Biophysical Characterization and Formulation of F1 V,
a Recombinant Plague Antigen
AJIT JOSEPH M. D SOUZA, BRANDI M. FORD, KEVIN D. MAR, VINCE J. SULLIVAN
BD Technologies, 21 Davis Dr., Research Triangle Park, North Carolina 27519
Received 28 July 2008; revised 6 October 2008; accepted 6 November 2008
Published online 30 December 2008 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.21652
ABSTRACT: The recombinant plague antigen, F1 V, was studied for its structural
characteristics using several biophysical techniques. A larger apparent molecular
weight relative to its calculated molecular weight obtained from size exclusion chro-
matography, an unusually large Rg obtained from MALS, and ANS dye binding studies
which indicate that all hydrophobic regions of the protein are exposed to solvent
demonstrated that F1 V exists like a disordered protein with a worm-like conformation.
The pH-solubility profile of F1 V showed a solubility minimum at pH 5, close to its pI,
consistent with the lack of repulsive forces that result in aggregation. Thus, in contrast
to most globular proteins that exhibit a secondary and a tertiary structure, F1 V seems
to lack tertiary structure and like an unfolded protein is more prone to aggregation via
hydrophobic interactions. Despite this, when renatured gradually using descending
guanidine hydrochloride concentration dialysis, in the presence of Mgþ2, a surfactant
and arginine hydrochloride at a pH of 7.5, F1 V appears to populate predominantly in its
monomeric state. ß 2008 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci
98:2592 2602, 2009
Keywords: F1 V; plague; vaccine; biotechnology; protein aggregation; protein
formulation; circular dichroism; fluorescence spectroscopy; light scattering
INTRODUCTION dimer of a heptamer.2 Further, the researchers
suggested that several ionic interactions are
F1 V is a 53.2 kDa synthetic recombinant plague involved in maintaining subunit interactions
antigen comprising the capsular F1 and virulence- both within and between the 7-mer building
associated V proteins joined by a dipeptide spacer. blocks. As an alternative to whole-cell vaccines,
The sequence of F1 V is listed in Scheme 1. immunization with purified recombinant Y. pestis
The F1 protein (15.5 kDa) naturally occurs as proteins, including fraction 1 (F1), was proposed
the only subunit of a capsular-like structure that and protective immunization following intramus-
functions as a fibrous virulence organelle on the cular administration of F1 capsular antigen was
surface of the Yersinia pestis. In recent years shown by Simpson et al.3 and Andrews et al.1 This
recombinantly produced F1 capsular protein approach, while effective against a wild-type
has been shown to confer protective immunity strain, proved ineffectual in conferring protective
against plague.1 Its association in solution has immunity against infection from strains lacking
been characterized using mass spectrometry as a a capsule.1 Therefore, several other virulence
factors were tested for their ability to elicit an
effective immune response and factor V was
Correspondence to: Vince J. Sullivan (Telephone: 919-597-
identified as an appropriate antigen due to its
6495; Fax: [phone]; E-mail: [e-mail])
high protective efficacy and abundant natural
Journal of Pharmaceutical Sciences, Vol. 98, 2592 2602 (2009)
expression during infection.4 The V antigen is a
ß 2008 Wiley-Liss, Inc. and the American Pharmacists Association
2592 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009
BIOPHYSICAL CHARACTERIZATION AND FORMULATION 2593
MADLTASTTA TATLVEPARI TLTYKEGAPI TIMDNGNIDT ELLVGTLTLG GYKTGTTSTS 60
VNFTDAAGDP MYLTFTSQDG NNHQFTTKVI GKDSRDFDIS PKVNGENLVG DDVVLATGSQ 120
DFFVRSIGSK GGKLAAGKYT DAVTVTVSNQ EFMIRAYEQN PQHFIEDLEK VRVEQLTGHG 180
SSVLEELVQL VKDKNIDISI KYDPRKDSEV FANRVITDDI ELLKKILAYF LPEDAILKGG 240
HYDNQLQNGI KRVKEFLESS PNTQWELRAF MAVMHFSLTA DRIDDDILKV IVDSMNHHGD 300
ARSKLREELA ELTAELKIYS VIQAEINKHL SSSGTINIHD KSINLMDKNL YGYTDEEIFK 360
ASAEYKILEK MPQTTIQVDG SEKKIVSIKD FLGSENKRTG ALGNLKNSYS YNKDNNELSH 420
FATTCSDKSR PLNDLVSQKT TQLSDITSRF NSAIEALNRF IQKYDSVMQR LLDDTSGK 478
Scheme 1. The amino acid sequence of F1 V. Initial part represents the sequence of
capsular protein F1, the dipeptide linker is shown in bold and the rest is the sequence of
virulence associated protein V.
multifunctional virulence factor that helps the formulation excipients yielded F1 V in largely a
bacterium establish itself in the host by assisting monomeric form at pH 7.5.
