13
Cenozoic extinctions account for the low diversity of extant gymnosperms compared with angiosperms Michael D. Crisp 1 and Lyn G. Cook 2 1 Division of Evolution, Ecology and Genetics, Research School of Biology, The Australian National University, Canberra, ACT 0200, Australia; 2 School of Biological Sciences, The University of Queensland, Brisbane, Qld 4072, Australia Author for correspondence: Mike Crisp Tel: +612 6125 2882 Email: [email protected] Received: 6 June 2011 Accepted: 29 July 2011 New Phytologist (2011) 192: 997–1009 doi: 10.1111/j.1469-8137.2011.03862.x Key words: angiosperms, diversification, extinction, fossil record, gymnosperms, molecular dating. Summary We test the widely held notion that living gymnosperms are ‘ancient’ and ‘living fossils’ by comparing them with their sister group, the angiosperms. This percep- tion derives partly from the lack of gross morphological differences between some Mesozoic gymnosperm fossils and their living relatives (e.g. Ginkgo, cycads and dawn redwood), suggesting that the rate of evolution of gymnosperms has been slow. We estimated the ages and diversification rates of gymnosperm lineages using Bayesian relaxed molecular clock dating calibrated with 21 fossils, based on the phylogenetic analysis of alignments of matK chloroplast DNA (cpDNA) and 26S nuclear ribosomal DNA (nrDNA) sequences, and compared these with published estimates for angiosperms. Gymnosperm crown groups of Cenozoic age are significantly younger than their angiosperm counterparts (median age: 32 Ma vs 50 Ma) and have long unbran- ched stems, indicating major extinctions in the Cenozoic, in contrast with angiosperms. Surviving gymnosperm genera have diversified more slowly than angiosperms during the Neogene as a result of their higher extinction rate. Compared with angiosperms, living gymnosperm groups are not ancient. The fossil record also indicates that gymnosperms suffered major extinctions when climate changed in the Oligocene and Miocene. Extant gymnosperm groups occupy diverse habitats and some probably survived after making adaptive shifts. Introduction Extant gymnosperms (comprising conifers, cycads, gneto- phytes and Ginkgo) are commonly described as ‘ancient’ (Hill & Brodribb, 1999; Liao et al., 2004; McLoughlin & Vajda, 2005; Keppel et al., 2008; Xiao et al., 2010; A ´ lvare ´z-Yepiz et al., 2011), especially by comparison with the angiosperms, even though both groups are most proba- bly sisters (Finet et al., 2010; Magallon, 2010; Smith et al., 2010) and of the same evolutionary age. Several reasons underlie the perception that gymno- sperms are ancient. First, gymnosperms have an excellent fossil record (Norstog & Nicholls, 1997; Hill & Brodribb, 1999; Stockey et al., 2005; Doyle, 2006; Hermsen et al., 2006), with crown group fossils dating back to the first appearance of Cordaitales in the Palaeozoic c. 310 Ma (reviewed in Won & Renner, 2006). By contrast, there is not a single undisputed angiosperm fossil older than the Cretaceous (< 145 Ma; Doyle & Endress, 2010; Friis et al., 2010). Second, gymnosperms are considered to be highly conservative in their morphology, with a number of species being referred to as ‘living fossils’ because they are morpho- logically very similar to Mesozoic fossils, for example, Metasequoia glyptostroboides, Ginkgo biloba and Wollemia nobilis (Niklas, 1997), Ephedra (Rydin et al., 2004) and cycads in general (Norstog & Nicholls, 1997). By contrast, the extant clade of angiosperms has produced an astonish- ing diversity of life forms and physiology (Feild & Arens, 2005; Crepet & Niklas, 2009; Brodribb & Feild, 2010). Although extant gymnosperms are ecologically diverse (Hill & Brodribb, 1999), the group declined dramatically in dominance and abundance through the mid–late Cretaceous (Crane, 1987; Lupia et al., 1999; Cantrill & Poole, 2005), apparently contracting to marginal habitats under competition from the physiologically innovative angiosperms (Midgley & Bond, 1991; Feild & Arens, 2005; Berendse & Scheffer, 2009; Brodribb & Feild, 2010). Third, gymnosperms have much lower species diversity New Phytologist Research Ó 2011 The Authors New Phytologist Ó 2011 New Phytologist Trust New Phytologist (2011) 192: 997–1009 997 www.newphytologist.com

Cenozoic extinctions account for the low diversity of extant gymnosperms compared with angiosperms

Embed Size (px)

Citation preview

Cenozoic extinctions account for the low diversity ofextant gymnosperms compared with angiosperms

Michael D. Crisp1 and Lyn G. Cook2

1Division of Evolution, Ecology and Genetics, Research School of Biology, The Australian National University, Canberra, ACT 0200, Australia; 2School of

Biological Sciences, The University of Queensland, Brisbane, Qld 4072, Australia

Author for correspondence:Mike Crisp

Tel: +612 6125 2882

Email: [email protected]

Received: 6 June 2011

Accepted: 29 July 2011

New Phytologist (2011) 192: 997–1009doi: 10.1111/j.1469-8137.2011.03862.x

Key words: angiosperms, diversification,extinction, fossil record, gymnosperms,molecular dating.

Summary

• We test the widely held notion that living gymnosperms are ‘ancient’ and ‘living

fossils’ by comparing them with their sister group, the angiosperms. This percep-

tion derives partly from the lack of gross morphological differences between some

Mesozoic gymnosperm fossils and their living relatives (e.g. Ginkgo, cycads and

dawn redwood), suggesting that the rate of evolution of gymnosperms has been slow.

• We estimated the ages and diversification rates of gymnosperm lineages using

Bayesian relaxed molecular clock dating calibrated with 21 fossils, based on the

phylogenetic analysis of alignments of matK chloroplast DNA (cpDNA) and 26S

nuclear ribosomal DNA (nrDNA) sequences, and compared these with published

estimates for angiosperms.

• Gymnosperm crown groups of Cenozoic age are significantly younger than their

angiosperm counterparts (median age: 32 Ma vs 50 Ma) and have long unbran-

ched stems, indicating major extinctions in the Cenozoic, in contrast with

angiosperms. Surviving gymnosperm genera have diversified more slowly than

angiosperms during the Neogene as a result of their higher extinction rate.

• Compared with angiosperms, living gymnosperm groups are not ancient. The

fossil record also indicates that gymnosperms suffered major extinctions when

climate changed in the Oligocene and Miocene. Extant gymnosperm groups

occupy diverse habitats and some probably survived after making adaptive shifts.

Introduction

Extant gymnosperms (comprising conifers, cycads, gneto-phytes and Ginkgo) are commonly described as ‘ancient’(Hill & Brodribb, 1999; Liao et al., 2004; McLoughlin &Vajda, 2005; Keppel et al., 2008; Xiao et al., 2010;Alvarez-Yepiz et al., 2011), especially by comparison withthe angiosperms, even though both groups are most proba-bly sisters (Finet et al., 2010; Magallon, 2010; Smith et al.,2010) and of the same evolutionary age.

Several reasons underlie the perception that gymno-sperms are ancient. First, gymnosperms have an excellentfossil record (Norstog & Nicholls, 1997; Hill & Brodribb,1999; Stockey et al., 2005; Doyle, 2006; Hermsen et al.,2006), with crown group fossils dating back to the firstappearance of Cordaitales in the Palaeozoic c. 310 Ma(reviewed in Won & Renner, 2006). By contrast, there isnot a single undisputed angiosperm fossil older than theCretaceous (< 145 Ma; Doyle & Endress, 2010; Friis et al.,

2010). Second, gymnosperms are considered to be highlyconservative in their morphology, with a number of speciesbeing referred to as ‘living fossils’ because they are morpho-logically very similar to Mesozoic fossils, for example,Metasequoia glyptostroboides, Ginkgo biloba and Wollemianobilis (Niklas, 1997), Ephedra (Rydin et al., 2004) andcycads in general (Norstog & Nicholls, 1997). By contrast,the extant clade of angiosperms has produced an astonish-ing diversity of life forms and physiology (Feild & Arens,2005; Crepet & Niklas, 2009; Brodribb & Feild, 2010).Although extant gymnosperms are ecologically diverse (Hill& Brodribb, 1999), the group declined dramatically indominance and abundance through the mid–lateCretaceous (Crane, 1987; Lupia et al., 1999; Cantrill &Poole, 2005), apparently contracting to marginal habitatsunder competition from the physiologically innovativeangiosperms (Midgley & Bond, 1991; Feild & Arens,2005; Berendse & Scheffer, 2009; Brodribb & Feild, 2010).Third, gymnosperms have much lower species diversity

NewPhytologist Research

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009 997www.newphytologist.com

than angiosperms (c. 900–1200 vs 250 000–270 000 ormore, respectively; Hill, 1998b; Magallon & Castillo,2009; Bell et al., 2010; Christenhusz et al., 2011; Earle,2011), exemplified by many high-level taxa with one(Ginkgoideae, Sciadopityaceae, Metasequoia, Wollemia) or afew species, and no gymnosperm genus has more than 120species. However, the perception that gymnosperms are‘ancient’ and ‘relictual’ because they have few species isbased on a fallacy that the species-poor sister group is themore ancient or primitive (Crisp & Cook, 2005; Baum &Offner, 2008; Omland et al., 2008).

