Xylan is the principal hemicellulose in the secondary cell walls of eudicots and in the primary and secondary cell walls of grasses and cereals. The biosynthesis of this important cell wall component has yet to be fully determined although a number of proteins have been shown to be required for xylan synthesis. To discover new genes involved in xylan biosynthesis we explored the psyllium (Plantago ovata Forsk) seed mucilaginous layer through EST profiling. This tissue synthesizes large amounts of a complex heteroxylan over a short period of time. By comparing abundant transcripts in this tissue with abundant transcripts specifically present during secondary cell wall formation in Arabidopsis thaliana, where glucuronoxylan biosynthesis is pronounced, we identified two Arabidopsis genes likely involved in xylan biosynthesis. These genes encode proteins containing a Domain of Unknown Function (DUF) 579 and were designated IRREGULAR XYLEM (IRX) 15 and IRX15-LIKE (IRX15-L). We obtained Arabidopsis T-DNA knockout lines for the two genes and analyzed their lower stems for changes in neutral monosaccharide composition. No changes were observed in each of these mutants, although the irx15 irx15-L double mutant displayed a moderate reduction in stem xylose. Further characterization of the irx15 irx15-L mutant revealed irregular secondary cell wall margins in fiber cells and a lower xylan degree of polymerization. Through these studies we conclude that IRX15 and IRX15-L function in a redundant manner and are involved in xylan biosynthesis.
In higher plants xylan is found as a group of polymers with a β-(1–4)-linked xylosyl (Xyl) backbone with various side chain modifications and has been grouped into homoxylan, glucuronoxylan, (arabino)glucuronoxylan, (glucurono)arabinoxylan, arabinoxylan, and complex heteroxylans. In woody species such as poplar and in many dicots such as Arabidopsis the xylan in the secondary wall is a glucuronoxylan, the side chain modifications being glucuronic acid (GlcA) or 4-O-methyl glucuronic acid (MeGlcA) at the Xyl O-2 position, and the GlcA or MeGlcA ([Me]GlcA) to backbone Xyl ratio being approximately 1:10. The xylan backbone is additionally acetylated at some O-2 or O-3 positions (Ebringerova and Heinze, 2000; Ebringerova et al., 2005). In some species, including Arabidopsis thaliana, a unique oligosaccharide, an 4-β-d-Xyl-(1–4)-β-d-Xyl-(1–3)-α-l-Rha-(1–2)-α-d-GalA-(1–4)-d-Xyl, is found terminating the reducing end of the glucuronoxylan backbone (Peña et al., 2007; York and O’Neill, 2008). Complex heteroxylans are structurally very diverse and can be found in cereals and certain seeds of dicots where the xylan backbone is highly branched with Xyl-, Ara-, Glc-, and Gal-containing monosaccharide, disaccharide, and trisaccharide side chains in substitution patterns highly dependent on the source (Ebringerova and Heinze, 2000; Ebringerova et al., 2005). Currently, we do not have a complete understanding of the enzymes involved in the biosynthesis of xylans.
Considerable effort has been made to delineate the glycosyltransferases involved in glucuronoxylan biosynthesis in Arabidopsis. A successful strategy has been to screen mutagenized Arabidopsis plants for altered xylem anatomy or reduction in stem mechanical strength (Turner and Somerville, 1997; Zhong et al., 2005). A complementary approach relies on the observation that many proteins shown to be involved in xylan biosynthesis have an expression pattern that is correlated with secondary cell wall formation. This makes co-expression analysis a powerful tool to gain additional insights into xylan biosynthesis and has been performed extensively in Arabidopsis utilizing publicly available microarray data from many different tissues (Brown et al., 2005; Persson et al., 2005; Ko et al., 2006) and in poplar utilizing microarray data from five different stages of developing wood (Aspeborg et al., 2005; Ubeda-Tomas et al., 2007). Additionally, differential expression of orthologous genes between cereals and eudicots has been used to identify genes involved in xylan biosynthesis (Mitchell et al., 2007).
The above studies have lead to the identification of a number of Arabidopsis mutants with impaired glucuronoxylan biosynthesis. One class of such mutants exhibits shorter xylan chains in the wall and substantial reductions in xylan xylosyltransferase activity in isolated microsomal fractions. These mutants are in genes that encode glycosyltransferases from two different families as defined by the CAZy database (Cantarel et al., 2009). IREGULAR XYLEM (IRX) 9, IRX9-Like (IRX9-L), IRX14 and IRX14-L are members of glycosyltransferase family 43 while IRX10 and IRX10-L are members of glycosyltransferase family 47. The double knockout mutants irx9 irx9-L, irx14 irx14-L and irx10 irx10-L display very similar phenotypes having strong reductions in plant growth and little or no detectable glucuronoxylan in stem inflorescences. It has further been shown that IRX9 and IRX9-L are not redundant with IRX14 and IRX14-L (Brown et al., 2009; Wu et al., 2009, 2010; Lee et al., 2010). Experiments with wheat microsomes show that IRX14 and IRX10 co-immunoprecipitate implying that they form a protein complex (Zeng et al., 2010). These phenotypes make these genes strong candidates for components of the xylan synthase (Brown et al., 2007, 2009; Lee et al., 2007a; Peña et al., 2007; Wu et al., 2009).
Apart from the backbone forming activities, other putative glycosyltransferases have been shown to be involved in glucuronoxylan biosynthesis. Investigations of IRX7/FRAGILE FIBER (FRA) 8, FRA8 HOMOLOG (F8H), IRX8/GALACTURONOSYLTRANSFERASE (GAUT) 12 and PARVUS indicate roles for these in synthesizing the reducing end structure (Brown et al., 2007; Lee et al., 2007b, 2009; Peña et al., 2007; Persson et al., 2007) and recent studies of GLUCURONIC ACID SUBSTITUTION OF XYLAN (GUX) 1 and GUX2 from CAZy glycosyltransferases family 8 establish these proteins as putative xylan glucuronyltransferases (Mortimer et al., 2010), while and members from CAZy glycosyltransferases family 61 have been proposed to catalyze the addition of side chain modifications to the xylan backbone (Mitchell et al., 2007).
