Cylindrocyclophane biosynthesis involves functionalization of an unactivated carbon center

Article abstract of DOI:10.1021/ja308318p

The cylindrocyclophanes are a family of natural products that share a remarkable paracyclophane carbon scaffold. Using genome sequencing and bioinformatic analyses, we have discovered a biosynthetic gene cluster involved in the assembly of cylindrocyclophane F. Through a combination of in vitro enzyme characterization and feeding studies, we confirm the connection between this gene cluster and cylindrocyclophane production, elucidate the chemical events involved in initiating and terminating an unusual type I polyketide synthase assembly line, and discover that macrocycle assembly involves functionalization of an unactivated carbon center.

Full text of DOI:10.1021/ja308318p

Communication
pubs.acs.org/JACS
Cylindrocyclophane Biosynthesis Involves Functionalization of an
Unactivated Carbon Center
Hitomi Nakamura, Hilary A. Hamer, Gopal Sirasani, and Emily P. Balskus*
Department of Chemistry & Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, United States
S
* Supporting Information
ABSTRACT: The cylindrocyclophanes are a family of
natural products that share a remarkable paracyclophane
carbon scaffold. Using genome sequencing and bioinfor-
matic analyses, we have discovered a biosynthetic gene
cluster involved in the assembly of cylindrocyclophane F.
Through a combination of in vitro enzyme characterization
and feeding studies, we confirm the connection between
this gene cluster and cylindrocyclophane production,
elucidate the chemical events involved in initiating and
terminating an unusual type I polyketide synthase
assembly line, and discover that macrocycle assembly
involves functionalization of an unactivated carbon center.
iving organisms such as bacteria and plants produce a wide
Lvariety of secondary metabolites, many of which have
complex chemical structures. Natural products with unusual
and synthetic chemists. Studying the biosynthesis of remarkable
chemical functionality often reveals unprecedented enzymatic
Figure 1. The cylindrocyclophane family of natural products and
previous biosynthetic studies. (A) Structures of cylindrocyclophanes
isolated from C. licheniforme strains ATCC 29204 and 29412. (B)
Feeding studies performed by Bobzin and Moore revealed the
molecular architecture are of interest to both chemical biologists
reactivity that may be further applied in combinatorial biosyn-
polyketide origin of the cylindrocyclophanes. (C) Biosynthetic
thesis and metabolic engineering.1 Unique natural product
structures also inspire the development of new methodology and
strategies for chemical synthesis.2 The cylindrocyclophanes are a
family of cytotoxic natural products from photosynthetic
cyanobacteria that contain a highly unusual [7.7]paracyclophane
ring system.3 This scaffold is unique among natural products, and
the biochemical transformations involved in its assembly are
currently unknown. Here we report the identification of a gene
cluster from the cyanobacterium Cylindrospermum licheniforme
that is involved in cylindrocyclophane production. Using a
combination of in vitro enzyme characterization and feeding
studies, we have discovered that assembly of the paracyclophane
macrocycle employs unusual biosynthetic logic, including
functionalization of an unactivated carbon center.
hypothesis put forth by Bobzin and Moore.4
experiments did not provide insight into the exact structure of the
monomeric precursor, Bobzin and Moore proposed that the C7/
C20 position was pre-functionalized for eventual C−C bond
formation during monomer construction on a polyketide
synthase (PKS) assembly line.
The unusual molecular framework of the cylindrocyclophanes
has inspired several elegant total syntheses.6 Though synthetic
chemists have recognized and admired the complexity-
generating power of the proposed biosynthetic route to the
[7.7]paracyclophane skeleton, no chemical synthesis to date has
been able to exploit this disconnection. Understanding
cylindrocyclophane biosynthesis at the molecular level should
therefore not only reveal new enzymatic reactions and
biosynthetic logic but also inspire chemists seeking to emulate
the elegance of Nature’s synthetic approach.
