Efficient Long-Distance Gene Flow into an Isolated Relict Oak Stand

Abstract

Geographically isolated and small populations outside a species' central distribution range are likely to be of major importance to a species' ability to quickly adjust its distribution range to global change dynamics. Gene flow from the outside plays a pivotal role in the fate of these marginal populations. It has been proposed that spatial fragmentation and perceived geographic isolation do not necessarily reflect a loss of genetic connectivity in tree species. However, the spatial limits of long-distance gene flow, as well as its magnitude and impact, are still generally unknown. In the present study, we analyzed long-distance pollen-mediated gene flow into an isolated relict stand consisting of 7 individuals of Quercus robur L. based on a total sample of 177 trees and 9 microsatellite loci. We show that pollen-mediated gene flow across more than 80 km in this wind-pollinated tree species contributed at least 35% of all successful pollinations in the investigated isolated and small oak stand at the eastern limit of the species' distribution. The observed pollen immigration shaped the genetic diversity of acorn progenies in the stand and might explain the comparably high genetic diversity in the persisting adult population.

distribution range dynamics, landscape fragmentation, reproductive connectivity, Quercus robur

In often rapidly changing climates, the mechanisms that allow tree species to survive are puzzling and still poorly understood. To date, fossil and genetic data have been used to generate hypotheses regarding high migration capabilities and associated range dynamics (e.g., Petit et al. 2002 for white oaks; Cheddadi et al. 2006 for Picea abies; Tollefsrud et al. 2008 for Pinus sylvestris). However, 2 studies found that relict tree stands and forest patches can persist outside of a species distribution range over long periods of time (Anderson et al. 2006; Shepherd et al. 2007). Furthermore, such a persistence scenario was proposed for a dominant European forest tree (Magri 2010). These findings challenge previous hypotheses that have focused on high migration rates. They suggest that small and persisting stands may be key to rapid range dynamics and the continuing survival of tree species (Hampe and Petit 2005; Petit et al. 2008).

A scenario proposing the persistence of isolated populations raises questions regarding long-term survival despite very small population size, the extent of local adaptation, and the existence and impact of long-distance gene flow (Bridle and Vines 2006; Lopez et al. 2008). A theoretical framework of evolution in marginal populations has been developed, describing scenarios of local adaptation, migration load, and genetic rescue. Central to these scenarios is the significance of long-distance gene flow. Successful external gene flow can rescue small populations from the effects of inbreeding and genetic drift (Willi and Fischer 2005; Pickup and Young 2008). Simultaneously, it might provide relict populations with the genetic diversity necessary for local adaptation (Aitken et al. 2008; Leimu and Fischer 2008). However, it could also limit local adaptation to suitable microhabitats. In this case, it might reduce the chances for persistence but will counteract the development of genetic constraints. These would reduce recolonization capabilities once the regional environment changes to more suitable conditions. Actual data on the existence and magnitude of successful long-distance gene flow into isolated marginal populations are generally lacking (Geber 2008), despite its pivotal role in adaptation processes and the resilience of tree species.

A pattern of substantial pollen immigration into stands has been confirmed as basically general across wind-pollinated tree species, mating systems, and spatial configurations (Ashley 2010). Focusing on oaks, pollen immigration into fragmented oak stands ranged from approximately 20% to 70% (Dow and Ashley 1996, 1998; Streiff et al. 1999; Nakanishi et al. 2004, 2005, 2009; Valbuena-Carabaña et al. 2005; Ohsawa et al. 2006; Pakkad et al. 2008; Pluess et al. 2009; Chybicki and Burczyk 2010). The impact of life history and ecological factors on the extent of pollen immigration into oak stands remains unclear. These factors include species-specific reproductive characteristics, the stage of investigated offspring, stand structure, and the level of fragmentation. In comparisons of the magnitudes of pollen immigration among studies, however, the parentage reconstruction method employed needs to be taken into account. Studies applying the exclusion approach reported approximately 50–70% of pollen immigration due to the conservative assignment of local parental trees. Those studies using probability based approaches (FaMoz; Gerber et al. 2003 and Cervus; Marshall et al. 1998; Kalinowski et al. 2007) generally only observed approximately 20–50%. An exception was the reported 62–66% of pollen immigration reconstructed with the neighborhood-model approach (Chybicki and Burczyk 2010). Nothing is known about successful gene flow across longer distances in oaks because all previous studies investigated stands within 1 km of conspecific oak stands. Petit and Hampe (2006), Bittencourt and Sebbenn (2007), and Ashley (2010) summarized successful pollen immigrations across distances of generally a few kilometers in other wind-pollinated trees. Reported gene flow distances of >10 km in wind-pollinated trees were identified in a small number of studies of isolated tree populations (Kaufman et al. 1998; Schuster and Mitton 2000; Bacles et al. 2005; Robledo-Arnuncio and Gil 2005; Bacles and Ennos 2008).

