Báo cáo Y học: Stepwise adaptations of citrate synthase to survival at life’s extremes From psychrophile to hyperthermophile pot

Stepwise adaptations of citrate synthase to survival at
life’s extremes
From psychrophile to hyperthermophile
Graeme S. Bell
1
, Rupert J. M. Russell
2
, Helen Connaris
2
, David W. Hough
1
, Michael J. Danson
1
and Garry L. Taylor
1,2
1
Centre for Extremophile Research, Department of Biology and Biochemistry, University of Bath, UK;
2
Centre for Biomolecular
Sciences, University of St Andrews, St. Andrews, UK
The crystal structure of citrate synthase from the thermo-
philic Archaeon Sulfolobus solfataricus (optimum growth
temperature ¼ 85 °C) has been determined, extending the
number of crystal structures of citrate synthase from differ-
ent organisms to a total of five that span the temperature
range over which life exists (from psychrophile to hyper-
thermophile). Detailed structural analysis has revealed
possible molecular mechanisms that determine the different
stabilities of the five proteins. The key to these mechanisms is
the precise structural location of the additional interactions.
As one ascends the temperature ladder, the subunit interface

of this dimeric enzyme and loop regions are reinforced by
complex electrostatic interactions, and there is a reduced
exposure of hydrophobic surface. These observations reveal
a progressive pattern of stabilization through multiple
additional interactions at solvent exposed, loop and inter-
facial regions.
Keywords: citrate synthase; Sulfolobus; citrate synthase;
thermostability; crystal structure; ion networks.
Comparative structural analysis of the same protein isolated
from mesophiles and thermophiles have highlighted many
structural adaptations that confer protein thermostability
[1–6]. The importance of electrostatic interactions at specific
locations within the structure, and particularly the presence
of ion-pair networks, is a feature that is common to almost
all the hyperthermophilic proteins [7–10], although many
other additional differences such as improved hydrophobic
packing, compactness and additional hydrogen bonds have
been observed in other proteins.
For our analysis we have chosen the enzyme citrate
synthase (CS) (EC 4.1.3.7), which catalyses the condensa-
tion of oxaloacetate and acetyl-CoA to form citrate and
CoA. The enzyme from psychrophilic, mesophilic and
thermophilic sources has been intensively studied both
kinetically [11–13] and structurally [3,14]. Crystal structures
exist for CS from a psychrophilic Antarctic bacterium
Arthrobacter strain DS2-3R (growth optimum ¼ 31 °C)
[15], pig (37 °C) [16], and the Archaea Thermoplasma
acidophilum (55 °C) [17] and Pyrococcus furiosus (100 °C)
[18]. To extend our previous studies we have chosen the
organism Sulfolobus solfataricus, a thermophilic Archaeon

that optimally grows at 85 °C. The gene for Sulfolobus
solfataricus CS has been cloned and sequenced [19], and
over-expressed in E. coli. The purified recombinant protein
exists as a homodimer of M
r
¼ 81,000, with each monomer
comprising 379 amino acids. The following abbreviations
will be used for the CSs, including their optimal growth
temperatures: Arthrobacter: ArCS(31), pig: PigCS(37),
T. acidolphilum: TpCS(55), S. solfataricus:ScCS(85),and
P. furiosus: PfCS(100).
The structure of unliganded SsCS(85) reported in this
paper can now be entered into the temperature ladder of CS
structures, and fills in the gap between the 55 °C and 100 °C
enzymes. Six CS crystal structures from five host organisms
(Table 1) can now be used for comparative analysis in order
to identify some of the structural features that could confer
(hyper)thermostability in this enzyme ÔfamilyÕ.Ascanbe
seen from Table 1, the organisms span the range of
temperatures at which life is known to exist, and the
inherent stability of each CS, from in vitro measured half-
lives of thermal inactivation [19–21], increases with the
optimum growth temperature of the host cells. The structure
of the SsCS(85) is thus discussed in comparison with the
other CS structures, and trends in structural changes are
correlated with the increasing thermal stabilities across the
homologous series of enzymes. In terms of thermostability,
the enzymes fall into two broad classes based on the
temperature at which the half-life equals 8 min: the psychro-
phile and pig enzymes at the lower end with temperatures of

45 °Cand58 °C, and the archaeal enzymes at the upper end
with temperatures of 87 °C, 95 °Cand100°C.
MATERIALS AND METHODS
Crystallization and structure solution
Recombinant SsCS(85) was purified as described previously
[19]. Crystallization trials were carried out using the
hanging-drop vapour diffusion method using the Hampton
Research Screens. A single rod-like crystal of approximate
dimensions 2 · 0.1 · 0.1 mm grew in a 6-lL drop contain-
ing 2 lL SsCS(85) (10 mgÆmL
)1
)with10m
M
citrate and
Correspondence to G. Taylor, Centre for Biomolecular Sciences,
University of St. Andrews, St. Andrews, KY16 9ST, UK.
E-mail:
(Received 3 July 2002, revised 8 October 2002,
accepted 4 November 2002)
Eur. J. Biochem. 269, 6250–6260 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03344.x
CoA, 2 lLof100m
M
Tris/HCl, pH 7.2, containing 17%
(v/v) PEG 8K, and 2 lLof0.1
M
CaCl
2
.Thecrystalgrewin
a partially dried out drop after six months. X-ray data were
collected at room temperature on a 30-cm Mar image plate

detector. Diffraction extended to 2.7 A
˚
resolution. The
crystal was translated stepwise perpendicular to the beam to
maximize the completeness of the data and to overcome
radiation damage of the crystal. The data were reduced and
scaled using
DENZO
/
SCALEPACK
[22] (Table 2). The asym-
metric unit of the P2
1
unit cell contains two dimers with a
solvent content of 51%. The structure of SsCS(85) was
solved by molecular replacement using the program
AMORE
[23]. Because the crystallization solutions contained both
citrate and CoA, it was assumed that the closed form of
SsCS(85) had crystallized; therefore, initial attempts were
made to solve the structure using the closed structures of
Pf CS(100) or ArCS(31) as the search model, but this did not
produce any clear solutions. Attempts were subsequently
made using the open structure of the TaCS(55) dimer as the
search model. Using data in the resolution range of 15–6 A
˚
and a Patterson integration radius of 25 A
˚
, 50 solutions
from the rotation function were calculated. Using the same

resolution range for the translation search, the top solution
(33rd highest from the rotation search) had a correlation
coefficient (CC) of 32.0 and R-factor of 53.4% (compared
with the next highest peak with a background CC of 23 and
R-factor of 56%). This solution was fixed, and a solution
for the second dimer in the asymmetric unit was identified
(CC of 37.7% and R-factor of 51.9%, compared to the next
highest peak of 31 and 53%, respectively). After a rigid-
body refinement in
AMORE
of the two dimers, the final
solutions had a CC of 56.6 and R-factor of 41.3%. The
failure to find a solution using the closed form of the
homologous enzyme, but a clear solution with the open
forms, strongly suggested that the SsCS(85) had unexpect-
edly crystallised in the open, unliganded form.
Refinement and validation
The restrained refinement of SsCS(85) was performed using
REFMAC
[24]. The initial R-factor in
REFMAC
(after rigid
body refinement) was 48.3% (R
free
¼ 48.6%) and final
R-factor of 20.8% (R
free
¼ 28.5%) for all data from 20.0 A
˚
to 2.7 A

