Integrating Molecular Phylogenetic and Paleobotanical Evidence on Origin of the Flower

Author(s): James A. Doyle

Source: International Journal of Plant Sciences, Vol. 169, No. 7 (September 2008), pp. 816-843
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/10.1086/589887 .
Accessed: 23/04/2013 19:29

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to
International Journal of Plant Sciences.

http://www.jstor.org

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
Int. J. Plant Sci. 169(7):816–843. 2008.
Ó 2008 by The University of Chicago. All rights reserved.
1058-5893/2008/16907-0002$15.00 DOI: 10.1086/589887

INTEGRATING MOLECULAR PHYLOGENETIC AND PALEOBOTANICAL

EVIDENCE ON ORIGIN OF THE FLOWER
James A. Doyle1
Department of Evolution and Ecology, University of California, Davis, California 95616, U.S.A.

Inferences on the origin of the angiosperm flower require consideration of other seed plants, especially fossils.
Molecular data favor a relationship of Gnetales to conifers rather than to angiosperms, and both alternatives are
equally parsimonious in terms of the morphological data set presented here. However, if molecular relationships
among extant taxa are accepted, morphology still associates glossopterids, Pentoxylon, Bennettitales, and
Caytonia with angiosperms. Bennettitales had flowerlike structures, but if Caytonia is sister to angiosperms,
aggregation of fertile parts probably occurred independently in Bennettitales and angiosperms. These results and
developmental genetic data are consistent with homology of the angiosperm bitegmic ovule with the cupule of
glossopterids and Caytonia, while the carpel could represent a leaf and a cupule-bearing axillary branch. Origin of
an adaxial cross zone could produce a uniovulate, ascidiate carpel, as in living basal angiosperms. Stamens may
represent similar units bearing two microsynangia. However, ovulate structures of Pentoxylon and Bennettitales
are more difficult to interpret, and any homologue of the carpel wall in Caytonia is unclear. Further progress may
require better understanding of homologies in known fossils and/or recognition of closer stem relatives of
angiosperms. A proposed Cretaceous stem relative, Archaefructus, is more likely a crown-group angiosperm
related to Hydatellaceae (Nymphaeales).

Keywords: angiosperms, seed plants, paleobotany, phylogeny, flower, ovule.

Online enhancement: appendix.

Introduction and formally designated Mesangiospermae by Cantino et al.

(2007). The main uncertainty is whether Amborella and Nym-
Reconstruction of the first flower and its origin—what the phaeales form two successive lines or a clade, with more data
angiosperm flower and its component parts came from—are favoring the former arrangement (Zanis et al. 2002; Soltis
two separate but related questions. In this article, I approach et al. 2005; Moore et al. 2007). Using parsimony to optimize
the second question with phylogenetic methods, integrating characters on a tree based on sequences of three genes and
results of both morphological and molecular analyses. morphology, Doyle and Endress (2000) presented a list of esti-
From a phylogenetic point of view, the original morphology mated ancestral states for 108 characters, such as vesselless
of the angiosperm flower can be addressed by asking where wood; simple, ovate leaves with pinnate venation and chlor-
the phylogenetic tree of angiosperms is rooted. Even in the ab- anthoid teeth; unilacunar two-trace nodes; undifferentiated
sence of morphological information on outgroups, if the an- perianth; stamens with adaxial microsporangia; several ascidi-
giosperm tree can be rooted with molecular data from other ate carpels; and fleshy, indehiscent fruits. Pollen characters
living plants, it is possible to optimize (plot) morphological were further refined and evaluated by Doyle (2005), floral
characters on the tree using parsimony (or more probabilistic phyllotaxis by Endress and Doyle (2007), and leaf architecture
likelihood methods) and thereby estimate ancestral states. by Doyle (2007). Ronse De Craene et al. (2003) presented a
This ‘‘top-down’’ approach (Bateman et al. 2006) is easiest if similar analysis of floral characters. It can be expected that
there are several lines with similar character states that branch such results will be improved by integration of the increasing
off sequentially at the base of the tree before the main radia- numbers of Early Cretaceous floral fossils, many described by
tion of the group, as appears to be the case in angiosperms. Friis et al. (2006).
This situation may help overcome the valid objection that sin- The second question—what sort of structures in earlier
gle low-diversity ‘‘basal’’ groups are not necessarily ‘‘primi- plants the flower came from, how they were transformed into
tive’’ in all their characters (Crisp and Cook 2005). the perianth, stamens, and carpels, and how they were assem-
Molecular analyses have provided highly consistent and sta- bled into the flower—can be approached phylogenetically by
tistically robust results concerning the rooting of the angio- asking what the closest outgroups of angiosperms are. Exami-
sperms and the identity of the first branches—Amborella, nation of the seed- and pollen-bearing organs in these plants
Nymphaeales, and Austrobaileyales—that are basal to the and optimization of the relevant characters on a cladogram
vast majority of angiosperms, often called ‘‘core angiosperms’’ may provide an improved picture of the precursor structures
1
for the flower at the closest outgroup node. This ‘‘bottom-up’’
E-mail: [email protected].
approach (Bateman et al. 2006) is therefore a question of seed
Manuscript received December 2007; revised manuscript received May 2008. plant phylogeny and the place of angiosperms in it, and because

816

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 817

of the highly relict and divergent nature of living gymnosperm pothesis proposed by Gaussen (1946) and adopted by Stebbins
taxa, it requires consideration of fossil as well as living plants. (1974) and Doyle (1978; fig. 1A), based on Caytonia. The fe-
Unfortunately, relationships among seed plant lines are far male structures of Caytonia (to which the name was first ap-
more uncertain and fraught with controversy than relation- plied; Thomas 1925) consisted of an axis bearing fleshy
ships within angiosperms. Much of what we thought we knew multiovulate structures termed cupules, which Harris (1940,
10 years ago about seed plant phylogeny, based on morphol- 1951) and Reymanówna (1973) interpreted as a rachis bear-
ogy, has been thrown into doubt by molecular analyses. Reso- ing two rows of leaflets with ovules on their adaxial surface.
lution of these problems requires integration of molecular, The cupules are folded tip to base in an anatropous fashion,
morphological, and fossil data in a phylogenetic framework. like the bitegmic ovules of angiosperms. They could be trans-
Fortunately, the better information on rooting and relation- formed into a structure like a bitegmic ovule by reduction
ships within angiosperms helps to improve the situation by from several ovules to one. In other words, what is conven-
clarifying ancestral states in angiosperms and thereby counter- tionally called an ovule in angiosperms is actually a cupule
acting the long-standing problem of comparing outgroups with containing one ovule. The most serious problem is explaining
angiosperm taxa that were incorrectly assumed to be primitive, the carpel: the structure bearing the cupules in Caytonia is
such as comparisons by von Wettstein (1907) and others be- narrow and rachislike, not a flat, leaflike structure with mar-
tween Gnetales and ‘‘Amentiferae’’ (mostly Fagales sensu APG ginal or adaxial cupules that could be folded to enclose the
1998, 2003). cupules. In Doyle (1978), I suggested that the carpel was de-
rived from the rachis, which was presumably larger relative
to the cupules early in development, by paedomorphic reten-
Hypotheses on Seed Plant and Angiosperm Relationships
tion of its juvenile proportions followed by folding to enclose
Before the advent of phylogenetic (cladistic) methods, bot- the cupules. This general scenario was adopted by Crane
anists had proposed a plethora of hypotheses on seed plant (1985a).
relationships and the origin of the flower. Often these at- Stebbins (1974) actually favored glossopterids over Cayto-
tempted to identify what group was ‘‘ancestral’’ to the angio- nia as the closest relatives of angiosperms. This concept was
sperms, a cladistically invalid concept: only a species in the anticipated by Melville (1963), on the basis of subsequently
geologic past can be ancestral to a clade, and this is seldom if refuted reconstructions of glossopterids, and elaborated by
ever distinguishable from a plesiomorphic sister group; ances- Retallack and Dilcher (1981; fig. 1B). Glossopterids had one
tral groups are paraphyletic and would be broken up into mono- or several cupules (also called sporophylls) attached to the mid-
phyletic units in a cladistic classification. A more appropriate rib of a more or less unmodified leaf, together called a fertiliger,
question is which other taxa are the sister group and the next- bract-sporophyll complex, or leaf-cupule complex (Schopf
closest outgroups of the angiosperms. Many precladistic hypoth- 1976; Pigg and Trivett 1994; Doyle 1996, 2006). Here I use
eses have been eliminated by both morphological and molecular the term ‘‘leaf-cupule complex’’ (as in Doyle 2006). The cupules
phylogenetic analyses (e.g., those postulating that angiosperms could be transformed into bitegmic ovules by reduction in ovule
are polyphyletic), whereas others have reappeared in various number, as in the Caytonia hypothesis, but the leaf/bract
cladistic studies. Here I concentrate on results of phylogenetic would provide a better precursor for the carpel: as Stebbins
analyses and restrict comments on precladistic hypotheses to (1974) emphasized, it would not have to be enlarged, just
those that have been supported by phylogenetic analyses. folded around the cupule. This hypothesis would predict that
The first phylogenetic analysis of both living and fossil seed angiosperm bitegmic ovules were originally attached to the
plants was that of Crane (1985a), who concluded that the closest midrib or adaxial base of the carpel rather than on or near its
relatives of angiosperms were living Gnetales and Mesozoic margins, the position that most earlier authors assumed was
Bennettitales and Pentoxylon (both with more or less ‘‘cycadlike’’ ancestral.
leaves). Crane pointed out that all these plants have fertile Subsequent morphological cladistic analyses differed in
parts grouped into flowerlike structures. His results implied many respects, but they all associated Gnetales and Bennetti-
that the flower originated long before the origin of angio- tales with angiosperms. Doyle and Donoghue (1986) obtained
sperms, in their common ancestor with Gnetales, Bennettitales, trees like those of Crane (1985a) but with Gnetales closer to
and Pentoxylon, which led to proposal of the name ‘‘antho- Bennettitales than to angiosperms. In the analysis of Nixon
phytes’’ (previously used as a synonym of angiosperms) for et al. (1994), angiosperms were nested within Gnetales as the
the whole clade (Crane 1985b; Doyle and Donoghue 1986). sister group of Welwitschia and Gnetum, with Chloranthus
This is essentially an updated version of the hypothesis of Arber and Ceratophyllum basal in some most parsimonious trees
and Parkin (1907), who postulated that angiosperms were re- and Casuarina and other Fagales basal in others. Anthophytes
lated to Gnetales and Bennettitales (Pentoxylon was unknown as a whole were associated with coniferophytes rather than
at the time) and derived from a common ancestor with a with Mesozoic seed ferns, a result also obtained by Rothwell
‘‘proanthostrobilus’’ consisting of an axis bearing perianth- and Serbet (1994). Trees with angiosperms nested in Gnetales
like sterile appendages, microsporophylls, and megasporo- were also obtained by Hickey and Taylor (1996). As discussed
phylls, in that order. by Doyle (1994) and Hickey and Taylor (1996), these results
Equally important for homologies of floral parts, however, would imply that angiosperm flowers were originally simple
are the taxa that Crane’s (1985a) analysis identified as out- and derived from simple flowers like those of Gnetales and
groups of the anthophytes—three groups of so-called Mesozoic that more complex flowers of other angiosperms were derived
seed ferns, namely, glossopterids (actually Permian), Cayto- either by multiplication and elaboration of parts or by group-
nia, and corystosperms. This result appeared to support a hy- ing of simple flowers into pseudanthia, as proposed by von

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
818 INTERNATIONAL JOURNAL OF PLANT SCIENCES

Fig. 1 A, Proposed homologies between ovulate structures of Caytonia and angiosperms, from Doyle (1978), following Gaussen (1946) and
Stebbins (1974). B, Proposed transformation of the ovulate leaf-cupule complex of glossopterids into an angiosperm carpel, from Retallack and
Dilcher (1981). See text for discussion.

Wettstein (1907). In the analysis of Doyle (1996), Gnetales back to the Late Carboniferous or Permian but Gnetales and
were the closest living relatives of angiosperms, but Caytonia angiosperms are not known until the Mesozoic–Late Triassic
was their sister group rather than being situated lower as an for probable stem relatives of Gnetales and the Early Creta-
outgroup of anthophytes. This would support the homology ceous for crown-group Gnetales and angiosperms (Crane 1988,
of the Caytonia cupule and the bitegmic ovule, but it would 1996; Doyle 1996). However, in an increasing number of multi-
challenge the hypothesis that flowers are homologous in an- gene analyses, Gnetales are related to conifers, as often proposed
giosperms, Bennettitales, and Gnetales, since the large sporo- before cladistics (Bailey 1949; Eames 1952; Bierhorst 1971;
phylls of Caytonia were probably not grouped into anything Doyle 1978), and angiosperms are the sister group of all living
resembling a flower. gymnosperms. In most such trees, as first shown by Bowe et al.
All these hypotheses have been thrown into doubt by mo- (2000), Gnetales are nested within conifers, linked with Pinaceae—
lecular analyses. Molecular studies say nothing directly about the so-called gnepine hypothesis.
the relationships of fossil taxa, but they do address the hypoth- The paucity of molecular support for a relationship be-
esis that Gnetales are the closest living relatives of angiosperms. tween Gnetales and angiosperms casts strong doubt on the
All analyses have indicated that Gnetales are monophyletic, anthophyte hypothesis, but the variety of molecular trees
thus refuting the hypothesis that angiosperms are nested in might be taken as a caution against molecular triumphalism.
Gnetales (Nixon et al. 1994; Hickey and Taylor 1996). A few However, more sophisticated methods are showing progress
analyses of ribosomal DNA have supported a relationship of toward sorting out which alternatives are most likely. First,
angiosperms and Gnetales, but only weakly (Hamby and Zim- the branches leading to and within Gnetales are unusually
mer 1992; Rydin et al. 2002), and analyses of other genes and long (as measured by numbers of nucleotide substitutions),
sequences of several genes combined, some from all three ge- suggesting that rates of molecular evolution accelerated in
nomes, give different trees (see reviews in Donoghue and this line. This suggests that trees in which Gnetales are basal
Doyle 2000; Magallón and Sanderson 2002; Burleigh and might be a result of long-branch attraction, specifically be-
Mathews 2004; Qiu et al. 2007). In some molecular trees, tween Gnetales and the stem lineage of seed plants, due to re-
Gnetales are the sister group of all other seed plants, and an- versals of seed plant synapomorphies on the stem lineage of
giosperms are linked with cycads, Ginkgo, and conifers. This Gnetales. Second, the signal favoring the basal position of
arrangement is implausible from a stratigraphic point of view, Gnetales is primarily in third-codon positions, which evolve
considering that cycads, ginkgophytes, and conifers extend faster because they usually do not affect amino acid composition.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 819

