ArticlePDF Available

Infrafamilial classifications and characters in Araliaceae: Insights from the phylogenetic analysis of nuclear (ITS) and plastid (trnL-trnF) sequence data

Authors:

Abstract and Figures

Traditional classifications of Araliaceae have stressed a relatively small number of morphological characters in the circumscription of infrafamilial groups (usually recognized as tribes). These systems remain largely untested from a phylogenetic perspective, and only a single previous study has explicitly explored intergeneric relationships throughout this family. To test these infrafamilial classification systems, parsimony and Bayesian-inference analyses were conducted using a broad sampling of 107 taxa representing 37 (of the 41) genera currently recognized in core Araliaceae, plus five outgroup genera. Data were collected from two molecular markers, the internal transcribed spacers (ITS) of the nuclear rRNA genes and the intron and intergenic spacer found in the trnL -trnF region of the chloroplast genome. The results suggest that there are three major lineages of Araliaceae, and that these lineages generally correspond with the centers of diversity for the family. The Aralia and Asian Palmate groups are centered primarily in eastern and southeastern Asia, whereas the Polyscias-Pseudopanax group is found throughout the Pacific and Indian Ocean basins. Several poorly resolved lineages are placed at the base of core Araliaceae, and the geographic distributions of these clades are consistent with a hypothesized rapid radiation of Araliaceae, possibly correlated with the breakup of Gondwanaland. Comparison of molecular results with the traditional systems of classification shows almost no congruence, indicating that they inadequately reflect phylogenetic relationships. Moreover, the morphological characters employed in these classifications appear to be highly homoplastic, and are thus of little utility at the infrafamilial level.
Content may be subject to copyright.
Infrafamilial classifications and characters in Araliaceae:
Insights from the phylogenetic analysis of nuclear (ITS)
and plastid (trnL-trnF) sequence data
G. M. Plunkett
1
, J. Wen
2
, and P. P. Lowry II
3
1
Department of Biology, Virginia Commonwealth University, Richmond, Virginia, USA
2
Department of Botany, Field Museum of Natural History, Chicago, Illinois USA and Laboratory
of Systematic and Evolutionary Botany, Institute of Botany, Chinese Academy of Sc iences, Beijing, China
3
Missouri Botanical Garden, St. Louis, Missouri, USA and Muse
´
um National d’Histoire Naturelle, Paris,
France
Received October 25, 2002; accepted October 12, 2003
Published online: February 26, 2004
Springer-Verlag 2004
Abstract. Traditional classifications of Araliaceae
have stressed a relatively small number of morpho-
logical characters in the circumscription of infrafa-
milial groups (usually recognized as tribes). These
systems remain largely untested from a phyloge-
netic perspective, and only a single previous study
has explicitly explored intergeneric relationships
throughout this family. To test these infrafamilial
classification systems, parsimony and Bayesian-
inference analyses were conducted using a broad
sampling of 107 taxa representing 37 (of the 41)
genera currently recognized in core Araliaceae, plus
five outgroup genera. Data were collected from two
molecular markers, the internal transcribed spacers
(ITS) of the nuclear rRNA genes and the intron
and intergenic spacer found in the trnL-trnF region
of the chloroplast genome. The results suggest that
there are three major lineages of Araliaceae, and
that these lineages generally correspond with the
centers of diversity for the family. The Aralia and
Asian Palmate groups are centered primarily in
eastern and southeastern Asia, whereas the Poly-
scias-Pseudopanax group is found throughout the
Pacific and Indian Ocean basins. Several poorly
resolved lineages are placed at the base of core
Araliaceae, and the geographic distributions of
these clades are consistent with a hypothesized
rapid radiation of Araliaceae, possibly correlated
with the breakup of Gondwanaland. Comparison
of molecular results with the traditional systems of
classification shows almost no congruence, indicat-
ing that they inadequately reflect phylogenetic
relationships. Moreover, the morphological char-
acters employed in these classifications appear to be
highly homoplastic, and are thus of little utility at
the infrafamilial level.
Key words: Araliaceae, Apiales, ITS, trnL-trnF,
infrafamilial classification.
Introduction
Araliaceae are a medium-sized family of 47
genera and over 1,350 species. They are best
represented in the tropics, where nearly all of
the largest genera are found. For example, five
of the six genera with 50 or more species
(Schefflera, Polyscias, Dendropanax, Oreopan-
ax, and Osmoxylon; excepting only Aralia) are
almost exclusively tropical or subtropical,
Plant Syst. Evol. 245: 1–39 (2004)
DOI 10.1007/s00606-003-0101-3
although several smaller genera are best devel-
oped in the north temperate zone and the
subtropics (e.g. Panax, Hedera, Brassaiopsis,
Macropanax, Gamblea, and Oplopanax). Aral-
iaceae are also largely Old World, with major
centers of diversity in southeastern Asia and
the Pacific and Indian Ocean basins. New
World araliads include many species but
represent only a few genera, most of which
are also well developed in the Old World (e.g.
Aralia, Panax, Oplopanax, Dendropanax,
Pseudopanax). With the inclusion of Sciado-
dendron in Aralia by Wen (2002), Oreopanax is
now the only exclusively New World genus.
Over the past decade, great progress has
been made in resolving the placement of
Araliaceae among the major lineages of the
order Apiales (Henwood and Hart 2001,
Lowry et al. 2001, Plunkett 2001, Plunkett
and Lowry 2001, Plunkett et al. 2003) and in
understanding relationships within and among
closely related genera of Araliaceae (e.g. Wen
and Zimmer 1996, Mitchell and Wagstaff 1997,
Costello and Motley 2001, Eibl et al. 2001,
Plunkett et al. 2001). However, only a single
study (Wen et al. 2001a) has explored rela-
tionships at the intergeneric level across the
whole family. Because that study lacked rep-
resentatives from several key taxa and was
based on only a single molecular marker
(nuclear ITS), the authors did not attempt an
assessment of traditional classification systems.
In the present study, we build on the findings
of Wen et al. (2001a) to test the traditional
systems of intergeneric classifications in Aral-
iaceae and the evolution of the characters upon
which these systems have been based.
Most treatments of Araliaceae have
grouped genera into three or more tribes,
giving greatest weight to petal aestivation and
insertion to delimit infrafamilial groups, some-
times in conjunction with endosperm texture
and stamen number (and more rarely locule
number, inflorescence structure, and/or habit).
Of the several systems considered here, that of
Harms (1894–97) was the simplest and has
been the most-often cited. He divided the
family into three tribes strictly on the basis of
petal aestivation and insertion (Table 1). His
tribe Aralieae was defined as having imbricate
aestivation, whereas Schefflereae and Mackin-
layeae were valvate, and were distinguished
from one another by petal insertion, either
broadly- or narrowly- (clawed) inserted,
respectively. Earlier systems were similar, but
with additional groups recognized (Table 1).
Bentham (1867) presented nearly identical
tribes Aralieae and Mackinlayeae, but the
genera placed by Harms in Schefflereae were
instead treated as tribes Panaceae and Hede-
reae (with smooth vs. ruminate endosperm,
respectively), plus Plerandreae (in which sta-
men number exceeded petal number). In See-
mann’s (1868) system, differences in petal
aestivation were deemed worthy of recognition
at the family rank, with the imbricate Aralia-
ceae and the valvate Hederaceae. The strict
adherence to this system led to the transfer to
Hederaceae of several taxa traditionally re-
ferred to Apiaceae (e.g. Crithmum, Hydrocot-
yle), and the transfer of Mackinlayeae into
Apiaceae.
In the early twentieth century, Calestani’s
(1905) and Viguier’s (1906) systems were more
complex. Calestani (1905) treated Araliaceae as
a subfamily (‘‘Aralineae’’) within Apiaceae, and
used a combination of inflorescence, mating
system, floral, and fruit features to circumscribe
eight araliad tribes (Table 1). The chief effect
was to establish two additional tribes (Osmoxy-
leae and Meryteae), both of which would be
considered monogeneric using present-day gen-
eric circumscriptions. Viguier (1906) divided
Araliaceae into 10 tribes on the basis of five
principal characters (leaf type, pedicel articula-
tion, corolla aestivation, stamen number, and
endosperm texture), supplemented by anatom-
ical features and a few additional morpholog-
ical characters used to distinguish particular
infrafamilial groups. He maintained Mackin-
layeae with the same circumscription as previ-
ous authors, and expanded Plerandreae to
encompass the imbricate Sciadodendron in
addition to the valvate genera previously
included. Rather than mechanically applying a
few characters to define his other tribes, Viguier
2 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
Table 1. Comparisons of the major classification systems of Araliaceae.
Bentham 1867 * Seemann 1868 Harms 1894-97 Calestani 1905 ** Viguier 1906 y Hutchinson 1967 Tseng & Hoo 1982
Tribe Aralieae: Petals
± imbricate, base
broadly inserted
Tribe Mackinlayieae:
Petals valvate,
narrowed at base
Tribe Panaceae:
Petals valvate,
stamens and petals
isomerous,
endosperm smooth
Tribe Hedereae:
Petals valvate,
stamens and petals
isomerous,
endosperm
ruminate
Tribe Plerandreae:
Petals valvate,
stamens numerous
(more than petals),
styles absent, or
shield-like
(‘‘umbonem’’), or
connate in a cone
Family Araliaceae:
Petals imbricate
Tribe Aralieae:
Ovary
5-carpellate
(3-carpellate
by abortion); never
2-carpellate
Tribe Panaceae:
Ovary 2-
carpellate (or 3-
carpellate by
excess); never
4–1 carpellate
Family Hederaceae:
Petals valvate
Tribe Cussonieae:
Stamens and
petal isomerous;
ovary 2(-3)-
carpellate;
endosperm
ruminate
Tribe Horsfieldieae:
Stamens and
petal isomerous;
ovary 2(-3)-
carpellate;
endosperm
smooth
Tribe Hedereae:
Stamens and
petal isomerous;
ovary (3-)5-1-
carpellate;
Tribe Aralieae:
Petals ±
imbricate,
base broadly
inserted
Tribe
Mackinlayeae:
Petals valvate,
narrowed at
the base, with
±long, inflexed
apices
Tribe Schefflereae:
Petals valvate,
base broadly
inserted
Tribe Aralieae:
Inflor. subdivisions
similar; flowers
homomorphic,
seldom dioecious,
distinct; calyx
minute; petals
caducous,
imbricate,
base broadly
inserted;
stamens and
petals isomerous;
endocarp crustose
Tribe Mackinlayeae:
Inflorescence
subdivisions
similar; flowers
homomorphic,
distinct; calyx large;
petals caducous,
valvate, much
narrowed at the
base, inflexed at the
apex; stamens and
petals isomerous;
fruit a drupe
or sterigma
Tribe Myodocarpeae:
Inflor. subdivisions
similar; flowers
homomorphic,
distinct; sepals
large, imbricate;
petals
caducous, imbricate,
the base
Tribe Pseudopanaceae:
Lvs. palmately
compd.; pedicels
usually articulated;
petals valvate;
stamens isomerous;
endosperm smooth
Tribe Polyscieae: Lvs.
pinnately compd.;
pedicels articulated;
stamens isomerous;
endosperm smooth
or ruminate by
penetration
Tribe Schefflerineae:
Lvs. palmately
compd. or lobed;
pedicels
unarticulated;
stamens isomerous;
endosperm smooth
Tribe Hedereae:
Stamens isomerous;
ovary 2-1-carpellate;
endosperm ruminate
by digestion
Tribe Myodocarpeae:
Infl. panicle of
umbels; pedicels
articulated; petals
imbricate; stamens
isomerous; ovary
2-carpellate; fruits
with oil vesicles;
endosperm smooth
Tribe Cussonieae:
Infl. racemes,
spikes, or
panicles of
racemes or
spikes
Tribe
Anomopanaceae:
Infl. panicles of
cymuless
All the following:
Infl. of simple
umbels, compound
umbels, or heads,
sometimes further
arranged in
panicles or
racemes:
Tribe Plerandeae:
Stamens more
numerous than
petals; petals
valvate
Tribe Aralieae:
Stamens and
petals isomerous;
petals imbricate
Tribe Mackinlayeae:
Stamens and
petals isomerous;
petals valvate and
clawed at base
Tribe Panaceae:
Stamens and
petals isomerous;
Tribe Plerandreae:
Woody; lvs.
simple, margin
entire or toothed,
or palmately
divided to
compound; petals
5(-10), free
or rarely
calyptrate,
valvate; stamens
isomerous (or
rarely more
numerous than
petals); ovary (1-)
2-5(-1)-
carpellate.
Tribe
Tetraplasandreae:
Woody; lvs.
pinnately
compound, rarely
simple with
pinnatifid divisions;
petals 5(-16), free
(-calyptrate),
valvate; stamens
isomerous
(–6 times
number of petals);
ovary (1-)2-5(-18)-
carpellate
Tribe Mackinlayeae:
Woody; lvs. simple,
palmately
compound or
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 3
Table 1 (continued)
Bentham 1867 * Seemann 1868 Harms
1894-97
Calestani 1905 ** Viguier 1906 y Hutchinson 1967 Tseng & Hoo 1982
endosperm
ruminate
Tribe
Pseudopanaceae:
Stamens and
petal isomerous;
ovary (3-)5-1-
carpellate;
endosperm
smooth
Tribe Plerandreae:
Stamens
2-many times as
numerous as
petal-s; ovary
5-1-carpellate
sometimes narrowed;
stamens and petals
isomerous; endocarp hard,
with oil-bearing vesicles
Tribe Schefflereae:
Inflorescence subdivisions
similar; flowers homomorphic,
polygamous or hermaphroditic,
distinct; calyx mostly minute;
petals caducous, valvate,
base broadly inserted; stamens
and petals isomerous; endocarp
crustose or cartilaginous
Tribe Hedereae:
Inflor. subdivisions similar;
flowers homomorphic,
polygamous or hermaphroditic,
distinct; calyx minute, petals
caducous, valvate or subimbricate,
base broadly inserted,
stamens and petals
isomerous; fruit a berry
Tribe Plerandreae:
Inflor. subdivisions similar;
flowers distinct; calyx minute;
petals caducous, valvate,
± joined to the apex; stamens
numerous; endocarp crustose
Tribe Osmoxyleae:
Inflor. of three rays, the middle
one carrying a sterile,
bacciform flower, the others
umbels or capitula;
flowers distinct; calyx minute;
petals caducous, valvate, joined
at the base into one
gametopetalous corolla,
stamens numerous;
endrocarp boney
Tribe Plerandreae:
Subtribe
Plerandrinae:
Stamens and carpels
1 (> petals);
pedicels
unarticulated;
endosperm smooth
Subtribe
Reynoldsinae:
Fls. > 5-merous; lvs.
pinnately compd.;
pedicels
unarticulated;
endosperm smooth
Tribe Meryteae:
Monocaulous or
sparsely branched;
lvs. simple w/
thickened joints on
midvein; fls.
unisexual in heads
Tribe Mackinlayeae:
Fls. 5-6-merous;
petals clawed at base,
valvate; stamens
isomerous;
endosperm smooth
Tribe Panaceae: Herbs
w/ 1 verticel of
palmately compd.
lvs.; fls. 5-merous;
pedicels articulated;
ovary 2-3 carpellate
Tribe Eremopanaceae:
Pedicels articulated;
ovary 1-carpellate
petals valvate,
broadly
inserted at
base; endosperm
smooth
Tribe Hedereae:
Stamens and
petals isomerous;
petals valvate,
broadly inserted
at base;
endosperm
ruminate
pinnately
compound; petals
5, free, clawed at
base, valvate;
stamens 5; ovary
2(-4)-carpellate
Tribe Aralieae:
Woody, rarely
herbaceous; lvs.
pinnately to
multi-pinnately
compound, rarely
simple with
pinnatifid
divisions; petals
5(-12), free,
imbricate;
stamens
isomerous; ovary
(2-)5(-12)-carpellate
Tribe Panaceae:
Woody or
herbaceous; lvs.
simple (divided or
not) or palmately
compound; petals 5,
free, imbricate;
stamens 5; ovary
2(-4)-carpellate
4 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
relied on various combinations of features to
delimit them (Table 1).
