The radiation of the genus Cheirolophus (Asteraceae) in Macaronesia constitutes a spectacular case of rapid diversification on oceanic islands. Twenty species – nine of them included in the IUCN Red List of Threatened Species – have been described to date inhabiting the Madeiran and Canarian archipelagos. A previous phylogenetic study revealed that the diversification of Cheirolophus in Macaronesia started less than 2 Ma. As a result of such an explosive speciation process, limited phylogenetic resolution was reported, mainly due to the low variability of the employed molecular markers. In the present study, we used highly polymorphic AFLP markers to i) evaluate species' boundaries, ii) infer their evolutionary relationships and iii) investigate the patterns of genetic diversity in relation to the potential processes likely involved in the radiation of Cheirolophus. One hundred and seventy-two individuals representing all Macaronesian Cheirolophus species were analysed using 249 AFLP loci. Our results suggest that geographic isolation played an important role in this radiation process. This was likely driven by the combination of poor gene flow capacity and a good ability for sporadic long-distance colonisations. In addition, we also found some traces of introgression and incipient ecological adaptation, which could have further enhanced the extraordinary diversification of Cheirolophus in Macaronesia. Last, we hypothesize that current threat categories assigned to Macaronesian Cheirolophus species do not reflect their respective evolutionary relevance, so future evaluations of their conservation status should take into account the results presented here.
Citation: Vitales D, García-Fernández A, Pellicer J, Vallès J, Santos-Guerra A, Cowan RS, et al. (2014) Key Processes for Cheirolophus (Asteraceae) Diversification on Oceanic Islands Inferred from AFLP Data. PLoS ONE 9(11): e113207. https://doi.org/10.1371/journal.pone.0113207
Editor: João Pinto, Instituto de Higiene e Medicina Tropical, Portugal
Received: February 13, 2014; Accepted: October 21, 2014; Published: November 20, 2014
Copyright: © 2014 Vitales et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was subsidized by the Spanish Ministry of Science (projects CGL2010-22234-C02-01 and 02/BOS) http://www.idi.mineco.gob.es/ and the Generalitat de Catalunya (Ajuts a grups de recerca consolidats, 2009/SGR/00439) http://www.gencat.cat/. DV benefited from a FPU grant from the Spanish Ministry of Education http://www.mecd.gob.es. JP benefited from a Beatriu de Pinós postdoctoral fellowship with the support of the Secretary for Universities and Research of the Ministry of Economy and Knowledge (Government of Catalonia) http://www.gencat.cat/ and the co-fund of Marie Curie Actions (European Union 7th R&D Framework Programme) http://ec.europa.eu/. AG-F benefited from a postdoctoral contract (Project CGL2010-22234-C02-01). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
In the last two decades, the Macaronesian archipelagos (Canary Islands, Cape Verde, Azores, Madeira and Savages) have attracted much interest from researchers studying plant diversification and radiation processes . These volcanic islands provide a wide variety of ecological conditions, geological ages and geographical isolation scales –, which promote the existence of a mosaic of habitats that represent an excellent natural laboratory in which to study selection forces and evolutionary processes. The Macaronesian archipelagos have been recognized as a hotspot of plant diversity , and have rapidly become a popular model system for scientists to test many speciation hypotheses, both empirically and theoretically. The wide diversity of habitats found in Macaronesia – spanning from xerophytic coastal cliffs to subalpine belts – has served to demonstrate the role of adaptive radiation in a range of plant groups (e.g. Argyranthemum Webb, ; Sonchus L. alliance, ; Aeonium Webb & Berthel., ; Echium L., ; Tolpis Adans., ). Indeed, niche pre-emption through adaptive radiation is the most prevailing hypothesis to explain the high degree of endemism and monophyly within Macaronesian lineages , . Furthermore, during the geological history of Macaronesian archipelagos, several islands have emerged, disappeared and/or changed their relative geographical position (see ), promoting complex isolation-connection and colonisation processes between islands and between islands and the continent. This complexity in volcanic archipelagos has given rise to numerous study cases examining the relative importance of vicariance versus dispersal in shaping insular biotas –; the role of islands as regions from which taxa might colonise continents and other archipelagos ,  and the different stages of colonisation and radiation processes in relation to the ontogeny phases of oceanic islands , .
Occasionally, island radiations occur over a short period of time, resulting in ecologically and morphologically distinct taxa, but leading to poor molecular differentiation , –. These cases, in which explosive species radiation takes place are among the most interesting and least understood evolutionary events, maybe due to the difficulty in carrying out species level analyses –. Rapid island radiations have been generally associated to some features – such as small population size, release from previous ecological constraints, and adaptation to new niches – recurrently observed in a wide variety of species and island archipelagos . However, since traditional phylogenetic studies usually provide little resolution in delimiting taxonomical boundaries and untangling the relationships among these rapidly evolving species, the precise role of morphological, life history, and physiological traits and their genetic basis in explosive plant radiations remain essentially unsolved .
The Macaronesian Cheirolophus Cass. (Asteraceae) complex comprises 20 species (out of the 30 constituting the whole genus), with Cheirolophus massonianus (Lowe) A.Hansen & Sunding occurring in the Madeiran archipelago, and the remaining species distributed across the western Canary Islands (see Fig. 1). Although they are usually associated with humid basalt cliffs, a few taxa have adapted to inhabit very diverse ecological zones of the archipelago , . Most of the island Cheirolophus are subshrubs, shrubs or even arborescent shrubs, showing a clear increase in woodiness relative to their continental congeners and a general shift towards inflorescences with white to purple flowers arranged in a candelabrum-like pattern. Some species are relatively widespread throughout a single island (e.g. C. webbianus (Sch.Bip.) Holub from Tenerife), whereas others are found on two different islands (e.g. C. teydis (C.Sm.) G.López from Tenerife and La Palma and C. massonianus from Madeira and Porto Santo). However, most species are narrow endemics occurring in a small number of restricted localities with only a few hundred or less individuals. Consequently, nine Macaronesian species have been included in the IUCN Red List of Threatened Species  as vulnerable, endangered or critically endangered. In addition, 17 of the species or subspecies endemic to the Canary Islands are included in the 2010 Red List of Spanish Vascular Flora . In those cases where many related taxa are under threat and conservation strategies must be prioritized, taxonomic knowledge becomes of special relevance and efforts should be made to establish accurately the boundaries of the species concept. To achieve this goal, the recent “unified species concept” advocates the use of diverse lines of evidence (e.g. monophyly at one or multiple DNA loci, morphological diagnosis, ecological distinctiveness, etc.) so that a higher degree of corroboration in taxonomic delimitation is attained . In recent times, conservation genetics has become an essential approach for evaluating the status and the level of threat at the species and population levels, identifying the genetic basis of processes which may potentially lead into an extinction vortex and informing authorities about management priorities for endangered species and/or populations .
Colour coding circles correspond to genetic structure derived from Bayesian mixture analysis of AFLP markers implemented in BAPS.
Several investigations based on morphological, cytogenetic, isozymes and DNA sequencing data have been previously used to infer evolutionary relationships and taxonomy for the 29 currently recognized species of Cheirolophus –. These studies revealed that the genus arose in the Mediterranean region with subsequent dispersal towards Macaronesia, but they failed to reconstruct the relationships among insular congeners. A recent phylogenetic study featuring relaxed-clock dating analyses and diversification tests (involving sequences of various nuclear and plastid regions and sampling several populations per species, ) has evidenced a recent origin of Macaronesian Cheirolophus radiation. The age of this group was estimated to be 1.74 Ma (95%HPD 0.82–2.93), implying a diversification rate of 0.34–2.84 species Myr−1. Indeed, this diversification rate is comparable to those exhibited by the Hawaiian Bidens L. (0.3–2.3 species Myr−1) or Macaronesian Echium (0.4–1.5 species Myr−1), considered as the fastest plant radiations on volcanic islands documented to date . Bayesian phylogeographic and ancestral range analyses were also applied in the same study, highlighting the major role of allopatric speciation to explain current diversity in Macaronesian Cheirolophus . Geographical isolation among populations and long distance dispersal (both intra and inter-island colonisations) were proposed as the main forces driving the rapid diversification of the group, but introgression and emerging ecological adaptation were also suggested as additional factors reinforcing the speciation process. Despite such efforts, the employed nrDNA and cpDNA markers provided very low variability within the Macaronesian species, thus hampering accurate inferences about the role played by those evolutionary mechanisms potentially involved in the radiation of Cheirolophus.
Amplified fragment length polymorphism (AFLP) analyses have been found to provide insights into interspecific relationships in diverse animal and plant groups when other methodologies have failed in this attempt (e.g. –). Indeed, several authors have suggested that this technique can be especially useful in reconstructing phylogenetic relationships among species that have diverged or radiated recently –. Although some considerations must be taken into account for its use, AFLP are currently widely employed in molecular ecology and evolution research –. In fact, this DNA fingerprinting approach has proven to be particularly suitable for evaluating the genetic structure of plant species in different oceanic archipelagos, and hence elucidating the potential evolutionary forces – such as gene flow or genetic drift – that influence the distribution of genetic diversity among individuals, populations, and species –.
AFLP fingerprinting – complemented with morphological and other molecular data – was employed here with the main objective of unravelling the explosive radiation that Cheirolophus underwent in the Canarian and Madeiran archipelagos. We applied phylogenetic and population genetic approaches to (i) evaluate the taxonomic identity of the Macaronesian species and to (ii) disentangle the evolutionary relationships between them. Additionally, we studied the patterns of genetic diversity within and between populations to (iii) infer the role of the evolutionary processes potentially involved with this explosive radiation (i.e. geographic isolation, ecological adaptation and introgression).
