Lawrence Berkeley National Laboratory
Recent Work
Title
CRAGE enables rapid activation of biosynthetic gene clusters in undomesticated bacteria.
Permalink
https://escholarship.org/uc/item/2nt21413
Journal
Nature microbiology, 4(12)
ISSN
2058-5276
Authors
Wang, Gaoyan
Zhao, Zhiying
Ke, Jing
et al.
Publication Date
2019-12-01
DOI
10.1038/s41564-019-0573-8
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
1
US Department of Energy Joint Genome Institute, Berkeley, CA, USA.
2
Molecular Biotechnology, Department of Biosciences and Buchmann Institute for
Molecular Life Sciences, Goethe Universität Frankfurt, Frankfurt am Main, Germany.
3
Environmental Molecular Sciences Laboratory, Pacific Northwest
National Laboratory, Richland, WA, USA.
4
Environmental Genomics and Systems Biology Division, Lawrence Berkeley National Laboratory, Berkeley,
CA, USA.
5
Emery Pharma, Alameda, CA, USA.
6
Biological Systems and Engineering Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA.
7
LOEWE Centre for Translational Biodiversity Genomics, Frankfurt, Germany.
8
Center for Advanced Bioenergy and Bioproducts Innovation, Urbana, IL,
USA.
9
Global Institution for Collaborative Research and Education, Hokkaido University, Hokkaido, Japan.
10
These authors contributed equally: G. Wang,
Z. Zhao, J. Ke, Y. Engel, Y.-M. Shi. *e-mail: h.bode@bio.uni-frankfurt.de; yyoshikuni@lbl.gov
Microbial secondary metabolites (also called natural prod-ucts or
specialized metabolites) play important roles in modulating
biological systems and their host environ-
ments
1–3
. Their roles include intra- and inter-cellular communica-
tion
4
, host defence
5
and pathogenicity
6
. Systematic discovery and
characterization of secondary metabolites is therefore an essential
step towards understanding diverse ecosystems and exploiting their
chemical diversity for applications in health, food, agriculture and
the environment
7,8
. Recent advances in computational tools (for
example, antiSMASH
9
and PRISM
10
) have provided researchers
with unprecedented opportunities to mine massive sequence-data
spaces and to identify biosynthetic gene clusters (BGCs) responsible
for producing these secondary metabolites, and therefore a possible
blueprint for characterizing the biosynthesis of chemically diverse
secondary metabolites.
These genome analyses revealed that individual bacterial strains
may possess 20 to 50 BGCs
11,12
. However, around 90% of BGCs are
not usually expressed under standard conditions, and the regula-
tory triggers required to activate them are typically unknown
13
.
Although synthetic biology can, in principle, uncouple BGCs from
their native regulatory constraints
8,14–18
, current efforts have pri-
marily focused on transcriptional and translational optimization
in model chassis strains. However, successful activation of BGCs
also requires that all translated products fold properly and undergo
appropriate post-translational modifications, all substrates and
co-factors be available, and all intermediates and products be toler-
ated by the chassis strain used. Significant advances in host strain
engineering are needed to meet these complex requirements.
Evolutionary studies have often guided scientists in develop-
ing successful strategies for designing biological function
19,20
.
Large-scale comparative genome studies are revealing that BGCs
are probably evolved through horizontal gene transfer and subse-
quent modifications among closely related bacteria
21–26
. These BGC
exchanges are thought to be beneficial because they offer opportu-
nities to rapidly test the fitness effects of secondary metabolites
21
.
This idea also implies that recipient bacteria could utilize the BGCs
transferred from closely related bacteria more efficiently, as they
may share physiological conditions suitable for BGC activation,
but with greatly relaxed regulations
27
. Therefore, use of multiple
closely related bacteria as chassis for BGC expression could, in the-
ory, improve the efficiency of BGC activation. Although a lack of
systematic studies leaves some uncertainty, the general conclusion
from empirical studies in natural product research also supports
this assumption
28,29
.
