Diastereoselective SmI2 mediated cascade radical cyclisations of methylenecyclopropane derivatives - Syntheses of paeonilactone B and 6-epi-paeonilactone A

Article abstract of DOI:10.1039/b009513n

The SmI₂ mediated cascade cyclisations of several methylenecyclopropyl ketones have been examined and found to proceed with high diastereoselectivity, which is critically dependent on the presence of HMPA in the reaction. In one case the radica

Full text of DOI:10.1039/b009513n

View Article Online / Journal Homepage / Table of Contents for this issue
Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
Diastereoselective SmI2 mediated cascade radical cyclisations of
methylenecyclopropane derivatives—syntheses of paeonilactone B
and 6-epi-paeonilactone A
1
Raymond J. Boffey,a William G. Whittinghamb and Jeremy D. Kilburn*a
a Department of Chemistry, University of Southampton, Southampton, UK SO17 1BJ
b Syngenta, Jealott’s Hill International Research Centre, Bracknell, Berkshire, UK RG42 6EY
Received (in Cambridge, UK) 27th November 2000, Accepted 17th January 2001
First published as an Advance Article on the web 16th February 2001
The SmI2 mediated cascade cyclisations of several methylenecyclopropyl ketones have been examined and found
to proceed with high diastereoselectivity, which is critically dependent on the presence of HMPA in the reaction. In
one case the radical species at the end of the cascade sequence underwent an unexpected and highly stereoselective
dimerisation. The cascade methodology has been applied to a short synthesis of ( )-paeonilactone B and of
( )-6-epi-paeonilactone A.
Cascade radical cyclisation reactions have proved to be very
popular as a synthetic strategy as they allow the construction
of several C–C bonds in one step and can provide elegant syn-
thetic routes to complex polycyclic compounds and natural
products.1 Cascade reactions, initiated in particular by the
versatile lanthanide reagent SmI2,2 have also been a focus of
recent attention. The versatility of SmI2 is emphasised by the
fact that it can be used to generate both carbon centred radicals
and carbanions (by a further one-electron reduction of the rad-
ical species) and several cascade sequences utilising combin-
ations of both of these aspects have been described.3 Cascade
processes, initiated by SmI2, and using exclusively radical
intermediates4 are, however, complicated—and potentially
limited—by the fact that each radical intermediate can undergo
competitive reduction to the corresponding organosamarium
species, which may then effectively terminate the intended
cascade sequence. The rate constant for reduction of a primary
alkyl radical to give an organosamarium species in THF has
been estimated as 5 × 105–7 × 106 MϪ1 sϪ1 (depending on
HMPA concentration).5 Thus with SmI2 solutions of 0.1 M, a
radical cyclisation will need to have a rate constant ≥5 × 104 sϪ1
to compete effectively with the alternative reduction pathway.
In previous work we have developed radical cascade
sequences involving methylenecyclopropane derivatives, which
have provided novel routes to a range of bicyclic6 and spirocy-
clic compounds.7 The key step in these cascade sequences has
been the 5-exo cyclisation of a methylenecyclopropylpropyl
radical 1, followed by rapid ‘endo’ opening of the resulting
cyclisations and particularly to investigate whether it could be
used to mediate radical cascade sequences efficiently. In this
paper we wish to report full details of this study which has led
to short syntheses of paeonilactone B and 6-epi-paeonilactone
A, as well as uncovering an unusual stereoselective dimerisation
at the end of a radical cascade sequence.8
The paeonilactones A (4), B (5) and C (6) are all constituents
of the paeony root (the root of Paeonia albiflora Pallas) which
has been used extensively in Chinese and Japanese medicines
for the treatment of pain.9 Apart from their analgesic proper-
ties, the high density of stereocentres and oxygen functionality
around the cyclohexane nucleus has made the paeonilactone
family challenging synthetic targets.10
A retrosynthetic analysis of paeonilactone B (Scheme 2)
cyclopropylmethyl radical
2 to give methylenecyclohexyl
radicals 3 which can then be used in further bond-forming steps
(Scheme 1).
