Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
Contents lists available at ScienceDirect
Perspectives in Plant Ecology, Evolution and Systematics
journal homepage: www.elsevier.com/locate/ppees
Research article
Bees and evolution of occluded corollas in snapdragons and relatives
(Antirrhineae)
Beatriz Guzmán a,∗ , José María Gómez b,c , Pablo Vargas a
a
b
c
Dpto. de Biodiversidad y Conservación, Real Jardín Botánico, CSIC, Madrid, Spain
Dpto de Ecología Funcional y Evolutiva, Estación Experimental de Zonas Áridas, CSIC, Almería, Spain
Dpto de Ecología, Universidad de Granada, Granada, Spain
a r t i c l e
i n f o
Article history:
Received 11 September 2014
Received in revised form 18 July 2015
Accepted 21 July 2015
Available online 30 July 2015
Keywords:
Exploitation barriers
Floral evolution
Macroevolution
Melittophily
Ornithophily
Personate flower
a b s t r a c t
The tribe Antirrhineae is a natural group comprising 28 genera and over 320 species. In most Antirrhineae
the corolla is marked by the development of a prominent palate (personate flower) that sometimes
occludes the entrance of the corolla tube. Traditionally strong bees of different size have been considered
as the principal pollinators of occluded, personate flowers (snapdragon-type). Phylogenetic analyses (ITS
and ndhF regions) were conducted to gain insight into evolutionary changes in personate flower and
corolla occlusion. In addition, logistic regressions were performed in order to test the role of this type of
flower as a floral filtering morphology for pollinators. Historical reconstruction of the personate flower
supports its ancestral condition within Antirrhineae and a general pattern of recurrent corolla occlusion
shifts have prevailed since lineage diversification of the Antirrhineae. In addition, we found some evidence
of snapdragon-type corollas negatively affecting hummingbirds and insects other than bees. Part of this
outcome is due to predominance of bees as single visitors of Antirrhineae species with closed corollas (c.
65% of the species). The lack of significant correlation between bees and occluded, personate flowers is
interpreted as the ability of bees of visiting any type of flowers. The personate flower, particularly when
occluded, displays one of the most specialized corollas in pollinator exclusion by physical barriers.
© 2015 Geobotanisches Institut ETH, Stiftung Ruebel. Published by Elsevier GmbH. All rights reserved.
Introduction
Flower attributes include rewards (nectar and pollen) and
attractants (e.g. colour, scent, flower density) to floral visitors
and sometimes protection against undesirable ones. Mechanical
restrictions to the entrance to the corolla tube (e.g. trichomes,
scales, stamens and the keel of papilionaceus corollas) help some
angiosperms to efficiently reduce visits of non-pollinating animals (Willmer, 2011). The most occluded corolla of angiosperms
is arguably that of Antirrhinum and relatives (personate flower)
because it includes a physical barrier. The personate flower (from
Latin persona, mask) (Fig. 1 and Fig. S1) is characterized by having a mask-like palate, defined as a curvature of the lower lip
that more or less closes the corolla tube (Sutton, 1988). The personate flower has one of the most complex corollas in shape of
the angiosperms (Magallón and Vargas, 2014). Indeed, the bipartite corolla is a sophisticated form of protection that has been
traditionally considered an effective barrier protecting the nectar
∗ Corresponding author. Tel.: +34 914203017; fax: +34 914200157.
E-mail address: [email protected] (B. Guzmán).
produced at the base of the tube from undesirable floral visitors, when the palate completely closes the tube (Müller, 1929;
Sutton, 1988). Evolutionary success of the personate flower is manifested by its occurrence in disparate families within the order
Lamiales: Gesneriaceae (Codonoboea personatiflora, Didymocarpus antirrhinoides, Rhyncholossum medusothrix), Lentibulariaceae
(Genlisea, most Utricularia spp.), Linderniaceae (Stemodiopsis),
Orobanchaceae (Lamourouxia, Melampyrum, Pedicularis), Phrymaceae (some Mimulus) and Plantaginaceae (Antirrhinum, Linaria,
Nemesia, some Penstemon spp.) (Bentham, 1835; Fleischmann,
2012; Kadereit, 2004; Kampny, 1995; Kiew and Sam, 2012; Pennell,
1935). As the tribe Antirrhineae shows numerous personate flowers
with different degrees of corolla occlusion, it offers the opportunity
of testing the role of occluded corollas in filtering floral visitors.
Placed in the family Plantaginaceae, the tribe Antirrhineae is a
natural group comprising 28 genera (c. 320 species) mostly distributed in the Northern Hemisphere (DePamphilis et al., 1994;
Vargas et al., 2004, 2014). In most Antirrhineae genera (75% of
them) the corolla is marked by the development of a prominent palate and a pouch or spur, which contains the nectar (Fig.
S1). There is a wide variation in the type of Antirrhineae corolla
tube opening. For example, whereas in Antirrhinum (snapdragons)
http://dx.doi.org/10.1016/j.ppees.2015.07.003
1433-8319/© 2015 Geobotanisches Institut ETH, Stiftung Ruebel. Published by Elsevier GmbH. All rights reserved.
468
B. Guzmán et al. / Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
Fig. 1. Floral shape diversity and degrees of corolla occlusion of Antirrhineae species from the Old World (top) and the New World (bottom). The distribution of the
tribe Antirrhineae is also shown (A and J). Relative frequency of two qualitative floral traits (personate flower, corolla occlusion) (I) is also shown. Antirrhinum barrilieri
(B), Antirrhinum valentinum (C), Anarrhinum bellidifolium (D), Chaenorhinum grandiflorum (E), Chaenorhinum origanifolium segoviense (F), Kickxia spuria (G), Linaria elegans
(H), Lophospermum erubescens (K), Holmgrenanthe petrophila (L), Pseudorontium cyathiferum (M), Maurandella antirrhiniflora (N), Howelliella ovata (O), Mohavea breviflora
(P), Rhodochiton atrosanguineum (Q). Photographs by B. Guzmán (B, C and E–H, K and N), J. Martín (D), M. Williams (L), D. Valov (M), J. Chesnut (O), C. Christie (P), and
turning-earth.co.uk (Q).
