A Reevaluation of Seed Plant Phylogeny

50 Pages·1994·22 MB·

by  Kevin C Nixon

A REEVALUATION OF SEED
Kevin Nixon, 2 William
C. L. Crepet,2
and
PHYLOGENY Dennis Stevenson, 3 Else Marie A
PLANT Friis
1
Abstract
Seed plant phylogeny evaluated using a data set of 46 terminals (taxa) and 103 morphological and anatomical
is
characters. Cladistic analyses using the criterion of parsimony were performed on the complete data set as well as
on subsets of the data, e.g., excluding fossils and/or combining various complex taxa into single terminals. The results
support the placement of the cycads as the sister group of a monophyletic group that includes several fossil "seed
When
ferns" as well as extant Ginkgo, conifers, gnetopsids, and angiosperms. fossils were included, Bennettitales
(cycadeoids) were part of an "anthophyte" clade that included gnetopsids and angiosperms. Pentoxylon was a sister
taxon to the core anthophyte clade, in some, but not all, of the most parsimonious trees. Caytonia was not found to
be closely associated with the anthophyte clade, but instead was often associated as a sister taxon of the glossoptends,
and these two taxa were consistently outside of the Gin&go-conifer-anthophyte clade. In all most parsimonious trees
for all analyses, Ephedra was to the outside of a clade that included all angiosperm taxa, Gnetum, and Welwitscnia,
thus rendering the traditional gnetopsid clade paraphyletic. New information provided on the morphology of Caytonia
is
and some previous interpretations of homology of the caytonian "cupule" are rejected. The effects of sampling,
may
compartmentalization, and polymorphism are explored in these data, snowing how different results be obtained
when polymorphic or "summary" terminals are used. The need for more work on gnetopsids and fossil taxa is
suggested.
focus
Introduction we expect readers to
used characters, fully
cladograms
most intently on the topologies of the
In recent years, several phylogenetic analyses
our
we While understandable,
that present. this is
of seed plants have been published, some with
analyses,
present seed plant
position that at all
emphasis on is
explicit the relationship of angio-
molecular, are
& whether they are morphological or
sperms
to other seed plants (Hill Crane, 1982;
numerous prob-
& from
highly preliminary, and suffer
Crane, 1985; Doyle Donoghue, 1986a, 1987,
b,
& We we most important
and
1992; Loconte Stevenson, 1990). refer the lems. Thus, the second, feel
& and evaluate the
reader to Loconte Stevenson 990) for a sum- goal of this paper, is to discuss
1
(
broa
mary and comparison of the results of several of numerous problems encountered in such
prob-
these studies. In this paper we present a new cla- analyses, show the effects of some of these
partia
some or at least
distic analysis of seed plant phylogeny based on lems, and propose solutions,
over-
However, the
morphological characters. Unlike the Loconte and solutions, to some problems.
nw«
Stevenson analysis, our analyses include several riding message that we wish to convey is that
ha^e
analyses
such
fossil taxa as terminals, as did the analyses by Crane of the problems associated with
Ihe
&
point.
(1985) and Doyle Donoghue 1986a, not been adequately dealt with to this
(e.g., b,
e
large-sea
with
1987). problems associated
solutions to
additiona
may
require
This study encompasses two major goals. The analyses of diverse taxa
improvemen *
first, and most obvious, to provide a character theoretical advances as well as vast
is
^n
capabilities-
analysis of seed plants, followed by a cladistic anal- computer hardware and software
in
lya
ysis and a phylogenetic interpretation of those analyze efficient
re- particular, the inability to
taxa
sults. While we feel that we have provided some sets that are larger than a few score
B»J
which
new
insights, a few new characters, and what we "compartmentalize" analyses,
users to
parsimonio
more
feel are neutral codings (i.e., fewer phylo- result in solutions that are not globally
&
genetic hypotheses encoded) some
of previously (Nixon Carpenter, in press).
Pe*
We r
Hid
1 thank K. R. Pedersen and M. A. Gandolfo for assistance during various phases of this project,
Stevens and Gar
Rothwell for helpful and constructive comments on the manuscript.
2 L. H. Bailey Hortorium, New
Cornell University, Ithaca, York 14853, U.S.A.
New
3 York Botanical Garden, Bronx, New York 10458, U.S.A.
* Department of Paleobotany, Swedish Museum of Natural History, Stockholm, Sweden.
Ann. Missouri Bot. Gard. 81: 484-533 1QQ4
Number
Volume 81 3 Nixon et 485
al.
,
Seed Plant Phytogeny
1994
Methods
and numbered
Materials characters are starting with character
0. The majority of characters (80) were binary; of
SAMPLING
the remaining 23 multistate characters, 15 were
Fossil taxa were selected for inclusion on the treated as nonadditive (unordered) and eight were
basis of availability of published or original data, treated as additive (ordered), as discussed in Ap-
and whether or not we had confidence in proposed pendix A.
reconstructions based on separate fossil organs. In addition to the complete data set, several
Wherever possible, we checked characters as used modified data sets were also constructed and an-
in previous analyses against original descriptions alyzed. These included data sets that had certain
and revisions in the literature, and in some cases terminals or groups of terminals excluded com-
we reinvestigated fossil morphology when material pletely, and/or groups of terminals "condensed"
was available (e.g., Caytonia, some bermettitalean into single terminals. Terminal condensations were
DAD
A
taxa). These investigations resulted in differences automatically generated in the matrix editor
in some character codings relative to previous stud- (DADA386 ver. 0.87, Nixon, 1993b). The "strict"
DADA
ies. These differences are discussed below in the condensation option of was selected, in which
section describing fossils, and in Appendix A, which the new generated "condensed" terminal was scored
treats all characters used in the analyses. as polymorphic (*) for all characters that would be
Extant taxa were sampled to ensure represen- polymorphic (variable within the original terminals
DADA
tation of major seed plant groups. Within angio- within the group). translates polymorphic
sperms, taxa were selected to represent broad vari- cells to ambiguous (— ) upon submission to the
ation and to ensure representative sampling of Hennig86 daughter process, but retains the poly-
groups that have previously been hypothesized to morphic coding (*) in the working matrix. Terminal
DADA-
be morphologically were automated through the
similar to hypothetical primitive exclusions also
angiosperms. For example, Magnolia (Magnoli- Henning86 interface, assuring accuracy by retain-
aceae) represents the traditional Besseyan view ing a single master copy of the complete data
(bessey, 1897) of the primitive angiosperm, Cer- matrix.