Today, there is a huge difference in species diversitybetween gymnosperms and angiosperms that has developedover the approximately 350 Myr since their lineagesdiverged. This difference is more impressive when consider-ing that the gymnosperm crown (c. 315 Ma) is much olderthan the angiosperm crown (c. 220 Ma) (Smith et al.,2010). Although the Cretaceous rise of angiosperms todominance in the fossil record was abrupt (Crane, 1987),their early diversification was not necessarily rapid.Molecular dating indicates a much older crown age thanshown by fossils alone (Magallon & Sanderson, 2005;Magallon, 2010; Smith et al., 2010), and the oldest knownangiosperm fossils were already diverse, indicating an earlierperiod of evolution ‘in obscurity’ (Friis et al., 2010). It isnot surprising that angiosperm diversification rates havevaried through time and among lineages (Magallon &Sanderson, 2001; Magallon & Castillo, 2009), and similarvariation might be expected in gymnosperms, although noestimate has been made to date. Thus, it is not immediatelyclear whether the relatively low extant diversity of gymno-sperms is the result of their slow speciation (e.g. Gorelick &Olson, 2011) or high extinction (e.g. Hill & Brodribb,1999; Cantrill & Poole, 2005) or changes in both lineagesthrough time. The overall species diversity of gymnospermfossil taxa, especially conifers, does not appear to havedeclined dramatically through the Cretaceous (Lidgard &Crane, 1988; Niklas & Tiffney, 1994; Lupia et al., 1999;Crepet & Niklas, 2009), despite probable competition fromthe angiosperms. Furthermore, evolutionary rates, as mea-sured by nucleotide substitutions, are not necessarily slowerin gymnosperms than in angiosperms. Both synonymousand nonsynonymous substitution rates are faster, on aver-age, in gymnosperm organelles (mitochondrial andchloroplast DNA, mtDNA and cpDNA), but slower intheir nuclear genes (Willyard et al., 2007; Drouin et al.,2008). Faster rates might be expected in angiospermsbecause they include large groups of small plants with shortlife cycles, such as grasses, in contrast with gymnosperms.These life history traits reflect a short generation time, andboth theory (Bromham & Penny, 2003) and empiricalstudies on plants (Verdu, 2002; Smith & Donoghue, 2008)indicate a positive relationship between short generationtime and high nucleotide substitution rates. However, sub-

stitution rates in the enigmatic gnetophytes are so highacross all three genomes that they increase overall rate esti-mates for gymnosperms by 12–64% (Drouin et al., 2008)and confound the placement of the group in molecular phy-logenies (Qiu et al., 2007; Doyle, 2008). Thus, lowdiversity and morphological conservatism of gymnospermsdo not necessarily equate with a history of slow molecularevolution.

Recent molecular dating studies have indicated that someputatively ancient lineages in both angiosperms and gymno-sperms have surprisingly young (post-Eocene) extantradiations (crown groups): for example, genera ofAraucariaceae (Biffin et al., 2010a) and cycads (Treutlein &Wink, 2002), Gunnera (Bell et al., 2010), Livistona (Crispet al., 2010), geographical clades within Monimiaceae(Renner et al., 2010), Nothofagus (Cook & Crisp, 2005),Phyllocladus (Wagstaff, 2004) and Pinus (Willyard et al.,2007). Such paradoxes can arise when the stem age of anextant lineage (when it diverged from its closest living rela-tive) is much older than the crown age (when the extantclade began to diversify). It appears that many of theancient fossils attributed to extant gymnosperm taxa proba-bly belong to their stem groups, that is, are from extinctsisters of the crown group (Doyle & Donoghue, 1993;Biffin et al., 2010a; Doyle & Endress, 2010). To calibratemolecular dating, it is critically important to determinewhether a fossil belongs to the crown of the taxon to whichit has been nominally assigned, or is from outside thecrown. A stem group fossil incorrectly placed in the crowncould make the crown appear too old (or too young if viceversa). Erroneous assignment of a fossil to a crown com-monly results from the use of the criterion of overallsimilarity, which probably reflects shared ancestral characterstates rather than shared derived states. In this study, weplace a fossil at a crown node only if it shares a synapomor-phy with a clade that originated within the crown group(Crepet et al., 2004; Renner, 2005; Ho & Phillips, 2009)or has been unambiguously assigned to a subclade withinthe crown (Magallon & Sanderson, 2001; Ho & Phillips,2009).

Several previous molecular phylogenetic studies ingymnosperms have addressed the age and diversificationrates of individual gymnosperm clades (some examples arecited above), but comparative analyses across gymnospermsas a whole are lacking. In broad molecular dating studiesacross seed plants as a whole, the sampling of gymnospermshas been limited, especially for internal crown groups,because the focus has been on angiosperms (Magallon,2010; Smith et al., 2010). This is unfortunate because toomuch sampling emphasis on species-rich clades is likely tobias diversity estimates as a result of undersampling ofspecies-poor clades (Ricklefs, 2007). Here, we present anew phylogenetic dating analysis of gymnosperms, sam-pling terminal taxa more densely than in previous studies.

998 Research

NewPhytologist

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

We analyse nuclear ribosomal and chloroplast genes usinguncorrelated and correlated relaxed molecular clocks in aBayesian framework with a large number of internal fossil-based calibration points. Our primary question is: are livinggymnosperm crown groups as ancient as previouslythought, particularly in comparison with angiosperms? Wealso ask whether gymnosperm lineages have diversifiedmore slowly than angiosperm lineages of comparable age.

Materials and Methods

Sampling and alignment of DNA

A sample of 86 seed plant taxa was drawn from sequencesavailable in GenBank for the large subunit (26S) of nuclearribosomal DNA (nrDNA) and the chloroplast gene matK(Supporting Information Table S1). The aim was to repre-sent the main lineages of gymnosperms, down to andincluding most of the 84 genera (Christenhusz et al., 2011),and nine angiosperms were included as an outgroup.Gnetophytes were omitted from the phylogenetic analysisbecause of the well-known long-branch problem in thisgroup, making the estimation of their phylogenetic positionuncertain (Won & Renner, 2006; Doyle, 2008; Mathews,2009) and possibly biasing the relaxed molecular clock dat-ing analyses (Ickert-Bond et al., 2009). Instead, we addeddivergence time estimates from a separate analysis of gneto-phytes (Ickert-Bond et al., 2009) to our dataset to estimatecrown ages and diversification rates of gymnosperms(below). The DNA loci were chosen because they aresufficiently conserved to be alignable across distantly relatedtaxa, but sufficiently variable to resolve low-level relation-ships, and because of their availability for a broad sample ofseed plants. In five cases in which sequences of both lociwere not available for the same species, a composite termi-nal was assembled from closely related species. Where thiswas not feasible (e.g. because of the uncertain monophylyof the composite taxa), only a single locus was used, leavingseven gaps in the 26S partition and 10 in the matKpartition.

Sequences were aligned initially using MUSCLE (Edgar,2004), as implemented in the CIPRES Science Gateway(Miller et al., 2010), and the alignment was adjusted byhand using Se–Al (Rambaut, 1996). Only the open readingframe of matK was used and regions of uncertain alignment(when inspected visually) in 26S were excluded.

Phylogenetics and relaxed molecular clock dating

Initially, separate maximum-likelihood analyses of the 26Sand matK datasets were conducted using RAxML(Stamatakis et al., 2008), as implemented in CIPRES, withthe default GTR + G model and the fast bootstrap option.The results indicated no conflicts between supported clades

and the two DNA loci were combined for subsequentanalyses. Chronograms (phylogenies with branch lengthsscaled proportionally to time) were estimated using twoBayesian methods: BEAST ver. 1.6.1 (Drummond et al.,2006), with an uncorrelated log-normal molecular clock,and MrBayes ver. 3.1.2 (Ronquist & Huelsenbeck, 2003),with an autocorrelated clock using penalized likelihood inr8s ver. 1.7.1 (Sanderson, 2003). BEAST has the advantageof estimating the chronogram directly from the sequencesand clock parameters, and modelling a separate clock for eachpartition, whereas the second approach requires two stages:first, sampling of a posterior tree profile using MrBayes, andthen using the profile with r8s to estimate lineage divergencetimes. The latter does not allow a partitioned clock.

In both Bayesian analyses, a separate nucleotide substitu-tion model was used for each locus. For matK, this was aGTR + G model partitioned separately for the third andfirst + second codon positions (known as ‘SRD06’ inBEAST). For 26S, a single GTR + G model was used forall positions. A maximum-likelihood tree estimated usingRAxML, with branches smoothed using penalized likeli-hood in r8s, was employed as a starting tree for BEAST.Multiple Bayesian Monte Carlo Markov Chains were runin BEAST and MrBayes, and each was monitored usingTracer (Rambaut & Drummond, 2007) until judged tohave converged on the same plateau (after ‘burnin’) usingseveral criteria recommended by the authors of theprograms: all effective sample sizes were > 200 (both pro-grams); all average split frequency differences were < 0.01(MrBayes); a Bayes factor test (MrBayes; Kass & Raftery,1995) was used. In BEAST, which runs only a single MonteCarlo Markov Chain at a time, convergence usuallyrequired two runs of 50 M generations. In MrBayes, twoparallel runs, each with eight chains, usually ensured con-vergence within 5 M generations. Substitution rateheterogeneity was assessed in BEAST from the posteriordistributions of the parameters ‘coefficient of variation’ and‘ucld.stdev’, as recommended by Drummond et al. (2007).

