A New Mechanism for Decreasing Aggregation of
`Recombinant Human Interferon-g by a Surfactant:
`Slowed Dissolution of Lyophilized Formulations in a
`Solution Containing 0.03% Polysorbate 20
`
`SERENA D. WEBB,1 JEFFREY L. CLELAND,2 JOHN F. CARPENTER,3 THEODORE W. RANDOLPH1
`
`1Department of Chemical Engineering, Center for Pharmaceutical Biotechnology, University of Colorado,
`Engineering Center, Room ECCH 111, Boulder, Colorado 80309-0424
`
`2Genentech, Inc., 460 Pt. San Bruno Boulevard, South San Francisco, California 94080
`
`3Department of Pharmaceutical Sciences, School of Pharmacy, University of Colorado Health Sciences Center,
`Denver, Colorado 80262
`
`Received 16 March 2001; revised 19 July 2001; accepted 28 August 2001
`
`ABSTRACT: To study the mechanisms by which Tween 20 (polysorbate 20) used in a
`reconstitution solution affects the aggregation of lyophilized recombinant human
`interferon-g (rhIFN-g), we used four types of buffered formulations containing 0.4±
`5 mg/mL rhIFN-g in either 10 mM potassium phosphate or phosphate buffered saline:
`(1) without excipients, (2) with 5% sucrose, (3) with 0.03% polysorbate 20, or (4) with the
`combination of 5% sucrose and 0.03% polysorbate 20. After lyophilization, infrared
`spectroscopy was used to analyze the secondary structure of the protein in the freeze-
`dried solid. Each solid showed structural perturbation of the protein. Each formulation
`was reconstituted with water or a 0.03% polysorbate 20 solution. Aggregation of rhIFN-
`g after reconstitution was measured by optical density at A350, and recovery of soluble
`protein was determined by high-performance liquid chromatography and ultraviolet
`spectroscopy. After reconstitution with a 0.03% polysorbate 20 solution, aggregation
`levels in all formulations were either reduced or similar to those found after recons-
`titution with water. These results revealed the potential for recovery of native protein
`using the appropriate reconstitution conditions, even though the protein is non-native
`in the lyophilized state. Urea-induced unfolding with and without polysorbate 20 as
`measured by second-derivative ultraviolet spectroscopy indicated that a concentration
`of 0.03% polysorbate 20 lowered the free energy of unfolding for rhIFN-g (destabilizing).
`Polysorbate 20 also retarded refolding from urea solutions and increased aggregation.
`At a level of 0.03%, polysorbate 20 did not protect the protein against surface-induced
`aggregation during agitation. Dissolution times in water versus a 0.03% polysorbate
`20 solution were measured using a rotating disk electrode for lyophilized formulations
`containing an electrochemically reactive species. The presence of 0.03% polysorbate
`20 in the reconstitution solution nearly doubled the time required for dissolution of the
`phosphate buffered saline formulation, and the sucrose formulations dissolved 33±57%
`more slowly. Slowing the dissolution rates of lyophilized powders allows more time for
`
`Correspondence to: Theodore W. Randolph (Telephone: 303-
`492-4776; Fax: 303-492-4341;
`E-mail: randolph@pressure3.colorado.edu)
`
`Journal of Pharmaceutical Sciences, Vol. 91, 543±558 (2002)
`ß 2002 Wiley-Liss, Inc. and the American Pharmaceutical Association
`
`JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 2, FEBRUARY 2002
`
`543
`
`KASHIV EXHIBIT 1029
`IPR2019-00791
`
`Page 1
`
`

