Báo cáo khoa học: The crystal structure of a hyperthermostable subfamily II isocitrate dehydrogenase from Thermotoga maritima pdf

The crystal structure of a hyperthermostable subfamily II
isocitrate dehydrogenase from Thermotoga maritima
Mikael Karlstro
¨
m
1
, Ida H. Steen
2
, Dominique Madern
3
, Anita-Elin Fedo
¨
y
2
, Nils-Ka
˚
re Birkeland
2
and Rudolf Ladenstein
1
1 Center for Structural Biochemistry, Karolinska Institutet, NOVUM, Huddinge, Sweden
2 Department of Biology, University of Bergen, Norway
3 Institut de Biologie Structurale CEA-CNRS-UJF, Grenoble, France
The enzymes of hyperthermophilic organisms are
remarkably stable and can resist denaturation at tem-
peratures ranging from 80 °C to above 130 °C [1,2],
whereas their counterparts from mesophilic organisms
are usually denatured at around 50 °C. However, they
are often homologous and their catalytic mechanisms
are usually identical [3]. Moreover, the gain in free sta-
bilization energy in hyperthermostable proteins, com-

pared with their mesophilic homologues, is generally
small, especially at the growth temperature of the
organisms [4]. Typically, the net free energy of stabil-
ization in both mesophilic and hyperthermophilic pro-
teins, is 5–20 kcalÆmol
)1
, which is equivalent to only
a small number of weak intermolecular interactions
[3,5,6]. There is no single mechanism or structural fea-
ture that is responsible for the high thermotolerance of
hyperthermostable proteins [4,7]. The most common
determinants of hyperthermostability that have been
Keywords
ionic networks; isocitrate dehydrogenase;
thermostability; Thermotoga maritima
Correspondence
M. Karlstro
¨
m, Karolinska Institutet, NOVUM,
Centre for Structural Biochemistry,
S-141 57 Huddinge, Sweden
Fax: +46 8 608 9290
Tel: +46 8 608 9178
E-mail:
(Received 28 February 2006, revised
26 April 2006, accepted 2 May 2006)
doi:10.1111/j.1742-4658.2006.05298.x
Isocitrate dehydrogenase (IDH) from the hyperthermophile Thermotoga
maritima (TmIDH) catalyses NADP
+

- and metal-dependent oxidative
decarboxylation of isocitrate to a-ketoglutarate. It belongs to the b-decarb-
oxylating dehydrogenase family and is the only hyperthermostable IDH
identified within subfamily II. Furthermore, it is the only IDH that has
been characterized as both dimeric and tetrameric in solution. We solved
the crystal structure of the dimeric apo form of TmIDH at 2.2 A
˚
. The
R-factor of the refined model was 18.5% (R
free
22.4%). The conformation
of the TmIDH structure was open and showed a domain rotation of 25–
30° compared with closed IDHs. The separate domains were found to be
homologous to those of the mesophilic mammalian IDHs of subfamily II
and were subjected to a comparative analysis in order to find differences
that could explain the large difference in thermostability. Mutational stud-
ies revealed that stabilization of the N- and C-termini via long-range elec-
trostatic interactions were important for the higher thermostability of
TmIDH. Moreover, the number of intra- and intersubunit ion pairs was
higher and the ionic networks were larger compared with the mesophilic
IDHs. Other factors likely to confer higher stability in TmIDH were a less
hydrophobic and more charged accessible surface, a more hydrophobic
subunit interface, more hydrogen bonds per residue and a few loop dele-
tions. The residues responsible for the binding of isocitrate and NADP
+
were found to be highly conserved between TmIDH and the mammalian
IDHs and it is likely that the reaction mechanism is the same.
Abbreviations
AfIDH, Archaeoglobus fulgidus IDH; ApIDH, Aeropyrum pernix IDH; AUC, analytical ultracentrifugation; BsIDH, Bacillus subtilis IDH; EcIDH,
Escherichia coli IDH; HcIDH, human cytosolic IDH; HDH, homoisocitrate dehydrogenase; IDH, isocitrate dehydrogenase; PcIDH, porcine

heart mitochondrial IDH; PfIDH, Pyrococcus furiosus IDH; TmIDH, Thermotoga maritima IDH.
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2851
observed are more ionic interactions at the protein sur-
face and increased formation of large ionic networks
[8–10]. Electrostatic optimization and the reduction in
repulsive charge–charge interactions are crucial [11,12].
However, in many cases, the combined effects of
multiple, modestly stabilizing interactions seem to be
responsible for enhanced thermostability. Examples
include a reduction in the hydrophobic accessible sur-
face area, increased hydrogen bonding, stronger inter-
subunit interactions, loop deletions and structural
compactness [13–15].
In order to analyse thermotolerance comparatively
within one protein family, we chose isocitrate dehydrog-
enase (IDH), a metal-dependent (Mg
2+
or Mn
2+
)
enzyme in the tricarboxylic acid cycle which catalyses
the subsequent dehydrogenation and decarboxylation
of isocitrate to a-ketoglutarate using NAD
+
or
NADP
+
as a cofactor [16]. Owing to its central role in
metabolism, IDH is present in organisms from all three
domains of life, Archaea, Bacteria and Eukarya. Conse-

quently, IDH is also present in organisms that have
adapted to a wide range of growth temperatures making
it an attractive model enzyme for studying heat-adap-
tive traits. In a previous publication, we characterized
hyperthermostable IDHs from Thermotoga maritima
(TmIDH), Aeropyrum pernix (ApIDH), Pyrococcus
furiosus (PfIDH) and Archaeoglobus fulgidus (AfIDH)
with respect to phylogenetic affiliation, cofactor
specificity, thermostability and oligomeric state, and
identified three different subfamilies of IDH [17].
ApIDH, AfIDH and PfIDH showed high sequence
identity to IDH from the mesophile Escherichia coli
(EcIDH) and they formed, together with all known
archaeal IDHs as well as most bacterial IDHs,
subfamily I. Within subfamily II, the bacterial and
eukaryotic NADP
+
-IDHs were grouped into different
branches. However, TmIDH was separated from both
and represented the deepest branch of this subfamily.
ApIDH in subfamily I was described as the most
thermostable IDH, with an apparent melting tempera-
ture of 109.9 °C and TmIDH was described as the
only hyperthermostable IDH known within sub-
family II with an apparent melting temperature of
98.3 °C. Moreover, we identified a heterogeneous mix-
ture of tetrameric and dimeric species of TmIDH in
solution, in which the tetrameric form of TmIDH repre-
sented a unique oligomeric state of NADP-IDH. Here,
we report the crystal structure of TmIDH, representing

the first bacterial structure of an IDH from
subfamily II. The crystallographic structure is presented
in the dimeric form as crystallization trials of the
tetrameric form of TmIDH have been unsuccessful to
date.
In order to reveal possible determinants of the
increased thermotolerance of TmIDH, we compared
the structure of the dimeric form with the mesophilic
mammalian homologues porcine mitochondrial IDH
(PcIDH) and human cytosolic IDH (HcIDH) from the
same subfamily. The observed differences were then
compared with differences between ApIDH and its
mesophilic homologue EcIDH in subfamily I. Further-
more, the analysis was used as a guideline to design
specific mutants of TmIDH. Their properties are dis-
cussed with respect to the main mechanisms involved
in protein thermostabilization.
Results and Discussion
Purification of the dimeric form of TmIDH
We previously described TmIDH as a heterogenous
mixture of dimeric and tetrameric species [17].
Recently, the crystal structure of tetrameric homoisoci-
trate dehydrogenase from the hyperthermophilic bac-
terium Thermus thermophilus was solved and it was
suggested that formation of the tetramer was involved
in thermostabilization [18]. New analytical ultracentrif-
ugation (AUC) data on TmIDH confirmed that two
oligomeric species with S
20,W
values corresponding to

a tetramer ( 7.9 S) and a dimer ( 5.0 S) are present
in the preparation using the previously described puri-
fication protocol (data not shown) [17].
In order to separate the two oligomeric forms, a gel-
filtration step was added and the separate fractions
were analysed using AUC. The AUC data demonstra-
ted that dimeric and tetrameric species do not re-
equilibrate to a heterogenous mixture (Fig. 1). Thus,
the dimeric and tetrameric forms were not in equilib-
rium and could be separated. The two forms were con-
centrated and subjected to various crystallization trials.
To date, crystals of the tetrameric species of TmIDH
0
0,1
0,2
0,3
0,4
0,5
0,6
051015
Sedimentation coefficient (S)
c(s) arbit. units
Fig. 1. Sedimentation velocity analysis of dimeric Thermotoga mari-
tima IDH, (TmIDH) at 20 °C. The data recorded at 0.1 mgÆmL
)1
,in
50 m
M NaCl buffered with 50 mM Tris ⁄ HCl pH 8 were fitted using
SEDFIT software [62]. The single peak is centered on 5,4 S. This
value corresponds to the one expected for a pure dimer TmIDH.

