Molecular basis of elansolid biosynthesis: Evidence for an unprecedented quinone methide initiated intramolecular diels-alder cycloaddition/ macrolactonization

Article abstract of DOI:10.1002/anie.201006880

Total control: The key steps in the biosynthesis of elansolide A1, a new and structurally unique polyketide metabolite from the gliding bacterium Chitinophaga sancti, have been elucidated from feeding experiments, by analysis of the biosynthetic gene cluster, and through the synthesis of model substrates. These steps include an unprecedented dehydration reaction, an intramolecular Diels-Alder cycloaddition (IMDA), and an unusual macrolactonization.

Full text of DOI:10.1002/anie.201006880

Communications
DOI: 10.1002/anie.201006880
Biosynthesis
Molecular Basis of Elansolid Biosynthesis: Evidence for an
Unprecedented Quinone Methide Initiated Intramolecular Diels–
Alder Cycloaddition/Macrolactonization**
Richard Dehn, Yohei Katsuyama, Arne Weber, Klaus Gerth, Rolf Jansen, Heinrich Steinmetz,
Gerhard Hꢀfle, Rolf Mꢁller,* and Andreas Kirschning*
Elansolids A1/A2 (1)[1] and B1–B3 (2–4) and
the structurally unusual and highly reactive
elansolid A3 (5)[2] are new metabolites from
the gliding bacterium Chitinophaga sancti
(formerly Flexibacter spec.; Scheme 1).
While elansolid A2 (1*) shows antibiotic
activity against Gram-positive bacteria in
the range of 0.2 to 64 mgmLꢀ1 and cytotox-
icity against L929 mouse fibroblast cells with
an IC50 value of 12 mgmLꢀ1, the atropisomer
elansolid A1 (1) is significantly less
active.[2,3] The elansolids feature a bicyclo-
[4.3.0]nonane core which in the case of
elansolids A1/A2 is part of a 19-membered
macrolactone. Elansolid B1 is the corre-
sponding seco acid of elansolids A1/A2,
while the elansolids B2 and B3 are workup
artifacts that result from nucleophilic addi-
tion of methanol and NH3, respectively, to
Scheme 1. Elansolids A1/A2 (1/1*), A3 (5), and B1–B3 (2–4) as well as results from
feeding studies with 13C-labeled precursors. SAM: S-adenosylmethionine.
the quinone methide moiety in elansolid A3 (Scheme 1).
[*] Dr. R. Dehn,[+] A. Weber, Prof. Dr. A. Kirschning
Institut fꢀr Organische Chemie und Biomolekulares Wirkstoffzen-
trum (BMWZ), Leibniz Universitꢁt Hannover
Schneiderberg 1B, 30167 Hannover (Germany)
Fax: (+49)511-762-3011
Noteworthy, the unique bicyclo[4.3.0]nonane core may arise
from an intramolecular Diels–Alder cycloaddition (IMDA).[4]
Herein we describe the identification of the elansolid
biosynthetic gene cluster for elansolid biosynthesis plus
details on the unique aspects of elansolid biosynthesis and
focus particularly on the proposed IMDA cycloaddition and
macrolactonization. Conceptually we approach this topic
from two directions: a) identification and detailed analysis of
the biosynthetic polyketide synthase machinery including
feeding studies and b) synthesis of model precursors and
synthetic studies on the IMDA cycloaddition.
E-mail: andreas.kirschning@oci.uni-hannover.de
Dr. Y. Katsuyama,[+] Prof. Dr. R. Mꢀller
Helmholtz-Institut fꢀr Pharmazeutische Forschung Saarland
Helmholtz Zentrum fꢀr Infektionsforschung
and Universitꢁt des Saarlandes
Pharmazeutische Biotechnologie, Campus Gebꢁude C23
Postfach 151150, 66041 Saarbrꢀcken (Germany)
Fax: (+49)681-302-70201
Initial feeding studies with Chitinophaga sancti employing
isotopically labeled precursors revealed that the elansolids
are polyketide-derived metabolites (Scheme 1) exhibiting a
chorismate-derived p-hydroxybenzoic acid starter unit (see
Table S3 and Figure S9 in the Supporting Information).