In r8s, the optimal smoothing parameter value wasestimated using fossil-based cross-validation. For 10trees sampled at maximum separation across the MrBayesposterior set, cross-validation indicated maximum rate het-erogeneity among lineages in every case. Ages and theirconfidence intervals were estimated for selected nodes usingthe r8s profile command with the posterior tree set.

Calibration

Table S2 lists details of calibrations for the root height and21 internal nodes, numbered to correspond with the nodelabelling in Fig. S1. The mean root height constraint of355 Ma was based on Won & Renner (2006), whoassigned minimum and maximum ages of 325 and385 Ma, respectively, based on fossils with fused (shared by

NewPhytologist Research 999

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

all extant seed plants) and unfused integuments, respec-tively. This calibration is consistent with the review bySanderson et al. (2004) and the fossil-constrained estimatesby Magallon & Sanderson (2005) and Smith et al. (2010),but is c. 20 Ma younger than the estimate by Magallon(2010). In BEAST, multiple analyses were performed to testthe effect of using different prior distributions for internalcalibration, that is, none, log-normal and uniform. Log-normal priors were preferred for fossil-based calibrationsand normal priors for secondary calibrations, as recom-mended by Ho & Phillips (2009). Uniform calibrationswith hard minima (and a maximum set at the root height)were used for comparison with r8s, which does not imple-ment calibrations with soft bounds. Calibration limited tothe root height was used to compare with internal calibrationsand assess their influence on divergence time estimates.

The Cenomanian (83.5–99.6 Ma) fossils described asWiddringtonia americana (tribe Cupresseae; McIver, 2001)combine features of both cupressoid tribes (Cupresseae andCallitridae), including leaves that are difficult to distinguishfrom Tetraclinis (Callitridae; Stockey et al., 2005). As thefossil is relatively old, its potential influence on dating islarge, and so alternative placement at the stem of the genusand the crown of the subfamily was tested by cross-valida-tion in r8s (Near et al., 2005).

Lineage ages

To address the primary question (whether living gymno-sperm crown groups can be considered ‘ancient’), wecompared the ‘stemminess’ of clades in gymnosperms andangiosperms. Stemminess is the branch length (age) ofcrowns relative to their stems (reviewed by Smith, 1994): thelonger the stem relative to the ‘twigs’ (crown), the greater thestemminess of the clade. As clades are nested all the way fromthe tips to the root of the tree of life, such comparisons needto be made among independent (non-nested) lineages withina defined time slice. We selected lineages that crossed theCretaceous–Palaeogene boundary (K–Pg, 65.5 Ma), andwhose crowns diverged during the Cenozoic (e.g. Fig. 1).We chose this boundary because most gymnosperm generaappear to have crowns of Cenozoic age and diverged fromtheir sisters in the Mesozoic (Fig. 1). This choice is some-what arbitrary; however, it was the last major geologicalboundary and it excludes earlier and potentially confound-ing events, such as the rapid mid-Cretaceous radiation ofangiosperms, and also the end-Cretaceous extinction event,whose impact on angiosperms and gymnosperms is unclear,although apparently it was not as severe as the Cenozoicextinction events (Macphail et al., 1994; Niklas, 1997;Wing, 2004; Crepet & Niklas, 2009). Therefore, wesampled the stem age of a lineage at its last divergence thatwas estimated to be > 65.5 Ma, and the crown age at its firstdivergence that was estimated to be < 65.5 Ma. Mean crown

and stem age estimates were compared between 21 gymno-sperm and 110 angiosperm lineages using statistics implementedin Prism 5.0d for Mac OSX (�1994–2011, GraphPadSoftware Inc., San Diego, California, USA). Lineages with asingle terminal species were excluded: either they had a singleextant species or their crown node had not been sampled.The gymnosperm sample comprised all clades listed inTable S3, plus Gnetum and Zamia. Angiosperm divergencetimes were sampled from Bell et al. (2010) for comparabilitywith ours: both studies used BEAST analyses with log-nor-mal calibration priors. As the distributions of most lineageages failed normality tests (D’Agostino–Pearson test), thenonparametric Mann–Whitney U-test was used to assesswhether the stem and crown ages differed between gymno-sperms and angiosperms.

Diversification

It is already clear from the large disparity in species’ richnessthat gymnosperms have diversified more slowly than angio-sperms since their lineages diverged in the Carboniferous.This could be the net outcome of either a higher extinctionrate or a lower speciation rate (or both) in gymnosperms.The fossil record indicates large mid–late Cenozoic extinc-tion spikes in gymnosperms following low extinction ratesin both gymnosperms and angiosperms throughout theCretaceous, Palaeocene and Eocene (Niklas, 1997; Crepet& Niklas, 2009). The first and largest (‘exceptionally high’)extinction of gymnosperms in the Cenozoic occurred c.29 Ma, at a rate approximately seven times higher than inangiosperms, followed by similarly differential extinctions,but with lower peaks: c. 16 Ma and 7–5 Ma (Niklas, 1997,Fig. 8.14 and Table 8.3; Crepet & Niklas, 2009, Fig. 2).To attempt to tease speciation and extinction rates apart,we compared the diversification of gymnosperm and angio-sperm crown groups during the Cenozoic period of highgymnosperm extinction, that is, from 30 Ma to the present.Speciation rates (k) were estimated at different relativeextinction rates (e) using the methods of Magallon &Sanderson (2001), as implemented in LASER (Rabosky,2006) hosted on the R project site (http://www.r-project.org/). The sample comprised 18 non-nested crown groupsof gymnosperms and 28 of angiosperms (Table S3).Primary sources for crown ages were the present study(Fig. 1) for gymnosperms and Bell et al. (2010) for angio-sperms and, for species’ counts, were Earle (2011) forconifers, Hill (2004a) for cycads and Stevens (2001onwards) for angiosperms. These sources were supple-mented by more detailed studies for particular clades(Table S3). As published analyses of angiosperm diversifica-tion tend to emphasize examples of rapid diversification(e.g. Richardson et al., 2001; Biffin et al., 2010b; Valenteet al., 2010), the angiosperm sample was balanced byincluding examples of these, as well as low-diversity clades

1000 Research

NewPhytologist

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

from Nymphaeales, Chloranthales and Magnoliidae(Table S3). Overall, the sample comprised plant lineageswith a range of longevities and life forms. For each clade,we ensured that the age was estimated for the basal node ofthe crown, not an internal node, by checking publishedphylogenies of the individual clades (cited in Table S3).The method requires input of the extinction rate to specia-

tion rate ratio (e). As e is very difficult to estimate reliably(e.g. Rabosky, 2010), we followed the recommendation(Magallon & Sanderson, 2001; Magallon & Castillo, 2009)of using low (0.1) and high (0.9) values and comparingthe results. Estimated speciation rates (k) at the differentvalues of e were compared between gymnosperms andangiosperms using the nonparametric Mann–Whitney

KPa (65.5 Ma)

Time before present (Ma)

Angiosperms

Gymnosperms

Cycads

Conifers

Pinaceae

Podocarpaceae

Araucariaceae

Zamiaceae

Cycadaceae

Cupressaceae

*

*

*

** **

* *

*

*

***

*

* * *

**

*

*

*

*

*

*

***

* *

*

*

*

** * **

**

*

*

**

*

***

*

**

*

*

*

**

**

**

*

* *

*

Amborella trichopoda

Nuphar spp.

Schisandra spp.Juglans nigraPlatanus occidentalis

Magnolia denudataPeumus boldus

Acorus gramineusChloranthus spp.

Bowenia serrulataBowenia spectabilis

Encephalartos spp.

Encephalartos arenariusEncephalartos lebomboensis

Lepidozamia hopei

Macrozamia moorei

Macrozamia communisMacrozamia dyeri

Macrozamia lucida

Ceratozamia mexicana

Ceratozamia miquelianaCeratozamia norstogii

Chigua restrepoiZamia integrifoliaZamia pumila

Zamia lindeniiZamia skinneri

Zamia furfuraceaMicrocycas calocoma

Stangeria eriopus

Dioon eduleDioon spp.

Cycas revolutaCycas revolutaCycas rumphii

Ginkgo bilobaGinkgo biloba

Abies fabriCedrus atlanticaCedrus deodaraPicea asperataPicea breweriana

Pinus massonianaPinus nigra

Pseudotsuga menziesii

Agathis australisAgathis dammara

Wollemia nobilis

Araucaria araucanaAraucaria angustifoliaAraucaria heterophylla

Acmopyle pancheriDacrycarpus imbricatusPodocarpus macrophyllus

Lagarostrobus frankliniiSaxegothaea conspicua

Sciadopitys verticillata

Athrotaxis cupressoidesAthrotaxis laxifolia

Cryptomeria japonicaGlyptostrobus pensilisTaxodium distichum

Calocedrus decurrensPlatycladus orientalisCupressus funebrisJuniperus communisChamaecyparis lawsonianaChamaecyparis pisifera

Thuja plicataThujopsis dolabrata

Actinostrobus spp. Callitris spp.

