Clockwork Orangutan: microRNAs, thermoregulatory tradeoffs, and models of brain size evolution

Brain and liver are the two most energy-demanding tissues in mammals, each accounting for ~20% of resting metabolic rate (rMR) in humans (6% and 52%, respectively, in mouse) (Elia 1992; Wang et al 2012; Kummitha et al, 2014). The size of the human brain represents a four-fold increase relative to those of other great apes, which in turn are three-fold larger than the brains of gibbons (Smaers et al, 2021; Grabowski et al, 2023; Shao et al, 2023). Since the sizes of the four most thermogenic organs (kidney, heart, liver and brain) scale more or less strictly according body size (Wang et al, 2012), an increase in the absolute size of the brain cannot be accounted for by a decrease in any of the other three organs (“expensive tissue hypothesis” of Aiello and Wheeler, 1995) (Navarette et al, 2011). Consequently, for a given primate size and resting MR (rMR), the only way to account for the greater lifetime energy demand of a larger brain is to postulate gains in the reliability of the acquisition and assimilation of food, storage of energy reserves, gestation, lactation and infant care (Mascia-Lees et al, 1986; Koteja, 2000; Fonseca-Azevedo and Herculano-Houzel, 2012; Isler and van Schaik, 2012; Isler and van Schaik, 2014; Kuzawa et al, 2014; Pontzer et al, 2016; Simmen et al, 2017; Grabowski et al, 2023; for review, see Heldstab et al, 2022). However, any account of the change in energy balance associated with a larger brain would not be complete without an explanation of the associated change in thermoregulation: for a given body size and rMR, a larger brain implies more rapid internal heat generation accompanied by faster blood perfusion in order to maintain temperature homeostasis (Nybo et al, 2002; Kiyatkin, 2007; Wang et al, 2014). Interestingly, a model of the energy cost of action potential generation predicts a global minimum between 37º - 42ºC (Yu et al, 2012). 1 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint A comprehensive biophysical model of thermoregulation in homeothermic mammals envisions a thermogenic core, on the one hand and, on the other, a periphery combining the properties of transport, diffusion, convection and radiation (Porter and Kearney, 2009). This model predicts that for a mammalian rMR that scales as M0.75, the thermogenic core will scale as ~M 0.50 (Porter & Kearney, 2009). We postulated that the combination of four thermogenic organs of mammals serves as a thermoregulatory core (“core model”), estimated organ-specific cellular MRs (cMRs) consistent with physiological and phylogenetic models, and found that this joint “core” cMR scaled as predicted (M 0.52). We demonstrate that across Euarchontoglires, increases in the relative proportion of the brain in the core are offset by decreases in the relative proportion of the liver. Since the liver is the primary organ of gluconeogenesis, this observation is compatible with the energy reliability theory of brain size evolution. Physiologically-based cMRs do not correspond to log-based sMRs Assuming a power law model, we distinguish between specific MRs (sMRs): sMR = m x Mb, where b is the slope of log rMR/log M, and cellular MRs (cMRs): cMR = n x Mc, where c is the exponent of mass obtained by maximum likelihood fitting to the distribution of a physiologically-relevant trait, such as the number of microRNA families (mirFam ). Previously, we demonstrated that, unlike sMRs, cMRs are sensitive to body temperature sample differences in the range of mass (Fromm and Sorger, 2023). Figure 1. Log-based sMRs do not correspond to mirFam-based cMRs. Data were drawn from Fromm and Sorger (2023). Hatched lines represent the log-based slopes of the birds (sMR.64) and the mammals (sMR.60). For the purpose of illustration, a difference in Tb of 0.2º C was applied between pigeon and finch. Note the inversion of cMR slopes. 2 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Rates of cellular metabolism depend on traits that do not scale logarithmically, such as body temperature (T b) or the rate of protein synthesis, for which the species-specific number of microRNA families (mirFam) may serve as proxy (Fromm and Sorger, 2023). Furthermore, cMRs will vary by tissue type and differ between tissues that are renewable or post-mitotic. The rank order of cMRs can be matched to non-logarithmic traits by variable allometric scaling, while the rank order of sMRs is fixed according to species size (Figure 1). Cellular MRs determine not only the overall slopes of these relationships, but also the order of cMR/trait ratios (Figure 2), with implications for models of genetic and phenotypic covariance (Lande, 1979; Felsenstein, 1985; Hansen, 1997; Pagel, 1999; Svensson et al, 2021; Walter and McGuigan, 2023). Figure 2. Cellular MRs determine the species rank of cMR/mirFam ratios. Primate species were labeled in order of magnitude of cMR.75/mirFam, and their distributions plotted as cMR was varied. Note the changes in rank order in addition to the changes in overall slope. In order to develop a thermogenic core model of thermoregulation, we based our estimates of cMRkdn, cMRhrt, cMRlvr and cMRbrn in primates on three criteria: a) size-independent correlation with mirFam (physiological relevance); b) correspondence of the combined cMR of these organs with that of the thermogenic core predicted in the biophysical model of Porter & Kearney (2009) (thermodynamic relevance); c) utility for the unsupervised classification of major primate clades (evolutionary relevance). 3 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint We then examined how tradeoffs in the core between brain and liver were related to the divergence of orders and species of Euarchontoglires, and to the corresponding changes in body size and number of microRNA families. Finally, we demonstrate how a logistic model of the relationship between brain size and energy reliability supports the inference of size adaptation to a low energy regime by reduction in size, on the one hand, and thermoregulatory adaptation to the physiological limit on the rate of heat dissipation (hence rate of thermogenesis) in large-brained primates, on the other.

