Work carried out in Arabidopsis has shown that there are multiple enzymes involved in the synthesis of the glucuronoxylan backbone (IRX9, IRX10 and IRX14), substitution of the backbone (GUX1/GUX2) and synthesis of the reducing end oligosaccharide (IRX8, FRA8 and PARVUS) 30 31 40 . To determine whether there is conservation of the genes required for xylan biosynthesis between higher and lower plants, each of the Arabidopsis proteins proposed to function in the synthesis of glucuronoxylan were used to BLAST search the Physcomitrella genome. A single putative IRX10 homolog is present in Physcomitrella encoded by Pp1s7_455V6 (PpGT47A) (Table 1) which shows a high degree of conservation with the Arabidopsis and Populus trichocarpa IRX10 proteins (Figure 1). The N-terminal regions of the Arabidopsis and Populus IRX10 proteins are the most divergent domains and contain predicted signal peptides in AtIRX10, AtIRX10-L, PtGT47A-1 and PtGT47A-2 whereas the N-terminal region of PpGT47A has a predicted transmembrane domain in common with two of the other Populus homologs PtGT47D-1 and PtGT47D-4 43 44 (Table 2). Two putative homologs were found for AtIRX9/IRX9-L (Pp1s52_108V6.1 and Pp1s1_540V6.1), whereas three putative homologs were each identified for AtIRX14/IRX14-L (Pp1s248_13V6.1, Pp1s78_128V6.1 and Pp1s19_221V6.1), AtFRA8/F8H (Pp1s13_216V6.1, Pp1s217_58V6.2 and Pp1s315_20V6.1), and AtGUX1/GUX2 (Pp1s21_381V6.1, Pp1s351_36V6.1 and Pp1s223_30V6.1) (Figure 2). Interestingly, there were no closely related sequences found for AtIRX8. Although several sequences were identified that exhibited a 60% similarity with PARVUS, all of them exhibited a higher degree of similarity with other Arabidopsis proteins and so were not considered likely to be homologs of PARVUS (Table 1). The degree of similarity between the Arabidopsis and Physcomitrella homologs was between 25% and 48% except in the case of the PpGT47A which exhibited 76% similarity with AtIRX10 and AtIRX10-L. All genes are present in two or three copies encoding closely related proteins in both species, with the exception of PpGT47A which is unique (Figure 2). In all cases, duplicated or triplicated genes within a species encode proteins that cluster together (Figure 2). This suggests that gene duplications have occurred in both species after they split from each other. However, concerted evolution of related genes cannot be ruled out as a reason for close similarity, at least not in Physcomitrella, where it has been previously observed 45 . The high degree of sequence similarity exhibited between the Arabidopsis IRX10 and the Physcomitrella GT47A proteins further suggests that they may exhibit functional conservation and thus may give clues as to the evolutionary origin of the GX biosynthesis machinery in higher plants.

Gene models were considered putative homologs if the Arabidopsis protein that was used for the Blast was the Arabidopsis protein which exhibited most similarity. *Although three Physcomitrella gene models were identified with the corresponding protein exhibiting between 49.57 and 51.57% similarity to PARVUS, the sequences all exhibited stronger homology to at least 10 other Arabidopsis proteins, and were therefore not considered as obvious homologs. (ClustalW).

IRX10 related protein sequences from Arabidopsis, Populus and Physcomitrella were analysed using the Phobius programme (http://www.phobius.sbc.su.se) for the presence of N-terminal signal sequences and the locations of the hydrophobic predicted transmembrane core (H-region) flanked by the N- and C- regions.

To investigate whether expression of PpGT47A is associated with specific tissues or developmental stages of growth, the uidA reporter gene was fused to the 3^′ end of the PpGT47A gene. The construct (Figure 3A) was introduced into the Physcomitrella genome at the endogenous PpGT47A site via homologous recombination. The pattern of pGT47A expression in Physcomitrella was analysed by histochemical staining of 4 independent stable pGT47A-GUS lines, each of which showed a similar staining pattern. Expression was localised predominantly to tissues which were newly formed or undergoing expansion. Strong expression was observed in the apical region of the adult gametophore (Figure 3B), new branches forming on the side of the mature gametophore (Figure 3E) and in the tissue immediately basal to the immature sporophyte (Figure 3D). Staining was also observed in the antheridia (Figure 3C). A peak of expression was seen at the tips of the protonema side branches (Figure 3F) and the lateral buds of the chloronema (Figure 3G).

