Genetic diversity within populations
The model-based clustering method demonstrated that the genetic structure of the M. boninensis remnant trees did not correspond perfectly to the operational populations based on geographical distribution (Fig. 4). Although the OTJ operational population did coincide with the model based clustering, other operational populations contained multiple clusters. However, in determining operational populations to be used in species conservation, there are two considerations of practical importance: 1) the convenience of conducting controlled crossing using trees in close physical proximity, whilst knowing their individual genetic relatedness; and 2) the ability to sample seed that may have been open pollinated from nearby paternal trees. Therefore, it is important to understand the genetic composition at two levels: the operational population and the individual level. The genetic diversity levels in OTJ and CCJ were lower than in the other populations (Table 2). Although the OTJ population contains the secondly largest number of trees of all the operational populations, the genetic composition of each tree is very similar, as demonstrated by the model-based clustering (Fig. 4). This might be the result of a founding event, with a small number of maternal trees having produced the seed to establish the extant OTJ population. However, the low genetic diversity in the CCJ population is different. Four different genetic groups were identified in the CCJ population by model-based clustering analysis. Out of the four genetic groups, some of the trees comprising three of them (blue/red, yellow and green in Fig 4) are located in the northern part of Chichi-jima Island. This is an area inhabited by immigrant people. Trees in these same genetic groups were also present in KWK on Haha-jima Island. Therefore, these trees in CCJ were probably planted by people using seed sources from Haha-jima Island. In contrast, the vegetation of the south of Chichi-jima is, to date, well preserved although there was some disturbance during World War II. These trees have retained their endemic genetic composition (purple in Fig. 4) on Chichi-jima Island, which probably represents the native genetic composition of the species here. We only found nine remnant trees classified with the purple genetic composition in the CCJ population. Although the expected heterozygosity level of the nine remnant trees was smaller than that of CCJ (0.628 vs. 0.674), their allelic richness exceeded that of CCJ (2.476 vs. 2.429). This might be because a limited number of maternal trees were used as the seed source for planting the trees in the northern part of Chichi-jima Island, or because of bi-parental inbreeding of parent trees of this population.
On Haha-jima Island, we observed a similar pattern of genetic diversity as for the CCJ population. Trees on Haha-jima Island were divided into four operational populations. One of the current authors (Y. Hoshi) remembered that the species had been grown in plantations on this Island. He recollected that part of KWK and LPS were planted by humans. Although the origin of the IGM population was not clear, the small, relatively uniform trees in IGM are distributed in quite a small area. This might indicate that, whilst these trees were not planted, they are the result of simultaneous regeneration from a few maternal trees, such as those of an original plantation. The amounts of within population genetic diversity of plantation trees (LPS) and IGM were lower than those of SKM and KWK. According to the model-based clustering, we identified a similar pattern of genetic composition between the planted populations (part of KWK and LPS) and IGM – the clusters represented by mostly green and pink colours (Fig. 4). This might also indicate an origin from a limited number of maternal trees or bi-parental inbreeding. In contrast, the rest of the trees in the KWK and SKM populations exhibited a complicated genetic composition, and the SKM population contained the highest amount of genetic diversity of all the operational populations, despite having the smallest sample size. In terms of genetic diversity and size of tree (data not shown), most trees in the SKM population and part of the KWK population are likely to be natural remnants.
The evidence of population bottlenecks and inbreeding
The individuals in the SKM population exhibited many components of different clusters (Fig. 4). As a result of its genetic characteristics, size of trees and vegetation pattern on the Islands, we deduced that this population is the closest to being natural. When we compared the genetic diversity level between the OTJ (almost equivalent to cluster 6) and SKM populations, the number of alleles, as well as the observed and expected heterozygosity was lower in the former. According to one theoretical study, the amount of reduction in heterozygosity depends not only on the "bottleneck" size but also on the rate of population growth after passing through the bottleneck, while the loss of alleles largely depends on the size of the "bottleneck" . If the SKM population produced many generations after the bottleneck event, the heterozygosity level would be reduced. However, we have assumed that the SKM population produced few generations after population decline, because M. boninensis is known to be an extremely long-lived woody species . Although the SKM population has maintained a high level of genetic diversity to date, the decline in genetic diversity in OTJ might be the result of bottleneck events, such as much severe logging, long-term population decline, or catastrophic population destruction.
