Phylogenetic relationships and phylogeography of relevant lineages within the complex Campanulaceae family in Macaronesia

Abstract Macaronesia has long been recognized as a natural model for studying evolutionary processes in plant diversification. Several studies have attempted to focus on single lineages, and few have covered the diversification of a family across all the archipelagos. We used a comprehensive sample to clarify the phylogenetic relationships and the biogeographic history of the Macaronesian Campanulaceae. Hypotheses related to the colonization of these archipelagos will be used to examine the diversification patterns of different lineages. We sequenced the ITS region and six cpDNA markers (atpB, matK, petD, rbcL, trnL‐F, and psbA‐trnH) from 10 Campanulaceae species, including seven endemic species in Macaronesia. The phylogeny of these taxa was reconstructed using maximum parsimony, maximum likelihood, and Bayesian inference. To study the relationships within each lineage, haplotype networks were calculated using NeighborNet and TCS algorithms. Moreover, data were combined with fossil information to construct time‐calibrated trees for the Macaronesian Campanulaceae species. The phylogenetic analyses are largely congruent with current taxon circumscriptions, and all the endemic genera formed monophyletic clades, namely Azorina in Azores; Musschia in Madeira; and Campanula in Cape Verde. The Azorina clade and the Cape Verde endemic Campanula may share a common ancestor in North Africa, and the divergence was dated ca. 12.3 million years ago (Mya). The divergence of the Musschia clade began in the Pliocene ca. 3.4 Mya. Moreover, several examples of intraspecific variation were revealed among the native species with a clear geographic structured patterns, suggesting that cryptic diversity might exist within the native Macaronesian Campanulaceae when compared to the close mainland taxa (e.g., Campanula erinus, Trachelium caeruleum), but additional studies are needed to support the molecular data. This study highlights the power of combining data (e.g., phylogeny and divergence times, with species distribution data) for testing diversification hypotheses within the unique Macaronesian flora, providing useful information for future conservation efforts.


| INTRODUCTION
Islands are biologically simpler than continents and consequently provide perfect geographic and historical settings for the study of species colonization and diversification (Parent, Caccone, & Patren, 2008).
Macaronesia displays a high degree of plant endemicity (Carine, Santos Guerra, Guma, & Reyes-Betancort, 2010;Caujape-Castells et al., 2010;Cosner, Raubeson, & Jansen, 2004), related to its geographic location (and variable isolation), geological origin, and climatic history (Jardim & Menezes de Sequeira, 2008). Although most of the phylogenetic studies have been focused on single lineages with emphasis on Canaries, which displays ca. one-third of the endemic plant diversity of Macaronesia, some studies include two or more Macaronesian archipelagos (Carine, Francisco-Ortega, Santos-Guerra, & Russell, 2004;Mort et al., 2015;Moura, Carine, Malékot, et al., 2015;Moura, Carine, & Sequeira, 2015;Romeiras et al., 2011). Nonetheless, broad-scale studies focused on a plant family covering native and endemics and comparing distinct patterns of phylogenetic structure among the five archipelagos was not yet addressed for this hotspot region. To compare phylogenetic and distribution patterns within all the archipelagos, this study is focused on the Campanulaceae Juss. family, which is characterized by a great number of native species, including some endemic genera in Macaronesia.
Phylogenetic relationships within Campanulaceae remain highly controversial, and a complex biogeographic history has been recently reported in several studies (e.g., Cosner et al., 2004;Crowl et al., 2014;Eddie, Shulkina, Gaskin, Haberle, & Jansen, 2003;Haberle et al., 2009;Roquet et al., 2008). One of the most comprehensive studies on the Campanulaceae family was presented by Crowl et al. (2016), providing a broad phylogenetic and phylogeographic perspective which included chromosomal and morphological data. Previously, Crowl et al. (2014) had produced a first phylogenetic analysis conjointly applying several molecular markers used in former studies within the subfamily Campanuloideae, namely the chloroplast markers atpB, matK, petD, rbcL, and trnL-F and the nuclear region ITS. However, the authors concluded that ITS was inefficient on Campanuloideae due to difficulties on the alignment of sequences, high levels of homoplasy, and concerted evolution, although it provided information at the species level.
Alternatively, two single-copy nuclear loci from the PPR genes family (pentatricopeptide repeat: PPR11 and PPR70) provided independent estimations of relationships, uncovering hybridization events.
To date, only some of the native Macaronesian Campanulaceae species have been included in these phylogenetic studies, and the relationships among some of the genera were not yet clarified. For instance, paraphyly and polyphyly were found between Campanula L. and Wahlenbergia Schrad. ex Roth and conflicts between morphological and molecular data were reported for Azorean endemic Azorina, which was placed inside the Campanula clade in recent studies (Crowl et al., 2014;Olesen, Alarcón, Ehlers, Aldasoro, & Roquet, 2012;Roquet et al., 2008Roquet et al., , 2009).
The biogeographic history in this family, namely in Campanula, is complex and involves a considerable number of migrations (e.g., from the Balkans to western Asia) being particularly critical in its diversification due to orogenic activity that took place in this region during the Late Neogene, and which could have promoted isolation and allopatric speciation within/among lineages, followed by range expansion and posterior isolation to give rise to new endemics, as well as several long-distance (independent) dispersal events to Macaronesia (Roquet et al., 2009). Some phytogeographic studies were recently published by Mairal, Pokorny, Aldasoro, Alarcón, and Sanmartín (2015), , who suggested that for C. canariensis, which is widespread across several Canary Islands, the ocean is apparently less of a barrier than topographic relief within volcanic islands; and the paleo-islands of Tenerife have probably acted as both genetic refuge and sources of new diversity within and between islands. Moreover, Alarcón, Roquet, García-Fernández, Vargas, and Aldasoro (2013)  This study aimed to (1) determine the phylogenetic relationships between native and endemic Campanulaceae species occurring in the Azores, Madeira, Canaries, and Cape Verde, as well as their relationships with continental taxa; (2) determine the potential processes driving diversification in Campanulaceae within the Macaronesian region with divergence time estimations; and (3) contribute with genetic data to assist in future conservation plans.

