Thioesterase-like role for fungal PKS-NRPS hybrid reductive domains

Article abstract of DOI:10.1021/ja803078z

Fungal reduced polyketides possess diverse structures exploring a broad region of chemical space despite their synthesis by very similar enzymes. Many fungal polyketides are capped by diverse amino acid-derived five-membered rings, the tetramic acids and related pyrrolidine-2-ones. The known tetramic acid synthetase enzymes in fungi contain C-terminal reductive (R) domains that were proposed to release reduced pyrrolidine-2-one intermediates en route to the tetramic acids. To determine the enzymatic basis of pyrrolidine-2-one diversity, we overexpressed equisetin synthetase (EqiS) R domains and analyzed their reactivity with synthetic substrate analogs. We show that the EqiS R domain does not perform a reducing function and does not bind reducing cofactors. Instead, the EqiS R catalyzes a Dieckmann condensation, with an estimated k_cat ≈ 15 s^-1. This role differs from the redox reactions normally catalyzed by short chain dehydrogenase/reductase superfamily enzymes.

Full text of DOI:10.1021/ja803078z

Published on Web 07/25/2008
Thioesterase-Like Role for Fungal PKS-NRPS Hybrid
Reductive Domains
James W. Sims and Eric W. Schmidt*
Department of Medicinal Chemistry, UniVersity of Utah, 30 South 2000 East Rm 201,
Salt Lake City, Utah, 84112
Received April 25, 2008; E-mail: ews1@utah.edu
Abstract: Fungal reduced polyketides possess diverse structures exploring a broad region of chemical
space despite their synthesis by very similar enzymes. Many fungal polyketides are capped by diverse
amino acid-derived five-membered rings, the tetramic acids and related pyrrolidine-2-ones. The known
tetramic acid synthetase enzymes in fungi contain C-terminal reductive (R) domains that were proposed to
release reduced pyrrolidine-2-one intermediates en route to the tetramic acids. To determine the enzymatic
basis of pyrrolidine-2-one diversity, we overexpressed equisetin synthetase (EqiS) R domains and analyzed
their reactivity with synthetic substrate analogs. We show that the EqiS R domain does not perform a
reducing function and does not bind reducing cofactors. Instead, the EqiS R catalyzes a Dieckmann
condensation, with an estimated kcat ≈ 15 s-1. This role differs from the redox reactions normally catalyzed
by short chain dehydrogenase/reductase superfamily enzymes.
Introduction
Equisetin (1) is a polyketide tetramic acid from filamentous fungi
that exemplifies a broad group of structurally diverse, biosyntheti-
cally related tetramic acids and derivatives (Figure 1).1–4 Equisetin
is active in an array of assays and inhibits HIV-1 integrase,1,5–7
leading to its use as a model for the development of the clinically
approved integrase inhibitor, raltegravir.8,9 It is structurally and
biochemically related to lovastatin, the canonical antihypercholes-
teremia agent that underlies the multibillion-dollar statin drug
class.10–12 In addition, equisetin is biosynthetically related to a broad
array of active fungal tetramic acids and their derivatives, including
fusarin C (2),13 cytochalasins,14 and relatives such as chaetoglobosin
A (3),15 tenellin (4),16 and pseurotin A (5),17 that are important in
(1) Vesonder, R. F.; Tjarks, L. W.; Rohwedder, W. K.; Burmeister, H. R.;
Laugal, J. A. J. Antibiot. 1979, 32, 759–761.
(2) Phillips, N. J.; Goodwin, J. T.; Fraiman, A.; Cole, R. J.; Lynn, D. G.
J. Am. Chem. Soc. 1989, 111, 8223–8231.
(3) Royles, B. J. L. Chem. ReV. 1995, 95, 1981–2001.
(4) Schobert, R.; Schlenk, A. Bioorg. Med. Chem. 2008, 16, 4203–21.
(5) Konig, T.; Kapus, A.; Sarkadi, B. J. Bioenerg. Biomembr. 1993, 25,
537–545.
Figure 1. Representative polyketide-pyrrolidine-2-one metabolites from
fungi. Polyketide-derived portions are blue, whereas amino acid-derived
portions, including the pyrrolidine moieties, are red. Equisetin is the only
true “tetramic acid” shown. All compounds shown are synthesized by
homologous proteins.
(6) Singh, S. B.; Zink, D. L.; Goetz, M. A.; Dombrowski, A. W.;
Polishook, J. D.; Hazuda, D. J. Tetrahedron Lett. 1998, 39, 2243–
2246.
(7) Wheeler, M. H.; Stipanovic, R. D.; Puckhaber, L. S. Mycol. Res. 1999,
103, 967–973.
agriculture and medicine. These compounds contain either an intact
tetramic acid ring or reduced and highly modified pyrrolidine-2-
one rings.
The genetic basis of this structural diversity has been elusive.
Genes have been identified for an array of different tetramic acid
derivatives with variable carbon skeletons and pyrrolidine-2-one
modifications.18–24 In all of these cases, the predicted protein
consists of a polyketide synthase (PKS)-nonribosomal peptide
(8) Handy, S. T.; Chang, K. Abs. Pap. Am. Chem. Soc. 2000, 219, U109–
U109.
(9) Handy, S. T.; Omune, D.; Menon, G. Abs. Pap. Am. Chem. Soc. 2001,
222, U79–U79.
(10) Endo, A.; Monacolin, K J. Antibiot. 1979, 32, 852–854.
(11) Hendrickson, L.; Davis, C. R.; Roach, C.; Nguyen, D. K.; Aldrich,
T.; McAda, P. C.; Reeves, C. D. Chem. Biol. 1999, 6, 429–439.
(12) Kennedy, J.; Auclair, K.; Kendrew, S. G.; Park, C.; Vederas, J. C.;
Hutchinson, C. R. Science 1999, 284, 1368–1372.
(13) Gaddamidi, V.; Bjeldanes, L. F.; Shoolery, J. N. J. Agric. Food Chem.
