Molecular characterisation of a membrane-bound galactosyltransferase of plant cell wall matrix polysaccharide biosynthesis


For correspondence (fax +44 1786 464994; e-mail


Galactomannan biosynthesis in vitro is catalysed by membrane preparations from developing fenugreek seed endosperms. Two enzymes interact: a GDP-mannose dependent (1→4)-β-d-mannan synthase and a UDP-galactose dependent (1→6)-α-d-galactosyltransferase. The statistical distribution of galactosyl substituents along the mannan backbone, and the degree of galactose substitution of the primary product of galactomannan biosynthesis appear to be regulated by the specificity of the galactosyltransferase. We now report the detergent solubilisation of the fenugreek galactosyltransferase with retention of activity, the identification on gels of a putative 51 kDa galactosyltransferase protein, and the isolation, cloning and sequencing of the corresponding cDNA. The solubilised galactosyltransferase has an absolute requirement for added acceptor substrates. Beta-(1→4)-linked d-manno-oligosaccharides with chain lengths greater than or equal to 5 acted as acceptors, as did galactomannans of low to medium galactose-substitution. The putative galactosyltransferase cDNA encodes a 51282 Da protein, with a single transmembrane alpha helix near the N terminus. We have also confirmed the identity of the galactosyltransferase by inserting the cDNA in frame into the genome of the methylotrophic yeast Pichia pastoris under the control of an AOX promoter and the yeast alpha secretion factor and observing the secretion of galactomannan α-galactosyltransferase activity. Particularly high activities were observed when a truncated sequence, lacking the membrane-spanning helix, was expressed.


Primary plant cell walls are composed of a matrix of polysaccharides which provide structural strength, are central to the control of growth and provide a shield against pathogen attack. The cell wall is a major source of renewable raw materials and its properties underlie many of the characteristic eating properties of plant-based foods. Cell wall polysaccharides are assembled by the action of sugar nucleotide dependent glycosyltransferases attached (apart from cellulose synthases) to Golgi membranes. Despite the importance of plant cell wall synthesis, none of the glycosyltransferases have been purified and characterised at the molecular level (but see Note added in proof).

Galactomannans (α-(1→6)-galactosyl substituted (1→4)-β-d-mannans) are dominant constituents of the cell walls of legume-seed endosperms and closely similar molecules are components of many other cell walls (Reid 1997). Legume seed galactomannans are used in the food and other industries (Reid & Edwards 1995), and commercial functionality is affected significantly by galactose substitution (Dea & Morrison 1975). Galactosyltransferase specificity has been identified as a key factor in the regulation of galactose-substitution in galactomannan biosynthesis (Reid et al. 1995). We now report on the molecular characterisation of the membrane-bound α-galactosyltransferase of cell wall galactomannan biosynthesis in the fenugreek (Trigonella foenum-graecum L) seed endosperm.

At a certain stage of seed development, galactomannan biosynthesis is a major metabolic activity in the legume-seed endosperm (Edwards et al. 1992; Meier & Reid 1977; Reid & Meier 1970). Thus, membrane preparations from endosperms hand-dissected at that stage are highly active in catalysing galactomannan biosynthesis in vitro from the sugar nucleotide precursors GDP-Man and UDP-Gal (Edwards et al. 1989; Reid et al. 1995). Two enzymes interact – a (1→4)-β-d-mannan synthase and a specific (1→6)-α-d-galactosyltransferase. The mannan synthase catalyses the stepwise transfer of mannosyl residues to the growing non-reducing end of a mannan chain and the galactosyltransferase the transfer of galactose residues to a specific galactosylacceptor mannose residue at or near the growing mannan chain end (Edwards et al. 1989). A critical factor in the unambiguous demonstration of galactomannan biosynthesis was the ability to determine the extent and statistical distribution of galactosylsubstitution in labelled in vitro product polysaccharides. This was achieved by enzymatic fragmentation analysis (Edwards et al. 1989; Reid et al. 1995) using a pure structure-sensitive endo-(1→4)-β-d-mannanase (McCleary & Matheson 1983; McCleary & Matheson 1986) releasing diagnostic oligosaccharides from galactomannans.