| Study area
The study area includes all the Macaronesian archipelagos, namely the Azores archipelago with nine islands and some islets; Madeira archipelago comprising Madeira, Porto Santo, and the Desert as subarchipelago composed of three small islands; the Canary archipelago including seven main islands and six islets; and finally Cape Verde, the southernmost islands of Macaronesia, with 10 islands. These archipelagos are characterized by high mountain ranges with a great habitat diversity,

| Sampling
Leaf samples of 46 individuals were collected in natural populations, preserved in silica dried and stored in vacuum-sealed bags. Sampling was conducted in five Azorean islands, Madeira, Porto Santo, Deserta Grande and four Cape Verde islands, and complemented by existing specimens on the AZU (Herbário do Departamento de Ciências Agrárias da Universidade dos Açores) and ORT (Instituto Canario de Investigaciones Agrarias) herbaria (see Tables 1 and 2 for sampling information). Selection of target species followed the most recent available checklists for each archipelago (Ginovés et al., 2005; Menezes de Sequeira et al., 2012;Sánchez-Pinto et al., 2005;Silva et al., 2010).
The suspicious individuals of Musschia angustifolia, as well as samples from the new Campanula species described for Santo Antão (i.e., C. feijoana and C. hortelensis) were also included and listed on Table 1. Some Macaronesian Campanulaceae sequence data available on GenBank (http://www.ncbi.nlm.nih.gov/genbank) were also used on this study (see Table S1).