Materials and Methods
One hundred and seventy-two individuals from 29 populations of 20 Macaronesian Cheirolophus species were sampled, covering the whole taxonomic diversity recognized for the Madeiran and Canarian archipelagos (see Fig. 1 and Table S1 for further geographical details of sampling sites). One to 12 individuals per population were included in the study depending on the material availability and the uneven success of DNA extraction and AFLP procedures. We decided to incorporate those populations containing extremely scarce sampling (one or two individuals) because of their distinctiveness: (1) it was the only material available for the taxon (C. dariasii (Svent.) Bramwell; C. cf. webbianus); (2) the inclusion of that particular population was essential to understand the distribution of the species (C. teydis from La Palma; C. massonianus from Porto Santo); or (3) the additional subpopulation completes the sampling of extremely local species (C. metlesicsii Parada). Leaf material for DNA extraction was collected from plants in the field, dried in silica-gel and stored at room temperature. Insular Cabildos of Tenerife, Gran Canaria, La Gomera, La Palma and El Hierro provided all the necessary permits to collect samples from protected natural areas in the respective islands. The Canarian Council of Education, University and Sustainability issued the authorization to collect samples of the protected species listed on Table S1. Samples used of the endangered C. massonianus were provided by the Botanical Garden of Madeira.
DNA extraction, AFLP protocol and nrDNA sequencing
Genomic DNA was extracted from fragments of silica-gel dried leaf tissue following the protocol of Doyle and Doyle  with slight modifications. DNA samples were cleaned using QIAquick columns (Qiagen, Valencia, CA, USA) and their quality and DNA concentration were determined using NanoDrop ND-1000 spectrophotometry (ThermoScientific, Wilmington, DE, USA). The AFLP technique was carried out following the protocol described in Vos et al.  in accordance with the modified AFLP Plant Mapping Protocol (Applied Biosystems Inc. Foster City, CA, USA) using EcoRI and MseI with 500 ng of isolated genomic DNA per sample. Eighteen primer pair combinations were tested on six individuals from three different populations to screen those producing the most informative and readable profiles. Three pairs of primers were selected giving the most polymorphic and scorable polymorphic pattern: EcoRI-CTT/MseI-AC; EcoRI-CTC/MseI-AA; and EcoRI-CAG/MseI-AT. The success of each step was tested by running the PCR products on a 1.5% agarose gel. Fragments were run on an ABI Prism 3100 Genetic Analyzer (Applied Biosystems Inc.) with 10 µL of High Dye (deionized formamide) and 0.2 µL of GeneScan 500 ROX Size Standard per sample. Amplified fragments were genotyped as present/absent using GeneMarker AFLP/Genotyping software (version 1.9; SoftGenetics, LLC., State College, PA, USA).
For an initial scoring, all alleles within a range of 50 to 490 bp were considered. Afterwards, visual correction was carried out to eliminate erroneous peaks (low intensity or no reproducibility). AFLP error rates were calculated following . Twenty-five random samples per primer combination were replicated to ensure reproducibility, repeating all parts of the AFLP protocol (extraction, digestion, pre-selective and selective PCR). All alleles with an error rate >5% were eliminated. In order to test the occurrence of size homoplasy, we also calculated the correlation between AFLP fragment sizes and frequencies using AFLP-SURV 1.0 .
We also employed some nuclear ribosomal DNA sequences to examine the potential role of genetic introgression in the radiation of Macaronesian Cheirolophus. Only those populations (see Table S2) showing clear evidences of interspecific gene flow according to our AFLP data – together with some of their putative parental species – were studied. Sequencing procedures, GenBank accessions and any other information about the material and methods employed can be found in 
AFLP genetic diversity and population differences
The use of AFLP data (dominant markers) for estimating allelic frequencies implies the consideration of an outcrossing mating system and near random mating. This means that those populations would be under Hardy-Weinberg equilibrium . Cheirolophus has a predominantly outcrossing mating system and is pollinated by generalist insects, so one expects near random mating in the studied populations.
To estimate genetic diversity in each population showing more than one sampled individual, the following parameters were calculated: a) private alleles; b) rare alleles, present in <10% of the samples; and c) unbiased heterozygosity (Hj), calculated using TFPGA v.1.3 . Further measures of genetic diversity were estimated through: (i) the frequency-down-weighted marker values (DW) index of  using AFLPDAT ; (ii) the band richness (Br), which is the number of phenotypes expected at each locus, and can be interpreted as an analogue of the allelic richness, ranging from 1 to 2 ; and (iii) the percentage of polymorphic loci (PLP) with a significance of 1% (P = 0.99). Br and PLP indices were calculated according to the rarefaction method of Hurlbert , and conditioned to the smallest population size (N = 3) allowed by the software AFLPDIV v.1.0 (http://www.pierroton.inra.fr/genetics/labo/Software/Aflpdiv/). As a consequence, Br and PLP were not calculated for populations having less than three individuals. To assess whether genetic diversity indexes (i.e. Hj, DW, Br and PLP) differed among islands, diversity range values were compared in a non-parametric Kruskal-Wallis test using package RCMDR  implemented in R software .
To address questions of species delimitation and evolutionary relationships, AFLP data generated from the three selected primer pairs were combined into one large matrix and analysed together. Phylogenetic trees were reconstructed for this combined dataset using neighbor-joining (NJ) and Bayesian inference methods. NJ trees were built using Nei-Li distances  and 1000 replicates with NTSYS PC v.2.1 software . The support for specific nodes for the NJ tree was calculated using bootstrap  with 10000 replicates. Bayesian inference was performed in MrBayes v.3.1.2  using a F81-like model for restriction sites ,  and running four independent chains each of length 10 million sampling every 100 trees. Convergence was assessed using Tracer v.1.4.1  and a burn-in of 1 million trees was discarded. The remaining trees were used to construct a 50% majority rule consensus tree.
Additionally, non-tree-building approaches have been recommended to avoid conflicting phylogenetic signals in those cases in which the analyses of the data do not exhibit a bifurcating tree-like behaviour . Specifically, network methods have been proposed to resolve the uncertainty of these processes , . We used the Neighbor-Net method  carried out with SplitsTree v.4.10  to construct a distance-based network for the AFLP dataset using the Jaccard coefficient , which is restricted to shared band presence rather than shared absence.
Genetic structure: spatial patterns, Bayesian clustering and PCoA
We performed Mantel tests to evaluate the spatial effect in the genetic differences between populations using two similarity matrices. Genetic distance matrices – with and without C. massonianus from Madeira – were constructed with FST values between populations, and geographical matrices were calculated by the spatial distance (X and Y coordinates) between populations using ArcGIS v.9.1 (ESRI, Redlands, CA, USA). Mantel tests were performed with the vegan package  implemented in R software  using 100000 permutations and considering a p-value limit of 0.05. Because most populations were from the Canary Islands and only two of them are from Madeiran archipelago, distant from the remainder (>500 km), the Mantel test analysis was repeated without these populations.
Bayesian clustering analysis were carried out using STRUCTURE v.2.3 . We applied the admixture ancestry model and the correlated allele frequencies. Ten independent simulations were run for each possible number of genetic groups (K) from K = 1 to 29, using a burn-in period of 105 generations and run lengths of 5×105. To estimate the number of genetic groups (K) we selected the K value that maximizes the probability of the data L(K). We also considered the criterion proposed by Evanno et al.  to estimate the best value of K for our data set, based on the rate of change in the probability between successive K values, ΔK. Bayesian analyses of the genetic structure were also conducted with BAPS (Bayesian Analysis of Population Structure, Spatial Clustering of Groups, ), which uses stochastic optimization instead of Markov chain Monte Carlo to find the optimal partition. We performed a mixture analysis of individuals with the geographic origin of the samples used as an informative prior (‘spatial clustering of individuals') or without this prior (‘clustering of individuals’). BAPS simulations were run with the maximal number of groups (K) set from 1 to 29. Each run was replicated 10 times, and the results were averaged according to the resultant likelihood scores. The output of the mixture analyses were used as input for population admixture analysis , with the default settings in order to detect admixture between clusters.
Some authors have warned about a degree of over-splitting in the genetic structure analysed with AFLP markers (e.g. ) and have suggested the use of simpler statistical analyses to compare the pattern found by Bayesian methods. Similarities among individuals were also studied via Principal Coordinate Analysis (PCoA, ) using the Jaccard distance, in order to detect other possible relations that could not be visualized with assignment methods or phylogenetic analyses. This procedure was carried out with R software  using the vegan package .
Finally, we conducted analysis of molecular variance (AMOVA; ) using ARLEQUIN v.3.5  to estimate genetic differentiation following an alternative and widely used non-Bayesian approach that does not assume Hardy-Weinberg equilibrium or independence of markers. A first AMOVA analysis was implemented without taking regional structure into account. We also carried out four independent AMOVAs grouping the populations of Cheirolophus according to: i) their current taxonomic affiliation; ii) their island origin; iii) the most plausible model proposed by STRUCTURE; and, iv) the clustering proposed by BAPS. Pair-wise genetic distance between individuals using the square Euclidean distance was used for the AMOVA analyses, considering three levels: within populations, between populations within regions and between regions. Only two levels (within and between populations) were considered when no additional grouping structure was applied.
Morphological revision and analyses
Diagnostic morphological characters classically used in taxonomical treatments of Macaronesian Cheirolophus were also analysed in this study. Morphological traits were measured from herbarium vouchers, cultivated individuals and/or gathered from the literature depending on the species. In the present study, morphological data was only considered for a given species when at least three different individuals where available for measurements. We obtained accurate data from 16 species for six morphological variables widely employed to distinguish among Canarian Cheirolophus taxa (Table S3) (including leaf length, leaf width, size of capitula, plant size, floral colour and leaf shape). Qualitative traits (i.e. floral colour and leaf shape) were coded numerically (+1 for whitish and −1 for rose-colored flowers; +1 for entire, −1 for divided and 0 for intermediate/both leaf shapes). All the variables were standardized by subtracting the character mean from each species measure and then dividing by the character standard deviation. The resulting matrix was analysed by Principal Components Analysis (PCA) using RCMDR  implemented in R software .
