Journal of Pharmaceutical Sciences 105 (2016) 530e541

Contents lists available at ScienceDirect

Journal of Pharmaceutical Sciences

journal homepage: www.jpharmsci.org

Pharmaceutical Biotechnology

Addition of Monovalent Electrolytes to Improve Storage Stability of

Freeze-Dried Protein Formulations
Hiroshika Goshima 1, Kelly M. Forney-Stevens 1, Ming Liu 2, Ken K. Qian 3,
Madhusudan Tyagi 3, Marcus T. Cicerone 3, Michael J. Pikal 1, *
1
Department of Pharmaceutical Sciences, School of Pharmacy, University of Connecticut, Storrs, Connecticut 06269
2
Department of Nuclear Engineering, North Carolina State University, Raleigh, North Carolina 27695
3
National Institute of Standards and Technology, Gaithersburg, Maryland 20899

a r t i c l e i n f o a b s t r a c t

Article history: This study investigates the effect of low levels of electrolytes on storage stability in freeze-dried sucrose-
Received 17 June 2015 based protein formulations. Both bovine serum albumin and recombinant human serum albumin were
Revised 23 September 2015 freeze dried with sucrose and alkali halides (LiCl, NaCl, KCl, RbCl, and CsCl) at selected low levels. All
Accepted 5 October 2015
formulations were stored at 50
C and 65
C up to 2 months and then assayed for protein aggregation. The
Available online 30 December 2015
data demonstrate that low levels of LiCl and NaCl enhance stability. No obvious correlations with either
protein secondary structure or global dynamics (structural relaxation time) were found. However, good
Keywords:
correlations were found between stability and both free-volume hole size via positron annihilation
protein formulation
stability
lifetime spectroscopy (PALS) and fast dynamics by neutron scattering. Volume changes on mixing and
freeze drying the partial molal volume of salt were also studied in an effort to detect decreases in free volume. These
lyophilization data did not support the hypothesis that reduction in free volume was the primary mechanism for salt-
glass dynamics induced stabilization. Finally, a positive effect of postlyophilization annealing on stability was demon-
mobility strated. In summary, we find that small amounts of LiCl and NaCl significantly stabilize these proteins,
free volume which is a result at variance with conventional formulation wisdom.
PALS © 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
mean square displacement
neutron scattering

Introduction Tg but yet stabilizes disaccharide-based protein formulations sug-

Stability has historically been a challenge in protein formulation
development, and refrigerated storage is a standard requirement.
Freeze drying is commonly used to stabilize a protein product
nominally because drying removes the reactant water and signifi-
cantly immobilizes the protein system, but other stabilization
mechanisms may also operate.1 The first choice for the protein-
stabilizing agent is frequently sucrose.2-4 However, addition of
other components to the disaccharide-based formulation has also
been investigated, such as polyols, polymers, and amino acids.5-19
Nominally, polymers would seem to be good stabilizer candidates
because they increase the glass transition temperature (Tg).1,6-8
However, whether Tg is a reliable predictor of stability well below
Tg is doubtful.6 Indeed, the observation that addition of low levels of
several small molecules, such as glycerol17 and sorbitol,5 decreases
* Correspondence to: Michael J. Pikal (Telephone: þ1 860-486-3202).
E-mail address: michael.pikal@uconn.edu (M.J. Pikal).
gests that it is perhaps local or “fast dynamics” that is a better
predictor of stability than Tg or global dynamics. The addition of
such small molecules seems to antiplasticize “fast dynamics” and
therefore stabilizes.17
Electrolytes are relatively common in freeze-dried formulations,
as buffers and often as NaCl arising from prior purification steps.9
However, because NaCl and buffer salts depress the collapse tem-
perature,11 use of salts in formulations is normally restricted to very
low levels. In this study, we investigate addition of alkali chlorides,
such as sodium chloride, as stability enhancers in sucrose-based
formulations. Currently, the expected effect of an electrolyte on
stabilization is unclear. Some studies suggest that NaCl destabilizes
factor VIII9 and human growth hormone.10 However, another
report recommended NaCl as a bulking agent for factor VIII, sug-
gesting that at least crystalline NaCl did not significantly damage
stability.7
The primary objective of this study was to investigate the effects
of amorphous electrolytes on storage stability in sucrose-based
protein formulations. Electrolytes studied were the series of alkali

http://dx.doi.org/10.1016/j.xphs.2015.10.004
0022-3549/© 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541
metal chlorides, that is, LiCl, NaCl, KCl, RbCl, and CsCl. To include
divalent cations and anions was judged inappropriate for the pur-
poses of this study because those ions often have protein-specific
interactions that may significantly impact protein conformation
and therefore impact stability. For example, it has been reported
that the storage stability of phosphofructokinase has been
improved by adding Mg2þ,12 and coagulation factor VIII, somato-
tropin, and human serum albumin are affected by Ca2þ, Zn2þ, and
Cu2þ, respectively.13 Anions such as I and SCN tend to impact
protein structure according to the Hofmeister series, at least in
aqueous solution.14,15 Hence, we used the chloride anion and
monovalent cations, that is, Liþ, Naþ, Kþ, Rbþ, and Csþ, for the
purpose of studying “general effects” of ions on protein stability in
freeze-dried solids.
The secondary objective was to better understand the role of
electrolytes in stabilization based on two hypotheses of protein
stabilization in disaccharide matrices; one commonly discussed
mechanism is a thermodynamic stabilization mechanism, acting
by stabilizing the native conformation by having the stabilizer
hydrogen bond to the protein as does water and is termed the
“water substitution mechanism.” Here, stabilization results from
maintaining “native-like” structure. In the present work, corre-
lations between “structure” and stability were studied using
infrared spectroscopy.20,21 The other is a purely kinetic stabili-
zation mechanism which acts by suppressing molecular mobility,
often denoted the “glass dynamics mechanism.”3,5 Mobility was
studied by evaluation of both “global mobility,” as represented by
the structural relaxation time measured by isothermal calorim-
etry22,23 and by evaluation of “fast dynamics” by analysis of
neutron scattering.16,17 A number of protein systems have been
studied using neutron scattering, by which one can characterize
the local dynamics in the glassy solid. Generally, it has been
found that this local dynamics or “fast dynamics” correlates well
with storage stability of the protein in glassy systems.16-19 In
addition, recently, high-precision density measurement for glassy
materials was used to characterize the system free volume, which
is potentially a predictor of molecular mobility and therefore,
stability.24,25 Good correlations were found between free volume
(i.e., density) and storage stability.3 Thus, one might expect pre-
cision density measurements may be useful in understanding the
role of alkali metal salts in stabilization of proteins in the glassy
state. To interpret variations in density in terms of free volume,
we applied specific volume-based approaches; (1) volume change
on mixing and (2) partial molal volume of solute (i.e., salt) to
evaluate free-volume changes in glassy formulations on addition
of electrolyte. To directly measure size of the “holes” constituting
free volume, a positron annihilation lifetime spectroscopy (PALS)
study26-35 was carried out. In addition, we investigated the effect
of “post-lyophilization-annealing,” which can be executed after
the freeze-drying process by heating dried cakes to induce
structural relaxation36,37 or densification,25 resulting in
enhancement of storage stability. This thermal treatment has
been found to be effective not only for stabilization of small
molecules36,38 but also for stabilization of an IgG protein,39 so it
was deemed of interest to determine if annealing had any impact
on stability of formulations containing low levels of added salt.
As model proteins, bovine serum albumin (BSA) was selected
because that is a well-investigated protein, and large amounts of
protein were readily available for the experiments. Additionally, as
a highly purified protein was desirable in the protein aggregation
stability studies, recombinant human serum albumin (rHSA) was
used for the purpose of “validation” of the BSA results. Monovalent
electrolytes in various amounts were colyophilized with a 1:1
weight ratio of protein to sucrose. We found that small amounts of
LiCl and NaCl, that remain amorphous, significantly stabilize these
531
proteins, a result that is at variance with conventional formulation
wisdom.
Materials and Methods
Materials
Bovine serum albumin, sucrose, potassium chloride, rubidium
chloride, cesium chloride, potassium phosphate, and sodium sul-
fate were purchased from Sigma-Aldrich (St. Louis, MO). Lithium
chloride, sodium chloride, sodium phosphate monobasic anhy-
drous, sodium phosphate dibasic heptahydrate, and potassium
phosphate monobasic were purchased from Fisher Scientific
(Pittsburgh, PA). Recombinant HSA in sterile ultrahigh purity grade
was a product of Albumin Bioscience (Huntsville, AL).
Preparation of Freeze-Dried Protein Formulations
BSA and rHSA were dialyzed against 5-mmol/L sodium phos-
phate buffer at pH 7 using a membrane with MWCO: 6-8,000
(Spectrum Laboratories, Dominguez, CA). Protein concentrations
were determined by UV-VIS spectroscopy (Cary 50 Bio, Varian) at
280 nm and then filtered by 0.22 mm polyether sulfone Millex-GP
(Millipore, Billerica, MA) to prepare the protein-buffered solu-
tions of BSA and rHSA at 50 mg/mL and 5 mg/mL, respectively.
Separately, sucrose and the electrolyte, that is, LiCl, NaCl, KCl, RbCl,
or CsCl, were dissolved and also filtered before mixing with the
protein solution to prepare the protein-sucrose-electrolyte solu-
tions at 1:1 in weight ratio for protein:sucrose with various weight
fractions of electrolyte. All percentages described in this article
were the weight fraction or mole fraction relative to total solute
components. After dispensed into 5-mL tubing clear glass vials
(Schott, Lebanon, PA) and semistoppered with Daikyo Fluoroteck
stoppers (West Pharmaceutical, Lititz, PA), lyophilization was car-
ried out using a laboratory freeze dryer (Dura-Stop, SP Scientific).
Samples were frozen by cooling to 40
C and holding for 1 h, fol-
lowed by setting the shelf temperature to 33
C and the chamber
pressure to 60 mTorr (8 Pa) for primary drying. Although, given the
collapse temperature (Tc) is usually about 2
C higher than the glass
0
transition temperature of freeze concentrate (Tg ),40 primary drying
0
conditions for each formulation were designed based on each Tg in
an attempt to maintain the product temperature below the
assumed Tc, there were some cases where the product temperature
0
did increase above the Tg ; yet no sign of collapse or loss of quality
was observed even in the “worst case” of 9.0% LiCl-BSA and/or
0
sucrose formulation with Tg of 47
C, where a maximum product