injectosome insertion into the host cell, by
reducing the expression of the host cytokines,
tumor necrosis factor alpha and gamma interferon MATERIALS AND METHODS
in response to Yersinia infection locally5 7 and by
playing a role in the regulation of the low-calcium F1 V [(AAY23169.1) capsule protein fraction
response.8 10 1/virulence antigen fusion protein (synthetic
In designing an improved plague vaccine construct)] was obtained through the NIH Biode-
researchers at United States Army Medical fense and Emerging Infections Research Re-
Research Institute for Infectious Diseases sources Repository, NIAID, NIH: Yersinia
(USAMRIID) discovered that a subunit vaccine, pestis F1 V Fusion Protein. Monomer-enriched
called F1 V, comprised of F1 and V antigens fused F1 V (NR-2561), recombinantly produced from
together and produced recombinantly, provided Escherichia coli, was provided as a solution at a
high levels of protection against bubonic and concentration of 1.5 mg/mL of F1 V (69.8%
pneumonic plague caused by virulent Y. pestis Monomer, 19.4% Dimer and 10.8% Oligomers
strains with or without F1 capsule.11 Although as indicated by the manufacturer) in 20 mM
some skeptics posit that the immunogenicity arginine, 10 mM NaCl and 1 mM L-cysteine at
of F1 V may be muted because of the lack of pH 9.52. Dimer-enriched F1 V (NR-2563) was
large aggregates relative to F1,12 there is general provided as a solution of 1.5 mg/mL (27.1%
consensus that the F1 V and F1 produce monomer, 48.8% dimer and 24.5% multimer as
similar immune response as measured using indicated by the manufacturer) in 20 mM
ELISA. Moreover, side-by-side comparisons arginine, 10 mM NaCl and 1 mM L-cyteine at
of F1 V to cocktail vaccines containing separate pH 8.7. Tween1 80 was obtained from Spectrum
F1 plus V antigens in animal studies demon- Chemicals (Gardena, CA), arginine hydrochlor-
strated statistically better protection by the fusion ide, MgCl2, NaCl, CaCl2. Hydroxypropyl-b-cyclo-
protein antigen as compared to the separate dextrin (HP-b-CD) from Sigma Aldrich (St.
antigens.13 Louis, MO) and Pluronic F127 was obtained from
In solution F1 V exists as a heterogeneous BASF (Florham Park, NJ). Urea was obtained
mixture of monmeric, dimeric and oligomeric from Pierce and guanidine hydrochloride was
species ranging in molecular weights from 54 to obtained from EMD (Gibbstown, NJ) and Sigma
1000 kDa.13 The recent emphasis of the regulatory Aldrich. Dialysis cassettes (MWCO 10,000 kDa)
agencies on well characterized biologics has were obtained from Pierce (Rockford, IL). Mole-
motivated molecular biologists to construct cular weight standards were obtained from
isoforms with minimal propensity to aggregate Sigma Aldrich.
and formulation scientists to develop formulations
with minimal heterogeneity. Here we report
that F1 V exists as an disordered protein
Size Exclusion Chromatography (SEC)
exhibiting some secondary structure, somewhat
like a premolten globule.14 To formulate F1 V into
SEC was carried out using a TSK-GEL G3000SW
a relatively less heterogeneous mixture it was
(Sigma Aldrich) column at a flow rate of 0.65 mL/
necessary to use descending denaturant concen- min. The mobile phase was a mixture of 90 parts of
tration dialysis. The process along with the
300 mM NaCl, 100 mM, phosphate buffer (pH 7.5)
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009
2594 D SOUZA ET AL.
and 10 parts of isopropyl alcohol. Detection was a path length of 1 cm. An extinction co-efficient of
set to 280 nm. For all samples the injection volume 0.4679 (mg/mL) 1 cm 1, calculated based on the
was held constant at 100 mL. Standards of known sequence of F1 V, was used to determine the
molecular weight were used to calibrate the concentration of F1 V in solution.
column.
Secondary Structure Using Circular Dichroism
Molecular Weight Using SEC
The secondary structure of F1 V was assessed
An excellent linear correlation exists between the using a Jasco 810 circular dichroism spectro-
logarithm of molecular weight of proteins and polarimeter. Samples were prepared similarly to
the ratio of their elution volume, Ve, to the void those described in the section above, however they
volume, Vo, of the size exclusion column.15 were dialyzed for only 4 h and following dialysis
Therefore, to assess the size of F1 V the Ve/Vo their concentrations were adjusted to 200 mg/mL
ratio of various standard proteins were plotted using the appropriate buffer, prior to analysis.
against their molecular weights to generate a Spectra were recorded on 200 mg/mL solutions in
standard curve. The molecular weight of F1 V 0.1 cm path length cuvettes at a scan rate of
was interpolated from the standard curve. 20 nm/min with a band width of 1 nm and a 2 s
response time. Each spectrum was collected from
190 to 260 nm as an average of 3 measurements
and the corresponding buffer spectrum was
Light Scattering
subtracted from the sample to yield the final
Molecular weight (MW), RMS radius (Rg) and
spectrum. The spectra were fitted using Olis
hydrodynamic radius (Rh) of F1 V were deter-
Global WorksTM Software (Olis, Bogart, GA) to
mined with ASTRAß (Wyatt Technology Corp.,
assess the fractions of various secondary struc-
Santa Barbara, CA) using data collected from a
tures in F1 V.
multi-angle static light scattering (MALS) detec-
tor (DAWN1 EOS, Wyatt Technology Corp.)
placed in-line with a TSK-GEL G3000SW column
Tertiary Structure Using ANS
and an interferometric refractometer (OptiLab
Fluorescence Measurement
DSP, Wyatt Technology Corp.). A dn/dc of
0.185 mL/g was used for all calculations. The Extrinsic dye emission spectra of anilino-1-
hydrodynamic diameter was calculated using the naphthalene sulfonate (ANS) in F1 V solution
WyattQELSTM (Wyatt Technology Corp.) which at pHs varying from 3 to 10 were obtained on the
interfaces with the MALS. Quantmaster fluorescence spectrophotometer
(Photon Technologies, Inc., Birmingham, NJ).
