Strategies for Glycoengineering Therapeutic Proteins

Abstract

Almost all therapeutic proteins are glycosylated, with the carbohydrate component playing a long-established, substantial role in the safety and pharmacokinetic properties of this dominant category of drugs. In the past few years and moving forward, glycosylation is increasingly being implicated in the pharmacodynamics and therapeutic efficacy of therapeutic proteins. This article provides illustrative examples of drugs that have already been improved through glycoengineering including cytokines exemplified by erythropoietin (EPO), enzymes (ectonucleotide pyrophosphatase 1, ENPP1), and IgG antibodies (e.g., afucosylated Gazyva^®, Poteligeo^®, Fasenra™, and Uplizna^®). In the future, the deliberate modification of therapeutic protein glycosylation will become more prevalent as glycoengineering strategies, including sophisticated computer-aided tools for "building in" glycans sites, acceptance of a broad range of production systems with various glycosylation capabilities, and supplementation methods for introducing non-natural metabolites into glycosylation pathways further develop and become more accessible.

Keywords: N-glycans; biomanufacturing; glycoengineering; glycosylation; pharmacodynamics; pharmacokinetics; therapeutic.

FIGURE 1

Overview of mammalian N-glycosylation. (A) Step 1. The LLO 14-mer structure shown (GlcNAc₂Man₉Glc₃) is co-translationally transferred from dolichol phosphate to an asparagine residue of a nascent unfolded protein by oligosaccharyltransferase (OST) in the ER (Breitling & Aebi 2013). Step 2. Chaperone-mediated protein folding occurs concomitant with glucose trimming, generating a (in Step 3) a GlcNAc₂Man₉ or GlcNAc₂Man₈ structure that functions as an export signal for the transfer of successfully folded proteins to the Golgi (Helenius & Aebi 2001). (B) In the Golgi, further trimming of mannose residues occurs to produce a series of GlcNAc₂Man_n structures referred to as “high mannose”-type N-glycans, where n is typically between 3 and 6. (C) Also in the Golgi, one, two, or three GlcNAc residues are added to a GlcNAc₂Man_n structures, which can be further elaborated (e.g., with galactose and sialic acid, as shown) producing “hybrid” type N-glycans when a single GlcNAc is added to a GlcNAc₂Man_n structure. Hybrid N-glycans typically are low in abundance and have few known roles in therapeutic proteins. A larger proportion of N-glycans have GlcNAc residues added to both terminal mannose residues of the GlcNAc₂Man₃ structure, most frequently resulting in small biantennary structures (Werz et al., 2007) such as those shown in Panel (D), where the glycoprofile of IgG Fc domain N-glycans from one study (del Val et al., 2016) are shown rank ordered by their relative abundance. (E) A relatively small proportion (generally 5% or less) of N-glycans are further elaborated, resulting in epitopes such as (i) sialyl Lewis x (sLe^x), the H, A, and B blood type antigens [(ii), (iii), and (iv), respectively]; (v) tri- and (vi) tetra-antennary structures that can be unsialylated to fully sialylated (vii); and finally certain N-glycans have extended “LacNAc” repeats (four are shown) that can serve as preferred ligands for certain receptors, such as the hemagglutinin protein of the influenza virus (Ji et al., 2017), whereas glycans from EPO can have single LacNAc repeats (Cowper et al., 2018).

FIGURE 2

N-glycans influence the clearance of therapeutic proteins based (A) on size and (B) receptor-mediated clearance. (A) Unglycosylated EPO (top) is compared with naturally glycosylated EPO, which has three N-glycans at Asn24, N38, and N83 (middle) and with darbepoetin alfa, which has five glycans including those newly-added at Asn30 and N88 (bottom). The glycan structures depicted are representative of the experimentally-determined N-glycan profile of EPO (Cowper et al., 2018), in particular the structure shown in Figure 1E(vii). Protein models were generated using SWSS-MODEL software (Waterhouse et al., 2018) and modified to present N-glycan structures via CHARMM (Jo et al., 2008) and PyMOL (PyMOL Molecular Graphics System, Version 2.0, Schrödinger, LLC). The darbepoetin alfa sequence was obtained from the KEGG Drug database. (B) Lectin receptor-mediated clearance removes proteins from circulation through ASPGR binding of galactose-terminated glycans (top left); addition of sialic acid masks the galactose blocking binding and clearance (top right). Mannose-terminated glycans bind to mannose receptors on macrophages, dendritic cells, dermal fibroblasts, and keratinocytes, resulting in clearance or, in some cases, therapeutic activity [for example to treat Gaucher disease (Section 2.3.3)].

