A photoinducible β-hairpin

Article abstract of DOI:10.1021/ja0442567

A photochromic azobenzene linker was incorporated as a turn element into an amino acid sequence known to fold into a β-hairpin structure in aqueous solution. Oligomer formation when the linker was in its thermodynamically favored trans form prohibited str

Full text of DOI:10.1021/ja0442567

Published on Web 02/12/2005
A Photoinducible â-Hairpin
Andreas Aemissegger,† Vincent Kra¨utler,‡ Wilfred F. van Gunsteren,‡ and
Donald Hilvert*,†
Contribution from the Laboratorium fu¨r Organische Chemie and Laboratorium fu¨r Physikalische
Chemie, Swiss Federal Institute of Technology, ETH-Ho¨nggerberg, CH-8093 Zu¨rich, Switzerland
Received September 21, 2004; E-mail: hilvert@org.chem.ethz.ch
Abstract: A photochromic azobenzene linker was incorporated as a turn element into an amino acid
sequence known to fold into a â-hairpin structure in aqueous solution. Oligomer formation when the linker
was in its thermodynamically favored trans form prohibited structure determination. Light-induced
conformational change of the linker to the cis form led to the formation of monomers which exhibited a
well-defined â-hairpin structure as determined by 1H NMR. The rate of the light-induced cis-to-trans
isomerization of the azobenzene-containing peptide was 30% slower compared to the unsubstituted
chromophore. These results suggest that suitably substituted azobenzenes can be used as photoinducible
turn elements to investigate and control the folding and stability of â-sheets.
â-Hairpins are supersecondary structural elements consisting
of two adjacent antiparallel peptide strands connected by a short
loop. Although such structures often aggregate when removed
from their normal tertiary context, model sequences for â-hair-
pins have been designed over the past decade that fold
autonomously in aqueous solution.1-4 These systems have
provided valuable insight into the factors that influence the
folding and stability of â-sheet structures, one of the major
architectural elements in proteins. They are also useful for
probing protein-protein, protein-RNA, and drug-receptor
interactions.
peptidescontaining the enantiomeric LPro-Gly are disordered.9,10
By inserting multiple turn elements into peptides, it has even
been possible to construct nonaggregating three- and four-
stranded structures.11,12
Nonpeptidic elements can also induce antiparallel pairing of
peptide strands in water. Successful turn mimics include a
variety of conformationally restricted templates, such as diben-
zofurans, biphenyls and diphenylacetylenes.13-16 Photochromic
compounds, which undergo large conformational changes when
exposed to light of appropriate wavelength, could also be
interesting in this context, since they might permit reversible
conformational control of hairpin formation.
The overall stability of â-hairpins is determined by multiple
factors. These include interstrand hydrogen bonding, secondary
structure propensities of residues in the strands, and side chain-
side chain interactions. The loop residues, in particular, have
been found to play a crucial role in hairpin folding. For example,
the dipeptide LAsn-Gly, which shows the highest statistical
correlation with type I′ â-turns in crystalline structures,5 is an
effective inducer of hairpin structure in short peptides.6,7
Incorporation of D amino acids at the i + 1 residue of the turn
is especially beneficial because they intrinsically favor the
normal right-handed twist of â-sheets. In fact, the dipeptide
DPro-Gly has been shown to be superior to LAsn-Gly in
nucleating hairpin formation in aqueous solution,8 whereas
Of the many photochromic systems that have been described,
azobenzenes are potentially well suited for this type of applica-
tion. They have been extensively characterized, are readily
synthesized, and give rise to large changes in the distance
between moieties appended to the aryl rings upon interconver-
sion of the trans and thermodynamically less favored cis isomers
(Figure 1). Unsubstituted trans-azobenzene exhibits three typical
absorption bands at 440 nm (nfπ*), 314 nm (πfπ*), and 230-
240 nm (σfσ*);17 substituted compounds have similar spectra
with slightly shifted absorption bands (Figure 2). Irradiation at
the wavelength of the πfπ* transition converts the trans to
the cis isomer, while the reverse process can be achieved either
thermally or by irradiation at the wavelength of the nfπ*
transition.
† Laboratorium fu¨r Organische Chemie.
‡ Laboratorium fu¨r Physikalische Chemie.
(1) Blanco, F.; Ramirez-Alvarado, M.; Serrano, L. Curr. Opin. Struct. Biol.
1998, 8, 107-111.
(9) Haque, T. S.; Gellman, S. H. J. Am. Chem. Soc. 1997, 119, 2303-2304.
(10) Espinosa, J. F.; Gellman, S. H. Angew. Chem., Int. Ed. 2000, 39, 2330-
2333.
(11) Das, C.; Raghothama, S.; Balaram, P. Chem. Commun. 1999, 967-968.
(12) Schenck, H. L.; Gellman, S. H. J. Am. Chem. Soc. 1998, 120, 4869-4870.
(13) Diaz, H.; Espina, J. R.; Kelly, J. W. J. Am. Chem. Soc. 1992, 114, 8316-
8318.
(14) Diaz, H.; Kelly, J. W. Tetrahedron Lett. 1991, 32, 5725-5728.
(15) Diaz, H.; Tsang, K. Y.; Choo, D.; Kelly, J. W. Tetrahedron 1993, 49, 3533-
3545.
(2) Gellman, S. H. Curr. Opin. Chem. Biol. 1998, 2, 717-725.
(3) Searle, M. S. J. Chem. Soc., Perkin Trans. 2 2001, 1011-1020.
(4) Cochran, A. G.; Skelton, N. J.; Starovasnik, M. A. Proc. Natl. Acad. Sci.
U.S.A. 2001, 98, 5578-5583.
