Engineering the substrate specificity of the dhbe adenylation domain by yeast cell surface display

Article abstract of DOI:10.1016/j.chembiol.2012.10.020

The adenylation (A) domains of nonribosomal peptide synthetases (NRPSs) activate aryl acids or amino acids to launch their transfer through the NRPS assembly line for the biosynthesis of many medicinally important natural products. In order to expand the substrate pool of NRPSs, we developed a method based on yeast cell surface display to engineer the substrate specificities of the A-domains. We acquired A-domain mutants of DhbE that have 11- and 6-fold increases in k_cat/K_m with nonnative substrates 3-hydroxybenzoic acid and 2-aminobenzoic acid, respectively and corresponding 3- and 33-fold decreases in k_cat/K_m values with the native substrate 2,3-dihydroxybenzoic acid, resulting in a dramatic switch in substrate specificity of up to 200-fold. Our study demonstrates that yeast display can be used as a high throughput selection platform to reprogram the nonribosomal code of A-domains.

Full text of DOI:10.1016/j.chembiol.2012.10.020

Chemistry & Biology
Article
Engineering the Substrate Specificity
of the DhbE Adenylation Domain
by Yeast Cell Surface Display
Keya Zhang,1 Kathryn M. Nelson,2 Karan Bhuripanyo,1 Kimberly D. Grimes,2 Bo Zhao,1 Courtney C. Aldrich,2,
*
and Jun Yin1,
*
1Department of Chemistry, University of Chicago, Chicago, IL 60637, USA
2Center for Drug Design, University of Minnesota, Minneapolis, MN 55455, USA
*Correspondence: aldri015@umn.edu (C.C.A.), junyin@uchicago.edu (J.Y.)
http://dx.doi.org/10.1016/j.chembiol.2012.10.020
SUMMARY
2005). An NRPS extension module is composed of three core
domains including a condensation (C) domain, an adenylation
The adenylation (A) domains of nonribosomal (A) domain and a peptidyl carrier protein (PCP) domain, while
the initial loading module requires only the A and PCP domains.
The A-domain is responsible for the selection, activation, and
peptide synthetases (NRPSs) activate aryl acids or
amino acids to launch their transfer through the
loading of its cognate amino acid substrate onto the downstream
PCP-domain where it is covalently attached via a thioester
linkage (Figure 1A). The thiol moiety from the PCP-domain is
NRPS assembly line for the biosynthesis of many
medicinally important natural products. In order to
expand the substrate pool of NRPSs, we developed
a method based on yeast cell surface display to engi-
neer the substrate specificities of the A-domains.
We acquired A-domain mutants of DhbE that have
not from cysteine but rather from the terminal thiol moiety of an
˚
approximately 20 A long phosphopantetheine (Ppant) cofactor
moiety that modifies a conserved serine residue of the PCP-
domain. Movements of the PCP-domain and the Ppant arm
11- and 6-fold increases in kcat/Km with nonnative
substrates 3-hydroxybenzoic acid and 2-aminoben-
deliver the amino acid substrates to the C-domain that catalyzes
peptide bond coupling between substrate molecules loaded on
zoic acid, respectively and corresponding 3- and neighboring modules. This leads to the elongation of the peptide
chain in the N / C direction. NRPSs thus work in an assembly-
line fashion with the growing peptide chain being passed down-
stream from one module to the next until it reaches the last
module. The full-length peptide chain is released by a thio-
esterase domain, usually via macrocyclization, to afford the final
product.
33-fold decreases in kcat/Km values with the native
substrate 2,3-dihydroxybenzoic acid, resulting in
a dramatic switch in substrate specificity of up to
200-fold. Our study demonstrates that yeast display
can be used as a high throughput selection plat-
form to reprogram the ‘‘nonribosomal code’’ of A-
domains.
A-domains are the gatekeepers of NRPSs since they are
responsible for selection of the appropriate carboxylic acid
substrate, which is usually an amino acid but can also be an
aryl acid as found in siderophore NRPs, a fatty acid for lipopep-
tide NRPs, or a hydroxy acid for peptide ester containing
NRPs (Gulick, 2009; Sieber and Marahiel, 2005). A-domains
INTRODUCTION
Nonribosomal peptide synthetases (NRPSs) are large multifunc- are 55–60 kDa and contain a large N-terminal subdomain
tional enzymes that synthesize peptide natural products and a small C-terminal subdomain with the active site located
known as nonribosomal peptides (NRPs), which are structurally at the domain interface. These proteins are conformationally
diverse and possess many important medicinal activities (Cane dynamic and exist in an open unliganded conformation as well
et al., 1998; Clardy and Walsh, 2004). The anticancer agent bleo- as two different closed ligand-bound conformations. The
mycin; the immunosuppresant cyclosporin; and the antibiotics carboxylic acid substrate binding pocket is lined by 10 residues
vancomycin, daptomycin, and capreomycin are all examples of that comprise the first-shell interactions with the substrate and
nonribosomal peptide drugs approved by the Food and Drug define the ‘‘nonribosomal codes’’ that enable the in silico predic-
Administration. Since NRPSs do not use the mRNA-templated tion of A-domain substrate specificity (Challis et al., 2000; Sta-
¨
ribosomal machinery, they are not restricted to the 20 proteino- chelhaus et al., 1999; von Dohren et al., 1999). A-domains cata-
genic amino acids and often contain D-amino acids as well as lyze a two-step adenylation-thioesterification reaction; in the first
unnatural a-amino acids. Other conspicuous features of NRPSs step, they bind the substrate acid and ATP to afford a ternary
include N-methylation and cyclization of the peptide backbone, complex and then catalyze their condensation to afford an inter-
both of which serve to enhance proteolytic stability. NRPSs mediate acyl-adenylate (Sikora et al., 2010) (Figure 1A).
utilize a modular architecture where each module is responsible Following release of pyrophosphate, the C terminus of the
for the incorporation of one amino acid substrate into the final A-domain undergoes an approximately 140ꢀ rigid body rotation,
molecule (Fischbach and Walsh, 2006; Sieber and Marahiel, which enables binding of the downstream PCP-domain and
92 Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
Figure 1. DhbE Catalyzed Aryl Acid Activation and the Yeast Selection Scheme to Change the Substrate Specificity of DhbE
(A) DhbE catalyzes the condensation of DHB 1 with ATP to form DHB-AMP 2. The activated DHB is then transferred to the ArCP domain of DhbB to form
a thioester conjugate 3 with the Ppant group of ArCP. DhbE can also activate SA 4. SA-AMS conjugate 5 is a bisubstrate inhibitor of DhbE. Structures of 3-HBA 6
and 2-ABA 7 as examples of nonnative substrates of DhbE are also shown.
(B) Selection of the A-domain library displayed on the surface of yeast cells. AMS-biotin conjugated SA (8) is used in the model selection to test the binding of
wtDhbE on yeast cell surface with substrate-AMS conjugate. Compounds 9 and 10 have nonnative substrates 3-HBA 6 and 2-ABA 7 conjugated to AMS-biotin.
They were used in the selection of DhbE mutants by yeast cell surface display.
