• Open Access

Expression of rat β(1,4)-N-acetylglucosaminyltransferase III in Nicotiana tabacum remodels the plant-specific N-glycosylation


* Correspondence (fax +41 44 632 11 48; e-mail alexander.frey@micro.biol.ethz.ch)


Plant N-linked glycans differ substantially from their mammalian counterparts, mainly with respect to modifications of the core glycan, which typically contains a β(1,2)-xylose and an α(1,3)-fucose. The addition of a bisecting N-acetylglucosamine residue by β(1,4)-N-acetylglucosaminyltransferase III (GnTIII) is known to control the processing of N-linked glycans in mammals, for example by preventing α(1,6)-fucosylation of the core glycan. In order to outcompete plant-specific β(1,2)-xylose and α(1,3)-fucose modifications, rat GnTIII was expressed either with its native localization domain (GnTIII) or with the cytoplasmic tail, transmembrane domain and stem region (CTS) of Arabidopsis thaliana mannosidase II (ManII) (GnTIIIA.th.). Both CTSs targeted enhanced yellow fluorescent protein (eYFP) to a brefeldin A-sensitive compartment, indicative of Golgi localization. GnTIII expression increased the fraction of N-glycans devoid of xylose and fucose from 13% ± 7% in wild-type plants to 60% ± 8% in plants expressing GnTIIIA.th.. N-Glycans of plants expressing rat GnTIII contained three major glycan structures of complex bisected, complex, or hybrid bisected type, accounting for 70%–85% of the total N-glycans. On expression of GnTIIIA.th., N-glycans displayed a higher heterogeneity and were of hybrid type. Co-expression of A. thaliana ManII significantly increased the amount of complex bisected structures relative to the plants expressing GnTIII or GnTIIIA.th., whereas co-expression of human ManII did not redirect the pool of hybrid structures towards complex-type structures. The method described offers the advantage that it can be implemented in any desired plant system for effective removal of β(1,2)-xylose and α(1,3)-fucose from the N-glycan.


N-Glycosylation is a key determinant for the production of recombinant therapeutic proteins, as it influences factors such as stability, activity, pharmacokinetics, targeting and immunogenicity (Varki, 1993). Plant N-glycosylation differs from its mammalian counterpart with regard to the processing steps taking place in the Golgi apparatus (Lerouge et al., 1998; Faye et al., 2005; Saint-Jore-Dupas et al., 2007). These differences include the absence of β(1,4)-galactose, terminal sialic acid and core α(1,6)-fucose, but the incorporation of α(1,3)-fucose and β(1,2)-xylose residues in the core N-glycan, and, to a low degree, a terminal Lewis A epitope at one or two of the antennae of the plant N-glycans. The two modifications of the core glycan are generally considered to represent the main obstacle to the use of plants for the production of glycosylated human therapeutic proteins, as these two epitopes are known to induce the formation of antibodies (Bardor et al., 2003). Despite these difficulties, several recent attempts have demonstrated the feasibility of overcoming these limitations and of modifying the N-linked protein glycosylation pathway in plants accordingly. The strategies pursued so far have aimed to either introduce foreign enzymatic activities into plants using enzymes such as α(1,4)-galactosyltransferase or β(1,4)-N-acetylglucosaminyltransferase III (GnTIII), or to reduce/eliminate the activities of α(1,3)-fucosyltransferase [α(1,3)-FucT] and β(1,2)-xylosyltransferase [β(1,2)-XylT] (Palacpac et al., 1999; Strasser et al., 2004; Huether et al., 2005; Cox et al., 2006; Rouwendal et al., 2007; Sourrouille et al., 2008; Strasser et al., 2008). Using the latter approach, either randomly created knockout plants were selected, or specific enzymes were directly inactivated using RNA interference (RNAi). In this respect, the moss Physcomitrella patens has been shown to be a very useful production system, as its genome is amenable to directed knockouts of enzymes in the N-glycosylation machinery (Koprivova et al., 2004). Furthermore, RNAi technology has been used to reduce the enzyme activities of α(1,3)-FucT and β(1,2)-XylT in the aquatic plant Lemna minor, Nicotiana benthamiana and alfalfa (Cox et al., 2006; Sourrouille et al., 2008; Strasser et al., 2008).

An alternative approach to overcome the limitations of therapeutic protein production in plants imposed by plant-specific N-glycosylation is the suppression of the activities of α(1,3)-FucT and β(1,2)-XylT. Such a strategy aims to modify the N-glycans in such a manner that they are no longer accepted substrates for these two enzymes. In mammalian cells, but not in plants, GnTIII catalyses the introduction of a bisecting N-acetylglucosamine (GlcNAc) residue to the core glycan (Narasimhan, 1982). The presence of the bisecting GlcNAc prevents the action of α(1,6)-FucT, Golgi-resident mannosidase II (ManII) and GnTII in mammalian cells. Therefore, GnTIII exerts a high degree of control over the N-glycosylation in the Golgi apparatus (Schachter et al., 1983).

The over-expression of GnTIII has been shown to catalyse efficiently the introduction of GlcNAc residues into mammalian cells (Sburlati et al., 1998). Furthermore, on titration of GnTIII expression levels using a repressible promoter, the highest GnTIII expression led to the accumulation of non-fucosylated hybrid bisected products (Umaña et al., 1999). It may be speculated that the bisecting GlcNAc residue introduced by GnTIII may exhibit a similar function in plant cells. Support for this hypothesis comes from the work of Bencur et al. (2005), which showed that the presence of branching sugars, such as β(1,2)-xylose or α(1,3)-fucose, at the core inhibits the activities of core-modifying enzymes (Bencur et al., 2005). Preliminary data indicated that rat GnTIII can be functionally expressed in suspension cultures and hairy root cultures derived from Nicotiana tabacum SR1, as judged by the appearance of a bisecting GlcNAc in plant N-glycans (Karg, 2002). In a recent study, in which Arabidopsis thalianaβ(1,2)-XylT and α(1,3)-FucT were expressed in CHO cells, and tested for their activity with various substrates, both enzymes were strongly inhibited by the presence of the bisecting GlcNAc residue. Despite the insertion of a bisecting GlcNAc with high efficiency into plant N-glycans, the expression of human GnTIII did not alter the xylose levels, but produced a small decrease in the fucose levels (Rouwendal et al., 2007). In order to compete with the endogenous glycosylation machinery, the correct spatial distribution in the Golgi complex and hence the correct temporal order for its action are prerequisites. Experiments by Ferrara et al. (2006) elegantly highlighted this issue, showing that the exchange of the cytosolic, transmembrane and stem region (CTS) of GnTIII with a series of CTSs from glycosyltransferases, or the CTS from ManII, produced different effects on α(1,6)-fucosylation, the fusion of GnTIII to ManII CTS being the most effective in competing with the α(1,6)-FucT in CHO cells. The reduction of α(1,6)-fucose coincided with a shift of the N-glycans towards a prevalence of hybrid bisected structures. Co-expression of the Golgi-resident ManII with GnTIII led to a shift of the N-glycans towards mainly non-fucosylated complex bisected structures (Ferrara et al., 2006).

In this study, the expression of rat GnTIII in N. tabacum SR1 plants is described, together with the effect of its targeted expression, using the A. thaliana ManII CTS, on N-glycosylation, the ultimate goal being to generate fucose- and xylose-deficient N-glycans. Furthermore, the effect of the simultaneous co-expression of human or A. thaliana ManII was studied. To that end, vectors were created for the concomitant expression of rat GnTIII and ManII genes, using a strong promoter derived from Agrobacterium tumefaciens opine genes. Both CTSs directed an enhanced yellow fluorescent protein (eYFP) fusion to punctate structures in the cells. Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) analysis of glycoengineered plants revealed a complete redistribution of the endogenous plant N-glycans and a strong decrease in the level of xylosylation and fucosylation of plant N-glycans in plants expressing rat GnTIII.


