Protein Engineering vol.11 no.10 pp.867–872, 1998
Proteins from thermophilic and mesophilic organisms essentially
do not differ in packing
Andrey Karshikoff1 and Rudolf Ladenstein
Department of Biosciences at Novum, Karolinska Institutet, NOVUM,
S-14157 Huddinge, Stockholm, Sweden
1To
whom correspondence should be addressed
The role of the packing density in the elevation of thermal
stability of proteins from thermophilic organisms is widely
discussed in the literature. In the present study, this issue
was reconsidered in the scale of an unbiased set of protein
structures. Partial specific volumes, void and cavity volumes
were calculated for a set of 80 non-homologous proteins
and for 24 proteins from thermophilic organisms and
analysed in the context of their possible role in thermal
stabilization. The results showed that there is no significant
difference between the two sets in respect to the partial
specific volume and cavity volume. The proteins from
thermophilic organisms showed a slight tendency of increasing void volume, i.e. reducing the packing density. However
this observation was not confirmed by the comparison of
this parameter for proteins within different structural
families. The results suggested that neither the reduction
of the packing density nor the reduction of the packing
defects can be considered as a common mechanism for
increasing the thermal stability of the proteins from thermophilic organisms. Combining the result from this and
our previous study we concluded that the electrostatic
interactions seem to be a common factor regulating the
thermal tolerance of proteins from thermostable organisms.
Keywords: protein/packing/voids/volume/thermal stability
Introduction
There are two aspects to the thermal tolerance of proteins in
thermophilic organisms. The first one is the understanding of
the physical principles of thermal stability, which is related to
common academic questions, such as protein folding. The
second aspect, which is of special importance for the engineering of proteins of industrial interest, is the determination of
‘traffic rules’ (Rehaber and Jaenicke, 1992; Böhm and Jaenicke,
1994) responsible for the elevation of thermal stability of
proteins.
To our knowledge, Perutz and co-workers (Perutz and Raidt,
1975; Perutz, 1978) were the first to address this problem on
a molecular level. On the basis of a comparison of the
structures of ferredoxin and hemoglobin A2 from mesophilic
and thermophilic organisms they have proposed that the
increased thermal stability of thermophilic proteins is due to
a few extra salt bridges and hydrogen bonds. During the last
two decades, the increasing number of available amino acid
sequences and three-dimensional structures offered the opportunity for a detailed comparison of related proteins from
mesophilic and thermophilic organisms. This, as well as a
number of site-directed mutagenesis experiments (Argos et al.,
1979; Eijsink et al., 1992; Goward et al., 1994), suggested
© Oxford University Press
that thermal stability results from an additive series of small
improvements at many locations in the molecule without
significant changes in the tertiary structure. However, whether
there are dominating factors is a question that remains still open.
The two most frequently discussed reasons for the increased
thermal stability of proteins from thermophilic organisms are
better hydrogen bonding and hydrophobic internal packing
(Vogt and Argos, 1997; Vogt et al., 1997). Thus for instance,
Britton et al. (1995) have stated that the improved packing
within the buried core of the protein plays an important role
in maintaining stability at extreme temperatures. Russell et al.
(1994) have noticed that the cavity volume of citrate synthase
from Thermoplasma acidophilum is about three times smaller
than that of pig citrate synthase. Comparing the three-dimensional structures of glutamate dehydrogenase from hyperthermophilic, thermophilic and mesophilic bacteria and archaea,
Knapp et al. (1997) have also found that the proteins from
thermophilic organisms are characterized by a reduced cavity
volume. As far as internal cavities can be treated as packing
defects, one can conclude that the reduction of cavity volume
of the proteins from thermophilic organisms increases their
thermal stability. This conclusion is in accord with the experimental observation that the creation of cavities reduces the
hydrophobic effect and stability of proteins (Eriksson et al.,
1992). On the basis of a survey of water-soluble oligomeric
proteins, Chan et al. (1995) have shown that the solvent
accessible surface area of the AoR dimer is 17% less then
expected.
