Spoonful of sugar helps the medicine go down: biomanufacture in glycoengineered Pichia pastoris of the potentially therapeutic recombinant glycoprotein factor H
Devlin, John Patrick
Glycoengineering is a technology that could improve protein therapeutics. While protein glycosylation in general enhances solubility and stability, and reduces aggregation, immunogenicity and proteolysis, specific kinds of glycosylation may also be critical. For example, capping of glycans with N-acetylneuraminic acid (Neu5Ac) maximises circulatory half-life in humans. Moreover, some glycans directly participate in molecular recognition and other aspects of glycoprotein function. Glycoproteins produced by non-human mammalian cells carry glycans capped by N-glycolyl-neuraminic acid rather than Neu5Ac. Yet production in human cell lines is costly and slow, requires specialist facilities, produces low yields and is subject to additional regulations. Hence there is a case for glycoengineering alternative expression systems capable of rapid, low-cost, high-yield glycoprotein production. This report focuses on the glycoengineering of Pichia pastoris, a yeast, to produce recombinant human glycoprotein factor H (FH) bearing human-like glycans. FH is a potent down-regulator of the complement system. Mutations and SNPs in FH result in autoimmune diseases such as atypical haemolytic ureamic syndrome and age-related macular degeneration (AMD). Recombinant FH is an enticing therapeutic candidate for treating AMD, but high doses are required since FH is abundant (200-300 mg l-1) in normal human serum. Human FH (155 kDa), with eight sites of N-linked glycosylation and 40 disulphides, is a challenging target for recombinant production. Yet FH was previously expressed to 10s of milligrams in P. pastoris. In this study, methods were established to confirm that human plasma-derived (h)FH carries predominantly N-linked diantennary disialylated complex-type glycans, with monosialylated diantennary structures and triantennary structures in fucosylated and non-fucosylated forms, contributing to glycan heterogeneity. Functional comparison of native hFH, enzymatically desialylated (DeSia-) hFH and deglycosylated recombinant P. pastoris-produced (DeGly-r)FH showed that DeSia-hFH had the lowest affinity for complement protein C3b, its key target. Moreover, DeSia-hFH binds C3d, an opsonic C3b-breakdown product, whereas native hFH does not. DeSia-hFH had an improved ability to accelerate decay of the C3 convertase (an enzyme that cleaves C3 to C3b) compared to native hFH, but neither was as good as DeGly-rFH in this respect. In contrast, DeGly-rFH had reduced cofactor activity (for factor I-mediated degradation of C3b) compared to native hFH whereas DeSiahFH did not have reduced cofactor activity. These data suggest that sialylation of FH glycans may play a role in stabilising a conformation of circulating FH that is not fully effective, consistent with specificity for self-surfaces and resistance to bacterial hijack. Aiming eventually to produce human-like glycosylated FH in glycoengineered P. pastoris, the SuperMan 5 strain served as a starting point. While conventional strains of P. pastoris put hypermannosylated N-linked glycans on proteins, glycans on SuperMan 5-produced FH were shown to contain just five mannose (Man) residues. In further glycoengineering, and following unsuccessful efforts to use inABLE technology for this purpose, commercially available (GlycoSwitch) vectors were used to introduce genes encoding the glycosyltransferase enzymes N-acetylglucosamine (GlcNAc) transferase I (GnTI) and galactose (Gal) transferase. These catalysed the formation of a hybrid-type glycan containing an N-acetyllactosamine (Gal-β(1,4)-GlcNAc (LacNAc)) antennae on a five-mannose glycan. Then two more GlycoSwitch plasmids, containing genes encoding α-Mannosidase II (ManII) and GnTII, were introduced into P. pastoris to catalyse the formation of a second LacNAc antennae. MALDI-TOF analysis found the glycosylation of this strain to be heterogeneous, containing the humanised diantennary digalactosyl glycan as well as other endogenous yeast glycans. This strain was designated SuperGal. Large-scale expression of rFH with terminally galactosylated complex-type glycans (Gal-rFH) in SuperGal yielded 100s of milligrams of purified Gal-rFH. Yeast-type glycans were enzymatically removed from rFH and the remaining complex-type humanised glycans were sialylated with a recombinant bacterial α(2,6)-sialyltransferase from Photobacterium sp. expressed in E.coli. Purified sialylated (Sia-) and non-sialylated (Gal-) rFH expressed in SuperGal were functionally characterised in vitro using SPR-based assays. In C3b-binding assays Sia-rFH had lower affinity compared to Gal-rFH. Both bound with lower affinity than DeGly-rFH. A similar pattern of binding affinity was seen for C3d. In C3 convertase decay-acceleration assays, all rFH glycoforms performed equally well and had greater activity than hFH. Conversely, Sia-and Gal-rFH were shown to perform equally as well as hFH in CA assays, while all three versions outperformed DeGly-rFH. However, in vivo complement activity assay carried out in a FH-knockout mouse model showed that humanisation of the glycosylation of rFH did not significantly improve activity compared to DeGly-rFH. In addition, analysis of the circulatory half-life of rFH showed that humanisation did not improve half-life. Further engineering steps will be required to increase the complex-type glycan site occupancy on rFH with a view to improving circulatory half-life and efficacy. However, this study represents a significant step forward in developing a therapeutically useful source of rFH.