American Journal of Botany 86(7): 1027–1037. 1999.

PHYLOGENETIC RELATIONSHIPS OF THE

HAMAMELIDACEAE INFERRED FROM SEQUENCES OF
INTERNAL TRANSCRIBED SPACERS (ITS) OF NUCLEAR
RIBOSOMAL DNA1
JIANHUA LI,2,5 A. LINN BOGLE,3 AND ANITA S. KLEIN4
2Harvard University Herbaria, 22 Divinity Avenue, Cambridge, Massachusetts 02138;

3
Department of Plant Biology, University of New Hampshire, Durham, New Hampshire 03824; and
4Department of Biochemistry and Molecular Biology, University of New Hampshire, Durham, New Hampshire 03824

Intergeneric relationships in the Hamamelidaceae have long been controversial. In this study, sequences of the internal
transcribed spacers of nuclear ribosomal DNA were used to reconstruct the phylogeny for the Hamamelidaceae. Three major
clades were recognized in the ITS-based phylogenetic tree: (1) Mytilaria-Exbucklandia-Rhodoleia, (2) Disanthus, and (3)
the Hamamelidoideae. Within the Hamamelidoideae there were three well-supported lineages: (1) Corylopsis-Loropetalum-
Tetrathyrium-Maingaya-Matudaea, (2) Eustigmateae sensu Endress, plus Molinadendron-Dicoryphinae, and (3) Hamamelis-
Fothergilleae sensu Endress, excluding Matudaea and Molinadendron. The Exbucklandioideae sensu Endress were not
monophyletic, nor were the tribes in the Hamamelidoideae in their current circumscriptions except for the Corylopsideae.
Strap-shaped petals, apetaly, and wind pollination have evolved three times independently in the Hamamelidaceae s.s.
(Hamamelidaceae minus Altingioideae), suggesting that homoplasy should be considered in future classifications of the
family.

Key words: Hamamelidaceae; ITS DNA sequences; homoplasy; phylogeny.

There are 31 genera and .140 species in the Hama-
melidaceae (Endress, 1993; Zhang and Lu, 1995). This
family is known for its broad geographic distribution,
high number of endemics, monotypic or oligotypic gen-
era, and diverse morphology (Bogle, 1968; Endress,
1989a, b, c, 1993).
The geographic range includes eastern and southern
Africa, Madagascar, northeastern Australia, western, cen-
tral, and southeastern Asia, eastern North America, Cen-
tral America, and northern South America (Gentry, 1993;
Ulloa and Jørgensen, 1993; Zhang and Lu, 1995; Lozano-
Contreras, 1996; Rakotobe, 1996). Whereas most of the
genera occur in the Northern Hemisphere, five genera are
distributed only in the Southern Hemisphere, including
Trichocladus Pers., Ostrearia Baill., Neostrearia Smith,
Noahdendron Endress, Hyland et Tracey, and Dicoryphe
Du Petit-Thouars. More than half of the genera in the
Hamamelidaceae are endemic to very restricted areas. For
instance, Ostrearia, Neostrearia, and Noahdendron are
distributed only in the rainforests of northeastern Queens-
land, Australia; Maingaya Oliv. is endemic to northeast-
ern Malaysia; Fortunearia Rehd. et Wils., Sinowilsonia
1 Manuscript received 3 April 1998; revision accepted 26 March
1999.
The authors thank P. K. Endress, T. D. Omar, Z.-C. Luo, R. Saunders,
L. G. Saw, and Y.-L Qiu for providing leaf material or DNA for this
study; The Arnold Arboretum (MA), the Missouri Botanical Garden
(MO), the Morris Arboretum (PA), and Longwood Gardens (PA) for
granting them permission to collect material. The research was partially
supported by the University of New Hampshire Graduate Student Re-
search Enhancement Fund to JLI and by the Howard and Dorothy Pow-
ers Fund to ALB.
This is a portion of the doctoral thesis of the first author presented
to the Graduate School of the University of New Hampshire, Durham.
5 Author for correspondence, current address: Arnold Arboretum of
Harvard University, 125 Arborway, Jamaica Plain, MA 02130.
1027
Hemsl., Semiliguidambar Chang, and Shaniodendron
(Chang) Deng, Wei et Wang can be found only in small
areas of China. However, there are two genera in this
family whose species are widely and disjunctly distrib-
uted in several continents. One of them is Liquidambar
L., which can be found in western Asia, southeastern
Asia, southeastern North America, and Central America
(Bogle, 1968, 1986); the other disjunct genus is Hama-
melis L., which is composed of species distributed in
southeastern Asia, eastern North America, and northern
Mexico (Mione and Bogle, 1990). In the Hamamelida-
ceae, while many genera are either monotypic or oligo-
typic, there are several genera that contain more than ten
species, such as Corylopsis Sieb. et Zucc., Dicoryphe,
and Distylium Sieb. et Zucc. (Chang, 1979; Endress,
1993; Rakotobe, 1996).
Morphological characters in the Hamamelidaceae are
highly diverse. For example, leaves are persistent or de-
ciduous, simple and pinnately veined, or palmately lobed
and veined. Most species are bisexual, but some are an-
dromonoecious, and still others are monoecious. Flowers
are complete and five-merous in most genera, four-mer-
ous in several genera, and variable in others; a few genera
have an incomplete perianth, or are naked. Flowers are
insect-, or bird-, or wind-pollinated (Bogle, 1970; En-
dress, 1989a; Li, 1997).
Most of the traditional classification systems place the
Hamamelidaceae in the ‘‘Lower’’ Hamamelidae, which
includes Cercidiphyllaceae, Tetracentraceae, Trochoden-
draceae, Daphniphyllaceae, Platanaceae, Myrothamna-
ceae, and Eupteleaceae (Cronquist, 1981; Takhtajan,
1997). Endress (1977) emphasized this family as a con-
necting taxon between the ‘‘Lower’’ and the ‘‘Higher’’
Hamamelidae (e.g., Betulaceae, Fagaceae, and Juglan-
daceae). Some members of the Hamamelidaceae have
1028 AMERICAN JOURNAL
also been considered as linking taxa between the ‘‘lower’’
hamamelids and some basal elements of rosids and as-
terids (Hufford, 1992; Chase et al., 1993; Endress, 1993;
Morgan and Soltis, 1993). A more comprehensive study
(more taxa and more sources of data) is needed to further
assess the systematic position of the Hamamelidaceae.
Nevertheless, the Hamamelidaceae (including the Altin-
gioideae) appears to be monophyletic based on both mor-
phological (Hufford, 1992) and DNA sequence data (Li,
Bogle, and Klein, unpublished data).
The name Hamamelidaceae dates back to 1818 when
Brown recognized it as a natural group with four genera,
Hamamelis, Dicoryphe, Dahlia Thunb. (5 Trichocladus
Pers.), and Fothergilla. Since then, more than half a doz-
en classification systems have been proposed for the Ha-
mamelidaceae (see review in Tong, 1930; Bogle, 1968;
Chang, 1979; Endress, 1989c). However, the most com-
prehensive classification system for the Hamamelidaceae
was put forward by Harms (1930). As shown in Table 1,
based on morphological and anatomical characteristics,
Harms recognized five subfamilies, the largest of which,
the Hamamelidoideae, was further divided into five
tribes. However, he did not classify the two then little-
known genera Mytilaria Lecomte and Ostrearia. Schul-
ze-Menz (1964) transferred Sinowilsonia from the tribe
Distylieae into the Corylopsideae, which had included
Corylopsis and Fortunearia. Chang (1973, 1979) erected
a new subfamily Mytilarioideae for Mytilaria and Chunia
Chang.
Endress (1989c) reviewed the previous classification
systems, especially those of Harms and Chang, for the
Hamamelidaceae, and recognized four subfamilies: Altin-
gioideae, Exbucklandioideae, Rhodoleioideae, and Ha-
mamelidoideae (Table 1). The Exbucklandioideae are a
combination of the Exbucklandioideae Harms, Disan-
thoideae Harms, and the Mytilarioideae Chang. For the
largest subfamily Hamamelidoideae, Endress (1989b, c)
united the tribes Fothergilleae sensu Harms and Disty-
lieae sensu Harms, forming the tribe Fothergilleae sensu
Endress. Another major revision in Endress’s system is
that Fortunearia and Sinowilsonia were grouped with
Eustigma Gardn. et Champ., thus establishing a more in-
clusive Eustigmateae. Furthermore, Endress (1989c)
treated the five Southern Hemisphere genera (Dicoryphe,
Neostrearia, Noahdendron, Ostrearia, Trichocladus) as
one of the three subtribes in the tribe Hamamelideae.
As can be seen from the comparison of the major clas-
sification systems, fundamental questions concerning the
systematics of the Hamamelidaceae still remain. That is,
are the subfamilies, tribes, and subtribes monophyletic?
In other words, do the classification systems reflect nat-
ural relationships in the Hamamelidaceae?
Nucleotide sequences of a chloroplast gene, rbcL, have
been extensively used to examine plant phylogenies at
higher taxonomic levels (Chase et al., 1993; Qiu et al.,
1998). However, to resolve phylogenetic relationships
among closely related genera, a fast-evolving DNA frag-
ment, the nrDNA ITS region, has proven to be more use-
ful (Baldwin, 1992; Baldwin et al., 1995; Bogler and
Simpson, 1996; Downie and Katz-Downie, 1996; Kron
and King, 1996; Schilling and Panero, 1996; Soltis, John-
son, and Looney, 1996). A phylogenetic analysis of the
Hamamelidaceae using nrDNA ITS data has shown that
OF BOTANY [Vol. 86
this region is informative in resolving subfamilial and
tribal relationships (Shi et al., 1998). However, more than
half of the recognized genera were missing in Shi et al.’s
data set, and a study with a much broader sampling is
needed to address the intergeneric relationships in the Ha-
mamelidaceae.
Therefore, the objectives of this study were as follows:
(1) to reconstruct phylogenetic relationships of the Ha-
mamelidaceae using ITS sequence data, and (2) to eval-
uate previous classification systems of the family in terms
of the monophyly of the subfamilies, tribes, and subtribes
of the Hamamelidaceae.
MATERIALS AND METHODS
Plant material—Thirty-two species were sampled in this study (Table
2), representing 28 genera in the Hamamelidaceae. Three genera were
not available for analysis: Chunia, Semiliquidambar, and Embolanthera.
Vouchers for the sampled species are deposited in the Hodgdon Her-
barium of the University of New Hampshire (NHA). Sequences of both
ITS-1 and ITS-2 have been submitted to the GenBank, and the acces-
sion numbers are given in Table 2. The aligned sequences are available
from the first author and American Journal of Botany’s website.
Molecular techniques—Total genomic DNAs were extracted from
fresh or silica-gel-dried leaves or buds following the protocol of Doyle
and Doyle (1987). Polymerase chain reactiony (PCR) was conducted
in 0.2-mL thin-walled microcentrifuge tubes following the procedures
described in Li et al. (1997) and Li, Bogle, and Klein (1998). The PCR
primers were universal ITS4 and ITS5 of White et al. (1990).
The PCR products were purified following Li et al. (1997). The re-
sulting PCR product was used directly as a sequencing template. Se-
quencing reactions were carried out using Cycle Sequencing Kits and
following the manufacturer’s protocols (Applied Biosystems, Foster
City, California).
The sequencing primers were ITS2, ITS3, ITS4 and ITS5 of White
et al. (1990). The cycle sequencing products were then separated on
6% polyacrylamide gel using an Automated Sequencer 373A (Applied
Biosystems, Foster City, California) in the Sequencing Facility Center
of the University of New Hampshire (UNH).
For Liquidambar, Tetrathyrium, and Maingaya, 0.8% DMSO (di-
methylsulfoxide) was added to both the PCR and cycle sequencing re-
actions.
The chromatograms were analyzed using the SEQED program (Ap-
plied Biosystems, Foster City, California). Also, in order to assure cor-
rect base-calling, we overlapped sequences generated from adjacent
primers of either the same or opposite directions. The boundaries of
ITS-1 and ITS-2 were determined by comparing sequences of the 39
18S and 5.8S and the 59 end of the 26S ribosomal genes of Canella
winterana (GenBank accession number GBAN-L03844; the prefix
GBAN- has been added for linking online version of American Journal
of Botany to GenBank but is not part of the actual accession number).
Sequence alignment—The ITS sequences within a genus and among
closely related genera were easily aligned by sight. However, ITS se-
quences among genera of different subfamilies were not readily align-
able, thus resulting in some ambiguous regions. These ambiguous sites,
as identified by alignability by eye, can be eliminated from the data
matrix before a phylogenetic analysis is conducted (Downie and Katz-
Downie, 1996). This ‘‘culled’’ method tends to create clades where
internal relationships are not well resolved (Wheeler, Gatesy, and De-
salle, 1995; Soltis, Johnson, and Looney, 1996).
Another way of dealing with alignments with ambiguous sites is to
select a so-called optimal alignment. This approach involves consisten-
cy indices of the aligned sequences and the phylogenetic trees produced
based on the data. Bogler and Simpson (1996) used this ‘‘optimality’’
 July 1999]

