An engineered disulfide bridge mimics the effect of
calcium to protect neutral protease against local unfolding
Peter Du
¨
rrschmidt*, Johanna Mansfeld and Renate Ulbrich-Hofmann
Department of Biochemistry ⁄ Biotechnology, Martin-Luther University Halle-Wittenberg, Halle ⁄ Saale, Germany
The neutral protease from Bacillus stearothermophilus
belongs to a group of metalloendopeptidases that have
maximum activity at neutral pH. Some members of
this group are highly conserved in amino acid sequence
and tertiary structure and form the group of thermo-
lysin-like proteases (TLPs) with thermolysin as the best
characterized representative [1]. TLPs consist of 300–
319 amino acid residues and are organized into two
domains. They have one catalytic zinc ion, and
between two and four stabilizing calcium ions. The
X-ray structures of thermolysin [2] and the neutral
protease from B. cereus [3] have been resolved and
show great similarities. Because of the high degree of
sequence identity (85%) with thermolysin [4], a 3D
model of the neutral protease from B. stearothermophi-
lus (Fig. 1) was constructed, on the basis of the X-ray
structure of thermolysin, by homology modeling [5]
and has been successfully used in a number of
Keywords
autoproteolysis; disulfide; local unfolding;
neutral protease; stability
Correspondence
R. Ulbrich-Hofmann, Martin-Luther-
University Halle-Wittenberg, Department of
Biochemistry ⁄ Biotechnology, Institute of

obs
vs. GdnHCl concentration). The k
obs
of unfolding showed a difference of
nearly six orders of magnitude (DDG
#
¼ 33.5 kJÆmol
)1
at 25 °C) between
calcium saturation (at 100 mm CaCl
2
) and complete removal of calcium
ions (in the presence of 100 mm EDTA). Analysis of the protease variant
W55F indicated that calcium binding-site III, situated in the critical region
56–69, determines the stability at calcium ion concentrations between 5 and
50 mm. In the chevron plots the disulfide bridge in G8C ⁄ N60C shows a
similar effect compared with pWT as the addition of calcium ions, suggest-
ing that the introduced disulfide bridge fixes the region (near calcium
binding-site III) that is responsible for unfolding and subsequent autopro-
teolysis. Owing to the presence of the disulfide bridge, the DDG
#
is 13.2
kJÆmol
)1
at 25 °C and 5 mm CaCl
2
. Non-linear chevron plots reveal an
intermediate in unfolding probably caused by local unfolding of the loop
56–69. The occurrence of this intermediate is prevented by calcium concen-
trations of > 5 mm, or the introduction of the disulfide bridge

the introduction of two cysteine residues into the
pseudo-wild type enzyme (pWT), a mutant in which
the only naturally occurring cysteine residue at posi-
tion 288 was exchanged with a leucine residue without
significant influence on the specific activity, thermosta-
bility [14] or spectroscopic properties of the enzyme
[15]. The disulfide bond produced a shift in the half-life
at 92.5 °C from <0.3 min to 35.9 min and increased
the temperature, after which half of the initial activity
is lost within 30 min (T
50
), by 16.7 °C [13]. The combi-
nation of eight stabilizing mutations in the region
56–69, including the introduction of the disulfide
bridge, G8C ⁄ N60C, produced an enzyme variant
(boilysin) with a half-life of 170 min at 100 °C [16].
Calcium ions play an important role in the stability
of TLPs. From the X-ray structure of thermolysin
[17,18], one double-binding site (I and II) and two sin-
gle-binding sites (III and IV) have been derived [19].
Data on their binding constants, however, are contra-
dictory. Tajima et al. [20] postulated the same calcium
affinity for all binding sites of thermolysin, whereas
Voordouw & Roche [21] found two classes of calcium-
binding sites with a lower affinity for the double-bind-
ing site. By using crystal soaking studies, Weaver et al.
[22] showed an ordering of the affinity of the binding
sites as follows: I > III > IV ‡ II. Sequence align-
ments with thermolysin revealed four adequate cal-
cium-binding sites also for the neutral protease from

