High Yield Refolding and Purification
Process for Recombinant Human
Interleukin-6 Expressed in
Escherichia coli
Daisuke Ejima, Mayumi Watanabe, Yutaka Sato, Masayo Date,
Naoyuki Yamada, Yoshiyuki Takahara
Central Research Laboratories, Ajinomoto Company, Inc.; 1-1 Suzuki-cho,
Kawasaki-ku, Kawasaki 210-8681, Japan; telephone: +81-44-244-7178;
fax: +81-44-245-8584; e-mail: plr ejimad@te10.ajinomoto.co.jp
Received 3 April 1998; accepted 24 July 1998
Abstract: Recombinant human interleukin-6 (hIL-6), a
pleiotropic cytokine containing two intramolecular disul-
fide bonds, was expressed in Escherichia coli as an in-
soluble inclusion body, before being refolded and puri-
fied in high yield providing sufficient qualities for clinical
use. Quantitative reconstitution of the native disulfide
bonds of hIL-6 from the fully denatured E. coli extracts
could be performed by glutathione-assisted oxidation in
a completely denaturating condition (6M guanidinium
chloride) at protein concentrations higher than 1 mg/mL,
preventing aggregation of reduced hIL-6. Oxidation in
6M guanidinium chloride (GdnHCl) required remarkably
low concentrations of glutathione (reduced form, 0.01
mM; oxidized form, 0.002 mM) to be added to the solu-
bilized hIL-6 before the incubation at pH 8.5, and 22°C for
16 h. After completion of refolding by rapid transfer of
oxidized hIL-6 into acetate buffer by gel filtration chro-
matography, residual contaminants including endotoxin
and E. coli proteins were efficiently removed by succes-
sive steps of chromatography. The amount of dimeric
hIL-6s, thought to be purification artifacts, was decreased
by optimizing the salt concentrations of the loading ma-
terials in the ion-exchange chromatography, and gradu-
ally removing organic solvents from the collected frac-
tions of the preparative reverse-phase HPLC. These
refolding and purification processes, which give an over-
all yield as high as 17%, seem to be appropriate for the
commercial scale production of hIL-6 for therapeutic
use. © 1999 John Wiley & Sons, Inc. Biotechnol Bioeng 62:
301–310, 1999.
Keywords: interleukin-6; protein refolding; inclusion
body; aggregation; purification
INTRODUCTION
Interleukin-6 (IL-6) is a cytokine, which is involved in di-
verse biological activities such as proliferation, differentia-
tion, and maturation events in host target cells (Hirano et al.,
1985; Simpson et al., 1997). Clinical use of hIL-6 in cancer
therapy has been of great interest and functional agonists
and antagonists for therapeutic use in IL-6-associated dis-
eases are now in development (Ciapponi et al., 1997;
Correspondence to: Daisuke Ejima
© 1999 John Wiley & Sons, Inc.
Sporeno et al., 1996). Some groups have already reported
hIL-6 production systems in recombinant E. coli, which
can produce large amounts of insoluble inclusion bodies
(Brakenhoff et al., 1987; Rock et al., 1992; Tonouchi et al.,
1988; Yasukawa et al., 1990). However, problems concern-
ing the low yield in the refolding process owing to the
aggregate-prone property of hIL-6 (Simpson et al., 1997;
Ward et al., 1995) remain to be solved. This property also
makes it rather difficult to design a high-yield purification
process applicable to the commercial scale production of
hIL-6, which can efficiently remove residual contaminants
from host cells and degradation or modification impurities
of hIL-6 from the purified product.
Production of heterologous proteins as inclusion bodies
in engineered E. coli is a common technique used to obtain
valuable non-glycosylated proteins in large amounts. A re-
cent study on refolding of denatured lysozyme (Hevehan et
al., 1997) revealed that low concentrations of denaturants
(guanidinium chloride or urea) can suppress the aggregation
of refolding intermediates, even at high-protein concen-
trations and thus, increase the yields of refolded lyso-
zyme recovered. This study emphasized the availability of
non-denaturating concentrations of guanidinium chloride
(GdnHCl) for designing industrial-scale refolding pro-
cesses. However, denaturant-induced equilibrium unfolding
studies on murine IL-6 (Ward et al., 1995; Zhang et al.,
1997) have demonstrated that natively disulfide-bonded
IL-6 forms partially unfolded conformations, which have a
high tendency to self-associate at low denaturant concen-
trations. These studies also showed that fully disulfide-
reduced IL-6 exhibits a higher tendency to self-associate
than that of natively disulfide-bonded IL-6. Taken together,
these results suggest that the refolding of hIL-6 by incubat-
ing disulfide-reduced denatured forms in partially denatur-
ating conditions, as in the case of refolding lysozyme
(Hevehan et al., 1997), will in principle, fall into the low
efficacy category.
In this investigation, we have established a new refolding
process for hIL-6, which consists of two steps. In the first
CCC 0006-3592/99/030301-10
step, high protein concentrations of natively disulfide-
bonded hIL-6 were obtained from solubilized inclusion bod-
ies under fully denaturating buffer conditions. In the second
step, the native conformation of hIL-6 was obtained by
rapidly removing the denaturant from the oxidized hIL-6
solution without any dilution. Aggregation during the re-
folding process was effectively suppressed. Refolded hIL-6
was purified successively by ion-exchange, reverse-phase,
and gel filtration chromatographies with aggregation con-
trolled by unique techniques to give a good recovery and
provide enough high quality material to be evaluated in vivo
free of responses to residual endotoxin and immunogenic
impurities.
MATERIALS AND METHODS
Materials
Trifluoroacetic acid (spectroscopic grade), acetic acid, tris-
