Origin and Early Evolution of Angiosperms

23 Pages·2008·0.46 MB·English

Origin and Early Evolution of Angiosperms
Douglas E. Soltis,a Charles D. Bell,b Sangtae Kim,c
and Pamela S. Soltisd
aDepartmentofBotany,UniversityofFlorida,Gainesville,Florida,USA
bDepartmentofBiologicalSciences,UniversityofNewOrleans,
NewOrleans,Louisiana,USA
cNationalInstituteofBiologicalResources,Incheon,Korea
dFloridaMuseumofNaturalHistory,UniversityofFlorida,Gainesville,Florida,USA
Contributionsfrompaleobotany,phylogenetics,genomics,developmentalbiology,and
developmental genetics have yielded tremendous insight into Darwin’s “abominable
mystery”—the origin and rapid diversification of the angiosperms. Analyses of mor-
phological and molecular data reveal a revised “anthophyte clade” consisting of the
fossilsglossopterids,Pentoxylon,Bennettitales,andCaytoniaassistertoangiosperms.
Molecular estimates of the age of crown group angiosperms have converged on 140–
180millionyearsago(Ma),olderthantheoldestfossils(132Ma),suggestingthatolder
fossilsremaintobediscovered.Whetherthefirstangiospermswereforestshrubs(dark-
and-disturbedhypothesis)oraquaticherbs(wet-and-wildhypothesis)remainsunclear.
Thenear-basalphylogeneticpositionofNymphaeales(waterlilies),whichmayinclude
the well-known fossil Archaefructus, certainly indicates that the aquatic habit arose
early.Afterinitial,early“experiments,”angiospermsradiatedrapidly(≤5millionyears
[Myr]),yieldingthefivelineagesofMesangiospermae(magnoliidsandChloranthaceae
assisterstoacladeofmonocotsandeudicots+Ceratophyllaceae).Thisradiationul-
timatelyproducedapproximately97%ofallangiospermspecies.Updatedestimatesof
divergencetimesacrosstheangiospermsconductedusingnonparametricratesmooth-
ing,withoneormultiplefossils,wereolderthanpreviousreports,whereasestimates
usingPATHd8weretypicallyyounger.Virtuallyallangiospermgenomesshowevidence
ofwhole-genomeduplication,indicatingthatpolyploidymayhavebeenanimportant
catalystinangiospermevolution.Althoughthefloweristhecentralfeatureofthean-
giosperms,itsoriginandsubsequentdiversificationremainmajorquestions.Variation
inspatialexpressionoffloralregulatorsmaycontrolmajordifferencesinfloralmor-
phologybetweenbasalangiospermsandeudicotmodels.
Keywords: phylogeny;divergence;floralevolution;genomeevolution
Introduction as “an abominable mystery” in a well-known
quotation taken from a letter to J.D. Hooker.
Angiosperms represent one of the greatest
Throughcontributionsfrompaleobotany,phy-
terrestrial radiations. The oldest fossils date
logenetics,classicaldevelopmentalbiology,and
from the early Cretaceous (Friis et al. 2006),
modern developmental genetics (evo–devo),
130–136 million years ago (Ma), followed by
tremendous progress has recently been made
a rise to ecological dominance in many habi-
in elucidating the origin and diversification of
tats before the end of the Cretaceous. Dar-
theangiosperms.Weattempttosynthesizeand
win referred to the origin of the angiosperms
summarizemanyoftheserecentdevelopments,
noting their importance to the rapidly chang-
ing angiosperm paradigm. We also provide a
Addressforcorrespondence:DouglasE.Soltis,DepartmentofBotany,
UniversityofFlorida,Gainesville,[email protected]fl.edu possibleroadmapforfutureresearch.