the entrance to the tube is completely closed by a conspicuous
palate, in Rhodochiton the corolla is completely open, without
palate or even any trace of it (Fig. 1 and Fig. S1). Between
these extremes there exist species showing intermediate conditions (Fig. S1): well-developed palates that do not close the
mouth centrally because of the existence of a groove (e.g. Antirrhinum valentinum, Chaenorhinum grandiflorum, Pseudomisopates
rivas-martinezii, Sairocarpus coulterianus), inflated palates that
incompletely close the throat (e.g. most species of Chaenorhinum,
Mohavea spp., Maurandella antirrhiniflora), and non- or poorly
developed palates (e.g. Anarrhinum spp., Howelliella ovata, Maurandya spp.) that permit easy access to nectar through the corolla
tube (Fig. 1 and Fig. S1). Personate flowers with completely
occluded corollas are typically found in snapdragons (Antirrhinum)
(Nuttall, 1827). Snapdragon-type flowers can be found in both
the New World (Sairocarpus) and the Old World (Acanthorrhinum,
Albraunia, Antirrhinum, Asarina, Chaenorhinum, Cymbalaria, Linaria,
Misopates, Schweinfurthia). The occluded, personate flower of Antirrhineae has been traditionally related to bee-pollination. Cursory
field observations indicate that floral visitors are bees, hummingbirds and butterflies (Sutton, 1988). However, pollination is well
documented in only a few Antirrhineae species. Indeed, there
are only observational data of the floral visitors of 56 Antirrhineae species (c. 17%) and no information is available for about
11 genera (c. 40% of Antirrhineae genera). Nevertheless, pollinators have been studied in more detail in Antirrhinum (three
spp.; Vargas et al., 2010), Linaria (23 spp.; Arnold, 1982; BlancoPastor, 2014; Crawford, 2003; Fernández-Mazuecos et al., 2013a;
Sánchez-Lafuente, 2007; Stout et al., 2000; Valdés and Díaz, 1996),
P. rivas-martinezii (Amat et al., 2011), and Sairocarpus (two spp.;
Oyama et al., 2010).
This study aimed to reconstruct the evolution of the personate, occluded corolla of Antirrhineae. In particular, we investigated
the role of individual floral traits in filtering floral visitors
and taxa phylogenetic relationships. The particular objectives
of the present study are to: (1) characterize flower shape
and level of corolla tube occlusion of Antirrhineae genera and
major lineages; (2) reconstruct evolutionary changes of corolla
phenotypes, particularly personate and occluded corollas using
phylogenetic analyses; and (3) explore whether snapdragon-type
corollas filter actual floral visitors by conducting floral visitor
censuses.
Material and methods
Taxon and DNA sampling
Taxa recognized in this study followed the taxonomic treatment
of Sutton (1988) except for Nuttallanthus, which is circumscribed
within Linaria (Fernández-Mazuecos et al., 2013b). In addition
we included two more genera (Pseudomisopates, Nanorrhinum)
described after Sutton’s study (Ghebrehiwet, 2001; Güemes, 1997).
A total of 186 species (of around 300), 193 taxa, and 213 individuals (Fig S1) were analyzed for DNA sequencing and morphological
characterization. Based on a previous study (Fernández-Mazuecos
et al., 2013b), Lafuentea rotundifolia was included as outgroup.
Our molecular analyses combine previously published data with
newly generated sequences (Table S1). The nuclear ribosomal ITS
(193 taxa, 33 new generated sequences) and the plastid ndhF gene
(132 taxa, 110 new generated sequences) were analyzed. Methods for DNA extractions, PCR amplifications and sequencing were
described in Fernández-Mazuecos et al. (2013b) and Vargas et al.
(2014).
Morphological data and floral visitors
Two characters (personate flower, corolla occlusion) traditionally considered effective barriers preventing the access of
undesirable floral visitors to the nectar produced at the base of
the tube (Sutton, 1988) were scored for Antirrhineae taxa and the
sister species L. rotundifolia (Table S2). In the present study, personate flowers are considered those with a well-developed palate,
irrespective of the occlusion of the corolla tube entrance. Categorical character for personate flower (yes, no) and corolla occlusion
(occluded, partially occluded, open-throated) were scored from the
literature (Elisens, 1985; Sutton, 1988; Thompson, 1988) and our
own observation on living plants.
Floral visitors were scored from the literature (Table S3) and
by conducting floral visitor censuses in the field during the period
2003–2013. We carried out censuses in 11 Antirrhineae species
(seven genera) of which flower visitors were unknown, plus some
of the sister species L. rotundifolia. Censuses were conducted
during 500–800 min recording the identities of all visitors (see
Vargas et al., 2010). One specimen per insect species was captured
for identification. Qualitative data were recorded and floral visitors
B. Guzmán et al. / Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
were pooled into taxonomical groups (bees, beetles, bee-flies, butterflies, flies, hummingbirds, moths/hawk moths). We analyzed
qualitative data of floral visitors to merge data from censuses and
those from 24 Antirrhineae species taken from literature because
unfortunately those studies did not include visitation rates.
DNA sequence alignment and phylogeny reconstructions
All sequences were aligned using MAFFT (ver. 6; Katoh et al.,
2002) with default parameters, and further adjustments were made
by visual inspection. ndhF matrix was translated to amino acids in
MEGA5 (Tamura et al., 2011) to confirm conservation of the amino
acid reading frame, ensure proper alignment, and to check for premature stop codons.
Optimal models of nucleotide substitution were determined for
each sequence data set according to the Akaike Information Criterion (AIC) using jModelTest 0.1 (Posada, 2008).
Maximum likelihood (ML) bootstrap analyses and the inference of the optimal tree were performed with the program RAxML
(Stamatakis, 2006; Stamatakis et al., 2008) under the general time
reversible (GTR) nucleotide substitution model with among-site
rate variation modelled with a gamma distribution. Five hundred
independent searches starting from different maximum parsimony
initial trees were performed using RAxML version 7.2.7 on the
CIPRES portal teragrid (www.phylo.org; Miller et al., 2010). Clade
support was assessed with 5000 replicates of a multiparametric
bootstrapping. Bayesian Inference (BI) analyses were conducted
with MrBayes 3.1.2 (Ronquist and Huelsenbeck, 2003) on the
CIPRES portal teragrid (www.phylo.org; Miller et al., 2010), using
the best-fit models of nucleotide substitution determined for each
sequence data set with jModelTest (Posada, 2008). Two independent but parallel Metropolis-coupled MCMC analyses were
performed with default settings. Each search was run for 15 million
generations sampled every 1000th generations. The initial 25% of
samples of each Metropolis-coupled MCMC run were discarded as
burnin, and the remaining samples were summarized as 50% majority rule consensus phylograms with nodal support expressed as
posterior probabilities. The standard deviation of split frequencies
between runs was evaluated to establish that concurrent runs had
converged (<0.01). Tracer v1.5 (Rambaut and Drummond, 2007)
was used to determine whether the MCMC parameter samples
were drawn from a stationary, unimodal distribution, and whether
adequate effective sample sizes for each parameter (ESS > 100)
were reached.
