ARTICLE
`
`Factors Affecting Protein Refolding Yields in a
`Fed-Batch and Batch-Refolding System
`
`Gareth J. Mannall, Nigel J. Titchener-Hooker, Paul A. Dalby
`
`Department of Biochemical, Advanced Center for Biochemical Engineering, Engineering,
`University College London, Torrington Place, London, WC1E 7JE, United Kingdom;
`telephone: 44 20 7679 2962; fax: 44 20 7209 0703; e-mail: p.dalby@ucl.ac.uk
`
`Received 6 March 2006; accepted 31 January 2007
`
`Published online 15 February 2007 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/bit.21377
`
`ABSTRACT: The refolding of recombinant protein from
`inclusion bodies expressed in Escherichia coli can present
`a process bottleneck. Yields at industrially relevant concen-
`trations are restricted by aggregation of protein upon
`dilution of the denatured form. This article studies the effect
`of five factors upon the dilution refolding of protein in a
`twin impeller fed-batch system using refold buffer contain-
`ing only the oxidized form of the redox reagent. Such a
`buffer is easier to prepare and more stable than a buffer
`containing both reduced and oxidized forms. The five
`factors chosen were: bulk impeller Reynolds number,
`mini-impeller Reynolds number, injection rate of denatured
`protein, redox ratio, and guanidine hydrochloride (GdHCl)
`concentration. A 25 factorial experiment was conducted
`at an industrially relevant protein concentration using
`lysozyme as the test system. The study identified that in
`the system used, the guanidine hydrochloride concentration,
`redox ratio, and injection rate were the most important
`factors in determining refolding yields. Two interactions
`were found to be important: redox ratio/guanidine hydro-
`chloride concentration and guanidine hydrochloride con-
`centration/injection rate. Conditions were also found at
`which high refolding yields could be achieved even with
`rapid injection and poor mixing efficiency. Therefore, a
`comparative assessment was carried out with minimal
`mixing in a simple batch-refolding mode of operation,
`which revealed different behavior to that of fed-batch. A
`graphical (windows of operation) analysis of the batch data
`suggested that optimal yields and productivity are obtained
`at high guanidine hydrochloride concentrations (1.2 M) and
`redox ratios of unity or greater.
`Biotechnol. Bioeng. 2007;97: 1523–1534.
`ß 2007 Wiley Periodicals, Inc.
`KEYWORDS: refolding; chaotrope concentration; redox
`ratio; mixing; fed-batch
`
`Introduction
`
`Expression of protein as inclusion bodies offers several
`processing advantages over expression as soluble protein.
`Inclusion bodies are resistant to proteolytic enzymes, are
`easy to isolate, and the existence of the protein in an inactive
`form allows expression of proteins that are potentially toxic
`to the host cell (De Bernardez Clark et al., 1998; Li et al.,
`2004). Active protein is derived from inclusion bodies by a
`procedure, which starts with the inclusion bodies being
`released from the cells by mechanical or chemical
`lysis
`(Falconer et al., 1999). The inclusion bodies are then isolated
`by filtration or centrifugation. Purification is achieved via
`several washing steps, which remove non-specifically bound
`contaminants, typically cell debris. Purified inclusion bodies
`are then solubilized using high concentrations of chaotropic
`reagents such as urea or guanidine hydrochloride (GdHCl),
`in the presence of reducing agents (i.e., Dithiothreitol
`(DTT), b-mercaptoethanol (BME)) to break disulfide
`bonds. After
`solubilization,
`the denatured protein is
`refolded. This is most commonly achieved by diluting the
`chaotropic agent, in the presence of redox reagents, to
`produce native protein. The refolding step often represents a
`bottleneck and is a step where yields could be improved
`considerably (Buswell and Middelberg, 2003; De Bernardez
`Clark et al., 1998).
`Lysozyme has often been used as a model protein to study
`the effects of operating variables on refolding, as the folding
`pathway for this protein has been previously characterized in
`significant detail (Buswell and Middelberg, 2003; Kiefhaber,
`1995; Radford et al., 1992; Wildegger and Kiefhaber, 1997).
`The majority of refolding (approximately 86%) occurs via a
`single intermediate, whilst a small proportion (approxi-
`mately 14%) refolds by a fast pathway (Kiefhaber, 1995).
`The monomolecular refolding reaction competes for the
`intermediate with a multimolecular aggregation reaction
`
`Correspondence to: P.A. Dalby
`
`ß 2007 Wiley Periodicals, Inc.
