Abstract
Understanding the impact of management on the dispersal potential of forest tree species is pivotal in the context of global change, given the implications of gene flow on species evolution. We aimed to determine the effect of logging on gene flow distances in two Nothofagus species from temperate Patagonian forests having high ecological relevance and wood quality. Therefore, a total of 778 individuals (mature trees and saplings) of Nothofagus alpina and N. obliqua, from a single plot managed 20 years ago (2.85 hectares), were mapped and genotyped at polymorphic nuclear microsatellite loci. Historical estimates of gene dispersal distance (based on fine-scale spatial genetic structure) and contemporary estimates of seed and pollen dispersal (based on spatially explicit mating models) were obtained. The results indicated restricted gene flow (gene distance ≤ 45 m, both pollen and seed), no selfing and significant seed and pollen immigration from trees located outside the studied plot but in the close surrounding area. The size of trees (diameter at breast height and height) was significantly associated with female and/or male fertility. The significant fine-scale spatial genetic structure was consistent with the restricted seed and pollen dispersal. Moreover, both estimates of gene dispersal (historical and contemporary) gave congruent results. This suggests that the recent history of logging within the study area has not significantly influenced on patterns of gene flow, which can be explained by the silviculture applied to the stand. The residual tree density maintained species composition, and the homogeneous spatial distribution of trees allowed the maintenance of gene dispersal. The short dispersal distance estimated for these two species has several implications both for understanding the evolution of the species and for defining management, conservation and restoration actions. Future replication of this study in other Nothofagus Patagonian forests would be helpful to validate our conclusions.
Introduction
In the current context of rapid environmental changes and increasing logging activities, understanding the dispersal potential (i.e. pollen and seed gene flow) of forest species is critical for effective management and conservation planning (Frankham et al., 2012; Monthe et al., 2017). Pollen and seed dispersals are essential processes with far-reaching consequences for reproduction, population and community dynamics, neutral and adaptive evolution and, ultimately, population and species persistence (Robledo-Arnuncio et al., 2014). The assessment of pollen flow is also useful to predict the impact of landscape alterations on reproduction, thus allowing the design of proper conservation strategies (Craft and Ashley, 2010; Beghè et al., 2016). Therefore, gene flow, together with the main characteristics of the mating system (e.g. outcrossing and selfing rates, inbreeding), is considered one of the most important determinants of the genetic structure of plant populations and crucial to inform appropriate strategies for stand management (Sork et al., 2002; Marchelli et al., 2012; Monthe et al., 2017).
Silvicultural selection represents an anthropogenic force through which sometimes large numbers of trees, and the genes they contain, are removed from a natural system (Schaberg et al., 2008). The cleaning of young stands of inferior phenotypes or the removal of high-value trees (that likely contribute more to reproduction than smaller individuals) may impact genetic diversity in mature tree populations if the phenotype is correlated to a particular genotype, and also in the next generation, if the forest is regenerated naturally (Finkeldey and Ziehe, 2004; Paffetti et al., 2012). This decrease in tree density reduces effective population size, modifies the spatial distribution of the reproductive trees and increases distance between conspecifics, probably affecting the reproductive biology of remaining tree populations (Lefèvre, 2004; Cloutier et al., 2007; Monthe et al., 2017). However, genetic variation in natural regeneration also depends on the amount of gene flow from surrounding forests, and therefore the effects of reduced population size on levels of genetic variation may be balanced by gene flow for species with efficient dispersal means for pollen and diaspores (Finkeldey and Ziehe, 2004). Moreover, wind patterns might be modified since open forests often promote longer-range dispersal, indicating that patterns of gene flow are complex and not necessarily predictable (e.g. Hardy, 2009). Hence, the study of dispersal distances, as well as the relationship between fecundity and tree size, is an important parameter because the possible changes in gene flow patterns due to silvicultural treatments may modify the amount of genetic diversity and the spatial genetic structure within a stand (Rajora et al., 2000; Takahashi et al., 2000; Thomson et al., 2011; Piotti et al., 2012). However, the long lifespan and generation cycle of most forest tree species pose particular difficulties with regard to the monitoring of genetic structures, and the consequences of management on adaptive processes may became noticeable after several generations (Finkeldey and Ziehe, 2004).
To study gene dispersal patterns within populations, two main categories of molecular-based genetic methods can be distinguished, indirect and direct methods. Indirect genetic methods that rely on the fine-scale spatial genetic structure (FSGS) of populations provide an indication of the mean parent–offspring distance (at migration–drift equilibrium), averaged over several generations (i.e. historical gene flow patterns), without distinguishing the respective role of pollen and seed dispersal (Vekemans and Hardy, 2004). Some indirect genetic methods could also provide contemporary gene flow (seed and/or pollen movement) estimates (Smouse et al., 2001; Robledo-Arnuncio et al., 2006; Grivet et al., 2009). On the other hand, direct genetic methods provide estimates of contemporary gene flow (seed and pollen movement), by modelling forward dispersal kernels (i.e. the probability density function describing the distribution of the post-dispersal locations relatively to the source point (Nathan et al., 2012)) using parentage probabilities of offspring under a spatially explicit neighbourhood model (Chybicki and Burczyk, 2010). Model-based approaches also provide inference on how reproductive success can be explained by attributes of individuals (e.g. their size) through the analysis of selection gradients (Chybicki, 2018). Comparing historical and contemporary gene dispersal can provide insights on the demographic history of the population. This is because similar estimates are expected if the population and gene dispersal processes remained stable over several generations, while recent disturbances should affect mostly contemporary gene dispersal estimates (Oddou-Muratorio and Klein, 2008; Monthe et al., 2017).
