Transgenic hybrid aspen (Populus tremula L. × tremuloides Michx.) lines were generated as described previously (Ratke et al., 2015). The lines harbored the codon-optimized cDNA of Hypocrea jecorina AXE (HjAXE), GenBank accession CAA93247 (Margolles-Clark et al., 1996) cloned behind the wood-specific promoter in pK-GT43B-GW7 (Ratke et al., 2015), and were denoted as WP:CE5. Twenty independent transgenic lines were screened in vitro for expression of HjAXE, and the 11 best lines were further screened in the greenhouse. The three most highly expressing lines were finally selected and grown in the greenhouse along with WT control for 8 weeks. The growth conditions were as follows: light photoperiods 18 h [using HQI-TS 400W/DH metal halogen lamps (Osram, Munich, Germany) to supplement daylight when necessary], 20/15°C (day/night) temperatures, and 60–70% relative humidity. The plants were watered daily, fertilized once per week with Rika-S (Weibulls Horto, Hammenhög, Sweden) and shifted weekly to avoid any position effects. Stem height was periodically measured, and the average internode length for internodes 19–35, and the stem diameter for internodes 20 and 40 were determined at the time of harvest.

Total RNA was extracted from developing xylem tissue of hybrid aspen by using the Cetyl Trimethyl Ammonium Bromide (CTAB) extraction method (Chang et al., 1993). The cDNA was synthesized from 1 μg of RNA using a cDNA biosynthesis Bio-Rad kit (Bio-Rad Laboratories AB, Sundbyberg, Sweden). Diluted cDNA (20–30 times) was used for transcript analysis of transgenic plants. The expression was normalized to ubiquitin (Potri.005G198700) and tubulin (Potri.001G464400), and presented relative to the levels in the lowest-expressing line as previously described (Pawar et al., 2017a). The primers of reference and target genes are provided in Supplementary Table S1.

Soluble and wall-bound fractions of proteins were isolated from the developing wood of transgenic and WT hybrid aspen using the method described by Biswal et al. (2014) and tested for acetyl esterase activity using naturally acetylated aspen wood components as esterase substrates (Margolles-Clark et al., 1996). Aspen wood powder (particle size < 50 μm) was suspended in 50 mM sodium citrate buffer (pH 5.0) at 20 g/L and 2 μL of this suspension was incubated with 10 μg of total protein extract in a total volume of 400 μL (in the same buffer) for 24 h at 45°C. After reaction, the mixtures were denatured for 5 min at 100°C, centrifuged briefly and 10 μL of the supernatants were analyzed for the content of acetic acid by using a K-ACET kit (Megazyme, Bray, Ireland). Reaction mixtures containing denatured (10 min at 100°C) instead of fresh protein were used as negative controls. Results are presented as μmol of acetic acid produced from wood powder by 1 mg of protein in 1 h at 45°C. Three trees per each transgenic line and per WT were analyzed.

Wood from internodes 19–35 was freeze-dried, and then ground to a rough wood powder (particle size < 0.5 mm), which was then ball-milled to a fine wood powder as previously described (Gandla et al., 2015). The fine wood powder was analyzed by using Fourier transform infra-red (FTIR) spectroscopy and pyrolysis gas chromatography combined with mass spectrometry (Py-GC/MS) as previously described (Gandla et al., 2015; Pawar et al., 2017a). The data were analyzed by using SIMCA-P (Umetrics AB, Umeå, Sweden).

The acetyl content was determined according to Gille et al. (2011) by saponification of the fine wood powder. The released acetic acid was analyzed by using HPAEC (high-performance anion-exchange chromatography) as previously described (Wang et al., 2018).

Alcohol insoluble residue (AIR) of the fine wood powder was prepared (Pawar et al., 2017a). The AIR was used to determine Klason lignin, acid-soluble lignin (ASL), Updegraff cellulose, and trimethylsilyl (TMS) monosaccharides content of non-cellulosic polysaccharides as described by Gandla et al. (2015).

