Enzymatic hydrolysis of the gelatinous layer in tension wood of Salix varieties as a measure of accessible cellulose for biofuels

Background Salix (willow) species represent an important source of bioenergy and offer great potential for producing biofuels. Salix spp. like many hardwoods, produce tension wood (TW) characterized by special fibres (G-fibres) that produce a cellulose-rich lignin-free gelatinous (G) layer on the inner fibre cell wall. Presence of increased amounts of TW and G-fibres represents an increased source of cellulose. In the present study, the presence of TW in whole stems of different Salix varieties was characterized (i.e., physical measurements, histochemistry, image analysis, and microscopy) as a possible marker for the availability of freely available cellulose and potential for releasing d-glucose. Stem cross sections from different Salix varieties (Tora, Björn) were characterized for TW, and subjected to cellulase hydrolysis with the free d-glucose produced determined using a glucose oxidase/peroxidase (GOPOD) assay. Effect of cellulase on the cross sections and progressive hydrolysis of the G-layer was followed using light microscopy after staining and scanning electron microscopy (SEM). Results Tension wood fibres with G-layers were developed multilaterally in all stems studied. Salix TW from varieties Tora and Björn showed fibre G-layers were non-lignified with variable thickness. Results showed: (i) Differences in total % TW at different stem heights; (ii) that using a 3-day incubation period at 50 °C, the G-layers could be hydrolyzed with no apparent ultrastructural effects on lignified secondary cell wall layers and middle lamellae of other cell elements; and (iii) that by correlating the amount of d-glucose produced from cross sections at different stem heights together with total % TW and density, an estimate of the total free d-glucose in stems can be derived and compared between varieties. These values were used together with a literature value (45%) for estimating the contribution played by G-layer cellulose to the total cellulose content. Conclusions The stem section-enzyme method developed provides a viable approach to compare different Salix varieties ability to produce TW and thus freely available d-glucose for fermentation and biofuel production. The use of Salix stem cross sections rather than comminuted biomass allows direct correlation between tissue- and cell types with d-glucose release. Results allowed correlation between % TW in cross sections and entire Salix stems with d-glucose production from digested G-layers. Results further emphasize the importance of TW and G-fibre cellulose as an important marker for enhanced d-glucose release in Salix varieties.