˚
1,2
. Tight non crystallographic symmetry (NCS)
restraints for both main-chain and side-chain were used
initially and six cycles of refinement carried after which the
R-factor was 36.3% (R
free
¼ 40.5%). Keeping the tight
NCS restraints, individual isotropic B-factor refinement was
then carried out, bringing the R-factor down to 24.7% (R
free
31.2%), after which the NCS restraints were gradually
loosened and the four monomers were built independently.
NCS restraints were controlled in
PROTIN
and, during the
refinement procedure, side-chain followed by main-chain
restraints were gradually loosened, with a final round
removing the NCS restraints continuing to lower the R
free
value.
The first two residues at the N-terminus and last seven
residues of the C-terminal arm were not seen in the poorly
defined electron density of these parts of the structure in all
four monomers. One conflict with the sequence data was
residue 57, which had been assigned as arginine and was
found from analysis of the electron density map to be a
proline (this is a totally conserved proline in all the other
known CSs). The position of the small domain with respect
to the large domain in SsCS(85) is the same as previously

observed in TaCS(55) [17]; this together with the absence of
density for substrates in the active site, supports the previous
speculation that the structure is the ÔopenÕ form of the
enzyme. After 24 rounds of refinement in
REFMAC
, the final
R-factor was 20.8% (R
free
¼ 28.5%) (Table 2). The quality
ofthefinalelectrondensityisshowninFig.1.
Table 2. Table displaying native data collected and refinement statistics
for the SsCS. Data in parentheses correspond to the high resolution
data shell (2.82–2.71 A
˚
).
Space group P2
1
Unit cell dimensions a ¼ 77.3 A
˚
,b¼ 97.9 A
˚
,
c ¼ 119.3 A
˚
, b ¼ 107.6°
Resolution limit 2.7 A
˚
Data completeness 88.6% (91.2%)
R
merge

7.2% (22.9%)
I/rI 9.37 (3.25)
Total No. of reflections 148169
Unique reflections 46758
R-factor 20.8%
Free R-factor 28.5%
No. protein atoms 11 742
Rmsd bond lengths (A
˚
) 0.009
Rmsd bond angles (°) 0.032
Table 1. CS structures used for analysis.
Source
organism
Optimum growth
temperature
(°C)
CS
Temperature (°C) at which
the half-life equals 8 min
Substrates in
crystal structure
Data resolution
(A
˚
)
Arthrobacter DS23R 31
a
45 Citrate and CoA 2.1
Pig 37 58 Citrate only

Citrate and CoA
2.7
2.0
Thermoplasma acidophilum 55 87 – 2.5
Sulfolobus solfataricus 85 95 – 2.7
Pyrococcus furiosus 100 100 Citrate & CoA 1.9
a
It should be noted that, although Arthrobacter DS23R was isolated from a habitat temperature of approximately 0 °C, this organism
displays a relatively high optimum growth temperature; therefore, although it is described here as psychrophilic, it should perhaps more
correctly be referred to as psychrotolerant.
Ó FEBS 2002 Citrate synthase from psychrophile to hyperthermophile (Eur. J. Biochem. 269) 6251
The Ramachandran plot shows that for the four mono-
mers in the asymmetric unit, 91.3% of residues lie in the
most favoured regions, with only 8.7% in the allowed
regions and no residues appearing in the generously allowed
or disallowed regions (excluding glycine and proline
residues). Atomic coordinates have been deposited in the
Protein Databank with accession code 107x
3
.
RESULTS
Overall structural comparisons
All the eukaryotic
4
, archaeal and Gram-positive bacteria CSs
are homo-dimeric structures with each monomer consisting
of a large and small domain. In addition, they are almost
entirely a-helical; the pigCS contains 20 a-helices (A-T) with
theenzymefromArCS(31), TaCS(55), SsCS(85) and
PfCS(100) containing 16 a-helices, all of which have an

equivalent in pigCS (helices A, B, H, and T are not present)
(Figs 2 and 3). Of the 16 equivalent helices, the large domain
comprises 11 helices (C-M and S) and the small domain five
helices (N-R). The small domain has been classed as
residues 217–321 inclusive for SsCS(85).
The active sites of CSs comprise residues from both
monomers and therefore CS is only active as a dimer,
stressing the importance of maintaining dimeric integrity as a
prerequisite for activity. Binding of citrate and CoA to the
active site has been discussed in detail for PfCS(100) and
ArCS(31), and the differences with respect to the pig enzyme
noted [15,18]. The SsCS(85) structure has no substrate
bound, but the location of active site residues can be
identified by comparison with the liganded PfCS(100)
structure. The citrate-binding residues comprising three
arginine residues, R267 (helix P), R338 (helix S) and R358¢
(where the prime denotes the residue of the second mono-
mer), and three histidine residues, H183 (loop K-L), H218
(loop M-N) and H258 (loop O-P), are equivalent to those
found in PfCS(100). The binding residues for the three
Fig. 2. Structurally based sequence alignment of the five CSs discussed.
Helices A to T are shaded, and the location of the small domain is
indicated by lowercase sequence letters. The three catalytic residues are
indicated by fl. The three arginines and histidines involved in binding
citrate are marked with a C. The residues involved in binding the three
phosphates of CoA are marked with an A. The sequence numbering is
shownatthestartofeachline.
Fig. 1. Stereo-diagram showing a typical region of the final 2Fo-Fc electron density map contoured at 1 r.
6252 G. S. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002
phosphate groups of CoA are likely to be K250 (loop O-P)