AND DISCUSSION The proportion of brain mass in the core varies reciprocally with that of liver Of six possible tradeoffs In a core consisting of four thermogenic organs, one major trade off occurs consistently among primate clades, that between brain and liver (Figure 1). A dampened tradeoff between heart and kidney occurs approximately in phase with the major tradeoff. Figure 3. The proportion of brain mass in the core varies reciprocally with that of the liver. Brain and liver proportions in the core vary reciprocally across 30 additional Euarchontoglires, together with a diminished reciprocity between heart and kidney (Supplementary Figure S1). The observations in primates are summarized geometrically in Figure 4. It is noteworthy that while relative core size (~ M0.50) is predicted to decrease with body size (~ M0.75) (Porter & Kearney, 2009), the relative core size in humans is greater than that in baboons, which have less than half their body mass. 4 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Figure 4. Variation across primates in core size and composition. The radius of each circle represents a spherical volume corresponding to organ or core mass, scaled to body size (equal circle diameters). Estimation and physiological relevance of thermogenic organ cMRs We have previously presented evidence that the number of microRNA families is a statistically and mechanistically credible proxy for the global rate of protein synthesis, which drives 20% of cellular ATP turnover in mammalian cells (Buttgereit and Brand, 1995). We further showed that mirFam is an attractor for whole animal cMR across the major mammalian clades (Monotreme, Atlantogenata, Laurasiatheria and Euarchontoglires) in a model of stabilizing selection (Fromm and Sorger, 2023). Absent other selective forces, we expected that the diversification of Euarchontoglires would be associated with a steady increase in cMR variance with respect to the distribution of mirFam, owing to genetic drift and the relaxation of selection. Figure 5 compares the dependence of cMR for each organ on organ size and mirFam, together with the log-based sMR for each organ. Note the similar trends and inversions among Kidney, Heart and Brain, on the one hand, and Liver and Body, on the other. Figure 5. Dependence of cMR on √Morg and mirFam in 23 primates. √M, a natural scale for cMRs, yields P values several orders of magnitude lower than a log scale. Log-based sMR slopes are shaded green, cMR slopes are shaded pink (see text for detailed explanation). 5 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Figure 6 summarizes these results and compares them to regressions on 27 species of Glires. The Y-axis is an ascending scale of P values, which can be interpreted as the degree of release of the constraint of the association with mirFam, hence the following order of evolutionary conservation: Kdn, Hrt < Brn < Lvr, Bdy. This series comports well with the relative “amplitudes” in the periodic pattern of variation in relative core proportion (Figure 1). Furthermore, the shift to a lower range of P values from Glires to Primates is consistent with a shorter time frame for the operation of genetic drift. Comparing the plots for Glires and Primates, there is a leftward shift in the dynamic ranges for liver, and brain cMRs, i.e. towards lower allometric exponents, indicating greater dependence on surface area (Figure 6). Given the higher size range among primates, this shift may signal a thermoregulatory adaptation corresponding to a faster rate of heat dissipation. This interpretation is supported by the observation that morphological types, a function of aspect ratios, were significantly related to the number of microRNA families, but not to genome size or the diversity of protein domains (Deline et al, 2018). Figure 6. Dependence of cMRorg on √Morg and mirFam yields a coherent pattern of variation. Curves were fit by the spline method (lambda = 0.00). Note the difference in P scales. The primate dataset omitted the largest animal and a small heterotherm (Homo sapiens and Cheirogaleus medius). Most striking is the inversion of the relationship with mirFam of liver cMRs (lower P values associated with higher cMRs), compared to those of kidney, heart and liver (Figure 4). The cMRbdy ~ Mbdy correlation tracks that of liver, implying that liver is the organ most affected by thermoregulatory adaptation to changes in animal size. Assuming an operating range of cMR.70 - cMR.80 for core and whole body, we estimated the relevant organ cMRs as cMR.70kdn, cMR.70hrt, cMR.70brn and cMR.80lvr, corresponding to the lowest P values in this range, and tested the validity of this combination against thermodynamic and phylogenetic models. 