Ann.N.Y.Acad.Sci.1133:3–25(2008).(cid:2)C 2008NewYorkAcademyofSciences.
doi:10.1196/annals.1438.005
3
4 AnnalsoftheNewYorkAcademyofSciences
Fossils 1999; Soltis et al. 1999a; Chaw et al. 2000;
Magallo´n & Sanderson 2002; Burleigh &
Paleobotany has at least three crucial roles Mathews 2004). Furthermore, there are no
to play in resolving the origin and early diver- known fossils representing unequivocal stem
sification of the flowering plants and address- group angiosperms (i.e., angiosperms that
ing the following questions: (1) What lineage clearly attach below the basal node leading to
isancestraltotheangiosperms?(2)Whatwere Amborella,Nymphaeales,andallotherlivingan-
the first angiosperms, and how have charac- giosperms).Hence,determiningtheclosestrel-
tersevolved?Fossildiscoveriesandtheanalysis ativesofangiospermsisamajorchallenge.
of existing data have made significant contri- The demise of the anthophyte hypothesis
butions to our understanding of the possible profoundly altered views of angiosperm ori-
ancestorsoffloweringplants(seemostrecently gins; accompanying this demise is a need for
Friis et al. 2007), as well as to the early diversi- alternatives. If no extant gymnosperm is sis-
fication of angiosperms. (3) When did the ma- ter to angiosperms, what fossil lineages, if any,
jor clades of angiosperms diversify, on the ba- are closely related to flowering plants? Doyle
sis of fossil calibrations and molecular-based (2001) and Soltis et al. (2005) conducted a se-
methods? ries of analyses of morphological and molec-
ular data sets involving extant taxa, as well as
fossils,toreassesstheclosestrelativesofthean-
Ancestors of Flowering Plants
giosperms. These analyses yielded similar re-
The closest relatives of angiosperms remain sultsandrevealedarevised“anthophyteclade.”
amystery.Phylogeneticanalysisofmorpholog- In this clade Caytonia is sister to angiosperms,
ical features conducted in the 1980s suggested followed by Bennettitales, Pentoxylon, and glos-
that Gnetales were the closest living relatives sopterids(Fig.1).Hence,fossilssuchasCaytonia
of angiosperms (e.g., Crane 1985; Doyle & andBennettitalesmaybetheclosestrelativesof
Donoghue 1986). Cladistic analyses of extant angiosperms (Fig. 1). However, the recent dis-
and fossil taxa recovered a clade of Bennet- coverythatgnetophytesandBennettitalesshare
titales, Pentoxylon, Gnetales, and angiosperms features (Friis et al. 2007) suggests that further
(Crane1985;Doyle&Donoghue1986).Doyle attention be given to circumscribing and eval-
and Donoghue (1986) named this clade the uating a revised anthophyte clade. Although
“anthophytes”inreferencetotheflowerlikere- these analyses are useful, one of the biggest
productive structures in all members. The an- remaining challenges facing evolutionary bi-
thophyte hypothesis, that angiosperms belong ologists is ascertaining the fossil lineages that
to this clade and are possibly sister to the gne- represent the closest relatives of angiosperms.
tophytes, concomitantly had a profound effect DoCaytonialesandBennettitalesactuallyrep-
on views of the evolution of the angiosperms. resent these lineages? Are there key fossil links
Forexample,acceptanceoftheanthophytehy- thatremaintobediscovered?
pothesisstimulatedthereinterpretationofchar-
acter evolution (Frohlich 1999; Donoghue &
The Earliest Angiosperms: Fossils
Doyle2000),includingtheoriginofthecarpel
anddoublefertilization(Friedman1994;Doyle There are two highly divergent views on
1998). the general habit of the earliest angiosperms:
Molecular phylogenetic studies, however, woody and terrestrial or herbaceous and
later revealed that Gnetales are not the sister aquatic. This has actually been a longstand-
grouptotheangiospermsbutthatextantgym- ing debate (reviewed in Soltis et al. 2005). The
nosperms form a clade, with Gnetales most hypothesis that the earliest angiosperms were
likely associated with conifers (e.g., Qiu et al. woodyissupportedbythefactthatmostbasal
Soltisetal.:OriginandEarlyEvolutionofAngiosperms 5
knownfossilangiospermswereaquatic.Archae-
fructusrepresentsperhapstheoldest,mostcom-
plete angiosperm fossil (Sun et al. 2002); it is
estimated to be approximately 115–125 mil-
lionyears(Myr)old.Onthebasisofmorphol-
ogy, it clearly was aquatic. The phylogenetic
placement of the fossil Archaefructus as sister to
all extant angiosperms (Sun et al. 2002), plus
the near-basal phylogenetic position of extant
Nymphaeales (water lilies, below), lends sup-
port to the view that the aquatic habit arose
earlyinangiospermsandthatperhapstheear-
liestangiospermswereaquatic.However,later
analyses (e.g., Friis et al. 2003) questioned the
placement of Archaefructus as sister to all ex-
tant angiosperms; some analyses of Friis et al.