TABLE 1. Comparison of two major classification systems of the Hamamelidaceae (? 5 uncertain, N/A 5 unapplicable).

Harms (1930) Endress (1989c)

Genus Subfamily Tribe Subfamily Tribe Subtribe

Disanthus Maxim. 1. Disanthoideae III. Exbucklandioideae

LI

Hamamelis L. 2. Hamamelidoideae 1. Hamamelideae IV. Hamamelidoideae I. Hamamelideae I. Hamamelidinae

Loropetalum R. Br. ex Reichb. 2. Hamamelidoideae 1. Hamamelideae IV. Hamamelidoideae I. Hamamelideae II. Loropetalinae
Tetrathyrium Benth. 2. Hamamelidoideae 1. Hamamelideae IV. Hamamelidoideae I. Hamamelideae II. Loropetalinae
Embolanthera Merr. 2. Hamamelidoideae 1. Hamamelideae IV. Hamamelidoideae I. Hamamelideae II. Loropetalinae
Maingaya Oliv. 2. Hamamelidoideae 1. Hamamelideae IV. Hamamelidoideae 1. Hamamelideae II. Loropetalinae
Dicoryphe Du Petit-Thouars 2. Hamamelidoideae 1. Hamamelideae IV. Hamamelidoideae I. Hamamelideae III. Dicoryphinae
Trichocladus Pers. 2. Hamamelidoideae 1. Hamamelideae IV. Hamamelidoideae I. Hamamelideae III. Dicoryphinae
Ostrearia Baill. ? ? IV. Hamamelidoideae I. Hamamelideae III. Dicoryphinae
Neostrearia L. S. Smith N/A N/A IV. Hamamelidoideae I. Hamamelideae III. Dicoryphinae
Noahdendron Endress, Hyland & Tracey N/A N/A IV. Hamamelidoideae I. Hamamelideae III. Dicoryphinae
Eustigma Gard. & Champ. 2. Hamamelidoideae 2. Eustigmateae IV. Hamamelidoideae II. Eustigmateae
Fortunearia Rehd. & Wils. 2. Hamamelidoideae 3. Corylopsideae IV. Hamamelidoideae II. Eustigmateae
Sinowilsonia Hemsl. 2. Hamamelidoideae 5. Distylieae IV. Hamamelidoideae II. Eustigmateae
Corylopsis Sieb. & Zucc. 2. Hamamelidoideae 3. Corylopsideae IV. Hamamelidoideae III. Corylopsideae
ET AL.—PHYLOGENETICS OF

Molinadendron Endress 2. Hamamelidoideae 4. Fothergilleae IV. Hamamelidoideae IV. Fothergilleae

Fothergilla Murray 2. Hamamelidoideae 4. Fothergilleae IV. Hamamelidoideae IV. Fothergilleae
Parrotiopsis Schneider 2. Hamamelidoideae 4. Fothergilleae IV. Hamamelidoideae IV. Fothergilleae
Parrotia C. A. Mey. 2. Hamamelidoideae 5. Distylieae IV. Hamamelidoideae IV. Fothergilleae
Sycopsis Oliv. 2. Hamamelidoideae 5. Distylieae IV. Hamamelidoideae IV. Fothergilleae
Distyliopsis Endress N/A N/A IV. Hamamelidoideae IV. Fothergilleae
Distylium Sieb. & Zucc. 2. Hamamelidoideae 5. Distylieae IV. Hamamelidoideae IV. Fothergilleae
Matudaea Lundell N/A N/A IV. Hamamelidoideae IV. Fothergilleae
Rhodoleia Champ. ex Hook. 3. Rhodoleioideae II. Rhodoleioideae
Exbucklandia R. W. Br. 4. Exbucklandioideae III. Exbucklandioideae
HAMAMELIDACEAE

Chunia Chang N/A N/A III. Exbucklandioideae

Mytilaria Lecomte ? ? III. Exbucklandioideae
Altingia Nor. 5. Altingioideae I. Altingioideae
Semiliquidambar Chang N/A N/A I. Altingioideae
Liquidambar L. 5. Altingioideae I. Altingioideae
1029
1030 AMERICAN JOURNAL OF BOTANY [Vol. 86

TABLE 2. Species sequenced for nrDNA ITS for this analysis.

