A prodrug resistance mechanism is involved in colibactin biosynthesis and cytotoxicity

Article abstract of DOI:10.1021/ja312154m

Commensal Escherichia coli residing in the human gut produce colibactin, a small-molecule genotoxin of unknown structure that has been implicated in the development of colon cancer. Colibactin biosynthesis is hypothesized to involve a prodrug resistance strategy that entails initiation of biosynthesis via construction of an N-terminal prodrug scaffold and late-stage cleavage of this structural motif during product export. Here we describe the biochemical characterization of the prodrug synthesis, elongation, and cleavage enzymes from the colibactin biosynthetic pathway. We show that nonribosomal peptide synthetases ClbN and ClbB assemble and process an N-acyl-d-asparagine prodrug scaffold that serves as a substrate for the periplasmic d-amino peptidase ClbP. In addition to affording information about structural features of colibactin, this work reveals the biosynthetic logic underlying the prodrug resistance strategy and suggests that cytotoxicity requires amide bond cleavage.

Full text of DOI:10.1021/ja312154m

Communication
pubs.acs.org/JACS
A Prodrug Resistance Mechanism Is Involved in Colibactin
Biosynthesis and Cytotoxicity
Carolyn A. Brotherton and Emily P. Balskus*
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, United States
S Supporting Information
*
ABSTRACT: Commensal Escherichia coli residing in the
human gut produce colibactin, a small-molecule genotoxin
of unknown structure that has been implicated in the
development of colon cancer. Colibactin biosynthesis is
hypothesized to involve a prodrug resistance strategy that
entails initiation of biosynthesis via construction of an N-
terminal prodrug scaffold and late-stage cleavage of this
structural motif during product export. Here we describe
the biochemical characterization of the prodrug synthesis,
elongation, and cleavage enzymes from the colibactin
biosynthetic pathway. We show that nonribosomal peptide
synthetases ClbN and ClbB assemble and process an N-
acyl-D-asparagine prodrug scaffold that serves as a substrate
for the periplasmic D-amino peptidase ClbP. In addition to
affording information about structural features of col-
ibactin, this work reveals the biosynthetic logic underlying
the prodrug resistance strategy and suggests that
cytotoxicity requires amide bond cleavage.
olibactin is a small-molecule genotoxin produced by certain
Figure 1. Colibactin biosynthesis may involve a prodrug resistance
mechanism. (A) The pks gene cluster and other NRPS−PKS-containing
gene clusters harboring putative prodrug scaffold synthesis and cleavage
machinery. Genes encoding NRPSs predicted to synthesize prodrug
scaffolds (pink) and peptidases that may cleave prodrugs (green) are
highlighted. (B) Postulated biochemical logic underlying the prodrug
resistance mechanism in the zwittermicin, xenocoumacin, and colibactin
biosynthetic pathways.
C
strains of Escherichia coli, including commensals found in
1
the human gut. This secondary metabolite is biosynthesized by a
hybrid nonribosomal peptide synthetase−polyketide synthase
(
NRPS−PKS) assembly line encoded within the pks genomic
+
island (Figure 1A). E. coli harboring the pks island (pks ) damage
1
2
the DNA of eukaryotic cells both in vitro and in vivo, inducing
double strand breaks and disrupting the cell cycle. This biological
activity has raised concerns that colibactin production by gut
microbes might contribute to human disease. This hypothesis is
supported by recent work demonstrating that pks+ E. coli
promote tumorigenesis in a mouse model of colitis-associated
line components and a tailoring enzyme encoded by the cluster
closely resemble genes involved an antibiotic resistance
mechanism that had been first postulated to occur in
3
5
colorectal cancer. Understanding the mechanism and con-
zwittermicin biosynthesis and was recently shown using in
sequences of colibactin-induced DNA damage is critical, since an
vivo experiments to be operative in the assembly of
xenocoumacin. This resistance mechanism, which is known as
6
estimated 21% of humans harbor colibactin-synthesizing E. coli
3,4
within their intestinal microbiota. However, the structure of
colibactin is currently unknown, as isolation attempts to date
have been unsuccessful. Here we report the biochemical
characterization of ClbN and ClbB, NRPS modules involved in
the initiation of colibactin biosynthesis, and ClbP, a peptidase
responsible for tailoring of the natural product scaffold during
export. This work reveals structural features of colibactin and
provides molecular evidence for the involvement of a prodrug
strategy in its production and cytotoxicity.
the prodrug strategy, is hypothesized to be conserved and
widespread, as homologues of the enzymatic machinery
implicated in this process are encoded in many otherwise
6
,7a
unrelated secondary metabolite biosynthetic gene clusters.