To establish whether the function of the PpGT47A enzyme is conserved between Arabidopsis and Physcomitrella, the PpGT47A gene under control of the 35S cauliflower mosaic virus (CaMV) constitutive promoter was expressed in the Arabidopsis irx10 irx10-L double mutant background. The PpGT47A gene partially rescued the irx10 irx10-L phenotype with plants forming a short inflorescence stem without the need for a protective growing cover that is normally required by the double mutant 35 (Figure 4A). Rosette diameters of complemented plants were increased by 2- to 3-fold and the inflorescence stems were between 3- and 28-fold taller compared to the irx10 irx10-L plants grown under a plastic cover (Figure 4B). Safranin staining was performed on basal stem sections of the Arabidopsis wild-type, irx10 irx10-L double mutant and complemented plants to identify regions of lignification. In Arabidopsis wild-type plants, the cell walls of the xylem vessels and interfascicular fibers were heavily stained around the whole cell (Figure 4C). In contrast, the irx10 irx10-L plants exhibited only a weak signal at the corners of the cell junctions indicating initiation of lignification, with virtually no evidence of secondary cell wall deposition in any other regions of the cell walls (Figure 4D). Irregular and variable safranin staining could be detected throughout the xylem and interfasicular fiber cell walls of stem sections from the complemented plants although the staining was less intense than in the wild-type. Despite the apparent increase in secondary cell wall production, the cell walls of irx10 irx10-L plants expressing PpGT47A were thinner than wild-type cell walls and displayed a collapsed xylem vessel phenotype (Figure 4C,D).

Monosaccharide analysis was performed on cell wall material isolated from stems of the complemented plants to determine the biochemical basis for the partial complementation phenotype (Figure 5). There were no significant differences in the levels of the monosaccharides measured between the complemented lines and the irx10 irx10-L double mutant. In particular, it is noteworthy that the level of xylose remained low in the PpGT47A expressing lines in contrast to the results obtained when the Arabidopsis irx10 irx10-L double mutant was complemented with the Arabidopsis IRX10 gene 35 .

Arabidopsis irx10 and irx10-L knockout mutants have proven to be informative in elucidating the function of the encoded proteins. In order to further investigate the function of PpGT47A, Physcomitrella gt47A knockout mutants were generated. A construct was made in which the nptII coding sequence was inserted into the PpGT47A gene deleting part of exons 2 and 4 and the whole of exon 3. The construct was introduced into the Physcomitrella genome via homologous recombination resulting in the deletion of the central region of the endogenous PpGT47A gene to create a likely null knockout mutant. Knockouts were analysed by PCR using primers spanning the recombination site to identify lines in which the endogenous PpGT47A gene had been disrupted. The gametophores of the knockout lines were visually similar to wild-type with no obvious morphological changes when grown on BCD media (Additional file 1: Figure S1). Further experiments were carried out to test whether varying the composition of the growth media by addition of different sugars or hormones under different light conditions (see materials and methods for details) revealed any phenotypic changes in the knockout lines but none were observed (data not shown). Monosaccharide analysis of gametophore cell walls from the knockout lines (Figure 6) did not show any significant differences compared to the wild-type. Of particular interest is the observation that there was no difference in xylose content between knockout and wild-type gametophores in contrast to the results obtained for the Arabidopsis irx10 and irx10-L mutants 35 .

