Microsatellite analysis of populations of the endangered tree Gomortega keule suggests pre-Columbian differentiation

ABSTRACT Temperate forests have been affected extensively by human activities, resulting in land cover changes and population fragmentation. However, these anthropogenic effects can be superimposed onto the natural history of species, making it difficult to determine which effect is more important for a particular species. Gomortega keule is an endangered tree that is found in one of the world’s biodiversity hotspots in central–south Chile. Human activities have significantly impacted on the original habitat in this region in recent years and are commonly considered to be the main cause of the scarcity of this species. However, aspects of the natural history of this evergreen tree may also help to explain its present-day genetic structure. In this study, we undertook microsatellite genotyping of the two southernmost populations of G. keule, which are 7.5 km apart and well isolated from other populations. We found that there was genetic differentiation between these populations, suggesting that they exhibited at least some differentiation before becoming isolated, most likely before human activities first impacted the region some two centuries ago. Molecular estimates of their divergence time supported a more ancient differentiation of the populations than would be explained by human activities alone. It is possible that their isolation may have followed the extinction of megafaunal seed dispersers around 12,000 years before present in this region, as indicated by fruit characteristics, the absence of recruitment by seedlings and the existence of clonal trees.


Introduction
Temperate forests have been significantly affected by human activities, particularly agriculture and forestry, resulting in land use changes and the fragmentation of existing species. However, the life history traits and natural history of species can also be key elements that affect population isolation and hence the population genetics of trees (Bacles & Jump 2011).
The coastal forests of central-south Chile lie within one of the 25 biodiversity hotspots on Earth (Myers et al. 2000). This region served as a refugium during the Quaternary glacial period, with many species surviving as a result of the temperatures being moderated by the ocean (Villagrán & Armesto 2005). Indeed, the distributions of several woody species are currently restricted to latitudes 36°-40°, which has been attributed to this natural history (Villagrán & Armesto 2005). However, over the last two centuries, this region has also been extensively affected by fires, timber exploitation, land clearance and the introduction of exotic forestry plantations (Armesto et al. 2010), and consequently the population fragmentation and reproductive isolation that is observed in many native woody species in these forests is commonly attributed to anthropogenic disturbance.
Gomortega keule (Mol.) Baillon is an endangered species that is found in central-south Chile. It exhibits a distribution pattern of small, isolated populations over a narrow geographical area near the coast ( Figure 1) and occurs in only one of the national protected areas in Chile. Martín & Sánchez (2000) provided details of 22 locations where the species exists, most of which have less than 100 individuals. Wood extraction activities and land clearance for agriculture occurred on a small scale before the arrival of Europeans in central Chile (Armesto et al. 2010), and indigenous people had some impact on the native forests, particularly in the main river basins. However, it is unlikely that G. keule populations were strongly affected by these activities because this evergreen tree mainly grows in mountainous areas, often in ravines and on pronounced slopes, where the impacts of indigenous human populations were restricted. Over the last two centuries, however, the areas in which G. keule occurs have been considerably affected by the logging of trees and especially the setting of fires for land clearance (Serra et al. 1986;Armesto et al. 2010), as well as the establishment of exotic timber plantations in more recent years, which has led to a reduction in native forest cover from 21% in 1975 to 7% in 2000 in the region where the species occurs (Armesto et al. 2010). Furthermore, individual G. keule trees have been found as stump shoots, indicating that the parental trees had been damaged or logged (Serra et al. 1986) and suggesting the loss of individuals in recent decades. In a genetic study of 11 populations using inter simple sequence repeat markers, García-Gonzáles et al. (2008) concluded that human activities caused the isolation of populations of G. keule. However, this assertion should be considered with caution as this is a long-lived species occurring in a habitat that has experienced only relatively recent anthropogenic disturbance.
The ecology of G. keule has some interesting features. The flowers are pollinated by syrphid flies (Lander et al. 2009). Although barochory has been suggested (Quesne & Stark 2006), there have been no studies on seed dispersion. Interestingly, the fruit are among the largest in the Chilean flora and are edible. Janzen & Martin (1982) proposed the idea of 'evolutionary anachronism' for several Central American species, such as Gymnocladus dioicus (Zaya & Howe 2009), and this could also apply to G. keule. If this was the case, the fruit of G. keule would have been dispersed by animals that once existed in central Chile, such as Palaeolama, Antifer, Equus, Milodontidae and Gomphoteriidae (Labarca & López 2006). Indeed, the edible pulp and very hard endocarp that encloses the seed ( Figure 2) appear to fit the megafaunal dispersion model (Guimarães et al. 2008). Today, the fruit of these trees rot on the forest floor in the absence of animals that are able to eat the complete fruit and transport the seed. In some regions, domestic animals (cows, pigs and goats) eat the flesh of the fruit but leave the seed. Although small seedlings can be found, they do not grow further, possibly due to foraging by domestic or wild (rabbits, hares) animals, or because of the ecological conditions that occur near settlements (humidity, light, litter alteration). Hence, with the extinction of the megafauna in South America around 12,000 years before present (yr BP) (Barnosky & Lindsey 2010), G. keule may have lost one class of its natural seed dispersers, causing it to become confined to areas that are close to the stands and populations where it remains today.
In this study, we investigated whether the isolation of populations of G. keule occurred much longer ago than could be explained by human activities in the last 200 years. To do this, we evaluated the genetic differentiation of two spatially close but distinct populations based on the notion that if these populations showed significant genetic differentiation, a process of genetic differentiation must have already been underway before any superimposed human activities in recent centuries. This could argue for a more complex situation than suggested by García-Gonzáles et al. (2008), with the megafaunal extinction around 12,000 yr BP potentially contributing to this process.
The two southernmost populations of G. keule were selected for this study, which are only 7.5 km apart and in the same valley, but nearly 70 km from the next nearest population. Importantly, there is a clear geographical separation between these populations today, with no G. keule trees between. Although most populations of G. keule are generally of a similar size to the study populations (60-80 individuals over < 1 hectare), we chose these particular populations because they occur in an area with a more limited degree of recent anthropogenic influence, being relatively distant from current human settlements. By contrast, populations near big cities (e.g. Concepción) or agricultural areas (northern area of distribution) are likely to have been affected for a longer period and more deeply by human activities.