`

`544
`
`WEBB ET AL.
`
`the protein to refold while it decreases the maximum concentration of the protein at the
`dissolution interface, thus reducing the total amount of aggregation. ß 2002 Wiley-Liss,
`Inc. and the American Pharmaceutical Association J Pharm Sci 91:543±558, 2002
`Keywords:
`rehydration; protein stabilization; surfactant; protein formulation;
`dissolution rates; dissolution model
`
`INTRODUCTION
`
`Most native proteins are only marginally stable,
`with a free energy of stabilization (DGN ˆ > Dˆ
`Gnative Gdenatured) of only about 50 15 kJ/mol.1
`A protein formulated in aqueous solution is
`particularly susceptible to conformational changes
`in its native structure, often leading to physical or
`chemical degradation. Therefore, proteins are
`commonly lyophilized to achieve long-term stabi-
`lity.2,3 Lyophilization produces rapidly dissolving
`powders, allows storage at higher temperatures,
`and reduces conformational mobility due to
`slowed molecular motion.
`The process of lyophilization often damages
`proteins. The addition of excipients and carefully
`designed lyophilization cycles are generally requi-
`red to yield high recoveries of native protein after
`lyophilization. The majority of protein lyophiliza-
`tion research has been directed toward protecting
`the protein during the freezing and drying steps.3
`Minimal effort has been directed toward the
`reconstitution step.4,5 Reconstitution occurs
`under conditions of temperature and pressure
`clearly different from reversing the path of the
`freeze dryer. Therefore, different methods for
`stabilizing the protein may be required. By adding
`excipients, such as amino acids, heparin, dextran
`sulfate, and surfactants to the reconstitution
`solutions for lyophilized proteins, Zhang et al.4,5
`showed that recovered activities could be increased
`and levels of aggregate formation decreased compa-
`red with using water alone. However, the mechan-
`ism(s) by which these additives reduced protein
`aggregation during reconstitution was not clear.
`The purposes of the current study were to (a)
`determine whether the aggregation of rhIFN-g is
`decreased by the use of polysorbate 20 in a
`reconstitution medium; (b) identify the mechan-
`ism(s) by which polysorbate 20 affects the aggrega-
`tion level; and (c) compare the effect of polysorbate
`20 included in the lyophilization formulation
`versus its addition during reconstitution.
`The protein selected for this study, recombi-
`nant human interferon-g (rhIFN-g, has been
`shown to aggregate after acid denaturation,6 after
`attempted refolding,7±9 and above pH 5 during
`
`thermal unfolding.7±12 It also aggregates during
`refolding from guanidine hydrochloride13 or urea.10
`Phosphate buffer systems were selected be-
`cause they cause signi®cant protein structural
`perturbation during lyophilization, resulting in
`aggregation upon reconstitution. Phosphate salts
`can crystallize during freezing,14 creating pH
`changes that could perturb rhIFN-g structure.
`Phosphate buffered saline (PBS) was used
`because of its propensity for large pH changes
`during freezing, whereas potassium phosphate
`buffer undergoes very little change. Sucrose was
`added in some of the formulations to partially
`stabilize the protein during lyophilization.3,15±18
`Polysorbate 20 was selected speci®cally because its
`addition to a reconstitution medium had been
`shown to reduce aggregation of keratinocyte growth
`factor and interleukin-2 after reconstitution.4,5
`
`MATERIALS AND METHODS
`
`Protein and Reagents
`
`Pharmaceutical-quality rhIFN-g expressed in
`Escherichia coli was produced and puri®ed at
`Genentech, Inc., South San Francisco, CA. Snake-
`skinß pleated dialysis tubing (7000 MWCO) and
`10% polysorbate 20 (Tween 20 in Surfact-Ampsß
`20) were purchased from Pierce, Rockford, IL.
`High-purity sucrose was purchased from Pfan-
`stiel, Waukegan, IL. Sodium chloride, potassium
`chloride, sodium phosphate (dibasic), and potas-
`sium phosphate (monobasic and dibasic salts)
`were purchased from Fisher Scienti®c, Atlanta,
`GA. Potassium ferrocyanide trihydrate was pur-
`chased from Sigma, St. Louis, MO. The Non-
`Interfering Protein Assayß was purchased from
`Bioworld Laboratory Essentials, Dublin, OH, and
`used according to the manufacturer's Protocol I
`instructions. All reagents were ACS reagent
`grade or higher quality. Millipore water was used
`in the preparation of solutions. Formulation
`solutions containing protein were ®ltered with
`low protein-binding ®lters (polyvinylidine di¯uor-
`ide) from Whatman, Clifton, NJ, with pore sizes of
`0.22 mm. Buffer solutions were ®ltered using a
`0.22 mm ®lter from Millipore, Bedford, MA.
`
`JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 2, FEBRUARY 2002
`
`Page 2
`
`

