A galactosyltransferase acting on arabinogalactan protein glycans is essential for embryo development in Arabidopsis


  • Naomi Geshi,

    1. Department of Plant Biology and Biotechnology, University of Copenhagen, Copenhagen, Frederiksberg C, Denmark
    Search for more papers by this author
  • Jorunn N. Johansen,

    1. Institut National de la Recherche Agronomique, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    2. AgroParisTech, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    Current affiliation:
    1. Institute of Immunology, Oslo University Hospital HF, Rikshospitalet, Oslo, Norway
    Search for more papers by this author
  • Adiphol Dilokpimol,

    1. Department of Plant Biology and Biotechnology, University of Copenhagen, Copenhagen, Frederiksberg C, Denmark
    Search for more papers by this author
  • Aurélia Rolland,

    1. Institut National de la Recherche Agronomique, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    2. AgroParisTech, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    Current affiliation:
    1. Anjou Recherche Semences, UMR1191 Physiologie Moleculaire des Semences, Angers Cedex 01, France
    Search for more papers by this author
  • Katia Belcram,

    1. Institut National de la Recherche Agronomique, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    2. AgroParisTech, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    Search for more papers by this author
  • Stéphane Verger,

    1. Institut National de la Recherche Agronomique, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    2. AgroParisTech, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    Search for more papers by this author
  • Toshihisa Kotake,

    1. Division of Life Science, Graduate School of Science and Engineering, Saitama University, Saitama, Japan
    Search for more papers by this author
  • Yoichi Tsumuraya,

    1. Division of Life Science, Graduate School of Science and Engineering, Saitama University, Saitama, Japan
    Search for more papers by this author
  • Satoshi Kaneko,

    1. Food Biotechnology Division, National Food Research Institute, Tsukuba, Ibaraki, Japan
    Search for more papers by this author
  • Theodora Tryfona,

    1. Department of Biochemistry, Cambridge University, Cambridge, UK
    Search for more papers by this author
  • Paul Dupree,

    1. Department of Biochemistry, Cambridge University, Cambridge, UK
    Search for more papers by this author
  • Henrik V. Scheller,

    1. Department of Plant Biology and Biotechnology, University of Copenhagen, Copenhagen, Frederiksberg C, Denmark
    2. Joint BioEnergy Institute, Feedstocks Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
    Search for more papers by this author
  • Herman Höfte,

    1. Institut National de la Recherche Agronomique, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    2. AgroParisTech, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    Search for more papers by this author
  • Gregory Mouille

    Corresponding author
    1. Institut National de la Recherche Agronomique, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    2. AgroParisTech, Unité Mixte de Recherche 1318, Institut Jean-Pierre Bourgin, Saclay Plant Sciences, Versailles, France
    • Department of Plant Biology and Biotechnology, University of Copenhagen, Copenhagen, Frederiksberg C, Denmark
    Search for more papers by this author

For correspondence (e-mail gregory.mouille@versailles.inra.fr).


Arabinogalactan proteins (AGPs) are a complex family of cell-wall proteoglycans that are thought to play major roles in plant growth and development. Genetic approaches to studying AGP function have met limited success so far, presumably due to redundancy within the large gene families encoding AGP backbones. Here we used an alternative approach for genetic dissection of the role of AGPs in development by modifying their glycan side chains. We have identified an Arabidopsis glycosyltransferase of CAZY family GT31 (AtGALT31A) that galactosylates AGP side chains. A mutation in the AtGALT31A gene caused the arrest of embryo development at the globular stage. The presence of the transcript in the suspensor of globular-stage embryos is consistent with a role for AtGALT31A in progression of embryo development beyond the globular stage. The first observable defect in the mutant is perturbation of the formative asymmetric division of the hypophysis, indicating an essential role for AGP proteoglycans in either specification of the hypophysis or orientation of the asymmetric division plane.


Arabinogalactan proteins (AGPs) are a class of abundant and highly diverse cell-surface proteoglycans. AGPs are thought to play an essential role in many aspects of growth and development (Seifert and Roberts, 2007; Ellis et al., 2010). The glycan part of AGPs, which represents up to 90% of the mass of the molecule, is complex and heterogeneous. It consists of a β–1,3-linked galactan backbone substituted at O6 with variable side chains rich in galactose and arabinose (type II arabinogalactan). Various structures have been found in distinct species, tissues and developmental stages (Tsumuraya et al., 1988; Tan et al., 2004; Tryfona et al., 2010, 2012). In addition, Estévez et al. (2006) used synthetic AGP peptides expressed in Arabidopsis to show that glycosylation patterns are highly regulated during development. Glycosylation of AGPs by glycosyltransferases (GTs) occurs in the Golgi apparatus. More than ten GTs are required to synthesize the various linkages that have been identified in AGP glycans (Tan et al., 2004). So far, only one of those activities (represented by two α–1,2-fucosyltransferases, FUT4 and FUT6) has been biochemically characterized (Wu et al., 2010), and only in crude plant membrane preparations not purified proteins.

