Biosynthesis of pipecolic acid by RapL, a lysine cyclodeaminase encoded in the rapamycin gene cluster

Article abstract of DOI:10.1021/ja0587603

Rapamycin, FK506, and FK520 are immunosuppressant macrolactone natural products comprised of predominantly polyketide-based core structures. A single nonproteinogenic pipecolic acid residue is installed into the scaffold by a nonribosomal peptide synthetase that also performs the subsequent macrocyclization step at the carbonyl group of this amino acid. It has been assumed that pipecolic acid is generated from lysine by the cyclodeaminases RapL/FkbL. Herein we report the heterologous overexpression and purification of RapL and validate its ability to convert L-lysine to L-pipecolic acid by a cyclodeamination reaction that involves redox catalysis. RapL also accepts L-ornithine as a substrate, albeit with a significantly reduced catalytic efficiency. Turnover is presumed to encompass a reversible oxidation at the α-amine, internal cyclization, and subsequent re-reduction of the cyclic Δ¹-piperideine-2-carboxylate intermediate. As isolated, RapL has about 0.17 equiv of tightly bound NAD⁺, suggesting that the enzyme is incompletely loaded when overproduced in E. coli. In the presence of exogenous NAD⁺, the initial rate is elevated 8-fold with a K _m of 2.3 μM for the cofactor, consistent with some release and rebinding of NAD⁺ during catalytic cycles. Through the use of isotopically labeled substrates, we have confirmed mechanistic details of the cyclodeaminase reaction, including loss of the α-amine and retention of the hydrogen atom at the α-carbon. In addition to the characterization of a critical enzyme in the biosynthesis of a medically important class of natural products, this work represents the first in vitro characterization of a lysine cyclodeaminase, a member of a unique group of enzymes which utilize the nicotinamide cofactor in a catalytic manner.

Full text of DOI:10.1021/ja0587603

Published on Web 03/01/2006
Biosynthesis of Pipecolic Acid by RapL, a Lysine
Cyclodeaminase Encoded in the Rapamycin Gene Cluster
Gregory J. Gatto, Jr.,† Michael T. Boyne, II,‡ Neil L. Kelleher,‡ and
Christopher T. Walsh*,†
Contribution from the Department of Biological Chemistry and Molecular Pharmacology,
HarVard Medical School, Boston, Massachusetts 02115, and Department of Chemistry,
UniVersity of Illinois at Urbana-Champaign, Urbana, Illinois 61801
Received December 26, 2005; E-mail: christopher_walsh@hms.harvard.edu
Abstract: Rapamycin, FK506, and FK520 are immunosuppressant macrolactone natural products comprised
of predominantly polyketide-based core structures. A single nonproteinogenic pipecolic acid residue is
installed into the scaffold by a nonribosomal peptide synthetase that also performs the subsequent
macrocyclization step at the carbonyl group of this amino acid. It has been assumed that pipecolic acid is
generated from lysine by the cyclodeaminases RapL/FkbL. Herein we report the heterologous overex-
pression and purification of RapL and validate its ability to convert L-lysine to L-pipecolic acid by a
cyclodeamination reaction that involves redox catalysis. RapL also accepts L-ornithine as a substrate, albeit
with a significantly reduced catalytic efficiency. Turnover is presumed to encompass a reversible oxidation
at the R-amine, internal cyclization, and subsequent re-reduction of the cyclic ∆1-piperideine-2-carboxylate
intermediate. As isolated, RapL has about 0.17 equiv of tightly bound NAD+, suggesting that the enzyme
is incompletely loaded when overproduced in E. coli. In the presence of exogenous NAD+, the initial rate
is elevated 8-fold with a Km of 2.3 µM for the cofactor, consistent with some release and rebinding of NAD+
during catalytic cycles. Through the use of isotopically labeled substrates, we have confirmed mechanistic
details of the cyclodeaminase reaction, including loss of the R-amine and retention of the hydrogen atom
at the R-carbon. In addition to the characterization of a critical enzyme in the biosynthesis of a medically
important class of natural products, this work represents the first in vitro characterization of a lysine
cyclodeaminase, a member of a unique group of enzymes which utilize the nicotinamide cofactor in a
catalytic manner.
Introduction
Rapamycin, FK506, and FK520 are macrolactone natural
products isolated from streptomycetes (Figure 1).1-3 Originally
identified as antifungal agents, they have since gained recogni-
tion as potent immunosuppressants and are currently in use or
under investigation for use in a number of therapeutic areas,
including transplant medicine, dermatology, and cardiology.4-7
While these compounds influence different signaling cascades
within the cell,8 they share a common initial step: interaction
† Harvard Medical School.
‡ University of Illinois at Urbana-Champaign.
(1) Vezina, C.; Kudelski, A.; Sehgal, S. N. J. Antibiot. 1975, 28, 721-726.
(2) Kino, T.; Hatanaka, H.; Miyata, S.; Inamura, N.; Nishiyama, M.; Yajima,
T.; Goto, T.; Okuhara, M.; Kohsaka, M.; Aoki, H. J. Antibiot. 1987, 40,
1256-1265.
Figure 1. Structures of rapamycin, FK506, and FK520. The pipecolyl
moiety is highlighted in red.
(3) Hatanaka, H.; Kino, T.; Miyata, S.; Inamura, N.; Kuroda, A.; Goto, T.;
Tanaka, H.; Okuhara, M. J. Antibiot. 1988, 41, 1592-1601.
(4) Kahan, B. D. Lancet 2000, 356, 194-202.
with a class of proteins known as the FK506-binding proteins
(FKBPs). FKBPs are members of the peptidyl-prolyl cis-trans-
isomerase (PPIase) family;9 the pipecolyl moieties of rapamycin
and FK506/FK520 serve as substrate mimics within the active
sites of these enzymes.10-12
(5) Reynolds, N. J.; Al-Daraji, W. I. Clin. Exp. Dermatol. 2002, 27, 555-
561.
(6) Antin, J. H.; Kim, H. T.; Cutler, C.; Ho, V. T.; Lee, S. J.; Miklos, D. B.;
Hochberg, E. P.; Wu, C. J.; Alyea, E. P.; Soiffer, R. J. Blood 2003, 102,
1601-1605.