High levels of pollen immigration originating from sources across a potentially large and genetically diverse geographic area might counteract the detrimental effects of forest fragmentation (Kramer et al. 2008). Previous studies investigating genetic diversity in anthropogenic or naturally fragmented oak stands have drawn conflicting conclusions. Although some report that diversity levels are maintained in fragmented systems (Chung et al. 2002; Craft and Ashley 2007; Pakkad et al. 2008), others observed reduced diversity levels in fragments (Sork et al. 2002; Fernández-M and Sork 2007; Kittelson et al. 2009). The reasons for the contrasting results are not clear from the literature. Apart from specific landscape configurations, differences in sample design, molecular markers, and methodology might play a role. In summary, there is still no consensus regarding the conditions and distances under which the observed pollen-mediated connectivity is able to counteract the effects of forest fragmentation and geographic isolation.

The goal of the present study was to 1) evaluate the existence and extent of successful long-distance gene flow into an isolated oak stand and 2) examine its impact on genetic diversity. For this, estimates of long-distance gene flow were based on parentage reconstructions using microsatellite markers. Genetic diversity measures of isolated marginal Quercus robur stands were compared with stands from the main distribution range as well as to the offspring cohort.

Materials and Methods

Pedunculate oak (Q. robur L.) is an ecologically and commercially important forest tree that occurs throughout most of Europe (Ducousso and Bordacs 2004). In the easternmost part of the European temperate forest zone, it is the last remaining hardwood tree species under increasingly dry and exposed conditions. It constitutes, together with birch, the tree islands in the transition zone to the Pontic steppes (Walter and Breckle 1986). Quercus robur–dominated forests cover large areas west of the Southern Ural Mountain chain. The species, however, is absent from the steppes east of the Ural Mountains. Only 2 remnant stands of pedunculate oak are known in the eastern foothills near Sibay (Republic of Bashkortostan, Russia). These are 30 km apart with the next closest known oak occurrences being 80 km to the southwest and west, 150 km to the northwest, and 300 km to the north across the Ural Mountains (700–1600 m in height). The 2 stands are considered to be relicts of a more eastern distribution of oak forests during warmer and more humid postglacial periods.

All adult trees within the 2 oak stands east of the Ural Mountains were sampled and included in the study. The stand west of Kuseevo consisted of approximately 20 mature oak trees, the second stand northwest of St Sibay only of 7 trees. In this stand, acorns were present in the fall of 2007. They represented 2 cohorts: green acorns directly collected from the seed tree (Tree 1), that is, originating from the flowering period in 2007 and brown acorns collected from the ground. Embryos of these acorns were mummified, thus those seeds represented either acorns aborted earlier that year or acorns that were produced during previous years.

Genetic diversity and differentiation of the 2 Bashkirian stands were compared with 2 Central European stands of pedunculate oak in northern Germany (Behlendorf and Berkenstrücken, Kreisherzogtum Lauenburg, Schleswig-Holstein) including 42 and 69 adult trees, respectively.

