Drying-induced variations in physico-chemical properties of amorphous pharmaceuticals and their...

Drying-Induced Variations in Physico-Chemical Properties ofAmorphous Pharmaceuticals and Their Impact on Stability (I):Stability of a Monoclonal Antibody

AHMAD M. ABDUL-FATTAH,1 VU TRUONG-LE,2 LUISA YEE,2 LAUREN NGUYEN,2

DEVENDRA S. KALONIA,1 MARCUS T. CICERONE,3 MICHAEL J. PIKAL1

1Department of Pharmaceutical Sciences, University of Connecticut, Storrs, Connecticut 06269

2Formulation Development, MedImmune Vaccines, Mountain view, California 94043

3Polymers Division, National Institute of Standards and Technology (NIST), Gaithersburg, Maryland 20899

Received 6 August 2006; revised 7 November 2006; accepted 21 November 2006

Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20859

ABSTRACT: The present study was conducted to investigate the impact of dryingmethod and formulation on the storage stability of IgG1. Formulations of IgG1 withvarying levels of sucrose with and without surfactant were dried by different methods,namely freeze drying, spray drying, and foam drying. Dried powders were characterizedby thermal analysis, scanning electron microscopy, specific surface area (SSA) analysis,electron spectroscopy for chemical analysis (ESCA), solid state FTIR, and molecularmobility measurements by both isothermal calorimetry and incoherent elastic neutronscattering. Dried formulations were subjected to storage stability studies at 408C and508C (aggregate levels were measured by size exclusion chromatography initially and atdifferent time points). Both drying method and formulation had a significant impact onthe properties of IgG1 powders, including storage stability. Among the drying methods,SSA was highest and perturbations in secondary structure were lowest with the spray-dried preparations. Sucrose-rich foamshad the lowest SSAand the lowest protein surfaceaccumulation. Also, sucrose-rich foams had the lowest molecular mobility (both fastdynamics and global motions). Stability studies showed a log-linear dependence ofphysical stability on composition. Preparations manufactured by ‘‘Foam Drying’’were the most stable, regardless of the stabilizer level. In protein-rich formulations,freeze-dried powders showed the poorest storage stability and the stability differenceswere correlated to differences in secondary structure. In stabilizer-rich formulations,stability differences were best correlated to differences in molecular mobility (fastdynamics) and total protein surface accumulation. � 2007Wiley-Liss, Inc. and the American

Pharmacists Association J Pharm Sci 96:1983–2008, 2007

Keywords: composition heterogeneity; electron spectroscopy for chemical analysis;glass dynamics; incoherent elastic neutron scattering; molecular mobility; proteinstabilization; protein secondary structure; structural relaxation time; specific surfacearea; surface coverage

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 8, AUGUST 2007 1983

intended to imply that the materials or equipment identifiedare necessarily the best available for the purpose.

Correspondence to: Michael J. Pikal (Telephone: (860)-486-3202; Fax: (860)-486-4998; E-mail: [email protected])

Journal of Pharmaceutical Sciences, Vol. 96, 1983–2008 (2007)� 2007 Wiley-Liss, Inc. and the American Pharmacists Association

Official contribution of NIST; not subject to copyright inthe U.S. Certain commercial equipment, instruments, ormaterials are identified in this article in order to specifythe experimental procedure adequately. Such identificationis not intended to imply recommendation or endorsement bythe National Institute of Standards and Technology, nor is it

INTRODUCTION

Drying proteins from aqueous solution to form amolecular dispersion of a protein in a glassymatrix has now become a standard practice in thepharmaceutical industry for stabilization.1–4

Freeze drying is widely used for parenteraladministration,1–3 whereas spray drying iswidely used for pulmonary delivery.4,5 Otherdrying methods have emerged as alternate dryingmethods to both such as spray coating,6 super-critical fluid technology,7 and spray freezedrying,8 a drying method that combines theatomization principle of spray drying and thefreeze drying process to produce powders withdifferent characters than those produced byeither conventional freeze drying or spray drying.Drying methods based on the principle of foamingand rapid evaporation of water (foam drying)were developed in the 1960s for preservation ofvaccines and bacteria.9–12 Now, these methods ormodifications of these methods are also beinginvestigated as alternate drying methods.13–15

Regardless of the drying method, the outcome is adry amorphous formulation of mainly the proteinand a stabilizer, hopefully with acceptable stabi-lity. However, systematic studies of the effect ofprocessing variations on storage stability arelacking. Because of differences in process, includ-ing vast differences in thermal history, physicaland chemical properties, including storage stabi-lity, are likely to vary with preparation method.For example, recent evidence has shown thatthermal history differences may impact degrada-tion rates in small molecule pharmaceuticalsystems.16,17

Physical and chemical stabilization upon longterm storage is of fundamental concern to thepharmaceutical scientist because of potentialalterations in immunogenicity, toxicity, and effi-cacy.18 The choice of stabilizer is usually limited toa small molecular mass sugar such as trehalose orsucrose, or a mixture of a small molecular masssugar and a polymeric excipient such as dex-tran.1,2,19 There are two different proposedmechanisms by which these excipients arebelieved to stabilize proteins in the amorphousstate, one is thermodynamic and the other iskinetic. The water substitution hypothesis pro-poses that stabilization results from nativestructure preservation by hydrogen bondingbetween the stabilizer and specific sites at thesurface of a protein, thereby substituting for thethermodynamic stabilization function ofwater lost

during drying.1,2,20 It seems unlikely that thiseffect, being a thermodynamic stabilizationmechanism, would vary dramatically with theprocess. The glass dynamics hypothesis proposesthat stabilization results from suppression ofglobal motions, also known as a-relaxations, thatmay be relevant to protein instability via theformation of a rigid, glassy matrix.21,22 a-relaxa-tions occur mainly due to translational androtational motions (rotation of the entiremolecule), and therefore strongly influence thediffusion of reactivemolecular species.23 Degrada-tion involving diffusion is strongly affected bychanges inmolecularmobility around glass transi-tion temperature.24 Since global relaxation isimpacted by thermalhistory,17 one expects processvariations that provide significantly differentthermal histories to impact stability.

Studies have been interpreted in terms of boththe ‘‘water substitute’’ concept and in terms ofglass dynamics. Cleland et al.25 explained the log-linear composition dependence of stability of amonoclonal antibody in terms of the water sub-stitution hypothesis. The authors observed thatthe physical stability of the lyophilizedmonoclonalantibody was correlated to the retention of sec-ondary structure in the solid state, and it wasassumed that the ‘‘water substitute’’ role of thesugar was responsible for stabilization of thenative structure. On the other hand, susceptibilityto aggregation has also been interpreted in termsof to the glass dynamics theory. Duddu et al.26

found a correlation between aggregation of amonoclonal antibody in lyophilized formulationsand molecular mobility, as measured by enthalpyrelaxation time. Inanother study,Yoshiokaet al.27

found that aggregation of bovine serum albumin(BSA) in the solid state increased as molecularmobility measured by proton nuclear magneticresonance (1H NMR) increased.

There are several examples in the literaturewhere neither the glass dynamics nor the watersubstitution hypothesis separately or together canfully explain stabilization.28 For example, nocorrelation was observed between stabilizationand matrix mobility in dried formulations ofinsulin and poly(vinyl)pyrrolidone,29 or in severalenzyme systems.30,31 Furthermore, in theseenzyme systems, excellent preservation of enzymeactivity was observed even above the glass transi-tion temperature (Tg).

30,31 This is surprising sinceabove Tg, global motions are relatively high andonewould expect high instability. Therefore, otherfactors seem to be responsible for stabilization.

1984 ABDUL-FATTAH ET AL.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 8, AUGUST 2007 DOI 10.1002/jps

Some of the problem may involve the assumptionthat it is the global relaxation process, which islinked to viscosity, which is most relevant to thedegradation dynamics of interest. It has beenproposed that dynamics on a smaller time scalethan global motions may be important for proteinstabilization in glasses.23,32,33 b-relaxations areexamples of such fast dynamics which generallyinvolve localized motion of specific portions of themolecule.34 They can occur over a broad range oftimescales, anywhere between vibrational androtational motion timescales.23,32–34 It is arguedthat b-relaxations influence diffusion of smallmolecules such as gases in systems of largermolecules, and possibly control small amplitudeproteinmotions that can result in protein aggrega-tion.23,32,33 Therefore, one would expect minimalcontribution of global motions in reactions(chemical or physical) controlled mainly by b-relaxations.24,34 Cicerone et al.23 showed that theaddition of small amounts of glycerol to anamorphous matrix of horse radish peroxidase(HRP) and poly(vinyl)pyrrolidone (PVP) resultedin increased stability ofHRP, even thoughTg of thesystemwas lowered. Improvement in stabilitywasconsistent with a decrease in mobility associatedwith b-relaxation (as measured by elastic in-coherent neutron scattering). A similar effect wasobserved with yeast alcohol dehydrogenase.Therefore, it appeared that glycerol, a plasticizerfor global motions, acted as an antiplasticizer forthe b-motions. Furthermore, if b-motions arerelevant to stability, suppressing such b-motionswould be the key to stabilization. Therefore, acomplete analysis of stability in terms of ‘‘glassdynamics’’ needs to also consider the possible roleof ‘‘fast dynamics.’’

Specific surface area is one material propertythat is likely to vary greatly with preparationmethod.35,36 A few studies have shown thatformulations with a high specific surface area(SSA) had inferior storage stability relative to thesame formulations with lower SSA prepared byother drying methods.35,36 For example, Saneet al.35 observed that the storage stability of afreeze dried antibody formulation was better thanthat of the same formulation spray dried. Proteinsecondary structure, as measured by Ramanspectroscopy, was the same in both formulationsbut SSA of the spray-dried formulation washigher. Carpenter et al.36 studied the storagestability of spray freeze dried and freeze driedrecombinant human interferon-g (rhIFN-g) for-mulation with trehalose in the presence and

absence of Tween 20. Second derivative FTIRspectra showed that protein secondary structurewas similar in all preparations. The spray freeze-dried formulation without Tween 20 had thehighest aggregation rate constant, consistent withits highest SSA and highest protein surfaceconcentration (as measured by electron spectro-scopy for chemical analysis, ESCA). The lyophi-lized formulation with Tween 20 had the lowestaggregation rate constant, consistent with its lowSSA and lowest protein surface concentration.None of these studies, however, provided amechanistic interpretation of these trends. It isbecoming well known36–43 that there may be asignificant separation of stabilizer from proteinduring spray drying. The surface region becomesrich in protein and the interior therefore mustbecome richer in stabilizer.36–43 The use of asurfactant has been shown to moderate theaccumulation of protein at the surface for spraydried samples.36,37,39–43 Such an effect, which iscompositional or chemical heterogeneity, refers tothe separation of chemical components duringdrying without phase separation.36,38–43 Develop-ment of compositional heterogeneity from a solu-tion initially uniform in composition requiresmolecular mobility, at least at some stage of theprocess. That is, one must have the molecularmobility to support themutual diffusion needed tocreate separation of components. Such heteroge-neity could arise during freezing for a freeze-drying process and during much of the dryingprocess for a spray dried material. Although thereis spatial variation in composition in freeze driedmaterials, it does appear that composition varia-tion in spray dried materials can be an order ofmagnitude larger39 thus, both SSA and proteinaccumulation at the surface are dependent ondryingmethod, and both factorswould be expectedto lead to inferior stability for spray dried powders.