Samples were prepared as described above
followed by the addition of 5 mLof a6mM solution
pH-Solubility Profile
of ANS to 200 mL at 200 mg/mL F1 V solution.
Buffers of pHs ranging from 3 through 8, in
Excitation was set at 375 nm and emission spectra
increments of 1 unit, were prepared by mixing
were monitored from 400 to 600 nm in 1 nm
0.2 M Na2HPO4 and 0.1 M citric acid in appro-
increments. The slit widths were set at 4 nm. Each
priate proportions.16 Buffers of pH 9 and 10 were
emission spectrum was an average of three
prepared by adjusting the pH of a 10 mM solution
collected spectra. The emission scans were
of glycine. Each of the 8 aliquots (120 mL each) of
obtained at 2.58C intervals from 108C to 908C
the monomer-enriched F1 V stock solution was
with a 3 min equilibration period.
diluted with a buffer (380 mL) to yield a 360 mg/mL
concentration. Each mixture was then transferred
to a dialysis cassette (MWCO 10,000, Pierce) and
Effect of Denaturant on the Relative Concentrations
dialyzed against the respective buffer overnight
of Monomer, Dimer, and Oligomers of F1 V
for 15 h under refrigeration. The dialysate
was filtered through 0.2 mm filters (PALL) and To determine if F1 V can be dissociated to a
measured for concentration of soluble F1 V. monomer using denaturants urea and guanidine
Measurements were made on an Agilent 8453 hydrochloride, 20 mL of the stock solution of
UV spectrophotometer using quartz cuvettes with dimer-enriched F1 V (1.5 mg/mL) was mixed with
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009 DOI 10.1002/jps
BIOPHYSICAL CHARACTERIZATION AND FORMULATION 2595
varying volumes of 8 M denaturant (urea or
guanidine hydrochloride) and water for injection
to produce solutions containing 150 mg/mL of F1
V in varying concentrations of denaturant. The
relative concentrations of the various species of
F1 V were monitored using SEC.
Descending Denaturant Concentration Dialysis
Lyophilized dimer-enriched F1 V (1.5 mg) was
dissolved in 3 mL of denaturation buffer composed
of 7.2 M guanidine hydrochloride, 50 mM MgCl2,
20 mM arginine hydrochloride solution and 0.1%
W/V Pluronic1 F-127. This was transferred to a
10,000 MWCO dialysis cassette (Pierce) and
Figure 1. An excellent linear correlation exists
equilibrated for 2 h in 250 mL of denaturation
between the logarithm of molecular weight of proteins
buffer contained in a flow-through beaker. Follow-
and the ratio of their elution volume, Ve, to the void
ing equilibration the denaturation buffer was
volume, Vo, of the size exclusion column. The molecular
gradually replaced by the renaturation buffer at
weight of monomeric F1 V was calculated to be about
the rate of 4 mL/min while maintaining the total 130 kDa using this correlation.
volume of the beaker at 250 mL. A total of 4 L of
renaturation buffer was required to drive the
denaturant concentration to less than 1 mM.
monomers, dimers and oligomers of F1 V, respec-
tively. Based on MALS data, the molecular weight
of monomeric F1 V was determined to be 55 kDa
RESULTS
(MWcalculated ź 53 kDa), and the root mean square
radius (Rg) was determined to be 10 nm. The
Molecular Weight Using SEC
hydrodynamic radius (Rh) was determined by
dynamic light scattering to be 3.0 nm. The
Size exclusion chromatography separates pro-
experimentally determined MW and Rh are close
teins on the basis of hydrodynamic volume. Both
to the expected values for F1 V. The Rg, however,
molecular weight and three dimensional shape
contribute to the degree of retention. The stan-
dard curve based on the retention volumes of the
various globular protein standards and the linear
fit along with the equation describing it are shown
in Figure 1 (data shown in Tab. 1). Using this
equation the molecular weight of monomeric F1 V
can be calculated to be about 130 kDa (Tab. 1),
about 2.5 times its actual molecular weight. This
suggests that the three dimensional shape of
F1 V is such that it does not allow it to diffuse into
the pores of the stationary phase as easily as other
globular proteins of a similar molecular weight.
Further this indicates a linear, extended con-
formation of the F1 V molecule.