atophyllum (Ceratophyllaceae) the branch in The data sets generated by the above procedures
first
we
various morphological (Les, 1988) and rbcL anal- included numerous combinations that are not
The
yses, and Chloranthus (Chloranthaceae), Piper reporting here because of space limitations.
(Fiperaceae), and the monocots the possible basal results that we report here are based on the fol-
groups (e.g., Burger, 1977, 1981). The "Amen- lowing five matrices:
tiferae," 46 103 characters;
considered basal in the "Englerian" view, Complete analysis: taxa,
I.
*ere represented by Chrysolepis (Fagaceae), Bet- 23.61% ambiguous, of which 1.36 are polymor-
"to (Betulaceae), and Casuarina (Casuarinaceae). phic (ambiguity 17.51% excluding 11 within-an-
Urimys
92-102).
Winteraceae)
represents the modified Bes- giosperm characters,
(
seyan hypothesis of Cronquist 1981) and Fossils excluded: 31 taxa, 97 characters;
(e.g.,
II.
0.32%
others, and Calycanthus (Calycanthaceae) the "first 12.31% ambiguous, of which are poly-
ranch" of 7.69% excluding 11 angio-
the morphological analyses of Loconte morphic (ambiguity
&
Stevenson 92-102).
(1991). Other angiosperm taxa were sperm characters,
29 83
eluded in an effort to represent major groups and Single angiosperm terminal: taxa,
III.
1.93%
Provide connecting among 21.77% ambiguous, of which
structure the other ter- characters;
"Nnals;
e.g., Platanus and Hamamelis, based on are polymorphic.
angiosperm terminal:
Previous analyses, appear to be reasonably placed IV. Fossils excluded, single
tween ^e 6.61% ambiguous, of which
66
"higher hamamelids" (Fagaceae, Bet- 14 taxa, characters;
aceae,
j Casuarinaceae in our analyses; see Nixon, are 3.59% polymorphic.
y angiosperm and co-
and
the lower hamamelids Trocho- V. Fossils excluded, single
) (e.g.,
17.8% am-
endron, 57 characters;
9
Tetracentron) and Because terminal: taxa,
magnoliids. nifer
e wanted 1.54% polymorphic.
to combine our data with existing rbcL biguous, of which are
ata in a separate project (Albert et 1994), Note that the number of informative characters
al.,
rnpnng No characters were
decisions were further restricted a priori for reduced data sets varies.
condensed
y the availability of r6cL data for taxa in our removed or receded other than for the
used
analyses. number of characters
Differences in
terminals.
The when taxa are re-
"complete" reduced variation
data included taxa (46 reflects the
set all
^
some characters
finals) and moved and/or condensed. Thus,
characters (103) as presented in
e „«i,.^^ri tUaf varv onlv among terminals that
Following Henning86,
the convention of
1
•
486 Annals of the
Missouri Botanical Garden
= — = =
Table 1. Data matrix. Characters 0-59. ? unknown; inapplicable; uncertain homology; * =
.
polymorphism. All treated as missing data.
5 10 15 20 25 30 35 40 45 50 55
I I I I I I
Aneurophyton 0007010001000 ???0000??0?0-0-*07??00?00-00?0000-0-?-?--0?---
Archaeopteris 00071 1001 1210-???000000000-0-0000?001 0002000000-0-000- -0?-?-
Lyginopteris 0.071 10000210-77700000001 1000000007000. 100000-0-000--0777?
11
Medul Losaceae 0.07210000210-77720000001 10*0000007770. 101010-0-000--0777?
11
Callistophyton 0107100000210-77710000001. 1100070000000031011 10- 10000--0011?
Corystosperm 0177120001200-7771010000101 10000070010003001 110-11000--0777?
Lepidopteris ????????????7-???10?0???*11 10007070010003001 110-0-000--0777?
Tatarina 011??????????-???10?00??0111000?0?00?00?30?1110-?????--2??7?
Glossopterids 010- 130101200-777100000001 101070000. *00. 0711 10-1 1077--2????
1
Caytonia ???????????7?-????00?0??0. -1010107001000311 10-1 1000--0777?
11
Williamsoniella 0107777777777-777700000701 110000171. 01 102101 110-0- 171 100777?
Cycadeoidea 01071001001 10-777200000111110000171. 011021011 10-0-111 100777?
Wi iamsonia 01071001001 10-7772000001 110000170.01 102101 110-0-1?11?0?7??
11
1 1
Pentoxyl on 01 1723001 1200-7770000000011 10000-000. 00. 0001 110-0-.00--07???
Cordaites
0117110111200-777100000000-0-000000010000011011-10000--00???
