Introduction

Spatial genetic structure (SGS), i.e. the non-random distribution of genotypes in space, may result primarily from isolation-by-distance (IBD) due to limited gene flow or from evolutionary forces such as local selection and single mutation events (Vekemans and Hardy 2004; Troupin et al. 2006). For neutral genes, SGS is mainly attributed to limited gene dispersal via pollen, seeds or other propagules. When significant SGS is found, gene dispersal (σ) can be estimated (Vekemans and Hardy 2004). The main factors that have an influence on SGS are discussed in a large body of literature, and a strong impact of biological characteristics of species on SGS has been detected (Wright 1943, 1946; Knowles 1991; Epperson and Chung 2001; Epperson 2003; Vekemans and Hardy 2004; Scotti et al. 2008). In many forest tree species, SGS has been investigated and in numerous cases no or only weak SGS was found (Epperson 2003) except when tree species with heavy and gravity-dispersed seeds were analysed (e.g. Geburek and Tripp-Knowles 1994). In most conifers SGS is lacking or weak at short distances due to long-distance pollen and seed dispersal, high level of outbreeding and/or strong selection against inbred individuals (Epperson 2003; Vekemans and Hardy 2004; Gapare et al. 2005). An allozyme analysis of Picea mariana at the local level showed a nearly random distribution of genotypes; a significant positive correlation was detectable only in one locus (Knowles 1991). In Norway spruce SGS at low distance classes is similarly expected to be very low. Burczyk et al. (2004) for example found that in a seed orchard of Norway spruce 83% of successful fertilizations were due to pollen that came from outside a 20 m vicinity. The average seed dispersal distance in Norway spruce is approximately 50 m depending on wind velocity (Kohlermann 1950) but has been estimated to be much larger in open areas (Piotti et al. 2009), thus strongly counteracting the spatial aggregation of similar genotypes. In P. abies, SGS has been studied in terms of a morphological trait (colour polymorphism of female conelets), as well as allozyme and molecular markers. Strong SGS on a landscape level was detected in female flower colour, which is probably controlled by one or only few genes. Generally high positive values (Moran’s I up to 0.4) in short-distance altitudinal classes and generally negative values (up to −0.1) in long-distance classes were estimated (Geburek et al. 2007). These findings were explained by IBD as well as by selection due to a different thermoregulatory function of different conelet colours. Contrary to these findings, three geographically distinct spruce populations analysed by allozymes had a very weak SGS or a random genetic pattern (Geburek 1998) while two nuclear sequence-tagged microsatellite markers were positively autocorrelated in two populations (Geburek et al. 1998).

The altitude at which a population grows may affect its SGS. For Norway spruce, knowledge in this respect is still limited. Several influence factors may alter patterns of SGS at high elevations, including reduced seed set (Danielewicz and Pawlaczyk 2007), patchiness of tree distribution and reduced seed shadow overlap (Young and Merriam 1994), increasing incidence of vegetative propagation (Tiefenbacher 1989; Danielewicz and Pawlaczyk 2007), and male-biased sexual function (Obeso 2002; Ortiz et al. 2002). In a number of studies the genetic diversity and SGS within one altitudinal band and potential influence factors have been investigated. One of the few studies that investigated altitudinal variation is the study of Bergmann (1978), who identified clinal variation at an allozyme locus associated with latitude and altitude respectively in six P. abies populations. However, altitude as an influential factor on the mating system, genetic diversity and SGS has not, in our opinion, been given sufficient study. We address the following questions: Is there an increase in SGS with altitude? Do selection and environmental parameters affect the genetic diversity and population structure along an altitudinal transect at the local level?

Materials and methods

Sampling, field measurements and DNA extraction

Three putatively autochthonous naturally regenerated populations were chosen along an altitudinal cline on the Hauserberg (Mayrhofen, Tyrol/Austria; 11°51.874′ E, 47°10.089′ N): population 1 (ca. 800 m asl), population 2 (ca. 1,200 m asl), population 3 (ca. 1,600 m asl). This forested mountain area is mainly stocked with pure Norway spruce forests, but stand structure differs due to management regimes. Selected stands were ca. 60 (population 1), 140 (population 2) and 100 (population 3) years old, with approximately 400, 250 and 300 trees per hectare, respectively. Spatial distance between each of the plots was approximately 1,000 m. Though this spatial distance between plots was relatively small, morphological divergence and adaptation to altitude were evident, e.g. in branching and growth patterns. On each of the three more or less square plots, the XY-coordinates of 150 contiguous trees were mapped and the diameter at breast height (dbh) was measured. A clumped/aggregated spatial tree distribution (data not shown) typical for naturally regenerated stands was indicated for all populations by the Clark-Evans Index (Clark and Evans 1954) and the uniform angle index (Gadow et al. 1998). Needle samples of these trees were collected by shotgun and stored at −20°C until DNA extraction. Genomic DNA was isolated using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany). Purified DNA was stored until use at 4°C.