Arabinogalactan protein (AGP) proteoglycans have been implicated in a number of cellular processes. For instance, xylogen, a hybrid molecule with an AGP and a lipid-transfer domain, has the ability to induce xylem formation in Zinnia elegans cell cultures (Motose et al., 2004). Deglycosylation of this molecule abolished the activity, demonstrating the importance of the glycan for its function. Other studies provided correlative evidence for a role for specific AGPs in plant development based on immunolabeling with monoclonal antibodies against AGP carbohydrate epitopes (Seifert and Roberts, 2007). For instance, certain AGP epitopes have well-defined temporal and spatial expression patterns during embryogenesis (Pennell et al., 1991; Coimbra and Salema, 1997; Stacey et al., 1990). In addition, somatic embryogenesis was inhibited or stimulated by certain AGP carbohydrate epitopes (Kreuger and van Hoist, 1995; McCabe et al., 1997). These findings show that AGP glycans undergo various changes during embryogenesis. However, as the precise epitope structure for these antibodies is not known, critical features of the molecular function of the AGPs remain to be determined.

Here we present evidence that the Arabidopsis gene At1g32930 encodes a β–galactosyltransferase (named AtGALT31A) that is involved in elongation of β–1,6-galactan side chains on AGPs. A mutation in this gene causes an arrest in embryo development at the globular stage. The first observable defect during embryogenesis is an abnormal asymmetric cell division of the hypophysis.


AtGALT31A is a galactosyltransferase that elongates β–1,6-galactan side chains of AGPs

As part of our study on the function of AGP glycan structures, we focused on CAZY family GT31 (www.cazy.org). This family has 33 members in Arabidopsis, 20 of which contain a putative galactosyltransferase domain (Pfam01762, http://pfam.sanger.ac.uk/). One family member GALT1, encoded by At1g26810, has been identified as a β–1,3-galactosyltransferase that is involved in formation of the Lewis a epitope (Strasser et al., 2007) on N–linked glycans. This family may therefore comprise candidates for synthesis of β–linked galactosyl residues on AGP glycan chains. Here we focused on the protein encoded by At1g32930. This polypeptide of 399 amino acids contains a single transmembrane domain at its N–terminus (residues 13–35, predicted by the TMHMM server; http://www.cbs.dtu.dk/services/TMHMM/) and a C–terminal galactosyltransferase domain (Pfam01762, residues 144–341), but lacks the lectin domain that is present in GALT1 (Figure 1; Strasser et al., 2007). At1g32930 transcript levels are low throughout development in Arabidopsis and higher in the root xylem, elongating zones (Brady et al., 2007), the shoot apex (Winter et al., 2007) and during seed development from ovule to the globular stage (Genevestigator, https://www.genevestigator.com/gv/plant.jsp). The overall expression pattern is therefore consistent with a role for At1 g32930 in plant development.

Figure 1.

Structure of GT31 genes. Domain structure of GT31 family members in Arabidopsis. Both are type II membrane proteins with a single predicted membrane-spanning domain at the N–terminus (filled block). GALT1 (At1 g26810, 643 amino acids long) contains both a galectin domain (PF00337, residues 172–362) and a galactosyltransferase domain (PF01762, residues 406–589), while At1 g32930 (399 amino acids long) only contains a galactosyltransferase domain (PF01762, residues 144–341).

To investigate the biochemical function of At1 g32930, epitope-tagged full-length protein or the soluble catalytic domain were expressed in Nicotiana benthamiana and Escherichia coli, respectively. Recombinant proteins were affinity-purified prior to enzyme assays (Figure 2a). To identify the donor substrate, we tested the GT activity of the recombinant protein using a broad range of NDP-14C-sugars. As an acceptor, we used heat-treated microsomes from N. benthamiana expressing a Gum arabic glycopeptide motif 8 times repeat conjugated to GFP (GAGP8–GFP) synthetic peptide (Xu et al., 2005), which encodes an arabinogalactan (AG) glycomodule.

Figure 2.

Identification of the donor substrate and acceptor of AtGALT31A. (a) Western blot analysis of the purified recombinant proteins expressed in E. coli (left) or N. benthamiana (right). Recombinant proteins of the expected size (approximately 75 kDa in both systems) were detected using antibodies against GST (left) or GFP (right). (b) Substrate mixture I (UDP-14C-Xyl, UDP-14C-Glc, GDP-14C-Man and GDP-14C-Fuc) and substrate mixture II (UDP-14C-Gal, UDP-14C-GlcA and UDP-14C-GlcNAc) were incubated with control (N. benthamiana transformed with p19) and recombinant AtGALT31A (N. benthamiana transformed with p19 and the AtGALT31A–GFP fusion construct) proteins purified using GFP antibody. A synthetic glycomodule (GAGP8–GFP, Xu et al., 2005) expressed in N. benthamiana was used as the AGP acceptor (Figure S1). The amount of precipitated 14C-product is presented as ratio to the control level (100%). Values are means and +/− SD (= 3). (c) Each individual substrate in mixture II was tested by the same assay. Values are means and +/− SD (= 3). (d) Purified ATGALT31A expressed in E. coli was incubated with radish AGP treated with arabinofuranosidase in the presence of UDP-14C-Gal. The resulting 14C-product was treated with buffer, P. chrysosporium exo-β–1,3-galactanase, endo-β–1,6-galactanase, or endo-β–1,6-galactanase and P. chrysosporium exo-β–1,3-galactanase together, and released 14C-oligosaccharides were separated by size exclusion chromatography. Materials isolated in fractions 19–20, 21–22 and 23–24 after endo-β–1,6-galactanase treatment were further separated by TLC (inset). Galactose (Gal) and β–1,6-linked galactobiose (Gal2) were detected as the major products, indicative of production of β–1,6-linked galacto-oligosaccharides by ATGALT31A.

