Advertisement

Finding the Needle in the Haystack: High-Resolution Techniques for Characterization of Mixed Protein Particles Containing Shed Silicone Rubber Particles Generated During Pumping

Published:December 08, 2020DOI:https://doi.org/10.1016/j.xphs.2020.12.002

      Abstract

      During the manufacturing process of biopharmaceuticals, peristaltic pumps are employed at different stages for transferring and dosing of the final product. Commonly used silicone tubings are known for particle shedding from the inner tubing surface due to friction in the pump head. These nanometer sized silicone rubber particles could interfere with proteins. Until now, only mixed protein particles containing micrometer-sized contaminations such as silicone oil have been characterized, detected, and quantified. To overcome the detection limits in particle sizes of contaminants, this study aimed for the definite identification of protein particles containing nanometer sized silicone particles in qualitative and quantitative manner. The mixed particles consisted of silicone rubber particles either coated with a protein monolayer or embedded into protein aggregates. Confocal Raman microscopy allows label free chemical identification of components and 3D particle imaging. Labeling the tubing enables high-resolution imaging via confocal laser scanning microscopy and counting of mixed particles via Imaging Flow Cytometry. Overall, these methods allow the detection and identification of particles of unknown origin and composition and could be a forensic tool for solving problems with contaminations during processing of biopharmaceuticals.

      Keywords

      Introduction

      Biopharmaceuticals are experiencing a rapid growth in pharmaceutical industry with a total of 316 active licenses in 2018. Currently, 40% of the 6000 or more products in global clinical development are biopharmaceuticals - mostly proteins - due to their high target specificity.
      • Walsh G.
      Biopharmaceutical benchmarks 2018.
      Because of their complex structure, proteins are sensitive to different kinds of stress like shear, oxidation or interfacial adsorption. This may lead to aggregate formation, which can potentially trigger immune responses.
      • Rosenberg A.S.
      Effects of protein aggregates: an immunologic perspective.
      • Moussa E.M.
      • Panchal Jainik P
      • Moorthy Balakrishnan S
      • et al.
      Immunogenicity of therapeutic protein aggregates.
      • Mahler H.-C.
      • Müller R.
      • Frieß W.
      • Delille A.
      • Matheus S.
      Induction and analysis of aggregates in a liquid IgG1-antibody formulation.
      In addition, regulatory guidelines for particle limitations from the European Pharmacopeia
      European Directorate For The Quality Of Medicine (EDQM)
      2.9.19 Particulate contamination: Sub-visible particles.
      and the USP
      United States Pharmacopeial Convention
      788 Particulate matter in injections.
      need to be fulfilled for product approval. To ensure product safety and quality, particle levels should be kept low.
      • Rosenberg A.S.
      Effects of protein aggregates: an immunologic perspective.
      ,
      • Moussa E.M.
      • Panchal Jainik P
      • Moorthy Balakrishnan S
      • et al.
      Immunogenicity of therapeutic protein aggregates.
      ,
      • Singh S.K.
      • Afonina Nataliya
      • Awwad Michel
      • et al.
      An industry perspective on the monitoring of subvisible particles as a quality attribute for protein therapeutics.
      Particles in biopharmaceutical products can be of protein origin, may derive from excipients
      • Weinbuch D.
      • Ruigrok M.
      • Jiskoot W.
      • Hawe A.
      Nanoparticulate impurities isolated from pharmaceutical-grade sucrose are a potential threat to protein stability.
      or from a foreign source during processing or storage such as rubber from stoppers, glass from vials, fibers from filters, or silicone oil droplets from syringes.
      • Bee J.S.
      • Randolph T.W.
      • Carpenter J.F.
      • Bishop S.M.
      • Dimitrova M.N.
      Effects of surfaces and leachables on the stability of biopharmaceuticals.
      • Gerhardt A.
      • Mcgraw Nicole R.
      • Schwartz Daniel K.
      • Bee Jared S.
      • Carpenter John F.
      • Randolph Theodore W.
      Protein aggregation and particle formation in prefilled glass syringes.
      • Krayukhina E.
      • Tsumoto K.
      • Uchiyama S.
      • Fukui K.
      Effects of syringe material and silicone oil lubrication on the stability of pharmaceutical proteins.
      Interactions between protein molecules and external particles can lead to protein aggregation,
      • Bee J.S.
      • Chiu David
      • Sawicki Suzanne
      • et al.
      Monoclonal antibody interactions with micro- and nanoparticles: adsorption, aggregation, and accelerated stress studies.
      for example shed stainless steel particles from the piston pumps serve as aggregation nuclei via adsorption and structural change of protein molecules at the surface.
      • Tyagi A.K.
      • Randolph Theodore W.
      • Dong Aichun
      • Maloney Kevin M.
      • Hitscherich Carl J.R.
      • Carpenter John F.
      IgG particle formation during filling pump operation: a case study of heterogeneous nucleation on stainless steel nanoparticles.
      ,
      • Bee J.S.
      • Davis M.
      • Freund E.
      • Carpenter J.F.
      • Randolph T.W.
      Aggregation of a monoclonal antibody induced by adsorption to stainless steel.
      Pumps are essential during the downstream and fill & finish processes. The generated protein particles depend on the applied pumping parameters and the type of pump.
      • Meireles M.
      • Aimar P.
      • Sanchez V.
      Albumin denaturation during ultrafiltration: effects of operating conditions and consequences on membrane fouling.
      By now, the most used pumps are rotary piston pumps, time-pressure fillers, and peristaltic pumps. Particularly rotary piston pumps can quickly generate high amounts of protein particles
      • Cromwell M.E.M.
      • Hilario E.
      • Jacobson F.
      Protein aggregation and bioprocessing.
      ,
      • Nayak A.
      • Colandene J.
      • Bradford V.
      • Perkins M.
      Characterization of subvisible particle formation during the filling pump operation of a monoclonal antibody solution.
      by disruption of the protein film built on the stainless steel surface.
      • Kalonia C.K.
      • Heinrich Frank
      • Curtis Joseph E.
      • Raman Sid
      • Miller Maria A.
      • Hudson Steven D.
      Protein adsorption and layer formation at the stainless Steel− solution interface mediates shear-induced particle formation for an IgG1 monoclonal antibody.
      In comparison, peristaltic pumps generate significantly less protein aggregates.
      • Nayak A.
      • Colandene J.
      • Bradford V.
      • Perkins M.
      Characterization of subvisible particle formation during the filling pump operation of a monoclonal antibody solution.
      ,
      • Bausch U.J.
      Impact of Filling Processes on Protein Solutions.
      ,
      • Chandavarkar A.S.
      Dynamics of Fouling of Microporous Membranes by Proteins.
      Often peristaltic pumps are preferred for filling lines because of their complete disposable flow path that helps to avoid cross contamination. Her and Carpenter showed that protein aggregation in peristaltic pumps depends on tubing type and formulation.
      • Her C.
      • Carpenter J.F.
      Effects of tubing type, formulation, and postpumping agitation on nanoparticle and microparticle formation in intravenous immunoglobulin solutions processed with a peristaltic filling pump.
      It is assumed that the aggregation is more likely driven by interfacial stress
      • Chandavarkar A.S.
      Dynamics of Fouling of Microporous Membranes by Proteins.
      ,
      • Denkinger S.N.
      Modelle zur Simulation des Abfüllprozesses biologisch-pharmazeutischer Arzneimittel.
      than by shear stress.
      • Thomas C.R.
      • Geer D.
      Effects of shear on proteins in solution.
      Peristaltic pumps are mostly equipped with silicone tubings due to cost and availability, but these tubings shed particles. If used in medical devices they lead to inflammation and fibrosis due to accumulation in lysosomes of macrophages.
      • Leong A.S.-Y.
      • Disney A.P.
      • Gove D.W.
      Spallation and Migration of silicone from blood-pump tubing in patients of hemodialysis.
      • Bommer J.
      • Gemsa D.
      • Waldherr R.
      • Kessler J.
      • Ritz E.
      Plastic filing from dialysis tubing induces prostanoid release from macrophages.
      • Laohapand T.
      • Osman E.M.
      • Morley A.R.
      • Ward M.K.
      • Kerr D.N.S.
      Accumulation of silicone elastomer in regular dialysis.
      Saller
      • Saller V.
      • Matilainen Julia
      • Grauschopf Ulla
      • Bechtold-Peters Karoline
      • Mahler Hanns-Christian
      • Friess Wolfgang
      Particle shedding from peristaltic pump tubing in biopharmaceutical drug product manufacturing.
      et al. showed that different amounts of nanometer-sized silicone rubber particles abrade from the inner tubing wall depending on surface roughness. These particles can enter the final product and potentially interact with proteins leading to immune responses in patients.
      While intensive studies have already focused on the silicone oil composite protein particles and their immunogenicity,
      • Van Beers M. M. c.
      • Gilli F.
      • Schellekens H.
      • Randolph T.W.
      • Jiskoot W.
      Immunogenicity of recombinant human interferon beta interacting with particles of glass, metal, and polystyrene.
      • Chisholm C.F.
      • Baker Abby E.
      • Soucie Kaitlin R.
      • Torres Raul M.
      • Carpenter John F.
      • Randolph Theodore W.
      Silicone oil microdroplets can induce antibody responses against recombinant murine growth hormone in mice.
      • Chisholm C.F.
      • Soucie Kaitlin R.
      • Song Jane S.
      • et al.
      Immunogenicity of structurally perturbed hen egg lysozyme adsorbed to silicone oil microdroplets in wild-type and transgenic mouse models.
      silicone rubber also has a potential for triggering immune responses. An early study of Anderson et al., reports on protein adsorption to silicone rubber followed by macrophage activation and leading to cytokine expression.
      • Anderson J.M.
      • Ziats Nicholas P.
      • Azeez Ahmad
      • Brunstedt Michael R.
      • Stack Steven
      • Bonfield Tracey L.
      Protein adsorption and macrophage activation on polydimethylsiloxane and silicone rubber.
      For the differentiation between protein particles and contaminations, a variety of methods have been established for example silicone oil droplets can be distinguished from protein particles by flow imaging, resonant mass measurements
      • Gambe-Gilbuena A.
      • Shibano Y.
      • Krayukhina E.
      • Torisu T.
      • Uchiyama S.
      Automatic identification of the stress sources of protein aggregates using flow imaging microscopy images.
      • Weinbuch D.
      • Zölls S.
      • Wiggenhorn M.
      • et al.
      Micro-flow imaging and resonant mass measurement (Archimedes)-Complementary methods to quantitatively differentiate protein particles and silicone oil droplets.
      • Strehl R.
      • Rombach-Riegraf V.
      • Diez M.
      • et al.
      Discrimination between silicone oil droplets and protein aggregates in biopharmaceuticals: a novel multiparametric image filter for sub-visible particles in microflow imaging analysis.
      or flow cytometry.
      • Ludwig D.B.
      • Trotter J.T.
      • Gabrielson J.P.
      • Carpenter J.F.
      • Randolph T.W.
      Flow cytometry: a promising technique for the study of silicone oil-induced particulate formation in protein formulations.
      However, contaminations in protein particles can be detected only in the micrometer size range.
      • Ludwig D.B.
      • Trotter J.T.
      • Gabrielson J.P.
      • Carpenter J.F.
      • Randolph T.W.
      Flow cytometry: a promising technique for the study of silicone oil-induced particulate formation in protein formulations.
      ,
      • Probst C.
      Characterization of protein aggregates, silicone oil droplets, and protein-silicone interactions using imaging flow cytometry.
      Identifying the formation of heterogeneous particles upon peristaltic pumping is associated with two main burdens. First, the nanometer size of the contaminant makes its identification by light microscopy-based techniques such as micro flow imaging impossible. Second, the rubber particles do not show spectral features like auto fluorescence or deep color, which could enable identification. Methods combining optical information with chemical identification at high resolution may enable to overcome this problem. In the past, Raman microscopy
      • Lankers M.
      • Munhall J.
      • Valet O.
      Differentiation between foreign particulate matter and silicone oil induced protein aggregation in drug solutions by automated Raman spectroscopy.
      and Imaging Flow Cytometry (IFC)
      • Probst C.
      Characterization of protein aggregates, silicone oil droplets, and protein-silicone interactions using imaging flow cytometry.
      showed great potential to resolve the mixed particles of micrometer sized silicone oil droplets and micrometer sized protein particles. While Raman microscopy is able to analyze native particles in a label free way through the spectral differences of the two species, fluorescence labeling of at least one component after particle formation allows counting by IFC and high-resolution images by confocal laser scanning microscopy (CLSM). CLSM can provide sharp images by collecting the emitted photons from the sample’s fluorophore utilizing a spatial pinhole which reduces out of focus light. Refining these methods could allow analysis of nanosized contaminants in larger protein aggregates.
      In this study, we demonstrate the detection and chemical resolution of heterogeneous particles consisting of aggregated protein and nanometer sized silicone rubber particles. Protein particles generated through peristaltic pumping through two silicone tubings of different surface characteristics and shedding propensity were characterized by turbidity, flow imaging, and quantitative laser diffraction (qLD). Raman microscopy, CLSM, and imaging flow cytometry methods were developed to identify and classify mixed particles and to resolve the distribution of the small non-proteinaceous contaminants within the larger protein particles. Adsorption of mAb to silicone particles was evaluated to gain insight into the mixed protein particle formation. The newly developed methods possess great potential for solving scientific problems on micro- and nanometer-based contaminations of protein aggregates to ensure quality and safety of biopharmaceuticals.