fuels particularly bioethanol and biogas [2][3][4][5][6]. Features that make Salix interesting for biofuel include its rapid growth, the possibility for growth on barren and poor soil, relatively low lignin content in contrast to other wood species, and its high annual volume production per hectare [7].
An important parameter for selection of any plant/ wood species for biofuel production is based on the lignocellulose material recalcitrance, which is related to the chemistry and relation between the major biomass polymers cellulose, hemicelluloses and lignin. In wood species, the cellulose content normally resides within the range 40-50% in both hardwoods and softwoods, while lignin, and hemicelluloses tend to be inversely correlated [8]. In woody lignocellulose, the main polymers are interrelated at molecular and ultrastructural levels with cellulose primarily in contact with the hemicelluloses that are primarily bound to lignin [9], the arrangement the ultimate supramolecular basis for recalcitrance. Recalcitrance in bioenergy plants is also related to total lignin content and is why low lignin containing and extractive free plant species like fast-growing Salix and its close relative Poplar have shown to be the most attractive options for biofuels. During bioethanol production, cellulose in the biomass is usually converted to fermentable sugars through thermal-or chemical pretreatments. This is followed by enzymatic hydrolysis with enzymes (i.e., cellulases), the glucose simultaneously converted by yeast or conducted in a separate process [1]. Thus, for bioethanol production, the most important and limiting factors will be the total cellulose content and its accessibility in the lignocellulose biomass structure for enzymatic hydrolysis.
Salix is a member of the Salicaceae family, and numerous varieties have been selected and commercially developed over the years for their rapid growth and volume production per hectare [10][11][12]. Salix spp. has a diffuse to semi-ring-porous wood anatomy characterized by the presence of fibres, parenchyma cells and radially orientated vessels (often clustered in 2-3 s) in different ratios depending on species and variety. Like many hardwoods, Salix spp. develop abnormal (i.e., reaction) wood under particular circumstances during which a characteristic tissue and cellular type known as tension wood (TW) containing gelatinous fibres (i.e., G-fibres) are produced [13]. In certain hardwoods (e.g., poplar, willow, oak, birch) this reaction is particularly prominent in tree branches with TW produced on the upper side of branches, the response helping to maintain the branch in a perpendicular orientation to the axial stem [14]. TW also occurs in young developing stems (e.g., 1-4-yearold plants) of Salix and its close relative Poplar [15,16]. Studies have further shown enhanced TW development in environments, where stems are exposed to strong adverse winds suggesting differences in both gravimetric responses and possibly genetic differences [15,17].
Anatomically, TW is composed of characteristic G-fibres with the most characteristic type producing an additional cellulose-rich layer, known as the gelatinous layer (G-layer) on the lumen side of the secondary cell walls of fibres [14,18,19]. G-fibres with gelatinous layers are cellulose-rich and in contrast to the other cell wall layers forming the outer secondary cell wall, the G-layer either lacks or has very low lignin levels in many hardwood species [14,20]. Early reports suggested a cellulose (as d-glucose) content as high as 98.5% [20] in Populus tremula, although more recent estimates of isolated G-layers from P. alba give values in the region of 75-78% together with matrix materials of hemicelluloses [21,22]. The cellulose in the G-layer is reported as highly crystalline with X-ray diffraction showing 60% crystallinity with monoclinic 1β cellulose crystals [23]. Various microscopy approaches have further shown G-layers in different hardwoods to consist morphologically of microfibrils and microfibril aggregates (macrofibrils; ranging from 10 to 60 nm) held in an interconnected honeycomb gel construction, with mesopores of the order 2-50 nm [16,[24][25][26] alternatively 6-12 nm [27]. All these morphological aspects are consistent with free access to enzymatic attack from the fibre cell lumen when exposed.
Recent studies with Salix have shown the presence of TW in stems processed for biofuel production using cellulase hydrolysis to confer higher bioethanol yields [15]. While some debate exists on what is more important, the total cellulose content or cellulose accessibility or a combination of both [28], the presence of TW can be recognized as an important marker for biofuel potential. Salix varieties prone to producing large amounts of TW could, therefore, be seen as a favorable option for biofuel production. When assessing different varieties of Salix for biotechnology and biopolymer yields, it is normal practice to comminute entire stems with bark by chipping/refining for traditional wet chemical analysis or more recently use high-throughput chemical screening, both approaches involving random sampling of the total initial Salix variant biomass [29]. The former approach enables the processing and comparison of large number of species/varieties and reflects industrial situations. However, while these approaches overcome the problems of processing large numbers of samples, only very small amounts of biomass are used (often mgs). Thus, any differences in tissue type in stems processed under these conditions would tend to be lost. For example, stems showing TW also attributes generally show opposite-(OW, i.e., the wood formed on the opposite side of the TW tissue) and normal wood (NW, i.e., wood without reaction tissue) [14,18] and thus bulk chemical analysis may only show small differences in polymeric chemical composition without some form of pre-separation. Similarly, analysis of only discrete regions from stems as frequently done for basal stem regions is generally argued as not representative of the situation along the stem. Thus even if a Salix variety was shown to have significant TW with increased cellulose content and accessibility at basal height above ground, this may not necessarily be related to the entire stem and Salix variant genotype or phenotypic response.
In the present study, we developed a method to assessed entire stem cross sections from different Salix varieties and applied a commercial cellulase to determine accessible cellulose as an indirect method to quantify the amount of TW tissue and readily accessible d-glucose. Thus, the higher the d-glucose concentration determined, the greater the amounts of TW fibres and accessible d-glucose and the most promising varieties for accessible sugar and bioethanol production.

