Cladistic analysis of morphological characters in Hypericum (Hypericaceae)

Abstract

Hypericum is a worldwide distributed genus with almost 500 species, including the medically used apomictic species H. perforatum. It is one of the few large plant genera where alpha taxonomy is nearly completed. To conduct a formal cladistic analysis of the genus, we coded 89 morphological characters for all described taxa, and analyzed the data with parsimony and Bayesian methods. The obtained trees indicate Hypericum to be monophyletic, if the monotypic genus Santomasia is included, and Lianthus as the sister group. The arrangement of the remaining genera of Hypericaceae included in the analysis is in congruence with molecular phylogenies. Apomorphic characters supporting the relationships of the genera are pointed out. The cladistic analysis revealed four groups within Hypericum: a basal grade containing Mediterranean species and three big clades containing most of the diversity of the genus. The borders of the Mediterranean Sea as part of the late Tethys Ocean are hypothesized as a probable area of origin for the genus. As indicated in the presented tree, a shrubby habit appears to be the ancestral state within Hypericum from which trees and herbs evolved, and apomixis originated at least three times independently within the genus.

Full text

1495
Nürk & Blattner •
Cladistic analysis of Hypericum
TAXON 59 (5) • October 2010: 1495–1507
INTRODUCTION
The genus Hypericum L. (St. John’s wort, Hypericaceae
Juss.) comprises almost 500 species of small trees, shrubs and
herbs occurring in temperate parts of the world, with a centre
of diversity in temperate regions of Eurasia. The species of the
genus can be recognized by their opposite exstipulate entire or
gland-fringed leaves, the presence of glandular secretions, and
choripetalous yellow flowers with stamens mostly in fascicles.
Typical habitats range from dry rocky places to moist wood-
land–meadow borders or fens and swamps. Hypericum is absent
only from habitats that are extremely hot, cold or dry, and is
rarely found in water other than in very shallow depths. In the
tropics, it is almost always confined to high altitude habitats
(Robson, 2003).
Hypericum and eight other genera have been treated as sub-
family Hypericoideae Engl. within Clusiaceae Lind. (Guttiferae
Juss.). Molecular studies, however, indicate that such a broadly
circumscribed Clusiaceae is paraphyletic as a result of a sister
group relationship between Hypericoideae and Podostemaceae
Kunth. (Chase & al., 2002; Gustafsson & al., 2002). Thus, the
current classification of flowering plants (APG III, 2009) splits
Clusiaceae into three families, one of which (Hypericaceae)
matches the former Hypericoideae.
Keller (1895, 1925) first attempted a comprehensive classi-
fication of the genus, followed by Kimura (1951). Both classifi-
cations were, however, unsatisfactory in several ways (Robson,
1977, 2003). Robson (1977) provided a revision of the genus,
and proposed a new classification, defining 30 sections. This
publication was the first in a series of monographs of subgroups
of Hypericum in which detailed information on characters for
species descriptions are given (Robson, 1981), as well as the
formal taxonomy of sections and species (Robson, 1985, 1987,
1990, 1996, 2001, 2002, 2006, 2010a,b). Thirty-six sections
have been to date described and 472 species have been rec-
ognized (Table 1). Thus, the genus is one of the few big plant
genera where alpha taxonomy will soon be complete.
The work of Robson (1977 onwards) attempts to arrive at “a
more natural system for the genus” (Robson, 1977: 306). It has
included data from studies on morphology, distribution, floral
anatomy, and to certain extent cytology. Based on hypothesized
evolutionary trends for 26 major characters (Table 2), an evolu-
tionary scenario was proposed. The resulting classification was
presented in a genealogical scheme showing suggested relation-
ships of the sections (Robson, 1977: fig. 1; 1981: fig. 2) and the
distribution of certain characters among these groups (Robson,
1981: figs. 6, 12, 16, 19, 22, 25, 27, 28, 29, 54). Based on this
genealogical network, it was hypothesized that Hypericum
could have evolved in Africa and spread to America, Asia and
Australia before break-up of Gondwana (Robson, 1977). This
vicariance hypothesis, however, is in conf lict with a probable
age of Hypericaceae of about 74 million years (Ma) according
to molecular phylogenies (Stevens, 2001; Davis & al., 2005),
as the final break-up of West Gondwana (South America and
Africa) took place in the lower Cretaceous about ≥105 Ma ago
(McLoughlin, 2001).
Cladistic analysis of morphological characters in Hypericum
(Hypericaceae)
Nicolai M. Nürk & Frank R. Blattner
Leibniz Institute of Plant Genetics and Crop Research (IPK), 06466 Gatersleben, Germany
Author for correspondence: Nicolai M. Nürk, nuerk@ipk-gatersleben.de
Abstract
Hypericum is a worldwide-distributed genus with almost 500 species, including the medically used, facultative apo-
mictic species H. perforatum. It is one of the few large plant genera for which alpha taxonomy has been completed and most
species have been described. To conduct a formal cladistic analysis of the genus, we coded 89 mor phological characters for
all described taxa and analyzed the data for the species using parsimony and Bayesian methods. The obtained trees indicate
that Hypericum is monophyletic if the monotypic genus Santomasia is included, and that Lianthus is its sister. The arrange-
ment of the remaining genera of Hypericaceae included in the analysis is in congruence with molecular phylogenies. Within
Hypericum the cladistic analysis revealed a basal grade containing Mediterranean species and three big clades containing most
of the diversity of the genus. In contrast to earlier assumptions, we found no indication for an African origin of Hypericum,
but assume that the genus evolved in what today is the Mediterranean area. Our phylogenies indicate a shrubby habit to be the
ancestral state within Hypericum from which species with tree-like and herbaceous habit evolved, and that apomixis originated
at least three times independently within the genus.
Keywords
biogeography; cladistic analysis; evolution; Hypericum ; phylogeny; St. John’s wort
Supplementary Material
Figures S1A–C are available in the free Electronic Supplement to the online version of this ar ticle
(http://www.ingentaconnect.com/content/iapt/tax).
M o r p h o lo g y

1496
TAXON 59 (5) • October 2010: 1495–1507Nürk & Blattner •
Cladistic analysis of Hypericum
Hypericum perforatum L. (Common St. John’s wort) is
known as a source of hypericin, and extracts of this plant are
sold as a treatment of mild to moderate depression. A con-
siderable amount of research on the occurrence of secondary
compounds in this species in par ticular, and in members of
the genus in general has been conducted (Avato, 2005; But-
terweck & Schmidt, 2007). Over the last decade, H. perfo-
ratum has also become a subject of interest as a model plant
for apomixis (asexual seed formation) research (Matzk &
al., 2001, 2007; Barcaccia & al., 2007; Schallau & al., 2010).
Reproductive biology in Hypericum is quite diverse and at
least 16 facultative pseudogamous apomictic species occur-
ring in three different sections have been described (Matzk
& al., 2003).
Comparative analysis of reproductive modes in the ge-
nus, as well as character evolution and historical biogeogra-
phy necessitates knowledge of the phylogenetic relationships
of Hypericum. Few studies using molecular approaches have
been published so far (Crockett & al., 2004; Park & Kim,
2004; Heenan, 2008), all including relatively few species, and
few or distantly related outgroup representatives. The work
of Crockett & al. (2004) was based on nuclear rDNA spacer
(ITS) sequences from 50 species, representing eleven sections
with focus on H. sect. Ascyreia (12 species) and sect. Myrian-
dra (24 species plus one undefined sample), and used Clusia
rosea Jacq. as outgroup. Heenan (2008) used the dataset of
Park & Kim (2004) plus ITS sequences of three taxa native to
New Zealand. Park & Kim (2004) included 36 species from ten
Table .
Classification of the genus Hypericum L. (sensu Robson) in numerical order of sections, subsections and series. In the second column
the phylogenetic status of sections revealed in the cladistic analysis is marked by: mt = monotypic, mp = monophyletic, pa = paraphyletic and po
= polyphyletic, ? = no evidence for monophyly or paraphyly (i.e., all species are placed on the same polytomy together with other taxa). In case
parsimony (MP) and Bayesian (BI) analysis revealed different phylogenetic status, two notifications are given, separated by MP/BI. The amount
of species per section (bold), subsection (regular) and series (italics), general distribution and citation for the systematic treatment of the section is
given in the remaining columns.
Section
Subsection
Series
Phylo-
genetic
status
[MP/BI]
Statistic
support
[bs/pp]
No. of
species
Distribution Systematic treatment
1. Campylosporus (Spach) R. Keller mp /0.58 10 Tropical & SE Africa + adjacent isl., SW Iran Robson, 1985: 178
2. Psorophytum (Spach) Nyman mt 1Spain (Balearic Isl.) Robson, 1985: 202
3. Ascyreia Choisy po/? 43 SE Europe, W to SE Asia, S China Robson, 1985: 206; 2001: 49
4. Takasagoya (Y. Kimura) N. Robson po 5Japan (Ryuku Isl.), Taiwan, Philippines Robson, 1985: 288
5. Androsaemum (Duhamel) Gordon mp
51/0.99
4Macaronesia, W & S Europe to Iran & Yemen Robson, 1985: 297
6. Inodora Stef. mt 1NE Turkey, Georgia Robson, 1985: 314
6a. Umbraculoides N. Robson mt 1Mexico (Oaxaca) Robson, 1985: 317
7. Roscyna (Spach) R. Keller mp
57/1.00
2Central to E Asia, NE America Robson, 2001: 52
8. Bupleuroides Stef. mt 1NE Turkey, Georgia Robson, 2001: 49
9. Hypericum po 42 Europe, NW Africa, Asia, NW America;
introduced (H. perforatum) into many other
parts of the world
Robson, 2002: 66
1. Hypericum po/? 19 Robson, 2002: 66
1. Hypericum po/? 12 Robson, 2002: 66
2. Senanensia N. Robson ?/po 7Robson, 2006: 28
2. Erecta N. Robson ? 23 Robson, 2006: 42
9a. Concinna N. Robson mt 1U.S.A. (northern California) Robson, 2001: 61
9b. Graveolentia N. Robson po/? 9SE Canada, eastern U.S.A. to Guatemala Robson, 2006: 79
9c. Sampsonia N. Robson mp
57/1.00
2NE India to S Japan Robson, 2001: 63
9d. Elodeoida N. Robson pa/? 5E & SE Asia (China to Kashmir) Robson, 2001: 66
9e. Monanthema N. Robson pa/? 7E & SE Asia (China to Sri Lanka) Robson, 2001: 75
10. Olympia (Spach) Nyman mp /0.71 4S Balkan peninsula, W Turkey, Aegean Isl. Robson, 2010a: 18
11. Campylopus Boiss. mt 1S Bulgaria, NE Greece, NW Turkey Robson, 2010a: 30
12. Origanifolia Stef. mp /1.00 13 Turkey, Georgia, Syria Robson, 2010a: 34
13. Drosocarpium Spach ?/mp /0.69 11 Madeira, Mediterranean to W Caucasus Robson, 2010a: 54
14. Oligostema (Boiss.) Stef. po 6Europe, Macaronesia, Mediterranean Robson, 2010a: 90
15. Thasia Boiss. mt 1Greece, Bulgaria, Turkey Robson, 2010a: 109
16. Crossophyllum Spach ?/mp /0.99 3N Aegean region, Turkey, Caucasus Robson, 2010a: 109

