SEARCH

SEARCH BY CITATION

Keywords:

  • Heterothermy;
  • Heterothermy Index;
  • homeothermy;
  • mammals;
  • thermoregulation;
  • Thermoregulatory Scope

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Aim

The ability of endotherms to physiologically regulate body temperature (Tb) is presumed to be important in the adaptive radiation of birds and mammals. Recently, attention has shifted towards determining the extent and energetic significance of Tb variation documented in an ever-expanding list of species. Thus, we provide the first global synthesis of ecological and evolutionary correlates of variation in mammalian Tb.

Location

World-wide

Methods

We conducted a phylogenetically informed analysis of Tb variation using two complementary metrics, namely Thermoregulatory Scope (TS) and Heterothermy Index (HI), that treat Tb variation as a continuous variable. We included morphological (e.g. body mass), ecological (e.g. food habits) and environmental (e.g. latitude) correlates in the analysis.

Results

Among 560 mammal species included in the TS analysis, Tb relates most strongly to body mass (included in all models), season (relative parameter weight: 0.95), absolute latitude (0.80) and hoarding behavior (0.72), with small-bodied, high latitude and non-hoarding species expressing the most Tb variation. Small-bodied and high latitude species also express a greater range of thermoregulatory patterns than large-bodied and low latitude species. Results were generally similar in HI analysis, but in summer the extent of heterothermy decreases with latitude.

Main conclusions

Mammalian heterothermy is related to evolutionary history, climate conditions constraining minimum Tb, resource conditions mediating energy supply for maintaining high Tb, and latitudinal variation in the nature of seasonality. Our analysis further shows that traditional classification of mammals as hibernators, daily heterotherms or homeotherms is clouded or possibly misleading.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Endothermy was one of the most significant steps in the evolutionary history of mammals and birds (Lovegrove, 2012b). Although the maintenance of an elevated, approximately constant body temperature (Tb) via internal heat production is the defining feature of endotherms, many birds and mammals are characterized by pronounced daily and/or seasonal fluctuations in Tb (Angilletta et al., 2010). Previously, small variations in Tb over daily and seasonal time scales were often attributed to thermoregulatory imprecision or outright failure of the thermoregulatory system, but they are increasingly interpreted as adaptive variation, more often the result of a change in thermoregulatory strategy or optima rather than a failure to thermoregulate (Humphries et al., 2003; Grigg et al., 2004; Cooper & Geiser, 2008; Angilletta et al., 2010; Careau et al., 2012).

Variation in Tb has important implications for our understanding of how endotherms maintain energy balance and survive across a wide range of climates and environments. Across species, Tb accounts for much of the variation in metabolic rate (MR) (Gillooly et al., 2001), and even within a species, small fluctuations in Tb have substantial effects on MR and likely on performance (Angilletta et al., 2010). Given that core Tb of some endotherm species can vary by nearly 40 °C over a few hours (Barnes, 1989) while others may not vary by more than 0.5 °C in a lifetime (McNab, 2008), differential expression of heterothermy (in this case, defined as any reversible fluctuations in Tb) is an extremely important factor in the ecologies and life histories of endotherms (Grigg et al., 2004; Lovegrove, 2012b). Many pages have been filled with evaluations of the ecological importance of MR (often in the context of the Metabolic Theory of Ecology; Brown et al., 2004). However, the importance of variation in Tb has been largely ignored in allometric and phylogenetic comparisons of endotherms, in part because most multispecies comparisons use basal MR (BMR) measurements that exclude consideration of MR outside of thermoneutral air temperatures (i.e. where BMR balances heat loss) and so-called normothermic Tb values (i.e. Tb values associated with thermoneutrality and classical endothermic homeothermy).

Past comparative research on variation in endotherm Tb has been dominated by categorical classifications of Tb variation. Categorical classifications separate different Tb patterns into discrete groups (i.e. homeothermy, daily torpor and hibernation) according to the amplitude and period of Tb fluctuations and, at times, their thermoregulatory causes or metabolic consequences (Geiser & Ruf, 1995; Geiser, 2004; Heldmaier et al., 2004). There is often overlap between these categories and there is extensive variation within these categories as well as extensive variation in the application of the definitions. Amplitude or period thresholds that successfully separate species into different categories in one context often do not perform as well in different contexts. Further, taxa considered either homeothermic or heterothermic have rarely been included together in comparative analyses (Boyles et al., 2011a). Finally, despite apparent phylogenetic patterns in Tb, few previous studies have used modern phylogenetically informed analyses to differentiate evolutionary and ecological factors affecting patterns in Tb (but see Lovegrove, 2012b). An alternative approach is to quantify Tb as a continuous variable so that all species can be included in one analysis. The Heterothermy Index (HI) recently proposed by Boyles et al. (2011b) is an example of such a continuous approach, intended to facilitate quantitative assessment of Tb patterns in endotherms while avoiding arbitrary thresholds often used to differentiate torpid and euthermic intervals. As research moves forward in this field, we will undoubtedly see more examples of continuous metrics of heterothermy, including a new metric we introduce herein.

We provide an analysis of mammalian Tb patterns that includes taxa from across the thermoregulatory spectrum and employs the most thorough phylogenetically informed analysis to date. Our goals are to (1) evaluate global patterns in mammalian Tb in reference to assumptions often made in thermoregulatory literature (e.g. Tb fluctuates more among small mammals and those found at high latitudes) and (2) determine if the recent interpretation of mammalian heterothermy as a plesiomorphic trait with a single evolutionary origin (Malan, 1996; Lovegrove, 2012a) is supported by a phylogenetically informed, species-level analysis of these patterns.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

We assessed predictors of Tb variation in mammals using two continuous metrics. First, we used published Tb measurements to calculate a new metric (Thermoregulatory Scope; TS) describing 560 mammal species. Second, we calculated HI values for a subset of these species for which adequate data were available (HI requires adequate sampling covering the whole range of Tbs expressed by the animal). Both metrics are similar in that they focus on deviation in Tb away from a high set point. They differ because TS is a proxy for the thermoregulatory range (flexibility) that a species can display and accounts for the magnitude of deviations, while HI is a proxy for the level of variation actually displayed in the wild and accounts for both the magnitude and duration of deviations away from a high set point. In other words, TS roughly represents an estimate of the fundamental thermoregulatory niche of species while HI represents the realized thermoregulatory niche. Thus the two metrics, while complementary, provide different information about variation in Tb among endotherms.