      Materials and Methods

       Materials

      33.4 mg/mL monoclonal IgG1 antibody (Isoelectric point: 8.0-8.3) in 20 mM histidine buffer pH 5.4 served as model monoclonal antibody (mAb). Buffers were prepared using highly purified water (HPW) from an Arium pro DI Ultrapure Water System (Sartorius Stedim Biotech GmbH, Goettingen, Germany). Histidine (Applichem, Darmstadt, Germany) or potassium dihydrogen phosphate (Grüssing GmbH, Filsum, Germany) were dissolved and pH was adjusted either with hydrochloric acid (VWR, Darmstadt, Germany) or potassium hydroxide (Bernd Kraft GmbH, Duisburg, Germany). Buffers were filtered through 0.2 μm cellulose acetate filters (47 mm ø, Sartorius Stedim Biotech GmbH) and mAb formulations through 0.2 μm polyethersulfone membrane syringe filters (VWR).
      The employed Pt-cured silicone tubings Accusil (Watson-Marlow, Falmouth, United Kingdom) and Versilic (Saint-Gobain, Taunton, MA, USA) had an inner diameter of 6.0 mm and a wall thickness of 2.1 mm. Tubing sets were assembled by connecting two 20 cm long pieces for the area in the pump head to 35 cm long tubing pieces via polypropylene Y-connectors (Kartell, VWR). To mimic production conditions tubing sets were rinsed with 5 L HPW at 80 °C and steam sterilized (121 °C, 15 min, 2 bar).
      Silicone rubber microparticles KMP-594 (average particle size of 4.6 μm) were kindly donated by Shin-Etsu Chemical Co. Ltd, Tokyo, Japan. Chemicals were obtained as follows: [[(3,5-dimethyl-1H-pyrrol-2-yl) (3,5-dimethyl-2H-pyrrol-2-ylidene)methyl]methane] (difluoroborane) (Bodipy) from ABCR (Karlsruhe, Germany), chloroform and ethanol from VWR; polysorbate 20 (PS20) from Merck (Darmstadt, Germany), and sodium azide from Acros Organics (Geel, Belgium).

       Fluorescent Labeling of Tubing and Silicone Rubber Particles

      Autoclaved tubing was filled with either 1 mg/mL Bodipy in EtOH for IFC or 2 mg/mL Bodipy in 20:80 (v/v) chloroform:EtOH for CLSM. After 24 h incubation, solvent residues were removed using a vacuum oven (Memmert, Büchenbach, Germany) for 3 h at 300 mbar and for 21 h at 10 mbar at 25 °C. For IFC measurements, 100 mg silicone rubber microparticles were suspended in 1 mL of 1mg/ml Bodipy solution in EtOH and processed as described above. After drying, the particles were resuspended in histidine buffer and centrifuged once at 12,800 g for 10 min for washing. The labeled silicone beads were collected and resuspended in 20 mM histidine pH 5.4 with 0.1% PS20 at a concentration of 4 mg/mL.

       Sample Preparation

      All pumping experiments were conducted in a laminar flow hood to avoid external particle contamination. Tubing sets were pre-rinsed by pumping 20 L HPW with a Flexicon PD12 peristaltic pump (Watson-Marlow Flexicon, Ringsted, Denmark) in continuous mode at 180 rpm. These settings resulted in a flow rate of 23 mL/s at an occlusion pressure of approximately 1.3 bar upon operation in air (manometer from WIKA Alexander Wiegand SE & Co. KG, Klingenberg, Germany, accuracy class 2.5). For sample preparation, 45 mL of formulation buffer followed by 45 mL of 1 mg/mL antibody solution were circulated 20 times at 180 rpm. Protein concentration was verified by UV absorption at 280 nm using a Nanodrop Micro-Volume UV-Vis spectrometer (Nano Drop 2000, Thermo Scientific, Wilmington, USA) and an extinction coefficient of 1.51 cm2/mg. As a reference for silicone free protein particles, 1 mL of a 1 mg/mL antibody solution was shaken for 2 min in a CapMix device (3M Espe, Neuss, Germany) or stirred for 1 h at 300 rpm with a Variomag Poly 15 (H + P LABORTECHNIK, Oberschleißheim, Germany). As a bacteriostatic agent, 0.01% sodium azide was added to the samples for the IFC measurements.

       3D Laser Scanning Microscopy

      Surface roughness of the inner tubing walls was visualized using a 3D laser scanning microscope (3D-LSM) Keyence VK-X200 equipped with a CF Plan ELWD 50x objective (Keyence GmbH, Neu-Isenburg, Germany). Micrographs of small tubing pieces were captured with the VK Viewer software in ‘Expert Mode’ at the standard settings. Two pictures were stitched and three representative regions of 100 × 100 μm2 were used for surface roughness calculations using the MultiFileAnalyzer version 1.3.1.120. Tubing curvature was corrected for via the correct tilt–sec curved surface function and artifacts were eliminated by a medium height cut level.

       Scanning Electron Microscopy (SEM)

      Silicone particles fixed on an aluminum sample grid were carbon sputtered under vacuum and analyzed by a Helios NanoLab G3 UC scanning electron microscope (FEI, Hillsboro, Oregon, USA) at 2.0 kV and a working distance of approximately 10.5 mm.

       Specific Surface Area

      The specific surface area of the silicone particles was determined by an Autosorb 1 (Quantachrome, Odelzhausen, Germany) equipped with a liquid nitrogen bath at -196 °C using krypton gas adsorption and Brunauer-Emmet-Teller (BET) isotherm analysis. Approximately 1.5 g were outgassed for at least 2 h at 150 °C. An 11-point gas adsorption curve was measured (p/p0 ratio of approximately 0.05-0.3). A multipoint Brunauer-Emmet-Teller fit was performed with the Autosorb 1 software.

       Turbidity

      A sample volume of 1.8 mL was analyzed using a Nephla turbidimeter (Dr. Lange, Duesseldorf, Germany). Data is presented in formazine nephelometric units (FNU).

