The origins and diversification of C 4 grasses and savanna-adapted ungulates

Global Change Biology (2009) 15, 2397–2417, doi: 10.1111/j.1365-2486.2009.01860.x The origins and diversification of C4 grasses and savannaadapted ungulates Y A N I S B O U C H E N A K - K H E L L A D I *w , G . A N T H O N Y V E R B O O M w , T R E V O R R . H O D K I N S O N *, N I C O L A S S A L A M I N z, O L I V I E R F R A N C O I S § , G R A I N N E N Í C H O N G H A I L E * and V I N C E N T S AV O L A I N E N } k *Department of Botany, School of Natural Sciences, Trinity College Dublin, Dublin 2, Ireland, wDepartment of Botany, University of Cape Town, Private Bag X3, Rondebosch 7701, South Africa, zDepartment of Ecology and Evolution, University of Lausanne, CH-1015 Lausanne, Switzerland, §Institut National Polytechnique de Grenoble TIMC-IMAG, TIMB, Faculty of Medicine, F38706 La Tronche, France, }Royal Botanic Gardens, Kew, Richmond TW9 3DS, UK, kImperial College London, Silwood Park Campus, Ascot SL5 7PY, UK Abstract C4 grasses constitute the main component of savannas and are pervasive in other dry tropical ecosystems where they serve as the main diet for grazing animals. Among potential factors driving C4 evolution of grasses, the interaction between grasses and grazers has not been investigated. To evaluate if increased grazing pressure may have selected for higher leaf silica production as the grasses diverged, we reconstructed the phylogeny of all 800 genera of the grass family with both molecular (combined multiplastid DNA regions) and morphological characters. Using molecular clocks, we also calculated the age and number of origins of C4 clades and found that shifts from C3 to C4 photosynthesis occurred at least 12 times starting 30.9 million years ago and found evidence that the most severe drop in atmospheric carbon dioxide in the late Oligocene (between 33 and 30 million years ago) matches the first origin of C4 photosynthesis in Chloridoideae. By combining fossil and phylogenetic data for ungulates and implementing a randomization procedure, our results showed that the appearance of C4 grass clades and ungulate adaptations to C4-dominated habitats match significantly in time. An increase of leaf epidermal density of silica bodies was found to correspond to postulated shifts in diversification rates in the late Miocene [24 significant shifts in diversification (Po0.05) were detected between 23 and 3.7 million years ago]. For aristidoid and chloridoid grasses, increased grazing pressure may have selected for a higher leaf epidermal silica production in the late Miocene. Keywords: C4 ecosystems, CO2 level, coevolution, evolutionary history, grasses, grazing, hypsodonty, phylogenetic trees, silica density, ungulates Received 1 October 2008 and accepted 11 November 2008 Introduction In recent years, there has been increased interest in understanding the evolutionary histories of C4 plants because of their impact on ecosystem functioning, climates (Cerling et al., 1997; Pagani et al., 2005) and the evolution of animals such as hominids and mammals (Janis, 1993; Macfadden & Cerling, 1994). Compared Correspondence: Yanis Bouchenak-Khelladi, Department of Botany, School of Natural Sciences, Trinity College Dublin, Dublin 2, Ireland, tel. 1 27 21 650 3772, fax 1 27 21 650 4041, e-mail: yanis.bouchenak-khelladi@uct.ac.za r 2009 Blackwell Publishing Ltd with the widespread C3 photosynthetic pathway, the more complex but less common C4 pathway is advantageous to plants growing in warmer climates and under conditions of carbon dioxide (CO2) limitation (Sage, 2004). The C4 pathway involves the formation of C4-dicarboxylic acids in the mesophyll cells. These acids then diffuse to the bundle sheath cells and release CO2 that is subsequently catalyzed by RUBISCO (Hatch, 1971). C4 plants have an advantage over C3 plants because they attain higher photosynthetic nitrogen use efficiency (Ehleringer & Monson, 1993). However, the C4 mechanism is not uniform among C4 plants (Taub & Lerdau, 2000). There are three distinct 2397 2398 Y. B O U C H E N A K - K H E L L A D I et al. biochemical pathways, each performed by different C4 species; namely NAD-ME, NADP-ME and PCK pathways after the bundle sheath decarboxylation enzyme used in each (Hatch, 1987). Despite representing only about 3% of all flowering plants, the grass family (Poaceae) alone contains ca. 4500 C4 species (i.e. 60% of all C4 plants; Sage, 2004). Among C4 plants, grasses are ecologically most important because of their dominance in many terrestrial biomes including savannas, tropical grasslands, some temperate grasslands, and most disturbed landscapes in warmer regions of the world. Recent advances in grass phylogenetics have resolved the major relationships at subfamilial and tribal levels (GPWG, 2001; Sanchez-Ken et al., 2007; Bouchenak-Khelladi et al., 2008; Sungkaew et al., 2009). Two major clades have been resolved, the BEP (Bambusoideae, Ehrhartoideae and Pooideae) and PACCMAD clades (Panicoideae, Arundinoideae, Chloridoideae, Centothecoideae, Micrairoideae, Aristidoideae and Danthonioideae). BEP contains exclusively C3 species while PACCMAD includes C3 species, as well as all known C4 grasses. The latter are mainly distributed within Aristidoideae, Chloridoideae and Panicoideae, and include representatives of NAPD-ME, NAD-ME and PCK biochemical type, respectively (Hattersley & Watson, 1992; GPWG, 2001; Hodkinson et al., 2007a, b; Sanchez-Ken et al., 2007; Bouchenak-Khelladi et al., 2008). In a recent study, Stromberg (2005) suggested that external factors triggered changes in vegetation structure during the late Oligocene or early Miocene, promoting the spread of open-habitat grasses. Among the several potential environmental influences on the ecological success of C4 grasses, climate change and low CO2 levels during the Cenozoic are the most commonly discussed (Sage & Monson, 1999; Stromberg, 2005; Christin et al., 2008; Vicentini et al., 2008). Keeley & Rundel (2003) proposed that increased fire frequency may have led to the rise of C4 grasslands. However, the ecological dominance of C4 grasslands did not coincide with the first evolutionary appearance of the C4 photosynthetic pathways in grass lineages. The interaction between low CO2 levels and frequent fires may also have promoted the spread of grassland at the expense of forest trees (Bond et al., 2003). Sage (2001, 2004) suggested that low atmospheric CO2 in the Oligocene causes high photorespiration rates, which facilitated the origin of C4 photosynthesis to proceed. It is documented that C4 grasses and savannas gained ecological dominance during the last 8 million years (Cerling et al., 1997; Beerling & Osborne, 2006), but the evolutionary origins of C4 photosynthesis in grasses needs further study (Kellogg, 2000; Roalson, 2007; Christin et al., 2008). Genetic comparisons suggest that C4 photosynthesis first arose in grasses during the Oligocene (Kellogg, 2000; Sage, 2004; Christin et al., 2008; Vicentini et al., 2008), but low taxonomic sampling has limited the scope of these studies. Christin et al. (2008) showed that C4 photosynthesis might have evolved 17 or 18 times, apparently being promoted by the decline of CO2 in the Oligocene. Vicentini et al. (2008) sampled 97 species, with an emphasis on Panicoideae, and found four to five origins for C4 photosynthesis in grasses also occurring in the Oligocene. However, as suggested by Sage (2001, 2004), the importance of low CO2 levels at the Oligocene transition is not necessarily the sole driver for C4 evolution, but rather a precondition. Studies on additional ecological factors are needed to understand the origins of this biological innovation (Roalson, 2008). The role of herbivory in the promotion of C4 grasses has not been investigated in detail, even though the spread of C4 grasslands may have been associated with increased grazing rates through the Miocene (Chapman, 1996). Sage (2001) suggested an evolutionary interaction between grasses and herbivores as a factor driving C4 evolution. A major and rapid radiation of vertebrate herbivores is documented to have occurred between 20 and 10 million years ago (MacFadden & Cerling, 1994; Hassanin & Douzery, 1999), along with a near simultaneous rise to ecological dominance of grasses (Stromberg, 2005), suggesting that grasses coevolved tightly with vertebrate herbivores (Stebbins & Crampton, 1961). The shifts from low- to high-crowned teeth in ungulates are likely to be related with shifts in their diet toward more fibrous vegetation (Jernvall et al., 1996) and shifts from browsing to grazing (Janis et al., 2002). The evolution of antiherbivore defense mechanisms may have occurred in response to the diversification of the world’s megaherbivore fauna (Coughenour, 1985; Chapman, 1996). Silica bodies are thought to reduce palatability, digestibility and the nutritional value of the forage grasses (Coughenour, 1985; Ellis, 1990; Chapman, 1996; Massey & Hartley, 2006), and are among the few substances capable of inducing morphological changes to animal mouthparts (Piperno, 2006). One might expect that C4 grasses increased their toughness and abrasiveness to resist grazing by, for instance, increasing the densities of silica bodies in their leaves (Massey & Hartley, 2006). In response to this defense, grazer ungulates may have evolved longer-lasting highercrowned cheek teeth that enable them to consume abrasive plant tissues, but evidence for such an evolutionary interaction between grasses and grazers remains elusive (Stromberg, 2006). In an attempt to test the hypothesis that grazing pressure selected for increased phytolith production r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S through time as grasses diverged, we produced a new thoroughly sampled phylogenetic tree for all 800 