Design strategies for the sequence-based mimicry of side-chain display in protein β-sheets by α/β-peptides

Article abstract of DOI:10.1021/ja306311r

The sophistication of folding patterns and functions displayed by unnatural-backbone oligomers has increased tremendously in recent years. Design strategies for the mimicry of tertiary structures seem within reach; however, a general method for the mimicry of sheet segments in the context of a folded protein is an unmet need preventing realization of this goal. Previous work has shown that 1→1 α→β-residue substitutions at cross-strand positions in a hairpin-forming α-peptide sequence can generate an α/β-peptide analogue that folds in aqueous conditions but with a change in side-chain display relative to the natural sequence; this change would prevent application of single β-residue substitutions in a larger protein. Here, we evaluate four different substitution strategies based on replacement of αα dipeptide segments for the ability to retain both sheet folding encoded by a parent α-peptide sequence as well as nativelike side-chain display in the vicinity of the β-residue insertion point. High-resolution structure determination and thermodynamic analysis of folding by multidimensional NMR suggest that three of the four designs examined are applicable to larger proteins.

Full text of DOI:10.1021/ja306311r

Article
pubs.acs.org/JACS
Design Strategies for the Sequence-Based Mimicry of Side-Chain
Display in Protein β‑Sheets by α/β-Peptides
George A. Lengyel and W. Seth Horne*
Department of Chemistry, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, United States
S
* Supporting Information
ABSTRACT: The sophistication of folding patterns and
functions displayed by unnatural-backbone oligomers has
increased tremendously in recent years. Design strategies for
the mimicry of tertiary structures seem within reach; however,
a general method for the mimicry of sheet segments in the
context of a folded protein is an unmet need preventing
realization of this goal. Previous work has shown that 1→1 α→
β-residue substitutions at cross-strand positions in a hairpin-
forming α-peptide sequence can generate an α/β-peptide
analogue that folds in aqueous conditions but with a change in
side-chain display relative to the natural sequence; this change would prevent application of single β-residue substitutions in a
larger protein. Here, we evaluate four different substitution strategies based on replacement of αα dipeptide segments for the
ability to retain both sheet folding encoded by a parent α-peptide sequence as well as nativelike side-chain display in the vicinity
of the β-residue insertion point. High-resolution structure determination and thermodynamic analysis of folding by
multidimensional NMR suggest that three of the four designs examined are applicable to larger proteins.
apoptosis.6 More recently, sequence-based design has begun
INTRODUCTION
■
to expand beyond helix mimicry to other structural contexts,
Unnatural-backbone oligomers that fold like proteins, often
such as sheet7 and a fold with less regular secondary structure.8
termed “foldamers,”1 offer the exciting possibility of manifesting
The development of unnatural oligomers that mimic the
complex tertiary folds of proteins is a frontier scientific
challenge. We believe that sequence-based modification of
proteinlike biological functions on protease-resistant scaffolds.2
Early work on the use of these molecules to mimic natural
proteins relied heavily on structure-based approaches.2f Such
natural proteins can provide a general and readily implemented
methods depend on the synthesis and characterization of
strategy to accomplish this goal. Success in such an approach
diverse, de novo designed sequences; as some understanding of
the fundamental relationships between sequence and folding in
requires rules for the modification of all the secondary structure
elements commonly encountered in proteins: helices, tight
a given unnatural backbone emerges, rational design can be
turns, loops, and sheets. Design principles for helix mimicry by
used to create an entity with both a desirable biological
function and a protease-resistant backbone. Recent years have
seen the emergence of an alternate approach for the design of
functional foldamers in which a natural protein sequence serves
α→β residue substitution are now well-established,4−6 and
other methods for helix replacement have shown promise.9 A
number of turn replacements have been reported to be effective
in folded proteins,10 and loops in a tertiary structure can be
as the starting point for design of its own mimic. Termed
tolerant to significant modification.11 Building on pioneering
“sequence-based design,” the latter method is best exemplified
in its application to prepare α/β-peptides3 (oligomers
early work on sheet folding behavior of β-residues in organic
solvent,12 we have recently begun to examine the applicability
composed of mixtures of natural α-amino acid residues and
of sequence-based α→β backbone modification to the mimicry
unnatural β-amino acid residues) that fold and function like
of protein sheets in water (Figure 1).7
proteins.
In previous work, we compared the sheet-folding propen-
The basic idea underlying sequence-based design is that
sities of 16 different β-residues when inserted into a short α-
natural proteins will manifest their inherent folding behavior
when a native sequence of side chains is displayed on an
unnatural backbone, provided the backbone modification is
made judiciously. The first applications of sequence-based
peptide hairpin model system; the β-residues had systematically
varied substitution pattern, stereochemistry, and ring con-
straint.7 Of the monomers examined, only three led to hairpin
folding in aqueous buffer when introduced as single α→β
design focused on α-helical prototype sequences and led to
replacement at cross-strand positions the prototype α-peptide.
significant advances in the development of oligomers with
complex folding patterns and functions, including formation of
helix-bundle quaternary structures4 and high-affinity binding to
Received: June 28, 2012
Published: September 4, 2012
biological receptors involved in HIV-cell fusion5 and
© 2012 American Chemical Society
15906
dx.doi.org/10.1021/ja306311r | J. Am. Chem. Soc. 2012, 134, 15906−15913

Journal of the American Chemical Society
Article
were purified, lyophilized, dissolved in 10 mM phosphate buffer (pH
8.9, 5% v/v DMSO), stirred until analytical high-performance liquid
chromatography (HPLC) and mass spectrometry (MS) showed
complete conversion to the cyclic disulfide (1−2 d), and then
repurified. Purification was accomplished by preparative HPLC on a
C18 column using gradients between 0.1% TFA in water and 0.1% TFA
in acetonitrile. All peptides were >95% pure as determined by
analytical HPLC on a C18 column. Identities were confirmed by mass
spectrometry using a Voyager DE Pro matrix-assisted laser desorption
ionization time-of-flight (MALDI-TOF) instrument (Supporting
Information, Table S1).
