Annual Variations in Norway Spruce Xylem Studied Using Infrared Micro-spectroscopy

: In temperate environments, ring width, cell size and cell wall thickness within the xylem of trees are known to be affected by climate conditions. Less is known about the effect of climate conditions on the chemical characteristics of the xylem, which are important for the susceptibility of the tissue towards fungal infections as well as for the degradability of the material within the forest ecosystem. We explored the use of infrared microspectroscopy to investigate the possible effects of temperature and drought on the relative amount of cell wall biopolymers, i.e. the ratios between cellulose, hemicellulose and lignin in the earlywood xylem cell walls of Norway spruce ( Picea abies (L.) Karst.) in temperate forests. Drought and warm temperatures were signiﬁcantly correlated to the hemicellulose to lignin ratio of the earlywood formed the following year, perhaps due to a reduced amount of stored resources being available for xylem formation.


Introduction
Current climate scenarios for northern Europe in the later part of the 21th century predict longer growing seasons, dryer summers and warmer autumns [1,2]. It is well-known that spring and/or summer droughts affect conifer tree growth negatively in a temperate climate [3][4][5][6][7], and that cambium phenology is related to climate [8,9]. It is also known that xylem characteristics, like for example cell size and cell wall thickness, are related to ecological factors, of which climate is one [10][11][12][13][14][15][16][17][18]. However, less is known about the possible effects of climate on the biopolymer composition of the xylem formed. Xylem cell dimensions and cell wall thicknesses affect physical wood properties such as density [19] as well as tree physiology via e.g. water conductance [20]. Cell wall chemistry on the other hand is important for the biodegradability of the wood, as for example a high lignin content generally correlates to high recalcitrance of plant biomass [21]. Based on this knowledge, it is possible that climate indirectly might affect both the susceptibility of the living tree to fungal attacks as well as the biodegradability of dead wood within the forest ecosystem. We consequently find it relevant to study the possible relations between climate and xylem cell wall biopolymer composition.
In temperate conifers, the chemical composition differs between earlywood and latewood: Earlywood generally contains more lignin and less cellulose and hemicellulose than latewood expressed as weight percent of dried wood [22,23]. This is an effect of differences in xylem structure, where earlywood consists of larger and more thin-walled cells, and thus contains more lignin-rich middle lamella per gram of dry matter. The difference in cell wall chemistry between earlywood and latewood implies that the amounts of earlywood and latewood present within each ring may give rise to confounding effects rather than climate signals only in studies where the chemical composition of whole tree rings were analyzed. Therefore, microspectroscopy appears to be a good alternative in studies where the aim is to clarify the effects of temperature and precipitation on xylem cell wall composition, as these techniques offer the spatial resolution required [24]. This approach seems not to have been attempted until recently [25]. Chemical imaging studies of conifer tracheid secondary cell wall have shown significantly higher relative lignin content in earlywood than latewood [26,27], emphasizing the need to study these two wood types separately in order to minimize the noise when attempting to link cell wall chemistry to climate signal.
In the present work we used Fourier transform infrared (FT-IR) microspectroscopy to test the hypothesis that drought and temperature of the current and preceding growth seasons affect the chemical composition of earlywood from P. abies. Studies of this species grown under temperate conditions have indicated that a warm autumn is associated with reduced tree vitality the following year [7,28], perhaps due to a lower amount of non-structural carbohydrates being stored over the winter. Presumably this signal would be strongest in the earlywood, and therefore, FT-IR microspectroscopy was applied to track variations in relative amounts of cellulose, hemicellulose and lignin in earlywood cell walls.

Experimental Site and Xylem Samples
The wood discs used as sample material came from an experimental site in a plantation forest in Frederiksborg, Denmark (latitude 55 • 57'21.3" N, longitude 12 • 21'14.1" E) established in 1964 with a mean annual temperature of 8.2 • C and an annual precipitation of 605 mm (period 1961-2012) [29]. The former use of the area was as cropland and the soil texture is sandy loam with a subsoil clay content of 23% [30]. The field trial includes 10 non-native conifer species and two native broadleaved species. Stem discs from 1.3 m above ground were acquired during the winter 2012-2013 from three P. abies co-dominant trees by thinning. The discs were oven dried at 103 • C prior to analysis. For further information, please see [7].

Temperature and Drought Data
Average monthly temperature, precipitation and potential evapotranspiration were obtained from the 0.5 • grid observational data from CRU TS v.3.22 from 1997 to 2012 [31]. Monthly drought index (DI) was estimated as monthly precipitation minus monthly potential evapotranspiration [32,33]. Low DI corresponds to dry conditions.

