Stabilization of a (ba)
8
-barrel protein by an engineered disulfide
bridge
Andreas Ivens
1
, Olga Mayans
2
, Halina Szadkowski
3
, Catharina Ju¨ rgens
1
, Matthias Wilmanns
2
and Kasper Kirschner
3
1
Universita
¨
tzuKo
¨
ln, Institut fu
¨
r Biochemie, Ko
¨
ln, Germany;
2
EMBL c/o DESY, Hamburg, Germany;
3
Biozentrum, Universita
¨

8
-barrel
proteins; stabilizing disulfide bonds; protein e ngineering.
Indoleglycerol phosphate synthase (IGPS) is a (ba)
8
-barrel
protein with an N-terminal extension of 48 residues. In
Escherichia coli, IGPS (eIGPS) is the N-terminal domain of
a monomeric, bifunctional enzyme, where the C-terminal
domain is phosphoribosyl anthranilate isomerase (ePRAI),
folded into another (ba)
8
-barrel [1]. The catalytic efficiencies
of the engineered separated domains are virtually identical
to those in the bifunctional enzyme [2]. eIGPS is, however,
more labile than ePRAI. The catalytic activity of eIGPS
decays at 55 °C w ith a half-life o f 0 .5 min [3]. In contrast,
ePRAI activity decays at 60 °C with a half-life of 100 min
(R. Sterner, Institut fu
¨
r Biochemie, Universita
¨
tzuKo
¨
ln,
Germany, personal communication). The eIGPS domain,
in turn, is also more labile than eIGPS in the native
bifunctional protein [4,5,6].
In contrast to eIGPS [1], the IGP synthases from the
hyperthermophiles Sulfolobus solfataricus (sIGPS [7]) and

crosslink the ba1andba8 modules.
MATERIALS AND METHODS
DNA manipulations and sequence analysis
Preparation o f DNA samples, digestion with restriction
endonucleases, agarose gel electrophoresis, and DNA
Correspondence to A. Ivens, Universita
¨
tzuKo
¨
ln, Institut fu
¨
r
Biochemie, Otto-Fischer-Str. 12–14, D-50674 Ko
¨
ln, Germany.
Fax: +49 221 4706 731, Tel.: +49 221 470539,
E-mail:
Abbreviations: IGP, indoleglycerol phosphate; CdRP, 1-(o-carboxy-
phenylamino)-1-deoxy-
D
-ribulose-5-phosphate; PRA, N-phos-
phoribosyl anthranilate; ePRAI, PRA isomerase domain from
Escherichia coli; eIGPS, IGP synthase domain from Escherichia coli;
eIGPS-PRAI, indoleglycerol phosphate synthase–phosphoribosyl-
anthranilate isomerase bifunctional protein from Escherichia coli;
etrpC, gene encoding eIGPS; sIGPS, IGP synthase from Su lfolobus
solfataricus; tIGPS, IGP synthase from Thermotoga maritim a;Nbs
2
,
5,5¢-dithiobis(2-nitrobenzoic acid); ASA, accessible surface area.

+
B
+
/e14
–
(McrA
–
) D(lac-proAB) thi
gyrA96(Nal
r
) endA1 hsdR17 (r
K
–
m
K
+
) relA1 supE44
recA1]. The expression vector pET21a(+), where protein
production is under control of the T7-RNA-polymerase
promoter [17], was used for expression of etrpC mutants.
Oligonucleotides
The following PCR primers were used t o amplify the etrpC
gene from the vector pMc-C/F, which contains the bifunc-
tional eIGPS:ePRAI gene [2]. The 5¢ primer was used as a
mutagenic primer for replacing Thr3 by Cys (bold letters
indicate the mutated codon). T3C 5¢ primer, 5 ¢-CGAGGG
TAA
CATATGCAATGCGTTTTAGCGAA-3¢;etrpC3¢
primer, 5¢-CCACGCGTC
AAGCTTCATACTTTATTC-3¢.

