do noit drop - cavitation in vials

RESEARCH ARTICLE Pharmaceutical Biotechnology

Do Not Drop: Mechanical Shock in Vials Causes Cavitation, ProteinAggregation, and Particle Formation

THEODORE W. RANDOLPH,1 ELISE SCHILTZ,1 DONN SEDERSTROM,2 DANIEL STEINMANN,3 OLIVIER MOZZICONACCI,3

CHRISTIAN SCHONEICH,3 ERWIN FREUND,4 MARGARET S. RICCI,4 JOHN F. CARPENTER,5 CORRINE S. LENGSFELD2

1Department of Chemical and Biological Engineering, University of Colorado Boulder, Boulder, Colorado2Department of Mechanical and Materials Engineering, University of Denver, Denver, Colorado3Department of Pharmaceutical Chemistry, University of Kansas, Lawrence, Kansas4Process Development, Amgen, Inc., Thousand Oaks, California5Department of Pharmaceutical Sciences, University of Colorado Denver, Aurora, Colorado

Received 22 August 2014; revised 22 September 2014; accepted 9 October 2014

Published online 21 November 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.24259

ABSTRACT: Industry experience suggests that g-forces sustained when vials containing protein formulations are accidentally droppedcan cause aggregation and particle formation. To study this phenomenon, a shock tower was used to apply controlled g-forces toglass vials containing formulations of two monoclonal antibodies and recombinant human growth hormone (rhGH). High-speed videoanalysis showed cavitation bubbles forming within 30 s and subsequently collapsing in the formulations. As a result of echoing shockwaves, bubbles collapsed and reappeared periodically over a millisecond time course. Fluid mechanics simulations showed low-pressureregions within the fluid where cavitation would be favored. A hydroxyphenylfluorescein assay determined that cavitation producedhydroxyl radicals. When mechanical shock was applied to vials containing protein formulations, gelatinous particles appeared on the vialwalls. Size-exclusion chromatographic analysis of the formulations after shock did not detect changes in monomer or soluble aggregateconcentrations. However, subvisible particle counts determined by microflow image analysis increased. The mass of protein attached tothe vial walls increased with increasing drop height. Both protein in bulk solution and protein that became attached to the vial walls aftershock were analyzed by mass spectrometry. rhGH recovered from the vial walls in some samples revealed oxidation of Met and/or Trpresidues. C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci 104:602611, 2015Keywords: stability; simulations; protein formulation; protein aggregation; oxidation; particle size

INTRODUCTION

To ensure drug product quality, it is important to retain thechemical and conformational purity of proteins during storage,shipping, and delivery to patients. Of the many chemical degra-dation paths and physical instabilities to which proteins maybe subjected, aggregation is one of the most prevalent.1 Thepresence of aggregated proteins in certain cases has been as-sociated with reduced efficacy, induction of unwanted immuneresponses, and anaphylactic shock.2,3

One largely unrecognized cause of protein damage is me-chanical shock to the primary drug product container. Vialsand syringes that are dropped by patients or caregivers mayexperience severe mechanical shock, but unless the containershatters, the contents are typically administered without fur-ther examination. When dropped primary containers hit a solidsurface, the solutions inside experience mechanical shock, andthe resulting shock waves and associated fluid mechanics maycause cavitation. Cavitation is the formation, growth, and col-lapse of cavities in a liquid.46 Cavities formed by mechanicalforces imposed by the high g-forces that arise from the me-chanical shock of dropping are unstable, and rapidly implode.

Correspondence to: Theodore W. Randolph (Telephone: +303-492-4776;Fax: +303-492-4341; E-mail: [email protected])

This article contains supplementarymaterial available from the authors uponrequest or via the Internet at http://wileylibrary.com.

Journal of Pharmaceutical Sciences, Vol. 104, 602611 (2015)C 2014 Wiley Periodicals, Inc. and the American Pharmacists Association

When the cavities violently collapse, hot spots are created.48

These extremely short-lived and localized hot spots are lo-cal regions of extremely high temperatures and pressure thatmay reach thousands of Kelvin and hundreds of atmospheres,respectively.48 During the collapse of a cavity, hydrogen andhydroxyl radicals can also be formed.4,911

As a result of the high temperatures, high pressures, andfree radical formation that occur during cavitation, proteinmolecules can be damaged.10 For example, cavitation result-ing from therapeutic ultrasound has been shown to causestructural changes in several different proteins including cy-tochrome c, lysozyme, bovine serum albumin, trypsinogen,RNAse, and "-chymotrypsinogen.1214 Furthermore, hydroxylradicals produced during cavitation have been shown to cross-link proteins and cause formation of protein particles.15,16

There have been observations of shipping- and handling-induced aggregation and particle formation in formulationstherapeutic proteins contained in their final drug product vialsand prefilled syringes.17,18 Furthermore, we hypothesize that ifa patient or caregiver were to drop a vial onto a hard surface,undetected cavitation could occur, fostering formation of proteinaggregates and proteinaceous particles. To simulate shock thatcould occur as a result of shipping and/or accidental dropping,a mechanical drop tester was used to apply controlled mechani-cal shock to glass vials containing formulations of two differentmonoclonal antibodies or recombinant human growth hormone(rhGH). High-speed video photography was used to record theformation and collapse of cavitation bubbles and fluid flows over

602 Randolph et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:602611, 2015

RESEARCH ARTICLE Pharmaceutical Biotechnology 603

submillisecond timescales. The formation of free radicals thatformed during cavitation was confirmed by fluorescence spec-troscopy using a hydroxyphenylfluorescein assay. Damage tothe proteins was quantified bymeasuring aggregate levels withsize-exclusion chromatography, and subvisible particle countsmade by microflow imaging. The protein adhering to the vialwalls following shock treatment was also quantified. Chemicaldegradation of the proteins was assessed by high-performanceliquid chromatography and mass spectrometry.