1985, 33, 652–654.
(16) McInnes, A. G.; Smith, D. G.; Wat, C.-K.; Vining, L. C.; Wright,
J. L. C. Chem. Commun. 1974, 281–282.
(17) Bloch, P.; Tamm, C.; Bollinger, P.; Petcher, T. J.; Weber, H. P. HelV.
Chim. Acta 1976, 59, 133–137.
(18) Song, Z.; Cox, R. J.; Lazarus, C. M.; Simpson, T. J. Chembiochem
2004, 5, 1196–1203.
(14) Smith, G. F.; Ridler, M. A.; Faunch, J. A. Nature 1967, 216, 1134–
1135.
(15) Silverton, J. V.; Kabuto, C.; Akiyama, T. Acta Crystallogr. 1978, B34,
588–593.
9
10.1021/ja803078z CCC: $40.75
2008 American Chemical Society
J. AM. CHEM. SOC. 2008, 130, 11149–11155 11149

A R T I C L E S
Sims and Schmidt
known NRPS R domains and virtually all of their relatives in the
short chain dehydrogenase/reductase (SDR) superfamily catalyze
redox reactions, this proposal appeared to be strongly supported.33
Like these characterized reductive NRPS proteins, EqiS and other
pyrrolidine NRPSs contain adenylation (A) and thiolation (T)
domains, which activate amino acids as covalent enzyme-linked
thioesters,34 as well as condensation (C) domains, which catalyze
peptide bond formation.35 The NRPS proteins thus follow the
standard domain order CATR. Recently, the tenellin PKS was
heterologously expressed, resulting in tetramic acid formation.36
This result supported a second possibility, that tetramic acids
themselves are directly offloaded by this enzyme group via a
Dieckmann condensation (Figure 2). In this scenario, the genetic
basis for compound diversity would be less clear. However, as this
result was obtained in vivo, it remained possible that a reduction-
oxidation process occurred within whole fungal cells.
Since understanding the mechanism of tetramic acid formation
would allow improved drug discovery and biosynthetic engineering
in this important compound class, we set out to test these and other
possibilities using the equisetin biosynthetic gene cluster eqi from
F. heterosporum.19 Synthesis of intermediate analogs and kinetic
analysis using purified equisetin synthetase (EqiS) protein domains
revealed that the EqiS R domain is a Dieckmann cyclase, not a
reductase. The proposed chemical mechanism bears some similarity
to that catalyzed by thioesterase enzymes in PKS and NRPS
metabolism, indicating a new role for fungal NRPS R domains.
These results have implications for the evolution of the diverse
fungal polyketide-pyrrolidine-2-one compound family.
Figure 2. Proposed biogenesis of equisetin and possible roles of R domain.
A polyketide portion (blue) is synthesized by the PKS, whereas serine (red)
is appended by the NRPS. A covalently enzyme-bound thioester intermediate
is the substrate for R, which would cleave the intermediate from the enzyme.
R could catalyze a reduction, leading to an aldol reaction followed by
oxidation to give equisetin. Alternatively, it could function as a Dieckmann
cyclase. The timing of N-methylation is unclear.
synthetase (NRPS) hybrid, terminating in an apparently typical
NRPS reductive (R) domain (Figure 2). The PKS proteins appear
to be fungal highly reducing iterative synthases, which use the same
set of domains in an iterative manner to cobble together carbon
skeletons from acetyl- and malonyl-CoA.11,12,25–27 Prediction of
the resulting polyketide structure using gene sequences is beyond
current art, as all of these PKS genes appear quite similar despite
vastly different carbon skeletons. Subsequently, the NRPS
module is proposed to append an amino acid, resulting in a beta-
ketoamide aminoacetyl-thioester. This enzyme-bound thioester
intermediate is primed for cleavage to yield the final pyrrolidine-
2-one derivative. While the polyketide structural diversity is
likely controlled by the PKS, the genetic basis for pyrrolidine-
2-one diversity remained unclear.
Experimental Methods
Expression of EqiS Domains. Detailed experimental methods
for generating vectors, chemical synthesis, and other procedures
can be found in the Supporting Information. C-terminal domain
constructs of EqiS were expressed in E. coli, resulting in proteins
that were purified to homogeneity using Ni-NTA resin followed
by FPLC with a Superdex SD200 size exclusion column. This
column was also used on a series of size standards, allowing the
size of eluted proteins to be accurately estimated. Expressed proteins
included ATR, TR, and R domains fused to N-terminal 6×-His
tags. Purity of proteins was assessed by FPLC and SDS-PAGE.
TR was analyzed by ESI-MS: ESI-MS [M + H2O] 52784
(C2341H3707N651O709S14 [M + H2O] calcd 52783).
Two proposals have been advanced to explain the diverse pyr-
rolidine-2-ones in this class of fungal polyketide metabolites. NRPS
R domains have been characterized in a number of cases to reduce
enzyme-bound thioesters using NAD(P)H as a cofactor, resulting
in aldehydes or alcohols.28–32 In the pyrrolidine-2-one family, it
was possible that a reductive process would lead to a universal
aldol intermediate (Figure 2). This aldol could undergo a variety
of different fates, including oxidation back to the tetramic acid,
oxidation at another site to yield pyrrolidine-2-one derivatives, or
Diels-Alder cyclization to yield the cytochalasin family. Since the
R Domain Putative Substrate and Product Synthesis. For
mechanistic analysis, substrate analogs that were capable or
incapable of undergoing a Dieckmann condensation were synthe-
sized. Briefly, these consisted of acetoacetyl-alanine and acetyl-
alanine, respectively. To imitate covalent enzyme attachment, these
substrates were attached as thioesters to N-acetylcysteamine or
coenzyme A (CoA) using the transthiolation method.37 Thioester
substrates were purified to homogeneity by HPLC and characterized
by NMR and MS. Standards of putative products of enzymatic
(28) Ehmann, D. E.; Gehring, A. M.; Walsh, C. T. Biochemistry 1999, 38,
6171–6177.
