FULL PAPER

DOI: 10.1002/chem.200900406

Bifunctional Catalysis by Natural Cinchona Alkaloids:

A Mechanism Explained
Clotilde S. Cucinotta,*[a] Monica Kosa,[a] Paolo Melchiorre,[b] Andrea Cavalli,[c] and
Francesco L. Gervasio*[d]
Dedicated to the Centenary of the Italian Chemical Society

Abstract: The use of bifunctional chiral

catalysts, which are able to simultaneously bind and activate two reacting
partners, currently represents an efficient and reliable strategy for the stereoselective preparation of valuable
chiral compounds. Cinchona alkaloids
such as quinine and quinidine, simple
organic molecules generously provided
by Nature, were the first compounds to
be proposed to operate through a cooperative catalysis. To date, a full
mechanistic characterization of the

dual catalysis mode of cinchona alkaloids has proven a challenging objective, due to the transient, non-covalent
nature of the involved catalystsubstrate interactions. Here, we propose a
mechanistic rationale on how natural
cinchona alkaloids act as efficient biKeywords: ab-initio calculations
alkaloids asymmetric synthesis
organocatalysis

reaction
mechanisms

[a] Dr. C. S. Cucinotta, Dr. M. Kosa

Computational Science, Department of Chemistry
and Applied Biosciences, ETH Zurich c/o USI-Campus
Via Giuseppe Buffi 13, 6900 Lugano (Switzerland)
Fax: (+ 41) 58-666-4817
E-mail: cucinotc@ethz.ch
[b] Dr. P. Melchiorre
Dipartimento di Chimica Organica A. Mangini
Alma Mater StudiorumUniversit di Bologna
Viale Risorgimento, 4, 40136 Bologna (Italy)
[c] Dr. A. Cavalli
Dipartimento di Scienze Farmaceutiche
Alma Mater StudiorumUniversit di Bologna
Via Belmeloro 6, 40126 Bologna (Italy)
Unit of Drug Discovery and Development
Italian Institute of Technology, Via Morego 30
16163 Genova (Italy)
[d] Dr. F. L. Gervasio
Computational Science, Department of Chemistry
and Applied Biosciences, ETH Zurich c/o USI-Campus
Via Giuseppe Buffi 13, 6900 Lugano (Switzerland)
Computational Biophysics Group, Structural Biology
and Biocomputing Programme
Spanish National Cancer Research Centre
c/Melchor Fdez. Almagro 3, 28029 Madrid (Spain)
E-mail: flgervasio@cnio.es
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.200900406.

Chem. Eur. J. 2009, 15, 7913 7921

functional catalysts by using a broad

range of computational methods, including classical molecular dynamics, a
mixed quantum mechanical/molecular
mechanics (QM/MM) approach, and
correlated ab-initio calculations. We
also unravel the origin of enantio- and
diastereoselectivity, which is due to a
specific network of hydrogen bonds
that stabilize the transition state of the
rate-determining step. The results are
validated by experimental evidence.

Introduction
Bifunctional catalysis has recently emerged as an important
strategy in small molecule asymmetric catalysis.[1, 2] This approach finds inspiration in the elegance and selectivity of
enzymatic catalysis,[3] which exploits a series of specific interactions that provide both pre-organization of the substrates and stabilization of the transition state (TS) structures.[4] In 1981, Wynberg and Hiemstra[5] conducted systematic studies on the enantioselective conjugate addition of aromatic thiols to cycloalkenones catalyzed by natural cinchona alkaloids such as quinine (Q) and quinidine (QD). These
studies, building on the pioneering work on proline-catalyzed Robinson annulation confirmed that simple naturallyoccurring organic molecules can act as efficient bifunctional
chiral catalysts[6] (Scheme 1).
On the basis of experimental evidence, they hypothesized
a bifunctional mode of catalysis that exploits both the quinuclidine moiety, via general base catalysis, and the hydroxyl
group on C9, via hydrogen-bond (H-bond) interactions, to
simultaneously activate the reagents.
This seminal concept was surprisingly underrated by the
chemical community for almost two decades, probably as a

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

7913

Scheme 1. Natural cinchona alkaloids and their dual activation tools.

consequence of the modest enantioselectivity generally observed in the Q-catalyzed asymmetric transformations.[7]
However, this report had greater consequences than expected, since it lies at the heart of asymmetric organocatalysis.[811] As Wynberg predicted,[5] chemists have recently
started to recognize the advantages of bifunctional catalysis
in preparation of enantiomers by catalytic chiral synthesis.
The engineering of semisynthetic organocatalysts has been
focused on the introduction of tunable H-bond donor
groups to improve the dual activation ability.[1214]
Although natural cinchona alkaloids can be considered
the prototype of bifunctional organocatalysts, their ability to
promote highly stereoselective processes exploiting a dual
activation path has not been clearly established. The mechanistic characterization of catalytic asymmetric transformations involving transient, non-covalent catalystsubstrate interactions is still not understood.
Recently, computational methods have proven to be very
useful in studying and understanding the bifunctionality and
reaction mechanism of thiourea-based organocatalysts.[1519]
Here, also encouraged by those successful studies, we carried out the first comprehensive computational study, using
a variety of computational methods and tools to elucidate
the mechanism of substrate activation by natural cinchona
alkaloids.
We studied the direct conjugate addition of 1,3-dicarbonyl
compounds to maleimides [Eq. (1)], recently discovered by
some of us, as the reaction model.[20]