Figure 2. Size exclusion chromatogram of dimer
enriched F1 V using the refractive index detector (solid
Light Scattering
trace). The dashed lines represent the molecular
A size exclusion chromatogram of the dimer
weights (right hand y-axis) associated with each of
enriched F1 V is shown in Figure 2. Three peaks
the three peaks. The monomer has a molecular weight
are observed, with elution times of approximately
of 55 kDa, the dimer a molecular weight of 107 kDa and
22.5, 19, and 15.5 min, which are assigned to oligomers range from 250 to 3000 kDa.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009
2596 D SOUZA ET AL.
is much higher than that of a typical 55 kDa
globular protein17 and suggests a more open,
extended structure. The molecular weight of the
dimer was determined to be 107 kDa, with a Rh of
5.2 nm and a Rg of 10 nm. The oligomer peak was
observed to contain species of varying molar
masses (Fig. 2) and therefore was not analyzed
further.
pH-Solubility Profile
Solubility, as the term is used here, is not meant
to imply thermodynamic solubility as there is a
kinetic element to this observation. It is important
to note that measurements were made after 15 h
of dialysis at a given pH. Despite the low
concentration tested in this study the solubility
of F1 V showed a clear dependence on the pH of
the medium (Fig. 3a). The pH of minimum
solubility was observed to be 5, near the
calculated pI (5.38) of F1 V. The solubility,
however, increased dramatically from pH 8 to
10 and correlates very well with a change in
the calculated net charge from 16 to 44. When
these same samples were further analyzed using
SEC, F1 V showed extensive aggregation at pHs
lower than 9 (Fig. 3b) with complete lack of
monomeric F1 V at pHs 5 and 6, around its charge
neutral state. It is apparent from this that the
association of F1 V into dimer/oligomers is due to
hydrophobic interactions that are enhanced when
Figure 3. (a) Effect of pH on the solubility of F1 V
F1 V has a net charge of zero, whereas at pH 10
(inclusive of monomer, dimer and oligomers). The pH of
where the F1 V molecule is highly charged
minimum solubility was observed to be 5, near the pI
aggregation may be limited due to charge charge
(5.38) of F1 V. (b) Effect of pH on the solubility and
repulsion. monomer/dimmer/oligomer equilibrium in F1 V. As the
pH decreases the solubility of total F1 V decreases and
the fractions of dimeric and oligomeric F1 V increase.
For clarity the curves have been offset by 70% along the
Secondary Structure Using Circular Dichroism
y-axis.
Figure 4 shows the CD spectra of F1 V at various
pHs. A clear increase in intensity is evident at
208 nm as a function of pH indicating a change in Although the fits were better at some pHs than
the secondary structures of F1 V with pH. These others, in general they were deemed good
data were fitted using CONTINLL, SELCON3 enough to draw some general conclusions.
and CDSSTR algorithms and 11 different basis From these data it is apparent that in solution
sets to estimate the fractions of the various the secondary structure is dominated by b-strands
secondary structures. A fit was deemed good if and distorted b-strands which increase at
it fit the data visually, had a normalized spectral pHs nears the pI of F1 V; correspondingly the
fits standard deviation of less than 0.1 and the all a-helical content decreases near the pI, but
the fractions of various secondary structures increases at pHs away from the pI. It is
added up to between 0.95 and 1.05. It was conceivable that the increased content of
observed that the CDSSTR algorithm along b-strands near the pI may in part be due to
with the SP37 basis set18 yielded the best fits. the presence of intermolecular aggregates that
The results of these fits are shown in Table 2. dominate the composition of the solution.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009 DOI 10.1002/jps
BIOPHYSICAL CHARACTERIZATION AND FORMULATION 2597
however, the maximum intensity is observed at
the lowest temperature suggesting that the sites
responsible for ANS binding are completely
accessible even at 208C (all panels in Fig. 5).
Further, the intensity of ANS fluorescence
decreases as the temperature increases. At pH 4
the fluorescence decreases to a baseline value by
408C and from there on tracks well with the blank
suggesting that the ANS may be dissociating from
the protein. The fluorescence decay therefore,
appears to be due to a combination of collisional
quenching by the solvent and the dissociation
of ANS from the protein. Although some other
transitions are seen in the fluorescence decay at
temperatures typically above 708C except at pH 5
Figure 4. Effect of pH on the secondary structure of
where the transition appears at 408C, they are not
F1 V as assessed using circular dichroism. The helical
consistent with unfolding transitions. The data is
content of F1 V increases with increasing pH. For
further complicated by the presence of multiple
clarity the curves have been offset by 25% along the
species of F1 V at varying concentrations at
y-axis.
each pH as seen from Figure 3b. Despite these
complexities, these data when taken together,
Tertiary structure using ANS fluorescence
suggest a lack of tertiary structure in F1 V.
It has been previously noted that the sole
tryptophan residue in F1 V is partially-to-com-
Effect of denaturant on the relative concentrations
pletely exposed to the solvent and therefore does
of monomer, dimmer, and oligomers of F1 V
not respond to changes in the tertiary structure
of F1 V.13 So to probe the tertiary structure an The heterogeneity of molar masses of F1 V in
ANS binding experiment was carried out. ANS solutions NR-2561 and NR-2563 is apparent from
is known to fluoresce much more vigorously their descriptions above. F1 V solutions are
when bound to, by way of hydrophobic and supplied at a pH close to 10, a pH that is typically
ionic interactions, or present in nonpolar regions not conducive for long term chemical stability
relative to an aqueous environment. Although of proteins and can cause pain and irritation
ANS has shown some intrinsic effect on the following intramuscular administration of the
structure of proteins19,20 it is not completely vaccine. Also, it has been speculated by some
without merit in detecting exposed hydrophobic that F1 V aggregates of the order of tetramers
portions on a protein.21 In case of most globular and higher are not particularly immunogenic
proteins the sites for ANS binding become whereas others suggest that they are more
accessible as the tertiary structure is disrupted immunogenic. The immunogenicity of monomeric
by the increasing temperature and the ANS dye and dimeric F1 V were shown to not differ
diffuses into the hydrophobic core of the protein; statistically in mouse models for pneumonic
binding of the ANS to the protein results in an plague.22 Therefore, to maximize the immuno-
increase in fluorescence intensity. In case of F1 V, genicity of the F1 V it was our goal to minimize
Table 1. Molecular Weights and Retention Volumes of Various Proteins
MW Standard MW (kDa) Log (MW) Ve (min) Vo (min) Ve/Vo
Apoferritin 443 5.65 18.53 15.69 1.18
B-Amylase 200 5.30 20.53 15.69 1.31
Alcohol dehydrogenase 150 5.18 22.40 15.69 1.43
BSA 66 4.82 24.30 15.69 1.55
Carbonic anhydrase 29 4.46 29.14 15.69 1.86
F1 V monomer 129.5a 22.66 15.69 1.44
MW, molecular weight; Ve, elution volume; Vo, void volume.
