2h as compared with Fig. 2d, which flies were visualized in the triangular chamber. However, the pattern of distribution is similar to that seen in the vertical triangular chamber and different from what would be seen in a random simulation (Fig. 2h– Kolmogorov-Smirnov test, P < 0.00001). Similarly, the SSI of a random simulation is close to 0, even in this chamber of a completely different size, shape and orientation (Fig. 2i).
For all of the experiments described below, we analyzed both the overall pattern of the histograms and the SSI. We show the SSIs here and the histograms in Fig. S1. To control for behavioral effects of environment seen in many other behavioral paradigms (e.g. associative learning Connolly & Tully 1998), each experiment was performed with a matched, internal control and in a temperature controlled environment. All attempts were made to keep the testing room at constant humidity; it was not feasible to humidify the test apparatus used here without disrupting the fly's behavior.
The most extensive previous studies of group formation show that flies are gregarious and cooperative, and that population size, presence of odorant, pheromones and genes affect this behavior (Lefranc et al. 2001; Tinette et al. 2004). Using the conditions we developed, we first tested whether the number of flies in the chamber would influence the SSI. Although 10 flies showed a lower SSI (Fig. 3a) and a statistically different distribution pattern as seen in the histograms (Fig S1a, supporting information), the distribution was constant using 20, 30 or 40 flies, (Fig. 3a and supporting information Fig. S1a) showing that above a certain threshold, social space did not vary with small changes in group size. For consistency, all further experiments were performed using 40 flies.
The effects of social experience
As for other behaviors (see the section Discussion), we tested whether social space would be similarly affected by social experience. Indeed, we find that male and female flies mated for 3 days showed higher SSI than virgins of same age (Fig. 3b and supporting information Fig. S1b). In addition, flies kept alone in vials for 7 days from age 3–4 days old show a lower SSI than controls, socially enriched for that same period of time (Fig. 3c and supporting information Fig. S1c). These data suggest that, similar to other more complex behaviors (Ganguly-Fitzgerald et al. 2006; Krupp et al. 2008; Svetec & Ferveur 2005; Svetec et al. 2005; Ueda & Wu 2009), social space is influenced by prior social conditions. The simplicity of our assay facilitates experiments to determine the genetic basis of this phenomenon (see below).
Odor and pheromone perception mutants
To test whether odor or pheromone perception might be involved in regulating social space between flies, we measured the SSI of mutants defective in olfaction. We tested paraslb1, a mutant in Cs background with broad olfactory defects: paralytic, which encodes a sodium channel (Lilly et al. 1994). We also tested two null alleles of Or83b (Or83b1 and Or83b2–Larsson et al. 2004), an odorant receptor, required for the perception of most odors and of the volatile pheromone cis-vaccenyl acetate (cVA), after outcrossing six times in our laboratory Cs background. We did not detect any differences between the SSI of these mutants and genetically matched controls (Fig. 4a,b and supporting information Fig. S1d,e).
Possible effects of vision
To test the possible effects of decreased visual cues on social space, we measured the SSI of flies in red light placed in a dark room. In this condition, flies still interact socially, but on average are farther from their nearest neighbor than controls, and thus show a decreased SSI (Fig. 4c and supporting information Fig. S1f). These data suggested that social space might depend at least partially on visual cues.
To determine whether more subtle changes in vision might affect social space, we used a mutant of the white gene (w1118Cs10) in which eye pigmentation is essentially absent (Green 2010). Although white mutants are not blind, they show defects in visual acuity because of the diffusion of light across adjacent photoreceptor arrays (Stark & Wasserman 1974), and outcrossed w1118Cs10 flies display modest but consistent defects in fast phototaxis as compared with genetically matched Cs (Fig. 4d). To control for possible effects on motor behavior in the phototaxis test, we used a standard climbing assay. Cs and w1118Cs10 do not show differences in climbing consistent with the notion that fast phototaxis is decreased because of decreased visual abilities in w1118Cs10 (Fig. 4e).
In social space assays, we found that the w1118Cs10 males and females come to rest farther apart than their genetically matched controls and thus, showed a lower SSI (Fig. 4f and supporting information Fig. S1g), similar to flies assayed in darkness (Fig. 4c). These data suggest that genetic defects in visual behavior may affect social space and more generally show the ability of our assay to detect changes in social behavior in a genetic mutant.
We have developed a simple method to use Drosophila to study a fundamental form of social behavior: social space in a social group. We have reproduced the earlier observation that individual flies' group together, and under the test conditions used here we did not observe sexual or aggressive behaviors, also consistent with the single previous study of social distribution in Drosophila (Navarro & del Solar 1975).
Social space is a group characteristic
Above the threshold of 10 flies/assay, we observed the same SSI with various group sizes. This apparent insensitivity to small changes in group size is similar to that described for some fish schools and groups of birds, as well as the locust hopper band (Mogilner et al. 2003; Uvarov 1928). Furthermore, the density of individuals inside a group under constant environmental conditions seems to be a species-specific characteristic (Mogilner et al. 2003). It will be interesting to determine whether the specific parameters for social space in D. melanogaster differ in other Drosophila species. The availability of sequenced genomes from a large number of closely related species may provide a unique opportunity for cross-species analyses. It may also be of interest to determine why flies behave in a different manner under very sparse conditions (10 flies/assay).
