Mechanistic insights on the magnesium(II) ion-activated reduction of methyl benzoylformate with chelated NADH peptide β-lactam models

Article abstract of DOI:10.1021/jo901236d

(Chemical Equation Presented) Mechanistic details of the Mg²⁺ ion-activated enantioselective reduction of methyl benzoylformate have been investigated at a B3LYP/6-31G* theory level, using peptide NADH models 1 rigidified with a β-lactam ring. Computation of the reaction pathway revealed important structural differences between the intermediate NADH/Mg ²⁺/ArCOCO₂R ternary complexes 3 and the corresponding transition states leading to enantiomeric methyl mandelates. Thus, ternary complexes showed the dihydronicotinamide moiety placed quasiequatorial to a seven-membered chelation pseudoplane including the two amide carbonyls and the Mg²⁺ cation, whereas productive transition states were strongly deformed with the dihydronicotinamide group oriented quasiaxial to the chelation pseudoplane. This chelation model was further applied to acyclic nonrigidified NADH models and, based on the fluxional mobility of the peptide chain bonds, experimental enantioselectivities were correctly predicted. Parallel experiments were also conducted in deuterated acetonitrile, using NMR techniques, to study the structure of the binary complexes 2 (NADH/Mg²⁺) and ternary complexes 3 (NADH/Mg²⁺/PhCOCO₂Me). Finally, owing to the incorporation of two diastereotopic trimethylsilyl NMR-tags in the β-lactam-NADH peptidomimetics, a nonproductive ternary complex predicted by calculations could be observed and its structure characterized on the basis of ROESY experiments and molecular modeling.

Full text of DOI:10.1021/jo901236d

pubs.acs.org/joc
Mechanistic Insights on the Magnesium(II) Ion-Activated Reduction of
Methyl Benzoylformate with Chelated NADH Peptide β-Lactam Models
ꢀ
Jesus M. Aizpurua,* Claudio Palomo,* Raluca M. Fratila, Pablo Ferron, Ana Benito,
ꢀ
Enrique Gomez-Bengoa, Jose I. Miranda, and Jose I. Santos
ꢀ
ꢀ
´ ´
ꢀ
Departamento de Quımica Organica-I, Universidad del Paıs Vasco, Joxe Mari Korta R&D Center. Avda,
Tolosa-72, 20018 San Sebastiaꢀn, Spain
jesusmaria.aizpurua@ehu.es
Received June 11, 2009
Mechanistic details of the Mg2+ ion-activated enantioselective reduction of methyl benzoylformate
have been investigated at a B3LYP/6-31G* theory level, using peptide NADH models 1 rigidified
with a β-lactam ring. Computation of the reaction pathway revealed important structural differences
between the intermediate NADH/Mg2+/ArCOCO2R ternary complexes 3 and the corresponding
transition states leading to enantiomeric methyl mandelates. Thus, ternary complexes showed the
dihydronicotinamide moiety placed quasiequatorial to a seven-membered chelation pseudoplane
including the two amide carbonyls and the Mg2+ cation, whereas productive transition states were
strongly deformed with the dihydronicotinamide group oriented quasiaxial to the chelation pseudo-
plane. This chelation model was further applied to acyclic nonrigidified NADH models and, based on
the fluxional mobility of the peptide chain bonds, experimental enantioselectivities were correctly
predicted. Parallel experiments were also conducted in deuterated acetonitrile, using NMR techni-
ques, to study the structure of the binary complexes 2 (NADH/Mg2+) and ternary complexes 3
(NADH/Mg2+/PhCOCO2Me). Finally, owing to the incorporation of two diastereotopic trimethyl-
silyl NMR-tags in the β-lactam-NADH peptidomimetics, a nonproductive ternary complex pre-
dicted by calculations could be observed and its structure characterized on the basis of ROESY
experiments and molecular modeling.
Introduction
zoylformates with a chiral coenzyme NADH model 1
(Scheme 1), a large number of such compounds have been
described.2 Various strategies have also been developed to
achieve directional and orientational control in these
redox reactions. Among the more successful, the use of
inherently chiral 4-methyl-1,4-dihydropyridines 1-(I)3 or
The nicotinamide adenine dinucleotide phosphate
NAD(P)H/NAD(P)+ redox system is of major importance
for several biological processes such as photosynthesis,
glycolysis, fatty acid synthesis, citric acid cycle, and amino
acid metabolism. Many organic and bioorganic chemists
have been challenged to mimic this type of biochemical
reaction, and since Ohno and co-workers1 reported the
first Mg2+-assisted asymmetric reduction of alkyl ben-
(2) Reviews on NADH models: (a) Burgess, V. A.; Davies, S. G.; Skerlj,
R. T. Tetrahedron: Asymmetry 1991, 2, 299–328. (b) Dupas, G.; Levacher, V.;
ꢀ
Bourguignon, J.; Queguiner, G. Heterocycles 1994, 39, 405–429. (c) Wang,
N. X.; Zhao, J. Synlett 2007, 2785–2791.
(1) (a) Ohnishi, Y.; Numakunai, M.; Ohno, A. Tetrahedron Lett. 1975,
3183–3184. (b) Ohnishi, Y.; Kagami, M.; Ohno, A. J. Am. Chem. Soc. 1975,
97, 4766–4768.
(3) (a) Ohno, A.; Ikeguchi, M.; Kimura, T.; Oka, S. J. Chem. Soc., Chem.
Commun. 1978, 328–329. (b) De Kok, P. M.; Bastiaansen, L. A.; van Lier,
P. M.; Vekemans, J. A.; Buck, H. M. J. Org. Chem. 1989, 54, 1313–1320.
DOI: 10.1021/jo901236d
r
Published on Web 07/31/2009
J. Org. Chem. 2009, 74, 6691–6702 6691
2009 American Chemical Society

