Genetic variation of the atp B- rbc L noncoding spacer of cpDNA in T. mongolica
In this study, we investigated the phylogeographical pattern and population structure of endangered T. mongolica in western Inner Mongolia. In total, 38 haplotypes were detected at the cpDNA atp B- rbc L locus in T. mongolica. The level of genetic variation of cpDNA is comparable to that of other endangered shrub plants, e.g., Dunnia sinensis (θ = 0.0022) , and Hygrophila pogoncalyx (θ = 0.00343) , but is lower compared to other endangered species, e.g., θ = 0.01018 for the cpDNA trn D- trn T spacer of Cunninghamia konishii , and θ = 0.01268 for the cpDNA atp B- rbc L spacer of Cycas taitungensis . The twofold lower nucleotide diversity to above endangered species in T. mongolica may be ascribed to its extremely small effective population size associated with the low seed set in wild (1.3 - 2.8%) .
Recent habitat loss has reduced the number and size of T. mongolica populations . Small populations of narrowly distributed species are expected to exhibit low levels of genetic variation, but high levels of genetic differentiation among populations, which were all detected in this species (Table 4) . Interestingly, different levels of genetic variation were detected in different populations. The HN and XD populations possessed more haplotypes and higher genetic diversity than others, whereas YKBLG population displayed genetic homogeneity (Table 1). The lack of genetic variability in some populations, e.g. SZS, MSG and TST, near threefold lower in nucleotide diversity, was likely associated with frequently human activities. In contrast, some populations of T. mongolica experienced relatively little disturbance due to low accessibilities [3, 10].
Our previous study revealed medium levels of genetic differentiation among populations of T. mongolica based on ISSR data . In contrast, in cpDNA spacer higher genetic differentiation was detected between populations than in ISSR fingerprinting. The difference may be highly associated with the reproductive characteristics of the species. It has been known that gene flow of seed plants occurs either via pollen prior to fertilization or seeds. In this study, T. mongolica is primarily pollinated by insects . Gene flow between populations via pollen would be limited by the migratory capacity of pollinators. In addition, seed dispersal of seeds from schizocarp, a dry fruit developing from four carpels, is constrained by gravity , likely resulting in most seed dispersal confined to short distances. With maternal inheritance and haploid nature, chloroplast DNA is suitable for estimating the contribution of seed movement to total gene flow , whereas, ISSRs represent nuclear DNA, mostly carried and dispersed by pollen dispersal . In this study, higher genetic differentiation between all populations in cpDNA than in ISSR is likely ascribed to limited seed dispersal.
The BEAST skyline plot for cpDNA spacer identified a recent population decline ever since sixteen thousand years before present likely associated with human destruction as T. mongolica has long been used as firewood (Figure 5) . Ecologically, this plant is still one of the dominant shrubs in Inner Mongolia. Through the analysis of skyline plot, we were able to recover the history of a long term human disturbance that caused a decline in population size of T. mongolica.
Another major factor that shaped the phylogeography and population demography is the frequent flooding of the Yellow River, the second longest river in China . The floods not only eroded river banks, but resulted in many habitats submerged, inevitably leading to population extinction. In addition to bank erosion, the Yellow River is well known for its heavy load of silt. Soil deposits elevate the riverbed and cause flows between natural levees. The river may break out of the levees into the surrounding lower flood plain and adopt a new route. Records indicate that the events have occurred about once every century . Such devastations caused dramatic changes of flora and fauna along the Yellow River. Geological records indicate that the river's levees were breached more than 1,500 times and its course changed 26 times in the last 3,000 years . Given such frequent flooding, T. mongolica would have experienced demographic fluctuations over and over. That is, severe periodical population bottlenecks followed by subsequent demographic expansion would elevate genetic drift effects and lead to a loss of genetic variation [33, 34].
Phylogeography and conservation of T. mongolica
In this study, gene genealogy of cpDNA in T. mongolica was recovered (Figures 2 and 3). Eight cpDNA clades were identified in the NJ tree. Most of the populations contained only one clade in the genetic composition, displaying a pattern of most genetic variation residing between populations. Nested contingency analysis discriminating the geographical associations of haplotypes and clades provides further insights into historical events that shaped the phylogeography (Figure 4). At the total cladogram, restricted levels of Dc vs. a large Dn illustrates restricted gene flow with isolation by distance as the primary process responsible for the present-day distribution of T. mongolica in Inner Mongolia. As cpDNA is maternally inherited, this inference indicates limited seed dispersal. Besides, long distance colonization was also observed in clade 3-1, a common phenomenon occurring over glacial maxima . Furthermore, past fragmentation observed in clade 3-2 was likely associated with the Yellow River flooding.
It is expected that endangered species that are narrowly distributed and own a small population size would have high risks of extinction, especially when gene flow between populations is restricted [1, 35]. Another consequence of a small-sized population is the susceptibility to inbreeding, which reduces heterozygosity and the performance of various fitness-related traits, thereby substantially increasing the probability of extinction [36, 37]. Given small sizes in the wild populations of T. mongolica, the maintenance of genetic diversity would be critical for the long-term survival of species in considering conservation strategies . Historical demographic events in a species play an important role in determining the present-day geographic structure of intraspecific genetic variation . In this study, given the high levels of population differentiation between populations and low levels of genetic diversity within populations of T. mongolica, for retaining the existing diversity, reservation regions covering major populations with high genetic variation should be established.
Habitat destruction and fragmentation would inevitably result in small and isolated populations [40, 41], and increase the risk of population extinction. Today, many natural habitats of T. mongolica around the Yellow River have been destroyed or altered due to human overexploitation. In order to develop an effective strategy to conserve species, the Evolutionarily Significant Unit (ESU) needs to be defined. Various criteria for ESUs have been suggested, including reciprocal monophyly , adaptive variation , and reproductive separation . Recognizing ESUs as reciprocally monophyletic groups ensures that the entire evolutionary heritage within species can be maintained and that populations belonging to different lineages can be managed separately . In T. mongolica, given high levels of genetic differentiation and reciprocal monophyly between most populations and lower genetic variation within populations, each population should be treated as different evolutionarily significant units for conservation (Figure 2).
From another view, conservation efforts can be targeted to genetic hot spots where populations have high levels of genetic diversity [17, 45]. Accordingly, HN and XD populations that owned higher genetic diversity near two-three folds than other populations (Table 1) can be recognized as genetic hot spots of T. mongolica. The concept of genetic and ecological exchangeabilities is also central to the definitions of ESU . Crandall et al.  emphasize that the ESU concept not only includes ecological data and genetic variation, but also considers the ecological and genetic exchangeabilities. In practice, the status of recent and historical genetic and ecological exchangeability between populations is considered. As the contemporary gene flow from the populations at genetic hot spots to other populations does not exist or cannot be determined, the populations at the genetic hot spots need to be treated as distinct conservation units for its unique genetic variation . Accordingly, hotspots at HN and XD of T. mongolica would represent different units for conservation, as seed dispersal between populations is limited.