Genome-edited rice deficient in two 4-COUMARATE:COENZYME A LIGASE genes displays diverse lignin alterations

Abstract

The 4-coumarate:coenzyme A ligase (4CL) is a key enzyme that contributes to channeling metabolic flux in the cinnamate/monolignol pathway, leading to the production of monolignols, p-hydroxycinnamates, and a flavonoid tricin, the major building blocks of lignin polymer in grass cell walls. Vascular plants often contain multiple 4CL genes; however, the contribution of each 4CL isoform to lignin biosynthesis remains unclear, especially in grasses. In this study, we characterized the functions of two rice (Oryza sativa L.) 4CL isoforms (Os4CL3 and Os4CL4) primarily by analyzing the cell wall chemical structures of rice mutants generated by CRISPR/Cas9-mediated targeted mutagenesis. A series of chemical and nuclear magnetic resonance analyses revealed that loss-of-function of Os4CL3 and Os4CL4 differently altered the composition of lignin polymer units. Loss of function of Os4CL3 induced marked reductions in the major guaiacyl and syringyl lignin units derived from both the conserved non-γ-p-coumaroylated and the grass-specific γ-p-coumaroylated monolignols, with more prominent reductions in guaiacyl units than in syringyl units. In contrast, the loss-of-function mutation to Os4CL4 primarily decreased the abundance of the non-γ-p-coumaroylated guaiacyl units. Loss-of-function of Os4CL4, but not of Os4CL3, reduced the grass-specific lignin-bound tricin units, indicating that Os4CL4 plays a key role not only in monolignol biosynthesis but also in the biosynthesis of tricin used for lignification. Further, the loss-of-function of Os4CL3 and Os4CL4 notably reduced cell-wall-bound ferulates, indicating their roles in cell wall feruloylation. Overall, this study demonstrates the overlapping but divergent roles of 4CL isoforms during the coordinated production of various lignin monomers.

Loss-of-function mutations of two 4-COUMARATE:COENZYME A LIGASE genes in rice result in distinctively different lignin compositional changes, demonstrating their diverse roles in lignin biosynthesis.

Introduction

The cinnamate/monolignol pathway produces diverse phenylalanine- and tyrosine-derived secondary metabolites that are vital for plant development and survival. One of the major end-products of this pathway is lignin, which is an abundant structural polymer in the secondary cell walls of vascular plants (Boerjan et al., 2003; Bonawitz and Chapple, 2010; Umezawa, 2010; Vogt, 2010; Barros et al., 2015; Umezawa, 2018; Dixon and Barros, 2019; Vanholme et al., 2019). Diverse metabolites derived from the pathway can serve as lignin monomers that combine to form complex and heterogeneous lignin polymers in cell walls via oxidative radical coupling (Freudenberg and Neish, 1968; Sarkanen and Ludwig, 1971; Higuchi, 1985; Ralph et al., 2004; Tobimatsu and Schuetz, 2019). The formulation of lignin monomers, which is the primary determinant of the structure and properties of lignin polymers, varies among the classes of vascular plants (Ralph et al., 2019). In general, lignins produced by most ferns, gymnosperms, and eudicots are derived mainly from monolignols, namely coniferyl, sinapyl, and p-coumaryl alcohols, which give rise to guaiacyl (G), syringyl (S), and p-hydroxyphenyl (H) units, respectively, in the final lignin polymers (Figure 1). On the other hand, lignins produced by monocotyledonous grasses (Poaceae) additionally use monolignol p-hydroxycinnamate conjugates, mainly γ-p-coumaroylated monolignols (coniferyl and sinapyl p-coumarate) (Figure 1) as well as γ-feruloylated monolignols (coniferyl and sinapyl ferulate), albeit at much lower amounts (Ralph et al., 1994; Ralph, 2010; Karlen et al., 2018). Furthermore, grass lignins include tricin, a member of the flavonoids. Tricin is incorporated into lignin polymer via radical coupling with monolignols and γ-p-coumaroylated monolignols, resulting in tricin-lignin (flavonolignin) units in the final lignin polymer (Figure 1; del Río et al., 2012; Lan et al., 2015; del Río et al., 2020; Lam et al., 2021). Accordingly, in grasses, there is substantial complexity in the use of the different lignin monomers derived from the cinnamate/monolignol pathway along with the flavonoid biosynthetic pathway branching from it (Figure 1). Because of the considerable interest in applying molecular breeding and bioengineering approaches to control lignin biosynthesis and increase the utility of the abundant grass biomass, the regulatory mechanisms of the cinnamate/monolignol pathway and the branching flavonoid biosynthetic pathway that coordinate the production of diverse lignin monomers in grasses have been the focus of the related research (Umezawa, 2018; Dixon and Barros, 2019; Halpin, 2019; Coomey et al., 2020; del Río et al., 2020; Umezawa et al., 2020; Lam et al., 2021).

Figure 1.

Proposed lignin biosynthetic pathways in grasses. PAL, phenylalanine ammonia-lyase; PTAL, phenylalanine/tyrosine ammonia-lyase; C4H, cinnamate 4-hydroxylase; C3H/APX, 4-coumarate 3-hydroxylase/ascorbate peroxidase; 4CL, 4-coumarate:CoA ligase; C3′H, p-coumaroyl ester 3-hydroxylase; HCT, p-hydroxycinnamoyl-CoA:quinate/shikimate transferase; CSE, caffeoyl shikimate esterase; CCoAOMT, caffeoyl-CoA O-methyltransferase; CCR, cinnamoyl-CoA reductase; CAld5H, coniferaldehyde 5-hydroxylase; CAD, cinnamyl alcohol dehydrogenase; CAldOMT, 5-hydroxyconiferaldehyde O-methyltransferase; PMT, p-coumaroyl-CoA:monolignol transferase; LAC, laccase; PRX, peroxidase; FA, ferulate; H, p-hydroxyphenyl; G, guaiacyl; and S, syringyl.

In the proposed cinnamate/monolignol pathway, 4-COUMARATE:COENZYME A LIGASE (4CL) plays a pivotal role in modulating the pathway flux by converting hydroxycinnamic acids, such as p-coumaric acid (pCA or 4-coumaric acid), caffeic acid, and ferulic acid (FA), into the corresponding coenzyme A (CoA) thioesters (Figure 1). In most vascular plant species, 4CLs are present in multiple isoforms and are encoded by small gene families. The 4CL isoforms may have different substrate specificities or diverse spatiotemporal gene expression patterns, enabling them to direct the pathway flux toward different end-products by supplying appropriate mixtures of the activated thioesters (Ehlting et al., 1999; Harding et al., 2002; Gui et al., 2011; Sun et al., 2013; Li et al., 2015; Rao et al., 2015; Lavhale et al., 2018). Because of their proposed roles in mediating the biosynthesis of lignin, as well as other metabolites (e.g. flavonoids), there has been substantial research on the functions of various 4CL isoforms in different plant species (Lee and Douglas, 1996; Ehlting et al., 1999; Harding et al., 2002; Tsai et al., 2006; Penning et al., 2009; Gui et al., 2011; Xu et al., 2011; Sun et al., 2013; Chen et al., 2013; Li et al., 2015; Rao et al., 2015; Zhou et al., 2015; Lavhale et al., 2018; Xiong et al., 2019). Moreover, 4CL represents a major target in crop breeding and bioengineering studies conducted to produce varieties with a low-lignin biomass with enhanced cellulose isolation or hydrolysis efficiencies (Hu et al., 1999; Voelker et al., 2010; Xu et al., 2011; Van Acker et al., 2013; Jung et al., 2016; Park et al., 2017; de Vries et al., 2018; Tsai et al., 2020).