CRAGE enables rapid activation of biosynthetic
gene clusters in undomesticated bacteria
Gaoyan Wang
1,10
, Zhiying Zhao
1,10
, Jing Ke
1,10
, Yvonne Engel
2,10
, Yi-Ming Shi
2,10
, David Robinson
1
,
Kerem Bingol
3
, Zheyun Zhang
1
, Benjamin Bowen
1,4
, Katherine Louie
1
, Bing Wang
1
, Robert Evans
1
,
Yu Miyamoto
1
, Kelly Cheng
1
, Suzanne Kosina
4
, Markus De Raad
4
, Leslie Silva
1
, Alicia Luhrs
5
,
Andrea Lubbe
5
, David W. Hoyt
3
, Charles Francavilla
5
, Hiroshi Otani
1,4
, Samuel Deutsch
1,4,6
,
Nancy M. Washton
3
, Edward M. Rubin
1
, Nigel J. Mouncey
1,4
, Axel Visel
1,4
, Trent Northen
1,4
,
Jan-Fang Cheng
1,4
, Helge B. Bode
1,7
* and Yasuo Yoshikuni
1,4,6,8,9
*
It is generally believed that exchange of secondary metabolite biosynthetic gene clusters (BGCs) among closely related bacte-
ria is an important driver of BGC evolution and diversification. Applying this idea may help researchers efficiently connect many
BGCs to their products and characterize the products’ roles in various environments. However, existing genetic tools support
only a small fraction of these efforts. Here, we present the development of chassis-independent recombinase-assisted genome
engineering (CRAGE), which enables single-step integration of large, complex BGC constructs directly into the chromosomes
of diverse bacteria with high accuracy and efficiency. To demonstrate the efficacy of CRAGE, we expressed three known and
six previously identified but experimentally elusive non-ribosomal peptide synthetase (NRPS) and NRPS-polyketide synthase
(PKS) hybrid BGCs from Photorhabdus luminescens in 25 diverse γ-Proteobacteria species. Successful activation of six BGCs
identified 22 products for which diversity and yield were greater when the BGCs were expressed in strains closely related to the
native strain than when they were expressed in either native or more distantly related strains. Activation of these BGCs dem-
onstrates the feasibility of exploiting their underlying catalytic activity and plasticity, and provides evidence that systematic
approaches based on CRAGE will be useful for discovering and identifying previously uncharacterized metabolites.
To enable the multi-chassis approach, genome engineering tools
such as phage integrases
30–33
can be useful. However, the utility of
these tools is limited to a small group of bacteria (Supplementary
Fig. 1). Additionally, integration efficiency significantly declines
as insert size increases
32
. To address these issues, we attempted to
develop chassis-independent recombinase-assisted genome engi-
neering (CRAGE) applicable to diverse bacterial species across
m
ultiple phyla (Fig. 1a). CRAGE is based on recombinase-assisted
genome engineering (RAGE) technology
34,35
. Use of RAGE has
allowed single-step integration of pathways comprising 48 kb
directly into the Escherichia coli chromosome without compromis-
ing integration efficiency. After simple antibiotic counter-selection,
t
he integration yield reached 100%. CRAGE extends RAGE by
enabling researchers to domesticate previously undomesticated
microbes, and substantially increases the chance of successful BGC
expression and discovery of previously uncharacterized secondary
metabolites.
Results
Developing a design principle for CRAGE. In RAGE, a landing
pad (LP) containing two mutually exclusive lox sites is first inte-
grated into a chromosome. The DNA constructs flanked by these
l
ox sites are then integrated into the LP, catalysed by Cre recombi-
nase. We modified RAGE
34,35
in three major ways to establish the
CRAGE design principle (Fig. 1a). First, we used transposon sys-
tems
36,37
(mariner system for Proteobacteria and Tn5 system for
Donor Recipient
Chromosome
Chromosome
pW17
T7RP
loxP lox5171
IR IR
KmR Cre
Transposase
Chromosome
Chromosome
loxP lox5171
pW34
AprR
Chromosome
Chromosome
Chromosome
b
Step 1: Landing pad integration
Step 2: BGC integration
IPTG
luxCDABE
IPTG
Ctrl
T7RP
IR IR
KmR Cre
Transposase
T7RP
loxP lox5171
IR IR
KmR Cre
BGC
T7RP
IR IR
KmR Cre
AprR
BGC
T7RP
AprR
BGC
a
XP01 P. luminescens subsp. laumondii TT01
XP02 P. luminescens subsp. luminescens
XP03 P. temperata subsp. khanii
XP04 X. nematophila corrig
XP05 X. doucetiae
EB01 D. zeae
EB02 D. solani
EB03 D. dadantii subsp. dadantii
EB04 D. dadantii subsp. dieffenbachiae
EB06 P. carotovorum subsp. odoriferum
EB10 S. odorifera
EB12 P. agglomerans strain Eh1087
EB13 E. pyrifoliae
EB15 E. oleae
EB16 Y. ruckeri
EB17 Y. bercovieri
EB18 Y. mollaretii
EB19 Y. aldovae
AM02 A. encheleia
AM03 A. salmonicida subsp. salmonicida
AM04 A. piscicola
AM05 A. salmonicida subsp. pectinolytica
PM01 P. simiae WCS417r
PM02 P. fluorescens Q8r1-96
PM03 P. putida KT2440
25 selected γ-Proteobacteria
Intensity (a.u.)
