Stereochemical requirements for β-hairpin formation: Model studies with four-residue peptides and depsipeptides

Article abstract of DOI:10.1021/ja960429j

Spectroscopic and crystallographic data are present for a series of tetrapeptides and analogous depsipeptides that can form a minimal β-hairpin (two intramolecular hydrogen bonds). These model compounds have been used to test the hypothesis that 'mirror image' β-turns promote β-hairpin formation. This hypothesis was inspired by a statistical survey of β-hairpins in globular proteins (Sibanda, B.L.; Thornton, J.M. Nature 1985, 316, 170), which showed that mirror image β-turns (type I' and type II'), although rare in general, are very commonly associated with β-hairpins containing a two-residue loop between the strand segments. Each of our four-residue molecules contains proline at the second position, to promote a central β-turn. The β-turn is induced to be either 'common' or 'mirror-image', relative to the outer residues, by choice of residue configuration (L vs D). In methylene chloride, end-capped tetrapeptide Ac-L-Val-D-Pro-D-Ala-L-Leu-NMe₂ folds largely into the β-hairpin conformation, while the diastereomer Ac-L-Val-L-Pro-L-Ala-L-Leu-NMe₂ displays little or no β-hairpin folding. For each diastereomer, the hydrogen-bonded driving force for β-hairpin folding is identical, and the dramatic difference in folding behavior therefore reflects a variation in the intrinsic conformational properties of the diastereomeric backbones. Similar behavior is seen for the diastereomeric peptide pair Ac-L-Val-D-Pro-Gly-L-Leu-NMe₂ vs Ac-L-Val-L-Pro-Gly-L-Leu-NMe₂, and for the analogous depsipeptides with a lactic acid or glycolic acid residue at the third position. Thus, our results show not only that mirror-image Pro-X turns strongly promote β-hairpin folding, but also that common β-turns strongly discourage formation of a β-hairpin with a two-residue loop.

Full text of DOI:10.1021/ja960429j

J. Am. Chem. Soc. 1996, 118, 6975-6985
6975
Stereochemical Requirements for â-Hairpin Formation: Model
Studies with Four-Residue Peptides and Depsipeptides
Tasir S. Haque, Jennifer C. Little, and Samuel H. Gellman*
Contribution from the Department of Chemistry, UniVersity of Wisconsin,
Madison, Wisconsin 53706
X
ReceiVed February 9, 1996
Abstract: Spectroscopic and crystallographic data are presented for a series of tetrapeptides and analogous
depsipeptides that can form a minimal â-hairpin (two intramolecular hydrogen bonds). These model compounds
have been used to test the hypothesis that “mirror image” â-turns promote â-hairpin formation. This hypothesis was
inspired by a statistical survey of â-hairpins in globular proteins (Sibanda, B. L.; Thornton, J. M. Nature 1985, 316,
1
70), which showed that mirror image â-turns (type I′ and type II′), although rare in general, are very commonly
associated with â-hairpins containing a two-residue loop between the strand segments. Each of our four-residue
molecules contains proline at the second position, to promote a central â-turn. The â-turn is induced to be either
“common” or “mirror-image”, relative to the outer residues, by choice of residue configuration (L vs D). In methylene
chloride, end-capped tetrapeptide Ac-L-Val-D-Pro-D-Ala-L-Leu-NMe2 folds largely into the â-hairpin conformation,
while the diastereomer Ac-L-Val-L-Pro-L-Ala-L-Leu-NMe2 displays little or no â-hairpin folding. For each
diastereomer, the hydrogen-bonded driving force for â-hairpin folding is identical, and the dramatic difference in
folding behavior therefore reflects a variation in the intrinsic conformational properties of the diastereomeric backbones.
Similar behavior is seen for the diastereomeric peptide pair Ac-L-Val-D-Pro-Gly-L-Leu-NMe2 vs Ac-L-Val-L-Pro-
Gly-L-Leu-NMe2, and for the analogous depsipeptides with a lactic acid or glycolic acid residue at the third position.
Thus, our results show not only that mirror-image Pro-X turns strongly promote â-hairpin folding, but also that
common â-turns strongly discourage formation of a â-hairpin with a two-residue loop.
Introduction
globular proteins.5,6 The shortest common loop is two residues;
in this case, the two loop residues are the second and third
residues of a â-turn. In crystalline proteins, two-residue
â-hairpin loops display distinctive conformational propensities,
as pointed out by Thornton et al. â-Turn conformations are
classified by the φ and ψ torsion angles of the middle two of
The combination of nonbonded forces that controls formation
of â-sheets in proteins is not yet well understood. Individual
amino acid residues have characteristic propensities to participate
in â-sheets, but the relative importance of such intrinsic
propensities and of “context” (i.e., nearest neighbors) is still
under debate.1 â-Sheets in proteins often form from discontinu-
ous polypeptide segments, unlike R-helices or â-turns, and this
feature may explain why there have been many model studies
5
the four residues (i.e., the two residues of the minimal
3g
â-hairpin’s loop). The majority of â-turns in folded proteins
adopt type I and type II conformations, but these “common”
turns are rare in two-residue â-hairpins. Turns that serve as
two-residue â-hairpin loops tend to be type I’ and type II’, where
the prime indicates that the φ and ψ torsion angles of the central
residues are opposite to those in the corresponding common
turns. (Type I’ and II’ are often called “mirror image” turns,
even though these conformations are actually diastereomerically
related to the type I and II turn conformations, respectively,
unless both central residues are glycine.) Sibanda and Thornton
rationalized the correlation between mirror image turns and
â-hairpins by noting that the natural twist of these rare turns
2
3
of R-helix and â-turn stability, but few such studies of â-sheet
1
stability. It is important to identify the most effective strategies
for assembling peptide strands into â-sheets, for rationalizing
natural protein folding patterns, and for designing new proteins.
4
The simplest way to bring two antiparallel strands together
is a short peptide segment between the C-terminus of one strand
and the N-terminus of the other. This strand-loop-strand
arrangement, referred to as a “â-hairpin”, is often observed in
X
Abstract published in AdVance ACS Abstracts, July 15, 1996.
(but not of the common type I and II turns) is compatible with
(1) (a) Kim, C. A.; Berg, J. M. Nature 1993, 362. 267. (b) Minor, D.
the natural twist between two strands of antiparallel â-sheet
L.; Kim, P. S. Nature 1994, 367, 660. (c) Minor, D. L.; Kim, P. S. Nature
5
b
formed from L-amino acids.
1
1
994, 371, 264. (d) Smith, C. K.; Withka, J. M.; Regan, L. Biochemistry
994, 33, 5510. (e) Smith, C. K.; Regan, L. Science 1995, 270, 980.
When a local conformational trend is observed in a folded
protein, it is difficult to know whether this trend results from
(
2) (a) Wojcik, J.; Altmann, K.-H.; Scheraga, H. A. Biopolymers 1990,
3
0, 121. (b) Scholtz, J. M.; Baldwin, R. L. Annu. ReV. Biophys. Biomol.
Struct. 1992, 21, 95. (c) O’Neil, K. T.; DeGrado, W. F. Science 1990,
50, 646. (d) Lyu, P. G.; Liff, M. I.; Marky, L. A.; Kallenbach, N. R.
(4) Leading references on â-sheet protein design efforts: (a) Kullmann,
W. J. Med. Chem. 1984, 27, 106. (b) Klauser, S.; Gantner, K.; Salgam, P.;
Gutte, B. Biochem. Biophys. Res. Commun. 1991, 179, 1212. (c) Quinn,
T. P.; Tweedy, N. B.; Williams, R. W.; Richardson, J. S.; Richardson, D.
C. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 8747. (d) Yan, Y.; Erickson, B.
W. Protein Sci. 1994, 3, 1069.
2
Science 1990, 250, 669. (e) Merutka, G.; Lipton, W.; Shalongo, W.; Park,
S.-H.; Stellwagen, E. Biochemistry 1990, 29, 7511. (f) Kemp, D. S.; Boyd,
J. G.; Mundel, C. Nature 1991, 352, 451.
(3) (a) Boussard, G.; Marraud, M. J. Am. Chem. Soc. 1985, 107, 1825.
(
b) Wright, P. E.; Dyson, H. J.; Lerner, R. A. Biochemistry 1988, 27, 7167.
(5) (a) Sibanda, B. L.; Thornton, J. M. Nature 1985, 316, 170. (b)
Wilmot, C. M.; Thornton, J. M. J. Mol. Biol. 1988, 203, 221. (c) Sibanda,
B. L.; Thornton, J. M. J. Mol. Biol. 1993, 229, 428.
(6) Short peptides folded into â-hairpins under non-aqueous conditions:
(a) Ueki, T.; Bando, S.; Ashida, T.; Kakudo, M. Acta Crystallogr. 1971,
B27, 2219. (b) Karle, I. L.; Kishore, R.; Raghothama, S.; Balaram, P. J.
Am. Chem. Soc. 1988, 110, 1958. (c) Awasthi, S. K.; Raghothama, S.;
Balaram, P. Biochem. Biophys. Res. Commun. 1995, 216, 375.
(
c) Falcomer, C. M.; Meinwald, Y. C.; Choudhary, I.; Talluri, S.; Milburn,
P. J.; Clardy, J.; Scherage, H. A. J. Am. Chem. Soc. 1992, 114, 4036. (d)
Liu, X.; Scott, P. G.; Otter, A.; Kotovych, G. Biopolymers 1992, 32, 119.