See also Figures S1–S14 and Tables S1 and S3.
insertion of the Ppant cofactor arm into the A-domain active site
As a first step to test our method, we selected the A-domain of
˚
(Gulick, 2009). Acylation of the Ppant moiety followed by release DhbE as a model system due to the availability of its 2.15 A X-ray
of the thioacylated PCP and AMP completes the catalytic cycle. cocrystal structure with bound acyl-adenylate (May et al., 2002),
Several strategies have been explored to modify the substrate availability of active-site directed inhibitors (Miethke et al., 2006),
specificity of A-domains in NRPS assembly lines to incorporate and detailed knowledge of the kinetic mechanism (Sikora et al.,
nonnative building blocks to synthesize analogs with improved 2010). DhbE is part of a three-module NRPS assembly line
or alternate biological activities (Cane et al., 1998; Nguyen comprising DhbE, DhbB, and DhbF responsible for the synthesis
et al., 2006). Domain swapping to produce chimeric proteins of the siderophore bacillibactin in Bacillus subtilis (May et al.,
was first used to alter A-domain specificity; however, this 2001). DhbE activates 2,3-dihydroxybenzoic acid (DHB, 1) to
approach suffers from nonoptimal interactions between noncog- form the DHB-AMP acyl-adenylate 2 and transfers it to the aryl
nate protein domains that often result in drastically reduced carrier protein (ArCP) domain of DhbB (Figure 1A). Another
catalytic turnover, as well as premature truncation of the nascent distinct advantage of DhbE is that it is not embedded in a multi-
peptide (Baltz, 2008; Fischbach et al., 2007; Robbel and domain NRPS but is a stand-alone protein, which loads the
Marahiel, 2010; Stachelhaus et al., 1995). In an effort to alleviate carrier protein domain in DhbB in an intermolecular fashion.
these problems, Walsh and coworkers used directed evolution Consequently, DhbE is amenable to kinetic characterization of
to improve the activity of chimeric assembly lines, which led to the entire adenylation-thioesterification reaction as one can
nearly 10-fold improvements in catalytic activity (Fischbach use stoichiometric amounts of its cognate ArCP domain of
et al., 2007). A complimentary approach to reengineer A-domain DhbB to obtain catalytic turnover. By contrast, most previous
specificity preserving the native protein-protein interactions reports for A-domain engineering where kinetic data are given
within an NRPS module has been to use the nonribosomal have only measured the adenylation partial reaction, as catalytic
peptide code and, more recently, computational structure- turnover was not possible with the given multidomain NRPS
based redesign to rationally mutate residues important for protein, and thus do not report on the kinetically and functionally
substrate recognition (Chen et al., 2009; Eppelmann et al., relevant overall reaction (Chen et al., 2009; Eppelmann et al.,
2002; Thirlway et al., 2012). In order to screen potential mutants 2002).
more comprehensively, Kelleher and co-workers developed
a novel workflow involving directed evolution to create a library RESULTS
of ꢁ1,000 mutants based on the ‘‘nonribosomal peptide code’’
and mass spectrometry to measure the ability of the resulting Engineer A-Domain Specificity by Yeast Cell Surface
mutants to support production of novel NRPs (Evans et al., Display
2011). In this study, we report a method to engineer the substrate Yeast cell surface display has been extensively used to engineer
specificity of A-domains by yeast cell surface display that the binding specificity of antibodies (Chao et al., 2006; Miller
takes advantage of high throughput fluorescence-activated cell et al., 2008). The yeast vector pCTCON2 expresses the antibody
sorting (FACS) for iterative rounds of selection of millions of library as a fusion to the yeast agglutinin protein Aga2p that is
A-domain mutants.
attached through disulfide bonds to Aga1p protein as part of
Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved 93

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
the yeast cell wall (Figure 1B). The yeast cell library is then incu- cell surface was able to bind 8 (Figure S1B). Thus yeast cell
bated with a fluorescently labeled antigen to allow the binding of surface display can be used to select A-domains based on their
antigen molecules to the antibody displayed on the yeast affinity with acyl-AMS probes.
surface. FACS is then used to isolate yeast cells displaying anti-
body mutants with high affinities with the antigen. To test if yeast Construct an A-Domain Library of DhbE for Yeast
selection can be used to engineer the substrate specificity of the Selection
A-domains, we cloned DhbE into the pCTCON2 vector to display To assess the substrate specificity of DhbE, we measured
the DhbE enzyme on the yeast cell surface. The Aga2p-DhbE the steady-state kinetics of the complete adenylation-thioester-
domain fusion also has a hemagglutinin (HA) tag and a Myc tag fication reaction catalyzed by DhbE using saturating con-
at the N and C termini, respectively, of the A-domain to enable centrations of the ArCP domain of DhbB in a coupled assay
the detection of A-domain displayed on the cell surface (Fig- that measures release of pyrophosphate (PPi). During this assay,
ure 1B). After inducing the yeast cell to express the Aga2p- PPi generated in the DhbE-catalyzed condensation reaction
DhbE fusion, we incubated the cells with a mouse anti-HA anti- between the aryl substrate and ATP is converted to Pi by inor-
body and a chicken anti-Myc antibody so that the antibodies ganic pyrophosphatase (Ehmann et al., 2000; Webb, 1992).
would bind to the peptide tags flanking DhbE on the cell surface. Subsequently, purine nucleoside phosphorylase catalyzes the
Cells were then washed and incubated with a mixture of goat phosphorolysis of the chromogenic substrate 7-methylthiogua-
antimouse antibody conjugated with Alexa Fluor 647 and goat nosine that can be monitored at 360 nm. After catalyzing the first
antichicken antibody conjugated with Alexa Fluor 488 to label round of aryl-AMP formation, DhbE remains inactive until the
DhbE displayed on the cell surface with fluorophores. Flow cy- aryl-AMP intermediate dissociates from the enzyme or is broken
tometry analysis of the yeast cells showed that more than 30% down by the transfer to DhbB. Consequently, the rate of PPi
of the cells were doubly labeled with Alexa Fluor 647 and 488 flu- release catalyzed by DhbE is a measure of the combined rate
orophores, indicating efficient display of DhbE on the yeast cell of aryl-AMP dissociation from the enzyme and aryl transfer to
surface (Figure S1A available online).