Thermoregulatory Scope

Data were collected from a large number of published sources (Appendix S2 in Supporting Information). Geiser & Ruf (1995) and Geiser (2004) were used as starting points to identify species with associated Tb data and their original sources. Lovegrove (2012b) and McNab (2008) were used as guides to identify additional species with Tb data available in the literature. To minimize bias in collecting data, particularly because all four initial sources focused on heterothermic species, mammalian families still missing in the dataset were identified using Mammals of the World (http://www.bucknell.edu/msw3/) and families were researched in a general Web of Knowledge search using genus names and ‘temperature’ in the topic line.

We defined TS as mean Tb – minimum Tb. Mean Tb was the mean Tb expressed while regulating around a given species' highest (typically active-phase) set point. Minimum Tb was the lowest Tb from which subjects have been shown to naturally rewarm. Thus, for a species expressing daily variation around a high temperature set point, minimum Tb was the lowest Tb observation recorded. For species which expressed torpor, the lowest torpid Tb was used, while mean Tb was the mean of minimal and maximal values measured in euthermic individuals. Priority was first given to studies using the most precise method of measurement (i.e. implanted preferred over orifice preferred over skin), and data from different types of measurements were not mixed. In a few cases, a skin temperature was taken over orifice or implanted recordings where minimum skin temperature was more than 5 °C below the other measurements and the value was taken to indicate torpor expression in the original study. If multiple studies used the same recording method, the study that recorded Tb over the longest time period, followed by the greatest number of subjects, was taken in priority. Two sources (one providing mean Tb and the other providing minimum Tb) for one species were combined for a few species if they employed the same method of Tb measurement. An important limitation of TS is its sensitivity to sampling effort because additional sampling is likely to increase, but can never decrease, the estimate. Accordingly, we use an index of sampling effort derived from publication intensity as a covariate in TS analyses. We further explored the sensitivity of our results to sampling effort by repeating our TS analyses on a smaller subset (n = 185) of more intensively studied species (i.e. those species for which at least two papers have been published).

Heterothermy Index

We collected data from a smaller subset of mammals for which adequate time-course data of Tb are available (Appendix S3). For this dataset, we focused our search on studies conducted under natural or semi-natural conditions. We first contacted authors of each study and requested representative raw data for at least one individual from each study site or treatment group. When attempted contact with authors failed (no email address or no reply), we extracted a representative set of Tb datapoints from digitized figures using DataThief III (version 1.1, http://www.datathief.org). We attempted to space points evenly on the time axis, but this method is obviously less accurate than using raw data, so it was used sparingly (10 populations in the winter dataset and 4 in the summer).

HI values integrate time and magnitude of deviations away from the Tb most commonly displayed during activity into a single, continuous metric. HI can be calculated as (Boyles et al., 2011b):

  • display math

Where Tb-opt is the optimal Tb for performance, Tb-i is the Tb measurement at time i and n is the number of times Tb is sampled. In essence, the metric calculates the magnitude of a heterothermic response in relation to Tb-opt an animal displays during a given time frame. Since HI values represent an estimate of the realized thermoregulatory niche, they are difficult to interpret if averaged across populations of a single species; thus, if multiple datasets or studies were available for a given species, they were calculated separately and included in the analysis as independent datapoints. For the same reason, we analyzed HI values recorded in summer and winter separately, instead of in a combined analysis with season as a covariate, as was done in the TS analysis. Note that in contrast to TS, a metric that only considers deviations below normal active temperatures, HI values considers deviations both above and below normal active temperatures.

Explanatory variables

Explanatory variables that we included in the dataset were body mass (logmass), latitude (abslat), average air temperature (avgtemp), temperature seasonality (seasontemp), total precipitation (totprecip), precipitation seasonality (seasonprecip), season, climatic zones (climate), food habits (food), food hoarding behavior (hoard), habitat substrate (substrate), publication intensity (rootpub; the number of Tb-related publications on each species, to account for the sampling effort bias in TS), laboratory or field (lab/field) and method of Tb collection (tbmeth). Explanatory variable values for a particular species occasionally vary between TS and HI analyses when data were taken from different studies at different locations. To consider species' relatedness in the analysis (Harvey & Pagel, 1991), we derived a phylogenetic tree that included all 560 mammal species with TS values. The main tree originated from Fritz et al. (2009) to which several additional species/subspecies were added where appropriate. The trees for summer and winter HI analyses were subsets of the tree used for the TS analysis. A complete explanation and the phylogeny and data sources for each variable can be found in the Supporting Information.

Data analysis

Distributions of several variables were highly right-skewed, thus we normalized them using log10 transformation. In four instances, average Tb was identical to minimum Tb; therefore we inserted the formula log10 (0.5 + TS) to transform all TS values. The distribution of publication intensity was also right-skewed but was transformed by taking the square root due to the abundance of zeros. We analyzed data using phylogenetic generalized least-squares (PGLS) models in R (v2.13.0; R Development Core Team 2009) using the ‘ape’ and ‘caper’ packages. In PGLS models, the value of the phylogenetic signal (λ) was estimated using maximum likelihood to optimally adjust the degree of phylogenetic correlation in the data. Values of λ can range from 0 to 1, where 0 indicates no correlation and 1 represents a very strong correlation.