On the other hand, Yip et al. (1995) have concluded that
the differences in thermal stability between mesophilic and
hyperthermophilic glutamate dehydrogenases cannot be associated with significant changes in packing density. In accordance
to this, the comparison of the accessible surfaces of thermostable glyceraldehyde-3-phosphate dehydrogenase (Korndörfer
et al., 1997) and glutamate dehydrogenases (Knapp et al.,
1997) has revealed that a reduction in surface area cannot be
considered as a major factor in determining the increased
thermal stability of these proteins. Apparently, these few
examples of correlation between thermostability and reduced
surface area reflect a quite controversial situation. Comparisons
of a possible correlation of increased packing density and
thermostability has resulted in a similar message: Vogt and
Argos (1997) have analysed the number of cavities and packing
density in different protein families containing at least one
member from thermophilic organisms and have found that
only in about half of the cases the increased packing density
correlates with the increased thermal stability.
In this short communication we present our analysis of a
possible correlation of packing density and increased thermal
stability of proteins from thermophilic organisms. Quantities
reflecting packing density, such as the partial specific volumes,
cavity and void volumes, were calculated for a set of 80 nonhomologous monomeric proteins and for a set of 24 proteins
from thermophilic and hyperthermophilic organisms. In addi867
A.Karshikoff and R.Ladenstein
Table I. PDB entry codes of the protein structures used in this work
From mesophilic organisms:
1CRN
9ins
2CTX
1HOE
2MCM
8RNT
2CDV
3CHY
1LPE
2I1B
3ADK
1SGT
1RHD
2GBP
1PHH
3PGK
2OVO
1UBQ
4CPV
1BP2
2RN2
9PAP
1FNR
2CPP
AAP1
1CC5
2TRX
2AZA
1MBD
1HNE
4APE
1PII
1ROP
2FXB
4FD1
3FGF
2SGA
2CNA
3TLN
1NPX
1NXB
351C
1YCC
1IF1
2FCR
1LTE
5CPA
1GLY
3MT2
3B5C
1TGI
3FXN
2ALP
1THM
1IPD
1PGD
1PI2
1PCY
1c2r
1END
1GKY
3BLM
2LBP
1COX
1R69
1SAR
1PAZ
2SNS
1COL
2CBA
1GOX
1LFI
1SN3
3FXC
GMF1
4CLN
8DFR
6ABP
1ALD
6CAN
From thermophilic organisms:
1CAA
2FXB
1RIS
1LDN
1BMD
1GD1
1AOR
1CYG
1PKP
1HDG
1CIU
1RIL
1XYZ
1SRY
2PRD
1IPD
3MDS
1PHP
1THM
1EFT
4PFK
(a)
1LNF
1GTM
aGlutamate
dehydrogenase from Thermus thermophilus (Knapp et al., 1997); co-ordinates under preparation for submission in PDB.
The entries are ordered according to the molecular weight.
tion, the void volumes of proteins in 14 families, showing
high structural homology with at least one member from
thermophilic microorganisms, were analysed. The contribution
of the voids to the internal packing was evaluated so that the
role of the folding and the chemical composition are excluded.
The results are discussed in the context of a possible role of
the cavity and void volumes for thermal stabilization of
proteins from thermophilic organisms.
Method of calculation
The volume of a protein molecule is calculated as a sum of
three terms,
V 5 Vp 1 Vv 1 Vc,
where Vp represents the volume occupied by the protein atoms
according to their van der Waals radii, Vv is the volume of
voids, which is the volume of the space embraced by the
molecular (contact) surface but not occupied by protein atoms,
and Vc is the volume of the internal cavities. An internal
cavity is defined as the space, which is sufficiently large to
accommodate at least one water molecule and is isolated from
the bulk so that water molecules cannot penetrate into this
space. Cavities which are smaller are treated as voids. The
water molecule was approximated by a spherical probe with
a radius of 1.4 Å.
The algorithm for calculation of the molecular volume uses
a simple grid integration. Initially, the protein molecule is
placed in a parallelepiped with a coarse cubic grid of 2 Å.
The edge lengths of the parallelepiped are equal to the
maximum lengths of the molecule in the three space directions
plus one probe layer, so that grid points on the walls belong
to the bulk. A flag is assigned to each node dependent on
whether it corresponds to the bulk or to one of the areas
specified for volume calculations (van der Waals moiety, voids
or cavities). All elementary cubes that belong to the bulk are
excluded from further considerations and the procedure is
repeated for a grid size of 1 Å. The final integration is
performed with a grid size of 0.2 Å.