PCR, genotyping and sequencing

Six expressed sequence tagged–simple sequence repeats (EST-SSRs) were used (Rungis et al. 2004): WS0019.M09, WS0022.B15, WS0023.B03, WS0073.H08, WS00111.K13 and WS00716.F13. All PCRs were done on a 96-well Piko Finnzymes Thermal Cycler (Finnzymes Oy, Espoo, Finland). The PCR mix for each sample contained 1x reaction buffer, 200 μM dNTPs, 1.5 mM MgCl2, 0.2 μM of each forward and reverse primer, 0.25 U Invitrogen Platinum Polymerase (Invitrogen, Carlsbad, USA) and ddH2O to reach a final volume of 10.0 μl. A volume of 0.7 μl of genomic DNA (ca. 100 ng) was finally added to the reaction. PCR profile was 3 min as initial denaturation at 94°C, followed by 30 cycles of denaturation at 94°C for 20 s, annealing at 53°C for 30 s and an extension step at 72°C for 30 s. Fragments were analysed on a Beckman Coulter CEQ8000 Genetic Analysis System (Beckman Coulter, Fullerton, CA, USA). Five homozygotes of each marker carrying different alleles were sequenced to verify the SSR nature of the markers. Prior to sequencing, PCR products were purified using the QIAquick PCR Purification Kit (Qiagen, Hilden, Germany). Direct sequencing was performed by a commercial provider using the same reverse primers.

Data analysis

GenAlEx v6.3 (Peakall and Smouse 2006) was used to calculate standard diversity indices for each population, including number of polymorphic loci, number of different alleles (N a), effective number of alleles, private alleles, observed (H 0) and expected heterozygosity (H e). Null allele frequency estimates were obtained with GENEPOP v4.0 (Raymond and Rousset 1995). Selective neutrality of the EST-SSR markers was assessed using the Ewens-Watterson neutrality test implemented in ARLEQUIN v3.5 (Excoffier and Lischer 2010). Exact tests for Hardy-Weinberg proportions and F-statistics were calculated with GENEPOP v4.0 (Raymond and Rousset 1995), and tests for linkage disequilibrium were performed in FSTAT v2.9.3.2 (Goudet 2001).

For the analysis of SGS, pairwise kinship coefficients (F ij ; Loiselle et al. 1995; Hardy 2003) were estimated using SPAGeDi v1.2 (Hardy and Vekemans 2002). These coefficients were calculated for each population separately for 4, 8 and 12 distance classes respectively under even sample size option and for five predetermined distance classes (max. distance 20, 40, 60, 80 and 100 m). Kinship coefficients were also calculated for the pooled data. To test the significance of observed values of F ij , 95% lower and upper confidence intervals were calculated with 10,000 permutations. Intensity of SGS was evaluated by the Sp statistic (Vekemans and Hardy 2004), Sp = b/(F 1 –1), where b is the regression slope of F ij on d ij (spatial distance), and F 1 is the average kinship coefficient between individuals in the first distance class.

Fixation indices and SGS were also calculated for each population after stratifying the sample trees into age classes using dbh as age proxy based on the relationship among age, diameter and height in Norway spruce (Mehtatälo 2004). Due to the fact that some dbh values were not present in all stands the chosen classification differs from stand to stand: trees below a dbh of 30, 40 and 30 cm were classified as “young” in populations 1, 2 and 3 respectively, and similarly trees above 44, 54 and 48 cm dbh were classified as “old”; trees between these values formed the intermediate dbh group and were not included in the calculations due to uncertainty in aging. For each of these dbh groups SGS was characterised as described above, however, locus WS0023.B03 (highest frequency of null alleles) was excluded. Significance of differences between observed values of fixation indices was tested by a Wilcoxon rank test.

