Woody plants are the most abundant source of terrestrial biomass and have been industrially used for construction, paper, energy and many materials and chemicals. Recently, wood has also been considered as a promising sustainable resource for the production of biofuels and other high-value products1. However, the efficient conversion of lignocellulose into simple sugars (glucose and xylose) has been stymied by its complex and recalcitrant structures. Therefore, recent studies have focused on optimizing wood cell wall characteristics and traits for conversion processes including more efficient enzymatic deconstruction using gene modification and metabolic engineering2.

Populus species (poplars, aspens and cottonwoods) are excellent models of woody plant biomass because of their worldwide distribution, fast growth, genomic resources, and technical advances such as the well-established transformation system3. Populus stems include many different cell types which differentiate from the cambium into vessel elements, fibers, and axial and radial parenchyma cells. The dominant material in woody biomass is the secondary xylem cell wall of fiber cells. In these cells, a thin and stretchy primary cell wall (PCW) layer is first deposited during cell expansion. Afterwards, a thick and rigid secondary cell wall (SCW) layer is deposited inside the PCW layer. The PCW is mainly constituted of cellulose, xyloglucan, and pectin with negligible amount of lignin, whereas the SCW consists of cellulose, xylan, mannan and lignin4,5. Genes encoding enzymes involved in the biosynthesis of cell wall components such as cellulose, hemicelluloses and lignin have been relatively well characterized in Populus species (reviewed by Ye and Zhong6).

Expression of the cell wall biosynthetic enzymes is controlled by a number of transcriptional factors (TFs). NO APICAL MERISTEM/ARABIDOPSIS TRANSCRIPTION ACTIVATION FACTOR/CUP-SHAPED COTYLEDON (NAC) and MYB families genes play critical roles in regulating xylem cell differentiation and cell wall thickening7,8. Homologous genes of Arabidopsis thaliana VASCULAR-RELATED NAC DOMAIN (VND)9, NAC SECONDARY WALL THICKENING PROMOTING FACTOR/SECONDARY WALL-ASSOCIATED NAC DOMAIN (NST/SND)10,11 and SOMBRERO (SMB) proteins designated as VNS (VND, NST/SND,and SMB RELATED) proteins12 are key regulators of secondary cell wall formation in xylem fibers, phloem fibers, and xylem ray parenchyma cells in Populus12,13,14,15,16. Among these, VNS09, VNS10, VNS11 and VNS12 corporately play important roles as a master switch regulators of cell wall formation in Populus17,18, but their downstream genes including several TFs are functionally yet uncharacterized. In addition to these studies of A. thaliana TF homologs in Populus species, more comprehensive analyses have focused on TFs expressed during the wood formation processes including secondary xylem differentiation, cell expansion along with PCW deposition, SCW deposition, and programmed cell death along with further lignification19,20,21. Transcriptomic analyses revealed more than 1000 TF genes that were potentially involved in the development of secondary xylem cells19,20. However, most of TFs are uncharacterized, with the exception of major TF families such as NAC and MYB22,23. Furthermore, yet only a few studies have succeeded in improving biomass quality by manipulating Populus TF genes24,25. Thus, our understanding of the TFs involved in the regulatory network system of Populus xylem cell wall formation and the potential for application in a bioengineering context is still limited.

Toward reducing wood recalcitrance and improving its enzymatic degradability, we focused on TF genes associated with wood formation as xylem cell wall modification targets, since most previous studies have focused on genes involved in cell wall biosynthesis or cell wall-degrading/-modifying enzymes in Populus26,27. In comparison, TFs can affect the synthesis of multiple cell wall components and thereby have greater effects. We selected 33 candidate TF genes that are highly up-regulated during wood formation, over-expressed them in hybrid aspen, Populus tremula × Populus tremuloides (Pt × t, wild-type clone T89), and analyzed their saccharification properties.


Selection of target TFs to manipulate xylem cell wall in hybrid aspen

To engineer xylem cell walls of hybrid aspen with improved enzymatic saccharification properties, first we explored the candidate TFs involved in xylem cell wall formation in Populus species. Targets were selected from genes which are downstream of a master switch regulator of SCW formation in Populus, VNS1017. Other targets were selected from genes spatially and temporally expressed during xylem formation19. As a result, a total of 33 TFs (Table 1) including 13 MYBs, 7 NACs and 2 TCPs were selected as target TFs for this study. A series of first and second master regulators such as NST/SNDs, VNDs, MYB002, and MYB02122,23 were eliminated from the target TFs, since these transcriptional activators were known to strongly modify SCW structures, affect plant growth negatively, and possibly lower biomass saccharification performance when overexpressed in planta17. Among the 33 TFs, MYB003 and MYB152 were characterized as transcriptional activators of SCW biosynthesis genes28,29,30, and MYB199 and KNAT7 were identified as transcriptional repressor for SCW formation in Populus31,32. The function of the other 29 TFs remains to be elucidated.

