Isolation of a metabolite from the pks island provides insights into colibactin biosynthesis and activity

Article abstract of DOI:10.1021/acs.orglett.5b00432

Colibactin is a structurally uncharacterized, genotoxic natural product produced by commensal and pathogenic strains of E. coli that harbor the pks island. A new metabolite has been isolated from a pks⁺ E. coli mutant missing an essential biosy

Full text of DOI:10.1021/acs.orglett.5b00432

Letter
pubs.acs.org/OrgLett
Isolation of a Metabolite from the pks Island Provides Insights into
Colibactin Biosynthesis and Activity
Carolyn A. Brotherton,† Matthew Wilson,† Gary Byrd,‡ and Emily P. Balskus*,†
†Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, United States
‡Small Molecule Mass Spectrometry Facility, Faculty of Arts and Sciences Division of Science, Harvard University, Cambridge,
Massachusetts 02138, United States
S
* Supporting Information
ABSTRACT: Colibactin is a structurally uncharacterized, genotoxic
natural product produced by commensal and pathogenic strains of E.
coli that harbor the pks island. A new metabolite has been isolated from
a pks+ E. coli mutant missing an essential biosynthetic enzyme. The
unusual azaspiro[2.4] bicyclic ring system of this molecule provides
new insights into colibactin biosynthesis and suggests a mechanism
through which colibactin and other pks-derived metabolites may exert
genotoxicity.
olibactin is a human-associated bacterial natural product
Cthat has been linked to disease. Exposure to certain strains
of Escherichia coli, including gut commensals and extraintestinal
pathogens, induces double-strand breaks in the DNA of host
cells.1 This genotoxic activity has been linked to a biosynthetic
gene cluster (the pks island) that encodes a hybrid nonribosomal
peptide synthetase (NRPS)−polyketide synthase (PKS) assem-
bly line (Figure 1A).1a Recent studies have shown that pks+ E. coli
promote tumorigenesis in a mouse model of colitis-associated
colorectal cancer (CRC) and that both colitis and CRC patients
have increased carriage of pks+ strains.2 These findings support
the theory that colibactin production affects cancer progression
in humans. The pks island is also found in the genomes of
sponge- and honeybee-associated bacteria, suggesting it may
influence other symbioses.3 Despite intense interest in its
activity, the active genotoxin encoded by the pks island
(colibactin) has not been isolated, severely limiting efforts to
study its biosynthesis, mode of action, and role in host−microbe
interactions.
We have gained structural information about colibactin by
studying the biosynthetic enzymes encoded by the pks island. We
Figure 1. (A) The pks island. NRPS (purple), PKS (brown), hybrid
NRPS/PKS (blue), and peptidase ClbP (green) are highlighted. (B)
recently characterized the chemical events associated with the
Established chemical events in colibactin biosynthesis. C = con-
colibactin self-resistance mechanism (Figure 1B).4 This process
densation, A = adenylation, T = thiolation, E = epimerization, KS =
involves the biosynthesis of an inactive metabolite or metabolites
(precolibactin) containing an N-acylated-D-asparagine intro-
duced by the first module of the assembly line (ClbN). Following
elongation by ClbB and the remaining assembly-line enzymes,
the prodrug motif is hydrolyzed by the periplasmic peptidase
ClbP, an activity that is necessary for production of active
genotoxin (colibactin). This biosynthetic proposal has been
supported by recent isolation efforts which yielded the prodrug
scaffold (1) and a potential assembly-line truncation product
(Metabolite A, 2) (Figure 1C).5 These studies also highlight the
challenges of isolating the active product(s) from the pks island.
ketosynthase, AT = acyltransferase, KR = ketoreductase, DH =
dehydratase, ER = enoyl reductase, CoA = coenzyme A. (C) Structurally
characterized metabolites from pks+ E. coli strains.