Moore et al. reported the isolation and structural character-
ization of cylindrocyclophanes A−F in the early 1990s (Figure
1A).3a,b Subsequent feeding experiments suggested a unique
biosynthetic pathway for constructing the [7.7]paracyclophane
skeleton (Figure 1B). The symmetric labeling pattern resulting
from acetate incorporation revealed the polyketide origin of
these natural products and led Bobzin and Moore to hypothesize
that their biosynthesis involved a C−C bond-forming, head-to-
tail dimerization reaction of an alkyl resorcinol intermediate
(Figure 1C).4 This type of tailoring event, which constructs two
sp2−sp3 C−C linkages and two stereocenters, has limited
precedent among biological transformations.5 Though feeding
Our strategy for locating the cylindrocyclophane biosynthetic
gene cluster involved whole genome sequencing followed by
bioinformatic searches for specific gene products likely to be
involved in natural product assembly. In addition to confirming
the polyketide nature of the cylindrocyclophanes, the original
Received: August 21, 2012
Published: October 29, 2012
© 2012 American Chemical Society
18518
dx.doi.org/10.1021/ja308318p | J. Am. Chem. Soc. 2012, 134, 18518−18521

Journal of the American Chemical Society
Communication
Figure 2. Discovery of the cyl gene cluster and proposed biosynthesis of the cylindrocyclophanes. (A) HMG-CoA synthase homologs involved in β-Me
group installation on polyketide scaffolds. (B) The putative cyl gene cluster. Each arrow represents the direction of transcription of an open reading
frame. See SI for complete cluster annotation. (C) Biosynthetic hypothesis for the assembly of the cylindrocyclophanes.
feeding studies revealed an acetate-derived β-methyl group
within the backbone of cylindrocyclophanes A−F (Figure 1B).4
This relatively unusual structural feature is introduced into
polyketide scaffolds using a highly conserved pathway containing
an enzyme homologous to hydroxymethylglutaryl-CoA (HMG-
CoA) synthase (Figure 2A).7 Reasoning that a closely related
enzyme should be involved in cylindrocyclophane assembly, we
used degenerate PCR to amplify putative HMG-CoA synthase
homologs from genomic DNA of C. licheniforme ATCC 29412, a
producer of cylindrocyclophanes A−F.3b,8 A 636 bp PCR
product was obtained, and its translated sequence revealed close
homology (92 and 90% identity at the amino acid level) to JamH
and CurD, cyanobacterial HMG-CoA synthases involved in β-
Me installation during jamaicamide9 and curacin10 biosynthesis
(Figure S1). Because genes responsible for natural product
assembly in microbes are typically co-localized, we hypothesized
that the C. licheniforme HMG-CoA synthase fragment should be
clustered with additional genes involved in cylindrocyclophane
biosynthesis.
Searches for this nucleotide sequence in C. licheniforme
genome sequence data generated by 454 pyrosequencing
revealed an identical sequence on contig 1024. In addition to a
full-length HMG-CoA synthase homolog, this 12.1 kb genome
fragment contained additional open reading frames (ORFs)
hypothesized to encode enzymes possessing functions consistent
with involvement in cylindrocyclophane assembly, including a
PKS and other enzymes involved in β-Me installation. Searches
for the remaining genes associated with addition of this
substituent identified two more contigs connected to cylin-
drocyclophane biosynthesis (111 and 112).11 We confirmed the
proximities of all three contigs by sequencing regions of the
genome spanning adjacent contigs, permitting assembly of a
single gene cluster. The putative cylindrocyclophane (cyl)
biosynthetic gene cluster contains 12 ORFs (Figure 2B),
including genes encoding the enzymatic machinery required
for installing the β-Me substituent (cylE, F, G, H). The cluster
also has two distinct types of PKS machinery, a short type I
modular PKS assembly line (cylD, H) and a type III PKS (cylI).
Two additional genes, predicted fatty acid adenylating enzyme
cylA and freestanding acyl carrier protein (ACP) cylB, likely
produce enzymes involved in initiating biosynthesis. The
remaining ORFs include homologs of a phospholipid methyl-
transferase (cylJ), a predicted flavin-dependent oxidoreductase
(cylL), a protein with a hemolysin-like binding region (cylK), and
a hypothetical protein (cylC). Detailed annotation of the cluster
(Table S1) revealed all of the enzymatic reactivity necessary for
biosynthesis of a putative cylindrocyclophane monomeric
precursor (8) and enabled us to generate a detailed hypothesis
for its assembly (Figure 2C).
