Promoter Choice Impacts the Efficiency of Plant Glyco‐Engineering

Glyco‐modulation of therapeutic proteins produced in plants has shown great success. Plant‐based expression platforms for tailored human‐like N‐glycosylation are based on the overexpression of foreign genes. However, drawbacks such as protein miss targeting, interference with endogenous glycosyltransferases, or with plant development hamper the widespread use of the technology. Here a technique that facilitates the generation of recombinant proteins with targeted N‐glycosylation at high homogeneity is described. It is focused on the synthesis of human‐type β1,4‐galactosylation by the overexpression of the human β1,4‐galactosyltransferase (GalT) in Nicotiana benthamiana. A GalT construct that targets the enzyme to the required late Golgi compartment (STGalT) is transiently co‐expressed with two pharmaceutically relevant glycoproteins. The impact of eight promoters driving the expression of STGalT is evaluated by mass spectrometry (MS) ‐based analyses. It is shown that five promoters (amongst them high expressors) induce aberrant non‐human glycosylation. In contrast, three promoters, considered as moderately active, regulate gene expression to levels leading to an improved efficiency of di‐galactosylation (and subsequent sialylation) on the reporter proteins. The results point to the importance of promoter choice for optimizing glycan engineering processes.


Introduction
Over the last decade, plants have emerged as a convenient, safe, and economical alternative system for the large-scale production of glycoproteins with therapeutic value. [1,2] The impact of different glycan modifications on the function of recombinant glycoproteins has been extensively reviewed highlighting the important advances in the field of glycobiology. [3] Functional roles of glycoproteins are often directly related to specific N-glycan structures. [4] Therefore, the efficacy of therapeutic glycoproteins can be maximized through modification of the N-glycan profiles.
In plants, numerous efforts have been undertaken to engineer therapeutic proteins with appropriate glycans to enhance their activity. [5,6] Plants are particularly amenable to glyco-engineering and allow the reconstruction of entire human glycosylation pathways. [6] Plant glyco-engineering encompasses the elimination of particular glycosidic linkages and the introduction of new ones. Attempts to establish plant-based expression platforms to tailor human-like N-glycosylation on recombinant proteins started with generation of mutants in different plant species that lack b1,2-xylosylation and core a1,3-fucosylation. [6] The incorporation of new enzymatic reactions in plants may be accomplished by genomic insertion or transient expression of foreign genes coding for specific glycosylation proteins. However, such proteins sometimes need further modifications to act in a targeted way. For example, the correct subcellular localization of glycosyltransferases, directed by its cytoplasmic tail, transmembrane domain and stem (CTS) region, [7] has profound implications on the final N-glycan profile. This was well illustrated by the efforts to achieve b1,4galactosylation in plants where incorrectly located b1,4-galactosyltransferases (GalT) resulted in the generation of aberrant (hybrid) structures. [6] Fully processed di-antennary b1,4-galactosylated structures were accomplished by targeting GalT to a late Golgi compartment using a chimeric version consisting of the CTS region of rat a2,6-sialyltransferase (ST) fused to the catalytic domain of GalT ( ST GalT). [8] Despite the enormous success, overexpression of proteins for glyco-modulation is not always straightforward. There are examples in the literature reporting drawbacks of gene overexpression including alterations on plant development and mislocalization of the target protein. [9][10][11] Previous data demonstrate that phenotypical modifications may occur upon stable in planta expression of foreign glycosyltransferases. Stable transgenic plant lines expressing ST GalT, exhibit phenotypes associated with the enzyme expression levels. [9] Galactosylated N-glycans, the acceptor substrate for protein sialylation, were generated either by transient or stable expression of ST GalT under the control of the strong 35S promoter. [8] In both cases, upon co-expression of the necessary enzymes to introduce a sialylation pathway (Sia), recombinant proteins were produced with a mixture of di-and monosialylated structures. [12,13] Incomplete processed glycans were attributed to the interference of ST GalT with endogenous proteins. [14] Such studies demonstrate that glyco-engineering is not a simple overexpression of a gene, but requires fine-tuning of various parameters such as correct subcellular localization and gene expression level. Promoter selection has become increasingly important to maintain appropriate level of transgene expression. [11,15] The availability of a broad spectrum of promoters that differ in their ability to regulate gene expression can dramatically increase the success of the technology for glycoengineering. Common promoters used in transgenic plants are derived from Agrobacterium tumefaciens, plants and from plant viruses. [15,16] The most widely used promoter is the cauliflower mosaic virus (CaMV) promoter of the 35S RNA. Promoters from A. tumefaciens nopaline synthase (nos), octopine synthase (ocs), and mannopine synthase (mas) genes are the most commonly used. Finally, plant constitutive promoters isolated from the highly expressing ubiquitin, actin, and rubisco genes provide valuable alternatives to viral and bacterial promoters.
Here we investigated the impact of different promoters driving ST GalT expression on the generation of human-type galactosylated N-glycans on plant-derived glycoproteins. We used eight different promoters to transiently express ST GalT in Nicotiana benthamiana and group them according to their capacity to modulate the glycosylation profile of two reporter proteins. We have established three groups of promoters that generate distinct galactosylation and sialylation profiles. Our findings clearly point to the importance of choosing the correct promoter to obtain targeted structures.