of each nucleotide triphos-
phate, 20 pmol of each primer, 0.1 lg of t emplate, 4 lLof
10 x Pfu reaction buffer and 2.5 U of Pfu -DNA polymerase
(Stratagene) in a total volume of 40 lL. The amplification
protocol for the production of megaprimers consisted of
3 min at 95 °C, followed by 35 cycles of 1 min at 95 °C,
2 min at 55 °C and 3 min at 72 °C. The megaprimers were
purified by electrophoresis on a 0.8% agarose gel. They
were used as templates in various dilutions and a t a reduced
annealing t emperature of 50 °C in a second PCR r eaction
with the 5 ¢ and 3¢ primers to yield the full-length mutated
gene. The resulting etrpC fragment was purified by electro-
phoresis on a 0 .8% agarose gel, digested with NdeIand
HindIII and purifi ed again. The fragment was then ligated
into a NdeI–HindIII digested and dephosphorylated
pET21a(+) vector, yielding the vector pET21a(+)-etrpC.
After transformation of E. coli BL21(DE3) with
pET21a(+)-etrpC, transformants w ere grown overnight in
2 m L Luria–Bertani medium [16], containing 0.1 lg amp-
icillinÆmL
)1
(Luria–Bertani/amp medium). The plasmids
were isolated and digested with NdeIandHindIII to screen
for clones with inserts. One positive clone was confirmed by
complete DNA sequencing.
Expression and purification of
E. coli
eIGPS(3–189)
The protein was expressed in E. coli BL21(DE3). Single
colonies harbouring the plasmid pET21a(+)-etrpC (3–189)

)1
, equilibrated with 10 m
M
potassium phosphate,
pH 7.5. After washing with equilibration buffer for 1.5 col-
umn vol., the column was eluted with a linear gradient from
10 to 300 m
M
potassium phosphate buffer pH 7.5, 1 m
M
EDTA. eIGPS(3–189) eluted, as determined by activity and
SDS/PAGE, at a phosphate concentration of 150 m
M
.
Fractions containing eIGPS(3–189) were pooled and
dialyzed overnight against 5 m
M
potassium phosphate
buffer pH 6.8, 100 m
M
KCl.
The dialysate was loaded onto a hydroxylapatite column
(2.5 · 25 cm, 122 mL) with 34.2 mLÆh
)1
, that had been
equilibrated w ith 5 m
M
potassium phosphate buffer pH 6.8,
100 m
M

NaCl, and eluted with
equilibration buffer at a flow rate o f 34.2 mLÆh
)1
. Fractions
with pure eIGPS(3–189) were concentrated by ultrafiltration
andstoredat)70 °C after dripping into liquid nitrogen.
Enzymatic assay for indoleglycerol phosphate synthase
Indoleglycerol phosphate synthase activity was assayed at
25 °Cin50m
M
Tris/HCl pH 7.5, 1 m
M
EDTA, with 40–
70 n
M
eIGPS and 3–5 l
M
CdRP. The reaction was started
by addition of the nonfluorescent substrate C dRP [ 20].
Appearance of IGP was measured continuously by its
fluorescence excited at 280 nm a nd emitted a t 350 nm.
Because IGP accumulates, t he progress curves were fitted to
the integrated M ichaelis–Menten equation that takes com-
petitive product inhibition into account [21]. The formation
of 1 lmol IGP per minute at 25 °C was defined as one unit
of activity (Table 1).
SDS/PAGE
SDS/PAGE was carried out according to the method of
Laemmli [22]. The stacking gel and separation gel contained
6 and 12.5% acrylamide, respectively. The protein samples