The finding that the Physcomitrella GT47A protein only partially rescues the Atirx10 irx10-L double mutant demonstrates that although the proteins are highly conserved at the sequence level, there are important differences which influence function. There are a number of possible reasons for these apparent functional differences. Firstly, the enzymatic activity of PpGT47A may be different to that of the Arabidopsis IRX10 or IRX10-L proteins. Despite the high degree of sequence conservation between the Arabidopsis and Physcomitrella IRX10 proteins, subtle amino acid differences could give an altered enzymatic activity but in the absence of a functional enzymatic assay this cannot currently be tested. Alternatively, PpGT47A could have the same enzymatic activity as the Arabidopsis IRX10 but the interaction between IRX9 and/or IRX14 to form a putative functional complex may not be optimal. A third possibility is that the targeting or subcellular localisation of PpGT47A is not correct when expressed in Arabidopsis. The highly divergent N-terminal domain that in the Arabidopsis IRX10 protein encodes a predicted signal peptide, in Physcomitrella instead forms a predicted transmembrane domain (Table 2). Thus, mislocalisation and/or a different conformation of the PpGT47A protein compared to AtIRX10 may combine to produce only a small amount of a functional GX synthesis complex or alternatively a complex with low activity in the partially complemented Arabidopsis plants.

The Arabidopsis IRX10 and Populus PtGT47A-1 genes are expressed within the vascular tissue made up mainly of cells rich in secondary cell walls. Bryophyte water conducting tissues in contrast do not form thickened cell walls and so it is unsurprising that the expression pattern of PpGT47A differs from that of the Arabidopsis homologs. The ability of PpGT47A to partially complement the Arabidopsis irx10 irx10-L double mutant plants shows that there is a degree of functional overlap between the Arabidopsis and Physcomitrella enzymes but that either activity, conformation or targeting of the enzymes prevent full rescue of the wild-type phenotype. Interestingly, the degree of similarity is higher between PpGT47A and AtIRX10-L than for AtIRX10 indicating that IRX10-L is closer to the ancestral form of the protein. The expression of PpGT47A in Physcomitrella tissues undergoing primary cell wall formation contrasts with that of the Arabidopsis IRX10 gene which is associated with tissues forming secondary cell walls. In vascular plants, the IRX10 family appears to have undergone a sub- or neo-functionalisation resulting in at least a subset of the genes becoming specialised for secondary cell wall formation.

Analysis of the monosaccharide sugar composition of cell walls isolated from the irx10 irx10-L lines partially complemented with the PpGT47A gene does not show any significant changes in comparison to the irx10 irx10-L double mutant indicating that either changes are in a minor component of the cell wall and thus below the detection limit or that the PpGT47A function is not related to the cell wall composition. Although current evidence strongly indicates that IRX10 and related proteins from other plants function in glucuronoxylan biosynthesis, there have been previous reports suggesting involvement of the tobacco NpGUT1, another member of the IRX10 family of proteins, in rhamnogalacturonan II (RGII) biosynthesis, adding GlcA to one of the side chains of the RGII polysaccharide 48 . The decrease in the level of GlcA reported for the Npgut1 tobacco mutant is greater than can be explained by a reduction in RGII levels alone arguing against NpGUT1 having a sole function in RGII synthesis. Interestingly, the Npgut1 mutant also exhibits a decrease in xylose content, implying that xylan synthesis could also be affected. Furthermore, the tobacco NpGUT1 protein is able to perform the same enzymatic function as the Arabidopsis IRX10, restoring xylan levels to those of the wild-type when fused with the N-terminal 71 amino acid domain of the AtIRX10 protein and expressed in the irx10 irx10-L background 35 . Although this result supports an involvement of NpGUT1 in glucuronoxylan biosynthesis, the possibility that members of the IRX10 family, including PpGT47A, could function in the synthesis of RGII or an RGII-like polymer cannot currently be discounted.

A major advantage of using Physcomitrella is the possibility to make clean loss of function mutants by homologous recombination. This approach was used to create a knockout mutant for the PpGT47A gene. However, the knockout mutant had no obvious phenotypes under the conditions tested, nor did monosaccharide analysis of the gametophore cell wall reveal any significant changes in the mutant (Figure 6). We conclude from this that loss of PpGT47A function either does not affect the cell wall composition, or alternatively that any such effects are too subtle to be detected in our analysis. It should be noted that RGII and xylan are both known to be minor components of the bryophyte cell wall 9 10 , and alterations in the composition of either or both polymers might therefore have escaped detection in our analysis of total cell wall material.