hydroximethylaminomethane (Tris), reduced and oxidized
glutathione (GSH and GSSG), ethylenediaminetetraacetic
acid (EDTA), hydrochloric acid, sodium hydroxide, and
Achromobacter protease I (lysylendopeptidase; EC
3.4.21.50, 4.5 AU/mg) were purchased from Wako Pure
Chemical Industries (Osaka, Japan). Guanidinium chloride
(GdnHCl) and HPLC-grade solvents were purchased from
Nakarai tesque (Kyoto, Japan). All buffers were prepared
with deionized water purified by the Milli-Q SP/UF system
(Millipore, Tokyo, Japan). All other chemicals were reagent
grade.
Recovery of Inclusion Bodies
Escherichia coli strain HB101 with an expression plasmid
containing the trp promoter followed by the mature hIL-6
sequence was used as the expression host (Yasueda et al.,
1990). The host cell was grown overnight in a 30 L fer-
mentor and the cultured-broth (20 L) obtained was directly
introduced into a high-pressure homogenizer (Type SHL-5;
Alfa-Laval) for cell disruption. The cell homogenate was
applied to a continuous centrifuge (type NO-U-5-HR;
Kansai Centrifugal Separator Manufacturing, Osaka, Japan)
at 7600g, and 1 L/min to recover the hIL-6 inclusion bodies.
The collected inclusion bodies were suspended in 500 mL
of washing buffer (20 mM Tris-HCl, 30 mM NaCl, pH 7.5),
and then centrifuged at 8000g for 15 min. The pellet of
inclusion bodies obtained was resuspended in 500 mL of 10
mM EDTA, pH 5.5, and stored at −80°C until use.
Solubilization of Inclusion Bodies
and Air-Oxidation
The pellet of inclusion bodies was solubilized in 6M
GdnHCl (13.8 mg/mL), adjusted to pH 5.5 with HCl and
allowed to stand for 2 h at room temperature. Solubilized
hIL-6 (1.5 mL) was rapidly diluted 10-fold to 1.38 mg/mL
with 10 mM Tris-HCl, pH 8.5, containing varying concen-
302 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 3, FEBRUARY 5, 1999
trations of GdnHCl (0.6 ∼ 6M), and then incubated for 16 h
at 22°C with gentle stirring. After incubation, each solution
was centrifuged and 12 ␮L of the supernatant part was
analyzed by reverse-phase HPLC for the yield of soluble
oxidized hIL-6.
Glutathione-Assisted Oxidation
The solubilized hIL-6 from above (13.8 mg/mL, pH 5.5)
was rapidly diluted 10-fold to 1.38 mg/mL in 6M GdnHCl.
Varying concentrations of reduced and oxidized glutathione
were then added to 15 mL of diluted hIL-6, keeping the
reduced/oxidized glutathione ratio at 5/1. Each supple-
mented solution was adjusted to pH 8.5 with 10 mM Tris-
HCl and incubated for 16 h at 22°C with gentle stirring.
Reverse-phase HPLC analysis for the yield of oxidized
hIL-6 was performed in the same way as air-oxidation.
Large-Scale Refolding by Glutathione-Assisted
Oxidation and Desalting Chromatography
The pellet of inclusion bodies containing 11 g of reduced
hIL-6 was solubilized in 12 L of 6M GdnHCl, pH 5.5, and
allowed to stand for 2 h at 22°C. Reduced and oxidized
glutathione were added into the solubilized hIL-6 at the
final concentrations of 0.01 mM and 0.002 mM, respec-
tively. The supplemented solution was adjusted to pH 8.5
with 10 mM Tris-HCl and incubated for 16 h at 22°C. After
the oxidation yield was confirmed by reverse-phase HPLC
analysis, the solution was directly applied to a gel filtration
column (Sephadex G-25 M, 44 cm internal diameter × 30
cm; Amersham Pharmacia Biotech, Tokyo, Japan) equili-
brated with 10 mM sodium acetate, pH 5.0, and the protein
was eluted with the same buffer at a flow-rate of 750 mL/
min. Eluted materials were detected by UV absorbance at
280 nm in all chromatography steps.
Purification of Oxidized hIL-6
Refolded hIL-6 (10 g) in 15 L of 10 mM sodium acetate, pH
5.0 was adjusted to 250 mM sodium acetate, pH 5.0 and
applied to a cation-exchange column (CM-Sepharose FF, 26
cm internal diameter × 10 cm; Amersham Pharmacia Bio-
tech, Tokyo, Japan) equilibrated with 10 mM sodium ac-
etate, pH 5.0. After the column was washed with 1 column
volume of the same buffer, hIL-6 was eluted with a linear
gradient of sodium acetate from 10 mM to 500 mM and a
pH gradient from 5.0 to 5.5 (final elution buffer) over 10
column volumes at a flow-rate of 0.5 L/min. Fractions of
500 mL were collected and assayed by analytical cation-
exchange HPLC by injecting 100 ␮L of each fraction.
Pooled hIL-6 fractions (1.2 g) in the cation-exchange
chromatography were adjusted to pH 4.0 with formic acid
and applied to a preparative reverse-phase HPLC column
(Vydac 214TPB10, 5 cm internal diameter × 25 cm, The
Separations Group, Hesperia, CA, USA) equilibrated with
0.1M sodium formate, pH 4.0. After the column was washed
with 1 column volume of the same buffer, hIL-6 was eluted
with a linear gradient of 0–60% acetonitrile in the elution
buffer (0.1M sodium formate, pH 4.0) over six column vol-
umes at a flow-rate of 50 mL/min. Five fractions were col-
lected, and assayed by the analytical reverse-phase HPLC
by injecting 50 ␮g of hIL-6 from each fraction. Four runs of
this chromatography were performed. Pooled hIL-6 frac-
tions (800 mL) were applied to a gel filtration column
(Sephadex G-25 M, 18 cm internal diameter × 18 cm, Am-
ersham Pharmacia Biotech, Tokyo, Japan) equilibrated with
20 mM acetic acid and 10% acetonitrile, and proteins were
eluted with the same solvent at a flow rate of 120 mL/min.
The hIL-6 (1760 mL) collected was kept at 22°C for 1 h and
then applied to another Sephadex G-25 M column (30 cm
internal diameter × 18 cm) equilibrated with 5 mM sodium
acetate, pH 4.5, and run with the same buffer at a flow-rate
of 350 mL/min.
Prior to the final purification step, hIL-6 was concen-
trated to greater than 8 mg/mL by cation-exchange chroma-
tography. One gram of solvent-free hIL-6 (5 mM sodium
acetate, pH 4.5) was adjusted to 175 mM sodium acetate, pH
