Production of complex multiantennary N-glycans in Nicotiana benthamiana plants.

In recent years, plants have been developed as an alternative expression system to mammalian hosts for the production of therapeutic proteins. Many modifications to the plant glycosylation machinery have been made to render it more human because of the importance of glycosylation for functionality, serum half-life, and the safety profile of the expressed proteins. These modifications include removal of plant-specific β1,2-xylose and core α1,3-fucose, and addition of bisecting N-acetylglucosamine, β1,4-galactoses, and sialic acid residues. Another glycosylation step that is essential for the production of complex human-type glycans is the synthesis of multiantennary structures, which are frequently found on human N-glycans but are not generated by wild-type plants. Here, we report both the magnICON-based transient as well as stable introduction of the α1,3-mannosyl-β1,4-N-acetylglucosaminyltransferase (GnT-IV isozymes a and b) and α1,6-mannosyl-β1,6-N-acetylglucosaminyltransferase (GnT-V) in Nicotiana benthamiana plants. The enzymes were targeted to the Golgi apparatus by fusing their catalytic domains to the plant-specific localization signals of xylosyltransferase and fucosyltransferase. The GnT-IV and -V modifications were tested in the wild-type background, but were also combined with the RNA interference-mediated knockdown of β1,2-xylosyltransferase and α1,3-fucosyltransferase. Results showed that triantennary Gn[GnGn] and [GnGn]Gn N-glycans could be produced according to the expected activities of the respective enzymes. Combination of the two enzymes by crossing stably transformed GnT-IV and GnT-V plants showed that up to 10% tetraantennary [GnGn][GnGn], 25% triantennary, and 35% biantennary N-glycans were synthesized. All transgenic plants were viable and showed no aberrant phenotype under standard growth conditions.

During the last decade, several research groups have explored different protein expression platforms, such as bacteria, yeasts, insect cells, and plants, for the production of cost-effective and safe biotherapeutics. For this purpose, plants offer several advantages over other expression systems. Compared to bacteria and mammalian cells, plant systems are considered as safe because of the absence of human pathogens, oncogenic DNA sequences, and endotoxins. Furthermore, the production capacity of transgenic plants is almost unlimited, as it depends only on the surface dedicated to the plant culture and the production costs are significantly lower than those of cell-based production systems (Twyman et al., 2003;. Recently, the first plant-made pharmaceutical, taligurase alfa, a form of the glucocerebrosidase enzyme for the treatment of Gaucher's disease, has been made available to patients in the United States and other countries under an Expanded Access Program approved by the U.S. Food and Drug Administration (http://www.protalix.com/glucocerebrosidase. html; Ratner, 2010).
One limitation shared by all nonmammalian production platforms relates to the fact that there are some striking differences in the N-glycan biosynthesis pathway compared with mammals, especially for what concerns the maturation of the N-glycans. Since correct glycosylation is of great importance for many pharmaceutical proteins, because of its effects on protein function, serum half-life, and the safety profile of the expressed proteins, attempts were made to humanize the glycosylation machinery of expression hosts.
The N-glycosylation pathway in plants is closely related to that of mammals in the sense that all the initial stages of the N-glycosylation pathway, up to the production of biantennary GnGn structures (Fig. 1), are highly conserved between plants and mammals. However, subsequent maturation of the N-glycans in plants and animals reveals important differences. Consequently, glycan chains of mammalian glycoproteins expressed in wild-type plants will undergo a different N-glycan processing. Plants lack bisecting GlcNAc, b1,4-Gal residues, sialic acid, and core a1,6-Fuc residues. Instead, they carry b1,2-Xyl and core a1,3-Fuc residues. Whereas mammalian cells can produce N-glycans with two or more terminal branches, plant N-glycans carry only two antenna structures. Although the lack of core a1,6-Fuc residues could be beneficial for effector activities of antibodies (Shields et al., 2002), the plant-specific b1,2-Xyl and core a1,3-Fuc sugar moieties may induce rapid clearance from circulation and cause a strong allergic reaction because of the presence of IgE antibodies directed against these epitopes (Gomord et al., 2010). The absence of b1,4-Gal residues is problematic for production of antibodies since these Gal residues increase complement-dependent cytotoxicity and complement binding (Raju, 2008). Another major drawback of plant N-glycosylation is the nonexistence of sialic acids on N-glycans. The sialic acids on N-glycans prevent rapid clearance of therapeutics from the bloodstream by the Gal-specific hepatic asialoglycoprotein receptor (Stockert, 1995).
