www.elsevier.com/locate/procbio

Process Biochemistry 42 (2007) 1259–1263

Short communication

Activity of recombinant GST in Escherichia coli grown

on glucose and glycerol

Rasmus Hansen 1, Niels Thomas Eriksen *

Department of Biotechnology, Chemistry, and Environmental Engineering, Aalborg University, Sohngaardsholmsvej 49, DK-9000 Aalborg, Denmark

Received 21 November 2006; received in revised form 2 May 2007; accepted 29 May 2007

Abstract

We have investigated production, solubility and activity of recombinant glutathione-S-transferase (GST) expressed in Escherichia coli BL21

grown in defined media with glucose or glycerol as carbon source. GSTwas predominantly expressed as a soluble protein on both carbon sources,

and 83–84% was found in the supernatant after cell lysis. In cultures grown on glucose, only 32 � 9% of the GST was active, while 76 � 13% of

the GST was active in cultures grown on glycerol. This shows that glycerol has the potential to increase the activity of soluble GST in E. coli

cultures in vivo.

# 2007 Elsevier Ltd. All rights reserved.

Keywords: E. coli; Glutathione-S-transferase; Glycerol; Activity

1. Introduction

Escherichia coli is the most widely used host for

expression of recombinant proteins at laboratory scale.

However, different recombinant proteins are produced with

variable yields and often they do not fold into active proteins

[1]. Yields of active, correctly folded recombinant proteins

are often improved by genetically linking the target protein

to a fusion partner that is expressed with a high yield in a

soluble form [2]. Glutathione-S-transferase (GST) is one of

the most commonly used, commercially available fusion

partners, which also allows protein quantification using GST

activity assays or GST specific antibodies, and purification

by affinity chromatography. Although GST supposedly

should increase the yields of active, soluble proteins, some

GST fusion proteins still form insoluble inclusion bodies

inside E. coli [3].

Folding of recombinant proteins may also be affected by the

medium in which the cells are grown. Organic osmolytes,

including glycerol have a stabilising effect on proteins in vitro

[4,5] and have been described as chemical chaperones [6].

* Corresponding author. Tel.: +45 96 35 84 65; fax: +45 98 14 18 08.

E-mail address: nte@bio.aau.dk (N.T. Eriksen).1 Present address: Faculty of Life Sciences, University of Manchester, Oxford

Road M13 9PT, UK.

1359-5113/$ – see front matter # 2007 Elsevier Ltd. All rights reserved.

doi:10.1016/j.procbio.2007.05.022

Addition of glycerol to growth media has increased the

solubility of recombinant proteins in E. coli [7,8] and in human

cell cultures in vivo [9]. Positive effects on protein solubility in

E. coli have also been observed in growth media supplemented

with sucrose [10], glycyl betaine [11] or proline [12]. As a

result of the increased solubility, also the activity of

recombinant proteins is increased by these osmolytes. In this

paper, we have compared activities of the common fusion

partner GST, which is expressed as a soluble protein also in the

absence of protein stabilising osmolytes, when E. coli cultures

are grown with glucose or glycerol as carbon sources. We also

investigated the effect of glycerol on GST fused to the ligand-

binding domain of mouse peroxisome proliferator-activated

receptor a, a fusion protein which forms insoluble inclusion

bodies [13].

2. Material and methods

2.1. Strains and constructs

E. coli BL21 was obtained from Novagen (Madison, USA). One strain was

transformed by the commercially available expression vector pGEX 5X-1

(Amersham Pharmacia Biotech) encoding glutathione S-transferase under

control of the tac promoter and b-lactamase. In a second transformed strain,

the gene encoding the ligand-binding domain (LBD) of mouse peroxisome

proliferator-activated receptor a (PPARa) was cloned into the pGEX 5X-1

vector to encode a fusion protein, GST-PPARa LBD. This construct was

denoted pGST-PPARa LBD [14].

Fig. 1. Concentrations of biomass (*), glucose (&), acetic acid (^), total GST

(~) and active GST (!), in batch culture of E. coli BL21 pGEX 5X-1 grown

with glucose as carbon source. The culture was induced at 4 h by addition of

0.1 mM IPTG, as indicated by the arrow.

