Content uploaded by James A. Doyle
Author content
All content in this area was uploaded by James A. Doyle on Sep 11, 2014
Content may be subject to copyright.
Seed Plant Phylogeny and the Relationships of Gnetales
Author(s): James A. Doyle
Source:
International Journal of Plant Sciences,
Vol. 157, No. 6, Supplement: Biology and
Evolution of the Gnetales (Nov., 1996), pp. S3-S39
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/2475206 .
Accessed: 11/09/2014 15:21
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 support@jstor.org.
.
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 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
Int. J.
Plant
Sci. 157(6
Suppl.):S3-S39. 1996.
?D 1996
by The University of
Chicago.
All
rights reserved.
1058-5893/96/5706S-0001$02.00
SEED PLANT
PHYLOGENY AND THE
RELATIONSHIPS OF
GNETALES
JAMES A. DOYLE
Section
of
Evolution and Ecology,
University
of
California, Davis, California
95616
Most
phylogenetic analyses of
morphological data
agree that Gnetales are a
monophyletic group related to
angio-
sperms
and Bennettitales.
However, they disagree
on whether these
groups (anthophytes)
are related to
coniferopsids
or to Mesozoic
seed ferns, and
thus on whether the
flowers of
Gnetales
are
primitively
simple or reduced.
Molecular
analyses
indicate that both
Gnetales and
angiosperms
are
monophyletic
but
disagree
on their
relationship.
The conclu-
sion of
Nixon et al.
(1994)
that
Gnetales are
paraphyletic, with
angiosperms nested within
them, is weakly
supported;
when
several questionable
embryological characters are
redefined
in
a neutral manner,
Gnetales are
inferred to be
monophyletic. Jurassic
reproductive structures associated with
linear
leaves and ephedroid
pollen
(Piroconites) consist
of a bract
and a scalelike
sporophyll covered with
Welwitschia-like
microsynangia or
ovules, recalling
the bract-
sporophyll complex of
glossopterids. An analysis of
seed plants
incorporating these fossils and
other new
data links
Gnetales with
Piroconites,
angiosperms
with
Caytonia,
and both
groups (plus
Bennettitales and
Pentoxylon)
with
glossopterids,
making up a clade
called the
glossophytes. These
results imply that
glossophytes originally
had glos-
sopterid-like leaves and
bract-sporophyll complexes, which were
transformed into carpels with
bitegmic
ovules
in
angiosperms,
but
reduced to
single,
terminal
ovules
in
Gnetales;
flowers arose
independently
in
the two lines.
The
common ancestor of
angiosperms
and Gnetales may be as old as
Permian, and some of their shared
advances,
such
as
double fertilization
(without
endosperm formation),
may
have arisen as
adaptations
to seasonal
temperate
climates in
Gondwana.
Introduction
Relationships of Gnetales
have
been almost as con-
troversial as those of the
angiosperms,
a
group
with
which
they
have
often been linked. In the
past
decade,
application of
phylogenetic
(cladistic)
methods to
both
morphological
and molecular
data has led to
progress
on this
topic,
although
at times the
conflicts
may
still
appear
to
outweigh
the
agreements. The
first section
of
this article reviews
current
hypotheses
on the
rela-
tionships
of
Gnetales
and the
results of cladistic anal-
yses
that bear on the
problem. Part of this
section fo-
cuses on an
analysis
by
Nixon
et
al.
(1994),
which
represents the
greatest
departure from
what seemed to
be an
emerging consensus
by
calling into
question
the
monophyly
of
Gnetales. This discussion
draws on an
analysis by
M. J.
Donoghue,
J. A.
Doyle,
and W.
E.
Friedman
(in
preparation)
designed to
evaluate
the
Nixon
et al.
study.
The
second section
presents a new
phylogenetic
analysis
of seed
plants, which
takes
into
account
recently described fossils
that
appear
to be re-
lated to
Gnetales and other new
data.
Hypotheses
on
relationships
of
Gnetales
PRECLADISTIC
VIEWS
Ideas on
the
relationships of
Gnetales have
been
closely
tied with
the
question
of
the
origin
of
angio-
sperms.
The
first two
detailed theories both
maintained
that
the two
groups
are
related,
but in
very
different
ways.
The
theory of Wettstein
(1907),
associated with
the
Englerian school of
angiosperm
systematics, proposed
that Gnetales were
ancestral
to
angiosperms.
In
cla-
distic
terms,
this
might
be
rephrased
as
saying
that
angiosperms
are
nested within
Gnetales,
or
that the
two
groups came from a
common ancestor
that was
Manuscript
received
February 1996;
revised
manuscript
received
July
1996.
typologically
like
Gnetales. Wettstein's
theory
was
tied
to his view that
Amentiferae,
such as
Juglandaceae,
Betulaceae, and
especially
Casuarina (remarkable for
its Ephedra-like
vegetative
appearance),
are the most
primitive
angiosperms
(in
cladistic
terms, basal and
probably
paraphyletic),
and their
aments of unisexual
flowers are
homologous with the
compound strobili of
Gnetales.
Bisexual
flowers
of other
angiosperms
would be
pseudanthia
derived
by aggregation
of sim-
ple, unisexual
flowers.
The
other theory,
that of
Arber and
Parkin (1907,
1908), is
best known
for
proposing that
angiosperms
are related
to
the
Mesozoic order
Bennettitales,
be-
cause
both
groups
have
flower-like
reproductive
struc-
tures. An
apparent
obstacle is
the fact that
Bennetti-
tales
do not have
anything like a
carpel:
the flower
terminates in an
ovuliferous
receptacle
covered with
single,
stalked
ovules,
intermixed
with
protective
in-
terseminal scales
(generally
believed to be
sterilized
ovules;
Harris
1954; Crane
1985).
Recognizing this,
Arber and Parkin
(1907)
did not claim that
angio-
sperms
were derived
from Bennettitales
but, rather,
that the two
groups
came from
a
common
ancestor
with
compound
micro-
and
megasporophylls;
the mi-
crosporophylls
would be
simplified
in
angiosperms,
the
megasporophylls
in
Bennettitales. It is less well
known that
they
thought
Gnetales
are
even more close-
ly
related to
angiosperms (Arber
and Parkin
1908).
However,
Arber
and
Parkin
differed from
Wettstein in
arguing
that the
flowers of
Gnetales are not
primitively
simple
but rather
reduced,
like the
flowers of Amen-
tiferae. One of their
arguments was
the
presence
of a
vestigial ovule in the
male flowers of
Welwitschia (see
Hufford
1996).
In
subsequent decades there
was movement
away
from the idea
that
Gnetales
are
related to
angiosperms.
For
example, Bailey
(1944,
1953) argued
that the ves-
sels in the
two
groups,
originally
considered evidence
S3
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S4 INTERNATIONAL JOURNAL OF PLANT SCIENCES
for a relationship, came from different kinds of tra-
cheids-with scalariform pitting in angiosperms,
and
with
circular bordered pitting
in
Gnetales (best
seen
in
Ephedra).
He
also showed that Gnetales
have
conifer-
like
primary xylem, with
no
scalariform pitting
in the
metaxylem,
a feature found otherwise
only
in
Ginkgo
and
conifers. This
might
be consistent with the
fact
that Gnetales have compound strobili, like the female
cones
of
conifers (Florin 1951). In addition, compar-
ative studies
of new
characters
(wood anatomy, pollen)
implied that the most primitive angiosperms
are so-
called magnoliids (e.g.,
in
having monosulcate pollen,
like many nonangiospermous seed plants), whereas the
amentiferous
groups
that Wettstein considered links
with Gnetales are derived (e.g., in having triporate pol-
len).
More
recently,
these views on relative advance-
ment have been
strengthened by
the
stratigraphic
se-
quence of angiosperm pollen, leaves, and floral
structures
in
the Cretaceous (Doyle 1969; Doyle
and
Hickey 1976;
Crane
et
al.
1995).
A
related
view
was
that of Eames
(1952),
who ar-
gued
that
Gnetales
are
diphyletic.
He
related
Ephedra
to
cordaites,
which
had
compound
strobili made
up
of
bracts and short shoots with bracteoles and
simple spo-
rophylls bearing
ovules
or
microsporangia.
In
Ephed-
ra,
the
ovule would
shift to
an apparent
terminal
po-
sition,
while the bracteoles fused
to
form the
outer
integument. However,
Eames related
Welwitschia
and
Gnetum
to
Bennettitales,
in
part
on
the
basis of
the
work of Florin
(1931)
on
stomata.
Welwitschia
and
Gnetum
have
syndetocheilic stomata,
with lateral sub-
sidiary
cells
inferred
to
be
derived from the same
mother
cell
as
the
guard cells,
a
type
found otherwise
only
in
Bennettitales and
angiosperms (where they
are
usually
described as
paracytic,
defined
on mature form
rather
than
development),
whereas
Ephedra
has
the
more
widespread haplocheilic type.
After
this,
it became
more
popular
to derive
angio-
sperms
from so-called Mesozoic seed
ferns, such
as
the
predominantly Jurassic genus Caytonia.
The
first
attempts (Thomas 1925) homologized
the
multiovulate
cupules of Caytonia with the carpels
of
angiosperms,
but this view foundered on evidence that the
cupules
are
arranged
like leaflets
on
the rachis of
a
sporophyll,
rather than like
leaves
on
a
stem
(Harris 1940).
How-
ever,
Gaussen
(1946),
Stebbins
(1974),
and
Doyle
(1978) proposed
a different
scenario: that the
carpel
was derived from the rachis
of
a
Caytonia megaspo-
rophyll, expanded
into
a
leaflike
structure,
with the
two rows of
cupules
transformed
into
angiosperm
ovules. Reduction
of
the
number of
ovules
per cupule
to one would
produce
a
structure
like
a
bitegmic
anat-
ropous ovule,
with the outer
integument
derived from
the
cupule
wall.
Harris
(1940, 1964)
and
Reymanowna
(1973) provided
evidence that
the
ovules of
Caytonia
are on
the
morphologically
adaxial side
of
the
cupule,
since the orientation of the
cupule implies
that it is
a
leaflet
folded
circinately
toward
the
adaxial side of
the
rachis. This
is consistent with the
position
of the inner
integument plus nucellus (which would correspond to
an individual Caytonia
ovule) relative to
the
outer
in-
tegument
in
the anatropous
bitegmic
ovules of
angio-
sperms (Doyle 1978; Umeda
et
al. 1994; Imaichi
et al.
1995).
Stebbins
(1974)
and Retallack
and
Dilcher
(1981)
proposed
a similar
scheme based on the Permian
glos-
sopterids
of
Gondwana,
which had
one
or more mul-
tiovulate "cupules" attached
to the
adaxial side
of a
leaflike bract. It is unclear
whether the cupules are
appendages
on an axillary shoot
that is
fused to
the
midrib of the bract
or
are morphologically part
of the
bract. In
the
latter case,
the cupule(s) would be anal-
ogous
to the adaxial
fertile
segment
of the
leaf in the
fern
family Ophioglossaceae.
Kato
(1990),
in
fact,
considered
these
structures homologous,
but a direct
relationship between glossopterids
and Ophioglossa-
ceae would involve
convergent origin (or loss)
of
a
prohibitive
number of seed
plant
advances.
In either
case,
studies
of
petrified
material
have
shown
that
the
cupules
are leaflike
in
anatomy, and
the
ovules
are
borne on their
adaxial
surface (Taylor
and
Taylor
1992; Pigg and
Trivett
1994),
rather than the abaxial
side,
as assumed
by
Retallack and Dilcher
(1981).
Since
the
cupules
of
glossopterids
are more leaflike
than the
closely
investing cupules
of
other
groups,
they
will be referred to hereafter as
"sporophylls,"
while
recognizing
that their
status as leaflike
appendages
is
problematic,
and
the
whole structure
will
be
called a
"bract-sporophyll complex."
As in
the
Caytonia
hy-
pothesis,
the
cupules (sporophylls)
would
be trans-
formed
into
bitegmic
ovules
by
reduction
in
ovule
number,
but the
carpel
could be derived from the sub-
tending bract simply by
folding,
with
no need for ex-
pansion.
These proposed homologies
of
angiosperm organs
are not
necessarily
mutually exclusive, since there
are
reasons to
suspect
that
Caytonia
and
glossopterids
are
related.
Not
only may
the
sporophylls
of
glossopterids
be
homologous
with the
cupules of Caytonia (Stebbins
1974),
but
Caytonia
also has leaves with four leaflets
that resemble
individual
simple
leaves
of
Glossopteris,
with
a midrib and
simple
reticulate
venation,
and
both
have saccate
pollen.
These ideas on angiosperms
may
not
seem
directly
relevant to
the
origin
of
Gnetales,
but
cladistic
analyses
and fossil
discoveries
discussed
below
suggest
that all the
groups
under discussion
are
related.
PHYLOGENETIC ANALYSES OF MORPHOLOGICAL
DATA
The
phylogenetic
analyses
of seed
plants by Crane
(1985)
and
Doyle
and
Donoghue (1986;
revised
by
Doyle
and
Donoghue
1992) attempted
to amass all
available characters
bearing
on seed
plant
relationships
and to
sort out
their
implications using
the
principle
of
parsimony.
One most
parsimonious
tree derived
from the data set
of
Doyle
and
Donoghue (1992)
is
shown
in
figure
1. This tree
represents
a
synthesis
of
the views of Arber
and
Parkin
(1907, 1908),
Gaussen
(1946),
and Stebbins
(1974).
Gnetales, angiosperms,
Bennettitales,
and
the
Early Cretaceous genus Pentox-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC RELATIONSHIPS OF GNETALES
S5
RNTHOPHYTES
CONIF HzSF
GNETS
Z
z
_
oe
z
z
cz J f
Treelenqth:
1 1 2
| Sporophy11s
Treelength:\\/
I 1
2
Mega comp
\\\V/ F1~~~~~~~sim mcom
/ _
~~~~~~~~~~~both
simple
Fig. I Representative most parsimonious
seed
plant tree
of
Doyle
and
Donoghue (1992),
with
shading showing
distribution
of the
spo-
rophyll character
and
diagrams summarizing
floral
morphology (Doyle 1994).
ELKI
=
Elkinsia,
MEDU
=
medullosans,
CALL
=
Callisto-
phyton, GINK
=
Ginkgoales,
CONI
=
conifers,
CORD
=
Cordaitales,
CYCA
=
Cycadales,
PELT==
Peltaspermum,
CORY
=
corystosperms,
GLOS
=
glossopterids,
CAYT
=
Caytonia,
ANGI
=
angiosperms,
PENT
=
Pentoxylon,
BENN
=
Bennettitales,
EPHE
=
Ephedra,
WELW
=
Welwitschia,
GNET
=
Gnetum,
CONIF
=
coniferopsids, MzSF
=
"Mesozoic
seed
ferns,"
GNETS
=
Gnetales.
ylon form a
clade,
named anthophytes because they all
have
flower-like reproductive structures. However,
the
anthophytes
are linked with
Caytonia
and
glossopter-
ids, implying that their common ancestor had Cayton-
ia-like sporophylls
and
cupules. The megasporophylls
would
be modified into carpels
in
angiosperms but
re-
duced to
single ovules
in
Bennettitales and Gnetales.
Flowers would be reduced still further
in
Gnetales: to
a
perianth
and
simple microsporophylls in the
male
flower,
to a
single
ovule
surrounded by
an
outer integ-
ument derived from the
perianth
in
the
female
flower.
Although these results form an appealingly coherent
scheme, Doyle
and
Donoghue (1986) already recog-
nized that they were far
from
definitive.
First, experiments designed to test alternative hy-
potheses,
conducted
by forcing
different sets
of taxa
together as clades, showed that many different
trees
are
almost
equally parsimonious.
These trees have
very different implications for floral homologies and
scenarios
of
floral evolution, explored in detail by
Doyle (1994).
In one set of
alternative trees, some
of
which were
only one step longer than the shortest trees, angio-
sperms were linked directly with either Caytonia
(Doyle
and
Donoghue 1986)
or
Caytonia plus glos-
sopterids (Doyle
and
Donoghue 1992),
and the result-
ing
clade
was linked with other anthophytes (fig. 2a).
Such
trees
may
be taken to
mean
that
anthophytes are
diphyletic,
or that
Caytonia and/or glossopterids
are
anthophytes, but
in
either case they contradict
the
orig-
inal anthophyte concept, since there is no evidence that
the sporophylls of Caytonia or glossopterids were ag-
gregated into flower-like structures. Either the flower
originated twice or it originated once and was lost in
Caytonia
and
glossopterids.
Still more different
were
trees that
Doyle
and Don-
oghue (1986) called "neo-englerian," in which antho-
phytes were linked with coniferopsids rather than Me-
sozoic
seed
ferns
and Gnetales
were basal
in
anthophytes.
These
were two steps longer
than the
shortest trees
in terms
of
the data set of
Doyle
and
Donoghue (1986),
but
only
one
step longer
in terms
of
Doyle and Donoghue's
later
study (1992; fig. 2b).
This arrangement
would
imply that the first anthophy-
tes had
axillary
fertile
shoots with
simple sporophylls
grouped into compound strobili,
as in
Gnetales and
conifers,
and
that more
complex
flowers
and
sporo-
phylls
in
angiosperms and Bennettiales were derived.
The
similarities between Gnetales
and
coniferopsids-
not only
the
strobili, but also
the
simple,
linear
leaves
and
xylem
features-would be
homologies
that were
reversed in other
anthophytes,
rather than conver-
gences,
as was the case in the most
parsimonious
trees.
Trees of a more or less neo-englerian type were also
found by Rothwell and Serbet (1994), who made
im-
portant contributions by including
more
taxa, new
characters of
primitive seed plants and coniferopsids,
and
new
observations on
Pentoxylon. However,
their
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S6 INTERNATIONAL JOURNAL OF PLANT SCIENCES
Conifero
Gnetales
CD,e LI Ca, a z
_
) C,z z
_
L
i,
0
Qa
t
3
-J
<
Z
L a
W
Z
El
El
E3 E3 E3
E
EW E3
E3
W
E E N
*
M E M
Cupule
ifilyginopterid
Treelenth: 113
none
anatropous
orthotrop
Conifero Gnetales
CD,
Ca _j
_j CD, V) z Ca z z z
tv~~~~~~~,
m m MEE3
m
03 EQ E
L
E
Leaf type
b
flall
dichot
cata+dichot
Treelength:
113
simple pinn
pinn
comp
E
I Iequivocal
Fig.
2
Alternative
slightly
less
parsimonious
trees
(one step longer
than the
shortest
trees)
based on
the
data
set of
Doyle
and
Donoghue
(1992). a, Angiosperms
linked
directly
with
Caytonia
and
glossop-
terids, showing
distribution of the
cupule
character.
b,
"Neo-engler-
ian"
tree, with anthophytes nested
within
"coniferopsids"
and
Gne-
tales basal
in
anthophytes,
showing
distribution of the leaf
character.
PRO
=
progymnosperms, DSF
(Devonian
seed
fern)
=
Elkinsia,
MED
=
medullosans, CAL
=
Callistophyton, CON= conifers, CRD
=
Cordaitales,
GIN
Ginkgoales, Conifero
=
coniferopsids,
CYC
=
Cycadales,
PEL
Peltaspermum,
CRS
corystosperms,
GLO
=
glossopterids,
CAY
=
Caytonia,
ANG
angiosperms,
PEN
=
Pentoxylon,
BEN
=
Bennettitales,
EPH
=
Ephedra,
WEL
=
Wel-
witschia, GNE= Gnetum.
trees differed
in
that
Pentoxylon
and Bennettitales
were basal
in
anthophytes and Gnetales were
directly
linked with
angiosperms.
Furthermore,
in
many
of
their most
parsimonious
trees
(fig. 3),
extant conifers
(the
sister
group
of
anthophytes)
were
separated
from
the Paleozoic conifer
Emporia, cordaites,
and
gink-
goes.
All
these
analyses yielded
the same
relationships
within
Gnetales,
with
Ephedra
the sister
group
of
Wel-
witschia and
Gnetum.
The
latter
were
linked
by syn-
apomorphies
such as more than one
order
of laminar
venation,
reduction of the
microgametophyte
to four
nuclei, tetrasporic
megagametophyte, free-nuclear egg,
and
a
feeder in the
embryo.
Second,
all the
analyses
discussed so far
pose
the-
oretical
problems
in
treating angiosperms
as
a
single
taxon. This
led
to
problems
in
scoring
the
group,
since
angiosperms vary
in
many
of
the characters
used,
such
as
spiral
vs.
opposite
leaves and
orthotropous
vs. anat-
A .4
- E s -
Gnetales
CLX. m - X0=X^a =L=D
Treelenqth:
191
| ,_ * veined
\// 10_0
~~~~~~dichotomous
Fig.
3 Representative most parsimonious tree
of
Rothwell
and
Serbet (1994), showing distribution of the leaf character.
ropous ovules. These variations required judgments as
to which
states
were ancestral. Doyle and Donoghue
(1986,
1992) scored
angiosperms
as a composite
of
Magnoliales and Winteraceae, which were widely
as-
sumed to be primitive at the time. However, subse-
quent developments
have put this assumption in more
and more doubt.
Even
if Amentiferae are rejected as
primitive angiosperms,
on the basis of such features as
triporate pollen, there are
magnoliids
with monosulcate
pollen that might suggest a very different angiosperm
prototype than Magnoliales and Winteraceae. The
most remarkable of these are Chloranthaceae, various-
ly associated with Laurales or Piperales, which
resem-
ble Gnetales in having opposite leaves, extremely sim-
ple flowers in spikes, and orthotropous rather
than
anatropous ovules. Chloranthaceae are particularly rel-
evant because fossil pollen, leaves, and fruits related
to
them are common in the first phases
of
the
Early
Cretaceous
angiosperm
record (Muller 1981;
Up-
church 1984; Walker and Walker 1984; Doyle
and
Hotton 1991; Pedersen et al. 1991; Crane et al. 1995).
If
the
first angiosperms were like Chloranthaceae, it
might
be
most
parsimonious
to
associate them directly
with Gnetales. Furthermore, molecular analyses dis-
cussed below (Hamby and Zimmer 1992; Chase et al.
1993; Doyle et al. 1994) place herbaceous rather than
woody magnoliid groups, such
as
Nymphaeales
or
Ceratophyllum,
at the base of the angiosperms.
A potential solution to
this
problem is to perform a
seed plant analysis that includes
a
sufficient sampling
of angiosperm taxa. There have been two such studies
based on morphology: by Doyle
et
al. (1994) and by
Nixon et al. (1994).
Doyle et al. (1994)
included
nine angiosperm taxa,
chosen to represent competing ideas on ancestral states
and to include enough intermediate groups to obtain
correct relationships within angiosperms. The shortest
trees were of the
previously
less parsimonious type
with glossopterids and Caytonia linked directly to an-
giosperms, and with Bennettitales,
Pentoxylon,
and
Gnetales
linked to this
lade.
The basal split in angio-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC RELATIONSHIPS
OF GNETALES
S7
sperms was into paleoherbs (herbaceous magnoliids,
such as Nymphaeales and Piperales, and monocots)
and woody magnoliids plus eudicots (groups with
tri-
colpate and derived pollen, which include about 95%
of dicot
species; Doyle and
Hotton
1991). However,
one step less parsimonious ("one-off")
trees showed
a variety of other arrangements, including some of the
type
found
by Doyle and Donoghue (1986, 1992), with
Caytonia
the sister
group
of
anthophytes
and Gnetales
linked with Bennettitales, and neo-englerian trees, with
anthophytes
linked with
conifers
and
Gnetales basal.
Relationships within angiosperms were highly vari-
able: sometimes Magnoliales were basal,
sometimes
Nymphaeales,
but
interestingly
not Chloranthaceae.
Still,
all these trees
indicated
that
both
angiosperms
and
Gnetales
are
monophyletic groups
and
the two are
each other's closest living relatives.
The analysis of Nixon et al. (1994) represents
a
greater departure from previous studies,
since it con-
tradicts the conclusion that
Gnetales
are
monophyletic.
Most
of their trees were
neo-englerian,
in that the clos-
est relatives of
anthophytes
were conifers and other
coniferopsids,
rather than
Caytonia
and
glossopterids.
The
exceptions
were trees
in
which other
Mesozoic
seed ferns
(corystosperms, peltasperms)
were inter-
polated
between conifers and
anthophytes.
In some
trees
(fig. 4a), angiosperms
were
nested
within Gne-
tales,
as the sister
group
of
Welwitschia
and
Gnetum.
The
basal branch
in
angiosperms
was
Chloranthus,
followed
by Ceratophyllum.
This
supports
the view
that
it was
wrong
to
assume
that the first
angiosperms
were like
Magnoliales
and
Winteraceae,
and it
sug-
gests
that the
conclusion
that
Gnetales
are
monophy-
letic
was
a
function
of this error. In other trees
(fig.
4b), angiosperms
were
again
associated
with
Welwit-
schia and
Gnetum,
but
Ephedra
was
lower,
below
Ben-
nettitales,
and the basal
angiosperm
line was Casua-
rina,
followed
by
other
hamamelids, recalling
the
views
of
Wettstein
(1907).
The
position
of
Pentoxylon
was
unstable:
in
trees
of the first
type,
it was basal in
anthophytes,
but
in
the second
type
it
was much
lower,
below cordaites.
ANALYSES OF MOLECULAR DATA
At this
point,
molecular
phylogenetic analyses
of
seed
plants
become relevant.
Since
molecular
analyses
include
only living plants, they
cannot decide
among
the
various seed
plant
trees
found
prior
to Nixon et
al.'s
(1994) study,
in which
angiosperms
were
always
the
closest
living
relatives
of
Gnetales.
These
trees dif-
fered
primarily
in
how fossil taxa such as
Caytonia,
glossopterids,
and Bennettitales fit
in
around the two
extant
groups (the radically
different scenarios for flo-
ral
evolution
implied by
these trees illustrate the im-
portance
of fossils in
reconstructing
character evolu-
tion,
even
when
relationships
of
extant
groups
are not
affected; Doyle
and
Donoghue 1992; Doyle 1994).
However,
molecular data do bear on whether
extant
Gnetales are a
monophyletic group,
as well as on ar-
rangements
within them and within
angiosperms.
The
Conifers
Gnet
FRngiosperms
H ~
-, ,l ,
cc
@,mn=F f
,
a
,
= K'=3X
s
< SJt
>.>J<7t o1
M>
"!
"V
6
1
i
L.
L,
t>
Vo Vol
LLJ 3
r
-!
>t >mt aJLt
C
C0.
L.=
=,
C
O
0
G.=
VI
-0
4,
= QF
n
(1994) a,
Eper th site
gru of
agosem
plu
Mewtci
that angiosperms
are
relte to
Gntls but
the do
.22
M
41
2 2
0
1
groups
are monophyletic.
2heoretically,
t
L.
of
Conifers
Gnet
Fbngiosperms
roplast-
g
r
t
-
a
4
ofg 4hs
Representative mos
pasioious treelns
of r Nionot-al
(9
a,
ph
th
sistae
goo
angiosperms pwelwimoch
andyGeticsimr,rop
and
Chiatranthu
ws
basal inanisem.bEhdrb-
lngowsenetiaers,
Andv
Cauaina
banisalrm
ianosperms.t
on
anlyses puished so far oli
n
all sort
the
vie
othat
angiospermnaeielte
to
Gneeuictales
buth Aethfeyrdo
netradwicthNixn
eth
alate.
(1994)in
indicatngithatebot
gr
oups4
ar
monophyleticm
4'
mL.'T.Thereicll,
the reul of
=vc
Nixoin
etuar.
mightbedutoe
to their inclupsi of
fosilrs,
whic
rcan
overrule
relationshipsf based
on modern etax
(Dnohu
et.
al. 1989 h v
Atere
iooreason
to
ets
dande
M
a
,OneLst
o stdie
has.
use
seuecs
of3
the chlo- .
ropas geerc.I the 475 an 9-aonaaye
oFig
Chs
etpal.e(1993),
ins whsiichusetedelns
wfNxnere
root
ed9).a
withcycads
Gnesitaes
andpo
angiosperms
plsWereitmono-
phylGeticsitrmrop
and
Chratphylumhws
basal
inanisem.bEpdrb-
angospenetiaers.
abov
thuais,baali
angispemssplitintson
anadyes
consisting sof
magndoliids
and
monocots
and
an-w
other
correspondin
toe
rlthed
euiots
wiethlsu Aetiferaeo
neteadwicthNixn eth alate.
Gnetalesn
andicatngiospermst
wereon eturn. relatbedutoe tof
twoi grolupsio
of consifes.
Anic
bcLn
studyrofe smlallernsetiof
taxad
by
Hasebe
etax
a.(Do992b),e
in.
which
seedvplantshere
iroot edawit two
fenesetand Marhaniehas
anedDN
sequences
werte chon-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S8
INTERNATIONAL
JOURNAL OF PLANT
SCIENCES
verted
into amino
acid
sequences, differed
in indicat-
ing that
angiosperms
and
gymnosperms as a
whole
are
sister
groups. Gnetales
were
nested within or
were the
sister
group
of other
gymnosperms,
depending
on
the
method of
analysis (maximum
likelihood
or
neighbor
joining,
a
distance method that
attempts to
compensate
for
unequal
evolutionary rates).
However, both
Gne-
tales and
angiosperms
were
again
monophyletic.
Has-
ebe et
al.
(1992b) performed
bootstrap
analyses
to
evaluate the
strength
of
these conclusions. Gnetales
were
monophyletic
at the
98%
bootstrap
level,
but
their
relationship
to
other
gymnosperms
was more
weakly
supported
(63%).
Curiously,
bootstrap support
for
monophyly
of
the
four
angiosperm
taxa was
only
50%.
Other
studies have been
based
on
nuclear
rDNA
se-
quences.
In
the
trees of
Hamby
and Zimmer
(1992)
and
Doyle
et al.
(1994),
in
which
Nymphaeales were
basal in
angiosperms, both
angiosperms
and
Gnetales
were
monophyletic,
but the
relationship
between them
was not
firmly
established.
In
the
60-taxon
analysis
of
Hamby
and Zimmer
(1992),
in
which seed
plants
were
rooted with
Equisetum
and
Psilotum,
the
sister
group
of
angiosperms was
cycads,
conifers, and
Ginkgo
when the data were
analyzed by
parsimony,
but
this
was favored over a sister
group
relationship
of
angio-
sperms
and
Gnetales
by only
one
step. Gnetales
and
angiosperms
were
sister
groups
when the
data were
analyzed
by neighbor
joining.
However,
in
a
parsi-
mony
analysis
of
the 71-taxon data set that
formed the
starting point
for
the
study
of
Doyle et
al.
(1994),
Gne-
tales and
angiosperms
were sister
groups,
although
this
was
favored over
the alternative
arrangement
by only
two
steps.
In
bootstrap analyses
of
this data
set,
in
which
49
of the
original
taxa
were
combined into
18,
Doyle
et al.
(1994)
found that
Gnetales were mono-
phyletic at the 99%
level, exceeded
only
by angio-
sperms
(100%).
No
pteridophytic taxa were
used
as
outgroups,
but when
seed
plants were rooted
with cy-
cads,
angiosperms
and
Gnetales
were
united at the
88%
bootstrap level,
one of
the
highest
values
in
the
tree.
Analyses of
chloroplast
rDNA
ITS
sequences (Go-
remykin
et al.
1996)
using neighbor
joining,
in which
seed
plants
were rooted with 16
pteridophytic
taxa and
Marchantia,
differed
in
associating
Gnetales with
Pi-
nus
(the
one conifer
included)
rather than
angiosperms,
although
the
bootstrap
support
for
this
relationship
was
fairly
weak
(50%-56%,
depending
on the
distance
measure
used).
However,
both
angiosperms and
es-
pecially
Gnetales were
strongly
supported
as mono-
phyletic
groups:
bootstrap
values
were
90%-92%
for
angiosperms
and
99%-100% for
Gnetales.
Interesting-
ly,
these
analyses agreed
with
those based on
nuclear
rDNA
(Hamby and
Zimmer
1992; Doyle et al.
