Scent mark signal investment predicts fight dynamics in house mice

Abstract

Signals mediate competitive interactions by allowing rival assessment, yet are often energetically expensive to produce. One of the key mechanisms maintaining signal reliability is social costs. While the social costs of over-signalling are well known, the social costs of under-signalling are underexplored, particularly for dynamic signals. In this study, we investigate a dynamic and olfactory-mediated signalling system that is ubiquitous among mammals: scent marking. Male house mice territorially scent mark their environment with metabolically costly urine marks. Competitive male mice are thought to deposit abundant scent marks in the environment. However, we recently identified a cohort of low-marking males that win fights. We hypothesized that there may be social costs imposed on individuals who under-invest in signalling. Here we find that scent mark investment predicts fight dynamics. Winning males that produce fewer scent marks prior to a fight engage in more intense fights that take longer to resolve. This effect appears to be driven by an unwillingness among losers to acquiesce to weakly signalling winners. We, therefore, find evidence for rival assessment of scent marks as well as social costs to under-signalling. This supports existing hypotheses for the importance of social punishment in maintaining optimal signalling equilibria. Our results further highlight the possibility of diverse signalling strategies in house mice.

1. Introduction

Signals of competitive ability play an important role in mediating rival assessment in aggressive contests [1–9]. However, signal production is often energetically expensive, and individuals face tradeoffs when investing in signalling effort relative to other life-history traits [10–12]. For example, increased signal investment can result in reduced gamete production [13–15], immune deficits [16,17] and higher risks of parasitism or predation [18–22].

In addition to production tradeoffs, there are social costs to signalling either too much or too little. Individuals that ‘over-signal’ their competitive ability receive heightened aggression from competitors [23–28], whereas individuals that ‘under-signal’ struggle to establish dominance relationships [26,28]. Such mismatches in signalled versus actual competitive ability muddle accurate rival assessment, resulting in escalated contests [26,28]. Receiver-dependent social punishment has been hypothesized as an important mechanism in maintaining optimal signalling equilibria [23]. While the social costs of over-signalling (i.e. ‘bluffing’ or ‘cheating’) have been well-examined, the social costs of under-signalling are under-studied, particularly for dynamic signals.

Here, we explore a dynamic and olfactory-mediated signalling system that is central to mammalian communication: scent marking [29–31]. Scent marks persist in the environment for long periods of time [32–35] and provide a record of social relationships that can be assessed by receivers [33,35–37]. Scent marks have been proposed as ‘cheat-proof’ signals of status owing to the metabolic and physical challenges of maintaining a scent-marked territory [35,36]. However, the intrinsic dynamism of scent marking complicates this ‘cheat-proof’ model. Individuals have the potential to flexibly adjust their signalling effort in response to the social environment, including strategic under-signalling. Despite the nearly ubiquitous nature of scent marking across mammalian species, we have a limited understanding of how susceptible scent marks are to inaccurate signalling.

In house mice (Mus musculus domesticus), urine marking is arguably the most prominent signalling modality. The generally accepted canon is that competitive males are aggressive, territorial and mark highly [38–41]. In addition to the costs of actively re-marking and patrolling a territory, urine marks themselves are metabolically costly in house mice [36,42–44]. Urine marking has previously been shown to have important life-history costs in house mice, as males that invest in marking earlier in life exhibit reduced body growth [42]. It is, therefore, generally assumed that urine marking is an honest indicator of a male's status and competitive ability [38–41,45–47]. Yet, we have recently tested this assumption and found it to be incomplete—urine marking prior to a contest did not predict wins or losses among size-matched rivals, in part owing to the presence of low-marking competitive males [48].

This surprising result led us to ask whether and how male house mice use scent mark information in competitor assessments. The objectives of this study were to: (1) test the hypothesis that scent mark signalling prior to a fight shapes contest dynamics, and (2) examine the potential social costs of under-signalling. We predicted that high-quality individuals that accurately signalled their competitive ability would quickly resolve their fights and therefore engage in less intense aggressive contests. By contrast, individuals that under-signalled their competitive ability would face the social costs of escalated aggressive encounters and experience delayed contest resolution.