Taxon GenBank
Species Abbr. Collector and voucher Source accessionsa

Altingia sp. ALT Y.-L. QIU China GBAN-AF015417-18

Corylopsis pauciflora S.& Z. CPA J.-H. LI 01 Arnold Arboretum, MA. GBAN-U65462
Corylopsis sinensis Hemsl. CSI J.-H. LI 02 Arnold Arboretum, MA. GBAN-U65461
Corylopsis spicata S.& Z. CSP J.-H. LT 03 Arnold Arboretum, MA. GBAN-U65463
Dicoryphe stipulacea Janume St. Hil. DIC A. Randrianasolo 543 Tulear, Madagascar GBAN-AF015419-20
Disanthus cercidifolius Max. DIS A. L. BOGLE Woodlanders, Inc. SC. GBAN-AF015421-22
Distyliopsis tutcheri Endress DOT A. L. BOGLE Woodlanders, Inc. SC. GBAN-AF019231-2
Distylium myricoides Hemsl. DIM A. L. BOGLE Woodlanders, Inc. SC. GBAN-U65464
Distylium racemosum S.& Z. DIR A. L. BOGLE Woodlanders, Inc. SC. GBAN-U65465
Eustigma oblongifolium Gardn. & EUS N.-J. CHUNG Taiwan GBAN-U65466
Champ.
Exbucklandia populnea (R.Br.) R.W. EXB A. L. BOGLE Manuka State Roadside Park, Hawaii GBAN-AF015423-24
Br.
Fortunearia sinensis R.& W. FOR J.-H. LI 04 Arnold Arboretum, MA. GBAN-U65467
Fothergilla major Lodd. FOT J.-H. LI Univ. of New Hampshire campus GBAN-AF015425-26
Hamamelis virginiana L. HVI J.-H. LI Univ. of New Hampshire campus GBAN-AF015656-57
Hamamelis vernalis Sarg. HVE J.-H. LI Arnold Arboretum, MA. GBAN-AF015655
Liquidambar styraciflua L. LIQ A. L. BOGLE Univ. of New Hampshire greenhouse. GBAN-AF055886
Loropetalum chinense (R.Br) Oliv. LOR A. L. BOGLE Missouri Bot. Gard. GBAN-AF015427-28
Maingaya malayana Oliv. MAI L. G. SAW Kepong, Malaysia GBAN-AF022241
Matudaea trinervia Lund. MAT P. K. ENDRESS Botanical Garden of Zurich, Switzerland GBAN-AF015437
Molinadendron guatemalense Endress MOL P. K. ENDRESS Botanical Garden of Zurich, Switzerland GBAN-AF015438
Mytilaria laosensis Lec. MYT Z.-C. LUO Guangxi, China GBAN-U65469
Neostrearia fleckeri Smith NEO P. K. ENDRESS Botanical Garden of Zurich, Switzerland GBAN-AF015439
Noahdendron nicholasii Endress, Hy-
land & Tracey NOA P. K. ENDRESS Botanical Garden of Zurich, Switzerland GBAN-AF015440
Ostrearia australiana Baill. OST P. K. ENDRESS Botanical Garden of Zurich, Switzerland GBAN-AF015441
Parrotia persica C.A.Mey PAR A. L. BOGLE Univ. of New Hampshire greenhouse GBAN-AF015443
Parrotiopsis jacquemontiana Rehd. PPS A. L. BOGLE Harvard Univ. campus GBAN-AF015442
Rhodoleia championii Hook. f. RHO A. L. BOGLE Lyon Arboretum, Hawaii GBAN-AF015429-30
Shaniodendron subaequale Deng, Wei
& Wang SHA Y.-L. QIU Jiangsu, China GBAN-AF015431-32
Sinowilsonia henryi Hemsl. SIN J.-H. LI 05 Arnold Arboretum, MA GBAN-U65468
Sycopsis sinensis Oliv. SYC A. L. BOGLE Woodlanders, Inc. SC. GBAN-AF015433-34
Tetrathyrium subcordatum Benth. TET R. Saunders Hong Kong GBAN-AF022242
Trichocladus crinitus Pers. TRI A. L. BOGLE Longwood Gardens, PA. GBAN-AF015435
GBAN-AF019233
a The prefix GBAN- has been added for linking the version of American Journal of Botany to GenBank but is not part of the actual GenBank

accession number.

method to analyze the phylogenetic relationships of the Agavaceae.
Many indices can be used to assess the optimality of sequence align-
ment, including the number of trees generated, the Consistency Index
(CI), Retention Index (RI), and Rescaled Consistency Index (RC). Bog-
ler and Simpson (1996), however, implied that the higher the RC was,
the better the alignment. This seems to be reasonable because the RC
index excludes characters that do little to the ‘‘fit’’ of the tree but inflate
the CI (Wiley et al., 1991).
In this study, the optimality method was employed to generate a data
matrix of the ITS sequences of the Hamamelidaceae. The alignment
that created trees with the highest RC index was considered as optimal
for both ITS-1 and ITS-2. Individual alignments were conducted using
the CLUSTAL option of the MEGALIGN program of DNA* software
package (DNA* Inc., Madison, Wisconsin).
Phylogenetic analysis—The resulting data matrices from the opti-
mality alignments were imported into the beta-test version of PAUP*
4.0d62 computer program for phylogenetic analyses, written by David
L. Swofford (1997) at the Smithsonian Institution, with permission.
Indels were treated as missing data because this coding strategy re-
tains information about substitutions that occur in other taxa in the indel
region. However, it does not convey the information regarding the evo-
lutionary event involved in the insertion or deletion (Platnick, Griswold,
and Coddington, 1991; Wojciechowski et al., 1993). Thus, a parsimony
analysis was also conducted treating gaps as the fifth character state.
All characters and their states were equally weighted in the parsimony
analyses. Sequence divergence was analyzed using the pairwise differ-
ence obtained from PAUP*. Due to the size of the data set and the
limitation of computer memory, the heuristic search option was used to
find the shortest trees with TBR (Tree Bisection and Reconnection)
branch swapping, MULPARS on, and STEEPEST DESCENT off.
It became impossible to produce reasonable sequence alignments
when we tried to use taxa outside the Hamamelidaceae as outgroups.
Therefore, in the parsimony analysis, Altingia and Liquidambar were
used as outgroups because: (1) phylogenetic analyses using ITS data
(Shi et al., 1998) and evidence from morphology and matK gene se-
quences (Li, 1997) have revealed that these two genera form a clade
sister to the clade containing the rest of the Hamamelidaceae, and (2)
the fossil record has shown that these genera are the most ancient mem-
bers in the Hamamelidaceae (Zhang and Lu, 1995). As a result, we
focus our analysis and discussion on the Hamamelidaceae s.s. (Hama-
melidaceae minus Altingioideae).
Both bootstrap and decay analyses were conducted using the PAUP*
program to test the relative strength of putative clades (Felsenstein,
1985; Bremer, 1988; Donoghue et al., 1992).
MacClade 3.03 (Maddison and Maddison, 1992) was used to trace
unambiguous changes along branches and to compare competing hy-
potheses concerning the relationships.
Both the skewness test (Huelsenbeck, 1991) and the permutation test
(Faith and Cranston, 1991) were conducted to evaluate the phylogenetic
July 1999] LI ET AL.—PHYLOGENETICS OF HAMAMELIDACEAE 1031

TABLE 3. Characteristics of the most parsimonious trees obtained by TABLE 4. Sequence characteristics of the two internal transcribed spac-
PAUP* heuristic searches of ITS-1 (1st line) and ITS-2 (2nd line) ers, separated and combined, in 32 species of the Hamamelidaceae.
sequence data sets using different combination of gap and gap
length penalties. Abbreviations: Consistency Index (CI), Retention Combined
ITS-1 ITS-2 (ITS-1 and ITS-2)
Index (RI), Rescaled Consistency Index (RC), number of steps
(Steps), number of trees found (Trees), Number of characters in the Length range (bp) 236–277 224–249 471–526
matrix (Nchar). The alignment considered optimal is in boldface Length mean (bp) 266.9 235 502.5
for ITS-1 and underlined for ITS-2. Aligned length (bp) 295 257 552
G1C content range (%) 57.5–69 61.3–68.2 60.8–68.3
Penalty G1C content mean (%) 63 64.9 64
(Gap/gap
length) Steps Trees Nchar CI RI RC Sequence divergence (%) 0.7–39.3 0.9–33.9 1.4–33.8
Number of indels 49 47 96
10/10 574 180 291 0.613 0.690 0.423 Number of variable sites 204 (69.2%) 159 (61.9%) 363 (65.8%)
438 126 259 0.564 0.628 0.354 Number of potentially in-
15/8 581 1008 291 0.608 0.689 0.419 formative sites 146 (49.5%) 117 (45.5%) 263 (47.6%)
439 270 259 0.563 0.629 0.354 Number of constant sites 91 (30.8%) 98 (38.1%) 189 (34.2%)
20/20 658 117 284 0.565 0.658 0.372 Transitions (average) 334 261 604
479 162 251 0.572 0.631 0.361 Transversions (average) 208 192 401
20/8 586 312 289 0.602 0.690 0.416 Transition/transversion 1.6 1.4 1.5
452 162 257 0.573 0.637 0.365 Skewness of tree length
distribution (g1 value
8/15 572 268 289 0.614 0.696 0.427 for 10 000 random
450 126 255 0.562 0.628 0.353 trees) 20.9 20.9 20.9
8/20 579 858 288 0.605 0.696 0.421
462 459 252 0.569 0.633 0.360
8/8 541 180 295 0.623 0.695 0.433
411 81 262 0.562 0.610 0.343 tween the two genera of Altingioideae and the Hama-
melidaceae s.s. was ;30% for both spacers (Table 5).