`
`Biotechnology and Bioengineering, Vol. 97, No. 6, August 15, 2007 1523
`
`APOTEX EX1042
`
`Page 1
`
`

`
`inactive aggregates. Previous
`insoluble,
`that produces
`studies on lysozyme have examined the independent effects
`of process factors on refolding yield,
`including GdHCl
`concentration, redox ratio, protein concentration, mixing
`intensity,
`reactor
`type,
`inclusion body contaminants,
`and pH (De Bernardez Clark et al., 1998; Hevehan and
`De Bernardez Clark, 1997; Katoh et al., 1999; Katoh and
`Katoh, 2000; Lee et al., 2002; Maachupalli-Reddy et al., 1997;
`Mannall et al., 2006; Willis et al., 2005).
`The effects of individual parameters on refolding yields
`have also been characterized for a number of proteins
`including ribonuclease, trypsinogen, and Fab fragments
`(Anfinsen and Haber, 1961; Buchner and Rudolph, 1991;
`Buswell et al., 2002; Fischer et al., 1993). However, few
`studies have looked at these parameters taken together in a
`factorial experiment (Buswell et al., 2002). Such factorial
`experiments can be used to determine the influence of each
`parameter upon a defined response such as refolding yield,
`and to quantify the degree of interactions between them.
`Interactions suggest either positive or negative cooperative
`effects, beyond the additivity of the parameters alone, and an
`understanding of
`them is key to designing optimized
`processes.
`The formation of native protein is favored at low protein
`concentrations where the rate of aggregation is minimized
`(Buswell et al., 2002). Consequently, the gradual addition of
`denatured protein to refolding buffer results in increased
`yields due to maintaining a high refolding efficiency at low
`protein concentrations and the presence of lower concen-
`trations of partially refolded protein over the duration of
`the addition (Katoh et al., 1999). In a previous study, we
`investigated the refolding of lysozyme using a fed-batch
`refolding method, conducted in a twin impeller reactor
`designed to achieve higher levels of dispersion of the injected
`protein relative to a single impeller system (Mannall et al.,
`2006). A refolding buffer containing only the oxidized form
`of the redox reagent (cystamine) was also used, such that a
`redox ratio was obtained upon addition of the denatured
`protein solution containing an excess of the reducing agent
`DTT. In this system protein concentration, chaotrope
`concentration, and redox ratio within the reactor are each
`directly linked to the injection volume such that redox ratio
`(reduced: oxidized) increases as the concentrations of
`protein and chaotrope rise. Typical dilution refolding
`methods add denatured protein to a buffer containing a
`predetermined ratio of oxidized and reduced redox reagent.
`An advantage of the variable redox ratio method over more
`commonly used methods is the greater stability to air
`oxidation of the refolding buffer, as it is already fully
`oxidized. Additionally, the variable redox approach ensures
`complete reduction of the protein disulfide bonds in the
`denaturation buffer, and then dilution into a fully oxidizing
`refold buffer.
`This article takes a systematic factorial approach to assess
`the importance and interactions of five key factors on the
`refolding yield of
`lysozyme, using the same fed-batch
`reactor. The five factors chosen were the: bulk impeller
`
`Reynolds number (Reb); mini-impeller Reynolds number
`(Rem); injection rate (Q); redox- ratio (R); and guanidine
`hydrochloride concentration (G). The effect of protein
`concentration was omitted as the decrease in refolding yields
`at increased protein concentration has been well docu-
`mented (De Bernardez Clark et al., 1998; Goldberg et al.,
`1991; Yasuda et al., 1998). The factorial design experiment
`varies each component of the refolding solution (guanidine
`hydrochloride concentration and redox ratio, to change the
`environment experienced by the denatured protein), and
`also each of the other factors, both independently and in all
`possible combinations. Our previous study (Mannall et al.,
`2006) used a twin impeller to examine the importance of
`dispersive mixing. The mini-impeller used may provide
`sufficient dispersion of protein molecules
`to prevent
`aggregation,
`increasing the selectivity of
`the refolding
`reaction over that of aggregation. For this reason, both
`impellers were tested for their effect on the refolding
`reaction. It was not the aim of this article to optimize the
`refolding of lysozyme per se, as this has been achieved
`elsewhere, but rather to understand the relative importance
`and interactions of the five factors, in fed-batch dilution
`using the redox system described. Interactions suggest that
`the effect of a given factor is dependent on the level of
`another factor. Knowledge of such interactions is critical for
`the effective design of efficient refolding strategies. It has
`been shown previously that the redox ratio is critical
`to refolding yields, with an optimum existing at about
`2:1 (reduced:oxidized) for lysozyme (De Bernardez Clark
`et al., 1998; Hevehan and De Bernardez Clark, 1997). This
`study seeks to establish whether the effect of redox ratio is
`still significant in a system where the effective redox ratio is
`defined by the volume of denatured protein added. It
`also investigates whether the injection rate still remains
`influential in the presence of very low initial redox ratios.