Table 1 List of transcription factors selected in this study.

To further characterize the 33 target TFs, their gene expression patterns during plant developments33 were analyzed. In short, almost all the TFs were highly expressed during xylem developmental processes in Populus (Figure S1). Next, using the AspWood database21, hierarchical cluster analysis of their gene expression during wood formation was analyzed together with 40 other cell wall synthesis-related genes26,27. This analysis revealed that the target TF genes were divided into three clusters (I, II and III; Figure S2). Cluster I contained 22 target TFs including previously characterized TFs such as Pt × tMYB003, Pt × tMYB152, Pt × tMYB199 and Pt × tKNAT728,29,30,31,32. Members of cluster I were highly expressed during SCW formation and wood fiber maturation. Cluster I also included 25 SCW biosynthetic enzymes associated with cellulose biosynthesis [five CELLULOSE SYNTHASEs (CESAs); CESA4, CESA7-A, CESA7-B, CESA8-A and CESA8-B], hemicellulose biosynthesis [six GLYCOSYLTRANSFERASEs (GTs); GT8D-1/IRREGULAR XYLEM 8 (IRX8), GT43A/IRX9, GT43B/IRX9, GT43C/IRX14, GT47A-1/IRX10 and GT47C/ /FRAGILE FIBER 8 (FRA8)], or lignin biosynthesis [p-COUMARATE 3-HYDROXYLASE (C3′H1), CINNAMATE-4-HYDROXYLASEs (C4H1 and C4H2), CINNAMYL ALCOHOL DEHYDROGENASE (CAD1), CAFFEOYL-CoA O-METHYLTRANSFERASEs (CCoAOMT1, CCoAOMT2 and CCoAOMT3), CINNAMOYL-CoA REDUCTASE (CCR7), CAFFEIC ACID O-METHYLTRANSFERASEs (COMT1 and COMT2), FERULATE 5-HYDROXYLASEs (F5H1 and F5H2), and p-HYDROXYCINNAMOYL-CoA:QUINATE/SHIKIMATE p-HYDROXYCINNAMOYLTRANSFERASEs (HCT1 and HCT6)]. In cluster II, six TFs (Pt × tERF123, PniMYB055, Pt × tTCP24, Pt × tWLIM2B, Pt × tPNAC161, and Pt × tMYB148) were induced at the initial stage of the xylem development. These TFs showed similar gene expression patterns to 14 PCW-related genes associated with cellulose biosynthesis (CESA1-A, CESA1-B, CESA3-A, CESA6-A and CESA6-B), xyloglucan biosynthesis [XYLOSYLTRANSFERASE (XXT1A) and CELLULOSE SYNTHASE-LIKE C-R (CSLC-R)], or pectin biosynthesis [GALACTURONOSYLTRANSFERASEs (GAUT1A and GAUT1B), GALACTURONOSYLTRANSFERASE-LIKEs (GATL3A and GATL3B], GALACTOSYLTRANSFERASEs (GALS1,GALS2A, and GALS2B]. Five TFs (Pt × tDOF4.6, Pt × tMYB192, Pt × tPNAC069, Pt × tMYB212, and Pt × tTCP3) were categorized into cluster III, which were highly expressed at later stage of xylem development and phloem tissues.

Screen of 33 TF-overexpressed poplar seedlings by enzymatic saccharification

The selected 33 TFs, fused with TagRFP at the C-terminal, were overexpressed in hybrid aspens by using a pGWB560 vector harboring the nuclear localization sequence (NLS), and pGWB560 harboring NLS tagged double-TagRFP (NLS-TagRFP) was used as a vector control. Expression of the introduced genes was confirmed by detection of the TagRFP-tag with real-time PCR analysis. TagRFP region of vector control plants was also amplified; we confirmed that no amplicon was detected in the wild type of T89 plants. Relative expression (to Pt × tUBQ) of the introduced genes ranged from 0.001 to 5 as shown in Fig. 1A. The harvested seedling stems of the successful transgenic hybrid aspens grown in a MS culture under aseptic condition were treated with a cell-wall degrading enzyme cocktail for 48 h and the released glucose amounts were measured by colorimetric methods via glucose oxidase reactions to calculate saccharification efficiency (Fig. 1B).