More recently, only two formal systems
have been widely cited. Hutchinson’s (1967)
classification was the first to use inflorescence
structure to define tribes (Table 1), but his
rigid interpretation of these characters led to
many difficulties, including the segregation of
clearly-related species into distinct genera to
accommodate his tribes (e.g. Parapentapanax
Hutch. was placed in tribe Cussonieae,
whereas Pentapanax was assigned to Aralieae).
Overall, however, Hutchinson’s (1967) system
had only a minor effect in establishing two new
tribes (Cussonieae and Anomopanceae), leav-
ing the remaining five much as Bentham (1867)
had circumscribed them a century before.
Tseng and Hoo (1982) relied largely on petal
aestivation and leaf shape, with two imbricate
tribes (palmate Aralieae and pinnate Pana-
ceae), and three valvate tribes (palmate Pler-
andreae, pinnate Tetraplasandreae, plus a
third tribe, Mackinlayeae, which included both
pinnate and palmate taxa with clawed petals).
This system helped to codify many of the ideas
proposed earlier by Eyde and Tseng (1971),
reflecting their fundamental division of the
family into pinnate vs. palmate groups.
Araliaceae have often been regarded as a
paraphyletic group in relation to Apiaceae, the
other major family of the order Apiales (e.g.
Calestani 1905, Thorne 1973, Judd et al. 1994).
However, recent studies of phylogenetic rela-
tionships at the ordinal level have helped to
identify a monophyletic group corresponding
largely to the traditional circumscription of
Araliaceae, sometimes referred to as ‘‘core
Araliaceae’’ or Araliaceae s. str. (Plunkett
et al. 1997, Plunkett 2001, Plunkett and Lowry
2001, Wen et al. 2001a). These studies suggest
that two groups of genera should be excluded
from core Araliaceae, tribes Mackinlayeae and
Myodocarpeae. Seemann (1868) treated Mack-
inlayeae as distinct from both Araliaceae and
Hederaceae, but most authors recognized this
tribe as a distinctive group within Araliaceae,
generally including Mackinlaya, Apiopetalum,
and the little known Pseudosciadium. Only
Table 1 (continued)
Tribe Meryteae:
Inflor. subdivisions similar;
dioecious, flowers dimorphic;
female flowers with
connate ovaries; calyx
lacking; petals continuous
with the ovary, persistent,
free, valvate; male flowers
with isomerous stamens;
endocarp crustose
* Bentham’s (1867) tribes were originally published as ‘‘series;’’
** Calestani (1905) treated Araliaceae as a subfamily (‘‘Aralineae’’) within Apiaceae;
Viguier’s (1906) tribes are presented with corrected suffices
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 5
Calestani (1905) and Viguier (1906) recognized
Myodocarpeae as a distinct group, defined as
araliads with articulated pedicels, imbricate
corollas, bicarpellate ovaries, and fruits with
oil vesicles. This tribe included only Myodo-
carpus, Delarbrea, and Porospermum (now
treated under Delarbrea; see Lowry 1986).
With the discovery and characterization of
fruiting specimens of the monotypic Pseudos-
ciadium, this genus was transferred from
Mackinlayeae to Myodocarpeae (Lowry
1986). Molecular studies based on a variety
of markers (e.g. rbcL, matK, ITS, 26S DNA
sequences; Plunkett et al. 1997, Plunkett and
Lowry 2001, G. Chandler and Plunkett un-
publ.) suggest that the exclusion of Mackin-
layeae and Myodocarpeae from Araliaceae is
necessary to render the family monophyletic,
reducing the number of genera to 41 (see
Table 2). Formal recognition of these two
groups as distinct families (Mackinlayaceae
and Myodocarpaceae) has recently been pro-
posed (Doweld 2001).
Within core Araliaceae, the phylogenetic
study of Wen et al. (2001a) made significant
progress in exploring intergeneric relation-
ships. Using nuclear ITS sequence data from
a large sample of araliaceous taxa (70 species in
40 genera), their analysis helped to demarcate
two major clades within the family, informally
labeled the ‘‘Eleutherococcus-Dendropanax-
Schefflera group’’ and the ‘‘Aralia-Polyscias-
Pseudopanax group.’’ In addition to these two
large clades, several lineages were poorly
resolved at the basal branches of the tree. To
test and refine the relationships suggested by
Wen et al. (2001a), we have assembled a
complementary data set based on a chloroplast
sequence region: trnL-trnF (which includes
both the trnL intron and the trnL-trnF
intergenic spacer). In addition, we have sub-
stantially increased our sampling to include a
near-comprehensive coverage of genera, along
with many additional species from large groups
that appear to be poly- or paraphyletic (e.g.
Schefflera, Polyscias, Gastonia). Several of the
genera sampled by Wen et al. (2001a) are now
placed in synonymy under other genera (viz.,
Brassaia, Dizygotheca, and Tupidanthus under
Schefflera; Euaraliopsis under Brassaiopsis;
Evodiopanax under Gamblea; and Pentapanax
under Aralia). Our current ‘‘working list’’
includes 47 genera of Araliaceae s. lat.
(Table 2). From this list, Stilbocarpa can be
removed (it belongs to the ‘‘Azorella clade’’ of
Apiaceae; Mitchell et al. 1999), as can the five
genera of tribes Mackinlayeae and Myodocar-
peae, leaving 41 genera in core Araliaceae
(Table 2), 33 of which were sampled by Wen et
al. (2001; rather than the 40 listed therein). Of
the eight genera not surveyed by Wen et al.
(2001a), we have been able to sample represen-
tatives from four genera (Harmsiopanax,
Merrilliopanax, Motherwellia, and Seemann-
aralia), leaving unsampled only Chengiopanax
(2 species) and three monotypic genera, Anak-
asia (A. simplicifolia Philipson), Woodburnia
(W. penduliflora Prain), and Cromapanax
(C. lobatus Grierson). We are now able to
re-evaluate the relationships suggested by Wen
et al. (2001a) using this expanded ITS data set
together with the newly derived trnL-trnF data
set. The resulting phylogenetic hypotheses are
used to address fundamental questions of
relationships in core Araliaceae, including
(1) the conformance of traditional classification
systems to phylogeny, (2) the utility of charac-
ters traditionally considered taxonomically
important at infrafamilial levels, and (3)
basic patterns of phylogeny and biogeography
within and among the major clades of the
family.
Materials and methods
One-hundred and eight accessions (representing
107 species) were sampled for both ITS and trnL-
trnF sequences (Table 3). Sources of plant tissue
and of previously published sequences are provided
in Table 3, along with GenBank accession numbers
for all sequences. This sample includes representa-
tives from 37 of the 41 genera recognized in Table 2
(excepting only Anakasia, Chengiopanax, Croma-
panax, and Woodburnia). Following recent treat-
ments, Pentapanax and Sciadodendron are now
included in Aralia (Wen 1993, 2001), Grushvitzkya
is now placed within Brassaiopsis (Wen et al. 2003),
6 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
Table 2. The 47 genera currently recognized in Araliaceae s. lat., including the six genera that molec ular
data suggest should be excluded from Araliaceae s. str. to render a monophyletic group
Genus # spp. Geographic distribution
Genera excluded from Araliaceae s. str.
Apiopetalum 2 New Caledonia
Delarbrea (incl. Porospermum) 6 New Caledonia, Vanuatu, Malesia (to Timor);
NE. Australia
Mackinlaya (incl. Anomopanax) 5 New Guinea, NE. Australia
Myodocarpus 10 New Caledonia
Pseudosciadium 1 New Caledonia
Stilbocarpa 3 New Zealand & suban tarctic islands
Genera included in Araliaceae s. str.
Anakasia 1 New Guinea
Aralia (incl. Coudenbergia, Hunaniopanax,
Megalopanax, Parapentapanax, Pentapanax,
Sciadodendron)
68 Asia; S., C. & N. America; West Indies
Arthrophyllum (incl. Eremopanax) 31 Nicobar and Andamon Islands Is., Malesia,
New Caledonia
Astrotricha 11 Australia
Brassaiopsis (incl. Wardenia, Euaraliopsi s,
Grushvitzkya)
25 China, Indo-Malesia
Cephalaralia 1 E. Aust ralia
Cheirodendron 6 Hawaii; Marquesas
Chengiopanax 2 China, Japan
Cromapanax 1 Bhutan
Cuphocarpus 5 Madagascar
Cussonia 25 trop., S. & E. Africa, Comoros, Yemen
Dendropanax (incl. Gilibertia) 70 trop. America, trop & subtrop Asia
Eleutherococcus (incl. Acanthopanx ) 30 E. Asia
Fatsia (incl. Diplofatsia, Boninofatsia) 3 Japan; Taiwan
Gamblea (incl. Evodiopanax) 4 E. Himalayas, China, Indo-China
Gastonia (incl. Indokingia, Peekeliopanax) 9 Madagascar; Comoros; Mascarenes; Seychelles;
Malesia; Australia
Harmsiopanax 3 W. Malesia
Hedera 15 Europe to N. Africa; trop & subtrop Asia
Heteropanax 7 India; China; Indo-China; Malesia
Kalopanax 1 China; Japan; Korea; Russian Far E
Macropanax (incl. Hederopsis) 14 China; Indo-Malesia
Merrilliopanax 4 NE India; SW China
Meryta (incl. Schizomeryta, Strobilopanax,
Botryomeryta)
25 Pacific islands
Metapanax 2 China; Vietnam
Motherwellia 1 NE Aust ralia
Munroidendron 1 Hawaii
Oplopanax (incl. Echinopanax) 3 E. Asia; W. N. America
Oreopanax 80 trop & subtrop. Ame rica; W est Indies
Osmoxylon (incl. Boerlagiodendron) 50 Malesia; Taiwan; Micronesia; Vanuatu
Panax 15 temp. N. America & E. Asia; Himalayas;
Indo-China
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 7
and Schefflera was broadened to encompass both
Plerandra and Tupidanthus (Frodin 1975, Lowry
1989, Lowry et al. 1989) (see Table 2). In the
present study, we employ species names from these
segregate genera either because formal taxonomic
transfers are lacking (e.g. for the species of
Plerandra) or because these species are more widely
known under their older names [e.g. Tupidanthus
calyptratus J. D. Hook. & Thomson rather than
Schefflera pueckleri (K. Koch) Frodin]. For most
large genera (e.g. Schefflera, Polyscias, Aralia),
multiple accessions were sampled representing
species from across most of their geographic and/
or taxonomic breadth. Four species from the
excluded tribes Mackinlayeae (Mackinlaya macros-
ciadea, Apiopetalum velutinum) and Myodocarpeae
(Delarbrea paradoxa, Myodocarpus fraxinifolius)
were used as outgroups (discussed above). All
molecular studies to da te suggest that the genus
Hydrocotyle (traditionally placed in Apiaceae) is
sister to or included within core Araliaceae, and
thus we analyzed sequences from three species of
this genus (H. verticillata for ITS, H. bowlesioides
for trnL-trnF, and H. vulgaris for both markers).
Total DNAs were extracted using the CTAB
method of Doyle and Doyle (1987) or using the
DNeasy Plant Mini kit (QIAGEN Inc.). Many ITS
sequences were already available from our earlier
study (Wen et al. 2001a), but for those that were
newly derived (Table 3), protocols follow Wen
et al. (1998, 2001a) or Plunkett et al. (2001) in
amplifying and sequencing the entire ITS region,
including the two spacers (ITS1 and ITS2) and the
intervening 5.8S rRNA gene. Primers ITS5 (for-
ward) and C26A (reverse) were generally used for
both amplification and sequencing, although some
taxa required additional or alternative primers,
including ITS4 (reverse) or internal primers ITS3
(forward) and the complement of N5.8S (reverse)
[= N5.8S-r: 5¢ TGC GTT CAA AGA CTC GAT
–3¢]. For trnL-trnF, the protocols of Eibl et al.
(2001) were followed, in which both the trnL intron
and the trnL-trnF intergenic spacer (IGS) were
amplified as a single amplicon using the ‘‘c’’ and
‘‘f’’ primers of Taberlet et al. (1991), althoug h
primer ‘‘f’’ has been modified for Araliaceae as
reported in Eibl et al. (2001) [modified primer ‘‘f’’
=5¢–AAC TGG TGA CAC GAG GAT TTT
CAG–3¢]. For some taxa, internal primers ‘‘d’’ and
‘‘e’’ (Taberlet et al. 1991) were also used for
amplification and/or sequencing. In both the ITS
and trnL-trnF regions, most mutations were base
substitutions, which facilitated manual alignment.
The ITS and trnL-trnF data matrices were each
analyzed separately using both maximum parsi-
mony (MP) and Bayesian inference (BI). In addi-
tion, a third combined matrix was also assembled
to assess the congruence (and thus combinability)
of the separate data sets. Results from the ILD or
partition homogeneity test (Farris et al. 1995, as
Table 2 (conti nued)
Genus # spp. Geographic distribution
Polyscias (incl. Bonnierella, Botryopanax,
Eupteron, Gelibia, Kissodendron,
Nothopanax, Palmerovandenbroekia,
Sciadopanax, Tieghemopanax)
150 Paleotropics (Africa, Indian Ocean basin,
Malesia; Australia; S. Pacific islands)
Pseudopanax (incl. Neopanax) 6 New Zealand, Tasmania, Chile, Argentina
Raukaua 3 New Zealand
Reynoldsia 5 Pacific islands
Schefflera (incl. Agalma, Brassaia, Crepinella
Didymopanax, Dizygotheca, Geopanax,
Neocussonia, Octotheca, Plerandra,
Sciadophyllum, Tupidanthus)
650+ Pantropical and subtropical; temp. China,
Japan, New Zea land
Seemannaralia 1 E. S. Africa
Sinopanax 1 Taiwan
Tetrapanax 1 Taiwan; W. & S. China
Tetraplasandra (incl. Pterotropia, Dipanax) 7 Hawaii
Trevesia (incl. Plerandropsis) 8 China; Indo-Malesia
Woodburnia 1 N. Myanmar (Burma)
8 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
Table 3. Names, sources, and GenBank accession numbers of ITS and trnL-trnF sequences from Ara-
liaceae and outgroup taxa. Herbarium acronyms follow Holmgren et al. (1990). Voucher and source for
ITS and trnL-trn F from each sample are identical unless otherw ise noted
Species Voucher/Origin
GenBank Accessions
ITS trnL-trnF
Outgroup
Apiopetalum velutinum Baill. P. P. Lowry 4700 (MO);
New Caledonia.
