Could clonality contribute to the northern survival of grey alder [ Alnus incana ( L . ) Moench ] during the Last Glacial Maximum ?

Reconstruction of the glacial and postglacial history of a species, including lifehistory traits, provides valuable insights into the relationships between ecological and genetic factors shaping phylogeographic patterns. Clonality appears as a trait of high importance for survival in northern refugia. In the present study, the phylogeographic structure of 24 Alnus incana (grey alder) populations and clonal structure in seven populations were analyzed utilizing four microsatellites markers. Palaeobotanical data were collected and combined with the genetic results in order to support the possibility that this species survived in northern refugia. Our study indicated that: (i) Balkan populations are the most divergent, which likely reflects their long-term in-situ existence, (ii) Western Alpine populations are genetically different from other European populations, which corresponds with palaeobotanical data, suggesting that this region served as a refugium for this species, (iii) the macrofossil data indicate that the Scandinavian and northeastern Polish populations are likely derived from the refugia located in northern latitudes, (iv) Western and Eastern Carpathian populations form separate groups, which indicate that both regions could serve as refugia. Clonality was an important factor in allowing grey alder to survive in northern regions during the last glaciation. However, this mode of reproduction has also influenced the population genetic structure, as we noted rather low level of gene diversity, HE = 0.386 and low allelic variability, A = 3.8, in this species.


Introduction
In the recent past, it was generally assumed that during the Last Glacial Maximum (LGM; 18-25 ka), suitable environmental conditions for tree species survival were restricted to the refugia located on the Balkan, Iberian, and Italian Peninsulas [1].An increasing number of phylogeographic studies have made it more evident that the premise of southern refugia and their key role in the postglacial history of trees is an oversimplification.For example, some species had the highest level of genetic diversity and allelic richness in the middle or northern latitudes, rather than in southern refugia as assumed by the established paradigm of the pattern of postglacial colonization [2].Additionally, macrofossils of the boreal trees such as Picea abies (L.) Karst.
(Norway spruce), Pinus sylvestris L. (Scots pine), and Betula pendula Roth (silver birch), as well as some mesophilous Quercus spp.(oaks) or Fagus sylvatica L. (common beech) suggest the persistence of tree stands close to the ice-sheet margins [3].Currently, the theory of Steward and Lister [4] on the role of cryptic northern refugia located in microenvironmentally favorable sites explains much of the observed distribution and genetic diversity of tree species.
Bhagwat and Willis [5] found that tree species that likely survived in northern micro-refugia were wind-dispersed, small-seeded, and habitat-generalist.Additionally, all those species were capable of clonal growth.Clonal reproduction can be considered as an alternative option in a life cycle in a situation when the sexual reproduction is reduced or absent due to environmental conditions and may have profound influence on spatial genetic structure [6].Clonal plant species dominate in the composition of plant communities of harsh environments, e.g., high latitudes and altitudes [7].The long-term survival of the Australian shrub, Erythroxylum pusillum Clarkson, was possible due to clonality [8].Population separated from the range core by the Gulf of Carpentaria (ca.500 km) is derived from a single ancestral genetic lineage established ca. 12 000 years ago.Another example comes from the Hawaiian population of Sphagnum palustre L., which was probably founded by a single founding episode ca.49-54 ka ago and clonal reproduction is the sole means of population growth and persistence [9].Also, in the early successional phase of modern glacier forelands, after initial seed recruitment, clonality is the major mode of reproduction and colonization [10].Öberg and Kullman [11] have stated, that the ability of Norway spruce to clonal reproduction was the major factor that enabled and enhanced the survival of this species during the early immigration phase into the Scandes at the very beginning of the Holocene.
Alnus incana (L.) Moench (grey alder) is a light-demanding, cold-tolerant, anemophilous and anemochoric tree species.Its natural range extends from the Fennoscandia, throughout the northern part of the Eastern European Plain, the Alps, the mountains of Central Europe, the Caucasus, and as far as to the Ural Mountains [12] (Fig. 1).It has been reported that grey alder from tree-limit and forest-limit in the Scandes reproduces only but extensively by root shoots and epicormic stem shoots [13].Palaeobotanical studies on the genus Alnus support the idea of northern refugia formed by the species [14].Recent reconstruction of the postglacial history of grey alder made with genetic markers indicated that the species was able to survive beyond the classic southern refugia in areas of Central Europe [15].Those populations were the main source for subsequent colonization of Fennoscandia and Eastern Europe.We assumed that the ability for clonal reproduction could largely facilitate the northern survival of A. incana during the LGM.Using nuclear microsatellite markers we aimed to analyze the phylogeographic pattern of the species in relation to the clonal structure of populations.Next, to test the obtained phylogeographic pattern we combined genetic results with palaeobotanical data collected in published sources.