2. Material and methods

(a) . Study system

To explore scent marking and aggressive behaviours we used male house mice (n = 62), as males will competitively urine mark and exhibit territorial aggression [33,39–41,46,49–51]. Experimental individuals were from two wild-derived inbred strains (NY2 and NY3) of house mice [52]. The progenitors of these strains were captured near Saratoga Springs, NY in 2013 by M.J.S. and are part of the same collecting effort that gave rise to the SarA/NachJ, SarB/NachJ and SarC/NachJ strains now available from the Jackson Lab. We note that these two lines represent genetically distinct families that are not close relatives to each other. We used two wild-derived strains because competitive behaviours are less pronounced in highly inbred and domesticated laboratory strains [53,54], and individuals within inbred strains tend to share identical urinary protein profiles [55]. At the time of experimentation all males were adult (3–5 months old) and sexually experienced. Mice were housed in an Animal Care facility at Cornell University with a 14 : 10 shifted light : dark cycle (dark cycle: 12.00–22.00), with food and water provided ad libitum. To reduce handling stress, mice were transferred between their home cage and the experimental arena using transfer cups [56].

(b) . Scent mark signalling and aggressive contests

In our previous work examining signal allocation decisions, we were surprised to find that scent-marking behaviour did not clearly predict wins or losses during fights and instead identified a cohort of low-signalling competitive males [48]. Together, these results led us to investigate the aggressive contests within this dataset in greater detail to better understand the relationship between signalling and competitive ability [48].

We placed males in an arena separated by a mesh barrier where they could see, hear and smell each other but were limited to minimal physical contact (figure 1a). This allowed us to measure male urine marking prior to a contest. After 30 min, we removed the mesh barrier and males engaged in a fight trial for an additional 30 min (figure 1a). Trials were performed on filter paper to prevent smearing of urine marks, for easier detection of urine deposition events. One day prior to experimentation we recorded male body weights to size-match individuals. As house mice are nocturnal, we conducted all experiments during the dark cycle between 12.00 and 17.00. Age- and weight-matched adult breeding males of distinct wild-derived strains (NY2 and NY3) were paired as competitors, resulting in a total of 31 pairs (n = 62). We, therefore, ensured that no two paired competitors were genotypically identical and that their scent marks were perceptibly different (i.e. characterized by unique major urinary protein profiles) [44,46,55,57,58]. We ear-clipped and bleached a patch of rump fur of one male in each pair a week prior to experimentation for easy identification of males within a pair (NY2 and NY3 strains are visibly indistinguishable).

Figure 1.

Trial design and recording methods. (a) Two-part trial design starting with a 30 min signalling trial where paired competitors were separated by a mesh barrier, and urine marking was measured. The mesh barrier was removed and males entered into the contest phase of the trial (fight trial) for an additional 30 min. (b) Urine depositions were recorded using thermal imaging. Urine exits the body hot and then cools below substrate temperatures, providing a distinct thermal signature. (c) For each fight trial, four aggressive behaviours were scored: wrestling, boxing, chases and hits. Wrestling and boxing were classified as intense attacks; chases and hits were classified as mild attacks.

All trials were recorded with a thermal imaging camera system (PI 640; Optris Infrared Sensing; 33° × 25° lens; approximately 3 Hz; thermal detection: 61–107°F (16–42°C)) and a security camera system (iDVR-PRO CMS; 1080 pixels; 30 frames s⁻¹). Thermal imaging allowed the detection of urine mark deposition events with fine spatio-temporal detail. Urine leaves the body hot (close to body temperature) and quickly cools to below the ambient substrate temperature, providing a distinctive thermal signature (figure 1b). The security camera system was used to visualize high-speed aggressive encounters. Both systems were used to cross-check for recording errors.

Videos were scored blindly using Behavioural Observation Research Interactive Software (BORIS) [59]. Urine depositions were scored as a clear hotspot in the focal mouse's trajectory that subsequently cooled below substrate temperature (figure 1b). Based on the total aggressive behaviours performed by each male, males were unambiguously classified as winners or losers (figure 2a and electronic supplementary material, figure S1). The following behaviours were scored in ‘fight trials’: chasing, hitting, boxing and wrestling bouts [60–62], based on which male initiated these behaviours (figure 1c and electronic supplementary material, figure S1A). Aggressive behaviours were further categorized as mild or intense based on the risk of injury (i.e. belly exposure and likelihood of bites occurring) and the extent of physical contact. Chases and hits were classified as mild attacks, while boxing and wrestling bouts were classified as intense attacks (figure 1c). Importantly, intense attacks are interactive behaviours, which require that the male receiving the attack actively defends itself rather than fleeing from the interaction. No mice experienced sustained injury in these trials.

Figure 2.

