Three unusual reactions mediate carbapenem and carbapenam biosynthesis [1]

Full text of DOI:10.1021/ja001723l

9296
J. Am. Chem. Soc. 2000, 122, 9296-9297
Communications to the Editor
lanic acid.12,13,14 These observations imply an evolutionary
relationship between clavam formation in Streptomyces cla-
Vuligerus and carbapenam/em biosynthesis. The protein encoded
by carB reveals similarities to enzymes that interact with acylCoA
derivatives, and carDE appear to give rise correspondingly to a
proline oxidase and a likely allied ferredoxin.9,15,16 In this work
we report the first functional analysis of the carbapenem gene
cluster to delineate the sequence of biochemical events to 2, 3,
and 4 (Scheme 1).
Three Unusual Reactions Mediate Carbapenem and
Carbapenam Biosynthesis
Rongfeng Li, Anthony Stapon, Joanne T. Blanchfield,‡ and
Craig A. Townsend*
Department of Chemistry, The Johns Hopkins UniVersity
3400 North Charles Street, Baltimore, Maryland 21218
ReceiVed May 22, 2000
Scheme 1
Members of the carbapenem family are important among the
â-lactam antibiotics for both their broad spectrum of antibiotic
activity and their relative resistance to most clinically encountered
â-lactamases.1 Since the isolation of thienamycin,2 more than 40
structurally related carbapenems have been identified. The
simplest among these is (5R)-carbapen-2-em-3-carboxylic acid
(4).3 It co-occurs in the Gram-negative bacteria Serratia
marcescens and Erwinia carotoVora with two saturated diaster-
eomers, 2 and 3.4,5 Unlike 4, these carbapenams have no
antibacterial activity.
Incorporation experiments have established that the bicyclic
nuclei of both thienamycin and 2-4 are derived from acetate
(â-lactam carbons) and glutamate (pyrrolidine ring),4,6 origins
clearly distinct from those to penicillin, cephalosporin,6,7 and
clavulanic acid.8 Recently, the gene cluster responsible for
carbapenem production has been identified in E. carotoVora. Of
the nine open reading frames (ORFs), five, carA-E, are thought
to be required for the production of 4, while carFG are involved
in a poorly understood self-resistance mechanism.9,10 Transforma-
tion of Escherichia coli with the entire cluster confered the ability
to produce 4, but genetic disruption of carD or carE resulted in
substantial loss of antibiotic production. Mutation of carA, carB,
or carC gave only a carbapenem-negative phenotype.10
Previous biochemical and genetic evidence has suggested
separately evolved biosynthetic pathways to the four known
classes of â-lactam antibiotics.11 The gene products of carA and
carC, however, show similarities to two enzymes, â-lactam
synthetase (â-LS) and clavaminate synthase (CS), respectively,
whose roles are well-characterized in the biosynthesis of clavu-
The carA, carB, and carC genes were cloned from E.
carotoVora genomic DNA by PCR amplification, and each was
inserted into the E. coli expression vector pET24a (Novagen).
The proteins were individually expressed in BL21(DE3)pLysS
and found to be soluble (Figure 1B, lanes 2, 4, and 6). The
construction of plasmids for their coexpression is shown in Figure
1A. The resulting plasmids, pET24a/carAB and pET24a/carABC,
contained a single T7 RNA polymerase promoter and a single
T7 terminator and each gene was preceded by an E. coli Shine-
Dalgarno sequence for ribosome binding. These plasmids were
used individually to transform BL21(DE3)pLysS, and the resulting
transformants were examined for overexpression by SDS-PAGE.
As shown in lanes 8 and 10 in Figure 1B, coexpression had no
adverse effect on the levels of soluble recombinant proteins in
either case.
The induction of â-lactamases in Bacillus lichenformis (ATCC
14580) by â-lactam antibiotics can be sensitively detected in a
colorimetric assay with nitrocefin.17 Fermentation of BL21
(pET24a/carABC) under standard conditions in LB medium
showed that the coexpression of these three genes relative to a
control led to the low level production of a nitrocefin-positive
compound. The titer of this compound, presumed to be 4, could
be enhanced to a level comparable to that of S. marcescens itself
when a modified medium was used.4 Once growth had reached
A600 ) 0.65, the temperature was reduced to 28 °C and IPTG
was added. After 5 h, the supernatant was extracted, and the
products were derivatized as their p-nitrobenzyl (PNB) esters. The
reaction residue was partially purified by silica gel chromatog-
raphy to remove excess PNB bromide.4 The crude PNB esters
* Address correspondence to this author. E-mail: Townsend@
jhunix.hcf.jhu.edu.
