Loss of function of folylpolyglutamate synthetase 1 reduces lignin content and improves cell wall digestibility in Arabidopsis

Background One-carbon (C1) metabolism is important for synthesizing a range of biologically important compounds that are essential for life. In plants, the C1 pathway is crucial for the synthesis of a large number of secondary metabolites, including lignin. Tetrahydrofolate and its derivatives, collectively referred to as folates, are crucial co-factors for C1 metabolic pathway enzymes. Given the link between the C1 and phenylpropanoid pathways, we evaluated whether folylpolyglutamate synthetase (FPGS), an enzyme that catalyzes the addition of a glutamate tail to folates to form folylpolyglutamates, can be a viable target for reducing cell wall recalcitrance in plants. Results Consistent with its role in lignocellulosic formation, FPGS1 was preferentially expressed in vascular tissues. Total lignin was low in fpgs1 plants leading to higher saccharification efficiency of the mutant. The decrease in total lignin in fpgs1 was mainly due to lower guaiacyl (G) lignin levels. Glycome profiling revealed subtle alterations in the cell walls of fpgs1. Further analyses of hemicellulosic polysaccharides by NMR showed that the degree of methylation of 4-O-methyl glucuronoxylan was reduced in the fpgs1 mutant. Microarray analysis and real-time qRT-PCR revealed that transcripts of a number of genes in the C1 and lignin pathways had altered expression in fpgs1 mutants. Consistent with the transcript changes of C1-related genes, a significant reduction in S-adenosyl-l-methionine content was detected in the fpgs1 mutant. The modified expression of the various methyltransferases and lignin-related genes indicate possible feedback regulation of C1 pathway-mediated lignin biosynthesis. Conclusions Our observations provide genetic and biochemical support for the importance of folylpolyglutamates in the lignocellulosic pathway and reinforces previous observations that targeting a single FPGS isoform for down-regulation leads to reduced lignin in plants. Because fpgs1 mutants had no dramatic defects in above ground biomass, selective down-regulation of individual components of C1 metabolism is an approach that should be explored further for the improvement of lignocellulosic feedstocks. Electronic supplementary material The online version of this article (doi:10.1186/s13068-015-0403-z) contains supplementary material, which is available to authorized users.


Background
One-carbon (C1) metabolism is important for synthesizing a range of biologically important compounds such as purines and amino acids [1]. C1 enzymes use tetrahydrofolate and its derivatives, collectively referred to as folates, as cofactors for many of the biochemical reactions that they catalyze [2]. Methionine is an important product of the C1 metabolic pathway because it is the direct precursor of S-adenosyl l-methionine (AdoMet), the universal methyl-group donor, which is crucial for the synthesis of a large number of secondary metabolites [3], such as lignin, betaines, phytohormones [4], and hemicellulosic 4-O-methyl glucuronoxylan (GX) [5].
Lignin is a complex, polymeric phenylpropanoidderived compound, which confers mechanical strength to the cell wall by cross-linking different polysaccharides, especially in fibers and tracheary elements [6]. The presence of lignin in secondary cell walls creates a hydrophobic environment in the vascular tissue for water conductance [7]. In plants, there are three major types of lignin monomers, namely p-coumaryl, coniferyl, and sinapyl alcohol [8]. Lignin biosynthesis is one of the most thoroughly investigated metabolic pathways in plants, but details on intermediate steps remain contentious and we have just started learning about the metabolic complexes that regulate flux through the phenylpropanoid pathway [9][10][11][12]. Recently, interaction of lignin with other wall matrix components has become a focus of cell wall research [13].
Lignin accounts for about 30 % of organic carbon on Earth [14] and is one of the major secondary compounds that utilizes a significant amount of AdoMet, the universal C1 pathway methyl donor. This has been shown in birch wood when the combined demands of several primary and secondary metabolites have been compared [3]. However, there is very little molecular evidence to show that these two pathways are directly linked. Although early radio-tracer experiments demonstrated that the methyl carbon of methionine, a precursor of AdoMet, was found in the O-methyl group of lignin [15], most of the studies to date have only provided indirect evidence linking C1 metabolism to lignin biosynthesis. For example, enriched transcripts of C1 metabolism genes in the lignin-rich vascular-tissues of various plant species were used to argue for a tight link between lignin biosynthesis and C1 reactions [16][17][18]. Only recently, mutant studies in model and crop plants are beginning to shed more light on the relationship between lignin biosynthesis and C1 metabolism. For instance, it was shown that a single-point mutation in the S-adenosylmethionine synthetase3 (SAMS3) gene resulted in decreased AdoMet and lignin content in Arabidopsis [19]. More recent research in maize has demonstrated that a mutation in the methylenetetrahydrofolate reductase (MTHFR) gene resulted in reduced lignin [20].
The supply of methyl units is also important for the synthesis of other major components of secondary walls such as the hemicellulose 4-O-methyl glucuronoxylan (GX) [21]. A study conducted on Arabidopsis glucuronoxylan-specific methyltransferase (AtGXMT1-an enzyme that catalyzes 4-O-methylation of the glucuronic acid substituents) showed that AdoMet is the donor of methyl units to AtGXMT1 [5]. Therefore, C1 metabolism is vital for maintaining the proper synthesis of various polymers that are important in modulating the structure and physical properties of the cell wall.
In plants, the synthesis of folates involves three cellular compartments, the cytosol, mitochondria, and plastids [22]. Polyglutamylation of folates is essential for the cellular retention of this family of compounds and results in more efficient reactions for many folate-dependent enzymes [23,24]. Most cellular folates carry a short polygamma-glutamate tail. Research has shown that removing these tails can reduce the total folate content up to 40 % in Arabidopsis [25] and cause major metabolic shifts and phenotypic changes in plants [26]. Folate polyglutamylation is carried out by the enzyme folylpolyglutamate synthetase (FPGS). Initial gene cloning, functional characterization, and cellular localization of the three isoforms of FPGS [FPGS1 (plastid), FPGS2 (mitochondria), and FPGS3 (cytosol)] in Arabidopsis were described by Ravanel et al. [27]. Recently, the roles of FPGS genes in Arabidopsis have been explored through mutant analysis [24,26,[28][29][30][31]. In addition to its impacts on early seedling development and root growth [24,26,30,31], mutation of FPGS1 caused changes in DNA methylation and the histone H3K9 dimethylation status of the Arabidopsis genome [29]. There is also a recent study showing that the maize brown midrib 4 (bm4) mutant, which is disrupted in a gene encoding a functional FPGS, has reduced lignin. The lower lignin content in the bm4 mutant further reinforces the importance of C1 pathway in lignin biosynthesis [32].
Previously, we showed that fpgs1 mutants in Arabidopsis led to reduced levels of methionine and other C1 metabolic intermediates in young seedlings [26]. As a result, primary roots of the seedlings failed to develop properly. Despite the early root developmental defects, fpgs1 mutants had above-ground growth comparable to wild-type plants [26]. Although there are recent reports that folate mutants in maize have reduced lignin [20,32], it is not clear whether lower lignin resulting from altered folate metabolism leads to a corresponding reduction in cell-wall recalcitrance. Here we show that loss of FPGS1 function in Arabidopsis leads to lower lignin and reduced cell-wall recalcitrance. The reduced lignin observed in fpgs1 mutants might not only be due to reduced flux of methyl units to lignin precursors, but is also a consequence of changes in the expression of genes associated with lignin biosynthesis and cell wall remodeling. These changes, in turn, result in plants with enhanced digestibility and sugar release efficiency, which are important requirements for efficient biofuel processing.