Results and discussion
The physical dimensions of 2-year growth of the Salix varieties used for studies varied from ca. 200 to 400 cm in height (Tables 1, 2, 3) and between 1.85 cm (Jorr) and 2.83 cm (Tora) at 40 cm above the ground to 1.05 cm (Tora) at 400 cm in diameter (debarked samples, Table 1). Only 2-year growth was studied with the majority of the Salix varieties ca. 60-100 cm taller including first year growth.
In the present work, emphasis was placed on Tora and Björn which are commercial Salix varieties. Tora showed the tallest stems with only Björn having slightly thicker stems (with or without bark) in 6 points measured along the stem axis (Tables 1, 2, 3). All physical measurements were made on debarked samples. Ten cm stem cylinders were used for section analyses of TW and the enzymatic and staining studies. For further comparison, a wood sample taken from an 8-year-old tree of S. viminalis growing locally in Uppsala, Sweden was also included in the study. All enzyme studies were performed on debarked samples.
Macerated Salix tension-, normal-and opposite wood fibres from 1.0 to 4.5 cm wide stems (or 2-year-old stems) have an average length of ca. 1.0 mm. With our Fiber-Master analysis of stems, 99% of the length weighted fibre length in S. viminalis was less than 1.5 mm and 96% less than 1.0 mm (unpublished observations). Thus, the thickness of the serial sections made (ca. 30 μm) was well   within the average range of full fibre length allowing free access of the enzyme preparation into the fibre lumina (i.e., to avoid the presence of fibres with closed ends).

Estimation of tension wood in Salix stem cross sections using image analysis
Presence of TW in cross sections of unstained Salix stems was visible prior to staining as cellular regions with glistening shiny appearance. For improved definition, however, three staining approaches were assessed initially including astra blue (single staining) and double staining with either safranin/astra blue and safranin/chlorazol black E (see "Materials and methods"; Fig. 1 Since areas of TW were frequently present on opposing sides of the same stem cross section (Figs. 2, 3), the induction of TW tissue was not unilateral and restricted from to one side as seen with TW in branches of hardwoods but rather multilateral (Figs. 2c, 3c). Thus the presence of opposite and normal wood was not possible to truly differentiate.

Density
The 2-year Salix samples examined showed a typical semi-diffuse anatomical structure with radially orientated vessels, fibres and parenchyma cells in the annual growth rings in cross section. Average absolute dry density measurements performed on debarked stem samples of Tora and Björn showed similar values of ca. 468 kg/ m 3 and ca. 473 kg/m 3, respectively (Tables 2, 3). Density decreased from ca. 522 to ca. 450 kg/m 3 in Tora and between ca. 521 and ca. 451 kg/m 3 in Björn with increasing height from 40 to 400 cm and 40 to 360 cm, respectively (Tables 2, 3). Density measurements at the lower stem heights are consistent with published reports for S. viminalis [29] and other Salix species [30,31], while the decrease in density with increasing height is normal for tree growth. Results, however, emphasize the variation that exists and importance of knowing the sample point along the stem.
Since density strongly reflects the anatomical composition, density measurements were also performed on enriched TW regions cut from stems of Tora and Björn from both first-and second growth rings  Table 4). Results for both Salix varieties showed a similar trend as the total stem with a decreasing total density between ca. 492-454 kg/m 3 for Tora and ca. 491-451 kg/m 3 for Björn with increase in stem height between 80 and 360 cm (Table 4). Density values for the first and second annual rings between 80, 200, 360 cm varied between Salix varieties with ca. 424 and 490 kg/m 3 respectively, recorded for Tora and ca. 451 and 509 kg/m 3 obtained for Björn for 1 st respectively, 2 nd growth rings ( Table 4). The secondary growth rings were both wider and composed of TW fibres with thicker G-layers. These variations emphasize the