Phylogenetic relationships—The parsimony analysis

information contained in the ITS data matrix. The skewness test was
implemented using the Random tree option of PAUP*, and 10 000
random trees were examined. The permutation test, as used in Plunkett,
Soltis, and Soltis (1997), was performed using the Permutation option
of PAUP* with 100 replicates and heuristic searches.
RESULTS
Sequence alignment—Among the seven pairs of
alignment parameters, Gps (Gap Penalty)/GLPs (Gap
Length Penalty), including 8/8, 8/15, 10/20, 10/10, 15/8,
20/8, and 20/20, the pair of 8/8 resulted in the alignment
of the highest RC index for ITS-1 and that of 20/8 pro-
duced the optimal alignment for ITS-2 (Table 3).
Sequence characteristics—The lengths of ITS-1 and
ITS-2 ranged from 236 to 277 bp (base pairs) and from
224 to 249 bp, respectively, in the Hamamelidaceae, and
the means were 267 bp for ITS-1 and 236 for ITS-2.
Several genera had roughly equal lengths of ITS-1 and
ITS-2, including Fortunearia, Eustigma, Molinadendron,
and Sinowilsonia. GC content ranged from 57 to 69%
with an average of 63% in ITS-1 and from 61 to 68%
with an average of 65% in ITS-2 (Table 4). Statistical
analysis (chi-square test) showed that the differences of
nucleotide composition in both ITS-1 and ITS-2 across
taxa, and between the two spacers were not significant
(P . 0.9).
The percentages of potentially informative sites were
49.5 and 45.5% for ITS-1 and ITS-2, respectively. The
alignment of the sequences required 49 and 47 indels in
ITS-1 and ITS-2, respectively (Table 4). In both spacers,
;60% of the indels were a single base in length. There
were three indels of ten or more bases in the ITS region:
two indels (12 and 26 bases) were in ITS-1, and one (ten
bases) in ITS-2. Pairwise sequence divergence in ITS-1
was generally higher than that in ITS-2, with averages of
18.5 and 16.6%, respectively. Pairwise divergence be-
with gaps as missing data using the optimality data of
ITS-1 and ITS-2 yielded 90 shortest trees of 1005 steps,
with a consistency index of 0.59. Figure 1 is the strict
consensus tree. Mytilaria, Exbucklandia, and Rhodoleia
formed the basal clade, while Disanthus was sister to the
clade containing the sampled genera of the Hamameli-
doideae. In the Hamamelidoideae, three major clades
were recognized: the Corylopsis clade, Hamamelis clade,
and the Trichocladus clade, among which the Corylopsis
clade was basal. Both Corylopsis and Hamamelis clades
were strongly supported, whereas the Trichocladus clade
was weakly supported (Fig. 1).
The Corylopsis clade consisted of two subclades, the
first of which was composed of three species of Cory-
lopsis; the second included Loropetalum, Tetrathyrium,
Maingaya, and Matudaea. Both subclades were strongly
supported (Fig. 1).
In the Trichocladus clade, four branches were recog-
nized: Dicoryphe, Trichocladus, the Eustigmateae sensu
Endress (1989b) Plus Molinadendron, and the three Aus-
tralian genera. Among the three Australian genera, Neos-
trearia was basally sister to the latter two genera. In the
clade of the Eustigmateae sensu Endress (1989b) plus
Molinadendron, Molinadendron and Sinowilsonia were
bound with a bootstrap value of 77% and a decay index
of three steps (Fig. 1).
The Hamamelis clade was essentially the Fothergilleae
sensu Endress (1989c), plus Hamamelis. There were four
clades whose relationships were not well resolved in the
phylogeny, including Hamamelis, Fothergilla, Parrotiop-
sis, and the well-supported Distylium group consisting of
Distylium, Distyliopsis, Parrotia, Shaniodendron, and Sy-
copsis (Fig. 1).
The phylogenetic analysis with gaps as the fifth char-
acter state generated 46 shortest trees of 1266 steps with
a consistency index of 0.6. The tree topology was the
same as the phylogeny in Fig. 1. with several exceptions
1032 AMERICAN JOURNAL OF BOTANY [Vol. 86

TABLE 5. Sequence divergence of ITS-1 and ITS-2 (below and above diagonal) in the Hamamelidaceae. The values are derived from pairwise
distances, calculated in PAUP* (mean distance 3 100), and gaps as missing data. Taxon abbreviations are as in Table 1.

Taxon CPA CSI CSP DIS DIC DIM DIR DOT EUS EXB FOR FOT HVE HVI LOR

CPA — 3 3.9 21.3 16.7 15.4 15.8 15.1 15.1 19.5 14.1 15 12 13.3 19
CSI 4.1 — 0.9 20.3 15.7 14 14 13.3 14.1 18.5 13.6 14 11.5 12.8 18
CSP 3.7 2.2 — 20.9 16.6 14.9 14.9 14.1 15 19 14.4 14.9 12.4 13.7 18.9
DIS 24.4 26.3 26.2 — 21.1 19.4 19.4 18.5 21.4 24.9 21.2 19.4 19.4 19.6 23.4
DIC 14.9 16.8 17.5 24.3 — 11.6 10.7 9.8 8.1 22.9 5.9 10.7 10.3 10 17.2
DIM 15.3 17.3 17.2 23.6 11.2 — 1.7 2.6 9.9 19.4 8.2 6.4 4.3 4.7 19.1
DIR 15.3 17.2 17.2 24 12 1.1 — 1.7 9.1 19.7 8.6 5.5 5.2 4.7 18.1
DOT 15.6 17.6 17.5 23.7 12 0.7 0.4 — 8.3 20.3 8.6 4.7 4.3 3.9 16.7
EUS 10.9 13.7 13.6 23.2 10.8 8 8.5 8.4 — 21.1 6 8.2 8.6 8.2 17.5
EXB 30.6 32.8 32.7 31.9 30.1 30.4 30.6 30.8 28.6 — 20.6 20.7 20.3 20.8 27.5
FOR 12.9 14.9 14.9 22.7 9.4 8 8.4 8.4 3.4 28.1 — 9.4 8.1 7.8 15.9
FOT 14 16.4 16 22.9 11.9 5.3 5.7 5.7 7.6 30.4 8.4 — 4.7 4.3 20.2
HVE 14.5 16.5 16.4 22.1 11.1 3.7 4.1 4.1 7.7 30.5 7.7 3.7 — 2.1 19.5
HVI 15.6 17.9 17.9 23.6 11.9 4.5 4.9 4.9 8.5 30.9 8.5 4.5 1.5 — 19.1
LOR 11.5 12.5 12.8 26.4 18 17.7 17.7 17.7 15.7 30.7 16.8 17.5 17.2 18.3 —
MAI 15.2 15.8 15 27.1 17.4 18.3 18.6 18.7 17.5 31.7 16.8 18.2 18.2 19.1 11.6
MAT 13.7 13.5 13.5 26 18.7 19.6 19.6 19.6 17 30.8 17.3 19 18.4 19.6 9.7
MOL 12.2 15 14.9 24.4 10.2 8.4 8.4 8.4 3.4 27.7 5.5 7.6 8.1 8.9 16.9
MYT 30 32.2 32.1 29.6 28.8 28.5 27.9 28.2 26 25.9 23.8 28.9 27.1 26.7 30.5
NEO 13.8 17.2 16.8 25.1 13.1 11.9 12.3 12.3 6.4 31 7.1 11.2 10.5 11.9 19.2
NOA 12.9 15.7 15.3 24.8 10.7 11.2 11.5 11.5 7.3 30.9 8 9.6 9.6 10.3 18.2
OST 11.6 14.9 14.5 25.1 10.4 10.4 10.8 10.8 7.7 32.5 8.5 9.2 9.2 10 17.6
PAR 16.8 18.7 18.6 24.8 12 1.9 3 2.6 8.6 29.7 8.4 5.7 3.4 4.1 18.4
PPS 16 17.9 17.9 23.6 13 6.3 7.1 7.1 8.5 29 9.3 6.4 5.6 6.3 18.7
RHO 28.7 30.8 31.5 28.8 28.2 30.1 29.4 29.7 27.2 19.1 25.7 28.8 28.2 28.6 29.3
SHA 16 18.3 18.3 23.6 10.8 2.2 3.3 3 9.4 30.8 8.5 6.1 3 4.5 18
SIN 13 15.9 15.8 26.1 10.7 8.8 9.2 9.2 3.8 30.9 5.1 9.3 8.5 9.3 17.4
SYC 17.5 19.4 19.3 26.5 13.5 3.4 4.1 3.7 9.8 31.5 9.8 7.5 5.2 5.6 18.7
TET 11.4 13.2 12.8 25.7 16.7 16.9 17.2 17.2 15.4 27.9 16.9 17.1 17.2 18.4 5.6
TRI 15.6 16.4 16.4 25.1 11.5 11.6 11.6 11.6 9.4 32.2 9.8 11.1 10.7 11.5 18.3
LIQ 31.9 32.6 32.6 33.2 34.1 30.4 29.6 30.4 27.2 34.1 27.2 32 30.9 31.2 33.2
ALT 35.5 37.4 36.9 36 35.6 32.7 32 32.8 31.4 39.3 30.5 34 32.9 32.9 37

