ABSTRACT Antibodies must bind their targets with high affinity and specificity to achieve useful therapeutic activity. They must also possess suitable developability properties (e.g. thermostability, solubility, viscosity, polyreactivity) to ensure favorable manufacturing, formulation, and in vivo performance. Both binding and developability properties are inherent to a given antibody amino acid sequence. Identification or selection of antibodies possessing suitable-binding characteristics is now routine, and de novo computational design models, trained on extensive complementarity-determining region sequence and structural data, are rapidly improving. Developability properties, however, remain difficult to predict largely due to insufficient training data, with empirical testing being heavily used to avoid challenges in late-stage antibody development. To fill this gap, we built a high-throughput antibody developability assay platform designed to generate the large datasets needed to train improved machine learning (ML) models. We optimized and automated known developability assays, and developed a robust integrated data analytics pipeline. Here, we report data on 246 antibodies – representing 106 approved, 135 clinical-stage, and 5 preregistration/withdrawn molecules – across a panel of 10 developability assays, in a “tidy data” format suitable for AI/ML modeling. We used these data to develop an XGBoost ML model that better predicts similarity to approved antibodies compared to conventional use of developability warning thresholds. Additionally, we confirm that preliminary predictive models do improve with more training data. Our high-throughput PROPHET-Ab platform enables data generation at the scale needed to develop improved ML models to predict antibody developability.

ABSTRACT Targeting checkpoint inhibitors is an effective therapy for treating cancer, with human programmed cell death protein 1 (hPD-1) being one of the most successful targets for developing antibody-based drugs. In this work, we isolated a panel of anti-PD-1 single-chain variable fragments with different binding and functional profiles from a fully synthetic human phage display library. Conversion of the best clone to hIgG1LALA and hIgG4PE formats, called UDIZ-007 and UDIZ-008, respectively, resulted in antibodies that effectively blocked the PD-1:PD-L1/L2 interaction and were highly selective as they did not cross-react with CD28 receptor family members. Doses of UDIZ-007 or UDIZ-008 at 10 mg/kg every 3 days for a total of six intraperitoneal administrations eradicated MC38-hPD-L1 colon tumors in B-hPD-1 transgenic mice for hPD-1 at day 17, with no relapse until the end of the study at day 56. Importantly, these antibodies bind hPD-1 in a unique region compared to the anti-PD-1 antibodies of known structure, which might have an impact on novel oncology indications when used as a standalone therapy or in combination with currently approved anti-PD-1 therapeutic antibodies. Therefore, UDIZ-007 and UDIZ-008 seem to be promising candidates for the development of antibody-based drugs targeting checkpoint inhibitors as a treatment for cancer.

ABSTRACT Since the approval of the first antibody drug in 1986, a total of 162 antibodies have been approved for a wide range of therapeutic areas, including cancer, autoimmune, infectious, or cardiovascular diseases. Despite advances in biotechnology that accelerated the development of antibody drugs, the drug discovery process for this modality remains lengthy and costly, requiring multiple rounds of optimizations before a drug candidate can progress to preclinical and clinical trials. This multi-optimization problem involves increasing the affinity of the antibody to the target antigen while refining additional biophysical properties that are essential to drug development such as solubility, thermostability or aggregation propensity. Additionally, antibodies that resemble natural human antibodies are particularly desirable, as they are likely to offer improved profiles in terms of safety, efficacy, and reduced immunogenicity, further supporting their therapeutic potential. In this article, we explore the use of energy-based generative models to optimize a candidate monoclonal antibody. We identify tradeoffs when optimizing for multiple properties, focusing on solubility, humanness and affinity and use the generative model we develop to generate candidate antibodies that lie on optimal Pareto fronts with respect to these properties.

ABSTRACT Progress in developing effective large-molecule therapies for neurological diseases is limited by exposure at the sites of action, beyond the blood–brain barrier (BBB). While transferrin receptor (TfR1)-mediated transport is gaining validation as a mechanism to deliver medicines of multiple modalities to the brain, there is much still to be learned about maximizing the potential of TfR1 targeting. We systematically vary for the first time affinity and valency of two anti‑TfR1 antibodies, which have distinct epitopes and pH sensitivities, to investigate their effects on cellular trafficking, biodistribution, safety, and pharmacokinetics. We establish how transcytosis and receptor degradation trend with affinity and connect these in-vitro functions to in-vivo behaviors in brain uptake and reticulocyte depletion. We identify unique anti-TfR1 antibody profiles for either short-term maximal or long-term sustained brain delivery and demonstrate the utility of these shuttles for different pharmacological applications. Our results show that bivalent anti-TfR1 antibodies can be equally effective in brain uptake as monovalent antibodies if engineered to have similar cell-surface TfR1 binding strength, with prolonged brain exposure and less severe adverse effects, but epitope, as well as affinity and valency, factors into selecting a shuttle with maximal performance. These results challenge the view that monovalent formats are inherently superior and instead establish that affinity, valency, and epitope can be tuned to select TfR1 shuttles optimized for different therapeutic needs.

