Evolution of the Earliest Horses Driven by Climate Change in the Paleocene-Eocene Thermal Maximum

See allHide authors and affiliations

Science  24 Feb 2012:
Vol. 335, Issue 6071, pp. 959-962
DOI: 10.1126/science.1213859


Body size plays a critical role in mammalian ecology and physiology. Previous research has shown that many mammals became smaller during the Paleocene-Eocene Thermal Maximum (PETM), but the timing and magnitude of that change relative to climate change have been unclear. A high-resolution record of continental climate and equid body size change shows a directional size decrease of ~30% over the first ~130,000 years of the PETM, followed by a ~76% increase in the recovery phase of the PETM. These size changes are negatively correlated with temperature inferred from oxygen isotopes in mammal teeth and were probably driven by shifts in temperature and possibly high atmospheric CO2 concentrations. These findings could be important for understanding mammalian evolutionary responses to future global warming.

Interest in how organisms respond to climate change has intensified in recent years with projected warming of ~2° to 4°C over the next century (1). Although models can be developed to predict evolutionary responses to warming of this magnitude, empirical examples must be drawn from fossil or historical records. Here we report a dramatic example of shifts in body size in the earliest known horses (family Equidae) during the Paleocene-Eocene Thermal Maximum (PETM) (~56 million years ago). The PETM is recognized in marine and continental records by an abrupt negative carbon isotope excursion (CIE) that lasted ~175 thousand years (ky), caused by the release of thousands of gigatons of carbon to the ocean-atmosphere system (2, 3). Some marine records suggest that although δ13C values shifted rapidly at the onset of the CIE in 21 ky or less (2), temperature increase was slower, peaking 60 ky or more into the CIE (4) at ~5° to 10°C above pre-CIE levels (5, 6). We use oxygen isotope values in mammal teeth as a proxy for local temperature change in the continental interior of North America, and we show that equid body size during the PETM was negatively correlated with temperature.

In extant mammals and birds (endotherms), closely related species or populations within a species are generally smaller-bodied at lower latitudes, where ambient temperature is greater (7). This relationship, known as Bergmann’s rule, is followed by ~65 to 75% of studied extant mammals (8, 9). The cause of Bergmann’s rule is usually attributed to thermoregulation and the optimization of body size (10) and/or the availability of food resources related to primary productivity (11). Bergmann’s rule predicts that average mammalian body size should decrease with warming climate, and smaller size in endotherms has even been suggested as a third “universal” response to warming, along with changes in phenology and species distribution (10). Declining body size has been attributed to warming over decadal and millennial scales in some living endotherms (12, 13), but many counterexamples also exist (10). Furthermore, it is difficult to distinguish natural selection (genetic change) from ecophenotypic plasticity (morphological response not genetically fixed) over such short time scales. The size change documented here was, however, sustained over thousands of generations, strongly suggesting that natural selection was the cause.

Previous studies lacked the stratigraphic resolution to recognize patterns in body size change within the PETM but demonstrated gross changes in size in several mammal lineages, based on first molar tooth area (14, 15). Size changes occurred in herbivorous ungulates (Perissodactyla, Artiodactyla, Condylarthra, and Tillodontia), Primates, and faunivores and omnivores (Creodonta, Carnivoramorpha, and Palaeanodonta), affecting both immigrant and endemic taxa (Fig. 1). These changes conform well to Bergmann’s rule in terms of the expected direction of size change. Quantifying published results, size reduction occurred in 10 Paleocene genera that ranged into the PETM, representing 38% of the range-through genera. This was followed by post-PETM size increases in eight of these genera, indicating that body size response was strongly taxon-specific (Fig. 1 and table S7). Post-PETM size increases also occurred in an additional eight genera, seven of which first appeared in the PETM (Fig. 1). Together these 16 genera represent a size increase in 40% of PETM genera that ranged into post-PETM biozone Wa-1 (Fig. 2).

Fig. 1

Summary of percent mean body size change in genera that exhibit change from the latest Paleocene to the PETM (left), and from the PETM to the post-PETM (right). No genus exhibits a size increase in the PETM or a decrease after the PETM. Compiled from published sources, except for Sifrhippus from this study. Asterisks indicate genera that first appear in the PETM. See table S7 for a summary of all PETM taxa and sources.

Sifrhippus [formerly Hyracotherium (16)] first appeared in North America and Europe during the PETM. Because of the lack of a plausible ancestor on these continents, it is widely thought to be an immigrant that crossed high-latitude dispersal routes opened by PETM warming (17). We use Sifrhippus to document mammalian body size change within the PETM. Sifrhippus is the most abundantly represented genus in new collections from the Cabin Fork area (~10 km2) of the southern Bighorn Basin, Wyoming, and the only one for which detailed stratigraphic and quantitative morphological data are available. We also isotopically sampled Sifrhippus, Coryphodon (Pantodonta; large archaic herbivorous ungulates), and Ectocion and Copecion (phenacodontid condylarths; herbivorous ungulates of uncertain affinities). The PETM at Cabin Fork is represented by a ~35-m-thick sequence of fluvial mudstones, floodplain soils (paleosols), and fluvial sandstones. We constructed an age model that assumes varying rates of sediment accumulation: Avulsion deposits (mudstones and thin sandstones) represent fast rates, and paleosols represent much slower rates [see the supporting online material (SOM)]. Local sections were correlated to a composite section (Fig. 2) using marker beds traced with a differential Global Positioning System unit (SOM).

