Decoupled ecomorphological evolution and diversification in Neogene-Quaternary horses

See allHide authors and affiliations

Science  10 Feb 2017:
Vol. 355, Issue 6325, pp. 627-630
DOI: 10.1126/science.aag1772

What drives divergence?

Horse evolution has long been held as a classic example of adaptive radiation. It has been thought that an increase in the height of cheek teeth opened up new grass resources, leading to divergence. Cantalapiedra et al., however, found that although the Equinae have experienced high levels of divergence, these splits do not appear to have been related initially to specific phenotypic changes. Instead, it seems that external environmental drivers and patterns of migration and isolation initiated population divergence, with phenotypic changes emerging once lineages had begun to divide.

Science, this issue p. 627


Evolutionary theory has long proposed a connection between trait evolution and diversification rates. In this work, we used phylogenetic methods to evaluate the relationship of lineage-specific speciation rates and the mode of evolution of body size and tooth morphology in the Neogene and Quaternary radiation of horses (7 living and 131 extinct species). We show that diversification pulses are a recurrent feature of equid evolution but that these pulses are not correlated with rapid bursts in phenotypic evolution. Instead, rapid cladogenesis seems repeatedly associated with extrinsic factors that relaxed diversity bounds, such as increasing productivity and geographic dispersals into the Old World. This evidence suggests that diversity dynamics in Equinae were controlled mainly by ecological limits under diversity dependence rather than rapid ecomorphological differentiation.

Rapid phenotypic evolution has long been taken to be an important factor in evolutionary radiations (1). In this model of adaptive radiation, changes of ecologically relevant traits should be faster in the early phases of the clade’s expansion (2), as lineages fill new adaptive zones (1, 3). Because change is predicted to slow down as these zones are filled (2), ecomorphological disparity among subclades and zones should be more partitioned than expected under a nonadaptive scenario (4). However, such an adaptive model may not be common in nature (2, 3, 5, 6), and accelerated diversification could result from extrinsic factors—such as geographical dispersals, increased productivity, or habitat heterogeneity (79)—that release diversity limits and promote speciation (10) without involving rapid ecomorphological divergence (3).

The radiation of equids in the Neogene has been cited as a textbook example of adaptive radiation for more than a century (11), as it is crucial in the development of evolutionary theory linking trait evolution and adaptive success (1, 12). The rich equid fossil record provides a suitable data set for testing these ideas within a phylogenetic framework. Much work has focused on the evolution of body size and dental morphology (12), as these two traits condense multiple dimensions (such as population density, range size, diet, and environmental pressures) of a species’ adaptive zone. Early studies based on dental proportions suggest that phenotypic change accelerated during an early Miocene radiation (1, 13, 14), although recent analyses show that body size disparity did not increase during diversification pulses (15). Yet, previous work has been nonphylogenetic and has not directly investigated the connection between diversification dynamics and phenotypic evolution.

In our study, we assessed tree-wide variation in speciation rates in the Neogene and Quaternary radiation of equids [from ~18 million years ago (Ma) to the present]. We then used phylogenetic maximum-likelihood trait modeling to evaluate the mode of evolution of body size and tooth crown height (hypsodonty) across clades, time, and space, as well as to directly question whether speciation and rates of phenotypic evolution are coupled in equine lineages. Finally, we used phylogenetically informed regressions [specifically, phylogenetic generalized least squares (PGLS) regression] to directly test for correlations among speciation rates, body size, and hypsodonty. Our data set incorporates a substantial amount of information from fossils (~95% of the 138 species considered are extinct), which considerably improves the ability of tree-based approaches to recover the past (3, 16).