The rough wood powder was sieved and the fraction with a particle size of 0.1–0.5 mm was used for pretreatment and saccharification. Reaction mixtures containing 50 mg wood (dry weight) and 1% (w/w) sulfuric acid were pretreated at 165°C for 10 min using a single-mode microwave system (Initiator Exp, Biotage, Uppsala, Sweden). Saccharification of pretreated and non-pretreated wood samples was performed by enzymatic digestion of 50 mg wood (dry weight) (or, for pretreated material, the solid residue remaining after the pretreatment of 50 mg wood) using a 1:1 (v/v) mixture of Celluclast 1.5L and Novozyme 188. The load of enzyme protein corresponded to 1 mg per 50 mg wood. The total mass of the reaction mixture was 1000 mg and the medium consisted of sodium citrate buffer (0.5 M, pH 5.2). Reaction mixtures were incubated for 72 h in 2-mL Sarstedt safe-seal micro-centrifuge tubes in an orbital shaker set at 170 rpm and 45°C. Aliquots withdrawn after 2 h were analyzed by using a glucometer (Accu-Chek Aviva, Roche Diagnostics, Risch-Rotkreuz, Switzerland), and data were used to calculate the glucose production rate (GPR). The yields of arabinose, galactose, glucose, mannose, and xylose in pretreatment liquids and in enzymatic hydrolyzates after 72 h incubation were determined using HPAEC, as previously described (Wang et al., 2018).

The surface area of non-pretreated and acid-pretreated sieved rough wood powder (0.1–0.5 mm) was analyzed with a single-point BET procedure using a TriStar 3000 analyzer (Micromeritics, Atlanta, GA, United States). A SmartPrep Degasser (Micromeritics) was utilized prior to the analysis with TriStar 3000 to remove potential adsorbed contaminants. The BET method is based on Langmuir theory and adsorption of nitrogen gas.

Simons’ staining estimates the accessibility of cellulosic materials to enzymes based on solute exclusion (Yu et al., 1995; Arantes and Saddler, 2011). A modified Simons’ staining assay (Chandra and Saddler, 2012) was used to analyze cellulose accessibility. The analysis was performed using the non-pretreated and acid-pretreated sieved rough wood powder (0.1–0.5 mm). Direct Blue (DB, Pontamine Fast Sky Blue 6BX) and Direct Orange (DO, Pontamine Fast Sky Orange 6RN) dyes were obtained from Pylam Products (Garden City, NY, United States).

JMP^® Pro program¹ with analysis of variance (ANOVA) was used for data analysis. Post hoc Dunnett-test was used to compare individual transgenic lines with the WT and contrast test was used to compare all transgenic lines with the WT.

Hybrid aspen lines 11, 14B, and 14C, expressing HjAXE (Margolles-Clark et al., 1996) harboring the plant signal peptide of aspen cellulase PtxtCEL9B3 under control of wood-specific promoter (Ratke et al., 2015), denoted WP:CE5, were grown in the greenhouse for 2 months. The three selected lines had the highest transgene expression from 20 obtained lines as described in Section “Materials and Methods.” Transgene expression in developing wood tissues was approximately two times lower in line 11 than in lines 14B and 14C (Figure 1A). Acetyl esterase activity was studied in wall-bound and soluble protein fractions using naturally acetylated aspen wood as a substrate (Margolles-Clark et al., 1996). This method was considered more specific for xylan acetyl esterases acting on polymeric xylan as compared to methods using synthetic esters as substrates. The activity was detected only in wall-bound protein fraction extracted from developing wood of WP:CE5 lines (Figure 1B). It was varying within the range of 0.56–0.73 μmol mg^–1 protein h^–1, consistent with the transgene expression levels. No activity was detected in wall-bound protein extracts of WT plants. No or negligible activity was recorded in the soluble protein fraction in transgenic lines and no activity was found in this fraction in the WT (Figure 1C). These data indicated that the expressed protein was active as acetyl esterase and associated with cell walls, as expected.

Plant morphology, height growth, and diameter growth of transgenic lines were similar to WT, whereas their internodes were shorter (Figures 1D–F). Since both transgenic and WT plants produced one leaf per internode, this indicates that the transgenic plants produced more leaves. Overall, the results show that the CE5 enzyme is expressed and active in the cell walls of selected transgenic lines without negatively affecting their growth or development.

To investigate potential effects of expression of CE5 AXE on the wood cell wall chemical composition, the contents of lignin and Updegraff cellulose, the yields of trimethyl-silyl (TMS) monosaccharides, and the composition of wood pyrolyzates were analyzed (Table 1). The contents of Klason and acid-soluble lignin, and Updegraff cellulose did not show consistent differences among the transgenic lines compared to WT (Table 1). Analysis using Py-GC/MS (Table 1) showed a small increase in H lignin units (which are a minor lignin component) in transgenic lines. For Py-GC/MS analysis, this was the only statistically significant difference for all transgenic lines. Thus, the wet chemistry and PyGC/MS results showed in agreement no distinct changes in crystalline cellulose or in the main lignin components.