5.0 and applied to a CM-Sepharose FF column (5 cm in-
ternal diameter × 2.5 cm) equilibrated with the 10 mM
sodium acetate, pH 5.0. Adsorbed hIL-6 was eluted step-
wise with 10 mM sodium citrate, pH 6.5 containing 50 mM
NaCl at a flow rate of 30 mL/min. Three runs of this chro-
matography were performed. Finally, 50 mL of concen-
trated hIL-6 was applied to a preparative gel filtration
HPLC column (Superdex-75 HR 60/600, 6 cm internal di-
ameter × 60 cm, Amersham Pharmacia Biotech, Tokyo,
Japan) equilibrated with 10 mM sodium citrate, pH 6.0 and
the proteins were eluted with the same buffer at a flow-rate
of 24 mL/min. Monomeric hIL-6 fractions were pooled,
filter sterilized (Millex GVHD-25020, Millipore, Tokyo, Ja-
pan) and stored at 5°C for characterization.
Analytical Methods
Analytical reverse-phase HPLC was performed using a
Vydac C4 column (214TP54, 4.6 mm internal diameter ×
250 mm, The Separations Group, Hesperia, CA, USA). The
column was equilibrated with 32% acetonitrile and 0.1%
trifluoroacetic acid/water, and the proteins were eluted with
a linear gradient of up to 60% acentonitrile and 0.1% tri-
fluoroacetic acid/water over 28 min at a flow rate of 1
mL/min at ambient temperature. The eluted materials were
detected by UV absorbance at 280 nm (all analytical HPLC
of hIL-6).
Analytical cation-exchange HPLC was performed using a
SP-NPR column (35 mm internal diameter × 30 mm, Tosoh
Corporation, Tokyo, Japan) equilibrated with 10 mM so-
dium acetate, pH 5.0, and the proteins were eluted with a
linear gradient of up to 0.5M sodium acetate, pH 5.5 over 5
min at a flow rate of 1 mL/min.
Analytical gel filtration was performed using a Superdex
75 HR 10/30 (10 mm internal diameter × 300 mm, Amer-
sham Pharmacia Biotech, Tokyo, Japan) equilibrated with
EJIMA ET AL.: REFOLDING AND PURIFICATION OF HUMAN INTERLEUKIN-6
10 mM sodium citrate, 8.7 mM sodium phosphate, pH 7.0
and the proteins were eluted with the same buffer at a flow
rate of 0.8 mL/min.
Sodium dodecylsulfate polyacrylamide gel electrophore-
sis (SDS-PAGE) was performed using the Phastsystem and
20% homogeneous separating gels (Amersham Pharmacia
Biotech, Tokyo, Japan) under non-reducing conditions ac-
cording to the manufacturer’s instruction. For GdnHCl-
containing samples, GdnHCl was removed using a small
prepacked desalting column (PD-10, Amersham Pharmacia
Biotech, Tokyo, Japan) equilibrated with 10 mM sodium
acetate, pH 5.0 according to the manufacturer’s instruction.
Gels were visualized by silver staining (diamine method).
Phosphorylase b (94 kDa), bovine serum albumin (67 kDa),
ovalbumin (43 kDa), carbonic anhydrase (30 kDa), soybean
trypsin inhibitor (20 kDa), and alpha-lactalbumin (14.4
kDa) were used as molecular weight markers.
Peptide mapping was performed by enzymatic digestion
(S/E ⳱ 100/1 in molar ratio) and analytical reverse-phase
HPLC. Purified hIL-6 (1 nanomole/21 ␮g in 10 mM sodium
citrate, pH 6.0) was diluted into 0.1 mL of the digestion
buffer (10 mM sodium phosphate, pH 7.0). Ten picomoles
Achromobacter protease I enzyme (Wako Pure Chemical
Industries, Osaka, Japan) dissolved in 0.01 mL of the di-
gestion buffer was added to the hIL-6 solution, and the
mixture was incubated for 12 h at ambient temperature. The
whole solution was then injected directly into a Vydac C4
(214TP54, 4.6 mm internal diameter × 250 mm, The Sepa-
rations Group, Hesperia, CA, USA) equilibrated with 0.1%
trifluoroacetic acid/water, and the peptides were eluted with
a linear gradient of 0–60% acetonitrile with 0.1% trifluoro-
acetic acid/water over 75 min at a flow rate of 1 mL/min.
The eluted peptide fragments were detected by UV absor-
bance at 215 nm. Amino acid sequences of detected pep-
tides were assigned by FABMS (JMS-HX110/HX110,
JEOL, Tokyo, Japan) and automatic Edman degradation
(model 470A, Applied Biosystems, Tokyo, Japan) of the
collected peptide peaks.
The concentration of purified hIL-6 was determined by
UV spectroscopy at 280 nm utilizing an extinction coeffi-
cient for hIL-6 of 0.47 mg−1 cm2 at 280 nm, which was
obtained experimentally using an hIL-6 protein standard
measured by quantitative amino acid analysis. The concen-
trations of crude hIL-6 were determined by analytical re-
verse-phase HPLC (above) using purified hIL-6 as a protein
standard.
Biological Assay
The hIL-6 titer was determined by an IgM-inducing assay
using SKW6-CL4 cells, as previously described (Hirano et
al., 1985).
Endotoxin and Escherichia coli Protein Contents
The amount of endotoxin in the hIL-6 samples was deter-
mined using the Lymulus amoebocyte lysate assay (Toxi-
color system, Seikagaku Kogyo Inc., Tokyo, Japan) accord-
ing to the manufacturer’s instruction.
303
 The E. coli protein (ECP) content of hIL-6 samples was
determined by ELISA using polyclonal anti-ECP antibod-
ies. The recombinant host cell was constructed by trans-
forming E. coli with a vector identical to the production
system but lacking the gene insert coding the product. They
were then cultured in a fermentor (3 L), disrupted, and
centrifuged in the same manner as the production. The pellet
obtained was solubilized, purified by desalting and cation-
exchange chromatographies using the same conditions for
hIL-6 except for the scales to yield the semi-purified ECP
mixtures. Rabbit and mouse polyclonal anti-ECP antibodies
were obtained using these ECP mixtures and a highly sen-
sitive sandwich ELISA system for ECP was constructed
using anti-ECP antibodies, biotin-labeled alkaliphosphatase
kit (AK 5001, Vectastain Vector Lab, Inc., Burlingame, CA,
USA) and BCIP/NBT phosphatase substrate system (Kap-
pel, Funakoshi, Tokyo, Japan). The detection limit of this
ELISA for ECP was 1 ng/mL.

