Biogeographic diversification of Eranthis (Ranunculaceae) reflects the geological history of the three great Asian plateaus

Abstract

The evolutionary history of organisms with poor dispersal abilities usually parallels geological events. Collisions of the Indian and Arabian plates with Eurasia greatly changed Asian topography and affected regional and global climates as well as biotic evolution. However, the geological evolution of Asia related to these two collisions remains debated. Here, we used Eranthis, an angiosperm genus with poor seed dispersal ability and a discontinuous distribution across Eurasia, to shed light on the orogenesis of the Qinghai–Tibetan, Iranian and Mongolian Plateaus. Our phylogenetic analyses show that Eranthis comprises four major geographical clades: east Qinghai–Tibetan Plateau clade (I-1), North Asian clade (I-2), west Qinghai–Tibetan Plateau clade (II-1) and Mediterranean clade (II-2). Our molecular dating and biogeographic analyses indicate that within Eranthis, four vicariance events correlate well with the two early uplifts of the Qinghai–Tibetan Plateau during the Late Eocene and the Oligocene–Miocene boundary and the two uplifts of the Iranian Plateau during the Middle and Late Miocene. The origin and divergence of the Mongolian Plateau taxa are related to the two uplifts of the Mongolian Plateau during the Middle and Late Miocene. Additionally, our results are in agreement with the hypothesis that the central part of Tibet only reached an altitude of less than 2.3 km at approximately 40 Ma. This study highlights that organismal evolution could be related to the formation of the three great Asian plateaus, hence contributing to the knowledge on the timing of the key tectonic events in Asia.

1. Introduction

The collisions of the Indian Subcontinent with Eurasia in the east and Arabian plate with Eurasia in the west are two important tectonic events in Asia [1]. The India–Eurasia collision started at about 61 Ma [2] and subsequently caused the deformation and uplift of the Qinghai–Tibetan Plateau (QTP) and the eastward lateral extrusion of Indochina [1,3,4]. In addition, the Mongolian Plateau (MP) is usually considered to be the consequence of the far-field effect of the continued India–Eurasia convergence [5]. The Arabia–Eurasia collision started at ca 35 Ma and resulted in the deformation and uplift of the Iranian Plateau (IP) region and the westward lateral extrusion of the Anatolia block [6]. The India–Eurasia and Arabia–Eurasia convergences have usually been regarded as separate collision processes, but they may have interacted by controlling the far-field Cenozoic deformation in Asia [1]. This interaction may have generated a northeast oriented zone of Central Asia right-slip faults (300–400-km wide and greater than 1500-km long), which encompassed the Zagros mountains, Tian Shan and Altai Shan orogens and western Mongolia (electronic supplementary material, figure S1). The uplifts and outward growths of the QTP, IP and MP regions and associated mountain belts have had a profound impact on regional and even global climates and environments [7–10]. Yet, the evolution of these three great Asian plateaus remains considerably controversial and is still poorly understood [1,11,12]. The tectonic events affected biotic evolution; in turn, evolutionary diversifications of organisms might could reflect the geological and related climate changes. However, with few exceptions [13,14], evolutionary studies of Asian organisms have focused on a single plateau, such as the QTP [15,16]. This precludes broad comparisons of the geological evolution of Asian plateaus associated with the India–Eurasia and Arabia–Eurasia collisions as a whole.