6 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint The biophysical model for thermoregulation of Porter & Kearney (2009) envisions a heat-generating “core” and an ellipsoid surround combining the properties of transport, convection, conduction and resistance. For a steady state resting MR that scales as M0.75, the corresponding rate of core thermogenesis is predicted to scale as ~ M0.50 (Porter & Kearney, 2009). As a test of thermodynamic relevance, we compared the predicted allometric scaling of theoretical ‘cMRcor’s consisting of the combination of four thermogenic organs, using either log-based sMRs or mirFam-based cMRs: sMR.Xcor = (rMRkdn + rMRhrt + rMRlvr + rMRbrn) / (Mkdn + Mhrt + Mlvr + Mbrn)X = (a x sMR.90kdn) + (b x sMR.86hrt) + (c x sMR.69lvr) + (d x sMR.84brn) or cMR.Xcor = (a x cMR.70kdn) + (b x cMR.70hrt) + (c x cMR.80lvr) + (d x cMR.70brn) Figure 7. The thermodynamic model of cMRcor corresponds to release from mirFam constraint. A. % variation of linear models of cMRcor predicted by combinations of log-based sMRs or mirFam-based cMRs of four thermogenic organs for 22 primates (SIMPLS method of partial least squares; deJong, 1993). Note the different scales for the sMR and cMR models. B.% variation of dependence of the ratio cMR.Ncor cMR.75bdy on mirFam (51 Euarchontoglires). The sMR model predicted a most probable allometric scaling for this 4-organ core of M0.77 (Figure 7), which corresponds to the crossover point for the inversion of slope with respect to mass between cMR.70 and cMR.80 (Figures 2 and 5). In contrast, the cMR model predicted two possible allometries, one on either side of this inversion (Figure 5), including one that closely matched the predicted allometric scaling of a thermogenic core in the ellipsoid model of thermoregulation (Porter & Kearney, 2009): cMR.52cor = 0.53 x cMR.70kdn + 0.38 x cMR.70hrt + 0.36 x cMR.80lvr + 0.09 x cMR.70brn 7 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint When regressed against mirFam, ratios of cMR.Ncor/cMR.75bdy exhibit a sharp minimum at N = 0.52, reinforcing our interpretation that the optimal thermodynamic model corresponds to a release from the constraint of the association with mirFam. As a further test of the proposed composition of cMRcor, we varied the combinations of organ cMRs and found that mirFam-based cMRs also generated the cMRcor corresponding most closely to that predicted by the biophysical model of Porter and Kearney (2009) (Figure 5). Figure 8. Comparison of core organ cMR combinations with respect to the allometric scaling of the corresponding core cMR. Each matrix represents one combination of organ cMRs, and each trace represents the % variation (R 2) for the predicted cMRcor (as in Figure 7). Two combinations (yellow shading) yielded the lowest optimal core allometric exponent, M0.52: [cMR.70kdn, cMR.70hrt cMR.80lvr, cMR.70brn] and [cMR.70kdn, cMR.70hrt cMR.85lvr, cMR.70brn] If our conjecture is correct regarding the relationship between evolutionary constraint and conservation of cMR correlations with mirFam, then the set of core organ cMRs that we have estimated should also be useful in the unsupervised classification of major primate clades. Expectation maximization of Gaussian (normal) mixture models allows the inference of a latent grouping variable, in this case phylogenetic clade. We applied this clustering method to the same 22 primates as in the previous analyses, After varying the cMRs for liver, the model with the lowest information criterion was again [cMR.70kdn, cMR.70hrt, cMR.80lvr, cMR.70brn] (Figure 6). 8 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Figure 9. Unsupervised classification of primate clades based on core cMR composition. Principal components were calculated from each cMR correlation matrix, and membership in two clusters estimated by expectation maximization of their joint probability distributions. Kidney, heat and brain cMR = cMR.70; liver cMRs varied between cMR.75lvr - cMR.85lvr (bold type). Ellipses were hand-drawn to include all members of the three phylogenetic clades (N = 22). While cMRs were needed to develop and test the core cMR concept, tradeoffs between liver and brain sMRs are also readily demonstrable with log-transformed data, provided that they ranked according to core size, not body size. Figure 10 compares the variation in organ size as a function of body mass (Figure 10A) or core mass (Figure 10C). Examination of the residuals of the corresponding regressions reveals a periodic pattern of variation for each organ, corresponding to their relative proportion of body mass (Figure 10B) or core mass (Figure 10D). The periodicity is a function of autocorrelation, as can be easily demonstrated when the analysis is limited to a single pair of organs (Figure 11). A consistent tradeoff between brain and liver extending across the entire range is apparent only when organ masses are standardized against core mass (Figure 10D versus Figure 10B). In other words, standardization of log Mbrn against log Mbdy precludes detection of this tradeoff (Navarette al, 2011). Tradeoffs in the evolution of organ cMRs conform to the core model of thermoregulation. The evolutionary histories of the four organ cMRs provides compelling evidence that they have been constrained by their relationship to the core. We examined how the tradeoffs between brain and liver were related to the divergence of orders and species of Euarchontoglires, and to the corresponding changes in body size and number of microRNA families (Figure 12). While the cMR.80lvr tree mirrored that of body size, the tree for cMR.70brn largely resembled that of mirFam (Figure 12), exhibiting a sharp contrast between the relative arrest in the rate of evolution of cMR.70brn upon the divergence of Glires, and acceleration leading up to the divergence of Catarrhini. Furthermore, relative to the Haplorrhini, the evolution of cMR.70brn was relatively arrested in the Lemuriform (Strepsirrhini) clade (Figure 12). The inference of an exceptional rate of brain size increase in humans attests to the validity of the variable rate model’s default expectation of a Brownian process of genetic drift (Pagel et al, 2004). 