Computationally Designed Bispecific Antibodies using Negative State Repertoires

Combinatorially restricted computational design of protein-protein interfaces to produce IgG heterodimers

Redesigning protein-protein interfaces is an important tool for developing therapeutic strategies. Interfaces can be redesigned by in silico screening, which allows for efficient sampling of a large protein space before experimental validation. However, computational costs limit the number of combinations that can be reasonably sampled. Here, we present combinatorial tyrosine (Y)/serine (S) selection (combYSelect), a computational approach combining in silico determination of the change in binding free energy (ΔΔG) of an interface with a highly restricted library composed of just two amino acids, tyrosine and serine. We used combYSelect to design two immunoglobulin G (IgG) heterodimers-combYSelect1 (L368S/D399Y-K409S/T411Y) and combYSelect2 (D399Y/K447S-K409S/T411Y)-that exhibit near-optimal heterodimerization, without affecting IgG stability or function. We solved the crystal structures of these heterodimers and found that dynamic π-stacking interactions and polar contacts drive preferential heterodimeric interactions. Finally, we demonstrated the utility of our combYSelect heterodimers by engineering both a bispecific antibody and a cytokine trap for two unique therapeutic applications.

Leveraging Artificial Intelligence to Expedite Antibody Design and Enhance Antibody-Antigen Interactions

This perspective sheds light on the transformative impact of recent computational advancements in the field of protein therapeutics, with a particular focus on the design and development of antibodies. Cutting-edge computational methods have revolutionized our understanding of protein-protein interactions (PPIs), enhancing the efficacy of protein therapeutics in preclinical and clinical settings. Central to these advancements is the application of machine learning and deep learning, which offers unprecedented insights into the intricate mechanisms of PPIs and facilitates precise control over protein functions. Despite these advancements, the complex structural nuances of antibodies pose ongoing challenges in their design and optimization. Our review provides a comprehensive exploration of the latest deep learning approaches, including language models and diffusion techniques, and their role in surmounting these challenges. We also present a critical analysis of these methods, offering insights to drive further progress in this rapidly evolving field. The paper includes practical recommendations for the application of these computational techniques, supplemented with independent benchmark studies. These studies focus on key performance metrics such as accuracy and the ease of program execution, providing a valuable resource for researchers engaged in antibody design and development. Through this detailed perspective, we aim to contribute to the advancement of antibody design, equipping researchers with the tools and knowledge to navigate the complexities of this field.

Design and engineering of bispecific antibodies: insights and practical considerations

Bispecific antibodies (bsAbs) have attracted significant attention due to their dual binding activity, which permits simultaneous targeting of antigens and synergistic binding effects beyond what can be obtained even with combinations of conventional monospecific antibodies. Despite the tremendous therapeutic potential, the design and construction of bsAbs are often hampered by practical issues arising from the increased structural complexity as compared to conventional monospecific antibodies. The issues are diverse in nature, spanning from decreased biophysical stability from fusion of exogenous antigen-binding domains to antibody chain mispairing leading to formation of antibody-related impurities that are very difficult to remove. The added complexity requires judicious design considerations as well as extensive molecular engineering to ensure formation of high quality bsAbs with the intended mode of action and favorable drug-like qualities. In this review, we highlight and summarize some of the key considerations in design of bsAbs as well as state-of-the-art engineering principles that can be applied in efficient construction of bsAbs with diverse molecular formats.

Engineering a tumor-selective prodrug T-cell engager bispecific antibody for safer immunotherapy

T-cell engaging (TCE) bispecific antibodies are potent drugs that trigger the immune system to eliminate cancer cells, but administration can be accompanied by toxic side effects that limit dosing. TCEs function by binding to cell surface receptors on T cells, frequently CD3, with one arm of the bispecific antibody while the other arm binds to cell surface antigens on cancer cells. On-target, off-tumor toxicity can arise when the target antigen is also present on healthy cells. The toxicity of TCEs may be ameliorated through the use of pro-drug forms of the TCE, which are not fully functional until recruited to the tumor microenvironment. This can be accomplished by masking the anti-CD3 arm of the TCE with an autoinhibitory motif that is released by tumor-enriched proteases. Here, we solve the crystal structure of the antigen-binding fragment of a novel anti-CD3 antibody, E10, in complex with its epitope from CD3 and use this information to engineer a masked form of the antibody that can activate by the tumor-enriched protease matrix metalloproteinase 2 (MMP-2). We demonstrate with binding experiments and in vitro T-cell activation and killing assays that our designed prodrug TCE is capable of tumor-selective T-cell activity that is dependent upon MMP-2. Furthermore, we demonstrate that a similar masking strategy can be used to create a pro-drug form of the frequently used anti-CD3 antibody SP34. This study showcases an approach to developing immune-modulating therapeutics that prioritizes safety and has the potential to advance cancer immunotherapy treatment strategies.

Noncovalent antibody catenation on a target surface greatly increases the antigen-binding avidity

Immunoglobulin G (IgG) antibodies are widely used for diagnosis and therapy. Given the unique dimeric structure of IgG, we hypothesized that, by genetically fusing a homodimeric protein (catenator) to the C-terminus of IgG, reversible catenation of antibody molecules could be induced on a surface where target antigen molecules are abundant, and that it could be an effective way to greatly enhance the antigen-binding avidity. A thermodynamic simulation showed that quite low homodimerization affinity of a catenator, e.g. dissociation constant of 100 μM, can enhance nanomolar antigen-binding avidity to a picomolar level, and that the fold enhancement sharply depends on the density of the antigen. In a proof-of-concept experiment where antigen molecules are immobilized on a biosensor tip, the C-terminal fusion of a pair of weakly homodimerizing proteins to three different antibodies enhanced the antigen-binding avidity by at least 110 or 304 folds from the intrinsic binding avidity. Compared with the mother antibody, Obinutuzumab(Y101L) which targets CD20, the same antibody with fused catenators exhibited significantly enhanced binding to SU-DHL5 cells. Together, the homodimerization-induced antibody catenation would be a new powerful approach to improve antibody applications, including the detection of scarce biomarkers and targeted anticancer therapies.