Figure 1
figure 1

Relative expression level of the transgene (TF-TagRFP) in transgenic hybrid aspens measured by real-time PCR using primer pairs specific to the RFP (A). pGWB560 harboring the NLS-TagRFP was used as a vector control. TagRFP regions fused with each introduced gene as well as vector control were amplified. No amplicon was detected in wild type of T89 plants. Released glucose amounts from biomass (wt/wt %) of seedling stems of the transgenic hybrid aspens grown in MS culture under aseptic condition after 48 h-treatment with enzyme cocktail (B). A series of TFs were sorted by average of released glucose (%) from different lines generated from the same genotype (filled circle).

Most of the tested transgenic aspens overexpressing a variety of TFs involved in SCW formation, such as MYB, NAC and KNOX proteins, showed a decline in the amount of released glucose compared to control. In contrast, four of the TF overexpressing lines exhibited increased release of glucose (Fig. 1B). All four lacked functional characterization in Populus previously: (1) Pt × tERF123ox overexpressing ETHYLENE RESPONSE FACTOR 123 (Pt × tERF123)34 exhibited almost 1.5-times glucose release (28.45% ± 7.05, mean ± s.d.), compared to the control (20.90% ± 0.37); (2) Pt × tZHD14ox overexpressing ZINC FINGER HOMEODOMAIN 14 (Pt × tZHD14)35 had ~ 1.3 times increase (26.45% ± 5.35) compared to the control; (3) Pt × tTCL1ox overexpressing TRICHOMELESS 1 (Pt × tTCL1)36, a MYB-related protein (a single repeat R3 MYB protein apart from R2R3 MYB family described above) showed the average of glucose amounts released from biomass (23.84% ± 3.40), which were also greater than that of control. The different lines (#7, #8, and #12) showed 23.46%, 20.65% and 27.42%, respectively. Considering the relative expression levels of the corresponding lines (0.0098, 1.4033 and 0.0086, respectively), the high ectopic expression of Pt × tTCL1 is likely the cause of the poor performance of line #8; (4) Pt × tWLIM2Box overexpressing WIDELY EXPRESSED LIN-11, Isl1 and MEC-3 (LIM)-STRUCTURAL DOMAIN PROTEIN 2B (Pt × tWLIM2B)37 showed a significant increase in two of the lines (#18 = 22.73% and #19 = 27.47%, respectively) but a significant decrease in line #15 (15.08%), which was also likely due to the high ectopic expression of Pt × tWLIM2B.

Enzymatic saccharification of Pt × tERF123ox and Pt × tZHD14ox grown in a greenhouse

To assess the enzymatic digestibility of mature xylem tissues, Pt × tERF123ox and Pt × tZHD14ox, the two transgenic aspen lines which showed the largest glucose releases in the initial saccharification screening as described above, were grown in a soil in a greenhouse with at least 3 biological replicates for further characterizations. In short, no significant phenotypic difference in stem heights and diameters was observed between the TF-overexpressing lines and the controls, except for Pt × tZHD14ox line #5 displaying apparently shorter stem heights than the controls (Fig. 2A–C). Based on anatomical stem sections stained with toluidine blue, xylem cell size and cell wall thickness were not changed between Pt × tERF123ox lines and the controls, except for Pt × tERF123ox line #3 (Fig. 2D–F). In case of Pt × tZHD14ox, compared to the controls, the xylem cell walls were thicker and thinner in line #1 and line #5, respectively (Fig. 2D–F). Overall, no apparent loss of biomass quantity was observed between the TF-overexpressing lines and the controls except for Pt × tZHD14ox line #5.

Figure 2
figure 2

Growth of Pt × tERF123ox, and Pt × tZHD14ox lines grown in the soil. Phenotypes of 79- and 69-days-old transgenic hybrid aspens compared with the control (expressing GFP-TUA6 alone) (A). Scale bar = 10 cm. Height (B) and diameter (C) of transgenic hybrid aspens and control plants. Cell wall thickness (D) and cell size (E) of wood fiber cells. Three hundred fiber cells were estimated from three trees per each genotype. Single and double asterisk(s) indicate P value < 0.05 and < 0.01, respectively in Student’s t- test when compared with the control plants. Anatomical observations of the stem sections stained with toluidine blue (F). Scale bar = 10 μm.

