вход по аккаунту


Improving Renaturation of Proteins from Inclusion Bodies.

код для вставкиСкачать
Dev. Chem. Eng. Mineral Process., I0(5/6),pp.581-599, 2002.
Improving Renaturation of Proteins from
Inclusion Bodies
Nicholas Kotlarski’, Brian K. O’NeillZ*,Eric H. Dunlop’
and Geoffrey L. F r a n ~ i s ” ~
Alpharma Animal Health Pty Ltd, 45 Poplar Road, Parkville, Victoria
3052, Australia
Cooperative Research Centrefor Tissue Growth and Repair,
Dept of Chemical Engineering, Adelaide University, South Australia
5005, Australia
GroPep Pty Ltd, P.0.Box 10065, Adelaide 5000, South Australia
The effect of vaving mixing rate and mode of dilution during refolding of proteins is
poorly characterized However, these two physical parameters are important design
considerations for large-scale processes. In this paper, an investigation of the
refolding of an analogue of Insulin-like Growth Factor I (LongkIGF-I; Wells et al.,
1994) is presented Partially puriJied inclusion bodies of the protein were isolated
from Escherichia coli (E. coli), dissolved following a standardized protocol and
refolded in either batch mode under different mixing rates or fed-batch mode at a
protein concentration of up to 2 g L-’. Rate constants for a widely accepted kinetic
model of refolding were regressed to transient protein concentration as a key
measure of performance. It was found that mixing rate did not significantly alter a
jirst-order rate constant of correct resolding, kN = 0.17 f 0.05 m i d . However, it did
increase an apparent second-order rate constant of aggregation, KA, @om
approximately 1.6 (without stirrin@ to 3.3 L g-’min-’ (rapidly stirred). Refolding at
high protein concentration in fed-batch mode gave approximately 5gold greater
maximal yield of renatured protein than in batch mode. This increase in yield of
active monomer was accompanied by a reduction in the amount of aggregate species
resolved by non-reducing SDS-PAGE. Results indicate that gradual addition of
denatured protein to give conditions that promote refolding is superior to rapid batch
dilution, but an increased rate of mixing reduces the yield of active product by
increasing the rate of aggregation. The kinetic model was shown to give a poor
description of refolding at high protein concentration in both batch and fed-batch
* Author for correspondence (
N. Kotlarski, B.K. O’Neill, E.H.Dunlop and G.L. Francis
A range of cell types are routinely used as hosts for expression of recombinant
protein, but the most common host organism is E. coli (Georgiou, 1988). High-level
expression of heterologous proteins within E. coli generally results in product
accumulation as insoluble inclusion bodies (Marston, 1986; Rudolph and Lilie, 1996).
While several in vivo strategies for avoiding formation of inclusion bodies have been
developed (e.g. Georgiou and Valax, 1996; Mitraki et al., 1991) they have not
provided an ideal alternative (Rudolph and Lilie, 1996). Accordingly, production by
expression as inclusion bodies is more common than soluble expression (Hevehan and
De Bernardez Clark, 1997).
Proteins sequestered within inclusion bodies require renaturation to impart
biological activity. The most common initial step in renaturation is complete
dissolution with strong chemical denaturants (Chaudhuri, 1994; Marston and Hartley,
1990; Rudolph and Lilie, 1996). Highly concentrated solutions of urea are the most
frequently used strong chemical denaturant (Chaudhuri, 1994).
Refolding is the second step in renaturation. The most common method for
refolding denatured protein is direct dilution in batch mode into a solution of
appropriate chemical composition (Fischer et al., 1993; Galliher, 199 1; Rudolph,
1990). However, in common with most methods, product loss due to aggregation is a
major problem (Chaudhuri et al., 1996; Fischer et al., 1993; Georgiou, 1988). There
are many publications reporting extensive empirical optimisation of chemical
additives to reduce aggregation of particular proteins. Despite these reports, dilution
to achieve an overall protein concentration in the micromolar range to avoid
aggregation is often recommended for the general case (Cleland and Wang, 1993;
Fischer et al., 1992; Rudolph, 1990). On a large-scale, such conditions are
unfavourable because they necessitate large processing volumes and incur significant
costs for equipment and consumables (Chaudhuri, 1994; Marston and Hartley, 1990).
Production of useful therapeutic proteins by E. coli, has placed increased emphasis
on the need to improve in vitro protein refolding efficiency. The key to increasing
refolding efficiency is to reduce aggregation at high overall protein concentration
(Goldberg et al., 1991). The use of reverse-micelles, two-phase mixtures, affinity
ligands, concurrent chromatographic separation, and rational protein engineering has
enabled achievement of high yields at high protein concentration. However, none of
these methods have been developed to the stage where they are generally applicable.
Conducting refolding by a series of dilutions, gradual fed-batch addition, and
continuous dilution reduce aggregation relative to single batch dilution without the
need to increase the use of expensive additives (Ambrosius and Rudolph, 1997; D o h
et al., 1995; Fischer et al., 1992; Galliher, 1991; Rudolph, 1990; Rudolph and Fisher,
1990). Recent economic design analyses of overall processes employing refolding in
different modes have identified fed-batch and continuous operation as superior to
batch mode (Chaudhuri et al., 1996; Kotlarski et al., 1997; Middelberg, 1996).
However, despite the simplicity and the observed benefits of alternative operating
modes, batch refolding remains the norm.