Structural mechanism of Fab domain dissociation as a measure of interface stability

Therapeutic antibodies should not only recognize antigens specifically, but also need to be free from developability issues, such as poor stability. Thus, the mechanistic understanding and characterization of stability are critical determinants for rational antibody design. In this study, we use molecular dynamics simulations to investigate the melting process of 16 antigen binding fragments (Fabs). We describe the Fab dissociation mechanisms, showing a separation in the V_H-V_L and in the C_H1-C_L domains. We found that the depths of the minima in the free energy curve, corresponding to the bound states, correlate with the experimentally determined melting temperatures. Additionally, we provide a detailed structural description of the dissociation mechanism and identify key interactions in the CDR loops and in the C_H1-C_L interface that contribute to stabilization. The dissociation of the V_H-V_L or C_H1-C_L domains can be represented by conformational changes in the bend angles between the domains. Our findings elucidate the melting process of antigen binding fragments and highlight critical residues in both the variable and constant domains, which are also strongly germline dependent. Thus, our proposed mechanisms have broad implications in the development and design of new and more stable antigen binding fragments.

Generation of bispecific antibodies by structure-guided redesign of IgG constant regions

Bispecific antibodies (BsAbs) form an exciting class of bio-therapeutics owing to their multispecificity. Although numerous formats have been developed, generation of hetero-tetrameric IgG1-like BsAbs having acceptable safety and pharmacokinetics profiles from a single cell culture system remains challenging due to the heterogeneous pairing between the four chains. Herein, we employed a structure-guided approach to engineer mutations in the constant domain interfaces (C_H1-C_L and C_H3-C_H3) of heavy and κ light chains to prevent heavy-light mispairing in the antigen binding fragment (Fab) region and heavy-heavy homodimerization in the Fc region. Transient co-transfection of mammalian cells with heavy and light chains of pre-existing antibodies carrying the engineered constant domains generates BsAbs with percentage purity ranging from 78% to 85%. The engineered BsAbs demonstrate simultaneous binding of both antigens, while retaining the thermal stability, Fc-mediated effector properties and FcRn binding properties of the parental antibodies. Importantly, since the variable domains were not modified, the mutations may enable BsAb formation from antibodies belonging to different germline origins and isotypes. The rationally designed mutations reported in this work could serve as a starting point for generating optimized solutions required for large scale production.

Extensive substrate recognition by the streptococcal antibody-degrading enzymes IdeS and EndoS

Enzymatic cleavage of IgG antibodies is a common strategy used by pathogenic bacteria to ablate immune effector function. The Streptococcus pyogenes bacterium secretes the protease IdeS and the glycosidase EndoS, which specifically catalyse cleavage and deglycosylation of human IgG, respectively. IdeS has received clinical approval for kidney transplantation in hypersensitised individuals, while EndoS has found application in engineering antibody glycosylation. We present crystal structures of both enzymes in complex with their IgG1 Fc substrate, which was achieved using Fc engineering to disfavour preferential Fc crystallisation. The IdeS protease displays extensive Fc recognition and encases the antibody hinge. Conversely, the glycan hydrolase domain in EndoS traps the Fc glycan in a "flipped-out" conformation, while additional recognition of the Fc peptide is driven by the so-called carbohydrate binding module. In this work, we reveal the molecular basis of antibody recognition by bacterial enzymes, providing a template for the development of next-generation enzymes.

Bispecific antibodies-effects of point mutations on CH3-CH3 interface stability

A new format of therapeutic proteins is bispecific antibodies, in which two different heavy chains heterodimerize to obtain two different binding sites. Therefore, it is crucial to understand and optimize the third constant domain (CH3-CH3) interface to favor heterodimerization over homodimerization, and to preserve the physicochemical properties, as thermal stability. Here, we use molecular dynamics simulations to investigate the dissociation process of 19 CH3-CH3 crystal structures that differ from each other in few point mutations. We describe the dissociation of the dimeric interface as a two-steps mechanism. As confirmed by a Markov state model, apart from the bound and the dissociated state, we observe an additional intermediate state, which corresponds to an encounter complex. The analysis of the interdomain contacts reveals key residues that stabilize the interface. We expect that our results will improve the understanding of the CH3-CH3 interface interactions and thus advance the developability and design of new antibodies formats.

Protein-protein interaction prediction with deep learning: A comprehensive review

Most proteins perform their biological function by interacting with themselves or other molecules. Thus, one may obtain biological insights into protein functions, disease prevalence, and therapy development by identifying protein-protein interactions (PPI). However, finding the interacting and non-interacting protein pairs through experimental approaches is labour-intensive and time-consuming, owing to the variety of proteins. Hence, protein-protein interaction and protein-ligand binding problems have drawn attention in the fields of bioinformatics and computer-aided drug discovery. Deep learning methods paved the way for scientists to predict the 3-D structure of proteins from genomes, predict the functions and attributes of a protein, and modify and design new proteins to provide desired functions. This review focuses on recent deep learning methods applied to problems including predicting protein functions, protein-protein interaction and their sites, protein-ligand binding, and protein design.

CDR loop interactions can determine heavy and light chain pairing preferences in bispecific antibodies

As the current biotherapeutic market is dominated by antibodies, the design of different antibody formats, like bispecific antibodies, is critical to the advancement of the field. In contrast to monovalent antibodies, which consist of two identical antigen-binding sites, bispecific antibodies can target two different epitopes by containing two different antigen-binding sites. Thus, the rise of new formats as successful therapeutics has reignited the interest in advancing and facilitating the efficient production of bispecific antibodies. Here, we investigate the influence of point mutations in the antigen-binding site, the paratope, on heavy and light chain pairing preferences by using molecular dynamics simulations. In agreement with experiments, we find that specific residues in the antibody variable domain (Fv), i.e., the complementarity-determining region (CDR) L3 and H3 loops, determine heavy and light chain pairing preferences. Excitingly, we observe substantial population shifts in CDR-H3 and CDR-L3 loop conformations in solution accompanied by a decrease in bispecific IgG yield. These conformational changes in the CDR3 loops induced by point mutations also influence all other CDR loop conformations and consequentially result in different CDR loop states in solution. However, besides their effect on the obtained CDR loop ensembles, point mutations also lead to distinct interaction patterns in the V_H-V_L interface. By comparing the interaction patterns among all investigated variants, we observe specific contacts in the interface that drive heavy and light chain pairing. Thus, these findings have broad implications in the field of antibody engineering and design because they provide a mechanistic understanding of antibody interfaces, by identifying critical factors driving the pairing preferences, and thus can help to advance the design of bispecific antibodies.