(5) Hutchinson, E. G.; Thornton, J. M. Protein Sci. 1994, 3, 2207-2216.
(6) de Alba, E.; Jimenez, M. A.; Rico, M. J. Am. Chem. Soc. 1997, 119, 175-
183.
(7) Ramirez-Alvarado, M.; Blanco, F. J.; Serrano, L. Nat. Struct. Biol. 1996,
3, 604-612.
(8) Stanger, H. E.; Gellman, S. H. J. Am. Chem. Soc. 1998, 120, 4236-4237.
(16) Kemp, D. S.; Li, Z. Q. Tetrahedron Lett. 1995, 36, 4175-4178.
(17) Griffiths, J. Chem. Soc. ReV. 1972, 1, 481-493.
9
10.1021/ja0442567 CCC: $30.25 © 2005 American Chemical Society
J. AM. CHEM. SOC. 2005, 127, 2929-2936
2929

A R T I C L E S
Aemissegger et al.
Scheme 1 a
Figure 1. Cis-trans isomerization of azobenzene.
a (A) Model hairpin sequences with turn element X. (B and C) Possible
hydrogen bonding patterns for the â-hairpin. Structural data for peptide 1a
is consistent with the pattern shown in B, whereas the azobenzene-containing
peptide cis-1b is better described by pattern C (see text).
Figure 2. Absorption spectra for 1b in the thermodynamic (‚‚‚) and
photostationary (s) state. The maxima for the πfπ* and nfπ* transitions
are at 314 and 426 nm, respectively.
the turn element in its cis configuration were anticipated to favor
a well-defined hairpin structure.
These properties have already been extensively exploited to
photoregulate the conformation of crown ethers18 and cyclo-
dextrins19 and the permeability of membranes.20,21 Azobenzene
chromophores have also been introduced into peptides and
proteins, both in side chains22-25 and in the backbone.26-28 The
ability of an azobenzene to template a â-turn conformation in
a cyclic peptide when in its cis form is particularly notable in
the present context.26 Here we show that azobenzenes can also
serve as photoswitchable â-turn mimics in open chain polypep-
tides, providing excellent conformational control in aqueous
solution.
Azobenzene units which have been previously incorporated
into peptide backbones generally contain p-substituted aryl
rings.26-28 However, visual inspection of models suggested that
this substitution pattern might not be optimal for a turn mimetic.
All possible symmetrical substitution patterns (Figure 1) were
therefore investigated by molecular dynamics (MD) simulation
with respect to their ability to template hairpin formation.29
Additionally, one and two methylene groups were allowed as
spacers between the functional groups and the aryl rings to
accommodate interstrand interactions between the appended
peptides.29
Results
The MD simulations indicated that amino acid 2b, with meta-
substituted aryl rings and a single methylene group spacer, was
one of the most promising turn mimetics. Because the substit-
uents are not in conjugation with the central NdN bond, thermal
cis-trans isomerization of meta-substituted azobenzenes should
be substantially slower than for para-substituted azobenzenes.30
This configurational stability was expected to be advantageous
for structural investigations by NMR spectroscopy.
Peptide Design. The 12-residue peptide 1a (Scheme 1)
derived from protein GB1 has been shown by Gellman and co-
workers to adopt a hairpin structure in aqueous solution.10 The
DPro-Gly segment at positions 6 and 7 nucleates hairpin
formation and was targeted for replacement by an azobenzene-
containing turn mimetic. Favorable interactions between the
hydrophobic cluster formed by the aromatic side chains and
Synthesis. As described for similar compounds,31 the syn-
thesis of 10 relies on the reaction of a nitrosobenzene 5 with an
aniline 8 (Scheme 2). Compound 5 was prepared by esterifi-
cation of commercially available m-nitrophenylacetic acid 3
under Mukaiyama conditions,32 followed by reduction of the
nitro group to the hydroxylamine with Zn dust in NH4Cl solution
and immediate reoxidation with FeCl3 to the nitroso compound.
(18) Shinkai, S.; Manabe, O. Top. Curr. Chem. 1984, 121, 67-104.
(19) Ueno, A.; Tomita, Y.; Osa, T. Tetrahedron Lett. 1983, 24, 5245-5248.
(20) Kano, K.; Tanaka, Y.; Ogawa, T.; Shimomura, M.; Okahata, Y.; Kunitake,
T. Chem. Lett. 1980, 421-424.
(21) Kano, K.; Tanaka, Y.; Ogawa, T.; Shimomura, M.; Kunitake, T. Photochem.
Photobiol. 1981, 34, 323-329.
(22) Pieroni, O.; Houben, J. L.; Fissi, A.; Costantino, P. J. Am. Chem. Soc.
1980, 102, 5913-5915.
(23) Willner, I.; Rubin, S. React. Polym. 1993, 21, 177-186.
(24) Liu, D.; Karanicolas, J.; Yu, C.; Zhang, Z. H.; Woolley, G. A. Bioorg.
Med. Chem. Lett. 1997, 7, 2677-2680.
(25) Kumita, J. R.; Smart, O. S.; Woolley, G. A. Proc. Natl. Acad. Sci. U.S.A.
2000, 97, 3803-3808.
(29) Kra¨utler, V.; Aemissegger, A.; Hu¨nenberger, P. H.; Hilvert, D.; Hansson,
T.; van Gunsteren, W. F. J. Am. Chem. Soc., in press.
(30) Asanuma, H.; Liang, X. G.; Komiyama, M. Tetrahedron Lett. 2000, 41,
1055-1058.
(26) Ulysse, L.; Cubillos, J.; Chmielewski, J. J. Am. Chem. Soc. 1995, 117,
8466-8467.