In order to correlate genetic AFLP data with morphological traits, we carried out Mantel tests and a generalized analysis of molecular variance (GAMOVA; ). Mantel tests were performed with R software  and the package vegan  computing 100000 permutations and considering a p-value limit of 0.05. Complementary to Mantel test, GAMOVA analyses provides a regression-based method of the analysis of molecular variance (AMOVA; ) that could be interpreted as a multiple regression. GAMOVA approach is an especially suited tool to identify, associate and evaluate the relationships between the differences in a phenotype or environmental variable of interest between some populations and the genetic variations among the same populations. The significance between genetic differentiation (i.e. FST) and six morphological traits (see above) was tested by running a regression matrix with a permutation test with 10000 repetitions.
AFLP genotyping and filtering
Initially, 371 alleles were obtained from automatic genotyping. After manual correction, error rates calculation, elimination of small and troublesome alleles and low intensity peaks, a final matrix with 249 (67.1%) alleles were considered for subsequent analyses. The final data set obtained showed an error rate of 3.2%, which is below the maximum error rate percentage accepted for good AFLP reproducibility (5%) .
Population genetics: diversity and differentiation
Within-population genetic diversity measures are shown in Table S1. Private fragments were scarce across the studied populations; only three were detected, one each in C. santos-abreui A.Santos, C. satarataensis (Svent.) Holub and C. tagananensis (Svent.) Holub populations. The frequency of rare fragments was, conversely, much higher (10.6%), but most of these were shared by two or more populations. Hj diversity values were similar in all populations, ranging between Hj = 0.0459 (C. duranii (Burchard) Holub), and Hj = 0.1539 (C. arboreus (Webb & Berthel.) Holub from Los Tilos) in populations with more than three sampled individuals, with a mean value of Hj = 0.07892. The frequency-down-weighted marker values (DW) index showed overall high values, but considerable differences were observed among populations (ranging from 8.65 in C. junonianus (Svent.) Holub var. isoplexiphyllus to 19.75 in C. webbianus). The percentage of polymorphic loci (PLP %) ranged from 10.1% in C. duranii to 77.0% in C. arboreus (Los Tilos), and band richness (Br) ranged from 1.08 in C. junonianus var. junonianus to 1.46 in C. arboreus (Los Tilos). The DW was the only diversity index showing significant differences among islands (Kruskal-Wallis chi-squared = 15.02, df = 5, p-value<0.05); La Palma populations showed the lowest DW values. Pair-wise FST comparisons were significant (Table S4; mean value FST±SD = 0.259±0.185 with all populations, mean value FST±SD = 0.302±0.051 with populations with more than three individuals sampled). The biggest difference was found between C. puntallanensis A.Santos and C. arbutifolius (Svent.) G.Kunkel (FST = 0.577). The pair-wise comparisons between populations of the same species were much smaller than those between species (e.g. mean FST = 0.055 between C. ghomerythus (Svent.) Holub populations).
Phylogenetic analyses of Macaronesian Cheirolophus based on AFLP data
Trees constructed with the combined AFLP data using NJ and Bayesian estimation had generally similar topologies. Tree-building analyses of AFLP data resulted in similar fully-resolved assignment of individuals into species classically identified by diagnostic morphological and ecological characters (e.g. Bayesian 50% majority rule tree in Fig. 2; other trees not shown). However, species relationships were poorly resolved and only a few groups were partially identified. All trees included a large, strongly supported clade grouping most species from La Palma plus C. duranii from El Hierro (posterior probability, PP = 0.99; Fig. 2). Other supported smaller lineages grouped species from the Taganana peninsula (C. tagananensis, C. anagaensis A.Santos, C. cf. webbianus and C. cf. sp. nova; PP = 0.99) or species from the Teno mountains (C. canariensis, C. burchardii Susanna; PP = 1.00), in both cases from Tenerife. The phylogenetic analyses also confirmed the close relationship between C. satarataensis and C. dariasii (PP = 1.00), until recently considered as subspecies of the same species. The two populations of C. teydis from Tenerife and La Palma grouped together but were not embedded in a resolved clade. Similarly, C. arbutifolius and C. falcisectus Svent. ex Montelongo & Moraleda (both from Gran Canaria) appeared closely related (PP = 0.84) but their phylogenetic position in the genus was not fully resolved.
Unrooted 50% majority rule tree from Bayesian analysis of the combined AFLP dataset for all Macaronesian species of Cheirolophus. Posterior probability values ≥ 70 are shown near each branch.
The Neighbor-Net (NN) analysis, although indicating considerable reticulation in the data, resolved most of the species groups according to their current taxonomic boundaries (Fig. 3). In addition, this NJ network approach clustered most of the species in agreement with the Bayesian reconstruction depicted in Fig. 2. The species from La Palma formed a cluster closely related to the species from El Hierro (C. duranii). Those inhabiting eastern Tenerife (C. webbianus, C. tagananensis, C. anagensis, C. metlesicsii and C. cf. sp. nova) and western Tenerife (C. canariensis and C. burchardii) were also segregated into regional clades already resolved in the Bayesian and NJ phylogenetic analyses. Neighbor-Net analysis also revealed some phylogenetic conflicts not identified in the tree building analyses. For example, the species from El Hierro (C. duranii) showed clear reticulation, with an intermediate position between La Palma cluster and a lineage grouping two species from La Gomera (C. satarataensis and C. dariasii). Finally, NN analyses suggested a degree of reticulated connection between C. ghomerythus (from La Gomera) and C. massonianus (from Madeira), but this potential relationship was not supported by any Bayesian approach.
Bayesian clustering and spatial analyses of the genetic structure
Using the matrix of inter-population FST distances, and the matrix of geographical distances in kilometres, the Mantel test indicated a significant correlation between genetic and geographical distances (r = 0.258; p-value<0.05). Similar results were obtained if the Madeiran populations were excluded from the analysis (Mantel test r = 0.326, p-value<0.05).
The Bayesian analysis of population genetic structure conducted with STRUCTURE found the highest L(K) and ΔK values for K = 2. This grouping separated La Palma species (plus C. duranii from El Hierro) – cluster A – from the rest of the species from Tenerife, La Gomera, Gran Canaria and Madeira (including the C. teydis population from La Palma) – cluster B –, showing high percentages of individual memberships (97% and 98%, respectively) for these predefined groups (Fig. 4). According to the STRUCTURE analyses, the population of C. teydis from La Palma, some individuals of C. arboreus, C. duranii from El Hierro and one specimen of C. metlesicsii were the only ones presenting considerable (>10%) levels of admixture among the two genetic groups detected (between 14.9% and 34.3%). BAPS results showed a more fragmented distribution, with K = 11 as the most plausible number of clusters (P = 0.995; Figs. 1 and 3). The mixture analyses with or without spatially informative priors resulted in congruent assignment of individuals and no individual reassignment between the populations was observed. The genetic structure revealed by BAPS proved to be highly congruent with the phylogenetic analyses carried out (Fig. 3), showing at the same time a clear geographical pattern. Except for the populations of C. teydis and C. massonianus, the occurrence of genetic clusters determined by BAPS appeared to be limited to single islands and, in most cases, even restricted to particular regions within the islands (Fig. 1). No admixed individuals were detected according to this Bayesian clustering approach.
Bayesian estimation of genetic structure within Macaronesian Cheirolophus according to the best model proposed by STRUCTURE (K = 2).
The results of non-hierarchical AMOVA indicated a highly significant level of genetic structure among populations (56.79%; df = 29; p-value<0.01, Table 1). The hierarchical AMOVA analyses also showed significant genetic differentiation explained at all levels and for all grouping schemes tested (all p-values<0.01, Table 1). The model that explained a larger fraction of variation among groups (38.63%) was the one considering the current taxonomic circumscription (K = 20). Using geographical origin (i.e. native island source, K = 7) as the grouping criterion revealed slightly lower, but also significant, genetic variance (23.59%) among sites. According to the structure proposed by the Bayesian clustering analyses, genetic differentiation among groups explained 18.73% of variation (K = 2; STRUCTURE) and 33.29% of variation (K = 11; BAPS).
The PCoA using the first three principal coordinates explained 32.8% of the total variation in the data and confirmed several relationships detected in the phylogenetic and cluster analyses (Fig. S1). The first coordinate (accounting for 18.2% of the total variation) distinguished two main groups of species: the western Canarian taxa occurring on La Palma and El Hierro clearly segregated from another main cluster formed by species inhabiting Tenerife, Gran Canaria, La Gomera and Madeira. The second coordinate (representing 7.9% of variation) segregated C. ghomerythus populations in a different cluster, confirming the distinctiveness of this species from the other taxa on La Gomera. The addition of this second axis also separated C. burchardii, C. canariensis and C. arbutifolius accessions in one discrete cluster and C. massonianus individuals in another. Finally, plotting a third coordinate (accounting for 6.7% of the total variation) with the first axis, led to the distinction of four discrete clusters of species from: i) Tenerife and Gran Canaria; ii) La Gomera; iii) La Palma and El Hierro; and iv) Madeira.
Morphological analyses and nrDNA sequencing data
The two first principal components represented 60% of the total morphological variation in the data set (Table S3). The first component (PC1) accounted for 31.4% of the total variance, showing the highest coefficients for flower colour and leaf shape traits (Table S3). PC2 represented another 28.4% of the total variance, having the highest coefficients for capitulum and leaf size and plant height traits. No correlation was found between genetic and morphological distances according to Mantel test (r = 0.00083, p-value>0.05), nor did any of the GAMOVA analysis show significant results for the morphological variables considered (all p-values>0.05).