Input data
The calculations were carried out for a set of 80 nonhomologous monomeric proteins with high resolution X-ray
structures, all available in the Protein Data Bank (PDB)
(Bernstein et al., 1977) and are listed in Table I. The entries
have been selected from the representative set of sequenceunbiased proteins proposed by Boberg et al. (1995). Structures
868
with a resolution .2.5 Å or with more than 10 missing atoms
were excluded from the data set. The set of structures of
proteins from thermophilic organisms contained 24 entries
(also given in Table I); 22 of them are non homologous
structures. In order to keep an equal representation of folding
patterns in the data sets, the calculations were performed for
monomers. For oligomeric proteins, the first subunit given in
PDB were used.
Test calculations
Different van der Waals radii may provide different results
(Rellick and Becktel, 1995). In order to assess the uncertainty
introduced by the use of van der Waals radii as parameters,
the volume calculations were performed for four different sets
of van der Waals radii. As far as the partial specific volume
is a directly measurable quantity, the most convenient way for
the assessment is to convert the calculated volumes into partial
specific volume, v° 5 0.6023V/M, where M is the molecular
weight. The test calculations were performed for eight proteins
with known three dimensional structure for which the experimental values of v° are available. The calculated values
together with the experimental data are listed in Table II. The
average deviations from the experiment vary only between 1.0
and 4.7%. The closest agreement with the experimental data
was obtained with the van der Waals radii set given by Kuhn
et al. (1995). Also, the results obtained with this set are very
close to those calculated by means of a more rigorous method,
based on Voronoi polyhedra (Harpaz et al., 1994). All results
discussed below are calculated with the van der Waals radii
given by Kuhn et al. (1995).
Results and discussion
Although protein volumes and packing density have earlier
been calculated and discussed by many researchers, we returned
to this topic again motivated by intensive discussions in the
literature which arise every now and then when a threedimensional structure of a thermostable protein is solved. If
improvement of packing is assumed to be a common mechanism for increasing thermal stability of proteins, physical
quantities, such as the partial specific volume or void volume,
should show a certain tendency distinguishing the proteins
from thermophilic and hyperthermophilic organisms from the
others. In order to detect possible trends we analysed the
properties of 24 proteins from thermostable organisms in the
context of a representative, unbiased set of 80 protein structures.
As far as the data set contains non-homologous proteins, we
Protein packing and thermal stability
Table II. Partial specific volumes (ml/g)
PDB code
Ribonuclease A, bovine
Lysozyme, chicken
Adenylate kinase, porcine
Concanavalin A, jack bean
Subtilisin, B. Licheniformis
Carbonic anhydrase B, human
Carboxypeptidase A, bovine
Malate dehydrogenase, porcine
Average deviation (%)
Experimental data
Calculated
a
b
a
b
c
d
0.702e
0.712g
0.740g
0.732g
0.731g
0.729g
0.733g
0.742g
0.707f
0.703f
0.694
0.707
0.720
0.728
0.720
0.732
0.723
0.740
1.0
0.678
0.696
0.704
0.713
0.711
0.717
0.713
0.726
2.7
0.658
0.677
0.685
0.697
0.696
0.701
0.699
0.732
4.7
0.663
0.683
0.690
0.701
0.699
0.706
0.703
0.713
4.6
0.730f
Van der Waals set from: a, Kuhn et al., 1995; b, Rashin et al., 1986; c, Chothia, 1975; d, Laskowski, 1995.
Data from: e, Chalikian et al., 1996; f, Rellick and Becktel, 1995; h, Gekko and Noguchi, 1979; g, Squire and Himmel, 1979.
ing to v° 5 0.729 ml/g. This value is lower than that given
earlier by Richards (1974) (about 0.75 ml/g), but is very close
to the average, 0.728 ml/g, calculated for the experimental
data listed in Table II. The slope calculated for the proteins
from thermostable organisms shows a somewhat higher partial
specific volume, v° 5 0.735 ml/g, which could even suggest
that these proteins are characterized by a lower packing
density. Practically, however, proteins from mesophilic and
thermophilic organisms are indistinguishable with respect to
v° because the observed difference is below the accuracy of
calculations.