Since the partition homogeneity test results have been shown
to be misleading (Barker and Lutzoni, 2002; Darlu and Lecointre,
2002) preliminary phylogenetic analyses of the two single DNA
regions data sets were examined. The ITS tree (Fig. S2) did not
result in strongly supported conflicts with respect to the ndhF
tree (Fig. S3). Accordingly, we followed a concatenation approach
and performed ML and BI analyses on a matrix of species with
sequences from both DNA regions (131 spp.). In addition, in four
species from four genera (Epixiphium, Holmegranthe, Linaria and
Lophosmpermum) were sequenced only for the ndhF region. They
were included because (1) there is information about floral visitors (Linaria glacialis, Lophospermum erectum) or (2) they belong to
monotypic genera (Epixiphium wislizenii, Holmegranthe petrophila).
Phylogenetic relationships were inferred as described above for
the ML analysis. The Bayesian analysis was conducted using BEAST
(v.1.5.4, Drummond et al., 2006; Drummond and Rambaut, 2007),
in order to obtain ultrametric trees (for subsequent analyses).
We used a birth-death model for the tree prior, an uncorrelated
relaxed molecular clock, a random starting tree, and the models
selected by jModeltest (Posada, 2008) for each DNA region. We
did two independent runs; each one for 120 million steps sampled
every 12,000th. Convergence to stationarity and effective sample
469
size (ESS) of model parameters were assessed using TRACER 1.5
(Rambaut and Drummond, 2007). Samples from both independent
runs were pooled after removing a 10% burnin using Log Combiner
1.5.4 (Drummond and Rambaut, 2007). We used a previous molecular estimated crown group age of Antirrhineae as a calibration
point (Vargas et al., 2014). We applied a normally distributed calibration point prior with a mean of 30.22 and standard deviation of
four million years.
Floral character evolution
We explored the evolutionary models better describing
the evolution of both personate flower and corolla occlusion.
Maximum-likelihood based measurement of phylogenetic signal
was performed using Pagel’s (Pagel, 1999) as recommended by
Münkemüller et al. (2012). We estimated and its model likelihood
score on the basis of 2000 Beast trees by using the “fitDiscrete” function in the R package geiger 1.99-1 (Harmon et al., 2008). After that,
we repeated the analysis constraining lambda to 1 and to 0 by using
the “transform” function. A significant departure from the model
with lambda 1 indicates that traits are not evolving according to
a Brownian motion (BM) model, whereas a significant departure
from the model with lambda 0 indicates the occurrence of phylogenetic signal in the evolution of traits (Nunn, 2011). Because
our dataset is very unbalanced, we tuned up the parameters of
the model to increase our confidence (number of iterations = 200,
fail = 1e + 300). All the analyses were performed under two models:
(1) Equal-Rates model (ER) of where a single parameter governs
all transition rates and (2) All-Rates-Different model (ARD) where
each rate is a unique parameter.
In addition, patterns of evolution of both floral traits were
explored using parsimony, ML and Bayesian approaches. Parsimony and ML optimizations were performed in Mesquite 2.75
(Maddison and Maddison, 2011) onto 2000 BEAST trees to account
for topological uncertainty. Parsimony mapping of ancestral states
was conducted assuming characters as unordered. Prior to perform
the ML reconstruction, we explored whether a single-rate model
by constraining all transition rates to be equal or an unrestricted
model with two (personate flower) and six (corolla occlusion)
transition rates best fits our data. All analyses were performed
in BayesTraits 1.0 (Pagel and Meade, 2007). Five hundred ML
optimizations were performed on each tree. The log Likelihood
Ratio Test did not favour any model (personate flower: LRT = 2.03,
d.f. = 1, p = 0.85; corolla occlusion: LRT = 5.97, d.f. = 5, p = 0.69).
Consequently, the Markov k-state 1 (Mk1) parameter model, with
equal probability for any particular character change, was selected
in order to infer ancestral states in Mesquite. For each state at each
node, the analysis calculated the number of trees on which such
state was reconstructed as uniquely best according to a decision
threshold of two log likelihood units (Maddison and Maddison,
2011). The estimated number of shifts between character states
was obtained using the “Summarize State Changes Over Trees”
application in Mesquite. To account for phylogenetic mapping
uncertainty, we further evaluated probabilities of ancestral states
calculated from 2000 BEAST trees using the MCMC method in
BayesMultiState, implemented in the BayesTraits 1.0 package
(Pagel and Meade, 2007). Ancestral states were only reconstructed
for 89 nodes, which were selected based on their posterior probability support values of the BEAST analysis (only those nodes with
PP ≥ 0.95 were analyzed). A reversible-jump (RJ) hyperprior with
a gamma prior (mean and variance seeded from uniform distributions from 0 to 10) was used to reduce uncertainty and arbitrariness
of choosing priors in the MCMC analysis. The ratedev parameter
was adjusted using the autotune option, which automatically
finds a rate deviation parameter which gives an acceptance rate of
approximately 30%. The “Addnode” command was used to find the
470
B. Guzmán et al. / Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
proportion of the likelihood associated with each of the possible
states at selected nodes. Three independent MCMC runs were
performed with 40 × 106 iterations. Chains were sampled every
5000 iterations after a burnin of 10 × 106 iterations. Because all
runs gave similar results, we here only report one of them.
(Fernández-Mazuecos et al., 2013b; Vargas et al., 2014). In addition,
all phylogenetic analyses (ML, Bayesian, individual vs multiple DNA
regions) were consistent with the monophyly of all genera provided
that Albraunia and Holzneria are included in Chaenorhinum (Figs. S2,
S3 and S5).
Phylogenetic logistic regression of floral visitors and
morphological variables
Evolution of personate flower and corolla occlusion
We evaluated the relationships between floral visitors (presence/absence of bee, beetle, bee-fly, butterfly, fly, hummingbird,
or moth as response variables) and two predictor variables
(personate flower, corolla occlusion) in 64 Antirrhineae species.