Generation of binary expression vectors for high expression levels

Titration studies in mammalian cells have shown a positive correlation between the expression level of GnTIII and a reduction in α(1,6)-fucosylation (Umaña et al., 1999), indicating that high expression levels are needed to reduce the modification of the core N-glycan. Therefore, the ability of several previously described viral promoters isolated from figwort mosaic virus, carnation-etched ring virus and peanut chlorotic streak virus (Maiti et al., 1997; Bhattacharyya et al., 2002, 2003; Jobling et al., 2003) to drive high levels of protein expression were evaluated using β-glucuronidase as a reporter. However, none of the tested promoters displayed significantly higher activity than the cauliflower mosaic virus (CaMV) 35S promoter (data not shown). Therefore, a chimeric A. tumefaciens-derived promoter comprising two or three repeats of the upstream activating sequence (UAS) of the octopine synthase promoter juxtaposed to the mannopine synthase promoter (Pmas) was devised to drive high levels of expression (Ni et al., 1995). These constructs were stably expressed in N. tabacum Bright Yellow-2 (BY-2) cells. Four to six individual clones were assayed for promoter activity. The promoter containing three UAS repeats fused to Pmas, termed UAS123mas, drove the highest β-glucuronidase expression levels, reaching, on average, a 30-fold higher reporter activity relative to the widely used CaMV 35S promoter (Figure 1). However, clonal variations of expression levels were obvious, which probably reflect differences in the site of integration or the number of integrated transgene copies. In order to reduce the effect of the integration site on gene expression, a synthetic matrix attachment region (MAR) was included upstream of the expression constructs (Van Der Geest et al., 2004).

Figure 1.

Evaluation of promoter activities using β-glucuronidase as reporter. Cauliflower mosaic virus (CaMV) 35S-β-glucuronidase (CaMV 35S), UAS13mas-β-glucuronidase (UAS13mas) and UAS123mas-β-glucuronidase (UAS123mas) constructs were stably transformed into Nicotiana tabacum Bright Yellow-2 (BY-2) cells. The promoter strength was measured using the fluorigenic substrate 4-methyl umbelliferyl glucuronide (MUG). Non-transformed wild-type BY-2 cells were used as controls (wt). Four to six clones were analysed for reporter activity. β-Glucuronidase activities of individual measurements, as well as mean values and standard deviations, are reported. RLU, relative light unit.

Based on these DNA elements, a set of six expression cassettes was created encoding GnTIII, either in its native configuration (GnTIII) or as a fusion of the A. thaliana ManII CTS, as described by Strasser et al. (2006), with the catalytic domain of GnTIII (GnTIIIA.th.) (Figure 2a). GnTIII was expressed alone or in combination with human ManII, with either its native CTS or the corresponding CTS of A. thaliana ManII, or the full-length A. thaliana ManII (Figure 2b). The binary expression vectors encoded native GnTIII (pAX67), GnTIIIA.th. (pAX70), native GnTIII together with ManII (pAX73), GnTIIIA.th. together with ManIIA.th. (pAX74) and A. thaliana ManII with rat GnTIII (pAX100) or GnTIIIA.th. (pAX101) (Figure 2c,d).

Figure 2.

(a) Schematic representation of rat β(1,4)-N-acetylglucosaminyltransferase III (GnTIII) and the hybrid enzyme GnTIIIA.th. GnTIIIA.th. is a fusion of amino acids 1–92 of the Arabidopsis thaliana mannosidase II (ManII), comprising the cytosolic tail, transmembrane domain and stem region (CTS), to amino acids 77–535 of rat GnTIII. A C-terminal Myc-tag is appended to both versions of the protein. (b) Schematic representation of human ManII, the hybrid enzyme ManIIA.th. and A. thaliana ManII. ManIIA.th. is a fusion of amino acids 1–92 of the A. thaliana ManII (CTS) to amino acids 110–1143 of human ManII. (c) Diagram of the T-DNA portion of the GnTIII expression vectors. The T-DNA contains a hygromycin resistance marker under the control of the nopaline synthase (nos) promoter (Pnos). GnTIII expression is driven by three repeats of the upstream activating sequence from octopine synthase (UAS) fused to the promoter of mannopine synthase (Pmas). An artificial matrix attachment region (MAR), composed of two identical juxtaposed units, is included upstream of the promoter. Restriction sites for the insertion of ManII expression cassettes are indicated. (d) Expression cassette driving the expression of ManII. The expression cassette is present in the GnTIII-expressing T-DNA as an SbfI, BamHI fragment, and the expression of ManII is under the control of the identical promoter. (e, f) Confocal laser scanning microscopy of BY-2 cells expressing enhanced yellow fluorescent protein (eYFP) fused to the CTS of rat GnTIII (e) or the CTS of A. thaliana ManII (f). Dot-like structures are visible for both fusion constructs (left panels), which disappear on treatment of the cells with brefeldin A (BFA) (right panels), suggesting that the punctate, brefeldin A-sensitive structures represent the Golgi apparatus. Schematic representations of the protein fusions used for confocal laser scanning microscopy are depicted above each panel.

In order to verify the localization of the enzymes to the secretory pathway, confocal laser scanning microscopy was performed. Fusion constructs comprising the N-terminal transmembrane domain and stem region of rat GnTIII or of A. thaliana ManII fused to eYFP were stably transformed into N. tabacum BY-2 cells (Figure 2e). Both N-terminal CTS sequences target the fluorescence to intracellular dot-like structures (Figure 2e,f, left panels), which are typically observed on expression of Golgi-localized proteins, such as the plant β(1,2)-XylT (Saint-Jore-Dupas et al., 2006). To more thoroughly verify these structures, the cells were treated with the Golgi destabilizing agent brefeldin A. On treatment of these cells with brefeldin A, the punctate structures disappeared and the fluorescence signal became diffuse (Figure 2e,f, right panels). These findings indicate that the N-terminal sequence of rat GnTIII is functional in N. tabacum BY-2 cells and targets the protein to a brefeldin A-sensitive compartment. Similarly, the N-terminal sequences of A. thaliana ManII target the reporter to a similar punctate, brefeldin A-sensitive structure, as observed previously on expression in A. thaliana (Strasser et al., 2006).

Generation of GnTIII-expressing N. tabacum plants

Using agrobacteria, 20, 20, 19, two, eight and two independent plant lines were generated expressing GnTIII, GnTIIIA.th., GnTIII and ManII, GnTIIIA.th. with ManIIA.th., GnTIII with A. thaliana ManII and GnTIIIA.th. with A. thaliana ManII, respectively. From these regenerated plantlets, 65%–100% were positive for GnTIII expression according to immunoblotting (data not shown). From the available plant lines, several were propagated further in order to select for homozygous N. tabacum lines: plants expressing native GnTIII (line NG20); two independent transformants expressing GnTIIIA.th. (RG2 and RG3); plants expressing native GnTIII together with ManII (NGM6); and those expressing GnTIIIA.th. in combination with ManIIA.th. (RGM1). Furthermore, from the plants expressing A. thaliana ManII in combination with native GnTIII (NGAM2) or GnTIIIA.th. (RGAM1), a single line from each was propagated. Seeds harvested from these plants were tested for homozygotes.

A key goal of this study was to find an answer to the question of whether the expression of GnTIII could outcompete the activities of α(1,3)-FucT and β(1,2)-XylT. To confirm the effect of GnTIII expression on N-glycan composition, plant material was analysed using an anti-horseradish peroxidase (anti-HRP) antibody. This antibody and affinity-purified fractions thereof have been used routinely to probe for the presence of α(1,3)-fucose and/or β(1,2)-xylose on the core glycan, α(1,3)-fucose representing the major antibody-eliciting epitope (Faye et al., 1993). Samples prepared from GnTIII-expressing plant material were less reactive to the anti-HRP antibody, relative to the samples of the wild-type plants, indicating that the expression of GnTIII can reduce α(1,3)-fucose and/or β(1,2)-xylose (Figure 3). Although a strong effect on N-glycosylation was observed, no clear correlation between the level of GnTIII expression and the effect on the xylose/fucose level was found. When probing the extract with an anti-GnTIII serum, a band of appropriate size and a prominent band of approximately 50 kDa were observed, which also reacted with an antibody specific for the C-terminal Myc-tag, indicating that the N-terminus of the protein is processed (data not shown). This processing of rat GnTIII has been described previously, and resulted in the generation of a 52-kDa fragment (Nishikawa et al., 1992).

