Biomaterials 31 (2010) 5191e5198

Contents lists available at ScienceDirect

Biomaterials
journal homepage: www.elsevier.com/locate/biomaterials

The incorporation of GALA peptide into a protein cage for an acid-inducible

molecular switch
Seung-Hye Choi, Kuiwon Choi, Ick Chan Kwon, Hyung Jun Ahn*
Biomedical Research Center, Korea Institute of Science and Technology, 39-1 Hawolgok-dong, Seongbuk-Gu, Seoul 136-791, South Korea

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

Article history: Caged proteins have been utilized as a biological container in a wide range of applications from material
Received 20 January 2010 science to biomedicine, and GALA peptide has been known to undergo coil-to-helix transition upon the
Accepted 9 March 2010 increased acidity. In this study, GALA synthetic peptide is incorporated to cage protein by genetic
Available online 31 March 2010
modification. Our engineered caged scaffold retains intact at the physiological pH but dissociate
completely at pH 6.0, and the dissociated subunits are re-assembled simply by neutralization to bio-
Keywords:
logical pH. This acid-induced dissociation has the potential as molecular switch in vivo as well as in vitro
Cage protein
so that the acid-sensitive caged proteins are applicable to drug delivery system for acidic target sites such
GALA peptide
Self-assembly
as tumor. Since our design depends on the conformational transition of GALA peptide, not on removal of
Molecular switch characteristic interface observed only in viral capsid-like protein, non-viral caged proteins can also be
Disassembly engineered to have molecular switching function. Therefore, this design for acid-sensitive scaffold would
broaden the width of applications in nanotechnology including biomimetic material synthesis and
biomedicine.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction acidic solutions [7,8,10]. However, any natural caged proteins have
not been reported to assemble at the physiological pH but revers-
Caged proteins are the spherical macromolecules that are ibly dissociate pH 6.0. Others have speculated that redesigning
precisely self-assembled from a limited number of subunits. This subunitesubunit interactions in the caged protein complex could
assembly is modulated by the extensive subunitesubunit interac- be a potential strategy for controlling its architecture and assembly
tions between the adjacent subunits. Container-like cage proteins [11]. In several viral capsids, the assembly profile was altered by
have played a role as the nanoscale platform in the biomimetic removing the N-terminus of the subunits, resulting in the virus
nanoparticle synthesis [1] and have provided a wide range of particle with a smaller volume [12,13]. Similarly by deletion of an
possible applications in biomedicine, including medical imaging embracing N-terminus arm, Wang and co-workers reported the
[2], and drug delivery [3]. Since the caged proteins have three first example that one caged protein, dihydrolipoamide acetyl-
distinct interfaces (the interior, exterior, and interface between the transferase (E2) could be engineered to correctly assemble at
subunits), chemical or genetic modifications at these interfaces physiological pH but irreversibly disassemble into aggregation at
have imparted designed functionality to the cage [2]. However, pH 5.0 [14,15]. However, this approach has the limitation that cage
despite recent advances in protein synthetic and computational protein without such N-terminally embracing arm cannot be
methods, it has been still a considerable challenge to redesign cage engineered.
proteins for the functionality. The N-terminal control of assembly, which modulates the
Several caged proteins have been reported that their assembly scaffold in viral capsid-like cages, is not the common feature in the
could be modulated by pH variation [4e8]. For example, CCMV ferritin-like [9,16] and closed shell cage-like cages [17e19]. For
(Cowpea Chlorotic Mottle Virus) and Norwalk capsids assemble at example, ferritin and Dps (DNA-binding protein from the starved
low pH and disassemble at high pH [4e6]. Ferritin proteins remain cells) caged proteins are composed of the compact subunits
intact over a wide range of pH [9], and reversibly disassembles without an embracing structural motif, so the deletion method is
under the acidic guanidine hydrochloride (GdnHCl) or strongly not applicable to these non-viral capsid-like cages for acid-induced
molecular switch.
In the applications of drug and gene delivery, the GALA peptide
* Corresponding author. Tel.: þ82 2 958 5938; fax: þ82 2 958 5909. has been designed to interact with membranes in a pH-sensitive
E-mail address: [email protected] (H.J. Ahn). manner [20,21]. This peptide is composed of the artificial synthetic

0142-9612/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biomaterials.2010.03.016
5192 S.-H. Choi et al. / Biomaterials 31 (2010) 5191e5198