Cycadaceae 0000221 100210-00020000001 10000000010003001 10-0-000--00010
1 1 1
Stangen'aceae 0000121 1001 10-00020000101 10000000010003101 110-0-000- -00010
11
Zamiaceae 0000-21 1001 10-00020000. 10*00000010003101 110-0-000--00010
01 11
Ginkgo
01 1013011 1200-000100000100-0-000000010013111110-0-000- -00010
Taxaceae 01 101 101 11200-000701000102-0- -000000100030101 11 -0-100- -001 10
Taxod Cupress 01 101 101 11200-000201*00102-0- -0000001*0031 101 11 -0-100- -001 10
Araucariaceae
0110110111200-000201*0010*-0-000000010003010111-0-100--00110
Pinaceae 1-*0000--00110
01 101 1011 1200--00201000102-0--0000001001301.il
Cephalotaxaceae
0110110111200-000201000102-0--00000010003010111-0-000--00110
Podocarpaceae
01101 10111200-00020100010*-0-00000001001301.111-*0000--00110
Ephedra
0201 13011 121 100010001 10102-0--00000.011 13101 100-0-101 1110100
Welwitschia
02011301112110001200110101011100112.01110101110-0-101111110
Gnetum
02111201112110001200110101121100112.11110100100-0-101100110
Chloranthus
.10113010-111101111011110112111110101-1111201*0-0-111010210
Piper
.10112010-1111011110010101221101101011011120110-0-111000210
Winteraceae
0101140101110-011110000101121101111110011120110-0-111010210
Calycanthus
02011301012111111110100101121101101110011121110-0-111000210
Eupomatia
01011201011111111110000101121101101110012121110-0-111100210
Magnolia
01011201011111111110001101121101111110012121110-0-111100210
Persea
0101130101211111111000010112110111117101.120100-0-11.-00210
—
Nymphaea
210112010 — 0-01101001 1101221101 111110012121110-0-111100210
L urn 110112010 11111000.10101021101101111011121110-0-111010210
i I i
Di llenia 0101120101111171100000010112110110111101112012000-111010210
Caltha
110112010-111101100001010122110110111101112012000-111010210
Trochodendron
0101120101110-011110000101121111111.1001112012000-111010210
Platanus
0101120101211101100000110122110110211101112012000-111010210
Hamamel
is 0101140101211101100000110112110110111111212012000-111010210
Chrysolepis
0101140101201101100000110112110110211111212012010-111010210
Betula
0101140101211101100000110112110110211111212012020-111000210
Casuan'na 1000210
0101 140101211101 10001 1010210--01 10001 --1212012020-11
Ceratophyllum --0210
210101010- --0-01 1000100100-0-00- 1120100-0- 11-
-001 1001
1
487
Number 3 Nixon et
al.
Volume 81
,
Seed Phylogeny
Plant
1994
60-102.
Table Continued. Characters
1.
60 65 70 75 80 85 90 95 100
I I I
I
?--- 0-007-0
Aneurophyton 00
7000-007-0
700-
Archaeopteris -00
Lyginopteris 7000-01 110000100001001000000770?
7000-0111000-101001101007000????
Medullosaceae
Callistophyton 7000-021 100007-1 101001707000????
Corystosperm 71100022001070- 110000071????????
?1?0-022?010?0-1100000?0??????7?
Lepidopteris
7170-022101070-1 10000070????????
Tatarina
?1?0-022*000?0-1000001?0??00????
Glossopterids
?1?0-02210?0?0-1000001?1??????7?
Caytonia
?1?0-0100010?101000?00?1???????7
iamsoniel
Ui la
1
0-010001071010000001 1?0???1?7
Cycadeoidea 71.
?1?0-0100010?101000000?1?0?????7
iamsonia
Wi
1 1
?1?0-0200000?0-1001001?1???????7
Pentoxyl on
10100?70?000????
100200.0070-1
Cordai tes 711
0000-071100000-10011001000000000
Cycadaceae
0100-011000100-10011101000000000
Stangeriaceae
0100-011000100-10011101000000000
Zamiaceae
0100-02.000000-11010001000000000
Ginkgo
170-020000000-11000001000000101
Taxaceae
1117020*00000-11000001000000101
Taxod Cupress
1111020010000-11100001000000101
Araucariaceae
1111020010000-11100001000000101
Pinaceae
1100020000000-11010001000000101
Cephalotaxaceae
11000200-0000-11010001000000101
Podocarpaceae
1100010001011111000001100000101
Ephedra
1100010001011110100002111111111
Welwitschia
1100010001011110000002111111111
Gnetum
--111-1101
110110110?.
00001102000001
Chloranthus 1.0-12.
1.0-12.0000110200000111111011070--10113112
Piper
1.0-12.1100110200000111011011010000111-000
Winteraceae
1.0-12.0100110200000111011011011000110-000
Calycanthus
00110-000
1100110200000111011011011.
Eupomatia l!o-12.
1.0-12.0100110200100111011011010000110-000
Magnolia
1.0-12.010011020000011101101100001101-3101
Persea
1000000110-000
110010020000011101101
Nymphaea 1.0-12.
1.0-12.11001102000001111110110000110013110
Lilium
1-000
100001001
1020000011101101
0-12. *0001
Dillenia
1.
0100110200000111011011000010010-000
Caltha 1^0-12.
1.0-12.11001102000001110110110010-0011-010
Trochodendron
1.0-12.00001102000001110110110000100114101
Platanus
102000001110110110111100112121
01001
Hamamelis l!o-12.
1.0-12.01001102000001110110110011110013121
Chrysolepis
01001002000001110110110111100012121
KO-12.
Betula
--000121-1
101101101.
1020000011
01001
Casuarina l!o-12.
000010720000011101101100011011-1100
CeratophyL l!o-12.
urn
I
488 Annals of the
Missouri Botanical Garden
are excluded in the condensed analyses. For ex- terminals, trees were rooted between cycads and
ample, characters that are relevant only within- the remainder of the extant seed plants, based on
angiosperms (92- 02) were operationally excluded the results of the full analyses and previously pub-
1
&
from analyses in which angiosperm taxa were con- lished analyses (Loconte Stevenson, 1990).
It
densed into a single terminal. should be noted that such rootings do not
affect
Additionally, as reported elsewhere (Albert et the topologies of the trees reported.
1994), the complete morphological data
al., set
was combined with rbch sequence data. For those Theoretical Considerations
analyses, fossil taxa were scored as ambiguous
missing data
rbch
(missing) for all data.