FIGURE 3

ENPP1 protein and glycoengineering. Improvements made to the pharmacokinetics of ENPP1 as reported by Stabach and coauthors (Stabach et al., 2021) are summarized in this figure. (A) First, in previous work (Albright et al., 2016), the enzyme was fused to the immunoglobulin Fc domain to increase protein recycling and serum recirculation through interactions with the neonatal Fc receptor (Albright et al., 2016); this “parent” construct had a serum half-life of 37 h and an AUC of 3,400 as depicted graphically in Panel (F). (B) Addition of an N-glycan site was achieved through the I256T mutation to ENPP1 resulting in addition of the glycan to Asn254; this newly added N-glycan approximately doubled serum half-life and octupled the AUC value. (C) Mutation of Met, Ser, and Thr (MST) that increase affinity for the neonatal Fc receptor (Vaccaro et al., 2005) were introduced into ENPP1-Fc Fc’s domain, further improving both serum half-life and AUC. Finally, two approaches to increase sialylation including (D) production of ENPP1-Fc in α2,6-sialyltransferase overexpressing CHO cells and (E) supplementation of the culture medium with the sialic acid metabolic precursor 1,3,4-O-Bu₃ManNAc sequentially further increased serum half-life (to a final value of 204 h) and the AUC value (to 45,000).

FIGURE 4

Carbohydrate epitopes relevant to therapeutic antibodies. (A) Sialic acid is found in human proteins in both α2,3-linkages (left) and α2,6-linkages (center); α2.6-linked sialic acid is critical for providing IgG antibodies with anti-inflammatory characteristics (Kaneko et al., 2006) whereas α2,3-linked sialic acid are effective at preventing ASPGR clearance (Ellies et al., 2002). The presence of the N-glycolylneuraminic acid (Neu5Gc, right) form of sialic acid on proteins produced in non-human mammalian cells can be pro-inflammatory (Tangvoranuntakul et al., 2003; Samraj et al., 2015), which may or may not be desired in a therapeutic protein. (B) The structure of the “α-Gal” trisaccharide epitope (left) is a major safety concern (Section 5.2.1); in human cells, the terminal alpha-linked galactose is not added to a glycan until the penultimate masking α1,2-linked fucose (right) is installed, preventing the synthesis of the “naked” immunogenic α-Gal epitope. Incidentally, the tetrasaccharide shown comprises the B-type blood antigen, whose present is a quality control parameter in IVIg therapy (Section 5.1.2).

FIGURE 5

Glycosylation-based antibody-drug conjugate (ADC) ligation strategies based on chemically-modified fucose (A) or sialic acid (B, C, and D). (A) Thiols can be installed into non-natural fucose using metabolic glycoengineering and used as “chemical handles” to ligate drug molecules to the Fc domain glycans of antibodies using thiol-reactive maleimides (Okeley et al., 2013). (B) Aldehydes can be selectively introduced into sialic acids by oxidizing the C8-OH groups; the aldehyde then can be conjugated to drugs using the hydrazino-iso-Pictet-Spengler (HIPS) reaction (Drake et al., 2014). (C) Metabolic glycoengineering can be used to install azido-sialic acids into glycans (Saxon & Bertozzi 2000), which can then be used to conjugate drugs to the antibody using dibenzocyclooctyne (DIBO) conjugation reactions (Li et al., 2014). (D) Alkyne groups can also be introduced into sialic acids through metabolic glycoengineering, which can then be conjugated using conventional copper catalyzed click chemistry (Du et al., 2009; Hong et al., 2010).

FIGURE 6

Overview of metabolic glycoengineering (MGE). Non-natural monosaccharide analogs capable of installing “chemical handles” into the N-glycans of therapeutic proteins include: (A) C6-modified fucose (B) C9-modified sialic acids, and (C) C2-modified ManNAc analogs, which are converted to N-acyl (C5) modified sialic acids before installation into N-glycans. (D) “High-flux” esterase-protected ManNAc analogs are now widely employed in MGE experiments to increase cell uptake and reduce the concentrations required for media supplementation from 30 to 75 mM (Yarema et al., 1998) to 100 μM or less (Jones et al., 2004; Kim et al., 2004; Almaraz et al., 2012).

FIGURE 7

Glycoforms of concern in bacterial, plant, fungal, and insect production systems. (A) Efforts to produce glycosylated recombination proteins in bacteria (Section 5.3.3) have resulted in the non-human glycan structure shown. (B) Mammalian N-glycans have α1,6-linked core fucose (right), which along with sialic acid, endow IgG antibodies with anti-inflammatory properties; plant cells (Section 5.3.4) produce N-glycan with α1,6-core fucose (center), and insect cells (Section 5.3.6) produce doubly-fucosylated N-glycans (right). (C) Xylose, a monosaccharide not present in mammalian N-glycans, is added to plant-produced N-glycans (Section 5.3.4). (D) Mannan synthesis in fungi (Section 5.3.5). (E) GalNAc incorporation in insects as compared to human galactose addition (Section 5.3.6).