The xylan synthase composition remains elusive because xylan synthase activity has not been established in vitro for any of the above proteins. The difficulty in obtaining a purified enzyme or a heterologously expressed protein with xylan synthase activity contrasts with the robust in vivo xylan synthase activity that can be measured in solubilized microsomes from various plants, suggesting that the xylan synthase is likely a protein complex (Brown et al., 2007, 2009; Peña et al., 2007; York and O’Neill, 2008; Wu et al., 2009). Identifying all genes that affect xylan accumulation in arabidopsis may be a necessary first step in assembling the components needed to demonstrate xylan synthase activity in vitro.
The expression correlation studies described above are useful but are limited by the fact that the correlations are to secondary cell wall formation and not to xylan biosynthesis. Gene expression studies in a tissue that primarily synthesizes xylan would likely yield more specific results. Many different plants form specialized tissues which accumulate one polysaccharide to high levels. Examples include galactomannan, which is produced in large quantities in the fenugreek endosperm as a storage polymer (Edwards et al., 1999; Dhugga et al., 2004), and xyloglucan, which is produced as a storage polymer in the cotyledons of the nasturtium seed. In both systems expression profiling has been used to elucidate genes involved in making these polymers. These studies led to the identification of the mannan synthase (Dhugga et al., 2004) and a strong candidate for the xyloglucan glucan synthase (Cocuron et al., 2007), respectively.
A previously unexplored system for expression profiling is the psyllium (Plantago ovata Forsk) seed mucilaginous layer. This tissue is a major constituent of the more commonly known psyllium seed husk product that forms a highly viscous gel upon water absorption and for this reason is used as a dietary supplement (Yu et al., 2009). The active component of the seed husk is a complex heteroxylan originating from the mucilaginous layer. It constitutes more than 60% of the dry weight of the seed husk and has an apparent structure with a β-(1–4)-Xyl backbone highly substituted with single Xyl at the O-2 positions and side chains of Ara-(1–3)-Xyl-(1–3)-Ara at the O-3 positions (Edwards et al., 2003; Fischer et al., 2004).
Here, we describe the use of transcriptional profiling of the psyllium mucilaginous layer to discover a new category of genes involved in xylan biosynthesis that affect xylan accumulation and the apparent xylan degree of polymerization (DP) in Arabidopsis.
Transcriptional profiling of the mucilaginous layer of the psyllium seed
To identify genes highly expressed during xylan biosynthesis we isolated mucilaginous layers from psyllium seeds at 10–16 days post anthesis (DPA) and determined the non-cellulosic neutral monosaccharide composition (Figure S1a–e). Based on these results and the feasibility of extracting enough high quality RNA from the developing layers we chose 6, 8, 10 and 12 DPA for EST analysis because xylan biosynthesis seemed pronounced at these stages. Using 1 μg total RNA and the SMART cDNA synthesis protocol we synthesized 15 μg of double-stranded cDNA from each stage. This cDNA was sequenced using a Roche GS-FLX resulting in a total of 896 726 ESTs with an average length of 211 bp. The ESTs were clustered using the cap3 software (Huang and Madan, 1999) and annotated by comparing each consensus sequence against the TAIR8 database of Arabidopsis protein sequences using the BLASTX program.
From the list of selected transcripts presented in Table 1 it is clear that genes involved in the biosynthesis of UDP-Ara, UDP-Xyl, and xylan are highly expressed. Homologs of the Arabidopsis RGP1, UXS6, IRX10, UXS2, and UXE1 are present among the 300 most abundant transcripts with UXS6 and IRX10 having the highest expression levels at 6002 and 6032 ppm. These five genes exhibit remarkably different expression profiles across the four stages as both strong increases (IRX10) as well as strong decreases (USX6) are seen during development. Transcripts with similar abundance represent genes involved in major cellular functions such as ubiquitination (UBQ1) and cytoskeleton formation (TUA6) (Table 1). For comparison, the guar mannan synthase was found in the seed endosperm at 2000 ppm (Dhugga et al., 2004) and in nasturtium the xyloglucan synthase was found in the developing seed cotyledons at 250 ppm (Cocuron et al., 2007). These observations indicated the prominence of xylan biosynthesis in this tissue and suggest that genes involved in xylan biosynthesis will be among those most abundantly expressed. Homologs of genes previously associated with xylan biosynthesis such as PARVUS, FRA8, IRX8, IRX9, and IRX14 were found with unexpectedly low abundance or not detected (Table 1). To investigate this further we subjected isolated mucilaginous layers from 10 to 12 DPA to 2D 13C–1H correlation (HSQC) NMR spectroscopy to test for the presence of the reducing end structure found in glucuronoxylans from some plants including Arabidopsis. This analysis revealed no detectable signals for the unique β-d-Xyl-(1–3), α-l-Rha-(1–2), and α-d-GalA-(1–4) residues (Figure S1f) that are readily identified by this method in Arabidopsis (Figure 6). The absence of detectable amounts of the xylan reducing end structure in psyllium mucilaginous layers may explain the low levels of expression of homologues of PARVUS, FRA8 and IRX8 in this tissue and support previous evidence indicating that these genes are involved in the biosynthesis of this structure (Brown et al., 2007; Lee et al., 2007b, 2009; Peña et al., 2007; Persson et al., 2007).
Table 1. EST frequencies of selected genes from psyllium mucilaginous layers
aAccording to abundance.