MATERIALS AND METHODS

Model Proteins

Two model proteins were used for the majority of the stud-ies: antistreptavidin IgG1 (provided by Amgen, Inc., ThousandOaks, CA) and rhGH. Antistreptavidin was formulated at ei-ther 1 or 35 mg/mL in 20 mM histidine buffer, pH 6. rhGH waspurified from inclusion bodies produced in E. coli19 and formu-lated at 1.75 mg/mL in 2 mM sodium phosphate, pH 7.4. Ina limited study to confirm video results, formulations contain-ing a second monoclonal antibody (mAb1) at a concentration of1mg/mL in a 20-mMhistidine buffer, pH 6, were also examined.

Sample Preparation and Mechanical Shock Treatment

Protein formulations were filtered with a 0.2-um syringe filterprior to shock treatment. Four milliliter sterilized borosilicateglass vials (West Pharmaceuticals, Lionville, Pennsylvania)were rinsed with filtered water, filled with 1 or 4 mL of proteinformulation and sealed at ambient pressure with chlorobutylstoppers (West Pharmaceuticals) that had been rinsed withfiltered water. To simulate the high g-forces and mechanicalshock that may be sustained when a drug vial is dropped, aLansmont Model 15D shock test system (Monterey, California)was used. Vials were mounted using a custom-designed holderin an upright position on the falling block of the shock testsystem, and supported by a TeflonR base. Sample vials weredropped from heights ranging from 1040 inches. Undroppedsamples (referred to as 0 in) in identical vials were used ascontrols.

High-Speed Video Analysis

In initial experiments, a Keyence (Osaka, Japan) VW-9000high-speed microscope was used to record images at a rate of6000 frames per second of the vials as they underwent mechan-ical shock. In additional studies, a Phantom high-speed camerawas used to record images at a rate of 66,666 frames per secondduring mechanical shock as vials containing either 1 or 3 mLof mAb1 formulation were dropped in the Lansmont shock testsystem.

Using the Phantom high-speed camera, high-speed imageswere also recorded of vials that were dropped by hand andcontacted a hard surface after a free fall of approximately 1 m.These vials struck the surface at various angles.

Hydroxyphenyl Fluorescein Assay to Detect Cavitation-InducedFree Radicals

In the presence of free radicals, hydroxyphenyl fluorescein(HPF) forms the highly fluorescent product fluorescein. HPF(Invitrogen, Carlsbad, California) was added to the 1 mg/mLprotein solutions to a concentration of 5 :M, and 1 mL ofthe resulting solution filled into 4 mL vials. Following shock

treatment, samples were placed in a fluorometer (SLM 8000;SLM, Urbana, Illinois). After excitation at 490 nm, fluorescenceemission intensity of the samples was measured at 515 nm. Afluorescein standard curve was used to determine the concen-tration of fluorescein in the samples.

Size-Exclusion Chromatographic Analysis of Soluble Aggregatesand Monomeric Protein

Samples were analyzed using size-exclusion chromatographyto determine the amount of monomeric protein and soluble ag-gregates before and after mechanical shock treatment. Prior toanalysis, samples were centrifuged for 10 min at 9300g to pel-let any insoluble material. Size-exclusion chromatography wascarried out using a Beckman HPLC system (Beckman Coul-ter Inc., Fullerton, California). For analysis of antistreptavidinsamples, the mobile phase was 100 mM sodium phosphate,300 mM sodium chloride, 0.01% sodium azide, pH 7, and aflow rate of 0.6 mL/min was used. For analysis of rhGH sam-ples, themobile phase consisted of 25mMammonium bicarbon-ate, 0.01% sodium azide, pH 7.4, and 1 mL/min flow rate wasused. For both of the proteins, 100 :L injections on a TSK-GELG3000SWXL column with an SW guard column were used. Ab-sorbance at 280 nm was measured.

Microflow Imaging Analysis of Subvisible Particulates

A benchtop FlowCAM (Fluid Imaging Technologies, Yarmouth,Maine) was used to image and count particles of sizes greaterthan or equal to 2 :m. A 10x objective and 100 :M flowcellwere used to analyze 215 :L of sample per assay. Betweensamples, the flowcell was flushed with water that had beenfiltered through a 0.2-:m filter.

Recovery and Quantification of Protein Adhered to Vial Wall afterCavitation

Two methods were used to recover protein that had adheredto vial walls. The first method used urea to dissolve proteinfrom the vial walls. Following shock treatment, all liquid wasremoved from the vials with a pipette. The vials were rinsedtwice with 1 mL formulation buffer to remove unadhered pro-tein, and 250 :L of 8 M urea were added to the vials. Sampleswere then incubated overnight at room temperature with mix-ing on a rotator such that the urea solution contacted the entirevial wall surface. The next day, the urea solution was diluted4x in order to be compatible with a standard Bradford totalprotein assay. Bradford reagent and the diluted urea solutionswere mixed in a 1:1 ratio, incubated at room temperature for5 min, and the absorbance at 595 nm was measured and com-pared with that for a placebo blank. A standard curve using thesame protein and buffer as that in the mechanically shockedsamples was used.