Figure 3.

Screening of primary transformants using Western blot analysis. β(1,4)-N-Acetylglucosaminyltransferase III (GnTIII) expression was detected using a serum raised against the luminal part of rat GnTIII. An additional band (GnTIII*), which is cross-reactive with the anti-GnTIII serum and an anti-c-Myc antibody, represents the C-terminal cleavage product of GnTIII. The bottom panel represents a Western blot of the identical sample using an anti-horseradish peroxidase (HRP) antibody, revealing the dramatic effects of GnTIII expression on N-glycosylation.

From the screened plant lines, several were propagated further and the N-glycans were analysed using MALDI-TOF-MS. In general, plant N-glycans of N. tabacum were composed of a mixture of mainly paucimannosidic structures (60%), and similar amounts of high-mannose, hybrid and complex sugars, each contributing roughly 10% to the total N-glycan pool (Table 1). In the glycoengineered plants, bisected versions of hybrid and complex sugars were also observed. The plant lines expressing GnTIII or GnTIIIA.th. produced a distinct glycan pattern, which also affected the amount of xylose and fucose on the core N-glycan. The expression of GnTIIIA.th. had a more dramatic effect on xylosylation and fucosylation of the core glycan than the expression of rat GnTIII (Table 1). Depending on the line, between 31% and 59% of the structures were devoid of both modifications of the N-glycan, whereas, in the case of GnTIII, the amount of structures lacking both fucose and xylose (15% and 7%) was not significantly different from that of wild-type plants (13% ± 7%). However, notably, in wild-type plants, the structures devoid of core xylosylation and fucosylation were of high-mannose type. Structures containing only xylose (Xyl1Fuc0) were more prevalent in GnTIII-expressing plants (31% and 39%) than in control plants (10% ± 1%). As it is believed that β(1,2)-XylT acts before α(1,3)-FucT, these findings indicate that GnTIIIA.th. acts before β(1,2)-XylT and α(1,3)-FucT, whereas rat GnTIII acts at a later stage in the N-glycan processing pathway, still allowing the action of β(1,2)-XylT on the N-glycan.

Table 1.  Effect of expression of rat β(1,4)-N-acetylglucosaminyltransferase III (GnTIII) and of GnTIII fused to the Arabidopsis thaliana mannosidase II (ManII) cytosolic tail, transmembrane domain and stem region (CTS) (GnTIIIA.th.) on the N-glycosylation of Nicotiana tabacum SR1 plants
Type of glycanGnTIIIGnTIIIA.th.SR1 wild-type (%)
NG6 (%)NG20 (%)RG2 (%)RG3 (%)RG14 (%)RG17 (%)
Xyl0Fuc01574859314513 ± 7
Xyl0Fuc154212614155 ± 6
Xyl1Fuc03931157192210 ± 1
Xyl1Fuc14159169371873 ± 2
Total fucose (Fuc)46623734503377 ± 8
Total xylose (Xyl)80893116564082 ± 2
Paucimannosidic12225462 ± 18
High-mannose1165759 ± 5
Hybrid942941262516 ± 10
Hybrid bisected43232812141 ± 1
Complex bisected30379617171 ± 1
Complex/hybrid bisected53512714323110 ± 9

In plants expressing GnTIIIA.th., hybrid structures were more prevalent, whereas complex structures were encountered more frequently in plants expressing GnTIII. This distribution of N-glycans mirrors the effects of expression of native GnTIII and of its targeting variants in mammalian cells (Ferrara et al., 2006), and raises the question of whether the co-expression of ManII could redirect the glycan pool of hybrid structures towards complex structures.

Characterization of homozygous GnTIII- and ManII-over-expressing plants

Western blot analysis confirmed that the effect of glycoengineering is stable over three generations (data not shown). For the analysis of glycan structures, N-glycans were isolated and analysed using MALDI-TOF-MS. Representative spectra of six different plant lines expressing GnTIII and ManII or their ManII CTS-targeted variants are depicted in Figure 4. The spectra of plants expressing GnTIII displayed three dominant peaks with m/z values of 1472, 1618 and 1821, accounting for roughly 70%–85% of the total N-glycans. Expression of GnTIIIA.th. produced a greater heterogeneity of N-glycan structures than did the expression of GnTIII (Figure 4b,d,f vs. Figure 4a,c,e). These patterns were observed irrespective of the presence and localization of human ManII, indicating that the co-expression of human ManII plays only a minor role in the reshaping of the N-glycosylation pathway in N. tabacum. In contrast with human ManII, the A. thaliana homologue caused profound changes in the glycosylation pattern, when co-expressed with GnTIII and GnTIIIA.th.. In plants expressing GnTIII and A. thaliana ManII, a dominant peak of m/z 1821 was detected, representing a bisected complex sugar containing fucose and xylose, coinciding with the disappearance of the peak at m/z 1472 (Figure 4e). A summary of the detected peaks and abundances is reported in Table 2.

Figure 4.

Representative matrix-assisted laser desorption ionization time-of-flight mass spectra from N-glycans liberated from total protein of homozygous plants expressing native β(1,4)-N-acetylglucosaminyltransferase III (GnTIII) (a), GnTIIIA.th. (b), GnTIII and ManII (c), the hybrid enzymes GnTIIIA.th. and ManIIA.th. (d), GnTIII in combination with Arabidopsis thaliana ManII (e) and GnTIIIA.th. in combination with A. thaliana ManII (f). The * in (a) designates two peaks, which represent K-adducts of sugars. They are at m/z of 1704.608 and 1836.650 next to peaks at 1688.910 and 1821.816, respectively.

Table 2.  Relative abundance of N-glycan structures isolated from total protein of homozygous glycoengineered SR1 tobacco plants using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS). The data report the mean value of six to nine measurements for glycoengineered plants and of three plants for the wild-type control
Glycan typeCompositionm/z Na+ adductGnTIII (%)GnTIIIA.th. (%) RG2GnTIIIA.th. (%) RG3GnTIII/ ManII (%)GnTIIIA.th./ ManIIA.th. (%)GnTIII/A. th. ManII (%)GnTIIIA.th./A.  th. ManII (%)SR1 wild- type (%)
  1. A.th.,Arabidopsis thaliana; Fuc, fucose; GlcNAc, N-acetylglucosamine; GnTIII, β(1,4)-N-acetylglucosaminyltransferase III; Hex, hexose; ManII, mannosidase II; n.d., not detected; Xyl, xylose.

