Stabilization of a cis amide bond in a host-guest complex

Article abstract of DOI:10.1021/ja9607805

Macrobicycle 12 has been synthesized and its binding properties with a range of N-acetyl amino acid carboxylates (as tetrabutylammonium salts) have been studied. While the binding results showed little selectivity for the various substrates investigated, detailed NMR studies have revealed that D-amino acid substrates bind predominantly on the outside of the macrobicycle cavity by a strong carboxylate-thiourea interaction, whereas L-amino acid substrates bind predominantly on the inside of the cavity also establishing a strong carboxylate-thiourea interaction but with the acetyl amide in a cis configuration. Molecular modeling studies suggest that the energetic penalty associated with adopting a cis amide configuration in the host-guest complex is compensated by intermolecular hydrogen bonds between the cis amide and the rim of the macrobicycle.

Full text of DOI:10.1021/ja9607805

10220
J. Am. Chem. Soc. 1996, 118, 10220-10227
Stabilization of a Cis Amide Bond in a Host-Guest Complex
Glen J. Pern´ıa,† Jeremy D. Kilburn,*,† Jonathan W. Essex,†
Russell J. Mortishire-Smith,*,‡ and Michael Rowley‡
Contribution from the Department of Chemistry, UniVersity of Southampton,
Southampton, SO17 1BJ, U.K., and Department of Medicinal Chemistry, Merck Sharp and
Dohme Research Laboratories, Neuroscience Research Centre, Terlings Park, Eastwick Road,
Harlow, Essex, CM20 2QR, U.K.
ReceiVed March 11, 1996X
Abstract: Macrobicycle 12 has been synthesized and its binding properties with a range of N-acetyl amino acid
carboxylates (as tetrabutylammonium salts) have been studied. While the binding results showed little selectivity
for the various substrates investigated, detailed NMR studies have revealed that D-amino acid substrates bind
predominantly on the outside of the macrobicycle cavity by a strong carboxylate-thiourea interaction, whereas L-amino
acid substrates bind predominantly on the inside of the cavity also establishing a strong carboxylate-thiourea interaction
but with the acetyl amide in a cis configuration. Molecular modeling studies suggest that the energetic penalty
associated with adopting a cis amide configuration in the host-guest complex is compensated by intermolecular
hydrogen bonds between the cis amide and the rim of the macrobicycle.
The energetic preference for a secondary amide to adopt a
trans configuration in preference to a cis configuration is of
primary importance in determining the secondary structure of
peptides.1 The energetic preference, for example, for trans
N-methylacetamide over the cis form, is estimated, based on
experimental work2 and on calculations,3 to be ∼10 kJ mol-1
in water and is little affected by change of solvent. As a
consequence cis amide bonds are rarely found in linear peptides
and are usually associated with proline residues.4 Non-prolyl
secondary amides can, however, be constrained to adopt the
cis configuration by incorporation into cyclic structures,5 and
Peggion et al. have reported that the linear heptadecapetide
bombolitin, which has an all-trans amide structure in free
solution, adopts a cis secondary Ile-Lys amide bond when
incorporated into a micelle.6 The discovery that the binding
proteins for the immunosuppressant agents cyclosporin A,
rapamycin, and FK506 are peptidy L-prolyl cis/trans isomerases
has further stimulated interest in the cis/trans isomerization of
peptides.7 In this paper we describe a macrobicyclic receptor
12 which selectively binds N-acetyl L-amino acid carboxylates
within the receptor cavity. In so doing, the acetyl amide adopts
a cis configuration, stabilized by two N-H‚‚‚OdC hydrogen
bonds to the side wall of the macrobicycle.8
Figure 1. Schematic of macrobicyclic receptor.
the intermolecular interactions involved in small molecule-
peptide complexes are of direct relevance to many biological
peptide-protein interactions and may also lead to new biosen-
sors, therapeutics and catalysts for peptide hydrolysis. In our
own efforts to develop novel receptors for amino acids and
peptides we have prepared a range of macrobicycles which
feature a specific binding site for carboxylic acid functionality
at the base of the cavity (represented schematically in Figure
1).8,10e
By incorporating additional amide functionality around the
rim of such a macrobicyclic structure, we intended to provide
further hydrogen bonding sites, suitably preorganized to interact
with guests such as amino acid derivatives, bound within the
macrobicyclic cavity. In macrobicycle 12 the carboxylate
binding site is provided by a thiourea moiety, which has been
shown to provide a strong binding site for tetraalkylammonium
carboxylates, even in relatively competitive solvents such as
The development of synthetic receptors for peptides and
amino acid derivatives9,10 is of considerable interest because
† University of Southampton.
‡ Merck, Sharp and Dohme.
X Abstract published in AdVance ACS Abstracts, September 15, 1996.
(1) Schultz, G. E.; Schirman, R. H. Principles of Protein Structure;
Springer-Verlag: New York, 1979.
(8) A preliminary account of some of this work has been published:
Pern´ıa, G. J.; Kilburn, J. D.; Rowley, M. J. Chem. Soc., Chem. Commun.
1995, 305-306.
(9) For recent reviews on molecular recognition, see: (a) Dowden, J.;
Kilburn, J. D.; Wright, P. Contemp. Org. Synth. 1995, 2, 289. (b) Molecular
Recognition (Tetrahedron Symposia No. 56). Hamilton, A. D., Ed.;
Tetrahedron 1995, 51. (c) Webb, T. H.; Wilcox, C. S. Chem. Soc. ReV.
1993, 22, 383.
(2) Radzicka, A.; Pedersen, L.; Wolfenden, R. Biochemistry 1988, 27,
4538-4541.
(3) Jorgensen, W. L.; Gao, J. J. Am. Chem. Soc. 1988, 110, 4212-4216.
(4) Stewart, D. E.; Sarkar, A.; Wampler, J. E. J. Mol. Biol. 1990, 214,
253-260.