a
Calculated.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009
2598 D SOUZA ET AL.
Table 2. The Fractions (F) of Various Secondary Structures of F1 V Generated by
Fitting the CD Spectra to Basis Set SP37 Using CDSSTR Algorithms
Normalized Spectral
pH Fa helix Fb stands Fb-turns Funordered Sum Fit Standard Deviation
3 0.23 0.38 0.19 0.19 0.99 0.08
4 0.25 0.36 0.20 0.20 1.01 0.04
5 0.19 0.44 0.18 0.22 1.03 0.05
6 0.20 0.40 0.19 0.22 1.01 0.04
7 0.22 0.35 0.22 0.21 1.00 0.08
8 0.25 0.34 0.20 0.21 1.00 0.09
9 0.22 0.38 0.19 0.21 1.00 0.01
10 0.24 0.35 0.20 0.21 1.00 0.02
The a-helix /b-strands fractions reported here are a sum of the fractions of a-helix/b-strands
and distorted a-helices/b-strands, respectively.
Fa helix ź Fa helix þ Fdistorted a helix and Fb standsź Fb strandþFdistorted b strand.
A fit is deemed good if it is visually acceptable, yields a sum of 1.0 and a normalized standard
deviation of <0.06.
this heterogeneity. To do this reproducibly, dimer concentration, however, was seen in the
regardless of the starting stock solution (either presence of increasing concentration of guanidine
NR-2561 or NR-2563), it was deemed necessary hydrochloride. All the species were dissociated
to dissociate the various F1 V species to a into the monomer at a guanidine hydrochloride
monomer and then formulate it to produce a less concentration of 7.2 M (Fig. 6b). The increas-
heterogeneous solution. It can be seen from ed effectiveness of guanidine hydrochloride, a
Figure 6a that the presence of 4 M urea causes charged denaturant, compared to urea, a neutral
a significant concentration of the oligomers to denaturant, indicates that the charge is important
dissociate. The dimers, however, remain unaf- in dissociating the F1 V dimer/oligomers; further
fected even at a concentration of 7.2 M. Thus, it is suggesting that charge may play an important
apparent that urea is not an optimal denaturant role in the association/dissociation of F1 V. Also,
for F1 V. A steady decrease in the oligomer and the high pH of guanidine solution may result in
a highly charged F1 V molecule that does not self-
associate due to charge charge repulsion.
Renaturation of F1 V
To reformulate F1 V at pH 7.5 with minimal
heterogeneity it was essential to renature the
F1 V after denaturation with 7.2 M guanidine
hydrochloride solution. Relative to procedures
like one-step dialysis, step-wise dialysis and
ultrafiltration, descending denaturant concentra-
tion dialysis was found to be the most practical
for renaturation. One-step dialysis resulted in a
solution with substantially oligomerized F1 V
with chromatograms similar to those in pH 7 and
Figure 5. Effect of temperature on the fluorescence
8 buffers shown in Figure 3b. Although, several
intensity due to ANS binding: with F1 V (filled squares)
excipients were tested to formulate the renatura-
and without F1 V (open circles). At all the pHs studied
tion medium we have limited our discussion to the
the intensity was maximum at the lowest temperature
and gradually decreased as the temperature increased. one that reduced heterogeneity the most.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009 DOI 10.1002/jps
BIOPHYSICAL CHARACTERIZATION AND FORMULATION 2599
not shown). It is conceivable that Mgþ2 interacts
strongly with the peptide bonds of the protein.
Therefore, 50 mM MgCl2 was incorporated into
the renaturation buffer. To minimize the inter-
molecular hydrophobic interactions, Tween1s,
Pluronic1 F127, and Brij1 were tested and
Pluronic1 F127 was observed to be the most
effective at a 0.1% W/V level (data not shown).
Arginine hydrochloride (20 mM) was incorporated
to further suppress aggregation; it is know to
effectively suppress protein protein interactions
while not significantly affecting the structure of
the protein.24 As can be seen in Figure 7 (black
trace) renaturation of F1 V using the renatura-
tion buffer described and descending denaturant
concentration dialysis results in a formulation
that is monomer-enriched (>75% monomer). In
liquid form, however, this formulation is stable
(without any increase in aggregates) for only
3 days at 48C and needs to be either lyophilized or
spray freeze dried for long term stability.