Reconfigurable asymmetric protein assemblies through implicit negative design

Asymmetric multiprotein complexes that undergo subunit exchange play central roles in biology but present a challenge for design because the components must not only contain interfaces that enable reversible association but also be stable and well behaved in isolation. We use implicit negative design to generate β sheet-mediated heterodimers that can be assembled into a wide variety of complexes. The designs are stable, folded, and soluble in isolation and rapidly assemble upon mixing, and crystal structures are close to the computational models. We construct linearly arranged hetero-oligomers with up to six different components, branched hetero-oligomers, closed C4-symmetric two-component rings, and hetero-oligomers assembled on a cyclic homo-oligomeric central hub and demonstrate that such complexes can readily reconfigure through subunit exchange. Our approach provides a general route to designing asymmetric reconfigurable protein systems.

Engineering a pure and stable heterodimeric IgA for the development of multispecific therapeutics

ABBREVIATIONS

CE-SDS: capillary electrophoresis sodium dodecyl sulfate; DSC: differential scanning calorimetry; FACS: fluorescence-activated cell sorting; FSA: full-sized antibody; Her2: human epidermal growth factor receptor 2; MFI: mean fluorescent intensity; OAA: one-armed antibody; PBS: phosphate-buffered saline; PDB: Protein Data Bank; SEC: size-exclusion chromatography; prepSEC (preparative SEC); RMSD: root-mean-square deviation; RU: resonance units; SPR: surface plasmon resonance; TAA: tumor-associated antigen; WT: wild-type.

Next generation Fc scaffold for multispecific antibodies

Bispecific antibodies (Bispecifics) demonstrate exceptional clinical potential to address some of the most complex diseases. However, Bispecific production in a single cell often requires the correct pairing of multiple polypeptide chains for desired assembly. This is a considerable hurdle that hinders the development of many immunoglobulin G (IgG)-like bispecific formats. Our approach focuses on the rational engineering of charged residues to facilitate the chain pairing of distinct heavy chains (HC). Here, we deploy structure-guided protein design to engineer charge pair mutations (CPMs) placed in the CH3-CH3' interface of the fragment crystallizable (Fc) region of an antibody (Ab) to correctly steer heavy chain pairing. When used in combination with our stable effector functionless 2 (SEFL2.2) technology, we observed high pairing efficiency without significant losses in expression yields. Furthermore, we investigate the relationship between CPMs and the sequence diversity in the parental antibodies, proposing a rational strategy to deploy these engineering technologies.

Dichotomous impact of affinity on the function of T cell engaging bispecific antibodies

BACKGROUND

Bispecific T cell engagers represent the majority of bispecific antibodies (BsAbs) entering the clinic to treat metastatic cancer. The ability to apply these agents safely and efficaciously in the clinic, particularly for solid tumors, has been challenging. Many preclinical studies have evaluated parameters related to the activity of T cell engaging BsAbs, but many questions remain.

MAIN BODY

This study investigates the impact of affinity of T cell engaging BsAbs with regards to potency, efficacy, and induction of immunomodulatory receptors/ligands using HER-2/CD3 BsAbs as a model system. We show that an IgG BsAb can be as efficacious as a smaller BsAb format both in vitro and in vivo. We uncover a dichotomous relationship between tumor-associated antigen (TAA) affinity and CD3 affinity requirements for cells that express high versus low levels of TAA. HER-2 affinity directly correlated with the CD3 engager lysis potency of HER-2/CD3 BsAbs when HER-2 receptor numbers are high (~200 K/cell), while the CD3 affinity did not impact potency until its binding affinity was extremely low (<600 nM). When HER-2 receptor numbers were lower (~20 K/cell), both HER-2 and CD3 affinity impacted potency. The high affinity anti-HER-2/low CD3 affinity BsAb also demonstrated lower cytokine induction levels in vivo and a dosing paradigm atypical of extremely high potency T cell engaging BsAbs reaching peak efficacy at doses >3 mg/kg. This data confirms that low CD3 affinity provides an opportunity for improved safety and dosing for T cell engaging BsAbs. T cell redirection also led to upregulation of Programmed cell death 1 (PD-1) and 4-1BB, but not CTLA-4 on T cells, and to Programmed death-ligand 1 (PD-L1) upregulation on HER-2HI SKOV3 tumor cells, but not on HER-2LO OVCAR3 tumor cells. Using this information, we combined anti-PD-1 or anti-4-1BB monoclonal antibodies with the HER-2/CD3 BsAb in vivo and demonstrated significantly increased efficacy against HER-2HI SKOV3 tumors via both combinations.

CONCLUSIONS

Overall, these studies provide an informational dive into the optimization process of CD3 engaging BsAbs for solid tumors indicating that a reduced affinity for CD3 may enable a better therapeutic index with a greater selectivity for the target tumor and a reduced cytokine release syndrome. These studies also provide an additional argument for combining T cell checkpoint inhibition and co-stimulation to achieve optimal efficacy.

BACKGROUND

Ensembles in solution as a new paradigm for antibody structure prediction and design

The rise of antibodies as a promising and rapidly growing class of biotherapeutic proteins has motivated numerous studies to characterize and understand antibody structures. In the past decades, the number of antibody crystal structures increased substantially, which revolutionized the atomistic understanding of antibody functions. Even though numerous static structures are known, various biophysical properties of antibodies (i.e., specificity, hydrophobicity and stability) are governed by their dynamic character. Additionally, the importance of high-quality structures in structure–function relationship studies has substantially increased. These structure–function relationship studies have also created a demand for precise homology models of antibody structures, which allow rational antibody design and engineering when no crystal structure is available. Here, we discuss various aspects and challenges in antibody design and extend the paradigm of describing antibodies with only a single static structure to characterizing them as dynamic ensembles in solution.

Recent Advances in the Molecular Design and Applications of Multispecific Biotherapeutics

Recombinant protein-based biotherapeutics drugs have transformed clinical pipelines of the biopharmaceutical industry since the launch of recombinant insulin nearly four decades ago. These biologic drugs are structurally more complex than small molecules, and yet share a similar principle for rational drug discovery and development: That is to start with a pre-defined target and follow with the functional modulation with a therapeutic agent. Despite these tremendous successes, this "one target one drug" paradigm has been challenged by complex disease mechanisms that involve multiple pathways and demand new therapeutic routes. A rapidly evolving wave of multispecific biotherapeutics is coming into focus. These new therapeutic drugs are able to engage two or more protein targets via distinct binding interfaces with or without the chemical conjugation to large or small molecules. They possess the potential to not only address disease intricacy but also exploit new therapeutic mechanisms and assess undruggable targets for conventional monospecific biologics. This review focuses on the recent advances in molecular design and applications of major classes of multispecific biotherapeutics drugs, which include immune cells engagers, antibody-drug conjugates, multispecific tetherbodies, biologic matchmakers, and small-scaffold multispecific modalities. Challenges posed by the multispecific biotherapeutics drugs and their future outlooks are also discussed.