We found evidence for repeated speciation bursts across Equinae, but none of these were associated with rapid ecomorphological evolution. Each subclade we considered shows an early expansion followed by a slowdown in diversification rates that results in a diversity plateau (14, 17). This pattern, which is mirrored in the early Miocene rise of American tribes (18 to 15 Ma) and when lineages entered the Old World (11 and 4.5 Ma) (Fig. 1, A and B), is consistent with a model of logistic growth with a finite upper limit to species richness (table S1) (8). If such a burst and slowdown dynamic was the result of early niche differentiation processes (an adaptive radiation) (3), we should expect early rapid ecomorphological divergence resulting from rapid trait evolution (2, 6). None of these early-phase expansions, however, were correlated with an early burst in body size or hypsodonty evolution (tables S5 to S14). Furthermore, rates of body size evolution were not significantly different in lineages exhibiting high and low speciation, whereas rates of hypsodonty evolution were significantly lower in lineages with fast speciation rates (Fig. 1, D and E, and table S15), pointing to a marked decoupling of diversification rates and the evolution of functional traits. In line with these findings and in contrast to previous notions (14), we found that rates of phenotypic evolution in American forms were marginally slower during their basal radiation (before ~15 Ma) than afterward (table S16).

Fig. 1 Speciation rates and ecomorphological evolution in Neogene-Quaternary horses.

(A) Chronogram showing polytomies and complete stratigraphic ranges (thin gray bars). Branches are colored according to the mean speciation rate. Vertical shaded bars highlight (from left to right) early Miocene, late Miocene, and Pleistocene. Gray circles indicate dispersals into the Old World. The inset shows the scale and density of speciation rate values across branches. myr, million years. (B) Log-scaled median diversity through time across 100 trees. Shaded regions represent the 95% quantile (see also fig. S1). Plio, Pliocene; Pl, Pleistocene. (C) Evolution of Equinae ecomorphospace plotted on the phylogenetic tree of Equinae during different time intervals. For the sake of clarity, Pseudhipparion simpsoni was left out of the plots. Note the log-transformed values on the y axes. (D) Rates of trait evolution reconstructed for each branch. (E) Rates of body size and hypsodonty evolution in lineages of the tree with high and low speciation rates.

We found that the subclades of Equinae substantially overlap each other in morphospace occupation across the entire analysis interval (see Figs. 1C and 2 and fig. S7). This differs from the among-subclades partitioned disparity expected in radiations that are adaptive, as observed in other clades (4, 18, 19). The observed disparity through time pattern supports the conclusion of our trait modeling approach. The explosive radiation of equine horses in the early Miocene and subsequent radiations as the group dispersed into the Old World were not spurred by early, rapid ecomorphological divergence (15).

Fig. 2 Subclade relative disparity through time (DTT) in the clade containing Equinae and Parahippus.

DTT measures the proportion of disparity of the whole clade held by each of the subclades whose ancestral lineages are present at a given time (4). DTT (solid line) is plotted for body size (A) and hypsodonty (B). The dashed line indicates the median DTT based on simulations of trait evolution under Brownian motion. The shaded area denotes the 95% range of the simulated data. The morphological disparity index (MDI) is the overall difference between the observed among-clades disparity and the null distribution. In adaptive radiations, disparity should be highly partitioned among clades, with each lineage holding little of the total disparity (little overlap in morphospace), and significant negative MDI values are expected. The significant P value in (B) suggests that the observed trend departs from the null expectation. DTT plots for different subclades are shown in fig. S7. mya, million years ago.

The early radiation of American equid tribes (18 to 15 Ma) has traditionally been explained as the direct outcome of morphological adaptation linked to the onset of grass-dominated habitats (17, 20). However, recent paleobotanical evidence suggests that grasslands were well developed in North America ~25 Ma (21), and recent microwear analyses show that the more primitive Parahippus forms already had a grass-dominated diet in the earliest Miocene, before 20 Ma (22). Our findings fit this environmental context and imply that the early Miocene radiation of equine lineages took place in the absence of rapid ecomorphological shifts (Fig. 1, C and D). One possible explanation is that grazing behavior evolved ahead of morphological adaptation (23) in early forms. Morphological change accelerated later on (Fig. 1, C to E, and table S16), probably to accommodate the new feeding style and as a response to enhanced competition within Equinae and with other clades. However, the appearance of the grazing behavior still predates the American radiation by several million years (22). In the absence of a clear intrinsic driver, we suggest that the early Miocene radiation of horses might be attributed to external factors, including higher productivity or increased biogeographic provincialism associated with climatic change (8, 9, 24, 25). These factors could have released diversity limits and promoted speciation by reducing competition (the crowding effect) among forms with largely overlapping ecomorphotypes and behavior (Fig. 1C).