ion, and the violet spheres indicate the
bound calcium ions with numbering of the
binding sites. The sensitive loop 56–69 is
shown in red, and the disulfide bridge
connecting positions 8 and 60 is shown in
yellow.
Engineered disulfide bridge mimics calcium effect P. Du
¨
rrschmidt et al.
1524 FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS
the engineered disulfide bond to conformational and
autoproteolytical stabilization. Particular attention is
paid to the role of calcium ions. For this reason, an
enzyme variant with a mutation in the vicinity of cal-
cium binding-site III (W55F) is included in the studies.
The results indicate two competing routes of auto-
proteolysis, one starting from locally unfolded mole-
cules and the other starting from globally unfolded
molecules.
Results
Screening of conditions for GdnHCl-induced
autoproteolysis
The unfolding of nonspecific proteases, such as TLPs, is
accompanied by autoproteolysis, rendering unfolding
irreversible. To screen conditions where autoproteolysis
becomes significant in GdnHCl-induced unfolding of
pWT and G8C ⁄ N60C, autodegradation of the enzymes
was determined at different protein concentrations (5–
100 lgÆmL
)1

Autoproteolysis becomes more significant as the con-
centration of GdnHCl increases. However, autoproteo-
lysis decreases again at GdnHCl concentrations of
>5 m for pWT and 7 m for G8C ⁄ N60C, and almost
no protein degradation was observed at 7–8 m
GdnHCl. Obviously, at high GdnHCl concentrations,
inactivation of the enzyme by unfolding is so fast that
autoproteolysis cannot occur. G8C ⁄ N60C behaves
similarly to pWT, but the curves in Fig. 3 are shifted
towards higher concentrations of GdnHCl.
Autoproteolysis is independent of the protein
concentration up to 5 m GdnHCl, as concluded from
degradation kinetics between 1 and 100 lgÆmL
)1
pWT
(Fig. 4). Hence, at GdnHCl concentrations of < 5 m,
unfolding is suggested to be the rate-limiting step for
autoproteolysis. In contrast, at GdnHCl concentrations
of > 5 m, the extent of autoproteolysis depends on the
protein concentration between 20 and 100 lgÆmL
)1
pWT (Fig. 4), showing that unfolding is not rate limit-
ing for autoproteolysis under these conditions. An
exact analysis of the interplay of unfolding and auto-
proteolysis, however, requires kinetic measurements, as
described below.
Influence of the disulfide bridge on the kinetics
of unfolding and autoproteolysis
Unfolding of pWT and G8C ⁄ N60C at the standard
calcium concentration of 5 mm was monitored by far-

resulting rate constants (k
obs
) were plotted in a semi-
logarithmic graph as a function of the GdnHCl concen-
tration (chevron plot) (Fig. 5).
The unfolding rates of secondary and tertiary struc-
ture coincided for both of the enzymes at all GdnHCl
concentrations. The autoproteolysis rates were identi-
cal to unfolding rates for G8C ⁄ N60C over the whole
range investigated (4–8 m GdnHCl) and for pWT at
GdnHCl concentrations of £ 5 m. All curves coincide
Fig. 4. Autoproteolysis kinetics of the pseudo-wild type (pWT)
neutral protease from Bacillus stearothermophilus in 5
M and 6.5 M
guanidine hydrochloride (GdnHCl) at different protein concentra-
tions. pWT [100 lgÆmL
)1
(s), 50 lgmL
)1
(e), 20 lgmL
)1
(,),
5 lgmL
)1
(h) and 1 lgmL
)1
(n)] was incubated in 50 mM
Tris ⁄ HCl, pH 7.5, containing 5 mM CaCl
2
and GdnHCl. After various

The data from Fig. 5 allow us to calculate the con-
tribution of the disulfide bridge to the increase in the
Gibbs free energy of activation of unfolding under
native conditions. Using the linear extrapolation model
[27], the rate constant for the unfolding of G8C ⁄ N60C
in the absence of denaturant is 1.09 ± 0.66 · 10
)9
s
)1
.
The corresponding rate constant for pWT, as deter-
mined experimentally by following the subsequent
autoproteolysis (unfolding was too slow to be meas-
ured directly), was 2.44 ± 0.76 · 10
)7
s
)1
. Following
Eyring’s equation:
k ¼
k
B
Á T
h
Á e
À
DG
#
RT
ð1Þ

