Molecular phylogeography and species distribution modelling evidence of 'oceanic' adaptation for Actinidia eriantha with a refugium along the oceanic-continental gradient in a biodiversity hotspot

Abstract

Background: Refugia is considered to be critical for maintaining biodiversity; while discerning the type and pattern of refugia is pivotal for our understanding of evolutionary processes in the context of conservation. Interglacial and glacial refugia have been studied throughout subtropical China. However, studies on refugia along the oceanic-continental gradient have largely been ignored. We used a liana Actinidia eriantha, which occurs across the eastern moist evergreen broad-leaved forests of subtropical China, as a case study to test hypotheses of refugia along the oceanic-continental gradient and 'oceanic' adaptation.

Results: The phylogeographic pattern of A. eriantha was explored using a combination of three cpDNA markers and 38 nuclear microsatellite loci, Species distribution modelling and dispersal corridors analysis. Our data showed intermediate levels of genetic diversity [haplotype diversity (h_T) = 0.498; unbiased expected heterozygosity (UH_E) = 0.510] both at the species and population level. Microsatellite loci revealed five clusters largely corresponding to geographic regions. Coalescent time of cpDNA lineages was dated to the middle Pliocene (ca. 4.03 Ma). Both geographic distance and climate difference have important roles for intraspecific divergence of the species. The Zhejiang-Fujian Hilly Region was demonstrated to be a refugium along the oceanic-continental gradient of the species and fit the 'refugia in refugia' pattern. Species distribution modelling analysis indicated that Precipitation of Coldest Quarter (importance of 44%), Temperature Seasonality (29%) and Mean Temperature of Wettest Quarter (25%) contributed the most to model development. By checking the isolines in the three climate layers, we found that A. eriantha prefer higher precipitation during the coldest quarter, lower seasonal temperature difference and lower mean temperature during the wettest quarter, which correspond to 'oceanic' adaptation. Actinidia eriantha expanded to its western distribution range along the dispersal corridor repeatedly during the glacial periods.

Conclusions: Overall, our results provide integrated evidence demonstrating that the Zhejiang-Fujian Hilly Region is a refugium along the oceanic-continental gradient of Actinidia eriantha in subtropical China and that speciation is attributed to 'oceanic' adaptation. This study gives a deeper understanding of the refugia in subtropical China and will contribute to the conservation and utilization of kiwifruit wild resources in the context of climate change.

Keywords: Actinidia eriantha; Climatic fluctuations; Oceanic–continental gradient; Phylogeography; Refugium; Subtropical China; ‘Oceanic’ adaptation.

Fig. 1

Geographical distribution of A. eriantha cpDNA haplotypes, BEAST-derived chronograms and TCS network. a Geographical locations of the 28 populations and distributions of 23 chloroplast haplotypes of A. eriantha examined in this study (the scale on the map represents meters above sea level). The three dashed lines correspond to the three population groups (“Southeast”, “Southwest” and “Main part”) identified by the program SAMOVA. b BEAST-derived chronogram of A. eriantha based on cpDNA sequences. Blue bars indicate 95% HPD credibility intervals for nodes of particular interest with ages (in Myr ago, Ma). Only bootstrap values higher than 50% are denoted above branches. c TCS-derived network of 23 chloroplast haplotypes. Each circle means a unique haplotype, with circle size reflecting its frequency. Small black circles mean missing haplotypes

Fig. 2

The results of bayesian skyline plots (BSP) and mismatch distribution analysis (MDA) of the “Main part” inferred from A. eriantha cpDNA. a Bayesian skyline plots (BSP) estimated using BEAST2 v. 2.4. The thick solid blue line is the median estimate, and the area delimited by the light blue broadband represents the highest posterior density (HPD) 95% confidence intervals for Ne. b Mismatch distribution analysis (MDA) estimated in Arlequin v. 3.5

Fig. 3

Genetic landscapes for A. eriantha: a genetic diversity based on Ae (No. of effective alleles); b genetic diversity based on UH_e (unbiased expected heterozygosity); c genetic divergence based on F_ST [F_ST = (H_t - H_e) / H_t, H_t means total expected heterozygosity, H_e means expected Heterozygosity]. The values of A_e, UH_e, F_ST have been standardized as [0,1]

Fig. 4

Population genetic structure of Actinidia eriantha.a Histogram of the Bayesian assignment for 28 populations (629 individuals) and the hierarchical Bayesian assignment for 17 populations (293 individuals) of A. eriantha based on genetic variation at 31 neutral EST-SSR loci using STRUCTURE. Each vertical bar represents one individual and its probability of membership for each of the K = 2 and hierarchical K = 4 clusters. b Geographic origin of the 28 A. eriantha populations and their color-coded grouping according to the STRUCTURE analysis. c The un-rooted NJ tree of 28 population revealed by 31 neutral nSSR data. d Principal coordinates analysis (PCoA) of A. eriantha based on their genetic distances (D_A) derived from 31 neutral nSSRs

Fig. 5

Seven evolutionary scenarios for five clusters of A. eriantha tested with approximate Bayesian computation (ABC) analyses using DIYABC. Prior distributions of model parameters were set for effective population size of five sampled clusters (N1–N5) and 4 founder clusters (N1F–N5F, except for N3F), duration of bottleneck after colonization event (DB) and relative timing of events in number of generations (t1–t4). Posterior probabilities of the seven scenarios obtained by logistic regression of 1% of the closest simulated datasets are shown on the top of each scenario. Scenario outlined in red is the best option

Fig. 6

Potential distributions of A. eriantha, West Distribution (WD) and East Distribution (ED) predicted using MaxEnt based on five bioclimatic variables representing the LIG, LGM, current and future climatic conditions, respectively. Warmer colors denote areas with a higher probability of presence. Dots show the extant occurrence points of the A. eriantha

Fig. 7

The dispersal corridors for A. eriantha based on chloroplast haplotypes at the LIG, LGM and current, respectively. The values of dispersal route have been standardized as [0,1]