Nothofagus alpina (Poepp. et Endl.) Oerst. (N. nervosa, raulí) and Nothofagus obliqua (Mirb.) Oerst. (roble pellín) are wind-pollinated, outcrossing and anemochorous tree species that produce high-quality timber in cold-temperate regions of South America. Natural hybridization between N. obliqua and N. alpina is known to occur (Donoso et al., 2004), especially within sympatric areas where both species co-occur (Gallo et al., 1997; El Mujtar et al., 2017). Most of the natural distribution of these mixed Nothofagus forests is located within the Lanín National Park Administration area (Sabatier et al., 2011), where different protection levels exist. Silvicultural management has been carried out within this National Park (40° 90´ S and 71° 21´ W) since the late 1980s, mainly by the application of the shelterwood system. This system consists of successive regeneration fellings which retain partial forest cover until the regeneration phase is complete (approximately 20 years). These Nothofagus forests seed cuts have been applied in mature stands to leave a homogeneous canopy cover of around 30–40 per cent to produce seeds for natural regeneration and to provide shelter for the young seedlings (Chauchard et al., 2003). Final cuts, the last cut of the shelterwood system, has not yet been conducted in the management areas of LNR since the regeneration objective (2500 saplings/ha of more than 2 meters height) has not yet been achieved.
Despite the economic and ecological importance of these forests, little is known about the processes determining their ecological and evolutionary dynamics. In particular, for gene flow, only contemporary estimates of pollen dispersal in a pure forest of N. alpina has been studied suggesting that pollination is mostly local (<35 m, Marchelli et al., 2012). To date, however, no study has been conducted to estimate both, historical and contemporary, patterns of gene flow in these two species. Comparative genetic diversity analyses of mature trees and natural regeneration in a mixed forest of N. alpina and N. obliqua showed no impact of management at the species level (Sola et al., 2016). This result suggests that the potential negative genetic impact (e.g. high inbreeding, low genetic diversity) of increasing interindividual distances resulting from harvesting could be counterbalanced by high immigration rates from unmanaged surrounding areas. However, fine-scale spatial genetic structure has not been studied for these Nothofagus species until now. The knowledge about historical and contemporary gene flow becomes even more relevant now given the future suitability predictions indicating that the distribution range of N. alpina could be drastically reduced, while only marginal populations at the edge of the distribution range of N. obliqua could be affected (Marchelli et al., 2017).
Therefore, our main objectives were (1) to determine whether silvicultural management (by seed cut) modifies the dispersal potential of two Nothofagus species by comparing historical (based on FSGS) and contemporary (based on spatially explicit models) estimates of gene dispersal distance in N. alpina and N. obliqua and (2) to test whether reproductive success is related to the diameter and/or height of individual trees. Confronting contemporary and historical dispersal estimates will allow us to investigate how pollen- and seed-mediated gene flow, respectively, shapes patterns of FSGS in N. alpina and N. obliqua. Based on the previous study of pollen flow in N. alpina (Marchelli et al., 2012), we expect rather restricted estimates of dispersal abilities in both Nothofagus species. Also, considering that the seed cut happened 20 years ago, and after that there have been no further interventions, we expect discrepancies between contemporary (established regeneration after the management) and historical (several generations ago) gene flow estimates.
Materials and methods
Study area, genetic data and tree traits
A mixed stand composed entirely of Nothofagus dombeyi, N. alpina and N. obliqua (20, 44 and 36 per cent, respectively, based on tree density; Sola et al., 2015), located on Cerro Quilanlahue (40° 8′ S, 71° 28′ W), at 930 m asl, within Lanín National Reserve (Neuquén, Argentina) was selected. Environmental, topographic and edaphic characteristics of the study plot have been previously described by Sola et al. (2016).
In 1993, silvicultural management of this stand was carried out, implementing a seed cut to achieve a canopy cover of around 40 per cent. This cut was not selective and aimed to maintain the original relative species composition, a homogenous spatial distribution and size relationships of mature trees (Supplementary Data Fig. S1, Fig. S2 and Table S1). Before and after this intervention, the site had no management. To evaluate gene flow patterns of N. alpina and N. obliqua, we used a 2.85-hectare rectangular plot (285 m x 100 m) within the stand. This plot was established in 2009 (ECOS, 2008) and has previously been used to characterize the impact of silvicultural management on the genetic diversity and species composition of a mixed Nothofagus forest (Sola et al., 2015, 2016). The genetic (nuclear microsatellite or simple sequence repeat, SSR), spatial (x and y coordinates) and dendrometric information (number of stems per hectare, basal area per hectare, diameter and tree heights and ages) of all mature trees and post-harvest regeneration from both species were therefore available (Sola et al., 2015, 2016). The mean and standard error of the age (based on wood ring counts, Sola et al., 2015) for N. alpina mature trees was 136 ± 86 years, whereas for N. obliqua was 187 ± 81 years. Post-harvest regeneration (individuals with root collar diameter < 10 cm and height < 4 m) was sampled from 14 transects (5 m width, 100 m length) systematically distributed every 20 m. The spatial coordinates of the regeneration corresponded to a systematic sampling grid every 10 m along transects, i.e. the regeneration near each grid point share the same x, y coordinates. The mean age of regeneration was 8 ± 3 and 7 ± 2 years for N. alpina and N. obliqua, respectively. Sola et al. (2015) determined that in the studied plot, (a) the forest was at a mature stage, and (b) mature trees were dominated by N. alpina, while this species represented a minority among the post-harvest regeneration.