0.1
M
potassium phosphate buffer at a given temperature
and irreversibly heat inactivated. Aliquots were taken at
certain time points and chilled on ice, until the remaining
activity was determined (in Tris, as described above) and
plotted against the incubation time. Kinetic data were
obtained a s d escribed above. Incubation buffer was 100 m
M
potassium phosphate, pH 7.5, 1 m
M
EDTA, 1 m
M
dithio-
threitol. D ithiothreitol was omitted in the case of oxidized
eIGPS(3–189).
Oxidation and reduction of the engineered disulfide
bridge in eIGPS(3–189)
For reduction of the disulfide bridge, the enzyme at a
concentration o f 1.5 mgÆmL
)1
was i ncubated f or 6 h at 4 °C
in 50 m
M
potassium phosphate pH 7.5, 300 m
M
NaCL,
10 m
M
dithiothreitol. For promoting the formation of the

EDTA. The following
extinction coefficients were used: e
TNB
(440 nm) ¼
9.22 m
M
)1
Æcm
)1
, eNbs
2
(325 nm) ¼ 17.38 m
M
)1
Æcm
)1
.
Excess amounts of both re ducing and oxidizing compounds
were removed before the measurements by gel filtration on
NAP columns (Pharmacia). A blank run was performed
with assay buffer before the protein was added to a final
concentration of 10–30 l
M
in a final volume of 1 mL. After
various time points the absorption at 440 nm was recorded.
CD spectra
CD spectra were monitored with a Jasco model J -720
spectropolarimeter, which was connected to a Philips SX
computer. The measurements were carried out in 0.05
M

Yield
%
Crude extract 2093 529 0.3 100
Anion exchange eluate 311 313 1.0 59
Hydroxylapatite eluate 208 208 1.0 40
Gel filtration eluate 102 239 2.3 45
Ó FEBS 2002 Stabilization of a (ba)
8
-barrel protein (Eur. J. Biochem. 269) 1147
out as reported for the wild-type monomeric eIGPS [25] in
50 m
M
potassium phosphate, pH 5 .0, 1.2
M
ammonium
sulphate and 5 m
M
EDTA.
Data were collected at the synchrotron radiation beam
line X11 (EMBL c/o DESY, Hamburg) from shock-frozen
crystals at 100 K using 30% (v/v) glycerol as cryoprotec-
tant. Data were recorded on a MAR-CCD detector in three
resolution sweeps to a maximum resolution of 2.1 A
˚
. The
crystals belong to the space group P6
3
22 and are affected by
strong pseudosymmetry. A large cell with dimensions a ¼
141.4 A

, had values of 25.0% R
merge
,
multiplicity 2.6% and 78% completeness). Structure solu-
tion was carried out by the molecular r eplace ment technique
(AMoRe [27]), using the eIGPS domain from the bifunc-
tional enzyme [1] as a search model. For refinement,
reflection data were divided into a working set and a t est set
(1057 reflections) u sing
FREERFLAG
. Refinement was carried
out using the
CNS
software [28] and included bulk solvent
correction, overall anisotropic B-factor scaling and
restrained, individual, isotropic B factor refinement. The
structure has been refined to a crystallographic R-factor of
24.1% (R
free
31.9%). The model includes protein residues
1–259, the CdRP compound and 187 solvent atoms. No
obvious interpretable electron d ensity can b e observed for
residues 1 and 2, so these were included as model s.
The coordinates and structure factors have been depo-
sited at the PDB with accession code 1218477 (1JCM).
RESULTS AND DISCUSSION
Design of disulfide bonds
Engineering of a new disulfide bond into eIGPS must take
into account the presence of three parental cysteines within
the s trands b