One of the most striking and unexpected features of the cyl
gene cluster is the abbreviated nature of the type I PKS assembly
line. CylD and CylH contain just two intact modules and are
therefore expected to perform only two elongation reactions.
Iterative function of these enzymes is unlikely as they both lack all
of the domains needed to form the fully reduced polyketide
backbone found in the final natural products. Given the size and
structures of the cylindrocyclophanes, we propose that biosyn-
thesis originates with recruitment of a free fatty acid. Activation
of decanoic acid by CylA and loading onto CylB to form thioester
7 could be followed by transfer to the first type I PKS CylD. This
PKS contains ketosynthase (KS) and acyltransferase (AT)
domains along with two tandem ACP domains, an arrangement
that is often associated with the β-Me group modification.7 We
hypothesize that CylD performs one elongation with malonyl-
CoA and the resulting β-ketothioester is further processed by
CylE, F, and G, which are predicted to catalyze addition of
acetate to the β-ketone and dehydration to form an α,β-
unsaturated thioester intermediate. We were unable to locate a
candidate freestanding ACP for use in β-Me installation but
noted that the second ACP domain on CylD is most closely
related in sequence to freestanding ACPs of this type, perhaps
indicating that it fills this role. The final two enzymatic activities
needed to complete construction of the chiral β-Me substituent
are encoded as domains on the second type I PKS CylH. CylH
contains enoyl-CoA hydratase (ECH) and enoylreductase (ER)
domains, which should catalyze decarboxylation and reduction of
the CylD product.7 The additional KS, AT, ACP, and
thioesterase (TE) domains of CylH constitute another PKS
module and should catalyze a second elongation with malonyl-
CoA. Interestingly, multiple sequence alignments of the CylH
TE domain with TE domains from characterized type I PKS
assembly lines indicate that it may lack a conserved catalytic
aspartate residue and could therefore be non-functional.12 We
propose that type III PKS CylI terminates the assembly line by
utilizing the CylH-bound β-ketothioester as a starter unit for
aromatic ring formation. Type III PKSs typically use CoA
18519
dx.doi.org/10.1021/ja308318p | J. Am. Chem. Soc. 2012, 134, 18518−18521

Journal of the American Chemical Society
Communication
thioesters as substrates and catalyze successive Claisen
condensations, cyclization of the resulting polyketide inter-
mediate, and aromatization. The action of CylI on the nascent
product of the CylD-CylH assembly line could generate the
resorcinol-type aromatic ring found in the cylindrocyclophanes.
Overall, the PKS machinery encoded by this biosynthetic gene
cluster is consistent with its role in cylindrocyclophane assembly.
We solidified the link between the cyl gene cluster and
cylindrocyclophane biosynthesis by characterizing the activities
of several of the encoded enzymes in vitro. Specifically, we
examined CylA and CylB, the enzymes predicted to be involved
in fatty acid recruitment, and CylI, the type III PKS. CylA, CylB,
and CylI were individually cloned from C. licheniforme,
overexpressed in E. coli, and purified. The ability of CylA to
activate and load decanoic acid onto holo-CylB was probed using
both HPLC and LC-MS (Figure 3A). Incubation of holo-CylB
amine (NAC) thioester analog (9) of the proposed CylI
substrate and evaluated its reactivity in the presence of CylI and
malonyl-CoA. A single new product peak was observed by HPLC
(Figure 4); isolation of this product from a large-scale enzymatic
Figure 4. In vitro characterization of CylI. HPLC time course for
formation of resorcinol 8 (230 nm).
reaction and characterization using NMR (1H, 13C) and HRMS
confirmed formation of resorcinol 8. The enzymatically
synthesized material was identical in all respects to a product
standard prepared via chemical synthesis. The reactivity of type
III PKS CylI supports its proposed role in cylindrocyclophane
biosynthesis, terminating a type I modular PKS assembly line in a
highly unusual fashion via aromatic ring formation. While a few
type III PKS enzymes and domains have been reported to accept
acyl-ACP substrates generated by iterative type I15 and type II16
fatty acid synthases, to our knowledge CylI is the first type III
PKS predicted to utilize an acyl-ACP substrate from a modular
assembly line.
