Figure 5 Genetic barriers obtained from BARRIER analysis based on cpDNA in the years of (a ) 2016 and (b ) 2022 ofPrimulina heterotricha populations. Delaunay triangulation (black lines) and inferred barriers (red, blue, and purple lines) separating the different original regions which show the geographic location of the genetic barrier.
The Mantel test showed a significant positive correlation between pairwise genetic distance and geographic distance for both nrDNA and cpDNA for the two study periods (Figure S2). The results of TajimaʼsD test and Fuʼs F s test are presented in Table 2 with the associated simulated P -values. The values for D andF s were positive for 2016 nrDNA sequences (D = 0.80474,P > 0.10; F s = 1.811), 2016 cpDNA sequences (D = 1.38922, P > 0.10; F s = 12.439), and 2022 cpDNA sequences (D = 2.2589, P < 0.05;F s = 11.3), but were negative for 2022 nrDNA sequences (D= -1.18345, P > 0.10; F s = -3.937) (Table 2). These results indicate that our sequences (with the only exception of 2022 cpDNA) are in agreement with the null hypothesis of constant population size and neutral evolution. The hierarchical mismatch analysis showed the distributions of differences for all populations (Figure S3), from which the hypothesis of demographic population expansion can be rejected.
| Discussion4.1 | Very quick effects of anthropogenic barriers on phylogeographical patternsThe evaluation of the effects of habitat fragmentation on animal and plant species has been traditionally addressed by comparing gene flow, genetic differentiation, and genetic structure of populations on fragmented and non-fragmented habitats simultaneously (Chung et al., 2014a; Gao et al., 2015; Schlaepfer et al., 2018). For example, Su et al. (2003) compared the genetic differentiation of populations separated by the Great Wall of China with those not subjected to this huge physical separation for six plants, and found attributable differences to the presence of the wall during 600 years. On a second example, the levels of genetic divergence in two crayfishes (Faxonius validus and Faxonius erichsonianus ) were significantly higher in streams impounded for 36–104 years than in non-impounded streams in Alabama, United States (Barnett et al., 2020). Direct comparisons (i.e. comparing before and after the fragmentation event the populations that have actually been subjected to such disturbance) are probably not available in the literature as it is assumed that the time elapsed from fragmentation should be very long, as enough generations should have passed to observe genetic changes. In their meta-analysis, Schlaepfer et al. (2018) found that effects are generally only observable after 50 years, although some exceptions could apply, as cases in which only 1–5 generations have passed. Herein we have measured the fragmentation effects of anthropogenic construction by using the same populations of a plant (Primulina heterotricha ) two times within six years (i.e. two generations of the study plant would have passed). We have detected a significant geographic structure in eight populations (Figures 2, 3), and AMOVA shows a high level of genetic variation between regions (Table 4), indicating significant genetic differentiation and limited gene flow among populations and regions. Both the phylogenetic tree and the STRUCTURE results indicated that there are three distinct genetic lineages, corresponding to three clades, i.e., northwest clade (NW region), southwest clade (SW region), and southeast clade (SE region) (Figures 2, 3). As one may expect for a species inhabiting a rugged mountain area (and where some rivers were dammed since the late 20th century), such a significant geographic structure was already detected before the expressway construction (2016). Although the effects are not still considerable (probably due to the insufficient time elapsed), we have been able to detect some changes in the genetic structure of P. heterotricha just a few years after the two expressways were completed. The three clades detected for P. heterotricha would likely predate the formation of the anthropogenic barriers of the Daguangba Reservoir and Expressways 1 and 2, as the range of this species, which occurs in mountains at relatively high elevations, is naturally fragmented by river valleys. As it can be observed in Fig. 1 (and Figures 4, 5), Expressway 1 has been constructed along the Changhua River (whose valley is, at some parts, up to 3 km wide) that crosses Hainan’s south-central mountain system in a SW–NE direction, and then the river turns into the NW, where it is dammed by the Daguangba Reservoir. Expressway 2 has been constructed taking advantage of the Tongshi River valley in spite of being much narrower. However, as our results show, the anthropogenic constructions would have intensified the isolation effects of the Changhua River and its associated valley, which has been cultivated since a long time ago (Xiao et al., 2012). In addition, conventional roads were built in these river valleys much earlier; for example, road G224, which was completed in 1954, runs nearly the same route as the two expressways. The changes in the significance of barriers before and after road construction in the BARRIER analysis (Figures 4, 5) suggest that geographic isolation caused by human constructions is key for understanding the present phylogeographical patterns of P. heterotricha. Most of the first barrier for both the nrDNA and cpDNA for the year 2016 corresponds to the separation between QX and WZ populations, which does not match with any of the anthropogenic constructions. In contrast, for the year 2022 the first barrier coincides with the dam and the expressways for both sequences. Another undisputable signal showing the important role of the anthropogenic barriers on the genetic structure of P. heterotricha is the sharp changes in the genetic