Using this acceptor, higher levels of 14C-sugar were incorporated using the substrate mixture UDP-14C-Gal, UDP-14C-GlcA and UDP-14C-GlcNAc (Mix II, Figure 2b) compared to a mixture of UDP-14C-Xyl, UDP-14C-Glc, GDP-14C-Man and GDP-14C-Fuc (Mix I, Figure 2b). Of the three nucleotide sugars in Mix II, UDP-14C-Gal was the most efficient donor substrate (Figure 2c). This identified the enzyme as a galactosyltransferase (GALT).

Next, we tested possible acceptor candidates in the assay. The GT31 family is a family of predicted inverting GTs. Inverting galactosyltransferases transfer d–galactose from UDP-α–d-galactose to the non-reducing end of acceptor sugar and create a β–linkage. Known wall polysaccharides with β–linked galactose are xyloglucan (β–1,2-Gal linked to Xyl), pectin rhamnogalacturonan I (β–1,2-Gal linked to Rha and β–1,4-Gal linked to Gal), pectin rhamnogalacturonan II (β–1,4-Gal linked to AcefA) and AGP (β–Gal linked to hydroxyproline, β–1,3-Gal linked to Gal and β–1,6-Gal linked to Gal). Glycolipids were not taken into account, as they are not detected in our assay. As xyloglucan GalTs, which belong to family GT47, have already been identified (Madson et al., 2003; Jensen et al., 2012), we investigated the remaining possibilities. First, we confirmed the presence of AG in the microsome acceptor from plants expressing GAGP8-GFP. We structurally characterized the AG modules by polysaccharide analysis using carbohydrate gel electrophoresis (PACE) and MALDI-ToF-MS. GAGP8-GFP was hydrolyzed using Irpex lacteus AG-specific exo-β–1,3-galactanase, which specifically cleaves the β–1,3-linked galactan backbone of AG regardless of the presence of side-chain substitutions (Kotake et al., 2009). From Figure S1(a), it is clear that the GAGP8-GFP glycomodule is susceptible to the exo-β–1,3-galactanase, resulting in a clear ladder of diverse oligosaccharides with degrees of polymerization from 1 to 8. The release of the Gal monosaccharide indicated the presence of unsubstituted main-chain β–1,3-galactan. MALDI-ToF-MS shows the release by endo-β–1,6-galactanase of oligosaccharides that are partly decorated with a pentose, probably Ara (Figure S1b). These results indicate an AG structure similar to AGPs from wheat flour (Triticum aestivum) (Tryfona et al., 2010) and Arabidopsis leaf (Tryfona et al., 2012), which consist of a β–1,3-linked galactan backbone substituted by β–1,6-galactan oligosaccharides. A Yariv radial gel diffusion assay showed Yariv reactivity, albeit weak (Figure S1c), suggesting the presence of β–1,3-galactan chains longer than seven residues (Kitazawa et al., 2013) on the GAGP8-GFP glycomodule. Taken together, these data confirm the presence of typical type II polysaccharides on the GAGP8–GFP protein that may provide acceptor sites for β–1,3- and β–1,6-galactosyltransferases.

We investigated further the ability of AG-related molecules to act as acceptors, and found that radish (Raphanus sativus L.) AGP treated with arabinofuranosidase (Tsumuraya et al., 1988) worked as an acceptor, whereas unsubstituted β–1,3-galactan did not (Figure S2). Treatment of the 14C-product formed on radish AGP with a highly specific endo-β–1,6-galactanase (Ichinose et al., 2008) released 14C-Gal and 14C-β–1,6-galactobiose (Figure 2d). These findings suggest that AtGALT31A acts as an AGP side chain-elongating β–1,6-GalT. Treatment of the products with Phanerochaete chrysosporium exo-β–1,3-galactanase alone released only a trace amount of 14C-galacto-oligosaccharides. Likewise, combined treatment with endo-β–1,6-galactanase and P. chrysosporium exo-β–1,3-galactanase did not release more 14C-labeled galacto-oligosaccharides than endo-β–1,6-galactanase alone (Figure 2d). This suggests that AtGALT31A elongates β–1,6-galactan and does not extend β–1,3-galactan.

We next tested well-defined short galacto-oligosaccharides as acceptors, such as β–1,6-galactobiose, β–1,6-galactotriose, β–1,3-galactobiose and β–1,4-galactobiose. 14C-Gal incorporation in a neutral product was measured after removal of unreacted UDP-14C-Gal using a strong anion exchanger, and the results are presented in Table 1. We observed an apparent increase of 14C-Gal in a neutral product in the presence of β–1,6-galactotriose (t test, = 0.02) using recombinant protein expressed in E. coli. Other combinations of recombinant protein and oligosaccharides did not show any statistically significant increase, although we cannot exclude the possibility that longer oligosaccharides, e.g. β–1,4-galactotriose, may work as acceptor as β–1,6-galactobiose was not significantly used as an acceptor but β–1,6-galactotriose was. We were unable to test this possibility as β–1,4-galactotriose is not commercially available.