temperature of 37 C was observed. Such unconventional phe-
nomena have also been observed in recent works.41,42 That is,
0
collapse temperature and Tg can become high, and one may
significantly exceed the collapse temperature measured by
freeze-drying microscopy and still not observe collapse when
freeze-drying in vials. Secondary drying was conducted at 40
C for
4 h before backfilling with nitrogen and sealing.
All samples freeze dried without visual collapse, and the water
content was determined to be less than 1% by Karl Fischer titration
(756 KF coulometer, Metrohm).
The glass transition temperatures (Tg) and the changes in heat
capacity at Tg (DCp) were measured by Modulated DSC (Q1000, TA
Instrument) at 1
C/min with modulating ±0.5
C every 100 s.
Samples were compressed into thin disks and placed in hermeti-
cally sealed aluminum pans. All sample handling was carried out in
a dry bag continuously purged with dry nitrogen or air (relative
humidity <3%). Glass transition temperatures were taken as the
midpoint of the heat capacity shift. To confirm amorphous before
and after storage, powder X-ray diffraction (PXRD; AXS D2 Phaser,
532 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541

Bruker) was used because cubic crystal structures, such as sodium Table 1
chloride, do not show birefringence with polarized light Density of the BSA-Sucrose (1:1) Formulation

microscopy. Value (g/cm3) SD (g/cm3) Average (g/cm3) Standard Error (g/cm3)

