American Journal of Botany 92(3): 544–557. 2005.
POLYPHYLY OF MUSSAENDA INFERRED FROM ITS AND
trnT-F DATA AND ITS IMPLICATION FOR GENERIC
LIMITS IN MUSSAENDEAE (RUBIACEAE)1
GRECEBIO D. ALEJANDRO,2,3 SYLVAIN G. RAZAFIMANDIMBISON,4,5
SIGRID LIEDE-SCHUMANN2
AND
Department of Plant Systematics, Bayreuth University, Universitätstr. 30, D-95440 Bayreuth, Germany; 3Research Center for the
Natural Sciences and College of Science, University of Santo Tomas, España, Manila, 1008 Philippines; 4Department of Systematic
Botany, Evolutionary Biology Centre, Uppsala University, Norbyvägen 18 D, SE-752 36, Uppsala, Sweden; 5The Bergius
Foundation at the Royal Swedish Academy of Sciences, P.O. Box 50017, SE-104 05, Stockholm, Sweden
2
Although recognition of Mussaenda as a separate genus has been widely accepted, its generic circumscriptions have always been
controversial. In this first molecular phylogenetic study focused specifically on Mussaenda sensu lato (s.l.) and its allied genera,
parsimony analyses were based on both ITS and trnT-F sequence data to (1) test the monophyly of Mussaenda s.l. as presently
circumscribed, (2) assess the phylogenetic relationships within the tribe Mussaendeae as currently delimited, (3) evaluate the phylogenetic value of the morphological characters traditionally and/or currently used to circumscribe Mussaendeae, (4) and make inferences
on the biogeographical origin of Mussaenda. Of the 63 trnT-F and 38 ITS sequences included in our studies, 52 and 36 sequences,
respectively, are newly published here. Our results highly support the polyphyly of Mussaenda s.l. as currently delimited but further
support the monophyly of Mussaendeae sensu Bremer and Thulin. The Malagasy Mussaenda are more closely related to Landiopsis
than they are to the African and Asian Mussaenda. Pseudomussaenda and the Afro-Asian Mussaenda clade are resolved as sister
groups. Aphaenandra is nested within the Afro-Asian Mussaenda clade. As a result, we merge Aphaenandra in Mussaenda, which is
now restricted to include only the African and Asian Mussaenda representatives. We describe a new genus Bremeria to accommodate
all Indian Ocean (Madagascar and the Mascarenes) Mussaenda species and make 19 new combinations. The newly delimited Mussaenda is diagnosed by reduplicate-valvate aestivation and glabrous styles, whereas Bremeria can be distinguished from the remaining
Mussaendeae genera by having both reduplicate- and induplicate-valvate aestivation and densely pubescent styles. Our studies strongly
suggest an African origin of the newly delimited Mussaenda. Finally, descriptions of the newly circumscribed Mussaenda and Bremeria
are provided.
Key words:
biogeography; Bremeria; ITS; Mussaenda; Mussaendeae; Rubiaceae; trnT-F.
Recent phylogenetic analyses within Rubiaceae (or coffee
family) based on the rbcL sequence data conducted by Bremer
and Thulin (1998) led to the reestablishment of the tribe Mussaendeae and proposition of the new tribal circumscriptions
for the tribe Isertieae. Mussaendeae, currently belonging to the
subfamily Ixoroideae sensu lato (s.l.) (Bremer et al., 1999),
comprises seven genera (Bremer and Thulin, 1998): Aphaenandra Miq., Heinsia DC., Landiopsis Capuron ex Bosser, Mussaenda s.l. Burm. ex L., Neomussaenda Tange, Pseudomussaenda Wernham, and Schizomussaenda Li. Mussaenda s.l. is
the most species-rich genus with ca. 163 species of small trees,
scandent or scrambling shrubs or true lianas. The genus is
mostly paleotropical and has its center of diversity in tropical
Asia with ca. 100 species, followed by tropical Africa with ca.
35 species (Bridson and Verdcourt, 1988), Madagascar with
ca. 24 species (S. Andriambololonera and S. Razafimandimbison, Missouri Botanical Garden and Bergius Foundation, respectively, unpublished manuscript), and the Mascarenes with
four species (Wernham, 1914; Andriambololonera and Razafimandimbison, unpublished manuscript). Mussaenda s.l. is
characterized by a combination of valvate aestivation, fleshy
or berry-like, indehiscent fruits, and numerous, small, reticulate seeds. Many species of Mussaenda s.l. (e.g., M. erythrophylla, M. incana, M. parvifolia, M. philippica) are commonly
cultivated in botanical gardens throughout the world because
of their beautiful, long-blooming, sturdy flowers with enlarged
calyx lobes.
Although recognition of Mussaenda as a separate genus has
never been challenged, its circumscription has always been
controversial (e.g., Miquel, 1857; Hooker, 1880; Kurz, 1887;
Schumann, 1891). Earlier authors disagreed as to whether
Mussaenda should include only the Asian and African species
with enlarged, petaloid calyx lobes or calycophylls (also called
semaphylls sensu Leppik, 1977) and fleshy, indehiscent fruits.
Miquel (1857) transferred the Asian capsular-fruited Mussaenda uniflora without enlarged calyx lobes, described by G. Don
(1834), to his new genus Aphaenandra. Similarly, both Wernham (1916) and Li (1943) described Pseudomussaenda and
Schizomussaenda, respectively, to accommodate all African
and another Asian capsular-fruited Mussaenda with enlarged
Manuscript received 7 May 2004; revision accepted 23 November 2004.
The authors thank Sylvie Andriambololonera, Petra De Block, Akiyo Naiki,
Christian Puff, Elmar Robbrecht, and Piet Stoffelen, who kindly provided leaf
material for the molecular work; Birgitta Bremer and two anonymous reviewers for comments and suggestions on the manuscript; Anna Bauer, Nahid
Heidari, Andreas Jürgens, and Angelika Täuber for help with sequencing;
Simon Malcomber for help with the implementation of the SH test; Ulrich
Meve for technical assistance; the University of Santo Tomas, Manila, Philippines for financial support during the field collecting by G.D.A. in the Philippines; Domingo Madulid of the Philippine National Museum for arranging
the collecting permits for G.D.A.; MEF (Ministère des Eaux et Forêts) and
ANGAP (Association Nationale pour la Gestion des Aires Protégées) in Madagascar for issuing collecting permits to S.G.R.; Missouri Botanical Garden
in Madagascar for arranging the collecting permits for S.G.R.; and the following herbaria and their staff for providing loans and/or access to collections:
BR, L, NY, P, PNHS, TAN, TEF, UPS, US, and WAG. This study was funded
by the Swedish Research Council grant to Birgitta Bremer and the Deutscher
Akademischer Austauschdienst grant to G.D.A.
1
544
March 2005]
ALEJANDRO
ET AL.—POLYPHYLY OF
calyx lobes. On the other hand, Candolle (1830), also endorsed
by Hooker (1880), Kurz (1887), and Schumann (1891), recognized a broad circumscription including Mussaenda species
with or without semaphylls and with dehiscent or indehiscent
fruits. This broad circumscription appears never to have gained
acceptance. Mussaenda s.l. as presently circumscribed (e.g.,
Robbrecht, 1988; Puff et al., 1993; Mabberley, 1997) includes
all Afro-Asian and Malagasy Mussaenda species with indehiscent fruits and with or without semaphylls. This situation
raises questions as to whether one of these three conflicting
generic limits circumscribes a monophyletic unit. The present
study is the first phylogenetic investigation to focus specifically on Mussaenda s.l. and its alliances.
Previous phylogenetic studies of some Rubiaceae groups
have shown that both the internal transcribed spacer (ITS) region of nuclear rDNA (e.g., Andreasen et al., 1999; Persson,
2000; Razafimandimbison and Bremer, 2001) and the trnT-F
region of chloroplast DNA (e.g., Razafimandimbison and Bremer, 2002) were useful for assessing phylogenetic relationships at both generic and tribal levels. The first objective of
the present study is to reconstruct robust phylogenies for Mussaenda s.l. and its allied genera using both the ITS and trnTF sequence data. The resulting phylogenies will then be used
to: (1) test the monophyly of Mussaenda s.l., (2) assess the
phylogenetic relationships within the tribe Mussaendeae as
currently delimited, (3) evaluate the phylogenetic value of the
morphological characters traditionally and/or currently used to
circumscribe Mussaendeae, and (4) make inferences on the
biogeographical origin of Mussaenda.
MATERIALS AND METHODS
Taxon sampling—Material was available for all genera currently placed in
Mussaendeae sensu Bremer and Thulin (1998) except for Neomussaenda. A
total of 37 Mussaendeae species, representing four individuals of Aphaenandra, three Heinsia species, three Pseudomussaenda and 25 Mussaenda s.l.
species, as well as one individual each of the two monotypic genera Landiopsis and Schizomussaenda, was included in our analyses. Twelve genera
(Acranthera, Gonzalagunia, Hippotis, Hoffmania, Isertia, Mycetia, Pauridiantha, Pentagonia, Pseudosabicea, Sabicea, Schradera, and Sommera) traditionally associated with Mussaendeae and Isertieae sensu Robbrecht (1988)
were also added in the trnT-F analysis to test the monophyly of Mussaendeae
sensu Bremer and Thulin (1998). Few representatives of Cinchonoideae sensu
Robbrecht (1988), Ixoroideae, and Rubioideae were additionally investigated
(see Appendix in Supplemental Data accompanying online version of this
article). The genus Luculia, which has been shown to be basal in Rubiaceae
(Bremer et al., 1999), was used as the outgroup to root the trnT-F tree. Origins
and voucher specimens are listed in Appendix.