| DNA extraction and amplification
DNA was extracted from silica dried leaves using a modified CTAB protocol (3 × CTAB) from Doyle and Dickson (1987). A DNeasy Plant Mini Kit (Qiagen, Crawley, UK) was used to extract DNA from herbarium specimens. DNA samples were kept in sterile deionized water at −20°C, after checking the quality and quantity using the spectrophotometer NanoDrop 2000 (Thermo Fisher Scientific).
Six chloroplast (cpDNA) regions were amplified: atpB, matK, petD, rbcL, trnL-F (Crowl et al., 2014) and psbA-trnH (Kress, Wurdack, Zimmer, Weigt, & Janzen, 2005); and one nuclear region (ITS) with the primers of Douzery et al. (1999). All products with the exception of primers, DNA, and pure water were provided by Biotaq PCR Kit (Bioline). The amplification reaction was performed in a T-Gradient thermocycler (Whatman Biometra). Amplification followed the protocol proposed by Carine et al. (2004), the complete ITS region was amplified with the following thermocycling program: initial denaturation for 1 min at 94°C, followed by 30 cycles of 1 min at 94°C, 1 min at 54°C, 3 min at 72°C, and a final extension step at 72°C for 8 min. For cpDNA markers, PCR conditions were set to an initial denaturation for 2 min at 94°C, followed by 30 cycles of 30 s at 94°C, 50°C for 30 s, and 1-min extension at 72°C, with a final extension of 5 min at 72°C.
The DNA fragments resulting from amplification were separated on agarose gel, 0.7%-1% in TBE buffer, stained with SafeView Classic Nucleic Acid Stain and were visualized with Visidoc-IT imaging system (UVP). A molecular marker 50-2,000 base pairs (Sigma-Aldrich) was used as reference. The amplification products were sequenced by STABVIDA, Lda (Portugal).

| Phylogenetic analysis
Sequence data were assembled, edited, and aligned using Geneious ver. 7.0.6 (Biomatters Ltd.) and the Geneious alignment algorithm. The alignments were then inspected and manually optimized.
Moore, and C. tysonii E. Phillips (Crowl et al., 2014). The obtained tree topology for each marker was then compared in order to detect discrepancies.
Maximum-parsimony (MP) analysis were conducted in PAUP* ver. 4.0 beta 10 (Swofford, 2003). The analysis used 1,000 heuristic searches, random stepwise addition, and TBR branch swapping. A strict consensus tree was calculated. jModelTest ver. 2.1.3 (Darriba, Taboada, Doallo, & Posada, 2012) was used to determine the best fitting model of sequence evolution based on the Akaike information criterion. In general, GTR (or its variations) was either the best model estimated or was among the top three selected models. A maximum-likelihood (ML) analysis was conducted using RAxML-HPC2 ver. 7.4.4 (Stamatakis, Hoover, & Rougemont, 2008) and the default settings on the CIPRES Science Gateway (Miller, Pfeiffer, & Schwartz, 2010) with 1,000 bootstraps and a partitioned dataset for the combined matrix.
Bayesian phylogenetic inference (BI) was obtained using The NeighborNet algorithm (Bryant & Mouton, 2004) as implemented in SplitsTree v4.0 (Huson & Bryant, 2006) was used with the default settings to visualize possible incongruences in the dataset. This method reduces the assumption that evolution follows a strictly bifurcating path and allows for the identification of reticulated evolution or incomplete lineage sorting among the dataset.
Levels of variation of target species were determined for each Macaronesian region, and statistical parsimony networks (Templeton, Crandall, & Sing, 1992) were produced with PopART ver. 1.7 using the TCS Network algorithm (Clement, Snell, Walke, Posada, & Crandall, 2002). The chloroplast markers included in the alignments were selected to obtain the highest possible number of haplotypes (Tables S1 and S2).
The new 310 sequences were submitted to GenBank (accessions numbers are listed in the appendix, Table S2).
T A B L E 1 Target native and naturalized Campanulaceae species in Macaronesia and their distribution and conservation status on the archipelagos. Distributions are indicated with islands from which samples are represented in this study shown with filled circles and nonsampled islands shown with open circles

IUCN Status
Canary Islands

| Divergence time estimation
Divergence times within the Campanulaceae family were estimated using the Bayesian MCMC algorithm implemented in BEAST v2.4.6 (Drummond, Suchard, Xie, & Rambaut, 2012). For this analysis, we used the combination of the ITS and three cpDNA markers (matK, rbcL, and petD), and The GTR model of sequence substitution was used for the dataset. To calibrate our phylogenetic tree, we used fossil T A B L E 2 DNA collection codes and localities for the populations used on this study  -Srodoniowa, 1977-Srodoniowa, , 1979) and applied to the node representing the last common ancestor of C. pyramidalis and C. carpatica. Therefore, a prior was applied to the root of the phylogenetic tree of this study and We used date ranges from the 95% highest posterior densities from Bell et al. (2010) to constrain the root of the tree (65-56 Mya). A relaxed lognormal molecular clock was used for all partitions, the Yule process was implemented for the two prior with a uniform model, and a random tree was used as the starting tree. The Bayesian MCMC was run for 10 8 generations, with one tree sampled every 1,000 generations.
To obtain more accurate divergence times for Macaronesian lineages, we used the coalescent species tree method StarBEAST2 in BEAST v2.4.6 (Drummond et al., 2012). These analyses were conducted for (1) Azorina, C. jacobaea and C. bravensis, (2) Musschia, and (3) W. lobelioides subsp. lobelioides. We used two combination of the ITS and four cpDNA markers (matK, rbcL, trnL-F and petD) and The GTR model of sequence substitution was used for all partitions. To calibrate the phylogenetic trees, we used the following secondary calibrations obtained in the first BEAST analysis, described on