Since aggregation of protein molecules occurs during the transition fiom unfolded
to native conformation (Chaudhuri et al., 1996; Cleland and Wang, 1993; Mitraki et
al., 1991; ZettlmeiBl et al., 1979), it is likely that the mass-transfer required for
Improving Renaturation of Proteins from Inclusion Bodies
changing the chemical environment may be a critical factor. In support of this
reasoning, the rate of solvent exchange from denaturing to native-favouring conditions
affected using physically distinct mechanisms has been demonstrated to effect the
yield of correctly refolded protein (Cleland and Wang, 1993). However, it has also
been shown to have a negligible effect if the chemical environment for refolding has
been optimized (Hejnaes et al., 1992). Rudolph and Lilie (1996) summarized that this
behaviour is protein-specific and depends on the relative stability of transient folding
intermediates. In cases where unstable folding intermediates are formed, the mixing
rate will influence the rate of mass-transfer irrespective of the operating mode. The
importance of adequate mixing during refolding by batch dilution has been noted by
several groups (Goldberg et al., 1991; Hevehan and De Bernardez-Clark, 1997;
Jaenicke and Rudolph, 1986). However, no experimental investigation examining
varied mixing rate during refolding by dilution in any mode has been reported.
The experimental work presented herein addresses two issues that are likely to be
important in large-scale refolding: the effect of mixing rate during batch dilution and
operation in the fed-batch mode. The renaturation system that was investigated
employed an 8 M urea solution under reducing conditions to denature LongR31GF-I
inclusion bodies. Refolding was conducted at a urea concentration of 2 M under
oxidising conditions. Reaction rate constants of correct refolding and aggregation and
the overall yield provided measures of performance. A series of refolding
experiments were conducted in batch mode under three different mixing rates and
compared statistically. Refolding in fed-batch mode was conducted repeatedly on a
small scale (< 1 g of fusion protein) and compared to control experiments conducted
using the batch mode.
Materials and Methods
The general chemicals and reagents used in all experiments were of laboratory grade
or higher. Crystalline extra pure grade urea (Merck, Darmstadt, Germany) and
MilliQTM reagent grade water (Millipore, Bedford, MA, USA) were used throughout.
Receptor-grade LongR31GF-I used as a control standard, was kindly provided by
GroPep Pty Ltd (Adelaide, SA, Australia). Unless noted, all operations were
conducted at room temperature. The error introduced due to pipetting in all
measurements was determined to be less than f 2%.
(i) Preparation of Inclusion Bodies
E. coli JMlOl containing the vector for expression of LongR31GF-I (Francis et al.,
1992) was cultured in a minimal glucose medium using a 20 L Chemap CF-2000
fermenter (Chemap AG, Voketswil, Switzerland). Product expression was induced at
an optical density (600 nm, ODm) of approximately seven by the addition of IPTG
(to 0.3 mM) and terminated 5.5 hours later by cooling the broth to below 10.0"C. The
final ODm of the broth was approximately 15. Cells were disrupted with five discrete
passes through a 15MR APV-Gaulin high-pressure homogenizer (CD valve) at a
maximum pressure of 55.2 MPa. 10 L of the homogenate was diluted to 20 L with
phosphate buffered saline (PBS) (1.17 g L-' NaCl, 1.57 g L-' KH2P04, 2.62 g L-'
N. Kotlarski, B.K. 0 'Neill, E.H. Dunlop and G.L. Francis
Na2HP04pH 6.9 f 0.1). The inclusion bodies were recovered using a Veronesi KLE160 solid-bowl disc-stack centrifuge operating at 8400 rpm with a feed rate of 0.4
L min-' (Veronesi Separatori, Bologna, Italy). The solids were recovered from the
bowl, resuspended in 20 L of PBS and centrihged two additional times. The
recovered solids were mixed thoroughly to give a homogenous paste and frozen in
aliquots at -20°C.
(ii) Dissolution of Inclusion Bodies
Thawed inclusion body paste was first diluted with water to 1.00 mL g-' and
thoroughly mixed. Dissolution was initiated by addition of 10 mL of freshly prepared
dissolution buffer per gram of paste to give a dissolution environment of 8 M urea, 0.1
M Tris-HC1 pH 9.1, 0.04 M Gly, 0.01 M EDTA, 0.016 M DTT. Thorough mixing
was maintained using a magnetic stirrer. Total dissolution was achieved rapidly with
the total protein and LongR'IGF-I concentrations remaining essentially constant
(P>0.95) from 5 to 60 minutes (results not shown).
(iii) High Performance Liquid Chromatography (HPLC)
Quantification of denatured and correctly folded LongR31GF-I was determined by
HPLC. Prepacked reverse-phase HPLC columns were obtained from Applied
Biosystems Inc., Foster City, CAYUSA (BrownleeTM, aquapore butyl, 7 pn particle
size, 300 A pore size, 100 x 2.1 mm). Stock feed solutions of 0.1% (v/v) TFA and
80% (v/v) acetonitrile, 0.08% (v/v) TFA were used to generate linear elution gradients
of 25 to 50% acetonitrile over 30 minute at a constant flow rate of 0.5 mLmin-'.
Elutions were monitored by UV absorbance at 214 nm. Frequent standard
LongR'IGF-I and protein-free samples were analysed to provide accurate extinction
coefficients and baselines. All samples were acidified (PH 5 2) using aqueous TFA to
prevent rearrangement of disulfide bonds (Weissman and Kim, 1991) and centrifuged
at 18 x lo3 g, for 3 minutes prior to loading. Samples were analysed within 48
hours and those that had to be stored for more than 12 hours were refrigerated at 4°C.
The applicability of this analytical technique to resolve LongR31GF-I in its native
conformation from mis- and un-folded isomers has been well established (Francis et
al., 1992; Milner et al., 1995). In this investigation, PeakFitTMv4.0 was used to
resolve the peak corresponding to LongR'IGF-I in its native-conformation (Jandel
Scientific, San Rafael, CAYUSA). Accuracy of the method to quantify nativeconformation LongR'IGF-I in heterogeneous refolding mixtures was validated using
receptor-grade LongR31GF-I in a range of spiking studies (results not shown).