FPGS1 is preferentially expressed in vascular tissues, consistent with its role in lignin biosynthesis
Lignification in plants occurs predominantly in the vascular tissues where secondary cell walls are formed. It has been shown previously that several C1 pathway genes that supply methyl units for lignin biosynthesis were enriched in the vascular tissues [17]. Consistent with previous reports, we found that the FPGS1 promoter fused to β-glucuronidase (GUS) (pFPGS1::GUS) was predominantly expressed in the vascular tissues of cotyledons, hypocotyls, roots of seedlings and inflorescence stems ( Fig. 1a-e). Based on both cross-and longitudinal sections of the transgenic plant inflorescence stems, GUS staining was mainly concentrated at the fascicular cambium region and the transition tissues from protoxylem to metaxylem (Fig. 1c-e).
The FPGS1 expression pattern was further examined using green fluorescent protein (GFP). The entire sequence of FPGS1 consisting of the 7-kb genomic DNA fragment was fused to GFP and transformed into the fpgs1 mutant. The transgenic lines carrying FPGS1 promoter::FPGS1-GFP (pFPGS1::FPGS1-GFP) construct revealed that FPGS1 expression was most distinct in the cytosol of developing vessel elements adjacent to the metaxylem (Fig. 1f ). We previously showed that pFPGS1::FPGS1-GFP is localized in both chloroplasts and the cytosol when transiently expressed in tobacco leaf epidermal cells [33]. This observation is consistent with earlier reports that the Arabidopsis FPGS1 is located in plastids and the cytosol [30]. The pFPGS1::FPGS1-GFP was shown to complement the seedling root architectural defects of the fpgs1 mutant; however, it localized mostly to the cytosol in roots [30] similar to observations made here in developing vessel elements. The observed FPGS1 expression in the vascular tissues from GUS and GFPreporter constructs was further supported by our earlier published in situ hybridization results [26] and publicly available microarray data sets [34] (Additional file 1: Fig.  S1).
Unlike FGPS1, expression of the other two Arabidopsis FPGS isoforms (i.e. FPGS2 and FPGS3) is not particularly high in stems compared to other tissues. FPGS2 is most highly expressed in leaves, whereas FPGS3 is most predominantly expressed in rosette leaves and floral organs as suggested by publicly available microarray data sets [34]. Publicly available microarray data sets also show that FPGS2 and FPGS3 have lower expression in vascular tissues compared to FPGS1 [34].