For gene flow analysis, we used 12 (N. alpina) and 11 (N. obliqua) polymorphic nuclear microsatellite (SSR) loci from the 15 SSRs previously genotyped (monomorphic loci were excluded) on both species (Supplementary Data Table S2, Sola et al., 2016). Considering that interspecific hybridization can influence dispersal estimations (Oddou-Muratorio et al., 2001), we excluded hybrid individuals from adult trees and regeneration based on our previous study on molecular identification and classification of hybrids (El Mujtar et al., 2017). We also excluded regeneration near plot edges (~20 m), i.e. a subsample of the central regeneration was used (68 out of 158 and 517 out of 813 for N. alpina and N. obliqua, respectively, Fig. 1), to avoid the overestimation of migration rates. The probability of excluding a putative parent pair (P3) estimates for each locus and for increasing combinations of loci was calculated, following the equation of Jamieson and Taylor (1997, Equation 3a), using Genalex (Peakall and Smouse, 2006).
Mapping of sampled individuals of Nothofagus species at the studied plot. Squares are mature trees of both species, and the central rectangle contains the subsampled central regeneration (empty circles). Circles of different sizes represent differences in regeneration density (larger circles represents more sapling captured). N = 109 and 84 mature trees and 68 and 517 seedlings of N. alpina and N. obliqua, respectively.
We used INEST 2.0 to jointly estimate unbiased multilocus inbreeding coefficient within the population (associated with the magnitude of FSGS), frequency of null alleles and presence of genotyping errors (Chybicki and Burczyk, 2009). We applied the individual inbreeding model (Bayesian approach) which has better statistical properties than the population inbreeding model (maximum likelihood approach). All analyses were performed under five models (nfb, nb, nf, fb, n) considering different combinations of three parameters: n (null alleles), f (inbreeding coefficient), b (genotyping errors) with Markov chain Monte Carlo of 500 000 cycles, burn-in period of 50 000 and thinning of 500. Then, the deviance information criterion (DIC) was used to determine which model was the best fit to the data. All mature trees and seedlings were used for estimating inbreeding coefficient, null allele frequency and presence of genotyping errors.
Indirect estimates of gene flow (based on fine-scale spatial genetic structure of populations)
Fine-scale spatial genetic structure was assessed for all mature trees in the population (N. alpina = 109 and N. obliqua = 84) following the procedure described in Vekemans and Hardy (2004). Pairwise kinship coefficients (Fij) were estimated between individuals using the Nason’s estimator of kinship coefficient (Loiselle et al., 1995), chosen due to its statistically robust properties (Vekemans and Hardy, 2004), using the software SPAGEDI 1.4c (Hardy and Vekemans, 2002). Estimations were performed considering 10 distance classes for which maximal distances are defined in such a way that the number of pairwise comparisons within each distance class is approximately constant (Hardy and Vekemans, 2002). As the number of pairwise comparisons for 10 distance classes were different between species (see Results), we also performed the analysis considering 17 and 5 distance classes for N. alpina and N. obliqua allowing to have similar number of pairwise comparisons in each class for both species. To analyse the FSGS, Fij values were regressed on the logarithm of the spatial distances between individuals, ln (dij), giving the regression slope b. Standard errors (SE) around mean Fij values within each distance class were obtained through a jackknife procedure consisting of deleting each locus one at a time. Furthermore, to illustrate the FSGS graphically, Fij values were averaged over the set of 10 distance classes. Statistical significance of Fij and b was determined under a 95 per cent confidence interval of Fij created by 10 000 permutations of individuals among distance classes. To quantify the strength of FSGS, the Sp statistic (Vekemans and Hardy, 2004) was computed as Sp = −b/(1–F1) where F1 is the mean pairwise kinship coefficient between individuals of the first distance class. Following Hardy et al. (2006), we present the standard error of b (calculated by jackknifing over loci) as an estimate of the variability of Sp. Finally, assuming drift–dispersal equilibrium, we estimated a characteristic gene dispersal distance, σg (σ2g is half the mean-squared parent–offspring distance), and the neighbourhood size, Nb = 4πde σ2g where de is an effective population density of reproductive individuals. For this purpose, we used an iterative procedure based on the restricted regression slope (br) of Fij on ln (dij) within a limited distance range σg < dij < 20 σg and the theoretical expectation that Nb = −(1–F1)/br under isolation by distance, following Vekemans and Hardy (2004). The effective population density de should be a fraction of dobs dependent on the variance in lifetime reproductive success among adults (Hardy et al., 2006). The observed density (dobs), which is the density of adult individuals with diameter at breast height > 10 cm per m2 in the plot, was 0.00397 and 0.00333 for N. alpina and N. obliqua, respectively. However, this density was measured 20 years after the seed cut. Accordingly, and considering that during the seed cut about half of the individuals of each species were removed (Sola et al., 2015), estimates of Nb and σg were performed using five effective densities (dobs*2, dobs*1.5, dobs, dobs/1.5, dobs/2). For a fixed de, lower and upper bounds for the 95 per cent confidence interval (CI) of Nb were computed as (F1–1)/(br + 2SEb) and (F1–1)/(br—2SEb), respectively, SEb being the standard error of the br estimates obtained by jackknifing over loci (when br < 2SEb, the upper bound was reported as infinite, ∞). The 95 per cent CI of σg was obtained similarly as √ Nb/(4πde) using the upper and lower Nb bounds (Hardy et al., 2006).