In contrast, the monosaccharide composition did show some changes in transgenic lines compared to WT (Table 1). The xylose (Xyl) content of the transgenic lines was ∼4% lower than in the WT, whereas the glucose (Glc) and rhamnose (Rha) contents were ∼18 and ∼8% higher, respectively. These results indicate changes in the composition of the matrix polysaccharides, a decrease in xylan content, and an increase in glucan in transgenic lines.

To further reveal potential subtle chemical changes in cell walls of WP:CE5 plants, diffuse reflectance Fourier-transform infrared (FTIR) spectroscopy was applied to the ground wood samples. The spectral data analyzed by orthogonal projections to latent structures-differential analysis (OPLS-DA) showed that there was a clear separation between transgenic hybrid aspen lines and the WT (Figure 2A). The bands at 1240, 1370, and 1740 cm^–1, which originate from C-O stretching, CH₂ bending and C = O stretching vibrations (Gorzsas et al., 2011), respectively, all present in acetyl ester groups, strongly contributed to this separation and these signals were less abundant in the transgenic plants than in the WT (Figures 2B,C). This result was similar to that obtained with the previously studied transgenic plants overexpressing AnAXE1 (Pawar et al., 2016, 2017a). On the other hand, the band at 1600 cm^–1, reflecting aromatic C = C vibrations found abundantly in lignin, and the region around 1640 cm^–1, reflecting water, were more abundant in WT plants, which was in contrast to results observed with AnAXE1-expressing Arabidopsis and aspen (Pawar et al., 2016, 2017a). The latter results suggest some changes in lignin structure and cell wall hydration that are specific to CE5 overexpressors.

Quantitation of acetic acid released through saponification of wood showed that the transgenic plants had reduced acetyl content by 10 to 16%, compared to the WT (Figure 3A).

The acetylation was further investigated using 2D HSQC NMR spectroscopic analysis of DMSO-extracted xylan (Supplementary Data Sheets S2, S3), which revealed the presence of acetylated and non-acetylated Xylp residues (Figure 3B). The signals shown in green were used to obtain the relative content of acetylated Xylp units whereas the signals shown in blue represent different non-acetylated Xylp. For the WT, 49% of the total Xylp units were acetylated. That included 22% monoacetylation at position C-2 (X2), 16% monoacetylation at position C-3 (X3), 6% di-acetylation at positions C-2 and C-3 (X23), and 5% acetylation at C-3 and meGlcA (X3G2) (Figure 3C). Transgenic lines exhibited reductions in signals from all acetylated Xylp, by 16% for X23, 14% for X2, 5% for X3G2, and 11% for X3. In total, the content of acetylated Xylp units was reduced by 12% in transgenic lines compared to the WT (Figure 3C). Moreover, there was a 44% increase in the content of non-acetylated Xylp units preceeding either non-acetylated or C-2 acetylated units (X-X/X2). The results suggest that the CE5 enzyme acted on Xylp positions 2 and 3, and could also deacetylate 2,3-double-acetylated Xylp as well as position 3 in glucuronosylated Xylp units of aspen wood xylan.

Moreover, to investigate the changes in the pattern of glucuronoxylan acetylation in the transgenic plants, heat-treated AIR samples from lines 14A and 14B and from the WT were treated with the AaGH10 endo-1,4-β-xylanase, and the released xylo-oligosaccharides (XOS) were analyzed by using oligosaccharide mass profiling (by AP-MALDI-ITMS; Chong et al., 2011). The treatment released acidic XOS with a degree of polymerization (DP) of three to seven and with up to seven acetyl groups (Figure 3D). The transgenic lines showed a prominent change in the distribution of acidic XOS compared to the WT. Whereas the most abundant acidic XOS in the transgenic lines had a DP of three and one acetyl group, the most abundant acidic XOS in the WT had a DP of four and two acetyl groups, although XOS with a DP of five with four acetyl groups and XOS with a DP of six with four or five acetyl groups were also common. In comparison with the WT, the distribution of acidic XOS products with a DP of four and five exhibited a shift toward less acetylated products for the transgenic plants. Furthermore, only the WT yielded noticeable quantities of acidic XOS with a DP of six or seven. This indicates that the glucuronoxylan of the transgenic plants was more accessible to the xylanase treatment and that the main products were less acetylated. The neutral XOS had a DP of two to five and up to five acetyl groups in both genotypes, but the transgenic lines had more DP 2 products without acetyl groups compared to WT (Figure 3D). The results indicate that the CE5 enzyme caused reduced acetylation of xylan for both neutral and meGlcA-substituted domains in glucuronoxylan.

