Constrained β-proline analogues in organocatalytic aldol reactions: The influence of acid geometry

Article abstract of DOI:10.1021/jo900840v

(Chemical Equation Presented) 7-Azabicyclo[2.2.1]heptane-2-carboxylic acid 11 was prepared in enantiopure form, and its catalytic potential in the direct aldol reaction between acetone and 4-nitrobenzaldehyde was assessed. The bicyclic system was found to

Full text of DOI:10.1021/jo900840v

Constrained ꢀ-Proline Analogues in Organocatalytic Aldol
Reactions: The Influence of Acid Geometry
Alan Armstrong,* Yunas Bhonoah, and Andrew J. P. White
Department of Chemistry, Imperial College London, South Kensington, London SW7 2AZ, U.K.
a.armstrong@imperial.ac.uk
ReceiVed April 24, 2009
7-Azabicyclo[2.2.1]heptane-2-carboxylic acid 11 was prepared in enantiopure form, and its catalytic
potential in the direct aldol reaction between acetone and 4-nitrobenzaldehyde was assessed. The bicyclic
system was found to be more selective than its monocyclic analogue ꢀ-proline 5b. A comparative density
functional theory study of proline 1, ꢀ-proline 5b, and 11 in the latter reaction revealed the origin of the
improved enantioselectivity of 11 over 5b. The geometry of the carboxylic acid group in the transition
states, which depended critically on pyrrolidine ring conformation, was found to play a key role.
SCHEME 1. Proline-Catalyzed Intermolecular Aldol
Reaction
Introduction
Amino acid organocatalysis has received much attention since
the discovery in 2000 by List, Lerner, and Barbas1 that proline
1 could catalyze the direct intermolecular aldol reaction (Scheme
1). This seminal work and subsequent efforts have had a
significant impact on the organic chemistry community, provid-
ing a rapid and efficient approach to the synthesis of chiral
molecules.2 Interestingly, the foundations of the organocatalytic
mode of action of proline were laid almost 40 years ago by
Hajos and Parrish3 and, independently, Eder, Sauer, and
Wiechert,4 in an intramolecular aldol cyclization. The scope of
transformations that proline can catalyze is truly remarkable and
includes the synthetically important aldol and Mannich reactions;
high diastereo- and enantioselectivities are generally observed.
Computational5-7 and experimental1,8 studies of the orga-
nocatalytic aldol reaction have provided support for a one-
proline mechanism based on enamine activation similar to the
accepted mechanism for enzymatic aldol reactions (Scheme 2).1
These investigations have established that the acid functionality
is essential for high catalyst activity and selectivity, potentially
through hydrogen-bonding activation of the electrophile and
stabilization of the transition states (List-Houk model).5 An
alternative mechanism involving oxazolidinones as product-
(1) List, B.; Lerner, R. A.; Barbas, C. F. J. Am. Chem. Soc. 2000, 122, 2395–
2396.
(2) Mukherjee, S.; Yang, J. W.; Hoffmann, S.; List, B. Chem. ReV. 2007,
107, 5471–5569.
(3) (a) Hajos, Z. G.; Parrish, D. R. German Patent DE 2102623, 1971. (b)
Hajos, Z. G.; Parrish, D. R. J. Org. Chem. 1974, 39, 1615–1621.
(4) (a) Eder, U.; Sauer, G.; Wiechert, R. German Patent DE 2014757, 1971.
(b) Eder, U.; Sauer, G.; Wiechert, R. Angew. Chem., Int. Ed. 1971, 10, 496.
(5) Bahmanyar, S.; Houk, K. N.; Martin, H. J.; List, B. J. Am. Chem. Soc.
2003, 125, 2475–2479.
(7) (a) Bahmanyar, S.; Houk, K. N. J. Am. Chem. Soc. 2001, 123, 11273–
11283. (b) Bahmanyar, S.; Houk, K. N. J. Am. Chem. Soc. 2001, 123, 12911–
12912. (c) Clemente, F. R.; Houk, K. N. Angew. Chem., Int. Ed. 2004, 43, 5766–
5768.
(8) (a) List, B.; Hoang, L.; Martin, H. J. Proc. Natl. Acad. Sci. U.S.A. 2004,
101, 5839–5842. (b) Hoang, L.; Bahmanyar, S.; Houk, K. N.; List, B. J. Am.
Chem. Soc. 2003, 125, 16–17. (c) Zotova, N.; Franzke, A.; Armstrong, A.;
Blackmond, D. G. J. Am. Chem. Soc. 2007, 129, 15100–15101. (d) Zotova, N.;
Broadbelt, L. J.; Armstrong, A.; Blackmond, D. G. Bioorg. Med. Chem. Lett. In
press. (e) Zhu, H.; Clemente, F. R.; Houk, K. N.; Meyer, M. P. J. Am. Chem.
Soc. 2009, 131, 1632–1633.
(6) Allemann, C.; Gordillo, R.; Clemente, F. R.; Cheong, P. H. Y.; Houk,
K. N. Acc. Chem. Res. 2004, 37, 558–569.
10.1021/jo900840v CCC: $40.75  2009 American Chemical Society
Published on Web 06/01/2009
J. Org. Chem. 2009, 74, 5041–5048 5041

Armstrong et al.
SCHEME 2. Proposed Mechanisms of Proline Catalysis
SCHEME 3. Synthetic Approach to Constrained ꢀ-Proline
Analogues
determining species has recently been proposed by Seebach,
Eschenmoser, and colleagues;9 in this model, the electrophile
is postulated to approach anti to the carboxylic acid group
without any hydrogen-bonding activation from the organocata-
lyst (Scheme 2).