(iv) Electrophoresis
Non-reduced samples were prepared by addition of at least an equal volume of double
strength sample buffer (0.9 M Tris-HC1 pH 8.45,24 % (v/v) glycerol, 8% (w/v) SDS,
1.5% (w/v) Coomassie Brilliant Blue, 0.5% (w/v) phenol red). To achieve reducing
conditions, 1.6 M DTT was added to a final concentration of 50 mM. All samples
were heated at 85OC for a minimum of two minutes and centrihged for three minutes
at 18 x 10' g, before loading.
Improving Renaturation of Proteinsfrom Inclusion Bodies
Precast 10-20% polyacrylamide gels were purchased from Novel Experimental
Technology, San Diego, CA, USA (8Ox8Ox1 mm, 10 well). TricineISDS running
buffer with a composition of 0.1 M Tns, 0.1 M tricine, 0.1% (w/v) SDS was used.
Novex M ~ k l wide
2 ~ range
protein standards were used as molecular-weight
markers (Novel Experimental Technology, San Diego, CA, USA).
Before staining, gels were incubated for 30 minutes in 2.5% (v/v) gluteraldehyde.
A solution of 0.1% (w/v) Coomassie R 350 (Pharmacia Biotech AB, Uppsala,
Sweden), 30% (v/v) methanol, 10% (v/v) acetic acid was used for staining.
Destaining was achieved using several changes of fresh destain solution (30% (v/v)
methanol, 10% (v/v) acetic acid).
(v) Total Protein Quantitation
A colorimetric assay was used to estimate the total protein concentration in refolding
mixtures and starting solutions. Protein dye concentrate was purchased from Bio-Rad
(Bio-Rad Laboratories, Hercules, CA, USA). At least a 1:lOO dilution of all samples
with 0.1M NaCI, 0.1% TFA was performed to ensure that the concentrations of
buffering chemicals and urea would not interfere with the assay, the chaotrope would
be diluted, and the protein concentration would be within the linear range of the assay.
200 pL of dye concentrate was well mixed with 800 pL of sample and incubated at
room temperature. The absorbance at 595 nm was measured after 15 minutes. The
total protein concentration was estimated relative to standard solutions of bovine
serum albumin (BSA). Samples for total soluble protein quantitation were quenched
in triplicate and analysed in duplicate.
Figure 1. Macroscopic kinetic scheme for protein refolding defining pathway and
rate constants.
(vi) Estimates of Rate Constants and Their Variance
Reaction rate constants for formation of active LongR31GF-I and aggregation were
used as measures of performance. A first-order equilibrium refolding reaction
competing with second-order aggregation was assumed (see Figure 1). This
mechanism has been used previously in simulation studies (Kotlarski et al., 1997;
Middelberg, 1996) and is mathematically consistent with those used by others
(Chaudhuri et al., 1996; Cleland and Wang, 1990; Kiefhaber et al., 1991; Matthiesen
et al., 1996; Zettlmeial et al., 1979). Reaction rate constants and their variances were
estimated by regression of experimental refolding data to this model using BMDP
Release 7 (BMDP Statistical Software, Inc., Los Angeles, CA, USA). Outlying data
points with standardized residual values greater than 2.5 standard deviations were not
N. Kotlarski, B.K. O’Neill, E.H. Dunlop and G.L. Francis
included in regression calculations (Kuehl, 1994). Patterning in residual values was
tested according to the method of Draper and Smith (1966).
Experiments Performed
I. Mixing Experiments
Six refolding trials were conducted to investigate different mixing conditions (repeats
of three stirring rates). After the standard dissolution (see above) had progressed for
30 minutes, a sample was diluted to refolding conditions (2 M urea, 0.1 M Tris-HC1
pH9.1, 1.6mM Gly, 0.01 M EDTA, 0.64mM DTT, 1.6 mM OX-p-ME). The
dilution was achieved by taking a 600 pL sample of the dissolved mixture and rapidly
injecting it into a buffered solution contained in a translucent polycarbonate sample
tube (25 mm internal diameter) to give a final volume of 15.0 mL.
The different mixing rates were set by the frequency of rotation of a 1.0 cm
magnetic flea immersed in the solutions coupled to a magnetic stirrer. Minimum
stirring conditions were taken as zero rotation. That is, dissolved protein was injected
into the appropriate diluent buffer, gently swirled for 5 s and then placed on a
magnetic stirrer for the duration of refolding. ‘Mild mixing’ conditions were taken as
the minimum stable speed of the magnetic stirrer (approximately 100 rpm). For
‘maximum mixing’ conditions, the speed of the magnetic stirrer was set at the
maximum stable speed of the flea rotating in the diluent buffer (approximately 500
rpm). For mild and maximum mixing conditions, refolding was initiated by dilution of
a 600 pL sample of dissolving mixture into already mixing diluent buffer.
11. Control Concentrated Batch Refolding Trials
After dissolution had progressed for 30 minutes, two 2.50 mL samples were diluted
25% (v/v) with a single rapid injection of each sample into well mixed solutions,
giving a chemical environment of 2 M urea, 0.1 M Tris-HC1 pH 9.1, 10 mM Gly,
10 mM EDTA, 4 mM DTT, 10 mM OX-p-ME. Mixing was maintained at the level
used as the maximum rate in the mixing studies.