Silviscan Density Measurements and Infrared Microspectroscopy
Two adjacent radial strips covering pith to bark and approximately 5 mm wide were cut from each of the three discs. One strip was characterized by use of the instrument Silviscan [34]. Using this instrument, density was measured by X-ray densitometry with a spatial resolution of 25 µm. Within each ring, positions having a wood density within 0-20% of the span from minimum to maximum wood density recorded for the ring was assigned to earlywood, the part within 80-100% of the span was assigned to latewood, and the remaining part was assigned to transition wood [35]. The Silviscan measurement also included micrographs of the cross sectional area and automatic image analysis resulting in estimates of earlywood width (EW), tangential and radial lumen diameter (LUM t and LUM r ) and cell wall thicknesses (CWT) for the three different wood types of each ring (for further information on the Silviscan equipment please refer to Evans & Ilic [34]).
The second radial strip was used for FT-IR micro-spectroscopy measurements. Subsamples containing three to six tree rings each (maximum subsample width of 10 mm) from the outermost 15 to 17 tree rings were isolated using a razor blade. A total of 33 sapwood tree rings were included, covering the calendar years 1997-2012 in three different trees. The subsamples were vacuum-saturated in water and cut into nominally 20-µm thick specimens from the radial face using a sliding microtome. Obtaining transmission spectra in the radial direction is unusual but ensured that each individual spectrum expressed a mixture of the cell wall layers present within each cell. The specimens were air-dried on glass slides for at least 1 h and mounted on cardboard frames for IR transmission measurements. A visual survey image (10.000 × 10.000 µm) and one IR image were recorded for each specimen using a Perkin-Elmer Spectrum Spotlight 400 FT-IR microscope (Perkin-Elmer, Waltham, Massachusetts, USA) equipped with a MCT detector. A new background (in air) was taken before each image was recorded with the settings: 8 cm −1 , 15 scans, 4000-720 cm −1 , interferometer speed 1 cm/s, and spatial resolution of 25 × 25 µm (one pixel). The settings used for imaging were the same, except that the number of scans per spectrum was reduced to eight. IR images, four lines in height and covering the whole width of the specimen, were captured ( Figure 1). All IR images were 4 × 25 = 100 µm in height, while the width varied depending on the total width of the tree rings included in the specimen. to 17 tree rings were isolated using a razor blade. A total of 33 sapwood tree rings were included, covering the calendar years 1997-2012 in three different trees. The subsamples were vacuumsaturated in water and cut into nominally 20-µ m thick specimens from the radial face using a sliding microtome. Obtaining transmission spectra in the radial direction is unusual but ensured that each individual spectrum expressed a mixture of the cell wall layers present within each cell. The specimens were air-dried on glass slides for at least 1 h and mounted on cardboard frames for IR transmission measurements. A visual survey image (10.000 × 10.000 µ m) and one IR image were recorded for each specimen using a Perkin-Elmer Spectrum Spotlight 400 FT-IR microscope (Perkin-Elmer, Waltham, Massachusetts, USA) equipped with a MCT detector. A new background (in air) was taken before each image was recorded with the settings: 8 cm −1 , 15 scans, 4000-720 cm −1 , interferometer speed 1 cm/s, and spatial resolution of 25 × 25 µ m (one pixel). The settings used for imaging were the same, except that the number of scans per spectrum was reduced to eight. IR images, four lines in height and covering the whole width of the specimen, were captured ( Figure 1). All IR images were 4 × 25 = 100 µ m in height, while the width varied depending on the total width of the tree rings included in the specimen. All spectra from all subsamples and all four scanlines were manually inspected. This was done to verify the quality of the spectra, to validate that the four spectra from the four individual scanlines could be regarded as replicates (Figure 1), and to locate the tree ring boarders in the subsamples. The individual spectra were manually assigned to the individual tree rings of the wood samples by measurement of the tree ring widths in the visual survey images using GIMP 2.8 (GNU Image Manipulation Program, free software available from gimp.org/downloads/). In addition, the spectra were used to locate the tree ring borders; as the spectra clearly showed a consistent shift in band pattern at 1000-1100 cm −1 at the tree ring border. The spectra were assigned to earlywood, transition wood or latewood based on the Silviscan density measurements, and only the earlywood spectra were included in the study [35]. In approximately half of the analyzed tree rings, thin patches in the specimens had resulted in spectra of low quality with signals from air rather than from wood. In order to make sure that all included radial positions were represented by four spectra, all four spectra obtained for the same radial position were excluded in case one or more spectra were of low quality.
To observe quantitative changes in xylem cell wall biopolymer composition between the studied annual rings, ratios of peak areas from the FT-IR spectra were evaluated. The lignin polymer contains a high number of aromatic functional groups not present in carbohydrates [23,36]. The FT-IR absorption band at 1508 cm −1 is assigned to aromatic skeletal vibrations, and the band can thereby be directly assigned to the lignin polymer with no contributions from carbohydrates in wood and other types of plant material [37].
It is by far more complex to assign FT-IR bands to the structural carbohydrates cellulose and hemicellulose present in the wood cell walls since both polymers are composed of monomer units of aldoses and ketoses (monosaccharides). The band at 1104 cm −1 is assigned to ring asymmetric valence All spectra from all subsamples and all four scanlines were manually inspected. This was done to verify the quality of the spectra, to validate that the four spectra from the four individual scanlines could be regarded as replicates (Figure 1), and to locate the tree ring boarders in the subsamples. The individual spectra were manually assigned to the individual tree rings of the wood samples by measurement of the tree ring widths in the visual survey images using GIMP 2.8 (GNU Image Manipulation Program, free software available from gimp.org/downloads/). In addition, the spectra were used to locate the tree ring borders; as the spectra clearly showed a consistent shift in band pattern at 1000-1100 cm −1 at the tree ring border. The spectra were assigned to earlywood, transition wood or latewood based on the Silviscan density measurements, and only the earlywood spectra were included in the study [35]. In approximately half of the analyzed tree rings, thin patches in the specimens had resulted in spectra of low quality with signals from air rather than from wood. In order to make sure that all included radial positions were represented by four spectra, all four spectra obtained for the same radial position were excluded in case one or more spectra were of low quality.
To observe quantitative changes in xylem cell wall biopolymer composition between the studied annual rings, ratios of peak areas from the FT-IR spectra were evaluated. The lignin polymer contains a high number of aromatic functional groups not present in carbohydrates [23,36]. The FT-IR absorption band at 1508 cm −1 is assigned to aromatic skeletal vibrations, and the band can thereby be directly assigned to the lignin polymer with no contributions from carbohydrates in wood and other types of plant material [37].
It is by far more complex to assign FT-IR bands to the structural carbohydrates cellulose and hemicellulose present in the wood cell walls since both polymers are composed of monomer units of aldoses and ketoses (monosaccharides). The band at 1104 cm −1 is assigned to ring asymmetric valence vibration in polysaccharides [38,39] and the band does thereby represent contributions from both cellulose and hemicellulose (holocellulose). The band at 1056 cm −1 is assigned to C-O valence vibration mainly from C 3 -OH in six membered pyranose rings [40]. All glucose units in cellulose have this C 3 -OH valence vibration, whereas the most abundant hemicellulose in softwood, galactoglucomannan, lack this vibration in some mannose units as the C 3 is partly substituted by acetyl groups. The second most abundant hemicellulose in softwood is arabino-4-O-methylglucuronoxylan. This type of hemicellulose does also lack monosaccharide units with C 3 -OH valence vibration as arabinose form a five membered furanose ring. In addition, the arabinose units are linked to the xylan backbone at the C 3 position [41].
In conclusion, the 1056 cm −1 vibration band is not specific for cellulose but the band intensity should be more influenced by quantitative change in cellulose than hemicellulose.
Hemicellulose differs from cellulose and lignin by containing unconjugated carbonyl groups in acetyl groups in softwood galactoglucomannan and carboxylic acid functional groups in methylglucuronic acid side groups in xylan [42]. In addition, unconjugated ester groups are present in lignin-carbohydrate complexes formed between carboxylic acid groups in methylglucuronic acid side chain of xylan and γ-hydroxyl on the side chain of lignin monomer units [42]. The absorption band at 1732 cm −1 is assigned to C=O stretch in unconjugated carbonyl groups [38,43]. The band at 1732 cm −1 can therefore be assigned to hemicellulose with no contributions from cellulose or lignin. However, an important point is that changes in relative band intensity cannot be used as conclusive evidence for changes in the amount of the hemicellulose polymer backbone unless it is assumed that the degree of acetylation in galactoglucomannan and the relative amount of methylglucuronic acid in xylan is unchanged. The degree of acetylation in galactoglucomannan is in the range 4.3-8.8% for different softwood species, whereas the degree of methylglucuronic acid attached to the xylose backbone is in the range 16-23% [23].