grass genera and used molecular clock techniques to date the origin of its clades. We also quantified the number, and geological timing, of shifts from C3 to C4 photosynthesis that have occurred during the evolution of grasses and compared our findings with those of Christin et al. (2008). Furthermore, using a complete species-level phylogenetic tree of ungulates (Bininda-Emonds et al., 2007), we dated the age of savanna-adapted species and correlated temporal changes in the silica densities of C4 grass leaves with changes in diet-related tooth morphology of ungulate fossils. Using a recently introduced statistical test, implemented as ‘shift.test’ in APTREESHAPE (Bortolussi et al., 2006), we also estimated when phylogenetic diversification of C4 grasses occurred. Materials and methods DNA extraction, amplification and sequencing We extracted DNA of 90 grass species (Appendix A). The rbcL and matK exons (encoding the large subunit of RUBISCO and maturase K, respectively) as well as the trnL-trnF (trnL intron and trnL-F intergenic spacer) region were sequenced for each species or downloaded from GenBank/EBI if available (Appendix A). Between 0.1 and 0.5 g of silica-gel or herbarium-dried leaf (Chase & Hills, 1991) or up to 1 g of fresh leaf (or seed) was used for DNA extraction. Total genomic DNA was prepared following the CTAB method (Doyle & Doyle, 1987). For herbarium material the CTAB protocol was modified by precipitating the DNA with propan-2-ol instead of ethanol and then storing samples at 20 1C for 4 weeks. All DNA extracts were purified by caesium chloride/ethidium bromide gradients. For herbarium samples, concentrated DNA extracts were obtained by cleaning 100 mL of dialyzed solutions through QIAquickt Spin Columns (QIAGEN Ltd, West Crawley, UK). The three DNA regions were amplified using a GeneAmps PCR System 9700 thermal cycler (ABI, Applied Biosystems, Warrington, Cheshire, UK) using the polymerase chain reaction (PCR). PCR reaction volumes (50 mL) included between 1 and 1.5 mL of template DNA (with DNA concentrations mostly ranging from 400 to 1200 ng mL1), 1 mL of a 0.4% bovine serum albumin solution, 0.5 mL of forward and reverse primers (100 ng mL1), 45 mL of 1.1  ReddyMixt PCR Master Mix [1.25 U Thermoprime Plus DNA polymerase, 75 mM Tris-HCl pH 8.8, 2.5 mM MgCl2, 0.2 mM for each dATP, CTP, GTP, TTP, 20 mM (NH4)2 SO4] and between 1.5 and 2 mL of sterile ultrapure water (MilliQ). Cycle sequencing reactions were carried out in a GeneAmp 2399 PCR System 9700 thermal cycler using the ABI BigDye Terminator v3.1. Cycle Sequencing Kits. After a series of cleaning with 250 mL of 70% ethanol, samples were suspended into 10 mL HiDit formamide (ABI) and run on a 3100 Automated DNA sequencer (ABI). Contigs were assembled using AutoAssembler 2.1 (ABI). DNA sequences were aligned unambiguously by eye. Five hundred and seventy-eight characters in the trnL-F region were excluded from the analyses as their alignment was ambiguous. Phylogenetic analyses We built a 90-taxon backbone tree based on Bayesian phylogenetic analyses of the three plastid DNA regions with 5060 molecular characters, comprising representatives of all but one (Puelioideae) subfamilies sensu the Grass Phylogeny Working Group (GPWG, 2001) and 24 of the 44 tribes. The substitution model used for the three different regions was determined using a hierarchical likelihood ratio test framework as implemented in MODELTEST 3.06 (Posada & Crandall, 1998). The optimal models identified were HKY 1 G 1 I (Hasegawa et al., 1985) for the rbcL data, TVM 1 G 1 I (Posada & Crandall, 1998) for the matK data and K81 1 G 1 I (Kimura, 1981) for the trnL-F data. The combined matrix was analyzed using Bayesian inference by partitioning the sequences by DNA region (Nylander et al., 2004). This allowed independent estimation of parameters for each partition. Site-specific rates of substitution were allowed to vary across partitions (ratepr 5 variable). The HKY 1 G 1 I model was used for the rbcL, and the more general GTR 1 G 1 I model (Yang, 1994) was used for the matK and the trnL-F data. The matK and trnL-F sequences were analyzed using the GTR substitution model as neither the TVM nor the K81 models can be implemented in MRBAYES 3.0b4 (Huelsenbeck & Ronquist, 2001); four parallel MCMC were run for 2 000 000 generations with trees sampled every 1000 generations. To assess the adequacy of sampling from the posterior probability distribution and to determine the burn-in, we plotted changes in likelihood over 2 000 000 generations using the sump function implemented in MRBAYES 3.0b4 (Huelsenbeck & Ronquist, 2001), and compared them across the four independent runs, which started from different, randomly chosen trees. The Bayesian consensus tree was then used as a constraint in a parsimony search in which all 800 grass genera (Watson & Dallwitz, 1992) were represented using morphological data (417 characters for each genus, excluding C3/C4-related characters), plus Elegia (Restionaceae) and Joinvillea (Joinvilleaceae) as outgroups (GPWG, 2001; Bouchenak-Khelladi et al., 2008). An initial parsimony search used 500 replicates of tree-bisection recon- r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2400 Y. B O U C H E N A K - K H E L L A D I et al. nection and kept a maximum of 15 trees per replicate. The consensus tree was then used as a starting tree for a second round of swapping, which resulted in 10 equally most parsimonious trees. For ungulates, we pruned the mammal supertree of Bininda-Emonds et al. (2007) and resolved bovid relationships with information from Bro-Jorgensen (2007) in order to minimize the number of polytomies because they increase the number of ambiguities in ancestral state reconstructions (Jordan et al., 2008). Molecular dating Because of nonconstancy of the molecular clock (likelihood ratio test: ln L 5 42450.46707c; ln L43104.07400 1 c; df 5 86, Po0.0001), we dated the 90-taxon tree with a relaxed Bayesian clock using MULTIDIVTIME (Thorne & Kishino, 2002), with five independent fossil calibration points. We used the grass spikelet from the Paleocene–Eocene boundary (Crepet & Feldman, 1991) to provide a minimum age of 55 million years for the crown node of the BEP 1 PACCMAD clade (i.e. all taxa excluding Anomochloa, Pharus and Streptochaeta; see Appendix A for details). The fossil of Cleistochloa (Dugas & Retallack, 1993) provided a minimum age of 14 million years for the crown node of Paniceae (i.e. Cenchrus, Digitaria, Echinochloa, Panicum and Pennisetum). The fossil of Distichlis (Dugas & Retallack, 1993) provided a minimum age of 14 million years for the crown node of Chloridoideae. The oldest grass fossils, silica bodies from dinosaur coprolites, have been found in India and are estimated to be between 65 and 67 million years old (Prasad et al., 2005). According to Prasad et al. (2005), several grass lineages were present at that time: early-diverging lineages (Anomochlooideae and Puelioideae) as well as members of the ‘spikelet’ group (BEP and PACCMAD members). There is considerable uncertainty over the age of the root node. These oldest fossils could potentially have been deposited long after the emergence of the lineages they represent. Therefore, we provided a minimum age for the crown group of grasses at 90 million years old because the ‘spikelet’ grass lineages are thought to have diversified between 15 and 20 million years after the divergence of Anomochlooideae and Puelioideae (Bremer, 2002; Christin et al., 2008). It suggests that the origin of grasses occurred at least 20 million years earlier than these oldest grass fossils found in India. We have taken a conservative 90 million years because of the age of the Prasad et al. (2005) fossil and estimations from Bremer’s work combined. In terms of sensitivity analysis, an alternative calibration was attempted (Bouchenak-Khelladi, 2007) without taking into account the Prasad et al. (2005) fossil. In this case, the dates of major clades were found to be approximately 10 million years younger. However, we felt it is impossible to ignore the novel findings of Prasad et al. (2005) because it constitutes the oldest grass fossils found to date. As required by the Bayesian analysis, we enforced the root of the tree with a maximum age; in this case we set the root node of the grasses to be o125 million years, which represents the age of the tricolpate pollen grain of eudicots (Drinnan & Crane, 1994). This chronogram provided 87 dated nodes that were mapped onto the 800-genus tree. Following Purvis (1995), dates for those nodes that did not possess an estimate (i.e. 713 nodes) were interpolated using a pure birth-model, under which a clade’s age is proportional to the logarithm of the number of lineages it contains. We used chronographer.pl and relDate.pl to perform this dating (BinindaEmonds et al., 2007). This was repeated with all the lower and upper confidence limits (95% confidence intervals) of the MULTIDIVTIME analysis. For ungulates we used the published supertree of mammals, which was already dated by the authors using the same method described above (Bininda-Emonds et al., 2007). Character optimization For grasses, we used the anatomical leaf ‘maximum cells-distant count’ character, defined as the interveinal distances, which range from one mesophyll cell for C4 species to two or more for C3 species (Hattersley & Watson, 1975). It is a reliable indication of the photosynthetic pathway possessed by grasses (Hattersley & Watson, 1975; Watson & Dallwitz, 1992). For ungulates, we reconstructed shifts between C3- and C4-related habitats on a dated phylogenetic tree of the ungulates using maximum likelihood (Pagel, 1999; Lewis, 2001). To identify which species occupy C4-dominated habitats (tropical/arid savannas and grasslands in Africa, South America and Asia