DNA extraction and amplification—Total DNA was extracted from fresh,
silica-gel dried leaf tissues (Chase and Hills, 1991) or herbarium material
using DNeasy Plant Mini kit (Qiagen, Hilden, Germany) and cleaned with
Qia-Quick PCR purification kit (Qiagen). For amplification and sequencing
of the trnT-F, the protocols are described in Razafimandimbison and Bremer
(2002).
The ITS region (ITS1, 5.8S gene, and ITS2) was amplified using primers
P17F (59-CTA CCG ATT GAA TGG TCC GGT GAA-39) and 26S–82R (59TCC CGG TTC GCT CGC CGT TAC TA-39) (Popp and Oxelman, 2001).
PCR cocktails were mixed as follows (25 mL): 15.3 mL dH2O, 2.5 mL 103
PCR buffer, 2.0 mL 25 mM MgCl2, 1.5 mL 2 mM dNTP, 1.0 mL of 10 mM
forward and reverse primers, respectively, 0.2 mL Taq DNA polymerase, and
1.5 mL DNA. The Q-solution (Qiagen) was also used as additive replacing
some of the water. PCR reactions were run on a Biometra UNO-Thermoblock
cycler with initial denaturation for 90 s at 978C, followed by 35 cycles of 20
MUSSAENDA (RUBIACEAE)
545
s 978C, 90 s 728C, 1 : 30 s 728C, finishing with 728C for 7 min. PCR products
were cleaned with Qia-Quick PCR purification kit (Qiagen).
Sequencing reactions were done using primers P16F (59-TCA CTG AAC
CTT ATC ATT TAG AGG-39) and P25R (59- GGG TAG TCC CGC CTG
ACC TG-39) (Popp and Oxelman, 2001) and the ABI PRISM Big Dye Terminator Cycle sequencing kit (Applied Biosystems, Bayreuth, Germany). All
sequencing was performed on an ABI Prism Model 310, version 3.0 sequencer.
Data analysis—The ITS and trnT-F sequences were assembled using the
Perkin Elmer Sequence Navigator, version 1.0.1 and Sequencher 3.1.1, respectively, and edited manually. All new sequences were submitted to EMBL,
and their accession numbers are in Appendix (see Supplemental Data for
online version of this article). We performed parsimony phylogenetic analyses
at two distinct but interrelated levels. We initially conducted a large-scale
phylogenetic analysis based on the trnT-F data, including the 36 taxa of Mussaendeae sensu Bremer and Thulin (1998), 12 genera previously placed in
Mussaendeae, 14 distantly related Rubiaceae taxa from Cinchonoideae, Ixoroideae s.l., and Rubioideae, and one outgroup taxon, for a total of 63 taxa.
The results of this analysis allowed us to select new outgroup taxa (Sabicea
diversifolia and Warszewiczia coccinea) from within Ixoroideae s.l. to root
both the ITS and combined ITS-trnT-F analyses of taxa from Mussaendeae
sensu Bremer and Thulin (1998). The parsimony analyses of the ITS, trnTF, and combined ITS-trnT-F data sets (excluding uninformative characters)
were performed with PAUP* version 4.0b (Swofford, 2000) on a Power Macintosh G3 computer using heuristic searches, with the MULTREES option on,
tree-bisection-reconnection (TBR) branch swapping, swap on best only in
effect, and 5000 random addition sequences. The heuristic search for the trnTF analysis could not be completed due to computational limitations. The trnTF data were then analyzed using the following settings: the MULTREES option off, nearest neighbor interchanges (NNI) branch swapping, and 10 000
random addition sequences. For the combined ITS-trnT-F data sets, we likewise searched for multiple islands of most-parsimonious trees (Maddison,
1991). In all analyses, characters were given equal weight, gaps were treated
as missing data, and phylogenetically informative indels were coded following
the simple gap coding method of Simmons and Ochoterena (2000). The consistency index (CI; Kluge and Farris, 1969) and retention index (RI; Farris,
1989) were calculated to estimate homoplasy. Bootstrap (BS; Felsenstein,
1985) values using 10 000 replicates, the MULTREES option off, NNI branch
swapping, and five random addition sequences were performed to assess relative support for the identified clades. Clades receiving a bootstrap support
of 50–69% were regarded as weakly supported, 70–85% as moderately supported, and 86–100% as strongly supported.
We statistically evaluated the combinability of the ITS and trnT-F data
partitions using the one-tailed Shimodaira-Hasegawa test (SH test; Shimodaira
and Hasegawa, 1999; Goldman et al., 2000) and the incongruence length
difference (ILD test; Farris et al., 1995), both implemented in PAUP*. We
perfomed maximum likelihood (ML) analyses of the ITS and trnT-F data,
respectively, using the GTR 1 G 1 I and the GTR 1 G substitution models,
which were selected by MrModeltest (Nylander, 2002) as the best models.
We subsequently conducted the SH tests, using resampling estimated by loglikelihood (RELL) optimization and 1000 bootstrap replicates, to compare
statistically the optimal ITS and trnT-F topologies, respectively, against two
alternative phylogenetic hypotheses: topology inferred from the ITS data constrained by the optimal ML topology of the trnT-F data (for the ITS matrix)
and topology from the trnT-F data constrained by the best topology from the
ITS data (for the trnT-F partition).
Incongruency test was performed using the incongruence length difference
(ILD test; Farris et al., 1995) to assess incongruencies between the ITS and
trnT-F data sets. This test uses the partition-homogeneity test as implemented
in PAUP* (Swofford, 2000). The heuristic search was set to 500 replicates
with 10 random addition sequence and NNI branch swapping. If the probability of obtaining a smaller sum of tree lengths from the randomly generated
data sets is lower (P # 0.05) than that of the original data sets, the null
hypothesis that the two data sets are homogenous is rejected and they are
interpreted as incongruent (Farris et al., 1995).
546
AMERICAN JOURNAL
Finally, additional SH tests were performed to test whether or not the optimal ML topologies of both the ITS (Fig. 1) and trnT-F (Fig. 2) trees were
significantly different from the alternative hypothesis constraining all sampled
African Mussaenda species monophyletic.
RESULTS
Sequence characteristics—The trnT-F sequences of the
sampled members of Mussaendeae varied from 1703 base
pairs (bp) (Heinsia crinita) to 1793 bp (Mussaenda latisepala).
The total GC content of the trnT-F Mussaendeae sequences
ranged from 30.82% (Pseudomussaenda flava) to 36.64%
(Mussaenda isertiana) and its average was 31.55%. The ITS
sequences of the sampled members of Mussaendeae varied
from 570 bp (all Malagasy Mussaenda included) to 596 bp
(Heinsia bussei and H. zanzibarica). The average total length
of ITS1 and ITS2 were 204 and 221 bp, respectively. From
all Mussaendeae sequences included, ITS2 (215–223 bp) was
longer than ITS1 (185–209 bp), consistent with the earlier report in Mussaenda erythrophylla (Andreasen et al., 1999). The
average length falls within the range for other angiosperms
(ITS1: 187–298 bp and ITS2: 187–252 bp; Baldwin et al.,
1995). The total GC content of the entire ITS region ranged
from 59.80% (Schizomussaenda dehiscens) to 63.76% (Landiopsis capuronii) and its average was 61.19%.
TrnT-F analysis—Of 63 trnT-F sequences included in our
studies, 52 are newly published here. The non-aligned trnT-F
sequences ranged from 1662 bp (Gonzalagunia affinis) to 1793
bp (Mussaenda latisepala). The trnT-F alignment of 63 taxa
consisted of 2263 positions, 45 (1.99%) of which were coded
as phylogenetically informative indels and 508 (22.45%) were
phylogenetically informative characters. Of these informative
characters, 314 (61.81%) were from the trnT-L spacer, 81
(15.94%) from the trnL intron, and 113 (22.24%) from the
trnL-F spacer. Within Mussaendeae, alignment of 37 taxa consisted of 1924 positions and contained 131 (7.31%) phylogenetically informative characters. Parsimony analyses of the 63
trnT-F sequences data resulted in 1410 equally parsimonious
trees (each 1207 steps long [L], CI 5 0.638, and RI 5 0.834).
In the strict consensus tree shown in Fig. 1, all investigated
members of Mussaendeae sensu Bremer and Thulin (1998)
formed a strongly supported (BS 5 100) monophyletic group.
Within the Mussaendeae clade, a total of four major clades
were resolved: (1) a strongly supported (BS 5 100) clade containing all sampled Heinsia species; (2) a highly supported (BS
5 100) monophyletic group comprising the sampled Pseudomussaenda species; (3) a moderately supported (BS 5 80)
clade forming three African Mussaenda species (M. afzelii, M.
grandiflora, and M. isertiana); and (4) a weakly supported (BS
5 63) clade containing all investigated Asian Mussaenda,
Aphaenandra uniflora, and five African Mussaenda species
(M. arcuata, M. elegans, M. erythrophylla, M. monticola, and
M. nivea). All sampled Malagasy Mussaenda, Landiopsis capuronii, and Schizomussaenda dehiscens were left unresolved.
Mussaenda s.l. as presently delimited was shown to be polyphyletic because the Afro-Asian Mussaenda species were not
directly related to the Malagasy Mussaenda species. Plus, the
Pseudomussaenda clade was resolved with high support (BS
5 90) as sister to the Afro-Asian Mussaenda clade. Furthermore, all sampled individuals of A. uniflora formed a strongly
supported (BS 5 99) monophyletic group, which was embedded within the Afro-Asian Mussaenda clade. Similarly, we
OF
BOTANY
[Vol. 92
perceived no support for the monophyly of the narrow circumscription of Mussaenda that included only the Afro-Asian
Mussaenda with semaphylls and dehiscent fruits (e.g., Miquel,
1857; Wernham, 1916). In contrast, the broadly circumscribed
Mussaenda including all species with and without semaphylls
and with dehiscent or indehiscent fruits (e.g., Candolle, 1830;
Hooker, 1880; Kurz, 1887; Schumann, 1891) was resolved
with high support (BS 5 98) as monophyletic. This clade was
resolved with strong support (BS 5 100) as sister to the Heinsia clade.