Most investigations into the development of novel amino acid
organocatalysts have focused on derivatizing the proline scaf-
fold.2 Studies on other scaffolds, especially where the pyrrolidine
ring has been constrained or the spatial relationship between
the amine and acid functionalities modified (e.g., ꢀ-amino acids),
are less common. A selection of such catalysts is shown in
Figure 1.10 Importantly, ꢀ-amino acids have shown remarkable
selectivity differences compared to proline-derived catalysts,
e.g., ꢀ-proline 5b promotes the classic organocatalytic Mannich
reaction with high anti diastereoselectivity,10c whereas proline
provides predominantly the syn product.11 cis-Pentacin 8 has
been successfully employed in the intramolecular aldol cycliza-
tion to generate the opposite enantiomeric product compared
to the proline-catalyzed reaction.10f Applications of ꢀ-amino
acids to the direct intermolecular aldol reaction remain
limited.10e,12
Our group has recently reported the preparation of constrained
ꢀ-proline analogue rac-1013 by an aza-Prins-pinacol rearrange-
ment strategy,14 and we wondered whether the free ꢀ-amino
acid would be capable of promoting organocatalytic reactions
in a similar manner to proline (Scheme 3). We envisaged that
the structural constraints imposed by the bicyclic system may
lead to better organization of the transition state and conse-
quently influence reaction stereoselectivity. In their study of 4,5-
methanoprolines, Hanessian and Houk15 indeed found that
constraints on the pyrrolidine ring impacted upon the planarity
of the iminium species formed during the intramolecular aldol
reaction and hence affected the enantiomeric excess of the final
product. The ꢀ-disposition of the carboxylic acid group on our
system should also eliminate catalyst decomposition via ir-
reversible decarboxylation, a common side reaction plaguing
proline-catalyzed aldol reactions.8c Furthermore, we anticipated
that the bicyclic amino acid would have somewhat better
solubility than proline in solvents commonly employed in
organocatalysis. To the best of our knowledge, there have been
noreportsontheuseofaminoacidsbasedonthe7-azabicyclo[2.2.1]-
heptane motif in organocatalysis, although Shinisha and Sunoj16
recently disclosed a density functional theory (DFT) study of
[2.2.1] and [2.1.1] bicyclic R-amino acids in the classic
organocatalytic aldol reaction. The calculations predicted ste-
reoselectivities higher than those using proline, although
experimental verification remains to be undertaken.
Results and Discussion
1. Catalyst Synthesis. We previously reported the prepara-
tion of N-Ts acid rac-10.13 Our synthetic strategy toward the
latter compound cannot easily be rendered asymmetric. We thus
initially attempted classical resolution of rac-10 by fractional
crystallization of its diastereomeric salt with various chiral
amines.17 Disappointingly, this proved unsuccessful in providing
enantiopure 10. Covalent attachment of a chiral auxiliary to the
racemic acid was more fruitful (Scheme 4); the Evans oxazo-
lidinone diastereomeric adducts 12 and 13 were separable by
flash chromatography. Cleavage of the chiral auxiliary from the
separated diastereomers gave the desired enantiomerically pure
(9) Seebach, D.; Beck, A. K.; Badine, D. M.; Limbach, M.; Eschenmoser,
A.; Treasurywala, A. M.; Hobi, R.; Prikoszovich, W.; Linder, B. HelV. Chim.
Acta 2007, 90, 425–471.
(10) (a) Catalyst 4: Terakado, D.; Takano, M.; Oriyama, T. Chem. Lett. 2005,
34, 962–963. (b) Catalyst 5a: Mitsumori, S.; Zhang, H.; Cheong, P. H. Y.; Houk,
K. N.; Tanaka, F.; Barbas, C. F. J. Am. Chem. Soc. 2006, 128, 1040–1041. (c)
Catalyst 5b: Zhang, H. L.; Mitsumori, S.; Utsumi, N.; Imai, M.; Garcia-Delgado,
N.; Mifsud, M.; Albertshofer, K.; Cheong, P. H. Y.; Houk, K. N.; Tanaka, F.;
Barbas, C. F. J. Am. Chem. Soc. 2008, 130, 875–886. (d) Catalyst 6: Kano, T.;
Takai, J.; Tokuda, O.; Maruoka, K. Angew. Chem., Int. Ed. 2005, 44, 3055–
3057. (e) Catalyst 7: Dwivedi, N.; Bisht, S. S.; Tripathi, R. P. Carbohydr. Res.
2006, 341, 2737–2743. (f) Catalyst 8: Davies, S. G.; Russell, A. J.; Sheppard,
R. L.; Smith, A. D.; Thomson, J. E. Org. Biomol. Chem. 2007, 5, 3190–3200.
(11) Notz, W.; Tanaka, F.; Watanabe, S.; Chowdari, N. S.; Turner, J. M.;
Thayumanavan, R.; Barbas, C. F. J. Org. Chem. 2003, 68, 9624–9634.
(12) Limbach, M. Tetrahedron Lett. 2006, 47, 3843–3847.
(13) Armstrong, A.; Bhonoah, Y.; Shanahan, S. E. J. Org. Chem. 2007, 72,
8019–8024.
(14) Armstrong, A.; Shanahan, S. E. Org. Lett. 2005, 7, 1335–1338.
(15) Cheong, P. H. Y.; Houk, K. N.; Warrier, J. S.; Hanessian, S. AdV. Synth.
Catal. 2004, 346, 1111–1115.
(16) Shinisha, C. B.; Sunoj, R. B. Org. Biomol. Chem. 2007, 5, 1287–1294.
(17) Amines used include (R)-1-phenylethylamine, (R)-1-p-tolylethylamine,
(R)-1-(4-methoxyphenyl)ethylamine, (R)-1-phenylpropylamine, (R)-1-(naphtha-
len-2-yl)ethylamine, (1R,2S)-2-(dimethylamino)-1-phenylpropan-1-ol, (-)-cin-
chonidine, and quinidine.
5042 J. Org. Chem. Vol. 74, No. 14, 2009

ꢀ-Proline Analogues in Organocatalytic Aldol Reactions
SCHEME 4. Preparation of Enantiopure Bicyclic Amino Acids 11a
a
n
Conditions: (i) pivaloyl chloride, Et3N, THF; (ii) (S)-4-isopropyl-oxazolidin-2-one, BuLi, THF; (iii) flash chromatography.
FIGURE 1. Constrained or ꢀ-amino acids employed in organocatalysis.