Now the question is ‘‘What can be the relation-ship between chemical heterogeneity and storagestability?’’ It is becoming established that storagestability of proteins in amorphous pharmaceuti-cals show a log-linear dependence on composi-tion.25,28,44–46 Assuming that the measured orobserved rate constant for degradation (kobs.) is asummation of contribution of stability from bothsurface and bulk protein, we may write:

kobs ¼ kSFPS þ kBFPB ð1Þ

where kS is the rate constant for decomposition ofsurface protein, kB is the rate constant fordecomposition of bulk protein, FPS is the fraction

STABILITY OF A MONOCLONAL ANTIBODY 1985

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 96, NO. 8, AUGUST 2007

of the surface protein and FPB is the fraction ofbulk protein. Thus, in a formulation with a largeprotein fraction and little stabilizer fraction at thesurface, the contribution of kSFPS to degradationwill increase compared with an identical formula-tion with lesser protein surface coverage. More-over, if stability has a log-linear dependence oncomposition, simple numerical calculations47

show that the loss of stability in the regiondepleted of stabilizer is greater than the gain instability in the core region enriched in stabilizer,the effect being greater in importance for systemsrich in stabilizer. Thus, in general, the overalldegradation rate would be higher for thosesamples that have a higher degree of componentseparation. Evidence, therefore, suggests thatSSA and possible protein surface accumulationshould be considered as additional factors inprotein stabilization.

Themain objective of this study was to evaluatethe impact of variations in drying process onstorage stability for selected formulations. For-mulations contained varying levels of sucrose asstabilizer, with orwithout surfactant.We assessedthe impact of drying method on surface composi-tion, specific surface area and molecular mobility(both slow and fast dynamics), and the correlationof these physical properties with stability. Here, amonoclonal antibody was used as a model protein.

MATERIALS AND METHODS

Materials

Purified recombinant humanized monoclonalantibody against integrin a2Vb (rhuMAb) orMedi-522, an IgG1, (1.76% mass fraction aggre-gate, 0.25% mass fraction fragments) was sup-plied as 100 mg/mL stock solution in 10 mMphosphate buffer (potassium phosphate mono-basic and dibasic) pH 6.0 by the MedImmunePurification Development Group (Santa Clara,CA). Chemicals and excipients were all used assupplied. Sucrose powder was purchased fromPfanstiehl Laboratories (Waukegan, IL). PluronicF68 (a hydrophilic nonionic surfactant with anHLB value >25 and a CMC of &8 mmol/L) waspurchased from Sigma. 13 mm 2 mL and 20 mm5mL type I borosilicate clear tubing glass vials, aswell as V2 F451 13 mm and V10 F451 20 mmsingle-vent Flurotec1 (Daikyo Seiko) lyophiliza-tion stoppers were purchased from West Pharma-ceuticals (Lititz, PA).

Preparation and Drying of Medi-522 Solutions

Using a stock solution of 80% by mass sucrose, theratio of rhuMAb to sucrose was adjusted to 4:1,2:1, 1:4, and 5:95 by mass. The total solids contentin solution was adjusted for each drying methodfor optimum drying by diluting with de-ionizedwater. Final pH was 6.1� 0.1 and phosphatebuffer concentration was 1 mM after dilution.Solids content for solutions to be freeze dried andspray dried were 5% by mass. Solids content forsolutions to be foam dried were (11 to 12)% bymass for protein-rich formulations and 25–30%by mass for sucrose-rich formulations. Formula-tions with a higher mass ratio of protein tostabilizer are labeled ‘‘protein rich’’ (i.e., 4:1 and2:1), while formulations with a higher mass ratiofor stabilizer are labeled ‘‘stabilizer-rich’’ or‘‘sucrose-rich’’ (i.e., 1:4 and 5:95). The 4:1 and1:4 formulations of rhuMAb with sucrose werealso prepared with Pluronic F-68. The concentra-tion of surfactant was adjusted to 0.2% by mass inthe final solution (well above critical micelleconcentration). The ratio of surfactant to protein,however, differed according to solids content insolution. In 4:1 protein:Sucrose formulation withsurfactant, the ratio of surfactant-to-protein was1:20 in solutions to be spray and freeze driedand was 1:38 in solutions to be foam dried. In1:4 protein:Sucrose formulation with surfactant,the ratio of surfactant-to-protein was 1:5 insolutions to be spray and freeze dried and was1:25 in solutions to be foam dried.

Spray Drying

Solutions were spray dried with a Buchi 190 MiniSpray Drier (Flawil, Switzerland) operatinginside an enclosure to maintain low humidityand good temperature control. The Buchiatomizer nozzle was replaced by a customdesigned atomizer nozzle designed to producehigh pressure effervescent atomization, a tech-nology borrowed from the diesel industry.48,49 Inshort, the idea is to bubble a small amount ofthe atomizing gas (in this case, nitrogen) into thestream of solution to be spray dried insidethe atomizer nozzle. As the atomizing gas andsolution mix is discharged from the orifice of theatomizer, the gas phase expands, breaking theliquid into small droplets of uniform dropletsize distribution. Smaller droplet sizes resultfrom this method, compared to those frompressure atomization (i.e., the normal Buchi



nozzle). There are claims that this methodprovides a high process yield without significantreduction in product purity.49 The solution feedrate was 2 mL/min, the inlet temperature was878C, the outlet temperature ranged from 50 to608C, relative humidity in the chamber wasmaintained below 5%, the atomizing gaswas nitrogen and the atomizing pressure was12.4 MPa (1800 psi). Spray dried powders weresubjected to a post drying time of 30 min inthe collector tube at 60–638C to reduce moisturecontent.

Freeze Drying

Solutions were freeze-dried using a FTS SystemsLyostar II (Stone Ridge, NY). Solutions werelyophilized in 5 mL serum vials with a fill volumeof 1.5 mL. Solutions were placed on shelvesprecooled at 58C. Shelf temperature was cooledat a rate of 2.58C/min to �408C and was held for90 min to freeze the solutions. Shelf temperaturewas raised to �308C at a rate of 1.38C/min forcarrying out primary drying (below the collapsetemperature). Collapse temperatures were about�348C or greater, depending on formulation.Shelf temperature was held at �308C for 48 hduring primary drying while maintaining achamber pressure of 66 Pa (50 mTorr) (forsublimation of ice). Average product temperatureduring primary drying was �378C to �388C. Theend of primary drying was marked by a rise inproduct temperature above the shelf temperaturewith a concurrent decrease in dew point to aconstant value. For secondary drying (evapora-tion of water in the amorphous phase), shelftemperature was heated to 108C at a rate of 0.18C/min. Temperature was maintained at 108C for280 min. Pressure during secondary drying wasmaintained at 66 Pa (50 mTorr). While 108C is nota typical secondary drying temperature for pro-teins, it is our experience that using highertemperatures in secondary drying usually leadsto very low moisture content in the final product(<0.5% by mass). This temperature was used inour studies to control the final moisture content tothe desired range for all processes, &1–2% bymass.

Annealing Followed by Freeze Drying

Some rhuMAb solutions were annealed in thefrozen state prior to primary drying. The purposeof this step was to produce samples of lower

specific surface area than the regular freezedrying cycle. This variation also allowed a test ofthe concept, which states that maintaining theprotein above a glass transition temperatureduring the process leads to instability. Annealingwas performed at �88C for 48 h (above glasstransition temperature of the freeze concen-trate—Tg0. Highest Tg0 was ��108C). Solutionswere re-frozen to �408C after annealing. Condi-tions for primary and secondary were the same asunder ‘‘Freeze drying.’’

Cakes produced upon freeze-drying (with andwithout annealing) were elegant with excellentretention of cake structure.

Foam Drying

Solutions were foam dried using a Virtis Genesisfreeze dryer equipped with a condenser model25EL (Gardiner, NY). Solutions were foam driedin 2 mL serum vials with a fill volume of 0.3 mL.Vials containing solutions were placed on shelvesat 158C and kept for 15 min to equilibrate.Foaming was initiated by reducing the chamberpressure to 66 Pa (50 mTorr) at a shelf tempera-ture of 158C. Chamber pressure used is signifi-cantly lower than vapor pressure of water atthe prevailing temperature to obtain foaming.Product temperatures initially decreased to&�20 to 58C, depending on the formulationcomposition, and remained at this temperatureduring most of primary drying. None of thesolutions froze, presumably due to the very highsolute content. The drying process is considered tobe composed of two drying phases: primary dryingand secondary drying. Primary drying was doneat a shelf temperature of 158C and a chamberpressure of 66 Pa (50 mTorr) for 1 h. This step is amuch shorter step than primary drying in afreeze-drying cycle and water is removed byevaporation. Secondary drying, a step to removeremaining moisture in amorphous phase, wasdone by raising shelf temperature to 338C at arate of 18C/min and maintained at that tempera-ture and at 33 Pa (25 mTorr) for 72 h. Theunusually long secondary drying time requiredfor foam drying is probably attributable to thevery low specific surface area of the foams (resultspresented later).

Adjusting Moisture Content in Dried Powders

Because the types of powders produced werequite different in their secondary drying



characteristics, the drying processes led to largevariations in moisture content (results notshown). In order not to convolute the effects ofmoisture variation on storage stability with theeffects due to variations in process and formula-tion, moisture contents in all formulations wereadjusted to be in the range of &1–2% by mass.Variability of moisture content within this rangewas associated with minimal variations in stabi-lity of IgG1 (results not shown,28). Powders withless than 1% by mass moisture, as measured byKarl Fischer titration, were re-hydrated in 10%relative humidity for 30–45 min and sealed undernitrogen. Powders with more than 2% by massmoisture were subjected to further vacuum dry-ing. Vacuum drying was performed in a VirtisGenesis freeze dryer equipped with a condensermodel 25EL (Gardiner, NY) at a shelf tempera-ture of 358C and a chamber pressure of 33 Pa(25 mTorr) for 1 h. Vials were sealed undernitrogen at the end of the drying cycle. Withsucrose-rich foams (where initial moisture con-tent was >2% by mass), additional drying timesfor up to 48 h were employed due to the very lowSSA of the dried foams.

Karl Fischer Moisture Determination

Residual moisture content of all formulations wasmeasured by direct injection using coulometricKarl Fischer titration (Denver Instrument Com-pany, Denver, CO). Powders were weighed andfilled into vials in a glove bag where a low relativehumidity (RH) was maintained (<2%) by flushingwith dry nitrogen. Powders were dissolved/dispersed in 2 mL of low moisture formamideand 0.5 mL of the solution was injected. Blankcorrections were applied. Standard deviation fromreplicate measurements was not more than 0.1%H2O.