AF229742 AY393698
Delarbrea paradoxa Vieill. ITS: P. P. Lowry 4766 (MO).
trnL-trnF: P. P. Lowry 4791
(MO); New Caledonia.
AF229750 AF382152
Hydrocotyle bowlesioides Mathias
& Constance
G. M. Plunkett 1373 (WS);
cultivated, Univ. of California
Botanical Garden.
n/a AY393699
Hydrocotyle verticillata Thunb. D. M. E. Ware 10036 (WILLI);
Virginia, USA.
AY389025 AY393700
Hydrocotyle vulgaris L. Valiejo-Roman et al., 1998. AF077895 n/a
Mackinlaya macrosciadea (F. Muell.)
F. Muell.
G. M. Plunkett 1365 (WS);
Queensland, Australia.
AF229744 AY393701
Myodocarpus fraxinifolius Brongn.
& Gris
P. P. Lowry 4803 (MO);
New Caledonia.
AY389026 AY393702
Araliaceae s. str.
Aralia apioides Hand.-Mazz. J. Wen 1149 (CS); Yunnan, China. U66704 AY393703
Aralia chinensis L. J. Wen 1136 (CS); Hainan, China. AF242256 AY393704
Aralia excelsa (Griseb.) J. Wen
[= Sciadodendron excelsum Griseb.]
F. Chi ang s.n. (CS); Mexico. AF242231 AY393780
Aralia humilis DC. J. Wen 4974 (CS), Arizona, USA. AF242230 AY393705
Aralia leschenaultii (DC.) J. Wen
[= Pentapanax fragrans (D. Don ) Ha]
J. Wen 4907 (CS); Godawari, Nepal. AY394569 AY393753
Aralia kingdon-wardii J. Wen, Lowry
& Es ser [= Pentapanax trifoliatus
Feng]
J. Wen 5069 (F); Yunnan, China. AY394570 AY393755
Aralia nudicaulis L. ITS: J. Wen 849 (A); North Carolina,
USA. trnL-trnF: J. Wen 1535 (CS);
Maryland, USA.
U41674 AF382157
Aralia plumosa Li [= Pentapanax
plumosus (Li) Shang]
J. Wen 3047 (CS); Yunnan, China. AF242255 AY393754
Aralia scopulorum Brandg. J. Wen 565 (OS); Baja California,
Mexico.
U66927 AY393706
Aralia spinosa L. ITS: J. Wen & H. Dong 976 (A);
Georgia, USA. trnL-trnF:
G. M. Plunkett 1371 (WS);
cultivated, Washington, USA.
U66928 AY393707
Arthrophyllum diversifolium Blume P. P. Lowry 5288 (MO); cultivated,
Bogor Botanic Garden.
AY389027 AY393708
Arthrophyllum ‘‘mackeei’’ Lowry,
ined.
P. P. Lowry 4670 (MO);
New Caledonia.
U63182 AF382158
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 9
Table 3 (conti nued)
Species Voucher/Origin GenBank Accessions
ITS trnL-trnF
Astrotricha pterocarpa Benth. G. M. Plunkett 1527 (MO),
Queensland, Australia.
U63189 AY393709
Astrotricha sp. nov. (‘‘Isabella Falls’’) G. M. Plunkett 1551 (MO);
Queensland, Australia.
U63190 AY393710
Brassaiopsis glomerulata Regel P. P. Lowry 4839 (MO); Vietnam. AY256901 AY393711
Brassaiopsis grushvitzkyi J. Wen,
Lowry & H.-T. Nguyen
[= Grushvitzkya stellata
N. T. Skvorts ova & L. V. Averyanov]
J. Wen 5864 (F); Ha Giang, Vietnam. AF551728 AY393729
Brassaiopsis hainla Seem. J. Wen 4905 (CS); Godawari, Nepal. AY389028 AY393712
Cephalaralia cephalobotrys Harms G. M. Plunkett 1519 (MO);
Queensland, Australia.
AF229762 AY393713
Cheirodendron platyphyllum
(Hook. & Arn.) Seem.
D. K. Harder 4072 (MO); Hawaii. AY389029 AY393714
Cheirodendron trigynum (Gaud.)
A. Heller
L. A. Johnson 92-110 (WS); Hawaii. AY389030 AY393715
Cuphocarpus aculeatus Decne. & Planch. P. P. Lowry 5013 (MO); Madagascar. AF229737 AF393920
Cussonia holstii Harms ex Engl. P. P. Lowry 4986 (MO); Tanzania. AY389031 AY393716
Cussonia paniculata E. May. P. B. Phili pson, 5263 (MO);
South Africa.
AY389032 AY393717
Cussonia thyrsifolia Thunb. P. B. Phili pson 5110 (MO);
South Africa.
AF229765 AY393718
Dendropanax arboreus (L.)
Decne. & Planch.
F. Chi ang s.n. (CS), Los Tuxtlas,
Mexico. trnL-trnF: G. M. Plunkett
1352 (WS); cultivated, Univ. of
California Botanical Garden.
AY389033 AY393719
Dendropanax trifidus (Thunb.) Makino J. Wen 2461 (CS); Tokyo, Japan. AF242238 AY393720
Eleutherococcus sessiliflorus
(Rupr. & Maxim.) S.Y. Hu
J. Wen 3131 (CS); Liaoning, China AF242227 AY393721
Eleutherococcus setchuanensi s Nakai ITS: Acc#: 536 84B; cultivated,
Arnold Arboretum. trnL-trnF:
Acc#: 963 84A cultivated,
Arnold Arboretum.
AF242252 AY393722
Eleutherococcus sieboldianu s (Makino)
Koidz.
Acc#: 463 78J; cultivated, National
Arboretum, Washington, D.C.
U63184 AY393723
Fatsia japonica (Thunb.)
Decne. & Planch.
ITS: A. Mitchell s.n. (CHR 502463);
cultivated, Christchurch, New Zealand.
trnL-trnF: G. M. Plunkett 1320 (WS );
cultivated, New York Botanical
Garden.
U63193 AY393724
Fatsia polycarpa Hayata P. P. Lowry 4968 (MO); Taiwan. AY389034 AY393725
Gamblea ciliata C. B. Clarke va r.
evodiifolia (Franch.) C.-B. Shang,
Lowry & Frodin
P. P. Lowry 4865 (MO); Vietnam. AF229766 AY393726
Gamblea ciliata C. B. Clarke va r.
evodiifolia (Franch.) C.-B. Shang,
Lowry & Frodin
J. Wen 3050 (CS); Yunnan, China. AF242228 AY393727
10 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
Table 3 (conti nued)
Species Voucher/Origin GenBank Accessions
ITS trnL-trnF
Gastonia rodriguesiana Marais Acc#: 662-86.06150, cultivated at
Royal Botanic Gardens, Kew.
AF382162 AY393728
Gastonia spectabilis (Harms) Philipson ITS: P. P. Lowry 5257 (MO);
West Papua, Ind onesia. trnL-trnF:
G. M. Plunkett 1537 (MO);
Queensland, Australia.
AY389036 AF382164
Harmsiopanax ingens Philipson P. P. Lowry 5266 (MO); Irian Jaya. AY389038 AY393730
Hedera helix L. ITS: J. Wen 2481 (CS); cultivated,
Colorado, USA. trnL-trnF:
G. M. Plunkett 1476 (VCU),
Virginia, USA.
AF242241 AY393731
Heteropanax fragrans (Roxb.) Seem. ITS: J. Wen 2213 (CS); Guangdong,
China. trnL-trnF: P. P. Lowry 4927
(MO); Vietnam.
AF242242 AY393732
Kalopanax septemlobus (Thunb.) Koidz. ITS: E. H. Wilson 1680 ; cultivat ed,
Royal Botanic Garden Kew.
trnL-trnF: G. M. Plunkett 1328 (WS );
cultivated, New York Botanical
Garden.
U63187 AY393733
Macropanax dispermus (Bl.) Ktze. P. P. Lowry 4940 (MO); Vietnam. AF229767 AY393734
Macropanax undulatus Seem. P. P. Lowry 4947 (MO); Vietnam. AY389039 AY393735
Merrilliopanax chinensis Li J. Wen 5068 (F); Yunnan, China. AY389040 AY393736
Meryta balansae Baill. P. P. Lowry 4733 (MO);
New Caledonia.
U63195 AY393737
Meryta ‘‘pedunculata’’ Lowry, ined. P. P. Lowry 4756 (MO);
New Caledonia.
U63194 AY393738
Meryta tenuifolia A. C. Sm ith P. P. Lowry 5516 (MO); Fiji. AY389041 AY393739
Metapanax davidii (Franch.) Frodin ex
J. W en & Frodin
J. Wen 1409 (CS); Hubei, China. AF242233 AY393740
Metapanax delavayi (Franch.) Frodin
ex J. Wen & Frodin
J. Wen 1217 (CS); Yunnan, China. AF242232 AY393741
Motherwellia haplosciadea F. Muell. G. M. Plunkett 1515 (MO);
Queensland, Australia.
AY389042 AY393742
Munroidendron racemosum
(C. N. Forbes) Sherff
G. M. Plunkett 1342 (WS); cultivated,
Missouri Botanical Garden.
AF229738 AY393743
Oplopanax elatus Nakai J. Wen 5407 (F), Jilin, Chi na. AY389043 AY393744
Oplopanax horridus Miq. M. M cBroom 5 (CS); Montana, USA. AY389044 AY393745
Oreopanax sanderianus Hemsl. G. M. Plunkett 1343 (WS); cultivated,
Missouri Botanical Garden.
AF242229 AY393746
Osmoxylon geelvinkianum Becc. G. M. Plunkett 1489 (MO); cultivated,
Flecker Botanic Garden..
AF229727 AY393747
Osmoxylon novo-guineense (Scheff.)
Becc.
D. Lorence 8157 (PTBG); cultivated,
National Tropical Botanical Garden.
AF229726 AY393748
Osmoxylon pectinatum (Merr.)
Philipson
Y.-Y. Huang 756 (HAST);
Green Island, Taiwan.
AY389045 AY393749
Panax pseudoginseng Wall. Hasebe 2218 (US); cultivated, Tokyo
Botanical Garden, originally from
Nepal.
U41693 AY393750
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 11
Table 3 (conti nued)
Species Voucher/Origin GenBank Accessions
ITS trnL-trnF
Panax quinquefolius L. J. Wen 1521, North Carolina, USA U41688 AY393751
Panax trifolius L. Kramer & Kramer s.n. (CS),
Ohio, USA.
U41698 AY393752
Plerandra insolita A. C. Smith P. P. Lowry 5503 (MO); Fiji. AY389047 AY393756
Plerandra vitiensis Baill. P. P. Lowry 5505 (MO); Fiji. AY389048 AY393757
Polyscias ‘‘abrahamiana’’ Lowry, ined. ITS: J.-N. Labat 3064 (P);
Madagascar. trnL-trnF:
J.-N. Labat 3065 (P); Madagascar.
AF229686 AF382148
Polyscias australiana (F. v. M.)
Philipson
G. M. Plunkett 1500 (MO);
Queensland, Australia.
AF229688 AF396410
Polyscias cf. cumingiana
(C. Presl) Fern.-Vill.
G. M. Plunkett 1357 (WS); cultivated,
Honolulu Botanical Garden.
AF242246 AY393758
Polyscias ‘‘crenata’’ (Pancher & Sebert)
Lowry, ined.
P. P. Lowry 4664 (MO);
New Caledonia.
AF229694 AF382169
Polyscias elegans Harms G. M. Plunkett 1495 (MO);
Queensland, Australia.
AF229698 AF382175
Polyscias joskei L. S. Gibbs A. C. Smith 7584 (US); Fiji. AF382944 AF38217 7
Polyscias lecardii (R. Vig.) Lowry ined. P. P. Lowry 4754 (MO);
New Caledonia.
AF229701 AF382178
Polyscias macguillivrayi Harms G. M. Plunkett 1536 (MO);
Queensland, Australia.
AF229707 AF396412
Polyscias mollis (Benth.) Harms G. M. Plunkett 1507 (MO);
Queensland, Australia.
AF229705 AF396413
Polyscias ‘‘orientalis’’ Lowry, ined. G. E. Schatz 3925 (MO); Madagascar. AF229708 AF393905
Polyscias sessiliflora Marais P. P. Lowry 4981 (MO); Re
´
union. AF229717 AF393912
Pseudopanax arboreus (Murr.)
Philipson
ITS: A. Mitchell s.n. (CHR 500663);
cultivated, Christchurch,
New Zealand. trnL-trnF:
G. M. Plunkett 1353 (WS);
cultivated, University of California
Botanical Garden.
U63165 AY393759
Pseudopanax crassifolius
(A. Cunn.) C. Koch
A. Mitchell s.n. (CHR 500661);
cultivated, Christchurch,
New Zealand.
U63168 AY393760
Pseudopanax ferox Kirk ITS: A. Mitchell s.n. (CHR 500660);
cultivated, Christchurch,
New Zealand. trnL-trnF: R. J. Bayer
NZ-01003 (CANB); New Zealand.
U63172 AY393761
Pseudopanax laetus (Kirk) Philipson A. Mitchell s.n. (CHR 500665);
cultivated, Christchurch,
New Zealand.
U63176 AY393762
Pseudopanax linearis (Hook. f.)
C. Koch
P. Heenan s.n. (CHR 500668);
cultivated, Christchurch,
New Zealand.
U63178 AY393763
12 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
Table 3 (conti nued)
Species Voucher/Origin GenBank Accessions
ITS trnL-trnF
Raukaua anomalus (Hook.)
A. D. Mitc hell, D. Frodin &
M. Heads
ITS: Acc# 18225 (CHR 500664);
cultivated, Christchurch,
New Zealand. trnL-trnF:
A. Mitchell s.n.
(CHR 527931); cultivated,
Christchurch, New Zealand.
U63164 AY393764
Reynoldsia sandwicensis A. Gray G. M. Plunkett 1359 (WS);
cultivated, Honolulu Botanical
Garden.
