Cross-strand coupling of a β-hairpin peptide stabilized with an Aib-Gly turn studied using isotope-edited IR spectroscopy

Article abstract of DOI:10.1021/ja0736414

Isotope-edited IR spectroscopy was used to study a series of singly and doubly ¹³C=O-labeled β-hairpin peptides stabilized by an Aib-Gly turn sequence. The double-labeled peptides have amide I′ IR spectra that show different degrees of vibrational coupling between the ¹³C- labeled amides due to variations in the local geometry of the peptide structure. The single-labeled peptides provide controls to determine frequencies characteristic of the diagonal force field (FF) contributions at each position for the uncoupled ¹³C=O modes. Separation of diagonal FF and coupling effects on the spectra are used to explain the cross-strand labeled spectral patterns. DFT calculations based on an idealized model β-hairpin peptide correctly predict the vibrational coupling patterns. Extending these model results by consideration of frayed ends and the hairpin conformational flexibility yields an alternate interpretation of details of the spectra. Temperature-dependent isotopically labeled IR spectra reveal differences in the thermal stabilities of the individual isotopically labeled sites. This is the first example of using an IR-based isotopic labeling technique to differentiate structural transitions at specific sites along the peptide backbone in model β-hairpin peptides.

Full text of DOI:10.1021/ja0736414

Published on Web 10/11/2007
Cross-Strand Coupling of a â-Hairpin Peptide Stabilized with
an Aib-Gly Turn Studied Using Isotope-Edited IR
Spectroscopy
Rong Huang,† Vladimir Setnicˇka,†,§ Marcus A. Etienne,‡ Joohyun Kim,†
Jan Kubelka,†,| Robert P. Hammer,‡ and Timothy A. Keiderling*,†
Contribution from the Department of Chemistry, UniVersity of Illinois at Chicago, 845 West
Taylor Street (m/c111), Chicago, Illinois 60607-7061, Department of Chemistry, Louisiana State
UniVersity, 232 Choppin Hall, Baton Rouge, Louisiana 70803-1804
Received June 4, 2007; E-mail: tak@uic.edu
Abstract: Isotope-edited IR spectroscopy was used to study a series of singly and doubly 13CdO-labeled
â-hairpin peptides stabilized by an Aib-Gly turn sequence. The double-labeled peptides have amide I′ IR
spectra that show different degrees of vibrational coupling between the 13C-labeled amides due to variations
in the local geometry of the peptide structure. The single-labeled peptides provide controls to determine
frequencies characteristic of the diagonal force field (FF) contributions at each position for the uncoupled
13CdO modes. Separation of diagonal FF and coupling effects on the spectra are used to explain the
cross-strand labeled spectral patterns. DFT calculations based on an idealized model â-hairpin peptide
correctly predict the vibrational coupling patterns. Extending these model results by consideration of frayed
ends and the hairpin conformational flexibility yields an alternate interpretation of details of the spectra.
Temperature-dependent isotopically labeled IR spectra reveal differences in the thermal stabilities of the
individual isotopically labeled sites. This is the first example of using an IR-based isotopic labeling technique
to differentiate structural transitions at specific sites along the peptide backbone in model â-hairpin peptides.
Introduction
to the intrinsic variability of â-sheet structures (parallel vs
antiparallel, in or out of register, twisted or flat). The tendency
Studies of structural characteristics of peptide models with
ordered secondary structures are important for developing
understanding of protein folding.1,2 Peptide models can show
fundamental characteristics of protein folding such as hydro-
phobic collapse,3 hydrogen bond (H-bond) formation,4 and side-
chain interaction5 in a simpler molecular framework which
allows isolation of the secondary and tertiary interactions.
Early studies in this field focused on model R-helical peptides
using designed peptide sequences to study helix formation and
stability.6-8 By contrast, it has been more difficult to design
and realize good â-sheet and/or â-hairpin peptide models due
of monomer structures to aggregate, which is a required property
for sheet formation, leads to relatively large and indeterminate
sizes.
However, during the past decade, many linear peptide designs
were shown to successfully adopt â-hairpin conformations,
which have facilitated detailed equilibrium and kinetic studies
of â-structures.9-15 A controversy has developed over the
relative importance of hydrophobic collapse and turn formation
for initiating and stabilizing hairpin formation. It is generally
accepted that characteristic H-bonds form later in the process.
Currently there are three major folding mechanisms proposed
for â-hairpin formation: The hydrophobic collapse model16
suggests that the initial step is not due to turn formation but to
the condensation of apolar residues. The “zipping” model17
suggests initial formation of a turn followed by H-bond
formation along the backbone. Finally a mixture of these models
† University of Illinois at Chicago.
‡ Louisiana State University.
§ Current address: Department of Analytical Chemistry, Institute of
Chemical Technology, Technicka 5, 166 28 Prague 6, Czech Republic.
| Current address: Department of Chemistry, University of Wyoming,
Laramie, WY, 82071.
(1) Callender, R. H.; Dyer, R. B.; Gilmanshin, R.; Woodruff, W. H. Annu.
ReV. Phys. Chem. 1998, 49, 173-202.
(2) Eaton, W. A.; Mun˜oz, V.; Hagen, S. J.; Jas, G. S.; Lapidus, L. J.; Henry,
E.; Hofrichter, J. Annu. ReV. Biophys. Biomol. Struct. 2000, 29, 327-359.
(3) Espinosa, J. F.; Gellman, S. H. Angew. Chem., Int. Ed. 2000, 39, 2330-
2333.
(9) Gellman, S. H. Curr. Opin. Chem. Biol. 1998, 2, 717-725.
(10) Ram´ırez-Alvarado, M.; Kortemme, T.; Blanco, F. J.; Serrano, L. Bioorg.
Med. Chem. 1999, 7, 93-103.
(11) Nesloney, C. L.; Kelly, J. W. Bioorg. Med. Chem. 1996, 4, 739-766.
(12) Smith, C. K.; Regan, L. Acc. Chem. Res. 1997, 30, 153-161.
(13) Stotz, C. E.; Topp, E. M. J. Pharm. Sci. 2004, 93, 2881-2894.
(14) Searle, M. S.; Ciani, B. Curr. Opin. Struct. Biol. 2004, 14, 458-464.
(15) Hughes, R. M.; Waters, M. L. Curr. Opin. Struct. Biol. 2006, 16, 514-
524.
(16) Dinner, A. R.; Lazaridis, T.; Karplus, M. Proc. Natl. Acad. Sci. U.S.A.
1999, 96, 9068-9073.
(17) Mun˜oz, V.; Henry, E. R.; Hofrichter, J.; Eaton, W. A. Proc. Natl. Acad.
Sci. U.S.A. 1998, 95, 5872-5879.
(4) Krantz, B. A.; Moran, L. B.; Kentsis, A.; Sosnick, T. R. Nat. Struct. Biol.
2000, 7, 62-71.
(5) Syud, F. A.; Stanger, H. E.; Gellman, S. H. J. Am. Chem. Soc. 2001, 123,
8667-8677.
(6) Rohl, C. A.; Baldwin, R. L. Methods Enzymol. 1998, 295, 1-26.
(7) Scholtz, J. M.; Baldwin, R. L. Annu. ReV. Biophys. Biomol. Struct. 1992,
21, 95-118.
(8) Williams, S.; Causgrove, T. P.; Gilmanshin, R.; Fang, K. S.; Callender, R.
H.; Woodruff, W. H.; Dyer, R. B. Biochemistry 1996, 35, 691-697.
9
13592
J. AM. CHEM. SOC. 2007, 129, 13592-13603
10.1021/ja0736414 CCC: $37.00 © 2007 American Chemical Society

