Ultra-high-diversity factorizable libraries for efficient therapeutic discovery.

The successful discovery of novel biological therapeutics by selection requires highly diverse libraries of candidate sequences that contain a high proportion of desirable candidates. Here we propose the use of computationally designed factorizable libraries made of concatenated segment libraries as a method of creating large libraries that meet an objective function at low cost. We show that factorizable libraries can be designed efficiently by representing objective functions that describe sequence optimality as an inner product of feature vectors, which we use to design an optimization method we call stochastically annealed product spaces (SAPS). We then use this approach to design diverse and efficient libraries of antibody CDR-H3 sequences with various optimized characteristics.

Antibody complementarity determining region design using high-capacity machine learning.

MOTIVATION: The precise targeting of antibodies and other protein therapeutics is required for their proper function and the elimination of deleterious off-target effects. Often the molecular structure of a therapeutic target is unknown and randomized methods are used to design antibodies without a model that relates antibody sequence to desired properties. RESULTS: Here, we present Ens-Grad, a machine learning method that can design complementarity determining regions of human Immunoglobulin G antibodies with target affinities that are superior to candidates derived from phage display panning experiments. We also demonstrate that machine learning can improve target specificity by the modular composition of models from different experimental campaigns, enabling a new integrative approach to improving target specificity. Our results suggest a new path for the discovery of therapeutic molecules by demonstrating that predictive and differentiable models of antibody binding can be learned from high-throughput experimental data without the need for target structural data. AVAILABILITY AND IMPLEMENTATION: Sequencing data of the phage panning experiment are deposited at NIH's Sequence Read Archive (SRA) under the accession number SRP158510. We make our code available at https://github.com/gifford-lab/antibody-2019. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MAA868, a novel FXI antibody with a unique binding mode, shows durable effects on markers of anticoagulation in humans.

A large unmet medical need exists for safer antithrombotic drugs because all currently approved anticoagulant agents interfere with hemostasis, leading to an increased risk of bleeding. Genetic and pharmacologic evidence in humans and animals suggests that reducing factor XI (FXI) levels has the potential to effectively prevent and treat thrombosis with a minimal risk of bleeding. We generated a fully human antibody (MAA868) that binds the catalytic domain of both FXI (zymogen) and activated FXI. Our structural studies show that MAA868 traps FXI and activated FXI in an inactive, zymogen-like conformation, explaining its equally high binding affinity for both forms of the enzyme. This binding mode allows the enzyme to be neutralized before entering the coagulation process, revealing a particularly attractive anticoagulant profile of the antibody. MAA868 exhibited favorable anticoagulant activity in mice with a dose-dependent protection from carotid occlusion in a ferric chloride-induced thrombosis model. MAA868 also caused robust and sustained anticoagulant activity in cynomolgus monkeys as assessed by activated partial thromboplastin time without any evidence of bleeding. Based on these preclinical findings, we conducted a first-in-human study in healthy subjects and showed that single subcutaneous doses of MAA868 were safe and well tolerated. MAA868 resulted in dose- and time-dependent robust and sustained prolongation of activated partial thromboplastin time and FXI suppression for up to 4 weeks or longer, supporting further clinical investigation as a potential once-monthly subcutaneous anticoagulant therapy.

Computational counterselection identifies nonspecific therapeutic biologic candidates.

Effective biologics require high specificity and limited off-target binding, but these properties are not guaranteed by current affinity-selection-based discovery methods. Molecular counterselection against off targets is a technique for identifying nonspecific sequences but is experimentally costly and can fail to eliminate a large fraction of nonspecific sequences. Here, we introduce computational counterselection, a framework for removing nonspecific sequences from pools of candidate biologics using machine learning models. We demonstrate the method using sequencing data from single-target affinity selection of antibodies, bypassing combinatorial experiments. We show that computational counterselection outperforms molecular counterselection by performing cross-target selection and individual binding assays to determine the performance of each method at retaining on-target, specific antibodies and identifying and eliminating off-target, nonspecific antibodies. Further, we show that one can identify generally polyspecific antibody sequences using a general model trained on affinity data from unrelated targets with potential affinity for a broad range of sequences.

Reconstruction of full antibody sequences in NGS datasets and accurate VL:VH coupling by cluster coordinate matching of non-overlapping reads.

Next-generation sequencing (NGS) is an indispensable tool in antibody discovery projects. However, the limits on NGS read length make it difficult to reconstruct full antibody sequences from the sequencing runs, especially if the six CDRs are randomized. To overcome that, we took advantage of Illumina's cluster mapping capabilities to pair non-overlapping reads and reconstruct full Fab sequences with accurate VL:VH pairings. The method relies on in silico cluster coordinate information, and not on extensive in vitro manipulation, making the protocol easily deployable and less prone to PCR-derived errors. This work maintains the throughput necessary for antibody discovery campaigns, and a high degree of fidelity, which potentiates not only phage-display and synthetic library-based discovery methods, but also the NGS-driven analysis of naïve and immune libraries.

