Tài liệu Báo cáo khoa học: Phage-display as a tool for quantifying protein stability determinants pptx

MINIREVIEW
Phage-display as a tool for quantifying protein stability
determinants
Joanne D. Kotz
1
, Christopher J. Bond
2
and Andrea G. Cochran
1
1
Department of Protein Engineering and
2
Medicinal Chemistry, Genentech, Inc., South San Francisco, CA, USA
To address questions of protein stability, researchers have
increasingly turned to combinatorial approaches that permit
the rapid analysis of libraries of protein variants. Phage-
display has proved to be a powerful tool for analyzing protein
stability due to the large library size and the robustness of the
phage particle to a variety of denaturing conditions. With the
B1 domain of protein G (GB1) and a camelid heavy chain
antibody as model systems, we are using phage-display lib-
raries to experimentally address questions that have generally
been addressed in silico, either through computational stud-
ies or statistical analysis of known protein structures. One
effort has focused on identifying novel solutions to repacking
the hydrophobic core of GB1, while maintaining stability
comparable to the wild type protein. In a second study,
a small set of substitutions in complimentarity-determining
region 3 was found to stabilize the framework of the camelid
antibody. Another major focus has been to obtain quanti-
tative data on b-sheet stability determinants. We have suc-

cessfully adapted a phage-display method for quantitating
affinities of protein variants (shotgun alanine scanning) to
analysis of GB1 stability. Using this method, we have ana-
lyzed the energetic contributions of cross-strand side chain–
side chain interactions. Finally, we discuss parameters to
consider in using phage-display to discriminate subtle sta-
bility differences among fully folded variants. Overall, this
method provides a fast approach for quantitatively addres-
sing biophysical questions.
Keywords: beta sheet; hydrophobic core; phage-display;
protein G; protein stability.
Introduction
Understanding determinants of protein stability is critical
both for predicting the tertiary structure of a protein from
an amino acid sequence, as well as for protein design.
Rather than characterizing individual proteins with single
mutations, or defined combinations of mutations, research-
ers have increasingly been using selection and screening
methods to investigate protein stability. In comparison to
the labor-intensive process of generating and characterizing
individual mutant proteins, these combinatorial approaches
offer the important advantage of simultaneously generating
libraries of protein variants, thus allowing a much larger
number of mutations to be investigated. However, inter-
preting the results from combinatorial experiments is not as
straightforward as characterizing individual proteins. Con-
sequently, results must be carefully assessed in light of the
library design and selection pressure applied. Each screening
or selection method, a number of which are discussed in this
review series, will have inherent advantages and limitations

that should be considered in addressing specific questions
of protein structure.
Phage-display is one selection technique that has been
successfully applied to investigating protein stability [1,2]. In
adapting phage-display from the more common selection
for binding affinity, investigators have focused on mutating
residues affecting protein stability, but not directly involved
in ligand binding (Fig. 1). Proteins are selected that retain
binding capacity, with the implicit assumption that a
properly folded protein is required for an intact binding
interface [3,4].
As a protein mutagenesis strategy, phage-display offers
a number of important advantages. The technology for
generating large libraries ( 10
10
members) has been well
developed [5], permitting the simultaneous characterization
of a relatively large number of mutants. In addition, the
high in vitro stability of the phage particle [6] permits the
use of a wide range of selection conditions. For example,
investigators have used high temperature [7–9] and denat-
urants [8,9] to increase selective pressure. Varying the
stringency of selection conditions by these methods allows
greater flexibility in experimental design and is particularly
relevant to questions of protein stability.
One limitation of the above approach is the requirement
for a known binding partner with a binding interface that
is unaffected by the mutations introduced. A number of
researchers have developed strategies for circumventing this
coupling of protein stability and function, relying on the