(e) Pietrzynski, G.; Rzeszotarska, B.; Kubica, Z. Int. J. Peptide Protein
Res. 1992, 40, 524. (f) Liang, G.-B.; Rito, C. J.; Gellman, S. H. J. Am.
Chem. Soc. 1992, 114, 4440. (g) Rose, G. D.; Gierasch, L. M.; Smith, J.
A. AdV. Protein Chem. 1985, 37, 1.
S0002-7863(96)00429-5 CCC: $12.00 © 1996 American Chemical Society

6
976 J. Am. Chem. Soc., Vol. 118, No. 29, 1996
Haque et al.
local or nonlocal forces. In order to determine whether the
correlation between mirror image turns and two-residue â-hair-
pin loops reflects an intrinsic conformational preference of
peptide backbones, we have examined the folding of tetra-
peptides and analogous depsipeptides in solution.7 A tetra-
peptide can adopt a “minimal â-hairpin” conformation, defined
by the presence of the 10- and 14-membered ring N-H--OdC
hydrogen bonds illustrated below. Selective incorporation of
D-residues allows us to predispose the inner two residues to
form a “mirror image” turn, relative to the outer two residues.
We probe the correlation between intrinsic â-hairpin propensity
and the presence of a “mirror image” turn conformation by
comparing the folding behavior of all-L molecules and their
heterochiral diastereomers.
Scheme 1
Chart 1
In these small molecules, intramolecular hydrogen bonding
provides a principal driving force for â-turn and â-hairpin
formation. For this reason, many of our studies have been
carried out in a relatively nonpolar solvent (methylene chloride),
which does not provide strong hydrogen bonding competition.
In each diastereomeric comparison discussed below, the hy-
drogen bonding options are identical between the isomers, but
the intrinsic folding propensity of the backbone varies between
isomers. Elucidation of these intrinsic folding propensities,
which should be largely independent of solvent, is the central
goal of these studies.
Our experimental approach requires that the central two
residues of the model compounds have a large tendency to adopt
â-turn conformations. The presence of L-proline at the second
of the four â-turn residues is well-known to promote formation
3g
of type I and type II turns, and all of our compounds contain
proline as the second residue. Further predisposition for â-turn
formation is achieved in one set of model compounds by using
an R-hydroxy acid residue rather than an R-amino acid residue
in the third position, which results in an ester rather than an
amide linkage between the second and third residues. End-
capped Pro-X dipeptides are known to equilibrate between
â-turn and γ-turn conformations (the latter involves a seven-
membered-ring hydrogen bond across the proline residue) in
3f
nonpolar solvents at room temperature, while the analogous
depsipeptides exist almost exclusively in the â-turn conformation
8
(
10-membered-ring hydrogen bond; Scheme 1). Crystallo-
8
graphic analysis of Pro-X depsipeptides indicates that the
R-hydroxy acid residue is a good structural mimic for an
R-amino acid residue, as one would expect in light of the similar
conformational properties of secondary amide and ester units.
The experiments below involve four tetrapeptides, containing
the sequences Val-Pro-Ala-Leu and Val-Pro-Gly-Leu, and five
depsipeptides, with Ala or Gly replaced by lactic acid (Lac) or
glycolic acid (Glyco), respectively. The terminal residues were
chosen based on their tendencies in proteins to occupy the first
and fourth positions of â-turns embedded in hairpins, and with
an eye toward maximizing solubility in nonpolar solvents.
These molecules are illustrated in Chart 1, along with the
shorthand designations used in the text.
5
Results and Discussion
(
7) A preliminary report has appeared on the depsipeptides discussed
here: Haque, T. S.; Little, J. C.; Gellman, S. H. J. Am. Chem. Soc. 1994,
16, 4105.
8) Boussard, G.; Marraud, M.; Neel, J.; Maigret, M.; Aubry, A.
Biopolymers 1977, 16, 1033.
Depsipeptides in Methylene Chloride. Figure 1 shows the
1
1
variation in the amide proton H NMR chemical shifts for
(
D
L
D
L
L
L
V PGlyco L and V PGlyco L in CD2Cl2 as a function of

Stereochemical Requirements for â-Hairpin Formation
J. Am. Chem. Soc., Vol. 118, No. 29, 1996 6977
D
L
D
D
L
D
D
V PGlyco L, and V P Lac L in CH2Cl2 (each depsipeptide
mM, i.e., below the onset of aggregation). We have previously
1
made extensive use of N-H stretch data obtained in dilute
CH2Cl2 solution to analyze intramolecular hydrogen bonding
9
in peptides, depsipeptides, and related molecules. These prior
studies allow straightforward interpretation of the data in Figure
2
. For both all-L depsipeptides, two bands of similar intensity
-
1
are observed, at 3422 and ca. 3310 cm . The former results
from a weak intraresidue five-membered ring N-H--OdC
interaction (“C5 interaction”; whether or not this interaction
10
constitutes a hydrogen bond has been a subject of debate).
-
1
The bands near 3310 cm arise from N-H involved in typical
amide-to-amide hydrogen bonds. Thus, a substantial proportion
of the N-H in the all-L depsipeptides is free of interresidue
hydrogen bonding, which is consistent with our hypothesis
(Scheme 2) that these depsipeptides experience considerable
â-turn formation but little or no â-hairpin formation. The
2 2
Figure 1. Amide proton NMR chemical shifts in CD Cl at room
temperature, as a function of the logarithm of depsipeptide con-
centration: (×) Val NH (“Val* NH”) and Leu NH (“Leu* NH”) of
D
L
D
L
L
L
V PGlyco L; (s, O) Val NH and Leu NH of V PGlyco L.
D
L
L
D
D
L
D
the logarithm of depsipeptide concentration. In order to
facilitate interpretation of this graph, we display the behavior
of the all-L depsipeptide with continuous lines, while the data
heterochiral depsipeptides V P Lac L and V PGlyco L,
-1
however, display only one N-H stretch band, at ca. 3296 cm ,
which is consistent with formation of strong intramolecular
amide-to-amide hydrogen bonds. The absence of a band
between 3400 and 3500 cm for these depsipeptides indicates
that both amide protons are fully hydrogen bonded, i.e., that
the â-hairpin conformations are fully populated under these
conditions (Scheme 2). (For these molecules, only the â-hairpin
folding pattern allows formation of two strong intramolecular
amide-to-amide hydrogen bonds simultaneously.)
D
L
D
for V PGlyco L are shown with symbols. The chemical shift
of an amide proton is usually very sensitive to hydrogen
bonding, and these data, spanning more than three orders of
magnitude in concentration, provide insight on the concentra-
tions at which aggregation (i.e., intermolecular hydrogen bond-
ing) occurs. Both amide protons of the all-L isomer shift
downfield above ca. 5 mM, indicating the onset of aggregation
at this concentration. In contrast, neither amide proton of
-
1
It has long been known that placement of a D-residue at the
third position of an otherwise all-L â-turn promotes formation
D
L
D
V PGlyco L displays any concentration dependence, which
suggests that this depsipeptide does not aggregate in the
concentration range examined. An alternative possibility, that
of a type II turn, and that L-Pro-D-Xxx sequences are particularly
11
strong type II turn inducers.
We therefore examined
D
L
D
V PGlyco L is fully aggregated at 0.01 mM, seems exceed-
D
L
D
D
V P Lac L, the central two residues of which should form a
ingly unlikely, given that each molecule contains just two
hydrogen bond donor sites.
mirror image turn relative to the outer residues. The rightmost
spectrum in Figure 2 shows N-H stretch region IR data for a
For a given amide proton, increased population of hydrogen
bonded states leads to a downfield shift (hydrogen bonding
equilibration in flexible molecules is rapid on the NMR time
scale, and the observed δNH reflects a population-weighted
average of contributing hydrogen bonded and non-hydrogen
bonded states). Comparisons between analogous protons in
diastereomers therefore provide insight on the extent of hydro-
gen bonding at the amide proton in question. The behavior of
the valine NH is of particular interest, since this proton can
participate in the 14-membered-ring hydrogen bond that defines
the minimal â-hairpin (involving the leucine CdO). For
1
mM solution of this depsipeptide in CH2Cl2. The major band
-1
occurs at 3308 cm , in the region for strong amide-to-amide
hydrogen bonds. Unlike the other two heterochiral depsipep-
tides, however, there is also a tiny absorbance at 3420 cm .
-1
Thus, the mirror image turn strongly promotes â-hairpin
D
L
D
D
formation in V P Lac L, although not quite to the extent
D
L
L
D
D
L
D
observed for V P Lac L and V PGlyco L.
Long-range nuclear Overhauser effects, detected via a
12
ROESY experiment, provided further evidence of â-hairpin
D
L
L
D
D
L
D
D
L
D
D
folding for V P Lac L, V PGlyco L, and V P Lac L in
CD2Cl2. This folding pattern requires that the two amide
protons lie near one another in space, and the expected Val
NH--Leu NH cross peak was observed for all three heterochiral
depsipeptides (Figure 3). No such cross peak was observed in
ROESY spectra of the all-L depsipeptides. It should be noted
that an NOE between protons that are so covalently distant from
one another is more strongly diagnostic of a compact conforma-
tion than are NOEs between protons on adjacent residues (the
latter are commonly adduced to support population of â-turns
and other secondary structures in short peptides). NH--NH cross
peaks between adjacent residues, for example, are consistent
with â-turn and R-helix conformations, but as Landis et al. have
D
L
D
V PGlyco L, Val NH appears at 8.25 ppm, while for the all-L
isomer, Val NH appears at 6.35 ppm, in the concentration-
independent limit. The latter chemical shift is characteristic of
a peptide backbone proton that experiences little or no hydrogen
bonding in CD2Cl2, while the former chemical shift indicates a
large amount of intramolecular hydrogen bonding.3 In contrast
to the 1.9-ppm separation between the Val NH’s, the Leu NH’s
of these diastereomers are only 0.4 ppm apart, which is
consistent with there being a substantial population of the â-turn
folding pattern (10-membered-ring hydrogen bond) in both
f,9
L
L
L
molecules. Thus, the δNH comparison between V PGlyco L
D
L
D
and V PGlyco L suggests that the all-L isomer experiences
significant â-turn folding, but little or no â-hairpin folding, while
the diastereomer in which the inner residues are predisposed to
form a mirror image turn, relative to the outer (Val and Leu)
residues, is largely folded into the â-hairpin. The analo-
13
recently shown, such relatively short-range interactions can
also arise from other conformations available to flexible
peptides.
