Content uploaded by Hanno Schaefer
Author content
All content in this area was uploaded by Hanno Schaefer on Oct 14, 2015
Content may be subject to copyright.
AMERICAN JOURNAL OF BOTANY 102 ( 10 ): 1 – 11 , 2015 ; http://www.amjbot.org/ © 2015 Botanical Society of America • 1
AMERICAN JOURNAL OF BOTANY
RESEARCH ARTICLE
e angiosperm ora of the Macaronesian region (comprising the
oceanic archipelagos of Azores, Madeira, Canaries, and Cape
Verdes) exhibits striking examples of evolutionary radiations in
groups such as the Aeonium Webb and Berthel. clade ( Mort et al.,
2007 ), Argyranthemum Webb ( Francisco-Ortega et al., 1996 ), and
Echium L. ( Böhle et al., 1996 ). At least in part as a consequence of
such evolutionary radiations, most Macaronesian endemic species
or subspecies of seed plants are restricted to only one island or ar-
chipelago ( Reyes-Betancort et al., 2008 ; Carine and Schaefer, 2010 ),
with many endemics occupying very restricted distributions within
a single island ( Reyes-Betancort et al., 2008 ). Only 10 Macarone-
sian endemic taxa may be considered widespread in so far as they
have distributions spanning three of the archipelagos with no taxo-
nomic di erences recognized. Six have Madeira–Canaries–Cape
Verdes distributions ( Asparagus scoparius Lowe, Dactylis smithii
Link subsp. hylodes Parker, Fumaria montana Schmidt, Lolium
canariense Steud., Patellifolia procumbens (C.Sm. ex Hornem.)
A.J.Scott, Ford-Lloyd and J.T.Williams, and Wahlenbergia lobelioi-
des (L.f.) Link subsp. lobelioides ), while four have distributions
spanning the Azores, Madeira, and the Canaries. Of these, three are
1 Manuscript received 22 May 2015; revision accepted 10 September 2015.
2 Department of Life Sciences, e Natural History Museum, Cromwell Road, London,
SW7 5BD;
3 Technische Universitaet Muenchen, Plant Biodiversity, Emil-Ramann Str. 2, D-85354
Freising, Germany;
4 Centro de Ciências da Vida, Universidade da Madeira, Campus da Penteada, 9000-390
Funchal, Portugal;
5 Jardín de Aclimatación de La Orotava (ICIA), C/ Retama n° 2 38400 Puerto de La Cruz,
Tenerife, Spain;
6 Department of Biology, Ecology and Evolution, Liège University, Bât. B22, Boulevard du
Rectorat 27, 4000 Liège, Belgium; and
7 Departamento de Biología Vegetal, Universidad de La Laguna, C/ Astrofísico Francisco
Sánchez, s/n 38071 Tenerife, Spain
8 Author for correspondence (e-mail: m.carine@nhm.ac.uk)
doi:10.3732/ajb.1500238
Are there any widespread endemic owering plant species in
Macaronesia? Phylogeography of Ranunculus cortusifolius 1
Bethany R. M. Williams 2 , Hanno Schaefer 3 , Miguel Menezes De Sequeira 4 , J. Alfredo Reyes-Betancort 5 , Jairo Patiño 6,7 , and Mark A. Carine 2,8
PREMISE OF THE STUDY: Oceanic island endemics typically exhibit very restricted distributions. In Macaronesia, only one endemic angiosperm species,
Ranunculus cortusifolius , has a distribution spanning the archipelagos of the Azores, Madeira, and Canaries. Earlier work suggested possible di erences
between archipelagos and the multiple origins of the species. This paper tests the hypothesis that R. cortusifolius is a single widespread Macaronesian
endemic species with a single origin.
METHODS: Chloroplast ( matK-trnK , psbJ-petA ) and ITS sequences were generated from across the distribution of R. cortusifolius . Relationships were investi-
gated using Bayesian inference and divergence times estimated using BEAST. Infraspeci c variation was investigated using statistical parsimony. The
general mixed Yule-coalescent model (GMYC) was further used to identify putative species boundaries based on maternally inherited plastid data.
KEY RESULTS: The hypothesis of multiple independent origins of R. cortusifolius is rejected. Divergence of the R. cortusifolius lineage from a western Medi-
terranean sister group in the late Miocene is inferred. Distinct genotypes were resolved within R. cortusifolius that are endemic to the Azores, Madeira, and
the Canaries. Four to ve putative species were delimited by di erent versions of the GMYC model.
CONCLUSION: Ranunculus cortusifolius is the result of a single colonization of Macaronesia. The large distances between archipelagos have been e ective
barriers to dispersal, promoting allopatric diversi cation at the molecular level with diversi cation also evident within the Canaries. Isolation has not been
accompanied by marked morphological diversi cation, which may be explained by the typical association of R. cortusifolius with stable and climatically
bu ered laurel forest communities.
KEY WORDS Azores; Canaries; divergence times; general mixed Yule-coalescent model; laurisilva; Madeira; Ranunculaceae; Ranunculus cortusifolius ;
speciation
http://www.amjbot.org/cgi/doi/10.3732/ajb.1500238The latest version is at
AJB Advance Article published on October 9, 2015, as 10.3732/ajb.1500238.
Copyright 2015 by the Botanical Society of America
2 • AMERICAN JOURNAL OF BOTANY
subspecies: Dracaena draco L. subsp. draco (although the native
status of this species in the Azores is still under debate; see Schaefer,
2003 ), Carex viridula Michx. subsp. cedercreutzii (Fagerstr.) B.
Schmid, and Rumex bucephalophorus L. subsp. canariensis (Steinh.)
Rech.f. Only one, the giant buttercup, Ranunculus cortusifolius
Willd., is recognized at speci c rank. e latter is the focus of this
paper.
Ranunculus cortusifolius is typical of damp habitats in Macaro-
nesia, o en occurring in association with edges and clearings of
laurisilva ( Lowe, 1857 ; Goepfert, 1976 ; Costa et al., 2012 ). It is wide-
spread in the region, occurring on all of the Canary Islands, on Ma-
deira, and on all of the Azorean islands, except Graciosa and Santa
Maria ( Jardim and Menezes de Sequeira, 2008 ; Izquierdo et al.,
2010 ; Silva et al., 2010 ).
e species was rst described from the Canaries by Willdenow
(1809) and subsequently recorded from Madeira by Lowe (1831)
who believed there to be su cient di erences between the Canar-
ian and Madeiran plants to justify the recognition of the latter as a
separate species ( R. grandifolius Lowe) based on size, pubescence,
and di erences in in orescence structure. However, this distinction
was not generally accepted by subsequent authors (e.g., Hooker,
1852 ; Masferrer, 1880 ). In the Azores, R . cortusifolius was rst col-
lected by Hochstetter ( Seubert and Hochstetter, 1843 ). Watson
(1870) used the name R. grandifolius to refer to plants from the ar-
chipelago and in so doing suggested a closer a nity with Madeiran
rather than Canarian plants although he also noted that Azorean
specimens appeared to be far more densely pubescent than their
Madeiran counterparts.
