bs_bs_banner
Botanical Journal of the Linnean Society, 2014, 174, 141–161. With 3 figures
Species limits and diversification in the Madagascar
olive (Noronhia, Oleaceae)
CYNTHIA HONG-WA1,2* and GUILLAUME BESNARD3
1
Department of Biology, University of Missouri–St. Louis, One University Blvd, St. Louis, MO
63121-4400, USA
2
Missouri Botanical Garden, PO Box 299, St. Louis, MO 63166-0299, USA
3
CNRS, ENFA, Université Paul Sabatier, Laboratoire Evolution et Diversité Biologique, UMR 5174,
31062 Toulouse 4, France
Received 10 December 2012; revised 2 July 2013; accepted for publication 17 August 2013
Studies of ecological and phenotypic diversity in adaptive radiations have greatly contributed to our understanding
of the patterns and processes of species diversification, whilst also challenging our assessment of the nature of
species. Here, analyses of bioclimatic, molecular and morphological data, interpreted in phylogenetic and geographical contexts, were carried out to understand species limits in the Madagascar olive (Noronhia, Oleaceae).
Most species hypotheses exhibit clear boundaries and are supported by at least one line of evidence, but a
contrasting pattern of high morphological and ecological variation with relatively low nucleotide sequence
divergence characterizes the diversification of Noronhia. This diversification was probably driven by fine-scale
ecological and evolutionary processes, as suggested by the poor fit with four models of species diversification of the
biota of Madagascar and the apparent lack of differentiation detected from large-scale bioclimatic data. Overall,
this study offers useful insights into the patterns of plant diversification in Madagascar, the understanding of
which requires good circumscription of species, improved knowledge of their distribution and operational models
of diversification that take into account the particular biology of plants. © 2013 The Linnean Society of London,
Botanical Journal of the Linnean Society, 2014, 174, 141–161.
ADDITIONAL KEYWORDS: adaptive radiations – bioclimatic data – morphology – multivariate analyses –
phylogeny – plant diversification.
INTRODUCTION
The unique and diverse biota of Madagascar testifies
to adaptive radiations of a variety of groups of
animals and plants (Goodman & Benstead, 2003;
Vences, 2005). The island has been isolated from
major landmasses since about 90 Mya (de Wit, 2003).
Although some of the biota are assumed to be Gondwanan relicts, most are thought to be derived from
Tertiary African and Asian colonizers (Yoder &
Nowak, 2006; Russell, Goodman & Cox, 2008; Warren
et al., 2010; Kuntner & Agnarsson, 2011; Reddy et al.,
2012; Buerki et al., 2013; Hong-Wa & Besnard, 2013).
In particular, a good proportion of the Malagasy
endemic plant genera originated in the Miocene
*Corresponding author. E-mail: cynthia.hong-wa@mobot.org
onwards and show strong affinities, in decreasing
order, with African, South-East Asian and Indian
plant taxa (Buerki et al., 2013). Levels of taxonomic
endemism and species diversity are high (Goodman &
Benstead, 2003), endemism being estimated to be
> 90% for non-volant and non-marine vertebrates and
> 80% for vascular plants (Goodman & Benstead,
2003; Callmander et al., 2011). The spatial pattern of
endemism is even more impressive, with many
species having narrow ranges and being known from
only one or a few localities (Schatz et al., 2000;
Goodman & Benstead, 2003; Vences et al., 2009;
Hong-Wa & Arroyo, 2012; Besnard et al., 2013).
The plant genus Noronhia Stadman ex Thouars
(Oleaceae), recently extended to include AfroMalagasy relatives formerly placed in Chionanthus L.
to accommodate phylogenetic relationships (Hong-Wa
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
141
142
C. HONG-WA and G. BESNARD
& Besnard, 2013), now comprises 54 described species
in the Malagasy Floristic Region (MFR, including
Madagascar, Comoros and Mascarenes; Takhtajan,
1986) that form a clade (the MFR clade; Hong-Wa &
Besnard, 2013). Madagascar is the current centre of
diversity of this genus with at least 46 endemic
species; eight other species are found in the Comoros
and Mascarenes, and a dozen species are found in
western and eastern tropical and southern Africa.
Colonization of Madagascar, from an African ancestor,
may have occurred in the late Cenozoic (c. 23 Mya),
and was followed by a burst of diversification starting
at c. 15 Mya with subsequent colonization of smaller
islands in the Indian Ocean (Hong-Wa & Besnard,
2013). Relationships in this group are currently
largely unresolved, and basal divergences seem to
have been relatively rapid, leaving little signal of
lineage separation (Hong-Wa & Besnard, 2013). In
contrast, the extent of morphological and ecological
variation in this group is extremely high (Perrier de
la Bâthie, 1951, 1952; C. Hong-Wa, pers. observ.).
Indeed, there is a great variation in traits, such as
leaf shape, size, texture and venation pattern, flower
shape, size, colour and arrangement, and fruit shape,
size and ornamentation. Noronhia also grows in arid
to humid habitats from sea level to > 2000 m, specializes in various substrates, such as limestone and
quartzite, and is a food source for various species of
lemur and probably also for birds. Furthermore,
Noronhia spp. often grow in sympatry, but show considerable fine-scale ecological differentiation related
to pollination and dispersal, edaphic properties and
topoclimatic conditions (Hong-Wa, 2012). Similar contrasting patterns have been observed in other Malagasy plant groups, such as Gaertnera Lam., Adenia
Forssk., Cyathea Sm., Coffea L. subgenus Coffea and
Pandanus Parkinson (Malcomber, 2002; Hearn, 2006;
Janssen et al., 2008; Anthony et al., 2010; Davis et al.,
2011; Buerki et al., 2012), which all exhibit wide
variation in morphology, but low genetic divergence.
Although the molecular study of the Malagasy flora
has yet to be intensified, large morphological variation in many plant groups has been well documented
and appears to be the signature of diversification
processes operating on the island.
Ecological variation, molecular evolution and morphological differentiation show contrasting patterns in
Noronhia, and this raises questions about the boundaries of the species it contains. Morphology was initially
used to recognize species in this group, which resulted
in a classification (Perrier de la Bâthie, 1952) that is
now outdated and unsatisfactory. Indeed, in spite of
the extent of morphological variation in Noronhia,
species recognition in this genus has been based
mainly on the shape and size of flowers, the length of
the corolla tube, the presence or absence of a corona,
the shape of the corona lobe, the shape of the anthers
and some vegetative features, such as leaf arrangement, shape, size and apex (Perrier de la Bâthie, 1952).
This focus on a few floral characters, even though
flowers of nine species were unknown at the time of
their description (Perrier de la Bâthie, 1949, 1952), has
made the identification of specimens of Noronhia challenging, especially when flowers are missing. Hence,
many specimens of the genus, intensively collected in
the last three decades, could not be assigned to any of
the currently described species and were left as a huge
pile of ‘Noronhia indet.’ in various herbaria (C. HongWa, pers. observ.). This has also been observed in other
plant groups, especially large genera, such as Memecylon L. (Melastomataceae) (Stone, 2012), and represents a real impediment to the study of the Malagasy
flora. Morphology remains important in species delimitation and most morphological plant species may correspond to reproductively independent lineages and
represent biologically real entities (Rieseberg, Wood &
Baack, 2006). However, boundaries of morphological
species may be obscure, for instance, in the case of
cryptic, plastic or polymorphic phenotypes (Duminil &
Di Michele, 2009). As such, an integrative approach
using multiple criteria (e.g. morphology, ecology and
genetics) has been increasingly applied to the species
delineation problem (Dayrat, 2005; Meudt, Lockhart &
Bryant, 2009; Zapata, 2010; Barrett & Freudenstein,
2011). Certainly, meaningful recognition of species,
considered as separately evolving lineages, can be
based on multiple lines of evidence under the general
lineage concept of species, but a single form of evidence
can be sufficient to delimit species in any one case (de
Queiroz, 2007). Each line of evidence, resulting from
an evolutionary process affecting lineage splitting and
divergence, may or may not appear at the same level,
in the same order or with the same magnitude (de
Queiroz, 2007; Padial & De la Riva, 2010). Therefore,
congruence between lines of evidence, although desirable, is not necessary to recognize species and incongruence is even expected (de Queiroz, 2007; Padial &
De la Riva, 2010). The recognition of species that
follows an integrative framework without the necessity of congruence has been referred to as ‘integration
by cumulation’ by Padial et al. (2010), an approach that
considers the additive value of each line of evidence in
distinguishing species. This approach is most appropriate in cases in which diversification occurred
recently and/or over a relatively short period of time
(Sites & Marshall, 2003; Padial et al., 2010). It is
opposed to the ‘integration by congruence’ approach, in
which agreement between at least two sources of
evidence is necessary to recognize species (Padial
et al., 2010). The two approaches are subject to overestimation and under-estimation of species number,
respectively.
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
Malagasy Noronhia, being a monophyletic radiation
and morphologically and ecologically diverse, represents an ideal system to study species diversification.
The mechanisms by which diversity arose in Noronhia,
and in fact most groups of organisms in Madagascar,
are largely unknown. Patterns of diversification and
endemism, mainly faunal, on Madagascar have been
explained by various hypotheses. In particular, Yoder
& Heckman (2006) proposed the ecogeographical constraint hypothesis to explain the east–west vicariance
that follows the sharp bioclimatic division of Madagascar into a humid east and a dry west. Raxworthy &
Nussbaum (1995) found the mountain massifs of
northern Madagascar to be centres of endemism, and
claimed that they have a large role in the generation
and maintenance of diversity in this region. Wilmé,
Goodman & Ganzhorn (2006) suggested that watershed contractions during past climatic oscillations led
to zones of isolation, thus promoting microendemism.
Angel (1942) and Martin (1972) proposed a zoogeographical zonation of Madagascar based on the distributions of its reptiles and lemurs, respectively, and
the role of rivers as barriers to dispersal. Recent
population genetic analyses confirm that rivers and
geographical distance are primary factors structuring
some rodent and lemur populations, although they
may not act as strict barriers to dispersal (e.g.
Quéméré et al., 2010; Rakotoarisoa, Raheriarisena &
Goodman, 2010). All of these hypotheses emphasize
the importance of physical barriers to gene flow in
species divergence. By contrast, the current climate
hypothesis (Pearson & Raxworthy, 2009) invokes a
strong influence of environmental gradients in driving
species divergence and in generating local endemism.
Overall, these hypotheses suggest a major role of
ecological diversification in lineage separation, and
assume a predominantly adaptive speciation. Moreover, the environment of Madagascar is particularly
heterogeneous (Dewar & Richards, 2007), and past
changes may have led to successive population fragmentations and reconnections for many
taxa (allopatry–sympatry oscillations). Such habitat
dynamics may promote speciation events by reinforcement during secondary contacts (e.g. Aguilée, Lambert
& Claessen, 2011).