IR-Based Isotopic Labeling Technique
A R T I C L E S
has been proposed.18 These differences arise from the complex
energy landscape for the formation of â-hairpins19 and the effects
of various unique peptide sequences on the relative weights of
different contributions to the process.20-22
factors such as the dynamics of the â-hairpin structure and the
effect of the specific sequence on the cooperativity between the
â-turn formation and the hydrophobic cluster formation. Kinetic
experiments (T-jump IR or T-jump fluorescence) also have some
controversy regarding interpretation of the relaxation in terms
of a single exponential.24,46 In response, isotopic labeling can
provide useful information to test the overall cooperativity and
local stability differences.
Optical spectra are sensitive to rapid structural change and
provide data reflecting the population of various intermediate
states, if they are spectrally distinct. Infrared (IR) and circular
dichroism (CD) spectroscopies are primarily used to monitor
secondary structure, and fluorescence is used to follow the
tertiary structure change. The disadvantage of these methods is
that they are low resolution, having no site selectivity, so that
only average structural change is normally monitored. Isotope-
edited IR spectroscopy can allow study of the thermal unfolding
process in a more site-specific manner, thus yielding a better
understanding of the peptide folding mechanism.37,47
The 16-residue â-hairpin derived from the protein G B1
domain was the first â-hairpin model system widely studied
using equilibrium IR spectroscopy23 and folding kinetics by
T-jump fluorescence.24 Another widely studied â-hairpin model
template (Trpzip), which is stabilized by hydrophobic interac-
tions of Trp residues, was proposed by Cochran,25 and has been
studied by IR,19,26 T-jump IR,27,28 T-jump fluorescence,29,30 and
2D-IR.31,32 We will present results of our isotopic labeling study
of Trpzip models separately (Huang et al., to be published).
We have previously presented spectra and theoretical models
D
for a series of â-hairpin peptides based on Pro-Gly and Asn-
Gly turn sequences incorporated into hairpin sequences designed
by Gellman.33 All these studies combined with NMR data
provided an increasing understanding of the stability and folding
mechanisms of â-hairpin peptides.34,35
While for many proteins the thermal unfolding often appears
to be a two-state process,36 for peptides, in many cases, this is
not an appropriate model.26,34,37-42 Although some â-hairpin
peptides are reported to favor a two-state transition,24,43,44 results
with various methods indicate a deviation from a “true” two-
state transition for several â-hairpin model peptides.19,34,39,41,45
NMR studies using isotopic labeling on different segments of
the peptide sequence provide evidence of multistate unfolding.38
The lack of two-state signature could be attributed to different
Isotope-edited IR spectroscopy has become a useful means
of developing site-specific structural and dynamic information
of polypeptides.47 It has been used to study structural and
coupling information in R-helix conformation,37,48-50 solvation/
desolvation,51-53 N-capping,54 stability and unfolding dyna-
mics,55-57 â-sheet formation and aggregation.58-66 We have
begun a series of studies on â-hairpins using isotopic labeling
and vibrational spectra67-69 for which this paper develops an
interpretation of cross-strand coupling. The most stable hairpin
turns reported incorporate a non-proteinic DPro-Gly sequence33,70,71
(18) Felts, A. K.; Narano, Y.; Gallicchio, E.; Levy, R. M. Proteins 2004, 56,
310-321.
(19) Yang, W.-Y.; Pitera, J. W.; Swope, W. C.; Gruebele, M. J. Mol. Biol. 2004,
336, 241-251.
(46) Xu, Y.; Wang, T.; Gai, F. Chem. Phys. 2006, 323, 21-27.
(47) Decatur, S. M. Acc. Chem. Res. 2006, 39, 169-175.
(48) Decatur, S. M.; Antonic, J. J. Am. Chem. Soc. 1999, 121, 11914-11915.
(49) Barber-Armstrong, W.; Donaldson, T.; Wijesooriya, H.; Silva, R. A. G.
D.; Decatur, S. M. J. Am. Chem. Soc. 2004, 126, 2339-2345.
(50) Huang, R.; Kubelka, J.; Barber-Armstrong, W.; Silva, R. A. G. D.; Decatur,
S. M.; Keiderling, T. A. J. Am. Chem. Soc. 2004, 126, 2346-2354.
(51) Starzyk, A.; Barber-Armstrong, W.; Sridharan, M.; Decatur, S. M.
Biochemistry 2005, 44, 369-376.
(20) Deechongkit, S.; Nguyen, H.; Jager, M.; Powers, E. T.; Gruebele, M.; Kelly,
J. W. Curr. Opin. Struct. Biol. 2006, 16, 94-101.
(21) Colombo, G.; DeMori, G. M. S.; Roccatano, D. Protein Sci. 2003, 12, 538-
550.
(22) Griffiths-Jones, S. R.; Sharman, G. J.; Maynard, A. J.; Searle, M. S. J.
Mol. Biol. 1998, 284, 1597-1609.
(23) Arrondo, J. L. R.; Blanco, F. J.; Serrano, L.; Goni, F. M. FEBS Lett. 1996,
384, 35-37.
(52) Fesinmeyer, R. M.; Peterson, E. S.; Dyer, R. B.; Andersen, N. H. Protein
Sci. 2005, 14, 2324-2332.
(24) Mun˜oz, V.; Thompson, P. A.; Hofrichter, J. Nature 1997, 390, 196-199.
(25) Cochran, A. G.; Skelton, N. J.; Starovasnik, M. A. Proc. Natl. Acad. Sci.
U.S.A. 2001, 98, 5578-5583.
(53) Walsh, S. T. R.; Cheng, R. P.; Wright, W. W.; Alonso, D. O. V.; Daggett,
V.; Vanderkooi, J. M.; DeGrado, W. F. Protein Sci. 2003, 12, 520-531.
(54) Decatur, S. M. Biopolymers 2000, 54, 180-185.
(55) Huang, C. Y.; Getahun, Z.; Wang, T.; DeGrado, W. F.; Gai, F. J. Am.
Chem. Soc. 2001, 123, 12111-12112.
(26) Wang, T.; Xu, Y.; Du, D.; Gai, F. Biopolymers 2004, 75, 163-172.
(27) Du, D.; Tucker, M. J.; Gai, F. Biochemistry 2006, 45, 2668-2678.
(28) Du, D.; Zhu, Y.; Huang, C.-Y.; Gai, F. Proc. Natl. Acad. Sci. U.S.A. 2004,
101, 15915-15920.
(56) Venyaminov, S. Y.; Hedstrom, J. F.; Prendergast, F. G. Proteins 2001, 45,
81-89.
(29) Snow, C. D.; Qiu, L.; Du, D.; Gai, F.; Hagen, S. J.; Pande, V. S. Proc.
Natl. Acad. Sci. U.S.A. 2004, 101, 4077-4082.
(57) Ramajo, A. P.; Petty, S. A.; Starzyk, A.; Decatur, S. M.; Volk, M. J. Am.
Chem. Soc. 2005, 127, 13784-13785.
(30) Yang, W.-Y.; Gruebele, M. J. Am. Chem. Soc. 2004, 126, 7758-7759.
(31) Smith, A. W.; Chung, H. S.; Ganim, Z.; Tokmakoff, A. J. Phys. Chem. B
2005, 109, 17025-17027.
(58) Brauner, J. W.; Dugan, C.; Mendelsohn, R. J. Am. Chem. Soc. 2000, 122,
677-683.
(32) Wang, J.; Chen, J.; Hochstrasser, R. M. J. Phys. Chem. B 2006, 110, 7545-
(59) Silva, R. A. G. D.; Barber-Armstrong, W.; Decatur, S. M. J. Am. Chem.
Soc. 2003, 125, 13674-13675.
7555.
(33) Stanger, H. E.; Gellman, S. H. J. Am. Chem. Soc. 1998, 120, 4236-4237.
(34) Hilario, J.; Kubelka, J.; Keiderling, T. A. J. Am. Chem. Soc. 2003, 125,
7562-7574.
(60) Petty, S. A.; Adalsteinsson, T.; Decatur, S. M. Biochemistry 2005, 44,
4720-4726.
(61) Petty, S. A.; Decatur, S. M. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 14272-
(35) Hilario, J.; Kubelka, J.; Syud, F. A.; Gellman, S. H.; Keiderling, T. A.
Biospectroscopy 2002, 67, 233-236.
14277.
(62) Petty, S. A.; Decatur, S. M. J. Am. Chem. Soc. 2005, 127, 13488-13489.
(63) Paul, C.; Axelsen, P. H. J. Am. Chem. Soc. 2005, 127, 5754-5755.
(64) Paul, C.; Wang, J.; Wimley, W. C.; Hochstrasser, R. M.; Axelsen, P. H. J.
Am. Chem. Soc. 2004, 126, 5843-5850.
(36) Jackson, S. E. Folding Des. 1998, 3, R81-91.
(37) Silva, R. A. G. D.; Kubelka, J.; Decatur, S. M.; Bour, P.; Keiderling, T. A.
Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 8318-8323.
(38) Santiveri, C. M.; Santoro, J.; Rico, M.; Jimenez, M. A. J. Am. Chem. Soc.
2002, 124, 14903-14909.
(65) Lansbury, P. T. J.; Costa, P. R.; Griffiths, J. M.; Simon, E. J.; Auger, M.;
Halverson, K. J.; Kocisko, D. A.; Hendsch, Z. S.; Ashburn, T. T.; Spencer,
R. G. Nat. Struct. Biol. 1995, 2, 990-998.
(39) Kuznetsov, S. V.; Hilario, J.; Keiderling, T. A.; Ansari, A. Biochemistry
2003, 42, 4321-4332.
(66) Halverson, K.; Sucholeiki, I.; Ashburn, T. T.; Lansbury, P. T. J. Am. Chem.
Soc. 1991, 113, 6701-6703.
(40) Lopez, M. M.; Chin, D.-H.; Baldwin, R. L.; Makhatadze, G. I. Proc. Natl.
Acad. Sci. U.S.A. 2002, 99, 1298-1302.
(67) Setnicˇka, V.; Huang, R.; Thomas, C. L.; Etienne, M. A.; Kubelka, J.;
Hammer, R. P.; Keiderling, T. A. J. Am. Chem. Soc. 2005, 127, 4992-
4993.
(41) Ahmed, Z.; Beta, I. A.; Mikhonin, A. V.; Asher, S. A. J. Am. Chem. Soc.
2005, 127, 10943-10950.
(42) Yamazaki, T.; Furuya, H.; Watanabe, T.; Nishiuchi, Y.; Nishio, H.; Abe,
A. Chem. Today 2006, 24, 55-58.
(68) Bour, P.; Keiderling, T. A. J. Phys. Chem. B 2005, 109, 5348-5357.
(69) Kim, J.; Huang, R.; Kubelka, J.; Bour, P.; Keiderling, T. A. J. Phys. Chem.
B 2006, 110, 23590-23602.
(43) Espinosa, J. F.; Mun˜oz, V.; Gellman, S. H. J. Mol. Biol. 2001, 306, 397-
402.
(70) Ragothama, S. R.; Awasthi, S. K.; Balaram, P. J. Chem. Soc., Perkin Trans.
1998, 2, 137-143.
(44) Streicher, W. W.; Makhatadze, G. I. J. Am. Chem. Soc. 2006, 128, 30-31.
(45) Dhanasekaran, M.; Prakash, O.; Gong, Y.-X.; Baures, P. W. Org. Biomol.
Chem. 2004, 2, 2071-2082.
(71) Zhao, C.; Polavarapu, P. L.; Das, C.; Balaram, P. J. Am. Chem. Soc. 2000,
122, 8228-8231.
9
J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007 13593