Structure and thermal stability of T. maritima M. Karlstro
¨
m et al.
2852 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet
have not been obtained, whereas the dimeric form pro-
duced good quality crystals that were used to solve the
structure.
Quality and description of the model
The crystal structure of TmIDH was solved by molecu-
lar replacement at 2.2 A
˚
and represents the apo form
of the enzyme. The atomic coordinates and structure
factors were deposited in the Protein Data Bank (entry
1ZOR). The model (Fig. 2A) was refined to a crystal-
lographic R-value of 18.4% and a free R-value of
22.3%. It was crystallized as a dimer in the asymmetric
unit with a solvent content of 54.8% corresponding to
a Matthews’ coefficient of 2.7 A
˚
3
ÆDa
)1
. The two sub-
units were related by a twofold noncrystallographic
rotation axis and are referred to as subunits A and B.
The space group was P2
1
2
1

2
1
and the unit cell parame-
ters a ¼ 62.5 A
˚
,b¼ 88.1 A
˚
,c¼ 180.9 A
˚
. The sym-
metry-related molecules in the crystal did not produce
the tetrameric form. Electron-density maps of subunit
A were of very high quality throughout the structure
determination, whereas density maps of the large
domain of subunit B lacked density for many side
chains. This was probably caused by a local disorder
in the crystal resulting from the flexibility of the large
domains in combination with the particular crystal
packing. However, these side chains were maintained
in the model with ideal conformations (as defined by
the rotamers of the o database) except for residues
B1–3, which were omitted because of dubious main
chain density. In subunit A, only three side chains
were lacking density and more solvent molecules were
built compared with subunit B. In total, 428 water
molecules were built. In both subunits, some residues
were modelled with two conformations with 50%
occupancy each (Lys A48, Lys A62, Glu A81, Lys
A84, Arg A109, Lys A183, Lys A220, Asn A240, Glu
A300, Arg A307, Arg A308, Arg A335, Glu A348, Glu

A
B
C
Fig. 2. (A) Ribbon representation of the Thermotoga maritima IDH
(TmIDH) dimer. The dimer forms through the association of the
small domains (subunit A: orange and subunit B: light blue) and the
formation of the clasp domain (A: pink and B: purple). Each large
domain (A: green and B: blue) is connected to the small domain via
a flexible hinge region. Both subunits were found in an open con-
formation. (B) Overlay of the large domains of TmIDH (green),
HcIDH (blue) and PcIDH (pink), showing the N- and C-termini and
the loop deletion of five residues between helices l and m in
TmIDH. The domains were superimposed separately because of
different conformations in the subunit. (C) Overlay of the small
domains including the clasp domains of TmIDH (green), Hc IDH
(blue) and PcIDH (pink) showing the loop deletion of three and four
residues in the clasp domain of TmIDH.
M. Karlstro
¨
m et al. Structure and thermal stability of T. maritima
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2853
B81, Glu B170, Asn B240, Glu B257, Arg B308 and
Glu B385). Four cis-peptides were identified, two of
which are cis-prolines (Pro376 in both subunits). The
rmsd values from ideal geometry were within reason-
able limits and the Ramachandran plot showed that
90.8 and 89.6% of the residues in subunits A and B,
respectively, fall within the most favourable region.
Asn159 in subunit B was found in a disallowed confor-
mation, which may be explained by unclear density

and its location in a sharp turn between strand M2
and M3. Table 1 summarizes the quality of the data
and the model.
Several crystal structures of different IDHs have
been reported: E. coli IDH [19] [EcIDH, PDB codes
3ICD (closed) and 1SJS (open)], Aeropyrum pernix
IDH [20] (ApIDH, PDB codes 1XGV, 1TYO and
1XKD) and Bacillus subtilis IDH [21] (BsIDH, PDB
code 1HQS) representing subfamily I and PcIDH [22]
(PcIDH, PDB code 1LWD) and HcIDH [23] [HcIDH,
PDB code 1T0L (closed) and 1T09 (open)] from sub-
family II. They are all homodimeric NADP
+
-IDHs
and their common folds are shared also by the crystal
structures of isopropylmalate dehydrogenase which
uses a substrate containing the malate moiety in com-
mon with isocitrate and belongs to the same family of
b-decarboxylating dehydrogenases [24–26]. TmIDH
showed highest structural similarity to PcIDH and
HcIDH, the two other members of subfamily II IDHs
for which the structure are known.
Hyperthermostable enzymes are often smaller than
their homologues from mesophiles. The amino acid
sequence of TmIDH is shorter than PcIDH and HcIDH.
TmIDH has 399 amino acid residues, whereas PcIDH
and HcIDH contain 413 and 414 residues, respectively.
However, all of the sequenced bacterial IDHs in this
subfamily (i.e. also those from mesophiles and psycro-
philes) have shorter sequences than the eukaryotic

IDHs. Therefore, the shorter sequence of TmIDH is
most likely a phylogenetic characteristic and not an
adaptation to higher temperature. All IDH subunits
consist of three domains: a large domain, a small
domain and a clasp domain (Fig. 2A). In TmIDH, resi-
dues 1–119 and 281–399 belong to the large domain, res-
idues 120–140 and 182–280 constitute the small domain
and residues 141–181 form the clasp domain. The large
domain is connected to the small domain by a flexible
hinge region. TmIDH was found in an open conforma-
tion with relative differences in the rotation of the large
domain of  30° compared with the closed ternary iso-
citrate–NADP
+
–Ca
2+
–HcIDH complex (PDB code:
1T0L) and of  25° compared with the binary iso-
citrate–PcIDH complex. Compared with the most open
subunit of the binary NADP
+
–HcIDH complex (PDB
code: 1T09), TmIDH was  6° more open. The clasp
domain in TmIDH was typical for subfamily II, with
two stacked four-stranded antiparallel b sheets instead
of the two antiparallel a helices beneath a single four-
stranded antiparallel b sheet characteristic for sub-
family I IDHs.
Because of different conformations, the large and
the small domains of each homologue were compared

separately. The rmsd values between the C
a
-carbons of
the large domain of TmIDH and that of PcIDH and
HcIDH were 0.92 A
˚
(using 231 C
a
-atoms) and 0.97 A
˚
(230 C
a
-atoms), respectively, when the nonconserved
loops in PcIDH and HcIDH were excluded. Superposi-
tion of the small domains, together with the clasp
domain, exhibited rmsd values between TmIDH and
PcIDH and HcIDH of 0.84 A
˚
(161 C
a
-atoms) and
0.88 A
˚
(159 C
a
-atoms), respectively. Overlays of the
different domains of TmIDH, HcIDH and PcIDH are
shown in Fig. 2B,C. Because of a transformation of
the secondary structure in the open form of HcIDH,
Table 1. Data collection and refinement statistics. Values in paren-

theses refer to data in the highest-resolution shell.
TmIDH
PDB code 1ZOR
Wavelength (A
˚
) 0.996
Resolution limits (A
˚
) 2.24–35.56
(2.24–2.29)
Mosaicity 0.5
Unit cell parameters a ¼ 62,5
b ¼ 88,1
c ¼ 180,a
a ¼ b ¼ c ¼ 90°
No. of unique reflections 49143(2500)
Redundancy 4.8
Completeness (%) 99.7 (100)
I ⁄ sigma (I) 15.84(3.32)
R
merge
(%) 6.6 (37.9)
Wilson B-factor 40.94
Space group P2
1
2
1
2
1
Refinement

No of non hydrogen atoms 6536
No of ions 2
Missing residues B1-3
Solvent molecules 428
Resolution range 2.24–35.56
(2.236–2.294)
R
cryst
(overall)% 18.4 (22.3)
R
free
(%) 22.3 (25.5)
Ramachandran plot (excl Gly and Pro)
Most favourable region subunit A ⁄ B (%) 90.8 ⁄ 89.6
Allowed regions A ⁄ B (%) 9.2 ⁄ 10.1
Disallowed regions A ⁄ B(%) 0⁄ 0.3
rmsd from ideal values
Bond lengths (A
˚
) 0.012
Bond angles (°) 1.251
Structure and thermal stability of T. maritima M. Karlstro
¨
m et al.
2854 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet
which is discussed more in detail below, the separate
domains of the closed form of HcIDH were used in
the superposition.
All secondary structural elements were conserved in
TmIDH, PcIDH and HcIDH. In total there were 16

a helices (44.9% of the residues in TmIDH) and 16
b strands (19.5% of the residues in TmIDH). In addi-
tion, one 3
10
helix was identified in TmIDH. The total
secondary structure content of 65.2% was only slightly
different from that of PcIDH (64.4%) and HcIDH
(66.2%). A structure-based sequence alignment is
shown in Fig. 3.
TmIDH has a loop deletion of three and four resi-
dues, compared with PcIDH and HcIDH, respectively,
between strands M2 and M3 in the clasp domain
(Fig. 2C). Other loop deletions were found between
helices g1 and g2 (one residue), strands E and D (one
residue) and helices l and m (five and four residues,
respectively, Fig. 2B). The two latter loop deletions
were also identified in several bacterial IDHs of meso-
philes from this subfamily and must be considered as
phylogenetic characteristics. The two first-loop dele-
tions, however, were not found in any mesophilic IDH.
This finding is in agreement with earlier observations
that proteins of hyperthermohiles often have shorter
loops [15,27]. In this way, fraying elements where the
structure might begin to unfold, are reduced.
Differences between the subunits
Subunits A and B of TmIDH were very similar. Super-
position of the small domains showed an rmsd of
0.27 A
˚
for the C

a
-atoms, whereas the rmsd between
the large domains was 0.32 A
˚
. The relative difference
in the rotation of the large domain between the two
subunits was 2.6°.
Dimer association
The interface in TmIDH was found to be similar to
that of PcIDH and HcIDH. The dimer associates
through the formation of the clasp domain and via
hydrophobic interactions between helices h and i in
both subunits to form a stable four-helix bundle.
The active site
The active site of the IDHs is formed in the cleft
between the small and the large domains and contains
residues from both domains and both subunits. The
substrate-binding residues of TmIDH were investigated
by superposition of the active site residues from each
domain of TmIDH and the closed substrate-bound
HcIDH and PcIDH separately. The putative isocitrate-
binding residues from the large domain, Thr77, Ser94
and Asn96, were very well aligned with rmsd values of
0.362 and 0.257 A
˚
between TmIDH vs. PcIDH and
HcIDH, respectively, for all 21 atoms. Arg100 and
Arg109 are also putative isocitrate-binding residues
but showed slightly different conformations in
TmIDH, most likely due to the absence of isocitrate in