Results

All sequences derived for locus verification contained a microsatellite motif. Interestingly, the sequence obtained for locus WS0023.B03 did not match the published sequence of Rungis et al. (2004). However, it contained a microsatellite identical to the sequence of a clone of Picea sitchensis (GenBank EF084044). All six utilized SSRs were highly polymorphic (Table 1). Over all loci and populations, a total of 146 alleles were observed. No linkage disequilibrium between loci was detected after Bonferroni correction (adjusted P <0.001). The highest number of alleles (34) was detected in WS0019.M09, while WS0073.H08 was the least polymorphic locus with only eight alleles. The number of private alleles ranged from seven to eight per population (Table 1). Their frequency ranged from 0.003 to 0.013. The average observed heterozygosity (H 0) was very high and ranged from 0.846 (±0.04) to 0.850 (±0.04) among populations. Most H e values exceeded H 0, and several deviations from Hardy-Weinberg equilibrium (HWE) were detected (Table 1). The main reason for deviations from HWE was probably the occurrence of null alleles, which was exemplified by amplification failure with primer WS00716.F13 in one individual from the third population. Null allele occurrence was significant in four of the primers in the GENEPOP calculations, and their frequency was estimated to lie in the range 0–0.067 (Table 1). Several deviations from neutral expectations were indicated by the Ewens-Watterson neutrality test; however, these did not consistently affect the same loci (Table 1). F ST over all populations was very low (0.002). The inbreeding coefficients (F) indicated small deviation from random mating in all three populations, decreasing significantly (P <0.05) with age (Fig. 1).

Table 1 Per locus and mean standard genetic diversity indices for the three and pooled Norway spruce populations at the Hauserberg site (Tyrol/Austria). Per population, 150 trees were genotyped (450 total)
Full size table
Fig. 1
figure 1

Inbreeding coefficient (F) calculated for two dbh/age classes and three populations of Norway spruce from different altitudes (800, 1,200 and 1,600 m asl)

Full size image

When pairwise kinship coefficients for eight distance classes under even sample size option were plotted (Fig. 2), no significant positive SGS was indicated in populations 1 and 2. Only one significant negative coefficient was estimated in a greater distance class of population 1. A marginally significant positive autocorrelation was detected at the lowest distance class in population 3 (high altitude). All slope parameter (b) values were non-significant. Non-significant patterns were also obtained for the other distance classes under even sample size option and for five predetermined distance classes (data not shown). When dbh was taken as a proxy for tree age, only in one case a significant positive spatial autocorrelation was obtained, namely for the smallest diameter class of population 1 using four distance classes with the even sample size option (Fig. 3). In all other dbh classes (small/young; large/old) no significant SGS was observed. The analysis of the pooled data showed a significant autocorrelation in the first distance class for 4, 8 and 12 distance classes, although Sp was very low (0.0005). All b-values for the full range of distance classes were highly significant (P <0.001). The results for four distance classes are shown in Fig. 4. The estimation of gene dispersal σ and effective neighbourhood size was not possible because no significant slope value was obtained when restricting the regression to σ > d ij  > 20σ (cf. Hardy et al. 2006).

Fig. 2
figure 2

Results of spatial genetic structure analysis in three Norway spruce populations along an altitudinal transect at location Hauserberg/Tyrol. Loiselle’s kinship coefficient is plotted for eight distance classes with respective 95% confidence intervals (dotted lines)

Full size image
Fig. 3
figure 3

Results of spatial genetic structure of young (dbh <30 cm) trees of Norway spruce in population 1 (800 m asl). Loiselle’s kinship coefficient (F ij ) plotted for four distance classes with respective 95% confidence interval (dotted lines)

Full size image
Fig. 4
figure 4

Results of spatial genetic structure analysis in Norway spruce at location Hauserberg/Tyrol for the pooled data of the three analysed populations (150 trees per plot). Loiselle’s kinship coefficient (F ij ) plotted for four distance classes with respective 95% confidence interval (dotted lines)