AF229739 AY393765
Schefflera actinophylla (Endl.) Harms ITS: G. M. Plunkett 1535 (MO);
Queensland, Australia. trnL-trnF:
G. M. Plunkett 1316 (WS);
cultivated, New York Botanical
Garden.
AF242245 AY393766
Schefflera baillonii R. Vig. P. P. Lowry 4782 (MO);
New Caledonia.
AF396420 AF396415
Schefflera candelabrum Baill. P. P. Lowry 4726 (MO);
New Caledonia.
AF229728 AY393767
Schefflera digitata J. R. Forst.
& G. Forst.
A. Mitchell s.n. (CHR 485642);
cultivated, Christchurch,
New Zealand.
U63188 AY393768
Schefflera elegantissima
(Veitch ex Mast.) Lowry & Frodin
P. P. Lowry 4715 (MO);
New Caledonia.
AY389050 AF396416
Schefflera gabriellae Baill. ITS: P. P. Lowry 4648 (MO);
New Caledonia. trnL-trnF:
P. P. Lowry 4792 (MO);
New Caledonia.
AF229731 AY393769
Schefflera impressa Harms J. Wen 5050 (F); Yunnan, China. AY389051 AY393770
Schefflera lukwangulensis (Tennant)
Bernardi
P. B. Phili pson 5163 (MO); Tanzania. AY389052 AY393771
Schefflera myriantha Drake N. A. Mwangulango 501 (MO);
Tanzania.
AY389053 AY393772
Schefflera pseudocandelabra R. Vig. P. P. Lowry 4795 (MO ),
New Caledonia.
AY389054 AY393773
Schefflera sp. D. Neill 11935 (MO); Ecuador. AY389055 AY393774
Schefflera sp. L. Al lorge 1087; French Guyana. AY389056 AY393775
Schefflera trevesioides Harms P. P. Lowry 4920 (MO); Vietnam.
[listed as S. hypoleucoides in
Wen et al. (2001a)]
AF229732 AY393776
Schefflera umbellifera Baill. P. P. Lowry 4808 (MO); Zimbabwe. AF242244 AY393777
Schefflera vieillardii Baill. ITS: P. P. Lowry 4747 (MO);
New Caledonia. trnL-trnF:
G. McPherson 17725 (MO);
New Caledonia.
AY389059 AY393778
Schefflera yunnanensis Li J. Wen 5028 (F); Yunnan, China. AY389060 AY393779
Seemannaralia haplosciadea (Seem.)
R.Vig.
P. B. Phili pson 5471 (MO);
South Africa.
AY389062 AY393781
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 13
implemented by PAUP*) indicated that the two
separate data sets were congruent, and so the
combined data set was also analyzed using both
MP and BI. The MP analyses were performed using
PAUP* (version 4, Swofford 2002), treating align-
ment gaps as missing data but scorin g 17 insertion/
deletion events (indels) as binary characters. In
early trials of the analyses, it was determined that
several large polytomies yielded many thousands of
shortest-length trees from each data matrix, quickly
exhausting computer memory. To explore the
greatest number of possible tree topologies, the
approach described in Plunkett et al. (2001, see also
Plunkett et al. 1997) was taken for each MP
analysis, in which an initial search of 1,000
replicates (with random addition, ACCTRAN
optimization, and MULPARS in effect) was exe-
cuted, saving no more than 100 trees per replicate.
The strict consensus from this analys is was then
used as a topological constraint for an additional
series of searches (1,000 replicates, saving no more
than 1,000 trees per replicate, and aborting each
replicate after having met this limit); only trees not
agreeing with the original constraint were saved.
Confidence in individual clades for the MP trees
was assessed using bootstrap analyses (Felsenstein
1985) in which 1,000 replicates were performed,
saving no more than 100 trees per replicate. Fitness
indices and nucleotide compositions for each data
set were assessed using MacClade 4 (Maddison and
Maddison 2000) and/or PAUP*.
The large size of the data matrix (108
terminals) made maximum likelihood analyses
impractical. Because Bayesian analysis combines a
likelihood approach together with confidence test-
ing in the form of posterior probabilities (Huelsen-
beck and Ronquist 2001), BI was employed as an
alternative to parsimony. Trees were constructed
using BI as implemented by the computer program
MrBayes (Huelsenbeck and Ronquist 2001). After
an initial run of 500,000 generations, it was
determined that stationarity was met by 50,000
generations. Thereafter, Bayesian analyses were
performed using 400,000 generations (saving trees
every tenth generation), employing three simulta-
neous chains for the Markov Chain Monte Carlo
analysis (the maximum possible when using over
200 MB of memory on a G4 Macintosh computer).
Table 3 (conti nued)
Species Voucher/Origin GenBank Accessions
ITS trnL-trnF
Sinopanax formosanus (Hayata) Li P. P. Lowry 4967 (MO); Taiwan. AF229768 AY393782
Tetrapanax papyriferus (Hook.)
K. Koch
ITS: A. Mitchell s.n. (CHR 502422);
cultivated, Christchurch,
New Zealand. trnL-trnF:
G. M. Plunkett 1344 (WS);
cultivated, Missouri Botanical
Garden
U63192 AY393783
Tetraplasandra oahuensis Harms D. Lorence 8158 (PTBG); cultivated,
National Tropical Botanical Garden.
AF229740 AY393784
Trevesia palmata Vis. ITS: G. Hao 901 (CS); cultivated,
South China Botanical Garden.
trnL-trnF: G. M. Plunkett 1329 (WS );
cultivated, New York Botanical
Garden.
AF242247 AY393785
Trevesia sundaica Miq. P. P. Lowry 5285 (MO); cultivated,
Bogor Botanic Garden.
AY389063 AY393786
Trevesia cf. valida Craib P. P. Lowry 5287 (MO); cultivated,
Bogor Botanic Garden.
AY389064 AY393787
Tupidanthus calyptratus J. D. Hook.
& Thomson [= Schefflera pueckleri
(K. Koch) Frodin]
G. M. Plunkett 1315 (WS);
cultivated, New York Botanical
Garden.
AF229769 AY393788
14 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
Likelihood parameters were set to the general time-
reversible substitution model and the gamma
distribution of variable sites for among-site rate
variation. Because BI employs explicit models of
sequence evolution, the 17 indels coded as binary
characters in MP analysis were not included in the
BI data matrices. In constructing the BI tree, the
‘‘sump’’ command of MrBayes was used to deter-
mine empirically the point at which stationarity
was reached. All trees produced prior to this point
were discarded from each of the analyses. The
remaining trees were then compared to create a
single BI tree with poster ior probabilities.
Results
Sequence characteristics. Across all 108 sam-
ples studied, the length of the entire ITS region
(ITS1, 5.8S, and ITS2) ranged from 610 to 628
bp, yielding 682 characters after alignment. Of
these, 276 characters were constant, 117 char-
acters were variable in a single taxon, and 289
were potentially parsimony-informative. The
trnL-trnF sequences ranged from 766 to 928 bp
long and yielded a data matrix of 1069
characters after alignment. Of these, 762 were
constant, 145 were variable in only a single
taxon, and 162 were potentially parsimony-
informative. The combined matrix, therefore,
was 1751 bp long, of which 451 (25.8%) bp
were potentially informative. Seventeen poten-
tially informative indels were added as binary
characters for the combined MP analyses (two
of which were derived from ITS data and 15
from trnL-trnF). Aligned data matrices are
available from the authors.
DNA divergences were estimated using
Kimura’s (1980) two-parameter method in
PAUP* and treating gaps as missing data.
For ITS, distances ranged from zero (between
Trevesia cf. valida and T. sundaica) to 34.8%
(between Mackinlaya macrosciadea and Hy-
drocotyle verticillata). When considering only
taxa from core Araliaceae, the highest diver-
gence was 16.9% (between Gastonia spectabilis
and Hedera helix). For the trnL-trnF data set,
the distances ranged from zero (between 21
pairs of species, all from core Araliaceae) to
10.4% (between Hydrocotyle bowlesioides and
Delarbrea paradoxa). If only core Araliaceae
are considered, the highest divergence among
trnL-trnF sequences was 4.2% (between Tet-
rapanax papyriferus and Astrotricha sp.-‘‘Isa-
bella Falls’’). In comparing the distance matrix
of ITS to that of trnL-trnF, it was possible to
determine the relative evolutionary rate of the
two markers. In core Araliaceae, ITS evolves
an average of 6.87 times faster than trnL-trnF.
Phylogenetic analyses. The MP analysis of
the ITS data set yielded 20,000 trees of 1,664
steps, with a consistency index (CI) of 0.358
(excluding uninformative characters), and a
retention index (RI) of 0.623. The trnL-trnF
analysis resulted in over 98,500 trees of 2,371
steps, with a CI of 0.370 (excluding uninforma-
tive characters) and an RI of 0.592. The
partition homogeneity test indicated that the
two data sets were congruent and therefore
combinable (p ¼0.99; values of p £ 0.05 indi-
cate significant heterogeneity). The MP analysis
based on the combined data set yielded 50,000
trees of 2,153 steps, with a CI of 0.414 (excluding
uninformative characters) and an RI of 0.660.
Strict consensus trees from each MP analysis are
presented in Figs. 1–3 along with values
obtained from the bootstrap (BS) analyses.
Because trees resulting from the Bayesian
inference (BI) analyses were visually congruent
with those resulting from the MP analyses, only
the combined BI tree is presented (Fig. 4), along
with posterior probability (PP) values.
The ITS data set used in this study is one-
third larger than that of Wen et al. (2001a), but
to the extent that the sampling of the two studies
overlap (66 terminals common to both), the
strict consensus trees from both studies appear
to be highly congruent. As in the earlier study,
the present ITS analysis resolves two major
clades (Fig. 1). The ‘‘Asian Palmate’’ group
(called the ‘‘Eleutherococcus-Dendropanax-
Schefflera’’ group by Wen et al. 2001a) is
well-supported (BS = 77%) and includes
Eleutherococcus, Kalopanax, Macropanax,
Metapanax, Brassaiopsis, Trevesia, Hedera,
Oreopanax, Sinopanax, Gamblea, Dendropan-
ax, Fatsia, Oplopanax, Heteropanax, Tetrapan-
ax, and several species of Schefflera. This clade
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 15
Astrotricha pterocarpa
Astrotricha
sp. Isabella Falls
Hydrocotyle verticillata
Hydrocotyle vulgaris
Delarbrea paradoxa
Myodocarpus fraxinifolius
Apiopetalum velutinum
Mackinlaya macrosciadea
ITS
MP
Gastonia rodriguesiana
Polyscias sessiliflora
Cuphocarpus aculeatus
Polyscias abrahamiana
Polyscias orientalis
Munroidendron racemosum
Reynoldsia sandwicensis
Tetraplasandra oahuensis
Gastonia spectabilis
Polyscias crenata
Polyscias joskei
Polyscias lecardii
Polyscias elegans
Harmsiopanax ingens
Polyscias mollis
97
45
56
52
93
98
71
100
99
64
90
100
Plerandra vitiensis
Plerandra insolita
Schefflera baillonii
Schefflera elegantissima
Schefflera gabriellae
Meryta balansae
Meryta pedunculata
Meryta tenuifolia
90
67
74
83
98
Arthrophyllum diversifolium
Arthrophyllum mackeei
Polyscias australiana
90
98
Polyscias cf. cumingiana
Polyscias macguillivrayi
100
Pseudopanax crassifolius
Pseudopanax linearis
Pseudopanax ferox
95
Pseudopanax arboreus
Pseudopanax laetus
99
Aralia apioides
Aralia nudicaulis
Aralia excelsa
Panax trifolius
Aralia leschenaultii
Aralia kingdon-wardii
Aralia chinensis
Aralia plumosa
Aralia spinosa
98
57
81
82
Aralia humilis
Aralia scopulorum
79
Panax quinquefolius
Panax pseudoginseng
41
21
Eleutherococcus setchuanensis
Kalopanax septemlobus
Eleutherococcus sieboldiana
Eleutherococcus sessiliflorus
Macropanax dispermus
Macropanax undulatus
Metapanax delavayi
Metapanax davidii
Brassaiopsis glomerulata
Brassaiopsis hainla
Brassaiopsis grushvitzkyi
Trevesia sundaica
Trevesia
cf. valida
Trevesia palmata
Hedera helix
Merrilliopanax chinensis
59
66
99
92
100
89
50
41
100
98
88
100
100
77
Fatsia japonica
Fatsia polycarpa
Oplopanax elatus
Oplopanax horridus
54
100
100
Dendropanax arboreus
Dendropanax trifidus
49
Gamblea ciliata
Gamblea ciliata
45
Oreopanax sanderianus
Sinopanax formosanus
45
Schefflera sp.-Ecuador
Schefflera sp.-Guyana
79
Osmoxylon geelvinkianum
Osmoxylon novo-guineense
Osmoxylon pectinatum
Cephalaralia cephalobotrys
Motherwellia haplosciadea
54
100
57
Cussonia thyrsifolia
Cussonia holstii
Cussonia paniculata
Seemannaralia gerrardii
69
99
Raukaua anomalus
Cheirodendron trigynum
Cheirodendron platyphyllum
74
91
Schefflera digitata
Schefflera candelabrum
Schefflera pseudocandelabra
Schefflera vieillardii
100
92
Schefflera lukwangulensis
Schefflera umbellifera
Schefflera myriantha
100
98
100
100
100
73
75
Schefflera actinophylla
Tupidanthus calyptratus
Schefflera yunnanensis
Schefflera trevesioides
Schefflera impressa
Heteropanax fragrans
Tetrapanax papyriferus
28
61
88
88
76
97
Brassaiopsis
Subgroup
Gamblea
Dendropanax
Polyscias s. lat.
Subgroup
Pacific Schefflera
Subgroup
Oreopanax + Sinopanax
Fatsia + Oplopanax
Pseudopanax
Schefflera § Schefflera
Cheirodendron + Raukaua
Cussonia + Seemannaralia
African-Malagasy Schefflera
Osmoxylon
Astrotricha
Asian Palmate GroupPolyscias-Pseudopanax Group
Aralia Group
Asian Schefflera
Subgroup
Neotropical Schefflera
Panax
Aralia
Eleutherococcus + Kalopanax
Macropanax + Metapanax
16 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
differs from that of Wen et al. (2001a) only by
the addition of the newly sampled Merrilliopan-
ax and Grushvitzkya, the latter of which has
been reduced to synonymy under Brassaiopsis
(Wen et al. 2003a.). The second major clade,
called the ‘‘Aralia-Polyscias-Pseudopanax’’
group by Wen et al. (2001a), comprises Aralia
(incl. Pentapanax and Sciadodendron), Panax,
Meryta, Pseudopanax, Polyscias, Reynoldsia,
Tetraplasandra, Munroidendron, Gastonia,
Cuphocarpus, Arthrophyllum and several species
of Schefflera (all identical to the findings of Wen
et al. 2001a), plus the newly sampled Harmsio-
panax and Plerandra. Support for this clade,
however, remains very poor (BS = 21%). On
the basis of results from the remaining analyses,
we recognize two distinct groups within this
clade, the Aralia group and the Polyscias-
Pseudopanax group. The larger lineages in the
ITS tree form two of eight clades in a broad
basal polytomy. Other clades in this polytomy
include Cheirodendron + Raukaua (BS =
74%), Cussonia + Seemannaralia (BS =
69%), Osmoxylon + Cephalaralia + Mother-
wellia (BS ¼< 50%), and three separate clades
comprised of species assigned to Schefflera [two
clades representing Schefflera sect. Schefflera
(Schefflera candelabrum + S. pseudocandela-
brum + S. vieillardii, BS = 100%, and
S. digitata); and an African-Malagasy clade of
Schefflera, BS = 98%]. A small clade with two
species of Astrotricha (BS = 100%) is sister to
the other members of core Araliaceae. The
entire core Araliaceae clade is weakly supported
(BS = 57%), but if it is considered together with
its sister group, Hydrocotyle, support is consid-
erably higher (BS = 75%).