Data Analysis
IR peak area ratios (1056/1508, 1104/1508, 1104/1732 and 1732/1508) were calculated for all spectra selected after the previously described quality control, and means were calculated for each ring from each tree. In addition, the mean value for each year (i.e., mean of the three trees) was calculated. The relationship between peak area ratios and cell dimensions were calculated based on Pearson's correlation both on tree level and across the three sampled trees. Plots were used to check significant linear relationships. The correlation between the mean peak area ratios and the monthly mean temperature as well as the cumulated DI were determined by calculation of Pearson's correlation coefficients. The correlations were determined both for the monthly average temperature in the previous year (June-December) and the current year (January-August) or the cumulated DI indices for different consecutive periods from March to August (DI3-DI8) in current-year and March to September (DI3-DI9) in previous-year. Finally, the previous and current year drought indices and monthly temperatures with significant (p < 0.05) correlations with the peak area ratios were identified. When calculating correlations across the three trees, only years for which IR data had been obtained for all three trees were included. The plots and MIXED procedure of the statistical software SAS were used to determine whether the relationships were linear [44,45]. The model used was defined as: where Peak ratio j is the mean IR peak ratio for each tree or the mean for the three sampled trees and year j, µ is the general mean, X nj is the climate variable n in year j (i.e., the climatic variables significantly correlated with the IR peak ratio), β n is the regression coefficients for climate variable n, X 2 nj is the squared climate variable n in year j, α n is regression coefficients for squared climate variable n, e j is the residual.