(Fig. 2). In the Trichocladus clade, Trichocladus was the
basal taxon, followed by Dicoryphe, the paraphyletic
Australian genera, and the monophyletic Eustigmateae
sensu Endress (plus Molinadendron). In the Hamamelis
clade, the clade of Hamamelis species was sister to the
Fothergilleae sensu Endress (minus Matudaea and Mol-
inadendron), in which Fothergilla was the basal taxon.
DISCUSSION
Sequence characteristics—In the angiosperms ITS-1
and ITS-2 are each less than 300 bp long: ITS-1 ranges
from 187 to 298 bp and ITS-2 from 187 to 252 bp (Bald-
win et al., 1995; Downie and Katz-Downie, 1996; Padg-
ett, 1997). The length of the two spacers in the Hama-
melidaceae fell within these ranges, but ITS-1 sequences
in most of the species sampled were in the high end of
the range, except for the four genera (Eustigma, Fortu-
nearia, Molinadendron, and Sinowilsonia) that shared
two unique deletions (12 and 26 bp). The ITS-2 in the
Hamamelidaceae was relatively less variable in length,
mostly in the range of 224–249 bp, which agrees with a
generalization that ITS-2 is more conservative than ITS-
1 (Hershkovitz and Lewis, 1996; Liston et al., 1996).
A permutation test for the data set from the optimality
alignment resulted in a probability of 0.01, indicating that
the ITS data contain significant phylogenetic signal (Faith
and Cranston, 1991). This is consistent with the skewness
of tree length distribution (Table 4).
Polyphyletic Exbucklandioideae sensu Endress—The
Exbucklandioideae sensu Endress (1989c) includes four
genera, Chunia, Disanthus, Exbucklandia, and Mytilaria.
The morphological synapomorphies of this subfamily in-
clude large, persistent stipules, palmate venation, and 4–
6 ovules per carpel. However, the phylogenetic analysis
based on the ITS DNA sequences provided a different
picture of this taxon (Chunia was not available for this
study). Exbucklandia and Mytilaria were grouped with
Rhodoleia, while Disanthus formed its own clade (Fig.
1). This indicates that the Exbucklandioideae sensu En-
dress (1989c) is polyphyletic. Furthermore, when the
three genera (Disanthus, Exbucklandia, and Mytilaria)
were forced into a monophylectic group, 35 more steps
were required.
Mytilaria, Exbucklandia, and Rhodoleia have been
previously recognized as subfamilies Mytilarioideae, Ex-
bucklandioideae, and Rhodoleioideae, respectively
(Harms, 1930; Chang, 1948, 1973, 1979; Tahktajan,
1997). In the ITS phylogeny, Mytilaria was allied with
the clade of Exbucklandia and Rhodoleia, suggesting a
close relationship among the three genera. Exbucklandia
and Rhodoleia also shared 32 unambiguous base substi-
tutions (Fig. 1). Interestingly, this pattern agrees with an
earlier taxonomic treatment (Reinsch, 1889) and a rbcL
phylogeny (Chase et al., 1993; Qiu et al., 1998). Never-
theless, these three genera differ greatly in morphology.
Mytilaria and Exbucklandia share palmate venation (vs.
July 1999] LI ET AL.—PHYLOGENETICS OF HAMAMELIDACEAE 1033

TABLE 5. Continued.

MAI MAT MOL MYT NEO NOA OST PAR PPS RHO SHA SIN SYC TET TRI LIQ ALT

11.2 12.1 15.3 23.6 17.6 14 15 14.6 13.6 21.9 15.7 18.2 15.9 12.5 14.4 31.2 30.9
9.1 10.8 13.5 24.3 16.6 13 14.4 12.7 12.7 20 13.9 16.4 14.1 12 14.8 31.8 31.5
9.9 11.7 14.4 23.8 17.5 13.9 15.3 13.6 13.5 20.4 14.8 17.3 15 12.9 15.7 32.3 31.9
23.2 22 20.2 27.6 22.2 20.2 21.2 18 19.4 26.9 18.5 23.2 19.3 19.9 20 29.3 30.5
15.1 14.8 8.9 25.1 9.4 8.5 8.1 10.3 9.6 24.4 9.9 9.3 10.7 16.8 9.3 34 33.1
16.9 17.9 10.3 25.7 13.4 11.6 11.1 2.1 4.8 23.9 2.6 12.9 3.4 18.6 11.2 32.1 33.1
15.6 17 9.8 25.2 12.5 10.7 10.2 1.3 3.9 23 1.7 12.9 2.5 17.2 11.1 31.6 32
15.6 17.1 10.8 24.8 12.5 9.9 9.4 1.3 3 23.1 1.7 12.1 0.4 17.3 10.3 31 31.5
15.6 16.5 8.5 25.8 11.2 7.7 7.7 8.6 8.4 23.6 9.1 9.4 9.1 17.7 10.3 30.1 31.9
21.3 23.8 19.7 27.2 24.5 24.5 21.9 19.3 20.6 15.9 20.5 24.7 20.7 22.4 23.7 30 31
14.2 18.2 5.9 26.9 9.8 6.8 8.1 9 7.9 23 9.4 7.6 9.4 17.5 10.2 30 31.2
16.7 17.4 11.5 24.6 12.4 10.2 9.8 4.3 3.9 24.4 5.5 11.9 5.1 18.5 11.6 32.3 32.7
16.4 17.1 12 26 12.9 10.3 9.8 3.9 3.5 24 5.2 12 5.2 17.8 9.9 33.1 33.6
16.1 16.8 11.6 26.2 13 10.4 9.5 3.4 3 23.7 4.7 12.1 4.7 17.4 10.3 32.9 33.3
19.9 11.2 15.8 32.2 17.3 14.5 17.6 17.6 18.2 28 17.1 19.6 17.5 14.1 19 29.7 30.8
— 14.7 14.5 24.4 16.9 14 15.5 15.1 14.6 22.7 16.3 18.8 16.4 15.1 17.5 29.9 30.9
11.2 — 14.7 27.2 17.5 14.7 15.1 16.6 15.4 23.7 17 17.2 17.9 9.4 17.7 28.1 28.5
18.2 17.8 — 27.9 12 10.2 9.7 10.3 11.3 23 10.7 7.6 11.6 16.7 11 30.8 32
32.7 31 25.6 — 26 25.8 25.8 25.1 24.6 30.4 24.4 29.1 25.6 25.1 25.6 32.3 30.1
21 21.1 6.8 30 — 9.4 8.1 12.9 12.7 24.4 12.5 13.6 13.3 15.5 12.4 30.1 30.6
18.9 20.1 7.6 30.5 6.7 — 5.5 10.3 9.6 24.3 10.7 10.6 10.7 14.4 11 30.8 31.7
18.3 19.2 8.1 30.8 7.3 5 — 9.8 10 23.4 10.2 10.6 10.2 15 10.2 29.4 30.4
19.5 20 8.9 27.5 12 11.9 11.2 — 2.6 22.6 1.3 12.4 2.1 16.8 10.7 31.7 32.1
20.5 20.3 8.9 28.8 12.2 12.5 11.5 5.2 — 23.7 3.9 11.7 3.9 16.8 10.9 30.4 30.8
29.5 28.3 25.8 23.8 29.1 30.3 28.6 29.1 27 — 23.3 24.9 23.9 22.8 22.7 33.8 33.1
18.3 19.2 9.8 27.9 12.3 12.3 11.6 1.8 6.3 29.1 — 12.8 2.5 16.2 11.1 31.4 31.4
18.7 19.2 4.7 26.8 8 9 9.4 9.3 10.1 29.6 10.2 — 12.9 20.5 11.9 33.9 34.7
19.5 21.1 9.4 29.1 14.3 13 12.7 3.4 7.9 30.5 3.7 10.6 — 18.1 11.1 31.8 32.3
11.2 10.5 17 28.7 19.9 18.9 18.3 16.9 18.3 28.5 16.5 17.9 18.4 — 18.4 30.1 30.2
18.5 19.1 8.9 30.4 11.5 9.2 9.6 13.1 12.2 30.5 13.5 11.1 13.4 18.2 — 32.5 33.1
33.2 31.8 26.8 31.7 33.5 31.8 31.2 30.6 31.1 31.2 31 29.5 31 32.9 32.4 — 5.4
36.5 36.3 31.3 36.8 36.1 33.8 34.3 32.9 34 36.7 33.7 31.8 34.7 37 35.1 15.5 —