The consensus tree based on trnL-trnF data
is largely similar, but less resolved (Fig. 2). The
generic composition of the Asian Palmate
group is identical with one exception (Moth-
erwellia appears within this clade rather than
in the Osmoxylon clade), but support is
extremely low (BS = 7%). The combined
Aralia-Polyscias-Pseudopanax group found in
the ITS tree is not resolved in the trnL-trnF
tree, but is instead represented by two distinct
clades: the Aralia group (Panax plus Aralia,
incl. Pentapanax and Sciadodendron;BS=
43%) and the Polyscias-Pseudopanax group
(Polyscias s. lat., the Pacific Schefflera sub-
group, and Pseudopanax; BS = 64%). The
other clades found in the large basal polytomy
of the trnL-trnF tree are also similar to those in
the ITS consensus, but not identical. For
example, Cheirodendron + Raukaua are not
fully resolved, and these genera are placed
together in a larger clade with Cephalaralia
and Schefflera section Schefflera (BS = 53%).
The trnL-trnF data also suggest that there is a
single clade uniting the African-Malagasy
Schefflera + Astrotricha + Hydrocotyle (albeit
very poorly supported; BS = 31%). Together,
core Araliaceae plus Hydrocotyle are again
very strongly supported (BS = 98%).
The consensus tree resulting from the com-
bined data set (Figs. 3, 4) is visually congruent
to that of Wen et al. (2001a), with the same two
major clades comprising virtually the same
genera. The Asian Palmate group is strongly
supported (BS = 86%, PP = 100%), and as in
the ITS tree, it excludes Motherwellia. The
Aralia group (BS = 67%, PP = 100%) and
Polyscias-Pseudopanax group (BS = 76%, PP
= 100%) form two distinct subclades, as in
the trnL-trnF tree, which together form a single
larger clade (BS = 40%; PP = 51%) as in the
ITS tree.
Discussion
Overall relationships in core Araliaceae. The
utility of ITS as a marker of phylogenetic
relationships in core Araliaceae was
Fig. 1. Strict consensus of 20,000 shortest trees resulting from the maximum parsimony analysis of 108 ITS
sequences; tree length = 1,664 steps; excluding uninformative characters, CI = 0.358; RI = 0.623. Bootstrap
percentages are provided above the branches. Names of clades discussed in the text are provided next to
brackets (dashed brackets indicate clades resolved in other analyses but left unresolved here)
b
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 17
Eleutherococcus sessiliflorus
Gamblea ciliata
Gamblea ciliata
Brassaiopsis grushvitzkyi
Heteropanax fragrans
Motherwellia haplosciadea
Tetrapanax papyriferus
Trevesia palmata
Trevesia sundaica
Trevesia cf. valida
Cuphocarpus aculeatus
Polyscias abrahamiana
Polyscias orientalis
Arthrophyllum diversifolium
Arthrophyllum mackeei
Polyscias australiana
Polyscias mollis
Polyscias
cf. cumingiana
Polyscias macguillivrayi
Meryta balansae
Meryta pedunculata
Meryta tenuifolia
Plerandra vitiensis
Plerandra insolita
Pseudopanax crassifolius
Pseudopanax linearis
Pseudopanax ferox
Schefflera gabriellae
Aralia humilis
Aralia nudicaulis
Aralia scopulorum
Aralia leschenaultii
Aralia plumosa
Aralia kingdon-wardii
Harmsiopanax ingens
trnL-trnF
MP
Dendropanax arboreus
Oreopanax sanderianus
Sinopanax formosanus
Fatsia japonica
Fatsia polycarpa
60
62
Kalopanax septemlobus
Metapanax delavayi
Metapanax davidii
Macropanax dispermus
Macropanax undulatus
97
56
42
Schefflera actinophylla
Schefflera impressa
Schefflera trevesioides
Schefflera yunnanensis
Tupidanthus calyptratus
93
Brassaiopsis glomerulata
Brassaiopsis hainla
88
Merrilliopanax chinensis
Hedera helix
63
Oplopanax elatus
Oplopanax horridus
95
7
Dendropanax trifidus
Schefflera sp.-Ecuador
Schefflera
sp.-Guyana
64
Polyscias elegans
Polyscias joskei
Polyscias lecardii
Polyscias crenata
56
78
Gastonia rodriguesiana
Polyscias sessiliflora
61
Gastonia spectabilis
Tetraplasandra oahuensis
Munroidendron racemosum
Reynoldsia sandwicensis
62
66
86
98
97
70
64
Pseudopanax arboreus
Pseudopanax laetus
86
Schefflera baillonii
Schefflera elegantissima
94
57
Panax quinquefolius
Panax pseudoginseng
Panax trifolius
80
Aralia excelsa
Aralia chinensis
Aralia spinosa
Aralia apioides
63
43
Schefflera candelabrum
Schefflera pseudocandelabra
Schefflera vieillardii
Schefflera digitata
Cheirodendron trigynum
Cheirodendron platyphyllum
Cephalaralia cephalobotrys
Raukaua anomalus
87
65
9553
100
Astrotricha sp. Isabella Falls
100
32
Schefflera lukwangulensis
Schefflera umbellifera
Schefflera myriantha
48
96
31
Seemannaralia gerrardii
Cussonia thyrsifolia
Cussonia holstii
Cussonia paniculata
62
62
Osmoxylon geelvinkianum
Osmoxylon novo-guineense
Osmoxylon pectinatum
98
100
100
98
Eleutherococcus setchuanensis
Eleutherococcus sieboldiana
64
Brassaiopsis
Subgroup
Gamblea
Dendropanax
Polyscias s. lat.
Subgroup
Pacific Schefflera
Subgroup
Oreopanax + Sinopanax
Fatsia
Pseudopanax
Schefflera § Schefflera
Cheirodendron + Raukaua
Cussonia + Seemannaralia
African-Malagasy Schefflera
Osmoxylon
Astrotricha
Oplopanax
Asian Palmate GroupPolyscias-Pseudopanax Group
Aralia Group
Asian Schefflera
Subgroup
Neotropical Schefflera
Panax
Aralia
Eleutherococcus + Kalopanax
Macropanax + Metapanax
Delarbrea paradoxa
Myodocarpus fraxinifolius
Apiopetalum velutinum
Mackinlaya macrosciadea
Hydrocotyle verticillata
Hydrocotyle bowlesioides
Astrotricha pterocarpa
18 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
demonstrated by Wen et al. (2001a), and is
reinforced by the present study. The trnL-trnF
region evolves at a much slower rate (an
average of 6.87 times more slowly than ITS),
but the tree based on trnL-trnF data is only
slightly less resolved. Moreover, the trees
resulting from the separate analysis of each
marker are highly congruent, as demonstrated
by both visual inspection and the partition
homogeneity test (p ¼0.99). These findings
indicate that the nuclear ITS region and the
plastid trnL-trnF region represent complemen-
tary molecular markers with which we may
explore evolutionary relationships in core
Araliaceae.
The molecular data presented here agree
substantially with earlier findings (Wen et al.
2001a) in placing the majority of genera in
three major clades, the Asian Palmate group,
the Aralia group, and the Polyscias-Pseudo-
panax group. In all analyses, near identical
complements of genera are found in each
group. The only exceptions are the inclusion
of Motherwellia in the Asian Palmate group
based on analyses of trnL-trnF data, and the
inclusion of Harmsiopanax in the combined
Aralia plus Polyscias-Pseudopanax group in
the ITS tree. Each tree also fails to resolve
relationships among these larger clades and a
series of smaller clades at or toward the base of
core Araliaceae. On the other hand, the
number and composition of these smaller
clades is identical (or nearly so) in the ITS,
trnL-trnF, and combined trees.
Comparison of the findings of our molec-
ular data to the infrafamilial classifications of
Araliaceae outlined in Table 1 reveals little
agreement (see Table 4). Both tribes Mackin-
layeae (Mackinlaya and Apiopetalum) and
Myodocarpeae (Myodocarpus and Delarbrea,
plus Pseudosciadium) can be eliminated from
consideration because they have been excluded
from core Araliaceae (discussed above; see also
Lowry et al. 2001, Plunkett 2001, Plunkett and
Lowry 2001). Thus modified, the widely-used
system of Harms (1894–97) would have recog-
nized only two tribes: Schefflereae (with val-
vate petals) and Aralieae (with imbricate
petals). Considering only those taxa recog-
nized in Table 2, tribe Aralieae comprises five
of the genera in our sample, which are placed
in two or three well-separated clades: Mother-
wellia (either at the basal polytomy or in the
Asian Palmate group), Harmsiopanax (at the
basal polytomy or in the Aralia plus Polyscias-
Pseudopanax group), and Aralia + Panax (the
Aralia group). Harms assigned all of the
remaining genera to his large tribe Scheffle-
reae, members of which can be found in almost
every clade of Araliaceae as revealed by the
present study (Table 4). The earlier systems of
Bentham (1867) and Seemann (1868) were
more elaborate, incorporating both stamen
number and endosperm texture to define
infrafamilial groups. The genera of Harms’s
Schefflereae were assigned by Bentham to three
tribes: (1) Panaceae (which Seemann treated as
Horsfieldieae plus Pseudopanaceae), (2) Hede-
reae (Seemann’s Cussonieae and Hedereae),
and (3) Plerandreae (treated in like manner by
Seemann). None of these segregate tribes
represent monophyletic groups in the trees
based on molecular data (Table 4). Bentham’s
definition of Aralieae was essentially the same
as that of Harms, but Seemann placed its
members in two tribes (Aralieae and Panaceae)
within a very narrowly defined Araliaceae (as
distinct from Hederaceae; see Table 1). Among
the three nineteenth century authors, both
Harms and Seemann placed Aralia (incl. Pen-
tapanax and Sciadodendron) with Panax, along
with a number of genera now excluded
from Araliaceae. In addition, Harms also
included Motherwellia and Horsfieldia
Fig. 2. Strict consensus of 98,500 shortest trees resulting from the maximum parsimony analysis of 108 trnL-
trnF sequences; tree length = 2,371 steps; excluding uninformative characters, CI = 0.370; RI = 0.592.
Bootstrap percentages are provided above the branches. Names of clades discussed in the text are provided next
to brackets (dashed brackets indicate clades resolved in other analyses but left unresolved here)
b
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 19
Combined
MP
Gastonia rodriguesiana
Polyscias sessiliflora
Cuphocarpus aculeatus
Polyscias abrahamiana
Polyscias orientalis
Munroidendron racemosum
Reynoldsia sandwicensis
Tetraplasandra oahuensis
Gastonia spectabilis
Polyscias joskei
Polyscias lecardii
Polyscias crenata
Polyscias elegans
Arthrophyllum diversifolium
Arthrophyllum mackeei
Polyscias australiana
Polyscias mollis
Polyscias
cf. cumingiana
Polyscias macguillivrayi
Plerandra vitiensis
Plerandra insolita
Schefflera baillonii
Schefflera elegantissima
Schefflera gabriellae
Meryta balansae
Meryta pedunculata
Meryta tenuifolia
Pseudopanax crassifolius
Pseudopanax linearis
Pseudopanax ferox
Pseudopanax arboreus
Pseudopanax laetus
Aralia leschenaultii
Aralia kingdon-wardii
Aralia chinensis
Aralia spinosa
Aralia plumosa
Panax quinquefolius
Panax pseudoginseng
Panax trifolius
Aralia humilis
Aralia scopulorum
Aralia apioides
Aralia nudicaulis
Aralia excelsa
Brassaiopsis glomerulata
Brassaiopsis hainla
Brassaiopsis grushvitzkyi
Trevesia sundaica
Trevesia
cf. valida
Trevesia palmata
Hedera helix
Oreopanax sanderianus
Sinopanax formosanus
Schefflera actinophylla
Tupidanthus calyptratus
Schefflera yunnanensis
Schefflera trevesioides
Schefflera impressa
Heteropanax fragrans
Tetrapanax papyriferus
Schefflera
sp.-Ecuador
Schefflera sp.-Guyana
Eleutherococcus sessiliflorus
Eleutherococcus setchuanensis
Eleutherococcus sieboldianus
Kalopanax septemlobus
Macropanax dispermus
Macropanax undulatus
Metapanax delavayi
Metapanax davidii
Fatsia japonica
Fatsia polycarpa
Oplopanax elatus
Oplopanax horridus
Dendropanax arboreus
Dendropanax trifidus
Merrilliopanax chinensis
Gamblea ciliata
Gamblea ciliata
Cussonia thyrsifolia
Cussonia holstii
Cussonia paniculata
Seemannaralia gerrardii
Schefflera umbellifera
Schefflera lukwangulensis
Schefflera myriantha
Harmsiopanax ingens
Schefflera candelabrum
Schefflera pseudocandelabra
Schefflera vieillardii
Schefflera digitata
Cheirodendron trigynum
Cheirodendron platyphyllum
Raukaua anomalus
Osmoxylon geelvinkianum
Osmoxylon novo-guineense
Osmoxylon pectinatum
Astrotricha pterocarpa
Astrotricha
sp. Isabella Falls
Cephalaralia cephalobotrys
Motherwellia haplosciadea
Hydrocotyle verticillata
Hydrocotyle vulgaris + H. bowlesioides
Delarbrea paradoxa
Myodocarpus fraxinifolius
Apiopetalum velutinum
Mackinlaya macrosciadea
99
96
97
52
100
98
99
100
100
100
36
10
63
72
7
54
100
100
59
98
76
91
98
68
38
15
89
86
99
48
52
61
100
72
97
99
67
96
91
100
100
100
96
100
55
100
72
82
89
98
99
100
74
76
40
67
83
97
93
69
62
61
93
100
51
99
87
86
99
100
98
43
100
51
100
100
100
100
99
Eleutherococcus + Kalopanax
Asian Schefflera
Subgroup
Brassaiopsis
Subgroup
Gamblea
Dendropanax
Polyscias s. lat.