The choice of the Equation (1) as the model appeared to

be suitable since, to date, it represents one of the few conjugate addition strategies in which natural cinchona alkaloids
can induce high levels of stereoselectivity.[21, 22] Moreover,
the highly stereoselective one-step construction of such
highly congested products (having two adjacent stereogenic
carbon atoms, one of which is quaternary by all carbon substitution) generally depends on the capability of the catalyst
to effectively provide a series of specific interactions, generating a highly structured TS. Kinetic studies have established
that the Q-catalyzed conjugate addition follows a first-order

7914

www.chemeurj.org

dependence on the catalyst, the nucleophile, and the electrophile. Furthermore, the experimental evidences: i) influence
of the reaction media, that is H-bond-acceptor or -donor solvents such as THF or MeOH or CH3CN, respectively, leading to a drastic decrease in stereoselectivity,[20] and ii) the
significantly higher reactivity and selectivity of Q compared
to the o-benzoyl protected derivative, point to an efficient
cooperative mode of catalysis. The observed structurereactivity and structurestereoselectivity relationships are consistent with the notion that Q acts as an efficient bifunctional
catalyst that exploits both the quinuclidine moiety and the
OH group to simultaneously activate and orient the Michael
donor and the acceptor by means of a network of H-bond
interactions.
We focused our attention on the reaction between acyclic
tert-butyl ketoester 1 and phenyl maleimide 2 catalyzed by
Q [Eq. (2)].

Our results support a bifunctional catalysis that proceeds

through a highly structured TS and account for the observed
stereochemical outcome of the process. Moreover, we found
that the formation of a specific H-bond between the hydroxyl group of Q and a carbonyl-oxygen of maleimide 2 stabilizes the transition state of the rate-determining step and is
responsible of the observed high level of stereo-control of
both relative and absolute configurations.
Catalytic cycle description: The reaction model investigated
herein is reported in Equation (2). The deprotonation of 1
by the quinuclidine nitrogen, the most basic site in the catalyst, promptly occurs and leads to the formation of a chiral
ion pair 4 (Figure 1). The binary complex 4 is the reactive
intermediate that can add to N-phenylmaleimide 2 in the
rate- and selectivity-determining step, affording an enolate
anion intermediate 6; a subsequent, fast proton-transfer
originates the 1,4-adduct 3 and regenerates the active catalyst Q. When the electrophilic maleimide 2 approaches this
binary complex 4, a noncovalent ternary complex 5 is likely
formed, which can assume several conformations.

Computational Details
The theoretical methods used to gain more insight into the reaction
mechanism varied from classical molecular mechanics, to density functional theory (DFT), to the mixed quantum mechanical/molecular mechanics (QM/MM) approach, to highly correlated post HartreeFock
methods. The first part of our study was devoted to find the most stable
conformations for the noncovalent complexes 4 and 5. The initial confor-

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 2009, 15, 7913 7921

Bifunctional Catalysis

FULL PAPER
mented in the CP2 K program.[40, 41] The use of QM/MM enabled us to decrease the needed computational resources while describing with sufficient accuracy both the CC coupling step, which is taken care by the
quantum subsystem, and the dispersive staking interaction which are well
described by the MM force-field. The quantum subsystem was modeled
both with the HCTH120 and BLYP exchange and correlation functionals.
The atoms of the MM region were described by the same GAFF potential used for the classical calculations (see Figure 2).

Figure 1. Cooperative catalysis in the quinine-catalyzed asymmetric conjugate addition. Mechanism of the reaction here investigated [Eq. (2)].
The noncovalent binary (4) and ternary (5) complexes are also shown.
The two main steps of the reaction, that is, the C C bond formation and
the protonation of the succinimide moiety, are highlighted. The formation of the zwitterionic binary complex (4) was almost barrierless.

mations of 4 and 5 were built using the coordinates of the isolated molecules, optimized at HF/6-31G* level. Force field parameters and atom
types were assigned using the Antechamber suite and the general Amber
force field GAFF.[23] The atomic point charges of each molecule were obtained using a restrained electrostatic potential (RESP[24]) fitting procedure.
To sample the conformations of 4 and 5, 100 and 130 ns MD simulations,
respectively, in the gas phase were performed. The temperature was kept
at 350 K, by a Langevin dynamics scheme.[25] Following the scheme of
ref. [26], snapshots were collected every 10 ps. Each snapshot was
quenched by rescaling the temperature to 0 K in 15 ps. The resulting
23 000 structures were clustered by using the algorithm described in
Daura et al.[27] with a RMSD cut-off of 1.5 . Geometry and relative stability of 4 were studied and compared with those of 5. The effect of the
solvent was taken into account by performing MD simulations of the ternary clusters in a box of CH2Cl2 (i.e., the solvent in which the reaction
was experimentally performed). The parameters of ref. [28] were used for
the solvent. The formation energy (at 0 Kelvin) of the most populated
ternary clusters was calculated with a correlated ab-initio method (RIMP2)[29] and an aug-cc-pVDZ basis set, using the Turbomole package.[30, 31] The effect of basis set superposition error was taken into account by performing full counterpoise calculations.[32, 33] The effect of
ester substituent in the stability of ternary cluster, was also highlighted
by studying the ternary complex built with the cyclic ethyl-ketoester (see
Supporting Information).
The formation energies DEMP 2 were calculated as the difference between
the total energy of the ternary clusters, in their MD quenched geometries, and the total energy of the isolated molecules (including ghost partners), in the same configuration assumed within each cluster. This computationally expensive method was needed to properly calibrate the delicate balance between electrostatic, dispersive and H-bonding interactions
which keep the complex together, including possible polarization effects.[34] First MP2 energies were compared with single-point B3LYP[35]
energies performed on the MM optimized geometries. Next the MP2 energies were compared with B3LYP, QM/MM-BLYP[36, 37] and QM/MMHCTH120[38] energies after full geometrical relaxations.
The mechanisms of the addition reaction were studied using the nudged
elastic band technique (NEB),[39] within the QM/MM scheme, as imple-

Chem. Eur. J. 2009, 15, 7913 7921

Figure 2. Partitioning of the system in the QM/MM calculations. The

light gray atoms are treated by DFT, the dark gray atoms are parameterized with GAFF force field.