(5) (a) Mierke, D. F.; Yamazaki, T.; Said-Nejad, O. E.; Felder, E. R.;
Goodman, M. J. Am. Chem. Soc. 1989, 111, 6847-6849. (b) Kessler, H.;
Anders, U.; Schudok, M. J. Am. Chem. Soc. 1990, 112, 5908-5916. (c)
Kukumaran, D. K.; Prorok, M.; Lawrence, D. S. J. Am. Chem. Soc. 1991,
113, 706-707. (d) Gramberg, D.; Robinson, J. A. Tetrahedron Lett. 1994,
35, 861.
(6) Bairaktari, E.; Mierke, D. F.; Mammi, S.; Peggion, E. J. Am. Chem.
Soc. 1990, 112, 5383.
(7) (a) Schreiber, S. L. Science 1991, 251, 283-287. (b) Fischer, G.;
Schmid, F. X. Biochemistry 1990, 29, 2205-2209.
(10) For recent work on synthetic receptors for amino acid and peptide
derivatives see, for example: (a) Schneider, H.-J. Angew. Chem., Int. Ed.
Engl. 1993, 32, 848. (b) Cristofaro, M. F.; Chamberlin, A. R. J. Am. Chem.
Soc. 1994, 116, 5089. (c) Kuroda, Y.; Kato, Y.; Higashioli, T.; Hasegawa,
J.; Kawanami, S.; Takahashi, M.; Shiraishi, N.; Tanabe, K.; Ogoshi, H. J.
Am. Chem. Soc. 1995, 117, 10950. (d) Gennari, C.; Nestler, H. P.; Salom,
B.; Still, W. C. Angew. Chem., Int. Ed. Engl. 1995, 34, 1765. (e) Waymark,
C. P.; Kilburn, J. D.; Gillies, I. Tetrahedron Lett. 1995, 36, 3051. (f) Flack,
S. S.; Kilburn, J. D. Tetrahedron Lett. 1995, 36, 3409. (g) Yoon, S. S.;
Still, W. C. Tetrahedron 1995, 51, 567.
S0002-7863(96)00780-9 CCC: $12.00 © 1996 American Chemical Society

Stabilization of a Cis Amide Bond
J. Am. Chem. Soc., Vol. 118, No. 42, 1996 10221
Scheme 1
dimethyl sulfoxide.11 The thiourea also has the potential to be
converted into a guanidinium group, which could again serve
as a carboxylate binding site. A biarylmethane unit forms the
rigidifying part of the rim of the macrobicycle, and chirality
and amide functionality are introduced Via two lysine deriva-
tives.
diisopropylethylamine in acetonitrile gave macrobicycle 12 in
variable yield but generally in the range 20-40%. The
conversion of diacid 11 to final macrobicycle 12 was carried
out without purification of the intermediates which were
unstable, particularly to chromatography. The synthesis is thus
short, reasonably efficient, and provides considerable scope for
variation of the various building blocks, which should allow
access to a range of related macrobicycles.
Binding studies with macrobicycle 12 were carried out with
various acylated amino acids as the tetrabutylammonium salts.
Attempts to determine binding constants by conventional NMR
titration experiments,15 with deuterochloroform as solvent, were
hindered by the fact that the various NH signals that were
monitored during the titration shifted into the aromatic region
of the NMR spectrum. Instead, by partitioning the guests
between water and chloroform we determined the distribution
of the various guests between the two phases (by integration of
the 1H NMR against an internal standard). Similarly the
distribution of guest between water and a chloroform solution
of the macrobicycle was determined, and thus we were able to
calculate binding constants (Table 1), by analogy with the picrate
extraction method developed by Cram.16
Inspection of the binding results indicated that macrobicycle
12 showed little selectivity for amino acid configuration (R or
S) or for the various amino acid side chains, with the possible
exception of the L-lysine derivative. Furthermore, binding of
simple carboxylates (benzoate, hexanoate) by macrobicycle 12
was essentially as strong as the binding of the more function-
alized amino acid derivatives, suggesting that binding of all the
substrates studied was almost entirely the consequence of a
strong interaction between the carboxylate and the thiourea unit.
At first sight, therefore, it seemed that our carefully prepared
macrobicycle was little more than an over-elaborate thiourea
derivative capable of binding a number of carboxylate salts.
The synthesis of macrobicycle 12 proved to be relatively
straightforward (Scheme 1). Methyl 4-methylbenzoate 1 was
converted to bromide 2 in good yield, provided care was taken
to avoid dibromination, by slow addition of slightly less than 1
equiv of bromine to the solution of 1, with irradiation by a 150
W bulb. Treatment of 4-bromobenzonitrile 3 with n-butyl-
lithium at -100 °C, followed by addition of trimethylborate,
cleanly gave the boronic acid 4. A palladium mediated Suzuki
coupling12 of bromide 2 with boronic acid 4, in dimethoxy-
ethane,13 gave the nitrile ester 5 in consistently good yields of
70-82%. Selective reduction of 5 with borane-methyl sulfide
complex gave the corresponding amine 6 which was coupled
to NR-tert-butyloxycarbonyl-Nꢀ-benzyloxycarbonyl lysine. Best
yields for this coupling were achieved using 1-(3-dimethylami-
nopropyl)-3-ethylcarbodiimide methiodide as coupling reagent.
Selective removal of the benzyloxycarbonyl protecting group
then gave amine 8, which was converted in two steps to the
thiourea 10, Via the isothiocyanate 9.14 Hydrolysis of the methyl
esters of 10 to give diacid 11 was best carried out using equal
volumes of 1 M LiOH‚H2O and dioxane at 45 °C. Conversion
to the corresponding bispentafluorophenyl ester, removal of the
tert-butyloxycarbonyl protecting groups and, finally, cyclization
by slow addition of the bisTFA salt to a refluxing solution of
(11) (a) Fan, E.; Van Arman, S. A.; Kincaid, S.; Hamilton, A. D. J. Am.