The long middle Miocene diversity plateau exhibited by New World clades was only surpassed with the first dispersal into the Old World ~11 Ma (Fig. 1B and fig. S5). Large dispersal events such as these have been predicted to increase diversity limits by providing new ecological arenas and spurring speciation (8, 26). Dispersals into the Old World always promoted speciation (Fig. 1, A and B) under disparate modes of evolution (tables S5 to S14, fig. S7, and data S1). For example, New and Old World hipparions show early-phase speciation (Fig. 1B) but under two completely distinct modes of phenotypic evolution that probably resulted from different environmental pressures (27) (fig. S7). Hipparions that dispersed across the Old World likely evolved from a large ancestor (fig. S3), and their size diffused five times faster around a larger optimal size (~250 kg) as compared with their American relatives (~100 kg) (Fig. 1C, fig. S7, and table S5). By contrast, hypsodonty of Old World hipparions changed six times slower than in New World hipparions, and the macroevolutionary optimum of the Old World animals barely departed from moderate scores over 10 million years (Fig. 1C, fig. S7, and table S6).

Geographic dispersals may affect morphological evolution dissimilarly in different clades. Whereas Old World hipparions show markedly different trait evolution than their American relatives (see above), American and Old World Equus are better modeled as a single clade (fig. S6). Trait evolution shifts were more likely to happen at the base of the Equus clade than when Equus entered the Old World (see data S1), which suggests that the Old World radiation of Equus did not require a substantial shift in phenotypic evolution. Size in Equini evolved under a macroevolutionary diffusion (table S5), confirming that larger sizes in this lineage did not arise by active selection (Cope’s rule) (15). In Equus, body size evolved slower and with a marked selection toward smaller sizes, and hypsodonty evolved faster under higher selective pressure than in the rest of Equini (fig. S6 and tables S5 and S6). Both traits followed more selected or constrained evolution in Equus than in other Equini lineages. Also, Equus show a tendency toward increased hypsodonty and decreased body size in more recent lineages that also show slower speciation (see PGLS results in tables S2 to S4; see also Fig. 1C). Taken together, these results are consistent with short-term responses to harsher environmental conditions and frequent shifts in resource availability in Pleistocene times, as previously noted (28, 29).

Our multilayered approach reveals a complex connection between ecological opportunities, diversification dynamics, and trait evolution (5, 6). Although early clade expansions are prevalent in Neogene horses, we found no evidence supporting a key role for ecomorphological divergence in these speciation pulses. Rather, ecologically relevant traits show completely disparate evolutionary modes during such diversification events (Fig. 1, C to E). Clade expansion probably occurred as a result of diversity limits being released due to extrinsic factors. Under this scenario and with the present data, it is difficult to tease apart the nature of the diversity limits rendering the subsequent slowing of the speciation rate (3). Given the substantial overlap in ecomorphological space of Equinae lineages (Figs. 1C and 2 and fig. S7), the signal of diversity dependence recovered in our analyses (table S1) may result, to a great extent, from intensifying within-clade competition and an attenuation of the factors that release diversity limits in the first place (3, 8, 9). Horses now join a growing body of fossil evidence characterized by diversity dependence as a recurrent pattern in macroevolution. Phenotypic evolution in Equinae is distinguished by a notable capacity for both short- and long-term shifts (high evolutionary rates in some clades, shifting macroevolutionary optima in others). Such shifts were certainly shaped by environmental pressures, but diversity dynamics was the domain of ecological limits.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S7

Tables S1 to S16

References (30107)

Data S1

References and Notes

  1. Acknowledgments: Data are available in the supplementary materials. This work was supported by the Humboldt Foundation (Germany) and the Spanish and Argentinian governments (projects CGL2010-19116/BOS, CGL 2015-68333P, AECID A/030111/10, ANPCYT-PICT-11-0561, and FONCYT PICT 2015-1512). We thank F. Bibi, J. Müller, M. Fortelius, J. Saarinen, and three anonymous reviewers for comments. We acknowledge technical assistance from J. M. Beaulieu, G. Slater, D. W. Bapst, N. J. Matzke, J. Uyeda, L. Harmon, and W. Gearty.
View Abstract

Stay Connected to Science

Navigate This Article