2.3. Chloroplast phylogenomics and diversification analyses

All sequence alignments were performed with MAFFT v7.471 (Katoh et al., 2019). Two different datasets were used to reconstruct Comastoma phylogeny: The first one included the complete genome of chloroplast while the second dataset included only protein‐coding sequences. Both the datasets included 35 ingroup and five outgroup species (details below) and were analyzed with maximum likelihood (ML) with IQ‐TREE v1.610 (Nguyen et al., 2014) and Bayesian inferences (BI) statistics MRBAYES v.3.2 (Ronquist & Huelsenbeck, 2003). Optimal substitution models were assessed with jMODELTEST using Akaike information criterion (Nguyen et al., 2014; Posada & Buckley, 2004). The ML analysis was performed using the GTR+F+R3 model and 1000 bootstrap replicates. The BI was conducted with two parallel runs of one million MCMCs, sampling after 1000 generations. A consensus tree was estimated after the first 25% of trees were discarded as burn‐in. In both ML/BI approaches, the trees were rooted with the outgroups including Calotropis procera (Asclepiadaceae), Carissa macrocarpa (Apocynaceae), Mitragyna speciose, Dunnia sinensis, and Emmenopterys henryi (Rubiaceae).

To assess the evolutionary history, we calibrated the divergence time of Comastoma and its related groups as implemented in BEAST (Drummond & Rambaut, 2007). For this analysis, we only used the whole chloroplast dataset in a concatenated fashion setting GTR as a substitution model with lognormal relaxed clock (Thomas et al., 2007).

We used 45‐Ma‐old infructescence and fruit fossils recovered for genus Emmenoptery (Rubiaceae) and seed fossil of 5 Ma age of Gentiana. Divergence times have been calibrated with sect. Cruciata (Gentianaceae) as this is the most recent common ancestor of genus Comastoma (Favre et al., 2016; Pirie et al., 2015).

The analysis was followed for 50 million generations with three independent MCMC runs, taking samples after every 5000 generations. We used the program Tracer v1.7.1 (http://tree.bio.ed.ac.uk/software/tracer/) to ensure that the ESS (effective sampling size) of each parameter was more than 200. After all parameters converged to exceed 200 ESS, we used Logcombiner v2.5.2 (https://www.onlinedown.net/soft/84600.htm) to combine all three runs. Finally, we used the program Tree Annotator v2.5.2 (Luo & Zhao, 2011) to generate the tree after burning the first 20% MCMC. The trees were displayed, annotated, and saved with the help of Figtree v1.4.0 (http://tree.bio.ed.ac.uk/software/figtree/).

3.2. Phylogenomics and diversification time

Both the ML and BI statistics recovered trees with the same topologies based on the whole chloroplast genome. Similarly, we recovered congruent trees based on only the protein‐coding sequence data. In all results, Comastoma species formed a clade sister to a couple of Lomatogonium species; together these taxa were sister to a clade containing other Lomatogonium species with Gentianella (Figure ​(Figure4).4). The basal node of this clade was highly supported in both ML and BI reconstructions, suggesting Comastoma shares a common ancestor with genus Lomatogonium. Depending on the subset of data used, there were subtle inconsistencies in support for some branches and conflicting relationships among Comastoma species. In the ML/BI trees based on the whole chloroplast genome sequence, Comastoma pedunculatum has the closest relationship with C. polycladum, and then, these two species clustered together with Comastoma falcatum, while C. pedunculatum first aggregated with C. falcatum and then C. polycladum in the ML/BI trees using the protein‐coding sequence. The robustness of the phylogenies of each branch is relatively high above 90%.

Open in a separate window

FIGURE 4

Phylogeny of Comastoma based on the maximum likelihood and Bayesian inferred trees (a) ML/BI tree based on complete chloroplast genomes of a concatenated data matrix. (b) ML/BI tree including only the protein‐coding sequences. A support rate less than 100 has been shown on the branches, where the first value represents the support rate of ML‐based statistics and the second value that of BI inferences