R. Hansen, N.T. Eriksen / Process Biochemistry 42 (2007) 1259–12631260

2.2. Growth, induction and on-line monitoring

Inocula were prepared from transformed cells stored at �80 8C. The cells

were grown overnight at 30 8C on solid LB medium (10 g tryptone L�1, 5 g

NaCl L�1, 15 g agar L�1) containing 100 mg ampicillin L�1. One colony was

transferred to 5 mL liquid LB medium containing 500 mg ampicillin L�1 and

incubated overnight at 30 8C. This culture was introduced into 100 mL of

defined growth medium containing 200 mg ampicillin L�1 and 5 g L�1 of the

same carbon source as used in growth experiments, and grown overnight at

30 8C before it was introduced into a 3 L Applikon BTS 05 bioreactor contain-

ing 2.5 L of defined medium [13] containing 10 g L�1 glucose or glycerol as

final cell density limiting substrate, 4 g L�1 (NH4)2SO4, 100 mg ampicillin L�1

and 0.5 mL L�1 of anti-foam A (Sigma–Aldrich). The bioreactor was equipped

with a Pt100 temperature sensor and autoclavable pH and oxygen electrodes

(Mettler Toledo). Temperature was kept at 30 8C to enhance correct folding of

the produced proteins [15], pH was maintained at 7.0 by titration with pulses of

1 M NaOH, and the aeration rate was 1.5 L min�1.

After an initial growth phase, the biomass reached a predetermined con-

centration (0.3 g L�1 of dry biomass) estimated on-line from the integrated

amount of added NaOH, and protein expression was induced by automatic

addition of IPTG to a final concentration of 0.1 mM from a cooled reservoir

[13]. Approximately three cell divisions were obtained during the protein

production phase before the carbon source was exhausted.

2.3. Biomass determinations

Biomass concentration was followed by apparent absorbance (OD) at

600 nm, and when necessary diluted in sterile growth medium to values below

0.4. OD measurements were compared to dry weight (DW) of biomass

measured after filtration onto pre-dried and pre-weighed 0.22 mm Millipore

filters and drying at 105 8C for 24 h.

2.4. Organic substrate and product analyses

Concentrations of glucose, glycerol and acetic acid were quantified using

HPLC: 50 mL of 0.22 mm filtered culture supernatant were added on an Aminex

HPX-87H column (Bio-rad) eluted with 0.4 mL min�1 of 0.5 mM H2SO4 at

30 8C. Detection was performed by a Knauer K-2300 refractive index detector.

2.5. Analyses of GST and GST-PPARa LBD

Cells were harvested by centrifugation and resuspended in 1 mL PBS buffer,

pH 7.5 containing 1 g lysozyme (Sigma–Aldrich) L�1, Protease Inhibitor

Cocktail (Sigma–Aldrich), 10 mM MgCl and 50 U mL�1 benzonase (Merck).

The cells were lysed by five repeated freeze–thaw cycles in liquid nitrogen and

at 30 8C, and the cell lysates were used for analyses of total GST and GST-

PPARa LBD contents. Centrifugation of the cell lysate at 10,000 � g for 20 min

separated GST and GST-PPARa LBD into a soluble fraction in the supernatant

and an insoluble fraction in the pellet.

GST and GST-PPARa LBD were purified from 600 mL of cell lysates,

which were loaded into MicroSpin columns containing Glutathione-Sepharose

beads (Amersham Pharmacia Biotech). Proteins were allowed to bind for 1 h at

4 8C and then washed twice with 600 mL PBS. Bound proteins were eluted with

200 mL, 10 mM reduced glutathione by centrifugation at 1000 � g for 1 min.

The purity of GST or GST-PPARa LBD was verified by SDS-PAGE. Protein

concentrations were estimated by the Bradford method [16] using bovine serum

albumin as reference.

Activities of the produced proteins in crude cells lysates or after purification

were measured by the GST Detection Module Assay (Amersham Pharmacia

Biotech), in which a conjugate between 1-chloro-2,4-dinitrobenzen and glu-

tathione is detected spectrophotometrically at 340 nm using an extinction

coefficient of 9.6 mM�1 cm�1.