Cladistic analyses were performed on Intel 486 Problems associated with missing data and /or
IBM PC
compatibles using Hennig86 version 1.5 polymorphism have not been discussed extensively
DOS
(Farris, 1988), run as a 16-bit daughter pro- in the literature, but interest in this topic has in-
DADA386 &
cess of ver. 0.87. All modified data sets creased in recent years (e.g., Nixon Davis, 1991).
DAD We
were generated directly within A, ensuring prefer to separate the concept of missing data
that characters were scored identically in sep- from that of missing values or general ambiguity.
all
arate analyses. Alternative character optimizations The former occurs when data are unknown or not
and trees for publication were produced using da- collected, and the latter when a cell is scored as a
dos ver. 1.4 (Nixon, 1993a). "missing" value for reasons other than missing
In order to increase the likelihood of finding a data, such as uncertain homology or polymor-
complete set of most parsimonious trees (Mickevich phism. Most currently available computer pro-
& & PAUP
Farris, 1982; Luckow Pimentel, 1985), taxon grams, including Hennig86 (Farris, 1988),
NONA
order in the larger data sets was pseudorandomly (Swofford, 1991), (Goloboff, 1993),
&
shuffled 100 times using the "autospin" option of MacClade (Maddison Maddison, 1992) and da-
DADA
(Nixon, 1993b) and directly submitted to dos (Nixon, 1993a), treat missing data computa-
Hennig86 using the "mh*" option. Trees were tionally the same during optimization and calcu-
had
collected from each run, and the final set was lation of tree length, e.g., as if the missing cell
analyzed with heuristic branch swapping ("bb*"). possible states present. Some programs that
all
The command DADA
autospin of automatically search most parsimonious cladograms (e.g..
for
removes redundant trees from replicate runs by PAUP and NONA) allow the coding of subset poly-
but not
filtering tree files through the Hennig86 "xsteps morphism states and 2 present,
(e.g.,
u" coded as
option. This provides an accurate count of the state while polymorphism must be
1),
number state-
of trees found, and allows generation of a ambiguous (missing), or equivalent to all
consensus from Hennig86. Al-
tree all the unique trees generated present, in the current version of
by & the prob-
replicate runs. though Nixon Davis (1991) discussed
Some coded as miss-
of the data sets that were generated by lem the context of polymorphism
in
are
problems
condensing groups had few enough Hennig86), these
terminals to ing data (e.g., for
coding a*
allow analysis by the "ie*" ("implicit enumera- not solved by subset polymorphism
do
problems
PAUP NONA.
tion") option in Hennig86, which guarantees and Other
to available in
polymorphic
when
find the complete set of most parsimonious trees occur, and possible that
is
it
states,
for the data. For these data sets (IV, V), replicate terminals possess only a subset of possible
tr^
randomized taxon order was unnecessary. PAUP and Hennig86 may report different
an^
Two extensive
non-seed-plant fossils (Aneurophyton, Ar- lengths for the same topology; thus,
Heiml
chaeopteris) were included in the full data matrix, correlated polymorphism may not allow ^
n
and were However, *
trees rooted between Aneurophyton and to find the most parsimonious trees.
^
ou
pointed
the remainder of taxa. All analyses were simulta- such conditions exist, the problems ^
m
even
neous (outgroups included without a priori as- Nixon & Davis 99 are most likely
( 1 1 ) ^_
with c
sumptions about topology Farris, 1972, 1982; Clark severe than the problems associated
&
Curran, 1986; Meacham, 1984, 1986; Nixon subset polymorphism as missing, and it is °
& wo
Carpenter, program
in press). Because outgroups were whether the results of either
subgr ^
treated monothetic
like all other taxa in the analysis, clado- stable to further analyses of
P^
tm>
grams were to
globally most parsimonious for the of polymorphic taxa. The best solution
charade
+
entire dataset (outgroup ingroup). The final lem to avoid complex multistate
is
cladograms were much possible-
rooted between as
the outgroup polymorphic terminals as
ter-
rf
ortior*
prop
minals and
ingroup. high
In analyses that excluded Although obvious that
fossil
it is
Number
3 Nixon
Volume 81 et 489
al.
,
Seed
1994 Plant Phylogeny
missing data introduce uncertainty into analyses, neously versus compartmentalized and/or
con-
We
it is not yet clear whether this uncertainty is pre- strained. present some examples of this prob-
how measure how much lem
dictable, to it, or missing using various modifications of our seed plant
data is allowable within a particular analysis before matrices below. Contrary to popular belief, this
is
the results should be rejected as unreliable. This a problem whether computer programs accept
topic presently the subject of experimental polymorphic codings or whether polymorphisms
is in-
vestigation (Nixon, in prep.). Preliminary results must be coded as ambiguous. Wherever possible,
we
indicate a predictable interaction of missing data have reduced polymorphism in our data by
and homoplasy; the effect of missing data ac- breaking terminals into units that are monomorphic
is
&
centuated by high levels of homoplasy. Data sets for each character (Nixon Davis, 1991).
numerous may
that include fossil taxa often have high Compartmentalization result in equally se-
levels of missing data as a result of imperfectly rious errors, which are more subject to researcher
preserved fossil organs and the difficulty in deter- bias, when polymorphic terminals are not coded as
mining homology of characters that cannot be polymorphic or ambiguous, but instead are coded
in-
vestigated equally in terminals. Thus, recent as having the "presumed" plesiomorphic state for
all
analyses of seed plant relationships have suffered the group represented by the terminal. In the Doyle
&
from high levels of missing data, often in excess Donoghue (1986a, b, 1987) analyses, angio-
of 20% of the total number of cells (e.g., 24% in sperms and some other diverse groups, such as
&
Doyle Donoghue, 1986b; 23.6% in our com- conifers and cycads, were treated as single ter-
i
plete analysis, or 17.5% excluding inapplicable minals, and several characters were scored either
states coded as ambiguous for some within-angio- as polymorphic (in those cases missing because of
sperm
characters). In our analyses, the vast ma- the computer programs used) or as the presumed
jority of cells scored as missing were due to lack primitive state for all angiosperms. Thus, for ex-
of information for fossil taxa; this reflected in ample, the "angiosperm" terminal in the above
is
much
the
lower levels of ambiguous data when analyses was scored as having alternate leaves,
fossils were excluded from the analyses. Polymor- bitegmic anatropous ovules, and spirally arranged
phism
in our analyses was low, ranging from 0.3 microsporophylls (= stamens), under the assump-
3.59%
to of the cells in the various matrices. The tion that these states were primitive in angiosperms.