1.3849 0.0034
Stability Studies 1.3859 0.0010
1.4051 0.0026
1.3820 0.0011 1.3915 0.0036
The BSA samples were stored at 50
C and 65
C for 2 months and
1.3980 0.0012
1 month, respectively, whereas the rHSA samples were stored at 1.3930 0.0016
50
C for 1.5 months. Samples were taken at selected times for assay.
For the annealing study, freeze-dried rHSA samples were annealed
at 75
C for 13 h before storage, and stored at 50
C for 1.5 months
ampoule, which was tightly sealed to avoid moisture exposure. The
with nonannealed samples as a reference or “control.” Storage
sample ampoule was introduced into the calorimeter controlled at
stability was evaluated by determining the increase in protein
40
C and the heat flow was measured for 72 h. The modified
aggregate as a function of the storage time. Size exclusion chro-
stretched exponential (MSE) function was fit to the data instead of
matography was performed using a TSKgel G2000SWXL column
using the Kohlrausch-Williams-Watts (KWW) equation to avoid
(7.8 mmID  30 cm, 5 mm, Tosoh Bioscience) to determine the
errors near t ¼ 0.22,23 The MSE equation can be expressed for power
percentage of aggregate at each time point. Freeze-dried samples
(heat flow), P (mW/g), with time in hours, as follows.
were reconstituted with purified water and appropriately diluted
and filtered before injection. All reconstituted samples were visu-      
ally clear after reconstitution. A mobile phase of sodium phosphate Tg  T $DCp b$t t b2
P ¼ 277:8  $ 1þ $ 1þ
buffer at pH 7 with NaCl was used for BSA samples, whereas po- t0 t1 t1
"    b1 #
tassium phosphate buffer at pH 7 with Na2SO4 was used for rHSA t t
analysis. The flow rate of 1 mL/min and the detection system by UV exp   1þ (2)
t0 t1
absorbance at 280 nm were used. The precision for a given sample
as well as between replicate samples was ±0.1%.
1=b ðb1Þ=b
tD ¼ t0  t1 (3)
Characterization of Protein Structure in Freeze-Dried Solid (Fourier
transform infrared spectroscopy [FTIR]) We evaluated values of effective KWW-stretched relaxation
time, tbD as this parameter is more robust and not as sensitive to
Secondary structure of the protein in the solid state was inves- systematic error as the effective KWW structural relaxation time,
tigated using a Nicolet Magna-IR system 560 spectrometer. Atten- tD.22
uated total reflectance (ATR) spectra were collected using a single
bounce, diamond crystal accessory because direct observation
without sample preparation is possible, and strong peak intensity is
obtained with a relatively small amount of sample.20 Freeze-dried
samples were moderately crushed under low-humidity condition Density Measurement by Gas Pycnometer
(<3% relative humidity) and quickly placed on the stage and
scanned 128 times at a resolution of 4 cm1. As the fast analysis Densities of freeze-dried samples were measured using an
time did not allow objectionable uptake of water, scans were run in AccuPyc 1330 helium pycnometer (Micromeritics, Norcross, GA) as
ambient condition. Raw spectra were baseline corrected using the described in previous literature.45 Calibration with a standard iron
Advanced ATR Correction.43 Second derivative spectra were created sphere of known volume followed by a verification run using
using OMNIC software (Thermo Nicolet, Madison, WI); the spectra crystalline sucrose of known density was performed before sample
were baseline corrected and area normalized over the amide I re- measurement for each day. The measured value was checked for
gion (1600-1700 cm1).20 The spectrum of native structure was consistency with the density of crystalline sucrose, 1.5854 ± 0.0048
evaluated by running the protein in a buffer solution. The similarity g/cm3.25 Otherwise, the sample cup which was dried overnight at
of spectra to the native was quantitated by using the area of overlap 65
C to completely eliminate moisture was exchanged, and the
method, as described in the literature.21 The width of the alpha runs were repeated. Once the accuracy of measurement was veri-
helix peak at half height, or full width at half maximum (FWHM), as fied, the sample powder was loaded in a sample cup, the inner and
a measure of structure that was recently suggested44 was calcu- outer surfaces of which were cleanly wiped to insure freedom from
lated based on curve fitting of the peak to the Gaussian function: particulates remaining. As freeze-dried materials were mostly
bulky and fluffy, repeated compression of the powder with a metal
( ) qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðx  bÞ2 rod was needed to load the desired amount of the sample. After
when; f ðxÞ ¼ a$exp  2
; FWHM ¼ 2c$ 2$lnð2Þ; confirming all surfaces of the cup being clean, the sample was
2c
placed in the chamber where helium gas of 134 kPa (19.5 psig) was
(1) filled and then equilibrated till not more than 34 Pa/min (0.005
psig/min) to give a density value by taking an average of 20 mea-
where a, b, and c represent fitting parameters. surements. The density value for amorphous sucrose was obtained
as 1.4988 ± 0.0021 g/cm3, which was consistent with the published
Investigation of Global Glass Dynamics by the Thermal Activity data.25 In addition, the density of the “base” BSA-sucrose formu-
Monitor (TAM) lation (no salt) was measured six times for different vials. As
summarized in Table 1, we obtained the average value of
Enthalpy relaxation is a measure of global dynamics and was 1.3915 g/cm3, with standard error of 0.0036, which was the same
directly measured by isothermal calorimetry (TAM model 2277, precision as found in the previous study (standard error ¼ 0.0036
Thermometric, Sweden). Working in a dry box, samples of about for amorphous sucrose),25 suggesting that the measurement pro-
200 mg were appropriately crushed and put into the stainless steel cedures are reliable for our protein-sucrose materials.
 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541
Analysis of Density Values to Characterize Free Volume
When comparing densities in a series of samples with variable
chemical composition with the intent to compare trends in free
volume, much if not most of the density variation can arise from
variation in atomic density, particularly for systems containing
heavy metal atoms, such as cesium. To avoid the variable atomic
density problem, we use density data to evaluate volume changes
on mixing and apparent molal volumes and examine these data for
trends that might suggest changes in free volume, as previously
discussed.45 With specific volume denoted as, v0, and given by the
0 the density (Eq. 6), to generate the partial specific volume. The
reciprocal of density, d, the volume change on mixing, Dvmix , may
be written,
X   
0 0
Dvmix ¼ v0  v0 ¼ ð1=dÞ  wi $ 1 di;0
i
0
where, v0 is the sum of the specific volumes of the pure compo-
nents i, with weight fraction of wi and its density of di,0. The vol-
umes of the atoms themselves, or van der Waals volumes, will then
cancel when forming the difference between final volume of the
mixture and the sum of the volumes of the individual components
in their pure states. Therefore, the change in specific volume on
mixing of two components may be related to free volume, v0 F by,
0 0 0 0
Dvmix ¼ v F  w1 $v1;0
F F
 w2 $v2;0
0
where, vi;0F denotes specific free volume of component i in pure
state, and thus, the free volume in the mixture may contain changes
in free volume of both components when mixed. If free volume is
reduced in the mixture relative to sum of the pure components, the
volume change on mixing should be negative. In our study, we take
component 1 as a BSA-sucrose matrix (or “solvent”) and compo-
nent 2 as a salt (or “solute”). In addition, at least as a first approx-
0
imation, we can neglect the v2;0 F because the pure salt is the
crystalline form with (nominally) essentially zero free volume. This
means loss of free volume in the systems corresponds to the loss of
free volume of the protein-disaccharide matrix, which should
produce a less mobile matrix and therefore, perhaps provide sta-
bilization. One might argue that we need to use the specific volume
of the pure amorphous salt instead of the crystalline salt because
the salt is not crystalline in the mixture. However, such data are not
readily available. Using amorphous salt as the “baseline” would of
course systematically lower volume changes on mixing.
As another method to characterize free-volume changes from
measured densities, we may evaluate the partial molal volume of
the “solute”, or in our case, the salt dissolved in the protein:sucrose
solvent. The partial molal volume is the volume change of the
system when one mole of solute is added to an infinite amount of
solvent. A comparison of the partial molal volume to the molar
volume of pure salt then allows an assessment of loss of free vol-
ume in the solvent on addition of solute. We may calculate apparent
specific volume of component 2 (i.e., salt) in the mixture, denoted
0
Fv , with total volume of the mixture (V), volume of component 1 in
pure state (V1,0), and mass of component 2 (m2).
0 V  V1;0
Fv ¼ ; or when the total mass is 1 gram;
m2
   spectrometer was operated in the fixed-window scan mode with
0 ð1=dÞ  w1 , 1 d1;0
Fv ¼
w2
0
The partial specific volume of salt ðV2 Þ is calculated from the
apparent specific volume data from the relationship,
533
0
0 vFv 0 0
V2 ¼ w2 $ þ Fv ; and when Fv
vw2
0 0
¼ intercept þ slope$lnðw2 Þ; V2 ¼ slope þ Fv (7)
If there is a loss of free volume by adding a salt, the apparent
specific volume should be less than the molar volume of the pure
salt, which we take as the salt in the crystalline form. To generate
partial molal volume data, Equation 7 was used. The slope of the
linear plot of apparent specific volume versus ln(w2) was com-
bined with the value of apparent specific volume calculated from
partial molal volume is simply the product of the partial specific
volume and the molecular weight of the solute (i.e., the salt).
Uncertainties in the calculated partial molal volumes were esti-
(4) mated from the standard error in the slope and the standard
error in the calculated apparent specific volume, using standard
propagation of errors methodology. As previously noted by
Chieng,45 the relation between partial molal volume and volume
change on mixing may be given by,
0
0 0 0 Dvmix
fv ¼ v2* þ v2;0
F
þ (8)
w2
That is, the qualitative trends for apparent specific volume and
volume change on mixing must be the same.
(5)
Positron Annihilation Lifetime Spectroscopy Measurement
PALS can be used as a means of direct measurement of size of
the free-volume “holes” in the amorphous solid. Recent studies
have been expanding applications of PALS for pharmaceutical
materials.29,31 The measurement is based on analyzing lifetime and
intensity of positronium, which in principle are sensitive to size and
concentration of free volume in a material.28 PALS methodology
and procedure of measurement were briefly discussed previ-
ously.45 Each of the collected spectra was best fitted with 3 positron
lifetimes with the third lifetime and its intensity attributed to the
annihilation of positronium. The mean pore diameter was calcu-
lated based on the Tao-Eldrup model which uses the assumption of
spherical microvacancies.26 The measurement was repeated 3
times per formulation.
Neutron-Scattering Measurements
The high-flux backscattering spectrometer at NIST Center for
Neutron Research was used to measure <u2> as a function of
temperature. In the neutron-backscattering experiment, samples
were probed by incident neutrons with wavelength of 6.271 Å
and kinetic energy of 2.08 meV with 0.85 meV full width at half
maximum energy resolution. Powders of freeze-dried samples at
approximately 400 mg each were spread onto aluminum foil
and then loaded into aluminum sample cells. The sample cells
were hermetically sealed with indium o-rings and tightened
with aluminum screws. The procedure was performed inside of
a helium glove box at room temperature, where the relative
humidity was approximately 0.1% to prevent absorption of
moisture by lyophilized powders. The high-flux backscattering
(6) the samples cooled from 324 K (51
C) to 4 K at 0.7 K per
minute. The elastic scattering intensity was analyzed to extract
<u2>, same as previously described.45 All samples were kept
at 20
C before use.
534 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541
Figure 1. X-ray diffraction patterns of the BSA-sucrose-NaCl formulations with
different amounts of electrolyte and after different storage temperatures after a month
(1M); the crystalline peak at about 32
indicates crystallization of NaCl; the percent-
ages and the temperatures give the weight fraction of NaCl and the storage temper-
ature, respectively.
Results and Discussion
Crystallization of Electrolyte
Because the electrolyte needs to be molecularly dispersed (here,
likely as an ion-pair perhaps involving charged groups on the
protein) in a single amorphous phase composed of protein and
sucrose to impact stability, it was necessary to examine the samples
for salt crystallization, both before and after storage. Figure 1 shows
the XRD patterns for 12.9% and 16.6% NaCl in the BSA formulations
after storage. Although no crystalline peak was found in BSA:NaCl
systems before storage, we observed crystalline NaCl in 12.9% NaCl
after 65
C, and in 16.6% NaCl at both 50
C and 65
C after one
month. Using peak intensity as a measure of degree of crystalliza-
tion, we conclude that crystallization is most complete in the 16.6%
sample at 65
C, followed by the 16.6% sample at 50
C, with lower
but still significant crystallization in the 12.9% sample at 65
C.
Greater crystallization at higher concentration and higher tem-
perature is expected. For 12.2% or more KCl in the BSA formulations,
crystallization of KCl was observed even before storage (data not
shown). All other formulations were confirmed as X-ray amor-
phous initially and throughout storage (data not shown).
Effect of Electrolyte on Protein Aggregation
Protein aggregation for all the formulations followed square root
of time kinetics, as normally seen in amorphous solids.5 Figure 2
illustrates the linearity in the plot of the percentage of aggregate
as a function of the square root of storage time. The rate constants
of aggregation were determined by fitting the square root of time
degradation equation to the stability data, and the resulting rate
constants are shown in Figure 3 as a function of salt content. With
BSA formulations, stability at 50
C was improved by adding LiCl,
Figure 2. A plot of the linearity observed for aggregation behavior as a function of
square root of time (n ¼ 1); open circle and broken line, rHSA-sucrose-CsCl (2.4%)
formulation stored at 50
C; filled circle and solid line, rHSA-sucrose formulation (no
salt) stored at 50
C; and square and thin line, HSA-sucrose-NaCl (11%) formulation
stored at 50
C. Error bars are not given because only one independent measurement
was made for each time point due to limited sample available. Assay errors are nor-
mally about ±0.1%.
NaCl, and KCl, and at 65
C, both LiCl and NaCl significantly
improved stability. As the fraction of electrolyte increased, stability
monotonically improved until about 10% for LiCl, NaCl, and KCl. For
KCl at 50
C and NaCl at 65
C, the data show a well-defined mini-
mum in rate constant around 10% salt (relative to total solids). We
suggest the lack of a large stabilization effect above 12.2% with KCl
is due, at least in part, to KCl crystallization both initially and during
the stability test. Crystallization likely also explains the loss of
stabilization noted for the 16.6% NaCl sample at 65
C (Fig. 3b). It is
obvious that RbCl and CsCl had no positive effect on stabilization.
The level of aggregation in our starting BSA was about 12%, and
to ensure that this high level of initial aggregation does not seri-
ously impact the rate constants for further aggregation and impact
the conclusions outlined previously, we performed equivalent
stability studies with the highly purified protein, rHSA, where the
initial aggregate was 0.36% (SD ¼ 0.04), to provide “validation” of
our surprising results with BSA, indicating that salt addition can
stabilize proteins. The rate constants obtained for rHSA were
similar to those obtained for BSA, and the qualitative conclusions
are much the same; addition of salts can stabilize. The modest
difference in rate constants and maximum stabilization obtained
may be due to the increased level of aggregation in the starting BSA
sample or may simply reflect the difference between the 2 proteins.
Thus, we do conclude that, at least for these sucrose formulations of
serum albumins, addition of monovalent electrolyte such as LiCl
and NaCl can improve storage stability. Because all salts investi-
gated were chloride salts, the stability differences obviously reflect
differences in the cations.
Destabilization Above Tg
Attention should be paid to the Tg in amorphous formulations
because the mobility and likely also stability will dramatically
change for samples stored near or above the glass transition re-
gion.1,8 Thus, the storage temperature is normally maintained
significantly below Tg. The effect of electrolyte on the Tg in the solid
 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541 535

