Assembly of asperlicin peptidyl alkaloids from anthranilate and tryptophan: A two-enzyme pathway generates heptacyclic scaffold complexity in asperlicin E

Article abstract of DOI:10.1021/ja308371z

Members of the asperlicin family of fungal metabolites produced by Aspergillus alliaceus are known potent CCK_A antagonists. Herein, we report the identification of the gene cluster responsible for directing their biosynthesis. We validate and probe the pathway by genetic manipulation, and provide the first biochemical characterization of the oxidative cyclization en route to the heptacyclic asperlicin E by reconstituting the activity of the FAD depend monooxygenase AspB. This report provides the first genetic characterization of a NRPS assembly line that efficiently activates two anthranilate building blocks and illustrates the remarkably efficient biosynthesis of the complex heptacyclic asperlicin E.

Full text of DOI:10.1021/ja308371z

Communication
pubs.acs.org/JACS
Assembly of Asperlicin Peptidyl Alkaloids from Anthranilate and
Tryptophan: A Two-Enzyme Pathway Generates Heptacyclic Scaffold
Complexity in Asperlicin E
Stuart W. Haynes,† Xue Gao,‡ Yi Tang,*,‡,§ and Christopher T. Walsh*,†
†Department of Biological Chemistry & Molecular Pharmacology, Harvard Medical School, 240 Longwood Avenue, Boston,
Massachusetts 02115, United States
‡Department of Chemical and Biomolecular Engineering, §Department of Chemistry and Biochemistry, University of California Los
Angeles, 420 Westwood Plaza, Los Angeles, California 90095, United States
S
* Supporting Information
the third building block) and that should set up formation of
ABSTRACT: Members of the asperlicin family of fungal
metabolites produced by Aspergillus alliaceus are known
potent CCKA antagonists. Herein, we report the
identification of the gene cluster responsible for directing
their biosynthesis. We validate and probe the pathway by
genetic manipulation, and provide the first biochemical
characterization of the oxidative cyclization en route to the
heptacyclic asperlicin E by reconstituting the activity of the
FAD depend monooxygenase AspB. This report provides
the first genetic characterization of a NRPS assembly line
that efficiently activates two anthranilate building blocks
and illustrates the remarkably efficient biosynthesis of the
complex heptacyclic asperlicin E.
the tetracyclic quinazoline benodiazepinedione core, e.g., in
asperlicins C and D (Figure 1) representing a dramatic
morphing of an Ant-Ant-Trp tripeptide framework. Second, the
presumed downstream asperlicin E metabolite has a fused
angular, heptacyclic 6-6-7-6-5-5-6 ring scaffold, representing
significant architectural complexity compared to the starting
amino acids. Here, we report identification of the asperlicin
gene cluster and characterization of a short, efficient two-
enzyme pathway to asperlicin E.
A. alliaceus strain (ATCC 20656) was obtained from ATCC
and grown in glucose minimal media (GMM) to induce
asperlicin production, allowing detection of known secreted
metabolites asperlicin C (expected 407.1503, found 407.1505),
asperlicin D (expected 407.1503, found 407.1504), asperlicin E
(expected 423.1452, found 423.1452) and the mature
metabolite asperlicin (expected 536.2292, found 536.2293) by
high resolution LC/MS analysis (structures in Figure 1).
Authentic samples of enantiomeric pairs of both regioisomeric
asperlicin C and asperlicin D were synthesized by microwave
mediated methodology previously described for the synthesis of
asperlicin C (see Supporting Information) and these were used
as synthetic standards to confirm the identity of those two
metabolites (Figures S5 and S6).10 Genomic DNA was
prepared and subjected to whole genome sequencing with
approximately 575-fold coverage on an Illumina machine
(Table S1).11 A bioinformatic search for anthranilate-selective
adenylation domains8 turned up an orf we have designated as
AspA (Figure 1, Tablea S2 and S3) as a proposed bimodular
anthranilate-activating NRPS (predicted domains A-T-C-A-T-
CT , where A = Adenylation, T = Thiolation, and C =
Condensation). Immediately adjacent is AspB (predicted to be
an FAD-dependent oxygenase) and on the other DNA strand
upstream AspC, predicted to be a valine-activating mono-
modular NRPS (A-T-C). This three gene organization is
reminiscent of the fumiquinazoline biosynthetic genes we have
previously described in Aspergillus fumigatus (Af12080,
trimodular NRPS that synthesizes and activates an Ant-Trp-
Ala tripeptide to perform a double cyclization/dehydration to
the tricyclic fumiquinazoline F;12 Af12060, FAD indole
spergillus fungal strains produce a variety of peptidyl
Aalkaloids that act as mycotoxins. Several of these fungal
alkaloids incorporate anthranilate as a nonproteinogenic amino
acid building block for two and three module nonribosomal
peptide synthetase (NRPS) assembly lines.1−4 Members of the
asperlicin family of metabolites (Figure 1), produced by several
strains of Aspergillus alliaceus, were of particular interest when
discovered in the 1980s as selective antagonists of the
cholecystokinin receptor CCKA.5,6 The major metabolite
asperlicin, with its peptide scaffold biosynthetically morphed
into a tetracyclic quinazoline [3,2-a][1,4] benzodiazepine-5,13-
dione core, was a lead scaffold for subsequent high affinity,
selective CCKA ligands.