highest levels were in those matrices with con- On the other hand, other polymorphic characters
ensed terminals, because condensations generate of angiosperms were scored as missing, or ambig-
Polymorphism for each character that varies among uous, such as presence of vessels. Given these
the condensed
terminals. choices in character coding, is not surprising that
it
&
the Doyle Donoghue (1986a, b, 1987) analyses
angiosperms
did not find a close relationship of to
COMPARTMENTALIZATION,
ARTIFICIAL TERMINALS
integuments,
N® Gnetales (with opposite leaves, single
POLYMORPHISM
and
orthotropous ovules, whorled microsporophylls,
^e
of the most serious problems previous foraminate- or simple-plated vessel elements). In
in
analyses
of seed plants has been the treatment of order to avoid such bias, we have sampled several
complex,
diverse groups as single terminals, which angiosperm taxa for observed characters, instead
en
presents when presuming an angiosperm ancestral form that
difficulty these terminals are of
Nymorphic
(variable within the group as a whole) predisposes placement of angiosperms in our clado-
0r
certain characters used Coding grams. Of course, such an approach has other
in the analysis.
^ mals undersampling
as polymorphic or ambiguous for these problems, mostly due to potential
&
aracters may Nixon Wheeler, 1991), but we consider
result in incorrect tree topologies, (see
a*
pointed out by Nixon & Davis (1991). In some these problems to produce less bias than does the
s, cladograms found when terminals are scored practice of designating ancestral states for poly-
as
Polymorphic
do not include any of the most morphic terminals.
rsimonious
cladograms when
found the terminal
J*
subdivided
into monomorphic The method
units.
HOMOLOGY DECISIONS
coding
diverse
terminals
as single units in
is
effect ***»,* any broad
Partmentauzation" and constraint of One of the most serious problems in
A
th
data problem
J (see Maddison et 1984). This method cladistic analysis of higher level taxa is the
al.,
^P
artr homology assessment among highly divergent
nentalization relaxes the application of of
Parsirnony, which Typically, the more closely related taxa are,
accounts for the disparate taxa.
f
may
0D homology be.
tained when such assessments of
analyses are run simulta- the easier
490 Annals of the
Missouri Botanical Garden
Seed plant analyses of necessity include taxa with of course such character interpretations
of
fossils
disparate morphology of reproductive and vege- are always open to debate. In cases where we
have
Some
tative structures. uncertainty exists in ho- interpreted certain homologies in fossil taxa differ-
mology assignment for some taxa for almost every ently from authors of previous analyses, we
discuss
character in the analysis. For instance, whether these differences in some detail in both the
dis-
angiosperm flowers are compound strobili, and cussion of characters and the discussion of the
fossil
We
whether carpels are "megasporophylls" are just a taxa. also have changed some of the terminal
few of the problems encountered in an analysis of units in the analyses in line with recent
fossil dis-
seed plants. In order to avoid decisions that are coveries so that there less character polymor-
is
laden with phylogenetic hypotheses, we have at- phism in the terminals. In these instances we have
much
tempted to code characters based as as pos- provided more detailed descriptions of the
fossils
on Whether new
sible direct observation of features. or in conjunction with establishing terminal units.
we
not have succeeded or failed in this effort will
be left to the judgment of the reader, and un- FOSSIL PROGYMNOSPERMS
doubtedly some vehemently argue complex
will that
Aneurophyton. Aneurophyton
typically includ-
is
models should be incorporated into the analysis.
ed in phylogenetic analyses of the seed plants as
An
example of such a complex model used by
an outgroup terminal, and we follow this tradition
previous authors the "cupule" of Caytonia, which
is
here. Other Aneurophytales, such as Tetraxylop-
has been coded as homologous with the second
known
and Rellimia, are better but appar-
teris
integument of angiosperms and Bennettitales by
more
ently derived in certain characters, e.g., pla-
&
some
previous authors (Crane, 1985; Doyle
&
(Bonamo Banks,
nation of reproductive structures
We
Donoghue, 1986a,
b, 1987). discuss our in-
1967; Bonamo, 1977, 1983), and should be in-
terpretation of this structure, both in terms of cod-
Aneurophyton char-
cluded in future analyses. is
ing and in terms of our cladistic results, in greater
acterized by three-dimensional branching systems
detail below.
sometimes decussate) ar-
with helical (or possibly
stems,
rangement of lateral branching, protostelic
Included Taxa
and
considerable development of secondary growth,
&
See Table 1 for a list of taxa and the complete terminal, fusiform sporangia (Serlin Banks 1978).
data matrix. Stems include a multilobed strand of mesarch pri-
secondary
mary xylem surrounded by pycnoxylic
The
TAXA
FOSSIL xylem and tracheids with bordered pits. ulti-
two- or
mate appendages are three-dimensional,
Fossil taxa were selected for inclusion in our
are often
three-times dichotomizing branchlets that
analyses on the basis of availability of data,
i.e., o
leaves
homologous with the
considered to be
completeness
of or presumed
fossils reconstruc-
some van-
There
more progymnosperms.
recent is
and
tions, relevance to the problem of seed plant
Aneurophy-
reproductive structures in
ability in
phylogeny. While numerous
angiosperm
are
fossils
lat-
aggregate fertile
some having
with taxa
now tales,
known,
relatively few are complete enough
or
towar
tendency
branches demonstrate a
eral that
different enough from modern
taxa to be relevant
somewha
be
appearing to
planation as well as
in a broad seed
plant analysis (see Crepet
et
al., Aneurophyton
In
Tetraxylop
circulate (e.g., teris).