Chem. Soc. 1993, 115, 369. (b) Smith, P. J.; Reddington, M. V.; Wilcox,
C. S. Tetrahedron Lett. 1992, 33, 6085.
(12) (a) Suzuki, A. Acc. Chem. Res. 1982, 15, 178. (b) Martin, A. R.;
Yang, Y. Acta Chem. Scand. 1993, 47, 221.
(13) Gronowitz, S.; Bobosik, V.; Lawitz, K. Chem. Scri. 1984, 23, 120.
(14) (a) Dyson, G. M.; George, H. J. J. Am. Chem. Soc. 1924, 1702. (b)
Rasmussen, C. R.; Villani, Jr., F. J.; Weaner, L. E.; Reynolds, B. E.; Hood,
A. R.; Hecker, L. R.; Nortey, S. O.; Hanslin, A.; Costanzo, M. J.; Powell,
E. T.; Molinari, A. J. Synthesis 1988, 456. (c) Scro¨eder, D. C. Chem. ReV.
1955, 55, 193.
(15) Wilcox, C. S. Frontiers in Supramolecular Organic Chemistry and
Photochemistry; Schneider, H.-J., Durr, H., Eds.; VCH: Weinheim, 1991;
pp 123-143.
(16) Kypa, E. P.; Helgeson, R. C.; Madan, K.; Gokel, G. W.; Tarnowski,
T. L.; Moore, S. S.; Cram, D. J. J. Am. Chem. Soc. 1977, 199, 2564.

10222 J. Am. Chem. Soc., Vol. 118, No. 42, 1996
Pern´ıa et al.
Table 2. Intermolecular NOEs and Distances for the Complex
Table 1. Association Constants for Macrobicycle 12 with Various
Tetrabutylammonium Carboxylates in CDCl3
between Macrobicycle 12 and Tetrabutylammonium
N-Acetyl-L-phenylalanine (at 285 K)
substrate
(tetrabutylammonium salt)
Ka/103 mol-1
-∆G/kJ mol-1
distance from
unrestrained
molecular
distance from
restrained
molecular
NOE upper
distancea
N-Ac-glycine
N-Ac-L-alanine
68.6 ( 7.9
16.9 ( 5.9
14.6 ( 2.0
22.0 ( 2.8
13.3 ( 7.0
9.6 ( 7.0
27.1 ( 0.3
23.7 ( 0.9
23.3 ( 0.3
24.3 ( 0.3
23.1 ( 0.4
22.3 ( 0.2
21.5 ( 0.2
22.7 ( 0.5
21.1 ( 0.3
28.7 ( 0.3
22.1 ( 0.5
26.6 ( 0.9
24.9 ( 0.3
crosspeak
assignment
constraint/Å
dynamics/Å
dynamics/Å
N-Ac-D-alanine
N-Ac-L-phenylalanine
N-Ac-D-phenylalanine
N-Ac-L-asparagine
N-Ac-D-asparagine
N-Ac-L-glutamine
NR-Ac-L-histidine
NR-Ac-L-lysine
Hk T Hw
Hk T Hx
Hk T Hy
Hk T Hz
Hl T Hw
Hl T Hy
Hl T Hz
H
H
H
H
5.0
5.0
4.0
5.0
5.0
5.0
5.0
4.0
4.0
5.0
4.0
4.0
4.0
4.0
4.0
5.6
5.3
4.0
4.5
3.4
4.8
4.6
5.2
4.3
4.0
3.1
4.1
4.7
4.7
3.2
3.2
4.9
5.7
3.8
3.5
4.2
5.3
5.0
3.7
4.5
3.7
4.5
3.9
5.0
4.3
4.0
3.3
4.3
4.6
4.5
3.0
3.2
5.3
5.7
3.6
3.4
4.4
6.8 ( 2.9
11.1 ( 8.0
5.8 ( 8.0
130.0 ( 20.0
8.9 ( 1.6
m T Hw
m T Hx
m T Hy
m T Hz
Cbz-glycylglycine
benzoic acid
hexanoic acid
55.4 ( 17.4
28.1 ( 3.3
Hn T Hw
Hn′ T Hw
Hn′ T Hx
Hn′ T Hy
Ho T Hi′
Ho T Hw
Ho T Hx
Ho T Hy
Ho T Hz
Hp T Hi′
3.0
>5.0b
>5.0b
3.0
3.0
4.0
a NOE distance upper bounds were determined from crosspeak
intensities by calibration against fixed interproton distances (geminal
protons and aromatic protons Hw T Hx and Hy T Hz). Constraints
were binned into 3.0, 4.0, and 5.0 Å categories. b No NOE crosspeak
was observed. c Macrobicycle 12 and L-phenylalanine carboxylate
1
showing H labeling.
Figure 2. Upfield shifts (ppm) for the various protons indicated in
1:1 complexes with macrobicycle 12, relative to the signals of the
unbound substrates.
However, 1H NMR spectra of 1:1 mixtures of macrobicycle
12 with the tetrabutylammonium salts of D- or L-N-acetyl alanine
or D- or L-N-acetyl phenylalanine showed markedly different
chemical shifts for the L-amino acid derivatives compared with
the D-amino acid derivatives, suggesting substantially different
modes of binding for the two enantiomeric series. In all the
complexes the signal for the thiourea NH was shifted substan-
tially downfield (2-3 ppm), relative to the uncomplexed
macrobicycle, indicating the formation of hydrogen bonds and
consistent with a strong carboxylate-thiourea interaction. 1H
chemical shifts for the aliphatic resonances of the D-alanine
derivative in the 1:1 complex (∼20 mM, >90% bound) were
not significantly perturbed from values observed for the
uncomplexed amino acid in CDCl3. In the 1:1 complex of
macrobicycle 12 with the L-alanine derivative, however, all four
alanine resonances were shifted dramatically upfield by up to
2.5 ppm compared with the uncomplexed amino acid (Figure
2).