Fig. 2

Comparison of PETM Cabin Fork records. (A) From left to right, epochs, mammalian biozones, formations, meter levels, marker beds, and δ13CE values for three common mammal genera. vPDB, Vienna Pee Dee belemnite standard. (B and C) δ13CE and δ18OE values for Coryphodon. vSMOW, Vienna standard mean ocean water standard. (D) Log-transformed measurements of first lower molar area (length × width) for Sifrhippus. Data points represent single individuals except where error bars (indicating 95% confidence of the mean for multiple samples from one individual) are shown. Solid colored lines show five-point moving averages; the gray area is the 95% envelope of uncertainty for each line. PALEO, Paleocene; Cf, Clarkforkian; Wa, Wasatchian. LIRB, SLIRB, and Br denote key marker beds.

The CIE at Cabin Fork is recorded in the carbonate component of mammalian tooth enamel (δ13CE) (Fig. 2, A and B) and in bulk organics and leaf wax n-alkanes (6, 18). δ13CE in mammalian herbivores reflects the δ13C value of the vegetation they consume, with predictable enrichment (19). Plants in turn track the δ13C value of atmospheric CO2, with influences from environmental factors such as humidity and vegetation density (20, 21). At Cabin Fork, phenacodontids (Ectocion and Copecion) record a negative shift of ~4.6 per mil (‰) in δ13CE at the onset of the CIE (Fig. 2A). This is consistent with estimates of atmospheric change of ~4.6‰ during the PETM from a leaf discrimination model (20) and ~4.0‰ from modeling of marine carbonate dissolution (2), indicating that phenacodontid δ13CE is primarily tracking atmospheric δ13C values, rather than environmental change.

Sifrhippus sandrae first appears at Cabin Fork near the base of the lowest intermittent red bed (LIRB) at 14.5 m (Fig. 2). The onset of the CIE in most mammal teeth also begins at the base of the LIRB (Fig. 2, A and B) but is recorded slightly lower (13.75 m) in dispersed bulk organics. The oldest specimens of S. sandrae had an average body size of ~5.6 kg, based on first lower molar area (SOM). Body size in S. sandrae progressively decreased from its first appearance at 14.5 m to the 41-m level, with a total reduction of ~30% over ~130 ky (P < 0.001) (Fig. 2D and SOM). Individuals at 41 m had an average body weight of ~3.9 kg and are among the smallest known horses. The dwarfing of S. sandrae was followed by a ~76% increase in body size during the recovery phase of the CIE, to an average size of ~7.0 kg (Fig. 2D).

The mode of evolution (random, static, or directional) for Sifrhippus body size change was determined using a moving window log rate interval (mwLRI) analysis, which is a modification of the standard LRI analysis (22) (SOM). Both methods assume that rates of change in a time series variable are inversely proportional to the interval of time over which rates are measured, because of the occurrence of small reversals in the variable. The relationship between rates of change and the lengths of intervals over which they are observed is used to determine evolutionary mode (22). Our mwLRI results indicate with 95% confidence that Sifrhippus body size directionally decreased from its first appearance to the 41-m level, after which stratigraphic resolution and sample sizes are insufficient to distinguish between directional and random evolutionary change. Thus, Sifrhippus experienced sustained selection for diminutive body size for ~130 ky.

To test whether body size change in Sifrhippus is significantly correlated with temperature, as predicted by Bergmann’s rule, we used δ18O values in Coryphodon enamel (δ18OE) as a proxy for change in mean annual temperature (MAT). Coryphodon was a large water-dependent or semi-aquatic mammal (21, 23). Studies of ecologically similar living mammals have shown that their δ18OE faithfully records the δ18O of surface water (24, 25), which in turn is strongly correlated with air temperature at mid- to high latitudes (26). Sifrhippus first lower molar area is negatively correlated with Coryphodon δ18OE values (P ≤ 0.05, SOM), suggesting that Sifrhippus body size decreased as ambient air temperature increased.