The genome of rice (Oryza sativa L.), a model grass species and an economically important grain crop, encodes five 4CL isoforms. Of these, Os4CL3 and Os4CL4 have been suggested to play more important roles in lignin biosynthesis than the other rice 4CLs (Gui et al., 2011; Sun et al., 2013; Liu et al., 2020). Recombinant Os4CL3 and Os4CL4 proteins both displayed high catalytic activities toward pCA, caffeic acid, and FA but no activity toward sinapic acid (Gui et al., 2011). In an earlier study, an antisense-based suppression of Os4CL3 expression in japonica rice (cv. Zhonghua 11) resulted in a decrease in lignin contents in the major aerial plant tissues, suggesting that Os4CL3 affects lignin biosynthesis; however, it is notable that Os4CL4 expression also decreased in the examined transgenic line (Gui et al., 2011). In another study, the silencing of Os4CL4 in indica rice (cv. Kasalath) seedlings also resulted in decreased lignin contents in the roots and leaves, implying Os4CL4 contributes to lignin biosynthesis along with Os4CL3 (Liu et al., 2020). The roles of 4CL isoforms in other grass species, including sorghum (Sorghum bicolor) (Saballos et al., 2008, 2012), switchgrass (Panicum virgatum) (Xu et al., 2011), and maize (Zea mays) (Xiong et al., 2019), have been elucidated by characterizing 4CL-deficient transgenic and mutant plants with obviously impaired lignin biosynthesis. Despite these extensive efforts, however, the contributions of multiple 4CL isoforms in the biosynthesis of diverse lignin monomers in grasses remain unclear, mainly because of the limited available information regarding the precise chemical structures of the lignins produced in 4CL-deficient grasses.

In this study, rice (japonica cv. Nipponbare) mutants harboring knockout mutations in Os4CL3 and Os4CL4 were generated via clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR associated protein 9 (CRISPR/Cas9)-mediated targeted mutagenesis, and the mutant cell walls were subjected to in-depth chemical structure analyses of cell walls using a series of wet-chemical and two-dimensional (2D) nuclear magnetic resonance (NMR) techniques. The results of these analyses revealed that loss-of-function mutations to Os4CL3 and Os4CL4 differentially affect the formation of lignin polymer units derived from diverse grass lignin monomers (i.e. the conserved non-γ-acylated monolignols and the grass-specific γ-p-coumaroylated monolignols and tricin), demonstrating the overlapping but divergent roles of the two 4CL isoforms during the coordinated production of various lignin monomers in grasses.

Results

Phylogenetic relationships and expression of rice 4CL genes

At the onset of this study, we revisited the phylogenetic relationships and expression of rice 4CL genes. An unrooted phylogenetic tree was constructed for the five 4CL isoforms in rice (Gui et al., 2011) and other known 4CL members from several grass species, such as S. bicolor (Saballos et al., 2012), Z. mays (Penning et al., 2009; Xiong et al., 2019), P. virgatum (Xu et al., 2011), Brachypodium distachyon (Sibout et al., 2017), and eudicot species, including Arabidopsis (Arabidopsis thaliana) (Ehlting et al., 1999; Li et al., 2015), Populus sp. (Harding et al., 2002; Tsai et al., 2006; Rao et al., 2015; Zhou et al., 2015), soybean (Glycine max) (Lindermayr et al., 2002), and tobacco (Nicotiana tabacum) (Lee and Douglas, 1996; Figure 2A and Supplemental Table S1). Consistent with earlier findings (Gui et al., 2011; Sun et al., 2013; Lavhale et al., 2018), Os4CL1, Os4CL3, Os4CL4, and Os4CL5 were clustered together in the grass Class III 4CL clade. Of these rice Class III 4CLs, Os4CL3 and Os4CL4 have highly similar protein sequences (82.6%) and were clustered together in a sub-clade containing the switchgrass Pv4CL1 (Xu et al., 2011), sorghum Bmr2 (Saballos et al., 2008, 2012), and maize Zm4CL1/bm5 (Penning et al., 2009; Xiong et al., 2019) implicated in lignin biosynthesis. In contrast, Os4CL2 was assigned to the grass Class IV 4CL clade that is more closely associated with the eudicot Class II 4CLs, which have been considered to participate in the biosynthesis of nonlignin phenylpropanoids (Gui et al., 2011; Sun et al., 2013; Lavhale et al., 2018).

Figure 2.

Phylogeny, gene expression, and genome editing of rice 4CLs. A, Phylogenetic tree of 4CL proteins. The unrooted phylogenetic tree of 4CL proteins from rice and other major grass and dicot species were built by neighbor-joining method with bootstrap values after 1,000 tests. Scale bar denotes 0.05 substitutions per site. Species abbreviations: Os, O. sativa; Sb, S. bicolor; Zm, Z. mays; Pv, P. virgatum; Bradi, B. distachyon; At, A. thaliana; Pto, Populus tomentosa, Pt, Populus tremuloides; Ptr, P. trichocarpa. Gm, G. max; and Nt, N. tabacum. Accession numbers of 4CL proteins used are listed in Supplemental Table S1. B, Expression of rice 4CL genes in wild-type rice culms at the heading stage as determined by RT-qPCR. A ubiquitin gene (OsUBQ5, LOC_Os01g22490) was used as an internal control. Values are mean ± standard deviation from three biological replicates (n = 3). C, Gene structures and mutation patterns in Os4CL3- and Os4CL4-deficient rice mutants generated via CRISPR/Cas9. The sgRNA target, protospacer adjacent motif, PAM, and inserted or deleted sequences are highlighted. WT, wild-type control line; os4cl3-a and os4cl3-b, Os4CL3-knockout lines; os4cl4-a and os4cl4-b, Os4CL4-knockout lines.

The involvement of the rice 4CL members in lignin biosynthesis was investigated further on the basis of an in silico expression analysis of the rice 4CL genes and other known/putative lignin biosynthetic genes in wild-type rice using the RiceXPro public DNA microarray database (Sato et al., 2013). Among the five rice 4CL genes, Os4CL3 and Os4CL4 were highly expressed along with other known/putative lignin biosynthetic genes in the developing culm, in which lignification typically occurs, whereas other rice 4CL genes (Os4CL1, Os4CL2, and Os4CL5) were expressed at much lower levels in the culm (Supplemental Figure S1). Consistently, the gene expression analysis via reverse transcription quantitative PCR (RT-qPCR) confirmed that Os4CL3 and Os4CL4 were highly expressed in wild-type rice culm at the heading stage, whereas the expression levels of the other three 4CL genes (Os4CL1, Os4CL2, and Os4CL5) were relatively low (Figure 2B). Collectively, these data support the earlier suggestion that both Os4CL3 and Os4CL4 are major 4CLs for lignin biosynthesis in rice (Gui et al., 2011; Liu et al., 2020).

Generation of Os4CL3- and Os4CL4-deficient rice mutants via CRISPR/Cas9

We generated rice mutant lines deficient in Os4CL3 and Os4CL4 by CRISPR/Cas9-mediated targeted mutagenesis. The CRISPR/Cas9 binary vectors (Mikami et al., 2015) harboring single-guide RNAs (sgRNAs) specific for the first exons of Os4CL3 and Os4CL4 were designed using the CRISPR-P program (Liu et al., 2017) and then inserted into embryogenic calli derived from Nipponbare rice. The genotyping of the regenerated plants by direct sequencing identified T₀ individuals with various insertion-deletion mutations (indels) in the targeted Os4CL3 and Os4CL4 sites. Consequently, we isolated T₁ individuals for two Os4CL3-knockout lines harboring a 1-bp insertion (os4cl3-a) and a 1-bp deletion (os4cl3-b) in the first exon of Os4CL3 as well as two Os4CL4-knockout lines (os4cl4-a and os4cl4-b) harboring different 1-bp insertions in the first exon of Os4CL4 (Figure 2C). The genotype and zygosity of individual plants were confirmed by direct sequencing and also by sequencing at least 12 randomly selected, subcloned PCR amplicons from the targeted genomic sites (Takeda et al., 2018, 2019). The indels in the isolated mutant lines caused frameshift mutations that result in the production of truncated nonfunctional 4CLs. We also confirmed there were no mutations in at least the three top-ranked potential off-target sites in any of the selected T₁ individuals (Supplemental Table S2).

Phenotype, vascular anatomy, and 4CL activity of Os4CL3- and Os4CL4-deficient rice mutants

The fully genotyped os4cl3 and os4cl4 mutant plants were grown along with the wild-type control plants in a greenhouse (Supplemental Figure S2A). Previous reports noted the impaired growth characteristics of Os4CL3- and Os4CL4-deficient rice transgenic lines (Gui et al., 2011; Liu et al., 2020). Consistent with the earlier observations, our genome-edited rice mutants of Os4CL3 and Os4CL4 displayed decreased biomass production and fertility compared with the wild-type control (Table 1). On the basis of a comparison with the os4cl3 mutants, the os4cl4 mutants showed more prominently impaired plant growth with decreases in plant height, culm length, panicle length, and tiller number compared with the wild-type control (Table 1). Overall, our data reflected the importance of the two rice 4CLs, especially Os4CL4, for plant growth and development.

