Differences in wood properties of Picea abies L. Karst. in relation to site of provenance and population genetics

Chemical composition

The chemical analysis of powdered wood allowed the quantitative assessment of the differences in chemical composition of investigated materials and is summarized in Table 2. The values of individual chemical components of Norway spruce wood for all locations are within the reference values for this species as reported in the literature (Tjoelker et al. 1998). The quantity of cellulose was ∼45.5% and was similar in all samples. The holocellulose content ranged from 67% to almost 72% and was the most varying component among all. Significant differences (almost 2.5%) in lignin content were observed between the material from northern Poland (➌) and Italy (➏). A positive Pearson correlation value was found between site altitude and lignin content (r=0.727). Moreover, 53% of the variation in the lignin content was explained by the increase of altitude (R²=0.529) at the significance level of 90% (P=0.101) through a consecutive regression analysis. This finding is in agreement with Gindl et al. (2001), who observed that trees growing at higher altitudes compensate for the thinner cell walls with increased lignin content to maintain mechanical integrity of the xylem.

The variation in extractives amount did not exceed 1.3% within all tested materials. These contents were inferior to those published in the literature (Tjoelker et al. 1998). Indeed, the extraction was carried out only with ethanol, while most of the literature references were based on ethanol-benzene mixture extraction. Slightly higher differentiations were noticed for substances soluble in 1% NaOH and varied in the range of ∼1.8%. A positive correlation between the extractive components soluble in hot H₂O, and the latitude was observed (r=0.656, R²=0.431, and P=0.157). Such results are in agreement with findings of Hakkila (1968), who observed that the average proportion of extractives slightly increase with increasing latitude in pine heartwood. However, the average total extractives’ proportion of sapwood remains constant. The most recent research published by Caron et al. (2013) state that yield class, latitude, longitude, and thinning do not have an impact on the total amount of acetone-soluble extractives.

The quantitative data related to the chemical composition of wood can be utilized for studies of taxonomy (Sokal and Sneath 1963; Sneath and Sokal 1973) and/or ecology (Legendre and Legendre 1983; Pielou 1984). The average “taxonomic” distance due to chemical content variations within all investigated populations is shown as a cluster plot in Figure 2a. A clear clustering of samples from northern populations of Finland and Estonia is noticed. All the other populations are grouped in the second cluster and reflect the geographical distances between sites.

Figure 2

Cluster plots of the taxonomic distances among six populations of Picea abies on the basis of chemical (a) and physical (b) wood characteristics.

It should also be noted that traditional “wet chemical analysis”, as discussed here, possesses a relatively large error margin and is not capable of providing refined qualitative information. Instead, clear evidences of variations to the molecular structure (as scrutinized with near-infrared spectra), due to the combined effect of tree provenance, environmental factors, and silvicultural practices, were previously reported by Sandak et al. (2011).

Physical properties

The summary of investigated physical wood properties is presented in Table 3 and Figure 3. Average wood densities are in the range of 0.39–0.52 g·cm^-3 and correspond to values attributed to Norway spruce (Figure 3a and b). The samples from Estonia (➋) and Finland (➊) possess the highest density values, which confirms the hypothesis of Zobel and van Buijtenen (1989) that growth rate and density of conifers are negatively correlated. The lowest densities were measured on the Italian wood originated from the Dolomite Mountains. It is a typical characteristic of resonance (tone) wood. Such wood, besides having superior acoustical characteristics, often possesses a fairly steady narrow or narrowing growth, with an average ring width of ∼1 to 2 mm. Other characteristic of tone wood is its low density associated with stiffness across the grain. Low ambient temperature and short grooving season negatively affect LW formation (Splechtna et al. 2000). It follows the supposition that the wood of trees growing at higher elevations has lower density (and LW width) compared to that of low altitudes. As a consequence, Norway spruce from the Paneveggio forest ➏ has a particular microstructure (low density and high stiffness at the same time) determining its superb acoustic properties (Nocetti and Romagnoli 2008).

Figure 3

Wood density (a), X-ray measured wood density (b), average ring width (c), and late wood ratio (d) of different provenances of Picea abies.