Fig. 1. Molecular volume versus molecular weight of the proteins from the
representative set (open circles) and for the proteins from thermophilic
organisms (solid circles).
believe that the observed dependencies for this set are a feature
independent of protein source and functional properties. After
excluding two entries, one glycosyltransferase (1CYG or
1CIU) and one glutamate dehydrogenase [1GTM or glutamate
dehydrogenase from Thermus thermophilus (Knapp et al.,
1997), see Table I], the set of proteins from thermostable
organisms becomes unbiased. Thus, the comparison of the
results for the two data sets should reveal systematic differences
if different packing densities or packing defects are inherent
to the thermostable proteins.
Partial specific volumes
In Figure 1 the molecular volumes calculated for the two
protein sets are plotted versus the molecular weight. As
expected, the dependence of the volume on the molecular
weight is linear (with r-factor of 0.999) with a slope correspond-
Cavity volumes
It is already known (Rashin et al., 1986; Hubbard and Argos,
1996) that the cavity volumes in proteins are negligibly small
in comparison to the molecular volume. The calculations
performed without considering the internal cavities resulted in
a reduction of v° by only 0.7% on average (data not shown).
In spite of the negligible contribution to the packing density
as a whole, cavities are energetically significant and influence
the stability of proteins (Eriksson et al., 1992; Ishikava et al.,
1993). With increasing of the molecular weight both the cavity
volume and cavity number increase (Hubbard and Argos, 1994).
In other words, increasing molecular weight is accompanied by
creation of energetically unfavourable packing defects. If one
assumes that the reduction of cavity volume and number is a
common mechanism for the elevation of thermal stability of
proteins, it should be pronounced at least for proteins with
higher molecular weight. The dependence of the cavity volume
on the molecular weight of the proteins from the representative
sets and from thermophilic organisms is shown in Figure 2.
The distribution of the calculated values for the two sets are
essentially identical, in spite of that the linear fits may suggest
a certain reduction of the cavity volume for the proteins from
thermophilic organisms. However, the correlation factors of
both fits are too low and prohibitive for allowing definitive
conclusions.
Void volumes
The void volume is a component which amounts to 20–25%
of the molecular volume and directly reflects the packing
density. It consists of two parts: the first part, V9v, results from
the spherical representation of the atoms and the probe. The
second one, Vv, which is of interest, results from the space
that arises from the imperfect packing of the folded protein.
It can be obtained simply by subtracting V9v from the total
void volume, Vv 5 Vv – V9v. The volume, V9v, was obtained
869
A.Karshikoff and R.Ladenstein
Fig. 2. Cavity volume versus molecular weight of the proteins from the
representative set (open circles) and for the proteins from thermophilic
organisms (solid circles). Dashed line, linear fit of the values for the
representative set; solid line, linear fit of the values for the proteins from
thermophilic organisms.
Fig. 3. Void volume, with the term resulting from the spherical
approximation excluded (see text), versus molecular weight of the proteins
from the representative set (open circles) and for the proteins from
thermophilic organisms (solid circles).
as the sum of the of the void volumes separately calculated
for the backbone and each side chain,
V9v 5 Vv(backbone) 1 ΣiVv(side chain i),
maintaining their native conformations. In this way the contribution of the folding and chemical composition are eliminated,
so that Vv contains only the void volume due to the burial of
the backbone and the individual side chains in the environment
of their neighbours. The calculations showed that Vv is 50–
60% of the total void volume and it reduces to 40–45% only
in a few cases for proteins with low molecular weight. For
both data sets, the dependencies of Vv on the molecular weight
are visually very close to linear and the values calculated for
thermophilic proteins seem to be uniformly distributed among
those of the representative data set (Figure 3). The slopes
(given in Å3 per atom instead per kDa) are 2.33 Å3/Na for the
representative data set and 2.43 Å3/Na for the proteins form
thermophilic organisms. The latter value was calculated for
the unbiased set of 22 structures. The slight difference in the
slopes suggests that thermophilic proteins are somewhat less
tightly packed. However, this result must be taken with caution
until more three-dimensional structures of proteins from thermophilic organisms become available.