Specific GST and GST-PPARa LBD concentrations in cell lysates were

estimated after separation of proteins on 12% (w/v) SDS-PAGE: 10 mL of cell

lysate were diluted with 10 mL of buffer pH 6.8 containing 0.2 M Tris, 2% (w/v)

SDS, 0.2% (w/v) bromophenol blue, 20% (v/v) glycerol and 40 mM DTT and

boiled for 5 min. After centrifugation, 10 mL were loaded onto the gel along

with three known amounts of purified GST or GST-PPARa LBD used as

standards. The gels were stained by Coomassie Brilliant Blue. The staining

intensities of 28 kDa (GST) or 60 kDa (GST-PPARa LBD) protein bands were

used to estimate the amounts of GST or GST-PPARa LBD loaded on the gel by

comparison to staining intensities of purified GST or GST-PPARa LBD

standards loaded on the same gel. The staining intensities were measured by

Totallab 1.11 image analysis software (Phoretics).

3. Results

Fig. 1 shows an example of a batch culture of E. coli BL21

pGEX 5X-1 grown on glucose as the growth limiting substrate

and induced by IPTG at a cell density of 0.3 g L�1. The

consumption of glucose correlated with the increase in cell

density, and also acetic acid accumulated and reached 0.4 g L�1

before it was re-metabolised upon glucose depletion. Acetic

acid was not detected in cultures grown on glycerol. Cultures

expressing GST-PPARa LBD showed similar growth char-

acteristics.

After inducer addition, the cells accumulated recombinant

GST (Fig. 1) or GST-PPARa LBD, which became the most

abundant proteins in the cells. GST was observed as strongly

stained protein bands located at 28 kDa on SDS-PAGE gels

(Fig. 2A), corresponding to the migration of purified GST.

Fig. 2B shows the standard curve between staining intensity and

concentration of purified GST from the gel in Fig. 2A, used to

estimate the concentration of total GST in each sample of cell

lysate loaded on the gel. Extracts of induced cultures of E. coli

BL21 pGST-PPARa LBD and purified GST-PPARa LBD both

showed strongly stained protein bands at 60 kDa.

GST activities in extracts of lysed cells also increased after

inducer addition. From these activities, the concentrations of

active GST (Fig. 1) or GST-PPARa LBD were estimated by

comparison to specific activities of purified GSTor GST-PPARa

LBD. The specific activity of purified GST, estimated from linear

regression analysis (Fig. 2B), was 4.7� 0.1 mmol min�1 g�1,

corresponding to a molar specific activity (turnover rate) of

138� 3 min�1. For purified GST-PPARa LBD, the specific

Fig. 2. (A) SDS-PAGE analysis of whole cell lysates from E. coli BL21 pGEX

5X-1 grown on glucose. The 28 kDa marker indicates the expected molecular

mass of GST. Lane 1: molecular weight markers, lanes 2–4: purified GST in

amounts of 0.25, 0.5 and 1.0 mg, respectively, lanes 5–9: whole cell lysates

harvested after 3, 5, 7, 8 and 9 h, respectively. The culture was induced at 4 h.

Lane 10: whole cell lysate of non-transformed E. coli BL21. (B) Activity of

purified GST (*) and purified GST-PPARa LBD (&) as function of the

concentration of purified protein and staining intensity of 28 kDa protein

bands of purified GST from Fig. 2A (^) as function of the concentration

of purified GST. Slopes of regression lines, 4.7 � 0.1 (�S.E.) and

2.6 � 0.1 mmol min�1 g�1, correspond to specific GST and GST-PPARa

LBD activities, respectively, and 3352 � 205 (�95% confidence limit-

s) L mg�1 used to convert staining intensities of 28 kDa protein bands

(Fig. 2A) into total GST concentrations (Fig. 1). Inset shows SDS-PAGE

analysis of cell lysates from E. coli BL21 pGEX 5X-1 and E. coli BL21 pGST-

PPARa LBD. The 28 and 60 kDa markers indicate the expected molecular

masses of GSTand GST-PPARa LBD, respectively. Lanes 1–3: cells harvested

after 2–3 h of induction, lane 1: whole cell lysate, lane 2: supernatant after

centrifugation of lysed cells, lane 3: resuspended pellet, lanes 4–6: cells

harvested at the end of the growth phase after 11–13 h of induction, lane 4:

whole cell lysate, lane 5: supernatant after centrifugation of lysed cells and

lane 6: resuspended pellet.

