Title thesis: Plants as a production platform for high-value proteins
Current treatments of inflammatory disorders are often based on therapeutic proteins. These proteins, so-called biopharmaceuticals, are isolated from a natural resource or, more often, made using cell based fermentation systems. The most common production platforms are based on the bacterium Escherichia coli, the yeast Saccharomyces cerevisiae or mammalian cell lines (mainly Chinese hamster ovarian (CHO) and murine myeloma (SP2/0) cells). Each platform has advantages and disadvantages and the protein to be produced largely dictates the choice of platform. Plants could provide a unique alternative production platform, as they combine the advantage of E. coli being economic with the advantages of mammalian cell lines being able to fold complex proteins, assemble heteromultimeric protein complex and, upon glyco-engineering, provide proteins with human type N-glycans. Furthermore, plant production is easily scaled up as the infrastructure is already in place due to our need for food and feed and plants have a limited risk of contamination with human pathogens. Transient transformation of N. benthamiana is a valuable plant production platform, as it is unmatched in terms of speed (matter of weeks).
This thesis describes the production of a variety of proteins and protein complexes in planta that are or may be used as biopharmaceuticals. Since 1982, plants can be genetically manipulated, which has lead to the production of many proteins in a variety of plants. Initially, the plants greatest drawback was limited protein yield. However, significant increases in yield have been achieved in the last two decades, predominantly by increasing transformation efficiency and/or level of transcription. Nowadays several plant expression systems exist that facilitate high protein production levels. Most experience is based on the production of antibodies, mainly of the IgG isotype, as these are often used as biopharmaceuticals. In general, IgG antibodies are produced on a scale of several grams per kilogram fresh weight. However, production levels of antibodies have shown to be variable and production levels of particular proteins, such as cytokines, have lagged behind. Cytokines form a large group of immune-signalling molecules and several cytokines have promising therapeutic potential. Their short in vivo half-life suggests an inherent instability, which is regarded the major production bottleneck.
In chapter 2 we describe the production of interleukin (IL)-10, a cytokine with immunosuppressive properties. We show that in contrast to mouse IL-10, human IL-10 multimerises extensively in planta. Both human and mouse IL-10 form homodimeric protein complexes through a mechanism referred to as 3D domain swapping. Only for human IL-10 3D domain swapping results in extensive multimerisation. By fusing IL-10 to green fluorescent protein (GFP) multimerisation was visualised, demonstrating that human IL-10 forms granules of organelle size. We discovered that for mouse IL-10 granule formation was prevented by N-glycosylation of the N-terminus. Introduction of this N-glycosylation site in human IL-10 partly prevented granule formation. The insertion of a glycine-serine linker between alpha helices D and E of human IL-10, allowing IL-10 to swap its own domains and form a stable monomer. This prevented granule formation completely and boosted protein yield 30-fold. However, the now comparable yields of human and mouse IL-10 were still low when compared to for instance IgG.
In the next two chapters we shifted the focus to the production of antibodies. Thus far the IgA isotype received little attention as candidate biopharmaceutical. However, the unique features of IgA, such as the ability to recruit neutrophils and suppress the inflammatory responses mediated by IgG and IgE, make it a promising antibody isotype for several therapeutic applications. Therefore, we compared the plant-based expression of IgA to IgG (Chapter 3). The variable regions of three antibodies commonly used in the clinic to treat inflammatory disorders (Infliximab, Adalimumab and Ustekinumab) were grafted on either an IgG1κ or IgA1κ backbone. Surprisingly, we achieved comparable high expression levels for all antibodies. The large variation in antibody yield found in literature is therefore likely due to differences in expression systems and experimental conditions, but not to the antibody idiotype used.
Evaluation of the secretion efficiency and N-glycosylation profiles revealed compelling differences between IgG and IgA antibodies when expressed in planta. Compared to IgG, IgA was poorly secreted. Poor secretion is most likely due to vacuolar targeting of IgA, as it was previously demonstrated that murine IgA contained a cryptic vacuolar targeting signal in the tailpiece. IgA also carried different N-glycans compared to IgG, which were never found to be a main fraction of a plant-produced protein. The predominant N-glycan on IgA lacked both the typical plant core α1,3-fucose and one terminal N-acetylglucosamine (GlcNAc). The core α1,3-fucose is most likely not added due to low accessibility of the core of the N-glycans. The lack of one GlcNAc may be due to inefficient addition of this sugar residue or (partial) β-hexosaminidase activity that may occur in both the vacuole and the apoplast. Whether the vacuolar targeting plays a role in the lack of the GlcNAc and what protein intrinsic properties influence partial GlcNAc addition or removal is unclear. We also showed that the N-glycosylation site in the IgA tailpiece does not always receive a N-glycan, which may be significant for secretory IgA formation as described in chapter 4.