Lignin content is reduced in fpgs1 mutants
Metabolic profiling revealed that levels of many phenolic acids (cinnamic, p-coumaric, ferulic, and sinapic acids), as well as phenylalanine and coniferyl alcohol, were significantly reduced in 7-day-old fpgs1-1 mutants compared to wild type ( [26]; Additional file 2: Fig. S2). These data prompted us to examine lignin content and composition in the fpgs1-1 mutants. To complement our lignin histochemical assays, lignin levels were quantified using the acetyl bromide (AcBr) method. We found a 17 % reduction in total lignin content in 35-day-old fpgs1 plants compared to controls (Fig. 2). Lignin content of mutants of the other two isoforms of the three-membered Arabidopsis FPGS family (fpgs2 and fpgs3) did not differ from that of the wild type (Fig. 2a). Two other fpgs1 mutant alleles, namely fpgs1-2 and fpgs1-3, also exhibited a significant reduction in total lignin (Fig. 2b). Additional evidence that a mutation in FPGS1 was the cause of the lower lignin levels in mutant plants was obtained by introducing the pFPGS1::FPGS1-GFP construct into the fpgs1-1 background. The fpgs1-1 plants expressing the pFPGS1::FPGS1-GFP construct restored lignin to wildtype levels (Fig. 2a). The thioacidolysis method was used to determine lignin composition in the fpgs1 mutants and wild type. Compared with the wild-type plants, a 22.5, 17.7, and 17.8 % reduction in G-lignin monomer was observed in the three fpgs1 mutant alleles accordingly. No significant changes in S-or H-lignin monomers were observed in all fpgs1 mutant alleles compared to the wildtype plants (Fig. 2b).

Transcript profiling of fpgs1 inflorescence stems reveals down-regulation of genes in the phenylpropanoid, methylation, and C1 pathways
To determine if changes in gene expression due to the FPGS1 mutation could partly explain the reduced lignin phenotype of the fpgs1 mutant, we conducted microarray analysis of 35-day-old inflorescence stem tissues using the ATH1 Affymetrix chip. Using a Bonferroni-corrected P value of 2.19202E−06 and a ratio of two as the cutoffs, there were 409 genes differentially expressed between the fpgs1-1 mutants and wild-type plants. In the mutant, 182 genes had higher expression and 227 had lower expression (Additional file 3: Table S1/Sheet 2). The most up-regulated gene in the mutant was a member of the R2R3-MYB transcription factor gene family. Although the specific function of this R2R3 gene is not known, several members of this family of transcription factors have been shown to regulate lignin biosynthesis [9,10]. Two genes encoding O-methyltransferases and two encoding glycine-rich family proteins were among the top 10 most down-regulated genes in the mutant. At a ratio cutoff of two, 11 putative phenylpropanoid pathway-related genes were among the differentially expressed genes, and the majority of these were down-regulated, including one putative caffeoyl-CoA 3-O-methyltransferase (CCoAOMT/At1g67980) (Additional file 3: Table S1/ Sheet 3). A total of 16,484 probe sets were detected in the stem tissues based on the "Presence" call in at least four out of the six stem samples or in all three replicates in either wild-type or the fpgs1 mutant based on the dCHIP program (http://www.hsph.harvard.edu/cli/complab/ dchip/). Among them, 135 probe sets were known or homologous to known lignin pathway enzymes, based on the strategy of Costa et al. [35] (see http://cahnrscms.wsu.edu/ibc/research/lewis/nsf/Pages/Genes.aspx) (Additional file 3: Table S1/Sheet 2/sheet3). Using a ratio cutoff of 1.5, the expression of 25 probe sets annotated as known or putative lignin pathway genes (18.5 %, 2.53 times enrichment over the background) were affected in the fpgs1 mutant. Similarly, 244 probe sets on the chip annotated as methyltransferases were detected in the stem tissue. Among the genes annotated as methyltransferases, the expression of 25 probe sets (10.3 %, 1.42 times enrichment over the background) was affected in the fpgs1 mutant with most of them down-regulated (Additional file 3: Table S1/Sheet 4).
Among the genes directly involved in the C1 metabolic pathway, expression of homocysteine S-methyltrans-ferase3 (HMT3), which encodes an enzyme that converts homocysteine to methionine in the S-methylmethionine (SMM) cycle, was six times higher in the mutant. Similarly, the expression of S-adenosylmethionine synthetase 3 and 4 (SAMS3, SAMS4) was also up-regulated. On the other hand, expression of SAMS1, SAMS2, and several AdoMet-dependent methyl transferases was significantly down-regulated (Table 1; Additional file 3: Table S1). We also observed that significant numbers of genes involved in the sulfur assimilation pathway were affected in the mutant. Among 28 genes involved in sulfur metabolism, 13 were affected in the fpgs1 mutant, with the majority of them up-regulated (Additional file 3: Table S1/Sheet 5).

Levels of the methyl group donor AdoMet and the degree of methylation of GX were reduced in fpgs1
AdoMet is the major methyl-group donor for numerous transmethylation reactions [36]. Methionine is the immediate precursor of AdoMet and we have shown previously that methionine levels are reduced in fpgs1 mutant seedlings [26]. To determine if the mutation in FPGS1 also affects AdoMet content, we quantified AdoMet by HPLC based on the method described by Castro et al. ([37]; Additional file 4: Fig. S3). Extracts from fpgs1-1 mutant stems had about 50 % less AdoMet content than extracts from wild-type stems (Fig. 3a).
GX and lignin are the major polymers in the secondary cell wall that contain methyl groups. To determine if the reduced AdoMet content found in the fpgs1-1 stems alters xylan methyletherification, we analyzed the structure of GX isolated from 35-day-old inflorescence stems of wild-type and fpgs1-1 mutant by 1 H-NMR spectroscopy, as described previously [5]. The NMR analyses showed that the degree of GlcA O-methylation of GX in fpgs1-1 was 10 % lower than wild type (Fig. 3b, c). Other GX structural features, including the amount of branching and degree of polymerization were identical.