We next intended to gain molecular insight into elansolid
biosynthesis by identifying the biosynthetic gene locus in
Chitinophaga sancti. As this genus has not been described to
harbor polyketide synthase (PKS) gene clusters, a cosmid
gene library of our producer strain was probed for the
presence of PKSs eventually enabling sequencing of the
elansolid biosynthetic gene locus (Figure S1A, Table S1;
Genbank accession number: HQ680975). In the course of
our study it became apparent that Chitinophaga pinensis
DSM2588 available from public strain collections had been
E-mail: rom@helmholtz-hzi.de
Dr. K. Gerth, Dr. R. Jansen, H. Steinmetz, Prof. Dr. R. Mꢀller
Mikrobielle Wirkstoffe, Helmholtz Zentrum fꢀr Infektionsforschung
Inhoffenstrasse 7, 38124 Braunschweig (Germany)
Prof. Dr. G. Hꢂfle
Helmholtz Zentrum fꢀr Infektionsforschung
Braunschweig (Germany)
[+] These authors contributed equally to this work.
[**] This work was funded by the Deutsche Forschungsgemeinschaft
(grant Ki 397/16-1). R.D. thanks the Fonds der Chemischen
Industrie for a PhD scholarship and Y.K. is indebted to the Alexander
von Humboldt Foundation for a postdoctoral fellowship. We thank
Shwan Rachid for the generation of the cosmid library and Silke C.
Wenzel for crucial comments on this manuscript.
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201006880.
3882
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2011, 50, 3882 –3887

genome-sequenced and contained a highly similar set of genes
(84.6% identity in nucleotide sequence with the cluster from
C. sancti; Table S2). We next found this strain to produce
elansolids, albeit at much lower titer (data not shown). The
biosynthetic machinery identified turned out to represent a
trans-AT PKS including all functionalities for b branching
(C32) and methylation (Scheme 2).[5,6] Detailed analysis of
the gene locus allowed us to delineate a biosynthetic scheme
(Scheme 2 and Figure S1B). The identified megasynthetase
incorporates many features characteristic of trans-AT sys-
tems, for example, unusual domain organization and modules
split across two subunits.[5] The assembly line consists of six
AT-less PKS subunits (J,K,O,P,Q,R) interacting with two
trans-AT functions encoded by elaB and elaC.
Based on the deduced domain architecture of the PKS as
well as phylogenetic analysis of the KS domains for their
predicted substrate specificity (Scheme 2 and Figure S2),[7] we
propose the biosynthetic model outlined in Scheme 2. Feed-
ing experiments plus the presence of a putative chorismate
lyase (ElaI) strongly indicate that p-hydroxybenzoic acid
serves as the starter unit. This starter is activated and attached
to the ACP of the loading module of ElaJ and subsequently
further processed in 12 consecutive elongation steps
(although this hypothesis does not correlate with the pre-
dicted substrate specificity of the KS in module 1 as shown in
Figure S2). Within the 14 identified elongation modules, the
KS domains of modules 9 and 14 do not contain the highly
conserved HGTGT motif essential for decarboxylative elon-
gation (Figure S3),[5] and are therefore assumed to be inactive.
Except for the extra ACP domains in modules 3 and 8, which
lack the essential serine residue (Figure S7), and the ElaK-ER
domain, which shows some mutations in the conserved region
(Figure S8), all other domains of the elansolid assembly line
are proposed to be functional during polyketide assembly.
After the second elongation step, geminal methyl groups are
presumably introduced by the MT domain in module 2 and
the trans-acting MT ElaS. However, we cannot exclude the
possibility that both methylations are catalyzed by the
internal MT domain, making ElaS superfluous for elansolid
biosynthesis, and it is also possible that ElaS acts at a later
stage of the biosynthesis.
Scheme 2. Proposed biosynthetic pathway of elansolid A3 (5) deduced from the elansolid biosynthetic gene cluster. The timing of the IMDA
reaction cannot be deduced with certainty and thus dehydration centered at C23 plus subsequent IMDA reaction might occur at a different stage
ꢀ
(see Scheme 4). Modules 9 and 14 do not insert extender units and are marked with an asterisk. Additionally, TE-catalyzed lactonization (via C23
OH) might take place prior to the IMDA reaction resulting in the formation of the hypothetical intermediate 9 (see below). AL: AMP ligase, ACP:
acyl carrier protein, KS: ketosynthase, DH: dehydratase, KR: ketoreductase, MT: methyltransferase, AT: acyltransferase, ER: enoylreductase, TE:
thioesterase; inactive domains are labeled with 8.
