A subdomain swap strategy for reengineering nonribosomal peptides

Article abstract of DOI:10.1016/j.chembiol.2015.04.015

Nonribosomal peptide synthetases (NRPSs) protect microorganisms from environmental threats by producing diverse siderophores, antibiotics, and other peptide natural products. Their modular molecular structure is also attractive from the standpoint of biosynthetic engineering. Here we evaluate a methodology for swapping module specificities of these mega-enzymes that takes advantage of flavodoxin-like subdomains involved in substrate recognition. Nine subdomains encoding diverse specificities were transplanted into the Phe-specific GrsA initiation module of gramicidin S synthetase. All chimeras could be purified as soluble protein. One construct based on a Val-specific subdomain showed sizable adenylation activity and functioned as a Val-Pro diketopiperazine synthetase upon addition of the proline-specific GrsB1 module. These results suggest that subdomain swapping could be a viable alternative to previous NRPS design approaches targeting binding pockets, domains, or entire modules. The short length of the swapped sequence stretch may facilitate straightforward exploitation of the wealth of existing NRPS modules for combinatorial biosynthesis.

Full text of DOI:10.1016/j.chembiol.2015.04.015

Article
A Subdomain Swap Strategy for Reengineering
Nonribosomal Peptides
Graphical Abstract
Authors
Hajo Kries, David L. Niquille, Donald
Hilvert
Correspondence
hilvert@org.chem.ethz.ch
In Brief
Nonribosomal peptide synthetases
(NRPSs) produce a plethora of bioactive
natural products that can be further
diversified by enzyme engineering. Kries
et al. successfully employ swapping of
short subdomains to transfer specificity
from one NRPS module to another. This
approach complements other
engineering strategies and may facilitate
combinatorial biosynthesis of novel
peptides.
Highlights
d
d
d
d
Specificity of nonribosomal peptide synthetases (NRPSs) is
encoded on subdomains
Subdomain swaps can be used to reprogram the specificity
of a dipeptide synthetase
An engineered construct successfully incorporates valine
into a dipeptide
Transplanting short subdomains may be advantageous for
combinatorial screening
Kries et al., 2015, Chemistry & Biology 22, 640–648
May 21, 2015 ª2015 Elsevier Ltd All rights reserved
http://dx.doi.org/10.1016/j.chembiol.2015.04.015

Chemistry & Biology
Article
A Subdomain Swap Strategy for
Reengineering Nonribosomal Peptides
Hajo Kries,1,2,3 David L. Niquille,1,2 and Donald Hilvert1,
*
1Laboratory of Organic Chemistry, ETH Zurich, 8093 Zu¨ rich, Switzerland
2Co-first author
3Present address: Department of Biological Chemistry, The John Innes Center, Norwich NR4 7UH, UK
*Correspondence: hilvert@org.chem.ethz.ch
http://dx.doi.org/10.1016/j.chembiol.2015.04.015
SUMMARY
another species (Enright et al., 1999). In other cases, the free-
standing progenitors may no longer exist but can be inferred
Nonribosomal peptide synthetases (NRPSs) protect from phylogenetic analyses and structural studies. The wide-
spread (b/a)8 fold of TIM barrel enzymes, for instance, is believed
to have arisen from fusion of two no longer extant half barrels
microorganisms from environmental threats by pro-
ducing diverse siderophores, antibiotics, and other
peptide natural products. Their modular molecular
structure is also attractive from the standpoint of
(Farı´as-Rico et al., 2014; Lang et al., 2000). Insertion-deletions
can also alter protein topology in a dramatic fashion (Grishin,
2001).
biosynthetic engineering. Here we evaluate a meth-
Protein engineering can be very successful when designers
odology for swapping module specificities of these
turn to nature for inspiration. For instance, introduction of point
mega-enzymes that takes advantage of flavodoxin-
like subdomains involved in substrate recognition.
mutations, screening, selection, and recombination in directed
evolution experiments is nothing more than a recapitulation of
Nine subdomains encoding diverse specificities
were transplanted into the Phe-specific GrsA initia-
¨
Darwinian evolution in the laboratory (Jackel and Hilvert, 2010).
In analogy to homologous recombination in nature, family shuf-
tion module of gramicidin S synthetase. All chimeras fling of genes takes advantage of the natural diversity in a protein
family to boost activity in laboratory evolution (Crameri et al.,
1998; Minshull and Stemmer, 1999). Circular permutation, too,
could be purified as soluble protein. One construct
based on a Val-specific subdomain showed sizable
adenylation activity and functioned as a Val-Pro
diketopiperazine synthetase upon addition of the
proline-specific GrsB1 module. These results sug-
gest that subdomain swapping could be a viable
alternative to previous NRPS design approaches tar-
geting binding pockets, domains, or entire modules.
is a tool that has been used for protein engineering, for example
to modulate substrate specificity of the ene-reductase old yellow
enzyme (Daugherty et al., 2013). In another impressive design
effort, a (b/a)8 barrel was illegitimately recombined with a (b/a)4
motif excised from a flavodoxin fold (Bharat et al., 2008), thereby
swapping four strands of the barrel. Only a few mutations were
needed to restore stability and ligand binding capability to the
artificial fusion protein (Eisenbeis et al., 2012). Mixing and match-
ing of protein fragments has also been successfully applied to
The short length of the swapped sequence stretch
may facilitate straightforward exploitation of the
wealth of existing NRPS modules for combinatorial multimodular natural product synthetases (Calcott and Ackerley,
2014; Cane et al., 1998; Williams, 2013).
Domains of modular nonribosomal peptide synthetases
(NRPSs) (Tanovic et al., 2008), polyketide synthetases, and fatty
biosynthesis.
INTRODUCTION
acid synthetases (Smith and Tsai, 2007) lend themselves to en-
gineering by fragment recombination because they are con-
The evolutionary history of proteins provides a rich source of nected like beads on a string. NRPSs, for example, are
ideas worthy of imitation by protein engineers (Glasner et al., multienzyme clusters in which the number and order of domains
2007). These ideas extend beyond single point mutations. Pro- usually determines the sequence of the synthesized peptide.
tein evolution ‘‘one amino acid at a time’’ (Bloom and Arnold, Adenylation (A), condensation (C), and thiolation (T) domains
2009; Tracewell and Arnold, 2009) alone cannot account for work together in modules that elongate nascent peptides by
the vast diversity of proteins found in the biological world. In successive addition of amino acids. Permutation of NRPS
addition to incremental sequence optimization, genetic events domains and individual modules has shown great promise for
that perturb protein folds more drastically present shortcuts to the purposeful, combinatorial biosynthesis of novel peptides
remote areas of the fitness landscape (Grishin, 2001; Lupas (Calcott and Ackerley, 2014; Cane et al., 1998; Kries and Hilvert,
et al., 2001). For example, gene fusions, circular permutations, 2011; Williams, 2013).
and illegitimate recombinations between unrelated genes (Lupas
In a seminal study, Stachelhaus et al. engineered surfactin A
et al., 2001) mix and match folding units of proteins. Compari- synthetase by transplanting A domains to alter amino acid spec-
sons of bacterial genomes show that protein domains from ificities (Stachelhaus et al., 1995). Module deletions (Mootz et al.,
one species are sometimes found as freestanding proteins in 2002), insertions (Butz et al., 2008), and fusions of unrelated
640 Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved

Figure 1. A Flavodoxin-like Subdomain Is
A
Responsible for Substrate Binding in
Domains
A
(A) Topological map of the A domain fold, adapted
from Conti et al. (1997). Circles and arrows
symbolize helices and b-strands, respectively.