Subgroup
Pacific Schefflera
Subgroup
Oreopanax + Sinopanax
Fatsia + Oplopanax
Pseudopanax
Asian Palmate GroupPolyscias-Pseudopanax Group
Aralia Group
Schefflera § Schefflera
Cheirodendron + Raukaua
Cussonia + Seemannaralia
African-Malagasy Schefflera
Osmoxylon
Astrotricha
Neotropical Schefflera
Panax
Aralia
Macropanax + Metapanax
20 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
(= Harmsiopanax). Molecular data provide
evidence only for a close relationship between
Aralia and Panax (the Aralia group).
Six of Calestani’s (1905) eight tribes and
eight of Viguier’s (1906) 10 tribes include
genera now placed within core Araliaceae.
Although these systems made use of a larger
number of characters, they do not delimit
tribes that correspond to the molecular group-
ings (Table 4). Apart from Calestani’s mono-
generic tribes (Osmoxyleae and Meryteae), the
remaining four tribes are each polyphyletic.
Three of his four largest tribes (Plerandreae,
Aralieae, Schefflereae) are represented by taxa
scattered throughout the molecular clado-
gram. His Hedereae is limited to taxa from
the Asian Palmate clade, but within this broad
clade, the genera of his Hedereae are not
monophyletic. Similarly, five of Viguier’s
larger tribes (Pseudopanaceae, Polyscieae,
Schefflereae, Hedereae, and Plerandreae) con-
tain genera found in three to 10 of our clades.
Members of his tribe Eremopanaceae are
found among three distinct clades. If current
generic circumscriptions are employed, his two
small tribes Panaceae and Meryteae each have
only a single genus, Panax and Meryta
(respectively). Our results place these two
genera within larger clades together with other
taxa (viz., Panax in the Aralia group, and
Meryta in the Pacific Schefflera subgroup).
In comparing the molecular trees to the
two more-recent treatments of Araliaceae,
Tseng and Hoo’s (1982) system fares better
than Hutchinson’s (1967). Beyond the many
clearly artificial genera recognized to fit Hutch-
inson’s rigid use of inflorescence characters, his
system contains numerous other inconsisten-
cies when compared to the trees based on
molecular data. When considering only the
genera of core Araliaceae, not one of Hutch-
inson’s tribes forms a monophyletic group in
the molecular trees. Tseng and Hoo’s (1982)
system compares somewhat more favorably.
Unlike earlier treatments, theirs incorporated
both habit and leaf shape to define tribes such
as Tetraplasandreae, which corresponds al-
most perfectly with the Polyscias s. lat. sub-
group in the molecular trees (differing only in
their placement of Heteropanax within Tetra-
plasandreae). Tseng and Hoo’s tribes Aralieae
and Panaceae are both polyphyletic, and as in
all previous systems (except Harms’), they
placed the closely related Aralia and Panax
in two distinct tribes. Finally, their clearly
artificial tribe Plerandreae has representatives
in almost every major clade of the molecular
trees (Table 4).
These comparisons suggest that infrafamil-
ial groups in Araliaceae cannot be recognized
solely on the basis of the few morphological
characters favored in most previous systems
(particularly when applied in a strict fashion).
For example, only one of the six treatments
outlined in Table 2 places Aralia and Panax in
the same tribe (that of Harms), even though all
recent studies have confirmed the close rela-
tionship of these genera (e.g. Wen and Zimmer
1996, Wen 2001, Wen et al. 2001a). The
traditional reliance on petal aestivation has
probably been greatly overemphasized, espe-
cially in light of the many genera with an
imbricate corolla that are now placed outside of
core Araliaceae (e.g. Myodocarpus, Delarbrea,
Stilbocarpa, and Aralidium). At present, imbri-
cate petals serve to define only the Aralia group,
and even then not uniquely, as several genera
found outside this clade also have imbricate
aestivation, including Seemannaralia, Mother-
wellia, Harmsiopanax, and Cephalaralia. Two
other characters have also been heavily
weighted in traditional classifications systems:
Fig. 3. Strict consensus of 50,000 shortest trees resulting from the maximum parsimony analysis of the
combined (ITS + trnL-trnF) data set from 108 taxa; tree length = 2,153 steps; excluding uninformative
characters, CI = 0.414; RI = 0.660. Bootstrap percentages are provided above the branches. Names of clades
discussedinthetextareprovidednexttobrackets(dashed brackets indicate clades resolved in other analyses
but left unresolved here)
b
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 21
Combined
BI
Schefflera candelabrum
Schefflera pseudocandelabra
Schefflera vieillardii
Schefflera digitata
Cheirodendron trigynum
Cheirodendn platphyllum
Raukaua anomalus
Cephalaralia cephalobotrys
Motherwellia haplosciadea
Cussonia thyrsifolia
Cussonia holstii
Cussonia paniculata
Seemannaralia gerrardii
Schefflera lukwangulensis
Schefflera umbellifera
Schefflera myriantha
Hydrocotyle verticillata
Hydrocotyle vulgaris + H. bowlesioides
Astrotricha pterocarpa
Astrotricha sp. Isabella Falls
Osmoxylon geelvinkianum
Osmoxylon novo-guineense
Osmoxylon pectinatum
Delarbrea paradoxa
Myodocarpus fraxinifolius
Apiopetalum velutinum
Mackinlaya macrosciadia
Cuphocarpus aculeatus
Polyscias crenata
Polyscias lecardii
Polyscias joskei
Polyscias elegans
Arthrophyllum diversifolium
Arthrophyllum mackeei
Polyscias australiana
Polyscias mollis
Polyscias cf. cumingiana
Polyscias macguillivrayi
Plerandra vitiensis
Plerandra insolita
Schefflera baillonii
Schefflera elegantissima
Schefflera gabriellae
Meryta balansae
Meryta pedunculata
Meryta tenuifolia
Pseudopanax crassifolius
Pseudopanax linearis
Pseudopanax ferox
Pseudopanax arboreus
Pseudopanax laetus
Aralia nudicaulis
Aralia excelsa
Aralia humilis
Aralia scopulorum
Harmsiopanax ingens
Munroidendron racemosum
Reynoldsia sandwicensis
Tetraplasandra oahuensis
Gastonia spectabilis
Gastonia rodriguensis
Polyscias sessiliflora
Polyscias abrahamiana
Polyscias orientalis
Aralia leschenaultii
Aralia kingdon-wardii
Aralia chinensis
Aralia plumosa
Aralia spinosa
Aralia apioides
Panax quinquefolius
Panax pseudoginseng
Panax trifolius
Eleutherococcus setchuanensis
Kalopanax septemlobus
Eleutherococcus sieboldianus
Eleutherococcus sessiliflorus
Macropanax dispermus
Macropanax undulatus
Metapanax delavayi
Metapanax davidii
Hedera helix
Schefflera actinophylla
Tupidanthus calyptratus
Schefflera yunnanensis
Schefflera trevesioides
Schefflera impressa
Heteropanax fragrans
Tetrapanax papyriferus
Brassaiopsis glomeratus
Brassaiopsis hainla
Brassaiopsis grushvitzkyi
Trevesia sundaica
Trevesia cf. valida
Trevesia palmata
Fatsia japonica
Fatsia polycarpa
Oplopanax elatus
Oplopanax horridus
Merrilliopanax chinensis
Dendropanax arboreus
Dendropanax trifidus
Gamblea ciliata
Gamblea ciliata
Oreopanax sanderianus
Sinopanax formosanus
Schefflera
sp.-Ecuador
Schefflera sp.-Guyana
100
100
56
53
96
51
100
97
100
100
100
98
100
100
67
100
100
57
100
100
100
100
100
100
99
100
100
100
100
100
100
100
65
60
62
100
79
58
100
99
66
100
100
97
98
85
55
100
99
100
97
98
99
99
76
100
100
100
100
100
100
94
100
100
94
100
66
51
88
100
100
100
100
100
100
100
100
96
100
100
55
100
100
Brassaiopsis
Subgroup
Gamblea
Dendropanax
Polyscias s. lat.
Subgroup
Pacific Schefflera
Subgroup
Oreopanax + Sinopanax
Fatsia + Oplopanax
Pseudopanax
Schefflera § Schefflera
Cheirodendron + Raukaua
Cussonia + Seemannaralia
African-Malagasy Schefflera
Osmoxylon
Astrotricha
Asian Palmate GroupPolyscias-Pseudopanax Group
Aralia Group
Asian Schefflera
Subgroup
Neotropical Schefflera
Panax
Aralia
Eleutherococcus + Kalopanax
Macropanax + Metapanax
22 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
endosperm texture and stamen number. Ben-
tham’s system serves to illustrate the discor-
dance between the presence (or absence) of
ruminate endosperm and underlying relation-
ships: neither his Hedereae (ruminate endo-
sperm) nor Panaceae (smooth endosperm)
approach monophyly (see Table 4). Likewise,
the use of stamen number (in relation to petal
number) has failed to define monophyletic
groups. Most systems have treated the presence
of numerous stamens as a defining character of
Plerandreae (as in the systems of Bentham,
Seemann, Viguier, and Hutchinson, but divided
into Plerandreae and Tetraplasandreae in Hoo
and Tseng’s system). Genera with this feature,
however, do not form a monophyletic group
even among the core members of the tribe (e.g.
Plerandra, Tupidanthus, Tetraplasandra, and
some species of Schefflera, see also Costello
and Motely 2001).
The strong agreement between trees based
on two independent molecular data sets (rep-
resenting both plastid and nuclear genomes)
stands in stark contrast to the traditional tribal
classifications of Araliaceae. This result mir-
rors findings described by many studies of the
closely related Apiaceae, where molecular
phylogenies based on many data sets have
converged on a unified picture of evolutionary
relationships that is at odds with all of the
published infrafamilial systems for that family
(e.g. Downie et al. 1996, 1998, 2001; Plunkett
et al. 1996b; Katz-Downie et al. 1999; Plunkett
and Downie 1999, 2000). As with Apiaceae,
the source of this problem may rest in the
relatively simplistic application of a small
number of morphological characters, biased
by preconceived notions of evolutionary trends
(Plunkett 2001). Downie et al. (2001) suggested
that resolution of tribal-level problems in
Apiaceae will require recognition of clades
largely on the basis of molecular data, with a
posteriori identification of morphological syn-
apomorphies that may be unique only as suites
of characters. This approach may also be
required for Araliaceae, although the vast
complexities plaguing the interpretation of
morphology in Apiaceae are less acute in
Araliaceae. Because published descriptions of
Araliaceae are often incomplete (due in large
part to the size and morphological complexity
of many species, the limitations of regional
treatments of widespread genera, and differing
notions of generic circumscriptions), we have
initiated construction of a database of mor-
phological characters from throughout the
family. To date, a comprehensive set of such
characters is not available. Until such time, we
recognize the clades outlined in Table 4 as
‘‘working groups’’ that are considered useful
for discussing informal taxonomic units and
for exploring evolutionary trends and relation-
ships.
Generic and intergeneric relationships: indi-
vidual assessments. The large degree of agree-
ment between trees based on plastid and
nuclear markers provides an opportunity to
examine evolutionary relationships among a
number of smaller groups within core Aralia-
ceae. A significant finding of this and other
recent studies is the non-monophyly of the two
largest genera in the family: Schefflera (ca. 650
species) and Polyscias (ca. 150 species). Given
that nearly two-thirds of all species in Arali-
aceae are currently assigned to these two
genera, this result has a profound effect on
our interpretation of intergeneric relationships
and character evolution in the family. After
assessing the status of Schefflera, Polyscias,
and their close relatives, a discussion is pro-
vided on other large or newly-discovered
groups in Araliaceae.
The Polyphyly of Schefflera: Over the past
thirty years, the definition of Schefflera has
been broadened to include all araliads having
once palmately compound leaves, and lacking
Fig. 4. TreebasedonBayesianInference(BI)analysisofthecombined(ITS+trnL-trnF) data set from 108
taxa. Posterior probabilities are provided above the branches. Names of clades discussed in the text are
provided next to brackets (dashed brackets indicate clades resolved in other analyses but left unresolved here)
b
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 23
Table 4. Clades in core Araliaceae based on molecular data compared to traditional tribal systems; see
Table 1 for definitions of tribes. Abbreviations of systems: Benth. = Bentham 1867; Seem. = Seemann
1868; Harms = Harms 1894-97; Cal. = Calestani 1905; Vig. = Viguier 1906; Hutch. = Hutchinson 1967;
H&T = Hoo and Tseng 1983. Abbreviations of tribes: Aral = Aralieae; Cuss = Cussonieae;
Erem = Eremopanaceae; Hed = Hedereae; Hors = Horsfieldieae; Mer = M eryteae; Osm = Osmoxyleae;
Pan = Panaceae; Pler = Plerandreae; Poly = Polyscieae; Pseu = Pseudopanaceae; Sc h = Schefflereae;
Tetpl = Tetraplasandreae. Dashes indicate genera not treated in a given system
Molecular Clades Benth. Seem. Harms Cal. Vig. Hutch. H.&T.
Asian palmate Group
Eleutherococcus Subgroup
Eleutherococcus Pan Pseu Sch Pseu Pan Pler
Kalopanax Sch Pan Pler
Macropanax Hed Cuss Sch Sch Hed Hed Pler
Metapanax ––––
Brassaiopsis Subgroup
Brassaiopsis Hed Cuss Sch Sch Sch Hed Pler
Grushvitzkya ––––
Trevesia Pan Pseu Sch Hed Sch Pan Pler
Hedera Hed Hed Sch Hed Hed Hed Pler
Oreopanax Subgroup
Oreopanax Hed Hed Sch Hed Hed Hed Pler
Sinopanax Pan Pler
Gamblea Subgroup
Evodiopanax Pan
Gamblea Sch Hed Hed Hed Pler
Dendropanax
Dendropanax Pan Pseu Sch Pan Pler
Fatsia Subgroup
Fatsia Pan Pseu Sch Hed Pseu Pan/(Hed) Pler
Oplopanax Sch Pan Sch Pan Pler
Asian Schefflera Subgroup
Schefflera (p.p.) Pler/
Pan
Psue/
Pler
Sch Sch (Sch)/
Pler
Cuss/Pler/
Pan/Hed
Pler
Heteropanax Hed Cuss Sch Sch Hed Hed Tetpl
Tetrapanax Hors Sch Pan Sch Pan Pler
Merrilliopanax Pan Pler
Aralia Group
Aralia (incl. Pentapanax) Aral Aral Aral Aral Poly Aral Aral
Sciadodendron Aral Aral Aral Pler Aral Aral
Panax Pan Pan Aral Pan Pan Pan Pan
Polyscias-Pseudopanax Group
Polyscias s. lat. Subgroup
Polyscias Pan Pseu Sch Sch Poly Pan Tetpl
Reynoldsia Pseu Sch Sc h Pler Cuss Tetpl
Tetraplasandra Pler Pler Sch Pler/Sch Pler Pler Tetpl
Munroidendron Cuss Tetpl
Gastonia Pan Pseu Sch Sch Pler Pan Tetpl
Cuphocarpus Pan Sch Poly Cuss Tetpl
Arthrophyllum Hed Sch Sch Erem Hed/(Pan) Tetpl
24 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
both articulated pedicels and prickles or arms
(Frodin 1975, 1986, 1989; Maguire et al. 1984;
Lowry 1989; Lowry et al. 1989). As a result of
this redefinition, many formerly distinct genera
are now treated within Schefflera, including
Agalma, Brassaia, Didymopanax, Dizygotheca,
Neocussonia, Octotheca, Plerandra, Scioda-
phyllum, and Tupidanthus. As such, the genus
has swelled to 650 species, representing half
of the species diversity across the family. In
two earlier studies of Araliaceae based on ITS
data (Plunkett et al. 2001, Wen et al. 2001a),
eight to nine species of Schefflera were sam-
pled. Even with this very limited coverage, the
species of Schefflera were found in at least
three unrelated clades in both studies. In the
present analysis, we have modestly increased
our sampling to 19 species of Schefflera
(including Plerandra and Tupidanthus). All
three data sets provide highly congruent results
suggesting at least five distinct clades: Asian
Schefflera, Neotropical Schefflera, Pacific Is-
land Schefflera, African/Malagasy Schefflera,
and Schefflera sect. Schefflera (which includes
the type species, S. digitata). In the separate
analyses (Figs. 1, 2), only four of the five
clades are fully resolved (sect. Schefflera par-
tially collapses in the ITS trees, and the Pacific
Island group collapses in the trnL-trnF trees),
but all five clades are resolved in the combined
trees (Figs. 3, 4).