ArCP. A previous study on EntE, a DhbE homolog in Escherichia
Next, we needed to develop a method to fluorescently label coli, which also activates DHB and loads it onto an ArCP,
the yeast cells displaying A-domain mutants with desired showed that the rate of DHB transfer to ArCP is more than
substrate specificity in order to select these cells from the 100-fold faster than the dissociation of DHB-AMP from the
A-domain library by FACS. Given the relatively low affinity of A- enzyme, due to the tight-binding nature of the intermediate
domains for their substrate acids (1 mM–1 mM) and inability to acyl-adenylate (Ehmann et al., 2000). We can therefore approx-
attach a biotin moiety conveniently without drastically affecting imate the rate of PPi release in the assay to that of substrate
substrate binding affinity, we elected to design a chemical probe transfer to the ArCP domain catalyzed by DhbE. Using the PPi
to report substrate recognition of the A-domains on the yeast cell release assay, we determined the specificity constants (kcat/Km)
surface that exploits the following: (1) the high affinity of for 2,3-dihydroxybenzoic acid (DHB, 1), 2-hydroxybenzoic
A-domains for their intermediate acyl-adenylate (acyl-AMP) as (SA, 4), 3-hydroxybenzoic acid (3-HBA, 6), and 2-aminobenzoic
a result of the bisubstrate nature of this intermediate that inter- acid (2-ABA, 7) as 1,100, 140, 22, and 5.5 minꢂ1mMꢂ1, respec-
acts with both the acid and ATP substrate binding pockets, (2) tively (Table 1). 3-HBA and 2-ABA exhibit 2% and 0.5% activity,
the ability to mimic the acyl-AMP and therefore generate a respectively, relative to the native substrate DHB. Based on
chemically stable probe by isosteric replacement of the phos- these results, we decided to use yeast cell surface display to
phate moiety for a sulfamate (acyl-AMS probe, wherein AMS engineer DhbE so that it can preferentially activate 3-HBA and
denotes adenosine monosulfamate, an isostere of AMP) (Ferre- 2-ABA over DHB and accordingly synthesized the correspond-
ras et al., 2005; Finking et al., 2003; Miethke et al., 2006; Somu ing acyl-AMS probes 9 and 10 (Figure 1A; see Supplemental
et al., 2006), (3) the potential to modify acyl-AMS probes at their Information for synthesis). For comparative purposes, we also
C-2 position for incorporation of a biotin moiety without compro- measured the steady-state kinetic parameters of DhbE using
mising binding affinity (Neres et al., 2008), and (4) the high the conventional ATP-PPi exchange assay that only measures
discrimination of acyl-AMS probes for their cognate A-domain the adenylation partial reaction. We demonstrated DhbE adeny-
(Qiao et al., 2007). Based on these design principles we prepared lates 3-HBA and 2-ABA, but at much lower rates than its native
salicyl-AMS probe 8 with biotin attached to the C-2 atom of the substrate DHB, and the relative trend in activity was identical
purine base through a long and flexible linker (Figure 1A; for to that observed with the PPi release assay (Figure S2A).
synthesis, see the Supplemental Information, Figures S5–S14,
The crystal structure of the DhbE A-domain in complex with
and Table S3) to report the binding to DhbE on the surface of DHB has been solved (Figure 2A), and it reveals a set of 10
yeast cells. Probe 8 contains salicylic acid (SA) rather than active-site residues of DhbE that serve as the specificity confer-
DHB due to the oxidative instability of the catechol in DHB. We ring nonribosomal code for DHB recognition: Asn235, Tyr236,
incubated 8 and chicken anti-Myc antibody with yeast cells to Ser240, Ala277, Gln304, Gly306, Val329, Val337, Tyr 339, and
allow their binding with wtDhbE on the cell surface. After washing Lys519 (Figure 2B) (May et al., 2002). Among them, the amide
the cell, a mixture of streptavidin-conjugated phycoerythrin (PE) moiety of the Asn235 side chain is engaged in hydrogen-bonding
and antichicken antibody conjugated with Alexa Fluor 488 were interactions with the 2-OH and 3-OH groups of DHB. The side
˚
added to the cells to detect the binding of 8 and the anti-Myc chain of Val337 is also in close distance (5.1 A) with the 2-OH
antibody to wtDhbE. Flow cytometry showed that more than group of DHB. Outside the residues involved in the nonribosomal
17% of the cell population was double positive with labeling of code, the imidazole nitrogen of His234 is within hydrogen-
both fluorophores, indicating that wtDhbE expressed on the bonding distance with both the carboxylate and 2-OH groups
94 Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
Table 1. PPi Release Rate of the Aryl Acid Adenylation Reaction Catalyzed by wtDhbE and Mutants
Ratio of kcat/Km with the Same
Enzyme and Substrate
Km (mM)
kcat (minꢂ1
)
kcat/Km (minꢂ1mMꢂ1
)
Substrate (Mutant/wtDhbE)
wtDhbE
DHB
4.3 0.4
14 2.2
25 7.7
62 17
4.61 0.09
1.90 0.07
0.56 0.02
0.34 0.03
1,100
140
22
SA
3-HBA
2-ABA
5.5
KZ4(Trp234His)
DHB
3.5 0.8
46 4.8
4.8 0.7
1.45 0.08
0.84 0.01
1.14 0.02
410
18
0.37
0.13
11
SA
3-HBA
240
KZ12(Trp234His)
DHB
38 12
56 14
3.5 0.3
1.25 0.14
0.37 0.03
0.12 0.01
33
6.6
34
0.03
0.047
6.2
SA
2-ABA
See also Table S1.
of DHB, and the methyl side chain of Ala333 is in close proximity truncations in the library (Figure 3B). In subsequent rounds of
˚
(3.5 A) to the 2-OH of DHB (Figure 2B). Since we intended to selection and FACS sorting, library cells were bound to probe
engineer DhbE to be specific for 3-HBA and 2-ABA, in which 9 and chicken anti-Myc antibody as the primary reagents, and
the 2-OH substitution of DHB is replaced with a hydrogen streptavidin-PE and antichicken antibody Alexa Fluor 488 as
atom or an amino group, we focused on improving the interac- secondary reagents. Sort gate covered the range of cells that
tions between DhbE and the C-2 functional groups in the aryl were the top 0.1%–1.0% in brightness for binding to 9 and the
acid substrates. We thus randomized His234, Asn235, Ala333, anti-Myc antibody (Figures 3C–3E). We also increased the selec-
and Val337 in DhbE to construct an A-domain library with tion stringency by decreasing the concentration of 9 from 1 mM in
a size of 5 3 106, large enough to cover all the mutants in a library round 3 to 0.1 mM in round 5. We observed a steady increase in
with four randomized residues (1.6 3 105) (Table S2).
the yeast cell population appearing at the diagonal of the flow
cytometry plot corresponding to the enrichment of doubly
labeled cells that displayed DhbE mutants with high affinity to
probe 9 (Figure 3).
Yeast Cell Surface Display to Identify DhbE Mutants that
Are Specific for 3-HBA Activation
We displayed the library on the surface of yeast cells and carried
After five rounds of selection with 9, 30 clones were
out iterative rounds of selection with acyl-AMS probe 9 contain- sequenced. Alignment with wtDhbE showed that clone KZ2 is
ing 3-HBA as the acyl moiety (Figure 1B). The yeast selection the dominant clone after yeast selection with 14 appearances
followed the protocol developed by Wittrup and Miller with among the 30 sequenced clones (Table 2). All selected DhbE
some modifications (Chao et al., 2006; Miller et al., 2008). In
the first round of selection we incubated the yeast cell library
with 3 mM 9. After washing the cells to remove unbound probes,
yeast cells labeled with biotin were collected by magnetic beads
coated with streptavidin. The use of magnetic-activated cell
sorting (MACS) allowed selection of a large quantity of yeast cells
(ꢁ1 3 1010) by binding to the biotinylated probe. The first round
of selection was to identify yeast cells based on the affinity of
probe 9 with the DhbE mutants displayed on the cell surface.
In the second round of selection, we bound the cells to anti-HA
and anti-Myc antibodies to enrich cells displaying full-length
DhbE in the library. Our DhbE was flanked by an N-terminal HA
tag and a C-terminal Myc tag when it was displayed as a fusion
with the Aga2 protein on the surface of yeast cells. Full-length
Figure 2. Substrate Binding Site of DhbE
DhbE should have both tags intact and be bound to both anti-
HA and anti-Myc antibodies (Figure 1B). The two types of anti-
bodies on yeast cell surface were detected by an antichicken
antibody conjugated to Alexa Fluor 488 and an antimouse anti-
body conjugated to Alexa Fluor 647. We used FACS to collect
the top 20% of cells that were doubly labeled by the two fluoro-
phores to eliminate cells displaying DhbE with N- or C-terminal
(A) Crystal structure of DhbE with the binding pocket of DHB shown in solid
ribbons and the rest of the protein in line ribbons (Protein Data Bank ID 1MD8)
(May et al., 2002).