NMR Sample Preparation and Data Collection. NMR samples
were prepared by dissolving peptide in 750−850 μL of degassed 50
mM phosphate, 9:1 H2O/D2O, pH 6.3 (uncorrected for the presence
of D2O) to a final concentration of 0.2−1 mM. 3-(Trimethylsilyl)-1-
propanesulfonic acid sodium salt (DSS, 50 mM in water) was added to
a final concentration of 0.2 mM. Each solution was passed through a
0.2 μm syringe filter, transferred to an NMR tube, and stored until
analysis. The NMR tube headspace was purged with a stream of
nitrogen prior to capping.
Figure 1. Change in side-chain display resulting from 1→1 α→β
residue substitution in each strand of a β-hairpin. Residues i and i + 2
(gray boxes and spheres) are on the same face in the natural backbone
(A) but on opposite faces in the unnatural backbone (B). Coordinates
are from published NMR solution structures of an α-peptide hairpin
and an analogue resulting from two α→β residue substitutions7.
NMR spectra of peptides were recorded on a Bruker Avance-700
spectrometer. Chemical shifts are reported relative to DSS (0 ppm).
Total correlation spectroscopy (TOCSY), nuclear Overhauser
enhancement spectroscopy (NOESY), and correlation spectroscopy
(COSY) pulse programs used excitation-sculpted gradient-pulse
solvent suppression. All experiments were obtained using 2048 data
points in the direct dimension and 512 data points in the indirect
dimension. TOCSY spectra were acquired with a mixing time of 80 ms
and NOESY spectra were acquired with a mixing time of 200 ms.
NMR measurements were performed at a temperature of 293 K unless
otherwise noted. We measured linear hairpin peptides with unnatural
backbones (6a−9a) at 278 K to maximize folded population and
facilitate comparison of folded stability. Natural backbone peptide 5a
was measured at both 278 and 293 K. NMR data at 293 K were used
for comparison among the α-peptide series (1a−5a), while data at 278
K were used for comparison with the unnatural backbone series (5a−
9a).
NMR Data Analysis and Structure Determination. The Sparky
software package (T. D. Goddard and D. G. Kneller, SPARKY 3,
University of California, San Francisco) was used to analyze two-
dimensional (2D) NMR data. Backbone chemical shift assignments
were generated (Supporting Information, Tables S2−S9), and each
cyclic peptide was analyzed for qualitative nuclear Overhauser effects
(NOEs) indicative of folding. Peptides that showed a high degree of
folding were fully assigned (Supporting Information, Tables S10−S13)
and inter-residue NOEs were tabulated. NOE integration values were
converted to distance restraints using eq 1:
Unfortunately, the 1→1 α→β residue substitution was found
unsuitable for modification of a larger protein sheet, due to a
fundamental change in side-chain display resulting from the
backbone modification (Figure 1).
The hydrogen-bonding pattern of a β-residue in an extended
strand typically differs from that of an α-residue in the same
context.7,12 As a result, the side chains of α-residues flanking the
point of α→β substitution end up on opposite faces of the
strand rather than residing on the same face as they do in the
natural backbone (Figure 1). None of the 16 monomers
examined in our prior study were able to suppress this change
in side-chain display resulting from single α→β replacement.
Although tolerated in a small hairpin, this change would abolish
folding in the context of a larger tertiary structure as it would
fundamentally alter the side-chain composition of the hydro-
phobic core of the protein. We therefore sought to devise and
test alternate β-residue substitution strategies that would retain
both sheet folding of a parent α-peptide sequence as well as
nativelike side-chain display in the vicinity of the insertion point
of the unnatural residue. We report here several designs that
accomplish both these goals.
MATERIALS AND METHODS
■
Monomer Synthesis. For full experimental details and character-
ization data, see the Supporting Information. β3-Monomers were
purchased as the free amino acid and Fmoc protected. β2-Amino acids
were synthesized using a route based on a diastereoselective Mannich
reaction as the key step.13 β2,3-Amino acid monomers were prepared
by either ring-opening of enantiomerically enriched thiazinones14 or
diastereoselective conjugate addition to an α,β-unsaturated ester.15
Peptide Synthesis. Peptides were synthesized using microwave-
assisted Fmoc solid-phase synthesis techniques on a MARS microwave
reactor (CEM). NovaPEG Rink amide resin or H-Glu(tBu) HMPB
NovaPEG resin was used as the solid support. Couplings were carried
out in NMP with a 2 min ramp to 70 °C and a 4 min hold at that
temperature, using 4 equiv of Fmoc-protected amino acid, 4 equiv of
HCTU, and 6 equiv of DIEA. Deprotections were performed with a 2
min ramp to 80 °C followed by a 2 min hold at that temperature, using
an excess of 20% 4-methylpiperidine in DMF. After each coupling or
deprotection cycle, the resin was washed three times with DMF.
Double coupling was performed at sequence positions following
proline or β-residues. N-terminal acetylation, when present, was
carried out on resin by treatment with 8:2:1 v/v/v DMF/DIEA/Ac2O.
Prior to cleavage from the resin, peptides were washed three times
each with DMF, dichloromethane, and methanol, and then dried.
Peptide cleavage was performed using 95% trifluoroacetic acid (TFA),
2.5% triisopropylsilane, and 2.5% water. Cysteine-containing peptides
I = cr6
(1)
where I is NOE intensity, c is a constant (determined using resolved
diastereotopic CH2 groups from Tyr, Gly, and/or Phe), and r is
distance.16 The distances were then sorted and classified as strong
(≤2.7 Å), medium (≤3.5 Å), weak (≤4.5 Å), or very weak (≤5.5 Å) to
generate distance restraints used for structure determination
(Supporting Information, Tables S14−S17).