A second method used guanidine hydrochloride solution toremove the protein from the vial walls. Following mechanicalshock treatment, the solution was removed from the vials witha pipette. Residual solution was then removed by rinsing thevials twice with 1 mL buffer. Finally, 0.5 mL of 7.5 M guanidinehydrochloride, 0.25 M Tris, pH 7.5, was added to the vials andincubated overnight at room temperature mixing on a rotatorsuch that guanidine could access all of the vial wall surface.The next day, the absorbance was measured at 280 nm to de-termine the amount of protein recovered from the vial walls.

DOI 10.1002/jps.24259 Randolph et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:602611, 2015

604 RESEARCH ARTICLE Pharmaceutical Biotechnology

Absorbance values were corrected by subtracting absorbancevalues for appropriate buffer blanks.

Tryptic Digestion/HPLC Analysis of Protein Degradation

Samples from the bulk solution and samples recovered from thewashed vial walls were enzymatically digested following shocktreatment using trypsin. For antistreptavidin samples, a proto-col similar to that developed by Ren et al.20 was used. All of thebulk solutionwas removed from the vials after shock treatment.The bulk solution was diluted into denaturation buffer (7.5 Mguanidine hydrochloride, 0.25 M Tris, pH 7.5) to a final concen-tration of 1 mg/mL and a final volume of 500 :L. After the solu-tion was removed from the vials, they were rinsed with buffer.Five-hundred milliliter of denaturation buffer was added to thevials and incubated overnight to recover protein from the vialwalls. The next day, 3 :L of 0.5 M dithiothreitol (DTT) wereadded to all samples and incubated at room temperature for 30min. Next, 7 :L of 0.5 M iodoacetic acid were added and incu-bated in the dark for 15 min. Four microliter of 0.5 M DTT wasadded to the samples and samples were buffer exchanged intodigestion buffer (0.1 M Tris, pH 7.5) using NAP-5 columns (GEHealthcare, Piscataway, New Jersey). Trypsin was added tothe samples in a 1:25 trypsinantistreptavidin ratio and incu-bated at 37C for 30 min. The reaction was quenched with 5 :Lof 20% formic acid. Reversed-phase chromatography analy-sis was conducted using a Phenomenex (Torrance, California)Jupiter C18 column on an Agilent (Palo Alto, California) 1100Series system. The flowrate was 0.2 mL/min and a linear gra-dient of 0%50% B (buffer A: 0.1% trifluoroacetic acid; bufferB: 90% acetonitrile; 0.085% trifluoroacetic acid) over 195 minwas used for elution. Absorbance at 215 nm was measured.

For rhGH samples, following shock treatment, all bulk so-lution was removed from vials, and the vials were rinsedwith buffer. Bulk solution and the vials were frozen sepa-rately at 80C and shipped on dry ice to The University ofKansas, where trypsin or chymotrypsin digestion and massspectroscopy analysis were performed on the samples. Sam-ples of the bulk solutions that had not been subjected to shocktreatment were used as controls. For the bulk solution sam-ples, 10 :L of sample were digested with 50 :L of trypsin orchymotrypsin solution (20 :g/mL in 0.1Mammoniumbicarbon-ate, pH 7.8) at 37C overnight. For analysis of protein adheringto vial walls, 50 :L of trypsin solution were added directly tothe vials and the protein adhering to the vial walls was digestedovernight at 37C. Following digestion, liquid chromatography-mass spectrometry (LCMS) analysis was run with a Presto-FF-C18 column. A 1-h gradient from 10%50% acetonitrile wasused followed by a shorter gradient to 80%acetonitrile forwash-ing. The chromatographic system used was an Acquity UPLCsystem (Waters Corporation, Milford, Massachusetts) coupledto a SYNAPT-G2 (Waters Corporation) mass spectrometer. Theflow rate was 20 :L/min. The SYNAPT-G2 instrument was op-erated for maximum resolution with all lenses optimized onthe [M+2H]2+ ion from the [Glu]1-fibrinopeptide B. The conevoltage was 45 V and AR was admitted to the collision cell. Thespectra were acquired using a mass range of 502000 amu. Thedata were accumulated for 0.7 s per cycle. Data were analyzedwith the ProteinLynx Global Server from Waters Corporation.Specific oxidation yields were calculated through the integra-tion of peak areas in the extracted ion chromatogram of themost abundant ion of the respective tryptic peptide. It was

assumed that the electrospray-ionized peptides of any oxida-tive modifications have similar mass spectrometric propertiesas the respective ions of the native peptides because of the smallmass difference.