CaCl
2
, which is below, at and above the standard
concentration of calcium ions used in studies on this
protease [8,25,26]. The observed rate constants of
unfolding proved to be very sensitive to the calcium
ion concentration. As Fig. 6 demonstrates, decreasing
the calcium ion concentration to 2 mm results in a
very distinct nonlinearity of curves in the chevron
plot. At 2 mm CaCl
2
, even G8C ⁄ N60C shows a non-
linear correlation between ln k
obs
and the GdnHCl
concentration, whereas at 100 mm CaCl
2
the devia-
tions from linearity in the semilogarithmic plot dis-
appear for both of the enzymes. Obviously, in chevron
plots low calcium ion concentrations promote nonline-
arity and high calcium ion concentrations promote
linearity.
The influence of calcium ions on the unfolding rate
constants was investigated, in greater detail, in the
presence of 7.25 m GdnHCl where autoproteolysis is
widely suppressed (Fig. 7). At this GdnHCl concentra-
tion, the addition of calcium ions is able to change
the unfolding rate constant over four orders of magni-
tude. The addition of EDTA resulted in a further

Kinetic measurements of autoproteolysis, CD and fluorescence
spectroscopy were performed as described in the Experimental
procedures. The error bars show the standard deviations of the k
obs
values fitted from the measuring data according to a first-order
reaction.
P. Du
¨
rrschmidt et al. Engineered disulfide bridge mimics calcium effect
FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS 1527
merely enhance the calcium affinity, but has an addi-
tional stabilizing effect.
The difference in the observed rate constants of
unfolding between pWT and G8C ⁄ N60C at calcium
ion concentrations of > 5 mm (Fig. 7) must be related
to the position of the introduced disulfide bridge that
is located in the vicinity of calcium binding-site III
(Fig. 1). To assign the contribution of bound calcium
ions at calcium binding-site III, the variant W55F, car-
rying a mutation in the respective binding site, was
included in the measurements. The specific activity of
W55F and the content of secondary structure meas-
ured by far-UV CD is comparable with that of pWT
(results not shown). Therefore, differences in the
unfolding behavior between pWT and W55F should
be caused by alterations in the affinity at calcium
binding-site III. The rate constants of unfolding of
pWT and W55F are similar for calcium ion concentra-
tions of < 5 mm and > 50 mm, but differ at concen-
trations between these values (Fig. 7). This means that

of unfolding were determined by fluorescence measurements, as
described in the Experimental procedures.
Fig. 7. Unfolding kinetics of the pseudo-wild type (pWT) neutral
protease from Bacillus stearothermophilus, of the disulfide bond
mutant G8C ⁄ N60C, and of the protease variant W55F, in the pres-
ence of 7.25
M guanidine hydrochloride (GdnHCl). The enzymes
(5 lgÆmL
)1
) were incubated in 50 mM Tris ⁄ HCl, pH 7.5, containing
CaCl
2
, as indicated, and 7.25 M GdnHCl. k
obs
values of unfolding
were determined by fluorescence measurements, as described in
the Experimental procedures.
Engineered disulfide bridge mimics calcium effect P. Du
¨
rrschmidt et al.
1528 FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS
absence of GdnHCl, the rate constant of unfolding of
pWT (k
0Ca2+
) was 3.01 ± 0.42 · 10
)3
Æs
)1
. In the pres-
ence of 100 mm CaCl

unfolding for the disulfide-containing variant under
native conditions in the presence of 100 mm CaCl
2
(Fig. 6) amounts to 1.76 · 10
)13
s
)1
, which corres-
ponds to a half-life of % 125 000 years. The different
unfolding rate constants for pWT and G8C ⁄ N60C
under native conditions at calcium saturation (100 mm)
confirm the conclusion, drawn above, that the intro-
duction of the disulfide bridge stabilizes the molecule
more than calcium ions.
The influence of isopropanol on the unfolding
rate constant
To measure unfolding without the interference of auto-
proteolysis, the addition of inhibitors to the enzymes
was examined. Inhibitors such as phosphoramidon or
o-phenanthroline are fluorescent and disturb the
applied spectroscopic techniques. The addition of iso-
propanol [IC
50
¼ 2.3% (v ⁄ v)], which is known to act
as an inhibitor of thermolysin [28], resulted in a
marked decrease (but not complete elimination) of
autoproteolysis without interfering with the spectro-
scopic methods. In the presence of 2 and 5 mm CaCl
2
,