Libocedrus bidwilliiLibocedrus yateensis

Pilgerodendrum uviferumAustrocedrus chilensisPapuacedrus papuana

Metasequoia glyptostroboidesSequoia sempervirensTaiwania cryptomeroidesCunninghamia lanceolata

Cephalotaxus harringtonia

Taxus baccataTorreya grandis

Fig. 1 Maximum credibility molecular chronogram (time tree) of gymnosperms from BEAST, with branch lengths proportional to time andlog-normal fossil-based calibrations. Asterisks on the branches indicate clade credibility ‡ 0.95. Bars at nodes indicate the 95% credibilityinterval of age estimates. The dashed grey vertical line indicates the Cretaceous–Palaeogene boundary (K–Pg), used to select lineages forcomparison (see text for full details).

NewPhytologist Research 1001

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

U-test. In particular, we tested the expectation that, ifhigher extinction rates alone account for the lower diversifi-cation rate of gymnosperms, speciation rates should beabout the same as in angiosperms. Therefore, we also com-pared k in angiosperms and gymnosperms with e = 0.1 and0.7, respectively, to reflect the seven times differential inCenozoic extinction rates inferred from the fossil record(Niklas, 1997).

Results

Phylogeny and molecular dating

The tree topologies from both BEAST (Fig. 1) andMrBayes (Fig. S1) were well resolved with strong support atmost nodes, and were identical except among closely relatedspecies within some genera, e.g. within Libocedrus,Macrozamia and Zamia. This indicates that the branchlength smoothing and calibrations applied in BEAST didnot perturb the topology. In addition, the root position wasthe same, even though no outgroup was used in BEAST,which produces an ultrametric rooted tree as a result of thebranch smoothing process (Drummond et al., 2006).Moreover, the topology was congruent with others derivedfrom more loci, but with fewer gymnosperm terminals (e.g.Won & Renner, 2006; Qiu et al., 2007; Rai et al., 2008;Zgurski et al., 2008; Magallon, 2010). One recent multi-nuclear gene analysis (Finet et al., 2010) placed Ginkgo sis-ter to cycads (rather than to conifers), but their samplingincluded only 11 gymnosperms. Using an alternative approachthat attempts to tease apart paralogous nuclear gene copies,Mathews et al. (2010) found radically different topologiesfrom all the above for both angiosperms and gymnosperms.

The placement of the W. americana calibration point attwo alternative nodes was tested using cross-validation inr8s, namely at the stem of Widdringtonia (node 22 inFig. S1) or the crown of Cupressoideae (node 18). Thedivergence time estimates from the shallower placement(node 22) were older throughout conifers, especially in theSciadopitys–Cupressaceae lineage (node 13 and its descen-dants), where they were 10–66 Myr older. Therefore,estimates for eight of the other nine calibrated nodes (exceptCalocedrus–Platycladus) in this clade were substantially olderthan their calibrated ages, thus failing the cross-validationtest (Near & Sanderson, 2004). Given this, and the observa-tion that W. americana shares features of both lineageswithin the crown of Cupressoideae, the calibration pointwas placed at the deeper node (18) in all other analyses.

Inferred divergence times and nucleotide substitutionrates varied with the branch smoothing method and calibra-tion model (Table 1). Using the results from the log-normal-calibrated model in BEAST (Table 1, first row) asthe baseline for comparison, r8s dates were generally olderand substitution rates were correspondingly slower. This

was probably partly a result of the use in r8s of hard min-ima, which consistently give older dates than do softminima when also used in BEAST (compare first three rowsin Table 1). However, even with no internal constraints, r8sconsistently gave older dates (except at the root, which wasconstrained) than did BEAST for the same nodes (for exam-ple, compare the fourth and last rows in Table 1).Exponential calibrations in BEAST gave slightly youngerdates than did log-normal calibrations, as expected, becausethe mode (highest probability) is younger in an exponentialdistribution with the same offset (minimum age). InBEAST, constraining only the root resulted in much youn-ger dates than under any internal calibration model,showing that calibration influences the branch smoothingmodel. The influence of internal calibrations can also beseen in the smaller error bars in the Cupressaceae clade(Fig. 1), which has nine calibration points (Fig. S1). Bycontrast, dates and rates were almost the same in r8swhether calibrated internally or only at the root (comparethe last two rows in Table 1), indicating that internal cali-brations exerted little influence in r8s. Given the crucial roleplayed by calibration in divergence time estimation (Ho &Phillips, 2009) and the superior capacity of BEAST to takeuncertainty into account in its divergence time modelling,we preferred the BEAST log-normal-calibrated divergencetimes for subsequent analyses.

Lineage ages

From the 21 gymnosperm and 110 angiosperm lineagesthat fitted the criterion of crossing the K–Pg boundary, wesampled mean estimates for the crown age, stem age andtheir difference (stem length) from the BEAST log-normal-calibrated posterior (cf. Fig. 1). The median ⁄ mean crownage estimate for gymnosperms (32 ⁄ 35.2 Ma) was youngerby 18 ⁄ 13.6 Myr than that for angiosperms (50 ⁄ 48.8 Ma).Conversely, gymnosperm stem nodes were significantlyolder, by 16.5 ⁄ 33.3 Myr. Stem lengths, the differencebetween stem and crown node ages, were much longer ingymnosperms than in angiosperms, by 39.5 ⁄ 46.3 Myr. Allthese differences were highly significant (Fig. 2, Table 2).To summarize, gymnosperm clades have significantly longerunbranched stems (stemmier) and shorter crowns (twiggier)than angiosperm clades.

Diversification

With respect to the post-Eocene (£ 30 Ma) sample of crowngroups, when the relative extinction rate (e) was assumed tobe the same in gymnosperms and angiosperms, the estimatedspeciation rate was significantly lower in gymnosperms, irre-spective of whether e was low (0.1) or high (0.9) (Table 3,Fig. 3). If we had included in our sample all of the extremelyrapid angiosperm radiations cited in the Materials and

1002 Research

NewPhytologist

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

Methods section, the difference in the estimated speciationrates between angiosperms and gymnosperms would havebeen even greater; thus, our results can be considered conser-vative in this respect. However, approximately equalspeciation rates were obtained in gymnosperms and angio-sperms when it was assumed realistically, based on the fossilrecord, that the relative extinction rate was seven timeshigher in gymnosperms (no significant difference,P = 0.645; compare first and last columns in Table 3 andFig. 3). Assuming an even larger extinction differential(e = 0.1 in angiosperms and e = 0.9 in gymnosperms)resulted in a higher speciation rate in gymnosperms than inangiosperms, although the difference was marginally signifi-cant (P = 0.042).

Nucleotide substitution rates

The use of relaxed molecular clock models (as opposed to astrict clock) was supported by evidence that rates varyamong lineages, similar to previous findings for seed plantsat this level (e.g. Magallon & Sanderson, 2005). Cross-vali-dation analysis in r8s indicated a maximal level of rateheterogeneity among lineages. In BEAST, a separate relaxedclock model was applied to each locus, allowing the assess-ment of the rate variation across the tree. The parameter‘coefficient of rate variation’ measures the standard deviationof rates in relation to the mean, with a value of zero indicat-ing a strict molecular clock. For both matK and 26S, the95% credibility intervals of both parameter distributionsexcluded zero, indicating non-clock-like substitution ratesand supporting the use of relaxed molecular clocks.

The covariance of rates measures the degree of rate corre-lation between neighbouring branches in the tree and, forthis parameter, the 95% credibility interval excluded zeroT

able

1Es

tim

ated

div

ergen

cetim

esan

dnucl

eotide

subst

itution

rate

sfo

rse

lect

ednodes

under

various

met

hods

and

calib

ration

model

s

Met

hod

Cal

ibra

tion

model

Div

ergen

cetim

es(M

aw

ith

95%

CI)

Subst

itution

rate

on

bra

nch

(es)

subte

nded

by

node

(subst

itutions

site

)1

Myr

)1

·10

4

with

95%

CI)

Cyca

sst

emC

ycas

crow

nM

acro

zam

ia

stem

Macr

oza

mia

crow

nC

alli

tris

stem

Cal

litr

is

crow

nC

ycad

ales

stem

Cyca

sst

emC

ycas

crow

nM

acr

oza

mia

+B

ow

enia

Macr

oza

mia

stem

Macr

oza

mia

crow

n

BEA

STLo

g-n

orm

al262

(251–2

82)

18.9

(5–4

2)

103

(70–1

43)

5.9

(1–1

7)

36.5

(30–4

8)

16.6

(7–2

8)

2.0

(0.5

–4.1

)3.7

(2.6

–4.9

)1.6

–4.6

(0.2

–12.3

)3.3

(1.0

–8.6

)1.9

(0.7

–3.3

)2.2

–2.3

(0.1

–10.0

)BEA

STEx

ponen

tial

252

(250–2

59)

17.8

(5–4

0)

90

(59–1

28)

5.6

(1–1

6)

31.5

(30–3

6)

14.4

(6–2

4)

1.8

(0.5

–3.4

)3.9

(2.8

–5.2

)1.7

–5.0

(0.2

–13.9

)3.5

(1.0

–8.7

)2.1

(0.8

–3.8

)2.5

–2.5

(0.1

–11.1

)BEA

STH

ard

min

ima

278

(250–3

13)