Modeling Immunity with Rosetta: Methods for Antibody and Antigen Design

Structure-based antibody and antigen design has advanced greatly in recent years, due not only to the increasing availability of experimentally determined structures but also to improved computational methods for both prediction and design. Constant improvements in performance within the Rosetta software suite for biomolecular modeling have given rise to a greater breadth of structure prediction, including docking and design application cases for antibody and antigen modeling. Here, we present an overview of current protocols for antibody and antigen modeling using Rosetta and exemplify those by detailed tutorials originally developed for a Rosetta workshop at Vanderbilt University. These tutorials cover antibody structure prediction, docking, and design and antigen design strategies, including the addition of glycans in Rosetta. We expect that these materials will allow novice users to apply Rosetta in their own projects for modeling antibodies and antigens.

Protein sequence design by conformational landscape optimization

The protein design problem is to identify an amino acid sequence that folds to a desired structure. Given Anfinsen's thermodynamic hypothesis of folding, this can be recast as finding an amino acid sequence for which the desired structure is the lowest energy state. As this calculation involves not only all possible amino acid sequences but also, all possible structures, most current approaches focus instead on the more tractable problem of finding the lowest-energy amino acid sequence for the desired structure, often checking by protein structure prediction in a second step that the desired structure is indeed the lowest-energy conformation for the designed sequence, and typically discarding a large fraction of designed sequences for which this is not the case. Here, we show that by backpropagating gradients through the transform-restrained Rosetta (trRosetta) structure prediction network from the desired structure to the input amino acid sequence, we can directly optimize over all possible amino acid sequences and all possible structures in a single calculation. We find that trRosetta calculations, which consider the full conformational landscape, can be more effective than Rosetta single-point energy estimations in predicting folding and stability of de novo designed proteins. We compare sequence design by conformational landscape optimization with the standard energy-based sequence design methodology in Rosetta and show that the former can result in energy landscapes with fewer alternative energy minima. We show further that more funneled energy landscapes can be designed by combining the strengths of the two approaches: the low-resolution trRosetta model serves to disfavor alternative states, and the high-resolution Rosetta model serves to create a deep energy minimum at the design target structure.

Computational design and experimental optimization of protein binders with prospects for biomedical applications

Protein-based binders have become increasingly more attractive candidates for drug and imaging agent development. Such binders could be evolved from a number of different scaffolds, including antibodies, natural protein effectors and unrelated small protein domains of different geometries. While both computational and experimental approaches could be utilized for protein binder engineering, in this review we focus on various computational approaches for protein binder design and demonstrate how experimental selection could be applied to subsequently optimize computationally-designed molecules. Recent studies report a number of designed protein binders with pM affinities and high specificities for their targets. These binders usually characterized with high stability, solubility, and low production cost. Such attractive molecules are bound to become more common in various biotechnological and biomedical applications in the near future.

Bringing the Heavy Chain to Light: Creating a Symmetric, Bivalent IgG-Like Bispecific

Bispecific molecules are biologically significant, yet their complex structures pose important manufacturing and pharmacokinetic challenges. Nevertheless, owing to similarities with monoclonal antibodies (mAbs), IgG-like bispecifics conceptually align well with conventional expression and manufacturing platforms and often exhibit potentially favorable drug metabolism and pharmacokinetic (DMPK) properties. However, IgG-like bispecifics do not possess target bivalency and current designs often require tedious engineering and purification to ensure appropriate chain pairing. Here, we present a near-native IgG antibody format, the 2xVH, which can create bivalency for each target or epitope and requires no engineering for cognate chain pairing. In this modality, two different variable heavy (VH) domains with distinct binding specificities are grafted onto the first constant heavy (CH1) and constant light (CL) domains, conferring the molecule with dual specificity. To determine the versatility of this format, we characterized the expression, binding, and stability of several previously identified soluble human VH domains. By grafting these domains onto an IgG scaffold, we generated several prototype 2xVH IgG and Fab molecules that display similar properties to mAbs. These molecules avoided the post-expression purification necessary for engineered bispecifics while maintaining a capacity for simultaneous dual binding. Hence, the 2xVH format represents a bivalent, bispecific design that addresses limitations of manufacturing IgG-like bispecifics while promoting biologically-relevant dual target engagement.

Computational stabilization of T cell receptors allows pairing with antibodies to form bispecifics

Recombinant T cell receptors (TCRs) can be used to redirect naïve T cells to eliminate virally infected or cancerous cells; however, they are plagued by low stability and uneven expression. Here, we use molecular modeling to identify mutations in the TCR constant domains (Cα/Cβ) that increase the unfolding temperature of Cα/Cβ by 20 °C, improve the expression of four separate α/β TCRs by 3- to 10-fold, and improve the assembly and stability of TCRs with poor intrinsic stability. The stabilizing mutations rescue the expression of TCRs destabilized through variable domain mutation. The improved stability and folding of the TCRs reduces glycosylation, perhaps through conformational stabilization that restricts access to N-linked glycosylation enzymes. The Cα/Cβ mutations enables antibody-like expression and assembly of well-behaved bispecific molecules that combine an anti-CD3 antibody with the stabilized TCR. These TCR/CD3 bispecifics can redirect T cells to kill tumor cells with target HLA/peptide on their surfaces in vitro.

The advent of de novo proteins for cancer immunotherapy

Engineered proteins are revolutionizing immunotherapy, but advances are still needed to harness their full potential. Traditional protein engineering methods use naturally existing proteins as a starting point, and therefore, are intrinsically limited to small alterations of a protein's natural structure and function. Conversely, computational de novo protein design is free of such limitation, and can produce a virtually infinite number of novel protein sequences, folds, and functions. Recently, we used de novo protein engineering to create Neoleukin-2/15 (Neo-2/15), a protein mimetic of the function of both interleukin-2 (IL-2) and interleukin-15 (IL-15). To our knowledge, Neo-2/15 is the first de novo protein with immunotherapeutic activity, and in murine cancer models, it has demonstrated enhanced therapeutic potency and reduced toxicity compared to IL-2. De novo protein design is already showcasing its tremendous potential for driving the next wave of protein-based therapeutics that are explicitly engineered to treat disease.