N-Ts acids 10 in good yield. The absolute configuration of
(1R,2S,4S)-10 was confirmed by X-ray crystallography (see
Supporting Information). Final removal of the tosyl group
completed the synthesis of enantiopure bicyclic amino acids
FIGURE 2. Conversion versus time graphs for the aldol reaction shown
in Scheme 1 catalyzed by ꢀ-proline 5b or (-)-11 in DMSO; 20%
catalyst used in all cases.
11.18
2. Catalyst Assessment. There have been no reports on the
use of ꢀ-proline 5b in the direct aldol reaction between acetone
and 4-nitrobenzaldehyde, although Barbas and colleagues10c
have shown that this amino acid is a highly efficient catalyst
for the anti-selective Mannich reaction. We assessed the catalytic
potential of 5b in the aldol reaction shown in Scheme 1 in both
DMSO and DMF.19 While good reactivity was observed (Figure
2), we found that ꢀ-proline afforded essentially racemic aldol
product (Table 1). It was clear that increasing the distance
between the secondary amine functionality and the carboxylic
acid by an extra carbon has a substantial negative impact on
enantioselectivity in the aldol reaction (cf. entries 1 and 2, Table
1), in stark contrast to the organocatalytic Mannich reaction.10c
Having established the effect on enantioselectivity of changing
the position of the carboxylic acid group, we turned our attention
to the investigation of structural constraints on the pyrrolidine
system through the use of optically pure amino acid (-)-11.
TABLE 1. Observed Enantioselectivities of Aldol Product 3 with
Amino Acids 1, 5b, and 11
aldol product ee (%)
entry
catalyst
DMSO
DMF
1a
2
3
Pro 1
ꢀ-Pro 5b
(-)-11
76
5
32
76
0
40
a Proline data from ref 1.
The bicyclic catalyst was found to be slightly less soluble than
ꢀ-proline as judged by visual inspection and exhibited lower
reactivity (Figure 2). Interestingly, the enantiomeric excesses
of the aldol product were found to be higher with (-)-11 than
with ꢀ-proline, although the former catalyst failed to reach the
selectivity levels observed with proline 1 (Table 1). Both (-)-
11 and 1 gave the same sense of asymmetric induction favoring
the (R)-aldol product, suggesting similar mechanisms of action
via the List-Houk transition state (Scheme 2).5 Our aldol results
thus show how constraining the pyrrolidine ring of ꢀ-proline
in a bicyclic framework can have a positive impact on
enantiomeric excess (from 0% to 40% ee in DMF).20 This
remarkable change in selectivity led us to conduct a comparative
study of proline, ꢀ-proline, and bicyclic amino acid 11 using
(18) Our optical rotation data for (-)-(1S,2R,4R)-11 was not in agreement
with that previously reported for its enantiomer (1R,2S,4S)-11 by Pandey, G.;
Laha, J. K.; Lakshmaiah, G. Tetrahedron 2002, 58, 3525–3534. ([R]25D-5.27
(c 1.2, MeOH)). These workers prepared the latter compound via an asymmetric
[3 + 2] cycloaddition involving Oppolzer’s acryloyl sultam but did not provide
proof of the absolute configuration of (1R,2S,4S)-11. We can only assume that
they derived the stereochemistry from models for facial selectivity of the
dienophile as put forward by Curran, D. P.; Kim, B. H. Tetrahedron 1993, 49,
293. In our work, the absolute configurations of both enantiomers of amino acid
11 are based on the X-ray structure of (1R,2S,4S)-10; there is no possibility of
racemization during the deprotection reaction leading to the free amino acid.
(19) ꢀ-Proline 5b was obtained from its commercial N-Boc derivative
following deprotection with TFA; see: Zhang, H. L.; Mifsud, M.; Tanaka, F.;
Barbas, C. F. J. Am. Chem. Soc. 2006, 128, 9630–9631.
(20) Other solvents attempted with amino acid (-)-11 include acetone (36%
ee), acetonitrile (34% ee), and NMP (18% ee); the reaction rates in these were
reduced compared to DMSO and DMF.
J. Org. Chem. Vol. 74, No. 14, 2009 5043

Armstrong et al.
SCHEME 5. Reaction Pathways Studied by DFT
DFT methods to gain further insight into the nature of the
transition states derived from each catalyst.
3. DFT Study. DFT calculations employing the B3LYP
functional21 have been shown to reliably predict product ratios
in a variety of proline-catalyzed reactions including the aldol,
Mannich, aminoxylation, and R-alkylation reactions.6,22 Such
studies can also help in the design of novel, more efficient
catalysts as evidenced by the work of Barbas, Houk, and
colleagues10b on the use of ꢀ-proline analogue 5a in the anti
Mannich reaction; DFT studies on 5a prior to catalyst synthesis
predicted selectivities of 95:5 dr and 98% ee.
Previous computational work on the organocatalytic aldol
reaction has revealed that the enantioselectivity is controlled
during the C-C bond formation between the enamine inter-
mediate and the aldehyde electrophile.5,23,24 A key proton
transfer from the carboxylic acid group to the developing
alkoxide anion helps in stabilizing the transition states (TS).
Thus the approach of the aldehyde is always syn to the acid
group. The enamine intermediate can adopt either the s-cis or
s-trans conformation, with the double bond either syn or anti
to the carboxylic acid, respectively. Additionally, the aldehyde
offers two prochiral faces (si and re). In the present study, we
have investigated the two enamines (anti and syn) and four
diastereomeric transition states (anti-re, anti-si, syn-re, and syn-
si) for each of the three amino acids in the gas phase and in
solvent (Scheme 5). The anti-re and syn-re structures lead to
the R-aldol product, whereas the anti-si and syn-si TSs lead to
the enantiomeric S-product. Geometry optimizations were
performed at the B3LYP level21 using the 6-31G** basis set.25
Single-point energies were then computed at the same level of
theory using the more flexible 6-311+G** basis set.
FIGURE 3. B3LYP/6-31G** optimized TS geometries for the proline-
catalyzed aldol reaction of Scheme 1. Distances are given in angstroms.
