Patterns of genetic diversity in three plant lineages endemic to the Cape Verde Islands

In an effort to better understand the evolution of the vascular plant flora of the Cape Verde Islands (Macaronesian Region), this study provides an updated checklist for the endemic vascular plants of the Cape Verde Islands and compares patterns of genetic diversity within three endemic plant lineages. The detected levels of genetic differentiation between islands indicate the existence of overlooked (cryptic) taxa in all three lineages, in the genus Umbilicus possibly at species level. These findings indicate that plant diversity in Cape Verde is higher than previously thought and highlights the need for additional studies.


Introduction
Efforts to conserve island floras and to understand their diversity are crucially dependent on baseline taxonomic knowledge: species are the unit of conservation actions, the focus of phylogenetic and phylogeographic work and the basic units for macro-ecological analyses. However, whilst island floras have been subject to study over a long period (e.g. Romeiras et al. 2014), and are often considered well explored (e.g. Joppa et al. 2011), recent discoveries of island taxa new to science have occurred even in groups of large organisms such as lizards and birds (Whittaker and Ferná ndez-Palacios 2007). Gray and Cavers (2013) showed that taxonomic effort expended contributes to the patterns investigated in theoretical biogeography and a recent survey of biologists working on oceanic islands suggested that 'Knowledge of the taxonomy, distribution and threat status of plants on oceanic islands is insufficient' (Caujapé-Castells et al. 2010). Clearly, the taxonomy of oceanic island plants is far from complete, and this is an important issue to address.
This paper considers recent developments in our understanding of the flora of the Cape Verde Islands and the potential of molecular data to provide further insights. The Cape Verdes are the most southerly archipelago of the Macaronesian Region that also comprises the archipelagos of the Azores, Canaries and Madeira in the North Atlantic. The Cape Verdes are a group of 10 volcanic islands located 1500 km southwest of the Canary Islands and 570 km west of the African mainland. Lying at tropical latitudes and in close proximity to the coast of Senegal, their flora is of mainly tropical African origin and, accordingly, it was proposed to group the Cape Verde flora with the palaeotropical floras of the Saharan Tropical region (Rivas-Martínez 2009). Within the archipelago, the islands form three clusters: (i) northern group (Santo Antão, Sã o Vicente, Santa Luzia and São Nicolau); (ii) southern group (Santiago, Fogo and Brava) and (iii) eastern group (Sal, Boavista and Maio) (see Fig. 1). The northern and the southern islands are characterized by high mountains [e.g. Monte Gordo (1304 m) in São Nicolau; Pico da Antónia (1392 m) in Santiago; Tope de Coroa (1979 m) in Santo Antão and Pico do Fogo (2829 m) in Fogo], offering a wide range of habitats over relatively short distances (Duarte and Romeiras 2009). The eastern islands are lower, drier and more homogeneous in their ecology. The islands' ages range from 25.6 to 21.1 Ma (for Sal and Maio, respectively) to ,6 Ma (for Brava), with ages decreasing from east to west (Doucelance et al. 2003). Only Fogo Island currently has volcanic activity, with the most recent eruptions occurring in 1995 and 2014.
Extensive taxonomic and collecting activity during the last decade of the 20th century resulted in the publication of a monograph on the Cape Verde endemic flora by Brochmann et al. (1997). A total of 82 endemic taxa were recognized and 5 distributional elements were identified: (i) northern; (ii) southern, which together form the (iii) western element; (iv) eastern and (v) ubiquitous, which comprises taxa distributed on at least one island of western and eastern elements (see Fig. 1). The vast majority of taxa that occur in the northern, southern and western distributional elements are montane; all eastern taxa are coastal; most of the geographically ubiquitous taxa are also ecologically ubiquitous, and more than half are also altitudinally ubiquitous.