(19) Sims, J. W.; Fillmore, J. P.; Warner, D. D.; Schmidt, E. W. Chem.
Commun. 2005, 186–188.
(29) Gaitatzis, N.; Kunze, B.; Mu¨ller, R. Proc. Natl. Acad. Sci. U.S.A. 2001,
98, 11136–11141.
(20) Eley, K. L.; Halo, L. M.; Song, Z.; Powles, H.; Cox, R. J.; Bailey,
A. M.; Lazarus, C. M.; Simpson, T. J. Chembiochem 2007, 8, 289–
297.
(30) Becker, J. E.; Moore, R. E.; Moore, B. S. Gene 2004, 325, 35–42.
(31) Kopp, F.; Mahlert, C.; Gru¨newald, J.; Marahiel, M. A. J. Am. Chem.
Soc. 2006, 128, 16478–16479.
(32) Read, J. A.; Walsh, C. T. J. Am. Chem. Soc. 2007, 129, 15762–15763.
(33) Oppermann, U.; Filling, C.; Hult, M.; Shafqat, N.; Wu, X.; Lindh,
M.; Shafqat, J.; Nordling, E.; Kallberg, Y.; Persson, B.; Jornvall, H.
Chem. Biol. Interact. 2003, 143-144, 247–253.
(34) Ehmann, D. E.; Shaw-Reid, C. A.; Losey, H. C.; Walsh, C. T. Proc.
Natl. Acad. Sci. U.S.A. 2000, 97, 2509–2514.
(35) Belshaw, P. J.; Walsh, C. T.; Stachelhaus, T. Science 1999, 284, 486–
489.
(36) Halo, L. M.; Marshall, J. W.; Yakasai, A. A.; Song, Z.; Butts, C. P.;
Crump, M. P.; Heneghan, M.; Bailey, A. M.; Simpson, T. J.; Lazarus,
C. M.; Cox, R. J. Chembiochem 2008, 9, 585–594.
(37) Vitali, F.; Zerbe, K.; Robinson, J. A. Chem. Commun. 2003, 2718–
2719.
(21) Schumann, J.; Hertweck, C. J. Am. Chem. Soc. 2007, 129, 9564–9565.
(22) Maiya, S.; Grundmann, A.; Li, X.; Li, S. M.; Turner, G. Chembiochem
2007, 8, 1736–1743.
(23) Bergmann, S.; Schumann, J.; Scherlach, K.; Lange, C.; Brakhage,
A. A.; Hertweck, C. Nat. Chem. Biol. 2007, 3, 213–217.
(24) Bohnert, H. U.; Fudal, I.; Dioh, W.; Tharreau, D.; Notteghem, J. L.;
Lebrun, M. H. Plant Cell 2004, 16, 2499–2513.
(25) Auclair, K.; Sutherland, A.; Kennedy, J.; Witter, D. J.; Van den Heever,
J. P.; Hutchinson, C. R.; Vederas, J. C. J. Am. Chem. Soc. 2000, 122,
11519–11520.
(26) Auclair, K.; Kennedy, J.; Hutchinson, C. R.; Vederas, J. C. Bioorg.
Med. Chem. Lett. 2001, 11, 1527–1531.
(27) Nicholson, T. P.; Rudd, B. A. M.; Dawson, M.; Lazarus, C. M.;
Simpson, T. J.; Cox, R. J. Chem. Biol. 2001, 8, 157–178.
9
11150 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Fungal Reductive Thioesterases
A R T I C L E S
reactions were synthesized as follows: The tetramic acid resulting
from acetoacetyl-Ala was synthesized by the conditions of Lacey,38
purified to homogeneity, and characterized by NMR and MS. A
putative aldehyde product of acetyl-Ala was synthesized by partial
reduction using DIBALH. The unstable intermediate was character-
ized by TLC, staining with 2,4-dinitrophenylhydrazine. Detailed
methods, including spectral data for reported compounds, are
available in the Supporting Information.
Covalent Loading of T Using Sfp and Synthetic Substrate
Analogs. The TR protein was characterized by loading the T domain
with various Coenzyme A analogs via Sfp.39,40 TR (180 µL, 16
µM) was incubated with Sfp (3 µL, 10 µM) in 50 mM Bis-Tris pH
6.0 or HEPES pH 6.0, 10 mM MgCl2, and 0.4 mM substrate at 30
°C for 1 h. Loading of fluorescent analog was measured by an
Odyssey Infrared Imaging System (Li-COR). Competition assays
were performed under identical conditions, incubating the acety-
lalanine-CoA or acetoacetylalanine-CoA derivative (0.4 mM) for
1 h prior to addition of Alexafluor-CoA. To test reactivity, the
pH was adjusted to 7 after 1 h and NADPH added. TR loaded
with acetylalanine-pantetheine: ESI-MS [M + H2O] 53236
(C2357H3733N654O717PS15 [M + H2O] calcd 53234).
Kinetic Analysis of R. The reductive domain was characterized
using synthetic substrate mimics. Enzymatic assays were performed
using 3.2 µM purified R or TR domains, 2-40 mM substrates, 1.5
mM NADPH, 10 mM MgCl2, 10 µM MnCl2, and 50 mM NaCl at
25 °C. Also, these reactions were reproduced in the presence and
absence of 1.5 mM NADH or NADPH. The assays were quenched
by addition of conc. H2SO4. The acidic aqueous mixture was
extracted with a slight excess of CHCl3 containing 5 ng/µL of
limonene as internal standard. The CHCl3 layer was analyzed by
GC-MS for product identification, or by FID for kinetic analysis.
Relative rates of reaction were determined by comparing the ratio
of the tetramic acid product to the internal standard limonene. A
standard curve was generated using synthetic tetramic acid at
varying concentrations. Both enzyme and substrate concentrations
were varied along with other controls to determine the rates of the
reaction.