Examining character evolution without accounting for correlations
among characters based on shared evolutionary history can introduce known biases (Felsenstein, 1985). Therefore, we performed
univariate Phylogenetic Logistic Regressions (PLRs) with Firth correction. Both the independent and the predictor variables were
categorical. The polytomous categorical variable corolla occlusion
was transformed into dummy variables. The regressions were
performed using a variance-covariance matrix of the species constructed using the BEAST MCC tree (see above section) in the R
package ape (Paradis et al., 2004). PLRs were run using the PloGReg.m function (Ives and Garland, 2010) implemented for Matlab
(Mathworks, Natick, MA, USA). A bootstrapping procedure involving 500 simulations was used to generate the 95% confidence
intervals and test for statistical significance of the slope of the
regression model. The parameters produced in LR are interpreted
in terms of odds ratios (OR), calculated by taking the exponential
of the parameter (B), which describe the strength of association
between predictor and response variable. The OR represents the
change in odds of the outcome of a binary response variable for a
one-unit change in the predictor variable.
Results
Floral data and floral visitors
Floral characteristics for 135 Antirrhineae taxa are shown in
Table S2. Fig. 1 reflects the diversity of floral morphology in the
tribe, respect to degrees of corolla tube occlusion. Flower was personate in 123 of the studied species (91.11%, Fig. 1I). The majority
(60.97%) of the personate-flowered species mostly showed completely occluded corollas, although partially occluded (35.77%) or
open-throated corollas (2.44%) were also found.
Information on actual floral visitors was compiled from 64 Antirrhineae species representing 21 genera (75%). Insects (88.88%) and
birds (11.11%) were the only visitors of Antirrhineae flowers (Fig. 2
and Table S4). Floral visitor systems vary from generalist (e.g. Anarrhinum bellidifolium) to specialist (e.g. Antirrhinum spp.), including
species of which no visitors were observed (e.g. Linaria micrantha).
Bees visited 84.37% of the Antirrhineae species (Fig. 2) both as single floral visitors (59.26%) or co-occurring with beetles (9.26%), flies
(7.40%), bee-flies (12.96%), butterflies (24.07%), moths (12.96%), and
hummingbirds (7.40%). One species (C. grandiflorum) was only visited by bee-flies, and hummingbirds were the unique visitors of four
species from the New World (Galvezia lanceolata, Gambelia juncea,
Lophospermum erubescens, and Mabrya acerifolia). Bees were the
single floral visitor of 64.70% of species with occluded, personate
flowers (Fig. S4).
Phylogenetic relationships of Antirrhineae
Sequence features of both ITS and ndhF data sets are summarized in Table S5. The ndhF + ITS tree (Fig. S5) retrieved
was consistent with previous molecular phylogenetic analyses
Results from analyses performed under ER and ARD models
were congruent. For the sake of simplicity, we show only results
from ER model (see Supplementary data for ARD model results).
Both personate flower and corolla occlusion evolution was consistent with Brownian motion (BM) evolution, because the likelihood
of models where lambda was estimated did not differ from those
where lambda was forced to equal one (see statistics in the Supplementary data). Besides, we detected phylogenetic signal in the
evolution of both floral traits, since the ML estimate of lambda differed from trees where lambda was forced to be zero (see statistics
in the Supplementary data).
A summary of historical character state reconstruction (ML and
parsimony results not shown), as applied to the two characters,
follows:
(a) Personate flower (Fig. 2A): Reconstruction analyses indicated
that the personate flower is the most likely ancestral state
for the tribe (Fig. 2A). The reconstruction showed independent
shifts away from the ancestral state in two Antirrhineae groups
(Anarrhinum clade and New World Maurandyinae) and in the
species H. ovata. At least one reversal to the ancestral state was
found within the New World Maurandyinae (M. antirrhiniflora).
The RJ MCMC approach sampled a model with equal transition rates in 98% of the iterations (personate to non-personate:
0.0115; non-personate to personate: 0.0116; Tables S6 and S7).
In addition, parsimony and ML showed that unequivocal transitions in both directions occurred with similar frequencies
(parsimony: 2.54/2.46; ML: 2.06/0.96) (Fig. 2A and Table S8).
(b) Corolla occlusion (Fig. 2B): Reconstruction analyses did not suggest that occluded corolla is the most likely ancestral state.
Occluded and open corollas would have evolved independently
at least eight and four times, respectively. This trait was evolutionary labile within Antirrhineae as also evidenced by the
parsimony and ML number of unequivocal transitions (Fig. 2B
and Table S8) among states. The RJ MCMC approach sampled more often (61.33%) a model with equal transition rates
between occluded/partially occluded corollas and between
partially occluded/open-throated corollas (Table S7). In contrast, overall, unequivocal changes from occluded → partially
occluded corollas significantly prevail over other transitions
(parsimony/ML, 6.91/2.05). Additionally, Bayesian transition
rates and parsimony/ML unequivocal transitions from occluded
to open corollas and vice versa were the less frequent (occluded
to open: 0.0024/1.50/0.50, open to occluded: 0.0079/1.48/0.00;
Tables S6 and S8).
Correlated evolution of floral visitors and phenotypic traits
Table 1 shows the phylogenetic univariate analyses of personate flower and corolla occlusion with six types of floral visitors.
Both personate flowers and occluded corollas seem to be positively
related with bees (Bpersonate = 1.22; Boccluded = 1.80) and negatively
with the remaining insects and hummingbirds (Table 1). However,
the occluded corolla was a significant predictor only for hummingbirds (B = −3.62, p < 0.01) and butterflies (B = −2.75, p < 0.01)
(Table 1).
B. Guzmán et al. / Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
471
Fig. 2. Results of the ancestral state reconstruction analyses for personate flower (A) and corolla occlusion (B) mapped onto the BEAST MCC tree obtained in the analysis
of ITS and ndhF sequences based on a Bayesian analysis over 2000 trees. Pie charts represent posterior probabilities of Bayesian inference character state evolution. The
character states found in the terminal taxa are indicated as circles of the respective colours. Floral visitor information plotted on the right side of the tree: 䊏 bee, 䊐 beetle,
bee-fly, 䊉 butterfly,
fly, hummingbird, moth, 5–6 insect visitor types, and
not seen. Evolutionary transition networks (inset; See Tables S6 and S8 for additional
results) are also shown. Arrows indicate the direction and relative proportion (given by line thickness) of Bayesian transition rates between states. Bayesian transition rates
(average) are provided next to arrows with MP and ML unequivocal transition numbers (average) in brackets. The circle size is proportional to the number of samples with
the given state. Parsimony and ML reconstructions performed in Mesquite 2.75 (Maddison and Maddison, 2011). Bayesian reconstruction performed in BayesTraits 1.0 (Pagel
and Meade, 2007).