A R T I C L E S
Huang et al.
Table 1. Model Peptide Sequences
resulting in a tight left-handed turn that matches the preferred
right-hand twist of the antiparallel strands.72 However, this is
not an optimal choice for isotopic substitution studies since the
Xxx-Pro amide linkage has an amide I′ band at ∼1611 cm-1
that spectrally overlaps the expected position of the 13CdO peak
of interest.34 In this study, we substituted an Aib residue
(aminobutyric acid or R,R-dimethyl glycine) into the Gellman
sequence (Scheme 1) to initiate turn formation and stabilize
hairpin formation73 which results in spectral properties suitable
for isotopic study.
a Underlined amino acid is 13CdO labeled (Scheme 1). b B ) Aib
(aminoisobutyric acid or R,R-dimethylglycine), O ) Orn (ornithine)
Scheme 1. Design Structure of Gellman A Peptide Showing
Numbering of Individual Residues in Sequence and
13CdO-Labeling Patterns (colored rectangles)
strands contain residues with very different (φ,ψ) angles due
to both a turn between them and frayed termini. Additionally
the â-strands are highly twisted.
DFT calculations based on a regular model geometry do not
include the structural variation possible in a dynamically
fluctuating hairpin. To overcome this limitation and better
simulate the real molecule, we carried out DFT calculations on
selected structures derived from long-time (∼100 ns) molecular
dynamics (MD) trajectories as well as from the NMR structure
of the peptide.69 Additionally we have used such trajectories to
provide ensembles of structures for which we carried out
classical (transition dipole) calculations of cross-strand coupling
of labeled positions.69 In this work, we will compare the
experimental spectra to interpretations derived from these
different theoretical simulations. A preliminary report on the
spectra of some of these labeled molecules has appeared.67
Density functional theory (DFT) calculations of model
peptides have long been used to simulate IR and other peptide
vibrational spectra.50,68,72,74-81 The use of full optimization of
the structure while constraining the (φ,ψ) angles to yield a
specific model conformation has been successful in elucidating
spectral characteristics of higher-frequency, amide-centered
modes. Transfer of DFT calculated vibrational parameters allows
calculations of the vibrational spectra for very large peptides,
provided that they have regular repeating geometries.77,82 In
particular, the vibrational coupling (frequency splitting) of
specifically labeled 13CdO’s for both R-helical and â-sheet
peptides can be accurately predicted by DFT-based calcula-
tions,37,50,58,68,74,75 which is useful for interpreting the isotopically
labeled amide I′ spectral patterns in regular structures, but even
more important for using the spectra of isotopically labeled
peptides to elucidate their local structural changes. Lower-level
semiempirical quantum chemistry calculations32,81,83 have been
also used to predict the vibrational couplings. For â-hairpin
peptides realistic spectral simulations are complicated by the
fact that the structures are not regular, since the antiparallel
Materials and Methods
Peptide Synthesis and Purification. All peptides studied were
synthesized at Louisiana State University on a Pioneer Peptide
Synthesizer using four equivalents of amino acid and activation with
TBTU, HOBt (final concentration of 0.25 M), and diisopropylethy-
lamine (DIEA) at a final concentration of 0.5 M. A stepwise coupling
of each amino acid was obtained using the standard solid-phase Fmoc
coupling chemistry. Aib was double coupled using PyAOP as the
coupling agent and DIEA as the activator (4 equiv and 0.2 M final
concentration of PyAOP and amino acid) at an elevated temperature
of 50 °C due to the difficult coupling of the sterically hindered Aib.84,85
Additionally, the residue on the N-terminal side of the Aib, Val5 in
this case, was also double coupled with PyAOP at 50 °C as above,
due to the difficulty in acylating the hindered Aib amino group.84-86
Labeled peptides were synthesized by using Fmoc-protected, side-chain
protected (as appropriate) 1-13C-labeled Fmoc-blocked amino acids
purchased from Cambridge Isotope Laboratories (CIL), Inc. Fmoc-1-
13C-Ile-OH was prepared as previously described in our initial report.67
Once the sequence was complete, the peptide was removed from the
solid-phase resin by using a cleavage cocktail (by dissolving in 88:5:
5:2 TFA/phenol/water/TIPS for 2 h). Crude peptides were isolated by
precipitation into 10 volumes of cold ether, purified by reverse phase
HPLC (purity g95%) and characterized by MALDI-MS. For clarity,
we use the notations in Table 1 to designate the various positions labeled
in the peptides used in this study. Lys (K) was substituted for Orn (O)
in the original Gellman sequence for labeling convenience and has no
structural or spectral consequence due to the similarity of K and O.
ECD Sample Preparation and Spectral Parameters. For ECD
experiments, peptide solutions of ∼0.15 mg/mL (∼0.1 mM) in 20 mM
phosphate buffer (pH ≈ 7) were prepared. Far-UV CD spectra were
(72) Bour, P.; Keiderling, T. A. J. Mol. Struct. (THEOCHEM) 2004, 675, 95-
105.
(73) Aravinda, S.; Shamala, N.; Rajkishore, R.; Gopi, H. N.; Balaram, P. Angew.
Chem. 2002, 114, 4019-4021.
(74) Kubelka, J.; Huang, R.; Keiderling, T. A. J. Phys. Chem. B 2005, 109,
8231-8243.
(75) Kubelka, J.; Keiderling, T. A. J. Am. Chem. Soc. 2001, 123, 6142-6150.
(76) Kubelka, J.; Keiderling, T. A. J. Am. Chem. Soc. 2001, 123, 12048-12058.
(77) Kubelka, J.; Silva, R. A. G. D.; Bour, P.; Decatur, S. M.; Keiderling, T. A.
In Chirality: Physical Chemistry. Hicks, J. M., Ed.; ACS Symposium Series
810; American Chemical Society: Washington DC, 2002; pp 50-64.
(78) Bour, P.; Keiderling, T. A. J. Phys. Chem. B 2005, 109, 232687-23697.
(79) Jalkanen, K. J.; Jurgensen, V. W.; Claussen, A.; Rahim, A.; Jensen, G.
M.; Wade, R. C.; Nardi, F.; Jung, C.; Degtyarenko, I. M.; Nieminen, R.
M.; Herrmann, F.; Knapp-Mohammady, M.; Niehaus, T. A.; Frimand, K.;
Suhai, S. Int. J. Quantum Chem. 2006, 106, 1160-1198.
(80) Choi, J.-H.; Kim, J.-S.; Cho, M. J. Chem. Phys. 2005, 122, 174903-174911.
(81) Choi, J.-H.; Hahn, S.; Cho, M. Int. J. Quantum Chem. 2005, 104, 616-
634.
(84) Fu, Y.; Etienne, M. A.; Hammer, R. P. J. Org. Chem. 2003, 68, 9854-
9857.
(82) Bour, P.; Sopkova, J.; Bednarova, L.; Malon, P.; Keiderling, T. A. J.
Comput. Chem. 1997, 18, 646-659.
(83) Hahn, S.; Ham, S.; Cho, M. J. Phys. Chem. B 2005, 109, 11789-11801.
(85) Fu, Y.; Hammer, R. P. Org. Lett. 2002, 4, 237-240.
(86) Martin, L.; Ivancich, A.; Vita, C.; Formaggio, F.; Toniolo, C. J. Pept. Res.
2001, 58, 424-432.
9
13594 J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007