(9) (a) Dado, G. P.; Gellman, S. H. J. Am. Chem. Soc. 1993, 115, 4228.
1
7
L
L
L
L
(b) Gallo, E. A.; Gellman, S. H. J. Am. Chem. Soc. 1993, 115, 9774. (c)
Gallo, E. A.; Gellman, S. H. J. Am. Chem. Soc. 1994, 116, 11560.
gous H NMR data previously reported for V P Lac L and
D
L
L
D
V P Lac L imply a similar pattern of folding behavior for this
(10) (a) Avignon, M.; Huong, P. V.; Lascombe, J.; Marraud, M.; Neel,
diastereomeric pair. These conformational deductions are
summarized in Scheme 2.
J. Biopolymers 1969, 8, 69. (b) Burgess, A. W.; Scheraga, H. A.
Biopolymers 1973, 12, 2177.
(
11) Boussard, G.; Marraud, M.; N e´ el, J. J. Chim. Phys. 1974, 71, 1081.
Figure 2 shows FT-IR data from the N-H stretch region for
(12) Bothner-By, A. A.; Stephens, R. L.; Lee, J.; Warren, C. D.; Jeanloz,
L
L
L
L
D
L
L
D
L
L
L
all five depsipeptides, V P Lac L, V P Lac L, V PGlyco L,
R. W. J. Am. Chem. Soc. 1984, 106, 811.

6
978 J. Am. Chem. Soc., Vol. 118, No. 29, 1996
Haque et al.
Figure 2. N-H stretch FT-IR data for 1 mM depsipeptide samples in CH
2
Cl
2
L
at room temperature, after subtraction of the spectrum of pure
L
L
L
L
-1
D
L
D
-1
L
L
L
CH
2 2
Cl . From left to right: V P Lac L, maxima at 3422 and 3317 cm ; V P Lac L, maximum at 3297 cm ; V PGlyco L, maxima at 3422
-
1
D
L
D
-1
D
L
D
D
-1
and 3306 cm
;
V PGlyco L, maximum at 3295 cm
; V P Lac L, maxima at 3420 and 3308 cm .
Scheme 2. Major Folding Patterns of Depsipeptides in Methylene Chloride and Acetonitrile
relative to methylene chloride: acetonitrile is a modest hydrogen
bond acceptor, and DMSO is a strong acceptor. N-H stretch
L
L
L
L
D
L
L
D
L
L
L
region IR data for V P Lac L, V P Lac L, V PGlyco L,
D
L
D
and V PGlyco L in CH3CN (Figure 4) show that folding
behavior in this solvent is similar to that observed in CH2Cl2.
(CH3CN IR data were obtained with 10 mM solutions; we
assume that there is no aggregation under these conditions, by
extrapolation from the methylene chloride results.) Each of the
all-L depsipeptides displays two maxima in the N-H stretch
-
1
region, at ca. 3370 and 3310 cm . The lower energy bands
are very similar in position to those observed for the same
molecules in CH2Cl2, and we attribute these bands to N-H
moieties engaged in intramolecular amide-to-amide hydrogen
bonds (presumably the 10-membered-ring â-turn hydrogen
-
1
bond). The higher energy bands are ca. 50 cm lower than
the analogous bands in CH2Cl2; these bands in Figure 4 are
assigned to N-H groups interacting with CH3CN. (Solvent-
exposed amide groups in CH3CN display an N-H stretch band
-1
that is 50-60 cm lower in energy than the corresponding band
14
in CH2Cl2; this solvent-induced shift presumably results from
weak CH3CN--H-N hydrogen bonds.) Thus, there appears to
be considerable â-turn but little or no â-hairpin for the all-L
depsipeptides in CH3CN, as was observed in CH2Cl2.
D
L
L
D
Figure 3. The amide proton portion of the 500-MHz ROESY spectrum
The IR data for heterochiral depsipeptides V P Lac L and
V PGlyco L in CH3CN (Figure 4) are nearly identical to the
of depsipeptide DV P Lac L, 1 mM in CD
L
L
D
at 297 K. Data were
D
L
D
2
Cl
2
acquired on a Varian Unity instrument using the ROESY pulse sequence
that is part of the standard Varian software, with 2048 data points along
data obtained in CH2Cl2 (Figure 2), indicating that the â-hairpin
induced by a mirror image turn is sufficiently robust to resist
hydrogen bond competition from surrounding CH3CN mol-
ecules. (We have previously observed that CH3CN can disrupt
t
2
, 128 t
was 0.50 s, accompanied by a relaxation delay of 1.20 s and a
homospoil” relaxation pulse. The spectral width was 6000 Hz.
1 1
increments, and 112 scans per t increment. The mixing time
“
14
less stable intramolecular hydrogen bonding patterns. ) The
population of â-hairpin conformations by the heterochiral
depsipeptides in acetonitrile is supported by ROESY data: the
hairpin-characteristic Val NH--Leu NH cross peak is observed
Additional details may be found in the Experimental Section. Proton
assignments were made on the basis of TOCSY data.
Depsipeptides in Acetonitrile and DMSO. These two
solvents provide an increase in hydrogen bonding competition
D
L
L
D
D
L
D
for both V P Lac L and V PGlyco L, although the intensity
(
13) Landis, C. R.; Luck, L. L.; Wright, J. M. J. Magn. Reson. B 1995,
(14) Gellman, S. H.; Dado, G. P.; Liang, G.-B.; Adams, B. R. J. Am.
Chem. Soc. 1991, 113, 1164.
1
09, 44.

Stereochemical Requirements for â-Hairpin Formation
J. Am. Chem. Soc., Vol. 118, No. 29, 1996 6979
Figure 4. N-H stretch FT-IR data for 10 mM depsipeptide samples in CH
3
CN at room temperature, after subtraction of the spectrum of pure
3
CN. From left to right: V P Lac L, maxima at 3370 and 3319 cm ; V P Lac L, maximum at 3298 cm ; V PGlyco L, maxima at 3368
L
L
L
L
-1
D
L
L
D
-1
L
L
L
CH
-
1
D
L
D
-1
and 3304 cm
;
V PGlyco L, maximum at 3296 cm .
Figure 5. N-H stretch FT-IR data for 10 mM depsipeptide samples in DMSO at room temperature, after subtraction of the spectrum of pure
L
L
L
L
-1
D
L
L
D
-1
L
L
L
-1
DMSO. From left to right: V P Lac L, maximum at 3259 cm ; V P Lac L, maximum at 3281 cm ; V PGlyco L, maximum at 3258 cm
;
D
L
D
-1
V PGlyco L, maximum at 3276 cm
.
Table 1. Cis:Trans Ratios
depsipeptide
is weaker for the latter. A weak NH--NH cross peak is observed
also for V P Lac L in CD3CN, but not for the all-L depsipep-
tides.
D
L
D
D
CD Cl2
2
CD CN
3
DMSO-d6
L
L
L
L
V P Lac L
6:94
7:93
a
5:95
a
4:96
12:87
5:95
15:85
25:75
L
L
L
L
D
L
L
D
L
L
L
DMSO IR data for V P Lac L, V P Lac L, V PGlyco L,
D
L
L
L
D
V P Lac L
a
a
a
a
D
L
D
L
L
and V PGlyco L are shown in Figure 5. N-H stretch signals
lose their diagnostic spectral dispersion in this solvent, because
amide-to-DMSO hydrogen bonds and strong amide-to-amide
hydrogen bonds give rise to bands in the same region of the IR
spectrum. Nevertheless, an interesting pattern is observed
among the data in Figure 5: the all-L depsipeptides display a
maximum at 3258 cm , while the maxima for V P Lac L
and V PGlyco L are ca. 20 cm higher in wavenumber. The
CH2Cl2 and CH3CN data show that amide-to-amide hydrogen
V PGlyco L
D
D
L
D
V PGlyco L
L
D
D
V P Lac L
3:97
a
Cis rotamer not detected.
that the proline exist as the trans rotamer. Table 1 shows the
L
L
L
L
D
L
L
D
-
1
D
L
L
D
cis:trans ratios for the depsipeptides V P Lac L, V P Lac L,
L
L
L
D
L
D
D
L
D
-1
V PGlyco L, and V PGlyco L in the three solvents employed
in these studies. The ratios in methylene chloride and aceto-
nitrile are consistent with our contention that the heterochiral
sequences promote â-hairpin formation, but these data do not
in themselves constitute strong evidence for this deduction, since
the amount of cis proline is always low. Greater amounts of
cis proline are observed in DMSO than in the other two solvents,
which is consistent with the general trend that increasing solvent
polarity favors the cis form. Interestingly, the all-L molecules
show lower cis:trans ratios than the heterochiral molecules in
DMSO. The origin of this trend is unclear, but the cis:trans
ratio cannot be related to â-hairpin formation in DMSO, since
we have independent evidence for some population of â-hairpin
-
1
bonded N-H occurs in the range 3290-3300 cm . Therefore,
the difference between the all-L depsipeptides and their hetero-
chiral isomers in DMSO suggests that significant â-hairpin
D
L
L
D
D
L
D
folding is retained by V P Lac L and V PGlyco L in this
solvent, while the all-L isomers are largely unfolded and
hydrogen bonded to DMSO. A weak ROESY Val NH--Leu
15
NH cross peak was observed for DV P Lac L in DMSO-d6,
L
L
D
which is good evidence for the existence of some â-hairpin
structure in this very competitive solvent. No NH--NH cross
D
L
D
D
L
D
D
peak was detected for V PGlyco L or V P Lac L in DMSO-
d6.