G-layer structure, presence of lignin and suitability for enzyme hydrolysis
Using the sliding microtome approach for sectioning, the G-layer was most often detached from the inner secondary cell wall in studied TW fibre cross sections (Figs. 4a-d, 6). This is typically an artefact of sectioning [16,32] caused by the weak attachment of the G-layer to the outer secondary wall, likely caused through differences in microfibril angle (MFA) at the interface between the two layers (i.e., G-layer was almost perpendicular to the fibre axis) and the S2 wall (ca. 80-90°) and its lack of lignification (see below). For the purpose of the present study, however, this has less importance as all stem sections were cut in a similar way and any artefacts derived would have been similar for all cross sections and Salix varieties. In addition, it is assumed that sectioning would have allowed for improved enzyme penetration and subsequent hydrolysis from both the inside (i.e., interface of the outer secondary wall) and outside (i.e., lumen side) of the G-layer during treatments. SEM observations of control sections further confirmed detachment of the G-layer from the outer lignified secondary cell walls during sectioning (Fig. 6). Previous studies on Poplar, a closely related wood species have also shown differences in MFA, the mesoporous and cellulose aggregate structure of the G-layer as well as its strong affinity for cellulase binding domains [16]. Both the light microscopy and SEM observations confirmed a typical G-fibre organization with a secondary wall structure of S1 + S2 + G layers as previously reported for Salix varieties [33]. While the G-layer varied considerably in thickness between growth rings and both early-and latewood, the G-layer appeared as an "open" rather than compact structure as previously reported for other wood species with similar S1 + S2 + G cell wall structure [16,26].
Light microscopy of TW fibres typically showed strong staining of the G-layer with astra blue in single and double staining approaches consistent with lack of lignin [34] and open structure (Fig. 4a, b). Lack of lignin in the G-layer was confirmed by negative reactions with both Weisner and Mäule reagents (Fig. 4c-h), the former staining middle lamellae and vessel secondary walls strongly reddish-pink for cinnamaldehyde units (i.e., guaiacyl lignin) and the latter the outer secondary walls in G-fibres and control normal fibres for syringyl lignin units (Fig. 4e-h). Previous studies have further confirmed the presence of a variety on non-cellulosic components, particularly 1-4-β-d galactan in the G-layer of Salix spp. [33]. Both the open structure and lack of lignification of the G-layer together with cross sections and open fibres enhanced the possibility of both enzyme penetration and hydrolysis.