∆E† values refer to relative energies at CPCM(DMSO)/B3LYP/6-
311+G**//B3LYP/6-31G**.
been previously reported.24 These were reoptimized at the
B3LYP/6-31G** level in the present work and then used for
single-point energy calculations at higher levels. The computa-
tions revealed that the syn enamine 14a was slightly more stable
than its anti counterpart 14b, seemingly due to steric interactions
between the bulkier methyl group and the carboxylic acid in
the latter. The enamine nitrogen adopted a nearly planar
geometry in both conformers due to favorable overlap of the
lone pair with the alkene system (see Supporting Information).23
Our results were in agreement with those reported previously.24
The calculated structures of the four diastereomeric transition
states 15a-15d are shown in Figure 3. Similar theoretical
studies have been carried out by Sinisha and Sunoj,16 although
the coordinates of the optimized TS geometries of 15a-15d
were not disclosed. Our results paralleled those previously
reported for isobutyraldehyde and acetaldehyde acceptor
electrophiles.23,24 Transfer of the carboxyl proton to the alkoxide
was observed in all cases; the three atoms engaging in the
H-bond were almost co-linear (θ ) 172-173°). We found that
3.1. Proline System. The optimized geometries of the anti
and syn enamines formed between proline and acetone have
(21) (a) Becke, A. D. J. Chem. Phys. 1993, 98, 5648–5652. (b) Lee, C. T.;
Yang, W. T.; Parr, R. G. Phys. ReV. B 1988, 37, 785–789.
(22) Fu, A. P.; List, B.; Thiel, W. J. Org. Chem. 2006, 71, 320–326.
(23) Arno, M.; Domingo, L. R. Theor. Chem. Acc. 2002, 108, 232–239.
(24) Rankin, K. N.; Gauld, J. W.; Boyd, R. J. J. Phys. Chem. A 2002, 106,
5155–5159.
(25) Hehre, W. J.; Radom, L.; Schleyer, P. V.; Pople, J. A. Ab Initio
Molecular Orbital Theory; Wiley: New York., 1986.
5044 J. Org. Chem. Vol. 74, No. 14, 2009

ꢀ-Proline Analogues in Organocatalytic Aldol Reactions
TABLE 2. Computed Absolute (∆Eabs†) and Relative (∆E†)
Activation Barriers for the Reaction of Scheme 1 at CPCM(DMSO)
B3LYP/6-311+G**//B3LYP/6-31G**
/
a b
a b
,
∆E†
†
,
catalyst
TS
∆Eabs
1
15a
15b
15c
15d
17a
17b
17c
17d
19a
19b
19c
19d
3.47 (6.37)
4.98 (8.15)
6.10 (9.18)
0.00 (0.00)
1.51 (1.78)
2.18 (2.84)
2.87 (3.28)
5.54 (2.54)
0.48 (1.30)
0.06 (0.20)
0.00 (0.00)
0.00 (0.00)
0.78 (1.72)
5.88 (6.56)
4.90 (5.33)
6.80 (9.62)
5b
11.49 (14.24)
6.43 (13.01)
6.25 (11.96)
6.19 (11.76)
5.84 (8.39)
6.63 (10.10)
12.02 (15.03)
11.05 (13.80)
(-)-11
a Energies in kcal/mol b Values in brackets refer to gas-phase
†
†
calculations (∆H298abs and ∆H298
) at B3LYP/6-311+G**//B3LYP/
6-31G**.
transition states derived from the anti enamine were more stable
than those arising from the syn conformer. The syn-re and syn-
si structures 15c and 15d tended to have increased pyrrolidine
ring puckering as indicated by the data in Figure 3; in these
structures, the C3-carbon was out-of-plane, forcing the car-
boxylic group to adopt a pseudoaxial orientation. Houk and co-
workers5 have shown in similar studies that the anti TSs are
further stabilized by additional weak electrostatic interactions
between the partially positive R′-H and the forming alkoxide
anion. In our anti transition structures, the NCHδ+ · · ·Oδ-
distances varied from 2.39 to 2.42 Å. The pseudoequatorial
disposition of the aromatic moiety of the aldehyde away from
the carboxylic acid group gave rise to an almost completely
staggered arrangement around the forming C-C bond in the
anti-re structure 15a, making the latter 1.78 kcal/mol lower in
energy than the anti-si TS 15b in the gas phase (Table 2).
Inclusion of solvent effects led to a stabilization of all four
structures but did not significantly change their relative energies.
The product ratio26 computed from the solvent calculations (92:
8) was in good agreement with experimental enantiomeric
excesses (76%).1
3.2. ꢀ-Proline System. The effect of changing the position
of the carboxylic acid group was next assessed. Calculated
enamine structures of ꢀ-proline showed that the syn enamine
16a was also more stable than the anti conformer 16b, although
the energy difference was smaller than with proline; no
significant interactions were noticed between the enamine and
carboxylic groups. As with the proline enamines, a nearly planar
nitrogen was observed.
Calculation of the transition state structures 17a-17d revealed
interesting results (Figure 4). To achieve the necessary proximity
for proton transfer, three of the four structures (17a, 17c, and
17d) had puckered pyrrolidine rings such that the carbon bearing
the carboxylic acid was out-of-plane. Such distortion put the
acid group in a pseudoaxial disposition. All four structures
showed some electrostatic interaction between the R- or R′-H
and the alkoxide anion (NCHδ+ · · ·Oδ- distances ) 2.08-2.47
Å), although these were stronger in the syn transition states.
Iminium planarities were similar to those observed in the proline
FIGURE 4. B3LYP/6-31G** optimized TS geometries for the ꢀ-pro-
line-catalyzed aldol reaction of Scheme 1. Distances are given in
angstroms. ∆E† values refer to relative energies at CPCM(DMSO)/B3LYP/
6-311+G**//B3LYP/6-31G**.
system, as judged by dihedral angles. However, in contrast to
the proline case, transition states arising from the syn enamine
were more stable. A significant non-ideal geometry of proton
transfer (θ ) 163°) was seen in the anti-re TS 17a with
increased puckering of the pyrrolidine ring. We believe these
factors are responsible for the high energy of the latter structure.