and K306 (loop Q-R), R259 and K262 (both loop O-P), with
the third phosphate being co-ordinated by R355¢ from the
second monomer. The catalytic residues H218, H258 and
D313 (loop Q-R) are also present in SsCS(85) and are in
a similar position to the PfCS(100) residues. It is likely
therefore that SsCS(85) binds substrates in a similar man-
ner to PfCS(100) and that the mechanism of catalysis is the
same.
The dimer interface of all the CSs is made up of two parts
and comprises residues solely in the large domain; the main
part is the eight a-helical sandwich of four antiparallel pairs
of helices (F, G, M and L), with the second being the
additional interaction of N- and C-terminal regions (Fig. 3).
The pigCS is different from the other four CSs in terms of
the topology of the C-terminal region. In the other four, the
C-terminal arm of one monomer wraps around the other
monomer, clasping the two together [18], and results in
more extensive interactions, including those with the
N-terminus. It is important to note that as the C-terminal
arms of the TaCS(55) and SsCS(85) are not complete in the
structures (see below), there may be additional interactions
present that have not been observed. This also suggests that
the C-terminal arm seems to be ordered only in the presence
of substrates.
Sequence and structural statistics
Pairwise sequence alignments were carried out using the
program
BESTFIT
from the Wisconsin
GCG

sequence analysis
package, and superposition was carried out using the least
squares fit in
O
[25] for fitting of alpha-carbon atoms
(starting from three conserved atoms). These statistics are
listed in Table 3.
Sequence identities between the various CSs for which
3D-structures have been determined range from 20%
[eukaryotic
5
vs. bacterial or archaeal) to 60% (SsCS(85)
and TaCS(55)]. These identities are reflected in the root
mean square (RMS) deviations between the alpha-carbons
of the structures, with the most similar structures being the
TaCS(55) and SsCS(85), and with the PfCS(100) and
ArCS(31) pair also showing a very low RMS deviation. As
some structures are in the open conformation and some
have substrates bound, the large and small domains of each
enzyme were compared separately (Table 3); in general,
such an analysis shows the same trend as that for the whole
dimer but the small domains tend to be more highly
conserved. As is suggested later, this may correlate with
differences particularly relating to the dimer interface, to
which the small domain does not contribute, and may reflect
the fact that the majority of the substrate-binding and
catalytic residues are from the small domain.
The molecular mechanisms underlying protein thermal
stability
In our comparison of CS atomic structures from organisms

spanning a wide-range of growth temperatures, the deter-
mination of the SsCS(85) structure fills an important gap
between the enzyme from Thermoplasma (55 °C) and
Pyrococcus (100 °C). With the structure reported in this
paper, we can now look for trends in the structures that
might correlate with the increasing thermostabilities of these
enzymes. However, the complex nature of the stabilization
of a protein structure lends itself to many types of
comparative analysis, and the results presented below are
those where significant differences exist between the struc-
tures. Other types of analysis (e.g. of hydrogen bonds and
helix capping) have been performed but are not included
Fig. 3. Schematic drawings of CS. Fromtoptobottom:ArCS(31),
PigCS(37), TaCS(55), SsCS(85) and PfCS(100). The right hand col-
umn represent views obtained by rotating the images in the left hand
column by 90° about a horizontal axis. N- and C-terminii are denoted
by blue and red spheres, respectively. The small domains (helices N to
R) are coloured in a lighted shade. Catalytic residues, and citrate and
Co-A where appropriate, are shown in ball-and-stick representation.
All figures were created using
BOBSCRIPT
[51] and
GL
_
RENDER
(L. Esser
and J. Deisenhofer, unpublished results).
Ó FEBS 2002 Citrate synthase from psychrophile to hyperthermophile (Eur. J. Biochem. 269) 6253
due to there being no significant differences between the
different structures.

Compactness and surface characteristics
The accessible surface area was calculated using the program
GRASP
[26] and the volume and cavity detection were
determined using the program
VOIDOO
[27] with a probe
radius 1.4 A
˚
and grid spacing of 0.75 A
˚
. All calculations for
closed structures were carried out in the absence of substrate,
and the results are summarized in Table 4.
ArCS(31), TaCS(55) and PfCS(100) have very similar
surface areas, with that of SsCS(85) being slightly higher;
however, all four enzymes have a considerably smaller
surface area and volume than the pigCS(37), even when
deleting the first 35 residues from the pig enzyme (these 35
amino acids comprise helices A and B, which are absent in
the other CSs being considered). A similar pattern to the
total accessible surface area (ASA)
6
is found when compar-
ing the overall volume, with pigCS(37) having a consider-
ably larger volume than the other CSs (again, even when
calculated with the N-terminally deleted structure).
However, it is also notable that the smallest volume is
exhibited by the psychrophilic CS (8.36 · 10
4

A
˚
3
). All the
CSs have a similar percentage of atoms buried, although the
hyperthermophilic PfCS(100) exhibits the highest with
54.5%.
Examination of the exposed hydrophobic area shows a
more obvious trend. Despite all the archaeal and bacterial
CSs having a similar overall ASA, there is quite a difference
in hydrophobic exposure when comparing ArCS(31) with
the other CSs; the closed conformation of the ArCS(31) has
7854 A
˚
2
overall exposed hydrophobic area (representing
29% of the total ASA) compared with the closed PfCS(100)
with only 4942 A
˚
2
(18% of total ASA). Thus, on average,
ArCS(31) exposes 23 A
˚
2
per hydrophobic residue compared
with 16 A
˚
2
per hydrophobic residue in PfCS(100). The
total amount (A

˚
2
) of hydrophobic surface area shows a
decrease as the thermostability of the protein increases, a
trend observed in other structural comparisons [4]. SsCS(85)
also follows the trend observed in PfCS(100), with the
elimination of all cavities capable of accommodating a
solvent molecule, indicating that this is a prerequisite for
maintaining integrity at high temperatures. The number of
internal cavities (and their total volumes in A
˚
2
calculated by
VOIDOO
) are 1 (104), 6 (476), 3 (218), 3 (184), 0 (0) and 0 (0)
Table 3. Overall comparison of primary and 3D structures of CSs. In the top half of the table the RMS deviations between Ca atoms (in A
˚
)aregiven
for complete dimers, the large domain and the small domain, with the number of contributing pairs of Ca atoms in parantheses. In the bottom half
of the table, the percentage sequence identities and similarities are shown, the latter in parentheses.
Enzyme
(open)
ArCS(31)
(closed) PigCS(37) PigCS(37) TaCS(55) SsCS(85) PfCS(100)
ArCS(31) – 2.27 (560) 2.12 (630) 1.97 (604) 1.94 (610) 1.32 (719)
1.74 (242) 1.53 (252) 1.57 (259) 1.07 (262)
1.79 (96) 1.59 (90) 1.51 (90) 1.27 (92)
PigCS(37) (open) – – 1.19 (730) 1.95 (651) 1.88 (646) 2.15 (550)
2.16 (533) 2.08 (519) 2.04 (631)
PigCS(37) (closed) 27% (50%) – – 1.81 (233) 1.79 (232) 1.84 (245)