       Detection of Subvisible Particles

      For an estimation of the particle size distribution and the total particle amount, the samples were analyzed by quantitative laser diffraction (qLD) using the Aggregates Sizer with WingSALD bio software version 3.2.2 (Shimadzu Corporation, Kyoto, Japan) in a batch cell with 5 mL sample volume. The calculations were based on a material specific refractive index of 1.46, an imaginary index of 0.1, and a protein particle density of 1.32 g/cm3.
      • Folzer E.
      • Khan Tarik A.
      • Schmidt Roland
      • et al.
      Determination of the density of protein particles using a suspended microchannel resonator.
      The cut-off level of noise was set to 500.
      Samples were also analyzed with a FlowCAM® 8100 (Fluid Imaging Technologies, Inc., Scarborough, ME, USA) equipped with a 10x magnification cell (81 μm × 700 μm). A sample volume of 150 μL was used and images were collected at a flow rate of 0.15 mL/min, an auto image frame rate of 28 frames/s and a sampling time of 60 s, which resulted in an efficiency value higher than 70%. Particles were identified using the following settings: 3 μm distance to the nearest neighbor, particle segmentation thresholds of 13 and 10 for the dark and light pixels, respectively. Particle size was reported as the equivalent spherical diameter using the VisualSpreadsheet® 4.7.6 software.

       Raman Microscopy

      3 μL samples were transferred to a self-constructed microfluidic channel with a width of 50 μm and a depth of 160 μm. Vibrational spectra were collected by a WITec alpha 300R microscope (WITec GmbH, Ulm, Germany), equipped with a 532 nm DPL laser (Cobolt AB, Solna, Sweden). The laser was focused through a 63× oil immersion objective (Zeiss Plan-Apochromat SF25 63x/1.4; Carl Zeiss AG, Germany) into the static liquid. The true-power module ensured stable laser output, which was set between 40 and 50 mW. Large area map configurations were performed by collecting single point Raman spectra with a pixel size of 0.25 – 0.5 μm (x/y axis) and 1 μm (x/z axis) using a high-performance CCD camera with a spectral resolution of 4 cm-1. Dependent on particle morphology and map configuration, exposure times were set between 0.2 s and 4 s to minimize sample damage.
      Raman spectral maps were processed by spike removal, baseline correction (penalized spline), and normalized to the area under the curve. Hyperspectral data was unmixed via vertex component analysis (VCA)
      • Nascimento J.M.P.
      • Dias J.M.B.
      Vertex component analysis: a fast algorithm to unmix hyperspectral data.
      using the software package Imagelab (Epina GmbH, Austria). A resampling factor of three was used and the spectral range above 3041 cm-1 was excluded to speed up calculation time. Concentration maps of the three main components (endmembers) show the lateral distribution of each component within the investigated area.

       Confocal Laser Scanning Microscopy

      Samples were fixed on glass slides by mixing 100 μL of sample with one drop of FluorSave™ Reagent (Merck Millipore Calbiochem, Darmstadt, Germany). Confocal images and z-stacks were acquired using a Leica TCS SP8 SMD microscope (Leica Microsystems, Wetzlar, Germany) equipped with a Leica PL APO 63x/1.20 water immersion objective at the 496 nm laser line for excitation. Images were analyzed using ImageJ. Images of pumped protein particles were composed z-stacks for better solution.

       Imaging Flow Cytometry

      Samples were run either in native form, stained with ProteoStat, or with ProteoStat and Bodipy (Amnis® Protein Aggregate & Silicone Oil Detection Kit, Luminex Corp, Seattle, US). Fluorescent labeling was performed by mixing 2 μL 50x dye cocktail containing ProteoStat or ProteoStat and Bodipy in 1x Assay Buffer with 98 μL sample. Final concentrations for ProteoStat and Bodipy were 0.75 μmol/L and 94 nmol/L, respectively. After at least 10 min incubation, samples were analyzed with an Amnis FlowSight® imaging flow cytometer (Luminex, Austin, USA) equipped with a 20× magnification objective allowing 1 μm pixel resolution and a 60 μm wide field-of-view. The assay was based on the method developed by Probst.
      • Probst C.
      Characterization of protein aggregates, silicone oil droplets, and protein-silicone interactions using imaging flow cytometry.
      The 785 nm side scatter (SSC) excitation laser was set to 70 mW and the 488 nm fluorescence excitation laser to 60 mW. BF in Ch01 (457/45, center wavelength/band width) and Ch09 (582/25) intensity is set automatically by the instrument software to achieve consistent background. Images were collected as followed: SSC in Ch06 (emission 762/35); Bodipy in Ch02 (528/65); ProteoStat in Ch04 (610/30); Bodipy precipitates in Ch05 (702/85). All events detected were recorded in high sensitivity mode without user threshold for 120 s, equal to 2.52 μL sample volume. All data was analyzed using IDEAS® 6.2 (Luminex, Seattle, USA) image analysis software.

       Adsorption of Protein to Silicone Particles

      Identification of interfacial induced protein aggregation and quantification of adsorbed protein on silicone particles was evaluated by mixing 400 mg silicone particles with 1.5 mL protein stock solution (c = 0.5; 1.0; 5.0 mg/mL) and rotating up to 48 h (SU1100, Sunlab Sustainable Lab Instruments, Mannheim, Germany) at 25 rpm at room temperature. Samples were centrifuged at 12,800 g for 10 min and analyzed by SEC using a TSKgel G3000SWXL column (Tosoh Bioscience GmbH, Stuttgart, Germany) and an Agilent 1100 series HPLC system equipped with a UV/Vis detector operated at 210 nm (Agilent Technologies, Santa Clara, CA, USA). The mobile phase consisted of 150 mM potassium phosphate buffer pH 6.5 at a flow rate of 0.5 mL/min at room temperature.

       Adsorption of Silicone Particles to Protein Aggregates

      To detect spontaneous association of free silicone particles with protein aggregates, a 2 mg/mL silicone microparticle suspension in histidine buffer was sonicated for 10 min. The suspension was mixed 1:1 with a stirred 2 mg/mL mAb solution diluted 1:50. This mixture was analyzed by flow imaging. For CLSM, stirred protein samples were mixed 1:1 with Bodipy stained silicone microparticles or silicone particles generated by buffer pumping through Bodipy stained Versilic tubing for 7 h. Mixed samples were incubated for 1.5 h before analysis.

       Statistics

      Results are reported as mean values with standard deviation (SD) of three independent experiments. Data was analyzed using GraphPad Prism (Version 5.02 for Microsoft Windows, Graph Pad Software, San Diego, USA).

      Results

       Surface Characterization of Tubing Material

      The inner silicone tubing surface was characterized using 3D-LSM (Fig. 1), as the surface roughness is an indicator for the degree of particle shedding during pumping.
      • Saller V.
      • Matilainen Julia
      • Grauschopf Ulla
      • Bechtold-Peters Karoline
      • Mahler Hanns-Christian
      • Friess Wolfgang
      Particle shedding from peristaltic pump tubing in biopharmaceutical drug product manufacturing.
      Surface roughness is described by the arithmetical mean height of the surface (SA) and the maximum height (SZ). While fresh Accusil tubing features a smooth surface (SA = 0.25 ± 0.03 μm; SZ = 2.93 ± 0.42 μm), Versilic exhibits a markedly higher surface roughness (SA = 3.00 ± 0.19 μm; SZ = 23.08 ± 2.57 μm) with a wavelike surface structure.
      Figure thumbnail gr1
      Figure 13D-LSM images (a) and surface roughness (b) of the inner wall of Accusil and Versilic tubing.

       Particle Formation Upon Pumping

      Subsequently, protein solutions and buffer as control were pumped and analyzed for particle formation with three different techniques. Turbidity served as a qualitative indicator for particle formation, qLD to count in the nanometer to micrometer range and flow imaging to count and study the morphology of micrometer size particles (Fig. 2). All pumped solutions were characterized by a substantial increase in particle numbers compared to the controls. Turbidity and particle numbers after pumping buffer were comparable for both silicone tubings with around 0.6 FNU and 4000 particles ≥1 μm/mL. The signal intensity in qLD was below the cut-off level of noise for pumped buffer. The degree of particle formation was affected by the tubing material. Pumped mAb samples contained significantly higher particle numbers than the pumped buffer. Accusil tubing material resulted in lower particle levels than Versilic tubing material, with 1.8 ± 0.2 versus 5.6 ± 0.1 FNU and 81,562 ± 5441 versus 450,709 ± 8301 particles ≥1 μm/mL and 8.5·106 ± 0.6·106 versus 5.1·107 ± 0.2·107 particles ≥0.3 μm/mL.
      Figure thumbnail gr2
      Figure 2Particle analysis by turbidity (a), flow imaging (b) and qLD (c) of pumped buffer and mAb after 20 cycles through Accusil and Versilic tubing. Representative images from flow imaging of pumped mAb (d).
      Morphological evaluation (Fig. 2d) was only possible for particles larger than 10 μm, as smaller particles appear blurry. The number concentration of the particles per ml larger than 10 μm increased drastically from less than 20 per ml to 200 per ml in buffer for both tubing materials, and 2565 ± 233 and 18,816 ± 244 per ml in mAb solution for Accusil and Versilic, respectively. Particles exhibit a structure like ripped off protein film fragments folded up in the medium. Such type of particle was not present in the pumped buffer controls. In the following, samples pumped in Versilic tubing were used for a detailed microscopy-based analysis due to the higher particle concentration, whereas samples pumped in Accusil tubing were used for IFC to quantify mixed particles in optimum industry settings.