Such distortions were not observed with proline, where the anti-
re TS 15a was in fact the most stable diastereomeric structure.
Steric interactions were also observed between the aromatic
portion of the aldehyde and the carboxylic group in the anti-si
structure 17b. More interestingly, the energy difference between
the lowest transition states 17c and 17d with ꢀ-proline was only
0.20 kcal/mol in the gas phase (Table 2). Our results thus show
that modification of the position of the acid group leads to
drastically reduced facial selectivity for the aldehyde. Inclusion
of solvent effects led to a stabilization of all the transition states
by 13-17 kcal/mol. Importantly, the syn-re TS 17c was lowered
to a greater extent than the syn-si structure 17d such that the
energy difference between the two transition states in solvent
was very small (0.06 kcal/mol), giving a calculated product ratio
of 53:47. This was in remarkable agreement with our experi-
mental enantiomeric excesses (5%). Interestingly, the anti-si TS
17b was found to be only 0.48 kcal/mol higher in energy than
the syn-si TS 17d, suggesting that it might contribute to the
overall ee. Consideration of 17b in the ee calculation led to a
(26) Product ratios were calculated using absolute rate theory: ln(k1/k2) )-
∆∆G/RT. See ref 5. Product ratios computed from solvent calculations were in
better agreement with experimental data than those derived from gas-phase
(∆H298† and ∆G298†) calculations, although the same general trend in the energies
of the transition states were observed in all cases.
J. Org. Chem. Vol. 74, No. 14, 2009 5045

Armstrong et al.
FIGURE 5. B3LYP/6-31G** optimized enamine structures formed
between acetone and amino acid 11. Distances are given in angstroms.
∆E values refer to relative energies at CPCM(DMSO)/B3LYP/6-311+G**//
B3LYP/6-31G**.
product ratio of 62:38, still significantly lower than the proline
value. Finally, the activation barrier in solvent was higher than
that calculated for proline (Table 2). This was partly due to the
greater change in pyrrolidine ring conformation required in the
ꢀ-proline syn enamine to achieve the syn-si transition state
geometry.
3.3. Bicyclic ꢀ-Proline System. With the structural factors
responsible for the complete erosion of enantioselectivity with
ꢀ-proline established, we turned our attention to the study of
the effect of constraining the pyrrolidine ring in a [2.2.1] bicyclic
system. The calculated structures of the enamines formed by
the bicyclic amino acid revealed the syn conformer 18b to be
more stable (Figure 5). In stark contrast to the monocyclic
catalysts, the bicyclic enamines were severely pyramidalized
at nitrogen (σ ) 138°). In both conformers, the disposition of
the carboxylic acid proton seemed to suggest intramolecular
H-bonding with the bridging nitrogen atom (Figure 5).
FIGURE 6. B3LYP/6-31G** optimized TS geometries for the 11-
catalyzed aldol reaction of Scheme 1. Distances are given in angstroms.
Only selected hydrogen atoms are shown. ∆E† values refer to relative
energies at CPCM(DMSO)/B3LYP/6-311+G**//B3LYP/6-31G**.
The calculated structures of the four diastereomeric transition
states 19a-19d are shown in Figure 6. We found that the
structures derived from the anti enamine were more stable,
analogous to the proline system. The geometry of proton transfer
was not ideal for the syn-re TS 19c; severe deviation from
linearity was noticed (θ ) 163°). Furthermore, no significant
interaction was observed between the partially positive R- and
ꢀ′-hydrogen atoms and the alkoxide anion; these factors account
for the high energy of the syn-re TS. Although the syn-si
structure 19d was less prone to these destabilizing effects, it
suffered from high strain due to distortion of the bicyclic system
to achieve ideal proton transfer geometry. This distortion is best
illustrated by comparison of the angle λ and the distance between
the carboxylic and bridgehead carbon atoms (µ) across the four
structures. The destabilization of syn-si was further augmented
by an eclipsing interaction between the carboxylic oxygen and
the bridgehead carbon atom.
Analysis of the anti structures 19a and 19b revealed good
geometry for proton transfer in both cases (θ ) 171°).
Interestingly, the orientation of the carboxylic acid group seemed
to be assisted by the axial ꢀ-hydrogen. Both transition states
enjoyed further stabilization due to electrostatic interactions
between the partially positive ꢀ′-hydrogen and the forming
alkoxide anion (CHδ+ · · ·Oδ- distances ) 2.29-2.33 Å). The
anti-re structure 19a was more stable than its anti-si counterpart
19b by 1.72 kcal/mol in the gas phase (Table 2), with a
completely staggered arrangement around the forming C-C
bond. We attribute the higher energy of the anti-si TS to a steric
clash between the aromatic fragment and the carboxylic group,
along with extra gauche interactions at the forming C-C bond.
As with the previous catalysts, the inclusion of solvent effects
led to stabilization of all four transition states (by 14-16 kcal/
mol). Interestingly, the energy difference between 19a and 19b
was considerably reduced (0.78 kcal/mol) compared to the gas-
phase values. The calculated product ratio in solvent (79:21)
was in fair agreement with experimental data (32% ee). Greater
accuracy can potentially be achieved by conducting full
geometry optimization in the presence of solvent rather than
single-point calculations. These have not been carried out in
the present study because of their extremely high computational
cost. Nevertheless, the DFT results described above successfully
rationalize the experimentally observed trend in selectivity with
the three amino acids and show how the accessibility of the
various transition states involved in the aldol reaction changes
with catalyst modification.