Aizpurua et al.
JOCArticle
SCHEME 1. General Pathway for the Reduction of Alkyl Ben-
zoylformates with Chiral NADH Models 1 in the Presence of
Magnesium Cation
3-sulfinyl-1,4-dihydropyridines 1-(II)4 can be mentioned.
However, chelation control through bidentate models 2,
retaining the original 1,4-dihydronicotinamide and including
some ancillary Mg2+-coordinating groups (e.g., CdO, OH),
remains the closest artificial system to mimic the enzymatic
chiral field of biological NADH reductions.5
The general reaction mechanism (Scheme 1) is assumed to
involve the coordination of chiral NADH models 1 with
magnesium ion to afford complexes 2, which are the actual
reducing agents of alkyl benzoylformates through the unst-
able ternary entities 3.1a,2b The NADH/Mg2+/PhCOCO2R
entities formed from chelated NADH models 2 are often
defined indistinctly as “transient ternary complexes” or “tran-
sition states” and are assumed to govern the enantioselectivity
of the hydrogen transfer step. However, despite the excellent
enantiomeric excesses attained in many instances, the struc-
tural reasons invoked to explain the sense of the facial
discrimination remain vague and purely empirical (Figure 1).
For example, (a) the coordination geometry of magnesium
ion varies from tetrahedral to octahedral, (b) the complexa-
tion sites of NADH models seldom include the nitrogen atom
of the dihydropyridine ring5 or peptide carbonyls,6,7 (c) the
benzoylformate ketoester moiety is represented interacting
with the magnesium cation either in a s-trans conformation8
or in a chelated s-cis conformation7,9 and, (d) most models
include a carbonyl-activating interaction of the magnesium
cation with the ketone group of alkyl benzoylformates, in line
FIGURE 1. Original drafts of ternary entities 3-(I-VI) proposed
by several authors to explain the facial discrimination observed for
the reduction of alkyl benzoylformates with chelated chiral NADH
models. “Hsyn rule” (bottom): the carboxamide CdO group and
the transferring hydrogen are syn-oriented in the enzymatic transi-
tion states.
with the mechanism accepted for alcohol dehydrogenase
enzymes.10 Indeed, it is generally assumed that most of the
deformations found in the enzyme-bonded 1,4-dihydronico-
tinamide moiety can be translated to the enzyme-free NADH/
Mg2+/PhCOCO2R models from the ternary complexes
formed by NADH coenzyme, Zn2+-containing L-lactate de-
hydrogenase enzyme and pyruvate substrate. According to
X-ray analysis10b the enzyme-bound carboxamide group is
about 30° out-of-plane with respect to the dihydropyridine
ring. In addition, extensive theoretical calculations11 suggest
(4) (a) Imanishi, T.; Hamano, Y.; Yoshikawa, H.; Iwata, C. J. Chem.
Soc., Chem. Commun. 1988, 473–475. (b) Obika, S.; Nishiyama, T.;
Tatematsu, S.; Miyashita, K.; Imanishi, T. Tetrahedron 1997, 53, 3073–3082.
(c) Li, J.; Liu, Y. C.; Deng, J. G. Tetrahedron: Asymmetry 1999, 10, 4343–
4347. (d) Li, J.; Liu, Y. C.; Deng, J. G.; Li, X. Z.; Cui, X.; Li, Z. Tetrahedron:
Asymmetry 2000, 11, 2677–2682.
(5) Ohno, A.; Kimura, T.; Yamamoto, H.; Kim, S. G.; Oka, S.; Ohnishi,
Y. Bull. Chem. Soc. Jpn. 1977, 50, 1535–1538.
(6) For previous conformationally restricted NADH peptides, see:
(a) Baba, N.; Oda, J.; Inouye, Y. J. Chem. Soc., Chem. Commun. 1980,
815–817. (b) Seki, M.; Baba, N.; Oda, J.; Inouye, Y. J. Am. Chem. Soc. 1981,
103, 4613–4615. (c) Seki, M.; Baba, N.; Oda, J.; Inouye, Y. J. Org. Chem.
1983, 48, 1370–1373.
(7) Saito, R.; Naruse, S.; Takano, K.; Fukuda, K.; Katoh, A.; Inouye, I.
Org. Lett. 2006, 8, 2067–2070.
(8) Talma, A. J.; Jouin, P.; De Vries, J. G.; Trootswijk, C. B.; Werumeus
Buning, G. H.; Waninge, J. K.; Visscher, J.; Kellogg, R. M. J. Am. Chem.
Soc. 1985, 107, 3981–3997.
(10) (a) Eklund, H.; Branden, C. I. In Biological Macromolecules and
Assemblies; Wiley: New York, 1985. (b) Eklund, H.; Samama, J. P.; Jones,
T. A. Biochemistry 1984, 33, 5982–5996.
(11) According to the original Benner’s model, this phenomenon would
be caused by a “reverse anomeric effect” that is unlikely to occur in 1,4-
dihydronicotinamides without N-glycosidic groups, see: (a) Nambiar, P. K.;
Stauffer, D. M.; Kolodziej, P. A.; Benner, S. A. J. Am. Chem. Soc. 1983, 105,
5886–5890. (b) Wu, Y. D.; Houk, K. N. J. Am. Chem. Soc. 1991, 113, 2353–
2358. (c) Almarsson, O.; Bruice, T. C. J. Am. Chem. Soc. 1993, 115, 2125–
2138. (d) Luo, J.; Bruice, T. C. J. Am. Chem. Soc. 2001, 123, 11952–11959.
(9) (a) Kanomata, N.; Nakata, T. Angew. Chem., Int. Ed. Engl. 1997, 36,
1207–1211. (b) Kanomata, N.; Nakata, T. J. Am. Chem. Soc. 2000, 122,
4563–4568.
6692 J. Org. Chem. Vol. 74, No. 17, 2009

Aizpurua et al.
JOCArticle
that the 1,4-dihydropyridine ring adopts a flat boat confor-
mation in the enzymatic process, facilitating the stereospecific
transfer of the pseudoaxial C4-H hydride placed in a syn
disposition with respect to the carboxamide CdO. This “Hsyn
transfer rule” is generally accepted to be at the origin of the
facial discrimination and subsequent enantioselectivity in the
nonenzymatic reduction of alkyl benzoylformates with
NADH/Mg2+ systems.12
Unfortunately, most of the structures 3-(I) to 3-(VI) used
to describe enzyme-free ternary entities from chelated chiral
NADH models are approximative drafts rather than actual
molecular modeling structures. As a matter of fact, no
accurate ab initio computational study has been reported
yet detailing the structures and relative stabilities of the
Mg2+-centered ternary complexes and the transition states
that account for the experimental enantioselectivities attai-
ned in NADH model-based nonenzymatic reductions.13
Considerable efforts have also been devoted to provide
spectroscopic evidence for the structural elucidation of the
intermediates involved in the enzyme-free enantioselective
reduction of R-keto esters with chiral NADH models. For
example, the complexation between metal ions and several
NADH models to form the rather stable binary complexes 2
has been extensively investigated with use of NMR, UV, IR,
and fluorescence techniques2b,8,14 and the structural features
of some of them have been established in detail. In contrast,
the very few examples in the literature so far dealing with the
direct detection of ternary entities 315 concern only achiral
NADH mimetics.
FIGURE 2. The “β-Lactam Scaffold-Assisted Design” (β-LSAD)
approach to NADH models: formal insertion in the native peptide
of a carbon atom (CR-H þ H-N f CR-CH2-N) provides
pseudopeptides 5 rigidified around the ψ (≈ 120°) torsion angle.
Definition of dihedral angles: R, (C2-C3-C4-N5); φ, (C4-N5-
C6-C7); and ψ, (N5-C6-C7-N8).
We present herein a detailed computational model to
explain the enantioselective reduction of methyl benzoylfor-
mate with peptide-containing NADH mimetics under none-
nzymatic conditions. The proposed model is based on the
combination of two structural features: first, the chelation of
NADH peptides with Mg2þ cation to form a seven-mem-
bered ring (Figure 2), and second, the application of the
Freidinger’s lactam peptidomimetic16 approach to promote
the covalent rigidification of the amide linkage in NADH
peptide ligands. In particular, using specially designed chiral
β-lactam NADH models that favor stereodifferentiation
with respect to the corresponding open peptides, we have
conducted a combined experimental and DFT computa-
tional study that has allowed us to establish for the first time
the quantification of the relative stabilities and structural
characteristics of the chelated ternary complexes peptide-
NADH/Mg2þ/PhCOCO2Me and the transition states there-
of. Furthermore, incorporation of diastereotopic silyl tags in
these β-lactam NADH models has permitted the detection of
a NADH/Mg2þ/PhCOCO2Me ternary complex with NMR
spectroscopy techniques.
(12) Vasse, J. L.; Levacher, V.; Bourguignon, J.; Dupas, G. Tetrahedron:
Asymmetry 2002, 13, 227–232.
(13) For a semiempirical (MNDO-PM3) study of the Mg2þ-mediated
reduction of methyl benzoylformate with achiral dihydronicotinamide, see:
(a) Toyooka, Y.; Matsuzawa, T.; Eguchi, T.; Kakinuma, K. Tetrahedron
1995, 51, 6459–6474. For theoretical studies on conformational features of
NADH analogues, see: (b) Donkersloot, M. C. A.; Buck, H. M. J. Am.
Chem. Soc. 1981, 103, 6554–6558. (c) Brewster, M. E.; Pop, E.; Huang, M.-J.;
Bodor, N. Heterocycles 1994, 37, 1373–1415. (d) Okamura, M.; Mikata, Y.;
Yamazaki, N.; Tsutsumi, A.; Ohno, A. Bull. Chem. Soc. Jpn. 1993, 66, 1197–
1203. (e) Obika, S.; Nishiyama, T.; Tatematsu, S.; Miyashita, K.; Imanishi,
T. Tetrahedron 1997, 53, 3073–3082. (f) De Luca, G.; Marino, T.; Mineva, T.;
Russo, N.; Toscano, M. J. Mol. Struct. (THEOCHEM) 2000, 501-502, 215–
220. (g) Zhong, H.; Bowen, J. P. J. Mod. Graph. Model. 2005, 24, 1–9.
(14) (a) Hughes, M.; Prince, R. H. J. Inorg. Nucl. Chem. 1978, 40, 703–
712. (b) De Kok, P. M. T.; Donkersloot, M. C. A.; van Lier, P. M.;
Meulendijks, G. H. W. M.; Bastiaansen, L. A. M.; van Hooff, H. J. G.;
Kanters, J. A.; Buck, H. M. Tetrahedron 1986, 42, 941–959. (c) Zehani, S.;
Lin, J.; Gelbard, G. Tetrahedron 1989, 45, 733–740. (d) Tamagaki, S.;
Simojo, Y.; Mimura, T.; Tagaki, W. Bull. Chem. Soc. Jpn. 1989, 62, 1593–
1600. (e) Wu, Y. D.; Houk, K. N. J. Org. Chem. 1993, 58, 2043–2045. (f)
Results and Discussion
NADH Model Design. Among the diverse families of
NADH models, nicotinamide peptides 417 (Figure 2) are
known to represent the closest imitation of the enzyme/
coenzyme complex and were expected to provide valuable
information concerning the NADH/substrate/coenzyme in-
teraction. From a structural viewpoint, they are also expe-
cted to arrange around the Mg2þ cation in a doubly coordi-
nated fashion within a seven-membered chelation pseudo-
plane. However, NADH peptides 4 could still enjoy
considerable conformational freedom because of the rotations
ꢀ
ꢀ
ꢀ
Bedat, J.; Ple, N.; Dupas, G.; Bourguignon, J.; Queguiner, G. Tetrahedron:
Asymmetry 1995, 6, 923–932. (g) Leroy, C.; Levacher, V.; Dupas, G.;
ꢀ
Queguiner, G.; Bourguignon, J. Tetrahedron: Asymmetry 1997, 8, 3309–
3318. (h) Vitry, C.; Bedat, J.; Prigent, Y.; Levacher, V.; Dupas, G.; Salliot, I.;
ꢀ
ꢀ
Queguiner, G.; Bourguignon, J. Tetrahedron 2001, 57, 9101–9108.
(15) (a) Fukuzumi, S.; Nishizawa, N.; Tanaka, T. Chem. Lett. 1983, 1755–
1758. (b) Ohno, A.; Yamamoto, H.; Oka, S. Bull. Chem. Soc. Jpn. 1981,
3489–3491.
(16) (a) Freidinger, R. M.; Veber, D. F.; Perlow, D. S.; Brookas, J. R.;
Saperstein, R. Science 1980, 210, 656–658. (b) Freidinger, R. M. J. Med.
Chem. 2003, 46, 5553–5566.
(17) For extended NADH peptide models, see: (a) Endo, T.; Hayashi, Y.;
Okawara, M. Chem. Lett. 1977, 391–394. (b) Endo, T.; Kawasaki, H.;
Okawara, M. Tetrahedron. Lett. 1979, 23–26. (c) Saito, R.; Naruse, S.;
Takano, K.; Fukuda, K.; Katoh, A.; Inouye, I. Org. Lett. 2006, 8, 2067–2070.
J. Org. Chem. Vol. 74, No. 17, 2009 6693