1991,
for a review of fossil flowers). Thus, we have
&
ult«-
germanicum Weyland the fertile
Krausel
not included any angiosperm
fossils in our analyses. two
VkO
and tn
mate branchlets dichotomize once,
The
descriptions of key fossil taxa below are not U*
rows of
two
branchlets curve inward and bear
intended to be complete, but they emphasize
in- Aneu-
All
form sporangia on the inner surfaces.
terpretations of fossil characters that differ from have
v»
preservation
have
rophytales that sufficient
&
previous
analyses (e.g., Crane, 1985; Doyle
&
a
homosporous (Taylor J
been found be
to
Donoghue,
1986a, b, 1987, 1992). Sometimes the
1993).
differences in interpretation are based on new
in-
commonly ccfl
formation, either from recent literature or direct Archaeopterids. Archaeopteris is
trU
recon
reinvestigation our genus
(e.g., reinvestigation of the "cu- sidered to be a composite /
( f
like
fron^
pule"
of Caytonia). Other includes
differences reflect re- from more than one fossil) that
s
interpretation based on the same data available to branching systems and trunks of varying •
other workers; in such cases, we hope that we have is a historically significant genus becausePiWt £
simplified interpretations and made them concept of
less de- mately was the basis of the
e
pendent on associa
a priori phylogenetic hypotheses, but nosperms." Beck (1960) was able to
Number 3
Volume 81 Nixon
et al. 491
,
Seed
1994 Plant Phytogeny
wood
coniferous secondary of the trunks (Callixy- Sternberg), and several types of microsporangiate
bn) with frondlike plana te branching systems (Ar- organs (Crossotheca, Telangium, Telangiopsis, and
chaeopteris) based on similar anatomical details. Feraxotheca) that are pinnate and laminar with
The stem has a eustele of mesarch vascular bundles abaxial sporangia Crossotheca and Feraxo-
in
with pyncnoxylic secondary wood having distinc- theca, but terminal in Telangium and Telangiop-
tive bands of circular bordered pits on the radial sis (compression fossils with the characteristics of
walls of the tracheids (e.g., Beck, 1960). Carluccio petrified Telangium). Stems, leaves, and cupules
et al. (1966) demonstrated that the ultimate have distinctive capitate epidermal glands (used by
&
branching systems of planate, variously dissected Oliver Scott, 1904, to recognize the "seed
webbed leaves, were actually spirally arranged on ferns"). Stems have anastomosing radially elongate
&
all orders of branching. This supported the inter- bands of fibers in the outer cortex (Oliver Scott,
pretation that Archaeopteris had planate lateral 1904). There considerable variability within the
is
branching systems and not fronds as originally family Lyginopteridaceae as usually circum-
it is
& &
thought. Anatomical studies by Beck (1971) and scribed (Stidd Hall, 1970; Taylor Millay,
Scheckler (1978) have refined further our under- 1981; Galtier, 1988) and relationships within and
standing of Archaeopteris morphology, anatomy, outside of the family will be better understood as
and ontogeny and have demonstrated that branches new fossil evidence becomes available. In our anal-
originated from axial primordia and not from ax- yses we have used a broad composite profile for
illary buds. the lyginopterids as our terminal, but anticipate
Various hypotheses link the Archaeopteridales that in future analyses this terminal would be di-
to taxa with large frondlike leaves (the "cycado- vided into subunits based on observed variation
phytes") based on presumed homology of the within the group.
lat-
eral planate branching system Some previous analyses have interpreted
with fronds, as well
as to simple-leaved branch systems in conifero- branching in Lyginopteris as axillary as opposed
& We
Phytes (Meeuse, 1963; Beck, 1971). More re- to dichotomous (Galtier Holmes, 1982).
cently,
Rothwell (1982) has suggested an interpret the branching pattern of the lyginopterids
alter-
native
origin of conifers from Callistophyton-\ike as dichotomous, but with occasional close dichot-
ancestors based on common mimic branching
pollen structure and omies that superficially axillary
& We
Platyspermic ovules, among other characters. (see Taylor Millay, 1981). do not know of
Reproductive branching
structures of Archaeopteris are any definitive evidence of axillary in this
usiform
sporangia borne adaxiaUy in rows along group. At any rate, because of the diversity within
terminal some
fertile leaflike appendages that are ter- the group, it is possible that taxa referred to
"wnally
dichotomous had dichotomous branching while
(e.g., Beck, 1971). Ar- as lyginopterids
c
*aeopteris species are considered to be either others had axillary branching, or a mixture of both
°mosporous or heterosporous, and has been forms of branching. Thus, we have scored our
it
suggested unknown presence
that as taxa become known lyginopterid terminal as for the
fossil better
sPecies will be discovered to have been hetero- or absence of axillary buds/branching.
^°rous
(Phillips may
et 1972). Megasporangia
al.,
*
larger (Arnold, Medullosans are characterized by
1935), or smaller (Phillips et Medullosans.