Direct estimates of gene flow (based on spatially explicit neighbourhood model)
The estimates of the seed and pollen dispersal kernels were based on the genotypes and spatial locations of all the 109 and 84 mature trees and the 68 and 517 seedlings of N. alpina and N. obliqua, respectively. We applied the seedling neighbourhood model implemented in NMπ software (Chybicki, 2018). This is the probability model of the genealogy of a sample of offspring given their genotypes and spatial positions, as well as the information about candidate parents (i.e. genotypes, spatial positions and, if available, phenotypes) (Oddou-Muratorio et al., 2005; Burczyk et al., 2006). We estimated ms, the seed migration rate; mp, the pollen migration rate; s, the selfing rate; δs, the average seed dispersal distance; δp, the average pollen dispersal distance; bs, the shape of seed dispersal kernels; and bp, the shape of pollen dispersal kernels. We also tested whether the diameter at breast height (DBH) and height of candidate parents affected the probability of being a true mother or father using these variables as selection gradients in NMπ. Selfing rate for both species converged to zero in all preliminary tests, although we tested several starting values. Therefore, this parameter was set equal to zero for the final estimation. At each step, before the estimation of the dispersal kernel parameters, genotyping errors were estimated with NMπ (setting the starting genotyping error rate at 0.01). This step allowed us to account for genotyping errors on parameter estimations. The two-component model was compared to the best fitting standard model using the Akaike information criterion (Akaike, 1973), which for a k-parameter model is defined as AIC = 2 (k-Ln(L)) (the smaller AIC, the better model fitting). Beginning with the simplest model of random mating as the null hypothesis, we tested the models with varying numbers of parameters and the shape parameter of dispersal kernel (b). The complete estimation procedure was repeated with different neighbourhood sizes (unlimited (the entire sampled area), 100 and 75 m of radius).
Comparisons of direct and indirect estimates of gene dispersal
In a two-dimensional space, σ2g = σ2s + 1/2σ2p (Crawford, 1984), where σs and σp refer to effective seed and pollen dispersal, respectively. Using this equation, we obtained an estimate of gene dispersal from direct estimates of seed and pollen dispersal to be compared with indirect gene dispersal estimate (Oddou-Muratorio and Klein, 2008). The parameters δs and δp, from the direct method, were converted to σs and σp according to Oddou-Muratorio and Klein (2008).
Results
Genetic data analysis
The probability of exclusion (P3 excluding a putative parent pair) exceeded 99 per cent above 9 loci for both species. For inbreeding unbiased estimates, the model showing the lowest value of DIC, for N. alpina and N. obliqua, was ‘nb’ which included null alleles and genotyping failures. However, inbreeding was not a significant factor in either of the two species (Fis = 0). Significant frequency of null alleles was detected for two (Notho214, 13 per cent, and NnBio11, 37 per cent) and four (Oakpum64, 11 per cent, NnBio37, 58 per cent, NnBio111, 47 per cent, and NnBio11, 54 per cent) loci for N. alpina and N. obliqua, respectively (Supplementary Data Table S2). However, considering the low effective number of alleles of both species (Supplementary Data Table S2), loci with null alleles were not excluded for indirect and direct estimations as in general they were also the most polymorphic. Performance of molecular markers was not enough for the analyses when loci with null alleles were excluded.
Indirect estimates of gene flow (based on spatial genetic structure of populations)
In agreement with models of isolation by distance, a significant linear decrease of Fij-coefficient with the logarithm of spatial distance was detected in both species; P-value of the slope (b parameter) was 0.012 and 0.007 for N. alpina and N. obliqua, respectively. Fine-scale spatial genetic structure was significant in mature trees of both species with individuals more related than expected by chance for the first distance class (<40 m, Fig. 2). Finally, the shape of the kinship curves was slightly similar for both species, with only significant regression at the shortest distance (Fig. 2). N. alpina computations of σg and Nb converged for values of de ≥ dobs, whereas in N. obliqua convergence was only obtained for de = dobs*2 (Table 1, Supplementary Data Table S3). In addition for N. alpina, the only species for which comparisons were possible, gene dispersal distance was slightly higher for low than for high effective density (Supplementary Data Table S3).
Analysis of FSGS in N. alpina and N. obliqua forest based on SSR data. Solid lines indicate the mean kinship coefficient per distance class (Fij) and dashed lines the limits of its 95 per cent confidence interval.
Estimates of fine-scale spatial genetic structure (FSGS) and historical gene flow for each species: average kinship coefficient between individuals of the first distance class (F1), FSGS intensity (Sp) and its standard error in parentheses, neighbourhood size (Nb), gene dispersal distance (σg) and 95 per cent confidence interval for effective density (de = dobs*2).
| Species | N. alpina | N. obliqua | |
|---|---|---|---|
| Parameter | Estimate | Estimate | |
| FSGS parameters | F1 | 0.0195 * (0.0044) | 0.0280 * (0.0147) |
| Sp | 0.0096 (0.0031) | 0.0144 (0.0057) | |
| Gene dispersal parameters | Nb | 200 (62–313) | 65 (38–363) |
| σg | 45 (25–56) | 28 (21–66) |
| Species | N. alpina | N. obliqua | |
|---|---|---|---|
| Parameter | Estimate | Estimate | |
| FSGS parameters | F1 | 0.0195 * (0.0044) | 0.0280 * (0.0147) |
| Sp | 0.0096 (0.0031) | 0.0144 (0.0057) | |
| Gene dispersal parameters | Nb | 200 (62–313) | 65 (38–363) |
| σg | 45 (25–56) | 28 (21–66) |
See Supplementary Data Table S3 for results corresponding to other tested de values. P-values: non-significant for P > 0.05; * for P ≤ 0.05
Estimates of fine-scale spatial genetic structure (FSGS) and historical gene flow for each species: average kinship coefficient between individuals of the first distance class (F1), FSGS intensity (Sp) and its standard error in parentheses, neighbourhood size (Nb), gene dispersal distance (σg) and 95 per cent confidence interval for effective density (de = dobs*2).