1994)
in
placing
Nymphaea
at
the base
of
the
angiosperms.
Most
of these studies
are
consistent with each
other
and with
morphological
analyses
concerning
relation-
ships
within
Gnetales,
with
Ephedra the sister
group
of
Welwitschia and Gnetum. An
exception
is the anal-
ysis
of
Hasebe et
al. (1992b), based on rbcL DNA
data
converted to
amino
acid
sequences, which indicated
that Gnetum
is
the sister
group
of
Ephedra
and Wel-
witschia.
However,
when
Hasebe
et al.
(1992a)
ana-
lyzed
the original DNA
sequences, they obtained the
standard
result.
EVALUATION OF GNETALEAN
MONOPHYLY
The results
of
Nixon
et
al.
(1994) deviate
sufficient-
ly
from
those
of
other
analyses
to
require special
con-
sideration.
The
following
discussion summarizes a
more
detailed examination of
Nixon
et
al.'s data set
by
M. J.
Donoghue,
J. A.
Doyle,
and W. E.
Friedman
(in
preparation).
First,
it
appears that
morphological support
for the
conclusion that
Gnetales
are
not
monophyletic
is
weak.
This
was
already suggested by the
study
of
Albert et
al.
(1994),
who combined the Nixon
et
al.
(1994)
data
set with rbcL data
for
the
same taxa. In trees
based
on
standard nucleotide
data, Gnetales
and
angiosperms
were
monophyletic,
and
Gnetales were linked with
Bennettitales
rather than
angiosperms.
More remark-
ably,
the
outgroups
of
anthophytes
were not conifer-
opsids
but
Caytonia and
glossopterids.
This
may
seem
paradoxical, since
there
are no
molecular data
on
Cay-
tonia.
However,
it
can
be
explained
if
molecular data
favoring
the
monophyly
of
Gnetales
are
strong enough
to overcome
the
morphological data
and
pull
Gnetales
together
into
a
clade,
as
a
result
forcing
angiosperms
lower
in the
anthophytes.
Then
angiosperms
can
affect
basal
states
in
anthophytes,
which
they
could not
when
they
were nested within
Gnetales.
If
this means that
basal
anthophyte
states are more
like those of
Cayton-
ia,
the
result
becomes understandable.
M. J.
Donoghue et al.
(in
preparation)
did
bootstrap
and
decay analyses
of
the Nixon et
al.
(1994)
data set
in
order to test
the strength of
their
conclusions, and
they
tested the relative
parsimony
of
alternative
hy-
potheses
by
using
the
constraints
option
in
PAUP
(Phylogenetic
Analysis Using
Parsimony; see
Swof-
ford
1990).
The
bootstrap
value of
the
angiosperm-
Welwitschia-Gnetum clade
is
only 54%,
and it
breaks
up (decays)
in
trees that are
only
one
step
less
parsi-
monious than
the
shortest trees.
In these
one-off
trees,
angiosperms
are
linked with
Bennettitales,
and
this
clade
is
linked with
Welwitschia
and
Gnetum;
the
shortest
trees
in
which
Gnetales are
monophyletic
are
two off. Other
experiments
bear
on
relationships
of
anthophytes
as
a
whole: it "costs"
only
one
step
to
link
anthophytes
with Caytonia and
glossopterids
rath-
er
than with
conifers.
Second,
the Nixon et al.
(1994)
result is a function
of
questionable
analysis
of
several characters.
Nixon
et al.
criticized
previous
studies for
theory-laden
char-
acter definitions that
depend
on
complex
evolutionary
hypotheses,
sometimes
rightly
so.
However,
many
of
their
characters
have similar
biases.
If
these
characters
are
redefined in a more
neutral
fashion,
the
hypothesis
that Gnetales
are
monophyletic
becomes
more
parsi-
monious.
The most critical
characters are the
five
syn-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS OF
GNETALES S9
apomorphies that unite angiosperms,
Welwitschia, and
Gnetum
(problems
with other characters are discussed
in the appendix).
Two synapomorphies concern
microgametophytes.
First is the number of nuclei: five or more in
Ephedra
and
other groups (one or two prothallials, a
tube nu-
cleus,
a
stalk cell, and two sperm); four in
Welwitschia
and
Gnetum;
and three in
angiosperms (a
tube
nucleus
and two sperm). The reason that this
character unites
Welwitschia
and
Gnetum
(four nuclei)
with
angio-
sperms (three nuclei) is that Nixon et al.
(1994)
or-
dered
the
states, so
that a transition from five
nuclei
to three
takes
two
steps. The second
synapomorphy
is
loss of the stalk cell.
The first problem is that there are
apparently two
ways
to
go
from five nuclei to four: one seen
in
Wel-
witschia
and
Gnetum, the other
in
Taxodiaceae and
other advanced conifers
(Sporne 1965; Singh
1978).
In
development
of
the microgametophyte,
each
divi-
sion
gives
rise to a
sterile nucleus and another that
ultimately produces
the
sperm.
In
Taxodiaceae,
the
tube nucleus is
produced by
the first
division,
implying
that
the
prothallial
was lost and the stalk nucleus
is
still
present.
In
Welwitschia
and
Gnetum,
the tube
nu-
cleus is
produced by
the
second
division,
implying
that
the first sterile nucleus is a
prothallial
and
the
stalk
cell is
missing. However,
there
is no
reason
to
assume
that nuclei
in
angiosperms
were lost
in the same
order
as
in
Welwitschia and
Gnetum,
rather than as
in
co-
nifers,
or
by
a
single step. Second,
if
the
nuclei are
identified
correctly,
loss
of
the stalk cell in
Welwit-
schia and
Gnetum
is
synonymous
with reduction to
four
nuclei,
and
angiosperms
would
inevitably
have
lost
the stalk cell
in
going
to three
nuclei, whether
by
the same route or a different one. Whether this sort of
nucleus-by-nucleus argument
for
homologies
is valid
can be debated.
However,
if
it is not
accepted,
there is
even less basis for
ordering numbers
of
nuclei or re-
taining
the stalk cell character. These
arguments
imply
that number of nuclei
is
best
treated
as three
unordered
states,
and
the
stalk cell
character
should be
eliminated
as
redundant.
Two
other
synapomorphies
involve
megagameto-
phyte
characters: loss
of
archegonia,
and
a
shift
from
alveolar to nonalveolar
development.
These are related
to a
third
character,
free-nuclear
egg,
seen
in
Welwit-
schia and Gnetum.
Ephedra
retains
the basic
seed
plant condition,
with
a
large gametophyte
and normal
archegonia; early development
is
free-nuclear,
but then
cell walls form
around
the
nuclei.
Presence or
absence
of
archegonia
is a
temptingly objective
character,
but
if
associated structures
are
considered the conditions
in
angiosperms
and
Welwitschia-Gnetum
actually ap-
pear very
different. The
highly
reduced
embryo
sac of
angiosperms
has no
identifiable
archegonia,
but
the
egg
is a
normal
cell,
flanked
by
the
synergids,
whereas
in
Welwitschia
and
Gnetum the base
of
the
megaga-
metophyte
is
cellular and
the
apex
free-nuclear,
and
free nuclei function
directly
as
eggs.
Lack of
arche-
gonia may
be
a
function
of the
pattern
of
cellulariza-
tion in the two
groups:
it
is
truncated
relative to
prim-
itive seed
plants in both, but it proceeds from both
ends in
angiosperms, and from the base in Welwitschia
and Gnetum. There is no basis for
saying
whether one
of these
states
preceded the
other or
both
came
inde-
pendently from
the
basic state.
If
so,
absence of
ar-
chegonia and
free-nuclear egg
are
best treated
as as-
pects
of an
unordered three-state megagametophyte
character:
cellular, with normal archegonia; apex and
egg
free-nuclear; center free-nuclear, egg cellular, but
no neck cells.
The alveolar character
expresses
the fact that cell
walls
form
regularly
around individual nuclei
in
most
seed plants,
but irregularly around several nuclei
in
Welwitschia and
Gnetum, resulting
in
multinucleate
cells that later become
polyploid.
It seems valid to
treat
this
as
independent
of the
previous character,
but
it is questionable whether
angiosperms
should be
scored
like
Welwitschia
and
Gnetum. Antipodals
and
synergids
are
derived
by centripetal
formation of cell
walls around
single nuclei,
as
in
primitive
seed
plants.
A less
ambiguous
distinction
might
be
between
for-
mation
of
uninucleate and multinucleate
cells,
which
would contrast
Welwitschia
and Gnetum with other
seed plants,
including angiosperms.
The fifth
putative
synapomorphy
involves
embryo-
genesis, which
is
primitively
free-nuclear
in
seed
plants but cellular
in
angiosperms, Welwitschia,
and
Gnetum.
Here the
problem
is
Ephedra,
in which
eight
free
nuclei
are
produced
after
fertilization, but
each of
these forms an
embryo by
cellular divisions. The old
interpretation
(accepted by Doyle and Donoghue 1986)
is
that this
is
a retention of free-nuclear
embryogene-
sis.
However,
Friedman
(1992, 1994)
showed that the
free
nuclei
are
actually
derived from two
fertilizations,
and
he
regarded
the situation
in
Ephedra
as an
aut-
apomorphy.
However initial events in
Ephedra
are
in-
terpreted,
its
later
embryogenesis
is
more like that of
angiosperms,
Welwitschia,
and
Gnetum.
A less
prob-
lematic distinction is whether each
embryo
is derived
from several
free
nuclei or a
single
uninucleate
cell,
which
would
associate
angiosperms
and all
three
Gne-
tales.
Only
a few conservative
changes
in
these
characters
are
needed
to obtain trees
in
which Gnetales are mono-
phyletic.
For
example,
if the
stalk cell character
is
eliminated and
Ephedra
is scored
as
unknown
for em-
bryogeny,
trees with Welwitschia and Gnetum linked
with
angiosperms
and trees with
Gnetales
monophy-
letic become
equally parsimonious. If,
in
addition,
ei-
ther the
microgametophyte
character is redefined as
unordered
or
angiosperms
are
rescored
as
unknown
for
megagametophyte cellularization, monophyly
of
Gne-
tales is
favored. When all the characters
in
question
are
reinterpreted
along
the
lines
suggested,
trees in
which
Gnetales
are
monophyletic
are favored
by
two
steps,
and
the
bootstrap
value for
Gnetales
is
75%.
This is
not
overwhelming,
but
it is
the
eighth-highest
percentage
on
the tree-for
example,
it is the same as
the
bootstrap
value
for
cycads,
a
group
whose mono-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S10 INTERNATIONAL
JOURNAL OF PLANT SCIENCES
phyly has rarely been questioned.
In some most
par-
simonious trees, Caytonia and glossopterids are the
closest relatives of anthophytes, although not
in
all.
These results imply that the analysis of Nixon et
al.
(1994) does not strongly contradict other evidence that
Gnetales are monophyletic, and that the different
re-
sults of previous analyses are not simply a function
of
assumptions about the first angiosperms.
These observations leave unresolved the relation-
ships of anthophytes as a whole, whether to Mesozoic
seed ferns or to coniferopsids. Nixon et al. (1994)
re-
jected one of the characters of Crane (1985) and Doyle
and Donoghue (1986, 1992) that favored
a
link with
Mesozoic seed ferns, the potential homology
of
the
reflexed cupule of Caytonia with the anatropous
bi-
tegmic
ovule of
angiosperms.
Based
on
valuable new
observations
on
fossils,
Nixon et al.
argued
that the
"stoma" of the
Caytonia cupule (which
should cor-
respond to the part of the angiosperm micropyle
formed
by the outer integument,
or
exostome)
is
formed
by both the lip
of
the cupule
wall and the cu-
pule stalk,
whereas
the
outer
integument
of
angio-
sperms
forms
a
continuous
ring, "except
in
such cases
when the funiculus
is
adnate
laterally
to
the
(outer-
most) integument" (p. 497). However, the exceptions
are
not
trivial: they are closely comparable
to the con-
dition
in
Caytonia,
and
they
occur
in
groups
that are
likely to be basal in angiosperms. In many angio-
sperms, as illustrated by
Robinson-Beers et al.
(1992)
in
Arabidopsis,
each
integument
does
grow up
as
a
complete ring around
the
nucellus,
and
the
whole
struc-
ture is
reflexed,
such
that the
outer
integument passes
between the inner
integument
and the
funicle.
In
con-
trast, Umeda et
al.
(1994)
and Imaichi et al.
(1995)
showed that the
outer integument
in
four families
of
woody magnoliids grows out as
a
U-shaped
hood on
one
side,
such
that it
goes only partway
around the
ovule,
and the
exostome
is
formed by the outer integ-
ument on one side and
the funicle
on
the
other,
like
the
stoma of
Caytonia.
This distinction was
recognized by
Taylor (1991)
as
the "apo" and "syn" states
of his
"continuity
between the
integument
and funiculus"
character
(with
intermediate "'semi" and "hemi"
states). Taylor's systematic survey
indicates
that
the
"syn"
state
predominates
in
anatropous magnoliids
and
monocots,
the
"apo"
state
in
the eudicot clade.
RELEVANCE
OF FOSSIL
GNETALES
Firmer answers to
these questions may depend
on
reconstruction
and incorporation
of
new fossil relatives
of
Gnetales. This
can
be
put
in
terms of the distinction
between the
crown
group,
which includes the
most
re-
cent common
ancestor
of
the living members of a
group
and
its
derivatives,
and
the stem lineage leading
to this
(Jefferies 1979; Doyle
and
Donoghue 1993).
As
reviewed
by Crane (1988, 1996), Cretaceous
ephedroid pollen
and
megafossils
are
becoming
better
known,
but
they may
not
say
much
about outgroup
relationships,
since
they represent crown-group Gne-
tales.
What
is
needed is
groups
on
the gnetalean stem
lineage, with character combinations intermediate be-
tween those of Gnetales and other taxa. Fortunately,
there are older fossils that may represent groups
of this
sort.
Perhaps most interesting are Piroconites (pollen-
producing organs) and Bernettia (ovulate structures),
described
from the
Early
Jurassic of
Germany by
Kirchner
(1992)
and
van
Konijnenburg-van
Cittert
(1992). For simplicity, I will refer to all these parts
as
Piroconites. The associated leaves (Desmiophyllum)
are
parallel-veined (like Welwitschia
leaves without
cross-veins) and opposite (Crane 1996), as in extant
Gnetales. The
pollen-producing
structures consist
of a
scalelike "microsporophyll"
attached
to
the
top
of
a
leaflike bract; its adaxial side
is
covered with Welwit-
schia-like trilocular
microsynangia,
which
contain
ephedroid pollen.
The
ovulate structures
are similar
but have presumed
ovule scars
on
the adaxial surface
of
the "megasporophyll,"
which
is more
intimately
fused
to the bract. These fossils are
most
suggestive
of
the reproductive structures
of
glossopterids,
which
had similar sporophylls
with adaxial ovules on
top
of
a
leaflike bract.
To
simplify
discussion and
compari-
sons
of
these organs, the ovule-bearing
scale will be
referred to as
a
sporophyll,
the
subtending
leaflike
or-
gan
as a
bract,
and the combined structure
as
a
bract-
sporophyll complex,
as
in
glossopterids. However,
it
should be recognized that the homologies
of
these
or-
gans
are far from established.
Another relevant fossil is
Dechellyia,
described
by
Ash
(1972)
from the Late
Triassic,
associated with
poorly understood cones containing ephedroid pollen.
This
plant
had
opposite,
linear leaves and what
appear
to be
winged
seeds
borne in
opposite pairs
at the
tips
of
leafy
shoots.
It is
possible
that each
winged
seed is
derived from
a
bract-sporophyll complex
like that of
Piroconites, by
reduction
of
the ovules
to one
(Doyle
1994). Schilderia, a Late Triassic wood with no vessels
but
Ephedra-like ray
structure
(Daugherty 1934;
Crane
1988; Doyle 1994), may
also
represent
a stem relative
of
Gnetales,
but it has not been
associated
with other
organs.
Reanalysis
of seed
plant phylogeny
The
way
to test the
implications
of these fossils is
to
add
them to
an
analysis
of
seed
plants. Although
some
may
feel it
is
time for a moratorium on seed
plant analyses, sufficient
new data and
critiques
of
characters have accumulated that it would
be indefen-
sible simply to introduce Piroconites into any existing
data set without
making
other
changes.
The
starting
point
for the
present analysis
was
the
nine-angiosperm
data set of
Doyle et al. (1994), which included fossil
as
well
as
extant
taxa,
modified
by incorporation
of
new characters and data from Nixon et al.
(1994),
Rothwell and Serbet
(1994),
and others. The data ma-
trix
and definitions
of
taxa and characters are
presented
in
the
appendix.
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS
OF
GNETALES
Sll
TAXA
In choosing
taxa,
I
simplified
the
analysis
by omit-
ting progymnosperms.
All
recent
analyses
indicate
that
seed plants
are monophyletic,
with
"coniferopsids"
nested
among "seed
fern" groups.
In this
respect,
the
study
of
Rothwell
and Serbet
(1994)
is
especially
sig-
nificant,
since they
introduced
important
new
data
on
Paleozoic
fossils.
On
the basis
of their results,
the
pres-
ent trees were
rooted
with
the Late
Devonian
genus
Elkinsia,
one
of
the
oldest
and
most
primitive
seed
ferns,
reconstructed
in
considerable
detail
by
Serbet
and
Rothwell
(1992).
In other
seed
fern groups,
I
replaced
"higher
lygi-
nopterids,"
which
included
Heterangium,
with
Lygi-
nopteris,
in
part
because
these two
genera
form
adja-
cent
branches
rather than
a
clade
in the
trees
of
Rothwell
and
Serbet
(1994).
Although
this
may
have
a
slight
effect
on
inferred character
evolution
near the
base
of the
tree,
it
seems
unlikely
to
affect
relation-
ships
among
advanced
groups.
I
added
the
Permian
seed fern
Autunia,
reconstructed by
Kerp
(1988),
a so-
called
peltasperm
that appears
intermediate
between
Callistophyton
(Pennsylvanian)
and Peltaspermum
(Permian-Triassic)
in
having
saccate
pollen
but
simple
sporophylls.
Since
so
many
of
the conflicts
among
recent
anal-
yses
depend
on
possible
relationships
between
antho-
phytes
and
conifers,
I included
all
six
groups
of mod-
ern conifers, plus
the
Paleozoic
genus
Emporia
(Lebachia
lockardii
of
Mapes
and Rothwell
1984),
used
by
Rothwell
and Serbet
(1994).
Emporia
was
separated
from modern
conifers
in most trees
found
by
Rothwell
and Serbet
(fig.
3),
a
result
that is hard
to
accept
in view
of
the many apparently
transitional
Permian
and Triassic
forms
described by
Florin
(1951).
This dissociation
of Paleozoic
and
modern co-
nifers could
be a result
of
the
fact
that
Rothwell
and
Serbet
included
only
three
extant conifer groups
(Pin-
aceae,
Podocarpaceae,
Taxaceae).
In the resulting
trees,
the
basal
branch
in
extant conifers
was Taxaceae,
which
are least like Emporia.
However,
other
charac-
ters
(such
as
pollen,
microgametophytes,
and tracheid
structure)
suggest
that
Taxaceae are
nested
well
within
conifers,
linked with
Cephalotaxus
(Hart
1987).
Nixon
et
al.
(1994)
included the
six
modem
conifer
groups,
but
their
results
could
be
incorrect
because they
omit-
ted Paleozoic forms.
Inclusion
of
these
seven
taxa re-
quired
introduction
of characters
potentially
uniting
conifers, previously
excluded
as
autapomorphies
(such
as one-veined
leaf,
bilateral
fertile
shoot,
compound
female but
simple
male
cones,
and tiered
proembryo),
plus
characters potentially
useful within
conifers.
The
latter were
drawn from
Hart
(1987)
as well as Roth-
well
and
Serbet
(1994)
and Nixon et
al.
(1994).
I
included
11
angiosperm
taxa,
chosen
to represent
the
spectrum
of
possible
basal
groups
and potential
links
among
them.
To
the
nine taxa
in
Doyle
et
al.
(1994),
I
added
Eupomatia,
which has
an
uncertain
position
near the base
of
Magnoliales
or
Laurales, pos-
sibly
linked
with Austrobaileya
and/or
Calycanthaceae
(cf.
Nixon
et al.
1994).
I
also
added
a taxon
repre-
senting
"higher"
or
"core" Laurales,
scored
on
the
assumption
that Monimiaceae
sensu
lato form
a
basal
paraphyletic
series
below
Hernandiaceae
and
Laura-
ceae (Doyle
et al. 1994).
I
did not
include
Cerato-
phyllum,
despite
the
fact that
it
is basal
in the
rbcL
analyses
and
near-basal
(just
above Chloranthus)
in
many
of the
trees
of Nixon
et al.
(1994),
because
so
many
of its
characters
are
uninterpretable.
It
might
be
argued
that
this biases
against
results
like those
of
Nix-
on et
al. (1994).
To
test this,
I analyzed
the Nixon
et
al. data
set with
Ceratophyllum
removed.
This
exper-
iment
did
not
affect
the conclusion
that
Gnetales
are
nonmonophyletic.
It did have
some effect
on relation-
ships:
all most parsimonious
trees
were of
the
type
with
Casuarina
basal
in
angiosperms
and
Ephedra
be-
low
Bennettitales
(fig.
4b).
However,
trees
with Chlor-
anthus
basal
and
Ephedra
the
sister
group
of
angio-
sperms,
Welwitschia,
and
Gnetum
(fig.
4a)
were
only
one step
longer.
I reunited
several
groups
that
were broken
up
by
Rothwell
and Serbet (1994)
or Nixon et
al.
(1994),
since these
(and
other)
analyses
have
confirmed that
they
are
monophyletic.
These
groups
are the medul-
losans
Quaestora
and Medullosa,
the
cycads
Cycas,
Zamia,
and
Stangeria,
and the
cordaites
Mesoxylon
and Cordaixylon.
Scoring
of these
groups
is a
simple
consensus
of the component
taxa
when there
are
only
two
of
these,
but
in
the
case of
cycads
I
assumed
that
Cycas
is
the
sister
group
of
Zamia
and
Stangeria,
as
found
by
Nixon et al.
(1994)
and molecular
analyses
(Hamby
and
Zimmer
1992;
Chase et
al.
1993).
Scoring
of
the
eudicots
was based
on
ranunculids,
Nelumbo,
and Platanus,
which
molecular
analyses
indicate
are
at or
near
the
base
of the clade (Chase
et al. 1993;
Drinnan
et
al. 1994).
Of
potential
fossil relatives
of
Gnetales,
I
included
Piroconites but
not
Dechellyia,
since the
latter
has
many
more unknown
characters
and
poses
more
prob-
lems in
morphological
interpretation.
CHARACTERS
Characters
are defined
and documented
in
the
ap-
pendix.
However,
some
generalities
on
policy
and rea-
sons for
changes
from
previous
treatments
are
appro-
priate
here.
Important
new data
from Nixon
et
al. (1994)
include
evidence
that the
supposed
pinnate
sporophylls
of
most
Mesozoic
seed ferns are actually
branches
bear-
ing
simple,
paddle-like
structures
(cf.
Yao
et al.
1995),
and that the
androecium
of
Ephedra
and
Welwitschia
consists
of
two branched microsporophylls
rather
than
six
simple
ones
(confirmed
by
Hufford 1996).
New
characters
or
character
states
include bifid
micropyle,
which
depends
on
platyspermy
but
is not redundant
with the anatomically
defined
character
(some
platys-
permic
seeds are
nonbifid),
and
which
has
the
advan-
tage
of
being
recognizable
when no
anatomy
is
pre-
served
(see
appendix
for discussion
of the
concept
of
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S12
INTERNATIONAL
JOURNAL
OF PLANT SCIENCES
platyspermy,
which
I
have
retained but
modified in
response to
the critique
of Rothwell
and Serbet
[1994]). Another
is the distinction
between
endo- and
ectokinetic
microsporangial
dehiscence, treated
as ad-
ditional states
of the previous
endothecium
character.
Contributions
of Rothwell
and Serbet
(1994)
in-
clude new
ovule characters
(sealing, pollen
chamber)
and
their demonstration
that the so-called ovulate
heads of Pentoxylon
(Carnoconites)
are
not axes but
rather sporophylls
with
seeds
on
both
sides.
I
rescored
other
characters
of
Pentoxylon
on
the
basis of Bose
et
al.
(1985),
who refuted certain
previously
reported an-
thophyte-like
anatomical
features (multiseriate
rays,
scalariform secondary xylem
pitting).
Previous
character definitions
were
criticized by
Nixon et
al. (1994) as excessively
theory-laden
and
dependent
on
complex
evolutionary hypotheses.
I
have
attempted
to
correct
cases where
these
criticisms
seem
valid, but,
as
anticipated by
Nixon et al. (1994,
p. 490),
I
would
defend
others as cases where
homology judg-
ments can
and
should
be made
a
priori on
the
basis of
positional
and/or developmental
criteria. There
is
al-
ways a tension
between
assessing
homology
at the
stage
of
character
analysis
and at the
stage
of
analysis
of
congruence
with other
characters
by
means of
par-
simony.
Excessive reliance
on a
priori
morphological
criteria risks
overlooking
cases
where
the
same struc-
ture has
changed
in
position
or
development,
whereas
coding
structures the
same
despite morphological
dif-
ferences risks
equating
superficial
similarities that
arose by convergence.
In
general,
I have
tried
to
define
characters
in
a
simpler
fashion, recognizing
states
as
mutually
exclusive conditions
found
in
different
groups, without
specifying
as stringently
the exact
morphological
homologies
of
the structures
involved
or
the changes
involved
in
going
from one
state
to
another.
A character
that
I
redefined in
response
to
Nixon et
al.
(1994)
is
microsporophyll
morphology
in
angio-
sperms.
The lateral
position
of
the
two
pairs
of
micro-
sporangia
suggests that
angiosperm
stamens
are
de-
rived from
pinnate
structures.
Doyle
and
Donoghue
(1986, 1992)
therefore scored
them as
pinnate,
but
this
can
certainly
be
challenged
as
making
excessive
as-
sumptions
about transformations.
On
the other
hand,
stamens differ
from the
simple
microsporophylls
of
cordaites,
conifers,
and
ginkgoes,
which
have
nothing
like the central
connective that
separates
the
two
pairs
of
microsporangia.
To
express
these
observations,
I
have treated the
angiosperm
stamen
as a third state
of
the
microsporophyll
character.
This means
that
it could
be derived
by
one
step
from
either
a
conifer-like
or a
more
complex microsporophyll.
Doyle
et al.
(1994)
treated two
pairs
of
sporangia
as a
separate
character
(an angiosperm
synapomorphy),
but
the
present
solu-
tion
has
the
same
effect in
uniting angiosperms
while
making
fewer theoretical
assumptions.
A case where
I
retained
a
definition
questioned by
Nixon et al.
(1994)
is the
comparison
of
the Caytonia
cupule
with
the outer
integument
of the
anatropous
bitegmic
ovule.
As discussed
above,
Nixon et
al. re-
jected
this homology
on positional
grounds,
but
their
rejection
was based
on a
misconception
about
angio-
sperm
ovules.
It is worth repeating
that coding
these
structures
the same
does not guarantee that
the
groups
having
them are
related; such
scorings
only state
that
the
similarities are potentially
homologous. Whether
it
is concluded that
they are
homologous
still depends
on
congruence
with other characters,
as determined
by
parsimony
analysis.
The fact that there
are such char-
acters
in Caytonia
and angiosperms,
even
in the data
set of Nixon et al.
(1994),
is shown by
the analysis
of
Albert
et al. (1994),
who combined the
Nixon et
al.
morphological
data set with
rbcL data
and found
that
anthophytes
did
associate
with Caytonia.
A
further
test,
based
on removal of the most
controversial
re-
productive
characters
from
the matrix,
is presented
be-
low.
I also followed previous
definitions
for the character
expressing
leaf
organization
and
major
venation.
Doyle
and Donoghue (1986,
1992;
also Doyle
et
al.
1994)
lumped
simple, pinnately
veined
leaves (as
in
Pentoxylon, glossopterids,
Gnetum,
and
woody
mag-
noliids)
with compound
leaves
of the
type
found
in
cycads
and
Bennettitales,
where
each
leaflet
has
par-
allel veins that come
from the rachis
of the
frond,
as
"simple
pinnate,"
and
separated
these
from
fernlike
compound
leaves,
where each leaflet
has its own
mid-
rib
and
secondaries.
Compound
leaves
show a wide
range
of
morphology;
if all the variations were
ex-
pressed
as
states,
the
number
of
states
would be large,
and
many
of
them
would
vary
within taxa. It
is desir-
able
to
lump
these conditions
into
a
smaller
number
of
states, which
would ideally
be
monomorphic
within
taxa
(cf.
Nixon and
Davis
1991).
Nixon
et
al.
(1994)
made a seemingly objective
dis-
tinction
between
simple
and
compound
leaves. How-
ever,
the
difference between
simple
and cycad-like
compound
leaves
may
be
a
result
of
interruption
of the
marginal
meristem,
whereas
the difference between
the
femlike and
cycad
types
may
involve
more
changes
in
development.
This
view
is
supported
by patterns
of
variation within taxa.
Living cycads
all have
com-
pound
leaves,
but several fossils
had
simple
leaves
of
the
Taeniopteris
type.
Some
of these
are
controversial
as
cycads,
such
as those associated with
Permian
cy-
cad-like
sporophylls
(Mamay
1976),
but others
occur
in
Bjuvia
and Nilssonia (=
Beania
=
Androstrobus),
which are securely assigned
to
the
group
(Stevenson
and Artabe
1995).
Nilssonia
itself shows all
degrees
of dissection
(Harris
1964).
Most Bennettitales
have
compound
leaves,
but the
Taeniopteris
type
occurs
in
Williamsoniella.
In
contrast,
femlike
compound
leaves
are unknown in Bennettitales
and
occur only
in
Stan-
geria
and
Bowenia in
cycads;
these
appear
to be
nested
within
the
group
and
thus
do
not
affect its
inferred
basic state.
Conversely,
none of the
groups
scored
as
femlike have
members
with
cycad-like
leaves.
These
observations
suggest
that although
variation between
simple
and
cycad-type
leaves
may
be
significant
at a
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS OF GNETALES
S13
lower taxonomic
level, it is not at the present
level of
analysis.
Another case
involving alternative ways of
defining
character
states,
discussed
above,
is
embryogeny.
Here
again
I
sought
a definition
that would allow
key taxa
to be
scored more
readily.
Rather than
defining
the
character
in
terms
of cellular vs.
free-nuclear,
for
which the
state
in
Ephedra is ambiguous,
I
have
dis-
tinguished whether
each embryo
is
derived by
cellular
divisions from a
single
uninucleate cell or from several
free
nuclei.
Another example
concerns the bipartite
outer integ-
ument
of
Gnetales.
Because this develops
from two
primordia and
corresponds in position to the
perianth
of
the
male
flower,
Doyle
and
Donoghue
(1986, 1992;
also
Doyle et al.
1994) treated
it
separately
from the
"cupule"
character,
which described
ovule-enclosing
envelopes
of
other
kinds
(none; lobed,
lyginopterid-
type cupule;
anatropous Caytonia cupule
or
bitegmic
angiosperm ovule;
orthotropous
but unlobed
cupule,
as
in
Bennettitales), for which
Gnetales were
scored as
lacking
a
cupule.
This
introduced
two
steps between
Gnetales
and
angiosperms
or Bennettitales-loss
of
the
cupule, plus
origin
of
the outer
integument.
This
might
be valid
if
evolution went
in
this
direction,
but
it
seems too biased
against
the
opposite
transition:
from
the
envelope
of Gnetales to the
cupule
of Ben-
nettitales
or
the outer
integument
of an
orthotropous
bitegmic ovule,
which
might
be
accomplished
by
a
simple change
to
a
ringlike primordium.
On the other
hand,
the
solution
of
Nixon et al. (1994),
lumping
all
kinds of
envelope
as one character and using a
separate
character
to
distinguish simple
and
bipartite,
equates
structures that are different
enough
in
position
and de-
velopment
to cast serious
doubt
on
their
homology.