Full size image

Discussion

The genetic structure and diversity along an altitudinal transect in Norway spruce and, in particular, differences in SGS have been the topic of this study. Although weak but significant SGS was detected in rare cases, our results indicate that the hypothesis of an increase in SGS in populations along an altitudinal cline has to be rejected. Against our expectations changes in the mating system of Norway spruce due to elevation were not reflected in measures of SGS. Spatial genetic structure is influenced by numerous ecological and evolutionary factors, e.g. mating system, dispersal of pollen and seed, neighbourhood size and demographic processes. Our results indicate that SGS in Norway spruce is mainly influenced by effective gene dispersal, strengthening previous findings (e.g. Piotti et al. 2009). Spruce pollen is very effectively dispersed due to low average pollen grain weight (111 × 10−9 g) and low sedimentation velocity (5.6 cm s−1) (Eisenhut 1961). In many cases, Picea pollen is effectively distributed over many kilometres (Xie and Knowles 1994). Recently, Piotti et al. (2009) have shown that gene flow is extensive. In an isolated Norway spruce stand, only 11.1% of the juveniles were sired from local trees, and mean effective seed dispersal ranged from 40 m to more than 800 m. In more than 90% of the cases, seed flow exceeded a distance of 100 m, indicating gusty winds during seed fall (cf. Kohlermann 1950). On the other hand, reproductive success was highly skewed in the study of Piotti et al. (2009).

Although a significant kinship coefficient was found in the first distance class from the highest elevation in our study, such a weak spatial pattern is certainly not biologically meaningful. Probably more pronounced SGS at the highest population is lacking because trees still predominately reproduce sexually at 1,600 m asl and seed shadows were still overlapping. In addition, strong winds may increase pollen and seed dispersal distances at high altitudes even if fewer mothers are effectively contributing (cf. Piotti et al. 2009). Our results resemble those of Leonardi et al. (1996) who studied a naturally regenerated uneven-aged P. abies population in the eastern Italian Alps. They mostly found a random distribution of isozyme genotypes. Weak SGS was observed in our study in younger trees (small dbh) of the population at the lowest altitude where tree age was also the lowest among plots. Ontogenetic effects on SGS are likely when natural selection acts against inbred and other consanguineous individuals. This was also found in a natural stand of Abies sachalinensis, where significant SGS was present in log regeneration and decreased sharply in older age classes (Lian et al. 2008). SGS will also decrease rapidly over time due to factors such as density-dependent mortality and micro-site selection (González-Martínez et al. 2006). Generation overlap and the influence of selective processes at the population level (e.g. overdominance selection on a few genes) likewise contribute to the decrease in SGS in natural populations (Doligez et al. 1998). In contrast to the results within plots, weak but significant SGS was found when these were treated as one population, albeit the Sp value was very low (0.0005), conforming to an IBD pattern. No attempt was made to infer gene dispersal parameters due to a lack of autocorrelation in the higher distance classes. Future studies in more and more widely spaced plots in undisturbed (if available at all) spruce stands would be needed to get unbiased indirect estimates of gene flow parameters in Picea abies. This also highlights the methodological problems in gene flow studies of the species, problems that are shared with other common stand-forming conifers, e.g. Pinus sylvestris (cf. Burczyk and Koralewski 2005; Robledo-Arnuncio et al. 2004).

For the interpretation of our results, stand history and possible human impact also have to be considered. McCauley et al. (1996) described two principal ways of gene flow: (1) dispersal among established populations and (2) gene flow from an already established population into a new habitat. In connection with this finding, Knowles et al. (1992) investigated the effect of differing establishment histories on SGS in two tamarack (Larix laricina) populations. Significant SGS was observed in the population established from a clear-cut forest site (with recruitment from the felled stand) while the population established on abandoned farmland did not show significant SGS. Harvesting or natural disturbances create open areas for colonization as described in Knowles (1991), Leonardi and Menozzi (1996) and Gömöry et al. (2006). It takes several generations of natural regeneration to build up a pattern of significant spatial autocorrelation as SGS will increase over generations due to limited pollen and seed dispersal within stands (Epperson 2003; Vekemans and Hardy 2004). This is also supported by more recent findings (Troupin et al. 2006). Marquardt and Epperson (2004) evaluated the spatial and population genetic structure based on nuclear microsatellites in old growth and second growth populations of Pinus strobus. The results showed a weak genetic structure in the old growth forest and a random distribution in the second growth forest. Similarly, Scotti et al. (2008) found significant SGS in P. abies with a mitochondrial marker within stands that had been gradually regenerated, but not at the edge of the forest where regeneration had been established in a short period. The stands investigated in this study originated from natural regeneration which was common practice in the Tyrol region at the time of stand establishment; whether seedlings had established before or after the cutting of the mature trees remains unknown, therefore we cannot rule out that regeneration took place after a clear cut, which would also help to explain the observed patterns.

