ArticlePDF Available

Systematics of Fraxinus (Oleaceae) and evolution of dioecy

Authors:
  • County Administrative Board of Jönköping

Abstract and Figures

Phylogenetic relationships among 40 of the 43 recognized species of Fraxinus L. (Oleaceae) were estimated on the basis of 106 nuclear ribosomal ITS sequences. ITS trees resulting from maximum likelihood (ML), maximum parsimony (MP) and Bayesian inference (BI) are congruent and identify six distinct lineages. These clades allow establishing sections with high molecular and morphological support. The basal resolution generally has low ML bootstrap and MP jackknife support, but the Bayesian posterior probabilities are high for certain relationships. An independent data set of combined sequences from the chloroplast rps16 and trnL-F regions contains few informative sites but corroborate most of the relationships in the ITS tree. The molecular phylogeny is discussed in the light of morphological and other data and a revised infrageneric classification with six sections are presented. The subgenera and subsections are abandoned and the section Pauciflorae is a new combination. Fraxinus quadrangulata and Fraxinus anomala are united with Fraxinus dipetala in the section Dipetalae and Fraxinus platypoda is transferred to the section Fraxinus. Fraxinus chiisanensis, Fraxinus spaethiana and Fraxinus cuspidata are treated as incertae sedis. A sectional key is given, together with a systematic list of the 43 recognized species, with common synonyms and distribution. Breeding system and other traits mapped on the phylogeny show that dioecy has three separate origins, and in each case followed after the transition from insect to wind pollination. In one instance dioecy evolved from hermaphroditism via androdioecy and twice via polygamy.
Content may be subject to copyright.
ORIGINAL ARTICLE
Systematics of Fraxinus (Oleaceae) and evolution of dioecy
Eva Wallander
Received: 1 August 2007 / Accepted: 21 December 2007 / Published online: 18 April 2008
Ó Springer-Verlag 2008
Abstract Phylogenetic relationships among 40 of the 43
recognized species of Fraxinus L. (Oleaceae) were esti-
mated on the basis of 106 nuclear ribosomal ITS
sequences. ITS trees resulting from maximum likelihood
(ML), maximum parsimony (MP) and Bayesian inference
(BI) are congruent and identify six distinct lineages. These
clades allow establishing sections with high molecular and
morphological support. The basal resolution generally has
low ML bootstrap and MP jackknife support, but the
Bayesian posterior probabilities are high for certain rela-
tionships. An independent data set of combined sequences
from the chloroplast rps16 and trnL-F regions contains few
informative sites but corroborate most of the relationships
in the ITS tree. The molecular phylogeny is discussed in
the light of morphological and other data and a revised
infrageneric classification with six sections are presented.
The subgenera and subsections are abandoned and the
section Pauciflorae is a new combination. Fraxinus qua-
drangulata and Fraxinus anomala are united with Fraxinus
dipetala in the section Dipetalae and Fraxinus platypoda is
transferred to the section Fraxinus. Fraxinus chiisanensis,
Fraxinus spaethiana and Fraxinus cuspidata are treated as
incertae sedis. A sectional key is given, together with a
systematic list of the 43 recognized species, with common
synonyms and distribution. Breeding system and other
traits mapped on the phylogeny show that dioecy has three
separate origins, and in each case followed after the tran-
sition from insect to wind pollination. In one instance
dioecy evolved from hermaphroditism via androdioecy and
twice via polygamy.
Keywords Fraxinus Oleaceae ITS Phylogeny
Key Classification Taxonomy Section Pauciflorae
Introduction
The genus Fraxinus L., the ashes, comprises 43 species
occurring in temperate and subtropical regions of the
northern hemisphere. The two main distribution areas are
North America (20 species) and eastern Asia (20 species).
Three species occur in Europe and western Asia. Fraxinus
is one of 24 extant genera of Oleaceae (the olive family)
and sole member of the subtribe Fraxininae, which is sister
group to the subtribe Oleinae in the tribe Oleeae (Wal-
lander and Albert 2000). The genus was described by
Linnaeus in 1753 and since then over 450 taxa have been
described, most of which are regarded as synonyms today.
The latest and most thorough monograph of the entire
genus includes 64 species (Lingelsheim 1920), and more
taxa have been described since then.
The genus is monophyletic (Wallander and Albert 2000)
and unique in the Oleaceae by mostly having relatively
large imparipinnate leaves and one-seeded samaras. Most
of the species are large or medium-sized trees, but some are
shrubs in dry areas. There is much variation in leaf mor-
phology (shape, texture, number of leaflets, leaflet margin,
petiolule length, indumentum, epidermal papillae, rachis
wings, etc.) and intraspecific variations in these features
have been the cause of most synonyms. As it is the char-
acteristic of nearly all taxa of Oleaceae, the small flowers
have only one pistil and two stamens. The corolla may be
lacking or consists of four (rarely two), white, linear, and
E. Wallander (&)
Nature Conservancy, Jo
¨
nko
¨
ping County Administration Board,
SE 551 86 Jo
¨
nko
¨
ping, Sweden
e-mail: eva.wallander@f.lst.se
123
Plant Syst Evol (2008) 273:25–49
DOI 10.1007/s00606-008-0005-3
free (rarely fused) petals. The synsepalous calyx is small,
cup-shaped, and usually dentate, or lacking. The petali-
ferous and insect-pollinated flowers are (with two
exceptions) borne in large showy panicles that emerge
together with the leaves from terminal buds. The apetalous
flowers, which are wind-pollinated, occur in lateral or
terminal inflorescences and emerge before the leaves
unfold. The syncarpous ovary contains four ovules, two in
each locule, but normally develops into a one-seeded
samara.
The genus has traditionally been divided into two
sections or subgenera on the basis of morphology
(Table 1). The section or subgenus Fraxinus (Fraxinaster
is an invalid name according to ICBN 22.1) comprised all
taxa with lateral inflorescences, whereas the section or
subgenus Ornus comprised taxa where the flowers are
borne in terminal panicles together with the leaves. The
section Fraxinus was further divided into five subsections
on the basis of presence or absence of calyx, number of
petals, and winged or unwinged leaf petioles. The section
Ornus was subdivided into two subsections, Ornus (Eu-
ornus is an invalid name according to ICBN 21.3) with
petals and Ornaster without petals. This classification has
remained relatively stable, although some authors after
Lingelsheim (1920) have chosen other ranks and cor-
rected the names for some infrageneric groups. The
classification of a few taxa has been uncertain, e.g.
Fraxinus chiisanensis, Fraxinus cuspidata, Fraxinus
platypoda, and Fraxinus anomala, but most of the dif-
ferences between the classifications have concerned the
number of included species and different opinions about
the synonymy.
There is a variety of pollination and breeding systems
within the genus Fraxinus. This situation offers an
interesting case for studying the evolution of traits related
to reproductive biology. About one-third of the species
are entomophilous and two-thirds are anemophilous. Most
of the anemophilous species are dioecious or polygamous.
A few of the entomophilous species are hermaphrodites,
but the majority are androdioecious. This is considered to
be a very rare breeding system (e.g. Charlesworth 1984;
Pannell 2002), but in Oleaceae, and particularly in
Fraxinus, there are many species that are morphologically
androdioecious.
Jeandroz et al. (1997) published the first molecular
phylogeny of Fraxinus, which was based on ITS-1 and
ITS-2 sequences of the nuclear ribosomal DNA from 20
species. Because they only included less than half of the
number of species, and left out representatives of the two
sections Pauciflorae and Sciadanthus, their phylogeny
could not be used for the purpose of this study. Although
they used their phylogenetic tree to map some floral
characters (presence or absence of calyx and/or corolla), a
Table 1 Different classification schemes of Fraxinus, including the
revised one proposed in this study
Lingelsheim (1920), Rehder (1940), Dayton (1954), Miller (1955)
Sect. Fraxinaster DC.
Subsect. Bumelioides (Endl.) Lingelsh.
Subsect. Melioides (Endl.) Lingelsh.
Subsect. Sciadanthus (Coss. et Dur.) Lingelsh.
Subsect. Dipetalae Lingelsh.
Subsect. Pauciflorae Lingelsh.
Sect. Ornus (Neck.) DC.
Subsect. Euornus Lingelsh.
Subsect. Ornaster (Koehne et Lingelsh.) Lingelsh.
Vassiljev (1952)
a
Subgenus Fraxinaster (DC.) V. Vassil.
Sect. Melioides (Endl.) Pfeiff.
Sect. Bumelioides (Endl.) Pfeiff.
Subgenus Ornus (Boehm.) Pers.
Sect. Euornus Koehne et Lingelsh.
Sect. Ornaster Koehne et Lingelsh.
Nikolaev (1981)
Subgenus Fraxinus
Sect. Fraxinus
Subsect. Paniculatae E. Nikolaev
Subsect. Racemosae E. Nikolaev
Sect. Melioides (Endl.) Pfeiff.
subsect. Melioides (Endl.) Lingelsh.
Subsect. Sciadanthus (Coss. et Dur.) Lingelsh (incl. subsect.
Pauciflorae Lingelsh.)
Sect. Dipetalae (Lingelsh.) E. Nikolaev
Subgenus Ornus (Boehm.) Pers.
Sect. Ornus (Boehm.) DC.
Sect. Ornaster Koehne et Lingelsh.
Wei (1992)
a
Subgenus Fraxinus
Sect. Fraxinus
Sect. Melioides (Endl.) Pfeiff.
Sect. Sciadanthus Coss. et Dur.
Subgenus Ornus (Boehm.) Pers.
Sect. Ornus (Boehm.) DC.
Sect. Ornaster Koehne et Lingelsh.
Wallander (this study)
Sect. Dipetalae (Lingelsh.) E. Nikolaev
Sect. Fraxinus
Sect. Melioides (Endl.) Pfeiff.
Sect. Ornus (Boehm.) DC.
Sect. Pauciflorae (Lingelsh.) E. Wallander
Sect. Sciadanthus (Coss. et Dur.) Lingelsh.
a
Vassiljev (1952) and Wei (1992) did not include Dipetalae or
Pauciflorae because these sections are not native to the regions they
covered. Earlier classifications are detailed by Miller (1955). Several
author citations have been corrected due to errors in original
publications
26 Plant Syst Evol (2008) 273:25–49
123
number of coding errors made their interpretation of
character evolution incorrect.
The main objective of this study in the genus Fraxinus
was twofold: first, to estimate the phylogeny of the entire
genus Fraxinus on the basis of molecular data and; sec-
ond, to use this phylogenetic estimate to study the
evolution of wind pollination and related traits in the
genus. In particular, I have been interested in how
the evolution of unisexual flowers correlates with that of
wind pollination. During the course of the phylogenetic
work, new relationships among taxa were discovered
which together with a deeper morphological study led to a
revised infrageneric classification of Fraxinus. This article
presents a well-supported phylogeny of the genus Fraxi-
nus on the basis of DNA sequences from the nuclear
ribosomal ITS and two chloroplast regions, and a revised
classification. Some traits relating to pollination systems
are mapped on the tree and briefly commented, but the
interpretations of the floral evolution stemming from this
work, in relation to transitions between pollination sys-
tems, are discussed in more detail elsewhere (Wallander
2001).
Materials and methods
Materials
First, a thorough inventory of the Fraxinus taxa accepted
in various recent treatments and their putative synonyms
was undertaken to come up with a reasonable represen-
tation of taxa to include in this study. This inventory was
based on original descriptions, floristic treatments (Sar-
gent 1949; Little 1952; Vassiljev 1952; Dayton 1954;
Murray 1968; Franco and Rocha Afonso 1972; Nakaike
1972; Grohmann 1974; Scheller 1977; Yaltirik 1978;
Nikolaev 1981; Vines 1984; Yamazaki 1993; Wei and
Green 1996), regional monographs (mainly Standley
1924; Miller 1955; Hara 1956, 1982; Kitagawa 1979; Sun
1985; Green 1991), and the only monograph of the genus
(Lingelsheim 1920). Many taxa were studied in the field
and in botanical gardens and arboreta (Go
¨
teborg Botanical
Garden and Arboretum, New York Botanical Garden,
Missouri Botanical Garden and Shaw Arboretum, Palermo
Botanical Garden, Kyoto Botanical Garden, and the Royal
Botanic Gardens at Kew). Over 1,000 herbarium speci-
mens from BM, C, E, GB, K, MO, NY, S, and UPS
(acronyms according to Index Herbariorum) were also
studied. This work resulted in a provisional list of about
50 species, 43 of which I eventually accepted (on the
basis of morphology, molecular data, and opinions of
other authors, as explained in Discussion’) (Table 2).