Fig. 3. (A) Specific concentrations of total GST in two induced cultures of E.

coli BL21 pGEX 5X-1 grown on glucose (*, &), two not induced cultures

grown on glucose (!), three induced cultures grown on glycerol (*,&,5) and

two not induced cultures grown on glycerol (~). (B) Specific concentrations of

active GST in two induced cultures of E. coli BL21 pGEX 5X-1 grown on

glucose (*, &), two not induced cultures grown on glucose (!), three induced

cultures grown on glycerol (*, &,5) and two not induced cultures grown on

glycerol (~). (C) Relationships between specific concentrations of active and

total GST in two induced cultures of E. coli BL21 pGEX 5X-1 grown on glucose

(*, &) and three induced cultures grown on glycerol (*, &, 5). Slopes of

regression lines (0.32 � 0.9 (�S.E.) and 0.76 � 0.13 for cultures grown on

glucose or glycerol, respectively) indicate fractions of active GST.

R. Hansen, N.T. Eriksen / Process Biochemistry 42 (2007) 1259–1263 1261

activity was 2.6� 0.1 mmol min�1 g�1 and the turnover rate

was 156� 8 min�1.

GST was expressed as a soluble protein found predomi-

nantly in the supernatant after centrifugation of lysed cells

(Fig. 2B, inset). The soluble fractions of GST were similar on

both carbon sources, 83 � 6% on glucose and 84 � 13% on

glycerol. GST-PPARa LBD was expressed as an insoluble

protein associated with the pellet (Fig. 2B, inset). This pattern

was observed shortly after inducer addition as well as at the end

of the growth phase.

Fig. 3A shows the specific concentrations of total GST in

cultures of E. coli BL21 pGEX 5X-1 grown on glucose,

compared to cultures grown with glycerol as the carbon source,

and cultures which were not induced. The non-induced cultures

grown on both carbon sources contained 1–2 mg g�1 of protein

with a molecular mass of 28 kDa similar to GST. Induced

cultures accumulated 70–140 mg g�1 of GST. There were no

clear differences in the specific GST concentrations between

cultures grown on glucose or glycerol.

The activity of GST in non-induced cultures corresponded to

a specific GST concentration of 0.4–1.8 mg g�1 on both carbon

Fig. 4. Specific concentrations of active GST-PPARa LBD as function of total

GST-PPARa LBD in two induced cultures of E. coli BL21 pGST-PPARa LBD

grown on glucose (*, &) and one induced culture grown on glycerol (*).

Slope of regression line = 0.96 � 0.07 (�S.E.) indicate the fraction of active

GST-PPARa LBD.

R. Hansen, N.T. Eriksen / Process Biochemistry 42 (2007) 1259–12631262

sources. This was similar to the specific concentration of non-

induced GST detected by SDS-PAGE. Addition of IPTG

resulted in increased GST activity. The final specific

concentrations of active GST reached approximately

25 mg g�1 in cultures grown on glucose (Fig. 3B). In

comparison, induced cultures grown on glycerol accumulated

considerably higher specific GST activities reaching 55–

120 mg g�1 of active GST. Fig. 3C shows the relationship

between activity and total specific concentration of GST in

induced cells. The GST produced in cells grown of glycerol was

significantly more active than the GST produced from cells

grown on glucose (t-test, P = 0.02). From linear regression

analysis, it was estimated that only 32 � 9% of the GST

produced in cells grown on glucose was active while 76 � 13%

of the GST produced in cells grown on glycerol was active.

In induced E. coli BL21 pGST-PPARa LBD, the specific

concentration of total GST-PPARa LBD reached 90–

121 mg g�1 with no clear differences in expression level

between cultures grown on glucose or glycerol. In contrast to

cultures expressing free GST, also the specific activities of

GST-PPARa LBD were similar in cultures grown on both

carbon sources, and 96 � 7% of the GST-PPARa LBD was

expressed as active protein (Fig. 4).

4. Discussion

Glycerol was able to enhance the activity of recombinant

GST in E. coli B21 and this enhancement was neither the result

of differences in concentration nor in solubility of the protein.

The specific concentrations of total GST, which accumulated in

cultures grown on glucose or glycerol, were similar, and 83–

84% of the GST was soluble in all cultures. Still, the fraction of

GST that was expressed as an active protein was 2.4 times

higher in cultures grown on glycerol compared to cultures

grown on glucose. High activities of recombinant proteins are

generally important. The activity of GST is particularly

important when GST is used as a fusion partner since activity

assays and affinity chromatographic methods for quantification

and purification of the fusion proteins depend on the GST

activity.