Approximate positions of binding pocket residues
are indicated in red. The subdomain of the protein
containing all variable binding pocket residues is
shown in blue.
A236
C331
I330
W239
D235
K517
(B) Cartoon structure of the subdomain in rainbow
colors from the N terminus (blue) to the C terminus
(red). The surface of the residues lining the L-Phe
binding pocket is shown as a red mesh. The sec-
ondary structure elements are characteristic of a
flavodoxin-type fold.
C-terminal
domain
A301
I299
A322
T278
2
1
3
4
5
6
(C) Homology score of an alignment of GrsAA with
all A domains from GrsB. The homology score
was calculated with jalview (Waterhouse et al.,
2009) taking into account the physicochemical
similarity of residues (Livingstone and Barton,
1993) and averaged over a window of 30 sequence
positions. Binding pocket residues are shown as
red diamonds.
B
C
10
9
8
7
6
5
4
3
2
1
3
4
5
6
0
100
200
300
400
500
600
position in alignment
modules (Duerfahrt et al., 2003; Mootz et al., 2000) identified determining binding pocket region. Transplantation of such sub-
additional degrees of freedom for combinatorial biosynthesis. domains into a phenylalanine-encoding A domain affords a
Mechanistic investigations of individual domains helped to chimera that functions in peptide formation.
circumvent unproductive module combinations (Belshaw et al.,
1999). Later, the discovery of small, terminal domains that RESULTS
mediate noncovalent module communication enabled the func-
tional connection of noncognate modules (Hahn and Stachel- Engineering Strategy
haus, 2004).
We scanned the GrsAA adenylation domain from the first module
Cutting and pasting individual domains or entire modules has of gramicidin S synthetase for a sequence stretch suitable for
become standard practice in NRPS engineering (Baltz, 2012; specificity engineering of NRPSs. Such a sequence stretch
Calcott et al., 2014; Duerfahrt et al., 2003; Mootz et al., 2000, should meet several criteria. First, all residues relevant for sub-
2002; Schauwecker et al., 2000; Schneider et al., 1998; Stachel- strate recognition in the A domain should be included. Second,
haus et al., 1995). Nevertheless, this approach is not always it should be short. Third, the tertiary structure should be
robust. In the past, the focus has been on changing the order compact. And fourth, functionally relevant domain interfaces
of the beads on the string—essentially the individual C, A, and should be avoided.
T domains—because the flexible linkages between them are
Inspection of the GrsAA structure (PDB: 1AMU) (Conti et al.,
believed to be rather insensitive to changes (Udwary et al., 1997) suggested a 132-residue-long segment encompassing
2002). However, the bead on a string metaphor is not necessarily the active site (T221 to I352) as a possible binding subdomain
the best guide for NRPS engineering. Spatial and temporal coor- (Figure 1). This protein fragment has a flavodoxin-like topology
dination of domain interactions is still poorly understood. As a (Figure 1): a parallel, five-stranded b-sheet sandwiched by a-he-
consequence, domain and module swaps often result in low ac- lices with the first two strands of the sheet in inverted order
tivities (Fischbach et al., 2007; Schneider et al., 1998; Stachel- (Eisenbeis et al., 2012). A sixth strand at the C terminus was
haus et al., 1995). Alternatively, synthetases can be tailored by included because it is part of the same b-sheet. The flavodoxin
fine-tuning individual domains, usually the specificity-deter- fold has been classified as one of nine ancestral protein folds
mining A domains (Eppelmann et al., 2002; Evans et al., 2011; to which all fold diversity observed today can be traced (Cae-
´
Kries et al., 2014; Thirlway et al., 2012), but the specificity tano-Anolles et al., 2007). NRPS adenylation domains belong
changes that can be achieved are typically conservative. to the ANL superfamily of enzymes (Gulick, 2009), which also in-
Here, we investigate an alternative strategy for modulating A cludes acyl-CoA synthetases and firefly luciferase. It is tempting
domain specificity in NRPSs that takes advantage of modularity to speculate that the evolutionary history of the ANL superfamily
on a subdomain structural level. We identify a compactly folded might have begun with a freestanding flavodoxin-like subunit
subdomain of the A domain that encompasses the specificity- that later became embedded within a larger scaffold.
Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved 641

Table 1. Michaelis-Menten Parameters of NRPS Modules
Enzyme
GrsB2
Substrate
L-Val
k
cat (minꢀ1
)
KM (mM) kcat/KM (mMꢀ1 minꢀ1
21 4.5 0.7
160 30 0.3 0.1
n.d. 0.07 0.03
)
10
5
100 30
50 20
n.d.a
3
sdV-GrsA L-Val
sdV-GrsA L-Phe
Michaelis-Menten parameters were measured for the adenylation partial
reaction in a pyrophosphate exchange assay (Otten et al., 2007). Errors
are given as the SD of at least four independent measurements with
different batches of protein.
an.d., not determined due to the absence of substrate saturation.
0
GrsA, where X stands for the amino acid specificity of the subdo-
main (X = P, V, O, L). The recombination partners and GrsAA had
pairwise sequence identities in the range of 44%–48% (Table S1)
and showed relatively low sequence conservation in the region
of the subdomain (Figure 1C; Table S1). An additional set of
five subdomains was obtained from several other organisms,
yielding constructs sdX2-GrsA (X = R, Q, L, F, and W). For the
second set, recombination partners possessing similarly high
levels of identity to GrsAA as the first set (38%–48%; Table S1)
but different specificities were retrieved from an annotated
Figure 2. Pyrophosphate Exchange Assay with Subdomain-
Swapped GrsA Variants
Exchange between 1 mM 32P-labeled pyrophosphate (PPi) and 5 mM ATP was
measured for 2 hr catalyzed by 5 mM enzyme in the presence of 1 mM amino
acid substrate. Background without amino acid is subtracted. Error bars
indicate the SD of four replicates with one batch of protein. With sdV-GrsA
additionally purified by anion exchange chromatography, PPi exchange in-
creases to 18%.
¨
collection of partial A domain sequences (K198 to T334) (Rottig
et al., 2011) by a BLAST search (Altschul et al., 1990).
All nine constructs were expressed as phosphopantetheiny-
The sequence of the binding subdomain comprises only 24%
of the 558 A domain residues, but includes nine of ten binding lated holoenzymes in Escherichia coli strain HM0079 (Gruene-
pocket residues correlated with substrate specificity (Challis wald et al., 2004) and purified by NiNTA affinity and anion
et al., 2000; Stachelhaus et al., 1999). In the absence of informa- exchange chromatography. Each construct was isolated as
tion about NRPS dynamics, it is unclear whether the binding sub- soluble protein in quantities ranging from 5 to 21 mg/l (GrsA,
domain is part of interaction networks relevant for substrate 26 mg/l; Figures S1A and S1B). Successful incorporation of the
recognition and catalysis. In the only available crystal structure phosphopantetheine prosthetic group into sdV-GrsA was indi-
of an entire module (SrfA-C; PDB: 2VSQ) (Tanovic et al., 2008), cated by electrospray ionization mass spectrometry (ESI-MS)
inevitably a static picture, the binding subdomain does not (Figure S2: calc., 128,848 Da; meas., 128,880 Da). Corroborating
engage in interdomain contacts. The structure of an A-T bido- evidence was obtained by MALDI-tandem mass spectrometry
main suggests, however, that certain subdomain residues are (MS/MS) analysis of the trypsin digested protein, which
probably involved in T domain binding (Sundlov et al., 2012).