Aizpurua et al.
JOCArticle
TABLE 1. Preparation of NADH β-Lactam Peptidomimetics 5a-f
and 8
TABLE 2. Asymmetric Reduction of Methyl Benzoylformate with
NADH Models 5a-e and 8
NADH
model
reaction
time [h]a
yield
[%]b
ee
major
entry
[%]c enantiomer
1
2
3
4
5
6
7
5a
5b
5c
5d
5e
5f
8
16
16
18
18
16
18
16
95
66
>99
>99
49
73
78
82
90
90
90
60
S
S
S
S
S
S
S
94
60
aReactions were quenched with water when complete consumption of
the model NADH was observed by 1H NMR. bCalculated on the basis of
c
model NADþ formation. Determined by HPLC, using the Chiralcel
OD stationary phase.
Synthesis of β-Lactam NADH Peptidomimetics and Bio-
mimetic Reduction of Methyl Benzoylformate. On the basis of
the previous work from our laboratory, the synthesis of
NADH β-lactam peptidomimetics 5 (Table 1) was addressed
starting from the readily available enantiopure N-[bis-
(trimethylsilyl)methyl]-β-lactams 6a-f.19 Hydrogenolytic
cleavage of the oxazolidinone moiety afforded the corre-
sponding R-amino-β-lactams20 which were next coupled
with nicotinyl chloride hydrochloride in the presence of
diisopropylethylamine (DIPEA) to give the corresponding
nicotinamide derivatives 7a-f. Quaternization of the pyr-
idine nitrogen atom with methyl iodide and subsequent
regioselective reduction with sodium dithionite provided
the desired NADH models 5a-f. Interestingly, this metho-
dology also permitted the incorporation of a second
β-substituent in the azetidinone ring to afford the
R,β-disubstituted models 5e-f. Finally NADH peptidomi-
metic 8, the desilylated analogue of model 5b, was prepared
similarly from 7b after desilylation with tetrabutylammo-
nium fluoride in THF at reflux in 64% isolated yield (for full
experimental details, see the Supporting Information).
Enantioselective reduction of methyl benzoylformate 9
with the NADH models was investigated in the presence of
magnesium perchlorate in acetonitrile at room temperature,
and the results obtained are summarized in Table 2. A
uniform stereoinduction tendency was observed in all
instances affording the S-(þ) enantiomer of methyl mande-
late 10 as the major reaction product. The enantiomeric
excess was increased by bulkier R-substituents or cis-R,β-
disubstitution on the azetidinone ring. Finally, in an experi-
ment conducted to rule out any particular stereodirecting
bias arising from the bis(trimethylsilyl)methyl group, the
desilylated model 8 was also tested for reduction, providing
aOverall yield from 6a-f. bOverall yield from 7a-f.
around R, φ, and ψ torsion angles within the 1,4-dihydroni-
cotinamide and the R-amino acid peptide chain. In a seminal
work, Inouye demonstrated that the enantiomeric excess of
the mandelate reduction products can be enhanced by incor-
porating the torsionally blocked proline amino acid into the
NADH peptide models.6 We reasoned that the incorporation
of a planar constraint by means of a lactam mimetic might
reduce the flexibility around the chelation pseudoplane, thus
enabling a more accurate interpretation of computational
calculations (by minimizing the energy contribution due to
conformational mobility, see below) and facilitating the inter-
pretation of interproton NOE interactions in NMR experi-
ments. To this end, we selected the minimal-sized and
puckering-free β-lactams 5 designed according to the
“β-LSAD” principle (see Figure 2).18 This constraint fixed
the ψ angle to ∼120°, strongly limiting the rotation around φ.
In addition, NMR-tagging could be easily accomplished
by placing two diastereotopic trimethylsilyl groups [R =
CH(SiMe3)2] internally oriented toward each face of the
chelation pseudoplane.
(19) (a) Palomo, C.; Aizpurua, J. M.; Legido, M.; Galarza, R.; Deya,
P. M.; Dunogues, J.; Picard, J. P.; Ricci, A.; Seconi, G. Angew. Chem., Int.
Ed. Engl. 1996, 35, 1239–1241. (b) Palomo, C.; Aizpurua, J. M.; Legido, M.;
Mielgo, A.; Galarza, R. Chem. Eur. J. 1997, 3, 1432–1441. (c) Palomo, C.;
Aizpurua, J. M.; Ganboa, I.; Benito, A.; Cuerdo, L.; Fratila, R. M.; Jimenez,
A.; Loinaz, I.; Miranda, J. I.; Pytlewska, K. R.; Micle, A.; Linden, A. Org.
Lett. 2004, 6, 4443–4446.
(18) (a) Palomo, C.; Aizpurua, J. M.; Benito, A.; Miranda, J. I.; Fratila,
R. M.; Matute, C.; Domercq, M.; Gago, F.; Martın-Santamaria, S.; Linden,
´
A. J. Am. Chem. Soc. 2003, 125, 16243–16260. (b) Palomo, C.; Aizpurua,
J. M.; Balentova, E.; Jimenez, A.; Oyarbide, J.; Fratila, R. M.; Miranda, J. I.
Org. Lett. 2007, 9, 101–104.
(20) In the case of R-(2-methylnaphthyl)-β-lactams 6d and 6f a partial
hydrogenation of the naphthalene moiety was observed and the correspond-
ing 5,6,7,8-tetrahydronaphthyl derivatives were obtained.
6694 J. Org. Chem. Vol. 74, No. 17, 2009