These conjectures have been supported by maximum likeli- (e.g., placement of Nymphaeales in the basal ‘‘ANITA’’ grade),
hood analyses (Sanderson et al. 2000; Magallón and Sander- but it did find several groups that had not been recognized
son 2002; Soltis et al. 2002; Burleigh and Mathews 2004), (Winteraceae with Canellaceae, Aristolochiaceae and Lactoris
which should theoretically be less sensitive than parsimony with Piperales) or had been differently constituted (Magno-
analysis to long-branch effects. Specifically, most data sets liales and Laurales) before cladistic studies. The recent discov-
that yield Gnetales-basal trees when analyzed with parsimony ery that Hydatellaceae are related to Nymphaeales rather than
give gnepine trees when analyzed with likelihood methods. to monocots was supported by both molecular and morpho-
Many molecular systematists think that this issue is now re- logical cladistic analyses (Saarela et al. 2007). Examples could
solved; this may be premature, but morphologists and paleo- be multiplied from other taxa.
botanists should consider seriously what the molecular results
imply if they are correct. It is in this spirit that I undertook
the analyses presented in this article.
Objectives and Approach
First, some general remarks about what gnepine trees do
not imply may be useful. They do not mean that angiosperms Most of this article is based on my recent phylogenetic
and gymnosperms were separately derived from Devonian analysis of seed plants (Doyle 2006), concentrating on its im-
‘‘progymnosperms’’ or that angiosperms extend back to the plications for the origin of the flower and its parts and dis-
Carboniferous. There could be any number of extinct seed cussing some aspects not previously treated. However, I also
plant lines attached to the stem lineage leading to living seed incorporate more recent advances in understanding of basal
plants and to the lineage leading to angiosperms. Such angiosperms and Gnetales, bring up to date the discussion of
plants would be called ‘‘gymnosperms’’ under the traditional developmental genetic evidence on homologies of the angio-
typological definition ‘‘plants with naked seeds,’’ which led sperm ovule, and evaluate the suggestion that the Early Cre-
Cantino et al. (2007) to propose the more restrictive name taceous fossil Archaefructus (Sun et al. 1998, 2002) is a stem
Acrogymnospermae for the molecular-based crown clade that relative of living angiosperms. Problems concerning the ori-
includes all living gymnosperms. In the phylogenetic nomen- gin of other angiosperm features, such as leaf morphology,
clature of Cantino et al. (2007), the ‘‘gymnospermous’’ stem pollen, and siphonogamy, are discussed in Doyle (2006).
relatives of angiosperms would be members of the stem-based One of my goals in Doyle (2006) was to evaluate whether
clade Pan-Angiospermae but not of Angiospermae, a name re- the conflict between molecular and morphological evidence on
stricted to the crown clade. Identification of angiosperm stem the relationships of angiosperms and Gnetales was as strong as
relatives is essential for understanding the origin of features it seems. I began with my previous seed plant data set (Doyle
such as the carpel and the flower; living acrogymnosperms are 1996), itself an update of that of Doyle and Donoghue (1986),
so divergent from angiosperms (and each other) in so many and modified it in several ways. Some modifications were based
characters that they can say very little about the origin of char- on doubts concerning homologies of characters that seemed to
acteristic angiosperm structures. link angiosperms and Gnetales (Donoghue and Doyle 2000).
The only way to identify fossil stem relatives of angio- For example, I had coded both taxa as having a tunica layer in
sperms is by consideration of their morphological characters, the apical meristem, but in angiosperms the tunica is two cells
preferably analyzed with cladistic methods. This is a daunting thick, whereas in Gnetales it is only one cell thick. This could
task in light of the apparent failure of morphological cladistics be evidence that the two conditions are not homologous. I
to infer the correct relationships of Gnetales. Scotland et al. therefore redefined this character as having three unordered
(2003) argued that such cases mean that morphological cla- states (tunica absent, one cell thick, and two cells thick), so it
distics is obsolete for phylogeny reconstruction (which would is equally parsimonious for the two tunica types to arise inde-
effectively preclude integration of fossils into phylogenies; pendently or one from the other. In addition, new similarities
Wiens 2004). However, this viewpoint is overly pessimistic. had been recognized between Gnetales and conifers. Carlquist
The fact that morphological cladistics was wrong about an- (1996) showed that Gnetales are like conifers and Ginkgo in
giosperms and Gnetales does not mean that it was wrong having a torus in the pits on their vessel walls. Conifers have
about everything or that it was no improvement over precla- a distinctly tiered stage in embryo development, which I had
distic reasoning. In fact, morphological cladistics provided an- scored as absent in Gnetales, but actually, embryos of Gne-
swers to many previously controversial questions that have tales are more or less tiered at a comparable stage (Martens
been confirmed by molecular data. Before cladistics, it was of- 1971; Singh 1978). Reassessments of characters that are more
ten suggested that coniferophytes and other seed plants (cy- directly related to the flower problem, especially potential ho-
cadophytes) were separately derived from progymnosperms, mologies of different types of cupules and the angiosperm bi-
but both morphological and molecular analyses indicate that tegmic ovule, are considered in the ‘‘Discussion.’’
coniferophytes are nested within other seed plants. It was Other changes in Doyle (2006) were related to molecular
widely claimed that angiosperms, Gnetales, and conifers were results concerning the rooting of the angiosperms. The taxon
polyphyletic (with Taxaceae not related to other conifers), but sampling of angiosperms in Doyle (1996) included two Mag-
both kinds of analysis agree that these taxa are monophyletic noliales (now placed well above the base of the angiosperms—
(if conifers are defined so as to include Gnetales if the latter seven nodes higher in the combined analysis of Doyle and En-
are nested within them, as in the definition of Cantino et al. dress [2000]) but only two members of the basal ANITA grade,
2007). The morphological analysis of Doyle and Endress not including Amborella. In Doyle (2006), I replaced these with
(2000) failed to recover some relationships inferred from mo- a more up-to-date sampling of ANITA groups and mesangio-
lecular data and morphological and molecular data combined sperms.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
820 INTERNATIONAL JOURNAL OF PLANT SCIENCES

Here I also add Hydatellaceae, formerly thought to be poa- have been autapomorphies of core Nymphaeales in Doyle
lian monocots (placed in Centrolepidaceae until studies by (2006) but also occur in Hydatellaceae: protoxylem lacunae,
Hamann [1975, 1976]) but now linked with Nymphaeales absence of vascular cambium, boat-shaped pollen, and oper-
(Saarela et al. 2007). The reproductive structures of these min- culum in the seed. A new character is presence or absence of
ute aquatics appear to be inflorescences of unisexual flowers floral bracts, which are absent in Archaefructus (as stressed by
consisting of one stamen or one carpel. Except for their linear Sun et al. [2002]) and Hydatellaceae but present in all other
leaves, most morphological features of Hydatellaceae resemble angiosperms in this data set.
those of Nymphaeales rather than those of Poales: the sto- Other changes are reorganizations of characters and states
mata are anomocytic rather than paracytic, the pollen is mono- based on unpublished work of J. A. Doyle and P. K. Endress.
sulcate rather than porate, and the seed has a minute embryo, Lack of fibers or sclerenchyma in the pericyclic region was
starchy perisperm, a palisade exotesta, and an operculum formed treated as a state of the pericycle character by Doyle and En-
by enlargement of cells of the inner integument, as in Nym- dress (2000) but eliminated in Doyle (2006) because it was an
phaeales. The carpels are ascidiate, and the embryo sac is autapomorphy of core Nymphaeales. Because this condition
four-nucleate (Hamann 1975; Friedman 2008), as in most also occurs in Hydatellaceae, I have reinstated it but as a sepa-
other members of the ANITA grade (Williams and Friedman rate character, on the presumption that loss of sclerenchyma
2004). Because Stevens (2007) has enlarged Nymphaeales to deserves recognition as a major change independent of transi-
include Hydatellaceae, I henceforth refer to Nymphaeales in tions between the two patterns of fiber and sclerenchyma dis-
the older sense of Cabombaceae and Nymphaeaceae as ‘‘core tribution. I split leaf phyllotaxis, formerly a single three-state
Nymphaeales.’’ character, into two binary characters, one for alternate versus
Consideration of Archaefructus is an outgrowth of a collab- opposite and the other for presence of at least some axes with
oration with Peter Endress (Doyle and Endress 2007) on ex- distichous phyllotaxis, on the presumption that the distinction
pansion and revision of the morphological data set of Doyle between opposite and alternate phyllotaxis is more fundamen-
and Endress (2000) and its use to evaluate the position of Early tal than variation between spiral and partly or wholly distichous.
Cretaceous fossils in the phylogeny of living angiosperms. Pre- I also split the stigma surface character, previously defined as
liminary results indicate that the most parsimonious position papillae (0) unicellular only (or stigma smooth), (1) some or
for Archaefructus within angiosperms is in Nymphaeales, all uniseriate pluricellular, or (2) some or all pluriseriate pluri-
linked with Hydatellaceae (Doyle and Endress 2007; Endress cellular (including multicellular protuberances). Structures de-
and Doyle, forthcoming), but because this analysis includes scribed as pluriseriate papillae intergrade with protuberances
only angiosperms, it does not address the hypothesis that that are not obviously papillar, and uniseriate papillae and
Archaefructus is a stem relative of angiosperms. Inclusion of protuberances occur together in Amborella and Trimenia (En-
Archaefructus in the present study is intended to test this hy- dress and Igersheim 1997), suggesting that they represent in-
pothesis. Because the morphology of the fertile shoot of Ar- dependent transformations. I have also redefined the uni- and
chaefructus is controversial, I have analyzed the position of pluricellular papilla types to refer to the emergent portion of
Archaefructus using two alternative sets of character scorings, the papilla, which appears to be a more consistent distinction.
one assuming that the shoot is a bisexual flower or preflower Other changes represent adjustments to the reduced taxon
(Sun et al. 2002) and the other assuming that it is an inflores- sampling in angiosperms, such as simplification of the carpel
cence of reduced unisexual flowers (Friis et al. 2003). sealing and fruit wall characters.
I made several changes in floral characters to reduce prob-
lems and bring definitions into line with analyses of Eklund
Material and Methods et al. (2004), Endress and Doyle (2007), and J. A. Doyle and
P. K. Endress (unpublished data). In Doyle (2006), I recognized
The data set is presented in the appendix in the online edi- four perianth characters: phyllotaxis, number of whorls, merism
tion, with a list of characters and their states and changes (¼merosity), and differentiation of the outer cycle. These are
from Doyle (2006). For documentation and reasoning behind all informative with a larger taxon sample but are less so with
character definitions and taxon scorings that have not been the present reduced number of angiosperms and not entirely
modified, readers are referred to Doyle (2006). Some modifi- independent of each other. Thus, irregular merism overlaps with
cations and decisions on scoring of Hydatellaceae and Ar- spiral phyllotaxis. If irregular is eliminated as a state of the
chaefructus are treated in the appendix, but those requiring merism character and spiral taxa are scored as unknown (for
more explanation or argumentation are discussed in this sec- inapplicable), merism becomes uninformative: among whorled
tion. As in Doyle (2006), most characters concerning inflores- taxa in the data set, the only departures from trimery are dimery
cence and floral organization and the morphology of stamens, in Winteraceae and polymery in the Barclaya-Nymphaeoideae
carpels, and whole bitegmic ovules and seeds derived from clade nested in core Nymphaeales. Similarly, number of whorls
them have been scored as unknown outside the angiosperms is problematic in spiral taxa, and among whorled taxa the
because they cannot be applied without making questionable only deviations from having more than two whorls and no
assumptions on homology. perianth occur in some but not all members of taxa (two
Several characters have been added in order to express previ- whorls in Cabombaceae but more in Nymphaeaceae, one in
ously uninformative features of core Nymphaeales, Hydatellaceae, Hedyosmum but none in other Chloranthaceae, and one or
and Archaefructus or as a result of ongoing reevaluation of two in Asaroideae). I therefore replaced these three charac-
characters by J. A. Doyle and P. K. Endress (unpublished data). ters with one character for presence or absence of perianth
Some are characters from Doyle and Endress (2000) that would and one for spiral or whorled phyllotaxis. The perianth dif-

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 821

ferentiation character of Doyle (2006), which distinguished rophylls borne on axes with no other appendages (Townrow
outer cycle not differentiated or forming a continuum with 1960; Kerp 1988). Other taxa have fertile shoots with basal
inner parts from sepaloid, was too closely tied to phyllotaxis, cataphylls or other sterile appendages that might be compared
since spiral perianths tend to show a continuity of forms and with a perianth, such as fertile branches of corystosperms
whorled ones more abrupt transitions (Endress and Doyle (Axsmith et al. 2000; Klavins et al. 2002), axillary fertile
2007; Endress 2008). I therefore modified this character to short shoots of cordaites and Paleozoic conifers such as Le-
distinguish all perianth parts sepaloid from outer sepaloid bachia and Emporia (Florin 1951; Mapes and Rothwell 1984),
and inner petaloid. Because this distinction can be applied male and possibly female ‘‘flowers’’ of Pentoxylon (Bose et al.
only to flowers with more than one perianth cycle, I scored 1985), and ‘‘flowers’’ of most but not all Bennettitales (Crane
Chloranthaceae (one cycle or none) and Asaroideae (one cy- 1985a, 1988). However, there are other fossil seed plants that
cle or two) as 0/1. lack strobili or flowers, to which this distinction is difficult
Similarly, I modified androecium characters by adding a or impossible to apply. In some, the seeds were borne on nor-
character for one stamen per flower, as in Chloranthaceae (re- mal leaves (medullosans and Callistophyton; Rothwell 1981);
constructed as ancestral in the family by Eklund et al. [2004]) in others, they were borne on appendages that differed from
and Hydatellaceae, and deleting androecium merism. If irreg- vegetative leaves but were probably not grouped into special
ular merism is eliminated as redundant with spiral phyllo- strobili or flowers (glossopterids and Caytonia). Cycads are a
taxis, the only deviation from the trimerous state is polymery special case. Female plants of the basal genus Cycas produce
in Winteraceae. regular leaves, cataphylls, and megasporophylls in alternating
Outside angiosperms, the only changes in scoring are zones along the stem, but male Cycas plants and all plants of
based on the reinterpretation of reproductive structures in the other genera produce strobili, with sterile cataphylls on
Gnetales by Mundry and Stützel (2004). In a developmental the peduncle in Cycas, Dioon, Encephalartos, Lepidozamia,
study, these authors concluded that the cup bearing six mi- and Macrozamia (Hermsen et al. 2006; D. W. Stevenson, per-
crosynangia in the male ‘‘flower’’ of Welwitschia, previously sonal communication), which are relatively basal, according
interpreted as either six simple sporophylls or two pinnate to molecular analyses (Rai et al. 2003; Chaw et al. 2005).
sporophylls, actually consists of two lateral strobili, each I expressed these distinctions with three unordered states:
with an apical meristem and three simple sporophylls. They (0) fertile appendages not borne on special shoots, (1) special
interpreted the male ‘‘flower’’ of Ephedra as consisting of fertile shoots without sterile appendages (no perianth), and
two similar units with four simple sporophylls, and they (2) special fertile shoots with sterile appendages (perianth). I
proposed plausible homologies with parts in the female scored extant conifers as unknown, on the assumption that
‘‘flowers’’ of both genera. I have therefore rescored microspo- their cone scales are derived from whole fertile short shoots,
rophyll morphology as simple (1) rather than either pinnate/ and Bennettitales as 1/2. To test implications of the assump-
paddle shaped or simple (0/1). Mundry and Stützel argued tion that glossopterids and Caytonia did not have flowers or
that their interpretation may also apply to Gnetum, but be- strobili, I scored them as 0. I considered two alternative scor-
cause this has not been confirmed, I have scored this charac- ings of cycads, to take into account the different conditions
ter in Gnetum as unknown. This reinterpretation also affects in male and female plants of Cycas. One (0/2) assumes that
the microsporophyll fusion character, previously scored as it would be most appropriate to score Cycas as having the
fused. The two lateral strobili are fused basally into a cup, nonstrobilar condition of the female plants, and the other (2)
but because it is unclear whether the individual microsporo- assumes the strobilar condition of the male plants. Although
phylls are fused to each other or borne separately around the the interpretation of Ephedra and Welwitschia by Mundry
apex, I have rescored this character as unknown in all Gne- and Stützel (2004) implies that the male strobili lack sterile
tales. Under the Mundry and Stützel interpretation, the num- appendages, I scored Gnetales as having special fertile shoots
ber of sporangia per microsporophyll in Ephedra is usually with sterile appendages, on the basis of the presence of an
two, as in Pinaceae and Podocarpaceae, not more than two, outer envelope around the ovule in the female ‘‘flowers.’’
as scored in Doyle (2006). However, because some species Scoring of Hydatellaceae generally follows that by Saarela
have more than two microsporangia (Martens 1971; Hufford et al. (2007), who added Hydatellaceae and two poalian mono-
1996) and the ancestral number is not established (S. Ickert- cots to the morphological data set of Doyle and Endress (2000),
Bond, personal communication), I have rescored Ephedra as along with characters for porate pollen (not represented in
uncertain (0/1) for this character. the present data set) and four- versus eight-nucleate embryo sac
Although I did not score presence or absence of a perianth (already used by Doyle [2006]; confirmed by Friedman [2008]).
outside angiosperms, I attempted to address the problem of Data on general morphological characters were from Hamann
its origin by defining a related character applicable across (1975, 1976); those on vegetative anatomy were from Cutler
seed plants, which I did not use in the actual phylogenetic (1969), Cheadle and Kosakai (1975), and Behnke (2000);
analyses because of problems in homology of the fertile or- those on pollen were from Linder and Ferguson (1985); and
gans (e.g., sporophylls in some taxa may correspond to cu- those on seedlings were from Cooke (1983). Unless otherwise
pules in others). If specialized, determinate, fertile shoots in indicated, scorings of Saarela et al. (2007) are followed for
other seed plants are considered potentially homologous with those characters that occur in the present data set. Scoring of
angiosperm flowers, some taxa have shoots bearing only spo- seed plant characters that were not included by Saarela et al.
rophylls that might be compared with flowers that lack a peri- is based on the same references. Rudall et al. (2007) provided
anth, namely, Ginkgoales and probably peltasperms (Autunia valuable new data on several characters, for example, show-
and Peltaspermum), which had paddle-shaped or peltate spo- ing that the inflorescences are reduced thyrses (indeterminate