In the screening of the original 11 transgenic lines, the three lines (11, 14B, and 14C) showing the highest transcript levels for the transgene were also the only lines that gave improved glucose yield in enzymatic saccharification of non-pretreated wood, suggesting that a threshold of transgene expression was needed to achieve measurable saccharification gains. Analytical enzymatic saccharification of non-pretreated wood (Figure 4A) resulted in a glucose yield of 137 mg/g for the WT (34.1% glucan recovery based on Updegraff cellulose). For the transgenic lines (average for 11, 14B, and 14C) there was a 27% improvement in glucose yield compared to the WT (glucose yield: 174 mg/g; glucan recovery: 44.6%). Although the transgenic lines contained similar fractions of mannose and galactose as the WT and although they had 5% lower content of xylose, enzymatic saccharification of non-pretreated transgenic lines resulted in higher yields of xylose, mannose, and galactose compared to the WT (11–21% improvement, Figure 4A). Glucose production rate analysis after 2 h of enzymatic hydrolysis also pointed toward higher rates of glucose release for transgenic plants (Supplementary Figure S1). The results show that expression of CE5 AXE improved the digestibility of non-pretreated transgenic aspen with regard to both cellulose and hemicellulose.

With acid pretreatment, the transgenic aspen gave 8% higher glucose yield in the pretreatment liquid than the WT (data not shown), and also 3% higher glucose yield in the enzymatic saccharification (Figure 4B). Most of the xylan was hydrolyzed in the acid pretreatment (data not shown), which resulted in higher xylan recoveries for the transgenic aspen (59.1%) than for the WT (55.6%) even though the total xylose yield was not affected (Figure 4C). When the sugars released in both the pretreatment liquid and the enzymatic hydrolyzate are combined, the WT reached 65.7% glucan recovery and 58.6% xylan recovery, whereas the transgenic aspen reached 70.2% glucan recovery and 61.7% xylan recovery. The results show that the transgenic aspen had improved glucan recovery and yields, and indicate that the xylan in the transgenic aspen was easier to degrade during the pretreatment.

Two methods were employed to compare the properties of the cell wall nanostructure of the transgenic plants and the WT, viz. Brunauer–Emmett–Teller (BET) analysis (Brunauer et al., 1938) and Simons’ staining (Chandra and Saddler, 2012). While BET analysis was carried out using air-dried wood powder or pretreated wood powder, Simons’ staining was carried out using wood powder or pretreated wood powder suspended in water to simulate the conditions during a saccharification reaction.

BET analysis, which measures physical adsorption of nitrogen gas to the solid phase, provided information about the specific surface area (Table 2). On an average, both non-pretreated and pretreated wood of the transgenic plants exhibited larger surface area than the WT. The average surface area was 11% larger for the transgenic plants than for the WT for both materials. The increase of the surface area after pretreatment agrees with the fact that the susceptibility to enzymatic saccharification also increased after pretreatment.

The Simons’ stain assay (Figure 5) is based on two dyes: blue (DB), which has a molecular mass of 993 Da, and orange (DO), which is filtered to obtain its high-molecular-mass (>100 kDa) fraction (Yu et al., 1995). Using these dyes, the accessible surface area and the porosity of the samples can be evaluated. For non-pretreated wood, transgenic plants exhibited 10% higher adsorption of DO than the WT, and 8% higher adsorption of total dyes (Figure 5A). However, this difference between the transgenic plants and the WT disappeared after pretreatment (Figure 5B). Comparing the adsorption ratio DO/DB, the transgenic plants showed 7% higher values than the WT without pretreatment. All samples showed a decreased DO/DB ratio after pretreatment, as the pretreatment resulted in increased adsorption of DB.

Data from both BET and Simons’ staining agree that transgenic plants had an advantage over the WT with regard to structural properties, such as larger surface area and better accessibility. In summary, the activity of the CE5 AXE enlarged the surface area of the transgenic aspen and improved its accessible surface area, which was especially evident for non-pretreated wood.