The prodrug resistance strategy involves installation of a
structural motif at the N-terminus of a secondary metabolite that
is removed in the final stages of biosynthesis (Figure 1B). This
prodrug scaffold, which consists of an N-acyl-D-asparagine
residue, is thought to be assembled by an initiating NRPS
We initiated our studies of colibactin biosynthesis by
performing a detailed annotation and bioinformatic analysis of
the genes in the pks island. We recognized that certain assembly
Received: December 12, 2012
©
XXXX American Chemical Society
A
dx.doi.org/10.1021/ja312154m | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Communication
module containing condensation (C), adenylation (A), thio-
lation (T), and epimerization (E) domains and extended by an
enzymatic assembly line. After release of the nascent natural
product from the assembly line, a periplasmic D-amino peptidase
removes the prodrug scaffold to give the active molecule. This
process has been previously characterized in xenocoumacin
biosynthesis; deletion of the gene encoding the peptidase XcnG
Multiple sequence alignments revealed that this domain
possesses amino acid motifs characteristic of starter C domains
10
involved in lipoinitiation (Figure S32). Starter C domains
employ multiple strategies for N-acylation, including the use of
11
fatty acyl thioesters bound to trans-acting T domains and the
12
utilization of fatty acyl-CoA thioesters. We hypothesized that
ClbN would accept fatty acyl-CoA substrates because the pks
island does not encode any homologues of fatty acid-activating
enzymes. We were also interested in the specificity of this
reaction, as the isolation of multiple prexenocoumacins
suggested the potential for promiscuity in the initial N-acylation
abolished antibiotic production and generated uncleaved
6
“
prexenocoumacins” that lacked activity. Although there is in
vivo evidence for the biochemical logic underlying this resistance
mechanism, there has been no comprehensive in vitro character-
8
6
ization of the prodrug synthesis and cleavage enzymes. We
event. To address these questions, we examined the activity of
decided to study the activities of the putative initiating NRPS and
D-amino peptidase from the pks cluster, anticipating that an
understanding of the prodrug resistance mechanism in colibactin
biosynthesis not only would illuminate important structural
features of the natural product but also could reveal strategies for
preventing cytotoxin production.
ClbN using a liquid chromatography−mass spectrometry (LC−
MS)-based assay. Incubation of holo-ClbN with L-asparagine,
ATP, and individual even-numbered straight-chain saturated
fatty acyl-CoA substrates revealed that the enzyme. Accepted
fatty acyl-CoAs with chain lengths ranging from 6 to 14 carbons
(Figure S6). We further examined substrate specificity using a
competition assay, incubating ClbN with a mixture of C to C
We first examined ClbN, the NRPS module predicted to be
responsible for construction of the N-terminal prodrug motif
4
18
fatty acyl-CoAs. Strikingly, under these conditions, ClbN
preferentially utilized myristoyl-CoA (C ) for N-acylation
(
Figure 2A). The structures of prexenocoumacins revealed
14
(
Figure 2C). To explore whether this substrate specificity
might reflect in vivo activity, we used LC−MS to compare
metabolites produced by strains expressing either ClbN or an
empty expression vector. We identified both C and C N-acyl-
12
14
asparagine in the strain expressing ClbN; these metabolites were
not found in the control strain (Figure S11). Moreover, we were
unable to detect shorter chain N-acyl-asparagine products in
either strain. These preliminary results suggest that the selectivity
observed in the in vitro competition assay may reflect ClbN’s
substrate preference in vivo. Finally, we verified that the ClbN E
domain is functional by performing assays in D O and using LC−
2
MS to confirm the incorporation of deuterium into L-Asn
resulting from exchange during epimerization (Figures S13−
13
S16).