TmIDH. However, the isocitrate-binding residues from
the small domain, Arg132, Tyr139 and Lys208¢ (the
prime indicates the neighbouring subunit of the dimer)
were still well aligned with rmsd values of 0.525 and
0.950 A
˚
between TmIDH vs. PcIDH and HcIDH,
respectively, for all 32 atoms. The metal-coordinating
residues Asp247¢, Asp270 and Asp274 were also struc-
turally similar with rmsd values of 0.745 and 1.056 A
˚
for all 24 atoms between TmIDH vs. PcIDH and
HcIDH, respectively. (The equivalent of Asp247¢
shows quite a different conformation in HcIDH which
is likely due to the absence of isocitrate in TmIDH.)
In the metal (Mg
2+
or Mn
2+
)-binding site, a Na
+
ion was built as it was the only positively charged ion
present in the crystallization buffer and no remaining
|F
o
| ) |F
c
| electron density could be observed at 2.0
sigma after it was built. The Na
+

ion was coordinated
by six oxygen ligands in both subunits. Two waters
(waters 185 and 427 in subunit A and waters 284 and
289 in subunit B), the carbonyl oxygen of Asp270 and
Asp247¢ represented the equatorial ligands, whereas
Asp270 and Asp274 were the axial ligands.
Cofactor binding
The b-decarboxylating dehydrogenases share a unique
cofactor-binding site that differs from the well-known
Rossmann fold found in many other dehydrogenases
[19,24,28]. Only a few amino acid residues appear to
be responsible for the discrimination between NAD
+
and NADP
+
[29–32]. The residues which interact with
the 2¢-phosphate of NADP
+
are responsible for the
discrimination between NAD
+
⁄ NADP
+
.
All of the residues involved in binding of the
NADP
+
in the ternary HcIDH complex (PDB code
1T0L) were found to be conserved in TmIDH except a
lysine (Lys260 in HcIDH) which interacts with the

2¢-phosphate of NADP
+
. This lysine was replaced by
Arg255¢ in TmIDH and might still be important for
the discrimination between NAD
+
and NADP
+
. Pre-
sumably, the other residues interacting with the 2¢-
phosphate in TmIDH are Arg308, His309 and
Gln252¢. However, close to Arg308 and His309 there
was another arginine (Arg312) in TmIDH directed
towards the supposed 2¢-phosphate of NADP
+
.In
M. Karlstro
¨
m et al. Structure and thermal stability of T. maritima
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2855
Structure and thermal stability of T. maritima M. Karlstro
¨
m et al.
2856 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet
PcIDH and HcIDH, this arginine is replaced by a Glu
(318) and Met (318), respectively. It is therefore poss-
ible that the cofactor binding is slightly different in
TmIDH compared with PcIDH and HcIDH. However,
all other residues putatively involved in binding of
NADP

+
were conserved structurally. The adenine
ring-binding residues correspond to His303, Val306,
Thr321 and Asn322, whereas Gly304, Thr305 and
Val306 should interact with the 5¢-phosphate of the
adenine. Thr77 and Arg82 are equivalent to the resi-
dues which bind the hydroxyl groups of the nicotina-
mide ribose. Thr75, Lys72 and Asn96 most likely bind
the amide group of the nicotinamide. The rmsd
between TmIDH and the NADP
+
–isocitrate–HcIDH–
complex (PDB code: 1T0L) was only 0.359 A
˚
for all
92 atoms of these residues which do not bind the
2¢-phosphate. Between TmIDH and PcIDH without
NADP
+
, the rmsd was 0.615 A
˚
for all 92 atoms.
The structural conservation of the isocitrate- and
NADP
+
-binding residues suggests that isocitrate and
NADP
+
binding is highly conserved between TmIDH
and the mammalian IDHs and that these enzymes

most likely share the same catalytic mechanism. For a
description of the mechanism, see Karlstro
¨
m et al. [20]
and Hurley et al. [16].
Putative phosphorylation site
In EcIDH, phosporylation of Ser113 in the so-called
‘phosphorylation loop’ by IDH kinase ⁄ phosphatase is
proposed to inactivate the enzyme by blocking the
binding of isocitrate to the active site both sterically
and by electrostatic repulsion of the c–carboxyl group
of isocitrate [33–36]. Whether this regulatory mechan-
ism exists in other IDHs is not clear. In the closed
ternary complex of HcIDH (PDB code: 1T0L), the
conserved helix i in the small domain is similar to that
observed in all other NADP
+
–IDH structures and is
part of the four-helix bundle at the dimer interface.
However, in the absence of isocitrate, the enzyme has
adopted an open conformation (PDB code: 1T09) and
helix i is found unwound into a loop conformation
where Asp279 interacts with Ser94 of the large domain
in the active site. This interaction mimics the phos-
phorylation of the equivalent serine in EcIDH which
inhibits the binding of isocitrate and makes the enzyme
inactive. It has been postulated that the new interac-
tion in HcIDH is competing with isocitrate in binding
to the active site and a self-regulatory mechanism of
activity is thereby provided [23]. In the open form of

TmIDH without isocitrate that we report here, this
helix is maintained and the self-regulating mechanism
is therefore not supported. The serine in the ‘phos-
phorylation loop’ is, however, conserved in TmIDH
and PcIDH (Ser94 and Ser95, respectively).
Thermostability
By sequence comparison, PcIDH and HcIDH are 51.3
and 52.2% identical to TmIDH, respectively. Neverthe-
less, the structural homology makes them suitable for a
comparison in order to identify differences which can be
related to thermostability. The apparent melting tem-
perature (T
m
)ofPcIDH was determined to 59.0 °C, i.e.
39.3 °C lower than the apparent T
m
of TmIDH. Below,
we try to relate the large T
m
difference to structural
determinants that presumably cause this highly
increased thermotolerance of TmIDH. The revealed dif-
ferences between TmIDH and its mesophilic homo-
logues in subfamily II are thereafter related to the
differences observed between ApIDH and its mesophilic
homologue EcIDH in subfamily I. The latter compar-
ison was made between the open Ap IDH (PDB code
1TYO) and the open EcIDH (PDB code 1SJS) and
might be slightly different from comparison between
ApIDH and the closed EcIDH done previously [20]. The

sequence identity between TmIDH and ApIDH is only
22.3% (a sequence alignment is shown in Fig. 3).
The major driving force in protein stability and fold-
ing is considered to be ‘the hydrophobic effect’ which
results in the burial of most of the hydrophobic resi-
dues in the protein core [37,38]. The hydrophobic
effect explains why many hyperthermostable proteins
show a significant increase in the number of buried
hydrophobic residues at the core or at subunit interfa-
ces, and is sometimes reflected by more hydrophobic
residues in the sequence [39–42]. This trend could also
be observed in TmIDH which has 42.9% hydrophobic
residues, whereas PcIDH and HcIDH have 39.5 and
39.4%, respectively (Table 2). However, many differ-
Fig. 3. Structure-based sequence alignment of TmIDH with porcine mitochondrial IDH (PcIDH, PDB code 1LWD), human cytosolic IDH
(HcIDH, PDB code 1T0L), Aeropyrum pernix IDH (ApIDH, PDB code 1TYO) and Escherichia coli IDH (EcIDH, PDB code 1SJS). The residues
occurring within structurally equivalent regions are boxed. Helices and strands appear as cylinders and arrows. Conserved residues are
shown in green, positions showing conservation of polar or charged character are in bold, those showing conservation of hydrophobic char-
acter are in yellow and residues showing a conservation of small size have smaller font. Sequence numbering according to TmIDH is red.
Secondary structure elements were given the nomenclature as implemented in EcIDH [19].
M. Karlstro
¨
m et al. Structure and thermal stability of T. maritima
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2857
Table 2. Characteristics of TmIDH, HcIDH, PcIDH, ApIDH and EcIDH.
TmIDH open HcIDH open HcIDH closed PcIDH closed ApIDH open EcIDH open
PDB code 1ZOR 1T09 1T0L 1LWD 1TYO 1SJS
Apparent melting temperature (°C) 98.3 N. d. N. d. 59.0 109.9 52.6
No of amino acids per subunit 399 414 414 413 435 416
Hydrophobic residues

a
(%) 42.9 39.4 39.4 39.5 46.2 44.4
Polar residues
b
(%) 25.5 32.8 32.8 32.4 27.6 28.4
Charged residues
c
(%) 31.6 27.8 27.8 28.1 26.2 27.2
Resolution (A
˚
) 2.24 2.7 2.4 1.85 2.1 2.4
rmsd of C
a
versus TmIDH large ⁄ small & clasp domain (A
˚
) – 0.85 ⁄ nd 0.97 ⁄ 0.88 0.92 ⁄ 0.84 2.27 ⁄ 1.57 nd
No of hydrogen bonds in subunit A 391 364 382 400 361 339
No of hydrogen bonds per residue in subunit A 0.98 0.88 0.92 0.97 0.84 0.82
No of hydrogen bonds (SS)
d
per residue in subunit A 0.12 0.08 0.09 0.12 0.08 0.04
No. of hydrogen bonds (SM)
e
per residue in subunit A 0.16 0.14 0.17 0.19 0.15 0.15
No. of hydrogen bonds (MM)
f
per residue in subunit A 0.70 0.66 0.66 0.66 0.61 0.62
No. of intersubunit hydrogen bonds 30 31 40 26 27 22
No. of intrasubunit ion pairs in subunit A at a 4 ⁄ 6 ⁄ 8A
˚