In fact, neither V nor ∆Vv are strongly linear functions of
the molecular weight. In Figure 4, the values of ∆Vv/Na (called
here specific void volume) are plotted versus the molecular
weight. The distribution of the values of the partial specific
volumes is essentially similar (data not shown). In our earlier
study (Spassov et al., 1995) we suggested that monomeric
globular proteins that form a well defined hydrophobic core
870
Fig. 4. Specific void volume versus molecular weight of the proteins from
the representative set (open circles) and for the proteins from thermophilic
organisms (solid circles). The dashed line defines the regions of the proteins
with under developed hydrophobic core (left hand side) and for proteins
with developed hydrophobic core (right hand side).
Protein packing and thermal stability
Table III. Values of the specific void volume (∆Vv/Na) for proteins within families of structurally homology
PDB code
Rubredoxin
1CAA
6RXN
1RDG
5RXN
7RXN
8RXN
Glutamate dehydrogenase
1GTM
Res.
Source
Topt
Monomers
1.8
1.5
1.4
1.2
1.5
1.0
Pyrococcus furiosus
Desulfovibro desulfuricans
Desulfovibro gigas
Clostridium pasteurianum
Desulfovibro vulgaris
Desulfovibro vulgaris
100
37
37
37
35
35
*1.63
1.70
1.86*
1.83
1.78
1.73
Pyrococcus furiosus
Thermotoga maritima
Clostridium symbiosum
100
80
37
*2.00
2.09*
2.04
*2.01
2.05*
2.03
Thermotoga maritima
Bacillus stearothermophilus
Escherichia coli
human
lobster
80
55
37
37
20
2.47*
2.37
2.44
2.36
*2.21
2.52
2.72*
2.60
*2.46
2.56
Thermus thermophilus
human
Pseudomonas Ovalis
Escherichia coli
Mycobacterium tuberculosis
75
37
37
37
37
2.36*
2.29
2.28
2.29
*2.12
2.45
2.42
*2.37
2.70*
2.52
Thermus flavus
porcine
pig
Escherichia coli
75
37
37
37
2.40
2.38
*2.33
2.53*
2.53*
2.51
*2.46
–
Thermus thermophilus
Escherichia coli
75
37
2.19*
*2.14
–
–
Thermus thermophilus
Escherichia coli
75
37
*2.38
2.43*
–
–
Clostridium thermocellum
Escherichia coli
60
37
2.61*
*2.39
2.65
–
Th. thermosulfurigenes
Bacillus stearothermophilus
Bacillus circulans, strain 251
Bacillus circulans, strain 8
60
55
35
35
*2.46
2.49
2.53
2.54*
–
–
–
–
Bacillus stearothermophilus
Lactobacillus casei
mouse
pig
porcine
dogfish
55
37
37
37
37
15
*2.28
2.48
2.29
2.40
2.60*
2.39
Bacillus stearothermophilus
Saccharomyces cerevisiae
55
30
2.52*
*2.14
–
–
Bacillus thermoproteolyticus
Bacillus cereus
55
30
*2.27
2.34*
–
–
Bacillus thermoproteolyticus
Peptococcus aerogenes
Clostridium acidurici
55
37
28
1.72*
1.64
*1.63
–
–
–
Thermoactinomyces vulgaris
50
1.97
–
Bacillus
Bacillus
Bacillus
Bacillus
37
30
30
26
1.99*
1.89
*1.86
1.98
–
–
–
–
1GDM
3-phosphate dehydrogenase
1HDG
2.5
1GD1
1.8
1GAD
1.8
3GPD
3.5
4GPD
2.8
Superoxide dismutase
3mds
1.8
1ABM
2.2
3SDP
2.1
1ISA
1.8
1IDS
2.0
Malate dehydrogenase
1BMD
1.9
4MDH
2.5
1MLD
1.9
1EMD
1.9
Ribonuclease H
1RIL
2.8
2RN2
1.5
Hydrolase
2PRD
2.0
1INO
2.2
Xylanhydrolase
1XYZ
1.4
2EXO
1.8
Glycosyltransferase
1CIU
2.3
1CYG
2.5
1CDG
2.0
1CGT
2.0
Lactate dehydrogenase
1LDN
2.0
1LLC
3.0
2LDX
3.0
5LDH
2.7
9LDT
2.0
6LDH
2.0
Phosphoglycerate kinase
1PHP
1.6
3PGK
2.5
Thermolysin
1LNF
1.7
1NPC
2.0
Ferredoxin
2FXB
2.3
1FDX
2.0
1FCA
1.8
Thermitase
1THM
1.4
Subtilisin
1SCA
2.0
2ST1
1.8
1SUP
1.6
1ST3
1.4
Oligomers
–
–
–
–
–
–
D-glyceraldehyde
Licheniformis Carlsberg
amyloliquefaciens BPN9
amyloliquefaciens BPN9
Lentus
*2.70
–
2.73
2.58
3.13*
2.70
The thermostable partners are given in bold; the hyperthermostable partners are underlined. Asterisks denote the minimum (left hand side) and maximum (left
hand side) values of ∆Vv/Na.