Laboratory Procedures

DNA was extracted from leaves of adult trees, as well as from embryonic tissue removed from acorns following a protocol by Dumolin et al. (1995). Extracted DNA was adjusted to 10 ng/μl per sample. Two sets of multiplexed microsatellite primers were used to genotype individuals (Lepais et al. 2006). Set C included primers for microsatellite loci QrZAG112, QrZAG96, QpZAG110, and QrZAG11, set D the loci QrZAG87, QrZAG7, QrZAG20, QrZAG5b, and QrZAG65. Forward primers were fluorescently labeled as described in Lepais et al. (2006) with dye TET replacing dye NED. PCR reactions and cycling conditions were conducted with final volumes of 15 μl generally using 3 μl of the DNA extract. Amplified fragments were cleaned using a 4 M LiCl solution. Two microliters of PCR product and 8 μl loading buffer containing 0.25 μl size standard 550 (GE Healthcare) were mixed and loaded into a MegaBACE 500 (GE Healthcare) for separation and detection of amplified fragments. MegaBACE reads were analyzed using Fragment Profiler 1.2 (GE Healthcare) and proposed peaks and genotypes were verified visually. Based on those manual checks, locus QrZAG39 was removed from the analyses because its primers were not specific enough for the present sample. QrZAG65 clearly had nearly null (very low allele peaks) and null alleles in all populations, Russian and German. For this locus in all homozygotes one allele was replaced by missing data, resulting in a conservative estimate of long-distance gene flow. All population genetic analyses were based thus on 9 microsatellite loci. All genotypes identified in this study were checked against a database of all oak genotypes produced within our lab to detect potential contamination by Q. robur DNA from studies of other populations performed in the lab.

Data Analyses

The 9 microsatellite loci provided an exclusion probability of >99.99% for parent pair reconstruction across the data set. First, it was checked whether maternal seed trees in the stand St Sibay could be assigned to each acorn collected in this stand, that is, one mismatched acorn was identified and removed from further analyses. Assignment of maternal trees was conducted using an exclusion approach (E) and maximum likelihood-based reconstructions as implemented in Cervus 3.0 (ML, Marshall et al. 1998; Kalinowski et al. 2007). Acorns were counted as possessing a local pollen donor under the exclusion approach, if one or more fitting parent pairs were identified within the stand at St Sibay. The exclusion approach was applied in addition to the maximum likelihood approach to evaluate ML parameter sets and allow comparisons with published exclusion-based estimates of pollen immigration. The results of Cervus are robust with regard to uncertainties in allele frequency estimations. However, the sample sizes were so small that the calculated estimates of allele frequencies based on the heterogeneous pool of the 2 adult samples and the acorn sample (including haplotypes of external pollen donors) might deviate quite strongly from actual frequencies.

Three approaches were conducted and their results combined to obtain a conservative estimate of the minimum number of acorns resulting from pollen immigration. First, the average number of foreign alleles per acorn was calculated, as well as the number of acorns with more than 3 foreign alleles. Alleles were defined as “foreign” if they were present in the acorn generation but not in the adult population at St Sibay. Second, the minimum number of acorns was identified for which no compatible pollen donor could be found within the adult populations of St Sibay and Kuseevo. For this, the natural logarithm of the combined likelihood ratio (LOD scores) of all possible parent pairs of the adult trees in these 2 stands for each acorn were calculated using Cervus. Because all acorns were collected in St Sibay, parent pair combinations with seed trees from Kuseevo were subsequently removed from the distribution. Acorns were assumed to be the result of long-distance pollen immigration if all possible local parent pairs for a specific acorn were unlikely, that is, had LOD scores in the lower 90% of the “local” LOD score distribution (i.e., LOD ≤ −4.25). Third, exclusion and maximum likelihood–based reconstructions were conducted to assign pollen donors to the seeds. The brown acorns were collected from the ground. Thus, seed tree assignment was questionable and parentage analyses were conducted for all green and brown acorns. Paternity analyses with the green acorns collected directly from the seed trees produced similar results. Significance of parental assignment at a confidence level of 95% was evaluated using simulations with the following settings: All adult oaks in the 2 stands were used as candidate parents (27 trees) and set as 100% of potential parent trees, self-fertilization was allowed, the proportion of loci typed was set to 0.929, and the error due to mistyping and mutations was set to 0.03. The critical delta value (delta > 0.00) was obtained by simulating 50 000 offspring. This delta at the lower range limit is probably due to the small number of adult trees in the population and the large number of foreign alleles in the offspring cohort. The effective number of pollen donors and their diversity in the 2 acorn cohorts (green and brown) were calculated according to Nielsen et al. (2003) (k_e³) and Simpson (1949) (1/D), respectively.