cut-off 31 ⁄ 58 ⁄ 96 24 ⁄ 49 ⁄ 78 29 ⁄ 56 ⁄ 85 28 ⁄ 55 ⁄ 77 29 ⁄ 57 ⁄ 85 27 ⁄ 50 ⁄ 77
No. of intrasubunit ion pairs per residue at a 4.0 A
˚
cut-off 0.078 0.058 0.070 0.075 0.067 0.065
Total no. of ion pairs per monomer at a 4 ⁄ 6 ⁄ 8A
˚
cut-off 35 ⁄ 63.5 ⁄ 24.5 ⁄ 53 32 ⁄ 61 30 ⁄ 58 31 ⁄ 63 29 ⁄ 55
Total no. of ion pairs per residue at a 4 A
˚
cut-off 0.088 0.059 0.077 0.073 0.071 0.070
Net charge (dimer) + 6 + 5 + 5 + 20 + 1 ) 19
No. of 3 ⁄ 4 ⁄ 5-member intrasubunit networks
in the monomer at a 4.0 A
˚
cut-off
1 ⁄ 1 ⁄ 23⁄ 1 ⁄ 04⁄ 1 ⁄ 04⁄ 1 ⁄ 03⁄ 1 ⁄ 06⁄ 0 ⁄ 0
No. of 3 ⁄ 4 ⁄ 5 ⁄ 6 ⁄ 7 ⁄ 8 ⁄ 9 ⁄ 10-member intrasubunit networks
in monomer A at a 6.0 A
˚
cut-off
2 ⁄ 2 ⁄ 2 ⁄ 0 ⁄ 0 ⁄ 1 ⁄ 0 ⁄ 15⁄ 2 ⁄ 0 ⁄ 1 ⁄ 1 ⁄ 0 ⁄ 1 ⁄ 04⁄ 1 ⁄ 1 ⁄ 0 ⁄ 0 ⁄ 0 ⁄ 2 ⁄ 05⁄ 1 ⁄ 2 ⁄ 1 ⁄ 0 ⁄ 1 ⁄ 0 ⁄ 04⁄ 2 ⁄ 0 ⁄ 0 ⁄ 2 ⁄ 0 ⁄ 0 ⁄ 08⁄ 1 ⁄ 2 ⁄ 0 ⁄ 1 ⁄ 0 ⁄ 0 ⁄ 0
No. of intersubunit ion-pairs at a 4 ⁄ 6A
˚
cut-off 8 ⁄ 11 1 ⁄ 86⁄ 10 4 ⁄ 64⁄ 12 4 ⁄ 10
Intersubunit 3 ⁄ 4 ⁄ 6 ⁄ 8-member networks at a 4.0 A
˚
cut-off 2 ⁄ 2 ⁄ 0 ⁄ 00⁄ 0 ⁄ 0 ⁄ 00⁄ 2 ⁄ 0 ⁄ 02⁄ 2 ⁄ 0 ⁄ 02⁄ 1 ⁄ 0 ⁄ 02⁄ 2 ⁄ 0 ⁄ 0
Intersubunit 3 ⁄ 4 ⁄ 5 ⁄ 6 ⁄ 7-member networks at a 6.0 A
˚
cut-off 1 ⁄ 2 ⁄ 0 ⁄ 0 ⁄ 12⁄ 0 ⁄ 0 ⁄ 0 ⁄ 11⁄ 2 ⁄ 0 ⁄ 1 ⁄ 02⁄ 0 ⁄ 0 ⁄ 2 ⁄ 01⁄ 0 ⁄ 0 ⁄ 2 ⁄ 02⁄ 2 ⁄ 0 ⁄ 0 ⁄ 4

Intersubunit 8 ⁄ 9 ⁄ 10 ⁄ 15 ⁄ 23-member networks at a 6.0 A
˚
cut-off 0 ⁄ 1 ⁄ 0 ⁄ 0 ⁄ 00⁄ 0 ⁄ 0 ⁄ 0 ⁄ 01⁄ 0 ⁄ 2 ⁄ 0 ⁄ 00⁄ 0 ⁄ 2 ⁄ 0 ⁄ 00⁄ 0 ⁄ 0 ⁄ 1 ⁄ 10⁄ 0 ⁄ 0 ⁄ 0 ⁄ 0
Volume (· 10
4
A
˚
3
) 15.5 15.9 15.7 15.4 15.8 15.2
Accessible surface area of dimer (A
˚
2
) 32780 35564 32564 31820 32916 32527
Surface ⁄ volume ratio 0.211 0.224 0.207 0.207 0.208 0.214
Buried surface at dimer interface (% of dimer) 16.7 13.7 18.5 17.6 14.9 15.6
Distribution of hydrophobic ⁄ polar ⁄ charged area at accessible
surface of dimer (% of dimer)
54.4 ⁄ 19.1 ⁄ 26.5 55.7 ⁄ 23.9 ⁄ 20.4 55.4 ⁄ 24.6 ⁄ 20.0 54.6 ⁄ 24.2 ⁄ 21.1 53.2 ⁄ 23.5 ⁄ 23.3 53.7 ⁄ 21.6 ⁄ 24.7
Distribution of hydrophobic ⁄ polar ⁄ charged area
at dimer interface (% of interface)
69.5 ⁄ 18.4 ⁄ 12.1 67.1 ⁄ 22.8 ⁄ 10.1 64.5 ⁄ 22.2 ⁄ 13.3 73.1 ⁄ 18.7 ⁄ 8.2 72.4 ⁄ 16.1 ⁄ 11.5 69.2 ⁄ 18.9 ⁄ 11.9
a
Hydrophobic residues: A,V,L,I,W,F,P,M.
b
Polar residues: G,S,T,Y,N,Q,C.
c
Charged residues: R,K,H,D,E.
d
SS, side chain–side chain hydrogen bonds.
e

SM, side chain–main chain hydro-
gen bonds.
f
MM main chain–main chain hydrogen bonds.
Structure and thermal stability of T. maritima M. Karlstro
¨
m et al.
2858 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet
ences between proteins from hyperthermophiles and
mesophiles are observed on the surface of the protein
subunits. The contribution of hydrophobic residues at
the protein surface will affect the stability of the pro-
tein unfavorably in aqueous solution, whereas hydro-
philic residues will help to solvate the protein and
thereby stabilize it. Accordingly, the accessible surface
area of hyperthermostable proteins is commonly more
polar or charged and less hydrophobic [8,9,11,43–45].
Surface and interface characteristics
First, the open form of TmIDH reported here, was
compared with the open form of HcIDH. However,
one must bear in mind that two of the helices (i and i¢)
in the interfacial four-helix bundle are unfolded in the
open HcIDH, giving the surface and the subunit inter-
face a different character. Therefore, we show data for
the open HcIDH, the closed HcIDH and the closed
PcIDH in Table 2. It was found that TmIDH has a
6.1% increase in charged accessible surface area, a
1.3% decrease in hydrophobic surface area and 4.8%
decrease in polar surface area compared with the open
HcIDH (Fig. 4A and Table 2). This is in line with ear-

lier comparisons between enzymes from hyperthermo-
philes and mesophiles [3,42]. In this context, Ap IDH is
unusual having a less charged and more polar access-
ible surface than its mesophilic homologue EcIDH.
However, the hydrophobic part is still reduced by
0.5% compared with the open EcIDH (Table 2). The
small decrease in the hydrophobic part of the surfaces
might seem insignificant. However, using 25 calÆ
mol
)1
ÆA
˚
)2
for the hydrophobic contribution to the free
energy of folding applied on the difference in accessible
hydrophobic area in TmIDH (which was 2005 A
˚
2
),
gives a free energy difference of  50 kcalÆmol
)1
indica-
ting a contribution to stability in the expected order of
magnitude [46,47].
Moreover, a reduction of the total solvent-exposed
surface area has also been associated with increased
stability [48,49]. The solvent-exposed surface area of
TmIDH was 32 780 A
˚
2

which is considerably smaller
than the 35 564 A
˚
2
of the open form of HcIDH. How-
ever, this is partially due to the shorter sequence of
TmIDH. Therefore, the surface-to-volume ratio was
determined [50]. For TmIDH it was 0.211, whereas for
the open HcIDH it was 0.224, indicating a slight
decrease in the accessible surface. The same trend was
observed between ApIDH (0.208) and EcIDH (0.214).
In principle, a decrease in the accessible surface
might be due to increased burial of the molecular sur-
face at the subunit interface. However, both TmIDH
and ApIDH were found to have a smaller relative
interface area compared with their respective homo-
logues. Because of the unwinding of helix i belonging
to the interfacial four-helix bundle in the open HcIDH,
we preferred to compare the interface area of TmIDH
with that of the closed HcIDH and PcIDH. It was
found that 16.7% of the total molecular surface of the
two TmIDH subunits was buried at the interface,
whereas 18.5 and 17.6% of the closed HcIDH and
PcIDH surfaces were buried, respectively. The inter-
face of ApIDH was proportionately smaller than both
the open and the closed EcIDH (Table 2) [20].
The interface of TmIDH showed a 5% increase in
hydrophobic area and a 1.2% decrease in charged area
compared with the closed HcIDH (Fig. 4B, Table 2).
A

B
Fig. 4. (A) Distribution of hydrophobic, polar and charged accessible
surface area (ASA) of the different IDHs. (B) Distribution of hydropho-
bic, polar and charged area at the interface of the different IDHs.
M. Karlstro
¨
m et al. Structure and thermal stability of T. maritima
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2859
A similar trend was found also when the interface of
ApIDH was compared with that of the open homo-
logue EcIDH. However, compared with the closed
PcIDH, the interface of TmIDH was 3.5% less hydro-
phobic and 3.9% more charged (Table 2). Part of the
hydrophobic contribution at the subunit interface in
TmIDH involves a cluster of four methionines located
in the interfacial four-helix bundle (Met272 and
Met275 from both subunits). This methionine cluster
was not present in PcIDH or HcIDH. However, below
the four-helix bundle, on the top of the clasp domain,
there was another, almost identical, intersubunit four-
membered methionine cluster (formed by Met176 and
Met178 from both subunits), which was found to be
conserved in these three IDHs. The additional methi-
onine cluster in TmIDH might be important for the
subunit interaction. It was also found that helix h of
the interfacial four-helix bundle might be stabilized in
TmIDH by three Ala residues in a row instead of one
in PcIDH and two separate Ala residues as in HcIDH.
Alanine is considered to be the most optimal helix-
forming amino acid [51,52].