Enzymatic hydrolysis of G-layer cellulose in Salix stem cross sections-morphological observations
Initial studies were performed to determine optimal conditions for using the cellulase complement Cellic CTec2 on Salix stem cross sections. Using the recommended enzyme dosage (i.e., enzyme concentration 20µL/10 mL citrate buffer), temperature and pH [35] and 80-100 mg biomass dry weight, we found by monitoring d-glucose release and progressive G-layer removal using light microscopy observations, that 3 days was optimal (see "Materials and methods"). We also found that pre-drying cross sections at 103 ± 2 °C for dry wt determinations before enzyme treatment did not reduce d-glucose production. Shorter periods of enzyme incubation resulted in partially degraded G-layers, but indicated enzyme hydrolysis and surface erosion from directions of both the fibre lumen and interface between the outer fibre secondary cell walls and the G-layer. Occasional fibrous material of partially hydrolyzed G-layers was also found on the sections incubated for shorter periods (Fig. 6d). Light microscopy of stained sections showed almost complete removal of TW bands and G-layers after 3-day incubation leaving only a very thin residue layer in same fibres (Fig. 5a-h). Similarly, SEM observations after 3-day enzyme treatment of cross sections from stems at 40, 200 and 360 cm height showed the absence of G-layers (Fig. 6c, d, g, h, k, l, o, p). Light-and SEM microscopy was used to observe the lignified wood cells for evidence of cell wall hydrolysis (e.g., cavity formation, surface erosion). No morphological evidence for hydrolysis and attack of lignified cell walls in fibres and vessel secondary cell walls and middle lamellae was found using this approach suggesting that enzyme digestion after 3 days was primarily directed at the non-lignified celluloserich gelatinous layers in the cross sections. Additional observations were made on 1-day enzyme-treated radialand tangential longitudinal sections cut from Salix stems with TW and G-fibres. This showed similar evidence for the surface hydrolysis of G-layers (not shown). Storage materials such as starch granules also remained in vasicentric parenchyma cells surrounding vessels after enzyme treatment (Fig. 6n). Observations indicate the G-layers were readily accessible to hydrolysis by the cellulase and that digestion was primarily from both sides of the G-layers in sections. The lack of morphological evidence for attack of the cell walls of other cellular elements (i.e., vessels, parenchyma cells, outer fibre secondary cell walls) provides further evidence for the importance of biomass pre-comminution and need for the exposure of cellulose during cellulase hydrolysis of lignocellulose. Despite using cross sections, little attack was apparent of the fibre cut surfaces, suggesting that the enzyme complement used had greater endo-cellulase than exocellulase activity on the Salix samples. Figure 7 shows quantification of the progressive release of d-glucose after Cellic CTec2 treatment over 3 days for 4 Salix varieties and one variety after fertilization (Tora + F). Values of R 2 ranged from 0.78 (lowest) for Björn and 0.95 for Tora showing increased d-glucose production with Salix stem height for all varieties (Fig. 7). Additional R 2 values included 0.64 for Loden, 0.84 for Jorr and 0.86 for Tora + F (Fig. 7). Results show , pre-hydrolyzed (i.e., pre-cellulase treated) cross sections of Tora (d-glucose 4 mg/g dry wt) and an 8-year field Salix sample (d-glucose 116 mg/g dry wt). The contrasting d-glucose production with the bleached softwood pulp and pre-hydrolyzed Salix cross sections is consistent with lack of lignin and freely available cellulose in the former and lack of freely available cellulose through lignification in the latter. Results with the field Salix TW sample are also consistent with freely available G-layer cellulose and the analysis method, although it is not known if age can change cellulose availability. Results indicate variations in available G-layer along the stem and presumably reflect cellulose accessibility. The value of d-glucose release (i.e., ca. 156 mg/g dry wt) for Tora at 400 cm approaches that for the bleached softwood pulp, where accessibility for cellulose should not be impeded. Minimal d-glucose release for the prehydrolyzed cross sections similarly reflect the lack of freely available cellulose, while the value for the 8-year Salix field sample indicates the method will also be viable for older stem material (Fig. 8). Comparison of the final release of d-glucose from Tora and Björn shows the latter Salix varieties to release greater amounts between 40 and 320 cm and in the latter higher release at 360 cm and above (Fig. 9, Tables 2, 3). Additional analysis of d-glucose release from Salix varieties Tora plus fertilization, Loden and Jorr with stem diameters of 1.62-1.72  (Table 5) showed stems with larger diameter to release more d-glucose (e.g., Björn). d-glucose release at height 200 cm showed variations between the different varieties (Table 6), although this was probably more likely related to the final height attained as the upper regions always gave highest release of d-glucose and the Salix varieties showed differences. Figure 10 shows a scatterplot matrix of the four variables, stem height, d-glucose production, percentage TW and density. Strong positive correlations were given between stem height, d-glucose and percentage TW all of which correlated negatively with density.

