Adaptive radiation within emerging environments is believed to underpin the diversification of many animal groups (
1). Radiating lineages can diverge rapidly (
2), but this may be difficult to detect if phylogenetic analyses are poorly calibrated or models are misspecified (
3,
4). Often the result is a dichotomy between ancient (short-fuse) and younger, more rapid (longer-fuse) diversification models, which can describe starkly different scenarios of ecologically driven radiation (
4).
Over the past 15 million years, Australia has undergone a shift from a mesic to a largely arid continent (
5). This environmental transition has been closely linked with the evolution of a diverse, aridity-adapted biota (
6), which includes the kangaroos and wallabies (Macropodidae;
Fig. 1A), the most abundant and diverse marsupial herbivores to evolve. But the timing and drivers of macropodid evolution have been difficult to resolve because of a patchy fossil record and imprecise divergence dating (
7). Nonetheless, most phylogenies infer that the grass-feeding kangaroos (Macropodini), constituting more than one-third of macropodid diversity, underwent a pronounced late Miocene diversification (
7–
9), 3 million to 8 million years before Australian grasslands emerged (
10).
In North America and Eurasia, Neogene grassland expansion ushered in the diversification of herbivores with high-crowned molar teeth, most notably ungulates (
11,
12). Ungulate success has been linked to the improved resistance of their teeth to the elevated dental wear characteristic of grass-based diets (
13). Similarly, extant grass-feeding kangaroos have higher-crowned molars (
14) (
Fig. 1B) and are much more diverse than their browsing counterparts. This suggests that grass exploitation has been a key factor in macropodin success (
7).
To test the roles of late Miocene aridity and Pliocene grassland expansion in the adaptive radiation of kangaroos, we measured molar crown height and macrowear (figs. S1 and S2) for >1000 macropodoid (kangaroos and relatives) specimens from the modern fauna and 93 fossil assemblages (table S1) of late Oligocene to late Pleistocene age (
15). Late Oligocene balbarines (plesiomorphic lophodont macropodoids) and basal macropodids have low-crowned dentitions and low macrowear levels (
Fig. 2, B and C) indicative of low-abrasion diets. Crown height disparity increased in the early Miocene (
Fig. 2B), probably as a response to more exploitation of fibrous plant resources, especially given that this coincides with the evolution of shearing bilophodont macropodid dentitions (
16). Increased middle Miocene balbarine macrowear seems to capture an attempt by these plesiomorphic kangaroos to switch to more abrasive plant resources near the middle Miocene climatic optimum (
Fig. 2, A and C), shortly before their extinction.
Unexpectedly, there is no late Miocene increase in crown height or macrowear (
Fig. 2, B and C) analogous to those interpreted as responses to aridity among Northern Hemisphere herbivores (
11,
17). Instead, late Miocene macropodids, represented by early sthenurines (short-faced kangaroos) and macropodines (modern kangaroos and wallabies), have even lower-crowned and less worn teeth than earlier macropodoids. This is indicative of browse- or even fungi-based diets (
Fig. 1B) among central Australian faunas, which should have been among the first to experience grassland expansion (
18).
Marked increases in crown height and macrowear across the Miocene-Pliocene transition (
Fig. 2, B and C) herald a major adaptive shift in kangaroo evolution. Macropodine crown height increased by up to 40% in as little as 3 million years, similar to rates measured among European and North American ungulates (
12,
17). The Pliocene shift toward more heavily worn teeth is not explained by taphonomic biases (fig. S7) or the weak correlation with increased longevity (fig. S8). Instead, it closely tracks macrowear differences between extant browsing and grass-feeding kangaroos (fig. S9). Simulations that vary the position of dietary selection across the kangaroo phylogeny (fig. S6) point to an adaptive transition at the base of Macropodini as the driver of rapid crown height evolution (table S7). The associated increases in dental wear and crown height strongly suggest that this adaptive transition was linked with intensified selection for dental durability. Increased crown height was probably favored because it delayed loph collapse, when bilophodont molars are relegated to a crushing rather than cutting modality (
19). We infer that increased grass exploitation was the ecological driver because the most pronounced trait shifts (
Fig. 2B) align with isotopic and pollen evidence for mid-Pliocene grassland expansion (
10) (
Fig. 3A). However, the relationship between diet and crown height (or macrowear) among extant kangaroos (
Fig. 1B and fig. S9) implies that most Pliocene macropodines were probably consuming both grass and dicot leaves, consistent with enamel δ
13C values from the chinchilla assemblage (
20). Not until the Pleistocene are macropodins differentiated from other macropodines by heavily worn, higher-crowned teeth (
Fig. 2D), which suggests a much more recent origin of specialized, grazing diets among kangaroos than among Northern Hemisphere herbivores such as horses (
12). The link between generalist Pliocene kangaroo diets and rapid dental evolution highlights how opportunistically consumed foods can be potent drivers of dietary adaptation (
21).