Aromatic interactions
Aromatic interactions are usually identified using a cut-
off distance of 7 A
˚
between the aromatic ring centres
[53]. The role of additional aromatic interactions in
increasing the thermostability of proteins has been dis-
cussed elsewhere [53,54]. However, Tm IDH was found
to have a decreased fraction of aromatic residues
(10.8%) compared with HcIDH (12.1%) and PcIDH
(12.7%). However, a cluster of aromatic residues invol-
ving Phe205, Phe188, Phe219, Phe223, Tyr215, Tyr241
and His133 was identified in the small domain of
TmIDH. All of these residues were conserved in PcIDH
and HcIDH except Phe205, which has a central posi-
tion in this cluster (Fig. 5C). A few other nonconserved
A
B
C
Fig. 5. (A) The mutation D389N decreased the apparent melting
temperature (T
m
)ofTmIDH by 21.8 °C. At a cut-off distance of
4A
˚
, Asp389 formed a conserved ion pair with Lys29, connecting
both termini with each other. At a 6 A
˚
cut-off, the ion pair was
involved in a nonconserved four-member ionic network with

Lys318 and Glu385. (B) Arg186 made interactions with Glu182,
Glu225 and Glu226 and was part of a five-member ionic network.
However, mutation R186M did not affect the apparent T
m
of
TmIDH. (C) Mutation F205M reduced the apparent T
m
by 3.5 °C.
Phe205 was found to have a central position in an aromatic cluster
in the small domain involving Phe188, Phe219, Phe223, Tyr215,
Tyr241 and His133. All but Phe205 were conserved within sub-
family II. The result confirms that aromatic clustering plays a role in
increasing the apparent melting temperature of proteins.
Structure and thermal stability of T. maritima M. Karlstro
¨
m et al.
2860 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet
aromatic residues were identified within 7 A
˚
of the resi-
dues of the mentioned cluster, e.g. Tyr218, Tyr243,
Tyr135, His174 and His216 but they were all pointing
away from the nearby aromatic residues and their
involvement in the cluster was unclear.
Charged residues and ionic interactions
An ion pair is usually defined as an interaction
between charged groups with a distance of up to 4.0 A
˚
between the charged nonhydrogen atoms [55]. An
increased number of ion pairs and larger ionic net-

works are the most common features of hyperthermo-
stable proteins observed in comparisons with
mesophilic proteins [8,56]. Furthermore, according to
model calculations and theoretical considerations, the
stabilizing contribution of ion pairs increases at higher
temperatures, because the desolvation penalty is
decreased with higher temperatures, mainly due to a
decrease in the dielectric constant of water [3,57]. In
addition, it has been claimed that the dielectric con-
stant of thermophilic proteins is increased, reducing
the dielectric difference between the protein and the
solvent, and thus the desolvation penalty, even more
[58]. That the contribution of each ion pair can vary
significantly depending on the local environment and
the solvation, means that the interaction might also be
considerable above the default 4.0 A
˚
cut-off [11,12].
Therefore, we analysed the number of ion pairs using
cut-offs of 4.0, 6.0 and 8.0 A
˚
and combined the analy-
sis with results from differential scanning calorimetry
and site-directed mutagenesis of certain ion pairs in
order to detect interactions contributing the most to
the thermotolerance of TmIDH. Comparison of ion
pairs and hydrogen bonds was based on subunit A in
all cases because many side chains lacked electron den-
sity in the large domain of subunit B in TmIDH. The
quality of the electron-density map of the interface-

containing small domain of subunit B was, however,
fine, which allowed safe analysis of the number of
intersubunit ion pairs.
The total number of charged residues per monomer
was higher in TmIDH: 126 (31.6% of the residues), than
in PcIDH and HcIDH: 116 (28.1%) and 115 (27.8%),
respectively (Table 2). The total number of ion pairs per
monomer (including half of the intersubunit ion pairs in
the dimer) was also slightly higher in TmIDH when a
cut-off distance of 4 A
˚
was used; 35 compared with 30
and 32 in PcIDH and the closed form of HcIDH,
respectively. However, because Tm IDH was in the open
form, it was more appropriate to compare it with the
open form of HcIDH. The total number of ion pairs in
the open monomer of HcIDH was only 24.5. The differ-
ence was clearer when the fractions defined by ion pairs
per residue were compared; 0.088 in TmIDH compared
with 0.073, 0.077 and 0.059 in PcIDH, HcIDH (closed)
and HcIDH (open), respectively (Table 2). This finding
is in agreement with other hyperthermostable proteins.
However, it has previously been found that only at a
cut-off of at least 4.2 A
˚
a significant increase in ion pairs
in ApIDH compared with the closed EcIDH is observed
[20]. This is also true in the present comparison of
ApIDH with the open EcIDH where the fraction is
0.071 and 0.070 using a cut-off of 4.0 A

˚
, but 0.087 and
0.079 at a 4.2 A
˚
cut-off for ApIDH and EcIDH, respect-
ively.
The total number of intersubunit ion-pairs per dimer
at a cut-off of 4 A
˚
was eight in TmIDH compared
with four, six and one in PcIDH, HcIDH (closed) and
HcIDH (open), respectively. Two of the intersubunit
ion pairs were found in each active site, formed by
Lys208 and Asp270 in TmIDH belonging to the four-
helix bundle. These residues were also part of a con-
served four-member ionic network (see ionic network
section below) and were involved in the binding of
Mg
2+
-isocitrate. Another four of the intersubunit ion
pairs were found in the clasp domain where two non-
conserved 3-member ionic networks were located in
both TmIDH and PcIDH. In TmIDH, each network
contained two intersubunit ion pairs (between Glu151
and Glu153 of one subunit and Arg157 of the other
subunit) and was extended to two 7-member intersub-
unit ionic networks when a cut-off of 6 A
˚
was used.
This extension was not observed in PcIDH at 6 A

˚
and
the 3-member networks only contained one intersub-
unit ion pair each. The final intersubunit ion pairs in
TmIDH at a cut-off of 4 A
˚
were formed on each sub-
unit by Lys122 located on a loop on the small domain
and Glu375 which was found on a loop on the large
domain of the other subunit. The closed HcIDH con-
tained in total six intersubunit ion pairs. All of these
intersubunit ion pairs were gone in the open HcIDH
because of the domain rotation and unfolding of helix
i and i¢ belonging to the four-helix bundle. However, a
new intersubunit ion pair was formed by Lys212 and
Asp273. This increase in intersubunit ion pairs has
been commonly observed in hyperthermostable pro-
teins [52]. In ApIDH however, a cut-off of 6.0 A
˚
was
needed to reveal a higher number of intersubunit ion
pairs compared with EcIDH (Table 2).
Ionic networks
TmIDH was found to have larger ionic networks than
PcIDH and HcIDH. The organization of ion pairs into
large networks is a characteristic feature of hyperther-
M. Karlstro
¨
m et al. Structure and thermal stability of T. maritima
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2861

mophilic proteins [44]. Because of cooperativity, ionic
networks contribute more to the stabilization energy of
the protein than do the sum of the pairwise interac-
tions [59,60]. In TmIDH, we identified one 3-member,
one 4-member and two 5-member intrasubunit net-
works in subunit A using a cut-off of 4 A
˚
, whereas in
PcIDH and the closed HcIDH, we found four 3-mem-
ber and one 4-member intrasubunit networks in the A
subunit. In addition, a conserved 4-member intersub-
unit network was identified in each active site of all
three IDHs and two 3-member intersubunit networks
were found in the clasp domain of TmIDH and
PcIDH. The number of intrasubunit networks in the
open HcIDH was only slightly different compared with
the closed form, whereas the intersubunit networks
were absent at a 4.0 A
˚
cut-off (Table 2). Except for the
large difference in intersubunit ion pairs and intersub-
unit ionic networks between TmIDH and the open
form of HcIDH, the remaining difference between
TmIDH, PcIDH and both HcIDHs was the 5-member
networks found in TmIDH. One was located in the
small domain and involved Glu182 and Arg186 on
helix f and Glu225, Glu226 and Lys229 on helix g. The
other was found in the large domain and was com-
posed of of Glu17 (helix a), Arg82, Glu81 and Glu85
(helix c2) and Arg307 (helix j1). The largest networks

in ApIDH and EcIDH are 4-membered at a 4.0 A
˚
cut-
off. However, at a 4.2 A
˚
cut-off, three seven-member
ionic networks were revealed in the ApIDH dimer,
whereas only two 5-member networks appeared in the
open EcIDH. These observations support earlier find-
ings that hyperthermophiles often have more or larger
ionic networks. When a cut-off distance of 6.0 A
˚
was
used, the size of the ionic networks was increased only
slightly in TmIDH compared with HcIDH and PcIDH
(Table 2). However, using a 8 A
˚
cut-off resulted in one
44-member, one 43-member and two 26-member net-
works in TmIDH involving both subunits, whereas
only one 26-member, one 23-member and one 20-mem-
ber network was found in the open HcIDH dimer and
one 43-member and one 39-member network in the
closed PcIDH dimer. These results suggest a higher
degree of charge compensation in TmIDH and that
cooperation of weak ionic interactions in networks
might also be important for thermostabilization.
Hydrogen bonds
A small increase in the fraction of hydrogen bonds per
residue was observed in TmIDH (0.98) compared with

PcIDH (0.97), the closed HcIDH; (0.92) and the open
HcIDH (0.88). However, the determination of hydro-
gen bonds is highly resolution dependent.
Site-directed mutagenesis
The mutant D389N was found to have an apparent
T
m
of 76.5 °C, i.e. a considerable decrease of almost
22 °C less than the recombinant wild-type form of
TmIDH. Asp389 was located on helix m at the C-ter-
minus and made an ionic interaction with Lys29 at a
4A
˚
cut-off (Fig. 5A). This ion pair is conserved in
PcIDH and HcIDH and is presumably involved in the
protection of both termini from thermal unfolding
which partially explains the large impact of this muta-
tion. At a cut-off of 6.0 A
˚
, the ion pair was extended
to a four-member ionic network in TmIDH through
additional interactions with Glu385 and Lys318. These
interactions could not be observed in PcIDH and
HcIDH, indicating that the stabilizing effect of ion
pairs is dependent on their local environment. There
were also several other charged residues close by
Asp389 which apparently makes it unusually sensitive
to mutation. Another mutant of Tm IDH which was
investigated is D36N ⁄ D341N. It was found to have an
apparent T