(7) Moses, J. W.; Leon, M. B.; Popma, J. J.; Fitzgerald, P. J.; Holmes, D. R.;
O’Shaughnessy, C.; Caputo, R. P.; Kereiakes, D. J.; Williams, D. O.;
Teirstein, P. S.; Jaeger, J. L.; Kuntz, R. E. N. Engl. J. Med. 2003, 349,
1315-1323.
(9) Schiene-Fischer, C.; Yu, C. FEBS Lett. 2001, 495, 1-6.
(10) Albers, M. W.; Walsh, C. T.; Schreiber, S. L. J. Org. Chem. 1990, 55,
4984-4986.
(8) Schreiber, S. L.; Crabtree, G. R. HarVey Lect. 1995, 91, 99-114.
9
3838
J. AM. CHEM. SOC. 2006, 128, 3838-3847
10.1021/ja0587603 CCC: $33.50 © 2006 American Chemical Society

Lysine Cyclodeaminase Activity of RapL
A R T I C L E S
The biosynthesis of most of the core macrocycle of rapamycin
by its producer strain, Streptomyces hygroscopicus, proceeds
via a multimodular array of polyketide synthases (PKSs)
encoded by the rapA, rapB, and rapC genes.13 The nonprotei-
nogenic pipecolic acid residue is added to the completed
polyketide chain by the action of a single nonribosomal peptide
synthetase module, RapP, which is also presumed to be
responsible for the macrolactonization of the chain (Figure
2A).14-16 Incubation of the rapamycin producer strain with
labeled metabolites had implicated lysine as the source of the
pipecolate moiety,17 a proposal supported by the identification
of the rapL gene within the rapamycin gene cluster. On the
basis of its homology to the ornithine cyclodeaminases (OCDs)
(Figure 2B), RapL was predicted to act as a lysine cyclodeami-
nase, catalyzing the direct formation of L-pipecolic acid from
L-lysine (Figure 2C).18 Additional evidence for this role of RapL
came from the observation that chromosomal disruption of the
rapL gene yields a S. hygroscopicus strain which requires the
exogenous addition of L-pipecolate for rapamycin production.19
According to this mechanism, the R-amino group of L-ornithine
is first oxidized to an imine in an NAD+-dependent manner.
Attack of the δ-amine at the imino carbon yields a tetrahedral
intermediate which subsequently undergoes loss of ammonia
to form the cyclic imino acid, ∆1-pyrroline-2-carboxylic acid.
Reduction of the imino group by the previously formed NADH
yields the product, L-proline, and recycles the cofactor back to
oxidized NAD+.
In most enzyme-catalyzed reactions that require NAD(P)+
or NAD(P)H, the pyridine nucleotide acts as a cosubstrate. The
mechanisms shown in Figure 2B and 2C are unusual in that
the nicotinamide cofactor is recycled to its original oxidation
state. Catalytic usage of NAD+, previously referred to as a
“complex NAD+-dependent transformation”,30 has been pro-
posed in several other enzymes, including UDP-galactose
4-epimerase,31 S-adenosylhomocysteine hydrolase,32 and myo-
inositol-1-phosphate synthase.33 Characteristic of the mecha-
nisms of these enzymes is the oxidation of substrate by NAD+
to transiently activate a bond that is otherwise not reactive.
Typically, these enzymes are multimeric and bind cofactor very
tightly so as to prevent the dissociation of NADH during the
reaction cycle.
Because of its unusual mechanism and its importance in the
biosynthesis of a key functional component of the macrolide
immunosuppressant rapamycin, we sought to characterize the
in vitro enzymatic activity of RapL. We report here the
heterologous expression and purification of RapL from E. coli
and demonstration of its lysine cyclodeaminase activity. RapL
discriminates L-lysine from L-ornithine by approximately 100-
fold and requires exogenous NAD+ for full activity. Using
spectroscopic and chromatographic methods, we have deter-
mined that the purified enzyme is present in three forms:
NAD+-bound, NADH-bound, and cofactor-free. In addition, the
similarity between the mechanisms of RapL and the OCDs has
been confirmed through the use of isotopically labeled substrates
and mass spectrometry. Finally, we have characterized the
inhibitory properties of several compounds previously proposed
to interrupt RapL function in precursor-directed biosynthetic
experiments.34-36
In addition to RapL and its ortholog in the FK506- and
FK520-producer strains, FkbL, only two other lysine cy-
clodeaminases have been identified: VisC and TubZ, from the
biosynthetic pathways of virginiamycin S20 and tubulysin,21
respectively. While little is known about the in vitro activity of
lysine cyclodeaminases, there have been a number of published
reports on the identification and characterization of OCDs.
Costilow and co-workers first described such activity on enzyme
purified from Clostridium sporogenes.22-24 They identified
NAD+ as a key cofactor in the transformation of ornithine to
proline and noted that the R-amino group, as opposed to that at
the δ-position, is lost during the reaction.24 Further enzymatic
characterization was reported on OCDs identified in the Ti
plasmids of the tumor-inducing bacterium Agrobacterium
tumefaciens.25-27 In addition, the cocrystal structure of Pseudomo-
nas putida OCD with NADH and ornithine was recently
published.28 On the basis of these studies, a reaction mechanism
for OCD had been proposed29 and later revised (Figure 2B).28
(11) Fischer, S.; Michnick, S.; Karplus, M. Biochemistry 1993, 32, 13830-
13837.
(12) Van Duyne, G. D.; Standaert, R. F.; Karplus, P. A.; Schreiber, S. L.; Clardy,
J. J. Mol. Biol. 1993, 229, 105-124.
Materials and Methods
(13) Schwecke, T.; Aparicio, J. F.; Molna´r, I.; Ko¨nig, A.; Khaw, L. E.; Haydock,
S. F.; Oliynyk, M.; Caffrey, P.; Corte´s, J.; Lester, J. B.; Bo¨hm, G. A.;
Staunton, J.; Leadlay, P. F. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 7839-
7843.
Materials. Chemically competent TOP10 and BL21(DE3) E. coli
were obtained from Invitrogen; pET28a was purchased from Novagen.
All restriction endonucleases and T4 DNA ligase were obtained from
New England Biolabs. Synthetic DNA oligonucleotides were purchased
from Integrated DNA Technologies and used without further purifica-
tion. PCRs were run with Pfu Turbo DNA polymerase from Stratagene,
on a Stratagene RoboCycler Gradient 96. Pre-cast SDS-PAGE gels
and protein molecular weight markers were obtained from BioRad. Ni-
NTA chromatography resin was purchased from Qiagen. L-[U-14C]-
(14) Nielsen, J. B.; Hsu, M.-J.; Byrne, K. M.; Kaplan, L. Biochemistry 1991,
30, 5789-5796.