On the basis of this list, at least two representatives of
each recognized species, and in some cases subspecies,
were chosen for DNA sequencing. In addition, some taxa
of uncertain status (putative synonyms) were chosen to
estimate their relationships.
Fresh or silica-gel dried leaf or seed material from many
taxa cultivated in botanical gardens, and from field col-
lections made in Spain, Italy, the USA, Mexico, China, and
Japan, were used for the molecular work. Vouchers for
these are deposited at GB. Materials from the aforemen-
tioned herbaria were also used. The final number of
ingroup specimens sequenced for this study was 89, which
represents 40 of the 43 recognized species. Amplification
failed for several specimens each of the three missing
species (Fraxinus griffithii, Fraxinus malacophylla, and
Fraxinus baroniana). Seventeen ITS-1 and ITS-2 sequen-
ces from the study by Jeandroz et al. (1997) were taken
from GenBank and included in the analysis. Five outgroup
taxa were chosen from the closely related subtribes Ligu-
strinae and Oleinae, on the basis of the Oleaceae phylogeny
of Wallander and Albert (2000). All vouchers and Gen-
Bank accession numbers are listed in the Appendix’, and
all accepted species and relevant synonyms are given with
authors in Table 2.
Molecular methods
DNA extraction, PCR amplification, and automated
sequencing were mainly done using methods and equip-
ment described by Wallander and Albert (2000). In
addition, a few samples were extracted using the DNeasy
Plant Mini kit (QIAGEN) and PCR performed using the
HotStarTaq
Ò
Master Mix kit (QIAGEN) without modifi-
cations. Amplifications were carried out in a GeneAmp
Ò
PCR System 9700 (Perkin-Elmer Applied Biosystems) and
PCR products sequenced on a CEQ
TM
8000 Genetic
Analysis System (Beckman Coulter). The ITS4 and ITS5
primers designed by Nickrent et al. (1994) and Wojcie-
chowski et al. (1993) were used to amplify the entire ITS
region of the nuclear ribosomal DNA. In some difficult
cases, the ITS-1 and ITS-2 regions were amplified and
sequenced separately using the internal primers ITS2 and
ITS3 of Wojciechowski et al. (1993).
The forward and reverse ITS sequences were assem-
bled and edited using Sequencher
TM
4.1.2 (Gene Codes
Corporation, Ann Arbor, MI, USA). Consensus sequen-
ces were aligned using the alignment feature in
Sequencher and then manually adjusted. Furthermore, the
rps16 and trnL-F intron sequences of ten Fraxinus spe-
cies from the study by Wallander and Albert (2000) were
combined with five new ones (obtained in the same way)
and two outgroup species into one data matrix, repre-
senting all sections. The alignments can be obtained from
GenBank.
Plant Syst Evol (2008) 273:25–49 27
123
Table 2 Revised infrageneric classification of Fraxinus (Oleaceae) listing the 43 accepted species and their geographical distribution, together
with common synonyms or those mentioned in this study
Sections and species Geographic distribution Synonyms
Section Dipetalae (Lingelsh.) E. Nikolaev
F. anomala Torr. ex S. Wats. SW USA F. lowelli Sarg., F. potosina T. S. Brandeg.
F. dipetala Hook. and Arn. SW USA F. jonesii Lingelsh., F. parryi Moran, F. trifoliata (Torr.) Lewis
and Epling
F. quadrangulata Michx. C and E USA, C Canada
Section Fraxinus
F. angustifolia Vahl S and C Europe to Central Asia F. oxycarpa Willd., F. oxyphylla M. Bieb. (nom. illeg.), F. pallisiae
A.J. Willmott, F. potamophila Herder, F. sogdiana Bunge,
F. syriaca Boiss.
F. excelsior L. N and C Europe to W Russia F. coriariifolia Scheele
F. mandshurica Rupr. China, Japan, Korea, E Russia F. nigra ssp. mandshurica (Rupr.) S. S. Sun
F. nigra Marsh. E USA, E Canada
F. platypoda Oliv. China
Section Melioides (Endl.) Lingelsh.
F. americana L. E USA and E Canada F. biltmoreana Beadle
F. berlandieriana DC. SW USA, Mexico
F. caroliniana Mill. SE USA F. cubensis Griseb.
F. latifolia Benth. W USA F. oregona Nutt.
F. papillosa Lingelsh. SW USA, Mexico
F. pennsylvanica Marsh. C and E USA, Canada
F. profunda (Bush) Bush SE USA F. tomentosa Michx. f. (nom. rej.)
F. texensis (Gray) Sarg. SW USA (Texas)
F. uhdei (Wenzig) Lingelsh. C America, Hawaii F. cavekiana Standley and Steyerm., F. chiapensis Lundell,
F. hondurensis Standley
F. velutina Torr. SW USA, Mexico F. attenuata M. E. Jones, F. pistaciaefolia Torr., F. toumeyi Britt.
Section Ornus
(Boehm.) DC.
F. apertisquamifera Hara Japan
F. bungeana DC. China
F. floribunda Wall. Himalaya, E Asia F. insularis Hemsl., F. retusa Champ. ex Benth.
F. griffithii C. B. Clarke SE Asia F. ferruginea Lingelsh., F. formosana Hayata, F. philippinensis Merr.
F. lanuginosa Koidz. Japan
F. malacophylla Hemsl. China, Thailand F. retusifoliolata Feng ex P. Y. Bai
F. ornus L. C and E Mediterranean
F. paxiana Lingelsh. Himalaya, China F. sikkimensis (Lingelsh.) Hand.-Mazz., F. suaveolens
W.W. Smith
F. raibocarpa Regel C Asia
F. sieboldiana Blume China, Japan, Korea
F. trifoliolata W. W. Smith China
F. baroniana Diels China
F. chinensis Roxb. E Asia F. japonica Blume ex K. Koch, F. rhynchophylla Hance
F. longicuspis Sieb. and Zucc. Japan
F. micrantha Lingelsh. Himalaya
Section Pauciflorae (Lingelsh.) E. Wallander, stat. nov.
F. dubia (Willd. ex Schult. and
Schult. f.) P. S. Green and M. Nee
Mexico, Guatemala F. petenensis Lundell, F. schiedeana Schlecht. and Cham.
F. gooddingii Little SW USA, N Mexico
F. greggii A. Gray SW USA, Mexico
F. purpusii Brandegee Mexico, Guatemala F. bicolor Standley and Steyerm., F. vellerea
Standley and Steyerm.
28 Plant Syst Evol (2008) 273:25–49
123
Phylogenetic analyses
The final ITS data matrix contained 106 ingroup sequences,
representing all but three of the 43 recognized species and
some of their putative synonyms, plus five outgroup
sequences. A parsimony analysis were performed using
heuristic searches in PAUP* 4.0b10 (Swofford 2002). The
search consisted of TBR branch-swapping of 99 random
addition sequence replicates limited to a maximum of
1,000 trees saved per replicate. All characters were con-
sidered to be unordered and given equal weight. Although
some indels appeared to be phylogenetically informative,
they were not separately coded. Gaps were treated as
missing data. Multistates were interpreted as uncertainties
(even if many appeared to be true polymorphisms). Branch
support was evaluated through parsimony jackknifing using
XAC (James S. Farris, Swedish Museum of Natural His-
tory, Stockholm) with 1,000 replicates, each with ten
random addition sequence replicates and non-rotational
branch-swapping. An additional output with GC values
(Goloboff et al. 2003) provided a measure of group fre-
quency minus greatest frequency of a conflicting group.
In addition to the parsimony analyses, a maximum like-
lihood analysis was performed as implemented in the
PhyML online web server (Guindon et al. 2005). Because of
computational limitations, the 111-taxon dataset had to be
reduced to 108. The three OTUs removed were all sequences
that were nearly identical duplicates of another conspecific
sequence. Although the GTR + G nucleotide substitution
model was selected by MrModeltest 2.2 (Nylander 2004), the
GTR model and other parameters estimated were used as
input values. The reliability of the internal nodes was esti-
mated through 100 parametric bootstrap replicates.
Node support was also established using Bayesian
inference, as implemented in MrBayes 3.1 (Huelsenbeck
and Ronquist 2001; Ronquist and Huelsenbeck 2003). First,
the partition homogeneity test (Farris et al. 1994), as
implemented in PAUP* version 4.0b10 (Swofford 2002),
was applied to test the congruence of results produced
independently from the three partitions defined for the data
set (ITS1: matrix positions 1–269, 5.8S: 270–428, and
ITS2: 429–663). For this purpose, a heuristic search was
performed with 1,000 replicates, 100 random addition
sequences, TBR branch swapping, and saving up to 50 trees
per replicate. MrModeltest 2.2 (Nylander 2004) was used to
determine the optimal nucleotide substitution model for
each of the three regions analyzed. Following the recom-
mendations of recent works (Pol 2004; Posada and Buckley
2004), the evolutionary models chosen by the Akaike
information criterion, the GTR + G model for ITS1 and
ITS2 and the k80 + I model for the nearly invariable 5.8S
region, were then incorporated into a MrBayes block in the
input file. The program performed two simultaneous runs
until the average standard deviation of split frequencies
became lower than 0.01. For each run, eight Metropolis-
coupled Markov chain Monte Carlo (MCMCMC) chains
were initiated, sampling every 500 generations, saving
branch lengths, and using other default settings.
The chloroplast data set (15 ingroup and two outgroup
sequences of the rps16 intron and the trnL-F region com-
bined) was analyzed separately to provide an independent
estimate of the relationships between the sections. The
parsimony analysis consisted of 1,000 random addition
sequence replicates and the number of saved trees per
replicate was not limited. A maximum likelihood analysis
was also carried out using the PhyML online web server.
The HKY nucleotide substitution model was selected by
MrModeltest 2.2 (Nylander 2004) and the parameters
estimated were used as input values. ITS sequences for the
same 17 species were analyzed in the same way (but with
the GTR model in the ML analysis) and the variable
characters of the chloroplast data set were then optimized
onto the ML tree using the software MacClade 4.08.
Morphological data
Herbarium material was studied for all species and some
were also studied in the field. Some information was also
Table 2 continued
Sections and species Geographic distribution Synonyms
F. rufescens Lingelsh. Mexico
Section Sciadanthus (Coss. et Dur.) Lingelsh.
F. hubeiensis S. Z. Qu, C. B. Shang
and P. L. Su
China
F. xanthoxyloides (G. Don) DC. N Africa to China F. dimorpha Coss. and Dur.
Incertae sedis
F. cuspidata Torr. SW USA, Mexico
F. chiisanensis Nakai Korea
F. spaethiana Lingelsh. Japan
Plant Syst Evol (2008) 273:25–49 29
123
gathered from the literature (see references for materials).
Characters considered important in previous classifications
(inflorescence position, number of petals, presence or
absence of calyx, and presence or absence of leaf rachis
wings) are mapped on a summary of the phylogenetic trees.
Pollination and breeding system are also mapped to show
how breeding systems have evolved in relation to polli-
nation system and related floral traits. MacClade 4.08 was
used to optimize the traits on the tree. Both ACCTRAN
and DELTRAN resolving options were tested, but since
there were no differences only one tree is shown.
Results
Phylogenetic analyses and congruence
between inference methods
The sequence characteristics for all data sets and partitions
are presented in Table 3. The partition homogeneity test did
not reach significance for the rejection of congruency
among the data partitions (P = 0.50). The Bayesian anal-
ysis was stopped after ten million generations, i.e. long after
the standard deviation of split frequencies of the two par-
allel runs had reached the critical value (0.01). The burn-in
phase was defined as the first 1 million generations, long
after the standard deviation of split frequencies had reached
0.05. Figure 1 shows the majority-rule consensus tree of
36,002 trees (sampled from a total of 40,002 trees from both
parallel runs) estimated using Bayesian inference (BI).