Reversibly glycosylated polypeptide 1 (RGP1)
Ubiquitin extension protein 1 (UBQ1)
UDP-glucuronic acid decarboxylase 6 (UXS6)
Irregular xylem 10 (IRX10)
Tubulin alpha-6 chain (TUA6)
UDP-glucose 6-dehydrogenase (putative)
UDP-glucuronic acid decarboxylase 2 (UXS2)
Domain of unknown function 579 (DUF579)
UDP-D-xylose 4-epimerase 1 (UXE1)
Cellulose synthase 1 (CESA1)
Cellulose synthase-like A2 (CSLA2)
Cellulose synthase 8 (CESA8)
Cellulose synthase 3 (CESA3)
Cellulose synthase 2 (CESA2)
Cellulose synthase 9 (CESA9)
Fragile fiber 8 (FRA8)
Cellulose synthase-like C5 (CSLC5)
Irregular xylem 14 (IRX14)
Irregular xylem 14-like (IRX14-L)
Irregular xylem 9 (IRX9)
Irregular xylem 8 (IRX8)
Other cell wall biosynthetic genes identified were CESA1, CESA8, CSLA2, CESA3, and CSLC which were found with a ranking of 360 or less (Table 1). This may reflect a low level of cell wall biosynthesis in the mucilaginous layers or, more likely, these expression levels represent contamination during tissue dissection from the developing, mannan-storing endosperm enclosed by the mucilaginous layer (Figure S1c).
As a way to discover new proteins involved in xylan biosynthesis we examined the psyllium EST data for sequences that were highly abundant in psyllium and where the closest homologs in Arabidopsis are primarily expressed in tissues making secondary cell wall. One such transcript encoded a homolog of the Arabidopsis gene At3g50220 (58% amino acid sequence identity). This transcript was highly abundant in psyllium with 606 ESTs, while, in Arabidopsis, At3g50220 has a specific secondary cell wall expression (Ko et al., 2006). Additionally, At3g50220 shows a highly significant co-expression with IRX10 (r =0.92; BAR expression angler, http://bar.utoronto.ca/ntools/cgi-bin/ntools_expression_angler.cgi; Toufighi et al., 2005). These results suggest that At3g50220 and its psyllium homolog are involved in xylan biosynthesis.
AT3G50220 has a single Domain of Unknown Function (DUF) 579 domain
The majority of the At3g50220 protein consists of a single Domain of Unknown Function (DUF) 579. This protein domain consists of approximately 180 amino acids and is broadly represented within Eukaryotes, identified in 75 different species, while a single bacterial sequence has been identified. In Arabidopsis, 11 proteins have been identified, each with a single DUF579 domain (the Pfam database: http://pfam.janelia.org/family/PF04669; Finn et al., 2010). Of these, 10 proteins, including At3g50220, form a family of structurally similar members each having a similar size (282–329 amino acids) (Figure S2) and having a single predicted N-terminal transmembrane domain (Figure S3) (TMHMM Server v. 2.0; http://www.cbs.dtu.dk/services/TMHMM/; Krogh et al., 2001). AT3G50220 has high sequence similarity to AT5G67210 (92% identity) and a very similar expression pattern (Figure S4; Schmid et al., 2005). The remaining eight members of the family share 37–52% sequence identity with AT3G50220. Based on the data reported below and the results of a similar study by Brown et al. (2011; this issue), the proteins AT3G50220 and AT5G67210 are designated IRREGULAR XYLEM (IRX) 15 and IRX15-LIKE (IRX15-L), respectively.
Disruption of both IRX15 and IRX15-L causes a decrease in cell wall xylose in the stem
To investigate the role of IRX15 and IRX15-L in xylan biosynthesis we identified T-DNA insertional alleles disrupting each of the genes (GK_735E12 for irx15; FLAG_532A08 for irx15-L). RT-PCR showed no detectable message for IRX15 and IRX15-L in irx15 and irx15-L, respectively (Figure S5a,b). These two lines were subsequently analyzed for changes in cell wall non-cellulosic neutral monosaccharide composition in the lower stem and siliques where the genes are most highly expressed. In neither irx15 nor irx15-L were any differences detected in the neutral monosaccharide composition nor did they show any altered growth phenotypes compared with Col-0 or Ws ecotypes (Figure S5c,d).
Based on the high amino acid identity between IRX15 and IRX15-L and their similar expression patterns we considered the two proteins likely to be redundant in function and so we crossed irx15 and irx15-L in order to determine the phenotype of the double mutant. To obtain a better comparison for the double mutant, we also crossed Col-0 and Ws and obtained hybrid lines, in this paper designated Col-0 × Ws.
Analysis of the non-cellulosic neutral monosaccharide composition of the lower stems of 7 weeks old plants showed a decrease in Xyl in the double mutant compared with Col-0, Ws, and Col-0 × Ws. The reduction in Xyl was 41 μg Xyl per mg alcohol-insoluble residue (AIR) compared with Col-0 × Ws, corresponding to 28% of the total amount of Xyl. The reduction in stem Xyl in irx15 irx15-L was verified by complementing the mutant using a genomic region of IRX15-L. As show in Figure 1, reintroduction of the IRX15-L wild-type allele into the irx15 irx15-L background (irx15 irx15-L IRX15-L) increased the Xyl to wild-type levels. The altered monosaccharide composition in irx15 irx15-L stems had no effect on crystalline cellulose content, lignin content, or stem height (Table S1).
In the lower stem, the most abundant cell wall carbohydrate apart from cellulose is xylan. The 28% reduction in Xyl in the monosaccharide composition therefore suggested a reduction in xylan, a result consistent with the psyllium IRX15 homologue being highly expressed in the xylan rich mucilaginous layers. Alternatively, a reduction in xyloglucan could be responsible for the loss of Xyl; however this seems less likely as no corresponding reductions in glucose, galactose and fucose, which represent the three other monosaccharides making up this polymer, were observed.
