Electronic supplementary material to:
When to initiate torpor use? Food availability times the transition to winter phenotype in a tropical heterotherm
Pauline Vuarin 1, Melanie Dammhahn 2, Peter M. Kappeler 2,3 and Pierre-Yves Henry 1
1 Mécanismes adaptatifs et Evolution (MECADEV UMR 7179), Sorbonne Universités, Muséum National d'Histoire Naturelle, Centre National de la Recherche Scientifique, 1 avenue du Petit Château, 91800 Brunoy, France
2 Behavioral Ecology & Sociobiology Unit, German Primate Centre, Leibniz Institute for Primate Research, Kellnerweg 4, 37077 Göttingen, Germany
3 Department for Sociobiology/Anthropology, Johann-Friedrich-Blumenbach Institute for Zoology & Anthropology, University of Göttingen, Kellnerweg 6, 37077 Göttingen, Germany
Present address:
M. Dammhahn
Animal Ecology, Institute for Biochemistry and Biology, University of Potsdam, Maulbeerallee 1, 14469 Potsdam, Germany
Corresponding author: P.-Y. Henry
Mécanismes adaptatifs et Evolution (MECADEV UMR 7179), Sorbonne Universités, Muséum National d'Histoire Naturelle, Centre National de la Recherche Scientifique, 1 avenue du petit Château, 91800 Brunoy, France
e-mail:
ESM1. Test for differences in ecological and physiological conditions of grey mouse lemurs between the two plots
Objective
Our experimental design was based on the comparison of torpor patterns between a control plot (locally known as N5) and a plot with food supplementation (locally known as CS5). To evaluate if differences between the two plots may confound the effect of food supplementation in the treatment plot, we tested for differences in ecological (population density, vegetation structure, food availability, air temperature) and physiological (grey mouse lemurs’ body constitution, torpor use) conditions between the two plots.
Materials and methods
Population density. Densities were computed from trapping data based on minimum number of alive animals (M. Dammhahn, unpublished data).
Vegetation structure. In each plot, 15 sampling points, located at the north-east corner of crossing between trails, were randomly picked. For each sampling point, four indices of vegetation structure were collected. Undergrowth density was scored using a 40 x 40 cm checkerboard divided into 16 black-and-white squares. The checkerboard was hold at two meters from the observer, at a height of 1.5 m. The undergrowth density score was the number of squares that were at least half hidden by vegetation. Following the point-quarter sampling method (Ganzhorn et al. 2003), three indices of tree (i.e., trunk diameter at breast height, DBH, of 5 cm or more) density were taken: the distance to the closest living tree was indicative of the density of the forest, the distance to the closest dead tree with DBH ≥ 15 cm was indicative of shelter density for grey mouse lemurs (Schmid 1998) and the DBH of the closest tree was indicative of the average age of trees. All indices were taken four times per sampling point (i.e., once per quarter), and averaged to produce one value per sampling point.
Food availability. Insect availability was used as a proxy for food availability. It was measured with two Malaise traps (Bioform, Germany, bidirectional surface of ca. 1.5 m²) (after Townes 1962), one per plot, which were set for one week at the cross of two small trails (n = 8 samples in the dry season, n = 6 samples in the wet season per plot). Samples were collected twice per month. We took all samples to the research station, where we identified insects to order, counted them and assigned them to size classes (Kunz 1988). Because we kept samples in ethanol for further taxonomic classification, we calculated dry weight from length, using a power function for all adult insects as weight[mg]=b0+(length[mm])^a with Ln(b0) = 3.071 and a = 2.2968 (after Ganihar 1997). Hence we obtained an index of total insect dry biomass for each sample.
Air temperature. Air temperature was recorded using temperature loggers (iButtons, Maxim Integrated, USA) every 20 min throughout the study period, following the methodology described in the main manuscript. Eight temperature loggers were set per plot: four were fixed to a trunk to record outside air temperature and four were placed inside a tree hole to record air temperature in typical rest shelters of grey mouse lemurs. Mean (Ta mean) and minimal (Ta min) values were averaged per iButton.
Body constitution of grey mouse lemurs. For all captured grey mouse lemurs, body size was indexed by head width, and body condition was calculated from body mass and head width (as in Vuarin et al. 2013).
Torpor parameters before the experiment. The three torpor parameters (Ptorp: probability of individuals entering torpor, Dtorp: torpor bout duration and Tsk min: minimal skin temperature) are the same variables as those described in the main manuscript (see material and methods).