DIP/Dpr interactions and the evolutionary design of specificity in protein families

Differential binding affinities among closely related protein family members underlie many biological phenomena, including cell-cell recognition. Drosophila DIP and Dpr proteins mediate neuronal targeting in the fly through highly specific protein-protein interactions. We show here that DIPs/Dprs segregate into seven specificity subgroups defined by binding preferences between their DIP and Dpr members. We then describe a sequence-, structure- and energy-based computational approach, combined with experimental binding affinity measurements, to reveal how specificity is coded on the canonical DIP/Dpr interface. We show that binding specificity of DIP/Dpr subgroups is controlled by "negative constraints", which interfere with binding. To achieve specificity, each subgroup utilizes a different combination of negative constraints, which are broadly distributed and cover the majority of the protein-protein interface. We discuss the structural origins of negative constraints, and potential general implications for the evolutionary origins of binding specificity in multi-protein families.

Designing protein structures and complexes with the molecular modeling program Rosetta

Proteins perform an amazingly diverse set of functions in all aspects of life. Critical to the function of many proteins are the highly specific three-dimensional structures they adopt. For this reason, there is strong interest in learning how to rationally design proteins that adopt user-defined structures. Over the last 25 years, there has been significant progress in the field of computational protein design as rotamer-based sequence optimization protocols have enabled accurate design of protein tertiary and quaternary structure. In this award article, I will summarize how the molecular modeling program Rosetta is used to design new protein structures and describe how we have taken advantage of this capability to create proteins that have important applications in research and medicine. I will highlight three protein design stories: the use of protein interface design to create therapeutic bispecific antibodies, the engineering of light-inducible proteins that can be used to recruit proteins to specific locations in the cell, and the de novo design of new protein structures from pieces of naturally occurring proteins.

Advances in protein structure prediction and design

The prediction of protein three-dimensional structure from amino acid sequence has been a grand challenge problem in computational biophysics for decades, owing to its intrinsic scientific interest and also to the many potential applications for robust protein structure prediction algorithms, from genome interpretation to protein function prediction. More recently, the inverse problem - designing an amino acid sequence that will fold into a specified three-dimensional structure - has attracted growing attention as a potential route to the rational engineering of proteins with functions useful in biotechnology and medicine. Methods for the prediction and design of protein structures have advanced dramatically in the past decade. Increases in computing power and the rapid growth in protein sequence and structure databases have fuelled the development of new data-intensive and computationally demanding approaches for structure prediction. New algorithms for designing protein folds and protein-protein interfaces have been used to engineer novel high-order assemblies and to design from scratch fluorescent proteins with novel or enhanced properties, as well as signalling proteins with therapeutic potential. In this Review, we describe current approaches for protein structure prediction and design and highlight a selection of the successful applications they have enabled.

Bispecific T-Cell Redirection versus Chimeric Antigen Receptor (CAR)-T Cells as Approaches to Kill Cancer Cells

The concepts for T-cell redirecting bispecific antibodies (TRBAs) and chimeric antigen receptor (CAR)-T cells are both at least 30 years old but both platforms are just now coming into age. Two TRBAs and two CAR-T cell products have been approved by major regulatory agencies within the last ten years for the treatment of hematological cancers and an additional 53 TRBAs and 246 CAR cell constructs are in clinical trials today. Two major groups of TRBAs include small, short-half-life bispecific antibodies that include bispecific T-cell engagers (BiTE®s) which require continuous dosing and larger, mostly IgG-like bispecific antibodies with extended pharmacokinetics that can be dosed infrequently. Most CAR-T cells today are autologous, although significant strides are being made to develop off-the-shelf, allogeneic CAR-based products. CAR-Ts form a cytolytic synapse with target cells that is very different from the classical immune synapse both physically and mechanistically, whereas the TRBA-induced synapse is similar to the classic immune synapse. Both TRBAs and CAR-T cells are highly efficacious in clinical trials but both also present safety concerns, particularly with cytokine release syndrome and neurotoxicity. New formats and dosing paradigms for TRBAs and CAR-T cells are being developed in efforts to maximize efficacy and minimize toxicity, as well as to optimize use with both solid and hematologic tumors, both of which present significant challenges such as target heterogeneity and the immunosuppressive tumor microenvironment.

Protein engineers turned evolutionists-the quest for the optimal starting point

The advent of laboratory directed evolution yielded a fruitful crosstalk between the disciplines of molecular evolution and bio-engineering. Here, we outline recent developments in both disciplines with respect to how one can identify the best starting points for directed evolution, such that highly efficient and robust tailor-made enzymes can be obtained with minimal optimization. Directed evolution studies have highlighted essential features of engineer-able enzymes: highly stable, mutationally robust enzymes with the capacity to accept a broad range of substrates. Robust, evolvable enzymes can be inferred from the natural sequence record. Broad substrate spectrum relates to conformational plasticity and can also be predicted by phylogenetic analyses and/or by computational design. Overall, an increasingly powerful toolkit is becoming available for identifying optimal starting points including network analyses of enzyme superfamilies and other bioinformatics methods.

A robust heterodimeric Fc platform engineered for efficient development of bispecific antibodies of multiple formats

Bispecific monoclonal antibodies can bind two protein targets simultaneously and enable therapeutic modalities inaccessible by traditional mAbs. Bispecific formats containing a heterodimeric Fc region are of particular interest, as a heterodimeric Fc empowers both bispecificity and altered valencies while retaining the developability and druggability of a monoclonal antibody. We present a robust heterodimeric Fc platform, called the XmAb® bispecific platform, engineered for efficient development of bispecific antibodies and Fc fusions of multiple formats. First, we engineer a purification solution for proteins containing a heterodimeric Fc using engineered isoelectric point differences in the Fc region that enable straightforward purification of the heterodimeric species. Then, we combine this purification solution with a novel set of Fc substitutions capable of achieving heterodimer yields over 95% with little change in thermostability. Next, we illustrate the flexibility of our heterodimeric Fc with a case study in which a wide range of tumor-associated antigen × CD3 bispecifics are generated, differing in choice of tumor antigen, affinities for both tumor antigen and CD3, and tumor antigen valency. Finally, we present manufacturing data reinforcing the robustness of the heterodimeric Fc platform at scale.