In the past, plants have proven to be quite flexible in tolerating the humanization of their N-glycan biosynthesis machinery and many of the differences in N-glycosylation have been addressed. Plant-specific b1,2-Xyl and core a1,3-Fuc were removed (Koprivova et al., 2004;Strasser et al., 2004Strasser et al., , 2008Cox et al., 2006), whereas bisecting GlcNAc (Rouwendal et al., 2007), b1,4-Gal residues (Palacpac et al., 1999;Bakker et al., 2001), and sialic acid residues (Castilho et al., 2010) were introduced. However, one important structural feature still missing for the expression of fully humanized proteins in plants is the ability to synthesize multiantennary N-glycan structures. These multiantennary N-glycan structures indirectly lead to an increased serum half-life of therapeutic proteins, as shown for, e.g. human erythropoietin and folliclestimulating hormone, since the presence of multiple antennae increases the size of the molecule sufficiently to avoid rapid renal clearance. Obviously, the requirement remains that the extra branches are also modified with Gal and sialic acid, because otherwise the molecule will be rapidly cleared through the liver's receptors for nonsialylated proteins (Morell et al., 1971;Egrie and Browne, 2001).
N-Acetylglucosaminyltransferase IV (GnT-IV) and GnT-V are responsible for synthesizing multiantennary N-glycans. In humans, GnT-IV and -V activity is encoded by two genes (GnT-IVa and -IVb, and GnT-Va and -Vb), and their expression was shown to be tissue dependent (Yoshida et al., 1998;Kaneko et al., 2003). However, in plants these enzymes are not found. In this study, we introduced the gene sequences for these enzymes in Nicotiana benthamiana plants to achieve the synthesis of multiantennary N-glycans. Both GnT-IV and GnT-V introduce an extra GlcNAc to the GnGn-glycan necessary for synthesis of triantennary N-glycans. Whereas the GnT-IV enzyme transfers a GlcNAc from UDP-GlcNAc to the biantennary oligosaccharide and produces triantennary N-glycans with a b1,4-GlcNAc on the a1,3-Man arm, the GnT-V enzyme will transfer a GlcNAc from the same activated donor to biantennary N-glycans and creates triantennary N-glycans with a b1,6-GlcNAc on the a1,6-Man arm (Fig.  1). In this article, we show the successful magnICONbased transient as well as stable expression for both enzymes in wild-type N. benthamiana plants. The enzymes were targeted to the Golgi apparatus by using plantspecific localization signals (LSs). Expression of each enzyme individually generated both isomers of triantennary N-glycans, and the combined expression of GnT-IV and GnT-V resulted in both tri-and tetraantennary N-glycans. Furthermore, these modifications were also combined with the RNA interference (RNAi)-regulated knockdown of fucosyltransferase (FucT) and xylosyltransferase (XylT), resulting in multiantennary N-glycans lacking the typical plant epitopes. The successful trans- formation and expression of both enzymes again demonstrates the great flexibility of plants for altering their N-glycosylation machinery without any negative effect on plant phenotype under standard growth conditions.

RESULTS
MagnICON-Based Transient Introduction of Triantennary N-Glycans in Wild-Type and XylT/FucT RNAi N. benthamiana Plants To obtain triantennary N-glycans on the endogenous proteins of N. benthamiana, these plants were transiently transformed with the human GnT-IVa, GnT-IVb, and GnT-Va genes. These genes are responsible for the addition of GlcNAc residues on N-glycans and thus contribute to the synthesis of multiantennary N-glycan structures. The GnT genes were introduced in wild-type plants as well as in plants with downregulated expression for both XylT and FucT (RNAi;Strasser et al., 2008). To achieve the highest levels of multiantennary glycans, the catalytic domains (CDs) of the GnTs were targeted to the Golgi apparatus using the plant-specific LSs from XylT and FucT. To swiftly test the effect of these LSs for each of the GnT genes and the activity of these human enzymes in a plant system, a set of magnICON provectors were constructed: 5# provectors containing different LSs and 3# provectors containing different CDs (Marillonnet et al., 2005). All possible fusions of LS-CD were tested by agroinfiltrating plants with all possible mixtures of 5# and 3# provectors (Supplemental Table S1).