Scheme 1
In order to extend the scope of this chemistry we chose to
investigate the 5-exo cyclisation of methylenecyclopropyl ketyl
radicals, and in particular sought to use this chemistry for the
synthesis of the paeonilactone family of natural products. In so
doing we wished to explore the potential for using SmI2 in such
Scheme 2
suggested that the cis-fused bicyclic methylenecyclohexane 7
could be prepared by a 5-exo cyclisation of methylene-
cyclohexyl radical 8 onto a pendant alkyne, and 8 could, in
DOI: 10.1039/b009513n
J. Chem. Soc., Perkin Trans. 1, 2001, 487–496
This journal is © The Royal Society of Chemistry 2001
487

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
View Article Online
turn, arise from cyclisation of ketyl radical 10 onto a methyl-
enecyclopropane unit with subsequent ‘endo’ ring opening of 9.
Whether such a sequence would prove to be diastereoselective
and provide the correct relative stereochemistry of the tertiary
alcohol required for the natural product was one of the key
aspects of the proposed investigation.
The synthesis of paeonilactone B began with addition of
lithiated methylenecyclopropane11 to aldehyde 12, to produce
the desired alcohols as a readily separable mixture of diastereo-
isomers 13 and 14 (Scheme 3). The relative stereochemistry for
known preference for the cyclisation of cycloalkyl radicals
onto tethered alkenes and alkynes to give cis-fused bicyclic
systems.14
In order to rationalise the observed diastereoselectivity we
repeated the cyclisations under identical conditions, but
replacing HMPA with the less effective chelator DMPU.15
These cyclisation reactions gave the bicyclic products with
reduced overall yields and required a larger excess of SmI2 (~6
equiv.) for consumption of starting material (Scheme 4).
Notably, for the cyclisation of 16, the diastereoselectivity
was reduced (ratio 18:19, 1.5:1), whereas for the cyclisation of
17 the diastereoselectivity was seemingly unaffected (ratio
19:18, >30:1). In the absence of either DMPU or HMPA the
cyclisation was, as expected, a poor reaction. Thus 16 gave
an overall yield of ~20% of 18 and 19, but with a reversal of
stereoselectivity (ratio 18:19, 1:1.3), while cyclisation of 17,
under these conditions, yielded none of the desired bicyclic
compounds.
The selectivity observed for the cyclisation of 16 in favour of
18, in the presence of HMPA, in which the tertiary alcohol and
ether oxygen are cis in the bicyclic product, might be the con-
sequence of chelation control from the weakly basic propar-
gylic ether oxygen to the samarium() bound to the ketyl
radical. However, the decrease in selectivity for the cyclisation
of 16 as HMPA is replaced by the weaker chelator DMPU, and
reversal of selectivity when neither is present (leaving the even
less effective chelator, THF, as the available samarium ligand),
effectively rules out this possibility. It seems probable that the
first step of the cyclisation of 16, which effectively sets the
relative stereochemistry of the product, proceeds through a
chair-like transition state, allowing the propargyl ether sub-
stituent to adopt a pseudo-equatorial position (Scheme 5).
Scheme 3 Reagents and conditions: (i) BuLi, THF, Ϫ78 ЊC; (ii) 12;
᎐
(iii) NaH, DMPU, THF; (iv) HC᎐CCH2Br; (v) TsOH, acetone, H2O.
᎐
the two diastereoisomers was established by X-ray crystallo-
graphic structure analysis of the p-nitrobenzoate ester 15
derived from alcohol 14.12 Alkylation of the alcohols gave the
correspondingpropargylethers,‡andsubsequentketaldeprotec-
tion provided the two diastereomeric cyclisation precursors, 16
and 17 respectively, in essentially quantitative yield.
Treatment of ketone 16 with SmI2, under standard condi-
tions13 (slow addition of 16 to 2.2 equiv. SmI2, ButOH,
HMPA, THF, 0 ЊC) gave the desired bicyclic products as a
readily separable mixture of diastereoisomers, 18 and 19, in
57% and 6% isolated yields respectively (ratio 18:19, 10:1 by
analysis of the 1H NMR of the crude reaction mixture)
(Scheme 4). In contrast, treatment of diastereoisomeric ketone
Scheme 5
As a consequence of the bond angles of the methylenecyclo-
propyl group, the alkene appears to be essentially staggered
between the ketyl radical oxygen and the ketyl methyl group.