Statistical analyses. Vegetation structure, air temperature and body constitution parameters were analysed with linear models including the fixed effect of the plot (control versus experimental). The model for food availability also included the effect of the season (dry versus wet). Torpor parameters were analysed with the statistical models described in the main manuscript.
Results
Undergrowth was 1.4 times denser in the control than in the experimental plot (Table S1; see also Schwab & Ganzhorn 2004). None of the other documented ecological and physiological parameters differed between the two plots, but body mass and body condition (Table S1). Although grey mouse lemurs had the same size in both plots, they were heavier and in better condition in the control plot than in the experimental plot (Table S1).
Conclusion
Only two parameters differed between the two plots: undergrowth density and grey mouse lemurs body condition. Since torpor use depends on body condition (Vuarin et al. 2013), we had to assess if our results on the effect of food supplementation on torpor use could be confounded by differences in body condition at the start of the experiment (see ESM 2). In absence of a confounding effect of initial differences in body condition, no other major ecological difference between the two plots can be suspected to confound the effect of food supplementation.
Table S1. Comparison of the control and food supplemented forest plots regarding population density, vegetation structure, food availability, air temperature, grey mouse lemurs’ body constitution and torpor parameters before the beginning of the food supplementation experiment.
* Even though 7 individuals were equipped with collar-mounted temperature loggers in the food supplemented plot before the experiment, only one did enter torpor, so that Dtorp values are averaged from a single individual for that plot.
ESM 2. Robustness assessment for effects of food supplementation on torpor use to differences in grey mouse lemurs’ body condition between plots before the experiment
Background and objective
In the present study, the experimental paradigm and data are inappropriate to assess the role of body condition on torpor use. The unique objective of our experimental setup was to increase food availability to individuals of one plot (experimental plot), while maintaining individuals of another plot with natural food availability (control plot). Food supplemented animals likely used this supplementary food to increase body fat deposits, i.e. their body condition (Vuarin & Henry 2014). Unfortunately, we could not document the temporal pattern of body condition throughout the experiment: animals were recaptured only once, at the end of the experiment, and have not been weighed at recapture. But, since (i) daily torpor use depends on body condition (the better the body condition, the deeper torpor bouts; Vuarin et al. 2013), and (ii) body condition was higher at the start of the experiment in the control plot than in the experimental plot in the present study (ESM 1), we still had to assess whether this initial difference in body condition between plots could have confounded the effect of food supplementation on torpor use. If the statistical effects of the plot on torpor parameters remained unchanged despite the inclusion of explanatory effects involving body condition at capture in statistical models, it could be concluded that differences in torpor use between control and experimental individuals were not a consequence of initial differences in body condition between plots, and therefore could confidently be interpreted as a direct influence of food supplementation on torpor use.
Materials and methods
To assess the robustness of our results to differences in initial body condition index (BCI) between plots at the start of the experiment, first, we tested for a potential effect of BCI, and interactions with other fixed effects, on the three torpor parameters. To do so, we added the following terms to final models presented in Table 1: (1) an additive effect of BCI, because torpor use may have depended on initial BCI, (2) an interaction between treatment and BCI effects, because the effect of food supplementation may have been stronger for individuals with lower body condition, (3) an interaction between time and BCI effects, because the effect of BCI at the start of the experiment may have weakened throughout the study period, and (4) an interaction between effects of treatment, time and BCI because BCI-dependency of the response to treatment could vary over time. Note that the fixed effect of BCI and its interactions were tested for torpor duration (Dtorp) and minimal skin temperature (Tsk min), but that only the fixed effect of BCI could be included for the probability of individuals entering torpor (Ptorp), because statistical models did not converge when interactions including BCI were added. Second, we assessed how robust was the strength of effects retained in final models (effects slope and significance in Table 1) to inclusion of BCI effects.
Results
Neither additive, nor multiplicative effects of BCI could be statistically detected for the three torpor parameters (Table S2). Hence, the robustness of effects retained in final models (Table 1) was re-assessed with final models of Table 1 to which we added only an additive term for initial differences in BCI (Table S3). Signs, strengths and statistical significances of effects presented remained very similar, despite the inclusion of this BCI term.