Sampling and genetic analysis
Grey alder was sampled in 24 locations covering the Central European part of the species range from the Scandinavian Peninsula to the Balkans (Tab. 1, Fig. 1).Leaves were collected from 28-32 individuals per population.In total, 721 individuals were subjected to genetic analysis.The clonal structure was investigated in seven populations: Bulgaria, Balkans, (B_1), the Austrian Alps (A_1), the French Alps (A_2), the Italian Alps (A_3), the Scandinavian Peninsula, Finland (S_2, S_3), and the Scandinavian Peninsula, Norway (S_6).Grey alder in those localities formed homogenous groves/ thickets along the streams and the ramets were sampled along a transect.Within the transect, each ramet was separated from another by ca. 10 m and the transect was expanded until the number of ramets approached 30.

Genetic diversity, differentiation, and clonality
Genetic diversity measured as expected heterozygosity (H E ), observed heterozygosity (H O ), and average number of alleles per locus (A) were estimated in Fstat v. 2.9.3.2 [17].Null allele frequencies were obtained following the expectation-maximization (EM) algorithm using software INEst 1.0 [18].Conformity to Hardy-Weinberg (H-W) genotypic proportions was tested using the exact test in GenePop v. 4.0 [19].The multilocus inbreeding coefficient F IS was evaluated using Fstat, and INEst was used to calculate the corrected value of F IS which considers null alleles.

Tab. 1 Location and parameters of genetic variability of 24
A multilocus unbiased estimate of Wright's fixation index (F ST ) was calculated as a measure of genetic differentiation in Fstat [20].Null alleles may lead to overestimation of F ST , thus the global F ST of Weir using the ENA correction [21] was calculated in FreeNA.Pairwise genetic differentiation (F ST ) between all pairs of populations was calculated using Arlequin 3.5 and the significance of the differentiation was tested using 3000 permutations [22].
Clonal structure in selected sites was assessed by genotype diversity according to the formula R = (G − 1)/(N − 1) [23], where G is the number of total genotypes detected and N is the number of individual ramets collected along a transect.GenAlEx 6.5 software was used for inferring the clonal structure, that is, the number of unique multilocus genotypes, the number of multiplied genotypes, and the probability of the identical genotype arising in the result of independent sexual reproduction (p gen ) [24].

Geographic structure
Principal coordinate analysis (PCoA) was performed using GenAlEx.Changes in the diversity parameters values within populations were estimated by linear regressions for H E , A, and F IS with latitude, longitude, and altitude (Statistica v.11).Bayesian methods implemented in Geneland v. 2.13.1 were used to investigate geographic structure [25].Georeferenced genotypes were used as prior information in the estimation procedure which included 100 independent runs, 200 000 iterations each, saving every 50th with a K set from 1 to 25. Owing to the presence of null alleles, a null allele model was used.The optimum number of K for the dataset is displayed by Geneland as the modal value along the MCMC chain.Analysis of molecular variance (AMOVA) for groups defined by Geneland was performed in Arlequin and significance was verified using 3000 permutations [22].

Collection of palaeobotanical data
In palaeobotanical analysis we considered mainly macrofossils that have more reliable taxonomical resolution than pollen data [26], which is especially crucial in case of alders.The species determination of pollen within alders is not routine, although possible [27].However, separation of the pollen of A. incana and A. glutinosa (A.glutinosa / A. incana-type) is very difficult and is practiced only by some authors [28].
The A. incana / Alnus sp.palaeobotanical dataset was restricted chronologically to the last full glacial (including LGM), LG (Late Glacial) and early Holocene records derived from published literature and from the publicly available Eurasian Macrofossil Database at the University of Oxford [29].If there were several records of A. incana / Alnus sp. for a site, only the oldest one was included in the dataset and shown on the generated map.Due to limited access and problems with exploring original publications used by Tallantire for a map of the distribution of alders fossil remains in Fennoscandia, these data were not included in our map [30].Those who are interested may refer to his paper.
Three different categories of data were collected to produce a map illustrating the distribution of fossil records: 1 -macrofossils specifically determined as A. incana, 2 -macrofossils determined as Alnus sp., 3  or charcoal fragments which are not identifiable to the species level [31].A small number of the records were fruits, fruit scales, cones, or other remains that are badly preserved and thus devoid of diagnostic features.This category of data was included in our study, especially if A. incana was proposed as the most probable species by the author reporting on a site.We used some selected pollen records of A. glutinosa / A. incana-type.We chose only those stands where the author of the data attributed them specifically to the local presence of A. incana supporting such statements by well-substantiated arguments.