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
822 INTERNATIONAL JOURNAL OF PLANT SCIENCES

axes bearing lateral cymes). Many characters are scored as follicles, but they did not actually report dehiscence. The
unknown in Hydatellaceae because the plant body is so re- seeds appear to have a palisade exotesta; Sun et al. (1998,
duced that state distinctions are inapplicable or questionable 2002) described the surface as consisting of epidermal cells
(vegetative buds, short shoots, nodal anatomy, leaf teeth, with cutinized anticlinal and periclinal walls.
laminar venation, and characters of secondary xylem and Searches for most parsimonious trees used PAUP (Swofford
phloem). 1990), with 100 heuristic search replicates, stepwise random
Scoring of Archaefructus is based on descriptions of Ar- addition of taxa, and tree bisection reconnection branch
chaefructus liaoningensis and Archaefructus sinensis by Sun swapping. In all cases, the Late Devonian seed fern Elkinsia
et al. (1998, 2002), with confirmation of paired stamens and (Serbet and Rothwell 1992) was specified as outgroup. The
carpels by Friis et al. (2003). One of the two alternative scor- relative parsimony of alternative arrangements was evaluated
ings (Archaefructus inf), following Friis et al. (2003), as- by searching for trees consistent or inconsistent with appropri-
sumes that the fertile shoot was a raceme of male and female ate constraint trees or by moving taxa manually with MacClade
flowers consisting of usually two stamens and one or two (Maddison and Maddison 2003). Many of these analyses in-
carpels. The other (Archaefructus flo), following Sun et al. volved ‘‘backbone constraint’’ trees of extant taxa and Elkinsia
(2002), assumes that the shoot was a bisexual flower or pre- only; fossil taxa attach to this backbone at the most parsimo-
flower with paired stamens below and carpels above. Sun nious location in terms of their morphology. Bootstrap analyses
et al. (2002) speculated that the apparent paired stamens were performed with PAUP, using 1000 bootstrap replicates,
may actually be single, branched sporophylls, simplified from with each replicate consisting of one heuristic search with
the pinnately branched sporophylls of many fossil seed plants closest taxon addition sequence, holding five trees at each
but less so than stamens of living angiosperms. To acknowl- step, and retaining 100 trees per replicate but allowing branch
edge and test this intriguing hypothesis, I have scored micro- swapping to continue beyond this limit in order to increase
sporophylls of Archaefructus as pinnate under the flower the probability of finding shorter trees. Decay analyses
interpretation. Morphology of Archaefructus eoflora (Ji et al. (Bremer support; Bremer 1988; Donoghue et al. 1992) were
2004), which may represent either a smaller species or a youn- conducted by searching for trees that were one, two, three, or
ger stage of A. sinensis, is generally consistent with that of A. (when memory was sufficient) four steps longer than the
liaoningensis and A. sinensis. Ji et al. (2004) interpreted seeds shortest trees and observing which clades disappeared in the
of A. eoflora as orthotropous, but the orientation is not clear resulting strict consensus. Decay values for the few groups
in their illustrations. Figure 2C of Sun et al. (1998), showing that remained after these searches were determined by search-
the end of a seed of A. liaoningensis, is more suggestive of an ing for shortest trees not consistent with a constraint tree in
anatropous ovule. which the group formed a clade but other relationships were
In their cladistic analysis, Sun et al. (2002) scored the finely unresolved.
dissected leaves of Archaefructus as having one order of dichot- I performed three main sets of analyses. The first were un-
omous laminar venation, as in cycads (except Bowenia) and constrained parsimony analyses of the data set without Ar-
Ginkgo, which, along with Ephedra and Pinus, were the out- chaefructus. Second, to test the implications of molecular
groups in their data set. Friis et al. (2003) compared the ternate trees for the angiosperm question, were constrained analyses
organization with that of Ranunculales and the Early Creta- with living taxa fixed in the arrangement seen in trees of the
ceous fossil Vitiphyllum, but in the present data set it is most gnepine type (Bowe et al. 2000). Third were analyses includ-
similar to the pinnately compound condition of Paleozoic ing Archaefructus, with extant taxa constrained to the molec-
seed ferns. To preserve the spirit of the comparisons of Sun ular arrangement. The relative parsimony of other positions
et al. (2002) as well as possible in terms of the present data of Archaefructus was evaluated by searching for trees that
set and to test their implications, I have scored the leaves as were one to three or four steps less parsimonious, filtering
pinnately compound with one order of open dichotomous the resulting trees with constraint trees to determine different
laminar venation. sorts of relationships obtained, or by moving Archaefructus
Judging from the figures of Sun et al. (2002), the stamens manually to other positions with MacClade (Maddison and
of Archaefructus may have a typical angiosperm morphology, Maddison 2003). I repeated most analyses without Hydatellaceae
but this is not well enough demonstrated to score microspo- and with appropriately modified constraints to evaluate the
rangial position, number, fusion, and dehiscence. Although effects of addition of this group. I used MacClade (Maddison
Friis et al. (2003) questioned whether the bodies that Sun and Maddison 2003) to reconstruct character evolution by
et al. (2002) described as pollen were in fact pollen grains, parsimony optimization of characters on trees and to iden-
because of their irregular size and shape, I have provisionally tify characters that support various clades and alternative
assumed that they are pollen and scored them, on the basis arrangements of taxa.
of the most convincing figure (fig. 2F) in Sun et al. (2002), as
boat shaped and monosulcate, with a continuous tectum and Results
unknown infratectal structure. It is unclear whether the ovules
were anatropous or orthotropous, so I have scored this and The unconstrained analysis of the new data set without
related characters as unknown. Sun et al. (1998, 2002) de- Archaefructus yielded 40 most parsimonious trees of 346
scribed the carpels as conduplicate (¼plicate), but in extant steps (strict consensus in fig. 2, with bootstrap and decay
carpels of similar appearance this cannot be determined with- values). These form two islands, of 24 and 16 trees, that dif-
out developmental or anatomical evidence (Friis et al. 2003; fer most significantly in the relationships of angiosperms and
Endress 2005). Sun et al. (1998, 2002) described the fruits as Gnetales.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 823

Fig. 2 Strict consensus of 40 most parsimonious trees of 346 steps obtained from the unconstrained analysis, with decay and bootstrap support
values for nodes. Numbered brackets indicate clades with bootstrap values of >50% that did not occur in the consensus of most parsimonious
trees (e.g., Caytonia plus angiosperms). GNET ¼ Gnetales.

Trees of the first island (consensus in fig. 3A) are of the an- In the second island, relationships among angiosperms in the
thophyte type, with Pentoxylon, Gnetales, Bennettitales, and taxon set are the same as those found when Hydatellaceae
angiosperms linked with Caytonia, glossopterids, and the were included (fig. 3C).
whole combined clade nested in coniferophytes. Pentoxylon The analysis with extant taxa constrained to the molecular
and Gnetales form either two successive branches or a clade. (gnepine) arrangement resulted in 18 trees of 364 steps. In the
There are two different rootings of the angiosperms. In all strict consensus (fig. 4), relationships at the base of the crown
trees, Winteraceae are linked with a ‘‘paleoherb’’ clade consist- clade that includes all living seed plants are poorly resolved.
ing of core Nymphaeales plus Hydatellaceae (Nymphaeales) This is a function of the existence of two islands, of six and
and Saururaceae plus Asaroideae (Piperales). In 16 trees, the 12 trees, and the ‘‘jumping’’ of Callistophyton between very
Winteraceae-paleoherb clade is sister to the remaining angio- different positions in trees of the second island. In the first is-
sperms, with Illicium plus Schisandraceae, Austrobaileya, Tri- land (representative tree in fig. 5A, with leaf organization op-
menia, Chloranthaceae, and Amborella branching pectinately timized on the branches), the outgroups of angiosperms are
in that order. In the other eight trees, Amborella is basal, fol- the same as in the unconstrained trees where Gnetales were
lowed pectinately by Chloranthaceae, Trimenia, Austrobaileya, nested in conifers (fig. 3C): Caytonia is the sister group of
Illicium plus Schisandraceae, Winteraceae, and the paleoherb the angiosperms, Bennettitales are the second outgroup, and
clade (fig. 3B). a clade consisting of glossopterids and Pentoxylon is basal in
In trees of the second island (consensus in fig. 3C), Gnetales the pan-angiosperms. All variation among trees concerns the
are nested within conifers, as in many molecular analyses, but arrangement of cordaites and the platyspermic seed fern taxa
are linked either with Araucariaceae or with Araucariaceae, Callistophyton, corystosperms, Autunia, and Peltaspermum
Cupressaceae (including Taxodiaceae), Cephalotaxus, and Taxa- within the acrogymnosperm clade. In the second island (fig.
ceae rather than with Pinaceae. The closest outgroups of angio- 5B), relationships among the platyspermic seed ferns are still
sperms are glossopterids, Pentoxylon, Bennettitales, and Caytonia, more unstable, with Callistophyton attached below the com-
with Caytonia the sister group of angiosperms. Nymphaeales mon ancestor of all living seed plants in five trees and nested
(including Hydatellaceae) and Piperales are either two successive among the other taxa in the other seven. However, all trees
basal branches or a basal clade, and Winteraceae are sister to the in this island have the same arrangement at the base of the acro-
remaining taxa. gymnosperms, with cycads, Pentoxylon, and glossopterids di-
An unconstrained analysis with Hydatellaceae removed verging pectinately in that order. The comparable constrained
yielded 56 trees of 330 steps, which represent two islands of analysis without Hydatellaceae yielded 18 trees of 349 steps,
40 and 16 trees. Outgroup relationships in the two islands are with the same outgroup relationships found when Hydatella-
the same as those found when Hydatellaceae were included ceae were included (figs. 4, 5).
(fig. 3A, 3C). In the first island, core Nymphaeales and Piperales The analyses including Archaefructus, with relationships
form a paleoherb clade, but Winteraceae are not always among living taxa constrained to the molecular arrangement,
linked with this clade; again, Amborella is basal in eight trees. support the view that Archaefructus is a crown-group angiosperm

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM
All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 825

Fig. 4 Strict consensus of 18 most parsimonious trees of 364 steps obtained from the analysis with the arrangement of living taxa fixed to a
molecular backbone constraint tree, with decay and bootstrap support values. GNET ¼ Gnetales.

related to Nymphaeales. However, the relative parsimony of Archaefructus is sister to living angiosperms are three steps
alternative relationships varies with the two interpretations of longer than the most parsimonious trees, but several positions
the reproductive structures. Characters supporting these re- lower in the tree are only two steps longer. When Hydatellaceae
sults are most conveniently presented in the ‘‘Discussion.’’ are removed, Archaefructus is associated with core Nymphaeales
When Archaefructus was scored as having an inflorescence in the shortest trees, but trees in which it is sister to living an-
of unisexual flowers, the analysis yielded 18 trees of 371 giosperms are only two steps longer.
steps, in which Archaefructus is the sister group of Hydatel-
laceae (fig. 6). Otherwise, these trees are identical to those
found without Archaefructus (figs. 4, 5). The next-best posi-
Discussion
tions for Archaefructus are as sister to core Nymphaeales,
which is one step less parsimonious (372 steps), and as sister
General Phylogenetic Implications
to both Hydatellaceae and core Nymphaeales, which is two
steps less parsimonious. A position as sister to all living an- These results differ most significantly from those of Doyle
giosperms is five steps less parsimonious (376 steps). Several (2006) in that trees with Gnetales nested in ‘‘anthophytes’’
positions near the base of crown-group seed plants (which and in conifers are equally parsimonious, whereas previously,
would imply that the seed-containing structures are conver- trees of the former type were favored by one step. This
gent with angiosperm carpels) have the same score (fig. 6). change is primarily a result of scoring Ephedra and Welwit-
When Hydatellaceae are removed from the analysis, it is schia as having simple microsporophylls rather than either
most parsimonious to link Archaefructus with core Nym- pinnate to paddle shaped or simple, following the interpreta-
phaeales, but a position as the sister group of living angio- tion of Mundry and Stützel (2004). As a result, nesting Gnetales
sperms is only two steps less parsimonious. among anthophytes (which have more complex microsporo-
The analysis with Archaefructus scored as having a bisexual phylls) adds a step that did not occur when Gnetales were
flower or preflower yielded 36 trees of 371 steps, in which scored as uncertain. Other changes in character definition
Archaefructus is the sister group of either Hydatellaceae or and scoring had no evident effect on inferred outgroup rela-
core Nymphaeales (fig. 7). Again, relationships outside the tionships, except for the appearance, in the first island, of
angiosperms are not affected. The shortest trees in which trees in which Pentoxylon and Gnetales form a clade rather

Fig. 3 Strict consensus trees for the two islands of most parsimonious trees found in the unconstrained analysis (fig. 2). A, Consensus of the 24
trees in the first island, with Gnetales (GNET) in the ‘‘anthophyte’’ clade. B, One of two most parsimonious topologies of angiosperms in the first
island, with Amborella basal. C, Consensus of the 16 trees in the second island, with Gnetales in conifers.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 Fig. 5 A, Representative tree of island 1 from the analysis with the molecular backbone (fig. 4), with nodes not found in all trees of the island
indicated by arrows and with shading of branches showing the inferred evolution of leaf organization (character 31; see appendix in the online
edition). B, Representative tree of island 2 from the same analysis. GNET ¼ Gnetales.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 827

Fig. 6 Representative most parsimonious tree from the analysis with Archaefructus added, with its fertile shoot interpreted as an inflorescence
of unisexual flowers and with the arrangement of living taxa fixed to the molecular backbone. Positions of Archaefructus up to five steps less
parsimonious than the best position (assuming the arrangement of other taxa shown) are indicated with MP ¼ most parsimonious, þ 1 ¼ one step
longer than most parsimonious, þ 2 ¼ two steps longer, etc. GNET ¼ Gnetales.

than two successive branches. Addition of Hydatellaceae had relevant to the relationships of angiosperms and Gnetales and
no effect on outgroup relationships. did not modify the obsolete taxon sampling of angiosperms.
These results strengthen the view expressed in Doyle (2006) As in Doyle (2006), another major result is that trees with
that morphology does not strongly contradict molecular data Gnetales nested in conifers, found both with and without
on the position of Gnetales, as previously believed, but rather backbone constraints (figs. 3C, 4, 5), identify the same previ-
is deeply ambiguous: Gnetales share morphological advances ously implicated fossil taxa as the most probable relatives of
both with Bennettitales and angiosperms and with conifers, angiosperms (¼other Pan-Angiospermae of Cantino et al.
half of which are homoplastic. The nearly complete lack of 2007): glossopterids, Pentoxylon, Bennettitales, and Caytonia.
resolution among crown-group seed plants in the consensus The resulting pan-angiosperm clade corresponds to the ‘‘glos-
(fig. 2) does not mean that morphology is ambiguous for all sophytes’’ of Doyle (1996), except for the elimination of
aspects of seed plant phylogeny. Inspection of the consensus Gnetales. This result is surprisingly insensitive to relation-
trees of the two islands (fig. 3A, 3C) shows that most of the ships within angiosperms, at least with the several rootings
lack of resolution is due to jumping of Gnetales between their inferred from this data set or specified by the molecular back-
two most parsimonious positions; many other relationships bone tree. In theory, correct identification of outgroups might
are consistent in the two kinds of trees. Other reflections on require correct rooting of the angiosperms, as well as vice
the similar but less complete shift in relative parsimony of the versa (as argued by Doyle et al. [1994] and Bateman et al. [2006]),
two positions of Gnetales between Doyle (1996) and Doyle but this problem may be less severe in practice. However, the
(2006) were presented in the latter article. In an analysis that low statistical support values for the relationships obtained
also began with the Doyle (1996) data set and improved the (a common situation with fossils, because of the large propor-
sampling of taxa and characters among basal seed plants, Hilton tion of missing data) indicate that the inferences made here on
and Bateman (2006) found a greater parsimony difference be- outgroups should be viewed as only a marginally favored hy-
tween trees of the two kinds, but they made fewer changes pothesis.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
828 INTERNATIONAL JOURNAL OF PLANT SCIENCES

Fig. 7 Same as fig. 6, but with the fertile shoot of Archaefructus interpreted as a bisexual flower or preflower. Positions of Archaefructus up to
three steps less parsimonious than the best position are indicated as in fig. 6. GNET ¼ Gnetales.