Estimation of total d-glucose release from stems of different Salix varieties and relation with total cellulose
For bioethanol production, one of the most important criteria is the total native cellulose present in the stem biomass and its accessibility for first stage enzyme (cellulase) hydrolysis [1]. With cellulose accessibility, this is accepted as regulated by biomass recalcitrance, where both lignin type and concentration can play an important role [36]. Previous studies on Salix varieties using wet chemical analysis have reported total cellulose contents ranging between 41.6 and 55.9% (i.e., mean 45%) with corresponding lignin levels between 13.8 and 28.0% [2,[37][38][39][40][41][42]. Therefore, for our calculations on the possible contribution played by G-layer cellulose to total stem cellulose concentration in our samples, we used 45% cellulose. From our staining experiments with stem cross sections, the gelatinous layer appeared nonlignified and thus the cellulose should be freely available for enzymatic hydrolysis, and, therefore, should not be affected by recalcitrance. Using these criteria and a total stem volume of 40-400 cm, the total calculated free d-glucose available that could be produced through the enzymatic treatment of the G-layer in TW in the Tora stems analyzed would lie in the region ca. 16% of the total cellulose (Table 7). If the total stem growth between 0 and 400 cm is used (i.e., not considering the primary shoot growth above), then the calculated value is ca. 14%. The corresponding values for Björn between 80 and 320 cm (% TW were not calculated at 40 and 360 cm; i.e., only the 1st annual ring was present in the latter) were ca. 21% and 20%, respectively (Table 8). Additional analysis of the 8-year-old Salix field sample (i.e., contained only TW) taken as control and at only one point in the stem was ca. 26% (Table 7). These results indicate that between ca. 15 and 21% of the total cellulose (i.e., 45%) in Tora and Björn stems   maybe contributed by accessible cellulose in the gelatinous layers of TW fibres. Brereton et al. [43], using a 3-month-old laboratory cultivated Salix variant with tension induction, found that isolated pure TW after milling to fine particles could release ca. 250 mg/g dry wt glucose with enzymatic saccharification, which was 66% higher than the corresponding control (released ca. 151 mg/g dry wt glucose). For the genotype Tora, which is the same Salix variety used in our study, the glucose yield was 150 mg/g dry wt after 3 years growing under low reaction wood inducing conditions, and 200 mg/g dry wt after 4 years growing under high reaction wood inducing conditions [2]. Sassner [3] also recorded a very high enzymatic glucose yield from 4-year-old Tora that released 55.6 g glucose/100 g dry wt after steam pretreatment, sulphuric acid impregnation and subsequent enzymatic hydrolysis. TW is normally recognized as an abnormal wood trait developed by many angiosperm trees that allows them to maintain its branches in a perpendicular position to the main trunk. TW development in branches is thought to be induced as a gravitational response [13], while the mechanism in young, rapidly growing coppice trees like Salix is unknown. Studies have shown, however, that , percentage TW in stem cross sections (%) and density (kg/m 3 ) for Tora. The strength of the relationship is given by the correlation coefficient value R. Positive correlations were given between stem height and d-glucose and percentage TW and negative correlations between stem densities (total), stem height, d-glucose and percentage TW both genotype and phenotypic response can be important for coppice trees, with one study showing a fivefold increase in glucose release from a Salix variety grown under control-and adverse windy conditions [15]. Quantitative analysis of percentage TW in stems in the present study grown at the same site showed variations with stem height and between annual growth rings. Similarly, it was difficult to separate the microdistribution of TW into classical zones of TW, opposite wood and normal wood as TW was frequently developed multilaterally around stems.
The present study further emphasizes the important contribution that TW and G-layer cellulose may have on the total cellulose content of Salix grown under field situations, a result consistent with previous studies and important with respect to optimizing biofuel production from coppice trees [15,17,44]. The use of stem cross sections provides a novel approach, whereby discrete stem regions and entire stems can be assessed directly for quantifying the total TW G-cellulose content in stems, thereby providing information on its accessibility as a readily available source of hydrolysable cellulose. Observations further confirmed the important role played by lignin for recalcitrance in the secondary cell walls of both the outer cell wall layers in G-fibres and other cell elements that remained  undigested during cellulase treatments in comparison to the non-lignified gelatinous layer of TW fibres. This result further confirms the importance of biomass comminution and cellulose exposure for cellulases during biomass processing. Results, therefore, indicate a very important contribution can be played by presence of TW and G-layer cellulose and that this contribution likely varies with genotype and prevailing environmental conditions.

Conclusions
Physical, histochemical/microscopy, image analysis and enzymatic studies were performed on 2-year-old rapidly growing Salix varieties to quantify the presence of tension wood (TW) along stems, characterize TW G-fibres and determine the likely contribution played by gelatinous layer cellulose to total stem cellulose; A method involving exposure of cross sections cut from the stems of different Salix varieties was developed and allowed determination of the role played by the G-layer cellulose from TW fibres; Density, microscopy (LM/SEM) and image analysis methods confirmed multilateral TW development in all stems and varieties with increasing % TW negatively correlated with density but positively correlated with stem height and cellulose (as d-glucose); G-layer cellulose in TW fibres exposed in cross sections was readily hydrolyzed by cellulase producing ca. 44 and 38 kg/m 3 of d-glucose on average for entire stems of Salix varieties Björn and Tora, respectively, with the d-glucose release increasing with stem height; By correlating stem TW volume with average total cellulose content from previous chemical analysis (i.e., literature value. 45%), it was estimated that G-layer cellulose contributed between 16 and 20% of the total cellulose yield in the two Salix varieties (Tora, Björn) studied.