Ranunculaceae, commonly known as the buttercup family, is an important herbaceous element of mountain ecosystems in the Northern Hemisphere [17]. The family originated in the Lower Cretaceous and experienced a burst of diversification during the Upper Cretaceous (ca 83–74 Ma) [18]. Owing to their seeds with poor dispersal abilities, the ranunculaceous species with follicles are often reported to be geographically highly isolated [18] and have been used to infer East Asian floristic evolution [19,20]. In this study, we focus on the biogeographic diversification of the genus Eranthis, an early spring ephemeral, known as winter aconite in Europe. Recent molecular clock estimates suggest a stem age of 56–59 Ma for the genus [18]. Eranthis mainly inhabits humid environments under deciduous forests of temperate Eurasia (figure 1a). The fruits of Eranthis are dehiscent follicles (figure 1b). Seeds may be simply gravity dispersed owing to lacking obvious adaptation to wind or animal dispersal, and hence are not expected to disperse over long distance or through physical barriers. The species of the genus are found throughout the Eurasian continent, but their distribution is discontinuous and each species is restricted to a particular biogeographic region, i.e. South Europe, South Turkey, the west and east edges of the QTP, MP or Northeast Asia (figure 1c; electronic supplementary material, table S1). Furthermore, species of the genus occur both in lowlands and at high elevations in different geographical regions (electronic supplementary material, figure S2). The present-day distribution of Eranthis may be linked to the geological evolution of Asia, especially the formation of the three great Asian plateaus and the extrusion of the Anatolia block.

Figure 1.

Phylogenetic relationships of Eranthis inferred from the combined plastid and nuclear sequences. Numbers above branches are bootstrap values (BS > 50%) and Bayesian posterior probabilities (PP > 0.80). Asterisks indicate BS = 100% and PP = 1.0. (a) An early spring Eranthis plant under the deciduous forest. (b) Dehiscent follicles of Eranthis. (c) The map shows sampling locations, detailed sampling information is presented in electronic supplementary material, table S3. Photographs by Andrey S. Erst. (Online version in colour.)

Given the inferred Late Palaeocene age and extensive Eurasian distribution of the genus, a poor dispersal ability, as well as the restriction of each species to particular biogeographic regions, Eranthis is a key biological model to infer the geological history of Asia. Here, we present a molecular phylogeny for Eranthis that includes all currently recognized species. By integrating phylogenetic, molecular dating and biogeographic methods, we investigate the historical biogeography of the genus, and thereby aim to illuminate the likely evolutionary histories of the three great Asian plateaus.

2. Material and methods

(a). Samples and sequences

We sampled 34 individuals of Eranthis (Cimicifugeae, Ranunculaceae), representing all 12 species (including one undescribed new species) [21]. We included nine species covering all three other genera of Cimicifugeae, as well as three species of Asteropyreae, Caltheae and Adonideae as the outgroups [18]. Materials used in this study were mainly collected from herbarium specimens except for a few from the field. Voucher information and GenBank accession numbers are listed in electronic supplementary material, table S3.

Five DNA regions, including plastid (trnL-F, trnH-psbA, matK, rbcL) and nuclear (ITS) loci were sequenced. Laboratory procedures, sequence handling and phylogenetic inference followed Xiang et al. [19]. Sequences were aligned using Geneious 10.1.3 [22], and adjusted manually. The final alignments contained 46 taxa with 1001 bp for trnL-F (excluding three ambiguous regions with 114 nucleotides), 251 bp for trnH-psbA (excluding one ambiguous region with 73 nucleotides), 1210 bp for matK, 1263 bp for rbcL and 668 bp for ITS. Phylogenetic analyses were conducted using maximum likelihood and Bayesian inference methods in RAxML 8.0 [23] and MrBayes 3.2.1 [24], respectively. The plastid and nuclear datasets were each first analysed independently. As there were no significant topological incongruences (greater than or equal to 70% bootstrap percentage and greater than or equal to 0.95 posterior probability) between plastid and nuclear datasets (electronic supplementary material, figure S3 versus electronic supplementary material, figure S4), all subsequent analyses were performed for the combined five-locus dataset only.

