Tiexin Wang

Antibody development, delivery, and efficacy are influenced by antibody-antigen affinity interactions, off-target interactions that reduce antibody bioavailability and pharmacokinetics, and repulsive self-interactions that increase the stability of concentrated antibody formulations and reduce their corresponding viscosity. Yet identifying antibody variants with optimal combinations of these three types of interactions is challenging. Here we show that interpretable machine-learning classifiers, leveraging antibody structural features descriptive of their variable regions and trained on experimental data for a panel of 80 clinical-stage monoclonal antibodies, can identify antibodies with optimal combinations of low off-target binding in a common physiological-solution condition and low self-association in a common antibody-formulation condition. For three clinical-stage antibodies with suboptimal combinations of off-target …

Early identification of antibody candidates with drug-like properties is essential for simplifying the development of safe and effective antibody therapeutics. For subcutaneous administration, it is important to identify candidates with low self-association to enable their formulation at high concentration while maintaining low viscosity, opalescence, and aggregation. Here, we report an interpretable machine learning model for predicting antibody (IgG1) variants with low viscosity using only the sequences of their variable (Fv) regions. Our model was trained on antibody viscosity data (>100 mg/mL mAb concentration) obtained at a common formulation pH (pH 5.2), and it identifies three key Fv features of antibodies linked to viscosity, namely their isoelectric points, hydrophobic patch sizes, and numbers of negatively charged patches. Of the three features, most predicted antibodies at risk for high viscosity, including …

Antibody-drug conjugates (ADCs) have experienced a surge in clinical approvals in the past five years. Despite this success, a major limitation to ADC efficacy in solid tumors is poor tumor penetration, which leaves many cancer cells untargeted. Increasing antibody doses or co-administering ADC with an unconjugated antibody can improve tumor penetration and increase efficacy when target receptor expression is high. However, it can also reduce efficacy in low-expression tumors where ADC delivery is limited by cellular uptake. This creates an intrinsic problem because many patients express different levels of target between tumors and even within the same tumor. Here, we generated High-Avidity, Low-Affinity (HALA) antibodies that can automatically tune the cellular ADC delivery to match the local expression level. Using HER2 ADCs as a model, HALA antibodies were identified with the desired HER2 expression-dependent competitive binding with ADCs in vitro. Multi-scale distribution of trastuzumab emtansine and trastuzumab deruxtecan co-administered with the HALA antibody were analyzed in vivo, revealing that the HALA antibody increased ADC tumor penetration in high-expression systems with minimal reduction in ADC uptake in low-expression tumors. This translated to greater ADC efficacy in immunodeficient mouse models across a range of HER2 expression levels. Furthermore, Fc-enhanced HALA antibodies showed improved Fc-effector function at both high and low expression levels and elicited a strong response in an immunocompetent mouse model. These results demonstrate that HALA antibodies can expand treatment …

Chinese hamster ovary (CHO) cells, predominant hosts for recombinant biotherapeutics production, generate lactate as a major glycolysis by‐product. High lactate levels adversely impact cell growth and productivity. The goal of this study was to reduce lactate in CHO cell cultures by adding chemical inhibitors to hexokinase‐2 (HK2), the enzyme catalyzing the conversion of glucose to glucose 6‐phosphate, and examine their impact on lactate accumulation, cell growth, protein titers, and N‐glycosylation. Five inhibitors of HK2 enzyme at different concentrations were evaluated, of which 2‐deoxy‐ d‐glucose (2DG) and 5‐thio‐ d‐glucose (5TG) successfully reduced lactate accumulation with only limited impacts on CHO cell growth. Individual 2DG and 5TG supplementation led to a 35%–45% decrease in peak lactate, while their combined supplementation resulted in a 60% decrease in peak lactate. Inhibitor …

Glycoproteomic analysis of three Chinese hamster ovary (CHO) suspension host cell lines (CHO-K1, CHO-S, and CHO-Pro5) commonly utilized in biopharmaceutical settings for recombinant protein production is reported. Intracellular and secreted glycoproteins were examined. We utilized an immobilization and chemoenzymatic strategy in our analysis. Glycoproteins or glycopeptides were first immobilized through reductive amination, and the sialyl moieties were amidated for protection. The desired N- or O-glycans and glycopeptides were released from the immobilization resin by enzymatic or chemical digestion. Glycopeptides were studied by Orbitrap Liquid chromatography–mass spectrometry (LC/MS), and the released glycans were analyzed by Matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF). Differences were detected in the relative abundances of N- and O-glycopeptide types, their …

Therapeutic antibody development requires selection and engineering of molecules with high affinity and other drug-like biophysical properties. Co-optimization of multiple antibody properties remains a difficult and time-consuming process that impedes drug development. Here we evaluate the use of machine learning to simplify antibody co-optimization for a clinical-stage antibody (emibetuzumab) that displays high levels of both on-target (antigen) and off-target (non-specific) binding. We mutate sites in the antibody complementarity-determining regions, sort the antibody libraries for high and low levels of affinity and non-specific binding, and deep sequence the enriched libraries. Interestingly, machine learning models trained on datasets with binary labels enable predictions of continuous metrics that are strongly correlated with antibody affinity and non-specific binding. These models illustrate strong tradeoffs …

The N‐glycan pattern of an IgG antibody, attached at a conserved site within the fragment crystallizable (Fc) region, is a critical antibody quality attribute whose structural variability can also impact antibody function. For tailoring the Fc glycoprofile, glycoengineering in cell lines as well as Fc amino acid mutations have been applied. Multiple glycoengineered Chinese hamster ovary cell lines were generated, including defucosylated (FUT8KO), α‐2,6‐sialylated (ST6KI), and defucosylated α‐2,6‐sialylated (FUT8KOST6KI), expressing either a wild‐type anti‐CD20 IgG (WT) or phenylalanine to alanine (F241A) mutant. Matrix‐assisted laser desorption ionization‐time of flight mass spectrometry characterization of antibody N‐glycans revealed that the F241A mutation significantly increased galactosylation and sialylation content and glycan branching. Furthermore, overexpression of recombinant human α‐2,6 …

In mammalian cells, N-glycans may include multiple N-acetyllactosamine (poly-LacNAc) units that can play roles in various cellular functions and properties of therapeutic recombinant proteins. Previous studies indicated that β-1,3-N-acetylglucosaminyltransferase 2 (B3GNT2) and β-1,4-galactotransferase 1 (B4GALT1) are two of the primary glycosyltransferases involved in generating LacNAc units. In the current study, knocking out sialyltransferase genes slightly enhanced the LacNAc content (≥4 repeats per glycan) on recombinant EPO protein. Next, the role of single and dual-overexpression of B3GNT2 and B4GALT1 was explored in recombinant EPO-expressing Chinese hamster ovary (CHO) cells. While overexpression of B4GALT1 slightly enhanced the levels of large glycans on recombinant EPO, overexpression of B3GNT2 in EPO-expressing CHO cells significantly decreased the recombinant EPO LacNAc …