High-mannoseGlcNAc2Hex4Xyl0Fuc01095.370n.d.≤ 2n.d.n.d.≤ 2n.d.n.d.2
GlcNAc2Hex5Xyl0Fuc01257.422≤ 2≤ 23 ± 2≤ 2≤ 2≤ 2≤ 22
GlcNAc2Hex6Xyl0Fuc01419.475≤ 2≤ 26 ± 8n.d.≤ 2≤ 2≤ 2≤ 2
GlcNAc2Hex7Xyl0Fuc01581.528≤ 2n.d.n.d.n.d.≤ 2n.d.n.d.n.d.
GlcNAc2Hex8Xyl0Fuc01743.581n.d.≤ 2n.d.n.d.≤ 2n.d.n.d.≤ 2
HybridGlcNAc3Hex3Xyl0Fuc01136.396≤ 2n.d.n.d.≤ 2≤ 2n.d.6 ± 2≤ 2
GlcNAc2Hex4Xyl1Fuc01227.412n.d.≤ 22 ± 2n.d.n.d.n.d.n.d.2
GlcNAc3Hex3Xyl1Fuc01268.438≤ 2n.d.n.d.≤ 2≤ 2≤ 2≤ 23
GlcNAc3Hex3Xyl0Fuc11282.454n.d.≤ 2n.d.n.d.≤ 2n.d.≤ 24
GlcNAc3Hex4Xyl0Fuc01298.449n.d.2 ± 43 ± 4n.d.3 ± 3n.d.8 ± 3≤ 2
GlcNAc3Hex3Xyl1Fuc11414.496≤ 2≤ 21 ± 1≤ 2≤ 26 ± 2≤ 215
GlcNAc3Hex4Xyl1Fuc01430.491n.d.n.d.n.d.n.d.n.d.n.d.n.d.≤ 2
GlcNAc3Hex4Xyl0Fuc11444.507n.d.≤ 2n.d.n.d.n.d.n.d.n.d.≤ 2
GlcNAc3Hex5Xyl0Fuc01460.502≤ 220 ± 319 ± 5≤ 212 ± 7n.d.14 ± 8≤ 2
GlcNAc3Hex4Xyl1Fuc11576.549n.d.n.d.n.d.n.d.n.d.n.d.n.d.≤ 2
GlcNAc3Hex5Xyl1Fuc01592.544n.d.n.d.n.d.n.d.n.d.n.d.n.d.≤ 2
GlcNAc3Hex5Xyl0Fuc11606.560n.d.n.d.n.d.n.d.≤ 2n.d.n.d.≤ 2
Hybrid bisectedGlcNAc4Hex4Xyl0Fuc01501.528≤ 28 ± 46 ± 4≤ 213 ± 10n.d.≤ 2n.d.
GlcNAc4Hex4Xyl1Fuc01633.571≤ 2≤ 2±≤ 2≤ 2n.d.≤ 2n.d.
GlcNAc4Hex4Xyl0Fuc11647.586n.d.≤ 22 ± 3n.d.2 ± 2n.d.4 ± 1n.d.
GlcNAc4Hex5Xyl0Fuc01663.581n.d.19 ± 817 ± 9n.d.20 ± 12n.d.4 ± 2n.d.
GlcNAc4Hex4Xyl1Fuc11779.629≤ 2≤ 2n.d.n.d.≤ 2n.d.≤ 2n.d.
GlcNAc4Hex5Xyl1Fuc01795.623n.d.≤ 2n.d.n.d.≤ 2n.d.n.d.n.d.
GlcNAc4Hex5Xyl0Fuc11809.639n.d.8 ± 78 ± 4n.d.8 ± 6n.d.6 ± 3n.d.
Complex/hybrid bisectedGlcNAc4Hex3Xyl0Fuc01339.4762 ± 1≤ 23 ± 3≤ 24 ± 3n.d.2 ± 1n.d.
GlcNAc4Hex3Xyl1Fuc01471.51826 ± 26 ± 56 ± 324 ± 59 ± 6411 ± 5n.d.
GlcNAc4Hex3Xyl0Fuc11485.533≤ 214 ± 78 ± 4≤ 26 ± 426 ± 1≤ 2
GlcNAc4Hex3Xyl1Fuc11617.57624 ± 211 ± 37 ± 432 ± 211 ± 318 ± 310 ± 3≤ 2
Complex bisectedGlcNAc5Hex3Xyl0Fuc01542.555≤ 2≤ 2n.d.≤ 2≤ 2n.d.≤ 2n.d.
GlcNAc5Hex3Xyl1Fuc01674.5977 ± 1≤ 2n.d.7 ± 1≤ 25 ± 1≤ 2n.d.
GlcNAc5Hex3Xyl0Fuc11688.613≤ 2≤ 2n.d.≤ 2≤ 2≤ 2≤ 2n.d.
GlcNAc5Hex3Xyl1Fuc11820.65523 ± 34 ± 52 ± 232 ± 52 ± 251 ± 65 ± 2n.d.
PaucimannosidicGlcNAc2Hex2Xyl1Fuc0 903.306≤ 2n.d.n.d.n.d.n.d.n.d.n.d.≤ 2
GlcNAc2Hex2Xyl0Fuc1 917.3222 ± 5n.d.2 ± 4n.d.n.d.n.d.n.d.n.d.
GlcNAc2Hex3Xyl0Fuc0 933.317n.d.n.d.n.d.n.d.n.d.n.d.n.d.≤ 2
GlcNAc2Hex3Xyl1Fuc01065.359n.d.n.d.n.d.n.d.n.d.≤ 2n.d.10
GlcNAc2Hex3Xyl0Fuc11079.375n.d.n.d.n.d.n.d.n.d.n.d.≤ 24
GlcNAc2Hex3Xyl1Fuc11211.417n.d.n.d.n.d.n.d.≤ 25 ± 22 ± 238

Xylose and fucose levels varied as expected between the different plant lines, as well as between the individual plants of a plant line (Figure 5a). Although plants expressing GnTIIIA.th. had low levels of fucose and xylose, plants expressing GnTIII exhibited only a reduced core fucose level. For plants expressing GnTIII alone or in combination with ManII, three distinct pools were detectable, and an increase in core fucose levels was observed on co-expression of human ManII and A. thaliana ManII. The main fraction of sugars from control plants contained xylose and fucose (73% ± 2%). Only 13% ± 7% of the detected N-glycans were devoid of both modifications, but these were mainly of high-mannose type. In the lines expressing native GnTIII and native GnTIII in combination with ManII, the amount of Xyl0Fuc0 was not decreased relative to that of wild-type plants. However, the fraction of Xyl0Fuc0 was increased drastically to averages of 53% ± 5% and 60% ± 8% in the progenies of the independent plant lines expressing GnTIIIA.th. (RG2.34 and RG3.16/25, respectively), and to 58% ± 10% in plants co-expressing GnTIIIA.th. and ManIIA.th. In contrast, the co-expression of A. thaliana ManII increased the degree of fucosylation and xylosylation, and 42% ± 10% of the structures were devoid of both core modifications. In the plant lines expressing GnTIII, total xylose levels were not reduced relative to those of control plants; indeed, a slight increase was observed on co-expression of either one of the two ManII forms. In contrast, the fucose level was decreased in GnTIII-expressing plant lines; however, the effect was reversed on co-expression of ManII. In plants expressing GnTIIIA.th., the amount of fucosylation was decreased on average to 31% ± 9%, with a minimum of 14% in a single plant. The structures containing xylose were even more dramatically reduced, reaching an average of 19% ± 4%, with a minimal amount of 13% in one plant.

Figure 5.

N-Glycan analysis of total protein extracts prepared from wild-type and homozygous glycoengineered F2 plants. (a) Effect of β(1,4)-N-acetylglucosaminyltransferase III (GnTIII) expression on the overall content of α(1,3)-fucose and β(1,2)-xylose in N-glycan preparations of homozygous plant lines, as determined using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS). Expression of GnTIIIA.th. most efficiently reduces the amount of fucose and xylose. The data points represent individual samples of a population. The peaks obtained in MALDI-TOF-MS have been assigned to high-mannose (b), hybrid (c), hybrid bisected (d), complex/hybrid bisected (e), complex bisected (f) and paucimannosidic (g) sugars, and their relative contributions to the total N-glycans are reported. Data were collected from six to nine measurements for glycoengineered plants and from three wild-type plants. Mean values and standard deviations are reported. Insets represent the respective structures, which may be further modified with α(1,3)-fucose and β(1,2)-xylose. The position of the bisecting N-acetylglucosamine (GlcNAc) is indicated by a grey square.

Glycoengineering of plant N-glycosylation using GnTIII completely changed the relative contributions of the different sugar types (Figure 5b–g, Table 3). A striking difference in the glycoengineered plants was the almost complete disappearance of paucimannosidic structures in the extracts. Although, in control plants, this type of sugar accounted for 62% ± 18% of the total sugars, paucimannosidic sugars were almost absent in all of the plant lines expressing GnTIII or GnTIII and ManII. The other types of sugars appearing in the extracts from wild-type plants were of high-mannose, hybrid or complex type, contributing 9% ± 5%, 16% ± 10% and 10% ± 9%, respectively, to the total N-glycans. High-mannose-type sugars were present at either similar or reduced levels in the glycoengineered plants relative to the wild-type control plants.