RESULTS

Production and Refolding of hIL-6

Recombinant hIL-6 was produced as insoluble inclusion
bodies in the engineered E. coli strain HB101 containing an
expression plasmid for mature hIL-6 sequence, as previ-
ously reported (Yasueda et al., 1990). After completion of
the fermentation, the cultured broth was directly introduced
into a high pressure homogenizer, and insoluble inclusion
bodies recovered by continuous centrifugation. The pellet
collected was resuspended in Tris-HCl buffer, centrifuged
again, and the washed inclusion bodies finally suspended in
10 mM EDTA, pH 6.0 before being stored at −80°C until
use.
Frozen inclusion bodies were thawed, and then solubi-
lized in 6M GdnHCl, pH 5.5 for 2 h at ambient temperature
without any disulfide reducing agents. Formation of the
disulfide bonds at a concentration of 1.38 mg/mL denatured
hIL-6 under several GdnHCl concentrations was examined
in 15 mL volumes by air-oxidation. Each solution was cen-
trifuged and obtained supernatant part was assayed by ana-
lytical reverse phase HPLC (Figs. 1,2). As shown in Figure
1B, most of extracted hIL-6 had reduced disulfide bonds
before oxidation. It was found that the yields of oxidized
hIL-6 after a 16 h incubation at 22°C were remarkably
dependent on the GdnHCl concentrations used (Fig. 2). As
the GdnHCl concentration increased from 0.6M, the yield of
oxidized hIL-6 first decreased to give the minimum yield at
3M (12%), and then increased to give the maximum yield at
6M (87%). Solutions with GdnHCl below 4M became tur-
bid after incubation for 16 h due to the presence of aggre-
gation. The oxidation yield with 1M GdnHCl (62%) was
distinctly lower than the maximum yield (87% in 6M), but
the purity of the oxidized material as judged by reverse
phase HPLC (Fig. 1F) was much better than that of the
maximum yield (Fig. 2). Partially denaturating GdnHCl
concentrations (2 ∼ 4M), which are commonly used in oxi-
Figure 1. Air oxidation of denatured hIL-6. HIL-6 extracted in 6M
GdnHCl (13.8 mg/mL) was diluted to 1.38 mg/mL in 10 mM Tris-HCl, pH
8.5 containing each concentration of GdnHCl. The solutions (15 mL) were
incubated at room temperature for 16 h, and 12 ␮L of each was assayed by
reverse-phase HPLC. Arrows and a dotted arrow show oxidized and re-
duced hIL-6, respectively.
dative refolding processes of many kinds of denatured pro-
teins (Fischer et al., 1993), were not effective for refolding
of reduced hIL-6.
Applications of thiol/disulfide redox buffer systems (glu-
tathione system) in 6M GdnHCl were examined in the same
15 mL volumes (Fig. 3). Combinations of reduced and oxi-
dized glutathione (GSH and GSSG) were added to the oxi-
dation solutions, keeping the molar ratio of GSH/GSSG at
5/1, and the mixtures incubated for 16 h at 22°C, similar to
the air-oxidation. The highest yield was obtained at 0.01
mM GSH/0.002 mM GSSG (105%, Fig. 3B), and this ma-
terial showed a molecular weight equivalent to that of pu-
rified hIL-6 on SDS-PAGE (Fig. 4, lane 3). The reason why
the yield was above 100% may be due to unresolved impu-
rities in the oxidized hIL-6 peak. The combination of 1 mM
GSH/0.2 mM GSSG, which is one of the popular conditions
used in oxidative refolding of denatured proteins (Fischer et
al., 1993), increased the yield a little (95%, Fig. 3C) as
compared with that of air-oxidation (91%, Fig. 3A). How-
ever, the oxidized hIL-6 obtained under this condition