greater susceptibility to proteolysis of unfolded proteins. An
Correspondence to A. G. Cochran, Department of Protein
Engineering, Genentech, Inc., 1 DNA Way, South San Francisco,
CA, 94080, USA. Fax: + 1 650 225 3734, Tel.: + 1 650 225 5943,
E-mail:
Abbreviations: CDR3, complimentarity-determining region 3; GB1,
B1 domain of protein G; scFv, single chain variable fragment;
V
H
, variable heavy chain.
(Received 5 January 2004, revised 18 February 2004,
accepted 5 March 2004)
Eur. J. Biochem. 271, 1623–1629 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2004.04076.x
accompanying review by Bai & Feng will discuss the
significant progress that has been made in developing these
methods [10].
In this review, we will focus on studies in our laboratory
investigating the effect of mutations on the stability of the
B1 domain of Streptococcal protein G (GB1). One aspect
of GB1 stability that we have addressed is the tolerance for
mutations in the core of the protein. These studies provide
a platform for comparing the results of combinatorial
experimental studies with a system that has been well
characterized computationally [11–14]. This work relies
upon the binding of properly folded GB1 to the immuno-
globulin Fc fragment to separate the few functional variants
from a large number of unfolded proteins. A similar
strategy is used by Bond et al. to characterize residues
necessary for protein stability in single chain camelid
antibodies [15]. By comparing the results of these studies

using different proteins but a similar experimental design,
we discuss the extent to which the most stable proteins are
selected.
Additionally, our laboratory has extended phage-based
stability selections to include quantification of the relative
contributions of amino acid substitutions to protein stabil-
ity. Using the b-sheet of GB1 as a model system, we have
asked whether a number of variants, all of which are folded
under the conditions of the selection, can be recovered
differentially based on varying stabilities. Remarkably, we
have found that, at least in some circumstances, a quanti-
tative correlation to biophysical data can be obtained from
a statistical analysis of selected phage populations [16]. We
also discuss experiments addressing the physical basis of this
selection and the range of stabilities that can be differen-
tiated.
Selecting the most stable protein variants
Two studies, one of GB1 and one of a camelid antibody,
have employed phage-display to identify the most stable
clones from a large pool of unfolded proteins. In both cases
stable clones are identified from the selection and stability is
confirmed when the individual proteins are characterized.
However, it is not clear whether the globally most stable
proteins encoded in the libraries are identified. The successes
and limitations of this strategy, and ideas for increasing the
selective pressure, are discussed below.
Repacking the GB1 core
We have begun to investigate by phage-display the tolerance
to substitution in the hydrophobic core of GB1. One goal of
this study was to compare phage-based strategies to

computational methods. Based on the definitions of Mayo
and coworkers, our library focused on core residues 5, 7 and
30, having less than 10% of the side-chain surface exposed
to solvent [11], and boundary residues 16, 18 and 33, those
which lie at the interface between the buried core and
surface residues [12]. These residues are spatially close to one
another and potentially in contact (Fig. 2). However, this
library does not exactly duplicate the computational library
because in the experimental system, residues near the Fc
binding interface cannot be varied (for instance, core
residues 3 and 39, and boundary residues 25 and 29 that
were all changed in a hyperstable GB1 variant [12]). Based
on the computational studies, core positions 5 and 30 were
expected to be intolerant to substitution. In contrast, the
core position 7 and boundary positions 16 and 18 have been
shown to tolerate substitutions, with some variants even
showing increased stability over the wild type protein
[11,12,14].
Following three rounds of selection at room temperature,
a consensus sequence began to emerge (Table 1). In
agreement with previous work, the wild type residues were
predominantly observed at positions 5 and 30, whereas
alternative residues appeared to be tolerated, or even
preferred, at all other library positions. For instance,
arginine and tryptophan were frequently observed at
positions 16 and 18, and tryptophan was preferred at
position 33. Following two additional rounds of selection,
a few particular sequences began to dominate the library.
Three individual GB1 variants were expressed and purified
for biophysical characterization; one of these differed only