Genetic diversity of the adult populations at both stands, as well as the offspring cohorts, was described using the number of genotypes and allele classes present, allelic richness (El Mousadik and Petit 1996), levels of heterozygosity, and significance of deviations from Hardy–Weinberg equilibrium (HWE; exact tests were performed using a burn-in of 50 000 steps and Monte Carlo Markov chains (MCMC) of 500 000 steps). Calculations of all diversity measures and the Garza–Williamson index (GWI) were performed using Arlequin 3.11 (Excoffier et al. 2005). An excess of heterozygosity was tested using the 2-phase model as implemented in the program BOTTLENECK (Cornuet and Luikart 1996). Reported are the results of the sign test and the Wilcoxon test. Census sizes were too low for the standardized differences test.

Locus-by-locus analyses of molecular variance (AMOVAs) (Excoffier et al. 1992) were conducted to gain insight into levels of differentiation and distribution of covariance components as implemented in Arlequin 3.11 (Excoffier et al. 2005). AMOVAs were performed based on the number of different alleles (“F_st-like”) and significance of the results was tested using 20 000 permutations using a confidence level of 0.05. The MCMC analyses for the exact tests ran for 500 000 steps after a burn-in of 50 000 steps.

Results

The microsatellite loci analyzed were highly polymorphic within the Russian data set (Table 1). Loci showed 9–21 alleles across all 66 samples. On average 10.11 ± 3.41 alleles were found per locus in the sample at Kuseevo. At St Sibay 2–9 alleles were found per locus within the 7 adult trees (on average 5.78 ± 2.20). All loci that could be tested by Cervus showed for the pooled Russian samples nonsignificant deviations from HWE with F values being small and mostly negative. For 3 loci, no test was performed (QrZAG7, QrZAG5b, and QrZAG65), due to too much missing data in the genotypes of the brown acorn cohorts.

Table 1

Diversity measures within the analyzed stands and offspring cohorts

Stand	N	Gfull	Gunique	Atotal	Alocus	Arichness	Arange	Hobs	Hexp	GWI
Kuseevo	20	8	8	91	10.11 ± 3.41	50.9	38.22 ± 22.14	0.85 ± 0.12	0.81 ± 0.09	0.30 ± 0.12
St Sibay	7	2	2	52	5.78 ± 2.20	39.7	32.44 ± 16.51	0.75 ± 0.29	0.73 ± 0.22	0.20 ± 0.09
St Sibay green acorns	23	21	21	92	10.22 ± 2.57	44.6	39.11 ± 17.39	0.77 ± 0.16	0.69 ± 0.15	0.28 ± 0.08
St Sibay brown acorns	16	6	6	50	5.56 ± 2.36	34.2	28.00 ± 15.00	0.74 ± 0.27	0.63 ± 0.21	0.21 ± 0.08
Behlendorf	42	32	32	151	16.78 ± 6.14	60.7	54.22 ± 20.30	0.78 ± 0.18	0.81 ± 0.17	0.31 ± 0.04
Berkenstrücken	69	54	54	188	20.89 ± 6.62	64.5	60.00 ± 16.60	0.81 ± 0.18	0.82 ± 0.19	0.34 ± 0.05

N: number of adult trees or acorns; G_full: number of fully identified multilocus genotypes (no missing data); G_unique: number of unique multilocus genotypes; A_total: number of allele classes present; A_locus: number of allele classes per locus averaged across all loci with standard deviation (SD); A_richness: average allelic richness across all loci with SD; A_range: average size range of alleles across all loci with SD; H_obs/H_exp: average observed/expected heterozygosity across all loci with SD (average across all loci with SD). SDs were calculated with regard to sampling loci, not individuals.