The effects of social experience
Social space is influenced by environmental factors including mating status and prior social isolation. Social interactions, such as mating or aggression, or lack of thereof, such as social isolation, are known to trigger a variety of behavioral responses and physiological modifications in animal species. The effects of mating in flies have been studied extensively (Villella et al. 2008). Mating is known to change the physiology and behavior of female flies, through the effect of sex peptides (Kubli 2003). It is possible that similar pathways could account for the relative lack of social behavior that we observe in virgin females, but it remains unclear why male virgins also appear to be less social. Isolation is also known to affect courtship and courtship memory (Krupp et al. 2008). Other effects of isolation include modification of neuronal excitability (Ueda & Wu 2009), chemical communication (Kent et al. 2008; Levine et al. 2002), sleep patterns (Ganguly-Fitzgerald et al. 2006), olfactory memory (Chabaud et al. 2009), circadian rhythm (Krupp et al. 2008) and aggression in females (Wang et al. 2008).
Social space does not seem to require odor or cVA perception
To begin a genetic analysis of social space, we tested paraslb1, a mutant with broad olfactory defects (Lilly et al. 1994), and two null alleles of Or83b an odorant receptor required for the perception of most odors as well as the volatile pheromone cVA (Or83b1 and Or83b2–Larsson et al. 2004). Both the formation of groups in Drosophila and their subsequent exploratory behavior of groups depend on known olfactory cues (Lof et al. 2009; Tinette et al. 2004). cVA has been shown previously to mediate spatial aggregation at long-range (food coattractant in a bioassay –Bartelt et al. 1985) and engender repulsion, aggression and some sexual behaviors at shorter range (Bartelt et al. 1985; Billeter et al. 2009; Liu et al. 2011; Wang & Anderson 2010). We do not detect a difference in social space in either para or Or83b mutants, suggesting that classical odors and pheromones, including cVA, do not mediate this behavior. These data suggest that social space may employ different mechanisms than a number of other well-characterized social behaviors in the fly.
A possible requirement for vision
Further experiments will be needed to test whether novel pheromones and/or odor receptors (Benton et al. 2009; Yew et al. 2009) mediate social space or if either hearing or taste play a role in social space as shown recently for other social behaviors (Ejima & Griffith 2008; Montell 2009). In this study, we show that social distance increases when assays are performed in the absence of visible light, and that the w1118Cs10 mutant shows a similar deficit. Together these observations suggest that vision might be important for social space; however, we note that w1118Cs10 has been implicated in a variety of fly behaviors (Anaka et al. 2008; Campbell & Nash 2001; Ejima & Griffith 2008; Lee et al. 2008; Lloyd et al. 2002; Svetec et al. 2005; Yarali et al. 2009; Zhang & Odenwald 1995) that may or may not reflect the function of white in the retina, and that white may also regulate amine levels in downstream neurons (Borycz et al. 2008; Yarali et al. 2009). It therefore remains possible that some behavioral effects of w1118Cs10 are because of activity outside of the eye and elsewhere in the nervous system. Thus, our data must be considered preliminary until additional genetic experiments determine whether social space is truly dependent on vision vs. another sensory modality. We note that mutant in white also decreases the ability of flies to cooperate during search behaviors (Tinette et al. 2004) but it remains unclear whether this phenotype results from a visual defect or another unrelated deficit caused by white.
Our data using white, para and Or83b show that the assay we report here is amenable to the analysis of genetic mutants and will facilitate a systematic genetic investigation of this behavior. In addition, its robust nature, low cost and simple method of analysis should allow large scale genetic screens that would be cumbersome with more elaborate assays of social space. Social space is a fundamental social behavior and is shown by all known species. We suggest that further genetic analysis of social space might prove useful for understanding the initiation of a variety of other more complex social behaviors, as well as pathological states relevant to human disease (Adamo & Tesson 2008; Sulzer 2007) (Bolduc et al. 2010; Gottesman & Gould 2003).
We thank Tim Lebetsky for his constructive comments. This work was supported in part by the UCLA-NIH Bridge Program to M.-T.C.; by PSC-CUNY awards, jointly funded by The Professional Staff Congress and The City University of New York; by a 2007 Young Investigator award from NARSAD ‘The World's Leading Charity Dedicated to Mental Health Research’, by a Pilot Grant award from the UCLA Center for Autism Research and Treatment (CART) with funding by the National Institute of Health (NIH) grant (STAART – U54 MH068172, PI: M. Sigman and D. Geschwind), and by a training support from the UCLA Cousins Center at the Semel Institute for Neurosciences with funding by the NIH grant (T32-MH18399) to A.F.S. and NIH grant (R01 MH076900) to D.E.K.