amino acid repeat of a Glu-Ala-Leu-Ala [20], and one sequence In the present studies, the engineered proteins were expressed
example of GALA peptide is WEAALAEALAEALAEHLAEALAEALEALA, in Escherichia coli expression systems and purified after genetic
of which the sequence length can vary to longer or shorter than 30 modification according to our strategy. The resulting proteins were
amino acids. The most remarkable feature of this peptide is its characterized by size exclusion chromatography, dynamic light
reversible conversion (Fig. 1) from a random coiled structure to an scattering, SDS polyacrylamide gel electrophoresis, and Circular
a-helix when the pH is reduced from 7.0 to 5.0 [20,21]. This unusual Dichroism. To utilize the coil-to-helix transition of GALA by acid,
behavior originates from the ordered repetition of the apolar and various lengths of GALA peptides were incorporated into the scaf-
polar residues. At the neutral pH, electrostatic repulsions between fold and investigated.
the carboxylic acid moieties of the glutamic acids are expected to
destabilize the a-helix, whereas at pH 5.0, the neutralization of 2. Materials and methods
these groups should promote a-helix formation, resulting in the
2.1. Design and construction of ferritin-GALAn hybrid protein
localization of the hydrophobic leucine residues on one side of the
a-helix and the pH-titratable hydrophilic glutamic residues on The human ferritin light chain gene was amplified by the polymerase chain
the other side. Moreover, as the glutamic acids are protonated, the reaction using human cDNA library as template. Three different length of GALA
hydrophobicity of the glutamic acid side chain increases [20]. The peptides were chosen to be inserted to E-helix truncated ferritin, and depending on
the composition of amino acids, three ferritin-GALA hybrids, ferritin-GALA2, ferritin-
pH value at the midpoint of maximum change of the helical content GALA4, and ferritin-GALA6 were named fGALA2, fGALA4, fGALA6, respectively. The
in GALA peptide was reported to be 6.0, and pKa value of glutamic forward oligonucleotide primers for fGALAn (n ¼ 2, 4, and 6, respectively) constructs
acid in polyglutamic acid was close to 6.0 [21,22]. Circular dichroism were 50 -G GAA TTC CAT ATG AGC TCC CAG ATT CGT CAG-30 and shared for each of
(CD) measurements also demonstrated the increase of an a-helical fGALAn hybrids. The reverse oligonucleotide primers for fGALA2, fGALA4, and fGALA6
constructs were 50 -CCG CCG CTC GAG TTA AGC TTC AGC TAA AGC TTC AGC TAA AGC
conformation in GALA peptide as the pH was decreased from 7.5 to
CTC CGG GCC-30 , 50 -CCG CCG CTC GAG TTA TTC AGC TAA AGC TTC AGC TAA ATG TTC
5.0 [21]. These data indicated that GALA peptide could undergo coil- AGC TAA AGC TTC AGC TAA AGC CTC CGG GCC-30 , 50 -CCG CCG CTC GAG TTA TTC AGC
to-helix transition near pH 6.0. In this study, we designed cage TAA AGC TTC AGC TAA AGC TTC AGC TAA AGC TTC AGC TAA ATG TTC AGC TAA AGC
protein genetically to utilize GALA peptide as acid-inducible TTC AGC TAA AGC CTC CGG GCC-30 respectively, where the bases in bond represent
molecular switch (Fig. 1). the NdeI and XhoI restriction enzyme cleavage sites. Amplified wild-type ferritin
DNA was templated to remove the whole C-terminal fragment region ranging
We selected human ferritin light chain for a cage model. This between the E-helix domain (161Gly-170Leu) and C-terminus (175Asp). And three
cage protein differed in viral capsid-like proteins in that there was different length of GALA genes were substituted for those eliminated C-terminal
no N-terminal embracing arm (Fig. 2A and B) and therefore, we fragment, resulting in fGALAn hybrid construct. The amplified DNA was inserted into
could not use deletion method. Ferritin is typically spherical capsid the NdeI/XhoI e digested expression vector pET-28a(þ) by T4 DNA ligase. This vector
construction adds a 20-residue tag (MGSSHHHHHHSSGLVPRGSH) to the amino-
with 432 point symmetry [23]. Cage structure of ferritin has 24
terminus of the gene product in order to facilitate protein purification. Each of
subunits with each subunits composed of four-helix bundle (which constructed expression vectors were sequenced and confirmed.
included A, B, C, and D helix) and E-helix (linked at the C-terminal
end) (Fig. 2A). A fifth short a-helix is tilted towards the bundle axis 2.2. Protein expression and purification
and masks the one end of the cylindrical bundle. The recombinant
The proteins were over-expressed in E. coli BL21(DE3) cells. Cells were grown at
human ferritin heavy chain has been subjected to numerous
310 K to an OD600 of 0.5 in LB medium containing 50 mg/mL kanamycin and protein
mutational studies to investigate in vivo reassembly. Although in expression was induced by 1.0 mM isopropyl-b-D-thiogalactopyranoside (IPTG). Cell
vitro stability was markedly decreased, assembly was not pre- growth continued at 293 K for 18 h after IPTG induction and cells were harvested by
vented by the deletion of the first 13 N-terminal residues (D1e13), of centrifugation at 4200g (6000 rev min1; Sorval GSA rotor) for 10 min at 277 K. The
the last 22 C-terminal residues (D161e182) [24], or by various cell pellet was resuspended in ice-cold lysis buffer (50 mM TriseHCl pH 8.0, 100 mM
sodium chloride, 1 mM phenylmethylsulfonyl fluoride) and homogenized with an
substitutions in the four-helix bundle, in the hydrophilic channels,
ultrasonic processor. The crude lysate was centrifuged at 70 400g (30 000 rev min1;
or in the inner and outer surfaces of the cage [25]. Beckman 45Ti rotor) for 1 h at 277 K and the recombinant protein in the supernatant
Cage proteins with acid-inducible molecular switch are poten- fraction was purified in two chromatographic steps. The first step utilized the
tially applicable to drug delivery systems [2,26]. For example, hexahistidine tags by metal-chelate chromatography on Ni-NTA resin (Qiagen).
Next, size exclusion chromatography (SEC) was performed on a Superdex 200 10/
capsulated drug within cage protein can avoid side effects and
300 GL column (GE Healthcare) previously equilibrated with buffer containing
physicochemical instability in vivo. And drug release is expected to 50 mM TriseHCl (pH 8.0), 100 mM sodium chloride, and 1 mM mercaptoethanol. The
be accomplished by dissociation of cage after arrival at acidic target homogeneity of the purified protein was assessed by SDS-PAGE. The protein solution
sites [26e29]. was concentrated using an YM10 ultrafiltration membrane (Amicon). The protein
concentration was estimated using Bradford assay method with bovine serum
albumin as a standard.