Control Reactions. R exhibited unusual behavior in that it was
relatively thermostable (Figure S5, Supporting Information). To
ensure that the observed reaction was due to enzymatic activity,
Qiagen Protease was used in a series of control reactions. Qiagen
Protease in water was either boiled for 20 min at 100 °C or left on
ice. The R domain (16 µM) was treated with 8 µL (50 mU) of
boiled or unboiled protease for 1 h at rt. These were subsequently
used in enzyme assays as control reactions.
To test for cofactor binding, R (16 µM 20 µL) was mixed with
urea (9 M, 80 µL) and incubated at rt for 1 h. The resulting mixture
was analyzed on a HP UV-vis spectrophotometer and by ESI-MS
in comparison to authentic standard of NADPH and NADP+. The
limits of NADPH detection were estimated to be <30 nM with
similar injection volumes, providing excellent sensitivity.
Product Analysis. Chemical products from enzymatic reactions
were detected in a number of ways. Ketoamide derivatives were
readily observed on TLC plates following staining with ferric
chloride, while aldehydes could be detected on TLC using 2,4-
dinitrophenylhydrazine. Volatile standards could be readily observed
by GC-MS and confirmed on the basis of the parent ion mass.
Thus, several different, sensitive detection methods were used to
observe the products of enzyme reactions. These methods were used
with many different enzyme reaction conditions, and only tetramic
acid products were detected. In addition, data concerning degrada-
tion of synthetic intermediates was obtained both via kinetic analysis
and by examining 1H NMR spectra following different treatments.
Formation of tetramic acid from thioester derivatives could be
clearly observed by NMR. Thus, conditions for enzyme reactions
were determined following numerous chemical and enzymatic
preliminary investigations.
NADPH Consumption. The reaction mixture used to test for
NADPH consumption was identical to that used in reductive domain
characterization, which has been previously described above.
Consumption was analyzed by monitoring loss of absorbance at
340 nm in 40 µL reactions in a 96 well plate in triplicate on a
Vector3V automated UV plate reader (Perkin-Elmer) for 2 h with
a 5 min delay between measurements. Reactions of 100 µL were
observed using a HP UV-vis spectrophotometer. Every 5 min, an
aliquot of the reaction was diluted to 1:100 and measurements were
taken. Following loading of TR, the pH was increased to 7.0 and
1.5 mM NADPH was added. This mixtures was immediately placed
in SpectraMax M5 (Molecular Devices) automated plate reader,
monitoring change in absorbance 340 nm for 1 h. A buffer only
sample, lacking protein, served as a negative control.
Results
Heterologous EqiS NRPS Domains Are Stably Expressed
and Folded. To assess R domain function three protein
constructs of EqiS were expressed and purified as soluble,
N-terminal 6xHis-tagged proteins, including ATR, TR, and R.
Unfortunately, ATR existed in several oligomeric states that
complicated analysis (Figure S3, Supporting Information). By
contrast, both TR and R proteins were monomers that were more
uniform and stable than ATR (Figure S5, Supporting Informa-
tion), so they were used for further analysis. Proper folding of
TR was tested by enzymatically loading a fluorescent-CoA
derivative using Sfp, a broad substrate pantetheinyltransferase.
TR treated with this derivative was brightly fluorescent, indicat-
ing that TR contained a competent thiolation domain (Figure
3). Both TR and R were further analyzed by CD and mass
spectrometry. HPLC-ESI-MS analysis of the tryptic digest
of TR and R showed that these were the predicted proteins (data
not shown). Both pantetheine-modified TR and unmodified TR
were also analyzed as intact proteins by ESI-MS, confirming
that TR was pantetheinylated to afford nearly 100% holoprotein
(Figure 3). CD spectra of the TR and R show that both are
structured proteins with melting temperatures of >90 and 75
°C, respectively (Figure S5, Supporting Information). Thus, TR
and R were used to test the mechanism of tetramic acid
formation in equisetin.
EqiS R Domain Produces Tetramic Acids and Not Reduced
Intermediates. On the basis of several considerations, there were
three likely possibilities for the formation of equisetin (Figure
2). First, R could catalyze a reduction, yielding an aldehyde
that would aldol cyclize. Subsequent reoxidation by R or another
protein would give the tetramic acid. Second, R could directly
catalyze a Dieckmann condensation to yield the tetramic acid.
Third, R may be uninvolved in tetramic acid formation.
Synthetic substrate analogs were designed to tease apart these
mechanistic possibilities (Figure 4). All of the analogs are
thioesters, resembling the presumed enzyme-linked thioester
substrate of wild-type R.31,32 Ala was used in place of Ser for
ease of synthesis. Finally, Ala was modified by N-acetylation
or N-acetoacetylation. The acetyl derivative was designed so
that it could in principle be reduced by R, but the pKa of the
acetate methyl is too high to allow aldol condensation in this
(41) Geoghegan, K. F.; Dixon, H. B.; Rosner, P. J.; Hoth, L. R.; Lanzetti,
A. J.; Borzilleri, K. A.; Marr, E. S.; Pezzullo, L. H.; Martin, L. B.;
LeMotte, P. K.; McColl, A. S.; Kamath, A. V.; Stroh, J. G. Anal.
Biochem. 1999, 267, 169–184.
(42) Dorrestein, P. C.; Van Lanen, S. G.; Li, W.; Zhao, C.; Deng, Z.; Shen,
B.; Kelleher, N. L. J. Am. Chem. Soc. 2006, 128, 10386–10387.
(38) Lacey, R. N. J. Chem. Soc. 1954, 850–854.
(39) Sieber, S. A.; Walsh, C. T.; Marahiel, M. A. J. Am. Chem. Soc. 2003,
125, 10862–10866.
(40) La Clair, J. J.; Foley, T. L.; Schegg, T. R.; Regan, C. M.; Burkart,
M. D. Chem. Biol. 2004, 11, 195–201.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11151

A R T I C L E S
Sims and Schmidt
Figure 4. Synthesis of substrate analogs to test mechanistic possibilities.