Sample populations
Plant materials were obtained from the Maule and Bío-Bío regions of Chile (Table 1)   were growing from old stumps, whereas the remainder were young trees or shoots (trunk diameters < 10 cm) growing under the canopy in a plantation of Pinus radiata. These young individuals were derived from the stumps of old trees that had been cut before or at the time of plantation establishment and so would be expected to have the same genotypes as the original logged trees. No seedlings were seen during collection. By contrast, population B (n = 63) mainly comprised very old trees, some of which were the oldest representatives of this species we found (trunk diameters c. 2 m). As for population A, no seedlings were observed in association with this population. Samples were collected from all of the G. keule trees that could be found in each of these populations, with only one sample being collected from each group of shoots or trunks derived from an individual tree stump in population A. In addition, 14 individuals were sampled from a population in the northern area of the species' distribution (190 km away from populations A and B) as an outgroup (group C).

Sample collection and DNA extraction
Young leaves were collected from each tree or alternatively the cambium was sampled if leaves could not be found for a particular tree. The collected tissues were dried in sealable plastic jars containing self-indicating silica gel crystals, which were renewed as required, and stored at − 20°C. DNA extraction followed Lander et al. (2007), whereby 5 mg diethyldithiocarbamic sodium salt, 10 mg PVP-40000 and 5 µL β-mercaptoethanol were added per 1 mL of cetyltrimethylammonium bromide solution that had been pre-heated to 65°C. This solution (500 µL) was then added to each sample for DNA extraction.