We then explored whether the N-acyl-asparagine scaffold
synthesized by ClbN was processed further by the colibactin
NRPS−PKS assembly line. The logic underlying the prodrug
strategy requires that the next module be an NRPS in order to
construct an amide bond that can serve as an eventual peptidase
Figure 2. ClbN is an NRPS involved in constructing the prodrug
scaffold. (A) Biosynthetic hypothesis for the activity of ClbN. (B) ATP−
3
2
[
P]PP exchange assay for ClbN substrate specificity. (C) Extracted
i
ion chromatograms of products obtained from the LC−MS competition
assay with ClbN, L-Asn, and equal amounts of C −C fatty acyl-CoA
4
18
substrates. All traces have been scaled to the same intensity.
substrate. To find a candidate module, we used bioinformatic
D
analyses to identify C domains, which are C domains that
L
prodrug scaffolds consisting of an N-acyl chain of variable length
appended to an invariant D-asparagine. In xenocoumacin
biosynthesis, this structure is likely constructed by the first
module of NRPS XcnA. ClbN, the homologous NRPS from the
pks island, was cloned from E. coli CFT073, overexpressed as an
follow E domains and accept upstream D-amino acyl thioester
6
10,14
electrophiles.
The C domain of ClbB, a two-module NRPS−
D
PKS hybrid, is the only C domain in the pks island (Figure
L
S33), indicating that it likely follows ClbN in the assembly line
(Figure 3A). To test this hypothesis, we cloned, overexpressed,
and purified the NRPS portion of ClbB (ClbBNRPS), which
contains C, A, and T domains, from E. coli CFT073. A BODIPY−
CoA loading assay confirmed the successful posttranslational
modification of apo-ClbBNRPS (Figure S17) and an ATP−
N-His -tagged fusion, and purified. A boron dipyrromethene−
6
9
coenzyme A (BODIPY−CoA) loading assay confirmed that the
T domain of apo-ClbN could be post-translationally modified to
the holo protein [Figure S3 in the Supporting Information (SI)].
Examination of the ClbN A domain’s substrate specificity using
32
[ P]PP exchange assay revealed that its A domain accepts
i
32
the ATP−[ P]PP exchange assay revealed selective activation of
L-asparagine (Figure 2B). This selectivity matched that observed
previously for ZmaO, the homologous NRPS module from the
multiple amino acids (Figure 3B). ClbBNRPS displayed the
highest activity with L-alanine but also activated L-valine, L-serine,
and glycine. This promiscuity extended to T domain loading, as
radiometric assays showed tethering of both L-alanine and L-
valine (Figure S18).
We examined the ability of ClbBNRPS to process substrates
from ClbN by incubating both holo enzymes with L-asparagine,
L-alanine, ATP, and a variety of fatty acyl-CoA substrates. We
i
5
zwittermicin biosynthetic gene cluster. Finally, a radiometric
1
4
loading assay using C-labeled L-asparagine demonstrated that
the T domain of holo-ClbN is charged with L-asparagine in the
presence of Sfp and ATP (Figure S5). Together, these results
confirm that ClbN utilizes L-asparagine, consistent with its
predicted role in prodrug scaffold construction.
detected N-acylated dipeptide products in the presence of C and
8
We next investigated whether the ClbN C domain could
acylate the amino substituent of the loaded L-asparagine.
C14 fatty acyl-CoAs using LC−MS (Figure 3C). This result
indicates that ClbB does accept substrates from ClbN, validating
B
dx.doi.org/10.1021/ja312154m | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Communication
whether this difference in cytotoxicity is directly linked to
catalysis of amide bond cleavage, as the activities of ClbP and
ClbPpep have not been examined with substrates containing the
prodrug scaffold.
