Social hierarchy is established and maintained with distinct acts of aggression in male Drosophila

Social interactions pivot on an animal’s experiences, internal states, and feedback from others. This complexity drives the need for precise descriptions of behavior to dissect the fine detail of its genetic and neural circuit bases. In laboratory contests, Drosophila males reliably exhibit aggression, conventionally scored using lunges as a proxy, allowing its study with the many available genetic tools. Here, we use an explicit approach to identify the onset and reversals in hierarchical relationships and observe that distinct aggressive acts precede, concur, and follow dominance. We find that lunges are inadequate in establishing social dominance. Lunges rather appear to reflect the dominant state of a male and help in maintaining his social status. Lastly, we characterize the recurring structure of aggression that emerges through subsequent reversals in dominance. Collectively, this work provides a framework for studying the complexity of agonistic interactions in flies enabling its neurogenetic basis to be understood with precision.


INTRODUCTION 24
Agonistic interactions important for establishing social hierarchies over food, territory, or 25 mates often progress as a sequence of stereotyped acts. Like other goal-driven 26 behaviors [1], the steps along this sequences are triggered by context and modified by 27 past experience [2,3], through largely unknown mechanisms that are coupled to an 28 animal's internal physiological state [4]. Determining these mechanisms remains 29 difficult, however, because social exchanges proceed as a dynamic, reciprocal, and 30 continual feedback loop of interactions between two (or more) decision-making 31 individuals that progresses over both moment-by-moment and protracted times scales 32 [5]. Intra-male adversarial contests in the fruit fly Drosophila melanogaster seem well 33 suited for studying such complex social phenotypes, due to its established study in a 34 laboratory settling [6,7] and the multitude of techniques available for manipulating and 35 recording gene and neuron function [8]. Nonetheless, efforts to explain which acts lead 36 to dominance hierarchies [9,10] have suffered from a lack of consensus on methods 37 and results for studies of aggression (as discussed within [11,12]). 38

39
Here we use an explicit approach to characterize the temporal relationship among 40 various aggressive acts and the establishment of social hierarchy between pairs of male 41 flies in arenas previously developed for high-throughput phenotyping [13,14]. Using this 42 approach, we observe that the acts of fencing, boxing, and tussling appear well 43 positioned in time to be related to the establishment of social hierarchy, whereas 44 lunging and chasing its maintenance. Unexpectedly, the commonly studied aggressive 45 act of lunging, near-universally, follows the onset of social hierarchies and thus is unlikely involved in its establishment. We carried out a simple social manipulation and 47 observed that males in contests between mixed pairs of genotypically high versus low 48 lunging lines became dominant at chance levels. This result suggests that the total 49 number of lunges executed by an individual is inadequate in explaining his social 50 outcome. We further observed that lunging persists as dominant males confront 51 unfamiliar opponents in unexplored settings. Lastly, we report that a faction of the 52 aggressive acts surveyed also intensify through subsequent reversals in social 53 dominance adding to a fuller appreciation of this complex, recurrent social exchange. 54 Together, this work provides a framework for untangling which aggressive acts causally 55 relate to the establishment of social dominance and those that are a consequence 56 57