placed Archaefructus with water lilies. The re-
cent discovery that Hydatellaceae are part of
the water lily clade (Saarela et al. 2007), and
hence a new branch near the base of the an-
giosperms, greatly increases the morphologi-
caldiversityencompassedbytheNymphaeales
clade. This finding raised the possibility that
Nymphaeales were once much more diverse,
and could have encompassed Archaefructus, as
wellasothernow-extinctlineages(Doyle,sub-
mitted). Hence, it now seems prudent to keep
Figure 1. Revised views of the phylogeny of
an open mind regarding the placement of Ar-
seed plants showing putative closest relatives of an-
chaefructus—it represents well the difficulty in
giospermsobtainedbyusingboththemorphological
matrixofDoyle(1996)andmoleculardata(reviewed placingfossillineagesthataremorphologically
inSoltisetal.2005).Arevised“anthophyte”cladeis distinct,withnoclearsynapomorphieswithex-
depicted.(A)TreemodifiedfromSoltisetal.(2005), tanttaxa.
inwhichmoleculardataforseedplantsarecombined
Another early angiosperm fossil (unnamed)
with the morphological matrix of Doyle (1996). (B)
wasconsideredapossiblewaterlilyrelativeby
TreemodifiedfromDoyle(2001),inwhichamolec-
ular constraint is used, placing Gnetales with other Friis et al. (2001). This fossil is dated at ap-
extantgymnosperms. proximately 125–115 Myr old and was used
as evidence to support the antiquity of the
angiosperms are woody, and all gymnosperms Nymphaeales lineage. This putative ancient
are woody, as are the fossil lineages that are water lily fossil is therefore extremely im-
consideredmostcloselyrelatedtoangiosperms portant in discussions of the diversification
(e.g.,Caytoniales,Bennettitales;seeearlierdis- of Nymphaeales, as well as the angiosperms.
cussion). Amborella, the sister to all other living A phylogenetic analysis, using the morpho-
angiosperms,iswoody,asaremembersofAus- logical data set of Les et al. (1999) for ex-
trobaileyales, another early-branching lineage tant Nymphaeales, placed the fossil as sister
oflivingangiosperms. to Nymphaeales; synapomorphies with extant
An aquatic origin of angiosperms is sup- Nymphaeales included a syncarpous gynoe-
ported by the fact that several of the earliest cium,aperigynousorepigynousperianth,and
6 AnnalsoftheNewYorkAcademyofSciences
a central protrusion of the floral apex be- been the enormous progress made in eluci-
tweenthecarpels.However,neitherofthelast dating angiosperm evolutionary history (re-
two characters is found consistently through- viewed in Judd & Olmstead 2004; Soltis &
out Nymphaeales (Friis et al. 2001). The un- Soltis 2004; Soltis et al. 2005). We provide a
named fossil exhibits features shared by both summarytreeofourcurrentunderstandingof
Nymphaeales and Illicium (Austrobaileyales) angiospermrelationships(Fig.2).Phylogenetic
(Friisetal.2001;Gandolfoetal.2004;Yooetal. analyses based on an ever-increasing number
2005)andoccursinasitewithabundantfossil of gene sequences have clearly and unequiv-
seeds attributed to Illiciales (Friis et al. 2001). ocally established the first branches of extant
The recent molecular analyses and dating ex- angiosperm diversity. Many studies have re-
perimentsofYooetal.(2005)raisethepossibility vealedstrongsupportforthesisterrelationship
that the unnamed fossil of Friis et al. may have of Amborellaceae, followed by Nymphaeales,
actuallybeenpartofanancientassemblagethat and then Austrobaileyales to all other extant
included Nymphaeales and Austrobaileyales. angiosperms (Mathews & Donoghue 1999,
That is, the unnamed fossil probably occupies 2000; Parkinson et al. 1999; Qiu et al. 1999,
adeeperplaceontheangiospermtreethanthe 2000, 2005; Soltis et al. 1999b, 2000, 2004;
branch leading to water lilies. Endress (2008) Barkman et al. 2000; Graham & Olmstead
recently reached the same conclusion on the 2000;Grahametal.2000;Soltisetal.2000;Za-
basisofareexaminationofthemorphologyof nisetal.2002,2003;Borschetal.2003;Hiluetal.