From a molecular perspective, a preliminary analysis by Schaefer
(2003) , using the nuclear ribosomal internal transcribed spacer re-
gion (ITS), suggested considerable variation at the molecular
level between three R. cortusifolius specimens from Tenerife in the
Canary Islands and three from the Azores (two from Flores and one
from Terceira). Analyzing these sequences in conjunction with a
single sequence of R. creticus L., an eastern Mediterranean species
previously considered closely related to R. cortusifolius based
on morphological similarities (e.g., Lowe, 1857 ; Bramwell and
Richardson, 1973 ), Schaefer (2003) found that Azorean specimens
were more similar to R. creticus than to Canarian specimens. is
led to the suggestion that R. cortusifolius might not be a single spe-
cies. However, the sampling of both individuals across the distribu-
tion of R. cortusifolius , and of continental outgroup taxa, was
inadequate to obtain a robust answer.
More recently, wider phylogenetic studies of Ranunculus by Paun
et al., (2005) and Emadzade et al., (2011) have placed R. cortusifo-
lius within a clade of species from the western Mediterranean, with
R. creticus resolved in a di erent clade. Paun et al. (2005) , using the
split of Ranunculus (Ranunculeae) and Xanthorhiza (Dichocar-
peae) to calibrate their phylogeny, suggested that R. cortusifolius
diverged from its sister group 0.8 million years ago (Ma). Using an
expanded data set and more calibration points, Emadzade and
Hörandl (2011) estimated the split to have occurred at 2.8 Ma (±
2.3 Ma). Signi cantly however, Paun et al. (2005) , Emadzade et al.
(2011) , and Emadzade and Hörandl (2011) all used just a single ac-
cession of R. cortusifolius in their respective analyses (from Tenerife in
each case), leaving the potential multiple origins of R. cortusifolius
suggested by Schaefer’s (2003) preliminary study unresolved.
is paper reports phylogeographic and model selection analy-
ses of plastid and nuclear sequences that examine the status and
evolution of R. cortusifolius . Speci cally, we aimed to (1) test the
hypothesis that multiple colonization and convergent evolution in
the Macaronesian islands explain the evolution of this species and
(2) determine whether molecular data suggest unrecognized diver-
si cation and speciation within Macaronesia. Finally, we discuss
factors that may help to explain the diversity patterns observed.
MATERIALS AND METHODS
Sampling — Forty-three accessions of Ranunculus cortusifolius
were included in the study: 21 from the Canary Islands, 11 from
Madeira, and 11 from the Azores. ese accessions were chosen so
as to represent the geographical spread of R. cortusifolius localities
in Macaronesia. Accessions from all islands with known R. cortu-
sifolius populations except São Miguel (Azores) and La Palma (Ca-
nary Islands) were included. Sequences from 50 Ranunculus species
of the IX Tethyan clade of Emadzade et al. (2011) (which includes
R. cortusifolius ) were used to provide a phylogenetic framework to
investigate whether R. cortusifolius is an exclusive lineage. To inves-
tigate divergence times of clades within R. cortusifolius and of
R. cortusifolius from its sister clade, matK sequences from 201 Ra-
nunculus species sequenced by Emadzade et al. (2011) were used,
together with Clematis ganpiniana (H.Lév. & Vaniot) Tamura and
Clematis patens C.Morren & Decne. as outgroups. e selection of
taxa follows Emadzade and Hörandl (2011) . Voucher information
and GenBank accession numbers are listed in Appendix S1 (see
Supplemental Data with the online version of this article).
DNA extraction, ampli cation, and sequencing — DNA extraction
and ampli cation of the ITS region followed the protocol of Carine
et al. (2004) . e psbJ-petA region was ampli ed using the primers
of Shaw et al. (2007) , and the matK/trnK region was ampli ed using
primers trn k710f and trn K3r ( Paun et al., 2005 ). e cycling condi-
tions detailed in Shaw et al. (2007) were used for both plastid re-
gions. BSA was added at 4% per volume to samples that proved
di cult to amplify. PCR products were puri ed using QIAquick
PCR Puri cation Kit (Qiagen, Crawley, UK) according to the man-
ufacturer’s instructions and cycle sequencing using the PCR prim-
ers was performed using Big Dye Terminator Ready Reaction Mix
(Applied Biosystems, Carlsbad, California, USA). Due to the length
of the matK/trnK region the internal primer mat K700f, designed by
Paun et al. (2005) , was used for sequencing in addition to the PCR
primers. Sequences were generated on Applied Biosystem 3730xl
DNA Analyzer automated sequencer (Applied Biosystems, Foster
City, California, USA).
Lasergene Navigator (DNAStar, Madison, Wisconsin, USA) was
used to edit and assemble sequences, and alignments were per-
formed by eye using the similarity criterion ( Simmons, 2004 ) and
the program Se-Al (version 2.0, Rambaut, 1996 ). Gaps were coded
as missing data, and indels that were shared by more than one
taxon were coded as single characters following the “simple indel
coding method” ( Simmons and Ochoterena, 2000 ). Four data sets
were assembled.
Data Set One— Designed to test whether R. cortusifolius is an exclu-
sive lineage, the data set comprised ITS and plastid sequences of
samples of R. cortusifolius together with those of the 50 species of
the IX Tethyan clade de ned by Emadzade et al. (2011) . Ranuncu-
lus amblyolobus Boiss. & Hohen., R. cappadocicus Willd., R. buhsei
Boiss., R. breyninus Crantz, and R. brachylobus Boiss. and Hohen.
OCTOBER 2015 , VOLUME 102 • WILLIAMS ET AL. —PHYLOGEOGRAPHY OF RANUNCULUS CORTUSIFOLIUS • 3
were defined as outgroups following the topology of Emadzade
et al. (2011) . e data set is provided in Appendix S2 (see online
Supplemental Data).
Data Set Two (plastid sequences) and Data Set ree (ITS sequences)—
Used to investigate patterns of sequence diversity within R. cortusi-
folius using statistical parsimony, the data sets are provided in
online Appendices S3 and S4, respectively.