304 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 3, FEBRUARY 5, 1999

Figure 2. Yield of soluble oxidized hIL-6 in varying concentrations of
GdnHCl (Fig. 1).
showed a contaminating band with a molecular weight a
little higher than that of natively disulfide-bonded hIL-6 on
SDS-PAGE (Fig. 4, lane 4). The amount of this contaminant
band increased (Fig. 4, lane 5) and the oxidation yield de-
creased to 81% (Fig. 3D) with the combination of 5 mM
GSH/1 mM GSSG. Peptide-mapping and Electrospray Ion-
ization mass spectrometry analyses of this contaminant
band revealed that it was probably due to an oxidation ar-
tifact which had one mixed disulfide bond with glutathione
at Cys(44) or Cys(50) (data not shown). Consequently, the
optimal oxidation condition was obtained using fully dena-
turating buffer containing 6M GdnHCl, 0.01 mM GSH, and
0.002 mM GSSG, at pH 8.5.
Large scale oxidation (12 L) was carried out almost quan-
titatively and the oxidized hIL-6 (0.9 mg/mL) was directly
applied to a Sephadex G-25 desalting column equilibrated
with 10 mM sodium acetate buffer, pH 5.0 (Fig. 5). The
fraction (15 L) indicated by the arrow was collected and
analyzed by reverse-phase HPLC (inserted chromatogram).
Almost 95% of hIL-6 loaded was recovered (0.68 mg/mL)
as one major peak with trace impurities as shown by ana-
lytical reverse-phase HPLC, and it had comparable bioac-
tivity to purified hIL-6 in the IgM inducing assay (data not
shown). Therefore, it was concluded that refolding of hIL-6
from inclusion bodies had been complete.
Purification and Characterization of
Refolded hIL-6
Cation-Exchange Chromatography
Refolded hIL-6 (10 g in 10 mM sodium acetate buffer, pH
5.0) was adjusted to 250 mM sodium acetate, pH 5.0 and
applied to the CM-Sepharose FF column equilibrated with
10 mM sodium acetate buffer, pH 5.0. Adsorbed hIL-6 was
EJIMA ET AL.: REFOLDING AND PURIFICATION OF HUMAN INTERLEUKIN-6
Figure 3. Glutathione-assisted oxidation of reduced hIL-6. Oxidation
conditions were similar to those of air-oxidation (Fig. 1) except additions
of reduced and oxidized glutathione (GSH and GSSG). Solutions were
incubated for 16 h at 22°C. The yield of oxidized hIL-6 was assayed by
analytical reverse-phase HPLC, shown inserted in each chromatogram.
GSH/GSSG: A, air-oxidation; B, 0.01 mM/0.002 mM; C, 1.0 mM/0.2 mM;
D, 5 mM/1 mM.
eluted with a linear gradient of up to 0.5 M sodium acetate,
pH 5.5 (Fig. 6). Forty-eight percent of the loaded hIL-6 was
recovered (4.8 g), and it showed a homogeneous peak in the
analytical ion-exchange HPLC (inserted A in Fig. 6) with a
more than 1000-fold decrease in the residual endotoxin con-
tent (Table I). When refolded hIL-6 (dissolved in 10 mM
acetate, pH 5.0) was directly added to the CM-Sepharose FF
column and eluted under the same conditions, dimeric hIL-6
appeared as a purification artifact and the yield of hIL-6
decreased to less than 35% (inserted B in Fig. 6).
Reverse-Phase HPLC
The hIL-6 collected (1200 mg) from ion-exchange chroma-
tography was adjusted to pH 4.0 with formic acid, and
loaded on a 50 mm internal diameter reverse-phase column
equilibrated with 0.1M sodium formate, pH 4.0. Adsorbed
hIL-6 was eluted with a linear gradient of 0–60% acetoni-
trile in 0.1M sodium formate, pH 4.0 (Fig. 7), and then
assayed by analytical reverse-phase HPLC (inserted in Fig.
7). Fraction 3, which showed an almost homogeneous hIL-6
305

Figure 4. SDS-PAGE analysis of glutathione-assisted oxidation of
hIL-6. Oxidized hIL-6 (Fig. 3) were desalted by PD-10 (Materials and
Methods) and 400 pg of each were analyzed using Phastsystem and 20%
homogeneous separating gels (Amersham Pharmacia Biotech) under non-
reducing conditions. Gels were visualized with silver staining (diamine
method). Arrows indicate the mixed disulfide form. Lane 1: Low molecular
weight markers; Lane 2: air-oxidation. Lane 3: GSH/GSSG, 0.01 mM/
0.002 mM; Lane 4: GSH/GSSG, 1 mM/0.2 mM; Lane 5: GSH/GSSG, 5
mM/1 mM; Lane 6: Purified hIL-6.
peak, was pooled and stored below 5°C. Product-related
impurities were effectively removed to side streams (frac-
tions 1, 2, 4, and 5) and residual endotoxin content in the
pooled fractions was below the detection limit (<0.006 EU/
mL). Four runs were performed with an average yield of
75% resulting in recovery of 3.6 g of hIL-6 from the 4.8 g
loaded (cation-exchange chromatography).
Figure 5. Desalting chromatography of large-scale oxidation. Oxidized
hIL-6 (12L) was directly introduced to a Sephadex G-25M column (44 cm
internal diameter × 30 cm) equilibrated with 10 mM sodium acetate buffer,
pH 5.0 and eluted with the same buffer. Fraction (15L) indicated by the
horizontal bar was collected and analyzed by reverse-phase HPLC (in-
serted chromatogram).
306 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 3, FEBRUARY 5, 1999
Removal of Solvents
The removal of acetonitrile and sodium formate used in
reverse-phase HPLC was performed by a two-step gel fil-
tration chromatography. HPLC fractions containing 3.6 g
hIL-6 were first applied on a Sephadex G-25 column equili-
brated with 20 mM acetic acid and 10% acetonitrile (Fig.
8A). HIL-6 (3.2 g) was eluted as a non-symmetrical peak.
The hIL-6 collected was kept at 22°C for 1 h, and then
introduced to the second Sephadex G-25 column equili-
brated with 5 mM sodium acetate buffer, pH 4.5 (Fig. 8B).
The amount of hIL-6 was 2900 mg, and it almost all showed
the monomeric form on analytical gel filtration HPLC (Fig.
8C). When reverse-phase HPLC fractions were directly
added to the second Sephadex G-25 column using 5 mM
sodium acetate buffer, almost half of the collected hIL-6
turned to be dimers (inserted in Fig. 8C).
Gel Filtration HPLC
Before being applied to the final purification step, hIL-6
was concentrated to greater than 8 mg/mL by cation-
exchange chromatography. Fifty milliliters of concentrated
hIL-6 (almost 400 mg/50 mL) were loaded on to the pre-
parative gel filtration HPLC column (Superdex-75 HR 60/
600) equilibrated with 10 mM sodium citrate, pH 6.0 as a
final purification step (Fig. 9). Monomeric hIL-6 (fraction
4) was separated completely from the hIL-6 dimers, which
were mainly formed in the two steps of the gel filtration for
removing HPLC solvents and the following concentration
steps. The hIL-6 fragment [Pro(1) − Asp(140)] formed by
acid-hydrolysis between Asp(140) and Pro(141) during the
storage of reverse phase HPLC fractions was eluted be-
tween the hIL-6 dimers and hIL-6 in fraction 2, suggesting
that the cleaved hIL-6 had formed dimers (data not shown).
Contaminating E. coli proteins (ECP) were eluted in the
fractions just before monomeric hIL-6 (inserted values in
Fig. 9). Six runs were performed, and a total of 1870 mg of
purified monomeric hIL-6 was recovered. All of these pu-
rifications are summarized in Table I.
Characterizations of Purified hIL-6
The homogeneity of the purified hIL-6 was confirmed by
high loaded reverse-phase HPLC (Fig. 10A) and SDS-
PAGE (Fig. 10B). Its primary structure was confirmed by
peptide mapping using Achromobacter protease I as a di-
gestion enzyme (Fig. 11). Purified hIL-6 containing very
low amounts of endotoxin and ECP (<0.006 EU/mg hIL-6
and 3ng/mg hIL-6, respectively; Table I) showed a biologi-
cal activity of 5 U/ng hIL-6 in the IgM inducing assay.
DISCUSSION
Refolding of hIL-6
Low recovery yields in protein refolding processes are often
due to aggregation of unfolded or partially folded proteins
Table I. Summary of hIL-6 refolding and purification. Purity was determined by analytical reverse-phase HPLC. All analytical conditions are described
in Materials and Methods.