1497
Nürk & Blattner •
Cladistic analysis of Hypericum
TAXON 59 (5) • October 2010: 1495–1507
17. Hirtella Stef. po/? 30 W Mediterranean & S Europe to Altai Robson, 2010b: 135
1. Stenadenum N. Robson po 12 Robson, 2010b: 139
2. Platyadenum N. Robson po 18 Robson, 2010b: 162
1. Lydia Sennikov ?5Robson, 2010b: 162
2. Scabra N. Robson po 3Robson, 2010b: 170
3. Abbreviata Sennikov po 10 Robson, 2010b: 175
18. Taeniocarpium Jaub. & Spach po/pa 28 Europe, Mediterranean to Iran & Mongolia Robson, 2010b: 193
19. Coridium Spach mp /1.00 6Mediterranean, Alps, Caucasus Robson, 2010b: 239
20. Myriandra (Spach) R. Keller mp
56/1.00
29 E & central North America to Honduras,
Bermuda & Caribbean Isl.; introduced (?)
into the Azores
Robson, 1996: 92
1. Centrosperma R. Keller po/? 14 Robson, 1996: 94
2. Pseudobrathydium R. Keller mt 1 Robson, 1996: 112
3. Suturosperma R. Keller po 7 Robson, 1996: 113
4. Brathydium (Spach) R. Keller pa 2 Robson, 1996: 122
5. Ascyrum (L.) N. Robson mp 5 Robson, 1996: 124
21. Webbia (Spach) R. Keller mt 1Canary Isl., Madeira Robson, 1996: 133
22. Arthrophyllum Jaub. & Spach po/mp 5S Turkey, Syria, Lebanon Robson, 1996: 137
23. Triadenioides Jaub. & Spach pa/? 5S Turkey, Syria, Lebanon, Socotra [Yemen Isl.] Robson, 1996: 141
24. Heterophylla N. Robson mt 1Turkey (NW & W-central Anatolia) Robson, 1996: 146
25. Adenotrias (Jaub. & Spach) R. Keller mp
96/1.00
3S Morocco to Mediterranean Robson, 1996: 147
26. Humifusoideum R. Keller po/? 12 Tropical & S Africa, Madagascar, SE to E Asia Robson, 1996: 153
27. Adenosepalum Spach po/? 25 Canary Isl., Madeira, Europe, Africa, SW Asia Robson, 1996: 170
1. Aethiopica N. Robson po/? 7 Robson, 1996: 172
2. Pubescentes N. Robson po/? 6 Robson, 1996: 181
3. Caprifolia N. Robson po/? 3 Robson, 1996: 189
4. Adenosepalum po/? 9 Robson, 1996: 193
28. Elodes (Adans.) W. Koch mt 1Azores & W Europe Robson, 1996: 208
29. Brathys (Mutis ex L. f.) Choisy po/? 87 Central & South America, Caribbean Isl.,
SE Canada & eastern U.S.A. (south to Florida)
Robson, 1987: 12; 1990: 12
1. Styphelioides N. Robson mp
74/1.00
2 Robson, 1990: 16
2. Phellotes N. Robson po 32 Robson, 1990: 16
3. Brathys po 39 Robson, 1990: 27
4. Spachium R. Keller po/? 14 Robson, 1990: 29
30. Trigynobrathys (Y. Kimura) N. Robson po/? 52 South America to S Canada, E to SE Asia,
the Hawaiian Isl., Australia, New Zealand,
Africa; introduced into Europe
Robson, 1990: 47
1. Connatum (R. Keller) N. Robson po 27 Robson, 1990: 51
2. Knifa (Adans.) N. Robson po 25 Robson, 1990: 95
Four hundred and fifty-seven species in thirty-six sections have been described in the monograph (Robson, 2007 onwards), plus one unnamed
species (17. H. sp.) from section Brathys, which has not been included in the analyses due to missing character descriptions (Robson, 1990: 22).
Furthermore, 14 species have been described additionally by several authors: H. dogobadanicum Assadi (H. sect. Campylosporus, Iran), Iran. J.
Bot. 2: 89 (1984); H. fosteri N. Robson (H. sect. Ascyreia, China), Acta Phytotax. Sin. 43: 271 (2005); H. wardianum N. Robson (H. sect. Ascyreia,
China), Acta Phytotax. Sin. 43: 273 (2005); H. enshiense L.H. Wu & F.S. Wang (H. sect. Hypericum, China), Acta Phytotax. Sin. 42: 76 (2004); H.
chejuense S.-J. Park & K.-J. Kim (H. sect. Hypericum subsect. Erecta, Korea), Novon 15: 258 (2005); H. jeongjocksanense S.-J. Park & K.-J. Kim
(H. sect. Hypericum subsect. Erecta, Korea), Novon 15: 260 (2005); H. hubeiense L.H. Wu & D.P. Yang (H. sect. Elodeoida, China), Acta Phytotax.
Sin. 42: 74 (2004); H. austroyunnanicum L.H. Wu & D.P. Yang (H. sect. Elodeoida, China), Acta Phytotax. Sin. 40: 77 (2002); H. haplophylloides
Halácsy & Bald. (section “24a.” Hyplophylloides N. Robson [in prep.: Hypericum monograph part 9], Albania), Verh. Zool.-Bot. Ges. Wien 42:
576 (1893); H. huber-morathii N. Robson (H. sect. Adenosepalum, Turkey), Notes Roy. Bot. Gard. Edinburgh 27: 197 (1967); H. minutum Davis
& Poulter (H. sect. Adenosepalum, Turkey), Notes Roy. Bot. Gard. Edinburgh 21: 182 (1954); H. formosissimum Takht. (H. sect. Adenosepalum,
Turkey, Armenia, Iran), Zametki Sist. Geogr. Rast. 9 (1940). These names are accepted by the authors, but not included in the monograph yet, nor in
the analysis.
Table .
Continued.
Section
Subsection
Series
Phylo-
genetic
status
[MP/BI]
Statistic
support
[bs/pp]
No. of
species
Distribution Systematic treatment