Figure 3. Effect of electrolyte on storage stability by plotting the rate constant against the weight fraction of electrolyte; (a) BSA-sucrose-electrolyte formulations stored at 50
C for
2 months, (b) BSA-sucrose-electrolyte formulations stored at 65
C for 1 month, (c) rHSA-sucrose-electrolyte formulations stored at 50
C for one and half a month; Error bars show
standard error of the rate constant. LiCl (diamond); NaCl (square); KCl (triangle); RbCl (filled circle); CsCl (open circle).

state is given by the data in Table 2. Note that the water contents are
all very low, less than 1%, so water contenteinduced variations in Tg
between samples are not significant in the context of this discus-
sion. Although LiCl increased the Tg, the impact of NaCl is minimal.
Other electrolytes reduced Tg, the amount of decrease depending
on the weight fraction. For KCl at 9.1%, the midpoint Tg was
decreased to 74
C. Here, the onset Tg was 66
C, which is very close
to the storage temperature of 65
C. These trends may arise from
charge density differences between the cations and thus reflect the
impact of electrostatic interactions on Tg. It seems likely that the
poor stability of the KCl systems at 9% KCl and higher is a result of
the proximity to the glass transition and the enhanced mobility at
these conditions. As previously demonstrated, Tg is a significant
factor to be considered for design and interpretation of storage
stability studies.1,8 In the following discussions, we focus on con-
ditions where the formulations are stored well below Tg.
Secondary Structure of Protein by FTIR
By comparing the similarity of the protein structure in the solid
to the native structure, the effect of electrolyte on the protein
conformation in the solid was investigated. Figure 4 shows second
derivative spectra for the rHSA formulations. Analysis of the spec-
troscopic data and stability data did not reveal any relation
between the protein structure and the stabilizing effect of elec-
trolyte. As summarized in Table 2, there were no large differences
between the electrolytes; the range of variation in the area of
overlap was 0.74-0.82 for BSA formulations and 0.70-0.79 for rHSA
formulations, with no obvious correlation with protein stability. As
well, the width at half height of the alpha helix peak band (FWHM
in Table 2) did not indicate any systematic relationship between the
conformation and the stability; the FWHM varied “randomly” in
the range of 12.1-15.2 cm1 for the different formulations.
Global Mobility by TAM
The effect of electrolyte on global mobility below Tg has been
studied in the BSA formulations. Stretched relaxation times ob-
tained using the MSE equation are summarized in Table 3. Before
data analysis, it was confirmed that no crystallization of electrolyte
occurred during the measurement (i.e., at 40
C for 72 h). The un-
certainties given are standard errors evaluated from the replicate
runs, normally duplicates. Most of the differences are within the
sum of standard errors and are obviously not significant. The
relaxation time showed no significant trend with increasing salt
with only two possible exceptions. For both LiCl and KCl at 9%, there
seemed to be a small decrease in relaxation time that corresponds
to a small systematic increase in molecular mobility. However,
536 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541