(b). Divergence time estimation

Divergence time was estimated in BEAST 2.4.0 [25]. We set an 83 Ma (95% HPD: 77–88 Ma) constraint for the root, which is the estimated age for the split of Cimicifugeae and Adonideae [18], with a normal distribution. Palaeoactaea nagelii, fruits from the Late Palaeocene Almont and Beicegel Creek floras of North Dakota, USA, is close to extant Actaea based on fruit size and shape and details of seed morphology and anatomy [26,27]. In accordance with previous studies [18,28,29], we gave a lower bound of 56 Ma for the stem group age of Actaea. Dating analyses were conducted under the GTR + I + Γ model for each DNA partition separately. Two independent runs of 100 million generations were carried out, sampling every 5000 generations and discarding the first 25% as burn-in. The maximum clade credibility (MCC) tree with median node ages was produced using TreeAnnotator 2.4.0 [25].

We used Bayes factors [30] calculated by marginal likelihoods derived from path sampling [31] to compare four clock models (strict, exponential, lognormal and random) and then two distinct tree speciation priors (Yule and birth–death). Under the optimal clock model (electronic supplementary material, table S4) and tree prior (electronic supplementary material, table S5), we carried out four dating analyses by using four distinct prior distributions for the calibrating fossil (exponential, lognormal, normal and uniform). These four analyses yielded highly similar time estimates for Eranthis (electronic supplementary material, table S6). Bayes factors suggested normal prior distribution as optimal for the calibrating fossil (electronic supplementary material, table S7); results from this analysis were thus reported and used for subsequent analyses. For comparative purposes, we also estimated divergence times using a pruned dataset consisting of 24 taxa (one accession per species). The run strategy was the same as the one for the full-data analysis. Both the full-data and species-level analyses highly agree in time estimates for Eranthis (electronic supplementary material, table S6).

To test the robustness of the age estimates, we also used the molecular dataset of Eranthis in combination with DNA substitution rate. Four plastid loci were linked in a single clock and a general plastid rate calculated by Zurawski et al. [32] was used. A normal distribution with a mean value and standard deviation of 0.001 and 0.0001 were specified in ucld.mean. For the nuclear ITS, we selected the substitution rate of herbs used by Wojciechowski et al. [33]: a normal distribution with a mean value and standard deviation of 0.0035 and 0.0002. The run settings were the same as the one using the full dataset and the fossil calibration under the optimal prior distribution. The dating results in combination with DNA substitution rate are highly congruent with those that uses only the calibrating fossil (electronic supplementary material, table S8).

(c). Ancestral range reconstruction

We estimated the ancestral ranges of Eranthis under the DEC and DIVALIKE models, implemented in the R package BioGeoBEARS [34]. The three outgroups were pruned. Based on the present-day distribution of Eranthis and the floristic regions it covers, we defined five bioregions (figure 1a): Mediterranean region (including South Europe and the Anatolia peninsula), west QTP, east QTP, Northeast Asia (including eastern Siberia, northeastern China, Korea and Japan) and MP. North America was also coded as a bioregion, in which a few species of Actaea occur. We performed the analyses in several different ways: without the maximum number of ancestral areas (maxarea) constrained and with the maxarea constrained to four, three or two. Using the Akaike information criterion, DIVALIKE was selected to be the best-fitting model for each of these four analyses (electronic supplementary material, table S9). Under the DIVALIKE model, the results of the four analyses were largely similar based on cumulative ancestral range reconstructions [35]. Each species of Eranthis is distributed in one bioregion, the ancestral range reconstruction constraining maxarea to two was thus reported and discussed. To exclude the impact of outgroup areas, we also performed biogeographic analyses by pruning all outgroups. Both biogeographic estimations with or without outgroups entirely agree with the assigned ancestral areas for Eranthis (figure 2 versus electronic supplementary material, figure S6).

Figure 2.