adjacent disulfide bridge, thus impairing catalysis.
One of the preferred, new disulfide bonds requires the
double substitution T3C/R189C, and fi xes t he N-terminus
to the barrel core of this variant, designated eIGPS(3–189).
The disulfide bond is accessible (ASA values of T3 and
R189 are 54 and 68 A
˚
2
, respectively) and fortuitously
mimics one o f the e xtra salt bridges in both sIGPS [7] and
tIGPS [3], which are missing in eIGPS [1]. However, the
replacement R189C in eIGPS disrupts the parental short-
range salt bridge of R189 to E169 on helix a5
.
The other selected disulfide variant involves the double
substitution I64C/M240C (ASA values o f I 64 and M240 are
27 and 0 A
˚
2
, respectively), and is de signated eIGPS(64–
240). M240 is an invariant but solvent-inaccessible residue
that anchors t he short h elix a8¢ to the core o f the protein [1].
This proposed disulfide crosslinks the loops b1a1andb8a8,
which are widely separated i n sequence but adjacent i n space
(Fig. 1 ), thus clamping the barrel between the N- and
C-terminal modules ba
1
and ba
8
. This disulfide bond is

2-mercaptoethanol; Fig. 2B), and with an overall yield of
45% (Table 1).
When the (3–18 9) protein was analyzed by SD S/PAGE
under nonreducing conditions, two bands of about equal
strength were observed (Fig. 2A). Because the spontaneous
formation of the disulfide bond was apparently incomplete,
total oxidation was achieved by incubating the protein with
an excess of Ellman’s reagent (Nbs
2
[32]), leading to a single,
but faste r migrating band (Fig. 2C). SDS micelles decorated
with disulfide-bonded proteins have a smaller hydrodynamic
volume than those d ecorated with the corresponding dithiol
forms [33], and therefore migrate more rapidly.
In contrast, during culture at 22 °C of the cells producing
the ( 64–240) va riant, most of the protein partitioned into the
insoluble fraction of the cell homogenate. Attempts to
purify this variant by first solubilizing the pre cipitate in
guanidinium chloride in the presence or absence of
dithiothreitol, an d then dialyzing a gainst phosphate buffer
[5], failed to yield significant amounts of soluble material.
Apparently, parallel substitution of residues 6 4 and 240 by
cysteines prevents the correct folding of the enzyme. The
available IGPS structures [1,7] suggest t hat replacement of
the long and hydrophobic side chain of the buried and
invariable residue M240 by the short, polar side-chain of
cysteine m ay disrupt the abundant hydrophobic i nteractions
at the C-terminus and hinder the correct folding of the
protein. Stu dies with this variant w ere therefore not pursued
further.

differences compared to the wild-type eIGPS in the vicinity
of the substitutions. Additional electron density was
observed, however, in the active site of the protein. It can
be modelled as the CdRP substrate (data not shown), but a
detailed description including a comparison to related
Fig. 2. Partial oxidative closure of the engineered (3–189) disulfide
bond. SD S PAGE in a bsence of mercaptoethanol. Lane A, purified
and spontaneously oxidized, variant (3–189); l ane B, as in (A), but with
dithiothreitol in the sample buffer; lane C, as in (A), but with Nbs
2
;
lanes M, marker proteins with the given M
r
-values (kDa).
Fig. 3. 2F
obs
-F
calc
/a
calc
electron density map
showing the disulfide bond contoured at 1.0 r.
The newly introduced cysteine residues C3 and
C189 are l abelled (black dots, Sulfur atoms).
Ó FEBS 2002 Stabilization of a (ba)
8
-barrel protein (Eur. J. Biochem. 269) 1149
complex structures will be reported elsewhere. Perhaps this
unexpected feature is responsible for the observed incom-
plete autoxidation of eIGPS(3–189) shown in Fig. 2A. The