9 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Figure 10. Reciprocal variation between liver and brain is a function of core size. A and C: Variation of organ size relative to body size and core size in primates. B and D: Residual variation relative to body size and core size in primates. 10 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Figure 11. Tradeoffs between liver and brain as a function of their combined mass. The “phase shift” between the residuals of liver and brain with respect to the core (Figure 7D) can be isolated by repeating the regression with respect to Log (Mlvr + Mbrn), demon- strating that the periodicity depends on autocorrelation. Figure 12. Time tree and variable rate phenograms of Log 2Mbdy, mirFam, cMR.80lvr and cMR.70brn. The node values for Log 2Mbdy were also applied to the time tree, in order to contrast the branch lengths. Bayes Factors for the variable rate trees ranged from 15.18 (Log 2M) to 37.94 (mirFam ). The two smallest Platyrrhini, one of which (Callithrix pygmaea) is a suspected heterotherm (T attersall, 2012), exhibited exceptional increases in cMR.80lvr, consistent with adaptation to a low energy regime 11 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint and increased reliance on gluconeogenesis. However, the most striking contrasts are the opposing cMRlvr - cMRbrn tradeoffs between Primates and Glires (Figure 12). Heart cMR.70 generally tracked the difference between Glires and primates in body size evolution, with a contrast in evolutionary rates between an acceleration among Catarrhini and the two largest Glires (Lagomorphs), on the one hand, and the relative arrest among the smaller Glires, on the other (Figure 13). Relative kidney cMR.70 levels generally corresponded to those of heart, while the relative rates of evolution were generally diminished except for the accelerated rate for human cMR.70kdn, suggesting adaptation to an altered landscape of opportunities for hydration versus risks of dehydration. Figure 13. Variable rate phenograms comparing the evolution of kidney and heart cMRs to the evolution of body size. 12 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint SUMMARY AND CONCLUSIONS Linearization of distributions of body mass and metabolic rate by log transformation has been de rigueur for phylogenetic comparative studies. Unlike log-derived sMRs, the cMRs of core thermogenic organs can be estimated according to physiological, thermodynamic and phylogenetic criteria. They can also be linearized with respect to the corresponding rate of thermogenesis of a four-organ core (Figure 14). Only the latter method reveals the inverse relationship between brain and liver cMRs. Countless comparative studies of whole body rMR in mammals have yielded sMR exponents ~ 0.75. Our analysis reveals that cMR.75 corresponds approximately to the midpoints of the dynamic ranges of the mirFam correlations with cMRlvr ~ √Mlvr and cMRbrn ~ √Mbrn (Figure 6). Importantly, the interval between cMR.70 - cMR.80 represents the region of minimal mirFam constraint on the core cMR (cMR.75cor). In other words, “Kleiber’s Law” for the log-based allometric scaling of mammalian rMR (Kleiber, 1934; Balllesteros et al, 2018) corresponds to the zone of weakest association between the core rate of thermogenesis and the global microRNA milieu. The recent addition of hundreds of primate genome sequences (Kuderna et al, 2024) has been accompanied by ever more fine-grained studies of the covariance of traits suspected to influence the evolution of brain size. For the reasons elaborated in the Introduction, we submit that a fundamental assumption underlying these analyses – that brain energy expenditure varies with log-transformed brain mass or volume in a manner consistent with thermoregulatory constraints – is not justified. Figure 14. The thermodynamic origin of tradeoffs among core organ cMRs. A. Species organ cMRs are proportional to the predicted species rate of core thermogenesis. B. Linear regressions of individual organ cMRs on the cMR of the model thermogenic core. Of the six possible tradeoffs in cMR among the four thermogenic organs, why is the brain-liver tradeoff the most significant? In a prolonged state of fasting, the liver is the primary organ of gluconeogenesis: 13 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint the glucagon-dependent reduction of amino acid backbones fueled by free fatty acids released from adipose tissue. If an evolutionary increase in reliability of the energy supply meant a decrease in the frequency of prolonged fasting, hence reliance on gluconeogenesis, retaining the ancestral liver size and cMR would have been inefficient. This dynamic would have been reinforced if brain enlargement meant greater reliability of the energy supply. We propose the following example of a cMR-based model of brain size evolution based on a simple feedback loop between brain size