2.3. Analysis of oligomeric state of proteins

After purification, each of all proteins in this study was incubated for 1 h at room
temperature in the pH range 3.0e8.0, and then analyzed by size exclusion chro-
matography (Superdex 200 10/300 GL column). Oligomeric states were judged from
the elution volume (Ve) and bands on denatured SDS-PAGE. Regardless of buffer
condition, all equilibrated buffer contained 100 mM sodium chloride and 1 mM
mercaptoethanol, and injection volume on SEC did not exceed about 2 mL. Before
analyzing proteins in this research, standard protein curve was prepared so that
24-mer of wild-type ferritin (Mr 440 kDa), BSA (Mr 67 kDa), and Cytochrome C (Mr
14 kDa) were eluted at Ve ¼ 9e10 mL, 15e16 mL, and 20e21 mL, respectively.
Fig. 1. Schematic diagram of acid-inducible disassembly of engineered protein cage by Protein elution profiles on SEC were monitored by measuring the absorbance at
GALA peptide. GALA peptides (yellow) are introduced to pore sites on cage scaffold 280 nm. When needed, the protein samples were concentrated on Centricon 30
without loss of assembly. Coil-to-helix transition of GALA (magenta) is triggered by membrane (Amicon).
acid and induces the conformational change around artificial pores, leading to disso-
ciation of caged particles. Our engineered protein scaffold retains intact at physio- 2.4. Dynamic light scattering (DLS) and circular dichroism (CD)
logical pH but reversibly dissociate at pH 6.0. The dissociated subunits re-associate
simply by returning to physiological pH. Each subunits of caged protein is represented DLS experiments were performed using a model DynaPro-801 instrument from
by hexagons, and yellow and magenta colours on protein subunits denote the random Protein Solutions (Lakewood, New Jersey). The data were measured at 297 K with
coiled and helical structures in GALA, respectively. the protein at 1 mg/mL concentration in a variety of buffer conditions containing
 S.-H. Choi et al. / Biomaterials 31 (2010) 5191e5198 5193

100 mM sodium chloride and 1 mM mercaptoethanol. CD scans were performed on

a Jasco J-715 spectropolarimeter equipped with a Jasco peltier temperature
controller. We used far-UV circular dichroism to characterize the protein folding/
unfolding and thermal stability. Samples at concentration of about 0.075 mg/mL in
50 mM potassium phosphate (pH 6.0 and 7.0, respectively) and 100 mM NaCl were
scanned between 190 and 260 nm at 25  C at a scanning speed of 10 nm/min in
0.1 cm path length quartz cells. To evaluate thermal stability, we monitored the
ellipticity at 222 nm from 5  C to 100  C at a heating rate of 0.5  C/min.

2.5. Electrophoresis and dialysis

Denaturing electrophoresis was carried out on 15% poly-acrylamide gels, and

proteins were stained with Coomassie Blue or silver stain (Bio-Rad). Samples were
diluted with an 5-fold concentrated sampling buffer (0.25 M Tris/HCl buffer, pH 6.8),
which contained SDS, b-mercaptoethanol, and glycerol. After destaining, if needed,
densitometry was performed on a Computing Densitometer (Molecular Dynamics).
We changed pH by buffer exchange with dialysis membrane tube (Mr. cutoff
3000 Da, Millipore) at room temperature for at least 4 h.

3. Results and discussion

3.1. Design of protein scaffold for acid-sensitive disassembly

Structural analysis of ferritin in atomic levels revealed that two

kinds of pores existed around the 3- and 4-fold axis, respectively
(Fig. 2B). Around the 4-fold axis, six pores, which were composed of
the adjacent four E-helices, were investigated for possible incor-
poration of GALA peptide (Fig. 2B). To substitute the E-helix with
GALA peptide (Fig. 2B and C), we linked GALA peptide to the end of
DE loop (160Ala) by genetic modification (Fig. 3A and B). SDS-PAGE
analysis revealed that alteration of the native pores by deletion of
the E-helices did not affect the expression and solubility of the
subunits (Fig. 3C). Since the sequences of DE loop were not modi-
fied and sequentially followed by GALA, we expected that GALA
could form the artificial pores on behalf of the E-helices (Fig. 3C).
To examine if the sequence length of GALA peptide affected on
acid-sensitive disassembly, three kinds of ferritin-GALA hybrids
with variation of Glu-Ala-Leu-Ala amino acid repeat were prepared
by the same redesign (fGALA2, fGALA4, and fGALA6 respectively)
(Fig. 3A). Each constructs showed the high expression and solu-
bility levels, as did the native ferritin (Fig. 3D), implying that genetic
modification for incorporation of GALA peptide did not hinder the
correct folding of subunits. Because lysis buffer maintained pH 8.0,
a random coiled structure of GALA peptide at least did not affect the
folding process of the subunits.

3.2. Correct assembly of native ferritin and ferritin-DE mutant at

a wide range of pH 4.0e8.0

Generally, ferritin caged proteins have been reported to disso-

ciate only under the strongly acidic condition and reversibly
re-associate at or above neutral pH [30e32]. Horse spleen apo-
ferritin dissociates under the extreme pH conditions below 2.8 and
reversibly re-associates at pH value above 3.0 [30,31]. Rhodobacter
capsulatus bacterioferritin is 100% 24-mer over the pH range
3.0e10.0 and dissociates at pH below 2.0 as well [32]. These
assembly behaviors have been utilized in the medical imaging so
that GdIII chelates, GdHPDO3A could be entrapped within ferritin