III. Fed-batch Refolding
Repeated fed-batch refolding trials were conducted using two GBC LC 1110 pumps to
feed dissolving inclusion body mixture and diluent buffer at a ratio of 1:3. Both
pumps were calibrated: one to deliver 8 M urea solution at 0.13 mL min-’ and the
other to deliver diluent buffer at 0.39 mL min-I. Fresh dissolution buffer (8 M urea,
0.1 M Tris-HC1 pH 9.1, 40 mM Gly, 10 mM EDTA, 16 mM DTT) and diluent buffer
(0.1 M Tris-HC1 pH 9.1, 10 mM EDTA, 13.3 mM OX-p-ME) were prepared. 1.25
mL of the dissolution buffer and 3.75 mL of the diluent buffer were mixed in a 50 mL
beaker and set stirring with a 2.0 cm flea on a magnetic stirrer. After dissolving the
inclusion bodies for 6 minutes, the pump feeding dissolution buffer was primed with
the dissolved material and returned to its set-point flow rate (0.13 mL min-I). Eight
minutes after initiating dissolution the primed feed lines were placed in the refolding
mixture and feeding commenced. After 60 minutes both pumps were stopped and
refolding continued for a further 4 hours. Throughout refolding, stirring speed was
maintained constant at 50% of ‘maximum’.
Improving Renaturation of Proteinsfiom Inclusion Bodies
Results and Discussion
I. Mixing Experiments
The purpose of conducting refolding trials in batch mode at Werent mixing rates, but
consistent chemical conditions, was to determine whether any effects on overall yield
and reaction rates could be identified.
For all three mixing rates dilution of the dissolved inclusion body mixture into
refolding buffer gave an opaque solution. The volume of insoluble material separated
fiom acid-quenched samples for analysis by HPLC showed a clear trend with time;
reducing over the first hour and subsequently increasing as refolding proceeded. The
minimum volume was obtained fiom samples quenched after refolding for
approximatelyone hour under all three mixing conditions.
Figure 2. Reducing SDS-PAGE of material precipitated during refolding of
LongkIGF-I inclusion bodies, dissolved using the standard method, at the maximum
stirring rate. Lane 1, Mark12 wide range protein standard Lanes 2-9, 10 pL of
refolding samples quenched at 0.50, 2.0, 7, 40, 150, 420, 600, and 1080 minutes after
the start of rejolding; lane 10: LongR'IGF-I standard.
A representative example of a SDS-PAGEseparation under reducing conditions of
the recovered insoluble material for one of the six refolding trials is presented in
Figure 2. The resolved bands indicate that the total amount of insoluble protein in the
acid-quenched samples followed the trend in total volume recovered. Although many
protein species were present in the insoluble material, LongR31GF-I predominated.
The approach used did not establish whether the insolubility occurred under the
refolding conditions, or as an artefact of the acid-quenching operation. With the
exception of the first sample (0.50 min), containing significant amount of material at
approximately 50 and 60 ma, the ratio of product to other contaminants remained
relatively constant throughout refolding. This suggests that the insolubility was likely
to be due to aggregation under refolding conditions, or at the very least, that the
proteins were changing state and hence solubility in acidic solution. It cannot be
N. Kotlarski, B.K. 0 'Neill, E.H. Dunlop and G.L. Francis
definitively concluded whether the proteins were restricted to self-association or
aggregated due to inter-molecular association. The constant ratio of product, 18 kDa,
and 40 kDa species is compatible with non-specific association during refolding as it
is most unlikely that three different species would have the same transient acidinsolubility behaviour. The band at 18 kDa may be a dimer of LongR31GF-Ithat was
not dissociated in reducing SDS-load buffer, and the 40 kDa species is likely to be an
outer membrane protein.
The total protein concentration present in the six refolding trials was quite
consistent within the range of 0.60 to 0.75 g L-'. There was no significant trend in the
overall protein concentration over time for any of the refolding trials. The standard
deviation in total protein concentration measurements was less than approximately 5%
of the mean. Constant total soluble protein concentration indicates that the fraction of
protein that aggregated was not significant relative to the overall amount. Further
evidence that aggregation was insignificant during prolonged incubation was provided
by HPLC analysis.
0.75 rnln
Elutlon tlmo-m
Figure 3. A typical time-series of HPLC profiles for refolding without stirring. The
peak corresponding to elution of reduced, denatured LongkIGF-I ( 0 ) is indicated in
the profile of the 0.75 min sample. Native species (N) is the predominant peak fiom
8.2 min onwards. Profiles are presented for an elution time of 9 to 28 minutes. AN
profiles have been normalized to accountfor the actual volume and dilution analysed.
Improving Renaturation of Proteins from Inclusion Bodies
A representative series of HPLC profiles of samples taken during the course of the
refolding without mixing over 24 hours is presented in Figure 3. Early time-points
reveal that an intermediate species forms rapidly which eluted early in the gradient,
followed by many poorly resolved species. As refolding progressed, the amount of
this rapidly formed intermediate decreased and native species accumulated to reach a
maximum after 80 to 150 minutes. A species that was resolved as a shoulder before
native LongR31GF-Iappeared upon W e r incubation. Appearance of this shoulder
was accompanied by a slight decrease in the amount of native species, suggesting that
slow structural rearrangement of the native species occurred. The rapid appearance of
a transient intermediate during refolding is consistent with previous studies receptorgrade LongR31GF-I, and is known to be due to a native-like conformation with two
disulfide bonds formed by the [18-611 and [6-481 cysteine pairings (Milner et al.,
1995). The observed trends in HPLC profiles for the experiments under mild and
maximum mixing conditions were consistent with those presented in Figure 3.
Table I. Rate coytants for the appearance of LongR31GF-I under dflerent mixing
conditiom. Estimates of rate constants and standard deviations were
regressed to the three-state refolding model using BMDP for each data set.
Muring rate
kN (min-I)
KA (mL rng-'min-')
k3 (min-')
kN (min-')
KA (mL mg-lrnin-l)
k~ (min-I)
kN (min-I)
KA (mLmg"min-')
k~ (min-I)
First repeat
Rate constant
Second repeal
Rote constant
8.8x lo4
-5.4x lo4
7.6x lo4
-3.9 10"
-3.9x 10-4
The concentration of LongR31GF-I present in each of the refolding mixtures
examined in the present study is presented in Figure 4. The curves have been
regressed to each set of data assuming the three-state refolding model using BMDP.