Analogs 6-9 were synthesized, as were standards 10 and 11. If R operates
as a typical NRPS reductase, reduced products should be observed. R
catalyzes the Dieckmann condensation to give 11, which was the only
observed product in any enzymatic reaction with R.
Acetyl- and acetoacetyl-Ala thioesters were synthesized, using
both CoA and SNAC thiols (Figure 4 and Figure S7, Supporting
Information). The acetylated derivatives were relatively stable,
while the acetoacetyl derivatives were unstable at pH > 7.5.
Standards of predicted products were synthesized, including a
tetramic acid and an acetyl-alaninyl aldehyde (data not shown).
The tetramic acid was stable and readily purified to homogene-
ity, while the unstable aldehyde was only transiently present in
solution and was detected by TLC.
Both the TR and R domains were analyzed for reactivity with
substrate analogs. The purified R domain was treated with the
acetoacetylalanine-CoA substrate, both with and without NAD-
PH. The reactions were monitored by TLC and GC-MS (Figure
5). Under these conditions, tetramic acid was produced, as
verified by comparison with an authentic standard. At pH 8,
nonenzymatic background formation of tetramic acid was rapid,
while at pH 6, no tetramic acid was formed either enzymatically
or nonenzymatically. Experiments to determine the pH sensitiv-
ity are described in the enzyme kinetics section. Therefore, all
further reactions were performed at pH 7.0, a condition giving
negligible background but rapid enzymatic turnover to give the
tetramic acid. Addition of either CoA or SNAC thioesters of
acetoacetylalanine afforded tetramic acid at approximately
equivalent rates. However, the acetylalanine derivatives were
not substrates for the reaction.
No aldehyde or aldol products were observed by GC-MS
or by TLC under any condition, using either the acetyl or
acetoacetyl analogs. With the acetyl aldehyde analog, a synthetic
standard was available that clearly stained with DNP on TLC
plates. No such standard was available for the acetoacetyl
analog. However, no aldehyde was detected by TLC. In addition,
the acetoacetyl moiety could be readily observed by TLC
staining with ferric chloride. No peak, aside from starting
material or tetramic acid, was observed. Moreover, no peak
corresponding to a reduced product or a reduced product from
which water was eliminated was ever detected by GC-MS.
Instead, we only observed tetramic acids by both TLC and
GC-MS. Addition of NADPH was not required for production
of tetramic acid, but tight binding of NADPH could not be ruled
out. Thus, two likely mechanisms remained for the role of the
reductive domain: either this enzyme catalyzes a Dieckmann
Figure 3. Covalent modification of T domain. a-c are ESI-MS of whole
proteins, y-axis is % abundance, x axis is m/z from 52200 to 54200 Da.
Smaller peaks at +178 are likely due to gluconoylation of the His-tag.41,42
(a) ESI-MS of purified T-R didomain. (b) ESI-MS of T-R loaded with
acetylalanine-CoA using Sfp pantetheinyltransferase. (c) ESI-MS of
acetylalanine-loaded T-R following incubation with NADPH at pH 7. No
reductive cleavage occurs, and the shift of 2 Da for 1 of the 2 peaks is
within error. (d) Coomassie staining (top) and fluorescence analysis of T-R
competition loading experiments. The protein in the left lane was treated
with acetylalanine-CoA and Sfp for 1 h, followed by treatment with
fluorescent CoA. The protein at right was treated with buffer and Sfp for
1 h, then with fluorescent CoA. This shows that T-R is ∼100% covalently
loaded by Sfp.
context. Thus, this substrate would allow direct monitoring of
a thioester reduction process. The acetoacetyl derivative could
be reduced, yielding aldol products, or alternatively it is also a
substrate for the Dieckmann condensation.
9
11152 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Fungal Reductive Thioesterases
A R T I C L E S
Figure 5. Dieckmann condensation catalysis by R. (a) GC-MS trace of tetramic acid standard 11, extracting ions at m/z ) 155. (b) Mass pattern of 11.
(c) GC-MS extracted for m/z ) 155 of R reaction with 9. Tetramic acid is produced and exhibited the same spectrum as in (b). (d) GC-MS extracted for
m/z ) 155 of a negative control lacking enzyme. (e) Velocity of tetramic acid formation. x-axis: substrate concentration in mM; y-axis: rate of product
formation in mM min-1
.
condensation, or it recycled NADPH to both reduce and oxidize
the substrate. While the recycling mechanism seems convoluted,
sugar epimerase enzymes use this mechanism and are homolo-
gous with R.43–45
Therefore, the phosphopantetheine attachment site was com-
pletely blocked with acetyalaninyl-phosphopantetheine. This
reaction was repeated with the acetoacetylalanine-CoA com-
pound, also demonstrating complete conversion to the holopro-
tein. ESI-MS of the intact protein further established that these
proteins were modified as predicted. In addition to providing a
rapid assay for apoprotein loading, the fluorescent competition
assay was useful to quickly assess Sfp functionality, revealing
that the enzyme was functional from pH 5 to 9. This pH range
was critical to avoid conditions amenable to tetramic acid
synthesis or thioester hydrolysis.
The loaded TR domains were used to define the effects of
NAD(P)H cofactors using ESI-MS. The loading reactions with
Sfp were done at pH 6.0, since we had previously determined
that R did not react with acetoacetyl derivatives at pH < 6.5.
(Experiments to determine pH sensitivity are described in the
next section.) The acetylalanine-CoA loaded TR showed a shift
in mass of 453 AMU, corresponding to modification with a
phosphopantetheinyl-acetylalanine. After increasing the pH to
7 and incubating with equimolar NADPH or excess or NADPH,
no significant change in mass was observed by ESI-MS (Figure
3; Figure S9, Supporting Information). When excess NADH
was used, a mass was observed consistent with no modification
of substrate and binding to one equivalent of NADH. Consump-
tion of NAD(P)H was also monitored by UV, which showed
no change in absorbance at 340 nm greater than background.