The Crystallography and NMR system (CNS) software package
was used to generate three-dimensional (3D) structures.17 Patches
were written for β-residues and accompanying linkages. Distance
restraints calculated as above were used in 100 simulated annealing
runs using default CNS parameters for protein NMR. Structures
including NOE distance-restraint violations >0.5 Å were discarded,
and the 10 lowest energy structures were obtained. The minimum
energy average of these 10 structures was inspected to identify H-
bonding contacts. These contacts were then included in an additional
restraint file, and the annealing process was repeated over 500 runs to
generate an ensemble of 10 low-energy structures and a minimized
average structure for each peptide. In the case of peptides where the
NMR ensemble had more than one distinct family of structures,
minimized average coordinates for each family were calculated.
15907
dx.doi.org/10.1021/ja306311r | J. Am. Chem. Soc. 2012, 134, 15906−15913

Journal of the American Chemical Society
Article
Calculation of Folding Equilbria by NMR. Fraction folded from
chemical shift deviation (f Hα) was calculated using experimentally
determined Hα chemical shifts (δHα) in eq 2:
δHα,observed − δHα,unfolded
δHα,folded − δHα,unfolded
fHα
=
(2)
where δHα,observed is the chemical shift of a particular Hα in the
unknown peptide (e.g., 1a), δHα,unfolded its chemical shift in an N- or C-
terminal fragment (e.g., 1c or 1d), and δHα,folded its chemical shift in a
disulfide-bridged cyclic analogue (e.g., 1b). Reported values of f are
averages calculated using chemical shift data for residues 4, 11, and 13.
Fraction folded from separation of diasterotopic Gly Hα’s (f Gly) was
calculated using eq 3:
ΔδHα/Hα′,observed
ΔδHα/Hα′,folded
fGly
=
(3)
where ΔδHα/Hα′,observed is the chemical shift difference between Gly
Hα’s in an unknown peptide (e.g., 1a) and ΔδHα/Hα′,folded the
corresponding difference in a disulfide-bridged cyclic analogue (e.g.,
1b).
The equilibrium constant for the folding equilibrium (Kfold) and
corresponding free energy of folding (ΔG°fold) were calculated from
fraction folded using eqs 4 and 5:
Figure 2. Four substitution strategies for the replacement of αα
dipeptide segments in a β-strand. Designs I and II involve 2→1
substitution (αα→β), while III and IV involve 2→2 substitution
(αα→ββ). Designs II and IV employ conformationally constrained
syn-β2,3-monomers, while I and III use more flexible β3- and/or β2-
residues.
f
Kfold
=
1 − f
(4)
(5)
ΔGfold = −RTln(Kfold
)
incorporating flexible β3- or β2-residues versus more con-
strained syn-β2,3-monomers.
Experimental uncertainty for a folded population determined by Hα
chemical shift deviation (f Hα) was estimated using the standard
deviation of the mean for populations based on residues 4, 11, and 13.
In design I, an αα dipeptide is replaced by a single β-residue
(β2 or β3), resulting in the deletion of two atoms from the
protein backbone (amide N and carbonyl C). One side chain is
also omitted, and the retained side chain (i.e., Xxx1 or Xxx2) at a
given site is determined by which residue from the replaced
dipeptide is more intimately tied to folding. In design II, two α-
residues are condensed into a single conformationally
restrained syn-β2,3-monomer, and both parent side chains are
retained. In designs III and IV, an αα dipeptide is replaced by a
two-residue ββ segment, resulting in the insertion of two
additional backbone C atoms. The positioning of side chains is
selected to retain the up/back pattern along the strand. Design
III accomplishes this through a β3β2-segment, while design IV
makes use of β2,3-residues with extra structure-promoting
methyl groups on spacer C atoms. In all cases, the absolute
stereochemistry of the β-residue(s) employed was selected to
best match the L configuration of the replaced αα dipeptide.
Model System Selection and Design of α/β-Peptide
Variants. With hypotheses about four possible strategies for
sheet α→β backbone modification in hand, we next looked for
a model system18 in which to systematically compare designs
I−IV. We had several requirements for properties of an ideal
model sequence. First, we sought an oligomer small enough to
be amenable to analysis by 1H-detected multidimensional NMR
of samples lacking isotopic labels. Second, we needed a sheet-
forming sequence with appropriate folded stability. Our
previous work provided some qualitative insights into relative
folding propensities of different types of β-residues incorpo-
rated into an α-peptide hairpin with an unnatural turn-
promoting D-Pro;7 however, rigorous thermodynamic analysis
of folding free energies was precluded there by lack of proper
controls to quantify folded population. We sought a system in
the present study that would provide a folded hairpin
population of 60−70% in the prototype sequence with ready
access to related control sequences for 0% and 100% folded
Error for a folded population determined by Gly Hα separation (f Gly
)
was estimated by assuming 0.01 ppm error in NMR peak assignment.
The above values were used in standard error propagation based on
eqs 3−5 to give uncertainties for fGly, Kfold, and ΔGfold. A lower bound
for the folded population of peptide 8a was calculated using eq 3, an
estimated minimum measurable glycine separation value of 0.03 ppm,
and an estimated value for the fully folded state of 0.322 ppm (average
observed for peptides 5b, 6b, 7b, and 9b).
RESULTS AND DISCUSSION
■
Backbone-Substitution Strategies for Mimicry of
Side-Chain Display in a Protein Sheet. As detailed in our
prior work, single α→β residue replacement at matching
positions along each strand of a hairpin-forming α-peptide can
lead to an analogue that retains folding in buffered aqueous
solution, albeit with a change in side-chain display (vide
supra).7 Of 16 β-residue types examined in that study, the two
with the highest propensity for manifesting the sheet fold of the
parent α-peptide sequence were a singly substituted β3-residue
and a doubly substituted syn-β2,3-residue. Based on the above
observations, we conceived four design strategies (Figure 2)
that we envisioned would lead to nativelike display of side
chains after sequence-based modification of a larger α-peptide
sheet.
By contrast to the 1→1 (α→β) residue substitution
examined previously, designs I−IV represent four different
ways of replacing an αα dipeptide with one or more β-residues.
