A behavioral syndrome of competitiveness in a non-social rodent

Animals compete for limited resources and the outcome of intraspecific competition should be determined by individual variation in behavioral traits, such as aggressiveness, and dominance status. Consistent among-individual differences in behavior likely contribute to competitiveness and predispose individuals to acquire specific dominance ranks. In a step towards better understanding these functional links, we studied trait integration into behavioral syndromes, using 26 captive male bank voles (Myodes glareolus). We repeatedly assessed boldness in an emergence test, exploration in an open-field test, aggressiveness in staged dyadic encounters, and the among-individual correlations between these behaviors. We further related these personality traits to dominance rank, from quantifying urine marking value (UMV), as marking in bank voles is related to dominance rank. We found repeatable variations in boldness, exploration, aggressiveness, and UMV, which were correlated at the among-individual level. Aggressiveness tended to be negatively correlated with body condition, a proxy for fitness. Thus, key personality traits and social rank are functionally integrated into a behavioral syndrome of intraspecific competitiveness.

Methods
Study animals

We conducted this study on captive animals from July-February of 2021 and 2022 in the animal facilities of the Animal Ecology Group in Potsdam (Germany). Study animals were male bank voles (Myodes glareolus) captured from the wild in a small forest patch near Potsdam (Germany)(n = 20) or born in captivity (n = 6) as F1 of wild-caught parents. Prior to the experiments, all animals spent 3 to 5 months in the laboratory and were sexually mature. Voles were housed in standard polycarbonate cages (type III and type IV) provided with wood shavings and hay as bedding, cardboard rolls as shelter, and a running disk for enrichment. Water and food pellets (Ssniff, NM, V 1244-0) were available ad libitum. 

Personality assessment

Emergence test and open-field test

We tested 26 males twice in two established behavioral tests to quantify among-individual differences in activity, exploration, and boldness (Mazza et al 2019, Schirmer et al 2019). A circular arena (100 cm diameter surrounded by 35 cm high walls) with a white floor connected with an opaque tube (11 x 30 cm) with a sliding door was used in both tests. By drawing on the floor, the arena was visually divided into a circular middle area (70 cm in diameter) where the vole was fully exposed and a border area (15 cm wide) by the wall, representing a safer area. Each area was additionally visually divided into eight sections.

We quantified boldness-related behaviors via an emergence test, also named dark-light test. The individual was placed in the tube for one minute of habituation, after which the door to the arena was opened by pulling a string from the neighboring room. Based on direct observation (or via video analysis), one observer (FE) measured two behavioral variables: the time to emerge with the head including ears (latency head) and the latency to emerge with the full body excluding the tail (latency body). We quantified activity- and exploration-related behaviors via open-field tests. In this test, individuals were placed in an inverted beaker at the edge of the arena and left to habituate for one minute. Thereafter the beaker was lifted from the neighboring room and the individual was observed for 10 minutes. We recorded the activity (running, jumping, scanning, grooming) instantaneously every 10 seconds, latency to enter the center part of the arena with the full body excluding the tail (latency center), number of crossings of the full body excluding the tail into the middle part (crossings), and the number of different sections entered with its full body excluding the tail (sections).

Staged dyadic encounters (SDE)

We ran staged dyadic encounters (SDE) between various dyads formed by the 26 male bank voles, to measure among-individual variation in behavior towards a conspecific. A total of 26 took part in three SDE against three different opponents, except for one male for which only two encounters (with two different opponents) could be performed. To create dyads, we matched individuals of similar weight allowing a maximal weight difference of 6 g (corresponding to ≤ 25% of the body mass).

SDEs were performed in a rectangle box (56.5 x 36.5 x 59.5 cm) divided by a wall into two similar-sized compartments (27 x 36.5 x 59.5 cm). The dividing wall had a wire grid and a sliding door at the bottom to allow visual, acoustic, and olfactory contact between the two compartments after the sliding door was pulled up (see Figure SI1 in online resource 1). To simplify the video analysis of the placement of voles in the arena, we divided the area into three equally big zones (with zone one closest to the divider, zone two in the middle, and zone three furthest away from the divider) with lines on a paper placed at the bottom of the arena. Animals were weighed immediately before testing. We placed dyads in the setup for one minute of habituation with the divider down. The arena was then moved to the test room for another five minutes of habituation. The observer then raised the sliding door from the neighboring room and the voles were left to have visual, acoustic, and olfactory contact for 30 minutes. 

One observer (FE) quantified movements of the focal vole between the three zones of the arena, where movements towards the divider and the opponent vole were defined as ‘approaches’ (Figure SI1.A1-A4 in online resource 1) and movements away from the divider and the opponent as ‘departures’ (Figure SI1.D1-D4 in online resource 1). We quantified, approaches and departures by counting crossings of the focal vole between zones with the whole body excluding the tail. For each movement between the zones, we additionally specified if the vole moved to or from the side of the arena in which the opponent vole was located (left or right of the vertical middle), i.e. if the vole crossed from one side over the arena, over the midline, to the other side of the arena with its full body (excluding the tail). For each of these crossings, we also noted down the zone position of the opponent vole. Furthermore, we quantified the time spent with its whole body (excluding the tail) in the section closest to the divider (‘zone one seeker’) and time spent in this section while the opponent also occupies zone one in his compartment (‘opponent seeker’). We combined variables into fewer variables explaining departures and approaches to opponen, including four variables of number of approaches, four variables of number of departures, and one variable for all crossings (‘totcross’). The four approach variables were: A1) number of all approaches to zone one (closest to the divider) disregarding opponent position, A2) number of approaches to opponent (to zone one when the opponent is on the same side and approaches to the same side as the opponent within zone one), A3) number of approaches to zone one only when the opponent also is in zone one, A4) number of approaches to opponent (to zone one when the opponent is in the same side of zone one and approaches within zone one to the side the opponent vole is located in zone one). The four departure variables are the opposite movement (i.e. from zone one or opponent) to the above-described approach movements (from here on referred to as ‘D1’, ‘D2’, ‘D3’, ‘D4’). See figure SI2 in online resource 1 for a visual demonstration on how the variables were combined.