Scanning Electron Microscopy (SEM) studies

SEM was utilized to characterize morphology ofselected spray-dried, freeze-dried, and foam-driedpreparations. Samples were mounted on siliconwafers mounted on top of double-sided carbontape on an aluminum SEM stub. The sample wasthen sputter coated in a Denton sputter coater for60–90 s at 100 Pa (75 mTorr) and 42 mA withgold:palladium. Images were taken with a PhilipsXL30 ESEM operated in high vacuum modeusing an Everhart-Thornley detector to capturesecondary electrons for the image composition.

Accelerating voltage was set at 20 kV using aLaB6 source. Working distance was 5–6 mm.

Modulated Differential ScanningCalorimetry (MDSC) Studies

MDSC studies were performed to determine theglass transition temperature (Tg) of the differentdried formulations with and without surfactant toaid in proper design of storage stability studiesand to characterize thermal behavior of thedifferent preparations. A Q-1000 DifferentialScanning Calorimeter equipped with a DSCrefrigerated cooling system, TA instruments,Inc. (New Castle, DE) was used. The powderedsample (5–15 mg) was filled and pressed intocompact pellets, placed in aluminum DSC pansand hermetically sealed using a sample encapsu-lation press. The entire process was performed ina glove box, with RH maintained under 2%.Nitrogen was used as a purge gas for MDSC at30 cm3/min. All MDSC measurements were donein triplicate or more. The MDSC protocol was asfollows:

� Equilibrate at 08C.� Modulate þ/� 18C every 120 s.� Ramp 28C/min to 1008C.

Powder Density Measurements

The density (r) (g/cm3) of dried powders wasmeasured with nitrogen pycnometry (Micro-meritics AccuPyc 1330 Gas Pycnometer,Norcross, GA). Powders were weighed into 1-cm3

sample aluminum sample cells and the cells weremounted into the analysis chamber. The numberof purges was 10, the purge fill pressure wasadjusted to 134 kPa (19.5 psig), the number ofruns was adjusted to 5, the run fill pressure wasadjusted to 134 kPa and the equilibration ratewas set to 34 Pa/min (0.005 psig /min). The gaspyconometer was tested for precision, accuracy,and inter-day variability using crystallinesodium chloride. The published r value forcrystalline NaCl is 2.17 g/cm3.50 Our measure-ments gave a r value of 2.182� 0.026 g/cm3 (n¼ 5replicates on 5 separate days). Additionally,reproducibility (inter-vial variability) of r mea-surements on four different amorphous samples(different sucrose-based formulations labeledA, B, C, D) gave 1.039� 0.008 g/cm3 (A),1.590� 0.019 g/cm3 (B), 1.433� 0.004 g/cm3 (C),and 1.570� 0.011 g/cm3 (D).



Specific Surface Area (SSA) Measurement

BET (Brunauer Emmett Teller) Specific surfacearea (SSA) analysis was performed using aMicromeritics, FlowSorb II 2300 BET surfacearea analyzer (Norcross, GA). Sample size wasat least 100 mg. Powder samples were degassedfor at least 3 h in the Flow Prep oven at 258C usingkrypton purge prior to measurement. Singlepoint calibration was performed prior to takingsurface area measurements using 0.1 mol Kr gas.Percent relative standard deviation (%RSD) forSSA measurements (replicate samples) rangedbetween 1% and 7% for preparations with SSA�0.5 m2/g, but was higher for preparations withvery low SSA of �0.1 m2/g (Tab. 2).

Electron Spectroscopy for ChemicalAnalysis (ESCA) Studies

ESCA was used to probe elemental composition ofthe surfaces of dried formulations, with ananalysis depth of &50 A. Survey spectra werecollected with a VG ESCALAB MK II seriesSpectrometer (pass energy¼ 60 eV). An AlKa

radiation (emission current of 11 kV anodevoltage and 34 mA emission current) from AlKa

X-ray source was directed onto a circular region ofapproximate diameter of 3 mm. Pressure wasmaintained in analysis chamber at less than 10�5

Pa (<10�8 Torr). Measurements were done ontriplicate samples. The peak areas from thespectra were converted to elemental atomicconcentrations on the surface of each sample.%RSD for ESCA measurements (replicatesamples) did not exceed 4% for protein-richsystems and ranged between 4% and 13.5% forsucrose-rich systems (Tab. 2). Details of dataanalysis are described elsewhere.47,51

Fourier Transform Infra Red (FT-IR)Spectroscopy Studies

FT-IR spectroscopy studies were performed usinga Nicolet Magna IR 760 Spectrometer (Thermo-Nicolet, Madison, WI). Studies were performedon both rhuMAb in solution (i.e., native conforma-tion) &10–15 mg/mL and in all dried powdersaccording to procedures described in the litera-ture.20,52–56 A total of 256 scans and a resolutionof 4 cm�1 were used for each spectrum. Thescanning range was 4000 to 400 cm�1. Thespectrum for native, aqueous rhuMAb (used as a

control) was obtained by placing a 15 mg/mLsolution between 2 CaF2 windows separated by a6 mm Mylar spacer. The spectrum of rhuMAb insolution was corrected by subtracting the spectraof liquid water and water vapor as describedelsewhere.57 The solid samples were preparedby pressing a ground homogeneous mixture of100 mg KBr with 2–3 mg of the dried proteinformulation into a pellet (disk) using a stainlesssteel die. Effects of water vapor and carbondioxide in the chamber were eliminated bysubtracting the background spectrum. The secondderivative IR spectra of the samples wereobtained using Omnic software version 6.0a(Nicolet, Madison, WI). The amide I region1600 to 1700 cm�1 of the spectra (smoothed byseven point smoothing) were baseline correctedand area normalized, then overlaid to comparethe spectra. Conformational changes of rhuMAbin dried samples were characterized by measur-ing the magnitude of change in secondarystructure of rhuMAb in amide I region relativeto that in solution. To quantitatively comparebetween the spectra of the different driedpreparations, the b-sheet ratio was calculatedfrom the two bands at 1690 cm�1 and 1640 cm�1

that presumably correspond to different types ofb-sheet configurations.56,58 Additionally, thespectral correlation coefficient (r) was determinedto quantify spectral similarity between spectraof dried formulations and that of aqueoussolution,52,58 defined by the following formula:

r ¼P

ðxiyiÞffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPx2i

Py2i

q ð2Þ

where xi and yi are the intensity of the spectra forsolid samples and the spectra of protein insolution (native structure) at wave number (i),respectively. %RSD for determination of r fromFTIR studies (replicate samples) did not exceed2.4% (Tab. 1).

Molecular Mobility

Measurement of Structural Relaxation Time byIsothermal Microcalorimetry

Isothermal microcalorimetry studies wereperformed to measure directly the rate ofenthalpy relaxation during aging experi-ments.59,60 A thermal activity monitor (TAM)(Thermometric, Sweden), was employed to mea-sure the rate of enthalpy relaxation of some



representative spray dried, foam dried, and freezedried formulations. TAM experiments were per-formed at 508C for at least 60 h. Samples wereloaded into stainless steel cylinders in a dry bagpurged with dry nitrogen to minimize moistureuptake during loading. Glycine of roughly thesame mass as the sample was used as an inertreference for the solid samples to minimize noise.The sample and reference containers were low-ered to a thermal equilibrium position in thecalorimeter and allowed to equilibrate with thecalorimeter for about 30 min. The sample andreference were then slowly lowered into themeasuring zone of the calorimeter and powerwas recorded (mW/g) as a function of time (h).Structural relaxation times were determined forthe tested formulations from the power-time data.Structural relaxation times are reported here as

ln tb (ln of stretched time) since tb is a more robustmeasure of a relaxation time constant.59,60 Detailsof data analysis are described in literature.59,60

Fast Dynamics from IncoherentElastic Neutron Scattering

Elastic incoherent neutron scattering measure-ments were performed on seven selected prepara-tions at the National Institute of Standards andTechnology (NIST) Center for Neutron Researchon the High Flux Backscattering Spectrometer(HFBS) on the NG2 beam line.61 A beam of coldneutrons with a kinetic energy of 2.08 meV�50 meV and a wavelength of 6.27 A was focusedonto the sample of interest. The accessiblemomentum transfer (Q) range was 0.25–1.75 A�1.

Elastic scattering intensity was recorded as a

Table 1. FTIR Studies: Measures of ‘‘Native-Like’’ Structure in the Solid State

Formulation TreatmentCorrelation

Coefficient (r)**b-Sheet

Ratio�SD*

Solution (Reference) 0.28

Medi522:Sucrose 4:1 LYO 0.79 0.81� 0.02ANNLYO 0.80 0.72� 0.02Spray dried 0.86 0.60� 0.04Foam dried 0.82 0.68� 0.00

Medi522:Sucrose 4:1(with Pluronic F68)

LYO 0.82 0.68� 0.02

ANNLYO 0.82 0.72� 0.02Spray dried 0.89 0.50� 0.02Foam dried 0.83 0.62� 0.02

Medi522:Sucrose 2:1 Spray dried 0.88 0.49� 0.01Foam dried 0.84 0.61� 0.01



LYO 0.92 0.43� 0.01

ANNLYO 0.90 0.48� 0.01Spray dried 0.95 0.38� 0.01Foam dried 0.85 0.61� 0.03


A high correlation coefficient and/or a low b-sheet ratio means high native-like structurecontent.

*SD, Standard deviation for multiple measurements.**Standard deviation for r values did not exceed 0.02.Spectral correlation coefficients and b-sheet ratios, defined by

�-sheet ratio ¼Depth of band ca:1690 cm�1

Depth of band ca:1640 cm�1



function of Q while the sample was heated at0.5 K/min from 40 K to 333 K (below the glasstransition temperature of the tested formula-tions). The Q-dependence of the incoherent elasticscattering Iinc.(Q, o¼ 0) was analyzed in terms ofthe Debye-Waller factor (a harmonic oscillatormodel), and the hydrogen-weighted mean-squareatomic displacement hm2i was determined atdifferent temperatures. Here, we report hm2i, inunits of A2, at 408C and 508C. In a numberof cases, it has been demonstrated that theamplitude of the mean square atomic displace-ments scales inversely with log(t) where t is thecharacteristic time for protein degradation inglassy systems.32,33

The HFBS accesses wave vectors covering0.25 A�1>Q> 1.75 A�1 (i.e., length scale range of3.6 A to 25 A).Motions seenat lowerQ (length scalerange of 3.6 A to 9 A) are not well-fit by a harmonicapproximation. The concept of atomic motionsbeing represented by a harmonic oscillator is

better founded in the high Q range (length scaleranges >9 A). Scattering events that arise frommotions occurring on a time scale of approximately5 nanoseconds are preferentially detected by theHFBS instrument. Only elastic events are directlyobserved, but total scattering is conserved, soinelastic scattering intensities are easily inferred.Fundamentals of this technique anddetails of dataanalysis are described in the literature.23,32,33,61