The four designs are meant to explore the impact of two
orthogonal variables on sheet folding after modification: (1)
backbone length relative to the replaced αα segment and (2)
backbone conformational preorganization in the introduced β
residue(s). The variable of backbone length is addressed by
applying 2→1 (αα→β) or 2→2 (αα→ββ) residue substitution.
The role of conformational preorganization is examined by
15908
dx.doi.org/10.1021/ja306311r | J. Am. Chem. Soc. 2012, 134, 15906−15913

Journal of the American Chemical Society
Article
states. This range would allow for measurable population
changes by NMR due to either increase or decrease in folded
stability resulting from backbone substitution.
Peptide 1a (Figure 3), the 16 residue C-terminal segment of
the first immunoglobulin-binding domain of Streptococcal
concert with appropriate control sequences to quantify folded
populations.22
In order to apply the above NMR population analyses for
proposed model systems 2a−5a, we prepared three variants of
each peptide to be examined (Figure 3). Cyclization through
intramolecular disulfide bridge involving Cys residues appended
to the N and C termini provided fully folded control peptides
(2b−5b).23 Dividing the 16 residue sequences in half produced
8 residue N-terminal fragments (2c−5c) and C-terminal
fragments (2d−5d) as unfolded control oligomers. We
synthesized the 16 peptides in series 2−5 by Fmoc solid
phase methods and subjected each to a panel of NMR
experiments (TOCSY, COSY, NOESY) in pH 6.3 phosphate-
buffered aqueous solution. Peptide 4a showed significant signal
1
broadening in the H NMR (attributed to aggregation), so its
folded population was not calculated. In our analysis of the data
obtained for model systems 2, 3, and 5, we found that Hα CSDs
of residues in hydrogen-bonding positions remote from the
turns were most diagnostic of folding. As noted in previous
work on a related hairpin model system, chemical shift values
for Hα’s in non-hydrogen-bonded positions were impacted in
unpredictable ways by their proximity to cross-strand aromatic
residues in the folded state.24
Figure 3. Hairpin model systems derived from the protein GB1.
Hydrophobic core residues from the folded GB1 tertiary structure are
shown in bold. For each sequence examined, four peptides were
prepared to quantify the folded population by NMR: the parent
sequence (1a−5a), a cyclic derivative with two added terminal Cys
residues in a disulfide bridge (1b−5b), an N-terminal fragment (1c−
5c), and a C-terminal fragment (1d−5d).
We determined folded populations for peptides 2a, 3a, and
5a using both Hα CSD and Gly Hα/Hα′ separation (Table 1).
Table 1. Population Analysis for Folding Equilibria of α-
Peptide Model Systems 2a−5a
a
folded population
b
c
peptide
Hα CSD
Gly Hα/Hα′ separation
protein G (GB1), has been shown to form a β-hairpin in
aqueous solution;19 however, it posed some problems that
prevented its use in the present study. The folded state of
peptide 1a is only marginally stable, with a reported hairpin
population of ∼40% at 278 K.19 Moreover, in our hands, the
large number of threonine residues led to poor chemical-shift
dispersion and complicated NMR analysis of its folding
equilibrium. In an effort to find a model sequence with more
favorable properties, we examined peptides 2a−5a (Figure 3), a
series of mutants of sequence 1a. Peptides 2a and 4a are
previously reported variants of 1a with improved folded
stability.20 Sequence 3a is a mutant of 2a with three
modifications intended to destabilize the folded state:
replacement of two sheet-promoting Thr residues with Ala
and elimination of an interstrand salt bridge involving the C
terminus. Peptide 5a is variant of 4a with two destabilizing
Thr→Ala substitutions.
Prior work has established multiple techniques for the
analysis of folding equilibria of short β-hairpin peptides by
multidimensional NMR.20 One commonly employed method is
the comparison of a subset of Hα chemical shifts (δHα) from a
sequence of unknown folded stability to corresponding δHα
values for the same sequence in a random coil conformation;
the resulting chemical shift deviations (CSDs) increase in
magnitude with folded population. When combined with δHα
values for a fully folded control peptide, CSDs can be used to
experimentally determine folded population.19,21 A related
method for analysis of hairpin folding by NMR involves
measuring the separation of diastereotopic Hα protons in a Gly
residue. In a well-folded structure, these two protons experience
different chemical environments on the NMR time scale; thus,
greater Hα/Hα′ separation is indicative of a larger folded
population. Like CSD analysis, this measure can be used in
2a
3a
4a
5a
97 6%
66 5%
ND
94 6%
68 6%
ND
55 2%
64 6%
a
Determined by NMR analysis of a sample in 50 mM phosphate, 9:1
H2O/D2O, pH 6.3 at 293 K. In each case, the cyclic disulfide peptides
2b−5b were used as a fully folded control. Monitored by the change
in Hα chemical shift relative to corresponding unfolded control
sequences 2c−5c and 2d−5d; uncertainties represent the average
population over residues 4, 11, and 13. Monitored by the chemical
shift separation of diastereotopic protons Hα and Hα′ in Gly10.
b
c
Based on these data, we ruled out model system 2 as being too
stable. Sequences 3a and 5a showed similar folded populations,
although peptide 5a had better chemical shift dispersion and
less ambiguity in resonance assignments. Model system 5 was
therefore selected as the starting point for backbone
modification. NMR data for the model peptides indicated
that the folded populations determined by Hα CSD and glycine
Hα/Hα′ separation were comparable within error. We employed
Gly separation in subsequent population analyses to minimize
the impact of changes to sequence register and hydrogen
bonding resulting from different α→β residue replacements in
designs I−IV.
Peptides 1a−5a share a conserved cluster of four hydro-
phobic residues (bold in Figure 3) that are packed in the core
of the GB1 tertiary structure. Our goals in evaluating designs I−
IV when applied for backbone substitution in sequence 5a were
to determine (1) whether the display of the hydrophobic core
tetrad is retained after αα→β or αα→ββ modification and (2)
the thermodynamic impact of backbone modification on
folding if it is tolerated. Application of design methods I−IV
15909
dx.doi.org/10.1021/ja306311r | J. Am. Chem. Soc. 2012, 134, 15906−15913

Journal of the American Chemical Society
Article
to model hairpin 5a generated a series of α/β-peptide
analogues (6a−9a, Figure 4). Since the modifications employed
Figure 5. Glycine Hα/Hα′ separation for observed in the NMR of α
peptide 5b and α/β-peptide analogues 6b−9b. NMR experiments
were performed in 50 mM phosphate, 9:1 H2O/D2O, pH 6.3 at 293 K.