Another intriguing change from Doyle (2006) occurred in analysis of Doyle and Endress (2000) strongly supported the
angiosperms, in the first island from the unconstrained analy- molecular arrangement, indicating that the contrasting morpho-
sis, where the number of most parsimonious trees increased logical results are due to homoplasy: similarities between Amborella
from eight to 24 (fig. 3A). Trees in which Nymphaeales are and Chloranthaceae (such as a single pendent, orthotropous
sister to all other angiosperms (as in all unconstrained trees ovule) and between Nymphaeales and Piperales (palmately
of Doyle 2006) disappear (Nymphaeales s. l. are always veined leaves and other features that may be functionally re-
linked with Piperales and Winteraceae), but eight trees ap- lated to herbaceous habit).
pear in which Amborella is basal, as in molecular analyses In those unconstrained trees in which Gnetales are nested
(fig. 3B). This rooting has not been seen in previous purely in conifers (fig. 3C), relationships within angiosperms changed
morphological analyses of seed plants. Plesiomorphic states only slightly from those found in the most comparable analy-
of Amborella that support this position are lack of vessels, al- sis of Doyle (2006), where conifers and Gnetales were con-
ternate phyllotaxis, lack of oil cells, and continuous tectum. strained to form a clade but other relationships were left
Several other aspects of the unconstrained trees with Am- unspecified. In that analysis, (core) Nymphaeales were basal,
borella basal conflict with molecular results. As in the mor- followed by Piperales, and Winteraceae were either basal in
phological analyses of Doyle and Endress (2000) and Eklund the remaining taxa or linked with Piperales, but in the present
et al. (2004), in which Amborella was specified a priori as analysis, Nymphaeales and Piperales form either two succes-
outgroup, Chloranthaceae are basal in the remaining angio- sive branches or a basal clade, and Winteraceae are sister to
sperms, followed by Trimenia, and Nymphaeales are linked the remaining taxa.
with Piperales. In contrast, molecular analyses exclude Chlo- Bateman et al. (2006) cited Doyle (2006) as finding that
ranthaceae from the basal ANITA grade and place them in constraining other seed plants into a molecular arrangement
the strongly supported mesangiosperm clade (Cantino et al. resulted in a shift toward angiosperm relationships found in
2007), which contains all other angiosperms, and place molecular studies, but this perception was apparently due to
Nymphaeales above or with Amborella in the ANITA grade. confusion with the analysis in which relationships in angio-
In both cases, the combined morphological and three-gene sperms were constrained to a molecular tree. In fact, the oppo-

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 829

site is true. With the present data set, the most parsimonious trees of the first island (fig. 5A), the fernlike, pinnately com-
trees with Amborella basal in angiosperms (as in molecular pound leaves of Callistophyton, corystosperms, and pelta-
analyses) are of the anthophyte type (fig. 3B), which conflicts sperms can be either homoplastic or homologous with those
with molecular data, whereas trees constrained to agree with of earlier seed ferns. In the latter case, simple pinnate leaves
molecular studies in having Amborella basal and Gnetales in would be derived twice, in cycads and the pan-angiosperm
conifers (not shown) are three steps longer (349 steps). line. This scenario is favored if one assumes that simplification
In the following discussion, I concentrate on the constrained more often produces similar forms than does secondary elabo-
analyses (figs. 4, 5). Trees from the unconstrained analyses ration.
have many interesting features (some discussed in Doyle
2006), but, as already noted, combined analyses show that the
General Floral Organization
conflicting morphological relationships in angiosperms are
strongly overruled by molecular data (Doyle and Endress The distinction between strobili and flowers is rather arbi-
2000). Molecular relationships within conifers (not consider- trary (see review in Bateman et al. 2006); both are special,
ing Gnetales) also have strong statistical support (Quinn et al. determinate short shoots bearing closely spaced fertile ap-
2002), including the relationship of Podocarpaceae with pendages, with or without associated sterile appendages.
Araucariaceae rather than Pinaceae, the only difference from Both can be distinguished from cases in which fertile append-
the morphological result. Bateman et al. (2006) described this ages are borne on shoots that are not strongly differentiated
procedure as accepting an unparsimonious tree, but although from vegetative shoots. I use ‘‘strobilus’’ for specialized fertile
the constrained tree is less parsimonious in terms of morphol- shoots in which the axis is relatively elongate at the time of
ogy, the Doyle and Endress (2000) study indicates that the pollination and ‘‘flower’’ for shoots in which the axis is much
angiosperm portion is more parsimonious in terms of mor- shorter relative to the whole structure (Doyle 1994). How-
phology and molecular data combined. The weakest aspect of ever, some have used the term ‘‘flower’’ for both types of
the molecular constraint tree, the position of cycads at the structure, for example, Florin (1951) in describing the fertile
base of the acrogymnosperms, has only minor effects on the short shoots of cordaites and early conifers. Some have de-
patterns of character evolution discussed here (Doyle 2006). fined flowers as restricted to angiosperms (Gifford and Foster
A true ‘‘total-evidence’’ analysis of both morphological and 1989; Loconte and Stevenson 1990), but this would leave no
molecular data might be theoretically preferable, with fossils term to express the similar appearance of the fertile structures
scored as unknown for molecular characters. However, the of Bennettitales (and Gnetales under older interpretations,
present procedure may be a valid approximation of such an but not if the ‘‘flowers’’ are compound strobili; Mundry and
analysis, and it may be of interest as a heuristic exercise that Stützel 2004).
asks what the implications of molecular analyses are if they Retallack and Dilcher (1988) reconstructed the leaf-cupule
are correct. complexes of glossopterids as borne on fertile short shoots,
Of the constrained trees, I have chosen one from the first but Pant and Singh (1974) illustrated two leaf-cupule com-
island (fig. 5A). First, trees of this island are more consistent plexes attached to a normal stem, and in any case vegetative
with stratigraphic evidence, since they entail a shorter gap (ghost leaves were borne on both long and short shoots (Pant and
lineage; Norell 1992) in the record of the pan-angiosperm Singh 1974). Given the large size of the pinnate sporophylls of
clade. All the constrained trees imply that the pan-angiosperm Caytonia, it seems unlikely that they were grouped into flow-
line diverged before the Late Carboniferous, the age of the old- erlike structures. This would be supported by a specimen, fig-
est known acrogymnosperms (cordaites and conifers). However, ured by Thomas (1925) and reexamined by Retallack and
in trees of the second island (fig. 5B), no pan-angiosperms are Dilcher (1988), of a megasporophyll attached to a slender
known until the Late Triassic (Bennettitales, Caytonia), whereas stem, but more specimens in organic connection would be de-
in trees of the first island, the pan-angiosperm clade includes sirable for confirmation. Microsporophylls of many Bennetti-
the glossopterids, which are abundant in the Permian. tales were quite large, but they often remained attached to the
Second, trees of the second island (fig. 5B) imply that the floral axis. If the present analysis is correct in identifying Cay-
leaf type in the common ancestor of living seed plants was tonia as the sister group of the angiosperms, it implies that ag-
simple pinnate, a term used here to include both simple, pin- gregation of fertile parts into flowers occurred independently
nately veined leaves of the Taeniopteris type and the com- in Bennettitales and angiosperms or that aggregation occurred
pound leaves of living cycads, in which the leaflets have in the common ancestor of these groups and Caytonia re-
parallel or dichotomous venation (for discussion of the dis- verted to a nonfloral state.
tinction between the simple pinnate and pinnately compound Whether the initial aggregation of fertile parts into a flower
states, see Doyle 1996). The fernlike, pinnately compound occurred before or after the divergence of Caytonia, an impor-
leaves of Callistophyton, corystosperms, and peltasperms (Au- tant question is whether it also involved sterile appendages
tunia and Peltaspermum) must be derived from this type that became the perianth. Optimization of characters on cur-
rather than being homologous (symplesiomorphic) with the rent angiosperm phylogenies, without considering outgroups,
similar leaves in basal seed ferns. In those trees in which Cal- implies that the most recent common ancestor of all living an-
listophyton is basal to all living seed plants, the close similari- giosperms had a perianth (Doyle and Endress 2000; Ronse De
ties between its leaves and those of corystosperms and Craene et al. 2003). This conclusion still holds with the addi-
peltasperms (which differ mainly in having rachial pinnules) tion of Hydatellaceae, which lack a perianth, because this
must also be homoplastic. Both scenarios would require a re- group is located two nodes above the base of the angiosperms.
markable reversal to the ancestral leaf type. In contrast, in In another near-basal group, Chloranthaceae, the genera As-

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
830 INTERNATIONAL JOURNAL OF PLANT SCIENCES

carina, Sarcandra, and Chloranthus have no perianth, but He- one pathway. For simplicity, in the following discussion I use
dyosmum, their sister group, has three small tepals (Endress ‘‘strobili’’ for both flowers and strobili and ‘‘cataphylls’’ for all
1987; Eklund et al. 2004). A perianth is also reconstructed as associated sterile appendages.
ancestral if one assumes that the closest outgroups of angio- Mapping the three-state character devised to address this
sperms lacked a perianth; it is most parsimonious to assume question on the preferred tree (fig. 9) underlines the ambiguity
that the perianth arose in the common ancestor of living an- of existing data on this question. Under all trees and with
giosperms and was lost in Hydatellaceae and other taxa with both scorings of cycads, the ancestral state for seed plants as a
no perianth. This is illustrated by a tree from the combined whole is no strobili, but the situation in crown-group seed
analysis of Doyle and Endress (2000), with modifications based plants is complex. When cycads are scored as either lacking
on more recent molecular data (Endress and Doyle 2007) and strobili or having strobili with cataphylls (0/2), the state at the
with two hypothetical outgroups that lack a perianth added at crown-group node is equivocal: either no strobili or strobili
the base (fig. 8). with cataphylls (fig. 9). There are five equally parsimonious
Although these arguments imply that both fertile and sterile scenarios for the evolution of this character. In four scenarios,
appendages became aggregated on the line leading to angio- the common ancestor of the crown group lacked strobili, and
sperms, they do not say whether the two kinds of appendages strobili with cataphylls originated within cycads and in the
were aggregated at the same time or the fertile ones first; i.e., common ancestor of corystosperms, peltasperms, and conifero-
whether the first flower had a perianth. This question can be phytes; cataphylls were then lost independently in Ginkgoales
addressed only by considering outgroups, but, as already and peltasperms. In the pan-angiosperm line, strobili origi-
noted, this poses problems concerning homologies of the fer- nated three times, in Pentoxylon, Bennettitales, and angio-
tile structures and the fact that the sporophylls of some fossils sperms, with cataphylls in Pentoxylon and angiosperms but
(and female plants of Cycas) are not aggregated into flowers either (1) without or (2) with cataphylls in Bennettitales; (3)
or strobili. This raises the possibility that aggregation of spo- twice, with cataphylls, in Pentoxylon and the common ances-
rophylls occurred more than once and followed more than tor of Bennettitales and angiosperms, with a reversal to no

Fig. 8 Angiosperm tree from Endress and Doyle (2007), with addition of Hydatellaceae, showing the inferred evolution of presence or absence
of perianth. Two hypothetical outgroups without perianth added at the base. Nymph ¼ Nymphaeales, Aust ¼ Austrobaileyales, Chlor ¼ Chlo-
ranthaceae, Piper ¼ Piperales, Ca ¼ Canellales, Magnol ¼ Magnoliales.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 831

Fig. 9 Representative tree from the analysis with the molecular backbone (fig. 5A), showing the inferred evolution of a character (134 [see
appendix in the online edition], not used in phylogenetic analyses) that distinguishes fertile appendages not borne on differentiated shoots from
strobili or flowers with and without perianth or comparable sterile appendages. GNET ¼ Gnetales.

strobili in Caytonia; or (4) once, with cataphylls, and with of angiosperms, especially its outer integument, is an equally
separate reversals to no strobili in glossopterids and Caytonia. serious problem, and the two questions are closely linked.
Under the scenario 5, strobili with cataphylls originated in the Several of the advances incorporated in Doyle (2006) con-
common ancestor of living seed plants, and there were inde- cern morphology and potential homologies of the angiosperm
pendent reversals to no strobili in female Cycas, Callistophy- bitegmic ovule and the ovule-bearing structures of fossil seed
ton, glossopterids, and Caytonia. Alternative 5 is the only plants, particularly those called cupules. The term ‘‘cupule’’
most parsimonious scenario when cycads are scored as having has been used in many taxa, but it is widely recognized that
strobili with cataphylls. Under both scorings of cycads, if Cal- cupules of the first seed plants, which were dichotomously or-
listophyton is placed in the corystosperm-peltasperm clade or ganized structures borne at the tips of special fronds or parts
below crown-group seed plants, it is most parsimonious to as- of fronds (Kidston 1924; Galtier 1988; Retallack and Dilcher
sume that the common ancestor of living seed plants had stro- 1988; Serbet and Rothwell 1992), were probably not homolo-
bili with cataphylls. gous with cupules of groups such as Caytonia and corystosperms,
This exercise underlines the slippery nature of the distinctions which appear to be laminar, with seeds on one surface, and
among flowers, strobili, and less differentiated fertile shoots. are therefore either leaves (sporophylls) or leaflets. This has
However, it does suggest that the common ancestor of Caytonia been confirmed by phylogenetic analyses of seed plants (cf.
and angiosperms had either fertile appendages borne on un- figs. 2–5), which place several lines with no cupules and seeds
specialized shoots or flowerlike shoots with both fertile organs on more or less unmodified leaves (such as medullosans, Cal-
and sterile appendages that could be homologous with a peri- listophyton, and cycads) between basal seed plants with di-
anth. It provides no support for a scenario in which flowers chotomous cupules (such as Elkinsia and Lyginopteris) and
were derived from special fertile shoots with no associated ster- later groups with cupules of the laminar type. Recently, it has
ile appendages. Further clarification may require recognition become clearer that laminar cupules were of two kinds: in
of closer angiosperm outgroups, as discussed further below. some the seeds were abaxial on the cupule lamina, and in
others they were adaxial.
In Permian and Triassic peltasperms (Autunia and Pelta-
Homologies of the Bitegmic Ovule and the Carpel
spermum), the seeds were borne on the abaxial side of spoon-
The origin of the carpel is often considered the most impor- shaped or peltate structures. Some authors interpreted these
tant aspect of the origin of the flower. However, as discussed as leaflets of a pinnate leaf (Townrow 1960; Doyle and Don-
in the ‘‘Introduction,’’ the origin of the peculiar bitegmic ovule oghue 1986; Retallack and Dilcher 1988), but it has been

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
832 INTERNATIONAL JOURNAL OF PLANT SCIENCES

shown that they were attached helically to an axis and were

therefore presumably sporophylls (Meyen 1987; Kerp 1988;
Nixon et al. 1994). In Triassic corystosperms, it has also
been uncertain whether the systems bearing the helmet-shaped
cupules were branches with simple sporophylls modified into
cupules (Thomas 1933) or pinnately compound sporophylls
with ovule-bearing leaflets (Harris 1951; Doyle and Dono-
ghue 1986; Retallack and Dilcher 1988). However, studies on
compressed and silicified material from Antarctica (Axsmith
et al. 2000; Klavins et al. 2002) have shown that they were
branches with spirally arranged cupules, plus a few scalelike
appendages below. Furthermore, in the vascular bundles in
the cupule, the xylem was located toward the outside of the
cupule, identifying that side as adaxial, and the phloem was
toward the inside, implying that the seeds were borne on the
abaxial surface, as in peltasperms.
In Caytonia, Harris (1940, 1951) and Reymanówna (1973)
argued that the cupules were adaxially folded leaflets, based
especially on their relation to the strongly dorsiventral rachis.
The cupules were attached to the flatter and presumably adax-
ial side of the rachis and enrolled toward its midline in circi-
nate fashion, implying that the enclosed seeds were adaxial
(fig. 1A). In contrast, Retallack and Dilcher (1988) reconstructed
the cupules as abaxially folded, on the basis of Thomas’s
(1925) specimen, mentioned above, of a sporophyll attached
to a stem. However, after examination of this specimen, I was
not convinced that the orientation of the sporophyll could be
determined. Whether or not Harris was correct about the ori- Fig. 10 Interpretation of positional relationships in the leaf-cupule
entation of Caytonia (a question discussed further below), ad- complex of glossopterids, with abaxial surfaces indicated in black.
axial seeds are known in at least one Triassic cupule-bearing
structure, Petriellaea (Taylor et al. 1994), from the position of
xylem and phloem in vascular bundles of the cupule wall. is that the cupule is the adaxial fertile segment of a three-
Adaxial ovule position has also been demonstrated in glos- dimensional leaf, as seen in the eusporangiate fern order Ophio-
sopterids. The cupules (or sporophylls) of glossopterids, which glossales. Kato’s suggestion that the adaxial fertile segment of
are attached to the midrib of a more or less unmodified leaf, the Ophioglossales, the glossopterid cupule, and the angiosperm
have been reconstructed as having seeds on the surface facing bitegmic ovule were homologous is ruled out by molecular
the leaf (Schopf 1976; Retallack and Dilcher 1981; fig. 1B), evidence that Ophioglossales are not related to seed plants (Pryer
but this has become controversial. Work by Taylor and Taylor et al. 2001; Rothwell and Nixon 2006), but these structures
(1992) on isolated silicified cupules with preserved vascular might be morphologically similar.
bundles showed that the seeds were attached to the xylem side In angiosperms, if the bitegmic ovule was derived from a
of the cupule and therefore adaxial. They took this to mean cupule, by reduction of the number of ovules to one, this cu-
that the seed-bearing surface faced away from the subtending pule was apparently of the type with adaxial ovules. In other
leaf. Taylor (1996) also argued that this orientation would be words, the nucellus plus the inner integument (¼original seed
necessary for pollination of the ovules, a questionable argu- plant ovule) is borne on the adaxial side of the outer integu-
ment in light of the fact that pollen reaches the ovules of coni- ment (¼cupule wall). This has nothing to do with the adaxial
fers even though they are borne on cone scales that are barely position of the bitegmic ovule on the carpel. As noted by
separated at the time of pollination. However, careful studies Frohlich (2003), when there are vascular bundles in the outer
of impressions split in various planes reaffirm that the side of integument, the phloem is to the outside (Svoma 1997), iden-
the cupule bearing the ovules faced the leaf (McLoughlin tifying the outer surface as abaxial and the inner as adaxial.
1990; Adendorff 2005). As Frohlich (2003) recognized, the finding that ovules are ab-
These observations can be most simply reconciled by inter- axial in the cupules of corystosperms (Axsmith et al. 2000;
preting the cupule as a sporophyll borne on an axillary branch Klavins et al. 2002) contradicts the homology between corys-
(fig. 10) on the opposite side from the subtending leaf, like the tosperm cupules and angiosperm bitegmic ovules postulated
adaxial prophyll of monocots and many magnoliids, such that in the original version of his ‘‘mostly male’’ hypothesis (Froh-
the adaxial surface faced the leaf (Retallack and Dilcher 1981, lich and Parker 2000).
fig. 3B). Another view, that the cupule is a modified branch Nixon et al. (1994) questioned the homology of the Cayto-
(cladode; Schopf 1976; Retallack and Dilcher 1981, fig. 3C) nia cupule and the angiosperm bitegmic ovule because of their
has become more difficult to maintain in light of its typically observations on Caytonia, which showed that the lip of the
foliar anatomy (Gould and Delevoryas 1977; Taylor and Tay- cupule extends only to either side of the stalk rather than
lor 1992). A more exotic possibility, suggested by Kato (1990), forming a complete ring. However, as discussed in Doyle