A high number of alleles and extremely high levels of heterozygosity were observed, supporting previous findings in P. abies (Tollefsrud et al. 2009). The study of Rungis et al. (2004) showed that genomic SSRs (17 loci; mean H 0 = 0.72) are more polymorphic than EST-SSRs (25 loci; mean H 0 = 0.65). Hence, it may be presumed that employing conventional nuclear SSRs in the studied material would have resulted in even higher estimates. Although few deviations from neutral expectations were detected in the Ewens-Watterson neutrality test, these were neither consistent among markers nor populations, and therefore neutrality for the chosen markers can be assumed. A low level of inbreeding and a locus-dependent slight heterozygote excess or deficit were observed. In general, excess of homozygotes compared to HW proportions may be either the result of a Wahlund effect, consanguineous mating, drift, selection or may be an artefact due to null alleles. In our material only a Wahlund effect can be excluded, since samples were derived from autochthonous populations which are likely in mating contact as suggested by our results. All other factors may play a role; however, we presume that null alleles are most significant. This seems justified as we observed a null homozygote in our dataset, and also relatively high frequencies of null alleles were detected in previous studies of the same species. For example, Yazdani et al. (2003) found a high number of null alleles with SSRs in controlled crosses of Norway spruce, and frequencies of null alleles ranged from 0.042 to 0.288 in populations from northern Europe (Tollefsrud et al. 2009), while we estimated up to 7% for one marker. Our results show that null alleles in Norway spruce are present even with EST-SSRs, although apparently at lower frequency (cf. Rungis et al. 2004). Inbreeding coefficients decreased with age (dbh), conforming to the expectation that with increasing age heterozygous individuals have a selective advantage as demonstrated in Pinus sylvestris (Burczyk 1991), Fagus sylvatica (Raddi 1993) or Carapa procera (Doligez and Joly 1997).

Among the studied populations, genetic characteristics (allele frequencies, effective and private alleles and heterozygosity) similarly resulted in very low though significant pairwise F ST values (0.0004–0.0035) and correspondingly high estimates of migrants per generation (N m = 60–100). Thus, differentiation at these neutral markers among the populations was negligible, and no biologically meaningful differentiation was present. Therefore it is valid to consider the three populations as part of a single population strengthening our assumption of a natural origin and immigration mode. Of course, these results do not imply that among the analysed stands there is no phenotypic divergence, which probably relies on selection for a few genes (Savolainen et al. 2007). Previous studies using SSRs in Norway spruce have shown that the among-population variation at neutral markers is generally low, even at a large scale. Scotti et al. (2006) calculated a low F ST (0.00–0.11) based on nuclear SSRs in four Italian populations. In 37 northern European populations of Picea abies, Tollefsrud et al. (2009) found F ST values in the range of only 0.02–0.03 in nuclear SSRs, while in a range-wide study based on allozymes a slightly higher value (F ST = 0.052) was found (Lagercrantz and Ryman 1990). Even populations from Central and Northern Europe representing two divergent lineages of the species (Tollefsrud et al. 2008) showed only moderately divergent allozyme markers (F ST = 0.12; Muona et al. 1990).

In Austria, the macrospatial allozyme patterns of 29 high-elevation populations of Norway spruce revealed that Norway spruce populations from the same altitudinal band are genetically very similar (F ST = 0.012; Geburek 1999). The results from the local scale of this study indicate that the genetic structure of Norway spruce populations along an altitudinal cline is very similar as well. A related result was obtained for Pinus sylvestris by Fournier et al. (2006). A high level of intraspecific genetic variation is linked to a wide distribution area, characterised by differing environmental conditions typical for Picea abies (Bergmann and Gregorius 1979). While Müller-Starck et al. (2000) always found low allelic richness in the highest populations in three altitudinal clines, our data do not support such a finding. Gugerli et al. (2001) pointed out a similar lack of genetic differentiation in Norway spruce across elevations as observed in their study on a small geographical scale. The high genetic diversity of Norway spruce found in this and previous studies underlines the high adaptive potential and plasticity of this tree species. In conclusion, our analysis provided insights into the genetic structure along an altitudinal transect in Norway spruce at the local level. Gene flow from adjacent populations was very high, overcoming effects on SGS due to potential altitude-dependent changes of the mating system.