‡ Present address: The University of Queensland, Department of Pharmacy,
Brisbane, Australia.
(1) Neu, H. C. Curr. Opin. Infect. Dis. 1994, 7, 3-10.
(2) Kahan, J. S.; Kahan, F. M.; Goegelman, R.; Currie, S. A.; Jackson,
M.; Stapley, E. O.; Miller, A. K.; Hendlin, D.; Mochales, S.; Hernandez, S.;
Woodruff, H. B.; Birnbaum, J. J. Antibiot. 1979, 32, 1-12.
(3) Parker, W. L.; Rathnum, M. L.; Wells, J. S.; Trejo, W. H.; Principe, P.
A.; Sykes, R. B. J. Antibiot. 1982, 35, 653-660.
(4) Bycroft, B. W.; Maslen, C.; Box, S. J.; Brown, A.; Tyler, J. W. J.
Antibiot. 1988, 41, 1231-1242.
(5) Bycroft, B. W.; Chhabra, S. R. J. Chem. Soc., Chem. Commun. 1989,
1989, 423-425.
(6) Williamson, J. M.; Inamine, E.; Wilson, K. E.; Douglas, A. W.; Liesch,
J. M.; Albers-Schonberg, G. J. Biol. Chem. 1985, 260, 4637-4647.
(7) Baldwin, J. E.; Bradley, M. Chem. ReV. 1990, 90, 1079-1088.
(8) Li, R.; Khaleeli, N.; Townsend, C. A. J. Bacteriol. 2000, 182, 4087-
4095.
1
were further purified by HPLC, whereupon H NMR analysis
(12) Bachmann, B. O.; Li, R.; Townsend, C. A. Proc. Natl. Acad. Sci.
U.S.A. 1998, 95, 9082-9086.
(13) McNaughton, H. J.; Thirkettle, J. E.; Zhang, Z.; Schofield, C. J.; Jensen,
S. E.; Barton, B.; Greaves, P. J. Chem. Soc., Chem. Commun. 1998, 1998,
2325-2326.
(9) McGowan, S. J.; Sebaihia, M.; Porter, L. E.; Stewart, G. S. A. B.;
Williams, P.; Bycroft, B. W.; Salmond, G. P. C. Mol. Microbiol. 1996, 22,
415-426.
(14) Marsh, E. N.; Chang, M. D.-T.; Townsend, C. A. Biochemistry 1992,
31, 12648-12657.
(15) Margolin, W.; Bramhill, D.; Long, S. R. J. Bacteriol. 1995, 177, 2892-
2900.
(10) McGowan, S. J.; Sebaihia, M.; Leary, S. O.; Hardie, K. R.; Williams,
P.; Stewart, G. S. A. B.; Bycroft, B. W.; Salmond, G. P. C. Mol. Microbiol.
1997, 26, 545-556.
(16) Boynton, Z. L.; Bennet, G. N.; Rudolph, F. B. J. Bacteriol. 1996,
178, 3015-3024.
(11) Townsend, C. A. Biochem. Soc. Trans. 1993, 21, 208-213.
(17) Sykes, R. B.; Wells, J. S. J. Antibiot. 1985, 38, 119-121.
10.1021/ja001723l CCC: $19.00 © 2000 American Chemical Society
Published on Web 09/07/2000

Communications to the Editor
J. Am. Chem. Soc., Vol. 122, No. 38, 2000 9297
5 with PNB bromoacetate provided the vinylogous amide 6 in
86% yield. Subsequent reduction with sodium cyanoborohydride/
acetic acid yielded the protected amino diacids 7 and 8 in a 1:2
ratio, which were separable by silica gel chromatography.
Catalytic hydrogenation, followed by treatment with TFA, yielded
the two diastereomeric amino diacids 1 and 9 in pure form.
Scheme 2a
a Reagents and conditions: (a) PNB bromoacetate, CH3CN, 40 h. (b)
triphenylphosphine, triethylamine, CH2Cl2, 20 h. (c) sodium cyanoboro-
hydride, acetic acid, CH3CN. (d) Pd(OH)2/C, H2, EtOH. (e) trifluoroacetic
acid.