Table 1.

Growth characteristics of Os4CL3- and Os4CL4-deficient rice plants

Trait	WT	os4cl3-a	os4cl3-b	os4cl4-a	os4cl4-b
Plant height (cm)1	119.1 ± 10.2a	119.1 ± 6.8a	113.4 ± 6.0a	87.9 ± 2.3b	80.5 ± 2.9b
Culm length (cm)2	78.7 ± 5.4a	80.3 ± 3.4a	74.4 ± 4.1a	51.25 ± 2.5b	46.5 ± 4.5b
Panicle length (cm)	23.3 ± 1.5a	20.2 ± 1.2a	20.3 ± 1.8a	12.9 ± 0.7b	11.2 ± 2.6b
Tiller number	12.0 ± 1.7a	9.33 ± 2.1a	7.7 ± 0.6b	9.3 ± 2.1a	6.0 ± 1.0b
Biomass (g)3	70.0 ± 11.3a	38.2 ± 7.4b	42.9 ± 5.4b	18.6 ± 4.6c	10.3 ± 5.0c
Fertility (%)	83.5 ± 5.7a	61.3 ± 3.1b	78.5 ± 4.8ab	29.7 ± 5.6c	18.1 ± 6.6c

Values are mean ± standard deviation from biological replicates (n = 3). Different letters on values indicate significant differences and significant differences over wild-type control are highlighted in bold (one-way ANOVA with Tukey’s test, P < 0.05). WT, wild-type control line; os4cl3-a and os4cl3-b, Os4CL3-knockout lines; and os4cl4-a and os4cl4-b, Os4CL4-knockout lines.

¹Length measured from the soil surface to the tip of the top leaf.

²Length measured from soil surface to the panicle base.

³Dry weight of all aerial parts excluding panicles.

The vascular anatomy and lignin deposition of the os4cl3 and os4cl4 mutants were examined by analyzing their culm cross-sections treated with the Wiesner (phloroglucinol-HCl) lignin staining solution (Supplemental Figure S2B). Although there was no obvious abnormality in the overall vasculature (e.g. collapsed xylem cells and thinner secondary cell walls), the reddish coloration (i.e. lignin signal) following staining with the Wiesner reagent was much weaker for the cell walls of both os4cl3 and os4cl4 lines than for the wild-type control cell walls. We noted that the reddish staining was lower for the os4cl3 plant cell walls than for the os4cl4 plant cell walls. These results suggested that lignin deposition was hindered in the os4cl3 and os4cl4 mutants, but especially in the os4cl3 mutant, as further examined below.

We also tested CoA ligation activities toward different cinnamic acids in crude protein extracts prepared from the developing stems of the os4cl3 and os4cl4 mutant seedlings (5 weeks old). We confirmed that, as in the culms at the heading stage (Figure 2B), Os4CL3 and Os4CL4 were predominantly expressed over the other 4CL genes in the seedling stem samples collected from the wild-type control line (Supplemental Figure S3). As listed in Table 2, crude protein extracts prepared from the wild-type control line displayed overall comparable CoA ligation activities toward pCA, caffeic acid, and FA; essentially, no activity was detected in reactions using cinnamic acid and sinapic acid as substrates. Compared with the wild-type control, the specific activities toward pCA, caffeic acid and FA were all markedly reduced in os4cl3 and os4cl4 mutant lines (Table 2), which could be attributed to the loss-of-function of the targeted 4CL isoform in each mutant line. Apparently, the reductions in the activities detected in the os4cl3 mutants were more prominent than the reductions detected in the os4cl4 mutants (Table 2), which is in line with the difference in the transcript abundances of the two 4CL genes (Os4CL3 > Os4CL4) in the wild-type rice (Supplemental Figure S3). The relative activities toward pCA, caffeic acid and FA within each genotype were slightly different but overall similar between the wild-type control and os4cl3 and os4cl4 mutant lines, that is, as did the wild-type control, the 4CL mutant lines displayed overall comparable substrate specificities toward the three cinnamic acid substrates (Table 2).

Table 2.

CoA ligase activity of crude protein extracts from Os4CL3- and Os4CL4-defecient rice plants

Substrate	CoA ester formation rate (nmol/min/µg protein)
	WT	os4cl3-a	os4cl3-a	os4cl4-a	os4cl4-b
pCA	11.7 ± 0.3a (1.0)	4.8 ± 0.1d (1.0)	4.5 ± 0.3d (1.0)	7.6 ± 0.5c (1.0)	8.6 ± 0.3b (1.0)
Cinnamic acid	N.D. (0.0)	N.D. (0.0)	N.D. (0.0)	N.D. (0.0)	N.D. (0.0)
Caffeic acid	10.5 ± 0.3a (0.9)	5.8 ± 0.5c (1.2)	5.9 ± 0.8c (1.3)	7.6 ± 0.8b (1.0)	8.7 ± 0.5b (1.0)
FA	8.9 ± 0.3a (0.8)	4.1 ± 0.5c (0.9)	5.0 ± 0.4c (1.1)	6.8 ± 0.6b (0.9)	7.3 ± 0.6b (0.8)
Sinapic acid	N.D. (0.0)	N.D. (0.0)	N.D. (0.0)	N.D. (0.0)	N.D. (0.0)

Crude protein extracts were prepared from stems of 5-week-old seedlings. Values are mean ± standard deviation from biological replicates (n = 3). Different letters on the values indicate significant differences in the CoA ligation activities toward each substrate and significant differences over wild-type control are highlighted in bold (one-way ANOVA with Tukey’s test, P < 0.05). Numbers in parentheses are relative activities compared with pCA within each genotype. WT, wild-type control line; os4cl3-a and os4cl3-b, Os4CL3-knockout lines; os4cl4-a and os4cl4-b, Os4CL4-knockout line; N.D., not detectable.

Cell wall structures of Os4CL3- and Os4CL4-deficient rice mutants

To further investigate the effects of the loss-of-function mutations to Os4CL3 and Os4CL4 on lignin biosynthesis and cell wall chemical structures, cell wall residue (CWR) samples prepared via a serial solvent extraction of the culm tissues of the os4cl3 and os4cl4 mutants (fully genotyped T₁ homozygous individuals) and the wild-type control were subjected to in-depth cell wall structural analyses using chemical and 2D NMR methods. We could not collect sufficient CWR samples from os4cl4-b plants because of their low biomass (Table 1). Therefore, we report the results obtained for os4cl3-a, os4cl3-b, os4cl4-a, and wild-type control cell wall samples.

Lignin content

The lignin content of the os4cl3 and os4cl4 mutant cell walls were first evaluated using the Klason lignin assay. Compared with the wild-type control, the Klason lignin level (corrected according to ash and protein contents) was significantly lower in both os4cl3-a (by ∼23%) and os4cl3-b (by ∼14%), whereas the Klason lignin level of the os4cl4-a cell walls was not significantly different from that of the wild-type control (Figure 3A). Next, we also estimated the lignin content changes in terms of the changes in the total yields of the lignin-derived monomeric compounds released by thioacidolysis and derivatization followed by reductive cleavage (DFRC) (Figure 3, B and C). Both of these methods involve the cleavage of the major β–O–4 linkages in lignin polymers and the release of quantifiable amounts of monomeric compounds (Figure 3D). Consequently, relative to the wild-type yields, the total yields of the lignin-derived monomeric compounds released by thioacidolysis (Figure 3B) and DFRC (Figure 3C) decreased significantly in the two os4cl3 lines (by ∼61% [thioacidolysis] and 38% [DFRC] in os4cl3-a and by ∼55% [thioacidolysis] and 37% [DFRC] in os4cl3-b) as well as in os4cl4-a (by ∼30% [thioacidolysis] and 14% [DFRC]). These data indicated that the monolignol-derived phenylpropane units in the lignin polymers decreased in both os4cl3 and os4cl4 mutant cell walls, suggestive of the roles for Os4CL3 and Os4CL4 in monolignol biosynthesis. The less prominent reductions in the Klason lignin content than those estimated by the total monomeric compounds yields in thioacidolysis and DFRC might be due to proportional increases in the lignin polymer units other than the monolignol-derived phenylpropane units, for example, pCA and tricin units, as further examined below.