Angew. Chem. Int. Ed. 2011, 50, 3882 –3887
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3883

Communications
The C-terminal ER domain in the downstream
module (module 3) is a typical example of unusual
domain arrangements in trans-AT PKS.[5] However,
ER sequence analysis indicates its similarity to non-
functional ER domains from the bryostatin pathway.[5]
This may imply that this domain is inactive and thus the
double-bond reduction is carried out by ElaB. In
analogy to other polyketide biosynthetic pathways[6]
a
set of five proteins (E,F,L,M,N) is encoded in the
elansolid biosynthetic gene cluster catalyzing the
b alkylation at C17. This modification is carried out
during chain assembly while the intermediate is bound
to module 5, which might explain the presence of
tandem ACP domains. Modules 6/7 and 8/9 represent
dehydrating bimodules often found in trans-AT PKS.[5]
The configurations of the double bonds generated
during elansolid biosynthesis as well as the configu-
rations at C7 and C9 correlate well with those
predicted[8,9] for the respective KR domains (Figure S5
and S6). Especially the predicted double-bond geo-
metries (C16–C19) as shown in Scheme 2 are crucial
for the subsequent IMDA reaction (see below and
Scheme 3). Interestingly, module 10 harbors an extra
DH domain which apparently does not catalyze the
dehydration of the respective b-hydroxy group. We
speculate that this domain instead catalyzes the
dehydration of the OH group at C23 (elansolid
numbering), resulting in the quinone methide moiety
(Scheme 2 and Scheme 4).
Scheme 4. Biosynthetic considerations for IMDA cycloaddition starting from
open-chain (6 and 7; cases 1a, 1b) and macrocyclic precursors (9 and 10;
cases 2a, 2b); for clarity, the configurations of the stereogenic centers are not
represented.
Notably, the SAM-dependent formation of the
geminal methyl groups at C22, most likely on the
nascent b-keto intermediate, excludes a subsequent
and subsequently hydrolyzed by the TE domain at the final
module to yield deoxyelansolid A3. As the final step of
elansolid A3 biosynthesis, ElaG, a cytochrome (CYT) P450
monooxygenase, is supposed to catalyze the oxidation at C20.
An alternative scenario would be the further extension of the
linear chain without an IMDA reaction; this would result in
product 9 which would be eventually formed by TE-catalyzed
lactonization. As this macrolactone would also be prone to
IMDA cycloaddition (Scheme 1), both routes implicate the
presence of highly reactive quinone methide intermediates in
several consecutive biotransformations.
In brief, the analysis of the PKS provides a logical scenario
for the biosynthesis of deoxyelansolid A3, which is thus the
presumed final product of the assembly line. The proposed
unusual dehydratase reaction setting up the IMDA reaction is
most intriguing and warrants further analysis. We cannot
exclude that this dehydration is carried out by another
dehydratase present in the cluster and the order of the
reactions may differ from that in Scheme 2 including the
Scheme 3. all-E IMDA precursor 11, its positional isomer 12, and the
endo transition states TS-I and TS-II.[12]
standard PKS dehydration reaction. The quinone methide
intermediate may subsequently undergo IMDA cycloaddition
to form the bicyclo[4.3.0]nonane core (see below). However,
it cannot be excluded that PKS processing continues with the
linear-chain polyketide (see Scheme 4). In the former case,
the quinone methide cycloadduct would be further extended
IMDA reaction which might occur on a “shorter” intermedi-
ate coupled to ElaO or ElaP.