The guess paths for the NEB were obtained by optimizing the endpoints,
and linearly interpolating their coordinates to obtain the intermediate
images. Each path was defined by up to eight replicas, corresponding to
an average distance between the images below 1.7 . The path minimizations were performed in the climbing image implementation,[39] keeping
the endpoints fixed.
To estimate the role of substituents in stereoselective addition and assess
the reaction mechanism, full (i.e., non-QM/MM) DFT simulation on
smaller model systems were performed. Those were obtained by including only the atoms directly involved in the reaction (see Figures 1, 6 and
Figure S6, Supporting Information). The bifunctional quinine catalyst
was modeled by separate NMe3 and MeOH groups. The B3LYP TS were
located with or without an H-bond between Q and 2, both with and without the presence of NMe3. The relative activation energies were compared.
All the energies are in kcal mol
correction.

and do not include zero-point energy

Experimental Section
General methods: See Supporting Information for further details.
Synthetic protocol: The reaction of Equation (2) was carried out in undistilled solvent without any precautions to exclude water. In an ordinary
vial equipped with a magnetic stirring bar, Q (19.4 mg, 0.06 mmol) was
dissolved in CH2Cl2 (0.6 mL). After addition of b-ketoester 1 (52 mg,
0.3 mmol), the tube was closed with a rubber stopper and the mixture
was stirred at room temperature for 10 min. Then, N-phenylmaleimide 2
(62.5 mg, 0.36 mmol) was added in one portion and stirring was continued for 72 h. Then the crude reaction mixture was diluted with CH2Cl2
(2 mL) and flushed through a plug of silica, using CH2Cl2/Et2O 1:1. Solvent was removed in vacuo, and the crude mixture was analyzed by
1
H NMR analysis, showing that the reaction reached completion. The residue was purified by flash chromatography (hexane/AcOEt 80:20) to
yield the desired 1,4-adduct 3 as a white solid (93 mg, 90 %).

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

www.chemeurj.org

7915

C. S. Cucinotta, F. L. Gervasio et al.

The diastereomeric ratio was determined to be 9:1 by 1H NMR analysis

on the crude mixture (d.r. 9:1, determined by integration of two sets of
1
H NMR signal: dmajor 2.28 ppm, dminor 2.31 ppm, s/dmajor 3.41 ppm (dd, J =
6.4, 9.6 Hz, 1 H), dminor 3.51 ppm (dd, J = 6.0, 9.6 Hz, 1 H). The enantiomeric ratio of the major diastereomer was determined to be 93:7 by
HPLC analysis on a Chiralpak AD-H column: 75:25 hexane/iPrOH, flow
rate 0.75 mL min 1, l = 214, 254 nm; major diastereomer (86 % ee):
tminor = 11.8 min, tmajor = 12.3 min. [a]D
RT = + 17.78 (c = 1.06, CHCl3, 86 %
ee); 1H NMR (400 MHz, CDCl3): d = 1.50 (s, 9 H), 1.64 (s, 3 H), 2.28 (s,
3 H), 2.67 (dd, J = 6.4, 18.4 Hz, 1 H), 2.99 (dd, J = 9.6, 18.4 Hz, 1 H), 3.41
(dd, J = 6.4, 9.6 Hz, 1 H), 7.297.32 (m, 2 H), 7.367.40 (m, 1 H), 7.44
7.48 ppm (m, 2 H); 13C NMR (100 MHz, CDCl3): d = 19.6 (CH3), 27.1
(CH3), 27.8 (CH3, 3 C), 32.9 (CH2), 45.1 (CH), 62.3 (C), 83.6 (C), 126.5
(CH), 128.6 (CH, 2 C), 129.1 (CH), 131.9 (C), 169.8 (C), 174.9 (C), 176.7
(C), 205.0 ppm (C); HRMS: m/z: calcd for C19H23NO5 : 345.15762; found:
345.1575.

In complex 5 a, one of the carbonyl oxygens of 2 is Hbonded to the hydroxyl group of the quinine. Stacking interactions also play a relevant role in stabilizing the cluster:
the phenyl substituent of 2 is located with a parallel displaced geometry (centroidcentroid distance is 4.09 ) in
front of the aromatic hetero-naphthyl-like (hn-l) plane of
the quinine (see Figures 3, 4 and Figure S1, Supporting Information).