Table 1. Galactosyltransferase assay using oligosaccharide acceptors
  Escherichia coli Nicotiana benthamiana
  1. Purified recombinant enzymes expressed in E. coli or N. benthamiana were incubated with UDP-14C-Gal in the presence of oligosaccharide acceptor. The level of 14C-Gal (pmol) in the reaction mixture after removal of unreacted UDP-14C-Gal using a strong anion exchanger is shown (means ± SD, = 3). The statistical significance of differences compared with control was determined using an unpaired Student's t test (a= 0.3, b= 0.02, c= 0.09).

β–1,3-Gal20.08 ± 0.050.11 ± 0.093.22 ± 0.323.58 ± 0.22
β–1,4-Gal20.09 ± 0.050.10 ± 0.06Not testedNot tested
β–1,6-Gal20.10 ± 0.050.21 ± 0.12a3.74 ± 0.643.60 ± 0.38
β–1,6-Gal32.29 ± 0.203.84 ± 0.50b2.68 ± 0.823.67 ± 0.36c

To confirm in planta the role of AtGALT31A in elongating β–1,6-galactan, we analyzed the glycans of AGPs of three independent lines over-expressing a functional GFP–AtGALT31A fusion protein. AGPs of 10-day-old light-grown seedlings were purified, and a linkage analysis of the glycosyl moiety was performed. All three lines showed an increase in the relative amount of 1,6-linked Gal in the AGP fraction, and minor changes in other monosaccharide linkages (Figure 3). Transgenic line 3, displaying the highest amount of 1,6-linked Gal, also showed a decrease of 1,2,5-linked Ara and 1,3-linked Gal, reflecting additional alterations in the AGP glycans. These data are in full agreement with the activity shown for the recombinant enzyme, and confirm that AtGALT31A elongates β–1,6-galactan on AGP glycans.

Figure 3.

Glycosidic linkage analysis of Yariv-precipitable fractions of 10-day-old light-grown seedlings. Lines 1, 2 and 3 correspond to three independently transformed plants (atgalt31a mutant complemented using the 35S:GFP-At1g32930cDNA construct). WS is the Wassilewskija accession. Values are means ± SD (= 3 biological repeats).

GFP–AtGALT31A accumulates in the Golgi apparatus and unidentified organelles

The construct used for complementation of the mutant (35S:GFP-AtGALT31A) was used for transient expression in N. benthamiana. The AtGALT31A protein was localized in mobile subcellular compartments (see Movie S1) that co-localized with a marker protein of the Golgi apparatus (mannosidase I–mCherry, Figure 4). In conclusion, AtGALT31A is present at least in the Golgi apparatus, the compartment in which glycan-synthesizing GTs are expected to be active.

Figure 4.

The AtGALT31A–GFP fusion protein accumulates in the Golgi apparatus. Imaging of AtGALT31A–GFP (cyan) and the Golgi marker mannosidase I–mCherry (magenta, Nelson et al., 2007), both transiently expressed in epidermal cells of N. benthamiana. Scale bars = 10 μm.

AtGALT31A is required for progression of embryogenesis beyond the globular stage

We isolated a mutant harboring a T–DNA insertion in the 9th exon of the AtGALT31A gene. Of 24 genotyped plants, we observed eight wild-type plants, 16 plants heterozygous for the mutation, and none homozygous for the mutation. The absence of plants homozygous for this mutation suggests gametophytic or embryo lethality. This was confirmed by examining the siliques of heterozygous plants. At 3 days post-anthesis, the developing seeds in the siliques appeared normal, while at a later stage, 25% of the seeds became white and died, leaving empty spaces in mature siliques. To define the stage at which the embryo was aborted, we examined embryos from 3 to 10 days post-anthesis. At 10 days post-anthesis, corresponding to the heart stage under our growth conditions, 25% of the embryos were arrested (154 seeds in four siliques of four individual plants, Figure 5a). The same proportion of embryos was arrested at a later stage. This segregation (25%) pattern indicates that the mutant phenotype is monogenic and recessive.

Figure 5.

AtGALT31A is essential for embryogenesis. (a) Nomarski images of embryos in siliques from +/atgalt31a plants. The percentage of embryos of the two stages (heart stage (left) and globular stage (right) observed in +/atgalt31a siliques is indicated at the top right of the images. Over 100 embryos were observed for each stage. (b) Confocal section of a propidium iodide-stained embryo at the globular stage. Embryos showing an abnormal oblique division plane in the hypophysis (indicated by a white arrow) are only found in +/atgalt31a plants. (c) Detection of GUS transcript by in situ hybridization. Developing seeds from transgenic plants carrying a construct containing a 2 kb fragment upstream from the AtGALT31A coding sequence fused to the GUS reporter gene were used. The sectioned immature seeds were hybridized with a full-length sense probe (right) and an antisense probe (control, left) to reveal the presence of the GUS transcript. Scale bars = 10 μm.