* 972) than complicated but fun-
' the microsporangia. Fertile ultimate their large bifurcate fronds,
Ranches noncu-
are borne helically on penultimate damentally eustelar stem anatomy, large
&
Ranches that are arranged and complex synangia (Stewart
in planate branching pulate ovules,
*/s ems. Devonian seeds on branching systems have Delevoryas, 1956; Millay & Taylor, 1979; Taylor
**n
&
associated with Archaeopteris & Taylor 1993; Stewart Rothwell, 1993). They
plants (Ar-
-like
nold,
1935). compressions and pet-
commonly preserved as
are
and reproductive
and vegetative
rifactions their
F0SSIL
SEED
PLANTS
and
anatomy and morphology have been carefully
^frnopterids. 1955; Stew-
Delevoryas,
Lyginopteris has been treated in extensively studied (e.g.,
&
previous & 1956; Millay Taylor, 1979;
Delevoryas,
cladistic analyses of seed plants as art
&
1969; Eg-
rnposite taxon 1965, 1971; Taylor Eggert,
(reconstruction) based on several Taylor,
&
* e Pendently & Mapes Rothwell, 1980;
1979;
described taxa. These include gert Rothwell,
fossil
&
°r compressed eustelar stems Lyginop- Stidd, 1981, 1990; Dufek Stidd, 1981; Rothwell
'
(
t h petrified ovules (Lagenostoma) within lobed & Eggert, 1986), although complete suites of veg-
^Panulate known
cupules (Calymmatotheca), com- etative and reproductive characters are not
fronds stem taxa. Interpretations
(Sphenopteris hoeninghausii of the petrified
for
all
492 Annals of the
Garden
Missouri Botanical
of the stem anatomy and the origin of the syn- ceae and cordaites and even conifers (e.g., Roth-
&
angiate prepollen organs have been subjects of well, 1982; Stewart Rothwell, 1993). This
is
&
considerable controversy (Eggert Rothwell, 1979; one of the questions best answered in a phylogenetic
&
Rothwell Eggert, 1986; Stidd, 1990) and there context and we address the possibility of a
callis-
variation in key vegetative characters within the tophytacean-conifer relationship below.
is
Medullosa
genus as presently circumscribed.
it is
endocentrica Baxter (Baxter, 1949) appears to Cordaites. Cordaites constitutes a conspicuous
many
have a different growth habit (vine vs. tree) and group of plants in Late Paleozoic floras of
mode of branching the only medullosan with the Northern Hemisphere. The earliest descriptions
(it is
& We
Hamer
axillary branching; Rothwell, 1988). of cordaites were based on the large strap-shaped
have encoded the characters of medullosans with or lanceolate leaves of Cordaites and associated
treelike growth habit that do not have axillary wood and reproductive organs from Europe (Feist-
branching. In future analyses, pending availability mantel, 1876; Grand-Eury, 1877; Renault, 1879).
of characters, we will add the axillary branching Subsequently, a number of new cordaitalean taxa
medullosans as an additional terminal Our have been established from other areas and incor-
unit.
coding of characters has not been affected by con- porated in three distinct families: Cordaitanthaceae
troversies over the origin of the polystelic stem of (Cordaitaceae), Rufloriaceae, and Vojnovskyaceae
medullosans or about the homology of the large (Meyen, 1984, 1988). The Cordaitanthaceae in-
and complex pollen-bearing organs. clude Euramerican taxa, while the two other fam-
are based on Angaran forms. There is consid-
ilies
Callistophyton. The Callistophytaceae were erable structural variation among members of the
first
&
recognized and circumscribed by Delevoryas three families, particularly regarding the repro-
Morgan (1954) when they demonstrated that pet- ductive organs. The Euramerican cordaites are
Angaran forms because
rified axes superficially similar to Cordaites were better understood than the
actually very similar in morphology to the stems of abundant petrified material. Reviews of taxa
Meyen
of Lyginopteris. The vegetative and reproductive included in the three families were given by
anatomy and morphology have subsequently been (1988) and Rothwell (1988). The Cordaitantha-
&
Mesoxylon), pith
carefully described. Stidd Hall (1970) were able ceae include wood (Cordaixylon,
to demonstrate the association between ovules of casts (Artisia), roots (Amyelon, Stelastellara),
& cones (Cor-
the genus Callospermarion (Eggert Delevoryas, leaves (Cordaites), male and female
1960) and fronds of Callistophyton. Subsequent daitanthus), male cones (Gothania), seeds (eg..
and
Samaropsis),
studies by Rothwell (1972a, 1975, 1980, 1981) Cardiocarpus, Mitrospermum,
have made the genus one of the best known of the pollen (Felixipollenites, Florinites, Sullisaccites).
Paleozoic pteridosperms. There appear to be at least two groupings of taxa:
Cordaixylon-Cordaites-CordaitantJius-
Callistophyton reconstructed as a shrubby the
is
in
considered
plant with spirally arranged fronds with Florinites-Cardiocarpus complex,
axillary
Mesoxylon-Cordaites-
branches and
occasional adventitious roots at the the present study, and the
n-
Sullisaccites-
nodes. Synangia and ovules were borne on Gothania-Felixipollenites /
the
group-
other
abaxial surfaces of pinnules. Ovules were not borne trospermum complex, but there are
Rothw
in cupules and were platyspermic* with the nucellus ings not yet fully understood (e.g., ^
8g.
& IV
1985,
free from the inner integumentary layer except at Warner, 1984; Trivett Rothwell,
the base. Synangia consisted of a ring of laterally Rothwell, 1988).
**
branching
fused elongate sporangia. Pollen-bearing organs The were profusely
in- cordaites tall,
Warnj
& &
clude Callandrium (Rothwell
(Stidd Hall, 1970) and Ida- or smaller trees or shrubs
spi^
& wood and
nothekion (Millay Eggert, 1970). Pollen in these 1984), with eustelic, pyenoxylic
arej^;
organs and pollen found in the micropyles of Cal- Stems of Cordaixylon
phyllotaxis.
"
S c0
lospermarion-type ovules is of the Vesicaspora while those of Mesoxylon are mesarch.