Parentage Reconstructions

Maternal seed trees were determined within the collection stand St Sibay for all but one green acorn. All green acorns collected from Tree 1 shared one allele per locus with that tree. All brown acorns collected from the ground between Trees 2 and 3 had Tree 2 as one of the reconstructed parents. The single green acorn collected from Tree 2 showed allele shifts by 1–2 size classes and possibly null alleles in comparison with its collection tree. This acorn was removed from the analyses because seed transfer into the stand seemed unlikely. This acorn was the cause for estimating a relatively high minimum error rate of 0.026 for the data set.

A high level of pollen-mediated gene flow into the stand is supported by the presence of an average of 39% foreign alleles per locus that are not found in the adult tree population of St Sibay. On average, 2.57 ± 1.63 foreign alleles were observed per individual acorn, and of 37 acorns, 21 (57%) seeds possessed 3 or more foreign alleles (Table 2). Local parent pairs within the 2 stands (maternal tree located at St Sibay, pollen donor at St Sibay or Kuseevo) were found to be extremely unlikely for 41% of all acorns across cohorts (Table 2). In comparison, acorns with parent pairs assigned at a confidence level of 95% showed LOD scores of >0.8. Parent pairs could be assigned to only 16% of all acorns within the stand near St Sibay using a strict exclusion approach. The likelihood approach assigned parent pairs to 51% of all acorns at a confidence level of 95% by allowing for mismatches between parents and offspring due to mutations, PCR or mistyping errors. Finally, as a conservative approach to define the extent of long-distance gene flow into the stand, all 3 criteria were taken together. Based on these combined criteria, 35% of all acorns and 52% of green acorns still were the result of long-distance gene flow.

Table 2

Estimates of the pollen immigration fraction for the offspring cohorts at St Sibay

Offspring cohort	No. of offspring	No. of maternal seed trees assigned	Av. no. of foreign alleles per acorn (±SD)	No. of acorns with ≥3 foreign alleles	All potential local parent pairs with LOD ≤ −4.25	No assignment of local pollen donor		Combined criteria
						E	ML
Green acorns	24	23	3.26 ± 1.32	19 (83%)	14 (61%)	22 (96%)	16 (70%)	12 (52%)
Brown acorns	14	14	1.43 ± 1.45	2 (14%)	1 (7%)	9 (64%)	2 (14%)	1 (7%)
Total	38	37	2.57 ± 1.63	21 (57%)	15 (41%)	31 (84%)	18 (49%)	13 (35%)

The average number and SD of acorn alleles that were not present in the adult tree population of St Sibay (foreign alleles) are given per acorn, as well as the number of acorns with 3 or more foreign alleles. Parentage within the stand was rejected if assignment of all potential local parent pairs resulted in LOD scores of −4.25 or smaller for all pairs (error level = 0.03). Parentage reconstructions were based on an exclusion approach (E) allowing no mismatches and a maximum likelihood approach (ML) taking into account sources of potential errors (error level = 0.03). Reported are the numbers of acorns for which no local pollen donor could be assigned at a confidence level of 95%. In the combined approach, only those acorns were defined as resulting from long-distance gene flow if they fulfilled all 3 criteria: 1) 3 or more foreign alleles, 2) all trio LOD scores were below −4.25, and 3) no parent pair was reconstructed (E and ML). Results were calculated after exclusion of one mismatching acorn.

A comparison of the 2 acorn cohorts showed that in the green acorn cohort nearly all successful pollen originated from pollen donors outside the stand (96% E, 70% ML). In the cohort of brown acorns, pollen immigration amounted to 64% (E) and 14% (ML). The large difference between E and ML reconstructions in this cohort was probably due to a larger amount of missing data in the brown acorn cohort. Higher levels of pollen immigration were also observed for the other measures of gene flow. Similarly, the numbers and diversities of reconstructed pollen donors (ML) differ for the 2 acorn cohorts that seem to represent 2 half-sib families (green acorns = collection Tree 1, brown acorns = reconstructed Tree 2). Both, the number of effective pollen donors (k_e³ = 21) and their diversity (1/D = 7) are much higher in the green cohort than in the brown (k_e³ = 4.8; 1/D = 3.8). In the green acorn cohort, each acorn had a different pollen donor, whereas in the brown cohort most pollen donors (assuming Tree 2 is the seed tree) contributed 2–5 acorns. In agreement with a lower fraction of pollen immigration and diversity found in the brown acorn cohort, selfing was also more frequent in this cohort (2 acorns = 14%) than in the green acorn cohort (1 acorn = 4%) based on the reconstructed parent pairs at the 95% confidence level.