Characterization of irx15 irx15-L using plant cell wall carbohydrate-specific antibodies
To investigate if irx15 irx15-L showed alterations in xylan or xyloglucan we performed immunohistochemical analysis on cross sections of resin imbedded stems using xylan (LM10 and LM11; McCartney et al., 2005) and xyloglucan (LM15; Marcus et al., 2008) specific monoclonal antibodies. For each of the three antibodies xylem and interfascicular fibers labeled strongly while LM15 also labeled the cell walls in the pith, phloem, cambium and cortex. LM11 and LM15 revealed no discernable differences between Col-0 × Ws and irx15 irx15-L sections (Figure S6). The LM10 antibody, however, showed a stronger signal for irx15 irx15-L compared with Col-0 × Ws throughout both xylem and the interfascicular region (Figure 2a,b). In cell walls of the interfascicular fibers in both Col-0 × Ws and irx15 irx15-L the labeling was strongest on the inner side of the secondary walls when labeled with LM10 (Figure 2c,d) while the staining was found uniform throughout the secondary cell wall when labeled with LM11 (Figure 2e,f).
An alteration in the LM10 labeling pattern suggested that the xylan had undergone changes in irx15 irx15-L to allow an increase in staining intensity, rather than the expected decrease. As an alternative approach to the immunohistochemical investigations we proceeded by extracting pectin and hemicelluloses from stem AIR and subsequently quantifying the extracted polymers with the LM10, LM11 and LM15 antibodies using dot blots.
Aliquots of AIR from Col-0 × Ws and irx15 irx15-L were extracted with 4 m KOH, spotted onto nitrocellulose membranes, and probed with the respective antibodies. Using this assay, LM10 epitopes could be detected in spotting concentrations down to 3 mg ml−1 AIR in Col-0 × Ws samples while no signal was obtained for irx15 irx15-L (Figure 3a). LM11 epitopes were detected down to 0.4 mg/ml AIR spotting concentration in Col-0 × Ws while irx15 irx15-L displayed a 10-fold reduction in signal. When the dot blot assay was performed using LM15 the two genotypes gave equivalent results with signals detected down to 0.5 mg ml−1 AIR spotting concentration (Figure 3).
These results indicate no changes in the xyloglucan pool in irx15 irx15-L, whereas changes were detected when probing for xylan.
The ultrastructure of irx15 irx15-L secondary cell walls is abnormal
The LM10 histochemical labeling in irx15 irx15-L led us to consider the ultrastructure of the cell wall. Transmission electron microscopy of irx15 irx15-L revealed clear morphological changes in the secondary cell walls of the interfascicular fibers where the strongest LM10 signal in irx15 irx15-L was observed. The inner surface of the wall in irx15 irx15-L was highly irregular (Figure 4b,c) whereas smooth walls were observed in both Col-0 × Ws, and irx15 irx15-L IRX15-L (Figure 4a,d). No changes in ultrastructure or cell morphology were found for irx15 irx15-L in the secondary cell wall of xylem vessels. Other cell types in the stem tissue were normal as determined using toluidine blue staining and light microscopy (Figure S7).
The irregular inner surface of the cell walls of the interfascicular fibers in irx15 irx15-L resembles those reported for the same cell types in irx9 and irx14 (Peña et al., 2007; Lee et al., 2010). These mutants have reductions in stem Xyl of approximately 50% and have shorter xylan chains (Brown et al., 2007, 2009; Peña et al., 2007; Lee et al., 2010). The fact that irx15 irx15-L shows similar ultrastructural phenotypes to these well characterized xylan mutants is another indication that xylan formation may be altered in the irx15 irx15-L double mutant.
Sequential extraction of cell wall polysaccharides
Isolation and analysis of the different cell wall polysaccharides allows for an evaluation of possible structural changes in the xylan. We therefore performed serial extractions of stem AIR using CDTA, Na2CO3, 1 m KOH, and 4 m KOH, and subsequently determined the neutral monosaccharide composition of these fractions including the insoluble residue. The results are displayed in Figure 5.
The reduction in Xyl for irx15 irx15-L in the neutral monosaccharide composition of AIR (Figure 1) was found in the 4 m KOH fraction with as much as 50% less Xyl in irx15 irx15-L compared with Col-0 × Ws and irx15 irx15-L IRX15-L. Reductions are also observed in the residue. Conversely, the Na2CO3 fraction displayed a 50% increase in Xyl for irx15 irx15-L. The CDTA and 1 m KOH fractions were found to have similar amounts of Xyl among the three genotypes.
The 1 m KOH fraction, which contained approximately 30% of the solubilised material, mainly consists of xylan (Peña et al., 2007) as indicated by the 90% Xyl content of this fraction. It was therefore surprising to find irx15 irx15-L having wild-type levels of Xyl in this fraction as our monosaccharide analysis on AIR showed a 28% reduction in Xyl in irx15 irx15-L. However, the three xylan mutants irx9, irx10 irx10-L and irx14 also showed relatively higher amounts of Xyl in the 1 m KOH fraction, 75–80% the amount of wild-type Xyl, compared with AIR material where the Xyl content was 43, 35 and 60% the amount of wild-type Xyl, respectively (Brown et al., 2007, 2009). In another study of irx9, Peña et al. (2007) found 65% of the extractable xylan in the 1 m KOH fraction for wild-type plants while it was found to be 90% for irx9. The xylans in irx9, irx10 irx10-L and irx14 have been shown to be shorter while other structural features, such as the reducing end structure and [Me]GlcA to backbone Xyl-ratio, are similar to those in wild-type plants (Peña et al., 2007; Brown et al., 2009; Wu et al., 2009; Lee et al., 2010). The shorter xylan chains in these mutants may lead to the increased alkaline extractability. The similar alteration in xylan extractability suggests that the length of the xylan in irx15 irx15-L may be altered, as also supported by the higher degree of extraction into Na2CO3. A difference in irx15 irx15-L is that it has more Xyl than irx9, irx10 irx10-L and irx14 and so a correspondingly larger amount of xylan is released in the 1 m KOH fraction.
NMR spectroscopy on the 1 m KOH extracted fraction
The resemblance of the extraction profiles for irx9, irx10 irx10-L and irx14 to the extraction profile for irx15 irx15-L suggests that the irx15 irx15-L xylan is shorter. To determine xylan DP we performed 2D 13C–1H correlation (HSQC) NMR spectroscopy on the 1 m KOH fraction from Col-0 × Ws, irx15 irx15-L and irx15 irx15-L IRX15-L.
