J. Zhang et al.
the reaction did not occur at all under conditions A, but did
give the [4+3] cycloadduct 4ja in 62% yield under condi-
tions B (Table 2, entries 17 and 18). The structures and rela-
tive stereochemistries of 4ga and the major diastereomer of
3ga were further confirmed by single-crystal X-ray diffrac-
tion analysis (Figure 1).[12]
Scheme 2. Plausible mechanisms for this catalytically tunable, regioselec-
tive 1,3-dipolar cycloaddition.
alkyne,[14] rather than the carbonyl, and, thus, mediates nu-
cleophilic attack to give oxonium-containing vinyl–gold in-
termediate C (cycle 2). This is followed by the regioselective
homo-Michael addition of nitrone 2 at the more substituted
position of the cyclopropyl ring,[6,2a,15] which produces furan-
yl–gold intermediate D; in turn, this, upon ring closure,
gives the formal [4+3] cycloadduct and regenerates the gold
catalyst. The high diastereoselectivity of this reaction may
result from the more favorable chair-like conformation of
intermediate D. An alternative, plausible catalytic cycle
(cycle 3) for the formation of the formal [4+3] cycloadduct,
could not be ruled out by the results described so far. In this
catalytic cycle, a key carbocationic furanyl–gold intermedia-
te E would be formed. Subsequent nucleophilic addition of
the nitrone to the carbocation would produce intermediat-
e D, which, in turn, would give the final product by the
same ring closure step as in cycle 2.
Figure 1. X-ray crystal structures of compounds 3ga (upper) and 4ga
(lower).
In order to gain an insight into the mechanism of the for-
mation of compounds 4, a control experiment without the
addition of the nitrone was performed. This showed that the
cycloisomerization of 1-(1-alkynyl)cyclopropyl ketone 1a
into the trisubstituted furan is very slow under conditions B,
indicating that the nitrone is involved in the cyclopropyl
ring-opening step, which is consistent with Schmalzꢂs
work.[2a] Further studies (Scheme 3) on the kinetic resolu-
tion[6f] of a racemic mixture of 1a and the transformation of
optically active ent-1a provided further strong supporting
evidence for the proposed cycle (cycle 2, Scheme 2). For ex-
ample, the reaction of racemic 1a with nitrone 2a under the
catalysis of the (R)-C3-TunePhos-derived[16] gold complex
was quenched, before completion, after 29 h and gave a
55% yield of ent-4aa in 38% ee and 19% of ent-1a, in
84% ee, was recovered. However, if the reaction was al-
lowed to go to completion a racemic mixture of cycloadduct
4aa was formed. Interestingly, optically active ent-1a react-
ed smoothly with nitrone 2a under the catalysis of
Furthermore, the gold-catalyst loading can be reduced to
only 0.2 mol% and the reaction still proceeds smoothly with
no reduction in either the yield, or the diastereoselectivity
of a 2.5 mmol scale reaction, which makes this method more
practical. For example, the formal [4+3] cycloaddition reac-
tion of 1a (2.5 mmol) with 2a (2.75 mmol) can be performed
in 10 min to give a 94% yield of 4aa.
Plausible mechanisms that account for these regioselec-
tively tunable cycloadditions are depicted in Scheme 2.
Under conditions A (cycle 1), the ScACHTNUTRGNEUNG(OTf)3/1,10-phenanthro-
line complex coordinates selectively to the carbonyl group
giving intermediate A, which makes the cyclopropane reac-
tive enough to regioselectively react with the nitrone and
form enolate intermediate B.[6] Subsequent ring closure
gives the formal [3+3] cycloadduct and regenerates the cata-
lyst. The major diastereomer may be formed from the more
favorable chair-like conformation of intermediate B. In con-
trast, the cationic gold(I) species preferentially binds to the
6148
ꢁ 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2010, 16, 6146 – 6150
Zhang, Yanqing