Glycome profiling reveals subtle alterations in the cell walls of fpgs1 inflorescence stems
Changes in lignin composition typically lead to corresponding modifications in other components of the cell wall [38][39][40][41][42]. To determine if there were any alterations in the overall extractability of non-cellulosic glycan epitopes in the fpgs1-1 mutant, glycome profiling of the cell wall material (alcohol insoluble residues; AIR) isolated from 35-day-old inflorescence stems of wild-type and fpgs1-1 mutants was conducted (Fig. 4a). The glycome profiles of the fpgs1-1 mutants were mostly similar to those of wild-type lines. However, a subtle reduction in the overall abundance of xyloglucan (non-fucosylated and fucosylated) and xylan epitopes was observed in the 1 M KOH and chlorite extracts of fpgs1-1 mutant walls (see white dotted blocks in Fig. 4a). Overall, when the total amounts of carbohydrate released during the various extractions were quantitated, it was noted that mild extractions released higher amounts of carbohydrate from cell walls of fpgs1-1 inflorescence stems (significantly higher in oxalate and marginally higher in carbonate extracts) indicating that the extractability of cell walls from fpgs1-1 was slightly higher than from wild type. Differences between fpgs1-1 and wild-type wall extracts were also noted with respect to the binding of individual antibodies (see white arrows in Fig. 4a). Based on replicated data of glycome profiling, several mAbs that exhibited significant (student t test, P < 0.05) altered binding responses to the extracts from fpgs1-1 mutants' cell walls in comparison to wild type indicated changes in the abundances of their epitope structures in the extracts from mutants. For instance, in mutant cell wall extracts, enhanced binding responses were noted for some mAbs, including CCRC-M96 (belonging to the non-fucosylated xyloglucan-5 clade) in oxalate, chlorite, and PC 4 M KOH extracts; CCRC-M107 (belonging to the arabinogalactan-2 clade) in oxalate and carbonate extracts; and CCRC-M131 (belonging to the homogalacturonan  Table S3). These data, in general, suggest potential cell wall alterations in fpgs1-1.
To determine if these minor differences in glycome profiling between wild type and fpgs1-1 are reflected in situ, immunolabeling of cross-sections of 35-day-old inflorescence stems was conducted using monoclonal antibodies directed against a fucosylated xyloglucan epitope (CCRC-M1), a xylan epitope (CCRC-M152), an arabinogalactan side chain epitope of RG-I (CCRC-M25/ RG-I/AG) and a non-methylesterified homogalacturonan epitope (CCRC-M38). All of these mAbs labeled wildtype and fpgs1-1 mutant stems in a similar fashion, indicating no major compositional differences in the cell wall epitopes recognized by these antibodies between wildtype and the fpgs1 mutants (Fig. 4b). It is worth noting that CCRC-M25 showed a significant difference in the glycome profiles for the PC 4 M KOH extracts, but such differences were not obvious in the in situ imaging.
Since homogalacturonan is another major sink for methyl groups, the difference in pectin methylation was examined by looking at the glycome profiling data for JIM5 and JIM7, antibodies that distinguish pectin with different degrees of methylation [43,44], neither showed significance differences between wild type and fpgs1-1. This observation was consistent with immunolocalization of stem inflorescence sections, which also did not reveal differences in labeling intensities by these two antibodies between wild type and fpgs1-1. Taken together, our results suggest that subtle changes in wall extractability (as reflected in the glycome profiles) in the mutant are not accompanied by obvious changes in wall epitope composition, including pectin methylation.

Saccharification efficiency is enhanced in fpgs1 mutants
Mature whole plants (stem and leaves) were analyzed for sugar release. Sugar release from the cell wall residues (CWRs) without acid pretreatment was 41 and 19 % more in the fpgs1-1 and fpgs1-3 plants, respectively, when compared to wild-type plants (Fig. 5a). Enzymatic hydrolysis of the CWR with acid pre-treatment showed 23 and 37 %  higher sugar release in the fpgs1-1 and fpgs1-3 mutants, respectively, compared to wild-type plants (Fig. 5b). Accordingly, the saccharification efficiency of the mutant increased around 14-20 % without acid pre-treatment and around 18-34 % with acid pre-treatment (Fig. 5c,  d). It is worth pointing out that the sugar release of the pFPGS1::FPGS1-GFP complemented lines (FPGS1c) was similar to the wild type, providing evidence that the improved saccharification efficiency in the fpgs1 mutant is indeed directly attributable to the mutation.
An independent test was performed on wild-type and fpgs1 CWRs using more sensitive HPLC methods. Although no major structural changes in the cell wall were noticed, a significant (5.88 %) increase in the glucan composition was observed in the fpgs1-1 mutant (Additional file 6: Fig. S4A). Glucose and xylose released from hydrothermal pretreatment and co-hydrolysis were measured and we found a 24 and 8 % higher release of glucose and xylose, respectively, in the fpgs1 mutant when samples were hydrothermally pretreated (Additional file 6: Fig. S4B), and 19 and 13 % higher release of glucose and xylose, respectively, when samples were digested without any pretreatment (Additional file 6: Fig.  S4C).