`Observations from the factorial study suggested that it
`would be valuable to study batch refolding as a comparison
`to fed batch systems. The second part of this study uses a
`batch high throughput study to understand how the
`behavior of the system changes under batch conditions,
`and to observe if optimal conditions might be achieved in a
`simpler less expensive system. A graphical windows of
`operation approach is used to determine the influence of
`conditions on yield and time to yield.
`
`Materials and Methods
`
`Materials
`
`Chicken egg white lysozyme, lyophilized powder approx.
`50,000 units/mg protein, was purchased from Sigma-
`Aldrich Company Ltd (Poole, Dorset, England, UK).
`All additional chemicals were sourced from Sigma
`Aldrich Company Ltd (Poole) and were of at least reagent
`grade.
`
`1524
`
`Biotechnology and Bioengineering, Vol. 97, No. 6, August 15, 2007
`
`DOI 10.1002/bit
`
`Page 2
`
`

`
`Twin Impeller Reactor
`
`All experiments used a twin impeller reactor as described
`previously (Mannall et al., 2006); a schematic of its design is
`detailed in Figure 1. It comprised a 0.25 L capacity tank with
`heating/cooling jacket: DT¼ 63 mm, HT¼ 80 mm. The tank
`was baffled so as to increase the efficiency of mixing with
`four baffles W¼ 7 mm. The bulk impeller was a six blade
`Rushton impeller DI¼ 24 mm, DD¼ 19 mm, WB¼ 6 mm,
`HB¼ 2 mm. The mini-impeller was a two blade paddle
`impeller: DI¼ 10 mm WB¼ 6 mm, HB¼ 2 mm at
`approximately 858 to the horizontal, 16 mm above the
`bulk impeller blade, 5 mm off center from the bulk impeller,
`and 5 mm below the injection point. The injection point was
`at 708 to the horizontal.
`The bulk impeller was driven by a Heidolph R2R-2000
`motor (Heidolph Instruments, Schwabach, Germany). The
`1
`mini-impeller was driven by a Eurostar digital, IKA
`Labortechnick stirrer (Eurostar, Staufen, Germany). Spec-
`trophotometric measurements were performed on a
`Genesys6 spectrophotometer (Thermo-Spectronic, Roche-
`ster, NY). Temperature was measured using a 2000 series
`temperature probe (Jencons Scientific, Leighton Buzzard,
`Bedfordshire, UK). A water bath was used to keep the
`reactor at a defined temperature (SU6 water bath, Grant
`
`Figure 1. Scale Down Twin-Impeller stirred tank reactor. The stirred tank
`reactor has two impellers:
`the bulk impeller controls bulk flow within the
`reactor; the mini-impeller a 2 blade paddle positioned just below the injection
`point rapidly disperses materials upon injection into the reactor. Reactor configura-
`tion: DT¼ 63 mm, HT¼ 80 mm. Bulk impeller dimensions: DI¼ 24 mm, DD¼ 19 mm,
`WB¼ 6 mm, HB¼ 4.5 mm. The bulk impeller was 21 mm from the bottom of the tank.
`Mini-impeller dimensions: DI¼ 10 mm WB¼ 6 mm, HB¼ 2 mm, at approximately 858 to
`horizontal, 16 mm above bulk impeller blade, 5 mm off center from bulk impeller, 0.5 mm
`below injection point. Injection point: 858 to the horizontal.
`
`Instruments Ltd, Cambridge UK). The pump utilized for all
`additions experiment was a peristaltic P-1 pump (GE
`Healthcare, Uppsala, Sweden).
`
`Denaturation and Reduction of Lysozyme
`Lysozyme (16 0.1 mg mL
`1) was denatured and reduced
`in 8 M GdHCl, 32 mM DTT, 50 mM Tris, 1 mM EDTA pH 8
`for 2 h at 37 8C. Aliquots of denatured lysozyme were stored
`at 808C prior to use. Once thawed, aliquots were not
`refrozen. Denatured lysozyme concentrations were mea-
`sured at 280 nm after 1:100 dilution in acetic acid (0.1 M)
`1
`1 mg
`and using an extinction coefficient of 2.37 mL cm
`(lysozyme) (De Bernardez Clark et al., 1998). The denatured
`state was confirmed by reverse phase high pressure
`(performance) liquid chromatography (RP-HPLC).