Maximally 5 (0 E, 5 ML) of the reconstructed pollen donors were assigned to the second relict oak stand east of the Ural Mountains at Kuseevo consisting of 20 adult trees, 30 km north of the investigated stand at St Sibay. However, this result depended very much on the error rate used in the ML parentage assignment. Assuming that pollination from Kuseevo is unlikely, the employed error rate in the ML reconstructions provides a minimum estimate of long-distance gene flow.

Genetic Diversity

The adult populations of the 2 oak fragments in Bashkortostan showed a reduced genetic diversity in comparison to the German stands in the central part of the distribution range (Table 1). However, the 2 fragments were still amazingly diverse in comparison to the German stands. The genetic diversity of the offspring cohorts is not strongly reduced in comparison to the maternal stand at St Sibay (brown acorns) or even increased (green acorns). No significant deviations from HWE were detected for most of the investigated microsatellite loci in the 2 adult populations or in the offspring cohorts from St Sibay. Significant deviations from HWE per locus (Kuseevo: 1 locus, St Sibay: 3 loci, St Sibay green acorns: 3 loci, and St Sibay brown acorns: 5 loci) were due to excesses of observed heterozygotes apart from locus QrZAG65 at Kuseevo. An excess of heterozygotes was also observed for the 2 Bashkirian stands and both acorn cohorts if all loci were combined. This excess of heterozygotes probably indicates historical pollen immigration and not heterozygote advantage because it was already present in the acorn cohorts. Both Bashkirian stands differ from each other just as much as from the German stands (Table 3). The observed significant differentiation between the 2 Bashkirian stands is probably due to the very low number of adult trees at St Sibay, showing the effect of genetic drift. Global differentiation among the 4 oak stands is highly significant (scenarios ((K, S), (BD, BS)): F_st = 0.035, P ≤ 0.001 and (K, S, BD, BS): F_st = 0.054, P ≤ 0.001).

Table 3

Pairwise differentiation of the 2 Bashkirian (Kuseevo, St Sibay) and 2 northern German (Behlendorf, Berkenstrücken) stands of Quercus robur based on F_st (lower triangle)

	Kuseevo	St Sibay	Behlendorf	Berkenstrücken
Kuseevo	—	≤0.001	≤0.001	≤0.001
St Sibay	0.099	—	≤0.001	≤0.001
Behlendorf	0.067	0.102	—	0.06
Berkenstrücken	0.067	0.102	0.003	—

Significance was evaluated using 1000 permutations (upper triangle).

A small ratio of the number of alleles to the range in allele sizes (GWI; Garza and Williamson 2001) and an excess of heterozygosity (Hk; Cornuet and Luikart 1996) in the adult population would point to recent bottleneck events. Although Hk is more likely to detect less severe, very recent events in populations with originally small θ, the GW index is more likely to identify long-lasting bottlenecks in populations with prebottleneck large θ that made a demographic recovery. The GWI of the stand at Kuseevo with its 20 adult oak trees lay in the range of the 2 German stands. The GWI of the stand at St Sibay was lower. However, this stand has a sample size of only 7 adult trees. Both stands showed the expected L-shaped frequency distributions and neither the sign tests (P values Kuseevo: 0.99, St Sibay: 0.057) nor the Wilcoxon tests (P values Kuseevo: 0.94, St Sibay: 0.90) of Cornuet and Luikart's Hk (1996) for the 2 stands were significant.