Cis/Trans Proline Rotamer Ratios. Adoption of a â-turn
or â-hairpin conformation by our four-residue molecules requires
(
15) Higashijima, T.; Tasumi, M.; Miyazawa, T. Biopolymers 1977, 16,
1259.

6
980 J. Am. Chem. Soc., Vol. 118, No. 29, 1996
Haque et al.
L
D
D
L
L
L
L
L
V P A L and V P A L in CD2Cl2 as a function of the
logarithm of peptide concentration. The all-L peptide aggregates
L
D
D
L
above ca. 1 mM, while V P A L does not appear to aggregate
L
D
D
L
significantly between 0.01 and 50 mM. Val NH of V P A L,
which would be involved in the hairpin-diagnostic 14-membered-
ring hydrogen bond, is the furthest downfield of the six amide
protons, in the concentration-independent limit. This observa-
tion suggests that there is substantial population of the â-hairpin
conformation by V P A L. In contrast, Val NH of the all-L
isomer is the most upfield of the six amide protons, which
implies that there is little or no â-hairpin folding in this case.
L
D
D
L
Figure 6. Ball-and-stick representation of the solid state conformation
L
D
D
L
D
L
D
As required for â-hairpin formation, Ala NH of V P A L
is quite far upfield, consistent with there being little or no
internal hydrogen bonding to this amide proton. Ala NH of
the all-L isomer, on the other hand, is the most downfield of
this peptide’s amide protons, in the concentration-independent
limit, suggesting significant hydrogen bonding at this position.
The Ala NH chemical shift of the all-L isomer is similar to that
observed for the dipeptide Ac-L-Pro-L-Ala-NHMe under identi-
of V PGlyco L, as determined by X-ray diffraction. The two amide
protons are engaged in intermolecular hydrogen bonds (not shown).
Numbering is given for nitrogen and oxygen atoms only.
3
f
cal conditions. There is a significant population of the γ-turn
folding pattern across the proline residue for this dipeptide in
3
f
methylene chloride at room temperature (seven-membered-
ring hydrogen bond involving the Ala NH; see Scheme 1). Thus,
L
L
L
L
the L-Pro-L-Ala segment of V P A L apparently exhibits
behavior similar to that observed for the isolated dipeptide.
Scheme 3 summarizes the conformational behavior we ascribe
L
L
L
L
L
D
D
L
to V P A L and V P A L in methylene chloride; this
hypothesis is supported by additional data outlined below, and
applies to the glycine-containing tetrapeptides, as well.
L
L
L
Concentration-dependent δNH data for V PG L and
Figure 7. Ball-and-stick representation of the solid state conformation
L
D
L
D
L
D
D
V PG L in CD2Cl2 are shown in Figure 9. The Val NH
of V P Lac L, as determined by X-ray diffraction. The two intramo-
lecular hydrogen bonds are indicated by dashed lines. Numbering is
given for nitrogen and oxygen atoms only.
chemical shifts (concentration-independent limit) are diagnostic
of â-hairpin formation, and in this feature the glycine-containing
peptides parallel the alanine-containing peptides. For the all-L
isomer, Val NH is far upfield, in the range typical of non-
L
D
L
hydrogen bonded peptide NH groups, but for V PG L, Val
NH is far downfield. There are some differences between the
L
D
L
data sets in Figures 8 and 9, however. First, V PG L is more
L
D
D
L
prone to aggregate than V P A L. Second, in the concentra-
tion-independent limits, Leu NH is further downfield than Gly
NH for V PG L, while Leu NH was upfield of Ala NH for
L
L
L
L
L
L
L
V P A L. This latter difference presumably results from the
fact that, in the competition between â- and γ-turn folding
patterns, the â-turn is relatively more favorable for Ac-L-Pro-
3
f
Gly-NHMe than for Ac-L-Pro-L-Ala-NHMe. Thus, the data
Figure 8. Amide proton NMR chemical shifts in CD
temperature, as a function of the logarithm of peptide concentration:
×) Val NH (“Val* NH”), Leu NH (“Leu* NH”), and Ala NH (“Ala*
NH”) of V P A L; (s, O) to Val NH, Leu NH, and Ala NH of
2 2
Cl at room
in Figure 9 suggest that the conformational hypothesis outlined
in Scheme 3 applies to V PG L and V PG L, but that the
L
L
L
L
D
L
(
position of the â- vs γ-turn equilibrium varies between
L
D
D
L
L
L
L
L
L
L
L
V P A L and V PG L.
L
L
L
L
V P A L.
L
L
L
L
L
D
D
L
N-H stretch region IR data for V P A L, V P A L,
L
L
L
L
D
L
V PG L, and V PG L (Figure 10) support the conclusion that
conformations for heterochiral but not all-L depsipeptides
D
L
L
D
D
L
L
D
the all-L tetrapeptides experience little or no â-hairpin formation,
(
ROESY data for V P Lac L and IR data for V P Lac L
L
D
D
L
L
D
L
D
L
D
while V P A L and V PG L are largely folded into the
and V PGlyco L).
â-hairpin conformation (Scheme 3). For each of the all-L
Depsipeptide Crystal Structures. Crystals suitable for
X-ray diffraction were obtained for both all-L depsipeptides and
-
1
peptides, the major band appears at ca. 3420 cm , which, as
discussed above, indicates an absence of strong amide-to-amide
hydrogen bonding. These peptides display only small or
moderate bands in the strong intramolecular hydrogen bonding
D
L
D
D
L
D
D
for V PGlyco L and V P Lac L. The two all-L depsipep-
D
L
D
tides and V PGlyco L crystallized in extended conformations,
with only intermolecular hydrogen bonds formed in the solid
state (the solid state conformation of V PGlyco L is shown
D
L
D
-1
region (3323 and 3311 cm ). In contrast, the major bands for
D
L
D
D
L
D
D
L
L
D
L
in Figure 6). V P Lac L, however, crystallized in a â-hairpin
conformation, as shown in Figure 7 (L-Pro-D-Lac φ and ψ
angles, sequentially: -65.0°, 125.2°, 76.8°, 14.5°; these torsion
V P A L and V PG L fall in the region that is characteristic
of strong intramolecular amide-to-amide hydrogen bonding.
Both of these heterochiral peptides have a minor band at higher
3g
-1
L
D
D
L
-1
angles correspond to the expected type II turn ).
wavenumber, 3420 cm for V P A L and 3435 cm for
L
D
L
Tetrapeptides in Methylene Chloride. Figure 8 shows the
V PG L. If hairpin folding is dominant in these heterochiral
molecules, then the Ala or Gly NH should not engage in
1
variation in the amide proton H NMR chemical shifts for

Stereochemical Requirements for â-Hairpin Formation
J. Am. Chem. Soc., Vol. 118, No. 29, 1996 6981
Scheme 3. Major Folding Patterns for Tetrapeptides in Methylene Chloride and Acetonitrile
Although the long-range signals are relatively weak, the data
in Figures 11 and 12 indicate that these cross peaks do not result
from transferred NOE effects. These long-range cross peaks
are not observed for the all-L peptides in CD2Cl2. The all-L
peptides did display an NOE cross peak between the amide
protons of the third and fourth residues (Ala NH--Leu NH or
Gly NH--Leu NH), which is conventionally interpreted to
indicate the presence of â-turn conformations. As mentioned
above, however, this relatively short-range interaction can arise
1
3
from other types of conformations as well.
Tetrapeptides in Acetonitrile and DMSO. N-H stretch
L
L
L
L
L
D
D
L
L
L
L
region IR data for V P A L, V P A L, V PG L, and
Figure 9. Amide proton NMR chemical shifts in CD
temperature, as a function of the logarithm of peptide concentration:
2 2
Cl at room
L
D
L
V PG L in CH3CN (Figure 13) imply considerable population
(
×) Val NH (“Val* NH”), Leu NH (“Leu* NH”), and Gly NH (“Gly*
of â-hairpin conformations for the heterochiral peptides, but
little internal hydrogen bonding for the all-L isomers (Scheme
L
D
L
L
L
L
NH”) of V PG L; (s, O) Val NH, Leu NH and Gly NH of V PG L.