22.1

(6–4

9)

1113

(74–1

61)

6.8

(0.9

–19)

45.8

(30–6

5)

19.0

(7.4

–34)

1.8

(0.4

–4.0

)3.6

(2.4

–4.9

)1.4

–3.9

(0.1

–10.9

)3.0

(0.8

–7.8

)1.7

(0.6

–3.0

)2.0

–2.1

(0.1

–8.7

)BEA

STN

one

exce

pt

root

156

(89–2

38)

10.1

(3–2

4)

50.5

(25–8

6)

2.8

(2–2

7)

21.7

(12–3

3)

9.3

(4–1

6)

1.6

(0.3

–3.4

)6.6

(3.2

–10.7

)3.7

–10.8

(0.2

–25.9

)6.7

(1.3

–14.6

)4 (1

.2–7

.5)

5.8

–6.0

(0.2

–17.9

)r8

sw

ith

MrB

ayes

pro

file

Har

dm

inim

a254

(246–2

63)

39.7

(7–7

2)

150

(129–1

70)

22.7

(11–3

4)

37.8

(30–4

6)

12.6

(8–1

7)

5.7

(4.7

–6.7

)2.4

(1.9

–2.9

)0.8

–1.1

(0.5

–1.4

)2.0

(1.5

–2.4

)0.9

(0.5

–1.2

)0.4

–0.7

(0.1

–1.8

)

r8s

with

MrB

ayes

pro

file

None

exce

pt

root

251

(237–2

65)

37.4

(9–6

7)

145

(123–1

67)

24.0

(3.8

–45)

25.2

(18–3

3)

8.4

(5.2

–12)

5.4

(4.7

–6.1

)2.4

(1.9

–2.9

)0.8

–1.2

(0.1

–2.4

)2.0

(1.5

–2.4

)0.9

(0.6

–1.3

)0.4

–0.8

(0.2

–1.9

)

Abbre

viat

ion

CIden

ote

scr

edib

ility

inte

rval

(BEA

ST)

or

confiden

cein

terv

al(r

8s)

.

0

100

200

300

G A G A G AStem lengthCrown ageStem age

Tim

e (M

a)

P ≤ 0.0001P ≤ 0.0001

P = 0.0030

Fig. 2 Comparison between lineage age distributions ofgymnosperms (G, n = 21) and angiosperms (A, n = 110). The greyhorizontal line indicates the Cretaceous–Palaeogene boundary, usedto select lineages for comparison (see text for details). Sampledlineages are non-nested, that is, independent. Box plots show 50percentiles (boxes), median (horizontal line), 95 percentiles(whiskers) and outliers (dots). Probability values are from Mann–Whitney two-tailed tests.

NewPhytologist Research 1003

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

for matK, but included zero for 26S. Although this appearsto indicate that the autocorrelation of rates is weak or absentin 26S, the covariance of rates in the uncorrelated log-normal relaxed clock model might not be effective in detect-ing rate autocorrelation (Ho, 2009). This measure hasconsistently indicated a lack of rate autocorrelation across awide range of datasets (Ho, 2009).

Overall substitution rates varied with the branch smooth-ing method and calibration model (Table 1). In the r8sanalysis using hard minima, high rates were inferred at dee-per nodes, slowing substantially towards the tips, forexample, in the Cycas and Macrozamia lineages (Table 1).This could indicate a systematic bias in r8s. By contrast, inthe BEAST log-normal-calibrated analysis, there was lessrate variation along these lineages and no directional trendalong lineages.

Discussion

Extant gymnosperm groups are not ancient

Relative to angiosperm crown groups in the Cenozoic era,those of gymnosperms are significantly younger, with med-ian ages in each group being 50 and 32 Ma, respectively(Table 2, Fig. 2). Whereas 50% of these angiosperm crownsare at least as old as the early Eocene, 50% of the gymno-sperm crowns are of Oligocene age or younger (Table 2,Fig. 2). Radiations are especially young in the archetypically‘ancient’ cycad genera, with estimated crown ages of c.19 Ma in both Cycas and Ceratozamia, 14 Ma inEncephalartos and 6 Ma in Macrozamia. The conifers of thefamily Araucariaceae are iconic ‘dinosaur’ plants, but theircrown age was estimated at only 36 Ma, with Wollemia

Table 3 Comparison of estimated speciation rates in crown groups of angiosperms and gymnosperms that are £ 30 Myr old

Relative extinction rate (e) = 0.1 Relative extinction rate (e) = 0.9 Relative extinction rate (e) = 0.7

Angiosperms Gymnosperms Angiosperms Gymnosperms Gymnosperms

n 28 18 28 18 18Speciation rate (k)

Median 0.209 0.124 1.179 0.518 0.289Mean 0.349 0.166 1.713 0.721 0.387SD 0.301 0.151 1.298 0.729 0.365

Mann–Whitney test (two-tailed)A, e = 0.1 – 127.0 na 161.0 231.0G, e = 0.1 0.0051** – 12.0 na naA, e = 0.9 na < 0.0001**** – 109.0 naG, e = 0.9 0.0417* na 0.0013** – naG, e = 0.7 0.645, ns na na na –

Sampled crown groups are non-nested, that is, independent. Speciation rates (k = net divergences Myr)1) were calculated assuming a relativeextinction rate (e = ratio of extinction rate to speciation rate) of 0.1, 0.7 or 0.9. In the lower half of the table, values above and below thediagonal are Mann–Whitney U and probability, respectively. A, angiosperms; G, gymnosperms. Asterisks indicate significance at alpha levels:ns, P > 0.05; *, P < 0.05; **, P < 0.01 and ****, P < 0.0001; na, irrelevant comparisons.

Table 2 Comparison of stem and crown ages in gymnosperm (G) and angiosperm (A) lineages

Stem age (Ma) Crown age (Ma) Stem length (Myr)

G A G A G A

n 21 110 21 110 21 110Median 95 78.5 32 50 70 30.5Mean 117.8 84.5 35.2 48.8 82.1 35.8SD 47.7 16.4 19.4 12.9 55.5 21.9

D’Agostino–Pearson omnibus normality testK2 12.79 45.46 3.899 12.78 8.309 21.58P 0.0017 < 0.0001 0.1423 0.0017 0.0157 < 0.0001Passed test? (a = 0.05) No No Yes No No No

Mann–Whitney test (two-tailed)U 503.5 682.0 510.0P <0.0001 0.0030 <0.0001

Stem length is the difference between stem and crown ages.

1004 Research

NewPhytologist

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

diverging from its sister group, Agathis, only at 18 Ma. Thus,the last common ancestor of extant Araucariaceae apparentlyexisted only after the nonavian dinosaurs became extinct.

Gymnosperm lineages are more stemmy than those ofangiosperms (Table 2). Taking into account the reverse dif-ference in crown ages, there is a very large and significantdifference in lineage stem lengths, with those of gymno-sperms being at least 40 Myr longer, on average (Table 2,Fig. 2). Furthermore, in gymnosperms, the median stemlength is 70 Ma, which is more than double the median ageof the twiggy crowns (32 Ma). Long stems such as these havebeen hypothesized to result from extinction and, if multipleco-existent lineages have long stems with crowns of similarage, this could indicate an extinction event driven by a majorenvironmental change (Crisp & Cook, 2009). Are thetwiggy gymnosperm crowns that did not begin to rediversifybefore the Oligocene survivors of an end-Eocene extinctionevent, as indicated by the fossil record (Niklas, 1997; Crepet& Niklas, 2009), when the global climate became sharplycooler and more seasonal (Zachos et al., 2001; McGowranet al., 2004)? If higher extinction rates alone accounted forthe lower extant diversity of gymnosperms, speciation rateswould be about the same as in angiosperms.

Low gymnosperm diversity is explained by highextinction rates, not low speciation rates

It is clear from the huge disparity in extant species diversitythat, over the last 355 Myr, gymnosperms have diversifiedat a much lower net rate than angiosperms. The net diversi-fication rate is the difference between the rates of speciation(generation of new lineages) and extinction (loss of old

lineages), and the low diversity of extant gymnosperms hasbeen alternatively attributed to a low speciation rate(Gorelick & Olson, 2011) and to a high extinction rate(Hill, 1998b; Hill & Brodribb, 1999; Burgoyne et al.,2005; Friis et al., 2005). However, speciation and extinc-tion rates are very difficult to tease apart retrospectively.The estimation of the extinction rate from molecular phy-logenies alone is problematic, and it has been suggested thatthis should be performed in conjunction with the fossilrecord (Hunter, 1998; Doyle & Endress, 2010; Quental &Marshall, 2010; Rabosky, 2010). In fact, the fossil recordstrongly implicates extinction in reducing gymnospermdiversity, despite studies that suggest little change in gym-nosperm diversity through the Cretaceous, Palaeocene andEocene, and notwithstanding the decline in gymnospermabundance relative to that of angiosperms (Niklas, 1997;Lupia et al., 1999). Seed plant extinctions occurred at theCretaceous–Palaeogene boundary, but did not appear todifferentially affect major clades of gymnosperms andangiosperms (Macphail et al., 1994; Niklas, 1997; Wing,2004; Crepet & Niklas, 2009). Much more significantextinctions of gymnosperms occurred during the Cenozoic,especially c. 29 Ma and 16 Ma (Niklas, 1997, Fig. 8.14,Table 8.3; Crepet & Niklas, 2009). In addition, based onthe Southern Hemisphere fossil record, it has been sug-gested that the sharp cooling and drying of the globalclimate at the end of the Eocene caused the extinction ofseveral conifer and cycad lineages (Hill, 1998b; Hill &Brodribb, 1999; Hill, 2004b). Southern Hemisphere coni-fer genera that became extinct in Australia in the mid–lateCenozoic, but survived to the present elsewhere, are physio-logically adapted to very wet climates that no longer exist inAustralia, whereas the surviving genera are all more droughttolerant (Hill, 2004b, Fig. 4). This suggests strong climaticniche conservatism in gymnosperms, such that only thosethat could track suitable habitat survived (Hill, 2004b;Donoghue, 2008). Nevertheless, some of the lineages thatsurvived and rediversified appear to have undergone success-ful adaptive shifts into new environments (Hill & Brodribb,1999). For example, both Callitris (Paull & Hill, 2010) andMacrozamia (Carpenter, 1991) have shifted from wet forestinto dry sclerophyll habitats, where they have diversified.