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 833

(1996), this condition is similar to that in many anatropous step process. The first step, beginning with the dichotomous
angiosperm ovules, in which the outer integument is interrupted sporophyte body of the first vascular plants, resulted in main
by the funicle, a condition called ‘‘synbi’’ (Taylor 1991) or stems with simple dichotomous, overtopped lateral append-
‘‘semiannular’’ (Umeda et al. 1994; Imaichi et al. 1995; En- ages, as in ‘‘trimerophytes,’’ or basal euphyllophytes (Kenrick
dress and Igersheim 1997; Igersheim and Endress 1997; Ya- and Crane 1997; Euphyllophyta [Cantino et al. 2007]), such
mada et al. 2001a). From the analysis of Doyle and Endress as the Early Devonian genus Psilophyton. This presumably
(2000), the semiannular condition can be reconstructed as an- occurred in the common ancestor of the euphyllophyte clade,
cestral in angiosperms, assuming that bitegmic ovules were which includes ferns, sphenophytes, ‘‘progymnosperms,’’ and
originally anatropous rather than orthotropous. seed plants. The second step involved planation of whole
Additional evidence comes from developmental genetic branch systems of the progymnosperm type, which bore nu-
studies in Arabidopsis based on phenotypes of developmental merous dichotomous appendages, into pinnately compound
mutants and localization of gene expression. Genes of the fronds (Beck 1970; Doyle and Donoghue 1986; Kenrick and
YABBY family are required for laminar growth and differen- Crane 1997; Doyle 1998). The systematic distribution of
tiation of abaxial tissues in leaves and other lateral organs YABBY genes suggests that they were involved in the second
(Bowman 2000; Floyd and Bowman 2007). One of these, step but not the first. This would be consistent with the hy-
INO (¼INNER NO OUTER because mutants have an inner pothesis presented here that the outer integument is a cupule
but no outer integument), is expressed in the outer epidermis derived from an ovule-bearing leaf, itself borne on an axillary
of the outer integument (Villanueva et al. 1999; Balasubra- branch, which was earlier derived from a compound leaf by
manian and Schneitz 2000; Meister et al. 2002; Skinner et al. simplification, like the vegetative leaves of glossopterids.
2004) but not in the inner epidermis or the inner integument. These observations can be related to the present phyloge-
This identifies the outside of the outer integument as abaxial, netic results by plotting the position of the original ovule rel-
consistent with its derivation from a laminar structure with an ative to the laminar structure on which it is borne, whether
ovule on its adaxial surface (as noted by Meister et al. [2002]; this is called a sporophyll or a cupule (fig. 11). This implies
Frohlich [2003]; Yamada et al. [2003]; Skinner et al. [2004]). that ovules in the first seed plants (seed ferns) were apical on
However, Sieber et al. (2004) questioned this interpretation frondlike leaves, first in cupules at the tips of branched
because the class III HD-Zip (homeodomain-leucine zipper) fronds or segments of fronds (Kidston 1924; Galtier 1988;
gene PHB, which is involved in determining adaxial identity Retallack and Dilcher 1988; Serbet and Rothwell 1992) and
in leaves, is expressed on the inside of the inner integument later without cupules at the tips of pinnae. With the topology
but not on the inside of the outer integument. They suggested under discussion (fig. 11), because of the variation in ovule
that this may indicate an origin of the two integuments from position in acrogymnosperms, ovule position at the base of
one by splitting. This hypothesis would greatly widen the both crown seed plants and acrogymnosperms is equivocal
field of comparison of angiosperms with outgroups to include (any of the four states). However, ovules became marginal in
taxa with unitegmic ovules not borne in cupulelike structures. cycads and abaxial in Callistophyton, corystosperms, and
Since my earlier review of this topic (Doyle 2006), under- peltasperms, either once or twice; in corystosperms and pel-
standing of the genetic control of ovule polarity has improved, tasperms the sporophylls were modified into what are called
and new data strengthen the interpretation of the outer integ- cupules. It is equivocal whether the apical position of ovules
ument as a separate laminar structure. Genes of the KANADI in coniferophytes (evident in cordaites and Ginkgoales but
family, which along with YABBY genes specify abaxial devel- obscured in extant conifers by transformation of the fertile
opment in leaves (Eshed et al. 2001, 2004), are required for short shoots into cone scales) is ancestral or derived from an ab-
normal growth of both the inner (ATS) and the outer (KAN1 axial position. In other most parsimonious arrangements where
and KAN2) integuments, and ATS is known to be expressed Callistophyton, corystosperms, and peltasperms form a clade,
on the outside of the inner integument (McAbee et al. 2006). abaxial position originates once from apical; with other topol-
McAbee et al. (2006) suggested that class III HD-Zip genes ogies, it originates once and reverses to apical in coniferophytes.
are adaxial determinants in both integuments, as well as in In contrast, the two fossil taxa with adaxial ovules—glossopterids
leaves, but they stressed that this has not yet been demon- and Caytonia—are in the pan-angiosperm clade, associated
strated. However, D. Kelley (personal communication) has with angiosperms, where the presumed equivalent of the origi-
found that not only PHB but also the class III HD-Zip gene nal seed plant ovule is adaxial relative to the outer integument
PHV is expressed on the inside of the inner integument. The (and with Pentoxylon and Bennettitales, where ovule position
fact that both YABBY and KANADI (KAN1 and KAN2) genes is problematic and was scored as unknown). The inferred re-
are involved in development of the outer integument but that lationships confirm that the cupules of these groups are not
only a KANADI gene (ATS) is involved in development of the homologous with those of corystosperms and peltasperms.
inner integument has been taken as further evidence that the These relationships are not based solely on this character; they
inner and outer integuments have different origins (McAbee are also supported by such angiosperm-like features as scalari-
et al. 2006). form pitting in the secondary xylem, siphonogamy, and reduc-
Floyd and Bowman (2007) noted that YABBY genes have tion of the megaspore membrane in Bennettitales and reticulate
been found in all investigated seed plants but not in lyco- venation and ‘‘flat’’ stomata without raised guard cell poles
phytes or ferns, and they speculated that this reflects the in- (Harris 1932; Doyle and Donoghue 1986; Barbacka and Bóka
dependent origin of leaves in seed plants and other lines. 2000) in Caytonia.
These observations can be related to the hypothesis that ori- Setting aside Bennettitales and Pentoxylon, in which repro-
gin of the ancestral fernlike leaf of seed plants was a two- ductive morphology is more obscure, possible homologies of

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
834 INTERNATIONAL JOURNAL OF PLANT SCIENCES

Fig. 11 Representative tree from the analysis with the molecular backbone (fig. 5A), showing the inferred evolution of ovule position on the
supporting laminar structure (character 85; see appendix in the online edition). GNET ¼ Gnetales.

the ovulate structures in glossopterids, Caytonia, and angio- pous structure (fig. 13), as seen in Caytonia (figs. 1A, 12E–12G),
sperms can be discussed with reference to several diagrams in followed by reduction of the number of ovules per cupule to
which abaxial surfaces of leaflike structures are indicated in one in angiosperms. Amborella and Chloranthaceae have or-
black (fig. 12). For angiosperms (fig. 12D), I have shown an thotropous rather than anatropous bitegmic ovules, and if
ascidiate carpel with one apical ovule attached to the adaxial outgroups are not considered, it is equivocal whether the an-
cross zone, as in many living basal angiosperms (Endress and cestral state in angiosperms was anatropous or orthotropous
Igersheim 2000; Buzgo et al. 2004). The positional corre- (Doyle and Endress 2000). In the present analysis, it is most
spondences between this type of carpel and the glossopterid parsimonious to interpret orthotropous ovules as derived
leaf-cupule complex (fig. 12A) are particularly close: the car- from anatropous ovules (fig. 13), because the only outgroups
pel wall corresponds to the subtending leaf, the outer integu- reported to have an orthotropous cupule or bitegmic ovule are
ment to the cupule with its adaxial ovules, and the adaxial some Bennettitales (Harris 1954; Crane 1985a; Pedersen et al.
cross zone to the area of fusion of the leaf and the axillary 1989; Friis et al. 2007). As discussed further below, Rothwell
cupule-bearing branch. This is consistent with the view of and Stockey (2002) and Stockey and Rothwell (2003) have in-
many developmental geneticists (Skinner et al. 2004) that the terpreted Bennettitales as having no outer envelope, and I
carpel wall and the placenta are separate structures. At the therefore scored them as uncertain (1/3). If I had scored Ben-
time when Stebbins (1974) and Retallack and Dilcher (1981) nettitales as having an orthotropous cupule, the ancestral state
proposed homologies of this sort, the correspondences with in angiosperms would be equivocal: either orthotropous or
angiosperms seemed less close, because most authors as- anatropous. Consistent with the view that the ovule orienta-
sumed that the ancestral carpel had two rows of laminar or tion of Amborella is derived, it is not completely orthotropous
marginal bitegmic ovules. However, the analysis of Doyle at maturity, and in both Amborella and Chloranthus, the in-
and Endress (2000) reconstructed the most parsimonious an- tegument is semiannular and dorsiventral early in develop-
cestral condition as either more than one ovule (as in core ment (Yamada et al. 2001b).
Nymphaeales, Austrobaileya, and Schisandraceae) or one (as Even if a uniovulate carpel is assumed to be ancestral in
in Amborella, Trimenia, Illicium, and Chloranthaceae), and angiosperms, the fact that ovules are arranged in two rows in
with the addition of Hydatellaceae, which have one apical several near-basal groups (and many more derived taxa) may
ovule, it becomes most parsimonious to reconstruct the uni- seem difficult to reconcile with the hypothesis that their pre-
ovulate condition as ancestral (fig. 13, right). cursor structure was borne on an axillary branch. However,
The present results imply that the next step in the modifica- the laminar to median (‘‘dorsal’’) placentation of the near-
tion of the cupule was its circinate incurvation into an anatro- basal Cabombaceae and Nymphaeaceae (Taylor 1991; Iger-

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 835

Fig. 12 Alternative interpretations of the homologies of ovulate structures in glossopterids (A–C), an ascidiate angiosperm carpel (D), and
Caytonia (E–G), with abaxial surfaces indicated in black. See text for discussion.

sheim and Endress 1998) might be more consistent with an tween Permian glossopterids and the first known crown-group
original axillary position. Under any scenario, ovule number angiosperms of the Early Cretaceous. This requires that plants
and position must have been highly labile since the beginning with one or more cupules apparently borne in the axil or on
of the angiosperm radiation. This lability was cited by Froh- the midrib of a leaf or bract persisted through the Triassic and
lich and Parker (2000; also Frohlich 2003) as support for Jurassic without being recognized in the fossil record. How-
their ‘‘mostly male’’ hypothesis that ovules (¼cupules) were ever, given the remarkable but poorly understood diversity of
transferred ectopically to organs that had previously been mi- Late Triassic seed plants, most graphically seen in the Molteno
crosporophylls, but it might simply indicate an early break- Formation of South Africa (Anderson and Anderson 2003),
down in positional relationships within angiosperms. the idea that such forms exist but have escaped recognition
Other uncertainties concern the diversity of ovulate struc- may not be so implausible.
tures in glossopterids (cf. Pigg and Trivett 1994; Taylor Compared with glossopterid cupules, the cupules of Cayto-
1996). Some glossopterids (e.g., Lidgettonia; Thomas 1958; nia are more similar to the bitegmic ovules of angiosperms
Surange and Chandra 1975; Schopf 1976; Retallack and in their anatropous form, and the ovules inside are more
Dilcher 1981; fig. 1B) had more than one cupule attached to angiosperm-like in having no pollen chamber, a thick nucel-
the midrib of a leaf. As discussed in Doyle (2006), there are lar cuticle, and no megaspore membrane. However, the structure
several possible interpretations of the morphology of these that bore the cupules is hard to explain in terms of the homolo-
structures (fig. 12B, 12C). In addition to the possibility that gies proposed here between glossopterids and angiosperms.
the cupules were part of an adaxial fertile leaf segment anal- Harris (1940, 1951) and Reymanówna (1973) interpreted
ogous to that of Ophioglossales (Kato 1990), one alternative the cupule-bearing structure of Caytonia as a pinnate sporo-
is that the cupules were leaflets of a pinnate sporophyll borne phyll and the cupules as adaxially enrolled leaflets. However,
on a branch (fig. 12B), and another is that they were several this is uncertain because there is little evidence on how the
sporophylls borne on an axillary branch, twisted to face the structure was attached to the plant, except for the specimen
subtending leaf (fig. 12C). of Thomas (1925) cited by Retallack and Dilcher (1988) and
All attempts to homologize parts in glossopterids and an- discussed above. If it was a sporophyll borne on a normal
giosperms must contend with the long stratigraphic gap be- stem (fig. 12E), the situation would be different from that in

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
836 INTERNATIONAL JOURNAL OF PLANT SCIENCES

Fig. 13 Representative tree from the analysis with the molecular backbone (fig. 5A). Left, inferred evolution of cupules and comparable
envelopes (character 87; see appendix in the online edition); right, angiosperm portion of the same tree, showing the inferred evolution of number
of ovules per carpel (character 100). GNET ¼ Gnetales.

glossopterids, where cupules were apparently borne on an ax- that are usually interpreted as sterilized ovules. Crane (1985a)
illary branch. Existence of such a structure in the common an- proposed that each stalked ovule was a reduced cupule-bearing
cestor of Caytonia and angiosperms might be more consistent sporophyll, following the interpretation of Harris (1954) that
with the rachis expansion hypothesis for origin of the carpel some Late Triassic Bennettitales had a cupule, e.g., Bennetti-
(Gaussen 1946; Doyle 1978; fig. 1A). Glossopterids and Cay- carpus crossospermus and Vardekloeftia, which was restudied
tonia might be more comparable if the Caytonia structure was by Pedersen et al. (1989). However, in studies of petrified mate-
a pinnate sporophyll borne on an axillary branch (fig. 12F); it rial, Rothwell and Stockey (2002) and Stockey and Rothwell
could then correspond to the pinnate sporophyll model for (2003) argued that the layer in Williamsonia and Cycadeoidea
multicupulate glossopterids (fig. 12B). Perhaps the two groups that some authors had interpreted as a cupule is the sacrotesta of
could be most easily compared if the Caytonia structure was a single integument, not a morphologically distinct organ, and
not a sporophyll but rather an axillary branch with cupules they questioned whether a cupule existed in any Bennettitales.
derived from simple sporophylls (fig. 12G), as in the second Because Pedersen et al. (1989) provided seemingly convincing
model for multicupulate glossopterids (fig. 12C). This hypoth- evidence for a discrete envelope around the ovule in Varde-
esis would conflict with the dorsiventrality of the presumed kloeftia, especially the fact that the tubular micropyle clearly
rachis emphasized by Harris (1940, 1951; fig. 1A). It would protrudes through a hole in the cuticle of the structure in question,
also remove Harris’s main argument that Caytonia ovules I scored Bennettitales as uncertain (1/3) for the cupule character.
were borne on the adaxial side of the cupule, although it would A more exotic suggestion is that the ovuliferous receptacle
not disprove that they were. Deciding among these alterna- was a single radialized sporophyll bearing many ovules
tives may require discovery of more specimens that show at- (Doyle and Donoghue 1986; Crane 1988; Doyle 1994). The
tachment of the cupule-bearing structures to stems. However, finding of Rothwell and Serbet (1994) that the supposed ovu-
even if such specimens are found, it may not be easy to deter- liferous shoots of Pentoxylon had leaflike bilateral anatomy
mine the exact morphological relationships at the point of at- illustrates that fertile stems and sporophylls with seeds on
tachment in the absence of developmental data. both surfaces cannot always be easily distinguished. Because
It is even less clear whether and how Bennettitales might fit of these problems, I scored Bennettitales as unknown for spo-
into this scheme. They had numerous stalked ovules attached rophyll morphology and ovule position. As a result, these
to an ovuliferous receptacle, mixed with interseminal scales characters had no direct effect on inferred relationships, but