Results and Discussion

Structure and energetic of the non covalent complexes: MD
simulations and cluster analysis were performed on the
binary and ternary complexes 4 and 5, reported in Figures 1
and 3 (see also Figure S1, Supporting Information). Neither
the formation energies of the ternary complex 5, nor the stereochemistry of the reaction adducts could be reliably predicted by extrapolating those of the binary complex 4 (see
Figure S2 in the Supporting Information). However, we
found some geometrical correspondence between the most
populated clusters of 4 and 5 (as shown below).
Here, we focused on the ternary complex 5. A detailed
cluster analysis allowed us to find several (meta)stable conformations for the ternary complex.
In Figure 3, the four most populated and stable conformations of 5 (5 ad), which represent the 83 % of the whole
130 ns of MD trajectory, are reported. Notably, during the
MD simulations, Q always remained in its gauche-open conformation,[42] in agreement with previous studies.[4346]
In 5 ad, both the planar moieties of the reactants (maleimide 2 and enolate ion 1) are in close proximity (Figure 3
and Figure S1, Supporting Information). The stereochemical
outcome of the process depends on the face approached by
the reactants. The reaction channels departing from complexes 5 ab lead to the formation of the experimentally observed R quaternary stereocenter in the reaction product 3,
while those departing from complexes 5 cd lead to the formation of an S stereocenter. Furthermore, depending on
which carbon of 2 is engaged in the addition process, one of
the possible diastereoisomeric products will be obtained.
Notably, the inter-conversion among clusters 5 ab to 5 cd
occurred on times of the order of 50 ns. This allowed us to
individuate well defined clusters, from which the reaction
channels depart.
In the ternary complexes 5 ad, the three molecules are
kept in close proximity by electrostatic, dispersive, and Hbond interactions. The catalyst Q interacts with the enolate
ion through both the protonated quinuclidine nitrogen and
the OH group.

7916

www.chemeurj.org

Figure 3. Ternary clusters. Most stable ternary complexes 5 a ( 149 kcal

mol 1), 5 b ( 150 kcal mol 1), 5 c ( 148 kcal mol 1), and 5 d ( 149 kcal
mol 1) are reported in part a), b), c), and d), respectively (DEMP2 refers
to the formation energy). The dashed lines show H-bond between Q and
molecular moieties of 1 and 2, which are explicitly indicated for sake of
clarity.

Figure 4. Cluster 5 a, representing the starting configuration for the addition reaction. This reaction proceeds through two steps, schematized with
arrows: the C C bond formation and the protonation of the succinimide
moiety of intermediate 6, after the release of the proton from the activated quinuclidine. The stereochemistry of the final products is also indicated.

In cluster 5 b, carbonyl groups of 1 form H-bonds with

amine and hydroxyl group of Q. Dispersive interactions between the phenyl group of 2 and the CH2 substituent of the
amine also stabilize 5 b. 1 and Q are rotated by 2408 with respect to their conformation in cluster 5 a. The tert-butyl substituent faces both the OMe tail of Q and its CH2 substituent.

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 2009, 15, 7913 7921

Bifunctional Catalysis

FULL PAPER

The arrangement of 2 and 1 in 5 c is almost the mirror

image of 5 b, apart from slightly different distances between
residues. The H-bonds between 1 and Q are the same of
cluster 5 b with the notable difference of the hydroxyl group,
which now is bound to the carbonyl adjacent to the tertbutyl substituent. The phenyl group of 2 is engaged in a dispersive interaction with the unsaturated tail of quinine. 1
and Q are rotated by 1808 with respect to their conformation in cluster 5 b.
In cluster 5 d, the two carbonyl O atoms of 1 form Hbonds with both the amine and the hydroxyl group of Q.
Amine group of Q forms H-bonds also with one of the carbonyl groups of 2. The cluster is additionally stabilized by a
stacking interaction between the aromatic ring of 2 and the
hn-l group of Q. The geometry is parallel displaced (centroidcentroid distance is 4.19 ). 1 and Q, are rotated by
608 with respect to their conformation in cluster 5 c.
Remarkably, the H-bond between the hydroxyl group of
Q and 2 is only observed in the ternary complex 5 a, in
which the mutual arrangement of the molecules leads to the
formation of the experimentally observed R,S final product
3. As shown below, this H-bond induces a polarization
effect in the p system (electrophilic activation), which activates the addition to the Michael acceptor by stabilizing the
addition TS. Moreover, it selects the carbon of 2 involved in
the nucleophilic attack with 1. By contrast, H-bond formed
between the hydroxyl group of Q and carbonyl groups of 1
(as in clusters 5 bd) decreases the nucleophilicity of this
latter, disfavouring the addition.
If this scenario holds true, it would validate a bifunctional
mode of catalysis by natural cinchona alkaloids, the mechanistic path originally proposed by Wynberg.[5]
In Table 1 the total formation energies (in kcal mol 1) of
the ternary complexes 5 ad, calculated at the RI-MP2 level
(DEMP2) and the corresponding energies calculated using
GAFF (DEGAFF) are reported. While the difference between
the energies calculated with the two methods is relatively
small (the maximum relative error is 2.5 %), the relative stability is different. The DEMP2 are more favorable for cluster
5 ab than for 5 cd. The opposite is true for the DEGAFF, in
which cluster 5 cd are slightly more favored. This discrepancy is due to the fact that the classical energies underesti-

Table 1. Formation energies in kcal mol of the ternary clusters 5, calculated with the GAFF force field (GAFF DETOT) and with RI-MP2
(MP2 DETOT). Electrostatic and van der Waals contributions to the
GAFF formation energy are labeled GAFF DEES and GAFF DEVDW,
respectively. Restricted Hartree Fock and correlation contribution to the
MP2 formation energy are labeled MP2 DEES and MP2 DECorr, respectively.
Cluster