Plant materials and harvesting
The 2-year-old Salix stems used were derived from a short rotation plantation experiment on arable land in Uppsala, Central Sweden (59° 49′ N, 17° 39′ E) [11], September 2020. The Salix varieties were taxonomically distinct at species or genotype level and included two full siblings 'Björn' (Salix schwerinii E. Wolf. × S. viminalis L.) and 'Tora' (S. schwerinii × S. viminalis) as well as 'Jorr' (S. viminalis) and 'Loden' (S. dasyclados Wimm.), the latter most distinct in terms of taxonomy from the other three varieties [11]. The stems were selected from a randomized plot, cut ca. 10 cm above the soil surface, transported to the laboratory, and maintained at -20° C until further use. Physical features (stem length and breadth) were recorded in the laboratory before further analyses.
Stems were sawn at 10 cm intervals from the basal end of the stem (Fig. 2).

Density measurements
Absolute dry densities of wood along Salix stems were calculated for debarked cylinders of ca. 2 cm sawn every 40 cm from the basal end of the stem (Fig. 1, Tables 2,  3). The 2 cm wood cylinders were dried in an oven at 103 ± 2 °C overnight and absolute dry weight measured next day. The volume of individual wood samples was then measured by the water displacement method [45]. Wood density (ρ) was determined as the mass (i.e., dry wt.) divided by volume and given in kg/m 3 . To determine the effect of TW on wood density, areas rich in TW (ca. 0.3-5 mm in cross section × 2.0 cm long), from Tora and Björn at 80, 200, 360 cm above ground, were carefully cut from the primary-and secondary growth rings of stems (Table 4) and processed as described above. Serial cut samples (2-3) were used for each measurement and each sample was measured three times.

Sectioning and staining of stem cross sections for tension wood
Entire stem cross sections (ca. 15-30 µm thick, Fig. 1) were cut using a sledge sliding microtome (Leica 1300 sledge microtome). Serial sections were stained initially with 1% w/v astra blue or double stained with either 1% w/v astra blue + 0.1% w/v safranin, 1% w/v chlorazol black E + 0.1% w/v safranin) [46] to visualize presence of TW. Chlorazol black E and astra blue stain the gelatinous layer (G-layer) of TW fibres strongly black or blue. The combination of safranin and chorazol black E gave best staining ( Fig. 1) for subsequent image analysis. Entire stem cross sections were stained and scanned using an Epson Perfection Pro 750 film scanner with a pixel resolution of 2400 dpi and areas of TW marked. Thereafter, using Adobe Photoshop CC 2017, the pixel values of the selected TW area and the whole section area can be determined. The percentage area of TW in entire stem cross sections was then quantified using the equation: area of TW in pixels / area of the whole section in pixels × 100%. From this, the total volume in kg/m 3 for individual sections of the stem could be calculated. Detailed analyses of individual TW regions and fibres after staining was conducted using a Leica DMLB light microscope (Leica Microsystems, Wetzlar, Germany) equipped with an Infinity X-32 digital camera (Deltapix, Samourn, Denmark) to confirm presence/absence of TW and G-layers. To visualize lignin distributions, fresh sections were stained using the Wiesner and Mäule reactions [47,48]. For the Wiesner reaction, sections were treated with 2% v/v phloroglucinol in ethanol and mounted in 6 M hydrochloric acid. For the Mäule reaction, sections were stained with 1% w/v potassium permanganate for 5 min, washed (5 min in water), treated with 2 M HCl for 5 min, washed (5 min in water) and mounted in ammonium hydroxide. Sections were observed directly after Wiesner and Mäule reactions, as the colour reaction is unstable. All stains were purchased from Sigma-Aldrich (St. Louis, USA).
Serial semi-thin cross sections of stems (ca. 15-30 µm thick, Fig. 1) were also cut for the staining and enzyme treatments using a commercial cellulase complement (Cellic CTec2, Novozymes [35]) and the light-and electron microscopy studies described below. In addition, radial-and transverse longitudinal sections (ca. 15-30 µm thick) cut from Salix stem regions showing TW were treated with cellulase to visualize changes in the morphology of the G-layer during enzymatic hydrolysis.