3
). For the all-L peptides, the absorbance maximum appears at
hydrogen bonding, which would explain the origin of these
minor bands. In order to elucidate the origin of the 3420-cm
-1
ca. 3370 cm , characteristic of solvent-exposed NH moieties
-
1
L
L
L
-1
in CH3CN. V PG L displays a shoulder at ca. 3300 cm ,
indicating a small population of conformations containing strong
intramolecular amide-to-amide hydrogen bonds (given the band
widths, one cannot rule out the presence of a small population
of such hydrogen bonds in V P A L). For each of the
heterochiral peptides, the absorbance maximum falls in the
region attributable to NH involved in strong amide-to-amide
hydrogen bonds; the maxima are nearly identical to those
observed in CH2Cl2 (Figure 10). V PG L also displays a
shoulder in the region for solvent-exposed NH in CH3CN, and
the breadth of the band for V P A L suggests that this peptide
displays significant absorbance in the solvent-exposed region,
too. (For these peptides, â-hairpin formation requires that one
of the three NH groups be solvent exposed.)
band of LV P A L, we prepared a version containing 15N-
D
D
L
D
L
15
L
D
labeled alanine, V P( N) A L. As expected, the isotopic
-
1
substitution caused by 3420-cm maximum to shift to lower
energy16 (3412 cm ; rightmost spectrum in Figure 10), but this
-1
L
L
L
L
-1
shift revealed a small shoulder at 3432 cm , which we attribute
to a small amount of non-hydrogen bonded Val and/or Leu
N-H.
1
7
NOESY data obtained in CD2Cl2 support our deduction that
L
D
L
L
D
D
L
L
D
L
V P A L and V PG L are folded into â-hairpin conforma-
tions to large extents, while the all-L isomers are not (Scheme
L
D
D
L
). For both LV P A L and V PG L, a long-range Val
D
D
L
L
D
L
3
NH--Leu NH cross peak was observed, as was an additional
long-range cross peak between C-terminal and N-terminal
methyl groups. The buildup of the long-range cross peaks is
shown in Figures 11 and 12, along with the buildup of closer
protons, Ala NH--Leu NH and Gly NH--Leu NH, respectively.
L
L
L
L
L
D
D
L
N-H stretch region data for V P A L, V P A L,
L
L
L
L
D
L
V PG L, and V PG L in DMSO are shown in Figure 14.
For three of the four molecules, the maximum occurs at ca.
-
1
(
16) For a localized A-B stretch, the band position can be estimated
3267 cm , which is similar to the position of the N-H stretch
bands observed for the all-L depsipeptides in DMSO (Figure
5). V PG L, however, shows a maximum at 3286 cm ,
which is similar to that observed for the heterochiral depsipep-
tides in DMSO. Thus, the V PG L data suggest that there
may be some â-hairpin folding even in the face of strong
hydrogen bonding competition from the solvent. No long-range
interactions were detected in NOESY spectra acquired for the
tetrapeptides in DMSO-d6.
from the equation
L
D
L
-1
-1
1/2
ν ) (2πc) [k(M + M )/M M ]
A
B
A
B
where c is the speed of light, k is the force constant of the A-B bond, MA
is the mass of atom A, and MB is the mass of atom B. (Silverstein, R. M.;
Bassler, G. C.; Morrill, T. C. Spectrometric Identification of Organic
Compounds, 5th ed.; John Wiley & Sons: New York, 1991; p 93.) This
L
D
L
1
5
-1
calculation predicts a localized N-H stretch to be ca. 12 cm lower in
1
4
energy than a N-H stretch.
17) Macura, S.; Ernst, R. R. Mol. Phys. 1980, 41, 95.
(

6
982 J. Am. Chem. Soc., Vol. 118, No. 29, 1996
Haque et al.
Figure 10. N-H stretch FT-IR data for 1 mM peptide samples in CH
2
Cl
2
at room temperature, after subtraction of the spectrum of pure CH
2
Cl
2
.
L
L
L
L
-1
L
D
D
L
-1
L
L
L
From left to right: V P A L, maxima at 3420 and 3323 cm ; V P A L, maxima at 3420 and 3324 cm ; V PG L, maxima at 3425 and 3311
-
1
L
D
L
-1
D
L
L
15
D
-1
-1
cm ; V PG L, maxima at 3435 and 3310 cm
;
V P ( N)A L, shoulder at 3432 cm and maxima at 3412 and 3321 cm
.
D-residues will prove to be useful for de novo design of â-sheet
1
8
proteins.
Since the hydrogen bonded drive for â-hairpin formation is
identical between diastereomers, the conformational differences
we observe provide insight on the intrinsic folding propensities
of the diastereomeric peptide or depsipeptide backbones. These
intrinsic folding propensities are presumably determined by
torsional strain and other nonbonded repulsions that develop
as the peptide backbone folds back upon itself, and by
1
9
conformational entropy.
(We assume that differences in
solvation between diastereomers are energetically small, in all
solvents.) Our results suggest that the correlation between
mirror-image turns and two-residue loop â-hairpins observed
1
1
Figure 11. Intensities of selected H- H NOESY cross peaks as a
L
D
D
L
2 2
function of mixing time for a 1 mM sample of V P A L in CD Cl .
Intensity is in arbitrary units, and is normalized between experiments.
Further details may be found in the Experimental Section.
5
in crystalline proteins arises from competition between the
intrinsically favored twist for two adjacent strands of â-sheet
and the intrinsically favored backbone conformations for â-turns,
with the former winning out.
The similarity of trends in the peptide and depsipeptide series
validates our initial assumption that the depsipeptides are
reasonable models for the conformational behavior of authentic
peptide backbones. The â-hairpin fold appears to be modestly
more stable in the heterochiral depsipeptides than in the
analogous peptides, perhaps because of the equilibrium involv-
ing the γ-turn shown in Scheme 1, which is unique to the
peptides.
2
. A Common â-Turn Is an Impediment to Formation
of a Two-Residue â-Hairpin Loop. This conclusion is based
on the behavior of the all-L tetrapeptides and depsipeptides in
methylene chloride, where failure to form a â-hairpin requires
that the hydrogen bonding potential of Val N-H and Leu CdO
remain unsatisfied. It is commonly assumed that stabilizing
any â-turn necessarily promotes â-hairpin formation;4
however, our results show that the type of â-turn is crucial,
and that the most common turn conformations for all-L peptides
1
1
Figure 12. Intensities of selected H- H NOESY cross peaks as a
L
D
L
2 2
function of mixing time for a 1 mM sample of V PG L in CD Cl .
Intensity is in arbitrary units, and is normalized between experiments.
Further details may be found in the Experimental Section.
a,b,20
Conclusions
1
. Mirror-Image â-Turns Stabilize Two-Residue Loop
(types I and II) are antithetical to formation of a two-residue
â-Hairpins. Determining the intramolecular hydrogen bonding
pattern is most straightforward in methylene chloride, among
the solvents we have examined, and the intramolecular hydrogen
bonding pattern identifies the backbone fold. Internal hydrogen
bonding information obtained via IR and NMR shows that the
heterochiral peptides and depsipeptides are completely or almost
completely folded into â-hairpins in methylene chloride, while
the all-L isomers experience little or no â-hairpin folding
â-hairpin loop. The relationship between â-turn conformation
and â-hairpin stabilization explains at least in part why several
recently reported dipeptide-mimetic units that are locally achiral
(18) A recently described â-sheet protein design (ref 4d) includes the
use of D-residues at positions intended to form two-residue â-hairpin loops.
(19) We recently reported that the â-hairpin conformation of four-residue
depsipeptide Ac-Gly-Pro-Glyco-Gly-NMe2 in CH2Cl2 is only about 40%
populated at room temperature: Gardner, R. R.; Gellman, S. H. J. Am.
Chem. Soc. 1995, 117, 10411. Comparison of this result with an observation
(
Schemes 2 and 3). These conclusions are strongly supported
D
L
D
by observation of long-range NOE enhancements for the
heterochiral peptides and depsipeptides, but not for their all-L
analogues. The minimal â-hairpins stabilized by mirror-image
turns persist, at least partially, even when hydrogen bonding
competition from the solvent undercuts the driving force for
folding. Thus, the stabilization of two-residue loop â-hairpins
by mirror-image â-turns appears to be quite significant energeti-
cally, and our results suggest that selective incorporation of
reported here, complete population of â-hairpin for V PGlyco L, indicates
that side chains, with the correct configuration, are required on the terminal
residues for complete â-hairpin formation. These terminal residue side
chains presumably decrease the number of non-hairpin conformations
available to the backbone, thus decreasing the loss of conformational entropy
associated with adoption of the hairpin folding pattern.
(
20) Searle, M. S.; Williams, D. H.; Rackman, L. C. Nature Struct. Biol.
1
995, 2, 999. For â-hairpin folding, in water, of a 16-residue peptide that
lacks proline, see: Blanco, F. J.; Rivas, G.; Serrano, L. Nature Struct. Biol.
1994, 1, 584.

Stereochemical Requirements for â-Hairpin Formation
J. Am. Chem. Soc., Vol. 118, No. 29, 1996 6983
3 3
Figure 13. N-H stretch FT-IR data for 10 mM peptide samples in CH CN at room temperature, after subtraction of the spectrum of pure CH CN.
L
L
L
L
-1
L
D
D
L
-1
L
L
L
From left to right: V P A L, maximum at 3368 cm ; V P A L, maximum at 3324 cm ; V PG L, maximum at 3371 and shoulder at 3303
cm ; V PG L, shoulder at 3371 cm and maximum at 3314 cm-1
-
1
L
D
L
-1
.
Figure 14. N-H stretch FT-IR data for 10 mM peptide samples in DMSO at room temperature, after subtraction of the spectrum of pure DMSO.
L
L
L
L
-1
L
D
D
L
-1
L
L
L
-1
L
D
L
From left to right: V P A L, maximum at 3265 cm ; V P A L, maximum at 3267 cm ; V PG L, maximum at 3265 cm ; V PG L, maximum
-
1
at 3286 cm
.
strongly promote â-hairpin conformations in solution:21 in the
absence of a chiral center, the distinction between “common”
and “mirror-image” turns disappears.
than in the heterochiral isomer, but full rationalization of our
observations will require a more detailed analysis.
