by the OTBS group of the Michael acceptor to explain the
formation of 4′a (Figure 3).
Table 2. Radical Cyclization
Figure 3. Proposed approach for the formation of 4′a.
solvent
time (h)
yielda (%)
1
2
3
4
5
MeCN
EtOH
MeOH
AcOH
3
To address this problem, we reasoned that increasing the
polarity of the solvent should favor the “uncomplexed”
approach of the nucleophile on the less hindered face of the
Michael acceptor. Indeed, it appears that the proportion of
the desired isomer 4a increases with the polarity of the
solvent (Table 1). The best selectivity was obtained using a
3.5
3.5
1.5
1.5
10
20-41
50
10:1 CF3CH2OH/AcOH
a Yields are calculated from 4a only.
2), as the yield increased with the solvent acidity.12,13 How-
ever, the results obtained in acetic acid were not reproducible,
which is probably due to the presence of acid-sensitive func-
tionalities such as the acetonide. A switch to 2,2,2-trifluo-
roethanol, displaying an intermediate pKa value between
those of methanol and acetic acid, allowed us to isolate the
tricyclic compound 3 in a reasonable and reproducible 50%
yield. Yet, it is necessary to use acetic acid as cosolvent in
order to dissolve the copper acetate.
Table 1. Michael Addition
Noteworthy is the fact that isomer 4′a seems to lead only
to degradation products, whereas 4a undergoes cyclization,
affording 3 as a single diastereomer. The configuration of
the carbon bearing the ethoxycarbonyl group could not be
determined by NOE experiments. We anticipated that a
cis,trans ring junction was more likely than a trans,trans
ring junction for this 5,6,5 ring system.
In order to resolve this stereochemical ambiguity, we
pursued the synthesis by a Luche reduction. We hoped that
cerium(III) chloride would be complexed by the oxygens of
the diketone 3, thus shielding its â-face and leading to a
diastereoselective reduction from the R-face. This would
allow us to generate the last stereocenter of the A-ring of
hexacyclinic acid with the desired stereochemistry. The
reaction afforded a mixture of hemiketal 13 and alcohol 14
(Scheme 4).14 It seems that cerium(III) chloride induces the
formation of hemiketal 13, which is then reduced to 14. The
hemiketal acts as a protecting group for the ketone located
on the C-ring, rendering the reaction not only diastereose-
lective15 but also regioselective.
solvent
Et2O
1:4 THF/toluene
4:1 THF/toluene
4:1 DMF/THF
4:1 DMF/THF
yield (%)
4/4′
1
2
3
4
5
60
50
46
43
60
1:7
1:2.5
2.5:1
7:1
7:1a
a Reaction run with 3 equiv of enol ether 5 and 3.3 equiv of n-BuLi.
mixture of DMF and THF (4:1), leading to a mixture of 4a
and 4′a in a 7:1 ratio. Noteworthy is the importance of a
careful monitoring of the temperature to obtain reproducible
diastereomeric ratios. Besides, the yield could be increased
to 60% using 3 equiv of enolate instead of 2 equiv.11
With an optimized Michael addition step in our hands,
we then turned our attention to the radical cyclization, using
Mn(OAc)3 to generate the radical from the â-keto ester. This
step was carried out on the unseparated mixture of 4a and
4′a. Once again, the choice of solvent proved crucial (Table
We assumed that hemiketals 13 and 14 were stabilized
by the presence of a hydrogen bond between the hydroxyl
(12) (a) Snider, B. M. Chem. ReV. 1996, 96, 339-363. (b) Melikyan, G.
G. Org. React. 1997, 49, 427-675. For a recent example of Mn(III)-
promoted radical cyclization, see: (c) Tan, X.; Chen, C. Angew. Chem.,
Int. Ed. 2006, 45, 4345-4348.
(13) When the reaction was carried out in nucleophilic solvents such as
MeOH or EtOH, hemiketals of type 13 were formed, which partly explains
the observed low yields.
(11) An inseparable mixture of ketone 8 and Michael acceptor 6a was
obtained along with adducts 4a and 4′a, which prevented us from calculating
a yield based on recovered starting material.
(14) Full conversion to alcohol 14 has not been optimized.
(15) The stereochemistry of 14 was proven by NOESY experiments.
Org. Lett., Vol. 10, No. 1, 2008
47
Toueg, Julie
