Tài liệu Báo cáo khoa học: Selection of stably folded proteins by phage-display with proteolysis docx

MINIREVIEW
Selection of stably folded proteins by phage-display with proteolysis
Yawen Bai and Hanqiao Feng
Laboratory of Biochemistry, National Cancer Institute, Bethesda, MD, USA
To facilitate the process of protein design and learn the basic
rules that control the structure and stability of proteins,
combinatorial methods have been developed to select or
screen proteins with desired properties from libraries of
mutants. One such method uses phage-display and proteo-
lysis to select stably folded proteins. This method does not
rely on specific properties of proteins for selection. There-
fore, in principle it can be applied to any protein. Since its
first demonstration in 1998, the method has been used to
create hyperthermophilic proteins, to evolve novel folded
domains from a library generated by combinatorial shuffling
of polypeptide segments and to convert a partially unfolded
structure to a fully folded protein.
Keywords: hydrophobic repacking; phage-display; protein
design; proteolysis.
Introduction
There are two basic biophysical issues in protein design. One
is to find mutations that make proteins thermodynamically
more stable. The other is to find an amino acid sequence for
a polypeptide chain that will fold to a target structure. The
first issue is important for developing therapeutic drugs and
useful enzymes in industry. The second issue is more critical
for learning and testing the principles of protein folding.
Although significant progress has been made towards
rational design of proteins with simple motifs [1–3], it is
still difficult to design native-like proteins with globular
structures [4]. In addition, it is still not completely clear how

the stability of a protein is encoded in the protein’s sequence
and how individual amino acid residues contribute to
stability. Thus, combinatorial approaches to select or screen
proteins with the desired properties from libraries of mutant
proteins have been sought [5–8]. Phage-display coupled with
proteolysis for selection of stably folded proteins was one
such recently developed method. It was first demonstrated
in 1998 by two research groups [9,10] based on the following
considerations: (a) stably folded and well structured proteins
should be more resistant to protease digestion than those
less stable and poorly folded; (b) M13 and fd phages are
resistant to cleavage by many proteases; (c) the surface g3p
proteins of phage are needed for bacteria infection, which
allows the coupling between the cleavage of inserted guest
proteins and the loss of phage infection. In these demon-
strations (Fig. 1A), guest variants of a protein (mutants of
barnase and ribonuclease T4, respectively) with different
thermodynamic stability were inserted into the region
between the C-terminal region and the two N-terminal
domains of the g3p. After several rounds of protease
digestion and amplification of the library, variants with high
thermodynamic stability were enriched over those that have
low thermodynamic stability.
A similar approach but with a different selection
strategy (Fig. 1B) was developed later [11,12]. His-tagged
guest proteins were fused at the N-terminal of the g3p
protein. Selection of phages with uncut proteins was made
using Ni coated chips and monitored using surface
plasmon resonance. In this study, the authors aim to
demonstrate that stably folded protein structures can be

obtained by focusing on the design of a hydrophobic core:
a core-directed protein design approach. This design
process has three steps: first, generation of multiple core
mutants of the target protein; secondly, display of
mutants on the phage surface; and finally, selection for
stably folded mutants by challenging the system with
protease. The concept was demonstrated by studying the
core packing of ubiquitin. Eight hydrophobic core
residues in ubiquitin were mutated randomly and simul-
taneously with all 20 amino acids. The mutants were
displayed on the surface of the phage and challenged with
the protease chymotrypsin. The selected sequences were
found to be very close to the wild type, consistent with the
hypothesis that a hydrophobic core may be used to direct
protein design for globular proteins. The authors conclu-
ded that the best solution to the core-packing problem for
ubiquitin is the natural wild type sequence, or residue
combinations extremely close to it. This result is similar
to the earlier conclusion obtained using combinatorial
computational methods [13,14]. Intriguingly, however, all
selected proteins are less stable than the wild type, and
wild type protein is not selected, suggesting that other
factors may prohibit the selection of the most stable
proteins. One possible reason is that there are protease
digestion sites in the loop region that might unfold
locally (Table 1).
Correspondence to Y. Bai, Laboratory of Biochemistry, National
Cancer Institute, Building 37, Room 6114E, Bethesda, MD 20892,
USA. Fax: + 1 301 402 3095, Tel.: + 1 301 594 2375,
E-mail:

Abbreviations: Bc-Csp, cold shock protein from Bacillus caldolyticus;
Bs-CspB, cold shock protein from Bacillus subtilis;cytb
562
,
cytochrome b
562
.
(Received 5 January 2004, revised 11 February 2004,
accepted 5 March 2004)
Eur. J. Biochem. 271, 1609–1614 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2004.04074.x
Towards selection of stable proteins
Despite the significant effort that has been made toward
studying the stability of proteins, it is still not fully
understood how the stability of a protein is encoded in its
sequence and how individual amino acid residues contribute
to stability. To learn the factors that stabilize the proteins,
researchers have recently become interested in studying the
proteins from thermophilic organisms. These proteins are
stable at very high temperatures. Thus, it is hoped that the
rules for stabilizing proteins may be revealed after compar-
ing the thermophilic proteins with the mesophilic proteins.
This is, however, complicated by the fact that a lot of
neutral mutations exist, which makes it difficult to find
which mutation or combination of mutations is important
for stability. To gain insights into this issue, Martin et al.
[15] used phage-display with proteolysis to convert the
mesophilic cold shock protein, Bs-CspB, from Bacillus
subtilis to a hyperthermophilic protein in a relatively
controlled manner. Mesophilic Bs-CspB differs from its
thermophilic counterpart Bc-Csp from Bacillus caldolyticus

at 12 surface-exposed positions. In their study, six of these
positions were randomized by saturation mutagenesis, in
which any of the 20 amino acids can occur at each of the
six positions. Selection was made under two different
conditions: in the presence of guanidinium chloride and at
elevated temperature. Several of the selected mutants are
significantly more stable than the naturally thermostable
homolog Bc-Csp, and the best variant reaches the stability
of Tm-Csp (the homolog from the hyperthermophile
Thermototoga maritime). Interestingly, this variant differs
from Tm-Csp at five positions and from Bc-Csp at all six
randomized positions, indicating that proteins can be
strongly stabilized by many different sets of surface
mutations. Furthermore, the selection is found to be
dependent on selection conditions. In the ionic denaturant
(guanidinium chloride) solution, nonpolar surface inter-
actions were optimized, whereas at elevated temperatures
variants with improved electrostatics were selected, pointing
to different strategies for stabilization at the protein surface.
Pedersen et al. [16] also attempted a similar experiment to
seek stable mutants of barnase. Using a subset of codons
that only encode hydrophobic residues, a library of barnase
mutants was made by randomizing the residues at the 17
positions that are different from those in the homologue
protein binase. The library was then challenged with trypsin.
Among the 20 clones selected, 10 were studied for their
stability. None of the selected mutants was found to be
more stable than the wild type barnase. This result has been
attributed to possible local unfolding in barnase (Table 1).
Towards selection of protein structures

It has been suggested that proteins occurring in nature have
been evolved by the assembly of nonhomologue genes. For
small protein domains, they may have evolved by assembly
and/or exchange of small gene segments, leading to
diversification of the domain architecture and even genera-
tion of an entirely new fold. Riechman & Winter [17] have
investigated this proposal. Using phage-display and pro-
teolysis, selected stably folded proteins from a phage library
in which the DNA encoding the N-terminal half of a
b-barrel domain (from cold shock protein CspA) was
substituted with fragmented genomic Escherichia coli DNA.
The phage library was then challenged by several proteases.
Table 1. Summary of proteins studied using phage-display with proteolysis. In barnase, R110 is the last residue, R59 and K62 are in the loop region.
In Ubiquitin, residues F45 and Y59 are in the loop region. TS, tagged selection; SIP, selectively infective phage.
Molecule Protease Positions Stability Structure change Cutting site in loops Method
Barnase Trypsin surface/core decrease no yes SIP
Ubiquitin Chymotrypsin core decrease no yes TS
RNaseT1(4A) Trypsin/Chymotrypsin/Pepsin surface increase no no SIP
Apocyt b
562
Arg-c core increase yes no TS
CspA Trypsin/Thermolysin surface/core increase yes no SIP
CspB Trypsin surface increase no no SIP
Fig. 1. Two different ways for selecting stably folded proteins using
phage-display with proteolysis. (A) Selectively infective phage (SIP) uses
the fact that the N-terminal domains (N1, N2) of the minor coat
protein (g3p) are responsible for binding and infection in E. coli.Thus,
incorporation of a library of target proteins between the N-terminal
domain and the C-terminal (CT) domain allows a protease-based
selection because proteolysis of the target protein also removes the