Greater aridity in the PETM could also have contributed to diminished body size by lowering primary productivity. Both floras and paleosols in the Bighorn Basin suggest increased aridity during at least parts of the PETM (6, 20, 27). To test this, we used two aridity proxies. The first is based on the difference between mean δ18OE values in aridity-sensitive and aridity-insensitive mammals (24). Coryphodon should be aridity-insensitive because of its probable water dependence (21, 23), whereas Sifrhippus is the taxon most likely to be aridity-sensitive, because it has the highest average mammalian δ18OE value, suggesting that it consumed leaves in open areas where leaf water was evaporatively 18O-enriched. Increased aridity should result in higher Sifrhippus δ18OE values and greater separation between it and Coryphodon δ18OE (Fig. 3A). Our second aridity proxy estimates mean annual precipitation (MAP) based on paleosol major oxides (Fig. 3C and SOM). Both proxies suggest drier conditions at the beginning of the CIE, followed by wetter conditions starting at ~20 m (~68 ky into the PETM), with a return to drier conditions by ~38 m (~108 ky into the PETM). Overall, there is poor agreement between Sifrhippus body size change and the aridity proxies. Both proxies indicate a shift to wetter conditions while body size in Sifrhippus is decreasing, which is counter to expectations if the primary cause of dwarfing was lowered productivity caused by increased aridity.

Fig. 3

Aridity and precipitation proxies. (A) Comparison of mean δ18OE values for aridity-insensitive Coryphodon (brown diamonds) and aridity-sensitive Sifrhippus (gold squares). Data are in 5-m bins. Brown circles are singletons of Coryphodon. (B) Aridity proxy curve based on (A), showing mean differences between Sifrhippus and Coryphodon δ18OE. Greater difference implies greater aridity. Error bars show 95% confidence of the mean, offset in (A) by 1 m to avoid overlap. (C) MAP proxy based on paleosol major oxides from a nearby correlative section (HW16 section, SOM).

Our results are consistent with mammalian dwarfing driven by warming, but temperature alone may be an insufficient explanation. Although body mass in living mammals is highly correlated with MAT in the Nearctic [coefficient of determination (R2) = –0.75], this relationship weakens above ~11°C and reverses at higher temperatures in the Neotropics (9). MAT was well above ~11°C in the latest Paleocene and PETM of the Bighorn Basin (6). Furthermore, ~25 to 35% of living mammals deviate from Bergmann’s rule (8, 9), and it is likely that at least some mammal lineages would have gotten larger during the PETM if MAT were the only controlling factor.

Another possible cause for body size decrease in the PETM is elevated atmospheric partial pressure of CO2 (Pco2) (28), which might covary with temperature. In many extant plants, elevated CO2 increases biomass but reduces nitrogen and protein content in leaves and can elevate phenol levels, yielding cellulose-rich vegetation that is less nutritious and harder for herbivores to digest (29). Ultimately, this should result in slower growth and reproductive rates in herbivorous mammals (30), conceivably resulting in selection for smaller body size. Although this mechanism could have reduced body size in herbivores, size also decreased among PETM carnivores (Fig. 1), which must be explained by an indirect response, such as selection for smaller predators because of smaller prey (31). Recent modeling of rates of carbon release during the PETM shows the largest increase in Pco2 at the onset of the CIE, followed by lower concentrations later in the event (2). This is inconsistent with a Pco2-driven decrease in body size, because Sifrhippus was smallest near the end of the main phase of the PETM. Although elevated Pco2 could have been a contributing factor, our results favor temperature as the primary driver of dwarfing in Sifrhippus.

PETM warming was similar in magnitude to that predicted by some global models over the next century (1) but occurred at a much slower rate and began from a warmer late Paleocene baseline. Nevertheless, some generalizations applicable to future warming may still be relevant. Diminished body size in some mammal species, along with changes in ecology and physiology, might be expected in response to warming. The pattern of dwarfing seen in the PETM mirrors recent reductions in body size in endotherms that have been attributed to anthropogenic warming (10, 12). Although the rate of present warming is much faster than during the PETM, and mammals may not respond in exactly the same manner, the dramatic response to warming observed in PETM equids provides a measure of possible responses to future warming in modern mammals.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S4

Tables S1 to S9


References and Notes

  1. Numerous authors have shown the use of “Hyracotherium” to be invalid for North American equids. Thus, the species “Hyracotheriumsandrae (PETM) and “H.” grangeri (post-PETM) were assigned to the new genera Sifrhippus Froelich 2002 and Arenahippus Froelich 2002, respectively. We found, however, that characters used to separate Sifrhippus from Arenahippus are highly variable and not useful for generic identification. Thus, we refer both species to Sifrhippus pending formal revision.
  2. Acknowledgments: We thank T. Bown, P. Gingerich, B. MacFadden, K. Rose, E. Sargis, and S. Strait for helpful discussions and advice; J. Curtis, B. Tucker, and A. Baczynski for help with isotope lab work; J. Bourque and A. Hastings for specimen preparation; and P. Koch and two anonymous reviewers for helpful comments. Supported by NSF grants EAR-0640076 (J.I.B., J.K., R.S.), EAR-0719941 (J.I.B.), EAR-0717892 (S.L.W.), EAR-0718740 (M.J.K.), and EAR-0720268 (F.A.M.). Data used in this paper are available in the SOM.
View Abstract

Navigate This Article