The information obtained from the gene cluster analysis
prompted us to analyze the role of quinone methide moieties
and the unprecedented IMDA cycloaddition in greater detail
and relate them to macrolactonization. If the quinone
3884
www.angewandte.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2011, 50, 3882 –3887

methide moiety is not formed by dehydration during PKS
assembly and the IMDA cycloaddition takes place at a later
stage of polyketide assembly, open-chain seco acid 6 (allyl
benzyl alcohol) and—with respect to the gene analysis more
likely—allyl alcohol 7 can be envisaged as possible open-
chain precursors (cases 1a and 1b in Scheme 4). At this stage
the vinylogous quinone methide 8, the key intermediate of
this biosynthetic sequence, is generated which would directly
undergo IMDA cycloaddition to give tetrahydroindane
quinone methide 5 (elansolid A3). In a tandem fashion, this
highly reactive species would cyclize by a Michael-type attack
of the carboxylate onto the quinone methide moiety. As a
result elansolid A 1 is formed. If water, methanol, or ammonia
serve as nucleophiles, elansolids B1–B3 (2–4) are generated,
instead.
In principle it cannot be excluded that IMDA substrate 8
originates from either the 24-membered macolactone 9 or
alternatively the 26-membered macolactone 10 after lactone
activation (cases 2a and 2b; Scheme 1). Intermediate 9 is
clearly favored over 10 by the proposed biosynthetic outlined
above. This scenario would require a PKS-associated cyclizing
thioesterase (TE) instead of the hydrolyzing TE as shown in
Scheme 2. Bioinformatic analysis currently does not allow a
distinction between TEs that preferentially hydrolyze or
lactonize. Importantly, in both cases 1 and 2 presented in
Scheme 4 the Diels–Alder cycloaddition generates elanso-
lid A3 (5) which would then directly cyclize through a
Michael-type addition of the carboxylate onto the quinone
methide moiety and result in elansolid A (1).
Based on the feeding experiments and the principal
considerations mentioned above we consequently addressed
the question of the IMDA cycloaddition[10] by preparing the
synthetic model compound all-E triene 11 (resembling IMDA
precursors 6 and 10) and the simplified regioisomer 12
(resembling precursors 7 and 9). It must be noted that model
substrate 11 was designed first to be closely related to the
natural system in order to study all factors of diastereocontrol
during IMDA cycloaddition in more detail.[11] The E config-
uration of all olefinic double bonds would result in the correct
relative configuration at C16, C19, C23, and C24[12] via the two
possible endo transition states TS-I and TS-II. The stereo-
center at position 20 has to exert diastereofacial control.[13]
The synthesis of allyl alcohol 22 commenced with allyl
chloride 18[14] which was transformed into allyl alcohol 19 by a
standard sequence that included a-allylation of 17, reduction/
oxidation followed by Horner–Wadsworth–Emmons (HWE)
olefination with phosphonate 13, ester reduction, O-silylation
of the intermediate allyl alcohol, and finally PMB removal
(Scheme 5). The methylidene moiety in 19 was epoxidized
under Sharpless conditions with good enantiomeric excess
and the epoxy alcohol was then reductively ring-opened to
yield a 1,2-diol.[15] Formation of the cyclic PMB acetal and
cleavage under reductive conditions gave alcohol 20. Oxida-
tion furnished an aldehyde which was fused with phosphonate
16 by a HWE olefination, and the resulting alkyne 21 was
finally desilylated to afford allyl alcohol 22. Phosphonate 16[16]
was prepared from vinyl iodide 15 (from 14[17]) which was first
elaborated by a Sonogashira–Hagihara alkynylation followed
by Appel bromination and Michaelis–Arbusov reaction.