Discussion

This study investigated 2 strongly isolated relict tree stands that were situated 80 km outside the eastern limit of the continuous distribution range of Q. robur and 30 km apart from each other. The St Sibay stand studied in detail received a considerable influx of successful external pollen despite its geographic isolation and the Southern Ural Mountain range as potential barrier to gene flow between it and the main distribution range. In oaks, no studies of strongly isolated stands exist that included direct measures of long-distance gene flow. However, studies of geographically strongly isolated stands of other wind-pollinated tree species have reported more or less extensive long-distance gene flow. The most similar geographic setting was described for a relict stand of P. sylvestris in central Spain (Robledo-Arnuncio and Gil 2005). This stand of only 36 adult trees was located outside the species' main distribution range with the next closest conspecific stands 30 km and 60 km away. Here, a fraction of 4.3% of long-distance pollen immigration was reconstructed using an exclusion approach. A similarly high amount of pollen immigration (6.5%, exclusion approach) was reported for an isolated stand of Pinus flexilis in Colorado (United States; Schuster and Mitton 2000). This stand consisted of 455 adults with the next closest stands 2 and 5 km away and a distance of 100 km to the main distribution range. The authors of both studies suggested that these levels of long-distance gene flow already have a considerable long-term evolutionary impact. In a study of Cecropia obtusifolia, 27.1% and 9.6% of all offspring pollen donors were located in stands 6 and 14 km, respectively, from the seed-collection stand employing the exclusion approach (Kaufman et al. 1998). Long-distance pollen immigration at a comparable level to the present study was estimated into a catchment area with 3 small remnant Fraxinus excelsior stands using several methods (Bacles et al. 2005; Bacles and Ennos 2008). Between 43% and 68% of pollen immigration was observed with pollen either originating from 2 additional remnants in the catchment or from stands more than 10 km away. So far, all studies of geographically strongly isolated populations of wind-pollinated trees have reported at least low levels of long-distance gene flow.

The reconstructed pollen immigration of 84% (E) and 51% (ML) into the stand St Sibay is high in comparison with values reported in other oak studies. Even the 35% of immigration estimated by the conservative combined criterion is still very high if compared with the relict pine stands. However, by now several studies have shown that small fragments act as pollen traps, that is, the smaller the stand size, the larger the fraction of pollen immigration (e.g., for wind-pollinated trees and herbs: Sork and Smouse 2006; Fénart et al. 2007; Fernández-M and Sork 2007; Bacles and Ennos 2008; Wang et al. 2010). The stand at St Sibay was an open, monospecific oak stand of only 7 adult trees of small to intermediate size. It was situated in a small dip near the crest of a north–south running hill range within a steppe landscape. The crowns of the trees did not form a closed canopy and not all individuals might flower synchronously and in all years, as is the general pattern in oaks. In the present data set of the 7 trees, only 4 contributed as pollen donors to the collected acorn crop. It, thus, seems unlikely that during the flowering period a local pollen cloud of high concentration was formed. Accordingly, pollen competition between local and external pollen will have been comparably low. This is apparently still the case, if one takes into account that concentrations of viable pollen in arriving air masses might be low after long-distance transport (Schueler et al. 2004).

Long-distance gene flow into the stand was variable between the 2 cohorts, though always ranging from “present” to “considerable.” It is unclear whether the brown acorns were aborted acorns from the flowering period 2007 or remnants of previous flowering seasons of recent years. They were mummified and infected by insects and fungi. In the first case, the higher fraction of within-stand pollination, as well as a higher selfing rate in this cohort, suggests that abortion might be at least partly due to inbreeding. In the second case, if the cohort of brown acorns represented acorn crops of previous years, long-distance gene flow, while annually variable, might not be an exceptional event. However, for insights into the frequency and variability of long-distance gene flow into the St Sibay stand, data from several reproductive seasons are needed.

Grasslands in which trees are highly visible dominate the eastern foothills of the Southern Ural Mountains. To our knowledge, apart from the second stand at Kuseevo, no further small and isolated oak stands exist in the area. The reconstructed contribution of oaks at Kuseevo to pollen immigration into St Sibay depended very much on employed method and error rate. Taking into account the presumably limited pollen emission of this relatively small stand, it seems more likely that immigrant pollen originated mainly from the central area of oak forests west of the Ural Mountains. Similarly, the substantial genetic diversity in the acorn cohorts points to extensive and diverse pollen sources in the main distribution range. In years with suitable spring meteorological conditions, oak forests covering extensive areas produce large amounts of pollen, which are released into the air and could be transported by prevailing wind systems over long distances and at considerable altitudes into the steppes. Long-distance transport of pollen clouds has been reported for several herb and tree species (e.g., beech, ragweed) across Europe and from boreal North America into the tundra environment of Greenland using pollen count data from sampling stations (Ranta et al. 2006; Cecchi et al. 2007; Belmonte et al. 2008; Rousseau et al. 2008). Here, we show that such long-distance dispersed pollen was still able to successfully fertilize female flowers after the transport and produced a considerable fraction of the acorn crop at St Sibay.