Table 2
Effect of Electrolytes on the Glass Transition Temperatures (Tg) and the Changes of Heat Capacity at Tg (DCp) for the Protein Formulation Including Sucrose and Different
Electrolytes

Formulation Electrolyte (%w/w) Tga (
C) DCpb (J/g$K) Area of Overlap FWHM (cm1)

BSA and sucrose (1:1) N/A 0 90 0.38 0.77 13.7

LiCl 2.4 91 0.41 0.79 12.4
4.8 94 0.40 0.80 12.9
6.9 94 0.40 e e
9.0 100 0.35 0.82 14.1
NaCl 2.4 94 0.31 0.81 13.1
4.8 92 0.39 0.76 12.2
6.9 88 0.45 e e
9.0 95 0.50 0.80 13.2
11.1 93 0.40 0.77 13.2
13.0 97 0.40 0.78 13.2
16.6 91 0.31 0.79 13.6
KCl 2.4 87 0.49 0.74 12.8
4.8 86 0.40 0.77 12.8
6.9 78 0.42 e e
9.1 74 0.39 e e
RbCl 2.4 93 0.30 e e
9.0 86 0.34 e e
CsCl 1.0 93 0.36 0.75 12.1
2.4 87 0.32 0.81 13.9
4.7 89 0.32 0.76 12.6
9.1 83 0.26 0.78 12.2

Formulation Electrolyte (%w/w) Tgc (
C) DCpc (J/g$K) Area of Overlap FWHM (cm1)

rHSA and sucrose (1:1) N/A 0 91 0.28 0.76 12.1

LiCl 2.4 88 0.33 0.78 13.0
4.7 89 0.29 0.71 13.0
7.0 95 0.28 0.76 13.3
NaCl 2.4 89 0.37 0.73 13.3
4.7 89 0.29 0.71 12.9
7.0 91 0.35 0.70 12.6
9.0 89 0.37 0.70 14.6
11.0 82 0.35 0.79 13.1
KCl 2.4 83 0.41 0.76 12.6
6.9 82 0.33 0.78 12.9
RbCl 2.4 87 0.30 0.71 15.2
6.9 83 0.27 0.71 15.1
CsCl 2.4 88 0.30 0.73 13.1
6.9 82 0.28 0.79 12.9
a
The maximum standard deviation was ±2.6
C.
b
The maximum standard deviation was ±0.50 J/g$K.
c
Analysis was not repeated due to limited sample amount.

these changes are very minor, as well as uncorrelated with stability,
likely due to scatter in the data, and we conclude that the stabili-
zation effect by electrolyte cannot be attributed to changes in global
dynamics. Recent studies suggest that global mobility in the glass
matrix often does not correlate well with global mobility, at least
well below the Tg.3,5
Effect of Electrolyte on Density and Its Interpretation as Free-Volume
Change
Table 4 summarizes the density data, the volume change on
mixing the salt with the base BSA-sucrose system, true, and the
calculated partial molal volumes. As expected, density increased as
the amount of salt increased because adding salt means adding
atoms with high atomic density.
To characterize changes in free volumes, we examined volume
change on mixing and partial molal volume. Volume changes on
mixing are plotted against molality in Figure 5a. Volume change
on mixing should be negative if the free volume of the protein-
disaccharide matrix decreases with addition of salt. Values are
very small (within almost ±0.010 cc/g), with an apparent slight
0
“V-shape” trend. However, given the definition of Dvmix , this
value must approach zero as the concentration of added solute
(i.e., salt) approaches zero. This appears not to be the case in
Figure 5a, particularly for LiCl, NaCl, and KCl. Because all values
must be zero at zero molal, the shape of the curve must actually
be “S-shaped,” with a maximum around 0.5 molal and a distinct
minimum around 1.5-2.0 molal, at least for LiCl, NaCl, and KCl. Of
course, uncertainty in the data is significant, and had we used
0
amorphous densities for the pure salts to calculate Dvmix , the
entire curve would be displaced lower (amorphous densities are
not available). However, this complex shape with both a
maximum and minimum would not be changed and therefore
seems to be real. Evidently, a low concentration of salt introduces
some free volume, but then higher levels decrease free volume
until the minimum. Part of the explanation of the trend past the
maximum at about 0.5 mol may involve the concept of two types
of free volume, as discussed for polymers32-35; one type of free
volume is multidimensionally interconnected spaces, where the
size becomes relatively large. This type of free volume is observed
in polymer crystals and termed “interstitial free volume.” The
other appears as many discrete and irregularly shaped holes due
to disordered packing in the amorphous phase. That is usually
called “local free volume,” also denoted as “holes.” The scale of
the local free volume is subnanometer, being too small to
contribute significantly to the transport of whole molecules, but
some holes can affect the mobility of chain segments of the
polymer.32,33 We note that the local free volume has been studied
 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541 537

Table 3
Effect of Electrolyte on the Structural Relaxation Time in the BSA-Sucrose-Electrolyte
Formulation at 40
C, Best Fit by MSE Equation

Electrolyte %w/w b tbD

None 0.0 0.26 22 ± 3

LiCl 2.4 0.22 25 ± 5
4.7 0.26 21 ± 2
6.9 0.25 19 ± 3
9.0 0.23 17 ± 0.1
NaCl 2.4 0.31 21 ± 3
4.7 0.23 25 ± 4
6.9 0.23 20 ± 3
KCl 6.9 0.30 17 ± 0.5
9.1 0.28 13 ± 1
RbCl 2.4 0.27 22 ± 4
9.0 0.28 20 ± 3
CsCl 1.0 0.23 24 ± 2
4.7 0.33 24 ± 1

mixing and partial molal volume are related by Equations 7 and 8,

Figure 4. Second derivative FTIR spectra of rHSA. Native structure refers to the spec-
trum of the protein in phosphate buffer solution at pH 7.0. All other samples were
freeze-dried solid of rHSA-sucrose-electrolyte formulation; the fraction and the sort of
the electrolyte were written in the graph.
using PALS.32-35 In the next section, we show some PALS data
obtained for three of our formulations.
As seen in Figure 5b, the partial molal volume of salt decreases
as salt is added for all systems, and at least for LiCl, NaCl, and KCl,
passes through a shallow minimum around 1.5-2.0 mol, similar to
the volume changes on mixing. Because the volume change on
and of course the same density data are used to evaluate each
0
quantity, similarity in trends for Dvmix and partial molal volume are
expected. Partial molal volume is volume occupied by one mole of
salt when added to an “infinite” volume of the BSA-sucrose system.
Partial molal volume may be less than the molar volume of crys-
talline salt (with zero free volume) if free volume of system is
decreased on addition of salt. Around the minimum in partial molal
volume, for LiCl, NaCl, and KCl, the partial molal volume is less than
the molar volume of the crystalline salt. However, for RbCl and CsCl,
the partial molal volumes decrease below the crystalline molar
volume at much lower concentrations. Figure 6 compares partial
molal volumes, molar volumes of crystal and (estimated) glassy
salt, and stability at 9% salt, which is near the concentration of
maximum stabilization for those salts that stabilize. Density of the
glass was estimated at 17% less than the corresponding crystal
density, as for silica versus quartz. Partial molal volumes for LiCl,