restricted), we filtered the Animal Diversity Website database (Myers et al., 2006), and considered that savanna-adapted ungulates are grazers; that is, ungulates with hypsodont teeth and squared-up jaw. We optimized habitats onto the dated supertree of ungulates and identified shifts in habitat preference. To infer the nodes at which shifts in photosynthetic pathways and habitat preference occurred, we optimized these characters with the Mk1 model of evolution (Lewis, 2001), implemented in MESQUITE v1.12 (Maddison & Maddison, 2006). We evaluated whether these characters were distributed randomly with regard to the phylogeny by using a common and robust test implemented in a parsimony framework to evaluate clumping, in MACCLADE v.4.08 (Maddison & Maddison, 2005). We compared the observed length of the opti- r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S mized character against a null distribution obtained by reshuffling the character randomly 1000 times on the tree. We also used randomization tests to evaluate whether the mean ages of the origins of C4 photosynthesis in grasses, and of transitions to open habitats in ungulates, were more or less recent than would be expected on the basis of chance alone. The null model assumes that the forward and backward rates of change for these traits (i.e. C4 or open habitats) are uniform over the entire grass phylogeny, and over the ungulate phylogeny. The appearance of C4 in grasses, and of open habitats in ungulates, was estimated by maximum likelihood, using a two-parameter model that is similar to the Assymp2p model implemented in MESQUITE v1.12 (Maddison & Maddison, 2006). For our purpose, the trait was assumed to have appeared as soon as its state probability at an ancestral node exceeded 0.5. This test is not used in order to link C4 evolution with the evolution of ungulate habitat preference but rather evaluates whether the estimated age of appearances of C4 pathway and savanna-adapted grazers were due to a signal in our data rather than to chance alone. Finally, we used 3591 fossils from the comprehensive Neogene record of the Old World mammals (Jernvall & Fortelius, 2002) to compute the abundance of artiodactyls and perissodactyls with brachydont, mesodont or hypsodont teeth recorded in three million-year windows. This allowed us to estimate changes in the relative abundance of hypsodonts through time. Silica density Mature leaf samples (one per species, as per those used for the DNA work) were fixed in FAA (ethyl alcohol 70%, formaldehyde 40%, acetic acid; 90 : 5 : 5). Abaxial epidermal scrapes were prepared by treating tissues for 30 s in 3.5% sodium hypochlorite and manually scrapping off mesophyll with a scalpel blade. Samples were stained for 5 min with red safranin and alcian blue. Epidermal scrapes were washed in water and placed in ethanol (50%, 75%, 90%, 96% and 100%, successively) for 1 min each before being transferred to xylene, mounted on microscopic slides and photographed. Using an Olympus CAST Stereology Systems (Olympus, Centre Valley, PA, USA), we computed an index of silica density per unit of surface by counting the numbers of silica bodies in 10 random 100 mm  100 mm quadrats and multiplied this value by the average surface area of 10 silica bodies. To test whether the density of silica bodies showed an increase in C4 grasses, we used a generalized estimating equation test (Paradis & Claude, 2002). The phylogeny of grasses was used as a covariable, and the effect of C3 and C4 grasses on the density of silica bodies was tested by a General Linear 2401 Model approach, which expresses observed random variables as a function of unobserved constants or random variables (Nelder & Wedderburn, 1972). Silica density was modelled using a Gamma distribution. Shifts in diversification rates Phylogenetic diversification rate shifts were detected on the complete generic-level phylogenetic tree (800 genera representing 10 176 species) following the approach described in Moore et al. (2004). The taxonomic level at which the tree was reconstructed was the genus level: the number of species in each genus was attached to the 800 taxa-tree. Shifts were detected using a likelihoodbased approach that evaluated the relative fit of models with one- or two-rate parameters distributed over different parts of a three-taxon tree. The key quantity to compute was a likelihood ratio that represented the relative fit of the one- and two-rate parameter models to the observed diversity partition. This was assessed by the difference in the natural logarithm of the respective likelihood values in homogeneous and heterogeneous diversification rate models, similar to the test statistic implemented in the computer program SYMMETREE (Moore et al., 2004). However, our sample size exceeded the 6000-taxon limit imposed by SYMMETREE. Therefore we used mathematical arguments taken from the theory of branching processes to derive a simpler formula for the logarithm of the likelihood ratio, which have been implemented in the R package APTREESHAPE (Bortolussi et al., 2006). The details of the derivation of the new formula can be found in Appendix B. Results The DNA sequences used to build the 90-taxon grass phylogeny are given in Appendix A and deposited in GenBank/EBI (EF125079–EF137619). The Bayesian consensus tree is shown in Appendix C. It was a wellresolved tree with all major subfamilies being monophyletic and well supported (1.00 Posterior Probabilities, PP hereafter; Appendix C). The sister relationship of BEP and PACCMAD was strongly supported (1.00 PP; Appendix C). The subfamilial relationships were well supported (1.00 PP; Appendix C), except for the monophyly of the Aristidoideae 1 Arundinoideae 1 Chloridoideae 1 Danthonioideae clade (0.59 PP; Appendix C). Arundinoideae (mainly C3 grasses) were the earliest-diverging lineage (0.59 PP; Appendix C), followed by Aristidoideae (C4, 0.92 PP; Appendix C), with Danthonioideae (C3) being sister to Chloridoideae (C4, 0.97 PP; Appendix C). Panicoideae (C4) were sister to the exclusively C3 Centothecoideae (1.00 PP; Appendix C). Within Panicoideae, tribes Andropogoneae and r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2402 Y. B O U C H E N A K - K H E L L A D I et al. r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S Paniceae were both monophyletic (1.00 PP; Appendix C). Two early-diverging lineages, Anomochlooideae and Pharoideae, were resolved, but Puelioideae were not sampled in this study. This well supported phylogenetic tree was then used for a subsequent constrained parsimony inference including all grass genera. Of the 417 morphological and anatomical characters used to infer the comprehensive generic-level phylogenetic tree, 360 were parsimony informative. After a series of TBR swapping, the 10 shortest trees found had 9983 steps with a consistency index of 0.04 and a retention index of 0.93. The phylogenetic tree is summarized in Fig. 1 and is also available online at TREEBASE (SN4105, http://www.treebase.org/treebase/). According to our molecular dating, grasses originated in the Cretaceous (about 90 million years ago, mya hereafter; Fig. 1). The BEP clade originated in the Paleocene whereas the younger PACCMAD clade originated in the late Eocene (57 and 40 mya, respectively; Fig. 1). Within the PACCMAD clade, the two largest subfamilies (Chloridoideae and Panicoideae, which include all the C4 grass species except those in Aristidoideae) originated between 25 and 30 mya in the early Miocene–late Oligocene (Fig. 1). The ungulate chronogram is shown in Fig. 4 and is also available online at TREEBASE (SN4105). Artiodactyla were monophyletic and appeared in the late Cretaceous (about 75 mya). Bovideae (Alcelaphineae, Antilopineae, Bovineae, Caprineae, Cephalopineae, Hippotragineae and Reduncineae) and Cervideae (Capreolineae and Cervineae) originated in the late Oligocene (around 30 mya; Fig. 4). Bovideae and Cervideae diversified at about the same time, in the early–middle Miocene (25– 15 mya; Fig. 4). Perissodactyla were retrieved as outlying lineages from a common ancestor with Artiodactyla at least 90 mya (Fig. 4). Equideae, Tapirideae and Rhinocerotideae originated in the late Paleocene (around 60 mya) and diversified by the early Miocene (around 25 mya; Fig. 4). We found that the distributions of the characters of interest, photosynthetic pathway and C4-habitat adaptation, were heavily clumped (P 5 0.0001) when optimized onto the phylogenetic trees (Fig. 2). For grasses, we found that shifts from C3 to C4 photosynthesis 2403 occurred at least 12 times (labelled 1–12 on Fig. 1), starting in the Oligocene 30.9 mya when the CO2 level also underwent its most abrupt decline (Pagani et al., 2005) (Table 1; Fig. 3). In the middle–late Oligocene, one Fig. 2 (a) Distribution of tree lengths (x-axis) for C4/C3 optimization on the grass tree. (b) Distribution of tree lengths (x-axis) for C4/C3-habitats optimization on the ungulate tree. The observed length is represented as thick vertical bars vs. lengths after each of 1000 random reshuffling (occurrence on y-axis). The observed tree length is significantly shorter indicating that the trait is clumped in the phylogeny (P 5 0.0001). Fig. 1 Comprehensive generic-level chronogram of the grasses showing the timing of origin of the major clades and subfamilies. Panel A includes Aristidoideae, Arundinoideae, Chloridoideae and Danthonioideae. Panel B includes tribes Andropogoneae, Arundinelleae and Paniceae and subfamily Centothecoideae. Panel C includes Anomochlooideae, Bambusoideae, Ehrhartoideae, Pharoideae, Pooideae and Puelioideae. A summary tree is provided in each panel and refers to panels A, B and C with the two major clades BEP and PACCMAD. BEP refers to the BEP clade sensu the GPWG (2001) and PACCMAD refers to the PACCMAD clade sensu