Expanding the Boundaries of Biotherapeutics with Bispecific Antibodies

Bispecific antibodies have moved from being an academic curiosity with therapeutic promise to reality, with two molecules being currently commercialized (Hemlibra® and Blincyto®) and many more in clinical trials. The success of bispecific antibodies is mainly due to the continuously growing number of mechanisms of actions (MOA) they enable that are not accessible to monoclonal antibodies. One of the earliest MOA of bispecific antibodies and currently the one with the largest number of clinical trials is the redirecting of the cytotoxic activity of T-cells for oncology applications, now extending its use in infective diseases. The use of bispecific antibodies for crossing the blood-brain barrier is another important application because of its potential to advance the therapeutic options for neurological diseases. Another noteworthy application due to its growing trend is enabling a more tissue-specific delivery or activity of antibodies. The different molecular solutions to the initial hurdles that limited the development of bispecific antibodies have led to the current diverse set of bispecific or multispecific antibody formats that can be grouped into three main categories: IgG-like formats, antibody fragment-based formats, or appended IgG formats. The expanded applications of bispecific antibodies come at the price of additional challenges for clinical development. The rising complexity in their structure may increase the risk of immunogenicity and the multiple antigen specificity complicates the selection of relevant species for safety assessment.

Structural and biochemical characterization of the Cutibacterium acnes exo-β-1,4-mannosidase that targets the N-glycan core of host glycoproteins

Commensal and pathogenic bacteria have evolved efficient enzymatic pathways to feed on host carbohydrates, including protein-linked glycans. Most proteins of the human innate and adaptive immune system are glycoproteins where the glycan is critical for structural and functional integrity. Besides enabling nutrition, the degradation of host N-glycans serves as a means for bacteria to modulate the host's immune system by for instance removing N-glycans on immunoglobulin G. The commensal bacterium Cutibacterium acnes is a gram-positive natural bacterial species of the human skin microbiota. Under certain circumstances, C. acnes can cause pathogenic conditions, acne vulgaris, which typically affects 80% of adolescents, and can become critical for immunosuppressed transplant patients. Others have shown that C. acnes can degrade certain host O-glycans, however, no degradation pathway for host N-glycans has been proposed. To investigate this, we scanned the C. acnes genome and were able to identify a set of gene candidates consistent with a cytoplasmic N-glycan-degradation pathway of the canonical eukaryotic N-glycan core. We also found additional gene sequences containing secretion signals that are possible candidates for initial trimming on the extracellular side. Furthermore, one of the identified gene products of the cytoplasmic pathway, AEE72695, was produced and characterized, and found to be a functional, dimeric exo-β-1,4-mannosidase with activity on the β-1,4 glycosidic bond between the second N-acetylglucosamine and the first mannose residue in the canonical eukaryotic N-glycan core. These findings corroborate our model of the cytoplasmic part of a C. acnes N-glycan degradation pathway.

Engineering IgG-Like Bispecific Antibodies-An Overview

Monoclonal antibody therapeutics have proven to be successful treatment options for patients in various indications. Particularly in oncology, therapeutic concepts involving antibodies often rely on the so-called effector functions, such as antibody dependent cellular cytotoxicity (ADCC) and complement dependent cytotoxicity (CDC), which are programed in the antibody Fc region. However, Fc-mediated effector mechanisms often seem to be insufficient in properly activating the immune system to act against tumor cells. Furthermore, long term treatments can lead to resistance against the applied drug, which is monospecific by nature. There is promise in using specific antibodies to overcome such issues due to their capability of recruiting and activating T-cells directly at the tumor site, for instance. During the last decade, two of these entities, which are referred to as Blinatumomab and Catumaxomab, have been approved to treat patients with acute lymphoblastic leukemia and malignant ascites. In addition, Emicizumab, which is a bispecific antibody targeting clotting factors IXa and X, was recently granted market approval by the FDA in 2017 for the treatment of hemophilia A. However, the generation of these next generation therapeutics is challenging and requires tremendous engineering efforts as two distinct paratopes need to be combined from two different heavy and light chains. This mini review summarizes technologies, which enable the generation of antibodies with dual specificities.

Computational Protein Design with Deep Learning Neural Networks

Computational protein design has a wide variety of applications. Despite its remarkable success, designing a protein for a given structure and function is still a challenging task. On the other hand, the number of solved protein structures is rapidly increasing while the number of unique protein folds has reached a steady number, suggesting more structural information is being accumulated on each fold. Deep learning neural network is a powerful method to learn such big data set and has shown superior performance in many machine learning fields. In this study, we applied the deep learning neural network approach to computational protein design for predicting the probability of 20 natural amino acids on each residue in a protein. A large set of protein structures was collected and a multi-layer neural network was constructed. A number of structural properties were extracted as input features and the best network achieved an accuracy of 38.3%. Using the network output as residue type restraints improves the average sequence identity in designing three natural proteins using Rosetta. Moreover, the predictions from our network show ~3% higher sequence identity than a previous method. Results from this study may benefit further development of computational protein design methods.

Integrating linear optimization with structural modeling to increase HIV neutralization breadth

Computational protein design has been successful in modeling fixed backbone proteins in a single conformation. However, when modeling large ensembles of flexible proteins, current methods in protein design have been insufficient. Large barriers in the energy landscape are difficult to traverse while redesigning a protein sequence, and as a result current design methods only sample a fraction of available sequence space. We propose a new computational approach that combines traditional structure-based modeling using the Rosetta software suite with machine learning and integer linear programming to overcome limitations in the Rosetta sampling methods. We demonstrate the effectiveness of this method, which we call BROAD, by benchmarking the performance on increasing predicted breadth of anti-HIV antibodies. We use this novel method to increase predicted breadth of naturally-occurring antibody VRC23 against a panel of 180 divergent HIV viral strains and achieve 100% predicted binding against the panel. In addition, we compare the performance of this method to state-of-the-art multistate design in Rosetta and show that we can outperform the existing method significantly. We further demonstrate that sequences recovered by this method recover known binding motifs of broadly neutralizing anti-HIV antibodies. Finally, our approach is general and can be extended easily to other protein systems. Although our modeled antibodies were not tested in vitro, we predict that these variants would have greatly increased breadth compared to the wild-type antibody.

