The effects of challenge or social buffering on cortisol, testosterone, and antler growth in captive red deer (Cervus … – Nature.com

Observation of farmed red deer took place in a deer facility belonging to the Institute of Animal Science (V...V.) at Podlesek, Praha, Czech Republic (500302.2"N 143537.1"E).

Setting up a classical control is a problem in a study like this. Without it, we could not fully distinguish whether the social environment causes differences in hormone levels and antler growth or whether specific individuals (with particular hormonal levels and antler growth characteristics) are more or less likely to aggregate with others. Without a full control, it is not possible to discount the hypothesis that internal characteristics of individual males could influence our results. We have previously shown that antler size has high repeatability in subsequent seasons during ontogeny56. Moreover, even though the males had shown behavioural plasticity, their individual attitude to seeking or avoiding interaction had been maintained despite the modifications of the social environment49.

In addition to our main experimental season (Season 1), we also had data on hormones and antler sizes from the following two seasons (Season 2 and Season 3). We, therefore, presumed that if the cause of the results were based on characteristics which were individual-specific, then the hormonal and antler values recorded for individual males in this study in Season 1 should correlate with those of the same individuals in the following season or seasons. On the other hand, if the hormonal and antler characteristics displayed by individual males in Season 1 were primarily a consequence of the males social tactics, the repeatability of hormonal and antler values between Season 1 and Season 2 should be lower than that between Season 2 and Season 3, when the males lived in the same environmental conditions.

Seventeen semi-tame red deer males (one male aged 9, seven aged 6, six aged 3, and three aged 2) belonging to the same bachelor group since birth were available at the beginning of the observation at the facility, within an area of approximately 4ha. This area was divided into six enclosures. Each enclosure (about 0.7ha large) contained a shelter (a wooden, roofed barn with one side permanently open with the entrance of approximately 24 m2), a water reservoir, and a mud pool for wallowing. During the main observation period (from 17th April to 28th August in Season 1, the period of antler growth), all enclosures were interconnected by two (in the first and last enclosures) or three permanently opened gates in other enclosures allowing the deer to move around and aggregate with or separate from others. In contrast, for control purposes, in Seasons 2 and 3, the same males were kept in three interconnected enclosures, each 0. 7ha in size, i. e. their living space in Seasons 2 and 3 represented about 50% of the area in Season 1. The animals fed predominantly on the natural pasture of the enclosures and were supplemented with hay (ad libitum) and occasionally also with potatoes, beets, apples, pears, barley and oats. The animals were identified with coloured, numbered collars and in Season 1 equipped also with GPS collars (Lotek Wireless Inc. GPS_3300, position readings with an error of less than 5m). When this study ceased, all animals stayed at the facility for future investigations.

According to European and Czech laws, the experimental deer facility is an accredited research centre for the ethical use of animals in research (60444/2011- MZE-17214). The experimental proposal no. MZe 1297 was approved by the Animal Care and Use Committee at the Ministry of Agriculture of the Czech Republic. We confirm that all methods were carried out following relevant guidelines and regulations and are reported in accordance with ARRIVE guidelines.

In Season 1, observations were designed to record agonistic interactions between animals when competing for supplemental food. At the time of observation, deer were fed a mixture of soya, barley, oats and a mineral/vitamin premix, which amounted to an average of 0.7kg/day/animal. When the supplemental food was presented, it usually attracted all males regardless of whether they were otherwise in groups or associated individually in the paddocks. Thus, at the time of provision of the supplementary food, all the males met together even if otherwise they preferred to avoid encounters with other individuals.

The food was carried to the observation place in a wheelbarrow and presented in several piles to encourage mild competition over a scarce resource (serious competition was prevented to preserve good welfare of the stags). The food piles were tipped from the wheelbarrow in 8 or 9 piles about 2m apart, in order to encourage competition without exacerbating it. This method has already been proven valid in previous studies e. g.,15,49, etc. Each observation session took place in the morning (between 9.00 a.m. and 11.30 a.m.) and ranged from 20 to 60min (depending how long the deer stayed at the site of supplementary feeding). In Seasons 1, 2, and 3, observations took place from 1 to 5 times per week between 1st May and 28th August (with an average equal to 3). In total, the deer interactions were observed for 37h in Season 1, 30h in Season 2, and 15h in Season 3. All deer were semi-tame and started to compete over the food as soon as it had been presented, running from one pile to another trying to eat as much as possible. When a feeding deer was challenged by others, it either escaped to another pile or defended itself. All the observations were made into a voice recorder and then transcribed into a table using Microsoft Excel. We recorded any occurrence of an approach of one male to another, any attack, threat gesture, or fighting, which caused an apparent displacement of the approached individual49,59. As in previous studies reviewed in55, we determined dominance status for each pair of males on the basis of the agonistic interactions observed. We regarded as dominant the males who won more agonistic encounters than they lost in any dyad, as subordinate the ones who lost more often than they won within the dyad, and as indifferent the males in a dyad with no agonistic interactions.