The maximum likelihood (ML) analysis yielded a tree
that is largely identical to the BI tree. The maximum par-
simony (MP) analysis resulted in 47,863 most
parsimonious trees. The strict consensus of these trees
contains the same major clades (=sections) as obtained by
BI and ML, but with poor resolution between them. The
MP and ML trees are not shown, but their jackknife and
bootstrap support values, respectively, are shown along
with the posterior probabilities on the BI tree (Fig. 1).
There is no conflict between the trees yielded by MP
compared to BI and ML, in the sense that there are no
major clades that are strongly supported in one tree but
contradicted in the other. All three inference methods find
the same major clades, which are strongly supported by
bootstrap, jackknife as well as posterior probability values,
but the difference between the results lies in that the strict
consensus of the MP trees is unresolved at the base
whereas most of the basal resolution of the Bayesian tree
(Fig. 1) is strongly supported.
In the following, support will be referred to as strong for
a posterior probability [0.91 or a jackknife and bootstrap
value [88%. These values have been shown to represent
Table 3 Sequence characteristics of the ITS data (106 ingroup and five outgroup taxa) and the combined chloroplast data set of the trnL-F
region and the rps16 intron (15 ingroup and 2 outgroup taxa)
Data parameters ITS region ITS1 ITS2 5.8S trnL-F + rps16
Sequence length range (bp)
a
611–638 238–252 205–228 159 1677–1690
Aligned length (bp) 663 269 235 159 859 + 831
Mean GC content (%) (range) 61 (54–67) 63 (54–74) 63 (53–71) 53 (50–55) 35 (34–36)
Number of informative sites (within ingroup) 220 (185) 114 (99) 99 (80) 7 (6) 17 (14)
Parsimony result
Number of MP trees 47863 18
Length of MP trees 918 70
RI (retention Index) 0.86 0.97
Likelihood result
Likelihood -6005 (-2801
b
) -2826
Some results from the parsimony and maximum likelihood analyses of the data sets are also listed
a
Sequences that are not full length (i.e. a stretch missing in the beginning and/or end) are excluded from the count
b
Likelihood value for the ITS tree with the same 17 taxa as in the chloroplast data set
Fig. 1 Majority rule consensus tree resulting from the Bayesian
analysis of 111 ITS sequences representing 40 Fraxinus species and
five Oleaceae outgroup species. For relevant nodes only, Bayesian
posterior probabilities are shown above the branches and parsimony
jackknife/maximum likelihood bootstrap values for the same data set
below the branches. A dash is shown instead of a jackknife value for
those branches that are collapsed in the strict consensus of the most
parsimonious trees. Asterisks denote clades that have additional
support from indels (that were not separately coded). In the few cases
where multiple sequences representing the same species do not group
together, they have been assigned numbers (13) to provide an option
to trace their vouchers in the Appendix’. GB after taxon names
indicates that those sequences were taken from GenBank (study by
Jeandroz et al. 1997). Names within parentheses are original
determinations of some vouchers, which are here regarded as
synonyms. Sectional assignments are according to the revised
classification in this study
c
30 Plant Syst Evol (2008) 273:25–49
123
Ligustrum vulgare
Syringa vulgaris
Forestiera acuminata
Osmanthus fragrans
Phillyrea latifolia
F. dipetala
F. dipetala
F. dipetala (F. jonesii)
F. anomala
F. anomala
F. quadrangulata
F. quadrangulata
F. quadrangulata GB
F. chiisanensis
F. cuspidata GB
F. cuspidata
F. cuspidata
F. spaethiana
F. spaethiana
F. spaethiana
F. americana
F. americana GB
F. uhdei
F. uhdei
F. uhdei
F. papillosa
F. papillosa
F. latifolia
F. latifolia
F. latifolia GB
F. americana (F. biltmoreana) GB
F. caroliniana 2
F. pennsylvanica 2
F. texensis 2
F. velutina
F. velutina
F. profunda (F. tomentosa) GB
F. pennsylvanica GB
F. velutina GB
“F. anomala” GB
F. berlandieriana
F. berlandieriana
F. caroliniana 1
F. caroliniana (F. cubensis)
F. pennsylvanica 1
F. texensis 1
F. dubia
F. dubia
F. dubia?
F. purpusii 3
F. dubia? (F. schiedeana)
F. greggii 2
F. greggii? 3
F. rufescens
F. gooddingii
F. greggii 1
F. purpusii 1
F. purpusii 2
F. hubeiensis
F. xanthoxyloides
F. xanthoxyloides
F. nigra
F. nigra
F. nigra GB
F. platypoda GB
F. mandshurica
F. mandshurica GB
F. mandshurica
F. platypoda
F. excelsior ssp. coriariifolia
F. excelsior
F. excelsior var. diversifolia
F. angustifolia (F. pallisiae) 2
F. angustifolia ssp angustifolia
F. angustifolia ssp oxycarpa 2
F. angustifolia ssp oxycarpa 1
F. angustifolia (F. sogdiana)
F. angustifolia (F. oxyphylla) GB
F. angustifolia (F. syriaca) GB
F. angustifolia (F. pallisiae) GB
F. angustifolia (F. potamophila)
F. angustifolia ssp syriaca
F. angustifolia (F. pallisiae) 1
F. raibocarpa
F. raibocarpa
F. trifoliolata
F. paxiana
F. paxiana (F. sikkimensis)
F. sieboldiana
F. sieboldiana
F. apertisquamifera
F. apertisquamifera
F. lanuginosa
F. lanuginosa
F. bungeana
F. bungeana
F. bungeana
F. ornus
F. ornus
F. floribunda
F. floribunda (F. retusa)
F. longicuspis
F. longicuspis
F. micrantha
F. micrantha
F. chinensis GB
F. chinensis
F. chinensis (F. rhynchophylla)
F. chinensis (F. japonica)
F. chinensis (F. japonica)
F. lon
g
icuspis GB
1.0
*
1.0
1.0
1.0
0.92
1.0
0.90
1.0
*
1.0
1.0
1.0
0.97
1.0
1.0
1.0
Melioides
Ornus
Fraxinus
Pauciflorae
Dipetalae
Sciadanthus
.99*
1.0
59/53
92/84
1.0
0.1 substitutions/site
0.56
100/92
96/95
<50/44
<50/62
98/100
81/70
100/100
61/90
-/40
98/100
100/100
100/100
68/80
1.0
.98
51/80
<50/45
93/95
92/90
76/70
-/60
Plant Syst Evol (2008) 273:25–49 31
123
minimal values required for a 95% confidence interval of a
node under certain circumstances (Zander 2004).
Congruence between the two independent data sets
The ML tree of the combined rps16 and trnL-F chloroplast
data set is shown in Fig. 2a. The MP analysis of the same
data set resulted in 18 MP trees (data characteristics in
Table 3) and the strict consensus of these trees (not shown)
is largely congruent with the ML tree. The only difference
in topology between them is that Fraxinus greggii and
F. chiisanensis are resolved as a sister group to the section
Melioides in the ML tree whereas they are unresolved in
the MP tree.
The ML chloroplast tree is largely congruent with, but
less resolved than, the ITS tree produced by ML (Fig. 2b)
based on the same taxon sampling. Here the variable sites
of the chloroplast data are mapped on the internal branches
to visualize the congruence between the data sets in another
way. Of the 17 informative sites of the cp data, 15 are
unambiguously changing on the internal branches of the
ML ITS tree. Only F. greggii has a position that is not
congruent in the chloroplast compared to the ITS tree. In
both trees, Fraxinus raibocarpa (which is classified under
the section Ornus on the basis of morphology) is resolved
as sister to the three sections Fraxinus, Sciadanthus and
Ornus, instead of sister to only the rest of section Ornus (as
in the complete ITS tree). The positions of Fraxinus spa-
ethiana and F. cuspidata are not resolved in the chloroplast
tree, but the section Dipetalae, in particular, receives
strong support in both trees.
Congruence between the phylogenetic estimate
and other data
The ITS trees produced by all three methods of phylo-
genetic inference identify six major clades (the sections
indicated in Fig. 1). These clades are all well supported
by MP jackknife, ML bootstrap values as well as
Bayesian posterior probabilities, except for the low sup-
port for the inclusion of F. raibocarpa in the section
Ornus. These clades also agree well with morphological
data, which are given for each section in the taxonomic
discussion below, and some characters are mapped on the
tree in Fig. 3. A few unexpected relationships were dis-
covered in the analyses, such as the three species in
section Dipetalae and the positions of F. cuspidata and
F. platypoda. However, when viewing morphology and
other data such as leaf flavonoid patterns, ecology,
and geographical distribution in the light of molecular
data, most of the previously unreliably placed taxa could
be classified as discussed below.
Melioides
Dipetalae
Ornus
Fraxinus
Sciadanthus
Pauciflorae
95
97
Syringa vulgaris
Phillyrea latifolia
Fraxinus cuspidata
Fraxinus spaethiana
Fraxinus dipetala
Fraxinus anomala
Fraxinus quadrangulata
Fraxinus americana
Fraxinus pennsylvanica
Fraxinus chiisanensis
Fraxinus greggii
Fraxinus raibocarpa
Fraxinus xanthoxyloides
Fraxinus chinensis
Fraxinus ornus
Fraxinus excelsior
Fraxinus ni
g
ra
89
65
80
72
85
89
76
95
a
Syringa vulgaris
Phillyrea latifolia
Fraxinus dipetala
Fraxinus anomala
Fraxinus quadrangulata
Fraxinus americana
Fraxinus pennsylvanica
Fraxinus chiisanensis
Fraxinus cuspidata
Fraxinus spaethiana
Fraxinus greggii
Fraxinus raibocarpa
Fraxinus chinensis
Fraxinus ornus
Fraxinus excelsior
Fraxinus nigra
Fraxinus xanthoxyloides
Melioides
Dipetalae
Ornus
Fraxinus
Sciadanthus
Pauciflorae
3
3
2
1
2
2
2
99
78
100
98
51
67
94
96
43
44
32
24
40
39
b
Fig. 2 a Maximum likelihood tree of the chloroplast data set (rps16
intron and trnL-F region) from 15 Fraxinus and two Oleaceae
outgroup species. Bootstrap values are shown above the branches.
b Maximum likelihood tree of a reduced set of the ITS data with the
same taxon sampling as the chloroplast data set. Bootstrap values are
shown below the branches. Numbers above branches are number of
unambiguous changes in the chloroplast data supporting that branch
(not shown at tips). Sectional assignments are according to the revised
classification in this study
32 Plant Syst Evol (2008) 273:25–49
123
There are two different 8 bp indels in the ITS data set
and one in the trnL-F data set. These indels were not
separately coded in the data matrices, but an a posteriori
mapping of them on the tree shown in Fig. 1 gives addi-
tional support for the following clades. The section
Dipetalae is supported by an 8 bp deletion in the trnL-F
data, all taxa of the section Pauciflorae share an 8 bp
deletion in the ITS data, and the four Eurasian species of
the section Fraxinus are also supported by another 8 bp
deletion in the ITS data.
Two or more accessions of the same species did not
group together in six cases, three in the section Melioides
and three in Pauciflorae (not counting some sequences
from the study by Jeandroz et al. (1997) that may be
misidentified). This includes the ITS sequence of an
additional specimen of Fraxinus americana (added at the
last minute and not included in the Bayesian analyses
presented in Fig. 1) that does not group with the other two
F. americana in the tree. Instead, it appears in the Fraxinus
pennsylvanica clade.