(27) Renner, C.; Behrendt, R.; Sporlein, S.; Wachtveitl, J.; Moroder, L.
Biopolymers 2000, 54, 489-500.
(31) Ulysse, L.; Chmielewski, J. Bioorg. Med. Chem. Lett. 1994, 4, 2145-
2146.
(28) Renner, C.; Cramer, J.; Behrendt, R.; Moroder, L. Biopolymers 2000, 54,
501-514.
(32) Saigo, K.; Usui, M.; Kikuchi, K.; Shimada, E.; Mukaiyama, T. Bull. Chem.
Soc. Jpn. 1977, 50, 1863-1866.
9
2930 J. AM. CHEM. SOC. VOL. 127, NO. 9, 2005

A Photoinducible
â
-Hairpin
A R T I C L E S
Scheme 2 a
state. Irradiation of the peptide in a Rayonet photoreactor
equipped with RPR-3500 lamps and separation of the two
isomers by preparative HPLC afforded the cis isomer in greater
than 98% purity.
Analytical ultracentrifugation at 4 °C shows the cis isomer
to be monomeric with an apparent molecular weight of 1688
( 48 Da, in good agreement with the calculated molecular
weight of 1604 Da considering the error introduced by modeling
the azobenzene unit as two phenylalanines in the calculation of
the partial specific volume of the peptide. When the temperature
of the rotor was raised to 40 °C, the absorption intensity
gradually decreased (data not shown) due to precipitation of
the trans isomer that is formed upon thermal isomerization of
the sample.
The rate of the cis to trans isomerization of peptide 1b was
determined by monitoring the increase in the πfπ* absorption
band at 323 nm. To assess the influence of the peptide on the
isomerization, compound 2b was studied under identical condi-
tions. The latter was prepared by treating 10 with 20% piperidine
in DMF for 1 h and purification by analytical HPLC. Irradiation
in the Rayonet photoreactor gave the cis form of the water
soluble amino acid in >90% purity as judged by analytical
HPLC. Because the dark reaction of azobenzenes is strongly
dependent on the pH,33-38 the reactions were carried out in 20
mM acetate buffer at pH 3.8. In the absence of light at 25 °C,
peptide 1b and amino acid 2b thermally isomerize with a rate
constant of about 5 × 10-7 s-1, corresponding to a half-life of
>2 weeks. Photoisomerizations were carried out at 25 °C at
431 and 420 nm for the cis peptide and 2b, respectively. A rate
constant of (1.5 ( 0.1) × 10-4 s-1 was obtained for the peptide,
which can be compared to a value of (2.2 ( 0.5) × 10-4 s-1
for the amino acid alone.
NMR Experiments. The trans form of peptide 1b forms a
highly viscous solution at 4 °C at a concentration of 9 mg/mL
(5.6 mM). The NMR spectrum of this sample shows extensive
signal broadening (Figure 3A), most likely due to extensive
aggregation of the peptide. Four-fold dilution of the sample and
increasing the temperature to 25 °C allowed a reasonably well-
resolved 1D proton spectrum to be recorded. However, signifi-
cant noise in the 2D spectra and the presence of 10% of the
peptide containing the cis form of the linker precluded assign-
ment of peaks to residues and determination of a three-
dimensional structure.
In contrast, the cis form of the peptide dissolves readily at a
concentration of 5.8 mg/mL (3.6 mM) and yielded spectra with
sharp and well resolved signals at 4 °C (Figure 3B). A full set
of DQF-COSY, TOCSY, and NOESY spectra was recorded.
UV data showed that thermal isomerization was negligible for
the duration of the NMR analysis (data not shown).
a Conditions: (a) tBuOH, 2-chloro-1-methylpyridinium iodide, Bu3N; (b)
NH4Cl, Zn, then FeCl3; (c) Fmoc-Cl, Hu¨nig’s base; (d) H2, PtO2; (e) AcOH;
(f) TFA, CH2Cl2
Aniline 8 was obtained from commercially available 3-nitroben-
zylamine hydrochloride 6 via a two-step sequence, involving
introduction of an Fmoc protecting group followed by reduction
of the nitro group. Although catalytic reduction by palladium
on charcoal leads to partial deprotection of the amine, 8 was
obtained in quantitative yield using PtO2 instead. Coupling of
5 and 8 in glacial acetic acid afforded the fully protected
azobenzene amino acid 9 in 57% yield. Cleavage of the tert-
butyl ester group with TFA in dichloromethane gave the turn
mimetic 10 in a form suitable for standard Fmoc solid-phase
peptide synthesis.
The synthesis of the azobenzene-containing peptide 1b was
carried out on a 0.1 mmol scale using standard conditions (20%
piperidine in DMF for Fmoc deprotection, HOBt/HBTU for
activation, diisopropylethylamine as base, and N-methylpyrro-
lidinone as solvent) on a Rink amide resin. The peptide was
cleaved from the resin with TFA/H2O/TIPS 93:5:2 and purified
by reversed-phase HPLC.
Resonance Assignment and Structure Determination.
Numerous NOE signals between residues that are not adjacent
in sequence are evident in the spectra of cis peptide 1b (see
Supporting Information for a full list of chemical shifts and NOE
(33) Sokalski, W. A.; Gora, R. W.; Bartkowiak, W.; Kobylinski, P.; Swora-
kowski, J.; Chyla, A.; Leszczynski, J. J. Chem. Phys. 2001, 114, 5504-
5508.
(34) Sanchez, A. M.; Barra, M.; de Rossi, R. H. J. Org. Chem. 1999, 64, 1604-
1609.
Biophysical Characterization. As judged by analytical
HPLC, 90% of peptide 1b has a trans configuration at
thermodynamic equilibrium; 10% adopts the cis configuration.