Microsatellite genotyping
Each sample was genotyped using seven microsatellite primer pairs. Five of these (Gk-1, Gk-30, Gk-31, Gk-35 and Gk-39) were developed by Lander et al. (2007), and the remaining two (CS2 and CS8) were developed by S. Mayes following the method of Haddrill et al. (2002). The procedure to develop primers CS2 and CS8 involved the digestion and polymerase chain reaction (PCR) amplification of genomic DNA, hybridisation to filters with artificial simple sequence repeat oligonucleotides, elution and amplification. Rather than using cloning, a proportion of the enriched library was pyrosequenced using a 1/16 run (non-titanium reagents; 454 Life Sciences).
To visualise the DNA, 1% agarose gels were prepared in 1× TAE buffer and ethidium bromide was added to a final concentration of 0.5 µg/mL before casting the gels. The PCR products with fluorescently labelled primers were then visualised on 2% agarose gels in 0.5× TBE buffer that were stained with ethidium bromide to a final concentration of 0.1 µg/mL.
The forward primers were directly labelled with D2, D3 or D4 fluorescent dye (WellRED; Sigma Aldrich). The optimum annealling temperature for PCR for each labelled primer was then determined in a Px2 Thermal Cycler using an annealing temperature gradient from 45°C to 60°C across the block. PCR for each sample was performed using a 20 µL reaction volume containing 2 µL template DNA, 2.5 µL buffer, 0.2 µL deoxynucleotide, 0.05 µL bovine serum albumin (10 µg/µL), 0.02 µL of each forward (labelled) and reverse primer (100 µM), 15 µL autoclaved H 2 O and 0.2 µL Taq polymerase. Master mixes for multiple reactions were made wherever possible. The reaction conditions were 94°C for 3 min, 30 cycles of 94°C for 1 min, 1 min at the specific annealling temperature for each primer, 72°C for 1 min, and final extension at 72°C for 20 min. The annealing temperatures were 60.9°C (Gk-1), 65.0°C (Gk-30, Gk-31 and Gk-35), 56.8°C (Gk-39) and 52.7°C (Gk-44, CS2 and CS8). A total of 152 genotypes were analysed for seven loci using capillary electrophoresis (CEQ 8000 Genetic Analysis System; Beckman Coulter). The PCR products were pooled according to their different fluorescent tags, with a maximum of three different coloured samples being analysed per well.

Microsatellite data analysis
The peaks that were produced by capillary electrophoresis were thoroughly examined for each locus and assigned to classes according to Amos et al. (2007) using a software tool developed by the same authors (FLEXIBIN).
For each sample, matrices were generated using allele size to score co-dominant markers at each locus. Polymorphic loci (Laurentin 2009) were used to generate unbiased estimates of heterozygosity (Nei 1978) and investigate population structure.

Population genetic analysis
Samples that were considered to have originated from the same individual were excluded from the population analysis. This included any trees that were both < 10 m apart and had identical genotypes in the binary matrix. In total, 19 distinct genotypes were identified among 42 samples, resulting in 23 samples (42 minus 19) being excluded. Of the remaining 129 samples that were included in the population analysis, 62 were from population A, 53 from population B and 14 from group C, all of which were assessed for all seven loci.
The number of alleles per population and per locus, and the allelic richness and private allelic richness per population were calculated with HP-Rare (Kalinowski 2005), using the rarefaction method to account for unequal sample sizes. The observed (H O ) and expected (H E ) heterozygosity for each locus and population were then computed with Arlequin 3.5 (Excoffier et al. 2005) using formulae from Nei (1987), and values of H O and H E for the combined populations were calculated with the software GDA using formulae from Nei (1978).
According to Laurentin (2009), co-dominant markers can be used to assess the diversity between groups with methodologies based on allele frequencies, such as Wright's F-statistics and Nei's parameters. Therefore, the inbreeding coefficient (F IS ) for each locus and population was computed using FSTAT (Goudet 1995) according to Weir & Cockerham (1984). The pairwise fixation index (Rho ST or ρ ST ) was calculated for each population using Genepop 4.0.5.3 (Rousset 2008), following which a second version of the fixation index (F ST ) was calculated under the stepwise mutation model using Arlequin.
Exact tests for deviations from Hardy-Weinberg equilibrium were performed with Arlequin and Genepop using the Markov chain method (forecast chain length: 1,000,000; dememorisation steps: 100,000). Linkage disequilibrium was tested separately for each population using Arlequin and a Bonferroni correction was applied to the P values. The frequency of null alleles was calculated assuming simultaneous inbreeding for each locus and population with the software INEst (Chybicky & Burczyk 2009) using either the population or individual inbreeding model (1,000,000 iterations).
The similarity index of Nei & Li (1979) was used to evaluate the genetic relationship among individuals, and cluster analysis (UPGMA) and principal coordinate analysis (PCoA) with Euclidean distances were used to visualise these relationships, both of which are appropriate for this type of study (Laurentin 2009). PCoA makes no assumptions about the distribution of the data or population genetics and the Euclidean distance does not consider the common absence of an allele as a shared characteristic (Kloda et al. 2008).
Cluster analysis and PCoA were performed using the MVSP software (Multivariate Statistical Package version 3.13; Kovach Computing Services, Anglesey, UK). A scatter plot was generated with the two largest eigenvalues along the first two axes and a dendrogram was produced with UPGMA using Nei & Li's index (Nei & Li 1979) as a measure of similarity. All 152 samples, which included those samples that were considered to have originated from the same individual, were included in the PCoA and cluster analysis.
Population structure was assessed using a Bayesian cluster analysis with the software Structure 2.3.4 (Pritchard et al. 2000). Analyses were performed using the admixture model with independent allele frequencies (20 runs), with burn-in and simulation lengths of 300,000 and 1,000,000 iterations, respectively. The optimal value for K was estimated by calculating the statistic ΔK (Evanno et al. 2005). These procedures were initially followed for all samples and then for all samples except group C in an independent set of runs.
The divergence time of populations A and B was estimated with the software IMa2 (Hey & Nielsen 2007), which estimates population-genetic parameters by calculating posterior probabilities in a Bayesian sampling framework. A total of 100,000 genealogies were sampled to estimate the joint posterior probability distribution of the divergence time parameter. Generation times of 25, 100 and 175 years were used, as suggested by Jones et al. (2013) for long-lived trees, and a mutation rate of 0.000316 was used, as proposed by Tamaki et al. (2016) for trees. More than 100 runs were made to adjust the priors (values of q = 20, m = 500 and t = 2), allowing for burn-in periods of 127-340 million steps. For each generation time, 13 final estimates were made using different seed numbers, and the mean values of the divergence time (yr BP) and the lower and upper bounds of the estimated 95% highest posterior density intervals were calculated.