871
A.Karshikoff and R.Ladenstein
have at least a molecular weight above about 14 kDa (see also
Privalov, 1989). If we formally assume the molecular mass
of 14 kDa as a limit below which proteins may have an
underdeveloped hydrophobic core, two regions of the distribution of specific void volume in the scale of molecular weight
can be distinguished. The first one, defined by the proteins
below this limit, is broad, with values of ∆Vv/Na between 1.1
and 2.6, whereas the second one, corresponding to proteins
with well developed hydrophobic core, is narrower, with values
of ∆Vv/Na close to the average. If our hypothesis (Spassov
et al., 1995) is correct, one can conclude that the reduction of
the values of ∆Vv/Na for the small proteins is more likely due
to the lack of buried area (underdeveloped hydrophobic core)
than due to a higher packing density. For proteins with a well
developed hydrophobic core the specific void volume remains
independent of the molecular weight. The proteins form
thermophilic organisms essentially follow this principle, i.e.
they do not differ from all other proteins (the representative
data set) in terms of the parameter ∆Vv/Na.
Proteins from structurally homologous families
The analysis of the specific void volume in the context of a
representative, unbiased data set did not show a clear trend
distinguishing thermostable from the other proteins, or if there
is a trend, it is of reduction of packing. We compared this
parameter also for proteins within different structural families
with at least one member from thermophiles. The results for
monomers and oligomers are listed in Table III. In six cases
the proteins from thermostable organisms have the lowest
values of ∆Vv/Na, i.e. they are more tightly packed than their
mesophilic partners within one family. In six other cases,
however, the opposite tendency is observed. This result is
essentially similar to that obtained by Vogt and Argos (1997).
Thus, the packing density is not a dominant factor contributing
to the increased thermal stability of the proteins from thermophilic organisms.
In our previous study (Spassov et al., 1995) we have found
that the optimization of both the electrostatic and protein–
solvent interactions correlates very well with the thermal
stability of proteins. It has been shown that for small proteins
the electrostatic optimization (the reduction of the repulsive
contacts) is a dominant factor, whereas with increasing molecular weight the role of the optimization of protein–solvent
interactions becomes dominant. According to the definition
(Spassov et al., 1995), this means that thermostable proteins
bury the hydrophobic moiety more efficiently. The results
obtained in the present study suggest that this is not accompanied by increasing packing density. More likely, this is related
to an exposure of polar and charged groups and thus creating
favourable conditions for the formation of stabilizing hydrogen
bonds and salt bridges at the protein surface. Indeed, in all
families, we have noticed a systematic increasing in the number
of salt bridges in thermostable proteins (Spassov et al., 1995).
Combining the results of the two studies, we conclude that
the increasing of the number of salt bridges together with the
increasing of the optimization of the electrostatic interactions
represent one of the most important mechanisms for the
elevation of the thermal stability of proteins.
References
Argos,P., Rossman,M.G., Gran,K.M., Zuber,H., Frank,G. and Tratschin,J.D.
(1979) Biochemistry, 18, 5698–5703.
872
Bernstein,F.C., Koetzle,T.F., Williams,G.J.B., Meyer,E.D.J., Brice,M.D.,
Rogers,J.R., Kennard,O., Shinaniushi,T. and Tasumi,M. (1977) J. Mol.
Biol., 112, 535–542.
Boberg,J., Salakoski,T. and Vihinen,M. (1995) Protein Engng, 8, 501–503.