Mussaendeae sensu Bremer and Thulin (1998) was resolved
with strong support (BS 5 94) as sister to Sabiceeae (represented by Sabicea diversifolia and Pseudosabicea becquetii)
and placed within Ixoroideae s.l. The remaining Mussaendeaeassociated genera included in our study were resolved with
high support in the three subfamilies: both Mycetia and
Schradera in Rubioideae sensu Bremer and Manen (2000);
Hoffmania, Gonzalagunia, and Isertia all in Cinchonoideae
sensu stricto (s.s.) (Bremer et al., 1995; Bremer et al., 1999);
and Hippotis, Pentagonia, and Sommera all in Ixoroideae s.l.
(Bremer et al., 1999; Rova et al., 2002). Finally, Acranthera
and Mussaendopsis were placed with strong support in Rubioideae and Ixoroideae s.l., respectively.
ITS analysis—A total of 38 ITS sequences were included
and 36 are newly published here. The aligned matrix contained
655 positions and 103 (15.72%) were phylogenetically informative, eight (1.22%) of which were coded as phylogenetically informative indels. Of these informative characters, 48
(46.60%) were from the ITS1, 53 (51.46%) from the ITS2,
and only two (1.94%) from the 5.8S gene. A parsimony analysis of the ITS data resulted in 524 equally parsimonious trees
(L 5 279, CI 5 0.599, and RI 5 0.811). In the strict consensus tree shown in Fig. 2, all investigated members of Mussaendeae sensu Bremer and Thulin (1998) resolved four major
clades: (1) the Heinsia clade (BS 5 100); (2) a strongly supported (BS 5 100) clade containing Landiopsis and the sampled Malagasy Mussaenda species; (3) the Pseudomussaenda
clade (BS 5 98); and (4) a moderately supported (BS 5 72)
clade containing Aphaenandra uniflora and all sampled African and Asian Mussaenda. Similar with the trnT-F tree (Fig.
1), the Heinsia clade was resolved as a sister to a clade containing the other sampled members of Mussaendeae. The Landiopsis-Malagasy Mussaenda clade was resolved with high
support (BS 5 90) as sister to a moderately supported clade
forming all investigated Schizomussaenda, Pseudomussaenda,
and all Afro-Asian Mussaenda species. Finally, all sampled
individuals of A. uniflora and three African Mussaenda (M.
afzelii, M. grandiflora, and M. isertiana) constituted strongly
supported (BS 5 99 and 100, respectively) monophyletic
groups.
Combined analysis—The results of the SH and partitionhomogeneity tests (Tables 1, 2, respectively) both showed that
the ITS and trnT-F data sets were significantly incongruent.
Visual inspection of the trnT-F and ITS trees (Figs. 1, 2) revealed topological conflicts regarding the placement of the
sampled Asian Mussaenda and Schizomussaenda dehiscens.
The sampled Asian Mussaenda together with five African
Mussaenda (M. arcuata, M. elegans, M. erythrophylla, M.
monticola, and M. nivea) were resolved with weak support (BS
5 63) as a monophyletic group in the trnT-F tree. In contrast,
these Asian Mussaenda species together with three African
March 2005]
ALEJANDRO
ET AL.—POLYPHYLY OF
MUSSAENDA (RUBIACEAE)
547
Fig. 1. Strict consensus tree derived from 1410 equally parsimonious trees based on the phylogenetic analysis of trnT-F sequence data. Numbers above
nodes are bootstrap support values .50%. The thin horizontal bar at the top corresponds to the outgroup, and thick bars indicate clades resolved within
Mussaendeae. Arrows indicate the positions of Landiopsis capuronii and Schizomussaenda dehiscens. Brackets indicate tribal limits of Sabiceeae (Sab) and
Mussaendeae and subfamilial limits.
Mussaenda (M. afzelii, M. grandiflora, and M. isertiana)
formed a strongly supported (BS 5 90) clade in the ITS tree.
In the ITS tree (Fig. 2), S. dehiscens left unresolved within a
poorly supported (BS 5 64) clade that also contains the Pseudomussaenda and the Afro-Asian Mussaenda subclades. In
contrast, this species was left unresolved outside the Pseudo-
mussaenda-Afro-Asian Mussaenda clade in the trnT-F tree
(Fig. 1). The two data sets became significantly congruent (P
5 0.294, Table 1) when S. dehiscens and all sampled Asian
Mussaenda species were excluded. When we excluded S. dehiscens and restored all sampled Asian Mussaenda species, the
two data sets were still significantly congruent (P 5 0.140,
548
AMERICAN JOURNAL
OF
BOTANY
[Vol. 92
Fig. 2. Strict consensus tree derived from 524 equally parsimonious trees based on the phylogenetic analysis of ITS sequence data. Numbers above nodes
are bootstrap support values .50%. The vertical thin bar indicates the outgroups, and thick bars indicate clades resolved within Mussaendeae. Arrows indicate
the positions of Landiopsis capuronii and Schizomussaenda dehiscens.
TABLE 1.
Partition
ITS
trnT-F
Log likelihood scores of ITS and trnT-F partitions combinability implementing the Shimodaira-Hasegawa (SH) test (*P , 0.05).
Constraint
Score
(2InL)
Difference
(2InL)
Optimal ML topology
constrained by trnT-F topology
Optimal ML topology
constrained by ITS topology
2664.339 64
2730.967 74
4669.527 04
4759.832 01
best
66.628 10
best
90.304 97
Significance
(P)
0.000*
0.004*
March 2005]
TABLE 2.
ALEJANDRO
ET AL.—POLYPHYLY OF
MUSSAENDA (RUBIACEAE)
549
Results from the incongruence length difference (ILD) test.
Taxa included
P values
All sampled Mussaendeae taxa 1 two outgroups (Sabicea diversifolia and Warszewiczia coccinea)
All Schizomussaenda dehiscens 1 Asian Mussaenda species 1 Aphaenandra uniflora excluded
Only Schizomussaenda dehiscens excluded
All sampled Asian Mussaenda species 1 Aphaenandra uniflora excluded
0.018
0.294
0.140
0.032
Table 1). In contrast, when we excluded all sampled Asian
Mussaenda species and restored S. dehiscens, they became significantly incongruent (P 5 0.032, Table 1). Based on this
evidence we combined the two data sets (excluding S. dehiscens) in one large matrix, which comprised 2579 bp (including
coded indels); 229 (8.87%) of these 2579 bp were parsimonyinformative characters. Parsimony analyses of the combined
ITS-trnT-F of 36 taxa resulted in three islands containing 240
most equally parsimonious trees (L 5 462, CI 5 0.660, and
RI 5 0.840). The strict consensus tree shown in Fig. 3 was
almost fully resolved and retained almost the same large
monophyletic groups found in both the trnT-F and ITS trees
(Figs. 1, 2). The sampled African Mussaenda were resolved
in two separate clades: the weakly supported (BS 5 64) African clade A (containing M. arcuata, M. elegans, M. erythrophylla, M. monticola, and M. nivea); and the strongly supported (BS 5 100) African clade B (forming M. afzelii, M.
grandiflora, and M. isertiana), which was resolved with high
support (BS 5 89) as sister to all the investigated Asian Mussaenda species. The African Mussaenda clade A collapsed in
both the trnT-F and ITS trees (Figs. 1, 2).
Furthermore, the results of SH tests additionally showed that
the optimal ML topologies of both the ITS (Fig. 1) and trnTF (Fig. 2) trees were not significantly different from the alternative hypothesis constraining all sampled African Mussaenda
species monophyletic (Table 3).
DISCUSSION
Data sets comparison within Mussaendeae—Although the
trnT-F region is three times longer than the ITS region, the
latter yields more informative characters (15.72%) than the
former (7.31%), consistent with the conclusions of Razafimandimbison and Bremer (2002) on Naucleeae s.l. Our results
additionally show that the trnT-L spacer (with 102 variable
sites) and the trnL-F spacer (with 37 variable sites) are more
variable than the trnL intron (with 32 variable sites), also consistent with Razafimandimbison and Bremer (2002). The trnTL spacer also has more phylogenetically informative characters
(64.88%) than the trnL-F spacer (27.43%), further suggesting
that these three regions evolving at different rates are useful
for inferring phylogenetic relationships at different taxonomical levels of Rubiaceae (see also Meve and Liede [2002, 2004]
for Apocynaceae, Gentianales).
Causes of incongruence between the ITS and trnT-F trees
within Mussaendeae sensu Bremer and Thulin (1998)—The
results of the partition-homogeneity tests show that Schizomussaenda dehiscens causes the significant difference between
the trnT-F and ITS data sets (Table 2) despite its unresolved
positions in both the trnT-F and ITS trees (Figs. 1, 2). The
results of both the SH (Table 1) and ILD (Table 2) tests seem
to indicate that the incongruence regarding the placement of
the sampled Asian Mussaenda species in the trnT-F and ITS
data sets is simply due to lack of enough resolution within the
two data sets.