The
solution
adopted
here
is
to treat the gnetalean con-
dition
as a fifth state of the
"cupule"
character rather
than
a
separate
character. This
would
allow
derivation
of
the
gnetalean outer
integument
from
a
bennettita-
lean cupule or an
angiosperm
outer
integument, or vice
versa, by
a
single
step.
Characters
of the
ovule-bearing
structures are
worth
examining
both
for
their
importance
in
establishing
re-
lationships
and for
their long-standing interest
to evo-
lutionary
morphologists.
I
redefined
megasporophyll
states as
pinnate
(including partly
three-dimensional)
for
early
seed
ferns, cycads, Caytonia,
and
most
an-
giosperms (i.e.,
taxa with
ovules
in
two
rows); simple
and
paddle-like
with several ovules
for
peltasperms,
Pentoxylon,
glossopterids,
and
Piroconites;
and
single,
stalked
or
sessile
ovules
for
coniferopsids
and
extant
Gnetales.
Following
Rothwell
and
Serbet
(1994),
I
added a character for
ovule
position:
abaxial in
Autu-
nia,
Peltaspermum,
and
Callistophyton;
adaxial in
glossopterids,
Piroconites, Caytonia,
and
angiosperms.
As noted
above,
this
scoring
of
angiosperms
refers
not
to
position
of
the
bitegmic
ovules on
the
carpel but,
rather,
to
position
of
the
nucellus plus inner
integu-
ment,
which
presumably corresponds
to the
original
seed
plant ovule,
relative
to the
outer
integument
(Doyle
1978; Doyle
and
Donoghue 1986). Although
this
character
is
potentially very
important
in
assessing
the
homology
of
similar-looking
structures, it was pre-
viously
omitted because data
were too sporadic and
problematic.
However, it is now
documented
in
more
taxa, such
as
glossopterids, based
on
orientation
of
the
xylem
and
phloem
in
vascular strands in the
sporo-
phyll (Taylor
and Taylor 1992;
Pigg
and Trivett
1994),
as well as on
compression
material
(Pant
and
Nautiyal
1984).
The
"megasporophyll"
and
"cupule"
characters as
thus defined
are
not
rigidly
tied to a classical
sporo-
phyll
concept. Rather,
the former describes the struc-
ture
on
which
the
ovule is
borne
(which
is
usually
a
lateral
appendage on a stem
in
the morphological
sense, but not
always assuredly
so,
as
in
the case
of
the
bract-sporophyll complex
of
glossopterids
or
the
"terminal" ovule
of modern
Gnetales),
whereas the
latter describes
any closely
investing envelope.
Thus
glossopterids
are
scored
as
having
paddle-like sporo-
phylls but no
cupules.
Morphologically,
there is some
potential
overlap
between these characters.
For ex-
ample, transformation of the
bract-sporophyll complex
of
glossopterids into the
pinnate
sporophyll with anat-
ropous cupules
of
Caytonia
would
equate paddle-like
"sporophylls"
with
"cupules,"
and what
may
be a
leaf
plus
an
axillary
branch with a
"sporophyll." Although
this scenario violates classical
organ
concepts,
I
see no
reason to assume that it
is
morphologically impossible,
and
it
can
be
interpreted
as
involving
two
independent
steps-transformation
of a
structure with dorsal and
ventral
parts
into one with
parts
in two
lateral
rows,
and
folding
each
flat
ovule-bearing part
into
an
anat-
ropous cupule.
Similarly,
under one
interpretation
im-
plicit
in the
scoring
of
the
sporophyll
character
(0/1),
corystosperms
have
ovules
on
paddle-like structures
and
anatropous
cupules,
both of
which are the same
organ.
However,
the
first
character
can
be viewed as
describing
the kind of
appendage
on
which
the
ovule
is borne, the
second its folded and
reflexed character.
A
particularly vexing
case
concerns Bennettitales.
Crane
(1985)
and
Doyle
and
Donoghue (1986, 1992)
interpreted
the ovuliferous
receptacle as an axis bear-
ing simple
sporophylls
with
single,
terminal
ovules,
often
if
not
always
surrounded
by
a
cupule. However,
another
hypothesis
(Doyle
and
Donoghue 1986;
Crane
1988; Doyle
1994)
is
that
the ovuliferous
receptacle
is
a
sporophyll
that
became
secondarily
radial
(and ter-
minal),
as inferred for
Pentoxylon
by
Rothwell and
Serbet
(1994).
If
so,
it
might
be
compared
with either
a
paddle-like
sporophyll,
as
in
glossopterids,
Pentox-
ylon,
and
Piroconites,
or a
pinnate
sporophyll,
as in
Caytonia
and
angiosperms.
Because of
this
range
of
possible
homologies,
I
scored
Bennettitales as un-
known.
Further
study
of
the vascular
architecture
of
bennettitalean flowers
might
shed
light
on
this
prob-
lem.
In
many
groups,
microsporophyll morphology
is
correlated with
megasporophyll
morphology,
so
treat-
ing
the two as
separate
characters
might overweight
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S14
INTERNATIONAL
JOURNAL OF
PLANT
SCIENCES
what is actually
a
single character. The most important
exceptions are cordaites, in which the megasporo-
phylls are sometimes branched but the microsporo-
phylls are simple; Gnetales, which have single sessile
ovules but branched microsporophylls (Nixon et al.
1994; Hufford 1996); and angiosperms, in which the
carpels (when multiovulate)
are
pinnately organized
but the stamens are simple.
The
solution adopted here
is to distinguish the two kinds of simple, one-veined
microsporophylls (i.e.,
in
coniferopsids
and
angio-
sperms)
from
all
more
complex types.
This
emphasizes
those cases in which morphology of the male and
fe-
male structures is not similar, without duplicating cor-
related changes
in
the two structures
in
other groups.
Several
characters pose potential problems
of what
may be called the "Maddison effect" (Maddison
1994): biases that arise when one character represents
the
presence
or
absence of
a
structure,
another
char-
acter different features of the structure. Problems occur
when the
structure originates independently
in
differ-
ent
groups. Scoring
the
state
in
the second character
as unknown
in
groups that lack the structure favors
arrangements
in
both groups
in
which
the second
char-
acter
has
the
same
basal state. However,
there
is
usu-
ally
no
reason to assume
that
the structure would
begin
with
the
same state
if it
originated independently. One
solution
is
to treat absence of the structure
and
its
var-
ious
forms as an
unordered multistate character, but
this weakens
presence
of
the
structure
as
a
potential
synapomorphy. Fortunately, most potential examples
of
this effect
in
the present
data
set concern
floral and
other characters of angiosperms, which
involve
struc-
tures that do not exist
or
whose
homologies
are
unclear
in
other groups.
If
angiosperms
are
monophyletic,
scoring outgroups
as unknown for
such characters
does
not
pose
a
problem (cf.
Nixon et al.
1994),
so
this is the solution
adopted.
I
rejected
some old
characters
and
some new ones
proposed by
Nixon et al.
(1994)
and
Rothwell and
Serbet
(1994)
because
of
problems
in definition
and
inconsistencies
in
available
data. An
example
is
simple
vs. ramiform
pollen tube (Nixon et al. 1994). Nixon
et al. scored
cycads, ginkgoes,
and conifers
as
rami-
form,
Gnetales and
angiosperms
as
simple. However,
Friedman (1987, 1993) described pollen tubes
of
cy-
cads
as
usually
unbranched
and noted
exceptions
to
the
predominant
condition in
all
groups except Ginkgo
(branched)
and Gnetales
(unbranched). Satisfactory
use
of
this
character
may require
more
systematic
data
on
conifers and
angiosperms,
and
probably
new
pri-
mary
observations.
A character that
I
considered
eliminating
is
thick
nucellar cuticle, documented by Harris (1954) in Me-
sozoic seeds and
included by Crane (1985)
and
Doyle
and
Donoghue (1986, 1992;
also
Doyle et
al.
1994)
as
a
potential synapomorphy
of
glossopterids, Caytonia,
and
anthophytes.
Use
of
this character is
plagued by
conflicting
data in
the literature.
Thus
Singh (1978)
contradicted Harris
in
stating
that
cycads
and
Ginkgo
have
a
thick nucellar
cuticle,
and Nixon
et
al.
(1994)
and Rothwell and Serbet (1994) scored
Paleozoic
groups differently. This may be a
function of study
method and/or inconsistent criteria for "thick" and
"thin." Harris's (1954) observations were based on
identification of two
cuticle layers
with different
cell
patterns in macerated compressions, the
inner (nucel-
lar) layer being either about as thin as or
notably thick-
er than the outer (inner cuticle of the
integument),
whereas most
Paleozoic
seeds
have
been studied as
petrifactions. To use this character
consistently across
seed
plants,
it would be desirable
to
study
all
taxa with
the same
techniques.
As
a
temporary
solution,
I have
scored the character for groups studied
by Harris
(1954) and other authors using his criteria
(e.g., Crane
1985) and treated other groups as
unknown.
ANALYSES
Heuristic searches for most parsimonious
trees were
performed with PAUP (Swofford 1990).
Most analyses
involved 50
replicates with stepwise
random addition
of
taxa
and
TBR
branch
swapping,
to increase
the
probability of discovering different
"islands" of trees
(Maddison 1991).
The relative
parsimony
of alterna-
tive
hypotheses
was
investigated by
searching
for trees
consistent with constraint
trees in
which
particular
taxa
were
specified
as
forming
a
clade.
Subsets
of most
parsimonious trees were often categorized
by filtering
trees consistent
or
inconsistent
with
particular
con-
straints.
Bootstrap analyses (Felsenstein 1985)
were
per-
formed with
PAUP, using
300
bootstrap
replicates.
In
order to avoid
prohibitive computation
times,
each
rep-
licate involved
only one analysis
with
closest
addition
sequence,
TBR
branch
swapping, holding
five trees at
each
step,
and
maxtrees set at
200.
This
is
less
likely
to
have found most
parsimonious
trees
for
each
rep-
licate than would several
replicates
and a
higher
max-
trees
value,
but
in
view of the
approximative
nature of
bootstrap analysis
in
general, any
distortions are
prob-
ably inconsequential.
Decay analyses (Bremer 1988;
Donoghue et
al.
1992),
which
determine
the minimum
number
of
ad-
ditional
steps
at which a
given
clade
found
in
the most
parsimonious trees
breaks
down,
were
performed
with
PAUP
by searching heuristically
for
trees
less
than
or
equal
to
a
given
number
of
steps
and
examining
which
clades
remain
in
the
strict consensus. With
one-off
trees,
the
3288
trees obtained were found
in
the first
three
of
50
replicates
with
stepwise
random
addition
of taxa and TBR
branch
swapping;
this search
was
probably exhaustive. Searches
for
longer trees in-
volved
three to
seven
single-replicate
analyses
for a
given range
in
tree
lengths,
each
saving
10,000
trees.
These are less
certain to
be
exhaustive,
but searches
for
trees constrained to show
plausible alternative
re-
lationships
were
consistent with
the
results obtained.
MacClade
(Maddison
and
Maddison
1992)
was
used
to
study
character
evolution and character
support
for
clades
on the
trees obtained from
PAUP.
In
describing
the
results,
when
I
say
that
particular
character
states
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC RELATIONSHIPS
OF
GNETALES S15
Platysperms
Glossophytes
Coniferopsids Gnet
Rngiosperms
k4
c
'A
s 1,
0
E
_'La
c a a
c a
.r
.
a
X
'A
uXX
|~~~
A
Fig.
52
preen
'anayi,_
wit dea
nmer,
et.
n
oosrpfe
quencies of
clades.L
MacClaye
as
i
uabg
us
changes
on
i
Ve
.on
dI 7
%I
y
d
u
d
sing telde.
pasionos
res all of
dhich belog t
asngl
dlsl3
[np
ad bo
dl
ml
dl
4
l6
dd
997
Donoghu
5989).
d2
46
d5
94
d4
91
Fig.
5
Strict consensus
of most
parsimonious
trees
found in
the
present
analysis,
with
decay
numbers
(dc,
etc.)
and
bootstrap
fre-
quencies
of clades.
unite
a
group,
these
states are those indicated
by
MacClade as unambiguous
changes
on
the
correspond-
ing
branch of the tree.
In some
cases,
a derived state
was
restricted to a
particular
dade but
MacClade
con-
sidered
its
point
of
origin
ambiguous
because more
than
one
other
state occurred
in
the closest
outgroups;
I
describe such states as
probably uniting
the
cdade.
Results
Figure
5
shows
a
strict
consensus
of
the 123
most
parsimonious
trees,
all
of which
belong
to
a
single
island
(Maddison
199
1),
plus
decay
numbers
and
boot-
strap
frequencies
for
clades.
The
consistency
index
(CI)
of
these trees
(a
measure of
homoplasy)
is
0.49,
which
is almost
exactly average
(on
the
regression
line
of
CI
vs.
number of
taxa)
for
36
taxa
(Sanderson
and
Donoghue
1989).
Examples
of most
parsimonious
trees
are
shown
in
figure
6.
Many
of
the variations
concern the
region
delimited
below
by
medullosans and above
by
two
groups:
(1)
coniferopsids
(cordaites,
ginkgoes,
coni-
fers) and
(2)
a
dade
consisting
of
glossopterids, Cay-
tonia,
and
taxa
that made
up
the
anthophytes
in
previous analyses, together
designated
the
glosso-
phytes.
The taxa affected
by
these variations
are there-
fore
Callistophyton,
cycads,
corystosperms,
and
the
peltasperm genera
Autunia and
Peltaspermum.
Rela-
tionships among
these
groups
fall
into two
main
pat-
terns:
(1)
Callistophyton
basal,
followed
by
cycads,
with
corystosperms
and
peltasperms
on
the
conifer-
opsid
line
(fig.
6b);
(2)
cycads
directly
linked
with
glossophytes,
and
the
remaining
groups
variously
dis-
tributed
on
the
glossophyte
line,
on the
coniferopsid
line,
or
below the
two
(Callistophyton
is
basal
in
55%
of
these
trees,
but
not
the
rest;
fig.
6a,
c).
Other
vari-
ations
occur within
coniferopsids
(whether
cordaites
or
ginkgoes
are
basal)
and
the
glossophytes,
in
which
the
basal
group
may
be
glossopterids
(39
trees),
Pentox-
Glossophytes
Coniferopsids
Gnet
Angiosperms
0, 4,
v
-
n
'
S. V
n
x
:
E
%
X ^ *
2
2~~~~~~~1
>W
0
Glossophytes
n sipl pinSn g ; c e i
Cl~ ~ ~ ~ ~ -0
0
'
0.49
1
pa0Lm.te
_j *Z:u
3 v
i
F<:0
w
IV
UoO
4<
iL L) L7
cw
:3
s
ea L) i
L uC F
:,3
-A
<_>
E
\\\\\\\\Q& a
~~~~~~~~~~Leaf
type
\\\\\\\\G&> t>Z I I
~~~~~pinn
comp
b \\\\\\\\Q* ./;/ EZ
~~~~~~~~~~~~sirnple
pinn
I Treelength:
247
|d\\ \ \ \
\toamous
one veined
\ZZ
uncertain
equiyocal
Glossophytes
Coniferopsids
Gnet
Rngiosperms
'A
%4
c:
%
:X* <oa
L
2 2
.49
2
22
\\\ \\\\\
./ -p
ci
i
2
qioa
VIo
(26
trees),MA
00
0
or-
Bentiae (5
Lre)
Mn
Meea
internal
,
arrangmentsoccur
Rather tha attmpin
ro
disus the
chrate sup-
p22
c com2
por
foT
lteeareelemengthn
247i
vaidichtmpli-
Glossoph!Jtes
Coniferopsids
Gnet
Angiosperms
'
22V0 0 2
9
A
52~~52522 i2r
25.
A2 25.0 i
'A
ig6a-c, Rersettvems pasioniostee
fondinth
prsn aayi,
showing disribtio
ofth lafcactr
ylon
(26 trees), or Bennettitales
(58 trees),
and
several~~~~~~~~~~~~~~~~~'
L,:3wL
C 9L
internal arrangements occur.~~~~~~~~inncom
Rahr
hnatepiniomicsstechrcer
sp-n
porelnthfralths
ageet
n thirvaidichtmpli-
cations for character
evolution,
I will
concentrlate
o
one. suse
thatRepisenmorie
consiastentniostheeknowndag
dis-th
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S16
INTERNATIONAL JOURNAL OF PLANT SCIENCES
tributions (fig. 6a). Thus trees
with
glossopterids
(Permian) basal in the glossophytes
are more
consis-
tent with the stratigraphic record than those
with
Pen-
toxylon (Jurassic-Early Cretaceous) or Bennettitales
(Late Triassic-Cretaceous) basal.
The latter
(fig. 6c)
are
doubly implausible
from a
stratigraphic point
of
view, since
the
older glossopterids
are three nodes
above Bennettitales,
linked with
Pentoxylon.
When
glossopterids are basal in the glossophytes, Callisto-
phyton
is
immediately above
the
medullosans,
but
the
position of cycads, corystosperms, and peltasperms
varies.
I
have chosen
a
tree in which coniferopsids
are
the
first
branch above Callistophyton, and corysto-
sperms and peltasperms
are near the
base
of
the
glos-
sophyte
line. This is more
plausible stratigraphically
than trees in which
the
Permian and Triassic
corysto-
sperms and peltasperms form
a
paraphyletic
series on
the coniferopsid line, below cordaites
and
Emporia,
which are the oldest known taxa
in
the
platysperm
clade
(Late Carboniferous).
At the base of
the
seed
plants
is
the series of
taxa
found in
previous
seed
plant analyses-lyginopterids,
medullosans, and a
clade
consisting
of
Callistophyton,
other
platyspermic groups,
and
cycads, together
called
the platysperms-supported by high decay
and boot-
strap values.
Medullosans are linked
with
platysperms
by loss
of the
original lyginopterid cupule,
loss
of
the
central column
in
the lagenostome,
thick
sarcotesta,
vascularized
nucellus,
and bilateral
pollen symmetry.
Characters
uniting
the
platysperms
in
figure
6a are a
change
from bifurcate to unbranched rachis
(reversed
in
corystosperms), abaxial microsporangia, bilateral
(platyspermic) seeds,
bifid
integument,
sealed micro-
pyle,
and saccate
pollen
with
honeycomb-alveolar
ex-
ine
structure and a sulcus
(lost
in cordaites and Em-
poria).
Of
these,
bifid
integument
does not arise
here
when
cycads
are located
immediately
above Callisto-
phyton (fig. 6b),
and its
point
of
origin
is
equivocal
in
other trees.
As
in
previous
studies
(cf.
Doyle 1988),
the
status
of
saccate
pollen
as
a
synapomorphy
of
the
platy-
sperms
is
uncertain because
of
the
large number
of
members
that
lack air sacs. These
may
be
either
prim-
itively
or
secondarily nonsaccate, depending
on the
tree.
For
example,
in
trees of
the
type
shown
in
figure
6a
(fig. 7a),
sacs arise at
the
base
of
platysperms
and
are lost in three
lines,
but in trees
of
the other
types
(fig. 7b)
sacs
originate
five times.
In
both kinds of
tree,
sacs
originate independently (whether
as
a
conver-
gence
or a
reversal)
in
glossopterids
and
Caytonia. Of
these,
the case of
glossopterids is
less
certain, since
their sacs are
homologous
with
those of
lower
groups
in one-off
trees in which
cycads and peltasperms
are
arranged differently.
As
already noted, cycads may
be
located just above
Callistophyton,
one node
above their
basal
position
in
some
trees of
Crane
(1985)
and
Doyle
and
Donoghue
(1986, 1992; Doyle 1988)
and
those
of Nixon
et al.
(1994), or nested more deeply within the platysperms,
recalling
other trees
of
Doyle and Donoghue (1986,
Glossophgtes
Coniferopsids
Gnet
flngiosperms
Treelnath 24 |
0
4
'q
M
0
7
a,
.m
co Z 2
A
-;;
aOOOOOOOEEEOEEOOE--OOOOOOOOOOOOEJOOOO
Treelength
2
247
2
22s
E
Fig.272Most
pasmoiu 2re
2hw infg2aadbhwn
C19 0.492; Doyle
1988)
andthoseofRothwellandabsent
(1994). The
platysperm
clade,
including
cycads, has
one of the highest decay
numbers (five steps) and
bootstrap values
(94%p)
in
the tree. These results and
those
of other
recent analyses (Doyle and
Donoghue
1992; Nixon
et al. 1994; Rothwell and
Serbet
1994)
contradict the commonly
suggested direct relationship
between medullosans and
cycads, seen in some trees
of Crane (1985)
and Doyle and Donoghue
(1986). The
shortest trees found when
cycads were forced together
with medullosans
are five steps longer than
the shortest
trees (this
accounts for the decay value
of the
platy-
sperms). The
unstable position of cycads
reflects the
fact that they
have several features that
may be either
primitive or
reversals, such as nonbifid integument
and
nonsaccate pollen, and their
uncertain seed symmetry
(Crane 1988; Doyle and
Donoghue
1992).
Coniferopsids are united
by dichotomous leaf ye-
nation, scalelike microsporophylls,
and endokinetic
microsporangial dehiscence.
In some trees (fig. 6a),
ginkgoes are linked with
conifers, based on endarch
leaf traces and loss of nucellar
vasculature. In others
(fig. 6b, c),
ginkgoes are basal and cordaites
are
linked
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS OF GNETALES
S17
with conifers, based on a shift from secretory cavities
to secretory
cells and
origin
of saccate
pollen;
these
are trees
in
which sacs originate
several
times
and the
outgroups of coniferopsids include nonsaccate taxa
such
as
Peltaspermum (fig. 7b). Although Doyle
and
Donoghue (1992) found some trees with ginkgoes
linked
directly
with
Peltaspermum,
as
proposed by
Meyen (1984), trees showing this relationship
are
now
four steps less parsimonious.
In
all these trees, the
lack of a
sulcus
in
pollen
of
cordaites
and
Emporia
is a
reversal,
rather than
prim-
itive, as generally assumed (see, e.g., Millay and Tay-
lor
1976),
a
problem
also
encountered
and discussed
by
Rothwell and Serbet
(1994).
The same is true
of
other
supposedly primitive
states
in
cordaites:
mesarch
leaf
traces, apical microsporangia, and (in trees
where
ginkgoes
are
basal,
such as in
fig. 6b, c)
lack of a well-
developed embryo
in
the seed. These
reversals are dif-
ficult to accept not only on functional grounds, but
also
because
cordaites are the oldest known members
of
the
platysperm
clade.
Trees in which
coniferopsids
are
basal in
the platysperms, such that
nonsulcate
pollen
may
be
primitive
in
cordaites
and
Emporia,
are
two
steps
less
parsimonious.
Contrary
to Rothwell
and
Serbet
(1994),
modern
co-
nifers are
linked with Emporia,
based on one-veined
leaves, dorsiventral
female short
shoots (=
cone
scales),
reflexed
ovules,
and reduced sarcotesta.
Com-
pound
female
strobili
are another
synapomorphy
in
trees
such as
that
in
figure 6a,
but these
may
arise
lower when
cordaites
are linked
with
conifers.
Extant
conifers are united by woody ovuliferous
cone
scales,
loss of
the
lagenostome, siphonogamy,
and
possibly
resin canals (the point where these originate is
con-
fused
by
variation between
secretory
cells and
cavities
in
the outgroups). Loss
of scalariform
metaxylem pit-
ting,
seen in
Ginkgo
but not
Emporia
and
cordaites,
is
another
synapomorphy
of
extant conifers when cor-
daites are the
sister
group
of conifers as a
whole,
but
this loss
may
occur lower when
ginkgoes
are linked
with
conifers.
All
trees imply that
Pinaceae and Po-
docarpaceae
are
primitive
in
having
saccate
pollen,
whereas
Taxaceae and other nonsaccate
groups
are
ad-
vanced
(fig. 7a, b).
This
difference from the results of
Rothwell
and
Serbet (1994)
can
be explained because
Taxaceae, instead
of
being
basal in
conifers,
are united
with
Cephalotaxus,
based on
spiral tertiary thickenings
in
the tracheids and successive
microspore cytokinesis.
These two taxa are in turn linked with Araucariaceae
and
Taxodiaceae,
which
Rothwell and Serbet did not
include in
their
analysis,
based not
only
on loss of
sacs, but
also on
fusion of the bract and cone scale,
radial
pollen symmetry (further modified to global
in
Taxaceae),
and
granular
exine
structure, and with Tax-
odiaceae
in
particular based
on
four-nucleate
micro-
gametophytes.
Relationships among the platyspermic groups be-
tween
coniferopsids
and
glossophytes
are
poorly
re-
solved
because
of
the small
number
of
potential syn-
apomorphies, aggravated by conflicts among
characters
and
the fact that these groups
have
high
proportions of missing data. Thus
in
some trees (figs.
6b, c; 7b) the
Permian
peltasperm Autunia
(Kerp
1988) and Triassic corystosperms are linked, based
solely on their saccate pollen, but this assumes that
sacs were originally absent
in
platysperms.
Converse-
ly, the loss of sacs is
a
synapomorphy
of
Peltasper-
mum, cycads, and anthophytes
in
figures 6a
and 7a
(cf. Doyle 1988).
The fact that
Autunia, Peltasper-
mum,
and
corystosperms
are
usually adjacent (plus
sometimes
Callistophyton) agrees
with
Meyen's
(1984) concept that these groups are derivatives of the
same
Permian
radiation.
However,
trees
in which
these
groups
are
paraphyletic relative to coniferopsids imply
that
coniferopsids
are also derived
from
this
radiation,
which conflicts with
stratigraphic
evidence.
Corystosperms,
which were
associated with antho-
phytes
in
the analysis
of
Crane (1985), pose special
problems
because
of
uncertainties
in
morphological
in-
terpretation.
In
the
present trees,
their
"anthophyte"
features (reduced megaspore wall, anatropous cupules)
are
convergences with Caytonia
and
anthophytes.
Yao
et al.
(1995) questioned
an association of
corysto-
sperms with anthophytes,
based
on
their demonstration
that the
microsporangiate
structures
are axes
bearing
simple sporophylls
rather than
pinnately compound
leaves.
This
finding
is
not responsible
for
the present
result, since my definition
of
the microsporophyll
char-
acter
does
not
distinguish
between
pinnate
and
paddle-
like.
A
more critical
question
is
whether the
ovule
is
on
the
abaxial or
adaxial side
of
the
hoodlike
cupule.
Since this
is
uncertain,
I
scored the character as either
state
(1/2),
but the
relationships
obtained
assume that
it
is abaxial. This seems
likely
if the
cupules
are
simple
sporophylls
on a
stem, analogous
to
the
microsporan-
giate structures,
since
they
are
folded downward with
respect
to the axis.
If, however,
the ovule is shown to
be
adaxial,
this would weaken the
present
result and
strengthen
a
relationship
of
corystosperms
with the
glossophytes.
The other so-called Mesozoic seed fern
groups,
glossopterids (actually Permian)
and
Caytonia,
are
as-
sociated with the taxa
previously
called
anthophytes.
I
have
named
this clade the
glossophytes
in order
to
emphasize
the
plesiomorphic
status
of
glossopterids
within it.
In
trees
in
which
glossopterids
are
basal,
glossophytes
are
united
by
adaxial
ovules, apical
or
adaxial
microsporangia,
loss of the
lagenostome,
and
thick
nucellar
cuticle.
Loss of
secretory
cavities
may
be another
synapomorphy (this
is
equivocal
in
some
trees because both cavities and canals
occur
in
the out-
groups), except
when
Bennettitales, which have canals,
are
basal,
and
these
are
homologous
with the
canals
of
cycads.
When
glossophytes
are
associated with cor-
ystosperms
and
cycads,
both
of which have
many-trace
nodes,
a
shift to two-trace
nodes
(as
in
Pentoxylon,
Gnetales,
and
probably glossopterids)
is another
syn-
apomorphy (fig. 8a).
When
Bennettitales
are
basal
in
the
clade,
adaxial
ovules
and
thick nucellar cuticle are
equivocal
as
glossophyte synapomorphies,
since Ben-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S18 INTERNATIONAL
JOURNAL OF
PLANT SCIENCES
Glossophgtes
Coniferopsids Gnet Rngiosperms
C-
.
'a ? , E *
.4
rc
g
=?<oL:s>5
ssc.=>>.LoX.
x~~~~
0I
-A
_
Z.
U2 w
- o<XosXs3 u
Q. min
Z
a. E:
-C
JQ<
ui
E
DO1 1 DOD0
ODDE
MO
DaEoEE
0
*DEMMEE
E3
Node
a
z
z z z z
z O z z
* z z
a C3 * 2
* *
* * C3
* * *
aone trace
I Treelength:
247 |mant
trace
ICI:O0.491
two
trace
three
trace
uqncerta1J
eqoivocal
Glossophgtes
Coniferopsids
Gnet
Rngiosperms
LC
00 o
.L
o
._ o s c
g
.
Et
>
soo<XosX3
4,
2
W3JQ<
-j
'I in
zzzzzzz
zz2z
z
C
:
-C
ui
E
Treelenth:
24
free lobes
| Cl:
0.49
|
IEJ
simplefi
bifid
Itubularl
\ r
I I
~~~~~~~~~equivocal
6lossophytes
Coniferopsids
Gnet
Rngiosperms
.9
IA
0 '
i
'A
LeC
| apicalVmart
Cl:
0.49
1
\ \
'abaxl'
Fig. 8 Most
parsimonious
tree shown in fig.
6a, showing distri-
bution of the
(a) nodal anatomy,
(b) integument apex,
and (c) ovule
position
characters.
nettitales
are unknown
or mixed
for these
characters.
Despite
its large
number of apparent
synapomorphies,
the glossophyte
clade
decays
in two steps, and
its boot-
strap
value is only
46%S.
However, this
may reflect
the
relatively
weak association
of Pentoxylon,
which
is
linked with
cycads in some
two-off
trees.
Glossophytes have
other
characteristic
features
whose status
depends
on outgroup
relationships.
When
they
are
linked
with cycads
and other Mesozoic
seed
ferns, simple
pinnate
leaves
and loss of the
bifid
in-
tegument
apex (seen
in
all
platysperms
discussed
so
far
except
cycads)
are
synapomorphies
of
cycads
and
glossophytes
(figs.
6a, 8b).
However,
when glosso-
phytes
are
lower
on
the
tree,
as the
sister
group
of the
coniferopsid
line
(fig.
6b),
their nonbifid integument
is
primitive
(as it
is
in
cycads),
and
the point
of
origin
of
simple
pinnate
leaves is
equivocal.
Although
ad-
axial
ovules
are a unique advance
of
glossophytes
(with
the possible
exception
of corystosperms),
the or-
igin of
this condition
poses problems.
When
glosso-
phytes
are
the
sister
group
of the
coniferopsid
line
(fig.
6b)
or
linked
with
cycads
alone
(fig. 6c),
ovule
posi-
tion
in
their ancestors
may have
been
either abaxial
or
apical-marginal.
However,
when Mesozoic
seed
ferns
with abaxial ovules
are on
the line
leading
to
glosso-
phytes (fig.
6a),
there must be
a
transition
from
abaxial
to adaxial ovules
(fig.
8c). Since
this
may
be hard
to
visualize as
a
single-step
change,
it
may
argue
for a
lower position
of the glossophytes.
When
glossopterids
are basal
in the
glossophytes
(fig. 6a,
b),
their sister
group
consists
of the former
anthophytes
plus
Caytonia,
which
is linked with an-
giosperms.
As
discussed below,
this
is consistent
with
the
views
of
Gaussen
(1946),
Stebbins
(1974),
Doyle
(1978),
and Retallack
and
Dilcher
(1981)
on carpel
and
cupule
homologies.
Since
the
Caytonia-antho-
phyte clade
corresponds
to
the former
anthophyte
clade
in
a
phylogenetic
sense,
as all the derivatives of
the
most
recent
common ancestor
of its former mem-
bers,
I
will
continue
to call
it
the anthophytes.