Ten days after infiltration, the transfected leaves were harvested and the composition of the N-glycans on endogenous proteins was determined by matrixassisted laser desorption ionization time of flight (MALDI-TOF) mass spectrometry (MS). The results of this analysis are summarized in Table I and show that all combinations in all plant backgrounds yielded the synthesis of GnGnGn structures, except for fucGnT-IVb expressed in the wild-type background. In the wild-type background, GnGnGn glycans decorated with Xyl, Fuc, or both were found, whereas in the RNAi background, these were absent except for a small peak of GnGnGnF glycans that is consistent with previously reported incomplete inhibition of FucT activity (Strasser et al., 2008). Furthermore, these results revealed that the amount of triantennary N-glycans was much higher in the XylT/FucT RNAi background than in the wild-type background ( Table I). The constructs containing the XylT LS yielded a higher percentage of triantennary N-glycans as compared to the FucT signal (22.3% compared to 14.6% and 20.4% compared to 11.8% for GnT-IVa and GnT-Va, respectively), except for xylGnT-IVb that yielded a slightly lower amount of GnGnGn structures compared to fucGnT-IVb in a XylT/FucT RNAi background (Table  I). This experiment indicated that both GnT-IV (-a and -b) and GnT-Va expression yielded the synthesis GnGnGn structures ( Fig. 2; Supplemental Figs. S1 and S2).
Because MALDI-TOF MS analysis does not allow to distinguish between the different conformations of the introduced GlcNAc residues, all samples were analyzed with liquid chromatography-electrospray ionization (LC-ESI) MS ( Fig. 3; Supplemental Figs. S3 and S4). The linkage of the introduced GlcNAc was determined from the difference in elution time for samples obtained after GnT-IV infiltration compared to GnT-V infiltrated leaves. Figure 3 shows the results of the LC-ESI MS analysis for the transient expression of fucGnT-IVb and fucGnT-Va in XylT/FucT RNAi N. benthamiana. GnT-IV expression yielded the triantennary glycans GlcNAcb1-2Mana1-6(GlcNAcb1-2 (GlcNAcb1-4)Mana1-3)Manb1-4GlcNAcb1-4GlcNAc-Asn (Gn[GnGn]; Fig. 1), whereas in GnT-V expressing lines the triantennary glycans were GlcNAcb1-6 (GlcNAcb1-2)Mana1-6(GlcNAcb1-2Mana1-3)Manb1-4GlcNAcb1-4GlcNAc-Asn ([GnGn]Gn-glycans; Fig.  1). The peak representing triantennary Gn [GnGn] for GnT-IV clearly eluted after the [GnGn]Gn peak for GnT-V. This divergence in elution time was indicative of the difference in conformation of the introduced GlcNAc residue. These results show unequivocally that the introduction of GnT-IV and -V in plants resulted in the synthesis of triantennary glycans according to the expected activity of the introduced enzymes.

Stable Introduction of Triantennary N-Glycans in Wild-Type and XylT/FucT RNAi N. benthamiana Plants
To test whether triantennary structures could also be produced in stably transformed N. benthamiana plants Table I. MALDI-TOF MS analysis of the N-glycans on endogenous proteins of transfected wild-type and XylT/FucT RNAi N. benthamiana plants The construct names are combinations of a LS and a GnT CD. The xyl and fuc prefixes refer to the XylT and FucT LSs. For all combinations of LS-CD, the relative abundance of bi-(GnGn) and triantennary (GnGnGn) N-glycans is indicated. /, Nontransfected background. MALDI-TOF MS analysis revealed the composition of the N-glycan pool on the endogenous proteins of the analyzed samples (Table II). A comparison of the percentage of triantennary N-glycans present on endogenous plant proteins after expression of each construct in a wild-type background and a RNAi background showed that the RNAi background yielded approximately a 10-fold increase of triantennary N-glycans compared to the wild-type background, except for the fucGnT-Va construct. Thus, the RNAi background seems more suitable for expression of the hybrid GnTs. The data also indicated an effect of the LS. The hybrid GnTs with the XylT LS yielded a higher percentage of triantennary N-glycans, except for the GnT-IVa hybrids for which both LSs yielded a high percentage of triantennary N-glycans. When comparing constructs in which different CDs of the GnTs were used, the GnT-IVa constructs scored best, followed by GnT-IVb and GnT-Va constructs. These observations correlate with those seen in the transient expression experiments.