New classification
On the basis of the morphological study and information
from the molecular phylogeny, I recognise 43 species of
Fraxinus (Table 2). Forty of these are classified into six
F. dipetala
F. quadrangulata
F. anomala
F. chiisanensis
F. americana
F. uhdei
F. papillosa
F. latifolia
F. caroliniana
F. texensis
F. pennsylvanica
F. velutina
F. profunda
F. berlandieriana
F. dubia
F. purpusii
F. rufescens
F. raibocarpa
F. paxiana
F. trifoliolata
F. sieboldiana
F. apertisquamifera
F. lanuginosa
F. bungeana
F. ornus
F. floribunda
F. micrantha
F. excelsior
terminal infl.
corolla
dioecy
dioec
y
dioecy
androdioecy
polygamy
hermaphroditism
no corolla
corolla
F. cuspidata
F. spaethiana
calyx
no calyx
hermaphroditism
lateral inflorescences
pollination system
entomophily
anemophily
F. greggii
F. gooddingii
F. longicuspis
F. chinensis
F. xanthoxyloides
F. hupehensis
F. nigra
F. angustifolia
F. platypoda
F. mandshurica
no
no corolla
rachis
winged
rachis winged
corolla
unwinged leaf rachis
geographic distribution
Old World
New World
Fraxinus
Pauciflorae OrnusMelioides
Sciad-
anthus
Dipetalae
Fig. 3 Summary of the Fraxinus phylogeny with some morpholog-
ical characters mapped. It is based on the tree in Fig. 1, but F.
cuspidata is here placed in a more parsimonious position based on
morphological data (see Discussion’). Some species relationships
within the sections Melioides and Pauciflorae were arbitrarily
resolved. The outgroup is not shown here. Characters mapped were
considered important in previous classifications (inflorescence posi-
tion, presence or absence of corolla, presence or absence of calyx, and
presence or absence of leaf rachis wings). Breeding systems are also
mapped to show how they have evolved in relation to pollination
mode and related floral traits. Pollination mode is shown as black
(anemophily) or white (entomophily) lines. Geographic distribution is
shown with black (New World) or white (Old World) boxes below the
taxon names. Sectional assignments are according to the revised
classification in this study
Plant Syst Evol (2008) 273:25–49 33
123
sections (Dipetalae, Fraxinus, Melioides, Ornus, Pauci-
florae and Sciadanthus), and three species are left unplaced
(F. chiisanensis, F. spaethiana, and F. cuspidata).
Although the sectional circumscription mostly is the same
as in previous classifications (Table 1), the subgenera are
abandoned and three species (F. anomala, Fraxinus qua-
drangulata, and F. platypoda) need to be transferred to
other sections as discussed further below. Some diagnostic
characters of the sections are found in the key (Fig. 4) and
in Fig. 3.
Floral evolution
Some morphological characters presumed to be taxo-
nomically informative in previous classifications and
those found interesting for the evolution of pollination
1. Inflorescences emerging with the leaves on current year shoots from terminal buds; flowers
with four, free, white petals, or apetalous, always with calyx; hermaphrodites or
androdioecious; 15 Old World species ………………………..….…… section Ornus
1. Inflorescences emerging from lateral buds on previous year’s shoot, before or at the same
time as the leaves emerge from terminal buds; flowers with 2-4 united petals or apetalous,
with or without calyx; hermaphrodites, polygamous, or dioecious; New World or Old
World species
2. Stems quadrangular; flowers mostly bisexual, with 2 united petals or without petals; 3
New World species ………………………………………..……. section Dipetalae
2. Stems terete; flowers uni- or bisexual, with 4 united petals or without petals; New World
or Old World species
3. Corolla with 4 united petals; 1 New World species ………………… F. cuspidata
3. Flowers without petals; New World or Old World species
4. Shrubs or small trees; leaf rachis winged; flowers with calyx; polygamous
5. Few-flowered, cymose panicles; samaras small (1.5-2.5 cm); 5 New World
species ………………………………………………… section Pauciflorae
5. Many-flowered and dense, cymose panicles; samaras large (3-5 cm); 2 Old
World species …………………………………………. section
Sciadanthus
4. Mostly large trees; leaf rachis not winged; flowers with or without calyx; dioecious
or polygamous
5. Flowers without calyx or with small and/or deciduous calyx; polygamous or
dioecious with rudimentary stamens; seed cavity of samara flattened; 4 Old
World species and 1 New World ……………………….. section Fraxinus
5. Flowers with calyx; strictly dioecious; seed cavity of samara terete; 10 New
World s
p
ecies …………………………………………… section Melioides
Fig. 4 Key to the sections of
Fraxinus. Two of the three
species incertae sedis,
F. chiisanensis (Korea) and
F. spaethiana (Japan), are not
covered in this sectional key
because some of their traits are
found in both the sections
Fraxinus and Melioides
34 Plant Syst Evol (2008) 273:25–49
123
systems are mapped on the summary tree in Fig. 3. There
are no differences between the ACCTRAN or DELTRAN
optimizations. The ancestral states (determined through
outgroup comparison) are unwinged leaf rachis, lateral
inflorescences, hermaphrodite flowers with corolla and
calyx. The calyx has been lost once (in section Fraxinus)
and the corolla three times (near the base of the tree and
within each of the sections Dipetalae and Ornus). Corolla
also appears to have been regained once in the section
Ornus, coinciding with a shift from lateral to terminal
inflorescences, but later lost again in the wind pollinated
species. Dioecy has evolved via polygamy in the sections
Melioides and Fraxinus and via androdioecy in the sec-
tion Ornus. In all cases, dioecy followed after the
transition from insect to wind pollination.
Discussion
The phylogenetic estimate and its support
The ITS phylogeny identifies six major clades (the sections
Dipetalae, Melioides, Pauciflorae, Fraxinus, Sciadanthus
and Ornus), which mostly receive strong support from
bootstrapping, jackknifing as well as Bayesian posterior
probabilities. But the relationships between the sections are
less clear. The resolution of the ITS tree receives high
support from posterior probabilities but much less so from
ML bootstrapping and MP jackknifing. Typically, Bayesian
posterior probabilities are higher than jackknife or bootstrap
values (Cummings et al. 2003; Erixon et al. 2003; Zander
2004), but in the present case the Bayesian support values
are between 0.9 and 1.0 for clades that have around 50% or
less jackknife/bootstrap support. This appears not to be
uncommon, for example in the study by Martins (2006).
The lack of basal resolution and/or support for the resolu-
tion also seems to be common in comparable ITS studies
(Barker et al. 2004; Plunkett et al. 2004; Denk et al. 2005;
Tate et al. 2005; Galicia-Herbada 2006; Martins 2006).
Several workers acknowledge the problems (paralogy,
pseudogenes, etc.) with using the multi-copy nuclear ITS
region for phylogenetic analyses (see review by A
´
lvarez
and Wendel 2003; Bailey et al. 2003, but also Baldwin
et al. 1995; Barker et al. 2004; Denk et al. 2005; Tate et al.
2005). Especially when polyploidization and hybridization
events are relatively recent, the problems may be accen-
tuated. This is the case in the section Melioides, where
several species are polyploid and hybridization appears to
confound the phylogenetic result. For example, an addi-
tional sequence of F. americana did not group with the
other but instead in the F. pennsylvanica clade. Both these
specimens (i.e. not counting the sequence from Jeandroz
et al. 1997) have been verified as F. americana and they
even originate from the same locality (Shaw Arboretum,
MO, USA). However, both F. americana and F. pennsyl-
vanica occur in that locality. It is thus possible that one of
them represents a polyploid hybrid with F. pennsylvanica
and therefore ends up in the F. pennsylvanica group. This
demonstrates the likelihood that there is intraspecific var-
iability of the ITS repeat in the genome, that have not yet
homogenized through concerted evolution, and chance
may decide which copy amplify in the PCR (A
´
lvarez and
Wendel 2003). To verify this, cloning the PCR products is
necessary. This was not done, however, because it was not
an aim of this study to go deeper into the relationships
within the section Melioides. Heterogeneity of paralogous
sequences and selective amplification might also be caus-
ing the pattern found within the section Pauciflorae, where
two or more accessions of some species do not group
together, although polyploidy is unknown here.
It is strongly advised to complement the ITS data with
other sources of data, such as low-copy nuclear genes and/
or single-copy chloroplast introns (Barker et al. 2004).
Thus, in an attempt to corroborate the results from nuclear
ITS data with chloroplast data, I analyzed a combined data
set with trnL-F and rps16 intron sequences from 15
‘backbone’ species that represented all sections and the
incertae sedis. Previous works (Gielly and Taberlet 1994;
Wallander and Albert 2000) indicated that there was too
little variation in these chloroplast regions to be useful for
phylogenetic studies in the genus Fraxinus. After adding
sequences from five more species, in addition to the ten
already published (Wallander and Albert 2000), the number
of informative characters increased from 12 to 14. There is
not enough phylogenetic information in the two slow
evolving chloroplast regions to shed any light on the basal
resolution of Fraxinus. Even so, the chloroplast data could
corroborate the shared ancestry of the three species in the
section Dipetalae as well as the relationship between the
section Ornus and the other European sections (Fig. 2a).
The slight difference in topology between the ITS trees in
Fig. 1 and 2b (where the section Fraxinus appears para-
phyletic in the latter) may be an artefact of the low taxon
sampling. The incongruent position of F. greggii in the
chloroplast tree compared to the ITS tree may also be due to
this, or due to the paralogy problem described earlier.
The ITS phylogeny generally shows very good con-
gruence with morphological data and together they provide
a reliable estimate of the relationships within the genus
Fraxinus. The close relationship between the three species
in the section Dipetalae was unexpected but is strongly
supported, not only by the ITS and chloroplast data
(including an 8 bp long deletion in the trnL-F data set) but
also because a closer investigation revealed that the three
species share two unique morphological characters (qua-
drangular stems and hermaphrodite flowers in leafless
Plant Syst Evol (2008) 273:25–49 35
123
lateral inflorescences). There is weak molecular support for
the inclusion of F. raibocarpa in the section Ornus, but
there is high morphological support for this group in that
all species share terminal inflorescences and a corolla with
four free petals (except for the secondary loss of petals in a
few species). On the basis of the GC value (Goloboff et al.
2003), the low molecular support in this case appears not to
be due to a conflicting position. The section Pauciflorae is
supported by an 8 bp long deletion in the ITS sequences
and several unifying morphological characteristics (the
most distinctive being the small leaves with winged
rachises, in combination with small fruits). Within the
section Fraxinus, which is strongly supported by posterior
probability and bootstrap values but only weakly supported
by jackknife values, the Eurasian species also share an
indel. The section Sciadanthus is a strongly supported
sister to the section Fraxinus in all analyses. The section
Melioides is also a strongly supported monophyletic group.
The relation between the three species incertae sedis,
which form a clade in some analyses and is strongly sup-
ported by posterior probability values but ambiguously by
bootstrap and jackknife values, is difficult to interpret since
F. cuspidata is spuriously placed here. In this case, I
speculate that none of the inference methods have been
able to accurately resolve the true relationships because of
their relatively divergent DNA sequences. It is possible that
this may be attributable to the phenomenon known as long-
branch attraction (Felsenstein 1978), which may affect
parsimony as well as likelihood based methods such as
Bayesian inference (e.g. Archibald et al. 2003; Zhang et al.
2006). Therefore, in Fig. 3, F. cuspidata has been placed in
a more parsimonious position on the basis of morpholog-
ical and other evidence. These relationships are all
discussed in detail in Taxonomy and morphology in
Fraxinus’’ .
Jeandroz et al. (1997) presented a phylogeny on the basis
of maximum parsimony and neighbour-joining analyses of
ITS data. They did not include any representatives from the
sections Pauciflorae or Sciadanthus, and only one repre-
sentative of the petaliferous taxa in section Ornus, and the
tree has a poorly supported basal resolution. The topology is
not congruent with that obtained by Bayesian inference
(Fig. 1). Other major discrepancies include the misplace-
ment of F. anomala in their study, which is probably
because of a misidentification, as evidenced by the position
of their sequence obtained from GenBank in Fig. 1.
Systematics and new classification
There is a remarkable congruence between the lineages
identified in the molecular phylogeny and traditional ideas
of relationships within the genus. A few discrepancies exist
and I therefore present a revised infrageneric classification
of the genus Fraxinus (Table 2) consisting of six sections:
Dipetalae, Fraxinus, Melioides, Ornus, Pauciflorae, and
Sciadanthus. In previous classifications (Table 1), the
genus
Fraxinus has been divided into two subgenera or
sections, Ornus and Fraxinus, on the basis of inflorescence
position. The section Ornus, with terminal inflorescences,
is clearly monophyletic and derived from taxa with lateral
inflorescences. These latter taxa, on the other hand, do not
from a monophyletic group. Thus, a more natural infra-
generic classification is achieved by having only six
sections. It may seem unnecessary to recognize six sections
within such a relatively small genus as Fraxinus. However,
the six sections are each clearly monophyletic and well
separated from each other morphologically, as well as
forming molecularly distinct lineages. A case could be
made for including the small section Sciadanthus with the
section Fraxinus, although I have kept them separate here
for traditional reasons.