Microvictoria (Gandolfo et al. 2004)—the fossil 2003; Kim et al. 2004; Stefanovic´ et al. 2004;
differs in key features from extant genera of Leebens-Macketal.2005;Stefanovic´etal.2005;
Nymphaeaceae. Jansen et al. 2007; Moore et al. 2007). In most
ThisunnamedfossilofFriisetal.(2001)and analyses, the monotypic Amborella is followed
Archaefructus certainly indicate that the aquatic bythewaterliliesorNymphaeales(whichcom-
habitaroseearlyinangiospermevolution.This priseNymphaeaceae)andCabombaceae(APG
view is further supported by the near-basal II2003)andHydatellaceae(Saarelaetal.2007);
placementofNymphaeales(waterlilies)inphy- althoughafewinvestigationsplacedAmborel-
logenies of extant taxa. However, there are laceae plus Nymphaeales as sister to all other
undoubtedly earlier, as yet undiscovered, an- angiosperms(reviewedinSoltisetal.2005),re-
giospermfossils.Inthisregard,molecularesti- centanalysesinvolvingcompleteornearlycom-
mates for the age of the angiosperms are con- plete plastid genome sequences have placed
verging;mostrecentestimatesareintherange Amborella alone as sister to all other extant an-
of 140–180 Ma (Bell et al. 2005), suggesting giosperms (e.g., Jansen et al. 2007; Moore et al.
anageforfloweringplantsthatissubstantially 2007).Althoughthisnowappearstobethefa-
olderthanthedateof132Maonthebasisofthe vored topology, it should be confirmed using
knownfossilrecord.Thesemolecularestimates nuclearsequencedata.
suggest, in fact, that the earliest angiosperms An exciting recent molecular finding is the
mayhaveariseninthelateJurassic,ratherthan placement of Hydatella (Hydatellaceae) in the
the early Cretaceous, and that the oldest an- Nymphaeales clade (Saarela et al. 2007). Hy-
giospermfossilsarestillundiscovered. datellaceae had been placed in the monocots
(Poales), but molecular data revealed instead
that the family is part of the early-diverging
Phylogenetics angiospermcladeNymphaealeswithhighsup-
port values. Hence, this is a dramatic phylo-
Resolving the Big Picture
genetic shift. This result is important in that it
One of most exciting recent developments greatlyexpandsthemorphologicaldiversityen-
in the study of angiosperm evolution has compassed by the Nymphaeales. For example,
Soltisetal.:OriginandEarlyEvolutionofAngiosperms 7
Figure2. Summarytopologyofcurrentviewofdeep-levelangiospermrelationships.This
topology is based on the analyses of the three-gene, 567-taxon data set (Soltisetal.2000,
2007a) with modifications based on the recent analyses of nearly complete plastid genome
datasets(Jansenetal.2007;Mooreetal.2007).
Rudalletal.(2007)proposedthatthemostplau- After Amborella and Nymphaeales, an Aus-
sibleinterpretationofthereproductive unitsis trobaileyales clade of Illiciaceae, Austrobailey-
that each such unit represents an aggregation aceae, and Trimeniaceae is the subsequent
of reduced, unisexual, apetalous flowers; this sister group to all remaining angiosperms.
type is different from flowers of Nymphaeales Whereas these basalmost branches are well
(Rudall et al. 2007). However, Hydatella has supported and resolved, relationships among
ascidiate carpel development, consistent with the remaining major lineages of angiosperms
placementinNymphaeales.Theplacementof havebeenunclearanddifficulttoresolve,with
Hydatella within Nymphaeales raises the pos- three-, five-, nine-, and even 11-gene data sets
sibility that this clade was once far more di- (see references above). These remaining an-
versemorphologicallyandincreasestheplausi- giosperms,orMesangiospermae(Cantinoetal.