Figure 3.

Lignin content and compositional analysis of Os4CL3- and Os4CL4-deficient rice plants. A, Lignin content determined by the Klason lignin assay. B and C, Yield and ratio of lignin degradation monomers released via thioacidolysis (B) and derivatization followed by reductive cleavage, DFRC (C). D, Lignin degradation monomers released via thioacidolysis and DFRC. Values in (A), (B), and (C) are means ± standard deviation of biological replicates (n = 3). Different letters indicate significant differences (one-way ANOVA with Tukey’s test, P < 0.05). H, p-hydroxyphenyl; G, guaiacyl; S, syringyl; WT, wild-type control line; Os4cl3-a and os4cl3-b, Os4CL3-knockout lines; and Os4cl4-a, Os4CL4-knockout line.

Lignin composition

Thioacidolysis and DFRC were also used to determine the lignin compositional changes in the os4cl3 and os4cl4 cell walls. As shown in Figure 3D, thioacidolysis releases thioethylated H- (H_thio), G- (G_thio), and S-type (S_thio) monomeric compounds from the H-, G-, and S-type β–O–4 units, respectively, regardless of their γ-acylation status (Lapierre et al., 1986). On the other hand, DFRC releases quantifiable amounts of nonacylated (γ-free) monomeric compounds (H_DFRC, G_DFRC, and S_DFRC) and γ-acylated monomeric compounds (Gʹ_DFRC and Sʹ_DFRC) from the corresponding non-γ-acylated and γ-p-coumaroylated β–O–4 units, respectively, providing useful information about the γ-acylation status of lignin polymers in grasses (Lu and Ralph, 1997, 1999; Karlen et al., 2018). The thioacidolysis and DFRC results indicated that compared with the corresponding yields from the wild-type control cell walls, the yields of all the G-type (G_thio, G_DFRC, and Gʹ_DFRC) and S-type (S_thio, S_DFRC and Sʹ_DFRC) monomeric compounds from the os4cl3 cell walls decreased significantly, whereas there were no or small changes to the yields of the H-type monomeric compounds (H_thio and H_DFRC) (Figure 3, B and C). The depletion was greater for the G-type monomeric compounds than for the corresponding S-type monomeric compounds, resulting in higher S_thio/G_thio and S_DFRC/G_DFRC ratios for the os4cl3 cell walls than for the wild-type cell walls (Figure 3, B and C). Overall, these results indicated that Os4CL3 contributes broadly to the formation of the major G- and S-type monolignol-derived lignin units, including those derived from the non-γ-acylated and γ-p-coumaroylated monolignols. On the other hand, thioacidolysis of the os4cl4 cell walls recorded a significant decrease in the yield of the G-type monomeric compound (G_thio) but not in the yields of the S-type (S_thio) and H-type (H_thio) monomeric compounds (Figure 3B). In agreement with this result, DFRC detected no significant changes in the yields of the S-type (S_DFRC and Sʹ_DFRC) and H-type (H_DFRC) monomeric compounds released from the os4cl4 cell walls (Figure 3C). DFRC detected a significant decrease in the yield of the non-γ-acylated G-type monomeric compound (G_DFRC) but not in the yield of the γ-acylated G-type monomeric compound (Gʹ_DFRC) from the os4cl4 cell walls (Figure 3C). These data collectively indicated that Os4CL4 contributes to the formation of the G-type monolignol-derived lignin units, particularly those derived from the non-γ-acylated G-type monolignol, that is, coniferyl alcohol.

Cell-wall-bound hydroxycinnamates and polysaccharides

Next, we analyzed the cell-wall-bound hydroxycinnamates (i.e. FA and pCA) in the rice cell walls by quantifying the free FA and pCA yields following a mild alkaline treatment of the CWR samples. The cell-wall-bound FA content was lower in the os4cl3 and os4cl4 cell walls than in the wild-type cell walls (Table 3), suggesting that both Os4CL3 and Os4CL4 contribute to the biosynthesis of the FA precursor for cell wall feruloylation, which mostly occurs in the hemicellulose (arabinoxylan) fraction (Ralph, 2010). The cell-wall-bound pCA content decreased significantly in the os4cl3 cell walls but conversely increased in the os4cl4 cell walls (Table 3). We also investigated the abundance of cell wall polysaccharides by performing a neutral sugar analysis according to the two-step acid-catalyzed hydrolysis method. We detected statistically significant increases in the amounts of cellulosic glucose (glucose released from trifluoracetic acid-insoluble fractions) released from the os4cl3-b mutant cell walls and hemicellulosic xylose released from the os4cl3-a mutant cell walls, relative to the corresponding wild-type control levels (Table 3).

Table 3.

Cell wall sugar and hydroxycinnamate analyses of culm cell walls from Os4CL3- and Os4CL4-deficient rice plants

Component	WT	os4cl3-a	os4cl3-b	os4cl4-a
Neutral sugars (mg/g CWR)
Cellulosic glucose1	365.5 ± 27.2b	389.0 ± 14.2b	421.1 ± 22.9a	381.6 ± 7.2b
Hemicellulosic glucose2	34.2 ± 2.4b	45.2 ± 2.2a	31.3 ± 1.7b	33.9 ± 3.3b
Xylose2	164.2 ± 0.9a,b	176.5 ± 5.7a	170.4 ± 4.8a,b	157.8 ± 9.1b
Arabinose2	28.0 ± 1.8a	29.8 ± 0.3a	29.0 ± 1.7a	27.5 ± 1.5a
Galactose2	13.7 ± 1.5a	14.8 ± 1.2a	13.8 ± 1.1a	15.7 ± 0.4a
Cell-wall-bound hydroxycinnamates (mg/g CWR)
p-Coumarate (pCA)	19.8 ± 1.2b	14.2 ± 1.5c	16.2 ± 1.5c	24.7 ± 0.6a
Ferulate (FA)	3.6 ± 0.28a	2.4 ± 0.3b	2.8 ± 0.1b	2.8 ± 0.1b

Values are mean ± standard deviation from biological replicates (n = 3). Different letters on values indicate significant differences and significant differences over wild-type control are highlighted in bold (one-way ANOVA with Tukey’s test, P < 0.05). WT, wild-type control line; os4cl3-a and os4cl3-b, Os4CL3-knockout lines; and os4cl4-a, Os4CL4-knockout line.

¹Glucose released from the trifluoroacetic acid-insoluble cell wall fractions.

²Monosaccharides released from the trifluoroacetic acid-soluble cell wall fractions.

2D NMR analysis of whole cell walls

To further clarify the changes to the cell wall chemical structures in the os4cl3 and os4cl4 mutants, we conducted a solution-state 2D ¹H–¹³C heteronuclear single-quantum coherence (HSQC) NMR analysis on the culm cell wall and lignin samples prepared from the os4cl3 and os4cl4 mutants and the wild-type control. First, we obtained 2D HSQC NMR spectra for the whole cell walls by directly swelling ball-milled rice CWR samples in the dimethyl sulfoxide (DMSO)-d₆/pyridine-d₅ NMR solvent system (Kim and Ralph, 2010; Mansfield et al., 2012). The HSQC NMR spectra for the rice cell walls included signals from lignin, p-hydroxycinnamate, and polysaccharide components typical of grass cell walls (Figure 4A). The peak assignments, which were determined on the basis of published chemical shift data (Kim and Ralph, 2010; Mansfield et al., 2012; Kim et al., 2017; Lam et al., 2017; Tarmadi et al., 2018; Tobimatsu et al., 2019), are listed in Supplemental Table S3. For a semi-quantitative examination of the relative proportions of lignin and polysaccharide units in the rice cell walls, we performed volume integration analysis of the well-resolved, major lignin, p-hydroxycinnamate and polysaccharide signals. The reported signal intensity data are relative intensities normalized according to the sum of the integrated signals, reflecting the proportional amount of each component in the rice cell walls (see “Materials and methods”).

Figure 4.