| Species | N. alpina | N. obliqua | |
|---|---|---|---|
| Parameter | Estimate | Estimate | |
| FSGS parameters | F1 | 0.0195 * (0.0044) | 0.0280 * (0.0147) |
| Sp | 0.0096 (0.0031) | 0.0144 (0.0057) | |
| Gene dispersal parameters | Nb | 200 (62–313) | 65 (38–363) |
| σg | 45 (25–56) | 28 (21–66) |
| Species | N. alpina | N. obliqua | |
|---|---|---|---|
| Parameter | Estimate | Estimate | |
| FSGS parameters | F1 | 0.0195 * (0.0044) | 0.0280 * (0.0147) |
| Sp | 0.0096 (0.0031) | 0.0144 (0.0057) | |
| Gene dispersal parameters | Nb | 200 (62–313) | 65 (38–363) |
| σg | 45 (25–56) | 28 (21–66) |
See Supplementary Data Table S3 for results corresponding to other tested de values. P-values: non-significant for P > 0.05; * for P ≤ 0.05
Direct estimates of gene flow (based on spatially explicit neighbourhood model)
Genotyping errors higher than 10 per cent were estimated with NMπ at three and four loci for N. alpina and N. obliqua, respectively, and were considered for further estimations (Supplementary Data Table S2). Estimations considering different neighbourhood sizes converged to similar values (Supplementary Data Table S4). Therefore, we report here only results of unlimited radius (Table 2). The best model according with AIC criterion was that including the exponential dispersal function for pollen and seed in both species (likelihood and relative weight of the models for AIC were 1 and 0.657 for N. alpina and 0.642 for N. obliqua). This model indicates that the estimated seed and pollen dispersal kernels have a shape parameter of 1, being therefore highly local. Both species showed (i) short pollen and seed dispersal distances, (ii) higher pollen than seed dispersal distance (2.4 and 1.3 times for N. alpina and N. obliqua, respectively) and (iii) considerable seed and pollen migration from outside the studied plot (Table 2).
Direct estimates of contemporary gene flow parameters of the best fitting model using the Akaike information criterion and 95 per cent parametric bootstrap confidence interval (CI): mean distances of seed (δs) and pollen (δp) dispersal, shape parameter of the seed (bs) and pollen (bp) dispersal kernels, rate of seed (ms) and pollen (mp) migration from outside the study plot, selection gradient of tree diameter and height on female (γd, γh) and male (βd, βh) fertility.
| Species | N. alpina | N. obliqua | |||||
|---|---|---|---|---|---|---|---|
| Parameter | Estimate | CI− | CI+ | Estimate | CI− | CI+ | |
| Seed dispersal and female fertility | ms | 0.25 | 0.07 | 0.44 | 0.41 | 0.33 | 0.50 |
| δs | 19.15 | 14.00 | 30.29 | 27.83 | 22.41 | 36.70 | |
| b s | 1.00 | – | – | 1.00 | – | – | |
| γd | 0.93 | 0.53 | 1.33 | 0.61 | 0.20 | 1.02 | |
| γh | – | – | – | 1.30 | 0.86 | 1.74 | |
| Pollen dispersal and male fertility | mp | 0.45 | 0.15 | 0.75 | 0.57 | 0.41 | 0.74 |
| δp | 46.08 | 27.41 | 144.39 | 36.58 | 26.46 | 59.24 | |
| b p | 1.00 | – | – | 1.00 | – | – | |
| βd | – | – | – | – | – | – | |
| βh | 1.88 | 0.72 | 2.98 | 0.61 | 0.20 | 1.02 | |
| Species | N. alpina | N. obliqua | |||||
|---|---|---|---|---|---|---|---|
| Parameter | Estimate | CI− | CI+ | Estimate | CI− | CI+ | |
| Seed dispersal and female fertility | ms | 0.25 | 0.07 | 0.44 | 0.41 | 0.33 | 0.50 |
| δs | 19.15 | 14.00 | 30.29 | 27.83 | 22.41 | 36.70 | |
| b s | 1.00 | – | – | 1.00 | – | – | |
| γd | 0.93 | 0.53 | 1.33 | 0.61 | 0.20 | 1.02 | |
| γh | – | – | – | 1.30 | 0.86 | 1.74 | |
| Pollen dispersal and male fertility | mp | 0.45 | 0.15 | 0.75 | 0.57 | 0.41 | 0.74 |
| δp | 46.08 | 27.41 | 144.39 | 36.58 | 26.46 | 59.24 | |
| b p | 1.00 | – | – | 1.00 | – | – | |
| βd | – | – | – | – | – | – | |
| βh | 1.88 | 0.72 | 2.98 | 0.61 | 0.20 | 1.02 | |
All estimates were performed considering unlimited neighbourhood. (−) means that the best fitting model was obtained with the shape parameter (b) fixed at 1 and with no effect of tree height on female (γh) fertility of N. alpina and of tree diameter on male (βd) fertility of both species
Direct estimates of contemporary gene flow parameters of the best fitting model using the Akaike information criterion and 95 per cent parametric bootstrap confidence interval (CI): mean distances of seed (δs) and pollen (δp) dispersal, shape parameter of the seed (bs) and pollen (bp) dispersal kernels, rate of seed (ms) and pollen (mp) migration from outside the study plot, selection gradient of tree diameter and height on female (γd, γh) and male (βd, βh) fertility.