When membrane preparations from developing fenugreek seed endosperms were treated with a range of detergents, only digitonin, Triton X-100 and NP-40 gave high galactosyltransferase activity in 100 000 g supernatants. Detection of galactosyltransferase activity after solubilisation was absolutely dependent upon the addition of acceptor substrates. Manno-oligosaccharides (1→4)-β-linked with five or more mannose residues acted as acceptors, the acceptor efficiency increasing with chain length up to the limits of solubility of the oligosaccharides (Table 1). Treatment of the product oligosaccharides with a pure α-galactosidase from germinated guar seeds (McCleary 1983) showed that the galactosyl linkages were α. The galactomannans from locust bean (Ceratonia siliqua) and guar (Cyamopsis tetragonoloba) also acted as acceptor substrates, and enzymatic fragmentation analysis showed that the newly formed galactosyl linkages were (1→6)-α. Locust bean galactomannan (low galactose-substitution; Man/Gal = 3.5) was a more efficient acceptor substrate than guar galactomannan (medium galactose-substitution; Man/Gal = 1.6). The galactomannan of fenugreek (high galactose-substitution; Man/Gal = 1.1) did not act as an acceptor. The positional specificity of transfer to acceptors will be the subject of a later communication.

Table 1.  Transfer of 14C-galactose residues from UDP-14C-Gal to manno-oligosaccharides
Oligosaccharide chain lengthGalactosyltransfer (μm)
  • Incubation: 2 h, 30°. UDP-Gal: 0.8 mm. Oligosaccharides: 1 mm.

  • *

    The nona- and decasaccharides were fully soluble only on heating. Some precipitation could have occurred during incubation, thus lowering the effective acceptor concentration.


Conventional purifications of the detergent solubilised galactosyltransferase were not possible due to low tissue amounts. Consequently, a variety of correlative methods were used to associate galactosyltransferase activity with particular protein bands on gels. Although all approaches indicated an association with a protein of Mr about 50K, the most decisive was a combination of non-denaturing isoelectric focussing and SDS-PAGE. With Triton X-100 solubilised preparations, this (Fig. 1) gave an exact correlation between activity and a protein band with apparent Mr = 51K and pI = 6.5. N-terminal and internal sequence data were obtained from the protein, and degenerate oligonucleotide primers (Fig. 2) were designed. RNA was obtained from hand-isolated fenugreek endosperms actively synthesising galactomannan, and RT-PCR techniques were used to obtain two partial, overlapping cDNA clones. One (approximately 1000 bp) was amplified using N-terminal primer NTP2S and internal primer GT3A4 (Fig. 2) and the other (approximately 500 bp) was obtained by 3′-RACE PCR using internal primer GT3S4. The N-terminal primer was designed to the extreme terminus of the protein. A single clone of about 1500 bp, with the same sequence as the composite, was obtained by RT-PCR using Pfu polymerase and primers designed to the extreme 5′ terminus of the 1000 bp clone (5′ GCGACGAAATTTGGTTCCAA 3′) and the 3′ terminus of the 500 bp clone (5′ GCTAAT- ATCATCACCACCTTC 3′). The sequence (Fig. 3) had an open reading frame of 1314 bp which encoded a 438 aa protein of predicted Mr = 51282 and pI = 6.65. The encoded protein included all the N-terminal and internal peptide sequences obtained by direct sequencing of the 51 K protein band. The absence of a methionine residue at the N-terminus of the purified protein indicated post-translational modification. Using 5′ RACE-PCR, a small amount of cDNA sequence was obtained which overlapped with the 5′ end of the sequence in Fig. 3 and extended beyond it. This showed clearly that the N-terminal alanine of the mature protein was immediately preceded by a methionine residue, which was presumably the start of translation. Secondary structure calculations (Rost & Sander 1993; Rost & Sander 1994; Rost et al. 1995) predicted, with high probability, a single hydrophobic membrane-spanning α-helix near the N-terminal end of the protein (double-underlined in Fig. 3). Many Golgi proteins, including some glycoprotein glycosyltransferases, are membrane-anchored in this way (Paulson & Colley 1989). Protein database searching revealed weak homologies with known and putative galactosyltransferases from the fission yeast Schizosaccharomyces pombe. Notably there was 24% similarity at the amino acid level over a 103 amino acid overlap with an α-(1→2)-galactosyltransferase known to be involved in the O-linked oligosaccharide modification of proteins transported through the Golgi stack, and to transfer galactose from UDP-Gal to a variety of mannose-based acceptors (Chappell et al. 1994).

Figure 1.

Separation of Triton X-100 solubilised enzyme by isoelectric focussing in a non-denaturing agarose gel.

Correlation of galactosyltransferase activity with pH gradient shows maximum activity at pH 6.5. Alignment of activity with proteins separated in a second dimension by SDS-PAGE shows exact correlation between the peak of enzyme activity and a protein band of apparent molecular weight 51 K.

Figure 2.

Design of degenerate oligonucleotide primers to protein sequence obtained from the 51 K protein.