Monophyly of Mussaendeae sensu Bremer and Thulin
(1998)—Our results strongly support (BS 5 100, Figs. 1, 3;
BS 5 89, Fig. 2) the monophyly of Mussaendeae sensu Bremer and Thulin (1998). Although we have not been able to
find any morphological synapomorphy to diagnose the tribe,
the combination of the following morphological characters
commonly found in Mussaendeae can be used to characterize
it: bifid stipules, shaggy trichomes, terminal inflorescences,
heterostyly, semaphylls, corolla lobes with tail-like projections, discoid placentae, and fruits with tanniniferous idioblasts
(Bremer and Thulin, 1998). The trnT-F tree (Fig. 1) further
corroborates the placement of Mussaendeae in Ixoroideae, also
in agreement with Bremer and Thulin (1998) and Rova et al.
(2002).
Placements of some traditionally Mussaendeae-associated
genera in Rubiaceae—The placement of Isertia, Gonzalagunia, and Hoffmania in Cinchonoideae s.s. (Bremer and Thulin,
1998) is further corroborated by our trnT-F tree (Fig. 1). Similarly, the position of Pauridiantha, Mycetia, and Schradera
in Rubioideae (e.g., Bremer and Thulin, 1998; Andersson and
Rova, 1999; Bremer and Manen, 2000; Rova et al., 2002) are
also supported by our results. Four Mussaendeae-associated
genera (Ecpoma, Pentaloncha, Stipularia, and Temnopteryx),
which are not included in the present study due to lack of
material, are tentatively placed by Andersson (1996) in Sabiceeae based on morphology. Recently, the placement of Sommera and Pentagonia (traditionally considered of Cinchonoideae affinity) in Ixoroideae s.l. (Bremer et al., 1999; Rova et
al., 2002) is further corroborated by our trnT-F tree (Fig. 1).
The position of Acranthera within Rubiaceae has always
been controversial since its original description. Acranthera
was originally described by Arnott, but it was Meisner (1838)
who validly published it in his survey of Rubiaceae. The genus
was traditionally placed in Mussaendeae (e.g., Meisner, 1838;
Hooker, 1873; Baillon, 1880; and Schumann, 1891) of Cinchonoideae because of its terminal inflorescences, valvate corolla aestivation, pluriovular-bicarpellate ovaries, and fleshy,
indehiscent fruits, features found in Mussaenda. However,
Acranthera always has simple and entire stipules, homostylous
flowers, corolla completely glabrous inside, stamens inserted
at the base of corolla tube, anthers forming a sheath around
the style, and the secondary pollen presentation (Bremekamp,
1947). We agree with Bremekamp (1947) that placing
Acranthera in Mussaendeae with bifid stipules, heterodistylous
flowers, and anthers attached at least inside of densely pubescent corolla tubes would make this tribe rather heterogeneous,
morphologically. As a result, Bremekamp (1966) removed
Acranthera from Mussaendeae and placed it in its own tribe
Acranthereae Bremekamp ex Darwin within Ixoroideae. Our
550
AMERICAN JOURNAL
OF
BOTANY
[Vol. 92
Fig. 3. Strict consensus tree derived from 240 equally parsimonious trees based on the phylogenetic analysis of combined ITS and trnT-F sequence data.
Numbers above nodes are bootstrap values .50% and below nodes are branch lengths. Vertical thin bar indicates outgroups, and thick bars indicate clades
resolved within Mussaendeae.
TABLE 3.
Partition
ITS
trnT-F
Log likelihood scores for two alternative tree topologies using Shimodaira-Hasegawa test (P , 0.05).
Constraint
Score
(2InL)
Difference
(2InL)
Optimal ML topology
African Mussaenda (clades A and B) monophyletic
Optimal ML topology
African Mussaenda (clades A and B) monophyletic
2664.339 64
2677.728 26
4669.527 04
4674.911 43
best
13.388 62
best
5.384 40
Significance
(P)
0.092
0.370
March 2005]
ALEJANDRO
ET AL.—POLYPHYLY OF
trnT-F tree (Fig. 1) places Acranthera as sister to the other
Rubioideae taxa included in this study.
Razafimandimbison and Bremer’s (2001) study based on the
rbcL sequence data placed Mussaendopsis in Ixoroideae s.l.,
consistent with our present findings. Our study strongly indicates that Mussaendopsis belongs to the strongly supported
clade containing Pentagonia, Hippotis, Condaminea, Sommera, Warszewicsia, and Calycophyllum. Mussaendopsis has
enlarged calyx lobes similar to those found in all investigated
genera in Mussaendeae as defined here. However, the genus
can easily be recognized by its intrapetiolar stipules (Puff and
Igersheim, 1994), and it is so far the only Asian member of
the clade. This clade was also previously identified by Bremer
(1996) and Rova et al. (2002) and is morphologically distinct
from the remaining tribes of Ixoroideae s.l.
Polyphyly of Mussaenda s.l.—The analyses presented
(Figs. 1–3) all support the monophyly of the broadly circumscribed Mussaenda that includes Aphaenandra, Landiopsis,
Pseudomussaenda, and Schizomussaenda. This circumscription maximizes nomenclatural stability because Aphaenandra,
Pseudomussaenda, and Schizomussaenda were originally described as Mussaenda species. However, it makes Mussaenda
highly heterogeneous, morphologically (e.g., with four types
of corolla aestivations: imbricate [Landiopsis], induplicate-valvate [Neomussaenda, Pseudomussaenda, and Schizomussaenda], reduplicate-valvate [Afro-Asian Mussaenda], and induplicate-reduplicate-valvate [Malagasy and Mascarene Mussaenda]; with dehiscent and indehiscent fruits).
Our analyses strongly support the polyphyly of Mussaenda
s.l. as presently delimited. A constrained parsimony analysis
of the combined data sets forcing Malagasy Mussaenda to be
monophyletic with African and Asian Mussaenda results in 60
equally most parsimonious trees, each 501 steps long. These
trees are 39 steps longer than the trees generated from the
unconstrained analyses and therefore are not the most parsimonious solution. Our results indicate that Mussaenda s.l.
needs to be recircumscribed. Here, we restrict Mussaenda to
include only the Afro-Asian Mussaenda and Aphaenandra and
recognize the Malagasy and Mascarene Mussaenda at generic
level. This scenario is consistent with the arguments put forward by Wernham (1914) and Bremekamp (1937) that the
Malagasy and Mascarene Mussaenda are distinct from the African and Asian Mussaenda species because of lack of enlarged calyx lobes and their relatively large flowers. It makes
both the Afro-Asian Mussaenda and the Malagasy Mussaenda
clades homogeneous, morphologically, and also reflects the
distinctness of these two groups, as well as Landiopsis, Pseudomussaenda, and Schizomussaenda. Furthermore, this involves some nomenclatural changes only for the Indian Ocean
(the Malagasy and Mascarene) Mussaenda species. The Malagasy Mussaenda clade is diagnosed by two morphological
synapomorphies: reduplicate- (each lobe folded inward and its
entire inner surface in contact with its adjacent lobes; Robbrecht, 1988: 84) and induplicate- (each lobe folded inward
and its entire inner surface in contact with its adjacent lobes,
Robbrecht, 1988: 84) valvate aestivation (Fig. 4) and densely
pubescent styles. Accordingly, we describe a new genus Bremeria to accommodate all Malagasy and Mascarene Mussaenda species. This generic name honors Professor Birgitta Bremer, who has dedicated her life to the study of Rubiaceae and
whose contributions have changed the views of the classifications of this large family. Furthermore, the circumscription
MUSSAENDA (RUBIACEAE)
551
of Mussaenda by Miquel (1857), also endorsed by Wernham
(1916) and Bremekamp (1937) that restricted Mussaenda to
the Afro-Asian Mussaenda with semaphylls and fleshy indehiscent fruits is not supported by our results, as the two African and Asian Mussaenda species (M. arcuata and M. pubescens, respectively) without semaphylls are both grouped
together with the other sampled Afro-Asian Mussaenda species with semaphylls. The Afro-Asian Mussaenda clade received high support (BS 5 100) in our combined tree (Fig.
3), and its members can be diagnosed by their reduplicatevalvate aestivation and ovary walls with laticiferous cells. We
propose here a much narrower circumscription of Mussaenda,
which includes all Afro-Asian species only.
Phylogenetic relationships and generic limits within Mussaendeae sensu Bremer and Thulin (1998)—Heinsia clade—
Our combined tree (Fig. 3) provides strong support (BS 5
100) for the monophyly of Heinsia, represented here by three
species (H. bussei, H. crinita, and H. zanzibarica). Recognition of Heinsia at the generic level has been widely accepted.
The genus can easily be recognized by a combination of deeply bifid stipules, imbricate corolla aestivation, fleshy and indehiscent fruits, and numerous exotesta cells mostly with wellprotruding tuberculate thickenings along both the radial and
inner tangential walls. Accordingly, its current generic status
should be retained. The morphology-based phylogeny by Andersson (1996) resolves Heinsia as sister to Aphaenandra, a
relationship not supported by our results, which place Heinsia
as sister to a clade formed by all sampled members of Mussaendeae.