Evolution of Eranthis with associated tectonic events. Timetree is modified from the electronic supplementary material, figure S3, generated in BEAST under the best-fitting clock model, tree prior process and fossil prior distribution. Pie charts show the cumulative probabilities for estimated ancestral ranges obtained under the DIVALIKE model, implemented in BioGeoBEARS. (a) The QTP region experienced the first uplift around 40 Ma and its central area reached an altitude of less than 2.3 km. (b) A second uplift of the QTP region and the resulting adjacent mountain buildings and Asian interior desertification occurred at about 26 Ma. (c) Uplift of the IP region and extrusion of the Anatolia block happened at almost the same time, around 12 Ma. (d) A second uplift of the IP and MP regions occurred around 6 Ma. The northeast-trending grey area in (d) shows the interacted zone between the India–Eurasia and Arabia–Eurasia collisions [1]. A series of palaeo-maps were reconstructed from http://www.odsn.de/odsn/index.html (link ‘Plate Tectonic Reconstructions'), and tectonic structures are modified from Yin [1]. AB, Anatolia block; AS, Altai Shan; IP, Iranian Plateau; LB, Lunpola basin; LM, Lüliang mountains; LT, Lhasa terrane; MP, Mongolian Plateau; QL, Qinling; QT, Qiangtang terrane; QTP, Qinghai–Tibetan Plateau; TS, Tian Shan; ZM, Zagros mountains. (Online version in colour.)

(d). Ancestral elevation reconstruction

We obtained locality data of the 11 known Eranthis species from the following herbaria: B, E, IRK, K, LE, MO, NS, NSK, P, PE, TI and TK (herbarium acronyms follow Index Herbariorum: http://sciweb.nybg.org/science2/Index-Herbariorum.asp), as well as from our field investigations. Localities without accurate coordinates were omitted. In total, we obtained 279 occurrence records with locality information (electronic supplementary material, table S2). For each record, we extracted the elevational variable from the calibration dataset for global land areas at 1-km resolution [36] using the ‘raster’ function in the ‘raster’ package of R [37]. The mean and boundaries of the 95% elevational range for each species was then calculated. We inferred the ancestral elevation of Eranthis using a squared-change parsimony method in Mesquite 3.61 [38]. This method is relevant for analysing continuous characters [39]. The MCC tree was used and a single accession per species of Eranthis was kept. For comparative purposes, we also inferred the ancestral elevation using the phylogenetic independent contrast (PIC) method, which is a Brownian-motion based estimator. The PIC analysis was conducted using the ‘ace’ function in the ‘ape’ package of R [33]. These two analytical methods generated highly congruent results (electronic supplementary material, figure S7). The parsimony reconstruction was reported and used for discussion (figure 3).

Figure 3.

Ancestral reconstruction of the mean and boundaries of the 95% elevational range using Mesquite. Box and whisker plot indicates the elevational range of each species. Numbers above the branches are the mean and boundaries of the 95% elevational range (in metres) for ancestral nodes. Branch colours correspond to the four clades presented in figure 1. (Online version in colour.)

To test the robustness of the ancestral elevation inferences, we also used elevation information from the literature. The evolutionary reconstructions for boundaries of the elevational interval were carried out using the parsimony method in Mesquite. Both elevation reconstructions using the information extracted based on the occurrence records and from the literature yielded similar results (figure 3 versus electronic supplementary material, figure S8).

3. Results

(a). Phylogeny

Maximum likelihood and Bayesian inference analyses resulted in identical trees (figure 1). Both supported the monophyly of Eranthis and recovered Actaea as its sister group. Eranthis includes two major monophyletic groups (I and II) and four well-supported clades (I-1, east QTP; I-2, North Asia; II-1, west QTP; II-2, Mediterranean), each defined largely by geography. Within clade I-2, the MP taxa clustered together and were nested within a clade of Northeast Asian taxa. The relationships at the species level within Eranthis are well resolved (figure 1).

(b). Divergence times

The divergence time estimates for Eranthis from the BEAST analysis under the optimal clock model (electronic supplementary material, table S4), tree prior (electronic supplementary material, table S5) and fossil prior distribution (electronic supplementary material, table S7) are shown in figure 2, and more details are provided in electronic supplementary material, figure S3 and electronic supplementary material, table S6. The genus Eranthis started to diversify at ca 41 Ma (95% highest posterior density (HPD): 25–54 Ma). Within group I, the east QTP taxa (I-1) diverged from the North Asian taxa at ca 26 Ma (95% HPD: 14–41 Ma). The split of the MP subclade and its Northeast Asian sister group occurred at ca 17 Ma (95% HPD: 8–29 Ma), and the MP subclade diversified around 6 Ma (95% HPD: 2–13 Ma). Diversification of group II occurred at ca 12 Ma (95% HPD: 4–24 Ma). The split between southern European and Anatolian species occurred at ca 6 Ma (95% HPD: 2–13 Ma).