       Component Analysis Using Raman Microscopy

      Fig. 3 shows the chemical image of a representative protein particle with a size of approximately 12 × 12 μm.2 The surrounding of the particle within the region of interest (ROI) appears clear without any visible contaminants. 1600 spectra were acquired in a ROI of 20 x 20 μm.2
      Figure thumbnail gr3
      Figure 3Calculated Raman spectra (left) and concentration maps (right) of the three detected endmembers: Raman spectral band intensities are allocated to the background (glass and water) (a), silicone nanoparticles (b) and protein (c). Scale bar represents 5 μm.
      The different components glass, water, protein and silicone are called endmembers. The three endmembers were extracted from the Raman data via vertex component analysis (Table 1). The concentration map of the first endmember represents the surrounding of the particle (background). The estimated Raman spectra of the first endmember shows a strong Raman band at 1100 cm-1 which is attributed to glass. The Raman bands at 1640 cm-1, and 3400 cm-1 are assigned to bending and stretching vibrations of water. The Raman spectrum of the second endmember indicates an intense silicone signal besides of a weaker protein signal; hence, this component represents a mixture of silicone rubber and protein. The Raman fingerprint of silicone occurs at four distinct Raman wavenumbers which are all present in the extracted spectrum of the second component: the distinct band at 490 cm-1 evoked by ν(Si-O-Si) vibrations,
      • Österle W.
      • Giovannozzi A.
      • Gradt T.
      • et al.
      Exploring the potential of Raman spectroscopy for the identification of silicone oil residue and wear scar characterization for the assessment of tribofilm functionality.
      the Raman band at 710 cm-1 attributed to (Si-CH3) symmetric stretch vibrations, and the strong Raman bands at the wavenumbers at 2900 cm-1 and 2970 cm-1 representing C-H stretching vibrations.
      • Howell N.K.
      • Arteaga G.
      • Nakai S.
      • Li-Chan E.C.Y.
      Raman spectral analysis in the C-H stretching region of proteins and amino acids for investigation of hydrophobic interactions.
      The extracted Raman spectrum of the second endmember shows also protein specific Raman bands such as the phenylalanine band at 1000 cm-1 and the protein intrinsic amide I band at 1670 cm-1.
      • Lippert J.L.
      • Tyminski D.
      • Desmeules P.J.
      Determination of the secondary structure of proteins by laser Raman spectroscopy.
      ,
      • Tfayli A.
      • Piot O.
      • Draux F.
      • Pitre F.
      • Manfait M.
      Molecular characterization of reconstructed skin model by Raman microspectroscopy: comparison with excised human skin.
      The concentration map of the second endmember shows that silicone-characteristic Raman bands occur at four different spots of the scanned area. The concentration map of the third endmember represents the microscopically visualized particle. The extracted Raman spectrum of the third endmember shows exclusively protein specific Raman bands. The Raman band at 1000 cm-1 is assigned to phenylalanine.
      • Lippert J.L.
      • Tyminski D.
      • Desmeules P.J.
      Determination of the secondary structure of proteins by laser Raman spectroscopy.
      The amide III band is apparent at 1240 cm-1, whereas the tryptophan-characteristic bands are visible at 760 cm-1 and 1345 cm-1.
      • Tobin M.C.
      Raman spectra of crystalline lysozyme, pepsin, and alpha chymotrypsin.
      ,
      • Dupaix A.
      • Bechet Jean-Jacques
      • Yon Jeannine
      • Merlin Jean-Claude
      • Delhayet Michel
      • Hill Max
      Resonance Raman spectroscopic studies of the interactions between trypsin and a competitive inhibitor (ligand-enzyme interactions/serine protease/vibrational spectroscopy).
      The C-H deformation and stretching modes are represented at 1450 cm-1 and 2940 cm-1, respectively.
      • Nakamura K.
      • Era Seiichi
      • Ozaki Yukihiro
      • Sogami Masaru
      • Hayashi Tomoya
      • Murakami Masataka
      Conformational changes in seventeen cystine disulfide bridges of bovine serum albumin proved by Raman spectroscopy.
      ,
      • Howell N.K.
      • Arteaga G.
      • Nakai S.
      • Li-Chan E.C.Y.
      Raman spectral analysis in the C-H stretching region of proteins and amino acids for investigation of hydrophobic interactions.
      The Raman band at 1670 cm-1 is assigned to the amide I vibration of the protein.
      • Tfayli A.
      • Piot O.
      • Draux F.
      • Pitre F.
      • Manfait M.
      Molecular characterization of reconstructed skin model by Raman microspectroscopy: comparison with excised human skin.
      Table 1Raman Band Assignment of Each Component.
      ComponentAssignmentWavenumber [cm-1]Ref.
      Backgroundglass1100
      • White W.B.
      • Minser D.G.
      Raman spectra and structure of natural glasses.
      δ(O-H) water1640
      • Tominaga Y.
      • Fujiwara A.
      • Amo Y.
      Dynamical structure of water by Raman spectroscopy.
      ν(O-H) water3400
      • Cárcamo J.J.
      • Aliaga A.E.
      • Clavijo R.E.
      • Brañes M.R.
      • Campos-Vallette M.M.
      Raman study of the shockwave effect on collagens.
      Tryptophan760
      • Takeuchi H.
      Raman structural markers of tryptophan and histidine side chains in proteins.
      Phenylalanine1000
      • Lippert J.L.
      • Tyminski D.
      • Desmeules P.J.
      Determination of the secondary structure of proteins by laser Raman spectroscopy.
      Amid II1240
      • Barron L.D.
      • Wen Z.Q.
      • Hecht L.
      Vibrational Raman optical activity of proteins.
      ProteinCα−H Tryptophan1345
      • Tobin M.C.
      Raman spectra of crystalline lysozyme, pepsin, and alpha chymotrypsin.
      δ(C-H)1450
      • Nakamura K.
      • Era Seiichi
      • Ozaki Yukihiro
      • Sogami Masaru
      • Hayashi Tomoya
      • Murakami Masataka
      Conformational changes in seventeen cystine disulfide bridges of bovine serum albumin proved by Raman spectroscopy.
      Amid I1670
      • Tfayli A.
      • Piot O.
      • Draux F.
      • Pitre F.
      • Manfait M.
      Molecular characterization of reconstructed skin model by Raman microspectroscopy: comparison with excised human skin.
      ν(C-H)2940
      • Howell N.K.
      • Arteaga G.
      • Nakai S.
      • Li-Chan E.C.Y.
      Raman spectral analysis in the C-H stretching region of proteins and amino acids for investigation of hydrophobic interactions.
      ν(Si-O-Si)490
      • Österle W.
      • Giovannozzi A.
      • Gradt T.
      • et al.
      Exploring the potential of Raman spectroscopy for the identification of silicone oil residue and wear scar characterization for the assessment of tribofilm functionality.
      Silicone rubberν(Si-CH3)710
      • Österle W.
      • Giovannozzi A.
      • Gradt T.
      • et al.
      Exploring the potential of Raman spectroscopy for the identification of silicone oil residue and wear scar characterization for the assessment of tribofilm functionality.
      C-H2900
      • Howell N.K.
      • Arteaga G.
      • Nakai S.
      • Li-Chan E.C.Y.
      Raman spectral analysis in the C-H stretching region of proteins and amino acids for investigation of hydrophobic interactions.
      C-H2970
      • Howell N.K.
      • Arteaga G.
      • Nakai S.
      • Li-Chan E.C.Y.
      Raman spectral analysis in the C-H stretching region of proteins and amino acids for investigation of hydrophobic interactions.
      ν stretching vibration δ bending vibration.
      Table 2Method Development for IFC Measurements.
      Bodipy Post-Staining ApproachBodipy Pre-Staining Approach
      Labeling of components:
      SiliconePost-Staining with BodipyPre-Staining Tubing with Bodipy
      Protein AggregatesPost-Staining with ProteoStatPost-Staining with ProteoStat
      Detection of mixed particlesNot possiblea) Silicone containing protein particles:

      Sample without ProteoStat staining

      b) Total amount of protein aggregates: Sample labeled with ProteoStat
      The unmixing of the hyperspectral data through VCA extracts endmember spectra. The according concentration maps visualize the spatial distribution of each component within the ROI, for example the distribution of silicone nanoparticles. In this example, the silicone particles appear to be located at the edge of the protein structure rather than within the protein particle.
      More Raman images of protein structures were recorded to investigate the position of the silicone nanoparticles relative to the protein (Fig. 4a). The overlay of protein and silicone concentration maps reveals a random distribution of silicone particles (at the edge and in the middle of the protein structure). However, the 2-dimensional image does not reveal whether silicone particles are located on top, bottom or in the middle of the protein. Therefore, depth scans (z-stacks) were performed visualizing the spatial distribution of silicone nanoparticles in three dimensions (Fig. 4b).
      Figure thumbnail gr4
      Figure 4Microscopic images of particles in brightfield and calculated concentration maps (a) and z-scan (z-difference: 1 μm) through a single particle (b). Silicone majorities (red) were present around and inside the protein core (green). The scale bar corresponds to 5 μm.
      Chemical depth profiles are achieved by recording Raman images at different heights/sample layers yielding a more comprehensive analysis of the silicone particle location. The depth profiles of the protein particles indicate that silicone was not only located on the protein surface but also inside the particle.

       Fluorescence-Based Particle Identification Using CLSM

      Complementary to the observations by Raman microscopy, fluorescence-based methods were employed to identify and quantify the mixed particle species. The tubing material was stained before its use by incubation in an organic Bodipy solution. CLSM enables the analysis of the particle morphology and the distribution of the silicone nanoparticles inside the mixed particle. Only a single staining was possible in CLSM due to the quick photobleaching of ProteoStat (Fig. 5).
      Figure thumbnail gr5
      Figure 5Representative brightfield and fluorescence micrographs of protein particles generated by stirring stained with 1 μmol Bodipy and generated during pumping through Bodipy stained Versilic tubing.
      Protein particles independent of stress condition exhibited an irregular shape. Pure protein particles did not feature any intrinsic fluorescence, while staining these particles with 1 μmol Bodipy resulted in a diffuse fluorescence of the whole particle. In contrast, the pumped samples contain protein particles with several intense Bodipy positive signals attributed to silicone. Silicone particles varied in size, ranging from a few nanometers up to 2 μm.