GPS collars measured inter-individual distances between males in Season 1 only. Positions were programmed to be recorded once per hour. This enabled us to obtain records of inter-individual distances during the observation period with an average of 90.404.6m (meanSE) per dyad (n=272) over the observation period, producing a reliable picture of mean inter-individual spaces whole period. In Seasons 2 and 3, we did not use GPS collars and made no detailed spatial observations as done in Season 1.

In all three study years, we weighed the males once a month (5 times between April and August), collected blood samples for the hormone analysis in a physical restraining facility (crush). All deer involved were used to this procedure and had undergone it since birth. No chemical restraint was used besides physical restraint. In Season 1, when collecting blood samples, we downloaded GPS records from data loggers. In all seasons, we measured the antlers after casting, as previously described e. g.,60 and used the total antler length, the final sum of the length of all tines, points and beams divided by 261, as a dependent variable.

Analyses of T and C concentrations were performed in the laboratories of ELISA development, s.r.o. (Velk ernoseky, the Czech Republic). T concentration was measured by RIA Kit from Beckam Coulter (code IM1087). T antibody for this RIA Kit is species-nonspecific. The radioimmunoassay of T is a competitive assay. Before the assay, plasma samples were extracted with ethyl ether; the solvent was evaporated, and the dry residues were re-dissolved in the recovery buffer of the kit. The re-dissolved extracts and calibrators were then incubated with 125I-labeled T, as a tracer, in an antibody-coated tube. The concentration range was up to 23ng/mL, the assay's detection limit was 0.1ng/mL, intra-assay-precision was 8.6%, and inter-assay was 11.9%. The recovery of the extraction step was 90%.

C concentration was determined by RIA Kit from Beckman Coulter (code IM1841) previously validated only in cattle62,63. C antibody for this RIA Kit is also species-nonspecific. The radioimmunoassay of C is a competitive assay. Samples and calibrators were incubated in monoclonal antibody-coated tubes with 125I-labeled cortisol tracer. The concentration range was up to 2000nM, the assay's detection limit was 5nM, intra-assay-precision was 9.4%, and inter-assay was 12.6%.

All data were analysed with the aid of the SAS System (SAS, version 9.4).

Previous studies have shown that it is essential in assessing relationships between social behaviour and physiology to record and analyse measured characteristics in as much detail as possible e. g.,6. Therefore, we preferred to analyse the inter-individual pairwise relationships rather than rely upon any form of summarized values.

(i) In the main observation period (Season 1), for each male, we collected for each observation the inter-individual mean distance (meters) from each of the herd mates (i.e., 16 inter-individual distances per male). A cluster analysis (PROC CLUSTER, with TYPE=NOMINAL and METHOD=HIERARCHICAL) was used to divide the mean inter-individual distances resulting into two groups, Associates (males keeping together) and Distant (those living apart). According to their involvement in interactions during the feeding competition, these latter (distant) dyads were further subdivided as Indifferent (i.e., no interaction within the dyad was recorded), or Non-associates (i. e., dyads keeping mutual distance, interacting during the feeding competition only). In conclusion, three levels of mutual relationship between the individual stags were considered: Associates (keeping together), Non-associates (keeping distance but interacting when meeting during the feeding) and Indifferent (keeping distance, non-interacting). For each male, we then calculated the "Proportion of Associates" of all dyadic relationships (% of individuals from all 16 possible dyadic groups who were identified as Associates of the focal individual) and the Proportion of Non-Associates relationships (% of individuals identified as Non-Associates within any dyadic group). At this point, however, it should be pointed out that the classification of Associates, Non-Associates and Indifferent concerns dyadic distances, not the categorization of males. Thus, each individual could be Associate with one male, Non-Associate with another male, and Indifferent with other males. It, therefore, depended on whom the focal individual had interactions with, and which conspecifics preferred more than others.