different colours around 3- and 4-fold symmetry axis, respectively (left and right). Pore
structures around 4-fold symmetry axis were magnified by dashed circle after rotated
by 90 along the horizontal axis. For clarity, one monomer in the relation of the 4-fold
symmetry is abbreviated and three E-helices are coloured by green. (C) Incorporation
of GALA peptide into protein cage. The view point is the same as Fig. B. Random coiled
Fig. 2. Ribbon diagram of monomer subunit and native pore structure in ferritin caged
structure of GALA peptide in pH 7.4 (yellow) does not hinder assembly of protein cage,
protein, and schematic diagram of fGALA hybrid around symmetry axis. (A) Subunit is
but transition to helical structure at pH 6.0 yields molecular repulsion (magenta)
composed of four-helix bundle and fifth E-helix. Each of five helices are displayed with
between GALA peptide in narrow artificial pores. This figure is prepared with PyMOL
different colours. (B) Individual subunits of 24-mer caged particle are represented with
(DeLano, 2002).
5194 S.-H. Choi et al. / Biomaterials 31 (2010) 5191e5198

Fig. 3. Various mutants of ferritin caged proteins and secondary structure analysis. (A) Summary of various ferritin mutants and GALA peptide sequences. GALA sequences are
denoted by one-letter code. (B) Secondary structure analysis of human ferritin light chain reveals five helices on sequences. The cylinders correspond to helices and deleted E-helix
region is highlighted with a rectangular. (C) Denatured SDS-PAGE analysis for E-helix-deleted ferritin shows the high level of expression and solubility in E. coli system. () and (þ)
represent before and after IPTG induction, respectively. (D) Solubility test of fGALA hybrids in E. coli expression system. fG2, 4, and 6 represent each of fGALA2, 4, 6 hybrids.

interior by dissociation at pH 2.0 and subsequently re-association at difference in the elution volume of 24-mer between native and
pH 7.0 [10]. However, we are not aware of any ferritin proteins that ferritin-DE. The hydrodynamic diameters measured by DLS for the
remain assembled at the neutral pH but disassemble at pH 6.0. native and ferritin-DE were 12.4  0.4 nm and 12.9  0.8 nm at the
At first, we examined the pH-dependent disassembly of the pH 7.0, respectively, and there was no noticeable difference in size
native human ferritin light chain using size exclusion chromatog- between the native and ferritin-DE. But in the pH value below 4.0,
raphy (SEC) and denatured SDS-PAGE. Each protein samples were DLS analysis could not measure the exact size due to the high
dialyzed at room temperature against the different pH values polydispersity (more than 90%), which seemed to be another
ranging from 3.0 to 8.0, and then concentrated to 1 mg/mL by evidence of disassembly.
ultrafiltration membrane. To make the same environmental
condition as protein samples, the SEC column was pre-equilibrated 3.3. Effect of GALA peptides on cage assembly in pH 7.0 and 8.0
with corresponding buffer and 2 mL of samples was loaded on
every injection. Below pH 4.0, most ferritin existed as unassembled Because our engineered protein should retain intact cage
building blocks as judged from the elution volume, Ve ¼ w20 mL assembly at the physiological pH, we investigated the effect of
after size exclusion chromatography, and any cage structures were GALA peptides on assembly of the cage at or above neutral pH. Each
not observed (Fig. 4). However, in the pH range of 4.0e8.0, the peak
appeared in the earlier elution volume (Ve ¼ w9 mL) and corre-
sponded to the size of 24-mer ferritin (Fig. 4), indicating the native
ferritin subunits self-assembled and retained its cage structure in
these pH range. This result demonstrated that disassembly in
human ferritin light chain was triggered only by harsh acidic
condition (below pH 4.0), as were other ferritin proteins.
Since we tried to build the artificial pores by alteration of the
native pores, we needed to examine if the native pore structures
modulate assembly of cage. Therefore, pH-dependent profile of the
truncated ferritin (ferritin-DE) was investigated with variation of
pH. The elution peaks resembled those of the native ferritin (Fig. 4)
at a wide range of pH values, as ferritin-DE assembled intact at pH
4.0e8.0 and dissociated below pH 4.0. This suggested that we could
alter the pore-forming structural motif without loss of assembly.
Fig. 4. Size exclusion (SEC) chromatography profile of native ferritin and mutant
Especially, because the E-helices were positioned in the hollow ferritin-DE. Straight lines correspond to SEC profiles at various pH range (4.0, 5.0, 6.0,
sphere of the cage, the outer diameter of ferritin-DE was expected 7.0, and 8.0), and dashed lines do at pH 2.5 and 3.0. There was no difference in elution
not to differ in that of the native ferritin, and actually there was no profiles between native and ferritin-DE.
 S.-H. Choi et al. / Biomaterials 31 (2010) 5191e5198 5195

Fig. 5. SEC profiles of fGALA4, 6 hybrids. (A) Oligomeric states of fGALA4, 6 hybrids are analyzed on SEC column at various pH solutions. Population of each species is analyzed on
denatured SDS-PAGE. Circles, squares, and triangles denote the different oligomeric states of protein particles, respectively. In the view of acid-dependant profile, only fGALA2 hybrid
coincided with the native ferritin rather than other two hybrids. (B) Superimposed SEC profiles of each fGALA4, 6 hybrid at various pH conditions. Whenever protein samples were
injected, their volume and concentration were adjusted to the equal value for comparison between profiles. (C) Decreased peak areas corresponding to 24-mer in Fig. B are plotted
at various pH conditions to predict the transition pH range of caged particles. The calculated area at 24-mer peaks is postulated to be proportional to the amount of assembled cage.