3. The Ability of Intramolecular Hydrogen Bonds To
Induce Folding Is Limited, Even in Nonpolar Environments.
The lack of significant â-hairpin formation among the all-L
peptides and depsipeptides in methylene chloride indicates that
closing the 14-membered-ring hydrogen bond costs more in
terms of induced strain in the all-L backbones than would be
gained by hydrogen bond formation. We have previously
argued that chemists and biochemists tend to overemphasize
the significance of hydrogen bonds in noncovalently controlled
â-Hairpin conformations have recently been identified in two
short all-L peptides (9 and 16 residues) in aqueous solution,
20,22
both containing a proline residue in the turn region.
In both
cases, the antiparallel â-strands are connected by a three-residue
loop rather than a two-residue loop. We propose that the rigid
L-proline residue in each of these peptides precludes a mirror-
image turn, and therefore disfavors a two-residue â-hairpin loop.
Consistent with this hypothesis, proline is rare in two-residue
9a
5
b
structure; the dramatic distinction in folding properties between
the all-L molecules and their heterochiral isomers in methylene
chloride is a good example of the perils of a hydrogen bond-
centered view. It is difficult or impossible, upon visual
inspection of drawings or physical models, to identify the
features of the all-L backbones that oppose 14-membered-ring
hydrogen bond formation of the four-residue molecules de-
scribed here.
â-hairpin loops in crystalline proteins.
Simple molecular
L
L
L
L
D
L
L
D
mechanics comparison of V P Lac L and V P Lac L sug-
gests that closure of the 14-membered hydrogen bonded ring
requires greater internal steric repulsion in the all-L depsipeptide
(21) (a) D ´ı az, H.; Tsang, K. Y.; Choo, D.; Espina, J. R.; Kelly, J. W. J.
Am. Chem. Soc. 1993, 115, 3790. (b) Gardner, R. R.; Gellman, S. H. J.
Am. Chem. Soc. 1995, 117, 10411. (c) Kemp, D. S.; Li, Z. Q. Tetrahedron
Lett. 1995, 36, 4175, 4179.
(
22) (a) Blanco, F. J.; Jim e´ nez, M. A.; Herranz, J.; Rico, M.; Santoro,
J.; Nieto, J. L. J. Am. Chem. Soc. 1993, 115, 5887. (b) Constantine, K. L.;
Mueller, L.; Andersen, N. H.; Tong, H.; Wandler, C. F.; Friedrichs, M. S.;
Bruccoleri, R. E. J. Am. Chem. Soc. 1995, 117, 10841. (c) Friedrichs, M.
S.; Stouch, T. R.; Bruccoleri, R. E.; Mueller, L.; Constantine, K. L. J. Am.
Chem. Soc. 1995, 117, 10855.
Experimental Section
Infrared Spectroscopy. Spectra were obtained on a Nicolet Model
740 FT-IR spectrometer. IR samples were prepared under anhydrous
conditions; solvents were distilled from CaH and stored over molecular
2

6
984 J. Am. Chem. Soc., Vol. 118, No. 29, 1996
Haque et al.
sieves, compounds and glassware were dried under vacuum overnight,
solid was purified by silica gel chromatography (eluting with 50:49:1
ethyl acetate/chloroform/methanol) to yield 0.125 g (0.381 mmol, 58%)
of the desired dipeptide as a white solid.
and solutions were prepared under an N atmosphere. The pure solvent
2
spectrum for a particular solution was subtracted from the sample
spectrum prior to analysis. Peaks in the amide NH stretch region were
baseline corrected, and analyzed without further manipulation.
NMR Spectroscopy. 1. Aggregation Studies. One-dimensional
spectra for aggregation studies were obtained on either a Varian Unity-
r-Hydroxy Acid Coupling. HO-L-Lac-D-Leu-NMe : Boc-D-Leu-
2
NMe (0.500 g, 1.94 mmol) was deprotected as described above. The
2
activated ester was prepared by addition of 0.253 g of HOSu (2.20
mmol) followed by 0.453 g of DCC (2.20 mmol) to a CH
of L-lactic acid (0.192 g, 2.14 mmol). A white precipitate formed soon
after DCC addition. The solution of H N-D-Leu-NMe and triethyl-
amine in CH Cl was then transferred into the activated ester solution.
2 2
Cl solution
5
00 or a Bruker AM-500 spectrometer. Samples for aggregation studies
were prepared by serial dilution from the most concentrated sample
typically 40 to 50 mM). Dry compounds were dissolved in CD
Cl
previously dried over 3 Å molecular sieves, and samples were prepared
with dry glassware under an N atmosphere.
. Conformational Analysis. NMR samples for conformational
analysis were prepared by dissolving the dry compound in dry
deuterated solvent under an N atmosphere. Samples were then
2
2
(
2
2
2
2
After stirring the resulting solution overnight, the white solid was
filtered off, and the remaining solution was concentrated to a clear
colorless oil. Purification by silica gel chromatography (eluting with
2
2
4% MeOH in CHCl
colorless oil (1.66 mmol, 86%).
HO-D-Lac-D-Leu-NMe : A modified coupling procedure was used
in obtaining this product, since D-lactic acid was purchased as the
lithium salt, which is insoluble in CH Cl . The coupling reaction was
run in 3:2 DMF/CH Cl . The desired product was obtained in a yield
3
) yielded 0.381 g of the desired product as a clear
2
degassed by the freeze-pump-thaw method, and the NMR tubes were
sealed under vacuum. All two-dimensional NMR spectra were obtained
on a Varian Unity-500 spectrometer. Proton signals were assigned via
2
2
2
2
3
12
17
TOCSY spectra, and ROESY or NOESY spectra provided the data
used in the conformational analyses. Two-dimensional spectra were
acquired using standard Varian pulse sequences and hypercomplex
2
2
of 67% after silica gel chromatography.
Ester Bond Formation. Boc-L-Pro-D-Lac-L-Leu-NMe : HO-D-
2
24
phase cycling (States-Haberkorn method ), and the data were processed
with Varian VNMR version 4.3 software. TOCSY spectra were
Lac-L-Leu-NMe (0.130 g, 0.565 mmol) was dissolved in 20 mL of
2
CH Cl , followed by 0.133 g of Boc-L-Pro-OH (0.620 mmol). The
2
2
recorded with 2048 points in t
6 scans per t increment. ROESY and NOESY spectra were recorded
with a similar number of t and t points unless otherwise noted, and
between 32 and 120 scans per t increment, depending on the
1
, 256 or 512 points in t
2
, and either 8 or
resulting clear solution was cooled to 0 °C. 4-(Dimethylamino)pyridine
(0.009 g, 0.074 mmol) was added, followed by 0.128 g of DCC (0.620
mmol), and the solution was stirred at 0 °C for 1 h, then allowed to
warm to room temperature and stirred for another 4 h. The solution
turned cloudy 25 min after DCC addition. The precipitate was filtered
off, and the solution was concentrated under vacuum. The resulting
solid was purified by silica gel chromatography (eluting with ethyl
acetate) to afford 0.158 g of the product as a white solid (0.370 mmol,
65%).
1
2
1
2
2
concentration of the sample. The width of the spectral window
examined was between 4500 and 6000 Hz. Sample concentrations for
two-dimensional spectra were 1 mM in CD
and 10 mM in DMSO-d
. NOESY Buildup Plots: NOESY spectra were obtained for five
2 2 3
Cl , 5 or 10 mM in CD CN,
6
.
3
different mixing times (ranging from 250 to 1250 ms) for each buildup
curve. Intensities were normalized between spectra in the following
manner. For the observed NOE interaction at a particular mixing time,
the volume integrations of the two cross peaks were averaged. The
average NOE volume was then divided by the average of the volume
integrations of the two corresponding diagonal peaks. The resulting
normalized NOE intensity was multiplied by an arbitrary number (in
the cases discussed here, 1000). Normalized NOEs for a specific
interaction were then plotted as a function of mixing time to obtain a
buildup plot.
Capping Procedure. Ac-D-Val-L-Pro-L-Lac-D-Leu-NMe : Tetra-
2
peptide Boc-D-Val-L-Pro-L-Lac-D-Leu-NMe was deprotected with 4
2
N HCl/dioxane as described above. Triethylamine (0.16 mL, 1.14
mmol) was added to the solid, and the mixture was dissolved in 5 mL
of CH Cl . The solution was cooled to 0 °C, and 0.047 mL of acetic
2
2
anhydride (0.500 mmol) was added to the stirring solution. The solution
was stirred at 0 °C for 1 h, then allowed to warm to room temperature
and stirred another 2 h. The organic solution was washed with 20 mL
of 10% HCl (aq), then 2 × 20 mL of 5% NaHCO (aq), and dried with
3
MgSO . The solution was filtered and concentrated under vacuum.
4
X-ray Crystallography. Crystals suitable for X-ray analysis were
obtained by vapor diffusion (over several days) of hexane into a solution
of the particular depsipeptide in ethyl acetate. In several cases it was
necessary to use dried solvents and grow the crystals under anhydrous
conditions.
The resulting crude solid was purified by silica gel chromatography
(eluting with 4% MeOH in CHCl ) to yield 0.152 g of the final product
3
as a white solid (0.325 mmol, 86%). Running the capping reaction in
methanol or with acetyl chloride instead of acetic anhydride resulted
in a reduced yield of the final product.