Table 3.  Effect of mannosidase II (ManII) co-expression on N-glycosylation in Nicotiana tabacum SR1 plants expressing rat β(1,4)-N-acetylglucosaminyltransferase III (GnTIII) or GnTIIIA.th.
Type of glycanGnTIIIGnTIIIA.th.SR1 wild-type (%)
(%)ManII (%)A.th. ManII (%)RG2* (%)RG3* (%)ManIIA.th. (%)A.th. ManII (%)
  • A.th., Arabidopsis thaliana.

  • *

    For plants expressing GnTIIIA.th., progenies of two independent transgenic lines (RG2 and RG3) were included in the study.

Xyl0Fuc011 ± 21 ± 23 ± 353 ± 560 ± 858 ± 1042 ± 1013 ± 7
Xyl0Fuc15 ± 51 ± 24 ± 325 ± 621 ± 516 ± 421 ± 55 ± 6
Xyl1Fuc037 ± 332 ± 512 ± 27 ± 69 ± 211 ± 616 ± 610 ± 1
Xyl1Fuc148 ± 566 ± 780 ± 616 ± 510 ± 615 ± 520 ± 673 ± 2
Total fucose (Fuc)52 ± 467 ± 784 ± 441 ± 1131 ± 931 ± 742 ± 677 ± 8
Total xylose (Xyl)84 ± 497 ± 392 ± 523 ± 119 ± 426 ± 1036 ± 1282 ± 2
Paucimannosidic3 ± 7   05 ± 3   02 ± 4    04 ± 662 ± 18
High-mannose1 ± 1   02 ± 21 ± 29 ± 74 ± 43 ± 29 ± 5
Hybrid3 ± 21 ± 29 ± 524 ± 928 ± 1119 ± 1333 ± 1116 ± 10
Hybrid bisected4 ± 11 ± 21 ± 137 ± 635 ± 645 ± 2320 ± 81 ± 1
Complex bisected32 ± 443 ± 757 ± 65 ± 53 ± 33 ± 410 ± 31 ± 1
Complex/hybrid bisected54 ± 455 ± 623 ± 432 ± 1024 ± 629 ± 1029 ± 910 ± 9

In plants expressing native GnTIII, most sugars were of complex bisected type (32% ± 4%) or of complex/hybrid bisected type (54% ± 4%). At the level of the individual structures, plants expressing GnTIII produced mainly four structures: two complex bisected GlcNAc5Hex3 forms, the major form comprising xylose and fucose (23% ± 3%), and two structures with the composition GlcNAc4Hex3, either comprising both core xylose and fucose (24% ± 2%) or only core fucose (26% ± 2%). Small amounts of high-mannose, hybrid, complex bisected and paucimannosidic sugars were also detectable (Tables 2 and 3).

Although native GnTIII produced mainly complex sugars, targeting its expression with A. thaliana ManII CTS produced a greater variety of observed sugar structures. The most abundant sugars were of hybrid type, with or without bisecting GlcNAc. In the two independent lines expressing GnTIIIA.th., 37% ± 6% and 35% ± 6% were hybrid bisected structures and 24% ± 9% and 28% ± 11% represented hybrid sugars. Similar fractions (32% ± 10% and 24% ± 6%) were assigned to the complex/hybrid bisected sugars. Furthermore, a small amount of high-mannose and complex bisected structures appeared (Table 3). The main structures detectable in these two lines were a hybrid structure (GlcNAc3Hex5Xyl0Fuc0: 20% ± 3% and 19% ± 5%), a hybrid bisected structure mainly lacking xylose and fucose – GlcNAc4Hex5Xyl0Fuc0 (19% ± 8% and 17% ± 9%), a fucosylated derivative thereof (8% ± 7% and 8% ± 4%) and the species GlcNAc4Hex4Xyl0Fuc0 (8% ± 4% and 6% ± 4%). Furthermore, GlcNAc4Hex3 structures comprising fucose (14% ± 7% and 8% ± 4%), xylose (6% ± 5% and 6% ± 3%) or xylose and fucose (11% ± 3% and 7% ± 4%) were present.

In the plant co-expressing GnTIII and human ManII, the amount of complex bisected structures increased to 43% ± 7% (P < 0.01, vs. GnTIII), the major structure being GlcNAc5Hex3Xyl1Fuc1, comprising 32% ± 5% of the total N-glycans (Table 2). This increase was compensated for by a decrease in high-mannose, hybrid, hybrid bisected and paucimannosidic structures (Table 3). On co-expression of A. thaliana ManII, this increase was even more pronounced. The structure GlcNAc5Hex3Xyl1Fuc1 (m/z 1821) comprised 51% ± 6% (P < 0.01, vs. GnTIII) of all N-glycans, with a strong accompanying decrease in GlcNAc4Hex3Xyl1Fuc0 (m/z 1472, 11% ± 5% vs. 26% ± 2%).

In the plant lines expressing GnTIIIA.th. and ManIIA.th., a similar distribution of N-glycans to that in plants expressing only GnTIIIA.th. was observed: hybrid bisected structures accounted for 45% ± 23%, hybrid structures for 19% ± 13% and complex/hybrid bisected structures for 29% ± 10% of the total sugars. However, a slight increase in hybrid bisected structures and a concomitant decrease in hybrid structures was observed (Table 3). The main peaks were hybrid bisected structures, accounting for 20% ± 12% (GlcNAc4Hex5Xyl0Fuc0) and 13% ± 10% (GlcNAc4Hex4Xyl0Fuc0). The other N-glycan types were encountered less frequently than in the absence of ManII.

Plants co-expressing GnTIIIA.th. and A. thaliana ManII displayed two main differences from plants expressing GnTIIIA.th. and ManIIA.th.. There was an increase in hybrid structures completely lacking core modifications (33% ± 11%) and a decrease in hybrid bisected structures (20% ± 8%) (Table 3). In particular, the GlcNAc4Hex5Xyl0Fuc0 and GlcNAc4Hex4Xyl0Fuc0 structures (4% ± 2% and ≤ 2% respectively) were strongly reduced relative to plants expressing GnTIIIA.th. or GnTIIIA.th. and ManIIA.th.. However, a twofold increase in complex bisected structures (10% ± 3%) was observed (P < 0.01).


One of the main goals of our glycoengineering approach was to study whether the activities of N. tabacumβ(1,2)-XylT and α(1,3)-FucT could be outcompeted by the expression of GnTIII. In contrast with approaches using knockouts or RNAi, a technique such as that described here can be applied to any desired plant system without the need for specific DNA sequence information. In mammals, the processing of hybrid structures to complex structures is strongly affected by the presence of a bisecting GlcNAc. The presence of the bisecting GlcNAc represents a stop signal for the further processing of hybrid structures by the Golgi ManII, GnTII and core α(1,6)-FucT, thus producing hybrid bisected structures (Schachter et al., 1983). Therefore, the relative abundance of GnTIII vs. ManII dictates the fate of the N-glycans, channelling the N-glycans into hybrid or complex structures. Thus, appropriate localization and dosage seem to be key for successful glycoengineering using GnTIII.

Several different tools were employed to ensure the high-level and stable expression of GnTIII and ManII in N. tabacum plants. This was achieved by employing the strong A. tumefaciens-derived UAS123mas promoter and a synthetic MAR element. Furthermore, GnTIII was expressed with either its native CTS sequence or C-terminally appended to the CTS of A. thaliana ManII. The strategy to target a heterologous glycosyltransferase in the plant Golgi has been described previously (Bakker et al., 2006), where a hybrid enzyme comprising the XylT localization sequences fused to the β(1,4)-galactosyltransferase was successfully used to reduce the amount of xylose and fucose. Confocal microscopy analysis of BY-2 cells expressing eYFP fused to the CTS of rat GnTIII or of A. thaliana ManII indicated that both CTSs are functional in BY-2 cells, targeting eYFP to a brefeldin A-sensitive subcellular compartment, indicative of correct targeting to the Golgi (Figure 2e,f). Previous studies have confirmed that the A. thaliana ManII CTS targets a green fluorescent protein (GFP) fusion to the Golgi compartment, its localization partially overlapping with that of the Lewis A epitope (Strasser et al., 2006). The combination of the different strategies allowed us to interfere effectively with the plant intrinsic N-glycosylation machinery, thereby decreasing the amount of fucosylated and xylosylated N-glycans of N. tabacum (Figure 5a).