To confirm that this phenotype was caused by the mutation in AtGALT31A, we complemented the embryo-lethal phenotype using the 35S:GFP-AtGALT31A construct, and recovered homozygous mutant lines with no visible phenotype. In conclusion, the embryo-lethal phenotype is caused by the mutation in the AtGALT31A gene. These complemented lines were also used to confirm the in vivo function of the AtGALT31A protein.

To further characterize the mutant phenotype, we observed embryos from heterozygous parents after propidium iodide staining. Embryos up to the globular stage appeared normal. The first observable defect was an abnormal asymmetric formative division of the hypophysis, the uppermost cell of the suspensor (indicated by an arrow in Figure 5b). We also observed subsequent ectopic divisions in the daughter cells of the hypophysis (i.e. the basal cell and the lens cell), which were never observed in wild-type embryos (Figure S3). AtGALT31A is expressed during early embryo development (BAR Arabidopsis eFP browser, http://bar.utoronto.ca/welcome.htm). Transcriptomic analysis on microdissected samples from developing Arabidopsis seeds shows AtGALT31A transcript in the embryo at the pre-globular stage and in the embryo and the suspensor at the globular stage (BAR Arabidopsis eFP browser, http://bar.utoronto.ca/welcome.htm). Suspensor expression of the transcript was confirmed in plants expressing the reporter gene uidA (encoding GUS) driven by the 2 kb upstream sequence of AtGALT31A (Figure 5c and Figure S4). We used this construct for the in situ hybridization as endogenous transcript levels were below the detection limits of this technique. This expression pattern is consistent with a requirement for AtGALT31A in progression of embryo development beyond the globular stage. In addition, the results suggest that one or more correctly glycosylated AGPs are required either for specification of the hypophysis within the suspensor or control of the first asymmetric formative division within this cell.


AtGALT31A possesses galactosyltransferase activity elongating β–1,6-galactan side chains of AGPs

Recombinant AtGALT31A showed 14C-Gal transferase activity to products digestible by an AGP-specific endo-β–1,6-galactanase. One of the acceptors used in this study, radish AGP treated with arabinofuranosidase, is relatively well characterized. Tsumuraya et al. (1988) reported the presence of 3–branched, 6–branched, 3,6–branched and terminal Gal (6, 44, 16 and 11 mol%, respectively; AGP IV–J fraction in Table III of Tsumuraya et al., 1988). Furthermore, Kotake et al. (2004) reported the release of β–1,6-linked galacto-oligosaccharides with degrees of polymerization from 2 to 8 upon partial digestion with recombinant endo-β–1,6-galactanase from Trichoderma viride (Kotake et al., 2004). A second AGP acceptor for recombinant AtGALT31A, GAGP8–GFP was characterized here, and its treatment with Irpex lacteus β–1,3-galactanase resulted in release of β–1,6-galactan with degrees of polymerization from 2 to 8 from this protein (Figure S1a). Therefore, both acceptors for AtGALT31A contain rather long β–1,6-galactan in the side chains. It is likely that recombinant AtGALT31A transferred 14C-Gal to long β–1,6-galactan as the product was digested by endo-β–1,6-galactanase (Figure 2d). This interpretation is also supported by the higher level of incorporation of 14C-Gal into β–1,6-galactotriose compared to β–1,6-galactobiose. An alternative possibility, initiation of the side chain from O6 positions of the β–1,3-galactan main chain, is unlikely as such a branching point is not a substrate for the endo-β–1,6-galactanase used and β–1,3-galactan did not work as acceptor. The putative site of addition of 14C-Gal in radish AGP by AtGALT31A is indicated in Figure 6. The modification of the glycans of AGPs synthesized in plants ectopically expressing AtGALT31A is in agreement with this conclusion. The β–1,6-galactosyltransferase activity of a family GT31 member may at first sight appear surprising as other members of this family have been identified as β–1,3-galactosyltransferases (Qu et al., 2008). However, the existence of a β–1,6-glycosyl transferase activity among members of another β–1,3-glycosyltransferase family already has been observed (i.e. a β–1,6-GlcNAc transferase in Trypanosoma brucei, Michael Ferguson, College of Life Sciences, University of Dundee, Dundee, personal communication).

Figure 6.

Schematic AG model structure for radish AGP treated with arabinofuranosidase (based on linkage analysis described in Tsumuraya et al., 1988 and Kotake et al., 2004). The supposed site of action by AtGALT31A towards the non-terminal Gal in the β–1,6-galactan side chain is indicated by the arrow.

Control of AGP glycan structures is essential for early steps in embryogenesis

This study provides evidence for the requirement of one or more correctly glycosylated AGPs for progression of embryogenesis beyond the globular stage. Atgalt31a mutant embryos are arrested at the globular stage and no defects are observed prior to this stage. The first observable defect is the aberrant asymmetric formative division of the hypophysis, suggesting a role for AGPs either in specification of the hypophysis or control of asymmetric targeting of the division plane to the mother cell wall. The expression of the AtGALT31A gene in the suspensor at the globular stage is also consistent with such a role.

