Modeling heterothermic fitness landscapes in a marsupial hibernator using changes in body composition

Abstract

Hibernation is an adaptive strategy that allows animals to enter a hypometabolic state, conserving energy and enhancing their fitness by surviving harsh environmental conditions. However, addressing the adaptive value of hibernation, at the individual level and in natural populations, has been challenging. Here, we applied a non-invasive technique, body composition analysis by quantitative magnetic resonance (qMR), to calculate energy savings by hibernation in a population of hibernating marsupials (Dromiciops gliroides). Using outdoor enclosures installed in a temperate rainforest, and measuring qMR periodically, we determined the amount of fat and lean mass consumed during a whole hibernation cycle. With this information, we estimated the daily energy expenditure of hibernation (DEE_H) at the individual level and related to previous fat accumulation. Using model selection approaches and phenotypic selection analysis, we calculated linear (directional, β), quadratic (stabilizing or disruptive, γ) and correlational (ρ) coefficients for DEE_H and fat accumulation. We found significant, negative directional selection for DEE_H (β_DEEH = - 0.58 ± 0.09), a positive value for fat accumulation (β_FAT = 0.34 ± 0.07), and positive correlational selection between both traits (ρ_{DEEH × FAT} = 0.24 ± 0.07). Then, individuals maximizing previous fat accumulation and minimizing DEE_H were promoted by selection, which is visualized by a bi-variate selection surface estimated by generalized additive models. At the comparative level, results fall within the isometric allometry known for hibernation metabolic rate in mammals. Thus, by a combination of a non-invasive technique for body composition analysis and semi-natural enclosures, we were characterized the heterothermic fitness landscape in a semi-natural population of hibernators.

Keywords: Correlational selection; Dromiciops; Hibernation; Isometric scaling; Quantitative magnetic resonance; Selection surface.

Fig. 1

Time series of body temperatures (red, from intraperitoneal data-loggers) and ambient temperatures (black, from environmental data-loggers), in five individuals with intraperitoneal data-loggers. The change in body mass during the experiment is shown in each plot. Dotted lines represents the winter period. See text for details and methods

Fig. 2

a Relationship between torpor bout duration and the hibernation period of the five individuals with data-loggers presented in Fig. 1. b Relationship between torpor bout duration and ambient temperature for the same individuals. c Maximum torpor bout duration, extracted from the time series of T_B (a), associated with the hibernation period. The dataset was adjusted to a parable (adjusted R² = 0.90, p = 0.0001), for which the equation is indicated in the graph. Confidence interval of 95% was shown

Fig. 3

Changes in body mass and composition (total water, lean mass and fat mass, in grams; within bars) during the experimental period (means; error bars represent standard errors). a Represents the “periodic” treatment and b shows the “undisturbed” treatment (= only initial and final qMR measurements). Numbers above bars show sample size, and numbers within bars indicate means in grams

Fig. 4

Daily energy expenditure of hibernation (DEE_H) determined in individuals that were moved periodically to the laboratory for measurements of qMR and individuals that were measured only at the beginning and end of the experiment, showing a net increase of 0.39 kJ day⁻¹ in periodically disturbed animals. Significant values are shown after an ANCOVA using lean mass as covariate (F_1,29 = 6.2; p = 0.019)

Fig. 5

Time progression of fat (a), lean mass (b,) and energy consumption as DEE_H (c), expressed as individual reaction norms. Boxes represent medians and range. Red crosses show non-survivor animals. Mean ambient temperature at the experimental site is plotted as reference in b

Fig. 6

a Survival curve (± 95% confidence intervals, dotted line) of D. gliroides hibernation in our semi-natural enclosures. b Survival as binomial variable, in function of DEE_H (standardized traits to mean = 0 and SD = 1). The linear coefficient of selection is shown for DEE_H, and was found to be significant (logistic regression with p = 0.045)

Fig. 7

Correlational selection maximizing low DEE_H and high fat accumulation, represented by a fitness surface modeled using generalized additive models (adj-R² = 0.82; deviance explained = 84%) and integrated smoothing parameters on the dataset using days of survival as fitness proxy. See “Results” and Table 4 for details

Fig. 8

Comparative analysis of daily energy expenditure of hibernation (DEE_H) using data from literature and our results for D. gliroides (modified from Nespolo et al. 2022b). Body composition during hibernation was recorded as body mass changes during hibernation for each hibernator