(B) Detailed structure of DHB binding site showing key active-site residues as
the nonribosomal code for DHB recognition. Names of the residues random-
ized in the DhbE library are in red.
See also Figure S4.
Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved 95

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
Figure 3. Sorting of the DhbE Library Displayed on the Yeast Cell Surface by Binding to Biotin-Linked 3-HBA-AMS 9
(A) Sorting with magnetic beads coated with streptavidin. Binding of cells to 9 and an anti-Myc antibody was analyzed by flow cytometry.
(B) FACS sorting by antibody binding to the HA and Myc tags flanking DhbE on the cell surface.
(C–E), FACS by binding to 9 and an anti-Myc antibody. Percentages of doubly labeled cells were shown in the flow cytometry plots. Red frames in the plots
represent the sorting gates used for the collection of yeast cells.
clones have His234 mutated to a Trp residue, while Val337 is selection can recognize 3-HBA for the adenylation reaction
invariable and unchanged from the wild-type sequence. There but that the mutants are unable to transfer the activated aryl
is also a tendency for Ala333 to be unchanged, while it is re- substrate to the ArCP.
placed by Ser and Thr residues in some of the selected clones.
We speculated that the His234Trp mutation, persistent in
Asn235 is more flexible, with Ala in KZ2 as the most preferred mutants KZ1–4, might have a detrimental effect on substrate
mutation. Other residues such as Ser, Thr, Cys, Val and Gln loading to the ArCP. The crystal structure of DhbE in complex
were also selected at position 235 to bind to 9. We expressed with DHB and AMP shows the imidazole side chain of His234
the DhbE mutants (KZ1–4) that appeared multiple times among directly interacting with the DHB carboxylate and the 2-OH
the sequenced clones and assayed their catalytic activities. group (Figure 2B) (May et al., 2002). Furthermore, the His234
The ATP-PPi exchange assay showed that DhbE mutants side chain shields the DHB substrate from a tunnel through
KZ1–3 have similar activities as wtDhbE for activation of which the Ppant group of PCP may approach the DHB-AMP
3-HBA to form 3-HBA-AMP while KZ4 catalyzes 3-HBA activa- intermediate for thioester formation. While the His234Trp muta-
tion at a rate about 10-fold higher than wtDhbE and mutants tion was selected because it may improve the binding between
KZ1–3 (Figure S2B). Also, KZ4 has a Km of 4 mM with 3-HBA, the DhbE mutants and the aryl acid substrate, the Trp side chain
only slightly larger than the Km of wtDhbE with the native may block access of the Ppant thiol to the DHB adenylate due to
substrate DHB (2.5 mM) (Table S1). We then tested the transfer its considerably larger size relative to His. To recover the ArCP
of 3-HBA to the ArCP domain of DhbB catalyzed by the DhbE loading activity of the DhbE mutants, we mutated Trp234 in
mutants. To our surprise, while wtDhbE gave about 10% loading KZ1–4 back to His and generated mutants KZ1–4(Trp234His).
of the ArCP domain with 3-HBA after 30 min (Figure 4A), none of The ATP-PPi exchange assay showed that all four mutants
the mutants could catalyze the loading of ArCP with 3-HBA (Fig- catalyze 3-HBA adenylate formation, with KZ4(Trp234His) still
ure 4B). These results confirm that the DhbE mutants from yeast possessing the strongest activity among the mutant clones (Fig-
ure S2C). A MALDI assay showed that within 30 min, the
KZ4(Trp234His) mutant can load 60% of the ArCP with 3-HBA,
Table 2. Alignment of DhbE Mutants Selected by Yeast Cell
while other mutants and wtDhbE can load less than 10% of the
Surface Display with AMS Conjugated Probes 9 and 10
ArCP (Figure 4C). These results show that the Trp234His reverse
Residue Number
mutation enabled substrate transfer from mutant DhbE to the
234
H
235
N
333
A
337
V
Number of
ArCP domain of DhbB, and the KZ4 mutant with Trp234His
mutation is significantly more active in loading 3-HBA onto
ArCP than wtDhbE.
wtDhbE
Times Selected
Clones selected by 9
We then used the PPi release assay to characterize the
kinetics of KZ4 (Trp234His) for catalyzing 3-HBA activation and
transfer to the ArCP (Figure 5; Table 1). We found that KZ4
(Trp234His) has a Km of 4.8 mM and a kcat of 1.14 minꢂ1 with
3-HBA. In comparison, wtDhbE has a Km of 25 mM with 3-HBA,
KZ1
KZ2
KZ3
KZ4
KZ5
KZ6
KZ7
KZ8
6
14
4
W
W
W
W
W
W
W
W
S
A
N
Q
Q
T
A
A
S
T
V
V
V
V
V
V
V
V
2
2
S
A
A
A
5.2-fold higher than KZ4 (Trp234His); and a kcat of 0.56 minꢂ1
,
1
2-fold less than KZ4 (Trp234His) with 3-HBA. This suggests
that DhbE mutants enriched by yeast selection with the
3-HBA-AMS probe 9 acquired higher binding affinity with the
nonnative substrate 3-HBA. As a result, the specificity constant
1
C
V
1
Clones selected by 10
(kcat/Km) of KZ4 (Trp234His) with 3-HBA is 240 minꢂ1mMꢂ1
,
KZ11
KZ12
10
9
W
W
D
D
T
T
R
K
which is 11-fold higher than wtDhbE with 3-HBA. The kcat/Km
of KZ4 (Trp234His) with the native substrate DHB is
410 minꢂ1mMꢂ1, which is nearly 3-fold lower than wtDhbE with
See also Table S2.
96 Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
Figure 4. MALDI Analysis of Transfer of 3-HBA Substrate to the ArCP Domain Catalyzed by wtDhbE and the KZ4 Mutants
(A–C) Reactions were allowed to proceed for 30 min at room temperature. (A) wtDhbE has some activity of loading 3-HBA to the ArCP domain of DhbB. (B) KZ4
mutant is defective in 3-HBA loading to ArCP. (C) The KZ4 (Trp234His) mutant can efficiently load 3-HBA to ArCP.
DHB (Table 1). Thus, KZ4 (Trp234His) makes a nearly 30-fold The yeast selection enriched A-domain mutants with high bind-
switch in substrate specificity.
ing affinities with the acyl-AMS probes. In this way, the selected
A-domain mutants may have better recognition of the nonnative
substrates and enable their loading onto the ArCP domains in the
NRPS assembly line. We demonstrated that yeast cell surface
Yeast Cell Surface Display to Identify DhbE Mutants
Specific for 2-ABA Activation
We also carried out selection with the same DhbE library dis- display is a very efficient platform to identify A-domain mutants
played on yeast cell surface and the acyl-AMS probe 10 contain- with altered binding specificities using nonnative substrates
ing 2-ABA as the acyl moiety (Figure 1). After five rounds of selec- conjugated to AMS. In just five rounds of selection of a yeast
tion following the same protocol as employed with probe 9, we library of 5 3 106 clones, we were able to identify distinct sets
found that the yeast library converged to two very similar DhbE of DhbE mutants that can bind to probes 9 and 10, containing
mutants: KZ11 and KZ12 (Figure S3; Table 2). Both mutants 3-HBA and 2-ABA as the acyl moieties, respectively.