Similar trends were observed in the 1:1 complexes of
macrobicycle 12 with the L- and D-phenylalanine derivatives.
In the complex formed with the D-phenylalanine derivative only
minor differences in chemical shifts were detected for the
“bound” substrate compared with the uncomplexed phenylala-
nine derivative in CDCl3 (Figure 2). For the L-phenylalanine
derivative however, the acetyl methyl, NH, CRH, and CâH
resonances were shifted dramatically upfield (Figure 2) although
the aromatic resonances for both the guest and macrobicycle
were minimally perturbed. This suggests that edge-face and
face-face π-π interactions are unlikely to contribute signifi-
cantly to association, and indeed it is unlikely that the aromatic
side chain is internalized to any great degree.
outside of the macrobicycle, while the L-amino acid derivatives
are bound within the cavity afforded by the macrobicycle. The
large observed upfield shifts are presumably a consequence of
shielding by the aromatic rings which compose the walls of
the cavity.
Detailed 2D NMR studies were carried out (at 285 K and
300 K) to probe the solution conformation of these complexes.
Complete assignments for the 1:1 complexes with macrobicycle
12 and the L-phenylalanine and the L-alanine derivatives were
obtained using DQF-COSY17 and TOCSY18 experiments to
define spin systems which were connected using data from
ROESY19 spectra. While NOESY20 experiments performed on
the 1:1 complexes failed to exhibit useful inter- or intramolecular
NOE crosspeaks (presumably due to unfavorable correlation
times), ROESY spectra obtained at 285 K contained a wealth
of structure-defining intermolecular contacts (Table 2).21
(17) Jeener, J.; Meier, B. H.; Backmann, P.; Ernst, R. R. J. Chem. Phys.
1979, 71, 4546.
(18) Rance, M. J. Magn. Reson. 1987, 74, 557.
(19) Bax, A.; Davis, D. G. J. Magn. Reson. 1985, 63, 207.
(20) Rance, M.; Sorensen, O. W.; Bodenhausen, G.; Wagner, E. R. R.;
Wu¨thrich, K. Biochem. Biophys. Res. Commun. 1983, 117, 479.
(21) To eliminate possible errors in interpretation caused by artefacts or
spurious intensities, ROESY spectra were acquired with three different
transmitter offsets (7.5, 5.0, and 2.5 ppm), each with two mixing times
(200 and 400 ms). Each spectrum exhibited qualitatively comparable
intensities for the key structure defining ROEs. Spin-lock field strength
and temperature also had a minimal effect on the relative ROE intensities.
The observed chemical shifts are consistent with binding of
the D-amino acid derivatives to the thiourea moiety on the

Stabilization of a Cis Amide Bond
J. Am. Chem. Soc., Vol. 118, No. 42, 1996 10223
Table 3. Intramolecular NOEs and Distances for Tetrabutyl
Ammonium N-Acetyl-L-phenylalanine in the Complex with
Macrobicycle 12 (at 285 K)
distance from
unrestrained
molecular
distance from
restrained
molecular
NOE upper
distancea
constraint/Å
crosspeak
assignment
dynamics/Å
dynamics/Å
Hk T Hl
Hk T Hm
Hh T Ho
Hk T Hp
Hk T Hq
Hl T Hn’
Hl T Ho
5.0
3.0
4.0
5.0
5.0
3.0
4.0
3.0
3.9
2.9
3.4
5.0
5.8
2.9
4.6
2.8
3.8
2.6
3.5
4.9
5.6
2.8
4.2
2.9
H
m T Ho
a NOE distance upper bounds were determined from crosspeak
intensities by calibration against fixed interproton distances (geminal
protons and aromatic protons Hw T Hx and Hy T Hz). Constraints
were binned into 3.0, 4.0, and 5.0 Å categories.
Figure 3. ROESY experiment (at 300 K) on 1:1 complex of 12 with
N-Ac-L-Phe-CO2- showing intermolecular NOEs between the aliphatic
protons of the guest and the aromatic sidewalls of the host.
For the 1:1 complex with the L-phenylalanine derivative, the
phenylalanine ring is apparently close to the B ring of the
macrobicycle as evidenced by crosspeaks from Hy and Hz to
the phenylalanine ortho protons Ho but not from Hw or Hx.
The upfield shifted acetyl protons (Hk) of the amino acid guest
gave NOEs of varying intensity to all four macrobicycle
aromatic signals and intraresidue NOEs to the phenylalanine
ring. Similarly there are a number of NOEs of varying
intensities from the the NH (Hl), R proton (Hm), and the benzylic
protons (Hn, Hn') of the L-phenylalanine derivative to the
macrobicycle aromatic signals (Figure 3).
In contrast, ROESY spectra for the 1:1 complex with the
D-phenylalanine derivative indicated that there were no signifi-
cant intermolecular contacts between the macrobicycle and the
guest. The corresponding ROESY spectra for the 1:1 complex
with the L-alanine derivative similarly indicated numerous
intermolecular contacts particularly to the biarylmethane side-
walls of 12, but the ROESY spectra for the 1:1 complex with
the D-alanine derivative was essentially devoid of intermolecular
crosspeaks. These differences are again consistent with binding
of the D-amino acid derivatives to the thiourea moiety on the
outside of the macrobicycle, while the L-amino acid derivatives
are bound within the cavity afforded by the macrobicycle.
For the 1:1 complex with the L-phenylalanine derivative, a
full set of intramolecular NOEs for the guest molecule were
also identified (Table 3).
Surprisingly, the acetyl methyl (Hk) gave crosspeaks to all
three aromatic signals Ho, Hp, and Hq and a particularly strong
NOE to the CRH (Hm) but only a weak NOE to the NH (Hl)
(Figure 4).