With the expression of the hybrid enzyme GnTIIIA.th., up to 60% of structures lacked both xylose and fucose. In contrast, the expression of rat GnTIII using the native CTS sequence only resulted in a decrease in fucose but not xylose. Recent experiments by Rouwendal et al. (2007) have shown that, using CaMV 35S-driven GnTIII expression, the efficient introduction of a bisecting GlcNAc residue can be achieved; however, despite a high degree of bisection, little effect on α(1,3)-fucose levels and no effect on β(1,2)-xylose levels were observed. This indicates that the effect of GnTIII expression on fucosylation and xylosylation may rely in part on the appropriate localization of the enzyme in the Golgi, therefore allowing it to act on the substrate before β(1,2)-XylT and α(1,3)-FucT. A. thaliana XylT is highly active on hybrid structures, and its activity is greatly reduced on complex N-glycans (Pagny et al., 2003; Bencur et al., 2005). The hybrid structure GlcNAc3Man3 is preferred over GlcNAc3Man5 (Bencur et al., 2005). XylT activity is abolished by the presence of a bisecting GlcNAc residue when presented on a complex sugar. Similarly, the activity of A. thaliana FucT is reduced by a bisecting GlcNAc (Rouwendal et al., 2007). Furthermore, the preceding activity of either one of the core-modifying transferases on the substrate reduces the affinity for the substrate of the other transferase (Bencur et al., 2005). Overall, these data, in conjunction with our in vivo results, show that XylT acts only poorly on the substrate after GnTIII, as observed on expression of GnTIIIA.th.; however, GnTIII is able to process xylosylated substrates, as observed on expression of native GnTIII.

In addition to its catalytic function of inserting a bisecting GlcNAc residue, the presence of GnTIII may play an alternative role. The high levels of GnTIII expression seem to block the N-glycan modifications in the Golgi apparatus. Redirecting GnTIII using ManII CTS leads to an accumulation of hybrid bisected and hybrid sugars. Two mechanisms may be responsible for this phenomenon. On the one hand, hybrid bisected sugars are not processed by ManII in mammalian cells (Schachter et al., 1983). Similarly, the activity of A. thaliana ManII is reduced to approximately 20% if the substrate contains a bisecting β(1,2)-xylose (Strasser et al., 2006). However, it is not deducible whether the bisecting GlcNAc has the same inhibitory effect on the activity of endogenous ManII as does the bisecting β(1,2)-xylose. On the other hand, GnTIIIA.th. competes with endogenous ManII for access to the substrate and/or space in the early Golgi compartment. A reduced activity of endogenous ManII would therefore prevent the action of GnTII and halt N-glycan processing at the level of the hybrid structures. A similar Golgi processing blocking phenomenon has also been observed on expression of galactosyltransferase fused to the A. thaliana XylT CTS. Expression of this hybrid enzyme leads to a very strong increase in the amount of hybrid structures, reaching up to 50% of the total structures. Furthermore, a doubling or even tripling in the amount of high-mannose structures relative to control plants has been observed (Bakker et al., 2006), a phenomenon not seen in our experiments.

The fact that both mechanisms contribute to the accumulation of hybrid and hybrid bisected structures can be derived from the MALDI-TOF-MS analysis. Although, in the case of wild-type plants, hybrid structures were almost exclusively composed of GlcNAc3Hex3Xyl1Fuc1, the hybrid bisected and hybrid counterparts in GnTIIIA.th. plants contained mostly four or five mannose residues: GlcNAc4Hex5Xyl0Fuc0 (19% ± 8% and 17% ± 9%), GlcNAc4Hex4Xyl0Fuc0 (8% ± 4% and 6% ± 4%) and GlcNAc3Hex5Xyl0Fuc0 (20% ± 3% and 19% ± 5%). This further substantiates the concept that trimming by ManII is strongly inhibited.

The over-expression of Golgi ManII in mammalian cells reverses this phenotype, redirecting hybrid bisected N-glycans towards complex bisected structures (Ferrara et al., 2006).

In contrast with mammalian cells, the over-expression of human ManII is unable to efficiently reduce the amount of hybrid and hybrid bisected structures in plants. The hybrid and hybrid bisected structures in GnTIIIA.th. plants co-expressing ManIIA.th. still contained mostly four or five mannose residues: GlcNAc4Hex5Xyl0Fuc0 (20% ± 12%), GlcNAc4Hex4Xyl0Fuc0 (13% ± 10%) and GlcNAc3Hex5Xyl0Fuc0 (12% ± 7%).

In the lines expressing both GnTIIIA.th. and ManIIA.th., the total amount of hybrid bisected structures was increased by approximately 10% relative to plants expressing GnTIIIA.th. In contrast, the co-expression of A. thaliana ManII reduced the amount of hybrid bisected structures to 20% ± 8% and produced a twofold increase in complex bisected structures. At present, one can only speculate on the specificity of A. thaliana ManII for bisected hybrid structures. Human ManII and the A. thaliana homologue share 38% identity and 55% similarity (Strasser et al., 2006). However, as A. thaliana ManII evolved in a system lacking bisecting GlcNAc, it is not foreseeable whether its activity is also inhibited by the presence of a bisecting GlcNAc.

It might be argued that the high prevalence of hybrid and hybrid bisected sugars would present an obstacle to the use of the plant lines as a production host for therapeutic proteins, such as the immunoglobulin G (IgG) molecule. However, it has been shown that the nature of the N-glycan, of either hybrid or complex type, on the IgG molecule has, at most, only a slight effect on the antibody-dependent cellular cytotoxicity (ADCC) (Ferrara et al., 2006; Kanda et al., 2007), whereas a large increase in IgG binding to FcγRIIIa, leading to a large increase in ADCC, is caused by the absence of core fucose (irrespective of its linkage), as well as the presence of a bisecting GlcNAc (Shinkawa et al., 2003; Hodoniczky et al., 2005; Cox et al., 2006; Ferrara et al., 2006). Therefore, the elimination of fucose and, presumably, also xylose from the N-glycans of plant-produced antibodies is crucial, not only to prevent an immunogenic reaction against these plant-specific epitopes, but also for improved activity of a plant-produced antibody. Furthermore, it has been shown that the prevalence of hybrid or hybrid bisected sugars reduces the complement-dependent cytotoxicity of an IgG (Ferrara et al., 2006; Kanda et al., 2007). As complement activation is implicated in the causation of first infusion reactions in vivo (Winkler et al., 1999; Van Der Kolk et al., 2001), IgG bearing hybrid or hybrid bisected glycans may be preferred.

Experimental procedures

Modification of pCAMBIA1200

All oligonucleotides for polymerase chain reaction (PCR) were obtained from Microsynth (Balgach, Switzerland) and are shown in Table S1 (see ‘Supporting Information’). The enhanced CaMV 35S promoter driving the expression of the resistance marker gene hptI was replaced with the nopaline synthase promoter (Pnos) isolated from pBI121 (Clontech, Palo Alto, CA, USA). Pnos was PCR amplified with oligonucleotide pair OAF001 and OAF003, attaching, in addition, the first 30 bp of the hptI gene to the 3′ end in order to enable promoter exchange. The promoter was replaced by digestion with BstXI and AatII and insertion of Pnos, generating pAX36.

A synthetic MAR, composed of two identical tandem repeats, MAR1 and MAR2, was created as described previously (Van der Geest et al., 2004). Repeats were assembled by the dimerization of oligonucleotide pairs and subsequent PCR amplification. Dimers were formed with the following oligonucleotide pairs: dimer A, Sm1A and Sm2; dimer B, Sm1B and Sm2; dimer C, Sm3 and Sm4; dimer D, Sm5 and Sm6; dimer E, Sm7 and Sm8A; dimer F, Sm7 and Sm8B. Flanking oligonucleotides (Sm1A and Sm8A for MAR1, Sm1B and Sm8B for MAR2) contained additional sites for subcloning. MAR1 was composed of dimers A, C, D, E (HindIII, BamHI) and MAR2 of dimers B, C, D, F (BglII, SbfI/EcoRI). The PCR products were inserted into pCR2.1-Topo (Invitrogen, Basel, Switzerland), and the DNA sequences were verified. MAR1 and MAR2 were joined in two consecutive steps in pSL1180 (Pharmacia, Uppsala, Sweden) using HindIII, BamHI (MAR1) and BglII, EcoRI (MAR2) into the BamHI, EcoRI sites of the intermediate vector, generating the full-length MAR element (pAX54). The full-length MAR element from pAX54 was inserted as a HindIII-PstI fragment into the corresponding sites of pAX36, generating pAX49.