| Species | N. alpina | N. obliqua | |||||
|---|---|---|---|---|---|---|---|
| Parameter | Estimate | CI− | CI+ | Estimate | CI− | CI+ | |
| Seed dispersal and female fertility | ms | 0.25 | 0.07 | 0.44 | 0.41 | 0.33 | 0.50 |
| δs | 19.15 | 14.00 | 30.29 | 27.83 | 22.41 | 36.70 | |
| b s | 1.00 | – | – | 1.00 | – | – | |
| γd | 0.93 | 0.53 | 1.33 | 0.61 | 0.20 | 1.02 | |
| γh | – | – | – | 1.30 | 0.86 | 1.74 | |
| Pollen dispersal and male fertility | mp | 0.45 | 0.15 | 0.75 | 0.57 | 0.41 | 0.74 |
| δp | 46.08 | 27.41 | 144.39 | 36.58 | 26.46 | 59.24 | |
| b p | 1.00 | – | – | 1.00 | – | – | |
| βd | – | – | – | – | – | – | |
| βh | 1.88 | 0.72 | 2.98 | 0.61 | 0.20 | 1.02 | |
| Species | N. alpina | N. obliqua | |||||
|---|---|---|---|---|---|---|---|
| Parameter | Estimate | CI− | CI+ | Estimate | CI− | CI+ | |
| Seed dispersal and female fertility | ms | 0.25 | 0.07 | 0.44 | 0.41 | 0.33 | 0.50 |
| δs | 19.15 | 14.00 | 30.29 | 27.83 | 22.41 | 36.70 | |
| b s | 1.00 | – | – | 1.00 | – | – | |
| γd | 0.93 | 0.53 | 1.33 | 0.61 | 0.20 | 1.02 | |
| γh | – | – | – | 1.30 | 0.86 | 1.74 | |
| Pollen dispersal and male fertility | mp | 0.45 | 0.15 | 0.75 | 0.57 | 0.41 | 0.74 |
| δp | 46.08 | 27.41 | 144.39 | 36.58 | 26.46 | 59.24 | |
| b p | 1.00 | – | – | 1.00 | – | – | |
| βd | – | – | – | – | – | – | |
| βh | 1.88 | 0.72 | 2.98 | 0.61 | 0.20 | 1.02 | |
All estimates were performed considering unlimited neighbourhood. (−) means that the best fitting model was obtained with the shape parameter (b) fixed at 1 and with no effect of tree height on female (γh) fertility of N. alpina and of tree diameter on male (βd) fertility of both species
Regarding selection gradients, significant positive effects on fertility were detected in both species. DBH had a positive effect on female fertility, being the mean and standard error of DBH for N. alpina and N. obliqua mature trees of 53 ± 31 and 72 ± 23 cm, respectively. The height of the tree had a positive effect on male fertility being the mean and standard error of height of 24 ± 8 and 31 ± 8 meters for N. alpina and N. obliqua mature trees, respectively. Additionally, a significant positive effect on female fertility was detected for height in N. obliqua (Table 2) (frequency distribution of diameters and heights in Supplementary Data Fig. S3).
Comparisons of direct and indirect estimates of gene dispersal
The values of σs and σp, estimated according to Oddou-Muratorio and Klein (2008), were 16.58 and 39.91 for N. alpina and 24.10 and 31.68 for N. obliqua. Therefore, the estimate of gene dispersal from direct estimates of seed and pollen dispersal were for N. alpina σg = √ (16.582 + ½ 39.912) = 32.73 and for N. obliqua σg = √ (24.102 + ½ 31.682) = 32.90. The gene flow estimate was, therefore, 37 per cent lower for direct than for indirect method in N. alpina and 15 per cent higher for direct than for indirect method in N. obliqua. However, these gene flow values lay within the range of indirect estimates (Table 1 and Supplementary Data Table S3).
Discussion
Gene dispersal in Nothofagus species
Estimations of gene dispersal from both direct and indirect methods gave congruent and similar results in both Nothofagus species, suggesting that the seed cut performed 20 years ago within this area has not significantly impacted patterns of gene dispersal distance. This result can be explained by the regularity of the seed cut applied and suggests that the residual tree density maintaining species composition and the homogeneous spatial distribution of remaining trees probably allowed the maintenance of gene dispersal.