Discussion
Increasing cell wall digestibility has been achieved using various means [45,46] but, mainly through the manipulation of genes encoding enzymes in the lignin biosynthetic pathway or those that directly modulate their expression Means with * are statistically significant over WT (Tukey's test, P < 0.05) [38][39][40][41][42]. Novel approaches are also being sought to reduce recalcitrance and methods such as partial substitution of normal monolignols with phenolic precursors or novel monolignols to increase carbohydrate accessibility, increase digestible sugar contents in the feedstock, and reduce substrate-derived fermentation inhibitors, all resulted in reduced recalcitrance [45,46]. Here, we report that the FPGS1 gene, a known component of C1 metabolism that supplies methyl units to lignin precursors and other cell wall components, can be an alternative target for reducing lignin content and cell wall recalcitrance. Although defects in growth were observed during early seedling development [26,30,31], such defects become less obvious in mature plants (Additional file 7: Fig. S5).
The FPGS1 localization in the xylem suggested a possible involvement of FPGS1 in lignin biosynthesis, as the xylem is a major site for this pathway. Other genes, such as Acaulis5 [47], SAMS3 [19], COMT, and CCoAOMT [48][49][50][51][52], which are directly involved in xylem development or lignification, were also expressed in the vascular tissues. In wild-type Arabidopsis stem tissues, transcript levels of FPGS1 were 27 times and 7 times more abundant than FPGS2 and FPGS3, respectively. This indicates that FPGS1 is the dominant FPGS isoform in the stem (Additional file 3: Table S1). The lower expression of FPGS2 and FPGS3 in stems and vascular tissues relative to FPGS1 may explain why their corresponding mutants did not show reduced stem lignin. However, we cannot discount the possibility that other tissues in fpgs2 and fpgs3 mutants have reduced lignin.
In this study, we recorded approximately a fifty percent reduction of AdoMet in fpgs1-1 stem tissues, which corresponded to about a 17 % reduction in lignin content. While a smaller impact on downstream product is generally expected due to concentrations, enzyme K M and regulation, we cannot discount the possibility that there may be alternative sources of methyl units for the lignin biosynthetic pathway. In addition to those supplied by the other functional FPGS isoforms (FPGS2 and FPGS3), other sources of methyl units could be obtained through the S-methylmethionine (SMM) cycle. Through the tandem action of Met S-methyltransferase (MMT) and Hcy S-methyltransferase (HMT), the SMM cycle transports SMM from source tissues to sink tissues to be converted to methionine, which further replenishes the depleted AdoMet pools [3,53,54]. In the fpgs1 mutant stems, transcripts of HMT3 were up-regulated six times, indicating potential compensation of methyl units through the SMM cycle in the fpgs1.
The direct impact of the FPGS1 gene product on AdoMet content could also be due to the fact that within the C1 pathway, methionine synthases cannot use monoglutamylated folates as methyl donors [55,56]. Due to lower AdoMet levels in fpgs1, the conversion of cinnamic acid into monolignols via O-methylation reactions catalyzed by CCoAOMT or COMTs (Additional file 8: Fig. S6); [6,57,58] could be compromised. The reduced lignin content in fpgs1 knockout plants mostly showed reduced levels of G-lignin and is reminiscent of previous reports where down-regulation of CCoAOMT resulted in less lignin because of the reduction mainly in G units in different species [6,[48][49][50][51][52][59][60][61][62][63]. In the fpgs1 mutant, transcript levels of CCoAOMT1 (AT4G34050) or COMT1 (AT5G54160) were not affected. It is worth noting, however, that transcripts of several OMTs and one putative CCoAOMT (AT1G67980) were significantly down-regulated in the fpgs1 stem tissues (Additional file 3: Table S1). Although, direct involvement of these OMTs in the lignin biosynthesis has not been established, it is likely that some of these genes are involved in cell wall biosynthesis. Elucidating the function of these OMT genes will be the subject of future research.
Relationships among metabolic pathways have been shown in previous studies through transcriptomics analysis. For example, Loizeau et al. [64] found that antifolate methotrexate (MTX) treatment on Arabidopsis suspension cells induced significant changes in the expression of 4538 genes encoding FPGS, methionine synthases, methylene-THF reductases, S-adenosylmethionine synthetases, and methionine S-methyltransferases. Changes in the expression of these genes were associated with a significant modification in the distribution of C1 derivatives, resulting in the reorientation of C1 units towards the synthesis of critical biological compounds [64].
Recently, system-wide analysis of lignin biosynthesis pathway mutants also revealed that shikimate, phenylpropanoid, and methyl donor pathways are tightly coregulated [41].
The essence of plastidial FPGS1 is not just restricted to the phenylpropanoid pathway and C1 metabolism, but also for the sulfur homeostasis in plastids. In support of this notion is the observation that changes in expression of nearly 13 out of the 28 genes involved in sulfur metabolism, including up-regulation of a few sulphate transporter (SULTR) genes, occurred in the fpgs1 mutants. Changes in the expression of SULTR genes are typically indicative of sulfur starvation [65]. It is noteworthy that methionine salvage and AdoMet play essential roles in sulfur, ethylene, and polyamine biosynthetic pathways [66]. The altered expression of genes involved in sulfur metabolism in fpgs1 was not surprising because the plastid is the major organelle for sulfur metabolism and the two sulfur-containing amino acids, Met and cysteine, are key players in the C1 pathway. Sulfur deprivation in plastids could reduce the AdoMet content up to tenfold [67]. Changes in sulfur metabolism in the fpgs1 mutant were also reflected by the fact that a significant number of genes involved in methionine-derived glucosinolate biosynthesis were up-regulated. Sulfate assimilation and glucosinolate biosynthetic are known to be interconnected [68]. Collectively, these metabolic shifts may have both direct and indirect impacts on the synthesis of lignin and secondary cell wall components.
In stem tissues, other pathways involved in secondary cell wall formation, such as the xylan biosynthetic pathway, also require methylation reactions [5,69]. Our microarray analysis of the fpgs1 mutant showed that several AdoMet-dependent methyltransferases, especially O-methyltransferases, were down-regulated in the mutant. The 10 % reduction in the GlcA O-methylation of GX in the fpgs1-1 mutant further supports the involvement of FPGS1 in secondary cell wall synthesis. The subtle changes in cell wall polysaccharide composition revealed by glycome profiling also indicate that changes in the C1 pathway may affect cell wall structure.
Reduction of lignin content or alteration of lignin composition has been shown to be associated with increased saccharification efficiency and reduced recalcitrance in many crops. For example, down-regulation of COMT and CAD increased saccharification efficiency in switchgrass by 16.5 and 23 %, respectively [70][71][72]. Furthermore, modification of various genes in the lignin biosynthesis pathway from different species has clearly demonstrated lignin down-regulation [38][39][40][41][42][73][74][75][76]. Recent work in Arabidopsis using system biology approaches showed improved sugar release in most of the lignin downregulated plants [42]. The partial redundancy in FPGS function could be exploited for the generation of biofuel crops with lowered lignin content and minimal undesired effects on biomass yield. It should be noted, however, that FPGS2, which encodes the mitochondrial FPGS isoform, is required for efficient nitrogen utilization [28], and recently, hypocotyls showed reduced growth in darkgrown fpgs1 seedlings [31]. Thus, we cannot rule out the possibility that down-regulation of FPGS1 could also lead to other plant developmental changes that are only manifested under certain environmental conditions. Nonetheless, the fact that FPGS1 impacts lignin formation without any major adverse effects on plant growth makes it a promising target for improvement of lignocellulosic feedstocks for sustainable biofuel production.