Partial (anomeric region) 13C–1H correlation HSQC spectra are shown in Figure 6. Plots were standardized to approximately the same total xylan levels. The three terminal units (X, R, and G) from the reducing end group are identified and all have similar integrals in each individual spectrum. Also identified are [Me]GlcA units, and the xylan units bearing their substitution at C-2. Reducing-end xylan units are not seen in these samples as NaBH4 treatment reduced them to xylitol end-units. Correlations from other C/H pairs in these xylitol units are seen in other areas of the spectra (Figure S8), where the xylan non-reducing ends (XylNRE) are also seen and can be quantified by their unique C-4 and C-5 correlations, but the anomeric correlation coincides with those from xylan internal units (Figure 6). Assignments were made from NaBH4-reduced xylotriose and xylobiose along with assignments in previously published papers on similarly isolated xylans (Peña et al., 2007; Ishii et al., 2008, 2010; Kim and Ralph, 2010).
To determine the average DP of the glucuronoxylan the ratio between total Xyl residues to XylNRE were determined (Table 2). The DP of irx15 irx15-L is reduced to 48% of the value found for the control sample, Col-0 × WS, while the [Me]GlcA substitution degree and the frequency of the reducing end structure per xylan molecule were found to be essentially unchanged. A shift towards less GlcA over total [Me]GlcA was also identified in irx15 irx15-L. In the complementation lines the average DP and GlcA over total [Me]GlcA were found to be similar to wild-type levels (Table 2).
Table 2. Characterization by NMR of glucuronoxylan from Arabidopsis stems extracted with 1 m KOH
Reducing end structure/XylNRE
The degree of polymerization (DP) of the xylan backbone of the glucuronoxylan was determined from the ratio of total xylose (Xyl) residues to Xyl at the non-reducing end (XylNRE). The degree of (4-O-methyl)glucuronic acid ([Me]GlcA) modification of the xylan backbone was inferred from the Xyl/[Me]GlcA ratio. The amount of glucuronic acid (GlcA) compared with total [Me]GlcA was obtained as the GlcA/[Me]GlcA ratio. The frequency of the xylan reducing end structure per xylan molecule was obtained by the reducing end structure/XylNRE ratio. Estimates were obtained by contour integration in the 2D 13C–1H correlation NMR spectra (Figures 6 and S8).
Col-0 × Ws
irx15 irx15-L IRX15-L
Over-expression of IRX15-L leads to altered xylan properties revealed by carbohydrate dot blot assay
The complementation lines irx15 irx15-L IRX15-L were also tested in the carbohydrate dot blot assay as presented for Col-0 × Ws and irx15 irx15-L in Figure 3. When probing with the LM11 xylan specific antibody some of the irx15 irx15-L IRX15-L plants tested had wild-type levels while the remaining plants showed moderate to 10 times stronger signals than wild-type plants. Plants were tested in both the T1 and the T2 generation and plants with the strongest signals in the T1 generation also gave rise to plants with strong signals in the T2 generation indicating the correlation of the genotype to this measurement. To investigate this phenomenon we analyzed stems of 5-week-old plants for transcript levels of IRX15-L and performed carbohydrate dot blots on the same samples. T2 plants from three irx15 irx15-L IRX15-L lines were analyzed displaying control level to very strong LM11 signals (Figure 7). As evident from the figure there is a strong correlation between IRX15-L expression and signal generated with the LM11 antibody in the dot blot assay (r =0.94). Several plants showed more than a 20-fold higher IRX15-L expression levels and 10 times higher LM11 signal. The expression level of several other genes involved in xylan and secondary cell wall formation, namely IRX7, IRX9, IRX10, IRX10-L, IRX14, CESA7, CESA8 and 4CL, were measured from all three lines and were found comparable with Col-0 × Ws and irx15 irx15-L levels (Table S2) arguing that the changes seen in the LM11 dot blots are caused by over-expression of IRX15-L.
Additionally, we prepared 1 m KOH extracted xylan from pools of T2 plants from the three recombinant irx15 irx15-L IRX15-L lines and analyzed this by 2D 13C–1H correlation NMR spectroscopy. We also tested these enriched xylan preparations in the dot blot assay using LM11. We found no correlation between any of the various aspects of glucuronoxylan structure revealed by the NMR analysis and the corresponding LM11 dot blot signals for the three recombinant xylan samples (Table S3). One difference of these plants compared with the control sample was that the xylan DP values of 53, 63 and 64 measured for these plants are higher compared with the Col-0 × Ws level of 46, a change that is consistent with the NMR analysis presented in Table 2. The irx15 irx15-L IRX15-L plants therefore consistently form longer xylan than the Col-0 × Ws plants by approximately 25% (P =0.03).
This characterization of the irx15 irx15-L IRX15-L plants shows that over-expression of IRX15-L leads to a xylan specific change in Arabidopsis stems.
Sub-cellular localization of IRX15
The subcellular localization of IRX15 was determined in vivo by Agrobacterium-mediated transient expression in Nicotiana benthamiana (Figure 8). A fusion construct with yellow fluorescent protein (YFP) fused to the C-terminal of IRX15 (IRX15–YFP) and driven by the CaMV 35S-promoter was generated, transiently co-expressed with the Golgi marker STtmd-CFP (Boevink et al., 1998), and analyzed by confocal laser-scanning microscopy. Leaves expressing IRX15–YFP and STtmd–CFP fusion proteins were observed between 24 and 36 h post infiltration to observe the localization when the minimal amount of accumulation of fluorescent protein needed for detection was present. Under these conditions YFP fluorescent signal accumulated in oval dots of approximately 1 μm in size as well as a more heterogeneous population of bodies significantly smaller (Figure 8a). Superimposing the CFP signal from the Golgi marker (Figure 8b) with the YFP signal showed colocalization of the YFP signal with the Golgi apparatus. The smaller heterogeneous population showed marginal to no CFP signal and displayed a very dynamic behavior within the cell. The smaller vesicles were frequently found proximal to the larger Golgi bodies. These smaller vehicles could be found as highly mobile individual dots within the cytosol. The dynamic behavior and size of these structures are consistent with their being secretory vesicles.