Stability Studies and Assays

HPLC Assay Methodology

Total covalent and noncovalent IgG1 solubleaggregates were measured by size exclusionchromatography (SEC). The LC system consistedof a HP1100 HPLC pump equipped with a UVmultiple wavelength detector, an automaticsampling system and an in-line degasser. The LCcolumn used was a TosoHaas TSK Gel Analytical

Table 2. Surface Properties of Powders and Density Measurements

Formulation TreatmentPowder

Density (g/cm3)SSA

(m2/g)�SDProtein Surface

Coverage (%w/w)�SD

Estimated % ofTotal Protein onSurface�SE*

Medi522:Sucrose 4:1 LYO 1.338 1.22� 0.02 87.8� 0.5 1.0� 0.01ANNLYO ** 1.64� 0.03 89.3� 2.4 1.2� 0.01Spray dried 1.385 3.04� 0.13 78.5� 1.7 2.1� 0.08Foam dried 1.352 1.46� 0.03 87.5� 0.1 1.1� 0.02


LYO 1.349 1.03� 0.01 78.8� 0.7 0.7� 0.01ANNLYO 1.528 1.12� 0.04 67.9� 1.3 0.8� 0.02Spray dried ** 3.94� 0.18 47.1� 0.0 1.7� 0.06Foam dried 1.337 1.41� 0.06 71.3� 2.8 0.9� 0.04

Medi522:Sucrose 2:1 Spray dried 1.341 2.53� 0.02 70.7� 0.9 1.8� 0.02Foam dried 1.384 1.10� 0.03 80.3� 1.8 0.9� 0.03

Medi522:Sucrose 1:4 LYO 1.436 1.33� 0.08 45.8� 2.0 2.2� 0.12ANNLYO 1.475 0.60� 0.03 52.0� 2.1 1.2� 0.06Spray dried 1.470 2.06� 0.11 60.7� 2.7 4.6� 0.24Foam dried ** 0.04� 0.01 60.1� 2.3 0.1� 0.02


LYO 1.577 1.21� 0.02 16.3� 0.8 0.8� 0.03ANNLYO ** 0.81� 0.06 21.3� 2.9 0.7� 0.08Spray dried 1.423 1.96� 0.07 12.0� 0.7 0.9� 0.05Foam dried ** 0.04� 0.01 23.2� 1.7 0.03� 0.01

Medi522:Sucrose 5:95 LYO 1.573 1.25� 0.06 33.9� 3.9 6.4� 0.58ANNLYO 1.598 0.50� 0.03 52.3� 3.7 4.2� 0.27Spray dried 1.494 1.68� 0.05 51.8� 1.0 13� 0.34Foam dried 1.577 0.17� 0.02 54.8� 2.0 1.4� 0.26

All values based on replicatemeasurements (n¼2–3). An estimate of the total protein surface accumulation on the powder surfaceis also given, calculated from density, SSA and protein surface coverage (as described in the text).

*SE, standard error calculated from propagation of errors.**For regions marked, values not determined due to insufficient sample availability. However, rough estimates were made for

densities of these powders for calculation of estimated % of total protein on the surface, based on average density values for thedifferent other preparations of the same formulation.



Column G3000SWXL (5 m, 7.8 mm� 300 cm I.D.)preceded by a TSK Gel Guard Column SWXLcolumn (7 m, 6 mm� 4 cm I.D.). The mobile phaseconsisted of 0.1 mol/L dibasic sodium phosphate,0.1 mol/L sodium sulfate, and 0.05% by masssodium azide. pHwas adjusted to 6.8 with 1NHCl.The method was run isocratic at 0.8 mL/min. Thewavelength for detection was 280 nm, the columntemperature was ambient, and the injectionvolume was 25 mL. Samples were run in duplicate.The SEC method was tested for reproducibility(inter-vial variability and inter-day variability)and precision (formeasuring aggregate formation).%RSD for precision test (multiple injections fromthe same vial) was not more than 2.7%. %RSD forinter-vial variability was not more than 3%.Additionally, inter-day variability was assessedby storing two different samples (treated as controlsamples) after initial SEC assay at �208C for9 months after which the samples were re-assayed(samples were labeled LYOSB and ANNSE).Assay results for initial time points for thecontrol samples were 1.74� 0.03 (#LYOSB)and 1.77� 0.04 (#ANNSE). Assay results after9 months at �208C for the control samples were1.70� 0.04 (#LYOSB) and 1.66� 0.02 (#ANNSE).

Stability Studies and Characterizationof Degradation Rates

Aggregate levels (% aggregates) in all formula-tions were measured by the SEC-HPLC methodinitially and at three different time points duringa 6-month storage stability study. Stabilitystudies were conducted in stability ovens at40� 18C and at 50� 18C. Two to three vials ofeach preparation were assayed at each time point.

RESULTS AND DISCUSSION

Moisture Content and Glass Transition

Final adjusted moisture level in all dried pre-parations was in the range 1–2% by mass, asmeasured by Karl Fischer titration. None of thedried preparations showed birefringence whenexamined using polarized light microscopyindicating purely amorphous character.

The drying method did not have a significantimpact on glass transition behavior. In protein-rich formulations (i.e., 2:1 and 4:1 IgG1:Sucrose),no glass transition event could be detected byMDSC, consistent with what was observed byothers.28,62,63 In sucrose-rich systems, two major

transitions were observed in the presence andabsence of surfactant; a glass transition eventwithan associated weak enthalpy recovery followed byan exothermic transition most probably due tocrystallization (Fig. 1). Glass transition tempera-tures (midpoint) (Tg) were found in the range&55–668C, the exact value depending on formula-tion and preparation method and averaged61.0� 4.58C for all formulations. Differences inheat capacity change at Tg (DCP) were minimalbetween different drying processes, except forfoams. DCP in the foam dried preparationsaveraged 0.46� 0.05 J/(g.8C), but for all otherpreparations the average was 0.64� 0.06 J/(g.8C).The lower DCp at Tg would imply a higher heatcapacity in the glassy state for the foam-driedmaterial since theheat capacity aboveTg shouldbethe same for all materials. At the present time, wedo not have an explanation for this observed effect.

The sucrose-rich spray dried preparation withsurfactant showed multiple transitions (Fig. 2)indicative of phase separation, one of the phasesbeing a pluronic-rich phase. The first two transi-tions coincided with the same transitions of purepluronic (MDSC thermogram of pure pluronic notshown), followed by a Tg with significantly higherDCP. In the second heat cycle, only one transition(the main glass transition event seen in the1st heat cycle) was observed. Although MDSCthermogram of a protein-rich spray dried systemof 4:1 IgG1:Sucrose did not show a glass transitionfor a sucrose/protein phase, it did show anendothermic transition possibly originating frompluronic.

Effect of Drying Method (Preparation Method) andFormulation on Particle Morphology (SEM)

In this section and those that follow, the termformulation is used to denote a sample of givenchemical composition in the solid, and ‘‘prepara-tion’’ is used to denote a given drying method.

Powders displayed different morphologies,depending mainly on the method of drying andon the disaccharide content. Freeze dried powderswere composed of thin plates (Fig. 3A and B).Surfactant-free foam dried powders with highsucrose content were composed of very large thickplates (Fig. 3C). Morphology of the same powderswith surfactant was similar (figures not shown).Surfactant-free foam dried powders with highprotein content, on the other hand, were composedof much thinner and smaller flakes (Fig. 3D),suggesting a larger SSA. Morphology of the same



Figure 1. A typical MDSC thermogram for a dried sucrose-rich IgG1 formulation(freeze dried, spray dried or foam dried) showing three signals (total heat flow, reversingheat flow, and nonreversing heat flow signals).

Figure 2. Reversing heat flow signal for spray dried formulations with surfactant: (I)Spray dried 1:4 IgG1:Sucrose with pluronic—1st heat cycle shows two endothermicevents corresponding to surfactant before themainTg. The two endothermic events wereabsent in the 2nd heat cycle. (II) Spray dried 4:1 IgG1:Sucrose with pluronic—1st heatcycle shows an endothermic event attributable to surfactant.



Figure 3. Morphology of dried formulations as observed bySEM (A), A surfactant-freesaccharide-rich freeze dried preparation (B), A surfactant-free saccharide-rich freezedriedpreparationannealed in the frozen state (C),A surfactant-free saccharide-rich foam(D),A surfactant-freeprotein-rich foam(E),A surfactant-free saccharide-rich spraydriedpreparation (F), A surfactant-free protein-rich spraydriedpreparation.Magnification foreach image is given on the SEM panel.



powders with surfactant was similar (figures notshown). The influence of disaccharide content onparticlemorphologywas also apparentwith spray-dried powders. Surfactant-free, sucrose-rich spraydried particles were spherical with no dimples.Larger particleswerewrinkledandmanyparticleswere fused together (Fig. 3E). Particle fusionincreased dramatically with a large increase insucrose content. This ‘‘fusion’’ phenomenon maycontribute to a reduction in the measured SSA ofthe powder. Morphology of the correspondingpowders (1:4 IgG1:Sucrose) with surfactant wassimilar (figures not shown). Surfactant-free, pro-tein-rich spray dried particles were spherical withan abundance of dimples (Fig. 3F). Morphology ofthese powders did not change upon the addition ofsurfactant (figures not shown).

Effect of Drying Method and Formulation onProtein Secondary Structure in the Solid State

Figure 4A and B shows examples of FT-IR spectraof IgG1 in different dried formulations comparedto the spectrum of IgG1 (control) in aqueoussolution. All data represent area normalized baseline corrected second derivative FT-IR spectra inthe amide I region 1600–1700 cm�1. The amide Iband arises primarily due to the C––O stretchingvibration of the peptide bonds of the backbonestructure, which is sensitive to changes inprotein conformation and secondary structure.In solution, the FTIR spectrum of the rhMAb inthe amide I region is dominated by a strong bandat 1639 cm�1, the marker band for IgGs arisingfrom the native b structure.25,28,45,58 Anotherband at 1690 cm�1 is also due to the b structure,while bands at 1665 and 1680 cm�1 are due toturn structure.58 Upon drying IgG solutions,conformational changes (i.e., perturbation insecondary structure) occur, manifested by adecrease in the intensity and broadening of theband at 1639 cm�1, and a shift in the highfrequency b-sheet band to &1693 cm�1 accom-panied by an increase in the area of thisband.1,2,20,36,58,64 No intermolecular b-sheet struc-ture at 1622 cm�1 was observed, consistent withthe findings of Cleland et al.25 Secondary struc-ture by FTIR is significant because formulationswith more native-like structure are expected tohave better storage stability than formulationswith less native-like structure.1,2,65

A good correlation was observed between pro-tein secondary structure and composition of theformulation. As expected,25,28 preparations with

higher sucrose levels showed less perturbation insecondary structure, as evidenced by ‘‘morenative’’ appearing spectra (Fig. 4), as well ashigher spectral correlation coefficient values(Tab. 1), regardless of the drying method (Tab. 1).