Fraxinus cuspidata forms an ambiguously supported
clade together with F. chiisanensis and F. spaethiana in
some analyses. Because of the present uncertainties and
pending further data to elucidate their positions, I leave all
three species as incertae sedis. Possible affinities on the
basis of non-molecular data are discussed further below.
Previously, the section Ornus comprised two subsec-
tions (sensu Lingelsheim 1920): Ornus with petals and
Ornaster without. Jeandroz et al. (1997) concluded, on the
basis of their limited analysis, that the two subsections
were monophyletic. However, that conclusion was unjus-
tified when only F. ornus represented the subsection Ornus.
In the present study, where nearly all species are included,
subsection Ornus is shown to be paraphyletic with respect
to subsection Ornaster. I have therefore chosen not to
recognize the subsections here, as the apetalous taxa are
derived from petalous ones (discussed more under section
Ornus below).
In summary, the differences compared to previous
classifications are: (1) abandoning the subgeneric and
subsectional levels and only recognizing six sections, of
which the section Pauciflorae is a new combination, (2)
moving F. platypoda from the section Melioides to the
section Fraxinus, (3) moving both F. quadrangulata from
the section Fraxinus and F. anomala from the section
Melioides to the section Dipetalae, and (4) treating
F. chiisanensis, F. spaethiana and F. cuspidata as incertae
sedis. All taxonomic changes and morphological support
are discussed below.
Taxonomy and morphology in Fraxinus
The genus
Fraxinus comprises 43 species in six sections:
Dipetalae, Melioides, Pauciflorae, Sciadanthus, Fraxinus,
36 Plant Syst Evol (2008) 273:25–49
123
and Ornus. The first five sections, excluding Fraxinus
dipetala but including the incertae sedis F. spaethiana and
F. chiisanensis, comprise 26 wind-pollinated species, which
are characterized by apetalous flowers in inflorescences that
emerge from lateral buds before the terminal leaf buds open.
F. dipetala and the incertae sedis F. cuspidata also have
lateral inflorescences, but sympetalous, hermaphrodite and
insect-pollinated flowers. The 15 species in the Eurasian
section Ornus have terminal inflorescences, with four free
petals or without, and this section appears to be derived
from a common ancestor shared with the other Eurasian
sections Fraxinus and Sciadanthus. A taxonomic discussion
for each section is given below.
The section Dipetalae
The section Dipetalae now comprises three American
species, previously scattered in different sections. F. qua-
drangulata belonged to the section Fraxinus, F. anomala to
Melioides, and F. dipetala used to be the sole member of
Dipetalae. Having found strong molecular support (from
both chloroplast and nuclear sequences) for the relationship
between these three species, I investigated their morpho-
logical characteristics more closely. In Fraxinus, they are
unique in having quadrangular twigs (due to development
of corky ridges) and hermaphrodite flowers occurring in
leafless lateral inflorescences. Other morphological simi-
larities include oval to ovate shaped wings of the samaras,
contrasting to the more elongated wings characteristic of
the samaras of the other sections. Although the three spe-
cies are seemingly quite different in other morphological
characters, such as number of petals, number of leaflets,
and life form, they are clearly most closely related to each
other. In addition to the high support from ITS and chlo-
roplast data, they also share an 8 bp long deletion in the
trnL-F data set.
Fraxinus dipetala (Two-petal ash) is a shrub or small
tree restricted to southwestern USA. It is the only Fraxinus
species having two petals, which are united and tubular by
fusion with the filaments. Sometimes, the petals are lacking
and the two forms may occur in the same panicle (Vines
1984). The flowers are fragrant and occur in many-flow-
ered and showy inflorescences, which probably attract
insects. The anthers are relatively large and protrude from
the corolla, an indication that the flowers might be both
wind- and insect-pollinated. It shows some morphological
affinity to F. cuspidata in the united petals.
Fraxinus anomala (Single-leaf ash) is also a shrub or
small tree, predominantly occurring in southwestern USA.
It is the only Fraxinus species with simple leaves, or
occasionally 3–5 leaflets (F. anomala Torr. var. lowelli
(Sarg.) Little). It differs from the species in the section
Melioides, where it was formerly classified by Lingelsheim
(1920), Miller (1955), and Nikolaev (1981), in having
quadrangular twigs and bisexual or sometimes unisexual
flowers (Sargent 1949; Vines 1984). The flowers have a
persistent calyx, but lack corolla, and appear in lateral
panicles before or with the young leaves. They are appar-
ently wind-pollinated.
Fraxinus quadrangulata (Blue ash) is a large tree with
conspicuously quadrangular twigs, occurring in eastern and
central North America. It was previously classified in the
section Fraxinus (Bumelioides). The flowers are mostly
hermaphrodite (Sargent 1949; Miller 1955; Vines 1984),
apetalous, and presumably wind-pollinated.
The section Melioides
The section Melioides has strong molecular support. They
are all medium-sized to large trees, deciduous and dioe-
cious. The unisexual flowers are apetalous and wind-
pollinated. The female flowers consist of a calyx and one
pistil, and the male flowers of two stamens with elongated
anthers and a small calyx. There are no rudimental organs
of the opposite sex in the flowers (a unique synapomorphy
for this section). The calyx is persistent in the samaras,
which have a distinctly terete seed cavity (except Fraxinus
caroliniana). The wing may be decurrent along the seed
cavity or not. In addition, the presence of flavones in the
leaves (besides the plesiomorphic flavonols) is a synapo-
morphy for these species (see further below about flavones
in F. chiisanensis).
Many taxa have been described in the section Melioides
and different authors accept different numbers of species
(Sargent 1949; Little 1952; Miller 1955; Kartesz 1994;
USDA, NRCS 2007). On the basis of examinations of over
300 herbarium specimens, and studies of some species in
the field, I accept ten species (Table 2) with a wide dis-
tribution in North America. A few species occur in Mexico
and Central America as well. Identification of the species
in this section is difficult because of both morphological
and ecological variation, and delimitation of taxa is also
complicated by extensive hybridization and polyploidy
(Miller 1955). As stated earlier, the ITS is not suitable for
elucidating relationships within this section because of this.
Some species may also be poorly defined but it is beyond
the scope of the present study to go deeper into this matter.
Traditionally, the species have been divided into two
main complexes, distinguished primarily by the presence of
papillae on the lower epidermis of the leaflets (‘the white ash
complex’) or by the absence of papillae (‘the red ash com-
plex’) (Wright 1944c; Wilson and Wood 1959; Hardin and
Beckmann 1982). The white ash complex comprises
F. americana (and Fraxinus biltmoreana), Fraxinus
papillosa, and Fraxinus texensis, and the red ash com-
plex comprises Fraxinus berlandieriana, F. caroliniana,
Plant Syst Evol (2008) 273:25–49 37
123
Fraxinus latifolia, F. pennsylvanica, Fraxinus profunda, and
Fraxinus velutina (Miller 1955). To the extent that ITS data
can say anything reliable about this group, there is no clear-
cut division in the molecular data that supports these com-
plexes (Fig. 1). There is one monophyletic group, however,
but it does not entirely correspond to the red ash complex, as
circumscribed by Miller (1955), since F. texensis is found
within this clade and F. latifolia is found outside. Thus,
presence or absence of epidermal papillae appears not to be
phylogenetically informative in this section.
Fraxinus americana (White ash) is a large tree distributed
in eastern North America. This species complex is composed
of three ecotypes (Wright 1944a; Wilson and Wood 1959)
with different ploidy levels. It is diploid (2n = 46) in the
northern parts of its range and diploid, tetraploid (2n = 92),
or hexaploid (2n = 138) in the southern parts (Wright
1944a). Black and Beckmann (1983) found trees of all ploidy
levels within immediate vicinity of each other in North
Carolina. This might explain why the two accessions of
F. americana that came from the same locality (Shaw
Arboretum, MO, USA) did not group together. It is
hypothesized that F. biltmoreana, which is not recognized
here, is an allopolyploid hybrid between a tetraploid
F. americana and a diploid F. pennsylvanica (Miller 1955;
Santamour 1962, but see Hardin and Beckmann 1982).
Miller (1955) treated F. texensis (Texas ash) as a sub-
species of F. americana, noting that F. texensis is separated
ecologically, physiologically, and morphologically from
F. americana. The most distinctive characters of F. tex-
ensis, which only occurs in Texas and northern Oklahoma,
are that it has fewer leaflets and smaller samaras than those
of F. americana.
Fraxinus papillosa (Chihuahuan ash) is a small tree that
occurs in southwestern USA and Mexico. Like the eastern
and much larger F. americana it has a distinctly papillose
lower leaf epidermis, but in contrast the leaflets are sessile.
Fraxinus uhdei (Shamel ash or Tropical ash) is distrib-
uted in Mexico, Guatemala, and Honduras. It is also
cultivated in Hawaii and naturalized in Puerto Rico and
some other tropical areas. It was first described as a variety
of F. americana, but later raised to species level by Lin-
gelsheim (
1907). It is distinguished by its long-petiolulate
and long-acuminate leaflets without papillae on the lower
epidermis, and tropical distribution, and therefore treated
as a separate species.
Fraxinus pennsylvanica (Red or Green ash) is distrib-
uted in the central and eastern USA and Canada. It also
consists of at least three different ecotypes, but in contrast
to F. americana they are all diploid (Wright 1944b).
Miller (1955) treated both F. velutina (Velvet ash) and
F. latifolia (Oregon ash, synonym F. oregona) as subspe-
cies of F. pennsylvanica. There is a report of polyploidy in
F. velutina (2n = 92, Taylor 1945) and it may hybridize
with F. pennsylvanica (Wright in Little 1952). Both
F. velutina and F. latifolia are western species, but
F. velutina occurs only in southwestern USA and northern
Mexico and F. latifolia is restricted to N California,
Oregon, and Washington. Both species have pubescent
leaves, but F. latifolia is distinguished by its sessile leaflets
whereas those of F. velutina are petiolulate.
Fraxinus profunda (Pumpkin ash, synonym F. tomen-
tosa) is hexaploid (2n = 138) and thought to be an
autopolyploid of F. pennsylvanica or, like F. biltmoreana,
another hybrid of a tetraploid F. americana and a diploid F.
pennsylvanica (Miller 1955; Wilson and Wood 1959;
Wright 1962, 1965, but see Hardin and Beckmann 1982).
Its leaves, twigs, flowers, and samaras are all larger than
those of F. pennsylvanica and F. americana, but qualita-
tively similar to one or the other of these two species.
According to Miller (1955), it is doubtful whether F. pro-
funda should be recognized, as it is not clearly separated
from F. pennsylvanica other than by its ‘gigas’ characters
due to polyploidy. However, these characters make it
morphologically distinct and therefore
F. profunda is
generally regarded as a separate species.
Fraxinus berlandieriana (Mexican ash) occurs in the
southwestern USA and northern Mexico. Miller (1955)
treated it as a synonym of F. pennsylvanica. But apart from
its southwestern distribution, it also differs from F. penn-
sylvanica in being a much smaller tree (about 10 m),
having fewer pairs of leaflets (3–5 vs. 5–9), and a wing of
the samara that is decurrent to the base (Sargent 1949;
Vines 1984).
Fraxinus caroliniana (Carolina ash or Water ash) occurs
in the southeastern swamps of the USA and is an extremely
variable species (Little 1952; Hardin 1974). It differs from
the other species of this section in not having a terete seed
cavity. Fraxinus cubensis in Cuba is treated as part of the
F. caroliniana complex. F. caroliniana reportedly forms
hybrids with F. americana (Miller 1955).