Plant Material
Nicotiana benthamiana wild-type (WT) and the glycosylation mutant DXT/FT plants, with down-regulation of plant-specific N-glycan residues (core a1,3-fucose and b1,2-xylose), [17] served as host for the expression of recombinant proteins. Plants were cultivated in a growth chamber at a constant temperature of 24 C, 60% relative humidity, and a 16 h light/8 h dark photoperiod.

Expression Vectors
The binary vectors used for transient expression of the IgG1 monoclonal antibody (4E10) and alpha-1-antitrypsin (A1AT) were described previously. [8,18] The two reporter glycoproteins were co-expressed with a chimeric human b1,4-galactosyltransferase that targets the enzyme to a late Golgi compartment ( ST GalT [8] ). For sialylation, the reporter glycoproteins were additionally co-expressed with the mammalian genes for biosynthesis of CMP-sialic acid, Golgi transport and transfer of sialic acids to the protein (Sia). [12] To transiently express ST GalT we generated a set of binary vectors where gene expression is driven by promoters from different origins including viral, agrobacterium and plants. Corresponding promoter fragments ( Figure 1A) were PCR amplified and used to replace the 35S promoter of the original ST GalT expression vector. [8] In some constructs, the terminator sequence of the original vector was also substituted as depicted in Figure 1A.
Viral promoters include an 835 base pair (bp) fragment of CaMV35S promoter (35S), an enhanced variant with a tandem duplication of 231 bp of upstream sequences plus translational enhancer 5 0 -UTR from tobacco etch virus (2Â35S) and a 346 bp long fragment from the cestrum yellow leaf curling virus (CmYLCV) promoter (CmpC). [19] The promoter fragments were isolated by PCR from available binary plasmids. [19,20] A. tumefaciens promoters used to drive ST GalTexpression were from the octopine (ocs, 360 bp) and mannopine (mas, 387 bp) synthase genes. The promoter fragments were isolated from commercially available cloning vectors [20] using PCR.
Finally, plant promoters comprising constitutive promoters from Arabidopsis thaliana actin (Act, 1216 bp), ribulose-1,5bisphosphate carboxylase-oxygenase (Rubisco) small subunit (RbcS1, 1002 bp), and ubiquitin 10 (Ubi10, 628 bp) were isolated by PCR either from A. thaliana genomic DNA or from commercially available vectors. [20] All PCR fragments were amplified with primers with additional 5 0 -HindIII and 3 0 -XbaI restriction sites and inserted into the original ST GalT expression vector digested in the same way.