10
3
10
5
10
4
Fig. 1 | Chromosomal integration of BGCs mediated through CRAGE. a, Schematic for CRAGE, a genome engineering technology that allows complex
biological systems to be implemented in a broad range of microbial strains. We primarily focused on formulating the design principle combining all existing
technologies that have proven to work in a wide range of organisms. Step 1: A pW17 plasmid containing a mariner transposon and transposase was
generated. The transposon contained a Cre recombinase gene and a kanamycin-resistant gene (KmR) flanked by two mutually exclusive lox sites (loxP and
lox5171). In addition, a T7-RNA polymerase (T7RP) gene under the control of a lacUV5 regulon was incorporated into the transposon. The pW17 plasmid
was conjugated from donor E. coli strain BW29427 into the panel of recipient bacterial strains (Supplementary Tables 1 and 2), and the transposon was
integrated into their chromosomes. Step 2: A different plasmid, pW34 (R6Kr ori) or pW5 (BAC-based), encoding a BGC under control of the T7 promoter
and an apramycin-resistant gene (AprR) flanked by the two mutually exclusive lox sites, was conjugated into the recipient strain containing the LP. BGC
integration into the chromosome of these chassis strains was mediated through Cre recombinase activity. b, For 25 chassis strains containing only an LP
(control) and luxCDABE, luminescence activity was induced with four different IPTG concentrations (0, 0.01, 0.1 and 1 mM) and measured. All data were
generated from biological triplicates. The standard deviations were generally less than 10%. The colours used for b are coordinated to represent strains
classified in the same phylogenetic branches. a.u., arbitrary units.
Actinobacteria) to insert the LP into recipient bacteria chromo-
somes. Because transposon systems are commonly used in engi-
neering of diverse species ranging from prokaryotes to eukaryotes,
t
hey are suitable for the first step of domestication. Although the
transposon is randomly integrated into the chromosome of the
recipient strain, we can screen and select the transformants with the
LP integrated into the location minimally affecting the host strain’s
physiology.
Second, we used a Lac-T7 expression system
38
to control the
expression of BGCs. This system is orthogonal to native transcrip-
tion; genes under the control of a T7 promoter are not transcribed
un
less a T7 RNA polymerase (T7RNAP) is present. Using this
system, we can minimize the expression of BGCs whose prod-
ucts may be toxic to E. coli while we are assembling the BGC con-
structs. Additionally, although codon usage and ribosome binding
si
tes may need to be redesigned for each chassis strain to obtain
more optimal results, this design principle in general allows a high
degree of flexibility, so that any single construct in any CRAGE
strain can be expressed without re-cloning as long as the T7RNAP
is expressed under the control of promoters that function in the
recipient strains.
Several studies suggest that genomic integration location can
also affect the expression of integrated genes
34,39–41
. Therefore, inves-
tigating different integration locations is another viable approach
f
or exploring the effect of different expression levels on BGC activ-
ity. However, our previous study, as well as others, suggested that
t
he effect of different integration locations on enzyme and path-
way activity was generally small (at most less than about seven- to
eig
htfold, usually two- to threefold for enzyme activity and less than
two fold for pathway activity)
34,39–41
. In contrast, the Lac-T7 system
allows us to explore a much wider range of pathway activity (10-
to 1,000-fold) than we could if we explored according to integra-
tion location. Additionally, the approach of investigating different
in
tegration locations would complicate our workflow and make our
approach less attractive. Therefore, we specifically chose to use the
Lac-T7 system to explore the effect of different expression ranges
on BGC activity.
Third, we incorporated the origin of transfer (oriT) into both
the LP and BGCs carrying plasmids to use conjugation as a pri-
mary transformation method. Conjugation systems have been
u
sed for transformation of a wide spectrum of bacterial species
including those in the Proteobacteria, Actinobacteria, Firmicutes,
Bacteroidetes and Cyanobacteria phyla
42–44
. Although the present
study focuses on γ-Proteobacteria, our preliminary results demon-
strate that CRAGE can engineer bacteria across multiple phyla (for
exa
mple, α-Proteobacteria, β-Proteobacteria and Actinobacteria)
(Supplementary Figs. 1–3).