Data Set Four— is data set was used to investigate divergence
times and to test for speciation vs. population-level processes using
the general mixed Yule-coalescent model described by Pons et al.,
(2006 ; herea er GMYC model). It included matK sequences for a
total of 203 samples, drawn largely from Emadzade et al. (2011) and
Emadzade and Hörandl (2011) but with one R. cortusifolius se-
quence included for each of the matK / trnK haplotypes identi ed.
e data set is provided in online Appendix S5.
Data analyses — Phylogenetic analyses were performed using Data
Set One. e ITS and plastid data sets were analyzed separately as
well as combined, so that their topologies could be compared and
their congruence assessed.
e Akaike information criterion (AIC; Akaike, 1974 ) was used
to assess the best model of sequence evolution in the program
MrModeltest 2.2 ( Nylander, 2004 ) with the GTR+I model chosen
for the psbJ-petA partition and the GTR+I+G model chosen for the
ITS and matK/trnK partitions. Bayesian analysis ( Yang and
Rannala, 1997 ) was undertaken using the program Mr.Bayes 3.1.2
( Ronquist and Huelsenbeck, 2003 ). Two independent runs, each
comprising four Markov chains were run simultaneously for
1 000 000 generations, with sampling every 1000 generations. e
convergence of parameter estimates was judged by the split vari-
ance being less than 0.05 and the rst 250 trees were then discarded
as “burn-in”.
Relationships within R. cortusifolius were investigated using
Data Sets Two and ree, performing statistical parsimony analysis
( Templeton et al., 1992 ) of the R. cortusifolius accessions using the
program TCS 1.21 ( Clement et al., 2000 ). Samples for which only
incomplete short fragments could be sequenced were excluded
from the analysis. ey were subsequently grouped into genotypes
with other specimens sharing identical sequence.
e estimated divergence time between R. cortusifolius and its
sister clade as well as between clades within R. cortusifolius was in-
vestigated using Data Set Four in the program BEAST v1.7.5
( Drummond et al., 2012 ). ree of the calibration points from
Emadzade and Hörandl (2011) were used: the split between Ranun-
culus and Clematis (normal prior distribution around a mean of
46.6 Ma with a standard deviation of 2); the minimum age of Myo-
surus (exponential prior distribution with an o set of 23.0 Ma and
a mean of 1.0 Ma); and the divergence between R. carpaticola and
R. notabilis (0.914 Ma; see Emadzade and Hörandl [2011] for fur-
ther details of calibration points). e age of the Juan Fernández
Islands, also used by Emadzade and Hörandl (2011) , was omitted
from the current study since an island endemic taxon such as the
Juan Fernández endemic R. caprarum could have diverged from its
mainland relative long before colonizing the islands ( Renner,
2005 ). e GTR model with gamma distribution and invariant sites
had the best AIC score for the matK data set using MrModeltest 2.2
( Nylander, 2004 ). A relaxed clock with lognormal distribution was
employed for the analysis. In accordance with Emadzade and
Hörandl (2011) , a Yule prior was implemented for the branching
rates. Four independent Monte Carlo Markov chains (MCMC)
were run for 50 000 000 generations, with sampling every 5000 gen-
erations. Convergence and acceptable mixing of the sampled pa-
rameters was checked using the program Tracer 1.5 ( Rambaut and
Drummond, 2003 ). Twelve thousand and ve hundred trees were
discarded as “burn-in”, and the remaining trees from each of the
four runs were combined.
We used the GMYC model to delimit putative species (for re-
view, see Fujisawa and Barraclough, 2013 ) on the basis of the results
of the phylogenetic analysis of Data Set Four (the chloroplast [cp]
data). is ML method diagnoses species-level clades by identifying
the shi in diversi cation rate associated with the change from neu-
tral coalescent processes within species to branching among genetic
clusters that re ect the timing of speciation events ( Yule, 1924 ).
Speci cally, the null model assumes that the entire sample derives
from a single population undergoing a single coalescent process,
whereas the GMYC model classi es the observed branching time
intervals into two categories, as the result of either inter- or intra-
speci c processes of lineage sorting. A log-likelihood ratio test is
then used to assess which model provides a better t. More recently,
this method was extended to allow multiple thresholds to account
for variation in the depth of the coalescent–speciation transition
along individual branches of a phylogenetic tree ( Monaghan et al.,
2009 ). Although the multiple threshold version may show a certain
tendency to over-splitting, its use has been recommend to explore
delimitation changes when the assumptions of the single threshold
needs to be relaxed and when the sampling of individuals across a
representative occupancy range may be achieved ( Fujisawa and
Barraclough, 2013 ), as might be the case of Ranunculus across the
Macaronesian islands where di erent times have elapsed since their
formation. We applied the GMYC model to the maximum clade
credibility tree obtained from the divergence time analysis (above).
Following Ahrens et al. (2007) , we ran the analyses with identical
haplotypes reduced to single entities. In addition, we ran the analy-
ses without removing sister groups to the R. cortusifolius clade,
which were used as outgroups ( Powell et al., 2011 ). e set of best
GMYC models was determined using a modi ed Akaike informa-
tion criterion score corrected for small sample size (AICc; for de-
tails see Powell, 2012 ). e GMYC package also produced a
lineage-through-time (LTT) plot, which we visually evaluated for
changes in branching rate. Single and multiple-threshold GMYC
models were run using the SPLITS package ( Ezard et al., 2009 ) for
R ( R Development Core Team, 2011 ).
We also used a Bayesian version of the GMYC model (bGMYC;
Reid and Carstens, 2012 ) that accounts for error in phylogeny esti-
mation and uncertainty in model parameters (
Monaghan et al.,
2009 ). Following Reid and Carstens (2012) , we randomly sampled
100 trees from the posterior distribution of the BEAST analyses
and ran the GMYC analyses on each tree for 50 000 generations,
with a sampling frequency and discarding the first 10 000 gen-
erations as “burn in”. These analyses were performed using the
bGMYC package ( Reid and Carstens, 2012 ) for R ( R Development
Core Team, 2011 ).
RESULTS
The combined plastid matrix ( matK/trnK and psbJ-petA ) had a
total length of 2296 nucleotides with 43 variable positions and these
4 • AMERICAN JOURNAL OF BOTANY
distinguished 17 haplotypes within R. cortusifolius . Ten of the hap-
lotypes were found in the Canary Islands ( Fig. 1 ). One cluster of six
haplotypes was resolved in the Canaries centered on haplotype 7H.