Winners displayed more aggressive behaviours throughout the fight trial, while losers rarely displayed any aggression after the first 5 min. (a) Total aggressive behaviours performed by males that either won or lost the fight. (b) Histogram of the temporal distribution of aggressive behaviours performed by winners and losers over the fight trial duration. (c) Total mild versus intense aggression displayed by winners and losers. A linear mixed model was used to model relationships (electronic supplementary material, table S1) and a type III analysis of variance was used to test for overall effects. The dependent variable (no. of aggressive behaviours) was logarithmically transformed to meet assumptions for model residuals, and male identity (ID) was treated as a random effect in the model to account for the paired data structure. Significance codes: n.s., p > 0.05, ***p < 0.001. (a,c) Boxplot midlines: medians; box limits: upper and lower quartiles; whiskers: 1.5× interquartile range; points: outliers. (d,e) Histograms of intense versus mild aggression exhibited by winners over the course of the fight trial.

(c) . Statistical analyses

We conducted all statistical analyses in R 3.6.0 [63]. We used linear mixed-effects models to examine relationships between dependent and response variables (tables S1–S4). Models were fitted using the package lme4 [64]. The lmerTest package was used to calculate degrees of freedom (Satterthwaite's method) and p-values [65]. We used a type 3 analysis of variance (ANOVA) to test for overall effects of fixed factors and interactions in the models. Models with aggression counts as the dependent variable were logarithmically transformed to meet assumptions for model residuals (tables S1–S4). All models with paired data include male ID as a random effect (tables S1–S4). Model and ANOVA output details are provided for each model in tables S1–S4. One male (NY3-325) was excluded from the proportion-based analyses in figures 3 and 4 (models: M4-M6; tables S2 and S4) owing to a 0 value (this male performed no intense attacks). This male was further identified as a low-end outlier (using the outliers package in R). R script and data sheets used for all statistical analyses are provided in the supplementary material.

Figure 3.

Temporal fight dynamics and intensity vary with the initial signalling effort of winning males. (a) Estimated marginal means plot of the contest aggression levels (log-transformed) given the initial marking effort (no. of urine marks) and attack intensity (mild versus intense) for winning males. (b) Estimated marginal means plot of contest aggression levels (log-transformed) by marking group, attack intensity, and 15 min time bins (first and second half of the fight trial) for winning males. (a,b) Linear mixed models (M4 and M5: electronic supplementary material, table S2) were used to assess relationships, and type III analyses of variance were used to test for overall effects (electronic supplementary material, table S2). The dependent variable (number of aggressive behaviours) was logarithmically transformed to meet assumptions for model residuals, and male identity (ID) was treated as a random effect in the model to account for the paired data structure. One male was excluded from these analyses owing to a zero value.

Figure 4.

Proportion of intense aggression with signalling effort among winners. The proportion of intense-to-total attacks is plotted against the total number of urine marks deposited prior to the fight for all winning males. The linear model is fitted to the data (M6: electronic supplementary material, table S4). The following groups of males are labelled for visualization purposes: intensely aggressive low-marking winners (orange) and mildly aggressive high-marking winners (turquoise). One male was excluded from this analysis owing to a zero value.

3. Results

(a) . Contest outcomes and aggressive intensity

Fight outcome was unambiguous in all contest pairings (electronic supplementary material, figure S1C). Winners performed more aggressive behaviours than losers (figure 2a). This was true for both cumulative aggression (figure 2a), as well as for specific fight behaviours (electronic supplementary material, figure S1B). Across all fight trials, winners performed 21–268 total attacks, while losers performed 0–19 (electronic supplementary material, figure S1C). Within pairs, the difference in attack count ranged from 21 to 266, with an average attack difference of 126 ± 11 between competitors. These winner–loser relationships were typically apparent within the first 5 min, as losing males quickly halt aggression (figure 2b). Winners, on the other hand, rapidly escalated aggression with peak activity occurring at approximately 300 s, followed by a gradual decline (figure 2b). Males thus performed fast competitor assessments once they physically engaged. While males were weight-matched as closely as possible, some variation was inevitable (electronic supplementary material, figure S2A). We did not, however, detect body weight differences between winning and losing males across trials (t_1,60 = 0.4, p = 0.69; electronic supplementary material, figure S2B).