`
`Reverse Phase HPLC (RP-HPLC)
`
`Assessment of native lysozyme concentration was achieved
`using a method based upon that of Lee et al. (2002).
`A Jupiter C5 reversed phase column (5 mm, 300 A˚ ,
`150 mm 4.6 mm, Phenomenex, Macclesfield, UK) was
`used on a Beckman (High Wycombe, UK) Gold High
`Pressure Liquid Chromatography system, comprising a
`126 Pump Unit, 166 Detector unit, 507e Autosampler unit,
`and employing System GOLD software. A linear acetoni-
`trile-water gradient 0.1% TFA (v/v) (30–46% over 12 min)
`1 was run. samples (100 mL) were
`at a flow rate of 1 mL min
`applied to the column. A standard curve was generated and
`yields from refold samples calculated from this. Lysozyme
`concentrations were assessed at 280 nm using an extinction
`1 mg
`1 (De Bernardez Clark
`coefficient of 2.63 mL cm
`et al., 1998). Native lysozyme had a retention time of
`8.57 0.03 min, denatured lysozyme had a retention time of
`11.20 0.1 min.
`
`Effect of Guanidine Hydrochloride Concentration
`1) was diluted
`Denatured reduced lysozyme (16 mg mL
`1:15 in various refold buffers (six refold buffers in total as
`detailed in Table I) at ambient temperature and mixed
`vigorously for 30 s, and left at ambient temperature. After
`2 h, a sample (2 mL) was taken and quenched with 10% v/v
`TFA (200 mL). A further sample was taken at 24 h and left
`unquenched. Samples were assayed for native lysozyme
`using RP-HPLC.
`
`Factorial Experiments
`
`1) was refolded by
`Denatured reduced lysozyme (16 mg mL
`1:15 dilution at 258C, once for each of the combinations of
`different mini-impeller intensities (Rem), bulk-impeller
`intensities (Reb), injection rates (Q), redox ratios (R) and
`final GdHCl concentrations (G) as detailed in Table II.
`
`Mannall et al.: Factors Affecting Protein Refolding
`
`1525
`
`Biotechnology and Bioengineering. DOI 10.1002/bit
`
`Page 3
`
`

`
`Table I. Compositions of buffers utilized when studying the effect of GdHCl concentration on refolding yields.
`
`Buffer
`
`[Tris HCl] (mM)
`
`[Cystamine] (mM)
`
`[EDTA] (mM)
`
`[GdHCl] (M)
`
`Final [GdHCl] (M)
`
`1
`2
`3
`4
`5
`6
`
`50
`50
`50
`50
`50
`50
`
`4
`4
`4
`4
`4
`4
`
`1
`1
`1
`1
`1
`1
`
`0
`0.167
`0.47
`0.717
`0.967
`1.967
`
`0.53
`0.69
`0.97
`1.20
`1.44
`2.38
`
`pH
`
`8.0
`8.0
`8.0
`8.0
`8.0
`8.0
`
`Table II.
`
`25 factorial experimental design.
`
`Cond no.
`
`1
`2
`3
`4
`5
`6
`7
`8
`
`9
`
`10
`
`11
`
`12
`13
`
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`25
`26
`27
`28
`29
`30
`31
`32
`33
`34
`35
`36
`37
`38
`39
`40
`
`Reb
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`0
`0
`0
`0
`0
`0
`0
`0
`
`Rem
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`
`
`
`
`
`
`
`
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`þ
`
`
`
`
`
`
`
`
`0
`0
`0
`0
`0
`0
`0
`0
`
`Q
`þ
`þ
`þ
`þ
`
`
`
`
`þ
`þ
`þ
`þ
`
`
`
`
`þ
`þ
`þ
`þ
`
`
`
`
`þ
`þ
`þ
`þ
`
`
`
`
`0
`0
`0
`0
`0
`0
`0
`0
`
`R
`þ
`þ
`
`
`þ
`þ
`
`
`þ
`þ
`
`
`þ
`þ
`
`
`þ
`þ
`
`
`þ
`þ
`
`
`þ
`þ
`
`
`þ
`þ
`
`
`0
`0
`0
`0
`0
`0
`0
`0
`
`G
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`þ
`
`0
`0
`0
`0
`0
`0
`0
`0
`
`Y@2 h
`
`Y@24 h
`
`78
`41
`36
`30
`46
`46
`27
`33
`67y
`35y
`36y
`33y
`51
`41y
`29
`34
`76
`36
`35
`32
`54
`43
`32
`28
`70
`38
`41
`33
`48
`42
`33
`30
`56
`50
`57
`58
`59
`62
`62
`60
`
`92
`43
`52
`34
`96
`69
`50
`55
`94y
`38y
`75y
`40y
`100
`58y
`48
`57
`92
`39
`50
`41
`100
`65
`67
`49
`94
`40
`63
`42
`100
`63
`70
`51
`70
`75
`69
`67
`73
`77
`70
`73
`
`
`
`which were
`Each condition (Cond) was repeated once, except for
`repeated four times. Eight midpoints were performed to provide an estimate
`of the variance in the data. Experiments were performed in a randomized
`order. Reb bulk impeller Reynolds number, Rem mini-impeller Reynolds
`1), R, Redox ratio, G, final GdHCl (M)
`number, Q injection rate (mL min
`concentration (þ), upper level, (), lower level and (0) midpoint. Results
`expressed in terms of protein yield: (Y@2 h) quenched at 2 h, (Y@24 h)
`quenched at 24 h. All data above were used to calculate 95% confidence for
`each response. y average of replicated data.