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 837

if they were better understood, they might either support or as a sporophyll with ovules borne on both surfaces. Such a
undermine the homologies proposed here between angio- sporophyll might be compared with the cupule of glossopter-
sperms and glossopterids. ids, but this would require a shift of the cupule from its pre-
A new element in this discussion is the description by Friis sumed position on an axillary branch, loss or fusion of the
et al. (2007) of Early Cretaceous dispersed seeds with a mi- subtending leaf, or some equally major change in morpholog-
cropylar tube, as in both Bennettitales and Gnetales, and a ical relationships.
four-part outer envelope, which they compared with the bi-
partite envelope in Gnetales and the tripartite envelope in the
Origin of the Stamen
less completely reconstructed fossil order Erdtmanithecales
(producers of Eucommiidites pollen; Friis and Pedersen 1996; As in most discussions of the flower problem, I have em-
Mendes et al. 2008). In a cladistic analysis building on the phasized the origin of the carpel, but the morphology of the
seed plant data set of Hilton and Bateman (2006), in which angiosperm stamen also has to be explained, and it poses as
Friis et al. (2007) rescored Bennettitales as having a ‘‘partite many problems. Perhaps because the stamen looks simpler to
(valvate) outer envelope,’’ the dispersed seeds were sister to the human eye, it is easy to overlook the fact it is as unique
Gnetales in a clade including Bennettitales, Erdtmanithecales, in a seed plant context as the carpel. Because stamens have
and Gnetales, whereas previously Bennettitales and Gnetales four microsporangia, authors beginning with Thomas (1925)
were separated. have compared them with the microsynangia of Caytonia,
Friis et al. (2007) noted that their results may suggest a radi- which consisted of four sporangia. However, as noted by
cal reinterpretation of the bennettitalian ovuliferous recepta- Harris (1937), stamens have a distinctly bilateral symmetry,
cle as an axis bearing many simple ‘‘flowers’’ of the gnetalian with the four sporangia in two lateral pairs (thecae) sepa-
type. A problem for this hypothesis is the fact that the stalked rated by the connective, whereas in Caytonia, the sporangia
ovules of Bennettitales are borne among numerous intersemi- formed radial groups not subdivided by sterile tissue, which
nal scales rather than in the axils of obvious subtending were borne at the tips of a branched structure most com-
bracts, as in Gnetales. Taylor and Kirchner (1996) did attempt monly interpreted as a pinnate sporophyll. Angiosperm the-
to interpret the ovules and interseminal scales of Bennettitales cae are distinctive in other respects, such as the presence of a
in terms of gnetalian units, but Rothwell and Stockey (2002) subepidermal endothecial layer, but this is an autapomorphy
found that the vascular strands supplying both kinds of ap- that has no direct bearing on outgroup relationships. In order
pendages depart from the stele in the same phyllotaxis and to explain the positional relationships, it might be better to
have similar anatomy. Finally, it is difficult to explain the homologize the stamen with a whole Caytonia sporophyll, dras-
large, apparently pinnate microsporophylls of Bennettitales in tically reduced to only one synangium of two sporangia on ei-
terms of Gnetales, especially with evidence confirming that the ther side. Following this interpretation, Doyle and Donoghue
microsporophylls of Gnetales are simple structures (Mundry (1986) scored angiosperm stamens as pinnate sporophylls,
and Stützel 2004). Because of these uncertainties, I did not but in Doyle (1996), I replaced this with a separate, less in-
modify the scoring of Bennettitales in the present analysis. formative but more theory-neutral state.
Even if Gnetales are related to Bennettitales, this would not It is also noteworthy that the stamens of most basal angio-
rule out homologies among angiosperms, glossopterids, and sperms have adaxial sporangia, and it is most parsimonious to
Caytonia, since it would place two taxa with similarly divergent reconstruct this state as ancestral (Doyle and Endress 2000;
morphology on the same side branch from the line leading to this study). In this respect they are like microsporophylls of
angiosperms. If Bennettitales are rescored like Gnetales and Bennettitales, in which microsynangia were borne on the ad-
placed (with or without Gnetales) below angiosperms and axial side of the sporophyll. Glossopterids had a branched
Caytonia, as found here and by Friis et al. (2007; fig. 3B), it unit bearing clusters of microsporangia that was attached to
is still most parsimonious to homologize the anatropous en- the adaxial side of a leaf or bract, reminiscent of the female
velope in Caytonia with that in angiosperms. The bitegmic structures (Surange and Maheshwari 1970; Surange and
structure that Friis et al. described in their fossil seeds and Chandra 1975; Schopf 1976; Gould and Delevoryas 1977).
Gnetales is not easy to compare with that in angiosperms, This suggests the possibility that the angiosperm stamen was
considering the one-sided development of the angiosperm derived from a leaf plus an adnate axillary branch, as hypoth-
outer integument (Umeda et al. 1994; Imaichi et al. 1995; esized for the carpel. The male structures of Caytonia are less
Yamada et al. 2001a), even when the ovule is orthotropous readily comparable if they were pinnate sporophylls, but in the
(Yamada et al. 2001b). Their comparisons also provide no absence of organic connection, it is hard to rule out the possibil-
obvious homologue for the angiosperm carpel. ity that they were actually branches in the axil of a leaf or bract.
It is also difficult to explain the reproductive structures of
Pentoxylon in terms of a glossopterid prototype (leaf mor- Closer Stem Relatives: The Question of Archaefructus
phology and wood anatomy would be more compatible with
a relationship of the two groups). Crane (1985a) interpreted Even if glossopterids, Pentoxylon, Bennettitales, and Cay-
the headlike seed-bearing structures, several of which were tonia are related to angiosperms, there is a great morphologi-
produced near the apex of a short shoot (Bose et al. 1985), cal gap between them and angiosperms. The preferred tree
as axes covered with numerous reduced sporophylls, which (fig. 5A) implies that at least seven synapomorphies arose on
he compared with the ovuliferous receptacle of Bennettitales. the line between Caytonia and angiosperms (more than one
However, based on the U-shaped configuration of the vascu- order of laminar venation, stamens with two pairs of micro-
lar tissue, Rothwell and Serbet (1994) interpreted the ‘‘head’’ sporangia, endothecial anther dehiscence, columellar exine,

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
838 INTERNATIONAL JOURNAL OF PLANT SCIENCES

sculptured sulcus membrane, reduced endexine, and carpel), (glossopterids and Caytonia in the present data set). They
and angiosperms have additional synapomorphies that can- questioned whether the perianth character can be polarized
not be localized on the tree because the relevant characters are using modern outgroups, since it is not readily applicable in
not preserved in fossils (the characteristic morphology of the taxa that lack structures comparable with flowers (except
male and female gametophytes, double fertilization, etc.). Gnetales, which Sun et al. [2002] scored as polymorphic), and
This highlights the need for closer stem relatives of the angio- they noted that perianth-like appendages do occur in at least
sperms. one potential fossil outgroup, Bennettitales. They also argued
Some supposed pre-Cretaceous angiosperms might be such that there are aquatic crown-group angiosperms with leaves
plants, but their morphology and relationships are uncertain. like those of Archaefructus, such as Cabomba in the Nym-
Cornet (1986, 1989b) associated Late Triassic pleated leaves phaeales, and that loss of perianth and reduction in number of
of Sanmiguelia with structures that he interpreted as flowers floral parts are common trends in aquatics. When they added
and carpels. However, details in the highly compressed fossils Cabomba to the Sun et al. (2002) data set and scored out-
are problematic (Crane 1987), and the male structures look groups as unknown for perianth, they found that it was
more like ginkgophyte strobili, raising the possibility that the equally parsimonious to place Archaefructus either on the an-
female structures are homologous with the ovuliferous ‘‘cap- giosperm stem lineage or with Cabomba.
sules’’ of the ginkgophyte group Czekanowskiales. Wang et al. Crepet et al. (2004) rejected these criticisms on several
(2007) described female structures of the Jurassic genus Schmeiss- grounds. Obviously, the flowers of core Nymphaeales, which
neria as bicarpellate gynoecia, but this is based on uncertain have a well-developed perianth and are almost always bisex-
interpretations of difficult compression material and the fact ual, are very different from those of Archaefructus. Crepet
the ovules are enclosed rather than on a convincing morpho- et al. noted that Cabomba has peltate as well as dissected
logical analysis of the supposed gynoecium. Otherwise, the leaves and argued that it should have been scored as polymor-
plant appears to be typically ginkgophytic in its short phic for this character. They called the scoring of the leaves of
shoots and leaf architecture (Kirchner and van Konijnenburg- Cabomba as dissected an ‘‘empirically verifiable character
van Cittert 1994), contrary to the discussion of Wang et al. miscoding’’ (Crepet et al. 2004, p. 1673). However, this cod-
(2007). ing is valid if the character states are defined as most or all
Cornet (1989a) also described angiosperm-like monosulcate leaves on the plant dissected versus none at all. Surely origin
pollen from the Late Triassic, with a tectal reticulum sup- of leaf dissection can provide evidence of relationship in de-
ported by columellae, and similar pollen has been found in scendant species, even if it does not affect all leaves on the
Triassic rocks elsewhere (Doyle and Hotton 1991; Hochuli plant, when contrasted with the complete lack of dissection in
and Feist-Burkhardt 2004). However, well-preserved speci- vast majority of basal angiosperms. Crepet et al. correctly
mens have a uniformly thick, laminated endexine, as in non- noted that plausibility arguments based on analogies with
angiospermous seed plants (Cornet 1989a; Doyle and Hotton cases of floral reduction in other aquatics provide no specific
1991). This could mean that these grains were not related to phylogenetic evidence on the position of Archaefructus.
angiosperms, but it could also mean that they were angio- The point of Friis et al. (2003) was not that Archaefructus
sperm stem relatives (Doyle and Hotton 1991; Doyle 2001, belonged to Nymphaeales but rather that its outgroup position
2005). However, until these fossils are associated with other was weakly supported. However, the discovery that Hydatel-
plant parts, there is little way to evaluate what, if anything, laceae are related to Nymphaeales shows that there is more
they say about the origin of angiosperms and the flower. floral diversity in this line than previously imagined, includ-
A better preserved fossil that has been explicitly proposed as an ing flowers even more reduced than those of Archaefructus—
angiosperm stem relative is Archaefructus, from the Barremian- only one stamen or one carpel, with no perianth or subtending
Aptian of China, which had multiovulate carpels borne along bract. Saarela et al. (2007) therefore made a speculative sug-
an elongate axis (Sun et al. 1998), finely dissected leaves, and gestion that Hydatellaceae might be more derived relatives of
pairs of stamens below the carpels (Sun et al. 2002). The cla- Archaefructus.
distic analysis of Sun et al. (2002), who used other living This suggestion is supported by analyses of Doyle and En-
seed plants as outgroups, indicated that Archaefructus was dress (2007; Endress and Doyle, forthcoming), which link
sister to all living angiosperms, on the basis of leaf characters Archaefructus with Hydatellaceae in a data set of angiosperms
(one vein order, dichotomous laminar vein form, nonanasto- only, and by the present analysis (fig. 6). Assuming that the
mosing veins) and the absence of a perianth. Stuessy (2004) fertile shoot of Archaefructus is an inflorescence of unisexual
took this result as evidence that the carpel evolved before flowers, its most parsimonious position is with Hydatellaceae.
double fertilization and condensation of parts into a typical A position as the sister group of living angiosperms is five
flower. steps less parsimonious, and several positions much lower on
The interpretation of Sun et al. (2002) was questioned by the tree have the same score, because of similarities between
Friis et al. (2003), who argued that the fertile shoot was more the leaves of Archaefructus and seed ferns. Unequivocal syn-
likely an inflorescence of unisexual flowers, reduced as an ad- apomorphies of Archaefructus and Hydatellaceae (i.e., char-
aptation to aquatic conditions; this argument was based par- acter state changes unambiguously located at this node) are
ticularly on the fact that the stamens and often the carpels loss of floral bracts and loss of perianth. Sun et al. (2002) cited
occur in pairs. Friis et al. argued that two of the features sup- the absence of bracts below the carpels and stamen pairs as
porting the stem position of Archaefructus are equivalent, di- evidence that the fertile shoot of Archaefructus was a flower
chotomous venation and lack of anastomoses, and that some rather than an inflorescence, but this feature is shared with
potential fossil outgroups have the derived reticulate state Hydatellaceae (Hamann 1975; Rudall et al. 2007), as well as

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 839

with other taxa, such as Acorus and Araceae (Buzgo and En- with this ‘‘floral disintegration’’ hypothesis. For example, it
dress 2000; Remizova and Sokoloff 2003). Synapomorphies implies that the ancestral flower had a perianth (fig. 8), more
of Archaefructus and Nymphaeales as a whole (not counting than one stamen, and more than one carpel. However, be-
those of Hydatellaceae and core Nymphaeales that are not cause of the number of unisexual lines (including Amborella)
preserved in fossils) are (semi)herbaceous habit, short or long near the base of the tree, it is equivocal whether the ancestral
and narrow stamen base (rather than long and wide), boat- flower was bisexual or unisexual. Addition of Archaefructus,
shaped pollen, and palisade exotesta. The single order of lami- which has several ovules per carpel, also weakens the infer-
nar venation in Archaefructus would be more consistent with ence based on living taxa that one apical ovule was ancestral
a stem position than with a position in the crown group (it in angiosperms; the reconstructed ancestral ovule number be-
would undergo two changes across the tree rather than three), comes equivocal. These issues are more appropriately ex-
but its open venation would not, because venation in the next plored in the context of a broader analysis of angiosperms
outgroup, Caytonia, is reticulate. This illustrates the dangers (Endress and Doyle, forthcoming).
of relying on only living outgroups to polarize characters in Whether or not Archaefructus affects ideas about the first
divergent living groups. flower, it reveals important early trends in floral evolution. It
Lack of a perianth in Archaefructus also contributes to its is also significant for the ecological radiation of angiosperms.
inferred position in the crown group. Even though outgroups Especially if it is related to the Albian genera Vitiphyllum
were scored as unknown for this character (for inapplicable), and Caspiocarpus (Friis et al. 2003), which had similar but
if Archaefructus is separated from Hydatellaceae, this charac- less finely dissected leaves, it represents an important trend
ter undergoes three steps on the tree rather than two (the for invasion of Early Cretaceous aquatic ecosystems, repre-
other being in Saururaceae). But even if outgroups had been sented today only by Hydatellaceae and core Nymphaeales.
scored as lacking a perianth, as in Sun et al. (2002), a rela- Stuessy (2004) suggested that his hypothesis that the carpel
tionship of Archaefructus with Hydatellaceae would be four evolved before double fertilization and the flower, which he
steps more parsimonious than a position on the stem lineage. argued was supported by Archaefructus, could reconcile the
The discovery that Hydatellaceae are related to Nymphaeales apparent conflict between fossil evidence that angiosperms be-
increases support for the hypothesis that Archaefructus is a gan to radiate in the Early Cretaceous and older dates based
crown-group angiosperm. When Hydatellaceae are removed on molecular data. He proposed that molecular dates mark
from the analysis, it is still most parsimonious to group Ar- the origin of the first angiosperm feature (the carpel), as seen
chaefructus with the remaining Nymphaeales, but the relative in Archaefructus, whereas the Cretaceous radiation began af-
parsimony of a position as sister to living angiosperms im- ter all three advances had evolved. If the present analysis is
proves from five steps less parsimonious to only two. correct in placing Archaefructus in the crown group, Archae-
These results might be questioned because they assume fructus does not support this or any other scenario for the or-
that the fertile shoot of Archaefructus was an inflorescence. der of origin of angiosperm features, but even if it did,
As Sun et al. (2002) argued, perhaps it was a preflower more Stuessy’s proposal for reconciliation of fossil and molecular
primitive than anything today; for example, they speculated dates would be invalid. Molecular dates are based on living
that the supposed paired stamens could be remnants of the taxa only, so by definition they provide ages for crown groups,
branched male structures of other seed plants. With Archae- not earlier events that occurred on the stem lineage, such as
fructus scored along these lines (fig. 7), one of its most parsi- origin of the carpel in Stuessy’s scheme. There is every reason
monious positions is still with Hydatellaceae, supported by to assume that some angiosperm apomorphies evolved before
absence of perianth, but it is equally parsimonious to link it others rather than all at once and that the angiosperm radia-
with core Nymphaeales, and a stem position becomes less tion did not begin until a suite of apomorphies had accumu-
unparsimonious: it is three steps worse than a position in lated (Doyle and Donoghue 1993), but until closer stem
Nymphaeales rather than five. Features that would support a relatives are confidently recognized, the order in which these
stem position are one order of laminar venation and pinnate apomorphies evolved cannot be determined.
microsporophylls, whereas a link with Hydatellaceae is no
longer supported by absence of floral bracts. It may therefore Conclusions
be premature to rule out the possibility that Archaefructus is
an angiosperm stem relative. Even though we have whole These considerations underline the point that the present
plants of Archaefructus, we need to know more about its analysis should be regarded as a heuristic exercise that may
morphology before we can be sure what it is. suggest avenues for future research, not as a definitive account
Even if Archaefructus is a crown-group angiosperm and of the origin of the flower. They also emphasize how much we do
does not affect inferred outgroup relationships, it could affect not know about the morphology of critical fossils. However,
reconstruction of the ancestral flower. In a commentary on they also suggest that better information on these fossils could
Saarela et al. (2007), Friis and Crane (2007) suggested that result in major progress toward understanding of the origin of
the increasing number of taxa with simple flowers near the the flower and its parts. There are also even less well under-
base of the angiosperm tree, including not only Hydatella- stood fossils that may be relevant, such as the gigantopterids
ceae and Archaefructus but also Chloranthaceae and Cerato- of the Permian of China, which had leaf venation approaching
phyllum, raises the possibility that these plants represent a that of angiosperms and shared the triterpenoid oleanane
prefloral state. An alternative (Rudall 2007) is that the flower (Taylor et al. 2006) but have not been associated with repro-
was still poorly integrated and could easily lose its distinction ductive structures, and the diverse seed plants of the Late Tri-
from an inflorescence. The present analysis is more consistent assic Molteno flora of South Africa (Anderson and Anderson