Purification of GST or GST-PPARa LBD was based on

affinity chromatography in which the active site of GST binds

to immobilised glutathione. Only active proteins are retained by

this method [3], and the purified products, therefore, contained

only active GST or GST-PPARa LBD. The turnover rates of

purified GST and GST-PPARa LBD were similar (138 � 3 and

156 � 8 min�1, respectively). This supports that all purified

proteins were active, and that the specific activities of fully

active GST or GST-PPARa LBD are accurately described by

the relationships between activity and concentration of purified

proteins in Fig. 2B. The similar turnover rates also indicate that

the GST domain had a similar conformational stage when it was

expressed by its own or fused to PPARa LBD.

Expression of GST or GST-PPARa LBD were induced

relatively early in the exponential growth phase. This allowed

the cultures to pass through more than three cell generations

while expressing their recombinant proteins. These relatively

long protein expression phases were chosen to better allow

possible differences between cultures grown on glycerol

relative to cultures grown on glucose to be revealed. In

Fig. 1, it can be seen that total as well as active GST

accumulated in the cultures until the end of the growth phase, at

which stage all the glucose and also the acetic acid, which had

previously been excreted by the cells, were depleted. The

specific concentration of GST rapidly increased upon IPTG

addition, but levelled out at plateaus of 60–120 mg g�1

(Fig. 3B). The specific rates of GST expression have, therefore,

remained constant from the time of induction until the depletion

of either glucose or glycerol. A similar pattern in protein

accumulation was also observed in cultures expressing GST-

PPARa LBD.

Almost all GST-PPARa LBD was expressed as active

protein in cultures grown on glucose and no further increase in

the activity was observed in cultures grown on glycerol (Fig. 4).

Therefore, the effect of glycerol was not to increase the

turnover rate of already active proteins. Instead, glycerol must

have turned non-active GST into active GST. Interestingly, even

though GST-PPARa LBD was more active than GST in vivo,

the former protein was predominantly expressed as insoluble

inclusion bodies. Although inclusion bodies composed of

active proteins have also been observed in other studies [17],

they are generally considered to be aggregates of misfolded and

inactive proteins. It, therefore, seems likely that GST-PPARa

LBD aggregated in a semi-soluble form, in which only the

PPARa LBD domain of the fusion protein was insoluble.

Lower solubility of recombinant proteins in high cell density

cultures of E. coli grown on glucose compared to glycerol has

previously been linked to the accumulation of up to 10–

12 g L�1 of acetic acid when glucose was the carbon source [7].

E. coli BL21 do not produce acetic acid when grown on

glycerol at 30 8C [18]. Accordingly, we detected acetic acid

only in cultures grown on glucose, but only at concentrations up

R. Hansen, N.T. Eriksen / Process Biochemistry 42 (2007) 1259–1263 1263

to 0.4–0.6 g L�1. However, it is unlikely that acetic acid at such

low concentrations was the cause of the lower activity of GST in

cultures grown on glucose (Fig. 3). The concentrations of acetic

acid in our cultures were well below the approximately 5 g L�1,

at which point acetic acid becomes toxic to E. coli at pH 7 [19],

and GST-PPARa LBD was fully active in cultures grown on

glucose despite the accumulation of acetic acid.

The positive effects of using osmolytes, such as glycerol, in

growth media for E. coli, may be particular beneficial when

recombinant proteins are over-expressed, reach high intracel-

lular concentrations and cause molecular crowding in the

cytoplasm. This may challenge the capacity of the cells own

chaperones [20] and the fidelity of the protein folding reactions

[21]. GST and some GST fusion proteins are expressed so well

in E. coli that they become the dominant proteins in the cells.

High concentrations of these proteins and molecular crowding

in the cytoplasm are, therefore, inherently associated to GST

when it is expressed by itself or as a fusion partner. The results

presented in this paper suggest that glycerol has the potential to

increase the activity of over-expressed GST in vivo, even when

the proteins are produced in soluble forms on other substrates

than glycerol.

Acknowledgement

We thank Dr. Karsten Kristiansen, University of Southern

Denmark for supplying the bacterial strains.