`
`1526
`
`Biotechnology and Bioengineering, Vol. 97, No. 6, August 15, 2007
`
`Factor levels used in the factorial experiments are detailed in
`Table III. Eight replicates of the midpoint were performed.
`These were used to determine not only error in the data, but
`also how well the data fits with a linear model. The midpoint
`had mean of 57.9 (standard deviation 3.84 (6.63% error)
`and a mean of 71.8 (standard deviation 3.34 (4.65%
`error)) for samples quenched at 2 h and 24 h respectively. In
`all cases, samples were mixed for a total of 2 h, and then left
`for a further 22 h at ambient temperature. Samples (2 mL)
`were removed and quenched at 2 h and 24 h with 200 mL of
`10 % (v/v) TFA. Samples were assayed by RP-HPLC to
`determine refolding yields. Five refold buffers were prepared
`as detailed in Table IV.
`To further prove data replicability conditions 9,10,11,12,
`and 14 (Table II) were repeated a further three times. These
`points were chosen after the initial analysis of factorial data
`as they cover both high and low levels of the most significant
`factors. Samples quenched at 2 h and 24 h were assayed by
`both RP-HPLC and enzyme assay to demonstrate parity
`between methods. Average errors by STDEV for RP-HPLC
`data (factorial and repeats combined) were 6.36% for
`samples quenched at 2 h, and 4.5% for samples quenched at
`24 h.
`All of the factorial data for each response (yield at 2 h or
`yield at 24 h), including averaged data from replicates, were
`used to calculate a single confidence interval
`for that
`response. This was used to determine if the calculated effects
`were significant at a 95% threshold
`
`Enzyme Assay
`
`Quenched sample(67 mL) was diluted in 933 mL 50 mM Tris
`1 mM EDTA. Diluted sample (100 mL) was diluted in 2.9 mL
`
`Table III. Levels used in factorial experiments.
`Upper (þ)
`Lower ()
`
`Factor
`
`Reb
`Rem
`Q
`R
`G
`
`576
`2,000
`1
`1.23 mL min
`2:1
`1.2 M
`
`3,836
`0
`1
`0.115 mL min
`0.17:1
`0.533 M
`
`Midpoint (0)
`
`2,206
`1,000
`1
`0.669 mL min
`1.085:1
`0.866 M
`
`screening experiments.
`initial
`Values were set on the basis of
`Reb bulk impeller Reynolds number, Rem mini-impeller Reynolds
`1), R, Redox ratio, G, Final GdHCl
`number, Q injection rate (mL min
`concentration (M).
`
`DOI 10.1002/bit
`
`Page 4
`
`

`
`Table IV. Compositions of buffers used for factorial experiments.
`
`[Tris HCl] (mM)
`
`[Cystamine] (mM)
`
`[EDTA] (mM)
`
`Buffer
`
`1
`2
`3
`4
`5
`
`Redox ratio
`2:1 (þ)
`50
`4
`1
`2:1 (þ)
`50
`4
`1
`0.17:1 ()
`50
`25
`1
`0.17:1 ()
`50
`25
`1
`50
`5.67
`1.085:1 (0)
`1
`(þ) indicates upper level of factor, () indicates lower level of factor, (0) indicates the midpoint.