A second cluster, separated by eight mutations from the former,
comprised three Tenerife specimens from the Anaga massif, which
each had a unique haplotype (9H, 6H, 10H). One specimen from La
Gomera (4H) failed to connect to the remainder of the network.
e Madeiran and Canarian specimens were separated from each
other by a minimum of seven mutations. Within Madeira, three
genotypes were identi ed (11H–13H), separated from adjacent hap-
lotypes by one or two mutations. e four Azores genotypes (15H–
17H) were similarly separated from each other by a minimum of
one or two mutations. Azorean and Madeiran genotypes were sep-
arated by a minimum of four mutations ( Fig. 1 ).
e ITS region provided less variation. e 612-bp fragment
included 11 variable sites, which de ned ve ribotypes within R.
cortusifolius . No evidence of ITS paralogy was observed. In the
statistical parsimony network ( Fig. 2 ), a cluster was found com-
prising three ribotypes from the Canaries, separated by one or
two mutations. One of the Canarian ribotypes occurred on all is-
lands. e remaining two were found only in specimens from La
Gomera. e Canary Island cluster was separated from the two
ribotypes found in Madeiran and Azorean specimens by seven
mutations. One ribotype was shared by all Azorean and most
Madeiran samples. e second ribotype, distinguished from the
former by a single mutation, was restricted to two of the 11 indi-
viduals sampled from Madeira.
e lower levels of variation in the ITS region were re ected in a
less resolved and less well-supported topology in both parsimony
and Bayesian analyses. Nevertheless, no well-supported topological
incongruence between the ITS and plastid trees was observed (on-
line Appendix S6), and the two data sets were therefore combined.
Bayesian analysis of the combined data set strongly supports the
monophyly of R. cortusifolius accessions ( Fig. 3 , Bayesian posterior
probability (herea er PP) = 1) and a sister group relationship be-
tween R. cortusifolius and a clade comprising R. spicatus Desf., R.
gregarius Brot., R. bullatus L., R. olissiponensis Pers., R. pseudomille-
foliatus Grau and R. paludosus Poir. (PP = 1), the latter in keeping
with the results of Emadzade et al. (2011) . Within R. cortusifolius ,
there is strong support for a Canary Island clade (PP = 1) and a
Madeiran-Azorean clade (PP = 1). Within the former, a subclade
comprising the three accessions from the Anaga massif together
with a cultivated specimen used by Hörandl et al. (2005) and
Emadzade et al. (2011) was resolved with strong support (PP = 1); a
second subclade comprising all other Canarian accessions was also
strongly supported (PP = 1). Within the Madeiran-Azorean clade,
accessions from Madeira were resolved as monophyletic (PP = 0.96).
In the BEAST analysis, the mean evolutionary rate was 1.02 ×
10
−3 substitutions per site per million years (95% HPD: 1.2878 ×
10
−3 – 0.75923 × 10
−3 ) for the matK data set. e Yule process
birth rate was 0.169 (95% HPD: 0.1276–0.2126). e chronogram
is provided in online Appendix S7. e split between R. cortusifo-
lius and its sister clade is estimated to have occurred 7.43 Ma
(95% HPD: 4.38–10.75 Ma; Table 1 , see also Fig. 4A and B ). e
FIGURE 1 Plastid haplotypes in Ranunculus cor tusifolius . (A) Distribution of haplotypes by island in the Azores and Madeira (above) and Canary Islands
(below). The size of pie charts is proportional to the number of specimens sampled. (B) Haplotype network. Each line represents a single base pair
di erence. Haplotype numbers are referred to in the text.
OCTOBER 2015 , VOLUME 102 • WILLIAMS ET AL. —PHYLOGEOGRAPHY OF RANUNCULUS CORTUSIFOLIUS • 5
FIGURE 2 ITS ribotypes in Ranunculus cortusifolius . (A) Distribution of ribotypes by island in the Azores and Madeira (above) and Canary Islands (below).
The size of pie charts is proportional to the number of specimens sampled. (B) Ribotype network. Each line represents a single base pair di erence.
Canarian clade split from the Madeiran-Azorean clade 5.34 Ma
(95% HPD: 2.83–8.09 Ma). e clade only found in Tenerife di-
verged from the widespread Canary Island clade 4.64 Ma (95%
HPD: 2.31–7.20 Ma), while the split between the Madeira and
Azores clade is estimated to have occurred 2.87 Ma (95% HPD:
0.93–4.94 Ma).
Single and multiple threshold GMYC methods were applied to
the plastid data. For both versions, the likelihood of the GMYC
model ( L GMYCsingle = 7.615125 and L GMYCmultiple = 7.971705) showed a
marginally signi cant better t to the data than the null model of
uniform coalescent branching rates (L
0 = 2.795538; P = 0.049 and
P = 0.047, respectively). e multiple threshold GMYC model
yielded a marginally better t than the single threshold GMYC
model, as determined by AICc scores. Consequently, we refer to
multiple threshold results herea er, which delimited four putative
genetic entities within the R . cortusifolius clade: one on the Azores,
one on Madeira and two on the Canaries ( Fig. 4B ), corresponding
to the groups identi ed in the phylogenetic analysis. e bGMYC
species delimitation provided a less conservative estimate, suggest-
ing a total of ve putative species ( Fig. 4B ). e LTT plot ( Fig. 4C )
showed an approximately steady increase in lineage accumulation
with a sharp increase in diversi cation rate toward the present.
DISCUSSION
Origins of Ranunculus cortusifolius — Both MP and Bayesian anal-
ysis gave high support ( Fig. 3 , 96% BS and 1 PP) for the monophyly
of Ranunculus cortusifolius accessions, consistent with a single col-
onization event into Macaronesia. is pattern is shared by the
great majority of genera that contain Macaronesian endemics;
while multiple independent congeneric colonizations have been
documented in the endemic ora, they are few in number ( Díaz-
Pérez et al., 2008 ; Carine et al., 2010 ). Although no specimens from
São Miguel (Azores) or La Palma (Canary Islands) were included in
this study, the sampling otherwise included accessions from all
other islands on which R. cortusifolius is distributed. Schaefer’s
(2003) study, with a much more limited sample size, suggested R.
cortusifolius may not be an exclusive lineage based on marked dif-
ferences in the ITS sequences of samples from the Canaries and the
FIGURE 3 Bayesian consensus tree from the analysis of Ranunculus cortusifolius and allied Ranunculus species based on the combined plastid and ITS
data set (Data Set 3). Bayesian posterior probabilities of more than 0.5 are given above nodes. Samples of R. cortusifolius are pre xed with either Ca-
nary, Madeira, or Azores depending on the archipelago from where they were collected and are given an island code as follows: EH = El Hierro, LG =
La Gomera, T = Tenerife, GC = Gran Canaria, FU = Fuerteventura, LA = Lanzarote, CO = Corvo, FL = Flores, FA = Faial, PI = Pico, SJ = San Jorge, TE = Ter-
ceira. Numbers refer to sample numbers listed in Appendix S1.
