The physical dimensions of two-year growth of the Salix varieties used for studies varied from ca. 200 to 400 cm in height (Tables 1-3) 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 two-year growth was studied with the majority of the Salix varieties ca 60-100 cm taller including first year growth. Tora and Björn were chosen for detailed studies as these had greatest growth in length and diameter (Tables 1-3).
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 E black (See Materials and Methods; Figure 1). Of the three approaches, safranin/chlorazol black E was chosen for analysis of TW tissue area as it provided improved definition for quantification when scanning entire cross-sections. TW was present in Salix stem cross-sections as discrete bands of distinct purple/black staining (Figures 1-3) and as single fibres diffusely scattered in both the early- and latewood growth rings of all stems of Tora, Björn, Loden and Jorr. Only the TW bands were possible to estimate quantitatively. Light microscopy confirmed TW bands and presence of wood fibres with G-layers (G-fibres). Distribution of TW was not proportionate to first and second growth ring thickness and showed variability with some TW developing bands longer than half the stem circumference (Figures 1-3). Differences in total TW were evident in cross-sections and varied with stem height/diameter and between Salix varieties and in Tora ranged from ca. 18 % at 80 cm to ca. 45 % at 400 cm with corresponding values in Björn ca. 27 % (80 cm) and ca. 38 % (320 cm) for similar heights (Figures 2, 3). The percentage area of TW estimated in cross-sections was used to estimate the volume of TW in 40 cm cylinders cut from the stems (Figure 2 (left), Tables 2, 3) and together to estimate TW in entire stems. In both Salix varieties, the amount of TW calculated from cross-sections as a percentage of actual cylinder volume increased with height as the total area of cross-sections/volume of cylinders decreased (Figures 2, 3).
Since areas of TW were frequently present on opposing sides of the same stem cross-section (Figures 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 (Figures 2c, 3c). Thus the presence of opposite and normal wood was not possible to truly differentiate.
The two-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/m3 and ca. 473 kg/m3 respectively (Tables 2, 3). Density decreased from ca. 522 to ca. 450 kg/m3 in Tora and between ca. 521 to ca. 451 kg/m3 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  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/m3 for Tora and ca. 491-451 kg/m3 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/m3 respectively recorded for Tora and ca. 451 and 509 kg/m3 obtained for Björn for 1st respectively 2nd growth rings (Table 4). The secondary growth rings were both wider and composed of TW fibres with thicker G-layers. These variations emphasize the inherent differences that are equalized when the density of entire stems are measured. Light- and electron microscopy confirmed variations in the thickness of gelatinous layers especially between early- and latewood (Figure 6 b, f, j).
G-layer structure, presence of lignin and suitability for enzyme hydrolysis
By 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 (Figures 4 a-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 80-90o) 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 (Figure 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 . 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 . 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  and open structure (Figures 4a, b). Lack of lignin in the G-layer was confirmed by negative reactions with both Weisner and Mäule reagents (Figures 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 (Figure 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. . 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/10mL citrate buffer), temperature and pH  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 oC 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 (Figure 6d). Light microscopy of stained sections showed almost complete removal of TW bands and G-layers after 3 days incubation leaving only a very thin residue layer in same fibres (Figure 5a-h). Similarly, SEM observations after 3 days enzyme treatment of cross-sections from stems at 40, 200 and 360 cm height showed the absence of G-layers (Figure 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 cellulose-rich gelatinous layers in the cross-sections. Additional observations were made on 1-day enzyme-treated radial- and 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 (Figure 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 exo-cellulase activity on the Salix samples. .
Quantitation of accessible D-glucose in Salix stems using Cellic CTec2 and GOPOX assay
Figure 7 show 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 R2 ranged from 0.78 (lowest) for Björn and 0.95 for Tora showing increased D-glucose production with Salix stem height for all varieties (Figure 7). Additional R2 values included 0.64 for Loden, 0.84 for Jorr and 0.86 for Tora + F (Figure 7). Results show a good correlation providing evidence for a progressive and stepwise release of D-glucose and absence of inhibition. Figures 8, 9 and Tables 2, 3 show results for the final release of D-glucose following 3 days cellulase treatment of stem cross-sections from 40-400 cm for Tora and 40 to 360 cm for Björn. Results for both Salix varieties show an increase of D-glucose release with stem height with final values of ca. 156 (Tora) and ca. 118 (Björn) mg of D-glucose/g dry wt of Salix wood, respectively. Figure 8 further shows the release of D-glucose from material controls including a bleached spruce pulp (D-glucose 214 mg/g dry wt), 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 pre-hydrolyzed 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 (Figure 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 (Figure 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 . With cellulose accessibility, this is accepted as regulated by biomass recalcitrance where both lignin type and concentration can play an important role . Previous studies on Salix varieties using wet chemical analysis have reported total cellulose contents ranging between 41.6 to 55.9 % (i.e. mean 45 %) with corresponding lignin levels between 13.8 to 28.0 % [2, 37-42]. 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 non-lignified 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 - 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 15-21 % 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. , 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 . Sassner  also recorded a very high enzymatic glucose yields from four-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 , while the mechanism in young, rapidly growing coppice trees like Salix is unknown. Studies have shown however, that both genotype and phenotypic response can be important for coppice trees, with one study showing a 5 fold increase in glucose release from a Salix variety grown under control- and adverse windy conditions . 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.