Although these various hypotheses were formulated
to explain mainly animal species diversity, endemism
and diversification on Madagascar, they may also be
broadly applicable to the flora. In fact, in their paper
advocating Madagascar as a model region for species
diversification study, Vences et al. (2009) included
examples from plant taxa, albeit scarce, that were
consistent with the predictions of the various diversification mechanisms or diversity models [e.g. Coffea
(the Baracoffea alliance), Beccariophoenix; Maurin
et al., 2007; Shapcott et al., 2007; Davis &
143
Rakotonasolo, 2008]. These hypotheses, in particular
the ecogeographical constraint (ECH), riverine barrier
(RBH), watershed contraction (WCH) and current
climate (CCH) hypotheses, can also shed light on the
diversification of Noronhia. Indeed, phylogenetic
analyses could clarify whether: (1) species or clades of
the humid east separate from species or clades of the
dry west according to the ECH; (2) species or clades are
separated by major rivers according to the RBH and
WCH; or (3) they diversify along an environmental
gradient according to the CCH. The RBH and WCH,
although developed from lemur studies, may be particularly relevant for Noronhia, as this genus is predominantly dispersed by various kinds of lemur (e.g.
Eulemur, Lemur, Microcebus, Varecia etc.; Birkinshaw,
1999, 2001; Donati, Lunardini & Kappeler, 1999;
Simmen et al., 2006; Radespiel, 2007; Andriamaharoa,
Birkinshaw & Ludovic, 2010; Martinez, 2010; Thorén,
2011) and, perhaps, also by some birds, but little is
known about the latter. Thus, the distribution of
Noronhia, which covers the entire island, although as
many as 19 of the 46 currently described Malagasy
species have narrow ranges, may mirror that of its
lemur dispersers.
Specimens accumulated since the last taxonomic
treatment of Noronhia (Perrier de la Bâthie, 1952)
have greatly improved our understanding of the distribution patterns of species, which fall into three
range categories: narrow, moderate and wide. Widely
distributed species are mostly in the humid east,
whereas those with moderate and narrow ranges are
in the dry west and topographically complex areas of
the north and south, respectively (C. Hong-Wa,
unpubl. data; but see also Perrier de la Bâthie, 1951).
Narrow endemism may reflect sampling bias owing to
the difficulty of access to some areas, but may also
reflect the role of topographical complexity in the
diversification of this genus, as richness and endemism are highest in mountainous regions (Hong-Wa,
2012). Although taxon sampling remains challenging,
we estimated that ours was a fair representation of
the current state of knowledge on this group. Moreover, sympatric species are often more distinct from
each other than they are from allopatric species, thus
leading to potential underestimation of the actual
diversity. Indeed, under the various scenarios of
diversification discussed here, we can largely assume
some cryptic diversity that is certain to be a taxonomic challenge. Thus, far from being exhaustive, our
current estimate of Noronhia spp. provides a first
approximation of the diversity in this genus that can
be subsequently refined with further botanical, geographical and molecular sampling. However, the
radiation of Noronhia also provides an opportunity to
examine the aforementioned diversification processes
from a plant perspective and to appreciate the taxo-
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
144
C. HONG-WA and G. BESNARD
nomic complexity of a morphologically heterogeneous
but genetically less diverse group.
Our approach to understanding diversification in
Noronhia focused on phylogeny, taxonomy and biogeography. In particular, we used a multifaceted, integrated approach to: (1) reassess the phylogenetic
relationships among Noronhia spp. distributed in the
MFR with a denser taxon sampling than in our previous study (Hong-Wa & Besnard, 2013); (2) examine
patterns of morphological variation and species limits
in a phylogenetic context and across geographical
scales; and (3) evaluate the congruence between the
predictions of the four hypotheses of diversification
mentioned above with phylogenetic patterns in
Noronhia.
MATERIAL AND METHODS
STUDY GROUP
The basic taxonomy of Noronhia was confirmed using
the morphological characters employed by Perrier de
la Bâthie (1949, 1952). We then recorded additional
qualitative diagnostic features [e.g. plant habit, presence of indumentum, colour of stem, leaf, flower and
fruit, texture of leaf and fruit, leaf venation pattern,
flower arrangement, inflorescence structure (diffuse
vs. compact) and fruit shape, ornamentation, apex;
Supporting Information Table S1] that allowed a
better understanding of species limits and a refined
circumscription of species in Noronhia. They are also
suggestive of the extent of evolutionary changes in
this group that perhaps correlate with the colonization of novel environments. We used these additional
features, together with some previously recognized
characters, to sort available specimens from the MFR
into narrowly defined groups or ‘hypotheses of species’
(abbreviated from here onwards as species). Some
named species that could not be distinguished from
others on the basis of these qualitative characters
were combined when they were also not differentiated
in discriminant function analyses of quantitative
variables (data not shown). These included, for
instance, N. ecoronulata H.Perrier vs. N. alleizettei
Dubard, and N. cruciata H.Perrier and N. verticilliflora H.Perrier vs. N. verticillata H.Perrier. In total,
87 qualitatively defined groups were distinguished
{excluding two from the Mascarenes [N. ayresii
(A.J.Scott) Hong-Wa & Besnard and N. boutonii
(A.J.Scott) Hong-Wa & Besnard] for which highresolution images were seen, but no specimens were
available for study} and were named when they corresponded to any of the 54 currently described species
in the MFR; potentially new species are referred to as
‘Noronhia sp.’ and are followed by a number. These 87
putative species represent the operational taxonomic
units (OTUs) used in the analyses of patterns of
variation of bioclimatic, molecular and morphological
variables in this study (see below). Twelve OTUs
included only one or two specimens in our analyses.
Thirty-three morphologically different specimens
could not be assigned to any OTU. Some of these
probably represent new OTUs, but the absence of
either reproductive or highly distinctive characters
prevented us from confidently assigning them to any
group or considering them as distinct groups at the
time this study was conducted. They were initially
included in the multivariate analyses, but were subsequently removed for being uninformative. Additional specimens and molecular data will help to
clarify their identity.
PHYLOGENETIC
ANALYSES
Plastid (trnL-F, trnT-L, trnS-G, trnK-matK) and
nuclear (internal transcribed spacer, ITS) DNA
sequences were obtained from 68 of the 87 OTUs in
the MFR, 42 of which had been included in an earlier
study (Hong-Wa & Besnard, 2013). African relatives
were also included and other members of Oleaceae
were represented as outgroup taxa; specimens of subtribe Schreberineae (Comoranthus Knobl. and Schrebera Roxb.) were used to root the tree. In total, this
molecular dataset included 157 accessions, with 40 of
the 68 OTUs of Noronhia being represented by multiple individuals (Supporting Information Table S2).
Voucher specimens are listed in Table S2 and sampling localities are shown in Figure 1. Thirteen potentially new species remain to be sampled for DNA.
Other OTUs corresponding to the described species of
Noronhia and lacking molecular data are N. crassinodis H.Perrier, N. jeremii Hong-Wa & Callm., N.
leandriana H.Perrier, N. populifolia H.Perrier, N.
urceolata H.Perrier and N. verrucosa H.Perrier, all
from Madagascar; N. populifolia is only known from
the type specimen that is 106 years old.
Laboratory protocols, primers used and data preparation are described in Hong-Wa & Besnard (2013).
All sequences have been submitted to GenBank
(accession numbers are given in Table S2). Despite
inconsistencies between the plastid and ITS datasets
[P < 0.05 from the partition homogeneity test in
PAUP* 4.0b10 (Swofford, 2002)], topological discordances among MFR OTUs were considered to be soft
incongruences, as they were either weakly supported
in both trees (bootstrap < 50%) or were part of polytomies in one of them (data not shown, but see also
Hong-Wa & Besnard, 2013). The use of the incongruence length difference (ILD) test (Farris et al., 1995),
as implemented by the partition homogeneity test, in
the assessment of data partition combinability has
been discouraged because of its sensitivity (Yoder,
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
145
Figure 1. Geographical locations of the samples used for molecular analysis superimposed on different biogeographical
zonations of Madagascar. A, Sampling localities. B, Bioclimatic zones of Schatz (2000). Letters refer to different zones: D,
dry; H, humid; SA, subarid; SH, subhumid. C, Zoogeographical zonations/riverine barriers of Martin (1972); CP, central
plateau; E1 and E2, east 1 and east 2; N, north; NW, north-west; W1 and W2, west 1 and west 2, Sb, Sambirano. D,
Centres of endemism of Wilmé et al. (2006); numbers and letters indicate centres of endemism and zones of retreat–
dispersion (also hatched), respectively. E, Climate clusters of Pearson & Raxworthy (2009); the different clusters are
shown with different symbols in the legend.
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
146
C. HONG-WA and G. BESNARD
Irwin & Payseur, 2001; Barker & Lutzoni, 2002;
Darlu & Lecointre, 2002). The incongruence detected
in our dataset concerns the outgroups more strongly
than the MFR clade (see Hong-Wa & Besnard, 2013).
Therefore, we chose to follow a total evidence
approach and combined the two datasets into a single
matrix for subsequent analyses, as our focus here was
only on the MFR clade, especially the Malagasy
species. Statistics of the different alignments are
given in Supporting Information Table S3. Phylogenetic analyses of the combined dataset used
maximum likelihood (ML) conducted on RAxML
v7.2.6 (Stamatakis, 2006) and Bayesian inference (BI)
carried out on MrBayes v.3.2.1 (Ronquist &
Huelsenbeck, 2003) on the CIPRES portal (Miller,
Pfeiffer & Schwartz, 2010) using the same program
settings as those described in Hong-Wa & Besnard
(2013) and GTR+I+G as the best-fit model of nucleotide substitution for each partition (i.e. plastid and
ITS). Exceptions to the BI approach, which consisted
of two parallel runs of four chains (one cold and three
hot), included 75 million generations sampled every
7500th generation and a Markov chain Monte Carlo
(MCMC) that used an exponential prior of unity for
the shape parameter. Temperature was also reduced
to 0.02 to allow optimal chain swapping. Model
parameters were unlinked across partitions. Convergence of the two runs was checked with the online
application AWTY (Nylander et al., 2008). A 50%
majority-rule consensus tree was generated after discarding the first 20% of trees as a burn-in. Bootstrap
support (BS) values and posterior probabilities (PPs)
obtained from the ML and BI analyses, respectively,
are shown below (Fig. 2).
PATTERNS
OF MORPHOLOGICAL AND NICHE VARIATION
For this study, we measured 973 herbarium specimens from the MFR deposited at G, MO, P, TAN and
TEF (abbreviated according to the Index Herbariorum; Holmgren, Holmgren & Barnett, 1990) to evaluate patterns of morphological variation among the 87
OTUs across their geographical ranges. Thirty-six
quantitative characters (14 leaf, 11 flower and 11
fruit) were recorded (Supporting Information
Table S4). A ruler was used to measure the length and
width of leaf and fruit parts and a digital caliper for
diameter and thickness. All flower parts were measured under a dissecting microscope. For each character, three independent measurements from organs of
different sizes were taken from each specimen to
account for within-individual variation. The average
of these three measurements was assembled in a data
matrix that included specimens in rows and variables
in columns. Data were log-transformed and scaled to
unit variance before analyses in an attempt to ensure
normality and to reduce the effect of measurement
units.
