Insular woodiness (IW), referring to the evolutionary transition from herbaceousness towards woodiness on islands, has arisen more than 30 times on the Canary Islands (Atlantic Ocean). One of the IW hypotheses suggests that drought has been a major driver of wood formation, but we do not know in which palaeoclimatic conditions the insular woody lineages originated. Therefore, we provided an updated review on the presence of IW on the Canaries, reconstructed the palaeoclimate, and estimated the timing of origin of woodiness of 24 insular woody lineages that represent a large majority of the insular woody species diversity on the Canaries. Our single, broad-scale dating analysis shows that woodiness in 60-65% of the insular woody lineages studied originated within the last 3.2 Myr during which Mediterranean seasonality (yearly summer droughts) became established on the Canaries. Consequently, our results are consistent with palaeoclimatic aridification as a potential driver of woodiness in a considerable proportion of the insular woody Canary Island lineages. However, the observed pattern between insular woodiness and palaeodrought during the last couple of million years could potentially have emerged as a result of the typically young age of the native insular flora that is characterised by a high turnover.

We used the CIPRES scientific portal for maximum likelihood (ML) phylogeny inference using RAxML - HPC version 8 [1,2]. A rapid bootstrapping algorithm was implemented, examining 1000 pseudo-replicates. GTRCAT approximation was selected in accordance with RAxML recommendations for large trees following Izquierdo-Carrasco et al. [3]; the other default parameters followed Stamatakis [4]. The phylogenetic relationships of all species in the broad-scale aligned dataset were constrained at family level when conducting the RAxML analysis to (1) assure that the discriminatory power of matK and rbcL at the genus and species level does not come at the cost of correct deeper phylogenetic reconstruction, and (2) to reduce computational time. 

We took 42 age constraints across the angiosperm phylogeny were taken from Janssens et al. [5] (see Supplementary material, Table S5). The root node of angiosperms was calibrated to 138.5 Mya (million years ago) based on estimates by [6]. To assess the phylogenetic uncertainty in our dating estimates, we generated 1,000 bootstrap pseudo-replicates using the best supported ML topology as constraint. Each pseudo-replicate was individually dated using TreePL version 1.0, which is specifically designed to date large phylogenies [7]. A small smoothing parameter of 0.003 was applied to minimise the influence of the large heterogeneity of substitution rates across the broad angiosperm phylogeny, while the default setting was kept for all other parameters (for more details to justify the smoothing parameter, see Janssens et al. [5]). Changing the smoothing parameter to 1 or even 10 did not change the dating estimates for most of the clades studied (results not shown). The 1000 dated trees were summarised into a single consensus tree to calculate the 95% range for the dating estimates of each node using TreeAnnotator version 1.8.4 with default settings [8]. 

1.      Miller MA, Pfeiffer W, Schwartz T. 2010 Creating the CIPRES Science Gateway for inference of large phylogenetic trees. In 2010 Gateway Computing Environments Workshop (GCE), pp. 1–8. (doi:10.1109/GCE.2010.5676129)

2.      Stamatakis A. 2014 RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30, 1312–1313. (doi:10.1093/bioinformatics/btu033)

3.      Izquierdo-Carrasco F, Smith SA, Stamatakis A. 2011 Algorithms, data structures, and numerics for likelihood-based phylogenetic inference of huge trees. BMC Bioinformatics 12, 470. (doi:10.1186/1471-2105-12-470)

4.      Stamatakis A. 2016 The RAxML v8. 2. X Manual. 

5.      Janssens S et al. 2020 A large-scale species level dated angiosperm phylogeny for evolutionary and ecological analyses. Biodivers. Data J. 8, e39677. (doi:10.3897/BDJ.8.e39677)

6.      Magallón S, Gómez‐Acevedo S, Sánchez‐Reyes LL, Hernández‐Hernández T. 2015 A metacalibrated time‐tree documents the early rise of flowering plant phylogenetic diversity. New Phytol. 207, 437–453. (doi:10.1111/nph.13264)

7.      Smith SA, O’Meara BC. 2012 treePL: divergence time estimation using penalized likelihood for large phylogenies. Bioinformatics 28, 2689–2690. (doi:10.1093/bioinformatics/bts492)

8.      Rambaut A, Drummond AJ. 2015 TreeAnnotator v1.8.4 MCMC Output analysis.