(15) Ko¨nig, A.; Schwecke, T.; Molna´r, I.; Bo¨hm, G. A.; Lowden, P. A. S.;
Staunton, J.; Leadlay, P. F. Eur. J. Biochem. 1997, 247, 526-534.
(16) Gatto, G. J., Jr.; McLoughlin, S. M.; Kelleher, N. L.; Walsh, C. T.
Biochemistry 2005, 44, 5993-6002.
(17) Paiva, N. L.; Demain, A. L.; Roberts, M. F. Enzyme Microb. Technol. 1993,
15, 581-585.
(18) Molna´r, I.; Aparicio, J. F.; Haydock, S. F.; Khaw, L. E.; Schwecke, T.;
Ko¨nig, A.; Staunton, J.; Leadlay, P. F. Gene 1996, 169, 1-7.
(19) Khaw, L. E.; Bohm, G. A.; Metcalfe, S.; Staunton, J.; Leadlay, P. F. J.
Bacteriol. 1998, 180, 809-814.
(30) Frey, P. A. Complex pyridine nucleotide-dependent transformations. In
Pyridine Nucleotide Coenzymes, Part B; Dolphin, D., Poulson, R.,
Avramovic, O., Eds.; Wiley-Interscience: New York, 1987; pp 461-511.
(31) Frey, P. A. FASEB J. 1996, 10, 461-470.
(20) Namwat, W.; Kamioka, Y.; Kinoshita, H.; Yamada, Y.; Nihira, T. Gene
2002, 286, 283-290.
(21) Sandmann, A.; Sasse, F.; Mu¨ller, R. Chem. Biol. 2004, 11, 1071-1079.
(22) Costilow, R. N.; Laycock, L. J. Biol. Chem. 1971, 246, 6655-6660.
(23) Muth, W. L.; Costilow, R. N. J. Biol. Chem. 1974, 249, 7457-7462.
(24) Muth, W. L.; Costilow, R. N. J. Biol. Chem. 1974, 249, 7463-7467.
(25) Sans, N.; Schro¨der, G.; Schro¨der, J. Eur. J. Biochem. 1987, 167, 81-87.
(26) Sans, N.; Schindler, U.; Schro¨der, J. Eur. J. Biochem. 1988, 173, 123-
130.
(32) Turner, M. A.; Yang, X.; Yin, D.; Kuczera, K.; Borchardt, R. T.; Howell,
P. L. Cell Biochem. Biophys. 2000, 33, 101-125.
(33) Jin, X.; Geiger, J. H. Acta Crystallogr. 2003, D59, 1154-1164.
(34) Kojima, I.; Demain, A. L. J. Ind. Microbiol. Biotechnol. 1998, 20, 309-
316.
(35) Graziani, E. I.; Ritacco, F. V.; Summers, M. Y.; Zabriskie, T. M.; Yu, K.;
Bernan, V. S.; Greenstein, M.; Carter, G. T. Org. Lett. 2003, 5, 2385-
2388.
(36) Ritacco, F. V.; Graziani, E. I.; Summers, M. Y.; Zabriskie, T. M.; Yu, K.;
Bernan, V. S.; Carter, G. T.; Greenstein, M. Appl. EnViron. Microbiol. 2005,
71, 1971-1976.
(27) Schindler, U.; Sans, N.; Schro¨der, J. J. Bacteriol. 1989, 171, 847-854.
(28) Goodman, J. L.; Wang, S.; Alam, S.; Ruzicka, F. J.; Frey, P. A.; Wedekind,
J. E. Biochemistry 2004, 43, 13883-13891.
(29) Walsh, C. Enzymatic Reaction Mechanisms; W. H. Freeman & Co.: New
York, 1979.
9
J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006 3839

A R T I C L E S
Gatto et al.
Figure 2. (A) Schematic of pipecolate (red) incorporation into the rapamycin polyketide chain. Through the formation of pipecolyl-AMP and pipecolyl-
S-enzyme intermediates (not shown), RapP catalyzes the formation of two bonds in the final product (circled in blue) that are critical for immunosuppressant
activity. Post-PKS/NRPS tailoring steps complete the biosynthesis. (B) The overall reaction (left) and proposed mechanism (right) of OCD.22,24,28 In the
mechanism, the R-NH2 (red) and δ-NH2 (blue) are highlighted. (C) The presumed overall reaction and mechanism for RapL, a lysine cyclodeaminase.
Lysine and L-[U-14C]ornithine were obtained from Amersham Bio-
sciences. D-[1-14C]Lysine was obtained from American Radiolabeled
Chemicals. BAS-IIIs phosphorimager plates were obtained from Fuji.
DNA sequencing and quantitative amino acid analysis were performed
at the Molecular Biology Core Facilities at the Dana-Farber Cancer
Institute. UV/vis spectra were measured on an Hewlett-Packard 8453
spectrophotometer. (R)-(-)-Nipecotic acid and (S)-(+)-nipecotic
acid were obtained from TCI America. L-[R-15N]Lysine and
L-[ꢀ-15N]lysine were purchased from Cambridge Isotope Laboratories.
DL-[2,3,3,4,4,5,5,6,6-d9]Lysine and DL-[2,3,3,4,4,5,5,6,6-d9]pipecolic
acid were obtained from C/D/N Isotopes. All other materials were
purchased from Sigma-Aldrich.
Overexpression and Purification of RapL. The open reading frame
of the rapL gene was amplified from a cosmid containing the rapamycin
biosynthetic cluster (kindly provided by Dr. Christopher Reeves, Kosan
Biosciences) using the following primers: 5′-GGG TTT CCA TAT
9
3840 J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006

Lysine Cyclodeaminase Activity of RapL
A R T I C L E S
GCA GAC CAA GGT TCT GTG CCA GCG-3′ (forward) and 5′-GGT
ATA CGA ATT CCT ACA GCG AGT ACG GAT CGA GGA CG-3′
(reverse). NdeI and EcoRI sites are underlined. The resulting PCR
product was digested with NdeI and EcoRI, purified by agarose gel
electrophoresis, and ligated into similarly digested pET28a to generate
the expression construct, pET28a-RapL(N-His), encoding RapL fused
to an N-terminal hexahistidine tag of the following sequence: +H3N-
MGSSHHHHHHSSGLVPRGSH. The identity of pET28a-RapL(N-His)
was confirmed by DNA sequencing.