Figure 4. α/β-Peptides resulting from application of designs I−IV to
α-peptide model system 5a. For each sequence examined, four
peptides were prepared to quantify the folded population by NMR:
the parent sequence (5a−9a), a cyclic derivative with two added
terminal Cys residues in a disulfide bridge (5b−9b), an N-terminal
fragment (5c−9c), and a C-terminal fragment (5d−9d).
either extend or contract the backbone, we made matched αα
substitutions at cross-strand positions in the hairpin (residue
pairs Ala4Tyr5 and Phe12Ala13 in 5a). Residues Tyr5 and Phe12
are found in the core of the larger GB1 protein; we therefore
retained their functional groups in design I, which requires
deletion of a side chain. For each backbone modification, we
prepared unfolded and folded control sequences (6b−9b, 6c−
9c, and 6d−9d), as detailed above for the α-peptide model
system. Protected β-amino acids were prepared by various
known routes and used in standard Fmoc solid-phase synthesis
protocols (see the Supporting Information). Each peptide was
purified by reverse-phase HPLC, and its identity was confirmed
by mass spectrometry.
Structural Consequences of α→β Substitution. Disul-
fide-cyclized peptides 6b−9b were first analyzed to determine
the effects of designs I−IV on the folded structure of model
sequence 5b. The Gly Hα/Hα′ separation for three of the α/β-
peptides (6b, 7b, and 9b) was comparable to that of α-peptide
5b (Figure 5). In contrast, peptide 8b showed a markedly
smaller separation of the diastereotopic Hα’s; this observation
suggests design III cannot support the hairpin fold, even with a
structure-promoting disulfide bridge at the terminus. We reason
that the extra flexibility imparted by the two additional
methylene units resulting from αα→β3β2 modification in
design III destabilizes the hairpin structure.
We investigated the hydrogen-bonding pattern and side-
chain display in 6b−9b by analysis of long-range NOE contacts
between the two strands of the hairpin. Peptides 6b and 7b
showed a set of interstrand NOEs consistent with a nativelike
hairpin fold. In support of the hypothesis underlying our
design, the hydrogen-bonding pattern resulting from αα→β
substitution led to a nativelike display of the hydrophobic tetrad
from GB1 on one face of the sheet (Figure 6). Peptide 8b
showed no long-range NOE contacts consistent with hairpin
folding, consistent with its modest Gly Hα/Hα′ separation.
Figure 6. Key interstrand NOEs that establish hydrogen-bonding
register and side-chain display in α-peptide 5b and α/β-peptide
analogues 6b and 7b. Dashed lines indicate NOEs with ambiguous
assignments. NMR experiments were performed in 50 mM phosphate,
9:1 H2O/D2O, pH 6.3 at 293 K.
In contrast to the other α/β-peptides, 9b was only sparingly
soluble in the pH 6.3 buffer system employed for NMR analysis
(9b was soluble at 0.2 mM concentration, while other NMR
samples were typically prepared at 0.4−0.8 mM). The only
NOE correlations observed for 9b corresponded to very close
interproton contacts (intraresidue, sequential). However,
resonances were sharp and not consistent with the signal
broadening typically observed with H-bonding-mediated sheet
aggregation (e.g., as seen for model peptide 4a). We performed
concentration-dependent one-dimensional (1D) 1H NMR
experiments on 5b−9b, and none of the oligomers showed
15910
dx.doi.org/10.1021/ja306311r | J. Am. Chem. Soc. 2012, 134, 15906−15913

Journal of the American Chemical Society
Article
Figure 7. NMR solution structures of α-peptide 5b and α/β-peptide analogues 6b and 7b. Panels (A−C) depict minimized average coordinates
resulting from simulated annealing for 5b (A), 6b (B), and 7b (C). Panels (D−F) compare the display of the hydrophobic tetrad from GB1 (first
side-chain carbon shown as a gray sphere) in 5b (D), 6b (E), and 7b (F). NMR data used for structure determination were obtained in 50 mM
phosphate, 9:1 H2O/D2O, pH 6.3 at 293 K.
peak migration after 5× dilution of the sample used for 2D
analysis. We reason that the poor aqueous solubility of 9b may
arise from a large hydrophobic patch in the folded state
resulting from the insertion of four structure-promoting methyl
groups in the center of the sheet. Despite its modest solubility,
the glycine Hα/Hα′ separation data suggest that 9b adopts a
well-folded structure. We hypothesize that 9b forms a hairpin
with a nativelike display of the GB1 hydrophobic tetrad, but the
lack of direct NOE evidence precludes a conclusive test of this
hypothesis. Collectively, structural analysis for disulfide-cyclized
peptides 6b−9b suggests that three of our proposed designs for
α→β substitution (I, II, and IV) lead to nativelike side-chain
display when applied to prototype α-peptide sheet model
system 5b.