yielded the mass for the phosphopantetheinylated peptide
Subdomain swapping occurred at least once during the evolu- IGIKDNFYALGGDS(ppant)IK (calc., 2050.992 Da; meas.,
tion of natural NRPSs. Genetic analysis of the NRPS for hormao- 2050.869 Da) with the cofactor attached to Ser578, as expected.
mycin revealed A domains with nucleotide identities in the range A fragment with a mass corresponding to the unmodified peptide
¨
of 90% but different specificities (Hofer et al., 2011). Within these was not found. Together, these results clearly show that the
almost identical A domain genes, only a stretch of ca. 400 base variant created by subdomain swapping is efficiently phospho-
pairs (bp) encoding active site residues showed lower identities. pantetheinylated by the genomically encoded Sfp. Circular di-
This observation was interpreted as the likely result of a genetic chroism (CD) spectra of six constructs, including sdV-GrsA,
recombination event that transferred the specificity-determining resemble that of GrsA (Figure S1C). Evidently, despite the heavy
portion of one domain into the protein scaffold of another. mutational load of more than 80 amino acid substitutions and
Indeed, using the sequence boundaries inferred from the natural two to four insertions or deletions, structural integrity was not
recombination event (K205 to A322 in GrsAA numbering), evolu- substantially compromised.
tion-guided genetic recombination of hormaomycin domains in
the laboratory succeeded in transplanting A domain specificity Adenylation Activity
in three of five cases (Cru¨ semann et al., 2013).
We measured adenylation activities of all nine constructs using
a pyrophosphate exchange assay (Otten et al., 2007). In an
initial screen, sdV-GrsA, sdL-GrsA, sdF2-GrsA, and sdL2-GrsA
Constructs
The feasibility of exploiting subdomain swaps for NRPS engi- showed significant pyrophosphate exchange activity stimulated
neering with our structure-guided boundaries was tested with by the respective target amino acid at 1 mM (Figure 2). Kinetic
nine constructs based on the D-Phe-encoding initiation module analysis of sdV-GrsA revealed a slightly reduced kcat (50 minꢀ1
)
GrsA. In addition to the canonical AT domain, GrsA contains for valine and an elevated KM (160 mM) compared with GrsB2,
an E domain for epimerization of the amino acid. Subdomains the source of the subdomain (kcat = 100 minꢀ1; KM = 21 mM;
were excised from all four A domains of the second protein Table 1). As a consequence, the apparent second-order rate
from gramicidin S synthetase, GrsB, yielding constructs sdX- constant kcat/KM for the chimera was 15-fold lower than for
642 Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved

Figure 3. Specificity of sdV-GrsA Is Remi-
niscent of Wild-Type A Domains
A
C
120
100
80
60
40
20
0
120
100
80
60
40
20
0
Adenylation kinetics of (A) GrsA, (B) sdV-GrsA, and
(C) GrsB2 were measured with all proteinogenic
amino acids as substrates (abbreviated in one-
letter code) at 1 mM substrate concentration. Re-
actions with the preferred amino acid substrate
were quenched at 10%–15% conversion and all
others were normalized to this value. Error bars
depict the SD from four measurements with the
same enzyme batch.
(D)
A homology model of sdV-GrsAA (left) in
ACDEFGH I K LMNPQRSTV
W
Y
ACDEFGH I K LMNPQRSTVWY
comparison with the crystal structure of GrsAA
(right) (Conti et al., 1997). The protein is cut away
in the plane of the amino acid binding pocket.
Amino acid ligands are shown as spheres (left,
L-Val; right, L-Phe; green, carbon; red, oxygen;
blue, nitrogen) and protein sidechains as cyan
sticks.
B
D
120
100
80
60
40
20
0
Lys517
Lys517
Asp235
Asp235
ACDEFGH I K LMNPQRSTVWY
GrsB2. Detailed kinetic characterization of other active con- with GrsAA (Figure 3D), whereas L-Phe would experience severe
structs proved challenging probably due to low kcat values.
In analogous experiments, Cru¨ semann et al. (2013) obtained
steric clashes.
active A domains upon swapping subdomains within the hor- DKP Synthesis
maomycin NRPS. However, constructs with subdomains from The ability of the subdomain-swapped sdV-GrsA to communi-
unrelated synthetases showed no activity. The observation of cate with other domains and modules was tested in an artificial
substantial adenylation activity in sdF2-GrsA and sdL2-GrsA diketopiperazine (DKP) synthetase consisting of sdV-GrsA and
(Figure 2) demonstrates that species borders need not be pro- GrsB1, the second module from gramicidin S synthesis (Fig-
hibitive for a subdomain swap approach.
ure 4A). GrsB1 is a typical proline-specific elongation module
with C, A, and T domains. DKP synthetases provide a convenient
sensor for peptide formation activity because the required mod-
Adenylation Specificity
Confronted with a plethora of cellular metabolites, including ules are reasonably small and can be expressed and assayed
many amino acids, NRPSs have to be specific, a property gener- in vitro (Gruenewald et al., 2004; Kries et al., 2014; Stachelhaus
ally attributed to A domains (Villiers and Hollfelder, 2009), which et al., 1998). The cyclized DKP products are readily detected by
activate the appropriate amino acid and covalently tether it to the liquid chromatography (LC)-MS.
assembly line. We mapped the substrate specificity of GrsA,
When combined with GrsB1, sdV-GrsA stimulated significant
sdV-GrsA, and the subdomain donor GrsB2 with all 20 standard DKP formation. The expected product, D-Val-L-Pro DKP, was de-
amino acids at 1 mM substrate concentration (Figures 3A–3C). tected by LC-MS along with 25% unepimerized L-Val-L-Pro DKP
The specificity profile shows that the sdV-GrsA chimera is as se- (Figure 4B). The engineered sdV-GrsA/GrsB1 synthetase incor-
lective as the native GrsA domain under these conditions (Villiers porated L-Val with a kobs of 0.003 minꢀ1 (Figure 4C), 300 times
and Hollfelder, 2009) but less discriminating than GrsB2. L-Val slower than the wild-type GrsA/GrsB1 system, which produces
gave the highest activity, as expected by design, followed by D-Phe-L-Pro DKP with a kobs of 0.9 minꢀ1 (Kries et al., 2014).
L-Leu and L-Phe. However, because the kinetics were measured Despite the large structural transition from the native substrate
at a single substrate concentration, this specificity profile is only L-Phe to L-Val, all domains remain at least partially functional.
a snapshot of the complex situation in the cell. The preference of
sdV-GrsA for L-Val over L-Phe, in terms of kcat/KM parameters DISCUSSION
(Table 1), is roughly four-fold.