No observer was present in the room of the setup during the emergence test, open-field, and SDE. All tests were recorded using cameras (ABUS, Mini Dome Camera HDCC35560) and analyzed half blinded to the vole identity after tests had been conducted.

Urine marking value

In addition to the SDE described above, we ran a series of repeated paired trials to asses urine making behavior of 24 males. Males were tested in the same dyads as in the behavioral SDE test. This setup consisted of two boxes (27.5 x 40 x 19 cm) attached side by side without a floor. The joint walls were perforated with small holes (ca. 0.5 cm diameter) to allow olfactory but only limited visual contact between the voles in each box. Filter paper was placed under the boxes to capture urine markings. We placed dyads in the setup, and left them to interact for two hours. Using UV light, we evaluated the filter paper from each vole by dividing it into 5 cm x 5 cm squares (35 in total) and counted how many of these squares were marked with a few large blots of marking (typical for subordinate status) or with many squiggly lines (typical for dominant status). An individual only marks typical for subordinate status or dominant status. We calculated a urine marking value (UMV) expressed as the proportion of the number of squares marked with urine markings of any style, subordinate or dominant urine marking patterns, divided by the total number of squares. We used the original quantitative variable of the proportion of squares marked.

After each behavioral test (emergence, open-field, SDE, UMV assessment), the animal was transferred back to its home cage, and the test arenas were cleaned with ethanol. Each vole was exposed to any of the setups only once per day. 

Body condition

At the end of the dyadic encounters, one observer (FE) measured the head width (mm) of all animals and calculated body condition using the scaled body mass index suggested by Peig and Green (2009). 

Statistical analyses

To analyze among-individual correlations between behaviors expressed in the different tests we first estimated repeatability of all behavioral variables. Due to it's biomodal distribution, we first transformed the time spent in zone one in the SDE into a binomial variable using the function “cutoff” from the package “cutoff” (Choisy, 2015). We log-transformed all approaches, departures, and total crossings from the SDE as well as latency to emerge head and body from the emergence test, and arcsine square root transformed the proportion spent active in the open-field test and proportion of squares with urine marking (UMV). We estimated repeatability using the rpt function from the package rptR (Nakagawa and Schielzeth, 2013; Stoffel et al., 2017), using 1000 simulations to estimate confidence intervals and 1000 permutations to estimate p-values. We used Gaussian family distribution for variables from the SDE which were zero-inflated.

Second, we reduced repeatable variables in separate principal component analyses (PCA) for each test (emergence, open-field, and SDE) with oblimin rotation for the open-field test and SDE. PCA assumptions were checked by inspecting the correlation matrix, Bartlett test, and Kaiser-Mayer-Olkin (KMO) criterion (Field et al., 2012). We included two variables for the PCA of the emergence test, three variables in the PCA for the open-field test, and two variables in the PCA of the SDE (Table SI2). If one of these behavioral variables had a better distribution than both the principal component and the other behavioral variable, we used this original values for in the subsequent bivariate mixed models. We only included components from the PCA with Eigenvalues > 1 in subsequent bivariate mixed models (BMMs).

Third, to estimate the fixed effects we ran univariate linear mixed effect models using the lmer function from the lme4 package (Bates et al., 2015) or generalized mixed effect models using the glmmTMB function from the glmmTMB package (Brooks et al., 2017) for each response variable with individual identity as a random effect. We evaluated the effect of controlling for test occasion (i.e. first, second or third time tested) on all response variables, and for body condition on the composite variable from the SDE. Due to the large percentage of animals (38%) that did not move in the SDE, we additionally ran a GLMM on a binomial variable of ‘aggressiveness’, and an LMM on a subset of data with only individuals that move as well as controlling for zero-inflation to further assess this relationship.

Lastly, to investigate among-individual correlations between pairs of variables we ran bivariate mixed models (BMMs), using the ‘mcmcglmm’ function from the MCMCglmm package (Hadfield, 2010; following procedures described in Dingemanse and Dochtermann, 2013 and Houslay and Wilson, 2017). We ran six BMMs to evaluate pair-wise correlations between two composite variables obtained from the PCAs, crossings, and UMVs. All models included individual identity as a random effect and variable-specific fixed effects (test occasion for crossings and body condition and test occasion for component from SDE). To retain all data in the SDE, we used the observations of both individuals in a dyad as independent observation. We fixed the within-individual variance to 0 because we did not estimate all variables at the same time. We ran all models with different priors as suggested by Hadfield 2010 (two informative, two non-informative priors). Reported model results are based on a flat uninformative prior (V=diag(2), nu=1.002). We used 250,000 iterations, a thinning interval of 100, and a burn-in of 50,000. Finally, using the posterior distributions from the bivariate mixed models, we calculated repeatability for each dependent variable, pairwise among-individual correlations, and their credibility intervals based on Houslay and Wilson (2017).