Cloning of enhancer and promoter fragments

A promoter comprising three repeats of the UAS of the octopine synthase promoter fused to the UAS and promoter region of Pmas was isolated by colony PCR and assembled in pSL1180 (Pharmacia) (Ni et al., 1995).

The individual UAS elements and Pmas region were isolated by colony PCR from A. tumefaciens Ach5 (Ooms et al., 1981). The presence of a 3′ flanking BamHI site in the UAS was used for the assembly of the UAS trimer. The PCR fragments harboured flanking EcoRI, BclI (oligonucleotides OAF031, OAF039), BamHI, BglII (oligonucleotides OAF32, OAF033) and BamHI, EcoRV (oligonucleotides OAF032, OAF034) restriction sites. By joining the elements, all internal BamHI restriction sites were eliminated in the final plasmid. Additional BamHI, SbfI restriction sites were introduced at the 5′ end of the UAS trimer.

The 385-bp fragment of the Pmas sequence was isolated by colony PCR using oligonucleotides OAF035 and OAF036, and inserted into the EcoRV and XbaI sites of pSL1180, generating pAX57.

The UAS trimer and Pmas promoter element were joined by inserting the UAS trimer as an EcoRI, EcoRV fragment into the corresponding sites of pAX57, generating pAX60. Plasmid pAX60 harbours the chimeric promoter fragment (5′-EcoRI/SbfI/BamHI-UAS2-UAS1-UAS3-EcoRV-MAS-XbaI-3′), which was used for the expression of glycoengineering enzymes. A further plasmid was obtained lacking UAS2 (pAX59).

Generation of β-glucuronidase reporter constructs

For the analysis of the promoter activity, reporter constructs based on the CaMV 35S promoter β-glucuronidase expression cassette from pBI121 were constructed in pCAMBIA2200. To that end, the expression cassette, comprising the CaMV 35S promoter, β-glucuronidase and polyadenylation signal, was transferred into the EcoRI, HindIII sites of pCAMBIA2200. The original promoter was exchanged with SbfI-XbaI fragments containing the UAS12mas or UAS123mas promoter.

Generation of eYFP fusion constructs

For confocal laser scanning microscopy, fusion constructs comprising amino acids 1–149 of the N-terminal rat GnTIII protein were fused to eYFP (pAX111). Furthermore, amino acids 1–165 of GnTIIIA.th., comprising the A. thaliana ManII CTS and first 73 amino acids of the catalytic sequence of rat GnTIII protein, were fused to eYFP (pAX112). The StuI, BglII fragment comprising the catalytic domain of GnTIII was excised from pAX67 and pAX70, and replaced with a PCR-amplified eYFP carrying 5′StuI and 3′BglII restriction sites. The promoter was replaced with the CaMV 35S promoter using SbfI, XbaI.

Isolation of A. thaliana ManII sequences

The A. thaliana ecotype Columbia ManII CTS (sequence corresponding to amino acids 1–92) was isolated from genomic DNA by PCR, modifying the 3′ end to contain additional bases of GnTIII or ManII using oligonucleotide pairs SK105 and SK106 or SK105 and SK109, respectively (Strasser et al., 2006). In a second PCR step, short fragments of either GnTIII or ManII were created using oligonucleotide pairs SK107 and SK108 or SK110 and SK111, respectively. These fragments were assembled by PCR, generating fusions of A. thaliana ManII CTS with GnTIII (oligonucleotides SK105, SK108) or A. thaliana ManII CTS with human ManII sequences (oligonucleotides SK105, SK111), and inserted into pCR2.1-Topo (Invitrogen), yielding pSK120 (GnTIII) and pSK121 (ManII).

The coding sequence of A. thaliana Golgi ManII was isolated using a One-Step RT-PCR Kit (Qiagen, Basel, Switzerland) with the addition of Protector RNase Inhibitor (Roche, Rotkreuz, Switzerland). RNA was extracted from 20-day-old A. thaliana cv. Columbia plants using Trizol reagent (Invitrogen) and an RNeasy Plant Mini Kit (Qiagen). The first part of A. thaliana ManII was amplified with the primers SKAM2-1 and SKAM2-2, the second part with primers SKAM2-3 and SKAM2-4 and the third part with primers SKAM2-5 and SKAM2-6. The amplified parts were cloned into pCR2.1 TOPO vector (Invitrogen) and sequenced, yielding pSK149 (first part), pSK150 (second part) and pSK153 (third part).

To facilitate the assembly of the amplified fragments, an annealed linker consisting of SKlinkerfor and SKlinkerrev was cloned into the EcoRI/HindIII sites of pUC19, yielding pSK151. The first part in pSK149 was cloned into pSK151 using the XbaI/SacI sites, yielding pSK152. The second part in pSK153 was cloned into pSK152 using the SacI/BlpI sites, yielding pSK155. The third part in pSK150 was cloned into pSK155 using the BlpI/SalI sites, yielding pSK156 containing the assembled ManII.

Cloning of native GnTIII and ManII forms and of CTS exchange variants

A c-Myc-tagged rat GnTIII cDNA and the native human ManII cDNA served as starting material for the cloning of the expression vectors (Ferrara et al., 2006). Both cDNAs were inserted downstream of the CaMV 35S promoter, followed by the nos polyadenylation signal (nosPa), both elements isolated from pBI121, generating pSK103 (ManII) and pSK142 (GnTIII), respectively.

The GnTIIIA.th. gene (pSK132) was generated by insertion of an Ecl136II and RsrII fragment, comprising A. thaliana ManII CTS isolated from pSK121, into the HindII, RsrI sites of pETR1096. The GnTIIIA.th. gene was isolated from pSK132 using BamHI, the overhang was filled in with a Klenow fragment, digested with XbaI and inserted into the XbaI and Ecl136II sites of pSK142, generating pSK136.

ManIIA.th. was obtained by exchanging the native CTS of human ManII in pSK103 with the CTS of ManIIA.th. from pSK121 using XbaI, RsrI, generating pSK124.

A redundant BamHI site was deleted from pSK136 and pSK142 by digestion with BamHI, Klenow fill-in and self-ligation, generating plasmids pAX63 (GnTIII) and pAX68 (GnTIIIA.th.), respectively. All expression cassettes comprise the CaMV 35S promoter, GnTIII or ManII and the nospA: pSK103 (ManII), pSK124 (ManIIA.th.), pAX63 (GnTIIIA.th.) and pAX68 (GnTIII).

Generation of binary expression vectors for GnTIII and ManII

The promoter fragment was excised with PstI, XbaI from pAX60, and inserted into the SbfI, NheI or SbfI, XbaI sites of pAX63 and pAX68, respectively, generating pAX65 and pAX69.

The expression cassettes containing the UAS123mas promoter-gntIII-nosPa from the resulting plasmids were inserted into the BamHI, EcoRI sites of pAX49, generating the final expression vectors pAX67 (native GnTIII) and pAX70 (GnTIIIA.th.).

For the co-expression of GnTIII and human ManII, carrying either their native localization sequence or the A. thaliana ManII CTS, the coding sequence of ManII under the control of the UAS123mas promoter was inserted upstream of the promoter driving GnTIII expression.

The UAS123mas promoter from pAX60 was inserted into pSK103 (native ManII) and pSK124 (ManIIA.th.), generating pAX64 (UAS123mas-native ManII) and pAX66 (UAS123mas-ManIIA.th.), respectively. An additional nosPa from pSK101 was inserted as a SalI, BglII fragment into the SalI, BamHI sites of both GnTIII expression vectors.