472
B. Guzmán et al. / Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
Fig. 2 (Continued ).
B. Guzmán et al. / Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
473
Table 1
Phylogenetic logistic regression parameter estimates for the effects of corolla occlusion and personate flower on floral visitors of 64 Antirrhineae species. B, slope; OR, odd
ratio.
B
Bee
Personate flower (no as reference)
Yes
Corolla occlusion (open-throated as reference)
Occluded
Partially occluded
Bee-fly
Personate flower (no as reference)
Yes
Corolla occlusion (open-throated as reference)
Occluded
Partially occluded
Beetle
Personate flower (no as reference)
Yes
Corolla occlusion (open-throated as reference)
Occluded
Partially occluded
Butterfly
Personate flower (no as reference)
Yes
Corolla occlusion (open-throated as reference)
Occluded
Partially occluded
Fly
Personate flower (no as reference)
Yes
Corolla occlusion (open-throated as reference)
Occluded
Partially occluded
Hummingbird
Personate flower (no as reference)
Yes
Corolla occlusion (open-throated as reference)
Occluded
Partially occluded
Moth
Personate flower (no as reference)
Yes
Corolla occlusion (open-throated as reference)
Occluded
Partially occluded
Bootstrap mean
(95% CI)a
1.22
1.23 (−0.78, 2.90)
1.80
0.58
1.79 (−0.40, 4.43)
0.53 (−1.59, 2.31)
−0.51
−0.53 (−2.27, 1.38)
−1.76
−0.15
−1.99 (−4.12, 0.23)
−0.07 (−1.85, 1.55)
−1.17
−1.11 (−3.05, 0.79)
−0.24
−0.005
−0.008 (−1.93, 1.78)
−0.41 (−2.78, 1.84)
0.56
0.50 (−1.19, 2.22)
−2.75
−2.16
−2.75 (−4.85, −1.08)*
−2.09 (−4.31, −0.44)*
−0.22
−0.14 (−2.32, 1.46)
−0.09
−0.02
−0.92 (−2.54, 1.64)
0.11 (−1.27, 1.64)
−2.29
−1.80 (−4.22, 0.06)
−3.62
−1.32
−3.41 (−5.33, −0.44)*
−1.29 (−3.16, 0.27)
−0.69
−0.57 (−2.47, 1.51)
−1.14
−0.60
−1.41 (−4.03, 0.88)
−0.52 (−2.49, 1.21)
a
Parametric bootstrapping was performed by simulating 500 data sets to obtain means and confidence intervals and to test the null hypotheses that there is no phylogenetic
signal in the residuals (H0 : a = −4, 1-tailed test) and that the regression coefficients equal 0 (H0 : bi = 0, 2-tailed tests).
*
p ≤ 0.05.
Discussion
Association between plants and flower visitors has been historically proposed as a main factor driving the evolutionary change
of both flower and pollinator phenotypes (Darwin, 1877; Stebbins,
1970). Multidisciplinary approaches using techniques and methods
developed over the past three decades (e.g. molecular phylogenetics, phylogenetic comparative methods, network analysis) have
provided a reliable framework to test this association (Armbruster,
2014; Patiny, 2012; Smith, 2010). Yet, a strong evolutionary signal has been difficult to recover at a macroevolutionary scale since
these approaches are complex for large plant groups of species and
genera (Forest et al., 2014; Valente et al., 2012; van der Niet and
Johnson, 2012).
Evolution of occluded, personate corollas in Antirrhineae
The personate flower has been traditionally used in the taxonomic circumscription of Antirrhineae genera (see Sutton, 1988
for a review); however, little is known about the evolutionary
change of the diverse corolla phenotypes. Rothmaler (1943) considered the personate flower as a derived state based upon the basal
position of the ‘Maurandya Gruppe’ (genera Epixiphium, Lophospermum, Maurandella, Maurandya, Rhodochiton) in his evolutionary
scheme. In contrast, a cladogram based on morphological characters led Elisens (1985) to suggest an ancestral state for the personate
flower. Our reconstruction analyses support its ancestral condition
(Fig. 2A). Given that personate flowers can also be found in genera of some other families of Lamiales, convergence of this type of
flower is interpreted. This inference per se needs to accept shifts of
the personate flower several times or that the personate flower has
evolved independently in Lamiales.
Our phylogenetic reconstruction also reveals dynamic and complex transitions of the corolla occlusion in Antirrhineae. The
ancestral character state reconstruction showed that the occluded
corolla arose late in the course of flower evolution. An intermediate step appears to be also needed in shifts from corolla
occlusion to opening, and vice versa (Fig. 2B). Indeed, the main
clades of the phylogeny primarily showed the three corolla states
(Fig. 2B). Intermediate states have been also claimed for evolution
of long spurs in North American columbines (Whittall and Hodges,
2007). Floral evolution by intermediate states appear to be the rule
rather than the exception in the evolution of reproductive traits of
angiosperms (Endress, 2011; Specht and Bartlett, 2009), although
474
B. Guzmán et al. / Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
sudden shifts without intermediate steps have been described in
cultivated Antirrhinum (Keck et al., 2003; Manchado-Rojo et al.,
2012) and Linaria (Cubas et al., 1999). In our study, transition
rates from occluded/partially occluded corollas appear to have been
similar to transition rates from partially occluded/open-throated
corollas (0.0350–0.0371, Fig. 2B and Tables S6 and S7). Nevertheless, species distribution in the phylogeny shows that the occluded
corolla (c. 190 spp.) is more frequent than open-throated (c. 30
spp.) and partially occluded (c. 65 spp.) corollas. Further studies
are needed to test whether certain lineages with snapdragon-type
corollas diversified into a higher number of species in a short period
of time (radiations).