The X-ray radiograms and density profiles, as presented in Figure 4, demonstrated slightly different density patterns for each location. Relatively low values of density are noticed for samples from Italy ➏, Poland ➎, and Poland ➌. Accordingly, the reaction wood was most frequent in the samples originated from mountainous areas (Italy ➏ and Poland ➎). The zone of juvenile wood had relatively high density in most samples. A tendency to a gradual increase of density with the age of tree can also be noticed on several density profiles.

Figure 4

Examples of x-ray radiograms and density profiles for trees of different provenances: Finland ➊ (a), Estonia ➋ (b), Poland ➌ (c), Poland ➍ (d), Poland ➎ (e) and Italy ➏ (f). (Note: arrows indicate locations of the reaction wood.)

The size of logs (log radius) was not correlated with the age. The wood samples from the north of Poland ➌ were the youngest and had the highest standard deviation of the ring width. The moderately young trees from southern Poland ➎ were relatively large in diameter. On the contrary, Finnish samples with the smallest diameter were extracted from the oldest trees. The small annual ring widths in both, Scandinavia and high mountainous areas, were caused by a short vegetation period and severe climate, as well as presumably soil fertility and silviculture. In the case of samples from southern Poland ➎, the mean width of annual rings was more than twice as great in comparison to that from Finland. It was also noticed that even if two locations in Southern Poland (➍ and ➎) were close to each other (the distance of ∼5 km), relatively large differences between the physical characteristics were noticed. The wood in ➍ was taken from a steep mountain slope at the level of ∼800 masl. Samples of ➎ were collected from the flat land (∼500 masl) at the foot of the mountain. Narrower rings in wood from ➍ can be explained, therefore, by more severe climatic conditions and natural environment (steep slope – up to 30% – and less fertile soil-bonitation III).

Wood from Finland ➊ had the highest share of LW (Figure 3d). It is due to the short vegetation season and a more severe climate. Analogical observations were made for spruce wood from nearby Estonia ➋, Italy ➏, and Poland ➍. Splechtna et al. (2000) reported, however, the decline in the percentage of LW and the mean ring density with increasing elevation. In this context, contradictory information is published by different authors. Zobel and van Buijtenen (1989) reported that LW genesis is dependent on the water availability. Irrigated trees have a higher percentage of LW. This statement might be particularly valid for spruce form Italy where the precipitation was the highest (1260 mm annually). The wood from Poland ➌ and ➎ had the lowest LW ratio. A clear tendency to increase the share of EW with the increase of the ring width was observed. An analogous remark was reported for spruce by Zobel and van Buijtenen (1989), among others.

Robust discriminant analyses were performed based on density, LW ratio, and average ring width. The summary of results is presented in Figure 5 and shows a great range of differentiation in terms of site of origin. The first two canonical scores enabled a clear separation of samples from Italy ➏, Finland ➊, and the other provenances. The analysis of variance on the physical properties of samples, as shown in Table 3, evidenced a strong influence of the site on the average ring width (P=0.003), whereas both density and LW ratio were less correlated to the site location. It was also found that the age of trees affected the wood properties, especially the LW ratio (P=0.019).

Figure 5

Canonical scores plot resulting from the robust discriminant analyses of density, late wood ratio, and ring width.

The “taxonomic” distances (a cluster analysis) due to the physical wood characteristics are shown in Figure 2b. Three clusters can be defined here, with Poland ➌ being the most distant from other locations. It was found that wood from Finland ➊ and Italy ➏ had relatively similar characteristics.

Wood density and LW ratio were correlated to the geographical position of the sampling. It was found that 83% of the variation in wood density were explained by differences in the latitude (regression P=0.010), whereas 66% by the altitude above the sea level (P=0.048). Wood density increased at higher latitudes (r=0.913) and was inversely correlated to the altitude (r=-0.815). The regression between LW ratio and latitude was statistically significant (P=0.035), while latitude explained 71% of the total variation. LW ratio was directly correlated with the increase in latitude (r=0.842). On the contrary, the effect of altitude on this propriety was lower (r=-0.363) and indicated the decrease in the LW ratio with the increase in the altitude.