Table 4
Effect of Electrolyte on Density, Volume Change on Mixing, and Apparent Specific Volume of Electrolyte, in BSA:Sucrose Formulations; the Measurement Was Repeated
At Least 2 Times
0 0
Formulation; BSA:Sucrose (¼1:1) :Salt Salt (wt%) Density (g/cm3) Dvmix (cc/g) Fv (cc/g)

Average SD ±SD ±SD

N/A 0 1.3915 0.0089 e e

LiCl 2.4 1.3814 0.0069 0.008 ± 0.004 26 ± 0.3
4.8 1.4025 e 0.003 14
6.9 1.4168 0.0010 0.001 12 ± 0.3
9.0 1.4089 0.0057 0.010 ± 0.003 16 ± 1.4
NaCl 2.4 1.3852 0.0001 0.006 36 ± 0.2
4.8 1.3954 0.0004 0.007 30 ± 0.3
6.9 1.4284 0.0007 0.004 18 ± 0.3
9.0 1.4328 0.0002 0.000 21 ± 0.1
11.1 1.4329 0.0069 0.005 ± 0.003 23 ± 1.8
13.0 1.4377 0.0052 0.007 ± 0.003 24 ± 1.1
16.6 1.4460 0.0026 0.013 ± 0.001 25 ± 0.4
KCl 2.4 1.3905 e 0.003 20
4.8 1.3990 0.0088 0.003 ± 0.004 19 ± 3.5
6.9 1.4283 0.0057 0.006 ± 0.003 13 ± 1.5
9.1 1.4392 0.0004 0.007 13 ± 0.1
12.2 1.4583 0.0004 0.009 13 ± 0.1
14.8 1.4614 0.0016 0.005 ± 0.001 15 ± 0.2
16.7 1.4469 e 0.006 17
RbCl 2.4 1.4036 0.0021 0.001 ± 0.001 40 ± 5.3
4.8 1.4231 0.0054 0.002 ± 0.003 38 ± 6.7
9.1 1.4583 0.0075 0.004 ± 0.004 38 ± 4.7
CsCl 1.0 1.3787 0.0071 0.008 ± 0.004 120 ± 65
2.4 1.3881 0.0025 0.010 ± 0.001 50 ± 9
4.7 1.4109 0.0086 0.009 ± 0.004 13 ± 15
9.1 1.4679 0.0016 0.002 ± 0.001 16 ± 1