The analysis of nrDNA sequences from  evidenced a considerable number of heteromorphic sites (double peaks) within potentially introgressed species. Table S2 shows nine positions (four observed in ITS regions and five in ETS region) presenting multiple peaks on the electropherograms in seven Macaronesian species.
Species delimitation among Macaronesian Cheirolophus
The performance of diverse methods to analyse AFLP data sets has been found to be particularly important in evolutionary studies of plant radiations , , , . Most of the analyses conducted in this study to evaluate the genetic structure of Cheirolophus in Macaronesia revealed different but complementary patterns, supporting the complex evolutionary history previously suggested for the genus in the archipelago . Phylogenetic analyses based on AFLP were particularly useful in unravelling taxonomic boundaries among Macaronesian congeners. Both tree-building and network methods provided full support to the current species circumscription mainly based on morphological characteristics. Thus, our results corroborate the distinctiveness of these extraordinarily recent diverged species and support the utility of classical diagnostic characters employed in the taxonomical delimitation of Macaronesian Cheirolophus.
Geographic differentiation and long-distance dispersal
AFLP analyses were not able to thoroughly disentangle the phylogeny of Macaronesian Cheirolophus, but assignment methods provided further evidence on the evolutionary relationship between different lineages. The model proposed by STRUCTURE (K = 2) identified a major geographical division, with the western islands (La Palma and El Hierro) clustering independently from the Central-Eastern ones (La Gomera, Tenerife, Gran Canaria), that grouped together with Madeira (Fig. 4). This genetic barrier has been also found in PCoA analyses (Fig. S1) and is supported by the phylogenetic signal (Figs. 2 and 3). Close evolutionary relationships among species from La Palma and El Hierro have already been reported in some studies concerning diverse animal and plant taxa from Macaronesia –. However, the majority of phylogeographic studies on Canarian flora and fauna have inferred that endemics from El Hierro appear to be more closely related to species from La Gomera than to those from La Palma . Indeed, plastid DNA phylogeographic analysis has recently revealed a clear evolutionary connexion among El Hierro and La Gomera haplotypes , which provides evidence for a series of incongruent patterns among AFLP and cpDNA data (see discussion on introgression section below). A model with more clusters was proposed by BAPS (K = 11), suggesting as well significant influence of allopatric speciation in the evolutionary history of Macaronesian Cheirolophus. The species from Tenerife are grouped in three clusters corresponding to well differentiated regions of the island. One cluster comprised the taxa inhabiting mainly the eastern part of Tenerife (i.e. C. webbianus, C. tagananensis, C. anagaensis, C. metlesicsii and C. cf. sp. nov.). Another cluster groups C. canariensis and C. burchardii from the Teno mountains, at the western end of the island. This eastern/western pattern of geographic distribution in Tenerife has been recovered in other studies of several plant groups (see , for a review). In some cases, this segregation has been explained by disjunct evolution of lineages in two palaeo-islands of Tenerife , , currently corresponding to Anaga and Teno massifs at different ends of the island. According to previous phylogeographic and dating analyses , this geographic splitting of Cheirolophus lineages from eastern and western Tenerife could be related to connection-isolation cycles caused by volcanic activity during the last 2 My . A third cluster included the populations of C. teydis from Tenerife (Las Cañadas) and La Palma, all of them living around the subalpine zone (1800–2200 m) on both islands. The representatives from Gran Canaria were placed together by BAPS in a separate cluster. This result may support the recent hypothesis of a single colonisation event for this island and the subsequent diversification process giving rise to C. arbutifolius and C. falcisectus . Our Neighbor-Net and PCoA analyses pointed to an evolutionary closeness among Gran Canaria and Tenerife lineages, but we were not able to reconstruct their accurate phylogeographic relationship. Nonetheless, in agreement to the BAPS results, the species from La Gomera are grouped in two different clusters showing a clear geographic distribution pattern: one including C. ghomerythus from the northern coast of the island and the other one embedding C. satarataensis and its former subspecies C. dariasii, from the south and south-west of La Gomera. Certain evolutionary closeness among the two lineages from La Gomera was detected in the PCoA analyses, but their monophyly could not be confirmed. BAPS method also differentiated four geographically-related clusters of species in La Palma: C. junonianus from the south of the island; C. arboreus from the north and the west; C. santos-abreui, C. puntallanensis and C. sventenii (A.Santos) G.Kunkel from the north and north-eastern part of La Palma; and C. teydis from the summits of this island (and from Las Cañadas in Tenerife). Finally, C. duranii from El Hierro and C. massonianus from Madeira grouped in two separate clusters, suggesting as well that geographic isolation has been involved in the radiation of the genus.
According to these results, sporadic long-distance dispersal (LDD) events have to be considered to explain the numerous intra and inter-island colonisations resulting in the current distribution of genetic and taxonomic diversity in Macaronesian Cheirolophus. Unfortunately, straightforward evidences of LDD are very difficult to obtain , and we have not been able to provide direct proofs of these events occurring on our group of study. However, several supporting evidences (discussed below) suggest that Macaronesian Cheirolophus might have the potential to undergo successful LDD events. Birds and lizards have been found to be involved in seed long- dispersal of multiple species in the Canary Islands ,  and other oceanic archipelagos , . In the Galapagos and Azores, finches (granivorous birds from family Fringillidae) have been reported as legitimate seed dispersers of dry-fruited plants – including Asteraceae – implicated in LDD events between islands –. The goldfinch (Carduelis carduelis L.) has been observed in the Canary Islands feeding on Cheirolophus seeds , which suggests that bird-mediated dispersal could also be an important LDD mechanism in Macaronesian species. The role of birds in LDD events in Cheirolophus was also pointed by Garnatje et al. , who inferred that seabirds could have mediated the re-colonisation event from the Balearic Islands towards the continent in C. intybaceus (Lam.) Dostál. Finally, viable seeds from another Centaureinae species have been recovered from lizard guts discarded by predatory birds in the Canarian archipelago , supporting this kind of secondary seed dispersal as a likely mechanism involved in the LDD of Macaronesian Cheirolophus. All these data suggest that Cheirolophus achenes may be able to be transported through long distances but, given the lack of strong direct evidences in our case study and the inherent stochasticity of LDD process , speculation about potential role of these or other animals as LDD vectors must be made with extreme caution.
Furthermore, the capacity for successful long distance colonisation in Cheirolophus could also be enhanced by certain intrinsic biological features showed by this group of plants. The genus presents a pseudo-self-compatible mating system , that may be able to originate a sexually reproducing population from a single propagule, carrying at the same time more genetic variation than a seed from an autogamous population . In summary, the results of our AFLP analysis are consistent with a role of allopatric divergence in the radiation of Macaronesian Cheirolophus; a hypothesis apparently supported by additional bibliographic evidences suggesting a certain degree of ability for successful long-distance colonisation events in the genus.
Geographic-genetic correlation and limited gene flow in Macaronesian Cheirolophus
Genetic isolation has been proposed as a major factor determining plant speciation on oceanic islands . The significance found in the spatial explicit analysis (Mantel test) suggests that Macaronesian Cheirolophus are influenced by geographic-genetic correlation across species. In particular, our results indicate that the more closely evolutionary-related species are as well geographically closer to each other, which is in good agreement with a gene flow scenario dominated by short distance events . Regular seed dispersal in Cheirolophus has been reported to be limited to very short distances (see  for some species description). This genus shows relatively heavy and pappus-lacking cypselas that are unable to be transported by the wind, falling by gravity beside the mother plant during dissemination . Another important factor enhancing geographic-genetic correlation across species might be topographic isolation . Most Cheirolophus taxa inhabit deep ravines, coastal cliffs or steep slopes , , additionally preventing regular seed dispersal through long distances. The large number of Macaronesian Cheirolophus species inhabiting a few restricted and isolated locations (e.g. microendemisms such as C. anagaensis, C. burchardii, C. dariasii, C. falcisectus, C. junonianus, C. metlesicsii, C. puntallanensis, C. santos-abreui or C. tagananensis) could be related to the additive effect of poorly connected habitats with the limited seed dispersal capacity of Cheirolophus.
Our analysis across the whole distribution range of Macaronesian Cheirolophus revealed overall low levels of genetic diversity. This result is in accordance with the general expectation that endemic species, and particularly island endemics , exhibit lower levels of genetic diversity than widespread species , . Generally, microendemics occupying restricted and isolated populations (e.g. C. anagaensis, C. duranii, C. junonianus, C. massonianus, C. cf. sp. nova, or C. tagananensis) show lower genetic diversity levels than species with numerous, widely distributed populations (e.g. C. arboreus, C. ghomerythus, C. sventenii, or C. teydis). However, this expected pattern is contradicted by certain populations of widely distributed species (e.g. C. arbutifolius, C. satarataensis) also showing lower genetic diversity values. Considering the limited number of sampled populations per species and the low number of individuals available, these indices could be underestimating genetic diversity, particularly in these widespread species. We found only three populations showing private fragments, probably due to the low sampling size within some populations and the strict choice of polymorphic bands.
Some of the genetic diversity indices studied in the Macaronesian Cheirolophus can also be compared with the values reported by Garnatje et al.  in the Mediterranean complex of C. intybaceus, which includes some taxa distributed in the eastern Iberian Peninsula and the Balearic Islands. Diversification in the C. intybaceus complex presumably started in the same period as the Macaronesian radiation , but in the Mediterranean group it resulted in only four doubtful species. Heterozygosity levels detected in populations of Macaronesian species are significantly lower than in Mediterranean populations of the C. intybaceus complex. Lower genetic variation has been associated with species showing limited geographical distribution, smaller populations and exhibiting little gene flow –. Our results could be reflecting that Macaronesian Cheirolophus – distributed in small populations across islands with steeply dissected topography – have been comparatively more isolated than Mediterranean ones, thus contributing to their progressive genetic divergence. As it has been proposed for other plant groups by Ellis et al.  and Knope et al. , the combination of certain ability for sporadic long-distance colonisation – see the section above – and poor gene flow capacity – due to both intrinsic and geographic characteristics – could have played an important role enhancing the explosive diversification of Cheirolophus in Macaronesia.