N-terminal domains and prevents the infection of phage in E. coli.(B)
A library of target proteins with a tag can be fused to the g3p protein
on the surface of the phage. The tag can be a His-tag [11] or antibody-
binding proteins such as the protein AB-domain [18]. Removal of
unstable proteins by proteolysis also removes the tag and prevents the
phage associated with it from being selected for further infection of
E. coli.
1610 Y. Bai and H. Feng (Eur. J. Biochem. 271) Ó FEBS 2004
Four proteins selected from the library were soluble and
were characterized using NMR, CD and amide hydrogen
exchange. The CD spectra indicated formation of a b-sheet
structure consistent with the segment from the CspA.
Thermal melting of the selected proteins was cooperative.
The thermodynamic stability of the proteins ranged from
1.8 to 5.3 kcalÆmol
)1
. NMR spectra of these proteins
showed sharp peaks, suggesting folded proteins were
selected. Detailed structural information is needed to
demonstrate its final success.
In a more recent test for the core-directed design
proposal, Chu et al. [18] have converted a partially unfolded
state, apocytochrome b
562
, to a fully folded four-helix
bundle protein in the absence of any cofactors. In this
work, the authors used the method similar to that of
Finucane et al. [11] except that the protein A B-domain
instead of His-tag was used to select the folded proteins.
Cytochrome b

562
(cyt b
562
) is a four-helix bundle protein
with a heme holding the N- and C-terminal helices
(Fig. 2A). In the absence of heme, apocytochrome b
562
adopts a partially unfolded conformation with the
C-terminal helix largely unfolded while the other three
helices remain folded. To create a four-helix bundle protein
in the absence of heme, four residues at positions 7, 98, 102
and 106, that are expected to form a hydrophobic core and
substitute the heme, were mutated. Residue 7 was changed
to Trp to provide a fluorescence probe for studying the
protein’s physical properties. The other three positions were
randomly mutated. In addition, residue 99 in the region for
redesign was substituted with Arg to provide a specific
cutting site for protease Arg-c. This library of mutants was
displayed on the surface of phage and challenged with pro-
tease Arg-c to select stably folded proteins. The consensus
sequences in this selection showed some interesting results.
Hydrophobic residues occurred at position 98 while hydro-
philic residues occurred at positions 102 and 106. Never-
theless, the selected proteins were thermodynamically very
stable.
The structure of one of the selected proteins with Ile,
Asn and Gly at positions of 98, 102 and 106, was
characterized using multidimensional NMR. All four
helices were formed in the structure. Furthermore, site-
directed mutagenesis was used to change one of the two

hydrophilic residues to a hydrophobic residue. This muta-
tion increases the stability of the protein, suggesting that the
selection was not solely based on the protein’s global
stability. Based on the comparison between the NMR
structure of the selected protein and a crystal structure of
another mutant that has two hydrophobic residues substi-
tuting for the two hydrophilic residues, an interpretation for
the selection result is proposed. In the X-ray structure,
the hydrophobic interaction distorted the last turn of the
C-terminal helix, which may make the site for proteolysis
more accessible. We have recently obtained the high
resolution structure for the selected protein (Fig. 2B)
(H. Feng & Y. Bai, unpublished result). The structure
shows that the C-terminal end of the fourth helix moves
slightly and uses hydrophobic residues (Y101 and Y105)
that are originally packed between the fourth and the third
helices in the wild type protein, to participate in the new
hydrophobic core of the structure. The two hydrophilic
residues in the selected structure are now exposed, which
explains why hydrophilic residues were selected at these two
positions. This result confirms the idea of using a hydro-
phobic core to direct protein design. However, it also shows
that proteins can make subtle structural changes to find
alternatives to fulfil the hydrophobic interactions, which
makes it difficult to predict the selection result.
Effect of flexible loops and partially folded
intermediate on selection
Depending on the position of protease cutting sites in the
structure, the existence of flexible loops and partially
unfolded states could have a significant effect on the result