of three different fGALAn (n ¼ 2, 4, and 6) hybrids were dialyzed hybrids, not fGALA2, yielded the shifted elution peaks on SEC
against both pH 7.0 and 8.0, and subsequently each samples were column, implying the conformational transition from 24-mer to
concentrated to 1 mg/mL. 2 mL of samples were injected on the small building blocks (Fig. 5A and B). In this study, two fGALA4 and
pre-equilibrated SEC column with both pH 7.0 and 8.0 buffer, and fGALA6 hybrids were named fGALA4, 6 hybrids to distinguish this
collected fractions from the individual peaks were analyzed by behavior from that of fGALA2 hybrid because we did not notice any
denatured SDS-PAGE. At both pH values 7.0 and 8.0, all fGALAn difference between two fGALA4 and fGALA6 hybrids.
hybrids retained their cage structures as judged from the elution Two fGALA4, 6 hybrids revealed the apparent dissociation under
volume (Ve) (Fig. 5A). Especially, the elution peak Ve ¼ w9 mL weak acidic conditions as following. At pH 6.8, the peak corre-
coincided with that of the native ferritin at pH 4.0e8.0 (Fig. 4), so it sponding to Ve ¼ w20 mL in fGALA4, 6 hybrids, increased as much
corresponded to the size of 24-mer. Dynamic light scattering as the height of 24-mer peak decreased, and this revealed the
analysis also indicated all fGALAn hybrids to have molecular weight partial dissociation was triggered. Interestingly, at more acidic
of w500 kDa (12.5  0.3 nm in diameter) with a polydispersity condition, pH 6.5, another peaks of Ve ¼ w15e17 mL were subse-
range of 20.0e24.5%. Because the calculated monomer weight quently eluted with considerably high population after the 24-mer
including the N-terminal tag is about 22 kDa, it corresponded to peak (Fig. 5A). These fractions corresponded to the intermediate
24-mer cage particle. From these results, we could prove that all oligomers because they eluted faster than Ve ¼ w20 mL but slower
fGALAn hybrids existed as 24-mer at pH 7.0 and 8.0, and the arti- than Ve ¼ w9 mL. Since the heights of 24-mer and smallest
ficial pores at least did not disrupt the cage structure in these pH building block decreased at the same time, these fractions were
range. And the length of GALA in the artificial pores was not likely regarded as mixture of partially dissociated particles or subas-
to affect the assembly, either. sembly of the smallest building blocks (Fig. 5B). And the poly-
dispersity value of DLS analysis at pH 6.5 increased more than 90.0%
3.4. Acid-sensitive dissociation of protein cage in pH 6.0 so that it implied the species of cage particles was highly
heterogeneous.
To demonstrate the acid-sensitive disassembly predicted by our Remarkably, below and at pH 6.0, the caged particles of fGALA4, 6
designing, each fGALAn sample was prepared at various acidic hybrids dissociated completely so that only single peak was
conditions by dialysis as described above and then, analyzed on the observed at Ve ¼ w17 mL (Fig. 5A and B). According to comparison
SEC columns and denatured SDS-PAGE. In contrast to at pH 7.0 and with our standard markers, elution of this single peak was slower
8.0, as pH decreased from 7.0 to 6.0, two fGALA4 and fGALA6 than BSA (Mr 67 kDa) but faster than cytochrome C (Mr 14 kDa), so
5196 S.-H. Choi et al. / Biomaterials 31 (2010) 5191e5198

we could demonstrate that cage particles dissociated completely at 3.6. Unfolding of subunits coupled to dissociation of caged particles
pH 6.0 and below pH 6.0. DLS analysis also revealed that the
molecular weight was w70 kDa with 4.2  0.2 nm in diameter Since the ferritin proteins are composed only of the a-helices,
(whereas the calculated molecular weight of cage was about they have the characteristic profile of minima at 208 and 222 nm on
530 kDa), supporting disassembly of cage particles. When circular dichroism spectra. In the stability study of R. capsulatus
compared with the native ferritin in Fig. 4, our engineered ferritin bacterioferritin, the CD and tryptophan fluorescence emission
cage dissociated completely at pH 6.0, whereas the native ferritin spectra demonstrated that any secondary or quaternary structural
did below pH 4.0. Since most natural ferritin proteins dissociate in change were not observed in the pH range 3.0e8.0, but the a-helix
harsh acidic condition such as at pH 3.0 [7e9], this is the first content at or below pH 2.0 decreased substantially [32]. Another CD
example of complete dissociation of ferritin cage at pH 6.0 without spectra of the recombinant horse ferritin light chain reported
loss of assembly at neutral pH, and these results showed our design a decrease in the a-helix content at acidic conditions (at or below
could impart the acid-sensitive disassembly to the engineered cage pH 2.0) as well [31]. These reports explain unfolding of the subunits
protein. is coupled to dissociation of the native ferritin at strongly acidic
Since the individual peak areas on SEC profiles were propor- conditions although a great deal of ordered structures are still
tional to the population of the corresponding species, we tried to present in the subunits and unfolding mechanism still remains
estimate a transition pH value by comparison between SEC profiles. veiled [31,32]. Similarly, in the native ferritin of our scaffold, we
The decreased ratio of cage assembly was determined by calcula- have also observed the representative CD spectra of a-helical
tion of 24-mer peak areas at each of pH values (Fig. 5B and C). proteins at pH 6.0 and 7.0 (Fig. 6A), and a-helix content also
Although the aggregation tendencies of concentrated samples decreased substantially below pH 4.0 (data not shown).
increased at acidic conditions, such aggregates were removed by
the high speed ultracentrifugation before injection and only the
soluble fractions were analyzed on SEC column. Whenever
analyzing the proteins samples on SEC column, since we adjusted
the sample concentration and the injection volume to the equal
values (1 mg/mL and 2 mL, respectively) so that it was possible to
compare the elution profiles. A sigmoidal curve fitted to this data,
reveals that two fGALA4, 6 hybrids undergo a transition from an
assemble state to a disassemble state in the pH range of 6.0e7.0,
and w80% dissociation of cage particles was observed near pH 6.5
for two fGALA4, 6 hybrids (Fig. 5C).
On the other hand, fGALA2 did not dissociate its cage in the pH
range of 4.0e8.0, but disassembled completely to be small building
blocks only below pH 4.0 (data not shown). This acidity-dependant
disassembly of fGALA2 resembled that of the native ferritin (Fig. 4).
Relative to two fGALA4, 6 hybrids, the artificial pores in fGALA2 was
built by only two repetition of Glu-Ala-Leu-Ala amino acids, and
this implied that the length of GALA was too short to have any
stable a-helical structure or to yield molecular interaction between
the adjacent GALA peptides.