Geographic isolation may have been important for enhancing diversification but ecological adaptation is as well a common mechanism contributing – either in allopatry  or in sympatry  – to island plant speciation. According to Whittaker and Fernández-Palacios , the diversification of Cheirolophus in the Macaronesian archipelagos can be considered an example of non-adaptive radiation occurred on oceanic islands. Indeed, most species of this group exploit very similar niches in different islands, showing minor morphological differences. However, Macaronesian Cheirolophus present as well a few cases of taxa that have apparently adapted their morphology to the significantly diverse ecological conditions found on these archipelagos. Cheirolophus teydis is the only species of the genus that occupies the subalpine habitat, showing morphological adaptations to tougher ecological conditions (i.e. rosette-like disposed leaves with reduced laminas; waxy leaf cover; high number of smaller flowers; annual flowering shoots). The species from Gran Canaria – C. arbutifolius and C. falcisectus – are the result of a diversification process originated after a single colonisation of the island (see above). These species have diverged into different niches allopatrically within the same island – C. falcisectus inhabits more xeric habitats and shows clear leaf reduction while C. arbutifolius occupies more humid locations and present an arborescent habit and a larger leaf surface – thus suggesting an additional example of ecological differentiation. A similar case of ecological adaptation can be found in La Palma, where C. junonianus – from the south of the island and inhabiting significantly more arid localities than the rest of species from the northern part of this island – shows parallel morphological adaptation in size and leaf shape to drought conditions. Equivalent eco-morphological responses have already been reported in other Macaronesian plant taxa that have undergone an adaptive radiation process (e.g. Argyranthemum ). Interestingly, our morphology-genetics correlation analyses indicate that morphological similarity across the different Macaronesian Cheirolophus species is not affected by genetic similarity. Certainly, phylogenetic and clustering analyses are not congruent with a relationship between the main lineages and ecology: some species occupying different niches share the same or close genetic groups/lineages (e.g. C. falcisectus and C. arbutifolius; or C. junonianus and C. arboreus), whilst species showing similar habitats and morphological features frequently belong to very different genetic clusters/lineages (e.g. C. duranii and C. tagananesis; or C. ghomerythus and C. burchardii). These patterns suggest that the few cases of ecological adaptation (cited above) do not seem to be the result of a single eco-morphologic shift occurring at initial stages of the Macaronesian Cheirolophus diversification, but they may correspond to more recent, multiple and independent phenotype-environment differentiation processes.
Clearly, there are not enough data here to discard a vital role of selection in the radiation of Cheirolophus. We only measured a few morphological traits usually employed to delimitate taxonomically the Macaronesian species (Table S3), but other potentially important morphological and physiological features could have been missed. In addition, there is no accurate evaluation of the niches occupied by the different species, so there could also be fine-scale ecological variables differentiating habitats formerly considered as equivalent. Therefore, more precise inferences about the relative importance of ecological adaptation in this radiation process would require additional intraspecific sampling, supplementary morphological, physiological and ecological measurements as well as more appropriate tests (see , , ). Further studies applying these methodologies will improve our understanding of the role played by selection versus neutral differentiation in islands diversifications.
Evidence of interspecific gene flow
Introgression is another mechanism formerly proposed to have played a role in the evolutionary history of Macaronesian Cheirolophus . For instance, the species from Madeira, C. massonianus, was proposed to have come into contact with a continental congener, resulting in a chloroplast capture event. The genetic imprint of this hybridization event was not detected in the nrDNA regions sequenced in that former study, grouping C. massonianus within the Canarian clade and suggesting that introgression – via plastid transfer (see ) – did not affect the nuclear genome. From our AFLP data, pair-wise FST comparisons between Madeiran and Canarian populations showed similar genetic differentiation values to comparisons within Canarian populations (Table S4), apparently supporting the overall closeness among nuclear genomes of species from both archipelagos. Unfortunately, we did not include in this AFLP study any continental species putatively involved in C. massonianus hybridization, so our analyses do not allow assessment of whether introgression was limited to the chloroplast genome or also affected the nuclear genome.
In contrast, traces of genetic admixture were detected from AFLP data by STRUCTURE analysis in a population of C. teydis from La Palma (Fig. 4). Our results suggest that C. teydis originated in Tenerife and colonised subsequently La Palma, where genetic contact with other species from this island may occur. Plastid DNA analyses  support this hypothesis, pointing to a process of plastid capture to explain haplotypic differentiation among C. teydis populations from Tenerife and La Palma. In the same way, few C. arboreus individuals from populations in North-West La Palma show some genetic introgression among both genetic clusters defined by STRUCTURE. It has been reported that some specimens of C. arboreus from this part of the island exhibit morphological traits significantly distinct from the type . Indeed, one of these C. arboreus populations from NW La Palma (Bco. Briestas) shows a different cpDNA haplotype from the rest of populations . Another potential case of introgression could be affecting the species from El Hierro, C. duranii. This species is grouped together with the rest of taxa from La Palma according to our phylogenetic, PCoA and clustering analyses (Figs. 2, 4 and S1). In contrast, plastid DNA phylogenetic analysis , revealed a clear evolutionary closeness among C. duranii and the species from La Gomera. The Neighbor-Net analysis of our AFLP data mainly supports the evolutionary relationship among C. duranii from El Hierro and the rest of species from La Palma, but it also shows a faint reticulation signal between C. duranii and C. satarataensis from La Gomera (Fig. 3). In this case, certain genetic admixture – albeit very weak – can also be perceived from the STRUCTURE analysis, suggesting as well that this species from El Hierro may present genetic traces from both La Palma and La Gomera Cheirolophus species. Finally, the occurrence of introgression events during the radiation process could be also supported by the polymorphic sites observed in nrDNA sequences of the species here mentioned (see Table S2).
These results suggest that introgression might have played certain role in the evolutionary history of Cheirolophus in Macaronesia. However, similar patterns can be generated by ancestral polymorphisms and incomplete lineage sorting . The STRUCTURE analysis does provide indication of admixed genotypes in some species, but at K = 2 the parent species for introgression are impossible to determine, being perhaps a vestige from some ancestral polymorphisms. According to the phylogeographic analysis performed by Vitales et al. , La Palma and El Hierro were colonised by Cheirolophus from Tenerife Island, so the genetic cluster A – mainly found in La Palma and El Hierro – proposed by STRUCTURE should be derived from the putative older cluster B, predominant in Tenerife, La Gomera and Gran Canaria (see Fig. 4). The admixture signal observed in C. duranii or in NW C. arboreus populations fits well with this alternative scenario considering certain retention of ancestral polymorphisms. In contrast, the pattern observed in C. teydis from La Palma – showing considerable admixture signal from the derived cluster B – seems more difficult to explain by the only action of ancient polymorphism retention. Heteromorphic positions found in nrDNA are consistent with some genetic reticulation but they can also be attributed to retention of polymorphisms and incomplete concerted evolution , especially considering the rapidity of this radiation process. In summary, there are some evidence for genetic introgression during the diversification of Macaronesian Cheirolophus, but in most cases there are alternative possible explanations (e.g. ancestral polymorphisms), thus limiting our conclusions about the relative importance of reticulation in the evolutionary history of the group.
The genus Cheirolophus is illustrating one of the largest plant radiations in the Canary Islands . Having nine of the 20 extant insular species included in the IUCN Red List , Cheirolophus shows the highest proportion of endangered taxa for any species-rich lineage in this archipelago. Thus, attending to the major conservational interest of the group, we consider capital to analyse the results of this study from a conservation point of view. Even though Macaronesian Cheirolophus are the result of an exceptionally recent diversification, phylogenetic analyses based on AFLP confirmed the evolutionary identity of currently described endemics. Generally, the genetic diversity indexes calculated for the different populations and species of the group were found to be relatively heterogeneous, reflecting the complexity of this radiation process (see above). However, the frequency-down-weighted fragment values per population (DW) provided interesting and useful information about conservation biology of the numerous endangered representatives of the genus. This rarity index (DW), has been employed to assess the genetic distinctiveness of populations and species , and can also be used as an indicator of uniqueness and evolutionary relevance for conservation. Previous works suggest that species with high levels of unique genetic information are more likely to be threatened (e.g. , ). However, according to our results, current threat categories assigned to Macaronesian Cheirolophus species do not reflect their uneven evolutionary differentiation. Cheirolophus taxa considered in higher extinction risk categories (CR and EN) in the Spanish Red List of Endangered Flora  and IUCN Red List of Threatened Species  have relatively low DW values (e.g. C. arboreus, C. duranii, C. metlesicsii, C. santos-abreui). In contrast, those species showing higher DW values - therefore considered more genetically distinct – are assigned to lower extinction risk categories (VU) or even considered unthreatened (e.g. C. canariensis, C. satarataensis, C. webbianus). As resources for conservation are limited, their optimal allocation is essential . Prioritization may be especially important when dealing with rich groups of closely related endemics inhabiting biodiversity hotspots –, as is the case of Macaronesian Cheirolophus. We suggest that future evaluations of the endangered status of Macaronesian Cheirolophus should take into account the evolutionary distinctiveness results presented here. Predictably, the prioritization in resources allocation among Cheirolophus species would change if their genetic differentiation level is considered during conservation assessment process.