Thus, R was incapable of reducing this thioester.
NAD(P)H is Not Required for Tetramic Acid Synthesis. The
role of the cofactor was further analyzed by monitoring for
consumption by UV spectroscopy. Both reactions containing
acetoacetylalanine-CoA or acetylalanine-CoA were monitored
on an automated plate reader. Reactions were performed in
triplicate with varying substrate concentrations, from 0 to 4 mM,
with and without enzyme. All rates of consumption of NADPH
were equivalent, indicating that enzymatic consumption of
NADPH was negligible. However, this left the possibility that
NADPH was bound tightly enough to coelute with the protein
upon purification, again as seen in some representatives of the
related sugar epimerase family. To determine if this was the
case with EqiS, R was treated with 8 M urea or Qiagen Protease
for an hour, and the resulting mixtures were analyzed by
MALDI-MS and DAD-HPLC in comparison to standards. In
no case was NAD(P)(H) observed, which is inconsistent with
a tightly bound cofactor hypothesis. As our detection limit was
30 nM, we would have observed the cofactor even if only ∼5%
of the protein was cofactor-loaded. Also, when TR was
submitted for ESI-MS analysis, only the apoprotein was
observed, with no shift for any tightly bound cofactor. Thus,
EqiS R does not copurify with redox cofactors, ruling out the
recycling mechanism for tetramic acid biosynthesis.
Covalently Loaded TR Does Not Catalyze Reduction. The
above experiments were done using substrate analogs that were
not covalently loaded onto T. It remained possible that R could
reduce covalently loaded substrates that more closely resemble
the natural pathway intermediates. To test this possibility, TR
was covalently loaded with acetyl- and acetoacetyl-Ala-CoA
derivatives using the broad-substrate pantetheinyltransferase, Sfp
(Figure 3 and Figure S6, Supporting Information). A competition
experiment was performed to show that this reaction is specific
and complete. Acetylalanine-CoA or buffer only was preincu-
bated with TR and Sfp, followed by treatment with Alexafluor-
CoA. The resulting proteins were visualized by SDS-PAGE
followed by fluorescence analysis, revealing that the buffer-
only control was highly fluorescent, but no fluorescence could
be detected in TR that was pretreated with acetylalanine-CoA.
When the acetoacetylalanine-CoA substrate was used in
similar experiments, the acetoacetylalanine group was lost even
at pH 6, and only the pantetheine-bound enzyme was observed
by ESI-MS. This result strongly indicates that the protein likely
catalyzes tetramic acid synthesis even at much reduced pH when
the substrate is covalently enzyme-bound for the following
reason. When using noncovalent substrates (SNAC or CoA
derivatives), the thioesters are stable for hours with no detectable
degradation by NMR. Thus, the acetoacetyl group would not
normally be lost under these conditions. Covalent attachment
of a substrate to enzyme is known to greatly accelerate reactions,
since this covalent attachment much more closely imitates the
natural substrate and much of the Km factor is removed. Thus,
the fact that the acetoacetyl group is lost at this low pH when
covalently linked to enzyme further supports this as an enzyme-
catalyzed reaction. Because these are by necessity single-
turnover experiments, it is technically demanding to detect the
small molecule products of the reaction. However, in numerous
chemical analyses of the acetoacetyl derivatives, we in no case
detected the hydrolysis product instead of the tetramic acid.
(43) Liu, Y.; Thoden, J. B.; Kim, J.; Berger, E.; Gulick, A. M.; Ruzicka,
F. J.; Holden, H. M.; Frey, P. A. Biochemistry 1997, 36, 10675–10684.
(44) Berger, E.; Arabshahi, A.; Wei, Y.; Schilling, J. F.; Frey, P. A.
Biochemistry 2001, 40, 6699–6705.
(45) Mayer, A.; Tanner, M. E. Biochemistry 2007, 46, 6149–6155.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11153

A R T I C L E S
Sims and Schmidt
Moreover, NAD(P)H was not consumed, as measured by ex-
tremely sensitive UV techniques.
constant enzyme concentrations and varying substrate from 2
mM to 40 mM, a linear slope was observed, indicating that
saturation could not be achieved for this enzyme. A total of 60
independent measurements were used to obtain a reaction
velocity slope. Assuming linearity, we calculated a V/Km of
0.018 min-1 using the Michaelis-Menten approximation.
Because saturating substrate concentrations could not be achieved,
the lower limit for the rate of reaction could only be estimated
at kcat ≈ 15 s-1 using a Lineweaver-Burk plot. This rate
estimate falls within an order of magnitude of the rates of
reduction observed with sugar epimerases and NRPS R domains,
indicating a robust catalysis of tetramic acid formation. We
could not approach the Km limit in these experiments because
it was very high, exceeding our substrate saturation limit. These
are unnatural substrate analogs, and moreover, they are not
covalently linked to enzyme as natural substrates would be.
Thus, it can be expected that Km for the enzyme-tethered, natural
substrates would be much lower than that for the noncovalent,
unnatural substrates. The overall catalytic rate of reaction should
be much faster for the natural substrates. These experiments
were not designed to measure this natural rate, but rather to
show that the process is enzyme catalyzed. These experiments
indeed clearly indicate that the R domain catalyzes a Dieckmann
condensation, that cofactor is not required, and that an aldehyde
intermediate is not produced en route to the tetramic acid
equisetin.
EqiS R Catalyzes a Dieckmann Condensation. A reductive
role for R was ruled out by the above experiments, strongly
suggesting a Dieckmann mechanism to form the tetramic acid.
However, a caveat to these experiments is that the substrates
are analogs of the natural ones, and thus it was possible that
the natural substrates would be reduced. It was thus important
to obtain kinetic measurements to validate the Dieckmann
condensation enzyme activity. To provide further evidence that
this is indeed an enzyme catalyzed Dieckmann condensation,
studies on the rate of reaction were done (Figure 5). All
reactions were run as independent experiments in triplicate,
using acetoacetylalanine-SNAC as the substrate. The SNAC
derivative was selected because it was substantially easier
to synthesize and purify and had approximately the same
reactivity as the CoA.
