Augmented Binary Substitution: Single-pass CDR germ-lining and stabilization of therapeutic antibodies

Although humanized antibodies have been highly successful in the clinic, all current humanization techniques have potential limitations, such as: reliance on rodent hosts, immunogenicity due to high non-germ-line amino acid content, v-domain destabilization, expression and formulation issues. This study presents a technology that generates stable, soluble, ultrahumanized antibodies via single-step complementarity-determining region (CDR) germ-lining. For three antibodies from three separate key immune host species, binary substitution CDR cassettes were inserted into preferred human frameworks to form libraries in which only the parental or human germ-line destination residue was encoded at each position. The CDR-H3 in each case was also augmented with 1 ± 1 random substitution per clone. Each library was then screened for clones with restored antigen binding capacity. Lead ultrahumanized clones demonstrated high stability, with affinity and specificity equivalent to, or better than, the parental IgG. Critically, this was mainly achieved on germ-line frameworks by simultaneously subtracting up to 19 redundant non-germ-line residues in the CDRs. This process significantly lowered non-germ-line sequence content, minimized immunogenicity risk in the final molecules and provided a heat map for the essential non-germ-line CDR residue content of each antibody. The ABS technology therefore fully optimizes the clinical potential of antibodies from rodents and alternative immune hosts, rendering them indistinguishable from fully human in a simple, single-pass process.Monoclonal antibodies are a highly established technology in drug development and the majority of currently approved therapeutic antibodies are derived from immunized rodents (1). The advent of display libraries and engineered animals that can produce “fully human” antibody v-gene sequences has had a significant positive impact on antibody drug discovery success (1), but these technologies are mostly the domain of biopharmaceutical companies. Antibodies from wild-type animals that are already extant, or can be freely developed, will therefore continue to be a rich source of therapeutic candidates. In addition, phylogenetically distant hosts such as rabbits and chickens may become a valuable source of monoclonals with clinical potential against challenging targets (2, 3).Chimerization of murine antibodies can reduce anti-IgG responses in man (4), but murine v-domains may still have provocative T-cell epitope content, necessitating “humanization” of their framework regions (5, 6). Classical humanization “grafts” murine CDRs into human v-gene sequences (7), but this typically leads to significant reduction in affinity for target, so murine residues are introduced at key positions in the frameworks (a.k.a. “back-mutations”), to restore function (8). Importantly, humanized antibodies do elicit lower immunogenicity rates in patients in comparison with chimerics (9).Alternative humanization methods have also been developed based on rational design or empirical selection (10–17), but current methods still all suffer from flaws, such as: high non-germ-line amino acid content retention (5, 6); grafting into poorly understood frameworks (13); resource-intensive, iterative methods (15, 18); requirement for homology modeling of the v-domains, which is often inaccurate (19, 20), or a cocrystal structure with the target antigen (14). Methods that allow humanization into preferred frameworks can add numerous framework mutations (18, 21), which may destabilize the v-domains (22), encode new T-cell epitopes, or introduce random amino acid mutations in CDRs (12, 13) that can drive polyspecificity and/or poor PK properties (23).Critically, testing of protein therapeutics in monkeys has been shown to be nonpredictive of immune responses in man (24) and animal immunogenicity testing has been suggested to be of little value in biosimilar development (25). Current evidence suggests that the main risk factors for antibody immunogenicity in man are human T-cell epitope content and, to a lesser extent, T-cell independent B-cell responses (6). B-cell epitopes are challenging to predict and B-cell-only responses to biotherapeutics appear to be driven by protein aggregates (26). The key attributes to reduce antibody immunogenicity risk in the clinic appear to be: low T-cell epitope content, minimized non-germ-line amino acid content and low aggregation potential (27).In recent years, several reports have strongly suggested that CDRs might be malleable in ways that could not be predicted a priori. Random mutagenesis and reselection of a classically humanized rat antibody found that individual framework back mutations and CDR residues could revert to human germ-line sequence, while maintaining or even improving the function of the antibody (28). A number of humanization studies have now also shown that a small number of positions in the CDRs could be substituted for human germ-line residues, through a rational design cycle of reversion mutations (5, 29). In addition to these observations, a number of structural analyses have illustrated the common redundancy of sequence space in antibody binding interfaces. Despite typically large buried interfaces between antibodies and protein targets, only a subset of residues in the CDRs of antibodies usually makes contact with antigen (30–32). Alanine scanning of CDR loops has also shown that only a limited number of residues directly affect antigen binding affinity (33). Indeed, it has even been shown that redundant paratope space in a single antibody may be exploited to engineer binding specificity to two separate targets (34). Additionally, CDR loop structures are known to be restricted to a limited number of canonical classes, despite amino acid variation within those classes at specific positions (35–38). These observations led us to hypothesize that, under the right experimental conditions, a large proportion of residues in grafted animal CDRs could be concurrently replaced by the residues found at the corresponding positions in a given destination human germ-line v-gene.In this study, we generated combinatorial libraries on the basis of a design principle we have named “Augmented Binary Substitution” (ABS). Each library was based on a single starting antibody: rat anti-RAGE (28), rabbit anti-A33 (2), and chicken anti-pTau (3). These libraries were built into human germ-line frameworks of high predicted stability and solubility, then interrogated via phage display and screened to identify lead clones with epitope specificity and affinity equivalent to the parental clone. ABS proved to be a facile, rapid method that retains only the functionally required CDR content of the parental animal antibody, without the need for prior crystal-structure insight. Notably, this CDR germ-lining approach generated highly stable and soluble human IgGs, from multiple key antibody discovery species, that have minimized predicted human T-cell epitope content. The reproducibility of these findings across three antibodies from three disparate species demonstrates a fundamental plasticity in antibody paratopes that can be broadly exploited in therapeutic antibody optimization.