pinnate venation in Rhodoleia), but differ in nodal anat-
omy (multilacunar vs. trilacunar; Bogle, 1990), inflores-
cence type (spadix vs. head), and chromosome base num-
ber (x 5 13 vs. x 5 8; Pan and Yang, 1994). Rhodoleia
is distinct from Mytilaria and Exbucklandia in a suite of
characters, including bird-pollinated, asymmetric flowers,
and a chromosome base number of x 5 12. In the ITS
region, these three genera diverged greatly from their
common ancestor (Table 5, Fig. 1). It therefore seems
reasonable to continue to treat them as belonging to in-
dividual subfamilies (Harms, 1930; Chang, 1973, 1979;
Bogle, 1989; Endress, 1989b, c).
Disanthus was the immediate sister taxon to the Ha-
mamelidoideae in the phylogenetic tree (Fig. 1), reminis-
cent of the conclusions from previous morphological
analyses (Hufford and Crane, 1989; Pan, Lu, and Wen,
1991). The intermediate systematic position of Disanthus
between the Exbucklandia-Rhodoleia-Mytilaria clade and
the Hamamelidoideae in the ITS-based tree does not pro-
vide evidence for the hypothesis that this genus is most
primitive in the Hamamelidaceae, as has been suggested
by some authors (Reinsch, 1889; Harms, 1930; Tong,
1930; Cronquist, 1981; Pan, Lu, and Wen, 1991; Takh-
tajan, 1997).
In the previous nrDNA ITS phylogeny (Shi et al.,
1998), Exbucklandia is sister to the Hamamelidoideae,
whereas Disanthus and Rhodoleia form a clade. This re-
sult does not agree with morphological analysis (Hufford
and Crane, 1989), or the ITS analysis of this study, or
the chloroplast gene phylogenies (rbcL—Chase et al.,
1993; matK—Li, 1997; trn noncoding regions—Li et al.,
unpublished data). The difference is probably due to
long-branch attractions resulting from the absence of My-
tilaria or from the use of Corylus (Betulaceae, higher
Hamamelidae) as the outgroup in Shi et al.’s analysis. In
addition, we found DNA sequence differences for the
same species sampled in both this study and Shi et al.’s
(1998).
Monophyly of the Hamamelidoideae—It has been
widely recognized that the Hamamelidoideae is a mono-
phyletic group, characterized by a ballistic seed dispersal
mechanism and one mature seed per carpel (Bogle, 1968;
Endress, 1989a, b, c; Hufford and Crane, 1989). The
present analysis adds strong evidence to this hypothesis
because the clade of this subfamily received a 100%
bootstrap value and a decay index of more than five steps,
and was further supported by 36 unambiguous base sub-
stitutions (Fig. 1). This result is consistent with the pre-
vious ITS phylogeny (Shi et al., 1998).
The intergeneric relationships within the Hamameli-
doideae, however, have been contentious (Harms, 1930;
Schulze-Menz, 1964; Endress, 1989b, c). The relation-
ships among the five tribes recognized by Harms (1930)
and more recently revised by Endress (1989c) and others
are discussed below.
Corylopsideae—The concept of the tribe Corylopsi-
deae has been revised several times in the past several
decades. Harms (1930) delimited Corylopsideae as in-
1034 AMERICAN JOURNAL
Fig. 1. Strict consensus of 90 shortest trees of 1005 steps, generated
using nrDNA ITS sequences of the Hamamelidaceae treating gaps as
missing data. Taxon abbreviations are as in Table 2. Numbers above
and below branches are bootstrap percentages and decay values/number
of unambiguous changes, respectively. CI 5 0.59, and RC 5 0.39.
Grouping on the right side follow Endress (1989c). ALToideae 5 Al-
tingioideae, COReae 5 Corylopsideae, DICinae 5 Dicoryphinae, EU-
Seae 5 Eustigmateae, EXBoideae 5 Exbucklandioideae, FOTeae 5
Fothergilleae, HAMeae 5 Hamamelideae, HAMoideae 5 Hamameli-
doideae, LORinae 5 Loropetalinae, RHOideae 5 Rhodoleioideae.
cluding Corylopsis and Fortunearia based on their sim-
ilarity of leaf morphology, while Schulze-Menz (1964)
added Sinowilsonia to this tribe considering similar floral
structures of the latter to Fortunearia. However, Endress
(1989b, c) considered Corylopsideae as containing a sin-
gle genus, Corylopsis, characterized by the orbicular pet-
al.
In the ITS-based phylogenies (Figs. 1, 2), the three
species of Corylopsis sampled formed a well-supported
clade and were sister to the branch comprising Lorope-
talum, Tetrathyrium, Maingaya, and Matudaea, whereas
Fortunearia and Sinowilsonia were phylogenetically dis-
tant from Corylopsis. Therefore, Corylopsis is not closely
related to either Fortunearia or Sinowilsonia, as has al-
ready been shown in previous ITS analyses (Li, Bogle,
and Klein, 1998; Shi et al., 1998). Floral ontogeny and
embryological study also support the separation of Cor-
ylopsis from Fortunearia and Sinowilsonia (Li and Bo-
gle, 1998).
Hamamelideae—In Harms’s system (1930), all of the
OF BOTANY [Vol. 86
Fig. 2. Strict consensus of 48 shortest trees of 1266 steps, generated
using nrDNA ITS sequences of the Hamamelidaceae treating gaps as
the fifth character state. CI 5 0.6, RC 5 0.4. Numbers above and below
branches are bootstrap percentages and decay values, respectively.
genera in the Hamamelidoideae that have strap-shaped
petals were placed in the tribe Hamamelideae and no fur-
ther subdivisions were proposed. Endress (1989c), how-
ever, recognized three subtribes in the Hamamelideae, in-
cluding the Hamamelidinae, Loropetalinae, and the Di-
coryphinae. The Hamamelidinae contained a single ge-
nus, Hamamelis, characterized by the four-merous flower;
the Dicoryphinae included the five Southern Hemisphere
genera (Dicoryphe, Trichocladus, Ostrearia, Neostrearia,
and Noahdendron) and was defined by the distinct anther
dehiscence pattern; the Loropetalinae contained the re-
maining four genera, Loropetalum, Embolanthera, Main-
gaya, and Tetrathyrium.
The analysis of the ITS data placed the genera of the
Hamamelideae into three different clades (Fig. 1), sug-
gesting that the tribe is polyphyletic (as shown by Shi et
al., 1998). Maingaya, Tetrathyrium, and Embolanthera
have been suggested to be the ancestral stock of the Aus-
tralian hamamelids (Raven and Axelrod, 1974). However,
this study does not support that proposition because in
the phylogeny (Fig. 1) the monophyletic Australian gen-
era were distant from the lineage containing Maingaya
July 1999] LI ET AL.—PHYLOGENETICS OF
and Tetrathyrium (Embolanthera not available for this
study). The exclusive Southern Hemisphere distribution
and the unique pattern of anther dehiscence suggest a
close relationship of the three Australian and two African
genera (Endress, 1989a, b, c). However, in the strict con-
sensus trees (Figs. 1, 2), the relationships of the Austra-
lian and African genera were not well resolved, even
though these genera formed a clade in the 50% majority
consensus tree (tree not shown). More evidence is needed
to resolve relationships of the Southern Hemisphere gen-
era and thus to elucidate their biogeography.
Eustigmateae—The delimitation of the Eustigmateae
has expanded from monogeneric (Harms, 1930) to tri-
generic (Endress, 1989c). Eustigma is distinct in its small,
auriculate petals and greatly enlarged stigmatic surfaces
(Harms, 1930). The similarities of Eustigma with For-
tunearia in other morphological characters such as ped-
icellate, lenticellate fruits, and covering sepals led En-
dress (1989c) to the conclusion that Eustigmateae should
include Fortunearia and Sinowilsonia.
The ITS-based tree (Fig. 1) offers strong support for
Endress’s hypothesis and further corroborates the impli-
cation (Endress, 1967) that Molinadendron and Sinowil-
sonia are closely related. Interestingly, these four genera
shared nine unambiguous base substitutions (Fig. 1) and
two unique deletions totaling 38 bp (Li, Bogle, and Klein,
1998).
Combining the Distylieae and Fothergilleae—In the
phylogeny (Fig. 1), the clade of two Hamamelis species
was sister to a branch containing the Distylieae (Disty-