2D HSQC NMR analysis of whole culm cell walls from Os4CL3- and Os4CL4-deficient rice plants. A, ¹H–¹³C short-range correlation (HSQC) NMR spectra of ball-milled whole rice culm cell walls. Signal assignments are listed in Supplemental Table S3. B, Normalized signal intensities and ratios of major lignin, hydroxycinnamate, and polysaccharide units expressed as percentages of the sum of the listed signals (½S_2/6 + G₂ + ½P_2/6 + F₂ + ½T_2′/6′ + A₁ + U₁ + Gl₁ + X₁ + X′₁ + X′′₁ = 100). NMR analysis was conducted for culm CWR samples pooled from three biologically independent plants for each line. WT, wild-type control line. Os4cl3-a and os4cl3-b, Os4CL3-knockout lines. Os4cl4-a, Os4CL4-knockout line.

In the cell wall spectra of the os4cl3 and os4cl4 mutants, the monolignol-derived G- and S-type lignin aromatic signals (G and S) were relatively lower, whereas the polysaccharide signals, such as those from glucose (Gl), nonacetylated (X) and acetylated (Xʹ and X″) xylose, and arabinose (A) units, were relatively higher, compared with those observed in the wild-type control spectrum (Figure 4B); the H-type aromatic signals (H) were not included in the volume integration analysis because of the considerable overlap with protein-derived aromatic signals (Kim et al., 2017). These findings suggested that the abundances of the monolignol-derived lignin units decreased relatively over the cell wall polysaccharide contents (cellulose and arabinoxylan) in the os4cl3 and os4cl4 mutant cell walls when compared with the wild-type control cell wall (Figure 4B). The cell-wall-bound FA signals (F) decreased in the os4cl3 and os4cl4 mutant cell wall spectra, whereas the cell-wall-bound pCA signals (P) decreased only in the os4cl3 mutant cell wall spectrum (Figure 4B), which was consistent with the alkaline-releasable pCA and FA content data (Table 3), as well as the DFRC-based lignin γ-acylation data (Figure 3C). Furthermore, our HSQC NMR analysis detected the contrasting changes in the lignin-bound tricin units in the os4cl3 and os4cl4 mutant cell walls. Specifically, compared with the wild-type control spectrum, the tricin aromatic signals (T) increased and decreased in the os4cl3 and os4cl4 spectra, respectively (Figure 4B). This finding indicated that the abundance of the lignin-bound tricin units increased and decreased in the cell walls of the os4cl3 and os4cl4 mutants, respectively, demonstrating the distinct contributions of Os4CL3 and Os4CL4 to tricin-lignin biosynthesis.

2D NMR analysis of enzyme lignin fractions

For a more precise examination of the lignin structure, we performed a 2D HSQC NMR analysis of the lignin-enriched cell wall (enzyme lignin) fractions prepared from the rice culm CWR samples by the enzymatic removal of a large part of the cell wall polysaccharides (Mansfield et al., 2012; Tobimatsu et al., 2019).

The aromatic sub-regions (δ_C/δ_H, 135–90/8.0–6.0) of the HSQC NMR spectra of the rice enzyme lignin fractions included well-resolved, typical lignin signals from the monolignol-derived H-, G-, and S-type aromatic rings (H, G, and S) along with those from the pCA (P) and tricin (T) units incorporated into the grass lignin polymer (Figure 5A and Supplemental Table S3). The volume integration analysis of the major aromatic signals (normalized on a ½S_2/6 + G₂ + ½H_2/6 + ½T_2ʹ/6ʹ + ½P_2/6 = 100 basis) indicated that, compared with the wild-type control lignin spectrum, the S/G aromatic signal ratios (½S_2/6/G₂) increased considerably in the os4cl3 and os4cl4 lignin spectra (Figure 5C). The increase in the S/G aromatic signal ratio was more prominent for os4cl3 than for os4cl4, which was consistent with the thioacidolysis- and DFRC-derived S/G monomeric compound ratios (Figure 3). Further, similar to the alkaline-releasable pCA content (Table 3) and the DFRC-based lignin γ-acylation data (Figure 3C), the pCA signals (P) decreased in the os4cl3 lignin spectrum, but not in the os4cl4 lignin spectrum (Figure 5C). Additionally, in accordance with the results of the cell wall NMR analysis (Figure 4), the tricin signals (T) increased in the os4cl3 lignin spectrum but decreased in the os4cl4 lignin spectrum (Figure 5C).

Figure 5.

2D HSQC NMR analysis of culm enzyme lignin fractions prepared from Os4CL3- and Os4CL4-deficient rice plants. A and B, Aromatic (A) and aliphatic (B) sub-regions of ¹H–¹³C short-range correlation (HSQC) NMR spectra of culm enzyme lignin fractions. Contour coloration matches the substructures shown in each panel. Signal assignments are listed in Supplemental Table S3. C and D, Normalized signal intensities and ratios of major lignin aromatic units expressed as percentages on a ½S_2/6 + G₂ + ½H_2/6 + ½T_2ʹ/6ʹ + ½P_2/6 = 100 basis (C), and inter-monomeric linkages and end-units expressed as percentages on a I_α + II_α + ½III_α = 100 basis (D). NMR analysis was conducted for enzyme lignin samples prepared from culm CWR samples pooled from three biologically independent plants for each line. WT, wild-type control line; Os4cl3-a and os4cl3-b, Os4CL3-knockout lines; and Os4cl4-a, Os4CL4-knockout line.

The aliphatic sub-regions (δ_C/δ_H, 90–50/6.0–2.5) of the HSQC NMR spectra resolved the signals from the major inter-monomeric linkages in the lignin polymers, including β–O–4 (I), β–5 (II), and β–β (III, predominantly the tetrahydrofuran type) (Lu and Ralph, 2002), along with the signals from the non-γ-acylated (γ-free) (X1) and γ-acylated (X1ʹ) cinnamyl alcohol end-units (Figure 5B and Supplemental Table S3). The volume integration analysis of the lignin inter-monomeric linkage signals (I_α, II_α, and ½III_α normalized on a I_α + II_α + ½III_α = 100 basis) revealed the β–5 (II) signal was lower and the β–β (III) signal was higher in the os4cl3 and os4cl4 lignin spectra than in the wild-type control lignin spectrum (Figure 5D). This result was suggestive of a relative decrease and increase in the β–5 and β–β linkage types, respectively, in the os4cl3 and os4cl4 mutant lignins. A decrease in β–5 linkages and an increase in β–β linkages are typical consequences of the relative decrease and increase in the abundance of G and S units, respectively, in lignin polymers (Stewart et al., 2009; Anderson et al., 2015; Takeda et al., 2017). Therefore, this result is in accordance with the significant increase in the S/G unit ratios for the os4cl3 and os4cl4 mutant lignins, as determined earlier (Figure 3). Further, the signals from the γ-acylated cinnamyl alcohol end-units (½X_γnormalized on a I_α + II_α + ½III_α = 100 basis) were decreased in the os4cl3 lignin spectra but increased in the os4cl4 lignin spectra (Figure 5D), further supporting our notion that the grass-specific γ-p-coumaroylated lignin units are relatively reduced in the os4cl3 lignins but not in the os4cl4 lignins.

Overall, as summarized in Figure 6, our NMR and chemical analysis data indicate the contrasting changes in lignin composition in the os4cl3 and os4cl4 mutant cell walls compared with the wild-type cell walls (Figure 6A), reflecting the distinct roles of Os4CL3 and Os4CL4 in the cinnamate/monolignol pathway that produce the diverse grass lignin monomers (Figure 6B). This idea is discussed in more detail below.

Figure 6.

Summary of lignin compositional changes in Os4CL3- and Os4CL4-deficient rice culm cell walls and proposed roles of Os4CL3 and Os4CL4 in lignin biosynthesis. A, rough estimations of the abundances of major lignin polymer units in rice culm cell walls. Relative abundances were calculated based on yields of DFRC-derived lignin degradation monomers (H_DFRC, G_DFRC, S_DFRC, Gʹ_DFRC, and Sʹ_DFRC) (Figure 3C) and ratio of HSQC NMR signal intensities of lignin-bound tricin units (½T_2ʹ/6ʹ) over H/G/S aromatic units (½S_2/6 + G₂ + ½H_2/6) (Figure 5C), and normalized based on the sum of the lignin units in the wild-type control. Values indicated on mutant data are changes of total and each lignin unit compared with the wild-type levels. Tricin units could be overestimated because of the higher response factors of the tricin end-unit signals compared with those of the internal H/G/S unit signals in HSQC NMR (Lan et al., 2016). WT, wild-type control line. Os4cl3-a and os4cl3-b, Os4CL3-knockout lines. Os4cl4-a and os4cl4-b, Os4CL4-knockout lines. B, Proposed roles of Os4CL3 and Os4CL4 in the biosynthesis of monolignol and tricin lignin monomers during rice culm lignification.