Rate constants and estimates of variance for the data are presented in Table I. The
rate constants for correct refolding and aggregation have a standard deviation of 1020% of their nominal values. However, the estimates of variance in the frst-order
back reaction from native to intermediate, indicate that no confidence can be placed in
the rate constants for kB.
N. Kotlarski, B.K. 0 ’Neill,E.H.Dunlop and G.L. Francis
Time (min)
Figure 4. A time-course of Longff‘IGF-I concentration for the j h t 5 hours in six
refolding trials at three different mixing conditions. The curves show the
regressedfit to each set of data assuming the three-state refolding model
using BMDP to estimate the rate constants,
Table II. Nested analysis of variancefor the mixing experiments at three stirring
Due to mixing
Between duplicates
A nested analysis of variance using normalized data of the repeated experiments
was conducted (see Table 11). A significant amount of the regression error is due to
differences between ‘duplicate’ experiments (F-statistic = 10.9). However, the Fstatistic value of 16.3 conclusively demonstrates that there is a significant difference
between the three mixing conditions (P>0.999). The rate constants fiom the overall
pooled data were: kN=0.17 f 0.05 min-’, KA = 0.67f 0.55 Lg-lrnin-’, and
kB = 1.7 x lo”% 0.035 min-’ (0.902 P). While the variance in kN is acceptable, the
range for KA indicates that there is a broad variation in second-order rate constant.
The uncertainty in k~ clearly demonstrates that product loss during extended
incubation was negligible.
The analysis of variance presented in Table I1 does not assign the observed
difference in mixing conditions to a particular reaction. Covariance analysis revealed
that for all of the data sets, kN and KA were highly correlated (correlation coefficients
approaching 1.000 in some cases). Clearly, the back reaction fiom native to
intermediate state was not the cause of the observed difference. Since the variance in
Improving Renaturation of Proteinsfrom Inclusion Bodies
the estimate of kN from the overall pooled regression is relatively low, it is probable
that the different behaviour can be correlated to the aggregation reaction. To test this
hypothesis, a nested analysis of variance was conducted using the normalized data
obtained from 0 to 150 min samples with k~ = 0.17 mh-' and kB set to zero (see Table
111). A significant amount of the regression error was due to differences between
'duplicate' experiments (P>0.9999). However, the F-statistic value of 14.9 established
that the difference between the three mixing conditions could be adequately described
as increasing second-order aggregation with increasing mixing rate with a constant
first-order rate constant of correct refolding (P>0.97).
Table III. Nested analysis of variance for mixing experiments with kN = 0.17 min-'
and kB = 0 for normalized LongkIGF-I concentrations collected
between 0 and I50 minutes.
Due to mixing
Between duplicates
Control Concentrated Batch Refolding Trials
Two batch refolding trials were conducted at high protein concentration to provide
controls for comparison with fed-batch refolding. Within 30 s of initiating refolding a
significant amount of material had precipitated, but did not appear to change upon
further incubation. The amount of insoluble material recovered by centrifugation of
acid-quenched samples for HPLC did not change with time. The volumes of the
insoluble material were notably larger than for samples taken fiom dilute refolding
mixtures at corresponding times.
The total concentration of soluble protein in each of the refolding mixtures was
estimated to be 3.8 f 0.3 g L-'and 3.4 f 0.3 g L-' (P I 0.90). The amount of reduced,
denatured LongR31GF-I added at the start of refolding corresponded to an overall
concentration of 1.87 f 0.10 g L-' (quantification by HPLC). The HPLC profiles for
samples taken from each refolding mixture at corresponding times were quite
The yield of correctly renatured LongR31GF-I,relative to the amount of denatured
fusion protein initially added, is presented as a function of time in Figure 5 . The
amount of renatured product increased to a maximum of approximately 5% within 20
minutes, but by 40 minutes had decreased significantly. Further incubation resulted in
only a slight decrease in the amount of native LongR31GF-I. The maximum yield was
approximately 10-fold less than the yield obtained in the mixing experiments at lower
protein concentration, indicating that the bulk of the product aggregated during
Regression to the three-state kinetic model of refolding reveals that the simplified
pathway does not provide a good description of the observed behaviour. The
regressed curves (in Figure 5 ) follow the general trend, but fail to describe the rapid
decrease in yield observed once the maximum had been reached. The concentration
of renatured product appeared to be approaching a stable equilibrium, rather than
N. Kotlarski, B.K. 0 'Neill,E.H. Dunlop and G.L. Francis
trending to zero as the model predicts. The poor tit of the model is reflected in the
relatively large values for the standard deviations in kN and KA (see Table IV).
Comparison of variance validates that given the uncertainty in the estimates of rate
constants, the experiments were true repeats (F-statistic = 0.57).
0.06 1
First repeat
Second repeal
Time (mh)
Figure 5. Plot of relative LongPIGF-I renaturation versus time for two concentrated
refolding trials. The curves show regressed fits to the three-parameter
model of refolding.
Examination of the residuals reveals that they are highly patterned with two
positive and two negative groups. That is, the model initially over-estimates the rate
of accumulation of native LongR31GF-I,under-estimates the maximum yield, does not
account for the rapid decrease in concentration, and tends to zero rather than a steady
state. The probability of obtaining only two positive groups and two negative groups
of residuals in each set of data is less than 0.03, indicating that the simplified kinetic
model is systematicallyflawed.
111. Fed-batch Refolding
For both fed-batch refolding trials, continuous addition of the dissolved starting
material resulted in a gradual reduction in the transparency of the mixtures.
Incubation without further addition of denatured material did not alter the appearance
of the mixtures. The amount of insoluble material separated by centrifugation of the
acid-quenched samples for analysis by HPLC, was consistent with the trend in visual
appearance; the amount of insoluble material increased with time during feeding and
subsequently did not change significantly.