→
6 • AMERICAN JOURNAL OF BOTANY
OCTOBER 2015 , VOLUME 102 • WILLIAMS ET AL. —PHYLOGEOGRAPHY OF RANUNCULUS CORTUSIFOLIUS • 7
Azores. is suggestion is not supported by the current study al-
though we do nd marked di erences at the molecular level in
plants from these archipelagos.
Molecular variation within the R. cortusifolius lineage — Bayesian
and MP analysis of the combined plastid and nuclear data set both
indicated that R. cortusifolius specimens from the Canary Islands
form a clade ( Fig. 3 ; PP = 1) that is sister to a clade comprising ac-
cessions from Madeira and the Azores (PP = 1). e sister group
relationship receives good support (PP = 1). An unexpected nding
from this study was the resolution of Canarian accessions into two
strongly supported subclades ( Fig. 3 ). While one of the Canarian
clades comprises accessions from all islands sampled, the other is
restricted to plants from the Anaga massif in Tenerife. Together
with the Teno and Roque del Conde massifs, Anaga existed as an
independent island until 3 Ma ( Reyes-Betancort et al., 2008 ).
e two Canarian clades constitute from two to three of the four
to ve putative species delimited in the analysis of the plastid data
using the GMYC and bGMYC models ( Fig. 4 ), respectively. Within
the Madeira-Azores clade, phylogenetic analysis, Madeiran acces-
sions were resolved as monophyletic (PP = 0.96) but Azorean ac-
cessions were unresolved ( Fig. 3 ). Nevertheless, the plastid statistical
parsimony analysis ( Fig. 1 ) shows that all haplotypes were restricted
to a single archipelago and that haplotypes within an archipelago
were more closely related to each other than they were to haplo-
types from other archipelagos. Furthermore, the GMYC and bG-
MYC models identi ed distinct Azorean and Madeiran entities
based on the plastid data.
ese results are largely consistent with morphological di er-
ences alluded to by earlier authors. us, Lowe (1831) distinguished
Madeiran and Canarian plants, while Watson (1870) grouped Ma-
deiran and Azorean plants although drawing attention to morpho-
logical di erences between them. In general, plants from the
Canaries, together with high altitude plants from Madeira tend to
be smaller in stature than other plants in the complex, a trend also
mirrored in the basal leaf size. Leaf pubescence also di ers between
the three archipelagos: Azorean plants are the most densely pubes-
cent, followed by Madeiran plants, whereas Canarian plants are
typically only sparsely pubescent to subglabrous, although excep-
tions do occur (notably on El Hierro where small, densely pubes-
cent plants occurring on rocks have been described as a distinct
variety). Di erences are also apparent in the nature of the in ores-
cence that is corymbose in Azorean plants ( Fig. 5A, B ) but not, or
rarely so in Madeira and the Canaries ( Fig. 5C–E, F and G , respec-
tively). In light of the data presented here, a more narrow circum-
scription of taxa, supporting the taxonomic concepts and observations
of earlier workers such as Lowe (1831) and Watson (1870) would
appear to be appropriate although taxonomic changes would be
premature pending a full revision of the complex. It is notable that
no morphological di erences have been identi ed to date between
the Canary Island groups identi ed.
Evolution of R. cortusifolius — e results reported here agree with
the studies of Paun et al. (2005) and Emadzade and Hörandl (2011) ,
which placed R. cortusifolius within a largely western Mediterra-
nean clade. While the estimated age of the split between R. cortusi-
folius and its sister clade inferred in this analysis (7.43 Ma, 95%
HPD: 4.38–10.76; Table 1 ) is earlier than that proposed by Emadzade
and Hörandl (2011 : 2.8 Ma), the con dence intervals of the respec-
tive analyses overlap.
Paun et al. (2005) noted that the placement of R. cortusifolius in a
western Mediterranean clade indicates that the hypothesized close
relationship with the eastern Mediterranean R. creticus (e.g., Lowe,
1857 ), which had been considered evidence of the relictual nature of
R. cortusifolius ( Bramwell and Richardson, 1973 ), is most likely due to
a preference for similar shady habitats resulting in homoplasious leaf
shape. e putative relictual status of R. cortusifolius is therefore no
longer supported by a transcontinental disjunct sister species rela-
tionship. Nevertheless, it is notable that the divergence time analysis
places the split of R. cortusifolius from its closest relatives in the late
Miocene ( Table 1 ), a timing consistent with the hypothesis that the R.
cortusifolius lineage might have survived in Macaronesia as a relict of
the Palaeotropical geo ora that went extinct on the continent as a
consequence of the neogene climatic deterioration (e.g., Fernández-
Palacios et al., 2011 ). Including in the analysis additional western
Mediterranean and North African species that have not been sam-
pled to date would allow this hypothesis to be further tested.
Divergence time estimates have been established for several
Macaronesian endemic lineages that, in keeping with R. cortusifo-
lius , have distributions spanning multiple archipelagos. ese in-
clude Echium (7.93 ± 3.58 Ma), the Aeonium alliance (18.83 ±
1.81), Sonchus (13.20 ± 5.51Ma), Sideritis (11.92 ± 5.87 Ma), and
Crambe (14.89 ± 5.52) ( Kim et al., 2008 ). e age of the R. cortusi-
folius lineage is broadly similar to that of the Macaronesian en-
demic Echium clade but younger than the others considered
above. However, the divergence of Madeiran and/or Azorean
plants is hypothesized to have occurred earlier in R. cortusifolius
(5.35 Ma) than in any of the lineages considered except the Aeo-
nium alliance ( Aeonium alliance 8.47 ± 1.92 and 7.13 ± 1.89; Son-
chus 2.717 ± 2.15; Echium 0.282 ± 0.36; Sideritis 0.380 ± 0.43;
Crambe 1.839 ±1.82; Kim et al., 2008 ). At the same time, R. cortu-
sifolius exhibits a lower rate of diversi cation than any of the other
lineages considered. us, using a simple pure birth macro-evolu-
tionary model given by the formula: D = (ln N
i − ln N
i ) / T , where
N
i equals the number of extant species and N 0 equals the number
of species at time T , (the mean estimated age of the most recent
common ancestor in millions of years; Myr), Kim et al. (2008)
gave a diversi cation rate of 0.62 species per Myr for the Macaro-
nesian Echium lineage that contains 27 species. In contrast, we
found that in the R. cortusifolius lineage, that is broadly similar in
age, the rate is 0.26 species per Myr, if the four putative species
delimited using the GMYC model are treated as distinct taxa.