Thus the preference for conformer 20 over 21 may largely result
from the preference for the bulky OSm()(HMPA)n moiety to
also adopt a pseudo-equatorial position and avoid a 1,3 diaxial
interaction with HA. Replacement of HMPA with DMPU may
effectively reduce the steric bulk of the OSm()Ln moiety,15
leading to a lower selectivity for conformer 20. In the absence
of either HMPA or DMPU the ketyl methyl becomes sterically
dominant, leading to a reversal in selectivity.
In contrast, the first step of the cyclisation of 17 may well
proceed through a boat-like transition state, since a chair-like
transition state would force the propargyl ether substituent into
a severely hindered axial orientation. In the boat-like transition
state the alkene now appears to be largely eclipsed with either
the ketyl methyl group (22) or the ketyl radical oxygen (23).
Conformer 22 may now be preferred over 23 since it alleviates
the electronic repulsion between the ketyl oxygen functionality
and the alkene π-system,16 and this preference is unaffected by
replacing HMPA with DMPU.
Scheme 4
17 with SmI2, under identical conditions, gave the bicyclic
product 19 in 79% isolated yield, and only a trace of the
diastereoisomer 18 (ratio 19:18, >30:1).
The stereochemistry of 18 was ultimately confirmed by the
successful conversion to the natural product paeonilactone B
(vide infra), but the stereochemistry of both 18 and 19 was at
this stage assigned on the basis of NOE studies and on the
Completion of the synthesis of paeonilactone B 5 firstly
required protection of the tertiary allylic alcohol of 18 as the
‡ Propargyl = prop-2-ynyl.
488
J. Chem. Soc., Perkin Trans. 1, 2001, 487–496

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
View Article Online
triethylsilyl ether 24,17 followed by oxidation of the allyl ether to
the desired α-methylene lactone 25 using CrO3 and pyridine
(Scheme 6).18 The selective oxidation of the ostensibly more
hexane) then the cis-fused 7-syn product 31 will be the preferred
product whereas if the cyclisation proceeds with the butenyl
sidechain in an axial orientation then the cis-fused 7-anti
product 32 will be preferred (Scheme 7).
Scheme 7
The diastereomeric allyl ethers 33 and 34 were prepared
in analogous fashion to the corresponding propargyl ethers,
by alkylation of the separated alcohols 13 and 14 with allyl
bromide, and subsequent ketal deprotection (Scheme 8).
Scheme 6 Reagents and conditions: (i) Et3SiOTf, Et3N, CH2Cl2, 0 ЊC;
(ii) CrO3, pyridine, CH2Cl2, rt; (iii) O3, EtOH, Ϫ110 ЊC; (iv) Me2S;
(v) OsO4, THF, H2O or cat. K2OsO4ؒ2H2O, K2CO3, K3Fe(CN)6,
ButOH, H2O; (vi) PhSH, Et3N, CH2Cl2; (vii) O3, MeOH, Ϫ78 ЊC;
(viii) Me2S; (ix) CCl4, reflux; (x) HFؒpyridine, THF.
Scheme 8 Reagents and conditions: (i) NaH, DMPU, THF; (ii) allyl
bromide; (iii) TsOH, acetone, H2O.