5046 J. Org. Chem. Vol. 74, No. 14, 2009

ꢀ-Proline Analogues in Organocatalytic Aldol Reactions
J ) 4.0 Hz, 1H), 3.49 (dd, J ) 9.0, 5.0 Hz, 1H), 2.39 (s, 3H),
2.34-2.22 (m, 2H), 1.91-1.83 (m, 2H), 1.73 (dd, J ) 12.0, 9.0
Hz, 1H), 1.65-1.61 (m, 1H), 1.49-1.45 (m, 1H), 0.88 (d, J ) 7.0
Hz, 3H), 0.81 (d, J ) 7.0 Hz, 3H) ppm; 13C NMR (101 MHz,
CDCl3) δ 171.2 (C), 154.0 (C), 143.3 (C), 137.7 (C), 129.2 (2 ×
CH), 127.5 (2 × CH), 63.5 (CH2), 62.0 (CH), 58.8 (CH), 58.5 (CH),
47.9 (CH), 34.4 (CH2), 29.3 (CH2), 29.0 (CH2), 28.0 (CH), 21.4
(CH3), 17.8 (CH3), 14.6 (CH3) ppm; MS (CI) m/z 424 (M + NH4+),
407 (M + H+), 317, 310, 204, 132; HRMS calcd for C20H27N2O5S
(M + H+) 407.1641, found 407.1650.
Conclusion
Enantiopure amino acid 11 was prepared, and its catalytic
potential in the direct aldol reaction between acetone and
4-nitrobenzaldehyde was assessed. The bicyclic amino acid
showed reduced activity compared to that of its monocyclic
analogue ꢀ-proline but exhibited greater selectivity, although
it failed to reach the levels observed with the natural amino
acid proline. DFT calculations revealed that ꢀ-proline reacted
via its syn enamine but exhibited no preference for facial
selectivity of the aldehyde. Constraining the ꢀ-proline ring in a
bicyclic system led to stabilization of the anti transition states
compared to the syn counterparts. The anti enamine was able
to differentiate between the two prochiral faces of the aldehyde,
although the facial selectivity was reduced when compared to
that of the proline system. Our studies provide the first
computational rationalization of the poor selectivity of ꢀ-proline
in the classic organocatalytic aldol reaction and show how
constraining the pyrrolidine ring can modify the carboxylic acid
geometry and contribute toward improving enantioselectivity.
We believe that such rigidification strategies can be valuable
in the design and development of improved catalysts. The
remarkable agreement between the calculated and experimental
enantiomeric excesses provides further support for the List-Houk
transition state model and emphasizes the importance of
hydrogen bonding interactions provided by the carboxylic acid
group. The Seebach-Eschenmoser model does not implicate
such interactions in determining stereocontrol. Although we have
not carried out calculations on the latter model, it is satisfying
that stereocontrol in a large number of carbonyl R-functional-
izations has now been shown to be consistent with the
List-Houk model. Work is underway in our laboratory to
synthesize and study analogues of bicyclic amino acid 11 in
the aldol and related reactions.
The more polar diastereomer 13 (0.605 g, 40%) as a colorless oil:
1
[R]21 +72.0 (c 1.0, CHCl3); IR (film) 1772, 1700, 1387 cm-1; H
D
NMR (400 MHz, CDCl3) δ 7.74 (d, J ) 8.0 Hz, 2H), 7.24 (d, J 8.0
Hz, 2H), 4.41-4.37 (d, J ) 4.0 Hz, 1H), 4.30 (d, J ) 4.5 Hz, 1H),
4.28 (t, J ) 4.5 Hz, 1H), 4.24 (t, J ) 8.0 Hz, 1H), 4.18 (dd, J ) 9.0,
3.0 Hz, 1H), 3.48 (dd, J ) 8.5, 5.0 Hz, 1H), 2.55-2.49 (m, 1H), 2.30
(s, 3H), 2.18-2.10 (m, 1H), 2.01-1.93 (m, 1H), 1.90-1.82 (m, 1H),
1.76-1.69 (m, 1H), 1.52-1.47 (m, 2H), 0.85 (d, J ) 7.0 Hz, 3H),
0.71 (d, J ) 7.0 Hz, 3H) ppm; 13C NMR (101 MHz, CDCl3) δ 171.0
(C), 154.1 (C), 143.4 (C), 137.7 (C), 129.4 (2 × CH), 127.6 (2 ×
CH), 63.6 (CH2), 63.4 (CH), 59.2 (CH), 58.5 (CH), 48.0 (CH), 32.4
(CH2), 29.5 (CH2), 29.2 (CH2), 28.4 (CH), 21.5 (CH3), 17.9 (CH3),
14.6 (CH3) ppm; MS (CI) m/z 424 (M + NH4+), 407 (M + H+);
HRMS calcd for C20H27N2O5S (M + H+) 407.1641, found 407.1646.
(1R,2S,4S)-7-Tosyl-7-azabicyclo[2.2.1]heptane-2-carboxylic Acid
(1R,2S,4S)-10 and (1S,2R,4R)-7-Tosyl-7-azabicyclo[2.2.1]heptane-
2-carboxylic Acid (1S,2R,4R)-10. To a solution of oxazolidinone
12 (0.663 g, 1.63 mmol, 1.0 equiv) in THF/H2O (4:1, 40 mL) at 0
°C was added H2O2 (1 mL, 9.80 mmol, 30% wt in H2O, 6.0 equiv),
and the resulting mixture was stirred for 5 min. A solution of
LiOH ·H2O (0.137 g, 3.27 mmol, 2.0 equiv) in H2O (10 mL) was
then added, and the reaction mixture was allowed to warm to rt
and stirred for 5 h. Na2SO3 (0.6 M, 25 mL) and saturated aq
NaHCO3 (25 mL) were then added, and the resulting solution was
concentrated under reduced pressure. The resulting aqueous residue
was washed with chloroform (3 × 30 mL), acidified to pH 1 with
3 M HCl, and extracted with EtOAc (4 × 50 mL). The combined
EtOAc layers were dried (MgSO4) and evaporated under reduced
pressure to give acid (1R,2S,4S)-10 (0.387 g, 80%) as a white solid:
Experimental Section
mp 113 °C; [R]21D 0.0 (c 0.2, CHCl3); IR (nujol) 1700, 1155 cm-1
;
1H NMR (400 MHz, CDCl3) δ 9.13 (br s, 1H), 7.79 (2d, J ) 8.0
Hz, 2H), 7.27 (d, J ) 8.0 Hz, 2H), 4.52 (d, J ) 4.0 Hz, 1H), 4.29
(t, J ) 4.0 Hz, 1H), 2.59 (dd, J ) 9.0, 5.0 Hz, 1H), 2.42 (s, 3H),
2.29-2.23 (m, 1H), 2.01-1.91 (m, 2H), 1.72 (dd, J ) 12.0, 9.0
Hz, 1H), 1.51-1.47 (m, 2H) ppm; 13C NMR (101 MHz, CDCl3) δ
178.1 (C), 143.7 (C), 137.3 (C), 129.5 (2 × CH), 127.6 (2 × CH),
61.6 (CH), 58.9 (CH), 47.6 (CH), 34.6 (CH2), 29.3 (CH2), 29.1
(CH2), 21.6 (CH3) ppm; MS (CI) m/z 313, 296 (M + H+); HRMS
calcd for C14H18NO4S (M + H+) 296.0957, found 296.0962. The
absolute structure of (1R,2S,4S)-10 was confirmed by X-ray of a
single crystal grown from MeOH/H2O (see Supporting Information).