`
`[GdHCl] (M)
`
`0.717
`0
`0.717
`0
`0.358
`
`Final [GdHCl] (M)
`1.2 (þ)
`0.533 ()
`1.2 (þ)
`0.533 ()
`0.867 (0)
`
`pH
`
`8.0
`8.0
`8.0
`8.0
`8.0
`
`of Micrococcus lysodeikticus cell suspension (0.3 mg/mL in
`100 mM potassium phosphate buffer pH 7). The change in
`absorbance at 450 nm was tracked over 1.5 min, after a delay
`of 30 s to allow equilibration. Yields were calculated from
`specific activity in each of the buffers used.
`
`Batch Refolding Tests for Yield and Throughput at
`Various Redox-Ratio and GdHCl Concentrations
`
`suggested that high
`study
`fed-batch factorial
`The
`refolding yields could be achieved even with rapid injection
`and poor mixing efficiency. Comparative studies of the
`influence of two key parameters, redox ratio and GdHCl
`concentration, were therefore carried out for a simple
`batch refolding system. In this system, the final yields
`obtained for each condition, and the time taken to
`reach these yields, were both determined. Each refold was
`left for a total of 24 h to allow the reaction to proceed to
`practical completion. For the purposes of subsequent
`analysis, the maximum practicable yield achieved under
`each condition was assumed to be equivalent
`to that
`measured at 24 h.
`1) was
`Denatured and reduced lysozyme (15.9 mg mL
`refolded by 1:15 dilution into a series of buffers of varying
`GdHCl concentrations and redox ratios. The denatured
`protein was added in a single step and distributed by a
`magnetic stirrer for 5 min to ensure homogeneity. Samples
`(2 mL) were taken at 1, 2, 4, 8, and 24 h and quenched with
`200 mL of 10 % (v/v) TFA. Nine buffers were prepared as
`detailed in Table V.
`The maximum yield and the time to achieve maximum
`yield were used to produce contour plots. Exponential fits to
`
`the data using Eq. (1), were used to predict the time taken to
`achieve specific yields.
`
`y ¼ að1 ebtÞ
`
`(1)
`
`where a and b are constants, y is yield and t is refold time in
`hours.
`Contour plots were generated from the data for both yield
`and time to yield. The experimental data generated was
`fitted using the following expression so as to account for the
`possible effects and interaction between redox ratio and
`guanidine hydrochloride concentration (Eq. (2)).
`z ¼ b1 þ b2R þ b3G þ b4R2 þ b5G2 þ b6RG
`
`(2)
`
`where z represents yield or time to yield, R is redox ratio, G is
`guanidine concentration. The constants b1–b6 were deter-
`mined by multiple regression.
`the effect of
`The resulting contour plots predict
`combinations of operating conditions. They were used in
`this study to generate windows of operation to examine the
`trade-offs between the required levels of yield and the
`associated process time required to achieve these.
`
`Results & Discussion
`
`Effect of Guanidine Hydrochloride
`Concentration on Refolding Yields
`
`Of the five factors used in the factorial study, appropriate
`upper and lower levels of final GdHCl concentration were
`determined by experiment. Final chaotrope concentration
`
`Table V. Compositions of buffers used for windows of operation experiments.
`
`Buffer
`
`[Tris.HCl] (mM)
`
`[Cystamine] (mM)
`
`Redox ratio
`
`[EDTA] (mM)
`
`[GdHCl] (M)
`
`Final [GdHCl] (M)
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`
`50
`50
`50
`50
`50
`50
`50
`50
`50
`
`4
`4
`4
`5.67
`5.67
`5.67
`25
`25
`25
`
`2:1
`2:1
`2:1
`1.085:1
`1.085:1
`1.085:1
`0.17:1
`0.17:1
`0.17:1
`
`1
`1
`1
`1
`1
`1
`1
`1
`1
`
`0
`0.358
`0.717
`0
`0.358
`0.717
`0
`0.358
`0.717
`
`0.533
`0.867
`1.2
`0.533
`0.867
`1.2
`0.533
`0.867
`1.2
`
`pH
`
`8.0
`8.0
`8.0
`8.0
`8.0
`8.0
`8.0
`8.0
`8.0
`
`Mannall et al.: Factors Affecting Protein Refolding
`
`1527
`
`Biotechnology and Bioengineering. DOI 10.1002/bit
`
`Page 5
`
`

`
`in the reactor has been shown to affect yields in the refolding
`of lysozyme (De Bernardez Clark et al., 1998; Hevehan
`and De Bernardez Clark, 1997; Katoh et al., 1999; Lee et al.,
`2002). It is believed that high concentrations of GdHCl
`decrease both the rate of refolding and aggregation, but that
`aggregation is the more severely affected (De Bernardez
`Clark et al., 1998). Therefore, denaturant concentration
`must be low enough to permit protein molecules to
`refold, but high enough to prevent protein aggregation and
`to promote protein flexibility, which is required for
`structural reorganization (Tsumoto et al., 2004). An optimal
`concentration therefore exists where yields are maximized
`without significantly reducing the rate of refolding (De
`Bernardez Clark et al., 1998).