m
of 88.5 °C, i.e.  10 °C less than the
recombinant wild-type protein. The mutations of these
two aspartates most likely interfere with the neutraliza-
tion of the positive charges provided by Lys3, 5 and 7
at the very end of the N-terminus. The distances
between the aspartates and the lysines were between
4.5 and 7.2 A
˚
. This result suggests that electrostatic
compensation is crucial and confirms the importance
of stabilizing the N-terminus to prevent thermal
unfolding. TmIDH has the stabilization of its N-ter-
mini in common with ApIDH which has a disulfide
bond at the N-termini between Cys9 and Cys87. The
mutation C87S in ApIDH decreases the apparent melt-
ing temp by 9.6 °C [20]. These two mechanisms to sta-
bilize the N-terminus complement each other. Often,
proteins that are stabilized by disulfide bonds show a
lower degree of electrostatic optimization [61].
One of the five-member ionic networks in TmIDH
involved helices f and g located on the small domain.
Arg186 has a central position in this network and
interacts with three glutamates (182, 225 and 226)
(Fig. 5B). However, the mutation R186M did not
affect the apparent T
m
of TmIDH. In contrast, one of
the seven-member ionic networks in ApIDH was found
to play an important role for the increased thermal

stability [20]. Mutation R211M in ApIDH, which dis-
rupts this network, decreases the apparent T
m
by
11.3 °C. However, this network forms interdomain
interactions and is located in the cleft at the opposite
side of the active site, whereas in Tm IDH, the five-
member networks were separately located on the dif-
ferent domains. The position of the mutated network
Structure and thermal stability of T. maritima M. Karlstro
¨
m et al.
2862 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet
on the small domain, far away from the terminals,
might be part of the reason why this mutation did not
have any effect on the melting temperature.
However, mutation F205M, which disrupts the aro-
matic cluster in the small domain mentioned above
(Fig. 5C), reduced the apparent T
m
to 94.8 °C, i.e.
)3.5 °C compared with wild-type TmIDH. All residues
in the cluster except Phe205 are conserved in PcIDH
and HcIDH, which suggests that Phe205 plays a role
in the increased thermal stability of TmIDH.
Conclusions regarding thermostability
The main differences in TmIDH compared with its
mesophilic homologues were more ion pairs, larger
ionic networks, an increased number of intersubunit
ion pairs and stabilization of the N- and C-termini via

long-range ionic interactions. Furthermore, we found a
relative increase in hydrophobic residues and a more
hydrophobic interface, whereas the accessible surface
was less hydrophobic and more charged. The surface-
to-volume ratio was reduced slightly. With respect to
earlier studies on thermostability, these differences are
most likely important factors for enhanced thermosta-
bility. Site-directed mutagenesis and calorimetric meas-
urements confirmed that stabilization of the N- and
C-termini via electrostatic compensation and clustering
of aromatic residues are vital for the stability of
TmIDH. However, analysis of one mutant, R186M,
demonstrated that the location of the ionic network
and the local environment probably is essential,
because the mutation did not affect the apparent T
m
of TmIDH. Several properties of TmIDH were shared
in common with ApIDH, which is the most thermosta-
ble IDH: stabilized N-terminus, more ion pairs and
larger ionic networks, more hydrophobic residues in
total, a more hydrophobic interface, a less hydropho-
bic accessible surface area and a decreased surface-to-
volume ratio. However, TmIDH lacked interdomain
ionic networks at a cut-off of 4.0 A
˚
and the size of the
networks were not as dramatically increased at 4.2 and
6.0 A
˚
cut-offs like in ApIDH. By contrast, the number

of intersubunit ion pairs was higher and the accessible
surface was more charged in TmIDH.
Experimental procedures
Mutations and protein purification
Mutations in TmIDH were introduced by PCR-based
mutagenesis using the QuickChange Site-Directed Muta-
genesis Kit (Stratagene, La Jolla, CA). The primers used
are listed in Table 3 and the mutation sites are under-
lined. Incorporation of the mutation at the desired posi-
tion was confirmed by DNA sequence analysis of the
entire idh-coding region using an ABI 3700 DNA
sequencer. Mutated enzyme was produced in E. coli strain
BL21-Codon-Plus (DE3)-RIL (Stratagene) as described
previously for the recombinant wild-type enzyme and
purified using heat treatment and Red-Sepharose chroma-
tography [17]. The protein appeared as a single band on
a SDS-acrylamide gel, however, two bands appeared on a
native gel, corresponding to the dimeric and the tetra-
meric forms. The two oligomeric states were separated
using a Superdex 200 16 ⁄ 60 gel filtration column.
Analytical ultracentrifugation
AUC experiments were performed on the recombinant wt
TmIDH both before and after gel filtration (on the dimeric
and tetrameric fractions separately) using a Beckman
XLA analytical ultracentrifuge (Beckman Coulter, Roissy,
France) equipped with a UV scanning system, using a four-
hole AN-60 Ti rotor with double centrepieces of 1.20 cm
path length. Two hundred absorbance profiles were recor-
ded at 20 °C and 42 000 r.p.m., analysed using the sedfit
program [62]. The calculation of the corrected sedimenta-

tion coefficient at 20 °C in water (s
20,w
) was performed as
described previously [17].
Table 3. Primer sequences.
TmIDH mutant Primer sequence
R186M 5’-CCTGGAGAAATCGATCA
TGAGCTTCGCTCAGTCGTG-3’
5’-CACGACTGAGCGAAGCT
CATGATCGATTTCTCCAGG-3’
F205M 5’-AAAAGGTCGACATCTGG
ATGGCGACGAAAGACACGATC-3’
5’-GATCGTGTCTTTCGTCGC
CATCCAGATGTCGACCTTTT-3’
D36N + D341N D36N-1
5’-CATCCTTCCCTATCTC
AACATCCAGCTGGTTTACT-3’
D341N-1
5’-AAGGGGAGAACTC
AACGGAACACCGGAGG-3’
D389N 5’-CACTCTCGAAGAGTTCATA
AACGAAGTGAAGAAGAATCTC-3’
5’-GAGATTCTTCTTCACTTCGT
TTATGAACTCTTCGAGAGTG-3’
M. Karlstro
¨
m et al. Structure and thermal stability of T. maritima
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2863
Differential scanning microcalorimetric
measurements

Differential scanning calorimetry was carried out on the
dimeric fractions of purified recombinant wild-type TmIDH
and the mutants with a MicroCal MCS calorimeter con-
trolled using the mcs observer program (MicroCal, North-
ampton, MA). The samples were dialysed against the
reference buffer used in the experiment (50 mm potassium-
phosphate buffer, pH 7.5, 0.1 mm NaCl) and degassed for
20 min prior to the calorimetric analysis. A protein concen-
tration of 1.5 mgÆmL
)1
was used. The calorimetric scans were
carried out between 20° and 120 °C using a scan rate of 1
KÆmin
)1
. A constant pressure of 2 bar was applied in order
to avoid boiling at high temperatures. Each sample was
scanned a second time after the actual calorimetric scan to
estimate the reversibility of the unfolding transition.
Crystallization, data collection and data
processing
TmIDH was crystallized in 16% PEG 6000, 100 mm succi-
nate buffer pH 6.0, 200 mm sodium acetate and 50 mm
NaCl by using the vapour diffusion method and sitting
drops at 291 K. 2.5 lLof18mgÆmL
)1
protein was added
to an equal volume of the reservoir solution. Crystals
appeared after a few days and grew to a typical dimension
of 0.2 · 0.2 · 0.5 mm. Data were collected at the synchro-
tron beamline I711 at MAXLAB (Lund, Sweden). The

crystal was flash-frozen with boiling nitrogen at 100 K in
the presence of 20% ethylene glycol as cryoprotectant. A
MarCCD 165 detector and marccd v. 0.5.33 software were
used for data collection. The data were indexed, integrated
and scaled with the hkl package [63]. Simulated precession
pictures of the intensity data in space group P1, showed
Laue mmm symmetry and the systematic extinctions fol-
lowing h ¼ 2n, k ¼ 2n and l ¼ 2n pointed to space group
P2
1
2
1
2
1
. The dataset was completed by including all poss-
ible hkls and R
free
columns using unique [64]. Structure
factors were calculated with the program truncate [64].
Data collection parameters and processing statistics are
given in Table 1. The Matthews’ coefficient (V
m
) was
2.7 A
˚
3
ÆDa
)1
suggesting a dimer in the asymmetric unit.
Molecular replacement

Phase information was obtained in space group P2
1
2
1
2
1
by
molecular replacement with the program phaser [65] using
data up to 2.5 A
˚
. The different domains of IDH were
searched for separately because of different domain orienta-
tions found in other IDH crystal structures. The rotation
and the position of the dimer of the small domains, inclu-
ding the clasp domain, were located using the equivalent
domains of the porcine mitochondrial IDH structure (PDB
entry 1LWD) as search model. The large domains were
located using both the porcine mitochondrial IDH (PDB
entry 1LWD) and the human cytosolic IDH (PDB entry
1T0L) as multiple models. Maps were initially calculated
with phases obtained from a polyalanine model of porcine
IDH with the separate domains positioned on the MR solu-
tion. Only nonconserved amino acids were changed to alan-
ine. The sequence identity between TmIDH,PcIDH and
HcIDH is 51.3 and 52.2%, respectively.
Crystallographic refinement, density modification
and model building
The initial R-factor of the polyalanine model was 49.9%
(R
free

