Inter- and intra-island speciation and their morphological and ecological correlates in Aeonium (Crassulaceae), a species-rich Macaronesian radiation

Abstract

Background and aims: The most species-rich and ecologically diverse plant radiation on the Canary Islands is the Aeonium alliance (Crassulaceae). In island radiations like this, speciation can take place either within islands or following dispersal between islands. Aiming at quantifying intra- and inter-island speciation events in the evolution of Aeonium, and exploring their consequences, we hypothesized that (1) intra-island diversification resulted in stronger ecological divergence of sister lineages, and that (2) taxa on islands with a longer history of habitation by Aeonium show stronger ecological differentiation and produce fewer natural hybrids.

Methods: We studied the biogeographical and ecological setting of diversification processes in Aeonium with a fully sampled and dated phylogeny inferred using a ddRADseq approach. Ancestral areas and biogeographical events were reconstructed in BioGeoBEARS. Eleven morphological characters and three habitat characteristics were taken into account to quantify the morphological and ecological divergence between sister lineages. A co-occurrence matrix of all Aeonium taxa is presented to assess the spatial separation of taxa on each island.

Key results: We found intra- and inter-island diversification events in almost equal numbers. In lineages that diversified within single islands, morphological and ecological divergence was more pronounced than in lineages derived from inter-island diversification, but only the difference in morphological divergence was significant. Those islands with the longest history of habitation by Aeonium had the lowest percentages of co-occurring and hybridizing taxon pairs compared with islands where Aeonium arrived later.

Conclusions: Our findings illustrate the importance of both inter- and intra-island speciation, the latter of which is potentially sympatric speciation. Speciation on the same island entailed significantly higher levels of morphological divergence compared with inter-island speciation, but ecological divergence was not significantly different. Longer periods of shared island habitation resulted in the evolution of a higher degree of spatial separation and stronger reproductive barriers.

Keywords: Aeonium; Canary Islands; ancestral area reconstruction; biogeographical stochastic mapping; diversification; island biogeography; molecular dating.

Fig. 1.

(A) Section Greenovia: A. aizoon, a small (up to 15 cm high) rosette plant endemic to Tenerife, with ±20-merous nectary-lacking flowers. (B) Section Leuconium: A. urbicum subsp. meridionale, a tall (1.5–2 m high) monocarpic plant with unbranched woody stem and hundreds of white-reddish flowers. Both subspecies of A. urbicum are endemic to Tenerife. (C) Section Aeonium: A. arboreum subsp. holochrysum var. holochrysum, a laxly branched, spacious and tall (up to 1.5 m high) shrub with yellow flowers appearing in winter, distributed across El Hierro, La Palma and Tenerife. (D) Section Petrothamnium: A. sedifolium, a small (up to 25 cm high), densely branched shrublet with little thick and clavate leaves, native to La Palma, La Gomera and Tenerife. (E) Section Canariensia: A. cuneatum, a sturdy, herbaceous rosette plant with distinctly glaucous leaves endemic to the laurel forest of northern Tenerife (Teno and Anaga). (F) Section Chrysocome: A. smithii, a shrublet with densely hairy branches and leaves featuring conspicuous reddish tannic stripes.

Fig. 2.

Maximum likelihood phylogeny of Aeonium inferred from concatenated supermatrices (concatML). Numbers above branches indicate bootstrap values (only shown if at least one of them is ≥75) obtained from the complete dataset of all 4280 loci (all/RAxML; values to the left) and from the dataset only comprising the 2218 loci in the 320- to 500-nt length range (320–500/RAxML; values to the right). Dashes (–) signify clades that were supported in 320–500/RAxML but not recovered in all/RAxML. Branches arising directly from the root were shortened (branch length = 0.019) for the purpose of better visibility. Clade names and affiliations to sections are indicated to the right of taxon names. Topology and branch lengths correspond to the results obtained with the 320–500 dataset. p.p., pro parte (partly).

Fig. 3.

Species tree of Aeonium inferred using ASTRAL-III. Numbers above branches indicate bootstrap values (only shown if at least one of them is ≥75) obtained from the species-tree analysis of all 4280 individual RAxML gene trees and from those 2218 gene trees inferred from loci in the 320- to 500-nt length range (left and right values, respectively). The dash (–) signifies a clade that was supported in 320–500/ASTRAL but not recovered in all/ASTRAL. The topology corresponds to the results obtained with the 320–500 dataset. Information about geographical distribution, elevational range across the distribution area and thermo- and ombrotype is given to the right of taxon names. Missing data with respect to elevation, thermo- and ombrotype is indicated with dashed horizontal lines. Further to the right, morphological character states are coded and specified by the inset figure legend below. A. urbicum urb., Aeonium urbicum subsp. urbicum; A. urbicum mer., A. urbicum subsp. meridionale; A. canariense chr., A. canariense subsp. christii; A. canariense can., A. canariense subsp. canariense; A. canariense vir., A. canariense subsp. virgineum; A. canariense lat., A. canariense subsp. latifolium; A. arboreum rub., A. arboreum subsp. holochrysum var. rubrolineatum; A. arboreum arb., A. arboreum subsp. arboreum; A. arboreum hol., A. arboreum subsp. holochrysum var. holochrysum; Monanthes bra., M. brachycaulos; Monanthes mur., M. muralis. A. lindleyi *Aeonium lindleyi s.l., including both subspecies.