We sought to understand the chemistry of prodrug activation
in more detail by evaluating the ability of ClbP to cleave
substrates containing structural motifs generated by ClbN. Our
initial efforts focused on ClbPpep, which we cloned from E. coli
CFT073 and expressed periplasmically. Using high-performance
liquid chromatography (HPLC), we found that ClbPpep can
cleave substrates containing N-octanoyl-D-asparagine and either
L-alanine or L-valine (Figures S22 and S23). A version of ClbPpep
containing a point mutation of the catalytic serine residue
(
S95A) was not able to cleave these substrates, confirming that
activity is due to ClbPpep. As expected, ClbPpep did not accept an
N-octanoyl-L-asparagine-containing substrate (Figure S24).
However, model substrates containing N-myristoyl-D-asparagine
also were not cleaved by ClbPpep (Figure S27). This substrate
specificity contrasted sharply with the product distribution
observed in the ClbN competition experiment, leading us to
hypothesize that the behavior of the truncated enzyme in vitro
might not accurately reflect the activity of the full-length enzyme
in vivo.
We investigated this possibility by examining the ability of E.
coli expressing either ClbPpep or full-length ClbP to cleave model
substrates in vivo. We incubated cultures expressing each of these
proteins, as well as the S95A mutant of full-length ClbP and an
empty vector control, with model substrates and used LC−MS to
detect the cleavage products. Strikingly, we found that while both
Figure 3. The NRPS module of ClbB elongates the prodrug scaffold.
(
[
A) Biosynthetic hypothesis for the activity of ClbB
. (B) ATP−
NRPS
3
2
P]PP exchange assay for ClbB
substrate specificity. (C)
i
NRPS
Extracted ion chromatograms of products obtained from individual
LC−MS assays with ClbN; ClbB
; C , C , C , or C fatty acyl-CoA;
L-Asn; and L-Ala or L-Val. All traces have been scaled to the same
NRPS
4
6
8
14
intensity. (D) LC−MS competition assay with equal amounts of C and
8
C and C N-acyl-D-asparagine-containing substrates were
C
fatty acyl-CoA substrates. Each bar represents the mean ± standard
8
14
14
processed by full-length ClbP, ClbPpep exhibited no activity
(Figure 4). This behavior is consistent with the in vivo activities
error of the mean (SEM) of normalized peak areas for extracted ion
chromatograms from three experiments.
our prediction of assembly line order. ClbBNRPS did not generate
products containing C and C N-acyl substituents. We also
4
6
performed the assay with L-valine in place of L-alanine and were
able to detect the corresponding N-acylated dipeptide (Figure
3C). Finally, we ran a competition experiment in which both
enzymes were incubated with L-alanine and equal concentrations
of octanoyl- and myristoyl-CoA; again the myristoylated product
predominated (Figure 3D). Together, these data indicate that
although ClbB is promiscuous with respect to both the chain
length of the N-acyl group on the intermediate it accepts from
ClbN and the amino acid it uses for elongation, the selectivity of
the initial N-acylation by ClbN may control prodrug scaffold
structure. Furthermore, the use of multiple amino acids by
ClbBNRPS suggests the possibility that colibactin could be a set of
related metabolites with different N-terminal amino acids.
Having established the roles of ClbN and ClbB in synthesizing
and elongating the prodrug scaffold, we next examined whether
peptidase ClbP is capable of prodrug activation. Full-length ClbP
contains an N-terminal signal sequence that targets the protein to
the inner membrane, a periplasmic peptidase domain containing
Figure 4. Peptidase ClbP cleaves the prodrug signal sequence in vivo.
LC−MS assays detecting N-acyl-D- or -L-asparagine produced from
cleavage of model peptidase substrates (100 μM) by E. coli BL21(DE3)
strains are shown. Each bar represents the mean ± SEM of three
experiments.
observed for cleavage of prexenocoumacin B by heterologously
6
expressed XcnG and the isolated XcnG peptidase domain. The
ClbP S95A mutant was inactive, again indicating that cleavage is
due to ClbP’s peptidase function. Incubations with an N-
octanoyl-L-asparagine-containing substrate resulted in negligible
levels of hydrolysis and greater amounts of unreacted starting
material compared with the corresponding D-asparagine-
containing substrate (Tables S12−S14). As the extent of
hydrolysis was similar across all of the cultures, including the
ClbP S95A mutant, we attribute this reactivity to endogenous
peptidases. The strong preference of ClbP for substrates
containing D-asparagine over L-asparagine further confirms that
ClbB elongates N-acyl-D-asparagine. Finally, we examined the
activity of ZmaM, a homologous peptidase from the zwittermicin
biosynthetic pathway that can restore cytotoxicity to a clbP
7
the active site, and three C-terminal transmembrane helices.