at boundary residues and was based on the consensus
sequence from the third round of sorting, while two
represented the most dominant clones from the fifth round
of sorting. All three proteins underwent two-state thermal
Fig. 2. Mutation of the GB1 core. Core residues (red) and boundary
residues (blue) were randomized in one library. Other core residues [10]
are shown in gray. This figure and Fig. 5 were generated from the
NMR structure (PDB code 2GB1) [33] using
INSIGHT II
(Accelrys,
San Diego, CA).
Fig. 1. Phage-display as a method for selecting for protein stability.
Folded proteins are retained, based on the formation of the three-
dimensional structure necessary to form a functional binding interface.
Unfolded proteins cannot bind and therefore are not selected.
1624 J. D. Kotz et al.(Eur. J. Biochem. 271) Ó FEBS 2004
unfolding transitions. For the mutant based on the third
round consensus sequence, the melting temperature (T
m
)
was equal to that of wild type GB1 (81 °C). The T
m
was
somewhat reduced from wild type for the two dominant
fifth round clones (59 °Cand62°C; Table 1). In each case
the proteins were fully folded at room temperature and
retained IgG binding activity (J. D. Kotz & A. G. Cochran,
unpublished results).
To reduce the time required for computational analysis,
only hydrophobic residues were allowed at core positions

and only 16 residues were allowed at boundary positions
(with cysteine, methionine, glycine and proline excluded in
the in silico experiment) [11,12,14]. In contrast, in our
experimental system all 20 amino acids were encoded at each
position. Surprisingly, in the two clones that dominated the
library, amino acids disallowed in the computational study
were shown to result in stable, folded proteins. At the core
position 7, serine was observed in one of the frequently
observed clones. At boundary position 16, a proline
occurred in the second clone investigated (Table 1). These
results highlight the stability of the GB1 core to substitutions
that may seem energetically unlikely, or even irrational, but
that can be rapidly explored using a phage selection.
Design of a heavy chain antibody scaffold
A conceptually similar approach was employed by Bond
et al. in the design of a camelid heavy chain antibody
scaffold for use in constructing naı
¨
ve antibody libraries [15].
Here, the association of the variable heavy chain (V
H
)with
protein A was used as a surrogate for direct stability
measurements. The V
H
domains in camelid heavy chain
antibodies are most similar to the classical V
H
3 family and
as such bind protein A with micromolar affinity. Further-

more, the protein A binding site is distal to the former light
chain interface and involves residues within the b-sheet
structure (Fig. 3). As in the GB1-Fc system, the protein A–
antibody interaction requires a correctly folded molecule,
and therefore binding can be used as a direct readout for
antibody stability.
To adapt to the loss of the light chain, these heavy chain
scaffolds rely on portions of complimentarity-determining
region 3 (CDR3) to maintain structural integrity. This
additional role of CDR3 complicated the design of heavy
chain libraries for antigen binding selections by requiring a
scaffold in which the structural residues of CDR3 were
fixed. To identify these structurally important residues,
potential heavy chain CDR3 scaffolds were evaluated by
sorting a 17-residue CDR3 library against protein A.
Following three rounds of sorting, 335 clones were isolated
and sequenced. When purified proteins were individually
characterized, the four most frequently observed clones had
thermostabilities of approximately 60 °C, similar to those of
other camelid heavy chain antibodies [15]. A crystal
structure of one clone revealed that the residues selected at
both ends of the CDR3 loop are ordered and interact with
the former light chain interface, supporting the idea that
these residues are structurally important (Fig. 4). Con-
versely, the remainder of the loop is disordered, consistent
with the observed tolerance to substitution at these loop
positions (C. J. Bond, J. C. Marsters & S. S. Sidhu,
unpublished results; [15]).
To what extent are the most stable sequences selected?
In both of the above studies, the dominant clones

obtained after selection were shown to be stable and
well-folded proteins. However, in the GB1 study we failed
to identify new variants with significantly increased
stability compared to the starting protein. Furthermore,
the dominant round five clones were not as stable as the
wild type sequence and therefore were not the most stable
proteins encoded in the library. This observation high-
lights an important caveat when using phage-display. The
sequences selected at each round are influenced by a
variety of factors including codon usage, expression levels,
Fig. 3. Structure of the Protein A–Fab complex (PDB code 1DEE) [34].
The heavy chain is colored blue and the light chain green. Protein A
(orange) interacts with the heavy chain on the side opposite to the light
chain interface and is represented using the program
SWISSPDBVIEWER
.
Table 1. Experimental repacking of the GB1 core. A representative
sample of clones obtained after three rounds of selection for GB1
binding is shown at top. Proteins that were individually purified and
characterized are shown in bold.
Position
Round of
sorting
%of
clones T
m
(°C)
5 7 16 18 30 33
LLRRFW3 1
LLR WF W3 1