It must
be
recognized,
however,
that
this
group
does not
cor-
respond
to
the
original
anthophyte
concept
in
a
typo-
logical
sense,
since it includes taxa (Caytonia,
Piro-
conites)
that did not have flowers.
Synapomorphies
of
the
expanded
anthophyte
clade are reduced
megaspore
wall
and
granular
exine
structure (reversed
in
Cayto-
nia).
Endarch
leaf
traces,
multiseriate
rays (a
reversal),
and fused
microsporangia
are
synapomorphies
of the
groups
above
Pentoxylon,
which resembles
glossop-
terids
in
having
the alternative
states.
Of
these
char-
acters,
leaf trace maturation
and
ray
width
are un-
known
in
Caytonia.
In
trees
like that
in
figure
6a,
characters uniting
Caytonia
and
angiosperms
are flat
guard
cells, pinnate
megasporophylls,
anatropous
cupules (=
bitegmic
ovules),
and
loss
of nucellar vasculature. Caytonia
and
angiosperms
are
both
"angiophytes"
as
defined by
Doyle
and
Donoghue
(1993):
members of
a stem-
based taxon
(de Queiroz
and Gauthier 1990)
contain-
ing
all derivatives of the
line
leading
to angiosperms
since their common ancestor with
their
closest living
relatives
(Gnetales).
Doyle and
Donoghue
(1993) de-
fined
angiosperms
as a
node-based
taxon:
i.e.,
the
crown
group,
consisting
of
all
derivatives of the most
recent common ancestor of living angiosperms.
Cay-
tonia
may
be characterized as a
stem
angiophyte.
The
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC RELATIONSHIPS
OF GNETALES S19
direct association of Caytonia and
angiosperms is
rel-
atively weak;
its
bootstrap
value is
58%, and
it
decays
in only one step. Filtering one-off
trees for those in
which Caytonia and angiosperms
do not form a
clade
shows that this rapid decay
is
due
to the
existence
of
one-off trees
in which
Pentoxylon
is linked with glos-
sopterids and Caytonia
is the
sister
group of other an-
thophytes.
Within
angiosperms, Nymphaeales
are
basal,
fol-
lowed by other paleoherbs, the
rooting favored by
rDNA data (Hamby and Zimmer
1992; Doyle et al.
1994; Goremykin et al. 1996). The
bootstrap support
for
angiosperms
is
very high (97%),
but
not
as high
as
in
Doyle
et
al.'s (1994) morphological
and
rDNA
analyses
of extant
taxa
(100%),
and
the
group decays
in
four steps. This
is
because many
angiosperm syn-
apomorphies involve characters
that are unknown
in
fossils
(companion cells, gametophytes,
fertilization),
so
it is
equivocal whether they unite
just angiosperms
or
angiosperms plus Caytonia.
The
only unequivocal
synapomorphies of angiosperms
alone are several
vein
orders,
closed
carpel,
two
pairs
of
microsporangia,
and
reduced endexine.
Filtering four-off
trees
for those
in
which
angiosperms
are
not
monophyletic
confirms that
decay
of the
group
is
a
result of
trees in which
Cay-
tonia
is nested
within
angiosperms
(e.g., just
above
Nymphaeales).
As anticipated, Piroconites is the
sister group
of ex-
tant
Gnetales,
linked with them
by opposite leaves,
parallel venation,
and striate
pollen.
On
analogy
with
Doyle
and
Donoghue's (1993)
terminology
for
angio-
sperms and angiophytes,
Piroconites
and
extant
Gne-
tales
may
be
designated
as
gnetophytes
(a
stem-based
taxon);
Piroconites
is a stem
gnetophyte.
The three
modern
genera
are united
by
compound strobili, sin-
gle,
terminal
ovule, bipartite
outer integument (pre-
sumably
derived from
two
perianth
parts),
and more
or
less
fused
microsporophylls.
As
in
all
previous
anal-
yses,
Welwitschia and
Gnetum
form a clade within
Gnetales,
united
by
more
than
one
order of
laminar
venation, paracytic stomata,
astrosclereids, four-nucle-
ate
male
gametophyte,
female
gametophyte
with
free-
nuclear
eggs
and
irregular
cellularization,
and
embryo
feeder. Based on
their
bootstrap
value
(98%),
Gnetales
are one of the two
most
strongly
supported monophy-
letic
groups
in
the whole data
set,
along
with
Welwit-
schia
plus
Gnetum
(98%),
contrary
to
Nixon
et al.
(1994).
The
position of Bennettitales
is highly unstable,
since there are
few
characters
supporting
a
direct
link
with
particular anthophyte
taxa and conflicts
among
these. In
figure 6a,
Bennettitales
are
linked with
Gne-
tales
(so
that
they
are
stem
gnetophytes),
based
on
their tubular
micropyle (fig. 8b);
the
state
of
this
char-
acter
is
unknown
in
Piroconites,
a
potential
weakness
of this
arrangement.
Placement of
Bennettitales
on the
angiophyte
line
(fig. 6b)
is
supported
by scalariform
pitting
in
the
secondary xylem (a
character not known
in
Caytonia).
In such
trees, pinnate
sporophylls and
anatropous cupules
are
equivocal
as
synapomorphies
of Caytonia and angiosperms,
because
the
megaspo-
rophyll character was scored as unknown
in
Bennet-
titales and their orthotropous cupules may
be derived
from the anatropous type.
A
basal position of
Bennet-
titales (fig. 6c)
is
supported by
their
open
leaf venation
(a
reversal in
other trees)
and
secretory
canals
(sym-
plesiomorphic with cycads here, convergent
in
other
trees).
Synapomorphies of extant
Gnetales and
angio-
sperms involving characters that
are not known in fos-
sils, such as lignin chemistry (Maule reaction),
a tunica
layer in the apical meristem, and double fertilization
of
the Ephedra type, could go back
either to
their
most
recent common ancestor
or
to a more distant
common
ancestor with
glossopterids.
In either
case,
the
tree
pre-
dicts
that these features existed
in
Caytonia
and Ben-
nettitales (except
when the latter are basal glosso-
phytes). Similarly, some synapomorphies of
extant
angiosperms,
such as
companion cells,
reduced
male
and female
gametophytes,
loss
of the
secondary
sus-
pensor,
and
endosperm,
could
go
back to
Caytonia
(or
to Bennettitales
in trees like that
in
fig. 6b).
It is
also
equivocal
whether the vessels and nonscalariform
me-
taxylem of modern
Gnetales
originated
above or
below
Piroconites.
New data on these characters
in
any
of
the
relevant
fossils
(e.g.,
on
the "aleurone
layer"
de-
scribed
in
Caytonia
seeds
by
Harris
[1958])
could
strengthen
or
weaken the
present
scheme.
I
examined
the
relative support
for
major
alternative
hypotheses on relationships
of
anthophytes
and
Gne-
tales
by
means
of the constraints
option
in
PAUP. The
shortest
typical neo-englerian trees,
made
by
forcing
anthophytes together
with conifers and
excluding glos-
sopterids
and
Caytonia,
are
eight steps longer
than the
shortest trees
(seven steps
if
Pentoxylon
is excluded
from
the
anthophytes,
in
which case it is linked with
glossopterids, Caytonia, and corystosperms).
Forcing
angiosperms together
with
Welwitschia and Gnetum,
as
in
the trees
of Nixon
et
al.
(1994),
is still
worse, as
expected
from the
high bootstrap support
for Gnetales:
it
adds 10 steps. These
trees
differ
from
those
of Nixon
et
al.
in that
glossopterids
are
interpolated
between an-
thophytes
and lower
groups.
A
more
interesting
alternative
hypothesis,
which re-
sembles
a
synthesis
of the
glossophyte concept
and
the
neo-englerian
view that
anthophytes
are derived from
a
coniferopsid-like ancestor,
is
shown
in
figure
9.
This
tree
deserves
closer
examination because
it
is
only
one
step
less
parsimonious
than the shortest trees. "Coni-
feropsids"
form not a clade but rather
a
paraphyletic
series. As in the most
parsimonious trees,
dichoto-
mously
veined
leaves,
scalelike
microsporophylls,
and
endokinetic
microsporangial
dehiscence arise at the
base of
this
series.
Ginkgoes
are basal and conifers are
linked
with
the
remaining groups
based
on
a shift from
secretory
cavities to
cells,
saccate
pollen,
loss of the
ringlike
sterile cell in the
male gametophyte,
and
ep-
igeal
seed
germination (cf.
Loconte
and
Stevenson
1990;
Nixon et
al.
1994). Cordaites
are linked with
glossophytes by
reversals to mesarch leaf traces and
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S20
INTERNATIONAL JOURNAL
OF PLANT
SCIENCES
Glossophytes
Coniferopsids Gnet flngiosperms
".1
rreelvn2 2
g
c-02v
tribution
.
La
tece
ering the age of cordaies).
RelationshipsZwitin
clom-
sophytes are generally
similar to~~~~E~~
simple picsse
above.~~~~~~~~~~~~~~~~~~~~~sevle
4,0c7.
401.
.[:
equivoca
FIg
9eronmed stepvess
l parsimoiousl texerinmwhic
"conifropsds
aeparaphyetic
ransd
glssphthes cnesited withien
them,shoin dis-lt
trbuind tofste lef
charater l 19)
ITperformeseeatadtoa
experimentswsitne to
pdrob te
possibility that the
present results are due to exces-
sively theory-laden
comparisons between the ovule-
bearing structures of angiosperms and Caytonia. I re-
moved the four characters
most intimately involved in
these comparisons: megasporophyll morphology
(29),
appendicular
vs.
terminal ovule position (30),
leaf sur-
face on which ovules are borne (31), and enveloping
structures ("cupules";
33). The resulting 56 most par-
simonious trees correspond
closely to those derived
from the whole data
set, with coniferopsids monophy-
letic and the glossophyte
group still forming a
dlade,
linked with cycads (fig. 10). In 25 trees the
basal group
is glossopterids or
Pentoxylon, in the other 31 Ben-
nettitales, with glossopterids linked with
Pentoxylon,
as in figure 6c. Piroconites
is linked with extant Gne-
tales, Caytonia with angiosperms. Thus
a
significant
portion of the support
for the arrangement based on
the whole data set
is independent of the contested
megasporophyll and
cupule characters.
The second experiment
tested the contention that
earlier results were
due to incorrect assumptions on
ancestral states in
angiosperms, and if the first angio-
sperms were more like Chloranthaceae,
it might be
more parsimonious to nest them within Gnetales.
This
view is contradicted by the present analysis,
which
nests Chloranthaceae among the 10 other
angiosperms,
linked with core Laurales.
However, many of charac-
ters used within angiosperms might be challenged.
The
experiment
consisted
of substituting
Chloranthaceae
for angiosperms as a
whole.
The 16
resulting
trees
differ from those based on the whole data
set
in
that
Glossophytes
Coniferopsids
Gnet
flngiosperms
f~~~~~4
-0
-
,V A C
04 .:,
EV 3:
-_
.2
"
X3c
Mo<X
X
_w3uL
ax
,VO,
3OzL
Mu 3n
Eu
2 2
Fig. 10 Strict
consensus
of most parsimonious
trees
found after
removal
of four characters
related
to megasporophyll
and
cupule
morphology.
coniferopsids
may be
either monophyletic
or paraphy-
letic,
but
as
before the
glossophytes
form
a cade (fig.
11). Relationships
within
the
glossophytes
differ
in
that Caytonia
is
basal and
glossopterids
are the
second
branch,
followed
by Pentoxylon.
However,
Chloran-
thaceae
are
linked with
Bennettitales
rather
than Gne-
tales,
based
on paracytic
stomata,
scalariform
second-
ary xylem
pitting,
and orthotropous
bitegmic
ovules,
which
together
with other
nongnetalean
features
of
Chloranthaceae
(e.g.,
pinnately
veined
leaves)
out-
weigh
gnetalean
features such
as opposite
leaves
and
vessels. Scalariform
pitting
does
not
occur
in Gnetales,
while the
bipartite
envelope
that
might
be homolo-
gized
with the
chloranthaceous
outer
integument
arises
between
Piroconites
and
Ephedra,
and
paracytic
sto-
mata originate
in the
Welwitschia-Gnetum
clade.
Since
this
study
was stimulated
by
the
idea
that
Pi-
roconites
might
affect
the
position
of
Gnetales
(Doyle
1994),
I
also
removed Piroconites from the data set.
Coniferopsids
Gnet
Treele th: 87 antroou
VI:
0 orkhotropous
Fi
C
Representative
os
parsimonious
tre fdft su
tution
of
Chloranthateae
fort
opersmsnoastree
a
u
who le
tuino
hornhca
oragoprs
sawoe
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE
PHYLOGENETIC
RELATIONSHIPS
OF GNETALES
S21
Surprisingly,
this had
no effect on
the
arrangement
of
the
remaining groups.
The 76
trees obtained
are a
sub-
set of
those found
when Piroconites
is included:
the
glossophyte
clade
is intact;
the basal
branch is either
glossopterids,
Pentoxylon,
or
Bennettitales;
and Cay-
tonia
is
linked with angiosperms.
Furthermore,
the
shortest neo-englerian
trees (with
Pentoxylon
excluded
from
the
anthophytes)
are seven
steps
longer
than the
shortest trees,
the
same difference
found
when
Piro-
conites
was
included.
However,
this
does not
mean
that Piroconites
is worthless
in
understanding
the or-
igin of
Gnetales,
since it does
clarify the
course
of the
reduction
series
leading to
the modern
genera.
The
new results may
be a function
of more accurate
rela-
tionships
within
conifers,
new data on
Pentoxylon,
and
scoring
of the microsporophylls
of
Gnetales
as
branched rather
than
simple
(following
Nixon et
al.
1994).
Discussion
SCENARIOS
FOR ORIGIN OF GNETALES
Trees with
glossopterids
at the
base of the
glosso-
phytes
(fig. 6a,
b) suggest
the
possibility
that
Gnetales,
Caytonia,
angiosperms,
Bennettitales,
and
Pentoxylon
were
derived
from the radiation
of
glossopterids
in the
Permian
of Gondwana,
which had
a
cool,
seasonal,
postglacial
climate. This would
be a
synthesis
of the
view
of Retallack
and
Dilcher (1981)
that angiosperms
were derived from
glossopterids
and
the
conjecture
of
Schopf
(1976)
that Gnetales
had a similar
origin.
Typ-
ical glossopterids
apparently
died out
in the mass
ex-
tinction at the end
of
the
Permian
(Retallack
1995);
the
anthophyte
line would
be the sole survivor
of this
event.
Pentoxylon
is the one
glossophyte
that remained
in the
original
environment, being
restricted
to tem-
perate
Southern
Gondwana
in the
Early
Cretaceous
(Drinnan
and
Chambers
1985),
and it had the
most
cold-adapted
features,
some retained from
glossopter-
ids
(pycnoxylic
wood,
deciduousness),
some
new
(short
shoots).
The
other
groups
radiated
into more
tropical
areas
during the
Mesozoic (Doyle
and
Dono-
ghue
1986, 1993;
Vakhrameev
1991).
In the
present
trees, glossopterids
have two
auta-
pomorphies
(striate,
saccate
pollen),
which
imply
that
they
are a
relatively plesiomorphic
sister
group
of
an-
thophytes
rather than
a
paraphyletic
"ancestral
com-
plex"
from which
anthophytes
were derived.
However,
this
inference may
be
premature.
First,
as noted
above,
it "costs"
only
one
step
to
consider
saccate
pollen
a
plesiomorphic
feature.
Second, glossopterids
were
more
varied than the reconstruction
used here might
suggest
(Pigg
and
Trivett
1994).
A
relevant example
is their variation between one
and
several
sporophylls
("cupules")
per
bract-sporophyll
complex (Schopf
1976).
Considering
implications
for leaf evolution
(fig.
6),
these
results
imply
that the
glossopterid
leaf
may
be
basic for
the
whole
glossophyte
clade.
In all
trees,
sim-
ple pinnate
leaves are
ancestral,
whether as a new fea-
Glossophytes
Coniferopsids
Gnet
Rngiosperms
4
,
4,
M
M
W
c @=x W
0000
0''
6
v' 0000200E
0.........**
CC
04
5
21=
-
O-
50!
z11_1
op
| Treelenqth
247 1
\<
W //
~~~~~~~~~~~~~~rticl
Fig. 12 Most
parsimonious tree
shown in
fig. 6a, showing
distri-
bution
of the fine venation
character.
ture or
a
synapomorphy
with
cycads.
Caytonia
leaves
have four
leaflets
that resemble
whole
glossopterid
leaves;
these
are
secondarily
compound.
Mexiglossa
(Delevoryas
and
Person
1975),
a
Glossopteris-like
leaf
from
the Jurassic
of Mexico,
which
is associated with
Caytonia-like
microsporophylls,
could
be an interme-
diate.
Angiosperm
leaves would
be derived
from
a
glossopterid
type
by
increase
in the number
of vein
orders, perhaps
as a result
of the
origin
of diffuse
mer-
istematic
activity (cf.
Doyle
and Hickey
1976).
In the
gnetophyte
line,
major
venation
became
parallel
in Pi-
roconites
and modem Gnetales,
with
a reversal
to pin-
nate
in
Gnetum.
Optimization
of the
fine
venation
character is
uncertain,
but
when
glossopterids
are basal
in the
glossophytes
(fig.
12),
their reticulate
venation
may
be ancestral.
This
would
mean that
the
open
ve-
nation
of
Pentoxylon
and
Bennettitales is
secondary.
However,
these
inferences
depend
on
scoring
fine
ve-
nation
as unknown in
Piroconites,
which
may
have
rare
reticulations
(J.
H. A. van
Konijnenburg-van
Cit-
tert,
personal
communication),
and
Ephedra,
in which
the
two or
three
veins sometimes
fuse
(Foster
1972).
If
the venation
of these
groups
is scored
as
open,
the
basal
state for
glossophytes
as
a whole is also
open.
Under
alternative
one-off
trees
in which
glossophy-
tes are nested
within
coniferopsids
(fig.
9),
the
glos-
sopterid
leaf
type
would
be derived
from a "conifer-
opsid"
type-not
a
conifer needle,
but
a
parallel
or
dichotomously
veined
leaf,
as in cordaites or
ginkgoes.
A
possible
intermediate
might
be the
genus
Ganga-
mopteris,
which is like
a cordaite leaf
with reticula-
tions and the veins
crowded
together
toward
the
mid-
dle
(Schopf
1976).
If
glossopterids
are
basal in
glossophytes,
scenarios
for evolution of
the
ovule-bearing
structures
(fig.
13)
are
consistent with the
hypotheses
of
Stebbins
(1974)
and Retallack
and Dilcher
(1981).
Inferred
transfor-
mations are
diagrammed
in
figure
14.
The
compound
strobili and
simple
flowers of Gnetales
are
purely
con-
vergent
with those of
coniferopsids.
Their ancestors
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S22 INTERNATIONAL
JOURNAL OF
PLANT SCIENCES
Glossophytes
Coniferopsids
Gnet
Rngiosperms
-0
=A,?
4,
"
4"
t
,-xW?
c?E'
\\\\\
4,
V
/
2
A
2'0
%AA
o_
V
.~~
~ 2
2
-
\\\\\ _LX/
fl \ \
-
Megas.
ro
.
-
poros-
Teenth:
247
IE\\ \Pn_nsimle
ieqcierOaial
Glossophyjtes
Coniferopsids Gnet
Ringiosperms
2 2 ~~~
~~~~
2
4
22
2r
L
2
2, 4, m 0
L_
A,
.
-
2
A ? A
2 2
2O~* , ?*c-^fixE-n=
*
r1
? -
4,c
,
4,
9
s:
a do L) Z <n
C E:
_j
) -
w
X : s:ir
Treelen2th
247
=
p
a
\
t _
o~~~~~~~~~~ucrtainoou
characters. radial
lo
oht
rophyll
eromplexes
ofGglossopterd
flndgPiospcniters,
which were later'aggregatedAinto
flowers, andrthese
figo1e
Mos
parsioiu
trce
show
inrgae fign6tsown
distri-d
butionii
of
ehe
the
(a)dlmegaesporophyllsad()oueevlp
(f
cuple")
rohl opee
fgosopterids, etxln and Piroconites,aehmlgu
with those of peltasperms and corystosperms or
inde-
pendently derived from pinnate
sporophylls,
as in cy-
cads,
depends on arrangement of the outgroups.
As
noted above, the fact that ovules are abaxial in
pelta-
sperms but adaxial in glossopterids and
Piroconites
may argue against the view that their similar-looking
sporophylls are directly homologous.
Schopf (1976) suggested that the glossopterid
"spo-
rophyll"
was derived from a coniferopsid axillary
fer-
tile
shoot with many simple
megasporophylls,
but this
would conflict with its leaflike anatomy (Taylor
and
Taylor 1992). Although
the
bract-sporophyll
complex
may represent a leaf fused to its axillary shoot,
the
anatomical
data imply that the
sporophyll itself
is a
leaf. An alternative hypothesis, which is more
plausi-
ble if
glossophytes
are near
the base
of the platy-
sperms,
is
that the
bract-sporophyll complex
was
de-
rived from a fertile frond
of the type found
in the most
primitive seed ferns,
where
ovules were borne on
a
three-dimensional continuation
of the
frond rachis
(Serbet
and
Rothwell
1992;
Stewart
and Rothwell
1993).
In gnetophytes, it may
be plausible
to derive Piro-
conites from a glossopterid
prototype, but the changes
between Piroconites and
modern Gnetales
are more
problematic. The bract-microsporophyll
complex
of
Piroconites is comparable
to the branched microspo-
rangium-bearing
structure attached
to a
leaflike
bract
in
glossopterids (Schopf
1976). However,
this would
have to be reduced to
a much simpler structure, seen
in its most complex form
in Welwitschia, with three
trilocular
synangia.
Reduction
from
many microsynan-
gia to three is not hard
to imagine, but
all
sign of
the
bract would have
to
be lost
at the same time. The
Piroconites bract-megasporophyll
complex
would have
to be reduced
still more, to
a
single,
terminal
ovule.
The Triassic genus
Dechellyia (Ash 1972) might
rep-
resent
an intermediate
stage,
where
each of the
two
ovules at the branch
tip
has a
wing
that
might represent
the
bract,
but
again
there is no
obvious
homologue
of
the bract
in
modem Gnetales.
However,
the male flow-
er of
Welwitschia
has two mounds
at the base
of the
terminal
ovule
that
might
be
vestiges
of bracts
(Mar-
tens 1971;
Hufford
1996).
The fact there are two
of
these
might
be
explained
if
the ancestor of Gnetales
had opposite winged
ovules,
as
in
Dechellyia,
one of
which was
sterilized
while the other was shifted
to a
terminal position. Eames
(1952) postulated
a
similar
scenario
for
Ephedra,
based on
analogy
with the
pres-
ence of two
sporophylls
in the male
flower
and
with
reduction from two axillary
"ovules"
(= flowers)
to
one
apparently
terminal "ovule"
in
the
compound
fe-
male
strobili of some
species.
Turning
to other
anthophyte groups,
the
trees
in
fig-
ure 13
suggest
a
scenario
for the
origin
of the
angio-
sperm carpel
that
may
be
more
plausible
than
expan-
sion
and
folding
of the
Caytonia
rachis
(Gaussen
1946;
Doyle 1978). Caytonia
is on the
line
to angiosperms,
but with Piroconites situated on the
adjacent
branch
and
Pentoxylon
and
glossopterids
below
both
lines,
the
common ancestor
of
Caytonia
and
angiosperms
may
have had
glossopterid-like
bract-sporophyll complex-
es.
Presumably
these would have several
sporophylls
(future anatropous cupules
=
bitegmic ovules)
per
bract rather
than
one. The bract would
only
have
to
be folded
lengthwise
to
produce
a
carpel,
or
reduced
to
produce
a
Caytonia
sporophyll.
For
angiosperms,
this
corresponds
to the view of
Stebbins
(1974)
and
Retallack
and
Dilcher
(1981),
and
it is
analogous
to
the scenario of Kato
(1990),
who derived
glossopterid
and
angiosperm
structures from the leaf
of
Ophioglos-
saceae,
with its
adaxial
fertile
segment.
If
the bract-
sporophyll complex
is
ultimately derived
from an ax-
illary
shoot and
its
subtending leaf, the angiosperm
carpel
is
a
composite
structure,
rather than
a
sporo-
phyll
in a
strict sense.
A
weakness of this scenario is
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS
OF GNETALES S23
eHtanlt
Gne
tales
Piroconites
Dechellyia
Glossopterids
fingiosperms
Caytonia
Fig.
14
Scenario for evolution of
ovule-bearing
structures
in
Gnetales,
Caytonia,
and
angiosperms
from
a
glossopterid prototype
that it predicts that the
bitegmic ovules should be
borne on the midrib of the
carpel rather than at its
margins, as
in
most angiosperms.
However,
this
may
be
consistent with the laminar
position
of the ovules
in Nymphaeales (Taylor 1991;
Doyle 1994), which are
basal in trees based on nuclear
and chloroplast
rDNA
sequences (Hamby and
Zimmer
1992; Doyle et
al.
1994; Goremykin
et al. 1996)
and those found
in
the
present analysis.
In Bennettitales,
a
glossopterid
prototype for the
male structures
is
consistent with
the fact that
many
members
had flat
microsporophylls
with adaxial
spo-
rangia (compared with Piroconites
by van Konijnen-
burg-van
Cittert
[1992];
cf. Crane 1988). However,
the
presence
of an
orthotropous
cupule around the ovule
(Harris 1954; Crane 1985,
1988) is now more difficult
to
explain (cf. Doyle
1994).
In
the trees of Crane
(1985)
and
Doyle and
Donoghue (1986, 1992), where
anthophytes
were
linked with
corystosperms and/or
Caytonia,
an
anatropous
cupule originated
below the
anthophytes,
and
the
bennettitalean cupule
could be
derived from this
structure
by a change in orientation.
However, anatropous cupules
are now restricted to
Caytonia
and
angiosperms (fig.
13b);
their homo-
logues
in
other
lines were still laminar rather than
cu-
pule-like. It is tempting
to homologize
the bennettita-
lean ovuliferous
receptacle with
the multiovulate
sporophyll
of
Piroconites
and
the multiovulate head
of
Pentoxylon, which
is apparently
a modified
leaf
(Roth-
well and Serbet
1994). However,
this
would leave no
homologue
for the cupule
of Bennettitales. Unless
this
cupule originated
de
novo,
it
seems
necessary
to as-
sume that the
ovuliferous
receptacle
was derived from
a bract-sporophyll
complex
with numerous sporo-
phylls,
each reduced
to one
cupulate
ovule, again
with
loss of the bract.
At the same
time, the relationships
found here
ex-
plain the absence
of a cupule in Gnetales,
which posed
a
problem
when
cupules
were inferred to be basic in
anthophytes (Doyle
and
Donoghue
1986; Doyle 1994).
Based
on
the
present trees (fig.
13b), ovules
on
the
line leading to
Gnetales
were never surrounded by
a
typical cupule.
This also goes for
Pentoxylon,
assum-
ing that Nixon
et al. (1994) and
Rothwell and Serbet
(1994)
were
correct
in
rejecting
the hypothesis of
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S24
INTERNATIONAL
JOURNAL OF
PLANT
SCIENCES
Crane
(1985)
that
its so-called
sarcotesta is
actually
a
cupule.
With
Piroconites
on
the
gnetalean
line and
Caytonia
on
the
angiosperm
line, these
results
imply
that flower-
like
structures
originated
independently
in
Gnetales,
Bennettitales, and
angiosperms
(Doyle
1994). This
is
supported
by
evidence that some
Bennettitales
did
not
have
typical
flowers,
including
taxa
that may
be basal
in
the
group
(Crane
1988).
The
hypothesis that
glossopterids
are
paraphyletic
relative to
Gnetales
and
angiosperms
might
explain
the
weak,
conflicting
molecular
evidence
for
a
relationship
of
the two
living
groups,
especially
if
they
were
de-
rived from
different
glossopterids
(e.g., species with
one
vs. several
sporophylls
per
bract-sporophyll com-
plex).
Under
both
previous
and
present
trees, the com-
mon
ancestor
of the two
lines must be at
least as old
as Late
Triassic,
since
Bennettitales,
stem
gnetophytes,
and
Caytonia are known back to
this time
(Doyle
and
Donoghue
1993).
However,
glossopterids
apparently
died
out
in
the mass extinction at
the end of the
Perm-
ian
(Retallack
1995).
If
they are
paraphyletic, the com-
mon ancestor of
Gnetales and
angiosperms could
be
well down in the
Permian,
if
not Late
Carboniferous,
hardly more recent than the
common
ancestor
of co-
nifers and
cycads
(presumably
Early
Carboniferous).
These results also have
interesting
implications
for
scenarios
of
ecological evolution.
Doyle
and
Dono-
ghue
(1986)
argued
that
accelerations of the life
cycle
in
anthophytes
were
adaptations to
dry tropical
con-
ditions,
since
Jurassic
Bennettitales and
Early
Creta-
ceous
Gnetales and
angiosperms
apparently
preferred
such
climates
(Doyle
et al.
1982; Crane and
Lidgard
1990;
Vakhrameev
1991).
However, the new
trees
sug-
gest
another
possibility,
since
they
imply
that common
advances
of
angiosperms and Gnetales
may
go
all the
way
back to
glossopterids.
Most of
the
Carboniferous
taxa on the line
leading
to
glossophytes lived in the
tropical
rain
forests of the
Euramerican
province,
whereas
glossopterids
lived in
the seasonal
temperate
zone
of
Gondwana.
Figure
15
shows the
inferred evo-
lution of
ray
width,
which
is an
aspect
of
the
vaguer
distinction between
manoxylic
and
pycnoxylic
wood.
It
suggests
that
pycnoxylic wood
originated
in
the
tropics
(perhaps
as an
adaptation
to
increasing
aridity,
which
culminated in the
Permian)
but was
preadapted
to
cold
climates. An
analogous
pattern
is seen in
the
pycnoxylic
coniferopsid
line,
where
cordaites
and
Permian
conifers
were
tropical
but
ginkgoes
and extant
conifer
families
radiated
into
temperate
regions
of
both
hemispheres
in the
Mesozoic
(Vakhrameev
1991).
Per-
haps
some
of the
reproductive advances
of antho-
phytes,
such as
double
fertilization of the
gnetalean
type
and
epigeal
seed
germination
(which arose inde-
pendently
in
conifers),
originated
in
glossopterids
as
adaptations
to
a short
summer
growing season but
pre-
adapted
anthophytes
for later
radiation in
the season-
ally
arid
tropics.
However,
reinvasion
of
the tropics
was
accompanied
by
reappearance
of
multiseriate
rays
Glossophytes
Coniferopsids
Gnet
Rngiosperms
bi o
t,
a
of the
dtiob
twenanxyic
and p
wo od .
A
;
gin|o
scalariformsecyleminCth|
MltterOtw
Fig.
15-
Most- pasmoiu
tree
shw
Cnfi.6,sowig
isri
twe
aoxyi
and
pycnxyli
wood.
gi.o
sclrfr
seon
ar
xye
nte
atrto
Conclusion
Although
the
relationships
of
Gnetales
are
still
in-
completely
resolved,
both
morphological
and
molec-
ular
evidence
imply
that
they
are a
monophyletic
group, like
angiosperms,
and
morphological
data
in
particular
indicate
that
the
two taxa
are
somehow
re-
lated.
The
conclusion
of
Nixon
et
al.
(1994)
that
Gne-
tales are
paraphyletic
relative
to
angiosperms
is
weakly
supported
by
characters
that I
believe
can
be
rejected
on
morphological
grounds,
and
it is
strongly
contra-
dicted
by
molecular
data.
Probable
fossil stem
relatives
of
Gnetales,
such
as
Piroconites
and
Dechellyia,
sug-
gest
that
the
flowers
of extant
Gnetales
and
their
com-
ponent
structures
are
reduced
rather
than
primitively
simple.