We used a modified stepwise approach to model selection based on the Akaike information criterion (AIC) for the TS analysis and AIC corrected for small sample sizes (AICc) in the HI analyses. Briefly, in the TS analysis we first chose the best fit model considering only technical variables (rootpub, tbmeth and lab/field) and retained important variables for the second step. We then competed models containing retained technical variables as well as climatic variables (climate, abslat, avgtemp, seasontemp, avgprecip and seasonprecip) and retained all variables in models with some support (i.e. models with ΔAIC < 2). Third, using retained technical and climatic variables in all models, we competed models with various combinations of biological variables (season, hoard, food and substrate) and retained all variables in models with some support. Finally, we competed 127 models containing all of the variables retained from previous steps in various combinations. In the HI analysis, we omitted the technical variable step because there was no laboratory versus field variable and all other technical variables were retained in the first step of the TS analysis. This selection technique led to 127 competing models in the winter HI analysis but only 7 competing models in the summer HI analysis. To assess the sensitivity of TS results to sampling effort, we re-ran the TS analyses restricted to the 185 species with publication intensity index > 1. A detailed explanation of our overall model comparison approach is given in Supplementary Materials.

For influential variables, partial regression plots were generated to visualize the effect of each variable on TS and HI after accounting for all other important factors. We used the R package ‘quantreg’ to fit quantile regression lines. All regression coefficients (β ± SE) are reported as standardized coefficients in units of standard deviation and residual averages (μ ± SE) for specific taxa. We reconstructed the ancestral states for TS (on the raw scale) using Felsenstein's (1985) phylogenetic independent contrasts (PIC) in the R package ‘ape’ (function ace). We used jmp 9.0.0 (SAS Institute Inc., Cary, NC, USA) to estimate whether distributions of each variable were best described as unimodal or a mixture of two to five normal distributions. We evaluated relative fit of each distribution by comparing their respective AICc values.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Thermoregulatory Scope

A TS value was calculated for each of 560 species, covering all prototherian and metatherian orders, and all eutherian orders except Dermoptera. Species-level TS ranged from 0 to 40.5 °C, with a highly right-skewed distribution. The distribution was quadri-modal, however, as a mixture of four normal distributions fit the data better (AICc4 = 3283.90) than a unimodal distribution or mixtures of two, three or five normal distributions (AICc1 = 4273.96; AICc2 = 3332.53; AICc3 = 3295.28; AICc5 = 3290.34). TS had distinguishable peaks around 1.40, 8.21, 18.95 and 29.99 °C, although the peak at 1.40 °C was at least an order of magnitude higher than the other three (Fig. 1a).

figure

Figure 1. Distribution of (a) thermoregulatory scope (TS) for 560 mammalian species obtained from the literature, defined as average body temperature (Tb) expressed while regulating around the given species' highest set point minus minimum Tb exhibited by the species, (b) winter heterothermy index (HI), and (c) summer HI. The inset shows the data without the peak at 1.40 °C to make the other three peaks more discernible.

Download figure to PowerPoint

Families with very low average TS included Felidae (mean ± SE = 0.58 ± 0.13 °C, n = 8), Delphinidae (0.67 ± 0.13 °C, n = 6), Phalangeridae (1.09 ± 0.26 °C, n = 7), Bovidae (1.14 ± 0.18 °C, n = 14) and Canidae (1.15 ± 0.46 °C, n = 10). Families with high TS included Gliridae (32.94 ± 2.52 °C, n = 5), Cheirogaleidae (26.80 ± 0.80 °C, n = 5), Vespertilionidae (26.32 ± 1.37 °C, n = 34), Dipodidae (25.32 ± 5.80 °C, n = 5) and Sciuridae (20.78 ± 2.73 °C, n = 30). The average TS was greatest in metatherians followed by prototherians and eutherians (Table 1a). Within eutherian clades, Afrotheria, Laurasiatheria and Euarchontoglires showed similar TS, while Xenarthra was considerably lower (Fig. 2; Fig. S1; Table 1a).

figure

Figure 2. Phylogeny of 560 mammalian species used in the analysis showing the raw mean thermoregulatory scope (TS; height of black bar). Labels on the periphery denote Monotremata, Marsupialia, Afrotheria, Xenarthra, Laurasiatheria, and Euarchontoglires and internal shadings of the branches denote Prototheria (no shading), Metatheria (dark shading), and Eutheria (light shading). The tree was drawn using Interactive Tree Of Life (Letunic & Bork, 2011). See Fig. S1 for a color version.

Download figure to PowerPoint

Table 1. Taxon-specific means in (a) thermoregulatory scope (TS; in °C), (b) heterothermy index (HI) in winter, and (c) HI in summer for major subclasses and Eutherian clades in the dataset. For all traits, the mean was calculated on both raw values (mean ± SE) and standardized residuals from the best model (resid, with an overall mean of 0 and SD of 1).
Clades(a) Thermoregulatory Scope(b) Heterothermy Index (winter)(c) Heterothermy Index (summer)
NMean±SEResid±SENMean±SEResid±SENMean±SEResid±SE
Subclasses:                     
Prototheria49.75±6.190.46±0.390NA±NANA±NA0NA±NANA±NA
Metatheria679.83±1.260.22±0.081110.57±2.860.15±0.1610.65±NA−0.65±NA
Eutheria4898.52±0.50−0.03±0.044812.32±1.46−0.03±0.12254.26±0.880.03±0.10
Eutherian clades:                     
Afrotheria2911.06±2.050.40±0.1357.21±1.180.19±0.1961.83±0.730.05±0.13
Xenartha133.79±1.670.21±0.160NA±NANA±NA0NA±NANA±NA
Laurasiatheria2298.07±0.680.02±0.052112.58±1.97−0.34±0.13136.79±4.82−0.08±0.15
Euarchontoglires2188.95±0.81−0.16±0.062213.23±2.550.20±0.2261.24±0.180.24±0.21

Measurement method and publication intensity affected estimates of TS and were thus retained in all models to account for any bias that might arise from differences in sampling methodology and effort. Studies reporting skin and orifice temperatures tended, on average, to report more variation in Tb than studies recording temperature with implanted devices. There was also a tendency for more intensively studied species to be characterized by higher TS estimates than seldom studied species. The best fit model included abslat, hoard and season (in addition to logmass, tbmeth and rootpub), but had a wi of only 0.13 (i.e. it could be considered as 13% probable that it is the best model; Table 2a). Four other models were equally parsimonious (ΔAIC < 2), and all included season, hoard and abslat in combination with a different climatic variable (avgtemp, totprecip, seasonprecip and seasontemp). Individual variable weights obtained by model averaging support a nuanced interpretation. Season (parameter weight of 0.95), abslat (0.80) and hoard (0.72) were very important in determining TS, as were, to a lesser degree, avgtemp (0.35), totprecip (0.37), seasonprecip (0.30) and seasontemp (0.30).

Table 2. Model selection parameters (AIC = Akaike's information criterion; AICc = AIC corrected for small sample size; Δi = difference in AIC or AICc; wi = Akaike weight; accwi = cumulative wi; ER = evidence ratio) of phylogenetic least-squares models testing adaptive hypotheses on inter-specific variation in (a) thermoregulatory scope (TS) in 560 mammalian species, (b) winter heterothermy index (HI) in 59 populations of 50 species, and (c) summer HI in 26 populations of 23 species. Variables included in these models are season, hoarding behavior (hoard), absolute latitude (abslat), average air temperature (avgtemp), total precipitation (totprecip), precipitation seasonality (seasonprecip) and temperature seasonality (seasontemp). Log-transformed body mass, square-root transformed publication intensity, and Tb method were included in all models. In a and b, only 7 and 8 of the 127 final competing models are shown (with Δi < 3; note that all models were used in model averaging, see text), whereas all 7 final competing models are shown in c.
ModelAICΔiwiaccwiER
(a) TS ∼logmass+rootpub+tbmeth     
+season+hoard+abslat475.30.000.130.131.00
+season+hoard+abslat+avgtemp476.61.280.070.201.90
+season+hoard+abslat+totprecip476.81.460.060.262.07
+season+hoard+abslat+seasonprecip477.31.960.050.302.67
+season+hoard+abslat+seasontemp477.31.970.050.352.68
+season+abslat477.32.020.050.402.75
+season+hoard+abslat+avgtemp+totprecip478.22.880.030.434.23
ModelAICcΔiwiaccwiER
(b) winter HI ∼logmass+rootpub+tbmeth     
+avgtemp123.30.000.210.211.00
+abslat+avgtemp+totprecip123.50.250.180.391.13
+abslat+avgtemp123.70.410.170.551.23
+hoard+abslat+avgtemp+seasontemp124.71.390.100.652.01
+abslat+avgtemp+seasontemp124.71.440.100.752.05
+abslat+totprecip+seasonprecip124.81.500.100.852.12
+hoard+abslat+avgtemp+seasonprecip125.32.070.070.922.81
+abslat+totprecip+seasontemp+seasonprecip125.92.640.050.983.73
(c) summer HI ∼logmass+rootpub+tbmeth     
+abslat+avgtemp59.50.000.800.801.00
+abslat+avgtemp+hoard62.83.280.150.955.16
+avgtemp67.07.510.020.9742.74
+abslat67.78.210.010.9860.70
+hoard67.88.280.011.0062.95
+hoard+avgtemp70.811.320.001.00286.56
+hoard+abslat71.812.000.001.00402.93

Estimates and partial correlations were taken from the best model including six variables (logmass, tbmeth, rootpub, season, hoard and abslat) and a significant λ optimized at 0.703 (log-likelihood ratio test: χ2 = 50.95, P < 0.001). Body mass was strongly negatively related to TS (β ± SE = −0.52 ± 0.06; Fig. 3a). Non-parallel quantile regression lines show that variance in TS decreases as body mass increases (Fig. 3a). Absolute latitude, conversely, showed a positive relationship with TS (β ± SE = 0.10 ± 0.04; Fig. 3b), with increasing variance at high latitudes (Fig. 3b). Food hoarders had lower TS than non-hoarders (β ± SE = −0.21 ± 0.11; Fig. 3c), and TS values were higher in studies conducted in winter or both seasons than summer and studies where season was unclear (Fig. 3d). Standardized estimates for additional climatic variables (taken from the 2nd to 4th best models) indicated positive influence of avgtemp (β ± SE = 0.04 ± 0.05) and negative effects of totprecip (β ± SE = −0.03 ± 0.04), seasonprecip (β ± SE = −0.01 ± 0.03) and seasontemp (β ± SE = −0.01 ± 0.05), although large standard errors relative to estimates suggest these effects were not strong.

figure

Figure 3. Partial residual correlations between thermoregulatory scope (TS, log-transformed, corrected for phylogeny at λ = 0.70 and other variables in the best model, see Table 2a) and (a) log10 body mass, (b) absolute value of latitude (abs latitude), (c) food-hoarding category (N = no, Y = yes), and (d) season in which TS was quantified (S = summer, W = winter, W/S = winter and summer, and U = unknown). Lower and upper lines in panels a and b are 25th and 75th quantile regression lines, respectively.

Download figure to PowerPoint

With effects of logmass, tbmeth, rootpub, season, hoard and abslat accounted for, families that exhibited lower than expected residual TS were Soricidae (μ ± SE = −1.02 ± 0.19, n = 19), Muridae (−0.74 ± 0.09, n = 42), Canidae (−0.47 ± 0.16, n = 10), Bathyergidae (−0.41 ± 0.16, n = 9), Cricetidae (−0.38 ± 0.11, n = 53) and Macropodidae (−0.14 ± 0.23, n = 7). By contrast, families that exhibited higher than expected residual TS were Gliridae (1.12 ± 0.09, n = 5), Cheirogaleidae (0.72 ± 0.12, n = 5), Dipodidae (0.66 ± 0.30, n = 5), Tenrecidae (0.67 ± 0.24, n = 7), Vespertilionidae (0.58 ± 0.06, n = 34) and Dasyuridae (0.50 ± 0.10, n = 20). The average residual TS was lower in eutherians than prototherians and metatherians (Table 1a). Within Eutheria, the ancestral Afrotherian clade and the Xenarthans showed greater residual TS than the younger Laurasiatheria and Euarchontoglires clades (Table 1a). The estimated ancestral state for TS was 7.97 °C (95% confidence intervals: −18.86 and 34.19).

A TS analysis restricted to 185 species for which > 1 paper has been published did not alter the direction of any strong predictors but altered their relative strength, with food hoarding and season increasing and latitude decreasing in relative importance (Table S1).

Heterothermy Index – winter

Winter HI values were calculated for 59 populations of 50 mammalian species, all metatherians and eutherians. Species-level winter HI ranged from 0.27 to 35.51 °C, with a highly right-skewed distribution. The distribution was either bi- or tri-modal, as mixtures of two or three normal distributions fit the data better (AICc2 = 409.85; AICc3 = 409.73) than a unimodal distribution or a mixture of four normal distributions (AICc1 = 441.56; AICc4 = 432.77). The tri-modal distribution revealed that winter HI had distinguishable peaks around 1.51, 15.47 and 32.28 °C (Fig. 1b).

The families Bovidae, Ursidae and Canidae were each represented by a single species in the dataset with very low winter HI values (0.27, 0.47 and 0.78 °C, respectively). The only family with multiple species represented in the dataset that had a low average winter HI was Muridae (mean ± SE = 1.59 ± 0.14 °C, n = 5). Families with high average winter HI included Sciuridae (21.17 ± 3.39 °C, n = 11) and Vespertilionidae (17.01 ± 2.35 °C, n = 11). At the broadest level, average winter HIs were similar among metatherians and eutherians (Table 1b). Within eutherian clades, Laurasiatheria and Euarchontoglires showed similar winter HI, while Afrotheria was slightly lower (no data for Xenarthra; Fig. S2; Table 1b).

The best fit model included only avgtemp (in addition to logmass, tbmeth and rootpub), but had a wi of only 0.21(Table 2b). Five other models yielded ΔAICc < 2, which all included abslat in combination with different climatic variables (avgtemp, totprecip, seasonprecip and seasontemp). Only one model included the hoard variable. The variables avgtemp (parameter weight of 0.84) and abslat (0.79) were very important in determining winter HI, as were, to a lesser degree, totprecip (0.35), seasontemp (0.28), seasonprecip (0.23) and hoard (0.19).

The best model showed significant phylogenetic signal in winter HI (λ = 1.00; log-likelihood ratio test: χ2 = 53.38, P < 0.001), a negative relationship with body mass (β ± SE = −0.46 ± 0.22; Fig. 4a), and negative relationship with avgtemp (β ± SE = −0.14 ± 0.03). In the second-best model, there was a positive effect of abslat (β ± SE = 0.19 ± 0.11; Fig. 4b) and negative effect of totprecip (β ± SE = 0.04 ± 0.07). The fourth-best model indicated that food hoarders had lower winter HI than non-hoarders (β ± SE = −0.56 ± 0.40) and that winter HI is negatively correlated with seasontemp (β ± SE = −0.08 ± 0.11). However, with the exceptions of body mass and avgtemp, large standard errors relative to estimates suggest these effects were not important.

figure

Figure 4. Partial residual correlations between log-transformed winter heterothermy index (HI, corrected for phylogeny at λ = 1 and other variables in the best model, see Table 2b) and (a) log10 body mass and (b) absolute value of latitude (abs latitude), and log-transformed summer HI (corrected for phylogeny at λ = 0.80 and other variables in the best model, see Table 2c). Lower and upper lines in panels are 25th and 75th quantile regression lines, respectively.

Download figure to PowerPoint

With effects of logmass, tbmeth, rootpub and avgtemp accounted for, Muridae still exhibited lower than expected residual winter HI (μ ± SE = −0.85 ± 0.05, n = 5). Sciuridae still had a relatively high residual winter HI (0.87 ± 0.28, n = 11), but Vespertilionidae was somewhat less of an outlier on the raw scale (−0.32 ± 0.15, n = 11).

Heterothermy Index – summer

Summer HI values were calculated for 26 populations of 23 mammalian species. Species-level summer HI ranged from 0.38 to 19.83 °C. The small sample size precluded fitting of a mixture of normal distributions, but the right-skewed nature of the distribution is clear (Fig. 1c). Likewise, the small sample size made family-level comparisons difficult as most families were represented by only one species (but see Fig. S3).

After the first three steps of the model selection procedure, seven models were put into competition at the final stage (Table 2c). All of these models included logmass, tbmeth and rootpub as covariates. The best fit model had a wi of 0.80 and included abslat and avgtemp (Table 2c). The second-best model included an additional effect of hoard but had a ΔAICc = 3.28. Individual variable weights obtained by model averaging indicated that abslat (parameter weight of 0.97) and avgtemp (0.97) were important in determining summer HI, whereas the variable hoard was less important (0.17).

The best model showed significant phylogenetic signal in summer HI (λ = 0.80; log-likelihood ratio test: χ2 = 69.66, P < 0.001) and a weak negative relationship with body mass (β ± SE = −0.17 ± 0.18; Fig. 4c). The variables abslat and avgtemp were negatively correlated with summer HI (abslat: β ± SE = −1.15 ± 0.35; Fig. 4d; avgtemp: β ± SE = −1.34 ± 0.40).

There were 13 species for which HI was quantified in both winter and summer. HI values were significantly larger in winter (mean ± SE = 13.82 ± 3.09 °C) than in summer (mean ± SE = 2.72 ± 0.76 °C; paired t-test = 3.675; P = 0.003; Fig. S4); however, the correlation between HI in winter and summer was not significant (r = 0.22; P = 0.47). Although several species have similar HI values in both seasons (e.g. Nyctereutes procyonoides and Xerus inauris), others have very low HI in summer and very high HI in winter (e.g. Spermophilus parryii and Marmota broweri; Fig. S4).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Here we provide a global analysis of Tb patterns in mammals including taxa from across the thermoregulatory spectrum. Our analysis suggests the need for modification of our traditional understanding of thermoregulatory patterns. Modes in the distribution of TS may bear a loose relationship to classical divisions between homeotherms, those that express daily torpor, and hibernators, but the four distinguishable modes (around 1, 8, 19 and 30 °C) suggests the possibility of a fourth category. However, the extensive overlap among, and variation within, these categories suggests these are minor modes (with the exception of the lowest mode representing traditionally defined ‘homeotherms’) within an otherwise largely continuous distribution of mammal Tb variation. There were three minor modes in winter HI values around 1.5, 15 and 32 °C (the sample size was too small to conduct such an analysis for summer HI), but again, the distribution was largely continuous. The fact that most modes are only minor and distributions are relatively continuous emphasizes the biological value of treating mammal thermoregulation as a continuous variable and including animals from across this continuum into integrated phylogenetic analyses (Angilletta et al., 2010; Boyles et al., 2011a; Canale et al., 2012).

Although mammalian normothermic Tb varies interspecifically by approximately 10 °C, thought to represent a continuum of baso-, meso- and supraendothermy (Lovegrove, 2012b), our analysis includes all these categories in a single, standardized analysis. The strong phylogenetic signal detected in our analysis (λ = 0.703 for TS, λ = 1.00 for winter HI, and λ = 0.80 for summer HI) indicates the expressed extent of heterothermy is an evolutionarily constrained trait with a significant phylogenetic signal. Accordingly, several mammalian families showed consistently low variation in Tb (Felidae, Delphinidae, Phalangeridae, Bovidae and Canidae), while other families were highly heterothermic according to both metrics (Gliridae, Vespertilionidae, Dipodidae and Sciuridae).

The taxonomic patterns in TS that remained after accounting for body mass, climatic, ecological and methodological factors revealed that phylogenetically older groups tend to be more likely to exhibit large variations in Tb, both among subclasses and within eutherian clades (Table 1). This pattern is consistent with the idea that, in mammals, heterothermy is a plesiomorphic, ancestral condition with evolutionary roots in the transition from ectothermy to endothermy (Malan, 1996; Lovegrove, 2012b). This pattern was cloudy, over even reversed, in our two HI analyses, suggesting that while generally capable of using more heterothermy, phylogenetically older taxa do not always do so under natural conditions. Additionally, the significant contemporary effects of climatic and ecological factors on thermoregulatory patterns suggest that, beyond this tendency for plesiomorphic heterothermy, mammalian patterns of heterothermy underwent extensive adaptive modification during the Cenozoic in response to factors such as the colonization of seasonal, high-latitude environments (Lovegrove, 2012b). The net result is a diverse range of thermoregulatory patterns being scattered across the contemporary mammal phylogeny, combined with the possible phylogenetically conserved tendency for groups often thought of as basal (Lovegrove, 2012b) to be more heterothermic than those often thought of as derived (see Krell & Cranston, 2004 for a discussion of this terminology).

Still, after phylogeny was accounted for, mass, latitude and hoarding behavior were important predictors of TS while latitude and average temperature were important predictors of HI. The mean of both TS and winter HI declined with increasing body mass (there was a similar but weak relationship for summer HI), suggesting that smaller mammals are more heterothermic than large mammals. On the surface, this agrees with previous interpretations of patterns in Tb in heterotherms (Geiser & Ruf, 1995), but inclusion of traditionally defined homeotherms in this analysis modifies the interpretation of the body mass–heterothermy relationship. With all species included, the predominant pattern is that small mammals occupy the entire gradient from homeothermic to strongly heterothermic, whereas large mammals are consistently closer to homeothermy. The decreasing variance in both indices as body mass increases suggests that small mammals have more thermoregulatory options (i.e. greater interspecific flexibility) than larger mammals but are not necessarily more heterothermic as is often suggested. Several factors likely contribute to this pattern. First, small mammals are more vulnerable to resource shortages than large mammals because of their higher mass-specific MRs (Gillooly et al., 2001; White, 2011), reduced ability to travel to distant resources (Lovegrove, 2000) and higher thermoregulatory costs (Cooper & Geiser, 2008; White, 2011). Thus, reducing metabolism by lowering Tb set point is one important option by which small mammals can compensate for their greater energetic vulnerability during periods of resource shortage. On the other end of the spectrum, large-bodied species require longer to cool and disproportionately more energy to re-warm than small mammals (i.e. thermal inertia, Prothero & Jürgens, 1986). This likely limits both the capacity and necessity to reduce Tb as an energy-saving mechanism and thereby constrains expressed thermoregulatory patterns.

The increase in TS and winter HI values with increasing latitude may represent climatic constraints. Lovegrove (2003) noted that major climate differences between zoogeographical zones are primarily determined by latitude. With increasing latitude, mean annual temperature decreases, annual extremes increase, and temperature and precipitation patterns become more predictable (Scholander et al., 1950; Lovegrove, 2003). Thus, given that an animal's Tb cannot drop below ambient temperature (Ta) during heterothermy, there is greater opportunity for species at higher latitudes, experiencing colder Tas, to exhibit more variation in Tb. Interestingly, the latitude–heterothermy relationship was reversed in the summer HI analysis, with species closer to the equator showing more variation in Tb than more polar species. This could reflect increased variability in resource availability during summer at lower latitudes (possibly associated with a wet/dry seasonal cycle) and/or increased pressure on reproduction at high latitudes because of the short summer season. Many tropical species, like temperate species, experience periodic resource shortages, in this case often associated with highly unpredictable precipitation patterns during summer (Schmid & Ganzhorn, 2009). Earlier literature emphasized heterothermy as a phenomenon largely restricted to small species occupying cold environments; this interpretation is changing with more research attention on tropical species (McKechnie & Mzilikazi, 2011). In this context, it is notable that the importance of latitude as a predictor of TS was sensitive to sampling effort in our analyses, largely because there were few well-sampled tropical species in our dataset. Thus, while our study does identify a latitudinal gradient in patterns of heterothermy, which remains when variation in sampling effort is accounted for as a covariate, the pattern is characterized by extensive variation in thermoregulatory patterns throughout the gradient, is sensitive to the reduced sampling effort in tropical and equatorial regions, and is likely intertwined with other factors, particularly resource availability (Lovegrove, 2003).

Both TS and HI metrics indicate the method of energy storage also affects variation in Tb, although the effect was much more prevalent in high-ranking models in the TS analysis than in the HI analyses (possibly due to the relative rarity of food hoarders in HI datasets). Food hoarders, which store food for consumption at a later time, showed less variable Tbs than species that do not hoard. The storage of food for future use greatly enhances the chances that an animal will survive unfavorable environmental conditions because much more energy can be reserved in a cache than as fat (Humphries et al., 2003). Fat storage capacity is constrained by body size, which greatly limits maximum reserve size in small mammals. In comparison, the maximum energy reserve that can be derived from a food store is much larger. Given the likely costs of decreasing Tb (Humphries et al., 2003; Angilletta et al., 2010), species that hoard food may not need to decrease their Tb as much as non-hoarders as they have a more secure energy supply. In fact, even within food-hoarding species, the extent and quality of the food reserve can affect the expression of heterothermy, such that individuals with large hoards of high-quality food may express less torpor over winter than those with small, poor-quality hoards (Landry-Cuerrier et al., 2008).

Our analysis reveals patterns in Tb variation that have important implications for theories attempting to predict MR patterns in endotherms (Brown et al., 2004; Speakman & Król, 2010). Most commonly, evaluations of these patterns have been based on BMR (Speakman & Król, 2010), which, by definition, is measured in animals within their thermoneutral zones and therefore ignores the metabolic consequences of Tb variation. The many detailed studies of the metabolic consequences of daily and prolonged torpor, as well as the few detailed studies of the metabolic consequences of less pronounced Tb variation (e.g. Bennett et al., 1993; Geiser, 2004), suggest the extent of metabolic variation introduced by Tb variation will overwhelm most of the fine-scale metabolic differentiation focused on in resting MR (RMR) studies. The recent suggestion to use field MR (FMR) instead of BMR (Anderson & Jetz, 2005; Speakman & Król, 2010) is one way to incorporate Tb variation into metabolic comparisons of endotherms, but the key to realizing this opportunity is to incorporate measures of Tb variation into FMR studies, and to design FMR sampling regimes that capture the full range of thermoregulatory patterns expressed by free-ranging animals.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Patterns of mammalian heterothermy appear to be affected by resource availability (i.e. energy needs) and biophysical constraints (i.e. climate means), which is consistent with the hypothesis that fundamental metabolic niches of endotherms are dictated by climate and realized metabolic niches are shaped by resources (Landry-Cuerrier et al., 2008). Although our sample sizes were large, the TS dataset nevertheless represents only ∼10% of extant mammalian species and the HI datasets represent < 1%. Still, our analyses can be used to predict whether unstudied species are most likely to be characterized by low- or high-amplitude heterothermy. The two metrics revealed similar patterns of heterothermy, but some differences emerged in phylogenetic and latitudinal patterns that show each metric is likely to be useful, and the correct metric will depend on the context of the study and the question being asked.

This study offers a global analysis of the thermoregulatory continuum in mammals and points to important modifications of traditional interpretations of body mass–heterothermy and latitude–heterothermy relationships. Specifically, (1) smaller species are not necessarily more heterothermic than larger mammals, but they do show greater interspecific variation in thermoregulatory patterns, and 2) latitudinal gradients are not as straightforward as traditionally assumed, and may even be reversed with more equatorial species displaying more variation in Tb during summer than more polar species.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

We are especially grateful to the many authors that kindly provided raw data from their studies. We thank M. Landry-Cuerrier and J. Samson for help obtaining the climatic variables and O.R.P. Bininda-Emonds for help with the phylogeny. Mark Brigham and two anonymous reviewers provided comments that improved the paper.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information
  • Anderson, K.J. & Jetz, W. (2005) The broad-scale ecology of energy expenditure of endotherms. Ecology Letters, 8, 310318.
  • Angilletta, M.J., Jr, Cooper, B.S., Schuler, M.S. & Boyles, J.G. (2010) The evolution of thermal physiology in endotherms. Frontiers in Bioscience, E2, 861881.
  • Barnes, B.M. (1989) Freeze avoidance in a mammal: body temperatures below 0 °C in an arctic hibernator. Science, 244, 15931595.
  • Bennett, N.C., Jarvis, J.U.M. & Cotterill, F.P.D. (1993) Poikilothermic traits and thermoregulation in the Afrotropical social subterranean Mashona mole-rat (Cryptomys hottentotus darlingi) (Rodentia: Bathyergidae). Journal of Zoology (London), 231, 179186.
  • Boyles, J.G., Smit, B. & McKechnie, A.E. (2011a) Does use of the torpor cut-off method to analyze variation in body temperature cause more problems than it solves? Journal of Thermal Biology, 36, 373375.
  • Boyles, J.G., Smit, B. & McKechnie, A.E. (2011b) A new comparative metric for estimating heterothermy in endotherms. Physiological and Biochemical Zoology, 84, 115123.
  • Brown, J.H., Gillooly, J.F., Allen, A.P., Savage, V.M. & West, G.B. (2004) Toward a metabolic theory of ecology. Ecology, 85, 17711789.
  • Canale, C.I., Levesque, D.L. & Lovegrove, B.G. (2012) Tropical heterothermy: does the exception prove the rule or force a re-definition? Living in a seasonal world: thermoregulatory and metabolic adaptations (ed. by b.y.T. Ruf , C. Bieber , W. Arnold and E. Millesi ), pp. 2940. Springer-Verlag, Berlin, Heidelberg.
  • Careau, V., Réale, D., Garant, D., Speakman, J.R. & Humphries, M.M. (2012) Stress-induced rise in body temperature is repeatable in free-ranging eastern chipmunks (Tamias striatus). Journal of Comparative Physiology B, 182, 403414.
  • Cooper, C.E. & Geiser, F. (2008) The ‘minimal boundary curve for endothermy’ as a predictor of heterothermy in mammals and birds: a review. Journal of Comparative Physiology B Biochemical Systemic and Environmental Physiology, 178, 18.
  • Felsenstein, J. (1985) Phylogenies and the comparative method. The American Naturalist, 125, 115.
  • Fritz, S.A., Bininda-Emonds, O.R.P. & Purvis, A. (2009) Geographical variation in predictors of mammalian extinction risk: big is bad, but only in the tropics. Ecology Letters, 12, 539549.
  • Geiser, F. (2004) Metabolic rate and body temperature reduction during hibernation and daily torpor. Annual Review of Physiology, 66, 239274.
  • Geiser, F. & Ruf, T. (1995) Hibernation versus daily torpor in mammals and birds: physiological variables and classification of torpor patterns. Physiological Zoology, 68, 935966.
  • Gillooly, J.F., Brown, J.H., West, G.B., Savage, V.M. & Charnov, E.L. (2001) Effects of size and temperature on metabolic rate. Science, 293, 22482251.
  • Grigg, G.C., Beard, L.A. & Augee, M.L. (2004) The evolution of endothermy and its diversity in mammals and birds. Physiological and Biochemical Zoology, 77, 982997.
  • Harvey, P.H. & Pagel, M.D. (1991) The comparative method in evolutionary biology. Oxford University Press, Oxford.
  • Heldmaier, G., Ortmann, S. & Elvert, R. (2004) Natural hypometabolism during hibernation and daily torpor in mammals. Respiratory Physiology & Neurobiology, 141, 317329.
  • Humphries, M.M., Thomas, D.W. & Kramer, D.L. (2003) The role of energy availability in mammalian hibernation: a cost-benefit approach. Physiological and Biochemical Zoology, 76, 165179.
  • Krell, F.-T. & Cranston, P.S. (2004) Which side of the tree is more basal? Systematic Entomology, 29, 279281.
  • Landry-Cuerrier, M., Munro, D., Thomas, D.W. & Humphries, M.M. (2008) Climate and resource determinants of fundamental and realized metabolic niches of hibernating chipmunks. Ecology, 89, 33063316.
  • Letunic, I. & Bork, P. (2011) Interactive Tree Of Life v2: online annotation and display of phylogenetic trees made easy. Nucleic Acids Research, 39, W475W478.
  • Lovegrove, B.G. (2000) The zoogeography of mammalian basal metabolic rate. The American Naturalist, 156, 201219.
  • Lovegrove, B.G. (2003) The influence of climate on the basal metabolic rate of small mammals: a slow-fast metabolic continuum. Journal of Comparative Physiology B Biochemical Systemic and Environmental Physiology, 173, 87112.
  • Lovegrove, B.G. (2012a) A single origin of heterothermy in mammals. Living in a seasonal world: thermoregulatory and metabolic adaptations (ed. by b.y.T. Ruf , C. Bieber , W. Arnold and E. Millesi ), pp. 311. Springer-Verlag, Berlin, Heidelberg.
  • Lovegrove, B.G. (2012b) The evolution of endothermy in Cenozoic mammals: a plesiomoprhic-apomorphic continuum. Biological Reviews (Cambridge), 87, 128162.
  • McKechnie, A.E. & Mzilikazi, N. (2011) Heterothermy in Afrotropical mammals and birds: a review. Integrative and Comparative Biology, 51, 349363.
  • McNab, B.K. (2008) An analysis of the factors that influence the level and scaling of mammalian BMR. Comparative Biochemistry and Physiology A, 151, 528.
  • Malan, A. (1996) The origins of hibernation: a reappraisal. Adaptations to the cold: the tenth international hibernation symposium (ed. by F. Geiser , A.J. Hulbert and S.C. Nicol ), pp. 16. University of New England Press, Armidale.
  • Prothero, J. & Jürgens, K.D. (1986) An energetic model of daily torpor in endotherms. Journal of Theoretical Biology, 121, 403415.
  • R Development Core Team (2009) R: a language for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
  • Schmid, J. & Ganzhorn, J.U. (2009) Optional strategies for reduced metabolism in gray mouse lemurs. Die Naturwissenschaften, 96, 737741.
  • Scholander, P.F., Hock, R., Walters, V. & Irving, L. (1950) Adaptation to cold in Arctic and tropical mammals and birds in relations to body temperature, insulation, and basal metabolic rate. Biological Bulletin (Woods Hole), 99, 269271.
  • Speakman, J.R. & Król, E. (2010) Maximal heat dissipation capacity and hyperthermia risk: neglected key factors in the ecology of endotherms. Journal of Animal Ecology, 79, 726746.
  • White, C.R. (2011) Allometric estimation of metabolic rates in animals. Comparative Biochemistry and Physiology A, 158, 346357.

Justin Boyles is a physiological ecologist focusing on the evolutionary and ecological drivers of thermoregulation and energetics in mammals and birds.

Amy Thompson is a graduate student interested particularly in the behavioral ecology of small mammals. This paper comprises part of the research contributing to Amy's M.Sc. thesis.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
geb12077-sup-0001-si.docx32K

Appendix S1 Additional information on methods and statistical procedures.

geb12077-sup-0002-si.xls328K

Appendix S2 The Thermoregulatory Scope dataset and metadata (literature source).

geb12077-sup-0003-si.xls83K

Appendix S3 The Heterothermy Index dataset and metadata (literature source).

geb12077-sup-0004-si.docx678K

Appendix S4 Additional results (Table S1 and Figs S1–S4).

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.