Sanchez-Ken et al. (2007). Twenty-four significant shifts (D1–D24) in diversification (Po0.05) were detected. Table 3 provides inferred dates and confidence intervals for the shifts at nodes D1–D24. Twelve grass clades (1–12) were found to have shifted from C3 to C4 photosynthesis (C3 in blue, C4 in red; ages and confidence intervals for clades 1–12 are provided in Table 1). Horizontal bars indicate 95% confidence intervals from molecular dating. r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2404 Y. B O U C H E N A K - K H E L L A D I et al. Table 1 Twelve clades that have shifted to C4 photosynthesis and their ages in million years No. Clades Age Upper Bound Lower Bound 1 2 3 4 5 Chloridoideae* Aristida/Stipagrostis Centropodia Andropogoneae* Cenchrinae (i.e. Achlaena/Chaetium/Neurachne/Paraneurachne/Thyridolepis/Acritochaete/Trachys/ Anthephora/Chlorocalymma/Targidia/Chaetopoa/Beckeropsis/Chamaeraphis/Odontelytrum/ Xerochloa/Paratheria/Stenotaphrum/Taeniorhachis/Streptolophus/Cenchrus/Pennisetum/ Pseudochaetochloa/Spinifex/Zygochloa/Hymenachne/Louisiella/Oryzidium/Pseudoraphis/ Snowdenia/Melinis/Rhynchelytrum) Dissochondrus Scutachne Yakirra Ancestral node to Anthenanthia/Leptocoryphium/Baptorhachis/Stereochlaena/Centrochloa/ Leptosaccharum/Spheneria/Ophiochloa/Triscenia/Digitariopsis/Leptoloma/Yvesia/Tricholaena/ Mesosetum/Tatianyx/Anthaenantiopsis/Axonopus/Paspalum/Thrasya/Thrasyopsis/Urochloa/ Brachiaria/Camusiella/Cymbosetaria/Setaria/Paspalidium/Ixophorus/Whiteochloa/Eriochloa/ Eccoptocarpha, etc. Acostia/Alloteropsis/Megaloprotachne/Echinochloa/Thuarea/Leucophrys/Steinchisma Arthagrostis/Hydrothauma/Sacciolepis/Paractenium/Plagiosetum/Setariopsis Digitaria 30.9 15.2 4.1 28.6 18.9 36.9 20.4 4.8 34.4 24 24.9 10 3.3 22.9 14.7 2.9 4.6 13 18.7 3.6 5.7 16.2 23.2 2.2 3.4 9.7 14.2 9.3 8.4 9.7 11.4 11.4 11.7 7.2 7.2 7.7 6 7 8 9 10 11 12 *See Watson & Dallwitz (1992) for list of genera. shift from C3 to C4 occurred at the origin of the Chloridoideae (shift 1) and another at the origin of the Andropogoneae (shift 4) (Table 1, Fig. 1). In the early Miocene, one shift to C4 occurred at the origin of Aristidoideae (shift 2) and nine shifts occurred within Paniceae (shifts 5–12) (Table 1, Fig. 1), with several reversals to C3 photosynthesis (Fig. 1). For ungulates, we inferred that occupation of C4 ecosystems started 26.2 (28.2–23.7) mya within Bovideae, just a few million years after the appearance of the first C4 grasses (Tables 1 and 2). Subsequent occupations of C4 ecosystems within Bovideae, Cervideae and Perissodactyla occurred in the early to late Miocene (from 23 to 5 mya; Fig. 4 and Table 2). Randomization tests show that, on average, the C4 photosynthetic pathway in grasses originated earlier than would be expected on the basis of chance alone (P 5 0.002; Fig. 5). The time of occupation of C4-habitats by ungulates, Fig. 3 (a) Proportion of brachydont/mesodont (dark), and hypsodont (light) teeth recorded in three million-year intervals (throughout the Neogene of the Old World), showing that the occurrence of hypsodonty increased through the Miocene. (b) Partial pressure of atmospheric CO2 in the Paleogene (redrawn with permission Pagani et al., 2005; Henderiks & Pagani, 2008), showing that the most acute drop in atmospheric CO2 occurred in the late Oligocene. r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S 2405 Fig. 4 Comprehensive species-level chronogram of Artiodactyls and Perissodactyls from Bininda-Emonds et al. (2007). Shifts between C3- and C4-related habitats were optimized using maximum likelihood (Pagel, 1999; Lewis, 2001). C4-dominated habitat optimizations are shown in red. Table 2 provides inferred dates and confidence intervals for the shifts between C3- and C4-related habitats. however, did not differ significantly from random (P 5 0.086; Fig. 5). We found 24 significant shifts of phylogenetic diversification rates among the C4 lineages (Po0.05; D1–D24, Fig. 1; Table 3). The ages of these shifts are heavily biased toward recent time, with five shifts in the Pliocene and 19 shifts in the Miocene. The most recent shifts in phylogenetic diversification rate occurred within Panicoideae whereas the most ancient ones occurred in the early Miocene within Chloridoideae (Fig. 1). A linear regression analysis confirmed that hypsodonty increased significantly only in the last 22.5 million years (r2 5 0.77; Fig. 3). According to our results, shifts in phylogenetic diversification rates and increases in hypsodonty occurred concurrently starting in the Miocene and ending in the late Pliocene (Figs 1 and 3; Table 3). Finally, in contrast to the pattern for C3 lineages (especially BEP clade members), we found that silica density increased significantly in C4 lineages (C4 slope 5 1 0.31, P 5 0.006). More importantly, the greatest increases in silica density among the C4 grass lineages occurred in the middle to late Miocene (Figs 1 and 6), coincident with the increased proportion of hypsodont ungulates (Fig. 3). The greatest increases in silica density occurred at the origin of Aristidoideae lineages and within Chloridoideae (Fig. 6). Discussion Numbers of origins of C4 grasses Using one of the most comprehensive phylogenetic reconstructions of a large angiosperm family to date, we estimate that shifts from C3 to C4 photosynthesis in grasses occurred at least 12 times between the late Oligocene and the late Quaternary. Sinha & Kellogg (1996) used phylogenetic trees from Barker et al. (1995) and Clark et al. (1995) to infer four independent origins of C4 photosynthesis in the grasses. Subsequently, Kellogg (2001) mapped five origins using the GPWG (2001) phylogeny of the grasses, while Christin et al. (2008) identified 18 origins in the PACCMAD lineage alone. Vicentini et al. (2008) did not sample as many taxa as Christin et al. (2008) and the present study and, therefore, retrieved a fewer number of origins. Also, there are major differences in the phylogenetic tree topologies as Aristidoideae are found as the earliest-diverging lineage within the PACCMAD clade, and there is no r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2406 Y. B O U C H E N A K - K H E L L A D I et al. sister relationship between Panicoideae and Aristidoideae 1 Arundinoideae 1 Chloridoideae 1 Danthonioideae 1 Micrairoideae (Vicentini et al., 2008). The optimization of C4 vs. C3 photosynthetic traits onto the phylogenetic tree relies on the assumption that grass genera are monophyletic, an assumption that has been shown to be misleading by the polyphyly of Panicum and/or Eragrostis (Aliscioni et al., 2003; Roodt-Wilding & Spies, 2006). Christin et al. (2008) sampled 187 grass species, which counted for 65 out of 355 C4 genera according to Watson & Dallwitz (1992). We sampled more taxa (i.e. all 800 grass genera) but missed the origin of C4 within genera. One appearance of C4 in Christin et al. (2008) is due to the polyphyly of Panicum (Panicum prionitis and Panicum anceps), while a second appearance is due to the polymorphic status of the photosynthetic pathway of Neurachne, which exhibits both C3 and C4. Finally, the other four appearances reported by Christin et al. (2008) and missed by us, resulted from the differences in topology with the phylogenetic tree presented in this study. We resolved Aristida and Stipagrostis as sister genera, resulting in a single appearance of C4 for Aristidoideae (Fig. 1 and Table 1), whereas Christin et al. (2008) resolved Sartidia as sister to Stipagrostis, resulting in two independent origins (one in Aristida and one in Stipagrostis). Our analysis also located the clade comprising Danthoniopsis, Tristachya and Loudetia within Paniceae, whereas Christin et al. (2008) retrieved it within Centothecoideae, identifying an origin of C4 photosynthesis within that subfamily. Christin et al. (2008) also identified two independent origins in Echinochloa and Alloteropsis, whereas we found a single appearance of C4, these two genera being resolved within a single clade (Table 1). All the other shifts are consistent between the two studies. One major issue regarding the present comparison between the results of Christin et al. (2008) and our study is the differences in the methods used for phylogenetic inferences. Christin et al. (2008) inferred evolutionary relationship using two plastid DNA markers whereas our study used a set of plastid DNA sequences and morphological characters to reconstruct a comprehensive generic-level tree. The differences between Christin et al. (2008) and our study do not, however, affect the results presented here, as all the shifts not retrieved by us occurred more recently than the evolution of the grazer lineages considered here. They do not affect our comparative analysis with grazing mammals. Diversification of C4 lineages The origins of several lineages of C4 grasses are relatively ancient (28.6 and 30.9 mya for Andropogoneae and Chloridoideae, respectively; Fig. 1 and Table 1) and Table 2 Age of shifts by ungulates to C4-grass-dominated habitats in million years* Taxa Age Upper bound Lower bound Mazama gouazoupira Cephalopus rufilatus Cephalopus silvicultor Bos taurus/sauveli Dama dama Odocoileus Hippocamelus antisensis/Ozotoceros Hemitragus jayakari/hylocrius Blastocerus Boselaphus Axis axis Cephalopus zebra Cephalopus niger Giraffa Capra nubiana Taurotragus/Tragelaphus strepsiceros Tragelaphus scriptus Antilopinae 1 (Gazella, Antilope, Lithocranius, Antidorcas, Ammodorcas) Lama Syncerus Camelus dromedarius Diceros/Ceratotherium Phacochoerus Ammotragus Hyemoschus Tragelaphus imberbis Tragelaphus angasii Oryx/Addax/Hippotragus/ Alcelaphus/Connochaetes/ Damaliscus Hylochoerus Equus Raphicerus/Dorcatragus Reduncini (Kobus, Redunca, Pelea) Ourebia Neotragus moschatus/batesi Madoqua Aepyceros 0.1 1.4 2.9 3.9 4 5.1 5.2 5.5 6.3 7.3 7.9 8.2 8.2 8.3 