Expression of 4E10 and A1AT
Four-to-five week old plants were used for expression of 4E10 and A1AT by agroinfiltration. Agrobacteria containing binary constructs for the reporter glycoprotein were infiltrated at an OD 600 of 0.3 and glyco-modulating proteins at an OD 600 of 0.05, unless stated otherwise (1.0 OD 600 corresponds to 5 Â 10 8 cells mL À1 ). Three days post-infiltration 4E10 was purified by protein A based affinity chromatography [8] and A1AT was collected from the intercellular fluid (apoplast). [18]

Glycan Analysis
The N-glycan composition of 4E10 and A1AT was determined using reversed-phase liquid chromatography electrospray ionization mass spectrometry (LC-ESI-MS) of tryptic glycopeptides as described previously. [21]

Subcellular Localization by Confocal Laser Scanning Microscopy
For subcellular localization experiments, ST GalT was Cterminally tagged with green fluorescent protein (GFP) as described previously. [12] The fusion construct was transiently www.advancedsciencenews.com www.biotechnology-journal.com expressed in N. benthamiana leaf epidermal cells at different OD 600 (0.2, 0.05, and 0.005). GFP expression was monitored at two dpi using an upright Leica TCS SP5 confocal laser scanning microscope (CLSM). Post-acquisition image processing was performed in Adobe Photoshop CS6.