A role for specific AGP epitopes in embryogenesis has already been suggested by use of anti-AGP monoclonal antibodies (Seifert and Roberts, 2007). For instance, the JIM4 antibody reveals epitopes in the apical domain of late globular and early heart-stage embryos (Stacey et al., 1990). More interestingly in this context is the distribution of the JIM8 epitope, which, like the AtGalT31A transcript, is present in the suspensor in globular oilseed rape embryos (Pennell et al., 1991). The JIM8 epitope also plays a key role in somatic embryogenesis in carrot suspension cells (Daucus carota; McCabe et al., 1997). Indeed, somatic embryogenesis is initiated by an asymmetric cell division generating a JIM8-negative daughter cell and a JIM8-positive daughter cell. The latter cells become vacuolated and die, whereas JIM8-negative cells have embryogenic capacity but only in the presence of JIM8-positive cells or growth medium conditioned with these cells. JIM8-positive cells therefore appear to produce a soluble factor that is required for the progression of embryogenesis. An interesting possibility is that AtGALT31A is required for production of the JIM8 epitope or a related molecule, which in turn may provide the suspensor-derived signal that is required for the progression of embryo development.

Our future research will focus on identification of the AtGALT31-dependent AGP glycan epitope, and whether this epitope acts as a suspensor-derived factor that is required for the progression of embryo development. In addition, we will investigate the molecular basis of the AGP-dependent control of the asymmetric cell division in the hypophysis, its relationship to the progression of embryonic development, and whether AGPs play similar roles in other formative cell divisions.

Experimental Procedures

Plant materials and growth conditions

Arabidopsis thaliana accession Wassileskija (WS) was used as the wild-type. The At1 g32930 mutant line (FLAG_379B06) was isolated from the Versailles T–DNA insertion collection (http://dbsgap.versailles.inra.fr/portail/). The left flanking sequence of the T–DNA insertion site was amplified by PCR using the T–DNA-specific primer Tag5 (5′-CTACAAATTGCCTTTTCTTATCGAC-3′) and a At1g32930 gene-specific primer (5′-CCACACCATTGAAGCTGAAGA-3′). PCR products were sequenced, and the site of insertion was confirmed. A second At1g32930 gene-specific primer (5′-CTATGCCAAATTGCTCCATCG-3′) was used for genotyping of the non-mutated copy of the gene. Seeds were sown on soil, and plants were grown at 21°C under long-day conditions (16 h light/8 h dark).


Full-length At1g32930 cDNA (AtGALT31A) was PCR-amplified from an Arabidopsis cDNA as a template, and cloned into Gateway entry vector pDONR207 (Invitrogen, http://www.invitrogen.com) using two rounds of PCR using the gene-specific primers 5′-ATGGGAATGGGAAGGTATCAGA-3′ and 5′-GAAACTACTATGCCAAATTGCTCC-3′, and followed by the universal primers 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCGAAGGAGATAGAACCATG-3′ and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTCTCCACCTCCGGATC-3′ (Castelli et al., 2004, http://www-urgv.versailles.inra.fr/atome/protocols.htm). In order to determine the subcellular localization of the protein, the cDNA cloned in the entry vector (with stop codon) was transferred to the Gateway pGWB6 binary vector (Invitrogen), placing the GFP at the N–terminus of the encoded protein.

For expression in N. benthamiana as a fusion protein with GFP at the C–terminus of the encoded protein, the entry clone without the stop codon was transferred to the pGWB5 binary vector. The destination clone was transformed into Agrobacterium strain C58C1 pGV3850 (Shen and Forde, 1989). For expression in E. coli, a soluble truncated construct was created by PCR amplification using primers 5′-CACCAGGCTTTTGGCAAGTTTTGAAA-3′ and 5′-TTAGAAACTACTATGCCAAATTGC-3′ and the AtGAT31A entry clone as a template. The PCR product was cloned into the pENTR/D–TOPO vector using the pENTR directional TOPO cloning kit (Invitrogen). A control construct was created using oligonucleotides 5′-CACCTAATAG-3′ and 5′-CTATTA-3′, by heating them at 95°C for 5 min followed by cooling to room temperature. The double-stranded construct produced, carrying a 5′-CACC sticky end followed by two stop codons (TAATAG), was cloned into the pENTR/D–TOPO vector as described above, moved into a pDEST15 destination vector (Invitrogen) by LR reaction, and transformed into the E. coli BL21 (DE3) strain.

For construction of the prom:GUS fusion, a fragment of 1956 bp upstream from the start codon of the AtGALT31A coding sequence was PCR-amplified from wild-type (Wassilewskija) using primers 5′-GATACCTTCCCATTCCCATGGTTTC-3′ and 5′-GTCCATGGGTACGATAAATGGGTGAG-3′, and cloned into the vector pCAMBIa1391Z (http://www.cambia.org/daisy/cambia/585.html).