       Quantification of Mixed Species Using IFC

      While microscopic techniques are limited in the quantification of particles due to low sample throughput, we employed IFC to overcome these limitations. For sample generation, protein was pumped through Bodipy labeled tubing and aggregates were subsequently stained with ProteoStat to detect protein particles. Data was presented in bivariate plots of Bodipy (Ch02) versus ProteoStat (Ch04) fluorescence intensity. Gates were drawn to classify the particles as either protein aggregates or silicone particles. Addition of ProteoStat to the buffer did not result in the formation of fluorescent particles for example by precipitation.
      For the evaluation of the method, single particle type controls were run (Fig. S1) and demonstrated sufficient resolution of both signals. Labeled and unlabeled silicone beads showed a false positive signal in the protein aggregation gate. Such unspecific fluorescence is known for surfactant concentrations higher than 0.01%
      PROTEOSTAT ® protein aggregation assay for microplates or flow cytometry instruction manual.
      due to interaction of ProteoStat with hydrophobic regions of the PS20 micelles.
      • McClure S.M.
      • Ahl P.L.
      • Blue J.T.
      High throughput differential scanning fluorimetry (DSF) formulation screening with complementary dyes to assess protein unfolding and aggregation in presence of surfactants.
      Post-staining with Bodipy results in a Bodipy positive population for micrometer sized silicone beads and absence of Bodipy fluorescence in pure protein samples (Fig. S2). As free Bodipy could bind to protein aggregates pure protein particles could be detected false positive as mixed particles. Dye loading and thus fluorescence intensity of shed tubing particles could be adapted by the choice of the organic solvent which affects swelling and diffusion of dye into the tubing wall. Therefore, prestaining was considered as the more reliable preparation method for nanoparticle identification (Table 2).
      No fluorescent particles were detected in pumped buffer samples, while pumped protein samples showed signals in the protein aggregate population only after ProteoStat staining. Analyzing the protein samples unstained, a silicone particle population could be uncovered. ProteoStat seems to dominate the signals of silicone particles due to its high intensity (Fig. 6). For the detection of the mixed species, it is therefore necessary to evaluate the sample before and after staining with ProteoStat.
      Figure thumbnail gr6
      Figure 6Bivariate Plots for buffer (a), 1 mg/mL mAb pumped through untreated (b) and Bodipy pre-stained Accusil tubing (c) before (left) and after staining with ProteoStat (right).
      Samples were analyzed without additional silicone post staining for silicone nanoparticle containing protein particles and with additional ProteoStat post-staining to analyze the total number of protein particles (Fig. 7). In mAb solutions pumped through naive and Bodipy labeled tubing, 407,968 ± 21,068 and 290,016 ± 20,610 protein particles per ml were detected, respectively. Due to instrument noise and spectral spillover the measurements of buffer and mAb solution pumped through unlabeled tubing the instrument falsely indicated the presence of a few Bodipy stained particles. The mAb solution showed approximately 131,974 ± 7,279 Bodipy positive particles per ml after pumping through Bodipy labeled tubing indicating that 45% of the protein particles contained silicone particles.
      Figure thumbnail gr7
      Figure 7Number of detected mixed particles in buffer and 1 mg/mL mAb pumped through unstained and Bodipy stained Accusil tubing.

       Protein Adsorption to Silicone Particles

      Shed silicone particles could not be produced in a sufficient large scale to allow reliable statistical analysis. Therefore, spherical silicone model microparticles with a specific surface area of 0.388 ± 0.002 m2/g were utilized for adsorption studies (Fig. 8a). MAb molecules adsorbed quickly to the silicone particles (Fig. 8 b + c) reaching an equilibrium of approximately 1.5 mg/m2 after 24 h. Increasing mAb concentration only lead to a minor increase to 1.8 mg/m2. Induction of protein oligomers could not be found in the bulk solution.
      Figure thumbnail gr8
      Figure 8SEM images of model silicone particles (a). MAb adsorption to silicone particles over time (c = 0.5 mg/mL) (b) and as function of protein concentration after 24 h (c).
      In CLSM, no mixed particles were detected after incubation of preformed protein aggregates with Bodipy labeled silicone model microparticles (Fig. S3a). Furthermore, no decrease in total particle number concentration, which would indicate assembly of both components, could be found for incubated mixtures (Fig S3b).