While such speciation rate estimates are clearly dependent on the
reliability of the divergence time estimates used, the di erences
suggested do serve to highlight the di ering fate of island clades:
whereas some lineages have radiated extensively in expanding
their range to occupy multiple islands and archipelagos in the
Macaronesian region following colonization, others with a similar
overall distribution have not. One notable difference between
TABLE 1. Estimated divergence times for Ranunculus cortusifolius. The nodes
included below are shown in the cladogram in Fig. 2 discussed in the text.
Node, divergence
Mean estimated divergence,
Ma (95% HPD)
R. cortusifolius from sister clade 7.4 (4.3–10.7)
Canarian and Madeiran-Azorean lineages 5.3 (2.9–8.2)
Canarian lineages 4.6 (2.5–7.3)
Madeiran and Azorean lineages 2.6 (0.9–4.9)
Note: HPD = highest probability density.
8 • AMERICAN JOURNAL OF BOTANY
R. cortusifolius and species-rich and morphologically diverse lin-
eages such as the Macaronesian Echium lineage is their ecological
amplitude. As noted earlier, R. cortusifolius is restricted to rela-
tively humid habitats in Macaronesia and is commonly associated
with forest edges of the laurisilva vegetation. In contrast, Echium ,
in keeping with many of the other lineages that have radiated
extensively, has a much broader ecological range, occurring in a
diverse range of habitats from the coast to the summit of the high-
est islands and including taxa restricted to thermophilous habi-
tats. Domínguez-Lozano et al. (2010) found that thermophilous
and rocky habitats were likely centers of diversi cation in the Canary
Islands, which they attributed, in part, to a much higher degree of
FIGURE 4 Summary of dating and generalized mixed Yule-coalescent (GMYC) analyses for Ranunculus cortusifolius (See Appendix S7 for tree with
support values). (A) Ultrametric Bayesian maximum clade credibility tree of the complete plastid data set (Data Set Four) obtained using a relaxed
lognormal clock; the gray shading corresponds to the R . cortusifolius clade. (B) GMYC clusters delimited from the multiple-threshold model for
the R . cortusifolius clade are highlighted in red and labeled according to their respective archipelagos. An asterisk indicates the Bayesian general
mixed Yule coalescent (bGMYC) species delimitation, i.e. groups for which the Bayesian posterior probability that all members within a lineage
belong to one species is greater than 0.95. (C) Lineage through time plot for the R . cortusifolius clade with vertical lines indicating the points of
maximum likelihood t of the multiple-threshold GMYC model, i.e., the points of transition from interspecies (Yule) to intraspecies (coalescent)
branching events.
OCTOBER 2015 , VOLUME 102 • WILLIAMS ET AL. —PHYLOGEOGRAPHY OF RANUNCULUS CORTUSIFOLIUS • 9
FIGURE 5 Growth habit of selected plants of Ranunculus cor tusifolius from the Azores (A: Corvo, B:
Pico), from Madeira (C–E), and from the Canary Islands (F, G: Gran Canaria). Photographs by H.
Schaefer (A, B, F, G) and Miguel Sequeira (C–E).
historical fragmentation. ermophilous habitats, a relatively
young ecosystem, formed with the onset of the Mediterranean cli-
mate in the region, 2.5 Ma ( Blondel and Aaronson, 2005 ), are
likely to have experienced range contractions and expansions dur-
ing the Quaternary as a function of climate change ( Domínguez-
Lozano et al., 2010 ).
In contrast, laurisilva is thought to have been largely bu ered
from temperature changes and associated range contractions linked
to Pleistocene climatic cycles because of its mid-altitude location
( Domínguez-Lozano et al., 2010 ). e preference of R. cortusifolius
for moist habitats and its association with laurisilva habitats may
have contributed to the limited morphological
diversi cation observed in the R. cortusifolius
lineage as proposed for other laurisilva taxa
(e.g., Patiño et al., 2014 ).
In summary, the results of this study indi-
cate that between four and ve distinct lin-
eages can be identi ed in R. cortusifolius.
Distinct lineages occur on each of the archi-
pelagos on which R. cortusifolius occurs, with
two to three diverged lineages in the Canary
Islands. e data show a signature of allo-
patric speciation rather than of intraspeci c
diversi cation. At the interarchipelago level,
the large distances between archipelagos
(415–1113 km) have provided e ective barri-
ers to dispersal, promoting anagenetic diver-
sification (sensu Stuessy et al., 2006 ) at that
broad geographical scale, and a more nar-
row circumscription of species may be ap-
propriate in line with observations on
morphology by earlier authors. At the intra-
archipelago level, the failure of the R. cortusi-
folius lineage to expand into a range of
habitats, coupled with its association with
the relatively stable, climatically bu ered
laurisilva habitat, may explain the limited di-
versi cation observed. Further studies on
pollination and seed dispersal and on adap-
tive di erences between populations on the
di erent archipelagos would provide further
insights into gene ow patterns and specia-
tion processes.
ACKNOWLEDGEMENTS
e authors thank the Cabildos of the Ca-
nary Islands, Gobierno de Canarias, Região
Autónoma dos Açores, and Governo Regional
da Madeira for permission to collect plant
material. ey are also grateful to Katy Jones,
Juli Caujapé-Castells, Alison Mills, Mónica
Moura, and Arnoldo Santos-Guerra for
assistance collecting material used in this
study and Mark Simmons (Associate Editor)
and two anonymous reviewers for their
comments. is research was partially real-
ized with support from SYNTHESYS. J.P.
gratefully acknowledges nancial support
from the Belgian Funds for Scienti c Research
(FNRS) (grants 1.5036.11 and 2.4557.11) and the University of
Liège (Grant C 11/32).
LITERATURE CITED
Ahrens , D. , M. T. Monaghan , and A. P. Vogler . 2007 . DNA-based taxonomy
for associating adults and larvae in multi-species assemblages of chafers
(Coleoptera: Scarabaeidae). Molecular Phylogenetics and Evolution 44 :
436 – 449 .
Akaike , H. 1974 . A new look at the statistical model identi cation. IEEE
Transactions on Automatic Control 19 : 716 – 723 .
10 • AMERICAN JOURNAL OF BOTANY
Blondel , J. , and J. Aaronson . 2005 . Biology and wildlife of the Mediterranean
region. Oxford University Press, Oxford, UK.