Results
Previous studies have provided evidence that different CTS domains fused to human b1,4-galactosyltransferase lead to distinct sub-Golgi location of the enzyme with a great impact on the generation of galactosylated N-glycans in plants. [8,22] In addition to the CTS region, the expression level of GalT seems to correlate with aberrant galactosylation. [9] We aimed to investigate the impact of promoter choice in controlling ST GalT expression toward a homogenous galactosylation profile. To this end, we transiently co-expressed a monoclonal antibody 4E10 with ST GalT under the control of eight different promoters. Well described constitutive promoters of different origin were chosen, encompassing 35S, 2Â35S and CmpC (viral-origin), ocs and mas (bacterial-origin), Act, RbcS1, and www.advancedsciencenews.com www.biotechnology-journal.com Ubi10 (plant-origin, Figure 1A). 4E10 was immune affinity purified and glyco-profiling was performed by LC-ESI-MS. The results showed that ST GalT driven by all promoters is able to galactosylate 4E10 (up to 82%). However, overall glycosylation profile of 4E10 depended on the promoter. According to the generated glycoforms, the promoters were clustered in three groups: Group 1 (2Â35S and RbcS1) led to the most heterogeneous glycosylation profile of 4E10, harboring a mixture of at least seven glycoforms at similar levels. In addition to complex di-and mono-galactosylated (28% of AA and GnA) N-glycans, aberrant galactosylation (up to 50%, Man5A and Man4A, MA) were also detected ( Figure 1B). Using group two promoters (35S, CmpC and Ubi10), 4E10 clearly showed a dominant glycoform, namely di-galactosylated glycans (AA) which, together with complex mono-galactosylated structures (GnA) account of up to 50% of the overall glycosylation. Nevertheless, significant amounts (up to 35%) of hybrid and incompletely processed galactosylated forms (Man4A, MA) are also detected. When ST GalT is driven by group three promoters (mas, ocs, and Act), the galactosylation profile is more homogenous and 4E10 is decorated mainly with fully processed di-galactosylated (AA) structures (up to 46%). Together with complex mono-galactosylated glycans they account for 67% of all glycoforms. In addition, minor amounts of incompletely processed structures (17%, MA) but, no hybrid glycans were detected ( Figure 1B). The galactosylation profile of 4E10 originated from ST GalT driven by group three promoters mirror that of the native serum IgGs. Notably, 4E10 carry similar levels of non-galactosylated structures (MGn, GnGn, 20-30%) independently of the promoter. It is possible that these structures arise from cells not expressing ST GalT. Since galactose-terminated glycans are the acceptor substrate for the transfer of sialic acid we evaluated if the galactosylation profiles created by different promoters would account for differences in sialylation. It has been previously demonstrated that IgG sialylation is enhanced by core plant fucosylation. [23] For this reason, we used WT plants to co-express 4E10 with the genes necessary for protein sialylation [12] and the eight ST GalT variants described above. Sialylation using ST GalT driven by promoters from groups 1 to 2 was very similar. 4E10 was decorated mainly with di-sialylated glycans carrying core fucose and xylose (up to 35%, NaNaXF). Sialylated incompletely processed and hybrid structures account for about 37% of the glycoprofile (Figure 2). Contrastingly, 4E10 of group 3 driven ST GalT exhibit almost exclusively di-sialylated (>90%, NaNaXF) N-glycans ( Figure 2).
To investigate if these results can be extended to other proteins, 4E10 was replaced by human a1-Antitrypsin (A1AT) and DXT/FT glycosylation mutant was used as expression host. [17] MS profiles of recombinant A1AT show that promoters from groups 1 to 2 produce similar glycosylation profiles ( Figure 2). A1AT glyco-profiles using ST GalT under control of promoters from groups 1 to 2 exhibit di-sialylated forms (15 and 42%, respectively) with significant amounts of incompletely processed MNa (34 and 24%, respectively) and non-galactosylated forms. In contrast, group 3 promoters led to a homogenous di-sialylated A1AT profile (up to 71% NaNa, Figure 2).
Next, we evaluated if the use of different terminators could alter the galactosylation efficacy of 4E10. The terminator in the 2Â35S:: ST GalT (group 1) was substituted by ags and rbc, while the 35S and g7 terminators replaced the terminator on Act:: ST GalT (group 3) ( Figure 1A). No obvious variation of the galactosylation and sialylation efficiency of 4E10 was observed (data not shown).
Collectively, the results indicate that ST GalTdriven by groups 1 and 2 promoters interferes with the activity of endogenous enzymes thus generating unusual hybrid glycan structures, a phenomenon observed earlier by using CTS domains that target the enzyme to early/medial Golgi cisternae or when the expression level of ST GalT exceeds a certain threshold. [9,22] We speculate that the amount of agrobacterium used in agroinfiltration experiments was having an impact on the ST GalT expression level. To determine if the expression level of ST GalT could be regulated by changing the amount of agrobacterium in the infiltration mixture, we co-expressed 4E10 with a ST GalT construct representative of each group at different OD 600 (0.005-0.2). LC-ESI-MS analysis of purified 4E10 showed that the relative abundance of galactosylated versus non-galactosylated glycoforms increased with the concentration of ST GalT ( Figure 3A). Decreasing concentration of ST GalT (OD 600 0.2-0.005) resulted in higher amounts on non-galactosylated glycans (MGn and GnGn): 26-81% (group 1); 27-67% (group 2); and 33-62% (group 3).
Increasing OD 600 (0.005-0.2) of group 3 ST GalT resulted in an increase of galactosylation from 38 to 67% (AA and GnA) but, did not induce the synthesis of aberrant galactosylation (Man5A or Man4A) ( Figure 3A). It seems that decreasing the amount of ST GalT construct in the infiltration mixture reduces the number of expressing cells but, does not modulate the ST GalT expression level within a transformed cell. This was confirmed by analysis of the expression of ST GalT fused to GFP driven by the 35S promoter (35S:: ST GalT-GFP, Figure 3B) and infiltrated at different OD 600 (0.2-0.005). Despite not being a quantitative method, confocal laser microscopy revealed that the number of cells (or Golgi stacks per cell) expressing ST GalT is reduced when lower OD 600 is used for infiltration ( Figure 3B), thus accounting for an increase of nongalactosylated structures on 4E10.