Primer NTP2S was a 16-mer sense primer designed to the extreme N-teminal protein sequence. Primers GT3S4 and GT3A4 were sense and antisense primers designed to an internal peptide.

Figure 3.

Sequence of the full c1400 bp cDNA clone and deduced protein sequence.

N-terminal and internal protein sequence obtained by direct sequencing is underlined singly. The predicted membrane-spanning α-helix is underlined doubly.

Although the cDNA clone clearly corresponded to the 51 K protein and seemed likely to encode a galactosyltransferase, there was no direct evidence that the encoded sequence was indeed the galactomannan galactosyltransferase. To obtain this, an attempt was made to obtain expression of an enzymatically active protein in the methylotrophic yeast Pichia pastoris. In recognition that the full length putative galactosyltransferase protein might become membrane-associated in Pichia and thus poorly expressed and/or secreted, two different sequences were inserted in frame into the Pichia genome, under the control of an alcohol oxidase promoter and the Saccharomycesα-secretion factor. One comprised the entire 1314 bp open reading frame, namely bp 1–1314 in Fig. 3, and encoded the entire putative galactosyltransferase sequence. The second comprised a shorter sequence, bp 127–1314 in Fig. 3, which encoded a protein lacking the extreme N-terminal sequence and membrane-spanning helix. Successful transformation was obtained using constructs of both types. When galactomannan galactosyltransferase was assayed in 2 day culture media, control cultures showed no activity, cultures containing the full length construct showed low activity, and those containing truncated constructs showed very high levels of activity comparable with those in 100 000 g supernatants of detergent solubilised fenugreek membrane preparations (Table 2). In the assays, the UDP-Gal donor substrate was labelled with 14C in the galactosyl moiety, and the acceptor substrate was the low-galactose galactomannan from locust bean. When the labelled products of the transfer reaction were treated with the structure-sensitive endo-β-mannanase (McCleary & Matheson 1983) from A. niger, the only labelled digestion products were galactomanno-oligosaccharides diagnostic of legume seed galactomannans (Fig. 4).

Table 2.  Galactomannan galactosyltransferase activities in 10× concentrated 48 h culture supernatants from Pichia transformants, in relation to the activity in a typical Triton X-100 extract of fenugreek membranes (not concentrated)
SampleActivity (μmol l–1 h–1)Activity (relative to fenugreek membranes)
Triton X-100 extract10.41.00
(fenugreek membranes)
Supernatant, colony 81.60.015
(full-length insert)
Supernatant, colony 234.20.041
(full-length insert)
Supernatant, colony 2794.90.91
(truncated insert)
Supernatant, colony 291161.11
(truncated extract)
Supernatant, pPIC90.110.001
transformation (no insert)
Figure 4.

Digital autoradiogram of TLC-separated oligosaccharides released on enzyme digestion of polysaccharide product formed in galactosyltransferase assay of culture supernatant from recombinant Pichia transformed with truncated construct.

On digestion with A. niger endo-β-mannanase alone (outer lanes), the only labelled products co-migrated with galactomanno-oligosaccharides diagnostic of legume seed galactomannans (Edwards et al. 1989; Reid et al. 1995). When a trace of pure α-galactosidase from guar seeds was added, galactose was released (inner lane) confirming that the galactosyl link was α. M2G = galactosylmannobiose, M3G = galactosylmannotriose, M5G2 = digalactosylmannopentaose, O = galactomannooctasaccharides, N = galactomannononasaccharides, H = higher galactomannooligo- saccharides (Edwards et al. 1989; Reid et al. 1995).


To the best of our knowledge this is the first membrane-bound glycosyltransferase involved in the biosynthesis of non-cellulosic plant cell wall polysaccharides to be identified unequivocally and characterised at the molecular level (but see Note added in proof). Nor has any catalytically active protein product of the recently identified putative plant cellulose synthase (Pear et al. 1996) genes been reported. In the context of the galactomannans, this new galactosyltransferase is a key enzyme in the regulation of galactose content and distribution during biosynthesis (Reid et al. 1995). Its characterisation presents new opportunities to investigate the molecular basis of the biosynthetic control mechanisms responsible for the differences in galactomannan structure between different legume species. It also offers clear experimental strategies for the manipulation of crop legumes to produce improved ‘designer’ galactomannan gums. In the wider context of the biosynthesis of complex plant cell wall polysaccharides, the characterisation of this enzyme demonstrates that, despite many years of frustratingly slow progress, the membrane-bound glycosyltransferases of complex cell wall polysaccharide biosynthesis are amenable to purification and characterisation using direct biochemical approaches. Using related methods, and experimental systems in which in vivo synthesis rates are high, new classes of glycosyltransferase involved in pectin and hemicellulose biosynthesis will almost certainly now be purified and characterised. Given such new sequence information, genetic approaches will swiftly lead to the multiplication of new sequences encoding cell wall polysaccharide glycosyl transferases, to the alteration of activities in planta of key glycosyltransferase enzymes, and to plants with cell wall polysaccharide compositions altered in specific, predictable ways. In our view this will help overcome current relative ignorance of how alterations of cell wall molecular composition translate into changes in wall rheology and tissue properties, and thus hasten advances in understanding the role of the cell wall in key processes such as cell elongation, plant resistance to pathogenesis and the control of vegetable and fruit textural properties.