Conclusion
The present analysis of the Arabidopsis fpgs1 mutants conclusively demonstrates that specific expression of FPGS1 in the vascular tissues contributes to the supply of methyl units for methylation reactions in the phenylpropanoid pathway. The increase in saccharification efficiency in the fpgs1 mutant without dramatic negative effects on biomass traits points to an alternative target for improvement of lignocellulosic feedstocks. Our work also demonstrates that the essence of plastidial FPGS1 is not just restricted to the lignocellulosic pathway, but also associated with the sulfur and glucosinolate pathways, which furthers our understanding of the complex regulation of plant secondary metabolism.

Characterization of fpgs1 mutants and plasmid construction
The fpgs1-1 (originally designated as drh2 or atdfb-1) mutant was identified through a forward genetic screen as described previously [26]. Two additional T-DNA insertion lines (SALK_015472, and SAIL_556_G08) at the FPGS1 locus (AT5G05980) acquired from ABRC (https:// abrc.osu.edu) were also used in this study and renamed as fpgs1-2 and fpgs1-3, respectively. Plant growth and molecular characterization of all of these mutants were as described by Srivastava et al. [26]. Most of the analyses described here were conducted on the fpgs1-1 mutant unless otherwise indicated.
All constructs were made using Gateway ™ technology (Invitrogen, Now Life Technologies). The entry clones were obtained using the pENTR-D-TOPO vector (Invitrogen) and expression vectors using the Gateway system [77]. These were propagated in the E. coli strain Top10 (Invitrogen) and DH5α, respectively. The templates used to clone all genes were genomic DNA or cDNA derived from A. thaliana Col-0 RNA. The 2-kb promoter region (upstream of the ATG start codon of FPGS1) was amplified using Taq DNA polymerase (Invitrogen), cloned into pENTR-D-TOPO, and later fused to uidA GUS reporter gene via LR-reaction in the destination vector pMDC162 for tissue expression analysis. The constructs were transformed into Agrobacterium tumefaciens strain LBA4404. For complementation studies, a 7-kb genomic DNA fragment containing the pAtFPGS1-FPGS1 was amplified using gene-specific primers (Additional file 9: Table S2). This construct is comprised of 2 kb upstream promoter and 5 kb downstream regions of the ATG start codon. This was cloned into pMDC107 binary vectors and transformed into A. tumefaciens LBA4404. The fpgs1-1 mutant was then transformed using the floral-dip method [78]. Transgenic plants were selected on 25 μg ml −1 hygromycin plates and later propagated on soil. Around ten independent T2 transgenic lines were tested and analyzed for fluorescence and complementation.