Identifying hierarchical relationships 59
Upon introduction to experimental arenas, pairs of males whom had never fought 60 (henceforth, "naïve males") display independent and haphazard movement that may 61 include one or both males jumping, taking flight, and climbing the wall as they attempt to 62 escape. With time males settle into a period of exploration during which they eventually 63 meet, and commonly this initial contact leads to bouts of short, alternating pursuits that 64 are characteristically restricted to the floor (which is entirely made up of food; see 65 Methods). Predictably this engagement then progresses to a situation where one male 66 consistently pursues the other with the retreating male leaving the floor (resource) 67 presumably attempting to flee by climbing the wall. It is this explicit characteristic 68 sequence -of one male pursuing and a second attempting to climb the arena wall -seen in Figure 2). To confirm this finding we ran a second, independent group of 116 contests and carefully reviewed the incidences of lunging (both direct inspection with 117 slow playback to correct for false negative displays and semi-automatically for false 118 positives; see Methods for details). Again males lunged after the establishment of 119 dominance, with the majority of the lunges executed by the dominant male (see Figure  120 S3). 121 122 From the observation that lunging occurs after the onset of dominance we reasoned 123 that they were unlikely to play a role in its establishment. If this were true executing 124 more (or less) lunges should not influence whether an individual emerges dominant. To 125 test this, we set up contests with mixed pairs of males from genotypically high and low 126 lunging lines and measured the total number of lunges executed by those males that 127 emerged dominant. We saw no influence of total number of lunges on dominance 128 outcomes, even when males from mixed pairs executed dramatically lopsided amounts 129 of lunging (see fourth columns in Figure 3A Table S2 for additional examples and details regarding the various fly 131 lines). It was therefore unsurprising to observe that males from both the high and low 132 lunging genotypes also executed lunges following the establishment of dominance (see 133 Figure S4). Collectively, these results provide evidence that lunges do not establish 134 dominance, and more likely play a role in its maintenance. 135 For the majority of contests, we observed that only a single, stable hierarchical 140 relationship formed between pairs of naïve males (as shown in Figure 2 and Figure  141 S3). In approximately one in five contests, however, after the establishment of 142 dominance there were subsequent reversals in social status, with lunges (sometimes by 143 both males) clustering near each reversal. To better understand how these lunges 144 relate to reversals in status, we determined the identities of males executing each lunge 145 (see Methods). The majority of the lunges preceding a reversal in status were 146 displayed by the currently dominant male; whereas, dominant males executed nearly all 147 of the lunges following reversals (see Figure S3). This analysis provides evidence that 148 lunging reflects current and likely past dominance status, and further support the 149 interpretation that lunges play a role in maintaining a male's social status. 150 151 Both increased total time holding dominance and also the occurrence of reversals in 152 social status contributed to a persistent high level of lunging by individuals against 153 familiar opponents. Moreover, it has been reported that "winner" males lunged and 154 became dominant earlier in contests against naïve opponents [19]; and also, that males 155 appeared to change their fighting tactics as a consequence of winning or losing [20]. 156 Therefore, to test if lunging functions as a general tactic for holding on to dominance 157 irrespective of opponent or location, we paired the emergent dominant males (from a 158 first contest between two naïve males; see Figure 4A) with naïve males in unexplored 159 arenas (see Figure 4B). The most salient feature of these second contests was that 160 unlike for pairs of naïve males, that rarely lunged before the establishment of dominance, the majority of dominate males (although not their paired naïve opponents) 162 lunged prior to establishing dominance (see Figure 4C). The temporal sequence for all 163 other aggressive acts studied (including lunging after the initial establishment of 164 dominance) appeared comparable to those observed in contests between two naïve 165 males (data not shown). Finally, to examine if the precocious lunging helped dominant 166 males reassert dominance against unfamiliar males in unexplored arenas, we quantified 167 the number of contests in which dominant males that had lunged early were successful 168 in reasserting their dominance status. For the majority of these contests, previously 169 dominant males reasserted their dominance (see Figure 4D). 170

DISCUSSION 172
Social dominance forms when individuals yield to the agonistic advances of others, 173 often in the context of conflict over resources such as access to food, territories, and 174 reproductive opportunities [21]. However, the relationship between specific aggressive 175 acts and the establishment of social hierarchies remains contentious largely for they 176 appear so highly correlated, even with disagreements over which drives the other [22]. 177 In this work, we use an explicit approach to characterize with high temporal resolution 178 how a range of known acts of aggression relate to the establishment, maintenance, and 179 subsequent reversals of social hierarchy. 180