have His234 replaced by Trp, the same as the mutants from
We also found that the DhbE mutants from yeast selection
the selection with probe 9. The two mutants also have identical have Km values 5- to 18-fold lower than wtDhbE for binding to
Asn235Asp and Ala333Thr mutations. The only difference the nonnative substrates 3-HBA and 2-ABA (Table 1). By
between these two mutants is that Val337 is mutated to Arg in contrast, the kcat values of the DhbE mutants change very little
KZ11 and to Lys in KZ12. Initially selected KZ11 and KZ12 are with KZ4 catalyzing 3-HBA transfer with a kcat 2-fold higher
not active in transferring 2-ABA substrate to the ArCP domain than wtDhbE, and KZ12 catalyzing 2-ABA transfer with a kcat
of DhbB, similar to mutants KZ1–4. However, when we mutated 2-fold lower than wtDhbE. These results suggest that the
Trp234 back to His, KZ11(Trp234His) and KZ12(Trp234His) improvement in catalytic efficiency (kcat/Km) of the A-domain
became catalytically active for loading 2-ABA to the ArCP.
mutants with the nonnative substrates is mainly due to the
The PPi release assay showed that KZ12 (Trp234His) has a Km decrease in Km, an indication of better recognition of the nonna-
of 3.5 mM with the 2-ABA substrate, 18-fold lower than that of tive substrates by the enzyme active site. These results agree
wtDhbE with 2-ABA (62 mM) (Table 1). Due to the lower Km, the with previous work on engineering catalytic antibodies that
k
cat/Km of KZ12 (Trp234His) with 2-ABA (34 minꢂ1mMꢂ1) is 6.2- showed tighter binding with the transition state analog by the
fold higher than that of wtDhbE with 2-ABA. The kcat/Km of KZ12 antibody does not lead to a higher turnover rate (kcat) of the cata-
(Trp234His) with the native substrate DHB is 33 minꢂ1mMꢂ1
which is nearly 33-fold lower than wtDhbE with DHB (Table 1).
,
lytic reactions (Baca et al., 1997).
Our work on DhbE engineering also shed light on the mecha-
As a result, KZ12 (Trp234His) makes a 206-fold switch in sub- nism of substrate transfer catalyzed by the A-domain. Yeast
strate specificity. Kinetic analysis of KZ12 (Trp234His) showed selection of the A-domain library identified a His234Trp mutation
that the mutant has a kcat of 0.12 minꢂ1, similar to that of present in all the mutants (Table 2). The crystal structure of DhbE
wtDhbE-catalyzed 2-ABA activation (0.34 minꢂ1) (Table 1); how- in complex with DHB and AMP suggested that the Trp residue at
ever this kcat value is considerably lower than the kcat of wtDhbE this position may provide better stacking interactions between
with its native substrate, DHB (4.6 minꢂ1). This suggests that, the indole ring of the Trp side chain and the carboxylate and
while yeast selection based on affinity binding with bisubstrate sulfonamide groups of the aryl-AMS conjugate (Figure 2B).
analog probes may improve binding interactions between the However, the His234Trp mutation may also block the access
engineered enzyme and nonnative substrates to give a lower of the Ppant group from the ArCP to the aryl-AMP formed at
Km, such affinity-based selection may not be able to improve the enzyme active site. This notion is supported by the recently
the kcat of the engineered enzyme with nonnative substrates.
reported crystal structure of EntE, a homologous enzyme of
DhbE from E. coli., covalently tethered to the ArCP domain of
EntB, equivalent to the ArCP domain of DhbB, through a mecha-
nism-based inhibitor (Mitchell et al., 2012; Sundlov et al., 2012).
DISCUSSION
In this study, we performed yeast selection on an A-domain In the EntE-EntB structure, the Ppant arm of the EntB ArCP is in-
library to change the substrate specificities of the A-domains. serted into the substrate binding site of the EntE A-domain in
Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved 97

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
Figure 5. Kinetic Analysis of DHB and
3-HBA Transfer to ArCP by wtDhbE and KZ4
(Trp234His) Based on the PPi Release Assay
Michaelis-Menten plot to compare the initial
rates of PPi release catalyzed by wtDhbE and
KZ4(Trp234His) at varying concentrations of native
substrate DHB (A) and nonnative substrate 3-HBA
(B). Data are means
determinations.
ranges for at least three
Once the A-domain is engineered to
incorporate nonnative aryl or amino acid
substrates into the NRPS assembly line,
order to form a covalent bond with the inhibitor molecule bound nonnative substrates could be used for the synthesis of ‘‘unnat-
to the A-domain (Figure S4) (Sundlov et al., 2012). The side chain ural natural products.’’ A correlation could be drawn between
of His234 from EntE has to adopt a different conformation from engineering NRPS A-domains and the engineering of tRNA
the corresponding His234 in the DhbE structure to allow the synthetases, which are also adenylation domains for amino
Ppant group to extend into the EntE active site. A His234Trp acid activation (Wang et al., 2006). When tRNA synthetases are
mutation places a large indole ring in the path of Ppant that engineered to load unnatural amino acids on tRNA, they open
may block the entry of the Ppant group. Indeed, none of the the gate for the incorporation of unnatural building blocks into
mutants we assayed with the His234Trp mutation could load proteins. NRPS enzymes seem to be more discerning of nonna-
the aryl acid substrate onto the ArCP. We were able to recover tive substrates than the ribosome for protein biosynthesis. While
the transfer activity of the DhbE mutant by introducing a reverse early work showed that downstream modules may block the
mutation to put His234 residue back in the DhbE mutants (Fig- elongation of peptide intermediates with the incorporation of
ure 4C; Table 1). This suggests that yeast selection is best nonnative substrates (Pavela-Vrancic et al., 1999; Uguru et al.,
utilized to improve substrate recognition at the A-domain resi- 2004), more recent studies have successfully demonstrated
dues devoted to substrate binding, but is less useful in opti- the ability of reprogrammed A-domains to incorporate nonnative
mizing A-domain kinetics. More structural and mechanistic substrates into the final NRPs (Evans et al., 2011; Fischbach
insights on A-domain catalysis would certainly guide the future et al., 2007; Thirlway et al., 2012).
library design and selection.
The structural complexity of NRP natural products in many
Yeast selection of an A-domain library also provided new cases is restrictive to the production of analogs with improved
insights on DhbE domain recognition of aryl acid substrates medicinal activities by traditional synthetic chemistry. Histori-
(Ames and Walsh, 2010; Stachelhaus et al., 1999). Asn235 and cally, natural product analogs were generated by derivatization
Val337 are the nonribosomal code residues that were random- of the parent natural product. Combinatorial biosynthesis offers
ized in the DhbE library. When the yeast library was screened a complimentary method to generate analogs through genetic
against probe 9 for 3-HBA activation, KZ4 with Asn235 replaced engineering of the biosynthetic genes. Addition, deletion, or
by a Gln residue showed higher efficiency in adenylate formation replacement of genes in NRPS gene clusters has been used to
with 3-HBA (Table 2; Table S1). The Gln side chain is one CH2 reprogram NRPS enzymatic assembly lines to produce new
group longer than the Asn side chain in wtDhbE. It may extend analogs of the antibiotics surfactin, daptomycin, and echino-
further into the substrate binding pocket to fill the space that mycin (Nguyen et al., 2006; Stachelhaus et al., 1995; Watanabe
used to be occupied by the 2-OH substitute of DHB. The amide et al., 2009). Similar approaches have also been used to recon-
NH2 of the Gln residue may also form hydrogen bonds with the figure closely related gene clusters of polyketide synthases to
3-OH group in 3-HBA (Figure 2B). Yeast selection with probe 9 generate analogs of the antibiotic erythromycin (McDaniel
also showed that Val337 is indispensable for binding to 3-HBA. et al., 1999; Pfeifer et al., 2001). We expect our method for engi-
When the yeast library was selected for binding to probe 10, neering NRPS A-domains should greatly expand the scope of
Val337 is replaced by Lys or Arg with longer and positively the chemical building blocks for the combinatorial biosynthesis
charged side chains (Table 2). The Lys and Arg residues at this of NRPs. Engineered nonribosomal peptide synthesis can be
position may interact with the 2-NH2 group of 2-ABA. Ala333 used in parallel with engineered ribosomal peptide synthesis
was not assigned as a nonribosomal code residue for A-domain (Yamagishi et al., 2011) to generate peptide libraries of diverse
binding with the aryl acid substrates (Ames and Walsh, 2010). structures.