1.60 (82) 1.55 (82) 1.49 (90)
0.87 (719) 2.03 (581)
TaCS(55) 32% (54%) – 22% (48%) – 0.76 (256) 1.66 (247)
0.76 (104) 1.02 (95)
SsCS(85) 34% (55%) – 27% (50%) 59% (76%) – 1.94 (597)
1.53 (245)
1.03 (96)
PfCS(100) 40% (58%) – 31% (53%) 42% (62%) 46% (67%) –
Table 4. Accessible surface area (ASA) and volume statistics of CSs.
CS
ArCS(31)
(closed)
PigCS(37)
(open)
PigCS(37)
(closed)
PigCS(37)
(closed,
N-terminally
deleted)
TaCS(55)
(open)
SsCS(85)
(open)
PfCS(100)
(closed)
ASA ( · 10
4
A
˚

2
) 2.72 3.34 3.20 2.99 2.72 2.82 2.72
No. of atoms calculated for 5784 6888 6884 6344 5722 5879 5961
No. of atoms buried 3044 3469 3601 3307 2955 3014 3248
Atoms buried (%) 52.6 50.4 52.3 52.1 51.6 51.3 54.5
Volume ( · 10
4
A
˚
3
) 8.36 9.96 9.98 9.18 8.71 8.51 8.65
Total area hydrophobic exposed (A
˚
2
) 7854 6654 6246 – 6001 5513 4942
% Hydrophobic of total ASA 29 20 20 – 22 20 18
6254 G. S. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002
for ArCS(31), PigCS(37) open, PigCS(37) closed, TaCS(55),
SsCS(85) and PfCS(100), respectively.
The subunit interface: ion pairs, hydrophobicity
and complementarity
Ion pairs were classed as residues of opposite charge
situated 4.0 A
˚
or less apart [28]. Looking simply at the total
numbers of ion pairs, it can be seen that all the thermophilic
CSs have a greater total number of ion pairs than the pig
enzyme, but that the psychrophilic enzyme actually has the
greatest number of all (Table 5). Looking then at the trends
towards inter/intrasubunit ion pairs, PfCS(100) has the

most interfacial ion pairs but both ArCS(31) and pigCS(37)
have more intrasubunit interactions than the TaCS(55) and
SsCS(85), and therefore the location of these ionic inter-
actions may be particularly relevant.
With respect to the ionic interactions at the dimer
interface (Fig. 4), the five CSs show considerable variation.
Interactions in the pig enzyme are all unique with respect
to the other CSs and involve residues near the termini and
outer helices (F and L) of the eight-helical sandwich; the
central helices of the dimer interface have no ionic
interactions associated with them. In marked contrast,
the other four CSs all have ionic interactions associated
with the two internal helices of the interface (G and M) as
well as those involved with the N-terminus and the C-
terminal arm. Firstly, with respect to the interfacial helices,
one completely conserved ionic interaction in ArCS(31),
TaCS(55), SsCS(85) and PfCS(100) is that between a
conserved aspartate at the N-terminal end of helix M with
a lysine at the C-terminal end of helix M in the other
subunit (D205 and K218 in TaCS).Thelysineresidueis
not conserved in the pig enzyme. This leads to a single ion
pair at each end of helix M in TaCS(55), whereas in
SsCS(85) there is an additional salt bridge in close
proximity with the first one, between E89 (loop F-G) of
one monomer and K108 at the C-terminal end of helix G
in the other; thus in SsCS(85) both central helices have ion
pairs at either end. In PfCS(100), the G-M interhelical
electrostatic interactions are even more pronounced than
in either TaCS(55) or SsCS(85), with a five-residue ionic
network comprising H93 and R99 (helix G) and D113 in

loop G-I, in addition to the above-mentioned Asp-Lys
pair.
Interestingly, in the psychrophilic ArCS(31) the first
Asp-Lys ion pair is part of a four-residue network (in
conjunction with D95 and R98, both in helix G). The four
residue network in ArCS(31) only comprises two single
interactions directly across the interface, whereas the
PfCS(100) five-residue network has four such interactions.
This would suggest that the PfCS(100) networks contri-
bute considerably more to the intermolecular interactions
than they do in the psychrophilic enzyme. Even so, the
psychrophilic enzyme does have an intersubunit ionic
network, which may be related to cold-stability in the face
of diminished hydrophobic interactions at very low
temperatures [15]. The ionic interactions at the central
helices G and M of the five CS structures are shown in
Fig. 5.
The nature of the dimer interface of SsCS(85) differs from
PfCS(100) in that there is a higher degree of hydrophobic
interactions (Table 6). Similar levels to those in SsCS(85)
are also observed in TaCS(55). All the thermophilic CSs
show a low value for the Ôgap volume indexÕ,theratioofthe
gap volume to the accessible surface area of the interface, an
indication that these proteins exhibit greater surface com-
plementarity at the interface compared with ArCS(31) and
pigCS(37).
Finally, examining the part of the dimer interface near the
active site, all the archaeal CSs and the ArCS(31) have ionic
interactions that also tend to stabilize the N-terminus;
however, PfCS(100) certainly has the most extensive ionic

interactions with two four-residue networks (E189,D12,
R356¢ and R358¢) and two single ion pairs (E11-R353¢). A
cluster of six isoleucines, three from each monomer, was
observed in this region for PfCS(100) [18], indicative of a
strong hydrophobic interaction in this region. Two of the
three isoleucines are conserved in TaCS(55) and SsCS(85),
residues 15 and 357, suggesting a similar role in these
enzymes.
Ionic interactions at the termini
Ionic interactions may also be important for prevention of
fraying of the N- and C-termini; several of these interactions
Table 5. Total number of ion pairs and ion pair networks in CS. A and B refer to the two subunits of the dimer.
ArCS(31)
PigCS(37)
(closed) TaCS(55) SsCS(85) PfCS(100)
Total Ion Pairs 52 36 43 45 43
Intra A 21 12 18 21 14
Intra B 21 12 21 18 12
Inter AB 10 12 4 6 17
Intra A networks
3 residue 2 1 0 3 0
4 residue 1 0 1 1 1
5 residue 0 1 0 1 0
Inter AB networks
2 residue 5 7 2 4 9
3 residue 1 2 0 1 0
4 residue 1 1 1 0 1
5 residue 0 0 0 0 1
Ó FEBS 2002 Citrate synthase from psychrophile to hyperthermophile (Eur. J. Biochem. 269) 6255
interlink the two terminal regions, and therefore they may