5a

5b

5c

5d

GAFF DEES
GAFF DEVDW
GAFF DETOT
MP2 DEES
MP2 DECorr
MP2 DETOT

123
16
139
113
36
149

129
14
143
116
34
150

129
15
144
114
36
150

129
15
144
110
38
148

Chem. Eur. J. 2009, 15, 7913 7921

mate the electrostatic contribution. This is apparent for

Table 1, where the total energies are partitioned into electrostatic and dispersive contributions. Notably, the largest
error is observed for cluster 5 a, in which the polarization of
the p system, introduced when the H-bond is established between the OH of Q and one of the carbonyls of 2, is severely under-estimated by the classical force-field. This limitation of the classical non polarizable GAFF force-field required to evaluate the formation energies with the more accurate MP2 approach, including possible polarization effects.
Even if the most favorable MP2 binding energies are
found for the cluster that would lead to the most abundant
experimental products, the energy differences between clusters are too small to unequivocally select one complex as
the starting configuration for subsequent reaction. This is
true even considering the uncertainties in MP2 energies,
which are expected to be of the order a few kcal mol 1.[34] To
further assess this point, we calculated the internal energies
of 5 ad, using fully relaxed geometries with B3LYP and
QM/MM-HCTH120 and QM/MM-BLYP approaches. The
cluster energy differences remained in all cases negligible
and very close to those calculated at MP2 level. The same is
true for the reaction adducts. The energy differences of the
diastereomeric products (0.15 kcal mol 1 in favor of the RR
(SS) configuration), calculated after B3LYP geometry optimizations, were too small to explain the observed enrichment. Therefore, the reaction is kinetically controlled.
Some interesting information can be finally obtained comparing 4 and 5. We find that the geometries of the most
populated binary clusters 4 ad correspond to those adopted
by Q and 1 in the corresponding ternary complexes 5 ad.
We also find that 4 is more stable whenever the ester functionality is close to Q moiety, rather than positioned far
away from it (see Figure S2, Supporting Information). Thus,
a clever selection of the ketoester substituent could stabilize
a specific geometry of the ion pair intermediate, increasing
the selectivity of the reaction. A geometry control of the nucleophilic species is thus provided. A significant effect of the
size of the ester group on the stereoselectivity is also highlighted studying the addition of cyclic ethyl-ketoester to
maleimides[47] (see Figure S3, Supporting Information).
Reaction mechanism: To find the most energetically favorable reaction pathway(s), we calculated the potential energy
along various paths starting from the ternary clusters 5 ad,
using the NEB approach and QM/MM.
As shown in Figures 1 and 4, after the formation of the
zwitterionic non-covalent ternary complex (5), the reaction
proceeds via two steps: the C C bond formation and the
protonation of the succinimide moiety of intermediate 6,
after the release of the proton from the activated quinuclidine nitrogen.
We compared the activation energies of the paths departing from the clusters 5 ac (5 d does not lead to a stable intermediate), in which C C bond formation occurs before
maleimide protonation (it will be shown soon that this rep-

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

www.chemeurj.org

7917

C. S. Cucinotta, F. L. Gervasio et al.

resents the most favorable reaction mechanism). We first

studied the paths represented in Figure 1, originating from
complexes 5 a and 5 c, which would lead to the formation of
R and S quaternary stereocenters in the product 3. Apart
from cluster 5 a, where the selection of the b-carbon of 2
was determined by the electron-withdrawing action of the
H-bond between Q and 2, for each complex the criterion to
select which unsaturated carbon of 2 was attacked, was represented by the relative distance between the reaction centres.
In Figure 5, the comparison between the activation energies for the reaction pathways from 5 a and 5 c, is reported.
It can be noted that the energies of the TSs (TSa and TSc)

Figure 5. The reaction pathways obtained from 5 a and 5 c, that is, the
OH-catalyzed and uncatalyzed reaction mechanisms, respectively. The
TSs (TSa and TSc) geometries are reported along with the associated
energy (activation energies in kcal mol 1) calculated at QM/MM level of
theory using the NEB approach. 6 a (g) and 6 c (c) represent the
corresponding anionic intermediates. For sake of clarity, 6 a and 6 c structures are reported in the supplementary information.

were in both cases very close to the energies of the corresponding reaction intermediates (6 a and 6 c) obtained by
the CC coupling step. More interestingly, the activation
energy was 13.9 and 24.8 kcal mol 1 for the Q-OH assisted
and non-assisted reaction mechanisms, respectively. The remarkable energy difference (9.8 kcal mol 1) is due to the polarization effect of the H-bond and is mainly responsible for
a bifunctional activation mode of catalysis, which leads to
the formation of the R quaternary stereocenter. Within the
uncertainties of QM/MM-NEB calculations, this is in qualitatively good agreement with the experimentally observed
enantiomeric excess of 86 % (see also Figure S4, Supporting
Information).
The reaction pathway obtained departing from 5 a also
well explained the diastereomeric outcome of the reaction.
In addition to providing an electrophilic activation and stabilizing the TS, the specific H-bond interaction between Q
and a carbonyl group of 2 determines the stereoselectivity
of the process by directing the Maleimide approach and,

