BIONEWS
Antibody drugs and gene therapies are now harder to clean up than to make. A single purification platform from NC State handles both.
The most expensive, failure-prone step in making antibody drugs and gene therapies is no longer the biology. It is the cleanup. A 24-person team at North Carolina State University has re-engineered that step with a modular, flow-through platform that decouples impurity removal from product capture, letting a single architecture handle bispecific antibodies, Fc-fusion proteins, and adeno-associated viral (AAV) vectors, the delivery vehicles used in most gene therapies.
For two decades, the manufacturing progress lived upstream: bigger fermenters, denser cell cultures, more drug per liter. Higher titers meant more impurities moved downstream, and cleanup steps were tuned molecule by molecule. The modality mix is now shifting toward bispecific antibodies, Fc-fusion proteins, and AAV vectors, and the same impurities that once sat comfortably inside a process now show up as the proximate cause of the costly failures.
Wenning Chu, a research scholar at NC State and first author on the new paper, and colleagues describe a platform built around one architectural move: pull impurity clearance out of the polishing step and run it as a flow-through operation ahead of, or alongside, the molecule-specific capture step. The goal is to make impurity removal generic, so the modality-specific work, the part that traditionally takes a development team a year or more of bespoke column work, only has to handle the molecule itself.
For the protein workflow, the team built a CHO-cell (Chinese hamster ovary, the standard factory cell for biologic drugs) train that runs a peptide-ligand pre-capture resin paired with size-exclusion chemistry, then Protein A capture (the industry-standard antibody-binding step), then a single-use mixed-mode polishing resin. The reported numbers, all bench-scale: product yields above 70 percent, a final product pool of 19 to 23 grams per liter per hour, monomeric purity (the share of correctly assembled antibody) at roughly 99 percent, and host cell protein clearance (the amount of leftover cell debris in the final drug substance) of 4 to 11 parts per million. The industry-standard final pool runs closer to 15 milligrams per milliliter, so the throughput gap is real even at this scale.
The gene-therapy workflow runs a different train. It uses a mixed-bed absorbent, then either an AAV-specific resin or a single-use high-capacity chromatography membrane, with a 1:3 charcoal-to-resin ratio for impurity stripping. Recovery comes in around 50 percent, host cell protein lands at 350 nanograms per milliliter, and the dose-relevant impurity load stays under 100 nanograms per dose. The numbers are bench-scale and single-laboratory, but the architecture is the point: the same upstream impurity-removal logic holds for a CHO protein run and an AAV run.
Chu's earlier work on peptide ligands for AAV capture from HEK 293 lysates (the human cell line used to make most AAV) laid the technical foundation for the new platform. The trade-press framing of the same gap, per Genetic Engineering & Biotechnology News, treats purification as the new bottleneck, which is what the architecture is positioned against. The architecture itself, flow-through chemistry run modularly so impurity removal happens before the molecule-specific step ever has to see it, is the move.
The honest limits are real. These are bench-scale numbers from a single research group in a single paper. No third-party lab has run the workflow, no good manufacturing practice (GMP) line has reproduced it, and no regulator has weighed in. The CHO and AAV trains share an architectural philosophy but not a literal column, and translation to commercial scale is the part that historically breaks academic bioprocess results. Adoption at a contract development and manufacturing organization, or inside a large biopharma, would tell readers whether the platform is more than a paper.
The field is finally treating purification as a design problem rather than an afterthought. That shift, more than any single set of recovery numbers, is what the next round of papers and process trains will test.