To date, molecular analyses of the Cape Verde flora have been limited. Analyses of the relationships of Macaronesian lineages have focussed largely on the floras of the Azores, Canaries and Madeira [e.g. for reviews see Carine and Schaefer (2010), Caujapé -Castells (2011), Pé rez-de-Paz andCaujapé -Castells (2013)], and where Cape Verdean taxa have been included, the sampling has often been limited (e.g. Tornabenea, Spalik and Downie 2007;Lotus, Ojeda et al. 2014). However, molecular studies of Cape Verdean Campanula L. (Alarcó n et al. 2013) and Echium L. (Romeiras et al. 2007(Romeiras et al. , 2011 have sampled more extensively and revealed the geographic structure of genetic variation within those lineages. For Echium, a single colonization from the Canaries 5 Ma was inferred, with subsequent diversification within the Cape Verdes during the Pleistocene (,1.8 Ma) and a split between the 'southern' (E. hypertropicum, E. vulcanorum) and 'northern' (E. stenosiphon s.l.) island species (García-Maroto et al. 2009;Romeiras et al. 2011). Within Campanula, Alarcó n et al. (2013) inferred a recent divergence (1.0 Ma) of the Cape Verde endemic C. jacobaea from its sister species C. balfourii (endemic to Socotra). Within the archipelago, three groups were identified: two restricted to the northern islands (one each endemic to Sã o Nicolau and Santo Antã o) and a third restricted to the southern islands (Fogo, Santiago and Brava). Thus, both Romeiras et al. (2011) andAlarcó n et al. (2013) found molecular patterns within lineages that are consistent with the distributional elements previously recognized by Brochmann et al. (1997) based on analyses of distributional data. This paper has two goals. First, we provide a revised list of Cape Verde endemic taxa, discuss the new species discoveries and re-evaluate the distributional elements defined by Brochmann et al. (1997) in light of endemic taxa and distribution data published since 1997. Second, we investigate the patterns of genetic variation within three plant lineages endemic to Cape Verde to explore the geographical structuring of genetic variation in the highly fragmented insular landscape of this archipelago. We chose Globularia L., Cynanchum L. and Umbilicus DC. For each of these lineages, only one species is currently recognized in the Cape Verdes: Globularia amygdalifolia Webb (Plantaginaceae), Cynanchum daltonii (Decne. ex Webb) Liede & Meve (Apocynaceae) and Umbilicus schmidtii Bolle (Crassulaceae). Globularia amygdalifolia and U. schmidtii constitute western elements sensu Brochmann et al. (1997) with distributions spanning five and four islands, respectively. Cynanchum daltonii, considered a ubiquitous element, occurs on seven islands in a variety of habitats from sea level to high mountain areas. The three taxa also differ in their phytogeographic relationships: G. amygdalifolia has Canaro-Madeiran affinities (a Macaronesian element), C. daltonii Sudano -Zambesian -Sindian affinities (an African element) and U. schmidtii is a Mediterranean element (Brochmann et al. 1997). Sampling multiple accessions from across the distribution range of each species for both nuclear (ITS) and plastid DNA regions (matK, psbA-trnH, rbcL, trnL-F), we aim to determine whether a molecular variation in the three focal lineages correlates with geographic distance. , southern (i.e. Santiago, Fogo and Brava), western (including species simultaneously present in northern and southern islands), eastern (i.e. Maio, Sal and Boavista) and ubiquitous (including species present in both western and eastern islands). The distribution of multi-island endemics-MIEs (grey) and single-island endemics-SIEs (black) in each island (right); and within the five distributional elements (bar graph in the centre; the number of taxa is placed above each bar).