We estimated the speciation rates in gymnosperms andangiosperms at high and low assumed relative extinctionrates, given the extant diversity and estimated ages of crowngroups (Table 3, Fig. 3). By sampling crown groups£ 30 Myr old, we were comparing gymnosperms withangiosperms through a period in which the relative extinc-tion rate was approximately seven times higher ingymnosperms, according to the fossil record (Niklas, 1997;Crepet & Niklas, 2009). Speciation rates estimated in thisstudy were significantly lower in gymnosperms than inangiosperms when relative extinction rates were assumed tobe equal. However, this disparity disappeared when the

A G A GG

Spec

iatio

n ra

te (λ

)

0.0

5.0

4.0

3.0

1.0

2.0a a

b

c

d

e = 0.1 e = 0.1 e = 0.9 e = 0.9 e = 0.7

Fig. 3 Comparison of estimated speciation rates in crown groups ofangiosperms (A, n = 28) and gymnosperms (G, n = 18) that arepost-Eocene in age (£ 30 Ma). Sampled crown groups are non-nested, that is, independent. Speciation rates (k = net divergencesMyr)1) were calculated assuming a relative extinction rate (e = ratioof extinction rate to speciation rate) of 0.1, 0.7 or 0.9. Box plotsshow 50 percentiles (boxes), median (horizontal line) and 95percentiles (whiskers). Matching lower case letters indicatenonsignificant results of Mann–Whitney two-tailed tests (P > 0.05)and nonmatching letters indicate significant differences.

NewPhytologist Research 1005

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

relative extinction rate was (realistically) assumed to be highin gymnosperms and low in angiosperms. Thus, the combi-nation of evidence from diversification modelling and thefossil record strongly suggests that the difference in diversitybetween gymnosperms and angiosperms can be explainedby a higher extinction rate in gymnosperms, and it is notnecessary to hypothesize a differential speciation ratebetween the two groups over the last 30 Myr. Based on thefossil record, Niklas (1997, Fig. 8.13) indicated that specia-tion rates in gymnosperms and angiosperms were about thesame through the Cenozoic. However, a more detailedgraph (Crepet & Niklas, 2009, Fig. 2) shows substantiallyhigher speciation rates in angiosperms than in gymno-sperms. In the latter study, the difference between theestimated speciation (Fig. 2a) and extinction (Fig. 2b) ratesdoes not appear to equal the inferred diversification rates(Fig. 2c), and the authors admit to limitations in the data.

Crown age sampling

Crown ages can be underestimated by failure to sample taxafrom both sides of the divergence at the basal node of thecrown. It is more probable that some crown nodes were notsampled in the angiosperm data than in our sample of gym-nosperms, because Bell et al. (2010) made no explicit effortto sample Cenozoic angiosperm crowns, whereas wechecked all the gymnosperm crown nodes against species-level phylogenies. The effect of any such bias is unclearbecause most of the sampled angiosperm crowns were onlya little younger than the K–Pg boundary. Therefore, anunsampled older node in a given lineage could be closer tobut younger than the K–Pg boundary compared with thesampled node, in which case the bias would increase the agegap between gymnosperm and angiosperm crowns.Alternatively, an unsampled node could be older than theK–Pg boundary, in which case it would have been sampledas a stem node, not a crown node. This would tend toshorten angiosperm stems and increase the difference instem node ages and stem lengths between angiosperms andgymnosperms. In the case of the speciation rate compari-sons, we checked all selected crown groups to ensure thatthe crown node had been included.

The problem of stem calibrations

The use of probable stem group fossils for calibration is afrequently encountered problem (Doyle & Donoghue,1993; Steiper & Young, 2008; Ho & Phillips, 2009;Magallon, 2010). Calibration points can be placed only atnodes and, if a fossil is thought to represent an extinct sistergroup that diverged part-way along a long internode, thisleads to a difficult placement choice between either end,because either could involve a large error margin (Steiper &Young, 2008; Ickert-Bond et al., 2009; Magallon, 2010).

Probability distributions with soft bounds (e.g. normal) canallow the manipulation of this uncertainty in designing cali-bration priors (Ho & Phillips, 2009); however, the resultwill still favour the node to which the calibration is linked.A recent attempt to break long stems by inserting nodes forfossils and simulating the connecting branch lengths had aminimal effect (Magallon, 2010) and involved guesswork inestimating the branch lengths.

In the present study, we faced a difficult choice in placingthe fossil taxon Macrozamia australis (Carpenter, 1991), theoldest known representative of this genus, and probably astem group fossil because its leaf cuticle characters are notseen in any extant species. We forced a stem node place-ment of this fossil in BEAST by setting its age (c. 28 Ma) asa hard minimum prior for the stem node (8 in Fig. S1) inan analysis in which all other internal constraints were givenlog-normal priors. However, this made little difference(< 0.5 Myr) to the node age estimates compared with con-straining neither the stem nor crown of Macrozamia, forwhich the mean estimates were 103 Ma for the stem and5.9 Ma for the crown (Table 1). We also tried a log-normalstem node calibration with an offset of 28 Ma in BEAST,but this reduced the estimated stem node age by c. 21 Macompared with setting either a 28-Ma hard minimum or noconstraint at the stem node. This paradoxical last resultprobably reflects the diminishing probability of older datesspecified by a log-normal distribution compared with a hardminimum or no minimum (Ho & Phillips, 2009), and doesnot solve the problem of how to place a calibration pointpart-way along a long stem. Like Macrozamia, Ephedra hasa very long stem and short crown (Ickert-Bond et al.,2009), and these authors tried a similar approach to ourswith similarly inconclusive results. Clearly, the problem ofstem calibration requires attention in future research.

Conclusion

We have shown that, in contrast with angiosperms, manyliving gymnosperm lineages have surprisingly youngcrowns, dating from post-Eocene radiations. Both thisfinding and the low diversity of extant gymnosperms (alsorelative to angiosperms) can be explained by majorCenozoic extinction events affecting gymnosperms, whereasthere is equivocal evidence (at best) for differences inCenozoic speciation rates between gymnosperms and angio-sperms. This study shows that comparison of the data fromthe fossil record with results of molecular dating and ratemodelling can contribute to the teasing apart of speciationand extinction rates. However, substantial uncertaintiesremain in dealing with both sources of data: for example,how to place stem group calibrations more accurately inmolecular phylogenies, and how to more accurately esti-mate and reconcile speciation, extinction and netdiversification rates in the fossil record. An explanation of

1006 Research

NewPhytologist

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

why gymnosperms have been more prone than angiospermsto extinction remains a challenge, but studies such as thoseby Hill & Brodribb (2006) and Brodribb & Feild (2010)show how a combination of independent data sources (suchas ecophysiology) with those from phylogenetics and thefossil record can lead to new insights.

Acknowledgements

Funding from the Australian Research Council supportedthe research (project no. DP 0985473). We thank SusanneRenner and an anonymous referee for their useful sugges-tions for improving the manuscript.

References

Alvarez-Yepiz JC, Dovciak M, Burquez A. 2011. Persistence of a rare

ancient cycad: effects of environment and demography. BiologicalConservation 144: 122–130.

Baum DA, Offner S. 2008. Phylogenies and tree-thinking. The AmericanBiology Teacher 70: 222–229.

Bell CD, Soltis DE, Soltis PS. 2010. The age and diversification of the

angiosperms re-revisited. American Journal of Botany 97: 1296–1303.

Berendse F, Scheffer M. 2009. The angiosperm radiation revisited, an

ecological explanation for Darwin’s ‘abominable mystery’. Ecology Letters12: 865–872.

Biffin E, Hill RS, Lowe AJ. 2010a. Did kauri (Agathis: Araucariaceae)

really survive the Oligocene drowning of New Zealand? SystematicBiology 59: 594–601.

Biffin E, Lucas E, Craven L, Ribeiro da Costa I, Harrington M, Crisp

MD. 2010b. Evolution of exceptional species richness amongst lineages

of fleshy-fruited Myrtaceae. Annals of Botany 106: 79–93.

Brodribb TJ, Feild TS. 2010. Leaf hydraulic evolution led a surge in leaf

photosynthetic capacity during early angiosperm diversification. EcologyLetters 13: 175–183.