1498
TAXON 59 (5) • October 2010: 1495–1507Nürk & Blattner •
Cladistic analysis of Hypericum
Table .
Characters used for classifying the genus Hypericum listing the hypothesized character evolution (Robson, 1977) vs. direction of char-
acter evolution revealed in the cladistic analysis. Characters states in italic font highlight evolutionary directions being incongruent in the two
columns.
Evolutionary direction used for classification
(Robson 1977: 306 ff)
Evolutionary direction revealed in the cladistic analysis
Plesiomorphic
character state
Apomorphic
character state
Plesiomorphic
character state
Apomorphic
character state
Habit trees → shrubs → peren. → annuals shrubs
→ herbs
→ trees
Indumentum absent → present absent → present
Glands
pale → dark pale → dark
pale channels → pale dots pale channels → pale dots
dark dots → dark streaks or lines dark dots → dark streaks or lines
→ increase of dark secretory tissue ?
Stem 4-lined → 2-lined → terete terete → 2-lined → 4-lined
Leaves
sessile
→ shortly petiolate
sessile
→ shortly petiolate
→ amplexicaul → perfoliate → amplexicaul or perfoliate
deciduous → persistent deciduous → persistent → deciduous (?)
opposite → 3-whorled → 4-whorled opposite → 3-whorled or 4-whorled
parallel venation → reticulate venation pinnate → parallelodromus
Perianth 5-merous → 4-merous 5-merous → 3- or 4-merous
Sepals
persistent → deciduous persistent → deciduous
unequal → equal equal → unequal
free → united free → united
margin entire → dentate → ciliate → fimbriate margin entire → margin not entire
Petals
persistent → deciduous persistent → deciduous
asymmetric → symmetric asymmetric → symmetric
Stamen fascicles
persistent → deciduous persistent → deciduous → persistent
5 → 4
2+2+1
→ 5 free fascicles → broad ring (?)
free → variously united → narrow ring → reduction up to 5 single stamina
Placentation loosely axile
→ definitely axile
axile → loosely axile → parietal
→ parietal
Ovules per placenta ∞ → 2 → (? 1) ∞ → 2
Seeds narrowly winged → carinate → cylindrical cylindrical → carinate, or winged
Basic chromosome
numbers 12 → 7 (? 6) 9 or 10 → 8
→ 14 → 14

1499
Nürk & Blattner •
Cladistic analysis of Hypericum
TAXON 59 (5) • October 2010: 1495–1507
sections, with focus on species from Korea and Japan, and used
the two Thornea species occurring in Central America as out-
group (T. matudae (Lundell) Breedlove & E.M. McClint. and
T. calcicola (Standl. & Steyerm.) Breedlove & E.M. McClint.).
Phylogenetic trees from these studies are not comparable due
to different species sampling. Furthermore, the small and un-
representative number of taxa included in these analyses does
not allow one to infer the direction of character evolution, or to
reconstr uct the historical biogeography of the genus.
As a first step to understand the evolutionary history of
the genus, we generated a phylogenetic hypothesis of Hyperi-
cum based on morphological characters analyzed with cladistic
and Bayesian methods. We coded characters used in species
descriptions for these analyses. In Hypericum, a complete revi-
sion done by one person is available, which makes description
of characters largely comparable. We compared the obtained
phylogenetic trees against the current infrageneric classifica-
tion of Hypericum, and generated hypotheses for character
evolution, the origin of apomixis and the historical biogegraphy
to be tested in future analyses with molecular methods.
MATERIALS AND METHODS
A dataset was assembled for all 591 taxa of Hypericum
described in the monograph (Robson, 1977 onwards), includ-
ing 457 species, 70 subspecies, 13 varieties, 11 formae and 40
hybrids (see Table 1, and quotations within). For the cladistic
analysis, only the 457 species were used. Hybrids were ex-
cluded because cladistic methods produce only divergently
branching phylogenetic trees and thus cannot represent the
reticulate structures if an analysis includes hybrids (McDade,
1990), or may cause major topological changes if hybrids be-
tween distantly related parents are included (McDade, 1992).
Taxa below species level were excluded to keep the number of
accessions to an amount that allowed analysis to be run in a
reasonable time. Nine outgroup taxa were included (Santoma-
sia steyermarkii (Standley) N. Robson, Lianthus ellipticifolius
(H.L. Li) N. Robson, Triadenum japonicum (Blume) Makino,
T. breviflorum (Wallich ex Dyer) Y. Kimura, Thornea matudae
(Lundell) Breedlove & E.M. McClint., Vismia cayennensis
(Jacq.) Pers., Harungana madagascariensis Poir., Cratoxy-
lum arborescens Blume and C. celebicum Blume) represent-
ing eight of the nine genera of Hypericaceae (missing: Eliea
Cambess.), from all three accepted tribes (Hypericeae Choisy,
Vismieae Choisy, Cratoxyleae Benth.) according to Stevens
(2007).
Character coding. —
Eighty-nine characters consistently
used in species descriptions in the Hypericum monograph
(Robson, 1977 onwards) and in descriptions of other genera of
Hypericaceae (Stevens, 2007; Li & Robson, 2007) were cho-
sen (see Appendix). We concentrated mainly on characters
that were defined as discrete by Robson. Only three numeric
characters were included (characters 1, 65 and 74), which were
arbitrarily but consistently coded. The 86 remaining characters
are discrete (i.e., consistently used by Robson in the species
descriptions, see Discussion). Of those, 43 are binary (absent or
present), and 46 are multistate. The large number of multistate
characters is due to the different character states in the spe-
cies diagnoses used to describe these characters. Polymorphic
characters were coded as ambiguities (i.e., allowing variable
taxa to have multiple character states). This method is least
bias-causing according to Kornet & Turner (1999), and to be
preferred if ancestral states are unknown. Characters described
as multistate, but in practice non-additive (e.g., pollen grain
types) were coded as uncertain (character numbers 63, 70 and
89). The data matrix cells scored as missing data comprised
5.7% of the entire matrix.
Phylogenetic inference. —
Phylogenetic analyses were
performed to test the monophyly of Hypericum and of the sec-
tions within the genus, as well as to establish a hypotheses
of sectional relationships. All analyses were performed on a
dataset containing 466 species (457 Hypericum and 9 outgroup
species). Two Cratoxylum species were defined as outgroup in
all analyses following Wurdack & Davis (2009), who showed
this genus to be sister to other Hypericaceae. All character
states were treated as unordered. Parsimony analyses were per-
formed in PAUP* v.4.0b10 (Swofford, 2002), Bayesian analysis
with MrBayes v.3.1.2p (Ronquist & Huelsenbeck, 2003; Al-
tekar & al., 2004). Analyses were run on a Linux cluster. The
parsimony analyses (MP) followed a two-step heuristic search
approach modified from Blattner (2004), with equal weights for
all characters, and multistate character interpretation varying
depending on whether a state is “uncertain” or “polymorphic”.
In an initial MP analysis (1st run), starting trees were obtained
via 50,000 stepwise and random taxon additions, with only five
trees held at each step, using tree bisection-reconnection (TBR)
for branch-swapping, swapping on best trees only, and saving
only one optimal tree from each repetition, even if it was not
optimal overall. The 50,000 saved trees obtained were after-
wards ordered according to tree-length (scores), and the trees
with the ten lowest scores (most parsimonious trees; normally
more than one tree was found for each of those lowest scores)
were used as starting trees for ten separate second analyses
(2nd run). In the 2nd run only best trees were saved in the
TBR search, which was limited to finding 100,000 trees. The
strict consensus tree (see Fig. 1) was calculated from that run
revealing trees with the lowest scores. Statistical support of
the branches was tested with 100,000 bootstrap re-samples
(Felsenstein, 1985), using the ‘fast and stepwise’ procedure
of PAUP*. Tree lengths, consistency (CI) and retention index
(RI) were calculated in PAUP* v.4.0b10 on a Macintosh OS 9
computer. For Bayesian inference (BI), two runs were done
with eight chains each for 5 × 107 generations under the Mk
model for morphological data (Markov k model; Lewis, 2001),
using 0.05 as temperature, and sampling a tree every 1000
generations. The initial 35,000 trees per chain were discarded
as burn-in, and posterior probabilities were calculated from the
remaining 30,002 trees. Character state changes were analyzed
using the parsimony criterion in the program McClade v.4.06
(Maddison & Maddison, 2003). Visualization of results was
done using FigTree v.1.2.3 (Rambaut, 2006–2009). The data
matrix and the MP consensus tree (Fig. S1) can be obtained
from the corresponding author.

1500
TAXON 59 (5) • October 2010: 1495–1507Nürk & Blattner •
Cladistic analysis of Hypericum
Fig. .
Scheme of the strict consensus of most parsimonious trees, showing phylogenetic relationships of 457 Hypericum species and 9 outgroup
representatives (with one species from Lianthus, Santomasia, Thornea, Vismia, and Harungana, and two from Triadenum and Cratoxylum). For
relationships among outgroups see Fig. 2. Square brackets and section names mark the position of sections in the t ree. Section names in gray
highlight the sections belonging to the “Mediter ranean grade”. Section names in quotation marks are polyphyletic in the tree. Small numbers
to the left of section names and numbers mark the position of parts of poly phyletic sections. Symbols depict the general distribution of growth
habit, and the occurrence of apomixis withi n Hypericum.