To aid our ongoing efforts to understand colibactin biosyn-
thesis, we sought to identify additional metabolites derived from
this gene cluster. We describe here the isolation of Metabolite B
(3), a new molecule produced by a ΔclbP E. coli mutant. This
Received: February 10, 2015
Published: March 10, 2015
© 2015 American Chemical Society
1545
DOI: 10.1021/acs.orglett.5b00432
Org. Lett. 2015, 17, 1545−1548

Organic Letters
Letter
Table 1. NMR Data of Metabolite B in DMSO-d6
fragment
carbon
δC (type)
δH, multiplicity, J (Hz)
COSY
HMBC (1H to 13C)
2, 3
ROESY
I
1
14.0 (CH3)
22.1 (CH2)
31.3 (CH2)
0.85, t (7.3)
1.30−1.13, m
1.30−1.13, m
1.30−1.13, m
1.30−1.13, m
1.45, m
2
1
2
3
4−10
11
12
13
14
29.10, 29.06, 28.99, 28.88, 28.75 (CH2)
28.65 (CH2)
25.2 (CH2)
35.2 (CH2)
172.1 (C)
12
11, 13
11
13
2.08, m
12
11, 12, 14
12, 14NH
−
−
NH, 7.88, d, (8.1)
15
14
13, 16b
18NH
15
16
49.8 (CH)
37.4 (CH2)
4.47, m
14NH, 16a, 16b
15, 16b
16, 17, 18
15, 17, 18
15, 17, 18
a 2.45, dd (7.7, 15.1)
b 2.30, dd (7.7, 15.1)
−
15, 16a
14NH
17
18
171.4 (C)
170.3 (C)
a NH, 7.24, s
b NH, 6.81, s
−
17bNH
17aNH
NH, 7.48, d, (8.2)
19
18
21
15, 20, 21
II
19
20
21
22
44.1 (CH)
20.5 (CH3)
29.87 (CH2)
38.4 (CH2)
3.71, m
18NH, 20
19
20, 21
1.00, d, (7.3)
18NH, 19, 21
1.60, m
22a, 22b
21, 22b
21, 22a
18NH, 19, 20, 22a, 22b
a 2.92, ddd (6, 8.7, 17.3)
b 2.82, ddd (6, 8.7, 17.3)
21, 23
21, 23
21
21
III
23
24
25
26
27
28
197.7 (C)
169.2 (C)
128.9 (C)
10.8 (CH3)
45.5 (C)
−
−
−
1.96, s
24, 25, 27
29, 30
28a, 29a
−
13.7 (CH2)
a 1.49, m
b 1.41, m
a 1.49, m
b 1.41, m
−
28b, 29a, 29b
28a, 29a, 29b
28a, 28b, 29b
28a, 28b, 29a
26
30NH
26
29
30
13.7 (CH2)
169.6 (C)
28, 30
30NH
NH, 8.51, s
24, 25, 27
28b, 29b
metabolite contains a spiro-cyclopropane, a structural motif rare
among NRPS and PKS products and one never previously
observed in an E. coli metabolite. The structure of 3 suggests that
DNA or protein alkylation may be important for genotoxicity.
Overall, this work demonstrates the pks island encodes novel
biosynthetic activity, sheds light on colibactin’s biological
activity, and expands our knowledge of the biosynthetic
capabilities of E. coli.
We decided to isolate molecules from a clbP mutant for several
reasons. Our characterization of the colibactin self-resistance
mechanism led us to hypothesize that this deletion would lead to
the accumulation of inactive pks metabolites containing the N-
terminal prodrug motif. We postulated that these metabolites
might display increased stability relative to those lacking the
motif, as the primary amine released by prodrug hydrolysis could
react as a nucleophile in intra- or intermolecular decomposition
pathways. Additionally, we envisioned using the structure of the
prodrug motif to identify candidate metabolites for isolation.
We began our work by comparing the metabolite profiles of
whole-culture methanol extracts of E. coli DH10B expressing the
wild-type pks cluster (BACpks) or the pks cluster with clbP
knocked out (BACpksΔclbP) using the web-based XCMS
program (Table S1).6 Several metabolites were significantly
enriched in BACpksΔclbP extracts. The most abundant was the
known compound 2 (m/z 462.3386, [M + Na]+ of 439.3410)
(Figure 1C), which was identified by Crawford and co-workers as
the most abundant metabolite associated with two ΔclbP
strains.5b Another highly abundant molecule, Metabolite B (3)
(m/z 569.3775, [M + Na]+ of 546.3781), had not been isolated
previously. Consistent with earlier studies, the only metabolite
significantly enriched in the BACpks extracts was the hydrolyzed
prodrug motif (1) (m/z 343.2549, [M + H]+ of 342.2519).5a
When we fed BACpksΔclbP cultures with fully deuterated
myristic acid (myr-d27), both 2 and 3 were labeled (Figure S2).
Through LC−MS/MS analysis we discovered that, in addition to
forming fragments corresponding to loss of the prodrug motif,
unlabeled and myr-d27 labeled 3 shared a highly unsaturated
fragment (m/z 205.1326), which we determined from accurate
mass measurement had a molecular formula of C12H17N2O (9)
(Figures S3 and S4). From this information, we hypothesized
that 3 might contain a heterocycle and would provide an
intriguing target for isolation.