Microsatellite genotyping
The number of alleles that was found for each locus is presented in Table 2. All seven markers were polymorphic (major allele frequency < 0.95) according to Laurentin (2009). The allelic richness and private allelic richness (after rarefaction) were lower in populations A (2.19 and 0.34, respectively) and B (2.13 and 0.24) than in group C (2.63 and 1.61) ( Table 2). Private alleles were found for all loci and their occurrence was associated with the presence of rare alleles, as many of them showed low frequencies. Although sample size can affect the number of alleles when a sampling approach is used, to the best of our knowledge populations A and B represented the entire populations.
The level of heterozygosity that was identified in G. keule (Table 3) is similar to or lower than that reported for other endangered trees (Tamaki et al. 2008;Finger et al. 2011;Shepherd & Perrie 2011). The largest difference between H O and H E (> 0.2) occurred at locus Gk-31 for population A and locus Gk-35 for group C (Table 3), with exact tests for deviation from Hardy-Weinberg equilibrium being significant (Bonferroni corrected P < 0.002) at each of these loci. The inbreeding coefficient for each locus and population ranged from − 0.233 to 0.874 (Table 4), with the highest values (F IS > 0.8) being observed for locus Gk-31 in population A and locus Gk-35 in group C.
The pairwise population fixation index, ρ ST , was 0.045 for A-B, 0.882 for A-C and 0.882 for B-C, and F ST was 0.061, 0.407 and 0.382 for the same pairwise comparisons, respectively. Significant linkage disequilibrium after Bonferroni correction was present in one pairwise comparison for population A and one for population B. The estimated frequency of null alleles was significantly different from zero (P < 0.001) for locus Gk-31 in population A using the population inbreeding model and the individual inbreeding model with INEst (Chybicky & Burczyk 2009). Similarly, the frequency was also significantly different from zero for locus Gk-35 in group C (P < 0.001 using the population inbreeding model and P < 0.01 using the individual inbreeding model). Although the presence of null alleles could have some influence on the accuracy of the statistical analyses, it was not considered likely that it had a major influence on the overall results.