Most kangaroo phylogenies fit a “short-fuse” radiation model (
Fig. 3B), where macropodin generic splits, and many intrageneric splits, occur in the drying late Miocene (
7–
9). However, this fits poorly with evidence for a Pliocene adaptive shift and the absence of any known Miocene macropodins. Our data instead suggest that most modern macropodin genera emerged during the early Pliocene warm phase (
22) and then adaptively radiated during the arid late Pliocene and early Pleistocene as C
4-photosynthesizing grasses expanded (
10) (
Fig. 3, A and C). This scenario implies a rapid origination of the modern macropodin genera (
Lagorchestes,
Macropus,
Onychogalea,
Setonix,
Wallabia) within as little as a million years, with subsequent mid-Pliocene grassland expansion providing ecological opportunity for macropodin cladogenesis and hybridization (
23).
The limited extent of the Australian late Miocene fossil record means that a hitherto concealed pre-Pliocene macropodin radiation cannot be ruled out, but several lines of evidence favor a Pliocene “explosive” diversification model. First, despite >50 years of collecting, the late Miocene Alcoota assemblage of central Australia has yielded no macropodins but does include many specimens of the low-crowned dorcopsin
Dorcopsoides fossilis, sthenurine
Hadronomas puckridgi, and three as yet undescribed, low-crowned, non-macropodin kangaroos. Second, a rapid Pliocene radiation would help to explain the persistent phylogenetic lability of
Setonix,
Onychogalea, and
Wallabia. Third, a rapid Pliocene diversification would require fewer independent convergences on grass feeding, versus at least six required under a “short-fuse” scenario (fig. S10). Finally, trait simulations under a “short-fuse” hypothesis predict high-crowned macropodins well in advance of paleoenvironmental evidence for grassland expansion, whereas an “explosive” model much more closely tracks crown height dynamics observed in the fossil record (
Fig. 3D).
The late Cenozoic has been interpreted as a phase of waning browser diversity (
11), but we find that the archetypal browsing kangaroos, the Sthenurinae, were adaptively radiating through this interval. Macrowear and crown height data indicate that middle and late Pleistocene sthenurines were consuming more abrasive plants than earlier in the late Cenozoic, although low-abrasion browse remained their staple diet (
Fig. 2, B and C). This Pleistocene dietary expansion aligns with a doubling of sthenurine species richness (
24) at a time when Australia was becoming increasingly arid (
5,
25). The association of dietary change with dental adaptation and increasing species richness, amid deepening Pleistocene aridity, discounts the supposed importance of aridity-forced dietary change in sthenurine extinction (
26). Microwear, isotopic, and morphological evidence point to reliance of some middle and late Pleistocene sthenurine kangaroos on chenopod shrubs such as saltbush, which are adapted to low-rainfall, high-salinity conditions (
24,
27). We speculate that a Pleistocene expansion of chenopod biomass may explain why, unlike in North America (
11), Australian browser diversity increased in spite of declining ecosystem productivity.
Aridity has been widely implicated in recent terrestrial diversifications (
6,
28), but our data reveal a more dynamic picture, where warm-to-cool oscillations promote taxonomic diversification followed by ecological and morphological diversification. Warm-wet intervals may prime clades for rapid adaptation during ensuing arid intervals, perhaps by fostering trophic generalists that can radiate when ecological opportunity arrives (
29). Tests of this model, which leverage the Cenozoic record of oscillating climate, hold promise for uncovering how climatically driven ecological change promotes adaptive radiation.
Acknowledgments
We thank the hundreds of volunteers, students, and scientists who collected and prepared specimens. For specimen access and information, we thank Y.Y. Zhen, S. Ingelby, J. Louys, K. Travouillon, K. Spring, S. Hocknull, A. Rozefelds, R. Lawrence, J. Wilkinson, H. Janetzki, K. Butler, A. Camens, L. Nink, D. Pickering, T. Ziegler, R. Palmer, M.-A. Binnie, D. Stemmer, M. Siversson, H. Ryan, L. Umbrello, A. Yates, and P. Holroyd. Valuable comments and discussion were provided by L. Hlusko, D. Polly, three reviewers, and members of the Flinders Palaeontology Laboratory. R. Meredith kindly provided his divergence time tree. M. Rücklin generously supported completion of this project.
Funding: Australian Research Council grants to G.J.P. (DP110100726, FT130101728). A.M.C.C. was supported by an Australian Postgraduate Award.
Author contributions: A.M.C.C. and G.J.P. designed the study and collected data. A.M.C.C. analyzed data and A.M.C.C. and G.J.P. wrote the paper.
Competing interests: None.
Data and materials availability: The data are available in the supplementary materials and alongside the R code on Dryad (
https://doi.org/10.5061/dryad.7b423q5).