Scheme 5. Synthesis of allyl alcohol 22 and IMDA studies. Reagents
and conditions: a) 1. [Cp2ZrCl2], AlMe3, CH2Cl2, RT, 15 h, 2. I2, 50%;
b) Me3SiCCH, [Pd(PPh3)4], CuI, pyrrolidine, RT, 2 h, 80%; c) CBr4,
PPh3, CH2Cl2, RT, 30 min; d) P(OEt)3, microwave irradiation, 1008C,
30 min, 70% for two steps; e) 1. DIPA, nBuLi, THF, ꢀ788C to 08C,
then addition of 17, ꢀ788C!ꢀ408C, 2. addition of 18, TBAI, ꢀ408C
to RT, 82%; f) Dibal-H, THF, ꢀ788C!RT, 99%; g) PCC, CH2Cl2, RT;
h) NaH, (EtO)2P(O)CH2CO2Et (13), THF, 508C, then addition of
aldehyde (from g), 808C, 91% for two steps; i) 1. Dibal-H, THF,
ꢀ788C!RT, 2. TBSCl, imidazole, DMAP, CH2Cl2, RT, 99% for two
steps; j) DDQ, CH2Cl2, pH 7 phosphate buffer, RT, 93%; k) Ti(OiPr)4,
d-(ꢀ)-DET, tBuOOH, CH2Cl2, ꢀ258C, 93%, 95% ee (determined by
1H NMR spectroscopy after formation of the S Mosher ester[15]);
l) LiAlH4, THF, 08C to RT, 86%; m) 1. 4-MeO-C6H4-CH(OMe)2, PPTS,
CH2Cl2, RT, 2. Dibal-H, toluene, ꢀ788C!RT, 89% for two steps;
n) DMP, NaHCO3, CH2Cl2, 08C, 86%; o) NaHMDS, 16, THF, ꢀ788C,
then addition of aldehyde, ꢀ788C!RT, 84%, all-E/other iso-
mers=10:1; p) TBAF, THF, 08C to RT, 99%; q) 1. TPAP, NMO,
ꢀ
CH2Cl2, ꢀ308C; 2. RC6H4 MgBr, THF, TMEDA, ꢀ788C to ꢀ508C
(R=OMe, 75%; R=TMS, 81% for two steps); r) 24a,b, TPAP, NMO,
MS 4 ꢃ, CH2Cl2, ꢀ308C (67%; de=5:1). Compounds 25 were not
obtained. Abbreviations: Cp=cyclopentadienyl, DIPA=ethyldiisopro-
pylamine, TBAI=tetra-n-butylammonium iodide, Dibal-H=diisobutyl-
aluminum hydride, PCC=pyridinium chlorochromate, TBS=tert-butyl-
dimethylsilyl, DMAP=4-dimethylaminopyridine, DDQ=dichlorodicya-
noquinone, DET=diethyl tartrate, PPTS=pyridinium p-toluenesulfo-
nate, DMP=Dess–Martin periodinane, NaHMDS=sodium hexame-
thyldisilazide, TBAF=tetra-n-butylammonium fluoride, TPAP=tetra-n-
propylammonium perruthenate, NMO=N-methylmorpholine N-oxide.
Using the Ley–Griffith method[18] we next oxidized allyl
alcohol 22 at ꢀ308C to the corresponding aldehyde which was
treated directly with 4-methoxyphenylmagnesium bromide or
alternatively with 4-trimethylsilylphenylmagnesium bro-
mide[19] to afford allyl benzyl alcohols 23a,b and 24a,b,
respectively, as mixture of diastereomers (ca. 1:1 for both
examples; Scheme 5). These alcohols served as model qui-
none methide precursors (see allyl benzyl alcohol 6,
Scheme 4). In no case were Brønstedt or Lewis acids able to
Angew. Chem. Int. Ed. 2011, 50, 3882 –3887
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3885

Communications
force the formation of the desired IMDA products.[20]
Also all attempts to provide the allyl benzyl alcohols
23a,b or 24a,b with a better leaving group (acetate,
tosylate) failed. Only when trienes 24 were oxidized to
the corresponding enones did spontaneous formation
of IMDA products 26 occur with good diastereoselec-
tivity via the preferred TS-I.[21] The corresponding 4-
methoxyphenyl substituted alcohols 23 did not furnish
the desired IMDA products 25; complete decomposi-
tion was observed.
The configuration of the Diels–Alder products 26
was unambiguously determined by analysis of the H,H
coupling constants (J) and by conducting NOE experi-
ments; the data were compared with the corresponding
data collected for authentic elansolids.[1,23] The abso-
lute configurations at C16, C19, C23, and C24[13] were
identical to those in the elansolids.
Next, we prepared the model substrate 12 with
reversed positioning of the hydroxy group and the
double bond at C23–C25 (elansolid numbering). Acti-
vation should also lead to the key quinone methide
intermediate (resembling 10, Scheme 4). Thus, alde-
hyde 28[23] was condensed with p-methoxyacetophe-
none to yield enone 29 as a single E isomer (Scheme 6).
