Features in Backgrounds of Microscopy Images Introduce Biases in Machine Learning Analyses.

Subvisible particles may be encountered throughout the processing of therapeutic protein formulations. Flow imaging microscopy (FIM) and backgrounded membrane imaging (BMI) are techniques commonly used to record digital images of these particles, which may be analyzed to provide particle size distributions, concentrations, and identities. Although both techniques record digital images of particles within a sample, FIM analyzes particles suspended in flowing liquids, whereas BMI records images of dry particles after collection by filtration onto a membrane. This study compared the performance of convolutional neural networks (CNNs) in classifying images of subvisible particles recorded by both imaging techniques. Initially, CNNs trained on BMI images appeared to provide higher classification accuracies than those trained on FIM images. However, attribution analyses showed that classification predictions from CNNs trained on BMI images relied on features contributed by the membrane background, whereas predictions from CNNs trained on FIM features were based largely on features of the particles. Segmenting images to minimize the contributions from image backgrounds reduced the apparent accuracy of CNNs trained on BMI images but caused minimal reduction in the accuracy of CNNs trained on FIM images. Thus, the seemingly superior classification accuracy of CNNs trained on BMI images compared to FIM images was an artifact caused by subtle features in the backgrounds of BMI images. Our findings emphasize the importance of examining machine learning algorithms for image analysis with attribution methods to ensure the robustness of trained models and to mitigate potential influence of artifacts within training data sets.

Miniaturized Forced Degradation of Therapeutic Proteins and ADCs by Agitation-Induced Aggregation Using Orbital Shaking of Microplates.

Microplate-based formulation screening is a powerful approach to identify stabilizing excipients for therapeutic proteins while reducing material requirements. However, this approach is sometimes not representative of studies conducted in relevant container closures. The present study aimed to identify critical parameters for a microplate-based orbital shaking method to screen biotherapeutic formulations by agitation-induced aggregation. For this purpose, an in-depth methodological study was conducted using different shakers, microplates, and plate seals. Aggregation was monitored by size exclusion chromatography, turbidity, and backgrounded membrane imaging. Both shaker quality and liquid-seal contact had substantial impacts on aggregation during shaking and resulted in non-uniform sample treatment when parameters were not suitably selected. The well volume to fill volume ratio (Vwell/Vfill) was identified as an useful parameter for achieving comparable aggregation levels between different microplate formats. An optimized method (2400 rpm [ac 95 m/s2], Vfill 60-100 µL [Vwell/Vfill 6-3.6], 24 h, RT, heat-sealed) allowed for uniform sample treatment independent of surface tension and good agreement with vial shaking results. This study provides valuable guidance for miniaturization of shaking stress studies in biopharmaceutical drug development, facilitating method transfer and comparability between laboratories.

An Interlaboratory Comparison on the Characterization of a Sub-micrometer Polydisperse Particle Dispersion.

The measurement of polydisperse protein aggregates and particles in biotherapeutics remains a challenge, especially for particles with diameters of ≈ 1 µm and below (sub-micrometer). This paper describes an interlaboratory comparison with the goal of assessing the measurement variability for the characterization of a sub-micrometer polydisperse particle dispersion composed of five sub-populations of poly(methyl methacrylate) (PMMA) and silica beads. The study included 20 participating laboratories from industry, academia, and government, and a variety of state-of-the-art particle-counting instruments. The received datasets were organized by instrument class to enable comparison of intralaboratory and interlaboratory performance. The main findings included high variability between datasets from different laboratories, with coefficients of variation from 13 % to 189 %. Intralaboratory variability was, on average, 37 % of the interlaboratory variability for an instrument class and particle sub-population. Drop-offs at either end of the size range and poor agreement on maximum counts of particle sub-populations were noted. The mean distributions from an instrument class, however, showed the size-coverage range for that class. The study shows that a polydisperse sample can be used to assess performance capabilities of an instrument set-up (including hardware, software, and user settings) and provides guidance for the development of polydisperse reference materials.

Nanobubbles in Reconstituted Lyophilized Formulations: Interaction With Proteins and Mechanism of Formation.

Reconstitution of lyophilized disaccharide formulations results in the formation of nanosized air bubbles that persist in suspension for weeks. If proteins are present, interactions with nanobubbles may cause loss of monomeric protein and formation of subvisible particles. The goals of this work are to determine the mechanism(s) by which nanobubbles form in reconstituted lyophilized formulations and to develop strategies for reducing nanobubble generation. We hypothesize that nanobubbles are created from nanosized gas pockets within lyophilized solids, which become bubbles when the surrounding matrix is dissolved away during reconstitution. Nanosized voids may originate from small ice crystals formed within the concentrated liquid during freezing that subsequently sublime during drying. Nanobubble concentrations are correlated with the extent of mannitol crystallization during freezing. Nanosized ice crystals, induced by the release of water during mannitol crystallization, were responsible for nanobubble formation. The presence of trehalose or sucrose, in formulations with low mannitol concentrations, inhibited excipient crystallization during lyophilization and reduced nanobubble levels following reconstitution. Our results show a correlation between nanobubble formation and concentrations of insoluble IL-1ra aggregates, suggesting that minimizing nanobubble generation may be an effective strategy for reducing protein aggregation following reconstitution.

DEHP Nanodroplets Leached From Polyvinyl Chloride IV Bags Promote Aggregation of IVIG and Activate Complement in Human Serum.

