Aggregation, immune complexes and immunogenicity

The development of an immune response to a protein therapeutic may nullify its beneficial activity or result in adverse events. Immunogenicity is, therefore, a major concern for clinicians, regulatory authorities and the biopharmaceutical industry. These concerns are particularly acute for the treatment of chronic diseases, as opposed to cancer, that may require repeated exposure to therapeutic over extended cycles of remission/relapse. There are many parameters that may be contributory to immunogenicity; however, the “bête noire,” for the past decade has been aggregation.^1–3

The experimental basis for this assertion rests with early experimental observations, e.g., (1) prepare a solution of bovine serum albumin (BSA) and spin at high speed overnight; carefully harvest the top third of the tube (devoid of aggregates!) and administer intravenously to a rabbit, and the result is immune tolerance; (2) Re-suspend the pellet at the bottom of the tube and similarly administer, and the result is a vigorous anti-BSA response.⁴ An appreciation of this fundamental finding has recently been exploited in an attempt to circumvent the immunogenicity of alemtuzumab (Campath-1H), a humanised anti-CD52 antibody.⁵ A variant of this antibody with a charge reversal, Lys⁵³ to Asp⁵³ in the H2 loop of the complementarity-determining region (CDR), was shown to essentially abrogate binding to the CD52 antigen. Administration of a high dose of this variant to mice transgenic for the human CD52 antigen resulted in the induction of long-lasting tolerance (high dose tolerance) to subsequent cycles of alemtuzumab administration. The essential difference between the variant and alemtuzumab is that it did not form immune complexes although expressing five unmodified CDRs. A fundamental feature of an immune response is that immune complexes, formed in the initial phase of a response, can heighten the response to the target antigen.^6,7

This poses the question: What is the difference between aggregated forms of IgG and immune complexes?

Early in my career I sought to determine the differential biologic activities of the human IgG subclasses. I had access to monoclonal human IgG proteins, isolated from the sera of patient with multiple myeloma, which is a cancer of IgG producing plasma cells; however, the antigen binding specificity was unknown. Therefore, we generated artificial “immune complexes” by heat aggregation (63°C for 10–20 min) or cross-linking (with bis diazotized benzidene!). Such preparations allowed elucidation of the differential abilities of the IgG subclasses, e.g., to activate the complement cascade, detect the presence of cellular Fc receptors, induce phagocytosis. The physicochemical properties of such aggregates/“immune complexes” were ill-defined, except for size.⁸

Further insights into the differential biologic properties of immune complexes were obtained from a series of experiments reported from the laboratory of Peter Lachmann.^9,10 Defined immune complexes (IC) were used to evaluate the ability of the human antibody classes and subclasses to trigger the neutrophil respiratory burst and degranulation. A panel of chimeric mouse-human anti-5-iodo-4-hydroxy-3-nitrophenacetyl (NIP) monoclonal antibodies were generated and IC were prepared with NIP conjugated BSA. Neutrophil activation was shown to vary depending on factors such as antibody class and subclass, epitope density and antigen:antibody ratio. An important conclusion from these studies was that different outcomes, e.g., degranulation or respiratory burst, could be elicited by immune complexes formed by the same antibody isotype at differing antigen/antibody ratios.

A sophisticated theoretical model for the potential of a divalent antibody to form immune complexes with antigens of differing valency, together with predictions of the size and consequent sedimentation velocity, was developed by Jens Steensgaard.¹¹ Subsequently, we tested the theoretical model using human IgG as antigen and a panel of mouse monoclonal anti-human IgG heavy and light chain antibodies, at varying antigen/antibody ratios.^12,13 These studies showed that the immune complexes formed differed for each anti-IgG antibody employed, i.e., the epitope specificity was an important parameter. The influence of epitope specificity is illustrated by studies demonstrating significant differences in biologic activities of Type I and Type II anti-CD20 antibodies that appear to differ only marginally in epitope specificity.^14,15

These data suggest that, in addition to administering aggregate free antibody, we need also to consider the possible nature and characteristics of immune complexes that may be formed on first and continued exposure of a patient to a therapeutic antibody. The dilemma is that immune complexes are cleared by cells that degrade and present antigen, e.g., macrophages, dendritic cells. The possible significance of such studies may be illustrated for anti-tumor necrosis factor (TNF) antibodies. TNF forms a trimer and is therefore, potentially, trivalent and can form three dimensional immune complexes with divalent antibody. A study of the size of immune complex formed between TNF and infliximab, etanercept and a third anti-TNF antibody, at differing antigen/antibody ratios, showed that each antibody generated immune complexes with a unique size profile.¹⁶ It has been suggested that a fundamental difference between IgG-antigen complexes and aggregated IgG is that the CDRs are engaged in the former but exposed in the latter; however, x-ray crystal structural analysis of Fab-antigen complexes shows that all CDRs are not engaged in binding antigen for a majority of the complexes analysed.¹⁷

Antibody therapeutics are delivered at high doses such that <1% of an unnatural or degraded component (i.e., non-self) can represent a viable immunogenic dose, e.g., amino acid residue mis-incorporation, methionine oxidation, deamidation.^18,19 However, I would offer that defining the “natural” structure of a protein/glycoprotein is not a simple exercise; it cannot be unequivocally inferred from the gene sequence since the protein product is subject to multiple intra-cellular processes that include co- and post-translational modifications (CTM; PTM). In addition the assigned structure is determined for molecules that have been resident in bodily fluid(s), prior to isolation and purification employing multiple physicochemical protocols. By contrast recombinant human therapeutics are produced in xenogeneic tissue, including Chinese hamster ovary, mouse NS0 and Sp2/0 cells, that may yield product not having the necessary human type CTM, PTM or add non-human CTM, PTM. Following secretion the product is maintained in the culture medium, for an extended period of time during which it is exposed to products of both live and dead producer cells. It is then subject to rigorous down-stream processing, formulation, storage and a defined delivery protocol.

I would further argue that antibody therapeutics are intrinsically immunogenic since even “fully human” antibodies, whether generated from phage display libraries or transgenic “humanized” mice, are formed from combinations of heavy and light chains that would be recognized as anti-self during development of the endogenous immune repertoire and be eliminated to maintain self tolerance. Antibody therapy thus represents the administration of anti-self antibody (autoantibody!) and as such violates a fundamental protective feature of the immune system, namely tolerance to self. Autoimmunity can result from, or precipitate disregulation of, multiple immune or non-immune pathways. The reported incidence of anti-drug antibody responses has increased as progressively more sensitive assays have been developed; the consequences must be carefully evaluated and, in some cases, may be modulated by the administration of mild immunosuppressive agents.^20,21