The genus Gentianopsis is the most basal lineage followed by Halenia corniculata and then Swertia. The relationship among the four species in Swertia is relatively complex, for example, Swertia bimaculata has the closest relationship with H. corniculata rather than with another three species in Swertia. All the 14 species of Gentiana clustered in the same lineage with the species T. volubile (Figure ​(Figure4).4). Time to the most recent common ancestor for all the 35 species of Gentianaceae included coalesced at 21.81 Ma (Figure ​(Figure5).5). The most basal genus Gentianopsis diverged at about 17.89 Ma, followed by Swertia, Halenia, Gentianella, Lomatogonium, and Comastoma. The genus Comastoma diverged from its allied genus Lomatogonium at 7.70 Ma and is the most recently evolved group. Also, the genus Comasatoma further diverged into two subgroups divided the five species into two groups as Comasatoma jigzhiense and Comasatoma pulmonarium; and C. pedunculatum and C. polycladum and C. falcatum. The split between both the Comastoma species occurred about 4.19 Ma. Similarly, T. volubile diverged from Gentiana lineage at 9.45 Ma.

Open in a separate window

FIGURE 5

Molecular dating based on the complete chloroplast genome sequence. “P” represents Pliocene, “Q” represents Quaternary

4.3. Calibrated divergence time of Comastoma

The origin, distribution, and differentiation of many species have a close association with the change in the geology of the QTP (Ebersbach et al., 2017; Ren et al., 2015; Wang et al., 2009). Comastoma and its related groups are mainly distributed on the QTP. The uplift of the QTP is mainly divided into three stages. Based on the molecular clock hypothesis, the differentiation time of Gentianaceae has been estimated to be about 21.87 Ma and the divergence time between Comastoma and its related groups about 7.71 Ma. This time of differentiation is close to that of tectonic change of the QTP (Mosbrugger et al., 2018; Su et al., 2019; Xing & Ree, 2017). The subfamily differentiation of Gentianaceae and the emergence of Comastoma should be mainly influenced by the second and third stages of QTP uplifting. The second stage of QTP uplift occurred in 21–17 Ma, mainly in the southern part of the QTP and Himalayas, while the most massive uplift of the Tibetan Plateau occurred at 10–8 Ma, or more recently, that is after the Miocene (Spicer, 2017; Xing & Ree, 2017). With its rich and diverse ecosystem, the QTP has become a natural place for the convergence and fusion of various biota. In this area, there are multiple biological types and complex floristic components, which are ecosystems that are sensitive to climate change, but it is also the diversification hotspot of many natural populations and the cradle of speciation (Mosbrugger et al., 2018; Muellner‐Riehl, 2019). At 21–17 Ma, the rapid uplift of the QTP dramatically changed the surface pattern of the region, such as the emergence or disappearance of mountains, and river diversion. The complex geographical changes promoted the division of Gentianaceae into different genera in different directions after 21.87 Ma. Under the influence of long‐term geological and structural events, the climatic conditions for the survival of Gentianaceae plants also gradually changed. In the third stage of uplift, the barriers or channels of species diffusion were further accelerated, and the ecological and environmental gradients were expanded, which provided new niches for many plants of Gentianaceae, thus driving the differentiation of Comastoma and Lomatogonium in 7.71 Ma. Similarly, Gentiana appeared 9.45 million years ago, and then, subgenus differentiation occurred at 8.42 Ma, forming most of the species in the sect. Kudoa but except Gentiana stipitata. Then, in 7.03 Ma, the differentiation of sect. Cruciata happened. It shows that the emergence and rapid radiation of this genus are also obviously affected by the third stage of uplift in the QTP.

Comastoma subgenera diverged after 4.19 Ma, from the middle‐late Pliocene to early Quaternary. In this period, the subgenus differentiation of Comastoma is closely related to the influence of previous plateau uplifts (especially the third stage of uplift), but the role of Quaternary glaciers should not be ignored. In the Quaternary, the glacial and interglacial periods alternated periodically (Qiu et al., 2011). Climate changes that fluctuated with the glacial cycles contributed to the complex geographical environment on the QTP. The new niches and geographic isolation promoted the establishment of extraneous species from neighboring regions. At the same time, some species hybridized to produce new species. These factors led to the rapid radiation of the Comastoma. Changes in the monsoon cycle also presented new selective pressures asynchronous with the QTP uplift (Clift et al., 2008; Ji et al., 2005; Tada et al., 2016); such monsoon effects are often overlooked in biogeographic studies of this region. Therefore, we believe that the transformation of the monsoon climate may also be another essential driving force for the evolution of Comastoma and allied taxa.