Genetic diversity and differentiation
Tests for linkage disequilibrium between pairs of microsatellite loci in each population did not indicate significant relationships.Genetic variation at nSSR loci was low with the average expected heterozygosity (H E ) 0.386 and only 6.25 alleles per locus (Tab.2).Twenty-five different alleles were noted in total for 721 individuals.The highest gene diversity was reported for locus L5.5 (H E = 0.670) and lowest for locus L022 (H E = 0.237).The number of alleles per locus ranged from 9 to 4 in Locus L3.1 and Locus L5.4, respectively.In case of two loci, L3.1 and L5.4,significant departure from Hardy-Weinberg genotypic proportions was noted.However, these two loci were affected by the null alleles at most (Tab.2).Highest population differentiation was noted for the Locus L5.4 (F ST = 0.119) and lowest for L3.1 (F ST = 0.053).
Gene diversity (H E ) and number of alleles per locus (A) exhibited high variation among the studied populations.H E ranged from 0.619 in Population B_1 (Balkans, Rila) to 0.242 in Population S_2 (Scandinavian Peninsula, Finland), and A from 4.25 in B_3 (Balkans, Pirin Mts) and EC_3 (Eastern Carpathians, Bieszczady Mts) to 2.25 in A_2 (Alps, France) (Tab.1).The substantial frequency of null alleles and generally high values for the inbreeding coefficient (F IS ) were very evident.The corrected estimation of F IS in INEst, however, resulted in a decrease in the value of F IS in most of the populations (Tab.1).
The overall level of genetic differentiation, based on sequence variation at nSSR loci, was low but highly significant (F ST = 0.0836; p < 0.001).The majority of the pairwise F ST were also significantly greater than zero (data not shown).The highest value of pairwise F ST was 0.3246 and it was found in the population from Rila, Balkans (B_2) and the population from the western coast of Norway (S_4); the lowest differentiation was noted between two Finnish populations (S_2 and S_3).A high level of genetic differentiation was also observed between most of the populations from the Central and Northern Europe vs.Western Alpine populations, as well as between Balkan populations and the remaining stands in Europe.
Tab. 2 Characteristic of four microsatellite loci used in the study of the genetic structure of Alnus incana populations in Europe.

Geographic structure
A PCoA analysis conducted at the level of population resulted in the first two axes accounting for 39.7% and 26.8% of the total variation, respectively (Fig. 2).The ordination of the studied populations according to the first axis underlined the distinct character of the Alpine Population A_2 and revealed two major groups of populations.The first group contained Carpathian and Balkan populations with a single stand from Bohemian Massif, while the second one comprised all of the remaining stands from Central Europe and Scandinavian Peninsula.Testing the relationship between genetic structure coefficients and geographic distances (latitude, longitude, and altitude) revealed a significant decrease in gene diversity (H E ) with increasing latitude (R 2 = 0.46; p < 0.001) and altitude (R 2 = 0.30; p < 0.01).Generally, the most-southerly located populations presented the highest level of gene diversity (Tab.1).No significant relationships with geographic location were noted for the remaining parameters, i.e. average alleles per locus (A) and inbreeding coefficient (F IS ).Bayesian analysis detected 11 groups with a log of posterior probability of -1637 (Fig. 1).The inferred clusters were as follows: Cluster I -two Western Alpine populations (A_2 and A_3), Cluster II -three Balkan populations (B_1, B_2, and B_3), Cluster III -two Eastern Carpathian populations (EC_2 and EC_5), Cluster IV -one population from Finland (S_2) and one from Norway (S_6), Cluster V -population from the North European Plain (NEP_2) and population from Finland (S_1), Cluster VI -two Western Carpathian populations (WC_1 and WC_2), Cluster VII -two populations from western Norway (S_4 and S_5), Cluster VIII -population from Giant Mountains (BM), Cluster IX -three populations from the North European Plain (NEP_1, NEP_3, and NEP_4), Cluster X -Eastern Carpathian population (EC_1) and population from Finland (S_3), and Cluster XI -eastern Alpine population (A_1).AMOVA indicated that 6.4% of total variation resided between these groups and the differentiation was highly significant (p < 0.001).