Landiopsis-Malagasy Mussaenda (Bremeria) clade—Our
ITS (Fig. 2) and combined tree (Fig. 3) strongly support (BS
5 100) the monophyly of the Bremeria-Landiopsis group,
which is characterized by having much larger corollas compared to its sister-group (the Pseudomussaenda-Afro-Asian
Mussaenda clade) with smaller corollas. Landiopsis (Bosser
and Lobreau-Callen, 1998) and Bremeria are resolved with
strong support (BS 5 100) as sister genera in both the ITS
(Fig. 2) and combined (Fig. 3) trees, but this relationship collapsed in the trnT-F tree (Fig. 1). Landiopsis can easily be
recognized by its subsessile inflorescences, imbricate aestivation, lenticellate and dehiscent fruits, and nonperforate exine.
We have not found any morphological synapomorphy for Landiopsis and Bremeria. However, the former is restricted to dry
habitats in northern Madagascar, whereas the latter is confined
to the low and mid-altitude Malagasy and Mascarene rainforests. Therefore, we maintain the current generic status of Landiopsis.
Palynological studies of Bosser and Lobreau-Callen (1998)
showed evident affinities of Landiopsis to Mussaenda s.l. and
its alliances in having the same apertural system and microendosculptured nexine. As a result, Landiopsis was placed in
Isertieae sensu Andersson (1996). Our molecular results
strongly support the placement of Landiopsis in Mussaendeae
sensu Bremer and Thulin (1998).
Pseudomussaenda clade—Pseudomussaenda was originally
described by Wernham (1916) to accommodate all African
Mussaenda species with dry, capsular fruits and induplicatevalvate aestivation. Our analyses all perceive strong support
for the monophyly of Pseudomussaenda. Wernham (1916),
also endorsed by Robbrecht (1988), tentatively placed Pseu-
552
AMERICAN JOURNAL
OF
BOTANY
[Vol. 92
Fig. 4. Flower buds and an opened flower of Bremeria hymenopogoides showing induplicate- and reduplicate-valvate aestivation. A1, a distinct median
ridge on one of the corolla lobes of a flower bud; A2, a median ridge on one of the corolla lobes of the opened flower (reduplicate-valvate aestivation); B,
infolded part of the margin of one corolla lobe of the opened flower (induplicate-valvate aestivation).
domussaenda in the tribe Condamineeae (Cinchonoideae sensu
Robbrecht, 1988) because of its capsular fruits. Based on their
detailed morphological investigations on some African and
Asian Mussaenda Puff et al. (1993), however, concluded that
Pseudomussaenda belongs to Isertieae sensu Robbrecht (1988)
and it is more closely related to Mussaenda s.l. than it is to
the rest of Isertieae. They further argued that this close relationship is not sufficient to warrant their unification. Our combined tree (Fig. 3) resolves with high support (BS 5 99) the
Pseudomussaenda clade as sister to the Afro-Asian Mussaenda
clade. This is inconsistent with the conclusions of Wernham
(1916) and Robbrecht (1988) but consistent with Puff et al.
(1993). We have not found any morphological synapomorphy
for the Pseudomussaenda and Afro-Asian Mussaenda clade.
It is worth noting that our results appear to conflict with the
conclusions of Puff et al. (1993) because their generic circumscription of Mussaenda s.l. includes the Indian Ocean (Madagascar and Mascarene) Mussaenda species; such circumscription is strongly supported to be polyphyletic by our studies.
However, their studies were based only on some Afro-Asian
Mussaenda; therefore, our findings are actually consistent with
their conclusions. Puff et al. (1993) additionally show that the
ovary walls of the Afro-Asian Mussaenda always contain laticiferous cells, which are absent in Pseudomussaenda. Bridson
and Verdcourt (1988) argue that Pseudomussaenda can be distinguished from Mussaenda s.l. by having five filiform corolla
lobe appendages. However, Puff et al. (1993) show that the
filiform appendages are also present in some Mussaenda species. Plus, these features are also found in many Philippian
Mussaenda species (Alejandro, personal observation). Like
Schizomussaenda, Pseudomussaenda has valvate-induplicate
corolla aestivation and dry, capsular fruits. Schizomussaenda
are with long corollas (always .5 cm) and the inner walls of
seeds with small pits; Pseudomussaenda, however, are with
shorter corollas (always ,5 cm) and the inner walls of seeds
with conspicuous large pits (Puff et al., 1993). Plus, Pseudomussaenda is restricted to mainland Africa, whereas Schizomussaenda is exclusively Southeast Asian. Accordingly, we
maintain the current generic status of Pseudomussaenda.
Afro-Asian Mussaenda clade—The Afro-Asian Mussaenda
clade corresponds to our newly circumscribed Mussaenda s.s.
March 2005]
TABLE 4.
ALEJANDRO
ET AL.—POLYPHYLY OF
MUSSAENDA (RUBIACEAE)
553
List of genera accepted here and their synonyms, geographic distributions, and number of species.
Accepted genera
in Mussaendeae
Bremeria Razafim. and Alejandro
Heinsia DC.
Landiopsis Capuron ex Bosser
Mussaenda s.s. Burm. ex L.
Synonyms
Aphaenandra Miq.; Landia
Comm. ex Juss.
Neomussaenda Tange
Pseudomussaenda Wernham
Schizomussaenda Li
Geographic
distributions
Number of
species
Madagascar and the Mascarenes
Mainland Africa
northern Madagascar
Mainland Africa and Asia
28
4–5*
monotypic
132
Southeast Asia
Mainland Africa
southwestern China westwards to northern Myanmar
2
4–5*
monotypic
* Number of species taken from Mabberley (1997).
The morphological tree shown in Andersson (1996, p. 154)
resolves with strong support the Afro-Asian Mussaenda clade,
represented by three African Mussaenda, M. arcuata, M. glabra, and M. pubescens, as sister to a clade formed by Pseudomussaenda and Schizomussaenda. Our combined tree (Fig.
3), however, resolves with strong support (BS 5 99) the AfroAsian Mussaenda clade as sister to Pseudomussaenda, consistent with the conclusions of Puff et al. (1993).
The Southeast Asian genus Aphaenandra is nested with the
Afro-Asian Mussaenda clade (Figs. 1–3). Miquel (1857) originally described Aphaenandra based on A. sumatrana, which
he tentatively placed in Rondeletieae. Since then, both its identity and position within Rubiaceae have always been under
debate. Hooker (1873), also endorsed by Schumann (1897),
considered Aphaenandra as a dubious genus because of its
suffrutescent habit and mode of vegetative propagation via stolons, making it rather unique within Rubiaceae. De Voogd
(1929), endorsed by Jochems (1929), Craib (1932), Bremekamp (1937), and Robbrecht (1988), all emphasized the striking similarities between Aphaenandra and Mussaenda s.l.: bifid stipules, heterodistylous but functionally dioecious flowers,
upper half of the inside of corolla tubes covered with yellow
hairs, stamens inserted in or above the middle, styles with two
filiform stigmas, and peltate placentae. Craib (1932) reduced
A. sumatrana under synonymy of Mussaenda uniflora Wall.
ex G. Don. Bremekamp (1937), however, argued that both the
small suffrutescent habit and the vegetative propagation mode
of Aphaenandra are sufficient for retaining it as a separate
genus. Accordingly, he resurrected Aphaenandra from synonymy and subsequently made the new combination of
Aphaenandra uniflora (Wall. ex G. Don) Bremekamp. The
morphologically based phylogenetic study by Andersson
(1996, p. 154) resolves Aphaenandra as sister to Heinsia. This
sister-genera relationship is not supported by our results (Figs.
1–3) because all sampled individuals of Aphaenandra uniflora
form a strongly supported monophyletic group, which is always embedded within the Afro-Asian Mussaenda clade.
Therefore, our findings are consistent with Craib’s decision but
inconsistent with the conclusions of Bremekamp (1937). This
placement of Aphaenandra is further supported by morphological data, because it also has typical reduplicate-valvate aestivation, the same basic chromosome number (x 5 11), and
ploidy level (diploid) of the Afro-Asian Mussaenda (Puangsomlee and Puff, 2001). All of the calyx lobes of Aphaenandra
are subequal, a feature also found in some Afro-Asian Mussaenda species (e.g., M. arcuata and M. elegans). Furthermore, the same functionally dioecious flowers have also been
discovered in the Japanese Mussaenda parviflora (Naiki and
Kato, 1999) and most Philippian Mussaenda species (Alejan-
dro, personal observation). So, merging Aphaenandra in Mussaenda is not anomalous as Bremekamp (1937) claimed.
Based on all our evidence presented, we sink Aphaenandra in
the newly circumscribed Mussaenda s.s. Jochems (1929)
pointed out that the fruits of Aphaenandra had a dehiscent
opening in the end, splitting the fruit in two halves. We investigated about 40 specimens of Aphaenandra uniflora and
did not find any dehiscent, capsular fruits; all mature fruits
appear to be fleshy and indehiscent.
Schizomussaenda—Schizomussaenda dehiscens (Li, 1943)
has the same type of induplicate-valvate aestivation as that
found in both Neomussaenda and Pseudomussaenda and dry,
capsular fruits, which are also characteristics for both Landiopsis and Pseudomussaenda. However, S. dehiscens can be
diagnosed by having radial and inner tangential walls of exotesta cells with finely verrucose appearance and small pits
(Puff et al., 1993). In all analyses presented here, this species
is not nested within any of the well-circumscribed Mussaendeae genera we recognize here. Accordingly, we continue to
maintain its current generic status.