Principal component analysis (PCA) based on correlation matrices of quantitative morphological variables was performed using the package stats on R
2.15.0 (R Development Core Team, 2012) to assess
patterns of variation among the 87 qualitatively
defined OTUs. Only those axes with eigenvalues > 1.0
were extracted using the eigenvalue-one rule or the
Kaiser criterion (Quinn & Keough, 2002). As PCA is
insensitive to collinearity and its success actually
depends on the presence of correlated variables
(Quinn & Keough, 2002; Gotelli & Ellison, 2004), we
excluded none of them. In general, there were few
specimens of each OTU with both flowers and fruits,
and so analyses of flower and fruit variables were
carried out separately. In addition, measurements
were taken from mature organs to reduce size bias,
which led to additional instances of missing values.
Overall, we chose to maximize the number of OTUs
and individuals in each analysis when missing data
were an issue, thus occasionally excluding some characters. Results of the PCAs were plotted on twodimensional scatterplots for visualization. We also
ran a stepwise discriminant function analysis (DFA)
on SPSS (SPSS Inc.) to test the boundaries of the
OTUs and identify the most taxonomically useful
variables. To avoid spurious interpretation, we only
carried out the DFAs when the sample size permitted
the estimation of the group centroid.
Likewise, to identify the climate space and assess
patterns of niche variation for species, we analysed 19
bioclimatic variables and elevation (Supporting Information Table S5) using PCA, as this provides a quick
assessment of the relative positions of each species in
climate space (Zhu et al., 2012; and references
therein). We obtained the 19 bioclimatic variables and
elevation from the WorldClim database (Hijmans
et al., 2005), which contains a set of global climate
layers with a spatial resolution of c. 1 km × 1 km
generated through interpolation of climate data
(monthly total precipitation and monthly mean,
minimum and maximum temperature) obtained from
climate station records from 1950–2000. We used
ArcGIS (ESRI Inc., Redlands, CA, USA) to extract the
bioclimatic and elevation data for each specimen. We
tested for multicollinearity among variables, but did
not find any significant strong correlation (Pearson’s
r ≥ 0.90) at the 0.05 level, suggesting that there is some
variation in each variable that can potentially influence niche variation. Therefore, we did not remove any
bioclimatic variables to avoid excluding those with
potential biological relevance, acknowledging that
overfitting may lead to poor performance of the analysis. However, given the property of PCA mentioned
above, correlated variables are not expected to affect
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
147
Figure 2. Maximum likelihood tree for Noronhia based on the combined plastid and internal transcribed spacer (ITS)
dataset. Numbers above and below the branches are posterior probabilities (PP) and bootstrap support (BS), respectively.
Bold letters refer to supported clades. Dark grey represents African species and light grey indicates species occurring in
Madagascar and the surrounding islands (Comoros and Mascarenes).
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
148
C. HONG-WA and G. BESNARD
greatly the interpretability of our data. Moreover,
over-parameterization has been found to be less problematic than under-parameterization in model building (Warren & Siefert, 2011).
PATTERNS
OF DIVERSIFICATION
We used different biogeographical zonations of Madagascar to represent the four hypotheses of diversification on the island: (1) the bioclimatic zones of Schatz
(2000) for ECH, noting that the same four main types
of biomes summarizing the 14 bioclimatic zones of
Cornet (1974) were recognized by both Schatz (2000)
and Yoder & Heckman (2006); (2) the zoogeographical
zonations of Martin (1972) for RBH; (3) the centres of
endemism of Wilmé et al. (2006) for WCH; and (4) the
climate clusters of Pearson & Raxworthy (2009) for
CCH (Fig. 1, adapted from the respective articles).
Phylogenetic predictions could be derived for the four
hypotheses of diversification (Vences et al., 2009). In
particular, the ECH predicts an east–west partition
between clades or sister species given the sharp bioclimatic distinction between these two regions or, by
extension, a genetic break between clades or species
from different bioclimatic zones. The RBH predicts
genetic differentiation between clades or sister species
occurring on either side of major rivers or, by extension, a genetic break between clades or species from
areas separated by any biogeographical barrier, such
as a mountain range. The WCH predicts a genetic
differentiation among watersheds that served as zones
of isolation. It also predicts that sister species will
occupy contiguous watersheds. Likewise, the CCH
predicts a genetic break between clades or species
from different climate clusters. In the humid eastern
escarpment and the central highlands, sister species
are expected to occupy adjacent climate clusters along
elevational gradients, whereas, in the dry western
lowlands, sister species are expected to be separated
along a north–south gradient.
Overall, genetic differentiation is expected to
increase with geographical distance or the presence of
physical barriers to gene flow. We thus tested for
correlation between pairwise genetic and pairwise
geographical distances for each clade using a Mantel
test performed in the R program and the package
ade4 (Dray & Dufour, 2007). Pairwise genetic distances within a particular clade were estimated with
the package ape (Paradis, Claude & Strimmer, 2004)
on R by considering only a subset of the alignment
that included the members of that clade exclusively.
As the best-fit model of nucleotide substitution (GTR)
is not implemented on the package ape for the estimation of genetic distance, we used the next best
model available (K80) as an alternative. Similarly,
pairwise geographical distances within each clade
were obtained by calculating the Euclidean distance
between the coordinates of the OTUs using the
package stats on R. To assess the concordance
between the observed phylogenetic splits and the
phylogenetic and spatial patterns predicted by each of
the four hypotheses, we first identified the physical
barriers and biogeographical subdivisions in each
hypothesis and other barriers that seemed to us to be
effective. We then mapped these onto the phylogenetic
tree, considering only the supported clades. Taking
into account the ranges of species (Fig. 3), a match
between a phylogenetic split and a biogeographical
pattern is considered as support for the hypothesis it
represents. For instance, a split between sister
species or clades corresponding to an east–west partition supports the ECH if, and only if, the range of
species A is confined to the east and that of species B
to the west. As these hypotheses were formulated only
for mainland Madagascar, excluding smaller islands
such as Nosy Be or Sainte Marie, species or individuals occurring on these smaller islands were removed
from this assessment of spatial patterns of diversification. This approach is rather qualitative, but, combined with the Mantel test, offers preliminary
insights into the patterns of, and potential processes
involved in, the diversification of Noronhia.
RESULTS
PHYLOGENETIC
ANALYSES
The combined plastid (trnL-F, trnT-L, trnS-G, trnKmatK) and nuclear (ITS) dataset contained 4500 bp,
691 bp of which were variable and 438 bp of which
were potentially parsimony informative; the overall
mean sequence divergence was 5.3%. The statistics of
the different alignments are summarized in Table S3.
The analyses of the combined dataset resulted in
highly concordant Bayesian and likelihood phylogenetic hypotheses within Noronhia (Fig. 2). The inclusion of additional taxa of Noronhia in this study
supports previous findings that the genus is strongly
supported as monophyletic (BS = 100%, PP = 1). The
MFR taxa were clearly separated from African species
(BS = 96%, PP = 1), supporting the single radiation of
the genus on Madagascar, with subsequent colonization of surrounding islands. Relationships within this
MFR clade were characterized by short internodes
and large basal polytomies, and were resolved into
only 16 clades with moderate to high support values
(BS ≥ 70% and/or PP ≥ 0.85). These clades are identified here with capital letters (A–P; Fig. 2) and will be
referred to accordingly.
Within each clade, OTUs, usually represented by
multiple individuals, separated clearly from each
other (Fig. 2). For instance, the three OTUs (N. capu-
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
149
Figure 3. Maps showing the distribution of species falling into clades and analysed in a phylogenetic context (Figs S2–
S17).
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
150
C. HONG-WA and G. BESNARD
ronii Bosser, N. gracilipes H.Perrier and N. sambiranensis H.Perrier) forming clade B were all reciprocally
monophyletic. This was also the case in clades C, E,
F, I, J, K, L, N and P. By contrast, ambiguous relationships were observed in clades A, D, G, H, M and
O. Thus, in clade D, the three individuals representing N. linearifolia Boivin ex Dubard did not cluster
together. Similarly, the two individuals of N. boivinii
Dubard in clade H and the individuals of N. pervilleana (Knobl.) H.Perrier and N. sp32 in clade O
failed to form monophyletic groups, but were either
paraphyletic or formed parts of polytomies. Finally,
individuals of N. brevituba H.Perrier and N. linocerioides H.Perrier in clade M were not reciprocally
monophyletic, but formed a strongly supported mixed
clade. The northern population of N. brevituba, represented by N. brevituba 2 and N. brevituba 3 in the
tree (Fig. 2), is sufficiently distinct from the eastern
population, represented by N. brevituba 1, and may
deserve separate status. There were also instances in
which different individuals of one OTU occurred in
more than one clade or were parts of polytomies (e.g.
N. grandifolia H.Perrier, the phylogenetic placement
of which reflects some geographical patterning).
However, within the large polytomy (Fig. 2), individuals of the same OTUs generally clustered together
with high support values (e.g. N. crassiramosa H.Perrier, N. comorensis H.Perrier, N. decaryana H.Perrier
and N. louvelii H.Perrier).
The potentially new species of Noronhia included
here showed overall good genetic differentiation from
each other and from described species (Fig. 2). For
instance, clade C was composed of only two putative
species (N. sp11 and N. sp13), both of which were well
separated. Clades L and P also included putative
species (N. sp38, and N. sp2 and N. sp22, respectively) that were reciprocally monophyletic and distinct from the described species. However, such clear
patterns were not always recovered. In some cases,
individuals of these putative species did not cluster
(N. sp28 and N. sp30) or, if they did, were part of the
large polytomy (N. sp5, N. sp15, N. sp27).
PATTERNS
OF MORPHOLOGICAL AND NICHE VARIATION
Patterns of both morphological and niche variation
were interpreted in phylogenetic and geographical
contexts. The phylogenetic context was established by
focusing on clades with BS ≥ 70% and/or PP ≥ 0.90. In
total, 16 such clades were identified (Fig. 2), and
PCAs and DFAs of morphological and bioclimatic
variables were carried out independently on these
clades to estimate patterns of variation among closely
related OTUs. For OTUs that were part of polytomies
or that were not available for molecular study, analyses were conducted in a biogeographical context using
the subdivisions proposed by Wilmé et al. (2006), as
many species have narrow to moderate ranges confined to one or a few of these. Independent PCAs and
DFAs of morphological and bioclimatic variables were
performed within each of the 12 biogeographical units
to assess patterns of variation among co-occurring
OTUs.
Results of the PCAs of morphological and bioclimatic variables on the 16 phylogenetic clades and 12
biogeographical zones are summarized in Tables 1
and 2 and Supporting Information Figs S2–S31.