Purification step HIL-6 (g) Yield (%) Purity (%) Endotoxin (EU/mL) ECP (␮g/g hIL-6)

Solubilization 11 100 ∼60 >100000 N.D.*

Oxidation 10.8 98 ∼60 N.D. N.D.
Removal of GdnHCl 10.2 93 90 420 N.D.
Cation-exchange chromatography 4.8 44 97 0.25 3100
Reverse-phase HPLC 3.6 33 >99 <0.006 110
Removal of solvents 2.9 26 >99 <0.006 88
Concentration 2.4 22 >99 <0.006 94
Gel filtration HPLC 1.87 17 >99 <0.006 3

N.D. ⳱ not determined.

(Goldberg et al., 1991; Thatcher and Hitchcock, 1994). It is
known to be very important to suppress aggregation, espe-
cially in the early phase of the refolding process, to generate
high yields. In the preliminary refolding experiments of
hIL-6, we tried to recover the natively oxidized form by
rapidly diluting the denatured, reduced form into partially
denaturating buffers containing glutathione systems. How-
ever, aggregation of the materials seemed to take place dur-
ing the dilution and following incubation in all cases, and
the yields were very low (data not shown). Therefore, in this
study we first investigated the effects of a wide range of
denaturant concentrations (up to a completely denaturating
condition) on the air-oxidation of reduced hIL-6. Almost all
of the Cys residues in crude hIL-6 extracted from inclusion
bodies have free SH (Fig. 1B), and these reduced hIL-6 tend
to aggregate to give low yields on oxidative refolding using
the non-denaturating concentrations of GdnHCl (Figure 1E
∼ G). It was suggested that non-denaturating concentrations
of GdnHCl (2M ∼ 4M), which are known to be effective for
suppressing aggregation and promoting oxidative refolding
of many proteins (Fischer et al., 1993; Hevehan and Clark,
1997), may produce particular conformation in reduced
hIL-6. These aggregation properties with varying GdnHCl

Figure 6. Cation-exchange chromatography of hIL-6. Refolded hIL-6

(10 g) in 10 mM sodium acetate, pH 5.0 was adjusted to 250 mM sodium
acetate, pH 5.0 and applied to a cation-exchange column (CM-Sepharose
FF, 26 cm internal diameter × 10 cm) equilibrated with 10 mM sodium
acetate, pH 5.0. HIL-6 was eluted with a linear gradient to the elution
buffer (500 mM sodium acetate buffer, pH 5.5) over 10 column volume at
a flow rate of 0.5 L/min. The horizontal bar shows pooled hIL-6 fractions.
Inserted Figure A, cation-exchange HPLC analysis of pooled hIL-6 frac-
tions; inserted Figure B, cation-exchange chromatography when refolded
hIL-6 was directly applied to the column without addition of sodium ac-
etate.
Figure 7. Preparative reverse-phase HPLC of hIL-6. Twelve hundred
milligrams of hIL-6 pooled in ion-exchange chromatography were adjusted
to pH 4.0 with formic acid, applied to a 50 mm internal diameter reverse-
phase column equilibrated with 0.1M sodium formate, pH 4.0. HIL-6 was
eluted with a linear gradient to 0.1M sodium formate, pH 4.0 containing
60% acetonitrile. Five fractions (horizontal bar) were collected and assayed
by the analytical reverse-phase HPLC (inserted figure).