Enzymatic saccharification efficiency in the Os4CL3- and Os4CL4-deficient rice mutants

We assessed the cell wall digestibility of the os4cl3 and os4cl4 mutants by investigating the saccharification efficiency of their culm CWR samples after a treatment with a cellulolytic enzyme cocktail (without any chemical pretreatment), as described by Hattori et al. (2012). The results revealed that the enzymatic saccharification efficiency was significantly higher for the Os4CL3-knockout os4cl3 rice culm cell walls than for the wild-type control rice culm cell walls (Supplemental Figure S4), which could be at least partly associated with the reduced lignin polymer units (Figure 3). On the other hand, the Os4CL4-knockout os4cl4 cell walls displayed a decrease in the enzymatic saccharification efficiency compared with the wild-type controls (Supplemental Figure S4).

Discussion

Both Os4CL3 and Os4CL4 contribute to cell wall lignification in rice

This study demonstrated the similarity and diversity in the roles of Os4CL3 and Os4CL4, which are two closely related Class III 4CLs (Figure 2A), in rice cell wall lignification. Our results are in line with the earlier suggestion that the Class III 4CLs are responsible for lignin biosynthesis in grasses, with functions similar to those of the Class I 4CLs in eudicots (Gui et al., 2011; Sun et al., 2013; Lavhale et al., 2018; Liu et al., 2020). Considering the predominant expression of Os4CL3 and Os4CL4 in the lignin-developing tissues of rice (Figure 2B and Supplemental Figures S1 and S3) and the prominent effects of the loss-of-function mutations to the two 4CL genes on culm lignin contents and structures (Figures 3–5) as well as rice growth performance (Table 1), it is plausible that Os4CL3 and Os4CL4 are the major sources of the 4CL activity contributing to cell wall lignification in rice. This is supported by a recent study that revealed that the double-knockout of Os4CL3 and Os4CL4 is seedling lethal possibly because of the resulting substantial loss of the 4CL activity required for lignification (Liu et al., 2020). As previously noted (Gui et al., 2011; Sun et al., 2013; Lavhale et al., 2018), the two other rice Class III 4CLs, that is, Os4CL1 and Os4CL5 (Figure 2A), may mediate lignification at specific locations or times during plant development or they may be activated in response to biotic or abiotic stresses. On the other hand, the single rice Class IV 4CL, that is, Os4CL2 (Figure 2A), may participate in the biosynthesis of phenylpropanoids other than lignin, as has been proposed for the closely related Class II 4CLs in eudicots (Hu et al., 1998; Ehlting et al., 1999; Chen et al., 2014b; Li et al., 2015).

Divergent roles of Os4CL3 and Os4CL4 in monolignol and tricin biosynthesis

Our chemical and 2D NMR analyses of the os4cl3 and os4cl4 mutant cell walls revealed that loss-of-functions of Os4CL3 and Os4CL4 differently alter the composition of the monolignol-derived lignin polymer units, demonstrating the differential contributions of the two 4CLs in monolignol biosynthesis. The os4cl3 cell walls were reduced in the major non-γ-p-coumaroylated and γ-p-coumaroylated G and S lignin units, with more prominent reductions in G units than in S units (Figure 6A). In contrast, the os4cl4 cell walls were primarily decreased in the abundance of the non-γ-p-coumaroylated G units, with little change in other monolignol-derived lignin polymer units (Figure 6A). Collectively, these data support our notion that Os4CL3 contributes more broadly to the biosynthesis of the major monolignol-type lignin monomers in grasses, that is, the non-γ-p-coumaroylated G- and S-type monolignols (coniferyl and sinapyl alcohol) and the grass-specific γ-p-coumaroylated G- and S-type monolignols (coniferyl and sinapyl p-coumarate) (Figure 6B). In contrast, Os4CL4 contributes primarily to the biosynthesis of the non-γ-p-coumaroylated monolignol, particularly the G-type coniferyl alcohol, and its contribution to the biosynthesis of the grass-specific γ-p-coumaroylated monolignols is much less substantial than the contribution of Os4CL3 (Figure 6B).

Another intriguing difference between the os4cl3 and os4cl4 mutant cell walls was the contrasting changes in the grass-specific tricin units in lignin. Our analysis data determined that, along with the decreases in the abundance of monolignol-derived lignin units, there was a considerable decrease in the number of lignin-bound tricin units in the os4cl4 mutant lignin (Figure 6A), suggesting that Os4CL4 is important not only for monolignol biosynthesis but also for flavonoid tricin biosynthesis (Figure 6B). It is most likely that Os4CL4 contributes to tricin biosynthesis by converting pCA to p-coumaroyl-CoA for subsequent use in the downstream flavonoid biosynthetic pathway (Lam et al., 2021, 2022; Figure 1). However, in the os4cl3 mutant lignin, the abundance of the lignin-bound tricin units showed no reduction and relatively increased over the decreased G and S lignin units (Figure 6A), indicating that the contribution of Os4CL3 to tricin biosynthesis is minimal or at least considerably less than that of Os4CL4 (Figure 6B).

The contrasting contributions of Os4CL3 and Os4CL4 in the biosynthesis of the grass-specific lignin monomers, that is, γ-p-coumaroylated monolignols and tricin, emphasize the characteristic diversity of 4CL isoforms within grasses. Notably, our recent lignin biosynthesis studies also revealed the differential metabolic regulation of the biosynthesis of grass lignin monomers. For example, rice p-COUMAROYL ESTER 3-HYDROXYLASE (OsC3′H) and rice CONIFERALDEHYDE 5-HYDROXYLASE 1 (OsCAld5H1), which are the two key cytochrome P450 enzymes that catalyze aromatic hydroxylations during the construction of the G and S aromatic rings (Figure 1), are likely involved in the biosynthesis of the non-γ-p-coumaroylated monolignols, but not in the biosynthesis of the grass-specific γ-p-coumaroylated monolignols during lignification in rice (Takeda et al., 2018, 2019). In addition, OsMYB108, which is a rice R2R3-MYB transcription factor that negatively regulates lignin biosynthesis, is more involved in the biosynthesis of the grass-specific γ-p-coumaroylated monolignols and tricin than in the biosynthesis of the non-γ-p-coumaroylated monolignols (Miyamoto et al., 2019). The diverse roles of the two rice 4CL isoforms revealed in this study further corroborate our speculation that the cinnamate/monolignol pathway enzymes and transcription factors have divergently evolved to produce the grass-specific γ-p-coumaroylated monolignols and tricin apart from the non-γ-p-coumaroylated monolignols upon lignification.

At this time, it is difficult to specify the mechanism(s) that defines the differential contributions of Os4CL3 and Os4CL4 in monolignol and tricin biosynthesis. The 4CL enzyme assay on the crude protein extracts from the 4CL mutant stems indicated that loss of function of either Os4CL3 or Os4CL4 does not have a pronounced impact on the relative substrate specificities of the remaining 4CL proteins toward different cinnamic acid substrates (Table 2). In line with this result, recombinant Os4CL3 and Os4CL4 proteins reportedly have similarly high catalytic activities toward pCA, caffeic acid, and FA (Gui et al., 2011). Therefore, it is unlikely that the individual catalytic activities of Os4CL3 and Os4CL4 could be the major factor contributing to their preferences in the production of different lignin monomers. Nevertheless, earlier studies revealed that some 4CL isoforms in poplar (Populus trichocarpa) form enzyme complexes that modulate the relative efficiencies of CoA ligations of hydroxycinnamic acids and thereby affect the metabolic flux in monolignol biosynthesis (Chen et al., 2013, 2014a; Lin et al., 2015). Also, the importance of the formation of multienzyme complexes between different pathway enzymes (i.e. metabolons) has been emphasized in the biosynthesis of plant metabolites including monolignols and flavonoids (Jørgensen et al., 2005; Bassard et al., 2012; Nakayama et al., 2019; Lin et al., 2021). It would be therefore intriguing to test whether such protein–protein interactions involving 4CL exist and affect lignin formations in grasses. On the other hand, the microarray-based gene expression data suggested that Os4CL4 is more widely expressed during rice plant development than Os4CL3 (Supplemental Figure S1). Therefore, as proposed for the 4CL isoforms in Arabidopsis (Li et al., 2015), the differences in the spatiotemporal gene expression pattern of Os4CL3 and Os4CL4 may contribute to their different roles in monolignol and tricin biosynthesis for lignification in rice. Further high-resolution analysis of gene expression in specific locations and times during cell wall lignification is needed to investigate this aspect.