The absence of an obvious knockout phenotype could be due to several reasons. It is possible that the PpGT47A gene is functionally duplicated in Physcomitrella. Although this is unlikely since there are no close homologues of PpGT47A in the Physcomitrella genome 49 , the possibility exists that another more distantly related GT can provide the same function as PpGT47A. Alternatively, it is possible that the modifications catalyzed by PpGT47A are important in Physcomitrella only under certain conditions, such as abiotic stress, but not under standard laboratory conditions. In any case, the fact that PpGT47A is strongly conserved and shows no evidence of being a pseudogene suggests that it encodes a functional protein, a conclusion that is reinforced by its ability to partially complement the Arabidopsis irx10 irx10-L double mutant.

The Physcomitrella patens subsp patens v1.1 database and the Phytozome Physcomitrella database (http://www.phytozome.net/physcomitrella.php) were used to search for homologs using the Arabidopsis IRX10, IRX10-L, IRX9, IRX9-L, IRX14, IRX14-L, IRX8, PARVUS, GUX1 and GUX2 proteins. Sequences were aligned using Clustal-W software. Phylogenetic trees (Figure 2) were computed using the neighbour-joining method 46 as implemented in the Clustal-X software. The Phobius signal peptide and transmembrane prediction software was used for topological analysis (phobius.sbc.su.se) 44 47 .

Arabidopsis was grown under long day conditions (16h light/8h dark) at a maximum irradiation of 150 μmolm^-2s^-1, 60% humidity with 22°C day and 18°C night temperatures. Physcomitrella gametophores were grown on BCD media (0.001 M MgSO₄, 0.0189 mM KH₂PO₄, 0.01 M KNO₃, 0.045 mM FeSO₄, alternative trace element solution (TES) as described on http://moss.nibb.ac.jp/) and protonemata tissues were grown on MM media (BCD media with 5mM ammonium tartrate). Light, humidity, and temperature conditions employed were as for Arabidopsis.

PpGT47A was PCR amplified from Physcomitrella DNA using forward primer (F): 5^′-ggggacaagtttgtacaaaaaagcaggctgggagaattgggtgtttcg-3^′ and reverse primer (R): 5^′-ggggaccactttgtacaagaaagctgggtgtgttacaaatcattgcccc-3^′), cloned into pDONR207 and then transferred into the pEarleyGate 100 destination vector 50 using the Gateway cloning system. The PpGT47A overexpression construct was introduced into the Arabidopsis irx10 irx10-L(+/-) background 36 using the floral dip method 51 . Transformed lines were screened using BASTA (Hoechst Schering AgrEvo GmbH, Germany) selection and irx10 irx10-L double mutants identified by PCR 35 allowing lines containing the transgene in a homozygous irx10 irx10-L background to be selected in the T2 generation. Wild-type and transformed mutant seed were sown and treated as previously described 35 .

To generate the PpGT47A knock-out construct, the NptII gene was PCR amplified from the pMT164 vector, introducing an XmaI restriction site (F: 5^′-aattcccggggagtcaaag-3^′ and R: 5^′-atggatcgatgttaacatgc-3^′). PpGT47A was PCR amplified from Physcomitrella Gransden 2004 (F: 5^′-ggacatagaagcatgatgc-3^′ and R: 5^′-ggaatacaacacgattcc-3^′) and cloned into TOPO2.1 vector. The TOPO2.1 vector containing the PpGT47A gene was cut by XmaI and MfeI, the NptII cassette was cut with XmaI and EcoRI, and the two DNA fragments ligated using T4 DNA-ligase. The resulting construct was linearized by EcoRI before being transformed into Gransden 2004 wild-type protoplasts 52 using PEG mediated transformation as described previously 53 . Stable transformants were identified by selection on kanamycin-containing media for 2 weeks, followed by growth on selection free media for 2 weeks and then a further hygromycin selection for 2 weeks. Transformants were confirmed by PCRs spanning the recombination sites. Primers used for verification of the knockout construct were F1: tggtcaggagaatcatgc and F2: cggaagtaacagaatgagg, R1 ttgataactgtgggttacc and R2: caggtgacatgagactcg, and for the insert F: ttcgctcatgtgttgagc and R: aggcatcttcaacgatgg. All restriction enzymes used were FastDigest, Fermentas (St. Leon Rot, Germany).