As the informal group names connote,
these five clades correspond largely to geo-
graphic distribution. Moreover, most of them
are not closely related within Araliaceae. The
Asian and Neotropical Scheffleras are weakly
allied within the broader Asian Palmate clade.
SE Asian–Neotropical disjunctions are unu-
sual, but the same distribution can be found
elsewhere among the Asian Palmates (e.g. in
Dendropanax, and between Oreopanax and
Sinopanax). The Pacific Island group of
Schefflera is found within the broader Pseudo-
panax-Polyscias group. In most trees (Figs. 1,
3, 4) there is a well-resolved sister-group
relationship between the Pacific Island Sche-
ffleras and Meryta, an unexpected finding
given the simple (to pinnatifid) leaves and
exclusively dioecious mating system of the
latter. In both of the combined trees (Figs. 3,
Table 4 (conti nued)
Molecular Clades Benth. Seem. Harms Cal. Vig. Hutch. H.&T.
Pacific Schefflera Subgroup
Schefflera (p.p.) (Pan) (Pseu) Sch Sch Sch Pan Pler
Plerandra Pler Pler Sch Pler Pler Pler Pler
Meryta Pan Sch Mer Mer Pan Pler
Pseudopanax Pan Pseu Sch Pseu Pan Pler
Other Groups:
Schefflera sect. Schefflera Pan Pseu Sch Sch Sch Pan Pler
African-Malagasy Schefflera Pan Pseu Sch (Sch) (Sch) (Pan) (Pler)
Cheirodendron Group
Cheirodendron Pseu Sch Pseu Pan Pler
Raukaua Pseu
Cussonia Group
Cussonia Hed Cuss Sch Sch Hed Cuss/(Hed) Pler
Seemannaralia Pan Pler
Osmoxylon Pan Pseu Sch Osm Sch/He d Pan Pler
Cephalaralia Poly Aral Aral
Motherwellia Aral Aral Poly Aral
Harmsiopanax Pan Hors Aral Pan Sch Aral Pan
Astrotricha Pan Hors Sch Pan Pseu Pan Pler
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 25
4), the Schefflera + Meryta clade is sister to
the two clades of Pseudopanax. The two
remaining groups of Schefflera are found
among the many small clades forming the
large polytomy at the base of core Araliaceae.
Although sampling of Schefflera in the
present study remains very limited, the results
are corroborated (at least in part) by previous
studies of pollen morphology (Tseng and
Shoup 1978) and wood anatomy (Oskolski
1996, Oskolski and Lowry 2001), which sug-
gest a high degree of variation in Schefflera
relative to other genera in the family. More-
over, preliminary results from a more extensive
molecular study focused specifically on
Schefflera resolve the same five clades (Plunk-
ett, Lowry, D. G. Frodin and Wen, unpubl.).
These findings suggest that massive taxonomic
revisions may be needed to reflect phylogenetic
relationships among the species currently
treated in Schefflera.
The Paraphyly of Polyscias: In its current
circumscription, Polyscias is the second most
speciose genus in Araliaceae (150 spp.). As
with Schefflera, the definition of Polyscias has
been considerably broadened during recent
decades (Stone 1965a, b; Smith and Stone
1968; Bernardi 1971; Philipson 1978, 1979;
Lowry 1989; Smith 1985) and presently
includes all araliads with imparipinnate leaves
and articulated pedicels. Using this definition,
many formerly distinct genera are currently
treated under the synonymy of Polyscias (e.g.
Bonnierella, Botryopanax, Eupteron, Gelibia,
Kissodendron, Nothopanax, Palmervandenbroe-
kia, Sciadopanax, and Tieghemopanax). In
both the broad study of Araliaceae (Wen et
al. 2001a) and the more focused analysis of
Polyscias (Plunkett et al. 2001), ITS sequence
data confirmed the alliance of these taxa.
However, those studies, together with the
present analyses, indicate Polyscias (as cur-
rently defined) is paraphyletic. In fact, most of
the pinnate genera of Araliaceae fall within the
large clade labeled ‘‘Polyscias s. lat.’’ (Figs. 1–
4). Relationships in the Pseudopanax-Polyscias
group in both the ITS and trnL-trnF trees
(Figs. 1, 2) are not fully resolved, but the
combined trees (Figs. 3, 4) resolve a single
clade that includes all samples from Polyscias
and six additional genera (Arthrophyllum,
Gastonia, Cuphocarpus, Reynoldsia, Tetrapla-
sandra, and Munroidendron). In studies based
on ITS data alone, support for this clade was
poor (BS < 50%), and it was not possible to
separate the Pacific Island Schefflera + Mery-
ta + Pseudopanax from Polyscias s. lat.
(Plunkett et al. 2001). The addition of trnL-
trnF data in the MP analyses modestly
increases bootstrap support for Polyscias
s. lat. (BS = 55%), but the BI tree provides
evidence for a well-supported clade (PP =
97%), and both trees now exclude all samples
of Schefflera, Meryta, and Pseudopanax.
The nature of the paraphyly of Polyscias
indicates a need for major realignments among
all the genera in this clade. For example,
within Polyscias s. lat., Gastonia appears to be
polyphyletic (Figs. 1–4), with some species
associated with Indian Ocean basin Polyscias
(e.g. G. rodriguesiana) and others with Pacific
taxa (e.g. G. spectabilis). This finding was also
corroborated in studies employing broader
samplings from Gastonia (Vu 2001, Costello
2002). Moreover, Plunkett, N. Vu, and Lowry
(unpubl.) have found that Cuphocarpus is also
polyphyletic within the Indian Ocean lineages
of Polyscias. Reynoldsia, Tetraplasandra, and
Munroidendron form a single clade centered on
the Pacific islands, especially Hawaii (where
Tetraplasandra and Munroidendron are ende-
mic) (see also Costello and Motley 2001),
together with Gastonia spectabilis from the
western Pacific. Arthrophyllum (W Malesia to
New Caledonia) appears monophyletic, but is
sister to two species of Australian Polyscias
(P. australiana, P. mollis). Although our sam-
pling is not exhaustive, it appears that the
Polyscias s. lat. clade probably represents
nearly 200 species, all of which have impari-
pinnate leaves (with the exception of a few
unifoliolate species). In order to refine our
understanding of relationships within this
clade, two studies have already been completed
(Eibl et al. 2001, for Polyscias sect. Tieghemo-
panax, and Plunkett et al., unpubl., for the
26 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
Indian Ocean taxa), and additional analyses
are now under way. As Plunkett et al. (2001)
suggest, paraphyly in Polyscias could either be
addressed by further broadening the definition
of the genus or through generic recircumscrip-
tions. Expanding the genus to include all of the
segregates comprising the Polyscias s. lat. clade
would, however, render it so diverse morpho-
logically that the genus would be difficult if not
impossible to characterize. Moreover, as
pointed out by Plunkett et al. (2001), while
the adoption of a broadly defined Polyscias
would meet the requirement for monophyly, it
would simply relegate the taxonomic difficul-
ties to a lower level and would not help to
clarify relationships among the members of the
group. It therefore seems preferable to recog-
nize about six to eight realigned genera within
the Polyscias s. lat. clade, each of which could
be circumscribed to comprise a morphologi-
cally and geographically coherent assemblage.
Eleutherococcus + Kalopanax: Eleuthero-
coccus consists of 35 species from eastern
Asia (China, Korea, Japan, and eastern Rus-
sia) and the Himalayan region (Harms 1918;
Li 1942; Hoo and Tseng 1965, 1978; Kim and
Sun 2000). The genus has been highly contro-
versial taxonomically, especially concerning
its delimitation, infrageneric classification
(e.g. Harms 1918, Nakai 1924, Li 1942, Hoo
and Tseng 1978, Ohashi 1987), and species
circumscriptions. Nakai (1924) divided
Eleutherococcus sensu Harms into four genera:
Evodiopanax, Kalopanax, Acanthopanax, and
Eleutherococcus s. str. Hoo and Tseng (1978)
recognized Kalopanax as a distinct genus, but
otherwise followed Harms’s (1918) classifica-
tion. Ohashi (1987) concurred with Nakai’s
(1924) recognition of Evodiopanax and Kalo-
panax as distinct genera, but placed Acantho-
panax in Eleutherococcus. Eleutherococcus sect.
Sciadophylloides of Harms (1918) is now
recognized as Chengiopanax (Shang and
Huang 1993a). Data based on trnL-trnF
sequences suggest that Kalopanax is sister to
Metapanax + Macropanax, but support for
this relationship is very low (BS = 42%;
Fig. 2), and all of the remaining trees strongly
support the inclusion of Kalopanax within
Eleutherococcus (BS or PP = 98–99%; Figs. 1,
3, 4). Like Eleutherococcus, Kalopanax bears
prickly stems, but it differs by its simple (vs.
palmately compound) leaves and much larger
inflorescences. The relationships among
Eleutherococcus and its close allies should be
examined in greater detail using additional
markers and an expanded sampling.
Gamblea: Gamblea was long recognized as
an unarmed monotypic Himalayan genus (see
Harms 1894–97, Hutchinson 1967), although
Harms (1918) suggested that it was closely
related to Acanthopanax sect. Eleutherococcus
(= Eleutherococcus), which generally has
spines or prickles. Several unarmed taxa from
Japan, Korea and China initially placed in
various genera (including Acanthopanax,
Eleutherococcus, Kalopanax, and Panax) were
united by Nakai (1924) under Evodiopanax,to
which another species from Malaysia and
northern Sumatra was later added (Henderson
1933, see also Philipson 1979, Shang and
Huang 1993b). Shang et al. (2000), however,
showed that material previously assigned to
Evodiopanax is congeneric with the collections
from the Himalayas placed in Gamblea; they
excluded G. longipes Merrill from Myanmar,
which was later transferred to Aralia (Wen
et al. 2001b). Our molecular analyses confirm
that Gamblea represents a distinct lineage
within the Asian Palmate group, and that it
is not closely related to the Eleutherococcus +
Kalopanax group (Figs. 1–4).
Macropanax + Metapanax: Wen and
Frodin (2001) described Metapanax to include
two species, M. davidii and M. delavayi, from
China and Vietnam, both of which were
previously placed in various other genera,
including Panax (Franchet 1886, 1896), Notho-
panax (= Polyscias) (Harms 1900), and
Pseudopanax (Philipson 1965). Our analysis
strongly supports the monophyly of the genus
(BS = 100, Fig. 3).
Macropanax comprises 15 species with a
wide distribution from central China westward
to the Himalayan region (Sikkim, Bhutan and
Nepal) and southward to west Malaysia, with
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 27
one species in Java (Philipson 1979; Shang
1983, 1985b; Grushvitzky et al. 1987). The
center of diversity for Macropanax is SW
China and Vietnam, where 12 species occur.
Hoo and Tseng (1978) suggested that Macro-
panax might be related to Schefflera, but Wen
et al. (2001a) indicated a close relationship to
Metapanax based on ITS sequences, a finding
confirmed by the present study (Figs. 1–3).
The molecular data are consistent with several
lines of morphological evidence. Synapomor-
phies shared by Macropanax and Metapanax
include evergreen leaves, mostly dentate leaflet
margins, inflorescences comprising paniculate-
ly arranged umbels, articulated pedicels, and
bicarpellate ovaries.
The Brassaiopsis Subgroup: Brassaiopsis
(usually defined to include Euaraliopsis) com-
prises 25–30 species distributed in central
China to SE Asia and westward to the
Himalayan region. Hutchinson (1967) estab-
lished Euaraliopsis to accommodate those taxa
with simple, lobed leaves, restricting Brassai-
opsis to species with palmately compound
leaves. However, he failed to provide a Latin
description for his new genus, which was thus
invalidly published (Article 43.1 in Greuter
et al. 2000). In a little known paper, Banerjee
(1973) validly described Pseudobrassaiopsis to
replace Hutchinson’s nomen nudum. Hoo and
Tseng (1978) incorrectly accepted the generic
status of Euaraliopsis, but most other workers
(e.g. Philipson 1979, Shang 1985a, Hoˆ 1993)
placed it in synonymy under Brassaiopsis.
Brassaiopsis and Pseudobrassaiopsis share
many characteristics, including unjointed ped-
icels, bicarpellate ovaries, undivided stigmas,
and ruminate endosperm.
Hoo and Tseng (1978) suggested that (1)
Pseudobrassaiopsis (cited as Euaraliopsis) may
be closely related to Trevesia because they
share leaves whose lobes may have pseudo-
petiolules (see Hoo and Tseng 1978, pl. 1, f. 7);
(2) Brassaiopsis is perhaps related to or
derived from Schefflera; and (3) Trevesia, with
its 5–12-carpellate ovary, is more ‘‘primitive’’
than Pseudobrassaiopsis (= Euaraliopsis).