(c). Ancestral range reconstruction

The ancestral area reconstructions using the DEC and DIVALIKE models in BioGeoBEARS generated highly congruent results, although DIVALIKE was identified as the best-fitting model for our data (electronic supplementary material, table S9). Results under the DIVALIKE model are shown in figure 2. The genus Eranthis originated in East Asia, and its most recent common ancestor had likely spread to the west QTP and Mediterranean regions. A subsequent vicariance event led to group I in East Asia and group II in the west QTP, Anatolia and South Europe. Within group I, a vicariance event separated the east QTP clade (I-1) from the North Asian clade (I-2), and a dispersal event from Northeast Asia to the MP within clade I-2 took place. Within group II, a vicariance event isolated the west QTP clade (II-1) and the Mediterranean clade (II-2).

4. Discussion

Our study presents a phylogenetic tree for Eranthis with all 12 species (figure 1). Divergence time estimates suggest a crown-group age of 62 Ma (95%HPD: 57–68 Ma) for Cimicifugeae and a stem age of 57 Ma (95%HPD: 53–60 Ma) for Actaea, which are consistent with the estimates of Wang et al. [18] and Kadereit et al. [29], yet they are much older than those of Zhai et al. [40]. However, Zhai et al. [40] only used the oldest known angiosperm tricolpate pollen as the fossil calibration point to constrain the crown-group age of eudicots and estimated divergence dates within Ranunculaceae. Internal calibrations are regarded critical for obtaining accurate estimates [41]. Detailed comparisons among divergence times estimated in this study and previous studies are shown in electronic supplementary material, table S8.

(a). Spatio-temporal evolution of Eranthis

The results of our integrated molecular dating and ancestral range reconstruction analyses (figure 2) indicate that Eranthis originated in East Asia, and then dispersed to the west QTP and Mediterranean regions between 56 and 41 Ma. This period is prior to the interruption of the Turgai strait (an epicontinental sea from the Arctic Ocean to the Tethys Seaway) (ca 34 Ma) [42] and the demise of the Tarim Sea (an epicontinental sea in the western Tarim Basin of China) (ca 34–37 Ma) [43]. Thus, the migration through the north temperate regions of North Asia seems unlikely. A proto-Tibetan highland consisting of the Lhasa and Qiangtang terranes had been assumed to have attained an elevation of greater than 4.5 km before the collision of India with Eurasia, but these two high terranes were at least separated by an east–west oriented elongate lowland along the Bangong–Nujiang suture, represented by a string of basins at present, such as the Lunpola basin [3,8,11,44,45]. Fossil evidence suggests that a humid tropical or subtropical lowland flora existed in the central QTP until the latest Palaeocene to Late Oligocene [9,44,45]. Thus, it is more likely that ancestral populations of Eranthis would have migrated westward via this east–west lowland during the Eocene (figure 2a).