Aizpurua et al.
JOCArticle
FIGURE 3. Binary and ternary complexes of octahedral Mg2þ with methyl benzoylformate 9 and NADH β-lactam model 11 in acetonitrile.
Relative free energy values are in kcal/mol. B3LYP/6-31G*-optimized geometries of binary complex 15 and ternary complex 21 are shown
omitting all the hydrogen atoms, except the amide NH and the NADH methylene. Dihedral angles as defined in Figure 2.
the (S)-methyl mandelate in slightly lower enantiomeric
excess than its silylated β-lactam counterpart 5b (compare
entries 2 and 7).
Single-point calculations with the self-consistent reaction
field (SCRF) based on the IEF-PCM24 solvation model
(MeCN, ε = 36.64) were carried out at the B3LYP/6-
311þþG** level on the previously optimized transition
states and stationary points.
The relative free energies at 298 K and some representative
structural parameters of the divalent cationic binary com-
plexes 12-15 and the ternary complexes 16-25 are summar-
ized in Figure 3. They were formally built up from the
magnesium complex [Mg(NCMe)6]2þ by displacement of
two or four acetonitrile ligands with methyl benzoylformate
9 and/or the β-lactam NADH model 11. In all cases, the
bonding patterns were created assuming an octahedral coor-
dination geometry for the Mg2þ cation.25
Computational Analysis. To study the reaction mechanism
and to ascertain the origin of the stereoselectivity observed
for the reduction of methyl benzoylformate 9 with NADH
β-lactam peptidomimetics 5, we selected first the model
structure 11 (Figure 3). The flexible benzylic moieties in
compounds 5a-f and the bulky CH(SiMe3)2 were replaced
by the less demanding and more rigid two methyl groups to
shorten the computational time and limit the conformational
energy. For the initial model, all structures were optimized
by using the density functional B3LYP21 and the 6-31G*
basis sets as implemented in Gaussian 03.22 All energy
minima and transition structures were characterized by
frequency analysis. The energies reported in this work in-
clude zero-point vibrational energy corrections (ZPVE) and
are not scaled. The intrinsic reaction coordinates (IRC)23
were followed to verify the energy profiles connecting each
transition structure to the correct associated local minima.
The formation of the binary complexes [Mg(NCMe)4/
PhCOCO2Me]2þ showed Gibbs free energies of -20.6
(12), -14.5 (13), and -5.6 kcal/mol (14), respectively. Five-
membered cyclic structure 12, chelating together the magne-
ꢁ
(24) (a) Cances, E.; Mennucci, B.; Tomasi, J. J. Chem. Phys. 1997, 107,
3032–3047. (b) Cossi, M.; Barone, V.; Mennuci, B.; Tomasi J. Chem. Phys.
ꢁ
Lett. 1998, 286, 253–260. (c) Tomasi, J.; Mennucci, B.; Cances, E. J. Mol.
Struct. (THEOCHEM) 1999, 464, 211–226.
(21) (a) Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785–789.
(b) Becke, A. D. J. Chem. Phys. 1993, 98, 5648–5652. (c) Kohn, W.; Becke, A.
D.; Parr, R. G. J. Phys. Chem. 1996, 100, 12974–12980.
(22) Frisch, M. J.; et al. Gaussian 03, Revision D.03; Gaussian, Inc.,
Wallingford, CT, 2004.
(23) Gonzalez, C.; Schlegel, H. B. J. Phys. Chem. 1990, 94, 5523–5527.
(25) (a) Kluge, S.; Weston, J. Biochemistry 2005, 44, 4877–4885. (b) At
low concentration of Mg(ClO4)2 in acetonitrile most of the six coordination
sites of Mg are filled with CD3CN, whereas at high concentration one or
more sites are more likely filled with the ClO4- counterions, see: Cha, J. N.;
Cheong, B. S.; Cho, H. G. J. Phys. Chem. A 2001, 105, 1789–1796.
J. Org. Chem. Vol. 74, No. 17, 2009 6695