8.6 9.3 11.4 12.1 0.1 3 3.8 4.7 4.8 10.8 6.5 6.5 11 15.1 9.8 7 7 9.5 12.3 10.7 14.8 23.2 0.1 0.1 2 3 3.2 0.1 3.2 4.6 3.4 0.1 8.1 8.2 8.2 7.1 4.9 7.8 8.1 7.1 12.3 14.4 14.7 15.6 17.3 18.5 19.2 19.6 19.6 22.2 23.1 14.8 28.5 19.8 17.3 19.9 26.8 25.1 25.1 25.6 8.7 14.1 0.9 11.4 16.8 16.4 11.7 13.6 13.6 18.8 23.1 25 26.2 26.2 26.2 26.2 26.2 26.2 27.6 51.6 28.7 28.7 28.7 28.7 28.7 28.7 16.8 11.1 23.7 23.7 23.7 23.7 23.7 23.7 *This table provides the list of taxa that have shifted to C4grass-dominated habitats and the ages of these shifts (including upper and lower bounds). The habitats for each taxa come from querying the Animal Diversity Website (Myers et al., 2006). Depending on the experts, however, certain taxa could have been attributed to the other type of habitats (C3- vs. C4dominated). We list here such ambivalent taxa, although we note that this would not affect significantly our results (Camelus dromedarius, Cephalophus silvicultor, C. rufilatus, C. zebra, C. niger, Giraffa camelopardalis, Lama glama, L. guanicoe, L. pacos, Sylvicapra grimmia, Neotragus spp., Hyemoschus aquaticus, Hylochoerus, Rhinoceros unicornis). r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S 2407 coincident with late-Oligocene CO2 decline (Fig. 3), which is consistent with recently published studies by Christin et al. (2008) and Vicentini et al. (2008). However, our results also suggest that the diversification of these lineages was more recent, corroborating evidence for the recentness of C4 grasses dominance as revealed by phytolith (Stromberg, 2005) and carbon and oxygen isotope data (Zazzo et al., 2000; Tipple & Pagani, 2007). The ‘CO2 starvation hypothesis,’ which postulates that atmospheric CO2 fell below a critical threshold during the Miocene allowing C4 plants to expand in dominance (Cerling et al., 1997), has been shown to be inaccurate using multiple atmospheric proxy records (Osborne & Beerling, 2006). The importance of low CO2 is not so much as a driver of C4 evolution, as a precondition (Sage, 2004). Therefore, atmospheric CO2 decline was unlikely to have been the sole driver of worldwide C4 grass expansion during the Miocene (Beerling & Osborne, 2006). Phylogenetic diversifications detected within the PACCMAD clade occurred between the mid-Miocene and the late Quaternary (Fig. 1 and Table 3), suggesting that C4 grasses may have constituted only a minor part of grass species diversity, and possibly of total vegetation biomass, before the Miocene. Bond et al. (2003) suggested that low CO2 levels in concert with wildfire could have facilitated the replacement of C3 forests by C4 grassland and savannas leading to a spread of open, high-light environments in the late Tertiary. The Miocene C4 grass expansion thus probably represents an ecological response to disturbance rather than a change in specific atmospheric conditions (Osborne & Beerling, 2006), an idea supported by fluxes of black carbon in deep-sea records of the North Pacific Ocean in the late Miocene (Herring, 1985). Recent and detailed studies suggest that late- Fig. 5 Randomizations for (a) number of occurrences and (b) ages of C4 trait in grasses and C4-related habitats in ungulates. The null model assumes that the forward and backward rates of change for these traits are the same along the whole grass and ungulate evolutionary trees (observed data represented as thick vertical bars). To be conservative, 35 taxa for which their positions in the phylogeny were odd and did not match the general classifications from Watson & Dallwitz (1992) or Clayton & Renvoize (1986) were deleted from the tests (i.e. Ampelodesmos, Boissiera, Brachyelytrum, Buchlomimus, Chandrasekharania, Chasechloa, Chloachne, Crinipes, Danthoniastrum, Decaryella, Diandrolyra, Elytrophorus, Eriachne, Euthryptochloa, Gouldochloa, Habrochloa, Hydrochloa, Leptosaccharum, Luziola, Metcalfia, Microcalamus, Nematopoa, Notochloe, Ochlandra, Opizia, Phaenosperma, Pheidochloa, Poecilostachys, Pringleochloa, Pseudodanthonia, Pseudozoysia, Styppeiochloa, Zenkeria, Zizania, Zizaniopsis). P-values number of origins: grasses (P 5 0.001), ungulates (P 5 0.001); mean: grasses (P 5 0.002), ungulates (P 5 0.086). r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2408 Y. B O U C H E N A K - K H E L L A D I et al. Table 3 Nodes at which significant shifts in diversification were detected (Po0.05) among C4 grass lineages (see Fig. 2 for localization on the tree) Nodes Description D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D11 D12 D13 D14 D15 D16 D17 D18 D19 D20 D21 D22 D23 D24 Age (in million years) Upper bound Lower bound Period Capillepidium and the rest Cymbopogon and the rest Asthenochloa clade and the rest Bhidea and the rest Stenotaphrum clade and the rest Chaboissiea and the rest Anadelphia clade and the rest Schoenefeldia and the rest Eremochloa clade and the rest Pennisetum clade and the rest of Paniceae Stiburus and the rest Enteropogon and the rest Leucophrys/Steinchisma and the rest Tricholaena and the rest Hemisorghum/Pseudosorghum and the rest of Andropogoneae Sorghum clade and the rest of Andropogoneae Mesosetum/Tatianyx and the rest Sehima and the rest Afrotrichloris clade and the rest Pappophorum and the rest Tridens and the rest Cladoraphis/Redfieldia and the rest of Chloridoideae Swallenia/Vaseyochloa and the rest Crinipes clade and the rest of Chloridoideae Miocene C4 expansion was regionally heterogeneous rather than globally synchronous, indicating that local or regional environmental factors rather than global climatic factors drove C4 plant expansion mainly in warm regions of the globe (Tipple & Pagani, 2007; Edwards & Still, 2008). However, Kürschner et al. (2008) found that CO2 levels are tightly coupled with climate in the Miocene. Major CO2 fluctuations seem to have occurred during this period, which suggest that changes in Miocene terrestrial ecosystems (such as the expansion of grassland biomes and the radiations of terrestrial herbivores) may be linked to fluctuations in CO2 levels (Kürschner et al., 2008). Even though some C4 lineages may have responded to increased grazing pressure throughout the Miocene, global mean temperature could have affected the evolution of grasslands, and thus the evolution of the different photosynthetic pathways in grasses. Linking the evolution of ungulates and leaf epidermal silica densities of C4 grasses One of the commonly discussed nonclimatic factors that could have driven C4 grass evolution is herbivory (Coughenour, 1985; Belsky, 1986; Janis, 1993; Macfadden & Cerling, 1994; Chapman, 1996; Jernvall et al., 1996; 3.7 4.6 4.8 4.9 5.4 6.1 6.8 7.5 8.4 8.8 8.9 9.2 9.3 9.9 10.6 10.9 11.2 11.7 12 12.7 13 13.5 16.3 23 4.7 6.1 6.4 6.4 7.4 7.6 8.7 9.4 10.6 12 10.8 11.6 11.4 12.4 13.7 14 14 14.7 14.5 15.6 15.9 16.4 19.9 27.8 2.6 3 3.2 3.3 3.4 4.6 5 5.6 6.1 5.5 9.7 6.9 7.2 7.4 7.6 7.9 8.4 8.6 9.4 9.8 10.1 10.6 12.7 18.2 Pliocene Pliocene Pliocene Pliocene Pliocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Miocene Sage, 2001; Beerling & Osborne, 2006). The effects of herbivores on grasslands are complex, leading to both positive and negative feedbacks with regards to the origin and expansion of C4 ecosystems, such as savannas (Beerling & Osborne, 2006). Nonetheless, the evolution of C4 grasses led to the formation of new habitats in which we would expect grazing mammals, such as ungulates, to evolve and diversify. Results here indicate that the first appearance of C4 photosynthesis occurred at the origin of Chloridoideae (about 30 mya) and the first ungulate adaptation to C4-dominated habitats approximately at the origin of Bovideae (about 26 mya) (Figs 1 and 4, Tables 1 and 2). Hassanin & Douzery (2003) found that the evolutionary radiation of ruminants (Bovideae and Cervideae) occurred between 29 and 32 mya, after the Eocene/Oligocene transition. These dates are in agreement with our molecular dating suggesting that Chloridoideae and Panicoideae (representing about 95% of C4 species) evolved in the cool, dry conditions of the Oligocene. Subsequent C4 origins occured at 29, 19 and 15 mya for Andropogoneae, Paniceae and Aristidoideae, respectively (Fig. 1, Table 1). These coincide with adaptations to C4-dominated habitats for Bovideae subfamilies (i.e. Antilopineae, Bovineae, Caprineae, Hippotragineae and Reduncineae), Equideae and Suideae (Fig. 4, Table 2). Regarding r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S 2409 Fig. 6 Bayesian consensus phylogenetic tree used as a topological constraint for inferring the chronogram in Fig. 1. It shows the shift in silica densities among 90 grass lineages reconstructed using the parsimony criterion. The legend shows the different color patterns used for the reconstruction of ancestral states. Index of silica density per unit of surface was calculated by counting the numbers of silica bodies in 10 random 100 mm  100 mm quadrats and multiplied this value by the average surface area of 10 silica bodies. ARI, Aristidoideae; ARU, Arundinoideae; CEN, Centothecoideae; CHL, Chloridoideae; DAN, Danthonioideae; EDL, early-diverging lineages; EH, Ehrhartoideae. BEP sensu the GPWG (2001) and PACCMAD sensu Sanchez-Ken et al. (2007). Bovideae, it is thought that an early diversification of the family in a moist Eurasian environment was followed by immigrations of diverse members (such as Antilopineae among others) into drier habitats in African between 20 and 15 mya in the middle Miocene (Matthee & Davis, 2001). In the subfamilies of Bovideae (such as Antilopineae, Hippotragineae and Reduncineae), Matthee & Davis (2001) argued that adaptation to open habitats in the middle to late Miocene was accompanied by rapid speciation. Our results suggest that particular C4 grasses might have experienced selection from grazing pressure for higher leaf epidermal silica densities with the