Both Os4CL3 and Os4CL4 contribute to cell wall feruloylation

Besides the unique lignin substructures, another feature of grass cell walls that distinguishes them from eudicot cell walls is the incorporation of FA into hemicelluloses (arabinoxylan). Such cell wall feruloylation contributes to the formation of the arabinoxylan–arabinoxylan and lignin–arabinoxylan cross-linked units that substantially affect the physicochemical properties of grass cell walls (Ralph, 2010; Karlen et al., 2016; Hatfield et al., 2017; Coomey et al., 2020). On the basis of chemical and 2D NMR data for the os4cl3 and os4cl4 mutant cell walls, we detected sharp decreases in the cell-wall-bound FA content in both mutant cell walls (Table 3 and Figure 4). Therefore, Os4CL3 and Os4CL4 could mediate cell wall feruloylation (Figure 6B), most likely by providing the feruloyl-CoA precursor for the BAHD acyltransferase implicated in arabinoxylan feruloylation (de Souza et al., 2018, 2019). However, earlier investigations on the in planta functions of rice 4CLs in specific varieties, namely japonica Zhonghua 11 (Gui et al., 2011) and indica Kasalath (Liu et al., 2020), did not detect any substantial decrease in the cell-wall-bound FA content in the Os4CL3- and Os4CL4-downregulated plants. This differs from the findings of this study on the os4cl3 and os4cl4 mutants with the japonica cultivar Nipponbare background. This discrepancy may be due to the differences in the genomic backgrounds and/or the differences in the tissues and developmental stages of the plant materials analyzed.

Along with reductions in the cell-wall-bound FA content, the cell-wall-bound pCA content decreased in os4cl3, but increased in os4cl4 (Table 3 and Figure 4). Given that a majority of cell-wall-bound pCA in grasses exists in the γ-p-coumaroylated lignin units (Ralph, 2010), the marked reduction in the cell-wall-bound pCA in os4cl3 is in accordance with the reduced lignin γ-acylation in os4cl3 as determined by DFRC (Figure 3C). On the other hand, as DFRC did not detect any increase in the γ-p-coumaroylated lignin units in the os4cl4 cell walls (Figure 3C), the increased cell-wall-bound pCA in os4cl4 might be associated with an increase of p-coumaroylated arabinoxylan units. Indeed, some earlier studies reported increments of cell-wall-bound pCA in transgenic grasses in which, like in our 4CL mutants, the biosynthesis of cell-wall-bound FA is disrupted (de Souza et al., 2018, 2019); further study is, however, needed to clarify why cell-wall-bound pCA is increased particularly in os4cl4.

In conclusion, our data demonstrated that Os4CL3 and Os4CL4 both play major roles, but differentially contribute to channeling the metabolic flux of the cinnamate/monolignol pathway dedicated to cell wall lignification in rice culms. The distinct roles of the two 4CLs in the biosynthesis of the conserved non-γ-p-coumaroylated monolignols and the grass-specific lignin monomers (i.e. γ-p-coumaroylated monolignols and tricin) support our view that these grass 4CLs along with other lignin-associated enzymes and transcription factors have diversified to fine-tune the production of diverse and complex grass lignin polymer units. Along with previous research on 4CLs (Hu et al., 1999; Xu et al., 2011; Van Acker et al., 2013; Jung et al., 2016; Park et al., 2017; de Vries et al., 2018; Tsai et al., 2020), this study also supports the notion that 4CL could be a viable target for future breeding and bioengineering to produce new grass varieties with a low-lignin biomass and enhanced cell wall digestibility (Supplemental Figure S4). Further elucidating the evolutionarily diverse regulatory mechanisms of grass lignin biosynthesis and the functions and properties of grass lignin structures may increase our understanding of the evolution of cell wall structures during the divergence of vascular plants, with implications for optimizing grass biomasses for the sustainable production of biochemicals.

Materials and methods

Phylogenetic analysis

Protein sequences of 4CLs were retrieved from Phytozome (version 11) (Goodstein et al., 2012) or the National Center for Biotechnology Information GenBank database and then aligned using ClustalW (Larkin et al., 2007). An unrooted phylogenetic tree was constructed according to the neighbor-joining method with 1,000 bootstrap replicates using MEGAX (Kumar et al., 2018).

Gene expression analysis

In silico spatiotemporal gene expression data were retrieved from the Rice Expression Profile Database (RiceXPro) (Sato et al., 2013). For the RT-qPCR analysis, total RNA samples were extracted from culms of plants at the heading stage and stems of 5-weeks-old seedlings, and reverse transcribed into first-strand cDNA using random hexamer primers (Invitrogen, San Diego, California, USA) (Koshiba et al., 2013). Gene expression was analyzed using the 7300 Real-time PCR system (Applied Biosystems, Foster City, California, USA) and gene-specific primers (Supplemental Table S4). A rice ubiquitin gene (OsUBQ5) was used as an Internal control.

Generation of CRISPR/Cas9 mutant rice

For the CRISPR/Cas9-mediated mutagenesis, two sgRNAs targeting the first exon of Os4CL3 and another two sgRNAs targeting the first exon of Os4CL4 were designed using the CRISPR-P 2.0 program (Liu et al., 2017) and then integrated into the pZH_OsU6gRNA_MMCas9 vector (Mikami et al., 2015) using the oligonucleotides listed in Supplemental Table S4. Embryonic rice (O. sativa L. cv. Nipponbare) calli were transformed with the CRISPR/Cas9 binary vectors using Agrobacterium tumefaciens strain EHA101 cells (Hiei et al., 1994). The regenerated T₀ plants were genotyped and grown to maturity in a growth chamber as previously described (Lam et al., 2017). Selected T₁ mutant lines were further genotyped and grown to maturity in a greenhouse as previously described (Lam et al., 2017). For genotyping, genomic DNA was extracted from young leaves, after which the genomic region containing the target site was amplified by PCR using the primers listed in Supplemental Table S4 and then directly sequenced (Takeda et al., 2019). To further confirm the genotype and zygosity of the selected T₁ mutant lines, the PCR products were also cloned into the pCR4-TOPO TA vector (Invitrogen) for the subsequent transformation of Escherichia coli DH5α recipient cells; at least 12 propagated colonies were sequenced for each plant (Takeda et al., 2019). For the off-target analysis, genomic regions containing the three top-ranked potential off-target sites (Supplemental Table S2) as predicted by the CRISPR-P 2.0 program (Liu et al., 2017) were amplified by PCR using the primers listed in Supplemental Table S4 and then directly sequenced. Fully genotyped T₁ mutant lines and wild-type rice grown together were used for the phenotypic characterization and cell wall analyses.