have additional relevance to the strength of the subunit
interactions. The lengths of the C-terminal arms vary, with
ArCS(31) being six residues shorter than those of PfCS(100)
and TaCS(55), and five shorter than SsCS(85). ArCS(31)
has fewer interactions of the C-terminal arm with the other
monomer than the three thermophilic CSs (including one
ion pair that appears to anchor the end of the arm: R375-
E48¢ in PfCS(100)). R375 and E48 are conserved in
TaCS(55) and SsCS(85) suggesting the likelihood of this
ion pair being present at their C-termini. ArCS(31) also has
an arginine residue (R375) which interacts with E56¢ but, as
this residue is four residues from the end of the C-terminus,
there may be more chance of fraying of this terminal arm in
the psychrophile. Both N-termini in PfCS(100) also have an
interconnecting ion pair (K8-D16¢) but this is a three residue
interaction in ArCS(31) (K7, D15¢ and D359¢). SsCS(85)
also has several terminal interactions (E9,R259 and R355¢)
but TaCS(55) does not.
Loop regions: length and ionic interactions
It has previously been suggested that loop regions tend to be
the most flexible regions within a protein, and are therefore
often the first areas to be subject to proteolytic cleavage or
heat denaturation [29]. It is possible therefore that increased
thermostability may be achieved by shortening loops or by
additional interactions stabilizing these regions.
The equivalent loop regions of the five CSs have
therefore been compared (several extra loops are present in
the pig enzyme). Although some of the differences in loop
conformations (particularly near the active site) may be
due to the open or closed nature of the structures, it is

immediately obvious on comparison of the ionic interac-
tions that there is considerable difference between the five
enzymes. There is a distinct absence of ionic interactions in
the loops of the pig enzyme, with the thermophilic and
psychrophilic CSs showing more extensive interactions.
Many of these ionic interactions are involved with the
dimer interface as well as those that interconnect one loop
region with another, thus possibly ÔpinningÕ loops together.
The total number of ionic interactions present within the
loop and dimer interface regions of the five CSs is:
ArCS(31) 12, pigCS(37) 8, TpCS(55) 16, SsCS(85) 24 and
PfCS(100) 18. As in evident, the thermophilic CSs have the
highest number of these interactions, with SsCS(85) having
the most. However, the interactions in PfCS(100) tend to
be more complex, perhaps affording a greater degree of
stabilization in particular areas of the protein structure;
also, as detailed below, the PfCS(100) often has the
shortest loops, reducing the need for a large number of
ionic interactions in those specific regions. The results for
individual loops are:
Loop E-F. The loop in pigCS(37) is one residue shorter
than in ArCS(31), SsCS(85), and PfCS(100), and two
residues shorter than TaCS(55), but the loops superimpose
Fig. 4. Residues involved in ion-pair interactions at the dimer interface.
Spheres are drawn for at the Ca atoms of acidic (red) and basic (blue)
side-chains involved in intersubunit ionic interactions. From top to
bottom: ArCS(31), PigCS(37), TaCS(55), SsCS(85) and PfCS(100).
6256 G. S. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002
well and there are no ionic interactions in pigCS(37)
compared with a single ion pair in ArCS(31). TaCS(55)

has a four residue intramolecular network and SsCS(85) has
a five residue network that involves a residue at the
N-terminal end of the loop.
Loop I-J. All the enzymes have similarly large loops but
there are no ion pairs in pigCS(37) with one ion pair in
TaCS(55). ArCS(31), SsCS(85) and PfCS(100) all have
multiple ionic interactions linking loops I-J and J-K.
Loop J-K. The loop in pigCS(37) is slightly shorter than the
others and contains no interactions, whilst the TaCS(55)
loop has an ion pair. ArCS(31), SsCS(85) and PfCS(100)
have interactions linking this loop with the previous loop I-J.
Loop N-O. This appears to be a long and flexible loop in
ArCS(31) and contains six charged residues, but no ion
pairs. PigCS(37) also has a longer and more extended loop
than SsCS(85) and TaCS(55) (which both contain ion pairs)
and PfCS(100) has the shortest loop.
Loop O-P. Although not obvious from the length of loops
as designated by
PROMOTIF
[30], the loop in the pig enzyme
is considerably more extended than the others. ArCS(31),
pigCS(37) and PfCS(100) all have single ion pairs
stabilizing this loop (the interaction in the pigCS(37) loop
links it to loop B-C), with TaCS(55) and SsCS(85) loops
having multiple ionic interactions that link loops O-P and
K-L.
Loop P-Q. TaCS(55) and SsCS(85) loops both contain
ionic interactions. In the case of SsCS(85), this is in the form
of a three residue network linking it with loop J-K. This
loop is absent in PfCS(100).

Loop Q-R. This loop is shortest in ArCS(31) and it has
already been suggested that the reason for this is that it
seems to allow greater accessibility to the active site [15].
However, recent site-directed mutagenesis studies to
increase the length of this loop to mimic the situation in
the PfCS(100), reveal that the cold activity of the ArCS(31)
is not significantly compromised by the mutations [31].
DISCUSSION
The determination of the crystal structure of SsCS(85), and
its comparison with four other CSs from organisms that
essentially span the temperature range over which life exists,
have allowed a detailed structural analysis to be performed
to investigate the structural mechanisms underlying protein
thermal stability in this enzyme. This has been possible
because, in general, the 3D structures of the CSs are highly
similar and we have therefore not only been able to identify
specific differences, but have also succeeded in finding
trends in structural changes that correlate with increasing
thermostability of the individual proteins. These identified
structural differences are mainly concerned with the
protein’s compactness, both in general and in the loop
regions, and with the nature of the interactions at the dimer
interface, possibly indicating that the respective protein
thermostabilities are largely determined by these parts of the
protein.
General compactness
The compactness of heat-stable proteins has often been
found to be synonymous with their thermostability, and
can be described in a number of ways. There is a tendency
towards a smaller accessible surface area and volume when

comparing the thermophilic archaeal CSs with the pig-
CS(37), and the tendency towards fewer cavities should
also correlate with the improved hydrophobic packing of
these proteins. The increased complementarity of the dimer
interface, as measured by the gap volume index, in the
thermophilic enzymes may also be a significant feature.
Although the total percentage of atoms buried is similar
for all the CSs, the decreased burial of hydrophobic groups
of ArCS(31) compared with the other CSs probably reflects
the decreased entropic penalty of exposure of hydrophobic
side-chains at psychrophilic temperatures (reviewed by
[32,33]).
Loop regions
There is a tendency towards shorter (even absent) loop
regions in the thermophilic CSs, correlating with the
compactness of these proteins when compared with pig-
CS(37). This trend has also been highlighted by analysis of
mesophilic and thermophilic genome sequences, and was
suggested to be a general strategy for thermostabilization
[34]. However, many of the shorter loops in the thermophilic
CS are similarly short in ArCS(31) (apart from loop N-O).
A more dramatic difference in the loops is seen in the
Fig. 5. Diagram showing ionic interactions in the central helices (G and
M) of the dimer interface of ArCS, PigCS, TaCS, SsCS and PfCS.
Helices from different monomers are coloured blue and orange.
Table 6. The dimer interface of CS. Statistics are calculated using the
protein–protein interactions server (Jones and Thornton, 1995) for the
CS crystal structures with the C-terminal arm removed.
ArCS
PigCS