obs
values
of unfolding were determined by fluorescence measurements, as
described in the Experimental procedures.
P. Du
¨
rrschmidt et al. Engineered disulfide bridge mimics calcium effect
FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS 1529
results show that autoproteolysis of both pWT and
G8C ⁄ N60C occurs at GdnHCl concentrations of
< 7.5 m (Fig. 3). For pWT, unfolding is rate-limiting
up to % 5 m GnHCl (Figs 4 and 5), whereas at
higher GdnHCl concentrations (5.5–7.5 m), unfolding
becomes faster than subsequent autoproteolysis
(Fig. 5). In this intermediate range of GdnHCl concen-
tration, the rate of autoproteolysis depends on the
protein concentration (Fig. 4). For G8C ⁄ N60C in the
presence of 5 mm CaCl
2
, unfolding is rate-limiting at
GdnHCl concentrations of < 7.5 m.
At very high GdnHCl concentrations (> 7.5 m)
unfolding is so fast that no autoproteolysis can occur,
either with pWT or with G8C ⁄ N60C. Under these con-
ditions, both variants show the same unfolding rates
(Fig. 5).
Stability differences between pWT and G8C ⁄ N60C
emerge under conditions where autoproteolysis occurs
(< 7.5 m GdnHCl). These differences are connected
with evident deviations from linearity in the chevron

source of stabilization of the neutral protease from
B. stearothermophilus. Assuming that all four binding
sites for calcium ions are occupied at a saturating con-
centration of 100 mm CaCl
2
(Fig. 7), their contribution
to the Gibbs free energy of activation of unfolding is
33.5 kJÆmol
)1
. This difference in kinetic stability cor-
responds to the difference between a mesophilic and a
thermophilic enzyme [34]. From unfolding of pWT
and G8C ⁄ N60C as a function of added CaCl
2
in com-
parison to the unfolding of W55F, which is modified
near binding site III within the region 56–69, it can be
concluded that binding site III starts to become occu-
pied at % 5mm CaCl
2
and is saturated at 100 mm
CaCl
2
(Fig. 7). This binding site is formed by residues
55, 57, 59 and 61 [8] and has obviously the lowest
affinity of the four calcium-binding sites, as shown by
the W55F variant (Fig. 7). This finding is consistent
with the results of Veltman et al. [24] who showed that
the thermoinactivation of the neutral protease from
B. stearothermophilus is dramatically changed by muta-

,
respectively. Correspondingly, the disulfide bridge
yields an increase of the kinetic stability at this tem-
perature by 14.5 kJÆmol
)1
, which is similar to the value
obtained here at 25 °C.
High stabilization effects of disulfide bridges are
often observed for reversibly unfolding proteins and
are mostly attributed to the restriction of the degrees
of freedom in the unfolded state [35]. Following this
concept, disulfide bridges should not influence unfold-
ing processes. Indeed, there are only a few reports of
successful stabilization against irreversible unfolding
by disulfide linkage in the literature [36–38]. The high
stabilization effect of the disulfide bond in G8C ⁄ N60C
Engineered disulfide bridge mimics calcium effect P. Du
¨
rrschmidt et al.
1530 FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS
can best be explained by the assumption that unfolding
and ⁄ or autoproteolysis via the unfolding intermediate
is hampered. Similarly to the calcium ions occupying
calcium binding-site III, the loop region 56–69 is sta-
bilized by the disulfide bridge connecting residues 8
and 60. Engineered disulfide bridges producing a sim-
ilar effect on protein stability as calcium ions have
been also reported for subtilisin BPN¢ or alkaline pro-
tease [37,38]. However, under calcium saturation
(Figs 5 and 6), the introduced disulfide bridge seems to