The transgenic plants expressing the CE5 member HjAXE exhibited decreased xylan acetylation in their wood by 10–16% compared to the WT (Figure 3A), which is comparable with the reductions observed in previously studied plants expressing the CE1 enzyme AnAXE1 (Pawar et al., 2016, 2017a) or plants with suppressed RWA gene expression (Pawar et al., 2017b). While these reductions were relatively modest, they had significant impact on xylan digestibility by endoxylanases. Overexpression of HjAXE prominently affected the distribution of acidic xylo-oligosaccharides released by endoxylanase AaGH10 (Figure 3D). As in AnAXE1-expressing plants and RWA-suppressed plants, the 2D qHSQC NMR signals from HjAXE-expressing plants indicated that Xylp units were deacetylated at both C-2 and C-3 positions of Xylp (Figures 3B,C). This is consistent with previous reports on deacetylation by HjAXE in vitro where activity at positions 2 and 3 was detected, and a preference for monoacetylated Xylp (Biely et al., 1997; Hakulinen et al., 2000; Biely et al., 2011; Neumüller et al., 2015). It was suggested that the deacetylation of glucuronosylated Xylp would be difficult based on the structure of the catalytic site (Hakulinen et al., 2000). In agreement, no such activity was observed in vitro when acetylated oligosaccharides were used as substrates (Neumüller et al., 2015). However, a significant decrease in X3G2 units accompanied by significant decrease in internal Xylp preceeded by X3G2 (X3G2-X) in WP:CE5 (Figure 3C), which was not observed in any of the previously studied transgenic plants, suggested a possibility of deacetylation at C-3 position in the glucuronosylated Xylp units by HjAXE. This conclusion is in line with the observed increase in completely deacetylated 3X + MeGlc xylo-oligosaccharides observed in WP:CE5 lines (Figure 3D), which would be expected assuming that HjAXE could deacetylate glucuronosylated Xylp. While this is a novel observation, these results might reflect endogenous GH10 activity (Derba-Maceluch et al., 2015), which creates a free C-4 position at the non-reducing end enabling the migration of an acetyl group to this position (Mastihubova and Biely, 2004). However, we cannot exclude that HjAXE could exhibit additional specificities when acting on native xylan associated with cellulose microfibrils, especially that their cellulose-binding domain could precondition them on cellulose-bound substrates.

Transgenic plants expressing HjAXE exhibited good height and diameter growth in the greenhouse during a 2-month cultivation period. The only difference from WT was a shorter internode length, which indicates an increased leaf production. It would need to be further studied if such an increase could result in more biomass production by the transgenic lines, and if the field performance of these lines is satisfactory.

Xylan of transgenic plants was shown to be more accessible to hydrolysis by AaGH10 xylanase (Figure 3D), confirming previous observations of synergy between AXEs and xylanases in xylan hydrolysis (Biely et al., 1986). Interestingly, the WP:CE5 lines had reduced Xyl content in matrix polysaccharides (Table 1), indicative of increased hydrolysis of the deacetylated xylan in muro by cell-wall-residing native GH10 enzymes, xylanases and/or transglycosylases (Derba-Maceluch et al., 2015). Similar reductions were observed in other transgenic lines with reduced xylan acetylation (Pawar et al., 2017a, b). If cell-wall-residing xylan was indeed partially hydrolyzed by plant GH10 enzymes, then increased cell-wall porosity, and increased accessibility to cellulose leading to increased glucose yields would be expected. Our analyses demonstrate both.

Increased surface area and porosity of lignocellulose from transgenic plants with reduced acetylation was for the first time demonstrated in this study using BET analysis and Simons’ staining (Table 2 and Figure 5). We speculate that increased surface area and porosity is a direct consequence of removal of cell-wall-residing xylan by endogenous GH10 enzymes in the cell walls of transgenic plants. Interestingly, the decrease in xylan biosynthesis that reduces xylan content is also known to improve saccharification but only when coupled with reduction in xylan chain length (Lee et al., 2011a; Ratke et al., 2018).

A 27% increase in glucose yield and 11% increase in xylose yield in enzymatic saccharification of non-pretreated wood, as well as a 3% increase in glucose yield after acid pretreatment and enzymatic saccharification were observed in WP:CE5 lines (Figure 4). Similarly increased sugar yield of enzymatic saccharification without pretreatment was reported for transgenic aspen and Arabidopsis with reduced xylan acetylation by either supressing native RWA genes or by overexpressing CE1 AXE (Pawar et al., 2016, 2017a,b). Similarly, the positive influence was smaller after acid pretreatment (Pawar et al., 2017a, b). The observed reductions in recalcitrance could be a direct consequence of increased porosity and accessibility due to xylan deacetylation by CE5 AXE acting in close proximity to cellulose.

Moreover, altered xylan acetylation is thought to highly affect cell wall architecture since the acetylation pattern was shown to mediate xylan binding to cellulose microfibrils (Grantham et al., 2017) as well as xylan covalent linkages to lignin (Giummarella and Lawoko, 2016). This could be the basis of the apparently paradoxical observation that increased acetylation in mutant rice (Zhang et al., 2017) and in transgenic poplar (Yang et al., 2017) resulted in improved saccharification. Applying tools to probe cell wall porosity, and cellulose-xylan as well as lignin-xylan interactions in these acetylation-altered plants could possibly give more definite explanations regarding their recalcitrance behavior.