In peptides, the classic diagnostic for a cis amide bond is the
observation of a dRR(i, i+1) NOE.22 In the complex between
macrobicycle 12 and the L-phenylalanine derivative, this NOE
corresponds to that present between the acetyl methyl (Hk) and
the CRH proton (Hm) of the L-phenylalanine derivative and is
Figure 4. ROESY experiment (at 300 K) on 1:1 complex of 12 with
N-Ac-L-Phe-CO2 showing intramolecular NOEs for the aliphatic
protons of the guest substrate.
-
the most intense intramolecular NOE observed with a volume
two orders of magnitude larger than that between the acetyl
methyl (Hk) and the NH (Hl). The conformation of the
L-phenylalanine derivative in the complex was examined by a
systematic search (using SYBYL23 ) around the four rotatable
bonds, imposing distance constraints obtained from NOE
volumes. In each of the conformers which satisfied the NOE
distance constraints, the acetyl amide bond adopts a cis
configuration, driven largely by the short acetyl methyl (Hk) to
CRH (Hm) and aromatic (Ho, Hp, and Hq) distances. Inspection
of the ROESY spectra for the 1:1 complex with the D-
phenylalanine derivative showed intramolecular crosspeaks
consistent with the expected trans amide and were essentially
identical to the crosspeaks observed in the ROESY spectrum
of uncomplexed tetrabutylammonium salt of N-acetyl phenyl-
alanine. Again, inspection of the corresponding ROESY spectra
for the 1:1 complex with the L- and D-alanine derivatives
similarly conclusively showed that the L-amino acid guest is
bound with the acetyl amide in the cis configuration, but the
D-amino acid guest is bound with the acetyl amide in the trans
configuration.
In 2D spectra of the complex between 12 and the L-
phenylalanine derivative at 285 K, protons Hw, Hx, Hy, and Hz
can be seen to be in exchange with a second set of resonances
representing approximately 30% of the intensity of the major
form (Figure 5).
(22) Dyson, H. J.; Wright, P. E. 2-Dimensional NMR Spectroscopy,
Croasmun, Carlson, Eds.; VCH: 1994; p 672.
(23) Tripos, Inc.: St. Louis, Missouri.

10224 J. Am. Chem. Soc., Vol. 118, No. 42, 1996
Pern´ıa et al.
substrate, while the loss in symmetry is not reflected in the
number of resonances observed in NMR spectra of the complex.
Molecular modeling, however, has been used to probe the mode
of binding of the L-phenylalanine derivative within the macro-
bicyclic cavity. The geometry of the host-guest complex was
examined using a combination of simulated annealing calcula-
tions and free molecular-dynamics using the MacroModel
program.24 A modified version of the OPLS* force field25 was
adopted, and the effect of solvent was included through the use
of GB/SA chloroform.26 The L-phenylalanine derivative, with
a cis amide bond, was docked into the macrobicycle by eye,
such that the carboxylate group was bound to the thiourea
moiety, in accord with experimental observations, and then
energy minimized. Ten simulated annealing calculations of 1
ns involving slow cooling from 600 to ∼0 K were performed,
followed by a 5 ns simulation at 300 K to examine the behavior
of the complex at room temperature. The free molecular-
dynamics simulation was repeated with the distance between
the acetyl amide proton and carboxylate oxygen of the guest
restrained to 2 Å (this eliminates the unrealistically short contact-
distance usually observed between this atom pair which results
from the 1,4 van der Waals scale-factor used in the OPLS*
force-field). The 10 minimum energy structures obtained from
the simulated-annealing calculations all show similar binding
geometries (Figure 6).
Figure 5. Exchange peaks assigned in a 100 ms TOCSY spectrum (at
285 K) of the 1:1 complex of 12 with N-Ac-L-Phe-CO2 at 285K.
Starred (*) peaks denote exchange between the major and minor
conformations of the host described in the text. Circled peaks denote
through-bond correlations also observed in DQF-COSY experiments.
-
This second set of signals does not exhibit the intermolecular
NOEs to the L-phenylalanine derivative, observed for the major
form, and has chemical shifts comparable with those obtained
for the unbound macrobicycle and for the macrobicycle bound
externally (when bound externally the macrobicycle appears to
adopt a conformation close to that in the unbound state, as
evidenced by spectra of 1:1 complexes of the macrobicycle with
the D-phenylalanine and D-alanine derivatives which shows
minimal changes to the chemical shifts for the macrobicycle).
Furthermore, the resolved aliphatic resonances for the L-
phenylalanine derivative (Hk and Hn') represent only 70% of
the expected intensity compared with the signals for the
tetrabutylammonium countercation, although no exchange part-
ners for the L-phenylalanine derivative are evident in either the
1D or 2D spectra. One possible explanation for these results
is that the L-phenylalanine derivative is associated with the
macrobicycle both internally (∼70%) and externally (∼30%).
The loss in intensity of the aliphatic resonances from the
L-phenylalanine derivative could result from rearrangement of
the guest, involving conversion of the cis amide to the trans
form on transferring from internal to external binding, on a time
scale which leads to exchange broadening.
The structures differ in two respects. First, the hydrocarbon
chains adjacent to the thiourea moiety adopt different conforma-
tions, and, second, the aromatic ring of the L-phenylalanine
derivative is found to lie in two binding geometries over the
aromatic rings of the macrobicycle but excluded from the main
cavity. The L-phenylalanine derivative is bound to the macro-
bicycle by essentially a total of six hydrogen bonds. The acetyl
amide carbonyl oxygen is simultaneously hydrogen bonded to
two of the amide NH’s of the macrobicycle, while the
carboxylate function forms a total of four hydrogen bonds to
the thiourea moiety and the other two amide NH’s of the
macrobicycle. Of particular note, the acetyl amide NH of the
guest does not form any hydrogen bonds with the macrobicycle.