Discussion
Methods to transiently express high levels of recombinant proteins and simultaneously modulate their glycosylation pattern in plants have been previously described [5] and a stepby-step protocol was recently established. [24] The approach allows generating recombinant proteins with a largely targeted glycosylation profile. It is, however, necessary to point out the importance of fine-tuning gene expression for effective glycoengineering. For N. benthamiana, these steps include the correct subcellular localization of the foreign enzymes, [8,11] elimination of endogenous glycosylation enzymes, [17,25] and/or overexpression of exogenous glycosyltransferases. [14,23] Despite the great achievements, efficient N-glycan processing in vivo may sometimes need further "polishing". Previously we reported on the phenotypic variation of transgenic plants stable expressing a chimeric human b1,4-galactosyltransferase ( ST GalT [8,9] ), which was correlated to GalT expression levels. It also became clear that ST GalT expression over a certain threshold www.advancedsciencenews.com www.biotechnology-journal.com negatively impacts the generation of fully processed glycans and induce aberrant glycosylation. From the numerous factors that can influence gene expression, promoters are perhaps the most important player. Although strong promoters are beneficial to increase yields of recombinant proteins, they might not always be appropriate. Here we show that the ability of ST GalT to decorate glycoproteins with galactose (and consequently with sialic acid) largely depends on the promoter used to drive its expression. This observation was valid for two tested glycoproteins (4E10 and A1AT). The relative high amounts of incompletely processed N-glycans (Man5A and Man4A) indicate that promoters from group 1 most probably induce high expression of ST GalT, which, in turn competes with endogenous Golgi-a-mannosidase II (GMII) for the acceptor substrate. The same seems to be true for group 2 promoters, however, to a lesser extent. Interestingly, previous results have shown a comparable GFP expression when driven by CmpC, 35S, and ubiquitin promoters, [19] all placed in group 2 in this study. Also, fine tuning of the ST GalT expression by replacing the 2Â35S for a weaker constitutive promoter (Gpa) was required to achieve the synthesis of glycans carrying mono-and bi-antennary Lewis x motifs. [11] Group 3 promoters appear to induce ST GalT expression at levels most suitable for the synthesis of complex di-galactosylated (and as a consequence di-sialylated) glycans at large homogeneity. Collectively, we demonstrate that glycoengineering through overexpression of foreign glycosylation enzymes requires optimized gene expression, at least for galactosylation, a fact that has not been addressed sufficiently previously. Many efforts have been applied to precisely target glyco-modulating proteins to correct subcellular compartments Figure 2. N-glycan profiling of two recombinant glycoproteins co-expressed with the genes necessary for protein sialylation (Sia [12] ) and different constructs for ST GalT expression. Left, 4E10 was expressed in WT plants. Right, Alpha-1-antitrypsin (A1AT) was expressed in N. benthamiana DXT/FT and glycosite 3 (GP3) is shown as an example (similar results were obtained for GP2 and GP1). Representative spectra from the three groups of promoters are shown. Peaks and glycan cartoons were labelled in accordance to the Consortium for Functional Glycomics guidelines (www. functionalglycomics.org).
www.advancedsciencenews.com www.biotechnology-journal.com but, little attention has been paid on how to control their expression levels. Transient expression systems have been developed for rapid analysis of promoter function, and although the system may not provide definitive information on gene expression patterns, it is powerful tool to gauge promoter strength. [16] With this investigation we point to the importance of the choice of promoter to drive the expression key enzymes on plant glycoengineering. In comparison to other glycosyltransferases, GalT has a broader acceptor substrate specificity and efficiently acts on all substrates carrying a terminal b-linked GlcNAc. [26] This makes controlling the sub-cellular localization and expression of this enzyme a crucial step for successful glyco-engineering. With this optimized technology we contribute to the establishment of versatile plant expression systems as a platform for the production of a broad range of recombinant glycoproteins. www.advancedsciencenews.com www.biotechnology-journal.com