Consistent with the results obtained from the screening of seedling tissue (Fig. 1), the amount of glucose released from xylem biomass of both Pt × tERF123ox and Pt × tZHD14ox plants (except for Pt × tZHD14ox line #5) were increased after 48 h-treatment with the enzyme cocktail as compared to the controls (Fig. 3B). For Pt × tERF123ox, the descending order of the glucose release was line #8 > #3 > #1. In particular, the initial hydrolysis rates of the transgenic poplar biomass were significantly increased (Fig. 3A). In addition to glucose release, the amounts of xylose released after 48 h-treatment with the enzyme cocktail were also measured. Notably, xylose released from both Pt × tERF123ox and Pt × tZHD14ox biomass, again except for Pt × tZHD14ox line #5, were significantly increased (Fig. 3C). To further assess the enzymatic digestibility, released amount of glucose and xylose from a total of glucan (non-crystalline and crystalline glucan) and xylan (Table 2) were calculated as glucose yield and xylose yield (Fig. 3D,E), respectively. Interestingly, all the xylose yields of Pt × tERF123ox and Pt × tZHD14ox lines were approximately 1.5–2 times higher than the controls whereas the enhancement of glucose yield exhibited less significance. Collectively, these data suggested that overexpression of Pt × tERF123 and Pt × tZHD14 led to improved biomass digestibility most likely via alterations of xylem cell wall structure. Notably, there were some differences in the enzymatic digestion of the poplar seedling and mature xylem biomass tested. For example, glucose releases after 48 h enzymatic hydrolysis of the mature xylems were not increased as much as those of the seedlings. These different enzymatic saccharification performances between the seedling and mature xylem biomass might be due to the differences in the sample part analyzed (i.e., seedlings include xylem as well as phloem and pith whereas bark and pith were removed from mature xylem), xylem developmental stages and plant cultivation conditions (i.e., growth chamber vs greenhouse conditions).

Figure 3
figure 3

The enzymatic saccharification of Pt × tERF123ox, and Pt × t ZHD14ox lines grown in the soil, in comparison to the controls (GFP-TUA6). Released glucose (A,B) and xylose (C) amounts from biomass (wt/wt %) after 3 h (A) or 48 h (B,C)-treatments with cellulase cocktail. Glucose and xylose yields (D,E) calculated based on a total of glucan (non-crystalline glucan and crystalline glucan) and xylan measured in Table 2. Data are shown as mean ± standard deviations of biological triplicates. Single and double asterisk(s) indicate P value < 0.05 and < 0.01, respectively in Student’s t- test when compared with the control plants.

Table 2 Cell wall lignin and polysaccharide analyses of Pt × tERF123ox and Pt × tZHD14ox transgenic hybrid aspens grown in green house.

Chemical composition of Pt × tERF123ox and Pt × tZHD14ox mature xylem cell walls

To investigate the cause of the enhanced saccharification performance of Pt × tERF123ox and Pt × tZHD14ox, lignocellulose composition and structures of their xylem cell wall samples were investigated by a series of chemical analyses (Table 2). Lignin content of the aspen cell wall samples were determined by thioglycolic acid assay38, and lignin composition, i.e., syringyl (S), guiacyl (G) and p-hydroxyphenyl (H) aromatic unit ratio, was determined by analytical thioacidolysis39. In addition, cell wall polysaccharide composition was determined by quantification of monomeric sugars released via a two-step acid-catalyzed hydrolysis method39,40.

The cell wall samples from two of the three Pt × tERF123ox lines, i.e., lines #1 and #8, showed significantly reduced lignin contents, by 9.1% and 6.4%, respectively, compared to the controls, whereas the other line #3 also showed a decreased lignin level, by 6.1%, albeit with no statistical significance. In addition, all the three Pt × tERF123ox lines showed tendencies toward decreased xylan and mannan, and increased cellulosic crystalline and amorphous glucan levels. No significant change in the S/G/H ratio was observed between all the Pt × tERF123ox and control cell wall samples (Table 2). Overall, these chemical analysis data suggested that the overexpression of Pt × tERF123 may lead to increased glucan and reduced xylan barrier and lignin recalcitrance and thereby boost enzymatic hydrolysis of both cellulose and xylan polysaccharides.