The L and D isomers of pipecolic acid were separated in isocratic
conditions of 1 mM copper(II) sulfate in water.
Determination of NAD+ versus NADH Content of RapL. The
content of NAD+ and NADH bound to purified RapL was estimated
using an HPLC method based on that described by Yuan and
Borchardt.37 A quantity of 220 µL of an 85 µM solution of purified
RapL (19 nmol) was precipitated with 3 volumes of 95% ethanol. The
supernatant was flash frozen in liquid N2 and lyophilized for 16 h. The
residue was dissolved in 100 µL of water and analyzed by HPLC on
a Beckman-Coulter System Gold fitted with an Alltech Econosphere
C18 column (5 µm, 250 × 4.6 mm) with monitoring at 260 and 340
nm. NAD+ and NADH separated in isocratic conditions at 2.5%
methanol in 0.1 M sodium phosphate buffer (pH 7.0). Standard curves
were generated with authentic NAD+ and NADH so as to correlate
peak area with cofactor amount.
Characterization of RapL Inhibitors. The cyclodeaminase assay
was run as described above, with the concentration of L-[U-14C]lysine
fixed at 100 µM (10 nCi/µL) and the addition of either (R)-(-)-nipecotic
acid, (S)-(+)-nipecotic acid, or thiazolidine-2-carboxylic acid in
concentrations ranging from 10 nM to 10 mM in 10-fold increments.
Reactions were run in triplicate. Using Mathematica, reaction velocities
as a function of inhibitor concentration were fit to the following
expression:
The rapL expression plasmid was transformed into the E. coli strain
BL21(DE3) and grown in LB media supplemented with 30 µg/mL
kanamycin. Each liter of culture was inoculated with 10 mL of overnight
starter culture. Cultures were grown at 25 °C to an OD600 of 0.4-0.6,
induced with 60 µM IPTG, then incubated at 25 °C for an additional
16 h. Cells were harvested by centrifugation (20 min at 6000g),
resuspended in lysis buffer (400 mM NaCl, 25 mM Tris pH 8.0, 10%
glycerol), and lysed by two passages through a French press at 16 000
psi. Ni-NTA resin (1 mL per liter of culture) and 2 mM imidazole
were added to the clarified lysate and allowed to batch bind for 2 h at
4 °C. The resin was washed with 5 column volumes of 2 mM imidazole
in lysis buffer. Protein was eluted with a step gradient of increasing
imidazole concentrations in lysis buffer (5, 20, 40, 60, 200, 500 mM).
The RapL protein product typically eluted in the 40 and 60 mM
fractions, which were pooled and dialyzed at 4 °C in two steps: 4 h in
100 mM NaCl, 50 mM Tris pH 8.0, 1 mM EDTA, 10% glycerol;
followed by 4 h in 100 mM NaCl, 50 mM Tris pH 8.0, 1 mM TCEP,
10% glycerol. Concentrated protein solutions were stored at -80 °C.
Protein concentrations were determined by quantitative amino acid
analysis.
Vi
1
)
Vo
[I]
1 +
IC50
where Vi and Vo are the initial velocities in the presence and absence
of inhibitor, respectively, [I] is the inhibitor concentration, and IC50 is
the concentration of inhibitor required to yield a half-maximal
response.38
Characterization of RapL Cyclodeaminase activity. To determine
kinetic parameters with respect to lysine, L-[U-14C]lysine (10-500 µM;
4 nCi/µL) was incubated in a final volume of 25 µL with 1.7 µM RapL,
100 µM NAD+, 100 mM HEPES (pH 8.0), and 1 mg/mL BSA at 30
°C. Reactions were run in triplicate and initiated by the addition of
enzyme. At time points of 5, 15, and 30 min, 5 µL was withdrawn
from the reaction mixture, quenched in 10 µL of methanol, and stored
at -20 °C for at least 10 min. The quenched reactions were spun at
16 000g for 10 min, and 1 µL of the supernatant was spotted on a
cellulose thin layer chromatography (TLC) plate (Merck KGaA;
Darmstadt, Germany). The TLC plate was developed in 3:1 n-propanol:
30% NH4OH and exposed to a BAS-IIIs phosphorimager plate for 16
h at room temperature. Plates were read using a Typhoon 9400
phosphorimager (GE Healthcare), and the resulting image was analyzed
by ImageQuant v5.2 (Molecular Dynamics). When it was necessary to
run authentic L-pipecolic acid standard, unlabeled sample was used,
and that portion of the plate was stained with ninhydrin. A known
amount of radioactivity was overlaid onto the ninhydrin-stained
compound prior to autoradiography. The ratio of pipecolic acid (Rf )
0.44) to lysine (Rf ) 0.10) signals was used to calculate the amount of
product formed in each reaction. Initial velocity data were fitted to the
Michaelis-Menten equation in Mathematica (Wolfram Research) to
obtain estimates for kcat and Km.
Fmoc Derivatization of RapL Reaction Mixtures. One millmole
L-lysine was incubated at 30 °C with 10 µM RapL, 5 mM Tris (pH
8.0), 1 mg/mL BSA, and 100 µM NAD+ (total volume ) 75 µL) for
16 h. The reaction was quenched with two reaction volumes of
acetonitrile, stored at -20 °C for at least 10 min, and spun at 16 000g
for 10 min. Derivatization of the quenched reaction was carried out
using a procedure based on that described by Fabiani et al.39 One
hundred fifty microliters of the supernatant was withdrawn and added
to 50 µL of 200 mM borate (pH 8). Twenty microliters of 10 mM
9-fluorenylmethyl chloroformate was added, and the reaction was
allowed to proceed at room temperature for 30 min. To remove any
remaining derivatizing reagent, 40 µL of 100 mM 1-aminoadamantane
was added, and the mixture was kept at room temperature for an
additional 5 min. The reaction was centrifuged at 16 000g for 1 min,
then 150 µL of the supernatant was injected onto a Beckman-Coulter
System Gold HPLC system fitted with an Alltech Alltima C18 column
(5 µm; 250 × 4.6 mm) and equipped with an inline Jasco FP-2020
Plus fluorescence detector. Fmoc-derivatized compounds were separated
with a gradient of 30-100% acetonitrile in aqueous 0.1% trifluoroacetic
acid over 70 min and detected using both absorbance at 263 nm and
fluorescence emission at 630 nm (λex ) 263 nm).40 The peak
corresponding to Fmoc-pipecolic acid was collected, flash frozen in
liquid N2, lyophilized overnight, resuspended in 30% acetonitrile, and
analyzed by mass spectrometry.