G8C ⁄ N60C (for CaCl
2
‡ 5mm).
Experimental procedures
Chemicals
Casein was purchased from Merck (Darmstadt, Germany),
Abz-AGLA-Nba from Bachem (Heidelberg, Germany),
GdnHCl from ICN Biomedicals GmbH (Eschwege, Ger-
many), isopropanol from Sigma-Aldrich Chemie GmbH
(Deisenhofen, Germany), and tris(hydroxymethyl)amino-
methane (Tris) from Amersham Biosciences (Uppsala,
Sweden). All other reagents were the purest ones avail-
able.
Enzymes
All enzyme variants were produced as described previously
[13,14]. The mutation W55F was introduced by site-directed
mutagenesis using the QuickChangeÔ site-directed muta-
genesis kit (Stratagene, Heidelberg, Germany), the primers
5¢-GTTTTGCCCGGCAGCTTGTTTACCGATGGCGACA
ACCAA-3¢ (forward) and 5¢-TTGGTTGTCGCCATCG
GTAAACAAGCTGCCGGGCAAAAC-3¢ (reverse), and
the wild-type gene of the neutral protease from B. stearo-
thermophilus, cloned into pET-28b(+) (J. Mansfeld, unpub-
lished data). As checked by the progam what if [5], the
mutation W55F should not dramatically disturb the pro-
tein geometry. The sequence of the mutated plasmid was
verified by dideoxy sequencing before using the SalI frag-
ment to reconstruct the pGE501 variant for expression in
B. subtilis.
Immediately before use, the enzyme solution was dialyzed

Ca
2+
Fig. 9. Unfolding scheme. The native state with unoccupied cal-
cium binding-site III (N) unfolds via an intermediate state (N*),
which is susceptible to autoproteolysis. Under calcium-saturation
conditions the native state (N
Ca
2+
) unfolds globally to the com-
pletely unfolded state D. The introduced disulfide bridge,
G8C ⁄ N60C, restricts unfolding.
P. Du
¨
rrschmidt et al. Engineered disulfide bridge mimics calcium effect
FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS 1531
Abz-AGLA-Nba was measured at 25 ° C at a substrate con-
centration of 20 lm. The increase in fluorescence emission
at 415 nm after excitation at 340 nm was recorded over
10 min [41].
Detection of autoproteolysis in the presence
of GdnHCl
SDS ⁄ PAGE was used to follow autoproteolysis of the
enzyme at a protein concentration of ‡ 20 lgÆmL
)1
. To con-
centrate the protein and to remove GdnHCl, the samples
were treated with sodium deoxycholate [15]. The precipi-
tates were dried under vacuum, dissolved in SDS sample
buffer and separated on 15% (w ⁄ v) polyacrylamide gels
according to Laemmli [42]. The proteins were stained with

curves in the presence of GndHCl reflect the decrease of
native molecules, and autoproteolysis does not disturb the
measurements.
Kinetic analysis
Kinetic measurements of both unfolding and autoproteoly-
sis were started by the addition of protein to the GdnHCl-
containing solution. The progress curves were fitted to a
single exponential function, yielding the rate constants
k
obs
.
Acknowledgements
The authors thank G. Vriend for preparation of
Fig. 1, V. G. Eijsink for plasmid pGE501 encoding the
gene of the wild-type sequence of the neutral pro-
tease from B. stearothermophilus, and R. Golbik for
stimulating discussions. The financial support of the
Kultusministerium des Landes Sachsen-Anhalt and
the Deutsche Forschungsgemeinschaft is gratefully
acknowledged.
References
1 Rawlings ND & Barrett AJ (1995) Evolutionary families
of metallopeptidases. Methods Enzymol 248, 183–228.
2 Holmes MA & Matthews BW (1982) Structure of ther-
molysin refined at 1.6 A resolution. J Mol Biol 160,
623–639.
3 Pauptit RA, Karlsson R, Picot D, Jenkins JA, Niklaus-
Reimer AS & Jansonius JN (1988) Crystal structure of
neutral protease from Bacillus cereus refined at 3.0 A
˚