bilitythatthemorphologyofArchaefructuscould 2007),aremagnoliids,Chloranthaceae,mono-
be encompassed by this clade, as originally cots, Ceratophyllaceae, and eudicots. Analy-
proposed by Friis et al. (2003; see also Doyle, ses ofmultigenedatasetswith adequatetaxon
submitted). sampling haveprovided strong support for the
8 AnnalsoftheNewYorkAcademyofSciences
monophyly of each of these clades of an- lion years); therefore, resolving relationships
giosperms. However, different analyses have amongthesecladeshasbeendifficult.
depictedvariousrelationshipsamongthesefive Recent analyses (e.g., Moore et al. 2007;
clades,withonlyweaksupport. Jansen et al. 2007; Jian et al., 2008) demon-
In contrast to earlier studies, recent anal- strate that with enough sequence data (20,000
yses based on complete or nearly complete or more base pairs), the remaining, problem-
plastid genome sequences appear to resolve atic,deep-levelproblemsinthefloweringplants
relationships among these five lineages of can also be resolved. Especially promising
Mesangiospermae.AfterthegradeofAmborella, are analyses involving complete sequencing of
Nymphaeales, and Austrobaileyales, Chloran- the plastid genome (e.g., Jansen et al. 2007;
thaceaeandmagnoliidsmayformacladethat Moore et al. 2007), particularly given the cur-
is sister to a large clade of monocots and Cer- renteaseofcompleteplastidgenomesequenc-
atophyllaceaepluseudicots(Mooreetal.2007) ing with new sequencing technologies (e.g.,
(Fig. 2). The clade of monocots as sister to Moore et al. 2006). An alternative method
theeudicot+Ceratophyllumcladeisstronglysup- that is promising, and much less costly, is the
ported (Moore et al. 2007). Jansen et al. (2007) completesequencingoftheslowlyevolvingin-
alsofoundtheserelationships,exceptthattheir vertedrepeatoftheplastidgenome(Jianetal.,
initial analysis did not include Ceratophyllum. 2008), which is readily accomplished in most
The first three living branches were relatively angiosperm groups by using the polymerase
easy to recover in phylogenetic analyses, a re- chainreaction–basedASAP(amplification,se-
sult reported in some of the initial multigene quencing and annotation of plastomes; Dhin-
analysesofangiosperms(e.g.,Soltisetal.1999, gra&Folta2005)method.
2000; Qiu et al. 1999; Parkinson et al. 1999;
Mathews & Donoghue 1999, 2000). These re-
Implications of Phylogeny for the
sults suggested a more gradual diversification
Earliest Angiosperms
process (or greater extinction) at the base of
the angiosperm tree. However, after some ini- A well-resolved phylogeny for extant an-
tial early “experiments” in angiospermy, the giosperms has been used as a framework in
angiospermsrapidlyradiated,yieldingthefive an effort to reconstruct the habit, as well as
lineages of Mesangiospermae. This is an ex- other morphological features, of the earliest
citing example of an important molecular in- flowering plants (APG II 2003; Soltis et al.
sightintoDarwin’s“abominablemystery.”The 2005). As noted, Amborella, which is sister to
fossil record certainly supports the presence all other extant angiosperms, is woody, as
of many diverse lineages early in angiosperm are Austrobaileyales. Anatomical and physio-
evolution (e.g., Friis et al. 1994, 1999, 2000, logical features of Amborella (which lacks ves-
2001), but phylogenetic analyses of extant an- selelements)andAustrobaileyales(whichhave
giosperms indicate that the radiation respon- “primitive” vessel elements considered inter-
sible for nearly all extant angiosperm diversity mediatebetweentracheidsandvesselelements)
was not associated with the origin of the an- (Carlquist & Schneider 2001, 2002) prompted
giosperms but occurred after the diversifica- the hypothesis that the earliest angiosperms
tionofAmborella,Nymphaeales,andAustrobai- were understory shrubs, perhaps without ves-
leyales (Mathews & Donoghue 1999; Soltis sel elements, andwith lower transpiration and
etal.1999b).ThecladescomposingtheMesan- stem water movement than is typical of most
giospermae (monocots, Ceratophyllaceae, eu- extant angiosperms (Feild et al. 2000, 2003,
dicots,Chloranthaceae,andmagnoliids)repre- 2004). Such plants occupied habitats referred
sentthe“bigbang”ofangiospermevolution— to as “dark and disturbed” and prompted
they diversified rapidly (within only a few mil- the hypothesis that the first angiosperms were
Soltisetal.:OriginandEarlyEvolutionofAngiosperms 9
understory shrubs living in partially shaded estimatesofbetween140and180Ma(Sander-
habitatsthatdependedoncanopydisturbance son et al. 2004; Bell et al. 2005; Moore et al.