We have recently undertaken biosynthetic studies of related
Aspergillus peptidyl alkaloids in the acetylaszonalenin,
fumiquinazoline,1,7−9 and tryptoquialanine4 series. These
studies centered on the utilization of anthranilate as a
noncanonical aryl-β-amino acid starter unit in the NRPS
assembly lines in tandem with L-tryptophan and L-alanine as
additional building blocks. We deciphered the fungal code for
anthranilate selection by NRPS modules to enable identi-
fication of the relevant biosynthetic gene clusters among a large
number of NRPS candidates in aspergillus genomes.8
In this work, we have turned our attention to the A. alliaceus
asperlicin producers for two reasons. First, we predict that the
asperlicin NRPS assembly line should select and activate
anthranilate for both the first and second residues (with Trp as
Received: August 23, 2012
Published: October 2, 2012
© 2012 American Chemical Society
17444
dx.doi.org/10.1021/ja308371z | J. Am. Chem. Soc. 2012, 134, 17444−17447

Journal of the American Chemical Society
Communication
Figure 1. Schematic of asp cluster and structures of asperlicins.
epoxygenase; and Af12050, monomodular Ala-activating NRPS
to create the FQA metabolite from FQF).1 Notably however,
the Asp cluster represents the first genetic characterization of a
cluster responsible for the double incorporation of (the
noncanonical aryl-β-amino acid) anthranilate. It was therefore
surprising that AspA encodes only a bimodular NRPS (with
predicted domains A-T-C-A-T-CT) instead of the trimodular
domain one would have anticipated for construction of a
presumed Ant-Ant-Trp-enzyme tethered intermediate.
To validate the proposed function of this gene cluster, a
double crossover deletion of the bimodular orf aspA was
constructed (Figure S5)13,14 in the producing A. alliaceus strain,
resulting in loss of any detectable asperlicin metabolites,
including asperlicins C, D, E and asperlicin itself (Figure S4b).
Feeding of synthetic asperlicin C to the AspA mutant restored
production of asperlicin and asperlicin E, whereas feeding of
asperlicin D to this mutant did not result in the restoration of
any downstream products (Figure S4c,d).
AspB, the predicted FAD binding enzyme, is highly
homologous to Af12060 (Figure S1) from the fumiquinazoline
pathway that we have previously purified and demonstrated to
activate the pendant indole ring of FQF as an epoxide (that
may be in equilibrium with the hydroxyiminium tautomer of
the indole).9 Intriguingly, in this case, the product of AspB
(unlike the product of Af12060) is seen to undergo one of two
possible fates: either immediate intramolecular capture en route
to asperlicin E, or with the involvement of an additional
enzyme (AspC), further processing to asperlicin. As predicted,
by comparison to the fumiquinazoline pathway, when aspB is
deleted via double crossover (via the same methodology to
used to knock out aspA), production of asperilicin and
asperlicin E is abolished and levels of asperlicin C are
considerably increased (Figure S4e). AspB was then expressed
and purified from Escherichia coli as a yellow-pigmented
protein, characteristic of flavin-containing enzymes (Figure
S3). We compared purified AspB and Af12060 for their ability
to convert synthetic asperlicin C or D (and their enantiomers)
in the presence of proposed cosubstrate O2 to new products.