Philipson (1979) argued that recognition of
the two genera as distinct unnecessarily frag-
mented a coherent assemblage, and further
suggested a close relationship between Bras-
saiopsis s. lat. and Trevesia based on their
similar vegetative characters. Trevesia com-
prises 10 species with a wide distribution in
SW China, the Himalayas, and extending into
SE Asia. It can be distinguished by its greater
number of locules (5–12 vs. 2–5 in Brassaiop-
sis; Grushvitzky et al. 1984). Brassaiopsis
producta (Dunn) Shang, with ovaries of 3–5
carpels, may provide a bridge between the two
genera (Shang 1985a, b), although locule
number in core Araliaceae is considered rela-
tively plastic (Shang 1985a, Wen 1993, Wen
et al. 2001a). Grushvitzkya Skvortsova & Aver.
is a monotypic genus recently described from
northern Vietnam (Skvortsova and Averyanov
1994). Jebb (1998) considered it to be closely
related to Brassaiopsis and Trevesia, perhaps
forming a ‘‘link’’ between them, as suggested
by its 5-carpellate ovary (vs. 2–5-carpellate in
Brassaiopsis s. lat. and 6–12-carpellate in
Trevesia).
Our phylogenetic results clearly support
the close relationship among Brassaiopsis s.
lat., Grushvitzkya and Trevesia. When Brassai-
opsis and Trevesia were first recognized they
were distinguished by the number of locules.
As additional species were described, however,
the generic limits became blurred, a situation
rendered even more confusing with the recog-
nition of Grushvitzkya. A recent detailed phy-
logenetic analysis based on molecular data
(Wen et al. 2003) has shown that Grushvitzkya
is nested within Brassaiopsis and should be
included therein, and another study (A. Mitch-
ell and J. Wen, unpubl.) indicates that Pseudo-
brassaiopsis is not distinguishable from
Brassaiopsis s. str.
Oreopanax + Sinopanax: The ITS data
presented here (Fig. 2) suggest a sister-group
relationship between the monotypic Sinopanax
of Taiwan and the much larger Neotropical
genus Oreopanax (80 species). The trnL-trnF
phylogeny further indicates that these genera
form a clade with the Asian Fatsia and the
Neotropical Dendropanax arboreus. Sinopanax
28 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
formosana was originally described by Hayata
(1908) as a species of Oreopanax, with which it
shares a number of characters, including
palmately-lobed simple leaves, large terminal
panicles of small capitate inflorescences, and
ruminate endosperm. Li (1949), however,
argued that the Taiwanese species differs in
having a 2-carpellate (vs. 5-carpellate) ovary, a
hermaphroditic (vs. polygamo-dioecious or
polygamo-monoecious) sexual system, and
rather short (vs. long) styles. The status of
Sinopanax needs to be explored further with a
broader sampling of Oreopanax and Dendro-
panax (especially in the New World).
Heteropanax: This genus comprises ca.
five species from southern Asia and China,
and is characterized by bicarpellate ovaries,
divided styles, and leaves that are 2–5 times
pinnately compound. The phylogenetic posi-
tion of Heteropanax within Araliaceae has
been highly controversial (see detailed discus-
sions in Wen et al. 2001a). Our ITS and
combined data place Heteropanax within the
Asian Schefflera subgroup with low to moder-
ate BS support (Figs. 1, 3), but its position was
not resolved in the trnL-trnF phylogeny
(Fig. 2). Additional work is needed to
assess its position using a broader sampling
of both Heteropanax and Asian species of
Schefflera.
The Aralia Group: The phylogenetic posi-
tion of Panax has long been controversial.
Decaisne and Planchon (1854) placed Panax
within Aralia. Bentham (1867) likewise placed
the ginsengs (e.g. Panax quinquefolius and its
relatives) in Aralia, but maintained ‘‘Panax’’
with a very different circumscription that
included species with a woody habit and valvate
petal aestivation (species that are currently
included in Polyscias), and placing their broadly
defined Aralia (with imbricate petals) and
‘‘Panax’’ in different tribes. Hutchinson (1967)
also treated these two genera in different tribes,
but the taxa included in his ‘‘Panax’’ are highly
heterogeneous, including both the herbaceous
ginsengs (Panax s. str., as typified by
P. quinquefolius) and woody species that are
now placed in Polyscias (Philipson 1951).
Harms (1897) regarded Panax as derived
from an herbaceous member of Aralia. Simi-
larly, Hoo (1961) suggested that Panax may
have been derived from Aralia or Acanthopan-
ax (= Eleutherococcus). Wen (1993, see also
Wen and Zimmer 1996, Wen and Nowicke
1999, Wen 2001, Wen et al. 2001a) used
morphological, molecular, and pollen-ultra-
structural data to demonstrate a close rela-
tionship between Panax and Aralia, which
differ primarily in their leaf architecture (pal-
mately compound in Panax vs. bi- or tripin-
nately compound in Aralia) and number of
carpels (2–3 in Panax vs. 5–8 in Aralia),
However, 4–5-carpellate ovaries are common
in Panax as developmental abnormalities
(J. Wen, pers. obs.), and Aralia lihengiana
J. Wen, L. Deng and X. Shi, a recently
described species from western Yunnan, has
a 3-carpellate ovary (Wen et al. 2002). Panax
and Aralia share imbricate floral aestivation,
uniform endosperm, and overall pollen mor-
phology and ultrastructure (Wen and Nowicke
1999), and have similar floral vasculature
(Eyde and Tseng 1971).
Pseudopanax: This genus was established
by Koch (1859) to include woody araliaceous
taxa with 2–5-carpellate ovaries, free styles, a
polygamodioecious sexual system, and articu-
lated pedicels. Seemann (1866a) divided Ko-
ch’s Pseudopanax into three genera based on
the number of locules: Pseudopanax s. str. with
5-carpellate ovaries, Nothopanax Miq. with 2
carpels, and Raukaua with 3–4. Philipson
(1951) showed that Nothopanax was identical
to Polyscias and merged the two. Allan (1961)
described Neopanax to include six species from
New Zealand with a bicarpellate gynoecium
(e.g. N. anomalus and N. laetus), but Philipson
(1965) argued for uniting Neopanax with
Pseudopanax, and expanded the latter to
include two species each from China and one
from New Caledonia. Based on ITS sequences
and morphological data, Mitchell and Wag-
staff (1997) showed that Pseudopanax,as
defined by Philipson (1965), was polyphyletic,
prompting Mitchell et al. (1997) to narrow its
circumscription and reinstate Raukaua (see
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 29
below). Wen and Frodin (2001) transferred the
two Chinese species to a new genus, Metapan-
ax, and Eibl et al. (2001) showed that the New
Caledonian species belongs to Polyscias sect.
Tieghemopanax. The molecular data presented
here (Fig. 3) confirm these decisions. More-
over, they suggest that Pseudopanax (as cur-
rently defined) comprises two distinct clades
within the Pacific Schefflera subgroup (BS =
74%), supporting D. G. Frodin’s idea (pers.
comm.) that Neopanax should be reinstated for
the lineage represented by P. arboreus, P.
laetus and related species (not sampled herein),
which would further narrow the circumscrip-
tion of Pseudopanax and render it monophy-
letic.
Cheirodendron + Raukaua: The sister-
group relationship between Cheirodendron
and Raukaua was first suggested by Mitchell
and Wagstaff (1997), although most tradi-
tional classification systems have placed them
in the same tribe. Raukaua, as reinstated by
Mitchell et al. (1997), comprises three species,
all endemic to New Zealand. Cheirodendron
has 6 species, all of which are endemic to
Hawaii (Lowry 1990) except for C. bastardia-
num (Decne.) D. G. Frodin from the Marque-
sas Islands (Frodin 1990). Although Cheiro-
dendron has often been considered ‘‘isolated’’
in Araliaceae because of its opposite leaves, it
shares many character states with Raukaua.
For example, both genera have palmately
compound leaves (although these are restricted
mostly to the juvenile foliage in Raukaua),
paniculate inflorescences with oppositely ar-
ranged umbellules, and articulated pedicels.
Moreover, the flowers of both genera have
pentamerous perianths and androecia, 2–5-
carpellate ovaries, fruits with laterally-com-
pressed endocarps, and seeds with smooth
endosperm. The relationship between Cheiro-
dendron and Raukaua is moderately supported
by ITS data (BS = 74%; Fig. 1), but is not
fully resolved by trnL-trnF (Fig. 2). In the
combined trees (Figs. 3, 4), however, this
relationship is well supported (BS = 87%;
PP = 100%). Given the recent origin of both
Hawaii and the Marquesas, it seems likely that
the ancestors of Cheirodendron originally dis-
persed eastward from New Zealand (or else-
where in the southwestern Pacific). More
intensive sampling will be needed to determine
the relationship of the single species in the
Marquesas to its congeners in Hawaii.
Cussonia + Seemannaralia: This small
clade includes two genera that are endemic to
or centered in southern Africa. Cussonia
comprises ca. 25 species in Africa, Yemen
and the Comoro Islands (see Lowry et al.
1999), whereas Seemannaralia is a monotypic
genus restricted to two small regions of eastern
South Africa (Burtt and Dickison 1975).
Seemannaralia gerrardii was originally de-
scribed in Cussonia (Seemann 1866b) but
differs by its imbricate petal aestivation, which
prompted Viguier (1906) to erect a new genus,
reflecting the heavy emphasis placed on aesti-
vation type for distinguishing genera and tribes
(Burtt and Dickison 1975, see also Harms
1914). Seemannaralia also differs from most
species of Cussonia in having simple rather
than palmately compound leaves (although the
leaves are palmately veined and lobed). More-
over, their fruits are bicarpellate, dry, and
strongly flattened, producing wing-like mar-
gins, strongly resembling the mericarp of many
Apiaceae traditionally assigned to tribe Peuce-
daneae (e.g. Peucedanum, Heracleum). Burtt
and Dickison’s (1975) study demonstrated,
however, that these resemblances are superfi-
cial and represent a convergence. In Seemann-
aralia the flattening is caused by lateral
compression of both carpels, whereas in
Peucedaneae it is due to dorsal compression
of each mericarp. Lateral flattening is found in
the segregated araliad genera Mackinlaya and
Myodocarpus, and in several groups usually
placed in Apiaceae subfam. Hydrocotyloideae
(Hydrocotyle, Trachymene, Centella, Platysace,
and Xanthosia, see Tseng 1967, Burtt and
Dickison 1975). Recent molecular studies
indicate that most of these hydrocotyloids
should be excluded from the core group of
Apiaceae, with Hydrocotyle and Trachymene
allied to core Araliaceae, and Centella
and Platysace associated with the basally
30 Plunkett et al.: Infrafamilial classifications and characters in Araliaceae
branching Mackinlayeae group (e.g. Plunkett
and Lowry 2001, G. Chandler and Plunkett,
unpubl.). Although the gynoecium of See-
mannaralia is bicarpellate, it produces uni-
carpellate, single-seeded fruits, a character
unknown in Apiaceae but found in a small
number of unicarpellate araliads, including
Arthrophyllum and Cuphocarpus, which are
part of the Polyscias s. lat. clade and likewise
have pinnately compound leaves (Figs. 1–4,
see also Plunkett et al. 2001). In their study of
the anatomy and morphology of Seemannara-
lia, Burtt and Dickison (1975) concluded that
the genus could safely be placed within Aral-
iaceae, but they could not confidently suggest
any close relative. In comparing Seemannaralia
to other lineages, they noted that only Cusso-
nia combines ruminate endosperm, palmate (or
palmatifid) leaves, bicarpelly, and a congruent
geographic distribution. Our molecular data
support this relationship with low to moderate
bootstrap support (62–86%).
Characters considered taxonomically
important. Most classification systems of Aral-
iaceae divide the family into tribes using a
rather small number of morphological charac-
ters (Table 1). The present study offers a
picture of relationships sharply at odds with
these systems, suggesting that the morpholog-
ical features used in traditional classifications
may be poor predictors of phylogeny (at least
at the infrafamilial level). To examine hypoth-
eses of character evolution and to evaluate
their utility in recognizing infrafamilial groups,
we have mapped morphological character
states along the parsimony consensus tree
based on the combined molecular data set
(Fig. 3) using the ‘‘trace character’’ functions
in MacClade 4 (Maddison and Maddison
2000). On the basis of the classification systems
presented in Table 1, we have identified seven
morphological characters that have been con-
sidered important in most or all classifications
of Araliaceae: petal insertion, petal aestivation,
stamen number (relative to petal number),
endosperm texture, carpel number, leaf shape,
and pedicel articulation. Below we present a
brief discussion of how each of these features
appears to have evolved within core Aralia-
ceae.
Petal insertion and aestivation: These char-
acters figure prominently in all seven of the
classification systems presented in Table 1, but
only petal aestivation is variable among the
core genera of Araliaceae. Tribe Mackinlayeae
(Mackinlaya and Apiopetalum) was defined
largely on the basis of its clawed petals, a
feature shared with Apiaceae. The removal of
Mackinlayeae from core Araliaceae (Plunkett
and Lowry 2001), however, leaves the family
characterized entirely by broadly inserted pet-
als, with a single exception, Osmoxylon, where
the petals are basally united into a short corolla
tube (see Philipson 1979). The distribution of
petal aestivation among core Araliaceae is more
complex. This character was weighted heavily
in the nineteenth century systems, particularly
by Seemann (1868), who used it as the primary
feature to recognize two distinct families (the
valvate Hederaceae and the imbricate Aralia-
ceae). In mapping this character along the
molecular topology, valvate petals appear to be
the ancestral state for core Araliaceae (as they
are in most other groups of Apiales), and
imbricate petals have evolved multiple times in
the family. Regardless of how polytomies are
interpreted [either as multiple speciation events
(‘‘hard polytomies’’), or as uncertainties in
resolution (‘‘soft polytomies’’)], imbricate pet-
als arose no fewer than five times indepen-
dently: in each of the four clades of the basal
polytomy (Cephalaralia, Motherwellia, Harm-
siopanax, and Seemannaralia), and in the Aralia
group, within which there is also a reversal back
to valvate petals in the Pentapanax subclade.
Petal aestivation, perhaps in combination with
other features, may be useful in characterizing
relatively small intergeneric groups (such as
Aralia and its relatives), but it is clearly
homoplastic across the entire family.
Number of stamens: This character has
been used in most systems to circumscribe
tribes in Araliaceae. Our analysis shows that
isomery of stamens and petals is both ancestral
and very common in Araliaceae. An increased
number of stamens, particularly in comparison
Plunkett et al.: Infrafamilial classifications and characters in Araliaceae 31
to the number of petals, was used to define
tribe Plerandreae in all systems except those of
Harms (1894–97), who did not recognize
Plerandreae, and of Tseng and Hoo (1982),
whose definition of the tribe was more complex
(see Table 1). There is no indication, however,
that increased stamen number is phylogeneti-
cally informative at the family level. In fact,
increases in stamen number have occurred
independently five times: in the Osmoxylon
clade, in the Asian Schefflera subgroup (Sche-
fflera segregate genus Tupidanthus), in the
Pacific Schefflera subgroup (Plerandra, and
likely also the unsampled New Caledonian
Schefflera plerandroides, which is part of the
Dizygotheca group), and in two subclades of
Polyscias s. lat.
Endosperm texture: This character has
been