A homology model of sdV-GrsAA, constructed on the Swiss- Combinatorial biosynthesis of natural products is a powerful
Model server (Biasini et al., 2014) using the crystal structure of strategy employed by natural evolution. The notion that modular
the GrsAA PheA domain (PDB: 1AMU) (Conti et al., 1997) as a machines like NRPSs and PKSs are also a valuable resource for
template, rationalizes the substrate preferences of the chimera. biosynthetic engineering in the laboratory has attracted broad
After manually docking the L-Val substrate and AMP into the attention over the last two decades (Calcott and Ackerley,
structure, the complex was optimized with the Rosetta Relax 2014; Cane et al., 1998; Hur et al., 2012; Kries and Hilvert,
protocol (Leaver-Fay et al., 2011). In qualitative agreement with 2011; Williams, 2013). While the chemical logic underlying effi-
the measured substrate profile (Figure 3B), L-Val fits snuggly in cient assembly of polyketides and nonribosomal peptides has
the modeled binding pocket, which is contracted compared been elucidated in detail, modifying these mega-enzymes for
Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved 643

(Kries et al., 2013). Third, technological innovations are helping
to unravel the dynamic processes in natural product synthetases
(Dutta et al., 2014; Whicher et al., 2014). Many problems associ-
ated with combinatorial biosynthesis today may thus be allevi-
ated in the near future.
A
sdV-GrsA
GrsB1
A
C
A
E
T
T
What would a reliable platform for the biosynthetic diversifica-
tion of nonribosomal peptides look like? Juggling genetic frag-
ments several kilobases long is clearly not an easy task. Our
findings support the feasibility of an alternative approach based
on genetic parts—binding subdomains—that are only ꢁ400 bp
long and, in contrast to genes of individual A domains or entire
modules, amenable to gene synthesis. Modern sequencing
technologies are providing metagenomic data from heteroge-
neous environmental samples, circumventing the need for
monoclonal DNA constructs. As a consequence, sequence da-
tabases are continuously growing and include data from organ-
isms that cannot be cultivated (Venter et al., 2004; Wilson and
Piel, 2013). Subdomain swaps with synthetic genes provide a
potentially powerful means to exploit these databases for combi-
natorial biosynthesis.
H
L
A
C
A
E
T
T
D
H
B
2.0
1.0
L
1)
D
2.0
1.0
2)
L
L
Subdomain exchange in NRPSs is likely to be an ancient
evolutionary strategy. Cru¨ semann et al. (2013) have argued
that swapping subdomains by evolutionary recombination
events shaped the hormaomycin gene cluster. Based on the
presence of a flavodoxin-like folding unit embedded within the
A domain structure (Figure 1), we inferred slightly different sub-
domain boundaries than those deduced for hormaomycin.
Both frames are similar but our subdomain stretch is shifted by
approximately one secondary structural element toward the
C terminus. Interestingly, Cru¨ semann et al.’s more N-terminal
evolution-guided frame does not include two residues of the
specificity code proposed by Stachelhaus et al. (1999) (residues
330 and 331 in GrsAA). One of these residues (amino acid 331) is
also missing in the specificity code proposed by Challis et al.
(2000) because it was argued to point away from the substrate.
Adenylation activities and specificity profiles of subdomain-
swapped constructs created with the more N-terminal frame
indicate that these residues are not necessary for specificity
transplantation in every case. Nevertheless, the beta hairpin
motif containing these residues directly contacts the substrate,
whereas the first helix of the evolution-guided subdomain,
which is not included in our structure-guided frame, contains
only second-shell residues. The more C-terminal frame is
altogether more proximal to the substrate binding pocket. In
the future, experimentally comparing different frames from one
subdomain source may help delineate the best frame for efficient
swaps.
2.0
1.0
3)
7.4
7.8
8.2
8.6
9.0
elution time (min)
C
2.0
1.5
1.0
0.5
0
0
50
100
time (min)
150
200
250
Figure 4. Peptide Synthesis by sdV-GrsA/GrsB1
(A) Mechanism of DKP formation. A, adenylation domain; E, epimerization
domain; C, condensation domain; T, thiolation domain.
(B) Extracted ion chromatograms (m/z 197.12 0.5) of a synthetic standard
for D-Val-L-Pro DKP (1) and L-Val-L-Pro DKP (2), and an enzymatic reaction with
sdV-GrsA and GrsB1 (3).
(C) Kinetics of Val-Pro DKP formation monitored by LC-MS. Error bars depict
the SD of three independent measurements with three batches of protein.
Errors were below 11% when measurements were repeated with the same set
of proteins. The large batch-to-batch variation is probably due to impurities in
GrsB1, which was obtained in low yield.
Our results extend those of Cru¨ semann et al. (2013) and
provide a proof-of-principle demonstration that a subdomain
swap strategy can yield active and selective A domain chi-
the production of novel antibiotics and other drugs has proved meras. In particular, our experiments show that such chimeric
challenging (Wilkinson and Micklefield, 2007).
A domains can function in the context of a dipeptide synthe-
Efforts directed at biosynthetic engineering of natural prod- tase. Although the yields are still low, the kinetic parameters
ucts should be redoubled for several reasons. First, the number determined for sdV-GrsA (Table 1) indicate that adenylation is
of building blocks for biosynthetic engineering is growing expo- not the kinetic bottleneck for DKP synthesis: sdV-GrsA has a
nentially in the postgenomic era with every microbial genome kcat of 50 minꢀ1, which exceeds DKP formation rates measured
deposited in a public database. Second, the field of enzyme with TycA/TycB1 or GrsA/GrsB1 (Belshaw et al., 1999; Gruene-
design and redesign is maturing as increasingly robust methods wald et al., 2004; Kries et al., 2014) by more than an order
for the design and optimization of biocatalysts are emerging of magnitude. Hence, DKP formation in sdV-GrsA/GrsB1 is
644 Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved

EXPERIMENTAL PROCEDURES
probably limited by one of the partial reactions following
adenylation.
General Cloning
Even though catalysis in the chimeras we studied was either
impaired to some extent or completely destroyed, repair of these
constructs might be relatively straightforward. A subdomain
swap is expected to perturb only a limited number of residues
at the hydrophobic interface between the subdomain and the
peripheral structure. Consequently, computational modeling,
coupled with laboratory evolution, might be exploited to identify
and remove inadvertently created voids and clashes (Evans
et al., 2011; Fischbach et al., 2007; Villiers and Hollfelder, 2011).
More detailed guidelines describing the individual strengths
and weaknesses of binding pocket, subdomain, domain, and
module swap for NRPS engineering should be established in
the future based on comprehensive and quantitative data. We
speculate that binding pocket mutagenesis will be the preferred
instrument to achieve small structural changes in substrate pref-
erence (Eppelmann et al., 2002; Evans et al., 2011; Kries et al.,
2014; Thirlway et al., 2012) since this approach is least likely
to wreak havoc on the structural integrity of the synthetase. How-
ever, second-shell and long-range interactions that may be
crucial for larger changes in specificity are ignored by this
approach. If longer fragments are swapped, more profound
changes of the NRPS become feasible albeit at the expense of
possible disruptions of the surrounding machinery. Subdomain
swapping represents a compromise that allows transplantation
of virtually all the selectivity determining residues and minimiza-
tion of larger disturbances to the enzymatic machinery, but activ-
ity loss is still a significant risk. Selectivity filters in downstream
domains, for instance the C domain, may cause additional
problems.
Procedures for cloning and media preparation were adapted from standard
protocols (Russell and Sambrook, 2001). General cloning was carried out in
Escherichia coli strain XL1-Blue (Stratagene). Microsynth AG synthesized the
custom oligonucleotides that were used as PCR primer (Table S2) and per-
formed Sanger sequencing of all inserts amplified by PCR. Restriction en-
zymes and DNA polymerases were purchased from New England BioLabs
Inc. Detailed protocols for plasmid isolation, gel electrophoresis, transforma-
tion, PCR, and ligation can be found in the Supplemental Information.
pSU18_grsB2_AT
For the cloning of expression plasmid pSU18_grsB2_AT encoding the AT bido-
main of the internal module GrsB2 (Table S3), the gene was amplified by PCR
from Aneurinibacillus migulanus genomic DNA in two fragments a and b.
