Zahra Rattray

The manufacture of heterodimeric antibodies is complicated by chain mispairing, aggregation, and physical instability. The impact of assembly strategy on heterodimeric antibody critical quality attributes remains underexplored. Here, five heterodimeric, monospecific, bivalent antibodies containing asymmetric constant domain substitutions that drive heterodimeric pairing have been generated by either direct cellular expression or via 1:1 mixing of independently expressed heavy-light chain pairs followed by reduction and oxidation (REDOX-pairing). These were compared against each other and a standard IgG1 antibody variant of matched sequence using an analytical panel comprising liquid chromatography–mass spectrometry, size-exclusion chromatography, hydrophobic interaction chromatography, dynamic light scattering, asymmetric flow field-flow fractionation, and surface plasmon resonance. Stability profiles were compared following incubation under accelerated stress (50 °C) and refrigerated storage (4 °C) conditions. REDOX-pairing markedly reduced half-antibody impurities relative to co-expression whilst demonstrating equivalent sequence identity and stability. Importantly, binding affinity of the heterodimeric molecules was comparable to their native IgG1 format comparators. REDOX-pairing therefore represents a viable approach for the generation of heterodimeric antibodies, such as bi-specifics, that are comparable to directly expressed material but offers improved control for minimising contaminants.

1. Breast cancer is a leading cause of global cancer-related mortality in women, with annual cases projected to reach over 3.5 million by 2050 [1]. Breast cancer is typically characterized by disti...

Biparatopic antibodies (bpAbs) have two non-overlapping epitopes on a single antigen, offering a distinct therapeutic potential, yet their application to soluble cytokines remains underexplored. Herewithin, we engineered and characterized biparatopic antibodies targeting human interleukin-23 (IL-23) by combining the clinically validated variable regions of ustekinumab and guselkumab into 1 + 1 IgG-like and 2 + 2 Fab-IgG extended biparatopic formats. Constructs were generated using heterodimer promoting C_H3 mutations and redox repair assembly of half‑antibodies, expressed in HEK-2936E cells, and benchmarked against parental monospecific antibodies. An orthogonal analytical workflow, comprising sodium dodecyl sulfate polyacrylamide gel electrophoresis, liquid chromatography-mass spectrometry, analytical size-exclusion chromatography, surface plasmon resonance (SPR), and mass photometry was used to define critical quality attributes and characterize biparatopic target engagement. SPR kinetic analysis demonstrated picomolar affinities across constructs and revealed avidity‑driven enhancements in apparent affinity. A bespoke SPR dual‑engagement bridging assay confirmed simultaneous IL-23A and IL-12B binding, validating obligate biparatopic engagement. Mass photometry independently corroborated in trans immune complex formation and revealed stoichiometry-dependent multimerisation across the bpAb formats explored. Functional neutralization was measured using an IL-23 luciferase reporter assay and supported by a bespoke, two-parameter mechanistic model, which accurately predicted observed EC₅₀ shifts. Biparatopic formats exhibited equivalent potency relative to equimolar antibody mixtures, establishing a generalizable framework for engineering and modeling soluble cytokine‑targeting biparatopic therapeutics.

Adsorption-loaded silk nanoparticles exploit protein electrostatics and β-sheet-rich architecture to achieve high doxorubicin loading and pH-responsive release, while preserving silk structure.

Central to the success of therapeutic antibodies is the ability to elicit biological effects through mechanisms such as direct antagonism, recruitment of immune effector cells, or internalization o...

Herpesviruses infect nearly all humans and have long been implicated in autoimmune and chronic diseases, yet their immune interactions with host proteins have not been systematically characterized at population scale. We profiled immunoglobulin G reactivities to >4,600 herpesvirus peptides and >15,000 human proteins using multiplexed protein display in two Mass General Brigham Biobank cohorts (discovery n=1,289; replication n=763), with longitudinal electronic health record follow-up. We identified and replicated 3,943 FDR-significant associations across 93 autoantigens, including previously uncharacterized viral–autoantigen axes, such as PHLDA1 and ZNF550. Eleven autoantigens were predicted by viral peptide reactivities with >85% accuracy in independent validation; some exhibited shared viral–host sequence homology, consistent with possible molecular mimicry. Integrating immune reactivities with incident disease outcomes revealed virus-specific network architectures: cytomegalovirus formed the largest multimorbid network, Epstein–Barr virus converged on pleiotropic autoimmune hubs, and herpes simplex viruses formed smaller, partially overlapping autoreactive networks. These findings define a herpesvirus–autoantigen–disease network atlas and prioritize candidate viral–host immune axes for mechanistic investigation.

This study presents the development of frit-inlet asymmetric flow field-flow fractionation for analysing heterogeneous nanoparticle mixtures, introducing performance metrics beyond current ISO standards.

A stepwise optimisation strategy for ASO-loaded LNPs integrating microfluidic LNP manufacture with high-resolution CQA analysis, enables data-driven formulation refinement, and accelerates LNP design.

The number of nucleic acid therapeutics is set to grow within the pharmaceutical industry sector, deploying nanocarrier-based delivery systems as drug products. Poly(A) is a widely used model sequence used in lipid nanoparticle (LNP) formulations for which there are no reported critical quality attributes (CQAs) such as molecular weight, chain length, and impurity profile. In this study, we analyze Poly(A) from three different vendors to measure any existing differences in their CQAs. Poly(A) from these brands was encapsulated in SM102 LNPs using microfluidics to produce three branded Poly(A)-based LNPs. We utilized an orthogonal analytical pipeline approach for both Poly(A) drug substance and LNP drug product CQA evaluation, which included a combination of dynamic light scattering and flow field flow fractionation multiplexed with inline UV, dynamic, and multiangle light-scattering detectors. Similar purity (260/280) values of >3 were obtained across all three brands of Poly(A); however, distinct differences in molecular weight and chain length distributions were identified across Poly(A) brands, with this study representing the first to apply EAF4 methodology for in-depth characterization of model RNA drug substances. Brand A produced a smaller and broader molecular weight distribution, followed by Brand B, and then Brand C produced the largest molecular weight species and the most uniform molecular weight distribution. On encapsulation in LNPs, differences seen in Poly(A) CQAs did not translate to CQA differences in resultant LNPs. We show that a deeper understanding of drug substance CQAs and their subsequent impact on resultant overall drug product characteristics is needed on a case-by-case basis. We show correlations between analytical pipelines, with future work investigating the impact of RNA molecular weight in LNP formulations with different lipid compositions and using these correlations in AI or machine learning to further enhance our knowledge of the correlation between drug substance and resultant drug product CQAs.