lium, Distyliopsis, and Sycopsis) and Fothergilleae sensu
Harms (1930) (Fothergilla, Parrotia, and Parrotiopsis).
Neither the Distylieae nor the Fothergilleae formed a
monophyletic clade. Instead, a close but unresolved re-
lationship was shown among Sycopsis, Parrotia, and
Shaniodendron. This result agrees with the finding that
Sycopsis and Parrotia are interfertile, giving rise to the
hybrid taxon, Sycoparrotia (Endress and Anliker, 1968).
Thus, the Fothergilleae sensu Endress (1989c), a merger
of Distylieae Harms and Fothergilleae Harms, is sup-
ported.
Fothergilla and Parrotiopsis each formed their own
clades in the ITS phylogeny (Fig. 1). This is consistent
with the showy specialized structures they have evolved
for insect pollination: long, white, clavate stamen fila-
ments in Fothergilla and large, white subfloral bracts in
Parrotiopsis (Bogle, 1970; Endress, 1989a). These spe-
cializations do not occur in any other members of the
Fothergilleae sensu Endress (1989c) or in the Hamame-
lidaceae.
The apetalous genera of the Hamamelidoideae have
long been considered as monophyletic (Harms, 1930; En-
dress, 1989c), but the ITS phylogeny (Fig. 1) suggested
that loss of petals in this subfamily has evolved at least
three times independently: Matudaea in the Corylopsis
clade, Molinadendron in the Trichocladus clade, and the
Fothergilleae in the Hamamelis clade. Therefore, the
apetalous genera do not belong to a single monophyletic
group, and apetaly is homoplasious in this subfamily.
It is a novel relationship that Hamamelis was grouped
with Fothergilleae sensu Endress (excl. Matudaea and
HAMAMELIDACEAE 1035
Molinadendron). Interestingly, this pattern finds support
from leaf venation (Harms, 1930; Chang, 1979; Li and
Hickey, 1988). The systematic position of Hamamelis
was not resolved in the ITS phylogeny using gaps as
missing data (Fig. 1); however, in the phylogenetic tree
using gaps as the fifth character state (Fig. 2), the clade
of the two Hamamelis species was sister to the Fother-
gilleae sensu Endress. Also, Hamamelis differs greatly
from the Fothergilleae in a group of floral characters,
including showy, strap-shaped petals, bisexuality, and
nectariferous phyllomes (Endress, 1989a). Therefore,
Hamamelis should be isolated from the Fothergilleae in
taxonomic treatments.
The evolutionary relationships shown in the phyloge-
netic trees (Figs. 1, 2) have significant implications for
character evolution and biogeography. For example, apet-
aly and wind-pollination have evolved independently
more than three times in the Hamamelidaceae. The Amer-
ican Matudaea (Central and South America) and Moli-
nadendron (Central America) were suggested to be close-
ly related to Asian Maingaya and Sinowilsonia, respec-
tively. The Southern Hemisphere genera were shown to
be a potentially monophyletic group by the 50% majority
consensus tree. However, we believe that a well-resolved
phylogeny based on multiple sources of data and detailed
studies of the fossil record are needed to reveal a com-
plete picture of character evolution and biogeography of
the Hamamelidaceae.
In conclusion, at the subfamily level, this study rec-
ognized the monophyly of the Exbucklandioideae Harms,
Mytilarioideae Harms, Rhodoleioideae Harms, Disan-
thoideae Harms, and the Hamamelidoideae. Furthermore,
it suggested close relationships of Exbucklandia, Rho-
doleia, and Mytilaria and the paraphyly of Exbucklan-
dioideae sensu Endress (1989c). At the tribal and subtri-
bal levels, the monophyletic groups were the corylopsi-
deae sensu Endress (1989c); Eustigmateae sensu Endress
(1989c), but expanded to include Molinadendron; Foth-
ergilleae sensu Endress (1989c), expanded to include
Hamamelis; Dicoryphinae; and Loropetalinae, expanded
to include Matudaea. The Hamamelideae, however, were
polyphyletic. The ITS phylogeny also revealed that some
morphological characteristics commonly used to classify
the Hamamelidoideae have evolved several times inde-
pendently. These characteristics included strap-shaped
petals, apetaly, and wind pollination. Thus homoplasy
needs to be taken into account in future classifications of
the Hamamelidaceae so that natural relationships can be
reflected.
LITERATURE CITED
BALDWIN, B. G. 1992. Phylogenetic utility of the internal transcribed
spacers of nuclear ribosomal DNA in plants: an example from the
Compositae. Molecular Phylogenetics and Evolution 1: 3–16.
———, M. J. SANDERSON, J. M. PORTER, M. F. WOJCIECHOWSKI, C. S.
CAMPBELL, AND M. J. DONOGHUE. 1995. The ITS region of nuclear
ribosomal DNA: a valuable source of evidence on angiosperm phy-
logeny. Annals of the Missouri Botanical Garden 82: 247–277.
BOGLE, A. L. 1968. Floral vascular anatomy and the nature of the
hamamelidaceous flowers. Ph.D. dissertation, University of Min-
nesota, Minneapolis, MN.
———. 1970. Floral morphology and vascular anatomy of the Ha-
mamelidaceae: the apetalous genera of Hamamelidoideae. Journal
of the Arnold Arboretum 51: 310–366.
1036 AMERICAN JOURNAL
———. 1986. The floral morphology and vascular anatomy of the Ha-
mamelidaceae: subfamily Liquidambaroideae. Annals of the Mis-
souri Botanical Garden 73: 325–347.
———. 1989. The floral morphology, vascular anatomy, and ontogeny
of the Rhodoleioideae (Hamamelidaceae) and their significance in
relation to the ‘‘lower’’ hamamelids. In P. R. Crane and S. Black-
more [eds.], Evolution, systematics, and fossil history of the Ha-
mamelidae, vol. 1, Introduction and ‘‘lower’’ Hamamelidae. Sys-
tematics Association Special Volume, Number 40A, 201–226. Ox-
ford University Press, New York, NY.
———. 1990. Multilacunar nodal anatomy in Mytilaria (Hamameli-
daceae). Journal of the Arnold Arboretum 71: 111–118.
BOGLER, D. J., AND B. B. SIMPSON. 1996. Phylogeny of Agavaceae
based on ITS rDNA sequence variation. American Journal of Bot-
any 83: 1225–1235.
BREMER, K. 1988. The limits of amino acid sequence data in angio-
sperm phylogenetic reconstruction. Evolution 42: 795–803.
BROWN, R. 1818. Appendix, -B, 374–379. In C. Abel. Narrative of a
journey in the interior of China. Longman, London.
CHANG, H.-T. 1948. Additions to the hamamelidaceous flora of China.
Sunyatsenia 7: 63–74.
———. 1973. A revision of the hamamelidaceous flora of China. Sun-
yatsenia 1: 54–71.
———. 1979. Hamamelidaceae. In Florae Reipublicae Popularis Sin-
icae. 35 (2): 36–116. Science Press, Beijing.
CHASE, M. W., ET AL. 1993. Phylogenetics of seed plants: an analysis
of nucleotide sequences from the plastid gene rbcL. Annals of the
Missouri Botanical Garden 80: 528–580.
CRONQUIST, A. 1981. An integrated system of classification of flower-
ing plants. Columbia University Press, New York, NY.
DONOGHUE, M. J., R. G. OLMSTEAD, J. F. SMITH, AND J. D. PALMER.
1992. Phylogenetic relationships of Dipsacales based on rbcL se-
quences. Annals of the Missouri Botanical Garden 79: 333–345.
DOWNIE, S. R., AND D. S. KATZ-DOWNIE. 1996. A molecular phylogeny
of Apiaceae subfamily Apioideae: evidence from nuclear ribosomal
DNA internal transcribed spacer sequences. American Journal of
Botany 83: 234–251.
DOYLE, J. J., AND J. L. DOYLE. 1987. A rapid isolation procedure for
small quantities of fresh leaf tissue. Phytochemistry Bulletin 19:
11–15.
ENDRESS, P. K. 1967. Systematische Studie über die verwandschaftlich-
en Beziehungen zwischen den Hamamelidaceen und Betulaceen.
Botanische Jahrbücher für Systematik 87: 431–525.
———. 1977. Evolutionary trends in the Hamamelidales-Fagales-
group. Plant Systematics and Evolution (Supplement 1): 321–347.
———. 1989a. Aspects of evolutionary differentiation of the Hama-
melidaceae and the lower Hamamelidae. Plant Systematics and
Evolution 162: 193–211.
———. 1989b. Phylogenetic relationships in the Hamamelidoideae. In
P. R. Crane and S. Blackmore [eds.], Evolution, systematics and
fossil history of the Hamamelidae, vol. 1, Introduction and ‘‘lower’’