greatest increases occurring in the late Miocene (from 6.3 to 13.8 for Aristidoideae, and from 6.3 to 12.5 for Chloridoideae; Fig. 6). The appearance of Aristidoideae is dated at about 15 mya in the middle Miocene (Fig. 1), which is concordant with the time of C4-habitat adaptation of Antilopineae (Fig. 4, Table 2). Within Chloridoideae, recent lineages seem to have evolved increased silica densities about 15–20 million years after the origin of the subfamily (Fig. 6). Recently, Prasad et al. (2005) suggested that grasses may have experienced high levels of herbivory even before 65 mya, well before the ungulate radiation and the first appearance of the C4 metabolic pathway. It is plausible that even though late-Cretaceous grasses did produce leaf epidermal phytoliths, the largest increase in phytolith production only occurred in the late Cenozoic for chloridoids and aristidoids. Neither Prasad et al. (2005), nor the present study, provide robust lines of evidence to favour one scenario or the other. Conclusion Using the largest body of phylogenetic data to date for grasses and ungulate mammals, we found evidence r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2410 Y. B O U C H E N A K - K H E L L A D I et al. that: (i) the most severe drop in atmospheric CO2 in the Oligocene coincided with the first origin of C4 photosynthesis in grasses (in Chloridoideae); (ii) multiple origins of C4 photosynthesis were complemented by the evolution of mammal groups adapted to C4 grass ecosystems, especially in the last 10 million years when C4 grasses diversified and became ecologically dominant; and (iii) that chloridoid and aristidoid grasses have evolved increased leaf epidermal silica densities from the middle Miocene to the Quaternary, possibly in response to increased selective pressure imposed by grazers. Because detailed species level phylogenetic trees of extant taxa can be informative of past evolutionary process, the use of phylogenetic methods to reveal new relationships between plant functional traits and climatic/nonclimatic factors should constitute the next step for understanding evolutionary processes leading to C4 ecosystems dominance in the late Tertiary. Acknowledgements We thank W. Bond, T. Barraclough, M. Bidartondo, B. Chase, E. Edwards, M. W. Chase and T. Galewski for comments, J. Henderiks and M. Pagani for sharing data on CO2 concentrations, S. Renvoize and D. Clayton for advice, L. Watson and M. Dallwitz for sharing data on grass morphology, E. Kapinos and L. Csiba for technical assistance, IITAC, the Irish HEA and the Trinity Centre for High Performance Computing (Trinity College Dublin) for access to high-performance computing clusters for phylogenetic analyses and the European Commission (HOTSPOTS), the Leverhulme Trust (SABBiG), the Darwin Initiative, and Enterprise Ireland (SC/2003/0437) for funding. We also thank the editor Rowan Sage for his very helpful comments on the manuscript. References Aliscioni SS, Giussani LM, Zuloaga FO, Kellogg EA (2003) A molecular phylogeny of Panicum (Poaceae: Paniceae): tests of monophyly and phylogenetic placement within the Panicoideae. American Journal of Botany, 90, 796–821. Barker NPE, Harley E, Linder HP (1995) Phylogeny of Poaceae based on rbcL sequences. Systematic Botany, 20, 423–435. Beerling DJ, Osborne CP (2006) The origin of the savanna biome. Global Change Biology, 12, 2023–2031. Belsky AJ (1986) Does herbivory benefit plants? A review of the evidence. The American Naturalist, 127, 870–892. Bininda-Emonds ORP, Cardillo M, Jones KE et al. (2007) The delayed rise of present-day mammals. Nature, 446, 507–512. Bond WJ, Midgley GF, Woodward FI (2003) The importance of low atmospheric CO2 and fire in promoting the spread of grasslands and savannas. Global Change Biology, 9, 973–982. Bortolussi N, Durand E, Blum MGB, Francois O (2006) ApTreeshape: a statistical analysis of phylogenetic treeshape. Bioinformatics, 22, 363–364. Bouchenak-Khelladi Y (2007) Grass evolution and diversification: a phylogenetic approach. PhD thesis, Trinity College Dublin. Bouchenak-Khelladi Y, Salamin N, Savolainen V, Forest F, van der Bank M, Chase MW, Hodkinson TR (2008) Large multigene phylogenetic trees of the grasses (Poaceae): progress towards complete tribal and generic level sampling. Molecular Phylogenetics and Evolution, 47, 488–505. Bremer K (2002) Gondwana evolution of the grass alliance of families (Poales). Evolution, 56, 1374–1387. Bro-Jorgensen J (2007) The intensity of sexual selection predicts weapon size in male bovids. Evolution, 61, 1316–1326. Cerling TE, Harris JM, Macfadden BJ, Leakey MG, Quade J, Eisenmann V, Ehleringer JR (1997) Global vegetation change through the Miocene/Pliocene boundary. Nature, 389, 153–158. Chapman GP (1996) From extinct to present-day grasses. In: The Biology of Grasses (ed. Chapman GP), pp. 112–123. CAB International, Oxon, UK. Chase MW, Hills DM (1991) Silica gel: an ideal material for field preservation of leaf samples for DNA studies. Taxon, 40, 215–220. Christin PA, Besnard G, Samaritani E, Duvall MR, Hodkinson TR, Savolainen V, Salamin N (2008) Oligocene CO2 decline promoted C4 photosynthesis in grasses. Current Biology, 18, 37–43. Clark LG, Zhang W, Wendel JF (1995) A phylogeny of the grass family (Poaceae) based on ndhF sequence data. Systematic Botany, 20, 436–460. Clayton WD, Renvoize SA (1986) Genera Graminum, Grasses of the World. Her majesty’s stationery office (ed.), London, UK. Coughenour MB (1985) Graminoid responses to grazing by large herbivores: adaptations, exaptations, and interacting processes. Annals of the Missouri Botanical Gardens, 72, 852–863. Crepet WL, Feldman GD (1991) The earliest remains of grasses in the fossil record. American Journal of Botany, 78, 1010–1014. Doyle JJ, Doyle JL (1987) A rapid isolation procedure for small quantities of fresh leaf tissue. Phytochemistry Bulletin, 19, 65–66. Drinnan AN, Crane PR (1994) A synopsis of medullosan pollen organs from the Middle Pennsylvanian Mazon Creek flora of northeastern Illinois, USA. Review of Paleobotany and Palynology, 80, 235–257. Dugas DP, Retallack GJ (1993) Middle Miocene fossil grasses from Fort Ternan, Kenya. Journal of Paleontology, 67, 113–128. Edwards EJ, Still CJ (2008) Climate, phylogeny and the ecological distribution of C4 grasses. Ecology Letters, 11, 266–276. Ehleringer JR, Monson RK (1993) Evolutionary and ecological aspects of photosynthetic pathway variation. Annual Review of Ecology and Systematics, 24, 411–439. Ellis RP (1990) Tannin-like substances in grass leaves. Memoirs of the Botanical Survey of South Africa, Pretoria, South Africa. GPWG (2001) Phylogeny and subfamilial classification of the grasses (Poaceae). Annals of the Missouri Botanical Gardens, 88, 373–457. Harding EF (1971) The probabilities of rooted tree-shapes generated by random bifunction. Advances in Applied Probability, 3, 44–77. Harris TE (1964) The Theory of Branching Process. Springer-Verlag, Berlin. Hasegawa M, Kishino H, Yano T (1985) Dating the human-ape splitting by a molecular clock of mitochondrial DNA. Journal of Molecular Evolution, 22, 160–174. r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S Hassanin A, Douzery EJP (1999) The tribal radiation of the family Bovidae (Artiodactyla) and the evolution of the mitochondrial cytochrome b gene. Molecular Phylogenetics and Evolution, 13, 227–243. Hassanin A, Douzery EJP (2003) Molecular and morphological phylogenies of Ruminantia and the alternative position of the Moschidae. Systematic Biology, 52, 206–228. Hatch MD (1971) The C4-pathway of photosynthesis-evidence for an intermediate pool of carbon dioxide and the identity of the donor C4-dicarboxylic acid. Biochemistry Journal, 125, 425–432. Hatch MD (1987) C4 photosynthesis: a unique blend of modified biochemistry, anatomy and ultrastructure. Biochemica et Biophysica Acta, 895, 81–106. Hattersley PW, Watson L (1975) Anatomical parameters for predicting photosynthetic pathways in grass leaves. Phytomorphology, 25, 325–333. Hattersley PW, Watson L (1992) Diversification of photosynthesis. In: Grass Evolution and Domestication (ed. Chapman GP), pp. 38–116. Cambridge University Press, Cambridge, UK. Henderiks J, Pagani M (2008) Refining ancient carbon dioxide estimates: significance of coccolithophore cell size for alkenone-based pCO2 records. Paleooceanography, 23. Herring JR (1985) Charcoal fluxes into sediments of the North Pacific Ocean: the Cenozoic record of burning. In: The Carbon Cycle and Atmospheric CO2: Natural Variations Archean to Present (eds Sundquist ET, Broecker WS), pp. 419–442. American Geophysical Union, Washington, DC. Hodkinson TR, Salamin N, Chase MW, Bouchenak-Khelladi Y, Renvoize SA, Savolainen V (2007a) Large trees, supertrees and diversification of the grass family. Aliso, 23, 248–258. Hodkinson TR, Savolainen V, Jacobs SWL, Bouchenak-Khelladi Y, Kinney MS, Salamin N (2007b) Supersizing: progress in documenting and understanding grass species richness. In: Reconstructing the Tree of Life: Taxonomy and Systematics of Species Rich Taxa (eds Hodkinson TR, Parnell JAN), pp. 279– 298. Taylor and Francis, CRC Press, Boca Raton, FL, USA. Huelsenbeck JP, Ronquist F (2001) MrBayes: Bayesian inference of phylogenetic tree. Bioinformatics, 17, 754–755. Janis CM (1993) Tertiary mammal evolution in the context of changing climates, vegetation, and tectonic events. Annual Review of Ecological Systematics, 24, 467–500. Janis CM, Damuth J, Theodor JM (2002) The origins and evolution of the North American grassland biome: the story from the hoofed mammals. Palaeogeography, Palaeoclimatology, Palaeoecology, 177, 183–198. Jernvall J, Fortelius M (2002) Common mammals drive the evolutionary increase of hypsodonty in the Neogene. Nature, 417, 538–540. Jernvall J, Hunter JP, Fortelius M (1996) Molar tooth diversity, disparity, and ecology in Cenozoic Ungulate radiations. Science, 274, 1489–1492. Jordan GJ, Weston PH, Carpenter RJ, Dillon RA, Brodribb TJ (2008) The evolutionary relations of sunken, covered, and encrypted stomata to dry habitats in Proteaceae. American Journal of Botany, 95, 521–530. Keeley JE, Rundel PW (2003) Fire and the Miocene expansion of C4 grasslands. Ecology Letters, 8, 683–690. 