Conclusion
As expected, torpor variables did not depend on BCI (and interactions with other fixed effects), everything else being equal as in Table 1. Since the main effects of the plot remained overall the same despite the inclusion of a (non-significant) body condition effect in the analyses of torpor parameters, differences in torpor use between plots could not be attributed to differences in body condition at the start of the experiment, and therefore can be interpreted as a direct effect of food supplementation on torpor use.
The fact that BCI effects did not influence torpor use in this experiment does not contradict our former conclusions on the role of body condition in torpor flexibility (Vuarin et al. 2013): the central experimental paradigm of the present work is food supplementation. This supplementary amount of food availability is very likely to have resulted in an improved body condition of food supplemented individuals (Vuarin & Henry 2014). Therefore we suspect that differences in body condition between plots (and between individuals within the experimental plot) have rapidly disappeared after some days, or weeks, of supplemental feeding. Hence, body condition at the start of the experiment was unlikely to confound the long-term difference in torpor use, but it deserved to be verified statistically.
For all these reasons, we decided not to introduce the potential dependence of torpor use on body condition at capture in the analyses presented in the main manuscript.
Table S2. Effects of body condition at capture (BCI), and its interactions with treatment and time effects, on the probability to enter torpor on a given day (Ptorp), torpor duration (Dtorp, for torpid animals only) and minimal body temperature (Tsk min), for 20 free-ranging grey mouse lemurs. For each torpor parameter, BCI-related terms were added to the final model identified in Table 1 (i.e. a model that included all statistically significant terms, including non-significant additive terms involved in significant interactions).
Table S3. Slope estimate and statistical significance for effects influencing torpor parameters with final models as presented in Table 1, and with these same models but including an additive term allowing for an influence of BCI on torpor use.
Supplementary figures and tables
Figure S1. Grey mouse lemur (Microcebus murinus) equipped with a collar-mounted temperature logger, in Kirindy forest, western Madagascar (photograph: Pierre-Yves Henry).
Figure S2. Feeding platform fixed on a 1.2 m pole (aboveground) and equipped with a wire mesh preventing animals larger than grey mouse lemurs to forage at the platform (photograph: Pierre-Yves Henry). Based on trapping data (M. Dammhahn, unpublished data), 3 main aggregates of individuals were identified within the study plot, and 3 feeding platforms were placed per aggregate (home ranges of grey mouse lemurs are small, overlap, and are aggregated through space; Dammhahn and Kappeler 2008). The average distance between feeding platforms within an aggregate was 36 m (range: 31-54 m), and the average distance between groups of feeding platforms was 93 m (range: 80-131 m). Three nights of monitoring of each feeding platform, with custom-made automated RFID reader-loggers (based on OEM board LID-650, Euro I.D. Identsysteme GmbH & Co. KG, Weilerswist, Germany) connected to a square 21.0 x 18.5 x 2.5 cm Trovan antenna (ANT610F, Euro I.D) placed under the feeder, revealed that at least 7 out of the 11 marked individuals regularly frequented the feeding platforms.
Figure S3. Method for graphical determination of torpor use based on daily Tsk profiles. The threshold value of 30°C classically distinguishes shallow resting hypothermia from daily torpor (e.g., Schmid 2001; Munro et al. 2005; Laundry-Cuerrier et al. 2008). When using external temperature loggers (such as the collar-mounted loggers used in this study), the temperature sensor may not be in very close skin contact during the active phase as the animal moves and it can be influenced by air temperature, resulting in an oscillating Tsk curve. By contrast, the sensor is in permanent contact with the skin during the resting phase as the animal is curled up, resulting in a smooth curve. Therefore, oscillating parts of Tsk profiles are typical of activity whereas smooth parts are typical of rest (Dausmann 2012), when animals can enter into torpor. Profile A illustrates a Tsk profile where the timing of entrance and arousal from torpor are unambiguous regarding the threshold value of 30°C. Profile B illustrates a first case of ambiguous determination of the timing of torpor use. When Tsk oscillated around 30°C, we did not consider that the animal was torpid because the rapid Tsk fluctuations suggested regular movements of the animals. We considered that the animal was torpid only when it was obviously thermoconforming, i.e. when its Tsk started a continuous, smooth temporal decrease below 30°C, with a subsequent smooth return to euthermia (arousal). Torpor bout duration was calculated as the time elapsed between the first and the last Tsk values below 30°C for the time period considered as torpor use. Profile C illustrates another ambiguous case, where Tsk oscillated around 30°C but never smoothly decreased below 30°C. In that last case, it was considered that the animal did not enter torpor.