The total concentrations of soluble protein in the two batches of dissolved
inclusion bodies used for fed-batch refolding were 13.9 k 0.1 and 11.8 f 0.1 g L-'
(P 10.90) and the concentration of denatured LongR31GF-I (estimated by HPLC) in
Improving Renaturation of Proteins from Inclusion Bodies
these mixtures was 8.2 and 7.7 g L-’. The observed and nominal protein
concentrations (assuming all protein remained soluble) are presented in Table V. In
the first trial, the observed protein concentration was approximately 40% less than the
nominal value at both sampling times. In the second trial approximately 30% and
20% of the nominal total soluble protein was absent from the first and second samples,
respectively. These values confirm that the insoluble material observed during
refolding was likely to contain a substantial amount of aggregated LongR31GF-I.
The HPLC profiles of samples quenched at corresponding times during the
refolding of each batch of clean inclusion bodies were quite consistent. A time series
of HPLC profiles for the second refolding trial is presented in Figure 6. The
relative amount of the species eluting at the expected time of native LongR’IGF-I, and
the total area under the curves increased in the profiles up to 55 minutes. After the
completion of feeding (60 min.), the total area beneath the curves and the relative
amount of LongR31GF-I began to decrease.
500 mln
160 min
Sb mln
76 min
66 mln
56 mln
16 mln
8 mln
4 mR
Elution limo
Figure 6. A time series of HPLC profiles for fid-batch refolding. The expected
elution time of renatured LongR’IGF-I (Iv) corresponds to the predominant
species in all profiles, LIS indicated in the 300 minute profile. All profiles
are have been corrected to show an equivalent loading.
N. Kotlarski, B.K. 0 'Neill,E.H.Dunlop and G.L. Francis
The concentrations of native LongR31GF-I,determined by integration of the HPLC
profiles using PeakFitfM, are presented in Figure 7. The rapid decrease in the amount
of native protein between 60 and 100 minutes was clearly evident in both refolding
trials. The amount of native protein appeared to stabilize at approximately 0.2 g L-'.
This corresponds to a relative renaturation of approximately 10% at an overall
LongR31GF-I concentration of 1.75 g L-' (the average of the two trials). In the
concentrated batch refolding trials, the overall renaturation yield was approximately
4% at an overall LongR31GF-I concentration of 1.87 g L-' (see Figure 5). However, if
the fed-batch refolding had been quenched at the same time as feeding ceased, the
concentration of LongR31GF-I would have been at a maximum of 0.35-0.4 g L-'
corresponding to a renaturation yield of 20-23%; approximately 5-fold greater
renaturation than obtained with batch refolding.
Tim. (min)
Figure 7. Plot of correctly folded LongR'IGF-I concentration versus time for both
fed-batch refolding trials. The approximate Concentration predicted
j?om batch trials is shown as a solid line. Regressed curvesfor each data
set are shownfor kN = 0.12 min-I.
Figure 7 shows the predicted LongR31GF-I concentration during refolding, based
on the three-state kinetic model, using rate constants regressed fiom concentrated
refolding trials (the average of values presented in Table IV). The shape of the curve
during feeding follows the same trend as the experimental data, but at approximately
half the protein concentration. The model predicts that once feeding ceases there
should be a gradual decrease in the product concentration. However, the apparent
behaviour was a rapid decrease to a constant value. The discrepancy between
predicted and observed behaviour suggests that the rate of aggregation during fedbatch refolding was less than detected in the concentrated refolding trials. A possible
contribution to this altered behaviour is the different mixing conditions used for fedbatch refolding.
Improving Renaturation of Proteins from Inclusion Bodies
Table I K Estimates of rate constants and standard deviations regressed to the threestate refolding model using BMDP for refolding at high product concentration.
kN (min-')
KA (mL mg-'min-')
k,g (min-')
First repeat
Rate constant
Second repeat
Rate constant
Regression of the three-state kinetic model to the observed transient LongR31GF-I
concentration proved difficult. The additional degrees of freedom introduced by the
discontinuity in the feeding strategy allow a range of rate constants to describe the
data and KA and kN are highly correlated. To find a sensible converged solution, the
value of kN was fixed at the average rate constant regressed to the concentrated
refolding data (i.e. 0.12 rnin-'; see Table VI). Even then, the proportionately large
values for the standard deviation indicate that the model is unable to describe the
observed behaviour satisfactorily. The regressed curves for each data set are included
in Figure 7. The fed-batch refolding trials can be regarded as true repeats given the
relatively large uncertainty in the rate constants (F-statistic = 0.63).
Table K Observed and nominal (assuming all protein remains soluble) total protein
concentrations during fed-batch refolding trials. Feed concentrations were
13.9 and 11.8 g L-' in thefirst and second repeats, respectively.
40 min (g L-')
90 min. 7s L-i)
First repeat
1.68 f 0.32
1.85 f 0.31
Second repeat
1.73 f 0.40
2.05 f 0.40
Regression of the fed-batch data to the three-state kinetic model describes the
actual behaviour with similar inadequacies as in the case of concentrated refolding. A
plot of the residuals is presented in Figure 8. Analysis of the residuals by the method
of Draper and Smith (1966) confirms that the model provides a flawed description.
The probability of obtaining this grouping of residuals is less than 7% relative to a
normal distribution.