3.5. Re-assembly capability

The design strategy proposed by Wang and coworker reported

that their engineered cage protein could not re-assemble after
dissociation because disassembly of protein cage lead to aggre-
gation [14,15]. However, the re-assembly capability is important
in application to drug delivery system because drug or cargo
should be capsulated by the dissociated subunit during re-
assembly process. If dissociation and re-association process of the
caged particles are irreversible, it would be impossible to cap-
sulate the cargo inside cage particles. Hence, we tried to inves-
tigate if the disassembled subunits could re-assemble into caged
particle simply by buffer change. The corresponding peak of
dissociated subunits was pooled after the sample was injected on
SEC column at pH 6.0. To neutralize the sample, the pooled
fraction was dialyzed against the buffer containing 50 mM
TriseHCl (pH 7.0) and 100 mM NaCl, and then concentrated to
1 mg/mL. When the sample was injected again on SEC column
equilibrated at pH 7.0, we could observe the same 24-mer peak as
Fig. 5A, indicating that acid-sensitive dissociation was reversible
and the dissociated subunits could re-associate into caged parti- Fig. 6. Far-UV circular dichroism spectra for native and fGALA4, 6 hybrids. (A) CD
spectra of the native ferritin at pH 6.0 and 7.0. Since ferritin subunits are composed of
cles by pH neutralization. Therefore, this result shows the first only a-helices, caged scaffold shows two minima at 208 and 222 nm. (B) CD spectra of
example of engineered caged scaffold to be reversibly re-assem- fGALA4, 6 hybrids at pH 6.0 and 7.0. Loss of 208 nm minimum and a shift in 222 nm
bled by neutralization. minimum at pH 6.0 indicate unfolding events.
 S.-H. Choi et al. / Biomaterials 31 (2010) 5191e5198 5197

Fig. 7. SEC profiles of fGALA4, 6 hybrids at various protein concentrations. Equal

volume of protein samples with variation of its concentration (0.075, 0.15, 0.3, and
0.6 mg/mL, respectively) are injected on SEC column at pH 7.0. These profiles reveal
that assembly of caged particles is independent of protein concentration.

Interestingly, as we expected, our engineered cage protein

revealed the different CD profile at pH 6.0. Before triggering
dissociation, two fGALA4, 6 hybrids also had the similar CD profile to
that of the stable caged particles at pH 7.0 (Fig. 6B). However, when
disassembly process was accomplished at pH 6.0, a remarkable
decrease of a-helix content at 208 nm minimum and a shift in
222 nm minimum were observed, representing variation of
secondary and quaternary structure (Fig. 6B). Especially, robust
variation of minima at 208 nm accounted for unfolding process of
subunits. Since this variation of CD spectra meant unfolding of
subunits, and elution profile on SEC column explained dissociation
of cage, we could conclude that unfolding of the subunits in two
fGALA4, 6 hybrids were coupled to dissociation of caged particles.
The native ferritin did not show any CD spectral variation in pH 6.0
so that the unfolding was anticipated to be triggered by coil-to-
helix transition of GALA.