Luche reduction followed by acylation furnished allyl
acetate 30. Forced by the electron-donating properties
of the para methoxy group, 30 spontaneously under-
went a [3,3]-sigmatropic rearrangement, which was
promoted under the slightly acidic purification con-
ditions during silica gel column chromatography, to
furnish the styrene derivative 31. After deprotection
and oxidation, the resulting aldehyde was coupled with
phosphonate 16. The resulting triene 32 was treated
Scheme 6. Synthesis of triene rac-32 through a [3,3]-sigmatropic rearrangement
of allyl acetate rac-31 and Brønstedt acid promoted IMDA cycloaddition to
provide the bicyclo[4.3.0]nonane core rac-33. Reagents: a) p-methoxyacetophe-
none, NaOMe, MeOH, D, 59%; b) NaBH4, CeCl3, MeOH, 08C, 99%; c) AcCl,
NEt3, CH2Cl2, RT, 50%; d) TBAF, THF, RT, 64%; e) DMP, CH2Cl2, RT, 30%;
f) NaHMDS, 16, THF, ꢀ788C, 62% (all-E); g) TFA, wet dioxane, RT, 15 h
(55%). TFA=trifluoroacetic acid.
with trifluoroacetic acid in wet dioxane to smoothly provide
the IMDA product rac-33 as a single diastereomer. The polar
aprotic solvent was chosen based on the assumption that it
would stabilize the quinone methide cation 34. As a result of
the IMDA reaction the second quinone methide cation 35 is
formed which is trapped by water originating from the wet
carboxylate onto a quinone methide species when elanso-
lid A3 (5) was kept at room temperature in DMSO
(Scheme 7).[2] After one day elansolids A1 (1) and A2 (1*)
were detected in a 2:3 ratio and after another 13 days this
ratio had changed to 3:2.
1
solvent.[24] Analysis of the relevant H NMR coupling con-
stants (J) revealed that the cycloaddition proceeded through
an endo transition state yielding the correct relative config-
uration as present in elansolid A (1).[22] Interestingly, nucle-
ophilic attack of water onto intermediate 35 proceeds syn
relative to the alkynyl substituent at C16 resulting the same
relative configuration at C25 as that found in the elansolids.
This is indicated by the small vicinal coupling between H24
and H25 which is also observed in authentic samples of the
elansolids.[1,22]
Scheme 7. Direct conversion of elansolid A3 to elansolids A1/A2.
The synthetic model studies along with the gene cluster
analysis strongly underline our hypothesis, that either an allyl
alcohol similar to 7 (see Scheme 4) or, from our point of view
less likely, the corresponding lactone 9 serve as biosynthetic
substrates for the IMDA cycloaddition, whereas 6 or the 26-
membered lactone 10 can almost certainly be excluded as
precursors.[25] In fact, the conditions found for activation of 32
closely resemble the biological situation.
In conclusion, we have conducted a detailed study on the
PKS-based biosynthesis of the elansolids and have shed light
on unique and unprecedented aspects, namely the setup of the
IMDA reaction by an unusual dehydration mechanism, the
IMDA reaction itself, and the macrolactonization. In order to
achieve such
a detailed insight we utilized different
approaches and strategies such as feedings studies, analysis
of the responsible biosynthetic gene cluster, and the chemical
syntehsis of complex model substrates.
Finally, we gained strong support for the proposed unique
lactonization based on the nucleophilic addition of the
3886
www.angewandte.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2011, 50, 3882 –3887

Chem. Soc. 1968, 90, 3732 – 3738; b) J. M. Seco, E. Qunioa, R.
Riguera, Chem. Rev. 2004, 104, 17 – 117.
[16] T. Motozaki, K. Sawamura, A. Suzuki, K. Yoshida, T. Ueki, A.
Ohara, R. Munakata, K. Takao, K. Tadano, Org. Lett. 2005, 7,
2261 – 2264.
Received: November 2, 2010
Revised: January 6, 2011
Published online: March 29, 2011
Keywords: biosynthesis · intramolecular Diels–Alder reaction ·
[17] Z. Tan, E. Negishi, Org. Lett. 2006, 8, 2783 – 2785.
[18] We had to choose an oxidation method that works well at
temperatures as low as ꢀ308C
.
macrocyclizations · polyketides · quinone methides
because the intermediate aldehyde
spontaneously
underwent
an
[1] H. Steinmetz, K. Gerth, R. Jansen, N. Schlꢀger, R. Dehn, S.
Reinecke, A. Kirschning, R. Mꢁller, Angew. Chem. 2011, 123,
553 – 557; Angew. Chem. Int. Ed. 2011, 50. 532 – 536.