In contrast to the Pt × tERF123ox lines, the lignin content of the cell wall samples from the Pt × tZHD14ox lines appeared to be comparable or even higher than that of the control cell walls. Cellulosic crystalline glucan contents of the Pt × tZHD14ox cell walls seemed to be comparable or lower than that of the controls (Table 2). The two independent Pt × tZHD14ox lines, i.e., lines #1 and #5, both showed reduced mannan, and increased galactan and amorphous glucan levels. Meanwhile, similar to the Pt × tERF123ox lines, we observed no significant change in the S/G/H lignin unit ratio of the Pt × tZHD14ox cell walls compared to the control (Table 2). Collectively, in contrast to our observation with Pt × tERF123ox, the improved saccharification performance of Pt × tZHD14ox is not apparently associated with changes to the glucan, xylan and lignin contents of the cell wall.

Expression of lignin and cellulose biosynthetic genes in Pt × tERF123ox and Pt × tZHD14ox

To further investigate the cause of the lignocellulose component alterations in Pt × tERF123ox and Pt × tZHD14ox, we assessed the changes in the expression of lignin and cellulose biosynthetic genes in Pt × tERF123ox and Pt × tZHD14ox. Expression levels of 22 lignin biosynthetic genes, including those encoding phenylalanine ammonium (PAL), 4-coumarate-CoA ligase (4CL), C4H, CCR, CAD, HCT, C3′H, CCoAOMT, F5H, COMT, cafferoyl shikimate esterase (CSE), and also 7 CESA genes involved in cellulose biosynthesis were measured by real time PCR using RNA extracted from the seedling stems as samples (Fig. 4). In case of Pt × tERF123ox lines, compared with control lines, the relative expression of most of the lignin biosynthetic genes tested, except for PAL1 and PAL3, were comparable or lower albeit without statistical significance, which was overall in line with the reduced lignin content (Table 2). In contrast, a series of CESA genes appeared to be apparently upregulated in Pt × tERF123ox. In particular, CESA1-B, CESA3-A, CESA6-A and CESA6-B associated with PCW formation41 as well as CESA8-B associated with SCW formation12,42 were significantly upregulated (Fig. 4). The fold changes of the CESA expression were in a descending order of line #8 > #3 > #1, which was apparently correlated with the cellulose content and saccharification performance determined earlier (Table 2, Fig. 3). In case of Pt × tZHD14ox line #1, relative expressions of cellulose and lignin biosynthetic genes were comparable or slightly higher in Pt × tZHD14ox lines. On the other hand, in Pt × tZHD14ox line #5, expression levels of several lignin biosynthetic genes including 4CL, C4H, CCR, CAD, HCT, CCoAOMT and CSE were significantly reduced, whereas gene expression of CESA8-B for cellulose biosynthesis was significantly increased, compared to the control lines.

Figure 4
figure 4

Relative expression level of a total of 29 genes involved in lignin and cellulose biosynthesis in seedling stems of Pt × tERF123ox, and Pt × t ZHD14ox lines grown for one-month in MS culture under aseptic condition, in comparison to the controls (GFP-TUA6). Expressions were measured by real-time PCR by using gene specific primer pairs listed in Table S1, expression levels between samples were normalized by using 18S ribosomal RNA gene expression, and then relative expressions were calculated by using expression levels in control as 1. Data are shown as mean ± standard deviations of biological triplicates. Single and double asterisk(s) indicate P value < 0.05 and < 0.01, respectively in Student’s t- test when compared with the control plants.


In this study, we successfully generated and screened transgenic hybrid aspen overexpressing 33 TFs with a potential role in regulating xylem cell wall biosynthesis (Table 1; Fig. 1A). Four of the previously uncharacterized TFs (Pt × tERF123, Pt × tZHD14, Pt × tTCL1 and Pt × tWLIM2B) had increased saccharification of seedling tissue on average when overexpressed (Fig. 1B). Among them, all lines of Pt × tERF123ox and Pt × tZHD14ox which showed the excellent saccharification performance at the seedling screening stage were further grown to form mature xylem in the greenhouse. Consequently, most of the tested Pt × tERF123ox and Pt × tZHD14ox lines displayed significantly increased initial glucan hydrolysis rates (glucose release after 3 h-treatment of enzyme cocktail) albeit with less significant enhancement in the total glucan degradation (glucose release after 48 h-treatment of enzyme cocktail) by enzymatic saccharification (Fig. 3). Notably, all the Pt × tERF123ox and Pt × tZHD14ox lines displayed increased xylan degradations (xylose release after 48 h-treatment of enzyme cocktail) with 1.5–2.0 times higher xylose yields compared to the controls. Xylan is thought to be one of the limiting factors in enzymatic hydrolysis of cellulose possibly because xylan physically covers cellulose surface in lignocellulose and thereby hinders the access of cellulolytic enzymes to cellulose substrate43. Therefore, easier removal of xylan may have enhanced cellulose hydrolysis of the Pt × tERF123ox and Pt × tZHD14ox cell walls.