Principal coordinates (PCoA) plot of AFLP data for the Macaronesian Cheirolophus populations included in this study. Different symbols correspond to different populations as shown in the legend in the right side. Some populations groups that are well-differentiated and/or mentioned in the text are circled and named.
Sampling information and genetic diversity indexes assessed. Taxa, locality, code, number of sampled individuals (N), and genetic diversity indexes assessed by AFLP in 29 populations of Cheirolophus from Macaronesia. Genetic indices: number of private fragments (fu); heterozygosity (Hj); percentage of polymorphic loci for a standardised sample size of three (PLP 1%); band richness for a standardised sample size of three (Br); and frequency-down-weighted marker values index (DW).
Polymorphic positions in nrDNA sequencing of some Canarian Cheirolophus taxa. Positions in ITS and ETS sequences of some Canarian Cheirolophus specimens where more than one base is represented in a single amplification product, seen as subequal multiple peaks on the electropherograms. Data of sequences from Vitales et al. (2014).
Data from diagnostic morphological characters of Canarian Cheirolophus species. Character loadings in first two principal components for the analysis of Cheirolophus morphological data (high loadings are highlighted in boldface type).
Laia Barres, Roser Vilatersana and the Botanical Garden of Madeira are acknowledged for their contribution to supplying C. massonianus material. We also thank Miquel Veny for keeping the collections of living plants.
Conceived and designed the experiments: TG JV JP DV. Performed the experiments: RSC MFF DV. Analyzed the data: AG-F OH DV. Contributed reagents/materials/analysis tools: AS-G MFF RSC OH. Wrote the paper: AG-F TG JV JP DV. Reviewed and improved the manuscript: AS-G OH MFF RSC.
- 1. Whittaker RJ, Fernández-Palacios JM (2007) Island biogeography: Ecology, evolution and conservation. Oxford Uni. Oxford.
- 2. Humphries CJ (1979) Endemism and evolution in Macaronesia. In: Bramwell D, editor. Plants and islands. London. pp.171–200.
- 3. Carracedo JC (1994) The Canary Islands: an example of structural control on the growth of large oceanic-islands volcanoes. J Volcanol Geotherm Res 60: 225–241.
- 4. Fernández-Palacios JM, De Nascimento L, Otto R, Delgado JD, García-del-Rey E, et al. (2011) A reconstruction of Palaeo-Macaronesia, with particular reference to the long-term biogeography of the Atlantic island laurel forests. J Biogeogr 38: 226–246.
- 5. Caujapé-Castells J (2011) Jesters, red queens, boomerangs and surfers: a molecular outlook on the diversity of the Canarian endemic flora. In: Bramwell D, Caujapé-Castells J, editors. The biology of islands floras. Cambridge: Cambridge University Press. pp.284–234.
- 6. Francisco-Ortega J, Jansen RK, Santos-Guerra A (1996) Chloroplast DNA evidence of colonization, adaptive radiation, and hybridization in the evolution of the Macaronesian flora. Proc Natl Acad Sci U S A 93: 4085–4090.
- 7. Kim SC, Crawford DJ, Francisco-Ortega J, Santos-Guerra A (1996) A common origin for woody Sonchus and five related genera in the Macaronesian islands: molecular evidence for extensive radiation. Proc Natl Acad Sci U S A 93: 7743–7748.
- 8. Jorgensen TH, Olesen JM (2001) Adaptive radiation of island plants: evidence from Aeonium (Crassulaceae) of the Canary Islands. Perspect Plant Ecol Evol Syst 4: 29–42.
- 9. García-Maroto F, Mañas-Fernández A, Garrido-Cárdenas JA, Alonso DL, Guil-Guerrero JL, et al. (2009) Delta6-desaturase sequence evidence for explosive Pliocene radiations within the adaptive radiation of Macaronesian Echium (Boraginaceae). Mol Phylogenet Evol 52: 563–574.
- 10. Gruenstaeudl M, Santos-Guerra A, Jansen RK (2012) Phylogenetic analyses of Tolpis Adans. (Asteraceae) reveal patterns of adaptative radiation, multiple colonization and interspecific hybridization. Cladistics 1: 1–19.
- 11. Silvertown J, Francisco-Ortega J, Carine M (2005) The monophyly of island radiations: an evaluation of niche pre-emption and some alternative explanations. J Ecol 93: 653–657.
- 12. Kim S-C, McGowen MR, Lubinsky P, Barber JC, Mort ME, et al. (2008) Timing and tempo of early and successive adaptive radiations in Macaronesia. PLoS One 3. doi:10.1371.
- 13. Grehan J (2001) Biogeography and evolution of the Galapagos: integration of the biological and geological evidence. Biol J Linn Soc 74: 267–287.
- 14. Heads M (2010) The endemic plant families and the palms of New Caledonia: a biogeographical analysis. J Biogeogr 37: 1239–1250.
- 15. Renner SS, Strijk JS, Strasberg D, Thébaud C (2010) Biogeography of the Monimiaceae (Laurales): a role for East Gondwana and long-distance dispersal, but not West Gondwana. J Biogeogr 37: 1227–1238.
- 16. Carine MA, Russell SJ, Santos-Guerra A, Francisco-Ortega J (2004) Relationships of the Macaronesian and Mediterranean Floras: molecular evidence for multiple colonizations into Macaronesia and back-colonization of the continent in Convolvulus (Convolvulaceae). Am J Bot 91: 1070–1085.
- 17. Bellemain E, Ricklefs RE (2008) Are islands the end of the colonization road? Trends Ecol Evol 23: 461–468.
- 18. Stuessy TF, Jakubowsky G, Gomez RS, Pfosser M, Schluter PM, et al. (2006) Anagenetic evolution in island plants. J Biogeogr 33: 1259–1265.
- 19. Whittaker RJ, Triantis KA, Ladle RJ (2008) A general dynamic theory of oceanic island biogeography. J Biogeogr 35: 977–994.
- 20. Richardson JE, Weitz FM, Fay MF, Cronk QCB, Linder HP, et al. (2001) Phylogenetic analyses of Phylica L. (Rhamnaceae) with an emphasis on island species: evidence from plastid trnL-F and nuclear internal transcribed spacer (ribosomal) DNA sequences. Taxon 405–427.
- 21. Mort ME, Crawford DJ, Archibald JK, O'Leary TR, Santos-Guerra A (2010) Plant DNA barcoding: A test using Macaronesian taxa of Tolpis (Asteraceae). Nat Hist 59: 581–587.
- 22. Knope ML, Morden CW, Funk VA, Fukami T (2012) Area and the rapid radiation of Hawaiian Bidens (Asteraceae). J Biogeogr 39: 1206–1216.
- 23. Albertson RC, Markert JA, Danley PD, Kocher TD (1999) Phylogeny of a rapidly evolving clade: the cichlid fishes of Lake Malawi, East Africa. Proc Natl Acad Sci U S A 96: 5107–5110.
- 24. Linder HP (2008) Plant species radiations: where, when, why? Philos Trans R Soc Lond B Biol Sci 363: 3097–3105.
- 25. Glor RE (2010) Phylogenetic insights on adaptive radiation. Annu Rev Ecol Evol Syst 41: 251–270.
- 26. Mzighani SI, Nikaido M, Takeda M, Seehausen O, Budeba YL, et al. (2010) Genetic variation and demographic history of the Haplochromis laparogramma group of Lake Victoria–An analysis based on SINEs and mitochondrial DNA. Gene 450: 39–47.
- 27. Bromham L, Woolfit M (2004) Explosive radiations and the reliability of molecular clocks: island endemic radiations as a test case. Syst Biol 53: 758–766.
- 28. Kim S-C (2012) Mapping unexplored genomes II: genetic architecture of species differences in the woody Sonchus alliance (Asteraceae) in the Macaronesian Islands. J Plant Res 125: 125–136.
- 29. Bramwell D (1976) The endemic flora of the Canary Islands; distribution, relationships and phytogeography. In: Kunkel G, editor. Biogeography and ecology in the Canary Islands. The Hague: Junk Publishers. pp.207–240.
- 30. Bramwell D, Bramwell Z (2001) Wild flowers of the Canary Islands. 2nd ed. Madrid: Editorial Rueda.
- 31. IUCN (2012) IUCN Red List of Threatened Species Version 2012.2. Available: http://www.iucnredlist.org/. Accessed 1 February 2012.
- 32. Bañares A, Blanca G, Güemes J, Moreno JC, Ortiz S (2010) Atlas y libro Rojo de la Flora Vascular Amenazada de España. Adenda 2010. Madrid: Dirección General de Medio Natural y Política Forestal y Sociedad Española de Biología de la Conservación de Plantas.
- 33. De Queiroz K (2007) Species concepts and species delimitation. Syst Biol 56: 879–886.
- 34. DeSalle R, Amato G (2004) The expansion of conservation genetics. Nat Rev Genet 5: 702–712.
- 35. Dostál J (1976) Cheirolophus. In: Tutin TG, Heywod VH, Burges NA, Moore DM, Valentine DH, et al. editors. Flora Europaea. Cambridge: Cambridge University Press. pp.249–250.
- 36. Bremer K (1994) Asteraceae. Cladistics and classification. Portland: Timber Press.
- 37. Garnatje T, Susanna A, Messeguer R (1998) Isozyme studies in the genus Cheirolophus (Asteraceae: Cardueae-Centaureinae) in the Iberian Peninsula, North Africa and the Canary Islands. Plant Syst Evol 213: 57–70.
- 38. Susanna A, Garnatje T, Garcia-Jacas N (1999) Molecular phylogeny of Cheirolophus (Asteraceae:Cardueae-Centaureinae) based on ITS sequences of nuclear ribosomal DNA. Plant Syst Evol 214: 147–160.