The b-sheet ratio was more sensitive tochanges caused by variation in drying methodthan were the correlation coefficients (Tab. 1). Ahigh b-sheet ratio is indicative of less nativestructure, and vice versa. Therefore, in ourdiscussion of secondary structure, we will focuson b-sheet ratio to compare small structuraldifferences due to the drying method. The impactof drying method on structure can be summarizedas follows:

(1) Regardless of the composition and of thepresence of surfactant, spray dried prepara-tions showed more native-like structurethan the other preparations.

(2) Regardless of the composition and of thepresence of surfactant, annealing in thefrozen state did not impact the secondarystructure of freeze dried formulations, sug-gesting that even 48 h of annealing was notsufficient for a significant fraction of theprotein to undergo unfolding due to theice-water interface.

(3) In a protein-rich formulation of 4:1 IgG1:Sucrose, more perturbation of secondarystructure was observed with the freeze-dried preparations, as evidenced by higherb-sheet ratios. Upon addition of surfactantto this system, slight improvements in thesecondary structure were observed, sug-gesting the surfactant offered some protec-tive effect against the ice interface. A moresignificant improvement was observed,however, in the secondary structure of thespray dried preparation. However, giventhe relatively low fraction of protein on thesurface and the limited sensitivity of FTIRto detect minor populations, it is debatablewhether the observed correlation between1690:1639 cm�1 ratio and storage stabilitycan be interpreted wholly in terms ofstructural alteration at the ice:aqueousinterface or air:aquous interface. Detectionof conformationally altered subpopulationin the surface region would likely requirethat the surface region be two to threemolecules deep rather than one moleculedeep as assumed by our reported calculationof ‘‘%Total protein on surface’’ (Tab. 2).



While possible, we have no evidence for asurface region several molecules deep.

(4) In a sucrose-rich formulation of 1:4 IgG1:Sucrose, the foam dried preparation showedthe greatest perturbation in secondary

structure. Upon addition of surfactant tothis system, little improvement in thesecondary structure was observed and theorder for the conformational changesremained the same (Foam dried>All freeze

Figure 4. SecondderivativeFT-IR spectra in the amide I region for preparations of the1:4 IgG1:Sucrose formulation (Fig. 3A) and the 4:1 IgG1:Sucrose formulation (Fig. 3B).



dried>Spray dried). In a 5:95 IgG1:Sucrosesystem, differences between freeze driedand foam dried preparations were small.

Effect of Drying Method and Formulation onSpecific Surface Area (SSA), Surface Composition,and Powder Density

Specific Surface Area (SSA)

An effect of sucrose level on SSA was observedonly with foams and spray dried powders:increasing SSA was associated with an increasein IgG1 level (Tab. 2). It is possible that particlefusion in spray dried powders (Fig. 3E) contrib-uted to the reduction in the SSA with increasingsucrose level.

Drying method also had a large impact on SSA(Tab. 2). Regardless of the composition and of thepresence of surfactant, spray dried preparationshad the largest SSA.

We note that contrary to expectations, anannealing treatment during freezing did notsignificantly reduce SSA in a protein-rich systemof 4:1 IgG1:Sucrose. While there is no report in theliterature on the effect of annealing on the SSA ofprotein-rich formulations, there are reports on SSAchanges upon annealing stabilizer-rich systems66–69

ling normally allows larger ice crystals to grow atthe expense of small ones in a process calledOstwald ripening, thereby resulting in a productwith a significantly lower SSA than the sameformulation with no annealing treatment66–69 inprotein-richsystemswhere foamdriedpowdershadSSA similar to freeze dried powders; the SSA offoam dried stabilizer-rich powders was extremelylow compared to the other preparations. Annealingin the frozen state prior to primary drying resultedin a lowering of the SSA of the final freeze-driedpowder (Tab. 2). Small changes were observed inSSA of all preparations of the 1:4 IgG1:Sucrosesystemuponaddition of surfactant, and the order ofdecrease in SSA remained the same.

Protein Surface Coverage

From an analysis of the relative amount of thedifferent elements in the pure components and inamulti-component powder, one can determine themolecular composition in the surface compositionregion (i.e., surface coverage for each compo-nent).37,41,42,47,51 ESCA measurements showedthat pure IgG1 is composed of approximately70% by mass C, 15.2% by mass N, and 14.8 mass% O. Both sucrose and pluronic contain only

C and O. The N peak, therefore, is a directmeasure of the presence of IgG1 in the outer 50 Asurface region,36,51 and hence it is possible todetermine the protein surface coverage, alsotermed protein surface concentration.37,41,42,51

Surfactant-free formulations with 4:1, 2:1, 1:4,and 5:95 IgG1:Sucrose are loaded with 80, 66.7,20, and 5% by mass IgG1, respectively. Thus, itfollows that the protein surface coverage is &80,67, 20, and 5 mass % IgG1 if IgG1 werehomogeneously distributed in the correspondingformulations. Similarly, the surface layer of 4:1and 1:4 IgG1:Sucrose formulations with surfac-tant should contain approximately 77% by massand 19% by mass IgG1, respectively.

Protein surface coveragewas impacted stronglyby formulation composition and the presence ofsurfactant (Tab. 2). As expected and as is consis-tent with literature, protein surface coverageincreased as protein weight fractionincreased,37–39,51 irrespective of the dryingmethod. Protein surface coverage significantlydecreased upon the addition of surfactant, similarto what was observed previously.40–42 Addition-ally, protein surface coverage showed either asurface excess (i.e., more protein than would beexpected for a homogeneous sample) or surfacedeficiency (i.e., less protein than would be expect-ed for a homogeneous sample). Therefore, itappears that component separation occurred dur-ing drying, and that the most surface activecomponent in solution (assumed to be protein insurfactant-free formulations, and surfactant informulations containing pluronic) enriched thesurface of the drying solutions and hence the driedpowders.37,39–42,70,71

Drying method did not have a strong impact onprotein surface coverage. The main impact ofdrying method could be seen when comparingsucrose-rich freeze dried formulations annealed inthe frozen state versus no annealing treatment.Annealing in the frozen state resulted in anincrease in protein surface coverage. Similarresults were obtained by Fureby et al.39

Total Protein Surface Accumulation

Total protein surface accumulation reflects howmuch of the total protein (mass fraction) addedto the formulation has accumulated at thesurface of the powders (Tab. 2). Total proteinsurface accumulation is an important parametersince this environment is potentially a morereactive state for the protein. ESCA studies



have traditionally only reported protein surfacecoverage. Surface coverage is a good indicator ofthe degree of homogeneity but does not measurethe fraction of the protein in the sample, whichhas accumulated at the surface. For example in aformulation with high protein surface coveragebut very low SSA, the fraction of protein at thesurface would be very small and the stabilityimplications would therefore be small.

Total protein surface accumulation wasimpacted mainly by drying method. Regardless ofthe composition and of the presence of surfactant,spray dried preparations had the largest totalprotein surface accumulation, largely as a result oftheir high SSA. In protein-rich systems (with andwithout surfactant), differences in total proteinsurface accumulation between powders weremodest. Drying method had more of an impact ontotal protein surface accumulation in surfactant-free sucrose-rich formulations, where total proteinsurface accumulation decreased in the followingorder: Spray dried>Freeze dried powder (noannealing treatment)>Freeze dried powderannealed prior to lyophilization >> Foam driedpowder. Unlike protein-rich formulations, thetotal protein surface accumulation was lowestwith sucrose-rich foam dried powders, as a directresult of their very lowSSA. Addition of surfactantin the 1:4 IgG1:Sucrose preparations decreasedtotal protein surface accumulation. Annealingsucrose-rich systems in the frozen state prior toprimary drying resulted in a lower SSA and lowertotal protein surface accumulation, even thoughannealing caused an increase in protein surfacecoverage.

Effect of Drying Method and Formulationon Molecular Mobility

Relaxation times (tb at 508C) and the amplitude offast motions (hm2i at 408C and 508C—tempera-tures for storage stability studies) for someselected formulations are shown in Table 3.Formulation did not have a measurable impacton global (slow) motions in the systems studied,but did have a significant impact on the amplitudeof local motions (i.e., fast dynamics) (Tab. 3).Relaxation time—tb (h)—may be regarded asproportional to the inverse of molecular mobilityfor global motions.22,72 The higher the tb value,the lower the molecular mobility and vice versa.The neutron scattering parameter, hm2i (in unitsof A2), is a measure of the amplitude of fastmotions. The higher this value is, the greater

the mobility for motions on the nanosecond timescale.

Global Motions

Figure 5 shows the power data from isothermalmicrocalorimetry experiments at 508C as a func-tion of time for preparations of the 1:4 IgG1:Su-crose formulation. Note that with the saccharide-rich foam, an initial endothermic responsewas observed that dissipated after approximately4–5 h followed by a normal exothermic responsethat decayed with time. The same was observedwith the 5:95 IgG1:Sucrose foam. We are notcertain of the origin of this initial endothermicresponse and did not include this range of datafor estimation of tb. Data were analyzed startingat the 5 h time point for sucrose-rich foams.Additionally, different portions of the exothermicresponse starting at several time points�5 h wereanalyzed to determine consistency of tb valuesobtained by the analysis. tb values were essen-tially the same for all data sets. For example, witha 5:95 IgG1:Sucrose foam: tb value from t¼ 5–60 hwas 14.9 h, from t¼ 8–60 h was 14.7 h, fromt¼ 10–60 h was 15.5 h and from t¼ 15–60 h was15.2 h. The endothermic response was notobserved with protein-rich foams.

Within the uncertainties in the data and giventhe composition range explored, composition didnot have a significant impact on tb values for spraydried and freeze dried formulations (P> 0.1 fromBonferroni test). Chang et al.28 reported a max-imum value for tb at &a 1:1 ratio of sucrose toprotein (recombinanthuman serumalbuminandamonoclonal antibody). However, we did not workwith sugar-protein compositions in the middle ofthe composition spectrum, so it is possible that thelack of composition effect in our data was simplydue to the fact that we did not have data in therange of the maximum. On the other hand,composition did impact tb values for foams.Sucrose-rich foams showed significantly higher tb

values than the protein-rich foam.The impact of drying method on tb values

depended on composition. No significant differ-ences were observed between tb values of thedifferent preparations of 4:1 IgG1:Sucrose. Insucrose-rich systems, however, the foams had thehighest tb values, followed by spray-dried powdersand were lowest with the freeze-dried prepara-tions. Annealing during freezing did not impactthe tb values for the 1:4 IgG1:Sucrose freeze-driedpreparation. In the presence of surfactant, the



1:4 IgG1:Sucrose foamhad lower tbvalues than thespray dried preparation, an unexpected observa-tion. However, we note that DSC evidence forphase separation was observed for the latterpreparation. Perhaps phase separation had aneffect on tb, although the mechanism for thispostulated effect is far from obvious.

