The evolution of critical thermal limits of life on Earth

The study investigates how species’ thermal tolerance limits (upper heat and lower cold thresholds) have evolved across the tree of life and what mechanisms explain observed global patterns in thermal limits. The context is that geographical range boundaries often align with physiological thermal limits, making them critical for forecasting biodiversity responses under climate change. Notably, heat tolerance is relatively invariant across latitude, elevation, and phylogeny, whereas cold tolerance varies widely. The authors evaluate three non-mutually exclusive hypotheses: (1) deep-time climate legacies (niche conservatism leading species to retain ancestral climatic affinities based on palaeoclimates at clade origin), predicting both limits correlate with warm vs cold ancestry; (2) physiological boundaries/optima constraining evolution of thermal limits, especially upper limits, predicting an Ornstein–Uhlenbeck (OU) mode with stabilizing selection towards attractors and slower evolution of heat tolerance; and (3) adaptation to current climatic extremes, predicting a tight match between environmental minima/maxima and thermal limits with similar evolutionary rates for upper and lower limits. The purpose is to disentangle the relative support for these mechanisms using a large cross-taxon dataset and phylogenetic comparative methods to inform how much evolutionary rescue is possible under rapid warming.

Prior work shows large interspecific and clade-level variation in cold tolerance, but remarkable invariance in heat tolerance across latitude, elevation, and phylogeny. Thermal performance curves are left-skewed, implying greater fitness penalties at high temperatures, yet the global cause of constrained upper limits remains untested. Niche conservatism literature suggests ancestral climatic affinities can constrain occupation of new climates over evolutionary time. Physiological theory posits high temperatures destabilize membranes and proteins, potentially imposing hard boundaries on heat tolerance; oxygen limitation and conserved macromolecular thermal sensitivity have been proposed mechanisms. Empirical studies (e.g., drosophilids) report strong phylogenetic constraints and limited capacity to evolve heat resistance above ~39 °C. Environmental analyses show maximum temperatures vary less across space than minimum temperatures, and behavioral thermoregulation often buffers heat stress more than cold stress, implying weaker selection on heat tolerance. Together, these findings motivate tests of deep-time legacy, physiological boundary, and current-climate adaptation hypotheses using phylogenetically informed, multi-taxon analyses.

Data: The authors compiled experimentally derived thermal tolerance limits for >2000 species across terrestrial and aquatic realms from the GlobTherm database, including lethal and critical thermal limits for plants (photosynthetic plants and macroalgae) and ectotherms, and thermal neutral zone (TNZ) edges for endotherms. Sampling spans ~70°S to 70°N, reflecting terrestrial dominance of macroscopic diversity; aquatic and terrestrial data were pooled for ectotherms and endotherms (plants analyzed separately due to broader phylogenetic disparity). Thermal limits were analyzed separately for upper (heat) and lower (cold) traits. Proxies and predictors: To test deep-time climate legacy, each species was assigned a ‘thermal ancestry’ based on the predominant palaeoclimatic regime (glaciated vs warm) at the time its order originated (order-level origination times from TimeTree and related sources). Physiological boundary hypotheses were assessed via evolutionary tempo and mode across phylogenies, using order age and model fits. Adaptation to current climatic extremes was evaluated with contemporary minimum and maximum environmental temperatures experienced across species’ ranges (WorldClim for terrestrial, Bio-ORACLE for marine). Phylogenetic comparative analyses: For each group (ectotherms, endotherms, plants), the authors fit alternative evolutionary models to the distribution of thermal limits across the phylogeny: Ornstein–Uhlenbeck (OU) indicating attraction towards optima/boundaries, Brownian Motion (BM) indicating random walk, and White Noise (WN). Tempo (σ^2; °C per Myr) and mode (-log α, where lower values indicate stronger attraction) were estimated, with log-likelihoods compared among models. Analyses were performed at multiple taxonomic levels; results summarized at broad levels (Table 1), with supplementary analyses providing family-level and metric-specific detail. Machine learning: Random forest models were used to quantify variable importance of (i) current temperatures, (ii) clade age, and (iii) palaeoclimate of order origin (categorical) in explaining variation in upper and lower thermal limits for ectotherms, endotherms, and plants. Model performance (R^2) and predictor importance were reported with confidence intervals. Directional relationships were explored in supplementary figures. Biogeographic summaries: The distribution of thermal data points and palaeoclimate ancestry across latitudes and realms was mapped. Relationships between order age and thermal limits were visualized with density plots and traitgrams, and sample sizes per category were indicated.