DISCUSSION
F1 V is a fusion protein with a calculated
molecular weight of 53.2 kDa and a calculated
isoelectric point of 5.38. The tendency of F1 V to
oligomerize has been documented13 and can be
attributed to the functional activity of its F-1
fragment to form multimers.12 F1 V has been
Figure 6. (a) Effect of urea on the dissociation of
F1 V (150 mg/mL). At concentrations as high as
7.2 M the denaturant urea could not dissociate the
F1 V dimers. For clarity the curves have been offset
by 55% along the y-axis. (b) Effect of guanidine hydro-
chloride on the dissociation of F1 V (150 mg/mL). The
use of 7.2 M guanidine hydrochloride resulted in com-
plete dissociation of the dimer into monomer as seen
from these chromatograms. For clarity the curves have
been offset by 85% along the y-axis.
It is apparent from the biophysical evidence
presented here that F1 V is essentially a dis-
ordered protein with exposed hydrophobic sites.
The pH of the renaturation solution was adjusted
to 7.5 by using 20 mM tris buffer. Since divalent
cations like Caþ2 and Mgþ2 are known for their
ability to salt-in23 we tested the effect of these
Figure 7. Upon renturation with 20 mM Tris
cations on F1 V. Interestingly, however, Mgþ2
(pH 7.5), 50 mM MgCl2, 20 mM arginine, 0.1% W/V
was observed to be significantly more effective in
Pluronics1 F-127 using descending denaturant concen-
preserving the monomeric state than Caþ2, sug-
tration dialysis F1 V can be formulated into a mono-
gesting a more specific interaction of Mgþ2 with
mer-enriched (>75% monomer) form as seen in this
the protein than a general salting-in effect (data chromatogram.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009
2600 D SOUZA ET AL.
modeled as a globular protein that refolds into a while Rh/Rg ź 0.33 for long rodlike chains, with
stable conformation upon purification on a col- worm-like chains exhibiting intermediate values
umn13 and aggregation has been attributed to that approach the rodlike chain value at high r/
intermolecular disulfide links.22 It appears that L.26,27 For the dimer, we obtain Rh/Rg ź 5.2/10
this may be due to the authors use of urea as a 0.5, again suggestive of an extended worm-like
denaturant which does not induce complete chain as opposed to a nondraining random coil.
dissociation of the F1 V dimers and oligomers. The lack of change in the Rg of the dimer relative
As shown here a 7.2 M solution of guanidine to that of the monomer indicates that the
hydrochloride can completely dissociate F1 V into monomer, which exists as an open extended
its monomeric form. structure, may dimerize laterally, with the two
It is apparent that F1 V exists like an unfolded chains lining up along their sides to form a
protein from the results of three of the experi- structure for which the geometric dimensions are
mental methods employed in this study: the not significantly different from the monomer. This
unusually large apparent molecular weight model is in agreement with Powell et al. s13
obtained from SEC, the unusually large Rg proposal of a fibrillar aggregate structure for
obtained from MALS, and ANS dye binding F1 V.
studies which indicate that all hydrophobic The conjugation of the two F1and V proteins
regions of the protein are exposed to solvent. results in a protein that although distinct from
Such proteins can be analyzed by either random either F-1 or V exhibits aggregation behavior
coil or worm-like chain models.17,25,26 For a consistent with the physical parameters of the
random coil F1 protein. As seen from the pH-solubility profile
the solubility minimum of the protein occurs at
pffiffiffiffi
ffi
b N
pH 5, close to its pI. This correlation between
Rg ź
6
solubility minimum and pI, consistent with the
lack of repulsive forces that result in aggregation,
where b is the length of an individual segment and
is a fairly common occurrence in proteins. Our
N the number of segments in the chain.26 For
data, however, show that F1 V is not a globular
proteins, b equals 0.38 nm, the distance between
protein. In contrast to most globular proteins that
amino acid a carbons and N equals the number of
exhibit a secondary and a tertiary structure, F1 V
amino acid residues. Using this model for F1 V,
seems to lack tertiary structure and like most
with N ź 478,13 yields a Rg of 3.3 nm, smaller than
unfolded proteins is prone to aggregation via
experimentally observed 10 nm. Alternatively, for
hydrophobic interactions.
flexible high molecular weight proteins, the
The task of reformulating F1 V was under-
Kratky Porod worm-like chain model can be used
taken primarily to reduce the pH of the solution
to approximate the radius of gyration as
from 9.9 (as supplied by the manufacturer) to a
rffiffiffiffiffiffi
lower pH, namely 7.5, because the high pH (i) can
Lr
Rg ź
accelerate chemical degradation, ex. in a solution
3
at pH 10 the rate constant for the deamidation of a
where the contour length L ź bN, and r is the hexapeptide was shown to be about 350 times that
chain persistence length, a measure of chain at pH 7.528 and (ii) can cause significant pain and
stiffness and extension.25,27 Note that higher irritation following intramuscular injection. As
order terms in the worm-like chain model can seen in Figure 3b a simple buffer exchange over
be ignored for low r/L values. Using Rg ź 10 nm, 12 h results in a significantly aggregated protein
this model gives r ź 1.7 nm for F1 V. This value is mixture whose immunogenicity is questionable.