(closed) TaCS SsCS PfCS
Interface ASA ( A
˚
2
) 3403 4934 3154 3363 3698
% of total ASA 21.0 24.0 19.0 19.8 22.6
% polar atoms 34.8 37.8 32.6 32.3 39.2
% nonpolar atoms 65.2 62.2 67.4 67.7 60.8
Hydrogen bonds 44 42 24 28 54
Gap volume ( A
˚
3
) 10591 17164 6474 8474 9605
Gap volume index 1.52 1.74 1.03 1.26 1.29
Ó FEBS 2002 Citrate synthase from psychrophile to hyperthermophile (Eur. J. Biochem. 269) 6257
comparison of the ionic interactions in these regions: very
few are present in pigCS(37), but a large number occur in
the archaeal and bacterial proteins, with the thermophilic
CSs, particularly SsCS(85), having the most extensive
networks that cross-link loop regions. A significant increase
in the number of long-range (in sequence terms) electrostatic
interactions is also observed in SsCs(85) and PfCS(100),
where they serve to tether different parts of the structure
together. This compares with the observations of b-glyco-
sidase from S. solfataricus [35], which was shown to have
ionic interactions (specifically networks) over the surface
of the protein such that they cross-linked areas of
surface structure. Similarly, a mutational analysis of the
hyperthermostable indoleglycerol-phosphate synthase from
T. maritima, in which an ion-pair linking two a-helices was

disrupted, resulted in a less stable protein [36].
Ionic interactions and hydrophobicity
at the subunit interface
The archaeal and bacterial CSs have a higher total number
of ionic interactions than the pigCS(37), which in fact
exhibits the lowest percentage participation of charged
residues in ion pairs or networks of the five enzymes in the
comparison. The psychrophilic ArCS(31) actually has the
most ionic interactions, which we have suggested may be
related to cold stability [15], but with respect to subunit
association, PfCS(100) has the most extensive interactions
across the dimer interface whilst ArCS(31) has more than
either TaCS(55) or SsCS(85).
The eight-helical sandwich part of the dimer interface
shows a definite trend towards increasing hydrophobicity
going from ArCS(31) and pigCS(37) to TaCS(55) and
SsCS(85), and this may be indicative of the increasing
strength of the hydrophobic interaction with temperature,
at least to temperatures approaching 100 °C [37]. PfCS(100)
also has a greater degree of hydrophobicity in this region
than ArCS(31) and pigCS(37) but lower than the other two
thermophilic CSs, and this may be compensated by the
more extensive ionic interactions in the hyperthermophilic
protein. That is, the ionic interactions in the central helices
(G and M) of the eight-helical sandwich also show an
increase from none in pigCS(37), two single ion-pairs in
TaCS(55), four single ion-pairs in SsCS(85) and the two
five-residue networks in PfCS(100). ArCS(31) also has two
four-residue networks here, but these seem to be less
extensive than those in PfCS(100) (with fewer interactions

actually across the interface). PfCS(100) also has the
additional two four-residue networks near the active site
region (with the other archaeal and bacterial enzymes
displaying interactions to a lesser degree) as well as the
isoleucine cluster [18] which is partly conserved in TaCS(55)
and SsCS(85), again suggesting a better hydrophobic
packing than with either the mesophilic or psychrophilic
enzyme.
Finally, the parts of the dimer interface associated with
both termini seem to be stabilized by ionic interactions
particularly in the PfCS(100), but also to some degree in
ArCS(31) and SsCS(85).
These conclusions with respect to the ionic interactions at
the subunit interface and termini are supported by muta-
genesis studies [38]. Analysis of chimeric mutants between
the TaCS(55) and PfCS(100), where the large and small
domains were swapped, demonstrated that the determinants
of thermostability lie mainly with the large (subunit contact)
domain, possibly correlating with the trend of increasing
ionic interactions that are seen at the subunit interface as the
thermostability of the enzyme increases. Additionally,
mutagenesis of the PfCS(100) where we have disrupted
the ionic network at the subunit interface, and have
removed the C-terminal ionic interaction, support the role
of these electrostatic bonds in the stability of the enzyme. In
nearly all cases, the catalytic parameters of the mutants were
not significantly different from the wild-type enzyme,
supporting the contention that we have not grossly altered
the structure of the enzyme but have merely disrupted
stabilizing ionic interactions.

The importance of electrostatic interactions and their
precise location to stabilizing proteins has been shown in
other crystal structures of (hyper)thermostable proteins, as
discussed in the recent review by Karshikoff and Ladenstein
[10]. The most striking examples include glutamate dehy-
drogenase [39–41], glyceraldehyde 3-phosphate dehydro-
genase [42,43] and lumazine synthase [44]. Again, the
electrostatic strengthening of the intersubunit contacts is a
common theme in these proteins. Finally, computational
analyses [7,45,46] and genomic comparisons [8,9,47] add
further support to these findings.
Concluding remarks
The importance of the determination of the structure of the
SsCS(85) is principally that it ÔcompletesÕ aseriesofCS
structures from which we are now able to identify trends in
the structures of CSs that appear to be correlated with the
different degrees of thermostability. Our findings correlate
well with the growing number of studies that conclude that
ionic interactions stabilizing crucial areas of structure are
perhaps the most common
7
method of stabilization of
proteins at high temperatures, particularly for oligomeric
proteins. Recent thermodynamic studies on a mesophilic
and thermophilic pair of CheY proteins, have suggested that
a reduced change in heat capacity upon unfolding is a
possible indicator of thermostability [48], also supported by
studies on a mesophilic and thermophilic pair of Rnase H
proteins [49]. These studies suggest that it may be difficult to
dissect the contributions of individual interactions to