differences local to helix a0 s eem to occur in both red(3–189)
and o x(3–189), w ith respect to the wild-type. In summary,
our near-UV and far-UV CD as well as fluorescence
measurements confirm that n either the introduction of the
(3–189) disulfide bridge nor specific experimental c onditions
affect the structure of the eIGPS dom ain.
Thermostability
eIGPS can be reversibly unfolded by GdmCl in both Tris
[34] and phosphate [35] buffers. Red(3–189) displays the
same properties (data not shown). However, the unfolding
of ox(3–189) by GdmCl in the absence of dithiothreitol was
irreversible, presumably due to thiol-disulfide scrambling
[30]. Therefore, t he relative stability of the three forms could
only be estimated by irreversible thermal inac tivation
(Fig. 6 ) [14]. The results show that the maximal velocities
(V
max
¼ k
cat
· [E
0
]) of the three forms decay irreve rsibly
and exponentially at 50 °C. In contrast to eIGPS, ox(3–189)
is stabilized 13-fold, whereas red(3–189) is destabilized
fivefold, most likely due to the loss of the salt bridge E167-
R189. In other words, ox(3–189) is stabilized 65-fold over
the d ithiol form, which is the correct reference for estimating
Fig. 4. Far-UV CD spectra at 25 °C. s,parentaleIGPS;h, reduced
(dithiol); n, oxidized (disulfide) forms of the variant eIGPS (3–189).
Protein concentrations were between 10 and 21 l

determined in samples drawn at the indicated times and quenched on
ice. Half-lives: ox(3–189), 49 min; eIGPS, 3.7 min; red(3–189),
0.75 min .
1150 A. Ivens et al. (Eur. J. Biochem. 269) Ó FEBS 2002
the effect of closing this disulfide bridge. These results imply
that the engineered disulfide bond crosslinks parts of the
structure that p robably separate in t he parental protein
before the rate-determining step of i ts irreversible unfolding
is attained [15]. Thus, t he d isulfide-linked variant apparently
unfolds via a transition state that is different from t o that of
the wild-type eIGPS. Variation of the phosphate concen-
tration between 5 a nd 100 m
M
revealed that the kinetic
stabilities of eIGPS and its variants increase with i ncreasing
phosphate concentration (data not shown). These observa-
tions support t he idea that phosphate may serve as an
additionally stabilizing electrostatic clamp within the active
site. Note that phosphate specifically interacts with K55
(loop b1a1), G216 ( loo p b7a7) (not shown i n Fig. 1) a nd the
helix dipole of helix a8¢ [1], i.e. protein segments that are far
apart in the protein sequence but adjacent in s pace.
These considerations could also explain why the second
disulfide variant (64–240) does not fold properly, even in
the presence of phosphate. M240 is an invariant, solvent-
inaccessible residue (ASA ¼ 0A
˚
2
) that anchors t he short
helix a8¢ to the core of th e protein [1]. Merz et al. [36] have

obs
½ESð2Þ
where k
obs
¼ k[DTNB] is the observed first-order rate
constant. Both the total number of reactive cysteine
sulfhydryls as well as their average rate constant k as
expressed b y t he observed half-lives t
1=2
¼
ln 2
k½DTNB
are
presented in Table 2. The measurements were performed
in 50 m
M
phosphate buffer, when the active site is 97%
saturated with phosphate [38].
As determined by SDS/PAGE under nonreducing condi-
tions (as described in Fig. 2), no i ntermolecular bridges were
formed during the oxidation process, as there was no
evidence for e ither aggregation o r cross-linking. The possi-
bility t hat a further i ntramole cular disulfide bridge had
formed between C113 and C134 (cf. Fig. 1) was also
excluded b y measurements of the forms in 0.5% SDS, which
unfolds the proteins immediately and allows Nbs
2
to react
with all free thiols that were not accessible in the folded
state. No decrease of the maximally expected numb er o f free