Aizpurua et al.
JOCArticle
TABLE 3. Relative Energies (in kcal/mol) and Main Torsion Angles (deg) of the Transition Structures 26-33 Computed at the B3LYP/6-31G* Level for
NADH Peptide Mimetics 11 and 11A
entry NADH
TS
topologya config. 9 complex
ΔGqb,c
0ΔΔG‡
ΔHqb,c
Rd
φd
ψd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
11
11
11
11
11
11
11
11
11A
11A
11A
11A
11A
11A
11A
11A
26
27
28
29
30
31
32
33
(HR-re)
(HS-re)
(HR-si)
(HS-si)
(HR-si)
(HS-si)
(HR-re)
(HS-re)
(S)
(S)
(R)
(R)
(R)
(R)
(S)
(S)
(S)
(S)
(R)
(R)
(R)
(R)
(S)
(S)
16
16
17
17
18
18
19
19
16A
16A
17A
17A
18A
18A
19A
19A
22.4 (23.0) [16.0]e 0.0 [0.0]e 19.3 (20.2)
23.1 (syn)
-87.3 (s-clinal)
10.7 (syn)
62.0
60.2
58.7
62.9
4.3
-10.7
-11.9
1.4
68.4
70.0
70.6
69.2
-25.1
-79.6 -128.0
-132.7 -150.0
-79.5 -119.4
116.5
115.7
115.7
119.3
122.3
126.7
125.3
121.9
139.5
124.4
130.5
154.0
143.5
25.9 (26.5)
28.2 (28.2)
31.3 (31.3)
26.6 (27.8)
3.5
5.8
8.9
4.2
23.8 (24.8)
22.4 (24.0)
27.1 (28.7) -143.3 (anti)
25.0 (25.0)
132.0 (anti)
-20.9 (syn)
149.5 (anti)
19.6 (syn)
22.4 (syn)
-94.0 (s-clinal)
-5.9 (syn)
25.5 (26.7) [19.7]e 3.1 [3.7]e 23.4 (23.4)
28.5 (29.5)
26.8 (27.8)
6.1
5.4
26.4 (26.8)
25.6 (26.1)
26A (HR-re)
27A (HS-re)
28A (HR-si)
29A (HS-si)
30A (HR-si)
31A (HS-si)
32A (HR-re)
33A (HS-re)
20.3 (21.7) [12.5]e 0.1 [0.0]e 15.8 (18.0)
24.5 (25.9)
24.8 (27.2)
27.1 (29.5)
27.5 (27.8)
4.3
4.6
6.9
7.4
21.6 (23.7)
24.7 (24.7)
26.3 (26.3) -149.8 (anti)
25.0 (25.3)
131.2 (anti)
-21.6 (syn)
20.2 (20.4) [13.0]e 0.0 [0.5]e 15.7 (16.0)
25.9 (25.9)
26.3 (26.3)
5.7
6.1
21.8 (22.3) -179.5 (anti)
23.5 (22.9) -1.9 (syn)
aFace-selective interaction between the 1,4-dihydronicotinamide (hydride donor) and methyl benzoylformate 9 (hydride acceptor). bActivation
energies calculated from the ternary complexes 16-19. cValues in parentheses correspond to the activation energies calculated from the most stable
“productive” ternary structure 17 (entries 1-8) and 19A (entries 9-16). dDihedral angles as defined in Figure 2. eActivation energies from single-point
calculation at the B3LYP/6-311þþG** level including the IEF-PCM solvation model (MeCN, ε=36.64).
sium cation with the methyl benzoylformate ketone and the
ester carbonyl, was found to be the one conformationally
preferred among the bidentate complexes 12-14. This s-cis
conformation was 6 kcal/mol more stable than the s-trans
isomer 13 and 15 kcal/mol more stable than the ester-bound
isomer 14. Most of the computed Mg---OdC distances
(2.01-2.23 A) were within the expected range for this kind
of coordinated bond, except the Mg---OMe distance in 14
(3.67 A), which was too long to be considered a dative
covalent bond.
perature. The structural parameters of the ternary complexes
16-25 were also remarkably homogeneous. For example,
the computed Mg---OdC distances ranged in the interval
2.05 ( 0.02 A for the β-lactam carbonyl; 2.01 ( 0.02 A for the
1,4-dihydronicotinamide carbonyl; 2.14 ( 0.01 A for the
ketone; and 2.16 ( 0.01 A for the ester carbonyl. Dihedral
angles φ varied from 61.8° in complex 17 to 69.8° in complex
25 and, similarly, dihedral angles R were all close to (20°
(syn), lying from -18.8° in complex 21 to 12.7° in complex 16
(see the Supporting Information for details). As a conse-
quence, according to these calculations, the seven-membered
chelation pseudoplane of the binary complex 15 experienced
a practically negligible structural change after the methyl
benzoylformate ligand was incorporated to form the ternary
complexes 16-25.
Next, an extensive computational survey of the intramo-
lecular hydride transfer reaction step was undertaken with
the aim to characterize the possible transition states arising
from the ten ternary complexes 16-25. Although the precise
mechanism of this step is still controversial and one-electron
transfer may occur in some NADH models,26 a concerted
hydride transfer mechanism was considered by analogy with
the Hantzsch 1,4-dihydropyridine mechanism described by
Simon and Goodman.27 Not surprisingly, using the β-lactam
NADH model 11 only the four complexes 16-19 led to
canonical transition structures (TS-26-33, Table 3, entries
1-8), whereas complexes 20-25 presented distances
between the dihydronicotinamide diastereotopic hydrogens
(HR, HS) and the methyl benzoylformate ketone carbon
atom which were too long to attain potentially “productive”
transition states. It was also found that two different reaction
topologies were possible for the hydrogen transfer in each
ternary complex 16-19, yielding a total of 8 transition states
(for detailed structures, see the Supporting Information).
Nonetheless, as the configuration of the final product was
dictated solely by the re- or si-facial attack to the methyl
benzoylformate ketone group, only one methyl mandelate
The formation of the binary complex 15 was energetically
very favored, with a Gibbs free energy of -46.9 kcal/mol for
the exchange equilibrium (11 þ [Mg(NCMe)6]2þ f [15]2þ þ
2 NCMe). This complex included two Mg---OdC bonds
which were almost identical in length (2.02 and 2.06 A). It
also comprised a slightly distorted chelation pseudoplane
bearing a 1,4-dihydronicotinamide group oriented in a qua-
siequatorial disposition. The planarity of the seven-mem-
bered chelation pseudoplane in complex 15 could be gauged
from the dihedral angle φ (68.5) encompassing the four
central C4-N5-C6-C7 atoms from the fiducial dihydropyr-
idine ring. The syn disposition of the 1,4-dihydronicotina-
mide moiety could also be confirmed from the dihedral angle
R (-19.6) within the C2-C3-C4-N5 bonds (see Figure 3).
As mentioned above, the methyl benzoylformate molecule
formed a strongly preferred s-cis chelate with Mg2þ in the
binary complex 12 and, consequently, this ligand conforma-
tion was retained to build up the ternary structures 16-25
comprising all ten possible molecular arrangements of 9
around 15. After B3LYP/6-31G* optimization, small energy
differences (only up to 2.2 kcal/mol) were computed between
the most stable and the less stable optimized complexes.
Accordingly, structures 20 and 21 with the methyl benzoyl-
formate ligand 9 placed in the rear bottom quadrant position
were found to be slightly more stable than, say, the isomers
18 and 19 with the methyl benzoylformate ligand placed in
the front top quadrant. Moreover, the ΔG° of the exchange
equilibrium from the binary complex 15 to the most stable
ternary structure 21 ([15]2þ þ 9 f [21]2þ þ 2NCMe) was only
-5.3 kcal/mol at 25 °C, suggesting the potential for equilib-
rium between the different ternary complexes at such tem-
(26) Zhu, X. Q.; Li, H. R.; Li, Q.; Ai, T.; Lu, J. Y.; Yang, Y.; Cheng, J. P.
Chem. Eur. J. 2003, 9, 871–880.
(27) Simon, L.; Goodman, J. M. J. Am. Chem. Soc. 2008, 130, 8741–8747.
6696 J. Org. Chem. Vol. 74, No. 17, 2009

Aizpurua et al.
JOCArticle
FIGURE 4. Energy diagram (ΔGq) for the 8 computed pathways (TS-26/33) of the hydrogen-transfer step in the intramolecular reduction of
“productive” ternary complexes 16-19 to the (R) and (S) mandelate complexes. Gibbs energies are in kcal/mol and correspond to the energy
differences from the corresponding ternary complexes 16-19. The relative dispositions of the 1,4-dihydronicotinamide CdO and C2dC3 bonds
are noted as syn and anti. Some energy levels might not be depicted linearly.
enantiomer could be formed from each ternary complex
through two concurrent pathways (Figure 4).
conditions (>0 °C) required for the reaction to occur at a
measurable rate. They also were in excellent agreement with
the values (ΔGq ≈ 21 kcal/mol) reported for related transfor-
mations in enzyme-promoted systems.11c However, a slight
overestimation inherent to the computational method used
was noticed for the activation energy difference between the
transition states TS-26 and TS-31 (ΔΔGq ≈ 3.1 kcal/mol) with
respect to the ee values (73-90%) measured experimentally
for the best silylated β-lactam NADH models 5a-f. Finally,
introduction of the solvent factor with the IEF-PCM solvation
model (MeCN, ε = 36.64) and carrying out single-point
calculations at the B3LYP/6-311þþG** level conf-
irmed the activation energy difference (ΔΔGq ≈ 3.7 kcal/mol)
between the transition states TS-26 and TS-31 and validated
the values calculated in vacuo at the B3LYP/6-31G* level.
As outlined earlier, ongoing models to explain the facial
discrimination of nonenzymatic NADH/Mg2þ/PhCO-
CO2Me systems rely on structural observations (e.g., the
“Hsyn rule”) which are ovserved in free NADH models or, in
some cases, on model NADH/Mg2þ secondary systems. The
translation of such information to the corresponding ternary
complexes and transition states is, at least, not obvious.
Indeed, we have observed that activation energies calculated
either from each individual complex 16-19 or from the most
stable “productive” ternary complex 17 (ΔGq values are
Surprisingly, a reduction in the coordination number at
Mg2þ from 6 to 5 was observed for the product adducts 34
and 35. Transformation of the ketone group into an alkoxide
anion triggered a change from the octahedral to the trigonal
bipyramidal geometry by loosening the coordination with
the N-methylnicotinium amide carbonyl group. A similar
observation has been documented recently on Mg2þ ion
hexahydrate upon deprotonation of one of the ligands to
form the hydroxide anion.24 The energies of 34 and 35 were
2.9 and 6.6 kcal/mol less stable than their respective starting
ternary complexes 16 and 18. We attributed this discrepancy
to a possible inadequate computation “in vacuo” of the
charge redistribution from the Mg2þ center to the pyridi-
nium salt and magnesium (mono)alkoxide. An inspection of
the relative activation enthalpies and Gibbs free energies
revealed that transition state TS-26 (entry 1) represents the
lower energy pathway to the (S)-methyl mandelate 10, the
main stereoisomer obtained experimentally from the β-lac-
tam NADH peptidomimetics 8 and 5a-f. Transition state
TS-31 (entry 6 in Table 3) was the next most stable and
provided the most favorable pathway to (R)-methyl mande-
late. Both activation energies were in the range ΔGq ≈ 22.4-
25.5 kcal/mol, consistent with the experimental reaction
J. Org. Chem. Vol. 74, No. 17, 2009 6697