and the reliability of the energy supply: improvements in energy reliability facilitate increases in brain size, which in turn favor further improvements in energy reliability, up to the limits imposed by anatomy and physiology e.g. the diameter of the birth canal and the physiological limit on the rate of heat dissipation (Speakman and Król, 2010). This neutral model reduces several current hypotheses (Expensive Brain/Ecological Brain/Cognitive Buffer) (Heldstab et al. 2022) to one that does not presuppose any underlying mechanism. It is a classic logistic model, with cMRmax standing in for carrying capacity in a population logistic model: d(cMRbdy) / d(Vbrn ) = r • cMRbdy [1 - (cMRbdy / cMRmax)] where r is a rate constant and the range of cMRbdy has been standardized to 1.00. We applied this model to the primate brain size dataset of Grabowski et al (2023), fitting it to mirFam, a proxy for cMRbdy i.e. the overall rate of energy utilization (Figure 15A). Unexpectedly, the model revealed that relative brain size is a predictor of mirFam, rather than vice versa, lending credence to the proposition that larger brains were associated with improvements in the reliability of the energy supply. Given that the relative sizes of the three other thermogenic organs also fit logistic models for mirFam (Supplementary Figure S2), the question arises as to which relationship is the principal driver of adaptation. In phylogenetic regressions for organ cMRs ranging from cMR.60 - cMR.80, kidney, heart and liver cMRs varied independently of phylogeny (Pagel’s lambda = 0), while the variation of brain cMR.60 - cMR.80 with respect to mirFam was secondary to phylogenetic distance (Pagel’s lambda = 1 in each case), suggesting that the divergence of primate species largely coincided with changes in the relationship of mirFam to brain mass and cMR (Supplementary T able S3) (Pagel, 1999). Returning to the logistic model based on relative brain size, a variable rate phenogram for primate species was generated using the ratios from a Hubbert linearization of the logistic function (Figure 15B) (Hubbert, 1982). Thus, the color scale represents the rank on the logistic curve which, in general, tracks body size (compare with the Log 2M panel in Figures 12 and13). The three smallest species, a Lemuriform and a pair of Callitrichinae, exhibited an accelerated rate of descent in rank. Cheirogaleus medius is a heterotherm, and Callithrix pygmaea may also be heterothermic (T attersall, 2012), 14 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint indicating that the baseline of the logistic function represents a regime of minimal energy reliability, perhaps related to low calorie-density folivory or insectivory. Following this logic, we can ask what sort of constraint might be encountered near the plateau of this model? A likely candidate would be the physiological limit on the rate of heat dissipation (Speakman and Król, 2010; Fromm and Sorger, 2023). During exercise with progressive hyperthermia, human brain temperature increases in parallel with that of the body core, leading to hyperthermia and fatigue (Nybo, 2012; Nybo and Gonález-Alonso, 2015; T suji et al, 2016). Over a period spanning 700,000 years of human evolution, body mass and shape varied inversely with temperature, indicative of selection in response to thermoregulatory demands (Stibel, 2023). Figure 15. Logistic models of brain size evolution (data from Grabowski et al, 2023). A, C. For these idealized models (R 2 = 1.00), mirFam or mirFam•Tb(ºC) was that predicted by the original logistic regression model (R 2 = 0.90 and 0.86, respectively) (jitter applied). Filled symbols: heterotherms (blue), suspected heterotherms (purple). D = Galeopterus variegatus. B, D. Variable rate phenograms of the Hubbert linearization of the logistic functions. 15 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Previously we reported several observations indicating a reciprocal relationship in the evolution of body temperature and metabolic rate: in the case of primates, body temperatures (T b) were inversely correlated with mirFam (Fromm and Sorger, 2023). This conserved nature of this relationship can be visualized by ranking primate species according to the product of Tb x mirFam (Supplementary Figure S3). We hypothesized that this product is representative of the heat dissipation limit; in this case, accelerated evolution near the plateau of a Tb x mirFam logistic model might signal a thermoregulatory adaptation. This indeed appears to be have been the case for the divergence of hominid apes (Figure 15C, D) and accords well with the disproportionate increase in the mass of the 4-organ thermogenic core that accompanied the enlargement of the primate brain (Figure 4) and the leftward shift in the mirFam correlation profiles for brain and liver (Figure 6). If our interpretation of these models is correct, then the Bayesian inference of opposite evolutionary trends depicted on the phenograms, descending in rank on the mirFam tree, and ascending in rank on the tree for mirFam x Tb, would assign the ancestral primate a high rank with respect to energy reliability and a low rank with respect to the heat dissipation limit, a propitious combination for the founder of a new order of mammals.