IR-Based Isotopic Labeling Technique
A R T I C L E S
acquired between 185 and 250 nm at 50 nm/min, with 1-nm bandwidth
and 2-s response time on a JASCO J-810 spectrometer using a 1-mm
quartz cell (Starna, Inc). As a test for concentration effects, spectra
were also run at 20, 5, and 1 mg/mL. Final spectra were recorded as
an average of eight scans and baseline subtracted. Variable-temperature
experiments were done by monitoring the temperature of the sample
holder heated by flow from a water bath (Neslab RTE7DP) with a
1 °C/min ramp speed and a 5-min equilibration time under control of
a JASCO program. A set of 19 spectra over the range of 5-95 °C (in
steps of 5 °C) was obtained for sample and buffer under identical
experimental conditions, and subtractions for baseline correction were
performed afterward. No smoothing was performed on the spectra
before analysis.
IR Sample Preparation and Spectral Parameters. Purified pep-
tides were lyophilized against 0.1 M DCl/D2O solution (both DCl and
D2O from Sigma), redissolved in D2O, neutralized by adding small
aliquots of NaOD solution, and lyophilized again. Peptide solutions
for IR studies were prepared with a range of concentrations, ∼10-46
mg/mL (∼7-32 mM), by dissolving lyophilized peptides in 20 mM
deuterated phosphate buffer (pH 7, ≈ uncorrected). These peptides are
quite soluble under the above conditions, and the data presented were
normally obtained at ∼23 mg/mL with a 100-µm path length. As a
test for concentration effects, spectra were also run at 20, 5, and 1
mg/mL.
IR spectra were acquired on a DigiLab FTS-60A spectrometer.
Peptide solutions were sealed in a homemade demountable CaF2
window cell having either a 50- or 100-µm Teflon spacer. All spectra
(sample and background) were collected at a resolution of 4 cm-1 with
a zero-filling factor of 8. Temperature-variation experiments were
conducted with a homemade cell holder controlled from 5 °C to 95 °C
by flow from a bath (Neslab RTE 111). The experiments were realized
by simultaneous execution of two independent programs (DigiLab
Resolution Pro for spectral scan and NEScom for temperature ramp).
The sample was heated at 1 °C/min and equilibrated for 11 min, and
then a 16-minute data collect (940 scans co-added) was accumulated
before the next temperature step.
IR Spectral Simulations. A model â-hairpin peptide structure was
derived from structural parameters for two strands of the â-sheet domain
in the structure of intestinal fatty acid binding protein (PDB code: 1IFC)
as has been described separately.76,87 An NMR-based structure of the
U peptide studied here88 was also used to create another model for
comparison calculations. The methods used to fragment the structure
into smaller turn and strand components, to carry out DFT calculations
of the spectral property tensors (force fields, and atomic polar tensors),
and to transfer these onto the full hairpin structure have been previously
described in detail.34,69,82 DFT calculations of the force field and atomic
polar tensors were done at the BPW91/6-31G** level for just the peptide
framework (i.e., all Ala, no solvent) using the Gaussian 03 package.
This level of computation has well-known limitations, but we have
shown it gives qualitatively useful analysis of the amide IR spectra.69,74,78
IR spectral patterns were simulated by assigning to each computed
transition a band shape having a width of 10 cm-1 (fwhm) to best match
the experimental resolution.
Figure 1. Temperature-dependent ECD spectra of the unlabeled peptide
(U) measured from 5 to 95 °C. Peptide concentration is ∼0.15 mg/mL in
20 mM phosphate buffer. The path length of the cell is 1 mm. Arrows
indicate direction of ellipticity changes with increasing temperature, which
is color coded from violet (cold) to red (hot). (Inset) Ellipticity changes at
various peak positions (color correlated to the arrows, 188 nm - red, 201
nm - blue, 215 nm - black, and 230 nm - pink) for the unlabeled peptide
(U) shown with increasing temperature. Solid lines are the best fit for the
data.
(PME). The initial configuration was a folded model structure derived
from an NMR-determined structure,88 and the MD trajectories sampled
several unfolding and refolding events, particularly at high temperatures.
Conformations obtained from these trajectories provided alternative
structural models for the spectral calculations.69 For comparison with
low-temperature experimental data, the conformational fluctuations were
analyzed specifically for the parts of the trajectories where the peptide
maintains an overall folded nature.
Results
ECD Thermal Unfolding. ECD spectra of all peptides were
measured in 20 mM phosphate buffer (pH ≈ 7) to verify that
they all adopted a â-hairpin conformation. All the peptides
studied (the unlabeled and the labeled peptide variants) show
identical band shapes, which are characteristic of â-sheet spectra
(with a negative trough at ∼215 nm and a positive peak at ∼201
nm), but the positive peak is reduced in intensity due to overlap
with a negative peak at ∼188 nm, echoing some features found
in turns90 (Figure 1). These shapes are independent of concen-
tration up to 20 mg/mL, indicating freedom from aggregation
effects. The aromatic residue (Tyr) in the peptide sequence gives
rise to a weaker positive feature at 230 nm. Since isotopic
substitution should have no effect on the ECD spectra, the
identical nature of these spectra indicates all the peptides have
the same fold. Added TFE did not further stabilize the â-hairpin
conformation, but rather induced the peptide to form a partial
MD Simulations. Molecular Dynamics (MD) calculations were
carried out in explicit water at temperatures ranging from 280 to 500
K. At each temperature several trajectories were simulated up to 100
ns. The simulation box used contains about 1500 SPC water molecules
and the peptide of interest, which was modeled with OPLS-AA. An
NVT ensemble was used by employing a Berendsen thermostat with a
0.1 ps coupling time. Simulations were performed using Gromacs 3.0
and 3.3.89 The long-range interactions were implemented either with a
simple 1.0 nm cutoff or with 1.0 nm cut offs and Particle Mesh Ewald
(87) Kubelka, J. Ph.D. Thesis. University of Illinois at Chicago, Chicago, 2002.
(88) Masterson, L. R.; Etienne, M. A.; Porcelli, F.; Barany, G.; Hammer, R. P.;
Veglia, G. Biopolymers 2007, 88, 746-753.
(90) Rose, G. D.; Gierasch, L. M.; Smith, J. A. AdV. Protein Chem. 1985, 37,
1-109.
(89) Lindahl, E.; Hess, B.; Spoel, D. v. d. J. Mol. Model. 2001, 7, 306-317.
9
J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007 13595

A R T I C L E S
Huang et al.
Figure 2. Experimental IR spectra of (a) the unlabeled and single-labeled hairpin peptides and (c) the double-labeled peptides. The experimental spectra
were normalized to have a peak area of 1 between 1720 and 1550 cm-1. The original spectra of all peptides were measured at ∼23 mg/mL in 20 mM
deuterated phosphate buffer in a 100 µm path length cell. Simulated IR spectra of Ac-A12-NHCH3 by transfer from shorter fragments are in (b) for the
unlabeled peptide compared with single-labeled peptides A3, A5, and A10 and (d) for the double-labeled peptides A3A8, A3A10, and A5A8.
R-helical-like conformation with a double minimum at 222 and
208 nm and a more intense positive peak at 193 nm (data not
shown). This change is probably aided by the high helical
propensity of Aib.91
and glutamic acid (∼1568 cm-1) at neutral pH.56,92,93 On heating
up to ∼95 °C, the amide I′ band shifted to 1643 cm-1, consistent
with unfolding of the hairpin, and the two Arg side-chain
absorption bands converge to a broad band at high temperature.
Another Arg side-chain band appears at ∼1608 cm-1 at high
temperature.56,93
The thermal stability of the secondary structure of the hairpin
peptide in phosphate buffer was monitored by variable-temper-
ature ECD (Figure 1). The overall band shape is preserved
from 5 to 95 °C, but the characteristic peak intensities (for the
features at 188, 201, and 230 nm) are much lower at high
temperature, although there is little change in intensity for
the negative band (â-structure) at ∼215 nm. These results are
very similar to the previously reported ECD for the closely
related Gellman-designed 12mer â-hairpin peptide, DPG12
(RYVEVDPGKKILQ).33-35 The main positive peak shifts from
201 to 199 nm with increasing temperature and does not show
a clear isodichroic point, which suggests an increasing contribu-
tion from residues in a disordered conformation. The broad
thermal transition, as monitored by temperature-induced ellip-
ticity changes at various wavelengths (Figure 1, inset), shows
little curvature and suggests that the peptide does not undergo
a simple two-state unfolding process. The high-temperature band
shape indicates that the peptide retains some residual secondary
structure in solution even at 95 °C.
To expose the local unfolding process at specific sites, amide
13CdO isotopic variants were prepared with single labels on
V3, V5, and I10 and double labels on V3K8, V3I10, and V5K8.
However, before developing a model for the effect of confor-
mational changes on the 13C-labeled group vibrations, it is neces-
sary to first understand the origins of the isotopic vibrational
frequency shifts. The single labels provide data for identifying
the local, position-determined frequency (diagonal force-field
terms), and the double labels add the effects of the cross-strand
vibrational coupling (off-diagonal terms) to them. The single-
labeled V3 and V5 hairpins give rise to 13CdO sidebands at
∼1613 cm-1 (which for V5 is weak, but obvious with
deconvolution), and I10 shows a sideband at ∼1610 cm-1
(Figure 2a), or ∼20-23 cm-1 below the maximum of the
unlabeled hairpin. Additionally the main 12CdO band shifts
higher in frequency, increasing the 12C-13C gap, in a manner
dependent on the substitution position.
The double-labeled variants have quite different 13CdO
patterns (Figure 2c). V3I10 (whose labeled 13CdO groups form
a small, 10-member H-bonded ring) has a moderate intensity
peak at 1607 cm-1, while V3K8 (13CdO’s form a big,
14-member H-bonded ring) has a strong 13CdO peak at 1614
cm-1, or 7 cm-1 to higher frequency and more intense than
Experimental and Simulated IR Spectra. Experimentally,
the unlabeled â-hairpin gives rise to a broadened amide I′ IR
band (D2O solution) whose maximum is ∼1633 cm-1 at 5 °C,
which is consistent with formation of a partial antiparallel
â-sheetlike cross-strand interaction (Figure 2a, black line). The
shoulder peaks below 1600 cm-1 are contributions from the side-
chain-originating absorption bands of arginine (∼1585 cm-1
)
(92) Barth, A. Prog. Biophys. Mol. Biol. 2000, 74, 141-173.
(93) Graff, D. K.; Pastrana-Rios, B.; Venyaminov, S. Y.; Prendergast, F. G. J.
Am. Chem. Soc. 1997, 119, 11282-11294.
(91) Toniolo, C.; Crisma, M.; Formaggio, M.; Peggion, C. Biopolymers 2001,
60, 396-419.
9
13596 J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007