Treatment of allyl ether 34 with SmI2, as before in the pres-
ence of HMPA, gave a 1:1 mixture of diastereoisomeric
bicycles, 35 and 36 in 35% isolated yield23 (Scheme 9). A slightly
electrophilic cyclohexyl alkene of 25 proved to be impossible,
with both alkenes reacting rapidly with ozone at Ϫ110 ЊC in
EtOH to give 26 in almost quantitative yield. Even more frus-
tratingly, treatment of 25 with OsO4 led to 27 with dihydroxyl-
ation of just the α-methylene lactone, presumably due to steric
congestion around the cyclohexyl alkene. Instead, base medi-
ated Michael addition of thiophenol to 25 gave the thioether
28, which was then successfully ozonolysed to give the desired
ketone, with concomitant oxidation of the thioether to the cor-
responding sulfoxide 29. Thermal elimination of phenylsulfenic
acid19 then reinstalled the α-methylene lactone and deprotec-
tion of the resulting silyl ether 30 was successfully achieved
using pyridineؒHF20 to give ( )-paeonilactone B 5, whose
structure was confirmed by comparison of NMR and IR
spectroscopic data to those reported previously for the natural
paeonilactone.9
The SmI2 mediated cascade reaction of methylenecyclo-
propyl ketone 16 thus provides a short route to paeonilactone
B, and proceeds with high diastereoselectivity which is critically
dependent on the presence of HMPA. Clearly a synthesis of
paeonilactone A 4 could be achieved by a stereoselective reduc-
tion of the exocyclic alkene of paeonilactone B, but having
established that the SmI2 mediated cascade reaction of propar-
gyl ether 16 proceeds with high diastereoselectivity to give the
bicyclic ether 18 it was of interest to see if the cascade method-
ology could be extended to the cyclisation of the corresponding
allyl ethers which might, in principle, install the methyl sub-
stituent for paeonilactone A directly. The stereochemical out-
come of the cyclisation of cyclohexyl radicals onto a pendant
butenyl (or allyloxy) residue has been investigated in some
detail. Studies by RajanBabu21 and Beckwith and Page22 indi-
cate that if the cyclisation proceeds with the butenyl sidechain
in an equatorial orientation (relative to the pseudo-chair cyclo-
Scheme 9 Reagents and conditions: (i) SmI2, ButOH, HMPA, THF,
0 ЊC.
improved yield (40%) of the same 1:1 mixture was obtained by
carrying out the reaction at Ϫ78 ЊC. No other bicyclic products
could be isolated from these reactions (nor could they be
1
detected in the H or 13C NMR spectra of the crude reaction
mixture).
The stereochemical outcome of this cyclisation is readily
rationalised in terms of the model presented above (Scheme 5)
for the cyclisation of the analogous propargyl ether 17. Thus
the cyclisation of the initially formed ketyl radical may proceed
via a boat-like transition state to give methylenecyclohexyl
radical 37, as essentially a single diastereoisomer, which then
cyclises to give a 1:1 mixture of alkyl radicals 38 and 39, which
in turn are reduced to the organosamarium and quenched by
J. Chem. Soc., Perkin Trans. 1, 2001, 487–496
489

Evaluation Only. Created with Aspose.PDF. Copyright 2002-2021 Aspose Pty Ltd.
View Article Online
ButOH. Replacement of HMPA with DMPU in the reaction
(0 ЊC) led to a lower yield (20%) of bicyclic products, but with
no change to the diastereoselectivity, again mirroring the results
observed with the corresponding propargyl ether 17. Thus it
seems that cyclisation of cyclohexyl radical 37, under the condi-
tions used here, gives no stereoselectivity at the newly formed
chiral centre. The lower yields of the cascade process in com-
parison to the analogous propargyl ethers were also disappoint-
ing and indeed somewhat surprising since the rate constant for
cyclisation of the hex-5-enyl radical is greater than the corre-
sponding rate constant for cyclisation of the hex-5-ynyl rad-
ical24 and thus one might anticipate that cyclisation of 37
(or 42) would be more efficient than cyclisation of 8. Since no
identifiable byproducts were isolated from these cyclisations it is
not possible to speculate on the reasons for the relatively low
yields in the allyl ether cyclisations.
and 43, it is only 43 which dimerises. Even more remarkably, the
dimerisation occurs only between opposite enantiomers of 43,
to give the dimer as a meso isomer, and none of the ‘homo’
coupling of identical enantiomers is observed! Dimerisation
(Wurtz coupling) of organosamariums such as benzylsamarium
diiodide are well known2 and recently a non-stereoselective
dimerisation of a samarium-derived glucosyl radical has been
described.26 However the only example of dimerisation at the
end of a radical cyclisation sequence that we are aware of, was
reported by Molander and McKieet al.15 in which cyclisation of
44 led to a mixture of diastereomeric dimers 45 and 46, under
conditions of relatively low SmI2 concentration (Scheme 11).
Cyclisation of 33 in the presence of HMPA, on the other
hand, gave a single diastereomeric bicycle 40 in just 17%
isolated yield, but accompanied by a 25% yield of a dimeric
product 41, as a single diastereoisomer,25 with the same relative
stereochemistry for the bicyclic portion as for 40 (Scheme 10).