The spatio-temporal evolution of Eranthis may have resulted from the uplifts of the three great Asian plateaus and associated mountain belts, as well as associated environmental changes. The uplift of the QTP region induced by the India–Eurasia collision might have prevented the migration of Eranthis through the east–west lowland and thereby resulted in the deep divergence between groups I and II at ca 41 Ma (95% HPD: 25–54 Ma). Later, the QTP grew continuously upward and outward and caused the formation of Lüliang and Qinling orogens in the northeast [46,47] and northern Chinese desertification [48]. These events could have caused the extinction of Eranthis in the northern and central China and thereby resulted in the split between clades I-1 and I-2 around 26 Ma (95% HPD: 14–41 Ma). Contrary to the Northeast Asian taxa, the east QTP taxa are distributed at higher elevations (figure 3). Our analyses recovered one dispersal from Northeast Asia to the MP, suggesting Northeast Asia as the source for the MP biodiversity. The split between the MP subclade and its Northeast Asian sister group at ca 17 Ma (95% HPD: 8–29 Ma) is likely related to the uplift of the MP region, which was suggested to be the consequence of the far-field effect of the India–Eurasia and Arabia–Eurasia collisions [1]. The divergence of the west QTP clade (II-1) from the Mediterranean taxa (II-2) around 12 Ma (95% HPD: 4–24 Ma) is probably linked to the uplift of the IP region and the extrusion of the Anatolia block resulting from the Arabia–Eurasia collision.

(b). Biogeographic implications for the geological evolution of Asia

Because of its conserved habitat, poor dispersal ability and scattered geographical distribution across Eurasia, Eranthis possesses unique genetic imprints that reflect the history of geological and environmental changes in Asia during the Cenozoic.

Following the initial India–Eurasia collision in the Early Cenozoic (ca 61 Ma) [2], the uplift and outward growth of the QTP occurred and continued well into the Miocene [8]. The deep divergence between the east and west groups (I and II) took place in the central part of the QTP around 41 Ma, associated with the first uplift of the QTP (figure 2a). This time frame is in concert with the results of geological studies, which indicate that the east and central QTP and its northern margin were elevated around 40 Ma [49–51]. The origin of alpine ginger [52] and the divergence of west Chinese salamanders [53] were estimated to have occurred during the same period, both of which were also suggested to be related to this uplift of the QTP region. The palaeoelevational estimate for the hinterland of the QTP, i.e. a string of basins between the Lhasa and Qiangtang terranes, is crucial for understanding Tibetan landscape evolution, but has been intensely debated [1,9,11]. Our elevation reconstructions suggest that the most recent common ancestor of Eranthis likely occurred up to the highest altitude of about 1.8 km (figure 3). Accordingly, we hypothesize that after the first uplift (ca 40 Ma), the central QTP likely had an elevation of less than 2 km, likely about 1.8 km (figure 2a). This differs from the hypothesis that during the first uplift the QTP reached an altitude of at least 2.5 km [53]. Our estimate is compatible with the result of the magnetostratigraphic and radiochronologic dating that indicates that the elevations of central Tibet were generally low (less than 2.3 km) at 39.5 Ma [45]. Also, the palaeobotanical evidence suggests an elevation of less than 2.3 km for the Lunpola basin of central QTP at ca 25 Ma [11] and an elevation of 1.5–2.9 km for the Kailas Basin in the western part of the Lhasa terrane during the latest Oligocene [54].

The split between the east QTP clade (I-1) and the North Asian clade (I-2) (figure 2b; ca 26 Ma) suggests a second uplift of the QTP region and the resulting palaeoclimatic changes. This age fell in line with the deep splits of lineages in alpine ginger (ca 32 Ma, 95% HPD: 18–50 Ma) [52] and in spiny frogs (ca 27 Ma, 95% HPD: 19–36 Ma) [55], both of which were postulated to have been triggered by this rapid uplift. A few endemic plant genera of the QTP, such as Nannoglottis (Asteraceae) [56] and Gymnaconitum (Ranunculaceae) [57] also originated during this period. This uplift has not only significantly modified Asian physiognomy, but has also greatly affected the regional and even the global climate and biotic evolution [9,49]. During this period, the QTP extended outward and consequently gave rise to adjacent mountain buildings by a far-field effect, such as the Emei mountain in the east [58] and Lüliang and Qinling mountains in the northeast [46,47], the onset of Asian interior desertification (ca 25 Ma) [48] and the establishment of East Asian monsoon climate [59]. Palaeontological evidence shows that radical changes in vegetation and biotic components occurred in the TP and adjacent regions around the Oligocene–Miocene transition [9,60].