thermostability. This may be true for small monomeric
proteins such as CheY and RnaseH, but for oligomeric
proteins that make up > 85% of intracellular proteins, the
nature of the oligomer interface is key. Ionic networks at
interfaces, however, are not the exclusive means of gaining
thermostability, as the tetrameric triosephosphate isomerase
structure from P. furiosus has shown [50]. Nevertheless, for
the family of CSs presented here, increased ionic interac-
tions either between loops or at the dimer interface do
appear to correlate with increasing thermostability. The
results presented here lay the foundation for a suite of site-
directed mutagenesis experiments to investigate the precise
role of each of the sets of individual interactions in the five
dimeric CSs. Preliminary experiments that remove interac-
tions and destabilize the enzyme without affecting catalytic
activity have already been successfully carried out, but the
real ÔproofÕ of their importance is now to introduce ionic
bonds and networks into the less stable CSs to increase their
thermostability.
6258 G. S. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002
ACKNOWLEDGEMENTS
This work was supported by the Biotechnology and Biological Sciences
Research Council.
REFERENCES
1. Russell, R.J. & Taylor, G.L. (1995) Engineering thermostability:
lessons from thermophilic proteins. Curr. Opin. Biotechnol. 6, 370–
374.
2. Rees, D.C. & Adams, M.W.W. (1995) Hyperthermophiles: taking
the heat and loving it. Structure Folding Design. 3, 251–254.
3. Danson, M.J. & Hough, D.W. (1998) Structure, function and

stability of enzymes from the Archaea. Trends Microbiol. 6, 307–
314.
4. Maes, D., Zeelen, J.P., Thanki, N., Beaucamp, N., Alvarez, M.,
Thi, M.H., Backmann, J., Martial, J.A., Wyns, L., Jaenicke, R. &
Wierenga, R.K. (1999) The crystal structure of triosephosphate
isomerase (TIM) from Thermotoga maritima:acomparative
thermostability structural analysis of ten different TIM structures.
Proteins. 37, 441–453.
5. Sterner, R. & Liebl, W. (2001) Thermophilic adaptation of pro-
teins. Crit Rev. Biochem. Mol. Biol. 36, 39–106.
6. Vieille, C. & Zeikus, G.J. (2001) Hyperthermophilic enzymes:
Sources, uses, and molecular mechanisms for thermostability.
Microbiol. Mol. Biol. Rev. 65, 1–43.
7. Xiao, L. & Honig, B. (1999) Electrostatic contributions to the stab-
ility of hyperthermophilic proteins. J. Mol. Biol. 289, 1435–1444.
8. Kumar, S., Tsai, C.J. & Nussinov, R. (2000) Factors enhancing
protein thermostability. Protein Eng. 13, 179–191.
9. Cambillau, C. & Claverie, J M. (2000) Structural and genomic
correlates of hyperthermostability. J. Biol. Chem. 275, 32383–
32386.
10. Karshikoff,A.&Ladenstein,R.(2001)Ionpairsandthether-
motolerance of proteins from hyperthermophiles: a Ôtraffic ruleÕ for
hot roads. Trends Biochem. Sci. 26, 550–556.
11. Kurz, L.C., Drysdale, G.R., Riley, M.C., Evans, C.T. & Srere,
P.A. (1992) Catalytic strategy of citrate synthase: effects of amino
acid changes in the acetyl-CoA binding site on transition-state
analog inhibitor complexes. Biochemistry 31, 7908–7914.
12. Kurz, L.C., Drysdale, G., Riley, M., Tomar, M.A., Chen, J.,
Russell, R.J. & Danson, M.J. (2000) Kinetics and mechanism of
the citrate synthase from the thermophilic archaeon Thermo-

plasma acidophilum. Biochemistry 39, 2283–2296.
13. Evans, C.T., Kurz, L.C., Remington, S.J. & Srere, P.A. (1996)
Active site mutants of pig citrate synthase: effects of mutations on
the enzyme catalytic and structural properties. Biochemistry 35,
10661–10672.
14. Danson, M.J. & Hough, D.W. (2001) Citrate synthase from
hyperthermophilic Archaea. Methods Enzymol. 331, 3–12.
15. Russell,R.J.,Gerike,U.,Danson,M.J.,Hough,D.W.&Taylor,
G.L. (1998) Structural adaptations of the cold-active citrate syn-
thase from an Antarctic bacterium. Structure 6, 351–361.
16. Remington, S., Wiegand, G. & Huber, R. (1982) Crystallographic
refinement and atomic models of two different forms of citrate
synthase at 2.7 and 1.7 A
˚
resolution. J. Mol. Biol. 158, 111–152.
17. Russell, R.J., Hough, D.W., Danson, M.J. & Taylor, G.L. (1994)
The crystal structure of citrate synthase from the thermophilic
archaeon, Thermoplasma acidophilum. Structure. 2, 1157–1167.
18. Russell, R.J., Ferguson, J.M., Hough, D.W., Danson, M.J. &
Taylor, G.L. (1997) The crystal structure of citrate synthase from
the hyperthermophilic archaeon Pyrococcus furiosus at 1.9 A
˚
resolution. Biochemistry 36, 9983–9994.
19. Connaris, H., West, S.M., Hough, D.W. & Danson, M.J. (1998)
Cloning and overexpression in Escherichia coli of the gene
encoding citrate synthase from the hyperthermophilic Archaeon
Sulfolobus solfataricus. Extremophiles 2, 61–66.
20. Muir, J.M., Russell, R.J., Hough, D.W. & Danson, M.J. (1995)
Citrate synthase from the hyperthermophilic Archaeon,
Pyrococcus furiosus. Protein Eng. 8, 583–592.

21. Gerike, U., Danson, M.J., Russell, N.J. & Hough, D.W. (1997)
Sequencing and expression of the gene encoding a cold-active
citrate synthase from an Antarctic bacterium, strain DS2-3R. Eur.
J. Biochem. 248, 49–57.
22. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray diffrac-
tion data collected in oscillation mode. Method Enzymol. 276, 307–
326.
23. Navaza, J. (1994) AMoRe: an automated package for molecular
replacement. Acta Crystallogr. A50, 157–163.
24. Murshudov, G.N., Vagin, A.A. & Dodson, E.J. (1997) Refine-
ment of macromolecular structures by the maximum-likelihood
method. Acta Crystallogr. D53, 240–255.
25. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldaaard, M. (1991)
Improved methods for building protein models in electron density
maps and the location of errors in these models. Acta Crystallogr.
A47, 110–119.
26. Nichols, A., Sharp, K.A. & Honig, B. (1991) Protein folding and
association: insights from the interfacial and thermodynamic
properties of hydrocarbons. Proteins 11, 281–296.
27. Kleywegt, G.J. & Jones, T.A. (1994) Detection, delineation,
measurement and display of cavities in macromolecular structures.
Acta Crystallogr. D50, 178–185.
28. Barlow, D.J. & Thornton, J.M. (1983) Ion pairs in proteins.
J. Mol. Biol. 168, 867.
29. Daggett, V. & Levitt, M. (1993) Protein unfolding pathways
explored through molecular dynamics simulations. J. Mol. Biol.
232, 600–619.
30. Hutchinson, E.G. & Thornton, J.M. (1996) PROMOTIF – a
program to identify and analyze structural motifs in proteins.
Protein Sci. 5, 212–220.