In the course of the free molecular-dynamics simulation of
the host-guest complex (both with and without the 2 Å distance
restraint between the amide proton and the carboxylate oxygen
of the guest) the aromatic ring of the guest was observed to
interconvert between the two binding geometries identified in
the simulated annealing study but was observed to lie predomi-
nantly over the B rather than the A ring of the macrobicycle, in
accord with the NMR data. The number of hydrogen bonds
identified fluctuated between 4 and 6, with subtle changes in
Direct determination of the solution structures of the 1:1
complexes between macrobicycle 12 and L-amino acid substrates
using the NMR data is nontrivial, primarily because the 2-fold
symmetry of the macrobicycle is lost upon binding a chiral
-
Figure 6. Stereopair of the 10 minimum energy structures of the complex of 12 with N-Ac-L-Phe-CO2 obtained from the simulated annealing
calculations.

Stabilization of a Cis Amide Bond
J. Am. Chem. Soc., Vol. 118, No. 42, 1996 10225
the hydrogen-bonding pattern being observed. However, the
fundamental binding geometry of the macrobicycle remained
unaffected. A detailed comparison with the NOE distance
constraints was made by monitoring the appropriate H/H
distances during the simulation (Tables 2 and 3). For atoms
where a united-atom representation was used in the modeling,
hydrogen atoms were added in standard geometries. Average
distances were obtained through either r-6 averaging, or a
combination of r-6 and r-3 averaging, as described in the
Experimental Section. The agreement between simulation and
experiment is generally good. In the unrestrained simulation,
the molecular dynamics distances exceed the NMR values by
greater than 0.5 Å in seven cases out of a total of 29. Four
such violations exist in the restrained simulation. To determine
whether these results are converged, the restrained simulation
was repeated using a different starting configuration in which
the aromatic ring of the L-phenylalanine derivative was in an
alternative binding geometry. Analysis of this trajectory gave
average distances that agreed with the previous simulation to
within 0.2 Å and a total of only two violations with the
experimental data. The agreement between simulation and
experimental distances therefore supports the reliability of the
proposed structural model. In particular, comparison of the Hy
T Ho and Hz T Ho distances with the Hw T Ho and Hx T Ho
distances (from molecular dynamics) clearly shows the close
association of the aromatic ring of the guest with the B ring of
the macrocycle. A representative structure of the host-guest
complex is given in Figure 7, showing the preferred orientation
of the L-Phe aromatic ring and the hydrogen-bond pattern.
In addition to satisfying the NOE distance constraints, the
mode of binding derived from molecular modeling places the
benzylic methylene of the guest in close proximity to the biaryl
methane side wall, while the acetyl methyl group is shielded
by both the biaryl methane side wall and the phenyl residue of
the guest, consistent with the greater upfield shift for this acetyl
methyl signal (-2.10 ppm) compared to that observed in the
1:1 complex between 12 and the L-alanine derivative (-1.02
ppm). The model for the complex also places the amino acid
side chain directed away from the cavity of the macrobicycle
and explains why there is little side chain discrimination for
the amino acid derivatives, except perhaps for lysine where the
longer chain may allow the amino group to reach around and
establish a further hydrogen bonding interaction with the
macrobicycle. The structure presented in Figure 7 is therefore
consistent with the NMR data and in the absence of any further
experimental data constitutes the best available model for this
system.
alanine derivative binds to the thiourea through the side of the
cavity, with hydrogen bonds from the carboxylate to the thiourea
and to two of the amide NH’s in the sidewall of the macrobi-
cycle, forming a hydrogen bonding pattern similar to that already
identified for the L-phenylalanine derivative. For the D-substrate,
however, no hydrogen bonding to the amide group of the
substrate is observed. The structure proposed for the binding
of the D-phenylalanine derivative should be treated with caution,
however, owing to the absence of corroborating experimental
evidence (from NOE’s, etc.). L-Amino acid substrates, on the
other hand, bind predominantly on the inside of the cavity also
establishing a strong carboxylate thiourea interaction, but with
the acetyl amide in a cis configuration, as adjudged by the
dramatic upfield shifts observed in the 1:1 complexes, and the
strong inter- and intramolecular NOE interactions. The overall
binding constants for the L- substrates, determined from extrac-
tion experiments, are only slightly greater than those measured
for the D-substrates. Such a small energetic preference is
consistent with the observation that the L-substrates occupy the
cavity approximately 70% of the time and presumably bind on
the outside of the cavity (as for the D-substrates) the other 30%
of the time. The free energy difference between an amide in
the cis and the trans configuration is estimated to be ∼10 kJ
mol-1 2,3
This energetic penalty is paid for in the complex, at
.
least in part, by two hydrogen bonds from the amide NH’s in
the sidewall of the macrobicycle to the acetyl carbonyl of the
guest. In addition the carboxylate moiety of the guest appears
to form a total of four hydrogen bonds to the thiourea NH’s
and the two remaining amide NH’s in the side wall of the
macrobicycle. This view of course neglects additional entropic
costs on binding L-substrates in a much more ordered complex
than that for the D-substrates and neglects compensating positive
van der Waals interactions between the substrates and the
biarylmethane units of the macrobicycle.
In conclusion we have observed the stabilization of a cis
amide, in a host-guest complex which selectively internalizes
L-amino acid derivatives. The use of NMR and molecular
modeling has provided a detailed picture of the structure of the
host-guest complex in solution and will now allow the rational
design of selective receptors by incorporating structural features
which block the binding of D-amino acid substrates on the
exterior of the cavity.