Protective anti-prion antibodies in human immunoglobulin repertoires.

Prion immunotherapy may hold great potential, but antibodies against certain PrP epitopes can be neurotoxic. Here, we identified > 6,000 PrP-binding antibodies in a synthetic human Fab phage display library, 49 of which we characterized in detail. Antibodies directed against the flexible tail of PrP conferred neuroprotection against infectious prions. We then mined published repertoires of circulating B cells from healthy humans and found antibodies similar to the protective phage-derived antibodies. When expressed recombinantly, these antibodies exhibited anti-PrP reactivity. Furthermore, we surveyed 48,718 samples from 37,894 hospital patients for the presence of anti-PrP IgGs and found 21 high-titer individuals. The clinical files of these individuals did not reveal any enrichment of specific pathologies, suggesting that anti-PrP autoimmunity is innocuous. The existence of anti-prion antibodies in unbiased human immunological repertoires suggests that they might clear nascent prions early in life. Combined with the reported lack of such antibodies in carriers of disease-associated PRNP mutations, this suggests a link to the low incidence of spontaneous prion diseases in human populations.

Development of Anti-CD32b Antibodies with Enhanced Fc Function for the Treatment of B and Plasma Cell Malignancies.

The sole inhibitory Fcγ receptor CD32b (FcγRIIb) is expressed throughout B and plasma cell development and on their malignant counterparts. CD32b expression on malignant B cells is known to provide a mechanism of resistance to rituximab that can be ameliorated with a CD32b-blocking antibody. CD32b, therefore, represents an attractive tumor antigen for targeting with a monoclonal antibody (mAb). To this end, two anti-CD32b mAbs, NVS32b1 and NVS32b2, were developed. Their complementarity-determining regions (CDR) bind the CD32b Fc binding domain with high specificity and affinity while the Fc region is afucosylated to enhance activation of FcγRIIIa on immune effector cells. The NVS32b mAbs selectively target CD32b+ malignant cells and healthy B cells but not myeloid cells. They mediate potent killing of opsonized CD32b+ cells via antibody-dependent cellular cytotoxicity and phagocytosis (ADCC and ADCP) as well as complement-dependent cytotoxicity (CDC). In addition, NVS32b CDRs block the CD32b Fc-binding domain, thereby minimizing CD32b-mediated resistance to therapeutic mAbs including rituximab, obinutuzumab, and daratumumab. NVS32b mAbs demonstrate robust antitumor activity against CD32b+ xenografts in vivo and immunomodulatory activity including recruitment of macrophages to the tumor and enhancement of dendritic cell maturation in response to immune complexes. Finally, the activity of NVS32b mAbs on CD32b+ primary malignant B and plasma cells was confirmed using samples from patients with B-cell chronic lymphocytic leukemia (CLL) and multiple myeloma. The findings indicate the promising potential of NVS32b mAbs as a single agent or in combination with other mAb therapeutics for patients with CD32b+ malignant cells.

Biophysical properties of human antibody variable domains.

There are great demands on the stability, expression yield and resistance to aggregation of antibody fragments. To untangle intrinsic domain effects from domain interactions, we present first a systematic evaluation of the isolated human immunoglobulin variable heavy (V(H)) and light (V(L)) germline family consensus domains and then a systematic series of V(H)-V(L) combinations in the scFv format. The constructs were evaluated in terms of their expression behavior, oligomeric state in solution and denaturant-induced unfolding equilibria under non-reducing conditions. The seven V(H) and seven V(L) domains represent the consensus sequences of the major human germline subclasses, derived from the Human Combinatorial Antibody Library (HuCAL). The isolated V(H) and V(L) domains with the highest thermodynamic stability and yield of soluble protein were V(H)3 and V(kappa)3, respectively. Similar measurements on all domain combinations in scFv fragments allowed the scFv fragments to be classified according to thermodynamic stability and in vivo folding yield. The scFv fragments containing the variable domain combinations H3kappa3, H1bkappa3, H5kappa3 and H3kappa1 show superior properties concerning yield and stability. Domain interactions diminish the intrinsic differences of the domains. ScFv fragments containing V(lambda) domains show high levels of stability, even though V(lambda) domains are surprisingly unstable by themselves. This is due to a strong interaction with the V(H) domain and depends on the amino acid sequence of the CDR-L3. On the basis of these analyses and model structures, we suggest possibilities for further improvement of the biophysical properties of individual frameworks and give recommendations for library design.

Absolute quantification of monoclonal antibodies in biofluids by liquid chromatography-tandem mass spectrometry.