The most abundant N-glycan structure for the xylGnT-IVa, fucGnT-IVa, and xylGnT-IVb constructs in the RNAi background was the triantennary Gn[GnGn] N-glycan structure (Table II). The glycosylation pattern of plant xylGnT-IVa RNAi exhibited only two dominant glycan varieties, biantennary and triantennary N-glycans, whereas undesired oligoman- nosidic or hybrid-type N-glycan structures were only present at very low levels (Supplemental Table S2). The specificity of the different GnTs was confirmed using LC-ESI MS analysis and showed that all GnT-IV constructs produced Gn[GnGn] glycans, whereas GnT-V constructs yielded [GnGn]Gn glycans (data not shown). Interestingly, these plants showed no aberrant phenotype under standard growth conditions, except for the plant xylGnT-IVa 48-9 RNAi that was smaller and produced no viable seeds. To obtain stably expressed tetraantennary N-glycans in XylT/FucT RNAi and wild-type N. benthamiana plants, the GnT-IV and GnT-V plants that scored best in terms of producing triantennary N-glycans were crossed (Supplemental Table S3). The transgenic line xylGnT-IVa 48-9 RNAi that yielded the highest percentage of Gn[GnGn] structures was not used since this plant did not produce viable seeds.
Seeds of all crosses were sown and plants were screened to confirm genomic insertion of the hybrid GnT constructs using real-time PCR and Southern blotting (data not shown). The selected plants were analyzed for the composition of N-glycans on endogenous proteins using MALDI-TOF MS analysis (Table  III). The results showed that tetraantennary N-glycan structures could be produced in XylT/FucT RNAi plants with all possible combinations of GnT-IV and -V. No tetraantennary N-glycans were found in the wild-type background, again indicating that the RNAi background is more suitable for expression of the human hybrid GnTs. This was not unexpected since the parent wild-type plants that were used for these crossings did not have a high level of triantennary N-glycans (6.9% for xylGnT-IVa 50-18 and 2.2% for xylGnT-Va 57-1 in the wild-type background).
The data also indicate that all LS-CD combinations used in this experiment were functional and led to an N-glycan composition on endogenous proteins desired for further humanization of the N-glycosylation pathway with Gal residues and sialic acids. The best plant (best in terms of relative percentage of bi-, tri-, and tetraantenna structures), fucGnT-IVa/xylGnT-Va 224-51, presented 10% tetraantennary, 25% triantennary, and 35% biantennary N-glycan structures ( Fig.  4; Table III). Furthermore, the glycosylation pattern of this fucGnT-IVa/xylGnT-Va 224-51 exhibited only four abundant glycan varieties, mono-, bi-, tri-, and tetraantennary N-glycans, and only very low levels of oligomannosidic N-glycan structures. In addition, the introduction of human GnT-IV and -V was combined with the silencing of the plant XylT and FucT without negative effects on the phenotype of the transgenic plants under standard growth conditions.

DISCUSSION
In this article, we describe the production of multiantennary N-glycans on endogenous proteins of N. benthamiana plants, both after transient and stable transformation of the plant. For the transient expression, we used the magnICON expression system that was developed for fast and high-yield expression of (therapeutic) proteins in N. benthamiana plants. In this study, it was demonstrated that the magnICON provector system is suitable for domain-swapping experiments with glycosyltransferases and allowed a fast optimization of glycosyltransferase localization by using combinatorial libraries of different LSs and CDs (Choi et al., 2003), although other transient expression systems could also work. Despite the fact that the magnICON expression system yields high expression levels with recombinant proteins, the yield in terms of percentage of multiantennary N-glycans present after expression of the hybrid GnTs was much lower in the transient than in the stable transformants. This observation could be the result of the fact that the N-glycans of all endogenous proteins were analyzed. In the transient expression experiments, the GnT is only expressed for a certain time period in the plant. As a consequence, some proteins were glycosylated before the GnT was expressed and thus escaped the GnT action. Since it is not possible to distinguish between proteins that went through the Golgi while the hybrid GnT was present and those that did not, a mixture of proteins that were expressed before transformation and after transformation was analyzed. This could explain the lower relative amount of triantennary N-glycans in the transient transformants.