Results
Significant correlations between temperature and IR peak ratios and between DI and IR peak ratios were identified ( Table 2).
Regarding temperature, significant correlations across all trees were found for the 1732/1508 peak ratio (hemicellulose/lignin) and 1104/1508 (holocellulose/lignin). The correlation was positive between ratio 1732/1508 and the July, September and November temperatures of the previous year, indicating a higher hemicellulose content relative to lignin in a year following a warm summer or autumn. The correlation was positive between the 1104/1508 peak ratio and the June temperature of the current year, indicating a higher holocellulose content relative to lignin in years with a warm summer.

Discussion
Our finding that temperature stress is related to lignin content within the xylem cell walls agrees with some but not all earlier studies, albeit no study exactly like the present one could be found. Gindl et al. [46] found, based on data from 50 tree rings from five Picea abies (L.) Karst trees, that the lignin concentration in the secondary cell wall layer (S2) of the terminal latewood tracheids was lower during years with relatively lower temperatures in September-October. For another P. abies tree, Gindl & Grabner [47] found that the lignin concentration was markedly reduced in all latewood tracheids formed in a year with abnormally low temperatures in September to October. On the other hand Kilpeläinen et al. [48] found that the xylem (earlywood and latewood was not separated) produced in eight 15-year-old Pinus sylvestris (L.) trees grown under elevated temperatures had higher relative lignin content and lower relative hemicellulose content than in trees exposed to ambient temperatures. However, in a later study, including eight 20-year-old P. sylvestris trees grown under elevated temperatures, no significant effects of higher temperature on the relative content of cellulose, lignin and hemicellulose were found, but the relative content of acetone-soluble extractives was lower [49]. In a similar experiment, including six P. abies trees grown under elevated temperatures, Kostiainen et al. [50] found no effect on the relative content of lignin and of α-cellulose in the xylem (earlywood and latewood was not separated), but the relative content of extractable and soluble sugars decreased for the trees exposed to elevated temperatures.
Regarding our findings on the relation between drought stress and relative lignin content, a point to consider is that the lignin content is higher in the middle lamella and in cell corners than in the secondary cell wall. The recorded mean 1732/1508 ratio can therefore be expected to decrease if the earlywood radial tracheid diameter is reduced as more of the transmission spectra reflect lignin-rich regions (the spectra were obtained in the radial direction, see Figure 1). However, no significant relationship was found between the 1732/1508 ratio and cell dimensions when including all three sampled trees (Table 3). This indicates that the observed correlations between drought and warm weather and decreased lignin content relative to hemicellulose most likely is due to chemical changes in the cell wall material and not due to anatomical differences. Donaldson [51] found poorly lignified middle lamella and outer secondary walls in Pinus radiata (D. Don) specimens. These abnormalities were assumed to be the result of frequent water stress and possibly associated with nutritional stress. Nanayakkara et al. [52], on the other hand, found also for P. radiata, that drought stress significantly increases the relative content of arabinose, galactose, glucose and xylose in normal wood of one-year old plants. In a more general review on the effect of abiotic stresses on lignins, Cabane et al. [53] concluded that abiotic stress affects lignin biogenesis pathways.
The results of the current study indicate that climate might have an effect on tree growth not only via ring widths and cell dimensions, but also via xylem cell wall composition. We suggest that this aspect of tree growth also be included when studying the interrelations between xylem characteristics and growth conditions.

Conclusions
Our results indicate that warm summers and drought stress are related to the formation of earlywood xylem containing relatively less lignin than would otherwise be the case. The lower lignin content suggests that the wood formed is more susceptible to fungal degradation, both in the living tree and as dead wood. If these climate conditions become prevailing, the relation found could potentially contribute to a higher risk of forest stand decline and a speeding up of the carbon cycle within temperate forest ecosystems. However, the study is based on a very limited number of trees from only one location and results need to be confirmed by additional studies including more sites, more species and more trees. Furthermore, different types of models [54], i.e., not merely linear ones, could be explored when attempting to link climate and infrared peak ratios. Also, even though reduced lignin content is generally known to reduce recalcitrance in plant biomass, this notion needs to be confirmed to also be the case in this particular situation.