Böhle , U. R. , H. H. Hilger , and W. F. Martin . 1996 . Island colonization and evo-
lution of the insular woody habit in Echium L. (Boraginaceae). Proceedings
of the National Academy of Sciences, USA 93 : 11740 – 11745 .
Bramwell , D. , and I. B. K. Richardson . 1973 . Floristic connections between
Macaronesia and the East Mediterranean region. Monographiae Biologicae
Canariensis 4 : 118 – 125 .
Carine , M. A. , S. J. Russell , A. Santos-Guerra , and J. Francisco-Ortega . 2004 .
Relationships of the Macaronesia and Mediterranean oras: Molecular evi-
dence for multiple colonizations into Macaronesia and back-colonizations
of the continent in Convulvulus (Convolvulaceae). American Journal of
Botany 91 : 1070 – 1085 .
Carine , M. A. , A. Santos Guerra , I. R. Guma , and J. A. Reyes-Betancort . 2010 .
Endemism and evolution of the Macaronesian ora. In D. M. Williams and
S. Knapp [eds.], Beyond cladistics: e branching of a paradigm, 101–124.
University of California Press, Berkeley, California, USA.
Carine , M. A. , and H. Schaefer . 2010 . e Azores diversity enigma: Why are
there so few Azorean endemic owering plants and why are they so wide-
spread? Journal of Biogeography 37 : 77 – 89 .
Clement , M. , D. Posada , and K. A. Crandall . 2000 . TCS: A computer program
to estimate gene genealogies. Molecular Ecology 9 : 1657 – 1659 .
Costa , J. C. , C. Neto , C. Aguiar , J. Capelo , D. Espírito Santo , J. Honrado , C.
Pinto-Gomes , et al. 2012 . Vascular plant communities in Portugal (conti-
nental, the Azores and Madeira). Global Geobotany 2 : 1 – 180 .
Díaz-Pérez , A. , M. Sequeira , A. Santos-Guerra , and P. Catalán . 2008 . Multiple
colonizations, in situ speciation, and volcanism-associated stepping-stone
dispersals shaped the phylogeography of the Macaronesian red fescues
( Festuca L., Gramineae) . Systematic Biology 57 : 732 – 749 .
Domínguez Lozano , F. , J. Price , R. Otto , and J. M. Fernández-Palacios . 2010 . Using
taxonomic and phylogenetic evenness to compare diversi cation in two Island
Floras. Perspectives in Plant Ecology, Evolution and Systematics 12 : 93 – 106 .
Drummond , A. J. , M. A. Suchard , D. Xie , and A. Rambaut . 2012 . Bayesian phy-
logenetics with BEAUti and the BEAST 1.7. Molecular Biology and Evolution
29 : 1969 – 1973 .
Emadzade , K. , and E. Hörandl . 2011 . Northern Hemisphere origin, transoce-
anic dispersal, and diversi cation of Ranunculeae DC. (Ranunculaceae) in
the Cenozoic. Journal of Biogeography 38 : 517 – 530 .
Emadzade , K. , B. Gehrke , H. P. Linder , and E. Hörandl . 2011 . e biogeographi-
cal history of the cosmopolitan genus Ranunculus L. (Ranunculaceae) in the
temperate to meridional zones. Molecular Phylogenetics and Evolution 58 : 4 – 21 .
Ezard , T. , T. Fujisawa , and T. G. Barraclough . 2009 . SPLITS: Species’ limits
by threshold statistics. R package version 1. R Foundation for Statistical
Computing, Vienna, Austria. Available at http://R-Forge.R-project.org/
projects/splits/ .
Fernández-Palacios , J. M. , L. Nascimento , R. Otto , J. D. Delgados , E. García-del-
Rey , J. R. Arévalo , and R. J. Whittaker . 2011 . A reconstruction of Palaeo-
Macaronesia, with particular reference to the long-term biogeography of the
Atlantic island laurel forests. Journal of Biogeography 38 : 226 – 246 .
Francisco-Ortega , J. , R. K. Jansen , and A. Santos-Guerra . 1996 . Chloroplast
DNA evidence of colonization, adaptive radiation and hybridization in the
evolution of the Macaronesian ora. Proceedings of the National Academy of
Sciences, USA 93 : 4085 – 4090 .
Fujisawa , T. , and T. G. Barraclough . 2013 . Delimiting species using single-
locus data and the Generalized Mixed Yule Coalescent (GMYC) approach:
A revised method and evaluation on simulated datasets. Systematic Biology
62 : 707 – 724 .
Goepfert , G. 1976 . Phylogenetic studies on Ranunculus cortusifolius Willd.
(Ranunculaceae), Macaronesian endemic species. Botanical Journal of the
Linnean Society 72 : 161 – 170 .
Hooker , J. D. 1852 . Ranuculus cortusifolius : Cotus-leafed ranunculus or but-
tercup . Curtis’ Botanical Magazine 78 : t.4625.
Hörandl , E. , O. Paun , J. T. Johansson , C. Lehnebach , T. Armstrong , L. Chen ,
and P. Lockhart . 2005 . Phylogenetic relationships and evolutionary traits
in Ranunculus s.l. (Ranunculaceae) inferred from ITS sequence analysis.
Molecular Phylogenetics and Evolution 36 : 305 – 327 .
Izquierdo , I. , J. L. Martín , N. Zurita , and M. Arechavaleta [eds.]. 2010 . Lista
de species silvestres de Canarias (hongo, plantas y animals terrestres) 2009.
Consejería de Medio Ambiente y Ordenación Territorial, Gobierno de
Canarias, Tenerife, Spain.
Jardim , R. , and M. Menezes de Sequeira . 2008 . Lista das plantas vasculares
(Pteridophyta and Spermatophyta) . In P. A. V. Borges, C. Abreu, A. M. F.
Aguiar, P. Carvalho, R. Jardim, I. Melo, P. Oliveira, et al. [eds.], A list of
the terrestrial fungi, ora and fauna of Madeira and Selvagens archipelagos,
179–208. Direcção Regional do Ambiente da Madeira and Universidade dos
Açores, Funchal and Angra do Heroísmo, Portugal.
Kim , S.-C. , M. R. McGowen , P. Lubinsky , J. C. Barber , M. E. Mort , and A.
Santos-Guerra . 2008 . Timing and tempo of early and successive adaptive
radiations in Macaronesia. PLoS One 3 : e2139 .
Lowe , R. T. 1831 . Primitae faunae et orae Maderae et Portus Sancti.
Transactions of the Cambridge Philosophical Society 4 : 1 – 70 .