IMDA reaction at 08C to yield
tetrahydroindanes 27a,b as diaste-
reomers (1:1) resulting from both
endo transition states TS-I and TS-II.
[2] R. Jansen, K. Gerth, H. Steinmetz, S. Reinecke, W. Kessler, A.
Kirschning, R. Mꢁller, unpublished results.
[19] The p-TMS substituent was chosen as an alternative, because it
can be transferred readily into a phenolic OH group: R. L. Funk,
K. P. C. Vollhardt, J. Am. Chem. Soc. 1980, 102, 5253 – 5261.
[20] Instead, only deprotection of the PMB ether was observed.
[21] This high propensity to cyclization can be explained by 1) a low-
energy LUMO of the dienophile, 2) a high-energy HOMO of the
diene on account of conjugation to the alkyne, and 3) a double
Thorpe–Ingold effect caused by the gem-dimethyl group and the
tertiary alcohol. However, the gene analysis shown in Scheme 2
almost excludes the possibility of an IMDA precursor bearing a
keto group at C25.
[3] K. Gerth, H. Steinmetz, G. Hꢂfle, EP 2 093 212A1, 2009.
[4] E. M. Stocking, R. M. Williams, Angew. Chem. 2003, 115, 3186 –
3223; Angew. Chem. Int. Ed. 2003, 42, 3078 – 3115.
[5] J. Piel, Nat. Prod. Rep. 2010, 27, 996 – 1047.
[6] C. T. Calderone, Nat. Prod. Rep. 2008, 25, 845 – 853.
[7] T. Nguyen, K. Ishida, H. Jenke-Kodama, E. Dittmann, C.
Gurgui, T. Hochmuth, S. Taudien, M. Platzer, C. Hertweck, J.
Piel, Nat. Biotechnol. 2008, 26, 225 – 233.
[8] S. Donadio, L. Katz, Gene 1992, 111, 51 – 60.
[9] N. S. Scrutton, A. Berry, R. N. Perham, Nature 1990, 343, 38 – 43.
[10] K.-i. Tadano, Eur. J. Org. Chem. 2009, 4381 – 4394.
[11] Additionally, the presence of the alkyne group paves the way for
a biomimetic total synthesis approach.
[22] See the Supporting Information for details on the analytical
assignments.
[23] T. Lampe, H. M. R. Hoffmann, Tetrahedron Lett. 1996, 37, 7695 –
7698.
[12] Throughout the text atom numbering refers to the numbering of
in the natural product elansolid.
[24] Noteworthy, when the weaker acid AcOH in dioxane was
employed instead, no IMDA products could be detected.
[25] While this manuscript was under review, Piel and co-workers
disclosed a biosynthetic proposal for the elansolids, suggesting a
rearranged allyl benzyl alcohol (similar to 6) as the direct
substrate for the IMDA reaction. From our perspective, the
electronic requirements for the IMDA cycloaddition are fulfilled
only when the quinone methide moiety is involved as the key
biosynthesis intermediate: R. Teta, M. Gurgui, E. J. N. Helfrich,
S. Kꢁnne, A. Schneider, G. Van Echten-Deckert, A. Mangoni, J.
Piel, ChemBioChem 2010, 11, 2506 – 2512.
[13] To the best of our knowledge no examples of IMDA cyclo-
additions with a tertiary alcohol in allylic position to the diene
have been described so far. In simpler cases, the stereoinduction
of this center varies from moderate to excellent, see: a) M. P.
Edwards, S. V. Ley, S. G. Lister, P. D. Palmer, D. J. Williams, J.
Org. Chem. 1984, 49, 3503 – 3516; b) W. R. Roush, J. Am. Chem.
Soc. 1980, 102, 1390 – 1404.
[14] S. L. Boulet, L. A. Paquette, Synthesis 2002, 895 – 901.
[15] The enantiomeric excess was determined by the reacting epoxy
alcohol resulting from the SAE with (R)-Mosher chloride and
1
integrating the diastereotopic signals in the H NMR spectrum
of the corresponding ester: a) J. A. Dale, H. S. Mosher, J. Am.
Angew. Chem. Int. Ed. 2011, 50, 3882 –3887
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
3887