roughly 4.5 amino acids in length, and agrees well Therefore, descending denaturant concentration
with the persistence lengths obtained for a dialysis was undertaken to gradually remove the
number of unfolded proteins, which are often in denaturant while maintaining the F1 V mole-
the range of 2 10 amino acids.25 The large cules in the presence of the final formulation
persistence length of F1 V indicates that it is a buffer containing Mgþ2, a surfactant and arginine
highly extended protein. Further, we obtain a hydrochloride at a pH of 7.5. This approach
value of Rh/Rg ź 0.3 for the F1 V monomer. This resulted in a formulation that appears to populate
value agrees with those for worm-like and rodlike F1 V predominantly in its monomeric state.
chains, but not with that for tightly packed Interestingly, the renaturation process does not
flexible chains. For a random coil Rh/Rg ź 0.66, alter the hydrodynamic volume of monomeric
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009 DOI 10.1002/jps
BIOPHYSICAL CHARACTERIZATION AND FORMULATION 2601
protein. This formulation is stable for 3 days, tendency to aggregate is, perhaps, a result of its
substantially longer than was observed by a disordered nature. Despite these shortcomings it
simple buffer exchange and long enough to carry is possible to reduce the heterogeneity associated
out any secondary dehydration process like with F1 V by formulating it with an appropriate
lyophilization or spray-freeze drying. process and excipients. It is hoped that insights
Although the need for a dehydration step may gained from this analysis will aid in the develop-
appear to undermine the need for a reformulation ment of vaccines based on F1 V.
step, it should be recognized that the rates of
chemical degradation are sensitive to pH of the
initial solution even in a lyophilized product.28
ACKNOWLEDGMENTS
Song et al.28 showed that at a pH 10 the
deamidation rate constant of a hexapeptide was
The authors would like to thank Dr. Vicki Pierson
about 4 times that at pH 7.5 in lyophilized
of NIAID and NIH Biodefense and Emerging Infec-
powders maintained at 50% RH. Interestingly,
tions Research Resources Repository, NIAID, NIH
at acidic pHs the deamidation rate constants were
for donating the F1 V, David Nellis of SAIC-Fre-
even higher in the rubbery solid matrices than in
drick for insightful discussions and Dr. Bradford
solution which were comparable to glassy solid
Powell of USAMRIID for insightful conversations
matrices. Since, chemical degradation rates in
and manuscript review.
proteins can be dependent on pH of the initial
solution even in their dehydrated forms main-
taining the pH is important to preserve the
integrity of the protein molecule over the long-
REFERENCES
term. Secondly, it is well known that pain and
irritation following an intramuscular injection
1. Andrews GP, Heath DG, Anderson GW, Jr., Welkos
increases as the pH of parenterals deviates from
SL, Friedlander AM. 1996. Fraction 1 capsular
7.4 in either direction. So to minimize this pain
antigen (F1) purification from Yersinia pestis
and irritation the dehydrated powder has to be
CO92 and from an Escherichia coli recombinant
reconstituted such that resulting solution is at
strain and efficacy against lethal plague challenge.
or near pH 7.4. Therefore, it is important to
Infect Immun 64:2180 2187.
reformulate F1 V despite the eventual dehydra-
2. Tito MA, Miller J, Griffin KF, Williamson ED, Tit-
tion process.
ball RW, Robinson CV. 2001. Macromolecular orga-
Although most proteins are folded in their nization of the Yersinia pestis capsular F1 antigen:
Insights from time-of-flight mass spectrometry.
native state, some appear to be naturally unfolded
Protein Sci 10:2408 2413.
and may need to be so to exert their biological
3. Simpson WJ, Thomas RE, Schwan TG. 1990.
activity.29,30 It remains to be seen if this lack of
Recombinant capsular antigen (fraction 1) from
tertiary structure in F1 V is associated with any
Yersinia pestis induces a protective antibody
biological activity. Regardless, for protein-based
response in BALB/c mice. Am J Trop Med Hyg
vaccine candidates an open, extended structure
43:389 396.
may be advantageous as it may facilitate adsorp-
4. Anderson GW, Jr., Leary SE, Williamson ED, Tit-
tion onto aluminum salt adjuvants by maximizing
ball RW, Welkos SL, Worsham PL, Friedlander AM.
hydrogen bonding with adjuvant. Further, it
1996. Recombinant V antigen protects mice against
may help present a wider array of linear and
pneumonic and bubonic plague caused by F1-
conformational epitopes on the adjuvant surface. capsule-positive and -negative strains of Yersinia
pestis. Infect Immun 64:4580 4585.
These potential advantages, however, need to
5. Nakajima R, Brubaker RR. 1993. Association
be weighed against the tendency of extended
between virulence of Yersinia pestis and suppres-
proteins to aggregate.
sion of gamma interferon and tumor necrosis factor
alpha. Infect Immun 61:23 31.
6. Nakajima R, Motin VL, Brubaker RR. 1995. Sup-
CONCLUSION
pression of cytokines in mice by protein A-V antigen
fusion peptide and restoration of synthesis by active
This study allows for a much more detailed picture
immunization. Infect Immun 63:3021 3029.
of the structure of F1 V than previously known.
7. Une T, Brubaker RR. 1984. Roles of V antigen in
F1 V exists in solution as a disordered protein and
promoting virulence and immunity in yersiniae.
not as a globular protein as previously thought. Its J Immunol 133:2226 2230.
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009
2602 D SOUZA ET AL.
8. Motin VL, Nedialkov YA, Brubaker RR. 1996. unfolded cytochrome c to molten globule state as
V antigen-polyhistidine fusion peptide: Binding to a result of electrostatic interactions. Biochemistry
LcrH and active immunity against plague. Infect 38:13635 13642.