3.7. Effects of protein concentration and ionic strength Fig. 8. Temperature stability profiles of native ferritin and fGALA4, 6 hybrids. The
ellipticity value at 222 nm is monitored to evaluate the degree of unfolding by heating.
One variant with the substitution Lys169 / Arg in the (A) The native ferritin does not show any transition event over a wide range of
recombinant human ferritin heavy chain has been reported to exist temperature both at pH 6.0 and 7.0. The slow increase of molar ellipticity during
heating accounts for the high degree of stability in native ferritin. (B) In contrast to the
as dimer in low protein concentration, but to be readily reas-
native ferritin, fGALA4, 6 hybrid reveals unfolding with a sigmoidal profile (Tm
sembled after concentration by membrane ultrafiltration [33]. To 78.0  1.0  C) at pH 7.0. As pH is lowered more to pH 6.0, additional transition point is
examine if our fGALA4, 6 hybrids form caged particles only in the observed at 39.0  0.5  C. At pH 6.0, any transition event is not observed.
high concentrated state, we have compared the oligomeric states at
the different protein concentrations of 0.075, 0.15, 0.3, and 0.6 mg/
mL, respectively. Due to the relatively low solubility of the samples, apparent midpoint unfolding temperatures (Tm). The native
it was impossible to concentrate more than 1 mg/mL. At pH 7.0, ferritin proteins remain intact over a wide range of temperature
although protein concentrations were increased up to eight times and this high thermostability has enabled ferritin to be utilized
higher for each of two fGALA4, 6 hybrids, only the caged particles widely as nanocontainer in material synthesis [9].
were observed on the profiles of SEC column (Fig. 7), and assembly In the native ferritin of our study, we could not measure any
of two fGALA4, 6 hybrids was not affected by the protein concen- apparent melting temperatures at pH 6.0 and 7.0 (at which cage
tration. In addition, when the acidity increased, acid-sensitive assembly retained intact) because the ellipticity values increased
disassembly of two fGALA4, 6 hybrids did not depend on the protein slowly and gradually (Fig. 8A). Relatively small variation of ellip-
concentration either (data not shown). ticity in a wide range of temperature explained high thermosta-
If unfolding of the subunits is modulated by electrostatic forces, bility of the native protein. In contrast, our acid-sensitive scaffold
variation of ionic strength would affect the acid-sensitive disas- revealed the apparent sigmoidal curve with a Tm 78.0  1.0  C at
sembly of cage particles. To investigate the influence of ionic strength pH 7.0 (Fig. 8B) and these transition inflections could account for
on acid-sensitive disassembly, we increased the NaCl concentration unfolding of scaffold by heating. Although the different profile
from 100 mM to 500 mM or decreased from 100 mM to zero, but did between the native and our scaffold was anticipated to be due to
not observe any different behavior on SEC profiles (data not shown). incorporation of the artificial pores into scaffold, the acid-sensitive
Hence, we could conclude that assembly process of our protein scaffold was still thermostable at pH 7.0. Remarkably, at pH 6.5,
scaffold was not affected by either protein or salt concentration. which corresponded to the early transition state of disassembly,
acid-sensitive scaffold had two melting temperature at
3.8. Thermal stability of fGALA4, 6 hybrids 39.0  0.5  C and 75.0  1.0  C (Fig. 8B). According to the analysis of
SEC profiles on above, the acid-sensitive scaffold existed as mixture
Because incorporation of artificial pores had possibility to lower of several oligomeric states in the transition pH range so we
the thermostability of our scaffold, we monitored the ellipticity at interpreted this unfolding profile as superimposition of several
222 nm as the protein samples were heated and calculated profiles, which corresponded to each oligomeric species. The first
5198 S.-H. Choi et al. / Biomaterials 31 (2010) 5191e5198