`In setting the limits for final GdHCl concentration in
`the refold buffer,
`it is evident that an upper limit of
`approximately 4.2 M exists. This is defined by the C0.5 of
`lysozyme in GdHCl (the concentration of chaotrope at
`which half the protein is folded) (Ahmad and Bigelow,
`1982). This was measured at pH 7 rather than pH 8 (used in
`this experiment) and is therefore only an approximation of
`the C0.5 under conditions used in these experiments. The
`yields of refolded lysozyme in a batch system at various final
`GdHCl concentrations can be seen in Figure 2, at 2 h
`(quenched) and 24 h (unquenched) after the start of the
`refold step. The optimal final GdHCl concentration is
`1.2 0.1 M, which is in close agreement with the 1.3 M
`optimum found by others (De Bernardez Clark et al.,
`1998). The slight difference might be accounted for by the
`difference in redox ratios used (2:1 here versus approxi-
`mately 1.33:1 in previous work) and the use of cystamine in
`place of oxidized glutahthione. As discussed by these
`authors,
`the increasing yield up to 1.2 M GdHCl
`is
`predominantly due to a decreasing rate of aggregation, while
`
`the lower yields achieved above 1.2 M GdHCl concentra-
`tions are more likely due to the considerably slowed
`refolding rates. Extrapolation of refolding kinetic data from
`our labs suggests it may take 3 days for the reaction to go to
`completion at a final GdHCl concentration of 2.36 M (data
`not shown). In reactions where the GdHCl concentration
`was above the optimal 1.2 M, no aggregation was visible,
`although a series of peaks prior to the native peak were
`found on RP-HPLC analysis as observed in Figure 3. The
`combined areas of these peaks are presented in Figure 4 and
`most likely represent soluble misfolded protein, formed
`by incorrect disulfide pairings and may include mixed
`disulfides, though the possibility of these peaks representing
`soluble aggregates cannot be discounted as the methods
`used here cannot identify the true identity of these species.
`Activity measurements would likewise not be able to
`identify these species. The amount of these species appeared
`to increase with GdHCl concentration, with a sharp increase
`at 1.44 M GdHCl. Refolding of protein above 1.2 M
`GdHCl may be so slow that
`these incorrectly folded
`forms can accumulate. Comparison of RP-HPLC and
`activity measurement (Buswell and Middelberg, 2002)
`suggest that activity is associated with a non native peak.
`This is supported by our comparison of assay methods
`detailed in Figure 5A and B. Typically yields are
`approximately 10% higher by assay, this additional amount
`may account
`for non-native protein forms possessing
`activity.
`The unconventional redox system used in this study may
`also have an impact on the formation of incorrectly-folded
`proteins. The refolding buffer initially contains only the
`oxidized form of the redox reagent (cystamine), whereas the
`denatured protein contains only an excess of reducing agent
`(DTT). High levels of the oxidizing species present during
`
`Figure 2.
`Effect of GdHCl concentration upon refolding yield of lysozyme at a
`final concentration of 1.1 mg mL1, quenched at 2 h (*) and 24 h (*). The optimal
`GdHCl concentration appears to be 1.2 M with any further increase in GdHCl
`concentration reducing refolding yields.
`
`Figure 3. RP-HPLC trace, showing the retention of native protein (8.57 min) and
`misfolded species (6–8 min), presumed to be disulfide variants of the native
`structure.
`
`1528
`
`Biotechnology and Bioengineering, Vol. 97, No. 6, August 15, 2007
`
`DOI 10.1002/bit
`
`Page 6
`
`

`
`Figure 4.
`Effect of GdHCl concentration on the amount of misfolded protein
`species from the refolding reaction quenched at 2 h. The amount increases as the
`concentration of GdHCl increases with a sharp increase at 1.44 M.
`
`the initial stages of injection may cause rapid oxidation of
`free thiols, resulting in a greater number of
`incorrect
`disulfide-bond pairings than when using a conventional
`refolding buffer containing initially both redox components.
`The magnitude of this effect is expected to be inversely
`proportional
`to injection rate with the formation of
`misformed disulfides favored at slower injection rates.