Corolla occlusion is evolutionarily quite labile and not so perfect in excluding floral visitors (Fig. 2B and Fig. S4). Nototribic
pollination (i.e. pollen deposited on the back of the thorax of nectarfeeding insects) by bees has been demonstrated to result in effective
pollination within Antirrhineae (Macior, 1967; Sánchez-Lafuente
et al., 2011; Vargas et al., 2010). Prevention from less-effective pollinators (e.g. butterflies, flies) and pollen predators (beetles) may
have driven evolution towards complete corolla occlusion in many
snapdragons and relatives. Historical periods of low abundance of
bees may have also influenced evolution towards a loss of palate’s
function and thus shifts to partially occluded corollas that allow
butterflies or even flies to enter. Interestingly, the fact that certain
threatened species have open-throated (e.g. Anarrhinum fruticosum, Linaria nigricans) or partially occluded (e.g. A. valentinum, G.
lanceolata, Nanorrhinum kuriense, P. rivas-martinezii) corollas that
offer greater opportunity to a larger guild of pollinators require to
be explored at a microevolutionary scale.
corollas received visits from bees and more diverse assemblages of
floral visitors. In Utricularia, bees are the main pollinators of Utricularia reniformis, a species bearing occluded corollas (Clivati et al.,
2014), while bees, butterflies, moths, hawk moths and flies visit the
partially occluded corollas of Utricularia purpurascens (Hobbhahn
et al., 2006). In addition, bees are the only visitors of occluded
corollas of Melampyrum arvense (Kwak, 1988), whereas bees, flies,
butterflies and ants visit open-throated corollas of Melampyrum
roseum (Hiei and Suzuki, 2001). The close association between bees
and snapdragon-type flowers in different groups of angiosperms
lead us to suggest that bees have favoured evolution of similar
corolla phenotypes multiple times in the course of Lamiales evolution.
Conclusions
The personate flower, particularly when occluded, displays one
of the most specialized and effective corollas in floral visitor exclusion by physical barriers. However, we found low statistical support
for strong specialization in the snapdragons. Lack of statistical
significance of a high-specialization system, such as that of the
occluded, personate flower of Antirrhineae, suggests that either the
evolutionary pattern of corolla occlusion is not robust or current
methods not always provide statistically traceable patterns. Further studies integrating phylogenetic and ecological information,
the genetic basis of floral trait variation, and the role of abiotic factors in determining mutualistic and antagonistic interactions are
needed to unveil the complex processes underlying the successful
floral diversification of Antirrhineae.
Snapdragon-type corollas as an actual filter to floral visitors
Acknowledgements
Bee pollination (melittophily) has been claimed the main factor
enhancing reproductive success in Antirrhineae since only strong
bees of different size can easily access the corolla tube and deposit
large amounts of pollen on the stigma (Müller, 1929; Sutton, 1988;
Vargas et al., 2010). Our statistical examination of the relationship
between occluded corollas and floral visitor assemblages provided
some evidence of occluded, personate flowers negatively affecting
hummingbirds and insects other than bees (Table 1). This is congruent with the predominance of bees in snapdragon-type corollas
(64.70% of the species with this type of corollas were exclusively
visited by bees; Fig. S4). This type of corolla is also visited by
other animals: beetles (5.88%), butterflies (8.82%), generalist flies
(2.94%), moths (8.82%) and hummingbirds (2.94%). Indeed, the
palate acts as a physical barrier to most floral visitors but bees
in three Antirrhinum species (Vargas et al., 2010). This specialized corolla is clearly effective; however, it has weak points for
small corollas. Recent studies of the small flowers of the genus
Linaria (Blanco-Pastor, 2014; Fernández-Mazuecos et al., 2013a)
found heavy beetles, butterflies, and strong hawk moths entering
species with occluded corollas (e.g. Linaria algarviana, Linaria anticaria, Linaria spartea), although significantly less frequently than
bees. In addition, careful inspection of corolla occlusion in flowers such as those of Pseudomisopates revealed a corolla with an
inconspicuous pore opening that facilitated a relatively generalized
floral visitor assemblage (Amat et al., 2011). Our logistic regression
analyses indicate that floral visits by bees increase, although not
significantly, when the corolla is closed (Table 1). The lack of significant correlation between bees and occluded corollas may be
explained by the flexible bee behaviour. Though bees are the most
frequent visitors of this type of corollas (30 of the 54 spp.; Fig. S4),
they also commonly visit other types of flowers (54 of 64 spp.). Similarly, other species of Lamiales with personate flowers seem to fit
into this constraint pattern, in which flowers with occluded corollas are primarily visited by bees, while flowers with non-occluded
The authors thank Peter Endress and two anonymous reviewers
for helpful discussions of the manuscript; M. Agudo, E. Amat, M.
Fernández-Mazuecos, A. González-Posada, I. Liberal, C. Martínez,
D. Romero for field assistance; M. Carles-Tolrá for Diptera identification; Concepción Ornosa for bee identification; E. Carrió
(Chaenorhinum tenellum), J. Güemes (Acanthorrhinum ramosissimum), J. Quiles (Linaria pelisseriana), J.M. Martínez (Kickxia lanigera
and Kickxia spuria), J. Ramírez (Chaenorhinum rubrifolium subsp.
raveyi/L. rotundifolia) for providing locations; E. Cano for lab
assistance. We are grateful to the project Flora iberica IX (CGL201128613-C03-01) for providing herbarium specimens. This research
was supported by the Spanish Ministry of Science and Innovation
through project CGL2009-10031 to PV and by the Spanish Ministry of Economy and Competitiveness through a Juan de la Cierva
fellowship to BG.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.ppees.2015.07.
003
References
Amat, M.E., Vargas, P., Gómez, J.M., 2011. Pollen quality limitation in the Iberian
critically endangered genus Pseudomisopates (Antirrhinaceae). Plant Ecol. 212,
1069–1078.
Armbruster, W.S., 2014. Floral specialization and angiosperm diversity: phenotypic
divergence, fitness trade-offs and realized pollination accuracy. AoB Plants 6
(plu003).
Arnold, R.M., 1982. Pollination, predation and seed set in Linaria vulgaris
(Scrophulariaceae). Am. Mid. Nat. 107, 360–369.
Barker, F.K., Lutzoni, F.M., 2002. The utility of the incongruence length difference
test. Syst. Biol. 51, 625–637.
Bentham, G., 1835. Scrophularineae Indicae. James Ridgway and Sons, London.