49.5%). The orientations of the two subunits and the
domains were refined with rigid body refinement, decreasing
the R-factor to 49.1% (R
free
48.8%). Simulated annealing
in CNS [66] was applied using torsion angles. The starting
temperature was 3500°K and was decreased in steps of
25°K. The best results were obtained using tight NCS
restraints (300 kcalÆmol
)1
ÆA
˚
)2
) between the small domains
and between the clasp domains and loose restraints (50
kcalÆmol
)1
ÆA
˚
)2
) between the large domains. After grouped
B-factor refinement in CNS, using one B-factor for each
main-chain residue and one for each side chain, the R-factor
dropped to 37.6% (R
free
40.0%).
Model building was performed with the program o [67]
guided by weighted 2|F
o
| ) |F

c
| and |F
o
| ) |F
c
| maps. Each
model building session was followed by TLS- and
restrained refinement including isotropic B-factor refine-
ment of the individual atoms with the program refmac5
[64]. The rigid TLS groups for anisotropic thermal motion
were defined according to the domains of TmIDH. When
subunit A was built, it was rotated over subunit B and
adjusted according to the map. In some regions of subunit
B, where the density was unclear, density-modified and
averaged maps were used. Density modification and avera-
ging according to the twofold noncrystallographic sym-
metry was done with the program dm. NCS restraints
were optimized, taking into account possible differences
due to crystal contacts. Some residues were built with two
different conformations when negative |F
o
| ) |F
c
| density
was present for a certain conformation and positive
|F
o
| ) |F
c
| density for the other conformation and vice

versa.
Water molecules were added using the program arp_
warp [68] and kept only if density was present in both
2|F
o
| ) |F
c
| (if above 1 sigma) and |F
o
| ) |F
c
| maps (if
above 3.0 sigma), if a suitable hydrogen bond donor ⁄ accep-
tor was present and if the temperature factor was below
65 A
˚
2
. Some waters were added manually.
The stereochemical quality of the models was checked
with the program procheck [69] as implemented in ccp4
[64]. Superposition and determination of rmsd values
between subunits, domains and homologous structures were
performed with lsqman [70].
Structure and thermal stability of T. maritima M. Karlstro
¨
m et al.
2864 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet
Domains were considered as rigid bodies. Domain rotations
were defined as the rotation angle required to superimpose the
large domain in subunit A on the same domain in subunit B,

with the small domains of subunit A and B remaining in a
fixed superimposed position. Rotation angles were calculated
from the O rotation matrix by convrot (W. Meining, unpub-
lished). No significant translation component was detected.
A structure-based sequence alignment was performed
with the program stamp [71] and was based on the
C
a
-atom coordinates and secondary structural assignments
made using the program dssp [72].
Surface and volume calculations and analysis
of noncovalent interactions
Ion pairs and ionic networks were analysed using the pro-
grams contact [64] and ionstat (W. Meining, unpub-
lished) with a maximum distance of 4 [55], 6 or 8 A
˚
.
Accessible surface areas were calculated using CNS [66].
Ser-OG, Thr-OG1, Tyr-OH, Trp-NE1, Cys-SG, Asn-OD1,
Asn-ND2, Gln-OE1, Gln-NE2, Arg-NE atoms and all main
chain O and N atoms were defined as polar atoms. Lys-
NZ, Arg-NH1, Arg-NH2, His-ND1, His-NE2, Asp-OD1,
Asp-OD2, Glu-OE1, Glu-OE2, OXT were treated as
charged atoms. All other atoms were treated as hydropho-
bic. The water probe radius was 1.4 A
˚
. Built water mole-
cules were excluded from the model. The water-accessible
volumes of the structures were calculated with voidoo [73]
using a water probe radius of 1.4 A

˚
.
Hydrogen bonds were calculated using hbplus v. 3.15
[74] and the following default parameters: maximum distan-
ces for D–A, 3.9 A
˚
and for H–A, 2.5 A
˚
; minimum angles
for D–H–A, D–A–AA and H–A–AA were 90° [75].
Figure preparations
Figures 2 and 5 were made using pymol [76]. Figure 3 was
prepared using the program alscript [77].
Acknowledgements
We are grateful to Marit Steine Madsen, University of
Bergen, Norway, for expression and purification of
TmIDH. This work was supported by a PhD grant
from The So
¨
derto
¨
rn University College to MK and
by the Norwegian Research Council (Project no.
153774 ⁄ 420). Part of the work was financed by contri-
butions from NorFa to IHS.
References
1 Adams MW (1993) Enzymes and proteins from organ-
isms that grow near and above 100 degrees C. Annu Rev
Microbiol 47, 627–658.
2 Cowan DA (2004) The upper temperature for life – where

do we draw the line ? Trends Microbiol 12, 58–60.
3 Vieille C & Zeikus GJ (2001) Hyperthermophilic
enzymes: sources, uses, and molecular mechanisms for
thermostability. Microbiol Mol Biol Rev 65, 1–43.
4 Jaenicke R & Bo
¨
hm G (1998) The stability of proteins
in extreme environments. Curr Opin Struct Biol 8, 738–
748.
5 Dill KA (1990) Dominant forces in protein folding.
Biochemistry 29, 7133–7155.
6 Jaenicke R (1991) Protein stability and molecular adapta-
tion to extreme conditions. Eur J Biochem 202, 715–728.
7 Matthews BW, Weaver LH & Kester WR (1974) The
conformation of thermolysin. J Biol Chem 249, 8030–
8044.
8 Karshikoff A & Ladenstein R (2001) Ion pairs and the
thermotolerance of proteins from hyperthermophiles: a
‘traffic rule’ for hot roads. Trends Biochem Sci 26, 550–
556.
9 Yip KS, Stillman TJ, Britton KL, Artymiuk PJ, Baker
PJ, Sedelnikova SE, Engel PC, Pasquo A, Chiaraluce R
& Consalvi V (1995) The structure of Pyrococcus furio-
sus glutamate dehydrogenase reveals a key role for ion-
pair networks in maintaining enzyme stability at
extreme temperatures. Structure 3, 1147–1158.
10 Rees DC (2001) Crystallographic analyses of hyperther-
mophilic proteins. Methods Enzymol 334, 423–437.
11 Spassov VZ, Karshikoff AD & Ladenstein R (1995)
The optimization of protein–solvent interactions: ther-

mostability and the role of hydrophobic and electro-
static interactions. Protein Sci 4, 1516–1527.
12 Xiao L & Honig B (1999) Electrostatic contributions to
the stability of hyperthermophilic proteins. J Mol Biol
289, 1435–1444.
13 Vogt G & Argos P (1997) Protein thermal stability:
hydrogen bonds or internal packing? Fold Des 2, S40–
S46.
14 Delboni LF, Mande SC, Rentier-Delrue F, Mainfroid
V, Turley S, Vellieux FM, Martial JA & Hol WG
(1995) Crystal structure of recombinant triosephosphate
isomerase from Bacillus stearothermophilus. An analysis
of potential thermostability factors in six isomerases
with known three-dimensional structures points to the
importance of hydrophobic interactions. Protein Sci 4,
2594–2604.
15 Russell RJ, Ferguson JM, Hough DW, Danson MJ &
Taylor GL (1997) The crystal structure of citrate synthase
from the hyperthermophilic archaeon Pyrococcus furiosus
at 1.9 A
˚
resolution. Biochemistry 36, 9983–9994.
16 Hurley JH, Dean AM, Koshland DE Jr & Stroud RM
(1991) Catalytic mechanism of NADP
+
-dependent iso-
citrate dehydrogenase: implications from the structures
of magnesium-isocitrate and NADP
+
complexes.

Biochemistry 30, 8671–8678.
M. Karlstro
¨
m et al. Structure and thermal stability of T. maritima
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2865
17 Steen IH, Madern D, Karlstrom M, Lien T, Ladenstein
R & Birkeland NK (2001) Comparison of isocitrate
dehydrogenase from three hyperthermophiles reveals
differences in thermostability, cofactor specificity, oligo-
meric state, and phylogenetic affiliation. J Biol Chem
276, 43924–43931.
18 Miyazaki K (2005) Identification of a novel trifunctional
homoisocitrate dehydrogenase and modulation of the
broad substrate specificity through site-directed muta-
genesis. Biochem Biophys Res Commun 336, 596–602.
19 Hurley JH, Thorsness PE, Ramalingam V, Helmers
NH, Koshland DE Jr & Stroud RM (1989) Structure of
a bacterial enzyme regulated by phosphorylation, isoci-
trate dehydrogenase. Proc Natl Acad Sci USA 86, 8635–
8639.
20 Karlstrom M, Stokke R, Steen IH, Birkeland NK &
Ladenstein R (2005) Isocitrate dehydrogenase from the
hyperthermophile Aeropyrum pernix : X-ray structure
analysis of a ternary enzyme–substrate complex and
thermal stability. J Mol Biol 345, 559–577.
21 Singh SK, Matsuno K, LaPorte DC & Banaszak LJ
(2001) Crystal structure of Bacillus subtilis isocitrate
dehydrogenase at 1.55 A
˚
. Insights into the nature of

substrate specificity exhibited by Escherichia coli isoci-
trate dehydrogenase kinase ⁄ phosphatase. J Biol Chem
276, 26154–26163.
22 Ceccarelli C, Grodsky NB, Ariyaratne N, Colman RF
& Bahnson BJ (2002) Crystal structure of porcine mito-
chondrial NADP
+
-dependent isocitrate dehydrogenase
complexed with Mn
2+
and isocitrate. Insights into the
enzyme mechanism. J Biol Chem 277, 43454–43462.
23 Xu X, Zhao J, Xu Z, Peng B, Huang Q, Arnold E &
Ding J (2004) Structures of human cytosolic NADP-
dependent isocitrate dehydrogenase reveal a novel self-
regulatory mechanism of activity. J Biol Chem 279,
33946–33957.
24 Imada K, Sato M, Tanaka N, Katsube Y, Matsuura Y
& Oshima T (1991) Three-dimensional structure of a
highly thermostable enzyme, 3-isopropylmalate dehydro-
genase of Thermus thermophilus at 2.2 A
˚
resolution.
J Mol Biol 222, 725–738.
25 Hurley JH & Dean AM (1994) Structure of 3-isopropyl-
malate dehydrogenase in complex with NAD
+
: ligand-
induced loop closing and mechanism for cofactor
specificity. Structure 2, 1007–1016.