1
Synthesis. Depsipeptides and peptides were synthesized by standard
dicyclohexylcarbodiimide/N-hydroxysuccinimide (DCC/HOSu) solu-
Characterization. Final compounds were characterized by H and
13
C NMR, FTIR, and high-resolution mass spectroscopy. NMR spectra
25
13
tion-phase coupling procedures. Representative procedures are given
below.
are referenced to the solvent peak of the sample. C spectra were
1
13
acquired with H decoupling, and all C peaks reported are one-carbon
1
1
13
Formation of Activated Ester Boc-L-Ala-OSu. Boc-L-Ala-OH
singlets when H decoupled, unless otherwise noted. H and C NMR
spectra reported for characterization were obtained on purified samples
on a Varian Unity-500 NMR spectrometer or a Bruker AC-300 or AM-
500 NMR spectrometer. Ester CdO stretch and amide I IR bands
observed for solutions of compounds in methylene chloride are listed,
where “(sh)” indicates that the peak is a shoulder on an adjacent peak.
All final compounds were very pure, as no impurities were detectable
2 2
(0.150 g, 0.791 mmol) was dissolved in 5 mL of CH Cl . HOSu (0.091
g, 0.791 mmol) was then added to the solution, followed by 0.163 g of
DCC (0.791 mmol). Several minutes after DCC addition, the formation
of a white precipitate was observed. The solution was stirred at room
temperature for 1 h. The crude activated ester was then allowed to
react with the selected nucleophile.
1
13
Peptide Deprotection and Coupling Procedure. Boc-L-Ala-L-Leu-
via H or C NMR spectroscopy.
L
L
L
L
1
2
NMe : Boc-L-Leu-NMe
2
(0.170 g, 0.659 mmol) was dissolved in 2
2 2 2
Ac- V P Lac L-NMe : H NMR (CD Cl , 500 MHz) δ 0.92 (d, J
mL of 4 N HCl/dioxane, and the solution was stirred at room
temperature for 1 h. The remaining HCl/dioxane was then removed
by bubbling N through the solution. CH Cl (2-3 mL) was added to
2 2 2
the resulting solid, followed by 0.110 mL of triethylamine (0.791
mmol). The resulting solution was added to the solution containing
the activated ester Boc-L-Ala-OSu, and the combined reaction solution
was stirred overnight (10-12 h). The white solid was filtered off, and
the remaining solution was concentrated to yield a white solid. The
) 6.6 Hz, 3 H), 0.93 (d, J ) 6.9 Hz, 3 H), 0.97 (d, J ) 6.4 Hz, 3 H),
1.00 (d, J ) 6.7 Hz, 3 H), 1.44 (d, J ) 7.0 Hz, 3 H), 1.51 (m, 4 H),
1.98 (s, 3 H), 2.07 (m, 3 H), 2.29 (m, 1 H), 2.90 (s, 3 H), 3.02 (s, 3 H),
3.65 (m, 1 H), 3.82 (m, 1 H), 4.49 (dd, J ) 8.4, 6.1 Hz, 1 H), 4.63 (dd,
J ) 9.0, 6.7 Hz, 1 H), 4.87 (td, J ) 8.6, 5.0 Hz, 1 H), 5.09 (q, J ) 7.0
1
3
Hz, 1 H), 6.27 (d, J ) 8.7 Hz, 1 H), 7.13 (d, J ) 7.3 Hz, 1 H).
NMR (CD Cl , 125.8 MHz) δ 15.27, 17.96, 18.06, 18.16, 19.33, 22.08,
23.24, 23.38, 24.92, 25.38, 29.35, 31.02, 35.84, 37.17, 42.32, 47.89,
C
2
2
-
1
5
1
9.65, 71.56, 169.98, 170.12, 170.85, 171.89, 171.95. FTIR ν (cm )
(
(
23) Braunschweiler, L.; Ernst, R. R. J. Magn. Reson. 1983, 53, 521.
24) States, D. J.; Haberkorn, R. A.; Ruben, D. J. J. Magn. Reson. 1982,
+
754 (sh), 1744, 1674, 1644, 1632 (sh), 1509. EI-MS m/z (M )
468.2948, observed 468.2960.
: 1H NMR (CD
Cl , 500 MHz) δ 0.92 (d, J
) 6.6 Hz, 6 H), 0.93 (d, J ) 4.9 Hz, 3 H), 0.96 (d, J ) 6.8 Hz, 3 H),
40 4 6
calculated for C23H N O
4
8, 286.
25) Bodanszky, M.; Bodanszky, A. The Practice of Peptide Synthesis;
Springer-Verlag: New York, 1984.
D
L
L
D
(
Ac- V P Lac L-NMe
2
2
2

Stereochemical Requirements for â-Hairpin Formation
J. Am. Chem. Soc., Vol. 118, No. 29, 1996 6985
1
4
3
.32 (d, J ) 7.1 Hz, 3 H), 1.34 (m, 1 H), 1.88 (ddd, J ) 13.8, 11.1,
.3 Hz, 1 H), 1.92 (s, 3 H), 2.00 (m, 4 H), 2.27 (m, 1 H), 2.93 (s, 3 H),
.13 (s, 3 H), 3.68 (dt, J ) 10.3, 6.9 Hz, 1 H), 4.04 (dt, J ) 10.2, 6.6
13.6, 9.2, 4.3 Hz, 1 H), 1.63 (m, 1 H), 1.80 (ddd, J ) 14.0, 9.9, 4.4
Hz, 1 H), 1.92 (s, 3 H), 2.01 (m, 3 H), 2.07 (m, 1 H), 2.25 (m, 1 H),
2.91 (s, 3 H), 3.14 (s, 3 H), 3.64 (m, 1 H), 4.20 (m, 1 H), 4.26 (t, J )
8.7 Hz, 1 H), 4.34 (p, J ) 7.0 Hz, 1 H), 4.38 (dd, J ) 8.6, 4.0 Hz, 1
Hz, 1 H), 4.46 (dd, J ) 8.9, 4.4 Hz, 1 H), 4.61 (t, J ) 9.9 Hz, 1 H),
.95 (ddd, J ) 11.1, 8.5, 3.8 Hz, 1 H), 5.16 (q, J ) 7.1 Hz, 1 H), 7.64
4
H), 4.86 (m, 1 H), 6.32 (d, J ) 7.9 Hz, 1 H), 7.26 (d, J ) 7.6 Hz, 1
1
3
13
(d, J ) 8.5 Hz, 1 H), 8.40 (d, J ) 9.8 Hz, 1 H). C NMR (CD
2
Cl
25.8 MHz) δ 18.10, 18.63, 19.37, 21.56, 22.54, 23.44, 24.98, 25.30,
9.71, 30.93, 35.89, 37.34, 40.67, 47.44, 47.94, 56.34, 59.99, 70.68,
2
,
H), 7.56 (d, J ) 6.4 Hz, 1 H). C NMR (CD
2
Cl
2
, 125.8 MHz) δ
1
2
1
1
4
17.71, 19.04, 19.30, 21.85, 22.76, 23.41, 24.95, 25.28, 29.76, 30.29,
35.89, 37.43, 40.95, 47.78, 48.09, 48.97, 58.03, 61.84, 171.25, 172.13,
-
1
-1
70.08, 170.45, 170.91, 172.70, 173.45. FTIR ν (cm ) 1743, 1666,
644 (sh) 1629, 1539. EI-MS m/z (M ) calculated for C23
172.71, 172.82, 173.19. FTIR ν (cm ) 1680 (sh), 1656, 1635, 1533,
+
+
H
40
N
4
O
6
1507. HR-LSIMS; calculated for C23
H N O
41 5 5
+ H 468.3186, observed
68.2948, observed 468.2960.
468.3186.