To facilitate this effort we sought to increase production of pks
metabolites in our expression strains. We evaluated the effects of
different variables on metabolite production in BACpks cultures
using the concentration of the hydrolyzed prodrug motif as a
proxy for flux through the biosynthetic pathway (Figure S5). We
found that overexpression of the pks transcriptional regulator
1546
DOI: 10.1021/acs.orglett.5b00432
Org. Lett. 2015, 17, 1545−1548

Organic Letters
Letter
clbR resulted in a 5-fold increase in prodrug motif accumulation.7
Growth in terrific broth (TB) led to a further 4-fold increase
compared to growth in Luria−Bertani (LB) media. With these
results in hand, we chose to isolate 3 from a DH10B strain grown
in TB and harboring both BACpksΔclbP and pTrcHisA-ClbR-N-
His6. We grew 60 L of culture, obtaining 70 g of dried cell mass.
Extraction with methanol provided 10.6 g of crude extract that
was further fractionated using silica gel flash column chromatog-
raphy. The 1:1 ethyl acetate/methanol fraction was further
purified using two rounds of reversed phase HPLC to obtain ∼5
mg of 3.
which is connected by a carbonyl to an unsaturated heterocycle
(III, Figure 2B). We propose the absolute stereochemistry of 3
based on analogy to that of pks metabolites 1 and 2, as well as the
known biosynthetic activities of ClbN and ClbB. Connectivities
between and within fragments I−III were established by
correlation spectroscopy (gCOSY) correlations and homonu-
clear multiple bond correlation (gHMBCAD) spectroscopy
(Table 1 and Figure 2B). While the left-hand portion of 3 (I and
II) is similar to known pks metabolites, the right-hand fragment
(III) is a novel structure composed of an unusual azaspiro[2.4]-
heptenone heterocycle.
The spiro-cyclopropane ring was the most challenging
structural assignment encountered in our analysis. The proton
resonances at δ1.49 (28Ha and 29Ha) and δ1.41 (28Hb and
29Hb) showed COSY correlations only to each other. According
to heteronuclear single quantum coherence spectroscopy
(gHSQCAD), these protons were attached to a carbon with
the same chemical shift (δ13.7, 28C and 29C) and the same
phase as the methine and methyl carbons. Analysis by
distortionless enhancement by polarization transfer (DEPT)
13C NMR unequivocally confirmed this resonance as a
methylene (Figures S12−S14). In the DEPT spectrum, the
carbon resonance at δ13.7 matched the phase of the other
methylene carbons, and in the 1H-coupled DEPT spectrum, the
peak appeared as a negative triplet (1JC−H of 169 Hz).
High-resolution mass measurement of the protonated
molecular ion for 3 (m/z 547.3851) yielded a molecular formula
1
of C30H50N4O5 with 8 degrees of unsaturation. The H NMR
spectrum in DMSO-d6 (Figure S7) displayed resonances
suggesting the presence of five NH protons at δ8.51 (s), δ7.88
(d, J = 8.1 Hz), δ7.48 (d, J = 8.2 Hz), δ7.24 (s), and δ6.81 (s); two
methines at δ4.47 (m) and δ3.71 (m); a methyl singlet at δ1.96; a
methyl doublet at δ1.00 (J = 7.3 Hz); a methyl triplet at δ0.85 (J =
7.3 Hz); seven methylenes at δ2.92 (ddd, J = 6, 8.7, 17.3 Hz),
δ2.82 (ddd, J = 6, 8.7, 17.3 Hz), δ2.45 (dd, J = 7.7, 15.1 Hz), δ2.30
(dd, J = 7.7, 15.1 Hz), δ2.08 (m), δ1.60 (m), and δ1.45 (m); and
a broad multiplet corresponding to 20 protons at δ1.30−1.13.
13C NMR analysis in DMSO-d6 (Figure S8) revealed resonances
suggesting the presence of a carbonyl at δ197.7; five amide or
electron-deficient sp2-hybridized carbons at δ172.1, δ171.4,
δ170.3, δ169.6, and δ169.2; an olefin carbon at δ128.9; three
carbons adjacent to amide nitrogens at δ49.8, δ45.5, and δ44.1;
three carbons adjacent to carbonyls at δ38.4, δ37.4, and δ35.2;
ten alkyl carbons at δ31.3, δ29.87, δ29.10, δ29.06, δ28.99, δ28.88,
δ28.75, δ28.65, δ25.2, and δ22.1; and four methyl groups or
upfield methylenes at δ20.5, δ14.0, δ13.7, and δ10.8. Complete
analysis by both one-dimensional (1D) and two-dimensional
(2D) NMR allowed us to complete the structural assignment of
Metabolite B (Table 1, Figure 2A).