1501
Nürk & Blattner •
Cladistic analysis of Hypericum
TAXON 59 (5) • October 2010: 1495–1507
RESULTS
The dataset of 466 species (457 Hypericum and 9 out-
group species) revealed that all 89 characters were variable
and parsimony informative. The two runs of the BI analysis
did not converge completely during one month of calculation.
Therefore, only the last 30% of trees, where chains closed
in, were used to calculate the posterior probabilities. The BI
analysis resulted in a phylogenetic tree (not shown) with com-
pletely resolved placement of outgroups (see Fig. 2), but with
a large polytomy within Hypericum. Several well-supported
clades were placed along this backbone polytomy (see Table 1).
These clades were also found in the consensus tree of the
MP analyses, where resolution within Hypericum is generally
higher. For the MP analyses, the consensus trees produced
from the 100,000 trees of each of the ten analyses were all
inspected, and the 100,000 trees with shortest lengths were
chosen to calculate the consensus tree shown in Fig. 1 and
Fig. S1. The most parsimonious trees had a length of 1677
steps (CI 0.1094, RI 0.7958). Eight of the ten MP analyses
placed Lianthus N. Robson as sister to Hypericum, although
with <50% bootstrap support. Santomasia N. Robson groups
in all analyses within Hypericum (close to or within H. sect.
Ascyreia). This non-monophyly of Hypericum is also seen in
the results of the BI analysis, where Santomasia is also nested
within H. sect. Ascyreia.
Fig. .
Scheme of a par t of the strict consensus of MP trees, showing phylogenetic relationships of Hypericum with regard to the other genera of
Hyper icaceae included in the study. Dist ribution of morphological characters support ing the branches is marked on the tree by numbered black
rectangles, and the cor responding character states (apomorphies) are given in the text below. Numbers above the bra nches indicate bootst rap
values (bs in%) and posterior probabilities (pp) suppor ting the branches [bs|pp]. Rooting of the tree is that in Wurd ack & Davis (2009).

1502
TAXON 59 (5) • October 2010: 1495–1507Nürk & Blattner •
Cladistic analysis of Hypericum
In both analyses (MP and BI), relationships between the
remaining outgroups are completely resolved and statistically
supported (Fig. 2). The placement of Lianthus as sister to Hy-
pericum had little support (<50% bs, 0.94 pp), as well as Hy-
pericum itself (<50% bs, but 0.97 pp).
Within Hypericum four major groups can be recognized
that were present in all MP analyses but with <50% bootstrap
support (Fig. 1):
(1) A “Mediterranean grade” containing the monotypic
sections Inodora, Umbraculoides, Webbia and probably Het-
erophylla, as well as sections Arthrophyllum, Triadenioides
and Adenotrias. These seven sections had no fixed positions
(in the different MP analyses), but were always placed on initial
splits in the genus or as sister to one or several of the big clades.
(2) A clade comprising mainly Indo-Malayan species from
sections Ascyreia and Takasagoya together with Afrotropical
species from section Campylosporus, and the Mediterranean
species from section Psorophytum in a clade (here named “As-
cyreia s.l.”). The monophyletic, Nearctic section Myriandra,
together with the Palaearctic section Androsaemum is sister
to “Ascyreia s.l.”, and the Holarctic section Roscyna is placed
as sister to all of them. This clade is named the “Myriandra-
Ascyreia s.l.” group.
(3) A clade comprising the Neotropic section Brathys and
the mainly Neotropic section Trigynobrathys, named the “Bra-
thys s.l.” group.
(4) A mainly Palaearctic clade including section Hyperi-
cum and sections 8–19 (section numbers refer to Table 1), a
part of section Humifusoideum and sections Adenosepalum
and Elodes. This group is named “Euhypericum”. In the most
parsimonious tree (Fig. 1) an Indo-Malayan/Australasian part
of section Humifusoideum is placed as sister to all other species
of the genus. In other, less parsimonious trees, this part of sec-
tion Humifusoideum is included in the “Euhypericum” group.
Due to the large polytomy within Hypericum, there is less
resolution among the major clades in the BI tree. Twelve of the
thirty-six constituent sections within the polytomy, however,
are monophyletic and statistically supported (Table 1; Fig. S1).
Both the MP and BI analyses identified the same sections
as monophyletic (Table 1). Only H. sect. Crossophyllum and
sect. Arthrophyllum were unclear of polyphyletic in MP, and
monophyletic in BI. That is, 69% (MP) and 91% (BI) of the
sections were monophyletic or paraphyletic, and can so be said
to agree with their recognition by Robson, who expected and
accepted paraphyletic taxa (Robson, 1981: 66: “The published
sectional classification […] in Fig. 2, shows examples of sec-
tions with multiple derivates and more than one hidden example
of paraphyly […]”).
A comparison between the presumed direction of evolution
of characters (Robson, 1977) and the evolutionary direction
derived from the cladistic analyses revealed good agreement
(Table 2). However, several of the character states described
by Robson (1977) as being ‘primitive’ (i.e., plesiomorphic) ap-
pear apomorphic or ambiguous in the phylogenetic trees. As
an example, within Hypericum evolution from trees to shrubs
to herbs was postulated (Robson, 1977: 306). However, from
the phylogenetic tree (MP) the shrubby habit is plesiomorphic
within Hypericum, and real trees occur only in H. sect. Cam-
pylosporus that is nested within the “Ascyreia s.l.” group.
According to the results of our analysis, we hypothesize that
perennial herbs evolved from the shrubby habit at least four
times independently within the three major clades.
Apomixis was expected to have evolved several times in-
dependently, as apomictic species have been recognized in
three sections (H. sects. Ascyreia, Hypericum, Hirtella). In
both the MP and the BI tree, apomictic species occur in three
different clades.
DISCUSSION
Character coding. —
The coding of characters followed
Robson (1981). Several character states are defined in this pub-
lication as semantic discontinuities and are consistently used in
species descriptions (Robson, 1985 onwards). We studied these
descriptions and extracted the character states given within to
describe a species, using the code as defined in the Appendix.
Therefore, several de facto continuous characters were treated
as discrete characters (19 or even more of the 86 ‘discrete’
characters) – because they were already ‘coded’ by Robson (by
giving them a certain term) – while examining the specimens.
We excluded quantitative numeric characters (except for
three characters, see Materials and Methods) from the cladistic
analyses. The main problem was the lack of these measure-
ments in the newly published parts of the Hypericum mono-
graph (at the time we accumulated the data) that would lead
to a huge amount of missing character states in the dataset.
Furthermore, by excluding these characters we dismissed the
issues of (1) within-taxon variation, and (2) comparability of
numeric data between far related groups (what might be prob-
lematic in a genus containing so many species as it is in Hy-
pericum) (Fristrup, 2001).
Phylogeny of Hypericaceae. —
Phylogenetic analyses
could not confirm the monophyly of Hypericum (Fig. 1). Par-
simony (MP) and Bayesian (BI) approaches showed that the
monotypic genus Santomasia was included within Hypericum.
The grouping of the remaining genera of Hypericaceae, with
Cratoxylum as outgroup (Eliea was not included in the analy-
sis), followed by Vismia and Harungana together in a clade,
followed by Thornea and Triadenum in a grade and Lianthus
as sister to Hypericum (Fig. 2) is in accordance with the most
recent classification of Hypericaceae (Stevens, 2007), and re-
flects the grouping of genera included in the molecular analysis
of Wurdack & Davis (2009).
Santomasia appears in the phylogenetic trees (MP and
BI) always in close relationship to H. sect. Ascyreia. The main
reason to exclude Santomasia from Hypericum has been the
occurrence of fasciclodes (vestigial fascicles) between the
five free stamen fascicles (Robson, 1981). The absence of
such fasciclodes is certainly an apomorphy for Hypericum,
as five fasciclodes are found in Vismia and Harungana and
three fasciclodes in Lianthus, Triadenum, Thornea, Cratoxylum
and Eliea. In three species of Hypericum sect. Adenotrias and
in the monotypic sect. Elodes, however, three fasciclodes are