`

`Formulation Buffers and Filling Solutions
`
`Two buffers were used to prepare the formula-
`tions: 10 mM potassium phosphate, pH 7.5 or PBS
`containing 10 mM sodium phosphate (dibasic),
`2 mM potassium phosphate (monobasic), 137 mM
`sodium chloride, and 3 mM potassium chloride,
`pH 7.0. rhIFN-g was dialyzed into each buffer,
`and vials were prepared with 0.4 mg/mL rhIFN-g
`in one of four buffered solutions: (1) no excipients,
`(2) 5% sucrose, (3) 0.03% polysorbate 20, or (4) the
`combination of 5% sucrose and 0.03% polysorbate
`20. The concentration of polysorbate 20 was selec-
`ted because it is signi®cantly above the critical
`micelle concentration (0.007%) in water. Protein
`concentrations of 0.4, 1, and 5 mg/mL rhIFN-g in
`each buffer only were also lyophilized. All concen-
`trations identi®ed as percentages (i.e., sucrose and
`polysorbate 20) were prepared as w/v solutions.
`
`Lyophilization Procedure
`
`Lyophilization vials of 3 mL-volume (West Com-
`pany no. 6800-0316, ¯int glass, Lionville, PA) were
`®lled with 1 mL solution and loaded into an FTS
`Durastop freeze dryer, Stone Ridge, NY. Thermo-
`couples were placed in two vials. Vials were loaded
`randomly into the freeze dryer and equilibrated at
` 18C for 30 min. The shelf temperature was
`decreased by 2.58C/min to 458C. When the tem-
`perature of the samples reached 308C, the shelf
`temperature was held at 458C for an additional
`2 h. Primary drying was performed with a shelf
`temperature of 358C and a chamber pressure
`< 100 mTorr for approximately 40 h. The shelf
`temperature was increased during secondary dry-
`ing by 18C/min and chamber pressure was
`increased to 200 mTorr. Shelf temperature was
`held at 20, 0, and 208C for 4 h each, and increased
`to 30 and 408C for 1 h each to complete the cycle.
`
`Moisture Determination
`
`Random lyophilized samples from formulations in
`both buffers were prepared in a dry-nitrogen-
`purged glove box and analyzed for moisture
`content using the Karl Fisher method.19 A Mettler
`DL37 coulometric moisture analyzer
`(High-
`tstown, NJ) was used with Hydranal reagents
`(Reidel de Haen, Seelze, Germany).
`
`Protein Secondary Structure
`by Infrared Spectroscopy
`
`Lyophilized samples containing 0.2 mg of rhIFN-g
`were mixed with 300 mg of potassium bromide
`
`DECREASING AGGREGATION OF rhIFN-g
`
`545
`
`and pressed at 30 mPa into pellets for secondary
`structure analysis. Calcium ¯uoride windows
`separated by a 6 mm spacer were used for liquid
`samples. Spectra were collected at 258C with a
`Nicolet Magna-IRß 750 Series II spectrometer,
`equipped with a DTGS detector as described
`previously.20 For each spectrum, a 256-scan inter-
`ferogram was collected in single beam mode, with
`a 4 cm1 resolution using Omnicß (v. 2.1) software
`from Nicolet. The optical bench and sample
`chamber were continuously purged with dry air
`supplied from a Whatman model 75-52 IR purge
`gas generator (Haverhill, MA). Background liquid
`and gaseous water spectra were subtracted from
`the protein spectra according to previously estab-
`lished criteria.21 Second-derivative spectra were
`calculated using Nicolet Omnic software. The
`®nal protein spectra were smoothed with a 7-
`point function to remove white noise. All second-
`derivative spectra were baseline corrected using
`Galactic's GRAMS 386 software based on a
`previously described method,21 and were area
`normalized under the second-derivative amide I
`region, 1600±1700 cm1.22 The native control
`sample consisted of 20 mg/mL rhIFN-g in 5 mM
`sodium succinate, pH 5.0.
`
`Reconstitution Procedure
`
`Reconstitution was performed at room tempera-
`ture (238C). Stoppers were removed and samples
`were reconstituted in a random order with 1 mL of
`water or an aqueous 0.03% polysorbate 20 solu-
`tion added within 2 s to the center of the lyophi-
`lized plug by a micropipet. Vials were placed on a
`Barnstead/Thermolyne Labquake red blood cell
`suspender and mixed at 8 rpm for 4 min. To verify
`that this mixing process does not cause aggrega-
`tion, nonlyophilized controls for all four formula-
`tions in each buffer containing 0.4, 1, and 5 mg/
`mL rhIFN-g were mixed under identical condi-
`tions. Absorbances were read at A350 to evaluate
`optical densities before and after the mixing
`process was completed. After 4 min of mixing,
`aggregates were not detected in any of the control
`samples.
`
`Optical Density and Protein Concentration
`Measurements Using Ultraviolet (UV) Spectroscopy
`
`A Perkin-Elmer Lambda 3B UV/VIS spectro-
`photometer was used to measure absorbances at
`280 nm (A280) and 350 nm (A350). A280 was used to
`determine protein concentration, and A350/mg
`
`JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 2, FEBRUARY 2002
`
`Page 3
`
`