Genetic studies in E. coli have shown that clbP is required for
1
activity, and previous characterization of ClbP in vivo has
focused primarily on elucidating the elements of the protein
7b
required for restoring cytotoxicity to clbP mutants. These
experiments revealed that the N-terminal signal sequence and all
three transmembrane helices are necessary for complementation.
Surprisingly, the periplasmic peptidase domain of ClbP
(
ClbPpep), which has been overexpressed, purified, and
7a 7b
7a
crystallized, is unable to restore cytotoxicity. It was unclear
mutant. E. coli expressing ZmaM also processed model
C
dx.doi.org/10.1021/ja312154m | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Communication
substrates (Figure S30), providing further support for the
involvement of prodrug cleavage in generating active cytotoxin.
Notably, E. coli expressing ZmaM cannot cleave prexenocouma-
inhibiting ClbP. Such compounds would be useful for elucidating
the role that cytotoxin production plays in gut communities
containing pks E. coli and could provide a means of preventing
+
6
cin B. Our results suggest that this discrepancy arises from
colibactin-mediated carcinogenesis.
differences in metabolite structure, indicating that these
peptidases have evolved to recognize specific substrates.
Together with previous in vivo complementation experi-
ASSOCIATED CONTENT
Supporting Information
Experimental methods and additional data. This material is
available free of charge via the Internet at http://pubs.acs.org.
■
*
S
7
ments, these results provide evidence that colibactin formation
involves amide bond hydrolysis. Importantly, the differential
activity of ClbP and ClbPpep toward model substrates in vivo
parallels their ability to complement clbP mutants and restore
AUTHOR INFORMATION
Corresponding Author
balskus@chemistry.harvard.edu
Notes
■
+
7b
cytotoxicity to pks E. coli. This observation strongly suggests
that generation of active cytotoxin requires cleavage of the
prodrug scaffold generated by ClbN (Figure 5). The difference in
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Alex Sieg for help with initial experiments; Alan
Saghatelian and Sunia Trauger for assistance with LC−MS; Jo
Handelsman for providing Bacillus cereus UW85; and Harvard
University, the Corning Foundation, the Searle Scholars
Program, the Richard and Susan Smith Family Foundation
(
Newton, MA), and the Milton Fund for support. C.A.B.
acknowledges a fellowship from Novartis.
REFERENCES
■
(
1) Nougayred
̀
e, J. P.; Homburg, S.; Taieb, F.; Boury, M.;
Brzuszkiewicz, E.; Gottschalk, G.; Buchrieser, C.; Hacker, J.;
Dobrindt, U.; Oswald, E. Science 2006, 313, 848.
Figure 5. Proposed biochemical logic underlying the prodrug resistance
mechanism in colibactin biosynthesis. IM = inner membrane, OM =
outer membrane.
(
2) Cuevas-Ramos, G.; Petit, C. R.; Marcq, I.; Boury, M.; Oswald, E.;
Nougayrede, J. P. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 11537.
3) Arthur, J. C.; Perez-Chanona, E.; Muhlbauer, M.; Tomkovich, S.;
̀
(
̈
Uronis, J. M.; Fan, T. J.; Campbell, B. J.; Abujamel, T.; Dogan, B.;
Rogers, A. B.; Rhodes, J. M.; Stintzi, A.; Simpson, K. W.; Hansen, J. J.;
Keku, T. O.; Fodor, A. A.; Jobin, C. Science 2012, 338, 120.
the activities of the full-length enzyme and the truncated
peptidase domain is intriguing and could indicate that the
transmembrane helices influence the catalytic activity or that
interactions of both substrate and enzyme with the inner
membrane are important. In support of the latter hypothesis,
examination of the putative substrate binding pocket of ClbP
(
4) Johnson, J. R.; Johnston, B.; Kuskowski, M. A.; Nougayred
Oswald, E. J. Clin. Microbiol. 2008, 46, 3906.