Selecting model biosynthetic gene clusters. Members of the
genus Photorhabdus, as well as the related genus Xenorhabdus, are
endosymbionts of soil-borne nematodes and are known to have
potent bioactivity toward a wide range of insects and insect larvae;
some of these entomopathogenic complexes are used as biological
in
secticides in agriculture
45,46
. These bacteria also produce numer-
ous secondary metabolites to inhibit the growth of competing
micr
obes within their hosts
47,48
. Genomes of ~50 Photorhabdus and
Xenorhabdus species have been sequenced and are accessible in
public databases. Computational analyses suggest that each of these
genomes contains ~20 to 50 putative BGCs
11,12
, and that many BGCs
divergently evolved among species within these two genera
49,50
. The
combination of extensive genomic resources and rich metabolic
potential makes these species ideal test cases for a purpose-engi-
neered multi-chassis strategy for BGC characterization.
W
e selected a model set of 10 NRPS and NRPS–PKS hybrid BGCs
from Photorhabdus luminescens subsp. laumondii TTO1 (Table 1,
Fig. 2 and Supplementary Fig. 4)
18,51
. Except for BGC6, these BGCs
had been previously cloned and heterologously expressed in E. coli.
However, only two of those nine BGCs had been successfully acti-
vated, making this study an ideal benchmark to test the efficacy
o
f the multi-chassis approach mediated by CRAGE. The selected
panel of pathways included four BGCs as controls that had been
previously studied successfully by single-chassis approaches (BGCs
4 and 9)
18,52–56
or by promoter replacement approaches in a native
strain (BGCs 1 and 6)
50,57,58
, as well as six putative BGCs that were
not functional using conventional chassis approaches (BGCs 2, 3, 5,
7, 8 and 10)
18
.
Preparing phylogenetically diverse chassis strain panels using
CRAGE. We selected 31 species of γ-Proteobacteria representing 10
different genera (Fig. 1, Supplementary Fig. 4 and Supplementary
Tables 1 and 2). The panel consisted of several Xenorhabdus and
Photorhabdus species (XPs) and many other species that are evolu-
tionarily slightly distant from XPs (other Enterobacteria (EB) and
b
acteria in the Aeromonas (AM) and Pseudomonas (PM) genera).
The panel also allowed us to systematically investigate correlations
between evolutionary relatedness and physiological compatibili-
ties of different chassis:BGC combinations. Selection criteria also
in
cluded the availability of complete or draft genome sequences
and classification as biosafety level 1. Furthermore, all the selected
strains had EntD- and/or Sfp-type phosphopantetheinyl transfer-
ases (PPTases), required for modifying NRPS and PKS proteins to
co
nvert their inactive apo forms into the enzymatically functional
holo forms
59
.
The LP on a transposon was first randomly integrated into the
chromosome of each strain and the integration site was determined
Table 1 | BGCs used in this study
BGC Gene(s) Genomic location Size (kb) Products Refs.
1 plu0897–plu0899 1021652–1035887 14.2 Known (7,8), Unknown (9,10) 50,57
2 plu1113–plu1115 1277282–1299002 21.7 Unknown
3 plu1210–plu1222 1392054–1414406 22.4 Unknown
4 plu1881–plu1877 2239952–2221617 18.3 Known (1–3) 18,53
5 plu2316–plu2325 2713457–2744350 30.9 Unknown (19–21)
6 plu2670 3173674–3124571 49.1 Known (kolossins A–C) 58
7 plu3123 3662492–3646119 16.4 Unknown (4–6)
8 plu3130 3688463–3678528 9.9 Unknown (22)
9 plu3263 3880777–3865127 15.6 Known (11–18) 18,54–56
10 plu3526–plu3538 4167664–4122969 44.7 Unknown
by whole genome sequencing. This purpose-engineering procedure
was successful in 27 species representing all 10 genera selected.
To demonstrate the general potential of this collection of chassis
strains to efficiently integrate and express heterologous pathways,
we selected a 7 kbp bacterial luciferase (lux) operon composed of
five genes (luxCDABE) originally derived from P. luminescens
60
. The
operon, flanked by mutually exclusive lox sites, was conjugated into
the panel of recipient strains and stably integrated into the chro-
mosome. Following antibiotic counter-selection, successful inte-
gration and operon activity were assessed by quantifying luciferase
ac
tivity. In 25 species from all 10 genera, chromosomal integration
was successful and luciferase activity was inducible with isopropyl
β--1-thiogalactopyranoside (IPTG; Fig. 1b). Importantly, antibi-
otic counter-selection resulted in an integration efficiency of 100%
(S
upplementary Table 2).
For several strains, we obtained two to three variants with the LP
inserted into different genomic locations. To investigate the effect
of different integration locations on BGC activity, we integrated the
lux operon into these strain variants. As expected from our previous
study and others
34,39–41
, we found that integration location had little
effect on luminescence activity (Supplementary Fig. 5). Indeed, our
strategy of using a Lac-T7 expression system covered a broader
range of luminescence activity. Therefore, we decided not to con-
sider the effect of integration location on BGC activity further.