Biochim Acta 48, 63–67.
11 Mansfeld J & Ulbrich-Hofmann R (2000) Site-specific
and random immobilization of thermolysin-like
Engineered disulfide bridge mimics calcium effect P. Du
¨
rrschmidt et al.
1532 FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS
proteases reflected in the thermal inactivation kinetics.
Biotechnol Appl Biochem 32, 189–195.
12 Mansfeld J, Vriend G, Van den Burg B, Eijsink VG &
Ulbrich-Hofmann R (1999) Probing the unfolding
region in a thermolysin-like protease by site-specific
immobilization. Biochemistry 38, 8240–8245.
13 Mansfeld J, Vriend G, Dijkstra BW, Veltman OR, Van
den Burg B, Venema G, Ulbrich-Hofmann R & Eijsink
VG (1997) Extreme stabilization of a thermolysin-like
protease by an engineered disulfide bond. J Biol Chem
272, 11152–11156.
14 Eijsink VG, Dijkstra BW, Vriend G, van der Zee JR,
Veltman OR, van der Vinne B, van den Burg B, Kempe
S & Venema G (1992) The effect of cavity-filling muta-
tions on the thermostability of Bacillus stearothermophi-
lus neutral protease. Protein Eng 5, 421–426.
15 Durrschmidt P, Mansfeld J & Ulbrich-Hofmann R
(2001) Differentiation between conformational and
autoproteolytic stability of the neutral protease from
Bacillus stearothermophilus containing an engineered dis-
ulfide bond. Eur J Biochem 268, 3612–3618.
16 Van den Burg B, Vriend G, Veltman OR, Venema G &
Eijsink VG (1998) Engineering an enzyme to resist boil-

dent of calcium. FEBS Lett 405, 241–244.
24 Veltman OR, Vriend G, Berendsen HJ, Van den Burg
B, Venema G & Eijsink VG (1998) A single calcium
binding site is crucial for the calcium-dependent thermal
stability of thermolysin-like proteases. Biochemistry 37,
5312–5319.
25 Eijsink VG, Vriend G, Van Den Burg B, Venema G &
Stulp BK (1990) Contribution of the C-terminal amino
acid to the stability of Bacillus subtilis neutral protease.
Protein Eng 4, 99–104.
26 Nakamura S, Tanaka T, Yada RY & Nakai S (1997)
Improving the thermostability of Bacillus stearothermo-
philus neutral protease by introducing proline into the
active site helix. Protein Eng 10, 1263–1269.
27 Tanford C (1970) Protein denaturation. C. Theoretical
models for the mechanism of denaturation. Adv Protein
Chem 24, 1–95.
28 Muta Y & Inouye K (2002) Inhibitory effects of alco-
hols on thermolysin activity as examined using a fluor-
escent substrate. J Biochem (Tokyo) 132, 945–951.
29 Bhuyan AK & Udgaonkar JB (1998) Multiple kinetic
intermediates accumulate during the unfolding of horse
cytochrome c in the oxidized state. Biochemistry 37,
9147–9155.
30 Sauder JM, MacKenzie NE & Roder H (1996) Kinetic
mechanism of folding and unfolding of Rhodobacter
capsulatus cytochrome c2. Biochemistry 35, 16852–
16862.
31 Zaidi FN, Nath U & Udgaonkar JB (1997) Multiple
intermediates and transition states during protein

II. General properties. J Gen Physiol 30, 291–299.
P. Du
¨
rrschmidt et al. Engineered disulfide bridge mimics calcium effect
FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS 1533
40 Fujii M, Takagi M, Imanaka T & Aiba S (1983) Mole-
cular cloning of a thermostable neutral protease gene
from Bacillus stearothermophilus in a vector plasmid and
its expression in Bacillus stearothermophilus and Bacillus
subtilis. J Bacteriol 154, 831–837.
41 Nishino N & Powers JC (1980) Pseudomonas aeruginosa
elastase. Development of a new substrate, inhibitors,
and an affinity ligand. J Biol Chem 255, 3482–3486.
42 Laemmli UK (1970) Cleavage of structural proteins
during the assembly of the head of bacteriophage T4.
Nature 227, 680–685.
43 Wharton SA, Martin SR, Ruigrok RW, Skehel JJ &
Wiley DC (1988) Membrane fusion by peptide analo-
gues of influenza virus haemagglutinin. J Gen Virol 69,
1847–1857.
Engineered disulfide bridge mimics calcium effect P. Du
¨
rrschmidt et al.
1534 FEBS Journal 272 (2005) 1523–1534 ª 2005 FEBS