      Discussion

      Peristaltic pumps are routinely used in different stages of manufacturing and handling of biopharmaceuticals. Most manufacturing stages contain filtration steps for particle removal although there are some cases where filtration caused protein aggregation due to interfacial adsorption to the filter membrane.
      • Maa Y.F.
      • Hsu C.C.
      Investigation on fouling mechanisms for recombinant human growth hormone sterile filtration.
      ,
      • Bódalo A.
      • Gómez J.L.
      • Gómez E.
      • Máximo M.F.
      • Montiel M.C.
      Study of L-aminoacylase deactivation in an ultrafiltration membrane reactor.
      Fill & finish operations lack final purification steps except filtration. Especially during filling, pumps could be the source of particle burden. Previous studies highlighted that upon peristaltic pumping, tubing material sheds
      • Saller V.
      • Matilainen Julia
      • Grauschopf Ulla
      • Bechtold-Peters Karoline
      • Mahler Hanns-Christian
      • Friess Wolfgang
      Particle shedding from peristaltic pump tubing in biopharmaceutical drug product manufacturing.
      and protein aggregates can be formed.
      • Her C.
      • Carpenter J.F.
      Effects of tubing type, formulation, and postpumping agitation on nanoparticle and microparticle formation in intravenous immunoglobulin solutions processed with a peristaltic filling pump.
      Hence, potential formation of particles and aggregates is critical, as different size species may enter the final product in an unobstructed way.
      We wanted to learn whether particles composed of both aggregated protein and shed silicone can form. Such mixed structures are suspected to induce immune responses.
      • Van Beers M. M. c.
      • Gilli F.
      • Schellekens H.
      • Randolph T.W.
      • Jiskoot W.
      Immunogenicity of recombinant human interferon beta interacting with particles of glass, metal, and polystyrene.
      • Chisholm C.F.
      • Baker Abby E.
      • Soucie Kaitlin R.
      • Torres Raul M.
      • Carpenter John F.
      • Randolph Theodore W.
      Silicone oil microdroplets can induce antibody responses against recombinant murine growth hormone in mice.
      • Chisholm C.F.
      • Soucie Kaitlin R.
      • Song Jane S.
      • et al.
      Immunogenicity of structurally perturbed hen egg lysozyme adsorbed to silicone oil microdroplets in wild-type and transgenic mouse models.
      At first, methods are needed to detect mixed protein species which are suitable for process development and root cause analysis. We therefore pumped a mAb solution using silicone tubing. Two silicone tubings with different surface roughness were tested to have a look on the effect of higher silicone particle spallation on protein aggregate formation. Independent of tubing material, pump roller-induced generation of particles above 1 μm were observed at a particle density less than 200 particles/ml per pump cycle. According to Saller et al., the shed silicone particles are typically around 180 nm in size. They found that the amount of shed particles is linked to the surface roughness of the inner tubing wall, which accords to our findings.
      • Saller V.
      • Matilainen Julia
      • Grauschopf Ulla
      • Bechtold-Peters Karoline
      • Mahler Hanns-Christian
      • Friess Wolfgang
      Particle shedding from peristaltic pump tubing in biopharmaceutical drug product manufacturing.
      The tubing with higher roughness rendered increased levels of protein particles larger than 1 μm per pump cycle with approximately 23,000 compared to 4000 particles per ml. This observation could be linked either to the difference in silicone shedding, or to the increasing surface area with increasing roughness prone to interfacial effects. With higher surface roughness on the nano- and micrometer scale, protein adsorption is enhanced beyond the corresponding increase in specific surface area,
      • Akkas T.
      • Citak C.
      • Sirkecioglu A.
      • Güner F.S.
      Which is more effective for protein adsorption: surface roughness , surface wettability or swelling? Case study of polyurethane films prepared from castor oil and poly (ethylene glycol).
      • Rechendorff K.
      • Hovgaard M.B.
      • Foss M.
      • Zhdanov V.P.
      • Besenbacher F.
      Enhancement of protein adsorption induced by surface roughness.
      • Deligianni D.D.
      • Katsala N.
      • Ladas S.
      • Sotiropoulou D.
      • Amedee J.
      • Missirlis Y.F.
      Effect of surface roughness of the titanium alloy Ti-6Al-4V on human bone marrow cell response and on protein adsorption.
      which again may lead to increased protein aggregation and particle formation. It is assumed that protein particle formation is based on the detachment of the protein film from the tubing surface by roller movements
      • Her C.
      • Carpenter J.F.
      Effects of tubing type, formulation, and postpumping agitation on nanoparticle and microparticle formation in intravenous immunoglobulin solutions processed with a peristaltic filling pump.
      similar to effects known from piston pumps.
      • Kalonia C.K.
      • Heinrich Frank
      • Curtis Joseph E.
      • Raman Sid
      • Miller Maria A.
      • Hudson Steven D.
      Protein adsorption and layer formation at the stainless Steel− solution interface mediates shear-induced particle formation for an IgG1 monoclonal antibody.
      We speculate that, based on their morphology, the protein particles resemble fragmented protein film.
      Multicomponent species can be discerned if components differ in specific properties such as morphology or density. The standard application is microflow imaging to tell between silicone oil droplets and protein particles, addressing a combination of different optical properties. Micrometer particles with nanometer-size contaminations cannot be further analyzed by this technique. Here, high resolution techniques with the ability for chemical identification are needed. In this study, Raman microscopy and fluorescence microscopic imaging were employed to close this gap. The mechanism of mixed silicone and protein particle formation is still unclear. Mixed particles could form independently and attach to each other, or silicone particles could be nuclei for the formation of protein aggregates. Questions on the formation mechanism of the mixed protein particles cannot be answered with these methods but they can provide fundamental information on the composition of each detected particle and insight into contaminant distribution in protein particles.
      Raman spectroscopy offers a variety of applications for pharmaceutical industry in early drug development, characterization of drug delivery systems, in-process measurements, product performance and quality control for the final product.
      • Paudel A.
      • Raijada D.
      • Rantanen J.
      Raman spectroscopy in pharmaceutical product design.
      In brightfield microscopy, all particles resemble homogenous protein aggregates, without a hint of embedded or connected foreign species. By scanning the protein particles with Raman microscopy, the chemical distribution of the whole particle dimension was visualized. In a high number of particles, an additional chemical component besides protein could be identified. Heterogeneous particles showed spots of high Raman scattering at 490 cm-1, 710 cm-1, 2900 cm-1 and 2970 cm-1, which is linked to silicone rubber shed from the tubing material during pumping.
      • Jayes L.
      • Hard A.P.
      • Séné C.
      • Parker S.F.
      • Jayasooriya U.A.
      Vibrational spectroscopic analysis of Silicones:  A fourier transform-Raman and inelastic neutron scattering investigation.
      Heterogeneous particles were detected by Raman microscopy without sample staining. The unmixing of the hyperspectral data set via VCA provided a fast and reliable evaluation tool, which visualized the spatial distribution of silicone nanoparticles on the surface of a protein particle. The presented Raman method is applicable for a vast majority of particles with minor limitations for highly fluorescent or weakly Raman active species. Particles detected in brightfield mode exhibited a sufficiently strong Raman signal at moderate laser power without particle destruction, degradation or photobleaching. However, recording the Raman maps is a time-consuming task (roughly 2 h/particle), hence, the current method suffers from poor sample throughput.
      Complementary to Raman microscopy, mixed particles can be analyzed by fluorescence-based methods. This study aimed to differentiate protein and silicone using two non-covalently binding extrinsic fluorescence dyes that have selective affinity for each species with well-separated signals and require no purification step. Based on a previous study the combination of Bodipy and ProteoStat fits these requirements.
      • Probst C.
      Characterization of protein aggregates, silicone oil droplets, and protein-silicone interactions using imaging flow cytometry.
      Silicone itself has no autofluorescence and needs labeling for detection. Bodipy is a highly fluorescent hydrophobic dye in non-polar and polar media with sharp and narrow emission peak at 503 nm. The tubing was stained by incubation with Bodipy and the dye migration enhanced by swelling of the tubing through organic solvents.
      • Saller V.
      • Matilainen Julia
      • Rothkopf Christian
      • et al.
      Preservative loss from silicone tubing during filling processes.
      The Bodipy loading of the silicone matrix can also be tuned by the swelling properties of the organic solvent. This approach is limited to hydrophobic components and it must be assured that the pump properties of the tubing are not changed by residual solvent. Extrinsic labeling of the protein is necessary as the intrinsic fluorescence of proteins is not strong enough for detection. Covalent labeling of the protein a priori may change polarity, charge and ultimately surface binding, protein-protein interactions and protein aggregation.
      • Cockrell G.M.
      • Wolfe M.S.
      • Wolfe J.L.
      • Schöneich C.
      Photoinduced aggregation of a model antibody-drug conjugate.
      ,
      • Teske C.A.
      • von Lieres Eric
      • Schröder Magnus
      • Ladiwala Asif
      • Cramer Steven M.
      • Hubbuch Jürgen J.
      Competitive adsorption of labeled and native protein in confocal laser scanning microscopy.
      Furthermore, covalent dye binding is not relevant when it comes to screening of a drug product. The protein staining dye ProteoStat was added to the samples prior to analysis. ProteoStat as a proprietary fluorescent rotor dye exhibits increased fluorescence when interacting with protein aggregate structures over a wide range of excipients including low to medium surfactant concentrations.
      • McClure S.M.
      • Ahl P.L.
      • Blue J.T.
      High throughput differential scanning fluorimetry (DSF) formulation screening with complementary dyes to assess protein unfolding and aggregation in presence of surfactants.
      We could not find a reference to the use of CLSM for the characterization of protein aggregates in biopharmaceuticals. Unfortunately, ProteoStat dye was quickly bleached whereas Bodipy showed good photochemical stability. Protein particles from stirred samples containing no silicone particles exhibited diffuse staining of the whole protein particle in presence of free Bodipy due to interactions with the hydrophobic cavities of protein aggregates.
      • Marfin Y.S.
      • Aleksakhina E.L.
      • Merkushev D.A.
      • Rumyantsev E.V.
      • Tomilova I.K.
      Interaction of BODIPY dyes with the blood plasma proteins.
      ,
      • Dorh N.
      • Zhu Shilei
      • Dhungana Kamal B.
      • et al.
      BODIPY-based fluorescent probes for sensing protein surface-hydrophobicity.
      In contrast, particles generated by pumping showed finely distributed silicone rubber remains attached to the protein particle. The sharp images taken in a few seconds with high resolution allowed morphological analysis and identification of silicone.
      CSLM and Raman microscopy do not allow a fast quantification of protein particle concentration. Therefore, we set up an IFC method which could quantify both, protein-only and mixed protein silicone particles. IFC allows simultaneous acquisition of multiple spectral images in high quality and sensitivity for fast-moving objects by a charge-coupled device detector with a time-delay integration technology. Although post-staining of silicone with Bodipy was possible with micrometer-sized silicone beads, no mixed particles could be detected in the pumped samples using the post-staining approach. Additionally, post-staining with Bodipy comes with the risk of unspecific protein staining and interactions with the protein formulation. For micrometer-sized contaminants like silicone oil, post-staining with Bodipy may be less critical, as the fluorescence intensity would exceed any interferences from protein signal.
      • Probst C.
      Characterization of protein aggregates, silicone oil droplets, and protein-silicone interactions using imaging flow cytometry.
      Consequently, silicone tubing was pre-labelled via incubation with organic solvents containing a higher Bodipy concentration of 2 mg/mL. Even with pre-labelling, the low fluorescence signal of the silicone nanoparticles was overwhelmed by the ProteoStat dye staining the protein. To distinguish mixed species, the pumped samples were analyzed with and without ProteoStat staining, such that mixed particles were identified as Bodipy positive in the ProteoStat-unlabeled samples, and protein-only particles were identified as ProteoStat-positive events subtracted by the mixed particles. Free 180 nm
      • Saller V.
      • Matilainen Julia
      • Grauschopf Ulla
      • Bechtold-Peters Karoline
      • Mahler Hanns-Christian
      • Friess Wolfgang
      Particle shedding from peristaltic pump tubing in biopharmaceutical drug product manufacturing.
      silicone rubber particles were not detectable as their size is below the SSC detection limit of around 200 nm, however silicone could be detected when attached to protein aggregates, presumably as multiple silicone particles were attached to individual protein aggregates.
      • Probst C.
      Characterization of protein aggregates, silicone oil droplets, and protein-silicone interactions using imaging flow cytometry.
      Protein particles larger than 2 μm were shown to contain at least 45% mixed particles for the pumped protein sample. To resolve smaller protein particles containing small amounts of silicone rubber particles, higher laser power and magnification could be beneficial. This method can be used for a wide range of formulations, but care should be taken when using surfactant concentrations >0.01%. Interactions between the ProteoStat dye and the surfactant micelles can lead to increased background intensity resulting in a higher risk for false positive protein aggregates signal.
      With the herein developed methods, it is now possible to distinguish and quantify protein particles containing foreign micro- and nanometer-sized particles. Raman microscopy, CLSM, and IFC together can provide important insights, as we demonstrate here for protein particles formed upon peristaltic pumping in silicone tubes. Due to the limitations in microscopic resolution and the low fluorescence intensity of the nanometer-sized particles, these approaches are limited to micrometer-sized protein particles.
      Most foreign micro - and nanoparticle contaminations are coated by a protein monolayer and do not necessarily lead to protein particle formation.
      • Bee J.S.
      • Chiu David
      • Sawicki Suzanne
      • et al.
      Monoclonal antibody interactions with micro- and nanoparticles: adsorption, aggregation, and accelerated stress studies.
      But there are also materials like stainless steel particles, which induce conformational changes upon protein adsorption and lead to aggregation.
      • Tyagi A.K.
      • Randolph Theodore W.
      • Dong Aichun
      • Maloney Kevin M.
      • Hitscherich Carl J.R.
      • Carpenter John F.
      IgG particle formation during filling pump operation: a case study of heterogeneous nucleation on stainless steel nanoparticles.
      ,
      • Bee J.S.
      • Davis M.
      • Freund E.
      • Carpenter J.F.
      • Randolph T.W.
      Aggregation of a monoclonal antibody induced by adsorption to stainless steel.
      Spiking silicone particles did not induce an increase in protein aggregates in the course of a stability study,
      • Saller V.
      • Hediger C.
      • Matilainen J.
      • et al.
      Influence of particle shedding from silicone tubing on antibody stability.
      but the presence of surfactants in the tested formulations could have suppressed interfacial interactions. The chosen silicone rubber microparticles are formed from crosslinked linear dimethylpolysiloxane and resemble high purity silicone rubber. Density and surface charge should be very similar to the pt-cured tubing material, but the materials are not completely identical in material properties. We showed that approximately 1.8 mg/m2 mAb adsorb to silicone particles, which is consistent with a monolayer formation.
      • Bee J.S.
      • Chiu David
      • Sawicki Suzanne
      • et al.
      Monoclonal antibody interactions with micro- and nanoparticles: adsorption, aggregation, and accelerated stress studies.
      There were no signs for formation or preferred adsorption of higher molecular weight species in SEC. Furthermore, the experiments were conducted in the absence of interfacial shear. These findings support the hypothesis that elevated protein particle formation in tubing with higher surface roughness is linked to an increase in adsorption area rather than to increased silicone rubber particle shedding. The adsorption is driven through an interplay of hydrophobic and electrostatic interactions between the positively charge mAb and the negatively charged silicone rubber
      • Saller V.
      • Matilainen Julia
      • Grauschopf Ulla
      • Bechtold-Peters Karoline
      • Mahler Hanns-Christian
      • Friess Wolfgang
      Particle shedding from peristaltic pump tubing in biopharmaceutical drug product manufacturing.
      leading to an irreversibly bound monolayer.
      • Marsh R.J.
      • Jones R.A.L.
      • Sferrazza M.
      Adsorption and displacement of a globular protein on hydrophilic and hydrophobic surfaces.
      Interactions between preformed aggregates and foreign species may be protein dependent, for example lysozyme aggregates strongly attached on silicone oil droplets, mAb aggregates showed less interaction with silicone oil.
      • Probst C.
      Characterization of protein aggregates, silicone oil droplets, and protein-silicone interactions using imaging flow cytometry.
      In case of our mAb, we did not find any evidence for subsequent adsorption of the silicone particles to protein aggregates. Silicone rubber particles could therefore become part of the aggregates if there is simultaneous disruption of protein film and detachment of underlying silicone rubber during roller movement.