Physiological Data The masses and rMRs of the four thermogenic organs were obtained from the dataset of Wang et al (2012), itself mainly derived from the compilation of Navarette et al (2011) (Supplementary T able S1). MicroRNA Families Complements of microRNA families were obtained from the most complete assembly for each species using covariance model MirMachine (Umu et al, 2023), trained on the manually-curated database MirGeneDB2.1 (Fromm et al 2021). Phylogenetic Trees Using Mesquite © (2021) we derived ultrametric trees for 51 Euarchontoglires or 25 primates from the phylogenies of Alvarez-Carretero et al (2021) and Shao et al (2023), respectively. Variable Rate Phenograms Variable rate trees and ancestral node values were inferred using BayesTraits.V4 (Pagel, 1999; Pagel et al, 2004; Venditti et al, 2011) (software available at www.evolution.rdg.ac.uk). Bayes factors for variable rate trees were computed from the difference between the median of the top 1% likelihoods and the median likelihood of the trees with lowest 1% sum of changes to branch length (Fromm and Sorger, 2023). Phenograms were generated using Phytools (Revell, 2012). 16 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Statistics Linear, partial least squares, and logistic regressions and normal mixture clustering were carried out and visualized using JMP16 ®. All R2 values were adjusted for the number of predictors. Phylogenetic linear regressions were carried out using the R: CAPER package (Orme, 2013). DATA AVAILABILITY The original data with accompanying references and the results of all analyses referred to in the text are provided in the Supplementary Information (below). ACKNOWLEDGMENT BF is supported by the Tromsø forskningsstiftelse (TFS) [20_SG_BF ‘MIRevolution’]. AUTHOR CONTRIBUTIONS Conceptualization: TS; Methodology: TS; Investigation: BF and TS; Writing – Original Draft: TS; Writing – Review & Editing: BF and TS; Visualization: BF and TS; Funding Acquisition: BF . DECLARATION OF INTERESTS The authors declare no competing interests.

Aiello LC and Wheeler P . 1995. The expensive-tissue hypothesis: the brain and the digestive system in human and primate evolution. Curr Anthro 36: 199-221. Álvarez-Carretero S, T amuri A, Battini M, Nascimento F , Carlisle E, Asher R, Yang Z, Donoghue P and Dos Reis M. 2022. A species-level timeline of mammal evolution integrating phylogenomic data. Nature 602: 263-267. Ballesteros FJ, Martinez VJ, Luque B, Lacasa L, Valor E and Moya A. 2018. On the thermodynamic origin of metabolic scaling. Scientific Reports 8: 1448. Buttgereit F and Brand MD. 1995. A hierarchy of ATP-consuming processes in mammalian cells. Biochem J 312: 163-167. de Jong (1993). SIMPLS: An alternative approach to partial least squares regression. Chemometrics and Intelligent Laboratory Systems, 18: 251–263. Deline B, Greenwood JM, Clark JW, Puttick MN, Peterson KJ, Donoghue PCJ. 2018. Evolution of metazoan morphological disparity. PNAS 29: E8909-E8918. Elia M. Organ and tissue contribution to metabolic rate. 1992. In: Kinney JM, Tucker HN (eds), Energy metabolism: tissue determinants and cellular corollaries. pp 61-80. (Raven Press, NY). Pagel M. 1999. Inferring the historical patterns of biological evolution. Nature 401: 877-884. Felsenstein J. 1985. Phylogenies and the comparative method. Amer Naturalist 125: 1-15. 17 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Fonseca-Azevedo K and Herculano-Houzel S. 2012. Metabolic constraint imposes a tradeoff between body size and number of brain neurons in human evolution. PNAS 109: 18571-18576. Fromm B, Høye E, Domanska D, Zhong X, Aparicio-Puerta E, Ovchinnikov V, Umu SU, Chabot, PJ, Kang W, Aslanzadeh M, T arbier M, Mármol-Sánchez E, Urgese G, Johansen M, Hovig E, Hackenberg M, Friedländer MR and Peterson KJ. 2021. MirGeneDB 2.1: toward a complete sampling of all major animal phyla. Nucl Acids Res 50: D204-D210. Fromm B and Sorger T . 