to open up new sites for colonization. Out- 2007). This apparent convergence in age esti-
groupcomparisonsfavorawoodyhabitforthe mation is probably due to the increased num-
first angiosperms. All other “anthophytes” are ber of characters used, as well as the use of
woody; this includes extant gymnosperms, as new “relaxed clock” methods that allow rates
well as those fossils typically considered close to vary across the tree (e.g., penalized likeli-
relativesofangiosperms(e.g.,Caytoniales,Ben- hood, PATHd8, and Bayesian methods). Also,
nettitales). recent studies have used multiple fossils as ei-
However, the strictly aquatic water lily lin- ther calibrations or maximum and minimum
eage (Nymphaeales) follows Amborella as the ageconstraints.
subsequent sister to all other flowering plants. Todate,themostcomprehensivedivergence
As a result of the near-basal placement of time analysis for the angiosperms, for taxon
Nymphaeaceae, the habit of the first flower- sampling,isthatofWikstro¨metal.(2001).The
ing plants is reconstructed as equivocal on the results from Wikstro¨m et al. (2001) have subse-
basis of analyses of extant taxa (Soltis et al. quentlybeenusedasatemporalframeworkfor
2005). Nonetheless, the early appearance of many ecological studies (e.g., Slingsby & Ver-
thewaterlilylineage,togetherwiththediscov- boom 2006; Vamosi et al. 2006; Edwards et al.
ery that some of the oldest angiosperm fossils 2007; Webb et al. 2007), as well as “external”
are aquatic (e.g., Archaefructus), reinforced the calibration points for subsequent divergence
hypothesis that the earliest angiosperms may time analyses of groups that may lack reliable
have been aquatic—that is, “wet and wild” fossils (e.g., Crayn et al. 2006; Park et al. 2006).
(D. Dilcher, pers. comm.; see Coiffard et al. Although a landmark study, the analysis of
2007). Wikstro¨metal.sufferedinseveralmethodolog-
icalwaysthatwereunavoidableatthetimebe-
cause the necessary analytical tools were lack-
ing, namely, reliance on one calibration point
Molecular Data and Divergence
and the use of nonparametric rate smoothing
Times
(NPRS;amethodthathasbeenshowntobea
biasedestimatorofcladeages).
Dating Origin of the Angiosperms
Here we have reanalyzed the data set used
In an attempt to estimate the timing of an- by Wikstro¨m et al. by using multiple calibra-
giosperm origins, various authors have used a tion points and age constraints (21 total; Ta-
variety of different molecular data sets and es- ble 1) and a tree topology (Fig. 3) inferred
timation procedures. Although the oldest an- from a recent Bayesian analysis (Soltis et al.
giospermfossilsdateto132Ma,moleculares- 2007). To make our results as comparable as
timates have been somewhat older. Molecular possible to those of Wikstro¨m et al. (2001), we
estimateshavedatedangiospermoriginstothe usedtheirfossilcalibrationpointasafixedage
LowerJurassic(175–200Ma;Sanderson1997; inallanalysesandinferredagesbyusingNPRS.
Sanderson&Doyle2001;Wikstro¨metal.2001; We also used PATHd8 (Britton et al. 2007),
Belletal.2005;Magallo´n&Sanderson2005),to a recently proposed relaxed-clock method to
theTriassic(200–250Ma;Sanderson&Doyle estimate divergence times across angiosperms.