In this article, we report a new approach for protein design, which combines traditional structural modeling with machine learning and integer programming. Using this method, we are able to design antibodies that are predicted to bind large panels of antigenically diverse HIV variants. The combination of methods from these fields allows us to surpass protein design limitations that have been seen up to this point. We predict that if we tested these modified antibodies against HIV variants they would have greater neutralization breadth than any antibodies seen to this point.

The making of bispecific antibodies

During the past two decades we have seen a phenomenal evolution of bispecific antibodies for therapeutic applications. The 'zoo' of bispecific antibodies is populated by many different species, comprising around 100 different formats, including small molecules composed solely of the antigen-binding sites of two antibodies, molecules with an IgG structure, and large complex molecules composed of different antigen-binding moieties often combined with dimerization modules. The application of sophisticated molecular design and genetic engineering has solved many of the technical problems associated with the formation of bispecific antibodies such as stability, solubility and other parameters that confer drug properties. These parameters may be summarized under the term 'developability'. In addition, different 'target product profiles', i.e., desired features of the bispecific antibody to be generated, mandates the need for access to a diverse panel of formats. These may vary in size, arrangement, valencies, flexibility and geometry of their binding modules, as well as in their distribution and pharmacokinetic properties. There is not 'one best format' for generating bispecific antibodies, and no single format is suitable for all, or even most of, the desired applications. Instead, the bispecific formats collectively serve as a valuable source of diversity that can be applied to the development of therapeutics for various indications. Here, a comprehensive overview of the different bispecific antibody formats is provided.

Computational design of a specific heavy chain/κ light chain interface for expressing fully IgG bispecific antibodies

The use of bispecific antibodies (BsAbs) to treat human diseases is on the rise. Increasingly complex and powerful therapeutic mechanisms made possible by BsAbs are spurring innovation of novel BsAb formats and methods for their production. The long-lived in vivo pharmacokinetics, optimal biophysical properties and potential effector functions of natural IgG monoclonal (and monospecific) antibodies has resulted in a push to generate fully IgG BsAb formats with the same quaternary structure as monoclonal IgGs. The production of fully IgG BsAbs is challenging because of the highly heterogeneous pairing of heavy chains (HCs) and light chains (LCs) when produced in mammalian cells with two IgG HCs and two LCs. A solution to the HC heterodimerization aspect of IgG BsAb production was first discovered two decades ago; however, addressing the LC mispairing issue has remained intractable until recently. Here, we use computational and rational engineering to develop novel designs to the HC/LC pairing issue, and particularly for κ LCs. Crystal structures of these designs highlight the interactions that provide HC/LC specificity. We produce and characterize multiple fully IgG BsAbs using these novel designs. We demonstrate the importance of specificity engineering in both the variable and constant domains to achieve robust HC/LC specificity within all the BsAbs. These solutions facilitate the production of fully IgG BsAbs for clinical use.

A new approach for generating bispecific antibodies based on a common light chain format and the stable architecture of human immunoglobulin G1

Bispecific antibodies combine two different antigen-binding sites in a single molecule, enabling more specific targeting, novel mechanisms of action, and higher clinical efficacies. Although they have the potential to outperform conventional monoclonal antibodies, many bispecific antibodies have issues regarding production, stability, and pharmacokinetic properties. Here, we describe a new approach for generating bispecific antibodies using a common light chain format and exploiting the stable architecture of human immunoglobulin G₁ We used iterative experimental validation and computational modeling to identify multiple Fc variant pairs that drive efficient heterodimerization of the antibody heavy chains. Accelerated stability studies enabled selection of one Fc variant pair dubbed "DEKK" consisting of substitutions L351D and L368E in one heavy chain combined with L351K and T366K in the other. Solving the crystal structure of the DEKK Fc region at a resolution of 2.3 Å enabled detailed analysis of the interactions inducing CH3 interface heterodimerization. Local shifts in the IgG backbone accommodate the introduction of lysine side chains that form stabilizing salt-bridge interactions with substituted and native residues in the opposite chain. Overall, the CH3 domain adapted to these shifts at the interface, yielding a stable Fc conformation very similar to that in wild-type IgG. Using the DEKK format, we generated the bispecific antibody MCLA-128, targeting human EGF receptors 2 and 3. MCLA-128 could be readily produced and purified at industrial scale with a standard mammalian cell culture platform and a routine purification protocol. Long-term accelerated stability assays confirmed that MCLA-128 is highly stable and has excellent biophysical characteristics.

Rosetta:MSF: a modular framework for multi-state computational protein design

Computational protein design (CPD) is a powerful technique to engineer existing proteins or to design novel ones that display desired properties. Rosetta is a software suite including algorithms for computational modeling and analysis of protein structures and offers many elaborate protocols created to solve highly specific tasks of protein engineering. Most of Rosetta’s protocols optimize sequences based on a single conformation (i. e. design state). However, challenging CPD objectives like multi-specificity design or the concurrent consideration of positive and negative design goals demand the simultaneous assessment of multiple states. This is why we have developed the multi-state framework MSF that facilitates the implementation of Rosetta’s single-state protocols in a multi-state environment and made available two frequently used protocols. Utilizing MSF, we demonstrated for one of these protocols that multi-state design yields a 15% higher performance than single-state design on a ligand-binding benchmark consisting of structural conformations. With this protocol, we designed de novo nine retro-aldolases on a conformational ensemble deduced from a (βα)₈-barrel protein. All variants displayed measurable catalytic activity, testifying to a high success rate for this concept of multi-state enzyme design.

Protein engineering, i. e. the targeted modification or design of proteins has tremendous potential for medical and industrial applications. One generally applicable strategy for protein engineering is rational protein design: based on detailed knowledge of structure and function, computer programs like Rosetta propose the sequence of a protein possessing the desired properties. So far, most computer protocols have used rigid structures for design, which is a simplification because a protein’s structure is more accurately specified by a conformational ensemble. We have now implemented a framework for computational protein design that allows certain design protocols of Rosetta to make use of multiple design states like structural ensembles. An in silico assessment simulating ligand-binding design showed that this new approach generates more reliably native-like sequences than a single-state approach. As a proof-of-concept, we introduced de novo retro-aldolase activity into a scaffold protein and characterized nine variants experimentally, all of which were catalytically active.