`

`546
`
`WEBB ET AL.
`
`protein was used to identify the presence of
`protein aggregates.23,24 A350 was recorded imme-
`diately upon reconstitution, and A280 was recor-
`ded after centrifugation to remove any insoluble
`aggregates. A350 was also recorded after centrifu-
`gation to verify the absence of signi®cant optical
`density. Protein concentration was
`calcula-
`ted using 2ˆ0.75 mL mg 1cm1 for rhIFN-g at
`280 nm.25
`
`Soluble Aggregate Detection and Protein
`Concentration Using High-Performance Liquid
`Chromatography±Size Exclusion Chromatography
`(HPLC±SEC)
`
`A Dionex DX500 chromatography system with a
`GP40 gradient pump, HP1050 series auto sam-
`pler and an AD20 absorbance detector at 214 nm
`was used with a Tosohaas TSK-GEL G2000SWXL
`stainless steel column. A 1.2 M KCl mobile phase
`was used at a ¯ow rate of 0.2 mL/min. The native
`dimeric form of rhIFN-g eluted at 41 min, and
`soluble aggregates appeared between 33 and
`35 min. The column was calibrated with cyto-
`chrome C, carbonic anhydrase, albumen, and
`alcohol dehydrogenase. Native rhIFN-g controls,
`run at the beginning and end of each HPLC
`sample run, were used to calculate the concentra-
`tion of protein in the samples using peak areas.
`
`Measurement of pH in Freeze-Concentrated
`Formulations
`
`pH in frozen solutions was measured using the
`method of Anchordoquy and Carpenter.26 An
`Ingold pH electrode containing a low-temperature
`electrolyte, Friscolyte ``B,'' was calibrated by
`measuring pH and conductivity (mV) at 25, 10,
`and 08C, using three standard buffer solutions.
`The pH electrode was placed into a 15-mL falcon
`tube containing each solution to be measured.
`Each tube was placed into an ethylene glycol bath
`maintained at 158C or below. A type T thermo-
`couple was placed within 3 mm of the probe tip to
`monitor sample temperature. The pH was mon-
`itored until the value remained constant (at least
`30 min after ice formation in the sample).
`
`Protein Unfolding in Urea Followed by
`Second-Derivative UV Spectroscopy
`
`A stock solution of urea in PBS, pH 7.0, was
`prepared as per the method of Pace et al.27
`Solutions of 1 mg/mL rhIFN-g in urea with and
`
`JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 2, FEBRUARY 2002
`
`without 0.03% polysorbate 20 were prepared, held
`overnight at 2±88C, and analyzed the next
`morning. Triplicate samples at each solution
`condition were analyzed. UV scans were mea-
`sured from 310 to 250 nm with a scan rate of 15 nm
`per minute using a Perkin-Elmer Lambda 3B
`dual beam spectrophotometer. Data acquisition
`was made via a National Instruments (Austin,
`TX) model AT-MIO-16E-10 data acquisition board
`at a rate of 5 samples per second. National
`Instruments LabViewß software was used to
`control data acquisition and Microsoft Excelß to
`convert the wavelength and absorption data from
`volts to nanometers and absorbance units, respec-
`tively. Background spectra from samples contain-
`ing all solution components except protein were
`subtracted for all samples. The second derivatives
`of the absorption spectra (d2A/dl2) were calcu-
`lated in GRAMS/386 (v. 3.02) software (Galactic
`Industries) using the Savitzky-Golay method with
`a second-order polynomial smoothed over  2 nm.
`Unfolding of rhIFN-g was followed by change in
`extremum depth near 286 nm in the second-
`derivative absorption spectra.28 The native pro-
`tein spectrum has a minimum near 286 nm
`re¯ective of the microenvironments of tryptophan
`and tyrosine residues.28±31 The depth of this
`minimum is reduced, eventually becoming a
`maximum as the concentration of urea is increa-
`sed. The data are converted to the fraction of
`native protein (fN) as a function of urea by using a
`baseline correction for the pre- and post-transi-
`tion regions as described by the method of Pace
`et al.27 The free energy of unfolding, DG, at each
`urea concentration was calculated as follows:
` G ˆ RT ln K
`
`…1†
`
`Temperature (T) is reported in Kelvin and R is the
`gas constant. The calculation for the equilibrium
`constant, K, assumes dissociation of the native
`dimer (N) into two monomers (D)28:
`N $ 2D
`K ˆ 4No‰fN ‡ 1=fN 2Š
`
`…2†
`…3†
`
`where No is the initial concentration of the native
`protein, and fN is the fraction of native protein.
`Note that because of the concentration term in
`the calculation for K, units are introduced. For
`the calculation of DG, a reference state of 1 M
`rhIFN-g at 238C was assumed,28 making K
`dimensionless. DG at 0 M urea was calculated
`by linear extrapolation of DG as a function of urea
`
`Page 4
`
`