Aizpurua et al.
JOCArticle
FIGURE 5. Energy diagram (ΔGq) for the 8 lower activation-energy pathways, using NADH-Ala-NHMe mimetic 11A. Gibbs energies are in
kcal/mol and correspond to the energy differences from the corresponding ternary complexes 16A-19A. Some energy levels might not be
depicted linearly.
shown in parentheses in Table 3) were essentially identical.
This suggested that the relative stabilities of the different
productive ternary structures in the ground state play only a
limited role in determining the R/S configuration of the final
products, whereas the relative stabilities of the transition
states have a striking importance.
To understand why one particular syn transition state
structure is more stable than another one, we turned our
attention to the torsion angles φ and ψ, embodying the strain
of the seven-membered chelation ring. In particular, φ
remained practically unchanged for syn-TS-26 (φ = 62.0°)
with respect to the starting ternary complexes 16 (φ=62.1°)
and 18 (φ=67.3°), but adopted a quite different value for syn-
TS-31 (φ = -10.7°). A comparison of the breaking and
forming C-H bond lengths within the hydride transfer
trajectory also confirmed such differences, yielding a slightly
more reagent-like structure for the transition state TS-31
(Figure 4).
Seeking a confirmation of the foregoing observations, we
conducted further computational calculations under identi-
cal conditions as before, but using the nonlactamic NADH
peptide model 11A28 instead of 11 (Figure 5).
The results listed in Table 3 (entries 9-16) revealed two
nearly isoenergetic syn transition states, TS-26A and
TS-31A, as the pathways of choice to the corresponding
(S)- and (R)-methyl mandelates. Their φ dihedral angles were
now very similar in absolute value (62.0° and -79.6°,
respectively) to those of their corresponding ternary
complexes (16A, φ = 73.4°; 18A, φ = -80.4°). Again, the
final adducts 34A and 35A showed trigonal bipyramidal
To gain insight into the structural origins of the activation
energy differences observed for the transition states TS-26-
33, we carefully examined the correlation of ΔGq with the
torsion angles R, φ, and ψ. As shown in Table 3, R torsion
angles varied over a wide range depending on the conforma-
tion adopted by the 1,4-dihydronicotinamide group around
the C3-C4 bond to favor the attack of the HR or the HS
diastereotopic protons to the ketone prochiral face of methyl
benzoylformate 9. In all instances the syn conformations
(R ≈ (20°; entries 1, 3, 6, and 8) were preferred over the
synclinal (R ≈ (45/90) or anticlinal (R ≈ (130/160) con-
formations, which were destabilized by up to 3.5 kcal/mol
because of the disruption of OdC-CdC π-π conjugation.
These results agreed with previous calculations conducted on
model NADH molecules in the ground state,13g but raised
the question about the suitability of the “Hsyn rule” alone to
explain the actual origin of the enantioselectivity observed in
enzymomimetic reductions with NADH models. In fact,
both transition states syn-TS-26 (R=23.1°) and syn-TS-31
(R = -20.9°) obey the rule, but each of them leads to the
opposite methyl mandelate enantiomer.
(28) To facilitate comparison, we assigned the notation nnA to all the
ternary complexes and transition states resulting from the replacement of the
β-lactam 11 by the open NADH model 11A.
6698 J. Org. Chem. Vol. 74, No. 17, 2009

Aizpurua et al.
JOCArticle
SCHEME 2. Formation of Binary Complexes of 10 mM Methyl
Benzoylformate 9 with Magnesium Perchlorate in Deuterated
Acetonitrile
strain effect was best viewed by considering the β-lactam
transition states TS-26 and TS-31 as the result of the formal
insertion of a methylene bridge (CR-H þ H-N f CR-
CH2-N) into the transitions states TS-26A and TS-31A,
respectively. Since the CR-H and N-H protons of TS-26A
have a syn disposition, their blockage to form a β-lactam ring
(ψ ≈ 120°) caused little change of the central dihedral angle φ
and, consequently, no extra strain tension in TS-26. In
contrast, the anticlinal disposition of the CR-H and N-H
protons of TS-31A provoked a destabilizing puckering of the
chelation pseudoplane in TS-31, as evidenced by the dra-
matic change in the torsion angle φ.
NMR. A series of NMR studies, including chemical shift
changes, NOE, and diffusion experiments, were performed
next to study the complexation of magnesium perchlorate
with methyl benzoylformate 9, with NADH models 5b and
5e, and with combinations of the two. Neither 1H NMR nor
13C NMR spectra of methyl benzoylformate 9 showed
perceptible chemical shift changes after the addition of 1
equiv of Mg(ClO4)2 at room temperature in deuterated
acetonitrile (Scheme 2). Only a small downfield shift
(0.13 ppm) for the OMe group 13C signal was observed
which, in line with our calculations and previous observa-
tions,29 confirmed the weak intensity of the interaction of 9
with the Mg2þ cation. DOSY experiments conducted at
20 °C also confirmed these observations and afforded very
similar diffusion coefficients (2.50 ꢀ 10-9 m2/s for 9 and
2.35 ꢀ 10-9 m2/s for the binary complex 9/Mg2þ).
FIGURE 6. Comparison of B3LYP/6-31G*-optimized lower
energy transition states TS-26A and TS-31A (top) with their
β-lactam homologues TS-26 and TS-31 (bottom). Formal insertion
of a methylene bridge (CR-H þ H-N f CR-CH2-N) causes the
puckering of the chelation pseudoplane in TS-31, but not in TS-26.
All hydrogen atoms except the amide NH, CRH, and NADH C4H
were omitted for clarity.
geometry and the lengths of C-H bonds being cleaved or
formed in TS-26A and TS-31A were also identical in both
transition states. Likewise, the torsion angle ψ remained
practically constant (≈130°) in either ternary complexes or
transition states, and presumably had no influence on the
stereochemical outcome of the reaction.
According to the aforementioned computational results,
our modeling indicats the formation of quasiracemic mixt-
ures of enantiomers from flexible NADH peptide models. In
line with this prediction, Saito and Inoue7 described recently
that 1,4-dihydronicotinamide pseudopeptides containing an
(L)-R-amino ester residue reduce methyl benzoylformate 9 in
enantiomeric excesses below 23% at room temperature.
They noted that the R/S configuration of the major methyl
mandelate depends on the side chain of the (L)-R-amino acid
employed and also on the reaction temperature. The con-
formational changes that, obviously, were suspected to be
responsible for such switching17a of the chirality of the
products can now be rationally understood on the basis of
our chelation model.
Finally, comparison of the lower energy transition states
arising from the flexible NADH peptide model 11A and the
rigidified NADH β-lactam model 11 (Figure 6) reinforced
the idea that a syn disposition of the 1,4-dihydronicotina-
mide C2dC3-C4-N5 bonds (R ≈ (20°) was necessary but
not sufficient to fully explain the stereochemical behavior of
NADH/Mg2þ/PhCOCO2Me systems. An additional strain
energy caused by the deformation of the seven-membered
chelation ring to attain the different transition states from
the ternary complexes was found to be the actual reason of
the stereodiscrimination in NADH peptide mimetics. Such a
We next examined the binary complexes 5b/Mg2þ and 5e/
Mg2þ. The conformation of the bis(trimethylsilyl)methyl
moiety in NADH models 5b and 5e was dramatically
affected by the absence or presence of a β-substituent in
the azetidinone ring (Figure 7). As previously established in
our laboratory using X-ray and NMR techniques,30
the more stable conformer of β-unsubstituted β-lactams
(e.g., 5b) has a marked preference for a syn disposition of
the CH(SiMe3)2 methine C-H bond and the β-lactam
carbonyl, whereas the β-substituted analogues (e.g., 5e)
adopt an anti relative disposition, resulting in an inversion
of the preference of the pro-R and pro-S SiMe3 groups placed
above and below the azetidin-2-one ring plane, respectively.
1
The H NMR spectrum of the binary complex 5b/Mg2þ
recorded at room temperature (Figure 7A) was more com-
plicated than expected and showed two sets of signals and
four trimethylsilyl groups in the aliphatic region comprised
between 1 and -1 ppm. This NMR signal splitting was
attributed to the existence of an equilibrium between the
chelated conformers 36 and 37 (Figure 7), generated by the
(29) Hoshide, F.; Baba, N.; Oda, J.; Inouye, Y. Agric. Biol. Chem. 1983,
47, 2141–2143.
(30) Palomo, C.; Aizpurua, J. M.; Benito, A.; Cuerdo, L.; Fratila, R. M.;
Miranda, J. I.; Linden, A. J. Org. Chem. 2006, 71, 6368–6373.
J. Org. Chem. Vol. 74, No. 17, 2009 6699