Enzyme hydrolysis
All studies were conducted on debarked and washed (distilled water) serial stem sections. Initial studies were carried out on stem sections of Tora to determine optimal conditions for incubation with Cellic CTec2, a cellulase blend enzyme for producing d-glucose [35]. According to the manufacturers, the enzyme blend contains a cocktail of cellulases and recommended for industrial treatment of lignocellulose biomass [35]. This involved time-period, enzyme concentration, number of Salix sections sufficient for quantification of d-glucose production, and studies on the effect of pre-drying sections at 103 ± 2 °C for dry weight determination before enzyme treatment as well as observations on the progressive removal of the G-layer using light microscopy after staining with chlorazol black E and safranin. Detection of free d-glucose was done using the glucose oxidase/peroxidase (GOPOD) assay kit according to the manufacturer's instructions (Megazyme, Bray, Ireland). Stem cross sections were incubated in 50 mL Erlenmeyer flasks containing 20 µL Cellic CTec2 enzyme in 10 mL citrate buffer (pH 5), at 50 °C and 100 rpm (i.e., sufficient to keep sections slightly moving and separate from each other) for 3 days in an INNOVA 4000 rotary incubator (Instrument AB, Lambda, Sweden). Twenty microliters of 95% ethanol was used as bactericide. A minimum of three serial stem cross sections (i.e., ca. 80-100 mg wood, dry wt) were used starting at 40 cm above the ground from Salix varieties (Tables 2, 3, 5, 6; Figs. 2, 3). Additional enzymatic analyses were conducted on an 8-year field sample of S. viminalis (Fig. 8).
As positive control for TW, washed stem sections of Tora previously incubated with Cellic CTec2 and shown by light microscopy to have their G-layers in TW fibres hydrolyzed were used. An obvious control would have been opposite or mature wood from similar stem regions. However, as shown from the total distribution and percentage of TW in the Salix stems (Figs. 2, 3), separating such samples was difficult. An additional method control included bleached (softwood) chemical pulp fibres [49]. Results of enzyme activity are expressed as mg d-glucose produced per g dry wt of wood material from sections or in kg/m 3 when whole stems are compared. All experiments were made in duplicate, reflecting a minimum of three serial cross sections per flask depending on stem cross-sectional area. For enzymatic analysis of sections from the upper regions of stems, where the diameter of the stem was small (Tables 2, 3), a minimum of 5-10 sections was required for quantification of d-glucose.

Scanning electron microscopy
Scanning electron microscopy (SEM) was used to follow the progressive hydrolysis of TW G-layers as well as any hydrolytic effects of the cellulase complement on the cellular structure of other cell wall layers (i.e., secondary cell walls/primary, middle lamella regions). Entire semi-thin sections of stems after enzymatic treatment and controls were fixed in 3% v/v glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.2) for 3 h at room temperature. After washing in the same buffer (3 × 20 min), sections were post fixed in 1% w/v osmium tetroxide for 1 h. Following washing in water (3 × 30 min; and overnight), sections were dehydrated in a progressive series of ethanol [16] and finally acetone. Samples were dried in an Agar E3000 critical point dryer (Quorum Technologies Ltd., East Essex, UK) using liquid CO 2 as the drying agent, mounted on stubs, and coated with gold using an Emitech K550X sputter device (Emitech Ltd., Kent, UK) [16]. Observations were made using a Philips XL 30 ESEM (Philips, Eindhoven, Netherlands) operated at 10-20 kV with images recorded digitally.

Analysis of tension wood in stems from field experiment
For comparison, a stem from an 8-year-old growing tree of S. viminalis was cut to provide control wood materials for the enzymatic studies and calculations of d-glucose production.