The SbfI, SalI fragments from pAX64 and pAX66 were inserted upstream of nosPa into the SbfI, SalI sites, generating pAX73, containing native human ManII and rat GnTIII, and pAX74, containing the targeting variants ManIIA.th. and GnTIIIA.th., respectively. Both genes were under the control of the UAS123mas promoter.

The full-length A. thaliana ManII gene from pSK156 was inserted into XbaI, SalI of pAX66, replacing the human ManII gene. The cassette comprising the UAS123mas promoter-A. thaliana ManII encoding sequence was transferred into the SalI and SbfI sites of pAX73 and pAX74, generating pAX100 (native rat GnTIII and A. thaliana ManII) and pAX101 (GnTIIIA.th. and A. thaliana ManII), respectively.

Plant growth and transformation of N. tabacum cv. Petit Havana SR1

All media and additives were obtained from Duchefa (Haarlem, the Netherlands). N. tabacum cv. Petit Havana SR1 plants were grown on either Murashige and Skoog (MS) medium supplemented with 3% sucrose or in soil with a 16-h photoperiod and at 25 °C (Frey et al., 2004).

Plasmids were transferred into A. tumefaciens LBA4404 using the freeze–thaw method, but including a 16-h regeneration period in yeast mannitol broth (YMB) before plating on selection medium (Hoekema et al., 1983).

Tobacco leaf discs were transformed using A. tumefaciens strain LBA4404 according to a method published previously (Svab et al., 1975). A culture of agrobacteria was grown in 20 mL of YMB for 32 h at 28 °C. The cells were collected by centrifugation, and the resulting pellet was resuspended in RMOP (MS medium containing 30 g/L sucrose, 1 mg/L myo-inositol, 1 mg/mL thiamine, 0.1 mg/L 1-naphthalene acetic acid and 1 mg/L 6-benzylaminopurine) to obtain an optical density at 600 nm of approximately one. Leaves from plants grown in vitro were cut into discs (diameter, 1 cm) and stored in RMOP. For transformations, leaf discs were floated on RMOP containing agrobacteria and incubated for 5 min. The leaf discs were blotted dry on filter paper and incubated upside down on solid RMOP for 48 h. They were then transferred to RMOP containing hygromycin (50 mg/L) and timentin (250 mg/L), and incubated under regular growth conditions. The leaf discs were transferred every 2 weeks onto fresh selective plates. Shoots were excised and transferred onto solidified MS medium containing sucrose (30 g/L), hygromycin (50 mg/L) and timentin (250 mg/L). Root-forming shoots were transferred into soil and grown for 3 months at 25 °C with a 16-h/8-h light/dark cycle.

Growth and transformation of N. tabacum BY-2 cells

N. tabacum L. cv. BY-2 cells were maintained in MS medium with 30 g/L sucrose, 1 mg/L thiamine, 255 mg/L KH2PO4 and 0.222 mg/L 2,4-dichlorophenoxyacetic acid at 25 °C on a gyratory shaker at 110 r.p.m. in the dark. The culture was diluted 1 : 10 in fresh medium weekly.

A culture of agrobacteria was grown in 5 mL YMB for 32 h at 28 °C and harvested by centrifugation. The resulting pellet was resuspended in MS medium to obtain an optical density at 600 nm of approximately one; 30 mL of a 5-day-old BY-2 culture was mixed with 60 µL of 100 mm acetosyringone; 1 mL of the Agrobacterium suspension was added to the plant cells, and the suspension was incubated for 48 h in the dark at 25 °C.

To remove the agrobacteria from the BY-2 cells, the suspension culture was washed three times with MS medium supplemented with 250 mg/L timentin, and plated onto solidified growth medium containing selective antibiotics and 250 mg/mL timentin.

Isolation of the soluble protein fraction

Leaf material was extracted in homogenization buffer containing 50 mm 3(N-morpholino)propanesulphonic acid (MOPS), pH 7.5, 2% (v/v) Triton X-100, 1 mm MgCl2, 1 mm dithiothreitol, 10% (w/v) sucrose and 150 mm NaCl. Cellular debris and nuclei were separated from the soluble protein fraction and endomembranes after centrifugation at 3000 g and 4 °C for 5 min. The protein concentration was determined by the Bradford method using bovine serum albumin as the standard (Bradford, 1976).


Lysates were resolved under denaturing conditions on 10% sodium dodecylsulphate (SDS)-polyacrylamide gels (Laemmli, 1970). Separated proteins were electrophoretically transferred onto Immobilon-P nylon membranes (Millipore, Billerica, MA, USA) at 300 mA and room temperature. All incubation and wash steps were performed in 1% non-fat dry milk in tris(hydroxymethyl)aminomethane (Tris)-buffered saline containing 0.1% Tween-20. Polyclonal rabbit antisera against rat GnTIII, human ManII and a polyclonal rabbit anti-HRP IgG fraction were used. The secondary antibody was an anti-rabbit IgG HRP conjugate (Amersham Pharmacia, Bukinghamshire, UK). Chemiluminescent detection was performed with an ECL™ Western blotting detection reagent (Pierce, Rockford, IL, USA), according to the supplier's recommendations.

N-Glycan isolation and MALDI-TOF-MS analysis

Pepsin-digested protein was deglycosylated in 100 mm sodium acetate, pH 5.2, using 0.5 mU peptide N-glycosidase A (PNGase A) (Roche), and the released glycans were purified using C18 and cation exchange columns. The mass spectrum of the purified glycan preparation was acquired using an Autoflex MALDI-TOF mass spectrometer (Bruker Daltonics, Fällanden, Switzerland) in positive ion mode and operated in reflector mode. An m/z range of 900–2000 was measured. Peaks were assigned according to the calculated molecular masses of plant N-glycans, interpreting hexose, N-acetylhexosamine, deoxyhexose and pentose as mannose, GlcNAc, fucose and xylose, respectively.

β-Glucuronidase assay

BY-2 cells were lysed in extraction buffer [50 mm NaH2PO4, pH 7.0, 10 mm ethylenediaminetetraacetic acid (EDTA), 0.1% Triton X-100, 0.1% sodium lauryl sarcosine, 10 mmβ-mercaptoethanol]. Soluble protein extract was collected by centrifugation at 23 000 g and 4 °C for 20 min. Twenty-five microlitres of cell extract were added to 200 µL of extraction buffer containing 1 mm 4-methyl umbelliferyl glucuronide (MUG; Duchefa, Haarlem, the Netherlands). The reaction was incubated at 37 °C, aliquots of 25 µL were sampled and the reaction was quenched by the addition of 250 µL of 1 m Na2CO3. Fluorescence was measured with excitation at 365 nm and emission at 455 nm. Four to six individually selected BY-2 clones were analysed per promoter. The data are represented as the β-glucuronidase activity (relative light unit/min/mg protein) of individual clones, and the mean and standard deviations of different lines.

Confocal laser scanning microscopy

Cells for confocal laser scanning microscopy were diluted 1 : 10 in fresh medium and grown for 4 days. Brefeldin A (Fluka Chemie, Buch, Switzerland) was added from a stock solution of 5 mg/mL in dimethyl sulphoxide (DMSO) to a final concentration of 10 µg/mL, and the cells were incubated for 1 h. One millilitre of BY-2 culture was collected by centrifugation at 500 g for 1 min, and the cells were fixed in 3% paraformaldehyde in phosphate-buffered saline (PBS) for 10 min and washed three times in PBS. Imaging was performed on a Leica TCS SP1 confocal laser scanning microscope (Leica, Wetzlar, Germany) with the adequate excitation and emission filters.


This project (7074.3 LSPP-LS) was financed by the Swiss federal agency ‘Commission for Technology and Innovation’. We would like to thank Professor Wilhelm Gruissem for initial scientific discussions and for providing the glasshouse facility, Dr Johannes Fütterer for providing the N. tabacum BY-2 cells and A. thaliana genomic DNA, Doris Russenberger for providing the A. thaliana plants, Professor Christian Baron for providing the A. tumefaciens Ach5 strain, and Calogero L’Abate and Daniel Streich for technical assistance.