Many studies have investigated the relationship between lignocellulose structure and enzymatic saccharification performance of cell wall materials. Major factors that affect cell wall saccharification performance are thought to be glucan contents as well as the recalcitrant lignin content and structure which substantially prevents the access of hydrolytic enzymes to cellulose and hemicellulose substrates and subsequent hydrolysis reactions44,45. Indeed, our chemical analysis data showed that Pt × tERF123ox displayed relatively increased glucan and significantly reduced lignin content, albeit with no apparent change in lignin aromatic composition, compared to the control (Table 2). Thus, it is plausible that the improved saccharification performance of Pt × tERF123ox is due to the reduced lignin recalcitrance, besides the reduced xylan barrier. The alteration of cell wall composition in Pt × tERF123ox, i.e., reduced lignin and increased cellulosic glucan levels, could be attributed to the up-regulation of a series of cellulose synthase genes upon overexpression of Pt × tERF123 (Fig. 4). Previously, several Populus ERFs were reported to be involved in the formation of tension wood (TW) which typically produce cell walls with reduced lignin and increased cellulose compared to normal wood cell walls34. In addition, a member of ERF gene family in A. thaliana, AtERF035, was recently identified as an active regulator of PCW formation; it was demonstrated that overexpression of AtERF035 substantially enrich pectin and cellulose proportionally over lignin in the cell walls through induction of PCW formation41. Intriguingly, Pt × tERF123 was significantly up-regulated during the cell expansion together with PCW-associated enzyme genes (Figure S2), and CESA1-B, CESA3-A and CESA6-A associated with PCW formation were significantly upregulated (Fig. 4). Therefore, it is possible that the altered lignocellulose composition detected in the Pt × tERF123ox cell walls (Table 2) might be associated with cell wall alteration analogous to TW and/or PCW formation(s), although our current histochemical and cell wall chemical analysis data do not entirely corroborate this hypothesis (Fig. 2 and Table 2). Further rigorous analyses on the transcription network associated with Pt × tERF123 as well as cell wall structures of Pt × tERF123-overexpressing and/or -downregulated transgenic aspens are needed to clarify this aspect.

In contrast to Pt × tERF123ox, the improved saccharification of Pt × tZHD14ox cell walls is unlikely to be associated with compositional change of lignin, xylan and cellulose in the cell walls (Table 2). Besides lignin content and structure, several other factors, such as cellulose crystallinity, hemicellulose modifications, and covalent and non-covalent linkages between lignin and polysaccharides, have been proposed as potential factors that may affect cell wall saccharification performance46,47. Therefore, future study will need to focus on further structural analyses on the Pt × tZHD14ox cell walls to investigate these factors. In this context, we note that a homologous gene of Pt × tZHD14 in A. thaliana (AtZHD9/AtHB34) was expressed in various tissues including inflorescence stems35,48, but its function remains unclear. In this study, we observed that cell wall thickness was fluctuated along with the introduced expression level of Pt × tZHD14 (Fig. 2E), and contents of some hemicellulosic sugars in cell walls were also affected in the Pt × tZHD14ox poplar lines (Table 2). Given that Pt × tZHD14 is substantially up-regulated during SCW formation together with SCW biosynthetic genes (Figure S2, Fig. 4), it is plausible that Pt × tZHD14 is involved in xylem development in aspen. Nevertheless, as also noted for Pt × tERF123, further detailed analysis on the transcription network associated with Pt × tZHD14 is needed to elucidate its role in xylem cell wall development in Populus.

While we found 4 promising TF-overexpressing hybrid aspens with improved cell wall saccharification performance, the other 29 TF-overexpressing lines showing decreased saccharification performance (Fig. 1B) are also of interest for further investigation of their potential role in xylem development in Populus. As previously reported, overexpression of both strong transcriptional repressors and/or activators of SCW are expected to cause a decrease of enzymatic digestibility along with striking loss or accumulation of secondary cell wall in the xylem. For example, overexpression of the strong repressors of SCW formation, KNAT7 and MYB199, were reported to result in remarkably thin xylem Populus cell walls31,32. Also, in the case of strong activators, overexpression of MYB003 and MYB152 in planta deposited ectopic and/or thick cell walls28,29. Overall, our results suggested that only approximately 10% of the selected TF up-regulated during wood formation increased enzymatic saccharification, therefore, the first seedling screening method in this study was effective to find transgenic lines with the improve enzymatic digestibility. In Arabidopsis, a similar approach successfully identified several cell-wall-related TF genes that increase cell wall saccharification efficiency49.