In this paper we demonstrate that transcriptional profiling of psyllium mucilaginous layers during the period of complex heteroxylan formation is a useful strategy to identify genes involved in xylan biosynthesis. The approach is based on the rationale that pairs of homologous genes between psyllium and Arabidopsis that are highly expressed during both the development of the psyllium mucilaginous layer and during secondary cell wall formation in Arabidopsis are strong candidates for xylan biosynthetic genes. Using this strategy we identified a psyllium transcript encoding a protein with a DUF579 domain. By knocking out the two most closely related genes in Arabidopsis, IRX15 and IRX15-L, we have demonstrated that the irx15 irx15-L double mutant represents a new class of xylan deficient mutants, adding to the previously described xylan mutants; parvus, irx7, f8h, irx8, irx8-H, irx9, irx9-L, irx10, irx10-L, irx14, irx14-L, gux1 and gux2.
The DUF579 domains of IRX15 and IRX15-L bear no structural resemblance to any of the known or putative glycosyltransferases identified in the CAZy database (Cantarel et al., 2009). Although it is possible that the DUF579 motif is a glycosyltransferase, it seems less likely in that IRX15 and IRX15-L are considerably smaller proteins (322 and 317 amino acids including predicted transmembranic domains, respectively) than found for glycosyltransferases (none of the 39 Arabidopsis glycosyltransferases in CAZy family 47 are <400 amino acids; http://www.cazy.org/GT47_eukaryota.html; Cantarel et al., 2009). The irx15 irx15-L mutant is therefore the first xylan-deficient mutant not annotated as a glycosyltransferase.
The irx15 irx15-L mutant phenotype is less severe than those of previously described genes. No discernable changes in growth phenotype, including inflorescence stem height, could be identified. Other xylan mutants have collapsed or irregular xylem and minor (irx9, irx10 and irx14) to severe (parvus, irx7 and irx8) stem height reduction (Brown et al., 2005, 2007; Peña et al., 2007; Wu et al., 2010). Also the amounts of Xyl in AIR and xylan enriched fractionations are higher than seen in the previously identified xylan mutants (Brown et al., 2007, 2009; Peña et al., 2007).
Our study leads us to conclude that IRX15 and IRX15-L proteins function in a redundant manner in xylan biosynthesis. Of the other eight structurally homologous proteins found in Arabidopsis three have expression patterns consistent with a role in secondary wall biosynthesis, namely At1g09610, At1g33800, and At4g09990 (Figure S4). These proteins could function in a redundant manner to IRX15 and IRX15-L in xylan formation, which could explain the moderate xylan deficient phenotype found in irx15 irx15-L.
Despite the lack of growth phenotypes and moderate reductions in Xyl, the combined irx15 irx15-L mutations have a pronounced affect on the secondary cell wall of the interfascicular fiber cells as observed by electron microscopy. The irregular cell wall phenotype seen in irx9 and irx14 (Peña et al., 2007; Lee et al., 2010), also evident in the irx9 i14h and i9h irx14 double mutant combinations (Lee et al., 2010), are observed in irx15 irx15-L. Relative to wild-type levels the xylan DP values for irx15 irx15-L were 50% (this study), 30% for irx9 (Peña et al., 2007) and 36% for irx14 (Lee et al., 2010). It is tempting to speculate that the irregular cell wall phenotype observed in irx15 irx15-L, irx9 and irx14 is a consequence of the shorter xylan chains found in these mutants.
Finding a stronger LM10 immunohistochemical labeling of the irx15 irx15-L secondary cell wall was unexpected due to the observed reduction in xylan. The irregular cell wall phenotype found by transmission electron microscopy could alter the accessibility of the LM10 epitopes to increase binding. Such an effect is not seen with the LM11 labeling of irx15 irx15-L. A Similar difference in immunohistochemical labeling of secondary cell walls between LM10 and LM11 has been shown by Persson et al. (2007). In that study, irx8 showed a stronger labeling with LM11 than with LM10, while irx9 displayed wild-type LM10 labeling but a strong reduction of LM11 labeling (Bauer et al., 2006; Persson et al., 2007). In an attempt to circumvent issues with epitope accessibility in the cell wall we performed 4 m KOH cell wall extracts and quantified the extracted polymers using dot blots and carbohydrate specific antibodies. Strong reductions in signal for both LM10 and LM11 were found for irx15 irx15-L compared with the control. The changes are larger than would be expected due to the loss of xylan in irx15 irx15-L as measured by Xyl. This change implies that some property of the xylan has changed. A change in DP could explain the lower signal as shorter xylan chains may bind less efficiently to the blot.
The carbohydrate dot blot assay using the LM11 antibody showed stronger signals in Arabidopsis plants with high levels of IRX15-L transcripts. NMR analysis of pools of T2 plants showed increases in xylan DP but these increases did not correlate well with the increased LM11 signal from the dot blot assay of the same samples. The xylan DP from the NMR data is inherently an average of the distribution of the various xylan DPs in the samples. It is therefore possible that there is an altered distribution of DPs in the over-expressing plants that results in a stronger signal in the carbohydrate dot blots. As suggested above, increases in xylan DP could result in strong increases in the affinity of glucuronoxylan for the nitrocellulose membrane and in this way result in the strong increases in LM11 signals observed. Alternatively, over-expression of IRX15-L could lead to structural changes in the xylan leading to more LM11 epitopes. The altered properties of the xylan reflected in the LM11 dot blot assay of both the irx15 irx15-L and the irx15 irx15-L IRX15-L plants, the changes in xylan DP observed by NMR, and the finding that an IRX15 homologue is highly expressed in psyllium mucilaginous layers argues for a direct role for IRX15 and IRX15-L in xylan biosynthesis in Arabidopsis.
