Phylogenetic relationships and divergence patterns
We investigated the phylogenetic relationships of various genetic groups, considering the presence and absence of the hybrid ecotype ofC. subpubescens (SH; Figs. 2 and 3). When a hybrid ecotype arises between two distantly related parent ecotypes, the parent ecotype is expected to be located closer to the hybrid ecotype compared with its original position on the phylogenetic tree. In the Central American genus Aphelandra , the inclusion of hybrids between distantly related species in phylogenetic analyses caused a breakdown in cladistic structure and significant topological changes (McDade, 1992). Thus, we concluded that the phylogenetic tree lacking hybrid ecotype SH (Fig. 3b) represents the true phylogenetic relationships, and ecotype SG is considered to have originated from ecotype S in the Chichijima Islands, subsequently colonizing the Hahajima Islands as ecotype SG.
Based on the phylogenetic tree lacking the hybrid ecotype (Fig. 3b) and the population demography model (Model e1 in Fig. S2e), we determined that the ancestral ecotype ST in the Hahajima Islands is the most ancient. Subsequently, ecotype SD in the Hahajima Islands underwent divergence. This was followed by species P and G in the Chichijima Islands, as well as ecotype S. Finally, ecotype SG in the Hahajima Islands emerged from ecotype S in the Chichijima Islands. Population demography analysis revealed that ecotype ST diverged from the ancestral species approximately 170 kya, whereas the other species and ecotypes underwent nearly simultaneous divergence around 73–77 kya (Table S6). Given that 73–77 kya, corresponds to the time when Marine Isotope Stage (MIS) 5 (interglacial period) changed to MIS 4 (glacial period) (Martinson et al., 1987), estimated to have been a period of rapid cooling, cold weather may have triggered the simultaneous divergence of this taxon. Kadereit and Abbott (2021) reviewed studies examining divergence times from phylogenetic trees from all continents and major climatic zones, finding that many plant speciation events occurred in the Quaternary and suggesting that climate change during this period was the cause.
However, it should be noted that estimated divergence times are highly dependent on the generation time used, as these times are calculated by multiplying generation time by approximate generation time. Cultivation experiments involving ecotypes SD and SG confirmed that flowering occurred within 1–2 years of sowing (Setsuko S., personal observation). We consider a 5-year generation time to be suitable for estimating divergence time; however, to validate our findings, it is essential that comparative analyses are conducted using alternative methods, such as BEAST (Drummond A, 2007), which employs fossil data for a more precise age calibration.
In this study, species and ecotypes adapted to a dry environment (species P and ecotype SD) and forest understory and forest edge environments (species G and ecotypes S and SG) diverged at the same time from ecotype ST that constitutes the canopy of tall mesic forests. Although speciation timing may lack precision due to methodological challenges regarding glacial or interglacial periods, it is evident that simultaneous diversification occurred concurrently. Considering the characteristics of the divergent species/ecotypes, the timing of speciation likely aligns with the onset of aridification on the islands. Initially, species/ecotypes adapted to the forest understory and forest edge environment may seem unrelated to aridification. However, species G inhabits forests that are not tall (dry scrub) (Toyoda, 2014), ecotype S inhabits the edges of mesic forests, and ecotype SG does not flower in overly dark forests (Setsuko S., personal observation). Therefore, they are all considered maladaptive to taller forests where ecotype ST grows. This suggests that the environment has changed from a forest with high tree height to a forest with reduced tree height and increased forest edges (Olson et al., 2018), potentially caused by the aridification of the island. Examples of organisms rapidly altering their phenotypes upon aridification have been reported in animals, such as Darwin’s finches (Grant & Grant, 2006), and in plants, such as Mimulus andBrassica (Dickman et al., 2019; Johnson et al., 2022). Selection has also been observed on the HMGA2 gene, causing beak size variation during drought in Darwin’s finches (Lamichhaney et al., 2016), and multiple genes associated with drought response traits evolving during drought in Brassica (Franks et al., 2016; Johnson et al., 2023). In the Callicarpa genus in the Bonin Islands, rapid adaptation to aridification may have led to speciation. Therefore, future research will involve identifying genes associated with drought adaptation.
Surprisingly, we revealed that the differentiation timing of the five species/ecotypes, occurring 73–77 kya, and the migration of ecotypes Sm and STm in the Mukojima Islands from ecotypes S in the Chichijima Islands and ST in the Hahajima Islands, respectively, took place during approximately the same period, around 81–82 kya (Fig. 4; Table S6 and S10). Aridification usually hinders the successful fruit reproduction of previously abundant plants (Abobatta, 2021), leading to food shortages across the entire island; therefore, it is plausible that avian seed dispersers may have moved to new islands in search of food (Boyle & Conway, 2007).
Population size changes, exhibiting large increases of more than two orders, were observed for species/ecotypes P, G, and SG in model e1 (Fig. 4a; Tables S6 and S8). Dry scrub, the habitat of species P and G, may have increased due to aridification during the glacial period. Ecotype SG currently inhabits the understory of mesic forests throughout the Hahajima Islands and has the largest population of any ecotype in the Hahajima Islands (Setsuko et al. 2023). Compared with the Chichijima Islands, the area of mesic forests is larger in the Hahajima Islands (Shimizu, 1999), and the substantial increase in population size of ecotype SG may be due to successful adaptation to the mesic forest environment through migration from the Chichijima Islands to the Hahajima Islands. However, large decreases of more than two orders were observed in ecotype Sm and STm in model g2 (Fig. 4b; Tables S8 and S10). This may be due to the limited number of individuals that migrated from the original island populations to the Mukojima Islands (i.e., the founder effect).