Experiments were performed to determine the pH dependence
of this process. Many similar thioesterase-catalyzed reactions
affect substrates with similar or identical pKa’s to those of our
synthetic substrates, and thus it was known that there may be
some pH sensitivity to this reaction. For example, the “Claisen
cyclase” subclass of thioesterases catalyzes a reaction involving
groups with very similar pH-based activity to those reported
here,46 and even “normal” thioesterase macrocyclization reac-
tions are somewhat similar when amines are involved in terms
of their pH dependence. As alluded to above, the enzyme was
not reactive at pH 6 down to the detection limit (250 ng after
1 h reaction), whereas at pH 8, reactions in duplicate indicated
that the background rate in the absence of enzyme was ∼50%
of the enzyme-catalyzed rate at that pH. At pH 7, the background
rate was measured in a time-course experiment to be <5% of
the enzyme-catalyzed rate (Figure S9, Supporting Information).
A background rate was also obtained using protease-digested
enzyme, as described below and used in the calculation of the
enzyme-catalayzed rate.
An additional problem was the thermotolerance exhibited by
R. It was possible to kill the enzyme activity after extensive
boiling for >15 min, but 5 min at 100 °C did not completely
abolish the activity. Therefore, the stability of the enzyme was
measured by circular dichroism (Figure S5, Supporting Informa-
tion). R began to lose structure at 75 °C, whereas TR did not
become unstructured even at 90 °C. Therefore, both constructs
were relatively thermotolerant in comparison to many fungal
enzymes. Such thermotolerance is a common occurrence in
proteins even from mesophilic organisms, including fungi.47 In
order to obtain a relevant control, we chose instead to digest R
with Qiagen Protease. In this control, all of the buffers and the
amino acid content would be identical to the actual reaction,
but R would be fully degraded by the protease. Duplicate
reactions were set up in which protein was treated either with
Qiagen Protease or with a boiled control of Qiagen Protease.
By SDS-PAGE, the proteolyzed R domain was completely
degraded, while R treated with boiled protease was intact. The
control protein still produced the tetramic acid while the
proteolyzed solution did not. This showed unequivocally that
the reaction is protein catalyzed.
Discussion
NRPS R domains are found in a wide range of biosynthetic
pathways to natural products of biomedical and agricultural
importance.18–24 EqiS synthesizes equisetin from polyketide and
peptide building blocks, forming the tetramic acid ring via a
Dieckmann condensation catalyzed by R. By contrast, all of
the previously characterized NRPS R domains from bacteria
and fungi are reductases, catalyzing release of an enzyme-bound
thioester to yield an intermediate aldehyde. Within bacteria, the
myxochelins are produced from an aldehyde intermediate that
is produced by the R domain of MxcG in a reaction requiring
NAD(P)H.29 Lyngbyatoxin is synthesized by a 4-electron
reductive cleavage, requiring 2 equiv of NADPH, to yield an
alcohol.32 The nostocyclopeptide precursor is reductively cleaved
from the enzyme by NADPH, resulting in the formation of an
imine macrocycle.31 The fungal NRPS R domains have been
rigorously examined only in the case of Lys2 in fungal lysine
biosynthesis.28 This enzyme contains the ATR domain order,
activating and then reducing R-aminoadipic acid to the semi-
aldehyde with the aid of NADPH. Using in vivo expression,
evidence was obtained that the fungal PKS-NRPS proteins might
catalyze a Dieckmann condensation.36 Here, we obtained in vitro
evidence using purified proteins, confirming this reaction and
demonstrating that it takes place on the NRPS R domain.
This Dieckmann condensation was unanticipated, since bio-
informatic analysis predicts EqiS and related R domains to be
similar in structure and function to fungal and bacterial NRPS
R domains that are redox enzymes (Figure S8, Supporting
Information). The fungal hybrid PKS-NRPS R domains are
approximately 40% similar to characterized fungal Lys2 pro-
teins. The N-terminal NADPH binding site is well conserved;
however, homology essentially disappears in the C-terminal
substrate-binding region. On the basis of BLAST and conserved
domain searches, the tetramic acid R domains clearly fall into
the short-chain dehydrogenase/reductase (SDR) protein family,
which almost universally depends upon NAD(P)(H) cofactors.33
The rates of reaction were monitored by GC-FID using an
internal standard of limonene for increased accuracy. Using
(46) Fujii, I.; Watanabe, A.; Sankawa, U.; Ebizuka, Y. Chem. Biol. 2001,
8, 189–197.
(47) For example, see: Pasamontes, L.; Haiker, M.; Wyss, M.; Tessier,
M.; van Loon, A. P. Appl. EnViron. Microbiol. 1997, 63, 1696–1700.
9
11154 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Fungal Reductive Thioesterases
A R T I C L E S
Crystal structures for SDR proteins from different subfamilies
indicate that they are structurally similar, with all members
containing a Rossman fold and a catalytic triad composed of
Ser, Tyr, and Lys.48–51 These enzymes function in diverse redox
reactions on substrates including sugars, alcohols, steroids,
aromatic compounds, and amino acids.33 The catalytic triad and
NADPH binding site are all apparently found in the NRPS R
domains as well, based upon sequence alignment and analysis.