      Conclusion

      Peristaltic pumping generates mAb particles containing nanometer sized silicone rubber fragments, which can be identified by Raman microscopy, CLSM and IFC. The silicone rubber particles do not serve as source of aggregation but are covered by protein and distributed in the total large protein aggregates. Raman microscopy as label-free method and CLSM using Bodipy stained tubing material could proof the presence of silicone fragments in protein particles. IFC could reveal that nearly half of the protein particles contained silicone rubber. The study indicated that mixed silicone rubber and protein particle are co-generated during pumping instead of subsequent absorption of both species. The developed methods possess great potential for chemical identification of protein aggregates contaminated with non-proteinaceous material to support process development and forensic root cause analysis.

      Acknowledgments

      Thanks to Christine Probst from Luminex Corporation (Seattle, USA) for performing the IFC analyses and the scientific input on fluorescence-based analysis of mixed particles. Special thanks go to Haley Pugsley and María Gracia García Mendoza from Luminex Corporation (Seattle, USA) for performing the repeat measurements. Thanks to Coriolis Pharma for providing access to the Aggregate Sizer. David Bauer is kindly acknowledged for his scientific input in the analysis of Raman data.

      Appendix A. Supplementary Data

      References

        • Walsh G.
        Biopharmaceutical benchmarks 2018.
        Nat Biotechnol. 2018; 36: 1136-1145
        • Rosenberg A.S.
        Effects of protein aggregates: an immunologic perspective.
        AAPS J. 2006; 8: E501-E507
        • Moussa E.M.
        • Panchal Jainik P
        • Moorthy Balakrishnan S
        • et al.
        Immunogenicity of therapeutic protein aggregates.
        J Pharm Sci. 2016; 105: 417-430
        • Mahler H.-C.
        • Müller R.
        • Frieß W.
        • Delille A.
        • Matheus S.
        Induction and analysis of aggregates in a liquid IgG1-antibody formulation.
        Eur J Pharm Biopharm. 2005; 59: 407-417
        • European Directorate For The Quality Of Medicine (EDQM)
        2.9.19 Particulate contamination: Sub-visible particles.
        in: European Pharmacopoeia. 6th. Council of Europe, Strasbourg2008
        • United States Pharmacopeial Convention
        788 Particulate matter in injections.
        in: United States Pharmacopeia. United States Pharmacopeial Convention, United States2014
        • Singh S.K.
        • Afonina Nataliya
        • Awwad Michel
        • et al.
        An industry perspective on the monitoring of subvisible particles as a quality attribute for protein therapeutics.
        J Pharm Sci. 2010; 99: 3302-3321
        • Weinbuch D.
        • Ruigrok M.
        • Jiskoot W.
        • Hawe A.
        Nanoparticulate impurities isolated from pharmaceutical-grade sucrose are a potential threat to protein stability.
        Pharm Res. 2017; 34: 2910-2921
        • Bee J.S.
        • Randolph T.W.
        • Carpenter J.F.
        • Bishop S.M.
        • Dimitrova M.N.
        Effects of surfaces and leachables on the stability of biopharmaceuticals.
        J Pharm Sci. 2011; 100: 4158-4170
        • Gerhardt A.
        • Mcgraw Nicole R.
        • Schwartz Daniel K.
        • Bee Jared S.
        • Carpenter John F.
        • Randolph Theodore W.
        Protein aggregation and particle formation in prefilled glass syringes.
        J Pharm Sci. 2014; 103: 1601-1612
        • Krayukhina E.
        • Tsumoto K.
        • Uchiyama S.
        • Fukui K.
        Effects of syringe material and silicone oil lubrication on the stability of pharmaceutical proteins.
        J Pharm Sci. 2015; 104: 527-535
        • Bee J.S.
        • Chiu David
        • Sawicki Suzanne
        • et al.
        Monoclonal antibody interactions with micro- and nanoparticles: adsorption, aggregation, and accelerated stress studies.
        J Pharm Sci. 2009; 98: 3218-3238
        • Tyagi A.K.
        • Randolph Theodore W.
        • Dong Aichun
        • Maloney Kevin M.
        • Hitscherich Carl J.R.
        • Carpenter John F.
        IgG particle formation during filling pump operation: a case study of heterogeneous nucleation on stainless steel nanoparticles.
        J Pharm Sci. 2009; 98: 94-104
        • Bee J.S.
        • Davis M.
        • Freund E.
        • Carpenter J.F.
        • Randolph T.W.
        Aggregation of a monoclonal antibody induced by adsorption to stainless steel.
        Biotechnol Bioeng. 2010; 105: 121-129
        • Meireles M.
        • Aimar P.
        • Sanchez V.
        Albumin denaturation during ultrafiltration: effects of operating conditions and consequences on membrane fouling.
        Biotechnol Bioeng. 1991; 38: 528-534
        • Cromwell M.E.M.
        • Hilario E.
        • Jacobson F.
        Protein aggregation and bioprocessing.
        AAPS J. 2006; 8: E572-E579
        • Nayak A.
        • Colandene J.
        • Bradford V.
        • Perkins M.
        Characterization of subvisible particle formation during the filling pump operation of a monoclonal antibody solution.
        J Pharm Sci. 2011; 100: 4198-4204
        • Kalonia C.K.
        • Heinrich Frank
        • Curtis Joseph E.
        • Raman Sid
        • Miller Maria A.
        • Hudson Steven D.
        Protein adsorption and layer formation at the stainless Steel− solution interface mediates shear-induced particle formation for an IgG1 monoclonal antibody.
        Mol Pharm. 2018; 15: 1319-1331
        • Bausch U.J.
        Impact of Filling Processes on Protein Solutions.
        University Basel, Switzerland2008
        • Chandavarkar A.S.
        Dynamics of Fouling of Microporous Membranes by Proteins.
        Massachusetts Institute of Technology, United States1990
        • Her C.
        • Carpenter J.F.
        Effects of tubing type, formulation, and postpumping agitation on nanoparticle and microparticle formation in intravenous immunoglobulin solutions processed with a peristaltic filling pump.
        J Pharm Sci. 2019; : 1-11
        • Denkinger S.N.
        Modelle zur Simulation des Abfüllprozesses biologisch-pharmazeutischer Arzneimittel.
        Rheinische Friedrich-Wilhelms-Universität Bonn, Germany2010
        • Thomas C.R.
        • Geer D.
        Effects of shear on proteins in solution.
        Biotechnol Lett. 2011; 33: 443-456
        • Leong A.S.-Y.
        • Disney A.P.
        • Gove D.W.
        Spallation and Migration of silicone from blood-pump tubing in patients of hemodialysis.
        N Engl J Med. 1982; 306: 135-140
        • Bommer J.
        • Gemsa D.
        • Waldherr R.
        • Kessler J.
        • Ritz E.
        Plastic filing from dialysis tubing induces prostanoid release from macrophages.
        Kidney Int. 1984; 26
        • Laohapand T.
        • Osman E.M.
        • Morley A.R.
        • Ward M.K.
        • Kerr D.N.S.
        Accumulation of silicone elastomer in regular dialysis.
        Proc EDTA. 1982; 19: 143-152
        • Saller V.
        • Matilainen Julia
        • Grauschopf Ulla
        • Bechtold-Peters Karoline
        • Mahler Hanns-Christian
        • Friess Wolfgang
        Particle shedding from peristaltic pump tubing in biopharmaceutical drug product manufacturing.
        J Pharm Sci. 2015; 104: 1440-1450
        • Van Beers M. M. c.
        • Gilli F.
        • Schellekens H.
        • Randolph T.W.
        • Jiskoot W.
        Immunogenicity of recombinant human interferon beta interacting with particles of glass, metal, and polystyrene.
        J Pharm Sci Sci. 2012; 101: 187-199
        • Chisholm C.F.
        • Baker Abby E.
        • Soucie Kaitlin R.
        • Torres Raul M.
        • Carpenter John F.
        • Randolph Theodore W.
        Silicone oil microdroplets can induce antibody responses against recombinant murine growth hormone in mice.
        J Pharm Sci. 2016; 105: 1623-1632
        • Chisholm C.F.
        • Soucie Kaitlin R.
        • Song Jane S.
        • et al.
        Immunogenicity of structurally perturbed hen egg lysozyme adsorbed to silicone oil microdroplets in wild-type and transgenic mouse models.
        J Pharm Sci. 2017; 106: 1519-1527
        • Anderson J.M.
        • Ziats Nicholas P.
        • Azeez Ahmad
        • Brunstedt Michael R.
        • Stack Steven
        • Bonfield Tracey L.
        Protein adsorption and macrophage activation on polydimethylsiloxane and silicone rubber.