2001; Magallo´n & Sanderson 2005), or even LikeWikstro¨metal.(2001),wecalculatearange
the Paleozoic (300 or 350 Ma; e.g., Ramshaw ofagesonthebasisofbranchlengthestimates
etal.1972;Martinetal.1989).However,inre- from parsimony and maximum likelihood.
centyears,studiesestimatingtheageofcrown Under the maximum likelihood criterion,
groupangiospermsappeartobeconvergingon a general-time-reversible model of sequence
10 AnnalsoftheNewYorkAcademyofSciences
TABLE 1. MultipleCalibrationPointsandAgeConstraintsUsedinDivergenceTimeEstimations
Organismgroup Age(Myr) MRCAaof:
Altingia(Altingiaceae) 88.5–90.4 AltingiaandPaeonia
Hamamelidaceae 84–86 DaphniphyllumandItea
Cercidiphyllaceae 65–71 CercidiphyllumandCrassula
Divisestylus(Iteaceae) 89.5–93.5 RibesandItea
Ailanthus(Simaroubaceae/Rutaceae/Meliaceae) 50 AilanthusandSwietenia
Burseraceae/Anacardiaceae 50 BurseraandSchinus
Parbombacaceoxylon(Malvaless.l.) 65.5–70.6 ThymeaandBombax
Bedelia(Fagales) 84 CucumisandCarya
Paleoclusia(Clusia/Hypericum) 89–90 DicellaandMesua
Illiciospermum(Illiciales) 89–99 IlliciumandSchisandra
Diplodipelta 36 ValerianaandDipsacus
AngiospermCrownGroup 131.8(min) AmborellaandValeriana
Eudicots(crowngroup) 125 RanunculusandValeriana
Virginianthus(Calycanthaceae) 98–113 CalycanthusandIdiospermum
Unnamed(Chloranthaceae) 98–113 HedyosmumandChloranthus
Perisyncolporites(Malpighiales) 49 DicellaandMalpighia
Pseudosalix(Malpighiales) 48 IdesiaandPopulus
Cornales 86 CornusandNyssa
Platanaceae 98–113 PlatanusandPlacospermum
Buxaceae 98–113 DidymelesandBuxus
Bignoniaceae 49.4 CatalpaandVerbena
Bignoniaceae 35 CatalpaandCampsis
aMRCA,mostrecentcommonancestor.
evolution was used. To accommodate among- these values. However, the difference at any
site rate variation, we used a discrete gamma specificnodewasgreaterthan10Myrforonly
((cid:1)) distribution with four rate categories and three of the nodes. One explanation for the
accountedfortheproportionofinvariablesites differences is that many of the minimum age
(I)inthesubstitutionmodel(Fig.3). constraintsusedinourstudy(sevenof20)were
The results of these analyses for some of older than the ages inferred by Wikstro¨m et al.
the major early angiosperm groups are sum- (2001),pushingagesfurtherbackintimefrom
marized in Table 2. In almost all cases, our theseconstrainednodes.
NPRSdates,witheitheroneormultiplefossils, Althoughthereappearstobeaconvergence
wereolderthanthoseofWikstro¨metal.(2001), in the age estimates for angiosperms across
potentially because we are using a slightly different studies, new data (whether molecu-
differenttreetopology(Soltisetal.2007a),Like- lar or fossil), along with new estimation meth-
wise,acrossthenodesexaminedhere,PATHd8 ods, will help to further refine our ideas of
providedyoungerageestimatesundereachfos- the age and time frame of angiosperm ori-
siltreatment. ginsanddiversification.Withtheinterestinus-
The mean absolute difference across all inglargeangiospermphylogeniestoinvestigate
nodes using these methods was approximately questions concerning ecology and compara-
5 Myr, with nearly all differences being older tive biology, along with the continual refine-
in the multiple-calibration analysis than the mentinrelationshipsamongmajorangiosperm
single-calibration analysis. This finding is po- lineages, a new estimate of the ages of the
tentiallysignificantgiventhattheaverageboot- major clades is much needed and should be
strap estimate of standard errors was less than pursued.
Soltisetal.:OriginandEarlyEvolutionofAngiosperms 11
Figure 3. Divergence time estimates for major clades of angiosperms from a reanalysis of the three-
gene, 567-taxon data set by using multiple calibration points and methods (see text). Dates corresponding
tonumberednodesaregiveninTable1.
Extant Members of “Old” Lineages Eocene to early Oligocene (41.1–67.7 Ma),
Represent Recent Diversifications and the two extant genera of Cabombaceae
diverged during the Miocene (19.9–65.6 Ma).