Characterization of the endemic flora
We compiled a list of the Cape Verde endemic vascular plants, including the recently described species, together with species distribution data, arranged according to the five distributional elements established by Brochmann et al. (1997): northern, southern, western (including species simultaneously present in northern and southern islands), eastern and ubiquitous (including species present in both western and eastern islands) (Fig. 1). Each species was also characterized as either hygrophytic, mesophytic or xerophytic following Brochmann et al. (1997). We determined the occurrence of single-island endemics (SIEs) and multi-island endemics (MIEs) for each distributional element and ecological class.

Eco-geographic data of studied plant lineages
Globularia amygdalifolia and Umbilicus schmidtii occur in small populations of few individuals in northeastexposed areas over 400 m above sea level (a.s.l.). Umbilicus schmidtii populations are found in humid zones mainly associated with montane rupicolous vegetation and Globularia amygdalifolia is mainly found in montane scrub vegetation, remarking its presence in volcanic lapilli areas of Fogo Island. Cynachum daltonii mainly occurs in semi-arid rocky escarpments up to 900 m a.s.l., reaching more than 2000 m a.s.l. on Fogo. It is a characteristic component of the northeast-exposed coastal cliffs, where it forms large stands.

Molecular analyses
Sampling. Plant material for DNA extraction was collected and preserved in silica gel. Vouchers were deposited at LISC (Tropical Research Institute) and BM (Natural History Museum, London) herbaria [see Supporting Information].
Globularia amygdalifolia populations were sampled from four of the five islands on which it occurs: Sã o Nicolau (one population, five individuals), Santo Antã o (one population, two individuals), Fogo (two populations, five individuals) and Brava (two populations, five individuals) [see Supporting Information]. We did not sample plants from Santiago, since all individuals seen were cultivated.
For Cynachum daltonii, populations were sampled on all seven islands from which it has been recorded but DNA extraction from the samples from Santiago and For Umbilicus schmidtii, individuals were sampled from all four islands on which it occurs: Sã o Nicolau (one population, five individuals), Santo Antã o (three populations, five individuals), Santiago (one population, five individuals) and Fogo (two populations, five individuals).
Molecular methods. DNA extraction from silica gel-dried leaf material followed a DNeasy Plant Mini Kit protocol (Qiagen, Crawley, UK), with a further purification using QIAquick PCR Purification Kit (Qiagen, Crawley, UK).
One individual per island per species was selected for a first molecular screening with five DNA regions [i.e. ITS and four cpDNA (matK, psbA-trnH, rbcL, trnL-F)] in order to ascertain which regions are variable within each species. Thereafter, only variable regions were amplified and sequenced for the remaining samples.
From the nuclear genome, the ITS1 -5.8S -ITS2 region was amplified using ITS5 and ITS4 primers from White et al. (1990). From the chloroplast genome, four regions were sequenced: parts of the maturase K (matK) gene using the primers matKF_uni: 5 ′ -AAT TTA CGA TCH ATT CAT TCM ATW TTT CC-3 ′ and matKR_uni: 5 ′ -AGT TYT ARC ACA AGA AAG TCG AAR TAT ATA-3 ′ following Schaefer et al. (2011); the ribulose-1,5-bisphosphate carboxylase/ oxygenase (rbcL) gene using the primers 1F and 724R of Olmstead et al. (1992); the trnL-F spacer using the primers 'e' and 'f' of Taberlet et al. (1991) and the psbA-trnH spacer using the primers of Sang et al. (1997). DNA amplification was performed in a 2720 Thermal Cycler (Applied Biosystems) in 25 mL-volume reactions. Standard polymerase chain reaction (PCR) procedures were applied to carry out amplifications. We used 1.25 units of DreamTaq TM DNA polymerase, and BSA (0.4 mg mL 21 ) for all reactions. The PCR conditions were as follows: (i) 10 min pre-treatment at 94 8C, 28 cycles of 1 min at 95 8C, 1 min at 55 8C, 3 min at 72 8C and a final stage of 7 min at 72 8C, for ITS; (ii) 10 min pre-treatment at 94 8C, 30 cycles of 1 min at 96 8C, 3 min at 50 8C, 3 min at 72 8C and a final stage of 7 min at 72 8C, for the cpDNA region. Amplified products were purified with Sureclean Plus (Bioline, London, UK) and sent to STAB Vida-Investigação e Serviços em Ciências Bioló gicas, Lda (Monte da Caparica, Portugal) for Sanger sequencing. Sequences were deposited in GenBank under the accession numbers KP279325 -KP279464 [see Supporting Information].

Characterization of the endemic flora
Ninety-two taxa (including several subspecies) are currently considered endemic to the Cape Verde Islands (Table 1). The majority of endemic taxa ( 75 %) are distributed in the western islands, either in the (i) northern (34 %), (ii) southern (16 %) elements or in the (iii) western element (25 %); lower proportions are distributed in the eastern islands (4 %) or are ubiquitous elements (21 %) ( Fig. 1 and Table 1). The northern islands are also the richest in SIEs (Fig. 1). Twenty-four of the 31 taxa that constitute the northern element are SIEs (77 %). The remaining seven taxa of the northern element occur on two to three islands with none found on Santa Luzia (Table 1). Ten SIEs occur in the southern element (six SIEs in Fogo, three in Santiago and one in Brava), constituting 66 % of the southern element. Of the four eastern element taxa, only one is a SIE (25 %).