The in vitro characterization of CylA, CylB, and CylI connects
the cyl gene cluster to assembly of an alkyl resorcinol
intermediate that could be a monomeric precursor to the
cylindrocyclophanes. To further confirm that the fatty acid
activation pathway encoded by this cluster is directly connected
to cylindrocyclophane production, we performed feeding studies
with C. licheniforme, evaluating incorporation of isotopically
labeled precursors into the cylindrocyclophane skeleton using
LC-MS of culture extracts. Feeding of d19-decanoic acid
produced both d18- and d36-cylindrocyclophane F (Figure 5).
Figure 3. In vitro characterization of CylA and CylB. (A) HPLC assay
showing the loading of decanoic acid onto holo-CylB by CylA. (B) LC-
MS competition assay reveals selectivity for C10 fatty acid activation by
CylA and CylB.
with decanoic acid, CylA, and ATP resulted in formation of the
anticipated decanoyl-CylB thioester. This activity was dependent
on the presence of both ATP and CylA. CylA was not able to load
decanoyl-CoA onto holo-CylB, which suggests that CylA is a
fatty acyl-AMP ligase rather than a fatty acyl-CoA ligase (Figure
S8). The size selectivity of this fatty acid activation and loading
process was evaluated using an LC-MS-based competition assay.
Incubating CylA and holo-CylB with equal amounts of C6, C8,
C10, C12, and C14 saturated fatty acids resulted in highly selective
formation of decanoyl-CylB thioester (Figure 3B). We observed
only a small amount (<10%) of product derived from octanoic
acid. Overall, these experiments demonstrate that CylA and CylB
are capable of recruiting decanoic acid for cylindrocyclophane
biosynthesis in a size-selective manner. This exquisite selectivity
is consistent with the observation that all members of the
cylindrocyclophane natural product family share the same core
carbon skeleton. Activation of a free fatty acid for use as a starter
unit is rare among type I modular PKS assembly lines but has
been demonstrated previously in the biosynthesis of mycobacte-
rial fatty acids13 and several hybrid PKS/non-ribosomal peptide
synthetase systems.9,14
Figure 5. Incorporation of labeled decanoic acid into cylindrocyclo-
phane F by C. licheniforme.
Having confirmed that the in vitro activities of enzymes
involved in initiating cylindrocyclophane biosynthesis supported
our biosynthetic hypothesis, we next explored the chemistry
underlying assembly line termination. Evaluating the reactivity of
type III PKS CylI required accessing a suitable substrate via
chemical synthesis. We hypothesized that CylI would catalyze
formation of a resorcinol-type aromatic ring via reaction of an
ACP-tethered β-ketothioester intermediate with two molecules
of malonyl-CoA (Scheme S1). We synthesized an N-acetylcyst-
The masses corresponding to these single and double
incorporation products were not found in extracts from cultures
fed with unlabeled decanoic acid. Interestingly, the double
incorporation product predominates, which may indicate that
synthesis of a monomeric precursor and dimerization are tightly
coupled, a hypothesis that is consistent with our inability to
locate masses of putative monomeric intermediates in culture
extracts. Overall, the results of this feeding experiment verify that
18520
dx.doi.org/10.1021/ja308318p | J. Am. Chem. Soc. 2012, 134, 18518−18521

Journal of the American Chemical Society
Communication
decanoic acid is a biosynthetic precursor to the cylindrocyclo-
phanes and is assimilated into both halves of the natural product
skeleton. It also confirms that the eventual C7/C20 carbon atom
of the paracyclophane scaffold enters the biosynthetic pathway as
an unactivated methylene group. Construction of this unique
macrocyclic architecture therefore requires a C−H functionaliza-
tion event at some point in the biosynthetic pathway.