Aizpurua et al.
JOCArticle
FIGURE 7. Binary complexes of β-lactam NADH peptide models 5b and 5e with magnesium perchlorate in deuterated acetonitrile. Carbonyl
13C NMR chemical shift changes (in ppm) for 36-38 after the addition of Mg(ClO4)2 to 5b and 5e. Boxes A-F show magnifications of the
trimethylsilyl and carbonyl regions of the 1H NMR spectra (top) and 13C NMR spectra (bottom).
hindered rotation of the benzylic group between the bulky
pro-S SiMe3 group and the axial acetonitrile ligand. Actu-
ally, a total SiMe3 signal coalescence was observed for 36 and
37 above 50 °C and the two sets of signals were restored after
recooling the mixture to 20 °C (see the Supporting
Information). Further addition of a second equivalent of
magnesium perchlorate to the binary complex 5b/Mg2þ
resulted in a reunification of the SiMe3 signals into two close
singlets, which was interpreted as a rupture of complexes 36
and 37 to form a 5b/2Mg2þ nonchelated species (spectra not
shown).
dihydronicotinamide and the β-lactam shifted 1.95 and
3.39 ppm downfield, respectively, upon complexation to
36,31 whereas formation of 37 provoked an upfield shift of
-2.91 ppm of the β-lactam carbonyl carbon.
Owing to the spatial disposition of the CH(SiMe3)2
moiety in the β-substituted NADH model 5e, it was unlikely
that the complex 5e/Mg2þ could adopt a conformation
similar to that of 37 with the benzyl group arranged between
the pro-R SiMe3 group and the axial acetonitrile ligand. In
good agreement with this prediction, we observed no dia-
1
magnetic upfield shift for SiMe3 signals in the H NMR
spectrum of the binary complex 5e/Mg2þ (Figure 7D), and
only two signals for SiMe3 and CdO groups in the 13C
NMR spectra (Figure 7E,F). Taking into account these
observations, we assigned the structure 38 to the complex
1
Chemical shift changes in the H NMR and 13C NMR
spectra caused by the immobilization of the phenyl group in
37 provided additional information to confirm its structure.
First, a selective and large diamagnetic aromatic shielding
(-0.55 ppm) of the pro-S SiMe3 group distinguished it from
5e/Mg2þ
.
1
Next, we examined the interaction of methyl benzoyl-
formate 9 with β-lactam NADH/Mg2þ binary complexes.
Because of the large separation of the chemical shift
observed for SiMe3 groups in complex 37, the NADH
model 5b was considered to provide the best NMR
tagging and was selected for our observations. Experi-
ments were performed with an equimolar mixture of 5b/
Mg(ClO4)2/9 in CD3CN at -40 °C in order to slow down
the reduction reaction of 9 to (S)-methyl mandelate and
the diastereotopic pro-R SiMe3 in the H NMR spectrum
(Figure 7A). Second, the mixture of complexed conformers
36 þ 37 displayed four SiMe3 groups in the 13C NMR
spectrum (Figure 7B), resonating between -0.2 and
-0.3 ppm upfield from the signals of the free NADH model
5b. Finally (Figure 7C), the carbonyl carbons of the
(31) For similar observations on other NADH models, see: Marchelli,
R.; Dradi, E.; Dossena, A.; Casnati, G. Tetrahedron 1982, 38, 2061–2067.
6700 J. Org. Chem. Vol. 74, No. 17, 2009

Aizpurua et al.
JOCArticle
magnesium perchlorate and 9. A careful examination of
both 13C and 1H NMR spectra showed that the most
significant changes after the addition of 1 equiv of methyl
benzoylformate to the binary complex 5b/Mg2þ were
suffered by the two diastereotopic SiMe3 groups belong-
ing to the set of “sharp peaks”. Finally, measurement of
the diffusion coefficient of NADH model 5b from DOSY
experiments showed a significant change from free 5b
(1.42 ꢀ 10-9 m2/s) to the binary mixture 5b/Mg2þ (0.99 ꢀ
10-9 m2/s), but only a very small change was observed
upon addition of 9 (1.10 ꢀ 10-9 m2/s). The value of the
diffusion coefficient of the methyl benzoylformate in the
ternary complex was of 2.62 ꢀ 10-9 m2/s, again very
similar to the values observed for the free 9 and binary
9/Mg2þ
.
ROESY experiments proved to be particularly helpful to
unravel the structure of one of the components of the 5b/
Mg2þ/9 ternary system, which was finally assigned to the
complex 39 (Figure 8). Significant “intramolecular” NOE
interactions within the NADH model 5b (indicated with
dashed arrows and elipses) and “intermolecular” NOE bet-
ween 5b and the methyl benzoylformate ligand 9 (drawn with
plain arrows and squares) are detailed in Figure 8. Two key
intramolecular cross-peaks were observed for the benzyl
aromatic protons (Ar): one with the CH(SiMe3)2 methyne
and another with the pro-S SiMe3 group, in good agreement
with the previous chemical shift shielding observed for such a
group in the binary complex 37 (see Figure 7A). In contrast,
the pro-R SiMe3 group showed no NOE with the benzyl
aromatic protons (Ar) but, instead, a cross-peak with a signal
at 2.88 ppm (dotted ellipse) that corresponded to the dihy-
dropyridine HR proton. On the other hand, three significant
intermolecular NOE signals between the OMe group of
methyl benzoylformate 9 and the NADH model 5b were
also detected: a first one with the dihydropyridine HR pro-
ton, another one with the CH(SiMe3)2 methyne proton, and,
finally, a third one with the pro-R SiMe3 group placed below
the chelation pseudoplane.
The above results were in good agreement with the initial
computational studies conducted from NADH binary com-
plex model 15 (Figure 3), which anticipated the formation
of two sets of ternary complexes, “productive” and “non-
productive”, with methyl benzoylformate 9. In the latter
category, the ternary complex 21 with ligand 9 placed in
the rear bottom-quadrant position was predicted to be
the more stable. Complex 39 showed exactly the same
ligand disposition as 21 and its enhanced stability would
explain the unusual observation of the “intermolecular”
NOE cross-peak signals between 5b and 9. Indeed, such
signals were not observed or were extremely weak for
the “broad” set of peaks arising from 5b, likely because of
the lower stability of the “productive” ternary complex
entities. Finally, the ternary complex formed by 9 with
the R,β-disubstituted β-lactam model 5e also failed to
provide “intermolecular” NOE cross-peaks, and only the
signals corresponding to the binary complex 38 (Figure 7)
were recorded. Again, this observation was consistent
with the proposed model, since the disposition of both
SiMe3 groups in 38, pointing toward the β-lactam carbonyl,
would hinder the placement of the methyl benzoylformate
ligand in the rear bottom-half quadrant of the ternary
complex.
FIGURE 8. Selected extensions of the ROESY spectrum (500 MHz,
400 ms, CD3CN, -40 °C) of the ternary complex 39 (5b/Mg2þ
/
PhCOCO2Me). Intramolecular NOE’s correspond to dashed arrows
and are inside ellipses; intermolecular NOE interactions are indicated
with plain arrows and cross-peaks inside rectangles.
to prevent the decomposition of the NADH mimic. The
coexistence of two sets of signals in the 1H spectrum, one
constituted of broad signals and another of sharper
peaks, strongly suggested that the “ternary entity” 5b/
Mg2þ/9 was actually formed by an equilibrium of various
ternary complexes exchanging the methyl benzoylfor-
mate ligand 9 at different rates, in agreement with the
conformational complexity predicted by DFT calcula-
tions. In addition, the 13C chemical shifts of the NADH
carbonyls resonated at intermediate frequencies between
the free 5b and binary complex 5b/Mg2þ, consistent with
the formation of a new entity. These data were also
supported by the changes observed for the chemical
shifts of the dihydropyridine protons in the presence of
J. Org. Chem. Vol. 74, No. 17, 2009 6701