OTUs that clearly separate from others in morphospace and climatic space are represented by (+),
whereas those that occupy contiguous or overlapping
morphospace regions are indicated by (–) in Tables 1
and 2. Overall, the OTUs analysed here showed distinct patterns of variation. Individuals of most OTUs
formed a cluster distinct from other such clusters in
multivariate analyses of quantitative vegetative characters alone or a combination of quantitative vegetative and flower or fruit characters. In some cases,
distinction among OTUs was observed only in separate analyses of flower or fruit characters (Tables 1
and 2, Figs S2, S10 and S19). In other cases, the
presence of OTUs behaving as outliers as a result of
extreme variations (e.g. very long leaf blades, very
large flowers) obscured the patterns of variation of
the remaining OTUs by constraining them in one
space of the scatterplot. The exclusion of these outliers allowed us to focus on the patterns of variation of
the remaining OTUs (e.g. Figs S9, S14 and S16).
Of the 50 OTUs placed in clades in the phylogenetic
analyses, it is noteworthy that all but six OTUs
(N. boivinii, N. brevituba, N. linocerioides, N. luteola
var. ankaranensis H.Perrier, N. mangorensis H.Perrier and N. sp9) formed discrete morphological clusters in at least one analysis based on quantitative
vegetative, flower or fruit features only or some combination of these three datasets. Of the 37 OTUs
falling into large polytomies or lacking molecular
data, and thus analysed in a geographical context,
only four (N. verrucosa, N. sp5, N. sp15 and N. sp40)
showed overlapping patterns of variation. OTUs that
included one or two specimens only were also mostly
distinguishable. OTUs showing ambiguous morphologies, such as N. aff. candicans and N. candicans
H.Perrier, also appeared distinct, although only vegetative characters were available (Fig. S30A). These
two taxa were phylogenetically unrelated, but geographically sympatric. Noronhia aff. candicans was
also distinct from other OTUs occurring in the same
area (Fig. S30B). Likewise, N. sp21 is generally
similar to N. buxifolia, but both were distinct in the
morphospace (Fig. S31). Analyses of bioclimatic and
elevation variables showed a high degree of overlap
between the climate space of different OTUs. This is
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
151
Table 1. Summary of the results of principal component analysis (PCA) and discriminant function analysis (DFA) on
bioclimatic and quantitative morphological data of species belonging to clades A–P (Figs S2–S17). Signs indicate the
presence (+) or absence (–) of distinction in the two-dimensional scatterplots of the PCAs. [1] indicates that only a single
specimen was available in the analysis. Abbreviations: mi, material insufficient; NA, not applicable; nk, not known. Level
of significance of Wilks’ λ: *0.01 < P ≤ 0.05, **0.001 < P ≤ 0.01, ***P ≤ 0.001
Morphology
Bioclimate
Vegetative
Flower
Fruit
Clade
Species
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
A
grandifolia
introversa
capuronii
gracilipes
sambiranensis
sp11
sp13
candicans
linearifolia
ambrensis
broomeana
sp20
buxifolia
myrtoides
alleizettei
boinensis
tubulosa
sp21
boivinii
densiflora
mangorensis
ovalifolia
tropophylla
seyrigii
divaricata
sp34
minoriflora
luteola
ankaranensis
emarginata
oblanceolata
peracuminata
sp9
sp38
brevituba
linocerioides
verticillata
sp25
obtusifolia
edentata
lanceolata
cochleata
humbertiana
pervilleana
sp32
incurvifolia
insularis
sp2
cordifolia
sp22
–
–
–
–
–
–
–
–
–
+
+
+
–
–
–
–
–
–
–
–
–
–
–
–
–
– [1]
+
+
–
+
–
–
–
–
–
–
–
–
–
–
+
–
–
–
–
–
+
–
–
–
0.501***
–
–
+
+
+
+
+
+
+
+
+
+
+
+
–
+
–
+
–
+
–
+
–
–
+
+ [1]
+
+
–
+
+
+
–
+
–
–
+
–
+
+
+
–
–
–
+
–
+
–
–
+
0.367***
+
+
+
+ [1]
+
+
+ [1]
+ [1]
+
nk
+ [1]
+
+
+
+
+ [1]
+
+ [1]
–
mi
–
+
+
+
+
+ [1]
+
+
+
+
mi
+
nk
nk
±
±
+
nk
+
+
+
+
+ [1]
+
+
+
+ [1]
+
mi
nk
0.005*
–
–
+
+
+
+ [1]
+
+
+
+
+ [1]
nk
+
+
–
+
–
– [1]
–
mi
–
+ [1]
–
–
+
+ [1]
+
+
–
+
+
+
–
nk
–
–
+
+
mi
+
+
–
+
–
+
–
+ [1]
–
+ [1]
nk
0.569
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
0.053***
0.284***
0.249***
0.000***
0.453***
0.608***
0.071***
0.421***
NA
0.222***
0.321***
0.396***
0.136***
0.013***
0.000***
0.056***
0.003***
0.044***
0.269***
0.275***
0.174***
0.011***
0.646***
NA
0.124***
0.000*
0.053***
0.011***
0.069***
0.001***
NA
NA
NA
NA
0.098*
NA
0.416***
NA
NA
0.000***
NA
0.072***
0.000***
NA
NA
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
0.052***
NA
0.064*
NA
0.182*
NA
NA
0.697*
NA
0.024**
NA
NA
0.012*
NA
NA
152
C. HONG-WA and G. BESNARD
Table 2. Summary of the results of principal component analysis (PCA) and discriminant function analysis (DFA) on
bioclimatic and quantitative morphological data of species found within the 12 centres of endemism (CE) and northern
(RDN) and southern (RDS) retreat–dispersion watersheds of Wilmé et al. (2006), and the Comoros (COM) (Figs S18–F29).
Signs indicate the presence (+) or absence (–) of distinction in the two-dimensional scatterplots of the PCAs. [1] indicates
that only a single specimen was available in the analysis. Abbreviations: mi, material insufficient; NA, not applicable; nk,
not known. Level of significance of Wilks’ λ: *0.01 < P ≤ 0.05, **0.001 < P ≤ 0.01, ***P ≤ 0.001. Species occurring in more
than one watershed are shown in bold
Morphology
Bioclimate
Vegetative
Flower
Fruit
Zone
Species
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
CE1
aff. candicans
crassinodis
aff. crassinodis
longipedicellata
louvelii
sp17
sp28
sp30
thouarsii
crassiramosa
decaryana
jeremii
louvelii
verrucosa
sp4
sp5
sp6
sp8
sp15
sp26
sp36
sp40
crassiramosa
decaryana
louvelii
sp31
sp37
decaryana
sp15
sp26
sp14
sp15
sp16
sp18
sp26
leandriana
louvelii
urceolata
comorensis
leandriana
louvelii
populifolia
sp19
humblotiana
jeremii
sp19
sp36
sp39
–
–
–
+
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
+
–
–
+
+
+
–
–
–
–
–
–
+
–
+
+
–
+
–
+
–
–
+
–
NA
+
+
+
+
+
+
+
+
–
+
–
–
–
–
+
±
+
–
±
–
–
–
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
NA
nk
+
+
mi
+ [1]
nk
nk
nk
mi
nk
+
mi
+
nk
nk
±
nk
nk
±
+
+ [1]
nk
nk
mi
mi
+
+
mi
mi
mi
+
+
+
nk
+ [1]
+
+ [1]
+
+
+ [1]
mi
mi
nk
mi
mi
nk
mi
nk
NA
nk
–
– [1]
– [1]
mi
+
+ [1]
+ [1]
+ [1]
+ [1]
–
+ [1]
–
– [1]
+ [1]
–
+
–
– [1]
–
mi
nk
mi
mi
mi
mi
mi
mi
mi
mi
nk
mi
+ [1]
+ [1]
+ [1]
mi
mi
mi
mi
mi
+ [1]
+ [1]
+
+ [1]
+ [1]
–
– [1]
+
NA
CE2
CE3
CE4
CE5
CE8
CE9 +
COM
CE10
CE11
[1]
0.132***
NA
[1]
[1]
NA
[1]
NA
[1]
NA
[1]
NA
[1]
[1]
[1]
NA
[1]
NA
[1]
[1]
[1]
0.005***
[1]
NA
[1]
[1]
[1]
NA
[1]
NA
[1]
[1]
NA
[1]
NA
[1]
[1]
[1]
NA
[1]
NA
[1]
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
153
Table 2. Continued
Morphology
Bioclimate
Vegetative
Flower
Fruit
Zone
Species
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
PCA
DFA Wilks’ λ
CE12
thouarsii
crassinodis
aff. crassinodis
humblotiana
longipedicellata
sp27
sp30
decaryana
humblotiana
longipedicellata
louvelii
planifolia
urceolata
sp4
sp7
sp8
sp19
sp36
decaryana
leandriana
sp1
–
–
–
–
+
–
–
–
–
–
–
–
–
–
–
–
–
–
+
+
+
0.006***
+
+
–
–
–
+
–
+
+
+
+
+
–
+
–
+
+
–
–
–
+
0.001***
mi
+ [1]
mi
+ [1]
+
+
nk
+ [1]
+
mi
mi
mi
+ [1]
nk
+ [1]
nk
nk
mi
mi
+ [1]
+
NA
+
mi
+ [1]
mi
– [1]
+ [1]
+ [1]
+ [1]
mi
mi
+ [1]
+
mi
+ [1]
+ [1]
mi
mi
+
–
–
+
NA
RDN
RDS
[1]
NA
[1]
[1]
[1]
[1]
NA
[1]
[1]
NA
[1]
[1]
[1]
[1]
not surprising as Noronhia spp. often co-occur in
several forested areas. However, OTUs occupying the
same or similar climatic spaces could usually be differentiated in morphospace (e.g. Figs S3, S4 and S13).
Results of the DFAs are also summarized in
Tables 1 and 2, with a particular focus on the value of
Wilks’ lambda, which was assessed at the significance
level of 0.05. Wilks’ lambda is a test statistic whose
value ranges from zero to unity. A value close to zero
indicates that groups are well separated, and a value
close to unity suggests poor discrimination between
them. The results presented here concern an entire
clade or biogeographical zone, without excluding
overly distinct OTUs as in the PCAs. The group
centroid could not be calculated for OTUs with a small
sample size, resulting in ‘NA’ in our summary Tables 1
and 2. In general, OTUs distinct in the PCAs were also
well discriminated in the DFAs (e.g. clades B and F in
Table 1, Figs S3 and S7). In some cases, OTUs were
also well discriminated despite overlapping patterns
of variation in the PCAs. For example, members of
clades O and P showed little separation in the twodimensional scatterplot of the PCAs (Figs S16 and
S17), but were found to be significantly distinct in the
DFAs. Indeed, OTUs may overlap in some regions of
the morphospace, but still have different centroids.
Lastly, there were a few cases in which OTUs indistinct in the PCAs were also poorly discriminated in
NA
NA
NA
NA
NA
the DFAs (e.g. clades A and I, Figs S2 and S10), but
these occurred only for a subset of the variables, i.e.
either vegetative, flower or fruit. Similar results were
also found for the analyses of bioclimatic variables
(Tables 1 and 2). Overall, OTUs could be distinguished
on any one set of variables or their combination.