31. Gerike, U., Danson, M.J. & Hough, D.W. (2001) Cold-active
citrate synthase: mutagenesis of active-site residues. Protein Eng.
14, 655–661.
32. Lonhienne, T., Gerday, C. & Feller, G. (2000) Psychrophilic
enzymes: revisiting the thermodynamic parameters of activation
may explain local flexibility. Biochim. Biophys. Acta. 1543, 1–10.
33. Sheridan, P.P., Panasik, N., Coombs, J.M. & Brenchley, J.E.
(2000) Approaches for deciphering the structural basis of low temp-
erature enzyme activity. Biochim. Biophys. Acta. 1543, 417–433.
34. Thompson, M.J. & Eisenberg, D. (1999) Transproteomic evidence
of a loop-deletion mechanism for enhancing protein thermo-
stability. J. Mol. Biol. 290, 595–604.
35. Aguilar, C.F., Sanderson, I., Moracci, M., Ciaramella, M., Nucci,
R., Rossi, M. & Pearl, L.H. (1997) Crystal structure of the beta-
glycosidase from the hyperthermophilic archeon Sulfolobus sol-
fataricus: resilience as a key factor in thermostability. J. Mol. Biol.
271, 789–802.
36. Merz, A., Knochel, T., Jansonius, J.N. & Kirschner, K. (1999) The
hyperthermostable indoleglycerol phosphate synthase from Ther-
motoga maritima is destabilised by mutational disruption of two
solvent-exposed salt bridges. J. Mol. Biol. 288, 753–763.
37. Dill, K.A. (1990) Dominant forces in protein folding. Biochemistry
29, 7133–7155.
38. Arnott, M.A., Michael, R.A., Thompson, C.R., Hough, D.W. &
Danson, M.J. (2000) Thermostability and thermoactivity of citrate
synthases from the thermophilic and hyperthermophilic archaea,
Thermoplasma acidophilum and Pyrococcus furiosus. J. Mol. Biol.
304, 657–668.
39. Yip, K.S., Stillman, T.J., Britton, K.L., Artymiuk, P.J., Baker,
P.J., Sedelnikova, S.E., Engel, P.C., Pasquo, A., Chiaraluce, R. &

Consalvi, V. (1995) The structure of Pyrococcus furiosus gluata-
mate dehydrogenase reveals a key role for ion-pair networks in
maintaining enzyme stability at extreme temperatures. Structure 3,
1147–1158.
Ó FEBS 2002 Citrate synthase from psychrophile to hyperthermophile (Eur. J. Biochem. 269) 6259
40. Britton, K.L., Yip, K.S., Sedelnikova, S.E., Stillman, T.J., Adams,
M.W., Ma, K., Maeder, D.L., Robb, F.T., Tolliday, N., Vetriani,
C., Rice, D.W. & Baker, P.J. (1999) Structure determination of the
glutamate dehydrogenase from the hyperthermophile Thermo-
coccus litoralis and its comparison with that from Pyrococcus
furiosus. J. Mol. Biol. 293, 1121–1132.
41.Knapp,S.,deVos,W.M.,Rice,D.&Ladenstein,R.(1997)
Crystal structure of glutamate dehydrogenase from the
hyperthermophilic eubacterium Thermotoga maritima at 3.0 A
˚
resolution. J. Mol. Biol. 267, 916–932.
42. Pappenberger, G., Schurig, H. & Jaenicke, R. (1997) Disruption of
an ionic network leads to accelerated thermal denaturation of
D
-glyceraldehyde-3-phosphate dehydrogenase from the hyper-
thermophilic bacterium Thermotoga maritima. J. Mol. Biol. 274,
676–683.
43. Isupov, M.N., Fleming, T.M., Dalby, A.R., Crowhurst, G.S.,
Bourne, P.C. & Littlechild, J.A. (1999) Crystal structure of the gly-
ceraldehyde-3-phosphate dehydrogenase from the hyperthermo-
philic archaeon Sulfolobus solfataricus. J. Mol. Biol. 291, 651–660.
44. Zhang, X., Meining, W., Fischer, M., Bacher, A. & Ladenstein, R.
(2001) X-ray structure analysis and crystallographic refinement of
lumazine synthase from the hyperthermophile Aquifex aeolicus at
1.6 A

˚
resolution: determinants of thermostability revealed from
structural comparisons. J. Mol. Biol. 306, 1099–1114.
45. Elcock, A.H. (1998) The stability of salt bridges at high temper-
atures: implications for hyperthermophilic proteins. J. Mol. Biol.
284, 489–502.
46. de Bakker, P.I., Hunenberger, P.H. & McCammon, J.A. (1999)
Molecular dynamics simulations of the hyperthermophilic protein
sac7d from Sulfolobus acidocaldarius: contribution of salt bridges
to thermostability. J. Mol. Biol. 285, 1811–1830.
47. Das, R. & Gerstein, M. (2000) The stability of thermophilic pro-
teins: a study based on comprehensive genome comparison. Funct.
Integr. Genomics. 1, 76–88.
48. Deutschman, W.A. & Dahlquist, F.W. (2001) Thermodynamic
basis for the increased thermostability of CheY from the hyper-
thermophile Thermotoga maritima. Biochemistry. 40, 13107–13113.
49. Hollien, J. & Marqusee, S. (1999) A thermodynamic comparison
of mesophilic and thermophilic ribonucleases H. Biochemistry 38,
3831–3836.
50. Walden, H., Bell, G.S., Russell, R.J., Siebers, B., Hensel, R. &
Taylor, G.L. (2001) Tiny TIM: a small, tetrameric, hyperthermo-
stable triosephosphate isomerase. J. Mol. Biol. 306, 745–757.
51. Esnouf, R.M. (1997) BOBSCRIPT: an extensively modified ver-
sion of MOLSCRIPT that includes greatly enhanced coloring
capabilities. J. Mol. Graph. Model. 15, 132–143.
6260 G. S. Bell et al. (Eur. J. Biochem. 269) Ó FEBS 2002