For determination of kinetic parameters with respect to ornithine,
the same assay conditions were used as above, with the exception that
L-[U-14C]lysine was replaced with L-[U-14C]ornithine (0.2-1.3 mM;
25 nCi/µL). For determination of kinetic parameters with respect to
NAD+, the concentration of L-[U-14C]lysine was kept constant at 150
µM (10 nCi/µL), and the concentration of NAD+ was varied from 0.1
to 50 µM.
To determine the chirality of the pipecolic acid product, the RapL
reaction was run as above in the presence of 200 µM L-[U-14C]lysine
(20 nCi/µL). After 100 min, the reaction was quenched with 2 reaction
volumes of methanol. Ten microliters of this mixture was analyzed by
radio-HPLC on a Beckman-Coulter System Gold fitted with a Phe-
nomenex Chirex D-penicillamine column (5 µm, 250 × 4.6 mm) and
equipped with an inline IN/US â-RAM Model 3 radio flow detector.
Mass Spectrometric Analysis. Data were collected on a 7 T LTQ-
FTMS (Thermo-Finnigan) with direct infusion through an electrospray
source. Samples were diluted in 50 µL of water:methanol:acetic acid
(37) Yuan, C. S.; Borchardt, R. T. J. Biol. Chem. 1995, 270, 16140-16146.
(38) Copeland, R. A. Enzymes: A Practical Introduction to Structure, Mech-
anism, and Data Analysis, 2nd ed.; Wiley-VCH: New York, 2000.
(39) Fabiani, A.; Versari, A.; Parpinello, G. P.; Castellari, M.; Galassi, S. J.
Chromatogr. Sci. 2002, 40, 14-18.
(40) Bank, R. A.; Jansen, E. J.; Beekman, B.; te Koppele, J. M. Anal. Biochem.
1996, 240, 167-176.
9
J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006 3841

A R T I C L E S
Gatto et al.
either errors in the published sequence or genotypic variations
in the producer strains. Overexpression of His6-tagged RapL
from an E. coli BL21(DE3) strain bearing pET28a-RapL(NHis)
was achieved by induction with 60 µM IPTG and overnight
growth at 25 °C. The crude cell lysate was purified by Ni-
NTA chromatography, from which we obtained approximately
8 mg of >95% pure RapL per liter of culture (Figure 3).
Steady-State Kinetics and Substrate Specificity of the
RapL Cyclodeaminase Reaction. The activity of RapL was
measured by cellulose thin layer chromatography (TLC) with
14C-labeled L-lysine substrate using conditions similar to those
reported for A. tumefaciens OCD.25 When 200 µM L-lysine was
incubated with 1.7 µM RapL and 100 µM NAD+ at 30 °C, we
observed a new spot with a retention factor (Rf) consistent with
authentic pipecolic acid. The intensity of this spot increased in
a time-dependent manner (Figure 4A). By comparing the chiral
radio-HPLC elution profile of the RapL reaction with that of
cold L- and D-pipecolic acid standards, we are able to assign
the stereochemistry of the radiolabeled pipecolic acid product
as L (Figure 4B). Steady-state kinetic analysis of this reaction
yielded a Km for L-lysine of 46 ( 4 µM and a kcat of 0.61 (
0.02 min-1, resulting in a kcat/Km of 13 ( 1 mM-1 min-1 (Figure
4C). We further explored the substrate specificity of RapL by
replacing L-lysine with either L-ornithine or D-lysine. A linear
dependence of the initial velocity of proline formation upon
L-ornithine concentration was observed at substrate concentra-
tions up to 1.3 mM (Figure 4D), yielding a kcat/Km of 0.14 (
0.02 mM-1 min-1. No pipecolic acid formation was detected
when D-lysine was used in place of its enantiomer.
Figure 3. SDS-PAGE analysis (4-15% Tris-HCl) of RapL fused to an
N-terminal hexahistidine tag (expected molecular weight: 38.4 kDa).
(49:49:2, v:v:v), and FTMS scans were summed for 3 min at a flow
rate of 3 µL/min (∼180 scans). The data were analyzed using Qual
Browser (version 1.4 SR1) in Xcalibur to acquire accurate intact masses
(within 5 ppm).
Results
Cloning, Overexpression, and Purification of RapL. The
open reading frame of rapL was amplified from a cosmid
containing the entire rapamycin biosynthetic gene cluster (kindly
provided by Kosan Biosciences) and cloned into pET28a to yield
a construct designed to express RapL fused to an N-terminal
hexahistidine tag. Upon sequencing of this construct, seven
discrepancies with the published rapL nucleotide sequence13
were identified. Of these, two were silent (G285 f C and G357
f C), while the remaining five resulted in changes in the
predicted amino acid sequence: C148 f A (Pro50Thr), G250
f C (Glu84Gln), G305 f A (Gly102Asp), T379 f G
(Ser127Ala), and A385 f G (Thr129Ala). We confirmed that
these nucleotide substitutions were present in the original cosmid
and not introduced during the amplification reaction. Given the
solubility and catalytic activity exhibited by the expressed
protein product (vide infra), we presume these changes to reflect
Cofactor Requirements for Cyclodeaminase Activity. To
determine the cofactor specificity of RapL, initial velocities of
pipecolic acid formation were determined in the presence of
150 µM L-lysine, 1.7 µM RapL, and either 100 µM NAD+,
100 µM NADP+, 100 µM NADH, or no additional cofactor
(Figure 5A). The dependence of full cyclodeaminase activity
Figure 4. (A) Thin layer chromatography of L-lysine (lane 1) and L-pipecolic acid (lane 2) standards, and a time course of the RapL reaction at 0, 30, and
60 min (lanes 3, 4, and 5). (B) Chiral radio-HPLC of the RapL reaction run in the presence of 200 µM L-[U-14C]lysine, measured by 14C radioactivity (top
trace) and elution profile of cold DL-pipecolic acid standard using the same solvent conditions as the top trace, measured by UV absorbance at 254 nm
(bottom trace). (C) Plot of initial velocity versus L-lysine concentration, with the best-fit curve to the Michaelis-Menten equation. (D) Plot of initial velocity
versus L-ornithine concentration, with the best fit to a linear equation. For parts (C) and (D), error bars are drawn at one standard deviation.