PATTERNS
OF DIVERSIFICATION
There was a lack of correlation between genetic and
geographical distances among OTUs within each
clade (Table 3, Fig. 3), suggesting that geographical
distance or the presence of physical barriers to gene
flow alone was not a sufficient predictor of genetic
distance. Indeed, genetic distance was fairly high
within clades J, M, N and P (4.5%, 5.0%, 4.9% and
4.8%, respectively) regardless of geographical distance between members of the clades (16, 32, 217 and
314 km, respectively). Similarly, genetic distance was
relatively low within clades D and O (1.8% and 1.2%,
respectively), but their members were 43 and 383 km
apart, respectively.
There was also an overall lack of evidence in
support of the ECH, as there was no evidence of
strong phylogenetic and spatial fit with the predictions of the RBH, WCH and CCH (Supporting Information Fig. S1). Indeed, the ECH predicted four
bioclimatic subdivisions or at least an east–west par-
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
154
C. HONG-WA and G. BESNARD
Table 3. Results of Mantel tests between cladewise geographical and genetic distances among species within each clade.
Signs represent support for (+), contradictions with (–) or uncertainty over (+/–) the predictions of the four hypotheses.
CCH, current climate hypothesis; ECH, ecogeographical constraint hypothesis; RBH, riverine barrier hypothesis; WCH,
watershed contraction hypothesis; r = Pearson’s correlation coefficient
Genetic distance (%)
Clade
Mean geographical
distance (km)
Mean
Maximum
r
P
ECH
RBH
WCH
CCH
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
55
62
43
110
80
169
307
554
16
139
53
32
217
383
314
2
1.5
1.2
1.9
0.3
1.5
1.7
1.7
3.1
1.8
2.1
3
2.5
0.8
2.6
3.3
3
1.8
2.8
0.5
3
3
3.1
4.5
2.5
4.1
5
4.9
1.2
4.8
0.06
0.75
0.06
0.99
0.98
0.58
0.34
0.91
–0.70
0.93
–0.33
0.15
0.61
0.12
–0.23
0.29
0.15
0.30
0.17
0.35
0.05
0.03
0.13
0.85
0.32
0.80
0.21
0.06
0.17
0.75
+
–
–
–
–
–
+/–
–
–
+
+/–
–
+
–
–
+
–
–
+
–
–
+/–
–
–
+
+/–
–
–
–
–
+
–
–
+
–
–
+/–
–
–
+
+/–
–
–
–
–
+
–
–
–
–
–
+/–
–
–
+
+/–
–
+
–
–
tition, none of which was apparent in the phylogenetic tree, either among or within clades. Instead,
OTUs from the same region or occupying the same
bioclimatic zone tended to cluster together within
clades, but the overall pattern of the phylogenetic tree
was a mosaic. Likewise, the correspondence between
the phylogenetic tree and major genetic breaks predicted by the RBH was rather weak. In many
instances, OTUs occupying different sides of a river
were more closely related than OTUs occurring on the
same side. For example, N. capuronii and N. ambrensis, both occurring north of the Mahavavy river,
belong to different clades within which they were
related to N. sambiranensis and N. sp20, respectively,
both found south of the Mahavavy river (Fig. S1).
Similar patterns were also observed for the WCH,
despite the expected correlation between the distribution of Noronhia and that of its lemur dispersers,
on which this hypothesis was based. However, even
though OTUs from the same watershed were not
always closely related, and OTUs from different
watersheds were not always particularly distinct
(Fig. S1), geographical proximity seems to be of some
importance. There was, for instance, no clade with
OTUs from northern and southern Madagascar.
Instead, related OTUs were found within clusters of
geographically close watersheds, e.g. clade B (CE1,
CE10 and CE12) or clade H (CE2, CE4 and CE5)
(Fig. S1), suggesting a broadly biogeographically constrained differentiation. The CCH also did not obtain
strong support from the data; sister OTUs did not
always occur in adjacent clusters, but more often
would occupy the same cluster, e.g. clade C, or one
OTU would straddle two or more zones, e.g. in clade
J (Fig. S1).
DISCUSSION
EVIDENCE
FOR MULTIPLE DISTINCT SPECIES
In this study, we considered three lines of evidence as
operational criteria to recognize species: bioclimatic,
molecular and morphological distinctions. Nineteen
OTUs lacked molecular data, and so only two lines of
evidence could be used for them. Moreover, the combined dataset of plastid (trnL-F, trnT-L, trnS-G, trnKmatK) and nuclear (ITS) DNA regions resulted in large
polytomies, which meant that it was more difficult to
assess whether the OTUs that fell into polytomies
were distinct both morphologically and phylogenetically, a satisfactory criterion for the ‘integration by
congruence’ approach (Padial et al., 2010). Partially
unresolved topologies were obtained from the separate
analyses of both plastid and nuclear datasets (data not
shown, but see also Hong-Wa & Besnard, 2013). The
analyses of the combined dataset also resulted in
partially unresolved topologies, suggesting that the
data combination only marginally improved the phylogenetic resolution. To explain this pattern, we favour
rapid diversification as the most likely explanation,
given the occurrence of short interior branches and
unresolved basal polytomies (Kim et al., 1996;
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
Seelanan, Schnabel & Wendel, 1997; Wendel & Doyle,
1998; Malcomber, 2002), and the apparent imbalanced
diversity between the MFR Noronhia and its sister
lineage (Schluter, 2000), although hybridization
cannot be excluded (see Hong-Wa & Besnard, 2013).
Consequently, we used the ‘integration by cumulation’
approach (Padial et al., 2010), which has been recommended for lineages in which diversification was
recent and/or rapid relative to that of their sister
lineages (Sites & Marshall, 2003; Padial et al., 2010),
and incorporated various lines of evidence (bioclimatic,
molecular and morphological) to recognize species in
Noronhia. The strength of the various operational
criteria to evaluate the limits among the 87 OTUs
varied, the bioclimatic data being the least discriminating. Indeed, phylogenetically related OTUs usually
occupied the same broad climatic niche (Table 1,
Figs S2–S17); analyses carried out in the geographical
context obviously focused on OTUs from the same
region, and so could hardly show bioclimatic differentiation (Table 2, Figs S18–S29). Molecular data
showed clear differentiation among most OTUs despite
unresolved relationships at deeper levels (Fig. 2). Likewise, quantitative morphological data clearly discriminated most phylogenetically or geographically related
OTUs (Tables 1 and 2, Figs S2–S31). Despite differences in discriminating power, each operational criterion provided useful insights into species limits in
Malagasy Noronhia, especially in cases of unresolved
phylogenetic placement or obscure morphology.
Indeed, given the contingency of each line of evidence
in separately evolving lineages, concurrent support for
the same species is not expected and the lack of
support from a particular criterion does not justify the
rejection of that species as long as it is likely to
represent a distinct evolutionary trajectory (de
Queiroz, 2007; Padial & De la Riva, 2010). However,
their additive value allows species recognition and
permits the inference of major forces underlying
species differentiation.
Overall, 84 of the 87 initial OTUs could definitely
be recognized as species using the ‘integration by
cumulation’ approach (Table 4), and will be treated as
such in an upcoming taxonomic revision of Noronhia
(C. Hong-Wa, unpubl. data). OTUs that were ambiguous include the three pairs N. boivinii and N. mangorensis, N. luteola var. ankaranensis and N. sp9, and
N. brevituba and N. linocerioides (Table 1, Figs S9,
S13 and S14). The morphometric analysis barely supported their initial recognition as OTUs. Nevertheless, there was some genetic distinction between
members of each pair. In particular, accessions of
N. mangorensis and N. luteola var. ankaranensis were
phylogenetically cohesive and fairly distinct from
those of N. boivinii and N. sp9, respectively, suggesting perhaps cryptic or convergent morphology. As the
155
Table 4. Number of species recognized based on different
types of data and their combination
Number of species Number of species
recovered out of
recovered with
Lines of evidence 87 initial OTUs
additional data
Bioclimatic only
+ Molecular
+ Morphological
+ Molecular +
morphological
Molecular only
+ Morphological
Morphological
only
19
50
81
84
–
31
62
65
43 out of 68 with
molecular data
84
81
–
41
–
sequences for N. sp9 were obtained from a herbarium
specimen, our ability to make inferences about the
possible causes and implications of this distinction is,
however, limited. Further molecular sampling will
provide a better understanding of the relationship
between these two OTUs. There was also a genetic
distinction between the northern and eastern populations of N. brevituba, which may suggest distinct
entities. Similarly, N. sp5 and N. sp15 were not
clearly distinct in both climatic and morphological
spaces (Table 2, Fig. S19), but were genetically different (Fig. 2) and geographically distant. Noronhia sp15
is littoral and N. sp5 is inland, occurring above
1000 m elevation; both grow in the eastern part of
Madagascar. Thus, phylogenetic data allowed morphologically nearly similar OTUs to be distinguished.
Obviously, the lack of molecular or morphological data
can hamper species recognition or rejection. For
instance, N. sp40 did not clearly differ from N. sp15
on the basis of vegetative characters alone despite
distinctive morphology (Table 2, Fig. S19). In such
cases, the inability to discriminate the different OTUs
quantitatively is mostly attributable to insufficient
morphological data and lack of molecular data.
Hence, recognition is not rejected until further
samples are available.
MECHANISMS
OF SPECIES DIVERSIFICATION
Our results provided evidence for extensive morphological variation in Noronhia, coupled with poor
signals of basal divergence from the genetic markers
used here. This contrasting pattern suggests a rapid,
recent and/or incomplete radiation (Kim et al., 1996;
Malcomber, 2002; Dunbar-Co, Wieczorek & Morden,
2008). Whether this diversification is ecologically
mediated remains to be determined, but the lack of
differentiation in climatic space indicates that factors
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
156
C. HONG-WA and G. BESNARD
other than climate have played a major role in driving
this pattern. Likewise, the weak correlation between
the distribution of Noronhia and that of its lemur
dispersers, as assessed through the WCH, is rather
disappointing, and can perhaps be accounted for by
repeated but uncoupled contraction and expansion of
their respective ranges. In any case, a correlation,
although faint, still exists and suggests that some
patterns of distribution may be explained by
co-evolutionary processes. Specific studies on species
pairs and consideration of other potential dispersers,
such as birds, may provide useful insights into their
role in the diversification and distribution of members
of this genus. Other large-scale mechanisms suggested by the four models of diversification considered
here also appeared to be of little significance in the
diversification of Noronhia, given the overall poor fit
between the phylogenetic tree and the physical and
ecological barriers to gene flow (Fig. S1), and the lack
of strong correlation between geographical and
genetic distances (Table 3). Indeed, geographical isolation alone or the presence of physical barriers is not
sufficient to explain patterns of diversification in
Noronhia, as most species appeared to have diverged
without a strong isolating role of rivers, drought or
sharp ecogeographical features. Models invoking
parapatric divergence along environmental gradients,
such as the CCH (Pearson & Raxworthy, 2009), also
account for some of the observed patterns (Table 3),
and would presumably be more appropriate than the
other three models as the CCH closely resembles the
phytogeographical subdivisions of Humbert (1955). In
general, the apparent lack of differentiation between
different biogeographical zones and the pronounced
differentiation in the same zone may reflect the signature of past climatic fluctuations and forest dynamics (e.g. Aguilée et al., 2011; Buerki et al., 2013).