), it is likely that C3 and C189 of red(3–189)
are the two reactive cysteines. They are converted to the
corresponding disulfide via a mixed disulfide intermediate
[E(SH)STNB], which does not accumulate.
EðSHÞ
2
þ DTNB ! EðSHÞSTNB þ TNB ð3Þ
EðSHÞSTNB ! ES
2
þ TNB ð4Þ
The cysteine group in excess of C 3 and C189 that reacts
to 40% c ompletion in r ed(3–189) is likely the same as that
which reacts in eIGPS to 90%. The particularly slow
reaction of this parental cysteine in ox(3–198) to only 10%
completion must be due to the decreased structural
fluctuations of this form of the variant hindering the access
of DTNB by comparison to eIGPS.
Catalytic constants
As the active s ite is l ocated in a d epression at the C-terminal
end of the b-barrel, between the structured segments that
carry the newly introduced pairs of cysteines (see Fig. 1),
enzyme activity is a sensitive monitor for detecting changes
in both the structure and flexibility of the three enzymes.
Steady-state kinetic measurements were conducted in Tris
Table 2. Reaction of protein sulfhydryl groups with Nbs
2
. The protecting effect of th e introduced disulfide bridge.
Variants
Free sulfhydryl groups per protein chain
a

ÓFEBS 2002 Stabilization of a (ba)
8
-barrel protein (Eur. J. Biochem. 269) 1151
buffer and i n the absence of dith iothreitol. Measurements in
phosphate buffer are not feasible because phosphate is a
competitive inhibitor (K
i
¼ 2.8 m
M
[38]). The Michaelis
constants (K
CdRP
M
) of the two forms of both ox(3–189) and
red(3–189) are only % 15% smaller than that of eIGPS
(Table 3). The turnover numbers, however, are decreased t o
45% in red(3–189) and to 10% in ox(3–189). A s t he poor
activity of the thermostable IGPS from S. solfataric us at
low temperature is due to the rate-limiting release of the
product IGP [36], it i s likely that ox(3–189) is ‘constipated’
[36] by the rigidified structure. This finding suggests that
covalent crosslinking the helix a0 to the loop b6a6is
responsible for the retarded release of product in ox(3–189),
and i s supported by the decreased reactivity with Nbs
2
of the
single most reactive cysteine in o x(3–189), in contrast to
eIGPS (Table 2).
CONCLUSION
We have demonstrated that a mesophilic (b/a)

for designing the stereo figure. This work was supported by grant Nr.
31–45855.95 of the Swiss National Science Foundation (to K. K.).
REFERENCES
1. Wilmanns, M., Priestle, J.P., Niermann, T. & Jansonius, J.N.
(1992) Three-dimensional structure of the bifunc tional e nzyme
phosphoribosyl anthranilate isomerase: indoleglycerol phosphate
synthase from Escherichia coli r efined at 2.0 A
˚
resolution.
J. Mol. Biol. 223, 477–507.
2. Eberhard, M., T sai-Pflugfelder, M., Bolewska, K., Hommel, U .
& Kirschner, K . ( 1995) Indoleg lycerol p hosphate syntha se-
phosphoribosyl anthranilate isomerase: comparison of the
bifunctional enzyme from Escherichia coli with engineered
monofunctional domains. Biochemistry 34, 5419–5428.
3. Merz, A., Kno
¨
chel, T., Jansonius, J.N. & Kirschner, K. (1999)
The hyperthermostable Indolglycerol Phosphate Synthase
from Thermotoga maritima is destabilized by mutational dis-
ruption o f two solvent-exposed salt bridges. J. Mol. Biol. 288,
753–763.
4. Stehlin,C.,Dahm,A.&Kirschner,K.(1997)Deletionmuta-
genesis as a test of evolutionary relatedness of indoleglycerol
phosphate syn thase w ith ot her T IM barrel enzymes. FEBS Lett.
403, 268–272.
5. Darimont, B., Stehlin, C., Szadkowski, H. & Kirschner, K.
(1998) Mutational analysis of the active site of indoleglycerol
phosphate synthase from Escherichia coli. Protein Sci. 7, 1221–
1232.

