How hibernation in frogs drives brain and reproductive evolution in opposite directions

Abstract

Environmental seasonality can promote the evolution of larger brains through cognitive and behavioral flexibility but can also hamper it when temporary food shortage is buffered by stored energy. Multiple hypotheses linking brain evolution with resource acquisition and allocation have been proposed for warm-blooded organisms, but it remains unclear how these extend to cold-blooded taxa whose metabolism is tightly linked to ambient temperature. Here, we integrated these hypotheses across frogs and toads in the context of varying brumation (hibernation) durations and their environmental correlates. We showed that protracted brumation covaried negatively with brain size but positively with reproductive investment, likely in response to brumation-dependent changes in the socio-ecological context and associated selection on different tissues. Our results provide novel insights into resource allocation strategies and possible constraints in trait diversification, which may have important implications for the adaptability of species under sustained environmental change.

Keywords: anurans; brain size; evolutionary biology; hiberation.

Figure 1.. Allometric and seasonal variation in the species-specific tissue masses.

(A) Allometric slopes between the mass of each tissue and cubed snout-vent length (SVL³) so that proportionate scaling follows a slope of 1 on a log–log scale. Each point represents a species-specific mean value in breeding condition (N = 16). Relationships deviating from proportionate scaling (based on bootstrapped 95% confidence intervals) are highlighted in blue (steeper than unity) or red (shallower than unity). (B) Mean percent change with 95% confidence interval for body mass and each individual tissue of 50 anuran species with data from both shortly before and after brumation (=breeding), based on absolute tissue masses between stages and log-transformed to maintain symmetry and additivity (Törnqvist et al., 1985): log(post-brumation/pre-brumation) × 100. The transparent gray dots depict species-specific values. (C) Relationship between brumation period and percent mass change in the amount of body fat. (D) Relationship between brumation period and percent mass change in testis mass. Each point indicates a species.

Figure 2.. Effects of brumation duration on the relative tissue sizes.

Relationships between brumation duration and the relative mass of the brain (A), body fat (B), and testes (C) across males of 116 anuran species in breeding (post-brumation) condition. All axes are controlled for the snout-vent length and phylogeny.

Figure 3.. Effects of brumation duration on the relative tissue sizes.

Panels (A–D) depict the phylogenetic correlations (shown as phylomorphospace plots; Revell, 2012) between the relative masses of (A) brain and body fat, (B) brain and testes, (C) testes and body fat, and (D) testes and hindlimb muscles, respectively, across the 116 species (results in Appendix 1—table 14). The relative tissue masses represent the centered log ratios of the compositional data, and the lines connect the nodes of the underlying phylogeny, indicating that phenotypic correlations are not simply the result of phylogenetic clustering. The correlation coefficients and 95% confidence intervals are indicated. The loadings from a phylogenetic principal component analysis (Revell, 2012) on the same variables are also mapped as vectors onto biplots between (E) the first and second or (F) the second and third principal components. In all panels, the point colors reflect the species-specific brumation periods (see legend in panel A). Generally, where brumation was relatively shorter or absent, species also tended to have relatively larger brains, less body fat and smaller testes, respectively, consistent with the univariate analyses (Figure 2).

Figure 4.. Results of the averaged phylogenetic path model.

Visual representation of the average phylogenetic path model across 116 anuran species. Arrows reflect the direction of the path, with their widths being proportional to the standardized regression coefficients and colors indicating the sign (blue = positive, red = negative). Paths with 95% confidence intervals (CIs) excluding 0 (i.e., arrows highly probable) are drawn as solid arrows, all others as dashed, semi-transparent arrows. For simplicity and to avoid overparameterization, other organs were omitted in path models as they showed little covariation with brumation duration or brain size. All phenotypic traits were log-transformed, and all variables were controlled for body size via additional paths from log SVL. Although snout-vent length (SVL) had a strong effect on all variables (all β > 0.37), its thick blue arrows to each box are omitted in this figure only for visual clarity, but all path coefficients are presented with their 95% CI in Appendix 1—figure 7, with further details in Appendix 1—figure 6 and Appendix 1—table 6.

Appendix 1—figure 1.. Distribution of brumation durations (in days) across the 116 anuran species.

Each dot represents the species-specific mean ± 1 SE of the 5 years (2012−2016) that specimens were collected. For each year and species, all days with a collection site-specific ambient temperature below the experimentally quantified, species-specific temperature thresholds were summed to estimate the brumation period.

Appendix 1—figure 2.. Effects of brumation on breeding conditions.

Prolonged brumation shortened the breeding season (A), thereby increasing the probability of dense breeding aggregations (B). This effect was also supported by a trend toward higher mean population densities in species with a shorter breeding season (C).