Crude protein extraction and 4CL enzyme assay

Stem samples were collected from 5-weeks-old seedlings, immediately frozen in liquid nitrogen, and then stored at −80°C until use. Crude protein extraction (Shigeto et al., 2017) and enzyme assay reactions (Lee et al., 1997; Gui et al., 2011) were conducted according to the procedures previously described with minor modifications. Briefly, samples were ground in liquid nitrogen and homogenized in 2.5 mL extraction buffer (200 mM Tris–HCl [pH 7.8], 20 mM ascorbic acid, 5% [v/v] β-mercaptoethanol, and 30% [v/v] glycerol) in the presence of 3% (w/v) polyvinylpolypyrrolidone and 2% (w/v) AG 1-X2 Resin (Bio-Rad, Hercules, California, USA) on ice for 10 min (Shigeto et al., 2017). The tissue homogenates were centrifuged at 10,000 × g for 20 min at 4°C and the supernatant was recovered and filtered through a Millex 0.45-µm filter (Merck Millipore, Billerica, Massachusetts, USA). The protein concentration of the crude extract was determined by the Bradford method (1976). To test CoA ligation activity, 300 µL of the reaction mixture containing 10 µg crude protein, 100 mM Tris–HCl (pH 7.5), 5 mM MgCl₂, 5 mM ATP, 330 µM CoA, 50 µM substrate (pCA, cinnamic acid, caffeic acid, FA, or sinapic acid) was incubation at 30°C for 1 h. The formation of CoA ester was monitored by the change in UV absorption at 311 nm for cinnamic acid, 333 nm for pCA, 346 nm for caffeic acid, 345 nm for FA, and 352 nm for sinapic acid using a Shimadzu UV-2600 spectrophotometer (Shimadzu Co, Ltd, Kyoto, Japan) (Lee et al., 1997; Gui et al., 2011). Extinction coefficients previously established for CoA esters (Lee et al., 1997) were used to calculate CoA ligation activity.

Histochemical analysis

Specimens (∼1 cm) were manually excised from culms at the heading stage and then fixed in a formaldehyde/propionic acid/ethanol (3.5:5:50, v/v/v) solution, decolorized in an ethanol/acetic acid (6:1, v/v) solution, and embedded in resin using Technovit 7100 (Heraeus Kulzer GmbH, Wehrheim, Germany) as described by the manufacturer. Sections (80-µm thick) were sliced using an RX-860 microtome (Yamato Kohki Industrial, Saitama, Japan). For lignin staining (Lam et al., 2017), sections were treated with 1% (w/v) phloroglucinol in ethanol for 10 min, acidified with 17.5% (w/w) HCl, and incubated for an additional 10 min. They were then examined using an Olympus BX51 microscope (Olympus Optical, Tokyo, Japan).

Chemical analysis

Rice CWR samples were prepared from dried mature rice culms via an extraction involving a water–methanol–hexane solvent series (Yamamura et al., 2012). The Klason lignin assay was performed as described by Hatfield et al. (1994). The Klason lignin content was corrected on the basis of the ash (Hatfield et al., 1994) and crude protein (Darvill et al., 1980; Jin et al. 2006) contents. Analytical thioacidolysis (Yamamura et al., 2012; Yue et al., 2012), DFRC (Karlen et al., 2018; Takeda et al., 2019), analyses of cell-wall-bound pCA and FA contents via a mild alkaline hydrolysis (Yamamura et al., 2011), and an analysis of the polysaccharide composition according to the two-step trifluoroacetic acid/sulfuric acid-catalyzed hydrolysis (Lam et al., 2017) were performed as previously described.

2D NMR analysis

The culm CWR samples (∼300 mg) were added to ZrO₂ vessels containing ZrO₂ ball bearings for the fine ball-milling using the Planetary Micro Mill Pulverisette 7 (Fritsch, Idar-Oberstein, Germany) (600 rpm; 12 cycles of 10 min, with 5 min intervals). For the NMR analysis of the whole cell walls, aliquots of the ball-milled CWR samples ( ∼60 mg) were directly swelled in 600 μL DMSO-d₆/pyridine-d₅ (4:1, v/v) and then subjected to the 2D HSQC NMR analysis (Kim and Ralph, 2010; Mansfield et al., 2012). For the NMR analysis of the enzyme lignin fractions, the remaining ball-milled CWR samples (∼240 mg) were digested with crude cellulases (Cellulysin; Calbiochem, La Jolla, California, USA) as described by Tobimatsu et al. (2013). The obtained rice enzyme lignin fractions (∼30 mg) were dissolved in 600 μL DMSO-d₆/pyridine-d₅ (4:1, v/v) and then subjected to the 2D HSQC NMR analysis. The HSQC NMR spectra were acquired using an Avance III 800 system (800 MHz; Bruker Biospin, Billerica, Massachusetts,, USA) equipped with a cryogenically cooled 5-mm TCI gradient probe. Adiabatic HSQC NMR experiments were conducted using a standard program (hsqcetgpsp.3; Bruker Biospin) and previously described parameters (Kim and Ralph, 2010; Mansfield et al., 2012). Data were processed and analyzed using the TopSpin version 4.0 software (Mac; Bruker Biospin) as previously described (Tarmadi et al., 2018; Dumond et al., 2021; Lam et al., 2022). The central DMSO-d₆ solvent peaks (δ_C/δ_H, 39.5/2.49 ppm) were used for calibrating the chemical shift. Signal assignments are listed in Supplemental Table S3. For the volume integration analysis of the whole-cell-wall spectra (Figure 4), aromatic C2–H2 correlations from lignin (S_2/6 and G₂), hydroxycinnamates (P_2/6 and F₂), and tricin (T_2ʹ/6ʹ) units and anomeric C1–H1 correlations from polysaccharide units (Gl₁, Xy₁, Xyʹ₁, Xy″₁, Ar₁, Ga₁, and GU₁) were manually integrated and the S_2/6, P_2/6, and T_2ʹ/6ʹ signals were logically halved. Each signal was normalized according to the sum of the integrated signals (½S_2/6 + G₂ + ½P_2/6 + F₂ + ½T_2′/6′ + A₁ + U₁ + Gl₁ + X₁ + X′₁ + X″₁ = 100). For the volume integration analysis of the enzyme lignin spectra (Figure 5), C2–H2 correlations from S (S_2/6), G (G₂), H (H_2/6), pCA (P_2/6), and tricin (T_2ʹ/6ʹ) aromatic signals, Cα–Hα contours from β–O–4 (I_α), β–5 (II_α), and β–β (III_α) inter-monomeric linkage signals, and Cγ–Hγ contours from the cinnamyl alcohol end-units (X1′_γ) were manually integrated and the S_2/6, H_2/6, P_2/6, T_2ʹ/6ʹ, III_α, and X1′_γ signals were logically halved. The aromatic unit signals were normalized on a ½S_2/6 + G₂ + ½H_2/6 + ½T_2ʹ/6ʹ + ½P_2/6 = 100 basis, whereas the inter-monomeric linkage and the cinnamyl alcohol end-unit signals were normalized on a I_α + II_α + ½III_α = 100 basis.

Enzymatic saccharification assay

The CWR samples (∼10 mg) were destarched and hydrolyzed for 24 h using an enzyme cocktail consisting of Celluclast 1.5 L (Novozymes, Bagsvaerd, Denmark) (1.1 FPU), Novozymes 188 (Novozymes) (2.5 CbU), and Ultraflo L (Novozymes) (65 μg) in 50 mM sodium citrate buffer (pH 4.8) (Hattori et al., 2012; Lam et al., 2022). The liberated glucose was measured using the Glucose CII test kit (Wako Pure Chemicals Industries, Osaka, Japan).

Statistical analysis

Student’s t test and a one-way ANOVA followed by Tukey’s multiple comparison test were performed using GraphPad Prism version 8 (GraphPad Software, San Diego, California, USA).

Accession numbers

The accession numbers for the protein and gene sequences used in this study are listed in Supplemental Table S1.

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. In silico gene expression analysis of rice 4CL genes.

Supplemental Figure S2. Morphology and histochemical analysis of rice mutants.

Supplemental Figure S3. Gene expression analysis of 4CL genes in rice seedlings.

Supplemental Figure S4. Enzymatic saccharification of rice mutant cell walls.

Supplemental Table S1. Accession numbers of 4CL protein sequences.

Supplemental Table S2. Off-target analysis of Os4CL3- and Os4CL4-deficient rice.

Supplemental Table S3. Peak assignments in 2D HSQC NMR spectra.

Supplemental Table S4. Primers and oligonucleotides used in this study.

 

O.A.A., Y.T., and T.U. conceived the research and wrote the manuscript with help from all the other authors. O.A.A., Y.T., P.Y.L., A.F.M., T.M., Y.O., and K.O. designed and performed the experiments, and analyzed the data.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plphys/pages/general-instructions) is: Yuki Tobimatsu (ytobimatsu@rish.kyoto-u.ac.jp).