of selection. If the cutting sites are in the flexible loop of the
native structure, they could prevent the selection of stable
proteins. By examining the structures of the proteins studied
by the phage-display and proteolysis, we found that
protease cutting sites exist in the loop regions for both
cases (barnase and ubiquitin) in which the selection did not
produce very stable proteins (Table 1). A more serious
problem can arise from the existence of partially unfolded
states that have the protease digestion sites in their unfolded
regions (Fig. 3). This is because the mutations in the folded
regions of the intermediate do not significantly change the
relative population between the intermediate and the fully
folded state. Therefore, little evolution pressure can be
added for selection of stable proteins if mutations are made
in the folded region. To be able to select stable mutants
using phage-display and proteolysis, it is necessary that
the protease cutting sites be close to the mutation sites or
in the region that is exposed only upon global unfolding.
The stable region may be determined by the existence of the
slowest exchanging amide protons.
Fig. 2. Effect of structural change on the selection. (A) Structure of
cyt b
562
. Residues M7 and H102 are the ligands of heme. Heme is
represented with a red ellipse. (B) Hydrophobic residues (Y101 and
Y105) that were originally packed between the third and fourth helix in
the cyt b
562
have become part of the new hydrophobic core. Side
chains at positions 102 and 106 that face inside in cyt b

562
have become
exposed in the selected structure.
Ó FEBS 2004 Selection of folded proteins by phage-display (Eur. J. Biochem. 271) 1611
Design of native-like proteins
The major difficulty encountered in protein design has been
that designed proteins often have more heterogeneous
structures than those of typical natural proteins. The
initially designed structures often had the correct secondary
structure and topology but lacked the well-packed hydro-
phobic core that is characteristic of most natural proteins
[19,20]. Iterative experimental design processes are normally
required to achieve the final target [3,21]. This problem
becomes a more critical issue because the proteins designed
recently by computational methods have also failed in this
aspect. Nauli et al. [22] have redesigned the second b-hairpin
of the protein G B1-domain and obtained a protein that
is more stable than the wild type by  4kcalÆmol
)1
.The
structure of this protein has been solved using the X-ray
crystallography method [23]. It is found that the B-factors of
the mutated residues are much higher than those of other
residues, indicating that there are significant dynamic
motions in the redesigned structure, which may contribute
in part to the thermodynamic stability. We also examined
another computer-designed protein G B1-domain variant
by Malakauskas & Mayo [24]. This redesigned protein is
also more stable than the wild type by  4kcalÆmol
)1

.Inthis
case, the dynamic behavior of the redesigned protein is even
more dramatic. Several cross peaks that correspond to the
redesigned residues in the
1
H-
15
NHSQCspectrumhave
very weak intensities even though the structure of the
redesigned protein has been solved using NMR [24].
Examination of the mutations in the two computer
redesigned proteins shows that most of the mutations are
from polar to hydrophobic residues. Thus, the two designs
have essentially reversed the earlier de novo design practice,
in which polar residues were incorporated into the designed
hydrophobic core to obtain unique conformation at the
expense of protein stability [25]. Regarding this issue, it
should be noted that these redesigned proteins have 1D
1
H-
NMR spectra that look very much like those of native-like
structures. Therefore, it suggests 1D
1
H-NMR spectrum is
insufficient for determining whether a redesigned protein
has a more dynamic motion on a fine level and
1
H-
15
N

HSQC spectra may be a minimum requirement for char-
acterizing the dynamic behavior of redesigned proteins in
the future. As proteins with heterogeneous structures and
dynamic behavior in the native state are likely to be more
sensitive to protease digestion than those with well-packed
structures, phage-display coupled with proteolysis may be
useful for solving this difficult problem. The backbone
dynamics [26] and the 3D structure of the redesigned
apocyt b
562
determined by NMR clearly show that the
protein has a uniquely folded state.
Combinatorial computation versus
phage-display
Significant progress has been made using combinatorial
computation to design proteins [1,13,22,27,28]. The advant-
age of the computational methods is that they can examine
very large numbers of mutations [27]. The limitation of the
current computational methods, however, is that most of the
computer programs need to have the backbone conforma-
tions completely fixed in order to make the computation
efficient [22,27,29]. The fix of the backbone conformations
could potentially prevent selection of alternative attractive
structures that are slightly different in terms of backbone
conformation. Earlier work on the T4 lysozyme revealed
that over-packed core mutants typically responded by slight
alteration of the main chain, preserving near-ideal rotameric
side chain conformations [30]. Some efforts have been made
towards solving this problem. For example, backbone
freedom was considered in designing proteins by using