Neomussaenda—We are unable to get good material for
Neomussaenda; our efforts to get DNA from herbarium specimens were repeatedly unsuccessful. Neomussaenda was originally described by Tange (1994) to accommodate the species
of the genus Greenea, G. xanthophytoides, and his new species, N. kostermansiana. Neomussaenda can be distinguished
from the remaining Mussaendeae genera by a combination of
induplicate-valvate aestivation, seed exotestal cells with tuberculate inner wall, idioblasts filled with numerous minute
druses, and drupaceous fruits. Tange (1994) argued that Neomussaenda is closely related to Pseudomussaenda and Schizomussaenda based on the following characters: bifid stipules,
terminal thyrsoidal inflorescences, induplicate-valvate aestivation, and fruits with a splitting zone opposite the placenta.
Such relationship is inconsistent with the morphological tree
shown in Andersson (1996, p. 154), which placed Neomussaenda with high support, as sister to the rest of Mussaendeae
sensu Bremer and Thulin (1998). Tange’s hypothesis is also
not supported by our results from the combined data sets as
Pseudomussaenda is resolved as closely related to the AfroAsian Mussaenda clade. We continue to maintain the current
status of Neomussaenda until new data are available. All accepted genera and their synonymies, distribution, and number
of species are given in Table 4.
Evolution and phylogenetic utility of some morphological
features for Mussaendeae and Mussaenda s.s.—Life forms—
554
AMERICAN JOURNAL
There is a great variation of habit in Mussaendeae sensu Bremer and Thulin (1998) ranging from erect shrubs or trees to
scandent or scrambling shrubs, true lianas or small suffrutices.
All Bremeria, Landiopsis, Neomussaenda, and Schizomussaenda are typically erect shrubs or trees (e.g., some of the
Malagasy Bremeria species). In contrast, both scandent or
scrambling, and erect shrubs are commonly found in our newly delimited Mussaenda, with only few true lianas and two
suffrutescent species (M. uniflora and M. parva). Both scandent and erect shrub habits are found in Heinsia. Therefore,
the taxonomic usefulness of life forms in Mussaendeae is rather limited.
Breeding systems—The breeding system of Mussaendeae
sensu Bremer and Thulin is predominantly heterodistylous.
However, some African (e.g., M. chippi and M. tristigmatica,
Hallé, 1961) and Asian (e.g., M. parva, M. parvifolia, M. reinwardtiana, M. uniflora, Bremekamp, 1937; Puff et al., 1993)
Mussaenda are functionally dioecious, suggesting that dioecy
in the newly circumscribed Mussaenda has evolved from distyly. Puff et al. (1993) argue that heterodistyly in some AfroAsian Mussaenda species (e.g., M. sanderiana) and Schizomussaenda dehiscens may not always be stable. Their observations show that the anthers and stigmas of these plants are
not clearly separated spatially, indicating a possible reversal to
homostyly. Based on the evidence described, we conclude that
breeding systems of Mussaenda are evolutionarily labile. In
contrast, Bremeria, Heinsia, Neomussaenda, and Pseudomussaenda are invariably heterodistylous.
Corolla aestivation types—Both imbricate (Heinsia and
Landiopsis) and valvate aestivations are found within Mussaendeae sensu Bremer and Thulin. There are three types of
valvate aestivations: induplicate- (Neomussaenda, Pseudomussaenda, and Schizomussaenda), reduplicate- (the newly delimited Mussaenda), and induplicate- and reduplicate-valvate
(Bremeria) aestivations. Our findings show that both imbricate
and induplicate-valvate aestivations have evolved independently at least two times within Mussaendeae sensu Bremer
and Thulin, indicating that they should not be used alone for
diagnosing genera in this tribe. In contrast, the reduplicateand induplicate-reduplicate-valvate aestivations evolved only
once within the tribe, making them reliable characters for diagnosing our newly circumscribed Mussaenda and Bremeria,
respectively.
Semaphylls—About 67% of the members of Mussaendeae
have developed enlarged, petaloid calyx lobes, which probably
function as optical organs for attracting nectar or pollen-feeding insects from long distances. The presence and/or absence
of semaphylls were traditionally used as primary criterion for
delimiting Mussaenda (e.g., Bremekamp, 1937). Our studies
clearly show that semaphylls have evolved independently numerous times within Mussaendeae sensu Bremer and Thulin
(1998) and the newly delimited Mussaenda. On the other
hand, it is worth noting that 98% of our newly circumscribed
Mussaenda have enlarged calyx lobes. Also, all Mussaendeae
genera with dry, capsular fruits (Landiopsis, Pseudomussaenda, and Schizomussaenda) have semaphylls.
Fruit types—Fruit types were used to segregate Aphaenandra, Pseudomussaenda, and Schizomussaenda, all with dry
capsular fruits, from Mussaenda s.l. with fleshy, indehiscent
OF
BOTANY
[Vol. 92
fruits. Our results, however, indicate that fleshy fruits have
evolved independently at least five times and capsular fruits
at least three times within Mussaendeae sensu Bremer and
Thulin. Therefore, fruit types should not be used as primary
character for recognizing genera in this tribe; on the other
hand, they can be used to characterize genera of Mussaendeae
in combination with other characters. The drupaceous fruit is
only found in Neomussaenda and is a good character for recognizing this genus.
Biogeography of Mussaenda s.s.—The biogeographical
history of the newly circumscribed Mussaenda s.s. can be inferred based on the results presented in this study. Our combined tree (Fig. 3) strongly suggests an African origin of Mussaenda s.s., which appears to have started to diversify in mainland Africa, where a total of 35 species is currently present.
The Asian Mussaenda species seem to have descended from
an African progenitor that must have reached Asia via a longdistance dispersal event. The major radiations of Mussaenda
s.s. seems to have occurred only after the group began to colonize Asia, where ca. 97 species (73.48%) of the 132 Mussaenda are presently found. Despite the fact that the AfroAsian Mussaenda species are shown to be closely related
(Figs. 1–3), mainland Africa and Asia do not share in common
any Mussaenda species. On the other hand, the most widespread African Mussaenda species, M. arcuata, is the only
African Mussaenda species that has successfully reached the
Comoro islands, Madagascar, and the Mascarenes, probably
via stepping-stone dispersal.
Synopsis—Mussaenda Burm. ex L. in Sp. Pl.: 177 (1753);
Gen. Pl. ed. 5: 85 (1754). TYPE: Mussaenda frondosa L. (lectotype, designated by Jayaweera (1963: 239), Hermann s.n.,
BM).
Aphaenandra Miq., in Fl. Ned. Ind. 2: 341 (1857); in Blumea, Suppl. 1, 120 (1937). TYPE: Aphaenandra sumatrana
Miq.
Landia Comm. ex Juss., in Gen. Pl. 201 (1789). TYPE: not
designated.
Shrubs to small trees, scandent shrubs, lianas, or rarely suffrutices. Leaves opposite, decussate, small to large, petiolate
or rarely subsessile; blades ovate or elliptic, usually pubescent
especially on the midrib and veins underneath; stipules bifid,
persistent or deciduous, with few or many colleters in continuous rows and/or in groups of two at the base. Inflorescences
terminal cymose corymbs, glabrous or variously hairy, few to
many-flowered or rarely reduced to a single flower; bracts and
bracteoles few to numerous, entire or trilobed (lateral lobes
always shorter); flowers small, typically heterostylous, usually
5-merous, (sub)sessile or shortly pedicellate; calyx tubes extremely reduced or cup-shaped to shortly tubular or ovoid,
usually pubescent, the lobes extremely short to long, linear to
lanceolate or ovate, rarely foliaceous, occasionally with a single semaphyll, rarely absent or all developed into semaphylls
(e.g., M. philippica var. aurorae); colleters frequently in sinuses between calyx lobes; semaphylls white to creamy yellow
or, rarely, red, elliptic to ovate or orbicular; corolla tubes short,
cylindrical or infundibular, usually forming a distinctly swollen part around anthers, glabrous or pubescent outside and
with unicellular trichomes inside, the lobes reduplicate-valvate
in bud, spreading at anthesis, orange, yellow to red, or rarely
white, elliptic to ovate, rarely linear-lanceolate, abaxially pubescent and adaxially papillate, apical filiform appendages
March 2005]
ALEJANDRO
ET AL.—POLYPHYLY OF
usually present; stamens inserted immediately below the opening or above the middle of the tube in short-styled morphs and
around the middle in long-styled morphs; filaments short, anthers 5, included, bilobed at base, dorsifixed near base; 2-carpellate ovaries, rarely 3–4 carpels; placentae peltate; ovules
numerous, imbedded in fleshy placentae; styles slender, typically glabrous; stigma lobes bifid, included or semi-exserted
in long-styled morphs. Fruits fleshy, indehiscent berry-like, ellipsoid, obovoid to globose, glabrous or pubescent, calyx lobes
deciduous or persistent, warts present or rarely absent; seeds
numerous, endospermic; exotesta cells polygonal, outer tangential walls thin and 6 smooth, radial and inner tangential
walls thickened with large pits.
Number of species: 132 species (97 species in tropical Asia
and 35 species in mainland Africa).
Diagnostic characters: Mussaenda s.s. can easily be diagnosed by having smaller flowers (than Bremeria), 3–5(–8) centimeter long, with reduplicate-valvate aestivation.
Bremeria Razafim. and Alejandro, gen. nov. TYPE: Bremeria landia (Poir.) Razafim. and Alejandro, comb. nova. Basionym: Mussaenda landia Poir., in Lam. Encycl. 4: 392
(1797); D. C., Prodr. 4: 372 (1830); Bojer, H.M.: 165 (1837);
Bak., F.M.S.: 140 (1877); Cordem., F.R.: 503 (1895); R.E.