• Deep-time climate legacies: For ectotherms and terrestrial plants, species from orders that originated during glaciated palaeoclimates tend to tolerate colder temperatures (i.e., have lower cold tolerance limits) than species from warm-origin orders. Heat tolerance showed no consistent relationship with palaeoclimate ancestry. This legacy signal was not broadly evident for endotherms or aquatic plants.
• Evolutionary tempo: Across taxa, lower (cold) thermal limits evolved faster than upper (heat) limits, indicating asymmetric rates consistent with stronger constraints on upper limits. Example σ^2 (°C Myr−1): ectotherms upper 0.784 vs lower 1.224; endotherms upper 0.593 vs lower 2.067; plants upper 0.675 vs lower 1.183; plants & algae upper 1.366 vs lower 1.300 (Table 1). Differences were especially large for endotherms.
• Evolutionary mode: OU models fit substantially better than BM or WN across groups and limits, indicating evolution towards ‘attractor’ values. Attractor strength (lower -log α indicates stronger attraction) tended to be stronger for upper limits in endotherms and terrestrial plants (e.g., endotherms: upper -log α 1.262 vs lower 1.029; plants: upper 0.835 vs lower 0.494). Log-likelihoods favored OU over BM and WN across datasets (e.g., ectotherms upper: LnLik OU -1542.68 vs BM -1577.04; endotherms upper: -817.68 vs -885.84).
• Relative importance of drivers: Random forests showed current environmental temperatures were the strongest predictors of both upper and lower thermal limits across taxa. Model R^2: ectotherms ~0.74 (upper) and 0.72 (lower); endotherms ~0.18 and 0.13; plants ~0.83 and 0.52. Clade age also contributed substantially, sometimes rivaling current climate in ectotherms and endotherms. Palaeoclimate category had low importance overall and was only significant for ectotherms (importance ~5.9% for upper and 5.3% for lower), aligning directionally with expectations (cold-origin species tolerate colder temperatures; warm-origin species slightly higher heat tolerance).
• Biogeographic context: Most sampled species (~80%) descend from warm-origin orders; data span 70°S–70°N. Variation in thermal limits increased with clade age, especially in ectotherms and endotherms, though recent-time sampling was sparser.
• Implication for climate change: Given the historically slow evolution of upper thermal limits and evidence for attraction towards upper physiological boundaries, adaptive shifts in heat tolerance are unlikely to keep pace with the unprecedented rate of contemporary warming for most species.

The findings demonstrate that global patterns in thermal tolerance arise from a combination of adaptation to present-day climatic extremes, evolutionary constraints consistent with physiological boundaries, and, to a lesser extent, deep-time climate legacies. The superior fit of OU models and slower tempo of heat-limit evolution indicate selection towards attractor values or boundaries for upper tolerance, likely reflecting biophysical constraints (e.g., oxygen limitation, macromolecular stability) that limit further increases in heat tolerance. Conversely, cold limits show greater evolutionary lability, particularly in endotherms, consistent with more evolvable mechanisms (e.g., changes in insulation, body size) and greater environmental variability in minimum temperatures. The deep-time legacy signal in ectotherm cold tolerance suggests that ancestral climate at clade origin still influences present-day physiological limits, while the lack of such a signal for heat limits implies that upper boundaries are set by conserved physiological constraints rather than ancestral climates. The strong role of current temperatures supports niche tracking and contemporary adaptation shaping present-day limits, especially in ectotherms and plants. However, low explanatory power in endotherms reflects different determinants of TNZ edges compared with critical/lethal limits in other taxa. Together, these results address the research question by quantifying the relative contributions of hypothesized mechanisms and highlighting limited scope for evolutionary rescue via increases in upper thermal limits under rapid warming. This has broad relevance for forecasting biodiversity vulnerability, particularly for tropical ectotherms with narrow thermal safety margins.

This study synthesizes a global, multi-taxon dataset to show that species’ thermal tolerance limits are primarily shaped by current climatic extremes, constrained by evolutionary attractors consistent with physiological boundaries (particularly for heat tolerance), and modestly influenced by deep-time climate legacies in ectotherm cold tolerance. Lower thermal limits evolve faster than upper limits across taxa, and OU dynamics dominate over Brownian motion, indicating stabilizing selection or attraction towards optima/boundaries. Given the historically slow evolution and constrained nature of upper thermal limits, most species are unlikely to adapt rapidly enough to escalating heat under climate change. Future research should: (i) expand taxonomic and geographic coverage (especially aquatic taxa and underrepresented clades); (ii) refine mechanistic links between cellular/biophysical constraints and macroevolutionary attractors; (iii) integrate intraspecific variation and plasticity with macroevolutionary patterns; and (iv) couple physiological evolution with demographic and range-shift models to improve risk assessments under future warming scenarios.

• Sampling limitations: The dataset, while large, is an incomplete sample of the tree of life and is uneven across realms and taxa (fewer aquatic taxa), which may bias estimates and reduce power for some groups.
• Metric heterogeneity: Different thermal tolerance metrics (lethal, critical limits; TNZ edges) were combined, and these traits differ in their physiological meaning and determinants, especially for endotherms versus ectotherms/plants.
• Phylogenetic and model assumptions: OU support can reflect both stabilizing selection and niche conservatism; inferring specific evolutionary processes from model fit requires caution. Phylogenetic trees and divergence time estimates introduce uncertainty.
• Proxy-based inference: Deep-time legacy was assessed using order-level palaeoclimate categories and origination ages as proxies, which may obscure finer-scale evolutionary dynamics.
• Environmental predictors: Contemporary temperature layers are broad-scale proxies and may not capture microclimatic conditions or behavioral buffering effects experienced by organisms.