Bromham L, Penny D. 2003. The modern molecular clock. NatureReviews Genetics 4: 216–224.

Burgoyne PM, van Wyk AE, Anderson JM, Schrire BD. 2005.

Phanerozoic evolution of plants on the African plate. Journal of AfricanEarth Sciences 43: 13–52.

Cantrill DJ, Poole I. 2005. Taxonomic turnover and abundance in

Cretaceous to Tertiary wood floras of Antarctica: implications for

changes in forest ecology. Palaeogeography, Palaeoclimatology,Palaeoecology 215: 205–219.

Carpenter R. 1991. Macrozamia from the early Tertiary of Tasmania and a

study of the cuticles of extant species. Australian Systematic Botany 4:

433–444.

Christenhusz MJM, Reveal JL, Farjon A, Gardner MF, Mill RR, Chase

MW. 2011. A new classification and linear sequence of extant

gymnosperms. Phytotaxa 19: 55–70.

Cook LG, Crisp MD. 2005. Not so ancient: the extant crown group

of Nothofagus represents a post-Gondwanan radiation.

Proceedings of the Royal Society of London, B: Biological Sciences 272:

2535–2544.

Crane PR. 1987. Vegetational consequences of the angiosperm

diversification. In: Friis EM, Chaloner WG, Crane PR, eds. The originsof the angiosperms and their biological consequences. Cambridge, UK:

Cambridge University Press, 107–144.

Crepet WL, Niklas KJ. 2009. Darwin’s second ‘‘abominable mystery’’:

why are there so many angiosperm species? American Journal of Botany96: 366–381.

Crepet WL, Nixon KC, Gandolfo MA. 2004. Fossil evidence and

phylogeny: the age of major angiosperm clades based on mesofossil and

macrofossil evidence from Cretaceous deposits. American Journal ofBotany 91: 1666–1682.

Crisp MD, Cook LG. 2005. Do early branching lineages signify ancestral

traits? Trends in Ecology and Evolution 20: 122–128.

Crisp MD, Cook LG. 2009. Explosive radiation or mass extinction?

Interpreting signatures in molecular phylogenies. Evolution 63:

2257–2265.

Crisp MD, Isagi Y, Kato Y, Cook LG, Bowman DMS. 2010. Livistonapalms in Australia: ancient relics or opportunistic immigrants? MolecularPhylogenetics and Evolution 54: 512–523.

Donoghue MJ. 2008. A phylogenetic perspective on the distribution of

plant diversity. Proceedings of the National Academy of Sciences, USA 105:

11549–11555.

Doyle JA. 2006. Seed ferns and the origin of angiosperms. Journal of theTorrey Botanical Society 133: 169–209.

Doyle JA. 2008. Integrating molecular phylogenetic and paleobotanical

evidence on origin of the flower. International Journal of Plant Sciences169: 816–843.

Doyle JA, Donoghue MJ. 1993. Phylogenies and angiosperm

diversification. Paleobiology 19: 141–167.

Doyle JA, Endress PK. 2010. Integrating Early Cretaceous fossils into the

phylogeny of living angiosperms: Magnoliidae and eudicots. Journal ofSystematics and Evolution 48: 1–35.

Drouin G, Daoud H, Xia J. 2008. Relative rates of synonymous

substitutions in the mitochondrial, chloroplast and nuclear genomes of

seed plants. Molecular Phylogenetics and Evolution 49: 827–831.

Drummond AJ, Ho SYW, Phillips MJ, Rambaut A. 2006. Relaxed

phylogenetics and dating with confidence. PLoS Biology 4: e88.

Drummond AJ, Ho SYW, Rawlence N, Rambaut A. 2007. A rough guideto BEAST. Auckland, New Zealand: The authors.

Earle CJ. 2011. The gymnosperm database. [WWW document] URL

http://www.conifers.org/index.php [accessed on 6 June 2011].

Edgar RC. 2004. MUSCLE: multiple sequence alignment with high

accuracy and high throughput. BMC Bioinformatics 5: 113.

Feild TS, Arens NC. 2005. Form, function and environments of the early

angiosperms: merging extant phylogeny and ecophysiology with fossils.

New Phytologist 166: 383–408.

Finet C, Timme RE, Delwiche CF, Marleta F. 2010. Multigene

phylogeny of the green lineage reveals the origin and diversification of

land plants. Current Biology 20: 2217–2222.

Friis EM, Pedersen KR, Crane PR. 2005. When Earth started blooming:

insights from the fossil record. Current Opinion in Plant Biology 8: 5–12.

Friis EM, Pedersen KR, Crane PR. 2010. Diversity in obscurity: fossil

flowers and the early history of angiosperms. Philosophical Transactionsof the Royal Society of London, B: Biological Sciences 365: 369–382.

Gorelick R, Olson K. 2011. Is lack of cycad (Cycadales) diversity a result of a

lack of polyploidy? Botanical Journal of the Linnean Society 165: 156–167.

Hermsen EJ, Taylor TN, Taylor EL, Stevenson DW. 2006. Cataphylls of

the Middle Triassic cycad Antarcticycas schopfii and new insights into

cycad evolution. American Journal of Botany 93: 724–738.

Hill KD. 1998a. ‘Gymnosperms’ – the paraphyletic stem of seed plants.

In: McCarthy PM, ed. Flora of Australia, vol. 48. Ferns, gymnosperms,and allied groups. Melbourne, Australia: CSIRO Publishing, 505–526.

Hill RS. 1998b. The fossil record of cycads in Australia. In: McCarthy

PM, ed. Flora of Australia, vol. 48. Ferns, gymnosperms, and allied groups.Melbourne, Australia: CSIRO Publishing, 539–544.

Hill KD. 2004a. The cycad pages. [WWW document] URL http://

plantnet.rbgsyd.nsw.gov.au/PlantNet/cycad/Sydney: Royal Botanic

Gardens [accessed on 22 July 2011].

Hill RS. 2004b. Origins of the southeastern Australian vegetation.

Philosophical Transactions of the Royal Society of London, B: BiologicalSciences 359: 1537–1549.

NewPhytologist Research 1007

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

Hill RS, Brodribb TJ. 1999. Southern conifers in time and space.

Australian Journal of Botany 47: 639–696.

Hill RS, Brodribb TJ. 2006. The evolution of Australia’s living biota. In:

Attiwill P, Wilson B, eds. Ecology: an Australian perspective. South

Melbourne, Australia: Oxford University Press, 19–40.

Ho SYW. 2009. An examination of phylogenetic models of substitution

rate variation among lineages. Biology Letters 5: 421–424.

Ho SYW, Phillips MJ. 2009. Accounting for calibration uncertainty in

phylogenetic estimation of evolutionary divergence times. SystematicBiology 58: 367–380.

Hunter JP. 1998. Key innovations and the ecology of macroevolution.

Trends in Ecology and Evolution 13: 31–36.

Ickert-Bond SM, Rydin C, Renner SS. 2009. A fossil-calibrated relaxed

clock for Ephedra indicates an Oligocene age for the divergence of Asian

and New World clades and Miocene dispersal into South America.

Journal of Systematics and Evolution 47: 444–456.

Kass RE, Raftery AE. 1995. Bayes factors. Journal of the AmericanStatistical Association 90: 773–795.

Keppel G, Hodgskiss PD, Plunkett GM. 2008. Cycads in the insular

South-west Pacific: dispersal or vicariance? Journal of Biogeography 35:

1004–1015.

Liao ZH, Chen M, Guo L, Gong YF, Tang F, Sun XF, Tang KX. 2004.

Rapid isolation of high-quality total RNA from Taxus and Ginkgo.

Preparative Biochemistry & Biotechnology 34: 209–214.

Lidgard S, Crane PR. 1988. Quantitative analyses of the early angiosperm

radiation. Nature 331: 344–346.

Lupia R, Lidgard S, Crane PR. 1999. Comparing palynological

abundance and diversity: implications for biotic replacement during the

Cretaceous angiosperm radiation. Paleobiology 25: 305–340.

Macphail MK, Alley NF, Truswell EM, Sluiter IRK. 1994. Early Tertiary

vegetation: evidence from spores and pollen. In: Hill RS, ed. History ofthe Australian Vegetation: Cretaceous to Recent. Cambridge, UK:

Cambridge University Press, 189–261.

Magallon S. 2010. Using fossils to break long branches in molecular

dating: a comparison of relaxed clocks applied to the origin of

angiosperms. Systematic Biology 59: 384–399.

Magallon S, Castillo A. 2009. Angiosperm diversification through time.

American Journal of Botany 96: 349–365.

Magallon S, Sanderson MJ. 2001. Absolute diversification rates in

angiosperm clades. Evolution 55: 1762–1780.

Magallon SA, Sanderson MJ. 2005. Angiosperm divergence times: the

effect of genes, codon positions, and time constraints. Evolution 59:

1653–1670.

Mathews S. 2009. Phylogenetic relationships among seed plants: persistent

questions and the limits of molecular data. American Journal of Botany96: 228–236.

Mathews S, Clements MD, Beilstein MA. 2010. A duplicate gene rooting

of seed plants and the phylogenetic position of flowering plants.

Philosophical Transactions of the Royal Society of London, B: BiologicalSciences 365: 383–395.