Both geographical latitude and site altitude strongly affect climatic conditions and the tree growth environment, even if they have influenced wood density and LW ratio differently. Regulation of tree growth is controlled genetically, in response to environmental cues, such as day length, light quality, and/or ambient temperature. Heide (1974) showed that formation of the LW tracheids possessing thick walls and narrow lumens took place upon exposure to a short-day photoperiod. The photoperiod control can in some cases be omitted; the growth cessation can be induced even in long photoperiods through a combination of high day and low night temperature. Light and temperature are important environmental factors affecting all physiological processes as the annual growth of the tree is a continuum. These responses involve environmentally-dependent modulations of the expression of specific genes and enzyme activities (Asante et al. 2011). The opposite trends observed for latitude and altitude can therefore, be explained by the different modulation imposed on the physiological processes from day length-light-temperature variables.

Only 2% and 25% of the total variance, for altitude and latitude, respectively, explained the variation of average ring width. Ring width was only moderately correlated (negatively) to both latitude (r=-0.496) and altitude (r=-0.136). It can be assumed, therefore, that genetics (besides silviculture, stand density, soil fertility, age, among others) may affect the ring width within the six populations investigated. The influence of genetic differentiation was analyzed, therefore, in relation to the ring width, but considering also other wood properties of interest.

The SIMINT procedure of NTSYSpc was used for grouping the studied populations by physical characteristics, in an analogy to the chemical analysis discussed before. The resulting dendrogram is shown in Figure 2b. It is clear that the clustering differs significantly from that obtained by the analysis of chemical components (Figure 2a). The two-way Mantel test (Mantel 1967) was used for the statistical comparison of matrices used for clustering. The results showed very poor fit (r=-0.011, t=-0.044, P=0.516), indicating that differentiation of the six populations studied were related to other (than physical properties) factors. Moreover, such differentiations were related to chemical and physical properties independently.

Genetic variables

The number of alleles N, gene diversity GD, and fixation index F_IS were calculated for each locus and population (Table 4). The number of alleles per locus obtained in this research is lower than those described by other authors for P. abies. Acheré et al. (2005) found alleles per locus for the 25 microsatellites studied in the range of 5–55 (25 in average). Maghuly et al. (2006) detected 27 (in average) alleles within five loci. Allelic richness, total number of alleles in a population, is also reported in Table 4. The values ranged from 31 (Poland ➌) to 54 (Estonia ➋ and Italy ➏). The slightly lower number of scored alleles within this study could be explained by the dimension of the experimental matrix (plants/number of loci). The possibility of detecting new alleles increases with increasing number of analysis and the polymorphism of each locus.

F_IS index is a coefficient of consanguinity due to not-usual mating. The high values at each locus indicate a significant excess of intrapopulation homozygosity. Some differences in alleles among population were present in both locus UAPCA24 and SpAC1B8. Finland ➊ and Estonia ➋ show lower values in gene diversity and F_IS for these loci. Italy ➏, however, had the lowest values of F_IS at locus EAC1D10. For Finland ➊, the lowest value was found for overall genetic diversity (F_IS=0.497).

According to Maghuly et al. (2006), populations of P. abies from high elevation exhibit more genetic variation than samples from middle and low elevations. This was not confirmed in the present research as the stands located at the lowest (Estonia ➋) and highest (Italy ➏) elevations possessed 54 alleles.

The values of pairwise differentiation F_ST among populations, as summarized in Table 5, were very low especially in the case of Italy ➏ vs. Poland ➍ and ➎, as well as for Finland ➊ vs. Estonia ➋. It was also confirmed by UPGMA cluster analysis from genetic distances among populations. The UPGMA dendrogram shows clustering of populations independently from the country of provenance (Figure 6). Poland ➌ population is the most divergent from the others. Two other clusters can also be distinguished, the first, grouping the northern populations of Estonia ➋ and Finland ➊, the second, grouping Poland ➎ and ➍ and Italy ➏. The fact of clustering three out of six populations in a single cluster (Italy ➏, Poland ➍ and ➎) indicates a moderate level of genetic separation due to geographical provenance.

Figure 6

UPGMA cluster plot of the genetic distances among six populations calculated from SSR data.

The above clustering reflects the statistical differentiation of the studied populations. The P values resulting from the exact G-test are summarized in Table 6, where all the population pairs are compared at the genotypic level (number of shared genotypes). Each pair differs statistically when P≤0.05, assuming 95% level of confidence. Only Poland ➌ differed from all the other regions. Estonia ➋ was dissimilar from Finland ➊, as well as Italy ➏ was statistically different from Poland ➍ and ➎. No significant difference was noticed between the genotypic structures of samples from Estonia ➋ and Poland ➍. The set of the above observations is a corroboration of the five populations grouping within the same cluster as presented in Figure 6.