The product of CarB was identified from a fermentation of
BL21 (pET24a/carB). The supernatant was lyophilized, suspended
in 0.2 N acetic acid, and passed through a cation exchange column
(Bond Elut SCX, Varian) with aqueous pyridine, followed by
Figure 1. (A) Construction of plasmids pET24a/carAB and pET24a/
carABC. The solid boxes represent ribosome binding sites. H: HindIII;
N: NdeI; No: NotI; Xb: XbaI; Xh: XhoI. (B). SDS-PAGE gel showing
soluble proteins. Lane 1, 3, 5, 7, and 9: uninduced. Lane 2, 4, 6, 8, and
10: induced. Lane 11, molecular weight markers. CarA, CarB, and CarC
are indicated by arrows. The solid bars represent ribosomal binding sites
1
HPLC purification. The H and 13C NMR spectra of partially
purified 1 were coincident with spectra taken of authentic 1
obtained by independent total synthesis (Scheme 2). Further
analysis by selective DPFGSE-TOCSY 1H NMR spectroscopy21
demonstrated that the product of the fermentation was unambigu-
ously a single diastereomer identical to 1.
showed that BL21 (pET24a/carABC) gave 2, 3, and 4, identical
to the similarly esterified products isolated from S. marcescens.3,4
These data indicated that carABC contain all of the genetic
information necessary for the assembly of these compounds from
primary metabolic precursors available in E. coli.
Employing the conditions optimized for the fermentation of
BL21 (pET24a/carABC) and subsequent isolation, we turned our
attention to determining the possible products generated from
BL21 (pET24a/carAB). Unexpectedly, 1H NMR analysis showed
that 2, the carbapenam of opposite bridgehead stereochemistry
compared to 4, was the sole â-lactam product of this recombinant
strain. In accord with the proposed oxidative role for CarC,9,18
these findings suggested that this enzyme is responsible for the
oxidation of carbapenam 3 to carbapenem 4. To our surprise,
however, the results also indicated that CarC acts as an epimerase
inverting the configuration of the ring fusion.
In conclusion, we take these findings to support a pathway
(Scheme 1) in which 1 is stereospecifically formed by CarB from
primary metabolic precursors under the proper choice of fermen-
tation conditions. CarA, homologous to the â-lactam synthetase
active in clavulanic acid biosynthesis,12,13 is then proposed to close
this first dedicated intermediate to the significantly more strained
carbapenam 2. Of 2 and 3, it is noteworthy that the thermody-
namically favored exo-diastereomer 2 is formed in these two steps
limited by the free energy accessible from the hydrolysis of ATP
presumed to be coupled in this process22,23 and the restriction of
an L-amino acid precursor. In the most unexpected transformation
of this sequence, CarC, a probable non-heme iron oxygenase,
carries out both the anticipated desaturation to the carbapenem
nucleus of 4, and the thermodynamically unfavorable epimeriza-
tion of 2 to 3. The mechanistic and energetic implications of these
observations are the subject of further investigation.
On the basis of this result and the potential role of CarA in
closure of the â-lactam ring, we synthesized the proposed product
of CarB, amino diacid 1. As shown in Scheme 2, starting from
L-glutamate, thiolactam 5 was prepared according to a previously
published procedure in four steps.19 Eschenmoser coupling20 of
Acknowledgment. We thank Professor John W. Blunt (University
of Cantebury, NZ) and Dr. Eric W. Schmidt for their help in carrying
out the selective TOCSY NMR experiments. We are grateful to Brian
O. Bachmann for informed discussions, and to the National Institutes of
Health (AI14937) for financial support.
(18) McGowan, S. J.; Bycroft, B. W.; Salmond, G. P. C. Trends Microbiol.
1998, 6, 203-208.
JA001723L
(19) Petersen, J. S.; Fels, G.; Rapoport, H. J. Am. Chem. Soc. 1984, 106,
4539-4547.
(21) Sharman, G. J. Chem. Commun. 1999, 1999, 1319-1320.
(22) Frey, P. A.; Arabshahi, A. Biochemistry 1995, 34, 11307-11310.
(23) Bachmann, B. O.; Townsend, C. A. Biochemistry, manuscript submit-
ted.
(20) Roth, M.; Dubs, P.; Gotschi, E.; Eschenmoser, A. HelV. Chim. Acta
1970, 54, 710-734.