`

`concentration.27 Reported error for DG at 0 M
`urea is based on 95% con®dence limits on the
`linear extrapolation.
`
`Refolding of rhIFN-g in Urea
`
`One or three milligrams/milliliter rhIFN-g was
`equilibrated for 4 h at 238C in 3.5 or 5 M urea in
`PBS, pH 7.0. Solutions were rapidly diluted to a
`urea concentration of 1 M with PBS, pH 7.0, or
`polysorbate 20 in PBS, pH 7.0. The level of
`polysorbate 20 was adjusted such that the ®nal
`concentration upon dilution was 100 mM (0.012%).
`Optical densities of the solutions were recorded
`immediately at A350.23,24 Solutions were centri-
`fuged and protein concentration in the super-
`natant was determined by the Non-Interfering
`Protein Assay.
`
`Agitation Procedure for
`Surface-Induced Aggregation
`
`Solutions of 1-mL sample size containing 1 or 5 mg
`rhIFN-g in PBS, pH 7.0, with and without
`polysorbate 20 (0.03 or 0.1%) were added to
`1.7-mL Eppendorf tubes and rotated at 8 rpm
`and 238C for 0.25, 1, 1.5, 3, and 15 h. For some
`samples, PBS buffer was diluted to 1/15 its initial
`strength to determine ionic strength effects (from
`152 to 10 mM). Protein aggregation was mon-
`itored by the ratio of A350 to protein concentration.
`Samples were centrifuged and protein recovery
`in the supernatant was analyzed using the
`Non-Interfering Protein Assay.
`
`Dissolution Rate Determination
`
`We determined dissolution times in water versus
`a 0.03% Tween 20 solution electrochemically
`using a rotating disk electrode at 238C. The
`method is brie¯y described herein, whileas the
`details may be found elsewhere.32 Potassium
`ferrocyanide was lyophilized with the formula-
`tions that had produced aggregation after re-
`constitution (all
`formulations except 10 mM
`potassium phosphate, 5% sucrose/0.03% Tween
`20 in potassium phosphate and in PBS). Each
`lyophilized vial contained suf®cient potassium
`ferrocyanide to result in a concentration of 0.8 mM
`when reconstituted with 30 mL of solution. The
`platinum rotating disk electrode was submerged
`slightly below the surface of the reconstitution
`solution in a 50-mL beaker. A platinum counter
`electrode and Ag/AgCl/KCl (saturated) reference
`
`DECREASING AGGREGATION OF rhIFN-g
`
`547
`
`electrode were af®xed together near the inside
`edge of the beaker. An Analytical Rotator (model
`AFASRP) from the Pine Instrument Company
`was used to control the rotating disk speed. A Pine
`Instrument Company Bi-Potentiostat
`(model
`AFRDE5) was used to apply a constant potential
`(600 mV) while a lyophilized sample was added to
`the reconstitution solution. The solution also
`contained a background electrolyte, 0.15 M NaCl,
`which was necessary to prevent the migration of
`ionic species in a ®eld and therefore enable
`measurement of diffusive processes. Data acquisi-
`tion was made using a National Instruments
`(Austin, TX) model AT-MIO-16E-10 data acquisi-
`tion board at a rate of 50 samples per second, and
`via a Kipp and Zonen X-Y Recorder (type BD91)
`concurrently. National
`Instruments LabView
`software was used to control the electronic data
`acquisition. Rates of dissolution were measured in
`triplicate. As the lyophilized material dissolves,
`the ferrocyanide becomes solvated and begins to
`undergo the following oxidation reaction:
`6 !Pt
`
`Fe…CN†36 ‡ e
`Fe…CN†4
`
`…4†
`
`NaCl
`
`The Pt concentration of ferrocyanide in solution is
`directly proportional to the current, and both
`increase until dissolution is complete. Dissolution
`pro®les were found to ®t a ®rst order response,
`and a ®rst order time constant (t) was calculated
`for each sample type.
`
`RESULTS
`
`Lyophilization and Reconstitution
`of rhIFN-g in Phosphate Buffers
`
`Characterization of the Lyophilized Formulations
`
`Residual moisture in lyophilized protein formula-
`tions may impact the overall stability as well as
`the reconstitution process. The water content
`remaining in the samples after lyophilization
`ranged from 0.23 to 0.77 0.1% (two standard
`deviations). Random samples were tested from
`both the potassium phosphate and PBS buffer
`formulations, and vials containing 0.4±5 mg
`rhIFN-g were included. No apparent differences
`in water content were obvious between sample
`types, and therefore differences in formulations
`cannot be attributed to variations in water
`content in the lyophilized samples.
`Previous studies have shown that phosphate
`buffers may induce signi®cant damage to proteins
`during freeze-drying as the result of pH changes
`
`JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 2, FEBRUARY 2002
`
`Page 5
`
`