Experimental Section
General Method for Elucidating Binding Constants Following
Cram’s Procedure.16 Extraction Experiment in the Absence of
Host. CDCl3 (2.0 mL) was added to a sample of the substrate (N-
acetyl amino acid tetrabutyl ammonium salt, AA-‚TBA+) (typically
20 mg) in D2O (2.0 mL), and the biphasic mixture was thoroughly
mixed using a vortex machine. After separation of the two layers, a
400 µL aliquot of the D2O solution was removed by syringe and added
to a 400 µL aliquot of a standard solution (0.01 M) dioxane in D2O
and a 1H NMR spectrum of the resulting solution was recorded.
Similarly, a 400 µL aliquot of the CDCl3 solution was added to a 400
µL aliquot of dioxane (2.5 × 10-3 M) in CDCl3, and a 1H NMR
spectrum of the resulting solution was recorded. The amount, and hence
concentration, of the substrate in both the CDCl3 and the D2O phase
was determined by comparison of the integrals of signals from dioxane
and the substrate (using peaks from the tetrabutylammonium fragment).
From these measurements the distribution constant (Kdsdefined in eq
1) for the substrate was determined. Five independent experiments
were performed for each substrate to give an average value for Kd and
a standard deviation.
Thus, D-amino acid substrates (D-alanine and D-phenylalanine
derivatives) seem to bind predominantly on the outside of the
macrobicycle cavity by a strong carboxylate-thiourea interac-
tion (worth 23-27 kJ mol-1 of free binding energy).27 The
slight upfield shifts of the D-substrates observed in the 1:1
complexes might indicate that the substrates are bound within
the cavity to a finite extent but not substantially. A short series
of simulated annealing calculations suggest that the D-phenyl-
(24) MacroModel V5.0; Mohamadi, F.; Richards, N. G. J.; Guida, W.
C.; Liskamp, R.; Lipton, M.; Caufield, C.; Chang, G.; Hendrickson, T.;
Still, W. C. J. Comput. Chem. 1990, 11, 440.
(25) (a) Jorgensen, W. L.; Tirado-Rives, J. J. Am. Chem. Soc. 1988, 110,
1657. (b) Jorgensen, W. L.; Nguyen, T. B. J. Comput. Chem. 1993, 14,
195. (c) Duffy, E. M.; Severance, D. L.; Jorgensen, W. L. Isr. J. Chem.
1993, 33, 323. (d) Jorgensen, W. L. J. Phys. Chem. 1986, 90, 6379.
(26) Still, W. C.; Tempczyk, A.; Hawley, R. C. J. Am. Chem. Soc. 1990,
112, 6127.
(27) Using an identical extraction procedure to that used for the
tetrabutylammonium salts with the macrobicyclic, the binding constant
between tetrabutylammonium benzoate and dibenzyl thiourea in CDCl3 was
estimated to be (33.5 ( 13.3) × 103 mol-1 (-∆Ga ) 25.4 ( 1.3 kJ mol-1).
Extraction Experiment in the Presence of Host. An accurately
weighed sample of macrobicycle 8 (typically 5 mg) in CDCl3 (2.0 mL)
was added to a sample of the substrate (N-acetyl amino acid tetra-
butylammonium salt) (typically 20 mg) in D2O (2.0 mL), and the
biphasic mixture was thoroughly mixed using a vortex machine. After

10226 J. Am. Chem. Soc., Vol. 118, No. 42, 1996
Pern´ıa et al.
Figure 7. Stereopairs of a single configuration of the complex of 12 with N-Ac-L-Phe-CO2- taken from the restrained molecular dynamics simulation
viewed (a) from the side, (b) from above, and (c) with some of the macrobicycle structure removed to emphasize the intermolecular hydrogen
bonding network.
separation of the two layers, a 400 µL aliquot of the D2O solution was
removed by syringe and added to a 400 µL aliquot of a standard solution
The association constant for the macrocycle (M) and amino acid
substrates (Kaseq 3) is then the ratio of Kd and Ke (eq 4).
1
(5.0 × 10-3 M) of dioxane in D2O, and a H NMR spectrum of the
Kd
resulting solution was recorded. Similarly, a 400 µL aliquot of the
CDCl3 solution was added to a 400 µL aliquot of dioxane (2.5 × 10-3
[AA-]D O + [TBA+]D O
{
} [AA-‚TBA+]CDCl
(1)
{\
[AA-]D O + [TBA+]D O + [M]CDCl } [AA-‚M‚TBA+]CDCl (2)
3
2
2
3
1
M) in CDCl3, and a H NMR spectrum of the resulting solution was
Ke
recorded. The amount, and hence concentration, of the substrate in
both the CDCl3 and the D2O phase was determined by comparison of
the integrals of signals from dioxane and the substrate (using peaks
from the tetrabutylammonium fragment). From these measurements
the distribution constant (Kesdefined in eq 2) for the substrate was
determined following the analysis of Cram.16 Again, five independent
experiments were performed for each substrate to give an average value
for Ke and a standard deviation.
2
2
3
Ka
{
[AA-‚TBA+]CDCl + [M]CDCl
} [AA-‚M‚TBA+]CDCl (3)
3
3
3
Ka ) Ke/Kd
(4)
The following assumptions are made in the above analysis: (a)

Stabilization of a Cis Amide Bond
J. Am. Chem. Soc., Vol. 118, No. 42, 1996 10227
substrate salt is completely dissociated in D2O; (b) the substrate salt is
associated, but monomeric, in CDCl3; (c) essentially no macrobicyclic
receptor is distributed into the D2O phase; (d) the macrobicycle forms
a 1:1 complex with the amino acid substrates. Assumptions (a) and
(b) were validated by measuring Kd across a range of initial concentra-
tions (4-60 mM) of substrate in D2O, which gave constant values for
Kd, and by carrying out NMR dilution experiments for the substrates
which indicated no appreciable dimerization or aggregation. Assump-
tion (c) was validated since none of macrobicycle 12 could be seen in
occurs in the NMR experiment. Consequently, although the NMR
results give a single distance constraint for a given atom pair, the
simulation provides up to 12 distances to be monitored for a given
pair (the three Hk protons with the four Hw protons, for example). In
making the comparison with the NMR distance constraints, the
simulation distances were averaged according to:
-6 -1/6
rav ) ( Ri,j
)
where rav is the final averaged simulation distance for pairs i,j; Ri,j is
the instantaneous value of the interatomic separation of atom pairs i, j;
and ... indicates that an average is calculated over the course of the
molecular dynamics trajectory. The coordinates used in the trajectory
analysis were generated approximately every 1 ps. In the case of the
Hk protons, however, a single r-3 averaged distance was evaluated for
the interactions with other protons for each frame of the trajectory.