9
3842 J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006

Lysine Cyclodeaminase Activity of RapL
A R T I C L E S
Figure 5. (A) Plot of time-dependent product formation, measured at 150 µM L-lysine in the presence of either 100 µM NAD+ (red), 100 µM NADP+
(blue), 100 µM NADH (black), or no added cofactor (green). Fitted initial velocities are reported in units of µM min-1. (B) Plot of initial velocity, corrected
for the rate of product formation in the absence of cofactor, versus NAD+ concentration, with the best-fit curve to the Michaelis-Menten equation. Error
bars for both (A) and (B) are drawn at one standard deviation.
Figure 6. (A) UV/vis absorption spectrum of RapL. (B) HPLC separation of NAD+ and NADH standards (left panel: red and blue traces, respectively),
and elution profile of supernatant from denatured RapL (right panel).
upon the addition of NAD+ enabled the determination of steady-
state kinetics with respect to this component. After correction
for the low level of pipecolate formation in the absence of added
cofactor, Michaelis-Menten behavior was observed when initial
velocity was measured in the presence of varying concentrations
of NAD+ (Figure 5B). From these data, an apparent Km of 2.3
( 0.4 µM for NAD+ was obtained.
Identity of Cofactors Present in Purified RapL. Given the
low level of pipecolate formation by RapL in the absence of
added cofactor (Figure 5A), we sought to ascertain the profile
of cofactors bound to the purified recombinant enzyme. First,
UV/visible spectroscopy of RapL revealed an absorption peak
near 340 nm (Figure 6A). This peak is consistent with the
presence of enzyme-bound NADH since oxidized nicotinamide
has no absorption at this wavelength. Second, an HPLC
method37 designed to separate NAD+ from NADH was applied
to 19 nanomoles of RapL denatured with ethanol. As shown in
Figure 6B, peaks with retention times consistent with authentic
NAD+ and NADH were observed. Preparation of standard
curves correlating cofactor amount with peak area enabled the
quantitation of cofactor present in this RapL sample: 0.17 mol
NAD+/mol RapL and 0.35 mol NADH/mol RapL.
9
J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006 3843

A R T I C L E S
Gatto et al.
Figure 7. LTQ-FTMS analysis of the (A) Fmoc-L-pipecolic acid (left) and Fmoc-DL-d9-pipecolic acid (right) standards, and the isolated Fmoc-pipecolic
acid product from the cyclodeaminase reactions run in the presence of (B) unlabeled L-lysine, (C) DL-[2,3,3,4,4,5,5,6,6-d9]lysine, (D) L-[R-15N]lysine, and
(E) L-[ꢀ-15N]lysine. For parts (B), (C), (D), and (E), the top panel shows the substrate and proposed structure of the derivatized product, while the bottom
panel shows the mass spectrum of the isolated Fmoc-pipecolic acid product. 15N atoms are colored red, and deuterium atoms are colored blue.
9
3844 J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006

Lysine Cyclodeaminase Activity of RapL
A R T I C L E S
Figure 8. (A) Chemical structures of (()-thiazolidine-2-carboxylic acid, (S)-(+)-nipecotic acid, and (R)-(-)-nipecotic acid. (B) Schematic of precursor-
directed biosynthesis of rapamycin analogues after manipulation of RapL by genetic or chemical methods. (C) Dose-response curves for T2CA (gray),
(S)-(+)-nipecotic acid (blue), and (R)-(-)-nipecotic acid (red). Error bars are drawn at one standard deviation.
Dissection of RapL Mechanism Using Isotopically Labeled
Substrates. On the basis of its homology to the ornithine
cyclodeaminases, RapL is presumed to act by a mechanism
similar to that shown in Figure 2B. To test this proposed
mechanism (Figure 2C), we utilized mass spectrometric analysis
of the pipecolic acid formed when RapL is incubated with
isotopically labeled lysine substrates. Treatment of the depro-
teinized assay mixture with 9-fluorenylmethyl chloroformate
(Fmoc-Cl) converted any available primary and secondary
amines to the corresponding Fmoc-carbamate. From this reaction
mixture, the Fmoc-pipecolic acid product was easily purified
by reversed-phase HPLC. Analysis of derivatized product
isolated in this manner on a 7T LTQ-FT mass spectrometer
yielded an [M + H]+ of 352.153 (Figure 7B), identical to the
value obtained for the commercially available Fmoc-L-pipecolic
acid standard (Figure 7A, left).
unlabeled substrate (Figure 7B). By contrast, when L-[ꢀ-15N]-
lysine was used in the same procedure, the observed [M + H]+
of the isolated Fmoc-pipecolic acid peak was 353.150, 1 Da
greater than that observed for the R-labeled and unlabeled
reactions (Figure 7E).
In addition, we sought to determine if the hydrogen atom at
the R-position of lysine is exchanged with solvent over the
course of the cyclodeaminase reaction. Lack of observable
exchange would be consistent with the participation of this
hydrogen in the redox chemistry with the nicotinamide cofactor.
Toward this end, we ran the RapL assay in the presence of DL-
[2,3,3,4,4,5,5,6,6-d9]lysine. Examination of the isolated Fmoc-
pipecolic acid peak from this reaction yielded an [M + H]+ of
361.210 (Figure 7C). This value is identical to that observed
for the Fmoc-[2,3,3,4,4,5,5,6,6-d9]pipecolic acid standard (Figure
7A, right), obtained by Fmoc derivatization of commercially
available perdeutero-pipecolic acid.
To determine the selectivity of amine loss from the substrate,
the above assay method was applied to lysine labeled with 15
N
Inhibition of RapL by Nipecotic Acids and Thiazolidine-
2-carboxylic Acid. It has been previously reported that treatment
of rapamycin producer strains of S. hygroscopicus with either
(()-thiazolidine-2-carboxylic acid (T2CA)34 or (()-nipecotic
acid35,36 (Figure 8A) enables the formation of rapamycin variants
at either the R- or the ꢀ-position. Incubation of L-[R-15N]lysine
with RapL, with subsequent derivatization, yielded Fmoc-
pipecolic acid with an [M + H]+ of 352.153 (Figure 7D),
identical to that isolated from the reaction of RapL with
9
J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006 3845

A R T I C L E S
Gatto et al.
when supplied with pipecolic acid derivatives (Figure 8B).