The existence of geographical trends of genetic variation over the European Norway spruce natural range was observed by classical statistical methods (Borghetti et al. 1988; Lagercrantz and Ryman 1990). Investigations carried out by the genetic markers revealed differentiation between southern and central European populations (Vendramin et al. 2000) and identified 16 genetic zones (Bucci and Vendramin 2000). Other studies focused more specifically on the area between the northern and southern natural ranges, defined as European disjunction or “spruceless zone” (Nowakowska 2009). The zone covers the Middle Polish Plains where P. abies has a recently sparse distribution. Dering and Lewandowski (2009) described this area as a hybrid for Norway spruce originating from Carpathian and Russian Weichselian refuges. The results reported here show that trees from Poland ➌ belong to a genetic group of spruce well separated from all the others. This has been confirmed by the findings of other researches stating that Baltic area of Norway spruce in Poland is a part of the Scandinavian area. The southern part of Poland, however, belongs to the Hercyno-Carpathic zone, even if there are stands with Norway spruce trees being introduced by German and/or Austrian foresters before the 1^st World War.

The result of comparing the matrices of genetic similarity with the average ring width by a two-way Mantel test reveals that both sets of data are poorly correlated (r=-0.007, t=-0.023, P=0.491). No statistical correlation was also found between the matrices of genetic similarity and chemical contents (r=0.208) as well as physical properties (r=0.323). Thus, it was not possible to identify any marker, depending on provenance and associable to wood properties. The presence of alleles shared among all the provenances can be deduced as well from the Lynch’s similarity analysis dendrogram (Figure 7). All the cases, each representing a single analyzed tree, are not clustered in relation to the tree provenance. Individual samples from each of the six locations are dispersed throughout the entire dendrogram because of the large intrapopulation variability. Several studies have shown that genetic variation is relatively high and that the majority is partitioned within (rather than between) populations (Cottrell et al. 2003; Acheré et al. 2005). Maghuly et al. (2006) analyzed genetic diversity in different subpopulations of Norway spruce in Austria by nuclear SSRs finding more genetic variation and heterozygosity within populations than among populations. The geographical differentiation of genes is relatively low due to both the mating system and the high migration rate (Taberlet et al. 1998).

Figure 7

UPGMA dendrogram of the single plants from the six origins, produced by Lynch’s similarity coefficient.

Adaptive traits in forest trees have often both evolutionary and agronomic relevance with practical importance in lumber and pulp production (Peter and Neale 2004). Molecular approaches to genetically dissecting wood properties (and other adaptive traits in forest trees) have focused on quantitative trait locus (QTL) mapping (Casasoli et al. 2006). Although given the large genetic-to-physical distance in most conifers, identification of specific genes responsible for phenotypic variation in these species is unlikely via QTL mapping. Association genetics has become a favored genetic approach for studying quantitative traits in many organisms. It is taking advantage of historical recombination to identify trait-marker relationships on the basis of linkage. The essence of applying genetic approach is the identification of statistical associations between variation in relevant phenotypic traits and allelic polymorphism in known genes, although it is generally difficult to demonstrate that a particular allelic variant is, indeed, causally related to a phenotype (Weigel and Nordborg 2005). Association genetics has been recently applied in several studies on trees (González-Martínez et al. 2007; Denis et al. 2013; Guerra et al. 2013). The present study is based on only four microsatellites and low number of markers. Nevertheless, these were sufficient to discriminate plants among and within the studied populations. Only two couples had the same fingerprinting as visible in Figure 7.

Authors are aware that genetic variability of traits can be best studied in planned experiments where several genotypes are grown in the same site and climatic conditions, with an adapted experimental design. Our study cannot be used for direct association between molecular and morphological features also because molecular approaches based on neutral genetic diversity as microsatellites are not directly connected to selection. These markers can be used (in combination with others) for quantitative mapping of trait loci. Moreover, it is generally recognized that any information about variation at the population level would help in assessing the importance of evolutionary processes. Such knowledge (genetic parameters based on molecular markers for each genetic zone) might also be of great commercial potential for selection and breeding activities, as well as for forest certification.