Af12060 showed no activity (Figure S2). However, pure AspB
indeed converted both enantiomers of synthetic asperlicin C to
two distinct new products (Figure 2). High-resolution LC-MS
analyses of both products were consistent with the formation of
heptacyclic asperlicin E (or isomers) (expected 423.1452, found
423.1452), via intramolecular oxygenation and subsequent
cyclization. Preparative incubations of asperlicin C with AspB
allowed isolation of sufficient product for NMR spectroscopic
characterization and confirmation of the AspB enzymatic
Figure 2. HPLC analysis (at 274 nm) of in vitro enzymatic reaction of
AspB with enantiomeric pairs of asperlicin C and D.
product as asperlicin E by comparison with the previously
characterized natural product (Figures S7 and S9). The
instability of the product generated from the reaction of
AspB with the enantiomer of asperlicin C (ent-asperlicin C)
hindered full characterization. However, the data recorded (and
the marked acceleration of its degradation under mildly acidic
conditions) are consistent with the formation of an oxidized
product, which is stalled prior to intramolecular cyclization en
route to an asperlicin E type product (Figures S8 and S9).
Interestingly, the apparent inactivity of ent-asperlicin C to
intramolecular cyclization suggests that the stereochemistry of
the Trp side chain is vital to correctly position the amide in the
7-membered diazepinone for intramolecular attack and,
furthermore, is an indication of the precision of energetic
fine-tuning of the cascade leading to the formation of asperlicin
E.
By contrast asperlicin D was completely inactive with AspB
as was fumiquinazoline F. Correspondingly, the genetic
insertion to knockout AspB (noted above) results in the
accumulation of much more asperlicin C than asperlicin D
17445
dx.doi.org/10.1021/ja308371z | J. Am. Chem. Soc. 2012, 134, 17444−17447

Journal of the American Chemical Society
Communication
Figure 3. Asperlicin biosynthetic pathway en-route to heptacyclic asperlicin E, mediated by the action of AspA and AspB.
(Figure S4), which is consistent with the further metabolism of
asperlicin C (but not asperlicin D) to further downstream
products (asperlicin E and asperlicin) in the native producer.
Thus, only two enzymes, the bimodular NRPS AspA and the
FAD-enzyme AspB, suffice to take two anthranilates and one
tryptophan building block and assemble the constrained
heptacyclic peptidyl alkaloid scaffold of asperlicin E. Figure 3
shows the proposed mechanism of these two enzymes: AspA
generating a tetracyclic scaffold and AspB converting it via
intramolecular capture of an oxygenated intermediate to the
heptacyclic framework.
Such an acyclic Ant-Ant-Trp-S-pantetheinyl-thioester teth-
ered to the second thiolation (T2) domain, as shown in Figure
3, must then undergo double cyclization. The first cyclization is
likely to be from attack of the free amino group of Ant1 on the
Trp3 thioester carbonyl (N1 on carbonyl11) to yield the 6, 11-
bicyclic system while cleaving the tethered peptide chain from
the NRPS. Subsequent transannular attack of the N5 amide
nitrogen on carbonyl11 with dehydration would give the
tetracyclic asperlicin C framework. Alternative capture of
amide nitrogen N9 on carbonyl4 (with subsequent dehydration)
yields an alternate 6-6-7-6 regiosiomeric tetracyclic core in
asperlicin D (Figure 3). The presence of a mixture of asperlicin
C and D regioisomers in producer extracts is consistent with an
uncatalyzed transannular cyclization occurring after formation
of the 6,11 bicyclic intermediate and post release from the
NRPS active site. It appears as though the asperlicin D isomer
is a dead end side product since it is not processed further by
the flavoenzyme AspB.