The development of a quantification method for monoclonal antibodies in serum has been accomplished by high-performance liquid chromatography multiple reactions monitoring mass spectrometry. A human monoclonal antibody (HmAb) was used as the model protein for method development and validation. A peptide from the CDR3-region of its heavy chain was selected and used for quantifying the entire mAb. This signature peptide served as a template for the internal standard. Prior to mass spectrometric analysis approximately 50% of the total serum protein content was removed by albumin depletion. The accuracy of the method ranged between 99 and 112% in cynomolgus monkey serum. The intra-assay coefficient of variation (CV) was lower than 4% at 4 microg/mL and 200 microg/mL HmAb (n = 3). The CV at 400 microg/mL corresponded to 9% (n = 3). In addition, the interassay variation was investigated in a male cynomolgus serum pool and in a female cynomolgus serum pool. The CV for the male cynomolgus pool at 4 microg/mL HmAb was 7% (n = 3). The CV obtained from the female pool was 8% (n = 3), at 4 microg/mL. The dynamic range of the method was 3 orders of magnitude. After albumin depletion of 25 microL of serum, a lowest limit of quantification of 2 microg/mL HmAb was reached in both human and cynomolgus monkey samples.

Structure-based improvement of the biophysical properties of immunoglobulin VH domains with a generalizable approach.

In a systematic study of V gene families carried out with consensus V(H) and V(L) domains alone and in combinations in the scFv format, we found comparatively low expression yields and lower cooperativity in equilibrium unfolding in antibody fragments containing V(H) domains of human germline families 2, 4, and 6. From an analysis of the packing of the hydrophobic core, the completeness of charge clusters, the occurrence of unsatisfied hydrogen bonds, and residues with low beta-sheet propensities, positive Phi angles, and exposed hydrophobic side chains, we pinpointed residues potentially responsible for the unsatisfactory properties of these germline-encoded sequences. Several of those are in common between the domains of the even-numbered subgroups, but do not occur in the odd-numbered ones. In this study, we have systematically exchanged those residues alone and in combination in two different scFvs using the V(H)6 framework, and we describe their effect on equilibrium stability and folding yield. We improved the stability by 20.9 kJ/mol and the expression yield by a factor of 4 and can now use these data to rationally engineer antibodies derived from this and similar germline families for better biophysical properties. Furthermore, we provide an improved design for libraries exploiting the significant additional diversity provided by these frameworks. Both antibodies studied here completely retain their binding affinity, demonstrating that the CDR conformations were not affected.

Stability improvement of antibodies for extracellular and intracellular applications: CDR grafting to stable frameworks and structure-based framework engineering.

By combining the knowledge gained from an analysis of the biophysical properties of natural antibody variable domains, the effects of mutations obtained in directed evolution experiments, and the detailed structural comparison of antibodies, it has now become possible to engineer antibodies for higher thermodynamic stability and more efficient folding. This is particularly important when antibodies are to be used under conditions where the disulfide bonds cannot form, i.e., in intracellular applications (as "intrabodies"). We describe in detail two methods for the knowledge-based improvement of antibody stability and folding efficiency. While CDR grafting from a non-human to the most closely related human antibody framework is an established technique to reduce the immunogenicity of a therapeutic antibody, CDR grafting for stabilization implies the use of a more distantly related acceptor framework with superior biophysical characteristics. The use of such dissimilar frameworks requires particular attention to antigen contact residues outside the classical CDR definition and to residues capable of indirectly affecting the conformation of the antigen binding site. As a second alternative, the stability of a suboptimal framework can be improved by the introduction of point mutations designed to optimize key residue interactions. We describe the analysis methods used to identify such point mutations, which can be introduced all at once, while maintaining the framework features necessary for antigen binding. These rational approaches render the continued "rediscovery" of certain mutations by directed evolution unnecessary, but they can also be used in conjunction with such methods to discover even better molecules.

Biophysical properties of camelid V(HH) domains compared to those of human V(H)3 domains.

Camelidae possess an unusual form of antibodies lacking the light chains. The variable domain of these heavy chain antibodies (V(HH)) is not paired, while the V(H) domain of all other antibodies forms a heterodimer with the variable domain of the light chain (V(L)), held together by a hydrophobic interface. Here, we analyzed the biophysical properties of four camelid V(HH) fragments (H14, AMD9, RN05, and CA05) and two human consensus V(H)3 domains with different CDR3 loops to gain insight into factors determining stability and aggregation of immunoglobulin domains. We show by denaturant-induced unfolding equilibria that the free energies of unfolding of V(HH) fragments are characterized by Delta G(N-U) values between 21.1 and 35.0 kJ/mol and thus lie in the upper range of values for V(H) fragments from murine and human antibodies. Nevertheless, the V(HH) fragments studied here did not reach the high values between 39.7 and 52.7 kJ/mol of the human consensus V(H)3 domains with which they share the highest degree of sequence similarity. Temperature-induced unfolding of the V(HH) fragments that were studied proved to be reversible, and the binding affinity after cooling was fully retained. The melting temperatures were determined to be between 60.1 and 66.7 degrees C. In contrast, the studied V(H)3 domains aggregated during temperature-induced denaturation at 63-65 degrees C. In summary, the camelid V(HH) fragments are characterized by a favorable but not unusually high stability. Their hallmark is the ability to reversibly melt without aggregation, probably mediated by the surface mutations characterizing the V(HH) domains, which allow them to regain binding activity after heat renaturation.