LFRWFF3 1
LI R WL W3 1
LSTLFW3 1
LLTWFF3 1
L L R W F W R3 consensus 81
LSI K F L 5 30 59
LYP V F M5 5 62
LLTTFYWT 81
Ó FEBS 2004 Phage-display for quantifying protein stability (Eur. J. Biochem. 271) 1625
export to the surface of the phage and affinity for the
target protein. Thus, although a folded protein is a
minimum requirement, phage-display selection may not
depend solely on protein stability.
To address this limitation, the selection can be made more
dependent on protein stability by destabilizing the host. For
example, to test the range of turn sequences permitted
in stable proteins [7], DeGrado and coworkers generated
random turn sequences in GB1 host proteins of different
stabilities; these libraries were screened for IgG competent
binders at either room temperature or elevated temperature.
They observed no sequence preference for turns in the wild
type host, whereas clear sequence preferences were observed
in destabilized hosts. As the stringency of the screen
increased, the functional solutions increasingly resembled
the wild type sequence and turns that are most commonly
found in proteins. Highlighting the effectiveness of the
screen, they confirmed biochemically that IgG binders
obtained under the most stringent conditions were signifi-
cantly more stable than nonbinders [7]. In a different
approach, Plu

¨
ckthun and coworkers directly compared the
use of temperature or denaturant as a selective pressure to
improve the stability of single chain variable fragment
(scFv) by phage-display [8]. In their study, the phage-
displaying scFv variants were subjected to high temperature,
denaturant or no stress prior to selection. High-temperature
stress resulted in the most sequence convergence and the
most stable clone analyzed. The authors concluded from
this observation that, at least in their system, temperature
stress was much more stringent than incubation with
denaturant. However, this may not be general; instead, it
may be a consequence of the irreversibility of thermal
denaturation and the reversibility of chemical denaturation
for scFV [8]. In the case of GB1, with known hyperstable
variants, we intend to compare the results of increasing the
selective pressure through each of these methods. Hopefully,
as additional systems are investigated, a general under-
standing of the pressure needed for a given stringency of
selection will emerge.
Quantifying protein stability
A major focus of our laboratory has been extending the use
of phage-display to allow ranking of the stabilities of
individual proteins in a pool of folded variants. In addition,
a rapid method for quantitatively, and simultaneously,
characterizing a large number of mutants would greatly aid
in understanding the effects of complex interactions on
stability, for instance, the effect of long range tertiary
contacts on b-sheet formation. We discuss recent advances
in the use of phage-display to probe the energetics of b-sheet

formation, as well as progress in understanding important
experimental variables of the method.
Analyzing stability determinants in the b-sheet of GB1
The potential for obtaining quantitative biophysical infor-
mation from phage-display was suggested by a new method
called alanine shotgun scanning, which analyzes the ener-
getic contribution of residues at a binding interface. Sidhu,
Weiss and coworkers [18–20] have treated the observed
frequency ratios of residues at a given position (i.e. wild
type/alanine ratio) as an equilibrium constant, which is then
used to calculate the relative free energies of binding for
different protein variants. Relative energies calculated from
this distribution data have been shown to correspond
directly with data obtained for individual purified mutants
[18]. Thus, shotgun scanning provides a rapid method for
using phage-display to quantify changes in affinity.
In order to apply this method to the ranking of protein
stabilities, the well-established b-sheet model system GB1
seemed an ideal initial target [13]. A library was constructed
varying two cross-strand residues, 44 and 53 (Fig. 5). These
positions were guest sites in published mutagenesis studies
from which b-sheet propensity scales have been developed
[21–24]. In both protein mutagenesis and phage studies, the
host protein included the I6A mutation, resulting in the
destabilization of the protein by  2.5 kcalÆmol
)1
relative to
wild type GB1. Following a binding selection at room
temperature, individual clones were sequenced. From the
observed residue distributions at position 53, a phage-based