7918

www.chemeurj.org

more importantly, by selecting the electrophilic b-carbon

(with respect to the carbonyl group of 2 that is H-bonded
with Q) involved in the conjugate addition step. Thus, the rigidified geometry of the highly structured TSa was at the
basis of the high diastereoselectivity. Notably, the activation
energy of the subsequent step (TSa-2 - see Figure S5, Supporting Information), namely the protonation of the intermediate 6 a, was 3.2 kcal mol 1 higher than TSa and
6.4 kcal mol 1 lower than TSc. The activation energy of the
protonation step leading to the reaction product (3) was
3.4 kcal mol 1, while the highest activation energy of the
mechanism was 17.1 kcal mol 1. Interestingly, the rigidified
geometry of TS and the low barrier for maleimide protonation support an intra-complex proton transfer, rather than
proton exchange with the solvent. These results were assessed by QM/MM calculations at BLYP level of theory
(see Figure S4, Supporting Information).
Besides those starting from 5 a and 5 c, we studied the reaction pathway from 5 b, also leading to the major R,S-3
final product. The energy of the TS of this reaction pathway
was 17.0 kcal mol 1. Even if less stable than the OH activated TSa, TSb (namely, the transition state relative to the
path departing from 5 b, see Figure S5, Supporting Information) is stabilized relative to TSc by the presence of a notable interaction between both unsaturated C atoms of 2 and
1. This is due to the peculiar arrangement of the bulky tertbutyl ester substituent with respect to Q. This occurs despite
the arrangement of 2 and 1 in 5 c is almost the mirror image
of that assumed in 5 b and the interaction networks holding
the two complexes together are very similar.
As stated above, the reaction starting from 5 d was finally
investigated. It was characterized by a high barrier, since the
C C bonded intermediate was thermodynamically not
stable. Once the covalent intermediate was formed, it
promptly evolved back to the starting reactants.
We also calculated the energetics of the possible competitive reactions starting with the protonation of 2 and proceeding with C C bond formation. After relaxation with the
NEB we found a reaction path in which the C C bond formation was concerted with protonation of maleimide. The
TS energies are found to be higher than the stepwise C C
bond formation/deprotonation. In particular, the competitive addition pathway for 5 a has a relative barrier of
36 kcal mol 1. In Figure S5 (Supporting Information), inset
TSa-bis the geometry of the relative TS is shown. Similarly
we found high TS for the competitive concerted reactions
starting from 5 b, 5 c. This result is consistent with the fact
that the protonation of 2 deactivates the addition reaction
by decreasing its electrophilic character.
We finally excluded that, in reaction starting from 5 ad,
protonation of 1 occurred before C C bond formation.
Indeed it was not possible to locate thermodynamically
stable intermediates: in all cases, immediately after the
bond between (protonated) 1 and 2 was formed, it relaxed
back to the starting reagents.

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 2009, 15, 7913 7921

Bifunctional Catalysis

FULL PAPER

Reaction mechanism on model systems: To further proof

the importance of the catalyzing action of the Q hydroxyl
group on the addition reaction, and understand the role of
substituents [R1, R2, R3, and R4 in Eq. (1)], we performed a
series of DFT calculations on model systems that retain only
the salient characteristics of the reactant/catalyst ternary
complex (see Figure 6).

retical and experimental investigations support the proposal

that the Q-catalyzed process proceeds through a cooperative
mechanism. This study illustrates how computational techniques are becoming a very valuable tool to predict the catalysts behavior.[48] In analogy with previous reports on thiourea-based organocatalysts,[1517] we find that Q serves as a bifunctional catalyst, with its ammonium group binding the
enolate and its hydroxyl orienting the maleimide via H-bond.
The computational approach
adopted also well accounts for
the observed high diastereoselectivity of the process. Whereas the computational characterization of the cooperative enantioselective reactions that involve strong catalystsubstrate
interactions (e.g., via enamine
formation[2, 12] or Lewis acid/
Lewis base dative bonding[49])
has been possible in several
cases, such a type of complex
systems, involving transient,
non-covalent interactions that
lead to the stereoselective construction of highly congested
products, require large computational efforts.[13] On these
grounds, we believe that the reFigure 6. Catalyzed and uncatalyzed reaction pathways for the model system, including the NMe3 group.
liability and effectiveness of the
adopted method might be
The calculations on the model clusters were performed at
useful to explore other type of reaction systems.
the B3LYP/6-311 + GACHTUNGRE(d,p) level of theory. The model system
Our study involved two main steps: i) the construction of
a reliable non-covalent ternary complex; ii) the investigation
considered here, is shown in Figure 6. In 2 the phenyl group
of all energetically possible reaction pathways of a mechawas replaced by methyl group, while 1 was simplified reducnism involving three distinct chemical species.
ing the substituents. The bifunctional action of Q is modeled
Determining the most stable conformation of the non-coby two separate entities (MeOH and HNMe3 + groups).
valent ternary complex called for extensive MD simulations
As shown in Figure 6, the difference between the OH acand cluster analysis. This step of the study was of paramount
tivated and non activated TS results in an energy difference
importance as different conformations of the ternary comof 3 kcal mol 1, the catalyzed path being lower in energy.
plex lead to different stereoisomers. However, enantioselecThe anion intermediate of the uncatalyzed reaction could
tivity cannot be rationalized by the energy differences benot be located in this case.
tween the diastereomeric ternary complexes leading to the
These results confirmed the catalytic function of the OH
formation of the opposite enantiomers. The nature of the
group. The comparison with QM/MM-HCTH120 and QM/
catalyst employed in the conjugate addition has a great inMM-BLYP energies, showed that the action of the substitufluence on the distereomeric ratio of the reaction, Q being
ents destabilized the TS. This result further confirms that
the only one able to afford high level of selectivity. This acthe addition of carbon anions to alkenes is affected by a
counts for a kinetic controlled process. For instance, the use
subtle interplay between geometry and charge transfer
of a catalytic amount of Et3N affords compound 3 with a
during the reaction, as well as the polarization induced by
specific substituents.
55:45 distereomeric ratio. Since the reaction is kinetically
driven, several reaction paths have been investigated. The
activation energies calculated at QM/MM level of theory
using the NEB approach univocally pointed to the stereoisoSummary and Conclusions
mer experimentally observed as that kinetically accessible
one. A specific H-bond between one of the carbonyls of
A comprehensive mechanistic study on the asymmetric Mimaleimide 2 and the hydroxyl group of Q assisted the C C
chael addition of ketoesters to phenyl-maleimide, catalyzed
bond formation, stabilizing the TS (by 9.8 kcal mol 1) that
by natural cinchona alkaloids, has been reported. Both theo-