affinities of population XA, both for nrDNA and cpDNA (Figure 3), which may be caused by the isolation of Expressway 1 but specially Expressway 2. Although it is hard to discern the relative contribution of valleys and rivers on one hand, and expressways on the other hand, the above-described changes in the generic partners from 2016 to 2022 suggest that the two expressways might contribute to shaping the phylogeographical patterns of the island-endemic P. heterotricha within six years, corresponding to two generations of the plant. Geographical discontinuity including anthropogenic disturbance is an important factor in population differentiation by weakening or blocking gene flow in many plant species (Slatkin, 1985; Su et al., 2003; Kartzinel et al., 2013; Chung et al., 2014b). These effects are even greater for small herbaceous plants like P. heterotricha , with a poor dispersal potential associated with small seeds dispersed largely by water.4.2 | Effects of anthropogenic barriers on genetic structure
Although the different clades of P. heterotricha keep stable population dynamics, as revealed by mismatch distribution analysis (Figure S3), the plant would have suffered some detectable effects on its genetic structure, caused by the reduced exchange of genes. Notably, the sharing of ribotypes among the three groups of populations separated by the expressway network almost disappeared, and population genetic differentiation based on nrDNA increased both at the population level (F ST[2016]= 0.0045, F ST[2022] = 0.0054, P< 0.01) and at the region level (Table S3) from 2016 to 2022. In contrast, F ST remained invariable along time with cpDNA at the population level (0.0098 in 2016 vs. 0.0097 in 2022; Table S2) and even increased at the group level (F ST only decreased between NW and SE regions, which could be attributable to the Expressway 1 construction; Table S4). This incongruence between nrDNA and cpDNA could stem both from the different modes of inheritance of these two markers (nrDNA is biparentally inherited, cpDNA is only maternally inherited) and from the fact that pollen migration rates are often much higher than seed migration rates (Ennos, 1994; Petit et al., 2005). Thus, one could expect that the maternally-inherited cpDNA (which is only transmitted by seeds) would be much less sensitive to the fragmentation effects, particularly if we take into account that the anthropogenic barriers would have affected pollen flow in a much higher extent than seed flow (see below).
Genetic differentiation was higher between populations separated by the dam than populations separated by the expressways (Tables S3, S4), i.e. the Daguangba Reservoir showed a more obvious and significant barrier effect on gene flow. Two factors could explain such an observed pattern. Firstly, the reservoir was built in 1994 and nearly a period of 30 years of isolation would have accumulated much more barrier effects on plant dispersal than the newly completed expressways, with a six-year period of effects or two generations of P. heterotricha . Secondly, the dam has a water reservoir of about 6 km long and a water surface of 100 km2. Such physical impediment probably brings much more barrier effects than the expressways, which are normally only 20–30 m wide. For P. heterotricha , its seeds are small and may be largely dispersed by raindrops or water courses and its pollination needs small-sized insects such as Amegilla leptocoma and A. yunnanensis (Ling et al., 2017a). Such dispersal mechanisms would have been affected by the Daguangba Reservoir, as dams could restrict hydrochory (Andersson et al., 2000). Effects, however, would be much more pervasive regarding pollen dispersal, as the flying distances of these small insects would be hardly enough to connect the two banks of the reservoir (separated by up to 4 km).
Despite having less impact than the reservoir, the effects of expressways are still notable forP. heterotricha . In addition to the physical barrier posed by the two high-capacity roads, the rapid growth of traffic (especially after the end of the COVID-19 pandemic) with much more noise and accumulation of pollutants will increase the levels of disturbance. The fast-moving traffic and changes in wind conditions associated with expressways probably affected pollinator movements of P. heterotricha thus decreasing pollen flow. Negative effects of roads on the movement (and even survival) of pollinators have been often detected (Stephens et al., 2000; Bhattacharya et al., 2003; Baxter-Gilbert et al., 2015; Fitch & Vaidya, 2021; Dániel-Ferreira et al., 2022). Although road construction might change hydrological conditions (by affecting natural flow pathways and water quality, e.g. Buchanan et al, 2013), we have not been able to find examples in the literature of roads affecting hydrochory.
In conclusion, we provide evidence of the increased barrier effects of anthropogenic constructions such as reservoirs and expressways on genetic structure and phylogeographical patterns within just two generations of this plant. These disruptive effects involving habitat fragmentation may pose primary threats to population regeneration, genetic diversity, and change the evolutionary processes of plants, especially in endemic, threatened short-lived plants. To alleviate such negative pressure, we suggest establishing ecological corridors to enhance gene flow between populations separated by these anthropogenic barriers. These could include road tunnels and overpass woodlands, which enhance the dispersal of seeds, including the movement of insects and reptiles (such as lizards and frogs). Such ecological corridors might also increase the vegetation integrity and habitat continuity between the two sides of these barriers and thus be helpful for the long-term persistence of the island-endemic P. heterotricha and other rare and endangered species.