By using NOE-derived distance restraints and simulated
annealing, high-resolution NMR solution structures were
determined for peptides 5a, 5b, 6b, and 7b (Figure 7 and
Supporting Information, Figure S1). The calculations showed
good convergence. The ensemble of low-energy structures
obtained for peptides 5a, 5b, and 7b each showed excellent
internal agreement (Supporting Information, Figure S2). In the
case of peptide 6b, two families of structures were found: a
slightly bent hairpin as well as a hairpinlike structure with a
dramatic curve near the termini (Supporting Information,
Figure S2). We noted some degree of twisting near the termini
of all the cyclic systems, which may be necessary to
accommodate the disulfide bridge. We surmise the presence
of a minor folded population in 6b with a significant kink in the
backbone results from inclusion of a disulfide rather than
insertion of unnatural residues. The less distorted hairpin
structure of peptide 6b was used for further analysis.
residues in the natural protein, we generated overlays of each
with the C-terminal hairpin from the crystal structure of full-
length GB1 (Supporting Information, Figure S3).25 Root-
mean-square deviation (rmsd) values were calculated based on
overlay of the backbone and first side chain C atom for each
key hydrophobic residue. The rmsd values for α/β-peptides 6b
and 7b are 1.03 and 1.80 Å, respectively. As a point of
comparison, this rmsd is 1.18 Å for the corresponding overlay
of α-peptide hairpin 5b with the native protein. We attribute
the better agreement of 6b versus 7b to the hydrophobic cluster
of the native backbone to result from the higher degree of
flexibility in 6b. This increased flexibility may allow better
positioning of the hydrophobic side chains after contraction of
the backbone, a consequence of both designs I and II.
Thermodynamic Impact of α→β Substitution. NMR
analysis of the cyclic disulfide series 5b−9b provided strong
evidence that three of the four originally conceived designs for
backbone substitution (I, II, and IV) can support hairpin
folding with nativelike side-chain display when applied to a
natural sequence. We next sought to compare the thermody-
namic impact of the different substitution strategies on the
folding equilibrium of the linear hairpin peptide 5a. We
calculated the folded population and corresponding free energy
of folding ΔGfold of linear peptides 5a−9a at 278 K (Table 2)
using established methods. Cyclized peptides (5b−9b) and
strand sequences (5c−9c, 5d−9d) were used as fully folded
and unfolded controls, respectively. Since the cyclic disulfide
Table 2. Population Analysis for Folding Equilibria of α-
a
Peptide 5a and α/β-Peptide Analogues 6a−9a
peptide
folded population (%)
ΔGfold (kcal mol−1
)
For each NMR ensemble, we calculated a minimized average
structure and applied it to compare the display of the GB1
hydrophobic tetrad after backbone modification (Figure 7).
Gratifyingly, the four key hydrophobic residues in α/β-peptides
6b and 7b are all clustered on the same face of the sheet as the
natural α-peptide 5b. This observation validates the ability of
designs I and II to facilitate nativelike side-chain display in
backbone modified sheets.
5a
6a
7a
8a
9a
66
24
17
5
5
5
−0.37 0.14
0.65 0.15
0.90 0.19
>1.3
<9
22
4
0.69 0.14
a
Determined by NMR analysis of a sample in 50 mM phosphate, 9:1
H2O/D2O, pH 6.3 at 278 K. Folded population was quantified by the
chemical shift separation of diastereotopic protons Hα and Hα′ in the
internal Gly present in each sequence (see methods).
In order to quantitatively compare the ability of α/β-peptides
6b and 7b to mimic the display of the hydrophobic core
15911
dx.doi.org/10.1021/ja306311r | J. Am. Chem. Soc. 2012, 134, 15906−15913

Journal of the American Chemical Society
Article
control for design III (peptide 8b) did not appear to adopt a
hairpin conformation (vide supra), we estimated a lower bound
for measurable folded population of 8a (see methods). From
the above data, we tabulated the change in folding free energy
(ΔΔGfold) resulting from the application of each backbone
substitution strategy to α-peptide 5a (Figure 8).
protein β-sheets. We applied each of these designs to a model
α-peptide hairpin derived from the sheet domain of a protein
tertiary fold. NMR structural analysis shows that three of the
four designs, which are based on α→β residue replacement,
lead to α/β-peptides that adopt nativelike hairpin folds in
aqueous buffer. Importantly and in contrast to prior work,7
each modified backbone displays a key hydrophobic cluster
from the native sequence on one face of the unnatural sheet.
Thermodynamic analyses of folding equilibria show that the α/
β-peptide hairpins are destabilized by only 0.5−0.6 kcal/mol
per αα-segment substitution.
Based on the findings described here, we propose that
designs I, II, and IV may be tolerated in a larger protein
prototype that includes β-sheet secondary structures. The small
free-energy penalties to folding, significant in the context of a
marginally stable hairpin peptide, might be more easily
tolerated in a well-folded protein, where there is more free
energy to spare. However, it is important to note the gap in
complexity between a hairpin and a sheet in a tertiary context.
The development of foldamers with structural complexity of
typical protein tertiary folds is a significant and unmet challenge
in the field. The results we report here do not achieve this goal,
but they represent an important advance toward it. Taken along
with precedent on the sequence-based mimicry of helices,4−6
turns,10 and loops,11 the new insights into sheet mimicry
reported here lay the groundwork for the design of tertiary
structure mimics based on readily available amino acid
sequence information. We are currently working to develop
and test such a design strategy. These studies will provide a
crucial test of the design principles developed here for sheet
mimicry by establishing their applicability in the much more
demanding context of a folded protein tertiary structure.
Moreover, they will take us one step closer to the long-term
goal of a general methodology for the design of unnatural
analogues of biologically active proteins with improved activity
and/or biostability.
Figure 8. Thermodynamic impact of designs I−IV when applied to to
α-peptide 5a. The difference in folding free energy at 278 K relative to
5a (ΔΔGfold) is shown for α/β-peptides 6a−9a. The value for 8a is a
lower bound for the extent of destabilization.
The thermodynamic impact of designs I, II, and IV applied to
parent sequence 5a are identical, within error. Each αα segment
replacement results in a modest 0.5−0.6 kcal/mol destabiliza-
tion of the folded state relative to the natural peptide. Backbone
contraction (αα→β substitution) appears equally well-tolerated
whether the newly introduced β-residue is conformationally
flexible (β3/β2 in design I) or structure-promoting (β2,3 in
design II). We suggest that the similar folding energies of 6a
and 7a results from two competing and offsetting factors. The
possibility exists that deletion of two atoms from the native α-
peptide backbone results in an unfavorable repositioning of
nearby side chains. The syn-β2,3-monomers in 7a rigidify the
backbone in an extended conformation compatible with sheet
formation, but they may do so at the expense of locking in an
unfavorable side-chain arrangement. The less conformationally
restricted β2- and β3-residues in peptide 6a, while not structure
promoting, may allow for a more optimal positioning of side
chains that compensates energetically for the entropic penalty
resulting from enhanced backbone flexibility.