N-Ts acid (1S,2R,4R)-10 was prepared using the same procedure
as for (1R,2S,4S)-10: oxazolidinone 13 (0.605 g, 1.49 mmol, 1.0
equiv), H2O2 (0.91 mL, 8.94 mmol, 30% wt in H2O, 6.0 equiv),
LiOH ·H2O (0.125 g, 2.98 mmol, 2.0 equiv). (1S,2R,4R)-10 (0.312
(S)-4-Isopropyl-3-((1R,2S,4S)-7-tosyl-7-azabicyclo[2.2.1]heptane-
2-carbonyl)oxazolidin-2-one 12 and (S)-4-Isopropyl-3-((1S,2R,4R)-
7-tosyl-7-azabicyclo[2.2.1]heptane-2-carbonyl)oxazolidin-2-
one 13. To a suspension of N-Ts acid rac-10 (1.10 g, 3.73 mmol,
1.0 equiv) in THF (25 mL) at 0 °C was added dropwise
triethylamine (0.93 mL, 6.66 mmol, 1.8 equiv) followed by pivaloyl
chloride (0.69 mL, 5.55 mmol, 1.5 equiv). The resulting hetero-
geneous mixture was stirred at 0 °C for 25 min, allowed to warm
to rt for 40 min, heated to 40 °C for 30 min, and then cooled to
-78 °C.
n
Separately, BuLi (2.22 mL, 5.55 mmol, 2.5 M in hexanes, 1.5
equiv) was added to a solution of (S)-4-isopropyl-oxazolidin-2-
one (0.717 g, 5.55 mmol, 1.5 equiv) in THF (25 mL) at -78 °C,
and the resulting viscous mixture was stirred at that temperature
for 1.5 h. The lithiated oxazolidinone was then added dropwise
via syringe to the solution of the mixed anhydride prepared above
over 30 min. The resulting mixture was maintained at -78 °C for
a further 30 min and then allowed to warm to 0 °C for 30 min and
then to rt for 1.5 h. Saturated aq NH4Cl was then added, the solution
was concentrated, and the resulting aqueous residue was extracted
with EtOAc. Combined organics were washed successively with
saturated aq NaHCO3 and brine, dried (MgSO4), and evaporated
under reduced pressure. Flash chromatography (EtOAc/petrol 10:
90 f 35:65, carried out 3 times) afforded the following compounds.
The less polar diastereomer 12 (0.663 g, 44%) as a colorless
g, 71%) was obtained as a white solid: mp 178 °C; [R]21 0.0 (c
D
1
0.2, CHCl3); IR (nujol) 1700, 1155 cm-1; H NMR (400 MHz,
CDCl3) δ 7.78 (d, J ) 8.0 Hz, 2H), 7.25 (d, J ) 8.0, 2H), 4.50 (d,
J ) 4.0 Hz, 1H), 4.27 (t, J ) 4.0 Hz, 1H), 2.57 (dd, J ) 9.0, 5.0
Hz, 1H), 2.40 (s, 3H), 2.27-2.21 (m, 1H), 2.01-1.89 (m, 2H),
1.71 (dd, J ) 12.0, 9.0 Hz, 1H), 1.54-1.49 (m, 2H) ppm; 13C NMR
(101 MHz, CDCl3) δ 178.1 (C), 143.7 (C), 137.3 (C), 129.5 (2 ×
CH), 127.6 (2 × CH), 61.6 (CH), 58.9 (CH), 47.6 (CH), 34.6 (CH2),
29.3 (CH2), 29.1 (CH2), 21.5 (CH3) ppm; MS (CI) m/z 313 (M +
NH4+), 296 (M + H+); HRMS calcd for C14H18NO4S (M + H+)
296.0957, found 296.0961.
oil: [R]21D +48.0 (c. 1.0, CHCl3); IR (film) 1774, 1700, 1386 cm-1
;
1H NMR (400 MHz, CDCl3) δ 7.75 (d, J ) 8.0 Hz, 2H), 7.24 (d,
J ) 8.0 Hz, 2H), 4.40 (d, J ) 4.0 Hz, 1H), 4.39-4.36 (m, 1H),
4.25 (t, J ) 8.5 Hz, 1H), 4.18 (dd, J ) 9.0, 3.0 Hz, 1H), 4.15 (t,
(1S,2R,4R)-7-Azabicyclo[2.2.1]heptane-2-carboxylic Acid (-)-
11 and (1R,2S,4S)-7-Azabicyclo[2.2.1]heptane-2-carboxylic Acid
(+)-11.18 A mixture of the N-protected amino acid (1S,2R,4R)-10
J. Org. Chem. Vol. 74, No. 14, 2009 5047

Armstrong et al.
(0.700 g, 2.37 mmol, 1.0 equiv) and phenol (0.670 g, 7.12 mmol,
3.0 equiv) was heated under reflux in 48% aq HBr (20 mL) for
5 h. The reaction mixture was then allowed to cool to rt, diluted
with water (20 mL), and extracted with EtOAc. The aqueous layer
was separated and evaporated under reduced pressure to give a crude
brown solid that was dissolved in H2O and loaded on a Dowex
50WX8-100 ion-exchange resin (H+ form, activated with 1 M HCl).