2411 Kellogg EA (2000) The grasses: a case study in macroevolution. Review of Ecology and Systematics, 31, 217–238. Kellogg EA (2001) Evolutionary history of the grasses. Plant Physiology, 125, 1198–1205. Kimura M (1981) Estimation of evolutionary distances between homologous nucleotide sequences. Proceedings of the National Academy of Science USA, 78, 454–458. Kürschner WM, Kvacek Z, Dilcher DL (2008) The impact of Miocene atmospheric carbon dioxide fluctuations on climate and the evolution of terrestrial ecosystems. Proceedings of the National Academy of Science USA, 105, 449–453. Lewis PO (2001) A likelihood approach to estimating phylogeny from discrete morphological character. Systematic Biology, 50, 913–925. Macfadden BJ, Cerling TE (1994) Fossil horses, carbon isotopes and global change. Trends in Ecology and Evolution, 9, 481–486. Maddison DR, Maddison WP (2005) MacClade 4. Version 4.08 for OSX. Sinauer Associates, Sunderland, MA. Maddison WP, Maddison DR (2006) Mesquite: a modular system for evolutionary analysis. Version 1.12. http://mesquiteproject.org Massey FP, Hartley SE (2006) Experimental demonstration of the antiherbivore effects of silica in grasses: impacts on foliage digestibility and vole growth rates. Proceedings of the Royal Society B–Biological Sciences, 273, 2299–2304. Matthee CA, Davis SK (2001) Molecular insights into the evolution of the family Bovidae: a nuclear DNA study. Molecular Biology and Evolution, 18, 1220–1230. Moore BR, Chan KMA, Donoghue MJ (2004) Detecting diversification rate variation in supertrees. In: Phylogenetic Supertrees: Combining Information to Reveal the Trees of Life (ed. BinindaEmonds ORP), pp. 487–533. Kluwer Academic, Dordrecht, the Netherlands. Myers P, Espinosa R, Parr CS, Jones T, Hammond GS, Dewey TA (2006) The animal diversity web. Accessed December 2007 at http://animaldiversity.org Nelder J, Wedderburn R (1972) Generalized linear models. Journal of the Royal Statistical Society Series A, 135, 370–384. Nylander JAA, Ronquist F, Huelsenbeck JP, Nieves-Aldrey JL (2004) Bayesian phylogenetic analysis of combined data. Systematic Biology, 53, 47–67. Osborne CP, Beerling DJ (2006) Nature’s green revolution: the remarkable evolutionary rise of C4 plants. Philosophical Transactions of the Royal Society B, 361, 173–194. Pagani M, Zachos J, Freeman KH, Bohaty S, Tipple B (2005) Marked change in atmospheric carbon dioxide concentrations during the Oligocene. Science, 309, 600–603. Pagel M (1999b) The maximum likelihood approach for reconstructing ancestral character sates of discrete characters on phylogenies. Systematic Biology, 48, 612–622. Paradis E, Claude J (2002) Comparative data and GEE. Journal of Theoretical Biology, 218, 175–185. Piperno DR (2006) Phytoliths: A Comprehensive Guide for Archeologists and Paleoecologists. AltaMira Press, Lanham, MD, USA. Posada D, Crandall K (1998) Modeltest: testing the model of DNA substitution. Bioinformatics, 14, 817–818. Prasad V, Stromberg CAE, Alimohammadian H, Shani A (2005) Dinosaur coprolites and the early evolution of grasses and grazers. Science, 310, 1177–1180. r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2412 Y. B O U C H E N A K - K H E L L A D I et al. Purvis A (1995) A composite estimate of primate phylogeny. Philosophical Transactions of the Royal Society of London B, 348, 405–421. Roalson EH (2007) C4 photosynthesis: convergence upon convergence upon . . . Current Biology, 17, 776–778. Roalson EH (2008) C4 photosynthesis: differentiating causation and coincidence. Current Biology, 18, 167–168. Roodt-Wilding R, Spies JJ (2006) Phylogenetic relationships in southern African chloridoid grasses (Poaceae) based on nuclear and chloroplast sequence data. Systematics and Biodiversity, 4, 401–415. Sage RF (2001) Environmental and evolutionary preconditions for the origin and diversification of the C4 photosynthetic syndrome. Plant Biology, 3, 202–213. Sage RF (2004) The evolution of C4 photosynthesis. New Phytologist, 161, 341–370. Sage RF, Monson RK (1999) C4 Plant Biology. Academic Press, San Diego, CA. Sanchez-Ken JG, Clark LG, Kellogg EA, Kay EE (2007) Reinstatement and emendation of subfamily Micrairoideae (Poaceae). Systematic Botany, 32, 71–80. Sinha NR, Kellogg EA (1996) Parallelism and diversity in multiple origins of C4 photosynthesis in the grass family. American Journal of Botany, 83, 1458–1470. Stebbins GL, Crampton B (1961) A suggested revision of the grass genera of temperate North America. In: Recent Advances in Botany (IXth International Botanical Congress, Montreal, Canada), Vol. 1. Toronto University Press, Toronto, Canada. Stromberg CAE (2005) Decoupled taxonomic radiation and ecological expansion of open-habitat grasses in the Cenozoic of North America. Proceedings of the National Academy of Science USA, 102, 11980–11984. Stromberg CAE (2006) Evolution of hypsodonty in equids: testing a hypothesis of adaptation. Paleobiology, 32, 236–258. Sungkaew S, Stapleton CMA, Salamin N, Hodkinson TR (2009) Non-monophyly of the woody bamboos (Bambuseae: Poaceae): a multi-gene region phylogenetic analysis of Bambusoideae s.s. Journal of Plant Research, 122, 95–108. Taub DR, Lerdau MT (2000) Relationship between leaf nitrogen and photosynthetic rate for three NAD-ME and three NADP-ME C4 grasses. American Journal of Botany, 87, 412–417. Thorne JL, Kishino H (2002) Divergence time and evolutionary rate estimation with multilocus data. Systematic Biology, 51, 689–702. Tipple BJ, Pagani M (2007) The early origins of terrestrial C4 photosynthesis. Annual Reviews of Earth and Planetary Sciences, 35, 435–461. Vicentini A, Barber JC, Aliscioni SS, Giussani LM, Kellogg EA (2008) The age of grasses and clusters of C4 photosynthesis. Global Change Biology, 14, 1–15. Watson L, Dallwitz MJ (1992) The Grass Genera of the World. CAB International, Wallingford, UK. Yang ZH (1994) Maximum-likelihood phylogenetic estimation from DNA sequences with variable rates over sites: approximate methods. Journal of Molecular Evolution, 39, 306–314. Zazzo A, Bocherens H, Brunet M et al. (2000) Herbivore paleodiet and paleo-environmental changes in Chad during the Pliocene using stable isotope ratios of tooth enamel carbonate. Paleobiology, 26, 294–309. Appendix A Table A1 Table of taxa, voucher information, accession numbers for GenBank (underlined are those taxa sequenced for this study) and silica densities Taxa Anomochlooideae Anomochloa marantoidea Streptochaeta spicata Streptochaeta sodiroana Aristidoideae Aristida congesta Aristida adscensionis Stipagrostis zeyheri Arundinoideae Arundo donax Molinia littoralis Molinia caerulea Phragmites australis Bambusoideae Bambuseae Arundinaria tecta Silica density Vouchers rbcL matK trnL-F AF164381 AF164383 na na na EF137515 5.290 1.452 Hodkinson574TCD AF021875 SSP419949 na ACU31359 na SZU31378 na AF164412 na na na na 7.563 Hodkinsons.n.TCD ADU13226 na AJ746295 U29900 AF164408 na AF164411 AF144575 EF137516 EF137517 na EF137518 Chase1992K AJ746179 EF125165 na Hodkinson131TCD Chase9257K 12.524 4.496 7.371 3.655 3.408 Continued r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S Table A1 2413 Continued. Silica density Taxa Vouchers rbcL matK trnL-F Arundinaria giganteus Bambusa multiplex Bambusa vulgaris Borinda emeryi Chimonobambusa marmorea Chusquea circinata Chusquea coronalis Dendrocalamus barbatus Drepanostachyum falcatum Himalayacalamus cupreus Himalayacalamus asper Indocalamus latifolius Melocanna baccifera Olmeca recta Otatea acuminata Pseudosasa japonica Pseudosasa amabilis Sinobambusa tootsik Thamnocalamus spathiflorus Thyrsostachys siamensis Yushania maculata Yushania maling Olyreae Olyra latifolia Pariana parvispica Pariana radiciflora Raddia brasiliensis Centothecoideae Centotheca lappacea Thysanolaena maxima Chase1995K Chase2000K Chase1963K Kew1992-0401 Chase1982K na BAMCPRBCL na EF125079 AJ746176 CCU13227 na AJ746173 AJ746265 EF125081 na AJ746177 EF125082 AJ746269 AJ746271 na AJ726273 EF125086 EF125087 EF125088 na AJ746277 na EF125166 na EF125167 EF125168 na AF164389 EF125169 EF125170 na EF125172 EF125173 EF125174 EF137435 EF137436 EF137438 na EF137442 EF137443 EF137444 EF137445 na EF137522 na EF137524 EF137525 EF137526 na EF137527 EF137528 EF137529 EF137534 na EF137535 EF137536 EF137539 EF137542 na EF137545 EF137553 EF137554 EF137555 na EF137556 EF125090 EF125091 na AJ746275 AF164386 na AF164387 EF137439 EF137540 EF137543 na EF137548 2.835 4.901 Hodkinson235TCD G.Sanchez-Ken22010K/ HodkinsonK10TCD EF125092 TMU31380 EF137446 EF137433 EF137557 EF137520 2.803 3.179 Chase19521K/Chase9262K EF125096 na EF125099 AJ784821 EF125103 na AY632375 na na EF125104 na EF125105 EF125107 EF125108 ECU31104 na na AM235073 na na na AF312330 AF312331 AF312353 na AF144591 na AF164426 na na AF312354 EF137447 AF144594 AF144580 na AF312341 na na AF144601 na EF137560 na EF137562 EF137568 EF137570 na na na EF137571 EF137559 na EF137561 AY576674 EF137563 na na EF137565 na EF137569 EF137564 10.862 Chloridoideae Chloris virgata Chloris truncata Cynodon transvaalensis Spartina pectinata Tragus racemosus Tragus berteronianus Zoysia sp. Zoysia macrostachya Zoysia japonica Calamovilfa longifolia Calamovilfa gigantea Crypsis schoenoides Dinebra retroflexa Eleusine indica Eragrostis capensis Eragrostis bicolor Eragrostis mexicana ssp. virescens Sporobolus sp. Sporobolus indicus Enneapogon polyphyllus Stapleton1126TCD Xia345 Chase1987K Stapletons.n.TCD Chase2004K Chase1994K Xia383IBSC