Chem. Eur. J. 2009, 15, 7913 7921

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

www.chemeurj.org

7919

C. S. Cucinotta, F. L. Gervasio et al.

leads to the experimentally observed product. Thus we identified a dual role for the Q hydroxyl group. First, it orients
the maleimide scaffold for the nucleophilic attack by the ketoester; second, it is responsible for the electrophilic activation of the maleimide by means of a electron withdrawing
effect on the carbonyl b-carbon, an interaction that select
the carbon atom attacked by the ketoester. This mechanistic
scenario is strongly supported by experiments that have
shown that either the protection of the OH of Q with a benzoyl group, or the use of a polar solvent, that may interfer
with the H-bond network in the TS, drastically reduce the
stereoselectivity.
We finally highlighted the role of substituents in the addition, showing that the stereoselectivity of the process is affected by the action of the substituents. A clever selection of
those, could on one hand stabilize binary clusters, and, on
the other hand, fix the b-ketoester in a convenient position
for subsequent reaction.

[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]

Acknowledgements

[21]

We thank Dr. Matteo Masetti, Dr. Bernd Ensing, Prof. Victoria Buch
and Prof. Michele Parrinello for useful discussions. The CSCS is gratefully acknowledged for providing computational resources. Dr. Maria
Grazia Giuffreda is acknowledged for technical support.

[22]
[23]
[24]

[1] D. Cahard, J. A. Ma, Angew. Chem. 2004, 116, 4666 4683; Angew.
Chem. Int. Ed. 2004, 43, 4566 4583.
[2] C. Allemann, R. Gordillo, F. R. Clemente, P. H. Y. Cheong, K. N.
Houk, Acc. Chem. Res. 2004, 37, 558 569.
[3] K. Dranz, H. Waldmann, Enzymatic Catalysis in Organic Synthesis,
Wiley-VCH, Weinheim, 1995.
[4] R. B. Silverman, The Organic Chemistry of Enzyme-catalyzed Reactions, Academic Press, New York, 2002.
[5] H. Hiemstra, H. Wynberg, J. Am. Chem. Soc. 1981, 103, 417 430.
[6] Bifunctional activation by naturally-occurring organocatalysts was
previously proposed for the proline-catalyzed Robinson annulation,
see: a) Z. G. Hajos, D. R. Parrish, J. Org. Chem. 1974, 39, 1615
1621. Hajos and Parrish interpreted their results as a simplified
model of biological system in which (S)-proline plays the role of an
enzyme. Later, Agami and Kagan proposed the involvement of
two proline molecules in the transition state of this intramolecular
aldol reaction, and they included this example in their seminal
report on nonlinear correlation between catalyst enantiomeric
excess and product enantiopurity in asymmetric catalysis: b) C.
Agami, Bull. Soc. Chim. Fr. 1988, 3, 499 507; c) C. Puchot, O.
Samuel, E. Duach, S. Zhao, C. Agami, H. B. Kagan, J. Am. Chem.
Soc. 1986, 108, 2353 2357. It was only recently that the bifunctional
activation mode by a single molecule of proline in the aldol reaction
has been widely supported by experimental evidence and theoretical
investigations: d) B. List, L. Hoang, H. Martin, Proc. Natl. Acad.
Sci. USA 2004, 101, 5839 5842; e) S. Bahmanyar, K. N. Houk, H. J.
Martin, B. List, J. Am. Chem. Soc. 2003, 125, 2475 2479; f) C. Allemann, R. Gordillo, F. R. Clemente, P. H. Y. Cheong, K. N. Houk,
Acc. Chem. Res. 2004, 37, 558 569. For further studies that rationalize the findings of Kagan and co-workers in a manner that remains
compatible with the currently accepted one proline reaction mechanism and reconciles reports of both linearity (ref. [6e]) and nonlinearity (ref. [6c]) of proline-catalyzed aldolization, see: g) M. Klussmann, S. P. Mathew, H. Iwamura, D. H. Wells, Jr., A. Armstrong,
D. G. Blackmond, Angew. Chem. 2006, 118, 8157 8160; Angew.
Chem. Int. Ed. 2006, 45, 7989 7992; h) M. Klussmann, H. Iwamura,

7920

www.chemeurj.org

[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]

[35]
[36]

[37]
[38]
[39]
[40]
[41]
[42]