Moreover, the four models considered here have been
formulated for terrestrial faunal diversification in
Madagascar, and highly vagile species did not support
these models (Weyeneth, Goodman & Ruedi, 2011).
Thus, they may not be suitable for plants dispersed
by these animals. In fact, it appears that these
animal-based models of diversification have limited
application in plant lineages because of intrinsic differences between these two groups.
The explanation of plant diversification in Madagascar requires more complex models that take into
account the particularities of the life histories of
plants and the porosity of their genome. Indeed, patterns of diversification consistent with some of these
four models have been found in some Malagasy plant
groups (e.g. Andrianoelina et al., 2006; Maurin et al.,
2007; Shapcott et al., 2007; Davis & Rakotonasolo,
2008; all cited in Vences et al., 2009), but others can
hardly be explained by these alone (e.g. Malcomber,
2002; Janssen et al., 2008; Koopman & Baum, 2008;
Anthony et al., 2010). This is not surprising as speciation studies of island endemic plants have often shown
the importance of adaptive shifts in response to novel
environments and chromosome evolution as the main
drivers of radiation (Kim et al., 1996; Stuessy & Ono,
1998; Dunbar-Co et al., 2008; but see also Stuessy
et al., 2006). The patterns of variation in morphology
and ecology in Noronhia are comparable with those of
other prominent island plant radiations (e.g. Plantago
L., woody Sonchus L. alliance, Hawaiian silversword
alliance), the insights from which can be useful for
this study, and other Malagasy plant groups that show
considerable variation in morphology and habitat
preference with low genetic diversity (e.g. Coffea,
Cyathea, Gaertnera, Pandanus). Certainly, the likelihood that fine-scale ecological and evolutionary processes also contributed to the diversification of
Noronhia and most components of the Malagasy flora,
in addition to the coarse-scale processes invoked by
the four models, is very high.
A major evolutionary force in plant diversification,
having been demonstrated in ancient lineages and in
more recently formed species (Soltis & Soltis, 2009;
Wood et al., 2009) and allowing sympatric speciation
(Rieseberg & Willis, 2007), is polyploidy or wholegenome duplication. Although conclusive data are still
unavailable for Noronhia, the presence of multiple
nuclear gene copies (Hong-Wa, 2012) can hint at
polyploidy in some members of this genus, thus
accounting for some of its diversification. It can also
explain the distribution patterns of its species, many of
which have narrow geographical ranges (see Fig. 3),
and can account for the overall lack of phylogenetic
resolution, as recently diverged species may have
porous boundaries (Kane et al., 2009; and references
therein). Moreover, because of the presence of multiple
gene copies, polyploids are expected to exhibit less
inbreeding depression (Lande & Schemske, 1985),
which is often documented in small and isolated populations. Thus, there may have been recurrent polyploidization events in Noronhia, such as those reported
in narrow endemics of Olea (Besnard et al., 2008;
Besnard & Baali-Cherif, 2009). The pattern of diversity and endemism and the unresolved relationships in
Noronhia may also be explained by sympatric speciation with gene flow, as demonstrated in Howea Becc.
(Savolainen et al., 2006; Babik et al., 2009) and
Coprosma J.R.Forst. & G.Forst. (Papadopulos et al.,
2011), both from Lord Howe Island.
Many large plant genera in Madagascar (e.g. Dalbergia L.f., Diospyros L., Dombeya Cav., Grewia L.
and Memecylon; C. Hong-Wa, pers. observ.) exhibit
patterns of morphological and ecological variation
similar to those in Noronhia. Despite the lack of
molecular data for most of these groups, other exam-
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
ples, e.g. Gaertnera, Adenia, Cyathea, Coffea subgenus Coffea and Pandanus (Malcomber, 2002; Hearn,
2006; Janssen et al., 2008; Anthony et al., 2010;
Buerki et al., 2012), show that these variations are
usually coupled with low genetic diversity, suggesting
that the patterns observed in Noronhia are ubiquitous, or at least widespread, among diversified Malagasy plant lineages. Hence, this study provides useful
insights into understanding the diversification of
these taxa, whilst also emphasizing the need for more
operational models aligned to the biology of plants in
general. Addressing these gaps and developing more
studies on Malagasy plant phylogeography and diversification will help to fulfil the recommendation of
Vences et al. (2009) for the establishment of Madagascar as a model region for studies of both plant and
animal species diversification. In addition, improving
our perception of species boundaries and our knowledge of species distribution will contribute greatly to
our assessment of plant diversification in Madagascar
(Stone, 2012). So far, efforts to survey the Malagasy
flora have provided useful information relevant to
conservation and diversity estimates (Kremen et al.,
2008; Callmander et al., 2011; Gautier et al., 2012),
but have not been efficiently applied to understanding
evolutionary history, which thus remains an important avenue for future research.
IMPLICATIONS
FOR DIVERSITY ESTIMATES
Reconciling taxonomy and evolutionary patterns can
be challenging, especially when complex patterns of
diversification are involved. Indeed, the presence of
large polytomies made it difficult to assess the phylogenetic cohesiveness of some Noronhia spp., and convergent morphologies obscured the limits of others.
These situations lead to a taxonomic complexity that is
at least partly attributable to the mechanisms underlying the diversification of Noronhia. It also affects the
estimation of biological diversity. Clearly, considering
different types of data independently will produce
different outcomes in these instances. Although taxon
sampling is still an issue, the integrated analyses of
bioclimatic, molecular and morphological data applied
in this study provided useful insights into the complex
nature of divergence in Noronhia and allowed a better
assessment of species boundaries. Population genetic
analyses and anatomical and chromosomal data would
also provide additional information for better separation of the species of this genus and for unravelling
cryptic diversity.
This study indicates an almost two-fold increase in
species richness in this group since the last taxonomic
treatment 60 years ago (Perrier de la Bâthie, 1952).
This increase results mainly from many new collections accumulated since then, but also involves some
157
generic rearrangements (Hong-Wa & Besnard, 2013)
and rank changes (C. Hong-Wa, unpubl. data). Such
changes are not unique to Noronhia and taxonomic
studies of Malagasy plants often report a considerable
increase in species number. For instance, a study of
Memecylon revealed that most of the specimens collected in the last 25 years represented undescribed
entities and will incur a 70% increase in the diversity
of this genus (Stone, 2012). Similarly, at least a
hundred species remain to be described in Diospyros
(Ebenaceae), based on currently available herbarium
material (G. E. Schatz, pers. comm.), and > 50 new
Impatiens spp. have been described between 2003 and
2007, and the taxonomic work is still in progress (see
the Catalogue of the Vascular Plants of Madagascar
webpage: http://www.tropicos.org/Project/MADA).
Species discovery, in animal and plant groups, has
increased exponentially in Madagascar over the last
decade (see WWF Report, 2011), and will probably
continue for years to come as a result of better sampling, improved analytical techniques and more
researchers (Scheffers et al., 2012). Unfortunately,
gaps in geographical sampling still represent a potential bias, often leading to inflated diversity. This can
be addressed by applying an integrative approach,
which can be useful in deciphering variation, and thus
avoiding taxonomic confusion, in highly diverse
groups such as those mentioned above. Although not
often used in the assessment of the Malagasy flora,
such an approach has already permitted a refinement
of species hypotheses in a broad range of animal taxa,
such as lemurs, tenrecs and frogs (Olson, Goodman &
Yoder, 2004; Yoder et al., 2005; Groeneveld et al., 2009;
Vieites et al., 2009). Given the patterns of microendemism in Madagascar, the traditional approach to
plant species discovery on this island can particularly
benefit from such integrative and analytical assessments to provide robust hypotheses of species, which
are a prerequisite for effective conservation.
ACKNOWLEDGEMENTS
We are grateful to P. F. Stevens, E. A. Kellogg, A. J.
Miller, A. E. Zanne, the associate editor and two
anonymous reviewers for constructive comments. The
herbaria G, MO, P, TAN and TEF made their collections available for morphological study. T. Andriamihajarivo, C. Birkinshaw, R. Letsara, P. Lowry, F.
Rakotonasolo, R. Randrianaivo, A. Randrianasolo, F.
Ratovoson, S. Razafimandimbison, A. Razanatsima,
G. Schatz and S. Trigui provided leaf material for
molecular analyses. The US National Science Foundation (NSF DEB 1011208), Garden Club of America,
Intra-European fellowship (PIEF-GA-2008-220813)
and Harris World Ecology Center at UMSL provided
financial support. The Kellogg Laboratory at UMSL,
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
158
C. HONG-WA and G. BESNARD
EDB Laboratory of the ‘Laboratoire d’Excellence
(LABEX)’ TULIP (ANR-10-LABX-41) and CEBA
(ANR-10-LABX-0025), Missouri Botanical Garden in
St. Louis and its Madagascar Program, Madagascar
National Parks, Ministère des Eaux et Forêts de
Madagascar, Rio Tinto Madagascar (QMM) and Idea
Wild provided administrative, financial, logistic or
material support.
REFERENCES
Aguilée R, Lambert A, Claessen D. 2011. Ecological speciation in dynamic landscapes. Journal of Evolutionary
Biology 24: 2663–2677.
Andriamaharoa H, Birkinshaw CR, Ludovic R. 2010.
Day-time feeding ecology of Eulemur cinereiceps in the
Agnalazaha forest, Mahabo-Mananivo, Madagascar. Madagascar Conservation and Development 5: 55–63.
Andrianoelina O, Rakotondraoelina H, Ramamonjisoa
L, Maley J, Danthu P, Bouvet J-M. 2006. Genetic diversity of Dalbergia monticola (Fabaceae) an endangered tree
species in the fragmented oriental forest of Madagascar.
Biodiversity and Conservation 15: 1109–1128.
Angel F. 1942. Les lézards de Madagascar. Tananarive: Académie Malgache.
Anthony F, Diniz LE, Combes M-C, Lashermes P. 2010.
Adaptive radiation in Coffea subgenus Coffea L. (Rubiaceae)
in Africa and Madagascar. Plant Systematics and Evolution
285: 51–64.
Babik W, Butlin RK, Baker WJ, Papadopulos AST,
Boulesteix M, Anstett M-C, Lexer C, Hutton I,
Savolainen V. 2009. How sympatric is speciation in the
Howea palms of Lord Howe Island? Molecular Ecology 18:
3629–3638.
Barker FK, Lutzoni FM. 2002. The utility of the incongruence length difference test. Systematic Biology 51: 625–637.
Barrett CF, Freudenstein JV. 2011. An integrative
approach to delimiting species in a rare but widespread
mycoheterotrophic orchid. Molecular Ecology 20: 2771–
2786.
Besnard G, Baali-Cherif D. 2009. Coexistence of diploids
and triploids in a Saharan relict olive: evidence from
nuclear microsatellite and flow cytometry analyses. Comptes
Rendus – Biologies 332: 1115–1120.
Besnard G, Christin PA, Malé PJG, Coissac E,
Ralimanana H, Vorontsova M. 2013. Phylogenomics and
taxonomy of Lecomtelleae (Poaceae), an isolated Panicoid
lineage from Madagascar. Annals of Botany 112: 1057–1066.