The cultivation and the application of genetically modified organisms are controversial. Recent technology of genome editing has drawn attention to solve this problem since it will enables modulations of gene expression and function without leaving foreign gene in a cell50. For example, gene expression can be enhanced by altering cis sequence in the upstream region of associated TF genes, or by silencing negative effectors, and also protein function can be modulated by amino acid substitution. The biomass-recalcitrance-associated TFs identified in this study can be considered as promising targets to improve poplar cell wall properties using such genome editing strategies. Moreover, given that biomass conversion often needs pre-treatment before enzymatic saccharification1, future study may further investigate the saccharification performance of the TF-overexpressing aspen lines identified in this study using various pretreatment strategies.

Collectively, our strategy to target and screen an array of TFs up-regulated during wood formation successfully constructed transgenic hybrid aspens with reduced xylem cell wall recalcitrance of xylem cell walls without apparent biomass loss. The observation that all cell wall components were coordinately changed in the transgenic aspens in this study is likely due to our target TFs regulating xylem cell wall formation at a global level, not a single cell wall component. Furthermore, the identified TF genes such as Pt × tERF123 and Pt × tZHD14 are widely distributed in various angiosperms34,35, therefore, these may be new molecular breeding targets for the improved enzymatic digestibility of various biomass, especially other woody biomass crops, such as eucalyptus and willow.


Plant materials and growth

Sterile plants of P. tremula × P. tremuloides (wild-type clone T89) were propagated in half strength Murashige and Skoog (MS) medium (pH 5.7) containing 0.8% (w/v) agar at 25 °C under long-day conditions (18-h light at 200 µmol m-2 s-1/6-h dark)42. Hybrid aspens were transplanted into soil mixture (3:1 fertilized peat moss:vermiculite, v/v) and grown in a greenhouse at 20 °C under long-day conditions. Plants were fertilized once a week with 2000-fold diluted Hyponex 6–10-5 solution (HYPONeX Japan Corp., Ltd., Osaka, Japan). Plant height and stem diameter at 10 cm above the soil was measured weekly.


Gene expression patterns of the selected 33 TFs were investigated in the Populus tissue expression data33 and the AspWood database ( 21. We re-analyzed raw counts of RNA sequencing dataset (GSE81077)33 retrieved from NCBI Gene Expression Omnibus. Genes with at least one count-per-million (cpm) in all samples were retained and normalized with trimmed mean of M-value (TMM) using edgeR, ver. 3.18.151 in R software, ver.3.3.252. To perform gene clustering analysis, we retrieved primary cell wall- and secondary cell wall-related genes from the genome database of Populus trichocarpa ( The expression patterns of 33 TFs and cell wall-related genes were analyzed in the AspWood database and visualized by expression heatmap.

Generation of transgenic hybrid aspens

cDNAs of the target TF coding sequences were isolated from P. tremula × P. tremuloides or Populus nigra as described below, using the primers listed in Table S1. The coding sequences (without stop codon) were cloned into the pENTR/D-TOPO vector (Thermo Fisher Scientific, Waltham, MA, USA), and then transferred to the pGWB560 vector (C-terminal fusion with TagRFP) by the Gateway LR reaction53. pGWB560 harboring the NLS-TagRFP was used as a control vector. The constructed binary vectors were introduced into a hybrid aspen, which expresses soluble-modified RED-SHIFTED GREEN FLUORESCENT PROTEIN (smRSGFP)-tagged Arabidopsis thaliana ALPHA TUBULIN-6 (TUA6) (GFP-TUA6) driven by Cauliflower Mosaic Virus 35S (CaMV35S) promoter, by Agrobacterium-mediated transformation54,55. Ectopic expression of GFP-TUA6 did not cause any observable effects on plant growth or xylem structure, such as the length and width of wood fibers and vessel elements (Figure S3).

RNA extraction and real-time PCR

To check relative expression level of the transgene (TF-TagRFP) in transgenic hybrid aspens, leaves were harvested from sterile plants grown for one month, frozen in liquid nitrogen, and total RNA was isolated using an RNeasy Plant Mini Kit (Qiagen, Hilden, Germany) with in-column DNase I digestion. First-strand cDNA was synthesized using a High Capacity RNA-to-cDNA Kit (Thermo Fisher Scientific). Real-time PCR was performed using a StepOnePlus Real-Time PCR System with Power SYBRGreen PCR Master Mix (Thermo Fisher Scientific). Expression of the introduced genes was measured using TagRFP region-amplified primers (Table S1). A primer pair designed to target Pt × tUBQ was used as an internal standard (Table S1). No amplicon was detected in the control T89 and GFP-TUA6 plants when using the TagRFP primer pair in real-time PCR analysis.