We formulate the AspB mediated transformation of
asperlicin C to asperlicin E via formation of the transient
indole epoxide or a hydroxyiminium adduct (via a FAD-4a-
OOH electrophilic oxygen donation to the indole 2, 3-double
bond), with presumed alpha stereochemistry, as in the related
Af12060-mediated FQF to FQA conversion.1,9 That inter-
mediate would undergo intramolecular capture by the amide in
the 7-membered diazepinone ring of asperlicin C (but
seemingly not in the diasteromer generated with ent-asperlicin
C). The poor nucleophilicty of the amide -NH is overcome in
this system by generating a strong electrophile and by the
Given only one predicted Ant-activating module (module 1)8
and one Trp activating module (module 2) in the bimodular
AspA, we presume module 1 must act iteratively to select,
activate, and sequentially load two anthranilyl units onto the T1
domain. This could happen via an anthranilyl-anthranilyl-
dipeptidyl thioester on T1 before transfer to the amino group of
Trp tethered on T2, or could involve an anthranilyl thioester on
T1 to condense with an anthranilyl-Trp-thioester that has built
up on T2. The first condensation domain may be required for
the formation of the iterative buildup of the tethered tripeptidyl
thioester while the terminal C domain is likely required for the
cyclization/release steps.12 At this stage, it remains a possibility
that a further enzyme (such as a standalone C-domain to act in
trans) may be involved in the formation of Ant-Ant dipeptide if
such an enzyme is not clustered with the Asp cluster. However,
our bioinformatic investigation of the genome only identifies
one Ant-activating A-domain. We are therefore confident that
the Ant-activating A-domain of AspA must be acting iteratively.
17446
dx.doi.org/10.1021/ja308371z | J. Am. Chem. Soc. 2012, 134, 17444−17447

Journal of the American Chemical Society
Communication
effective high local concentration of the nucleophile due to the
intramolecular nature of the addition. The action of AspB
installs the −OH group and creates a N−C bond resulting in
the linkage of the bicyclic indole moiety to what had been the
tetracyclic core by the formation of a new 5-membered
pyrroline ring. This generates the 6-6-7-6-5-5-6 heptacyclic
framework of asperlicin E (where the underlined ring is the one
just created).
This is a remarkable two-enzyme pathway: AspA and AspB
start with three amino acids, two of them anthranilates and one
tryptophan, and produce the highly constrained seven ring
scaffold. Notable is the tandem use of two anthranilates to snap
shut a 6-6-7-6 tetracyclic core, followed by the action of AspB
and subsequent intramolecular capture that converts the
bicyclic indole to a tricyclic pyrroloindole moiety, which is in
turn fused to the four ring substructure.
ACKNOWLEDGMENTS
■
We thank Steven Malcolmson, Jared Parker and Thomas
Gerken for their helpful advice and discussion. We also thank
the NIH GM20011, GM49338 (to C.T.W.) and GM092217
(to Y.T.) for funding.
REFERENCES
■
(1) Ames, B. D.; Liu, X.; Walsh, C. T. Biochemistry 2010, 49, 8564.
(2) Takahashi, C.; Matsushita, T.; Doi, M.; Minoura, K.; Shingu, T.;
Kumeda, Y.; Numata, A. J. Chem. Soc., Perkin Trans. 1 1995, 2345.
(3) Yin, W. B.; Grundmann, A.; Cheng, J.; Li, S. M. J. Biol. Chem.
2009, 284, 100.
(4) Gao, X.; Chooi, Y. H.; Ames, B. D.; Wang, P.; Walsh, C. T.; Tang,
Y. J. Am. Chem. Soc. 2011, 133, 2729.
(5) Chang, R.; Lotti, V.; Monaghan, R.; Birnbaum, J.; Stapley, E.;
Goetz, M.; Albers-Schonberg, G.; Patchett, A.; Liesch, J.; Hensens, O.;
et al. Science 1985, 230, 177.
The third enzyme encoded in the gene cluster, AspC, is
proposed to be a valine activating monomodular NRPS, akin to
the alanine activating NRPS that captures the epoxygenated
FQF species and annulates it to fumiquinazoline A.1 To
generate asperlicin, AspC presumably provides a tethered valyl
thioester to compete with the intramolecular capture of the
hydroxyiminium ion that would yield asperlicin E as shown in
Figure 3. Instead, the NH2 of the valyl moiety tethered to AspC
could capture the oxidized intermediate, thus annulating the
indole moiety to yield asperlicin, a distinct scaffold where the
seven rings systems are in separate tetracyclic and tricyclic
substructures (Figure S10). Analysis of that competition will
require purified active AspC in subsequent studies. Meanwhile,
the aspC knockout mutant gives asperlicin E as the major
product (Figure S10), consistent with shutting down the
conversion of asperlicin C to asperlicin.
