What Happens to Biologics During Cleaning? New Data on Cleaning-Induced Protein Degradants

What happened

The authors report a bench-scale experimental validation of an existing parenteral PDE for cleaning-induced protein degradants, published in the Journal of Pharmaceutical Sciences (2025;114(2):866-877; DOI 10.1016/j.xphs.2024.10.041) and authored by Joseph R. Cohen, Marisa K. Joubert, Syeda Tabassum, Allyson Capili, Julia Carreon, Cathie Xiang, Siddharth Prabhu, A. Merlo, Dan Mytych, David G. Dolan and Ram Kouda (Amgen).

Multiproduct biologic manufacturing leaves protein residue on shared product-contact surfaces. Because those surfaces see extreme pH and temperature during cleaning, the residue is not intact drug substance but a mixture of fragments. Acceptance limits are still derived from MAC calculations anchored to a PDE - and for biologics the relevant PDE is the one for the degradant, not the active molecule.

Cohen et al. used a bench-scale cleaning method to generate degradants from a half-life-extended (HLE) BiTE molecule and a monoclonal antibody (mAb-1). Degradants were characterised alone and after spiking 650 ug of HLE BiTE-A or mAb-1 degradants into mAb-1 - into both a same-product and a cross-product background. The panel covered endotoxin testing, size exclusion chromatography (SEC), light obscuration by HIAC and micro-fluidic imaging (MFI). Immunogenic potential was assessed with a cell-line assay for immune activation and a human PBMC assay measuring T-cell activation, proliferation and cytokine release at 20 h and 7 days.

The authors report that the cleaning procedure caused significant degradation of the molecules, which were no longer in intact or active form. The presence of 650 ug of degradant carryover - alone or spiked into the same or a cross product - did not increase immunogenicity risk in the cell-based assays. Their conclusion: the current PDE of 650 ug/dose for cleaning-induced degradant carryover carries no immunogenicity risk to patients.

Why it matters

For small molecules, the PDE is a toxicity number and the MAC calculation follows from it. For biologics this logic breaks down: the active protein may have a PDE far below the sensitivity of the analytical methods and below the capability of the cleaning process, so a limit set on intact protein is neither measurable nor achievable. The practical resolution is to set the limit on the degradation state that actually survives cleaning. That is why 650 ug/dose matters: it sits roughly two orders of magnitude above the accepted PDE for an active therapeutic protein and converts an unverifiable low-ppm swab limit into something the process can demonstrate.

The value itself is not new. It traces to clinical experience with gelatin as a reference immunogen - gelatin is itself a protein hydrolysate produced under pH and temperature extremes and of non-human origin, a conservative surrogate for degraded human fragments. What this study adds is orthogonal support from human-cell immunogenicity assays rather than literature extrapolation.

What is particularly interesting

First, the study tests the degradant in the presence of the subsequently manufactured product, not in buffer alone - the configuration that matters for multiproduct changeover. The authors report that spiking into active mAb-1 gave no elevated response above the media-only background.

Second, the readout window is 20 h and 7 days, so the PBMC work probes the late adaptive phase (T-cell activation and proliferation), not merely an acute innate cytokine spike. An argument built only on the first hours of a cytokine assay is easy to attack in review; a 7-day endpoint is harder to dismiss.

Third, the analytical package is deliberately orthogonal: endotoxin for pyrogenic load, SEC for fragment distribution, HIAC and MFI for subvisible particles.

Limitations

• Entirely in vitro. Cell-based immunogenicity assays are predictive tools with known limits; no in vivo or clinical confirmation is presented.
• Degradants came from a bench-scale cleaning surrogate, not production-scale CIP. Worst-case equivalency to the plant cycle is argued, not demonstrated.
• Only one HLE BiTE and one mAb were evaluated, so modality coverage is narrow and protein-specific effects cannot be excluded.
• Residual intact protein in the degradant mixture and full assay cut-points sit behind the paywall and could not be verified.

Questions for our own programme

• Which of our products degrade to completion under our cleaning conditions, and do we hold SEC/HIAC/MFI data or only a total-protein assay?
• Is our PDE justification for inactivated residual protein documented as a distinct value from the active-molecule PDE?
• If we lean on a degradant-based limit, can we defend worst-case equivalence between our bench-scale study and plant CIP?

Wider context

This paper sits inside a well-established argument in biopharmaceutical cleaning validation: degrade first, set the limit second. Published positions on degraded-protein residues based on gelatin and on in vitro PBMC work have argued acceptable exposures for inactive fragments of 1 mg/dose and above, roughly 100x that for active protein. Parallel work on resin reuse and on clearance of degraded protein through purification reaches consistent conclusions from the process side.

The regulatory background is EMA's risk-based requirement for health-based PDE values in multiproduct facilities, adopted in 2015 and reinforced by PIC/S from 2018. For biologics that requirement demands an explicit position on what the residue is after cleaning.

Sources

• Cohen JR, et al. J Pharm Sci. 2025;114(2):866-877. DOI 10.1016/j.xphs.2024.10.041. PMID 39490658.
• PubMed abstract, PMID 39490658.
• BioProcess International: Acceptable Exposure Limit for Inactive Therapeutic Protein Fragments.
• BioProcess International: Cleaning Validation Acceptance Limits for Biological Process Residues, Part 1.
• PDA J Pharm Sci Technol 74(4):377.