Population structure
There was some genetic differentiation among the populations studied. Moderate genetic differentiation (F ST = 0.051-0.150; Yeh 2000) was found between populations A and B   (Figure 3) showed that there was some differentiation between populations A and B, which is consistent with their geographical location: these populations are only 7.5 km apart ( Figure 1) and each occupies < 1 hectare.
The dendrogram (Figure 4) supported the PCoA, with all samples except those from group C falling into a separate cluster, and samples from populations A and B generally being grouped into two partially overlapping clusters.
The Bayesian analysis with the software Structure ( Figure 5) appeared to confirm the groups that were identified by the PCoA and the dendrogram analysis. The graphical output of Structure for K = 2 indicated some degree of differentiation between populations A and B. When the analysis included group C, the highest ΔK corresponded to two clusters, placing populations A and B in the same cluster. However, the output for K = 3 also  Figure 3. Principal coordinate analysis for three populations of Gomortega keule using microsatellites for seven loci. Populations A (n = 75) and B (n = 63) are the southernmost populations of this species, while group C (n = 14) comprises individuals found in the northern area of the species' natural distribution. The graph was generated using MVSP 3.13 with the first two axes accounting for 30.5% of the variance (Axis 1, 19.4%; Axis 2, 11.1%).
suggested some differentiation between populations A and B, which is biologically meaningful.
The estimated divergence times for populations A and B and the 95% highest posterior density intervals were 890 (395-2908) yr BP for a generation time of 25 years, 3956 yr BP 263) for 100 years and 6784 yr BP (3185-20,951) for 175 years.

Megafaunal dispersal and natural history of G. keule
It has been proposed that the seeds of G. keule require megafaunal dispersal (Muñoz-Concha & Davey 2011), and so it is possible that populations A and B have become increasingly isolated following the extinction of megafauna from Chile, leading to their current level of differentiation. This idea that population differentiation was already underway before human activities affected these populations is supported by a number of observations from the present study. The data suggest that populations A and B, which are geographically very close, have a moderate level of genetic differentiation but are clearly more closely related to each other than to group C, which could argue for the previous existence of a more continuous population that is now fragmented. Populations A and B occur in the same valley, so fragmentation as a result of physical barriers alone can be ruled out. The minimum divergence times calculated in the present study are greater than 200 yr BP (minimum 395 yr BP with 95% confidence for a 25-year generation time), which would argue for pre-existing differentiation before any superimposed effect of human activities.
The PCoA showed a clearer separation of the populations than was found by García-Gonzáles et al. (2008), who used inter simple sequence repeat to analyse the genetic structure of 11 populations of G. keule, but only sampled some of the individuals from populations A and B (29 and 7, respectively). In addition to the larger number of individuals that were sampled in the present study, the use of co-dominant microsatellite markers improved the resolution of a number of the analyses and also allowed heterozygositybased indices to be calculated.
Differentiation between the examined populations of G. keule is demonstrated by the fixation index values, which lie within the range of those assessed with microsatellites for other endangered tree species, such as Magnolia stellata (Tamaki et al. 2008), Dalbergia monticola (Andrianoelina et al. 2009), Medusagyne oppositifolia (Finger et al. 2011) and Pseudopanax ferox (Shepherd & Perrie 2011). Furthermore, similar levels of genetic differentiation have also been observed among isolated populations of Myrtus nivellei, a relict species that is currently experiencing geographic range contraction (Migliore et al. 2013). It has previously been argued that clonality plays a role in maintaining unaltered genotypes over a long period of time (even thousands of years) in other woody species with relict and isolated populations, as discussed by Backs et al. (2015), Bradbury et al. (2016) and Migliore et al. (2013). Although there have been no reported studies on the role of clonality as a mechanism to maintain heterozygosity and prevent inbreeding, and no current evidence for the age of clones or trees of G. keule, the ability to maintain a population through off-shoots could be a key factor that explains the current level of variation in these populations and so warrants further investigation.
The generation times of trees can be greatly extended through vegetative (clonal) reproduction. Indeed, clones of relict tree species have been dated back to 3000-11,000 yr BP in Eucalyptus absita (Bradbury et al. 2016) and over 13,000 yr BP in Quercus palmeri (May et al. 2009). Therefore, as clones of G. keule were observed in the present study and by Lander et al. (2010), it is possible that this species has an extended generation time, making a generation time of 25 years unlikely.
The estimated time for the divergence of populations A and B is over 800 yr BP, placing the population differentiation process well before two centuries ago. Therefore, the isolation of these two populations of G. keule was probably well progressed before extensive anthropogenic disturbances first impacted the area around 200 yr BP (Armesto et al. 2010), indicating that population differentiation may have started in glacial times, well before human-mediated fragmentation of the area. This situation was also discussed for the endemic tree Nothofagus alessandrii by Torres-Díaz et al. (2007), who concluded that there was probably no gene flow between populations of this species before they were impacted by relatively recent human activities.
A long regeneration cycle suggests that the study populations have been isolated for a long time. The rare occurrence or complete absence of successful sexual regeneration as evidenced by a lack of seedling establishment in the G. keule populations sampled has also been observed in other relict tree species such as Myrtus nivellei (Migliore et al. 2013), Quercus palmeri (May et al. 2009), Quercus hinckleyi (Backs et al. 2015) and Eucalyptus absita (Bradbury et al. 2016). The failure of seedlings to establish and develop into new trees may indicate that the effective generation time is very long, but could also reflect the cessation of seedling recruitment due to the more recent introduction of grazing animals such as hares.
The fruit of G. keule matches the definition of megafaunal fruits that was introduced by Guimarães et al. (2008), being 4-7 cm in diameter and yellow, with a small number of large seeds (Figure 2). The species also matches the predictions of those authors, showing clumped spatial patterns, reduced geographical ranges and high levels of among-population structuring. Other traits that suggest that G. keule may have megafaunal seed dispersal characteristics include the lack of modern dispersal agents, large and well-protected seed, edible fruit pulp and a strong vegetative propagation ability, as previously discussed for Gymnocladus dioicus by Zaya & Howe (2009). In addition, the present-day rare occurrence or complete absence of recruitment by seedlings and the isolated distribution of individuals and populations further support the idea of megafaunal dispersal of G. keule.
A number of questions still need to be answered to better understand the population dynamics of G. keule. These include the juvenility and longevity of individuals, the ecological conditions that are required for seedling survival, and the minimum viable population size.