Vaugham, Mauritius Inst. Bull. 1: 46 (1937). TYPE: Mauritius,
Commerson s.n. (syntype, P–LA; isosyntype, P).
Bremeria ab aliis generibus Mussaendeae facile distincta est
aestvatione valvata-induplicata combinata cum aestivatione
valvata-reduplicata, stylis dense puberulis.
Shrubs to medium-sized trees. Leaves opposite, decussate,
usually pubescent and sometimes scabrous; stipules bifid,
sometimes divided to the base, pubescent on both sides or only
outside, deciduous, the colleters few or many and usually in
groups of two at the base or extending in or above the middle.
Inflorescences typically terminal, paniculate, sometimes reduced to a single flower; bracts and bracteoles present or, rarely, absent, usually entire or bilobed; flowers usually large, 5merous, short to long pedicellate; calyx tubes oblong or ovoid,
variously hairy, rarely glabrous, the lobes long-linear to subulate, mostly unequal, pubescent on both sides or rarely glabrous inside, often persistent; colleters few to numerous, usually between sinuses or sometimes along the margins of calyx
lobes; corolla tubes long, funnel-shaped, always evenly pubescent all over outside and with unicellar trichomes inside,
lobes with both induplicate-valvate and reduplicate-valvate
aestivation in bud, spreading at anthesis, white to pink, or
greenish at the base and at the top and reddish in the middle
(Bremeria landia), usually abaxially pubescent and adaxially
tomentose, apical filiform appendages typically present; stamens inserted at the throat or to midway on the tubes, filaments short, anthers included; ovaries 2-carpellate, styles slender, typically pubescent, stigmas subentire or shortly bilobed,
always included; ovules many per locule. Fruits large, fleshy,
indehiscent berries or drupes, always crowned by the persistent
calyx lobes; seeds numerous, with endosperm, thickened, pitted.
Number of species: 24 species in Madagascar and four species in the Mascarenes.
Diagnostic characters: Bremeria differs from Mussaenda s.s.
by its large flowers, 7–13(–15) cm long, without petaloid calyx
lobes, induplicate-reduplicate-valvate aestivation and densely
pubescent styles.
MUSSAENDA (RUBIACEAE)
555
New combinations—Here, we present our 19 new combinations that consist of 18 described Malagasy and one Mascarene Mussaenda species. All necessary lectotypifications of
Bremeria species will be published in the ongoing systematic
revision of the genus (Andriambololonera and Razafimandimbison, unpublished manuscript).
Bremeria asperula (Wernham) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda aperula Wernham, in J.
Bot. 52: 67 (1914). TYPE: Madagascar, Baron 493 (syntype,
BM, P).
Bremeria decaryi (Homolle) Razafim. and Alejandro, comb.
nova. Basionym: Mussaenda decaryi Homolle, in Not. Syst.
(Paris) 7: 3 (1938). TYPE: Madagascar, Domaine oriental,
Mont de Vatovavy, Perrier de la Bâthie 3988, Perrier de la
Bâthie 3994; Decary 4908, Decary 5455, Decary 4872, Decary 5562 (syntypes, P).
Bremeria erectiloba (Wernham) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda erectiloba Wernham, in J.
Bot. 52: 67–68 (1914). TYPE: Madagascar, Tanala, Ambohimitombo forest, Deans Cowan s.n.; Forsyth Major 274 (syntypes, BM, K, P; isosyntype, MO).
Bremeria fusco-pilosa (Baker) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda fusco-pilosa Baker, in J.
Linn. Soc., Bot. 21: 410 (1885). TYPE: Madagascar, Baron
2467, Baron 2470, Baron 6118 (syntypes, K, P).
Bremeria gerrardi (Homolle) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda gerrardi Homolle, in Not.
Syst. (Paris) 7: 4 (1938). TYPE: Madagascar, Gerrard 21–
6166, no. 37 (syntype, K).
Bremeria humblotii (Wernham) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda humblotii Wernham, in J.
Bot. 52: 70 (1914). TYPE: Madagascar, Humblot 617 (syntype, K, P).
Bremeria hymenopogonoides (Baker) Razafim. and Alejandro, comb. nova. Basionym: Mussaenda hymenopogonoides
Baker, in J. Bot. 20: 138 (1882). TYPE: Madagascar, forests
of the Tanala country, Baron 313 (holotype, K; isotype, P).
Bremeria lantziana (Homolle) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda lantziana Homolle, in
Not. Syst. (Paris) 7: 4 (1938). TYPE: Madagascar, Domaine
oriental, Matatane, Lantz s.n.; Decary 10999 (syntypes, P).
Bremeria latisepala (Homolle) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda latisepala Homolle, in
Not. Syst. (Paris) 7: 5 (1938). TYPE: Madagascar, Expos. Colon. Marseille s.n. (syntype, P).
Bremeria mauritiensis (Wernham) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda mauritiensis Wernham, in
J. Bot. 52: 66–67 (1914). TYPE: Mauritius, in sylvis, ad radices montium; Sur les hautes montagnes, Bojer s.n.; Blackburn s.n. (syntypes, K).
Bremeria monantha (Wernham) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda monantha Wernham, in J.
Bot. 52: 70 (1914). TYPE: Madagascar, between Tamatave and
Antananarivo, Meller s.n.; Thompson s.n. (syntypes, BM, K).
Bremeria perrieri (Homolle) Razafim. and Alejandro, comb.
nova. Basionym: Mussaenda perrieri Homolle, in Not. Syst.
(Paris) 7: 5 (1938). TYPE: Madagascar, Domaine oriental, rivière Anove, côte Est, Perrier de la Bâthie 3753; Decary 131
(syntypes, P).
Bremeria pervillei (Wernham) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda pervillei Wernham, in J.
Bot. 52: 67 (1914). TYPE: Madagascar, Baron 6373, Baron
5800; Hildebrandt 3003 (syntypes, P).
556
AMERICAN JOURNAL
Bremeria pilosa (Baker) Razafim. and Alejandro, comb.
nova. Basionym: Mussaenda pilosa Baker, in Kew. Bull. 105
(1895). TYPE: Madagascar, Baron 6179 (syntypes, K, P).
Bremeria punctata (Drake) Razafim. and Alejandro, comb.
nova. Basionym: Mussaenda punctata Drake, in Grandidier,
Hist. Pl. Madagascar t. 36: 447 (1897). TYPE: Madagascar,
Mahalougouloue?, Thompson s.n. (syntype, BM).
Bremeria ramosissima (Wernham) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda ramosissima Wernham, in
J. Bot. 52: 69 (1914). TYPE: Madagascar, Humblot 392 (syntypes, K, P).
Bremeria scabridior (Wernham) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda scabridior Wernham, in J.
Bot. 52: 71 (1914). TYPE: Madagascar, Baron 1505, Baron
3975 (syntypes, K).
Bremeria trichophlebia (Baker) Razafim. and Alejandro,
comb. nova. Basionym: Mussaenda trichophlebia Baker, in J.
Linn. Soc., Bot. 20: 166 (1882–1883). TYPE: Madagascar,
Baron 493 (syntypes, K, P).
Bremeria vestita (Baker) Razafim. and Alejandro, comb.
nova. Basionym: Mussaenda vestita Baker, in J. Linn. Soc.,
Bot. 20: 166 (1882–1883). TYPE: Madagascar, Betsileo-land,
Baron 55; Langley-Kitching s.n. (syntypes, K).
In conclusion, the present phylogenetic studies highly support the polyphyly of Mussaenda s.l. as currently circumscribed, whereas the monophyly of Mussaendeae sensu Bremer and Thulin (1998) is further supported. We describe a new
genus Bremeria to accommodate the Malagasy and Mascarene
Mussaenda species, merge Aphaenandra in Mussaenda s.s.,
which is now restricted to include only the African and Asian
Mussaenda species. The newly circumscribed Mussaendeae
contains seven genera: Bremeria, Heinsia, Landiopsis, Mussaenda s.s., Neomussaenda, Pseudomussaenda, and Schizomussaenda (Table 4). Many of the vegetative and reproductive
characters traditionally used to delimit genera in Mussaendeae
sensu Bremer and Thulin (1998) are shown to be unreliable
for group recognition because they have evolved independently several times within the tribe or even within Mussaenda
s.s. However, some of them still can be used in combination
with other characters to characterize genera. Our results suggest an African origin of both the newly delimited Mussaenda
s.s. and the Asian Mussaenda.
LITERATURE CITED
ANDERSSON, L. 1996. Circumscription of the tribe Isertieae (Rubiaceae). Opera Botanica Belgica 7: 139–164.
ANDERSSON, L., AND J. H. E. ROVA. 1999. The rps16 intron and the phylogeny of the Rubioideae (Rubiaceae). Plant Systematics and Evolution
214: 161–186.
ANDREASEN, K., B. BALDWIN, AND B. BREMER. 1999. Phylogenetic utility
of the nuclear rDNA ITS region in subfamily Ixoroideae (Rubiaceae):
comparisons with cpDNA rbcL sequence data. Plant Systematics and
Evolution 217: 119–135.
BAILLON, H. 1880. Histoire des plantes, vol. 7. Hachette, Paris, France.
BALDWIN, B. G., 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 phylogeny. Annals of the Missouri Botanical Garden 82: 247–277.
BOSSER, J., AND D. LOBREAU-CALLEN. 1998. Landiopsis Capuron ex Bosser,
genre nouveau de Rubiaceae de Madagascar. Adansonia 20: 131–137.