stability scale was calculated. Strikingly, this scale correlated
quantitatively with the published propensity scale derived
from thermal stability measurements [16].
Fig. 4. Structure of the evolved heavy chain scaffold [17]. The frame-
work regions are colored grey while CDR3 is colored red. Residues
critical to scaffold stablility, both at the former light chain interface and
inCDR3,areshowninyellow.Theimagewasgeneratedusing
SWISSPDBVIEWER
.
1626 J. D. Kotz et al.(Eur. J. Biochem. 271) Ó FEBS 2004
The use of an unusually trivial library, in which just two
surface positions were varied to all amino acids, permitted
the rapid analysis of the energetic contribution of side
chain–side chain interactions at a single surface site. This
analysis could be compared to earlier studies of residue
pairing in diverse b-sheets from the Protein Data Bank
[25–27]. These earlier studies indicated many statistically
significant deviations from random pairing, the nature of
which depended to some extent on the exact method of data
normalization. In contrast, the data obtained by the phage-
display method (that do not require normalization) sugges-
ted only minor energetic contributions from most side chain
interactions [16]. This rapid analysis of many amino acid
pairs demonstrates the power of using combinatorial
approaches to address questions of protein stability, where
a large number of interactions must be characterized to
understand the underlying trends.
Current work in our laboratory is directed toward
extending the studies of b-propensities and side chain–side
chain interactions in b-sheets. We are now creating a library

at positions 6 and 15 of the GB1 b-sheet. These two side
chains form a nonhydrogen bonded pair, unlike residues 44
and 53 whose backbone amides are hydrogen bonded to one
another. Nonhydrogen bonded pairs have different C
a
and
C
b
distances than hydrogen bonded pairs and therefore may
have different residue preferences and potential for side
chain–side chain interactions. Initial results suggest that the
selected residue distributions correlate well with a conven-
tional stability scale for position 6, and we are in the process
of analyzing the relationship between the two positions
(J. D. Kotz & A. G. Cochran, unpublished results).
Importantly, these results suggest that the phage-display
method will be generally applicable to quantitative compar-
isons of protein stabilities.
Importance of host stability for selection
The investigation of turn stability from DeGrado and
coworkers (discussed above) emphasized the importance of
host stability in tuning the stringency of a selection or screen
[7]. In our effort to rank protein variants of very similar
stabilities, we will probably need to achieve a balance
between a very stable host, whose folded population would
be predicted to be insensitive to stability changes, and a very
unstable host that would not allow characterization of
destabilizing mutations. In our current studies of positions 6
and 15 of the GB1 b-sheet, we have used hosts of two
different stabilities (wild type and a mutant destabilized by

 2kcalÆmol
)1
). By comparing results from the two host
proteins, we intend to characterize the host stability range
necessary for obtaining quantitative data. Surprisingly, we
have found that variants that should be fully folded at the
temperature of selection can nevertheless be discriminated,
raising questions about the basis for selection [16].
Physical basis for ordering stabilities
As described above, a binding selection relies on the
requirement that a protein be folded in order to functionally
interact with a binding partner. This is a powerful method
for the isolation of rare folded variants from a larger pool of
unfolded molecules [4]. However, in distinguishing a more
stable protein from many other stable proteins, the physical
basis for the selection is not as clear. The most straightfor-
ward idea is that each protein variant is in equilibrium
between the folded and unfolded state with only the folded
state being competent for binding [2]. That is, folding is
thermodynamically coupled to binding, resulting in an
apparent affinity change as the fraction of folded molecules
changes [28]. A second possibility is that many proteins in
the pool are fully folded but that enhanced protein stability
leads somehow to higher target affinity, increasing the
likelihood of recovery in the binding selection. Finally, more
stable proteins may be displayed at a higher level on the
surface of the phage, but then once displayed, retained with
equal probability during the interaction with the binding
partner. To distinguish changes in display level from
selections requiring interaction with the specific binding