Sampling strategy
Some of the samples that were found to be genetically identical were clearly shoots from an old tree that had since disappeared, as reported previously (Martín & Donoso 1996). However, some trees of G. keule that were growing < 2 m apart were found to be genetically different, allowing us to reject the assumption that close physical proximity indicates that individuals are clonally related. The intricate way in which the trees occupy space on the forest floor is also well illustrated by the case where a root (10 cm diameter) that was sampled from population B was found to be different from three individuals growing < 7 m away but identical to a large tree growing 15 m away. Therefore, developing an understanding of the spatial relationships within the population will be an important aspect of any future conservation planning.
The findings of the present study support the idea that vegetative propagation (i.e. regeneration by sprouting) is currently important in natural populations of G. keule, as stated by Martín & Donoso (1996). Some individuals that were physically close to each other had the same genetic profile and so were probably produced from a single tree. Although we did not undertake extensive sampling of shoots that were very close together, the observation of shoots with the same genetic profile may indicate that clonality has played an important role in the recent survival of G. keule trees.
Clonal production is a relatively common event in trees with reduced populations such as Eucalyptus absita (Bradbury et al. 2016) and Myrtus nivellei (Migliore et al. 2013), with some remarkably old individuals being found, as seen in Quercus palmeri (May et al. 2009) and Quercus hinckleyi (Backs et al. 2015). Since clonality effectively extends the reproductive cycle by maintaining the original genetic combinations, the estimated divergence times for G. keule may be underestimates.

Implications for conservation
Although the occurrence of small and increasingly isolated populations of G. keule seems to have preceded human history in the area, anthropogenic effects remain very important for the future conservation status of this species. It is apparent that there has been extensive destruction of the habitat, populations and trees of this species, as evidenced by the fact that most individuals are recently re-grown trees derived from old tree stumps with signs of destruction by humans (logging, fires).
Long-term isolation and genetic differentiation among populations may have important implications for conservation management strategies for G. keule and other species that occur in disjunct populations, particularly where the natural dispersal agents are no longer present. Conservation efforts should first be directed towards the protection of remnant populations and individuals, including the exclusion of cattle and hares from protected zones, and the careful management of light through canopy maintenance, particularly for individuals that are re-growing in current or former forestry plantations. Given the very limited levels of seedling survival at present, it will be difficult to maintain high genetic diversity through natural seed production. Therefore, seeds or vegetative material should be sampled from as many populations as possible to maximise the amount of genetic diversity that is captured for immediate ex situ conservation. However, the risks of outbreeding depression should also be evaluated.
The level of population differentiation that was observed in G. keule may offer a model for the conservation of other Chilean species for which the dispersal agents may be extinct, particularly Pitavia punctata and Jubaea chilensis, with a need to sample from all extreme populations of each of these. The southern populations of G. keule are very important from an agricultural and forestry perspective, as each isolated population contains valuable genetic resources for the future domestication and genetic breeding of the species, and may contain genetic information that is not found among other extant populations.