Computational Fluid Dynamics Calculations of MechanicalShock-Induced Cavitation

The transient behavior of fluid within a vial impacting a solidsurface was modeled using Fluents (ANSYS, Inc., Canonsburg,Pennsylvania) volume of fluid, axisymmetric solver with theReynolds Averaged Navier-Stokes turbulence model describedby the realizable k-g closure model. A pressure velocity coupledalgorithm was used to calculate the flow field with a second-order discretization for the momentum, turbulence, and energyequations. Vials were modeled as containing water in the bot-tom one-third of the vial with air at 1 atm in the remainingarea with appropriate meniscus height and shape. The wallboundaries were modeled with no-slip conditions using stan-dard wall functions to account for the viscous boundary layerat the surface of the vial. The temperatures for all the bound-ary conditions were set to 293K. The fluid used in the com-parison of simulated results to experimental data for modelvalidation was water at standard temperature and pressure.The Ideal Gas Law was selected to allow compressible behaviorof the gas phase and a user-defined function implemented forthe compressible behavior of the liquid phase. The compress-ibility of water as a function of pressure was calculated usingthe bulk modulus of water. A sliding wall mesh was utilizedto simulate the fall and rebound of the vial. Initial fluid veloc-ities matched the walls downward velocity at impact. Threeimpact cases were evaluated to simulate dropping a vial of con-taining water from a height of 90 cm onto soft and hardsurfaces: (1) a soft, 5-cm foam-padded surface, yielding an im-pact of 50 g, (2) a soft, 1-cm foam-padded surface, yielding animpact of 500 g, and (3) a hard metal surface, with a cor-responding impact of 5000 g. These gravitation accelerationswere estimated from high-speed images using the vial geome-try, recorded vial positions, and the video frame rate to deter-mine velocities immediately before and after contact. The 90-cmdrop heights were chosen to correspond drop distances that avialmight experiencewhen falling off a counter or out of a usershand.

A total of 21,219 quadrilateral elements were used to achievea high-quality computational mesh, resulting in an averagemesh density of 2072 per square millimeter. The maximumcell skewness was 0.50, and the minimum orthogonal qualityof 0.79. The maximum growth rate from one element to thenext was 10%. Convergence criteria for the computational do-mains were met when the residuals for pressure, momentum,and turbulence were below 103 and below 106 for the energyequation.

RESULTS

Indications of Cavitation

High-Speed Video Analysis

To visualize fluid motion and cavitation occurring in liquidswithin partially filled vials during shock treatment, high-speedvideo of antistreptavidin formulations in partially filled vialswas captured as the vials were dropped in the Lansmont Model

Randolph et al., JOURNAL OF PHARMACEUTICAL SCIENCES 104:602611, 2015 DOI 10.1002/jps.24259


Figure 1. Time lapse images of fluid in a partially filled vial as it contacts a solid surface after being dropped from a height of 40 inches. In thethird frame, recorded 500 :s after contact, two large cavitation bubbles become visible. These cavitation bubbles collapse rapidly, with completecollapse occurring around 1000 :s after the initial contact.

15D shock test system. Figure 1 shows selected time lapse im-ages of a representative vial during shock treatment; full videois available in the supplementary material. Time 0 representsthe time point when the vial hit the bottom platform of theshock tower and began to rebound. As the vial began to re-bound, the meniscus inverted, and a jet of liquid shot up towardthe top of the vial. As the jet formed, cavitation bubbles couldbe seen near the bottom of the vial at about 500 :s after impact.The cavitation bubbles quickly collapsed within a few hundredmicroseconds.

Similar cavitation was also observed when mAb1 solutionsin the same type of 4 mL vials were dropped in the LansmontModel 15D shock test system from a height of 40 inches (Fig. 2).Cavitation was evident in images recorded after 165 :s. Vialsfilled with 3 mL of mAb1 solution showed more extensive cav-itation than vials filled with 1 mL, and the cavitation bubblesthat were produced appeared to oscillate in size. After some ofthe bubbles collapsed, new cavitation events rapidly followed,suggesting that shockwaves emanating from the collapsing pri-mary cavitation sites and/or that echoes of the original shockwave caused additional bubble formation. Cavitation was notobserved when vials were filled with deionized water (Fig. 3).

When the orientation of the vial upon contact with a hardsurface was varied by hand-dropping vials containing 1 mg/mLmAb1 solutions and allowing them to free fall for approximately1m, cavitation was observed. In these vials, the pattern and po-sition of the cavitation bubbles were highly variable, dependingon the orientation of the vials at the point of contact with the

surface. Occasionally, vial breakage was observed as a result ofthe free-fall drops; no attempt was made to recover or analyzesamples in vials that broke. A representative sequence of high-speed video images for a square impact between the vial andthe surface is shown in Figure 4, and for an angled impact inFigure 5; full video sequences are provided in the supplemen-tary material. In these studies, where images were recorded us-ing the Phantom camera at very high speeds, cavitation couldbe observed within 30 :s after the falling vial contacted thehard surface.

HPF Assay

Because collapse of cavitation bubbles was expected to producefree radicals, HPF was used to detect hydroxyl radicals pro-duced during mechanical shock treatment. Upon reacting witha hydroxyl radical, HPF is converted to the fluorescentmoleculefluorescein.21 HPFwas added to vials prior to mechanical shocktreatment and the fluorescence intensity was measured aftertreatment. A fluorescein standard curve was used to determinethe concentration of fluorescein produced during shock treat-ment, and results are shown in Figure 6. Between 20 and 40 nMof fluorescein were detected in the vials after shock treatment.The quantity of fluorescein that is formed, is only a qualitativeindicator of the formation of free radicals, because the free rad-icals that are produced during cavitation would be expected toreact not only with HPF but also with the buffer and protein.Thus, it is likely that far more than 2040 nM of free radicalswere formed as a result of shock treatment.

Figure 2. Time lapse images of mAb1 solution in a fully filled vial as it contacts a solid surface after being dropped from a height of 40 inches.In the third frame, recorded 330 :s after contact, many large bubbles appear and grow until they begin to collapse in the seventh frame labeled990 :s. These cavitation bubbles collapse rapidly, with complete collapse occurring around 1300 :s after the initial contact. The TeflonR bumperon which the vials were mounted in the mechanical shock tester is visible in the bottom of each frame.