Although the
conclusion
that
Gnetales
and
an-
giosperms
are
related
to
glossopterids
is
far from
es-
tablished,
it is
sufficiently
supported
by
new data
that
it
should
be
taken into
account
in discussions
of
ho-
mologies
in
both
groups,
and it
may
serve
to
focus
attention
on the
morphology
of
structures
in
known
fossils
and
to
provide
a search
image
(not
necessarily
the
only
one) in
attempts to
recognize
relevant
new
fossils.
Better
information
on
critical
characters
in
fos-
sil
groups
such
as
glossopterids,
Caytonia,
Sanmigue-
ta,
Piroconites,
and
Dechellyia
could
resolve
some
of
the
present
uncertainties,
but
reconstruction
of
new
fossil
taxa
on
the
stem
lineages
of
Gnetales
and/or
an-
giosperms
could
be
far more
decisive.
Acknowledgments
I
wish to
thank
Geeta
Bharathan,
Michael
Dono-
ghue,
Vera
Ford,
Ned
Friedman,
Leslie
Gottlieb,
and
Michael
Sanderson
for
discussion
of problems
of
char-
acter
analysis;
Ned
Friedman,
Dean
Kelch,
Kathleen
Pigg,
and
Han
van
Konijnenburg-van
Cittert
for
use
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS
OF GNETALES
S25
Elkinsia
0??000000?????0000010000??0?000000??0?001?0??0000???00000000????O0?000?000?????000???????O?
Lyginopteris
0?00000000000?1100000000??1?000000??0?00100??0000???01100000????O0?0000000?????000???????O?
Medullosans
O?OOBOOOOOOOO?10?00000000?2?000010??0?00100??0000???01001011????00?0100000?0??0000???????O?
Callistophyton
0?0000100??00?11000000000?1?001010??0?10100??0000???12001111?????0?1111000000???00???????O?
Cordaitales
O?A002?000000?11010000010?0??00010??1?00010??2000???12001111?????0?0?11000?00?0000???????O?
Emporia
0?0004??000???1101000001??0?200110??1?200?0??1100???12001000?????0?0111000????0??????????1?
Pinaceae
00OA04??0001001101100001012020?110??1?11010??1110????2111100????10011110000000100000?010011
Podocarpaceae
000004??0001001101100001002A20?110??1?11010??1110????2111100????10011110000000100000?010011
Araucariaceae
OOOOBF?00001011101100001002020?110??1?10010??1111????20111000???1001002000?000100000?010011
Taxodiaceae
000004??0001001101100001002020?AlO??1?1001O??llAl????2A111000???10010020000020100000?010011
Cephalotaxus
000004??0001001101101001002020?010??1?10010??111????12111110????1011002000?02?100000?010011
Taxaceae
000004??0001001101101001002021?030??1?10010??00?????12111100????1?12202000?02?100000?010011
Ginkgoales
000102?000010011211000010010200010??1?AA010??0000???121011100???10011010000001000000?000010
Corystosperms
O?OOOOOOOOOlO?1111000001??l?AOC020??0?10000??0000???120?????O????1?111100000???????????????
Autunia
?????0100?0?????????????????101010??0?100?0??0000???12?????????????11110000????????????????
Peltaspermum
?????0100?0?????????????????101010??0?10000??0000???120?????O????0?110?0000????????????????
Cycadales
OO?OO1?OOOOOOOllOA1A0000002000AO1O??O?1OAOO??OOOO???Al11111O???10011010000001000000?000010
Glossopterids
0?0?01?100000?11B?000001??3?102010??0?BOO?O??000????llOlllO?l????0?111110000????O??????????
Caytonia
??0?00?1010?????????????????002020??0?10110??000????1101??001????1?111100000???????????????
Bennettitales
0?0001?00011??11010100000?2??0?030??0?2010A??000?????3AlllOlA???11?110200000???000???????1?
Pentoxylon
0?0101?000000?112?0000010?3?10?010??0?00000??000????110??1111????1?110200000???????????????
Piroconites
????22??0?0O?????????????????102010??0?201?0??000????1??????????????11021000????????????????
Ephedra
001022??0001011121100100003121?040??0?00101??200?????31111010???1102102100?000100010?100011
Welwitschia
001022?11011101121100110002121?040??0?00101??200????131111010????101102100001?1111?0?100111
Gnetum
001?21?110111111111?01?0003121?040??O?O??01??200????031111000????102202?00?01?111110?100111
Magnoliales
OOOO11?lllllAlllD1010100114100202100201012001000?010?1011100100111A1102000013?1020111101011
Eupomatia
000011?11111011111010100114?00202100203012001000??01?10111001001110?002000013?1020111101011
Austrobaileya
000021?11111011121010100104100202100202012001000?000?101110010011101103010123?1020111101011
Chloranthaceae
A00021?111110111210101001041?0?031?1B03012??0200?2?0?10111001?001101103011123?102011110101?
Core
Laurales
000021?111110111210101001141?0202101203012000000?001?10111001?001102202??1??3?102011110101?
Winteraceae
000001?11111111131010000104100202100203012000000?000?101110011101101003010023?1020111101011
Eudicots
A00003?lllAlOlllD10101001031002021A0213012000BOO?AAO?1O1110011AO110?003010123?1020111101011
Aristolochiac
100013?11101011131010110124100202110213012010000?111?1011100110011011030100?3?102011110101?
Piperales
100013?111?1011111010100104100?03110213012010200?2?0?10111001100110110300A123?lA20110101011
Nymphaeales
110003?lllOlA111310??00?103?002021102120120AOOOO?AAO?101110011101101102000023?1020110101010
Monocots
110OOE?111?10111110??10?12G100202110213012010000?110?10111001??011A11030?0013?1020111101011
Fig.
16 Data matrix used
in
the
present analysis.
A
=
0/1;
B
=
0/2;
C
=
1/2;
D
=
1/3;
E
=
2/3;
F =
2/4;
G
=
3/4
of unpublished data; and Larry Hufford
and Gar Roth-
well for constructive comments on the
manuscript.
Appendix
Taxa
and
characters
The data matrix for the
present analysis
is shown in figure
16. Where definitions and scoring of taxa and
characters are
the same as
in
Doyle and Donoghue (1986, 1992)
and Doyle
et
al.
(1994)
for seed
plants
and
Donoghue
and
Doyle (1989)
for
angiosperms, and are
not
contested by
Rothwell and Ser-
bet
(1994)
or Nixon et al.
(1994),
readers
are
referred
to the
earlier papers for references and detailed argumentation.
TAXA
1. Elkinsia
(=
"Devonian seed
fern" of
Doyle
and Don-
oghue 1992). Based
on
Serbet and Rothwell
(1992).
2.
Lyginopteris.
3.
Medullosans
(Quaestora
and
Medullosa). Assump-
tions on
characters
of
Quaestora based
on
Rothwell
and Ser-
bet
(1994).
4.
Callistophyton.
5. Cordaitales.
A
consensus of
Mesoxylon
and Cordaix-
ylon as reconstructed by Rothwell and Serbet
(1994).
6.
Emporia (originally
described as
Lebachia
lockardii;
Mapes
and
Rothwell
1984). Scoring
based on Rothwell
and
Serbet
(1994).
7.
Pinaceae.
8.
Podocarpaceae. Morphological and
molecular analy-
ses differ as to whether Saxegothaea or a group
of scale-
leaved genera is basal among the modem
representatives
(Kelch, in press, and personal communication).
Rissikia (Tri-
assic) and Mataia (Jurassic) are
assumed to be stem relatives
(e.g., in scoring pollen characters).
9. Araucariaceae.
10. Taxodiaceae (including Cupressaceae,
assumed to be
a derived subgroup; Hart 1987; Chase et
al. 1993). Sciadop-
itys might
be treated as
separate taxon,
but
in
the
study
of
Hart (1987) it was the sister group of Taxodiaceae,
and none
of the characters that changed between the
two groups were
used
in
the present analysis.
11. Cephalotaxus.
12.
Taxaceae.
13. Ginkgoales. Interpretation
of
the
ovulate structures
as strobili
bearing simple sporophylls
is
supported by
data
of Zhou and
Zhang (1992)
on Baiera
from
the Jurassic
of
China.
14.
Corystosperms.
Rothwell
and Serbet (1994)
based
cupule and
ovule characters
(e.g., round,
adaxial ovules)
on
cupules
described in
Taylor
and
Taylor
(1993), named
Pe-
triellaea by Taylor
et al.
(1994). However,
Taylor et al.
(1994) questioned the relationship
of these cupules to cor-
ystosperms,
based on absence of resin cavities and differ-
ences from the foliar
anatomy
of Dicroidium (the leaf of
corystosperms).
The
presence
of several ovules
per cupule
(more similar to Caytonia) also conflicts
with
the
original
compression
material of
corystosperms
from South Africa.
The
present
treatment thus
follows
Doyle
and
Donoghue
(1986), based
on
compressions. Doyle
and
Donoghue (1986,
1992)
based stem
characters
on
Rhexoxylon,
which
is
notable
for
having
internal
secondary growth,
but
Meyer-Berthaud
et al.
(1993)
also associated stems with external
secondary
wood
only (Kykloxylon)
with
corystosperms.
The
present
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S26
INTERNATIONAL
JOURNAL OF
PLANT SCIENCES
scoring
represents a consensus
of these two
stem types. Leaf
anatomical
characters rely on
Pigg (1990a),
microsporophyll
characters
on
Yao
et
al.
(1995).
15. Autunia
(formerly
Callipteris, plus
associated repro-
ductive
structures; Kerp
1988).
16.
Peltaspermum (=
Lepidopteris of
Nixon
et
al. 1994).
17.
Cycadales. These are
assumed to
include Bjuvia
and
Beania, which
had
Taeniopteris
and related dissected
leaf
types
(Nilssonia), since these
appear to be
nested within
the
group (Stevenson
and Artabe
1995), but not
necessarily pro-
posed
Carboniferous-Permian
cycad relatives
with
Taeniop-
teris leaves
(Mamay 1976).
18.
Glossopterids.
Important new data
on anatomy
are
drawn from
Pigg and Taylor
(1993) and
Pigg and
Trivett
(1994),
and data on
ovule position
are from
Taylor
and
Tay-
lor
(1992).
19.
Caytonia.
20.
Bennettitales.
21.
Pentoxylon. Changes
in
character
scoring
are based
on
Bose et
al.
(1985)
and Rothwell and
Serbet
(1994).
22.
Piroconites
(Bernettia, Desmiophyllum
gothanii).
Based
on
Kirchner
(1992),
van
Konijnenburg-van
Cittert
(1992), and
Crane (1996).
23. Ephedra.
24.
Welwitschia.
25.
Gnetum.
26. Magnoliales.
"Core
Magnoliales"
of
Donoghue
and
Doyle (1989),
consisting
of
Degeneria,
Myristicaceae,
An-
nonaceae
(in
which
Anaxagorea
is
assumed
to
be
basal;
Doyle
and Le Thomas
1994),
and
Magnoliaceae.
Scoring
for
variable characters
is
based
on
relationships
found
by
Don-
oghue
and
Doyle (1989) and
Doyle
and Le
Thomas
(1994),
in
which
Degeneria or Myristicaceae were each basal in
some most
parsimonious
trees,
but
Annonaceae and
Mag-
noliaceae
were
always
linked with
one
or
another of the
remaining taxa.
27.
Eupomatia.
28.
Austrobaileya.
29.
Chloranthaceae.
30. Laurales. "Core
Laurales"
(MON)
of
Donoghue
and
Doyle (1989),
consisting
of
Hortonia, Monimiaceae sensu
lato,
Gomortega,
Hernandiaceae,
and Lauraceae. Monima-
ceae are
assumed
to
be
paraphyletic,
as shown in
Doyle
et
al.
(1994).
31.
Winteraceae.
32.
Eudicots.
Based
on
Ranunculidae,
Nelumbo,
and
Platanus,
which are
(near)
basal
according
to
analyses
of
rbcL and
atpB
sequence
data
(Chase
et al.
1993;
Drinnan et
al.
1994).
33.
Aristolochiaceae
(including Saruma,
assumed
to be
basal).
34.
Piperales
(Saururaceae
and
Piperaceae,
but not Chlo-
ranthaceae).
35.
Nymphaeales
(Cabombaceae
and
Nymphaeaceae,
but not
Nelumbo).
36.
Monocots. Based
primarily
on
Acorus, Araceae,
Al-
ismatidae,
and
dioscoreids,
which
are
(near)
basal
according
to
morphological
and molecular
analyses (Duvall
et
al.
1993;
Bharathan and
Zimmer
1995;
Davis
1995).
CHARACTERS
DDZ
designates
character
numbers
in
the nine-angio-
sperm analysis
of
Doyle
et
al.
(1994);
NCSF,
characters of
Nixon et al.
(1994); RS,
characters of
Rothwell and
Serbet
(1994).
All
multistate characters
are unordered.
VEGETATIVE
MORPHOLOGY
1
(DDZ
58, modified).
Habit
(0) woody,
(1) (semi)her-
baceous
(secondary growth reduced or
absent). Donoghue
and
Doyle (1989)
defined the corresponding character
in
terms
of
woody vs.
herbaceous
and scored
taxa
with a
cam-
bium but
"anomalous"
secondary
wood
(some
Piperaceae,
Aristolochiaceae) as
unknown (cf.
Nixon
et
al. 1994). By
lumping the latter types with
herbaceous,
the
present
defi-
nition
indirectly supports the
concept
that
reduction
of
sec-
ondary
growth
is a
step toward
loss and allows more taxa
to be
scored as
monomorphic. Dividing the
character into
three ordered states
(woody, secondary growth
reduced,
ab-
sent)
would
reintroduce
problems
of
polymorphism.
Fur-
thermore,
total loss
is
correlated
with
the radicle character
(2);
together,
the
two characters
may express
well
enough
the
similarities and differences
among
taxa that exhibit
vary-
ing
degrees of
herbaceousness.
2
(DDZ
57).
Radicle
(0) persistent, (1)
replaced by
ad-
ventitious roots. Fossils are scored as
unknown; although
groups with main trunks and
radiating
roots almost
surely
had the normal seed
plant
ontogeny,
it
may be
premature
to
extend this
to
early
seed ferns with stem-derived
prop
roots
(Stewart
and Rothwell
1993).
With removal
of
progymnosperms, apical
vs.
axillary
branching (DDZ 1)
becomes uninformative. The main
prob-
lem
concerns
cycads,
which branch
dichotomously (Steven-
son
1988).
Nixon et al.
(1994)
scored this as the
outgroup
state,
but it is not
exactly
comparable
with the
pseudomon-
opodial
branching
of
progymnosperms
and is not
unequiv-
ocally basic
in
the
group
(Doyle
and
Donoghue
1992).
Con-
trary
to Nixon et al.
(1994),
this
viewpoint
does
not depend
solely
on
interpretation
of Nilssoniocladus
(Kimura
and
Sek-
ido
1975)
as a
cycad,
although
I
do view this as its most
likely
affinity. Doyle
et al.
(1994)
and Rothwell and
Serbet
(1994)
scored
Lyginopteris
and
medullosans as
axillary, but
Nixon
et al.
(1994)
scored them as
unknown,
because the
mode of
branching in
the
former
is
not
typically axillary,
and
branching
has
been observed in
only
one
species
of the
latter.
3
(DDZ
2). Axillary buds
(0) single, (1)
multiple. Roth-
well and
Serbet (1994) and
Nixon
et
al.
(1994) treated these
as two
states of
a
three-state
character also
including apical
branching
(progymnosperms).
Nixon et al.
ordered
this
char-
acter,
as
implicit
in
treatment of mode of
branching
and
number
of buds as
separate
characters by
Doyle
and
Dono-
ghue
(1986, 1992),
but Rothwell and
Serbet treated it
as
unordered,
because
they questioned
the
implicit
assumption
that
single axillary
branching
is more
easily derived from
apical
branching
than
multiple branching.
This
may
obscure
a
significant
similarity
between the
two
kinds
of
axillary
branching (an example
of the
general problem
discussed
by
Maddison
[1994]),
but
the issue
becomes irrelevant with re-
moval
of
progymnosperms.
Although
it
may
be
problematic
to
score
Lyginopteris and Medullosa as
having axillary
branching (Nixon
et
al.
1994),
when
they
do branch
they
do
so
singly (Rothwell
and Serbet
1994). Cordaites are scored
as
(0/1)
because
multiple
buds
occur in
Mesoxylon (Rothwell
and
Serbet
1994).
4
(NCSF
2, modified).
Vegetative
short
shoots
(0)
ab-
sent,
(1) present.
Nixon et al.
(1994)
defined this
character
as
presence
or
absence of short
shoots and scored
cordaites
and
all conifers as
(1),
presumably
because
they
have
fertile
short shoots.
However,
cordaites and
most conifers lack
veg-
etative
short shoots.
Since
vegetative
and fertile short
shoots
vary
independently, they seem best
considered
two
charac-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS
OF GNETALES
S27
ters. Fertile short shoots are questionable as
a character in
their own right, since they are closely correlated with simple
sporophylls. In glossopterids, Pigg and Taylor (1993)
re-
viewed evidence on internode variation and concluded that
it is unclear whether this reflects possession of true short
shoots or
simply
seasonal
variation;
I have therefore scored
the group as unknown. Nixon et al. (1994)
scored
corysto-
sperms as unknown,
but
enough data
have accumulated on
their stem
anatomy
to
doubt that
they
had
short
shoots.
They
also
scored Gnetum as having
short shoots.
In
climbing spe-
cies,
most or all
foliage
leaves are
borne
on
short
shoots.
However,
in
G. gnemon and G. montanum there
is little
dif-
ference
between long
and
short shoots
(Pearson 1929;
J.
A.
Doyle, personal observation), and
I
have therefore scored
the genus as unknown.
5
(DDZ 3). Phyllotaxy (0) spiral, (1) distichous, (2) op-
posite
or whorled. Nixon
et
al.
(1994)
scored Taxodiaceae-
Cupressaceae
and Araucariaceae
as
unknown.
However,
Cu-
pressaceae appear to be nested among Taxodiaceae
with
spiral leaves (Hart 1987;
Chase
et
al.
1993).
Since
the
po-
sition of
Agathis
in
Araucariaceae
is less
clear,
this
family
is scored (0/2). Medullosans are scored (0/2)
because of
op-
posite phyllotaxy
in
Quaestora (Rothwell
and Serbet
1994).
Pigg and Taylor (1993)
documented
spiral phyllotaxy
in
glossopterids
and
rejected suggestions
that
some members
had
whorled leaves.
Opposite phyllotaxy
in Piroconites is
based on Crane
(1996).
6
(DDZ 4, modified).
Leaves
(0) pinnately compound,
(1) simple
and
pinnately
veined
or
dissected
into
parallel-
veined
segments, (2) linear
or
dichotomous
with two or more
veins, (3) palmately
veined
(actinodromous
or
acrodro-
mous), (4) linear
with
one vein (rarely two; may
fork
api-
cally, as
in
Emporia). Addition
of state
(4),
as in Rothwell
and
Serbet (1994), separates
the needle-like leaves of
coni-
fers
(a potential synapomorphy
of
the
group)
from the
mul-
tiveined leaves of
ginkgoes, cordaites,
and
Welwitschia.
Fol-
lowing
Rothwell and Serbet
(1994),
I
have eliminated the
cataphyll aspect
of the
definition
of
Doyle
and
Donoghue
(1986, 1992; Doyle
et al.
1994). Although cataphylls appear
at the same
point
as the inferred
change
from
progymno-
sperm
branch
systems
to fernlike
fronds,
it
was
probably
premature
to consider one
event
an inevitable
correlate
of
the other.
In
any case, specifying
the
presence
of
cataphylls
would be
superfluous,
since
they
are universal
in
this data
set. The distinction between
pinnately compound
and
simple
pinnate
leaves is discussed in the text. Rothwell and
Serbet
(1994)
scored
Podocarpaceae
as
multiveined,
but
such
leaves
(Nageia) appear
to be nested
well within the
family (Kelch,
in
press, personal communication).
This is less
certain
in
Araucariaceae. Rothwell
and
Serbet
(1994)
and
Nixon
et
al.
(1994)
scored
Caytonia
as
simple pinnate,
like
glossopterids,
but
although
their
leaves are
otherwise
similar, Caytonia
dif-
fers from
glossopterids
in
having
four leaflets rather
than
a
simple leaf. Rothwell and Serbet also scored corystosperms
as
simple pinnate,
but
although they vary
in
leaf
complexity,
they
all
appear
to be
pinnately compound
as defined here
(cf.
Retallack
1977).
Rothwell
and Serbet scored
Ephedra
as
one-veined,
but because
it
varies between two and three
veins
(Foster 1972),
I
consider it multiveined. Monocots are
scored
(2/3)
because
palmately
veined
leaves
occur in
aroids,
alismatids,
and dioscoreids
(Bharathan
and Zimmer
1994),
and work
by
G. Bharathan
(personal communication)
indi-
cates that
many
of
these are more like dicots than
typical
monocots in their
development.
7
(DDZ 5).
Rachis
(0) bifurcate, (1) simple. Rothwell
and Serbet
(1994) expanded
this
character
(RS 6),
defined
as occurrence
of
dichotomies,
where the states recognized
here correspond
to "at
base but
not
distally"
and
"in
none,"
to include states for "in all"
(progymnosperms, ginkgoes)
and "in some" (Emporia). However,
the "in all" state over-
laps with the previous character,
and both
it and the
Emporia
state are autapomorphic in the
present data set. As
in
Doyle
and Donoghue (1986, 1992),
I
have
scored only pinnately
compound groups:
the
concept
is based on
a
contrast
seen
among
early
seed
ferns,
and its extension
to
more modified
leaf
types
is
problematic.
8
(DDZ 6). Laminar venation
(0) open, (1) reticulate.
Ephedra is rescored as unknown
(as
in
Nixon
et al.
1994)
rather than open,
because the two
or
three veins occasionally
fuse
laterally (Foster 1972).
Piroconites
is
scored
as un-
known because
anastomoses may occur
but
are
not
positive-
ly demonstrated (J.
H.
A.
van Konijnenburg-van Cittert, per-
sonal
communication).
Conifer
taxa with one-veined leaves
are rescored as
unknown,
following
Nixon
et
al.
(1994);
Ar-
aucariaceae
are scored as
open,
based on
the
multiveined
leaves of Agathis. Nixon
et al. scored medullosans as
un-
known, presumably
based on Linopteris (cf. Taylor
and
Tay-
lor
1993),
but
I
assume
that this
is
derived
within the
group.
Because of
a
typographical
error,
definitions
of
this and
the
following
character were confused in
Doyle
and Dono-
ghue (1992),
and this was
perpetuated
by
Rothwell and
Ser-
bet
(1994), although
without
consequence
for
relationships.
9
(DDZ 7).
Laminar vein
orders
(0) one, (1)
two or
more. Nixon
et al.
(1994)
used
a multistate character that
counts the midrib of
pinnately
veined
leaves as the first-order
vein. However,
this
equates
the condition in
Welwitschia,
with
parallel primaries
and finer
cross-veins,
with
that in
cycad-
and fernlike
leaves,
and
it
fails to
capture
the simi-
larity
between
Welwitschia and Gnetum in
having
the blade
covered
by
veins of more than one order.
In this
data set,
adnate-axillary stipules (DDZ 63) are
uni-
formly present only
in
Piperales,
chloranthoid teeth
(DDZ
64) only
in Chloranthaceae. Nixon et al.
(1994)
scored
Chloranthus, Magnoli Ia, Nymphaea,
and Platanus as
having
stipules,
but
the
stipules
of
each
of
these
groups
are
mor-
phologically
different.
VEGETATIVE
ANATOMY,
CHEMISTRY
10
(DDZ 8). Guard
cell poles (0) raised, (1) level with
aperture.
11
(DDZ 9, modified).
Stomata (0) anomocytic (hap-
locheilic), (1)
some
or
all
paracytic
(syndetocheilic).
Dono-
ghue
and
Doyle (1989; Doyle
et al.
1994) recognized
vari-
able
and
laterocytic
as a third state
(Upchurch 1984).
This
distinction
may
be
valuable
in
the context
of
primitive
an-
giosperms,
but
it obscures the fact that Chloranthaceae and
other
variable
taxa are
capable
of
producing paracytic
sto-
mata. This
ability
is
a major
departure
from the basic
con-
dition
in
seed
plants
as a
whole,
and
splitting
it
into
two
states
may
obscure
patterns
at the
present
broader level of
analysis. Furthermore,
laterocytic
and variable are
question-
ably equivalent. Tetracytic
in
Piperales
is
scored as un-
known, being
an
autapomorphy.
Donoghue
and
Doyle
(1989)
scored core
Laurales as
unknown,
but with
paracytic
and variable combined
they
can be scored
as
paracytic.
Nix-
on et al.
(1994)
scored all
angiosperms
as
syndetocheilic,
but
many
of the
groups
included
have anomocytic stomata,
which
correspond
to
haplocheilic
for the
present purposes
(Metcalfe
and Chalk
1957,
1979; Payne 1979; Cronquist
1981).
12
(NCSF 23).
Leaf
traces (0) mesarch, (1) endarch. As
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S28
INTERNATIONAL
JOURNAL
OF
PLANT
SCIENCES
noted
by Nixon
et al. (1994),
this is not
redundant
with
mat-
uration
of the
stem primary
xylem.
The mesarch
condition
in glossopterids
and
Pentoxylon
is confirmed
by
Pigg
(1990b)
and
Bose
et al. (1985).
Corystosperms
are rescored
as endarch, based
on
observations
of
Meyer-Berthaud
et
al.
(1993)
on
Kykloxylon.
Pigg (1990a)
described
traces
in cor-
ystosperm
leaves
(Dicroidium)
as
exarch
to
marginally
mes-
arch;
however,
this refers to
the lateral veins
and does not
contradict
Meyer-Berthaud
et
al.
(1993),
based
on
main
leaf
traces
in
better-preserved
material
(K.
B. Pigg, personal
communication).
13
(NCSF
33).
Foliar astrosclereids
(0) absent,
(1) pres-
ent.
Called
branched sclereids
in
Donoghue
and
Doyle
(1989);
a potential
synapomorphy
of
Welwitschia,
Gnetum,
and
some angiosperms
(Nixon
et
al.
1994).
Core Magnoli-
ales are
scored
(0/1),
since astrosclereids
are lacking
in
De-
generia
but present
in other
groups
(Metcalfe
1987;
Doyle
and
Le Thomas 1994).
They
occur
in
Nymphaeaceae
but
not
Cabombaceae (Cronquist
1981),
so
Nymphaeales
are scored
as
(0/1).
14
(DDZ
10). Apical
meristem (0)
without tunica,
(1)
with
tunica.
Nixon
et
al. (1994)
scored
Araucariaceae
as
(0),
but
Sporne (1965)
and
Gifford
and Foster
(1989)
indicate
that they have
a
tunica.
15
(RS
14). Cauline
protoxylem
(0)
one central
strand,
(1)
two
or more sympodia.
The derived
state characterizes
all
seed plants
above
Elkinsia
(Rothwell
and Serbet 1994).
16
(DDZ
11).
Stele
(0) protostele
or arcuate
primary
xy-
lem segments,
(1)
eustele of
more
or less round bundles.
This circumvents
deciding
whether
Medullosa
has a true
eustele and
thus
scoring
the
group
as
heterogeneous
(Quaes-
tora
is
protostelic,
Medullosa
has
arcuate
segments).
With
medullosans based
on
both
genera,
the related
character
of
internal
secondary
xylem ("polystele"
state
of
the corre-
sponding
character
in
Nixon
et al.
1994)
should be elimi-
nated, since
it
is polymorphic
in
medullosans
and corysto-
sperms
(Rhexoxylon
but
not
Kykloxylon;
Meyer-Berthaud
et
al.
1993),
leaving
only
Pentoxylon
monomorphic.
Nixon
et
al.
(1994)
scored
Cycadaceae
as
polystelic,
but even
ac-
cepting
this it would be another
polymorphism
when cycads
are treated
as a
single
taxon.
17
(DDZ
13, modified).
Nodes with
(0) one
trace to
each
leaf,
(1) more
than
three traces,
(2)
two traces
from adjacent
bundles,
(3)
three
traces.
Following
Rothwell and
Serbet
(1994),
I
have redefined
this character
in
terms
of
leaf
traces
rather than leaf gaps,
but I
have
maintained
the
distinction
among two,
three,
and
more
traces,
which they
eliminated.
Doyle
and Donoghue
(1992)
treated
the medullosan
condi-
tion,
with many
traces derived
from one solid
mass
or
sev-
eral arcs of
primary
xylem,
as
a
separate
state. Nixon et
al.
(1994)
apparently
misread
Doyle
and
Donoghue
as attrib-
uting
this state
to cycads.
Rothwell
and Serbet
(1994)
also
recognized
the distinctness
of Quaestora
and
Medullosa
by
assigning
them
a new state
(leaf
traces from
two
or
more
protoxylem
strands
or
bundles
over
a
length of
stem).
How-
ever,
with
medullosans
combined
as one
taxon,
this state
becomes
uninformative,
so I
have
scored medullosans
as
un-
known,
as
in
Doyle
et
al. (1994).
Rothwell and
Serbet
(1994)
scored Bennettitales as
unknown,
but I
follow
Nixon et al.
(1994)
in
treating
them
as
(0),
based on Williamsonia
and
Cycadeoidea.
Nixon
et
al.
scored
glossopterids
as
two-trace,
but
because
Pigg
and
Taylor
(1993)
could
not
determine
whether the two
leaf
traces
that
they
observed came
from
one
or two
stem
bundles,
I
have scored them as (0/2). Nixon
et al.
(1994)
scored
Nymphaeales
as
having
many traces,
but
Nymphaeaceae
are three-trace (Weidlich
1976),
and Cabom-
baceae
are difficult to
interpret
but not many-trace
(Moseley
et al. 1984).
I
have
rescored
eudicots
as
(1/3),
since
the
num-
ber
of
traces
in taxa considered
basal
varies
between
three
(Nelumbo,
some ranunculids)
and many
(other
ranunculids,
Platanus)
(Cronquist
1981).
18 (DDZ
12).
Primary
xylem
(0)
mesarch, (1)
endarch.
Nixon et
al.
(1994)
accepted corystosperms
and
Pentoxylon
as
mesarch, glossopterids
as
endarch,
but
Rothwell
and Ser-
bet
(1994)
scored
Pentoxylon
as mesarch,
corystosperms
and
glossopterids
as
(?).
In
Pentoxylon,
Bose et al. (1985)
men-
tioned
protoxylem
only
at the ends
of the
tangentially
elon-
gated
bands,
so
I
have scored
this taxon
as
unknown.
Scor-
ing
of
corystosperms
is
based
on
Meyer-Berthaud
et al.
(1993),
who
described
Kykloxylon
and several
species
of
Rhexoxylon
as endarch;
an exception
is
a small
stem
of R.
piatnitzkyi,
said to
be
mesarch.
In
the glossopterids
studied
by
Pigg
and Taylor
(1993),
primary
xylem was
poorly
pre-
served,
and maturation
could
not be characterized.
19 (DDZ
14).
Metaxylem
(0)
with scalariform
pitting,
(1)
without
scalariform
pitting.
Nixon
et al.
(1994)
scored
cordaites
as
(1),
but the
group
is known for
extensive
sca-
lariform pitting
at
the
ill-defined
transition
from the
primary
to
the
secondary
xylem
(Stewart
and
Rothwell
1993;
Roth-
well
and Serbet
1994).
They
also scored
Pentoxylon
as
(1),
but
Bose
et al.
(1985)
reported
scalariform
pitting
in
short
shoot bundles
and
leaf traces.
20
(DDZ
15).
Secondary
xylem
tracheids
with
(0)
cir-
cular bordered
pits
or
perforations
only, (1)
at least
some
scalariform
pits
or
perforations.
Nixon
et al.
(1994)
distin-
guished
scalariform
and
mixed
states,
but
no
taxa
in the
pres-
ent data set
are
exclusively
scalariform.
They
scored
Cyca-
daceae
as
(0),
Zamiaceae and Stangeriaceae
as
(1),
so
I
have
rescored
the combined
group as
(0/1).
I
follow
Nixon
et al.
in
rescoring
Pentoxylon
as (0),
based
on Bose
et al.