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
840 INTERNATIONAL JOURNAL OF PLANT SCIENCES

2003). Furthermore, continued progress in understanding the which this article is based; Phil Garnock-Jones and the
diversity and relationships of living and fossil basal angio- School of Biological Sciences, Victoria University of Welling-
sperms and the morphology of other fossil seed plants has ton, for providing a supportive environment and facilities;
brought the nature of the remaining problems into far clearer Dallas Mildenhall and GNS Science, New Zealand, for ac-
focus than was the case 25 years ago. cess to literature; Dior Kelley for discussion of developmen-
tal genetics and ovule evolution; Steffi Ickert-Bond for
Acknowledgments discussion of Ephedra; Dennis Stevenson for information on
cycad cones; Rose Prevec and Steve McLoughlin for unpub-
I wish to thank Thomas Denk for the invitation to present lished observations on glossopterids; and Peter Endress for
a talk at the symposium in honor of Else Marie Friis on valuable comments on the manuscript.

Literature Cited

Adendorff R 2005 A revision of the ovulate fructifications of Glossop- Buzgo M, PK Endress 2000 Floral structure and development of
teris from the Permian of South Africa. PhD diss. University of the Acoraceae and its systematic relationship with basal angiosperms.
Witwatersrand, Johannesburg. Int J Plant Sci 161:23–41.
Anderson JM, HM Anderson 2003 Heyday of the gymnosperms: Buzgo M, PS Soltis, DE Soltis 2004 Floral developmental morphol-
systematics and biodiversity of the Late Triassic Molteno fructifi- ogy of Amborella trichopoda (Amborellaceae). Int J Plant Sci 165:
cations. National Botanical Institute, Pretoria. 925–947.
APG (Angiosperm Phylogeny Group) 1998 An ordinal classification Cantino PD, JA Doyle, SW Graham, WS Judd, RG Olmstead, DE
for the families of flowering plants. Ann Mo Bot Gard 85:531–553. Soltis, PS Soltis, MJ Donoghue 2007 Towards a phylogenetic
——— 2003 An update of the Angiosperm Phylogeny Group clas- nomenclature of Tracheophyta. Taxon 56:822–846.
sification for the orders and families of flowering plants: APG II. Bot Carlquist S 1996 Wood, bark, and stem anatomy of Gnetales: a
J Linn Soc 141:399–436. summary. Int J Plant Sci 157(suppl):S58–S76.
Arber EAN, J Parkin 1907 On the origin of angiosperms. J Linn Soc Chaw SM, TW Walter, CC Chang, SH Hua, SH Chen 2005 A
Bot 38:29–80. phylogeny of cycads (Cycadales) inferred from chloroplast matK
Axsmith BJ, EL Taylor, TN Taylor, NR Cuneo 2000 New perspectives gene, trnK intron, and nuclear rDNA ITS region. Mol Phylogenet
on the Mesozoic seed fern order Corystospermales based on attached Evol 37:214–234.
organs from the Triassic of Antarctica. Am J Bot 87:757–768. Cheadle VI, H Kosakai 1975 Vessels in Juncales. II. Centrolepidaceae
Bailey IW 1949 Origin of the angiosperms: need for a broadened and Restionaceae. Am J Bot 62:1017–1026.
outlook. J Arnold Arbor Harv Univ 30:64–70. Cooke DA 1983 The seedling of Trithuria (Hydatellaceae). Vic Nat
Balasubramanian S, K Schneitz 2000 NOZZLE regulates proximal- (Blackburn) 100:68–69.
distal pattern formation, cell proliferation and early sporogenesis Cornet B 1986 The leaf venation and reproductive structures of a
during ovule development in Arabidopsis thaliana. Development Late Triassic angiosperm, Sanmiguelia lewisii. Evol Theory 7:231–
127:4227–4238. 309.
Barbacka M, K Bóka 2000 The stomatal ontogeny and structure of ——— 1989a Late Triassic angiosperm-like pollen from the Rich-
the Liassic pteridosperm Sagenopteris (Caytoniales) from Hungary. mond rift basin of Virginia, U.S.A. Palaeontogr Abt B Palaeophytol
Int J Plant Sci 161:149–157. 213:37–87.
Bateman RM, J Hilton, PJ Rudall 2006 Morphological and molec- ——— 1989b The reproductive morphology and biology of Sanmi-
ular phylogenetic context of the angiosperms: contrasting the guelia lewisii, and its bearing on angiosperm evolution in the Late
‘‘top-down’’ and ‘‘bottom-up’’ approaches used to infer the likely Triassic. Evol Trends Plants 3:25–51.
characteristics of the first flowers. J Exp Bot 57:3471–3503. Crane PR 1985a Phylogenetic analysis of seed plants and the origin
Beck CB 1970 The appearance of gymnospermous structure. Biol of angiosperms. Ann Mo Bot Gard 72:716–793.
Rev Camb Philos Soc 45:379–400. ——— 1985b Phylogenetic relationships in seed plants. Cladistics 1:
Behnke HD 2000 Forms and sizes of sieve-element plastids and 329–348.
evolution of the monocotyledons. Pages 163–188 in KL Wilson, DA ——— 1987 [Review of] Cornet, B., The leaf venation and repro-
Morrison, eds. Monocots: systematics and evolution. CSIRO, Mel- ductive structures of a Late Triassic angiosperm, Sanmiguelia
bourne. lewisii, Evolutionary Theory 7:231–309 (1986). Taxon 36:778–779.
Bierhorst DW 1971 Morphology of vascular plants. Macmillan, New ——— 1988 Major clades and relationships in the ‘‘higher’’ gymno-
York. sperms. Pages 218–272 in CB Beck, ed. Origin and evolution of
Bose MN, PK Pal, TM Harris 1985 The Pentoxylon plant. Philos gymnosperms. Columbia University Press, New York.
Trans R Soc B 310:77–108. ——— 1996 The fossil history of the Gnetales. Int J Plant Sci
Bowe LM, G Coat, CW dePamphilis 2000 Phylogeny of seed plants 157(suppl):S50–S57.
based on all three genomic compartments: extant gymnosperms are Crepet WL, KC Nixon, MA Gandolfo 2004 Fossil evidence and
monophyletic and Gnetales’ closest relatives are conifers. Proc Natl phylogeny: the age of major angiosperm clades based on mesofossil
Acad Sci USA 97:4092–4097. and macrofossil evidence from Cretaceous deposits. Am J Bot 91:
Bowman JL 2000 The YABBY gene family and abaxial cell fate. Curr 1666–1682.
Opin Plant Biol 3:17–22. Crisp MD, LG Cook 2005 Do early branching lineages signify an-
Bremer K 1988 The limits of amino acid sequence data in angio- cestral traits? Trends Ecol Evol 20:122–128.
sperm phylogenetic reconstruction. Evolution 42:795–803. Cutler DF 1969 Anatomy of the monocotyledons. IV. Juncales.
Burleigh JG, S Mathews 2004 Phylogenetic signal in nucleotide data Oxford University Press, Oxford.
from seed plants: implications for resolving the seed plant tree of Donoghue MJ, JA Doyle 2000 Seed plant phylogeny: demise of the
life. Am J Bot 91:1599–1613. anthophyte hypothesis? Curr Biol 10:R106–R109.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 841

Donoghue MJ, RG Olmstead, JF Smith, JD Palmer 1992 Phyloge- Floyd SK, JL Bowman 2007 The ancestral developmental tool kit of
netic relationships of Dipsacales based on rbcL sequences. Ann Mo land plants. Int J Plant Sci 168:1–35.
Bot Gard 79:333–345. Friedman WE 2008 Hydatellaceae are water lilies with gymnosper-
Doyle JA 1978 Origin of angiosperms. Annu Rev Ecol Syst 9: mous tendencies. Nature 453:94–97.
365–392. Friis EM, P Crane 2007 New home for tiny aquatics. Nature 446:
——— 1994 Origin of the angiosperm flower: a phylogenetic 269–270.
perspective. Plant Syst Evol Suppl 8:7–29. Friis EM, PR Crane, KR Pedersen, S Bengtson, PCJ Donoghue, GW
——— 1996 Seed plant phylogeny and the relationships of Gnetales. Grimm, M Stampanoni 2007 Phase-contrast x-ray microtomogra-
Int J Plant Sci 157(suppl):S3–S39. phy links Cretaceous seeds with Gnetales and Bennettitales. Nature
——— 1998 Phylogeny of vascular plants. Annu Rev Ecol Syst 29: 450:549–552.
567–599. Friis EM, JA Doyle, PK Endress, Q Leng 2003 Archaefructus: an-
——— 2001 Significance of molecular phylogenetic analyses for paleo- giosperm precursor or specialized early angiosperm? Trends Plant
botanical investigations on the origin of angiosperms. Palaeobota- Sci 8:369–373.
nist 50:167–188. Friis EM, KR Pedersen 1996 Eucommiitheca hirsuta, a new pollen
——— 2005 Early evolution of angiosperm pollen as inferred from organ with Eucommiidites pollen from the Early Cretaceous of
molecular and morphological phylogenetic analyses. Grana 44: Portugal. Grana 35:104–112.
227–251. Friis EM, KR Pedersen, PR Crane 2006 Cretaceous angiosperm flow-
——— 2006 Seed ferns and the origin of angiosperms. J Torrey Bot ers: innovation and evolution in plant reproduction. Palaeogeogr
Soc 133:169–209. Palaeoclimatol Palaeoecol 232:251–293.
——— 2007 Systematic value and evolution of leaf architecture Frohlich MW 2003 An evolutionary scenario for the origin of flow-
across the angiosperms in light of molecular phylogenetic analyses. ers. Nat Rev Genet 4:559–566.
Cour Forschungsinst Senckenb 258:21–37. Frohlich MW, DS Parker 2000 The Mostly Male theory of flower
Doyle JA, MJ Donoghue 1986 Seed plant phylogeny and the origin of evolutionary origins: from genes to fossils. Syst Bot 25:155–170.
angiosperms: an experimental cladistic approach. Bot Rev 52:321–431. Galtier J 1988 Morphology and phylogenetic relationships of early
——— 1993 Phylogenies and angiosperm diversification. Paleobiol- pteridosperms. Pages 135–176 in CB Beck, ed. Origin and evolution
ogy 19:141–167. of gymnosperms. Columbia University Press, New York.
Doyle JA, MJ Donoghue, EA Zimmer 1994 Integration of morpho- Gaussen H 1946 Les Gymnospermes, actuelles et fossiles. Trav Lab
logical and ribosomal RNA data on the origin of angiosperms. Ann For Toulouse Tome II Etud Dendrol, sect 1, vol 1, fasc 3, chap 5, 1–26.
Mo Bot Gard 81:419–450. Gifford EM, AS Foster 1989 Morphology and evolution of vascular
Doyle JA, PK Endress 2000 Morphological phylogenetic analysis of plants. WH Freeman, New York.
basal angiosperms: comparison and combination with molecular Gould RE, T Delevoryas 1977 The biology of Glossopteris: evidence
data. Int J Plant Sci 161(suppl):S121–S153. from petrified seed-bearing and pollen-bearing organs. Alcheringa
——— 2007 Integrating Early Cretaceous fossils into the phylogeny 1:387–399.
of Recent angiosperms. Page 150 in Plant Biology and Botany 2007 Hamann U 1975 Neue Untersuchungen zur Embryologie und Sys-
program and abstract book. Botanical Society of America, St. Louis. tematik der Centrolepidaceae. Bot Jahrb Syst 96:154–191.
Doyle JA, CL Hotton 1991 Diversification of early angiosperm pollen ——— 1976 Hydatellaceae: a new family of Monocotyledoneae.
in a cladistic context. Pages 169–195 in S Blackmore, SH Barnes, eds. N Z J Bot 14:193–196.
Pollen and spores: patterns of diversification. Clarendon, Oxford. Hamby RK, EA Zimmer 1992 Ribosomal RNA as a phylogenetic
Eames AJ 1952 Relationships of the Ephedrales. Phytomorphology tool in plant systematics. Pages 50–91 in PS Soltis, DE Soltis, JJ
2:79–100. Doyle, eds. Molecular systematics of plants. Chapman & Hall, New
Eklund H, JA Doyle, PS Herendeen 2004 Morphological phyloge- York.
netic analysis of living and fossil Chloranthaceae. Int J Plant Sci 165: Harris TM 1932 The fossil flora of Scoresby Sound East Greenland.
107–151. 2. Description of seed plants incertae sedis together with a dis-
Endress PK 1987 The Chloranthaceae: reproductive structures and cussion of certain cycadophyte cuticles. Medd Gronl 85:1–112.
phylogenetic position. Bot Jahrb Syst 109:153–226. ——— 1937 The fossil flora of Scoresby Sound East Greenland. 5.
——— 2005 Carpels in Brasenia (Cabombaceae) are completely Stratigraphic relations of the plant beds. Medd Gronl 112:1–114.
ascidiate despite a long stigmatic crest. Ann Bot 96:209–215. ——— 1940 Caytonia. Ann Bot, NS, 4:713–734.
——— 2008 Perianth biology in the basal grade of extant angio- ——— 1951 The relationships of the Caytoniales. Phytomorphology
sperms. Int J Plant Sci 169:844–862. 1:29–39.
Endress PK, JA Doyle 2007 Floral phyllotaxis in basal angiosperms: ——— 1954 Mesozoic seed cuticles. Sven Bot Tidskr 48:281–291.
development and evolution. Curr Opin Plant Biol 10:52–57. Hermsen EJ, TN Taylor, EL Taylor, DW Stevenson 2006 Cataphylls
——— Forthcoming Reconstructing the ancestral flower and its of the Middle Triassic cycad Antarcticycas schopfii and new insights
initial specializations. Am J Bot. into cycad evolution. Am J Bot 93:724–738.
Endress PK, A Igersheim 1997 Gynoecium diversity and systematics Hickey LJ, DW Taylor 1996 Origin of the angiosperm flower. Pages
of the Laurales. Bot J Linn Soc 125:93–168. 176–231 in DW Taylor, LJ Hickey, eds. Flowering plant origin,
——— 2000 Gynoecium structure and evolution in basal angio- evolution and phylogeny. Chapman & Hall, New York.
sperms. Int J Plant Sci 161(suppl):S211–S223. Hilton J, RM Bateman 2006 Pteridosperms are the backbone of seed
Eshed Y, SF Baum, JV Perea, JL Bowman 2001 Establishment of plant evolution. J Torrey Bot Soc 133:119–168.
polarity in lateral organs of plants. Curr Biol 11:1251–1260. Hochuli PA, S Feist-Burkhardt 2004 A boreal cradle of angiosperms?
Eshed Y, A Izhaki, SF Baum, SK Floyd, JL Bowman 2004 Asymmetric angiosperm-like pollen from the Middle Triassic of the Barents Sea
leaf development and blade expansion in Arabidopsis are medi- (Norway). J Micropalaeontol 23:97–104.
ated by KANADI and YABBY activities. Development 131:2997– Hufford L 1996 The morphology and evolution of male reproductive
3006. structures of Gnetales. Int J Plant Sci 157(suppl):S95–S112.
Florin R 1951 Evolution in cordaites and conifers. Acta Horti Igersheim A, PK Endress 1997 Gynoecium diversity and systematics
Bergiani 15:285–388. of the Magnoliales and winteroids. Bot J Linn Soc 124:213–271.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
842 INTERNATIONAL JOURNAL OF PLANT SCIENCES