Hamamelidae. Systematics Association Special Volume, Number
40A, 227–248. Oxford University Press, New York, NY.
———. 1989c. A suprageneric taxonomic classification of the Ha-
mamelidaceae. Taxon 38: 371–376.
———. 1993. Hamamelidaceae. In K. Kubitzki, J. G. Rohwer, and V.
Bittrich [eds.], The families and genera of vascular plants, II, Flow-
ering plants, Dicotyledons, 322–331. Springer-Verlag, Berlin.
———, AND J. ANLIKER. 1968. 3 Sycoparrotia semidecidua hybr. nov.
(5 Parrotia persica C. A. Mey. 3 Sycopsis sinensis Oliv.).
Schweizerische Beiträge zür Dendrologie 16–18: 6–28.
FAITH, D., AND P. CRANSTON. 1991. Could a cladogram this short have
arisen by chance alone? Cladistics 7: 1–28.
FELSENSTEIN, J. 1985. Confidence limits on phylogeny: an approach
using the bootstrap. Evolution 39: 783–791.
GENTRY, A. H. 1993. A field guide to the families and genera of woody
plants of northwest South America (Colombia, Ecuador, Peru): with
supplementary notes on herbaceous taxa. Conservation Internation-
al, Washington, DC.
HARMS, H. 1930. Hamamelidaceae. In A. Engler and K. Prantl [eds.],
Die natürlichen Pflanzenfamilien, ed. 2, vol. 18a, 303–345. Engel-
mann, Leipzig.
HERSHKOVITZ, M. A., AND L. A. LEWIS. 1996. Deep level diagnostic
OF BOTANY [Vol. 86
value of the rDNA-ITS region. Molecular Biology and Evolution
13: 1276–1295.
HUELSENBECK, J. P. 1991. Tree-length distribution skewness: an indi-
cator of phylogenetic information. Systematic Zoology 40: 257–
270.
HUFFORD, L. 1992. Rosidae and their relationships to other nonmag-
noliid dicotyledons: a phylogenetic analysis using morphological
and chemical data. Annals of the Missouri Botanical Garden 79:
218–248.
———, AND P. R. CRANE. 1989. A preliminary phylogenetic analysis
of the ‘‘Lower’’ hamamelidae. In P. R. Crane and S. Blackmore
[eds.], Evolution, systematics, and fossil history of the Hamame-
lidae, vol. 1, Introduction and ‘‘lower’’ Hamamelidae. Systematics
Association Special Volume Number 40A, 17–45. Oxford Univer-
sity Press, New York, NY.
KRON, K. A., AND J. M. KING. 1996. Cladistic relationships of Kalmia,
Leiophyllum, and Loiseleuria (Phyllodoceae, Ericaceae) based on
rbcL and nrITS data. Systematic Botany 21: 17–31.
LI, H.-M., AND L. J. HICKEY. 1988. Leaf architecture and systematics
of the Hamamelidaceae sensu lato. Acta Phytotaxonomica Sinica
96–110.
LI, J.-H. 1997. Systematics of the Hamamelidaceae based on morpho-
logical and molecular evidence. Ph.D. dissertation, University of
New Hampshire, Durham, NH.
———, AND A. L. BOGLE. 1998. Comparative embryology and floral
ontogeny of the Corylopsis complex (Hamamelidaceae). American
Journal of Botany 85 (6, supplement): 173 (Abstract).
———, ———, AND A. S. KLEIN. 1998. Phylogenetic relationships in
the Corylopsis complex (Hamamelidaceae) based on sequences of
the internal transcribed spacers of nuclear ribosomal DNA and mor-
phology. Rhodora 99: 302–318.
———, ———, ———, AND K. Y. PAN. 1997. Close relationship
between Shaniodendron and Parrotia (Hamamelidaceae), evidence
from ITS sequences of nuclear ribosomal DNA. Acta Phytotaxon-
omica Sinica 35: 481–493.
LISTON, A., W. A. ROBINSON, J. M. OLIPHANT, AND E. R. ALVAREZ BUYL-
LA. 1996. Length variation in the nuclear ribosomal transcribed
spacer region of non-flowering seed plants. Systematic Botany 21:
109–121.
LOZANO-CONTRERAS, G. 1996. Hallazgo de la familia Hamamelidaceae
en suramérica y descripción de una nueva especie de Matudaea de
Colombia. Revista de la Academia Colombiana Ciencias Exactas,
Fı́sicas y Naturales 20: 443–445.
MADDISON, W. P., AND D. R. MADDISON. 1992. MacClade: analysis of
phylogeny and character evolution, version 3.03. Sinauer, Sunder-
land, MA.
MIONE, T., AND A. L. BOGLE. 1990. Comparative ontogeny of the in-
florescence and flower of Hamamelis virginiana and Loropetalum
chinense (Hamamelidaceae). American Journal of Botany 77: 77–
91.
MORGAN, D. R., AND D. E. SOLTIS. 1993. Phylogenetic relationships
among members of Saxifragaceae sensu lato based on rbcL se-
quence data. Annals of the Missouri Botanical Garden 80: 633–
660.
PADGETT, D. J. 1997. A biosystematic monograph of the genus Nuphar
Sm. (Nymphaeaceae). Ph.D. dissertation, University of New Hamp-
shire , Durham, NH.
PAN, K.-Y., AND Q.-E. YANG. 1994. Karyotypes of Disanthus and My-
tilaria (Hamamelidaceae). Acta Phytotaxonomica Sinica 32: 235–
239.
———, A.-M. LU, AND J. WEN. 1991. A systematic study on the genus
Disanthus Maxim. (Hamamelidaceae). Cathaya 3: 1–28.
PLATNICK, N. I., C. E. GRISWOLD, AND J. A. CODDINGTON. 1991. On
missing entries in cladistic analysis. Cladistics 7: 337–343.
PLUNKETT, G. M., D. E. SOLTIS, AND P. S. SOLTIS. 1997. Clarification
of the relationship between Apiaceae and Araliaceae based on matK
and rbcL sequence data. American Journal of Botany 84: 565–580.
QIU, Y. L., M. W. CHASE, S. B. HOOT, E. CONTI, P. R. CRANE, K. J.
SYTSMA, AND C. R. PARKS. 1998. Phylogenetics of the Hamame-
lidae and their allies: parsimony analyses of nucleotide sequences
of the plastid gene rbcL. International Journal of Plant Sciences
159: 891–905.
RAKOTOBE, E. A. 1996. Le genre endemique malgache Dicoryphe Du
July 1999] LI ET AL.—PHYLOGENETICS OF
Petit-Thouars (Hamamelidacees): repartition et phytogeographie.
Biogéographie de Madagascar: 177–182.
RAVEN, P. H., AND D. I. AXELROD. 1974. Angiosperm biogeography
and past continental movements. Annals of the Missouri Botanical
Garden 61: 540–673.
REINSCH, A. 1989. Über die anatomischen Verhältnisse der Hamame-
lidaceae mit Rücksicht auf ihre systematisches Grüppierung. Bo-
tanische Jahrbücher für Systematik, Pflanzengeschichte und Pflan-
zengeographie 2: 347–395.
SCHILLING, E. E., AND J. L. PANERO. 1996. Phylogenetic reticulation in
subtribe Helianthinae. American Journal Botany 83: 939–948.
SCHULZE-MENZ, G. K. 1964. Rosales. In H. Melchior [ed.], A. Engler’s
Syllabus der Pflanzenfamilien, 12th ed., 2, 193–242. Borntraeger,
Berlin.
SHI, S.-H., H.-T. CHANG, Y.-Q. CHEN, L.-H. QU, AND J. WEN. 1998.
Phylogeny of the Hamamelidaceae based on the ITS sequences of
nuclear ribosomal DNA. Biochemical Systematics and Ecology 26:
55–69.
SOLTIS, D. E., L. A. JOHNSON, AND C. LOONEY. 1996. Discordance be-
tween ITS and chloroplast topologies in the Boykinia group. Sys-
tematic Botany 21: 169–187.
TAKHTAJAN, A. 1997. Diversity and classification of flowering plants.
Columbia University Press, New York, NY.
TONG, K. 1930. Studien über die Familie der Hamamelidaceae mit
HAMAMELIDACEAE 1037
besonderer Berücksichtigung der Systematik und Entwicklungsges-
chichte von Corylopsis. Bulletin of Department of Biology, College
of Science, Sun Yatsen University 2: 1–72.
ULLOA, U. C., AND P. M. JøRGENSEN. 1993. Arboles y arbustos de los
Andes del Ecuacor. A A U Reports 30, Aarhus University Press,
Risskov.
WHEELER, W. C., J. C. GATESY, AND R. DESALLE. 1995. Elision: a meth-
od for accommodating multiple molecular sequence alignments
with alignment-ambiguous sites. Molecular Phylogenetics and Evo-
lution 4: 1–9.
WHITE, T. J., T. BRUNS, S. LEE, AND J. TAYLOR. 1990. Amplification
and direct sequencing of fungal ribosomal RNA genes for phylo-
genetics. In M. Innis et al. [eds.], PCR protocols: a guide to meth-
ods and applications, 315–322.
WILEY, E. O., D. SIEGEL-CAUSEY, D. R. BROOKS, AND V. A. FUNK. 1991.
The compleat cladist—a primer of phylogenetic procedures. Spe-
cial Publication Number 19. Museum of Natural History, Univer-
sity of Kansas, Lawrence, KS.
WOJCIECHOWSKI, M. F., M. J. SANDERSON, B. G. BALDWIN, AND M. J.
DONOGHUE. 1993. Monophyly of aneuploid Astragalus (Fabaceae):
evidence from nuclear ribosomal DNA internal transcribed spacer
sequences. American Journal of Botany 80: 711–722.
ZHANG, Z.-Y., AND A. M. LU. 1995. Hamamelidaceae: Geographic dis-
tribution, fossil history and origin. Acta Phytotaxonomica Sinica
33: 313–339.