Lignin analysis
Lignin content and composition were measured in the post-flowering 35-day-old stem tissues by the AcBr [79] and thioacidolysis [80] methods. At this stage, plants had well-developed inflorescence stems, which was optimal for lignin quantification. Because 35-day-old plants were still mostly green, this time point was chosen so we could obtain quality RNA for gene expression data that corresponded to lignin analysis. Lignin-derived monomers were identified and quantified by GC/MS using a Hewlett-Packard 5890 series II gas chromatograph with a 5971 series mass selective detector (column: HP-1, 60 m × 0.25 mm × 0.25 μm film thickness). Mass spectra were recorded in electron impact mode (70 eV) with 60-650 m/z scanning range. Five biological replicates were included for each assay. For each biological replicate, mature inflorescence stems of 20 individual plants were pooled and assayed.

Microscope image acquisition
Confocal imaging was performed using a Leica TCS SP2 confocal microscope (Leica Microsystems) and the HCX PL APO 63 ×/1.2 W water-immersion objective. All images were analyzed using Leica Confocal Software. GFP channels were acquired by scanning sections using the 488-nm laser line for excitation. Emission of the GFP signal was detected between 500 and 530 nm. Microscopy of vascular tissues was carried out after preparing 100 µm longitudinal sections of 35-day-old fresh stem tissues using Vibratome 1000 plus. GUS staining was performed on 35-day-old stem tissues and 7-day-old whole seedlings as described in Srivastava et al. [26].

Affymetrix microarray and qRT-PCR analysis
The Affymetrix arrays and qRT-PCR experiments were performed using total RNA from 35-day-old inflorescence stem basal portion. Three biological replicates were included for both wild-type control and fpgs1 mutant and each replicate had ten pooled plants. Total RNA was isolated using an RNeasy Mini Kit (Qiagen, Valencia, CA). RNA quantification and QC was as described previously [81]. For each array experiment, 500 ng of total RNA was used for labeling using the Affymetrix GeneChip ® Arabidopsis ATH1 Genome Array (Affymetrix, Santa Clara, CA). Probe labeling, chip hybridization, and scanning were performed according to the manufacturer's instructions for IVT Express Labeling Kit (Affymetrix). Data normalization and analysis were conducted as described earlier [81]. The microarray data were submitted to ArrayExpress and the accession no. is E-MEXP-3915.
Real-time qRT-PCR was performed using previously described procedures [26]. The plant-specific EF4A2 (At1g54270) gene was used as a control for constitutive gene expression. The primers used for this analysis are presented in Additional file 9: Table S2. Three biological replicates and three technical replicates were used for each sample group.

Analysis of AdoMet levels
Two hundred mg lyophilized ground samples of wild type and the mutant stem tissue were put into 10 ml ethanol: water (70:30 v/v) mix and samples were centrifuged at 14,000 rpm for 15 min at 4 °C. Supernatant was collected and further processing and analysis of the sample were done as described by Castro et al. [37]. UPLC/FLR analysis of AdoMet derivatives was carried out using a Waters Acquity UPLC system (Waters) coupled to FLR detector (Waters). A Waters Acquity UPLC 2.1 × 150 mm, BEH C18 column was used for analysis at 20 °C. The mobile phase consisted of 0.1 M sodium acetate in HPLC grade water, pH 4.5 (A), and HPLC grade acetonitrile (B). Using an injection volume of 2.0 µl, separation was achieved using the following gradient: isocratic at 95.8 % A for 7.8 min; 95.8 % A to 50 % A in 6 min; isocratic at 50 % A for 1 min followed by re-equilibration at 95.8 % A for 7 min at a flow rate of 0.280 ml/min. The fluorescent AdoMet derivatives were monitored at excitation of 270 nm and emission of 410 nm with a PMT Gain setting of 1.00 and data rate of 5 pts/s. A total of five biological replicates were used, and 20 plants in total were pooled for each replicate. Identification of compounds was based on retention time of authentic standards. Data were processed using Empower 2 Software (Waters).

Preparation of glucuronoxylan oligosaccharides and 1 H-NMR spectroscopy
Alcohol insoluble residues (AIR) were obtained from 35-day-old inflorescence stems of wild-type and fpgs1-1 mutant and treated sequentially with specific endoglycanases as described [82]. The enzyme-treated residue was then extracted with 1 N KOH containing 1 % (w/v) NaBH 4 and then with 4 N KOH containing 1 % (w/v) NaBH 4 . The 1 and 4 N KOH soluble fractions were neutralized with glacial acetic acid, dialyzed against deionized water and lyophilized. Glucuronoxylan oligosaccharides were generated by treating the 1 N KOH soluble fractions for 24 h at 37 °C with a Trichoderma viride endoxylanase (M1, 0.04 units/10 mg polysaccharide; Megazyme).
The glucuronoxylan-derived oligosaccharides were characterized by 1 H NMR spectroscopy as described in Urbanowicz et al. [5]. Glucuronoxylan oligosaccharides (1-2 mg) were dissolved in D 2 O (0.25 mL, 99.9 %; Cambridge Isotope Laboratories, Andover, MA). Oneand two-dimensional NMR spectra were recorded at 298°K with a Varian Inova-NMR spectrometer (Agilent Technologies, Santa Clara CA) operating at 600 MHz for 1 H and equipped with a 5-mm NMR cold probe as described [5]. Chemical shifts were measured relative to internal acetone (δ 2.225). Data were processed using MestReNova software (Universidad de Santiago de Compostela, Spain). The extent of GlcA methylation was determined by integration of the signals corresponding to the anomeric protons of GlcA and methylated GlcA in the 1D NMR spectra. The areas of the overlapping signals were determined using the deconvolution method in MestReNova. The degree of GlcA O-methylation was calculated using the data obtained in the NMR analysis of GX oligosaccharides obtained by two independent endoxylanase treatments.