——— 1998 Gynoecium diversity and systematics of the paleoherbs. Mundry M, T Stützel 2004 Morphogenesis of the reproductive
Bot J Linn Soc 127:289–370. shoots of Welwitschia mirabilis and Ephedra distachya (Gnetales),
Imaichi R, M Kato, H Okada 1995 Morphology of the outer integu- and its evolutionary implications. Org Divers Evol 4:91–108.
ment in three primitive angiosperm families. Can J Bot 73:1242–1249. Nixon KC, WL Crepet, D Stevenson, EM Friis 1994 A reevaluation
Ji Q, H Li, LM Bowe, Y Liu, DW Taylor 2004 Early Cretaceous of seed plant phylogeny. Ann Mo Bot Gard 81:484–533.
Archaefructus eoflora sp. nov. with bisexual flowers from Beipiao, Norell MA 1992 Taxic origin and temporal diversity: the effect
western Liaoning, China. Acta Geol Sin 78:883–896. of phylogeny. Pages 89–118 in MJ Novacek, QD Wheeler, eds.
Kato M 1990 Ophioglossaceae: a hypothetical archetype for the Extinction and phylogeny. Columbia University Press, New York.
angiosperm carpel. Bot J Linn Soc 102:303–311. Pant DD, RS Singh 1974 On the stem and attachment of Glossop-
Kenrick P, PR Crane 1997 The origin and early diversification of land teris and Gangamopteris leaves. II. Structural features. Palaeontogr
plants: a cladistic study. Smithsonian Institution, Washington, DC. Abt B Palaeophytol 147:42–73.
Kerp JHF 1988 Aspects of Permian palaeobotany and palynology. X. Pedersen KR, PR Crane, EM Friis 1989 The morphology and phylo-
The West- and Central European species of the genus Autunia Krasser genetic significance of Vardekloeftia Harris (Bennettitales). Rev
emend. Kerp (Peltaspermaceae) and the form-genus Rhachiphyllum Palaeobot Palynol 60:7–24.
Kerp (callipterid foliage). Rev Palaeobot Palynol 54:249–360. Pigg KB, ML Trivett 1994 Evolution of the glossopterid gymno-
Kidston R 1924 Fossil plants of the Carboniferous rocks of Great sperms from Permian Gondwana. J Plant Res 107:461–477.
Britain. Mem Geol Surv G B Palaeontol 2:379–522. Pryer KM, H Schneider, AR Smith, R Cranfill, PG Wolf, JS Hunt, SD
Kirchner M, JHA van Konijnenburg-van Cittert 1994 Schmeissneria Sipes 2001 Horsetails and ferns are a monophyletic group and the
microstachys (Presl, 1833) Kirchner et Van Konijnenburg-Van Cittert, closest living relatives to seed plants. Nature 409:618–622.
comb. nov. and Karkenia hauptmannii Kirchner et Van Konijnenburg- Qiu YL, L Li, B Wang, Z Chen, O Dombrovska, J Lee, L Kent,
Van Cittert, sp. nov., plants with ginkgoalean affinities from the et al 2007 A nonflowering land plant phylogeny inferred from nu-
Liassic of Germany. Rev Palaeobot Palynol 83:199–215. cleotide sequences of seven chloroplast, mitochondrial, and nuclear
Klavins SD, TN Taylor, EL Taylor 2002 Anatomy of Umkomasia genes. Int J Plant Sci 168:691–708.
(Corystospermales) from the Triassic of Antarctica. Am J Bot 89: Quinn CJ, RA Price, PA Gadek 2002 Familial concepts and relation-
664–676. ships of the conifers based on rbcL and matK sequence compari-
Linder HP, IK Ferguson 1985 On the pollen morphology and sons. Kew Bull 57:513–531.
phylogeny of the Restionales and Poales. Grana 24:65–76. Rai HS, HE O’Brien, PA Reeves, RG Olmstead, SW Graham 2003
Loconte H, DW Stevenson 1990 Cladistics of the Spermatophyta. Inference of higher-order relationships in the cycads from a large
Brittonia 42:197–211. chloroplast data set. Mol Phylogenet Evol 29:350–359.
Maddison DR, WP Maddison 2003 MacClade 4: analysis of phylog- Remizova M, D Sokoloff 2003 Inflorescence and floral morphology
eny and character evolution, version 4.06. Sinauer, Sunderland, MA. in Tofieldia (Tofieldiaceae) compared with Araceae, Acoraceae and
Maddison WP 1994 Missing data versus missing characters in Alismatales s.str. Bot Jahrb Syst 124:255–271.
phylogenetic analysis. Syst Biol 42:576–581. Retallack G, DL Dilcher 1981 Arguments for a glossopterid ancestry
Magallón S, MJ Sanderson 2002 Relationships among seed plants of angiosperms. Paleobiology 7:54–67.
inferred from highly conserved genes: sorting conflicting phyloge- ——— 1988 Reconstruction of selected seed ferns. Ann Mo Bot
netic signals among ancient lineages. Am J Bot 89:1991–2006. Gard 75:1010–1057.
Mapes G, GW Rothwell 1984 Permineralized ovulate cones of Reymanówna M 1973 The Jurassic flora from Grojec near Kraków
Lebachia from late Palaeozoic limestones of Kansas. Palaeontology in Poland. II. Caytoniales and anatomy of Caytonia. Acta Palaeobot
27:69–94. 14:45–87.
Martens P 1971 Les Gnétophytes. Encyclopedia of plant anatomy. Ronse De Craene LP, PS Soltis, DE Soltis 2003 Evolution of floral
Vol 12, pt 2. Borntraeger, Stuttgart. structures in basal angiosperms. Int J Plant Sci 164(suppl):S329–S363.
McAbee JM, TA Hill, DJ Skinner, A Izhaki, BA Hauser, RJ Meister, GV Rothwell GW 1981 The Callistophytales (Pteridospermopsida): re-
Reddy, EM Meyerowitz, JL Bowman, CS Gasser 2006 ABERRANT productively sophisticated Paleozoic gymnosperms. Rev Palaeobot
TESTA SHAPE encodes a KANADI family member, linking polarity Palynol 32:103–121.
determination to separation and growth of Arabidopsis ovule Rothwell GW, KC Nixon 2006 How does the inclusion of fossil data
integuments. Plant J 46:522–531. change our conclusions about the phylogenetic history of euphyllo-
McLoughlin S 1990 Late Permian glossopterid fructifications from the phytes? Int J Plant Sci 167:737–749.
Bowen and Sydney basins, eastern Australia. Geobios 23:283–297. Rothwell GW, R Serbet 1994 Lignophyte phylogeny and the evolu-
Meister RJ, LM Kotow, CS Gasser 2002 SUPERMAN attenuates positive tion of spermatophytes: a numerical cladistic analysis. Syst Bot 19:
INNER NO OUTER autoregulation to maintain polar development of 443–482.
Arabidopsis ovule outer integuments. Development 129:4281–4289. Rothwell GW, RA Stockey 2002 Anatomically preserved Cycadeoi-
Melville R 1963 A new theory of the angiosperm flower. II. Kew Bull dea (Cycadeoidaceae), with a reevaluation of systematic characters
17:1–63. for the seed cones of Bennettitales. Am J Bot 89:1447–1458.
Mendes MM, EM Friis, J Pais 2008 Erdtmanispermum juncalense Rudall PJ 2007 The inflorescence-flower boundary: insights from
sp. nov., a new species of the extinct order Erdtmanithecales from three phylogenetically diverse case-studies. Page 167 in Plant Biology
the Early Cretaceous (probably Berriasian) of Portugal. Rev and Botany 2007 program and abstract book. Botanical Society of
Palaeobot Palynol 149:50–56. America, St. Louis.
Meng SW, AW Douglas, DZ Li, ZD Chen, HX Liang, JB Yang 2003 Rudall PJ, DD Sokoloff, MV Remizowa, JG Conran, JI Davis, TD
Phylogeny of Saururaceae based on morphology and five regions Macfarlane, DW Stevenson 2007 Morphology of Hydatellaceae,
from three plant genomes. Ann Mo Bot Gard 90:592–602. an anomalous aquatic family recently recognized as an early-
Meyen SV 1987 Fundamentals of palaeobotany. Chapman & Hall, divergent angiosperm lineage. Am J Bot 94:1073–1092.
New York. Rutishauser R, R Sattler 1987 Complementarity and heuristic value
Moore MJ, CD Bell, PS Soltis, DE Soltis 2007 Using plastid of contrasting models in structural botany. II. Case study on leaf
genome-scale data to resolve enigmatic relationships among basal whorls: Equisetum and Ceratophyllum. Bot Jahrb Syst 109:227–255.
angiosperms. Proc Natl Acad Sci USA 104:19363–19368. Rydin C, M Källersjö, EM Friis 2002 Seed plant relationships and

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions
 DOYLE—PHYLOGENY, FOSSILS, AND ORIGIN OF THE FLOWER 843

the systematic position of Gnetales based on nuclear and chloro- Taylor DW, G Kirchner 1996 The origin and evolution of the
plast DNA: conflicting data, rooting problems, and the monophyly angiosperm carpel. Pages 116–140 in DW Taylor, LJ Hickey, eds.
of conifers. Int J Plant Sci 163:197–214. Flowering plant origin, evolution and phylogeny. Chapman & Hall,
Saarela JM, HS Rai, JA Doyle, PK Endress, S Mathews, AD Marchant, BG New York.
Briggs, SW Graham 2007 Hydatellaceae identified as a new branch Taylor DW, H Li, J Dahl, FJ Fago, D Zinniker, JM Moldowan 2006
near the base of the angiosperm phylogenetic tree. Nature 446:312–315. Biogeochemical evidence for the presence of the angiosperm molec-
Sanderson MJ, MF Wojciechowski, JM Hu, T Sher Khan, SG Brady ular fossil oleanane in Paleozoic and Mesozoic non-angiospermous
2000 Error, bias, and long-branch attraction in data for two chloro- fossils. Paleobiology 32:179–190.
plast photosystem genes in seed plants. Mol Biol Evol 17:782–797. Taylor EL 1996 Enigmatic gymnosperms? structurally preserved Perm-
Schopf JM 1976 Morphologic interpretation of fertile structures in ian and Triassic seed ferns from Antarctica. Rev Palaeobot Palynol 90:
glossopterid gymnosperms. Rev Palaeobot Palynol 21:25–64. 303–318.
Scotland RW, RG Olmstead, JR Bennett 2003 Phylogeny reconstruc- Taylor EL, TN Taylor 1992 Reproductive biology of the Permian
tion: the role of morphology. Syst Biol 52:539–548. Glossopteridales and their suggested relationship to flowering
Serbet R, GW Rothwell 1992 Characterizing the most primitive seed plants. Proc Natl Acad Sci USA 89:11495–11497.
ferns. I. A reconstruction of Elkinsia polymorpha. Int J Plant Sci Taylor TN, GM Del Fueyo, EL Taylor 1994 Permineralized seed fern
153:602–621. cupules from the Triassic of Antarctica: implications for cupule and
Sieber P, J Gheyselinck, R Gross-Hardt, T Laux, U Grossniklaus, K carpel evolution. Am J Bot 81:666–677.
Schneitz 2004 Pattern formation during early ovule development Thomas HH 1925 The Caytoniales, a new group of angiospermous
in Arabidopsis thaliana. Dev Biol 273:321–334. plants from the Jurassic rocks of Yorkshire. Philos Trans R Soc B
Singh H 1978 Embryology of gymnosperms. Encyclopedia of plant 213:299–363.
anatomy. Vol 10, pt 2. Borntraeger, Berlin. ——— 1933 On some pteridospermous plants from the Mesozoic
Skinner DJ, TA Hill, CS Gasser 2004 Regulation of ovule develop-
rocks of South Africa. Philos Trans R Soc B 222:193–265.
ment. Plant Cell 16(suppl):S32–S45.
——— 1958 Lidgettonia, a new type of fertile Glossopteris. Bull Br
Soltis DE, PS Soltis, PK Endress, MW Chase 2005 Phylogeny and
Mus (Nat Hist) Geol 3:179–189.
evolution of angiosperms. Sinauer, Sunderland, MA.
Townrow JA 1960 The Peltaspermaceae, a pteridosperm family of
Soltis DE, PS Soltis, MJ Zanis 2002 Phylogeny of seed plants based
Permian and Triassic age. Palaeontology 3:333–361.
on evidence from eight genes. Am J Bot 89:1670–1681.
Umeda A, R Imaichi, M Kato 1994 Ovular development and mor-
Stebbins GL 1974 Flowering plants: evolution above the species
phology of the outer integument of Magnolia grandiflora (Magno-
level. Harvard University Press, Cambridge, MA.
liaceae). Am J Bot 81:361–367.
Stevens PF 2007 Angiosperm phylogeny Web site, version 3. http://
Villanueva JM, J Broadhvest, BA Hauser, RJ Meister, K Schneitz, CS
www.mobot.org/MOBOT/research/APweb/.
Gasser 1999 INNER NO OUTER regulates abaxial-adaxial pat-
Stockey RA, GW Rothwell 2003 Anatomically preserved Williamsonia
terning in Arabidopsis ovules. Genes Dev 13:3160–3169.
(Williamsoniaceae): evidence for bennettitalean reproduction in the Late
von Wettstein RR 1907 Handbuch der systematischen Botanik. Vol
Cretaceous of western North America. Int J Plant Sci 164:251–262.
Stuessy TF 2004 A transitional-combinational theory for the origin 2. Franz Deuticke, Vienna.
of angiosperms. Taxon 53:3–16. Wang X, S Duan, B Geng, J Cui, Y Yang 2007 Schmeissneria: a
Sun G, DL Dilcher, S Zheng, Z Zhou 1998 In search of the first missing link to angiosperms? BMC Evol Biol 7:14, http://www
flower: a Jurassic angiosperm, Archaefructus, from northeast China. .biomedcentral.com/1471-2148/7/14.
Science 282:1692–1695. Wiens JJ 2004 The role of morphological data in phylogeny recon-
Sun G, Q Ji, DL Dilcher, S Zheng, KC Nixon, X Wang 2002 Ar- struction. Syst Biol 53:653–661.
chaefructaceae, a new basal angiosperm family. Science 296:899–904. Williams JH, WE Friedman 2004 The four-celled female gameto-
Surange KR, S Chandra 1975 Morphology of the gymnospermous phyte of Illicium (Illiciaceae; Austrobaileyales): implications for
fructifications of the Glossopteris flora and their relationships. understanding the origin and early evolution of monocots,
Palaeontogr Abt B Palaeophytol 149:153–180. eumagnoliids, and eudicots. Am J Bot 91:332–351.
Surange KR, HK Maheshwari 1970 Some male and female fructifi- Yamada T, R Imaichi, M Kato 2001a Developmental morphology of
cations of Glossopteridales from India. Palaeontogr Abt B Palaeo- ovules and seeds of Nymphaeales. Am J Bot 88:963–974.
phytol 129:178–192. Yamada T, M Ito, M Kato 2003 Expression pattern of INNER NO
Svoma E 1997 Seed development and function in Artabotrys OUTER homologue in Nymphaea (water lily family, Nymphaea-
hexapetalus (Annonaceae). Plant Syst Evol 207:205–223. ceae). Dev Genes Evol 213:510–513.
Swofford DL 1990 PAUP: phylogenetic analysis using parsimony, Yamada T, H Tobe, R Imaichi, M Kato 2001b Developmental
version 3.0. Illinois Natural History Survey, Champaign. morphology of the ovules of Amborella trichopoda (Amborellaceae)
Takhtajan AL, ed 1988 Sravrintel’naya anatomiya semyan. Vol 2. and Chloranthus serratus (Chloranthaceae). Bot J Linn Soc 137:
Dvudol’nyye: Magnoliidae, Ranunculidae. Nauka, Leningrad. 277–290.
Taylor DW 1991 Angiosperm ovules and carpels: their characters Zanis MJ, DE Soltis, PS Soltis, S Mathews, MJ Donoghue 2002 The
and polarities, distribution in basal clades, and structural evolution. root of the angiosperms revisited. Proc Natl Acad Sci USA 99:
Postilla 208:1–40. 6848–6853.

This content downloaded from 152.14.136.96 on Tue, 23 Apr 2013 19:29:18 PM

All use subject to JSTOR Terms and Conditions