`

`548
`
`WEBB ET AL.
`
`upon freezing.33±36 To assess this effect on rhIFN-
`g, the secondary structure of the protein in the
`lyophilized formulations was characterized by
`infrared (IR) spectroscopy. The second-derivative
`IR spectra of the amide I region for lyophilized
`0.4 mg/mL rhIFN-g in 10 mM potassium phos-
`phate, pH 7.5, are shown in Figure 1A. Bands in
`the second-derivative IR spectrum of native
`rhIFN-g have been assigned previously.6 The
`band at 1682 cm1 is a combination of two bands
`at 1677 and 1684 cm1. The band at 1677 cm1 is
`assigned to extended chain, whereas the band at
`1684 cm1 is assigned to turn structure. The band
`appearing at 1633 cm1 is also composed of two
`bands at 1630 and 1635 cm1, both of which are
`
`Figure 1.
`(A) Second-derivative IR spectra of 0.4 mg/
`mL rhIFN-g lyophilized in 10 mM potassium phos-
`phate, pH 7.5. Formulations: 10 mM potassium phos-
`phate (- - - - - -); 5% sucrose (± ± ±); 0.03% polysorbate
`20 (Ð Ð); 5% sucrose and 0.03% polysorbate 20 (- - -);
`native control
`(Ð Ð, bold).
`(B) Second-derivative
`IR spectra of 0.4±5 mg/mL rhIFN-g lyophilized in
`10 mM potassium phosphate, pH 7.5. Symbols: 0.4 mg/
`mL (Ð Ð); 1mg/mL (± ± ±); 5 mg/mL (- - - - - -); native
`control (- - -, bold).
`
`JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 2, FEBRUARY 2002
`
`assigned to extended chain. Finally, the dominant
`band at 1656 cm1 corresponds to a-helix.
`Less-intense a-helix bands were seen in
`sucrose-free formulations with and without poly-
`sorbate 20. These losses were compensated for by
`an increase in turn structure (at 1684 cm1), and
`formation of b-sheet structure at 1695 cm1,37,38
`particularly in the excipient-free formulation. The
`general band broadening of the amide I region in
`these same formulations is indicative of hetero-
`geneity in the secondary structure of rhIFN-g
`after lyophilization.6,38,39 When sucrose is present
`in the 10-mM potassium phosphate formulations,
`native bands are retained to a greater extent,
`though there is still substantial perturbation of
`secondary structure in the dried solid.
`The IR spectra for 0.4±5 mg/mL rhIFN-g
`lyophilized in 10 mM potassium phosphate, pH
`7.5, are shown in Figure 1B. All samples show
`signi®cant perturbation of native protein struc-
`ture, as indicated by the dramatic reduction in
`helix band intensity in the IR spectra. As protein
`concentration increases, this perturbation increa-
`ses slightly but remains similar (within error)
`among the different concentrations.
`The second-derivative IR spectra of the amide I
`region for 0.4 mg/mL rhIFN-g lyophilized in PBS,
`pH 7.0, are shown in Figure 2A. In contrast to
`results shown in Figure 1A for samples lyophi-
`lized in 10 mM potassium phosphate, sucrose-
`containing formulations in PBS exhibit little to no
`improvement in retention of native a-helix struc-
`ture at 1656 cm1, although they do retain more
`extended chain structure at 1677 and 1635 cm1.
`The IR spectra for 0.4±5 mg/mL rhIFN-g in PBS,
`pH 7.0, are shown in Figure 2B. Clearly, the
`secondary structure after lyophilization is not
`dependent on protein concentration. Overall, the
`samples lyophilized in PBS (Fig. 2B) retained
`more a-helical content than the samples lyophi-
`lized in 10 mM potassium phosphate (Fig. 1B).
`To determine whether damage during lyophi-
`lization was possible due to pH changes occurring
`during freezing, we measured apparent pH values
`reached during freezing of the formulations. The
`change from the initial pH 7.5 in potassium
`phosphate was greatest when polysorbate 20 was
`present, in which it increased to 8.24 (see Table 1).
`The presence of sucrose caused only a slight
`decrease in pH in the phosphate buffer. Overall,
`the changes in pH were small in the potassium
`phosphate buffer as compared with PBS. In PBS,
`the largest change in pH during freezing, a
`decrease of nearly 3 pH units, again occurred in
`
`Page 6
`
`