The GUS reporter gene was fused to the PpGT47A gene in the Physcomitrella genome using the following strategy. Two fragments, one homologous to the DNA sequence upstream and one to the region downstream of the PpGT47A stop codon, were amplified from endogenous Physcomitrella Gransden 2004 DNA by PCR. The PCR for the upstream fragment introduced a BamHI and a XbaI site at the 5^′ end with the forward primer and the stop codon was disrupted with the reverse primer (F: 5^′-ggatccttctttctagtgacaatcgg-3^′ and R: 5^′-cgcggatccacaaatcattgccccaagg-3^′). The downstream fragment was amplified using F: 5^′-ctttcttggaatactcacc-3 and R: 5^′-ctggaacgcatctagacc-3^′ primers. The fragments were cloned separately into the Topo2.1 vector. The cloned downstream fragment was excised with NotI and XbaI and ligated into a modified Topo vector carrying the GUS gene 54 . The upstream PCR fragment was cut with BamHI and introduced into Topo vector carrying the downstream fragment and the GUS gene. All restriction enzymes were FastDigest, Fermentas.

The composition of the BCD media and the growth conditions in the light chamber were as previously described 45 . Clumps of subcultured protonema tissue were placed on BCD plates and grown for three weeks in continuous light at 25°C and then moved to short day conditions (8 hours light/16 hours dark at 15°C) and grown for three months. GUS staining was performed by incubating the moss tissue in X-gluc substrate solution as described by the Physcobase protocol (http://moss.nibb.ac.jp/). Stained tissue was analyzed with an Olympus SZX12 stereo microscope and images recorded using an Olympus XC30 camera.

Small clumps of subcultured protonema tissue (approximately 2 mm in diameter) were used to inoculate standard BCD plates and plates containing different additions. The additions were 0.15 M glucose, 0.15 M fructose, 0.15M sucrose, 0.15 M mannitol or 5 mM ammonium tartrate. The plates were placed vertically in an incubator under three different conditions, high light (30 μmol/m²s), low light (6 μmol/m²s) or complete darkness. In addition, morphology studies were carried out on Physcomitrella grown on BCD plates containing the following conditions and concentrations of hormones. High light: 7 μM BAP, 1 μM ABA, 0.5 μM IAA and 10 μM GA3. Low light: 0.5 μM BAP, 0.5 μM ABA, 0.5 μM IAA and 10 μM GA3. All hormones were from Sigma Aldrich (Stockholm, Sweden). Plates were observed for two months.

Basal stem regions from wild-type Arabidopsis plants measuring 30 mm in height or 6 weeks old, 9-week old mutant or complemented plants and 8-week old Physcomitrella gametophores grown on BCD media were used for monosaccharide analysis. Tissues were collected in 80% ethanol and stored at -80°C until being freeze dried (Modulyo, Edwards, West Sussex, UK). Dried material was ball milled in a beadmill (Retsch MM301, Haan, Germany) for 2×30s at 30 Hz. Alcohol insoluble residues (AIR) were obtained as previously described 39 . The AIR material was suspended in 0.1 M phosphate buffer, pH7 containing 0.01% sodium azide. Alpha-amylase (Roche, Indianapolis, USA) was added at a concentration of 1000U per 1g of cell wall material and the material was digested with gentle shaking for 24h at 37°C. The procedure was repeated once before the pellet was washed first with 0.1 M phosphate buffer pH 7, then with water and finally acetone. The material obtained was analysed using the TMS method 55 56 57 .

Basal stem segments were collected, fixed in FAA (5% Acetic acid, 50% ethanol, 5% formadehyde in dH₂O) and stored at 4°C until being sectioned using a vibratome (60 μm thickness) (Leica VT1000S, Germany), stained with 1:2 filtered safranin (1% in 50% ethanol): alcian blue (1% in H₂0, 1% formalin and 0.15% glacial acetic acid), rinsed in H₂O and mounted in 50% glycerol 58 .