IR-Based Isotopic Labeling Technique
A R T I C L E S
Table 2. Comparison of Experiment and Simulations for Labeled
â-Hairpin Peptides
presumably attributable to dynamic fluctuations between mul-
tiple conformations of the short, twisted â-strands arising from
disorder at the frayed ends and to the strain and distortion
involved in the turn.
experimental
simulated
computed
spectral peaka
spectral peaka
normal modesb
notation
12C
13C
difference
12C
13C
difference
13Cd
splitting
When singly substituted with 13C in the strands, the hairpin
exhibits an additional peak (due to the isotopic effect) that is
computed at ∼40 cm-1 lower-frequency than the main 12C
amide I band for V3 and I10. The peptide V5 has a single label
U
1633 N/A
1642 1613
1637 1614
1640 1610
N/A
29
23
30
28
1678 N/A
1678 1636
1678 1650
1679 1636
1684 1641
N/A
42
28
43
43
N/A
1635
1650
1637
1632 (w)
1640 (s)
1632 (s)
1640 (w)
1635 (m) -16
1651 (m)
N/A
N/A
N/A
N/A
+8
V3
V5
I10
V3K8 1642 1614
next to the turn and is computed to have a 14 cm-1 higher 13
C
V3I10 1643 (1627c)
1607
V5K8 1638 1616
16
36
22
1680 1633
1676 1636
47
40
-8
frequency than do V3 or I10, due to its poorly formed cross-
strand H-bond. This is of course specific to the model structure
chosen, and probably cannot be generalized. The diagonal FF
can obviously vary significantly with position in the hairpin,
which probably relates to local distortion in the model.
a Frequencies (in cm-1) are determined from the original spectra as
apparent maxima. No resolution enhancement used. b Frequency of indi-
vidual component modes (in cm-1). c This frequency corresponds to the
shoulder peak. d Intensity of the normal modes (in parentheses): s - strong,
m - medium, w - weak.
The spectra of the 13C disubstituted peptides were also
computed (Figure 2d). Although there are two components
(normal modes) associated with the isotopic labeled carbonyls,
only one 13C peak is apparent in the computed spectra because
of the relatively large difference in intensity for the two labeled
components as computed with our model. For the double-
labeling pattern forming a large H-bond ring (14-atom) such as
V3K8, which is in the middle of the hairpin, the higher-
frequency component is more intense. By contrast, the labels
forming a smaller ring (10-atom), such as V3I10 (in the middle
of the hairpin) and V5K8 (near the turn), have the lower
component predicted to be more intense. The simulated
13CdO peak positions of V3K8 and V3I10 result in an 8 cm-1
difference in the 13C absorbance maxima of these two peptides
due to their having opposite components of the pair being most
intense in the IR. These results for the 13CdO modes are also
summarized in Table 2. In our idealized simulation, the V5K8
13C peak is not the same as that for V3I10, suggesting its
13CdO coupling is somewhat different than in V3I10. This
again, as noted for the V5 case, could be the result of local
distortion of the structure near the turn. The computed 12C peak
frequencies for these labeled peptides do not vary much (except
for V3K8) due to the introduction of labels, but the 12CdO IR
intensities decrease in proportion to the increase in the number
of 13CdO groups.
V3I10. These 13CdO sideband results reflect predictions of our
theoretical model (see below). The V3I10 also has a weak
shoulder at ∼1627 cm-1, which could arise from the isotope-
substituted positions disrupting 12CdO coupling in the strands,
as would explain the weaker, but similar shoulder in I10. For
V5K8, the 13C peak is only evident as a weak shoulder at ∼1616
cm-1. These results are summarized in Table 2.
To better understand the role of isotopic labels on the IR
spectra of the hairpin peptides, we have simulated IR spectra
of a model â-hairpin initially using an idealized structure derived
from a protein â-sheet segment (pdb 1IFC)76 and compared them
to our experimental IR spectra of the above peptides. In Figure
2b, the simulated amide I IR spectrum of the unlabeled peptide
shows an intensity pattern typical of a sheet conformation,
having an intense peak at low frequency (computed at ∼1678
cm-1) and a weaker peak at high frequency (∼1737 cm-1). The
moderate band (1713 cm-1) between these extremes corresponds
to the non-H-bonded amide groups directed away from the other
strand (out at the vacuum in this model). Our simulated spectra
are not corrected for solvent (since they are derived from
vacuum calculations on ideal hairpin segments, with Ala-based
sequences) or for the expected errors for normal modes
calculated with a DFT FF; thus, the computed amide I′ values
are ∼45 cm-1 higher than the experimental 12CdO maxima.
Due to these solvation effects, in order to compare experimental
and simulated 13CdO effects, it is best to look at frequency
differences which are compiled in Table 2 for the 13CdO-
12CdO shift and for the 13CdO coupling patterns for double-
labeled species.
The 12CdO experimental values do correspond to the ex-
pected lower-frequency component of a twisted â-sheet con-
formation.72,75 The amide I′ frequency of a twisted pair of strands
would be higher than for an extended, multistranded â-sheet
with maximal coupling between strands.75 For this reason the
1633 cm-1 amide I′ maximum of the unlabeled hairpin is higher
than the amide I′ maximum observed at ∼1610 cm-1 in, for
example, poly-Lys â-sheets.94 The experimental IR spectrum
of the unlabeled peptide (Figure 2a) is broader than the simulated
12C amide I′ band but actually has less apparent dispersion,
implying less exciton coupling due to nonuniformity of the
â-strands. The broadness of the experimental 12CdO band is
In Table 2, we have summarized the observed and predicted
12C and 13C peak frequencies and their differences along with
computed normal-mode frequencies of the 13C labels. Due to
the absolute frequencies obtained for amide I′ modes with
vacuum simulations being significantly higher than experimental
values, difference frequencies offer the best basis of comparison.
Single-labeled peptides V3 and I10 have a larger experimental
12C-13C frequency difference than V5 (Table 2, column 4), but
this is due to the V5 12CdO frequency being lower, which
probably reflects less disruption of the â-strand exciton coupling
since the V5 label is close to the turn. In the same sense, the
double-labeled peptide V3I10 has the largest isotope shift (if
only considering the low-frequency 13C peak), followed by
V3K8 and finally V5K8. Simulated spectra (Table 2, column
7) show the same trends but not for the same reasons. The
single-labeled V5 peptide again has the smallest difference
frequency. In this case, the reduction is due to the 13CdO mode
being very high compared to the others. Most likely this unique
13CdO mode results from a difference in the V5 diagonal FF
contribution, possibly due to its location next to the turn. The
variations in the experimental 12CdO mode frequencies may
(94) Yasui, S. C.; Keiderling, T. A. J. Am. Chem. Soc. 1986, 108, 5576-5581.
9
J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007 13597