181
The approach introduced within this work has several advantages. It is easy to describe 182 and straightforward to implement. It accommodates that adversaries compete over The complexity of reversals in dominance requires more study. From a preliminary 210 characterization of a larger number of contests with reversals, we observed that lunging 211 appears as just one of several interacting aggressive acts likely involved in the process 212 (see Figure S5). Taken together, it is clear that further consideration is needed for a 213 more complete understanding of these dynamic, reversing, and escalating exchanges.   analyses. An example dataset shows the data used for statistical analyses. In order to 357 apply paired statistical tests, only contests allowing five minutes or more (always equal) 358 windows of time before and after the establishment of social hierarchy were used unless 359 otherwise noted. In the example shown only contests with five minutes of recorded data before and after establishment were used (blue shaded box, above; median frequency 361 of behavior for corresponding period denoted as bold, blue line, below). In this example, 362 data recorded earlier than five minutes preceding and following establishment and also 363 the data from the entire first five and last two contests were excluded. The exclusions 364 occasionally decreased the sample size, and thereby the statistical significance, yet 365 never qualitatively changed results. 366  (B) and one-minute (C) windows prior to and following establishment are noted above 399 plots (horizontal, green lines). In all cases the Wilcoxon signed rank test was used as 400 described in Figure S2.    standard media and were incubated in the same conditions as they were reared. 494

Aggression assay 503
Unless noted, aggression assays were performed similar to those previously described 504 [25] modified from [14]. Briefly, pairs of males were aspirated simultaneously into a 16-505 mm diameter x 10-mm tall enclosed arena staged in an environmentally-control room 506 held at 25°C and 45% humidity and their behavior was recorded for 20 minutes. To 507 encourage consistent aggressive interactions, the entire floor of the arena was 508 composed of a thin layer of apple juice-sugar agar made as 10g sucrose and 9g agar 509 boiled in 400mL 100% apple juice (Mott's, Plano, TX). To keep the quality of the agar 510 floor consistent, it was used either immediately after a 2-hour setting period or air-dried 511 for 1 hour, wrapped inside of plastic (Saran, Racine, WI), and held at room temperature until the following day. To impede flies from climbing, the wall of the arena was made 513 slippery by coating it with Fluon (BioQuip, Rancho Dominguez, CA). Similarly, to limit 514 flies from hanging from the ceiling, the lid of the arena was brushed with Sigmacote 515 (Sigma-Aldrich, St. Louis, MO), a transparent silicone paint allowing an unobstructed 516 view for recording from above. Both coatings were left to dry for at least 24 hours before 517 running experiments. 518 519

Aggression screen 520
The genetically high-and low-aggressive males used came from a P-element screen 521 only lines that maintained a stable phenotype after outcrossing, had normative levels of 524 activity as estimated by measuring their total distance travelled and total number of 525 jumps, and appeared otherwise healthy were further used. A line, 5.116, exhibiting 526 normative levels of aggression and activity was also identified in the screen. This 527 "standard" was used for all principal experiments reported within this work. Table S2  528 includes measures of the aggressive acts from the various lines used within the 529 manuscript.

Behavior classification 543
To catalog the changes in social hierarchy and also the aggressive acts, we used a 544 combination of manual and automatic classification methods (see Figure S1 and Table  545 S1 for methods, descriptions, and software used). In summary, the establishment and 546 reversals of social dominance and also the aggressive acts, fence, box, and tussle, 547 were manually annotated with VCode [28] and we used CADABRA [13] to automatically 548 classify wing flick, lunge, and chase. To assign specific lunges to identified males we 549 used a semi-automated method previously reported [29]. This software application uses 550 as inputs both original full-length movies and corresponding time-stamped behavioral 551 events (generated with CADABRA [13] in our case) and iteratively makes short movies 552 inclusive of each event, allowing users to accurately assess if or which male executed a 553 particular lunge. For the P-element screen, pairs of males were tracked and 554 automatically scored for total number of wing flicks, tussles, lunges, chases, and jumps 555 using CADABRA [13]. 556