Heterologous expression and affinity purification of recombinant protein

Transient expression of the construct co-infiltrated with p19 into N. benthamiana by Agrobacterium-mediated transformation was performed as described previously (Voinnet et al., 2003). Microsome preparation was performed as described by Geshi et al. (2002). Microsomal membranes containing protein concentrations of 5 mg/ml (Bradford, 1976) were solubilized by adding n–dodecyl β–maltoside to a final concentration of 5 mm. Affinity purification was performed by incubation with 0.8 μg anti-GFP antibody per 1 mg protein) at 4°C for 2–3 h with rotation prior to addition of 20 μl protein G agarose beads (50% slurry, pre-equilibrated in PBS) followed by incubation overnight at 4°C. The beads were collected by brief centrifugation (2500 g, 5 min., at 4°C) followed by three washing steps in PBS (5 min. per wash at 4°C). Finally, enzyme-immobilized beads (4–5 μl per assay) were suspended in an equal volume of PBS. Expression of a GST fusion construct in the E. coli BL21 DE3 strain, preparation of spheroplasts, and affinity purification of the recombinant enzyme were performed as described by Vuttipongchaikij et al. (2012), except that, instead of inoculation at 150 rpm at 15°C for 2 days, we prolonged the culture for 3 days without induction using isopropyl thio-β–d–galactoside. Recombinant proteins bound to the glutathione beads were suspended in PBS (75% slurry) and used for enzyme assays (0.2–0.3 μg protein based on Western blotting).

Substrates, acceptors and hydrolases for the enzyme assay

Preparation of the following polysaccharides/oligosaccharides and hydrolases was performed as described previously: radish AGP treated with arabinofuranosidase (Tsumuraya et al., 1988), β–1,3-galactan, β–1,3-galactobiose, β–1,6-galactobiose and β–1,6-galactotriose from larch arabinogalactan (Aspinall et al., 1958; Sekimata et al., 1989; Ichinose et al., 2009), exo-β–1,3-galactanase from P. chrysosporium (Ichinose et al., 2005) and Irpex lacteus (Kotake et al., 2009), α–arabinofuranosidase (EC; Tsumuraya et al., 1984), and endo-β–1,6-galactanase (EC from Trichoderma viride (Kotake et al., 2004) and Streptomyces avermitilis (Ichinose et al., 2008). β–1,4-galactobiose was obtained from Megazyme (www.megazyme.com). UDP-14C-Xyl, UDP-14C-Glc, UDP-14C-Gal, GDP-14C-Fuc, GDP-14C-Man, UDP-14C-GlcA and UDP-14C-GlcNAc were obtained from NEN ( http://www.dupont.com/). UDP-Xyl was obtained from CarboSource (http://www.ccrc.uga.edu/~carbosource/CSS_home.html), and other nucleoside diphosphate sugars were obtained from Calbiochem/Novabiochem (http://www.emdmillipore.com/chemicals/).

Enzyme assay

The synthetic glycomodule GAGP8–GFP construct, placed under the control of the 35S in the pBI121 vector (Xu et al., 2005), was kindly provided by Marcia Kieliszewski (Department of Chemistry and Biochem, Ohio University, Athens, OH). We transformed this plasmid into Agrobacterium strain C58C1 pGV3850 (Shen and Forde, 1989) and infiltrated N. benthamiana to prepare acceptors for AG glycosylation activity. For identification of donor substrates, combined or individual substrates (UDP-14C-Xyl, UDP-14C-Glc, GDP-14C-Man, GDP-14C-Fuc, UDP-14C-Gal, UDP-14C-GlcA and UDP-14C-GlcNAc; 0.1 mm concentration each, 277 Bq) were tested for the GT reaction in the presence of recombinant enzyme expressed in N. benthamiana immobilized on beads (4–5 μl), microsomes (5 μl) prepared from N. benthamiana after infiltration of the Agrobacterium strain harboring the GAGP8–GFP plasmid and heat-treated as acceptor (Orfila et al., 2005), 0.1 m HEPES/NaOH, pH 7.0, 10 mm MnCl2 in a total volume of 25 μl. The reaction was performed at 25°C for 16 h. The products were precipitated in the presence of 0.25 μl of 10 mg/ml horseradish peroxidase and 0.28 μl of 0.3% H2O2 (Kjellbom et al., 1997). The presence of 14C-sugars in the pellet was determined by scintillation counting after washing away non-specific radioactivity.

For identification of the acceptor, various polysaccharides/oligosaccharides were tested using the recombinant enzyme expressed in E. coli. In the assay described above, GAGP8–GFP microsomes were replaced by 5–10 μg polysaccharides (arabinofuranosidase-treated radish root AGP or β–1,3-galactan) or 20 mm oligosaccharides (β–1,3-galactobiose, β–1,6-galactobiose, β–1,6-galactotriose or β–1,4-galactobiose). Polysaccharides that had been partially digested by exo-β–1,3-galactanase were also tested. Reaction mixtures using polysaccharide acceptor were applied to TLC to exclude small compounds (mainly galactose and UDP-Gal) as described by Kotake et al. (2004). Silica material around the sample application area was collected, and suspended in 100 μl McIlvaine buffer (McIlvaine, T.C. 1921), pH 5.5, and treated with AG-specific hydrolases, exo-β–1,3-galactanase (20 mU) and/or endo-β–1,6-galactanase (1 mU) in the presence of 0.02% NaN3 for 16 h at 25°C. One unit of enzyme activity is defined as the amount of enzyme that released 1 micro mol of galactose/min from beta-1,3- and beta-1,6-galactan, respectively (Ichinose et al., 2005, Ichinose et al., 2008). Released 14C-oligosaccharides were separated by size exclusion chromatography with a Superose 12 column (http://www.gelifesciences.com/) using 50 mm ammonium formate, pH 5.5, at a flow-rate of 0.4 ml min−1 (Geshi et al., 2002). Fractions (0.8 ml) were collected at 2 min intervals, and the radioactivity was determined by scintillation counting. Fractions containing 14C-active materials were also analyzed by TLC (Kotake et al., 2004). Reaction products produced using oligosaccharide acceptors were passed through a anion resin Dowex 1 × 8 (chloride form, 200–400 mesh Sigma-Aldrich, http://www.sigmaaldrich.com/france.html), and the radioactivity of the 14C-sugars in the flow-through was determined by scintillation counting.