The section Pauciflorae
Fraxinus section Pauciflorae (Lingelsh.) E. Wallander,
comb. et stat. nov. (basionym: Fraxinus section Fraxinas-
ter subsection Pauciflorae Lingelsh., Engler’s Bot. Jahrb.
40:218. 1907), is a monophyletic group consisting of five
New World species that all occur in arid regions of the
southwestern USA, Mexico, and Guatemala. They are
shrubs or small trees with small coriaceous leaves. In
common with the two species of section Sciadanthus, the
leaves have winged rachises, but in contrast they have few-
flowered panicles. The flowers are polygamous, apetalous,
and wind-pollinated. The samaras have a persistent calyx.
Two species occur in the southwestern USA and
Mexico: F. greggii (Gregg’s ash) in Texas, Arizona, New
38 Plant Syst Evol (2008) 273:25–49
123
Mexico, and Mexico, and F. gooddingii (Goodding’s ash)
restricted to Arizona in the USA and Sonora in Mexico.
The latter was described by Little (1952), but later referred
to as a subspecies of F. greggii by Murray (1982), and has
an ITS sequence identical to one of the specimens of
F. greggii. It is doubtful whether F. gooddingii should be
retained as a separate species. Nevertheless, I have tenta-
tively accepted F. gooddingii because I have not seen more
than one herbarium specimen of it. Fraxinus rufescens,
occurring only in Mexico, is also related to these two
species but distinguished by its ferruginous-tomentose
shoots and inflorescences.
Fraxinus dubia and Fraxinus purpusii both occur in
Mexico and Guatemala and should not be so difficult to
distinguish on the basis of morphology. The leaflets of
F. purpusii have coarsely serrate margins whereas the
leaflets of F. dubia have entire margins. Nevertheless, the
analysis of several specimens each of F. dubia, including
its supposed synonym F. schiedeana (Green 1991),
F. purpusii, and F. greggii, indicates that the species
boundaries between them are not so clear (Fig. 1). An
alternative explanation for the pattern found may be,
although polyploidy and hybridization is unknown in this
section, that heterogenous paralogs of the ITS may have
been sequenced. Except for F. greggii, the Mexican species
are not well described or illustrated in the literature
(Standley 1924; but see the recent treatment by Rzedowski
2004) and I have only seen the type of F. purpusii.This
section is in need of a more thorough revision.
The section Sciadanthus
The section Sciadanthus consists of only two Old World
species: Fraxinus xanthoxyloides (Afghan or Algerian ash)
distributed from Morocco and Algeria in north Africa
through the Middle East to the Himalaya and China, and
Fraxinus hubeiensis which is a threatened species endemic
to the Hubei province in China (Ming and Liao 1998). They
are both small trees or shrubs, and characterized by apet-
alous flowers with calyx, except that the male flowers of
F. xanthoxyloides lack calyx. The flowers are polygamous
and wind-pollinated. They resemble the New World Pau-
ciflorae, but have many more flowers in their congested,
cymose panicles, and larger samaras. The leaves are rela-
tively small (7–15 cm) and possess a winged rachis. They
form a well-supported sister group to the section Fraxinus
and share the same main geographical distribution area.
The section Fraxinus
The section Fraxinus (invalid name Bumelioides) com-
prises five species, and all except F. nigra
are distributed in
Eurasia. In common with the species of the section
Melioides they are all relatively large and wind-pollinated
trees. In previous classifications, this group was charac-
terized by polygamous flowers without calyx, but now
F. platypoda, which has a reduced calyx, is included as
well. Fraxinus nigra has a deciduous calyx, but the other
three species have asepalous flowers. The male flowers
consist of two stamens, the hermaphrodite flowers of one
pistil and two stamens, and the female flowers of one pistil
and sometimes rudimentary stamens. They are distin-
guished from the strictly dioecious species of the section
Melioides in also having a flattened seed cavity of the
samara (vs. terete in the section Melioides) and foliar ter-
minal bud scales (vs. entire in the section Melioides)
(Whelden 1934). Although not all species have been
investigated, the two sections also differ from each other in
RFLP pattern (Morand et al. 2001) and foliar flavonoid
content (Harborne and Green 1980; Min et al. 2001).
Fraxinus excelsior (Common ash or European ash) is
distributed in the northern and central Europe and eastwards
to Volga river basin in western Russia. It displays a poly-
gamous breeding system, but recent studies indicate that it
might be subdioecious or functionally dioecious (Wallander
2001; FRAXIGEN 2005). Fraxinus coriariifolia, distin-
guished only by its variably pubescent shoots and leaves, is
found in Romania, Turkey, Caucasus, and northern Iran. It
may deserve recognition as a subspecies of F. excelsior,as
treated by Murray (1968) and Yaltirik (1978).
Fraxinus angustifolia (Narrow-leaved ash) includes a
complex of taxa which have not been fully clarified due to
extreme variation in morphology. After considering the
opinions of several authors (Anderson and Turrill 1938;
Metcalfe 1938; Vassiljev 1952; Franco and Rocha Afonso
1972; Scheller 1977; Yaltirik 1978; de Jong 1990), study-
ing herbarium material and some populations in the field,
and having seen the minimal variation among the ITS
sequences of these taxa, I have come to the conclusion that
Fraxinus oxycarpa, Fraxinus syriaca, Fraxinus pallisiae,
Fraxinus potamophila, and Fraxinus sogdiana should be
synonymized under F. angustifolia (see Yaltirik 1978, for
additional synonyms). The first two taxa may be retained as
subspecies together with the autonym, as they have been
treated by Franco and Rocha Afonso (1972) and Yaltirik
(1978). The leaf and shoot pubescence and leaflet mor-
phology within this complex are too variable to deserve
specific recognition. A more detailed analysis of the
F. angustifolia complex (unpublished ITS data involving
more specimens) reveals a nested relationship among the
taxa and there appears to be no monophyletic groups that
could be regarded as separate species. Following this view,
F. angustifolia s. l. has a wide distribution in the Medi-
terranean area and southeast Europe, through Turkey and
the Caucasus region, Southwest Asia, and east to the
Turkestan region (F. sogdiana).
Plant Syst Evol (2008) 273:25–49 39
123
Fraxinus angustifolia is closely related to F. excelsior
and they have also been shown to hybridize (Jeandroz et al.
1996; Raquin et al. 2002; Fernandez-Manjarres et al. 2006;
Heuertz et al. 2006). Fraxinus angustifolia differs mor-
phologically from F. excelsior (Fukarek 1960) and is
andromonoecious (Grunwald and Karschon 1984; Gyenova
1993; FRAXIGEN 2005). In contrast to all other taxa of the
genus, which have compound paniculate inflorescences,
F. angustifolia (including all its synonyms) has simple
racemes. Additional characters for distinguishing between
F. angustifolia and F. excelsior are listed by FRAXIGEN
(2005).
Fraxinus mandshurica (Manchurian ash) is most closely
related to F. platypoda and both species occur in China,
although F. mandshurica has a wider distribution into
eastern Russia, Korea, and Japan. F. mandshurica is
dioecious with rudimentary stamens in the pistillate flowers
(Yamazaki 1993, personal observation).
Fraxinus platypoda was previously placed in the section
Melioides (Lingelsheim 1920; Nikolaev 1981), despite its
Asian distribution, because of the presence of a small
calyx, at least in the bisexual flowers (Wei and Green
1996). Nevertheless, it exhibits all other characters shared
by the taxa in section Fraxinus, including the distinctly
flattened seed cavity of the samara, and molecular data give
strong support for its placement here. In fact, the ITS
sequences of F. platypoda and F. mandshurica are nearly
identical. Many authors (e.g., Nakaike 1972; Wei and
Green 1996) place F. spaethiana (endemic to Japan) in
synonymy under F. platypoda. Both of them have a dis-
tinctly swollen base of the leaf petiole clasping the shoot,
but F. platypoda differs from F. spaethiana in not having a
papillose epidermis on the abaxial side of the leaves. The
ITS sequences of F. platypoda and F. spaethiana are also
quite dissimilar and I recognize them as separate species. It
is possible that the previous placement of F. platypoda was
based on a mix-up with
F. spaethiana (which appears to be
related to Melioides).
Fraxinus nigra (Black ash) is an eastern North
American species, but there has never been any dis-
agreement over whether it belongs to this section or to the
American Melioides. It is morphologically similar to
F. mandshurica, and the latter has been referred to as a
subspecies of F. nigra by Sun (1985) and Green in Wei
and Green (1996). The flowers are polygamous and the
calyx is small and deciduous, or absent. The seed cavity
of the samara is flattened, similar to the other species in
this section.
Section Ornus
The molecular support for the section Ornus is high,
although with low support for the inclusion of
F. raibocarpa, and there is no doubt that it is a mono-
phyletic group. Within Fraxinus, the species are unique in
that the inflorescences are borne on current year shoots
together with the leaves, which emerge from terminal buds,
in contrast to lateral inflorescences on previous year’s
shoots in all the other sections.
The section Ornus as circumscribed here comprises 15
species, all distributed in Eurasia and with a concentration
in eastern Asia. Eleven of these species have four,
essentially free, petals that are only united at the very
base (compared to F. cuspidata). The flowers are fragrant
and produce a lot of pollen. Apparently, they are insect-
pollinated or both wind- and insect-pollinated (personal
observation). These 11 species constitute subsection
Ornus sensu Lingelsheim (1920). I could not obtain ITS
sequences from two petaliferous species of this section,
F. griffithii and F. malacophylla, but they undoubtedly
belong here. F. griffithii occurs in Japan, Taiwan, and the
Philippines and is an evergreen or semi-evergreen small
tree with large showy panicles. F. malacophylla is a
relatively large tree, distributed in southern China and
northern Thailand, and characterized by brown tomentose
leaflets. In these two species, the flowers are always
hermaphrodite (Yamazaki 1993; Wei and Green 1996).
This is also the case for F. raibocarpa, a shrub or small
tree in Central Asia with characteristically falcate sam-
aras. The remaining eight petaliferous species are all
androdioecious (Yamazaki 1993; Wei and Green 1996,
personal observation) and occur mostly in the Himalayan
mountains, China, and Japan. Of them, Fraxinus siebol-
diana (China and Japan), Fraxinus lanuginosa (Japan),
and Fraxinus apertisquamifera (Japan) form a strongly
supported group. The rest, including the European species
F. ornus (Manna ash), have no supported resolution
among them.
Four species are apetalous (previously subsection Orn-
aster): F. micrantha in the Himalaya,
F. longicuspis in
Japan, F. baroniana in China, and the more widespread
southeast Asian F. chinensis s. l. These species form a
well-supported monophyletic group, distinguished mor-
phologically from the rest of section Ornus in having no
corolla. They are mostly relatively large trees and although
the flowers and leaves emerge together, they flower before
the leaves are expanded and are wind-pollinated. Unfor-
tunately, I could not obtain any ITS sequence from
F. baroniana, but on the basis of morphology it undoubt-
edly belongs to this group. It is also apetalous, but differs
morphologically from the other species in being a smaller
tree and in having relatively narrow leaflets. Fraxinus mi-
crantha and F. longicuspis are androdioecious (Nakaike
1972; Yamazaki 1993, personal observation) and F. ba-
roniana and F. chinensis are dioecious (Wei and Green
1996). Some individuals of F. chinensis have rudimentary
40 Plant Syst Evol (2008) 273:25–49
123
stamens in their female flowers (Wei and Green 1996,
personal observation).
The ITS sequences of both Fraxinus japonica (Japan)
and Fraxinus rhynchophylla (Korea and northern China)
are practically identical to F. chinensis, and they are here
treated as synonyms to the latter species. Wei and Green
(1996) also treat F. japonica as a synonym of F. chinensis
Roxb. ssp. rhynchophylla (Hance) E. Murray. There are no
clear morphological boundaries between the taxa of the
F. chinensis complex (Nakaike 1972, personal observation,
but see Kang et al. 2002). F. chinensis has 2n = 46, 92, or
138 (Wright 1962; Nikolaev 1981) and the ploidy levels in
this species complex may explain some of the morpho-
logical variation.