References

[1] Villaverde A, Carrio MM. Protein aggregation in recombinant bacteria:

biological role of inclusion bodies. Biotechnol Lett 2003;25:1385–95.

[2] Hannig G, Makrides SC. Strategies for optimising heterologous protein

expression in Escherichia coli. Trends Biotechnol 1998;16:54–60.

[3] Frangioni JV, Neel BG. Solubilization and purification of enzymatically

active glutathione S-transferase (pGEX) fusion proteins. Anal Biochem

1993;210:179–87.

[4] Gekko K, Timasheff SN. Mechanism of protein stabilization by glycerol:

preferential hydration in glycerol-water mixtures. Biochemistry 1981;20:

4667–76.

[5] Sawano H, Koumoto Y, Ohta K, Sasaki Y, Segewa S, Tachibana H.

Efficient in vitro folding of the three-disulfide derivatives of hen lysozyme

in the presence of glycerol. FEBS Lett 1992;303:11–4.

[6] Yang D-S, Yip CM, Huang THJ, Chakrabartty A, Fraser PE. Manipulating

the amyloid-b aggregation pathway with chemical chaperones. J Biol

Chem 1999;274:32970–4.

[7] Strandberg L, Enfors S-O. Factors influencing inclusion body formation in

the production of a fused protein in Escherichia coli. Appl Environ

Microbiol 1991;57:1669–74.

[8] Leandro P, Lechner MC, de Almeida IT, Konecki D. Glycerol increases

the yield and activity of human phenylalanine hydroxylase mutant

enzymes produced in a prokaryotic expression system. Mol Gen Metabol

2001;73:173–8.

[9] Sato S, Ward CL, Krouse ME, Wine JJ, Kopito RR. Glycerol reverses the

misfolding phenotype of the most common cystic fibrosis mutation. J Biol

Chem 1996;271:635–8.

[10] Bowden GA, Georgiou G. Folding and aggregation of b-lactamase in

the periplasmic space of Escherichia coli. J Biol Chem 1990;265:

16760–6.

[11] Blackwell JR, Horgan R. A novel strategy for production of a highly

expressed recombinant protein in an active form. FEBS Lett 1991;295:

10–2.

[12] Ignatova Z, Gierasch LM. Inhibition of protein aggregation in vitro and in

vivo by a natural osmoprotectant. Proc Natl Acad Sci 2006;103:13357–61.

[13] Eriksen NT, Kratchmarova I, Neve S, Kristiansen K, Iversen JJL. Auto-

matic inducer addition and harvesting of recombinant Escherichia coli

cultures based on indirect on-line estimation of biomass concentration and

specific growth rate. Biotechnol Bioeng 2001;75:355–61.

[14] Kussman-Gerber S, Kratchmarova I, Mandrup S, Kristiansen K. Micro-

affinity columns for analysis of protein–protein interactions. Anal Bio-

chem 1999;271:102–5.

[15] Schein CH, Noteborn MHM. Formation of soluble recombinant proteins

in Escherichia coli is favored by lower growth temperature. Bio/Technol

1988;6:291–4, http://www.nature.com/nbt/journal/v6/n3/abs/nbt0388-

291.html.

[16] Bradford MM. A rapid and sensitive method or the quantification of

microgram quantities of protein utilizing the principle of protein dye

binding. Anal Biochem 1976;72:248–54.

[17] Garcıa-Fruitos E, Gonzalez-Montalban N, Morell M, Vera A, Ferraz RM,

Arıs A, et al. Aggregation as bacterial inclusion bodies does not imply

inactivation of enzymes and fluorescent probes. Microb Cell Factories

2005;4:27.

[18] Christensen ML, Eriksen NT. Growth and proton exchange in recombinant

Escherichia coli BL21. Enzyme Microb Technol 2002;31:566–74.

[19] Luli GW, Strohl WR. Comparison of growth, acetate production, and

acetate inhibition of Escherichia coli strains in batch and fed-batch

fermentations. Appl Environ Microbiol 1990;56:1004–11.

[20] Thomas JG, Banyez F. Protein misfolding and inclusion body formation in

recombinant Escherichia coli cells overexpressing heat-shock proteins. J

Biol Chem 1996;271:11141–7.

[21] van den Berg B, Ellis RJ, Dobson CM. Effects of macromolecular

crowding in protein folding and aggregation. EMBO J 1999;18:6927–33.