S. P. Mathew, D. H. Wells, Jr., U. Pandya, A. Armstrong, D. G.

Blackmond, Nature 2006, 441, 621 623.
K. Kacprzak, J. Gawronski, Synthesis 2001, 961 998.
P. I. Dalko, Enantioselective Organocatalysis: Reactions and Experimental Procedures, Wiley, New York, 2007.
D. W. C. MacMillan, Nature 2008, 455, 304 308.
P. Melchiorre, M. Marigo, A. Carlone, G. Bartoli, Angew. Chem.
2008, 120, 6232 6265; Angew. Chem. Int. Ed. 2008, 47, 6138 6171.
M. Movassaghi, E. N. Jacobsen, Science 2002, 298, 1904 1905.
M. S. Taylor, E. N. Jacobsen, Angew. Chem. 2006, 118, 1550; Angew.
Chem. Int. Ed. 2006, 45, 1520.
T. Marcelli, J. H. van Maarseveen, H. Hiemstra, Angew. Chem. 2006,
118, 7658 7666; Angew. Chem. Int. Ed. 2006, 45, 7496 7504.
S. J. Connon, Chem. Commun. 2008, 2499 2510.
S. J. Zuend, E. N. Jacobsen, J. Am. Chem. Soc. 2007, 129, 15872
15883.
A. Hamza, G. Schubert, T. Sos, I. Ppai, J. Am. Chem. Soc. 2006,
128, 13151 13160.
P. R. Schreiner, A. Wittkopp, Org. Lett. 2002, 4, 217 220.
P. R. Schreiner, Chem. Soc. Rev. 2003, 32, 289 296.
P. Hammar, T. Marcelli, H. Hiemstra, F. Himo, Adv. Synth. Catal.
2007, 349, 2537 2548.
G. Bartoli, M. Bosco, A. Carlone, A. Cavalli, M. Locatelli, A. Mazzanti, P. Ricci, L. Sambri, P. Melchiorre, Angew. Chem. 2006, 118,
5088 5092; Angew. Chem. Int. Ed. 2006, 45, 4966 4970.
B. Trk, M. Abid, G. London, J. Esquibel, M. Torok, S. C. Mhadgut, P. Yan, G. K. Prakash, Angew. Chem. 2005, 117, 3146 3149;
Angew. Chem. Int. Ed. 2005, 44, 3086 3089.
S. Lou, B. M. Taoka, A. Ting, S. E. Schaus, J. Am. Chem. Soc. 2005,
127, 11256 11257.
J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. Case, J.
Comput. Chem. 2004, 25, 1157 1174.
C. I. Bayly, P. Cieplak, W. Cornell, P. A. Kollman, J. Phys. Chem.
1993, 97, 10269 10280.
R. J. Loncharich, B. R. Brooks, R. W. Pastor, Biopolymers 1992, 32,
523 535.
F. L. Gervasio, P. Procacci, G. Cardini, A. Guarna, A. Giolitti, V.
Schettino, J. Phys. Chem. B 2000, 104, 1108 1114.
X. Daura, I. Antes, W. F. van Gunsteren, W. Thiel, A. E. Mark, Proteins 1999, 36, 542 555.
T. Fox, P. A. Kollman, J. Phys. Chem. B 1998, 102, 8070 8079.
D. Feller, E. D. Glendening, D. E. Woon, M. W. Feyereisen, J.
Chem. Phys. 1995, 103, 3526 3542.
R. Ahlrichs, M. Br, M. Hser, H. Horn, C. Klmel, Chem. Phys.
Lett. 1989, 162, 165 169.
F. Weigend, M. Hser, Theor. Chem. Acc. 1997, 97, 331 340.
S. F. Boys, F. Bernardi, Mol. Phys. 1970, 19, 553.
H. B. Jansen, P. Ros, Chem. Phys. Lett. 1969, 3, 140.
A good estimate of stacking energies can be obtained already at
MP2 level, using an appropriate basis-set. Here we used the aug-ccpVDZ basis set to give a reliable estimate of dispersive interaction,
in agreement with the values obtained at CCSD(T) level (cf. K. E.
Riley, P. Hobza, J. Phys. Chem. A 2007, 111, 8257 8263).
A. D. Becke, J. Chem. Phys. 1993, 98, 5648 5652.
a) C. Lee, W. Yang, R. G. Parr, Phys. Rev. B 1988, 37, 785; b) B.
Miehlich A. Savin, H. Stoll, H. Preuss, Chem. Phys. Lett. 1989, 157,
200.
A. D. Becke, Phys. Rev. A 1988, 38, 3098.
A. D. Boese, N. L. Doltsinis, N. C. Handy, M. Sprik, J. Chem. Phys.
2000, 112, 1670.
G. Henkelman, B. P. Uberuaga, H. Jnsson, J. Chem. Phys. 2000,
113, 9901 9904.
T. Laino, F. Mohamed, A. Laio, M. Parrinello, J. Chem. Theory
Comput. 2006, 2, 1370 1378.
J. VandeVondele, M. Krack, F. Mohamed, M. Parrinello, T. Chassaing, J. Hutter, Comput. Phys. Commun. 2005, 167, 103 128.
The geometry of the two conformers of Q was fully optimized by
means of DFT-based calculations at the B3LYP/6-31GACHTUNGRE(d,p)//B3LYP/
3-21Gd level of theory, showing that the gauche-open conformation

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Chem. Eur. J. 2009, 15, 7913 7921

Bifunctional Catalysis

[43]
[44]
[45]
[46]

FULL PAPER

is favored both in protonated and natural Q, by 0.53 and 0.59 kcal

mol 1, respectively.
G. D. H. Dijkstra, R. Kellogg, H. Wynberg, J. Org. Chem. 1990, 55,
6121 6131.
G. D. H. Dijkstra, R. M. Kellogg, H. Wynberg, J. S. Svedsen, I.
Marko, K. B. Sharpless, J. Am. Chem. Soc. 1989, 111, 8069 8076.
G. S. Cortez, S. H. Oh, D. Romo, Synthesis 2001, 1731 1736.
G. Vayner, K. Houk, Y. K. Sun, J. Am. Chem. Soc. 2004, 126, 199
203.

Chem. Eur. J. 2009, 15, 7913 7921

[47] The strong effect of the size of the ester group on the stereoselectivity of this process has been originally observed, see ref. [20].
[48] K. N. Houk, H. Y. Cheong, Nature 2008, 455, 309 313.
[49] L. P. C. Nielsen, C. P. Stevenson, D. G. Blackmond, E. N. Jacobsen,
J. Am. Chem. Soc. 2004, 126, 1360 1362.

 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Received: February 13, 2009

Published online: June 2, 2009

www.chemeurj.org

7921