Concerns regarding the impact of subvisible particulate impurities on the safety and efficacy of therapeutic protein products have led manufacturers to implement strategies to minimize protein aggregation and particle formation during manufacturing, storage, and shipping. However, once these products are released, manufacturers have limited control over product handling. In this work, we investigated the effect of di(2-ethylhexyl) phthalate (DEHP) nanodroplets generated in polyvinyl chloride (PVC) bags of intravenous (IV) saline on the stability and immunogenicity of IV immunoglobulin (IVIG) formulations. We showed that PVC IV bags containing saline can release DEHP droplets into the solution when agitated or transported using a pneumatic tube transportation system in a clinical setting. We next investigated the effects of emulsified DEHP nanodroplets on IVIG stability and immunogenicity. IVIG adsorbed strongly to DEHP nanodroplets, forming a monolayer. In addition, DEHP nanodroplets accelerated IVIG aggregation in agitated samples. The immunogenicity of DEHP nanodroplets and IVIG aggregates generated in these formulations were evaluated using an in vitro assay of complement activation in human serum. The results suggested DEHP nanodroplets shed from PVC IV bags could reduce protein stability and induce activation of the complement system, potentially contributing to adverse immune responses during the administration of therapeutic proteins.

Particle Formation and Aggregation of a Therapeutic Protein in Nanobubble Suspensions.

The generation of nanobubbles following reconstitution of lyophilized trehalose formulations has recently been reported. Here, we characterize particle formation and aggregation of recombinant human interleukin-1 receptor antagonist (rhIL-1ra) in reconstituted formulations of lyophilized trehalose. Particle characterization methods including resonant mass measurement and nanoparticle tracking analysis were used to count and size particles generated upon reconstitution of lyophilized trehalose formulations. In addition, accelerated degradation studies were conducted to monitor rhIL-1ra aggregation in solutions containing various concentrations of suspended nanobubbles. Reconstitution of lyophilized trehalose formulations with solutions containing rhIL-1ra reduced nanobubble concentrations and generated negatively buoyant particles attributed to aggregated rhIL-1ra. Furthermore, levels of rhIL-1ra aggregation following incubation in aqueous solution correlated with concentrations of suspended nanobubbles. The results of this study suggest that nanobubbles may be a contributor to protein aggregation and particle formation in reconstituted, lyophilized therapeutic protein formulations.

Formation of Stable Nanobubbles on Reconstituting Lyophilized Formulations Containing Trehalose.

During an investigation of subvisible particles found in lyophilized formulations of intravenous immunoglobulin, we used resonant mass measurement techniques and discovered the presence of nanobubbles (NBs) when a 5% trehalose formulation was reconstituted. This discovery prompted studies to characterize these NBs in placebo formulations as a function of processing conditions and solution compositions. Degassing the reconstituted solutions by applying vacuum removed micron-sized bubbles but did not substantially affect the concentration of NBs. Samples that were annealed in the frozen state before lyophilization had reduced surface areas and, on reconstitution, yielded fewer NBs. Trehalose formulations with added polysorbate 20 (PS20) and formulations with higher ionic strength also had smaller numbers of NBs. Zeta potentials of the bubbles were negative in each of the formulations tested, but the negative zeta potentials were decreased in magnitude with increasing ionic strength and with addition of PS20. When incubated at 4(°)C, the number of NBs was largely unchanged at the end of 11 days, whereas the number of micron-sized bubbles gradually decreased during the 11-day incubation. Because of their exceptional stability, NBs are expected to contribute to the numbers of submicron particles that can be detected in reconstituted lyophilized protein formulations.

Crystallographic Investigation of Imidazolium Ionic Liquid Effects on Enzyme Structure.

We present the first crystallographic insight into the interactions of an ionic liquid (IL) with an enzyme, which has widespread implications for stabilizing enzymes in IL media for biocatalysis. Structures of Bacillus subtilis lipase A (lipA) and an IL-stable variant (QM-lipA) were obtained in the presence of increasing concentrations of 1-butyl-3-methylimidazolium chloride ([BMIM][Cl]). These studies revealed that the [BMIM] cation interacts with surface residues through hydrophobic and cation-π interactions. Of specific interest was the disruption of internal stacking interactions of aromatic side chains by [BMIM], which provides structural evidence for the mechanism of enzyme denaturation by ILs. The interaction of [BMIM] and Cl ions with lipA was reduced by the stabilizing mutations Y49E and G158E in QM-lipA. Ultimately, these findings present the molecular basis for stabilizing enzymes from IL-induced inactivation, as well as the selection of ILs that are less denaturing.

Biodiesel transesterification kinetics monitored by pH measurement.

Quantification of a pH change that was observed over the course of the transesterification reaction that converts vegetable oil to biodiesel may provide a simple method to monitor the reaction. Transesterification of canola oil at 6:1 methanol to oil ratio with 0.5 wt.% KOH as catalyst was studied at 25, 35, and 45 °C. Reaction conversion was correlated to pH measurements and the results were shown to be in agreement with an independent measure of conversion using an enzymatic assay for glycerol. Rate constants obtained from these measurements are consistent with those in the literature. The measured pH change appears to be related to dilution of OH(-) ions as the oil is converted to products rather than to depletion of OH(-) due to reaction.