Böhm,G. and Jaenicke,R. (1994) Int. J. Pept. Protein Res., 43, 97–106.
Britton,K.L., Baker,P.J., Borges,K.M.M., Engel,P.C., Pasquo,A., Rice,D.,
Robb,F.T., Scandurra,R., Stillman,T.J. and Yip,K.S.P. (1995) Eur. J.
Biochem., 229, 688–695.
Chalikian,T.V., Totrov,M., Abagyan,R. and Breslauer,K.J. (1996) J. Mol. Biol.,
260, 588–603.
Chan,M.K., Mukund,S., Kletzin,A., Adams,M.W.W. and Rees,D.C. (1995)
Science, 267, 1463–1469.
Chothia,C. (1975) Nature, 254, 304–308.
Eijsink,V.G.H., Vriend,G., van der Burg,B., van der See,J.R. and Venema,G.
(1992) Protein Engng, 5, 165–170.
Eriksson,A.E., Baase,W.A., Zhang,X.-J., Heinz,D.W., Blaber,M., Baldwin,E.P.
and Matthews,B.W. (1992) Science, 255, 178–183.
Gekko,K. and Noguchi,H. (1979) J. Phys. Chem., 83, 2706–2711.
Goward,C.R., Miller,J., Nicholls,D.J., Irons,L.I., Scawen,M.D., O’Brien,R.
and Chowdhry,B.Z. (1994) Eur. J. Biochem., 224, 249–255.
Harpaz,Y., Gerstein,M. and Chothia,C. (1994) Structure, 2, 641–649.
Hubbard,S.J. and Argos,P. (1994) Protein Sci., 3, 2194–2206.
Hubbard,S.J. and Argos,P. (1996) Protein Engng, 8, 1011–1015.
Ishikava,K., Okumura,M., Katayanagi,K., Kimura,S., Kanaya,S., Nakamura,H.
and Morikawa,K. (1993) J. Mol. Biol., 230, 592–542.
Knapp,S., de Vos,W.M., Rice,D. and Ladenstein,R. (1997) J. Mol. Biol., 267,
916–932.
Korndörfer,I., Steipe,B., Huber,R., Tomschy,A. and Jaenicke,R. (1997) J. Mol.
Biol., 246, 511–521.
Kuhn,L.A., Swanson,C.A., Pique,M.E., Tainer,J.A. and Getzoff,E.D. (1995)
Proteins, 23, 536–547.
Laskowski,R.A. (1995) J. Mol. Graph., 13, 232–330.
Perutz,M.F. (1978) Science, 201, 1187–1191.
Perutz,M.F. and Raidt,H. (1975) Nature, 255, 256–259.
Privalov,P.L. (1989) Annu. Rev. Biophys. Biophys. Chem., 18, 47–69.
Rashin,A.A., Iofin,M. and Honig,B. (1986) Biochemistry, 25, 3619–3625.
Rehaber,V. and Jaenicke,R. (1992) J. Biol. Chem., 267, 10999–11006.
Rellick,L.M. and Becktel,W.J. (1995) Meth. Enzymol., 259, 377–395.
Richards,F.M. (1974) J. Mol. Biol., 82, 1–14.
Russell,R.J.M., Hough,D.W., Danson,M.J. and Taylor,G.L. (1994) Structure,
2, 1157–1167.
Spassov,V.Z., Karshikoff,A.D. and Ladenstein,R. (1995) Protein Sci., 4,
1516–1527.
Squire,P.G. and Himmel,M.E. (1979) Arch. Biochem. Biophys., 196, 165–177.
Vogt,G. and Argos,P. (1997) Folding Design, 2, S40–S46.
Vogt,G., Woell,S. and Argos,P. (1997) J. Mol. Biol., 269, 631–643.
Yip,K.S.P.,
Stillman,T.J.,
Britton,K.L.,
Artimiuk,P.J.,
Baker,P.J.,
Sedelnikova,S.E., Engel,P.C., Pasquo,A., Chiaraluce,R., Consalvi,V.,
Scandurra,R. and Rice,D.W. (1995) Structure, 3, 1147–1158.
Received February 16, 1998; revised June 1, 1998; accepted June 17, 1998