Fragment a was amplified with primer pair grsb2_a_f/grsb2_EcoRI_del_r. Frag-
ment b was amplified with primer pair grsb2_EcoRI_del_f/grsb2_b_r. The
primer grsb2_EcoRI_del_f introduces a silent mutation in order to delete an
EcoRI site inside the gene. Fragments a and b were assembled by PCR
using primer pair grsb2_a_f/grsb2_b_r. The full-length gene was ligated into
the pSU18 vector via the EcoRI and BamHI restriction sites. The resulting
plasmid encodes residues K1549 to G2083 of GrsB2 (UniProt: P0C063), an
additional N-terminal methionine and a C-terminal –GSRSH6 tag.
Subdomain-Swapped pSU18_mgrsA Constructs
Construction of plasmids pMG211_mgrsAA0, pSU18_mgrsA, and pTrc99a_
grsB1 has been described previously (Kries et al., 2014). In plasmids
pMG211_mgrsAA0 and pSU18_mgrsA, the subdomain stretch of grsAA is
flanked by restriction sites that can be harnessed for exchanging the subdo-
main. The subdomain encoding sequences of grsB were amplified from
A. migulanus genomic DNA by PCR using primers mXbeg_f and mXend_r,
where X indicates the specificity of the subdomain in grsB in one-letter amino
acid code (X = V, L, O, P). At the 50-end of the subdomain, the mXbeg_f primer
spans the EcoRI restriction site and the beginning of the subdomain. In order
to append a stretch between the 30-end of the subdomain and the SacI
restriction site, a second fragment was amplified with primers mend_f and
T7TR. The overlapping fragments were assembled by PCR and cloned into
pMG211_mgrsAA0 via the EcoRI and SacI sites. From pMG211_mXgrsAA0
constructs, AflII/SacI fragments encoding the subdomain were cut out and
cloned into pSU18_mgrsA.
Further information will be required to ascertain why some
swaps yield active enzymes and others do not. The best predic-
tor of success might be sequence identity of the donor and
acceptor A domains (Table S1). Structural similarity of the donor
and acceptor substrate likely plays a role as well (Figure 2). An
interesting question is whether subdomain swaps also transfer
binding elements for MbtH-like proteins, which often copurify
with A domains and are, in certain cases, required for activity
(Felnagle et al., 2010; Zhang et al., 2010). Structural analysis of
an A domain bound to an MbtH-like protein shows that the sub-
domain is far away from the binding interface (Herbst et al.,
2013). If binding subdomains do not participate in MbtH binding,
their transplantation might be more successful than domain or
module exchanges.
A second generation of subdomain swap variants was constructed with syn-
thetic, codon-optimized gene fragments ordered from ATG:biosynthetics
GmbH that encode subdomains for the substrates Phe, Trp, Leu, Gln, and
Arg (Table S4). Sequences of these subdomains were retrieved from data-
bases (Table S3).
Protein Production, Purification and Mass Spectrometry
Phosphopantetheinylated proteins were expressed in E. coli HM0079 (Grue-
newald et al., 2004) and purified by affinity chromatography on NiNTA similar
to a previously described procedure (Kries et al., 2014). Bacterial cultures
were grown in LB medium containing 20 mg/ml chloramphenicol for pSU18
constructs and 150 mg/ml ampicillin for pTrc99a_grsB1. Protein production
was induced by adding 250 mM isopropyl b-D-1-thiogalactopyranoside to
500 ml of culture incubated in a rotary shaker at 37ꢂC and 250 rpm. Cells
were harvested by centrifugation after 16–20 hr at 18ꢂC and lysed by sonicat-
ion in 50 mM Tris-HCl (pH 7.4), 0.5 M NaCl supplemented with 1 mg/ml
chicken egg white lysozyme, 1 mM tris(2-carboxyethyl)phosphine (TCEP),
and protease inhibitor (Sigma, P8849). The cleared lysate was applied to
NiNTA columns. The columns were washed with 50 mM Tris-HCl (pH 7.4),
0.5 M NaCl containing 20 mM imidazole, and 1 mM TCEP and eluted with
the same buffer containing 300 mM imidazole. Protein was washed with
assay buffer in Amicon Ultra-15 centrifugal filters (30 kDa cut-off, Millipore).
Concentrations were measured spectrophotometrically using predicted
extinction coefficients at 280 nm. For determining Michaelis-Menten param-
eters, protein was additionally purified by anion exchange chromatography
on Mono Q HR 10/10 columns connected to a Biologic Duo Flow fast protein
liquid chromatography system (Bio-Rad Laboratories Inc.). Protein was
SIGNIFICANCE
Our results suggest that a design strategy inspired by the
fold architecture of A domains can be a viable alternative
to previous NRPS design approaches focusing on domains,
modules, and binding pocket mutagenesis. Since binding
subdomains are considerably shorter than domains or
modules, subdomain swapping could pave the way for
NRPS engineering based on bioinformatic searches and
gene synthesis. A simplified procedure for combinatorial
biosynthesis of nonribosomal peptides based on subdomain
swaps may accelerate the development of new peptide
drugs in the future.
Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved 645

eluted with a gradient of 20–400 mM NaCl in 20 mM Tris-HCl (pH 8). The
molecular mass of sdV-GrsA was determined after extensive washing with
D-Val (Bachem) was activated with N,N0-diisopropylcarbodiimide (1.1 eq.)
and DIPEA (3 eq.) for 2 min and coupled to the deprotected L-Pro resin for
4 hr. After alternating washing with CH2Cl2 and dimethylformamide (each
3 3 5 ml for 2 min), the coupling step was repeated overnight. The resulting
Boc-protected dipeptide was deprotected with TFA in CH2Cl2 (1:1, v/v) and
washed with CH2Cl2 (3 3 5 ml for 2 min). The deprotected dipeptide was
cleaved from the resin by cyclization with acetic acid in CH2Cl2 (0.1 M) for
1 hr and the resin was washed with CH2Cl2. Evaporation of solvent in vacuo
yielded pure D-Val-L-Pro DKP as a white solid (88% yield) without further puri-
fication. The 1H-NMR spectrum was in good agreement with a published refer-
ence (Hendea et al., 2006). 1H-NMR (300 MHz, CD3OD): d 4.14 (dd, J = 9.9,
6.5 Hz, 1H), 3.50 (dd, J = 6.1, 0.8 Hz, 1H), 3.59–3.32 (m, 2H), 2.32–2.16
(m, 1H), 2.04 (dq, J = 13.6, 6.8 Hz, 1H), 1.96–1.68 (m, 3H), 0.91 (dd, J = 9.8,
6.8 Hz, 6H). LC-MS, 197.13 m/z (M + H+, 197.12 m/z).
0.1% acetic acid by ESI-MS on
a Bruker maXis ESI-Qq-TOF-MS. For
MALDI-MS/MS, the sample was precipitated with 10% trichloroacetic acid,
washed with cold acetone, dissolved in 10 mM Tris (pH 8.2) buffer supple-
mented with 2 mM CaCl2, and digested with 500 ng of trypsin for 34 min at
60ꢂC in a microwave. The resulting peptide fragments were measured on a
Bruker Ultraflextreme.