B. Guzmán et al. / Perspectives in Plant Ecology, Evolution and Systematics 17 (2015) 467–475
Blanco-Pastor, J.L., 2014. Análisis de los factores responsables de la evolución de
angiospermas durante el Cuaternario: un estudio macro y microevolutivo en
Linaria sect. Supinae. Universidad Pablo de Olavide, Seville, Spain (Ph.D.
Thesis).
Clivati, D., Cordeiro, G.D., Płachno, B.J., de Miranda, V.F.O., 2014. Reproductive
biology and pollination of Utricularia reniformis A. St. -Hil. (Lentibulariaceae).
Plant Biol. 16, 677–682.
Crawford, P.T., 2003. Biosystematics of North American Species of Nuttallanthus
(Lamiales). University of Oklahoma, Norman, Oklahoma, USA (Ph.D. Thesis).
Cubas, P., Vincent, C., Coen, E., 1999. An epigenetic mutation responsible for
natural variation in floral symmetry. Nature 401, 157–161.
Darlu, P., Lecointre, G., 2002. When does the incongruence length difference test
fail? Mol. Biol. Evol. 19, 432–437.
Darwin, C.R., 1877. The Different Forms of Flowers on Plants of the Same Species.
Murray, London.
DePamphilis, C.W., Atkinson, T.N., Elisens, W.J., 1994. Tribal relationships in
Scrophulariaceae subfamily Antirrhinoideae: insights from sequence variation
of the plastid-encoded gene rps2. Am. J. Bot. 81 (Suppl. 6), 152 (Abstract).
Drummond, A.J., Ho, S.Y.W., Phillips, M.J., Rambaut, A., 2006. Relaxed
phylogenetics and dating with confidence. PLoS Biol. 4, e88.
Drummond, A.J., Rambaut, A., 2007. BEAST: Bayesian evolutionary analysis by
sampling trees. BMC Evol. Biol. 7, 214.
Elisens, W.J., 1985. Monograph of the Maurandyinae
(Scrophulariaceae-Antirrhineae). Syst. Bot. Monogr. 5, 1–97.
Endress, P.K., 2011. Evolutionary diversification of the flowers in angiosperms. Am.
J. Bot. 98, 370–396.
Felsenstein, J., 1985. Confidence limits on phylogenies: an approach using the
bootstrap. Evolution 39, 783–791.
Fernández-Mazuecos, M., Blanco-Pastor, J.L., Gómez, J.M., Vargas, P., 2013a. Corolla
morphology influences diversification rates in bifid toadflaxes (Linaria sect.
Versicolores). Ann. Bot. 112, 1705–1722.
Fernández-Mazuecos, M., Blanco-Pastor, J.L., Vargas, P., 2013b. A phylogeny of
toadflaxes (Linaria Mill.) based on nuclear Internal Transcribed Spacer
sequences: systematic and evolutionary consequences. Int. J. Plant Sci. 174,
234–249.
Fleischmann, A., 2012. Monograph of the Genus Genlisea. Redfern Natural History
Productions, Poole, Dorset, England.
Forest, F., Goldblatt, P., Manning, J.C., Baker, D., Colville, J.F., Devey, D.S., Jose, S.,
Kaye, M., Buerki, S., 2014. Pollinator shifts as triggers of speciation in painted
petal irises (Lapeirousia: Iridaceae). Ann. Bot. 113, 357–371.
Ghebrehiwet, M., 2001. Taxonomy, phylogeny and biogeography of Kickxia and
Nanorrhinum (Scrophulariaceae). Nord. J. Bot. 20, 655–689.
Güemes, J., 1997. Pseudomisopates (Scrophulariaceae), un nuevo género ibérico. An.
Jard. Bot. Madr. 55, 492–493.
Harmon, L.J., Weir, J.T., Brock, C.D., Glor, R.E., Challenger, W., 2008. GEIGER:
investigating evolutionary radiations. Bioinformatics 24, 129–131.
Hiei, K., Suzuki, K., 2001. Visitation frequency of Melampyrum roseum var.
japonicum (Scrophulariaceae) by three bumblebee species and its relation to
pollination efficiency. Can. J. Bot. 79, 1167–1174.
Hobbhahn, N., Küchmeister, H., Porembski, S., 2006. Pollination biology of mass
flowering terrestrial Utricularia species (Lentibulariaceae) in the Indian
Western Ghats. Plant Biol. 8, 791–804.
Ives, A.R., Garland Jr., T., 2010. Phylogenetic logistic regression for binary
dependent variables. Syst. Biol. 59, 9–26.
Kadereit, J.W., 2004. The families and genera of vascular plants VII. Flowering
plants. Dicotyledons. Lamiales (except Acanthaceae including Avicenniaceae).
Springer-Verlag, Berlin Heidelberg.
Kampny, C.M., 1995. Pollination and flower diversity in Scrophulariaceae. Bot. Rev.
61, 350–366.
Katoh, K., Misawa, K., Kuma, K., Miyata, T., 2002. MAFFT: a novel method for rapid
multiple sequence alignment based on fast Fourier transform. Nucleic Acids
Res. 30, 3059–3066.
Keck, E., McSteen, P., Carpenter, R., Coen, E., 2003. Separation of genetic functions
controlling organ identity in flowers. EMBO J. 22, 1058–1066.
Kiew, R., Sam, Y.-Y., 2012. Codonoboea personatiflora (Gesneriaceae), a new species
from Peninsular Malaysia. PhytoKeys 18, 61–66.
Kwak, M.M., 1988. Pollination ecology and seed-set in the rare annual species
Melampyrum arvense L. (Scrophulariaceae). Acta Bot. Neerl. 37, 153–163.
Macior, L.W., 1967. Pollen-foraging behavior of Bombus in relation to pollination of
nototribic flowers. Am. J. Bot. 54, 359–364.
Maddison, W.P., Maddison, D.R., 2011. Mesquite: A Modular System for
Evolutionary Analysis. Version 2.75. http://mesquiteproject.org
Magallón, S., Vargas, P., 2014. Eudicotyledons: the greatest flower diversity in
Angiosperms. In: Vargas, V., Zardoya, R. (Eds.), The Tree of Life. Sinauer
Associates, Inc., Publishers, Sunderland, US, pp. 156–166.
Manchado-Rojo, M., Delgado-Benarroch, L., Roca, M.J., Weiss, J., Egea-Cortines, M.,
2012. Quantitative levels of Deficiens and Globosa during late petal
development show a complex transcriptional network topology of B function.