transition temperature, Tm 39.0  0.5  C, was likely to imply [7] Stefanini S, Finazzi-Agro A, Chiancone E, Antonini E. Binding of hydrophobic
compounds to apoferritin subunits. Effects on the polymerization state. FEBS
unfolding of the intermediate oligomers, which were less stable
Lett 1979;100:296e300.
than assembled cage particles. Moreover, this profile suggested [8] Gerl M, Jaenicke R. Assembly of apoferritin from horse spleen: comparison of
potentially another proof that the coil-to-helix transition of GALA the protein in its native and reassembled state. Biol Chem Hoppe-Seyler
could destabilize the assembly of caged particles. 1987;368:387e96.
[9] Harrison PM, Arosio P. The ferritins: molecular properties, iron storage func-
At pH 6.0, which corresponded to complete dissociation of our tion and cellular regulation. Biochim Biophys Acta 1996;1275:161e203.
scaffold, any apparent transition between folding and unfolding [10] Aime S, Frullano L, Geninatti Crich S. Compartmentalization of a gadolinium
was not observed (as we expected), because there were not complex in the apoferritin cavity: a route to obtain high relaxivity contrast
agents for magnetic resonance imaging. Angew Chem Int Ed Engl 2002;41
any stable secondary and quaternary structures (Fig. 8B). After (6):1017e9.
unfolding of the subunits and subsequent dissociation of cage [11] Varpness Z, Peters JW, Young M, Douglas T. Biomimetic synthesis of a H2
particles were triggered, dissociated form had the lowered ther- catalyst using a protein cage architecture. Nano Lett 2005;5:2306e9.
[12] Tang TH, Johnson JM, Dryden KA, Young MJ, Zlotnick A, Johnson JE. The role of
mostability than caged scaffold. However, our acid-sensitive scaf- subunit hinges and molecular “switches” in the control of viral capsid poly-
fold was sufficiently stable in a wide range of temperature before morphism. J Struct Biol 2006;154:59e67.
acidity-inducible disassembly was initiated. [13] Larson SB, Lucas RW, McPherson A. Crystallographic structure of the T ¼ 1
particle of brome mosaic virus. J Mol Biol 2005;346:815e31.
[14] Dalmau M, Lim S, Wang SW. Design of a pH-dependent molecular switch in
4. Conclusions a caged protein platform. Nano Lett 2009;9:160e6.
[15] Dalmau M, Lim S, Wang SW. pH-triggered disassembly in a caged protein
complex. Biomacromolecules 2009;10(12):3199e206.
In summary, we demonstrated the engineered cage protein can [16] Bozzi M, Mignogna G, Stefanini S, Barra D, Longhi C, Valenti P, et al. A novel
dissociate completely at pH 6.0 but reversibly re-assemble at pH non-heme iron-binding ferritin related to the DNA-binding proteins of the Dps
family in Listeria innocua. J Biol Chem 1997;272(60):3259e65.
7.0. In this study, ferritin scaffold was genetically modified by our [17] Koeck PJ, Kagawa HK, Ellis MJ, Hebert H, Trent JD. Two-dimensional crystals of
strategy so that artificial GALA peptides were incorporated into reconstituted beta-subunits of the chaperonin TF55 from Sulfolobus shibatae.
caged protein. Coil-to-helix transition of GALA peptide inducible by Biochim Biophys Acta 1998;1429(1):40e4.
[18] Trent JD, Kagawa HK, Yaoi T, Olle E, Zaluzec NJ. Chaperonin filaments: the
acid was suggested to impart acid-sensitive disassembly function archaeal cytoskeleton? Proc Natl Acad Sci U S A 1997;94(10):5383e8.
to cage protein. Our design strategy overcomes the limitation [19] Steinbüchel A, Marchessault RH. Biopolymers for medical and pharmaceutical
that depended on the specific structural motif of viral capsids. The applications, vol. 2. Wiley; 2005. p. 951e972.
[20] Subbarao Jr NK, Parente RA, Szoka Jr FC, Nadasdi L, Pongracz K. pH-dependent
disassembly and re-assembly process of caged protein were
bilayer destabilization by an amphipathic peptide. Biochemistry
sensitive to weak acid like pH 6.0 so that engineered cage protein 1987;26:2964e72.
would be potentially applicable as acid-inducible molecular switch. [21] Li W, Nicol F, Szoka Jr FC. GALA: a designed synthetic pH-responsive amphi-
pathic peptide with applications in drug and gene delivery. Adv Drug Deliv
Rev 2004;56:967e85.
Acknowledgements [22] Haas DH, Murphy RM. Design of a pH-sensitive pore-forming peptide with
improved performance. J Pept Res 2004;63:9e16.
[23] Wang Z, Li C, Eijenburg M, Soistman E, Ruble J, Wright B, et al. Structure of
This research was supported by the Pioneer Research Center human ferritin L chain. Acta Crystallogr 2006;D62:800e6.
Program (2009-0081523) and Global Research Laboratory Project [24] Levi S, Luzzago A, Cesareni G, Cozzi A, Franceschinelli F, Albertini A, et al.
Mechanism of ferritin iron uptake: activity of the H-chain and deletion
of MEST.
mapping of the ferro-oxidase site. A study of iron uptake and ferro-oxidase
activity of human liver, recombinant H-chain ferritins, and of two H-chain
deletion mutants. J Biol Chem 1988;263:18086e92.
Appendix [25] Santambrogio P, Levi S, Arosio P, Palagi L, Vecchio G, Lawson DM, et al.
Evidence that a salt bridge in the light chain contributes to the physical
Figures with essential color discrimination. Many of the figures stability difference between heavy and light human ferritins. J Biol Chem
1992;267:14077e83.
in this article have parts that are difficult to interpret in black and [26] Taillefer J, Brasseur N, van Lier JE, Lenaerts V, Le Garrec D, Leroux JC. In vitro
white. The full color images can be found in the on-line version, at and in vivo evaluation of pH-responsive polymeric micelles in a photody-
doi:10.1016/j.biomaterials.2010.03.016. namic cancer therapy model. J Pharm Pharmacol 2001;53:155e66.
[27] Ojugo AS, McSheehy PM, McIntyre DJ, McCoy C, Stubbs M, Leach MO, et al.
Measurement of the extracellular pH of solid tumours in mice by magnetic
References resonance spectroscopy: a comparison of exogenous (19)F and (31)P probes.
NMR Biomed 1999;12(8):495e504.
[28] Lee ES, Shin HJ, Na K, Bae YH. Poly(l-histidine)ePEG block copolymer micelles
[1] Kramer RM, Li C, Carter DC, Stone MO, Naik RR. Engineered protein cages for
and pH-induced destabilization. J Control Release 2003;90:363e74.
nanomaterial synthesis. J Am Chem Soc 2004;126:13282e6.
[29] Kim GM, Bae YH, Jo WH. pH-induced micelle formation of poly(histidine-co-
[2] Uchida M, Klem MT, Allen M, Suci P, Flenniken M, Gillitzer E, et al. Biological
phenylalanine)-block-poly(ethylene glycol) in aqueous media. Macromol
containers: protein cages as multifunctional nanoplatforms. Adv Mater
Biosci 2005;5:1118e24.
2007;19:1025e42.
[30] Crichton RR, Bryce CFA. Subunit interactions in horse spleen apoferritin.
[3] Flenniken ML, Willits DA, Harmsen AL, Liepold LO, Harmsen AG, Young MJ,
Dissociation by extremes of pH. Biochem J 1973;133:289e99.
et al. Melanoma and lymphocyte cell-specific targeting incorporated into
[31] Yoshizawa K, Mishima Y, Park SY, Heddle JG, Tame JRH, Iwahori K, et al. Effect
a heat shock protein cage architecture. Chem Biol 2006;13:161e70.
of N-terminal residues on the structural stability of recombinant horse L-chain
[4] Johnson JE, Speir JA. Quasi-equivalent viruses: a paradigm for protein
apoferritin in an acidic environment. J Biochem 2007;142:707e13.
assemblies. J Mol Biol 1997;269(5):665e75.
[32] Kilic MA, Spiro S, Moore GR. Stability of a 24-meric homopolymer: compar-
[5] Bancroft JB, Hills GJ, Markham R. Study of the self-assembly process in a small
ative studies of assembly-defective mutants of Rhodobacter capsulatus bac-
spherical virus. Formation of organized structures from protein subunits in
terioferritin and the native protein. Protein Sci 2003;12:1663e74.
vitro. Virology 1967;31(2):354e79.
[33] Santambrogio P, Pinto P, Levi S, Cozzi A, Rovida E, Albertini A, et al. Effects of
[6] Ausar SF, Foubert TR, Hudson MH, Vedvick TS, Middaugh CR. Conformational
modifications near the 2-, 3- and 4-fold symmetry axes on human ferritin
stability and disassembly of Norwalk virus like particles: effect of pH and
renaturation. Biochem J 1997;322:461e8.
temperature. J Biol Chem 2006;281(28):19478e88.