Figure 3. Time lapse images of deionized water in a partially filled vial as it contacts a solid surface after being dropped from a height of40 inches. There is no evidence for any bubble formation or cavitation in any of the images.



Figure 4. Time lapse images of a 1 mg/mL mAb formulation in a partially filled vial as it contacts a solid surface squarely after being droppedfrom a height of 1 m without the aid of a shock tower. In the third frame, recorded 30 :s after contact, many small bubbles rapidly appear anddisappear in an oscillatory fashion until they completely disappear in the ninth frame labeled 120 :s.

Figure 5. Time lapse images of fluid in a partially filled vial containing 1 mg/mL mAb as it contacts a solid surface at an angle after beingdropped from a height of 1 m without the aid of a shock tower. In the second frame, recorded 30 :s after contact, many small bubbles appear andrapidly disappear, until they almost completely disappear in the ninth frame labeled 270 :s.

Results of Computational Fluid Dynamics Calculations

Numerical simulations of the fluid behavior, in particularpressure-field characteristics, were conducted under each ofthe three impact cases (described in Materials and Methods)to assess whether the conditions for cavitation would be sat-isfied and in what regions of the vial such conditions mightoccur. Figure 7 shows the computed pressure fields within the

Figure 6. Concentration of fluorescein liberated from reaction ofp-phenylfluorescein with hydroxyl radicals generated during shocktreatment as a function of drop height. Concentrations were deter-mined from the difference in fluorescence between fluid inmechanicallyshocked vials and fluid in identical, untreated control vials.

vials. At the bottom of each of the vials is a low-pressure re-gion, with the magnitude of the low pressure being dependenton the drop conditions. Each of the three impact cases has thesame maximum pressure in the upper portion of the vial, butthe low-pressure magnitudes are greatly different. The vaporpressure of the water in these simulations is 3.5 kPa. Vialsdropped on soft surfaces, which produce shock that mightbe typical of that experienced by the contents of a vial thathas been protected by secondary packaging, are higher than3.5 kPa, whereas vials dropped on hard surface showed lowpressures well below the vapor pressure of water. This indi-cates that only the hard-impact, metal on metal case is likelyto cause cavitation.

Analysis of Bulk Liquid Within Mechanically Shocked Vials

Following shock treatment, the solution from dropped and con-trol nondropped vials was removed and analyzed for proteinaggregates, particles, and protein chemical damage.

Size-Exclusion Chromatography

Size-exclusion chromatography was used to measure the con-centrations of monomer and soluble aggregates in the bulk so-lution following shock treatment. For all sample types tested(1.0 mg/mL antistreptavidin, 35 mg/mL antistreptavidin, and1.75 mg/mL human growth hormone), no significant loss ofmonomer or increased level of soluble aggregates resultingfrom application of mechanical shock could be detected by size-exclusion chromatography (data not shown).



Figure 7. Numerical simulation of one-third filled vial with water being dropped from 90 cm impacting under the following conditions: (left) a2.0 inches foam-padded impact (50 g), (center) a 0.5 inch foam-padded impact (500 g), and (right) a metal surface (5000 g).

Particle Count and Size Distribution

FlowCAM microflow imaging analysis was used to count andsize particles of size equal to or greater than 2 :m in 1 mLsamples of each of three formulations (antistreptavidin at 1 or35 mg/mL, and rhGH at 1.75 mg/mL) following application ofmechanical shock to vials (Figs. 8a, 8c, and 8e). Shock treatmentresulted in an increase in the number of particles. Drop heightdid not appear to have a large effect on the number of particlescounted; all drop heights produced about the same number ofparticles for a given protein and protein concentration. For bothantistreptavidin and rhGH, a large fraction of the particlesthat formed after application of mechanical shock showed long,fibrillar aspect ratios (Figs. 8b, 8d, 8f).

In a separate experiment to determine the effect of fill vol-ume on particle formation, a set of samples was prepared byfilling vials with either 1 or 4 mL of a 1.75 mg/mL rhGH formu-lation. These vials were dropped from a height of 40 inches inthe Lansmont shock tester. In vials filled with 4 mL solution,cavitation was more apparent than in the vials with a 1-mL fillvolume. The concentrations in the solution of microparticles ofsize equal to or greater than 2 :m were measured before andafter shock treatment using the FlowCam instrument. Prior toshock testing, an average of 11,100 particles of size equal toor greater than 2 :m/mL was detected in the rhGH solutions.After a single drop, the particle concentrations increased to16,800 700 particles/mL in vials filled with 1 mL of solution,and to 31,300 3500 particles/mL in vials filled with 4 mL ofsolution.

Tryptic Digestion/RP-HPLC Analysis of Protein Degradation

Trypsin digestion followed by reverse-phase HPLC was used tomonitor chemical damage to the protein as a result of shocktreatment. The bulk solution from 35 mg/mL antistreptavidinsamples was digested with trypsin and analyzed on reversed-phase chromatography. No chemical differences in proteins re-maining in the bulk solution were detected in the dropped sam-ples compared with the control samples (data not shown).