algebraic parameterization of the backbone for proteins with
simple motifs [1] and by manipulating the relative orienta-
tions of super secondary structural elements [31]. A more
general method has also been explored by Desjarlais &
Handel [32]. Another concern is that computational
methods generally lack the consideration of multi-body
Fig. 3. Effect of a partially unfolded intermediate on the selection result.
If the cutting site for protease is in the unfolded region of an interme-
diate state, selection of stable proteins will not be achieved because these
mutants will not change the free energy difference between the inter-
mediate (I) and the native (N) states. U represents the unfolded state.
1612 Y. Bai and H. Feng (Eur. J. Biochem. 271) Ó FEBS 2004
interactions. Therefore, long range effects of a mutation,
which have been shown to be important even in small
proteins [33,34], are not considered in the calculation. The
calculated stabilities of selected proteins are not correlated
with those measured in experiments, suggesting a lack of
intrinsic consistency and reliability of the computational
methods [35]. In comparison with the computational
method, the major limitation of the phage-display and
proteolysis method is that the size of the library is relatively
small, permitting simultaneous mutations only at about six
positions for each library. This limitation may be alleviated
to some extent if the complementary nature of the side chain
interactions is considered. An advantage of the phage-
display method is that the backbone of the protein does
not need to be strictly defined and long range effects of
mutations are included automatically, which could explore
the structures that would be missed using computational
methods.

Perspectives
The current experimental results of using phage-display and
proteolysis to select stable folded protein structures clearly
indicate that this method is a powerful tool for protein
design. Further perfection of the method should help to
provide insights into understanding the forces that stabilize
proteins and to designing proteins with new folds. A more
promising aspect is to combine the computational method
with phage-display. For example, the computational
approach can be used to identify potentially important
positions for mutation while phage-display and proteolysis
can be used for the final selection.
References
1. Harbury, P.B., Plecs, J.J., Tidor, B., Alber, T. & Kim, P.S. (1998)
High-resolution protein design with backbone freedom. Science
282, 1462–1467.
2. Schafmeister, C.E., LaPorte, S.L., Miercke, L.J. & Stroud, R.M.
(1997) A designed four helix bundle protein with native-like
structure. Nat. Struct. Biol. 4, 1039–1046.
3. Walsh, S.T., Cheng, H., Bryson, J.W., Roder, H. & DeGrado,
W.F. (1999) Solution structure and dynamics of a de novo designed
three-helix bundle protein. Proc.NatlAcad.Sci.USA96,
5486–5491.
4. DeGrado,W.F.,Summa,C.M.,Pavone,V.,Nastri,F.&Lom-
bardi, A. (1999) De novo design and structural characterization of
proteins and metalloproteins. Annu.Rev.Biochem.68, 779–819.
5. Ness, J.E., Del Cardayre, S.B., Minshull, J. & Stemmer, W.P.
(2001) Molecular Breeding: The natural approach to protein
design. Adv. Protein Chem. 55, 261–286.
6. Wintrode, P.L. & Arnold, F.H. (2001) Temperature adaptation of

enzymes: lessons from laboratory evolution. Adv. Protein Chem.
55, 161–226.
7. Kametkar, S., Schiffer, J.M., Xiong, H., Babik, J.M. & Hecht,
M.H. (1993) Protein design by binary patterning of polar and non-
polar amino acids. Science 262, 1680–1685.
8. Michnick, S.W. (2001) Exploring protein interactions by inter-
action-induced folding of proteins from complementary peptide
fragments. Curr.Opin.Struct.Biol.11, 472–477.
9. Sieber, V., Pluckthun, A. & Schmid, F.X. (1998) Selecting proteins
with improved stability by a phage-based method. Nat. Biotechnol.
16, 955–960.
10. Kristensen, P. & Winter, G. (1998) Proteolytic selection for protein
folding using filamentous bacteriophages. Fold. Des. 3, 321–328.
11. Finucane,M.D.,Tuna,M.,Lees,J.H.&Woolfson,D.N.(1999)
Core-directed protein design. I. An experimental method for
selecting stable proteins from combinatorial libraries. Bio-
chemistry 38, 11604–11612.
12. Finucane, M.D. & Woolfson, D.N. (1999) Core-directed protein
design. II. Rescue of a multiply mutated and destabilized variant
of ubiquitin. Biochemistry 38, 11613–11623.
13. Lazars, G.A., Desjarlais, J.R. & Handel, T.M. (1997) De novo
design of the hydrophobic core of ubiqutiin. Protein Sci. 6,
1167–1178.
14. Wernisch, L., Hery, S. & Wodak, S.J. (2000) Automatic protein
design with all atom force-fields by exact and heuristic optimiza-
tion. J. Mol. Biol. 301, 713–736.
15. Martin, A., Kather, I. & Schmid, F.X. (2002) Origins of the high
stability of an in vitro-selected cold-shock protein. J. Mol. Biol.
318, 1341–1349.
16.Pedersen,J.S.,Otzen,D.E.&Kristensen,P.(2002)Directed