1503
Nürk & Blattner •
Cladistic analysis of Hypericum
TAXON 59 (5) • October 2010: 1495–1507
present between the stamen fascicles. These sections are sepa-
rated from the other taxa showing three fasciclodes according
to the classification of Robson (1977) and the results of this
analysis. In addition, the fasciclodes in H. sect. Adenotrias and
sect. Elodes have no connection to the vascular cylinder (stele),
whereas they do have a vascular connection in Cratoxylum.
The missing vascular connection, and mainly the position of
these taxa in the tree (Fig. 1 and Fig. S1), indicating that the
occurrence of fasciclodes within Hypericum is a case of par-
allel evolution. Similarly, Vismia and Harungana with five
fasciclodes are separated from Hypericum by clades with three
fasciclodes. Santomasia is certainly not closely related to Vis-
mia and Harungana (Robson, 1981: table 2), and the occurrence
of (five) fasciclodes in Santomasia might also have evolved in
parallel. It is not known if the fasciclodes of Santomasia are
connected to the stele or not. Santomasia and the “Ascyreia
s.l.” group within Hypericum share several other characters
such as the occurrence of f ive free stamen fascicles, loosely ax-
ile placentation, and cyathiform yellow flowers. Vegetatively,
Santomasia is most similar to H. roeperianum G.W. Schimp.
ex A. Rich. (H. sect. Campylosporus) (Norman Robson, pers.
comm.). If this resemblance indicates a close relationship, then
floral similarities (i.e., the fasciclodes) between Santomasia
and Hypericum sect. Adenotrias and sect. Elodes would have
evolved separately, as indicated by the cladistic analyses.
Phylogenetic inference within Hypericum. —
The parsi-
mony analyses revealed four groups within Hypericum. These
groups are present in all strict consensus trees of the MP analy-
ses, although with <50% bootstrap support.
(1) A “Mediterranean grade” (sections 6, 6a, and 21–25)
containing 17 in part local endemic species mainly distributed
in the Mediterranean Basin, the Canary Islands and on Socotra
(with the exception of the monotypic section 6a Umbraculoides
from Mexico). They are characterized by a deciduous shrubby
habit, the occurrence of (only) punctiform pale glands on leaves,
stamens and petals that are persistent after flowering (except
H. sect. Umbraculoides having deciduous petals), and axile
placentation. According to the results of this cladistic analysis,
all these character states are plesiomorphic within the genus.
(2) The “Myriandra-Ascyreia s.l.” group (containing sec-
tions 3 p.p. and 7 as sister to a clade containing sections 20 and
5 as sister to “Ascyreia s.l.”, containing section 1, 2, 3 p.p., and 4)
comprises 94 species, that is around 20% of the genus. This clade
is not characterized by uncontradicted apomorphies. Several
apomorphic characters support the monophyly of H. sect. Myri-
andra (e.g. sepals that are deciduous after flowering, androecial
elements arranged in a broad continuous ring, and pollen type
VII), or connect this section with the “Ascyreia s.l.” group (e.g.
deciduous petals and stamens, and loosely axile placentation).
Some species of H. sect. Myriandra, however, have developed
parietal placentation, which also occurs in H. sect. Brathys and
sect. Trigynobrathys, and some species of the “Ascyreia s.l.”
group have late deciduous (or nearly persistent) petals and sta-
mens (Norman Robson, pers. comm.). Thus, the association
between the monophyletic H. sect. Myriandra and the other
sections of this clade is uncertain. The possession of five free
fascicles is a synapomorphy for the “Ascyreia s.l.” group.
(3) The mainly Neotropic “Brathys s.l.” group (containing
section 29 p.p. in a grade, followed by section 30, including
three species from section 29), comprises 139 species, i.e. 30%
of the genus. Several characteristics separate this clade from
other members of Hypericum, such as a tendency towards mod-
ification/reduction of androecial elements from (a) an arrange-
ment in a narrow continuous ring to (b) 5 or 3 obscure fascicles
to (c) 5 free fascicles or 5 single stamens. They have parietal
placentation (that otherwise occurs only in several species of
H. sect. Myriandra, in one species of section Monanthema, and
in the monotypic H. sect. Elodes), and have pollen type VIII.
(4) The “Euhypericum” group (sections 8–19 and 26–28)
contains more than 45% of the diversity of Hypericum – 207
species belonging to 20 sections, most of which are native to
the Old World. The possession of dark glands, the dominance
of the herbaceous habit, and the arrangement of stamens in a
2 + 2 + 1 configuration (resulting in three visible stamen fas-
cicles) characterize members of this clade.
The existence of the “Mediterranean grade” and the mono-
phyly of the three big groups mentioned above must be further
confirmed by molecular data as they do not get convincing
statistical support in our analysis (although preliminary phy-
logenetic data of the nuclear rDNA ITS region support most of
these groups; Nürk & al., unpub.). The sections recognized by
Robson (2003) are either monophyletic or paraphyletic in our
analysis, but our results do not ref lect the sectional relation-
ships presented in Robson (1977, 1981, 2003). Hypericum sect.
Euhypericum Boiss. (Keller, 1925) does, however, include most
of the members of the “Euhypericum” group.
Comparing our results with previously published molecu-
lar approaches based on ITS sequences (Crockett & al., 2004)
reveals some congruency. “Euhypericum” is nearly identical
to clade A in Crockett & al. (2004), in which the monotypic
H. sect. Triadenioides is included and placed as sister to “Euhy-
pericum” taxa. Members of our “Myriandra-Ascyreia s.l.”
group are not monophyletic in Crockett & al. (2004), where
H. sect. Myriandra (named clade C) is sister to two clades
(named A and B) and where B that is mostly identical to our
“Ascyreia s.l.”. In their analysis, however, no taxa from H. sect.
Brathys or sect. Trigynobrathys were included and, therefore,
putative relationships cannot be clarified between sect. Myri-
andra and “Brathys s.l.” and sect. Myriandra and “Ascyreia
s.l.”, respectively. Some species of the “Brathys s.l.” group were
included in Park & Kim (2004). Their analysis of ITS sequences
focused mainly on East Asian species. Due to sparse sampling,
the MP tree presented in Park & Kim (2004) is neither really
comparable nor congruent with Crockett & al. (2004), nor the
results presented in this paper.
Character evolution. —
The absence of dark glands is
characteristic for most members of the “Mediterranean grade”,
the entire “Brathys s.l.” group and “Myriandra-Ascyreia s.l.”,
and the presence of dark glands is characteristic for “Euhy-
pericum”. Dark glands do occur, however, in some species
of H. sect. Campylosporus (nested within the “Ascyreia s.l.”
group; see Fig. S1). Some other morphological characters are
unique to this section, e.g. they have a tree-like habit or grow
as real trees (H. bequaertii De Wild.). Hypericum bequaertii

1504
TAXON 59 (5) • October 2010: 1495–1507Nürk & Blattner •
Cladistic analysis of Hypericum
was assumed to be the “most primitive” species of Hyperi-
cum (Robson, 1981: 73; 1985: 164), showing character states
that were assumed to be plesiomorphic, such as the possession
of only pale glands (Robson, 1985: 182). In this phylogenetic
analysis, however, several of these character states appear to
be apomorphic or, at least, homoplastic (Table 2). Dark glands
are present in Vismia, Harungana and Cratoxylum, but ab-
sent in Lianthus, Triadenum and Thornea. Thus, presence of
only pale glands (and absence of dark glands) seems to be a
plesiomorphic character for Hypericum and the development
of dark glands has apparently evolved in parallel in Vismia,
Harungana, Cratoxylum and several times independently in
Hypericum. Thus, the absence of dark glands does not neces-
sarily indicate that H. bequaertii is particularly “primitive”
within the genus (i.e. has only/mostly plesiomorphic character
states). In addition, the large flower of H. bequaertii, which is
described as cyathiform or campanulate and has extensively
fused stamen filaments, “could be either primitive or [dis-
play] specialisations associated with high-altitude conditions”
(Robson, 1985: 182). The comparison with outgroup species
having campanulate flowers and connate filaments, like Tri-
adenum with small campanulate flowers, indicates that these
characters evolved independently in both genera.
The evolution of the arrangement of stamens in fascicles
(Robson, 1981: f ig. 20) was hypothesized to have taken place
from five free fascicles towards various aggregations and re-
ductions (e.g. towards a 2 + 2 + 1 arrangement, or the reduction
towards five single stamens). In our analysis, however, the com-
parison with outgroup species indicates the 2 + 2 + 1 arrange-
ment to be the plesiomorphic character state within Hypericum.
The present reconstruction of the evolution of habit (Table
2) differs from previous hypotheses (Robson, 1977). A shrub
habit appears in this analysis to be plesiomorphic within Hy-
pericum, and plants with a tree-like or herbaceous habit evolved
several times from shrubby ancestors. The tree-like habit seems
to dominate in the tropics, as also shown in other plant families
(e.g., Blattner & Kadereit, 1995). Annuals are postulated to
have evolved from perennials in the South American “Brathys
s.l.” group.
Biogeography. —
The position of African Campylosporus
in the MP tree (Fig. 1 and Fig. S1), embedded in a clade contain-
ing mainly Palaearctic or Indo-Malayan species, contradicts the
hypotheses that it constitutes the most early diverging group
of the genus. In the MP analyses the Mediterranean sections
(H. sect. Androsaemum [Macronesia, Mediterranean and one
species in western Europe as far north as Scotland], sect. Ino-
dora [NE Turkey and Georgia], sect. Webbia [Canary Islands],
sect. Arthrophyllum [Turkey, Lebanon], sect. Triadenioides
[Socotra to the Levant], sect. Heterophylla [Turkey], and sect.
Adenotrias [circum-Mediterranean]), some of which are local
endemics, are always separated by initial splits from the re-
minder of the genus, and are placed within Hypericum as sister
to one (or more than one) of the three big groups. The sister of
Hypericum, Lianthus ellipticifolius, occurs in Yunnan, China.
These phylogenetic relationships indicate a geographical origin
of Hypericum in the area of the Mediterranean basin (and/or
eastwards thereof), perhaps as part of the late Tethys Ocean.
The present distribution pattern of several species of Hy-
pericum is puzzling, and requires further investigation us-
ing molecular tools. In detail, H. umbraculoides N. Robson
(H. sect. Umbraculoides) from Oaxaca, Mexico is placed in the
“Mediterranean grade” and is not related to other American
taxa in our analysis, nor in the scheme of sectional interrela-
tionships given in Robson (2003). Five species from H. sect.
Trigynobrathys (H. lalandii Choisy, H. globuliferum R. Keller,
H. humbertii Staner, H. scioanum Chiov. and H. oligandrum
Milne-Redh.) that are nested within the mainly Neotropic
“Brathys s.l.” group are native to Africa, and H. japonicum
Thunb. ex Murray (H. sect. Trigynobrathys) occurs in Asia
and Australasia. Section Roscyna (nested within the “Myri-
andra-Ascyreia s.l.” group) comprises two mainly Palaearctic
distributed species, one subspecies of which (H. ascyron L.
subsp. pyramidatum (Ait.) N. Robson) is native to Canada and
Eastern U.S.A. In H. sect. Humifusoideum, nine species are
native to Southeastern Asia, but three species occur in Af-
rica. No evidence for the monophyly of this section, however,
is revealed by our analysis. Hypericum sect. Graveolentia,
comprising nine species native to North or Central America
is placed within the “Euhypericum” group, which comprises
mostly Palaearctic taxa. Finally, H. scouleri Hook. from H. ser.
Hypericum occurs in western North America extending south
to Mexico and is placed in our analysis in a clade containing
most of the species of this series, which all occur in the Pa-
laearctic (Fig. S1).
During most of the Tertiary, the landmasses recognized
today as the African and South American continents were
much closer than at present and this might have facilitated
interchange between them (McLoughlin, 2001). If an origin
subsequent to the break-up of Gondwana were assumed for
Hypericum, several long-distance dispersal (LDD) events must
be invoked in order to explain the present distribution pattern.
Growing evidence exists, however, (e.g. for Malpighiaceae,
Davis & al., 2002; Hordeum [Poaceae], Blattner, 2006) that
relatively recent long-distance dispersals rather than old vicari-
ance of western Gondwana biotas might play a major role in the
formation of biogeographic disjunctions. The tiny seeds with a
sculptured testa, which are typical for all Hypericum species,
might easily be attached to dispersal vectors as migrating birds
(Robson, 1981).
CONCLUSIONS
The present analysis yielded several testable hypotheses re-
garding relationships between Hypericum and the other genera
of Hypericaceae, sectional relationships within the genus, and
character evolution and biogeography of Hypericum. (1) Hy-
pericum is monophyletic, if the monotypic genus Santomasia
is included. (2) Either Lianthus or a clade containing Triade-
num and Lianthus is sister to Hypericum. (3) Phylogenetically,
Hypericum is made up of a Mediterranean grade and three big
groups, although relationships among these groups are not yet
clearly resolved and these groups had <50% bootstrap sup-
port. Of special interest is the unclear affiliation of taxa of