After yeast selection with AMS conjugates to 3-HBA or 2-ABA,
Ala333 was replaced with Ser or Thr residues with hydroxyl SIGNIFICANCE
side chains that may form hydrogen bonds with the 3-HBA
and 2-ABA substrates. Overall, our results demonstrate that We describe a powerful method based on yeast cell sorting
A-domains can be reprogrammed by protein engineering to to engineer A-domain substrate specificity employing acyl-
recognize nonnative substrates. It would be interesting to see if AMS probes for affinity selection that enables one to screen
the A-domain mutants we acquired in this study incorporate millions of mutants, which is three orders of magnitude
3-HBA or 2-ABA substrates into the bacillibactin scaffold.
more than previously described. Using this method, we
98 Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
with antibodies for 45 min at 4ꢀC. The cells were then washed twice with 0.1%
BSA in TBS and stained with 5 mg/ml goat antimouse antibody conjugated with
Alexa Fluor 647 (Invitrogen A-21235) and goat antichicken antibody conju-
gated with Alexa Fluor 488 (Invitrogen, A-11039) in 0.1 ml 0.1% BSA in TBS.
The cell suspension was shielded from light and incubated at 4ꢀC for
30 min. After washing twice with 0.1% BSA in TBS, cells were analyzed on
a flow cytometer (LSRII, BD Biosciences) to count the number of cells that
were labeled with both fluorophores. Cells were also analyzed from control
labeling reactions in which primary antibodies were either excluded from the
reaction or cells were only labeled with primary and secondary antibodies
for just one of the affinity tags.
successfully switched the substrate specificity of DhbE for
nonnative substrates. We envision that the yeast selection
platform can also be used to engineer A-domain specificity
with amino acids by selection with the corresponding
aminoacyl-AMS probes. Our results also provided insight
into the substrate transfer to the ArCP domain as the
His234Trp mutation was catalytically incompetent in the
thioesterification reaction, but fully active in the adenylation
partial reaction. Activity could be fully recovered by intro-
ducing the reverse mutation, and these results are con-
sistent with structural studies recently reported from Gulick
and co-workers (Sundlov et al., 2012).
Cells displaying wtDhbE were also assayed by binding to acyl-AMS probe 8.
During the labeling with primary reagents, 10 nM 8 and 10 mg/ml chicken anti-
c-myc antibody (Invitrogen, A-21281) were used. In the following step, 5 mg/ml
streptavidin conjugated with PE (Invitrogen, S-32350) and 5 mg/ml goat anti-
chicken antibody conjugated with Alexa Fluor 488 (Invitrogen, A-11039)
were used as secondary reagents for the detection of wtDhbE on the cell
surface and the binding of wtDhbE to 8.
EXPERIMENTAL PROCEDURES
See the Supplemental Experimental Procedures for full synthetic procedures,
cloning, and kinetic characterization of the A-domain mutants.
Selection of the DhbE Library Displayed on the Yeast Cell Surface
The first round of selection of the yeast library was carried out with MACS. For
subsequent rounds of selection, FACS was used to identify yeast clones dis-
playing DhbE mutants that were bound to acyl-AMS probes 9 or 10 with high
affinity. For MACS selection, approximately 8 3 107 yeast cells displaying the
DhbE library were incubated with 9 or 10 in a total volume of 600 ml 0.1% BSA
in TBS. After 45 min at 4ꢀC with intermittent inversion to ensure thorough mix-
ing, cells were pelleted by centrifugation. Cells were then resuspended in fresh
0.1% BSA in TBS and pelleted again. This procedure was repeated twice to
remove biotin-linked affinity probe that was not bound to the yeast cells. After
washing, cells were mixed with 90 ml of streptavidin-coated microbeads
provided by the mMACS Streptavidin Starting Kit (Miltenyi Biotec, 130-091-
287) in a total volume of 800 ml of 0.1% BSA in TBS. Cell suspension was
shielded from light and incubated at 4ꢀC for 30 min. After the labeling reaction
was finished, the cells and magnetic beads were added to 30 ml of 0.1% BSA
in TBS. The cell suspension was pelleted by centrifugation at 500 g for 10 min.
The supernatant was aspirated, and the cell pellet including the magnetic
beads was resuspended in 500 ml 0.1% BSA-TBS. Yeast cells bound to
magnetic beads by biotin-streptavidin interaction were then captured by
a magnet according to manufacturer’s instructions, and the beads were
washed with 0.1% BSA in TBS. Cells bound to the magnetic beads were
eluted into 5 ml SDCAA medium supplemented with 100 mg/ml ampicillin
and 50 mg/ml kanamycin and were cultured at 30ꢀC overnight. In parallel,
library cells were bound to primary and secondary antibodies to evaluate the
display of DhbE mutants on the yeast cell surface.
Yeast Display of the DhbE Library
DhbE library in pCTCON2 was chemically transformed into YB100 yeast cells
following published protocols with some modifications (Gietz and Schiestl,
1991; Gietz and Woods, 2002). Briefly, yeast cells were first cultured in
200 ml YPD (20 g dextrose, 20 g peptone, and 10 g yeast extract in 1 l deion-
ized H2O, sterilized by filtration) to an optical density 600 (OD600) around 0.5 at
30ꢀC. The cells were then pelleted at 3,500 rpm for 5 min. Cells were subse-
quently washed by 20 ml TE (100 mM Tris base, 10 mM EDTA, pH 8.0) and
20 ml LiOAc-TE (100 mM lithium acetate in TE), before resuspension in approx-
imately 800 ml LiOAc-TE. A typical transformation reaction contained a mixture
of 1 mg pCTCON2 plasmid DNA, 2 ml denatured single-stranded carrier
DNA from salmon testes (Sigma Aldrich), 25 ml resuspended yeast competent
cells, and 300 ml polyethylene glycol (PEG) solution (40% [w/v] PEG 3350 in
LiOAc-TE). In order to achieve a library size of 106, 30 transformations were
set up in parallel. A control was also prepared in which the pCTCON2 plasmid
was excluded. Both the transformation reactions and the control were incu-
bated at 30ꢀC for 1 hr and then at 42ꢀC for 20 min. Cells in each transformation
were pelleted by centrifuging at 13,000 rpm for 30 s and resuspended in 20 ml
SDCAA medium (2% [w/v] dextrose, 6.7 g Difco yeast nitrogen base without
amino acids, 5 g Bacto casamino acids, 50 mM sodium citrate, and 20 mM
citric acid monohydrate in 1 l deionized H2O, sterilized by filtration). Yeast cells
were resuspended, pooled together into 1 l SDCAA medium and allowed to
grow at 30ꢀC over a 2-day period to an OD600 above 5. For long-term storage
of the yeast library, 20 ml of the yeast culture was aliquoted in 15% glycerol
stock and stored at ꢂ80ꢀC.