Distinguishing between potential constitutional isomers of
fragment III was achieved through analysis of NOE correlations
and chemical synthesis. Nuclear Overhauser effect spectroscopy
(ROESYAD) revealed multiple through-space correlations
within fragment III, including correlations from δ1.96 (26H)
to δ1.49 (28Ha and 29Ha) as well as correlations from δ1.41
(28Hb and 29Hb) to δ8.51 (30NH) (Figure S15). These data
suggested the lactam nitrogen and the allylic methyl carbon were
both adjacent to the spiro carbon (27C). To provide additional
support for this structural assignment, we synthesized a model of
fragment III (7) in five steps from Boc-protected 1-amino-1-
cyclopropane carboxylic acid (4) (Figure 2C). The key step in
this sequence was an intramolecular Knoevenagel condensation
that proceeded rapidly and cleanly to provide the desired
From our analyses, we concluded that 3 is comprised of an N-
myristoyl-D-asparagine residue (I, Figure 2B) joined by an amide
bond to a saturated linker derived from L-alanine (II, Figure 2B),
8
1
spirocycle in excellent yield. The H and 13C NMR spectra of
this molecule closely matched those of fragment III (Figures
S16−S17, Table S3). From this comparison we can conclude that
3 and 7 very likely possess related connectivity.
The chemical structure of Metabolite B provides information
about some of the biosynthetic events that follow construction of
the prodrug scaffold and elongation by ClbB (Scheme S1).
Specifically, the construction of 3 could begin with a thiolation
(T) domain-tethered thioester generated by ClbB. Decarbox-
ylative Claisen condensation with malonyl-CoA, amide bond
formation with 1-amino-1-cyclopropane carboxylic acid, and a
second Claisen condensation with malonyl-CoA would generate
an enzyme-tethered intermediate that could undergo an
intramolecular Knoevenagel condensation similar to that used
in the synthesis of 7. Thioester hydrolysis and decarboxylation
would provide 3. There are no homologues of known
cyclopropane biosynthetic enzymes encoded in the pks cluster,
suggesting that this structural feature may be made using new
biosynthetic logic.9 We do not yet know if 3 is produced by wild-
type pks+ E. coli and whether it is an intermediate or a shunt
product in colibactin biosynthesis.
Figure 2. Metabolite B (3) possesses an unusual spiro-cyclopropane
ring system. (A) The deduced structure of 3. (B) Key COSY and HMBC
correlations establishing the structure of 3. (C) Synthesis of model
compound 7.
To gain preliminary insights into whether the bicyclic ring
system present in Metabolite B is also found in precolibactin, we
tested the ability of peptidase ClbP to cleave 3 in an in vivo LC-
1547
DOI: 10.1021/acs.orglett.5b00432
Org. Lett. 2015, 17, 1545−1548

Organic Letters
Letter
Notes
MS assay (Scheme 1, Figures S18−S20). We measured the
amount of 1 in culture extracts of ClbP-expressing E. coli strains
The authors declare no competing financial interest.
Scheme 1. Metabolite B is Processed by Peptidase ClbP
ACKNOWLEDGMENTS
■
We thank Gregory Heffron (Harvard Medical School, Boston,
MA) for assistance with NMR experiments, Sunia Trauger
(Small Molecule Mass Spectrometry Facility, Harvard Univer-
sity, Cambridge, MA) for help with LC−MS analyses, and Li Zha
(Department of Chemistry and Chemical Biology, Harvard
University, Cambridge, MA) and Pedro Leao (Center for Marine
̃
and Environmental Research, University of Porto, Porto,
Portugal) for helpful discussions. We acknowledge financial
support from the Damon Runyon-Rachleff Innovation Award,
the Smith Family Award for Excellence in Biomedical Research,
and the Packard Fellowship for Science and Engineering.
incubated with 3, known ClbP substrate N-myristoyl-D-
asparagine-L-alanine-O-methyl ester,4 or DMSO. E. coli express-
ing full-length ClbP-C-His6 hydrolyzed both 3 and the control
substrate. In assays with 3, we did not identify masses
corresponding to the accumulation of primary amine 8 or
imine 9. Instead we observed a product 10 that had a mass
consistent with a chloride adduct (m/z 241.1108). Neither 1 nor
10 was found in assays with cells expressing the inactive mutant
ClbP-S95A-C-His6 or an empty vector. The ability of ClbP to
process Metabolite B supports the hypothesis that this
spirocyclic ring system is present in precolibactin and/or other
ClbP substrates produced by pks+ E. coli.