A R T I C L E S
Huang et al.
Figure 3. Original temperature-dependent IR spectra (left panels) and difference spectra (right panels) for double-labeled peptides V3K8, V3I10, and
V5K8. All peptide concentrations were ∼23 mg/mL in 20 mM deuterated phosphate buffer measured in a 100-µm path length cell. The difference spectra
were taken by subtracting the spectrum at 5 °C from each spectrum at higher temperatures. In all panels the spectra are color coded for temperature, from
blue at the cold (5 °C) extreme to red at the hot (95 °C) end, as suggested by the arrows.
reflect disruption of exciton coupling in the hairpin â-strands,
which would lead to lower-frequency amide I′ maxima in longer
â-strands. The comparisons made in this paper are for predicted
and observed bandshapes, which are the basis of comparison
of theory and experiment. Specific normal modes are not
resolvable, particularly since the 12C modes are all overlapped,
so their individual characteristics cannot be derived from
experiment. However, properties of selected decoupled modes
can be inferred from the 13C results. More details of the
computed mode structure are available in the Supporting
Information (Figures S1, S2).
The observed 12C-13C separations for the different label
patterns are qualitatively reflected in our model calculations,
but the details for specific modes under various labeling schemes
are not as well predicted. This may be a result of our use of an
idealized model. To address this, we have done DFT calculations
on structures derived from MD simulations and from NMR
experiments and have compared the predicted spectra based on
those structures with our ideal hairpin structure simulation as
previously reported (see Figure S1 in the Supporting Informa-
tion).69 Although the detailed mode distributions and magnitudes
obtained for simulations with these alternate structures are not
the same as for the ideal ones above, all models have a
qualitative consistency in terms of predicted mode distribution,
resulting in a 12C amide I′ maximum in the lower frequency
part of the band and an asymmetric shape. The key difference
is that the ideal structure (as simulated in Figure 2) has its most
intense mode (by a factor of 5) as one of the low-frequency
components of the exciton band, whereas the more distorted
structures result in a distribution of dipole strengths over the
exciton components that gradually fall in intensity from low to
high frequency (see Figure S1 in Supporting Information). This
more distributed intensity pattern leads to a broadened amide I′
absorption pattern in better agreement with experiment than seen
in Figure 2 for the idealized hairpin for the 12CdO band. As
for the isotope-labeled modes, all models have the same splitting
for V3K8 (big ring) and V3I10 (small ring) but vary substan-
tially in predicted coupling constant magnitude and yield less
consistent results for V5K8 (small ring), whose 13CdO modes
are dependent on the local conformation of the turn in each
structure (Figure S2, in Supporting Information).
Thermal Unfolding by IR. To study the relationship of
vibrational coupling and temperature-induced conformational
change, the temperature-dependent FTIR were measured for all
the peptides, and the amide I results for the double-labeled
hairpin peptides are shown graphically in Figure 3 as original
(left panels) and difference spectra (T - 5 °C) (right panels).
The thermal unfolding process is quite reversible (data not
9
13598 J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007

IR-Based Isotopic Labeling Technique
A R T I C L E S
shown, for single-labeled hairpins see Figure S3 in Supporting
Information). The main features of these spectra are (1) the
frequency shift and small intensity loss of the main 12C amide
I′ peak and (2) the significant intensity change of the 13C peak.
While the low-temperature spectra varied sharply, depending
on label positions as described above, at high temperature the
amide IR of V3K8, V3I10, and V5K8 are quite similar to each
other. As temperature increases, the peptide unfolds from a more
compact structure (â-hairpin) toward a less ordered structure
(partial random coil), as evidenced by the upward frequency
shift of the main 12C amide I′ peak for all peptides (Figure 3,
left panels). In the difference spectra this frequency shift appears
as a decrease of the intensity at ∼1630 cm-1 (negative
difference) and an increase at ∼1660 cm-1 (positive difference)
(Figure 3, right panels), and the apparent band components have
increased separation.
While the main 12C amide I′ peak has similar behavior for
all the labeled peptides, the 13CdO band has a more variable
thermal behavior. Of the three double-labeled peptides, V3K8
has a single 13C peak whose intensity change with temperature
dominates the difference spectrum. V3I10 has two new features
arising from the 13C labeling, one resolved, 1607 cm-1, and
the other a shoulder, ∼1627 cm-1. While the lower one is
certainly a 13CdO band, the higher one may be mostly 12CdO
in origin, reflecting the disruption of the exciton band by the
13C label. This is supported by the similar, though weaker,
pattern seen in the I10 spectrum. Despite the presumably
different mixing with the 12C modes, these bands have similar
Tm values. For V5K8, the 13C features are overlapped in the
difference spectra but are more evident in the normal spectra.
A new peak at ∼1608 cm-1, also evident in U and K5,
assignable to an Arg side-chain mode becomes resolved in going
from low to high temperature.56,93
Figure 4. 12C Amide I′ frequency shift with temperatures for U (black
[), V3 (red 4), V5 (pink 0), I10 (blue O), V3K8 (red 2), V3I10 (blue
b), and V5K8 (pink 9). Solid lines are the best two-state fits for the data.
Table 3. Thermodynamic Parameters for Two-State Model Fit of
Labeled Hairpin Unfolding
monitored by
amide I
frequency shift
monitored by
12C amide I
intensity change
monitored by
13C amide I
intensity change
Tm
(K)
∆
H
Tm
(K)
∆
H
Tm
(K)
∆
H
(kJ/mol)
(kJ/mol)
(kJ/mol)
The amide I′ peak frequencies of the labeled peptides for the
lowest experimental temperatures (T ) 5 °C) fall into two
groups (as shown in Figure 4, left axis). For V5 and V5K8, the
12CdO frequency is only ∼3-4 cm-1 higher than for the
unlabeled peptide (U), while for V3, V3K8, and V3I10, the
12C peak is about ∼8-9 cm-1 above U. The 12CdO peak for
I10 lies somewhat between the groups. These differences show
the effect of 13CdO substitution(s) on the exciton coupling of
12CdO’s. However, the temperature-dependence of the 12CdO
peak is consistent for all peptides, as seen from the overall
parallel behavior of their frequency shifts (Figure 4), which
suggests that the same process is operative. Since the amide I
frequencies for all the peptides do not converge even at 90 °C,
the peptide does not appear to have reached a high-temperature
steady state (disordered or unfolded form). This agrees with
the ECD thermal unfolding data (Figure 1) where the intensities
of the various bands changed but the spectral shapes did not.
The absence of an upper baseline for these curves may imply
less cooperativity in thermal unfolding; however, their obvious
curvature did make it possible to attempt a fit of the data with
a two-state model as an idealized, limiting test (Figure 4).
U
358 ( 16
337 ( 6
363 ( 13
354 ( 8
8.3 ( 1.9 353 ( 5
9.8 ( 2.0 352 ( 12
9.6 ( 1.6 362 ( 6
9.0 ( 1.1 361 ( 12
9.3 ( 1.5 350 ( 6
10.3 ( 2.2 347 ( 7
13.3 ( 1.5
8.3 ( 1.5 335 ( 4
12.6 ( 1.1
8.4 ( 1.2 331 ( 2
12.7 ( 1.9 344 ( 3
9.6 ( 1.4 343 ( 4
18.1 ( 2.6 339 ( 11
N/A
N/A
8.7 ( 1.1
a
11.2 ( 1.1
7.6 ( 0.5
10.7 ( 1.1
4.1 ( 0.9
V3
V5
I10
a
V3K8 349 ( 8
V3I10 340 ( 7
V5K8 362 ( 11 11.0 ( 1.8 352 ( 5
a Unrealistic fitting results were obtained.
By contrast, 13C peak frequency shifts (data not shown) for
V3I10 and V3K8 are of much less magnitude (∼1 cm-1 shift)
than for the main 12C peak and do not yield a useful measure
of the transition. V5K8 is different; the frequency does shift,
and the mode presumably changed character. The change has
less contribution from cross-strand coupling due to the distortion
of the H-bond next to the turn. However, their intensity changes
are directly a consequence of the cross-strand coupling, which
is lost on unfolding. Thus, we attempted to fit the apparent 13Cd
O intensities as obtained from the original amide I IR data (e.g.,
Figure 3 and Figure S3 in Supporting Information). There was
no advantage to use of the difference IR data for this analysis.
The results are shown in Figure 5, and the analysis is
summarized in Table 3.
From results of the fit, V5K8 would have a higher Tm than
V3K8 and V3I10 (Table 3). The 12C spectra have different
contributions from turn and strand residues, depending on the
positions of the labeled residues. Thus, the V5K8 12C result
will have more strand contribution, and its higher Tm suggests
that the turn has less stability than the â-strands of this peptide.
Variations in 12C peak and 13C peak intensities (Figure 5)
provide alternate probes of the thermal unfolding process, in
which the V3K8 and V3I10 13CdO-labeled modes gain intensity
by cross-strand coupling. Loss of intensity implies loss of order
or increase in distance between residues on opposing strands.
9
J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007 13599