To produce plant-made pharmaceuticals without potential immunogenic reactions when administered intravenously, N. benthamiana plants were generated in which the XylT and FucT genes were silenced via RNAi (Strasser et al., 2008). The transiently transfected and stably transformed GnT RNAi plants still exhibit some residual Fuc moieties on the endogenous N-glycans (Table III; Supplemental Tables S2 and S4; Strasser et al., 2008) that might be due to incomplete silencing.
Fortunately, these Fuc residues are not present on N-glycans of heterologously expressed proteins in the RNAi background (Strasser et al., 2008), which creates possibilities for further humanization of the N-glyco-sylation pathway in these plants. In our experiments, XylT/FucT RNAi plants exhibited 17% fucosylated N-glycans, while stably GnT-transformed XylT/FucT RNAi plants showed a significant reduction of fucosylated N-glycans (Table III; Supplemental Table S2). This trend can be observed for both XylT and FucT localized hybrid GnTs in the XylT/FucT RNAi background. A possible explanation for the reduction of the relative amount of fucosylated N-glycans could be that the introduced GnT is integrated in the Golgi apparatus at a position that is very close to the FucT enzyme (s). As a result, both glycosyltransferases are competing with each other to decorate the passing N-glycans. This was previously also observed with the expression of a hybrid b1,4-galactosyltransferase in Nicotiana tabacum (cv Samsun NN). The expression of the b1,4galactosyltransferase enzyme caused a strong reduction of N-glycans with potentially immunogenic corebound Xyl and Fuc residues (Bakker et al., 2006).
For the introduction of multiantennary structures on N-glycans of N. benthamiana plants, different components were used: Two different LSs and three different GnT-CDs were combined in all possible combinations in two different plant backgrounds. When comparing the transfections and transformations in the wild-type background and the XylT/FucT RNAi background, in terms of relative percentages of tri-and tetraantennary N-glycan structures produced on endogenous proteins (Tables I-III), it is clear that the activity of GnT-IV and GnT-V was more efficient in the RNAi plants. One explanation could be that the GnT-IV and -V constructs were targeted to the medial/trans Golgi with the XylT and FucT LSs. In wild-type plants, the XylT and FucT enzymes are present in the medial/trans Golgi (Saint-Jore- Dupas et al., 2006Dupas et al., , 2007Gomord et al., 2010), and, as a consequence, the introduced GnT enzymes will compete with the XylT and FucT enzymes for the same substrates. In the XylT/FucT RNAi plants, the XylT and FucT enzymes are absent from the Golgi. Consequently, there is no competition for the same substrates, which could explain the higher amount of multiantennary N-glycans as compared to the wild-type background. Alternatively, it is possible that the presence of Xyl and Fuc residues on the N-glycans resulted in a reduced efficiency of conversion by the GnT-IV and -V enzymes. Consequently, the XylT/FucT RNAi plants would also yield a higher relative percentage of triantennary N-glycans compared with wild-type plants.
Both the GnT-IVa and -IVb isoforms were introduced in N. benthamiana plants to achieve GnT-IV expression. Both genes are highly homologous to each other and exhibit 61% sequence identity at the amino acid level. Although they are homologs, the GnT-IVa isoform appears to be the most active form in terms of producing triantennary N-glycans, both in transient as well as in stable expression studies (Tables I and II). This difference in activity could be the direct effect of different expression levels for GnT-IVa and GnT-IVb hybrids. Our analyses showed that the expression Table III.

Results of the MALDI-TOF MS N-glycan analysis of crossed GnT-IV and GnT-V wild-type and XylT/FucT RNAi N. benthamiana plants
In the sample names, the hybrid GnT fusions of both crossed plants are given. The fuc and xyl prefixes refer to the XylT and FucT LSs. For all samples, the relative abundance of all occurring N-glycans is given. The N-glycan structures corresponding to the glycan abbreviations are illustrated in Figure 1 or see http://www.proglycan.com for an explanation of N-glycan abbreviations. WT, Wild type.  levels for the hybrid GnT-IVa constructs were higher than for GnT-IVb constructs (data not shown). GnT-IVa and -IVb expression has been studied extensively in human tissues and was shown to be tissue dependent (Yoshida et al., 1998). Nevertheless, when crossing stable GnT-Va expressing XylT/FucT RNAi plants with stable GnT-IVa or GnT-IVb expressing XylT/ FucT RNAi plants (Supplemental Table S3), similar results were obtained in terms of producing multiantennary N-glycan structures (Table III). A comparison of the different glycan profiles obtained after transient expression of each construct revealed that the constructs with the XylT LS yielded a higher relative amount of triantennary N-glycans as compared to the FucT signal, except for expression of xylGnT-IVb in a XylT/FucT RNAi background (Table  I; Supplemental Table S4). This trend was less obvious in the stable transformants for the production of triantennary N-glycans (Table II; Supplemental Table  S2). For the production of tetraantennary N-glycans, the trend got completely lost since both LSs gave similar results (Table III).