Immun 64:4313 4318. 20. Matulis D, Lovrien R. 1998. 1-Anilino-8-naphtha-
9. Skrzypek E, Straley SC. 1995. Differential effects of lene sulfonate anion-protein binding depends
deletions in lcrV on secretion of V antigen, regula- primarily on ion pair formation. Biophys J 74:
tion of the low-Ca2þ response, and virulence of 422 429.
Yersinia pestis. J Bacteriol 177:2530 2542. 21. Latypov RF, Liu D, Gunasekaran K, Harvey TS,
10. Price SB, Cowan C, Perry RD, Straley SC. 1991. Razinkov VI, Raibekas AA. 2008. Structural and
The Yersinia pestis V antigen is a regulatory pro- thermodynamic effects of ANS binding to human
tein necessary for Ca2(þ)-dependent growth and interleukin-1 receptor antagonist. Protein Sci
maximal expression of low-Ca2þ response virulence 17:652 663.
genes. J Bacteriol 173:2649 2657. 22. Goodin JL, Nellis DF, Powell BS, Vyas VV, Enama
11. Heath DG, Anderson GW, Jr., Mauro JM, Welkos JT, Wang LC, Clark PK, Giardina SL, Adamovicz
SL, Andrews GP, Adamovicz J, Friedlander AM. JJ, Michiel DF. 2007. Purification and protective
1998. Protection against experimental bubonic and efficacy of monomeric and modified Yersinia pestis
pneumonic plague by a recombinant capsular F1-V capsular F1-V antigen fusion proteins for vaccina-
antigen fusion protein vaccine. Vaccine 16:1131 tion against plague. Protein Expr Purif 53:63 79.
1137. 23. Arakawa T, Timasheff SN. 1984. Mechanism of
12. Miller J, Williamson ED, Lakey JH, Pearce MJ, protein salting in and salting out by divalent cation
Jones SM, Titball RW. 1998. Macromolecular orga- salts: Balance between hydration and salt binding.
nisation of recombinant Yersinia pestis F1 antigen Biochemistry 23:5912 5923.
and the effect of structure on immunogenicity. 24. Arakawa T, Ejima D, Tsumoto K, Obeyama N,
FEMS Immunol Med Microbiol 21:213 221. Tanaka Y, Kita Y, Timasheff SN. 2007. Suppression
13. Powell BS, Andrews GP, Enama JT, Jendrek S, Bolt of protein interactions by arginine: A proposed
C, Worsham P, Pullen JK, Ribot W, Hines H, Smith mechanism of the arginine effects. Biophys Chem
L, Heath DG, Adamovicz JJ. 2005. Design and 127:1 8.
testing for a nontagged F1-V fusion protein as 25. Bright JN, Woolf TB, Hoh JH. 2001. Predicting
vaccine antigen against bubonic and pneumonic properties of intrinsically unstructured proteins.
plague. Biotechnol Prog 21:1490 1510. Prog Biophys Mol Biol 76:131 173.
14. Uversky VN. 2002. Natively unfolded proteins: A 26. Goldenberg DP. 2003. Computational simulation of
point where biology waits for physics. Protein Sci the statistical properties of unfolded proteins. J Mol
11:739 756. Biol 326:1615 1633.
15. Whitaker JR. 1963. Determination of molecular 27. Berry GC. 1988. Remarks on a relagtion among the
weights of proteins by gel filtration on Sephadex. intrinsic viscosity, the radius of gyration, and the
Anal Chem 35:1950 1953. translational friction coefficient. J Polymer Sci,
16. Pearse A. 1980. Histochemistry, theoretical and Part B: Polymer Phys 26:1137 1142.
applied, 4 edition. Edinburgh: Churchill Living- 28. Song Y, Schowen RL, Borchardt RT, Topp EM.
stone. 2001. Effect of  pH on the rate of asparagine
17. Millett IS, Doniach S, Plaxco KW. 2002. Toward a deamidation in polymeric formulations:  pH -rate
taxonomy of the denatured state: Small angle scat- profile. J Pharm Sci 90:141 156.
tering studies of unfolded proteins. Adv Protein 29. Sickmeier M, Hamilton JA, LeGall T, Vacic V,
Chem 62:241 262. Cortese MS, Tantos A, Szabo B, Tompa P, Chen
18. Sreerama N, Venyaminov SY, Woody RW. 2000. J, Uversky VN, Obradovic Z, Dunker AK. 2007.
Estimation of protein secondary structure from DisProt: The database of disordered proteins.
circular dichroism spectra: Inclusion of denatured Nucleic Acids Res 35:D786 D793.
proteins with native proteins in the analysis. Anal 30. Vucetic S, Obradovic Z, Vacic V, Radivojac P, Peng
Biochem 287:243 251. K, Iakoucheva LM, Cortese MS, Lawson JD, Brown
19. Ali V, Prakash K, Kulkarni S, Ahmad A, Madhu- CJ, Sikes JG, Newton CD, Dunker AK. 2005. Dis-
sudan KP, Bhakuni V. 1999. 8-anilino-1-naphtha- Prot: A database of protein disorder. Bioinformatics
lene sulfonic acid (ANS) induces folding of acid 21:137 140.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 98, NO. 8, AUGUST 2009 DOI 10.1002/jps