1505
Nürk & Blattner •
Cladistic analysis of Hypericum
TAXON 59 (5) • October 2010: 1495–1507
Altekar, G., Dwarkadas, S., Huelsenbeck, J.P. & Ronquist, F. 2004.
Parallel Metropolis-coupled Markov chain Monte Carlo for Bayes-
ian phylogenetic inference. Bioinformatics 20: 407–415.
AGP (Angiosperm Phylogeny Group). 2009. An update of the Angio-
sperm Phylogeny Group classification for the orders and families
of flowering plants: APG III. Bot. J. Linn. Soc. 161: 105–121.
Avato, P. 2005. A survey on the Hypericum genus: Secondary metabo-
lites and bioactivity. Stud. Nat. Prod. Chem. 30: 603–634.
Barcaccia, G., Bäumlein, H. & Sharbel, T.F. 2007. Apomixis in St
John’s wort (Hypericum perforatum L.): An overview and glimpse
towards the future. Pp. 259 –280 in: Hörandl, E., Grossniklaus, U.,
van Dijk, P.J. & Sharbel, T.F. (eds.), Apomixis: Evolution, mecha-
nisms and perspectives. Ruggell: Gantner.
H. sect. Myriandra with the “Ascyreia s.l.” and the “Brathys
s.l.” groups, respectively. (4) Character state evolution is often
identical to evolutionary trends postulated earlier by Robson
(1977), but exceptions exist. For example, habit seems to have
changed several times from shrubs to herbs, and once from
shrubs to trees, within H. sect. Campylosporus alone. Thus,
shrubs and not trees are probably the ancestral state in Hyperi-
cum. (5) The historical biogeography of the genus has to be
newly re-evaluated, as the postulated center of origin in Africa
as part of the Gondwana continent does not fit the proposed
age of the family (~74 Ma) according to molecular data, nor the
time-frame for the break-up of Gondwana (>105 Ma ago). The
position of H. sect. Campylosporus in the “Ascyreia s.l.” group
contradicts the hypothesis that the species of this section are
sister to the reminder of the genus. Thus, our results imply a
non-African origin for Hypericum (probably in the area of to-
day’s Mediterranean region). (6) Intercontinental long-distance
dispersals may have occurred frequent within the genus, and
would provide explanations for the distribution of particular
taxa, namely in H. sects. Umbraculoides, Humifusoideum and
Trigynobrathys they provide explanations for taxon distribu-
tion. (7) Immigration into montane habitats in the tropics re-
sults in the evolution of trees in Hypericum.
Open questions for which we cannot formulate hypoth-
eses derived from our dataset are the ages of the genus and
its infrageneric entities. To answer these questions and to test
the hypotheses formulated above, future analyses employing
molecular markers are necessar y.
ACKNOWLEDGEMENTS
Special thanks are due to Norman Robson for his great help,
fruitf ul discu ssions and inspirations. We than k the herbar ia B, BM,
GAT and KYO for allowing us to use their collect ions. We thank
Mark Carine for hosting NM N during his stays at the Natural History
Museum London, Tobias Czauder na for help with the Linu x cluster,
Norman Robson, Maia Gurushid ze, Christina Baier, Peter Stevens
and Sara Crockett for carefully reading and commenting on the manu-
script. We thank the anonymous reviewers for helpful comments on the
manuscript, and the Deutsche Forschungsgemeinschaft for f inancial
support of t his work (gr ant BL 462/8).
LITERATURE CITED
Blattner, F.R. 2004. Phylogeny of Hordeum (Poaceae) as inferred by
nuclear rDNA ITS sequences. Molec. Phylog. Evol. 33: 289–299.
Blattner, F.R. 2006. Multiple intercontinental dispersals shaped the
distribution area of Hordeum (Poaceae). New Phytol. 69: 603
–614.
Blattner, F.R. & Kadereit, J.W. 1995. Three intercontinental disjunc-
tions in Papaveraceae subfamily Chelidonioideae: Evidence from
chloroplast DNA. Pl. Syst. Evol., Suppl. 9: 147–157.
Butterweck, V. & Schmidt, M. 2007. St. John’s wort: Role of active
compounds for its mechanism of action and efficacy. Wiener Med.
Wochenschr. 157: 356 –361.
Chase, M.W., Zmarzty, S., Lledó, M.D., Wurdack, K.J., Swensen,
S.M. & Fay, M.F. 2002. When in doubt, put it in Flacourtiaceae:
A molecular phylogenetic analysis based on plastid rbcL DNA
sequences. Kew Bull. 57: 141–181.
Crockett, S.L., Douglas, A.W., Scheff ler, B.E. & Khan, I.A. 2004.
Genetic profiling of Hypericum (St. John’s wort) species by nuclear
ribosomal ITS sequence analysis. Pl. Med. 70: 1–7.
Davis, C.C., Bell, C.D., Mathwes, S. & Donoghue, M.J. 2002. Laur-
asian migration explains Gondwanan disjunctions: Evidence from
Malpighiaceae. Proc. Natl. Acad. Sci. U.S.A. 99: 6833–8637.
Davis, C.C., Webb, C.O., Wurdack, K. J., Jaramillo, C.A. & Dono-
ghue, M.J. 2005. Explosive radiation of Malpighiales supports
a mid-Cretaceous origin of modern tropical rain forests. Amer.
Naturalist 165: E36–E65.
Felsenstein, J. 1985. Conf idence limits on phylogenies: An approach
using the bootstrap. Evolution 39: 783–791.
Fristrup, K.M. 2001. A history of character concepts in evolutionary
biology. Pp. 13 –35 in: Wagner, G.P. (ed.), The character concept in
evolutionary biology. San Diego, London: Academic Press.
Gustafsson, M.H.G., Bittrich, V. & Stevens, P.F. 2002. Phylogeny of
Clusiaceae based on rbcL sequences. Int. J. Pl. Sci. 163: 1045–1054.
Heenan, P.B. 2008. Three newly recognised species of Hypericum
(Clusiaceae) from New Zealand. New Zealand J. Bot. 46: 547–558.
Keller, R. 1895. Hypericum. Pp. 208–215 in: Engler, A. & Prantl, K.
(eds.), Die natürlichen Pflanzenfamilien, 3. Teil 6. Abteilung.
Leipzig: Engelmann.
Keller, R. 1925. Hypericum. Pp. 175–183 in: Engler, A. & Prantl, K.
(eds.), Die natürlichen Pflanzenfamilien, 2nd ed., vol. 21. Leipzig:
Engelmann.
Kimura, Y. 1951. Hypericaceae. In: Nakai, T. & Honda, M. (eds.), Nova
Flora Japonica, vol. 10. Tokyo: Sanseido.
Kornet, D.J. & Turner, H. 1999. Coding polymor phism for phyloge-
netic reconstruction. Syst. Biol. 48: 365–379.
Lewis, P.O. 2001. A likelihood approach to estimating phylogeny from
discrete morphological character data. Syst. Biol. 50: 913–925.
Li, X.-W. & Robson, N.K.B. 2007. Triadenum. Pp. 35–36 in: Wu, Z.Y.,
Raven, P.H. & Hong, D.Y. (eds.): Flora of China, vol 13. St. Louis:
Missour i Botanical Garden Press.
Maddison, D.R. & Maddison, W.P. 2003. McClade, version 4.06.
Sunderland, Massachusetts: Sinauer.
Matzk, F., Hammer, K. & Schubert, I. 2003. Coevolution of apomixis
and genome size within the genus Hypericum. Sexual Pl. Reprod.
16: 51–58.
Matzk, F., Meister, A., Brutovska, R. & Schubert, I. 2001. Recon-
struction of reproductive diversity in Hypericum perforatum L.
opens novel strategies to manage apomixis. Pl. J. 26: 275–282.
Matzk, F., Prodanovic, S., Czihal, A., Tiedemann, J., Arzenton,
F., Blattner, F.R., Kumlehn, J., Altschmied, L., Schuber t, I.,
Johnston, A., Grossniklaus, U. & Bäumlein, H. 2007. Genetic
control of apomixis: Preliminary lessons from Poa, Hypericum and
wheat egg cells. Pp. 159 –166 in: Hörandl, E., Grossniklaus, U., van
Dijk, P.J. & Sharbel, T.F. (eds.), Apomixis: Evolution, mechanisms
and perspectives. Ruggell: Gantner.
McDade, L. 1990. Hybrids and phylogenetic systematics I. Patterns of
character expression in hybrids and their implications for cladistic
analysis. Evolution 44: 1685–1700.