ABSTRACT With monoclonal antibodies becoming one of the largest classes of biopharmaceuticals, it is important to have curated data to train computational models that can accelerate their design. Despite the massive amount of mutagenesis data generated on antibody–antigen interactions, only a few small, well-curated datasets are available. This paper introduces AbAgym, a manually curated repository comprising approximately 324k mutations in antibody–antigen complexes, including approximately 10% of interface mutations, whose effects on antibody–antigen binding have been experimentally quantified through deep mutational scanning (DMS) experiments. We collected and curated 68 DMS datasets from the literature together with the three-dimensional structure of each antibody–antigen complex. We benchmarked the performance of established force field methods as well as recent machine learning models that predict the change in binding affinity upon mutation. The former achieved modest performance, whereas the latter performed only marginally better than random. Finally, our analysis of hotspot residues responsible for immune evasion highlights the importance of accounting for biological complexities, such as conformational changes or oligomeric states that influence antibody–antigen binding, which are often overlooked. Abagym is freely available for academic use at https://github.com/3BioCompBio/Abagym.

ABSTRACT Clinical and preclinical studies have confirmed insulin-like growth factor-1 receptor (IGF-1R) antagonism can reduce the inflammation and proptosis occurring in thyroid eye disease (TED). We assessed the preclinical pharmacology, pharmacokinetics, and pharmacodynamics of veligrotug (formerly VRDN-001), an anti-IGF-1R antibody in clinical development for TED. Veligrotug exhibited high-affinity binding to human IGF-1R protein (KD 0.55 nM) and IGF-1R endogenously expressed in HOCF cells (mean EC50 2.41 nM). Veligrotug did not bind to the insulin receptor in ELISA assays or inhibit insulin-mediated receptor phosphorylation in HepG2 cells. Binding epitope and antagonist properties were compared to teprotumumab (Tepezza®), a marketed anti-IGF-1R antibody. Mutational scan analysis demonstrated veligrotug and teprotumumab have overlapping but distinct binding epitopes. Veligrotug behaved as a full antagonist, providing near-complete inhibition of IGF-1 binding at ≥ 50 nM, in contrast to teprotumumab which plateaued at ~50% inhibition. Veligrotug provided near-complete inhibition of IGF-1R autophosphorylation and AKT phosphorylation, in contrast to partial inhibition by teprotumumab. Veligrotug pharmacokinetic parameters in cynomolgus monkeys were consistent with other human/humanized antibodies in monkeys: half-life was ~5–6 days, serum clearance was low (7.6‒12.9 mL/day/kg), and volume of distribution was low (64‒93 mL/kg). A robust pharmacodynamic response was observed after a single dose of veligrotug, with ~2.5-fold increases in IGF-1 levels that remained elevated throughout the dosing period for the 10 mg/kg and 50 mg/kg dose groups. Veligrotug’s pharmacologic, pharmacokinetic, and pharmacodynamic characteristics make it a good candidate for clinical development. Indeed, efficacy data at week 15 from two Phase 3 pivotal studies of veligrotug, THRIVE and THRIVE-2, showed statistically significant improvements in TED symptoms based on primary and secondary outcomes.