2023. Rapid adaptation of cellular metabolic rate to the microRNA complements of mammals and its relevance to the evolution of endothermy. iScience: 108740. Grabowski M, Kopperud BT , T suboi M and Hansen TF . Both diet and sociality affect primate brain-size evolution. Syst Biol 72: 404-418. Hansen TF . 1997. Stabilizing selection and the comparative analysis of adaptation. Evolution 51: 1341-1351. Heldstab SA, Isler K, Graber SM, Schuppli C and van Schaik. The economics of brain size evolution in vertebrates. Curr Biol 32: R697–R708. Hubbert MK. 1982. T echniques of prediction as applied to the production of oil and gas. In Gass, Saul I. (ed.). Oil and Gas Supply Modeling (proceedings of a symposium held at the Department of Commerce, Washington, DC, June 18–20, 1980). NBS Special Publication 631. Washington (DC): National Bureau of Standards. pp. 16–141. Isler K and van Schaik CP . 2012. Allomaternal care, life history and brain size evolution in mammals. J Human Evol 63: 52 - 63. Isler K and van Schaik CP . 2014. How humans evolved large brains: comparative evidence. Evol Anthropol 23: 65-75. Kiyatkin AE. 2007. Brain temperature fluctuations during physiological and pathological conditions. Eur J Appl Physiol 101:3-17. Kleiber M. 1932. Body size and metabolism. Hilgardia 6: 315-353. Koteja P . 2000. Energy assimilation, parental care and the evolution of endothermy. Proc Roy Soc B 267: 479-484. Kuderna LFK, Gao H, Janiak MC . . . Bonet TM. 2023. A global catalog of whole-genome diversity from 233 primate species. Science 380: 9-60913. Kummitha CM, Kalhan SC, Saidel GM and Lai N. 2014. Relating tissue/organ energy expenditure to metabolic fluxes in mouse and human: experimental data integrated with mathematical modeling. Physiol Rep e12159 (1-20). Kuzawa CW, Chugani HT , Grossman LI, Lipovich L, Muzik O, Hof PR, Wildman DE, Sherwood CC, Leonard WR and Lange N. 2014. PNAS 111: 13010-13015. 18 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Lande R. 1979. Quantitative genetic analysis of multivariate evolution, applied to the brain: body size allometry. Evolution 33: 402-416. Magalhães JP , Abidi Z, dos Santos GA . . . Ka Po T o P . 2024. Human Ageing Genomic Resources: updates on key databases in ageing research. Nucl Acids Res 52: D900-D908. Mascia-Lees FE, Relethford JH and Sorger T . 1986. Perspectives on permanent breast enlargement in human females. Amer Anthro 88: 423-428. Nybo L. 2012. Brain temperature and exercise performance. Exp Physiol 97: 333-339. Nybo L, Secher NH and Nielsen B. 2002. Inadequate heat release from the human brain during prolonged exercise with hyperthermia. J Physiol 545: 697-704. Nybo L and González-Alonso J. 2015. Critical core temperature: a hypothesis too simplistic to explain hyperthermia-induced fatigue. Scand J Med Sci Sports 25: 4-5. Navarette A, van Schaik CP and Isler K. 2011. Energetics and the evolution of human brain size. Nature 480: 91-93. Orme D, Freckleton RP , Thomas GH and Petzoldt T . 2013. CAPER: comparative analyses of phylogenetics and evolution in R. Methods Ecol and Evol 3: 141-151. Pagel M. 1999. Inferring the historical patterns of biological evolution. Nature 401: 877-884. Pagel M, Meade A and Barker D. 2004. Bayesian estimation of ancestral character states on phylogenies. Syst Biol 53: 673-684. Pontzer H, Brown MH, Raichlen DA . . . Ross SR. 2016. Metabolic acceleration and the evolution of human brain size and life history. Nature 533: 390-393. Porter WP and Kearney M. 2009. Size, shape, and thermal niche of endotherms. PNAS 106: 19666-19672. Revell LJ. 2012. phytools: an R package for phylogenetic comparative biology (and other things).