1506
TAXON 59 (5) • October 2010: 1495–1507Nürk & Blattner •
Cladistic analysis of Hypericum
McDade, L.A. 1992. Hybr ids and phylogenetic systematics II. The
impact of hybrids on cladistic analysis. Evolution 46: 1329–1346.
McLoughlin, S. 2001. The breakup history of Gondwana and its im-
pact on pre-Cenozoic f loristic provincialism. Austral. J. Bot. 49:
271–300.
Park, S.-J. & Kim, K.-J. 2004. Molecular phylogeny of the genus
Hypericum (Hypericaceae) from Korea and Japan: Evidence from
nuclear rDNA ITS sequence data. J. Pl. Biol. 47: 366–374.
Rambaut, A. 2006 –2009. FigTree, version 1.1.3. Available from http://
beast.bio.ed.ac.uk/FigTree.
Robson, N.K.B. 1977. Studies in the genus Hypericum L. (Guttiferae).
1. Infrageneric classification. Bull. Brit. Mus. (Nat. Hist.), Bot.
5: 295–355.
Robson, N.K.B. 1981. Studies in the genus Hypericum L. (Guttiferae).
2. Characters of the genus. Bull. Brit. Mus. (Nat. Hist.), Bot. 8:
55–226.
Robson, N.K.B. 1985. Studies in the genus Hypericum L. (Guttiferae).
3. Sections 1. Campylosporus to 6a. Umbraculoides. Bull. Brit.
Mus. (Nat. Hist.), Bot. 12: 163–325.
Robson, N.K.B. 1987. Studies in the genus Hypericum L. (Guttiferae).
7. Section 29. Brathys (part 1). Bull. Brit. Mus. (Nat. Hist.), Bot.
16: 1–106.
Robson, N.K.B. 1990. Studies in the genus Hypericum L. (Guttiferae).
8. Sections 29. Brathys (part 2) and 30. Trigynobrathys. Bull. Brit.
Mus. (Nat. Hist.), Bot. 20: 1–151.
Robson, N.K.B. 1996. Studies in the genus Hypericum L. (Guttiferae).
6. Sections 20. Myriandra to 28. Elodes. Bull. Brit. Mus. (Nat.
Hist.), Bot. 26: 75–217.
Robson, N.K.B. 2001. Studies in the genus Hypericum L. (Guttiferae).
4(1). Sections 7. Roscyna to 9. Hypericum sensu lato (part 1). Bull.
Brit. Mus. (Nat. Hist.), Bot. 31: 37–88.
Robson, N.K.B. 2002. Studies in the genus Hypericum L. (Guttif-
erae). 4(2). Section 9. Hypericum sensu lato (part 2): subsection
1. Hypericum series 1. Hypericum. Bull. Nat. Hist. Mus. Lond.
(Bot.) 32: 61–123.
Robson, N.K.B. 2003. Hypericum botany. Pp 1–22 in: Ernst, E. (ed.),
Hypericum: The genus Hypericum. Medicinal and Aromatic Plants
31, Industr ial Profiles. London, New York: Taylor & Francis.
Robson, N.K.B. 2006: Studies in the genus Hypericum L. (Clusiaceae).
Section 9. Hypericum sensu lato (part 3): subsection 1. Hypericum
series 2. Senanensia, subsection 2. Erecta and section 9b Graveo-
lentia. Syst. Biodivers. 4: 19–98.
Robson, N.K.B. 2010a. Studies in the genus Hypericum L. (Hyperi-
caceae). 5(1). Sections 10. Olympia to 15/16. Crossophyllum. Phy-
totaxa 4: 5–126.
Robson, N.K.B. 2010b. Studies in the genus Hypericum L. (Hyperi-
caceae). 5(2). Section 17. Hirtella to 19. Coridium. Phytotaxa 4:
127–258.
Ronquist, F. & Huelsenbeck, J.P. 2003. MrBayes 3: Bayesian phyloge-
netic inference under mixed models. Bioinformatics 19: 1572–1574.
Schallau, A., Arzenton, F., Johnston, A.J., Hähnel, U., Koszegi,
D., Blattner, F.R., Altschmied, L., Haberer, G., Barcaccia, G.
& Bäumlein, H. 2010. Identification and genetic analysis of the
APOSPORY locus in Hyperiucm perforatum L. Pl. J. 62: 773–784.
Stevens, P.F. 2001 onwards. Angiosperm phylogeny website, version
9, June 2008 [and more or less continuously updated since]. http://
www.mobot.org/mobot/research/APweb/.
Stevens, P.F. 2007. Hypericaceae. Pp. 194–201 in: Kubitzki, K. (ed.),
The families and genera of vascular plants, vol. 9. Berlin, Heidel-
berg: Springer.
Swofford, D.L. 2002. PAUP*: Phylogenetic analysis using parsimony
(*and other methods), version 4. Sunderla nd, Massachusetts:
Sinauer.
Wurdack, K.J. & Davis, C.C. 2009. Malpighiales phylogenetics: Gain-
ing ground on one of the most recalcitrant clades in the angiosperm
tree of life. Amer. J. Bot. 96: 1551–1570.
Appendix.
Character coding.
HABITAT: 1) Elevation: lowland (0 to ~1000 m a.s.l.) (n), montane (>1000 m a.s.l. and not <500 m a.s.l.) (y). 2) Humidity-preference: wet (plants standing in
water) (a), humid (b), dry (c). H ABIT: 3) Live for m: tree (a), shrub (b), perennial herb (c), annual herb (d). 4) Runners: absent (n), present (y). 5) Taproot: absent
(n), present (y). 6) Vegetative layer (aerial bulbils etc.): absent (n), present (y). 7) Wood parenchyma: absent (n), present (y). STEM: 8) Number of stem lines:
absent (a), two (b), three (c), four (d), six (e). 9) Stem ancipitous (i.e., complanate, t wo-edged): not ancipitous (n), ancipitous (y). 10) Stem terete (rounded at
node): not terete (n), terete (y). 11) GLANDS ON STEM – t ype: pale (a), red (b), black (c), amber (d), absent (eglandular) (e). 12) Indumentum: stem glabrous
(n), stem hair y (y). 13) Indumentum stellate (multicellular): unicellular trichomes (n), stellate t richomes (y). 14) Cortex exfoliating: persistent (cortex not ex-
foliating) (a), exfoliating in flakes (b), sheets or plates (c), strips (d), scales (e), irregularly (f). 15) Internodes: shorter than leaves (n), longer than leaves (y).
LEAVES: 16) Phyllotaxis: opposite (a), three-whorled (b), four-whorled (c). 17 ) Leaf ty pe: foliage (a), linear to "ericoid" (b), scale (<2 mm long) (c). 18) Inser-
tion: sessile (a), sub-sessile (or sub-petiolate: ≤0.5 long) (b), petiolate (>0.5 mm long) (c). 19) Adnation of opposite leaf bases: not clasping the stem (free) (a),
amplexicaul (stem-clasping) (b), perfoliate (connate pairs) (c). 20) Indumentum: leaves glabrous (n), leaves hairy (y). 21) Margin: not entire (denticulate to
ciliate) (n), entire (y). 22) Leaf persistence (deciduousness): persistent (evergreen plant) (n), deciduous (y). 23) Stomatal type: paracytic (n), anomocytic or
cyclocytic (y). 24) Leaf venation I – ty pe: parallelodromus (a), pinnate (b), one nerved (midrib only) (c). 25) Leaf venation II – adnation of pinnate veins: not
all later al veins ad nate among t hemselves (n), all lateral veins adnate among themselves (y). 26) Leaf venation III – tertiary reticulum: absent (n), present (y).
GLANDS ON LEAVES: 27) Laminar glands I – t ype: pale (a), red (b), dark (c), absent (d). 28) Lam inar glands II – shape: linear (not inter rupted) (a), streaks
and short lines (interr upted or striifor m) (b), small dots ( punctiform) (c). 29) Marginal glands – ty pe: pale (a), red (b), dark (c), absent (d). 30) Ventral glands:
absent (n), present (y). INFLOR ESCENCE: (Inf lorescences in Hyperic um are generally cymose (Robson, 1981: 83), but several modifications make it difficult
to describe inflorescences within Hypericum by single terms without loosing information. Therefore we coded inflorescences in a key-like way, following the
descriptions given in Robson (1981: 83ff).) 31) Ramif ication on terminal node: alternate (a), decussate (b), pleiochasial (c), not branched (uniflor) (d). 32) Posi-
tion of flowering branches: on terminal node only (n), flower ing branches also from lower nodes (y). 33) Ramif ication of inf lorescence branches above the
term inal node: alternate (a), decussate ( b), monochasial (c), dichasial (d), pseudo-dichotomous (e), "sympodial" (f), not bra nched (unif lor) (g). 34) Subsidiary
inf lorescence branches: not divided by only-leaf-bearing nodes from ter minal inf lorescence (n), divided by only-leaf-bearing nodes from terminal inflorescence
(y). 35) Bracts margin: entire (n), not entire (denticulate, ciliate or fimbr iate) (y). 36) Glands on bracts margin: absent (n), present (y). FLOWERS: 37) Corolla
type: stellate (or i nfundibulifor m = obconic) (a), cyathifor m (b), campanulate (c), tubular (d). 38) Ligulate outgrowth of petals: absent (a), ligula enti re (b),
ligula t rifid (c), ligula fringed (d). 39) Style: homostylous (n), heterostyous (y). 40) Merosity: pentamerous (a), tetramerous (b), trimerous (c), hexamerous (d).
SEPALS: 41) Con nation of sepals: free (a), united at base (b), ≥2/3 united (c). 42) Sepals persistence (deciduousness): persistent (n), deciduous (y). 43) Position
of sepals in fruit: erect (a), spreading (b), reflexed or recurved (c). 44) Margin of sepals: entire (n), not entire (f ringed) (y). 45) Sepal veins branched: not branched
(n), branched (y). 46) Dimorphism in sepals: equal (to subequal) (a), unequal (to subequal) (b), dimorphic ('markedly unequal') (c) [also coded: “subequal or
equal to unequal” (ab)]. GLANDS ON SEPALS: 47) Laminar gland s I – ty pe: pale (a), red (b), dark (c), absent (d). 48) Laminar glands II – shape: linear (not
inter rupted) (a), streaks and shor t lines (interrupted or striiform) (b), small dots (punctiform) (c). 49) Marginal glands I – type: pale (a), red (b), dark (c), absent
(d). 50) Ma rginal glands II – position: marginal (a), inf ramarginal (b), submarginal (c). 51) Margi nal glands: sessile (n), stipitate (raised) (y). PETALS: 52)
Color: yellow (a), pin k (b), white (c), greenish (d). 53) Corolla aestivation: imbricate (n), contorted (y). 54) Petals shape: symmetrical (n), asymmetrical (y). 55)
Petals pubescent on adaxial surface: not pubescent (glabrous) (n), pubescent (y). 56) Petals persistence (deciduousness): persistent (n), deciduous (y). 57)