(1985),
who observed
only
uniseriate
circular
pitting.
Nixon
et al.
(1994)
used an
additional character
contrasting
foraminate
and scalariform
vessel
perforations,
but
this
is correlated
with
tracheid
pitting
in the taxa included
here,
and it seems
reasonable
to combine
kinds of
perforations
and
presumably
homologous
pits
on the
end walls of tracheids.
21
(new
character).
Tertiary spiral
thickenings
in trache-
ids
(0)
absent, (1)
present.
A
feature
of Taxaceae
and Ce-
phalotaxus
(Hart
1987).
22
(DDZ
16).
Vessels (0)
absent,
(1) present.
23
(DDZ
59).
End
wall pits
or
vessel perforations
(0)
multiple,
(1)
simple.
24
(DDZ
17).
Rays
(0)
at least some
multiseriate,
(1)
all
uniseriate
or
biseriate.
Rejected
by
Rothwell
and Serbet
(1994)
because
of
continuous variation,
but
it
appears
that
all
groups
can
be
scored
unambiguously
as
the character
is
defined
here.
Nixon
et
al.
(1994)
added
manoxylic
vs.
pyc-
noxylic
as
a
separate
character,
but
this
is
correlated
with
ray type
in all
groups
in
their
matrix
except
Gnetales
and
angiosperms,
which
they
score
as
multiseriate
and
pycnox-
ylic.
However,
describing
these taxa
as
pycnoxylic,
especial-
ly
angiosperms,
is
debatable;
I
prefer
the
somewhat less sub-
jective
character of
ray
width. Bose et
al.
(1985)
state that
an
earlier
report
of
multiseriate
rays
in
Pentoxylon
was based
on
a disturbed area of
a
branching
bundle.
25
(DDZ
51).
Companion
cells
in
phloem
(0)
absent, (1)
present.
Absence
in
Pentoxylon
is inferred
from Sharma
and
Bohra
(1977).
26
(DDZ
60).
Sieve-tube plastid
inclusions (0)
starch,
(1)
PI
type,
(2)
Pll
type.
Nonangiospermous
groups
based
on
Behnke
(1981),
angiosperms
on Behnke
(1988).
Nixon
et al.
(1994)
scored Pinaceae
as
unknown,
but Behnke
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS OF
GNETALES
S29
(1981) described them as
P
type (similar but not identical
to
PI in angiosperms). Donoghue
and Doyle (1989) scored
core
Laurales as unknown, but
the
basic
state with the
ingroup
arrangement assumed
here is
PI.
27 (DDZ 18, 61,
RS 22, NCSF 17, 18,
modi-
fied). Secretory structures
(0) isolated
cells or
groups
of
cells, (1) cavities, (2)
canals, (3) absent, (4)
oil cells. This
definition follows
Rothwell and Serbet
(1994)
in
distinguish-
ing cells from cavities. Where
Nixon et al. (1994)
and Roth-
well
and Serbet (1994)
conflict (Lyginopteris, Peltaspermum,
glossopterids, Taxaceae,
Gnetum),
I
have followed
the
latter,
who agreed more
with other sources
(e.g., glossopterids:
Pigg
and
Taylor 1993;
Gnetales: Martens 1971).
Rothwell
and
Serbet
(1994)
scored angiosperms (treated
as one
taxon)
as
lacking secretory
structures, but Nixon et al. (1994)
scored several taxa as having
cavities, based
on
their
own
survey.
These occurrences correspond to presence
of oil
cells
in
both
leaves
and cortex
in
all taxa covered
by
Met-
calfe
(1987), plus Piperaceae,
listed as
having
cavities
by
Metcalfe and Chalk
(1983).
This
suggests
that cavities are
redundant with
oil
cells
in the
leaves,
treated
by
Nixon
et
al. as a
separate
character (NCSF 18). One
solution
might
be
to eliminate character 14 and
lump
oil cells with
cavities.
However,
oil cells differ
in their contents and structure
from
the cavities
of
the one
extant
nonangiosperm group
with cav-
ities (Ginkgo), which contain
mucilage. Hence
I
have
treated
oil
cells as
an additional
state of the
present
character.
I
have
provisionally assumed
that cavities
in
extinct
groups
are
more
like those of
Ginkgo
than
oil
cells,
and that isolated
cells
in Paleozoic
groups
are
not
oil
cells,
but this deserves
closer anatomical
study.
I
have scored
monocots as
(3/4),
since
oil
cells
occur
in
Acorus,
which
may
be basal
(Duvall
et al.
1993;
Davis
1995).
Nixon et al.
(1994)
scored
Nym-
phaeales
as
having
oil
cells,
but Metcalfe
and Chalk
(1957,
1983) and Cronquist (1981)
indicated
that
oil
cells are absent
in this
group;
consistent with this, Nixon et
al. scored them
as
lacking
cortical
secretory
structures.
I
eliminated
presence
or
absence
of
resins (NCSF
19),
which
Nixon
et al.
(1994)
scored as
present
in conifers and
corystosperms,
because of
questions
on its
applicability
to
fossils.
28 (DDZ 19). Lignin
with (0) no Maule reaction,
(1)
Maule reaction.
Many
but not all Podocarpaceae show
a
Maule reaction
(Gibbs
1957);
I
have scored them as
(0/1).
Another
chemical
character of Doyle et al. (1994),
ben-
zylisoquinoline alkaloids
(DDZ 62), is uninformative
in this
data
set.
It
appears
to
be
basic in
core
Laurales but varies
in
core Magnoliales and
basal eudicots (Nelumbo,
ranuncu-
lids,
but not
Platanus).
OVULATE
STRUCTURES
29
(DDZ 20, NCSF 67,
modified). Ovule-bearing
struc-
ture
(0) pinnate (ovules
or
"cupules"
in
two
rows on a
dor-
siventral
structure)
or
pinnate
with a
three-dimensional fer-
tile
portion, (1) simple,
paddle-like (ovules
not
in
two
definite
rows), (2) simple,
stalklike,
with one
ovule,
or ovule
sessile. See text for discussion and documentation
of
antho-
phyte groups.
This definition leaves out
the
fact that
"pad-
dles"
are associated
with a
bract
in
glossopterids
and Piro-
conites,
but this loss of information seems
unavoidable,
because there are
suspicions
that the same condition also
existed
in
several Mesozoic seed fern
groups (Nixon
et al.
1994),
but direct evidence is
lacking. Sessile,
as in modern
conifers
and
Gnetales,
is
combined
with stalked because the
ovules are
clearly
borne on a
shoot
potentially
homologous
with the fertile short shoot
of Emporia and cordaites. Cor-
daites are scored
as unknown because the
megasporophylls
vary between
one
and
more ovules
(Florin 1951;
Rothwell
and Serbet 1994), and
in
the latter case
they
are not
clearly
comparable
with either
pinnate
or
paddle-like.
Nixon
et al.
(1994) showed that the
ovulate structures
of
corystosperms,
previously interpreted as
bipinnate sporophylls
with cupules
corresponding to pinnules,
are actually
branch
systems,
with
the "pinnae" arranged
spirally. However,
it is not so clear
whether the pinnae are
in turn branches, with the
cupules
representing paddle-like
megasporophylls (coded
by Nixon
et al. as "peltate-enclosed"),
analogous
to the
simple
micro-
sporophylls,
or
sporophylls
with
the cupules representing
leaflets,
so I
have scored
the group (0/1). Chloranthaceae
and core
Laurales are
scored as unknown to
allow
equiva-
lence of their uniovulate
carpels with the condition
in Gne-
tales.
Nixon et
al. (1994)
scored Caytonia as having
peltate-
enclosed
structures,
but in contrast to other
Mesozoic seed
ferns these do
seem
to
be
part
of a
pinnate
structure,
based
on the dorsiventral character
of the rachis and the uniform
orientation of
cupules
toward one
(adaxial)
side
(Harris
1940, 1964).
The
comparable
character of Rothwell and
Ser-
bet
(1994;
RS
25),
which combines
mega-
and
microsporo-
phylls (as
in
Doyle
and
Donoghue 1992),
has
several
auta-
pomorphic
states
(Cordaixylon,
cycads, Pentoxylon)
and
needs revision
in
the
light of new data on microsporophylls
of
Gnetales.
A character of
Crane
(1985)
and
Doyle
and
Donoghue
(1986, 1992; Doyle
et al.
1994),
number
of
ovules
per
anat-
ropous cupule or potential
homologue,
was eliminated be-
cause
it
tends to associate
groups
scored as
having
one ovule
per cupule (corystosperms,
Bennettitales, angiosperms),
even
though
the
homology
of the
"cupules"
of these taxa
is
highly controversial (Nixon
et al.
1994).
30
(DDZ 21,
modified). Ovule (0)
on a lateral append-
age
or
sessile
but lateral
on
stem, (1)
terminal
on
stem. Ter-
minal
might
be treated as an additional state of the
previous
character,
but this would eliminate stalked or sessile ovules
as a
potential homology
between Gnetales and conifers, or
between
Taxaceae and other conifers
(a concept
accepted by
Nixon et al.
1994).
Eames
(1952) argued
that
Ephedra
dif-
fers
from
Welwitschia and Gnetum
in
having appendicular
rather than
terminal
ovules, and this was accepted
by Roth-
well and
Serbet
(1994).
However,
this view was
based
en-
tirely
on
analogies:
comparison
with the male
flowers
of
Ephedra,
where the microsporangia
had been
considered
ter-
minal
on
a cauline
column,
but
a
series
of
intermediates
in
the genus
shows
that
the
column
is formed
by
fusion of two
appendages;
with
fertile
shoots of
cordaites;
and
with vari-
ation
from
two flowers
(misleadingly
termed
"ovules")
to
one
apparently
terminal
flower
per
strobilus
in
Ephedra.
There
is
no
variation
between
one and two ovules
in
the
female flowers. Eames was
surely
correct
in
concluding
that
Ephedra
ovules were
ultimately
derived from
appendicular,
but he
gave
no
basis
for
contrasting
them with the
ovules
of
Welwitschia
and
Gnetum.
31
(RS 28, modified).
Ovule position
on
supporting
lam-
inar structure (0) apical
or
marginal, (1) abaxial, (2)
adaxial.
See text
for
discussion.
In
groups
with two
integuments,
"ovule"
refers
to
the
nucellus
plus
first
integument,
the
pre-
sumed
homologue
of
the
original
seed
plant
ovule.
Rothwell
and Serbet
(1994)
defined their
character
in terms of
position
on either
a
leaf
or a
stem,
but
I
prefer
to avoid
equating
positions
on
clearly
different
organs.
Thus Rothwell and
Ser-
bet scored taxa with ovules terminal on the axis and in
or-
thotropous cupules
as
terminal,
but I
have scored
them
as
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S30
INTERNATIONAL
JOURNAL
OF
PLANT SCIENCES
unknown. Similarly,
they scored
modern conifers
in terms
of ovule position
on the cone scale,
but because
the scale is
presumably a
shoot and position
on it does not
correspond
to position on
an original sporophyll,
I have
scored them as
unknown.
Rothwell and Serbet
(1994)
scored corystosperms
as adaxial, but
as discussed
in the list of taxa, this
was based
on cupules now
doubted
to be corystosperms (Petriellaea;
Taylor
et
al.
1994).
I have scored corystosperms
(1/2)
to
allow for either
orientation of
the
"cupules."
32 (RS 29).
Ovule orientation
(0) erect, (1)
inverted.
Again,
"ovule" refers to the nucellus
plus
first
integument.
Under this definition, the
inverted
condition,
found in
many
conifers,
cannot be equated
with the
anatropous
condition
in
bitegmic angiosperm
ovules, as
done by Nixon et
al. (1994).
Nixon
et
al. scored Araucariaceae,
Podocarpaceae,
and Tax-
odiaceae-Cupressaceae
as
erect.
However,
Araucariaceae
have
inverted
ovules (Sporne
1965;
Hart
1987);
Podocar-
paceae
are mixed,
but
taxa
with erect ovules
(e.g.,
Phyllo-
cladus,
Microstrobus; Hart 1987)
appear
to be nested within
the group (Kelch,
in press, personal
communication); and
Taxodiaceae
are mixed
(although
more
appear
to
be
erect
than were scored
as
such by
Hart
1987;
based
on Takaso
and
Tomlinson
1990).
33
(DDZ
22, modified). Ovule
(=
nucellus
plus
first in-
tegument) (0)
in radial,
lobed
"cupule," (1)
with no
closely
enclosing
structure, (2)
in
anatropous "cupule"
or outer
in-
tegument, (3)
in
orthotropous,
unlobed "cupule"
or outer
integument, (4)
in
bipartite
outer
integument
derived
from
two
primordia.
See text
for
discussion.
Following
Rothwell
and Serbet
(1994),
I
tentatively
assume
that the
Pentoxylon
sarcotesta is
not
a
cupule,
contrary
to
Crane
(1985)
and
Doyle and
Donoghue (1986,
1992; Doyle et
al.
1994).
34
(DDZ
54).
Closed
carpel
with
stigmatic
pollen ger-
mination
(0)
absent, (1) present.
This avoids
the
question
of
carpel
homology (treated
in
character 29),
while
distinguish-
ing angiospermy
from
less
complete
enclosure in other
groups.
35
(NCSF
98, modified).
Carpels (0) spiral
or
irregular,
(1)
whorled. Nixon et
al.
(1994)
scored
Nymphaea
as
spiral,
but
although
Nymphaeaceae
have
many carpels,
these
are
whorled
(Cronquist
1981).
They
scored Winteraceae as
whorled,
but
carpels
in this
group
are
spiral or,
more pre-
cisely, irregular
(Endress
1986).
Nixon et al.
(1994)
also
included characters
describing
carpel
number
and indeter-
minate vs. determinate
carpel
number,
but these
appear
to
interact
too
complexly
with
carpel arrangement
and each
other and to vary too
much
within
taxa. One
carpel
is
most
distinctive,
but
in
the
present
data set this characterizes
only
one taxon
(Chloranthaceae,
scored
as unknown for
the
pres-
ent
character).
36
(DDZ
76).
Ovules
per
carpel
(0) several,
(1)
one
api-
cal.
A
third state
in
Donoghue
and
Doyle (1989),
one
basal,
is
represented
only
in
Piperaceae
within
Piperales
and
My-
risticaceae
within
core
Magnoliales.
Since this state
is un-
informative
and the
other
members have several
ovules,
I
have scored
Piperales
and
core
Magnoliales
as
having
sev-
eral ovules. Nixon
et
al.
(1994)
extended a similar character
(NCSF 68)
to
nonangiospermous
groups,
but as
they
ac-
knowledged
this character is
unlikely
to
be
homologous
across seed
plants.
Characters 36 and 37 are
theoretically subject
to the
Mad-
dison effect
(Maddison 1994),
but
not if
angiosperms
are
monophyletic.
Fruit
type
(used
in
Doyle
et al.
1994)
is uninformative
in
this data set.
Only
core Laurales could be scored as having
drupes,
Aristolochiaceae as
dehiscent;
in
Piperales,
Pipera-
ceae have
drupes, but
Saururaceae
have capsules;
and in
Nymphaeales,
fruits are dry-indehiscent
in
Cabombaceae,
but berry-like
in
Nymphaeaceae
(Cronquist
1981).
MICROSPORANGIATE
STRUCTURES
37 (DDZ
20, 52,
modified).
Microsporophylls
(0) pin-
nate or
paddle-like,
(1) simple,
one-veined,
scalelike, (2)
simple, one-
(rarely
three-) veined,
with two
pairs
of longi-
tudinal
microsporangia.
As discussed
in the text,
microspo-
rophyll
morphology
in taxa scored
(0) is usually
similar
to
that of the megasporophylls.
Nixon
et
al. (1994)
combined
paddle-like
and
simple
in
their character
36 and scored
all
Mesozoic seed ferns
as
simple,
including Caytonia,
but
with-
out new observations
on
Caytonia,
this
seems too bold a
departure
from previous
interpretations
of this
group.
In
any
case,
this
interpretation
would not
affect
scoring
of the
pres-
ent
character.
I
have
scored Chloranthaceae
(0/2)
to allow
for
the
hypothesis
that the
three-lobed androecium
of
Chlor-
anthus is a pinnate
sporophyll (cf.
Nixon
et al.
1994).
38 (DDZ
68,
NCSF 96).
Stamens
(0)
laminar, (1) with
well-differentiated
filament. Nixon
et al.
(1994)
scored
Nym-
phaea
as
laminar,
but
although
this
genus
is
famous
for
in-
tergradation
of
stamens
with
petals,
normal
stamens
in this
and other
Nymphaeales
have a
well-differentiated filament
(Cronquist
1981).
Theoretical Maddison
effect
(Maddison
1994),
but
inconsequential
if
angiosperms
are
monophyletic.
39
(DDZ
25, NCSF
40,
modified). Microsporangia
(0)
terminal,
(1) abaxial,
(2) adaxial,
(3)
lateral.
Addition
of
a
lateral state
for
most
angiosperms
follows Nixon et
al.
(1994)
and lessens the problems
discussed
by
Doyle et
al.
(1994)
in
comparing
states
in
angiosperms
and other
groups.
However, angiosperms
with
strongly
abaxial
or
adaxial
spo-
rangia (Cronquist
1981; Endress
and Hufford
1989) are
scored
as such.
Nixon
et al.
(1994)
scored
Lyginopteris
as
unknown, because
of variation between terminal
and abaxial
in
lyginopterids,
but
I
follow Rothwell and
Serbet (1994)
in
scoring
it
as
terminal based on their
more
closely
circum-
scribed
concept.
Following
Rothwell
and Serbet
(1994)
and
Nixon et al.
(1994),
I
have rescored
Caytonia
as
abaxial
rather than
unknown.
Nixon
et al. (1994)
scored Ginkgo
as
abaxial,
but
I
have scored
Ginkgoales
(0/1),
because some
fossil
groups
are terminal. Rothwell and Serbet
(1994)
scored Taxaceae
as
terminal, but
the position
of
the sporan-
gia
on the
peltate sporophylls
can
be
interpreted
as abaxial,
as done
by
Nixon
et al.
(1994).
Nixon et
al.
(1994)
scored
Eupomatia
and
Magnolia
as
adaxial,
but
I
have scored
Eu-
pomatia
as
marginal,
since it varies between
marginal
and
slightly
adaxial,
and
Magnoliales
as
abaxial,
which is
the
state in
Liriodendron and magnolialean
families
other
than
Magnoliaceae
(seen
in
extreme
form in
Degeneria).
40
(new
character).
Microsporangia
per
sporophyll (0)
more than
two, (1)
two.
Conifers
as a whole
vary
consid-
erably (Hart
1987),
but
Pinaceae and
Podocarpaceae
consis-
tently
have
two
microsporangia,
which is unusual relative
to
seed
plants
as a whole. Nixon et
al.
(1994;
NCSF 39)
dis-
tinguished
many
from one to four
(Ginkgo,
Pinaceae,
Po-
docarpaceae,
Gnetales, angiosperms),
but this
equates
much
more
diverse structures. The
only
ambiguities
concern Gink-
goales,
which
vary
from two
in
Ginkgo
to more
in
fossils,
and
Gnetum,
which seems too
highly
modified
to
be inter-
preted.
41
(DDZ
26).
Microsporangia
(0) free, (1)
fused at least
basally.
Rothwell and Serbet
(1994)
scored
Caytonia
as
free,
but this contradicts
primary
reports
(Harris
1940, 1964).
Fol-
lowing
Rothwell and Serbet
(1994)
and
Nixon
et al. (1994),
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS
OF GNETALES
S31
I have rescored Pentoxylon as
free rather than unknown.
Cy-
cads are scored (0/1),
since
sporangia
are
free in
Cycas
but
fused
in
other groups (Nixon et al. 1994).
42 (NCSF 42). Microsporangial
dehiscence
(0)
ectoki-
netic, (1) endokinetic, (2) endothecial.
Data and concept
from Nixon
et
al. (1994).
If
endothecial
is a subset
of
en-
dokinetic, as stated by Nixon
et
al. (1994),
this would
tend
to associate angiosperms
with
Caytonia.
43 (DDZ 27, modified;
NCSF
38).
Microsporophylls (0)
free, (1) basally
fused.
A
partly
overlapping distinction
be-
tween
spiral
and whorled was used instead by Doyle
and
Donoghue (1986, 1992) and Rothwell
and Serbet (1994),
and as an additional character
by
Nixon
et al.
(1994;
NCSF
37). However, Gnetales
are
not whorled if
they
have
pairs
of branched
microsporophylls
rather than whorls
of
simple
ones,
and
their
paired
condition
could be a
consequence
of
their
general opposite phyllotaxy.
Nixon
et al. (1994)
scored
Bennettitales as fused,
but
microsporophylls
are
free in
Leg-
uminanthus (Crane 1988), so
I
have
treated Bennettitales as
uncertain.
Following
Bose
et al.
(1985)
and Nixon et al.
(1994),
I
have
scored Pentoxylon
as free rather
than un-
known. Nixon
et al. scored
Chloranthus
as
fused,
but
be-
cause it is
questionably comparable
to the
gnetalean
condi-
tion and the stamens in other
genera
are
free (Endress 1987),
I
have scored Chloranthaceae
as unknown.
44
(DDZ 67). Stamen
number
(0)
various, (1) multiples
of three. Nixon
et
al.
(1994)
scored Chloranthus
as trimerous
(NCSF 95),
but because of its
marked floral
asymmetry
it
is
questionably comparable
to other trimerous
forms;
in
any
case,
Chloranthaceae as
a whole are not trimerous.
45 (new character). Inner staminodes
(0) absent, (1)
present.
Used
by Donoghue
and Doyle (1989)
but not
by
Doyle et
al.
(1994)
because
it
was uninformative
in that
data
set.
Staminodes
are scored as
present
in
core
Magnoliales,
since
Magnoliaceae (no staminodes)
are nested
within the
group, Anaxagorea (with staminodes)
is
apparently
basal
in
Annonaceae
(Doyle
and Le
Thomas 1994),
and the
state
in
Myristicaceae (with
stamens fused
into a
column)
is
unde-
fined.
AGGREGATION,
ASSOCIATED STRUCTURES
46
(DDZ 28,
RS
23,
NCSF
62,
modified). Fertile ap-
pendages (0)
not
aggregated
or
in
simple strobili, (1) simple
male, compound
female
strobili, (2)
compound male and
fe-
male
strobili.
In
Doyle
and
Donoghue
(1986, 1992; Doyle
et al.
1994),
the conifer
state
(1)
was
autapomorphic
and
thus scored as
unknown, but
it is informative with inclusion
of several conifer taxa. Rothwell
and
Serbet (1994) recog-
nized
a
more
complex character, expressing
seven
patterns
of
aggregation,
but four states
are
autapomorphic
or absent
in the
present
data
set,
and the
distinction between
"not
ag-
gregated"
and
"simple
cones"
is
problematic
in
some fossils
scored as not
aggregated (i.e.,
the
situation
in
glossopterids
and
Caytonia
is not
directly
established,
while
peltasperms
and
corystosperms
show
partial aggregation).
Nixon et
al.
(1994) distinguished presence
and absence
of compound
me-
gastrobili
and
thus
equated
the conifer and cordaite condi-
tions,
a
potentially major
loss of
information.
They
scored
corystosperms
as
having compound
strobili,
but even if their
ovulate structures
represent
two orders of
branching, they
are less
condensed than
those of
cordaites or conifers.
They
scored other Mesozoic seed
ferns,
Pentoxylon, Bennettitales,
and
angiosperms
as
unknown,
but
despite
the
homology
questions
involved it seems fair to
say
that
these do not have
compound
strobili of the sort seen
in
conifers or cordaites.
They scored Taxaceae as unknown, but
I have scored these
as having simple strobili and let parsimony
sort out whether
this is secondary. Nixon
et al.
(1994)
used another character
for presence
or absence of
simple megastrobili
(NCSF 61),
but this is largely correlated
with deviations from
pinnate
sporophylls and
is
again questionably
scored
in
many
fossils.
The difficulty in applying this distinction
is
illustrated by
the
fact that Nixon
et
al. scored glossopterids,
Caytonia, pelta-
sperms,
and
corystosperms
as
aggregated,
whereas Rothwell
and Serbet (1994) scored
them as not aggregated.
47
(RS 24, modified). Symmetry
of ovuliferous
shoot
(0) radial, (1) bilateral (dorsiventral).
Radial
includes bira-
dial
(Gnetales).
Bilateral unites
Emporia
and
extant
conifers,
except Taxaceae. Rothwell
and
Serbet
(1994) recognized a
third
state
with
bilateral
penultimate
axes, but this
is an aut-
apomorphy
of cordaites
in
the
present
data
set.
48
(RS 27, modified). Ovuliferous
shoot (0)
with
distinct
appendages, (1)
cone scale
without distinct
appendages.
This
distinguishes
extant conifers
(except
Taxaceae)
from
Em-
poria.
Character 27
of
Rothwell and Serbet (1994)
includes
both the cone scale state
(ovule
on reduced
shoot)
and
the
terminal ovule of
Taxaceae
and Gnetales
(ovule
on
stem,
character
31
here). The "woody
cone" character
of
Nixon
et
al. (1994; NCSF 63),
which unites Taxodiaceae, Pinaceae,
and Araucariaceae,
is
harder
to define.
For
example,
Nixon
et al.
scored cordaites as
having
woody
cones and
Cephal-
otaxus
as
not,
but
it is not clear
why
these are
distinguished.
I have scored Taxodiaceae
(0/1)
because
of the
vestigial ap-
pendages
on the cone scales
of
Cryptomeria,
Taxaceae as
unknown because of
uncertain delimitation of
the
ovulifer-
ous shoot.
49 (new character). Bract and
axillary
female shoot
(0)
free, (1) fused. Character and
data
based primarily
on Hart
(1987; character 101). Following
Hart
(1987),
I
have scored
Cephalotaxus,
which has
highly
reduced
cone
scales,
as un-
known. Scoring of taxa outside conifers
raises questions of
homology;
for
example,
under
one interpretation of glossop-
terids,
but not
others, they
have
a
bract fused to
a fertile
short shoot (see text
for
discussion).
Therefore, only groups
with definite leaflike
sporophylls
are scored, and glossopter-
ids, Caytonia,
and
anthophytes are treated
as unknown.
Nixon et al.
(1994) distinguished
few vs.
many compound
cone units within conifers
(NCSF
64), but
this
seems too
continuous and variable within
taxa to be
applied consis-
tently.
In
their
trees, many
was
a
synapomorphy
of
Pinaceae
and
Araucariaceae,
but Paleozoic
forms such as
Emporia
imply
that
it
is
really plesiomorphic
in conifers.
Another character
used
by
Nixon et al.
(1994)
to resolve
relationships within conifers, cone scale
seed wing (NCSF
77), refers
to
morphologically
different
structures
in
Pina-
ceae
(part
of
the
cone
scale),
Araucariaceae
(bract),
and Wel-
witschia
(perianth
=
bracteoles).
50
(DDZ 65).
Perianth
(0)
more
than
two
whorls,
or
spiral-irregular, (1)
two
whorls, (2)
none. This
overlaps
somewhat
with
characters 35
(presence
of
perianth)
and
94
(perianth indeterminant/spiral
vs.
determinant/cyclic)
of Nix-
on et
al.
(1994).
The
only
conflict on
scoring
concerns
Eu-
pomatia,
which Nixon et al. scored as
spiral,
but which has
only
a
calyptrate calyx and petaloid
inner
staminodes
(En-
dress
1977),
which can be
rejected
as
part
of the
perianth
on
positional grounds. Eupomatia
is the
only taxon
with a
single perianth whorl,
so this state
would
be
uninformative.
51
(DDZ 66). Perianth symmetry
(0) various, (1)
trim-
erous.
52
(DDZ 75). Hypanthium (=
floral cup) (0) absent, (1)
present.
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S32
INTERNATIONAL JOURNAL
OF PLANT
SCIENCES
Characters 51-53
are
theoretically subject
to
the Maddi-
son effect (Maddison 1994),
but
not if
angiosperms
are
monophyletic.
OVULES,
SEEDS
53 (DDZ 29, modified).
Anatomical symmetry
of
ovule
(0) radial (radiospermic), (1)
biradial or bilateral (platy-
spermic).
This
character
takes into account the criticisms
of
the radiospermic-platyspermic
distinction by Rothwell and
Serbet
(1994),
based
on
intergradation
and
use of
inconsis-
tent
criteria,
while
attempting
to retain what seems
to
be a
significant contrast among early
seed plants (e.g., Lyginop-
teris
and Medullosa vs. cordaites
and Callistophyton) and to
extend
it
cautiously
to
more
derived groups.
Rothwell and
Serbet
substituted two characters,
number of
vascular
bun-
dles (RS 40)
and
(external)
ovule symmetry (RS 41). How-
ever, two states
in
the latter
are autapomorphic (bilateral
in
Emporia, asymmetric
in
Pinaceae).
Because external sym-
metry seems influenced by so
many correlative factors,
I
have
defined the
present
character primarily
in
terms
of
anat-
omy.
In their bundle number
character,
Rothwell and Serbet
coded absence of bundles
as a
third state,
but this is
too
likely
to be an incidental
consequence
of
reduction.
Instead,
I
have scored round
seeds
with
no
vascular bundles
or
other
anatomical markers,
and
unvascularized
flat
seeds
in
conifers
and Ephedra, whose shape
may
be
a
function
of
enclosing
structures,
as
unknown.
However,
I have
scored
well-pre-
served
biradial seeds
not
enclosed
on two sides
as
platy-
spermic
without
data
on
bundles
(peltasperms, Caytonia,
glossopterids, Pentoxylon).
Rothwell
and
Serbet scored cor-
ystosperms
as unvascularized and
round,
but
this
is
based
on
the questionably
related
Petriellaea
(Taylor
et al.
1994);
I
have
scored them
as platyspermic
based on
the bifacial
cuticle
in
compression
material (Thomas 1933).
Unlike
Rothwell and Serbet, I have
scored Ginkgoales,
which show
rare intraspecific deviation
from predominantly twofold
symmetry,
as
platyspermic.
Nixon
et
al.
(1994)
scored an-
giosperms
as
platyspermic,
Bennettitales
and Gnetales as
ra-
diospermic. However,
ovules of
Bennettitales
and
angio-
sperms (=
nucellus
plus
inner
integument) are
round
and
lack anatomical markers. According to Martens (1971),
Ephedra has no bundles in the
inner integument; in
Welwit-
schia,
two
go
to the
base
of the
integument
but not into
it;
in
Gnetum,
several go up
to the base
of
the free integument.
These data
support treating
the
genera
as done here and
by
Rothwell
and Serbet
(1994).
I
have
scored
Cephalotaxus
and
Taxaceae as platyspermic,
since
Sporne (1965) reports
two
bundles
in both.
54 (new character). Apex
of
integument (0)
free
lobes,
(1) simple, (2) bifid, (3) straight,
tubular.
This
combines
sev-
eral
characters
describing
mutually
exclusive
ovule
apex
morphologies in Crane (1985),
Doyle
and
Donoghue (1986,
1992; Doyle
et al.
1994), Rothwell
and Serbet
(1994),
and
Nixon
et
al.
(1994) (lobed:
RS
36;
bifid:
NCSF
76;
tubular:
DDZ
32, RS 37, NCSF 70).
Aspects of the
bifid
state
are
discussed
in the
text. Peltaspermum
and
corystosperms were
scored
as
unknown
by Doyle
and
Donoghue (1986, 1992)
and
tubular
by
Nixon
et
al.
(1994)
because
of their
protrud-
ing micropyle; however,
this is
bifid
and curved rather than
straight,
so
they
are
readily
scored
in
terms of the
present
definition. Nixon
et
al.
(1994)
scored Ephedra as both bifid
and tubular but
quoted
several
papers saying
that there is
no
evidence of a bifid
integument
in
gnetopsids.
55 (new character). Integument
(0) free from nucellus,
(1)
fused more
than
halfway
up
from the
base.