Glycome profiling and immunofluorescence labeling
Preparation of cell wall AIR, cell wall fractionation, and glycome profiling were carried out as described previously [83,84]. Plant glycan-directed monoclonal antibodies [85]; Additional file 10: Appendix S1) were obtained as hybridoma cell culture supernatants from laboratory stocks at the Complex Carbohydrate Research Center [CCRC, JIM, and MAC series; available from Carbo-Source Services (http://www.carbosource.net)]. For immuno-labeling experiments, 35-day-old stem from both wild type and fgps1-1 were fixed, sectioned, and immunolabeled as described by Avci et al. [86].

Determination of saccharification efficiency
Cell-wall residues (CWRs) used in lignin analysis were also used for sugar analyses as described previously [70,71] with slight modifications. Enzymatic saccharification of Arabidopsis samples (with and without acid pre-treatment) was performed following the analytical procedure of the National Renewable Energy Laboratory (LAP-009) (http://www.nrel.gov/biomass/analytical_procedures. html). To release total sugar in the medium, 100 mg CWR was incubated at 30 °C for 1 h in 1.5 ml 72 % (w/v) H 2 SO 4 and later autoclaved at 121 °C for 30 min after adding 42.5 ml deionized water. The material was centrifuged at 4000 rpm for 30 min and the supernatant was collected in a new tube followed by diluting the entire extract to 50 ml with deionized water. For the saccharifications without pretreatment, 100 mg CWR was directly incubated in 10 ml digestion solution (8.8 ml 0.1 M, pH = 4.8 sodium citrate buffer + 0.2 ml 2 % NaN 3 + 1.0 ml Enzyme Cocktail stock) at 50 °C for 72 h. The enzyme cocktail stock was prepared using 1.5 ml cellulase stock from Trichoderma reesei (Sigma-Aldrich, St. Louis, MO, USA) and 1.5 ml Novozyme 188 stock Cellobiase from Aspergillus niger (Sigma-Aldrich) in 45.5 ml sodium citrate buffer (pH = 4.8). For pretreated samples, 100 mg CWRs were pretreated with 2.5 ml of 1.5 % (w/v) H 2 SO 4 and these samples were autoclaved at 121 °C for 40 min. This acid-pretreated material was washed three times with 10 ml water to obtain neutral pH and after each washing, the material was centrifuged at 4000 rpm for 20 min to retain any CWRs from the supernatant. The remaining biomass was lyophilized for 24 h and treated with enzyme cocktail as described earlier. Monomeric sugars (glucose and xylose) released in the samples were analyzed spectrophotometrically using the phenol-sulfuric acid assay method [87]. The concentration in the original sample was calculated with a standard curve based on known d-glucose concentrations (Sigma-Aldrich).
Saccharification efficiency was determined as the ratio of sugars released by enzymatic hydrolysis to the amount of sugars present in the cell wall material before enzymatic hydrolysis. For each biological replicate, mature inflorescence stems of 20 individual plants were pooled and assayed. Five biological replicates were included for each assay.

High-throughput pretreatment and enzymatic hydrolysis (HTPH) and downscaled compositional analysis
The HTPH system at University of California, Riverside [88][89][90] was employed to determine whether the mutants exhibited altered sugar release compared to the WT control during pretreatment and enzymatic hydrolysis. Glucose and xylose release from hydrothermal pretreatment and co-hydrolysis were measured. First, hydrothermal pretreatment was conducted at 180 °C for 11.1 min. Then, the pretreated slurry was co-hydrolyzed 50 °C in 50 mM citrate buffer (pH 4.8) for 72 h. Accellerase ® 1500 cellulase (112.5 mg) and XY xylanase (37.5 mg) from Genencor were used to release glucan and xylan in the raw biomass. Glucans and xylans were determined using a downscaled compositional analysis method [88] that is nearly identical to conventional wet chemistry procedures [90], but uses 100 times less biomass. The entire process was performed in 1.5 ml high recovery glass vials (Agilent, Santa Clara, CA, USA) with 3 mg dry biomass, loaded into each vial by a Core Module Robotics Platform (Symyx Technologies, Sunnyvale, CA). A set of glucose and xylose sugar recovery standards (SRS) were run in parallel for correction of sugar degradation. Sugars were analyzed by Waters Alliance 2695 HPLC (Milford, MA, USA) equipped with an Aminex HPX-87H column (Bio-Rad, Hercules, CA, USA) and a refractive index detector.