Table VI. Estimates of rate constants and standard deviations from fed-batch
refolding trials (kN = 0.12 min-9.
kN ( m i d )
KA (mL mg-'min-')
k,g (min-')
First repeat
Std deviation
Rate constant
Assumed = 0
I .6
Second repeat
Std deviation
Rate constant
Assumed = 0
N. Kotlarski, B.K. O'Neill, E.H. Dunlop and G.L. Francis
Selected results of SDS-PAGE of samples taken from the dissolved starting
material and subsequent fed-batch refolding are presented in Figure 9. Lane 2 shows
a typical separation of the proteins present in the dissolving mixture of inclusion
bodies. Under reducing conditions, the notable change in the resolved bands was a
decrease in the amount of 18 kDa species and a corresponding increase in the band at
the correct position for monomeric LongR31GF-I. Lanes 3 and 4 show the resolution
of protein species present under reducing conditions in the refolding mixture for
samples taken at 60 and 95 minutes, respectively. There is a slight increase in the
relative amount of protein present in the band corresponding to dimers of LongR31GFI in the latter sample. Lane 9 shows the distribution of protein species present in
concentrated batch refolding under non-reducing conditions for comparison with the
results of fed-batch refolding.
Figure 8. Plot of standardized residualsfor bothfed-batch refolding trials.
Lanes 5 to 8 show a time-course of samples taken from the refolding mixture under
non-reducing conditions. After only 6 minutes, the ladder of bands with molecular
weights corresponding to aggregates of LongR31GF-I is evident. Allowing for the
greater volume of sample loaded, there is still significantly more protein in the sample
taken after 6 minutes than subsequent samples. In the samples taken at 44,62 and 96
minutes a significant amount of the protein can be seen trapped in the stacking gel. It
appears that much of the protein aggregated extensively and was too large to enter the
running gel. The relative amount of multimers in the samples taken at later times
appears to be much less than in the first sample. However, quantification of the ratio
of monomeric to dimeric, trimeric, and tetrameric forms by densitometry does not
susbstantiate this conclusion.
The relative amounts of multimeric to monomeric species in material refolded in
batch mode were significantly greater than when fed-batch operation was employed.
Improving Renaturation of Proteins from Inclusion Bodies
This indicates that aggregation of refolding monomers was reduced by gradual
addition of the denatured protein. Quantification of the amount of correctly renatured
LongR31GF-I by HPLC supports this observation. Comparison of the HPLC profiles
of batch and fed-batch refolding revealed that the relative proportion of correctly
renatured LongR31GF-I was greater when fed-batch refolding was employed.
However, analysis by SDS-PAGEshows that much less protein was resolved in the
fed-batch samples despite approximately the same overall protein concentration (see
lanes 7 and 9 in Figure 9). The measured total soluble protein concentration suggests
that a greater hction of the protein remained soluble during batch refolding. It
appears that fed-batch refolding resulted in a greater yield of correctly renatured
LongR31GF-I compared to batch refolding, despite apparently greater association of
monomeric and multimeric species. These observations add further weight to an
argument suggesting that the three-state refolding pathway is not an accurate
simplification for the refolding of LongR31GF-I despite being the most widely
accepted model. However, the simplified pathway did prove useful for comparing
refolding performance under different mixing rates at low protein concentration.
Hence, it may provide approximate (but adequate) predictions in mathematical
simulations in the absence of a superior model, particularly for operation in batch
mode and at low protein concentration.
Figure 9. SDS-PAGE of samples taken +om refolding mixtures. Lane I , protein
standard lane 2, dissolving mixture @er 70 min (6 pL, non-reducing); lane 3,
refolding at termination of feeding (6 pL, reducing); lane 4, refolding @er 95 min (6
pL, reducing); lane 5, refolding afier 6 min (I5 pL, non-reducingk lane 6, refolding
@er 44 min (15 pL, non-reducing); lane 7, refolding afier 62 min (15 pL, nonreducing); lane 8, refolding afier 96 min (15 pL, non-reducing); lane 9,
corresponding batch-refolded inclusion bodies afier 60 minutes (1 0 pL, nonreducing); lane 10, LongPIGF-I standard
N. Kotlarski, B.K. 0 'Neill, E.H. Dunlop and G.L. Francis
This work has presented a case study of two important issues that must be considered
when refolding is conducted by dilution on a large scale. In the absence of a known
kinetic refolding pathway, the generic scheme illustrated in Figure 1 can be regressed
to give an adequate description of the refolding of LongR31GF-I. However, the
observed behaviour of refolding at high protein concentration in batch and fed-batch
modes demonstrates that the actual refolding pathway is more complex than the
assumed simplified scheme.
The rate of mixing has been identified as a significant factor in determining the
yield of correctly renatured LongR31GF-I during refolding. Increased agitation
accelerated the rate of off-pathway aggregation and consequently increased product
loss. This finding suggests that an optimal mixing rate exists in order to achieve
maximum renaturation: adequate mixing is required to disperse the dissolved starting
material and avoid heterogeneity, but excessive mixing reduces overall yield by
increasing the rate of aggregation.
Fed-batch refolding at a constant feed rate gave superior yield than batch dilution
when conducted at high concentration. The duration of feeding and mixing rate were
not optimized, yet a five-fold increase in yield was obtained. This suggests fed-batch
operation as the favoured mode for large-scale refolding where high protein
concentration is desirable.
Apparent second-order rate constant of aggregation (L g-lmin-')
First-order rate constant of back reaction fiom native to intermediate (min-')
rate constant of correct refolding (min-')
measured at 600 nm.
Ambrosius, D. and Rudolph, R. 1997. Process for the reactivation of denatured protein. US Patent
5.6 18,927.
Chaudhuri, J.B. 1994. Refolding recombinant proteins: process strategies and novel approaches. Ann.
N.Y. Acad. Sci. 721: 374-385.
Chaudhuri, J.B., Batas, B., and Guise, A.D. 1996. Improving protein refolding yields by minimizing
aggregation. Ann. N.Y. Acad. Sci. 782: 495-505.
Cleland, J.L. and Wang, D.I.C. 1990. Refolding and aggregation of bovine carbonic anhydrase B: quasielastic light scattering analysis. Biochemistry 29: 11072-11078.