8:Q;
Renau^
type, monosaccate with a distal sulcus and two xylem are araucarioid (e.g., '
pits
& 1984;
The Warner,
lateral lobes. pollen often well preserved and 1912; Rothwell
is Scott,
jy
& conspicuous,
has been found with an intact branched pollen tube Rothwell, 1985). The pith is
tissue
parenchyma
(Rothwell, 1972b) and
with apparently intact with a septation of
ically
^
tulatc
spa
&
microgametophytes strap-shaped,
within the walls (Millay Eg- Foliage comprises large
an
^^
gert, 1970, 1974) that are similar to those in extant or narrow grasslike leaves (Cordaites), ^
**
conifers. These similarities have led to speculation in some species also small scale or needlel
ex
or
about the relationship between the Callistophyta- around buds, at the base of branches
Number 3 Nixon 493
Volume 81 et al.
,
Seed Plant Phylogeny
1994
&
along the branches (Rothwell Warner, 1984; to Cardiocarpus in most features, but the integ-
Rothwell, 1988). There is no mid-vein in Cor- ument vasculature forms a characteristic tracheal
daites, but numerous, densely spaced, dichoto- plate, and nucellar vasculature lacking (Taylor
is
&
mizing veins that run almost parallel in the leaf Stewart, 1964).
lamina.
A detailed account of the organization and Glossopterids. The glossopterids, the signal group
structure of the microsporangiate and ovulate of the Gondwana flora, are based on an imprecisely
cones of Cordaitanthus was given by Florin 950, understood assemblage of variously preserved fos-
1
(
1951). The pollen-bearing and seed-bearing cones including leaves, stems, roots, and a wide range
sils
are similar in basic form; they are compound, and of reproductive structures (e.g., Delevoryas, 1969;
&
consist of lateral branches borne on a central Pant, 1977; Schopf, 1976; Gould Delevoryas,
&
bracteose axis, either in a planate pattern or spi- 1977; Retallack Dilcher, 1981; Pigg, 1990;
&
An
rally arranged. Each lateral branch has a number Taylor Taylor, 1 993). excellent overview of
of spirally arranged scalelike sterile appendages pre- 1985 literature on glossopterids is provided by
interspersed with sporophylls. These are simple, Crane (1985) and there have been several inter-
laminar, with an apical cluster of four to six free esting and significant studies since that time, in-
microsporangia. Dehiscence by longitudinal cluding analyses of petrified material from Antarc-
is slits
& &
toward the center of the sporangial cluster. Pollen tica (Pigg, 1990; Taylor Taylor, 1993; Pigg
grains found in situ in Cordaitanthus are similar Taylor, 1990, 1993).
to dispersed grains assigned to Florinites. They The leaves have a multiveined midrib with anas-
are eusaccate, monosaccate, and inaperturate tomosing lateral venation that is not hierarchical
&
(Florin, 1936; Millay Taylor, 1974). Male ga- with respect to vein order diameters (e.g., Pigg,
metophytes
were observed in the pollen first by 1990), and our coding of venation reflects this
Henault (1902) and later described in detail by distinction. In this character, the leaves are similar
Horin
(1936). Four to five or six nuclei are ar- to those of Sagenopteris (considered to be the leaf
ranged
in a single row oriented towards the distal form of Caytonia).
pole. There a great deal of variation in the nature
is
The
megasporophylls are stalklike, and in the of ovuliferous and pollen-bearing organs. Repro-
"lore recent forms termed "glossopterid" include
are simple, bearing a single, ductive structures
terminal ovule (C sporangia on simple or
zeilleri Renault), or in the older seeds or pollen-bearing
forms
dichotomous, bearing two ovules (C. pseu- branched stalks subtended by foliar structures, with
dofluitans Kidston). Dispersed and seeds the stalks variously adnate to the foliar structures;
petrified
^sociated with Cordaitanthus are assigned to Car- sometimes the seeds are apparently borne directly
wcarpus. The some ovulate
ovules are platyspermic with a on the foliar structures. In petrified
*,ngle integument consisting of an inner sclerotesta structures, the ovules appear to be completely en-
and
an outer sarcotesta. A single vascular strand closed by a foliar organ. Within the envelope, the
en
|ers the ovule and gives rise to one pair of vas- orthotropous ovules are connected by a network
c
"'ar bundles that pass up through the integument of parenchyma cells that might be interpreted as
a" anot her pair of vascular bundles that pass into participating in pollination (e.g., Thomas, 1958;
&
e nucellus. The & 965; Gould Delevoryas, 1977;
nucellar and Pant Nautiyal,
cuticle distinct 1
is
&
fee rom tne integument and megaspore Pant, 1977; Taylor Taylor, 1993). Pollen-bcar-
thick
^mbrane.
branched axes
ing organs consist of variously ter-
ne microsporangia that con-
microsporangiate cone, Gothania, associ- minating in clusters of
* with Me.so.ry/oAi-type wood (Hirmer, 1933), tain bisac rate and apparently quasis.n < ate pollen
&
S,m,,ar to the compound cone of Cordaitanthus (Surange & Chandra, 1975; Gould Delevoryas,
J* general structure, but having sporangia 1977; Meyen, 1987). There has been some un-
differs in
anged
r about the na-
i n a single row at the apex of the spo- certainty expressed in the literature
^°P vll. Pollen grains found Gothania ture of these reproductive structures (i.e., foliar
in situ in
eusaccate and monosaccate, with a distinct vs. axillary). Based on the full range of forms, the
and
nature of the seed stalks, the
thT^ tri,ele S° ar simi,ar to dispersed pollen of apparent axillary
'
some
tf, opposing bracts in
"xi[j<>llcn,trs occurrence of additional
or Sullisaccites type (Millay
branch
U)r i9?4; & we regard these structures as axillary
Trivett Rothwell, 1985). Ovules taxa,
(r '
of
appendages,
1l 'rosf)ermum Q systems with variously modified fertile
are ]inked olhania by the
tQ
eserire Thus, encoding
of subtending bract. in
FelixipollenUes pollen in the micro- adnate to a
m
X Baylor & Glossoptms. «.,,fra>t to pre-
Taylor, 1993). They are similar the characters for <