regarding protein stability in vivo and its r elation with misfolding
diseases. Proteins 40, 58–70.
15. Clarke, J. & Fersht, A.R. (1993) Engineered disulfide bonds as
probes of the folding pathway of barnase: i ncreasing the stability
of proteins against the rate of denaturation. Biochemistry 32,
4322–4329.
16. Sambrook, J., Fritsch, E.E. & Maniatis, T. (1989) Molecular
Cloning: a Laboratory Manual. Cold Spring Harbor Laboratory
Press, Cold Spring Harbor, NY.
17. Studier, F .W., R o senberg, A .H., Du nn, J.J. & D ubendo rff, J .W.
(1990) Use of T7 R NA polymerase to direct e xpression of c loned
genes. Methods Enzymol. 185, 60–89.
18. Horton, R.M., Hunt, H.D., Ho, S.N., Pullen, J.K. & Pease, L.R.
(1989) Engineering hybrid genes without the use of restriction
enzymes: gene splicing by overlap extension. Gene 77, 61–68.
19. Sarkar, G. & Sommer, S.S. (1990) The ÔmegaprimerÕ method of
site-directed mutagenesis. Biotechniques 8, 404–407.
20. Kirschner, K., Szadkowski, H., Jardetzky, T.S. & Hager, V.
(1987) Phosphoribosyl anthranilate isomerase-indoleglycerol-
Table 3. Steady-state kinetic constants o f e IGPS, red(3–189) a nd o x
(3–189)
a
.
Variant
k
cat
(s
)1
)
K

assembly of the head o f b acteriophage T 4. Na ture 227, 680–685.
23. Bradford, M.M. (1976) A rapid and sensitive method for the
quantitation of microgram quantities of protein utilizing the
principle of prot ein-dye binding. Anal. Biochem. 72, 248–254.
24. Ellmann, G.E. (1959) Tissue sulfhydryl groups. Arch. Biochem.
Biophys. 82, 7 0–77.
25. Wilmanns, M., Schlagenhauf, E., Fol, B. & Jansonius, J.N.
(1990) Crystallization a nd structure solution at 4 A
˚
resolution of
the r ecombinant synthase domain o f N-(5¢-phosphoribosyl)-
anthranilate isomerase: indole-3-glycerol-phosphate synthase
from Escherichia c oli complexed t o a substrate a nalogu e. Pr otein
Eng. 3, 173–180.
26. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray dif-
fraction data collected in oscillation mode. Methods Enzymol.
276, 307–326.
27. Navaza, J. (1994) An automated package for molecular
replacemen t. Acta Crystallogr. A50, 157–163.
28. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W .L., Gros,
P., Grosse-Kunstleve, R.W., Jiang, J S., Kuszewski, J., Nilges,
N., Pannu, N.S., Read, R.J., Rice, L.M., Simonson, T. & War-
ren, G.L. (1998) Crystallography and NMR system (CNS): a
new software s ystem f or m acromolecula r structu re det ermina-
tion. Acta Crystallogr. D54, 905–921.
28a. Collaborative Computational Project, Number 4 (1994) The
CCP4 Suite: Programs for Protein Crystallography. Acta
Crystallogr. D50, 760–763.
29. Swodhamini, R., S rinivasan, N., Shoichet, B., Santi, D.V.,
Ramakrishnan, C. & Balaram, P. ( 1989) Stereochemical model-

38. Bisswanger, H., Kirschner, K., Cohn, W., Hager, V. & H ansson,
E. (1979) N-(5-Phosphoribosyl) anthranilate isomerase-indole-
glycerol-phosphate synt hase. 1 . A su bstrate analogue binds to
two different binding sites on the bifunctional enzyme from
Escherichia coli. Biochemistry 18, 5946–5953.
39. Sterner, R. & Liebl, W. (2001) Thermophilic adaptation of
proteins. Crit. Rev. Biochem. Mol. Biol. 36, 39–106.
Ó FEBS 2002 Stabilization of a (ba)
8
-barrel protein (Eur. J. Biochem. 269) 1153