| Phylogenetics
Parsimony analysis of the concatenated data matrix generated was rooted with Cyphia spp. for the seven markers used in this study: the ITS and the chloroplast markers matK, rbcL, psbA-trnH, trnL-F, petD, and atpB (parameters of the best parsimonious trees are described on Table S3).
Topological discrepancies between the nuclear and cp data were Nicolau, Brava, and also for Santiago with the ITS+cp dataset (Fig. S1).
Furthermore, the clade composed by the Cape Verde Campanula endemic species plus Campanula mollis, C. edulis, and Feeria angustifolia is sister to the A. vidalii clade and is well supported on the ITS tree and on the combined cp tree but without F. angustifolia (Figures 2 and 3). (2) all Cape Verde accessions (ML = 100%; MP = 88%; PP = 1; Figure 3).
Within the Madeira group, the populations of Porto Santo group distinctively with high support (ML = 93%, PP = 1; Figure 3). Lobelia urens is a fully supported monophyletic group, sister to the other clades in all the analyses.  evidenced by the substantial number of loops found in these points of the phylogenetic network (Figure 4).
Regarding haplotype diversity, for A. vidalii (Figure 5a), one ribotype was found to be endemic to São Jorge (H1). The other four ribotypes were shared between islands: H2 being endemic to the central group, H3 is shared among São Jorge and São Miguel, H4 is endemic to the western group, and H5 is shared by plants sampled from eastern group and some plants sampled from Pico.
In the C. erinus ribotype network (Figure 5b), H1 is shared between São Miguel and Madeira island, H2 is shared among Canary Islands, Madeira, and Santa Maria island from the Azores, and H3 is related to the sample of the mainland, from southern Europe.
Regarding the ITS network obtained for the Cape Verde Campanula

| Divergence time estimation
Date estimates for nodes within the Campanulaceae family in Macaronesia are presented in Figure 8 (see C1 to C9, Table 3). Our analysis indicates that the Azorina must have diverged from Campanula group composed by the Cape Verde endemics and their sister species C. mollis, C. edulis and F. angustifolia (A1; Figure 9, Verde Campanula and their sister species (C2, Figure 8, Table 3) was estimated to have occurred at 9.56 Mya (A2; Figure 9, Table 4). The split between C. bravensis and C. jacobaea (C3; Figure 8,  Figure 9, Table 4) and C. bravensis diverged from it at roughly 0.01 Mya (A5, Figure 9, Table 4).

| Phylogenetic analyses
Recent studies conducted within the Campanulaceae family have made significant progresses toward a robust phylogenetic hypothesis of the group (Crowl et al., 2014;. These authors concluded that previous studies including ITS have shown significant limitation in resolving species level relationships and providing accurate information on the placement of several genera (e.g., Jasione and Musschia).
In our study, these two genera resolved as monophyletic but a polytomy still persisted, however, our combined chloroplast markers resulted in a well-resolved tree regarding the relationships between the other target clades. Hybridization is an evolutionary force common in plants (e.g. Payseur & Rieseberg, 2016) and within the Campanulaceae family is a likely cause of the discrepancy between the nuclear and F I G U R E 2 ITS phylogeny. Best tree from maximum-likelihood analysis. Numbers above branches (≥70%) are maximum-likelihood (black) and maximum-parsimony bootstrap values (red). Number bellow branches (≥0.70) are Bayesian posterior probabilities. Sequences of taxa labeled with "(Gb)" were obtained on GenBank (Table S2) | 99