Fast Dynamics

Figure 6 shows hm2i values as a function oftemperature for different preparations of the1:4 IgG1:Sucrose formulation. Note that, asexpected, hm2i increases with increasing tempera-ture. At 313 and 323 K, hm2i values increase asprotein fraction increases in freeze dried formula-tions (Tab. 3). Just as formulation impactedfast dynamics, so did the drying method. In a1:4 IgG1:Sucrose formulation, the order of decrease

of hm2i values at both 313 K and 323 K was: Spraydried>Freeze dried &Freeze dried annealed inthe frozen state>Foam. Therefore, since themolecular mobility of the 1:4 IgG1: Sucrose foamis lowest by both measures, one would expect thisfoam to have superior stability relative to theother preparations, provided mobility differencesdominate stability differences. Note that thisprediction is opposite to the predictions madeearlier using FTIR results. It is most curious thatfor the 5:95 IgG1:Sucrose system, the foam hadfaster local dynamics at both 313 K and 323 K thanthe same formulation freeze dried.

Relationship between Global Motionsand Fast Dynamics

Overall, there was not a reasonable correlationbetween fast dynamics and global motions, either

Table 3. Molecular Mobility Studies: (1) TAM Experiments (for Global Motions) to Evaluate tb (h) of SomeLyophilized (LYO), Lyophilized Annealed in the Frozen State Prior to Primary Drying (ANNLYO), Spray Dried andFoam Dried Preparations of Different Formulations

Formulation Treatment

Global Motions(from TAM Studies

at 508C)

Fast Dynamics (fromNeutron Scattering

Studies)

Tg�T* tb (h)�SDhm2i (A)2 at

408Chm2i (A)2 at

508C

IgG1 (pure)*** LYO X** 1.22Medi522:Sucrose 4:1 LYO X** 1.64� 0.36 0.221 0.232

Spray dried X** 1.80� 0.21Foam dried X** 1.52� 0.03

Medi522:Sucrose 1:4 LYO 14.3 0.97� 0.04 0.131 0.138ANNLYO 6 0.96� 0.03 0.127 0.135Spray dried 15.5 1.97� 0.24 0.155 0.164Foam dried 14.4 17.2� 2.8 0.079 0.083


LYO 10.3 0.89� 0.02

Spray dried 13.5 9.18� 0.26Foam dried 13.9 7.65� 1.24

Medi522:Sucrose 5:95 LYO 12.1 0.86� 0.05 0.108 0.125Spray dried 15.5 2.29� 0.34Foam dried 8.4 16.6� 1.4 0.138 0.151

Sucrose LYO 14 0.109 0.121Spray dried 12 2.40� 0.42

b values for foam dried preparations (sucrose-rich) was in the range of 0.7–0.9. For all other preparations, b values ranged from(0.2 to 0.3) (including the protein-rich foam). (2) Incoherent elastic neutron scattering (NS) studies (for fast amplitude motions) toevaluate hm2i (A)2 at 408C and 508C.

(For gray shaded areas, values not determined).*T is the temperature at which TAM experiments were performed (i.e., 508C).**It was not possible to measure a Tg for these protein-rich systems usingMDSC. Using thermally stimulated current (TSC), a Tg

of �808C was obtained for the 4:1 IgG1:Sucrose system (1% moisture). Therefore, T�Tg �308C is a good first approximation forprotein-rich systems.

***Moisture content of pure freeze dried IgG1 was 0.51% w/w (i.e., AT LEAST 0.5% lower moisture content than all other freezedried formulations). Also reported tb value of the pure freeze dried IgG1 is for a single run.



with the variable being composition (at least infreeze dried powders, Figure 7) or drying method(Tab. 3). In fact, the data in Figure 7 show that asprotein fraction increases, amplitude of fastdynamics increases (i.e., greater mobility) butglobal mobility (i.e., reciprocal of tb) apparentlygoes through a minimum. Clearly, fast dynamicsand global dynamics are at least partiallydecoupled for these samples. The study of differ-ent types of molecular mobility and their impacton stability is currently attracting much interest.

Even though different types of mobility have beenstudied22,73 and some studies have examined thecorrelation between a given type of mobility andstorage stability23,33,34, data that examine thevariation of both types of mobility and stability ina controlled series of samples is unique to thisresearch.

In-Process Stability

The effect of drying method on initial physicalaggregation of IgG1 (i.e., In-process physicalstability), as assayed by size exclusion chromato-graphy, is summarized in Table 4. In-processaggregation showed only a small correlation tocomposition. Protein-rich formulations (4:1 and2:1 IgG1:Sucrose), regardless of the dryingmethod used, showed slightly higher aggregatelevels after drying than sucrose-rich formula-tions. Spray drying, a process known to bedamaging to many proteins vulnerable tothe air–water interface,36,74 did not seem to bedamaging to the IgG1, consistent with thepreservation of secondary structure as measuredby FTIR noted earlier.

Storage Stability (Physical Stability)and Physico-Chemical Properties

Degradation kinetics in amorphous pharmaceu-ticals is often described by ‘‘square root of time’’

Figure 6. Plots for Debye-Waller factors from inco-herent elastic neutron scattering for preparations of the1:4 IgG1:Sucrose formulation showing hm2i as a functionof temperature (K).

Figure 5. Plot fromTAMexperiments at 508C for preparations of the 1:4 IgG1:Sucroseformulation showing power (mw/g) as a function of time (h).



Figure 7. The correlation of protein fraction with global motions at 508C (in freezedried formulations and spray dried formulations) and fast dynamics at 508C (in freezedried formulations).

Table 4. In-Process Degradation of IgG1 in All Formulations and Rate Constants for Physical Aggregation—k (%P/month0.5)—after storage stability studies at 40 and 508C. Initial aggregate level in solutions pre-dried was(1.72� 0.04)%.

Formulation TreatmentInitial Levels of Aggregation

(In-Process)�SD*k at 408C�SE**ð%P=

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffimonth

pÞ

k at 508C�SE**ð%P=

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffimonth

pÞ

Medi522:Sucrose 4:1 LYO 2.08� 0.01 6.4� 0.2 12.9� 0.2ANNLYO 2.00� 0.01 6.4� 0.2 13.2� 0.2Spray dried 2.62� 0.03 5.5� 0.2 12.1� 0.2Foam dried 2.26� 0.03 5.5� 0.2 11.5� 0.2


LYO 1.91� 0.01 5.4� 0.4 11.4� 0.4

ANNLYO 1.91� 0.03 5.0� 0.4 10.7� 0.4Spray dried 2.02� 0.01 5.4� 0.5 12.6� 0.4Foam dried 2.57� 0.46 5.5� 0.4 10.7� 0.4

Medi522:Sucrose 2:1 Spray dried 2.03� 0.01 2.7� 0.2 6.8� 1.3Foam dried 2.21� 0.01 2.2� 0.2 7.9� 1.3

Medi522:Sucrose 1:4 LYO 1.79� 0.00 0.11� 0.01 0.31� 0.02ANNLYO 1.82� 0.01 0.05� 0.01 0.27� 0.02Spray dried 1.74� 0.01 0.08� 0.01 0.29� 0.02Foam dried 1.85� 0.05 �0.01 0.20� 0.02


LYO 1.74� 0.03 0.09� 0.01 0.24� 0.06

ANNLYO 1.74� 0.02 0.07� 0.01 0.32� 0.06Spray dried 1.69� 0.03 0.22� 0.02 1.13� 0.07Foam dried 1.78� 0.01 0.04� 0.02 0.22� 0.07

Medi522:Sucrose 5:95 LYO 1.79� 0.03 �0.01 0.02� 0.01ANNLYO 1.77� 0.04 �0.01 0.08� 0.01Spray dried 1.71� 0.01 0.15� 0.02 0.20� 0.02Foam dried 1.74� 0.02 �0.01 0.05� 0.02

Overall, the time dependence of degradation was more consistent with square-root-of-time kinetics. The F value for first orderkinetics was 304 versus 1063 for square-root-of-time kinetics at 408C and 165 versus 1018 at 508C.

*SD, Standard deviation from two or three vials.**SE: Standard error for k value (%P/month0.5) estimated from multiple linear regression using a GLMmodel.



or ‘‘stretched exponential’’ kinetics24,28,59,75,76,similar to the time dependence shown by relaxa-tion kinetics. For our data, a statistical analysisbased on F-values showed that the model withsquare-root of time kinetics fit the IgG1 stabilitydata more appropriately than did first orderkinetics (Tab. 4). The percent aggregates, %P,followed the linear equation75:

%P ¼ P0 þ kffiffit

pð3Þ

where P0 is the initial level of aggregates and k isthe apparent rate constant for physical aggrega-tion on the stretched time scale. Multiple linearregression was done using a general linear model(GLM) to plot and fit the assay data to the model(Eq. 3). Bonferroni tests were performed todetermine if differences between physical aggre-gation rate constants due to drying methodsare statistically significant. Additionally, pairedt-tests were performed to test if differences dueto the presence or absence of surfactant arestatistically significant.

Results for physical storage stability studies aresummarized in Table 4. Overall, differences instability between preparations of the same for-mulation due to varying drying method weresmall. Additionally, it is obvious that inmost cases408Cwas not high enough a temperature to revealsmall stability differences between different

preparations of the same formulation. On theother hand, more conclusive information wasobtained from the 508C data.

Effect of Formulation

As expected and as is consistent with literature,stability of IgG1 in the solid state showed alog-linear dependence on composition.25,28,44

Regardless of the drying method, stability of theformulations increased with sucrose weight ratio.In freeze-dried formulations, the storage stabilityof IgG1 significantly improved (at both 408Cand 508C) as the weight fraction of stabilizerincreased. This trend correlates with an improve-ment in protein secondary structure in the solidstate, a decrease in mass fraction of protein at thesurface of the formulations, a decrease in mole-cular mobility as measured by fast dynamics andan increase in density in sucrose-rich systems(Fig. 8, Tab. 2). The latter is the inverse of specificvolume, which is directly related to free volume.Therefore, one would expect a lower free volumeas sucrose weight fraction increases, assumingmean atomic masses are the same (calculatedto be approximately 7.6 for pure sucrose andapproximately 8 for pure IgG1). Moreover, onewould expect stability to improve with a decreasein the free volume. There was no correlationbetween improvement in stability and global

Figure 8. Relationship of stability (aggregation rate constants) in freezedried systemsof IgG1:Sucrose at 508C to physico-chemical properties: Stability correlated withmolecular mobility (fast dynamics) and protein secondary structure.



motions, unlike what has been observed withother protein systems,26,28 but this lack ofsensitivity to global motions may be a result ofthe limited number of compositions investigatedand the existence of a minimum in global mobilityat a concentration we did not study.28 A similarstability trend was observed with both spray-dried and foam-dried preparations.

The effect of surfactant varied. In a protein-richsystem of 4:1 IgG1:Sucrose, the stability ofboth freeze dried preparations improved withsurfactant which coincided with a slight improve-ment in secondary structure, as well as a sig-nificant decrease in protein surface coverage.However, there was no significant change in thestability of the spray dried or foam dried prepara-tions even though surfactant addition was asso-ciated with significantly lower protein surfacecoverage and improved secondary structure.Mole-cular mobility in these systemswas not measured.In a sucrose-rich system of 1:4 IgG1:Sucrose,the stability of the foam and both freeze driedpreparations were not significantly different thanthe same preparations without surfactant. Thespray dried preparation with surfactant showedinferior stability, as compared to the samepreparation without surfactant, perhaps as aresult of phase separation.