Plant J. 72, 294–307.
Miller, M.A., Pfeiffer, W., Schwartz, T., 2010. Creating the CIPRES Science Gateway
for inference of large phylogenetic trees. In: Proceedings of the Gateway
Computing Environments Workshop (GCE), New Orleans, pp. 1–8.
475
Müller, L., 1929. Anatomisch-biomechanische Studien an maskierten
Scrophulariaceen-blüten. Österr. Bot. Z. 80, 149–161.
Münkemüller, T., Lavergne, S., Bzeznik, B., Dray, S., Jombart, T., Schiffers, K.,
Thuiller, W., 2012. How to measure and test phylogenetic signal. Methods Ecol.
Evol. 3, 743–756.
Nunn, C.L., 2011. The Comparative Method in Evolutionary Anthropology and
Biology. University of Chicago Press, Chicago.
Nuttall, T., 1827. An Introduction to Systematic and Physiological Botany. Hilliard
and Brown, Cambridge.
Oyama, R.K., Jones, K.N., Baum, D.A., 2010. Sympatric sister species of Californian
Antirrhinum and their transiently specialized pollinators. Am. Mid. Nat. 164,
337–347.
Pagel, M., 1999. Inferring the historical patterns of biological evolution. Nature
401, 877–884.
Pagel, M., Meade, A., 2007. BayesTraits. Version 1.0. Reading, UK. http://evolution.
rdg.ac.uk
Paradis, E., Claude, J., Strimmer, K., 2004. APE: analyses of phylogenetics and
evolution in R language. Bioinformatics 20, 289–290.
Patiny, S., 2012. Advances in the study of the evolution of plant–pollinator
relationships. In: Patiny, S. (Ed.), Evolution of Plant–Pollinator Relationships.
Cambridge University Press, Cambridge, UK, pp. 458–468.
Pennell, F.W., 1935. The Scrophulariaceae of eastern temperate North America.
Monographs 1, 1–650.
Posada, D., 2008. jModelTest: phylogenetic model averaging. Mol. Biol. Evol. 25,
1253–1256.
Rambaut, A., Drummond, A.J., 2007. Tracer v1.4, Available from
http://beast.bio.ed.ac.uk/Tracer.
Ronquist, F., Huelsenbeck, J.P., 2003. MrBayes 3: Bayesian phylogenetic inference
under mixed models. Bioinformatics 19, 1572–1574.
Rothmaler, W., 1943. Zur Gliederung der Antirrhineae. Feddes Repert. Spec. Nov.
Regni Veg. 52, 16–39.
Sánchez-Lafuente, A.M., 2007. Corolla herbivory, pollination success and fruit
predation in complex flowers an experimental study with Linaria lilacina
(Scrophulariaceae). Ann. Bot. 99, 355–364.
Sánchez-Lafuente, A.M., Rodríguez-Gironés, M.A., Parra, R., 2011. Interaction
frequency and per-interaction effects as predictors of total effects in plant
pollinator mutualisms: a case study with the self-incompatible herb Linaria
lilacina. Oecologia 168, 153–165.
Smith, S.D., 2010. Using phylogenetics to detect pollinator-mediated floral
evolution. New Phytol. 188, 354–363.
Specht, C.D., Bartlett, M.E., 2009. Flower evolution: the origin and subsequent
diversification of the angiosperm flower. Annu. Rev. Ecol. Evol. Syst. 40,
217–243.
Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic
analyses with thousands of taxa and mixed models. Bioinformatics 22,
2688–2690.
Stamatakis, A., Hoover, P., Rougemont, J., 2008. A rapid bootstrap algorithm for the
RAxML web servers. Syst. Biol. 57, 758–771.
Stebbins, G.L., 1970. Adaptive radiation of reproductive characteristics in
angiosperms, I: Pollination mechanisms. Annu. Rev. Ecol. Syst. 1, 307–326.
Stout, J.C., Allen, J.A., Goulson, D., 2000. Nectar robbing, forager efficiency and seed
set: bumblebees foraging on the self incompatible plant Linaria vulgaris
(Scrophulariaceae). Acta Oecol. 21, 277–283.
Sutton, D.A., 1988. A Revision of the Tribe Antirrhineae. Oxford University Press,
London.
Tamura, K., Peterson, D., Peterson, N., Stecher, G., Nei, M., Kumar, S., 2011. MEGA5:
molecular evolutionary genetics analysis using maximum likelihood,
evolutionary distance, and maximum parsimony methods. Mol. Biol. Evol. 28,
2731–2739.
Thompson, D.M., 1988. Systematics of Antirrhinum (Scrophulariaceae) in the New
World. Syst. Bot. Monogr. 22, 1–142.
Valdés, B., Díaz, Z., 1996. Habitual autogamy in Linaria tursica Valdés et Cabezudo
(Scrophulariaceae). Flora 199, 329–333.
Valente, L.M., Manning, J.C., Goldblatt, P., Vargas, P., 2012. Did pollination shifts
drive diversification in Southern African Gladiolus? Evaluating the model of
pollinator-driven speciation. Am. Nat. 180, 83–98.
van der Niet, T., Johnson, S.D., 2012. Phylogenetic evidence for pollinator-driven
diversification of angiosperms. Trends Ecol. Evol. 27, 353–361.
Vargas, P., Ornosa, C., Ortiz-Sánchez, F.J., Arroyo, J., 2010. Is the occluded corolla of
Antirrhinum bee-specialized? J. Nat. Hist. 44, 1427–1443.
Vargas, P., Roselló, J.A., Oyama, R., Güemes, J., 2004. Molecular evidence for
naturalness of genera in the tribe Antirrhineae (Scrophulariaceae) and three
independent evolutionary lineages from the New World and the Old. Plant
Syst. Evol. 249, 151–172.
Vargas, P., Valente, L.M., Liberal, I., Guzmán, B., Cano, E., Forrest, A.,
Fernández-Mazuecos, M., 2014. Testing the biogeographical congruence of
palaeofloras using molecular phylogenetics: snapdragons and the
Madrean–Tethyan flora. J. Biogeogr. 41, 932–943.
Whittall, J.B., Hodges, S.A., 2007. Pollinator shifts drive increasingly long nectar
spurs in columbine flowers. Nature 447, 706–712.
Willmer, P., 2011. Pollination and Floral Ecology. Princeston University Press,
Princeton, NJ.