The divergence time estimates of Yoo TheseresultsindicatethatextantNymphaeales
et al. (2005) suggest that crown group diversifiedrelativelyrecently,whereasthestem
Nymphaeales date to the Eocene (44.6– lineage to Nymphaeales is old, on the ba-
67.9 Ma); Nymphaeaceae and Cabom- sis of a fossil attributed to Nymphaeales from
baceae split at that point. Extant genera of the Early Cretaceous (125–115 Ma; Friis et al.
Nymphaeaceae began to diversify in the late 2001). Although the Friis et al. fossil and a
12 AnnalsoftheNewYorkAcademyofSciences
TABLE 2. EstimatedCladeAges(Myr)
Cladeno.a Clade Wikstrometal.b NPRSe PATHd8c NPRSd PATHd8d
1 Angiosperms 158–179 166–187 152–171 165–188 164–173
2 153–171 159–179 132–155 161–180 149–158
3 147–165 151–172 127–142 154–174 138–146
4 —e 145–167 125–135 148–170 132–140
5 — 147–165 126–136 150–168 131–140
6 122–132 140–159 45–50 145–162 98
7 127–134 135–153 39–41 139–156 98
8 108–113 118–144 37–39 118–146 98
9 140–155 144–157 123–131 146–160 131–136
10 Eudicots 131–147 135–149 110–119 136–152 125
11 130–144 131–146 108–118 132–149 123–124
12 128–140 130–144 107–117 131–148 122–123
13 124–137 124–139 102–110 125–142 110–116
14 123–135 124–136 100–106 125–139 107–113
15 116–127 118–129 89–95 119–132 96–103
16 114–124 116–124 84–90 117–128 93–96
17 104–111 106–113 84–90 109–117 92–95
18 106–114 107–113 72–77 110–117 86
19 — 104–109 71–76 107–113 79–81
20 102–112 91–104 69–75 96–108 79–80
21 114–125 115–126 86–92 120–130 96–100
22 111–121 113–123 85–89 116–126 94–97
23 108–117 106–120 85–89 109–124 93–97
24 100–109 97–110 84 101–113 87–89
aCladenumbersrefertonumberednodesinFigure3.
bWikstrometal.datesaregivenintheirSupplementalData,availableonline.Becausewetriedtoestimatetheage
oftheroot,butwithconstraintsasnoted,itispossiblethatarangeofsolutionsmightexist.Foreachanalysis,wefixed
twoages:(1)theageoftherootnode(mostrecentcommonancestorofseedplants)at310Myrand(2)theageofthe
mostrecentcommonancestorofCucurbitalesandFagalesat84Myr.
cBasedonsinglefixedcalibrationofthemostrecentcommonancestorofCucurbitalesandFagalesat84Myr.
dBasedon20minimumageconstraintsinadditiontosinglefixedcalibrationofthemostrecentcommonancestor
ofCucurbitalesandFagalesat84Myr.
eNodenotcompatiblewithtreeused.
fossil attributed to Nymphaeaceae from the tures resembling those of Hedyosmum from the
middle Cretaceous (90 Ma; Gandolfo et al. Barremian–Aptian boundary, approximately
2004) may be best placed deeper in the an- 125 Ma (see Friis et al. 1994, 1999; Friis 1997;
giosperms, there is a distinct gap between the Doyleetal.2003;Eklundetal.2004).However,
origin of Nymphaeales and its diversification divergence time estimates based on molecular
intomodernlineages. data indicate that the extant genera of Chlo-
TheseresultsforNymphaealesindicatingre- ranthaceaearerelativelyyoung(i.e.,60–29Ma
centdiversificationinanancientlineageagree for Hedyosmum, 22–11 Ma for Chloranthus, and
withsimilarfindingsforthebasalangiosperms 18–9MaforAscarina;Zhang&Renner2003).
Chloranthaceae (Zhang & Renner 2003) and Similar results have been obtained for Illi-
Illicium (Illiciaceae; Morris et al. 2007). The cium (Illiciaceae). Morris et al. (2007) estimated
fossil record indicates clearly that Chloran- divergencetimeswithinIlliciumbyusingpenal-
thaceaerepresentoneoftheoldestangiosperm izedlikelihoodandmultiplecalibrationpoints.
lineages, with unequivocal reproductive struc- TheIlliciumcrowngroupappearstohavearisen