In press.
Besnard G, García-Verdugo C, Rubio de Casas R, Treier
UA, Galland N, Vargas P. 2008. Polyploidy in the olive
complex (Olea europaea L.): evidence from flow cytometry
and nuclear microsatellite analyses. Annals of Botany 101:
25–30.
Birkinshaw CR. 1999. The importance of the black lemur
(Eulemur macaco) for seed dispersal in Lokobe forest, Nosy
Be. In: Rakotosaminana B, Rasamimanana H, Ganzhorn
JU, Goodman SM, eds. New directions in lemur studies.
New York: Kluwer Academic Publishers, 189–199.
Birkinshaw CR. 2001. Fruit characteristics of species dispersed by the black lemur (Eulemur macaco) in the Lokobe
forest, Madagascar. Biotropica 33: 478–486.
Buerki S, Callmander MW, Devey DS, Chappell L,
Gallaher T, Munzinger J, Haevermans T, Forest F.
2012. Straightening out the screw-pines: a first step in
understanding phylogenetic relationships within Pandanaceae. Taxon 61: 1010–1020.
Buerki S, Devey DS, Callmander MW, Phillipson PB,
Forest F. 2013. Spatio-temporal history of the endemic
genera of Madagascar. Botanical Journal of the Linnean
Society 171: 304–329.
Callmander MW, Phillipson PB, Andriambololonera S,
Rabarimanarivo M, Rakotonirina N, Raharimampionona J, Chatelain C, Gautier L, Lowry II PP. 2011.
The endemic and non-endemic vascular flora of Madagascar
updated. Plant Ecology and Evolution 144: 121–125.
Cornet A. 1974. Notice explicative no. 55. Essai de cartographie bioclimatique à Madagascar. Paris: ORSTOM.
Darlu P, Lecointre G. 2002. When does the incongruence
length difference test fail? Molecular Biology and Evolution
19: 432–437.
Davis AP, Rakotonasolo F. 2008. A taxonomic revision of
the Baracoffea alliance: nine remarkable Coffea species from
western Madagascar. Botanical Journal of the Linnean
Society 171: 355–390.
Davis AP, Tosh J, Ruch N, Fay MF. 2011. Growing coffee:
Psilanthus (Rubiaceae) subsumed on the basis of molecular
and morphological data; implications for the size, morphology, distribution and evolutionary history of Coffea. Botanical Journal of the Linnean Society 167: 357–377.
Dayrat B. 2005. Towards integrative taxonomy. Biological
Journal of the Linnean Society 85: 407–415.
Dewar RE, Richards AF. 2007. Evolution in the hypervariable environment of Madagascar. Proceedings of the
National Academy of Sciences of the United States of
America 104: 13 723–13 727.
Donati G, Lunardini A, Kappeler PM. 1999. Cathemeral
activity of red-fronted brown lemurs (Eulemur fulvus rufus)
in the Kirindy forest/CFPF. In: Rakotosaminana B,
Rasamimanana H, Ganzhorn JU, Goodman SM, eds. New
directions in lemur studies. New York: Kluwer Academic
Publishers, 119–137.
Dray S, Dufour AB. 2007. The ade4 package: implementing
the duality diagram for ecologists. Journal of Statistical
Software 22: 1–20.
Duminil J, Di Michele M. 2009. Plant species delimitation:
a comparison of morphological and molecular markers.
Plant Biosystems 143: 528–542.
Dunbar-Co S, Wieczorek A, Morden CW. 2008. Molecular
phylogeny and adaptive radiation of the endemic Hawaiian
Plantago species (Plantaginaceae). American Journal of
Botany 95: 1177–1188.
Farris JS, Källersjö M, Kluge AG, Bult C. 1995. Constructing a significance test for incongruence. Systematic Biology
44: 570–572.
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
Gautier L, Chatelain C, Callmander MW, Phillipson PB.
2012. Richness, similarity and specificity of Madagascar
flora compared with sub-Saharan Africa. Plant Ecology and
Evolution 145: 55–64.
Goodman SM, Benstead JP. 2003. The natural history of
Madagascar. Chicago, IL: The University of Chicago Press.
Gotelli NJ, Ellison AM. 2004. A primer of ecological statistics. Sunderland, MA: Sinauer Associates.
Groeneveld LF, Weisrock DW, Rasoloarison RM, Yoder
AD, Kappeler PM. 2009. Species delimitation in lemurs:
multiple genetic loci reveal low levels of species diversity in
the genus Cheirogaleus. BMC Evolutionary Biology 9: 30.
Hearn DJ. 2006. Adenia (Passifloraceae) and its adaptive
radiation: phylogeny and growth form diversification. Systematic Botany 31: 805–821.
Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A.
2005. Very high resolution interpolated climate surfaces for
global land areas. International Journal of Climatology 25:
1965–1978.
Holmgren PK, Holmgren NH, Barnett LC. 1990. Index
herbariorum. Part 1: the herbaria of the world. The Bronx:
International Association for Plant Taxonomy.
Hong-Wa C. 2012. Diversification and coexistence in the
Madagascar olive (Noronhia, Oleaceae). PhD Dissertation,
University of Missouri, St. Louis, MO.
Hong-Wa C, Arroyo TPF. 2012. Climate-induced range contraction in the Malagasy endemic plant genera Mediusella
and Xerochlamys (Sarcolaenaceae). Plant Ecology and Evolution 145: 302–312.
Hong-Wa C, Besnard G. 2013. Intricate patterns of phylogenetic relationships in the olive family as inferred from
multi-locus plastid and nuclear DNA sequence analyses: a
close-up on Chionanthus and Noronhia (Oleaceae). Molecular Phylogenetics and Evolution 67: 367–378.
Humbert H. 1955. Les territoires phytogéographiques de
Madagascar. Année Biologique 31: 439–448.
Janssen T, Bystriakova N, Rakotondrainibe F, Coomes
D, Labat J-N, Schneider H. 2008. Neoendemism in Madagascar scaly tree ferns results from recent, coincident diversification bursts. Evolution 62: 1876–1889.
Kane NC, King MG, Barker MS, Raduski A, Karrenberg
S, Yatabe Y, Knapp SJ, Rieseberg LH. 2009. Comparative genomic and population genetic analyses indicate
highly porous genomes and high levels of gene flow between
divergent Helianthus species. Evolution 63: 2061–2075.
Kim S-G, Crawford DJ, Franscico-Ortega J, SantosGuerra A. 1996. A common origin for woody Sonchus and
five related genera in the Macaronesian islands: molecular
evidence for extensive radiation. Proceedings of the National
Academy of Sciences of the United States of America 93:
7743–7748.
Koopman M, Baum D. 2008. Phylogeny and biogeography of
tribe Hibisceae (Malvaceae) on Madagascar. Systematic
Botany 33: 364–374.
Kremen C, Cameron A, Moilanen A, Phillips SJ, Thomas
CD, Beentje H, Dransfield J, Fisher BL, Glaw F, Good
TC, Harper GJ, Hijmans RJ, Lees DC, Louis JED,
Nussbaun RA, Raxworthy CJ, Razafimpahanana A,
159
Schatz GE, Vences M, Vieites DR, Wright PC, Zjhra
ML. 2008. Aligning conservation priorities across taxa in
Madagascar with high-resolution planning tools. Science
320: 222–226.
Kuntner M, Agnarsson I. 2011. Biogeography and diversification of hermit spiders on Indian Ocean islands (Nephilidae: Nephilengys). Molecular Phylogenetics and Evolution
59: 477–488.
Lande R, Schemske DW. 1985. The evolution of selffertilization and inbreeding depression in plants. I. Genetic
models. Evolution 39: 24–40.
Malcomber S. 2002. Phylogeny of Gaertnera Lam. (Rubiaceae) based on multiple DNA markers: evidence of a rapid
radiation in a widespread, morphologically diverse genus.
Evolution 56: 42–57.
Martin RD. 1972. Adaptive radiation and behavior of the
Malagasy lemurs. Philosophical Transactions of the Royal
Society 264: 295–352.
Martinez BT. 2010. Forest restoration in Masoala National
Park, Madagascar: the contribution of red-ruffed lemur
(Varecia rubra) and the livelihoods of subsistence farmers at
Ambatoladama. PhD Dissertation, University of Minnesota,
St. Paul, MN.
Maurin O, Davis AP, Chester M, Mvungi EF, Jaufeerally
Y, Fay MF. 2007. Towards a phylogeny for Coffea (Rubiaceae): identifying well-supported lineages based on nuclear
and plastid DNA sequences. Annals of Botany 100: 1565–
1583.
Meudt HM, Lockhart PJ, Bryant D. 2009. Species delimitation and phylogeny of a New Zealand plant species radiation. BMC Evolutionary Biology 9: 111.
Miller MA, Pfeiffer W, Schwartz T. 2010. Creating the
CIPRES Science Gateway for inference of large phylogenetic
trees. Proceedings of the Gateway Computing Environments Workshop (GCE), 14 November 2010, New Orleans,
LA, 1–8. IEEE.
Nylander JAA, Wilgenbusch JC, Warren DL, Swofford
DL. 2008. AWTY (are we there yet?): a system for graphical
exploration of MCMC convergence in Bayesian phylogenetics. Bioinformatics Applications Note 24: 581–583.
Olson L, Goodman SM, Yoder AD. 2004. Illumination of
cryptic species boundaries in long-tailed shrew-tenrecs
(Mammalia: Tenrecidae; Microgale), with new insights into
geographic variation and distributional constraints. Biological Journal of the Linnean Society 83: 1–22.
Padial JM, De la Riva I. 2010. A response to recent proposals for integrative taxonomy. Biological Journal of the
Linnean Society 101: 747–756.
Padial JM, Miralles A, De la Riva I, Vences M. 2010. The
integrative future of taxonomy. Frontiers in Zoology 7: 16.
Papadopulos AST, Baker WJ, Crayn D, Butlin RK,
Kynast RG, Hutton I, Savolainen V. 2011. Speciation
with gene flow on Lord Howe Island. Proceedings of the
National Academy of Sciences of the United States of
America 108: 13 188–13 193.
Paradis E, Claude J, Strimmer K. 2004. APE: analyses of
phylogenetics and evolution in R language. Bioinformatics
20: 289–290.
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
160
C. HONG-WA and G. BESNARD
Pearson RG, Raxworthy CJ. 2009. The evolution of local
endemism in Madagascar: watersheds versus climatic gradient hypotheses evaluated by null biogeographic models.
Evolution 63: 959–967.
Perrier de la Bâthie H. 1949. Révision des Oléacées de
Madagascar et des Comores. Mémoire de l’Institut Scientifique de Madagascar, Série B 2: 275–310.
Perrier de la Bâthie H. 1951. Notes biologiques sur les
Oléacées de Madagascar et des Comores. Mémoires de
l’Institut Scientifique de Madagascar, Série B 3: 175–185.
Perrier de la Bâthie H. 1952. Oléacées. In: Humbert H, ed.