EJIMA ET AL.: REFOLDING AND PURIFICATION OF HUMAN INTERLEUKIN-6 307


Figure 8. Removal of solvents used in reverse-phase HPLC. HPLC frac-
tions containing hIL-6 (Fig. 7) were combined and introduced to the first
Sephadex G-25 column equilibrated with 20 mM acetic acid and 10%
acetonitrile (A). Collected hIL-6 (horizontal bar) was kept at 22°C for 1 h
and introduced to the second Sephadex G-25 column equilibrated with 5
mM sodium acetate buffer, pH 4.5 (B). Recovered hIL-6 (bar) was ana-
lyzed by analytical gel filtration HPLC (C). In Figure 8C, monomeric and
dimeric hIL-6 peaks were indicated by arrow and dotted arrow, respec-
tively. Inserted figure of C shows solvent-free hIL-6 obtained by one-step
gel filtration of HPLC fractions to 5 mM sodium acetate buffer.
concentrations (Fig. 2) seem to be consistent with that of the
unfolding intermediates of murine IL-6 reported by Ward et
al. (1995) and Zhang et al. (1997). With equilibrium dena-
turation studies, they observed that unfolding intermediates
are formed under partially denaturating conditions and these
tend to self-associate or aggregate. They also found that
unfolding intermediates derived from disulfide-reduced
form have a higher tendency to self-associate than that of
the oxidized form of murine IL-6. Intermediates prone to
self-association or aggregation may be a common feature of
IL-6s from some species. As shown in Figure 2, reduced
hIL-6 can be efficiently oxidized under completely denatur-
ating concentrations (6M) of GdnHCl. Therefore, it was
concluded that reduced hIL-6 should first be oxidized in 6M
GdnHCl, and second, refolded in aqueous solutions. One
disulfide bond of native hIL-6 exists between Cys(44) and
Cys(50), and the other between Cys(73) and Cys(83). These
bonds are formed by the nearest Cys pairs in the primary
sequence of hIL-6. Under completely denaturating condi-
tions, peptide chains tend to form disulfide bonds between
the two cysteines with less amino acid residues separating
them in the primary structure than other Cys pairs (Tatsumi
et al., 1994). This is thought to be one of the reasons why
the two native disulfide bonds of hIL-6 can be formed ef-
ficiently (Fig. 1I) at protein concentrations higher than 1
mg/mL in 6M GdnHCl and quantitatively (Fig. 3B) with the
308 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 3, FEBRUARY 5, 1999
Figure 9. Preparative gel filtration HPLC of hIL-6. Fifty milliliters of
concentrated hIL 6 was applied to a preparative gel filtration HPLC column
(Superdex-75 HR 60/600, 6 cm internal diameter × 60 cm) equilibrated
with 10 mM sodium citrate, pH 6.0 and developed with the same buffer at
a flow rate of 24 mL/min. Escherichia coli protein (ECP) contents of
collected fractions (horizontal bar) were assayed by ELISA (Materials and
Methods).
assistance of the redox-system. Application of hIL-6 oxida-
tion protocol to other particular proteins, which have similar
disulfide bond situations to those of hIL-6, may be possible.
Oxidized, but denatured, hIL-6 was refolded by rapid de-
salting chromatography from 6M GdnHCl into acetate
buffer, pH 5.0 (Fig. 5). Ninety-five percent of the loaded
hIL-6 was recovered without any substantial dilution, and
the hIL-6 collected showed the bioactivity equal to that of
authentic hIL-6. Additionally, most of process-related im-
purities were removed to the side streams in this chroma-
tography, with the purity of collected hIL-6 increasing to
more than 90% (reverse-phase HPLC). Breton et al. (1995)
reported that bioactive N-terminally truncated hIL-6 con-
taining one disulfide bond (deleting one bond between
Cys(44) and Cys(50) in intact hIL-6) could be obtained by
extraction of denatured hIL-6 from inclusion bodies pro-
duced in E. coli with 6M GdnHCl overnight, followed by
refolding using 10-fold rapid dilution with aqueous Tris-
HCl buffer. They obtained a lower yield of refolded hIL-6
using 7M urea as an extraction medium than using 6M
GdnHCl, and they ascribed this discrepancy in yield to the
different denaturating abilities between 6M GdnHCl and 7M
urea. However, they also observed that 6M GdnHCl and 7M
urea gave the same yield when they conducted similar de-
naturation and refolding experiments using the oxidized,
purified forms as starting materials. They did not discuss the
results of oxidation kinetics or refolding studies of reduced,
purified hIL-6. We demonstrated here that natively oxidized
hIL-6 can be refolded quantitatively by rapid desalting. We
also confirmed that yield in the air-oxidation of reduced
hIL-6 using 7M urea was lower than using 6M GdnHCl