In chapter 4 we describe the expression of this large heteromultimeric protein complex. Secretory IgA consists of two IgA complexes that are connected via the joining chain and associate with a part of the polymeric-Ig-receptor (called the secretory component). The challenge for sIgA expression is that assembly of four polypeptides (the alpha heavy chain, the light chain, the joining chain and the secretory component) in a 4:4:1:1 ratio is required. Previous studies on stable and transient expression of sIgA demonstrated that sIgA formation in plants is possible. However, sIgA was always accompanied by a large proportion of monomeric IgA as well as other assembly intermediates. The presence of assembly intermediates may be caused be unequal expression and/or stability of the individual components. However, we demonstrated that not the expression strategy, but protein complex assembly through impaired incorporation of the joining chain was the limiting step for sIgA production. Because joining chain incorporation depends on N-glycosylation of the IgA tailpiece we hypothesise that the partial N-glycosylation of the IgA tailpiece is the cause of inefficient joining chain incorporation. The efficiency of N-glycosylation may be improved by mutation of the glycosylation site N-X-S to an N-X-T site, as the latter glycosylation motif was shown to be more often occupied in a large-scale glycosylation site analysis.
Successful engineering of the plant glycosylation pathway to provide biopharmaceuticals with human N-glycan types has been achieved. Glyco-engineering also provides the opportunity to generate potentially interesting N-glycan types with regard to immunogenicity for vaccination purposes or protein trafficking for in vivo targeting of biopharmaceuticals. In chapter 5 we describe the expression of the Schistosoma mansoni egg antigen omega-1, an immunomodulatory protein with therapeutic potential. Its biological activity depends on its RNAse activity and its N-glycans that enable internalisation by human dendritic cells. Omega-1 cannot be isolated from natural resources in sufficient quantities to study its in vivo biology i.e. which N-glycan types facilitate full activity of omega-1. Therefore, we set up the plant-based production of this protein. S. mansoni-derived omega-1 predominantly carries diantennary N-glycans with a Lewis X motif on one or both antennae. Lewis X consists of a β1,4-linked galactose and α1,3-linked fucose residue attached to a terminal GlcNAc. This attachment is performed by β1,4-galactosyltransferase and α1,3-fucosyltransferase IXa that naturally do not occur in plants. Co-expression of these two glycosyltransferases with omega-1 resulted in a N-glycan profile comparable to S. mansoni-derived omega-1. However, it was necessary to control the expression of the β1,4-galactosyltransferase to contain this enzyme in the trans-Golgi compartment. Overexpression most likely leads to overflow of β1,4-galactosyltransferase to the medial-Golgi and galactosylation at this stage disturbs the activity of other glycosyltransferases resulting in hybrid N-glycan types.
We also observed that more than 90% of omega-1 was secreted to the apoplast, which facilitated efficient purification. This demonstrated that plants, as previously suggested, are not poor protein secretors. However, the underlying protein properties controlling secretion efficiency are currently unknown. Both omega-1 and IgG are very efficiently secreted under natural circumstances. Yet, despite the fact that we used the same signal peptide to facilitate secretion of IgG and omega-1, IgG was only secreted from plant cells to a maximum of 28%. More knowledge on protein secretion efficiency in plants may overcome cumbersome purification, currently the greatest bottleneck in the downstream processing of plant produced proteins.
Finally in chapter 6, we shifted focus from the expression of a particular protein to the effect of codon use on protein yield. We designed a codon optimisation strategy that, unlike other strategies, turned out to be surprisingly robust. This strategy is based on a general codon bias found in plants. Because this codon bias could be found among plant species, including monocots and dicots, and resulted in an increase in mRNA stability and translatability, we suspected that this codon bias arises from a selection pressure on the mRNA structure. We extended this general codon bias to representative species of other kingdoms of life and demonstrate that there is a selection pressure increasing mRNA stability and translatability (more protein per mRNA molecule). Stability was the result of an increase in the number of nucleotide bonds. However, there is a trade off between mRNA stability and translatability, and the nucleotide bonds in an mRNA should be well balanced over the entire molecule, making it ‘airy’, to ensure efficient translation.
Altogether we conclude that transformation efficiency and level of transcription are no longer limiting factors for protein yield upon plant-based expression. However, an increase in translation initiation and translation rate dictated by codon use may still provide an increase in protein production. Furthermore, we show that many proteins demonstrated specific production bottlenecks. The yield of human IL-10 was hampered by its extensive multimerisation, sIgA assembly is most likely limited by inefficient N-glycosylation of the tailpiece of IgA, both IgG and IgA are inefficiently secreted compared to omega-1 and IgA displayed an aberrant N-glycosylation profile. Therefore, more knowledge on how protein intrinsic properties influence protein yield and/or quality of heterologous produced proteins in plants should now be generated. Transient expression in N. benthamiana is an ideal tool to study these protein intrinsic properties that limit protein yield in heterologous expression.