Anthranilate is the key nonproteinogenic building block, in
combination with tryptophan, for Aspergilli to fashion a variety
of complex peptidyl alkaloid scaffolds including fumiquinazo-
lines,2 tryptoquialanines,4 azonalenins,3 and ardeemins15,16 in
short, efficient biosynthetic pathways. The tandem iterative
utilization of the nonproteinogenic anthranilate in the asperlicin
pathway, and related pathways builds a unique tetracyclic core
scaffold that can be further elaborated to architectural
complexity in related scaffolds seen in the benzomalvins,17,18
circumdatins,19−22 and sclerotigenin.23
(6) Liesch, J. M.; Hensens, O. D.; Zink, D. L.; Goetz, M. A. J.
Antibiot. 1988, 41, 878.
(7) Ames, B. D.; Haynes, S. W.; Gao, X.; Evans, B. S.; Kelleher, N. L.;
Tang, Y.; Walsh, C. T. Biochemistry 2011, 50, 8756.
(8) Ames, B. D.; Walsh, C. T. Biochemistry 2010, 49, 3351.
(9) Haynes, S. W.; Ames, B. D.; Gao, X.; Tang, Y.; Walsh, C. T.
Biochemistry 2011, 50, 5668.
(10) Liu, J. F.; Kaselj, M.; Isome, Y.; Chapnick, J.; Zhang, B.; Bi, G.;
Yohannes, D.; Yu, L.; Baldino, C. M. J. Org. Chem. 2005, 70, 10488.
(11) Ambry Genetics, California, USA. Assuming genome size of
approximately 30 Mb (by analogy with previously fully assembled
filamentous fungi).
(12) Gao, X.; Haynes, S. W.; Ames, B. D.; Wang, P.; Vien, L. P.;
Walsh, C. T.; Tang, Y. Nat. Chem. Biol. 2012, 8, 823.
(13) Nayak, T.; Szewczyk, E.; Oakley, C. E.; Osmani, A.; Ukil, L.;
Murray, S. L.; Hynes, M. J.; Osmani, S. A.; Oakley, B. R. Genetics 2006,
172, 1557.
(14) Szewczyk, E.; Nayak, T.; Oakley, C. E.; Edgerton, H.; Xiong, Y.;
Taheri-Talesh, N.; Osmani, S. A.; Oakley, B. R. Nat. Protoc. 2006, 1,
3111.
(15) Hochlowski, J. E.; Mullally, M. M.; Spanton, S. G.; Whittern, D.
N.; Hill, P.; McAlpine, J. B. J. Antibiot. 1993, 46, 380.
(16) Karwowski, J. P.; Jackson, M.; Rasmussen, R. R.; Humphrey, P.
E.; Poddig, J. B.; Kohl, W. L.; Scherr, M. H.; Kadam, S.; McAlpine, J. B.
J. Antibiot. 1993, 46, 374.
(17) Sun, H. H.; Barrow, C. J.; Sedlock, D. M.; Gillum, A. M.;
Cooper, R. J. Antibiot. 1994, 47, 515.
(18) Sun, H. H.; Barrow, C. J.; Cooper, R. J. Nat. Prod. 1995, 58,
1575.
(19) Rahbaek, L.; Breinholt, J.; Frisvad, J. C.; Christophersen, C. J.
Org. Chem. 1999, 64, 1689.
(20) Rahbaek, L.; Breinholt, J. J. Nat. Prod. 1999, 62, 904.
(21) Ookura, R.; Kito, K.; Ooi, T.; Namikoshi, M.; Kusumi, T. J. Org.
Chem. 2008, 73, 4245.
(22) Zhang, D.; Yang, X.; Kang, J. S.; Choi, H. D.; Son, B. W. J.
Antibiot. 2008, 61, 40.
(23) Joshi, B. K.; Gloer, J. B.; Wicklow, D. T.; Dowd, P. F. J. Nat.
Prod. 1999, 62, 650.
ASSOCIATED CONTENT
■
S
* Supporting Information
Experimental procedures, synthesis of asperlicin C and D,
isolation of genomic DNA, purification of all small molecules,
purification of AspB, details of enzymatic reactions, protocol of
genetic disruption, growth and extraction of fungal cultures,
enzymatic reactions with Af12060, estimation of bound flavin
to AspB, profiles of asperlicins with mutant strains, LC-HRMS
comparisons of synthesized asperlicins with standards and
NMR spectroscopic data. This material is available free of
charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
christopher_walsh@hms.harvard.edu; yitang@ucla.edu
Notes
The authors declare no competing financial interest.
17447
dx.doi.org/10.1021/ja308371z | J. Am. Chem. Soc. 2012, 134, 17444−17447