Density of crystalline salt: LiCl, 2.07 g/cm3; NaCl, 2.17 g/cm3; KCl, 1.988 g/cm3; RbCl, 2.8 g/cm3; CsCl, 3.988 g/cm3.46
538 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541
Figure 5. (a) Volume change on mixing versus molality (mol/Kg-solvent); LiCl (diamond); NaCl (square); KCl (triangle); RbCl (filled circle); CsCl (open circle). (b) Partial molal
volume of salt (symbols) and each molal volume of crystal (short bars); LiCl (diamond); NaCl (square); KCl (triangle); RbCl (filled circle); CsCl (open circle). The short heavy bars
represent the molal volume of the crystalline salts, in order of LiCl, NaCl, KCl, CsCl, and RbCl (from bottom to top).
NaCl, KCl, and RbCl are slightly less than crystalline molal volume,
but differences are almost within standard errors. The partial molal
volume for CsCl is much less than molar volumes of either glass or
crystal. Rate constant for degradation increases in the order, LiCl <
NaCl z KCl < RbCl z CsCl, the latter two salts having essentially the
same degradation rate as the system without any salt. Thus,
although in several cases, addition of salt causes loss in free volume,
particularly for CsCl, this salt has no impact on stability. Thus, it is
clear, from analysis of both the volume changes on mixing and the
partial molal volumes that the impact of free volume as assessed by
density measurements is not related in any simple way to stability.
Free-Volume Measurements by PALS
Physical parameters collected from PALS measurement are
summarized in Table 5. The positron lifetime as well as the mean
diameter of the free-volume pores, both of which have the same
trend because d3 was directly derived from t3, are given. The
standard error of the triplicate data was 0.02 ns for t3. The intensity
data are given because intensity is approximately related to the
number density of “holes.” Here, we assume, as is conventional,
that the positron lifetime is not very sensitive to chemical sur-
roundings of the vacancy and mainly reflects free-volume pore
sizes.30,31 Although the relationship can be complex and intensity
may be impacted by other factors as well,30 the qualitative trend
itself of the lifetime would not be impacted.
Figure 6. Comparison of properties at 9% salt: partial molal volume, molar volumes of
pure crystal and glass, and 10 times the aggregation rate constant at 50
C. Filled bars,
partial molal volume; heavy stripe, molar volume crystal; light stripe, estimated molar
volume amorphous; open bar, 10 rate constant for aggregation at 50
C.
In Figure 7, the correlation of the PALS parameters with stability
is illustrated. There is a good correlation of mean pore diameter of
the free volume with storage stability (R2 > 0.96), and the corre-
lation is sensible, at least in the series of the salt systems studied.
However, no monotonic relationship was observed between in-
tensity and stability. As PALS gives a direct measurement of rela-
tively small free volume (subnanometer), PALS provides a different
assessment of free volume than what is provided by density. That is,
PALS counts not all “free volume,” and it is the size of the “holes”
that is measured. Density (nominally) measures all free volume
accessible to helium in the gas pycnometric measurement.
Although more extensive data are needed for a definitive conclu-
sion, it does appear that PALS data may provide meaningful infor-
mation on free volume that is critically linked to long-term stability
in the glassy formulations.
Fast Dynamics by Neutron Scattering
Based on previous studies,3,45 we would expect a better corre-
lation between stability and fast dynamics, measured by neutron
scattering, than with global dynamics, as measured by enthalpy
relaxation time. This is exactly what we observe. Results from the
neutron scattering experiments with the rHSA:sucrose system
without salt and with 7% LiCl or with 7% CsCl are given in Figure 8.
The vertical axis gives the mean square amplitude of motion on a
nanosecond timescale, mostly by hydrogen-bearing groups of
atoms, and the horizontal axis is temperature. It should be noted
that the formulation without salt and the formulation with CsCl
added have essentially the same fast dynamics by this measure, but
the formulation with added LiCl shows much lower mean square
amplitude of motion, particularly in the range of room temperature
to the storage temperatures used in the stability studies conducted
in this study. The meaning is simply that the addition of LiCl
dampens the fast dynamics, or antiplasticizes mobility on the
nanosecond timescale. Note that the stability data for BSA and rHSA
systems (Figure 3) are well correlated with the fast dynamics re-
sults. Addition of LiCl stabilizes but addition of CsCl either has no
systematic effect (BSA) or destabilizes somewhat (rHSA). The PALS
data (Table 5, Figure 7) also show a trend of hole size with these
Table 5
Positron Lifetime, Intensity, and Calculated Mean Diameter of Free Volume by PALS;
the Measurement Was Repeated Three Times (±0.02 ns)
Sample t3 (ns) I3 (%) d3 (nm)
rHSA-sucrose (1:1) 1.644 10.5 0.495
rHSA-sucrose (1:1)-9% CsCl 1.667 8.2 0.500
rHSA-sucrose (1:1)-9% LiCl 1.592 7.5 0.484
 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541
Figure 7. Correlation of mean diameter of the local free volume with aggregation rate
constant. Open squares, intensity (I3 ÷ 5); filled circles, diameter (d3  3). Rate con-
stants were estimated from those for 7% salt formulations.
systems that is consistent with the fast dynamics results and sta-
bility. That is, small hole size is associated with smaller mean
square amplitude of motion and better stability. Perhaps, the cor-
relation with fast dynamics and/or PALS hole size data arises
because of differing electrostatic interactions due to different ion
size or is a reflection of “optimal volume” in modifying free volume.
However, the detailed mechanism is not clear.
Annealing Effect on Protein Stabilization in Salt Systems
After and/or during the drying process, samples that are held at
a given temperature, ambient or above but below Tg, can result in
significant “annealing effects,” thereby producing a significant
decrease in global mobility as well as significant stabilization
against degradation.39 The maximum annealing effect for mobility
Figure 8. Mean square amplitude of motion on a nanosecond timescale from neutron-
scattering experiments. Key: 1, rHSA-sucrose (1:1); 2, rHSA-sucrose (1:1) þ 7% LiCl; 3,
rHSA:Sucrose (1:1) þ 7% CsCl.
539
reduction commonly occurs at about 15
C-20
C below Tg.37,38 The
annealing temperature used reflects this expectation in that the
annealing temperature of 75
C averages 17
C below the Tg for the
samples studied. Figure 9 illustrates the result of the annealing
study, which was conducted for 9% LiCl, 9% NaCl, and 9% CsCl sys-
tems using rHSA. Rate constants were calculated from a series of
assay data by linearly fitting the equation for square root of time
kinetics to the aggregation data. Clearly, an impact of annealing on
protein stability was observed (Fig. 9) for all systems studied except
the 9% LiCl system. For the LiCl system, annealing did not decrease
the rate constant for degradation, perhaps because the stabilization
effect of LiCl is already superior to that of the other salts and
therefore has already reached a physical state where there might be
no room for additional stabilization. For annealed samples, the
order of stability was NaCl z LiCl > None > CsCl, which was
completely consistent with the stability study data set shown in
Figure 3. Hence, this study has demonstrated that not only can
addition of small amounts of LiCl and NaCl improve protein stability
in sucrose-based formulation, but further enhancement of storage
stability may be obtained by annealing operations.
Conclusions
This study has demonstrated that the addition of monovalent
electrolytes (alkali chlorides) to sucrose-based protein formula-
tions can improve storage stability. The variations of aggregation
rate could not be rationalized by either protein secondary structure
using FTIR or global glass dynamics detected by TAM. Changes in
free volume measured by volume changes on mixing and partial
molal volumes of salt in the protein:sucrose systems suggested
decreases in free volume for some systems, but those changes in
free volume were not well correlated with the impact of the salts on
stability. In an attempt to correlate stability with an alternate
measure of free volume, PALS measurements were carried out to
examine the correlation between stability and the mean diameter
of free-volume holes, as calculated from the lifetime of positro-
nium. Although the data set is sparse, hole size was well correlated
with stability data in the series: LiCl, CsCl, and no salt formulation,
suggesting that the relatively small and local free volume as
Figure 9. Results of annealing study with rHSA- and/or sucrose-based formulation.
Annealing was conducted at 75
C for 13 h (black bars), which was compared with
nonannealing samples (white bars). Each value of subtracting storage temperature
from Tg for each formulation was calculated using BSA and/or sucrose formulations;
15
C, 20
C, 8
C, 25
C, for none, 9% NaCl, 9% CsCl, 9% LiCl, respectively.
540 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541
measured by PALS hole size may be a good predictor of storage
stability in glassy formulations. Fast dynamics, as characterized by
the mean square amplitude of motion on a nanosecond timescale,
is consistent with the PALS data and stability; small hole size from
PALS means smaller amplitude of fast dynamics motion and better
stability. Additional data are needed to support this tentative
conclusion. Additionally, annealing was observed to further
improve stability in protein-sucrose-salt systems. The observation
that small amounts of LiCl and NaCl can significantly stabilize these