Bulk solution rhGH samples were also digested with trypsinand analyzed by mass spectroscopy. Both control samples

and samples of bulk solution dropped from a height of 40inches showed similar levels of oxidation of methionine-14,methionine-125,methionine-170, and tryptophan (about 1% ox-idation or less, data not shown).

Analysis of Protein Recovered from Vial Walls

Following shock treatment, gelatinous specks of protein wereeasily visible on the vial walls. The protein that adhered tothe vials walls in the form of these gelatinous globules wasrecovered and quantified and analyzed for chemical damage.

Quantification of Recovered Protein

The protein adhering to the vial walls was recovered usingguanidine or urea to remove the protein from the vial wallsand quantified using the absorbance at 280 nm and the Brad-ford assay. The amount of protein recovered per vial is shown inFigure 9. In all cases, guanidine was more effective at removingadherent protein from the vial walls than was urea. Althoughthe efficiency with which protein could be recovered from thevial walls was different depending on whether guanidine orurea were used, in both cases (and for both proteins tested) theamount of protein that adhered to the walls and could be re-moved for analysis increased as the drop height was increasedin the mechanical shock tester.

Trypsin Digestion

After recovering the protein with guanidine, trypsin or chy-motrypsin digestion and reverse-phase chromatography wereused to monitor any chemical modifications to the protein.Chemical damage as a result of shock treatment was not de-tected for the 35 mg/mL antistreptavidin samples (data notshown).

Following recovery from the vial walls, rhGH samples weredigested and analyzed by mass spectroscopy with focus on theoxidation of Met and Trp. Five series of samples were analyzedafter trypsin digestion, where large sample-to-sample varia-tions were recorded. For example, series 1 clearly indicateda significant increase of oxidation of Met-14 to Met sulfoxide(MetSO) and the single Trp residue compared with controls,



Figure 8. Particle counts and examples of particlemorphologies for shock-treated 1mg/mLantistreptavidin (a and b), 35mg/mLantistreptavidin(c and d), and 1.75 mg/mL human growth hormone (e and f). The black portion of each bar represents particles with diameters at least 2 : andless than 3 : ; the dark gray portion represents particles with diameters at least 3 : and less than 5 : ; the light gray portion represents particleswith diameters at least 5 : and less than 10 : ; and the white portion represent particles with diameters at least 10 : . Each bar represents theparticle count from one vial.

whereas series 2 showed only a significant increase of oxida-tion of Met-170. Series 3 and 4 failed to indicate any signif-icant increase of oxidation compared with controls, whereasseries 5 showed an increase in Met-14 oxidation only for vialscontaining 4 mL solution and dropped from 40 inches height.Series 5 was also analyzed after chymotrypsin digestion ofrhGH, which did not show significant oxidation of Met-14 toMetSO. It is possible that digestion by one enzyme versus an-other enzyme can bias toward certain products. However, atthis point, we can only conclude that oxidation of Met and Trpmay be possible, but cannot be consistently observed. This re-sult is consistent with the large sample-to-sample variations in

cavitation observed by the high-speed camera. In this context,it is also important to note that hydroxyl radicals, generatedvia cavitation, react rather unselectively with proteins target-ing all 20 amino acids, though with different rate constants.22

Upon reaction with Met, hydroxyl radicals initially generatesulfide radical cations, which subsequently decompose into"-(alkylthio)alkyl radicals, whereas MetSO is not a primaryproduct of Met oxidation by hydroxyl radicals unless superox-ide is generated simultaneously.23 It appears that, whereverobserved, MetSO formation in rhGH may be a consequenceof secondary processes, that is, peroxyl radical or peroxideformation.



Figure 9. Amount of protein recovered from the walls of dropped vials following treatment with urea (a, c, and e) or guanidine hydrochloride(b, d, and f). Dropped vials contained 1 mg/mL antistreptavidin (a and b), 35 mg/mL antistreptavidin (c, and d), or 1.75 mg/mL human growthhormone (e and f).

DISCUSSION

Inadvertent dropping of vials and syringes almost certainlyoccurs in the clinic and during in-home administration of ther-apeutic protein formulations. Clearly, such events have thepotential to cause cavitation, as in the present study, even therelatively minor shock incurred when vials dropped only 10inches onto hard surfaces caused cavitation. In turn, the vio-lent collapse of cavitation bubbles and the resulting free rad-icals, high temperatures, and secondary shock waves may belikely to cause localized oxidative and conformational damageto proteins.

In the numerical simulations, the pressure fields within flu-ids subjected to a high-acceleration collision (such as that typ-ical of a dropped vial) indicate the generation of high-pressure

shock waves and associated low-pressure, cavitation-prone re-gions along the bottom surface of the vial. The strength, loca-tion, and timing of this low-pressure region are consistent withthe high-speed imaging of large vapor bubble formation withina vial under similar conditions. The application of energy-absorbing materials such as those that might be used in sec-ondary packaging can mitigate this issue, as clearly demon-strated by the two soft-impact conditions. Under these condi-tions, neither a shock wave discontinuity in the pressure fieldnor a pressure region with pressures below the vapor pressureof water is observed. Future numerical studies are focusingon understanding the exact roles that dynamics (acceleration),vial shape, and fluid properties (density, surface tension, vis-cosity, etc.) play in establishing the shock wave and dictatingthe magnitude of the low-pressure region.