We further explored the intensity of aggressive behaviours initiated by competitors during contests and found that aggressive intensity has a significant interaction with fight outcome (M1: F_1,60 = 34, p < 0.0001; figure 2c and electronic supplementary material, table S1). Losers exhibit similarly few mild and intense aggressive behaviours (M1: t_1,60 = 1.5, p = 0.4; figure 2c and electronic supplementary material, table S1). By contrast, winners perform significantly more mild attacks than intense ones (M1: t_1,60 = −6.8, p < 0.0001; figure 2c and electronic supplementary material, table S1). Body weight did not have a strong effect on the number of aggressive behaviours performed by individuals (M2: F_1,59 = 3.3, p = 0.08; electronic supplementary material, table S1), and including body weight resulted in a worse model overall (M1 versus M2: electronic supplementary material, table S1). Given the low rates of aggressive attacks performed by the eventual contest losers, we focused on the dynamics of aggressive behaviours initiated by the ultimate contest winners. Among winners, the temporal dynamics reveal that the number of intense attacks steadily declines over the course of the fight (figure 2d). Mild attacks, however, remain elevated for longer and decline in frequency more slowly (figure 2e). These data indicate that while fight outcome is straightforward, attack frequency varies with intensity. Furthermore, there appear to be distinct temporal patterns for mild and intense fight behaviours.

(b) . The social costs of under-signalling: fight intensity and resolution

We next explored the relationship between scent mark signalling and fight dynamics among the winning males, as these individuals initiated the vast majority of aggressive behaviours (figure 2). The total number of attacks performed by winners did not differ with their marking levels (M3: F_1,28 = 0.26, p = 0.6; electronic supplementary material, table S2). However, when attack intensity is included in the model, some striking patterns emerge. Attack intensity strongly predicts aggression levels (M4: F_1,56 = 7.5, p = 0.008; electronic supplementary material, table S2), as winning males overall perform more mild attacks relative to intense ones (figures 2c and 3a). There is also a significant interaction between fight intensity and initial marking effort (M4: F_1,56 = 5.6, p = 0.02; figure 3a and electronic supplementary material table S2), indicating that not all contests are won the same way. We find that winning males that marked highly prior to the contest engage in fewer intense attacks and more mild attacks, whereas males that marked lowly engage in more intense aggression (figure 3a). Therefore, the signalling effort of males predicts the aggressive intensity of a fight (figure 3a). Notably, winner aggression was not predicted by the marking levels of losing males, nor was loser aggression shaped by the signalling effort of either losers or winners (electronic supplementary material, table S3).

We further examined the temporal dynamics of these behaviours. To do this we split the fight trial into two 15 min time bins, corresponding to the first and second half of the trial (figure 3b). Time bin has a significant effect on contest aggression (M5: F_1,112 = 5.3, p = 0.02; electronic supplementary material, table S2), with a significant two-way interaction between fight intensity and time bin (M5: F_1,112 = 5.7, p = 0.02; electronic supplementary material, table S2). In the first half of the trial, low-marking winners perform similar levels of mild and intense attacks (figure 3b). By contrast, high-marking winners perform dramatically more mild attacks relative to intense attacks (figure 3b). This difference in fight intensity between marking groups diminishes in the second half of the trial, such that all winners display higher rates of mild compared with intense aggression (figure 3b). The fights of low-signalling winners thus exhibit more severe escalation in the first 15 min, suggesting males that invest less in signalling effort take longer to resolve aggressive contests than males investing more in signalling effort.

Together, these data indicate that the proportion of intense aggressive behaviours males perform varies with signalling effort. We examined the proportion of intense to overall attacks relative to the scent marking behaviour of each winning male prior to the contest (figure 4). The proportion of intense attacks is predicted by initial signalling effort (M6: F_1,27 = 4.7, p = 0.04; electronic supplementary material, table S4). This analysis illustrates a striking delineation between low- and high-signalling competitive males (figure 4). Furthermore, it reveals what appear to be two distinct groups of males that each compose nearly a quarter of all winners: (1) intensely aggressive low-marking males and (2) mildly aggressive very high-marking males (figure 4).

4. Discussion

Here we have shown that despite fight outcome being overwhelmingly clear (figure 2), there are stark differences in contest dynamics associated with how males signalled prior to a fight. We find evidence for social costs to under-signalling in house mice, as low-marking winners experienced more intense fights and delayed contest resolution (figures 3 and 4). This suggests there are likely important tradeoffs underlying signal investment decisions in terms of competitor assessment and aggression. This finding is particularly relevant since in our prior work we identified a cohort of competitive yet stably low-marking male mice [48]. This study highlights the complex decisions animals face when determining their signal investment and willingness to engage in aggressive encounters. At any given moment, individuals confront metabolic resource limitations. Deciding when and where to invest these resources has important fitness consequences.