AGP purification and linkage analysis

AGPs were purified using Yariv reagent from 400 light-grown Arabidopsis seedlings grown in vitro for 10 days on minimal medium, as described previously (Schultz et al., 2000; Eudes et al., 2008). For each genotype, 100 mg purified AGPs were permethylated as described previously (Kim and Carpita, 1992). Permethylated polysaccharides were then hydrolyzed, reduced and derivatized as described by Burton et al. (2000). The derivatives were analyzed by GC/MS (Agilent Technologies, www.home.agilent.com). A BPX 70 column (length 25 m, internal diameter 0.22 mm; SGE Analytical Science, http://www.sge.com/) was used for all separations. Identification of the derivatives and deduction of the glycosidic linkages were based on both the elution order of standards and fragment ion signatures.

PACE analysis and Yariv radial gel diffusion assay

The PACE analysis was performed as described by Tryfona et al. (2010), and the Yariv radial diffusion assay was performed as described by Kitazawa et al. (2013) from microsomes prepared from N. benthamiana after infiltration with an Agrobacterium strain harboring the GAGP8–GFP plasmid.

Oligosaccharide sample clean-up, permethylation of AG polysaccharides and MALDI-ToF-MS analysis

Following the enzyme digestions and prior to permethylation, released peptides and enzymes were removed from the mix using reverse-phase Sep-Pak C18 cartridges (Waters, http://www.waters.com) as described previously (Tryfona et al., 2010). Dry samples were desalted using cation Dowex cations resins 50W×8 (hydrogen form, 50–100 mesh, Sigma-Aldrich) as described by Tryfona et al. (2010). Permethylation was performed as described by Ciucanu and Kerek (1984). Dry permethylated samples were resuspended in 100 μl MeOH and kept at room temperature for MALDI-ToF-MS analysis. Methylated MeOH-dissolved samples (5 μl) were mixed with 5 μl 2,5–DHB 2,5-dihydroxybenzoic acid matrix (10 mg/ml dissolved in 50% MeOH), and 1 μl of the mixture was spotted on a MALDI target plate and analyzed by MALDI-ToF/ToF-MS/MS (4700 proteomics analyzer, Applied Biosystems, http://www.appliedbiosystems.com) as previously described (Tryfona et al., 2010).

Phenotypic analysis

For histological analysis of embryos, ovules were processed as described previously (Mayer et al., 1993). Whole-mount preparations were used for quantitative analysis of embryo and seedling phenotypes. Embryos were mounted in a mixture of chloral hydrate/distilled water/glycerol (8:3:0.5), incubated for 45 min at 20°C, and viewed with Nomarski optics using a light microscope (Leica DM2500, http://www.leica-microsystems.com). Propidium iodide staining and observation of the embryo were performed as described previously (Truernit et al., 2008).

Laser scanning confocal microscopy

Images were collected using a spectral Leica SP2 AOBS confocal microscope (Leica Microsystems, http://www.leica-microsystems.com) or an LSM 710 confocal microscope (Zeiss, http://www.zeiss.com) equipped with an argon laser. Excitation was performed using an argon ion laser at 488 nm, and fluorescence was detected at 495–540 nm for imaging expression of GFP constructs and 560–620 nm for imaging expression of mCherry constructs. Each image shown represents either a single focal plane or a projection of individual images taken as a z–stack.

In situ hybridization

Tissue fixation, embedding, sectioning and in situ hybridization were performed as described previously (Nikovics et al., 2006). An antisense probe for the full-length open reading frame of GUS was synthesized in vitro and labeled with digoxigenin-UTP using a gel-purified PCR product that included the T7 RNA polymerase binding site as template. The two specific primers used for the production of the probe are GUS ATG (5′-ATGATGGGTCAGTCCCTTATGTTACGT-3′) and GUS + T7 (5′-TGTAATACGACTCACTATAGGGCTCATTGTTTGCCTCCCTGCTG-3′).


Support was provided by the European Commission Framework Programme FP6 projects AGRON-OMICS (grant number LSHG-CT-2006-037704) and WALLNET (grant number 512265) and FP7 project RENEWALL (grant number 211981), and the Office of Science, US Department of Energy (contract DE-AC02-05CH11231 with Lawrence Berkeley National Laboratory).