Lowe , R. T. 1857 . On Ranunculus creticus , R. cortusaefolius , and R. grandifolius
of authors. Journal of Botany and Kew Gardens 9 : 65 – 75 .
Masferrer , R. 1880 . Recuerdos botánicos de Tenerife. Datos para el estudio de la
ora canaria. Anales de la Sociedad Española de Historia Natural 9 : 309 – 369 .
Monaghan , M. T. , R. Wild , M. Elliot , T. Fujisawa , M. Balke , D. J. G. Inward , D.
C. Lees , et al . 2009 . Accelerated species inventory on Madagascar using co-
alescent-based models of species delineation. Systematic Biology 58 : 298 – 311 .
Mort , M. E. , D. E. Solits , A. Santos-Guerra , and J. Francisco-Ortega . 2007 .
Physiological evolution and association between physiology and growth
form in Aeonium (Crassulaceae). Taxon 56 : 453 – 464 .
Nylander , J. A. A. 2004 . MrModeltest v2. 2. Program distributed by the author.
Evolutionary Biology Centre, Uppsala University, Uppsala, Sweden.
Patiño , J. , M. A. Carine , J. M. Fernández-Palacios , R. Otto , H. Schaefer , and
A. Vanderpoorten . 2014 . e anagenetic world of spore-producing land
plants. New Phytologist 201 : 305 – 311 .
Paun , O. , C. Lehnebach , J. T. Johansson , P. Lockhart , and E. Hörandl . 2005 .
Phylogenetic relationships and biogeography of Ranunculus and allied
genera (Ranunculaceae) in the Mediterranean region and in the European
Alpine System. Taxon 54 : 911 – 930 .
Pons , J. , T. G. Barraclough , J. Gomez-Zurita , A. Cardoso , D. P. Duran , S. Hazell ,
S. Kamoun , et al . 2006 . Sequence-based species delimitation for the DNA
taxonomy of undescribed insects. Systematic Biology 55 : 595 – 609 .
Powell , J. R. 2012 . Accounting for uncertainty in species delineation during
the analysis of environmental DNA sequence data. Methods in Ecology and
Evolution 3 : 1 – 11 .
Powell , J. R. , M. T. Monaghan , M. Opik , and M. C. Rillig . 2011 . Evolutionary
criteria outperform operational approaches in producing ecologically rel-
evant fungal species inventories. Molecular Ecology 20 : 655 – 666 .
R Development Core Team . 2011 . R: A language and environment for sta-
tistical computing. Website http://R-Forge.R-project.org . R Foundation for
Statistical Computing, Vienna, Austria.
Rambaut , A. 1996 . Sequence alignment editor version 1.0, alpha 1 . University
of Oxford, Oxford, UK.
Rambaut , A. , and A. Drummond . 2003 . Tracer version 1.4. Available at http://
evolve.zoo.ac.uk/ .
Reid , N. M. , and B. C. Carstens . 2012 . Phylogenetic estimation error can de-
crease the accuracy of species delimitation: A Bayesian implementation of
the general mixed Yule-coalescent model . BMC Evolutionary Biology 12 :
196 .
Renner , S. 2005 . Relaxed molecular clocks for dating historic plant dispersal
events. Trends in Plant Science 10 : 550 – 558 .
Reyes-Betancort , J. A. , A. Santos-Guerra , I. R. Guma , C. J. Humphries , and M.
A. Carine . 2008 . Diversity, rarity and the evolution and conservation of
the Canary Islands endemic ora. Anales del Jardin Botanico de Madrid 65 :
25 – 45 .
Ronquist , F. , and J. P. Huelsenbeck . 2003 . MrBayes 3: Bayesian phylogenetic
inference under mixed models. Bioinformatics 19 : 1572 – 1574 .
Schaefer , H. 2003 . Chorology and diversity of the Azorean ora. Dissertationes
Botanicae 374 : 1 – 130 .
Seubert , M. , and C. Hochstetter . 1843 . Übersicht der Flora der azorischen
Inseln. Wiegmann’s Archiv für Naturgeschichte 9 : 1 – 24 .
OCTOBER 2015 , VOLUME 102 • WILLIAMS ET AL. —PHYLOGEOGRAPHY OF RANUNCULUS CORTUSIFOLIUS • 11
Shaw , J. , E. B. Lickey , E. E. Schilling , and R. L. Small . 2007 . Comparison of
whole chloroplast genome sequences to choose noncoding regions for phy-
logenetic studies in angiosperms: e tortoise and the hare III. American
Journal of Botany 94 : 275 – 288 .
Silva , L. , M. Moura , H. Schaefer , F. Rumsey , and E. F. Dias . 2010 . Vascular
plants (Tracheobionta) . In P. A. V. Borges, A. Costa, R. Cunha, R. Gabriel,
V. Gonçalves, and A. F. Martins [eds.], A list of the terrestrial and marine
biota from the Azores, 117–146. Princípia, Cascais, Portugal.
Simmons , M. P. 2004 . Independence of alignment and tree search. Molecular
Phylogenetics and Evolution 31 : 874 – 879 .
Simmons , M. P. , and H. Ochoterena . 2000 . Gaps as characters in sequence-
based phylogenetic analysis. Systematic Biology 49 : 369 – 381 .
Stuessy , T. F. , G. Jakubowsky , R. S. Gómez , M. Pfosser , P. M. Schlüter , T. Fer , B.
Y. Sun , and H. Kato . 2006 . Anagenetic evolution in island plants. Journal
of Biogeography 33 : 1259 – 1265 .
Templeton , A. R. , K. A. Crandall , and C. F. Sing . 1992 . A cladistic analysis
of phenotypic associations with haplotypes inferred from restriction en-
donuclease mapping and DNA sequence data. III. Cladogram estimation.
Genetics 132 : 619 – 633 .
Watson , H. C. 1870 . Botany . In F. D. C. Godman [ed.], Natural history of the
Azores or Western Islands, 124–300. J. Van Voorst, London, UK.
Willdenow , C. 1809 . Enumeratio plantarum Horti Regii Berolinensis, conti-
nens descriptiones omnium vegetabilium in horto dicto cultorum. Taberna
Libraria Scholae Realis, Berlin, Germany.
Yang , Z. , and B. Rannala . 1997 . Bayesian phylogenetic inference using DNA
sequences: a Markov chain Monte Carlo method. Molecular Biology and
Evolution 14 : 717 – 724 .
Yule , G. U. 1924 . A mathematical theory of evolution based on the conclu-
sions of Dr. J.C. Willis, FRS. Philosophical Transactions of the Royal Society
of London, B, Biological Sciences 213 : 21 – 87 .