BREMEKAMP, C. E. B. 1937. Notes on the Rubiaceae of tropical Asia. Blumea
1Supplement: 112–122.
BREMEKAMP, C. E. B. 1947. A monograph of the genus Acranthera Arn. ex
Meisn. Journal of the Arnold Arboretum 28: 261–308.
OF
BOTANY
[Vol. 92
BREMEKAMP, C. E. B. 1966. Remarks on the position, the delimitation and
the subdivision of the Rubiaceae. Acta Botanica Neerlandica 15: 1–33.
BREMER, B. 1996. Combined and separate analyses of morphological and
molecular data in the plant family Rubiaceae. Cladistics 12: 21–40.
BREMER, B., K. ANDREASEN, AND D. OLSSON. 1995. Subfamilial and tribal
relationships in the Rubiaceae based on rbcL sequence data. Annals of
the Missouri Botanical Garden 82: 383–397.
BREMER, B., R. J. JANSEN, B. OXELMAN, M. BACKLUND, H. LANTZ, AND
K.-J. KIM. 1999. More characters or more taxa for a robust phylogeny—
case study from the coffee family (Rubiaceae). Systematic Biology 48:
413–435.
BREMER, B., AND J. F. MANEN. 2000. Phylogeny and classification of the
subfamily Rubioideae (Rubiaceae). Plant Systematics and Evolution 225:
43–72.
BREMER, B., AND M. THULIN. 1998. Collapse of Isertieae, re-establishment
of Mussaendeae, and a new genus of Sabiceeae (Rubiaceae); phylogenetic relationships based on rbcL data. Plant Systematics and Evolution
211: 71–92.
BRIDSON, D. M., AND B. VERDCOURT. 1988. Isertieae. In R. M. Polhill [ed.],
Flora of tropical East Africa, Rubiaceae, part 2. 460–475. A. A. Balkema, Rotterdam, Netherlands.
CANDOLLE, A. P. DE. 1830. Prodromus systematis naturalis regni vegetabilis
4. Treutell & Würtz, Paris, France.
CHASE, M. W., AND H. H. HILLS. 1991. Silica gel: an ideal material for
preservation of leaf samples for DNA studies. Taxon 40: 215–220.
CRAIB, W. G. 1932. Florae Siamensis Enumeratio. A list of the plants known
from Siam with notes of their occurrence. Caprifoliaceae & Rubiaceae
(in part), vol. 2, 1–145. Siam Society, Bangkok, Thailand.
DE VOOGD, C. N. A. 1929. Aphaenandra. De Tropische Natuur 18: 110.
DON, G. 1834. A general history of the dichlamydeous plants, vol. 3, Calyciflorae: viii 1 867. London, UK.
FARRIS, J. S. 1989. The retention index and the rescaled consistency index.
Cladistics 5: 417–419.
FARRIS, J. S., M. KÄLLERSJÖ, A. G. KLUGE, AND C. BULT. 1995. Testing
significance of incongruence. Cladistics 10: 315–319.
FELSENSTEIN, J. 1985. Confidence limits on phylogenies: an approach using
bootstrap. Evolution 39: 783–791.
GOLDMAN, N., J. P. ANDERSON, AND A. G. RODRIGO. 2000. Likelihood-based
tests of topologies in phylogenetics. Systematic Biology 49: 652–670.
HALLÉ, N. 1961. Contribution à l’étude biologique et taxonomique des Mussaendeae (Rubiaceae) d’Afrique tropicale. Adansonia 1: 266–298.
HOOKER, J. D. 1873. Ordo LXXXIV, Rubiaceae. In G. Bentham and J. D.
Hooker [eds.], Genera plantarum ad exemplaria imprimis in herbariis
kewensibus servata defirmata, vol. 2, 7–151. Lovell Reeve, London, UK.
HOOKER, J. D. 1880. Flora of British India, vol. 3, Mussaendeae, 86–92.
Lovell Reeve, London, UK.
JOCHEMS, S. C. J. 1929. Aphaenandra. De Tropische Natuur 18: 153.
KLUGE, A. G., AND J. S. FARRIS. 1969. Quantitative phyletics and the evolution of anurans. Systematic Zoology 18: 1–32.
KURZ, S. 1887. Forest Flora of British Burma, vol. 2, Mussaenda, 55–58.
M/S Bishen Singh Mahendra Pal Singh, Dehra Dun, India and M/S Periodical Experts, Delhi, India.
LEPPIK, E. E. 1977. Calyx-borne semaphylls in tropical Rubiaceae. Phytomorphology 27: 161–168.
LI, H. L. 1943. Schizomussaenda, a new genus of the Rubiaceae. Journal of
the Arnold Arboretum 24: 99–102.
MABBERLEY, D. J. 1997. The plant-book—a portable dictionary of the vascular plants, 2nd ed. Cambridge University Press, Cambridge, UK.
MADDISON, D. R. 1991. The discovery and importance of multiple islands
of most-parsimonious tress. Systematic Zoology 40: 315–328.
MEISNER, C. D. F. 1838. Plantarum vascularium genera secundum ordines
naturales digesta 1, Acranthera, 162. Leipzig, Germany.
MEVE, U., AND S. LIEDE. 2002. A molecular phylogeny and generic rearrangement of the stapelioid Ceropegieae. Plant Systematics and Evolution 234: 171–209.
MEVE, U., AND S. LIEDE. 2004. Subtribal division of Ceropegieae. Taxon 53:
61–72.
MIQUEL, F. A. W. 1857. Flora van Nederlandsch-Indië, vol. 2, Aphaenandra,
341–342. Amsterdam, Netherlands.
NAIKI, A., AND M. KATO. 1999. Pollination system and evolution of dioecy
from distyly in Mussaenda parviflora (Rubiaceae). Plant Species Biology
14: 217–227.
NYLANDER, J. A. 2002. Mrmodeltest, version 1.1b. [Software program dis-
March 2005]
ALEJANDRO
ET AL.—POLYPHYLY OF
tributed online by the author]. Department of Systematic Zoology, EBC,
Uppsala University, Sweden, website, http://www.ebc.uu.se/systzoo/staff/
nylander.html.
PERSSON, C. 2000. Phylogeny of the Neotropical Alibertia group (Rubiaceae),
with emphasis on the genus Alibertia, inferred from ITS and 5S ribosomal DNA sequences. American Journal of Botany 87: 1018–1028.
POPP, M., AND B. OXELMAN. 2001. Inferring the history of the polyploid
Siline aegaea (Caryophyllaceae) using plastid and homoeologous nuclear
DNA sequences. Molecular Phylogenetics and Evolution 20: 474–481.
PUANGSOMLEE, P., AND C. PUFF. 2001. Chromosome numbers of Thai Rubiaceae. Nordic Journal of Botany 21: 165–175.
PUFF, C., AND A. IGERSHEIM. 1994. The character states of Mussaendopsis
Baill. (Rubiaceae-Coptosapelteae). Flora 189: 161–178.
PUFF, C., A. IGERSHEIM, AND U. ROHRHOFER. 1993. Pseudomussaenda and
Schizomussaenda (Rubiaceae): close allies of Mussaenda. Bulletin de
Jardin Botanique National de Belgique 62: 35–68.
RAZAFIMANDIMBISON, S. G., AND B. BREMER. 2001. Tribal delimitation of
Naucleeae (Cinchonoideae, Rubiaceae): inference from molecular and
morphological data. Systematics and Geography of Plants 71: 515–538.
RAZAFIMANDIMBISON, S. G., AND B. BREMER. 2002. Phylogeny and classification of Naucleeae s.l. (Rubiaceae) inferred from molecular (ITS, rbcL
and trnT-F) and morphological data. American Journal of Botany 89:
1027–1041.
ROBBRECHT, E. 1988. Tropical woody Rubiaceae. Opera Botanica Belgica 1:
1–271.
MUSSAENDA (RUBIACEAE)
557
ROVA, J. H. E., P. DELPRETE, L. ANDERSSON, AND V. ALBERT. 2002. A trnLF cpDNA sequence study of the Condamineeae-Rondeletieae-Sipaneeae
complex with implications on the phylogeny of the Rubiaceae. American
Journal of Botany 89: 145–159.
SCHUMANN, K. 1891. Rubiaceae. In A. Engler and K. Prantl [eds.], Die natürlichen Pflanzenfamilien, vol. 4, 1–156. Engelmann, Leipzig, Germany.
SCHUMANN, K. 1897. Rubiaceae. In A. Engler and K. Prantl [eds.], Die natürlichen Pflanzenfamilien, Nachträge [I] zum II–IV, 309–316. Engelmann, Leipzig, Germany.
SHIMODAIRA, H., AND M. HASEGAWA. 1999. Multiple comparisons of loglikelihoods with application to phylogenetic inference. Molecular Biology and Evolution 16: 1114–1116.
SIMMONS, M. P., AND H. OCHOTERENA. 2000. Gaps as characters in sequence-based phylogenetic analyses. Sytematic Biology 49: 369–
381.
SWOFFORD, D. L. 2000. PAUP*: phylogenetic analysis using parsimony (*
and other methods), version 4.0b. Sinauer Associates, Sunderland, Massachusetts, USA.
TANGE, C. 1994. Neomussaenda (Rubiaceae), a new genus from Borneo.
Nordic Journal of Botany 14: 495–500.
WERNHAM, H. F. 1914. The Mussaendas of Madagascar. The Journal of Botany 52: 64–72.
WERNHAM, H. F. 1916. Pseudomussaenda: a new genus of Rubiaceae. The
Journal of Botany 54: 297–301.