G
enomic integration of BGCs across diverse chassis strains. To
introduce the selected NRPS and NRPS–PKS hybrid BGCs (Table 1,
Fig. 2 a
nd Supplementary Fig. 4) into the phylogenetically diverse
set of chassis strains, BGC constructs were assembled. This process
was successful for 9 of 10 BGCs. We were unable to clone BGC6
because it has extensive internal repeats reflected by its overall
NRPS organization, which hampered homologous recombination
in yeast. BGC6 was also not cloned in a previous study conducted
by Fu and others
18
. The remaining nine BGCs were subsequently
transferred to the panel of recipient strains via conjugation. This
process was successful for almost all chassis:BGC combinations
in 24 of 25 chassis strains, with an average integration efficiency
of 57% as measured by simple antibiotics counter-selection and
subsequent colony PCR to confirm successful BGC integration
(Supplementary Tables 3 and 4). As expected, even the largest clus-
ters tested were integrated without compromising the success rate.
A
cross the BGC size range evaluated in this series (10–48 kbp),
integration efficiency remained nearly identical, demonstrating the
efficacy of CRAGE technology.
Using the purpose-engineered multi-chassis approach to rapidly
identify known products. We cultivated the panels of 24 chassis
strains harbouring each BGC at four different IPTG concentra-
tions. We then used liquid chromatography–high resolution mass
spectrometry (LC-HRMS) to measure secondary metabolites pro-
duced by each chassis:BGC combination. Controls included a con-
ventional heterologous expression system, the E. coli BL21(DE3)
s
train with a genome-integrated Bacillus subtilis PPTase (sfp)
61
(Supplementary Fig. 4).
Targeted analysis of LC-HRMS data for chassis strains express-
ing BGCs 1, 4 and 9 revealed successful production of every pre-
viously described metabolite in at least one of the chassis strains
exa
mined here
18,50,52–57
(Fig. 3 and Supplementary Figs. 6–8). For
example, BGC4 comprises five genes (plu1877–plu1881) encod-
ing NRPS and NRPS–PKS hybrid (Fig. 3a). Previous studies have
im
plicated the expression of BGC4 in the production of luminmy-
cin A (1), glidobactin A (2) and cepafungin I (3), all of which show
BGC2
plu1113 plu1115
BGC7
plu3123
BGC8
plu3130
BGC9
plu3263 (gxpS)
BGC5
plu2320 plu2321
plu2324
plu2319plu2316
plu2670 (kolS)
BGC6
BGC1
plu0898 plu0899plu0897
P
T7
T
T7
BGC4
plu1880 plu1878plu1881
BGC1
0
plu3535 plu3533 plu3530 plu3527plu3534plu3538 plu3532
plu1212 plu1213 plu1214 plu1218 plu1220
plu1222plu1210
BGC3
Acyl-homoserine-lactone acylase:Diamine N-acetyltransferase:
MbtH protein:
Oxalate decarboxylase: Diaminobutyrate aminotransferase:
MFS transporter:
Receptor: Thiazollnyl reductase component: Salicyl-AMP ligase:
Transposase:ATP-binding cassette:
Thioesterase component:
Hypothetical protein:
Non-ribosomal peptide/polyketide synthase:
Non-ribosomal peptide synthetase:
Polyketide synthase:
Fig. 2 | Design and architecture of BGC constructs. The design and gene architecture of each BGC construct. P
T7
and T
T7
correspond to a T7 promoter and
a T7 terminator, respectively.
![Fig. 4 | Expression of BGC7 in a panel of phylogenetically diverse chassis strains. a, A proposed scheme of secondary metabolite biosynthesis catalysed by BGC7 (plu3213). NRPS domains are shown as circles; letters represent adenylation (A), thiolation (T), dual condensation and epimerization (C/E) and thioesterase (TE) domains. C/E* indicates dual condensation and epimerization domains with non-functional epimerization domain. The structure of the main ririwpeptide A (4) was confirmed by synthesis and labelling experiments (Supplementary Figs. 11–15). b, Measured MS1 intensity for 4 (m/z 372.238 [M+2H]2+), ririwpeptide B (5) (m/z 243.238 [M+3H]3+) and ririwpeptide C (6) (m/z 350.730 [M+2H]2+) produced from the panel of phylogenetically](/figures/fig-4-expression-of-bgc7-in-a-panel-of-phylogenetically-1tgbp4ft.webp)