One intriguing possibility for macrocycle formation is direct
oxidative sp2−sp3 C−C bond-forming dimerization of 8
involving the resorcinol aromatic ring and the unactivated alkyl
chain. A conceptually related transformation is utilized in the
biosynthesis of the prodiginine alkaloid streptorubin B and has
been linked to the activity of Reiske oxygenase RedG.5 Notably,
no RedG homolog is found in the cyl cluster or in the C.
licheniforme genome sequencing data. If cylindrocyclophane
biosynthesis does employ direct oxidative C−C bond formation,
it must therefore be carried out using distinct enzymatic
chemistry. We hypothesize that one or more of the remaining
enzymes encoded by the cyl cluster (CylC, CylJ−L) may be
involved in paracyclophane formation, and the stage is now set to
elucidate the details of this process through further biochemical
characterization.17
In summary, we have obtained the first molecular insights into
the biosynthesis of the cylindrocyclophanes through identi-
fication of a biosynthetic gene cluster, in vitro characterization of
biosynthetic enzymes, and feeding studies. Assembly of potential
biosynthetic intermediate 8 involves a remarkable combination
of polyketide biosynthetic machinery: size-selective recruitment
of a fatty acid building block by CylA and CylB, elaboration of the
fatty acid by a type I modular PKS, and termination of the
assembly line by type III PKS CylI. This unusual initiation and
off-loading logic could potentially be applied in combinatorial
biosynthesis. Incorporation of labeled decanoic acid into the
cylindrocyclophane skeleton in vivo confirmed the role of this
metabolite in biosynthesis and established that assembly of the
macrocycle involves functionalization of an unactivated carbon
center. Elucidating the biosynthetic logic underlying para-
cyclophane construction and tailoring, including the precise
nature of the key C−C bond-forming event, are important future
challenges that should enrich our understanding of how Nature
constructs complex molecular architecture.
Christopher T. Walsh, whom we acknowledge for support and
helpful advice. We received financial support from Harvard
University, the Corning Foundation, and the Searle Scholars
Program. H.N. acknowledges fellowship support from the NIH
(GM095450) and the Hershel Smith Fellowship program.
REFERENCES
■
(1) (a) Walsh, C. T.; Fischbach, M. F. J. Am. Chem. Soc. 2010, 132,
2469. (b) Weeks, A. M.; Chang, M. C. Y. Biochemistry 2011, 50, 5404.
(2) Wender, P. A.; Miller, B. L. Nature 2009, 460, 197.
(3) (a) Moore, B. S.; Chen, J. L.; Patterson, G. M. L.; Moore, R. E.;
Brinen, L. S.; Kato, Y.; Clardy, J. J. Am. Chem. Soc. 1990, 112, 4061.
(b) Moore, B. S.; Chen, J. L.; Patterson, G. M. L.; Moore, R. E.
Tetrahedron 1992, 48, 3001. (c) Bui, H. T. N.; Jansen, R.; Pham, H. T. L.;
Mundt, S. J. Nat. Prod. 2007, 70, 499. (d) Chlipala, G. E.; Sturdy, M.;
Krunic, A.; Lantvit, D. D.; Shen, Q.; Porter, K.; Swanson, S. M; Orjala, J.
J. Nat. Prod. 2010, 73, 1529.
(4) Bobzin, S. C.; Moore, R. E. Tetrahedron 1993, 49, 7615.
(5) Sydor, P. K.; Barry, S. M.; Odulate, O. M.; Barona-Gomez, F.;
Haynes, S. W.; Corre, C.; Song, L.; Challis, G. L. Nat. Chem. 2011, 3,
388.
(6) (a) Smith, A. B., III; Kozmin, S. A.; Paone, D. V. J. Am. Chem. Soc.
1999, 121, 7423. (b) Smith, A. B., III; Kozmin, S. A.; Adams, C. M.;
Paone, D. V. J. Am. Chem. Soc. 2000, 122, 4984. (c) Hoye, T. R.;
Humpal, P. E.; Moon, B. J. Am. Chem. Soc. 2000, 122, 4982.
(d) Yamakoshi, H.; et al. Org. Biomol. Chem. 2009, 7, 3772. (e) Nicolaou,
K. C.; Sun, Y.; Korman, H.; Sarlah, D. Angew. Chem., Int. Ed. 2010, 49,
5875.
(7) Calderone, C. T. Nat. Prod. Rep. 2008, 25, 845.