partner (stability- or affinity-based), one could divide a
phage library in half and sort one half against a binding
partner and the other half against an expression tag.
A comparison of the sequences obtained from these two
different selections should reveal any display bias. Alter-
natively, a Western blot could be used to directly probe the
display levels of selected protein variants. To distinguish
affinity-based selections from those based solely on protein
stability, the affinities of purified mutant proteins for target
can be measured by standard methods (immunoassay or
surface plasmon resonance) at the temperature of the phage
selection (or at temperatures at which the variants are fully
folded); in studies with GB1, it does not appear that
sufficient affinity differences exist among the folded
variants to explain their differential selection (J. D. Kotz
& A. G. Cochran, unpublished results) [16,24].
Analyzing the less stable GB1 variants
Another potential limitation in quantifying stability by
phage-display results from the inherently larger number of
sequences observed for the more stable clones. For example,
in the analysis of side chain interactions at positions 44 and
53 in GB1, the hydrophobic amino acids are more
stabilizing, and thus they occur much more frequently than
Fig. 5. Quantifying b-sheet stability. The hydrogen-bonded pair (red)
and the nonhydrogen-bonded pair (blue) were varied in separate
phage-display libraries. Surrounding residues from the solvent exposed
face of the b-sheet are shown in gray.
Ó FEBS 2004 Phage-display for quantifying protein stability (Eur. J. Biochem. 271) 1627
other amino acids. Therefore, even with  1200 sequences,
many of the 400 possible pairwise combinations are not

expected to occur very frequently (if at all). As a result, even
though charged side chain–side chain interactions are
thought to be energetically significant [29,30] (and we did
observe a number of oppositely charged pairs), charged
amino acids did not occur frequently enough in our GB1
data set for observed pair correlations to be statistically
significant. One could imagine addressing this issue by
sequencing 10- to 100-fold more clones. However, it should
be possible instead to increase representation of these amino
acids by tailoring the initial library (by a different choice of
degenerate codon), thereby eliminating the dominant, more
stabilizing residues.
Conclusions and future prospects
It has long been appreciated that phage libraries are a rich
source of unexpected functional diversity. Because of the
very large library sizes that can be achieved, it is tempting to
maximize the number of positions varied and then carry out
many rounds of selection, in order to increase the chance of
identifying a rare, highly functional clone. Although it is
often possible to identify exciting new molecules, this
approach introduces a Ôblack boxÕ aspect to the use of
phage technology. In contrast, large scale screens can be
very useful when asking certain quantitative questions, for
example, what fraction of library members exhibits a
property of interest? Unfortunately, such screens generally
provide only crude measures of the degree to which a
variant exhibits the property. Thus, it is difficult to use
screens to answer many questions about proteins that are
traditionally addressed by conventional site-directed muta-
genesis and assays of purified variants. This is particularly

relevant to the study of protein stability, in which detailed
thermodynamic comparisons are often required. One solu-
tion is to design screens that yield a reliable, quantifiable
output parameter that reports on the stability of library
members [31]. We have chosen instead to modify our use of
phage selection technology.
The discovery that phage-display can be used quantita-
tively to report on affinity is an exciting new development
[18,20]. It appears that once nonfunctional background
phage are eliminated from a library through a few rounds
of selection, the remaining phage essentially represent an
equilibrium population of more and less functional vari-
ants, allowing the use of statistical thermodynamics to rank
free energy differences. Because DNA sequencing is now
relatively routine, it is straightforward to sample the
selected phage population sufficiently to identify residues
important for binding and to provide a reliable estimate of
how important they are. For binary mutagenesis (e.g. wild
type vs. alanine), only a few hundred sequences are needed
to characterize most interfaces [18], and an experiment of
this type can be conducted inexpensively in a couple of
weeks.
The extension of ÔshotgunÕ phage-display to selections for
folding should expand the utility of combinatorial muta-
genesis and complement existing methodology [32]. A major
advantage to the shotgun method is that it combines
strengths of traditional phage selections (e.g. amplification
and discrimination of functional variants) with those of high
throughput screens (e.g. adequate statistical sampling of
smaller libraries). However, there are complications encoun-