Molecular analyses
Globularia amygdalifolia-genetic variation was detected only in the ITS region. Two substitutions in the 659 bp ITS fragment defined three ribotypes, one of which was found only in plants from Fogo, the second restricted to Brava and the third shared between plants from the northern islands of Santo Antã o and Sã o Nicolau (Fig. 2A). The network supported a split between plants from the northern and southern islands.
Umbilicus schmidtii-genetic variation was observed in the matK (844 bp), psbA-trnH (289 bp) and trnL-F spacer (286 bp) regions (the ITS region was not successfully sequenced for most samples). A total of 11 substitutions, 9 indels and 1 inversion (8 bp) defined five haplotypes for the combined 1420 bp of plastid DNA. Three haplotypes were private to the northern island populations; of which two were found in plants from Santo Antã o (one comprising plants from Cova and the other comprising plants from Pedra Rachada/Delgadinho da Corda) and one was found in Sã o Nicolau. Three to four mutations separated the three northern island haplotypes. The two remaining haplotypes were found in plants from the southern islands. Both haplotypes occurred in the population on Santiago with one also found in Fogo. The two southern haplotypes are distinguished by 5 mutations with 13 mutations separating southern and northern haplotype groups (Fig. 2B).
Cynanchum daltonii-genetic variation was detected only in the ITS region: three substitutions were found in the 648 bp fragment and these defined two ribotypes (Fig. 2C). One was found in plants from Santo Antã o, Sã o Nicolau and Brava, and the second in plants from Fogo and Boavista.

Spatial patterns of endemism
The endemic vascular plant list for Cape Verde has increased from 82 (Brochmann et al. 1997) to 92 taxa, which is an 12 % increase in 18 years. Of the endemics more recently discovered or reclassified as endemics, one is a fern (Dryopteris gorgonea) and the remainder are angiosperms, of which six are recognized at species rank (Fagonia mayana, Helichrysum nicolai, Lotus alianus, Tornabenea ribeirensis, Solanum rigidum and Withania chevalieri) and three at subspecies rank (Echium stenosiphon subsp. glabrescens, Teline stenopetala subsp. santoantonai and Dracaena draco subsp. caboverdeana). Three of the 10 (Echium stenosiphon subsp. glabrescens, Lotus alianus and Tornabenea ribeirensis) have other conspecific Cape Verdean endemic taxa (Table 1) and their discovery reflects an enhanced understanding of lineages that have diversified within the Cape Verdes. The remaining taxa lack conspecific endemic taxa and may be considered anagenetic lineages sensu Stuessy et al. (2006). Their description has typically occurred within the context of wider revisionary or monographic studies. Solanum rigidum, for example, was previously considered introduced to the Cape Verdes from the Americas (Gonçalves 2002 sub S. fuscatum L.). However, monographic research on the genus has shown it to be a Cape Verde endemic that has rather been introduced into the Americas (Knapp and Vorontsova 2013). The discovery of the anagenetic element of the Cape Verde endemic flora is likely far from complete. Whilst many such taxa are widespread taxa of little conservation concern, this is not the case for all. For example, Helichrysum nicolai, described by Kilian et al. (2010), is restricted to the Alto das Cabaças range in the northeastern part of Sã o Nicolau. Its conservation status has not been assessed, but population sizes (very few tens) and the extent and area of occurrence

Genetic diversification within the plant lineages
The molecular analyses we present in this paper provide further insights into the patterns of endemic plant diversity in the Cape Verdes. Globularia amygdalifolia and Umbilicus schmidtii both constitute western elements of the flora (Table 1) and in both cases, differentiation is evident at the molecular level between plants from southern and northern islands (see Fig. 2A and B). Isolation and drift resulting from the small and fragmented distributions of these taxa coupled with the significant distances separating the northern and southern sub-archipelagos (230 km separates Santo Antã o from Brava; and 140 km Sã o Nicolau from Santiago, maximum and minimum distances, respectively) is likely to explain this pattern. A similar intraspecific pattern was documented in Campanula jacobaea (Alarcó n et al. 2013), which is also a western element. From a conservation perspective, it would be appropriate to treat northern and southern populations of western element taxa as distinct management units.
In the case of U. schmidtii, the differentiation of northern and southern populations appears to be consistent with differences observed in the inflorescence (e.g. morphological characters of the flowers, data not shown) suggesting the need of further research, in order to ascertain whether overlooked taxa occur in the archipelago.
Umbilicus is a small succulent herb, which is perhaps most likely to harbour cryptic diversity because it is an undercollected species, and most of the herbarium specimens lack important characters for identification. Its tiny seeds are probably wind-dispersed, but gene flow might be reduced due to the significant distance between the islands.
Some variation in leaf size and shape was also evident within G. amygdalifolia, but no clear geographical pattern in the morphological variation could be identified. Nevertheless, a morphological re-examination in this species within the context of the molecular results presented here would be appropriate.
Intraspecific variation was also found in C. daltonii with two distinct haplotypes. However, the distribution of the two haplotypes is incongruent with Brochmann et al.'s (1997) elements since one is distributed in the eastern island of Boavista and the southern island of Fogo whereas the other is distributed in the northern islands of Santo