Flore de Madagascar et des Comores. 166e famille. Paris:
Muséum National d’Histoire Naturelle, 1–89.
de Queiroz K. 2007. Species concepts and species delimitation. Systematic Biology 56: 879–886.
Quéméré E, Crouau-Roy B, Rabarivola C, Louis EE,
Chikhi L. 2010. Landscape genetics of an endangered
lemur (Propithecus tattersalli) within its entire fragmented
range. Molecular Ecology 19: 1606–1621.
Quinn GP, Keough MJ. 2002. Experimental design and data
analysis for biologists. Cambridge: Cambridge University
Press.
R Development Core Team. 2012. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. ISBN 3-900051-07-0.
Available at: http://www.R-project.org
Radespiel U. 2007. Ecological diversity and seasonal adaptations of mouse lemurs (Microcebus spp.). In: Gould L,
Sauther ML, eds. Lemurs: ecology and adaptation (developments in primatology: progress and prospects). New York:
Springer, 211–234.
Rakotoarisoa JE, Raheriarisena M, Goodman SM. 2010.
Phylogeny and species boundaries of the endemic species
complex, Eliurus antsingy and E. carletoni (Rodentia:
Muroidea: Nesomyidae), in Madagascar using mitochondrial
and nuclear DNA sequence data. Molecular Phylogenetics
and Evolution 57: 11–22.
Raxworthy CJ, Nussbaum RA. 1995. Systematics, speciation and biogeography of the dwarf chameleons (Brookesia;
Reptilia, Squamata, Chamaelontidae) of northern Madagascar. Journal of Zoology 235: 525–558.
Reddy S, Driskell A, Rabosky DL, Hackett SJ,
Schlenberg TS. 2012. Diversification and the adaptive
radiation of the vangas of Madagascar. Proceedings of the
Royal Society of London B 279: 2062–2071.
Rieseberg LH, Willis JH. 2007. Plant speciation. Science
317: 910–914.
Rieseberg LH, Wood TE, Baack EJ. 2006. The nature of
plant species. Nature 440: 524–527.
Ronquist F, Huelsenbeck JP. 2003. MrBayes 3: Bayesian
phylogenetic inference under mixed models. Bioinformatics
Applications Note 19: 1572–1574.
Russell AL, Goodman SM, Cox MP. 2008. Coalescent
analyses support multiple mainland-to-island dispersals in
the evolution of Malagasy Triaenops bats (Chiroptera: Hipposideridae). Journal of Biogeography 35: 995–1003.
Savolainen V, Anstett M-C, Lexer C, Hutton I, Clarkson
JJ, Norup MV, Powell MP, Springate D, Salamin N,
Baker WJ. 2006. Sympatric speciation in palms on an
oceanic island. Nature 441: 210–213.
Schatz GE. 2000. Endemism in the Malagasy tree flora. In:
Lourenço WR, Goodman SM, eds. Diversity and endemism
in Madagascar. Paris, ORSTOM: Mémoires de la Société de
Biogéographie, Société de Biogéographie, 1–9.
Schatz GE, Birkinshaw C, Lowry PP II, Randriatafika
F, Ratovoson F. 2000. The endemic plant families of
Madagascar project: integrating taxonomy and conservation. In: Lourenço WR, Goodman SM, eds. Diversity and
endemism in Madagascar. Paris, ORSTOM: Mémoires de la
Société de Biogéographie, Société de Biogéographie, 11–
24.
Scheffers BR, Joppa LN, Pimm SL, Laurance WF. 2012.
What we know and don’t know about Earth’s missing biodiversity. Trends in Ecology and Evolution 27: 501–510.
Schluter D. 2000. The ecology of adaptive radiation. Oxford:
Oxford University Press.
Seelanan T, Schnabel A, Wendel JF. 1997. Congruence
and consensus in the cotton tribe (Malvaceae). Systematic
Biology 22: 259–290.
Shapcott A, Rakotoarinivo M, Smith RJ, Lysakova G,
Fay MF, Dransfield J. 2007. Can we bring Madagascar’s
critically endangered palms back from the brink? Genetics,
ecology and conservation of the critically endangered palm
Beccariophoenix madagascariensis. Botanical Journal of the
Linnean Society 154: 589–608.
Simmen B, Sauther ML, Soma T, Rasamimanana H,
Sussman RW, Jolly A, Tarnaud L, Hladik A. 2006. Plant
species fed on by Lemur catta in gallery forests of the
southern domain of Madagascar. In: Jolly A, Sussman RW,
Koyama N, Rasamimanana H, eds. Ringtailed lemur
biology: Lemur catta in Madagascar (developments in primatology: progress and prospects). New York: Springer,
55–68.
Sites JW, Marshall JC. 2003. Delimiting species: a Renaissance issue in systematic biology. Trends in Ecology and
Evolution 18: 462–470.
Soltis PS, Soltis DE. 2009. The role of hybridization in plant
speciation. Annual Review of Plant Biology 60: 561–580.
Stamatakis A. 2006. RAxML-VI-HPC: maximum likelihoodbased phylogenetic analyses with thousands of taxa and
mixed models. Bioinformatics Applications Note 22: 2688–
2690.
Stone RD. 2012. Endemism, species richness and morphological trends in Madagascan Memecylon (Melastomataceae). Plant Ecology and Evolution 145: 145–151.
Stuessy TF, Jakubowsky G, Gómez RS, Pfosser M,
Schlüter PM, Fer T, Sun B-Y KH. 2006. Anagenetic
evolution in island plants. Journal of Biogeography 33:
1259–1265.
Stuessy TF, Ono M. 1998. Evolution and speciation of island
plants. Cambridge: Cambridge University Press.
Swofford DL. 2002. PAUP*. Phylogenetic Analysis Using
Parsimony (* and other methods). Version 4. Sunderland,
MA: Sinauer Associates.
Takhtajan A. 1986. Floristic regions of the world. Berkeley,
CA: University of California Press.
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161
DIVERSIFICATION OF NORONHIA
Thorén SIK. 2011. Comparative feeding ecology of two sympatric mouse lemurs (Microcebus spp.) in northwestern
Madagascar. PhD Dissertation, University of Veterinary
Medicine, Hannover.
Vences M. 2005. Madagascar as a model region for the study
of tempo and pattern in adaptive radiations. In: Huber BA,
Lampe KH, eds. African biodiversity: molecules, organisms,
ecosystems. Proceedings of the 5th International Symposium
on Tropical Biology, Museum Koenig. Bonn: Springer, 69–84.
Vences M, Wollenberg KC, Vieites DR, Lees DC. 2009.
Madagascar as a model region of species diversification.
Trends in Ecology and Evolution 24: 456–465.
Vieites DR, Wollenberg KC, Andreone F, Köhler J, Glaw
F, Vences M. 2009. Vast underestimation of Madagascar’s
biodiversity evidenced by an integrative amphibian inventory. Proceedings of the National Academy of Sciences of the
United States of America 106: 8267–8272.
Warren BH, Strasberg D, Bruggemann J, Prys-Jones
RP, Thébaud C. 2010. Why does the biota of Madagascar
region have such a strong Asiatic flavor? Cladistics 26:
526–538.
Warren DL, Siefert SN. 2011. Ecological niche modeling in
Maxent: the importance of model complexity and the performance of model selection criteria. Ecological Applications
21: 335–342.
Wendel JF, Doyle JJ. 1998. Phylogenetic incongruence:
window into genome history and molecular evolution. In:
Soltis DE, Soltis PS, Doyle JJ, eds. Molecular systematics of
plants II: DNA sequencing. Norwell, MA: Kluwer Academic
Publishers, 265–296.
Weyeneth N, Goodman SM, Ruedi M. 2011. Do diversification models of Madagascar’s biota explain the population
structure of the endemic bat Myotis goudoti (Chiroptera:
Vespertilionidae)? Journal of Biogeography 38: 44–54.
Wilmé L, Goodman SM, Ganzhorn JU. 2006. Biogeographic evolution of Madagascar’s microendemic biota.
Science 312: 1063–1065.
161
de Wit MJ. 2003. Madagascar: heads it’s a continent, tails it’s
an island. Annual Review of Earth and Planetary Science
31: 213–248.
Wood TE, Takebayashi N, Barker MS, Mayrose I,
Greenspoon PB, Rieseberg L. 2009. The frequency of
polyploid speciation in vascular plants. Proceedings of the
National Academy of Sciences of the United States of
America 106: 13 875–13 879.
WWF Report. 2011. Treasure Island: new biodiversity on
Madagascar (1999–2010). Available at: http://assets.wwf
.org.uk/downloads/madagascarspeciesreport.pdf
Yoder AD, Heckman KL. 2006. Mouse lemur phylogeography revises a model of ecogeographic constraint in Madagascar. In: Fleagle J, Lehman SM, eds. Primate
biogeography: progress and prospects. New York: Kluwer
Press, 255–268.
Yoder AD, Irwin JA, Payseur BA. 2001. Failure of the ILD
to determine data combinability for slow loris phylogeny.
Systematic Biology 50: 408–424.
Yoder AD, Nowak MD. 2006. Has vicariance or dispersal
been the predominant biogeographic force in Madagascar?
Only time will tell. Annual Review of Ecology Evolution and
Systematics 37: 405–431.
Yoder AD, Olson LE, Hanley C, Heckman KL,
Rasoloarison R, Russell AL, Ranivo J, Soarimalala V,
Karanth KP, Raselimanana AP, Goodman SM. 2005. A
multidimensional approach for detecting species patterns in
Malagasy vertebrates. Proceedings of the National Academy
of Sciences of the United States of America 102: 6587–6594.
Zapata F. 2010. Phylogenetics and diversification of Escallonia (Escalloniaceae). PhD Dissertation, University of Missouri, St. Louis, MO.
Zhu G, Bu W, Gao Y, Liu G. 2012. Potential geographic
distribution of brown marmorated stink bug invasion
(Halyomorpha halys). PLoS ONE 7: e31246.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher’s web-site:
Figure S1. Biogeographical zones recognized under the ecogeographical constraint (ECH), riverine barrier
(RBH), watershed contraction (WCH) and current climate (CCH) hypotheses mapped onto the phylogenetic tree.
Figures S2–S31. Results of principal component analyses of bioclimatic and morphological data carried out in
phylogenetic (Figs S2–S17) and geographical (Figs S18–S31) contexts. BIO, FL, FR, VEG, VEG + FL and
VEG + FR are bioclimatic, flower, fruit, vegetative, vegetative and flower, and vegetative and fruit variables,
respectively. Number (2) after an abbreviation refers to a second analysis with the same set of variables, but
excluding the species indicated in brackets.
Table S1. List of qualitative characters used to separate specimens into groups.
Table S2. List of operational taxonomic units (OTUs) included in this study with corresponding GenBank
accession numbers for the different loci.
Table S3. Statistics of the alignment.
Table S4. List of quantitative characters measured for morphometric analyses.
Table S5. Elevation and bioclimatic variables used in the principal component analyses and discriminant
function analyses.
© 2013 The Linnean Society of London, Botanical Journal of the Linnean Society, 2014, 174, 141–161