Experimental procedures

Membrane preparations

Fenugreek plants were grown (Edwards et al. 1992) and endosperms from developing fenugreek seeds actively synthesising galactomannan were obtained and homogenised (Edwards et al. 1989) as before. Membranes pelleting between 13 000 g and 100 000 g were retained.

Detergent treatment

Membranes were resuspended (homogeniser) in 100 mm Tris–HCl buffer pH 7.5 (12.5 μl per endosperm) containing EDTA (2 mm) and DTT (10 mm), mixed with an equal volume of 2% (v/v) detergent, placed on ice, homogenised briefly every 10 min for 30 min and spun at 100 000 g for 1 h.


Assays for membrane-bound galactosyltransferase have been described previously (Edwards et al. 1989). After detergent solubilisation, detection of galactosyltransferase activity required the addition of manno-oligosaccharides or galactomannans with relatively low degrees of galactose substitution. The products of transfer to manno-oligosaccharides were of too low molecular weight to be precipitable with ethanol. Incubation mixtures were therefore passed, by centrifugation, through short columns (0.5 ml microcentrifuge tubes pierced at the bottom) packed with DEAE-cellulose (200 μl) to remove most of the negatively charged sugar nucleotide substrate. The column effluent was freeze dried and taken up in a known volume of water (50 or 100 μl). Aliquots were subjected to scintillation counting. Further aliquots were separated on TLC (Reid et al. 1995) and the plates were subjected to quantitative digital autoradiography (Reid et al. 1995). The relative amounts of label associated with oligosaccharide reaction product and residual sugar nucleotide substrate were calculated from the autoradiographic data and used to correct the results of the scintillation counting.

Separation of Triton X-100-solubilised proteins by non-denaturing IEF

IEF gels (8 × 10 cm) were assembled as described in the Hoefer Technical bulletin no. 134, and cast using the Hoefer gel-caster SE245. The separation gel was prepared by mixing IsoGel agarose (FMC Bioproducts; 120 mg), sorbitol (2.4 g) and water (10.36 ml) and heating at 100°C for 10 min. After cooling to 65°C, 600 μl of a 2% (w/v) solution of TX-100 (Boehringer 789 704) and 600 μl of ampholyte mixture (a 4 : 1 vol:vol. mixture of pH 5.0–8.0 Ampholine, Sigma A5799 and pH 3.5–10.0 Ampholine, Sigma A5174) were added. Gels (6.7 cm wide sample well and 0.5 cm wide reference well alongside), were run in a Hoefer SE 250 vertical gel apparatus at 4°C. The sample consisted of 750 μl of detergent extract to which 45 μl of ampholyte mixture, 65 μl glycerol and 5 μl bromophenol blue (0.05% w/v in water) had been added. An overlay (40 μl of the ampholyte mixture, 40 μl of 2% detergent, 40 μl glycerol, 5 μl of bromophenol blue and 680 μl water) prevented direct mixing of the sample and the cathode buffer. The reference well contained coloured IEF standards (Bio-Rad; 2.5 μl) and overlay. Cathode and anode buffers were 20 mm NaOH and 6 mm phosphoric acid, respectively. After focussing at 200 V for 30 min and 600 V for 60 min the gel, attached to GelBond (FMC Bioproducts) was cut into strips parallel to the direction of current flow. The end strips were fixed and stained with Coomassie Blue. Separation of the focussed proteins in a second dimension was carried out by placing an entire gel strip (on GelBond) in the sample well of a 10% SDS gel (10 × 12 cm).