Clonal structure
A considerable amount of vegetative reproduction was noted in seven of the studied populations (Tab.3).The smallest genotypes/individuals ratio was recorded for the Finnish population (S_2; R = 0.414) where only 13 unique genotypes among 30 sampled ramets were found.The largest number of unique genotypes (27) was found in the population from Balkans (B_1; R = 0.896).The size of genets varied among studied populations with the average number of ramets per genet being 3.15.The most frequent number of ramets within the genet was 2 (17.3%).The largest genet with 13 ramets was detected in Population S_2 (Finland).The values of probability of emerging of multiplied genotypes in result of independent generative reproduction events were very variable in particular populations (Tab.S1).They were ranging from low (p gen = 6.6 × 10 −3 ) to considerably high (p gen = 2.528 × 10 −1 ), which reflects low number of loci used and their moderate polymorphism.

Distribution of fossil records
Definitive fossil remains of A. incana, and those with a high probability of being grey alder, were not frequent in the palaeoecological records (Fig. 3, Tab.S2).Our survey indicated the presence of this species in the higher latitudes of the Russian Plain during the last full glacial and probably (Alnus sp.) in the Hungarian Plain, Bohemia, and northern Carpathians (Slovakia); convincing arguments for the potential presence of grey alder in the LGM in southern Britain (the Bodmin Moore site) have been previously presented [32].An indication of the possible survival of A. incana in Central Europe and even further northward can be drawn from the location of the sites in which this species has been documented for the LG: Kašučiai Lake in Lithuania and Starunia in NW Ukraine.These findings led us to the opinion that several records of Alnus sp.macrofossils (mainly wood) of the same age from the area of Poland probably belong to grey alder as well.Pollen data also support the presence of this species in the LG in the Alps and in the mountains of Scandinavia [28,33] The early-Holocene appearance of A. incana in the middle and high latitudes deserves special consideration as well.In Kreekrak (the Netherlands), grey alder grew at about 11.4 ka cal.BP [34], wood of Alnus sp.dated to ca. 10.3 ka cal.BP was found at the KH1 site (Estonia) [35] and several occurrences of A. incana mega-and macrofossils dating to the 10th millennium cal.BP was reported from Scandinavia [36,37].

Discussion
Geographic structure and fossil records: glacial survival of grey alder Grey alder habitat requirements, particularly in thermal and precipitation parameters would potentially enable it to survive the LGM in stands located further to the north in Europe.In Scandinavia, fossil remains of boreal trees dated to the LGM and LG, and evidence of an early postglacial appearance of some mesophilous taxa, have been the subject of vivid debate for more than a decade [37][38][39][40][41][42].Based on analysis of modern and ancient mtDNA from Scandinavia, Parducci et al. [43] suggested that Picea abies and Pinus sylvestris survived the last glaciation as far as northwestern Norway.This result was contested by Birks et al. [42].The oldest fossil remains of grey alder in Tab. 3 The clonal structure of seven analyzed populations of Alnus incana.the Scandes are dated to 9.3 ka cal.BP [36,37].A pollen analysis of sediments from a nunatak lake located in central Norway indicated the possible occurrence of A. incana during the LG interstadial [33].The palaeobotanical data from NE Europe suggest its presence in the period of full glacial [29] and LG [44] in the vicinity of the ice-sheet.The presence of grey alder macroremains in the Kreekrak site (SW Netherlands) dating from the beginning of the Holocene may be another argument for the potential persistence of small, scattered populations of A. incana in a periglacial zone of Western Europe [34].Recently, Tzedakis et al. [45] have proposed a model of patchily distributed populations of trees that would be present up to 49° N in the Carpathians and in higher latitudes in the East European Plain during the LGM, which was possible due to a milder climate in the eastern part of the continent at the LGM.