D
L
D
D
: 1H NMR (CD
D
L
15
L
D
1
Ac- V P Lac L-NMe
2
2
Cl
2
, 500 MHz) δ 0.92 (d, J
2 2 2
Ac- V P( N) A L-NMe : H NMR (CD Cl , 300 MHz) δ 0.89
)
6.6 Hz, 6 H), 0.93 (d, J ) 6.9 Hz, 3 H), 0.94 (d, J ) 6.7 Hz, 3 H),
(d, J ) 6.3 Hz, 3 H), 0.90 (d, J ) 6.3 Hz, 3 H), 0.94 (d, J ) 6.6 Hz,
3 H), 1.02 (d, J ) 7.0 Hz, 3 H), 1.32 (dd, J ) 7.4, 2.9 Hz, 3 H), 1.41
(ddd, J ) 13.8, 9.0, 4.8, Hz, 1 H), 1.65 (m, 1 H), 1.86 (m, 1 H), 1.92
(s, 3 H), 2.01 (m, 4 H), 2.26 (m, 1 H), 2.91 (s, 3 H), 3.17 (s, 3 H), 3.62
(dt, J ) 10.1, 7.0 Hz, 1 H), 4.28 (m, 2 H), 4.43 (m, 2 H), 4.87 (ddd,
1
1
2
1
4
7
.48 (d, J ) 7.2 Hz, 3 H), 1.51 (m, 1 H), 1.67 (p, J ) 6.4 Hz, 1 H),
.85 (ddd, J ) 13.9, 9.5, 4.7 Hz, 1 H), 2.01 (m, 2 H), 2.04 (s, 3 H),
.15 (m, 3 H), 2.89 (s, 3 H), 3.18 (s, 3 H), 3.63 (td, J ) 8.5, 5.8 Hz,
H), 3.79 (td, J ) 7.9, 5.9 Hz, 1 H), 4.41 (dd, J ) 7.9, 4.4 Hz, 1 H),
.57 (t, J ) 9.0 Hz, 1 H), 4.90 (td, J ) 9.2, 5.2 Hz, 1 H), 5.07 (q, J )
J ) 10.3, 8.1, 4.4 Hz, 1 H), 6.15 (dd, J ) 90.8, 8.8 Hz, 1 H), 7.42 (d,
.0 Hz, 1 H), 7.82 (d, J ) 9.5 Hz, 1 H), 8.04 (d, J ) 8.6 Hz, 1 H). 13
C
J ) 7.7 Hz, 1 H), 7.81 (d, J ) 8.5 Hz, 1 H). C NMR (CDCl
13
, 75.4
3
NMR (CDCl
3
, 75.4 MHz) δ 17.40, 18.07, 18.84, 21.81, 22.45, 22.73,
MHz) δ 17.35, 18.72, 19.02, 21.57, 22.36, 23.05, 24.38, 24.74, 29.14,
15
13
2
5
1
4.37, 24.95, 28.37, 30.88, 35.71, 37.08, 40.26, 46.46, 47.07, 55.54,
8.80, 70.76, 169.84, 170.49, 170.86, 171.99 (2 C). FTIR ν (cm
749, 1665, 1641 (sh), 1630, 1538. (EI-MS m/z (M ) calculated for
29.49, 35.62, 37.09, 40.33, 47.45, 47.51, 48.11 (d, N- C J ) 17
15 13 15 13
-
1
)
Hz), 57.55, 61.20 (d, N- C J ) 12 Hz), 170.70 (d, N- C J ) 25
+
-1
Hz), 171.80, 172.30, 172.61, 172.96. FTIR ν (cm ) 1680 (sh), 1656,
1
5
+
C
23
H
40
L
N
4
O
6
468.2948, observed 468.2949.
1635, 1530. HR-LSIMS; calculated for C23
H
41 NN
O
4 5
+ H 469.3156,
L
L
1
2 2 2
Ac- V PGlyco L-NMe : H NMR (CD Cl , 500 MHz) δ 0.89 (d,
observed 469.3143.
L
L
L
1
J ) 6.7 Hz, 3 H), 0.91 (d, J ) 6.4 Hz, 3 H), 0.95 (d, J ) 6.4 Hz, 3 H),
Ac- V PG L-NMe : H NMR (CD Cl , 500 MHz) δ 0.88 (d, J )
2 2 2
0.97 (d, J ) 6.7 Hz, 3 H), 1.52 (m, 2 H), 1.58 (m, 1 H), 1.99 (s, 3 H),
2.05 (m, 2 H), 2.10 (m, 2 H), 2.29 (m, 1 H), 2.91 (s, 3 H), 3.05 (s, 3
6.9 Hz, 3 H), 0.90 (d, J ) 6.6 Hz, 3 H), 0.95 (d, J ) 6.6 Hz, 3 H),
0.96 (d, J ) 6.7 Hz, 3 H), 1.47 (m, 2 H), 1.59 (m, 1 H), 1.96 (m, 1 H),
1.99 (s, 3 H), 2.06 (m, 3 H), 2.19 (m, 1 H), 2.91 (s, 3 H), 3.06 (s, 3 H),
3.62 (m, 1 H), 3.71 (dd, J ) 16.9, 4.9 Hz, 1 H), 3.78 (m, 1 H), 4.05
(dd, J ) 16.9, 6.9 Hz, 1 H), 4.36 (t, J ) 6.5 Hz, 1 H), 4.63 (dd, J )
9.0, 6.4 Hz, 1 H), 4.94 (dt, J ) 8.6, 5.6 Hz, 1 H), 6.35 (d, J ) 9.0 Hz,
H), 3.67 (m, 1 H), 3.85 (m, 1 H), 4.43 (t, J ) 7.6 Hz, 1 H), 4.44 (d,
J ) 15.4 Hz, 1 H), 4.63 (dd, J ) 9.0, 6.6 Hz, 1 H), 4.73 (d, J ) 15.4
Hz, 1 H), 4.96 (dd, J ) 15.6, 7.0 Hz, 1 H), 6.35 (d, J ) 8.2 Hz, 1 H),
1
3
7
2
4
.40 (d, J ) 8.2 Hz, 1 H). C NMR (CD
2 2
Cl ) δ 17.78, 18.92, 21.99,
2.86, 24.38, 25.06, 28.90, 29.47, 31.33, 35.58, 36.93, 42.02, 46.74,
7.37, 55.29, 59.07, 62.35, 166.38, 170.08, 171.02, 171.28, 171.72.
13
1 H), 6.79 (bt, 1 H), 7.03 (d, J ) 7.2 Hz, 1 H). C NMR (CDCl
3
,
125.8 MHz) δ 17.87, 19.34, 22.05, 22.95, 23.14, 24.65, 25.14, 28.43,
-
1
+
FTIR ν (cm ) 1757, 1673, 1644, 1633, 1508. EI-MS m/z (M )
calculated for C22 454.2791, observed 454.2782.
31.29, 35.75, 37.06, 42.09, 42.99, 47.31, 47.79, 55.68, 60.51, 168.43,
-
1
H
38
N
4
O
6
170.10, 171.76, 171.87, 172.24. FTIR ν (cm ) 1694, 1673, 1645,
D
L
D
1
+
Ac- V PGlyco L-NMe
2
: H NMR (CD
2
Cl
2
, 500 MHz) δ 0.92 (d,
1630, 1534, 1507. EI-MS m/z (M ) calculated for C22
H N
39 5
O
5
453.2951, observed 453.2958.
Ac- V PG L-NMe : H NMR (CD Cl , 500 MHz) δ 0.91 (d, J )
2 2 2
L
D
L
1
6.6 Hz, 3 H), 0.92 (d, J ) 6.6 Hz, 3 H), 0.95 (d, J ) 6.6 Hz, 3 H),
1.00 (d, J ) 6.9 Hz, 3 H), 1.43 (ddd, J ) 13.7, 8.9, 4.6 Hz, 1 H), 1.64
(m, 1 H), 1.78 (ddd, J ) 14.0, 9.8, 4.5 Hz, 1 H), 1.94 (s, 3 H), 2.04
(m, 4 H), 2.21 (m, 1 H), 2.97 (s, 3 H), 3.15 (s, 3 H), 3.57 (dd, J )
17.3, 5.7 Hz, 1 H), 3.63 (m, 1 H), 4.02 (dd, J ) 17.4, 7.3 Hz, 1 H),
4.08 (m, 1 H), 4.35 (m, 1 H), 4.93 (ddd, J ) 10.2, 8.4, 4.5 Hz, 1 H),
6.56 (bt, J ) 6.0 Hz, 1 H), 7.45 (d, J ) 8.2 Hz, 1 H), 7.72 (d, J ) 8.4
Hz, 1 H). 13C NMR (CD
2 2
Cl , 125.8 MHz) δ 18.94, 19.34, 21.89, 22.63,
23.39, 24.94, 25.26, 29.66, 31.01, 35.98, 37.49, 41.13, 42.90, 47.39,
48.13, 57.61, 61.79, 168.99, 172.00, 172.40, 172.52, 173.20. FTIR ν
-1
+
(cm ) 1690, 1659, 1632, 1538, 1518. EI-MS m/z (M ) calculated for
+
C
22
H
39
N
5
O
5
+ H 454,3029, observed 454.3021.
Acknowledgment. This work was supported in part by the
H), 3.04 (s, 3 H), 3.60 (m, 1 H), 3.74 (m, 1 H), 4.33 (p, J ) 7.2 Hz,
Army Research Office and by the National Science Foundation.
T.S.H. was the recipient of a National Research Service Award
(T32 GM08923). The authors thank Dr. Doug Powell and Mr.
Kurt Schladetzky for their assistance in obtaining the crystal-
lographic data.
1
H), 4.51 (dd, J ) 8.0, 3.9 Hz, 1 H), 4.61 (dd, J ) 8.9, 6.3 Hz, 1 H),
.89 (td, J ) 9.1, 3.8 Hz, 1 H), 6.17 (d, J ) 9.5 Hz, 1 H), 6.71 (d, J
4
)
1
3
8.4 Hz, 1 H), 7.13 (d, J ) 7.3 Hz, 1 H). C NMR (CDCl
3
, 125.7
MHz) δ 17.93, 18.23, 19.27, 21.97, 22.96, 23.33, 24.70, 25.14, 27.93,
3
1
1
4
1.44, 35.74, 37.00, 42.37, 47.34, 47.80, 49.02, 55.68, 60.04, 170.08,
-1
71.03, 171.76, 172.03, 172.24. FTIR ν (cm ) 1675, 1645, 1627 (sh),
Supporting Information Available: Representative ROESY
and NOESY spectra for several compounds and structure factor
sets for the four crystal structures (45 pages). See any current
masthead page for ordering and Internet access instructions.
+
504. HR-LSIMS; calculated for C23
H N O
41 5 5
+ H 468.3186, observed
68.3189.
L
D
D
L
2 2 2
Ac- V P A L-NMe Cl , 500 MHz) δ 0.90 (d, J )
: 1H NMR (CD
6
1
.4 Hz, 3 H), 0.91 (d, J ) 6.1 Hz, 3 H), 0.96 (d, J ) 6.7 Hz, 3 H),
.03 (d, J ) 6.7 Hz, 3 H), 1.30 (d, J ) 7.6 Hz, 3 H), 1.39 (ddd, J )
JA960429J