A R T I C L E S
Huang et al.
Figure 6. MD trajectory results showing H-bond distance fluctuation for
a largely folded conformation at 300 K. Variations in the five H-bonds
shown in Scheme 1 are indicated as: V5CdO-K8N-H H-bond (red solid),
K8CdO-V5N-H H-bond (green solid), V3CdO-I10N-H H-bond (blue solid),
I10CdO-V3N-H H-bond (pink dotted), and R1CdO-Q12N-H H-bond (cyan
dotted). The R1CdO-Q12N-H H-bond fluctuation indicates the unfolding
of the terminal residues, and the red line shows the V5CdO-K8N-H H-bond
next to the turn also fluctuates. Sequential opening of the H-bonds on
unfolding is evident in this particular trajectory.
The stability of these Aib-Gly â-hairpins is comparable to
D
that of peptide models based on the Pro-Gly turn sequence.34
Both hairpins have very broad unfolding transitions that remain
incomplete at even 95 °C, confirming that some residual struc-
ture remains. Our simulations have shown these to occur in
different sequential steps of which “unzipping” by sequentially
losing H-bonds from the termini is one possibility (Figure 6).
Classical force field-based molecular dynamics (MD) simula-
tions have been carried out by several labs to develop residue-
level detailed understanding of â-hairpin formation.16,18,19,29,97-103
The results varied, depending upon force fields, solvent model,
simulation protocols, and the peptide system studied. However,
the trajectories obtained and their analyses have increased our
physical insights into â-hairpin structure in various compact
states as well as the corresponding folding mechanisms. In
addition, the MD simulations provide a basis for incorporating
some effects of structural fluctuations69 into quantum mechanical
calculations of the vibrational spectra. Such combined ap-
proaches can provide more realistic simulations and therefore
improve the interpretation of the experimental vibrational
spectra. Our approach utilizes MD simulations, in particular,
to sample local structural heterogeneity and conformational
fluctuation but does not yet incorporate dynamics directly. The
MD details have been reported with regard to spectral simula-
tion69 and will be communicated further regarding structural
variation in separate publications (Kim and Keiderling, manu-
script to be published).
Figure 5. Two-state fit for (a) 12C peak intensity change and (b) 13C peak
intensity change. Symbols are represented in the same way as in Figure 4.
Solid lines are the best fit.
Although V3K8 has the largest 13C peak intensity change over
this temperature range, the Tm values derived from analysis of
the V3K8 and V3I10 13CdO intensity changes are very similar,
and both are higher than the Tm from V5K8. Thus, the 13C and
12C results are consistent with the turn region having reduced
stability, presumably due to distortion or strain in the hairpin,
since V5K8 is next to the turn.
Discussion
Aib-Gly Is a Stable Turn-Promoting Hairpin Formation.
For this study Aib-Gly was substituted for DPro-Gly in the
original Gellman 12-residue â-hairpin peptide sequence95,96 to
avoid interference with the isotope-shifted IR band.34 The IR
spectra of the unlabeled peptide (U) confirm the ability of Aib-
Gly to promote a turn, which then enables cross-strand H-
bonding and leads to formation of a â-hairpin conformation as
previously reported using X-ray crystallography,73 and as now
confirmed by NMR structural analysis.88 The alignment of the
hairpin strands is further confirmed by the 13CdO labeling
patterns and their close agreement with theory for V3K8 and
V3I10. However, even the low-temperature folded state has
considerable distortion from regularity in the terminal residues
and at the turn as demonstrated by our MD results for folded
structures (see Figure S4 in Supporting Information) and by
the lack of well-defined cross-strand coupling for V5K8.
Our MD simulations for this peptide show the turn region to
remain somewhat compact (avoiding fully extended structures)
(97) Bolhuis, P. G. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 12129-12134.
(98) Garcia, A. E.; Sanbonmatsu, K. Y. Proteins 2001, 42, 345-354.
(99) Klimov, D. K.; Thirumalai, D. Proc. Natl. Acad. Sci. U.S.A. 2000, 97,
2544-2549.
(100) Ma, B.; Nussinov, R. Protein Sci. 2003, 12, 1882-1893.
(101) Tsai, J.; Levitt, M. Biophys. Chem. 2002, 101-102, 187-201.
(102) Zhang, J.; Qin, M.; Wang, W. Proteins 2006, 62, 672-685.
(103) Zhou, R.; Berne, B. J. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 12777-
12782.
(95) Haque, T. S.; Gellman, S. H. J. Am. Chem. Soc. 1997, 119, 2303-2304.
(96) Syud, F. A.; Espinosa, J. F.; Gellman, S. H. J. Am. Chem. Soc. 1999, 121,
11577-11578.
9
13600 J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007

IR-Based Isotopic Labeling Technique
A R T I C L E S
although quite variable in detailed conformation even for highly
folded structures. For short, lower-temperature trajectories (such
as seen in Figure 6, and as expanded in S4, Supporting
Information), the folded hairpin contains all the expected
H-bonds. The terminal bond and the turn H-bond show much
more fluctuation, as was noted above, until the hairpin unfolds.
structure (derived from PDB: 1IFC)67 but incorporated a small
shift of the strands relative to each other, the distance between
the oxygen atoms of the two labeled carbonyls were about the
same for the big and small rings, and the resultant splitting was
almost the same as reported previously.67 However, in both
cases, and in fact for all antiparallel models, opposite sign
patterns are obtained for the coupling of CdO pairs in small
(+) and large (-) H-bonded rings, even if only TDC is used
for the computational model.68,69 For the structures derived from
MD or NMR the frequency shift patterns are the same, but the
predicted intensity of the two components of the 13CdO band
are more similar than in the ideal case, which has maximal
intensity difference of in- and out-of-phase 13CdO modes due
to its local symmetry that derives from the two triamide-strand
basis used to construct the idealized model and transfer the
vibrational tensor parameters,68,82 as opposed to the fragmenta-
tion method used for the NMR and MD structures.69,78 We have
shown separately that the distribution of coupling constants
(from a TDC model) for an ensemble of compact partially folded
states had distinctly different peak values and dispersions for
the three double-labeled sequences.68,69 These dynamically
sensitive results again fit the differential spectra observed
between V3K8 (less 13C shift, higher intensity) and V3I10 (more
13C shift, lower intensity) that we saw (Figure 2) for the idealized
hairpins, which directly results from their having oppositely
signed 13CdO-13CdO coupling. The phase difference seen in
these two coupling patterns is critical, since it leads to the
frequency difference observed in all models, which is confirmed
by experiment. This characteristic isotopic shift additionally
confirms the proper antiparallel alignment of the strands by use
of IR data alone.
In experiments, we did see a strong 13C peak in V3K8 that
is higher in frequency than for V3I10, but we saw more complex
spectral features in the latter. The “extra” band in V3I10 (at
∼1627 cm-1) probably arises from a breakdown of the 12CdO
exciton coupling by 13CdO substitution in the middle of a strand
(especially the I10 position). One might ascribe it to nonde-
generate diagonal FF contributions due to the location of
13CdO oscillators, since I10 probably becomes involved with
the frayed N- and C-termini (as suggested by MD simulation,
which showed that the outer two H-bonds in this hairpin are
not stable69).
Our simulation also predicted that the two 13C-labeled
carbonyls forming a small ring near the â-turn (V5K8) are
nondegenerate because the two carbonyls are easily distorted
away from an orientation resulting in a good coupling.
Consequently, the frequencies computed for these two modes
are essentially the same as those of the individual single
carbonyls when labeled. The validity of this prediction was
confirmed by the weak shoulder seen for the 13CdO contribution
in V5K8, which is much the same as for V5. By extension, the
failure to see a clear coupled 13CdO band in the V5K8 labeled
hairpin confirms distortion of the turn region.
By contrast, higher temperature trajectories lead to “fully
unfolded” conformers which have no cross-strand H-bonds.69
Such MD trajectories begun from an unfolded hairpin (for
example Figure S5 in Supporting Information) occasionally
show refolding to a partially formed structure with frayed termini
and formation of a type I′ turn. However even when unfolded,
as seen in the bottom of Figure S5 in Supporting Information,
the φ,ψ angles of the Aib indicate that its conformation is mostly
contained in this part of Ramachadran space (φ ≈ 60°, ψ ≈
30°) with little correlation to the overall folding as represented
by the H-bond structure. Although the turn region is compact,
considerable fluctuation remains which leads to multiple
deformations of the V5CdO-K8N-H H-bond since the geometries
of these residues fluctuate widely.
Our simulations regularly show that the H-bond closest to
the turn separates and the residues close to the turn to have a
wider variance than those in the center of the antiparallel strands.
This is best seen in the H-bond distance distribution plots which
show the H-bond from K8N-H to V5CdO to have a broad
distribution and peak at ∼0.33 nm, while those in the center of
the hairpin are narrow and peak at ∼0.29 nm for T ) 300 K
(Figure S6 in Supporting Information).69 The result of this
fluctuation is a reduced average amide 13CdO coupling for the
labeled positions in V5K8. The trajectories we have simulated
imply the central residues of the hairpin strand region have
relatively high uniformity, in comparison to the fraying of the
termini and twists and the fluctuations of the turn even at
moderate temperatures. Consequently the V3I10 and V3K8
labeled peptides have strongly coupled 13CdO groups resulting
in well-defined and intense isotopic shifted bands.
Isotope Editing Probes Strength of Vibrational Couplings.
From simulation results, different isotopic labeling patterns
showed distinct spectral effects. In our model peptide, there are
two kinds of double 13C-labeling patterns: 14-member H-bond
ring (big ring) and 10-member H-bond ring (small ring), as seen
in Scheme 1. The labeling pattern resulting in a big ring gives
a
13C band with higher intensity and a higher frequency (less
separation from the 12C band, e.g., V3K8), while the alternate
labeling pattern resulting in a small ring gives a 13C band with
much lower intensity at a lower frequency (more separation from
12CdO, e.g., V3I10). That 12C modes mix at a low level with
13C modes is suggested by the strong 13C peak intensity seen
for V3K8, since the relative peak height is much larger than
the 13C:12C ratio (2:10) would suggest. By contrast, there appears
to be much less 13C:12C mode mixing for V3I10, which may
result from the intense component of the coupled isotope modes
being separated more from the main 12CdO peak.
The secondary effect of substituting 13C isotopes on the
12CdO peak frequency can also yield site-specific conforma-
tional information. We have done some calculations of spectral
patterns expected for structures derived from MD trajectories
that had either all five H-bonds formed or only two or three
H-bonds formed in the hairpin structure. For single-labeled
peptides, V3 and I10 are located in the midst of the remaining
In model â-sheet calculations,68 the predicted coupling of
13CdO for a large ring was less than for a small ring which is
reasonable because the distance between the two 13C-labeled
carbonyls is normally less in the small ring. Transition dipole
coupling (TDC) in particular, varies inversely with distance
(∼R-3), thus making the small ring coupling stronger. Since
our idealized â-hairpin model was based on a regular â-sheet
9
J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007 13601