The cis content increases to about 85% in the photostationary
(35) Sanchez, A. M.; de Rossi, R. H. J. Org. Chem. 1995, 60, 2974-2976.
(36) Lovrien, R.; Pesheck, P.; Tisel, W. J. Am. Chem. Soc. 1974, 96, 244-248.
(37) Wettermark, G.; Langmuir, M. E.; Anderson, D. G. J. Am. Chem. Soc.
1965, 87, 476-481.
(38) Lovrien, R.; Waddington, J. C. J. Am. Chem. Soc. 1964, 86, 2315-2322.
9
J. AM. CHEM. SOC. VOL. 127, NO. 9, 2005 2931

A R T I C L E S
Aemissegger et al.
Figure 4. RMSD values for backbone (O) and side chain (0) atoms for
the structures obtained by simulated annealing with respect to their average
structure. (A) Originally calculated structure based on all NOE restraints,
including the two misassigned values. (B and C) Clusters 1 and 2 obtained
after removal of the erroneous NOEs.
Figure 3. 1H NMR spectra of peptide 1b in 1:9 D2O/H2O 50 mM sodium
acetate buffer pH 3.6, 4 °C. (A) trans-1b at 5.6 mM, (B) cis-1b at 3.6 mM
resonance assignments). Although multiple NOE signals involv-
ing the ring protons of the azobenzene and F8 residues could
not be assigned unambiguously, a total of 53 other NOE distance
restraints were available for structure calculations. These
included strong intra- and interstrand signals. Simulated an-
nealing was performed with the program CNS.39
A total of 100 structures was generated from the NMR data.
The superposition of the 10 best structures (as judged by the
lowest overall energies) indicates that the peptide forms a hairpin
structure which has considerable flexibility at its termini and
also in the turn region (Figure 4A). In addition, the C-terminal
strand appears to be somewhat more flexible than its N-terminal
counterpart. Explicit solvent MD simulations were used for
further refinement of the structure to accommodate solvent and
electrostatic effects.29 Analysis of the resulting structures (Figure
5) revealed that two NOE distance restraints are incompatible
with the calculated structures. Both of these violations involve
the amide proton of residue F8. Reexamination of the NOESY
peak assignments showed an ambiguity and likely misassign-
ment in one of these cases, which occurs at the intersection of
Figure 5. Cross-eyed stereo superposition of the 10 best structures
(backbone atoms only) of the primary cluster (87% of the trajectory)
obtained in the explicit solvent MD simulation of peptide cis-1b.
two regions of nearly isochronous nuclei. As numerous NOEs
are observed between the side chains of residues W2 and Y4,
and residue Y4 shows a relatively strong NOE to the amide
proton of residue F8, the other restraint in question is most likely
the result of spin diffusion. This conclusion is supported by the
observation that this NOE is only present in the 400 ms NOESY
spectrum.
In the originally generated structures the side chain of residue
T9 has an anomalously low RMSD value compared to the other
amino acids in the peptide, suggesting that it is strongly
constrained (Figure 4A). However, this result appears to be an
artifact that arose because the two problematic NOEs were
included in the calculations. When the refinement was repeated
without them, a new set of 100 structures was obtained which
are evenly divided into two clusters of hairpins (Figure 6), each
(39) Brunger, A. T.; Adams, P. D.; Clore, G. M.; DeLano, W. L.; Gros, P.;
Grosse-Kunstleve, R. W.; Jiang, J.-S.; Kuszewski, J.; Nilges, N.; Pannu,
N. S.; Read, R. J.; Rice, L. M.; Simonson, T.; Warren, G. L. Acta
Crystallogr. 1998, D54, 905-921.
9
2932 J. AM. CHEM. SOC. VOL. 127, NO. 9, 2005

A Photoinducible
â
-Hairpin
A R T I C L E S
better defined. Whereas the ensemble of 10 lowest energy
structures from the initial calculation had an average RMSD to
the mean coordinates of 0.66 ( 0.22 Å for the backbone atoms
of all residues except the linker and 1.65 ( 0.35 Å for all heavy
atoms, these values dropped to 0.55 ( 0.19 Å/0.43 ( 0.17 Å
for the backbone and 1.17 ( 0.28 Å/0.99 ( 0.26 Å for the side
chains of cluster 1/cluster 2. For the latter calculations, the N-
and C-terminal residues were excluded as they are heavily
disordered.
Discussion
The properties of peptide 1b establish the feasibility of using
azobenzene chromophores as photoswitches to control the
conformation of open chain polypeptides in aqueous solution.
In its cis configuration, the meta-substituted azobenzene mimics
the dipeptide DPro-Gly in nucleating a stable and monomeric
hairpin structure. In contrast, the trans configured peptide does
not adopt a unique structure in accord with MD simulations29
and tends to form higher order aggregates. Until the latter
ultimately precipitate from solution, the peptide can be reversibly
switched between the cis and trans states by irradiation at an
appropriate wavelength.
The structure deduced for cis-1b is an ensemble of hairpins
with a well-defined stem and the expected residue pairings. In
addition to the usual flexibility at the N- and C-termini, the
linker appears to be considerably more mobile than conventional
type I′ and II′ turns, which are strongly constrained by hydrogen
bonds between the first and fourth residues of the turn. Two
major, roughly isoenergetic conformations are evident that differ
with respect to the orientation of the azobenzene linker either
above or below the plane of the hairpin (Figure 6C). Addition-
ally, although steric clashes prevent the aromatic rings from
simultaneously occupying the plane of the cis configured N-N
double bond, rotation around the C-N bond allows them to
explore multiple conformational states. The methylene spacers
at the meta positions of the azobenzene appear to insulate the
peptide strands somewhat from the linker so that only the
immediately flanking residues are affected by its flexibility.