- 39. Garnatje T, Garcia S, Canela MÁ (2007) Genome size variation from a phylogenetic perspective in the genus Cheirolophus Cass. (Asteraceae): biogeographic implications. Plant Syst Evol 264: 117–134.
- 40. Garnatje T, Hidalgo O, Vitales D, Pellicer J, Vallès J, et al. (2012) Swarm of terminal 35S in Cheirolophus (Asteraceae, Centaureinae). Genome 55: 529–535.
- 41. Vitales D, Garnatje T, Pellicer J, Vallès J, Santos-Guerra A, et al. (2014) The explosive radiation of Cheirolophus (Asteraceae, Cardueae) in Macaronesia. BMC Evol Biol 14: 118.
- 42. Koopman WJM, Zevenbergen MJ, Berg RG Van den (2001) Species relationships in Lactuca s.l. (Lactuceae, Asteraceae) inferred from AFLP fingerprints. Am J Bot 88: 1881–1887.
- 43. Després L, Gielly L, Redoutet B, Taberlet P (2003) Using AFLP to resolve phylogenetic relationships in a morphologically diversified plant species complex when nuclear and chloroplast sequences fail to reveal variability. Mol Phylogenet Evol 27: 185–196.
- 44. Richardson JE, Fay MF, Cronk QCB, Chase MW (2003) Species delimitation and the origin of populations in island representatives of Phylica (Rhamnaceae). Evolution 57: 816–827.
- 45. Guo Y-P, Saukel J, Mittermayr R, Ehrendorfer F (2005) AFLP analyses demonstrate genetic divergence, hybridization, and multiple polyploidization in the evolution of Achillea (Asteraceae-Anthemideae). New Phytol 166: 273–289.
- 46. Mendelson TC, Shaw KL (2005) Rapid speciation in an arthropod. Nature 433: 375–376.
- 47. Díaz-Pérez A, Sequeira M, Santos-Guerra A, Catalán P (2008) Multiple colonizations, in situ speciation, and volcanism-associated stepping-stone dispersals shaped the phylogeography of the Macaronesian red fescues (Festuca L., Gramineae). Syst Biol 57: 732–749.
- 48. Meudt HM, Lockhart PJ, Bryant D (2009) Species delimitation and phylogeny of a New Zealand plant species radiation. BMC Evol Biol 9: 111.
- 49. Joyce DA, Lunt DH, Genner MJ, Turner GF, Bills R, et al. (2011) Repeated colonization and hybridization in Lake Malawi cichlids. Curr Biol 21: R108–9.
- 50. Bussell JD, Waycott M, Chappill JA (2005) Arbitrarily amplified DNA markers as characters for phylogenetic inference. Perspect Plant Ecol Evol Syst 7: 3–26.
- 51. Koopman WJM (2005) Phylogenetic signal in AFLP data sets. Syst Biol 54: 197–217.
- 52. Althoff DM, Gitzendanner MA, Segraves KA (2007) The utility of amplified fragment length polymorphisms in phylogenetics: a comparison of homology within and between genomes. Syst Biol 56: 477–484.
- 53. García-Pereira MJ, Caballero A, Quesada H (2010) Evaluating the relationship between evolutionary divergence and phylogenetic accuracy in AFLP data sets. Mol Biol Evol 27: 988–1000.
- 54. Herrmann D, Poncet N, Manel S, Rioux D, Gielly L, et al. (2010) Selection criteria for scoring amplified fragment length polymorphisms (AFLPs) positively affect the reliability of population genetic parameter estimates. Genome 310: 302–310.
- 55. Crawford LA, Koscinski D, Keyghobadi N (2012) A call for more transparent reporting of error rates: the quality of AFLP data in ecological and evolutionary research. Mol Ecol 21: 5911–5917.
- 56. Gaudeul M, Rouhan G, Gardner MF, Hollingsworth PM (2012) AFLP markers provide insights into the evolutionary relationships and diversification of New Caledonian Araucaria species (Araucariaceae). Am J Bot 99: 68–81.
- 57. Fiz-Palacios O, Vargas P, Vila R, Papadopulos AST, Aldasoro JJ (2010) The uneven phylogeny and biogeography of Erodium (Geraniaceae): radiations in the Mediterranean and recent recurrent intercontinental colonization. Ann Bot: 871–884.
- 58. Papadopulos AST, Price Z, Devaux C, Hipperson H, Smadja CM, et al. (2013) A comparative analysis of the mechanisms underlying speciation on Lord Howe Island. J Evol Biol 26: 733–745.
- 59. Ley AC, Hardy OJ (2013) Improving AFLP analysis of large-scale patterns of genetic variation – a case study with the Central African lianas Haumania spp (Marantaceae) showing interspecific gene flow. Mol Ecol 22: 1984–1997.
- 60. Bensch S, Akesson M (2005) Ten years of AFLP in ecology and evolution: why so few animals? Mol Ecol 14: 2899–2914.
- 61. Meudt HM, Clarke AC (2007) Almost forgotten or latest practice? AFLP applications, analyses and advances. Trends Plant Sci 12: 106–117.
- 62. Bacon CD, Allan GJ, Zimmer EA, Wagner WL (2011) Genome scans reveal high levels of gene flow in Hawaiian Pittosporum. Taxon 60: 733–741.
- 63. Doyle J, Doyle JL (1987) Genomic plant DNA preparation from fresh tissue-CTAB method. Phytochem Bull 19.
- 64. Vos P, Hogers R, Bleeker M, Reijans M, Lee TVD, et al. (1995) AFLP: a new technique for DNA fingerprinting. Nucleic Acids Res 23: 4407–4414.
- 65. Bonin A, Bellemain E, Bronken Eidesen P, Pompanon F, Brochmann C, et al. (2004) How to track and assess genotyping errors in population genetics studies. Mol Ecol 13: 3261–3273.
- 66. Vekemans X, Beauwens T, Lemaire M, Roldán-Ruiz I (2002) Data from amplified fragment length polymorphism (AFLP) markers show indication of size homoplasy and of a relationship between degree of homoplasy and fragment size. Mol Ecol 11: 139–151.
- 67. Lynch M, Clarke AC (1994) Analysis of population genetic structure with RAPD markers. Mol Ecol 3: 91–99.
- 68. Miller MP (1997) Tools for population genetic analyses (TFPGA). Available: http://www.ccg.unam.mx/~vinuesa/tlem09/docs/TFPGADOC.PDF. Accessed 29 September 2012.
- 69. Schönswetter P, Tribsch A (2005) Vicariance and dispersal in the alpine perennial Bupleurum stellatum L. (Apiaceae). Taxon 54: 725–732.
- 70. Ehrich D (2006) AFLPDAT: a collection of r functions for convenient handling of AFLP data. Mol Ecol Notes 6: 603–604.
- 71. Coart E, Glabeke S Van, Petit RJ, Bockstaele E Van, Roldán-Ruiz I (2005) Range wide versus local patterns of genetic diversity in hornbeam (Carpinus betulus L.). Conserv Genet 6: 259–273.
- 72. Petit RJ, Mousadik A El (1998) Identifying populations for conservation on the basis of genetic markers. Conserv Biol 12: 844–855.
- 73. Fox J (2005) The R Commander: A basic statistics graphical user interface to R. Journal Stat Softw. 14: 1–42.
- 74. R Development Core Team (2012) R: A language and environment for statistical computing. Viena: R Foundation for Statistical Computing. Available: http://www.r-project.org/ Accessed 23 April 2012..
- 75. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution 38: 1358–1370.
- 76. Nei M, Li WH (1979) Mathematical model for studying genetic variation in terms of restriction endonucleases. Proc Natl Acad Sci 76: 5269–5273.
- 77. Rohlf FJ (1998) NTSYS-pc, Numerical Taxonomy and Multivariate Analysis, Exeter Software, Version 2.0. Available: http://www.exetersoftware.com/cat/ntsyspc/ntsyspc.html. Accessed 5 February 2012.
- 78. Felsenstein J (1985) Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783–791.
- 79. Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19: 1572–1574.
- 80. Luo R, Hipp AL, Larget B (2007) A bayesian model of AFLP marker evolution and phylogenetic inference. Stat Appl Genet Mol Biol 6: 11.
- 81. Felsenstein J (1981) Evolutionary trees from DNA sequences: a maximum likelihood approach. J Mol Evol 17: 368–376.
- 82. Rambaut A, Drummond AJ (2007) Tracer. Available: http//beast.bio.ed.ac.uk/Tracer. Accessed 12 September 2011.
- 83. Huson DH, Bryant D (2006) Application of phylogenetic networks in evolutionary studies. Mol Biol Evol 23: 254–267.
- 84. Vriesendorp B, Bakker FT (2005) Reconstructing patterns of reticulate evolution in angiosperms: what can we do? Taxon 54: 593–604.
- 85. Bryant D, Moulton V (2004) Neighbor-net: an agglomerative method for the construction of phylogenetic networks. Mol Biol Evol 21: 255–265.
- 86. Jaccard P (1901) Étude comparative de la distribution florale dans une portion des Alpes et du Jura. Bull Société Vaudoise Sci Nat 37: 547–579.
- 87. Oksanen J, Kindt R, Legendre P, O'Hara B, Simpson GL, et al. (2008) The vegan package version 1.15-0. Available: http://cran.r-project.org. Accessed 14 April 2013.
- 88. Hubisz MJ, Falush D, Stephens M, Pritchard JK (2009) Inferring weak population structure with the assistance of sample group information. Mol Ecol Resour 9: 1322–1332.
- 89. Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol Ecol 14: 2611–2620.
- 90. Corander J, Marttinen P (2006) Bayesian identification of admixture events using multilocus molecular markers. Mol Ecol 15: 2833–2843.
- 91. Gower JC (1966) Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika 53: 325–338.