In
extant
groups, this corresponds
roughly
to the
distinction
between
endochalazal vs. pachychalazal
ovule development
(Takaso
and Bouman 1986),
used
as
a
character
by
Nixon
et
al.
(1994; NCSF 72). Ovule
development has the disadvantage
of not being applicable
to
fossil
groups.
Based on analogy
with modern
seeds,
it
can
be suspected
that
forms
such
as
medullosans, cordaites,
and
Callistophyton,
with
the
integ-
ument free to
the
base,
were
endochalazal
(Singh
1978).
This
might reverse the picture
inferred from modern groups,
of
pachychalazal as primitive
and endochalazal
as
derived.
For
these
reasons,
I
prefer
to
use a
character
that
can be
scored
in
more taxa. Nixon et al.
(1994) scored all modern
conifers
as
pachychalazal,
but Mapes
and Rothwell
(1984)
described
Emporia as having a free
integument, as in cordaites.
Cham-
berlain
(1935)
and
Harris (1954) stated
that
the
integument
is free in Araucariaceae,
which agrees with the figure
in
Sporne (1965).
In Taxodiaceae, Sporne (1965),
Singh
(1978), and Takaso and
Tomlinson (1989, 1990) indicate
that
the
integument
varies from
free to
fused,
so
I
have scored
the
group as (0/1).
In
Taxaceae,
it is fused
by
chalazal
growth
in
both
Taxus and
Torreya (Sporne 1965;
Singh
1978).
In Mesozoic
groups,
I
have relied on
descriptions
of
ovule
cuticles by Harris
(1954).
Nixon et al.
(1994)
scored
Gnetales as
endochalazal, citing
Takaso and
Bouman
(1986).
However,
the
descriptions
in Takaso (1985)
and
Takaso
and
Bouman
(1986)
indicate
that
this refers to the
outer integu-
ment(s),
not the
inner
one,
which
is
fused
most
of
the
way
to
the nucellus
at
maturity, especially
in
Gnetum
(Martens
1971; Singh 1978).
56 (RS 43). Lagenostome
(0) present, (1)
absent.
This
and the following character
subdivide
the
previous
nucellar
apex character (DDZ
31), following
Rothwell
and Serbet
(1994).
Rothwell
and
Serbet
scored Bennettitales as
un-
known,
but
Harris
(1954)
described several bennettitalean
seeds as
having
a
thick,
cylindrical
nucellar cuticle with
no
prominent apical
extension,
so I
have scored
them
as (1).
They scored corystosperms
as
having a lagenostome,
but this
is presumably based on
Petriellaea (Taylor
et
al.
1994).
57
(RS 42, modified).
Pollen
chamber
(0) hydrasperman
(with column), (1)
nonhydrasperman
or absent. Rothwell
and
Serbet
(1994) recognized
absent
as a
third
state
(Taxaceae,
Bennettitales,
Welwitschia, angiosperms),
but
I
have
lumped
this with
(1)
because
of
intergradation
in
degree
of devel-
opment
of the
pollen
chamber
and
conflicting reports
in the
literature.
According
to
Singh (1978),
a
pollen
chamber
seems to
be
absent
in
Pinaceae
and
Podocarpaceae
(both
of
which
Rothwell and Serbet
scored
as
nonhydrasperman),
Ar-
aucariaceae,
and
most
Taxodiaceae
(except
Athrotaxis),
but
there is a
rudimentary
chamber
in
Cephalotaxus
and
Taxus
(which
Rothwell
and
Serbet
scored as
absent).
In
Welwit-
schia,
Martens
(1971)
described
a
small
chamber formed
by
breakdown of cells where
the
pollen germinates
and
dis-
agreed
with claims that this
is not a true
chamber.
It seems
premature
to score
Bennettitales,
and loss
of
a
pollen
cham-
ber
in
angiosperms
may
be
a
by-product
of
carpel
closure.
Presence
or
absence
of a column
appears
to be
a less am-
biguous
break in an
otherwise continuous
series. Rothwell
and Serbet
(1994)
scored
peltasperms
and
corystosperms
as
having
a
lagenostome,
but
this
seems
premature
in
the ab-
sence of more
definitely
associated petrified material.
58
(RS 38).
Micropyle (0)
not sealed after
pollination,
(1)
sealed. In
this
new character of
Rothwell and
Serbet
(1994),
the derived
state
appears
to
depend
on loss
of
the
hydrasperman pollen
chamber but
is not redundant
with
it
(Callistophyton
and
cordaites
are
sealed,
but
not
medullo-
sans
and
Emporia).
However,
another
new
character of Roth-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE-PHYLOGENETIC
RELATIONSHIPS
OF
GNETALES
S33
well and
Serbet
(1994),
sealing of
the
pollen
chamber (RS
44;
0
= sealed
by
central column,
1 = at apex,
2 =
not
sealed)
is less defensible
as an independent
character.
State
(1)
coincides
with
loss
of
the hydrasperman
pollen
chamber,
and state
(2)
occurs in
groups
where
the pollen
chamber
is
disappearing
and
states
are most
ambiguous.
59 (DDZ
30,
modified).
Sarcotesta
(0)
absent
or
unise-
riate,
(1)
multiseriate.
Rothwell
and
Serbet
(1994;
RS 39)
redefined
earlier
sacrotesta characters
to
include
the
unise-
riate
sarcotesta
of such
taxa
as Lyginopteris
and
Caytonia.
However,
this obscures
the
obvious
contrast
between
the
thick
sarcotesta of
medullosans
and cordaites
and its barely
visible
homologue
in Lyginopteris
and
conifers,
and
it means
that all
taxa
except
corystosperms,
Gnetales,
and angio-
sperms
are scored
as
having
a
sarcotesta.
Furthermore,
the
exceptions
are
suspiciously
correlated
with enclosure
of
the
seeds.
Nixon et
al.
(1994)
scored all
Mesozoic seed ferns
as
lacking
a
sarcotesta,
but
the situation
in
peltasperms
and
cor-
ystosperms
is
not established.
They
scored
Podocarpaceae
and Magnolia
as having
a
sarcotesta,
but
in
these groups
the
fleshy layer
is
not part
of the
first
integument,
but
rather
the
cone
scale
and the
outer
integument,
respectively.
I have
therefore scored
these
taxa as
(0).
60
(DDZ
33,
modified;
RS
45).
Nucellus
(0)
not
vas-
cularized,
(1)
vascularized
at least
at base.
Following
Roth-
well and
Serbet
(1994),
I
have redefined
this character
to
include
forms with a basal
pad
of
vascular
tissue,
based es-
pecially
on
intergradation
between
a
pad
and
a
more exten-
sive medullosan-like
system
in cordaites.
Data
are
primarily
from Rothwell and
Serbet
(1994),
except
for
corystosperms
(where
Rothwell and Serbet's
scoring
was
presumably
based
on
Petriellaea).
Their
treatment
of
Gnetales agrees
with
Martens
(1971)
on
presence
of
a
"vascular
cupule"
in
Ephedra
and Welwitschia
but not
in Gnetum. Rothwell
and
Serbet
(1994)
scored
angiosperms
as unknown,
but
although
a
survey
of
taxa
in
the data
set would be desirable,
I
have
provisionally
scored
them (0),
based
on reports
that the
ovule trace
usually
ends
abruptly
at the base of
the
chalaza,
and vasculature
in
the
nucellus is
rare and not connected
with
the
funicular strand
(Maheshwari
1950; Gifford
and
Foster
1989).
61
(DDZ
34).
Nucellar
cuticle (0)
thin, (1)
thick.
See
text
for discussion.
Data
from
Harris
(1962)
and Crane
(1985)
for Pentoxylon;
Pant
and Nautiyal
(1960,
1984)
and
Crane
(1985)
for
glossopterids;
Martens (1971)
and
Hill and
Crane
(1982)
for
Gnetales;
and
Harris
(1954)
for other taxa.
Although
other Bennettitales described
by
Harris
(1954)
have a
thick
nucellar cuticle,
Vardekloeftia
does not;
since
this
genus
may
be
basal
(Crane
1988),
I
have
scored
the
group
(0/1).
62
(DDZ
78).
Testa
(0)
multiplicative,
(1)
nonmultipli-
cative.
63
(DDZ
79).
Exotesta
(0)
normal, (1)
palisade.
64
(new
character).
Ruminations
in
the
seed coat
(0)
ab-
sent, (1)
present.
Used
in
Donoghue
and
Doyle
(1989)
but
not in
Doyle
et
al. (1994),
where it
was
uninformative.
Characters
62-64 are scored as
unknown outside angio-
sperms,
since
they
concern
the outer
integument,
whose ho-
mologies
are
problematic.
They
are
theoretically
subject
to
the Maddison effect
(Maddison
1994),
but
not if angio-
sperms
are
monophyletic.
65
(DDZ
41).
Megaspore
tetrad
(0)
tetrahedral, (1)
lin-
ear.
Data
on
groups
between
medullosans and extant
taxa
based
on studies on macerated material
could have
an im-
portant
effect. Presence
of a
linear
tetrad
in
Bennettitales
is
based
on
Crepet
and
Delevoryas (1972).
66 (DDZ
42,
modified).
Megaspore
wall
(0) thick,
(1)
thin
or
absent.
Crane
(1985)
and
Doyle and
Donoghue
(1986,
1992)
treated
this
as
a binary
character,
but
Doyle
et
al. (1994)
split
it into
an
ordered
three-state
character
(thick,
thin,
absent),
since
Gnetales
differ
from angiosperms
in
hav-
ing
a
thin
sporopollenin
wall
rather than none
at all
(Martens
1971).
However,
the
distinction
between
thin and
absent
is
hard
to make
in
fossils,
including
those
taxa most
critical
for
relationships.
In
Caytonia,
Harris
(1958)
showed that
the
layer
that
he
had earlier
identified
as the megaspore
wall
was part
of
a cellular
"aleurone
layer"
and
could
find
no
megaspore
wall
in
the
best-preserved
seeds,
where
he argued
a megaspore
wall would
have
been
seen
if present.
This
might be
a synapomorphy
of
Caytonia
and angiosperms.
However,
Harris
also failed to
find
a
definite megaspore
wall
in Bennettitales
and Pentoxylon,
but it is
less clear
whether
this
absence
is
real
or due to
imperfect
preservation.
Nixon
et
al.
(1994)
scored
all
conifers
as
thick,
but
Rothwell
and
Serbet
(1994)
scored
Taxaceae
as
thin;
since I have
been
unable to
resolve
this
discrepancy,
I have
scored Taxaceae
as
unknown.
POLLEN,
MICROGAMETOPHYTE
67
(new
character).
Microspore
cytokinesis
(0)
simul-
taneous,
(1)
successive.
Data
for conifers
and
Ginkgo
from
Hart (1987),
for Gnetales
from
Martens
(1971),
and
for
an-
giosperms
from Davis (1966)
and Kubitzki
et
al. (1993).
Descriptions
of
cycads
are somewhat
variable but
appear
to
be
nearer the simultaneous
extreme
(Rao
1961;
Singh
1978).
Lauraceae
are
successive,
but
I have
scored
core Laurales
(0)
because
Monimiaceae
(assumed
to
be
basal
and
para-
phyletic)
are simultaneous.
In Aristolochiaceae,
Aristolochia
is variable but
Asarum is simultaneous,
so
I
have
scored
the
family
as
(0).
Most monocots
are successive,
but Dioscorea-
ceae are
simultaneous.
68
(DDZ
35,
modified).
Pollen with (0) proximal
tetrad
scar, (1)
distal
sulcus
or round
germinal
area
(leptoma,
ul-
cus), (2)
no
aperture.
Nixon
et
al.
(1994)
split
out
tetrad scar
as a
separate
character,
but
except
in
medullosans,
which
they
scored as
having
both a
tetrad scar and
a
sulcus,
loss
of the scar is correlated
with
appearance
of
a
sulcus.
Fur-
thermore,
medullosans
are
not a clear
exception,
since
they
have two distal
grooves
that
are questionably
homologous
with a sulcus. These
were
apparently
not a
site of
pollen
tube
exit,
since
mature
microgametophytes
are known
inside
the
grain
(Stewart
1951;
Friedman
1993).
Nixon
et al. (1994)
scored
cordaites
as
having
both a tetrad
scar
and a
sulcus,
but this refers
to different
members of the
group.
Millay
and
Taylor (1976)
described Early
Pennsylvanian
cordaite
pollen
as trilete to
monolete,
and Middle
Pennsylvanian
pollen
as
lacking
a
proximal
suture,
but
in the
latter the
distal area of
attachment between
the
nexine and the sac,
which should
correspond
to
a
sulcus,
is
generally
irregular
in
shape.
Con-
sistent
with these
observations,
Rothwell
and
Serbet
(1994)
scored one of
the
cordaites
in their
analysis,
Mesoxylon,
as
having
a
proximal
tetrad scar,
and
the
other, Cordaixylon,
as
unknown.
Given
the
morphological
uncertainties
and the
stratigraphic
distribution of
the two
forms,
I have scored
the
combined
group
as
(0).
Apertures
in
conifers
were docu-
mented
by
Kurmann
and Zavada
(1994).
Donoghue
and
Doyle (1989)
assigned
the sulculate
pollen
of
Eupomatia
to
a
state that included sulcate,
inaperturate,
and
sulculate;
with
inaperturate
treated
as
a
separate
state,
I
have
scored
Eu-
pomatia
as
unknown to allow derivation
from
either
sulcate
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S34
INTERNATIONAL
JOURNAL
OF PLANT
SCIENCES
or inaperturate.
Tricolpate
pollen (eudicots)
is
autapomorph-
ic and thus
scored as
unknown.
Nixon
et al.
(1994)
treated leptoma
(poorly delimited
ap-
erture) as
a
separate
character
of
conifers
and cordaites
(NCSF 46),
in addition
to
presence
of
a
sulcus,
but
this has
the unwarranted
effect
of implying
that the
leptoma
is a
de-
rived type
of sulcus.
In
fact,
the distinction between
a
lep-
toma and
the sulcus of
saccate seed
ferns is
minimal.
69
(DDZ
36,
modified).
Pollen symmetry (0) radial,
(1)
bilateral, (2)
global. Doyle
and Donoghue
(1992)
noted that
global is
not equivalent
to radial; in
their data set
it was
limited
to Gnetum
and thus scored as unknown, but
here
it
occurs
in three taxa.
Adding global
symmetry
might
be
questioned
because the
global taxa
are all inaperturate,
but
the reverse
is not so,
since Ephedra
is inaperturate
but
strongly
bilateral (with
respect
to
the
polar axis).
Doyle
et
al.
(1994)
scored angiosperms
with radial
symmetry
as
un-
known because
radial was correlated
with
globose
shape
(Winteraceae)
and
three
colpi (eudicots),
but
the
globose
and
tricolpate
states have
been removed
from
this data
set,
so
radial symmetry
is no longer
redundant.
Nixon et al.
(1994)
scored cordaites
as bilateral
and Pinaceae and
Podocarpaceae
as unknown.
The older cordaite pollen
of
Millay
and Taylor
(1976) varies
between
trilete (radial)
and
monolete (bilater-
al);
the
younger
grains
without a tetrad
scar are
elliptical
but
hardly as
strongly bilateral
as
typical
bisaccate pollen.
Of
their
two
cordaites,
Rothwell and Serbet
(1994)
scored Me-
soxylon
as
unknown,
Cordaixylon
as bilateral.
In view of
this
marginal
situation,
I have scored cordaites
as unknown.
Some Podocarpaceae
have radial
trisaccate pollen,
but these
are
nested
within the
group
based
on both
morphological
and molecular data
(Kelch,
in
press,
personal
communica-
tion).
Pinaceae are strongly bilateral,
except
for Tsuga,
which has
unique operculate
pollen
and
is
probably
nested
within the
group (Hart
1987;
Chase
et al.
1993).
I
have
scored both families
as
bilateral.
Boat-shaped
vs.
globose
pollen (DDZ
69)
was
eliminated
because
it is unclear whether
shapes
of saccate and
sporelike
pollen
can be
compared
with
those
of
nonsaccate,
monosul-
cate
groups.
Outside
angiosperms,
the distinction
seems
too
correlated with
presence
or absence
of
sacs,
while
in
angio-
sperms
it is often
correlated
with size.
Pollen size
(DDZ
70)
was eliminated
because
comparisons
between groups
with
saccate
and
nonsaccate
pollen
are
problematic
(Doyle
et
al.
1994),
and deviations
from medium-sized are
rare and
often
correlated
with other
changes (e.g.,
loss of sacs
and
aper-
tures).
70
(DDZ
37).
Pollen
(0)
nonsaccate
or
subsaccate, (1)
saccate. Nixon
et al.
(1994)
scored
Pinaceae and Podocar-
paceae
as
polymorphic,
making
sacs
strictly
a cordaite
and
seed
fern feature.
However,
nonsaccate taxa
(Larix,
Pseu-
dotsuga)
appear
to be
nested within Pinaceae
(Hart
1987;
Chase et
al.
1993).
The one exception
in
Podocarpaceae,
Saxegothaea,
is
basal
according
to
an
analysis
of
morpho-
logical
data but not a molecular
analysis (Kelch,
in
press,
personal
communication),
and
fossil stem
relatives
of the
family
are saccate.
Doyle
et al.
(1994)
and Nixon
et al.
(1994) recognized
an
additional character
(DDZ
38, NCSF
49) for
eusaccate (al-
veolae detached from
nexine)
vs.
protosaccate
or
quasi-sac-
cate
(alveolae
continuous
from tectum to
nexine).
However,
Osborn
and
Taylor (1994)
have shown
that some
types
thought
to be
protosaccate
based
on
studies
of
compression
material
(notably corystosperms
and
Caytonia) are
in
fact
eusacccate.
In
the
present
data
set,
only Emporia
is
assuredly
protosaccate,
so the character is uninformative.
71 (DDZ
39, modified).
Infratectal
structure
(0) massive
or spongy
alveolar, (1)
honeycomb
alveolar, (2)
granular, (3)
columellar.
Unordered
for reasons
discussed in
Doyle
et al.
(1994).
I
have
combined
spongy
alveolar
with massive,
seen
in this data set
in
Lyginopteris
(Millay
et al.
1978),
because
both types
coexist in "lyginopterids"
(Stidd
et
al. 1985);
this
is consistent
with
the
scoring
of Lyginopteris by
Rothwell
and Serbet
(1994).
Nixon
et
al. (1994)
scored Cephalotaxus
as alveolar,
but Kurmann (1992)
indicated that it is granular.
They also
scored Pentoxylon
as uncertain,
but TEM
sections
by Osborn
et al.
(1991a)
are
fully
comparable
with those of
Bennettitales.
Doyle
et
al. (1992)
scored Nymphaeales
as
(2/3), based
on reports
of
columellar
structure
in
Cabom-
baceae (Osborn
et al. 1991b)
and granular
structure
in Nym-
phaeaceae
(Walker
1976; Takahashi
1992).
However,
the
columellae of Cabombaceae
are
irregular
and resemble
those
interpreted
as derived
from fused
granules
in
Annonaceae
(Le Thomas
1980).
In
view of
the marginally
columellar
structure
of Cabombaceae
and the
distinctly granular
struc-
ture
of Nymphaeaceae,
I have rescored the
whole
group
as
granular.
72 (DDZ
40). Exine
striations (0)
absent, (1)
present.
As
in
Doyle
and
Donoghue
(1992),
I
have
scored Gnetum
as
unknown,
because
TEM
studies
(Gillespie
and Nowicke
1992, personal
communication;
Kurmann
1992)
have
shown
that
it has a
reduced
sexine with tentlike
spines
that resemble
the striations of
Ephedra
and Welwitschia
in
being
formed
by
raising
of the tectum.
Nixon
et
al.
(1994) split
striations
into
global
(Gnetales)
and
proximal
(glossopterids,
Tatari-
na), but
proximal
is uninformative
in this
data
set,
and
the
difference
may
be a function
of
presence
or absence
of sacs.
73
(DDZ
71).
Tectum
(0)
continuous
or
finely perforate,
(1)
foveolate-reticulate.
Core
Laurales are scored
as
un-
known
because the
tectum is
represented
only
by spines.
Nixon et al.
(1994)
scored
the
striate pollen
of
Ephedra
and
Welwitschia
as "semitectate
or reticulate" for "tectum
form,"
like
many
angiosperms (NCSF
54).
It
is
difficult to
see
any
structural
similarity
in the tectum
of
these
groups.
In
reticulate angiosperms,
the
tectum
is interrupted
by large
lumina
(hence
the
term
semitectate),
and the reticulum
con-
sists of tectal
bridges
between the heads of
columellae,
whereas
in
Ephedra
and
Welwitschia
the
tectum
is contin-
uous,
and the striations are formed
by corrugation
of
the
entire tectum
(Van
Campo
and
Lugardon
1973;
Hesse
1984;
Kedves
1987;
Kurmann
1992;
Osborn
et al.
1993).
In
any
case,
the
presence
of
striations is redundant with
the
previ-
ous character.
Nixon
et al.
(1994)
included
another character
(NCSF 52)
for tectum
absent
vs.
clearly
defined,
with all
groups
except
Gnetales,
Bennettitales,
and
angiosperms
scored
as absent.
They
noted
that some
coniferophytes
have
a
loosely
asso-
ciated
outer
granular
layer
that
they
did
not consider a
tec-
tum.
This
is
true for nonsaccate
conifers,
but
it
is correlated
with the shift from alveolar to
granular
infrastructure,
al-
ready
treated as another
character.
Furthermore,
saccate co-
nifers and seed
ferns
and nonsaccate
groups
with alveolar
structure
(cycads,
Ginkgo)
have
a
solid,
continuous outer
layer
that
does
not
differ
in any obvious
way from
the tectum
of
anthophytes (Kurmann
1992;
Osborn
and
Taylor 1994).
74
(DDZ
73).
Supratectal spinules
(0)
absent, (1) pres-
ent. Gnetum is scored as (0),
since its
spines
are tentlike
structures
formed
by
raising
of the tectum rather than
supra-
tectal
outgrowths (Gillespie
and
Nowicke
1992, personal
communication;
Kurmann
1992).
75
(DDZ
72). Aperture
membrane
(0)
smooth
or
weakly
sculptured,
(1) conspicuously
sculptured.
Relatively unsculp-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE PHYLOGENETIC
RELATIONSHIPS
OF GNETALES
S35
tured apertures in conifers were illustrated by Kurmann and
Zavada (1994).
76 (DDZ 74). Endexine (0) uniformly thick (laminated),
(1) absent, (2) thin (nonlaminated), except under apertures.
Unordered for
reasons
discussed
in
Doyle
et al.
(1994).
Uni-
form thickness of
the
endexine
around
the
grain
is more
readily
established
in
fossils
than
presence
or absence of
laminations but correlated with the latter
in
living taxa and
many fossils (Doyle and
Hotton
1991; Osborn and Taylor
1994). Millay et al. (1978)
showed that
Lyginopteris (Cros-
sotheca)
has a
two-layered exine,
but the inner
layer
is rel-
atively thin
and not
differentially stained;
hence
I
have
scored the
group
as
unknown.
Zavada
(1991)
documented a
typical gymnospermous
endexine in
glossopterids.
77
(DDZ 43, modified). Microgametophyte
with
(0)
five
or more
nuclei, (1)
four
nuclei,
tube nucleus
produced by
the second division
(no
stalk
cell), (2)
four
nuclei,
tube
nu-
cleus
produced by
the first division
(no prothallials), (3)
three nuclei. See text for discussion.
Nixon
et al.
(1994)
scored all conifers as
(0),
but this
ignores
four-nucleate
co-
nifers (Spome 1965; Singh 1978).
Hart
(1987)
scored
Ce-
phalotaxus as having prothallial cells, but
this is contradicted
by Sporne (1965)
and
Singh (1978).
78
(RS 55).
Sterile
cell
(0)
colinear with other micro-
gametophyte cells, (1) ring-shaped.
Based
on
Friedman
and
Gifford
(1988).
Rothwell and Serbet
(1994)
scored
angio-
sperms, Gnetum,
and
Welwitschia
as
(0),
but
the sterile cell
is absent
in
angiosperms and putatively
so in
Gnetum
and
Welwitschia.
The character is
inapplicable
in
Cephalotaxus
and
Taxaceae
because prothallial cells,
around
which the
sterile cell forms the
ring,
are
absent.
I have therefore scored
these taxa as unknown.
79
(DDZ 44). Sperm
transfer
(0) zooidogamous, (1)
si-
phonogamous. As
in
Doyle and Donoghue (1986, 1992),
I
have scored fossils with a tetrad scar and no sulcus as zooi-
dogamous, on
the
assumption that although origin
of a
pol-
len tube may have preceded origin of a sulcus (Doyle 1988;
Friedman
1993), siphonogamy probably
did not
originate
until later. A
benchmark
is
provided by medullosans,
in
which
endosporic gametophytes
and
sperm
are known inside
the
pollen (Stewart 1951;
Friedman
1993).
Nixon et al.
(1994)
added another
character, suspended
vs.
penetrating
pollen tube, but,
as
noted
by
Rothwell and Serbet
(1994),
this is correlated with mode of
sperm
transfer
in
all cases
where both characters are known.
MEGAGAMETOPHYTE, FERTILIZATION,
EMBRYO
80
(DDZ 45). Megagametophyte (0) monosporic, (1)
tetrasporic.
Fossils with known
megaspore
tetrads
(see
char-
acter
55)
must
be
monosporic
and are scored
accordingly.
Piper
and
Lilium were united
by
the
tetrasporic
condition
in
Nixon et al.
(1994),
but the
corresponding
taxa
in
the
present
data set are variable
(in Piperales,
Saururaceae
are mono-
sporic; Cronquist 1981)
or
basically monosporic (monocots).
81
(DDZ 46, 55, modified). Megagametophyte (0) large,
cellular,
with
normal
archegonia; (1) large, apical part
and
egg free-nuclear; (2) eight-nucleate,
central
part free-nuclear,
egg
cellular but no neck cells.
See
text
for discussion. Roth-
well and Serbet
(1994) scored Callistophyton, cordaites, and
glossopterids
as unknown for their
archegonium
character.
However, typical archegonia
are
reported
in
Callistophyton
(Rothwell 1980),
cordaites
(Taylor
and
Taylor 1993),
and
glossopterids (Taylor
and
Taylor 1993; Pigg and
Trivett
1994). Although
some
stages
of
megagametophyte
devel-
opment
and
embryos
are known
in
Bennettitales, archegonia
have apparently
not
been
seen
(Crepet 1974;
Sharma
1974),
contrary to Doyle and
Donoghue (1986, 1992; Doyle
et al.
1994). However, observations
by Sharma (1974)
on mega-
gametophytes of Williamsonia,
which are
large
and appar-
ently cellularized basipetally
rather
than
acropetally,
suggest
that they
are
closer
to the basic seed plant pattern
than either
derived state, so
I
have
scored
them as
(0).
82 (RS 58, NCSF 85,
modified). Megagametophyte
cel-
lularization (0) enclosing
single nuclei, resulting
in
uninu-
cleate cells, (1) enclosing
several nuclei, resulting
in multi-
nucleate-polyploid cells.
See text for discussion.
Fossils
considered alveolar by both
Rothwell
and
Serbet (1994)
and
Nixon
et al.
(1994)
are scored
(0).
Rothwell and Serbet
scored Bennettitales as unknown,
but Nixon et
al.
scored
Cycadeoidea
and Williamsonia
as alveolar
(cf.
Sharma
1974);
I
have
extended this to the whole group.
83
(DDZ 48,
modified). Fusion of (0) only
one
sperm
with a female
gametophyte
nucleus, (1) regular
fusion of
both
sperm.
Established
in
Ephedra by
Friedman
(1990,
1992),
in
Gnetum by Carmichael
and Friedman (1995).
84
(DDZ 56,
modified).
Fertilization
producing
(0) dip-
loid zygote(s)
and
embryo(s),
(1) diploid zygote
and
embryo
plus triploid endosperm
tissue.
The
derived
states
in
char-
acters 83 and
84 constitute classic double
fertilization of the
angiosperm type.
Treating
these
characters
separately
has the
effect of
ordering
the
gnetalean
and
angiosperm
conditions.
This seems
justified
on the
grounds
that
they represent
two
different modifications
of
development; endosperm
forma-
tion
requires
not
only
a second
fertilization,
but also
the
origin
of a new kind of
tissue
(possibly
derived from one of
the two
embryos;
Friedman
1994).
Rothwell and Serbet
(1994)
rejected
a
three-state
analogue
of characters 83 and 84 as
uninformative,
since the
Ephedra
and
angiosperm
states
were
each known
in
only
one
taxon,
but
it
becomes informative
with inclusion of several
angio-
sperms
and
the
new
data
on
Gnetum.
85
(DDZ 81).
Nutritive tissue
in
seed
(0)
endosperm
plus perisperm, (1)
endosperm only. Nonangiosperm
groups
are
scored
as
unknown;
if their
condition,
where the nutritive
tissue is
gametophytic,
was included as a
state,
the shift to
either of the
angiosperm
states would be redundant with
character 84. This character is
theoretically subject
to the
Maddison effect
(Maddison
1994),
but not
if
angiosperms
are
monophyletic.
86
(DDZ 47,
modified). Embryo (0)
derived from sev-
eral free
nuclei, (1)
from a
single
uninucleate cell
by
cellular
divisions.
Replaces
free-nuclear vs. cellular
embryogenesis;
see text for discussion. Multiseriate vs.
uniseriate
suspensor
appears
to be correlated
with this character.
87
(RS 62,
modified). Proembryo (0)
not
tiered, (1)
tiered. Rothwell
and
Serbet (1994) treated
this conifer fea-
ture as a state of a
general
embryogenesis
character
analo-
gous
to 87.
However,
this
obscures
the fact that conifers
resemble cycads and Ginkgo
in
having
a free-nuclear phase
and
are in this
respect plesiomorphic
relative to anthophytes.
Araucariaceae are
marginal
in
having
a more
massive,
mul-
ticellular
embryo
than other
conifers, approaching
cycads
and
Ginkgo,
but the
embryo
is more
organized
into tiers than
it is
in
the
latter taxa
(Sporne 1965; Singh 1978).
I have
therefore scored Araucariaceae as
(1).
88
(new character).
Secondary suspensor (=
embryonal
tubes) (0) present, (1)
absent. A
secondary suspensor
in
var-
ious forms is
universal
in
nonangiosperm groups
(Singh
1978)
but
lacking
in
angiosperms;
the
variations
seem cor-
related
with
other characters
(87, 88).
89
(DDZ 49).
Feeder
in
embryo (0) absent, (1)
present.
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S36
INTERNATIONAL JOURNAL OF PLANT
SCIENCES
90 (RS 64, NCSF 89,
modified). Seeds
shed
(0)
without,
(1) with well-developed
embryo. Rothwell and Serbet
(1994)
defined their
corresponding character
in
terms of presence
or
absence
of
seed
dormancy, Nixon
et
al. (1994)
in
terms
of
post-vs. preshed
embryo maturity. Rothwell and Serbet
(1994) scored fossils
based on presence or absence of a well-
developed embryo
in
the seed; the present definition makes
this the primary
criterion. The only case where the two def-
initions do not coincide
is in cycads, which Rothwell
and
Serbet
scored as
unknown. Here the embryo may germinate
immediately upon
shedding (i.e., there
is
no dormancy),
but
it is developed when the
seed
is
shed (Gifford and Foster
1989). Rothwell and
Serbet (1994), Holt and
Rothwell
(1995), and W.
E. Friedman (personal communication) agree
that the
often-cited claim that fertilization
in
Ginkgo occurs
after the seeds
are
shed
(accepted by
Nixon
et al.
1994)
is
incorrect.
In
some fossils, failure to find
an
embryo could
be
due
to poor
preservation,
but
this is
unlikely
to be the
case for
Carboniferous
seeds that are known in
abundance;
in
scoring these
I
follow
Rothwell
and Serbet (1994).
91 (DDZ
50). Seed germination (0) hypogeal (= cryp-
tocotylar), (1)
epigeal (= phanerocotylar).
I
have
followed
Nixon et al.
(1994)
in
scoring
all
conifers, Eupomatia, and
monocots (as
in
Loconte
and Stevenson
1991)
as
epigeal.