When the backbone is extended (αα→ββ substitution), the
use of structure promoting syn-β2,3-monomers is essential to
retain the hairpin fold encoded by the parent sequence. The
high degree of conformational flexibility in each β3β2-segment is
likely responsible for the lack of any measurable folded
population for 8a. It is also possible that extension of the
backbone destabilizes the folded hairpin by disrupting diagonal
hydrophobic contacts26 between strands (e.g., Trp→Phe
interactions seen in 5b and 6b). The structure-promoting
monomers in 9a mitigate the energetic penalty resulting from
backbone conformational freedom but still prevent the diagonal
hydrophobic contacts that may stabilize the hairpin.
ASSOCIATED CONTENT
■
S
* Supporting Information
Figures S1−S3, Tables S1−S17, experimental methods, and
minimized average coordinates for the NMR solution structures
of 5a, 5b, 6b, and 7b. This material is available free of charge
via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
horne@pitt.edu
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
Funding for this work was provided by the University of
Pittsburgh.
REFERENCES
(1) Gellman, S. H. Acc. Chem. Res. 1998, 31, 173−180.
(2) (a) Hill, D. J.; Mio, M. J.; Prince, R. B.; Hughes, T. S.; Moore, J.
S. Chem. Rev. 2001, 101, 3893−4012. (b) Davis, J. M.; Tsou, L. K.;
Hamilton, A. D. Chem. Soc. Rev. 2007, 36, 326−334. (c) Goodman, C.
M.; Choi, S.; Shandler, S.; DeGrado, W. F. Nat. Chem. Biol. 2007, 3,
252−262. (d) Bautista, A. D.; Craig, C. J.; Harker, E. A.; Schepartz, A.
Curr. Opin. Chem. Biol. 2007, 11, 685−692. (e) Guichard, G.; Huc, I.
■
CONCLUSIONS
In summary, we have described here four different design
strategies for the sequence-based backbone modification of
■
15912
dx.doi.org/10.1021/ja306311r | J. Am. Chem. Soc. 2012, 134, 15906−15913

Journal of the American Chemical Society
Article
Chem. Commun. 2011, 47, 5933−5941. (f) Horne, W. S. Expert Opin.
Drug Discovery 2011, 6, 1247−1262.
(15) Langenhan, J. M.; Gellman, S. H. J. Org. Chem. 2003, 68, 6440−
6443.
(3) (a) Horne, W. S.; Gellman, S. H. Acc. Chem. Res. 2008, 41, 1399−
1408. (b) Pilsl, L. K. A.; Reiser, O. Amino Acids 2011, 41, 709−718.
(4) (a) Horne, W. S.; Price, J. L.; Keck, J. L.; Gellman, S. H. J. Am.
Chem. Soc. 2007, 129, 4178−4180. (b) Horne, W. S.; Price, J. L.;
Gellman, S. H. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 9151−9156.
(c) Giuliano, M. W.; Horne, W. S.; Gellman, S. H. J. Am. Chem. Soc.
2009, 131, 9860−9861. (d) Price, J. L.; Horne, W. S.; Gellman, S. H. J.
Am. Chem. Soc. 2010, 132, 12378−12387.
(16) Wuthrich, K. NMR in Structural Biology: A Collection of Papers by
̈
Kurt Wuthrich; World Scientific: River Edge, NJ, 1995.
̈
(17) (a) Brunger, A. T.; Adams, P. D.; Clore, G. M.; DeLano, W. L.;
Gros, P.; Grosse-Kunstleve, R. W.; Jiang, J.-S.; Kuszewski, J.; Nilges,
M.; Pannu, N. S.; Read, R. J.; Rice, L. M.; Simonson, T.; Warren, G. L.
Acta Crystallogr., Sect D: Biol. Crystallogr. 1998, 54, 905−921.
(b) Brunger, A. T. Nat. Protocols 2007, 2, 2728−2733.
(18) (a) Searle, M. S.; Ciani, B. Curr. Opin. Struct. Biol. 2004, 14,
458−464. (b) Nowick, J. S. Acc. Chem. Res. 2008, 41, 1319−1330.
(19) Blanco, F. J.; Rivas, G.; Serrano, L. Nat. Struct. Mol. Biol. 1994, 1,
584−590.
(5) (a) Horne, W. S.; Johnson, L. M.; Ketas, T. J.; Klasse, P. J.; Lu,
M.; Moore, J. P.; Gellman, S. H. Proc. Natl. Acad. Sci. U.S.A. 2009, 106,
14751−14756. (b) Johnson, L. M.; Horne, W. S.; Gellman, S. H. J. Am.
Chem. Soc. 2011, 133, 10038−10041. (c) Johnson, L. M.; Mortenson,
D. E.; Yun, H. G.; Horne, W. S.; Ketas, T. J.; Lu, M.; Moore, J. P.;
Gellman, S. H. J. Am. Chem. Soc. 2012, 134, 7317−7320.
(6) (a) Horne, W. S.; Boersma, M. D.; Windsor, M. A.; Gellman, S.
H. Angew. Chem., Int. Ed. 2008, 47, 2853−2856. (b) Lee, E. F.; Smith,
B. J.; Horne, W. S.; Mayer, K. N.; Evangelista, M.; Colman, P. M.;
Gellman, S. H.; Fairlie, W. D. ChemBioChem 2011, 12, 2025−2032.
(c) Boersma, M. D.; Haase, H. S.; Peterson-Kaufman, K. J.; Lee, E. F.;
Clarke, O. B.; Colman, P. M.; Smith, B. J.; Horne, W. S.; Fairlie, W. D.;
Gellman, S. H. J. Am. Chem. Soc. 2012, 134, 315−323.
