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