`
`Factorial Experiments
`
`Results from the factorial experiments are detailed in
`Table II and were used to calculate the impact of the factors
`upon refolding yields, which are summarized in Figures 6
`and 7. Discussion of these effects first considers significant
`single factor effects and then the most significant interac-
`tions.
`
`Major Effects and Interactions
`
`Final GdHCl Concentration
`
`Figures 6 and 7 show that final GdHCl concentration
`(G) influences the refolding yield measured at both 2 h and
`24 h, in a similar fashion to that noted in previous studies
`(Katoh et al., 1999), where
`increases
`in chaotrope
`concentration raised the refolding yield. High concentra-
`tions (1.2 M) of GdHCl will prevent aggregation of the
`protein, leading to greater amounts of soluble protein being
`
`Figure 5. A: Parity plot for enzyme assay versus HPLC for samples quenched at
`2h. B: Parity plot for enzyme assay versus HPLC for samples quenched at 24 h.
`
`Figure 6.
`Effect of factors for reaction samples quenched at 2 h, results are
`shown for a 95% confidence interval. Significant single factors GdHCl concentration G
`and redox ratio R. Significant two factor interactions were found between GdHCl
`concentration G and redox ratio R, and between injection rate Q and guanidine
`concentration G. None of the factors/interactions appear to be more significant than
`others. Bars indicate 95% confidence interval.
`
`Mannall et al.: Factors Affecting Protein Refolding
`
`1529
`
`Biotechnology and Bioengineering. DOI 10.1002/bit
`
`Page 7
`
`

`
`redox ratio (R) of reduced to oxidized species of the redox
`reagent, has a significant effect on refolding yield at both
`2 and 24 h (Figs. 6 and 7), even though the redox ratio
`changes considerably over the injection period. The two
`redox species act as a redox pair, allowing the shuffling of
`disulfides until the correct pairings are achieved. An optimal
`ratio is required to allow efficient disulfide shuffling. At
`low redox ratios (0.17:1), the concentration of cystamine
`(oxidized species) is too high, particularly in the initial
`injection period, to allow efficient disulfide exchange. This
`may lead to the rapid formation and trapping of incorrectly
`disulfide-paired proteins and reduced final yields. The effect
`of redox ratio is most obvious at high GdHCl concentrations
`(1.2 M), and an interaction between these two factors
`(RG) is confirmed in Figures 6 and 7. The lower rate of
`aggregation at high GdHCl affords the redox ratio present
`more time to promote the formation of correctly paired
`cysteines and hence the formation of native protein. By
`contrast, the rapid aggregation of protein at the low GdHCl
`concentration (0.533 M) decreases the time for which the
`effect of redox ratio can act and hence also decreases its
`impact upon refolding yields.
`
`Rate of Injection
`
`It has been shown previously that refolding yields can be
`improved by decreasing the rate of
`injection of
`the
`denatured protein (Katoh et al., 1999; Katoh and Katoh,
`2000). The formation of native protein is favored at low
`protein concentrations (Buwell et al., 2002) where aggrega-
`tion of folding intermediates is slower than refolding to the
`native protein. Similarly, the gradual addition of denatured
`protein to refolding buffer results in improved yields due to
`the presence of a lower concentration of partially refolded
`protein over the duration of the addition (Katoh et al.,
`1999).
`it is observed that increased
`From Figures 6 and 7,
`injection rates (Q) have a significant negative effect upon the
`refolding yields measured at 24 h, but has a positive effect at
`2 h. There is a considerable difference between the injection
`conditions used. Under a slow injection rate (0.115 mL
`1) complete addition takes just under 2 hours (1 h
`min
`1)
`55 min), whilst under a fast injection rate (1.23 mL min
`complete addition takes just over 10 min. This means that
`under fast injection conditions, the protein is at the final
`protein concentration for about 1 h 50 min before the 2 h
`sample is taken, whereas under slow injection conditions it is
`at this concentration for only 5 min. Therefore, at the 2-h
`time point, the fast injection rate is likely to give a greater
`yield than the slow injection rate, given the greater time
`available to finish refolding, after complete addition of the
`protein. However,
`increased injection rates have a sig-
`nificant negative effect upon yield at 24 h. Slow injection of
`the denatured protein minimizes the effective concentration
`of partially renatured intermediates that are present at any
`specific time. Therefore, the denatured protein entering the
`system is less likely to collide to form aggregates. It is
`
`DOI 10.1002/bit
`
`Figure 7.
`Effect of factors for reaction samples quenched at 24 h, results are
`shown