Presumably, these compounds act on RapL to block pipecolic
acid biosynthesis. To test this hypothesis directly, we determined
the ability of the two stereoisomers of nipecotic acid as well as
T2CA to inhibit the in vitro activity of RapL. Dose-response
curves were obtained by measurement of initial reaction velocity
in the presence of inhibitor concentrations spanning 7 orders
of magnitude from 10 nM to 10 mM (Figure 8C). The fitted
IC50 values for the stereoisomers of nipecotic acid were 680 (
70 µM and 4.8 ( 0.5 mM for the (S) and (R) configurations,
respectively. The fitted IC50 value for T2CA was 96 ( 16 µM.
place of its enantiomer. These data indicate a clear preference
for the expected L-lysine substrate, implying that RapL has
evolved a high degree of selectivity for alkyl chain length and
R-carbon stereochemistry.
The enzymatic logic to convert lysine into the nonproteino-
genic monomer pipecolic acid is akin to that used to convert
nonproteinogenic ornithine into proline: a cryptic redox process
that does not result in net change in the oxidation state between
substrate and products (Figure 2C). This strategy is part of a
reaction class, the so-called “complex NAD+-dependent trans-
formations”, in which the enzyme binds oxidized nicotinamide
so tightly that it typically does not dissociate at the end of each
catalytic cycle and, hence, functions as a prosthetic group. To
prevent dissociation of reduced cofactor and, as a result,
premature termination of the catalytic cycle, one presumes that
NADH should bind to the enzyme even more tightly than
NAD+. In parallel, upon oxidation of lysine, the nascent imino
acid must be held at the active site long enough to undergo an
uninterrupted series of transformations, including intramolecular
attack by the ꢀ-amine and loss of ammonia from the ensuing
tetrahedral adduct to arrive at the cyclic ∆1-piperideine-2-
carboxylate intermediate. Finally, this cyclic imine must be
reduced by the nascent NADH in the active site, regenerating
the starting oxidation state of the bound nicotinamide (NAD+)
and releasing the product.
For some enzymes that utilize NAD+ catalytically, including
mammalian S-adenosylhomocysteine hydrolase,44 UDP-galac-
tose 4-epimerase,45 ADP-â-L-glycero-D-manno-heptose 6-epi-
merase,46 and TDP-glucose 4,6-dehydratase,47 no addition of
cofactor is necessary for full enzymatic activity, a presumed
consequence of tight binding of the prosthetic group in the
enzyme active site. However, for other enzymes of this class,
including CDP-glucose 4,6-dehydratase,48 parasitic forms of
S-adenosylhomocysteine hydrolase,49,50 dehydroquinate syn-
thase,51 and the OCDs,22,26,27 full activity is dependent upon
added NAD+. RapL falls into this latter group, as removal of
exogenous NAD+ from the assay resulted in an 8-fold loss in
initial reaction velocity (Figure 5A). The apparent Km of NAD+
in the RapL reaction is 2.3 ( 0.4 µM, within a 5-fold range of
similar values reported for OCDs.22,26,27 To further examine the
identity of the cofactors bound to RapL, we determined the
NAD+/NADH content of the enzyme using spectroscopic and
HPLC-based methods. The absorption spectrum of RapL
revealed a peak near 340 nm, consistent with the presence of
bound NADH. Quantitation of cofactor content after denatur-
ation and chromatographic analysis indicated that RapL is
present in three forms: apo, NAD+-bound, and NADH-bound.
Given its catalytic mechanism, RapL can falter in at least
two ways during a reaction cycle. If the enzyme mistakenly
loses nicotinamide cofactor, either starting NAD+ or intermedi-
ate NADH, it will be stranded in a redox incompetent state.
Second, if any of the oxidized lysine intermediates diffuse out
Discussion
The core structure of the macrolide immunosuppressants
rapamycin, FK506, and FK520 is predominantly polyketide in
origin, with the exception of a single pipecolic acid residue.
This amino acyl group is installed by the nonribosomal peptide
synthetase RapP/FkbP and is crucial for the interaction of these
molecules with their targets, the FKBPs. The positioning of the
pipecolyl group as a substrate mimic within the PPIase active
site serves as a critical determinant for the subnanomolar affinity
of rapamycin for FKBP.41
Since the pipecolyl group is installed by an NRPS module,
one would anticipate that L-pipecolic acid is the substrate for
the activating adenylation step typically observed in this class
of megasynthetases. Indeed, this has been demonstrated in vitro
for both RapP and FkbP.14-16 Lysine has been implicated as
the precursor of this nonproteinogenic amino acid based on
feeding studies with isotopically labeled metabolites17 and the
identification of the rapL/fkbL genes within the biosynthetic
clusters for these compounds.18,42,43 RapL bears a close resem-
blance to ornithine cyclodeaminases, enzymes which directly
convert ornithine to proline. On the basis of this homology, it
has been proposed that RapL catalyzes the analogous conversion
of lysine to pipecolic acid. With an appreciation for the
importance of the pipecolyl group in the activity of these
compounds, the unique nature of the cyclodeaminase reaction,
and recent interest in the disruption of RapL for the engineered
biosynthesis of derivatized macrolides, we report here the
heterologous overexpression and purification of RapL and the
first in vitro characterization of its lysine cyclodeaminase
activity.
Using a radiographic TLC-based assay, we observed the time-
dependent conversion of L-lysine to L-pipecolic acid in the
presence of RapL and NAD+. In addition, we confirmed the
correct stereochemistry of the product by chiral radio-HPLC.
RapL exhibited Michaelis-Menten kinetic behavior with respect
to L-lysine, yielding a Km of 46 ( 4 µM, 5- to 100-fold tighter
than that observed for OCDs from C. sporogenes23 and A.
tumefaciens.26,27 Cyclization of L-ornithine by RapL was also
observed, albeit at a significantly reduced rate when compared
to the natural substrate. Although saturation kinetics were not
observed for L-ornithine, the catalytic efficiency (approximated
by kcat/Km) was reduced 100-fold relative to L-lysine. We did
not observe any product formation when D-lysine was added in
(44) Palmer, J. L.; Abeles, R. H. J. Biol. Chem. 1979, 254, 1217-1226.
(45) Wilson, D. B.; Hogness, D. S. J. Biol. Chem. 1964, 239, 2469-2481.
