Engineering Aggregation Resistance in IgG by Two Independent Mechanisms: Lessons from Comparison of Pichia pastoris and Mammalian Cell Expression

Abstract

Aggregation is an important concern for therapeutic antibodies, since it can lead to reduced bioactivity and increase the risk of immunogenicity. In our analysis of immunoglobulin G (IgG) molecules of identical amino acid sequence but produced either in mammalian cells (HEK293) or in the yeast Pichia pastoris (PP), dramatic differences in their aggregation susceptibilities were encountered. The antibodies produced in Pichia were much more resistant to aggregation under many conditions, a phenomenon found to be mainly caused by two factors. First, the mannose-rich glycan of the IgG from Pichia, while slightly thermally destabilizing the IgG, strongly inhibited its aggregation susceptibility, compared to the complex mammalian glycan. Second, on the Pichia-produced IgGs, amino acids belonging to the α-factor pre-pro sequence were left at the N-termini of both chains. These additional residues proved to considerably increase the temperature of the onset of aggregation and reduced the aggregate formation after extended incubation at elevated temperatures. The attachment of these residues to IgGs produced in cell culture confirmed their beneficial effect on the aggregation resistance. Secretion of IgGs with native N-termini in the yeast system became possible after systematic engineering of the precursor proteins and the processing site. Taken together, the present results will be useful for the successful production of full-length IgGs in Pichia, give indications on how to engineer aggregation-resistant IgGs and shed new light on potential biophysical effects of tag sequences in general.

Introduction

Antibodies and their derivatives have found a broad range of applications, from basic research to medical therapy. With the use of many different techniques for their generation,¹ several therapeutic immunoglobulins, usually in the whole immunoglobulin G (IgG) format, have been approved by the Food and Drug Administration and applied to the treatment of a wide range of diseases.2, 3, 4 One major challenge encountered in the development of antibody-based therapeutics, however, is their aggregation susceptibility under both formulation conditions5, 6 and in vivo, where aggregation is involved in a significant degradation pathway for monoclonal antibodies.⁷ Irreversible aggregation thereby leads to reduced bioactivity⁸ and presents an increased risk of immunogenicity.9, 10, 11, 12, 13, 14 Aggregation of even only a few percent of the administered therapeutic antibody can cause severe problems, as already low amounts of aggregates have long been known to cause anaphylactoid reactions.15, 16 Therefore, engineering IgG to reduce or eliminate its self-association is of great importance and might further extend the potential of therapeutic antibodies.

Since nearly all recombinant therapeutic antibodies are expressed in the IgG format, choices must be made in the molecular composition and expression system. Currently, all approved therapeutic antibodies are produced in mammalian cells,¹⁷ since only this expression system can introduce the complex glycosylation that is necessary for the interaction of the antibodies with some of the activating Fcγ receptors (FcγR).¹⁸ IgGs normally possess a single N-linked glycosylation at Asn297 in the C_H2 domain, maintaining these domains at a critical distance to bind to the FcγR.

In some cases, however, this FcγR interaction is neither needed nor wanted, thus opening the range of possible production hosts also to Pichia pastoris (PP). In the past, this methylotrophic yeast has been widely used for the expression of both secreted and intracellular eukaryotic proteins.¹⁹ While it is able to carry out efficient disulfide bond formation and isomerization using the eukaryotic secretion quality control machinery,19, 20, 21 Pichia expression generates mannose-rich glycosylation of IgG.²² The attached glycan moieties differ markedly from the product of mammalian cells, resulting in antibodies that lack FcγR binding capacity. However, recently, Pichia has also been engineered to introduce complex, human-like glycosylation.23, 24, 25, 26

Due to its respiratory growth, Pichia can grow to very high cell densities²⁷ and therefore express high levels of recombinant protein through efficient and tightly regulated promoters, accounting for up to 30% of the total cell protein.²⁸ Moreover, comparatively few endogenous proteins are secreted by this yeast system, which offers further advantages for subsequent protein purification and downstream processing.²⁹ Since Ridder et al. reported the first successful production of functional antibody fragments in Pichia,³⁰ more than 50 reports describing antibody expression in this system have been published.³¹ While most of these molecules were single-chain variable fragments and fusions thereof, several Fab fragments have been produced in Pichia as well.32, 33, 34 Interestingly, until today, Pichia has only been used as an expression host for a handful of full-length IgGs.25, 26, 35

Almost all heterologous proteins produced in Pichia have been expressed under the control of the methanol-inducible alcohol oxidase 1 (AOX1) promoter.20, 36 However, throughout this study, the constitutive glyceraldehyde-3-phosphate dehydrogenase (GAP) promoter³⁷ was used, since the ability to compare mutant proteins for their expression and aggregation tendencies was crucial. The GAP promoter is usually considered to result in lower protein yields compared to AOX1,³⁸ even though some reports describe the GAP promoter to be superior to AOX1 for certain proteins,39, 40 including Fab fragments.⁴¹ A potential reason could be the lower rate of expression that might better match the corresponding folding rate and, thus, reduces the stress level for the cells.

Previously, our laboratory analyzed sequence modifications in the variable domain in various antibody formats for aggregation propensity and conformational stability.42, 43, 44 Eukaryotic expression systems, where functional expression yields can be compared and mutant proteins can be isolated and biophysically characterized, are required to carry out such studies for whole IgGs. Therefore, it is vital that isogenic strains can be produced, only differing by the sequence of the protein to be tested. As a first step, we investigated homologous recombination in HEK293 cells⁴⁵ and Pichia. As Pichia does not maintain episomal vectors, the expression cassettes are generally designed for integration into the yeast genome, thus leading to stable cell lines expressing the protein of interest (POI). In initial experiments, we were very surprised to find large differences between the aggregation behaviors of purified IgG protein from both systems, even though their amino acid sequence was identical: the same recombinant IgG expressed in HEK293 cells was much more aggregation prone than its counterpart from Pichia.

Here, we describe our investigations of this phenomenon and pinpoint it to differences in the processing of the precursor proteins, as well as to differences between the unequal glycosylations. As an initial model system, an antibody that had originally been selected from the HuCAL library⁴⁶ by panning against myoglobin from horse skeletal muscle and subsequently been stability engineered was used.⁴⁷ For this study, it was converted to a full-length IgG and investigated for its processing and expression in mammalian HEK293 cells and in the yeast Pichia pastoris.