The second-round selection of the yeast library was to enrich cells display-
ing full-length DhbE mutants. The library cells amplified from the first round
of selection were labeled with 10 mg/ml mouse anti-HA antibody (Santa
Cruz Biotechnology, sc-7392) and chicken anti-c-myc antibody (Invitrogen,
A-21281). After washing the cells, secondary antibodies of goat antimouse
antibody conjugated with Alexa Fluor 647 (Invitrogen, A-21235) and goat anti-
chicken antibody conjugated with Alexa 488 (Invitrogen, A-11039) were added
to the cells in 0.1% BSA in TBS. After incubation, cells were washed twice with
0.1% BSA in TBS. Cells with the top 15%–20% of brightness labeled by both
fluorophores were collected by FACS.
To titer transformation efficiency, 10 ml of the resuspended yeast transform-
ants from either the library mix or control was serially diluted in SDCAA medium
and plated on Trp- plates—20 g agar, 20 g dextrose, 5 g (NH4)2SO4, 1.7 g Difco
yeast nitrogen base without amino acids, 1.3 g drop-out mix excluding Trp in
1 l deionized H2O, autoclaved. Yeast cells transformed with pCTCON2 plas-
mids were expected to appear within 2 days of incubation at 30ꢀC.
Model Selection of Yeast Cells Displaying wtDhbE
Yeast cell EYB100 was transformed with wild-type DhbE (wtDhbE) in
pCTCON2 and streaked on a Trp- plate. Yeast colonies grew up after two
days of incubation at 30ꢀC. Cells were scraped from the Trp- plate to inoculate
a 5 ml SDCAA culture that was allowed to shake at 30ꢀC to reach an initial
OD600 of 0.5. Cells were centrifuged at 3,000 rpm for 5 min and induced for
DhbE expression by resuspension in 5 ml SGCAA (2% [w/v] galactose, 6.7 g
Difco yeast nitrogen base without amino acids, 5 g Bacto casamino acids,
38 mM Na2HPO4 and 62 mM NaH2PO4 , H2O in 1 l deionized H2O, sterilized
by filtration). The yeast culture was shaken at 20ꢀC for 16–24 hr.
For analysis of DhbE display on the surface of yeast cells, 106 cells were
resuspended in 0.1 ml Tris-buffered saline (TBS) (25 mM Tris, pH 7.5,
150 mM NaCl) with 0.1% bovine serum albumin (BSA). Mouse anti-HA anti-
body (Santa Cruz Biotechnology, sc-7392) and chicken anti-c-myc antibody
(Invitrogen, A-21281) were used as primary antibodies, and they were added
to the cell suspension at a concentration of 10 mg/ml. The cells were incubated
In subsequent rounds of yeast selection, cells were labeled with affinity
probes
9 or 10 and 4 mg/ml chicken anti-c-myc antibody (Invitrogen,
A-21281) as the primary reagents, and 5 mg/ml streptavidin-PE (Invitrogen,
S-32350) and 5 mg/ml goat antichicken antibody conjugated to Alexa Fluor
488 (Invitrogen, A-11039) as the secondary reagents. The labeling reaction
was incubated at 4ꢀC for 30 min; then cells were pelleted, washed twice,
and resuspended in TBS with 0.1% BSA for sorting. The concentration of
probe 9 or 10 was decreased from 1 mM in the third round of selection to
0.1 mM in the fifth round of selection. The sorting gate also became more strin-
gent with the top 0.1% of doubly labeled cells collected in the fifth round of
selection. After the fifth round of cell sorting, the collected cells were grown
in an SDCCA medium to an OD600 around 0.5. Zymoprep II Yeast Plasmid
Miniprep Kit (Zymo Research, D2004) was then used to extract pCTCON2
plasmid DNA from the yeast cells. The plasmids were then transformed into
Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved 99

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
SS320 E. coli competent cells. Individual colonies were miniprepped, and the
plasmid DNA was sequenced to reveal the selected mutations in the DhbE
clones. The DNA sequences of the mutant DhbE clones were aligned by the
ClustalW algorithm in the Lasergene MegAlign software (DNAStar).
peptide synthetases. Chembiochem: A European Journal of Chemical Biology
4, 903–906.
Fischbach, M.A., and Walsh, C.T. (2006). Assembly-line enzymology for poly-
ketide and nonribosomal Peptide antibiotics: logic, machinery, and mecha-
nisms. Chem. Rev. 106, 3468–3496.
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.
SUPPLEMENTAL INFORMATION
Supplemental Information includes fourteen figures, three tables, and Supple-
mental Experimental Procedures and can be found with this article online at
http://dx.doi.org/10.1016/j.chembiol.2012.10.020.
Gietz, R.D., and Schiestl, R.H. (1991). Applications of high efficiency lithium
acetate transformation of intact yeast cells using single-stranded nucleic acids
as carrier. Yeast 7, 253–263.
ACKNOWLEDGMENTS
Gietz, R.D., and Woods, R.A. (2002). Transformation of yeast by lithium
acetate/single-stranded carrier DNA/polyethylene glycol method. Methods
Enzymol. 350, 87–96.
This work was supported by a lab startup grant from the University of Chicago
(to J.Y.) and a grant from the National Institutes of Health (AI070219 to C.C.A.).
We thank Professor K. Dane Wittrup of the Massachusetts Institute of Tech-
nology for providing the pCTCON2 plasmid and yeast strain EBY100. We
also thank S. Annie Gai and Tiffany F. Chen of the Wittrup Group and Satoe
Takahashi, Stephen Kron, Akiko Koide, and Shohei Koide of the University
of Chicago for helpful discussions.
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.
May, J.J., Wendrich, T.M., and Marahiel, M.A. (2001). The dhb operon of
Bacillus subtilis encodes the biosynthetic template for the catecholic sidero-
phore 2,3-dihydroxybenzoate-glycine-threonine trimeric ester bacillibactin.
J. Biol. Chem. 276, 7209–7217.
Received: August 31, 2012
Revised: October 19, 2012
Accepted: October 25, 2012
Published: January 24, 2013
May, J.J., Kessler, N., Marahiel, M.A., and Stubbs, M.T. (2002). Crystal struc-
ture of DhbE, an archetype for aryl acid activating domains of modular nonri-
bosomal peptide synthetases. Proc. Natl. Acad. Sci. USA 99, 12120–12125.
McDaniel, R., Thamchaipenet, A., Gustafsson, C., Fu, H., Betlach, M., and
Ashley, G. (1999). Multiple genetic modifications of the erythromycin polyke-
tide synthase to produce a library of novel ‘‘unnatural’’ natural products.
Proc. Natl. Acad. Sci. USA 96, 1846–1851.
REFERENCES
Ames, B.D., and Walsh, C.T. (2010). Anthranilate-activating modules from
Miethke, M., Bisseret, P., Beckering, C.L., Vignard, D., Eustache, J., and
Marahiel, M.A. (2006). Inhibition of aryl acid adenylation domains involved in
bacterial siderophore synthesis. FEBS J. 273, 409–419.
fungal nonribosomal peptide assembly lines. Biochemistry 49, 3351–3365.