The introduction of GnT-IVa, -IVb, and -Va was combined with the silencing of the plant-specific XylTs and FucTs without any negative effects on the phenotype of the plants under standard growth conditions. There was only one XylT/FucT RNAi plant in which the stable expression of xylGnT-IVa construct yielded the highest relative amount of triantennary N-glycans but did not produce viable seeds. Whether or not these two observations can be linked is not clear, but it seems unlikely since this observation was only made for one plant on a total of 350 plants. Possibly the hybrid xylGnT-IVa gene was inserted in a DNA region that is important for seed development. Nevertheless, this study again demonstrates the flexibility of plants in terms of manipulations in their glycosylation machinery.
The presented research is based on quantifications of newly introduced tri-and tetraantennary N-glycan structures in plants. When interpreting these quantitative data, one should take into account the possibility of b-N-acetylglucosaminidases degrading the newly formed tri-and tetraantennary N-glycans. This would decrease the abundance of tetra-and triantennary structures and increase the tri-and biantennary N-glycans, respectively. In experiments with Arabidopsis (Arabidopsis thaliana) b-N-acetylglucosaminidases, we have seen that the specificity of these degradative enzymes toward b1,2-, b1,4-, and b1,6linked GlcNAcs is similar (B. Nagels, A. Dedeurwaerder, K. Weterings, N. Callewaert, and E.J.M. Van Damme, unpublished data). There is consequently no reason to believe that tri-and tetraantennary glycans would be more prone to degradation than biantennary glycans. LC-ESI MS analysis of endogenous N-glycans of the plant fucGnT-IVa/xylGnT-Va 224-51 revealed the occurrence of a small portion of N-glycans with retention times indicating them to be degradation products of tetraantennary N-glycans. Similarly, small peaks for unusual biantennary N-glycans derived by degrada-tion of triantennary N-glycan structures were seen (data not shown). Due to this degradation, the amounts of tri-and tetraantennary N-glycans produced in our transgenic plants could be underestimated by maybe 10%.
The results obtained in this study indicate that all combinations tested between LSs and CDs of enzymes were functional. The best plant (best in terms of relative percentage of bi-, tri-, and tetraantennary structures) presented 10% tetraantennary, 25% triantennary, and 35% biantennary N-glycan structures ( Fig. 4; Table III). Since the N-glycan profile of serum proteins in humans contains around 10% triantennary and very little tetraantennary N-glycans (Callewaert et al., 2003), the best plant created in this study will be sufficient for most serum-type glycosylations but further optimalization may still increase the range of applicability.
Since terminal GlcNAc residues are required for the addition of Gal and subsequently sialic acids, these plants, containing 70% N-glycans with terminal GlcNAc, are highly suitable for further humanization. These modifications would allow to produce biopharmaceuticals requiring complex glycosylation to obtain a prolonged serum half-life, such as, e.g. human erythropoietin.

Plant Material
Nicotiana benthamiana wild-type and XylT/FucT RNAi (Strasser et al., 2008) plants and GnT transformed lines were grown in soil with a 16/8-h photoperiod at 225 mmol m 22 s 21 and a temperature of 24°C to 26°C during the light period and 20°C to 22°C during the dark period.

