(T1030-01-01) Using Flow Imaging Microscopy to Meet Particulate Matter Standards for Biotherapeutics

Purpose: Particles are ubiquitous in biotherapeutics and have been associated with significant changes in the safety and efficacy of these life-saving medicines. The United States Pharmacopeia (USP) requires researchers to monitor subvisible (2-100 μm) and visible ( >100 μm) particles in biologics. However, the compendial particle monitoring methods are insensitive to relevant particle types like protein aggregates and gray zone particles (i.e. visible particles that are unlikely to be observed via standard visual inspection processes). These methods also offer limited information about particle types and sources. Flow imaging microscopy (FIM) is high-throughput, light microscopy-based analytical technique that provides accurate particle concentration, size, and type information for both subvisible and visible particles. Capturing these measurements can help scientists design biotherapeutics with minimal particulate matter, ensuring compliance with pharmacopeia guidelines. This poster presents case studies to demonstrate how FIM measurements can help researchers monitor and control particulate matter in biotherapeutic samples.

Methods: The first case study involved using FIM to investigate the impact of surfactants on protein aggregation during shaking stress. Three simple formulations in PBS were prepared: one containing BSA, one containing polysorbate 80, and another containing both BSA and polysorbate 80. Each formulation was stressed via an accelerated shaking stress to generate subvisible particles. The resulting particles were then analyzed via FlowCam LO, an instrument that provides simultaneous FIM and light obscuration (LO) data. The latter is the compendial technique for monitoring subvisible particles via USP < 788 >. In the second case study, we analyzed samples via FIM and a simulated visual inspection process to compare the sensitivity of each method to visible particles. Six vials of simulated particle-free drug product in PBS were prepared. Three “test” vials were then spiked with visible particles, while the remaining “control” vials were spiked with additional PBS. Each vial was inspected in a lightbox for visible particles and then analyzed via an FIM instrument configured to analyze visible particles. The third case study involved using image analysis tools within FIM instrument software to distinguish between protein aggregate and silicone oil droplet particles. Protein aggregates were prepared by stressing an IVIg formulation in PBS via agitation. Silicone oil droplets were prepared by blending silicone oil in PBS to create an emulsion. Aliquots of each sample were imaged via FIM and the resulting particle images were then analyzed via VisualAI, a pre-trained AI tool for identifying images of protein aggregates and silicone oil droplets.

Results: In the first case study, FIM detected higher particle concentrations and larger particle sizes in all three samples than were detected by LO.  Comparing the different samples, the protein-only formulation contained higher particle concentrations than the other investigated formulations, suggesting extensive aggregation had occurred. In contrast, very few particles were observed in either the surfactant only or protein and surfactant formulations. The sharp decrease in particle concentrations upon the addition of surfactant suggests that, unsurprisingly, surfactant helps reduce protein aggregation during agitation stresses. This approach can be used to assess how formulation changes influence particle formation, enabling formulation scientists to design drug products that contain minimal particulate matter. In the second case study, visible particles could not be observed in either the control or test vials when inspected in a lightbox. When these samples were then analyzed via FIM, visible and gray zone particles were observed in the test vials but not the control vials. These results demonstrate how FIM can help researchers reliably detect visible and gray zone particles in biotherapeutics—even in scenarios where they are challenging to detect via visual inspection. In the third case study, VisualAI was capable of distinguishing between FIM images of protein aggregates and silicone oil droplets with over 90% accuracy. Operators can use this approach to reliably determine the concentration of both particle types in biotherapeutic samples. These measurements can be used to inform formulation and quality control decisions based on the concentrations of each particle type present in a sample in addition to the overall particle content.

Conclusion: In each of these case studies, FIM provided insight into the amount and types of particles in biotherapeutic samples. In the first and second case studies, FIM measured the amount of subvisible and visible particles, respectively, present in different samples—even particles that traditional particle analysis methods failed to detect. The third case study demonstrates how the particle images provided by FIM can be used to determine particle type and source, information that can help scientists influence the amount of particles in a sample. The insight provided by each case study can be used to design drug products that contain fewer particles, helping guarantee compliance with pharmacopeia standards.

Acknowledgements: Both Austin Daniels and Sigrid Kuebler are employed by Yokogawa Fluid Imaging Technologies, the manufacturer of the FIM instruments and software used in this poster.

Subvisible particles measured in different formulations following agitation stress. Dark shaded regions indicate FIM measurements and light shaded regions indicate LO measurements. (Left) particle counts measured in different size ranges. (Center) Particle size distributions measured for each sample. (Right) Collages of sample particle images captured by FIM.

(Left) Images of control and test vials taken inside a light box. No visible particles were observed while inspecting the particles in the lightbox. (Right) collages of particles observed via FIM in the control and test vials. Collages on the left contain subvisible particles 50-100 μm in size, while the collage on the right contain visible and gray zone particles >100 μm in size.

(Left and center) Collages of sample FIM images of protein aggregates and silicone oil droplets. (Right) confusion matrix showing the classification performance of VisualAI on FIM images known to contain a protein aggregate or silicone oil droplet.