        J Biomater Sci. 1996; 7: 159-169
        • Gambe-Gilbuena A.
        • Shibano Y.
        • Krayukhina E.
        • Torisu T.
        • Uchiyama S.
        Automatic identification of the stress sources of protein aggregates using flow imaging microscopy images.
        J Pharm Sci. 2020; 109: 614-623
        • Weinbuch D.
        • Zölls S.
        • Wiggenhorn M.
        • et al.
        Micro-flow imaging and resonant mass measurement (Archimedes)-Complementary methods to quantitatively differentiate protein particles and silicone oil droplets.
        J Pharm Sci. 2013; 102: 2152-2165
        • Strehl R.
        • Rombach-Riegraf V.
        • Diez M.
        • et al.
        Discrimination between silicone oil droplets and protein aggregates in biopharmaceuticals: a novel multiparametric image filter for sub-visible particles in microflow imaging analysis.
        Pharm Res. 2012; 29: 594-602
        • Ludwig D.B.
        • Trotter J.T.
        • Gabrielson J.P.
        • Carpenter J.F.
        • Randolph T.W.
        Flow cytometry: a promising technique for the study of silicone oil-induced particulate formation in protein formulations.
        Anal Biochem. 2011; 410: 191-199
        • Probst C.
        Characterization of protein aggregates, silicone oil droplets, and protein-silicone interactions using imaging flow cytometry.
        J Pharm Sci. 2020; 109: 364-374
        • Lankers M.
        • Munhall J.
        • Valet O.
        Differentiation between foreign particulate matter and silicone oil induced protein aggregation in drug solutions by automated Raman spectroscopy.
        Microsc Microanal. 2008; 14: 1612-1613
        • Folzer E.
        • Khan Tarik A.
        • Schmidt Roland
        • et al.
        Determination of the density of protein particles using a suspended microchannel resonator.
        Artic J Pharm Sci. 2015; 104: 4034-4040
        • Nascimento J.M.P.
        • Dias J.M.B.
        Vertex component analysis: a fast algorithm to unmix hyperspectral data.
        IEEE Trans Geosci Remote Sens. 2005; 43: 898-910
        • White W.B.
        • Minser D.G.
        Raman spectra and structure of natural glasses.
        J Non Cryst Sol. 1984; 67: 45-59
        • Tominaga Y.
        • Fujiwara A.
        • Amo Y.
        Dynamical structure of water by Raman spectroscopy.
        Fluid Phase Equilib. 1998; 144: 323-330
        • Cárcamo J.J.
        • Aliaga A.E.
        • Clavijo R.E.
        • Brañes M.R.
        • Campos-Vallette M.M.
        Raman study of the shockwave effect on collagens.
        Spectrochim Acta A Mol Biomol Spectrosc. 2012; 86: 360-365
        • Takeuchi H.
        Raman structural markers of tryptophan and histidine side chains in proteins.
        Biopolymers (Biospectroscopy). 2003; 72
        • Lippert J.L.
        • Tyminski D.
        • Desmeules P.J.
        Determination of the secondary structure of proteins by laser Raman spectroscopy.
        J Am Chem Soc. 1976; 98: 7075-7080
        • Barron L.D.
        • Wen Z.Q.
        • Hecht L.
        Vibrational Raman optical activity of proteins.
        J Am Chem Soc. 1992; : 784-786
        • Tobin M.C.
        Raman spectra of crystalline lysozyme, pepsin, and alpha chymotrypsin.
        Science. 1968; 161: 68-79
        • Nakamura K.
        • Era Seiichi
        • Ozaki Yukihiro
        • Sogami Masaru
        • Hayashi Tomoya
        • Murakami Masataka
        Conformational changes in seventeen cystine disulfide bridges of bovine serum albumin proved by Raman spectroscopy.
        FEBS Lett. 1997; 417: 375-378
        • Tfayli A.
        • Piot O.
        • Draux F.
        • Pitre F.
        • Manfait M.
        Molecular characterization of reconstructed skin model by Raman microspectroscopy: comparison with excised human skin.
        Biopolymers. 2007; 87: 261-274
        • Howell N.K.
        • Arteaga G.
        • Nakai S.
        • Li-Chan E.C.Y.
        Raman spectral analysis in the C-H stretching region of proteins and amino acids for investigation of hydrophobic interactions.
        J Agric Food Chem. 1999; : 924-933
        • Österle W.
        • Giovannozzi A.
        • Gradt T.
        • et al.
        Exploring the potential of Raman spectroscopy for the identification of silicone oil residue and wear scar characterization for the assessment of tribofilm functionality.
        Tribol Int. 2015; : 481-490
        • Dupaix A.
        • Bechet Jean-Jacques
        • Yon Jeannine
        • Merlin Jean-Claude
        • Delhayet Michel
        • Hill Max
        Resonance Raman spectroscopic studies of the interactions between trypsin and a competitive inhibitor (ligand-enzyme interactions/serine protease/vibrational spectroscopy).
        Proc Natl Acad Sci U S A. 1975; 72: 4223-4227
      1. PROTEOSTAT ® protein aggregation assay for microplates or flow cytometry instruction manual.
        (Available at:)
        www.enzolifesciences.com
        Date: 2018
        Date accessed: March 23, 2020
        • McClure S.M.
        • Ahl P.L.
        • Blue J.T.
        High throughput differential scanning fluorimetry (DSF) formulation screening with complementary dyes to assess protein unfolding and aggregation in presence of surfactants.
        Pharm Res. 2018; 35: 81
        • Maa Y.F.
        • Hsu C.C.
        Investigation on fouling mechanisms for recombinant human growth hormone sterile filtration.
        J Pharm Sci. 1998; 87: 808-812
        • Bódalo A.
        • Gómez J.L.
        • Gómez E.
        • Máximo M.F.
        • Montiel M.C.
        Study of L-aminoacylase deactivation in an ultrafiltration membrane reactor.
        Enzyme Microb Technol. 2004; 35: 261-266
        • Akkas T.
        • Citak C.
        • Sirkecioglu A.
        • Güner F.S.
        Which is more effective for protein adsorption: surface roughness , surface wettability or swelling? Case study of polyurethane films prepared from castor oil and poly (ethylene glycol).
        Polym Int. 2013; 62: 1202-1209
        • Rechendorff K.
        • Hovgaard M.B.
        • Foss M.
        • Zhdanov V.P.
        • Besenbacher F.
        Enhancement of protein adsorption induced by surface roughness.
        Langmuir. 2006; 22: 10885-10888
        • Deligianni D.D.
        • Katsala N.
        • Ladas S.
        • Sotiropoulou D.
        • Amedee J.
        • Missirlis Y.F.
        Effect of surface roughness of the titanium alloy Ti-6Al-4V on human bone marrow cell response and on protein adsorption.
        Biomaterials. 2001; 22
        • Paudel A.
        • Raijada D.
        • Rantanen J.
        Raman spectroscopy in pharmaceutical product design.
        Adv Drug Deliv Rev. 2015; 89: 3-20
        • Jayes L.
        • Hard A.P.
        • Séné C.
        • Parker S.F.
        • Jayasooriya U.A.
        Vibrational spectroscopic analysis of Silicones:  A fourier transform-Raman and inelastic neutron scattering investigation.
        Anal Chem. 2003; 75: 742-746
        • Saller V.
        • Matilainen Julia
        • Rothkopf Christian
        • et al.
        Preservative loss from silicone tubing during filling processes.
        Eur J Pharm Biopharm. 2017; 112: 109-118
        • Cockrell G.M.
        • Wolfe M.S.
        • Wolfe J.L.
        • Schöneich C.
        Photoinduced aggregation of a model antibody-drug conjugate.
        Mol Pharm. 2015; 12: 1784-1797
        • Teske C.A.
        • von Lieres Eric
        • Schröder Magnus
        • Ladiwala Asif
        • Cramer Steven M.
        • Hubbuch Jürgen J.
        Competitive adsorption of labeled and native protein in confocal laser scanning microscopy.
        Biotechnol Bioeng. 2006; 95: 58-66
        • Marfin Y.S.
        • Aleksakhina E.L.
        • Merkushev D.A.
        • Rumyantsev E.V.
        • Tomilova I.K.
        Interaction of BODIPY dyes with the blood plasma proteins.
        J Fluoresc. 2016; 26: 255-261
        • Dorh N.
        • Zhu Shilei
        • Dhungana Kamal B.
        • et al.
        BODIPY-based fluorescent probes for sensing protein surface-hydrophobicity.
        Nat Publ Gr. 2015; 5: 18337
        • Saller V.
        • Hediger C.
        • Matilainen J.
        • et al.
        Influence of particle shedding from silicone tubing on antibody stability.
        J Pharm Pharmacol. 2018; 70: 675-685
        • Marsh R.J.
        • Jones R.A.L.
        • Sferrazza M.
        Adsorption and displacement of a globular protein on hydrophilic and hydrophobic surfaces.
        Colloids Surf B Biointerfaces. 2002; 23: 31-42