(20) Fesinmeyer, R. M.; Hudson, F. M.; Andersen, N. H. J. Am.
Chem. Soc. 2004, 126, 7238−7243.
(21) Fesinmeyer, R.; Hudson, F.; Olsen, K.; White, G.; Euser, A.;
Andersen, N. J. Biomol. NMR 2005, 33, 213−231.
(22) Griffiths-Jones, S. R.; Maynard, A. J.; Searle, M. S. J. Mol. Biol.
1999, 292, 1051−1069.
(23) Tatko, C. D.; Waters, M. L. Protein Sci. 2003, 12, 2443−2452.
(24) Espinosa, J. F.; Gellman, S. H. Angew. Chem., Int. Ed. 2000, 39,
2330−2333.
(25) Frericks Schmidt, H. L.; Sperling, L. J.; Gao, Y. G.; Wylie, B. J.;
Boettcher, J. M.; Wilson, S. R.; Rienstra, C. M. J. Phys. Chem. B 2007,
111, 14362−14369.
(7) Lengyel, G. A.; Frank, R. C.; Horne, W. S. J. Am. Chem. Soc. 2011,
133, 4246−4249.
(8) Haase, H. S.; Peterson-Kaufman, K. J.; Lan Levengood, S. K.;
Checco, J. W.; Murphy, W. L.; Gellman, S. H. J. Am. Chem. Soc. 2012,
134, 7652−7655.
(26) Syud, F. A.; Stanger, H. E.; Gellman, S. H. J. Am. Chem. Soc.
2001, 123, 8667−8677.
(9) (a) David, R.; Gunther, R.; Baumann, L.; Luhmann, T.; Seebach,
D.; Hofmann, H.-J.; Beck-Sickinger, A. G. J. Am. Chem. Soc. 2008, 130,
15311−15317. (b) Lee, B.-C.; Zuckermann, R. N. ACS Chem. Biol.
2011, 6, 1367−1374.
(10) (a) Baca, M.; Kent, S. B. H.; Alewood, P. F. Protein Sci. 1993, 2,
1085−1091. (b) Viles, J. H.; Patel, S. U.; Mitchell, J. B. O.; Moody, C.
M.; Justice, D. E.; Uppenbrink, J.; Doyle, P. M.; Harris, C. J.; Sadler, P.
J.; Thornton, J. M. J. Mol. Biol. 1998, 279, 973−986. (c) Odaert, B.;
Jean, F.; Melnyk, O.; Tartar, A.; Lippens, G.; Boutillon, C.; Buisine, E.
Protein Sci. 1999, 8, 2773−2783. (d) Kaul, R.; Angeles, A. R.; Jager,
̈
M.; Powers, E. T.; Kelly, J. W. J. Am. Chem. Soc. 2001, 123, 5206−
5212. (e) Arnold, U.; Hinderaker, M. P.; Nilsson, B. L.; Huck, B. R.;
Gellman, S. H.; Raines, R. T. J. Am. Chem. Soc. 2002, 124, 8522−8523.
(f) Kaul, R.; Deechongkit, S.; Kelly, J. W. J. Am. Chem. Soc. 2002, 124,
11900−11907. (g) Tam, A.; Arnold, U.; Soellner, M. B.; Raines, R. T.
J. Am. Chem. Soc. 2007, 129, 12670−12671. (h) Liu, F.; Du, D.; Fuller,
A. A.; Davoren, J. E.; Wipf, P.; Kelly, J. W.; Gruebele, M. Proc. Natl.
Acad. Sci. U.S.A. 2008, 105, 2369−2374. (i) Fuller, A. A.; Du, D.; Liu,
F.; Davoren, J. E.; Bhabha, G.; Kroon, G.; Case, D. A.; Dyson, H. J.;
Powers, E. T.; Wipf, P.; Gruebele, M.; Kelly, J. W. Proc. Natl. Acad. Sci.
U.S.A. 2009, 106, 11067−11072.
(11) (a) Green, B. R.; Catlin, P.; Zhang, M. M.; Fiedler, B.; Bayudan,
W.; Morrison, A.; Norton, R. S.; Smith, B. J.; Yoshikami, D.; Olivera,
B. M.; Bulaj, G. Chem. Biol. 2007, 14, 399−407. (b) Valverde, I. E.;
Lecaille, F.; Lalmanach, G.; Aucagne, V.; Delmas, A. F. Angew. Chem.,
Int. Ed. 2012, 51, 718−722. (c) Reinert, Z. E.; Musselman, E. D.;
Elcock, A. H.; Horne, W. S. ChemBioChem 2012, 13, 1107−1111.
(12) (a) Krauthauser, S.; Christianson, L. A.; Powell, D. R.; Gellman,
S. H. J. Am. Chem. Soc. 1997, 119, 11719−11720. (b) Seebach, D.;
Abele, S.; Gademann, K.; Jaun, B. Angew. Chem., Int. Ed. 1999, 38,
1595−1597. (c) Karle, I. L.; Gopi, H. N.; Balaram, P. Proc. Natl. Acad.
Sci. U.S.A. 2001, 98, 3716−3719. (d) Karle, I.; Gopi, H. N.; Balaram, P.
Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 5160−5164. (e) Martinek, T. A.;
Toth, G. K.; Vass, E.; Hollosi, M.; Fulop, F. Angew. Chem., Int. Ed.
2002, 41, 1718−1721.
(13) (a) Chi, Y.; Gellman, S. H. J. Am. Chem. Soc. 2006, 128, 6804−
6805. (b) Chi, Y.; English, E. P.; Pomerantz, W. C.; Horne, W. S.;
Joyce, L. A.; Alexander, L. R.; Fleming, W. S.; Hopkins, E. A.; Gellman,
S. H. J. Am. Chem. Soc. 2007, 129, 6050−6055.
(14) Xu, X.; Wang, K.; Nelson, S. G. J. Am. Chem. Soc. 2007, 129,
11690−11691.
15913
dx.doi.org/10.1021/ja306311r | J. Am. Chem. Soc. 2012, 134, 15906−15913