(46) Read, J. A.; Ahmed, R. A.; Morrison, J. P.; Coleman, W. G., Jr.; Tanner,
M. E. J. Am. Chem. Soc. 2004, 126, 8878-8879.
(47) Zarkowsky, H.; Glaser, L. J. Biol. Chem. 1969, 244, 4750-4756.
(48) He, X.; Thorson, J. S.; Liu, H.-w. Biochemistry 1996, 35, 4721-4731.
(49) Parker, N. B.; Yang, X.; Hanke, J.; Mason, K. A.; Schowen, R. L.;
Borchardt, R. T.; Yin, D. H. Exp. Parasitol. 2003, 105, 149-158.
(50) Yang, X.; Borchardt, R. T. Arch. Biochem. Biophys. 2000, 383, 272-280.
(51) Srinivasan, P. R.; Rothschild, J.; Sprinson, D. B. J. Biol. Chem. 1963, 238,
3176-3182.
(41) Bierer, B. E.; Mattila, P. S.; Standaert, R. F.; Herzenberg, L. A.; Burakoff,
S. J.; Crabtree, G.; Schreiber, S. L. Proc. Natl. Acad. Sci. U.S.A. 1990, 87,
9231-9235.
(42) Motamedi, H.; Shafiee, A. Eur. J. Biochem. 1998, 256, 528-534.
(43) Wu, K.; Chung, L.; Revill, W. P.; Katz, L.; Reeves, C. D. Gene 2000,
251, 81-90.
9
3846 J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006

Lysine Cyclodeaminase Activity of RapL
A R T I C L E S
of the active site before re-reduction, then the enzyme will be
left in the NADH-bound state, incompetent for subsequent
catalytic cycles. For RapL overproduced from E. coli, both
deactivating pathways may be operant. As isolated, RapL has
only 0.5 equiv of nicotinamide bound, only one-third of which
is bound to NAD+. While exogenous addition of NAD+ can
stimulate turnover, it is not yet clear what fraction of bound
NADH may be displaced and, hence, what enzyme fraction may
be converted back to competency by added NAD+. Presumably,
the exogenous addition of ∆1-piperideine-2-carboxylate would
regenerate all the NADH-occupied RapL sites.
These two problems are not unique to RapL but show up to
different extents in other members of this class of enzymes
containing tightly bound, nondissociable NAD+ with redox
reactions occult in the overall stoichiometry. For example, a
similar mixture of apo and holo forms has been reported for
CDP-glucose 4,6-hydratase.48 In addition, Frey and co-workers
demonstrated that full activity of UDP-galactose 4-epimerase
could only be achieved after mild denaturation of the protein
and replacement of any bound NADH with NAD+.52 Further
dissection of these issues for RapL would be greatly aided by
the preparation of enzyme entirely in its apo form.
Using isotopically labeled substrates, postreaction Fmoc
derivatization, reversed-phase HPLC purification, and LTQ-FT
mass spectrometry, we have examined two aspects of this
mechanism: (a) loss of the amine at the R-position and (b)
retention of the hydrogen atom on the R-carbon. When RapL
was incubated in the presence of lysine labeled with 15N at the
R-position, no mass shift was observed in the purified pipecolic
acid product. However, when the label was present at the
ꢀ-position, the mass of the corresponding product was greater
than the unlabeled standard by 1 Da. These data clearly indicate
that the ꢀ-NH2 is retained during the course of the cyclization,
while the R-NH2 is likely lost as ammonia, consistent with the
observations of P. putida OCD by Muth and Costilow.24 To
determine if hydrogen exchange with solvent at the R-carbon
occurs, we incubated RapL with a form of lysine in which each
carbon-bound hydrogen has been replaced with deuterium. If
hydrogen exchange at the R-position (or at any other carbon)
of lysine does occur, one would expect the mass of the pipecolic
acid product to be lighter by at least 1 Da. However, for this
experiment, the observed mass of the pipecolic acid product is
consistent with retention of all nine deuterium atoms. This
observation strongly suggests that the deuterium abstracted from
the R-position of the substrate and transferred to NAD+ is the
same as that returned to the cyclic imino acid intermediate from
the transiently formed 4-[2H]NADH.
There has been much interest in the generation of biologically
active variants of rapamycin and FK506/FK520. One general
strategy toward the generation of such compounds has focused
on the alteration of the biosynthetic pathway by genetic53 or
chemical54 methods such that a requisite intermediate is no
longer produced. The bacterial strain is then supplemented with
analogues of the missing intermediate that can still be utilized
by the biosynthetic enzymes. This technique, precursor-directed
biosynthesis, has been applied to RapL both by genetic
disruption19 and by the treatment of producer strain with
presumed cyclodeaminase inhibitors: (()-nipecotic acid35,36 and
(()-thiazolidine-2-carboxylic acid (T2CA).34 To determine the
in vitro effectiveness of these compounds as RapL inhibitors,
we measured dose-response curves for the two stereoisomers
of nipecotic acid and T2CA over a broad range of inhibitor
concentrations. T2CA exhibited the lowest IC50 at 96 µM. As
one would expect based on the stereospecificity of the RapL
reaction, the two stereoisomers of nipecotic acid displayed
different activities, with an IC50 for (S)-(+)-nipecotic acid
approximately 7-fold lower than that of its enantiomer (680 µM
versus 4.8 mM). While all three compounds did impair
cyclodeaminase activity to varying degrees, their relatively
modest IC50 values leave open the possibility that more effective
RapL inhibitors have yet to be characterized.
Acknowledgment. This work was supported by grants from
the National Institutes of Health (GM020011 (to C.T.W.),
GM067725 (to N.L.K.), F32GM069169 (to G.J.G.), and
T32GM07283 (to M.T.B.)). We thank Christopher D. Reeves
at Kosan Biosciences for providing the rapL-containing cosmid,
and Annaleise Howard-Jones and Ellen Yeh for helpful dis-
cussions and critical reading of the manuscript.
JA0587603
(53) Weist, S.; Sussmuth, R. D. Appl. Microbiol. Biotechnol. 2005, 68, 141-
150.
(52) Liu, Y.; Vanhooke, J. L.; Frey, P. A. Biochemistry 1996, 35, 7615-7620.
(54) Thiericke, R.; Rohr, J. Nat. Prod. Rep. 1993, 10, 265-289.
9
J. AM. CHEM. SOC. VOL. 128, NO. 11, 2006 3847