These distances were then averaged using the r-6 procedure already-
described. This hybrid protocol is necessary since the Hk protons are
likely to be spinning more rapidly than every 200 ps.
1
the H NMR spectra of the D2O solutions, and assumption (d) was
validated by a Job plot28 for the complexation of the L-phenylalanine
derivative with macrobicycle 12.
Molecular Modeling Protocol. The complex formed between
macrobicycle 12 and the L-phenylalanine derivative was modeled using
the MacroModel program24 on a Silicon Graphics workstation. The
OPLS* parameter set was selected for these studies. However,
electrostatic and van der Waals parameters were assigned by comparison
with the original OPLS papers.25 The van der Waals and electrostatic
parameters used in the study are given in the supporting information.
Solvation was included in the simulations through the GB/SA con-
tinuum model of chloroform.26 All possible nonbonded interactions
were included in the calculations. The L-phenylalanine derivative, with
a cis amide bond, was docked into the host by eye, such that the
carboxylate group was bound to the thiourea moiety, in accord with
experimental observations, and then energy minimized. To determine
a range of plausible geometries for the complex, 10 successive
simulated-annealing molecular-dynamics simulations were performed.
Each simulation consisted of 1 ns of molecular dynamics, with slow
cooling from an initial temperature of 600 K to a final temperature of
∼0 K. A time-step of 1.5 fs was adopted, and all bonds were
constrained using the SHAKE algorithm.29 No restraints were applied
to force the complex to adopt a predetermined geometry. The 10
structures obtained at the end of the simulated annealing calculations
are energy minima. To characterize the behavior of the complex under
experimental conditions, one of the structures produced by the annealing
calculations was selected as the starting point for a 5 ns molecular-
dynamics simulation. The simulation was performed at a temperature
of 300 K in conjunction with a temperature-bath relaxation constant
of 0.2 ps.30 The OPLS* parameter set scales 1,4 van der Waals
The choice of r-6 over r-3 averaging is based on the time scales for
molecular motions within the system. From the Debye expression31
for the rotational correlation time
τrot ) 4πa3η/3kT
τrot for the host-guest system is estimated to be approximately 200
ps; a, the molecular radius, is estimated to be 7.5 Å; η, the solvent
viscosity, is 0.542 × 10-3 kg m-1 s-1 at 25 °C;32 k is the Boltzmann
constant; and T is the temperature. Inspection of the molecular
dynamics trajectories indicates that the macrocycle aromatic rings do
not flip in 5 ns. The host aromatic ring is observed to flip approximately
every 0.5-1.0 ns. However the influence of solvent viscosity was not
included in the calculations, and this can only serve to increase the
time over which ring-flips occur. Consequently molecular tumbling
is occurring faster than internal conformational transitions, and r-6
averaging is therefore appropriate.33 The exception to this is for the
methyl Hk protons for which a hybrid averaging protocol needs to be
adopted as noted above.
Acknowledgment. We are grateful to S. Thomas and N.
Aboitiz (MSD) for assistance acquiring NMR spectra and to
MSD and EPSRC for a studentship to G.J.P. J.W.E. is a Royal
Society University Research Fellow. J.W.E. would like to thank
Dr. A. Bonvin (Yale University) and D. C. Lim (Yale
University) for helpful discussions and assistance with the
computational studies.
1
interactions to /8th their full strength, and this results in the acetyl
amide-proton carboxylate-oxygen distance being unrealistically short.
Given the critical role of these hydrogen-bonding groups in mediating
binding, the effect of this force-field artefact was investigated by
repeating the 5 ns molecular dynamics simulation with a harmonic
restraint of 2 Å, 150 kcal mol-1 Å-2, applied to the amide-proton
carboxylate-oxygen distance. Hydrogen bonds were assigned in the
structures using the MacroModel default criteria. Furthermore, if a
hydrogen bond between the acetyl amide-proton and carboxylate-
oxygen was identified, it was discarded as being physically unrealistic.
To compare the simulation results with the NMR distance constraints,
hydrogen atoms were added to united-atom carbons in standard
geometries. A 5 ns simulation is too short for the formally equivalent
hydrogen atoms of the system to undergo full motional averaging, as
Supporting Information Available: Details of acquisition
of NMR data, full experimental details for the synthesis and
characterization of compounds 2 and 4-12, and nonbonded
parameters (charge, σ, ꢀ) for macrobicycle 12 and tetrabutyl-
ammonium N-acetyl L-phenylalanine, used in the molecular
modeling (12 pages). See any current masthead page for
ordering and Internet access instructions.
(28) (a) Connors, K. A. Binding Constants, The Measurement of
Molecular Complex Stability; Wiley: 1987; pp 24-28. (b) Job, A. Annales
de Chimie 1928, 9, 113.
(29) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J. Comput. Phys.
1977, 23, 327.
(30) Berendsen, J. H. C.; Postma, J. P. M.; van Gunsteren, W. F.; DiNola,
A.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684.
JA9607805
(31) Atkins, P. W. Physical Chemistry, 3rd ed.; OUP: Oxford, 1986; p
621.
(32) Weast, R. C. Handbook of Chemistry and Physics; CRC Press: Boca
Raton, FL, 1980.
(33) Tropp, J. J. Chem. Phys. 1980, 72, 6035.