Thus, the backbone conformation of the peptide strands in an
overlay of the 10 best structures from each cluster is strikingly
similar, with only residue K7 showing a significant displacement
due to the different azobenzene conformations (Figure 6C). The
side chains of residues 2 to 4 and 8 to 10 are completely
unaffected by linker flexibility and maintain essentially identical
packing in the two primary clusters.
Aside from differences in the turn region, the hairpin
structures deduced for cis-1b qualitatively resemble those
reported for 1a. Nevertheless, there is one important difference.
In 1a, strong backbone-backbone NOEs were observed between
the HR protons of residues W2 and V11 (corresponding to W2
and V10 in peptide 1b) and Y4 and F9 (corresponding to Y4
and F8).10,40 These are missing in cis-1b. Instead, a strong NOE
between HR of residue Q3 and HN of residue T9 is observed.
This result indicates that the backbones of the N- and C-strand
segments are both flipped by ca. 180° so that the HR protons
which face the opposite strand in one peptide are directed toward
the surrounding solvent in the other (Scheme 1B versus 1C).
Figure 6. (A and B) Cross-eyed stereo superposition of the 10 best
structures (backbone atoms only) of the two clusters obtained by simulated
annealing of peptide cis-1b. (C) Superposition of the polypeptide backbone
for best fit of the backbone heavy atoms for residues 2-5 and 7-10. The
radius of the cylindrical rods representing the polypeptide chains is
proportional to the mean backbone displacement per residue among the 10
lowest energy conformers of each cluster. The two representations differ
by a 90° rotation about the vertical axis.
containing approximately 45% of all structures, plus a variety
of outliers. None of the outliers is among the 50 best (lowest
overall energy) structures. The two main clusters do not differ
significantly in terms of overall energy. The per-residue RMSDs
observed for the side chain of residue T9 in the new structures
are in the range of the adjacent amino acids (Figure 4B and
4C). Additionally, the orientation of the side chains is much
(40) Espinosa, J. F.; Syud, F. A.; Gellman, S. H. Protein Sci. 2002, 11, 1492-
1505.
9
J. AM. CHEM. SOC. VOL. 127, NO. 9, 2005 2933

A R T I C L E S
Aemissegger et al.
Although 1a and cis-1b exhibit different interstrand hydrogen
bonding patterns, the hydrophobic side chains of W2, Y4, F8,
and V10 cluster similarly in both. This hydrophobic core is
believed to contribute to the stability of 1a in aqueous solution10
and likely plays a similar role in the azobenzene-containing
peptide. Moreover, investigations of dibenzofuran- and biphenyl-
type turn templates have shown that hydrophobic interactions
between the â-turn mimic and the side chains of amino acids
on one face of the hairpin can further stabilize sheetlike
structures in solution.41-43 Analogous interactions are evident
between the four hydrophobic residues and the photochromic
linker in one of the structure clusters obtained in the simulated
annealing calculations. In the second cluster, the azobenzene
bends to the opposite face of the sheet toward the hydrophobic
side chain of V5.
to a stable â-hairpin that is consistent with all the experimental
data (Figure 5). In agreement with the simulated annealing
results, the peptide strands are packed differently than in the
hairpin formed by parent peptide 1a,10,40 favoring the hydrogen
bonding arrangement shown in Scheme 1C over that in Scheme
1B. In contrast to the in vacuo calculations, however, only a
single family of structures is observed. The azobenzene unit
appears to be highly mobile and folds toward the hydrophobic
residues W2, Y4, F8, and V10 as seen in one of the clusters
obtained by simulated annealing. In the explicit solvent MD
simulation, the hydrophobic core formed by W2, Y4, and V10
is more compact than in the structures obtained by simulated
annealing. Moreover, the unrestrained MD simulation places
F8 and the turn element in close contact, a finding that is
supported by the numerous NOE signals involving the turn
element and residue F8 that were observed but could not be
assigned unambiguously. The interaction between F8 and the
chromophore favors the right-handed twist commonly observed
for â-sheets embedded in proteins (Figure 5).47
In summary, we have prepared an Fmoc-protected azobenzene
amino acid 10 that is compatible with solid-phase synthesis and
can be readily incorporated into peptides. Our results show that
its insertion into the backbone of a short open chain polypeptide
enables reversible control of peptide conformation in solution.
Like the central dipeptide in a conventional â-turn, this
chromophore is able, in its cis configuration, to preorganize two
peptide strands as a â-hairpin. In the specific case examined,
the azobenzene is much floppier than a dipeptide turn element
and favors a different set of interactions between the attached
strands. Nevertheless, based on its ability to organize the side
chains of a polypeptide spatially, this turn mimic could prove
useful as a photoregulatory element to control â-hairpin
formation in peptide hormones or larger proteins, including
enzymes.
NOE intensities were previously used to estimate that 50 to
75% of peptide 1a adopts the â-hairpin fold at 3 °C.10 The extent
of â-hairpin folding in the azobenzene-containing cis-1b cannot
be reliably estimated using the same NOE-based analysis
because it assumes a two state model in which the peptide is
either folded, so that interproton distances are well defined and
correlate with signal intensity, or unfolded, in which case the
protons are too far apart to give rise to an NOE. This two state
assumption is questionable if a preorganizing turn element forces
the two short peptide strands into a specific orientation, as is
the case here. In principle, HR chemical shifts offer an
alternative method of quantifying populations, as they are highly
sensitive to local secondary structure.44-46 However, reference
values are required for the fully folded and unfolded states, as
equilibration is rapid on the NMR time scale. Unfortunately,
the unfolded state of 1b, as modeled by the trans form, produces
aggregates, precluding collection of relevant NMR data. Al-
though we have not been able to reliably quantify the fraction
of the cis-1b peptide populating the â-hairpin, the popula-
tion of compact structures is likely to be comparable to that
observed for 1a based on the quality of the spectra, which show
no other species, the absence of aggregation even after several
weeks, and the preorganizing nature of the cis-configured
chromophore.
