Species integrity and genetic structure
Species delimitation is of critical importance in conservation biology because the numbers of species reflect attributes that can be compared across taxa and localities, such as threats, richness, endemism, and diversity [52, 53]. Species delimitation based on morphological traits alone can mask the presence of cryptic species or overestimate the diversification of a single species [54]. The integration of several approaches (e.g., species distribution, molecular data and morphological traits) improves our understanding of the patterns of speciation [55].
In Magnolia, some populations of M. pacifica were originally proposed as separate subspecies, which subsequently were raised to full species level [7, 10, 11]. According to this, the number of carpels per fruit and the number of stamens per flower distinguished M. pugana from M. pacifica [10]. While M. vallartensis differed from M. pacifica by the presence of a colorful abaxial base of outer petals, leaf shape and size, fewer carpels and more petals number [11]. Here, we did not find good accordance with this morphological distinction that has motivated the designation of three different species. This was supported by clustering results, the neighbor-joining tree and measures of allelic differentiation. Thus, we hypothesized that these populations are characterized by a combination of genetic clusters and local differentiation due to geographic restrictions on gene flow, which reflect the different levels of organization present in the genetic structure. However, a great proportion of the genetic variation affecting phenotypic adaptive traits cannot be detected with neutral markers, and it is well known that for plants in general, the diversity at phenotypic level is larger than at neutral marker level [56].
Different demographic and historical processes may have led to the observed pattern. Clustering results suggested two values of genetic partition: K= 2 and K= 6 for STRUCTURE and TESS, respectively. Although these genetic clusters explain a very small fraction of the genetic variation, and K= 2 had no statistical support, we observed, however, that for K= 2, D values also recovered a higher allelic differentiation among populations of these two clusters, suggesting a break in allele frequencies. The resulted genetic subdivision (K = 6) may also suggest a higher local differentiation within the Magnolia species complex by the interaction between the response of plants to local climates and the effects of forest fragmentation in the past.
The Mexican cloud forests are the remnants of temperate woody elements that migrated south from northern North America during the early Oligocene [4, 5, 14]. Studies have suggested that since their establishment, mountain regions have undergone repetitive latitudinal and altitudinal migration, promoting the expansion, contraction, and divergence of populations [57]. There are two hypotheses about the role that the Last Glacial Maximum played in the Neotropical cloud forests: the dry refugee scenario, in which species distributions were shifted by aridity (contraction and extinction) and the moist forest hypothesis, in which no significant decrease in precipitation favored downslope migration and population connectivity, while during warm interglacial periods, fragmentation can result in the development of isolation by distance pattern [58].
The inferred gene flow among groups of populations, in particular between two geographically distant populations (SMP and CCO) for which a close relationship was also observed by NJ, and the presence of genetic intermediates between two populations from the west (SJU and SSE) and one population (RVE) from the east cluster (Fig. 3), might support the population connectivity hypothesis. In this context, the presence of isolation by distance and the overall high allelic differentiation obtained among populations, due to limited pollen and seed dispersal under a fragmented landscape, provide additional support for this scenario. The isolation by distance pattern has been found in both temperate and tropical American magnolias [20, 21]. Thereby, we suggest that climatic changes in the recent past impacted forest distribution, leading to fragmentation of the forest in which Mexican Magnolia presently occur [23], and thus explaining the geographic distribution of the genetic variation. The influence of the past glacial cycles was also observed in the Japanese Magnolia kobus [59] and the relict species Platanus orientalis, whose distribution is strongly influenced by humidity and water availability, similar to magnolias [60].
High genetic diversity and no evidence of inbreeding
Population genetic studies of Magnolia species have shown different patterns of genetic diversity and gene flow [19, 20, 22, 23], suggesting that the genetic patterns are dependent on the evolutionary history of populations and the strength of stochastic forces that each Magnolia species had experienced. We found levels of genetic diversity comparable to those obtained for M. acuminata, a species distributed in Canada and the United States, and other species distributed in Japan, such as M. stellata and M. kobus. The high levels of genetic diversity found in these species were partly explained by historical dynamics [19, 59, 61].
In this sense, the moist forest scenario (discuss above) is consistent with the preservation of genetic diversity through historical gene flow between populations [58]. In addition, the long generation times of trees could buffer the loss of genetic variation within populations, contributing to the maintenance of levels of genetic diversity [19]. On the other hand, the lower allelic richness towards the eastern cluster may be explained by the conjunction of a low effective population size and the prevalence of genetic bottlenecks detected in these populations. If this hypothesis is correct, an enhanced effect of genetic drift in this geographic area may remove alleles within populations and directly disturb allele frequencies, resulting in a decrease of allelic richness eastward and reinforce genetic differentiation. Contrary to what can be expected in small isolated populations, no inbreeding was detected for the studied populations. This can be explained by the protogynous flower of Magnolia and the active movements of its pollinators [62, 63]. Nevertheless, taking into consideration the continuous natural forest fragmentation since glacial periods, and the self-compatibility observed in Magnolia species [63], it is also likely that inbreeding remains undetected because insufficient time has elapsed [65].
Conservation of Mexican Magnolia
It is well known that genetic variability is needed for the long-term viability of populations and species. In this study, we detected high levels of genetic diversity and high allelic differentiation with alleles at relatively low frequencies. Thus, in order to conserve the whole variation, we need to preserve most of the populations throughout the geographic distribution, which can only be achieved by the establishment of federal protection of these forests. At this point, integrating both in situ and ex situ collections are pivotal in the preservation of Magnolia species [15]. Ex situ collections of studied populations have been started by creating seed banks that subsequently may serve as a source of material for the reintroduction and/or reforestation of the same population of origin (in situ source population). Ex situ collections have been demonstrated to be essential to conservation strategies because they can protect an important fraction of the genetic variation, such as in Magnolia officinalis subsp. biloba, among which cultivated populations retain 95 % of the genetic variation of wild populations [65], or in the case of Magnolia vovidesii (formerly M. dealbata), in which ex situ collections have served for reintroduction, propagation and insitu restoration programs in Mexico [66].
At the moment, we did not detect any imminent genetic risks in the Magnolia trees studied in terms of the levels of genetic diversity and inbreeding coefficients. However, given the endangered status of Magnolia species in Mexico, their small effective population size, their scarcity and fragmented habitats, we warn about the negative effects that the persistent anthropic fragmentation may have in further generations. Rates of deforestation are increasing in Mexico [67], producing fragmented forests and isolated natural populations of trees that increase the probabilities of inbreeding and genetic drift that leads to inbreeding depression, genetic degradation and changes in the genetic structure [68, 69].