Transient expression shows IRX15 localizes to Golgi and other cytosolic bodies, possibly the trans-Golgi network. This verifies IRX15 as a membrane protein that, at least by transient expression, co-localized with the location of xylan biosynthesis. Although IRX15 and IRX15-L seem to be involved in determining xylan DP, they do not appear to be glycosyltransferases, as discussed above. However, given their subcellular localization, the IRX15 and IRX15-L proteins could be involved in xylan backbone elongation by interacting with the xylan synthase.
Identifying genes that are abundant in psyllium mucilaginous layers and expressed predominately in tissues that are synthesizing secondary cell walls is likely to identify a core set of genes required to synthesize xylan. An unexpected finding from our analysis of the psyllium mucilaginous layer is that genes from glycosyltransferases family 43 identified to be involved in xylan biosynthesis in Arabidopsis and Poplar, such as IRX9 and IRX14, are either expressed at a very low level or are not detected. This contrasts with the very high expression of the psyllium IRX10. It is therefore unlikely that these GT43 genes are required for psyllium heteroxylan biosynthesis. This is surprising because the IRX9, IRX10 and IRX14 activities in Arabidopsis each appear to be essential components of the xylan-backbone synthase activity (Brown et al., 2007, 2009; Lee et al., 2007a, 2010; Wu et al., 2009, 2010). Our sequence profiling in psyllium, however, argues that IRX9 and IRX14 are not in all cases obligatory components of the xylan synthase.
In parallel with our work Brown et al. have preformed a series of similar and complementary experiments to investigate the function of IRX15, IRX15-L and the three other members of this protein family with a secondary cell wall expression pattern. This work is present in the accompanying paper (Brown et al., 2011) where very similar conclusions are reached about their function in xylan biosynthesis.
Psyllium plants (Indian, Plantago ovata, Sand Mountain Herbs, AL, USA) were grown in soil in environmental growth chambers with a 16-h photoperiod, 22°C, 60% relative humidity, 140–150 μmol photons m−2 s−1.
Arabidopsis plants were grown in soil at 16-h photoperiod, 20°C, 60% relative humidity, 100–120 μmol photons m−2 s−1. Inflorescence stems were harvested as indicated for the different experiments between 7 and 13 weeks. Only the lower half of the primary stem was subjected to analysis.
Total RNA extraction from psyllium mucilaginous layers
Psyllium mucilaginous layers were obtained from seeds by hand dissection to separate the mucilaginous layers from the endosperm (Figure S1a–d). Total RNA was isolated as described by Cocuron et al. (2007).
cDNA library construction and 454 FLX sequencing
First-strand cDNA synthesis was performed using SuperScript II Reverse Transcriptase (18064-022; Invitrogen, http://www.invitrogen.com/) in combination with the SMART IV oligo (634903; Clontech, http://www.clontech.com/) and a modified version of the CDS III/3′ primer (5′-ctagaggccgaggcggccgacatgttttgtttttttttcttttttttttvn-3′). The cDNA was subsequently amplified by PCR using the Advantage 2 Polymerase Mix (639201; Clontech), 5′ SMART PCR primer (634903; Clontech) and the modified CDS III/3′ primer. Amplified cDNA was treated with Proteinase K, digested with SfiI and size-fractionated on ChromaSpin-400 Columns (636076; Clontech) to obtain DNA larger than 100 bp. A more detailed protocol is described in Appendix S1.
High-throughput cDNA sequencing was done by the DOE Joint Genome Institute using the Roche GS-FLX sequencer (454 Life Sciences, http://454.com/).
454 sequencing and assembly
ESTs from the different libraries were clustered using the CAP3 software (Huang and Madan, 1999) and the consensus sequences were compared with the predicted proteins from Arabidopsis using the BlastX program (Altschul et al., 1997).
Genomic DNA was prepared by use of the RNeasy plant mini kit (Qiagen, http://www.qiagen.com/). Homozygosity was verified by PCR for GK_735E12 using forward 5′-ctctcggcacaagaactcatc-3′ and reverse 5′-tggattcccaaataaacacttc-3′ primers with insert specific primer (LBo8409) 5′-atattgaccatcatactcattgc-3′, and for FLAG_532A08 using forward 5′-tcaaacgacagctaccacatg-3′ and reverse 5′-cggacttgtggatcagctaag-3′ primers with insert specific primer (LB4) 5′-cgtgtgccaggtgcccacggaatag-3′.
Molecular complementation of irx15 irx15-L
The full-length gene of At5g67210, including 1931 bp upstream and 1151 bp downstream of the coding region, were cloned by PCR from genomic DNA with the primers 5′-cgtctaaagcttgcagagcatctagctgtgtc-3′ and 5′-gccataggcgcgcccaagagagaaaatgtg-3′. The resulting PCR product was inserted as a HindIII–AscI fragment in the binary vector pGWB511 (Nakagawa et al., 2007) and the sequence was verified by sequencing both strands.
The pGWB511-At5g67210 construct was transformed into Agrobacterium tumefaciens GV3101 and the resulting strain was used to transform flowering irx15 irx15-L plants using previously described procedures (Clough and Bent, 1998).
Non-cellulosic neutral monosaccharide analysis
Harvested plant materials were lyophilized, ground into a fine powder, and washed three times with 70% ethanol, three times with 1:1 methanol-chloroform, and two times with acetone to obtain alcohol insoluble residue (AIR). The AIR was subsequently de-starched with 1.8 μg amylase (A6380; Sigma-Aldrich, http://www.sigmaaldrich.com/) and 0.02 U pullanase (P2986; Sigma-Aldrich) per 10–40 mg AIR.