Transient Expression Plasmids
For the introduction of multiantennary N-glycan structures in plants, six expression constructs were made based on fusions between the plant-specific Golgi LS of Arabidopsis (Arabidopsis thaliana) b-1,2-XylT or Arabidopsis FucT and the CD of human GnT-IVa, GnT-IVb, or GnT-V. The LSs were cloned into the tobacco mosaic virus-based 5# module pICH29590 and the CDs were cloned into the tobacco mosaic virus-based 3# provector pICH21595 (Marillonnet et al., 2005). The XylT LS, consisting of the cytoplasmic tail and transmembrane domain (96 bp), was amplified from the full-length Arabidopsis b-1,2-XylT gene, clone U13462 obtained from the Arabidopsis Biological Resource Center. The resulting PCR product of 123 bp was cloned into pCR2.1-TOPO (Invitrogen) and subsequently digested with the restriction enzyme BsaI and ligated into the BsaI sites of pICH29590 to create pTBN003. The FucT LS, consisting of the cytoplasmic tail and transmembrane domain (186 bp), was amplified from the full-length Arabidopsis FucTB gene, clone U16327 obtained from the Arabidopsis Biological Resource Center. The resulting PCR product of 213 bp was cloned into pCR-Blunt II-TOPO and subsequently BsaI digested and ligated into pICH29590 to generate pTBN004. The CD of GnT-IVa (1,527 bp), including the stem region, was amplified from a cDNA library derived from human HepG2 cells (Laroy et al., 2001). The resulting PCR product was cloned into pCR-Blunt II-TOPO (Invitrogen) and subsequently BsaI digested and ligated into the BsaI sites of pICH21595, generating pTBN011. The CD of GnT-IVb (1,548 bp), including the stem region, was amplified from a human placental cDNA library. The resulting PCR product was BsaI digested and ligated into the BsaI sites of pICH21595, generating pTBN010. The CD of GnT-Va (2,109 bp), including the stem region, was amplified from full-length GnT-Va codon optimized for N. benthamiana. The resulting PCR product was cloned into pCR4-TOPO and subsequently digested with BsaI and ligated into the BsaI sites of pICH21595, generating pTBN015. All primers used for generation of the provectors are listed in Supplemental Table S5. Subsequently, the 3# and 5# provectors were transformed into the Agrobacterium tumefaciens strain GV3101(pMP90). After plasmid preparation from transformed bacterial clones, the introduction of the vectors was confirmed by restriction enzyme digestion.

Plant Infiltration
For transient infiltration of all possible LS-CD combinations (Supplemental Table S1) in N. benthamiana, 10-mL Agrobacterium suspensions (grown to OD 600 2) were sedimented at 3,000g for 20 min. The pellets were resuspended in 10 mL of infiltration medium (10 mM MES + 10 mM MgSO 4 [pH 5.5] bottle top filtered [0.22 mm], containing 0.5% D-Glc and 100 mM acetosyringone). For infiltration of provector combinations, infiltration mixes consisting of a magnICON 3# provector (OD 600 0.2), a magnICON 5# provector (OD 600 0.2), and a magnICON vector containing a recombinase (OD 600 0.16) were made and resulted in in planta recombinations and fusions made between the LS and CD, allowing all possible fusions of LSs and CDs to be tested (Marillonnet et al., 2005). Leaves of 4-to 6-week-old N. benthamiana wild-type and XylT/ FucT RNAi (Strasser et al., 2008) plants were infiltrated by injecting the infiltration medium through the stomata on the lower epidermal surface of fully expanded leaves using a syringe without needle (Batoko et al., 2000). Ten days after infection, the transfected leaves were harvested.

Stable Expression Plasmids and Transformation
Six synthetic expression constructs were made (Entelechon GmbH) based on fusions between the plant-specific LS of XylT or FucT and the CD of GnT-IVa, GnT-IVb, or GnT-V. All constructs were optimized with the optimal codon use for N. benthamiana using Entelechon software. The synthetic hybrid fusions were cloned into a plant expression T-DNA vector, containing glyphosate tolerance, under the control of a cauliflower mosaic virus 35S promoter (Supplemental Table S6). The resulting recombinant vectors were transformed into the A. tumefaciens strain C58C1Rif(pGV4000). Leaf disc transformation (Regner et al., 1992) was used to transform N. benthamiana wild-type and XylT/FucT RNAi (Strasser et al., 2008) plants, resulting in 12 transgenic lines of 25 transgenic plants each.