Critically, they are found in EqiS and its relatives.
here, at least EqiS R does not perform a reductive mechanism,
disproving the universal aldol hypothesis. It remains unclear if
the other pathways, especially those leading to more reduced
pyrrolidine-2-one derivatives, go via a Dieckmann condensation
or arise through a reduction on R. In recent years, the genes
responsible for several tetramic acids and more modified
relatives have been characterized. Sequence and phylogenetic
analysis show that the fungal hybrid PKS-NRPS R domains
form a distinct branch separate from other SDR proteins (data
not shown), and thus these R domains can be easily recognized
and differentiated from their reducing relatives. Unfortunately,
there are no readily observable sequence cues that enable the
differentiation of R domains leading to the tetramic acids versus
the relatively reduced pyrrolidine-2-ones. It is possible that some
of the R domains catalyze reduction, whereas others catalyze
Dieckmann condensation, or the tetramic acid itself could be
the universal intermediate en route to fungal pyrrolidine-2-ones.
Further experiments are needed to differentiate these possibili-
ties. Like fungal PKSs, similar R domains generate diverse
chemical structures by evolutionary mechanisms that may subtly
different than those found in bacterial pathways.
Recently, genes for two bacterial tetramic acids have been
sequenced.59,60 These hybrid PKS-NRPS genes terminate with
a thioesterase (TE) domain in place of the fungal R. TE domains
are known to catalyze acid-base chemistry, and they likely are
the catalysts of the Dieckmann condensation in the bacterial
group. Here, we have identified and SDR superfamily member,
EqiS, that also functions as a thioesterase rather than a reductase.
This represents a new class of thioesterase domain in NRPS
metabolism. These results will be useful in the genetic synthesis
of new tetramic acids, which are often potently bioactive, and
in the discovery of new potential pharmaceuticals by genome
mining.
Some of the best characterized SDR proteins are UDP-D-
galactose epimerase (UDP-epimerase) and ADP-L-glycero-D-
mannoheptose 6-epimerase (AGME). Both enzymes use NAD+
in a catalytic fashion to epimerize sugar hydroxyls.52 A catalytic
triad of Lys, Tyr, and Ser generate an acidic Tyr with a pKa of
6.08.43–45 This drives the general acid-base mechanism by
removing a proton from the sugar hydroxyl, facilitating oxidation
by NAD+ to give NADH. The resulting ketone is then reduced
from the opposite face by the same NADH, which regenerates
cofactor.43 This catalytic triad and the cofactor-binding pocket
are also found in NRPS R domains. Cofactor binds extremely
tightly to UDP-epimerase resulting in a coelution with protein,53
and it is used in a catalytic fashion. In contrast, EqiS R does
not coelute with cofactor and does not need cofactor to catalyze
tetramic acid formation. In the epimerases, the catalytic Tyr
participates in acid-base catalysis. Since this triad is conserved
in NRPS R domains, including EqiS R, it is possible that this
residue is involved in a general acid-base mechanism to
catalyze Dieckmann condensation, generating tetramic acids.
The family of tetramic acids and derivatives produced by
fungi is highly variable. Five-membered rings resulting from
these pathways have variable redox states and are often subject
to further modifications, generating an enormous chemical
diversity from similar starting points.54–58 An attractive hypoth-
esis would be the formation of an aldol intermediate, formed
by the R domain, which could be enzymatically modified to
the oxidized tetramic acids and other derivatives (Figure S10,
Supporting Information). On the basis of the work presented
Acknowledgment. We thank Chad Nelson and Krishna Par-
sawar from the University of Utah Mass Spectrometry Core Facility
for obtaining protein mass measurements. Gene Masters, Kianoush
Sadre-Bazzaz and Chris Hill helped with protein purification.
Michael Kay and Philip Moos provided instrumental access and
advice. Jim Muller helped with GC–MS. This work was funded
by the Herman Frasch Foundation for research in agricultural
chemistry and by an NIH Training Grant in Microbial Pathogenesis
(MP T32) to J.W.S.
(48) Thoden, J. B.; Frey, P. A.; Holden, H. M. Protein Sci. 1996, 5, 2149–
2161.
(49) Thoden, J. B.; Frey, P. A.; Holden, H. M. Biochemistry 1996, 35,
5137–5144.
(50) Thoden, J. B.; Hegeman, A. D.; Wesenberg, G.; Chapeau, M. C.; Frey,
P. A.; Holden, H. M. Biochemistry 1997, 36, 6294–6304.
(51) Thoden, J. B.; Wohlers, T. M.; Fridovich-Keil, J. L.; Holden, H. M.
Biochemistry 2000, 39, 5691–5701.
Supporting Information Available: Complete experimental
details for all procedures, NMR spectra and spectral data, mass
spectra and data, and pyrophosphate assay results. This material
is available free of charge via the Internet at http://pubs.acs.org.
(52) Glaser, L. Epimerases. In The Enzymes; Boyer, P. D., Ed. Academic
Press: New York, 1972; Vol. 4, pp 355-380..
(53) Dutta, S.; Maiti, N. R.; Bhattacharyya, D. Eur. J. Biochem. 1997, 244,
407–413.
(54) Wright, J. L.; Vining, L. C.; McInnes, A. G.; Smith, D. G.; Walter,
J. A. Can. J. Biochem. 1977, 55, 678–685.
JA803078Z
(55) Vederas, J. C.; Tamm, C. HelV. Chim. Acta 1976, 59, 558–566.
(56) Duspara, P.; Jenkins, S. I.; Hughes, D. W.; Harrison, P. H. M. Chem.
Commun. 1998, 2643–2644.
(59) Yu, F.; Zaleta-Rivera, K.; Zhu, X.; Huffman, J.; Millet, J. C.; Harris,
S. D.; Yuen, G.; Li, X. C.; Du, L. Antimicrob. Agents Chemother.
2007, 51, 64–72.
(57) Harrison, P. H. M.; Hughes, D. W.; Riddoch, R. W. Chem. Common.
1998, 273–274.
(60) Bihlmaier, C.; Welle, E.; Hofmann, C.; Welzel, K.; Vente, A.;
Breitling, E.; Muller, M.; Glaser, S.; Bechthold, A. Antimicrob. Agents
Chemother. 2006, 50, 2113–2121.
(58) Fujita, Y.; Oguri, H.; Oikawa, H. Tetrahedron Lett. 2005, 46, 5885–
5888.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11155