In the bulk solution of the vials, there were no detectablechanges in monomer or soluble aggregate concentrations afterthe vial contents were subjected to mechanical shock. This isnot surprising, as the energy released upon collapse of a cavita-tion bubble is expected to be concentrated in an extremely smallvolume. There were also no visible insoluble aggregates formedin the bulk solution following mechanical shock treatment; allof the visible aggregates appeared to be attached to the vialwalls. However, there was an increase in the number of sub-visible particles of size greater than or equal to 2 :m detectedby microflow imaging analysis following shock treatment. Theability of microflow imaging techniques to detect protein ag-gregates at levels below the sensitivity limit for size-exclusionchromatographic analysis has been documented.24 Mechanicalshock treatment of samples formulated at the higher concen-tration of antistreptavidin generated more particles than didshock treatment of samples formulated at the lower concentra-tion of antistreptavidin (Fig. 8). Again, this might be expected ifonly protein in a small volume surrounding a collapsing cavita-tion bubble is damaged. Over the range of drop heights tested,the increase in subvisible particles in samples subjected to me-chanical shock compared with undropped samples was not de-pendent on drop height, perhaps suggesting a threshold valueof mechanical shock is required to create cavitation-associatedformation of particles.

Joubert et al.,25 characterized the aggregates of antibodiesexposed to various types of stress. On the basis of size andmorphology, the aggregates produced as a result of shock treat-ment most closely resemble the aggregates produced duringmechanical stresses such as agitation and stirring in theirclassification scheme. This might indicate that in addition tothe effects of cavitation-induced free radicals, the mechanicalstresses caused by cavitation could also play an important rolein generating protein aggregates.

After shock treatment, protein aggregates adhering to thevial walls were visible. The amount of protein recovered fromthe vial walls increased as drop height increases and was con-centration dependent. More protein was recovered from the vialwalls following shock treatment of 35 mg/mL antistreptavidinthan was recovered following shock treatment of 1 mg/mL an-tistreptavidin. The amount of adherent protein scaled roughlywith protein concentration. The protein that adhered to thewallof the vial after application of mechanical shock was difficultto remove, and resisted rinsing by buffer solutions. We specu-late that the adherent properties of the gelatinous globules onthe walls may have resulted from the potential oxidative dam-age and subsequent chemical cross-linking that resulted fromthe free radicals and high temperatures produced as cavitationbubbles collapsed, or frommechanical stresses caused by shockwaves associated with collapse of cavitation bubbles.

In vials filled with deionized water with no added proteinor buffer, we did not see video evidence of cavitation causedby the mechanical shocks that we applied. In previous work,we have shown that DNA molecules may serve as nucleationsites for cavitation.26 We speculate that in the present case,protein molecules may play a similar role. Likewise, surfactantmicelles, although not examined in this study, might also beable to serve as nucleation sites for cavitation.

In all of our studies, the cavitation phenomenon washighly variable from sample to sample. High-speed videoimages showed that the number and location of cavitationsites were different in every sample tested, perhaps as a

result of subtle differences in shape of the interior bot-tom surface of each vial. The sample-to-sample variationwas even more apparent in the samples that were hand-dropped and thus contacted the surface at a various randomangles.

The levels and types of chemical damage that were observedin the proteins were also quite variable. This variation was con-sistent not only with the variable cavitation levels that wereobserved by high-speed photography, but also with the inher-ent difficulties in sampling the damaged protein, which hadformed small, adherent, gelatinous particles on the vial wallsthat were difficult to rinse off with buffer, instead requiringurea or guanidine HCL to remove them.

Chemical changes to antistreptavidin adhering to the vialwalls were not detected, but in many rhGH samples where cav-itation had occurred extensive oxidation was also observed. Wehypothesize that this damage to rhGH is because of the reactiveoxygen species such as OH radicals produced during cavitationreacting with the protein. A possible explanation for the lackof oxidative damage detected in the antistreptavidin samplemight be the lack of sensitivity of our analytical technique to po-tential oxidative damage in the relatively large mAb molecule,wherein oxidative damage might have been dispersed over alarger number of residues. In addition, because our peptide-focused mass spectrometry technique is not sensitive to glycanchemical changes, damage would not be detected if the OHradicals reacted preferentially with the highly solvent-exposedglycosylated portion of the molecules.

CONCLUSIONS

Themechanical shock caused by dropping unprotected (e.g., notpackaged in protective secondary packaging) glass vials con-taining liquid formulations of therapeutic protein from heightsas low as 10 inches was sufficient to cause cavitation in theliquid, as evidenced high-speed video and the production of hy-droxyl radicals detected by HPF. Numerical simulations wereconsistent with high-speed video and showed that cavitationmay be mitigated by energy absorbing materials, such as thoseoften found in secondary packaging.

As a result of this cavitation, subvisible particles were cre-ated in the bulk solution of the vials. In addition, protein par-ticles were visible on the vial walls following shock treatment.The amount of protein adhering to the vial walls increased withdrop height. For rhGH, the protein adhering to the vial wallsfollowing shock treatment showed increased levels of oxidation.

The loss of monomeric protein because of cavitation likelywas not large enough to affect potency of protein in the bulksolution. However, the mechanical shock caused subvisible par-ticles to form, and such particles have the potential to affectimmunogenicity of the formulation.27 Thus, packaging, formu-lation excipient composition, and process conditions should bedesigned to minimize mechanical shock and the potential forcavitation.

ACKNOWLEGMENT

Funding for this workwas provided by Amgen, Inc. and throughthe NIH #5RO1 EB006006.