Adenylation Kinetics
Adenylation kinetics were recorded with a 96-well pyrophosphate exchange
assay as described previously (Kries et al., 2014; Otten et al., 2007). Reactions
were conducted at room temperature (22–24ꢂC) in B&W Isoplate-96 plates
(Perkin Elmer) in a volume of 60 ml of buffer (50 mM Tris-HCl [pH 7.4],
10 mM MgCl2, 1 mM TCEP, 0.1 mg/ml BSA) in the presence of 0.1 mM inor-
ganic pyrophosphate (PPi) labeled with ca. 0.02 mCi 32P-PPi (Perkin Elmer),
2 mM ATP (Sigma A7699), and varying enzyme and substrate concentrations.
In the activity screen (Figure 2), higher concentrations of PPi (1 mM) and ATP
(5 mM) were employed. Reactions were quenched with 60 ml of a 1.6% char-
coal suspension in an aqueous solution of 3.5% HClO4 and 3.5% Na4P2O7.
The charcoal was pelleted by centrifugation and washed twice with 200 ml
of water. After resuspending the charcoal in 150 ml of scintillation mix
(Optiphase Supermix, Perkin Elmer), scintillation was measured in a Micro-
plate Scintillation & Luminescence Counter TopCount NXT (Perkin Elmer).
Steady-state parameters were determined by fitting initial velocities of 32P-
PPi incorporation to the Michaelis-Menten equation in Kaleidagraph (Synergy
Software).
SUPPLEMENTAL INFORMATION
Supplemental Information includes supplemental protocols, two figures, and
four tables and can be found with this article online at http://dx.doi.org/10.
1016/j.chembiol.2015.04.015.
AUTHOR CONTRIBUTIONS
H.K., D.N., and D.H. designed the research; D.N. performed the experiments
including cloning, protein production, kinetic assays, CD spectroscopy, and
chemical synthesis with assistance from H.K.; H.K. performed sequence ana-
lyses and computational docking; H.K., D.N., and D.H. analyzed the data and
wrote the paper.
CD Spectroscopy
CD of the protein samples was measured on a JASCO J-715 spectropolarim-
eter from 200 to 260 nm in 0.5-nm steps at a scanning speed of 20 nm/min and
a response time of 2 s. Samples were kept at 25ꢂC in a quartz cuvette with a
pathlength of 0.2 cm. Protein was diluted in low-salt phosphate buffer
(10 mM Na2HPO4 [pH 7.5], 10 mM NaCl). A buffer blank was subtracted and
curves were smoothed by averaging ellipticities from four wavelengths.
Mean residual ellipticities (MRE) were calculated according to the following
equation: MRE = Qobs/(10 3 lcn), where Qobs is the ellipticity measured in de-
grees, l the optical pathlength in cm, c is the molar enzyme concentration, and
n is the number of protein residues.
ACKNOWLEDGMENTS
We gratefully acknowledge the gift of strain HM0079 from Professor Mohamed
¨
A. Marahiel (Philipps Universitat Marburg). We thank Professor Hans-Martin
Fischer (ETH Zu¨ rich) for assistance with the radiochemical experiments. We
also thank Louis Bertschi and Dr. Xiangyang Zhang (Mass Spectrometry Ser-
vice Facility, Laboratory of Organic Chemistry, ETH Zu¨ rich) and the Functional
Genomics Center Zurich for performing the MS analyses. This work was finan-
cially supported by the Studienstiftung des deutschen Volkes (H.K.), the
Stipendienfonds der Schweizer Chemischen Industrie (H.K.), the Schweizer
Nationalfonds and ETH Zu¨ rich.
DKP Formation Assay
Module sdV-GrsA was tested for DKP formation in combination with GrsB1 at
5 mM concentration of each protein in 50 mM HEPES buffer supplemented with
10 mM MgCl2, 100 mM NaCl, 5 mM ATP, 1 mM TCEP, and 1 mM L-Val and
L-Pro. HEPES buffer, MgCl2, and NaCl were added from a 20-fold stock solu-
tion adjusted to pH 8.0. Reactions were conducted at 37ꢂC in a volume of
500 ml. Aliquots of 100 ml were removed at the indicated time points and
quenched by heat denaturation for 3 min at 95ꢂC. For LC-MS analysis,
protein was removed by centrifugation for 5 min at 16,000 g. Samples were
analyzed on an Agilent HPLC system (1200) equipped with a 3 3 30 mm Agilent
RP-C18 column (3.5 mm particle size) connected to a Bruker maXis ESI-Qq-
TOF mass spectrometer. Comparison of peak areas in extracted ion chro-
matograms to a linear regression curve obtained with an authentic standard
allowed quantification of the DKP product (Figure 4C). Diastereomers of Val-
Pro DKP were separated on an Xbridge C18 column with 3.5-mm particle
size (Figure 4B).
Received: January 27, 2015
Revised: March 31, 2015
Accepted: April 15, 2015
Published: May 21, 2015
REFERENCES
Altschul, S.F., Gish, W., Miller, W., Myers, E.W., and Lipman, D.J. (1990). Basic
local alignment search tool. J. Mol. Biol. 215, 403–410.
Baltz, R.H. (2012). Combinatorial biosynthesis of cyclic lipopeptide antibiotics:
a model for synthetic biology to accelerate the evolution of secondary metab-
olite biosynthetic pathways. ACS Synth. Biol. 3, 748–758.
Belshaw, P.J., Walsh, C.T., and Stachelhaus, T. (1999). Aminoacyl-CoAs as
probes of condensation domain selectivity in nonribosomal peptide synthesis.
Science 284, 486–489.
DKP Standards
Bharat, T.A.M., Eisenbeis, S., Zeth, K., and Ho¨ cker, B. (2008). A ba-barrel built
by the combination of fragments from different folds. Proc. Natl. Acad. Sci.
USA 105, 9942–9947.
L-Val-L-Pro DKP was purchased from Bachem (G-4730) as a standard for LC-
MS. D-Val-L-Pro DKP was synthesized by standard Boc solid-phase peptide
synthesis on a Merrifield resin according to published protocols (Gisin and
Merrifield, 1972). All synthesis steps were conducted at room temperature in
a 15-ml reaction vessel equipped with a glass frit and a nitrogen inlet for agita-
tion. Boc-protected L-Pro precoupled to Merrifield resin (Bachem, 0.9 mmol/g,
200–400 mesh) was deprotected with trifluoroacetic acid (TFA) in CH2Cl2
(1:1, v/v) (2 3 5 ml for 15 min), washed with CH2Cl2 (3 3 5 ml for 2 min), neutral-
ized with diisopropylethyl amine (DIPEA) in CH2Cl2 (1:19, v/v; 2 3 5 ml for
3 min), and washed again with CH2Cl2 (3 3 5 ml for 2 min). Boc-protected
Biasini, M., Bienert, S., Waterhouse, A., Arnold, K., Studer, G., Schmidt, T.,
Kiefer, F., Cassarino, T.G., Bertoni, M., Bordoli, L., et al. (2014). SWISS-
MODEL: modelling protein tertiary and quaternary structure using evolutionary
information. Nucleic Acids Res. 42, W252–W258.
Bloom, J.D., and Arnold, F.H. (2009). In the light of directed evolution: path-
ways of adaptive protein evolution. Proc. Natl. Acad. Sci. USA 106, 9995–
10000.