Figure 10. Characterizations of purified hIL-6. (A) Fifty micrograms of
purified hIL-6 was assayed by analytical reverse-phase HPLC. Inserted,
amplified chromatogram. (B) One microgram of purified hIL-6 was as-
sayed by SDS-PAGE (silver staining). Lane 1, molecular weight marker;
Lane 2, hIL-6. All analytical conditions are described in Materials and
Methods.
(data not shown). Therefore, we supposed that the discrep-
ancy Breton reported may be caused by the different oxi-
dation rates of truncated hIL-6 extracted in 6M GdnHCl and
7M urea, as well as by the different denaturating strength of
two reagents.
Purification of Refolded hIL-6
Refolded hIL-6 was purified sufficiently to homogeneity by
successive chromatography steps (ion-exchange, reverse-
phase, and gel filtration HPLC) to be evaluated for its func-
tions in in vivo experiments (Table I). These refolding and
purification processes, with an overall yield as high as 17%,
make it possible for them to be used in the industrial scale
production of hIL-6 for therapeutic use. Refolded hIL-6
exhibits self-association characteristics during the purifica-
tion. Zhang et al. (1997) suggested that partially unfolded
EJIMA ET AL.: REFOLDING AND PURIFICATION OF HUMAN INTERLEUKIN-6
Figure 11. Peptide mapping of purified hIL-6. HIL-6 peptide fragments
obtained by Achromobacter protease I digestion were assayed by analytical
reverse-phase HPLC. Amino acid sequences of 11 HPLC peaks identified
by Edman degradation and FABMS are indicated by arrows. All analytical
conditions are described in Materials and Methods.
forms of murine IL-6 may accumulate and aggregate at high
concentrations even in the absence of denaturant. In ion-
exchange chromatography, salts (250 mM sodium acetate)
must be added to the material before loading to prevent the
increase of dimers (Fig. 6). The added salts may protect
hIL-6 retained on the ion-exchange ligand from changing its
conformation to partially unfolded forms. However, yield of
this step is less than 50% indicating the optimization of this
step remains. To increase the yield, changing the stationary
phase to other ion-exchangers, which can reduce undesir-
able interactions with hIL-6, should be considered as well as
optimization of other chromatographic parameters. Aceto-
nitrile used in the preparative reverse-phase HPLC must be
removed gradually in two steps to minimize hIL-6 dimer-
ization (Fig. 8). In other studies we have found that some
dimeric forms of hIL-6, which are distinguished from each
other by analytical ion-exchange HPLC, are produced by
the addition of acetonitrile to aqueous hIL-6 solution (un-
published data). Characterizations of hIL-6 dimers pro-
duced by the addition of acetonitrile are now under inves-
tigation. Ward et al. (1996) already reported dimeric hIL-6s
as purification artifacts and suggested their potential as
strong hIL-6 receptor antagonists. However, they have not
reported further on the characteristics of their dimeric hIL-6.
Previously, some neurotrophic factors were reported to be
able to change their forms (from homodimers to het-
erodimers between related molecules) in aqueous solutions
upon the addition of acetonitrile (Radziejewski et al., 1993).
Acetonitrile may have ability to change the conformation
of some cytokines and promote them to dissociate and as-
sociate.
309
 We are indebted to W. Nakamatsu and his team for the E. coli
fermentation. We also thank S. Takahashi for his stimulating
discussion on the thermodynamic phenomena. We also thank A.
Okano for the IgM-inducing assay.
References
Brakenhoff JPJ, De Groot ER, Evers RF, Pannekoek H, Aarden LA. 1987.
Molecular cloning and expression of hybridoma growth factor in Esch-
erichia coli. J Immunol 139:4116–4121.
Breton J, Fiura AL, Bertolero F, Orsini G, Valsasina B, Ziliotto R, Filippis
VD, Laureto PPD, Fontana A. 1995. Structure, stability and biological
properties of a N-terminally truncated form of recombinant human
interleukin-6 containing a single disulfide bond. Eur J Biochem 227:
573–581.
Ciapponi L, Miaone D, Scoumanne A, Costa P, Hanson MB, Svenson M,
Bendtzen K, Alonzi T, Paonessa G, Cortese R, Ciliberto G, Savino R.
1997. Induction of interleukin-6 (IL-6) autoantibodies through vacci-
nation with an engineered IL-6 receptor antagonist. Nature Biotechnol
15:997–1001.
Fischer B, Sumner I, Goodenough P. 1993. Isolation, renaturation, and
formation of disulfide bonds of eukaryotic proteins expressed in Esch-
erichia coli as inclusion bodies. Biotechnol Bioeng 41:3–13.
Goldberg ME, Rudolph R, Jaenicke R. 1991. A kinetic study of the com-
petition between renaturation and aggregation during the refolding of
denatured-reduced egg white lysozyme. Biochem 30:2790–2797.
Hevehan D, Clark EDB. 1997. Oxidative renaturation of lysozyme at high
concentrations. Biotechnol Bioeng 54:221–230.
Hirano T, Taga T, Nakano N, Yasukawa K, Kashamura S, Shimizu K,
Nakajima K, Pyun KH, Kishimoto T. 1985. Purification to homoge-
neity and characterization of human B-cell differentiation factor
(BCDF or BSFp-2). Proc Natl Acad Sci USA 82:5490–5494.
Radziejewski C, Robinson RC. 1993. Heterodimers of the neurotrophic
factors: Formation, isolation, and differential stability. Biochem 32:
13350–13356.
310 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 62, NO. 3, FEBRUARY 5, 1999
Rock F, Everett M, Klein M. 1992. Overexpression and structure-function
analysis of bioengineered IL-2/IL-6 chimeric lymphokine. Prot Eng
5:583–591.
Simpson RJ, Hammacher A, Smith DK, Matthews JM, Ward LD. 1997.
Interleukin-6: Structure-function relationship. Prot Sci 6:929–955.
Sporeno E, Savino R, Ciapponi L, Paonessa G, Cabibbo A, Lahn A. 1996.
Human IL-6 receptor super-antagonists with high potency and wide
spectrum on multiple myeloma cells. Blood 87:4510–4519.
Tatsumi E, Takahashi N, Hirose M. 1994. Denatured state of ovalbumin in
high concentrations of urea as evaluated by disulfide rearrangement
analysis. J Biol Chem 269:28062–28067.
Thatcher DR, Hitchcock A. 1994. Protein folding in biotechnology. In: RH
Pain, editor. Mechanism of protein folding. New York: Oxford Uni-
versity Press. p 229–261.
Tonouchi N, Oouchi N, Kashima N, Kawai M, Nagase K, Okano A, Matsui
H, Yamada K, Hirano T, Kishimoto T. 1988. High-level expression of
human BSF-2/IL-6 cDNA in Escherichia coli using a new type of
expression-preparation system. J Biochem 104:30–34.
Ward LD, Hammacher A, Howlett GJ, Matthews JM, Fabli L, Morits RL,
Nice EC, Weinstock J, Simpson RJ. 1996. Influence of interleukin-6
(IL-6) dimerization on formation of the high affinity hexameric IL-6
receptor complex. J Biol Chem 271:20138–20144.
Ward LD, Matthews JM, Zhang JG, Simpson RJ. 1995. Equilibrium de-
naturation of recombinant murine interleukin-6: Effect of pH, dena-
turants and salt on formation of folding intermediates. Biochem 34:
11652–11659.
Yasueda H, Nagase K, Hosoda A, Akiyama Y, Yamada K. 1990. High-
level direct expression of semi-synthetic human interleukin-6 in Esch-
erichia coli and production of N-terminus Met-free product. Bio/
Technol 8:1036–1040.
Yasukawa K, Saito T. 1990. A new system for the expression of human
interleukin-6 in Escherichia coli. Biotechnol Lett 12:419–424.
Zhang JG, Matthews JM, Ward LD, Simpson RJ. 1997. Disruption of the
disulfide bonds of recombinant murine interleukin-6 induces forma-
tion of a partially unfolded state. Biochem 36:2380–2389.