Materials and Methods
Materials. All commercially available reagents and solvents were
used as received. Organic extracts were dried over magnesium sulfate,
filtered, and concentrated using a rotary evaporator. Nonvolatile oils
and solids were dried under vacuum at 0.01 Torr. Reactions were
monitored by thin-layer chromatography on precoated E. Merck silica
gel 60 F254 plates (0.25 mm) and visualized either by short wave UV
or by staining with 10% ethanolic phosphomolybdic acid. Flash
chromatography was performed on Fluka silica gel 60. Chemical shifts
are expressed in ppm relative to tetramethylsilane or the solvent signal.
All melting points are uncorrected.
Unrestrained MD simulations with a physically meaningful
force field and explicit solvent were employed to provide a more
realistic assessment of the solution structure of the peptide.29
The conventional in vacuo simulated annealing process, which
was utilized to determine the structures described above, only
considers bonded and repulsive van der Waals interactions.
While this method predicts a well-defined hairpin structure for
cis-1b, it neglects electrostatic, hydrophobic, and solvent effects
that undoubtedly influence the ensemble of accessible confor-
mations accessible to the peptide in aqueous solution. As noted
above, hydrophobic interactions are likely to be extremely
important in this system.
General Methods. UV irradiation was carried out in a Rayonet
photoreactor using 350 nm fluorescent tubes or in an Aviv 202 CD
spectrometer at the desired wavelength with a bandwidth of 10 nm.
Irradiated peptides were handled in dimmed light and stored at -80
°C. Sedimentation equilibrium analysis was performed in a Beckman
Optima XL-A analytical ultracentrifuge with an An60Ti rotor at three
different rotor speeds at 4 °C in 10 mM acetate buffer, pH 3.8 with 90
mM NaCl. Observed molecular weights were determined using the
Ultrascan II package.48 The partial specific volume was calculated from
the amino acid composition, substituting the azobenzene linker by two
phenylalanine residues.
Starting either from the initially generated NMR model or
from a fully extended structure, the MD simulations converged
(41) Schneider, J. P.; Kelly, J. W. Chem. ReV. 1995, 95, 2169-2187.
(42) Nesloney, C. L.; Kelly, J. W. J. Am. Chem. Soc. 1996, 118, 5836-5845.
(43) Nesloney, C. L.; Kelly, J. W. J. Org. Chem. 1996, 61, 3127-3137.
(44) Wishart, D. S.; Sykes, B. D.; Richards, F. M. J. Mol. Biol. 1991, 222,
311-333.
(45) Wishart, D. S.; Sykes, B. D.; Richards, F. M. Biochemistry 1992, 31, 1647-
1651.
(46) Williamson, M. P. Biopolymers 1990, 29, 1428-1431.
Peptide Synthesis. Peptide synthesis was carried out on a 0.1 mmol
scale using standard Fmoc chemistry on a Rink amide resin (loading
(47) Chothia, C. J. Mol. Biol. 1973, 75, 295-302.
(48) Demeler, B. Ultrascan II, version 6.2; The University of Texas: San
Antonio.
9
2934 J. AM. CHEM. SOC. VOL. 127, NO. 9, 2005

A Photoinducible
â
-Hairpin
A R T I C L E S
0.54 mmol/g) on an Applied Biosystems 432A peptide synthesizer.
Cleavage from the resin and amino acid side chain deprotection was
done with a mixture of 2% triisopropylsilane and 5% H2O in TFA for
40 min and subsequent precipitation of the crude peptide from cold
diethyl ether. The peptide was purified by preparative reversed-phase
HPLC (Nucleosil 100-7 C18 250 × 21 mm2, 5% to 60% acetonitrile
over 45 min in water containing 0.1% trifluoroacetic acid at a flow
rate of 10 mL/min), and purity of the isolated isomers was confirmed
by analytical HPLC (Nucleosil 100-5 C18 250 × 4.6 mm2, 5% to
60% acetonitrile containing 0.05% TFA over 45 min in water containing
0.1% TFA at a flow rate of 1 mL/min.). Retention times for cis and
trans were 29.7 and 31.5 min, respectively. ESI MS m/z M+ calcd
1604.9, found 1604.9.
about 30 min. A total of 4 g of Zn dust was added very slowly. The
reaction mixture was stirred for 6 h at room temperature, filtered, and
slowly added to a solution of 12.4 g (48.9 mmol) of FeCl3‚6 H2O in a
2:1 mixture of H2O and EtOH at -10 °C. After 1 h the mixture was
allowed to warm to room temperature and stirred for another hour. It
was extracted 3 times with EtOAc, the organic extracts were dried over
MgSO4, and the solvent was evaporated. The remaining oil was purified
by flash chromatography (AcOEt/hexane 1:3) to obtain 4.74 g of a
dark green oil that was used without further purification.
(3-Nitrobenzyl)carbamic Acid 9H-Fluoren-9-yl Methyl Ester (7).
To a solution of 5.32 g (28.2 mmol) of 3-nitrobenzylamine hydrochlo-
ride and 24 mL (133 mmol) of diisopropylethylamine in 200 mL of
dichloromethane was added a solution of 7.29 g (28.2 mmol) of Fmoc-
Cl in 50 mL of dichloromethane. After stirring overnight at room
temperature, the solution was extracted 3 times with 10% HCl. The
organic phase was washed with sat. NaHCO3, then water and brine.