Our results raise important questions about the role of ClbP in
generating active genotoxin and the mechanism by which
colibactin production leads to DNA damage. We postulate that
prodrug cleavage could enhance the activity of 3 and related pks
metabolites in two ways: by releasing metabolites from the E. coli
membrane through removal of the lipophilic N-acyl chain and by
generating a more potent electrophile, an α,β-unsaturated
iminium with an adjacent cyclopropyl group, that could alkylate
DNA or a target protein.10 Intriguingly, this activation strategy
bears resemblance to, but is unique from, those involved in
modulating the activities of other DNA-alkylating natural
products that contain cyclopropanes,11 including the CC-
1065/duocarmycins12 (conformational change upon DNA
binding) and the illudins/acylfulvenes13 (activation via reductive
metabolism). Based on this precedent, we have tentatively
proposed a cyclopropane-opened structure for 10.
REFERENCES
■
(1) (a) 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. (b) Cuevas-Ramos, G.; Petit, C.
R.; Marcq, I.; Boury, M.; Oswald, E.; Nougayred
Sci. U.S.A. 2010, 107, 11537.
̀
e, J. P. Proc. Natl. Acad.
(2) 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.
(3) (a) Bondarev, V.; Richter, M.; Romano, S. Environ. Microbiol. 2013,
15, 2095−2113. (b) Engel, P.; Vizcaino, M. I.; Crawford, J. M. Appl.
Environ. Microbiol. 2015, 81, 1502.
(4) Brotherton, C. A.; Balskus, E. P. J. Am. Chem. Soc. 2013, 135, 3359.
(5) (a) Bian, X.; Fu, J.; Plaza, A.; Herrmann, J.; Pistorius, D.; Stewart, A.
F.; Zhang, Y.; Muller, R. ChemBioChem 2013, 14, 1194. (b) Vizcaino, M.
̈
I.; Engel, P.; Trautman, E.; Crawford, J. M. J. Am. Chem. Soc. 2014, 136,
9244.
(6) Tautenhahn, R.; Patti, G. J.; Rinehart, D.; Siuzdak, G. Anal. Chem.
2012, 84, 5035.
(7) Krumbholz, G. Ph.D. Thesis, Julius-Maximilians-Universitat
̈
Wurzburg, 2010.
̈
(8) Kato, T.; Sato, M.; Yoshida, T. Chem. Pharm. Bull. 1971, 19, 292.
(9) Thibodeaux, C. J.; Chen, W.; Liu, H. Chem. Rev. 2012, 112, 1681.
(10) (a) Wong, H. N. C.; Hon, M.; Tse, C.; Yip, Y.; Tanko, J.;
Hudlicky, T. Chem. Rev. 1989, 89, 165. (b) Bose, G.; Bracht, K.;
Bednarski, P. J.; Lalk, M.; Langer, P. Bioorg. Med. Chem. 2006, 14, 4694.
(11) (a) Wolkenberg, S. E.; Boger, D. L. Chem. Rev. 2002, 102, 2477.
(b) Gersch, M.; Kreuzer, J.; Sieber, S. A. Nat. Prod. Rep. 2012, 29, 659.
(12) (a) Gosh, N.; Sheldrake, H. M.; Searcey, M.; Pors, K. Curr. Top.
Med. Chem. 2009, 9, 1494. (b) Boger, D. L.; Garbaccio, R. M. Acc. Chem.
Res. 1999, 32, 1043. (c) Boger, D. L.; Johnson, D. S. Proc. Natl. Acad. Sci.
U.S.A. 1995, 92, 3642.
In conclusion, we have successfully isolated a complex
metabolite associated with the pks island that will guide further
studies of colibactin biosynthesis and biological activity. The
structure of Metabolite B raises intriguing questions about how
the azaspiro[2.4] bicyclic ring system is biosynthesized by the pks
assembly line enzymes and how this structure may contribute to
colibactin’s genotoxicity. Beyond its impact in the area of natural
product discovery, this work will aid future efforts to understand
how pks+ E. coli impact human health and disease.
(13) Tanasova, M.; Sturla, S. Chem. Rev. 2012, 112, 3578.
ASSOCIATED CONTENT
* Supporting Information
■
S
1
Full experimental and characterization data, including H and
13C NMR for all new compounds. This material is available free
of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
■
*E-mail: balskus@chemistry.harvard.edu.
1548
DOI: 10.1021/acs.orglett.5b00432
Org. Lett. 2015, 17, 1545−1548