McGowran B, Holdgate GR, Li Q, Gallagher SJ. 2004. Cenozoic

stratigraphic succession in southeastern Australia. Australian Journal ofEarth Sciences 51: 459–496.

McIver EE. 2001. Cretaceous Widdringtonia Endl. (Cupressaceae)

from North America. International Journal of Plant Sciences 162:

937–961.

McLoughlin S, Vajda V. 2005. Ancient wollemi pines resurgent. AmericanScientist 93: 540–547.

Midgley JJ, Bond WJ. 1991. Ecological aspects of the rise of angiosperms:

a challenge to the reproductive superiority hypotheses. Biological Journalof the Linnean Society 44: 81–92.

Miller MA, Pfeiffer W, Schwartz T 2010. Creating the CIPRES Science

Gateway for inference of large phylogenetic trees. In: Proceedings of theGateway Computing Environments Workshop (GCE), New Orleans.

New York, USA: Institute of Electrical and Electronics Engineers

(IEEE), 1–8.

Near TJ, Meylan PA, Shaffer HB. 2005. Assessing concordance of fossil

calibration points in molecular clock studies: an example using turtles.

American Naturalist 165: 137–146.

Near TJ, Sanderson MJ. 2004. Assessing the quality of molecular

divergence time estimates by fossil calibrations and fossil-based model

selection. Philosophical Transactions of the Royal Society of London, B:Biological Sciences 359: 1477–1483.

Niklas KJ. 1997. The evolutionary biology of plants. Chicago, IL, USA:

University of Chicago Press.

Niklas KJ, Tiffney BH. 1994. The quantification of plant biodiversity

through time. Philosophical Transactions of the Royal Society of London, B:Biological Sciences 345: 35–44.

Norstog KJ, Nicholls TJ. 1997. The fossil Cycadophytes. In: Norstog KJ,

Nicholls TJ, eds. The biology of the Cycads. Ithaca, NY, USA: Cornell

University Press, 169–201.

Omland KE, Cook LG, Crisp MD. 2008. Tree thinking for all biology:

the problem with reading phylogenies as ladders of progress. Bioessays30: 854–867.

Paull R, Hill RS. 2010. Early Oligocene Callitris and Fitzroya(Cupressaceae) from Tasmania. American Journal of Botany 97: 809–820.

Qiu YL, Li LB, Wang B, Chen ZD, Dombrovska O, Lee J, Kent L, Li

RQ, Jobson RW, Hendry TA et al. 2007. A nonflowering land plant

phylogeny inferred from nucleotide sequences of seven chloroplast,

mitochondrial, and nuclear genes. International Journal of Plant Sciences168: 691–708.

Quental TB, Marshall CR. 2010. Diversity dynamics: molecular phylogenies

need the fossil record. Trends in Ecology and Evolution 25: 434–441.

Rabosky DL. 2006. LASER: a maximum likelihood toolkit for detecting

temporal shifts in diversification rates from molecular phylogenies.

Evolutionary Bioinformatics Online 2006: 257–260.

Rabosky DL. 2010. Extinction rates should not be estimated from

molecular phylogenies. Evolution 64: 1816–1824.

Rai HS, Reeves PA, Peakall R, Olmstead RG, Graham SW. 2008.

Inference of higher-order conifer relationships from a multi-locus plastid

data set. Botany-Botanique 86: 658–669.

Rambaut A 1996. Se–Al: Sequence alignment editor. University of Oxford,

Department of Zoology, http://evolve.zoo.ox.ac.uk/. Accessed 21

August 2011.

Rambaut A, Drummond AJ 2007. Tracer v1.5 [WWW document]. URL

http://beast.bio.ed.ac.uk/Tracer. Accessed 21 August 2011.

Renner SS. 2005. Relaxed molecular clocks for dating historical plant

dispersal events. Trends in Plant Science 10: 550–558.

Renner SS, Strijk JS, Strasberg D, Thebaud C. 2010. Biogeography of the

Monimiaceae (Laurales): a role for East Gondwana and long-distance

dispersal, but not West Gondwana. Journal of Biogeography 37:

1227–1238.

Richardson JE, Weitz FM, Fay MF, Cronk QCB, Linder HP, Reeves G,

Chase MW. 2001. Rapid and recent origin of species richness in the

Cape flora of South Africa. Nature 412: 181–183.

Ricklefs RE. 2007. Estimating diversification rates from phylogenetic

information. Trends in Ecology and Evolution 22: 601–610.

Ronquist F, Huelsenbeck JP. 2003. MrBayes 3: Bayesian phylogenetic

inference under mixed models. Bioinformatics 19: 1572–1574.

Rydin C, Pedersen KR, Friis EM. 2004. On the evolutionary history of

Ephedra: Cretaceous fossils and extant molecules. Proceedings of theNational Academy of Sciences, USA 101: 16571–16576.

Sanderson MJ. 2003. r8s: inferring absolute rates of molecular evolution

and divergence times in the absence of a molecular clock. Bioinformatics19: 301–302.

Sanderson MJ, Thorne JL, Wikstrom N, Bremer K. 2004. Molecular

evidence on plant divergence times. American Journal of Botany 91:

1656–1665.

1008 Research

NewPhytologist

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com

Smith AB. 1994. Rooting molecular trees: problems and strategies.

Biological Journal of the Linnean Society 51: 279–292.

Smith SA, Beaulieu JM, Donoghue MJ. 2010. An uncorrelated relaxed-

clock analysis suggests an earlier origin for flowering plants. Proceedingsof the National Academy of Sciences, USA 107: 5897–5902.

Smith SA, Donoghue MJ. 2008. Rates of molecular evolution are linked

to life history in flowering plants. Science 322: 86–89.

Stamatakis A, Hoover P, Rougemont J. 2008. A rapid bootstrap

algorithm for the RAxML web servers. Systematic Biology 57: 758–771.

Steiper ME, Young NM. 2008. Timing primate evolution: lessons from

the discordance between molecular and paleontological estimates.

Evolutionary Anthropology 17: 179–188.

Stevens PF 2001 onwards. Angiosperm phylogeny website, version 9. [WWW

document] URL http://www.mobot.org/MOBOT/research/APweb/

[accessed on 6 June 2011].

Stockey RA, Kvacek J, Hill RS, Rothwell GW, Kvacek Z. 2005. The fossil

record of Cupressaceae s. lat. In: Farjon A, ed. A monograph ofCupressaceae and Sciadopitys. Kew, UK: Royal Botanic Gardens, 54–68.

Treutlein J, Wink M. 2002. Molecular phylogeny of cycads inferred from

rbcL sequences. Naturwissenschaften 89: 221–225.

Valente LM, Savolainen V, Vargas P. 2010. Unparalleled rates of species

diversification in Europe. Proceedings of the Royal Society of London, B:Biological Sciences 277: 1489–1496.

Verdu M. 2002. Age at maturity and diversification in woody

angiosperms. Evolution 56: 1352–1361.

Wagstaff SJ. 2004. Evolution and biogeography of the austral genus

Phyllocladus (Podocarpaceae). Journal of Biogeography 31: 1569–1577.

Willyard A, Syring J, Gernandt DS, Liston A, Cronn R. 2007. Fossil

calibration of molecular divergence infers a moderate mutation rate and

recent radiations for Pinus. Molecular Biology and Evolution 24: 90–101.

Wing SL. 2004. Mass extinctions in plant evolution. In: Taylor TN, ed.

Extinctions in the history of life. Cambridge, UK: Cambridge University

Press, 61–97.

Won H, Renner SS. 2006. Dating dispersal and radiation in the

gymnosperm Gnetum (Gnetales): clock calibration when outgroup

relationships are uncertain. Systematic Biology 55: 610–622.

Xiao LQ, Moller M, Zhu H. 2010. High nrDNA ITS polymorphism in

the ancient extant seed plant Cycas: incomplete concerted evolution and

the origin of pseudogenes. Molecular Phylogenetics and Evolution 55:

168–177.

Zachos J, Pagani M, Sloan L, Thomas E, Billups K. 2001. Trends,

rhythms, and aberrations in global climate 65 Ma to present. Science292: 686–693.

Zgurski JM, Rai HS, Fai QM, Bogler DJ, Francisco-Ortega J, Graham

SW. 2008. How well do we understand the overall backbone of cycad

phylogeny? New insights from a large, multigene plastid data set.

Molecular Phylogenetics and Evolution 47: 1232–1237.

Supporting Information

Additional supporting information may be found in theonline version of this article.

Fig. S1 Majority-rule consensus phylogram of gymno-sperms from MrBayes showing calibration points describedin Table S2.

Table S1 Taxonomic sample with GenBank accessionnumbers and species’ identities and gymnosperm terminalsarranged hierarchically according to Christenhusz et al.(2011)

Table S2 Calibrations for relaxed molecular clock dating

Table S3 Age and speciation rate estimates for selectedcrown groups of angiosperms and gymnosperms that are£ 30 Myr old, ranked by age

Please note: Wiley-Blackwell are not responsible for thecontent or functionality of any supporting informationsupplied by the authors. Any queries (other than missingmaterial) should be directed to the New Phytologist CentralOffice.

NewPhytologist Research 1009

� 2011 The Authors

New Phytologist � 2011 New Phytologist Trust

New Phytologist (2011) 192: 997–1009

www.newphytologist.com