After drying over MgSO4 and evaporation of the solvent, 10.7 g (quant.)
NMR Spectroscopy. NMR experiments involving peptides were
carried out at 25 °C (trans) or 4 °C (cis) in 1:9 D2O/H2O 50 mM sodium
acetate buffer pH 3.8 (uncorrected) on a Bruker DRX 500. The solvent
signal was suppressed by a 2 s presaturation delay. All 2D spectra were
recorded with 512 increments in F1 (TPPI) and 4k data points in F2.
These were processed to 1k data points in F1 and F2 with a cos2 filter
in F1 and a cos2 filter shifted by π/3 in F2. For phase sensitive DQF-
COSY, 32 scans per FID were recorded. For TOCSY, 48 scans per
FID were recorded with a 9 kHz spin lock (DIPSY2) and a mixing
time of 80 ms. For NOESY, 64 scans per FID were recorded with
mixing times of 100 and 400 ms, respectively.
1
of a white solid were obtained. Mp 152-153 °C. H NMR (DMSO-
d6, 400 MHz): δ 8.13-8.11 (m, 2 H), 8.01 (t, J ) 6.1 Hz, 1 H), 7.89
(d, J ) 7.5 Hz, 2 H), 7.70-7.61 (m, 4 H), 7.42 (t, J ) 7.4 Hz, 2 H),
7.34-7.30 (m, 2 H), 4.38 (d, J ) 6.8 Hz, 2 H), 4.32 (d, J ) 6.1 Hz,
2 H), 4.24 (t, J ) 6.7 Hz, 1 H). 13C NMR (DMSO-d6, 100 MHz): δ
156.3, 147.7, 143.7, 142.1, 140.7, 133.7, 129.8, 127.5, 126.9, 125.0,
121.7, 121.5, 120.0, 65.4, 46.7, 43.0. HRMS (MALDI, DHB) m/z (M
+ Na)+ calcd 397.1159, found 397.1160.
Hâ1-Hâ2 and Hꢀ1-Hú2 cross-peaks of F8 and W2, respectively,
were used as reference for NOE intensities. Peak assignments were
done with Sparky49 and initial structures were calculated via torsion
angle dynamics using CNS.39
(3-Aminobenzyl)carbamic Acid 9H-Fluoren-9-yl Methyl Ester
(8). In 500 mL of a 2:1 mixture of ethanol and 1,4-dioxane 10.7 g
(28.6 mmol) of 7 were dissolved. After addition of 100 mg of PtO2
under argon, the atmosphere was changed to H2 (1 bar). After 2 h of
vigorous stirring, the catalyst was removed by filtration over Celite.
After evaporation of the solvent, 9.9 g (quant.) of an off white solid
Kinetic Measurements. Isomerization kinetics were performed by
irradiating purified trans-1b in an Aviv 202 CD spectrometer at 320
nm with a bandwidth of 10 nm until the photostationary state was
reached. The isomerization back to trans-1b was then monitored by
UV using thermostated samples stored in the dark or with irradiation
at 420 nm. Rate constants were derived by fitting the data to a first-
order process.
1
were obtained. Mp 142-143 °C. H NMR (DMSO-d6, 400 MHz): δ
7.89 (d, J ) 7.5 Hz, 2 H), 7.75 (t, J ) 6.1 Hz, 1 H), 7.71 (d, J ) 7.4
Hz, 2 H), 7.44-7.40 (m, 2 H), 7.35-7.31 (m, 2 H), 6.94 (t, J ) 7.7
Hz, 1 H), 6.46 (s, 1 H), 6.43 (d, J ) 7.9 Hz, 1 H), 6.38 (d, J ) 7.5 Hz,
1 H), 5.00 (s, 2 H), 4.31 (d, J ) 6.9 Hz, 2 H), 4.22 (t, J ) 6.9 Hz, 1
H), 4.05 (d, J ) 6.1 Hz, 2 H). 13C NMR (DMSO-d6, 100 MHz): δ
156.2, 148.5, 143.8, 140.6, 140.2, 128.6, 127.5, 127.0, 125.1, 120.0,
114.4, 112.4 (2 isochronous signals), 65.3, 46.7, 44.0. HRMS (MALDI,
DHB) m/z (M + Na)+ calcd 367.1417, found 367.1417. Elemental
analysis C22H20N2O2, calcd: C, 76.72; H, 5.85; N, 8.13; O, 9.29.
Found: C, 76.82; H, 5.94; N, 8.04; O, 9.46.
MD Simulations. MD simulations were performed using the
GROMOS96 program,50,51 with the GROMOS 43A1 united-atom force
field51 and the SPC water model.52 Parameter details are published
elsewhere.29
3-Nitrophenylacetic Acid tert-Butyl Ester (4). A solution of 4.95
g (27.4 mol) of 3-nitrophenylacetic acid, 2.06 g (27.9 mmol) of tert-
butanol, and 16 mL (66.8 mmol) of tributylamine in 100 mL of toluene
was added to a suspension of 8.45 g (33.0 mmol) of 2-chloro-1-
methylpyridinium iodide in 50 mL of toluene. The mixture was refluxed
overnight. After removal of the solvent, the dark residue was purified
by flash chromatography (EtOAc/hexane 1:3) to obtain 5.73 g (88%)
(3-{3-[(9H-Fluoren-9-ylmethoxycarbonylamino)methyl]phenylazo}-
phenyl)acetic Acid tert-Butyl Ester (9). A solution of 4.74 g of 5 in
200 mL of glacial acetic acid was added to a solution of 7.36 g (21.4
mmol) of 8 in 200 mL of glacial acetic acid and stirred for 48 h at
room temperature. The solvent was removed in vacuo, and the
remaining gum purified by flash chromatography (AcOEt/hexane 1:3)
1
of a pale yellow oil. H NMR (CDCl3, 400 MHz): δ 8.16 (s, 1 H),
8.13 (d, J ) 8.1 Hz, 1 H), 7.62 (d, J ) 8.1 Hz, 1 H), 7.50 (t, J ) 7.9
Hz, 1 H), 3.65 (s, 2 H), 1.46 (s, 9 H). 13C NMR (CDCl3, 100 MHz):
δ 169.6, 148.3, 135.6, 129.3, 124.4, 122.1, 81.75, 42.0, 28.0. MS (EI)
m/z 237.2. Elemental analysis C12H15NO4, calcd: C, 60.75; H, 6.37;
N, 5.90. Found: C, 60.81; H, 6.45; N, 6.01.