5) Kevany, B. M.; Rasko, D. A.; Thomas, M. G. Appl. Environ.
Microbiol. 2009, 75, 1144.
̀
e, J. P.;
(
̈
(6) Reimer, D.; Pos, K. M.; Thines, M.; Grun, P.; Bode, H. B. Nat.
Chem. Biol. 2011, 7, 888.
(7) (a) Dubois, D.; Baron, O.; Cougnoux, A.; Delmas, J.; Pradel, N.;
Boury, M.; Bouchon, B.; Bringer, M.; Nougayrede, J.; Oswald, E.;
Bonnet, R. J. Biol. Chem. 2011, 286, 35562. (b) Cougnoux, A.; Gibold,
L.; Robin, F.; Dubois, D.; Pradel, N.; Darfeuille-Michaud, A.; Dalmasso,
G.; Delmas, J.; Bonnet, R. J. Mol. Biol. 2012, 424, 203.
7b
revealed a negative electrostatic potential, which suggests that
colibactin is positively charged. Positively charged amino acids
facilitate binding of myristoylated proteins to cell membranes as a
result of strong electrostatic interactions with acidic phospho-
̀
1
5
lipids. If precolibactin contains a myristoylated prodrug
scaffold attached to a positively charged core, it may be similarly
membrane-associated. Our work also confirms that the
biochemical logic employed by the colibactin prodrug resistance
mechanism is conserved across multiple biosynthetic pathways,
as ClbN produced a prodrug scaffold resembling that of the
prexenocoumacins and peptidase ZmaM processed model
substrates containing the colibactin prodrug scaffold.
(8) ClbPpep has previously been reported to cleave Gly-, D-Ala-, L-Ala-,
and L-Ala-L-Ala-p-nitroanilides with low catalytic efficiencies (k /K =
cat
m
−1
−1
7a
0
.005, 0.018, 0.020, and 0.040 μM min , respectively). We were
unable to detect any cleavage activity with D-Ala- and L-Ala-p-
nitroanilides using ClbPpep purified from periplasmic expressions.
(9) La Clair, J. J.; Foley, T. L.; Schegg, T. R.; Regan, C. M.; Burkart, M.
D. Chem. Biol. 2004, 11, 195.
(
10) Rausch, C.; Hoof, I.; Weber, T.; Wohlleben, W.; Huson, D. H.
BMC Evol. Biol. 2007, 7, 78.
11) Wittmann, M.; Linne, U.; Pohlmann, V.; Marahiel, M. A. FEBS J.
008, 275, 5343.
12) Kraas, F. I.; Helmetag, V.; Wittmann, M.; Strieker, M.; Marahiel,
M. A. Chem. Biol. 2010, 17, 872.
In summary, we have gained insights into the biochemical
events involved in colibactin assembly by characterizing the
activities of ClbN, ClbB, and ClbP, the enzymes involved in
synthesizing, elongating, and cleaving the colibactin prodrug
scaffold. By elucidating the biosynthetic logic underlying this
resistance strategy, we have obtained information about
structural features of colibactin, which should have an immediate
practical impact on efforts to isolate and characterize this natural
product. More importantly, we have also gained an under-
standing of the chemistry employed in generating the mature
cytotoxin. Ultimately, this knowledge could also guide the design
of small molecules that interfere with colibactin production by
(
2
(
(13) Stachelhaus, T.; Walsh, C. T. Biochemistry 2000, 39, 5775.
(14) Clugston, S. L.; Siber, S. A.; Marahiel, M. A.; Walsh, C. T.
Biochemistry 2003, 42, 12095.
(15) Sigal, C. T.; Zhou, W. J.; Buser, C. A.; McLaughlin, S.; Resh, M. D.
Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 12253. (b) McLaughlin, S.;
Aderem, A. Trends Biochem. Sci. 1995, 20, 272.
D
dx.doi.org/10.1021/ja312154m | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX