The decoupled nature of basal metabolic rate and body temperature in endotherm evolution

The origins of endothermy in birds and mammals are important events in vertebrate evolution. Endotherms can maintain their body temperature (T b) over a wide range of ambient temperatures primarily using the heat that is generated continuously by their high basal metabolic rate (BMR)1. There is also an important positive feedback loop as T b influences BMR1–3. Owing to this interplay between BMRs and T b, many ecologists and evolutionary physiologists posit that the evolution of BMR and T b must have been coupled during the radiation of endotherms3–5, changing with similar trends6–8. However, colder historical environments might have imposed strong selective pressures on BMR to compensate for increased rates of heat loss and to keep T b constant9–12. Thus, adaptation to cold ambient temperatures through increases in BMR could have decoupled BMR from T b and caused different evolutionary routes to the modern diversity in these traits. Here we show that BMR and T b were decoupled in approximately 90% of mammalian phylogenetic branches and 36% of avian phylogenetic branches. Mammalian BMRs evolved with rapid bursts but without a long-term directional trend, whereas T b evolved mostly at a constant rate and towards colder bodies from a warmer-bodied common ancestor. Avian BMRs evolved predominantly at a constant rate and without a long-term directional trend, whereas T b evolved with much greater rate heterogeneity and with adaptive evolution towards colder bodies. Furthermore, rapid shifts that lead to both increases and decreases in BMRs were linked to abrupt changes towards colder ambient temperatures—although only in mammals. Our results suggest that natural selection effectively exploited the diversity in mammalian BMRs under diverse, often-adverse historical thermal environments. Evolutionary change in basal metabolic rate was not coupled with body temperature across endothermic species, although the mechanism by which birds and mammals adapted to colder environments differed.

The origin of endothermy in birds and mammals are iconic events in vertebrate evolution.Endotherms can maintain their body temperature (Tb) over a wide range of ambient temperatures (Ta) using primarily the heat generated continuously by their high basal metabolic rates (BMR) 1 .There is also an important positive feedback loop in that BMR itself is influenced by Tb 1-3 .Owing to this, many ecologists and evolutionary physiologists posit that the evolution of BMR and Tb must have been coupled during the radiation of endotherms [3][4][5] , changing with similar trends [6][7][8] .However, colder historical environments might have imposed strong selective pressures on BMR to compensate for increased rates of heat loss and to keep Tb constant [9][10][11][12] .Thus, adaptation to cold Ta via BMR increases could have decoupled BMR from Tb and caused different evolutionary routes to the modern diversity in these traits.Here we show that BMR and Tb were decoupled in ~ 90% of mammalian and in ~ 36 % of avian phylogenetic branches.Mammalian BMR evolved with rapid bursts but without any long-term directional trend, whereas Tb evolved mostly at a constant rate and towards colder bodies from a warmer-bodied ancestor.Avian BMR evolved predominantly at a constant rateagain with no trend, whereas Tb evolved with much greater rate heterogeneity than BMR and there has been adaptive evolution towards colder bodies.Furthermore, rapid shifts leading to both BMR increases and decreases were linked to abrupt changes towards colder Ta but only in mammals.Our results suggest that natural selection effectively exploited the diversity of mammalian BMR under diverse, often adverse historical thermal environments.Phylogenetic statistical methods 13,14 now provide us with the opportunity to formally test whether BMR has been linked to Tb or Ta throughout the evolution of birds and mammals.
By accommodating for and identifying heterogeneity in the rate of phenotypic evolution these methods can detect and reconstruct accurate historical evolutionary processes 15 .
We first quantified and compared rates for BMR and Tb evolution along each branch of the time-calibrated phylogenetic trees of birds and mammals (henceforth branch-wise rates, r; see Methods).r measures how fast a trait evolved along an individual phylogenetic branch (r is a rate scalar by which the background rate is multiplied to increase or decrease the pace of evolution).If BMR and Tb were coupled during the evolution of endotherms, the amount of change along phylogenetic branches in both traits should be positively associatedwhere r is high in BMR we expect it to be high in Tb (Fig. 1 b).We tested this prediction against alternative evolutionary scenarios.Firstly, we cannot make any inferences about coupling or decoupling where there is no rate heterogeneity for both BMR and Tb (r = 1 for all branches in the tree for both traits; Fig. 1a).Secondly, we infer decoupled evolution if both traits show rate heterogeneity, but the magnitudes of rs are negatively correlated (i.e.branches evolving at a high rate for BMR are evolving at a low rate for Tb, and vice-versa, Fig. 1c).We suggest this scenario implies decoupled evolution because a negative correlation most likely implies that one trait tends to be conserved whilst the other evolved rapidly.Thirdly, we infer decoupled evolution if only one trait shows rate heterogeneity while the other evolved at a constant rate (Fig 1d and e) or if both traits show heterogeneity but the branch-wise rates are not associated (Fig 1f).
As BMR, body mass (Mass), Tb, and Ta are at least to some extent correlated in extant birds and mammals, and such correlations may vary between orders 16 , we estimated the branch-wise rates for BMR and Tb while accounting for their covariates across extant species using the phylogenetic variable-rates regression model 17 (henceforth variablerates; Methods).This approach allows for simultaneous estimation of both an overall relationship between, for instance, BMR as a function of Mass and Tb across extant species, and any shifts in rates (r) that apply to the phylogenetically structured residual variance in the relationship.In both birds and mammals, the variable-rates model significantly fits the data better than the constant-rate regressions, which assume a single rate across all branches (Methods; Table S1 to S8).The best fitting variable-rates model for mammalian BMR includes both Mass and Tb with a single slope for each trait estimated across all orders (Table S1 and S2).For mammalian Tb, the best fitted model includes Mass and BMR as covariates, also with a single slope across all orders (Table S3 and S7).In birds, the best model for BMR includes only Mass with a single slope for all orders (Table S4).Finally, the best fitted model for avian Tb includes Mass only in Columbiformes (Table S6).
The branch-wise rates estimated in the best fitting models shows that mammalian BMR evolved at a constant rate (r = 1) in just 11.2% of branches and at faster rates (r > 1) in 88.8% of branches (Fig. 2a).Mammalian Tb evolved at a constant rate in 70.3% of branches and faster rates in 29.7% of branches (Fig. 2b).In birds, BMR evolved at a constant rate in 90.5% of branches and at faster rates in 9.5% of branches (Fig. 2d).
Avian Tb evolved at a constant rate in 69 % of branches and at faster rates in 31% (Fig. 2e).When the branch-wise rates for BMR and Tb were compared, we found that in mammals, both traits evolved at a constant rate in 10.6% of branches (Fig. 3a consistent with Fig. 1a).In 60.2% of branches only one trait evolved at faster rates while the other trait diverged at a constant rate.This indicates that BMR and Tb evolved in a decoupled fashion along these branches (Fig. 3a consistent with Fig. 1d, e).We found that 29.2% of branches had an increased rate in both BMR and Tb.However, the magnitudes of the branch-wise rates were not significantly correlated (pMCMC [% of posterior distribution crossing zero] = 9%; Table S9; Fig. 3a consistent with Fig 1f).This also suggests decoupled evolution in those brancheslikely because of distinct selection pressures acting on BMR and Tb.On the other hand, both traits evolved at a constant rate in 63.8% of branches for birds (Fig. 3c consistent with Fig. 1a).In 32% of branches only one trait evolved at fast rates while the other trait diverged at a constant rate (Fig. 3c consistent with Fig. 1d, e).In the remaining 4.2% of branches, both traits evolved at faster rates, but the r magnitudes were not statistically correlated (pMCMC = 16.9%,Table S10, Fig. 3c consistent with Fig. 1f).
As rapid bursts in BMR evolution were not coupled with those in Tb evolution, we evaluated the alternative hypothesis postulating that BMR evolved in response to Ta.This hypothesis suggests that colder environments increase the rate of heat lost from organisms which is subsequently compensated by BMR increases [9][10][11][12] .These BMR increases could have occurred over long periods of time because of global cooling 18 generating a long-term directional trend in BMR during the radiation of mammals and birds.This expectation is in line with the Plesiomorphic-Apomorphic Endothermy Model [6][7][8] (PAE Model).By assuming that BMR and Tb are coupled in endotherms and that they both can be used as a proxy of the degree of endothermy, the PAE model predicts a general tendency towards higher endothermic levels through time (from basoendothermic ancestors, Methods) associated with the Cenozoic global cooling.Global cooling is not the only source of variation in Ta.Long-term directional increases in BMR might have also been driven by historical dispersals of endotherms towards higher latitudes 19 .In either case, if a long-term decrease in Ta drove adaptation via BMR elevation, and Tb followed the same trajectory (as assumed by the PAE model) we expect to find a positive correlation between the branch-wise rates of BMR and the branch-wise rates of Ta.With this in mind, we also expect a positive trend towards higher BMR and Tb values from basoendothermic ancestors and a negative trend towards lower Ta from warmer ancestral environments.We used the variable-rates model to estimate the branch-wise rates for Ta whilst accounting for latitude since, generally, Ta decreases from the equator to the poles (Methods; Table S11).
The variable-rates model significantly improved the fit to the Ta data over the constantrate regression model in both mammals and birds (Table S11).In 21.2% of mammalian branches Ta evolved at a constant rate, and with rate heterogeneity in the remaining 78.8%including 72.2% of branches with faster rates and 6.6% with slower rates (r < 1, Fig. 2c).This indicates that most ancestral mammalian lineages (72.2%) faced abrupt historical changes in their Ta, while far fewer lineages (6.6%, mostly bats) survived and continued existing in similar thermal environments.In birds, 77.6% of branches show faster rates of Ta change, 22.1% show changes at a constant rate, and in only a single branch the Ta changed at a slower rate (Fig. 2f).
When branch-wise rates of mammalian BMR and Ta evolution were compared, we found that they were coupled in 74.9% of branches (pMCMC = 0%; Table S12; Fig. 3b, consistent with Fig. 1b).To evaluate further if Ta decreases were linked to BMR increases in the 74.9% of mammals where both traits were coupled (i.e. to ascertain the direction of change), we evaluated the expected positive trend in BMR as a response to the long-term decrease in Ta.We conducted Bayesian phylogenetic regressions between extant values of these two variables (in turn) and the path-wise rates (sum of branch-wise rates along branches in the path from the root of the tree to each terminal species, Methods) 15 .We found a negative effect of path-wise rates on Ta across all mammals (Fig. 4b; Table S14), which supports a long-term directional trend towards habitats with lower Ta over time.
However, we did not find evidence for any trend in mammalian BMR evolution -BMR increases and decreases were equally likely in our sample (Table S14).Our results suggest that in colder environments, where resources were available to fuel metabolic elevation, selection favoured higher mammalian BMR 20 .Another possibility might be that BMR increase was a correlated response to direct selection on other physiological traits, like maximal metabolic capacities for thermogenesis, whose benefits outweigh the energetic cost of BMR elevation 20 .Otherwise, selection may have always favoured BMR decreases under an ever colder environment 20 .
In contrast to mammals, most avian branches that experienced rapid shifts in Ta did not show evidence of coupled changes in BMR -68.4% of branches had fast rates of Ta evolution but a constant rate of BMR evolution (Fig. 3d consistent with Fig. 1d, e).
Moreover, the small fraction of branches where BMR evolved at fast rates (9.5%) were not linked to rapid shifts in Ta (Fig. 3d consistent with Fig. 1f; Table S13).Avian BMR did not show a positive evolutionary trend despite the fact they also experienced colder environments over time (Fig. 4d; Table S15).Birds might not have responded to colder temperatures by changes in their BMR because their lower thermal conductance might have helped them retain internal heat 9 .Alternatively, other physiological strategies, such as torpor, may have been selected for under colder environments 21 .
Finally, we found a negative effect of path-wise rates on Tb in both mammals (Fig. 4a; Table S14) and birds (Fig. 4c; Table S15).This suggest thaton averageendotherms evolved towards colder bodies from warmer-bodied ancestors.These directional models predict a mean Tb of 35.3 °C and 40.4 °C in the most recent common ancestor (MRCA) of mammals and birds respectively (Fig. 4a, c), suggesting that early birds and mammals were mesoendotherm rather than basoendotherms (Methods).This result does not support that ancestral mammals could not attain Tb > 30 °C owing to the elevated metabolic rates necessary to compensate heat loss in cold environments 22 .However, if the Tb-Ta differential (ΔT) determines how hot early mammals were, we expect that the mammalian MRCA with a Tb of 35.3 °C could survive in an environment warm enough to have a low ΔT.Our model describing the negative trend in Ta predicts that the MRCA of mammals lived in an environment with 23 ºC on average (Fig. 4b), resulting in a ΔT of 15.3 ºC.This ancestral ΔT is very conservative compared with the ΔTs observed in extant mammals.For example, there are small mammals that achieve Tb higher than 39 °C (e.g. 16) that can survive in environments of 11 °C19 (ΔT = 28 ºC).Also, some larger mammals have stable Tb even in extreme environmental conditionsthe Artic hare (Lepus arcticus) can maintain its Tb of 38 °C16 in temperatures as low as -Taken together, our results reveal that BMR was not coupled with Tb across the evolution of endothermic species.As environments became colder, mammals survived by changing their BMR, while birds likely survived owing to their high thermal insulation.Evaluating the isolated and/or combined effect of environmental variables on physiological attributes has implications for evidence-based projections for the future 23 .In this sense, the previously unappreciated complexity, interplay and decoupled nature in the evolutionary history of BMR, Tb and Ta might point to undetected resilience of endotherms in the face of modern global challenges.over their evolutionary history.Path-wise rates had a significant negative effect in mammalian and avian Tb (pMCMC = 4% and 3%; n = 502 and 367 species) and in mammalian and avian Ta (pMCMC = 0 and 0; n = 2922 and 6142 species), both supporting a negative macroevolutionary trend 15 .Transparent and dark lines indicate the posterior distribution of slopes and the mean slope respectively, estimated from the Bayesian PGLS (Methods).

Methods.
Data.We used a time-calibrated phylogenetic tree of extant mammals (n = 3321) 24 , and the body mass (M), basal metabolic rate (BMR), and body temperature (Tb) taken from Clarke et al. 16 (n = 632).After identifying species in the tree that have trait information, we obtained a final mammalian dataset of 502 species, which includes representatives from 15 orders (SI).
For birds, we used the consensus time-calibrated tree from Rolland et al. 19 .This tree was inferred from the samples of trees provided by Jetz et al 25 .Data for BMR, Tb, and Mass were obtained from Fristoe et al 9 .After matching this database with the phylogenetic tree, we obtained a final sample of 164 species which includes representatives from 21 orders (SI).The dataset used to evaluate evolutionary trends in Tb (see below) is from Clarke & Rothery 26 , which contains 367 species with phylogenetic information.
Data for ambient temperature (Ta) and latitude for extant mammals and birds was extracted from Rolland et al. 19 These datasets include 2922 species of mammals and 6142 species of birds which have phylogenetic information.The Ta for extant endothermic species is the temperature of environments in which birds and mammals inhabit todaymeasured as the mean ambient temperature for the mid-point latitude of each species distribution (Rolland et al. 19 ).The Ta at which a species exists today may not be a heritable trait per se.However, the evolution of Ta can still be inferred using phylogenetic methods since habitat selection reflects species adaptations (traits) to some characteristics of the environment.This interrelationship should leave phylogenetic signal in the Ta at which endothermic species live.Accordingly, we found significant phylogenetic signal in the Ta of both mammals (PosteriorMean = 0.77; Bayes Factor = 665) and birds (PosteriorMean = 0.8; Bayes Factor = 1404).Furthermore, the phylogenetic signal for Ta is very high (=1) in birds and mammals, when estimated using the median-r scaled tree.
Inferring the branch-wise rates of evolution.We identified heterogeneity in the rate of We estimated the r values of BMR, Tb, and Ta evolution using the phylogenetic variablerates regression model in a Bayesian framework 17 .This model is designed to automatically detect shifts in the rate of trait evolution across phylogenetic branches while accounting for a relationship with another trait or traits across extant species values.This approach allows for simultaneous estimation of both an overall relationship between, for instance, BMR as a function of Mass and Tb across extant species, and any shifts in rates (r) that apply to the phylogenetically structured residual variance in the relationship.As residual variance is explained by shifts in rate across phylogenetic branches we can, for example, determine how much BMR has changed in the past (r) after accounting for their covariation with Mass and Tb in the present (the relationship between the values across extant species).Thus, if the amount of BMR change along individual phylogenetic branches were coupled with the amount of change of Tb, then we should find the r values of BMR to be positively associated with the r values of Tb.The branch-wise rates for Tb evolution can be estimated while accounting for its covariation with other traits or factor across extant species.Previous studies on the association between BMR and Tb using extant species values alone have not evaluated the association in evolutionary terms even when they use phylogenetic method.
We evaluated 24 phylogenetic variable-rates regression models and 24 phylogenetic constant-rate regression models (Table S1 to S8).Regression model selection was conducted using Bayes Factors (BF) via marginal likelihoods estimated by stepping stone sampling.BF is calculated as the double of the difference between the log marginallikelihood of the complex model and the simple model.By convention, BF > 2 indicates positive evidence for the complex model, BF 5-10 indicates strong support, and BF > 10 are considered very strong support 27 .We inferred the r values of BMR and Tb with the phylogenetic variable-rates regression models that best fit the data for our samples of mammals and birds (Table S7 and S8).We also estimated the r values for Ta after accounting for the effect of latitude of species distribution (Table S11) and, consequently, we accounted for the geographic variation of Ta across extant species distributions.We used BayesTraits v3.0 28 to detect the magnitude and location of r in a Bayesian Markov chain Monte Carlo (MCMC) reversible-jump framework, which generates a posterior distribution of trees with scaled branches lengths according to the rate of evolution.There is no limit or prior expectation in the number of the r branch-scalars, r numbers vary from zero (no branch is scaled) to n, where n is the number of branches in the phylogenetic tree.Regarding the values of each r parameter, we used a gamma prior, with  = 1.1 and  parameter rescaled in order to get the median of the distribution equal to one.With this setting, the numbers of rate increases and decreases proposed is balanced 13 .We ran 50,000,000 iterations sampling every 25,000 to ensure chain convergence and independence in model parameters in BMR and Tb analyses.We discarded the first 25,000 iterations as burn in.For the Ta analysis in mammals we ran 200,000,000 iterations sampling every 100,000, and we discarded the first 100,000 iterations as burn in.For Ta analysis in birds we ran 400,000,000 iterations discarding the first 100,000,000 as burn in, and we sampled every 200,000.Regression coefficients were judged as significant according to a calculated pMCMC value for each posterior of regression coefficients: where < 5% of samples in the posterior distribution crossed zero, this indicates that the coefficient is significantly different from zero.

Figure 1 .
Figure 1.Possible evolutionary scenarios between BMR and Tb given their branch-

Figure 2 .
Figure 2. Branch-wise rates (r) of BMR, Tb, and Ta on the mammalian and avian

Figure 3 .
Figure 3. Branch-wise rates (r) of BMR, Tb, and Ta in bivariate space for mammals

Figure 4 .
Figure 4. Mammals (a, b) and birds (c, d) evolved towards both colder Tb and Ta evolution along phylogenetic branches (branch-wise rates) by dividing the rate into two parameters: a background rate parameter ( 2 b) which assumes changes in the trait of interest (e.g.BMR) are drawn from an underlying Brownian process, and a second parameter, r, that identifies a branch-specific rate shift.A full set of branch-wise rates are estimated by adjusting the lengths of each branch in a time-calibrated tree (stretching or compressing a branch is equivalent to increasing or decreasing the phenotypic rate of change relative to the underlying Brownian rate of evolution).Branch-wise rates are defined by a set of branch-specific scalars r (0 < r < ) which transform each branch in order to optimize the phenotypic rate of change to a Brownian process ( 2 b r).If phenotypic change occurred at accelerated (faster) rates along a specific branch of the tree, then r > 1 and the branch is stretched.Decelerated (slower) rates of evolution are detected by r < 1 and the branch is compressed.If the trait evolves at a constant rate along a branch, then the branch will not be modified (i.e.r = 1).