For each subject we had available also other characteristics of interactions (listed in Table 2), between him and all other males such as the number of attacks, wins, losses, etc. For the analysis, we used the mean values of all quantifiable variables over the whole period for each subject and all its dyads. Having 17 males with 16 relationships each, we obtained 272 dyadic records in total.

To check for possible multicollinearity, we first calculated correlations (PROC CORR) between the individual metrics involved (Table 2). Significant correlation was found between the Bodyweight at the beginning of the observation and at the end of the entire experimental period (MayAugust; r=0.91, P<0.0001), between Bodyweight and Weight gain (r=0.83, P<0.0001), between Age and Bodyweight (at the beginning of the observation r=0.84, P<0.0001; and at the end of the observation r=0.71, P<0.0001. We subsequently made a judgment of the extent of collinearity by checking related statistics, such as Tolerance value, Variance Inflation Factor (VIF), Eigenvalue, and Condition Number and using TOL, VIF and COLLIN options of the MODEL statement in the SAS REG procedure. We discovered apparent collinearity between all variables characterizing agonistic interactions (i. e., Sum of all agonistic interactions of any given type, Wins, Losses, and Number of attacked conspecifics). When either of these characteristics entered the REG procedure alone, the lowest tolerance value did not drop below 0.13. The highest variance inflation value did not exceed the value of 7.5. Also, there was no case of small eigenvalues combined with the large corresponding condition number. So, there was no threat of other multicollinearity indicated by these results.

Across the models, where appropriate, count variables were log-transformed (natural logarithm transformation) to improve the normality of residuals and to reduce skewness.

Since the issues analysed in this study represent more complex causality, we used the information-theoretic approach (IT-AIC) for estimating the effects of the factors on dependent variables64.

Associations were subsequently sought between C concentrations (ii), T concentrations (iii), or total antler length as dependent variables and the remaining fixed factors (Table 2) using a multivariate General Linear Mixed Model (GLMM, PROC MIXED). To account for the repeated measures on the same individuals, all analyses were performed using PROC MIXED with ID of the individual male as a random effect. For each dependent variable, we constructed a set of multiple a priori hypotheses and added a Null model. Where appropriate, we included interaction terms. Specifically, for log-transformed C concentrations, we set up 38 alternative hypotheses, for log-transformed T concentrations 26 hypotheses, and for Total antler length 90 hypotheses (Supplementary Table S4). For each dependent variable (i.e., C, T, and total antler length) we generated all GLMMs in the Supplementary Table S4 and converted values of fit statistics.

Since the introduction of Akaike's Information Criterion (AIC), more information criteria have been developed with differing mathematical properties and philosophies of model selection65. We used expanded information criteria AIC, AICC, BIC, CAIC, and HQIC to select a true model, as recommended by Christensen65. Then we compared the candidate models by ranking them based on the information criteria being used (PROC RANK). The model with the lowest value (i. e. closest to zero) is considered to be the "best" model64,65. To see if the best model has merit, we compared our model to the null model for all dependent variables and all fitting criteria, showing delta (null best model) and a relative information loss [exp((nullbest)/2)], an approach adapted from Burnham and Anderson64.

The differences (i) between the Fit statistic values (the smallest values indicating the best fitting model) were sorted according to AIC values. Akaike weight wi can be interpreted as the probability that Mi is the best model (in the AIC sense, that it minimizes the KullbackLeibler discrepancy), given the data and the set of candidate models e. g.,64. For five models with the lowest AIC values, we therefore calculated AIC, Akaike weights wi, and for estimating the strength of evidence in favour of one model over the other we divided their Akaike weights wmin/wj (AIC Odds)64.

Associations between the dependent variable and countable fixed effects are presented by fitting a random coefficient model using GLMM as described by Tao et al.66. We calculated predicted values of the dependent variable and plotted them against the fixed effects with predicted regression lines.

Several statistical methods are typically used to show comparability or repeatability67. As previously49, we chose Lins concordance correlation coefficient68 using the SAS macro described by67 and Kendalls tau-b correlation coefficient to estimate a measure of association of the hormonal and antler measures of the same subjects between Season 1 Season 2, and between Season 2 Season 3. For computing Kendalls correlation coefficients and its confidence interval estimation, we applied macro by Looney69. From the Seasons 1 to 3, males group consisted of the same individuals. Decreasing N on comparisons between seasons (Table 4) reflected that some males were removed from the facility for other purposes (four in Season 2 and one in Season 3).

Go here to see the original:
The effects of challenge or social buffering on cortisol, testosterone, and antler growth in captive red deer (Cervus ... - Nature.com