proteins provides new insight into the impact of sodium chloride,
which is commonly used in biopharmaceuticals, on product quality.
Acknowledgments
The authors acknowledge partial support from NIH/NIBIB under
grant R01 EB006398-01A1. The authors are grateful to Dr. Jack
Gromek at the Institute of Materials Science (IMS) at University of
Connecticut for using PXRD and also would like to express appre-
ciation to Dr. Norman Chieng and Dr. Puneet Sharma for discus-
sions. This work used facilities supported in part by the National
Science Foundation under Agreement No. DMR-0944772. Certain
trade names and company products are identified in this article to
specify adequately the experimental procedure. Such identification
does not imply recommendation or endorsement by the National
Institute of Standards and Technology, or does it imply that the
instruments, materials, or suppliers identified are necessarily the
best available for the purpose.
References
1. Kasper JC, Winter G, Friess W. Recent advances and further challenges in
lyophilization. Eur J Pharm Biopharm. 2013;85:162-169.
2. Wang B, Tchessalov S, Warne NW, Pikal MJ. Impact of sucrose level on storage
stability of proteins in freeze-dried solids: I. Correlation of protein-sugar
interaction with native structure preservation. J Pharm Sci. 2009;98:3131-3144.
3. Wang B, Tchessalov S, Cicerone MT, Warne NW, Pikal MJ. Impact of sucrose
level on storage stability of proteins in freeze-dried solids: II. Correlation of
aggregation rate with protein structure and molecular mobility. J Pharm Sci.
2009;98:3145-3166.
4. Wang W, Singh S, Zeng DL, King K, Nema S. Antibody structure, instability, and
formulation. J Pharm Sci. 2007;96:1-26.
5. Chang L, Shepherd D, Sun J, Tang X, Pikal MJ. Effect of sorbitol and residual
moisture on the stability of lyophilized antibodies: implications for the
mechanism of protein stabilization in the solid state. J Pharm Sci. 2005;94:
1445-1455.
6. Katayama DS, Carpenter JF, Menard KP, Manning MC, Randolph TW. Mixing
properties of lyophilized protein systems: a spectroscopic and calorimetric
study. J Pharm Sci. 2009;98:2954-2969.
7. Osterberg T, Fatouros A, Mikaelsson M. Development of a freeze-dried albu-
min-free formulation of recombinant factor VIII SQ. Pharm Res. 1997;14:892-
898.
8. Yoshioka S, Aso Y. Correlations between molecular mobility and chemical
stability during storage of amorphous pharmaceuticals. J Pharm Sci. 2007;96:
960-981.
9. Jameel F, Tchessalov S, Bjornson E, Lu X, Besman M, Pikal MJ. Development of
freeze-dried biosynthetic factor VIII: I. A case study in the optimization of
formulation. Pharm Dev Technol. 2009;14:687-697.
10. Pikal MJ, Dellerman KM, Roy ML, Riggin RM. The effects of formulation vari-
ables on the stability of freeze-dried human growth hormone. Pharm Res.
1991;8:427-436.
11. Her L, Deras M, Nail SL. Electrolyte-induced changes in glass transition tem-
peratures of freeze-concentrated solutes. Pharm Res. 1995;12:768-772.
12. Carpenter JF, Crowe LM, Crowe JH. Stabilization of phosphofructokinase with
sugars during freeze-drying: characterization of enhanced protection in the
presence of divalent cations. Biochim Biophys Acta. 1987;923:109-115.
13. UniProtKB, protein knowledgebase. Available at: http://www.uniprot.org/
uniprot/. Accessed December 2, 2015.
14. Broering JM, Bommarius AS. Evaluation of Hofmeister effects on the kinetic
stability of proteins. J Phys Chem B. 2005;109:20612-20619.
15. Zhang L, Tan H, Fesinmeyer R, et al. Antibody solubility behavior in monovalent
salt solutions reveals specific anion effects at low ionic strength. J Pharm Sci.
2012;101:965-977.
16. Cicerone MT, Tellington A, Trost L, Sokolov A. Substantially improved stability
of biological agents in dried form: the role of glass dynamics in preservation of
biopharmaceuticals. Bioprocess Inter. 2003;1:36-47.
17. Cicerone MT, Soles CL. Fast dynamics and stabilization of proteins: binary
glasses of trehalose and glycerol. Biophys J. 2004;86:3836-3845.
18. Yoshioka S, Miyazaki T, Aso Y, Kawanishi T. Significance of local mobility in
aggregation of b-Galactosidase lyophilized with trehalose, sucrose or sta-
chyose. Pharm Res. 2007;24:1660-1667.
19. Zhong Q. Protein Stability in Sugar-Based Glass: Using Dynamics to Sort Causality
From Correlation. Cambridge, MA: World Lyophilization Summit; 2011.
20. Carpenter JF, Prestrelski SJ, Dong A. Application of infrared spectroscopy to
development of stable lyophilized protein formulations. Eur J Pharm Biopharm.
1998;45:231-238.
21. Kendrick BS, Dong A, Allison SD, Manning MC, Carpenter JF. Quantitation of the
area of overlap between second-derivative amide I infrared spectra to deter-
mine the structural similarity of a protein in different states. J Pharm Sci.
1996;85:155-158.
22. Liu J, Rigsbee DR, Stotz C, Pikal MJ. Dynamics of pharmaceutical amorphous
solids: the study of enthalpy relaxation by isothermal microcalorimetry.
J Pharm Sci. 2002;91:1853-1862.
23. Kawakami K, Pikal MJ. Calorimetric investigation of the structural relaxation of
amorphous materials: evaluating validity of the methodologies. J Pharm Sci.
2005;94:948-965.
24. Imamura K, Maruyama Y, Tanaka K, Yokoyama T, Imanaka H, Nakanishi K. True
density analysis of a freeze-dried amorphous sugar matrix. J Pharm Sci.
2008;97:2789-2797.
25. Kikuchi T, Wang B, Pikal MJ. High-precision absolute (true) density measure-
ments on hygroscopic powders by gas pycnometry: application to determining
effects of formulation and process on free volume of lyophilized products.
J Pharm Sci. 2011;100:2945-2951.
26. Itoha Y, Shimazu A, Sadzuka Y, Sonobe T, Itai S. Novel method for stratum
corneum pore size determination using positron annihilation lifetime spec-
troscopy. Int J Pharm. 2008;358:91-95.
27. Townrow S, Kilburn D, Alam A, Ubbink J. Molecular packing in amorphous
carbohydrate matrixes. J Phys Chem. 2007;111:12643-12648.
28. Xie W, Ju H, Geise GM, et al. Effect of free volume on water and salt transport
properties in directly copolymerized disulfonated poly(arylene ether sulfone)
random copolymers. Macromolecules. 2011;44:4428-4438.
29. Dlubek G, Shaikh MQ, Ratzkeb K, Pionteckc J, Paluchd M, Faupel F. Sub-
nanometre size free volumes in amorphous Verapamil hydrochloride: a posi-
tron lifetime and PVT study through Tg in comparison with dielectric
relaxation spectroscopy. Eur J Pharm Sci. 2010;41:388-398.
30. Tant MR, Hill AJ. Structure and Properties of Glassy Polymers. ACS Symposium
Series. 710. Washington DC: ACS Books; 1998.
31. Zelko R, Orban A, Suvegh K. Tracking of the physical ageing of amorphous
pharmaceutical polymeric excipients by positron annihilation spectroscopy.
J Pharm Biomed Anal. 2006;40:249-254.
32. Dlubek G, Piontech J, Kilburn D. The structure of the free volume
in poly(styrene-co-acrylonitrile) from positron lifetime and pressure volume
temperature (PVT) experiments. I. Free volume from the Simha-Somcynsky
analysis of PVT experiments. Macromol Chem Phys. 2004;205:500-511.
33. Dlubek G, Bondarenko V, Al-Qaradawi IY, Kilburn D, Krause-Rehberg R. The
structure of the free volume in poly(styrene-co-acrylonitrile) from positron
lifetime and pressure volume temperature (PVT) experiments. II. Local free
volume from positron annihilation lifetime spectroscopy (PALS). Macromol
Chem Phys. 2004;205:512-522.
34. Dlubek G, Stejny J, Lüpke TH, et al. Free-volume variation in polyethylenes of
different crystallinities: positron lifetime, density, and X-ray studies. J Polym
Sci. 2001;40:65-81.
35. Dlubek G, Pionteck J, Ratzke K, Kruse J, Faupel F. Temperature dependence of the
free volume in amorphous Teflon AF1600 and AF2400: a pressure-volume-
temperature and positron lifetime study. Macromolecules. 2008;41:6125-6133.
36. Abdul-Fattah AM, Dellerman KM, Bogner RH, Pikal MJ. The effect of annealing
on the stability of amorphous solids: chemical stability of freeze-dried mox-
alactam. J Pharm Sci. 2007;96:1237-1250.
37. Luthra SA, Hodge IM, Pikal MJ. Effects of annealing on enthalpy relaxation in
lyophilized disaccharide formulations: mathematical modeling of DSC curves.
J Pharm Sci. 2008;97:3084-3099.
38. Wang B, Pikal MJ. The impact of thermal treatment on the stability of freeze
dried amorphous pharmaceuticals: I. Dimer formation in sodium ethacrynate.
J Pharm Sci. 2010;99:663-682.
39. Wang B, Cicerone MT, Aso Y, Pikal MJ. The impact of thermal treatment on the
stability of freeze-dried amorphous pharmaceuticals: II. Aggregation in an IgG1
fusion protein. J Pharm Sci. 2010;99:683-700.
40. Pikal MJ, Shah S. The collapse temperature in freeze-drying: dependence on
measurement methodology and rate of water removal from the glassy state. Int
J Pharm. 1990;62:165-186.
41. Nunn S, Nishikida K. Advanced ATR Correction Algorithm. Waltham, MA: Thermo
Fisher Scientific Inc.; 2008. Application note: 50581.
42. Schersch K, Betz O, Garidel P, Muehlau S, Bassarab S, Winter G. Systematic
investigation of the effect of lyophilizate collapse on pharmaceutically
relevant proteins I: stability after freeze-drying. J Pharm Sci. 2010;99:2256-
2278.
 H. Goshima et al. / Journal of Pharmaceutical Sciences 105 (2016) 530e541 541

43. Tchessalov S, Dixon D, Warne N. 2010. Lyophilization Above Collapse. WO
2010017296 A1.
44. Xu Y, Carptenter JF, Cicerone MT, Randolph TW. Contributions of local mobility
and degree of retention of native secondary structure to the stability of re-
combinant human growth hormone (rhGH) in glassy lyophilized formulations.
Soft Matter. 2013;9:7855-7865.
45. Chieng N, Cicerone MT, Zhong Q, Liu M, Pikal MJ. Characterization of dynamics
in complex lyophilized formulations: II. Analysis of density variations in terms
of glass dynamics and comparisons with global mobility, fast dynamics, and
PALS. Eur J Pharm Biopharm. 2013;85:197-206.
46. National Center for Biotechnology Information. The PubChem Project. Available
at: http://pubchem.ncbi.nlm.nih.gov. Accessed December 4, 2015.