646 Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved

Butz, D., Schmiederer, T., Hadatsch, B., Wohlleben, W., Weber, T., and
Su¨ ssmuth, R.D. (2008). Module extension of a non-ribosomal peptide synthe-
tase of the glycopeptide antibiotic balhimycin produced by Amycolatopsis bal-
himycina. ChemBioChem 9, 1195–1200.
Grishin, N.V. (2001). Fold change in evolution of protein structures. J. Struct.
Biol. 134, 167–185.
Gruenewald, S., Mootz, H.D., Stehmeier, P., and Stachelhaus, T. (2004). In vivo
production of artificial nonribosomal peptide products in the heterologous host
Escherichia coli. Appl. Environ. Microbiol. 70, 3282–3291.
´
Caetano-Anolles, G., Kim, H.S., and Mittenthal, J.E. (2007). The origin of mod-
ern metabolic networks inferred from phylogenomic analysis of protein archi-
Gulick, A.M. (2009). Conformational dynamics in the acyl-CoA synthetases,
adenylation domains of non-ribosomal peptide synthetases, and firefly lucif-
erase. ACS Chem. Biol. 4, 811–827.
tecture. Proc. Natl. Acad. Sci. USA 104, 9358–9363.
Calcott, M.J., and Ackerley, D.F. (2014). Genetic manipulation of non-ribo-
somal peptide synthetases to generate novel bioactive peptide products.
Biotechnol. Lett. 36, 2407–2416.
Hahn, M., and Stachelhaus, T. (2004). Selective interaction between nonribo-
somal peptide synthetases is facilitated by short communication-mediating
domains. Proc. Natl. Acad. Sci. USA 101, 15585–15590.
Calcott, M.J., Owen, J.G., Lamont, I.L., and Ackerley, D.F. (2014). Biosynthesis
of novel pyoverdines by domain substitution in a nonribosomal peptide
synthetase of Pseudomonas aeruginosa. Appl. Environ. Microbiol. 80, 5723–
5731.
Hendea, D., Laschat, S., Baro, A., and Frey, W. (2006). Diastereoselective
alkylation of a proline-derived bicyclic lactim ether. Helv. Chim. Acta 89,
1894–1909.
Cane, D.E., Walsh, C.T., and Khosla, C. (1998). Harnessing the biosynthetic
Herbst, D.A., Boll, B., Zocher, G., Stehle, T., and Heide, L. (2013). Structural
basis of the interaction of MbtH-like proteins, putative regulators of nonriboso-
mal peptide biosynthesis, with adenylating enzymes. J. Biol. Chem. 288,
1991–2003.
code: combinations, permutations, and mutations. Science 282, 63–68.
Challis, G.L., Ravel, J., and Townsend, C.A. (2000). Predictive, structure-
based model of amino acid recognition by nonribosomal peptide synthetase
adenylation domains. Chem. Biol. 7, 211–224.
Ho¨ fer, I., Cru¨ semann, M., Radzom, M., Geers, B., Flachshaar, D., Cai, X.,
Zeeck, A., and Piel, J. (2011). Insights into the biosynthesis of hormaomycin,
an exceptionally complex bacterial signaling metabolite. Chem. Biol. 18,
381–391.
Conti, E., Stachelhaus, T., Marahiel, M.A., and Brick, P. (1997). Structural basis
for the activation of phenylalanine in the non-ribosomal biosynthesis of gram-
icidin S. EMBO J. 16, 4174–4183.
Crameri, A., Raillard, S.A., Bermudez, E., and Stemmer, W.P.C. (1998). DNA
shuffling of a family of genes from diverse species accelerates directed evolu-
tion. Nature 391, 288–291.
Hur, G.H., Vickery, C.R., and Burkart, M.D. (2012). Explorations of catalytic do-
mains in non-ribosomal peptide synthetase enzymology. Nat. Prod. Rep. 29,
1074–1098.
Cru¨ semann, M., Kohlhaas, C., and Piel, J. (2013). Evolution-guided engineer-
ing of nonribosomal peptide synthetase adenylation domains. Chem. Sci. 4,
1041–1045.
Ja¨ ckel, C., and Hilvert, D. (2010). Biocatalysts by evolution. Curr. Opin.
Biotechnol. 21, 753–759.
Kries, H., and Hilvert, D. (2011). Tailor-made peptide synthetases. Chem. Biol.
Daugherty, A.B., Govindarajan, S., and Lutz, S. (2013). Improved biocatalysts
from a synthetic circular permutation library of the flavin-dependent oxidore-
ductase old yellow enzyme. J. Am. Chem. Soc. 135, 14425–14432.
18, 1206–1207.
Kries, H., Blomberg, R., and Hilvert, D. (2013). De novo enzymes by computa-
tional design. Curr. Opin. Chem. Biol. 17, 221–228.
Duerfahrt, T., Doekel, S., Sonke, T., Quaedflieg, P., and Marahiel, M.A. (2003).
Construction of hybrid peptide synthetases for the production of a-L-aspartyl-
L-phenylalanine, a precursor for the high-intensity sweetener aspartame. Eur.
J. Biochem. 270, 4555–4563.
Kries, H., Wachtel, R., Pabst, A., Wanner, B., Niquille, D., and Hilvert, D. (2014).
Reprogramming nonribosomal peptide synthetases for ‘‘clickable’’ amino
acids. Angew. Chem. Int. Ed. Engl. 53, 10105–10108.
Lang, D., Thoma, R., Henn-Sax, M., Sterner, R., and Wilmanns, M. (2000).
Structural evidence for evolution of the b/a barrel scaffold by gene duplication
and fusion. Science 289, 1546–1550.
Dutta, S., Whicher, J.R., Hansen, D.A., Hale, W.A., Chemler, J.A., Congdon,
˚
G.R., Narayan, A.R.H., Hakansson, K., Sherman, D.H., Smith, J.L., et al.
(2014). Structure of a modular polyketide synthase. Nature 510, 512–517.
Leaver-Fay, A., Tyka, M., Lewis, S.M., Lange, O.F., Thompson, J., Jacak, R.,
Kaufman, K., Renfrew, P.D., Smith, C.A., Sheffler, W., et al. (2011).
ROSETTA3: an object-oriented software suite for the simulation and design
of macromolecules. Methods Enzymol. 487, 545–574.
Eisenbeis, S., Proffitt, W., Coles, M., Truffault, V., Shanmugaratnam, S., Meiler,
¨
J., and Hocker, B. (2012). Potential of fragment recombination for rational
design of proteins. J. Am. Chem. Soc. 134, 4019–4022.
Enright, A.J., Iliopoulos, I., Kyrpides, N.C., and Ouzounis, C.A. (1999). Protein
interaction maps for complete genomes based on gene fusion events. Nature
402, 86–90.
Livingstone, C.D., and Barton, G.J. (1993). Protein sequence alignments: a
strategy for the hierarchical analysis of residue conservation. Comput. Appl.
Biosci. 9, 745–756.
Eppelmann, K., Stachelhaus, T., and Marahiel, M.A. (2002). Exploitation of the
selectivity-conferring code of nonribosomal peptide synthetases for the
rational design of novel peptide antibiotics. Biochemistry 41, 9718–9726.
Lupas, A.N., Ponting, C.P., and Russell, R.B. (2001). On the evolution of
protein folds: are similar motifs in different protein folds the result of conver-
gence, insertion, or relics of an ancient peptide world? J. Struct. Biol. 134,
191–203.
Evans, B.S., Chen, Y., Metcalf, W.W., Zhao, H., and Kelleher, N.L. (2011).