Asymmetric Fc Engineering for Bispecific Antibodies with Reduced Effector Function

Asymmetric bispecific antibodies are a rapidly expanding therapeutic antibody class, designed to recognize two different target epitopes concurrently to achieve novel functions not available with normal antibodies. Many therapeutic designs require antibodies with reduced or silenced effector function. Although many solutions have been described in the literature to knockout effector function, to date all of them have involved the use of a specific antibody subtype (e.g., IgG2 or IgG4), or symmetric mutations in the lower hinge or CH2 domain of traditional homodimeric monospecific antibodies. In the context of a heterodimeric Fc, we describe novel asymmetric Fc mutations with reduced or silenced effector function in this article. These heteromultimeric designs contain asymmetric charged mutations in the lower hinge and the CH2 domain of the Fc. Surface plasmon resonance showed that the designed mutations display much reduced binding to all of the Fc gamma receptors and C1q. Ex vivo ADCC and CDC assays showed a consistent reduction in activity. Differential scanning calorimetry showed increased thermal stability for some of the designs. Finally, the asymmetric nature of the introduced charged mutations allowed for separation of homodimeric impurities by ion exchange chromatography, providing, as an added benefit, a purification strategy for the production of bispecific antibodies with reduced or silenced effector function.

Transient and stable CHO expression, purification and characterization of novel hetero-dimeric bispecific IgG antibodies

IgG bispecific antibodies (BsAbs) represent one of the preferred formats for bispecific antibody therapeutics due to their native-like IgG properties and their monovalent binding to each target. Most reported studies utilized transient expression in HEK293 cells to produce BsAbs. However, the expression of biotherapeutic molecules using stable CHO cell lines is commonly used for biopharmaceutical manufacturing. Unfortunately, limited information is available in the scientific literature on the expression of BsAbs in CHO cell lines. In this study we describe an alternative approach to express the multiple components of IgG BsAbs using a single plasmid vector (quad vector). This single plasmid vector contains both heavy chain genes and both light chain genes required for the expression and assembly of the IgG BsAb, along with a selectable marker. We expressed, purified, and characterized four different IgG BsAbs or "hetero-mAbs" using transient CHO expression and stable CHO minipools. Transient CHO titers ranged from 90 to 160 mg/L. Stable CHO titers ranged from 0.4 to 2.3 g/L. Following a simple Protein A purification step, the percentage of correctly paired BsAbs ranged from 74% to 98% as determined by mass spectrometry. We also found that information generated from transient CHO expression was similar to information generated using stable CHO minipools. In conclusion, the quad vector approach represents a simple, but effective, alternative approach for the generation of IgG BsAbs in both transient CHO and stable CHO expression systems. © 2017 American Institute of Chemical Engineers Biotechnol. Prog., 33:469-477, 2017.

Fc Engineering for Developing Therapeutic Bispecific Antibodies and Novel Scaffolds

Therapeutic monoclonal antibodies have become molecules of choice to treat autoimmune disorders, inflammatory diseases, and cancer. Moreover, bispecific/multispecific antibodies that target more than one antigen or epitope on a target cell or recruit effector cells (T cell, natural killer cell, or macrophage cell) toward target cells have shown great potential to maximize the benefits of antibody therapy. In the past decade, many novel concepts to generate bispecific and multispecific antibodies have evolved successfully into a range of formats from full bispecific immunoglobulin gammas to antibody fragments. Impressively, antibody fragments such as bispecific T-cell engager, bispecific killer cell engager, trispecific killer cell engager, tandem diabody, and dual-affinity-retargeting are showing exciting results in terms of recruiting and activating self-immune effector cells to target and lyse tumor cells. Promisingly, crystallizable fragment (Fc) antigen-binding fragment and monomeric antibody or half antibody may be particularly advantageous to target solid tumors owing to their small size and thus good tissue penetration potential while, on the other hand, keeping Fc-related effector functions such as antibody-dependent cellular cytotoxicity, complement-dependent cytotoxicity, antibody-dependent cell-mediated phagocytosis, and extended serum half-life via interaction with neonatal Fc receptor. This review, therefore, focuses on the progress of Fc engineering in generating bispecific molecules and on the use of small antibody fragment as scaffolds for therapeutic development.

Immunoglobulin Fc Heterodimer Platform Technology: From Design to Applications in Therapeutic Antibodies and Proteins

The monospecific and bivalent characteristics of naturally occurring immunoglobulin G (IgG) antibodies depend on homodimerization of the fragment crystallizable (Fc) regions of two identical heavy chains (HCs) and the subsequent assembly of two identical light chains (LCs) via disulfide linkages between each HC and LC. Immunoglobulin Fc heterodimers have been engineered through modifications to the CH3 domain interface, with different mutations on each domain such that the engineered Fc fragments, carrying the CH3 variant pair, preferentially form heterodimers rather than homodimers. Many research groups have adopted different strategies to generate Fc heterodimers, with the goal of high heterodimerization yield, while retaining biophysical and biological properties of the wild-type Fc. Based on their ability to enforce heterodimerization between the two different HCs, the established Fc heterodimers have been extensively exploited as a scaffold to generate bispecific antibodies (bsAbs) in full-length IgG and IgG-like formats. These have many of the favorable properties of natural IgG antibodies, such as high stability, long serum half-life, low immunogenicity, and immune effector functions. As of July 2016, more than seven heterodimeric Fc-based IgG-format bsAbs are being evaluated in clinical trials. In addition to bsAbs, heterodimeric Fc technology is very promising for the generation of Fc-fused proteins and peptides, as well as cytokines (immunocytokines), which can present the fusion partners in the natural monomeric or heterodimeric form rather than the artificial homodimeric form with wild-type Fc. Here, we present relevant concepts and strategies for the generation of heterodimeric Fc proteins, and their application in the development of bsAbs in diverse formats for optimal biological activity. In addition, we describe wild-type Fc-fused monomeric and heterodimeric proteins, along with the difficulties associated with their preparations, and discuss the use of heterodimeric Fc as an alternative scaffold of wild-type Fc for naturally monomeric or heterodimeric proteins, to create Fc-fusion proteins with novel therapeutic modality.

Turning a Negative into a Positive: Conversion of a Homodimer into a Heterodimer Using Negative State Repertoires

In this issue of Structure, Leaver-Fay et al. (2016) engineer bispecific antibodies using multistate computational protein design with negative state repertoires. In combination with additional mutations selected rationally, they produced antibodies that assembled as heterodimers with up to 93% purity.