ABSTRACT Studies in animal models are essential to expanding the scope of interventions evaluated for safety, immunogenicity, and efficacy in clinical trials. Ferrets (Mustela putorius furo) are a key small-animal model for examining acquisition, replication, transmission, and disease manifestation, with particular relevance in modeling diverse viruses that target the respiratory tract. However, despite use in studies of vaccine immunogenicity and protection, as well as passive antibody transfer, there is little data characterizing antibody and Fc receptor biology in this species. To address this gap, ferret Fcγ receptors (FcγRI, FcγRII, FcγRIII) were identified and recombinantly expressed and characterized for binding to recombinant ferret and human IgG. In general, ferret IgG bound each receptor with slightly higher affinity than the human IgG subclasses, which exhibited similar ferret receptor binding profiles as observed for human receptors (IgG1 and IgG3 > IgG4 > IgG2). N-linked glycosylation motifs on ferret receptors were typically occupied, and binding was dependent on IgG Fc glycosylation. While further insight into the expression patterns and activities of innate immune cells stimulated by IgG is still needed, these data define Fc – FcγR recognition patterns in ferrets to help support optimal clinical translation of passive and active immunization studies.

ABSTRACT First-generation cancer immunotherapies, such as high-dose interleukin-2 (IL-2), have demonstrated clinical efficacy, but are limited by significant systemic toxicities due to their broad expression of cytokine receptors. This has driven the iterative development of targeted cytokine delivery strategies. Early efforts focused on receptor-biased IL-2 variants designed to attenuate or abrogate IL-2 receptor α (IL-2 Rα/CD25) binding. Subsequently, the concept of “cis-targeting” has emerged as a strategy to deliver cytokines to specific immune cell populations, enhancing anti-tumor responses while mitigating systemic toxicity. This review highlights key common γ-chain cytokines (IL-2, IL-7, IL-15, and IL-21) as well as IL-12, providing an overview of their structures, receptors, as well as their distinct T cell functions. Furthermore, we specifically focus on the current landscape of engineered cytokine variants that facilitate targeted cytokine delivery in cis to specific T cells. By successfully restricting cytokine activity to specific T cell populations, cis-targeting approaches represent a promising strategy in the field, enabling efficient immunotherapies with improved tolerability and enhanced anti-tumor responses.

ABSTRACT Alpha-galactosylation, galactose-α-1,3-galactose (α-Gal), is always regarded as a critical quality attribute due to its potential to provoke immunogenic responses in patients. Consequently, monitoring alpha-galactosylation in therapeutic proteins is essential, but current analytical techniques fall short in terms of identification sensitivity and quantification accuracy. Specifically, the released glycan assay by hydrophilic interaction liquid chromatography-fluorescence-mass spectrometry, the gold standard for glycan separation/identification, faces challenges due to ambiguities with isomeric glycan structures. To address these challenges, we developed a comprehensive analytical method that enhances both the identification sensitivity and quantification accuracy for α-Gal. We developed an integrated workflow that combines an advanced mass spectrometry technique – parallel reaction monitoring triple-stage mass spectrometry – with exoglycosidase sample treatment. This approach generates structurally specific signature ions, enhances identification sensitivity, enables the separation of α-Gal from its isomers, and improves quantification accuracy. By employing this more sensitive analytical approach without ambiguity in assignment for common glycan structures found in monoclonal antibodies, the safety of therapeutic proteins can be better assured, effectively minimizing the risk of α-Gal-induced immunogenicity.

ABSTRACT T cell bispecific antibodies (TCBs) are an emerging class of cancer therapy that are typically designed for high binding affinity to CD3 and tumor antigen (TA). Using this approach, TCBs have demonstrated significant clinical efficacy, but they have also elicited cytokine release syndrome and off-target on-tumor toxicities. CD3 affinity-attenuation has recently been reported as an approach to maintain efficacy while reducing cytokine release, but the interdependence of CD3 affinity with other factors is often not systematically explored. For this purpose, we generated a series of TCBs comprising CD3 binders with varying affinities and TA binders with either high or low affinities, utilizing FOLR1 and CEACAM5 as tumor targets. The CD3 binders were classified into high, intermediate, low, and very low affine binders based on affinity measurements as well as functionality. Depending on the target, different combinations of binders showed the most advantageous profile of tumor-cell killing while coupled with lower cytokine secretion. For instance, within the FOLR1-TCBs series, CD3intermed exhibited a favorable profile compared to CD3high in vitro using cocultures and in vivo using humanized mice. For CEACAM5-TCBs, CD3low, along with CD3intermed, showed a favorable profile compared to CD3high in both in vitro and in vivo settings. Additionally, CD3low avoided on-target, off-tumor toxicity and reduced cytokine release in transgenic mice. Taken together, reducing cytokine release while maintaining adequate efficacy is possible through CD3 binder affinity attenuation, but optimizing cytokine release profiles by CD3 binder affinity-attenuation is dependent on additional parameters.