A R T I C L E S
Huang et al.
extended â-strand region; however, I10 is closer to the frayed
termini (in the case of two or three remaining H-bonds in the
hairpin), and thus I10 should have less effect on the 12CdO
exciton coupling than V3. This suggests that I10 would have
an amide I′ frequency closer to that of the unlabeled peptide
than would V3 (much as seen experimentally, Table 2). Since
V5 is located near the turn, it would have much less impact on
the interstrand coupling, and thus its 12CdO should be even
more like U (again as seen experimentally). Overall, the order
of breaking 12CdO exciton couplings due to single 13CdO
substitution is V3 > I10 > V5. Considering the double-label
cases, we see a shift up in frequency for the apparent 12CdO
band that mimics experiment for the more fully folded of the
MD structures modeled, but the differences in the patterns are
small between various structures. Our experimental data suggest
that the low-temperature vibrational coupling in our peptide is
dependent on the unique strand geometry that represents an
energy minimum which may be best reflected by a combination
(ensemble) of variably distorted (partially unfolded) models
(ideal â-hairpin and frayed â-hairpin).
Isotope-Editing Reveals Site-Specific Stability. Optical
spectroscopy can normally only monitor global structural
changes in proteins and peptides. With isotopic labeling, one
could in principle differentiate local conformational change at
a residue level (site specific), which could greatly improve the
understanding of protein-folding mechanisms and local structure.
Here we have used the labels to discriminate turn and strand
stability, being “regiospecific” through cross-strand coupling
rather than strictly site specific.
directly affected by 13C substitution. The 13CdO interruption
of the 12CdO exciton is smaller when labeled at V5K8; thus,
the 12CdO band in that case samples more of the antiparallel
strand and the Tm derived is higher than for V3K8 and V3I10
(Table 3). Thus, the 12CdO thermal changes in the latter cases
tend to reflect the contributions of the turn and terminal groups
more than for the V5K8 or for U, the unlabeled case. On the
other hand, the 13CdO peak intensity change reports on the
temperature variation of coupling of the labeled positions, thus
giving more direct site-specific stability information. The 13C
peak intensity change for V3K8 with increased temperature
demonstrates that the vibrational coupling is significantly
affected. There is strong vibrational coupling between the two
13CdO modes at low temperature (when the peptide is in a
â-hairpin conformation), and this is lost as the distance of the
two 13CdO groups is increased and/or the orientation is
randomized when the peptide unfolds.
On the contrary, if the vibrational coupling were weak, the
effect of temperature would be much smaller. Since the peptide
contains labels that act as dipolar oscillators in both folded and
unfolded states, the relevant change that accompanies unfolding
is loss of cross-strand coupling. This is what we see in Figure
3. V3I10 has two labels at positions that could be influenced
by fraying of the ends, and V5K8 has two labels at positions
close to the turn, whereas V3K8 is labeled in the antiparallel
strand. This results in differences between the low-temperature
spectra, even for the small-ring labeling in V5K8 and V3I10.
As temperature increases, the peptide unfolds so that all positions
become disordered, and there is little differentiation between
labeling sites; each is locally disordered, and there is no
significant cross-strand coupling. Thus, the final, high-temper-
ature spectra of all these peptides are very similar both for the
13CdO and 12CdO peaks, demonstrating unfolding of the
strands at high temperature. The 13CdO intensity changes have
roughly similar Tm for the three double-labeled hairpins which
suggests that these local labels sense the antiparallel strands
and, for this example, do not differentiate between one end or
the other.
Generally, a frequency shift of the amide I′ IR is expected
when proteins/peptides undergo a conformational change be-
cause different secondary structural components have charac-
teristic amide I frequencies. The low-temperature unlabeled
â-hairpin peptide 12C amide I′ peak at ∼1633 cm-1 shifts
gradually to ∼1643 cm-1 as temperature is increased, which
indicates a more dominant disordered conformation. Some of
this frequency shift must result from the change in solvation of
various sites changing environments from the folded peptide,
with cross-strand H-bonded residues, to the unfolded variant,
with primarily solvent H-bonded residues. Solvent shifts of
peptide IR frequencies have been well studied, and we have
explicitly modeled spectral changes in sheet structures, showing
that the exposed CdO groups on the edge will have the largest
changes.34,68,78 Those residues with H-bonds to other strands
will have less spectral shift upon interacting with solvent
molecules. The edge residues have variable access that is a
function of sequence, and these all contribute to the 12C band,
resulting in its width and part of its frequency shift on unfolding
that reflects more global (solvation, in this case) effects.
However, in our models, the 13C-labeled residues are all cross-
strand H-bonded at low temperature and thus have much less
solvent effect. Thus, they have the potential to provide local
probes of structure with less global impact.
Conclusions
We have systematically studied a series of â-hairpin peptides
with isotope-edited IR spectroscopy. The peptide with Aib-Gly
in the sequence as a turn-initiating segment shows very stable
â-hairpin conformation over a wide temperature range. Different
isotopically labeled variants of the â-hairpin peptide show
different patterns of vibrational coupling, which is affected by
the geometry of the peptide. Thermal unfolding experiments
reveal that the vibrational coupling is sensitive to temperature.
13C substitutions also affect residual 12C couplings as monitored
through frequency shifts and intensity variations. These effects
are correctly predicted by theoretical DFT-based simulations
for an idealized â-hairpin. In particular, the sign variation of
the vibrational coupling between the cross-strand amides in the
small and large H-bonding rings produces significant frequency
shifts of the 13C-labeled bands which can be used to confirm
the registration of the antiparallel strands in the hairpin.
IR spectra show a difference in behavior for the 12CdO
frequency and 13CdO intensity with temperature and thus can
provide site-specific information about the effect of the unfold-
ing on vibrational coupling. The main amide I frequency change
with temperature for the labeled peptides is indicative of a
temperature effect on the 12CdO exciton coupling since this is
Acknowledgment. This research was originally supported by
a grant from the National Science Foundation (CHE03 16014
and CHE07 18543 to T.A.K.) and from the National Institutes
9
13602 J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007

IR-Based Isotopic Labeling Technique
A R T I C L E S
of Health (AG17983 to R.P.H.) and ACS PRF (42027-AC4 to
R.P.H.). Parts of this study were developed while T.A.K. was
a Fellow of the John Simon Guggenheim Foundation. We thank
Catherine L. Thomas and Martha Juban in the LSU Protein
Facility for expert technical assistance in peptide synthesis and
purification and Ling Wu at UIC for help in confirming spectral
data.
Supporting Information Available: Detailed results from MD
simulations, QM spectral calulations, and temperature-dependent
IR spectra for the unlabeled and single-labeled peptides. This
material is available free of charge via the Internet at
http://pubs.acs.org.
JA0736414
9
J. AM. CHEM. SOC. VOL. 129, NO. 44, 2007 13603