Ecol Evol 3: 217–223. Shao Y , Zhou, Li F . . . Wu DD. 2023. Phylogenomic analyses provide insights into primate evolution. Science 380: 913-924. Simmen B, Pasquet P , Masi S, Koppert GJA, Wells JCK and Hladik CM. 2017. Primate energy input and the evolutionary transition to energy-dense diets in humans. Proc Roy Soc B 284: 20170577. Smaers JB, Rothman RS, Hudson DR, Balanoff AM, Beatty B, Dechmann DKN, deVries D, Dunn, JC, Fleagle JG, Gilbert CC, Goswami A, Iwaniuk AN, Jungers WL, Kerney M, Ksepka DT , Manger PR, Mongle CS, Rohlf FJ, Smith NA, Soligo C, Weisbecker V and Safi K. 2021. The evolution of mammalian brain size. Sci Adv 7: eabe2101 (1-11). 19 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Speakman JR and Król E. 2010. Maximal heat dissipation capacity and hyperthermia risk: neglected key factors in the ecology of endotherms. J Anim Ecol 79: 726–746. Stibel JM. 2023. Climate change credictive of body size and proportionality in humans. Evol Biol 50: 461-475. Svensson EI, Arnold SJ, Bürger R et al. 2021. Correlational selection in the age of genomics. Nat Ecol Evol 5, 562–573. T attersall GJ. 2012. Diurnal changes in metabolic rate in pygmy marmosets: Implications for sleep, torpor, and basal metabolism in primates. In: Ruf, T ., Bieber, C., Arnold, W., Millesi, E. (eds) Living in a Seasonal World, chapter 41: 471-480. Springer. T suji B, Hayashi K, Kondo N and Nishiyasu T . 2016. Characteristics of hyperthermia-induced hyperventilation in humans. Temperature 3: 146–160. Umu SU, Paynter VM, Trondsøn H, Buschmann T , Rounge TB, Peterson KJ and Fromm B. 2023. Accurate microRNA annotation of animal genomes using trained covariance models of curated microRNA complements in MirMachine. Cell Genomics 3, 100348. Venditti C, Meade A. and Pagel M. 2011. Multiple routes to mammalian diversity. Nature, 479, 393-396. Walter GM and McCuigan K. 2023. Phenotypic evolution: predicting the future. eLife 12: e91450 (1-3). Wang H, Wang B, Normoyle K, Jackson K, Spitler K, Sharrock MF , Miller CM, Best C, Llano D and Du R. 2014. Brain temperature and its fundamental properties: a review for clinical neuroscientists. Front Neurosci 307: 1-17. Wang ZM, Zhang J, Ying Z and Heymsfield SB. 2012. Organ-tissue level model of resting energy expenditure across mammals: new insights into Kleiber’s Law. ISRN Zool: 673050 (1-9). Yu Y , Hill AP and McCormick DA. 2012. Warm Body T emperature Facilitates Energy Efficient Cortical Action Potentials. PLoS Comp Biol e1002456: 1-16. 20 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint SUPPLEMENTARY INFORMATION T able S1. Summary of mirFam, mass and resting metabolic rates for organs, core and body. All physiological data were obtained from Wang et al (2012), most of which were compiled by Navarette et al (2011). The names of the indicated taxa were substituted with congeneric species on the phylogenies of Álvarez-Carretero et al (2022) and Shao et al (2023). 21 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint T able S2. Summary of mirFam, brain volume (ECV), body mass and temperature for 25 primates. Endocranial volume (ECV) and body mass were obtained from Grabowski et al, 2023. T emperatures were obtained from the AnAge database (Magalhães et al, 2024). 22 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Figure S1. The proportions of brain and liver vary reciprocally in a 4-organ core. Species include one Scandentia, two Lagomorpha, and 27 Glires, including three Sciuridae. Figure S2. Logistic models for the size evolution of four thermogenic organs. Symbols representing clades as in previous figures, with shaded symbols representing heterotherms. 23 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Figure S3. Conservation of mirFam x Tb and its relation to body size. A.Regardless of clade, the largest species deviate the least from the mirFam-log2M relationship. B.Visualization of the reciprocal variation of mirFam and Tb. 24 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint Supplementary T able S3. Summary of phylogenetic regressions of organ cMRs on mirFam. 25 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 23, 2024. ; https://doi.org/10.1101/2024.05.21.595052doi: bioRxiv preprint