`

`DECREASING AGGREGATION OF rhIFN-g
`
`549
`
`Figure 3. Second-derivative IR spectra of rhIFN-g at
`the apparent pH of freeze concentration in potassium
`phosphate and PBS buffers. Legend: native control (Ð
`Ð, bold); 10 mM potassium phosphate, pH 7.5 (Ð Ð);
`potassium phosphate, pH 8.2 (± ± ±); PBS, pH 7.0 (- - -
`Ð); PBS, pH 5.2 (- - - - - -).
`
`the presence of polysorbate 20. In contrast, in
`buffer alone, the pH decreased about 2 units.
`To assess potential damage to the protein
`induced by pH changes of the buffers during
`freezing, the pH of solutions of rhIFN-g in each
`buffer alone were adjusted to those measured in
`the freeze-concentration experiment. The second-
`derivative IR spectra of rhIFN-g in these solutions
`are shown in Figure 3. Alkalinization of potas-
`sium phosphate buffer from pH 7.5 to 8.2 caused a
`slight reduction in the intensity of the a-helix
`band at 1656 cm1. In contrast, acidi®cation of
`PBS from pH 7.0 to 5.2 caused a dramatic
`reduction in the intensity of the a-helix band
`concomitant with appearance of new bands at
`1620 cm1 and 1692 cm1. These changes were
`unexpected because the protein is stable in 5 mM
`sodium succinate, pH 5.0, and maintains native
`
`Figure 2.
`(A) Second-derivative IR spectra of 0.4 mg/
`mL rhIFN-g lyophilized in PBS, pH 7. Formulations:
`PBS, pH 7 (- - - - - -); 5% sucrose in buffer (± ± ±); 0.03%
`polysorbate 20 in buffer (Ð Ð); 5% sucrose and 0.03%
`polysorbate 20 in buffer (- - - Ð); native control (Ð Ð,
`bold). (B) Second-derivative IR spectra of 0.4±5 mg/mL
`rhIFN-g lyophilized in PBS, pH 7.0. Symbols: 0.4 mg/
`mL (Ð Ð); 1mg/mL (± ± ±); 5 mg/mL (- - - - - -); native
`control (Ð Ð, bold).
`
`Table 1. Determination of Apparent Frozen pH for Both Potassium Phosphate and
`PBS Buffer Formulations
`
`Formulation
`
`No excipients
`5% Surcrose
`0.03% Polysorbate 20
`5% Sucrose/0.03% Polysorbate 20
`
`10 mM Potassium
`Phosphate, pH 7.5
`
`PBS, pH 7.0
`
`8.17
`7.23
`8.24
`7.18
`
`5.21
`5.70
`4.33
`5.77
`
`Based on calibration of pH standards 4, 7, and 10 at 0, 10, and 258C, the error in measurement
`was < 0.1% for 95% con®dence.
`
`JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 2, FEBRUARY 2002
`
`Page 7
`
`