Chase1986K Chase1989K Chase1985K Hodkinson62BTCD Chase1996K Chase1978K/Stapleton138BK Xia384 K Chase1980K Hodkinson24BTCD A.M. de Carvalho4394 CEPEC Hodkinson528TCD Chase2003K Chase9265K Chase9274TCD Chase19533K/Chase9277K HodkinsonK13TCD Hodkinson61TCD/Chase9261K Chase19523K/Chase9264K Cronk8176TCD Hodkinson126TCD Chase9270K Chase9275K Chase9267K 2.302 2.653 2.795 5.505 0.618 2.313 3.434 2.743 4.666 3.340 3.942 4.528 5.822 1.581 4.843 2.511 3.015 6.227 2.995 6.012 4.492 4.216 3.908 11.393 2.052 6.421 2.957 3.828 Continued r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2414 Y. B O U C H E N A K - K H E L L A D I et al. Table A1 Continued. Taxa Enneapogon scaber Enneapogon glaber Danthonioideae Danthonia spicata Danthonia sp. Merxmuellera macrowanii Ehrhartoideae Oryza sativa Panicoideae Andropogoneae Andropogon gerardii Arthraxon sp. Coix lacryma-jobi Cymbopogon citratus Imperata cylindrica Miscanthus sinensis Miscanthus giganteus Saccharum officinarum Sorghum halepense Sorghum bicolor Spodiopogon sibiricus Zea mays Zea diploperrenis Paniceae Cenchrus incertus Cenchrus setigerus Digitaria sanguinalis Echinochloa esculenta Echinochloa crus-galli Echinochloa utilis Panicum virgatum Panicum capillare Pennisetum glaucum Pennisetum macrourum Pharoideae Pharus latifolius Pooideae Aveneae Arrhenatherum elatius Avena sativa Avena fatua Hierochloe odorata Holcus lanatus Koeleria pyramidata Phalaris arundinacea Trisetum flavescens Trisetum sp. Brachypodieae Brachypodium sylvaticum Bromeae Bromus ramosus Bromus inermis Silica density rbcL matK trnL-F ESU31103 na na AF312360 na na Chase9254K G.Sanchez-Ken22021K/Chase9256K DSU31102 na MMU31438 AF164409 na EF137451 na EF137573 EF137575 8.700 Hodkinson46TCD RICCPRBCL AF148650 EF137577 4.994 Kew 1969-19004 Hodkinson111TCD Hodkinson124TCD Hodkinson42TCD Chase19300K/Kew 1996-2520 Hodkinson5TCD s.n. HodkinsonK104TCD Kew 1966-54209 AJ 784818 EF125114 EF125116 EF125117 AJ784826 EF125118 na EF125120 EF125122 na AJ784833 ZMA86563 na af144577 EF137454 EF137458 EF137459 EF137462 na EF137466 EF137470 na AF164418 EF137474 ZMA86563 na AY116263 EF137578 EF137580 EF137581 AY116262 AJ426571 na AY116253 AY116244 na EF137586 na AY116260 6.056 3.641 8.070 1.126 1.830 0.590 Hodkinson 117TCD na CEHRBCL AJ746264 EF125128 na na EF125135 na PENCARBOXL na na EF137457 AF164421 na na AF164422 na AF164423 na EF137467 EF137579 na AY116268 na AY116269 na AY116267 na na ay116266 Hodkinson514TCD AY357724 AF164388 EF137587 7.612 Hodkinson27TCD EF137486 AF164395 na EF137503 EF137504 EF137505 AF164396 na EF137513 EF137591 na EF137592 EF137605 EF137606 EF137607 EF137615 na EF137619 0.295 0.713 Hodkinson31TCD Chase19307K/Kew 387-51.38701 Hodkinson25TCD Hodkinson10TCD Hodkinson20TCD D.Gowing060K Hodkinson18TCD AJ784823 ASTCPRBCL na AJ784828 AJ746279 AJ784825 AJ784827 AJ746276 na Hodkinson22TCD AJ746258 AF164400 EF137593 3.159 Hodkinson41TCD na BICHRBPCX na AF164398 EF137595 na 2.747 Vouchers Kew 1997-5606 Kew 1986-5307 Chase9279 K Hodkinson47TCD Hodkinson110TCD Chase19524K Kew 1969-19104 8.567 1.120 3.906 5.282 1.157 3.985 1.459 3.097 0.547 8.244 0.000 1.994 4.287 2.021 1.890 Continued r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S Table A1 2415 Continued. Taxa Glycerieae Glyceria fluitans Glyceria maxima Meliceae Melica uniflora Melica altissima Poeae Briza media Briza erecta Castellia tuberculosa Catapodium rigidum Cynosurus cristatus Dactylis glomerata Deschampsia caespitosa Festuca rubra Lolium perenne Parapholis incurva Poa trivialis Poa pratensis Sesleria caeruela Vulpia ciliata Vulpia myuros Stipeae Stipa dregeana var. dregeana Stipa offneri Stipa sp. Triticeae Agropyron sp. Agropyron mongolicum Elymus glaucescens Elymus virginicus Eremopyrum bonaepartis Eremopyrum orientale Hordeum vulgare Hordeum murinum Taeniatherum caput-medusae Triticum aestivum Outgroups Restionaceae Elegia sp. Elegia squamosa Elegia cuspeidata Joinvilleaceae Joinvillea plicata Joinvillea ascendens Silica density Vouchers rbcL matK trnL-F Chase10776K Hodkinson19 TCD AJ746290 na EF137500 na na EF137603 4.362 Hodkinson44TCD aj746263 na na AF164399 EF137611 na 2.853 M.F.Fay143K AJ746285 na EF125149 EF125150 EF125151 AY395535 EF125152 AJ746261 AJ746293 EF125154 AJ746301 na EF125156 EF125157 na na AF164401 EF137492 EF137491 na EF137494 EF137495 EF137498 EF137506 EF137508 na AF164402 EF137511 na AF164403 EF137594 na EF137596 EF137597 EF137599 EF137600 EF137601 EF137602 EF137619 AF533036 EF137616 na EF137617 na AY118103 0.553 SDU31442 na na na AF164407 na na na EF137618 1.850 EF125160 na EGCHRBPCX na EF125162 na AY137456 na EF125164 AY328025 EF137478 na EF137496 na EF137497 na X64129 na EF137512 AB042240 na AF519117 na AF519144 na AF519151 na af519126 AF519164 AF148757 1.276 L12675 na na na AF881526 na na na AF148735 na L01471 na na AF164380 na na na Chase19520K/Kew1995-4283 Hodkinson23TCD Hodkinson14TCD Hodkinson26TCD D.Gowing063K Hodkinson15TCD D.Gowing069K Chase19528K D.Gowing066K Hodkinson45TCD Kew 1986-852 HodkinsonK79TCD Hodkinson62TCD Chase19301K Chase19525K Chase19532K Appendix B Equations derived from Moore et al. (2004) to test for shifts in diversification rates as implemented in APTREESHAPE (Bortolussi et al., 2006). If the ERM branching process is initiated with a single species and allowed to run for a period of time t with a branching rate l, 1.680 1.109 0.579 0.407 1.778 2.106 1.210 1.660 1.770 0.886 0.000 2.786 4.021 1.201 3.774 0.748 the probability of realizing n species is, according to Harris (1964): Pðnjl; tÞ ¼ elt ð1  elt Þn1 : ðB1Þ Accordingly, the probability of realizing n species partitioned between the left and the right descendants of a single node with ‘ r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 2416 Y. B O U C H E N A K - K H E L L A D I et al. and r species, respectively, under the heterogeneous two-rate parameter model HA is n1 X Pðijl; tÞPðn  ijls ; tÞ; ðB2Þ Pð‘; rjHA Þ ¼ Pð‘jl; tÞPðrjls ; tÞ= i¼1 where l is the ancestral rate and ls the shifted rate. For notational convenience, we assume that the shift occurred on the right sister branch after the speciation event, but the derivation of the symmetric formula poses no conceptual difficulties. Now, introducing the parameters q ¼ 1  expðltÞ; qs ¼ 1  expðls tÞ: ðB3Þ Equation (B1) can be rewritten as Pð‘; n  ‘jHA Þ ¼ ðq‘ qn‘ s Þ= n1 X qi qsn1 : ðB4Þ i¼1 Using Q 5 q/qs and assuming l6¼ls, we obtain that Pð‘; n  ‘jHA Þ ¼ Q‘ ð1  QÞ=ð1  Qn Þ: ðB5Þ Under the null-model with uniform branching probability (HO: l 5 ls), we recover the classical result of Harding (1971) Pð‘; n  ‘jHO Þ ¼ 1=ðn  1Þ: ðB6Þ likelihood-based method implemented in APTREESHAPE (with P  0.05), and we pinpointed them onto the complete generic level phylogenetic tree. Diversification shifts might be more common in more recent time periods because there are a greater number of lineages represented, and hence greater opportunity for shifts to occur. It relates to the issue of multiple testing within clades. Multiple testing issues are usually solved by using Bonferroni’s corrections. Without corrections, it is natural to expect that larger clades would produce more shifts. Bonferroni’s corrections are very difficult to implement for the kind of nested testing used in our study because the tests are not independent. Instead, they are strongly correlated; the issue is not peculiar to the kind of analysis used there, it is the same for SYMMETREE (Moore et al., 2004) or for other approaches to detecting diversification rate shifts. We performed a brief sensitivity analysis to evaluate the impact of not using Bonferroni’s corrections. Simulating the null model having equal rates in all lineages, we counted how many shifts were detected when tolerating 1%, 5% and 10% type I errors. For clades with n 5 100 taxa, we found that the percentages of false positive were actually lower than predicted from the type I errors. Increasing the sample size to n 5 200 taxa, we found that the false positive rate increased, but still remained very low at the 1% confidence level (see Table below). According to Eqns (B5) and (B6), the log-likelihood ratio LRHA : HO can be written as LRHA : HO ¼ log Pð‘; n  ‘jHA Þ  log Pð‘; n  ‘jHO Þ; False-positive rate Type I error N 5 100 N 5 200 0.01 0.05 0.10 0.0015 0.0124 0.0430 0.0024 0.0295 0.0953 which is equal to LRHA : HO ¼ ‘ log Q  logð1  Qn Þ þ logðn  1Þ þ C; ðB7Þ with C a constant independent on ‘ and n. The P-values of the likelihood ratio test can be computed by using Harding’s result (i.e. sampling ‘ from the uniform distribution, allowing us to implement the computation of P-values efficiently). Finally, diversification rate shifts were tested on the basis of the D1 statistic, which involved computing likelihood ratios at two nested levels (Moore et al., 2004). In order to attenuate the effect of absence of correction for multiple testing, we assessed shifts that led to a minimum of a 100-fold increase in the ancestral rate. The type I errors were fixed to a 5% level. Pvalues were computed using 10 000 Monte-Carlo replicates of D1 under the ERM model. We compiled the results of the It may be true that a greater number of lineages imply detecting more shifts, but the simulations actually indicate that the within clade multiple testing issue may impact the number of false positive only very slightly with the type I error used in our study (1% level). Appendix C r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417 C 4 G R A S S E S A N D S AVA N N A A D A P T E D U N G U L A T E S 2417 Fig. A3 Bayesian consensus tree of 90 taxa (used as a topological constraint for inferring the comprehensive generic-level phylogenetic tree shown in Fig. 1), where posterior probabilities for each clade are shown above the branches. EHR, Ehrhartoideae and CEN, Centothecoideae. r 2009 Blackwell Publishing Ltd, Global Change Biology, 15, 2397–2417