MENEZES Et al.
chloroplast trees (Wendling et al., 2011). The significant quantity of loops that can be found in the NeighborNet network in this study, further leads to the hypothesis of early hybridization between different species along the evolutionary story of Campanulaceae, and a similar case of was reported by Romeiras, Vieira, et al. (2016) for the Macaronesian Beta ssp. (Amarathaceae family). This may be related to the high controversial taxonomy based in several phylogenetic studies with paraphyly and polyphyly observed in some genera that were reported by recent studies (e.g., Cosner et al., 2004;Crowl et al., 2014Crowl et al., , 2016Eddie et al., 2003;Haberle et al., 2009;Roquet et al., 2008).
Our results support a phylogenetically close relation of Azorina The long distances between the Macaronesian archipelagos may be responsible for high levels of diversity found on these F I G U R E 4 NeighborNet phylogenetic network of the concatenated ITS + cp dataset of Campanulaceae F I G U R E 3 Plastid phylogeny. Best tree from maximum-likelihood analysis of combined plastid dataset: matK, rbcL, psbA-trnH, trnL-F, petD, and atpB. Numbers above branches (≥70%) are maximum-likelihood (black) and maximum-parsimony bootstrap values (red). Number bellow branches (≥0.70) are Bayesian posterior probabilities. Sequences of taxa labeled with "(Gb)" were obtained on GenBank (Table S2) archipelagos, acting as effective barriers to dispersal and promoting allopatric speciation (Schaefer et al., 2011). However, within-island diversification (i.e., in A. vidalii) might be due to complex topographies and long eruptive episodes that happened on these islands (Borges Silva et al., 2016;Brown, Hoskisson, Welton, & Baez, 2006;Dias, Moura, Schaefer, & Silva, 2016;Juan, Emerson, Oromí, & Hewitt, 2000;Silva et al., 2015).
Considering the Cape Verde diversity, two endemic species (i.e., Canary Islands, and Cape Verde. In the past studies (e.g., Crowl et al., 2014;Cupido, 2009;Eddie et al., 2003;Haberle et al., 2009;Olesen et al., 2012 andRoquet et al., 2009), Wahlenbergia hederacea was placed related within the Jasione clade. Eddie and Cupido (2014) proposed a new generic name for W. hederacea as Hesperocodon hederaceus (L.) Eddie & Cupido, based on the last molecular studies and its main morphological characteristics, which are fundamentally campanuloid. Eddie and Cupido (2014) concluded that the capsule dehiscence mechanism of W. hederacea is essentially wahlenbergioid and differs from most campanuloids, considered together with Feeria and Jasione as "transitional" genera due to these characteristics intermediate between typical wahlenbergioids and typical campanuloids. However, our study, W. hederacea is within Wahlenbergia clade, being sister of W. lobelioides subsp. lobelioides. The Macaronesian endemic species may be the missing link that could provide the right relationship between W. hederacea and its respective clade. There can also be seen a quantity of loops on Figure 6 that report an ancestral hybridization between W. hederacea and W. lobelioides subsp.

lobelioides.
In what concerns the Madeiran T. caeruleum haplotype, the molecular patterns detected might be linked to a man-mediated founder effect (Frankham, 1997), considering records of the introduction of this species in the Madeira island ca. 1840 (Lowe, 1868). Bayesian analysis implemented in Beast, based on the concatenated dataset of ITS and cpDNA markers (matK, rbcL and petD), illustrating the estimated divergence ages at selected calibrated nodes. The geographic origin of each specimen is provided (right side) with a color code for continental areas and for the Macaronesian archipelagos. C1-C9 as described in Table 3. Mya, million years ago F I G U R E 9 Maximum-clade-credibility (MCC) time-calibrated tree of Azorina and Cape Verde Campanula in Macaronesia inferred and dated using a multispecies coalescent method (StarBEAST2), based on the two partitions dataset: ITS and cpDNA markers (matK, rbcL, trnL-F and petD), illustrating the estimated divergence ages at selected calibrated nodes. A1-A5 as described in Table 4. Mya, million years ago introduced and to the genetic diversity present in the population of origin (Frankham, 1997)