The loaded resin was washed with distilled water and then eluted
with 3 M aq NH3. The latter fractions were concentrated under
reduced pressure, the concentrate was then boiled for 5 min with
activated charcoal and filtered, and the filtrate was evaporated under
reduced pressure. The resulting semisolid was dissolved in toluene/
MeOH (2:1), the solvent was then evaporated, and this procedure
was repeated twice to give amino acid (-)-11 (0.240 g, 72%) as a
energy profile connecting the latter to the two associated minima.
Enthalpies at B3LYP/6-31G** were obtained from frequency
calculations and include zero-point vibrational energy (ZPVE).
Single-point energies were calculated using the more flexible
6-311+G** basis set. Solvation energies for reactants and transition
states were computed using a polarizable continuum model (SCRF-
CPCM)29 with a permittivity of 46.7, the value for DMSO, with
the united-atom Kohn-Sham (UAKS) radii. All quoted solvation
energies refer to the free energy of solvation with all electrostatic
el
terms, G . These have been previously employed in studies of
solv
the organocatalytic aldol reaction.16 Absolute activation barriers
†
†
∆H298abs and ∆Eabs (gas-phase and solvent, respectively) were
obtained as the energy difference between isolated reactants and
the corresponding transition state structures. ∆H298† values refer to
relative gas-phase activation enthalpies, and ∆E† values refer to
relative activation barriers in DMSO. Enamine starting geometries
were initially obtained at AM1 using Spartan ES v.2.0.0.30 The
structures were then optimized at higher levels of theory using
Gaussian 03. Measurements of σ (pyramidalization at nitrogen) and
ω (mean deviation from plane) were carried out using Spartan and
ChemBio3D Ultra 11.031 respectively.
white hygroscopic solid: [R]21 -13.0 (c 0.77, MeOH), -22.0 (c
D
1.0, H2O); IR (ATR) 2956, 2524, 1651, 1562, 1387 cm-1; 1H NMR
(400 MHz, CD3OD) δ 4.23 (m, 1H), 4.17 (br s, 1H), 2.62 (t, J )
7.0 Hz, 1H), 2.07-2.05 (m, 2H), 1.94-1.92 (m, 2H), 1.74-1.72
(m, 2H) ppm; 13C NMR (101 MHz, CD3OD) δ 178.9 (C), 62.9
(CH), 59.3 (CH), 47.3 (CH), 34.2 (CH2), 27.6 (CH2), 26.8 (CH2)
ppm; MS (CI) m/z 142 (M + H+); HRMS calcd for C7H12NO2 (M
+ H+) 142.0868, found 142.0871.
Amino acid (+)-11 was prepared using the same procedure as
for (-)-11: N-Ts acid (1R,2S,4S)-10 (0.400 g, 1.36 mmol, 1.0
equiv), phenol (0.383 g, 4.07 mmol, 3.0 equiv), 48% aq HBr (11
mL). (+)-11 (0.192 g, quantitative) was obtained as a white
hygroscopic solid: [R]20D +24.0 (c 1.0, H2O); IR (ATR) 2956, 2535,
Acknowledgment. We are grateful to the EPSRC, ORSAS
(award to Y.B.), Pfizer, and Merck for financial support of this
research. We also thank Imperial College High Performance
Computing Service for access to the HPC computer clusters.
Stimulating discussions with Prof. Donna Blackmond are
gratefully acknowledged.
1
1563, 1388 cm-1; H NMR (400 MHz, CD3OD) δ 4.23 (m, 1H),
4.17 (br s, 1H), 2.62 (br t, 1H), 2.06-2.04 (m, 2H), 1.94-1.92
(m, 2H), 1.74-1.72 (m, 2H) ppm; 13C NMR (101 MHz, CD3OD)
δ 180.0 (C), 63.1 (CH), 59.2 (CH), 48.0 (CH), 34.3 (CH2), 27.8
(CH2), 26.8 (CH2) ppm; MS (CI) m/z 142 (M + H+); HRMS calcd
for C7H12NO2 (M + H+) 142.0868, found 142.0871.
Computational Methods. All stationary points (reactants and
transition states) were optimized and characterized by frequency
analysis using hybrid density functional theory (B3LYP)21 and the
6-31G**25 basis set as implemented in Gaussian 03, rev. E.01.27
The unique imaginary frequencies for the transition states were
checked to ensure that the frequency indeed pertained to the desired
reaction coordinate. Intrinsic reaction coordinate (IRC)28 calcula-
tions were carried out on selected transition states to verify the
1
Supporting Information Available: Copies of H and 13C
NMR spectra for all new compounds, X-ray crystallographic
data (including CIF file) for (1R,2S,4S)-10, experimental
procedure for the organocatalytic aldol reaction, Cartesian
coordinates, electronic energies and selected dihedral angles of
all calculated structures, and complete citation for ref 27. This
material is available free of charge via the Internet at
http://pubs.acs.org
JO900840V
(29) (a) Cances, E.; Mennucci, B.; Tomasi, J. J. Chem. Phys. 1997, 107,
3032–3041. (b) Cossi, M.; Barone, V.; Cammi, R.; Tomasi, J. Chem. Phys. Lett.
1996, 255, 327–335.
(27) Frisch, M. J., et al. Gaussian 03, ReV. E01. See Supporting Information
for full citation.
(28) (a) Gonzalez, C.; Schlegel, H. B. J. Chem. Phys. 1991, 95, 5853–5860.
(b) Gonzalez, C.; Schlegel, H. B. J. Chem. Phys. 1989, 90, 2154–2161.
(30) Spartan ES, V. 2.0.0; Wavefunction, Inc.: Irvine, CA.
(31) ChemBio3D Ultra 11.0; CambridgeSoft Corporation: Cambridge, MA.
5048 J. Org. Chem. Vol. 74, No. 14, 2009