The non-cellulosic neutral monosaccharide composition of the wall matrix polysaccharides was obtained by treating de-starched AIR with trifluoroacetic acid and subsequent derivatization of the solubilized monosaccharides into their corresponding alditol acetates followed by quantification by GC-MS (Albersheim et al., 1967).
Transmission electron microscopy and immunohistochemical analysis
For immunohistochemical analysis samples were prepared as described by Freshour et al. (1996) using LR White Resin (14381; Electron Microscopy Sciences, http://www.emsdiasum.com/) as imbedding resin. Transverse sections of 3 μm were then prepared, fixed onto Vectabond-treated (SP-1800; Vectorlabs, http://www.vectorlabs.com/) microscope slides, blocked with Dulbecco’s phosphate-buffered saline (DPBS) 5% skim milk, labelled with LM10, LM11 or LM15 (1:20 in DPBS; PlantProbes, http://www.plantprobes.net/), washed with DPBS, incubated with FITC::anti-rat IgG (F-6258; Sigma), washed with DPBS, and finally slides were mounted with Citifluor AF1 (Citifluor Ltd, http://www.citifluor.co.uk/). Microscopy was performed using a laser confocal scanning microscope (FV1000D IM-IX81; Olympus, http://www.olympusamerica.com/).
Both transmission electron microscopy and immunohistochemical analysis were repeated two times on two sets of plants grown independently with three biological replicates in each experiment.
Quantitative dot blot analysis using carbohydrate specific antibodies
Aliquots of de-starched AIR were extracted with KOH with 1% (w/v) NaBH4 over night at room temperature. The supernatant was added acetic acid to reduce the pH and dilutions were done in 0.8 m KOH. Purified 1 m KOH stem carbohydrate preparations were dissolved directly in 0.8 m KOH. Spotting was performed onto nitrocellulose (162-0115; Bio-Rad, http://www.bio-rad.com/) using a Biomek 2000 robot (Beckman Coulter, http://www.beckmancoulter.com/). The carbohydrate dot blots were then blocked, incubated with primary antibody (LM10, LM11 or LM15), incubated with rabbit anti-rat antibody conjugated to horseradish peroxidase (P0450, Dako, http://www.dako.com/), and developed by chemiluminescence. A more detailed protocol is described in the Supporting Information.
Sequential extraction of stem cell wall material
Aliquots of 15 mg de-starched AIR from the lower half of primary stems from 7 weeks old plants were sequentially extracted with 50 mm CDTA pH 6.5, 50 mm NaCO3 10 mm NaBH4, 1 m KOH 10 mm NaBH4, and 4 m KOH 10 mm NaBH4. Supernatants and residue were recovered by centrifugation and NaCO3, KOH and residue fractions were adjusted to pH 5 with glacial acetic acid. All fractions were then dialyzed extensively in cellulose ester dialysis membranes (MWCO 100-500 D; Spectrum Labs) against water, and subsequently lyophilized.
2D 13C–1H HSQC NMR spectroscopy
The xylan extracts from the 1 m KOH 10 mm NaBH4 extractions, 3–4 mg, and psyllium mucilaginous layers in the form of ball-milled AIR (Retsch PM100, 12 ml stainless jar, five 10-mm balls, 600 rpm, 5 min), 30 mg, were dissolved in 4:1 DMSO-d6/pyridine-d5 (Kim and Ralph, 2010). NMR spectra were acquired on a Bruker Biospin Avance 700 MHz spectrometer (http://www.bruker-biospin.com/) fitted with a cryogenically cooled 5-mm TCI gradient probe with inverse geometry (proton coils closest to the sample). The central DMSO solvent peak was used as internal reference (δC 39.5, δH 2.49 ppm). The 13C–1H correlation experiment was an adiabatic HSQC experiment (Bruker standard pulse sequence ‘hsqcetgpsisp2.2’; phase-sensitive gradient-edited-2D HSQC using adiabatic pulses for inversion and refocusing) (Kupce and Freeman, 2007). HSQC experiments were carried out using the following parameters: acquired from 10 to 0 ppm in F2 (1H) with 2800 data points (acquisition time 200 msec), 200–0 ppm in F1 (13C) with 560 increments (F1 acquisition time 8 msec) of 64 scans with a 1 sec interscan delay; the d24 delay was set to 0.86 ms (1/8J, J = 145 Hz). Processing used typical matched Gaussian apodization in F2 and, squared cosine-bell and one level of linear prediction (32 coefficients) in F1. Volume integration of contours in HSQC plots used Bruker’s TopSpin 2.1 software.
The full-length coding region of IRX15 was cloned by PCR from genomic DNA with the primers 5′-cacgggggactctagaatgaagaacggatcaggg-3′ and 5′-catgaccgtcgacatagacaaagacgaaaccgaa-3′ and inserted as a XbaI–SalI fragment upstream of YFP in the binary vector pVKH18En6 (Batoko et al., 2000).
The pVKH18En6–IRX15–YFP fusion construct was transformed into A. tumefaciens GV3101 and co-infiltrated at an optical density (OD600) of 0.05 into leaves of 3-week-old N. benthamiana plants with a strain harboring the 35S:p19 construct for suppression of gene silencing (Voinnet et al., 2003) and a strain harboring the Golgi marker STtmd–CFP (Boevink et al., 1998) using previously described procedures (Voinnet et al., 2003). Infiltrated leaves were observed between 24 and 36 h after infiltration by laser confocal scanning microscopy (FV1000D IM-IX81; Olympus). Transient expression and microscopy was performed independently two times.
We thank Christa Pennacchio and Erika Linquist (DOE Joint Genome Institute) for high through put cDNA sequencing; Nick Thrower (Department of Biochemistry and Molecular Biology, Michigan State University, USA) for providing the bioinformatic expertise clustering of the cDNA libraries; and Chris Hawes (School of Life Sciences, Oxford Brookes University, UK) for the kind gift of the Agrobacterium strain harboring the STtmd–CFP construct. This work was funded by the DOE Great Lakes Bioenergy Research Center (DOE BER Office of Science DE-FC02-07ER64494).