REFERENCES

1. WangW, Singh S, ZengD,KingK,NemaS. 2007. Antibody structure,instability, and formulation. J Pharm Sci 96(1):126.2. Rosenberg AS. 2006. Effects of protein aggregates: An immunologicalperspective. AAPS J 8(3):E501-E507.3. Kessler M, Goldsmith D, Schellekens H. 2006. Immunogenicity ofbiopharmaceuticals. Nephrol Dial Transplant 21:912.4. Thompson L, Doraiswamy L. 1999. Sonochemistry: Science and en-gineering. Ind Eng Chem Res 38(4):12151249.5. Suslick KS, Flannigan DJ. 2008. Inside a collapsing bubble: Sonolu-minescence and the conditions during cavitation. Annu Rev Phys Chem59:659683.6. McNamara W, Didenko Y, Suslick K. 1999. Sonoluminescencetemperatures during multi-bubble cavitation. Nature 401(6755):772775.7. Flannigan D, Suslick K. 2005. Plasma formation and temperaturemeasurement during single-bubble cavitation. Nature 434(7029):5255.8. Didenko Y, McNamara W, Suslick K. 1999. Temperature of multi-bubble sonoluminescence in water. J Phys Chem A 103(50):1078310788.9. Pestman J, Engberts J, Dejong F. 1994. SonochemistryTheory andapplications. Recueil Des Travaux Chimiques Des Pays 113(12):533542.10. Riesz P, Kondo T. 1992. Free-radical formation induced by ultra-sound and its biological implications. Free Radic Biol Med 13(3):247270.11. Riesz P, Berdahl D, Christman C. 1985. Free-radical generationby ultrasound in aqueous and nonaqueous solutions. Environ HealthPerspect 64:233252.12. Marchioni C, Riccardi E, Spinelli S, DellUnto F, Grimaldi P, BediniA, Giliberti C, Giuliani L, Palomba R, Castellano A. 2009. Structuralchanges induced in proteins by therapeutic ultrasounds. Ultrasonics49(67):569576.13. Edwards J, James C, CoakleyW, BrownR. 1976. Effect of ultrasoniccavitation on protein antigenicity. J Acoust Soc Am 59(6):15131514.14. Gulseren I, Guzey D, Bruce B, Weiss J. 2007. Structural and func-tional changes in ultrasonicated bovine serum albumin solutions. Ul-trason Sonochem 14(2):173183.15. Suslick K, Grinstaff M, Kolbeck K, Wong M. 1994. Characteriza-tion of sonochemically prepared proteinaceous microspheres. UltrasonSonochem 1(1):S65S68.

16. Grinstaff M, Suslick K. 1991. Air-filled proteinaceousmicrobubblesSynthesis of an echo-contrast agent. Proc NatlAcad Sci USA 88(17):77087710.17. Ricci M, Donahue L, Ball N, Stackhouse N, Lee J, Krishnan S,Pallitto M. 2008. Silicone oil induced particle formation in prefilledsyringes. 236thACSNationalMeeting, August 1721. Philadelphia, PA:American Chemical Society.18. Bonk R EB, Huinker A, Pallitto M, Ricci M, Stackhouse S. 2009.Stabilized protein compositions. Patent WO2009099641A2.19. Fradkin AH, Boand CS, Eisenberg SP, Rosendahl MS, RandolphTW. 2010. Recombinant murine growth hormone from E. coli inclu-sion bodies: Expression, high-pressure solubilization and refolding, andcharacterization of activity and structure. Biotechnol Prog 26(3):743749.20. Ren D, Pipes G, Liu D, Shih L, Nichols A, Treuheit M, Brems D,Bondarenko P. 2009. An improved trypsin digestion method minimizesdigestion-induced modifications on proteins. Anal Biochem 392(1):1221.21. Setsukinai K, Urano Y, Kakinuma K, Majima H, Nagano T. 2003.Development of novel fluorescence probes that can reliably detect re-active oxygen species and distinguish specific species. J Biol Chem278(5):31703175.22. Davies M. 2005. The oxidative environment and protein damage.Biochim Biophys Acta 1703(2):93109.23. Schoneich C. 2005. Methionine oxidation by reactive oxygenspecies: Reaction mechanisms and relevance to Alzheimers disease.Biochim Biophys Acta 1703(2):111119.24. Barnard JG, Singh S, Randolph TW,Carpenter JF. 2011. Subvisibleparticle counting provides a sensitive method of detecting and quan-tifying aggregation of monoclonal antibody caused by freeze-thawing:Insights into the roles of particles in the protein aggregation pathway.J Pharm Sci 100(2):492503.25. JoubertMK, LuoQZ, Nashed-Samuel Y,Wypych J, Narhi LO. 2011.Classification and characterization of therapeutic antibody aggregates.J Biol Chem 286(28):2511825133.26. Lentz YK, Anchordoquy TJ, Lengsfeld CS. 2006. DNA acts as anucleation site for transient cavitation in the ultrasonic nebulizer. JPharm Sci 95(3):607619.27. Rombach-Riegraf V, Karle AC, Wolf B, Sorde L, Koepke S, GottliebS, Krieg J, Djidja MC, Baban A, Spindeldreher S, Koulov AV, KiesslingA. 2014. Aggregation of human recombinant monoclonal antibodiesinfluences the capacity of dendritic cells to stimulate adaptive T-cellresponses in vitro. PLoS One 9(1):e86322.


do noit drop - cavitation in vials

Documents