1507
Nürk & Blattner •
Cladistic analysis of Hypericum
TAXON 59 (5) • October 2010: 1495–1507
Apiculus on petals: absent (a), apical (b), lateral or subapical (c). 58) Margin of petals: entire (n), not entire (fringed, ciliate or denticulate) (y). GLANDS ON
PETALS: 59) Laminar glands I – type: pale (a), red (b), dark (c), absent (d). 60) Laminar glands I I – shape: linear (not interrupted) (a), streaks a nd short lines
(interrupted or striiform) (b), small dots (punctiform) (c). 61) Marginal glands – type: pale (a), red (b), dark (c), absent (d). 62) Marginal glands raised: not r aised
(sessile) (n), raised (stipit ate) (y). ANDROECIUM: 63) Configuration (arrangement of st amens): 5 free fascicles (a), 2+1+1+1 fascicles (b), 2+2+1 fascicles (c),
2+2+1 fascicles + 3 ster ile fascicles (d), narrow continuous ring (e), broad continuous ring (f), tetramer ring by elimination (g), 5 obscure fascicles (h), 3 obscure
fascicles (i), 5 single stamens (j). 64) Stamens persistence (deciduousness): persistent (n), deciduous (y). 65) Proportion of stamen length to petal length (sta-
men/petal): 0.20–0.59 (a), 0.60–0.90 (b), 0.91–1.50(–2.00) (c). 66) Interstaminal glands: absent (n), present (y). 67) Connation of stamina: free (a), shortly united
(b), united above middle (c). 68) Gland on anthers: absent (a), amber (b), black (c). 69) Fasciclodes (vestigial fascicles): absent (a), three fasciclodes (b), f ive
fasciclodes (c). 70) Pollen grains type: I (a), II (b), III (c), IV (d), V (e), VI (f), VII (g), VIII (h), IX (i), X ( j), XI (k). GYNOECIUM: 71) Placentation: parietal
(a), loosely axile (b), axile (c). 72) Number of seeds per ovar y: few (n), many (∞) (y). 73) Number of styles: f ive (a), four (b), three (c), two (d), six (e), seven (f ),
eight (g) [f, g: only in species H. pleiostylum C. Rodr.Jim.]. 74) Proportion of style leng th to ovary length (style/ovary): 0.01–0.59 (a), 0.60–0.99 (b), 1.00–1.59
(c), 1.60–1.99 (d), 2.00–2.99 (e), 3.00–3.99 (f), 4.00 –4.99 (–7.00) (g). 75) Union of styles: free (a), partly u nited (in flower) (b), complete union (also u nited in
fruit) (c). 76) Stigma shape: (sub-)globose (a), (sub-)capitate (also ‘rounded’, ‘tr uncate’, ‘peltate’) ( b), narrow or small (at least not capitate) (c), ellipsoid (d),
cylindrical (e), clavate (f ), infu ndibuliform (g). 77) Persistence of style: breaks off in fr uit (n), persists on fruit (y). FRUIT: 78) Type of fr uit: capsule (n), berry
(y). 79) Capsule aper ture mechanism: loculicidal (n), septicidal (y). 80) Propor tion of fr uit length to sepals length: shorter than sepals (a), equaling sepals (b),
exceeding sepals (c). 81) Fruit enclosed by twisted petals: not enclosed (n), enclosed (y). 82) Surface structure of capsules (Vittae & Vesicles): not vittate
(without str ipes) (a), vertical raised vittae (b), vertical vittae with glands (c), lateral (towards the margin of the single carpel) vit tae diagonal and dorsal vitt ae
vertical (d), swollen vittae (e), pale vesicles (‘vesiculate’, bubble-shaped) (f), black vesicles (g), only 1–2 pale vesicles (h). SEEDS: 83) Seed shape: cylindrical
(a), f usiform (spindle -shaped) (b), pyriform (pear-shaped) (c), ellipsoid (d), ovoid (e), clavate (club-shaped) (f), elongate and flattened (g). 84) Seed appendages
I – laterally carinate: not carinate (n), carinate (y). 85) Seed appendages II – ter minally winged: not winged (n), winged (y). 86) Seed appendages III – with a
distal expansion: absent (n), present (y). 87) Seed appendages IV – with an elaiosome (‘caruncu late’): absent (n), present (y). 88) Testa sculptur ing: reticulate
(also ‘linear-reticulate’, ‘ir regularly reticulate’, ‘linear-foveolate’ or ‘foveolate’) (a), scalarifor m (also ‘linear-scalarifor m’ or ‘ribbed-scalarifor m’) (b), papillose
(also ‘rugulose’ or ‘smooth’) (c). CYTOLOGY: 89) Chromosome numbers (n = …): 6 (a), 7 (b), 8 (c), 9 (d), 10 (e), 12 (f ), 14 (g), 16 (h), 18 (i), 19 (j).
For Bayesian phylogenetics, letters were changed to numbers (0–9). Only character nu mber 70 (pollen g rains type) had more than ten states, therefore we
excluded the pollen type XI in the Bayesian analyses (originally described as a probable pollen type in combination with pollen type X. Character st ate de-
scribed only for H. sect. Hirtella).
Appendix.
Continued.