No signatures of recent population size changes, that is, founder events, were found for both analyzed Russian stands. This suggests that the observed reduction in genetic diversity in the 2 stands is due to genetic drift. Thus, the Bashkirian fragments seem to be persisting relicts of a former, more Eastern distribution of pedunculate oak rather than recently founded populations of a current expansion. The genetic impact of long-distance pollen dispersal on stand persistence depends not only on its quantity but also on the genetic diversity of the incoming genes, that is, the diversity of the contributing pollen sources (Sork and Smouse 2006). Fernández-M and Sork (2007) showed that the genetic diversity of fragments of the Andean oak Quercus humboldtii was reduced despite high gene immigration. However, the investigated stand at St Sibay is much more geographically isolated and, in addition, smaller than all tree stands investigated to date. It will only receive the very far tails of realized pollen dispersal kernels due to its strong geographic isolation. This, however, means that the stand and, thus, sample a very diverse pollen pool (Bacles and Ennos 2008). In consequence, the stand is an example of a presumably rare connectivity pattern combining strong isolation with a high number of pollen sources (Sork and Smouse 2006). Sork and Smouse describe this pattern as possessing a low risk of losing its genetic diversity over time. In congruence with this, we found that the observed high influx of long-distance dispersed genes into the oak stand at St Sibay substantially contributed to the genetic structure and diversity of the offspring cohorts. The results suggest that genetic diversity is maintained or even increased during regeneration in the studied stand.

Studies of tree stands at species range margins and stands in regions of long-term scattered and fragmented distributions observed for several wind-pollinated tree species that local stands were genetically diverse and differentiation between stands was low even after long time periods (often several thousands of years) of fragmentation (Schuster and Mitton 2000; Mosseler et al. 2004; Bacles et al. 2005; Ledig et al. 2005). In oaks, this was reported for fragmented stands of sessile oak (Qurcus petraea (Matt.) Liebl.) at its range margin in Ireland (Muir et al. 2004) and bur oak (Quercus macrocarpa Michx.) tree islands in North American tallgrass prairies (Craft and Ashley 2007). However, in the latter system, young cohorts were found to be stronger differentiated between stands than old trees (Kittelson et al. 2009). Here, Kittelson et al. suggested as underlying cause reduced connectivity due to a drastic decrease in habitat. The magnitude of pairwise differentiation between the 2 German pedunculate oak stands in the present study is similar to measures of pairwise F_st between fragmented stands of sessile oak in Ireland and stands in western Europe (Muir et al. 2004). The higher differentiation between the German stands and the Bashkirian stand near Kuseevo can be explained by isolation-by-distance. The high differentiation between the 2 Bashkirian stands, as well as between St Sibay and the German stands, is probably due to the small sample size, that is, genetic drift, of the stand at St Sibay. Overall, the Bashkirian stands fit the general patterns of diversity and differentiation observed for long-term fragmented oak stands.

In conclusion, we observed that effective long-distance gene flow is extensive and contributes considerably to the genetic diversity of the analyzed offspring cohorts in the intensely studied small and isolated oak stand at St Sibay east of the Ural Mountains. The results suggest that the substantial impact of long-distance pollen immigration from very diverse pollen sources on genetic diversity is an important factor for the long-term persistence of this relict stand.

Funding

German Federal Ministry of Food, Agriculture and Consumer Protection (Cooperation in Agricultural Research Grant Germany/Russia); German Science Foundation (DE 1172/3-1).

We would like to thank I. Schulze for conducting the molecular lab work for this study and the reviewers of the manuscript for their helpful comments and suggestions that improved the study greatly.

Author notes

Corresponding Editor: Brian Murray