evolution of barnase stability using proteolytic selection. J. Mol.
Biol. 323, 115–123.
17. Riechmann, L. & Winter, G. (2000) Novel folded protein domains
generated by combinatorial shuffling of polypeptide segments.
Proc.NatlAcad.Sci.USA97, 10068–10073.
18. Chu, R.A., Takei, J., Knowles, J.R., Andrykovitch, M., Pei, W.,
Kajava, A.V., Steinbach, P., Ji, X. & Bai, Y. (2002) Redesign of a
four-helix bundle protein by phage display coupled with proteo-
lysis and structural characterization by NMR and X-ray crystal-
lography. J. Mol. Biol. 323, 253–262.
19. Regan, L. & DeGrado, W.F. (1988) Characterization of a
helical protein designed from first principles. Science 241,
976–978.
20. Kamtekar, S., Schiffer, J.M., Xiong, H., Babik, J.M. & Hecht,
M.H. (1993) Protein design by binary patterning of polar and
nonpolar amino acids. Science 262, 1680–1685.
21. Betz, S.F., Bryson, J.W. & DeGrado, W.F. (1995) Native-like and
structurally characterized designed alpha-helical bundles. Curr.
Opin. Struct. Biol. 5, 457–463.
22. Nauli, S., Kuhlman, B. & Baker, D. (2001) Computer-based
redesign of a protein folding pathway. Nat. Struct. Biol. 8,
602–605.
23. Nauli, S., Kuhlman, B., Trong, I.L., Stenkamp, R.E., Teller, D. &
Baker, D. (2002) Crystal structures and increased stabilization of
the protein G variants with switched folding pathways NuG1 and
NuG2. Protein Sci. 11, 2924–2931.
24. Malakauskas, S.M. & Mayo, S.L., (1998) Design, structure and
stability of a hyperthermophilic protein variant. Nat. Struct. Biol.
5, 470–475.
25. Lumb, K.J. & Kim, P.S. (1995) A buried polar interaction imparts

structural uniqueness in a designed heterodimeric coiled coil.
Biochemistry 34, 8642–8648.
26. Takei, J., Pei, W., Vu, D. & Bai, Y. (2002) Populating partially
unfolded forms by hydrogen exchange-directed protein engineer-
ing. Biochemistry 41, 12308–12312.
27. Dahiyat, B.I. & Mayo, S.L. (1997) De novo protein design: fully
automated sequence selection. Science 278, 82–87.
28. Looger, L.L. & Hellinga, H.W. (2001) Generalized dead-end
elimination algorithms make large-scale protein side-chain struc-
ture prediction tractable: implications for protein design and
structural genomics. J. Mol. Biol. 307, 429–445.
29. Desjarlais, J.R. & Handel, T.M. (1995) De novo design of the
hydrophobic cores of protiens. Protein Sci. 4, 2006–2018.
30. Baldwin, E.P. & Matthews, B.W. (1994) Core-packing constraints,
hydrophobicity and protein design. Curr. Opin. Biotechnol. 5,
396–402.
Ó FEBS 2004 Selection of folded proteins by phage-display (Eur. J. Biochem. 271) 1613
31. Su, A. & Mayo, S.L. (1997) Coupling backbone flexibility and
amino acid sequence selection in protein design. Protein Sci. 6,
1701–1707.
32. Desjarlais, J.R. & Handel, T.M. (1999) Side-chain and backbone
flexibility in protein core design. J. Mol. Biol. 289, 305–318.
33. Lockless, S.W. & Ranganathan, R. (1999) Evolutionarily
conserved pathways of energetic connectivity in protein families.
Science 286, 295–299.
34. Russ, W.P. & Ranganathan, R. (2002) Knowledge-based poten-
tial functions in protein design. Curr. Opin. Struct. Biol. 12,
447–452.
35. Mendes, J., Guerois, R. & Serrano, L. (2002) Energy estimation in
protein design. Curr. Opin. Struct. Biol. 12, 441–446.

1614 Y. Bai and H. Feng (Eur. J. Biochem. 271) Ó FEBS 2004