Baca, M., Scanlan, T.S., Stephenson, R.C., and Wells, J.A. (1997). Phage
display of a catalytic antibody to optimize affinity for transition-state analog
binding. Proc. Natl. Acad. Sci. USA 94, 10063–10068.
Miller, K.D., Pefaur, N.B., and Baird, C.L. (2008). Construction and screening of
antigen targeted immune yeast surface display antibody libraries. Curr.
Protoc. Cytom. 4, Unit 4.7.
Baltz, R.H. (2008). Biosynthesis and genetic engineering of lipopeptide antibi-
otics related to daptomycin. Curr. Top. Med. Chem. 8, 618–638.
Mitchell, C.A., Shi, C., Aldrich, C.C., and Gulick, A.M. (2012). Structure of
PA1221, a nonribosomal peptide synthetase containing adenylation and
peptidyl carrier protein domains. Biochemistry 51, 3252–3263.
Cane, D.E., Walsh, C.T., and Khosla, C. (1998). Harnessing the biosynthetic
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.
Neres, J., Labello, N.P., Somu, R.V., Boshoff, H.I., Wilson, D.J., Vannada, J.,
Chen, L., Barry, C.E., 3rd, Bennett, E.M., and Aldrich, C.C. (2008). Inhibition
of siderophore biosynthesis in Mycobacterium tuberculosis with nucleoside
bisubstrate analogues: structure-activity relationships of the nucleobase
domain of 50-O-[N-(salicyl)sulfamoyl]adenosine. J. Med. Chem. 51, 5349–
5370.
Chao, G., Lau, W.L., Hackel, B.J., Sazinsky, S.L., Lippow, S.M., and Wittrup,
K.D. (2006). Isolating and engineering human antibodies using yeast surface
display. Nat. Protoc. 1, 755–768.
Chen, C.Y., Georgiev, I., Anderson, A.C., and Donald, B.R. (2009).
Computational structure-based redesign of enzyme activity. Proc. Natl.
Acad. Sci. USA 106, 3764–3769.
Nguyen, K.T., Ritz, D., Gu, J.Q., Alexander, D., Chu, M., Miao, V., Brian, P., and
Baltz, R.H. (2006). Combinatorial biosynthesis of novel antibiotics related to
daptomycin. Proc. Natl. Acad. Sci. USA 103, 17462–17467.
Clardy, J., and Walsh, C. (2004). Lessons from natural molecules. Nature 432,
¨
Pavela-Vrancic, M., Dieckmann, R., Dohren, H.V., and Kleinkauf, H. (1999).
829–837.
Editing of non-cognate aminoacyl adenylates by peptide synthetases.
Biochem. J. 342, 715–719.
Ehmann, D.E., Shaw-Reid, C.A., Losey, H.C., and Walsh, C.T. (2000). The EntF
and EntE adenylation domains of Escherichia coli enterobactin synthetase:
sequestration and selectivity in acyl-AMP transfers to thiolation domain co-
substrates. Proc. Natl. Acad. Sci. USA 97, 2509–2514.
Pfeifer, B.A., Admiraal, S.J., Gramajo, H., Cane, D.E., and Khosla, C. (2001).
Biosynthesis of complex polyketides in a metabolically engineered strain of
E. coli. Science 291, 1790–1792.
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.
Qiao, C., Gupte, A., Boshoff, H.I., Wilson, D.J., Bennett, E.M., Somu, R.V.,
Barry, C.E., 3rd, and Aldrich, C.C. (2007). 50-O-[(N-acyl)sulfamoyl]adenosines
as antitubercular agents that inhibit MbtA: an adenylation enzyme required for
siderophore biosynthesis of the mycobactins. J. Med. Chem. 50, 6080–6094.
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.
Robbel, L., and Marahiel, M.A. (2010). Daptomycin, a bacterial lipopeptide
synthesized by a nonribosomal machinery. J. Biol. Chem. 285, 27501–27508.
Ferreras, J.A., Ryu, J.S., Di Lello, F., Tan, D.S., and Quadri, L.E. (2005). Small-
molecule inhibition of siderophore biosynthesis in Mycobacterium tubercu-
losis and Yersinia pestis. Nat. Chem. Biol. 1, 29–32.
Sieber, S.A., and Marahiel, M.A. (2005). Molecular mechanisms underlying
nonribosomal peptide synthesis: approaches to new antibiotics. Chem. Rev.
105, 715–738.
Finking, R., Neumuller, A., Solsbacher, J., Konz, D., Kretzschmar, G.,
Schweitzer, M., Krumm, T., and Marahiel, M.A. (2003). Aminoacyl adenylate
substrate analogues for the inhibition of adenylation domains of nonribosomal
Sikora, A.L., Wilson, D.J., Aldrich, C.C., and Blanchard, J.S. (2010). Kinetic and
inhibition studies of dihydroxybenzoate-AMP ligase from Escherichia coli.
Biochemistry 49, 3648–3657.
100 Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology
Reprogramming Adenylation Domain Specificity
Somu, R.V., Boshoff, H., Qiao, C., Bennett, E.M., Barry, C.E., 3rd, and Aldrich,
C.C. (2006). Rationally designed nucleoside antibiotics that inhibit siderophore
biosynthesis of Mycobacterium tuberculosis. J. Med. Chem. 49, 31–34.
upstream nonribosomal peptidyl thioester intermediates. J. Am. Chem. Soc.
126, 5032–5033.
¨
von Dohren, H., Dieckmann, R., and Pavela-Vrancic, M. (1999). The nonriboso-
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.
mal code. Chem. Biol. 6, R273–R279.
Wang, L., Xie, J., and Schultz, P.G. (2006). Expanding the genetic code. Annu.
Rev. Biophys. Biomol. Struct. 35, 225–249.
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.
Watanabe, K., Hotta, K., Nakaya, M., Praseuth, A.P., Wang, C.C., Inada, D.,
Takahashi, K., Fukushi, E., Oguri, H., and Oikawa, H. (2009). Escherichia coli
allows efficient modular incorporation of newly isolated quinomycin biosyn-
thetic enzyme into echinomycin biosynthetic pathway for rational design and
synthesis of potent antibiotic unnatural natural product. J. Am. Chem. Soc.
131, 9347–9353.
Sundlov, J.A., Shi, C., Wilson, D.J., Aldrich, C.C., and Gulick, A.M. (2012).
Structural and functional investigation of the intermolecular interaction
between NRPS adenylation and carrier protein domains. Chem. Biol. 19,
188–198.
Webb, M.R. (1992). A continuous spectrophotometric assay for inorganic
phosphate and for measuring phosphate release kinetics in biological
systems. Proc. Natl. Acad. Sci. USA 89, 4884–4887.
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.
Yamagishi, Y., Shoji, I., Miyagawa, S., Kawakami, T., Katoh, T., Goto, Y., and
Suga, H. (2011). Natural product-like macrocyclic N-methyl-peptide inhibitors
against a ubiquitin ligase uncovered from a ribosome-expressed de novo
library. Chem. Biol. 18, 1562–1570.
Uguru, G.C., Milne, C., Borg, M., Flett, F., Smith, C.P., and Micklefield, J.
(2004). Active-site modifications of adenylation domains lead to hydrolysis of
Chemistry & Biology 20, 92–101, January 24, 2013 ª2013 Elsevier Ltd All rights reserved 101