Aizpurua et al.
JOCArticle
Conclusions
under reduced pressure and the crude extract was purified by
flash chromatography (eluent: EtOAc).
Strongly rigidified β-lactam peptidomimetic NADH mod-
els have been used to investigate the ternary entities NADH/
Mg2þ/PhCOCO2Me taking part in the biomimetic reduction
of methyl benzoylfomate. According to computational cal-
culations at the B3LYP/6-31G* level, up to 10 stable ternary
complexes can be formed by the accommodation of an s-cis-
methyl benzoylformate ligand around a seven-membered
chelation pseudoplane encompassing the Mg2þ cation, the
β-lactam carbonyl, and the dihydronicotinamide carbonyl.
Calculations have also allowed an unambiguous identifica-
tion of the “productive” ternary complexes, as well as the
corresponding transition states leading to the (S)- and (R)-
methyl mandelates with activation energy differences about
3-4 kcal/mol. A comparison of the preferred transition
states of the rigidified NADH β-lactam model 11 and the
flexible peptide counterpart 11A has shown that a combina-
tion of two conformational features is necessary to attain
efficient stereodifferentiation: first, the syn disposition of the
1,4-dihydronicotinamide C2dC3-C4-N5 bonds (R ≈ (20°)
to reach the HR-re and HS-si topologies and, second, the
diastereomeric deformation of the chelation pseudoplane by
blockage of the torsion angles φ and ψ to bring the dihy-
dropyridine group from a quasiequatorial disposition in the
ternary complexes to a quasiaxial one in the transition
structures.
In agreement with theory, some R,β-substituted β-lactam
NADH mimetics synthesized yielded the expected (S)-meth-
yl mandelate enantiomer in up to 90% ee, albeit calculated
ΔΔGq values suggested higher enantiomeric excesses.
Furthermore, it has been shown that two diastereotopic
trimethylsilyl groups incorporated close to the β-lactam ring
can act as NMR-tags. This has permitted the identification
of the structure of an exceptionally stable “nonproductive”
ternary complex and its structural determination by ROESY
experiments including for the first time cross-peak signals
between an NADH model and methyl benzoylformate.
Studies concerning the extension of the proposed model
to the realm of monodentate NADH mimetics and to
different carbonyl substrates are currently underway in our
laboratory.
General Procedure for the Preparation of N-Methyl-1,4-dihy-
dronicotinamides 5a-f. To a solution of the corresponding
nicotinamide 7 (0.5 mmol) in CH3CN (5 mL) under N2 atmo-
sphere was added methyl iodide (3 mL). The reaction mixture
was stirred at 50 °C for 2-24 h. After that time, the solvents were
evaporated and the residue was dried under vacuum to afford
the pyridinium iodide as a hygroscopic solid that was used
without further purification. To a solution of the corresponding
pyridinium salt (0.5 mmol) in 12 mL of deoxygenated CH2Cl2/
MeOH (3/1) was added a deoxygenated solution of Na2S2O4
(10.0 mmol) in 0.5 M aqueous Na2CO3 (12.5 mL) and the
mixture was vigorously stirred under nitrogen in the dark for
16 h (the color of the solution quickly turned from orange to
yellow). The organic layer was separated, the aqueous phase was
extracted with CH2Cl2 (deoxygenated; 2 ꢀ 10 mL), and the
combined organic phase was dried (MgSO4) and concentrated
in vacuo to give the crude corresponding dihydropyridine,
which was used in the following reaction without further
purification. These compounds can be kept for 24-48 h at
-30 °C without oxidation or decomposition.
General Procedure for the Biomimetic Reduction of Methyl
Benzoylformate with the NADH Models 5a-f and 8. All reac-
tions were conducted in the dark, under a nitrogen atmosphere,
at room temperature, in sealed NMR tubes or flasks. In a typical
run, a solution of the NADH model (0.103 mmol, 1.1 equiv),
Mg(ClO4)2 (0.103 mmol, 1.1 eq., 23.0 mg), and methyl benzoyl-
formate (0.093 mmol) in 0.7 mL of deoxygenated CD3CN was
prepared in a dried NMR tube in the dark and the reaction was
followed by 1H NMR. After complete consumption of the
NADH model, H2O was added and the aqueous layer was
extracted with CH2Cl2 (4 ꢀ 2 mL). The combined organic layers
were dried over MgSO4 and the solvent was evaporated under
reduced pressure. The crude reaction mixture was submitted to
HPLC analysis although in some cases the resulting mandelate
was isolated by preparative thin layer chromatography (eluent:
hexanes/EtOAc). Enantiomeric excesses were determined by
HPLC, using chiral stationary phases (Chiralcel OD) and
hexanes/isopropanol as the mobile phase. Detailed reaction
conditions, yields, and HPLC conditions are provided in the
Supporting Information (Table S1).
ꢀ
Acknowledgment. We thank the Ministerio de Educacion
y Ciencia (MEC, Spain) (Project: CTQ2006-13891/BQU),
UPV-EHU, and Gobierno Vasco (ETORTEK inanoGUNE
IE-08/225) for financial support. Thanks are also due to
SGI/IZO-SGIker for the generous allocation of computa-
tional resources and to SGIker (UPV/EHU) for NMR
facilities. A grant from Gobierno Vasco to R.M.F. is
acknowledged.
Experimental Section
General Procedure for the Preparation of 3-Alkyl-1-[bis-
(trimethylsilyl)methyl]-3-nicotinamidoazetidin-2-ones 7a-f. To
a solution of the nicotinoyl chloride hydrochloride (3.6 mmol,
0.64 g) in 40 mL of dry CH3CN cooled to -5 °C under nitrogen
atmosphere were added N,N-diisopropylethylamine (DIPEA)
(24 mmol, 4.2 mL) and a solution of the corresponding R-amino-
β-lactam (3 mmol) in 10 mL of CH3CN. The reaction mixture
was stirred for 1 h at 0 °C, then at room temperature (5 to 18 h,
monitored by 1H NMR). After that time, CH2Cl2 (20 mL) was
added and the solution was washed with 0.1 M NaOH (20 mL),
NH4Cl (20 mL), and H2O (20 mL). The organic phase was
decanted and dried over MgSO4. The solvents were evaporated
Supporting Information Available: Preparation procedures,
physical and spectroscopic data for compounds 5a-f, 7a-f, and
8, HPLC chromatograms for ee determination, NMR spectra of
complexation and diffusion studies, Cartesian coordinates of all
computed stationary points, relative and absolute activation
energies for all reactions, and complete ref 22. This material is
available free of charge via the Internet at http://pubs.acs.org.
6702 J. Org. Chem. Vol. 74, No. 17, 2009