The section Ornus, as circumscribed by Lingelsheim
(1920) and others, encompassed two subsections. Although
subsection Ornaster is clearly monophyletic based on
strong molecular support and floral synapomorphies, in the
present phylogenetic classification there is no support for
recognising the subsection Ornus, as the former is derived
from the latter. I have therefore abandoned the subsections
within section Ornus, despite the fact that the two groups
are morphologically well distinguished (with or without
petals).
Species incertae sedis
The odd American species F. cuspidata (Fragrant ash,
Fresno) is a small entomophilous tree occurring in Mexico
and the southwestern USA. On the basis of its possession
of corolla it was included in the otherwise Eurasian section
Ornus (Lingelsheim 1920). Although having four petals, its
flowers are not similar to those of the species in this sec-
tion. The petals are united and form a tube about one-third
of the length of the corolla, not free as in the section Ornus.
The two stamens are united with the corolla tube and
shorter than the petals. The fragrant flowers are borne
terminally in lateral, leafy panicles developed from the
axils of the leaves of the previous year, not in terminal
panicles on current year shoots as in the section Ornus.
Nikolaev (1981) included F. cuspidata with F. dipetala in
the section Dipetalae. Although differing in petal number
(F. dipetala has two and F. cuspidata four petals), they are
both hermaphrodites with the plesiomorphic sympetalous
corolla fused with the filaments. They both have inflores-
cences on lateral shoots (which are not leafy in F. dipetala)
and are also the only two petaliferous species of Fraxinus
in America. The results from molecular data do not support
a close relationship with the section Ornus or the section
Dipetalae, but instead F. cuspidata is found in a poorly
supported clade with the other two unclassified species
F. chiisanensis and F. spaethiana. Although this is a highly
dubious position, F. cuspidata appears to not belong within
any of the previously described sections. On the basis of its
unique morphology within the genus, such as the four
united petals and inflorescence position, a case could be
made for placing it in a section of its own. However,
pending further data to elucidate its position I have tenta-
tively left it as incertae sedis.
The ITS sequences of F. chiisanensis and F. spaethi-
ana are quite divergent compared to the other taxa and
are only similar to the odd F. cuspidata. These three
species form a clade that is moderately supported by
Bayesian posterior probabilities but with less than 50%
bootstrap and jackknife support. In the taxon-reduced
chloroplast and ITS trees (Fig. 2a, b), one or all three of
them appear as closely related to the section Melioides.
The GC output of the complete ITS jackknife tree gives
weak support for this position due to grouping conflicts
and in Fig. 1 the clade occurs in a basal trichotomy.
Possible affinities on the basis of non-molecular data are
discussed below.
Fraxinus chiisanensis is a wind-pollinated tree, endemic
to Korea. The samaras have a persistent calyx and the leaf
rachis is not winged. Surprisingly, these features led the
auctor Nakai (1929) to conclude that it belonged to section
Dipetalae, which previously comprised only the American
F. dipetala. However, the combination of a persistent calyx
and non-winged rachis is also a feature of the section
Melioides, where it would fit much better. The relation-
ships of F. chiisanensis were discussed by Nakaike (1972),
who believed that the section to which this species belongs
could not be determined without flowers (which he had not
seen). Hypotheses on the putative hybrid origin of F. chii-
sanensis from F. mandshurica (section Fraxinus) and F.
chinensis (section Ornus) have been refuted by Noh et al.
(1999) on the basis of RFLP patterns and by Min et al.
(2001) on the basis of foliar flavonoids. Although the
apetalous flowers of F. chiisanensis appear to be interme-
diate in morphology between F. mandshurica and
F. chinensis (Min et al. 2001), these species have no other
close similarities. I also find it unlikely that two distantly
related species from such morphologically different sec-
tions could hybridize. On the basis of examinations of
flowers of F. chiisanensis, and photos by Min et al. (2001),
I noted that the lateral inflorescences have either apetalous
male flowers or hermaphrodite flowers with elongated
anthers, similar to those of section Melioides. Although
polygamous or androdioecious in appearance, the anthers
in the hermaphrodite flowers seem to be smaller than in the
male ones, suggesting functional dioecy. The samaras of
F. chiisanensis are similar to those of Melioides, with a
terete seed cavity. The geographical distribution and
presence of bisexual flowers, on the other hand, suggest a
relationship with F. platypoda and F. mandshurica in the
section Fraxinus. Recently, chemical support for a shared
Plant Syst Evol (2008) 273:25–49 41
123
ancestry of F. chiisanensis and the section Melioides has
come from studies of foliar flavonoids. A study by Min
et al. (2001) found the presence of advanced flavones in F.
chiisanensis, in common with F. americana and F. penn-
sylvanica of the section Melioides (Harborne and Green
1980; Black-Schaefer and Beckmann 1989). An expanded
study by Chang et al. (2002) confirmed that F. chiisanensis
shares the apomorphic flavones with several other species
of the section Melioides. Other taxa of Fraxinus that have
been investigated have only the plesiomorphic flavonols.
Fraxinus spaethiana is a large wind-pollinated tree,
endemic to Japan. Lingelsheim (1907, 1920) classified it
in the section Ornus, on the basis of a mistake that it had
four petals. It has lateral inflorescences with polygamous
and apetalous flowers. Calyx is present only in pistillate
flowers (Yamazaki 1993) and the samaras have a flattened
seed cavity. As mentioned earlier, it is morphologically
similar to the Chinese F. platypoda, and has been sug-
gested to be a synonym of this species (Nakaike 1972;
Wei and Green 1996). The molecular data, however,
places these taxa in widely separated positions. The fla-
vonoid study by Min et al. (2001) only found flavonols in
F. spaethiana. Thus, only the ITS sequence is similar to
F. chiisanensis.
The molecular data are not conclusive, but they indi-
cate a relationship between the three species and the
section Melioides. Some morphological and chemical
data also suggest a close relationship for at least F. chii-
sanensis with the species in the section Melioides. Min
et al. (2001) proposed that F. chiisanensis should be
included in the section Melioides due to its chemical
affinity with this section. Chang et al. (2002) found the
same chemical pattern, but on the other hand noted a
strong discontinuity between F. chiisanensis and the
species in the section Melioides in the shape of the ter-
minal bud scales, some leaf characters, and floral
sexuality. They also compared ITS sequences and con-
cluded that F. chiisanensis seems to be a highly
primitive species within the section Melioides and that
‘it was probably differentiated from the ancestor of this
group a long time ago’’. The results for F. spaethiana are
unclear due to the non-finding of flavones and its flat-
tened seed cavity. Because of these present uncertainties,
I leave both species as incertae sedis.
Evolutionary comments
Floral evolution
Jeandroz et al. (1997) traced the evolution of corolla and
calyx, and concluded that the floral evolution had been
homoplasious. However, owing to an incomplete
phylogeny as well as several errors in character state
coding (F. quadrangulata and F. nigra have a small and
deciduous calyx, and all included species of previous
subsection Ornaster have a calyx but no corolla), the
number of inferred character state changes are incorrect.
On the basis of my nearly complete phylogeny, as
depicted in Fig. 3, the presence of calyx and corolla is
ancestral in the genus. The corolla has been lost three
times, but it also appears to have been regained from an
apetalous state in the ancestor of section Ornus (before
lost again). This floral trait is correlated with pollination
mode and loss of corolla is one of a suite of characters
coupled to the anemophilous syndrome (Faegri and van
der Pijl 1979). The calyx has been completely lost only
once (within the section Fraxinus). In a few taxa of some
other sections, the calyx is only reduced. Simultaneously
with regaining the corolla, terminal inflorescences appear
to have evolved from lateral ones in the ancestor of
section Ornus.
Breeding system evolution
The species of Fraxinus display several, to some extent
intergrading, breeding systems. Hermaphroditism is
ancestral in the family Oleaceae (Wallander and Albert
2000) and appears to be the ancestral state in Fraxinus as
well. The three species of section Dipetalae, plus F. ma-
lacophylla, F. griffithii and F. raibocarpa of the section
Ornus, and F. cuspidata, are the only hermaphroditic
species. No less than ten species in section Ornus (eight
with and two without corolla) are androdioecious, at least
phenotypically (Yamazaki 1993; Wei and Green 1996,
personal observation). Androdioecy (separate male and
hermaphrodite individuals) is an extremely rare breeding
system (e.g. Charlesworth 1984; Pannell 2002). The genus
Fraxinus is therefore unusual among angiosperms in hav-
ing several androdioecious species. Several other genera of
the Oleaceae also have androdioecious species, e.g. Chio-
nanthus (Ueda 1996), Osmanthus (e.g. Wei and Green
1996), and Phillyrea (e.g. Vassiliadis et al. 2002), and this
high incidence is probably tied to their common phyloge-
netic history (Wallander 2001). Dioecy occurs in all ten
species of the section
Melioides, the two apetalous
F. chinensis and F. baroniana of the section Ornus, and in
F. mandshurica of the section Fraxinus. The remaining
four species of the section Fraxinus are morphologically
polygamous, as well as the five species of the section
Pauciflorae and the two species of the section Sciadanthus.
The conclusion drawn from the phylogenetic work is that
several of the wind-pollinated species have evolved dioecy
independently. Dioecy evolved once from hermaphrodit-
ism via androdioecy in the section Ornus, once via
polygamy in the section Fraxinus, and once in the ancestor
42 Plant Syst Evol (2008) 273:25–49
123
of the section Melioides. This gradual loss of first female
function resulting in male flowers and later loss of male
function resulting in functionally female flowers has hap-
pened repeatedly following the transition from insect to
wind pollination. This interesting trend in the evolution of
unisexual flowers is discussed in more detail elsewhere
(Wallander 2001).
Biogeography
It is beyond the purpose of the present paper to do a
thorough biogeographical analysis. However, on the basis
of the phylogenetic result and morphological studies, the
following biogeographical interpretation is made. Sup-
ported by fossil evidence (Call and Dilcher 1992), Fraxinus
is hypothesized to have originated in the North America
during the Eocene with two subsequent dispersal events
across one or both of the two land bridges into Eurasia (the
ancestor of F. chiisanensis and F. spaethiana, and the
ancestor of the sections Ornus, Fraxinus, and Sciadanthus)
and one dispersal event back to North America (F. nigra).
This hypothesis differs from scenario A proposed by
Jeandroz et al. (1997) because the dispersal of the ancestor
of F. chiisanensis and F. spaethiana was not accounted for
in their study.
Conclusions
The reliability of the phylogenetic hypothesis was asses-
sed in several ways. The ITS data were analyzed using
three different methods of phylogenetic inference, maxi-
mum parsimony, maximum likelihood, and Bayesian
analyses, and the clade support in the resulting trees was
evaluated through jackknife, bootstrap, and posterior
probability values, respectively. Although with variable
support for the basal resolution, no highly supported
clades were contradicted. An independent data set with a
smaller number of representative taxa, consisting of
combined chloroplast sequences from the rps16 and trnL-
F regions, was found to be largely congruent with the ITS
phylogeny. In addition, several morphological characters
support the major clades, which correspond well to the
sections in previous classifications. Thus, I conclude that
the result obtained here most likely represents a reliable
estimate of the phylogenetic relationships within the
genus Fraxinus.
The subgenus Fraxinus was found to be paraphyletic,
because subgenus Ornus is derived from it, and in the
revised classification I have abandoned the subgeneric
rank. I recognize six sections only and a total of 43 species.
Three species have been transferred to other sections to
accord with the phylogenetic results and three species are
treated as incertae sedis because their phylogenetic posi-
tions are still uncertain. Further data, including other DNA
regions and additional morphological and biochemical
data, may be able to elucidate their phylogenetic
relationships.
Morphological and chemical data were interpreted in the
light of the molecular phylogeny. Most of the traditional
taxonomical characters appeared to be informative. How-
ever, traits such as absence of corolla and/or calyx, which
are correlated with the evolution of wind pollination,
constitute parallel losses in some groups and do not reflect
a common phylogenetic history. Unisexual and bisexual
flowers occur in different combinations in most species and
the evolution of breeding systems shows a trend from
hermaphroditism via androdioecy or polygamy to dioecy.
The interpretation of the floral evolution, in relation to
transitions between pollination systems, is discussed in
more detail elsewhere (Wallander 2001).