The split between the west QTP clade (II-1) and the Mediterranean clade (II-2) (figure 2c; ca 12 Ma) suggests a rapid uplift of the IP region triggered by the Arabia–Eurasia collision, which served slab detachment in eastern Turkey and northwestern Zagros [6]. Our dating is highly consistent with the geological hypothesis that uplift, exhumation and shortening in the Zagros mountains and the IP region accelerated at about 15–12 Ma [6]. The Arabia–Eurasia collision displaced the Anatolia block westward relative to the IP, which occurred at about 15–11 Ma [61] or 7–5 Ma [62,63]. Contrary to the eastward lateral extrusion of Indochina that has been widely investigated [52,55,64], the westward lateral extrusion of the Anatolia block has yet to be examined using modern phylogenetic and biogeographic methods. Our data suggest that the uplift and tectonic extrusion should have occurred synchronously to generate the observed distribution pattern in Eranthis around 12 Ma (figure 2c). The split between the Anatolian and southern European species was estimated at ca 6 Ma, associated with the progressive uplift and expansion of the IP region. This is supported by geological studies, indicating an about 7–8 Ma uplift of the southern margin of the Anatolia block [65] and an about 5 Ma active mountain building of the Zagros mountains [66].

The split of the MP subclade and its Northeast Asian sister group occurred at about 17 Ma. The species in the MP are distributed at relatively higher elevations than the species in the Northeast Asia (figure 3). We hypothesize that the split was associated with geological events that separated the MP and Northeast Asia. Our estimated age is compatible with the geological hypothesis that the rapid uplift of the MP region occurred around 20–10 Ma [67,68]. The diversification of the MP subclade arose at about 6 Ma (figure 2d), which implies a second uplift and expansion of the MP region. This hypothesis is consistent with the geological hypotheses that the summit of the MP emerged at about 5–10 Ma [69] and that the uplift of the Gobi-Altai mountain range of Mongolia started at 5 ± 3 Ma [5]. The two discrete intervals of uplift of the MP region that our analyses revealed are concurrent with those of the rapid uplift and expansion of the QTP and IP regions (figure 2b,c) [6,69,70], and with those of the Tian Shan and Altai Shan [71,72]. The MP, Tian Shan and Altai Shan are all located in the Central Asian right-slip fault zone (electronic supplementary material, figure S1) [1]. Thus, we suggest that the two uplifts of the MP region in the Miocene likely represent two rapid responses to the interaction of the India–Eurasia and Arabia–Eurasia collisions.

In conclusion, our study provides the first case, to our knowledge, of an organismal group reconciling and illuminating the geological history of the QTP, IP and MP. The biogeography of Eranthis may have been shaped by multiple uplifts of these three Asian plateaus and the extrusion of the Anatolia block. Considering the wide credibility intervals of the estimated times of divergence, we urge caution interpreting these results and the geological hypotheses that we put forward should be further tested by studying other taxa that exhibit similar habitat requirements and dispersal abilities in Eurasia.

Data accessibility

The sequences reported in this paper have been deposited in the GenBank database (accession nos. MW716473–MW716501; MW722251–MW722357). The datasets supporting this article are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.0zpc866wz [73].

Authors' contributions

W.W. conceived the research. K.-L.X., A.S.E., H.-W.P., R.d.C.O., F.J. and T.V.E. carried out taxon sampling. K.-L.X., A.S.E. and J.Y. generated all the data. K.-L.X., A.S.E., J.Y., H.-W.P. and W.W. performed the data analyses. W.W. led the writing with contributions from all co-authors.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by the Strategic Priority Research Program of Chinese Academy of Sciences (grant no. XDB31000000), the National Natural Science Foundation of China (grant nos. 31770231, 31770233 and 32011530072), K.C.Wong Education Foundation (grant no. GJTD-2020-05) and the Russian Science Foundation (grant no. 19-74-10082).