The observed differences in aggressive intensity among competitive (i.e. winning) males could be driven by the winners themselves, their losing counterparts, or a combination of the two. A winner-driven explanation is that winners allocate more effort toward aggression rather than signalling, such that the total energy invested is constant across lower- versus higher-marking winners. Alternatively, losing males may be less inclined to back down during attacks initiated by weakly signalling males. Importantly, intense aggressive behaviours (i.e. wrestling and boxing) are highly interactive and require that losers actively defend themselves rather than flee from the encounter (i.e. chases). This lends support for the observed differences being loser-driven and alludes to a possible key difference between losing an encounter and submitting to a competitor.

The trial design limits males' ability to escape the interaction. In natural environments males would likely avoid prolonged encounters [66], and dominance relationships would be established through shorter repeated interactions. Nevertheless, the sustained encounters used in this study allowed us to observe temporal shifts in fight intensity. We find that low-signalling winners do not transition to more mild aggression until midway through the fight, whereas high-signalling winners start off relatively mildly. Fight resolution is therefore delayed when winners signal lowly prior to a fight. Similar to what has been observed in aggressive contests between paper wasps [26] and chameleons [28], we find that male house mice experience social costs as a result of inaccurately signalling their competitive ability. This is striking because the costs of under-signalling appear relatively high. ‘Scent-silent’ males engage in more intense fights, take longer to resolve dominance relationships, and incur greater risk of injury. More precise measures of the nature and magnitude of social costs, beyond altered fight dynamics in mice and other taxa, will be an important area for development in future studies.

Given the potential social costs, the existence of competitive low-signalling males suggests there may be fitness benefits to remaining scent-silent, at least under some socioecological conditions. Perhaps the most obvious benefit to reduced signalling is that it saves energy. Males might withhold signal investment to build up their metabolic reserves, as urine marking is energetically expensive [42,43,47,67]. In doing so, males may gain body mass and more effectively defend territories later in the season. This is plausible given that prior work has shown males that invest early in urine marking pay the cost of reduced body size [42]. The low signalling effort observed among competitive males could therefore reflect important features of life history in house mice, and potentially in many other species.

Males entering into the trials had no prior competitive experience. Males may strategically hold off scent mark investment until there are rival males present, suggesting population density may have large effects on signalling strategies [68–70]. This is particularly intriguing given prior hypotheses of urine marking as ‘cheat-proof’ [35,36]. These hypotheses emphasize the inability to deceptively over-report or bluff one's competitive ability but do not address the possibility of males under-reporting. Our results highlight the importance of investigating under-signalling strategies, as they may be more common than previously appreciated across taxa.

Another possibility is that male mice exhibit a spectrum of signalling strategies, including the classically described ‘territorial males’ that invest highly in marking as well as scent-silent ‘satellite’ males. In this scenario, low-signalling individuals might avoid detection by other males yet are competitive enough to mate, though reduced marking effort likely decreases the chances of obtaining mating opportunities [33,71,72]. Previous work described males reducing their scent marking after losing [41]. Indeed, in the trials reported here, we found that high-marking males dramatically reduce their marking efforts after losing [48]. However, this scenario does not readily explain the observed patterns of winning males that continue to mark infrequently. It may be that in time they would increase investment in scent marking, and shifting to high-marking is a slower process than downregulating after a loss [43,47]. Studies of scent-marking effort in more natural contexts are sorely needed.

We find evidence for direct social costs as a result of under-signalling one's competitive ability. This supports existing theoretical frameworks underlying the importance of social punishment in shaping patterns of signal investment [23,25,26]. Our results further provide evidence that urine marking is used in competitor assessments and appears to determine losers' willingness to submit. Furthermore, our findings highlight the possibility of diverse signalling strategies in house mice. As a dynamic signalling system, individuals may flexibly adjust scent mark investment depending on the social landscape and their energetic reserves. Diverse strategies may be more commonplace for dynamic signals across taxa than is currently recognized and warrants further investigation.

Ethics

All experimental protocols conducted at Cornell University were approved by the Institutional Animal Care and Use Committee (IACUC: Protocol no. 2015-0060) and were in compliance with the NIH Guide for Care and Use of Animals.

Data accessibility

Data sheets and R code used in all analyses are provided in the electronic supplementary material [73].

Authors' contributions

C.H.M.: conceptualization, data curation, formal analysis, investigation, visualization, methodology, supervision, writing—original draft, writing—review and editing; K.H.: data curation, methodology; M.F.H.: data curation, methodology; T.M.R: visualization; M.J.S.: supervision, funding acquisition, project administration, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed herein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This research was funded by a USDA Hatch Grant (NYC-191428; M.J.S.). The funders were not involved in the design of the study, the collection, analysis and interpretation of data, the writing of the manuscript or any decision concerning the publication of the paper.