- 92. Excoffier L, Smouse PE, Quattro JM (1992) Analysis of molecular variance inferred from metric distances among DNA haplotypes: Application to human mitochondrial DNA restriction data. Genetics 131: 479–491.
- 93. Excoffier L, Laval G, Schneider S (2005) Arlequin (version 3.0): an integrated software package for population genetics data analysis. Evol Bioinform Online 1: 47–50.
- 94. Nievergelt CM, Libiger O, Schork NJ (2007) Generalized analysis of molecular variance. PLoS Genet 3: e51.
- 95. Pompanon F, Bonin A, Bellemain E, Taberlet P (2005) Genotyping errors: causes, consequences and solutions. Nat Rev Genet 6: 847–859.
- 96. López-Sepúlveda P, Tremetsberger K, Ortiz MÁ, Baeza CM, Peñailillo P, et al. (2013) Radiation of the Hypochaeris apargioides complex (Asteraceae: Cichorieae) of southern South America. Taxon 62: 550–564.
- 97. Juan C, Oromi P, Hewitt GM (1995) Mitocondrial DNA phylogeny and sequential colonization of Canary Islands by darking beetles of the genus Pimelia (Tenebrionidae). Proc R Soc London B 261: 173–180.
- 98. Francisco-Ortega J, Barber JC, Santos-Guerra A, Febles-Hernández R, Jansen RK (2001) Origin and evolution of the endemic genera of Gonosperminae (Asteraceae: Anthemideae) from the Canary Islands: evidence from nucleotide sequences of the internal transcribed spacers of the nuclear ribosomal DNA. Am J Bot 88: 161–169.
- 99. Jorgensen TH (2002) The importance of phylogeny and ecology in microgeographical variation in the morphology of four Canarian species of Aeonium (Crassulaceae). Biol J Linn Soc 76: 521–533.
- 100. Sanmartín I, van der Mark P, Ronquist F (2008) Inferring dispersal: a Bayesian approach to phylogeny-based island biogeography, with special reference to the Canary Islands. J Biogeogr 35: 428–449.
- 101. Trusty JL, Olmstead RG, Santos-Guerra A, Sá-Fontinha S, Francisco-Ortega J (2005) Molecular phylogenetics of the Macaronesian-endemic genus Bystropogon (Lamiaceae): palaeo-islands, ecological shifts and interisland colonizations. Mol Ecol 14: 1177–1189.
- 102. Guillou H, Carracedo JC, Paris R, Pérez Torrado FJ (2004) Implications for the early shield-stage evolution of Tenerife from K/Ar ages and magnetic stratigraphy. Earth Planet Sci Lett 222: 599–614.
- 103. Juan I, Emerson B, Orom I, Hewitt G (2000) Colonization and diversification: towards a phylogeographic synthesis for the Canary Islands. Trends Ecol Evol 15: 104–109.
- 104. Ancochea E, Huertas MJ, Cantagrel JM, Coello J, Fuster JM, et al. (1999) Evolution of the Cañadas edifice and its implications for the origin of the Cañadas Caldera (Tenerife, Canary Islands). J Volcanol Geotherm Res 88: 177–199.
- 105. Nathan R (2006) Long-distance dispersal of plants. Science 313: 786–788.
- 106. Valido A, Nogales M (1994) Frugivory and seed dispersal by the lizard Gallotia galloti (Lacertidae) in a xeric habitat of the Canary Islands. Oikos 70: 403–411.
- 107. Nogales M, Delgado JD, Medina FM (1998) Shrikes, lizards and Lycium intricatum (Solanaceae) fruits: a case of indirect seed dispersal on an oceanic island (Alegranza, Canary Islands). J Ecol 86: 866–871.
- 108. Nogales M, Heleno R, Traveset A, Vargas P (2012) Evidence for overlooked mechanisms of long-distance seed dispersal to and between oceanic islands. New Phytol 194: 313–317.
- 109. Heleno RH, Olesen JM, Nogales M, Vargas P, Traveset A (2012) Seed dispersal networks in the Galápagos and the consequences of alien plant invasions. Proc Biol Sci 280: 20122112.
- 110. Guerrero AM, Tye A (2009) Darwin's Finches as seed predators and dispersers. Wilson J Ornithol 121: 752–764.
- 111. Heleno R, Blake S, Jaramillo P, Traveset A, Vargas P, et al. (2011) Frugivory and seed dispersal in the Galápagos: what is the state of the art? Integr Zool 6: 110–129.
- 112. Montelongo Parada V (1982) Cheirolophus metlesicsii sp. nov. una nueva especie de Asteraceae de Tenerife. Bot Macaronesica 10: 67–78.
- 113. Garnatje T, Pérez-Collazos E, Pellicer J, Catalán P (2013) Balearic insular isolation and large continental spread framed the phylogeography of the western Mediterranean Cheirolophus intybaceus s.l. (Asteraceae). Plant Biol: 1–10.
- 114. Padilla DP, González-Castro A, Nogales M (2012) Significance and extent of secondary seed dispersal by predatory birds on oceanic islands: the case of the Canary archipelago. J Ecol 100: 416–427.
- 115. Crawford DJ, Lowrey TK, Anderson GJ, Bernardello G, Santos-Guerra A, et al. (2009) Genetic diversity in Asteraceae endemic to oceanic islands: Baker's Law and polyploidy. In: Funk VA, Susanna A, Stuessy TF, Bayer RJ, editors. Systematics, Evolution, and Biogeography of Compositae. Vienna: International Association for Plant Taxonomy. pp.139–151.
- 116. Crawford DJ, Anderson GJ, Bernardello G (2011) The reproductive biology of island plants. In: Bramwell D, Caujapé-Castells J, editors. The biology of islands floras. Cambridge: Cambridge University Press. pp.11–36.
- 117. Ellis AG, Weis AE, Gaut BS (2006) Evolutionary radiation of the “stone plants” in the genus Argyroderma (Aizoaceae): unravelling the effects of landscape, habitat, and flowering time. Evolution 60: 39–55.
- 118. Kisel Y, Barraclough TG (2010) Speciation has a spatial scale that depends on levels of gene flow. Am Nat 175: 316–334.
- 119. Sventenius ER (1960) Las centaureas de la sección Cheirolophus en las islas macaronésicas. Anu Estud Atl 6: 219–236.
- 120. Frankham R (1997) Do island populations have less genetic variation than mainland populations? Heredity 78: 311–327.
- 121. Hamrick JL, Godt MJW (1989) Allozyme diversity in plant species. Plant population genetics, breeding and genetic resources. Sunderland: Sinauer Associates Inc. pp. 43–64.
- 122. Hamrick JL, Godt MJW (1996) Effects of life history traits on genetic diversity in plants species. Philos Trans R Soc Biol 351: 1291–1298.
- 123. Hamrick JL, Linhart YB, Mitton JB (1979) Relationships between life history characteristics and electrophoretically detectable genetic variation in plants. Annu Rev Ecol Syst 10: 173–200.
- 124. Lande R (1980) Genetic variation and phenotypic evolution during allopatric speciation. Am Nat 116: 463–479.
- 125. Warwick SI, Gottlieb LD (1985) Genetic Divergence and Geographic Speciation in Layia (Compositae). Evolution 39: 1236–1241.
- 126. Schluter D (2000) The ecology of adaptive radiation. Oxford: Oxford University Press.
- 127. Papadopulos AST, Baker WJ, Crayn D, Butlin RK, Kynast RG, et al. (2011) Speciation with gene flow on Lord Howe Island. Proc Natl Acad Sci U S A 108: 13188–13193.
- 128. Bouck A, Vision T (2007) The molecular ecologist's guide to expressed sequence tags. Mol Ecol 16: 907–924.
- 129. Noor MAF, Feder JL (2006) Speciation genetics: evolving approaches. Nat Rev Genet 7: 851–861.
- 130. Tsitrone A, Kirkpatrick M, Levin DA (2003) A model for chloroplast capture. Evolution 57: 1776–1782.
- 131. Holder MT, Anderson JA, Holloway AK (2001) Difficulties in detecting hybridization. 50: 978–982.
- 132. Bailey CD, Carr TG, Harris SA, Hughes CE (2003) Characterization of angiosperm nrDNA polymorphism, paralogy, and pseudogenes. Mol Phylogenet Evol 29: 435–455.
- 133. Fernández-Palacios JM (2008) The Canaries: an important biogeographical meeting place. J Biogeogr 35: 379–387.
- 134. Paun O, Schönswetter P, Winkler M, Intrabiodiv Consortium (Institution), Tribsch A (2008) Historical divergence vs. contemporary gene flow: evolutionary history of the calcicole Ranunculus alpestris group (Ranunculaceae) in the European Alps and the Carpathians. Mol Ecol 17: 4263–4275.
- 135. Purvis A, Gittleman JL, Cowlishaw G, Mace GM (2000) Predicting extinction risk in declining species. Proc Biol Sci 7: 1947–1952.
- 136. Redding DW, Mooers AØ (2006) Incorporating evolutionary measures into conservation prioritization. Conserv Biol 20: 1670–1678.
- 137. Arponen A (2012) Prioritizing species for conservation planning. Biodivers Conserv 21: 875–893.
- 138. Caujapé-Castells J, Tye A, Crawford DJ, Santos-Guerra A, Sakai A, et al. (2010) Conservation of oceanic island floras: Present and future global challenges. Perspect Plant Ecol Evol Syst 12: 107–129.
- 139. Forest F, Grenyer R, Rouget M, Davies TJ, Cowling RM, et al. (2007) Preserving the evolutionary potential of floras in biodiversity hotspots. Nature 445: 757–760.
- 140. Cadotte MW, Davies TJ (2010) Rarest of the rare: advances in combining evolutionary distinctiveness and scarcity to inform conservation at biogeographical scales. Divers Distrib 16: 376–385.