Population
In light of the above palaeobotanical data, populations from the Scandinavian Peninsula (S_1-S_6) and North European Plain (NEP_1-NEP_4) investigated in the present study could derive from micro-refugial populations located on mid-latitudes.However, our results indicate large genetic differentiation among those populations as they were grouped into different clusters.Based on our data we are not able to confirm a single refugium for those populations.According to Mandák et al. [15] grey alder inhabiting Northern Europe is derived from the ancestral populations that survived the LGM in Central Europe.Although our search for macrofossil records of A. incana involved an extensive survey of palaeobotanical literature, we did not find any data on macroremains in Southern and Southeastern Europe for the period of interest.Recent study of Douda et al. [14] indicated for Alnus only low pollen percentages on Southern European peninsulas.We assume, however, that the presence of distinct Alnus pollen in sediments in Greece, Slovenia, and Italy from the LGM and LG reflects the presence of this species in the southern refugia [46,47].Mandák et al. [15] reported a unique genetic diversity present in those southernmost European stands that exemplify the valuable "rear-edge populations" [48].
The Carpathians were probably another important refugium for A. incana.This has been previously suggested by Jankovská et al. [49] on the basis of a pollen analysis of glacial deposits in Slovakia and is supported by the LG macrofossil record from Starunia [50].According to PCoA, Balkan and Carpathian populations are genetically more related to each other and similarly distinct from the remaining European stands.Also, it seems that they had limited impact on current grey alder gene pool that corresponds well to recent colonization scenario presented by Mandák et al. [15].The authors also indicated that the Carpathian populations of A. incana derive from the Balkans, which may explain the close genetic affinity of both groups of populations revealed by PCoA.
PCoA and Bayesian clustering revealed also a split between the Western (WC_1, WC_2) and Eastern Carpathian (EC_1-EC_5) populations of A. incana (Fig. 1 and Fig. 2).A similar picture was shown in Abies alba Mill.[51] and Pinus mugo Turra [52].Additionally, a single population from the Sudetes (BM), a mountain range located between the Alps and the Carpathians, appeared to be genetically distinct from the Carpathian populations (WC_1, WC_2, EC_1-EC_5) and exhibited relatively high genetic diversity (Tab.1).This may indicate the long-term persistence of grey alder in this area, perhaps also during the LGM, as is suggested by recent genetic data [15].A similar geographically-driven pattern for Carpathian and Sudetian populations has been observed for Pinus mugo [53].
We noted substantial genetic differentiation in pairwise genetic comparisons between two populations from the Western Alps (A_2 and A_3) and the remaining European populations; this result was supported by Bayesian clustering as well.PCoA analysis had lower resolution power and mainly revealed the very distinct character of the Western Alpine population, A_2.The Western Alps have been previously defined as a refugium for Pinus sylvestris [54] and P. mugo [55].A continuous pollen presence for A. glutinosa / A. incana-type, starting from the Younger Dryas, has been reported in the northwestern Alps [56].David [28] based on pollen records stated that A. incana refugial populations existed somewhere in the Western Alps.However, a recent study of Mandák et al. [15] based on cpDNA and SSR markers did not reveal the existence of the divergent genetic linage of Alnus species in the Western Alps.

Clonality
In all studied populations, significant clonal structure was present.It proves that clonal reproduction is an important component of grey alder population dynamics.The low genotypes/ramets ratio, observed in the northernmost populations, imply that in suboptimal habitats this mode of reproduction serves as an efficient strategy for survival [13].The largest clone which comprised of 13 ramets was indeed found in population from the Scandinavian Peninsula, S_2.In our opinion, grey alder benefited from diverse advantages related to clonality, which was one of the important life-history traits enabling the species to survive the LGM in the northern micro-refugia, as it has been suggested with macrofossil data presented in this work.
It is generally expected that a lower level of genetic diversity will be found in clonal plants because sexual recombination, the major process generating genetic variation, is limited or temporarily absent.Theoretical predictions and field data show that clonal reproduction may negatively affect the genotypic diversity [57,58].We observed an overall low level of genetic diversity in A. incana, H E = 0.386, which was lowest in Northeastern European populations compared to the remaining stands (H E = 0.366 and H E = 0.432; p = 0.009).The allelic variability assessed with average number of alleles per locus (A) was also rather low with only 3.31 alleles per locus.A similar level of allelic variability at eight SSR loci was found in A. incana (A = 3.88), by Zhuk et al. [59], while in A. glutinosa the authors noted a higher value (A = 5.13).
In our opinion, the cost of prolonged clonality assisting the northern survival of gray alder could be a partial loss of genetic variability, especially in the northernmost populations.In the situation of environmentally enhanced predominance of the vegetative spread, the intraspecific competition may be intensified [60] and locally less adapted genotypes may become outcompeted [61,62] that might lead to reduction of genetic variability; in the most extreme situation a single genet may become fixed.

Conclusions
The combined palaeobotanical and genetic data presented in our work indicate that the glacial and postglacial history of grey alder include existence of the northerly located refugia and differential input of the refugia into the processes of colonization.We believe that clonality may have been crucial in formation of an effective northern refugia and the successful early recolonization of the species.Gray alder populations inhabiting the southern part of the European continent, the so called "rear-edge populations" [48], constitute the important sources of species genetic diversity.These Balkan populations probably represent the oldest, ancestral gene pool of the species and need to be protected from genetic impoverishment.

Fig. 1
Fig. 1 Geographic location and Bayesian clustering of the 24 Alnus incana populations analyzed in the current study; abbreviations of populations names are in Tab. 1.

Fig. 2
Fig. 2 Principal coordinate analysis (PCoA) of the 24 populations of Alnus incana; colors of the populations as in Bayesian clustering and abbreviations as in Tab. 1.