Effect of Drying Method

Drying method impacted protein storage stabilityin different ways, depending on composition. Animportant finding from storage stability studieson all systems studied with and without sur-factant is that there were, at most, minordifferences observed between the storage stabilityof freeze dried formulations annealed in thefrozen state and those with no annealing treat-ment. Note that annealing caused an increasein protein surface coverage, possibly due toincreased molecular mobility during annealingabove T0

g, thereby permitting more protein mole-cules (which are surface active) to migrate to theice–water interface. Therefore, it is expected thatconditions become more favorable for proteinunfolding due to higher mobility above T0

g andexposure to the ice surface for the duration ofannealing. However, FTIR studies showed thatannealed formulations did not undergo a signifi-cant change in secondary structure duringannealing. Moreover, one must also recognizethat annealing produces lower SSA, resulting in alower total protein surface accumulation, which

should stabilize. A study by Tang et al.77 showed arelatively high coupling between system viscosityand unfolding rate, especially as stabilizer ratioincreased. That is, as viscosity increases, unfold-ing slows greatly. In the freeze concentratedsystems studied, even well above T0

g, proteinunfolding was very slow on the time scale offreeze drying. Calculations showed that even at208C above T0

g, the estimated half-life for proteinunfolding of the two proteins studied was on thetime scale of years. Other studies have alsoreported that annealing did not have a damagingeffect on protein secondary structure.36,78 Here,we show that not only did annealing not alterprotein secondary structure, but also annealingdid not cause appreciable deterioration in storagestability.

In a 1:4 IgG1:Sucrose system, the foam was themost stable preparation at both 408C and 508C,even though foam dried material underwent moreperturbation in secondary structure than theother preparations. This improvement in stabilitycorrelated with a decrease in molecular mobilityon both scales (global motions and fast dynamics),along with a significant reduction in both SSAand total protein surface accumulation (Fig. 9A).Similar stability for spray-dried and freeze-driedpreparations correlates somewhat better with fastdynamics (i.e., similar amplitudes for spray driedand freeze dried samples) than global dynamics(i.e., higher mobility for freeze dried than spraydried samples).

In the 5:95 IgG1:Sucrose system, the composi-tion at which chemical heterogeneity effect isexpected to be large,79 there was a rough correla-tion between storage stability and total proteinsurface accumulation (Fig. 9B). Based onstructural relaxation time values, one wouldexpect the foam (which had the highest tb value)to be the most stable followed by the spray-driedpreparation followed by both freeze dried prepara-tions (the un-annealed freeze dried preparationthe lowest tb value). Based on protein secondarystructure, one would expect the spray-dried pre-paration to be the most stable. In fact, the spraydried preparation was the least stable preparationat both 408C and 508C. Lower stability with thespray-dried preparation correlated with its highertotal protein surface accumulation. While totalprotein surface accumulation was higher for thefreeze-dried preparations than the foam, stabilitydifferences between these three preparationsweresmall. Fast dynamics studies suggest that thefreeze-dried preparation (no annealing) had



slightly less molecular mobility than the foam.The preceding observation suggests that fastdynamics, which is not always well correlated tothe viscosity of the glass, represents an additionaldimension to consider in the relationship between

mobility and stability. In the 4:1 IgG1:Sucroseprotein-rich system, both lyophilized preparationsshowed poor stability relative to the spray-dried and foam-dried preparations. This poorerstability did correlate with greater perturbation

Figure 9. Relationship of stability (aggregation rate constants) at 508C to physico-chemical properties: Stability correlated with (A)molecularmobility (global motions andfast dynamics) and total protein surface accumulation in 1:4 IgG1:Sucrose and (B)molecular mobility (fast dynamics) and total protein surface accumulation in 5:95IgG1:Sucrose. (The connecting lines in A–B are for the reader’s convenience only and donot represent interpolations between points).



in protein secondary structure (Tabs. 2 and 4),but did not correlate with molecular mobilityand total protein surface accumulation (Tabs. 3and 4).

In both the 1:4 and 4:1 IgG1:Sucrose systemswith surfactant, the spray dried preparation wasthe least stable, but differences in stabilitybetween all 3 other preparations were small.Lower stability correlated roughly with a highertotal protein surface accumulation, but it iscurious that the higher instability for spray-driedmaterials was not the observation in systemswithout surfactant. Onewould expect the additionof surfactant would have the greatest stabilizationeffect for spray dried material, and therefore ifstability for spray dried material is comparablewithout surfactant, addition of surfactantshould lead to spray dried material being the moststable.

CONCLUSIONS

Formulation and drying method both influencedthe physical stability of IgG1. Stability of IgG1

improved as sucrose fraction increased, regard-less of the drying method. The impact of dryingmethod on stability varied depending on composi-tion. In general, annealing during freezing did nothave a significant impact on storage stability offreeze-dried samples. In a protein-rich system,differences in stability between preparationscorrelated with protein secondary structure. Insucrose-rich systems, differences in stabilitybetween preparations correlated with differencesin molecular mobility from fast dynamics (ratherthan molecular mobility associated with globalmotions) and also correlated roughly with totalprotein surface accumulation.

ACKNOWLEDGMENTS

We acknowledge MedImmune Vaccines forfinancial support for this project. We would alsolike to acknowledge Dr. Daniel Goberman fromthe Surface Sciences Laboratory in the Institutefor Materials Sciences at the University of Con-necticut for his assistance with use of the VGESCALAB MK II series Spectrometer. Finally,we would like to thank Dr. David Lechuga-Ballesteros and Dr. Yi Liang for taking the SEMimages.

REFERENCES

1. Carpenter JF, Pikal MJ, Randolph TW. 1997.Rational design of stable lyophilized protein for-mulations: Some practical advice. Pharm Res 14:969–975.

2. Rational Design of Stable Protein Formulations:Theory and Practice. 2002. In: Carpenter JF,Manning MC, editors. Pharmaceutical Biotechnol-ogy, vol. 13. New York, NY: Kluwer/Plenum. 203p.

3. Lyophilization of Biopharmaceuticals. 2004. In:Costantino HR, Pikal MJ, editors. Biotechnology:Pharmaceutical Aspects, vol. 2. Arlington, VA:AAPS Press. 686p.

4. Maa Y-F, Prestrelski SJ. 2000. Biopharmaceuticalpowders: Particle formation and formulation con-siderations. Curr Pharma Biotechnol 1:283–302.

5. Lee G. 2002. Spray-drying of proteins. PharmaBiotechnol 13 (Rational Design of Stable ProteinFormulations):135–158.

6. Maa Y.-F, Ameri M, Chen D. 2004. Spray-coatingfor biopharmaceutical powder formulations:Beyond the conventional scale and its application.Pharma Res 21:515–523.

7. Jovanovic N, Bouchard A, Jiskoot W. 2004. Stabi-lization of proteins in dry powder formulationsusing supercritical fluid technology. Pharma Res21:1955–1969.

8. Maa Y-F, Nguyen P-A, Hsu CC. 1999. Proteininhalation powders: Spray drying vs spray freezedrying. Pharma Res 16:249–254.

9. Annear DI. 1970. Recoveries of bacteria afterdrying and heating in glutamate foams. J hygiene68:457–459.

10. Annear DI. 1966. Recoveries of bacteria afterdrying in vacuo at a bath temperature of100 degrees C. Nature 211:761.

11. Annear DI. 1956. The preservation of bacteria bydrying in peptone plugs. J hygiene 54:487–508.

12. Truong-Le V. 2003. Preservation of bioactivematerials by freeze dried foam. (MedimmuneVaccines, Inc., USA). Application: WO. p. 72.

13. Mattern M, Winter G, Lee G. 1997. Formulationof proteins in vacuum-dried glasses. Part 1.Improved vacuum-drying of sugars using crystal-lizing amino acids. Eur J Pharma Biopharma 44:177–185.

14. Mattern M, Winter G, Lee G. 1999. Formulation ofproteins in vacuum-dried glasses. II. Process andstorage stability in sugar-free amino acid systems.Pharma Dev Technol 4:199–208.

15. Worrall EE, Litamoi JK, Ayelet G. 2000. Xerovac:An ultra rapid method for the dehydration andpreservation of live attenuated Rinderpest andPeste des Petits ruminants vaccines. Vaccine 19:834–839.

16. Hill SA, MacNaughtan W, Whitcombe MJ. 2005.The effect of thermal history on the maillard



reaction in a glassy matrix. J Agri Food Chem 53:10213–10218.

17. Abdul-Fattah A, Dellerman K, Bogner R, Pikal M.The effect of annealing on the stability of amor-phous solids: Chemical stability of freeze-driedmoxalactam. J Pharm Sci (in press).

18. Krishnamurthy R, Manning MC. 2002. The stabi-lity factor: Importance in formulation development.Curr Pharma Biotechnol 3:361–371.

19. Pikal MJ. 1994. Freeze-drying of proteins: Process,formulation, and stability. ACS Symposium Series,567 (Formulation and Delivery of Proteins andPeptides):120–133.

20. Carpenter JF, Prestrelski SJ, Dong A. 1998.Application of infrared spectroscopy to develop-ment of stable lyophilized protein formulations.Eur J Pharma Biopharma 45:231–238.

21. Franks F, Hatley RHM, Mathias SF. 1991. Materi-als science and the production of shelf-stablebiologicals. BioPharm (Duluth, MN, United States)4:38.

22. Pikal MJ. 2004. Mechanisms of protein stabiliza-tion during freeze-drying and storage: The relativeimportance of thermodynamic stabilization andglassy state relaxation dynamics. Drugs PharmaSci 137 (Freeze-Drying/Lyophilization of Pharma-ceutical and Biological Products):63–107.

23. Cicerone MT, Tellington A, Sokolov A. 2003.Substantially improved stability of biologicalagents in dried form: The role of glassy dynamicsin preservation of biopharmaceuticals. BioProcessInt 1:36–38.

24. Yoshioka S, Aso Y, Kojima S. 2004. Temperature-and glass transition temperature-dependence ofbimolecular reaction rates in lyophilized formula-tions described by the Adam-Gibbs-Vogel equation.J Pharma Sci 93:1062–1069.

25. Cleland J.L, Lam X, Carpenter J.F. 2001. A specificmolar ratio of stabilizer to protein is required forstorage stability of a lyophilized monoclonal anti-body. J Pharma Sci 90:310–321.

26. Duddu SP, Zhang G, Dal Monte PR. 1997. Therelationship between protein aggregation andmolecular mobility below the glass transitiontemperature of lyophilized formulations containinga monoclonal antibody. Pharma Res 14:596–600.

27. Yoshioka S, Aso Y, Kojima S. 1996. Determinationof molecular mobility of lyophilized bovine serumalbumin and g-globulin by solid-state 1H NMR andrelation to aggregation-susceptibility. Pharma Res13:926–930.