Screening of Stably Transformed and Crossed Plants
After leaf disc transformation, transformants were first selected based on glyphosate resistance. Subsequently plants were screened by real-time PCR to confirm genomic insertion of the hybrid GnT constructs and identify singlecopy plants. Next-generation plants and plants resulting from mutual crossings were also screened by real-time PCR to confirm the presence of the gene of interest and determine the copy number for each construct. Real-time PCR was performed on genomic DNA with the TaqMan universal PCR master mix (Applied Biosystems) using the 7500 fast real-time PCR system (Applied Biosystems). In each real-time run, a primer set and a probe for the target construct (the glyphosate resistance or GnT-Va CD) as well as a primer set and a probe for the endogenous control, N. benthamiana XylTg19b gene, were used. Sequences of all primers and probes are listed in Supplemental Table S5.
In every set of analyzed samples, two single-copy references and one wildtype sample were used as control samples. The amplification data were processed with the 7500 fast system SDS software. The copy numbers of all samples were calculated using the 2 2DDCt method (Livak and Schmittgen, 2001) and confirmed by Southern blotting. The plants were further analyzed via reverse transcription real-time PCR to identify the strongest GnT expressors. Total RNA was isolated from all single-copy plants using the RNeasy plant mini kit (Qiagen) and treated with RNase-free DNase (Qiagen) to eliminate genomic DNA contamination. RNA samples (1 mg) were used for the reverse transcription reaction using the high-capacity cDNA archive kit (Applied Biosystems). Relative real-time PCR was performed, using the 7500 fast real-time PCR system (Applied Biosystems), on the prepared cDNA with the SYBR green PCR master mix (Applied Biosystems). The N. benthamiana elongation factor1a gene was used as an endogenous control to normalize the amount of cDNA. To process the amplification data, the 7500 fast system SDS software was used. The expression levels were calculated relative to a wildtype, nontransformed sample. Primer combinations against the CDs of GnT-IVa, GnT-IVb, and GnT-Va were used (Supplemental Table S5).

MALDI-TOF and LC-ESI MS Analysis of N-Glycans on Endogenous Proteins
Leaves of transformed plants (0.5-1 g) were harvested and homogenized by blending in 3 mL of formic acid. The homogenates were incubated overnight with pepsin (167 mg/mL) at 37°C. The pepsin-digested samples were centrifuged for 8 min at 4,800g and the supernatant was applied to a DOWEX 50WX2-400 cation-exchange resin (Sigma-Aldrich). The column was washed with 2% HAc and subsequently glycopeptides were eluted using 0.6 M NH 4 Ac, pH 6. Elutions were concentrated and filtered over a G25 Sephadex column (Bio-Rad Laboratories). The eluted glycoproteins were dried and sedimented. The pellet was resuspended in 100 mL of 50 mM NH 4 Ac, pH 5, and incubated for 6 min at 96°C to inactivate the pepsin. Subsequently, the glycopeptides were digested overnight using 0.075 mU of PNGaseA (Proglycan) at 37°C. PNGaseA-digested glycopeptides were applied to a DOWEX 50WX2-400 cation exchange resin (Sigma-Aldrich). N-Glycans were eluted with 2% HAc, subsequently concentrated, and further purified by reversed-phase chromatography on a Strata C-18E column (Phenomenex). The resulting free N-glycans were subjected to MALDI-TOF-MS analysis and LC-ESI MS analysis to identify and quantify all glycan structures of the transiently and stably transformed plants (Strasser et al., 2004;Pabst et al., 2007). LC-ESI MS was performed in the negative ion mode to minimize in-source fragmentation.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Table S1. Overview of all generated LS-CD fusion products by combining 5# and 3# magnICON provectors.
Supplemental Table S2. Results of the MALDI-TOF MS N-glycan analysis of stably transformed wild-type and XylT/FucT RNAi N. benthamiana plants with all GnT-IV and -V constructs.
Supplemental Table S3. Crossing scheme of stably expressing GnT-IV with stably expressing GnT-V plants, both in the wild-type background and in the RNAi background.
Supplemental Table S4. Results of the MALDI-TOF MS N-glycan analysis of transient expression of all hybrid GnTs in wild-type and XylT/FucT RNAi N. benthamiana plants.
Supplemental Table S5. Overview of all primers and probes used for PCR, real-time PCR, and reverse transcription real-time PCR reactions.
Supplemental Table S6. Synthetic constructs and expression vectors for stable transformation of N. benthamiana plants with hybrid GnT-IV and -V.