Significant genetic correlation at the first distance class (< 40 m) was detected for both species, as in other Nothofagus species (Supplementary Data Table S5) and, a shallower slope of the kinship coefficients at further distances, explained by the local scale of the analysis (similar correlograms were obtained for species studied on local scales, Almeida Viera et al., 2010; Schroeder et al., 2014; Kitamura et al., 2018). The intensity of FSGS detected in N. alpina and N. obliqua (average Sp = 0.012, SD = 0.0034) was in the range of reported values for N. pumilio (0.0009–0.0630, Supplementary Data Table S5), but it was higher than the Sp for four outcrossing tree species with wind-dispersed seeds and pollen (Pinus strobus (0.01076), Acer saccharum (0.01016), Larix laricina (0.00450), Fraxinus excelsior (0.00196), Vekemans and Hardy, 2004). Likewise, historical and contemporary restricted gene flow was estimated in both species, along with contemporary restricted seed and pollen dispersal. Therefore, although the magnitude of FSGS usually results from restricted seed dispersal (Vekemans and Hardy, 2004), short-distance pollen dispersal could also explain FSGS in these Nothofagus species. Despite the existence of groups of related individuals, no significant inbreeding coefficients (selfing (s) = 0 for both species) were detected. This result agrees with the high degree of self-incompatibility previously reported for Nothofagus dombeyi, N. nitida and N. obliqua (Riveros et al., 1995) probably due to prezygotic barriers acting on the stigmatic surface (Torres and Puntieri, 2013). In addition, a significant but very low biparental inbreeding (tm-ts = 0.04) has been estimated for N. alpina (Marchelli et al., 2012).
We also detected significant seed and pollen immigration on both species. However, these migration events might be explained by short dispersal distance as our study plot/area belongs to a region of continuous Nothofagus forests. This interpretation agrees with the fitting of seed and pollen dispersal kernels to exponential functions (b = 1), but we cannot rule out that rare events of long dispersal distance also occur. This high rate of immigration contributes to gene flow, possibly reducing the negative genetic consequences (e.g. high inbreeding, low genetic diversity) of silvicultural management. Additionally, and relevant for conservation and evolution, the estimated seed migration rates refer to effective gene flow, i.e. to immigrant gametes whose establishment was successful. Similar levels of immigration rates have been reported for other wind-dispersed species such as Pinus attenuata, Quercus macrocarpa and Fagus sylvatica (Burczyk et al., 1996; Craft and Ashley, 2010; Oddou-Muratorio et al., 2011; Piotti et al., 2012).
Finally, the impact of tree diameter was only detected on female fecundity, whereas the impact of tree height was detected on male and female fertility. Plant size is often considered as a proxy of fecundity because larger individuals have more branches and therefore more reproductive organs (Torres et al., 2012). For example, in F. sylvatica and F. crenata tree diameter had an effect on female fertility (Asuka et al., 2005; Oddou-Muratorio et al., 2010). For South American Nothofagus species, seed production increased in stands with larger trees (Dezzotti et al., 2016). On the other hand, in wind-pollinated hermaphrodites, taller plants tend to be more male-biased than shorter plants (Sakai and Sakai, 2003), because they can disperse their pollen more extensively (Chybicki and Burczyk, 2013).
The existence of FSGS suggests possible persistent effects of habitat, biological or demographic limitations to gene flow. For example, in N. dombeyi and N. pumilio, patterns of FSGS have been related to the scale of episodic disturbance impacting regeneration establishment (Premoli and Kitzberger, 2005; Fajardo et al., 2016), and effects of habitat conditions have also been reported for N. antarctica and N. pumilio (Premoli and Steinke, 2008; Mathiasen and Premoli, 2013). In our study area, preharvest density (average value = 73 mature trees per hectare of each species, with an average distance of 12 m between trees) might have favoured mating among neighbouring trees, generating a FSGS fitting an isolation-by-distance model. This hypothesis has some support from the results of the indirect method. In N. alpina, for which comparisons were possible, gene dispersal distance was slightly higher for low than for high effective density.
In accordance with our results, limited seed dispersal for Nothofagus species has been reported based on field or laboratory experiments (Supplementary Data Table S6), suggesting that most seeds fall beneath mother trees. These values range from 6 to 250 m depending on the statistic reported (mean or maximal distance). Although restricted pollen dispersal is rather uncommon in wind-pollinated trees (Kremer et al., 2012), it has been however reported for some Fagaceae species (e.g. 8 m for Quercus oleoides, Deacon and Cavender-Bares (2015); 64.8 m for Q. lobata, Sork et al. (2002); 30.5 m for Q. robur, Moracho et al. (2016)). Among Nothofagus species, restricted pollen dispersal has previously been reported for N. alpina (< 35 m), based on an indirect method and genotyping of mother trees and seeds (Marchelli et al., 2012). However, the estimates of dispersal function they found were exponential power (b < 1) suggesting that other factors could determine the observed gene flow pattern. Stand density, intensity and direction of wind, seed and pollen characteristics could all influence seed and pollen dispersal (Moar and Myers, 1978; Facgri and Van Der Pijl, 1980; Austerlitz et al., 2007; García et al., 2015, 2017).
This is the first study focused on the simultaneous characterization of the fine-scale spatial genetic structure (FSGS) and reproductive system of two ecologically and economically important Nothofagus species. It was based on an intensive mapping, genotyping of mature trees and established regeneration (> 700 individuals) taken from Sola et al. (2016) and direct and indirect estimates of dispersal distances. Altogether, our results reveal short pollen and seed dispersal distance in both Nothofagus species and suggest that the regular shelterwood system has not significantly affected gene flow as we discussed. However, a change on global genetic diversity of the mixed forest under study has been reported because of the change on relative abundance of species in the post-harvest population that followed the implemented management (Sola et al., 2016). The conditions generated after logging benefit more N. obliqua than N. alpina, and therefore much more juvenile individuals of the first species were found (Sola et al., 2020). Considering that both species interchange genetic information and that the generation of hybrids individuals is common (Gallo et al., 1997; El Mujtar et al., 2017), we cannot rule out that the silvicultural system applied influence mating patterns and interspecific pollen and seeds dispersal distances. This question will be analysed in a future work considering analytical models taking into account possible hybridization and/or both genetic entities (species) as just one.