Appendix 1—figure 3.. Confirmatory path analysis on the links between environmental variation, the brumation duration and breeding activity and the formation of breeding aggregations.

Directed acyclic graphs representing 8 candidate models (A) formed the predictions for the path analysis, resulting in the average path model (B). The path coefficients (standardised regression coefficients) and their 95% confidence intervals are depicted in (C). The variable letters correspond to the first letters of the full variable names in panel B. The model comparisons are listed in Appendix 1—table 13. The results were qualitatively identical when substituting latitude with elevation, mean temperature, or the coefficient of variation in temperature (except that the effect of mean temperature on brumation duration was negative as predicted). Note that here we used ‘hibernation’ instead of ‘brumation’ to avoid conflicts in codes.

Appendix 1—figure 4.. Directional test of trait evolution between relative testis size and breeding conditions.

Transition rates between binary states for relative testis size and (A) the duration of the breeding season or (B) aggregation formation. Independent models are shown on the left, models with changes in relative testis size dependent on the breeding parameters are depicted on the right. For both pairs of variables, the independent model had the highest support (A: wAIC = 0.50, B: wAIC = 0.66), but the dependent models shown here were either not significantly less supported (A: ΔAIC = 0.83, wAIC = 0.33) or relatively weakly so (B: ΔAIC = 2.38, wAIC = 0.20). The remaining two models (changes in breeding parameters depending on those in relative testis size or both variables evolving interdependently) found much less support in both analyses (A: ΔAIC ≥ 3.40, wAIC ≤ 0.09, B: ΔAIC ≥ 3.64, wAIC ≤ 0.11).

Appendix 1—figure 5.. Pairwise partial correlations between the different tissues.

All correlations are controlled for snout-vent length and phylogeny, expressed both by ellipses (above diagonal) and the correlation coefficients (below the diagonal). All correlations with |rp| ≥ 0.19 had P < 0.05 before and after accounting for multiple testing based on the False Discovery Rate.

Appendix 1—figure 6.. Candidate path models.

Directed acyclic graphs representing 28 candidate models that were compared to disentangle the relationships between five traits through a phylogenetic confirmatory path analysis and multi-model inference. For spatial reasons, the digestive tract is termed gut, and we omitted the paths from snout-vent length to all five variables in each panel to control tissue masses for body size or to improve d-separation (in the case of brumation).

Appendix 1—figure 7.. Coefficients of the averaged phylogenetic path model.

The path coefficients (standardised regression coefficients) are depicted with their 95% confidence intervals, corresponding to the path analysis in Figure 4 (directed path models in Appendix 1—figure 6). The paths from SVL to other traits are included only to control for body size (in the case of tissues) or to improve d-separation (SVL → Brumation). For visual clarity, these are omitted in the directed acyclic graphs (Appendix 1—figure 6) and the final averaged model (Figure 4).

Appendix 1—figure 8.. Validation of the different estimates of brumation duration.

(A) Comparisons of field observations with predicted brumation periods based on experimental thermal thresholds and site-specific climate data. The dashed line represents unity. The inserted density plot indicates that twenty-five of the 30 estimated values were within ≤8 days of the field observations across the spectrum, with the remaining five estimates deviating by approx. 13, 26 (3×) or, for one outlier, 55 days. (B) Variation in species-specific brumation periods based on different approaches. The green dots are the same field-based data as in panel A. The black dots represent values that combine the experimental temperature thresholds with the site-specific climatic data (predicted brumation below the threshold). The red and blue dots are based on the same experimental thresholds, but this time following the air temperature profiles of 2 and 4°C below these thresholds, respectively, to account for the potential buffering effect of the soil for underground hibernacula. Since the green dots were most closely associated with the black ones (i.e., without the additional buffer), we focused primarily on those values, but conducted additional analyses using the shorter, buffered brumation periods with no qualitative change. Finally, the horizontal dashed line splits the species into those with brumation likely present (above) or absent (below) to generate binary brumation variables for more direct comparison with previous studies in other taxa. This cut-off of ≤40 versus >40 days was motivated by the relatively large gap in the sequence of brumation periods in both the black and blue dots and that anurans are likely to survive relatively short cold spells without special adaptations. This cut-off generated divergent datasets between the direct experimental thresholds (black: N = 91 species with and 25 without prolonged brumation) and buffered thresholds (blue: N = 47 species with and 69 without prolonged brumation).

Appendix 1—figure 9.. Comparison of air and burrow temperatures.

The difference between burrow and outside air temperature increased with burrow depth across the five species examined in a pilot study. The points reflect the means with standard errors around them based on four burrows per species.

Appendix 1—figure 10.. Phylogeny of the species used in the study.

Species with data on pre-brumation condition are labelled with double-asterisks (**). Residual trait sizes are derived from log-log linear regressions against snout-vent length.