Contrary
to Loconte
and Stevenson (1991), Endress (1983)
reported that
germination of Austrobaileya is epigeal.
Literature
cited
Albert
VA,
A
Backlund,
K
Bremer,
MW
Chase,
JR
Manhart,
BD
Mishler,
KC Nixon
1994 Functional
constraints and
rbcL
evi-
dence for land
plant
phylogeny. Ann Mo
Bot
Gard 81:534-567.
Arber
EAN,
J
Parkin 1907 On the
origin
of
angiosperms.
J
Linn
Soc
Bot 38:29-80.
1908
Studies on the evolution of the
angiosperms:
the
re-
lationship of the angiosperms to the
Gnetales.
Ann Bot
22:489-
515.
Ash
SR
1972 Late Triassic
plants
from the Chinle Formation
in
northeastern Arizona.
Palaeontology 15:598-618.
Bailey IW
1944
The
development
of
vessels
in
angiosperms
and
its
significance
in
morphological research. Am J
Bot 31:421-428.
1953 Evolution of
the tracheary tissue of land
plants.
Am
J Bot
40:4-8.
Behnke HD
1981
Sieve-element characters. Nord J Bot
1:381-400.
1988
Sieve-element
plastids, phloem
protein,
and evolution
of
flowering plants.
III.
Magnoliidae.
Taxon
37:699-732.
Bharathan
G, EA Zimmer
1995 Early branching
events
in
mono-
cotyledons:
partial 18S
ribosomal DNA
sequence
analysis. Pages
81-107 in PJ
Ruddall,
PJ
Cribb,
DF
Cutler,
CJ
Humphries,
eds.
Monocotyledons: systematics
and evolution.
Royal
Botanic Gar-
dens, Kew.
Bose
MN,
PK
Pal,
TM Harris
1985
The
Pentoxylon plant.
Philos
Trans R Soc
Lond Ser B Biol Sci
310:77-108.
Bremer K
1988 The limits
of amino acid sequence
data
in
angio-
sperm
phylogenetic
reconstruction. Evolution
42:795-803.
Carmichael
JS,
WE Friedman
1995
Double
fertilization
in
Gnetum
gnemon: the
relationship
between the cell cycle and
sexual repro-
duction. Plant Cell
7:1975-1988.
Chamberlain
CJ
1935
Gymnosperms: structure and
evolution.
Uni-
versity
of
Chicago Press,
Chicago; reprint,
Dover,
New York. 484
pp.
Chase
MW,
DE
Soltis,
RG
Olmstead,
D
Morgan,
DH
Les,
BD Mish-
ler,
MR
Duvall,
et
al 1993
Phylogenetics
of
seed plants: an anal-
ysis
of
nucleotide
sequences
from the
plastid gene
rbcL. Ann
Mo
Bot Gard
80:526-580.
Crane
PR
1985
Phylogenetic
analysis
of
seed
plants
and the
origin
of
angiosperms.
Ann
Mo
Bot
Gard 72:716-793.
1988 Major clades
and relationships in the
"higher" gym-
nosperms.
Pages
218-272 in CB
Beck,
ed.
Origin
and evolution
of
gymnosperms. Columbia
University Press,
New York.
1996 The fossil
history
of the Gnetales.
Int
J
Plant Sci 157:
S50-S57.
Crane
PR,
EM
Friis,
KR
Pedersen 1995 The
origin and early di-
versification
of
angiosperms.
Nature 374:27-33.
Crane
PR,
S
Lidgard
1990
Angiosperm
radiation and
patterns
of
Cretaceous
palynological
diversity. Pages
377-407 in
PD
Taylor,
GP
Larwood,
eds.
Major
evolutionary
radiations. Clarendon
Press,
Oxford.
Crepet
WL
1974
Investigations
of
North American
cycadeoids:
the
reproductive
biology
of
Cycadeoidea.
Palaeontogr
Abt
B
Palaeo-
phytol 148:144-169.
Crepet
WL,
T
Delevoryas
1972
Investigations of North American
cycadeoids:
early
ovule
ontogeny.
Am
J
Bot
59:209-215.
Cronquist
A
1981
An
integrated
system
of
classification of flow-
ering plants.
Columbia
University Press,
New York. 1262
pp.
Daugherty
LM
1934 Schilderia
adamanica:
a
new fossil wood from
the petrified forests
of
Arizona.
Bot Gaz 96:363-366.
Davis
GL
1966 Systematic
embryology of the angiosperms. Wiley,
New York. 528
pp.
Davis JI 1995
A
phylogenetic
structure of the monocotyledons, as
inferred from chloroplast
DNA restriction site variation,
and
a
comparison
of
measures of
clade
support. Syst
Bot 20:503-527.
Delevoryas T, CP
Person
1975
Mexiglossa varia gen.
et
sp. nov.,
a new
genus
of
glossopteroid
leaves from the Middle Jurassic of
Oaxaca,
Mexico.
Palaeontogr
Abt
B
Palaeophytol
154:114-120.
De Queiroz K, J Gauthier 1990
Phylogeny
as
a central
principle
in
taxonomy: phylogenetic
definitions of taxon names.
Syst
Zool
39:
307-322.
Donoghue MJ,
JA
Doyle 1989
Phylogenetic analysis
of
angio-
sperms
and the
relationships
of
Hamamelidae.
Pages
17-45
in PR
Crane,
S
Blackmore,
eds.
Evolution, systematics,
and fossil his-
tory
of
the
Hamamelidae. Vol 1. Clarendon
Press,
Oxford.
Donoghue MJ, JA Doyle, J
Gauthier, AG Kluge,
T
Rowe 1989 The
importance
of
fossils
in
phylogeny
reconstruction. Annu Rev Ecol
Syst 20:431-460.
Donoghue MJ, RG Olmstead, JF
Smith,
JD
Palmer
1992 Phylo-
genetic relationships of Dipsacales based on
rbcL sequences.
Ann
Mo
Bot
Gard
79:333-345.
Doyle JA 1969 Cretaceous
angiosperm pollen of the Atlantic Coast-
al Plain
and
its
evolutionary
significance.
J Arnold Arbor Harv
Univ
50:1-35.
1978
Origin
of
angiosperms. Annu Rev Ecol Syst 9:365-
392.
1988
Pollen evolution in seed
plants:
a cladistic
perspec-
tive. J
Palynol (G.
Thanikaimoni
Memorial Vol) 23-24:7-18.
1994
Origin
of the
angiosperm
flower:
a
phylogenetic per-
spective.
Plant
Syst
Evol
Suppl
8:7-29.
Doyle JA,
MJ
Donoghue
1986
Seed
plant phylogeny
and the
origin
of
angiosperms:
an
experimental
cladistic
approach.
Bot
Rev
52:
321-431.
1992
Fossils
and
seed
plant
phylogeny reanalyzed.
Britton-
ia
44:89-106.
1993
Phylogenies
and
angiosperm
diversification. Paleo-
biology
19:141-167.
Doyle JA,
MJ
Donoghue, EA Zimmer
1994 Integration
of
morpho-
logical
and ribosomal
RNA data
on
the
origin
of
angiosperms.
Ann
Mo Bot
Gard
81:419-450.
Doyle JA,
LJ
Hickey
1976 Pollen and leaves from
the mid-Creta-
ceous Potomac
Group
and their
bearing
on
early angiosperm
evo-
lution.
Pages 139-206
in
CB
Beck,
ed.
Origin
and
early
evolution
of
angiosperms.
Columbia
University Press,
New York.
Doyle JA,
CL Hotton
1991 Diversification of
early angiosperm pol-
len in a
cladistic context.
Pages
169-195 in
S
Blackmore,
SH
Barnes,
eds. Pollen and
spores:
patterns
of
diversification. Clar-
endon
Press,
Oxford.
Doyle JA,
S
Jardine,
A
Doerenkamp
1982
Afropollis,
a new
genus
of
early angiosperm pollen,
with
notes
on the
Cretaceous
paly-
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE PHYLOGENETIC
RELATIONSHIPS OF GNETALES
S37
nostratigraphy and paleoenvironments of Northern Gondwana.
Bull
Cent
Rech
Explor-Prod
Elf-Aquitaine 6:39-117.
Doyle
JA, A Le Thomas
1994 Cladistic analysis and
pollen
evo-
lution
in Annonaceae. Acta
Bot Gall 141:149-170.
Drinnan
AN, TC Chambers 1985 A reassessment of
Taeniopteris
daintreei from the Victorian
Early
Cretaceous: a member of the
Pentoxylales
and a
significant
Gondwanaland
plant.
Aust J Bot
33:89-100.
Drinnan
AN,
PR
Crane, SB Hoot
1994
Patterns of floral evolution
in
the
early diversification of
non-magnoliid
dicotyledons (eudi-
cots).
Plant Syst Evol Suppl
8:93-122.
Duvall
MR,
MT
Clegg,
MW
Chase,
WD
Clark,
WJ
Kress,
HG
Hills,
LE Eguiarte, et al 1993
Phylogenetic hypotheses
for the mono-
cotyledons
constructed
from
rbcL
sequence data.
Ann
Mo
Bot
Gard
80:607-619.
Eames AJ
1952 Relationships
of the
Ephedrales.
Phytomorphology
2:79-100.
Endress PK
1977 Uber Bliitenbau und Verwandtschaft der
Eupom-
atiaceae und Himantandraceae
(Magnoliales).
Ber Dtsch
Bot
Ges
90:83-103.
1983 Dispersal and distribution in some small archaic relic
angiosperm
families
(Austrobaileyaceae,
Eupomatiaceae,
Himan-
tandraceae,
Idiospermoideae-Calycanthaceae). Sonderbd Natwiss
Ver
Hambg
7:201-217.
1986 Reproductive structures and
phylogenetic
significance
of extant
primitive angiosperms.
Plant
Syst
Evol 152:1-28.
1987 The Chloranthaceae:
reproductive
structures and
phy-
logenetic
position.
Bot Jahrb
Syst
109:153-226.
Endress
PK,
LD
Hufford
1989
The
diversity
of
stamen
structures
and dehiscence
patterns
among Magnoliidae.
Bot J Linn Soc 100:
45-85.
Felsenstein
J 1985 Confidence
limits on
phylogenies:
an
approach
using
the
bootstrap.
Evolution
39:783-791.
Florin R
1931 Untersuchungen
zur
Stammesgeschichte
der Conif-
erales und Cordaitales.
Erster
Teil:
Morphologie
und
Epidermis-
struktur der
Assimilationsorgane
bei der rezenten
Koniferen.
K
Sven
Vetenskapsakad Handl,
ser
3,
10:1-588.
1951 Evolution
in
cordaites and conifers. Acta Horti Ber-
giani
15:285-388.
Foster
AS 1972 Venation patterns in the
leaves of
Ephedra. J Ar-
nold
Arbor Harv Univ
53:364-385.
Friedman WE
1987 Growth and
development
of the male
game-
tophyte
of
Ginkgo
biloba
within
the
ovule
(in
vivo).
Am J Bot
74:1797-1815.
1990 Sexual
reproduction
in
Ephedra
nevadensis
(Ephed-
raceae): further evidence of double fertilization in a
nonflowering
seed
plant.
Am J Bot
77:1582-1598.
1992 Evidence of
a
pre-angiosperm origin
of endosperm:
implications
for the evolution of
flowering plants.
Science
255:
336-339.
1993 The
evolutionary history
of the seed
plant
male
ga-
metophyte.
Trends Ecol & Evol
8:15-20.
1994 The evolution of
embryogeny
in seed
plants
and the
developmental origin
and
early history
of
endosperm.
Am J Bot
81:1468-1486.
Friedman
WE,
EM
Gifford
1988
Division of the
generative
cell
and late
development
in the male
gametophyte
of
Ginkgo
biloba.
Am
J
Bot 75:1434-1442.
Gaussen
H
1946
Les
Gymnospermes,
actuelles
et
fossiles.
Trav
Lab
For
Toulouse Tome
II
Etud
Dendrol,
sec
1,
vol
1,
fasc
3, chap 5,
pp
1-26.
Gibbs
RD
1957 The Maule
reaction, lignin,
and the
relationships
between
woody plants. Pages
269-312
in KV
Thimann,
ed. The
physiology
of forest trees.
Ronald,
New York.
Gifford
EM,
AS Foster
1989
Morphology
and evolution of vascular
plants.
Freeman,
New York.
626
pp.
Gillespie
LJ,
JW Nowicke
1992
Systematic implications
of
pollen
morphology
in
Gnetum. Eighth International
Palynological Con-
gress
(Aix-en-Provence),
Program
Abstract
53.
Goremykin
V,
V
Bobrova,
J
Pahnke,
A
Troitsky,
A
Antonov,
W
Martin
1996
Noncoding
sequences
from the
slowly evolving
chloroplast
inverted
repeat
in
addition
to rbcL data do not
support
gnetalean
affinities of
angiosperms.
Mol Biol
Evol
13:383-396.
Hamby RK, EA Zimmer
1992 Ribosomal RNA as a phylogenetic
tool
in
plant systematics.
Pages
50-91
in
PS
Soltis,
DE
Soltis,
JJ
Doyle, eds. Molecular
systematics of plants. Chapman &
Hall,
New York.
Harris
TM
1940 Caytonia.
Ann
Bot, NS,
4:713-734.
1954 Mesozoic
seed
cuticles.
Sven Bot Tidskr
48:281-291.
1958
The seed of
Caytonia.
Palaeobotanist
7:93-106.
1962 The
occurrence of the fructification Carnoconites
in
New Zealand.
Trans R Soc NZ Geol 1:17-27.
1964
The Yorkshire Jurassic
Flora.
Vol 2.
Caytoniales, Cy-
cadales and
pteridosperms. British Museum (Natural
History),
London. 191 pp.
Hart JA 1987
A
cladistic
analysis
of
conifers: preliminary
results.
J
Arnold Arbor
Harv Univ 68:269-307.
Hasebe M,
M
Ito,
R
Kofuji,
K
Iwatsuki,
K
Ueda 1992a Phyloge-
netic relationships in
Gnetophyta
deduced from
rbcL gene
se-
quences.
Bot
Mag Tokyo 105:385-391.
Hasebe
M,
R
Kofuji,
M
Ito,
M
Kato,
K
Iwatsuki,
K
Ueda 1992b
Phylogeny
of
gymnosperms
inferred from rbcL
gene sequences.
Bot
Mag Tokyo 105:673-679.
Hesse M 1984 Pollenkitt is
lacking
in
Gnetatae: Ephedra
and Wel-
witschia;
further
proof
for its restriction to the
angiosperms.
Plant
Syst
Evol
144:9-16.
Hill
CR,
PR
Crane 1982
Evolutionary
cladistics and the
origin
of
angiosperms. Pages 269-361
in
KA
Joysey,
AE
Friday,
eds. Prob-
lems of
phylogenetic
reconstruction. Academic
Press,
London.
Holt
BF,
GW Rothwell
1995
Phenology
and
germination
biology
of
Ginkgo
biloba.
Am J Bot
Suppl
82:20.
Hufford
L
1996
The
morphology
and evolution of male
reproduc-
tive structures of Gnetales.
Int
J
Plant Sci
157:S95-S112.
Imaichi
R,
M
Kato,
H
Okada 1995
Morphology
of the outer
integ-
ument in three
primitive
angiosperm
families. Can J Bot
73:1242-
1249.
Jefferies RPS
1979
The
origin
of chordates-a
methodological
es-
say. Pages
443-477 in
MR
House,
ed. The
origin
of
major
inver-
tebrate
groups.
Academic
Press, London.
Kato M
1990 Ophioglossaceae:
a
hypothetical archetype
for the
angiosperm carpel.
Bot J Linn Soc
102:303-311.
Kedves M
1987
LM and EM
studies on
pollen grains
of recent
Welwitschia mirabilis
Hook.
and
Ephedra species.
Acta
Bot
Hung
33:81-103.
Kelch DG In
press
Phylogeny of the Podocarpaceae based on mor-
phological
data.
Syst Bot.
Kerp
JHF 1988
Aspects
of Permian
palaeobotany
and
palynology.
X. The West- and Central
European species
of the
genus
Autunia
Krasser
emend. Kerp
(Peltaspermaceae)
and the
form-genus
Rhachiphyllum Kerp
(callipterid foliage).
Rev
Palaeobot
Palynol
54:249-360.
Kimura T, S Sekido 1975 Nilssoniocladus n.
gen.
(Nilssoniaceae,
n. fam.) newly
found from the
early
Lower Cretaceous of
Japan.
Palaeontogr
Abt
B
Palaeophytol
153:111-118.
Kirchner
M
1992
Untersuchungen
an
einigen Gymnospermen
der
frankischen
Rhat-Lias-Grenzschichten.
Palaeontogr
Abt
B
Palaeo-
phytol
224:17-61.
Kubitzki
K,
JG
Rohwer,
V
Bittrich,
eds 1993
The
families and
genera
of
vascular
plants.
Vol
2.
Flowering plants-dicotyledons:
magnoliid,
hamamelid and
caryophyllid
families.
Springer-Verlag,
Berlin. 653
pp.
Kurmann MH 1992 Exine
stratification
in
extant
gymnosperms:
a
review
of
published
transmission electron
micrographs.
Kew
Bull
47:25-39.
Kurmann
MH,
MS Zavada
1994
Pollen
morphological diversity
in
extant and fossil
gymnosperms. Pages
123-137
in MH
Kurmann,
JA
Doyle,
eds.
Ultrastructure
of
fossil
spores
and
pollen. Royal
Botanic
Gardens,
Kew.
Le
Thomas A
1980 Ultrastructural
characters of
the
pollen grains
of African Annonaceae and their
significance
for the
phylogeny
of
primitive angiosperms.
Pollen
Spores
22:267-342.
Loconte
H,
DW
Stevenson
1990 Cladistics of the
Spermatophyta.
Brittonia 42:197-211.
1991 Cladistics of the
Magnoliidae.
Cladistics
7:267-296.
Maddison DR
1991 The
discovery
and
importance
of
multiple
is-
lands of
most-parsimonious trees.
Syst
Zool
40:315-328.
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
S38
INTERNATIONAL JOURNAL OF PLANT SCIENCES
Maddison
WP
1994 Missing data versus
missing
characters in
phy-
logenetic analysis. Syst Biol
42:576-581.
Maddison WP, DR Maddison
1992 MacClade:
analysis
of
phylog-
eny
and
character
evolution. Version
3.
Sinauer, Sunderland,
Mass.
398 pp.
Maheshwari P 1950 An
introduction to the
embryology of angio-
sperms. McGraw-Hill, New York. 453
pp.
Mamay
SH 1976 Paleozoic
origin
of the
cycads.
US Geol
Surv
Prof
Pap 934: 1-48.
Mapes
G, GW Rothwell 1984 Permineralized
ovulate
cones of
Le-
bachia
from
late Palaeozoic limestones of Kansas.
Palaeontology
27:69-94.
Martens
P
1971 Les
Gnetophytes. Encyclopedia
of
plant anatomy,
vol
12. Borntraeger, Stuttgart. 295
pp.
Metcalfe CR
1987
Anatomy
of
the dicotyledons. 2d ed.
Vol
3.
Magnoliales, Illiciales, and Laurales
(sensu
Armen
Takhtajan).
Clarendon
Press,
Oxford. 224
pp.
Metcalfe
CR,
L
Chalk 1957
Anatomy
of the
dicotyledons.
Clar-
endon
Press, Oxford. 1500
pp.
1979 Anatomy
of
the
dicotyledons.
2d
ed. Vol
1.
System-
atic
anatomy
of leaf and
stem,
with
a
brief
history
of
the
subject.
Clarendon
Press,
Oxford. 276
pp.
1983 Anatomy
of
the
dicotyledons. 2d
ed. Vol
2.
Wood
structure and
conclusions of the
general
introduction. Clarendon
Press, Oxford. 297 pp.
Meyen
SV
1984 Basic features of
gymnosperm
systematics
and
phylogeny
as evidenced
by
the
fossil
record. Bot Rev 50:1-112.
Meyer-Berthaud B,
TN
Taylor,
EL
Taylor 1993
Petrified stems bear-
ing Dicroidium leaves from the Triassic
of
Antarctica. Palaeon-
tology
36:337-356.
Millay
MA,
DA
Eggert,
RL
Dennis 1978
Morphology
and
ultra-
structure of four
Pennsylvanian
prepollen types.
Micropaleontol-
ogy
24:305-315.
Millay
MA,
TN
Taylor
1976
Evolutionary
trends
in
fossil
gymno-
sperm
pollen.
Rev
Palaeobot
Palynol 21:65-91.
Moseley
MF
Jr, IJ Mehta, PS
Williamson,
H
Kosakai
1984
Mor-
phological
studies
of
the
Nymphaeaceae (sensu
lato).
XIII.
Con-
tributions
to
the vegetative and floral structure of
Cabomba.
Am
J
Bot
71:902-924.
Muller
J
1981 Fossil pollen
records of extant
angiosperms. Bot Rev
47:1-142.
Nixon
KC,
WL
Crepet,
D
Stevenson,
EM Friis
1994
A
reevaluation
of seed
plant phylogeny.
Ann Mo Bot Gard
81:484-533.
Nixon
KC, JI Davis 1991
Polymorphic taxa,
missing values and
cladistic
analysis.
Cladistics 7:233-241.
Osborn
JM,
TN
Taylor 1994
Comparative
ultrastructure
of
fossil
gymnosperm pollen
and its
phylogenetic implications.
Pages
99-
121 in MH
Kurmann,
JA
Doyle, eds.
Ultrastructure of fossil
spores
and
pollen. Royal Botanic
Gardens, Kew.
Osborn
JM,
TN
Taylor,
PR
Crane
1991a The ultrastructure of
Sahn-
ia
pollen
(Pentoxylales).
Am
J
Bot
78:1560-1569.
Osborn
JM,
TN
Taylor,
MR
de Lima
1993
The ultrastructure of
fossil
ephedroid pollen
with
gnetalean
affinities from the
Lower
Cretaceous
of
Brazil. Rev Palaeobot
Palynol
77:171-184.
Osborn
JM,
TN
Taylor,
EL
Schneider
1991b
Pollen
morphology
and
ultrastructure
of the
Cabombaceae: correlations with
pollina-
tion
biology.
Am J
Bot 78:1367-1378.
Pant
DD,
DD
Nautiyal
1960
Some
seeds
and
sporangia
of Glos-
sopteris
flora from
Raniganj Coalfield,
India.
Palaeontographica
B
107:41-64.
1984
On the
morphology and structure of
Ottokaria zeilleri
sp.
nov.-a
female
fructification of
Glossopteris.
Palaeontogr
Abt
B
Palaeophytol 193:127-152.
Payne
WW
1979 Stomatal
patterns
in
embryophytes:
their evolu-
tion,
ontogeny
and
interpretation.
Taxon 28:117-132.
Pearson HHW
1929 Gnetales.
Cambridge
University Press, Cam-
bridge.
194
pp.
Pedersen
KR,
PR
Crane,
AN
Drinnan,
EM Friis
1991
Fruits
from
the
mid-Cretaceous of North
America
with
pollen
grains of
the
Clavatipollenites type.
Grana
30:577-590.
Pigg
KB 1990a
Anatomically preserved
Dicroidium foliage from
the central
Transantarctic
Mountains. Rev
Palaeobot Palynol 66:
129-145.
1990b Anatomically preserved Glossopteris foliage
from
the
central
Transantarctic Mountains. Rev Palaeobot Palynol 66:
105-127.
Pigg KB,
TN
Taylor 1993 Anatomically preserved Glossopteris
stems with attached leaves
from the central
Transantarctic
Moun-
tains, Antarctica. Am
J
Bot 80:500-516.
Pigg KB,
ML
Trivett 1994 Evolution of
the
glossopterid gymno-
sperms
from Permian
Gondwana.
J
Plant Res
107:461-477.
Rao LN 1961 Life-history of Cycas circinalis L. Part I. Microspo-
rogenesis, male
and
female gametophytes and spermatogenesis. J
Indian Bot Soc 40:601-619.
Retallack GJ 1977 Reconstructing Triassic vegetation
of
eastern
Australasia: a
new
approach
for
the
biostratigraphy
of Gondwa-
naland.
Alcheringa 1:247-277.
1995 Permian-Triassic life crisis on land. Science 267:77-
80.
Retallack G, DL Dilcher 1981 Arguments for a glossopterid ances-
try
of
angiosperms. Paleobiology
7:54-67.
Reyman6wna M 1973 The Jurassic flora from
Grojec
near Krak6w
in
Poland. Part II. Caytoniales and anatomy
of
Caytonia.
Acta
Palaeobot 14:45-87.
Robinson-Beers K,
RE
Pruitt,
CS
Gasser
1992
Ovule development
in
wild type Arabidopsis and
two female-sterile mutants. Plant
Cell 4:1237-1249.
Rothwell
GW 1980
The
Callistophytaceae (Pteridospermopsida).
II.
Reproductive
features.
Palaeontogr
Abt
B
Palaeophytol
173:85-
106.
Rothwell
GW,
R
Serbet 1994
Lignophyte phylogeny
and the
evo-
lution of
spermatophytes:
a numerical cladistic
analysis. Syst
Bot
19:443-482.
Sanderson
MJ,
MJ
Donoghue
1989 Patterns of variation
in
levels
of
homoplasy.
Evolution
43:1781-1795.
Schopf
JM
1976
Morphologic interpretation
of fertile structures
in
glossopterid gymnosperms.
Rev Palaeobot
Palynol
21:25-64.
Serbet
R,
GW
Rothwell
1992
Characterizing
the
most
primitive
seed
ferns.
I. A reconstruction
of
Elkinsia
polymorpha.
Int
J
Plant
Sci
153:602-621.
Sharma
BD
1974 Ovule
ontogeny
in
Williamsonia Carr. Palaeon-
togr
Abt
B
Palaeophytol
148:137-143.
Sharma
BD,
DR
Bohra 1977 Structure of
phloem
in
some
plants
of Bennettitales
and
Pentoxylales
collected
from the
Rajmahal
Hills,
India.
Geophytology 7:214-216.
Singh
H 1978
Embryology
of
gymnosperms.
Handbuch
der
Pflan-
zenanatomie 10.
Borntraeger,
Berlin.
302
pp.
Sporne
KR
1965
The
morphology
of
gymnosperms.
Hutchinson
University Library,
London. 216
pp.
Stebbins
GL
1974
Flowering plants:
evolution above the
species
level. Harvard
University Press, Cambridge,
Mass.
399
pp.
Stevenson
DW
1988
Strobilar
ontogeny
in
the
Cycadales. Pages
205-224
in
P
Leins,
SC
Tucker,
P
Endress,
eds.
Aspects
of floral
development. Cramer,
Berlin.
Stevenson
DW,
AE
Artabe
1995
Fossil
and extant
cycads-
a cla-
distic
analysis.
Am J Bot
Suppl
82:82-83.
Stewart
WN
1951
A new
Pachytesta
from
the
Berryville locality
of southeastern Illinois. Am Midl Nat
46:717-742.
Stewart
WN,
GW Rothwell
1993
Paleobotany
and the
evolution of
plants.
2d ed.
Cambridge University Press, Cambridge.
521
pp.
Stidd
BM,
MO
Rischbieter,
TL
Phillips
1985
A
new
lyginopterid
pollen organ
with alveolate
pollen
exines. Am J Bot 72:501-508.
Swofford
DL
1990
PAUP:
phylogenetic analysis using parsimony.
Version 3.0.
Illinois Natural
History Survey, Champaign.
Takahashi
M
1992
Development
of
spinous
exine
in
Nuphar ja-
ponicum
De
Candolle
(Nymphaeaceae).
Rev Palaeobot
Palynol
75:317-322.
Takaso
T 1985 A
developmental study
of the
integument
in
gym-
nosperms.
3.
Ephedra distachya
L.
and
E.
equisetina Bge.
Acta
Bot Neerl
34:33-48.
Takaso
T,
F
Bouman 1986 Ovule and seed
ontogeny
in
Gnetum
gnemon
L.
Bot
Mag Tokyo
99:241-266.
Takaso
T,
PB Tomlinson
1989
Aspects
of cone and ovule
ontogeny
in
Cryptomeria (Taxodiaceae).
Am
J
Bot
76:692-705.
1990 Cone and ovule
ontogeny
in
Taxodium and Glyptos-
trobus
(Taxodiaceae-Coniferales).
Am J Bot
77:1209-1221.
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
DOYLE PHYLOGENETIC RELATIONSHIPS OF GNETALES S39
Taylor DW 1991 Angiosperm ovules and carpels: their characters
and polarities, distribution
in
basal clades,
and
structural
evolu-
tion. Postilla 208:1-40.
Taylor EL, TN Taylor 1992 Reproductive biology of the Permian
Glossopteridales
and their
suggested relationship
to
flowering
plants. Proc Natl Acad Sci USA 89:11495-11497.
Taylor TN, GM Del Fueyo, EL Taylor 1994 Permineralized
seed
fern cupules from the Triassic
of
Antarctica: implications
for cu-
pule and carpel evolution. Am J Bot 81:666-677.
Taylor TN, EL Taylor 1993 The biology and evolution of fossil
plants.
Prentice
Hall, Englewood Cliffs,
N.J.
Thomas HH 1925 The Caytoniales, a new group of angiospermous
plants from
the
Jurassic rocks
of
Yorkshire. Philos Trans
R Soc
Lond Ser
B
Biol Sci 213:299-363.
1933 On some pteridospermous plants
from the Mesozoic
rocks
of
South Africa. Philos Trans
R
Soc Lond Ser
B 222:193-
265.
Umeda
A,
R
Imaichi,
M
Kato
1994
Ovular development
and
mor-
phology of the outer integument of Magnolia
grandiflora
(Mag-
noliaceae).
Am J
Bot 81:361-367.
Upchurch
GR
Jr
1984
Cuticular
anatomy
of
angiosperm
leaves
from the Lower Cretaceous Potomac Group.
I.
Zone
I
leaves.
Am
J
Bot
71:192-202.
Vakhrameev VA 1991 Jurassic and Cretaceous floras and climates
of
the Earth. Cambridge University Press, Cambridge. 318 pp.
Van Campo M, B Lugardon
1973 Structure grenue infratectale
de
l'ectexine des pollens de
quelques Gymnospermes et Angio-
spermes. Pollen Spores 15:171-187.
Van Konijnenburg-van Cittert
JHA 1992 An enigmatic Liassic
mi-
crosporophyll, yielding Ephedripites
pollen. Rev
Palaeobot
Paly-
nol 71:239-254.
Walker
JW
1976
Evolutionary significance
of the exine in the
pol-
len of primitive angiosperms.
Pages 1112-1137 in
IK
Ferguson,
J Muller, eds. The evolutionary
significance of the exine. Academ-
ic Press, London.
Walker
JW,
AG
Walker
1984
Ultrastructure
of
Lower
Cretaceous
angiosperm pollen
and the
origin
and
early
evolution
of
flowering
plants. Ann Mo Bot Gard
71:464-521.
Weidlich
WH
1976 The organization
of the vascular system
in the
stems
of the Nymphaeaceae.
I. Nymphaea subgenera Castalia
and
Hydrocallis.
Am J
Bot
63:499-509.
Wettstein RR von 1907 Handbuch
der
systematischen
Botanik, II.
Band. Franz Deuticke, Wien.
577 pp.
Yao X,
TN
Taylor,
EL
Taylor
1995 The corystosperm pollen
organ
Pteruchus from the Triassic
of Antarctica.
Am J Bot
82:535-546.
Zavada MS
1991
The
ultrastructure
of
pollen
found in the
dispersed
sporangia of Arberiella (Glossopteridaceae).
Bot Gaz 152:248-
255.
Zhou Z, Zhang
B
1992 Baiera hallei Sze and associated
ovule-
bearing organs
from the Middle
Jurassic
of
Henan,
China.
Pa-
laeontogr Abt
B
Palaeophytol
224:151-169.
This content downloaded from 169.237.66.225 on Thu, 11 Sep 2014 15:21:28 PM
All use subject to JSTOR Terms and Conditions