Localisation of galactosyltransferase activity in IEF gels

Isogel agarose contains galactomannan which we showed, by enzymatic fragmentation analysis, to be of the low galactose type, and thus capable of acting, in situ, as an acceptor substrate for galactosyltransferase. A complete gel strip (on GelBond) was equilibrated in 200 mm Tris–HCl pH 7.5 for 10 min and then incubated in a mixture containing 50 mm Tris–HCl pH 7.5, 10 mm MnCl2, 0.1% (w/v) Triton X-100 and 800 μm14C-labelled UDP-Gal for 3 h at 30°C. The strip was then fixed in 40% (v/v) methanol/10% (v/v) acetic acid for 20 min and washed overnight in 40% methanol to remove unincorporated label with retention of the labelled galactomannan product within the gel. The washed gel strip was cut into 2 mm sections. Each section was removed from the GelBond, washed extensively with 40% methanol, dissolved in concentrated HCl (20 μl) and subjected to liquid scintillation counting.

Protein sequence from the 51K band

The material focussing at pI 6.2–6.8 was excised from an entire IEF gel and the gel sections were applied as the sample to an SDS-PAGE gel. To obtain N-terminal sequence, the gel was blotted onto Problott membrane (Applied Biosystems), the blot was stained lightly with Coomassie blue and the 51K band was excised for sequencing. Internal sequence information was obtained by proteinase digestion in-gel and separation of the peptides either by SDS-PAGE (Cleveland et al. 1977) or by HPLC (Rosenfeld et al. 1992).

cDNA cloning

Endosperms from 100 seeds (approximately 1 g) from pods harvested 32–35 days after anthesis were frozen (liquid N2) and ground in a mortar and pestle with liquid N2. RNA was prepared (Lopez-Gomez & Gomez-Lim 1992) and used in RT-PCR and RACE protocols (Frohman & Martin 1989). Amplified bands were cloned using the Original TA cloning kit (Invitrogen). When a proofreading DNA polymerase was used for amplification (Pfu, Stratagene) the 3′A-overhangs necessary for TA cloning were added by incubation with Taq polymerase and dATP prior to ligation.

Pichia expression

Primers GTEXP1S (5′ GTATCTCTCGAGAAAAGAGCGACGAAAT- TTGGTTCCAAA3′), GTEXP2S(5′GTATCTCTCGAGAAAAGAAA CTCCAACCCAAAATTCAAC3′) and GTEXP3A(5′ TTAATTCGC- GGCCGCCCTTTATGGTGATGCAGCGGGGTA 3′) were designed to allow PCR amplification of the entire galactosyltransferase sequence and of a truncated sequence with sequence extensions allowing cloning in-frame in the multiple cloning site of the Pichia expression vector pPIC9 (Invitrogen), using the Xho1 and Not1 restriction sites (italicised on primer sequences). Using plasmid DNA with the full-length sequence (Fig. 3) as template, primers GTEXP1S and GTEXP3A amplified a c1400 bp band which was cloned into pPIC9. Ampicillin-resistant clones were screened by PCR and by digestion with Xho1 and Not1. Primers GTEXP2S and GTEXP3A amplified an approximately 1300 bp fragment which was similarly treated. The pPIC9 constructs with the full-length and truncated sequences were linearised with Stu1. Competent cells of Pichia pastoris GS115 (Invitrogen) were prepared and transformed using the EasyComp (Invitrogen) kit. Putative transformants were selected by their ability to grow on histidine-free medium and were further screened by direct PCR amplification of colonies.

Assay for galactomannan galactosyltransferase activity associated with Pichia transformants

Single colonies of putative positive transformants were inoculated into 10 ml of BMGY (no methanol) medium in 50 ml conical tubes and grown at 30°C with continuous rotatory shaking (200 rpm) for 24 h (A600 about 2.7). Cells were harvested by centrifugation and resuspended in BMMY (containing methanol) medium to give an A600 value of 1.0. Samples (50 ml) were then cultured at 30°C for up to 3 days. Supernatants were collected, concentrated (×10) using Vivapore (Vivascience) membrane concentrators (7.5 K cut-off), and assayed for galactosyltransferase activity using locust bean galactomannan as galactosylacceptor. The assays (100 μl) contained 50 μl concentrated supernatant, 25 mm Tris–HCl buffer pH 7.5, 2 mm MnCl2, 0.2% (w/v) locust bean galactomannan and 800 μm labelled UDP-Gal, and were incubated at 30°C for 2 h. At the end of the incubation time glacial acetic acid (50 μl) was added and the mixture was heated at 100°C for 2 min. The galactomannan acceptor was precipitated by adding methanol to a final concentration of 70%, washed exhaustively with hot 70% methanol and either subjected to liquid scintillation counting or enzymatic fragmentation analysis.


This work was carried out with the support of a co-operative research grant awarded by the Biotechnology and Biological Sciences Research Council (BBSRC) UK.


  1. EMBL, GenBank and DDBJ nucleotide sequence database accession number AJ245478.