1
to obtain 6.74 g (57%) of a bright orange solid. Mp 64-66 °C. H
NMR (CDCl3, 500 MHz): δ 7.84-7.80 (m, 4 H), 7.75 (d, J ) 7.4 Hz,
2 H), 7.60 (d, J ) 7.4 Hz, 2 H), 7.49-7.45 (m, 2 H), 7.41-7.37 (m,
4 H), 7.30 (m, 2 H), 5.20 (bs, 1 H), 4.49-4.47 (m, 4 H), 4.24 (t, J )
6.9 Hz, 1 H), 3.63 (s, 2 H), 1.45 (s, 9 H). 13C NMR (CDCl3, 125
MHz): δ 170.5, 156.5, 152.9, 152.7, 143.9, 141.3, 139.6, 135.9, 132.0,
130.0, 129.5, 129.2, 127.7, 127.1, 125.0, 123.5, 122.4, 121.8, 121.4,
120.0, 81.1, 66.8, 47.3, 44.9, 42.4, 28.1. HRMS (MALDI, DHB) m/z
(M + Na)+ calcd 570.2363, found 570.2371. Elemental analysis
C34H33N3O4, calcd: C, 74.57; H, 6.07; N, 7.67; O, 11.69. Found: C,
74.47; H, 6.18; N, 7.56; O, 11.73.
3-Nitrosophenylacetic Acid tert-Butyl Ester (5). To a solution of
5.73 g (24.2 mmol) of 4 in 200 mL of methoxyethanol was added a
solution of 1.9 g (35.6 mmol) of NH4Cl in 50 mL of H2O. The solution
was thoroughly degassed by bubbling a stream of argon through it for
(49) Goddard, T. D.; Kneller, D. G. SPARKY 3, version 3.106; University of
California: San Francisco.
(50) Scott, W. R. P.; Hu¨nenberger, P. H.; Tironi, I. G.; Mark, A. E.; Billeter, S.
R.; Fennen, J.; Torda, A. E.; Huber, T.; Kru¨ger, P.; van Gunsteren, W. F.
J. Phys. Chem. A 1999, 103, 3596-3607.
(3-{3-[(9H-Fluoren-9-ylmethoxycarbonylamino)methyl]phenylazo}-
phenyl)acetic Acid (10). To a solution of 6.74 g (12.3 mmol) of 9 in
500 mL of dichloromethane was added 50 mL of trifluoroacetic acid.
After stirring for 36 h, 200 mL of water was added and the organic
layer was separated and washed with water and brine. After drying
over MgSO4, the solvent was removed to obtain 5.87 g (97%) of an
(51) van Gunsteren, W. F.; Billeter, S. R.; Eising, A. A.; Hu¨nenberger, P. H.;
Kru¨ger, P.; Mark, A. E.; Scott, W. R. P.; Tironi, I. G. Biomolecular
Simulation: The GROMOS96 Manual and User Guide; Verlag der
Fachvereine (vdf): Zu¨rich, 1996.
(52) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; Hermans, J.
Intermolecular Forces; Reidel: Dordrecht, 1981; pp 341-342.
9
J. AM. CHEM. SOC. VOL. 127, NO. 9, 2005 2935

A R T I C L E S
Aemissegger et al.
orange-brown solid. Mp 168-169 °C. 1H NMR (DMSO-d6, 400
MHz): δ 12.43 (bs, 1 H), 8.00 (t, J ) 6.1 Hz, 1 H), 7.89 (d, J ) 7.5
Hz, 2 H), 7.80-7.77 (m, 4 H), 7.72 (d, J ) 7.5 Hz, 2 H), 7.58-7.53
(m, 2 H), 7.49-7.39 (m, 4 H), 7.33-7.30 (m, 2 H), 4.38 (d, J ) 6.8
Hz, 2 H), 4.33 (d, J ) 6.1 Hz, 2 H), 4.25 (t, J ) 6.8 Hz, 1 H), 3.74 (s,
1 H). 13C NMR (DMSO-d6, 100 MHz): δ 172.4, 156.3, 151.9, 151.8,
143.8, 141.3, 140.7, 136.5, 132.6, 130.1, 129.3, 129.2, 127.5, 126.9,
125.0, 123.0, 121.5, 121.3, 120.4, 120.0, 65.3, 46.7, 43.4, 40.2. HRMS
(MALDI, DHB) m/z (M + Na)+ calcd 514.1737, found 514.1731.
Acknowledgment. We thank Dr. Marc-Olivier Ebert and
Prof. Bernhard Jaun for their help with the interpretation of the
NMR data.
Supporting Information Available: Tables with chemical
shifts for cis-1b and list of NOE resonance assignments. This
material is available free of charge via the Internet at
http://pubs.acs.org.
JA0442567
9
2936 J. AM. CHEM. SOC. VOL. 127, NO. 9, 2005