(8) We independently confirmed that this strain produces
cylindrocyclophane F. For isolation details see SI.
(9) Edwards, D. J.; Marquez, B. L.; Nogle, L. M.; McPhail, K.; Goeger,
D. E.; Roberts, M.; Gerwick, W. H. Chem. Biol. 2004, 11, 817.
(10) Chang, Z. X.; Sitachitta, N.; Rossi, J. V.; Roberts, M. A.; Flatt, P.
M.; Jia, J. Y.; Sherman, D. H.; Gerwick, W. H. J. Nat. Prod. 2004, 67,
1356.
(11) Descriptions of the bioinformatic analyses used to locate the gene
cluster and the strategy for cluster assembly can be found in the SI.
(12) The sequence of the CylH TE domain is most closely related to
type II editing thioesterases, which are typically freestanding enzymes
that catalyze hydrolysis at reduced rates. This homology suggests
another possibility for assembly line termination; the TE could be
functional but be outcompeted for off-loading by CylI.
(13) (a) Trivedi, O. A.; Arora, P.; Sridharan, V.; Tickoo, R.; Mohanty,
D.; Rajesh, S.; Gokhale, R. S. Nature 2004, 428, 441. (b) Arora, P.;
Goyal, A.; Natarajan, V. T.; Rajakumara, E.; Verma, P.; Gupta, R.;
Yousuf, M.; Trivedi, O. A.; Mohanty, D.; Tyagi, A.; Sankaranarayanan,
R.; Gokhale, R. S. Nat. Chem. Biol. 2009, 5, 166.
(14) Hansen, D. B.; Bumpus, S. B.; Aron, Z. D.; Kelleher, N. L.; Walsh,
C. T. J. Am. Chem. Soc. 2007, 129, 6366.
(15) (a) Austin, M. B.; Saito, T.; Bowman, M. E.; Haydock, S.; Kato, A.;
Moore, B. S.; Kay, R. R.; Noel, J. P. Nat. Chem. Biol. 2006, 2, 494.
(b) Miyanaga, A.; Funa, N.; Awakawa, T.; Horinouchi, S. Proc. Natl.
Acad. Sci. U.S.A. 2008, 105, 871.
(16) (a) Song, L.; Barona-Gomez, F.; Corre, C.; Xiang, L.; Udwary, D.
W.; Austin, M. B.; Noel, J. P.; Moore, B. S.; Challis, G. L. J. Am. Chem.
Soc. 2006, 128, 14754. (b) Chemler, J. A.; Buchholz, T. J.; Geders, T. W.;
Akey, D. L.; Rath, C. M.; Chlipala, G. E.; Smith, J. L.; Sherman, D. H. J.
Am. Chem. Soc. 2012, 134, 7359. (c) Nakano, C.; Ozawa, H.; Akanuma,
G.; Funa, N.; Horinouchi, S. J. Bacteriol. 2009, 191, 4916.
(17) Feeding studies to assess whether resorcinol 8 is an intermediate
in cylindrocyclophane biosynthesis were attempted, but were
complicated by the toxicity of 8 to C. licheniforme at low concentrations
(<10 μM); lack of incorporation could be attributed either to 8 being
off-pathway or to cell death. This observed toxicity does provide a
possible explanation for tight coupling of monomer generation and
macrocyclization.
ASSOCIATED CONTENT
* Supporting Information
■
S
Experimental details and characterization data. This material is
available free of charge via the Internet at http://pubs.acs.org.
Nucleotide sequence data have been deposited into GenBank
(accession no.: JX477167).
AUTHOR INFORMATION
Corresponding Author
balskus@chemistry.harvard.edu
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We acknowledge Katarzyna Hojczyk, Ivan Bochkov, and
Smaranda Craciun for help with early experiments. We received
assistance with LC-MS experiments from Sunia Trauger, Alan
Saghatelian, Nawaporn Vinayavekhin, and Tejia Zhang, and
advice concerning large-scale cyanobacterial cultivation from
Cyril Portmann. Preliminary work was carried out in the lab of
18521
dx.doi.org/10.1021/ja308318p | J. Am. Chem. Soc. 2012, 134, 18518−18521