Scheme 11
The dimerisation was not observed when higher concentrations
of SmI2 (as in our work described here) were used.
Clearly radical intermediate 43 is more stable than any
of the other radical intermediates formed in the cyclisation
sequence, and we believe this can be explained most readily
by the fact that this intermediate has both the OSm() and
the alkyl radical on the endo face of the bicyclic structure,
allowing stabilisation of the radical by interaction with the
samarium(). This would effectively ‘protect’ the radical from
further reduction by SmI2, allowing build up of the radical
species and eventual dimerisation. The exclusive formation of
the meso dimer is harder to rationalise. The dimerisation may
involve formation of a diradical intermediate from two
monomers 43 (e.g. bridging of two ketyl oxygens with two
samariums) followed by radical coupling to give the dimeric
product. For the formation of such an intermediate, approach
of identical enantiomers to each other may be impeded relative
to the approach of opposite enantiomers. Alternatively, the
structure of the rac diradical intermediate, if formed, may not
readily allow coupling of the two alkyl radicals and may be
slowly quenched to give 40, or it may equilibrate with a meso
diradical intermediate whose structure does allow radical
coupling.
Although the cyclisations of the allyl ethers failed to provide
the correct stereochemistry for paeonilactone A, the conversion
of the bicyclic ethers 35, 36 and 40 to diastereoisomers of the
natural product was none-the-less investigated. The diastereo-
meric mixture of alcohols 35 and 36 was converted to the
corresponding mixture of triethylsilyl ethers 48 in 70% yield,
whereas the single diastereomeric alcohol 40 was converted to
triethylsilyl ether 49 in only 32% yield, reflecting the difficulty
of protecting an alcohol on the endo face of a cis-fused bicyclic
system (Scheme 12).
Scheme 10 Reagents and conditions: (i) SmI2, ButOH, HMPA, THF,
0 ЊC.
Again, no other bicyclic or dimeric products could be
isolated from the reaction (nor could they be detected in the
1H or 13C NMR spectra of the crude reaction mixture) and
again the stereochemical outcome of this cyclisation is readily
rationalised in terms of the model presented previously for
the cyclisation of the analogous propargyl ether 16. Thus the
cyclisation of the initially formed ketyl radical may proceed via
a chair-like transition state to give methylenecyclohexyl radical
42, as essentially a single diastereoisomer. Radical 42 then
cyclises to give exclusively alkyl radical 43 which is either
reduced to the organosamarium and quenched to give 40, or
dimerises to give 41. The sequence therefore appears to be
highly stereoselective, although it does not give the desired
stereochemistry for the natural product paeonilactone A.
As before with the analogous propargyl ether 16, replace-
ment of HMPA with DMPU in the cyclisation led to a loss of
stereoselectivity in the first steps of the cascade, and thus to the
formation of both methylenecyclohexyl radicals 37 and 42,
which in turn gave a mixture of all three bicyclic products 35, 36
and 40, but no dimeric products were detected, indicating that
the presence of HMPA is essential for the formation of 41.
The formation of dimer 41 was quite unexpected since it is
formed under conditions of relatively high SmI2 concentration
where rapid reduction of the primary radical 43 and quenching
would be expected, as observed for intermediates 38 and 39. In
a further experiment we treated a 1:1 mixture of starting allyl
ethers 33 and 34 with SmI2, under the same conditions as
before, with HMPA, and obtained a mixture of all three bicyclic
products 35, 36 and 40 and the single dimeric product 41. Thus,
under conditions that generate all three primary radicals 38, 39
Ozonolysis of triethylsilyl ethers 48 did not produce the
desired cyclohexanone, but gave instead the dihydrobenzofuran
50 in 45% yield (Scheme 12). Similarly ozonolysis of triethylsilyl
ether 49 gave dihydrobenzofuran 50 in 38% yield accompanied
by a small amount of the desired cyclohexanone 51. The
dihydrobenzofuran 50 has previously been isolated as the major
product from the decomposition of the paeony root metabolite
52, when the latter was stored as a solution in CHCl3–MeOH
490
J. Chem. Soc., Perkin Trans. 1, 2001, 487–496