It is important to highlight that results regarding impact of silvicultural management on gene flow need to be considered carefully due to the limitations of the SSR markers we used (e.g. high frequency of null alleles, low polymorphism). Genotyping errors and null alleles could result in false parentage exclusion, overestimated migration rates, seed and pollen dispersal distances and biased FSGS patterns. We retained loci with null alleles as they were the most polymorphic and because the software used for the direct method (NMπ) take into account in their estimations genotyping errors due to mutation or lab mistake (Chybicki, 2018). However, we could not completely rule out possible biases. Highly polymorphic SSRs would allow a more refined evaluation of harvesting impact, but the studied species seem to have low levels of genetic diversity (Azpilicueta et al., 2004; Marchelli et al., 2008; El Mujtar et al., 2014), and until now such markers are not available. Therefore, new studies based on a higher number of SSRs or other kind of markers such as SNPs from massive genotyping (e.g. RADSeq, GBS) would be useful to validate our results.
Furthermore, the intense genotyping we applied constrains performing replications, and therefore results should be interpreted with caution as they could be affected by experimental biases, others than that of SSR markers, such as mortality of adult trees and size and design of sampling plot. Mortality of adult trees within the studied site could result in false migration events, while small and narrow sampling plot design (not circular but rectangular plot) could result on partly empty circular neighbourhood for the direct method, affecting parameter estimations. We recognized the limitations of our sampling plot design, and for that reason only central regeneration of the plot was used for estimations (Fig. 1), a decision supported by the analyses performed with all the regeneration that showed the same model fitting with greater immigration rates but similar short dispersal distances for both species (Supplementary Data Table S7). However, further replications of this study on others mixed Nothofagus forests under management would be useful to validate our conclusions.
Implications for management, conservation and restoration of N. alpina and N. obliqua
We showed at the studied plot that gene flow of N. alpina and N. obliqua is restricted and that diameter and height of trees are relevant for reproductive success. This work also indicated that at the studied plot, the current regular shelterwood system seems not to strongly affect gene flow (but see previous discussion about methodological limitations). Altogether these results indicate that the most commonly management applied on Nothofagus forests in Argentina that is performed at a stand scale, assuring homogeneously distributed trees in each unit and preserving some good and large trees to increase fecundity and dispersion, is actually a low-impact harvesting strategy. However, evaluations on (managed and unmanaged) mixed Nothofagus forests are required to determine environmental patterns of natural pollen and seed dispersal and interactions of environmental conditions and harvesting system on gene flow. For example, the significant fine-scale genetic structure detected in the study plot suggests that a harvesting strategy based on coarsely removal of entire patches of plants could not be appropriate, because it may affect the spatial structure of genetic diversity within the Nothofagus forest.
Likewise the presence of significant FSGS suggests that for future ex situ conservation, restoration and breeding strategies, seeds need to be collected from selected mothers located at a distance higher than gene dispersal (> 45 m in the study plot) to avoid collecting seeds from genetically related individuals. Moreover, considering that most vigorous trees contribute with a higher proportion of descendants, they should therefore be considered when pooling seeds from different trees to generate the plants for restoration activities.
Future suitability predictions based on spatially explicit zoning approach suggest that climate change might differentially affect N. alpina and N. obliqua (Marchelli et al., 2017). In these species, adapting to a changing climate is conditioned by the trade-off between having a long lifespan and high reproductive rate on the one hand and low dispersal capacity on the other. Any in situ conservation strategy should promote ecological connectivity to expedite natural migration.
Conclusions
The comparison of historical and contemporary gene dispersal estimates permits detecting the impact of human perturbations in previously stable habitats. Our results indicate that at the study area, (i) shelterwood systems with 40 per cent canopy retention have not significantly impacted on patterns of gene dispersal distance of N. alpina and N. obliqua; (ii) gene flow, both at the seed and pollen level, is not as extensive as it is generally assumed in wind-dispersed species; (iii) these limited dispersal abilities, and population dynamics and density, seem to be important factors of FSGS determination; and (iv) diameter and height of trees are relevant for reproductive success highlighting their importance on the definition of adequate management prescriptions. These results provide important information to plan management, conservation and restoration of Nothofagus resources based on improved knowledge of the spatial extent of dispersal, but further evaluations at different sites are required to have a clear picture of gene flow dynamics on N. alpina and N. obliqua forests.
Acknowledgements
We thank the staff at the Institute of Biosciences and Bioresources, National Research Council of Italy, particularly to Yoshiaki Tsuda for his kind help at laboratory activities, cátedra de Ordenación Forestal de la Universidad Nacional del Comahue, Grupo de Genética Ecológica y Mejoramiento Forestal del INTA Bariloche and at the Departamento Forestal del Parque Nacional Lanín, particularly M. Peñalba and M. Lara for assisting in the fieldwork. Thanks are also due to the reviewers for their constructive criticism and helpful comments on the manuscript.
Conflict of interest statement
None declared.
Funding
The Instituto Nacional de Tecnología Agropecuaria [Proyecto PNFOR 044001 Domesticación de especies forestales nativas, PE1104064 Aplicación de herramientas moleculares para el uso y la conservación de la diversidad genética forestal], the Programa de Mejoramiento de Especies Forestales (PROMEF) [BIRF 7520-AR] and CONICET (PIP 11220110100891), the Universidad Nacional del Comahue [Proyecto 04/S016 Ecología y manejo del bosque mixto de Nothofagus: un avance hacia la conservación] and the Institute of Biosciences and Bioresources, Division of Florence, National Research Council.