tered in studying folding by phage, namely the need for a
selection that indirectly, yet cleanly, reports on stability.
Furthermore, as discussed above, there are certain param-
eters (such as the stability of the starting host protein) that
may require optimization in order to achieve good results. It
is our belief that these concerns can be addressed and that
phage-display will prove a useful, and fast, way to test
quantitatively many hypotheses about protein structure.
References
1. Hoess, R.H. (2001) Protein design and phage display. Chem. Rev.
101, 3205–3218.
2. Forrer,P.,Jung,S.&Plu
¨
ckthun, A. (1999) Beyond binding: Using
phage display to select for structure, folding and enzymatic activity
in proteins. Curr.Opin.Struct.Biol.9, 514–520.
3. O’Neil, K.T., Hoess, R.H., Raleigh, D.P. & DeGrado, W.F.
(1995) Thermodynamic genetics of the folding of the B1
immunoglobulin-binding domain from Streptococcal protein G.
Proteins 21, 11–21.
4. Gu, H., Yi, Q., Bray, S.T., Riddle, D.S., Shiau, A.K. & Baker, D.
(1995) A phage display system for studying the sequence
determinants of protein folding. Protein Sci. 4, 1108–1117.
5. Sidhu, S.S., Lowman, H.B. & Wells, J.A. (2000) Phage display for
selection of novel binding peptides. Methods Enzymol. 328,
333–363.
6. Sieber, V., Plu
¨
ckthun, A. & Schmid, F.X. (1998) Selecting proteins
with improved stability by phage-based methods. Nat. Biotechnol.

16, 955–960.
7. Zhou, H.X., Hoess, R.H. & DeGrado, W.F. (1996) In vitro
evolution of thermodynamically stable turns. Nat. Struct. Biol. 3,
446–451.
8. Jung, S., Honegger, A. & Plu
¨
ckthun, A. (1999) Selection for
improved protein stability by phage display. J. Mol. Biol. 294,
163–180.
9. Martin, A., Sieber, V. & Schmid, F.X. (2001) In-vitro selection of
highly stabilized protein variants with optimized surface. J. Mol.
Biol. 309, 717–726.
10. Bai, Y. & Feng, H. (2004) Selection of stably folded proteins by
phage-display with proteolysis. Eur. J. Biochem. 271, 1609–1614.
11. Dahiyat, B.I. & Mayo, S.L. (1997) Probing the role of packing
specificity in protein design. Proc. Natl Acad. Sci. USA 94,
10172–10177.
12. Malakauskas, S.M. & Mayo, S.L. (1998) Design, structure and
stability of a hyperthermophilic protein variant. Nat. Struct. Biol.
5, 470–475.
13. Ross,S.A.,Sarisky,C.A.,Su,A.&Mayo,S.L.(2001)Designed
protein G core variants fold to native-like structures: sequence
selection by ORBIT tolerates variation in backbone specification.
Protein Sci. 10, 450–454.
14. Su, A. & Mayo, S.L. (1997) Coupling backbone flexibility and
amino acid sequence selection in protein design. Protein Sci. 6,
1701–1707.
15. Bond, C.J., Marsters, J.C. & Sidhu, S.S. (2003) Contributions
of CDR3 to V
H