`

`550
`
`WEBB ET AL.
`
`structure under these conditions.25,40 It is likely
`that the combination of high salt (152 mM) and
`low pH disrupt the helix dipoles resulting in
`protein denaturation. These bands at 1620 cm1
`and 1692 cm1 are indicative of intermolecular
`b-sheet associated with formation of protein
`aggregates,37,38 and have been used to monitor
`aggregation of rhIFN-g in the liquid state.6
`
`Reconstitution of the Lyophilized Formulations
`
`Formulations lyophilized in potassium phosphate
`and reconstituted in water showed either very
`little or no aggregation (Table 2). In samples in
`which aggregates were detected,
`inclusion of
`0.03% polysorbate 20 in the reconstitution solu-
`tion reduced aggregate levels. An exception
`occurred when rhIFN-g was lyophilized in the
`presence of 0.03% polysorbate 20. In this case,
`aggregation levels were unaffected by the further
`addition of polysorbate 20 to the reconstitution
`medium.
`When lyophilized samples of rhIFN-g were
`reconstituted with water alone, aggregates were
`present
`in all PBS formulations except
`the
`formulation containing both sucrose and polysor-
`bate 20, based on A350 and SEC results (Table 2).
`The greatest levels of aggregation were noted in
`samples lyophilized in PBS alone. The addition of
`sucrose to the PBS formulation reduced but did
`not eliminate aggregation upon reconstitution. In
`contrast, PBS formulations reconstituted with
`0.03% polysorbate 20 showed statistically insig-
`ni®cant levels of aggregation by protein recovery
`from SEC, and reduced light scattering at 350 nm.
`
`Mechanism of Stabilization by Polysorbate 20
`
`To determine the mechanism of stabilization by
`polysorbate 20, we investigated several possibi-
`lities. Polysorbate 20 may stabilize the native
`state of the protein by increasing its free energy of
`unfolding, or it may facilitate refolding to the
`native state. Alternatively, the surfactant may
`inhibit
`surface-induced denaturation during
`reconstitution. Another potential mechanism is
`the in¯uence of polysorbate 20 on the rate of
`reconstitution of the lyophilized protein. Each of
`these mechanisms was assessed in detail.
`
`Unfolding rhIFN-c in Urea
`
`The urea-unfolding curve for rhIFN-g was shifted
`to lower concentrations of urea in the presence of
`polysorbate 20 (Fig. 4). In both cases, the curves
`
`JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 91, NO. 2, FEBRUARY 2002
`
`showed sigmoidal, highly cooperative transitions
`typical of two-state processes. No precipitation of
`the protein occurred. A two-state, equilibrium
`model was assumed for thermodynamic calcula-
`tions. For the unfolding of rhIFN-g in PBS,
`DGˆ 60.6 2.0 kJ/mol was calculated versus
`53.5 3.2 kJ/mol in the presence of polysorbate
`20. These results suggested that polysorbate
`20 destabilized the protein under these conditions.
`
`Refolding rhIFN-c After Rapid
`Dilutions From Urea Solutions
`
`rhIFN-g was equilibrated in solutions of urea at
`concentrations of 3.5 and 5 M. These chaotrope
`concentrations correspond to fn & 0.5 and fn & 0,
`respectively, as shown by the unfolding curve in
`Figure 4. During rapid dilutions to 1 M urea, the
`introduction of polysorbate 20 increased A350/mg
`and decreased soluble protein remaining after the
`dilutions were completed (see Table 3). Appar-
`ently, polysorbate 20 did not facilitate refolding
`and instead fostered aggregation during refold-
`ing. In addition, higher concentrations of protein
`resulted in larger losses because of aggregation
`(Table 3).
`
`Surface-Induced Denaturation of
`rhIFN-c Resulting From Mild Agitation
`
`We hypothesized that polysorbate 20 might
`inhibit rhIFN-g aggregation during reconstitu-
`tion by competing with protein for access to air±
`water interfaces created during injection of
`reconstitution solution into vials. To test whether
`polysorbate 20 is effective at reducing potential
`denaturation at air/water interfaces, we exposed
`rhIFN-g to air±water interfaces by agitating
`solutions of rhIFN-g for 15 h. The percent of
`soluble protein recovered after agitation was
`invariant at both protein concentrations tested,
`in the presence or absence of 0.03% polysorbate
`20, and in solutions containing PBS or 15-fold
`diluted PBS (Fig. 5, inset). However, the inclusion
`of 0.1% polysorbate 20 yielded full recovery of the
`initial starting concentration of rhIFN-g after 15 h
`of agitation. After 15 h of agitation, A350/mg
`protein was higher in solutions containing 0.03%
`polysorbate 20 than in solutions without poly-
`sorbate 20 (Fig. 5). A350/mg protein was lower in
`solutions with lower ionic strength, i.e., 15-fold
`diluted PBS (Fig. 5), as well as in the solution
`containing 0.1% polysorbate 20. Therefore, the
`polysorbate 20 concentration used in the recon-
`stitution studies (0.03%) did not stabilize the
`
`Page 8
`
`

`

`DECREASING AGGREGATION OF rhIFN-g
`
`551
`
`aBuffertypes:Aˆ10mMpotassiumphosphate,pH7.5,andBˆPBS,pH7.0.ForsampleswithoutSECresults,proteinrecoveriesarecalculatedfromA280measurements
`
`recordedaftercentrifugation.RemainingrecoveriesarebasedonSECresults.Errorsare2standarddeviations.
`
`Ð
`
`Ð
`
`1.0
`0.2
`
`Ð
`
`Ð
`
`1.7
`1.4
`
`Ð
`
`Ð
`
`0.4
`0.0
`
`0.2
`0.1
`0.0
`0.3
`0.1
`0.0
`0.1
`0.0
`0.7
`0.8
`0.1
`0.1
`
`Ð
`
`Ð
`
`99.0
`99.8
`
`Ð
`
`Ð
`
`98.3
`98.6
`
`Ð
`
`Ð
`
`99.6
`
`100
`
`99.8
`99.9
`
`100
`
`99.7
`9