Cleland, J.L. and Wang, D.I.C. 1993. In vifro protein refolding, pp. 527-555. In: G. Stephanopoulos
(ed.), Biotechnology,vol. 3, Bioprocessing, 2nd edition. VHC Publishers Inc., New York.
Dorin, G., Gray, D.R., Chang, B.S.,Cowgill, C.A., and Milky, R.J. 1995. Process for purifying bacterially
produced M-CSF. US Patent 5,466,781.
Draper, N.R.and Smith, H. 1966. Applied regression analysis. John Wiley & Sons, New York,pp.95-98.
Fischcr, B., Perry, B., Sumner, l., and Goodenough, P. 1992. A novel sequential procedure to enhance the
renaturation of recombinant protein from Escherichiu coli inclusion bodies. Prot. Eng. 5: 593-596.
Fischer, B., Sumner, l., and Goodenough,P. 1993. Isolation, renaturation, and formation of disulfide bonds
of eukaryotic proteins expressed in Escherichia colt as inclusion bodies. Biotechnol.Bioeng. 41: 3-13.
Improving Renaturation of Proteins from Inclusion Bodies
Francis, G.L., ROSS,M., Ballard, F.J., Milner, S.J., Senn, C., McNeil, K.A., Wallace, J.C., King, R., and
Wells, J.R.E. 1992. Novel recombinant fusion protein analogues of Insulin-like Growth Factor (IGF)-I
indicate the relative importance of IGF-binding proteins and receptor binding for enhanced biological
potency. J. Mot. Endo. 8: 213-223.
Galliher, P.M. 1991. Continuous method of refolding proteins. US Patent 4,999,422.
Georgiou, G. 1988. Optimizing the production of recombinant proteins in microoganisms, AlChE J. 34:
Georgiou, G. and Valax, P. 1996. Expression of correctly folded proteins in Escbericbin coli. Current
Opinion in Biotechnology 7: 190-197.
Goldberg, M.E., Rudolph, R., and Jaenicke, R. 1991. A kinetic study of the competition between
renaturation and aggregation during the refolding of denatured-reduced egg white lysozyme.
Biochemistry 30: 2790-2797.
Hejnaes, K.R., Bayne, S., Nsrskov, L., Holmegarrd Serensen, H., Thomsen, J., SchWer, L., Wollmer, A,,
and Skriver, L. 1992. Development of an optimized refolding process for recombinant Ala-Glu-IGF1. Prot. Eng. 5: 597-806.
Hevehan, D.L. and De Bernardez Clark, E. 1997. Oxidative renaturation of lysozyme at high
concentrations. Biotechnol. Bioeng. 54: 22 1-230.
Jaenicke, R. and Rudolph, R. 1986. Refolding and association of oligomeric proteins. Meth. Enzymol.
131: 218-250.
Kiefhaber, T., Rudolph, R., Kohler, H.H., and Buchner, J. 1991. Protein aggregation in vibo and in vivo:
A quantitative model for the kinetic competition between folding and aggregation. Bioflecbnology 9:
Kotlarski, N., O’Neill, B.K., Francis, G.L., and Middelberg, A.P.J. 1997. Design analysis for refolding of
monomeric protein. AlChE J. 43: 2123-2132.
Kuehl, R.O. 1994. Statistical principles of research design and analysis. Wadsworth Publishing Company,
Belmont, California, pp. 116.
Marston, F.A.O. 1986. The purification of eukaryotic polypeptides synthesized in E. coli. Biochem. J.
240: 1-12,
Marston, F.A.O. and Hartley, D.L. 1990. Solubilizationof protein aggregates. Meth. Enzymol. 182: 264276.
Matthiesen, F., Hejnaes, K.R., and Skriver, L. 1996. Stabilization of recombinantly expressed proteins.
Ann. N.Y. Acad. Sci. 782: 413421.
Middelberg, A.P.J. 1996. The influence of protein refolding strategy on cost for competing reactions,
Chem. Eng. J. 61: 41-52.
Milner, S.J., Francis, G.L.,
Wallace, J.C., Magee, B.A., and Ballard, F.J. 1995. Mutations in the B-domain
of Insulin-like Growth Factor-I influence the oxidative folding to yield products with modified
biological properties. Biochem. J. 308: 865-871.
Mitraki, A,, Haase-Pettingell, C., and King, J. 1991. Mechanisms of inclusion body formation. In: G.
Georgiou and E. De Bemardez-Clark (eds), Protein Refolding. ACS Symposium Series 470: 3549.
Rudolph, R. 1990. In: H. Tscchesche (ed.), Modem methods on protein and nucleic acid analysis. Walter
de Gruyter, Berlin, pp. 149-171.
Rudolph, R. and Fischer, S. 1990. Process for obtaining renatured proteins. US Patent 4,933,434.
Rudolph, R. and Lilie, H. 1996. In viba folding of inclusion body proteins. FASEB 1. 10: 49-56.
Samuelsson, E. and Ulhkn, M. 1996. Chaperone-like effect during in vitro refolding of Insulin-Like
Growth Factor I using a solubilizing fusion partner. Ann. N.Y. Acad. Sci. 782: 486494.
Weissman, J.S. and Kim, P.S. 1991. Reexamination of the folding of BPTI: predominance of native
intermediates. Science 253: 1386-1393.
Wells, J.R.E., King, R.M.,and Francis, G.L. 1994. Growth hormone fusion proteins, methods of
production, and methods of treatment. US Patent 5,330,971.
ZettlmeiDI, G., Rudolph, R., and Jaenicke, R. 1979. Reconstitution of lactic dehydrogenase. Noncovalent
aggregation vs. reactivation. 1. Physical properties and kinetics of aggregation. Biochemistry 18:
Received 6 June 2001; Accepted after revision: 20 December 200 1.
Без категории
Размер файла
1 326 Кб
renaturation, improving, inclusion, protein, bodies
Пожаловаться на содержимое документа