28. Chang L, Shepherd D, Pikal MJ. 2005. Mechanismof protein stabilization by sugars during freeze-drying and storage: Native structure preservation,specific interaction, and/or immobilization in aglassy matrix? J Pharma Sci 94:1427–1444.

29. Yoshioka S, Aso Y, Miyazaki T. 2006. Negligiblecontribution of molecular mobility to the degrada-

tion rate of insulin lyophilized with poly(vinylpyr-rolidone). J Pharma Sci 95:939–943.

30. Buera MP, Rossi S, Chirife J. 1999. DSC confirma-tion that vitrification is not necessary for stabiliza-tion of the restriction enzyme Eco RI dried withsaccharides. Biotechnology Progress 15:577–579.

31. Davidson P, Sun WQ. 2001. Effect of sucrose/raffinose mass ratios on the stability of co-lyophi-lized protein during storage above the Tg. PharmaRes 18:474–479.

32. Cicerone MT, Soles CL. 2004. Fast dynamics andstabilization of proteins: Binary glasses of trehaloseand glycerol. Biophys J 86:3836–3845.

33. Cicerone MT, Pikal MJ, Chang L. 2005. Fastdynamics as a diagnostic for excipients in preserva-tion of dried proteins. Am Pharma Rev 8:22, 24–27.

34. Yoshioka S, Miyazaki T, Aso Y. 2006. b-Relaxationof insulin molecule in lyophilized formulationscontaining trehalose or dextran as a determinantof chemical reactivity. Pharma Res 23:961–966.

35. Sane SU, Wong R, Hsu CC. 2004. Raman spectro-scopic characterization of drying-induced struc-tural changes in a therapeutic antibody:Correlating structural changes with long-termstability. J Pharma Sci 93:1005–1018.

36. Webb SD, Golledge SL, Randolph TW. 2002. Sur-face adsorption of recombinant human interferon-gin lyophilized and spray-lyophilized formulations.J Pharma Sci 91:1474–1487.

37. Faeldt P, Bergenstahl B. 1994. The surface compo-sition of spray-dried protein-lactose powders. Col-loids and Surfaces, A: Physicochem Eng Aspects90:183–190.

38. Millqvist-Fureby A, Burns N, Bergenstahl B. 1999.Surface activity at the air-water interface inrelation to surface composition of spray-dried milkprotein-stabilized emulsions. Special Publication—Royal Soc Chem 227 (Food Emulsions andFoams):236–245.

39. Millqvist-Fureby A, Malmsten M, Bergenstahl B.1999. Surface characterisation of freeze-dried pro-tein/carbohydrate mixtures. Int J Pharma 191:103–114.

40. Millqvist-Fureby A, Malmsten M, Bergenstahl B.1999. Spray-drying of trypsin—Surface character-isation and activity preservation. Int J Pharma188:243–253.

41. Adler M, Lee G. 1999. Stability and surface activityof lactate dehydrogenase in spray-dried trehalose.J Pharma Sci 88:199–208.

42. Adler M, Unger M, Lee G. 2000. Surface composi-tion of spray-dried particles of bovine serumalbumin/trehalose/surfactant. Pharma Res 17:863–870.

43. Yu Z, Johnston KP, Williams RO. 2006. Sprayfreezing into liquid versus spray-freeze drying:Influence of atomization on protein aggregationand biological activity. Eur J Pharma Sci 27:9–18.



44. Breen ED, Curley JG, Shire SJ. 2001. Effect ofmoisture on the stability of a lyophilized huma-nized monoclonal antibody formulation. PharmaRes 18:1345–1353.

45. Maury M, Murphy K, Lee G. 2005. Spray-drying ofproteins: Effects of sorbitol and trehalose onaggregation and FT-IR amide I spectrum of animmunoglobulin G. Eur J Pharma Biopharma 59:251–261.

46. Chang L, Shepherd D, Pikal MJ. 2005. Effect ofsorbitol and residual moisture on the stability oflyophilized antibodies: Implications for themechanism of protein stabilization in the solidstate. J Pharma Sci 94:1445–1455.

47. Abdul-Fattah A, Pikal M, Lechuga-Ballesteros D.2005. Composition heterogeneity effects due tovariation in processing and formulation of Methio-nyl-human Growth Hormone. in AAPS NationalBiotechnology Conference. San Francisco, CA:

48. Sovani S.D, Chou E, Crofts J.D. 2001. Highpressure effervescent atomization: Effect of ambi-ent pressure on spray cone angle. Fuel 80:427–435.

49. Truong-Le V, Scherer T. 2005. High pressurespray-dry of bioactive materials. (MedimmuneVaccines, Inc., USA). Application: WO. p. 71.

50. Lide DR. 2002. Handbook of Chemistry andPhysics, 83rd Edition. Boca Raton, FL: CRC Press.pp. 2664 pp.

51. Faeldt P, Bergenstahl B, Carlsson G. 1993. Thesurface coverage of fat on food powders analyzed byESCA (electron spectroscopy for chemical analysis).Food Struct 12:225–234.

52. Kendrick BS, Dong A, Carpenter JF. 1996. Quanti-tation of the area of overlap between second-derivative amide I infrared spectra to determinethe structural similarity of a protein in differentstates. J Pharma Sci 85:155–158.

53. DePaz RA, Dale DA, Randolph TW. 2002. Effects ofdrying methods and additives on the structure,function, and storage stability of subtilisin: Role ofprotein conformation and molecular mobility.Enzyme Microbial Technol 31:765–774.

54. Dong A, Meyer JD, Carpenter JF. 2000. Compara-tive Fourier transform infrared and circulardichroism spectroscopic analysis of a1-proteinaseinhibitor and ovalbumin in aqueous solution. ArchBiochem Biophys 383:148–155.

55. Izutsu K-i, Aoyagi N, Kojima S. 2004. Protectionof protein secondary structure by saccharides ofdifferent molecular weights during freeze-drying.Chem Pharma Bull 52:199–203.

56. Heller MC, Carpenter JF, Randolph TW. 1999.Conformational stability of lyophilized PEGylatedproteins in a phase-separating system. J PharmaSci 88:58–64.

57. Dong A, Caughey WS. 1994. Infrared methods forstudy of hemoglobin reactions and structures.

Methods Enzymol 232 (Hemoglobins, Pt. C):139–175.

58. Prestrelski SJ, Tedeschi N, Carpenter JF. 1993.Dehydration-induced conformational transitions inproteins and their inhibition by stabilizers. BiophysJ 65:661–671.

59. Liu J, Rigsbee D.R, Pikal M.J. 2002. Dynamics ofpharmaceutical amorphous solids: The study ofenthalpy relaxation by isothermal microcalorime-try. J Pharma Sci 91:1853–1862.

60. Kawakami K, Pikal MJ. 2005. Calorimetric inves-tigation of the structural relaxation of amorphousmaterials: Evaluating validity of the methodolo-gies. J Pharma Sci 94:948–965.

61. Meyer A, Dimeo RM, Neumann DA. 2003. Thehigh-flux backscattering spectrometer at the NISTCenter for Neutron Research. Rev Sci Instrum 74:2759–2777.

62. Pikal MJ, Dellerman KM, Riggin RM. 1991. Theeffects of formulation variables on the stability offreeze-dried human growth hormone. Pharma Res8:427–436.

63. Angell CA. 1995. Formation of glasses from liquidsand biopolymers. Science (Washington, D. C.) 267:1924–1935.

64. Carpenter JF, Izutsu K.-i, Randolph TW. 2004.Freezing- and drying-induced perturbations ofprotein structure and mechanisms of proteinprotection by stabilizing additives. Drugs PharmaSci 137 (Freeze-Drying/Lyophilization of Pharma-ceutical and Biological Products):147–186.

65. Prestrelski SJ, Pikal KA, Arakawa T. 1995.Optimization of lyophilization conditions for re-combinant human interleukin-2 by dried-stateconformational analysis using Fourier-transforminfrared spectroscopy. Pharma Res 12:1250–1259.

66. Lu X, Pikal MJ. 2004. Freeze-Drying of Mannitol-Trehalose-Sodium Chloride-Based Formulations:The Impact of Annealing on Dry Layer Resistanceto Mass Transfer and Cake Structure. Pharma DevTechnol 9:85–95.

67. Pikal MJ, Shah S, Lang J.E. 1983. Physicalchemistry of freeze-drying: Measurement of sub-limation rates for frozen aqueous solutions bya microbalance technique. J Pharma Sci 72:635–650.

68. Randolph TW, Searles JA. 2002. Freezing andannealing phenomena in lyophilization: Effectsupon primary drying rate, morphology, and hetero-geneity. Am Pharma Rev 5:40,42,44–46.

69. Searles JA, Carpenter JF, Randolph TW. 2001.Annealing to optimize the primary drying rate,reduce freezing-induced drying rate heterogeneity,and determine Tg0 in pharmaceutical lyophiliza-tion. J Pharma Sci 90:872–887.

70. Elversson J, Millqvist-Fureby A, Elofsson U. 2003.Droplet and particle size relationship and shell



thickness of inhalable lactose particles duringspray drying. J Pharma Sci 92:900–910.

71. Elversson J, Millqvist-Fureby A. 2005. Particle sizeand density in spray drying-effects of carbohydrateproperties. J Pharma Sci 94:2049–2060.

72. Pikal MJ. 2002. Chemistry in solid amorphousmatrices: Implication for biostabilization, inAmorphous Food and Pharmaceutical Systems.Cambridge, UK: The Royal Society of Chemistry.pp. 257–272.

73. Yoshioka S, Aso Y, Kojima S. 2002. Differentmolecular motions in lyophilized protein formula-tions as determined by laboratory and rotatingframe spin-lattice relaxation times. J Pharma Sci91:2203–2210.

74. Maa Y-F, Nguyen P-AT, Hsu S.W. 1998. Spray-drying of air-liquid interface-sensitive recombinanthuman growth hormone. J Pharma Sci 87:152–159.

75. Pikal MJ, Rigsbee DR. 1997. The stability of insulinin crystalline and amorphous solids: Observation of

greater stability for the amorphous form. PharmaRes 14:1379–1387.

76. Yoshioka S, Aso Y, Kojima S. 2001. Usefulness ofthe Kohlrausch-Williams-Watts stretched expo-nential function to describe protein aggregation inlyophilized formulations and the temperaturedependence near the glass transition temperature.Pharma Res 18:256–260.

77. Tang X, Pikal MJ. 2005. Measurement of theKinetics of Protein Unfolding in Viscous Systemsand Implications for Protein Stability in Freeze-Drying. Pharma Res 22:1176–1185.

78. Sarciaux J-M, Mansour S, Nail SL. 1999. Effects ofbuffer composition and processing conditionson aggregation of bovine IgG during freeze-drying.J Pharm Sci 88:1354–1361.

79. Abdul-Fattah A, Kalonia D, Pikal M. Thechallenge of drying method selection for proteinpharmaceuticals: Product quality implications. JPharm Sci (in press).



Drying-induced variations in physico-chemical properties of amorphous pharmaceuticals and their...

Documents

Transcript of Drying-induced variations in physico-chemical properties of amorphous pharmaceuticals and their...