Directed evolution of the nonribosomal peptide synthetase AdmK generates
new andrimid derivatives in vivo. Chem. Biol. 18, 601–607.
Minshull, J., and Stemmer, W.P.C. (1999). Protein evolution by molecular
breeding. Curr. Opin. Chem. Biol. 3, 284–290.
¨
Farı´as-Rico, J.A., Schmidt, S., and Hocker, B. (2014). Evolutionary relationship
Mootz, H.D., Schwarzer, D., and Marahiel, M.A. (2000). Construction of hybrid
peptide synthetases by module and domain fusions. Proc. Natl. Acad. Sci.
USA 97, 5848–5853.
of two ancient protein superfolds. Nat. Chem. Biol. 10, 710–715.
Felnagle, E.A., Barkei, J.J., Park, H., Podevels, A.M., McMahon, M.D., Drott,
D.W., and Thomas, M.G. (2010). MbtH-like proteins as integral components
of bacterial nonribosomal peptide synthetases. Biochemistry 49, 8815–8817.
Mootz, H.D., Kessler, N., Linne, U., Eppelmann, K., Schwarzer, D., and
Marahiel, M.A. (2002). Decreasing the ring size of a cyclic nonribosomal pep-
tide antibiotic by in-frame module deletion in the biosynthetic genes. J. Am.
Chem. Soc. 124, 10980–10981.
Fischbach, M.A., Lai, J.R., Roche, E.D., Walsh, C.T., and Liu, D.R. (2007).
Directed evolution can rapidly improve the activity of chimeric assembly-line
enzymes. Proc. Natl. Acad. Sci. USA 104, 11951–11956.
Otten, L.G., Schaffer, M.L., Villiers, B.R.M., Stachelhaus, T., and Hollfelder, F.
(2007). An optimized ATP/PPi-exchange assay in 96-well format for screening
of adenylation domains for applications in combinatorial biosynthesis.
Biotechnol. J. 2, 232–240.
Gisin, B.F., and Merrifield, R.B. (1972). Carboxyl-catalyzed intramolecular ami-
nolysis. A side reaction in solid-phase peptide synthesis. J. Am. Chem. Soc.
94, 3102–3106.
¨
Rottig, M., Medema, M.H., Blin, K., Weber, T., Rausch, C., and Kohlbacher, O.
Glasner, M.E., Gerlt, J.A., and Babbitt, P.C. (2007). Mechanisms of protein
evolution and their application to protein engineering. Adv. Enzymol. Relat.
Areas Mol. Biol. 75, 193–239.
(2011). NRPSpredictor2-a web server for predicting NRPS adenylation domain
specificity. Nucleic Acids Res. 39, W362–W367.
Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved 647

Russell, D.W., and Sambrook, J. (2001). Molecular Cloning: A Laboratory
Manual (Cold Spring Harbor).
Tracewell, C.A., and Arnold, F.H. (2009). Directed enzyme evolution: climbing
fitness peaks one amino acid at a time. Curr. Opin. Chem. Biol. 13, 3–9.
Schauwecker, F., Pfennig, F., Grammel, N., and Keller, U. (2000). Construction
and in vitro analysis of a new bi-modular polypeptide synthetase for synthesis
of N-methylated acyl peptides. Chem. Biol. 7, 287–297.
Udwary, D.W., Merski, M., and Townsend, C.A. (2002). A method for prediction
of the locations of linker regions within large multifunctional proteins, and
application to a type I polyketide synthase. J. Mol. Biol. 323, 585–598.
Schneider, A., Stachelhaus, T., and Marahiel, M.A. (1998). Targeted alteration
of the substrate specificity of peptide synthetases by rational module swap-
ping. Mol. Gen. Genet. 257, 308–318.
Venter, J.C., Remington, K., Heidelberg, J.F., Halpern, A.L., Rusch, D., Eisen,
J.A., Wu, D., Paulsen, I., Nelson, K.E., Nelson, W., et al. (2004). Environmental
genome shotgun sequencing of the Sargasso Sea. Science 304, 66–74.
Smith, S., and Tsai, S.-C. (2007). The type I fatty acid and polyketide syn-
Villiers, B.R.M., and Hollfelder, F. (2009). Mapping the limits of substrate spec-
thases: a tale of two megasynthases. Nat. Prod. Rep. 24, 1041–1072.
ificity of the adenylation domain of TycA. ChemBioChem 10, 671–682.
Stachelhaus, T., Schneider, A., and Marahiel, M.A. (1995). Rational design of
peptide antibiotics by targeted replacement of bacterial and fungal domains.
Science 269, 69–72.
Villiers, B., and Hollfelder, F. (2011). Directed evolution of a gatekeeper domain
in nonribosomal peptide synthesis. Chem. Biol. 18, 1290–1299.
Waterhouse, A.M., Procter, J.B., Martin, D.M.A., Clamp, M., and Barton, G.J.
(2009). Jalview Version 2-a multiple sequence alignment editor and analysis
workbench. Bioinformatics 25, 1189–1191.
Stachelhaus, T., Mootz, H.D., Bergendahl, V., and Marahiel, M.A. (1998).
Peptide bond formation in nonribosomal peptide biosynthesis. Catalytic role
of the condensation domain. J. Biol. Chem. 273, 22773–22781.
Whicher, J.R., Dutta, S., Hansen, D.A., Hale, W.A., Chemler, J.A., Dosey, A.M.,
˚
Narayan, A.R.H., Hakansson, K., Sherman, D.H., Smith, J.L., et al. (2014).
Stachelhaus, T., Mootz, H.D., and Marahiel, M.A. (1999). The specificity-
conferring code of adenylation domains in nonribosomal peptide synthetases.
Chem. Biol. 6, 493–505.
Structural rearrangements of a polyketide synthase module during its catalytic
cycle. Nature 510, 560–564.
Sundlov, J.A., Shi, C., Wilson, D.J., Aldrich, C.C., and Gulick, A.M. (2012).
Structural and functional investigation of the intermolecular interaction be-
tween NRPS adenylation and carrier protein domains. Chem. Biol. 19,
188–198.
Wilkinson, B., and Micklefield, J. (2007). Mining and engineering natural-
product biosynthetic pathways. Nat. Chem. Biol. 3, 379–386.
Williams, G.J. (2013). Engineering polyketide synthases and nonribosomal
peptide synthetases. Curr. Opin. Struct. Biol. 23, 603–612.
Tanovic, A., Samel, S.A., Essen, L.-O., and Marahiel, M.A. (2008). Crystal
structure of the termination module of a nonribosomal peptide synthetase.
Science 321, 659–663.
Wilson, M.C., and Piel, J. (2013). Metagenomic approaches for exploiting un-
cultivated bacteria as a resource for novel biosynthetic enzymology. Chem.
Biol. 20, 636–647.
Thirlway, J., Lewis, R., Nunns, L., Al Nakeeb, M., Styles, M., Struck, A.-W.,
Smith, C.P., and Micklefield, J. (2012). Introduction of a non-natural amino
acid into a nonribosomal peptide antibiotic by modification of adenylation
domain specificity. Angew. Chem. Int. Ed. Engl. 51, 7181–7184.
Zhang, W., Heemstra, J.R., Jr., Walsh, C.T., and Imker, H.J. (2010). Activation
of the pacidamycin PacL adenylation domain by MbtH-like proteins.
Biochemistry 49, 9946–9947.
648 Chemistry & Biology 22, 640–648, May 21, 2015 ª2015 Elsevier Ltd All rights reserved