| Diversification of Campanulaceae in Macaronesia
The Macaronesian flora shows an important connection with Southwest Europe and the Northwest African flora, being the last one the main source of colonization events in many Macaronesian lineages (Sanmartín, Anderson, Alarcón, Ronquist, & Aldasoro, 2010). for Macaronesia inferred and dated using a multispecies coalescent method (StarBEAST2), based on the two partitions dataset: ITS and cpDNA markers (matK, rbcL, trnL-F, and petD), illustrating the estimated divergence ages at selected calibrated nodes. W1-W4 as described in Table 4. Mya, million years ago described for Pericallis D. Don in the Azores, Madeira, and Canaries which the origin of woodiness is correlated with ecological variation from open to species-rich habitats and the ancestor of Pericallis was probably an herbaceous plant adapted to marginal habitats of the laurel forest (Panero, Francisco-Ortega, Jansen, & Santos-Guerra, 1999).
Another pattern is reported for Echium L. which the islands colonization is related to the origin of perennial woodiness from herbaceous habit and was furthermore accompanied by intense speciation (Böhle, Hilger, & Martin, 1996).
In what concern to the Macaronesian endemic W. lobelioides

| Conservation approaches
The plant conservation strategy in Macaronesia has a clear focus on threatened endemic taxa, but only some of the study species were assessed and included in the IUCN Red List of Threatened Species (www.iucnredlist.org). As suggested by Romeiras, Catarino, Filipe, et al. (2016), new prioritization methods should consider the spatial and intra-archipelago genetic diversity of insular taxa. Due to their uncertain status, most of this study target taxa are not currently protected, with some exceptions. Musschia isambertoi is the rarest species of Musschia and was proposed the IUCN category "Critically Endangered" (CR,C2a(i,ii);D) by Sequeira et al. (2007), although included in a protected area (Natura 2000 EU PTDES0001) the recent uncontrolled increase of population numbers of the common goat drove this taxon to the edge of extinction. Moreover, Romeiras, Catarino, Gomes, et al. (2016) proposed the IUCN category of "Endangered" (B1ab(ii,iv)+2ab(ii,iv)) for C. bravensis and "Vulnerable" (B1ab(ii)+2ab(ii)) for C. jacobaea, and confirmed that the Cape Verde vascular plants have become more threatened and their conservation status has declined in the last years, mostly as a consequence of the increase in exotic species, habitat degradation, and human disturbance. Conservations plans are needed in order to preserve the Cape Verdean Campanulaceae (see for details Romeiras, Catarino, Gomes, et al., 2016).
The genetic diversity found within the studied taxa among the archipelagos and the several single-island haplotypes observed must be considered and protected. Although there is a clear lack of taxonomic revisions including insular taxa usually considered as native, further studies in population genetics structure and reproductive biology should also be conducted, as proposed by Silva et al. (2015) in a holistic approach to conservation of rare island plants. Furthermore, to define better conservation strategies of the Macaronesian endemic flora, prioritizing threatened species and conserving the entire extent of their natural ranges was recently recognized as a crucial step (Romeiras, Catarino, Gomes, et al., 2016;Romeiras, Vieira, et al., 2016).
Concerning the colonization status of the target Campanulaceae, as described in the regional checklists (Ginovés et al., 2005;Menezes de Sequeira et al., 2012;Sánchez-Pinto et al., 2005;Silva et al., 2010), the phylogenetic data generated in this study provides an additional knowledge to understand the doubtful status about the colonization of these taxa: (1) C. erinus status of "naturalized" for the Azores (Silva et al., 2010) and "doubtfully native" for the Canary Islands (Ginovés et al., 2005) should be changed to "native"; (2) T. caeruleum, listed as "doubtfully naturalized" for the Azores (Silva et al., 2010) needs an indepth study to assert a native status; and (3) W. lobelioides subsp. lobelioides, listed "doubtfully native" for Cape Verde, should be considered native.
This study shows that the spatial patterns of species differ among the studied Campanulaceae lineages as well as their evolutionary history within the Macaronesian archipelagos. Hence, future conservation measures should consider the existing inter and intra-archipelago genetic variation in the Campanulaceae family in Macaronesia, and accordingly, priority taxa should not be restricted to endemic lineages, but also include native threatened species.