When we estimated the pairwise kinships between individuals within a cluster and compared their average between clusters, the difference in average pairwise kinships between cluster 6 and all individuals was significant (Fig. 5). This meant that bi-parental inbreeding may have occurred during the bottleneck event in the OTJ population. However, the level of inbreeding was, supposedly, not strong, although expression of strong inbreeding depression for the species may have been masked by mass mortality of inbred individuals at the seedling stage. This speculation is also supported by the non-significant and negative fixation indices in cluster 6 and OTJ (Table 3). In contrast, potentially man-made populations, such as parts of KWK and LPS, and IGM, contained multiple clusters. However, most individuals within these populations were classified into two main clusters (green and pink), supported by the 90% confidence interval in the model-based clustering analysis (Fig. 4). Furthermore, we detected higher pairwise kinship values within clusters (green and pink) than those for all pairs of individuals (Fig. 5). This could be the consequence of a limited number of maternal trees producing the seeds from which the man-made populations were propagated. In contrast, although the mean of the pairwise kinship of SKM was somewhat high value, the mean of pairwise kinship of the putative natural populations and mixed background population, SKM and CCJ, is not significantly different from zero. This suggests that these populations might not have experienced intensive bi-parental inbreeding, as occurred in the putative man-made populations. If we collect seed from a few mother trees for the propagation of the next generation, even open pollination will potentially cause bi-parental inbreeding and a decline in the genetic diversity of the new generation. Therefore, it is very important to construct seed orchards using clones of trees from natural populations and/or to perform controlled crossing between them. Furthermore, some pairs in SKM population represented high level of pairwise kinships between them, which implied that we must consider selection of parental trees to construct seed orchards and combinations of parental trees for control crossing even if parental trees are only selected from the natural populations
Establishment of conservation units and a propagation program
Ex situ conservation would seem to be necessary to conserve M. boninensis, because it is difficult to obtain seeds free of hybridization with M. accidosa. The exception is the OTJ population . In addition, seedling establishment is extremely rare because of the competition with introduced alien species, such as Bischofia javanica . Therefore, selection of pure M. boninensis seedlings, controlled crossing and ex situ conservation are all necessary to ensure propagation of the species. To achieve this, we must consider Moritz's management units in order to select appropriate mother trees and we must undertake controlled crossing to obtain seed sources for propagation. In terms of genetic diversity, there was a large difference between the putative natural populations, with the exception of OTJ and IGM (south of CCJ, a part of KWK, most of SKM) and putative plantation trees. According to the model-based clustering analysis, planted individuals are classified into two main clusters, green and pink. This genetic differentiation, which is supported by the model-based clustering, should be used in selecting the MUs. We were able to identify at least seven main MUs for M. boninensis, which were OTJ, natural trees in CCJ, natural trees in KWK, SKM, man-made trees in CCJ and KWK and two major clusters including individuals in IGM and LPS (Green and Pink). Natural populations retained many elements of the different clusters. On the other hand, most of the plantation trees were simply classified into a single cluster. Hence, natural individuals were prime candidates for use as maternal trees in controlled crossing and ex situ conservation by grafting. According to analyses of pairwise FST and RST and the model-based clustering, we should avoid controlled crossing and instead should establish new seed orchards by open pollination of individuals from different clusters. To date, the OTJ population has been the only one to produce pure M. boninensis seeds. These seeds, therefore, have been the only ones used for propagation. There is a hybridization problem with the introduced species in the KWK and SKM populations. Because the MU of OTJ is definitely different from the other natural populations, we need to independently conduct controlled crosses or ex situ conservation (and subsequently develop seed orchards) within each cluster. Only then will propagated seedlings be available to use for propagation of the species in the forest according to the MUs identified in this study.