26 Wallon G, Kryger G, Lovett ST, Oshima T, Ringe D &
Petsko GA (1997) Crystal structures of Escherichia coli
and Salmonella typhimurium 3-isopropylmalate dehydro-
genase and comparison with their thermophilic counter-
part from Thermus thermophilus. J Mol Biol 266,
1016–1031.
27 Hardy F, Vriend G, van der Vinne B, Frigerio F,
Grandi G, Venema G & Eijsink VG (1994) The effect
of engineering surface loops on the thermal stability of
Bacillus subtilis neutral protease. Protein Eng 7, 425–
430.
28 Rossmann MG, Moras D & Olsen KW (1974) Chemical
and biological evolution of nucleotide-binding protein.
Nature 250, 194–199.
29 Chen R, Greer AF & Dean AM (1997) Structural con-
straints in protein engineering – the coenzyme specificity
of Escherichia coli isocitrate dehydrogenase. Eur J Bio-
chem 250, 578–582.
30 Hurley JH, Chen R & Dean AM (1996) Determinants
of cofactor specificity in isocitrate dehydrogenase: struc-
ture of an engineered NADP
+
-NAD
+
specificity-rever-
sal mutant. Biochemistry 35, 5670–5678.
31 Miyazaki K & Oshima T (1994) Co-enzyme specificity
of 3-isopropylmalate dehydrogenase from Thermus ther-
mophilus HB8. Protein Eng 7, 401–403.
32 Steen IH, Lien T, Madsen MS & Birkeland NK (2002)

Identification of cofactor discrimination sites in NAD-
isocitrate dehydrogenase from Pyrococcus furiosus. Arch
Microbiol 178, 297–300.
33 Garnak M & Reeves HC (1979) Phosphorylation of iso-
citrate dehydrogenase of Escherichia coli. Science 203,
1111–1112.
34 Borthwick AC, Holms WH & Nimmo HG (1984) The
phosphorylation of Escherichia coli isocitrate dehydro-
genase in intact cells. Biochem J 222, 797–804.
35 LaPorte DC & Koshland DE Jr (1982) A protein with
kinase and phosphatase activities involved in regulation
of tricarboxylic acid cycle. Nature 300, 458–460.
36 LaPorte DC (1993) The isocitrate dehydrogenase
phosphorylation cycle: regulation and enzymology.
J Cell Biochem 51, 14–18.
37 Kauzmann W (1959) Some factors in the interpretation
of protein denaturation. Adv Protein Chem 14, 1–63.
38 Dill KA (1990) The meaning of hydrophobicity. Science
250, 297–298.
39 Hennig M, Sterner R, Kirschner K & Jansonius JN
(1997) Crystal structure at 2.0 A
˚
resolution of phos-
phoribosyl anthranilate isomerase from the hyperther-
mophile Thermotoga maritima: possible determinants of
protein stability. Biochemistry 36, 6009–6016.
40 Lim JH, Yu YG, Han YS, Cho S, Ahn BY, Kim SH &
Cho Y (1997) The crystal structure of an Fe-superoxide
dismutase from the hyperthermophile Aquifex pyrophilus
at 1.9 A

˚
resolution: structural basis for thermostability.
J Mol Biol 270, 259–274.
41 Voorhorst WG, Warner A, de Vos WM & Siezen RJ
(1997) Homology modelling of two subtilisin-like
proteases from the hyperthermophilic archaea Pyrococ-
cus furiosus and Thermococcus stetteri. Protein Eng 10,
905–914.
42 Knapp S, Kardinahl S, Hellgren N, Tibbelin G, Schafer
G & Ladenstein R (1999) Refined crystal structure of a
superoxide dismutase from the hyperthermophilic
Structure and thermal stability of T. maritima M. Karlstro
¨
m et al.
2866 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet
archaeon Sulfolobus acidocaldarius at 2.2 A
˚
resolution.
J Mol Biol 285, 689–702.
43 Fersht AR & Serrano L (1993) Principles in protein sta-
bility derived from protein engineering experiments.
Curr Opin Struct Biol 3, 75–83.
44 Yip KS, Britton KL, Stillman TJ, Lebbink J, de Vos
WM, Robb FT, Vetriani C, Maeder D & Rice DW
(1998) Insights into the molecular basis of thermal stabi-
lity from the analysis of ion-pair networks in the
glutamate dehydrogenase family. Eur J Biochem 255,
336–346.
45 Auerbach G, Ostendorp R, Prade L, Korndorfer I,
Dams T, Huber R & Jaenicke R (1998) Lactate dehy-

drogenase from the hyperthermophilic bacterium
Thermotoga maritima: the crystal structure at 2.1 A
˚
resolution reveals strategies for intrinsic protein stabili-
zation. Structure 6, 769–781.
46 Chothia C (1984) Principles that determine the structure
of proteins. Annu Rev Biochem 53, 537–572.
47 Chothia C (1975) Structural invariants in protein fold-
ing. Nature 254, 304–308.
48 Fujinaga M, Berthet-Colominas C, Yaremchuk AD,
Tukalo MA & Cusack S (1993) Refined crystal structure
of the seryl-tRNA synthetase from Thermus thermophi-
lus at 2.5 A
˚
resolution. J Mol Biol 234, 222–233.
49 Chan MK, Mukund S, Kletzin A, Adams MW & Rees
DC (1995) Structure of a hyperthermophilic tungstop-
terin enzyme, aldehyde ferredoxin oxidoreductase.
Science 267, 1463–1469.
50 Goldman A (1995) How to make my blood boil. Struc-
ture 3, 1277–1279.
51 Argos P, Rossman MG, Grau UM, Zuber H, Frank G
& Tratschin JD (1979) Thermal stability and protein
structure. Biochemistry 18, 5698–5703.
52 Arnott MA, Michael RA, Thompson CR, Hough DW
& Danson MJ (2000) Thermostability and thermoactiv-
ity of citrate synthases from the thermophilic and
hyperthermophilic archaea, Thermoplasma acidophilum
and Pyrococcus furiosus. J Mol Biol 304, 657–668.
53 Burley SK & Petsko GA (1985) Aromatic–aromatic

interaction: a mechanism of protein structure stabiliza-
tion. Science 229, 23–28.
54 Kannan N & Vishveshwara S (2000) Aromatic clusters:
a determinant of thermal stability of thermophilic pro-
teins. Protein Eng 13, 753–761.
55 Barlow DJ & Thornton JM (1983) Ion-pairs in proteins.
J Mol Biol 168, 867–885.
56 Szilagyi A & Zavodszky P (2000) Structural differences
between mesophilic, moderately thermophilic and extre-
mely thermophilic protein subunits: results of a compre-
hensive survey. Structure 8, 493–504.
57 Elcock AH (1998) The stability of salt bridges at high
temperatures: implications for hyperthermophilic pro-
teins. J Mol Biol 284, 489–502.
58 Dominy BN, Minoux H & Brooks CL 3rd (2004) An
electrostatic basis for the stability of thermophilic pro-
teins. Proteins 57, 128–141.
59 Lebbink JH, Knapp S, van der Oost J, Rice D, Laden-
stein R & de Vos WM (1999) Engineering activity and
stability of Thermotoga maritima glutamate dehydrogen-
ase. II: Construction of a 16-residue ion-pair network at
the subunit interface. J Mol Biol 289, 357–369.
60 Horovitz A, Serrano L, Avron B, Bycroft M & Fersht
AR (1990) Strength and co-operativity of contributions
of surface salt bridges to protein stability. J Mol Biol
216, 1031–1044.
61 Spassov VZ, Karshikoff AD & Ladenstein R (1994)
Optimization of the electrostatic interactions in proteins
of different functional and folding type. Protein Sci 3,
1556–1569.

62 Schuck P (2000) Size-distribution analysis of macro-
molecules by sedimentation velocity ultracentrifugation
and lamm equation modeling. Biophys J 78, 1606–
1619.
63 Otwinowski Z & Minor W (1997) Processing of X-ray
diffraction data collected in oscillation mode. Methods
Enzymol 276, 307–326.
64 Collaborative Computational Project Number 4. (1994)
The CCP4 suite: programs for protein crystallography.
Acta Crystallogr D Biol Crystallogr 50, 760–763.
65 McCoy AJ, Grosse-Kunstleve RW, Storoni LC & Read
RJ (2005) Likelihood-enhanced fast translation func-
tions. Acta Crystallogr D Biol Crystallogr 61, 458–464.
66 Brunger AT, Adams PD, Clore GM, DeLano WL, Gros
P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J, Nilges
M, Pannu NS et al. (1998) Crystallography & NMR
system: a new software suite for macromolecular struc-
ture determination. Acta Crystallogr D Biol Crystallogr
54, 905–921.
67 Jones TA, Zou JY, Cowan SW & Kjeldgaard. (1991)
Improved methods for building protein models in elec-
tron density maps and the location of errors in these
models. Acta Crystallogr A 47, 110–119.
68 Lamzin VS & Wilson KS (1993) Automated refinement
of protein models. Acta Crystallogr D Biol Crystallogr
49, 129–147.
69 Laskowski RA, Moss DS & Thornton JM (1993) Main-
chain bond lengths and bond angles in protein struc-
tures. J Mol Biol 231, 1049–1067.
70 Kleywegt GJ (1996) Use of non-crystallographic sym-

metry in protein structure refinement. Acta Crystallogr
D Biol Crystallogr 52, 842–857.
71 Russell RB & Barton GJ (1992) Multiple protein
sequence alignment from tertiary structure comparison:
assignment of global and residue confidence levels.
Proteins 14, 309–323.
72 Kabsch W & Sander C (1983) Dictionary of protein sec-
ondary structure: pattern recognition of hydrogen-
M. Karlstro
¨
m et al. Structure and thermal stability of T. maritima
FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet 2867
bonded and geometrical features. Biopolymers 22, 2577–
2637.
73 Kleywegt GJ & Jones TA (1994) Detection, delineation,
measurement and display of cavities in macromolecular
structures. Acta Crystallogr D Biol Crystallogr 50, 178–
185.
74 McDonald IK & Thornton JM (1994) Satisfying hydro-
gen bonding potential in proteins. J Mol Biol 238, 777–
793.
75 Baker EN & Hubbard RE (1984) Hydrogen bonding
in globular proteins. Prog Biophys Mol Biol 44, 97–
179.
76 DeLano WL (2002) The PyMOL Molecular Graphics
System. DeLano Scientific, San Carlos, CA.
77 Barton GJ (1993) ALSCRIPT: a tool to format multiple
sequence alignments. Protein Eng 6 , 37–40.
Structure and thermal stability of T. maritima M. Karlstro
¨

m et al.
2868 FEBS Journal 273 (2006) 2851–2868 ª 2006 Karolinska Institutet