Acknowledgments This research was supported by the Lewis B.
and Dorothy Cullman Foundation, The Royal Swedish Academy of
Sciences, Kungliga och Hvitfeldtska Stiftelsen, Wilhelm och Martina
Lundgrens Vetenskapsfond, and Collianders stiftelse. I am grateful to
the following individuals and institutions for help in the field or for
providing me floral and/or leaf material: Tsutomu Enoki (University
of the Ryukyus, Okinawa, Japan), Shen Hailong (Northeast Forestry
University, Harbin, China), Woong-Ki Min (Seoul National Univer-
sity, Korea), Satoshi Nanami and Takashi Osono (Kyoto University,
Japan), Junko Okazaki (Osaka Kyoiku University, Japan), Ladislav
Paule (Technical University, Zvolen, Slovak Republic), and Xu Yo-
uming (Huazhong Agricultural University, Hubei, China). Herbaria
(BM, C, E, GB, K, MO, NY, S, and UPS) that lent material and gave
permission to extract DNA are gratefully acknowledged, and espe-
cially Mark Chase at the Jodrell Laboratory at Kew Gardens, who
provided the DNA extracts from living and herbarium material at K. I
am grateful to Alexandre Antonelli for running the Bayesian analyses
for me and I thank Steve Farris for doing the jackknife analyses, Peter
S. Green for valuable discussions, John Landon for taxonomical
advice, and Olga Khitun for translating Russian texts. Most of this
work was done when I was a Ph.D. student at the Department of Plant
and Environmental Sciences, Go
¨
teborg University, and I thank the
staff for all help I received. For critically commenting on previous
drafts, which significantly improved this paper, I thank A
˚
slo
¨
g Dahl,
Roger Eriksson, Bengt Oxelman, Johan Rova, and two anonymous
reviewers.
Appendix
See Table 4
Plant Syst Evol (2008) 273:25–49 43
123
Table 4 Voucher information and GenBank accession numbers for 90 specimens of Fraxinus and five outgroup taxa used for ITS sequencing
(the F. americana sequence marked asterisk was not included in the Bayesian analysis)
Taxon Voucher Origin GenBank accession
number (ITS)
GenBank accession
number (trnL-F + rps16)
Ingroup
F. americana U82906 + U82907
F. americana* Wallander 99 (GB) USA (Missouri,
Shaw Arb.)
EU314811
F. americana Wallander 101 (GB) USA (Missouri,
Shaw Arb.)
EU314812 AF231825 + AF225233
F. americana (F. biltmoreana) U82910 + U82911
F. angustifolia ssp. angustifolia Wallander 135 (GB) Tenerife (cult.
Puerto de la Cruz)
EU314813
F. angustifolia ssp. oxycarpa 1 Paule 44a18 (ZV) Turkey EU314814
F. angustifolia ssp. oxycarpa 2 Wallander 2 (GB) Italy (cult. Go
¨
teborg
Bot. Garden)
EU314815
F. angustifolia ssp. syriaca Samuelsson 1 (GB) Israel (Acco) EU314816
F. angustifolia (F. syriaca) U82872 + U82873
F. angustifolia (F. oxyphylla) U82868 + U82869
F. angustifolia (F. pallisiae) U82870 + U82871
F. angustifolia (F. pallisiae) 1 Paule 43a29 (ZV) Turkey EU314817
F. angustifolia (F. pallisiae) 2 Rechinger 10066 (S) Greece EU314818
F. angustifolia (F. potamophila) Wallander 88 (GB) Uzbekistan (cult.
Missouri Bot.
Garden)
EU314819
F. angustifolia (F. sogdiana) Elias 10008 (C) Tajikistan EU314820
F. anomala (wrong ID) U82914 + U82915
F. anomala Pinzl 10931 (NY) USA (Nevada) EU314821
F. anomala Rollins 1899 (GB) USA (Colorado) EU314822 AF231826
+ AF225234
F. apertisquamifera Kinoshita sn 1999-
07-14 (GB)
Japan (Fukui) EU314823
F. apertisquamifera Wallander 274 (GB) Japan (Saitama) EU314824
F. berlandieriana Jones 3595 (NY) USA (Texas) EU314825
F. berlandieriana Pringle 13584 (S) Mexico (Hidalgo) EU314826
F. bungeana King 168 (S) China (Hupei) EU314827
F. bungeana Tianwei & Zhaofen
228 (MO)
China EU314828
F. bungeana Wallander 406 (GB) China (Beijing) EU314829
F. caroliniana 1 Hill 11048 (NY) USA (Florida) EU314830
F. caroliniana 2 Massey & Boufford
4500 (MO)
USA (North
Carolina)
EU314831
F. caroliniana (F. cubensis) Rova 2261 (GB) Cuba EU314832
F. chiisanensis Min 264 + Min 304
(SNUA) (identical
ITS sequences)
South Korea EU314833 EU284157 + EU284157
F. chinensis U82884 + U82885
F. chinensis Wallander 116 (GB) China (cult.
Go
¨
teborg Bot.
Garden)
EU314834 AF231827 + AF225235
F. chinensis (F. japonica) Wallander 235 (GB) Japan (cult. Kyoto) EU314835
F. chinensis (F. japonica) Wallander 245 (GB) Japan (cult. Kyoto) EU314836
F. chinensis (F. rhynchophylla) Wallander 400 (GB) China
(Heilongjiang)
EU314837
F. cuspidata U82916 + U82917
44 Plant Syst Evol (2008) 273:25–49
123
Table 4 continued
Taxon Voucher Origin GenBank accession
number (ITS)
GenBank accession
number (trnL-F + rps16)
F. cuspidata Barneby 18368 (NY) USA (Arizona) EU314838
F. cuspidata Reichenbacher 1716
(MO)
USA (Arizona) EU314839 AF231828 + AF225236
F. dipetala Walker 1287 (NY) USA (California) EU314840
F. dipetala Wallander 180 (GB) USA (California) EU314841 AF231829 + AF225237
F. dipetala (F. jonesii) Thorne 58757 (NY) Mexico (Baja
California)
EU314842
F. dubia Breedlove 32784
(MO)
Mexico (Chiapas) EU314843
F. dubia Garcı
´
a 1456 (MO) Mexico (Chiapas) EU314844
F. dubia? Martı
´
nez & Soto
3718 (MO)
Mexico (Guerrero) EU314845
F. dubia?(F. schiedeana) Villanueva 274
(NY)
Mexico (Veracruz) EU314846
F. excelsior Wallander 159 (GB) Sweden (Go
¨
teborg) EU314847 AF231830
F. excelsior L. var. diversifolia Ait. Wallander 1 (GB) Sweden (cult.
Go
¨
teborg Bot.
Garden)
EU314848 + AF225238
F. excelsior ssp. coriariifolia Wallander 353 (GB) Romania (Tulcea) EU314849
F. floribunda Wallander 240 (GB) Japan (cult. Kyoto) EU314850
F. floribunda (F. retusa) Wallander 249 (GB) Japan (Okinawa) EU314851
F. gooddingii McGill & Lehto
20365 (NY)
USA (Arizona) EU314852
F. greggii 1 Annable 2379 (NY) USA (Arizona) EU314853
F. greggii 2 Johnston 7214 (S) Mexico (Coahuila) EU314854
F. greggii? 3 Diaz 406 (MO) Mexico
(Tamaulipas)
EU314855 AF231831 + AF225239
F. hubeiensis Xu Youming s.n.
2001–07 (WH)
China (Hubei) EU314856
F. lanuginosa Seino 2 (GB) Japan (Hokkaido) EU314857
F. lanuginosa Wallander 266 (GB) Japan (Saitama) EU314858
F. latifolia U82912 + U82913
F. latifolia Wallander 182 (GB) USA (California) EU314859
F. latifolia Wallander 322 (GB) USA (Washington) EU314860
F. longicuspis U82888 + U82889
F. longicuspis Im 10518 (NY) Japan (Shiga) EU314861
F. longicuspis Wallander 256 (GB) Japan (Saitama) EU314862
F. mandshurica U82874 + U82875
F. mandshurica Seino 1 (GB) Japan (Hokkaido) EU314863
F. mandshurica Wallander 396 (GB) China
(Heilongjiang)
EU314864
F. micrantha Bist 96 (S) India EU314865
F. micrantha Polunin et al. 4299
(UPS)
Nepal EU314866
F. nigra U82878 + U82879
F. nigra Rickson 239 (GB) USA (cult. Miami
University
Campus)
EU314867
F. nigra Wallander 105 (GB) USA (cult. Missouri
Bot. Garden)
EU314868 EU284158 + EU284163
F. ornus Wallander 38 (GB) Italy (Sicily) EU314869 AF231832 + AF225240
Plant Syst Evol (2008) 273:25–49 45
123
Table 4 continued
Taxon Voucher Origin GenBank accession
number (ITS)
GenBank accession
number (trnL-F + rps16)
F. ornus Wallander 216 (GB) Cult. NY Bot.
Garden
EU314870
F. papillosa Felger 94–288 (MO) Mexico (Chihuahua) EU314871
F. papillosa Tucker 2597 (S) Mexico (Chihuahua) EU314872
F. paxiana Wallander 187 (GB) China (cult.
Go
¨
teborg Bot.
Garden)
EU314873
F. paxiana (F. sikkimensis) Wallander 188 (GB) China (cult.
Go
¨
teborg Bot.
Garden)
EU314874
F. pennsylvanica U82902 + U82903
F. pennsylvanica 1 Wallander 83 (GB) USA (cult. Missouri
Bot. Garden)
EU314875
F. pennsylvanica 2 Wallander 103 (GB) USA (Missouri,
Shaw Arb.)
EU314876 EU284159 + EU284164
F. platypoda U82876 + U82877
F. platypoda Wallander 114 (GB) China (cult.
Go
¨
teborg Bot.
Garden)
EU314877
F. profunda (F. tomentosa) U82896 + U82897
F. purpusii 1 Breedlove & Thorne
30445 (NY)
Mexico (Chiapas) EU314878
F. purpusii 2 Breedlove 42154
(MO)
Mexico (Chiapas) EU314879
F. purpusii 3 Medrano et al.
11420 (MO)
Mexico (Oaxaca) EU314880
F. quadrangulata U82880 + U82881
F. quadrangulata Wallander 94 (GB) USA (Missouri,
Shaw Arb.)
EU314881
F. quadrangulata Wallander 98 (GB) USA (Missouri,
Shaw Arb.)
EU314882 AF231833 + AF225241
F. raibocarpa Regel s.n. July 1982
(S)
Tajikistan (Hissar) EU314883 EU284160 + EU284165
F. raibocarpa Sabirov s.n. 1955-
08-17 (MO)
Uzbekistan EU314884
F. rufescens Zamudio 3673 (MO) Mexico (Queretaro) EU314885
F. sieboldiana Takahashi et al.
1708 (MO)
Japan (Honshu) EU314886
F. sieboldiana Wallander 265 (GB) Japan (Saitama) EU314887
F. spaethiana Wallander 142 (GB) Japan (cult.
Go
¨
teborg Bot.
Garden)
EU314888
F. spaethiana Wallander 259 (GB) Japan (Saitama) EU314889 EU284161 + EU284166
F. spaethiana Wallander 320 (GB) Japan (cult.
Go
¨
teborg Bot.
Garden)
EU314890
F. texensis 1 Chase 3887 (K) USA (cult. Kew
Garden)
EU314891
F. texensis 2 Walker 1692 (NY) USA (Texas) EU314892
F. trifoliolata Forrest 15313 (E) China (Yunnan) EU314893
F. uhdei Boa s.n. 2001–08
(GB)
Colombia (cult.) EU314894
46 Plant Syst Evol (2008) 273:25–49
123
References
A
´
lvarez I, Wendel JF (2003) Ribosomal ITS sequences and plant
phylogenetic inference. Molec Phylogenet Evol 29:417–434
Anderson E, Turrill WB (1938) Statistical studies on two populations
of Fraxinus. New Phytol 37:160–172
Archibald JK, Mort ME, Crawford DJ (2003) Bayesian inference of
phylogeny: a non-technical primer. Taxon 52:187–191