Fig. 4.

Chronogram of Aeonium inferred using BEAST, based on the sequence dataset of those 357 loci in the 320- to 500-nt length range that had sequence information for at least 20 different samples (referred to as the min20tax dataset in the text). The topology inferred from the 320–500/RAxML analysis (Fig. 2) was used as starting tree for this BEAST analysis. Only posterior probabilities ≥0.85 are shown above branches. Mean age estimates are given for each supported node (posterior probability ≥0.95) in red font, and the blue error bars indicate the respective 95 % highest posterior density (HPD) intervals. For nodes 6 and 28, age estimates and error bars received paler colours for a better contrast with neighbouring nodes. The scale shows time in million years before present. See also Table 2 for a summary of the stem and crown age estimates of well-supported clades.

Fig. 5.

Co-occurrence matrix of all species and subspecies of Aeonium. Taxa are sorted in clades (see Fig. 2 for clade names) and their names are abbreviated by the first three letters of their species/subspecies/variety epithet. Boxes filled in ochre indicate co-occurrence of the two taxa defining the coordinates of the box, boxes filled in green indicate published natural hybrids between the two respective taxa. Letters in coloured boxes specify where the co-occurrence of two taxa or their hybrids are located; for the geographical codes of these letters see Table 1 or Fig. 3. Numbers in the diagonal boxes below the abbreviated taxon names summarize the number of co-occurrences and number of natural hybrids with other Aeonium taxa (left and right of the slash, respectively) for each taxon. Percentages marked ‘Co’ and ‘Hy’ indicate the proportions of co-occurrences and natural hybrids, respectively, for each pairing of two clades, where the number of possible pairings between all taxa of the two respective clades is considered 100 %. LIN^*, Aeonium lindleyi s.l., including both subspecies.

Fig. 6.

Ancestral area reconstruction of the dated Aeonium phylogeny (Fig. 4) obtained using the BAYAREALIKE+J model in BioGeoBEARS. Pie charts indicate the relative probability of each region or combination of regions for each node and each branch, and rectangles at the tips indicate the recent distribution area or combination of regions for each sampled taxon of the phylogeny. The branches arising directly from the root were shortened for better visibility. Dashed vertical lines represent the ages of Madeira (~5 myr), La Palma ( 1.7 myr) and El Hierro ( 1.1 myr). For the geographical code of letters at the tree tips see Table 1 or Fig. 3.

Fig. 7

Topology of the Canary Islands off the Atlantic coast of North Africa (inset map) and dispersal paths between islands in the evolution of Aeonium that were inferred by the majority of the BSMs. Arrowheads indicate directionality of dispersals, and arrows with arrowheads on both ends indicate ambiguity about directionality. Dotted arrows indicate dispersals with ambiguous region of origin. Numbers above arrows correspond to the number of times a certain dispersal path was taken. Letters in brackets correspond to the nodes and lineages listed below (see Fig. 2 for clade names and Fig. 4 for node numbers) that followed the corresponding dispersal path. For the letter code of island names see Table 1 or Fig. 3. (a) Node 17; Aeonium canariense subsp. christii, A. diplocyclum, A. spathulatum, A. valverdense. (b) Node 42. (c) Node 28. (d) Node 17; Aeonium canariense subsp. christii, A. diplocyclum, A. spathulatum. (e) Nodes 13 and 29; Aeonium sedifolium. (f) Node 32; Aeonium lindleyi subsp. viscatum. (g) Aeonium decorum. (h) Nodes 9 and 26; Aeonium appendiculatum, A. sedifolium. (i) Node 20. (j) Aeonium arboreum subsp. holochrysum var. holochrysum. (k) Node 22. (l) Nodes 14 and 27; Aeonium aureum. (m) Aeonium percarneum, A. spathulatum. (n) Node 36. (o) Aeonium balsamiferum. (p) Node 35; Aeonium korneliuslemsii.

Fig. 8.

Boxplots of REDs in (A) morphological characters and (B) ecological habitat characteristics between well-supported sister lineages (see Materials and methods section for details concerning selection of sister lineages). REDs are separated by the biogeographical correlate (inter- or intra-island) of the diversification event giving rise to the respective sister lineages. Raw data contributing to these boxplots are listed in Table 6. *Difference between means is significant (P < 0.05); n.s., difference between means is not significant (P > 0.05).