To measure relative expression level of genes involved in lignin and cellulose biosynthesis in Pt × tERF123ox, Pt × t ZHD14ox, the seedling stems were harvested from sterile plants grown for one month and stored in liquid nitrogen. RNA extraction and cDNA synthesis were conducted as described above, and expressions were measured by real-time PCR by using gene specific primer pairs for 22 lignin biosynthetic genes and 7 cellulose biosynthetic genes listed in Table S1. A primer pair designed to 18S ribosomal RNA was used as an internal standard (Table S1). Relative expressions were calculated by using average of expression levels of control lines as a standard (relative expression level of control = 1.0).

Enzymatic saccharification of transgenic hybrid aspens

For transgenic hybrid aspen grown in aseptic conditions, stem tissues were lyophilized and ground into fine powder using a crusher (µT-12, TAITEC Corporation, Saitama, Japan) at 2000 rpm for 30 s, repeated three times. For transgenic aspen grown in the greenhouse, harvested plant tissues were lyophilized and debarked, and the xylem tissues were ground into fine powder as described above. The saccharification assay was modified from the NREL protocol56. Briefly, powdered sample (10 mg) was mixed with 1 mL of 1 mg/mL enzyme cocktail in 50 mM sodium acetate buffer (pH 5.0) with 0.1% ampicillin. The mixture was incubated at 50 °C for 48 h with inversion mixing by rotator at approximately 30 rpm (RT-50, TAITEC). The enzyme cocktail included Cellulase from Trichoderma reesei ATCC 26921 (Merck KGaA, Darmastat, Germany) and Cellobiase from Aspergillus niger (Novozyme 188, Merck) with a protein concentration ratio of 4:1. The strain of Trichoderma reesei ATCC26921 is a well-known enhanced cellulase-producing mutant as QM9414 derived from a wild-type strain of QM6a57. The enzyme mixture includes cellulolytic enzymes such as cellobiohydrolase, endoglucanase and beta-1,4-glucosidase as well as hemicellulolytic enzymes such as mannanase and xylanase, xylosidase, arabinofuranosidase, and other enzymes to release glucose and xylose from cellulosic biomass58. To accurately measure released glucose, beta-1,4-glucosidase (cellobiase) was added to the cellulase mixture. Additional xylanase loading in the enzyme cocktail did not increase the glucose release from poplar seeding stem samples. The amount of released glucose and xylose in the supernatant was measured using the LABASSAY GLUCOSE (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) and D-Xylose assay kits (, Megazyme Ltd., Wicklow, Ireland), respectively.

Anatomical observation of hybrid aspen xylem tissues

Stem samples were harvested from the 20th internode and fixed in FAA solution (50% ethanol, 10% formaldehyde, and 5% acetic acid). Trimmed stem segments were dehydrated in a graded ethanol series (50%, 60%, 80%, 90%, and 100%), and were embedded in LR White Hard resin (TAAB, Aldermaston, UK) with 5% PEG400. Cross Sections (2-µm thick) were cut using a rotary microtome (RX-860; Yamato Kohki Industrial, Saitama, Japan) and then stained with a 0.5% (w/v) toluidine blue solution. Sections were imaged using a Leica DMi8 microscope with a Leica DFC7000T microscope camera (Leica Microsystems, Wetzlar, Germany). Cell wall thickness and cell size of wood fiber cells were measured using the ImageJ software (n = 100 for each plant)59,60.

Chemical analysis of hybrid aspen xylem tissues

The ground xylem powder was extracted sequentially by water and 80% (v/v) ethanol to generate extractive-free cell wall residues61. The thioglycolic acid lignin content assay38 and thioacidolysis lignin composition analysis38,39 were carried out as described previously. For determination of the non-cellulosic cell wall polysaccharide composition, the extractive-free cell wall residues were first hydrolyzed by trifluoroacetic acid. The released monosaccharides were converted into alditol acetates and subjected to analysis by gas chromatography–mass spectrometry with inositol acetate as an internal standard39,40. To determine crystalline glucan content, the remaining residues were treated with Updegraff reagent62, followed by complete hydrolysis using 72% (v/v) sulfuric acid63. Glucose released was quantified by the Glc CII test kit (FUJIFILM Wako Pure Chemical Corporation).