H domain stability and the design of mono-
body scaffolds for naive antibody libraries. J. Mol. Biol. 332,
643–655.
16. Distefano, M.D., Zhong, A. & Cochran, A.G. (2002) Quantifying
b-sheet stability by phage display. J. Mol. Biol. 322, 179–188.
17. Dahiyat, B.I. & Mayo, S.L. (1996) Protein design automation.
Protein Sci. 5, 895–903.
18. Weiss, G.A., Watanabe, C.K., Zhong, A., Goddard, A. &Sidhu, S.S.
(2000) Rapid mapping of protein functional epitopes by combi-
1628 J. D. Kotz et al.(Eur. J. Biochem. 271) Ó FEBS 2004
natorial alanine scanning. Proc. Natl Acad. Sci. USA 97,
8950–8954.
19. Vajdos, F.F., Adams, C.W., Breece, T.N., Presta, L.G.,
de Vos, A.M. & Sidhu, S.S. (2002) Comprehensive functional
maps of the antigen-binding site of an anti-ErbB2 antibody
obtained with shotgun scanning mutagenesis. J. Mol. Biol. 320,
415–428.
20. Morrison, K.L. & Weiss, G.A. (2001) Combinatorial alanine-
scanning. Curr. Opin. Chem. Biol. 5, 302–307.
21. Smith, C.K. & Regan, L. (1995) Guidelines for protein design:
The energetics of b-sheet side chain interactions. Science 270,
980–982.
22. Smith, C.K., Withka, J.M. & Regan, L. (1994) A thermodynamic
scale for the b-sheet forming tendencies of the amino acids.
Biochemistry 33, 5510–5517.
23. Minor, D.L. Jr & Kim, P.S. (1994) Context is a major determinant
of b-sheet propensity. Nature 371, 264–267.
24. Minor, D.L. Jr & Kim, P.S. (1994) Measurement of the b-sheet-
forming propensities of amino acids. Nature 367, 660–663.
25. Lifson, S. & Sander, C. (1980) Specific recognition in the tertiary

structure of b-sheets of proteins. J. Mol. Biol. 139, 627–639.
26. Wouters, M.A. & Curmi, P.M.G. (1995) An analysis of side-chain
interactions and pair correlations within antiparallel b-sheets: The
differences between backbone hydrogen-bonded and non-hydro-
gen bonded pairs. Proteins: Struct. Funct. Genet. 22, 119–131.
27. Hutchinson, E.G., Sessions, R.B., Thornton, J.M. & Woolfson,
D.N. (1998) Determinants of strand register in antiparallel
b-sheets of proteins. Protein Sci. 7, 2287–2300.
28. Ruan, B., Hoskins, J., Wang, L. & Bryan, P.N. (1998) Stabilizing
the subtilisin BPN¢ pro-domain by phage display selection: How
restrictive is the amino acid code for maximum protein stability?
Protein Sci. 7, 2345–2353.
29. Merkel, J.S., Sturtevant, J.M. & Regan, L. (1999) Sidechain
interactions in parallel b-sheets: The energetics of cross-strand
pairings. Structure 7, 1333–1343.
30. Lassila, K.S., Datta, D. & Mayo, S.L. (2002) Evaluation of the
energetic contribution of an ionic network to beta-sheet stability.
Protein Sci. 11, 688–690.
31. Lahr, S.J., Broadwater, A., Carter, C.W. Jr, Collier, M.L.,
Hensley, L., Waldner, J.C., Pielak, G.J. & Edgell, M.H. (1999)
Patterned library analysis: a method for the quantitative assess-
ment of hypotheses concerning the determinants of protein
structure. Proc.NatlAcad.Sci.USA96, 14860–14865.
32. Sauer, R.T. (1996) Protein folding from a combinatorial per-
spective. Fold. Des. 1, R27–R30.
33. Gronenborn, A.M., Filpula, D.R., Essig, N.Z., Achari, A.,
Whitlow, M., Wingfield, P.T. & Clore, G.M. (1991) A novel,
highly stable fold of the immunoglobulin binding domain of
Streptococcal protein G. Science 253, 657–661.
34. Graille, M., Stura, E.A., Corper, A.L., Sutton, B.J., Taussig, M.J.,

Charbonnier, J B. & Silverman, G.J. (2000) Crystal structure of a
Staphylococcus aureus protein A domain complexed with the Fab
fragment of a human IgM antibody: Structural basis for
recognition of B-cell receptors and superantigen activity. Proc.
NatlAcad.Sci.USA97, 5399–5404.
Ó FEBS 2004 Phage-display for quantifying protein stability (Eur. J. Biochem. 271) 1629