expanding the promiscuity of a natural-product...
TRANSCRIPT
Expanding the promiscuity of a natural-productglycosyltransferase by directed evolutionGavin J Williams, Changsheng Zhang & Jon S Thorson
Natural products, many of which are decorated with essentialsugar residues, continue to serve as a key platform for drugdevelopment1. Adding or changing sugars attached to suchnatural products can improve the parent compound’spharmacological properties, specificity at multiple levels2,and/or even the molecular mechanism of action3. Thoughsome natural-product glycosyltransferases (GTs) are sufficientlypromiscuous for use in altering these glycosylation patterns,the stringent specificity of others remains a limiting factor innatural-product diversification and highlights a need forgeneral GT engineering and evolution platforms. Herein wereport the use of a simple high-throughput screen based ona fluorescent surrogate acceptor substrate to expand thepromiscuity of a natural-product GT via directed evolution.Cumulatively, this study presents variant GTs for theglycorandomization of a range of therapeutically importantacceptors, including aminocoumarins, flavonoids andmacrolides, and a potential template for engineering othernatural-product GTs.
As an emerging method to differentially glycosylate natural products,glycorandomization uses the inherent or engineered substrate pro-miscuity of anomeric kinases (Fig. 1a, E1) and nucleotidyltransferases(E2) for the in vitro synthesis of sugar nucleotide libraries as sugardonors for natural-product GTs4. Although the successful glyco-randomization of various natural-product scaffolds (including glyco-peptides5, avermectins6 and enediynes7) has been reported, otherrecent antibiotic glycorandomization attempts have revealed thataminocoumarin and macrolide GTs (NovM and EryBV, respectively)accept only 2 alternative sugar nucleotides out of 25 to 40 potentialdonors tested8,9. Thus, though permissive GTs open new opportu-nities for drug discovery, the stringent specificity of other GTs remainsa limiting factor in natural-product diversification and highlights aneed for general GT engineering and/or evolution platforms. Despitethe wealth of GT structural and biochemical information10, attemptsto alter GT donor/acceptor specificities via rational engineering havebeen largely unsuccessful and primarily limited to sequence-guidedsingle-site mutagenesis11. Owing to a lack of high-throughput GTscreens and selections, successful reports to alter GT donor/acceptorspecificities via directed evolution are equally sparse. Although an
in vivo selection for the directed evolution of the sialyltransferase CstII(a unique member of the GT-A superfamily) was recently disclosed12,the directed evolution of any member of the structurally and func-tionally distinct GT-B superfamily has not been achieved.
One member of the GT-B superfamily13, the oleandomycin GT(OleD) from Streptomyces antibioticus, catalyzes the glucosylation ofoleandomycin (1) using UDP-Glc (2) as donor to produce glucoside 3(Fig. 1b)14. A recent mass spectrometry analysis of OleD specificity ledto the identification of a range of small aromatic phenolics as putativeOleD acceptors15. Notably, this panel included the fluorescentcoumarin 4-methylumbelliferone (4) (Fig. 1b). Given the results ofcoumarin-based glycosynthase assays16, we postulated that 4 wouldoffer the ability to directly assess OleD-catalyzed glycosyltransfer viafluorescence. Specifically, masking the C7 OH of 4 quenches fluores-cence17. Preliminary rate determinations indicated the conversion rateof 4 to glucoside 5 (with 2 as donor using wild-type OleD) to be300-fold less than the rate with the natural acceptor 1 (for example,Table 1 and Supplementary Fig. 1 online). Using the fluorescence-based assay, activity in Escherichia coli pET28/OleD crude cell extractswas reproducibly detected to be only Btwo-fold higher than activityin cell extracts prepared from cultures that do not overexpress OleD(Fig. 2a). These preliminary studies confirmed 4 as a weak substratefor OleD. More importantly, these studies validated the fluorescence-based GT assay and set the stage to alter OleD catalytic efficiencyand/or substrate promiscuity by directed evolution. A mutant OleDlibrary was constructed by error-prone PCR using wild-type OleD astemplate, such that each variant had (on average) one or two aminoacid mutations per gene product. A relatively small library (B1,000colonies) of variants was initially screened using the fluorescence-based GT assay. For screening, extract aliquots were incubated with 4and 2 and allowed to react for 3 h, after which the change influorescence intensity was measured.
Several potential positive hits were identified in the first round ofscreening (Fig. 2a), and three of these—designated 2C3, 7B9 and8B3—were selected for further analysis. DNA sequencing of these hitsrevealed that 2C3 has a single amino acid mutation, A242V, whereas7B9 and 8B3 each have two amino acid mutations, S132F/G340W andP67T/I112T, respectively. In order to assign functional mutationswithin 7B9 and 8B3, the corresponding four single mutants wereconstructed and characterized. A comparison of the specific activity of
Received 8 June; accepted 25 July; published online 9 September 2007; doi:10.1038/nchembio.2007.28
Laboratory for Biosynthetic Chemistry, Pharmaceutical Sciences Division, School of Pharmacy, National Cooperative Drug Discovery Program, University of Wisconsin-Madison, 777 Highland Avenue, Madison, Wisconsin 53705, USA. Correspondence should be addressed to J.S.T. ([email protected]).
NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 10 OCTOBER 2007 6 5 7
L E T T ERS
the wild-type and mutant OleDs using the substrate pair 4 and 2confirmed the first-generation variants 2C3, 7B9 and 8B3 to be moreactive than wild-type OleD (Fig. 2b) and revealed that the G340W(from 7B9) and I112T (from 8B3) mutations are each nonfunctional.
The remaining three functional mutations were subsequently com-bined by site-directed mutagenesis to provide the mutant P67T/S132F/A242V, the specific activity of which was determined to beB30-fold higher than that of wild-type OleD with 4 and 2 as acceptorand donor, respectively.
The wild-type and triple-mutant OleD were compared by deter-mining steady state kinetic parameters with 4 or 2 as variablesubstrates, as described in the Supplementary Methods online. Thewild-type enzyme could not be saturated with 4 (SupplementaryFig. 2 online), even at the solubility limit of 4 in water/DMSO(10 mM), which indicates that 4 is a poor substrate for wild-typeOleD. Nevertheless, a kcat/Km value of 0.18 mM–1 min–1 was deter-mined by linear regression. Saturation was observed by varying donor2 at a fixed concentration of 4 (10 mM) to provide an apparent Km of0.3 mM. However, the Vmax from this analysis is unlikely to representthe true kcat. The steady state kinetics of the triple-mutant P67T/S132F/A242V were very different from those of the wild-type enzyme.Saturation with 4 could be achieved to give an apparent Km of0.14 mM and a kcat of 1.48 min–1 (Supplementary Fig. 2). Anapparent donor Km of 0.03 mM (a ten-fold improvement over
O
OH
O
O
O
O
O OH
OHO
NMe2
OMe
OO
OH
O
O
O
O
O OH
O
NMe2
OMe
O
OHO
HO OH
OHO O
HOHO
O UDP
OH
Wild-type OleD
OH
OHO OO OHO O
HOOH
OVariant OleD
(Fluorescent) (Nonfluorescent)
Oleandomycin
(1)
UDP-Glc (2)
Glycosylated oleandomycin
(3)
4-Methylumbelliferone
(4)
4-Methylumbelliferyl
β-D-glucopyranoside
(5)
2
O
OHR
HO
OR
OPO32–
OR
O NDP
OR O
E1 E2 GTs
a
b
OH
Figure 1 Enzymatic glycosylation. (a) General overview of enzymatic
glycorandomization. E1 represents a flexible anomeric sugar kinase,
E2 represents a flexible sugar-1-phosphate nucleotidyltransferase, GT
represents a flexible glycosyltransferase and the gray oval represents a
complex natural-product scaffold. (b) The native macrolide glucosyl-
transferase reaction catalyzed by OleD (upper reaction), and the
4-methylumbelliferone (4) glucosylation reaction used for OleD
directed evolution (lower reaction).
Table 1 Wild-type and mutant OleD glucosylation rates with acceptors 1, 4 and 27–32
Enzyme
Acceptor WT P67T S132F A242VP67T/
S132F/A242VFold
improvementa
Oleandomycin (1) 700 200 1,487 70 13 0.02
4-Methylumbelliferone (4) 2.5 15.7 6.3 5.4 84 33
7-Hydroxycoumarin-4-acetic acid (27) NDb 0.22 0.08 0.16 3.6 >180
7-Hydroxycoumarin-3-carboxylic acid (28) 0.01 0.022 0.022 0.01 0.62 62
Novobiocic acid (29) 0.24 0.65 0.49 0.33 1.14 4.8
Kaempferol (30) 13.2 99.9 23 15.3 28 2.1
Daidzein (31) 4.2 16.0 5.5 8.5 24 5.7
Genistein (32) 3.4 28.8 8.8 14.1 26 7.7
O
OH
HO O
HN
O
OH
O
OOH
HO
OH
OH
O
O
HO
OH
O
OOH
HO
OH
OHO O
O
OH
OHO O
O
OH
27
28
29
30
31
32
7
7
7
7
7
7
5 3
4'
4'
4'5
Rates of glucoside formation are shown in nanomoles of product formed per minute per mg of enzyme. See Methods for assay conditions and detection details.aFold improvement of P67T/S132F/A242V compared with wild-type OleD. For glucosyltransfer to 1, standard error of the assay is ±12% of the rate. For transfer to 4 or27, 28, 29, 30, 31, 32, standard error is less than 10%. bND, not detected (o0.01 nanomoles of product formed per minute per mg of enzyme). See Figure 1 for structuresof acceptors 1 and 4.
L E T TERS
65 8 VOLUME 3 NUMBER 10 OCTOBER 2007 NATURE CHEMICAL BIOLOGY
wild-type OleD) was determined using a fixed concentration of 4 andvarying the concentration of 2. Moreover, the triple-mutant kcat valuesdetermined with either donor 2 or acceptor 4 as the variable substratewere in good agreement. Thus, in the context of 4 glucosylation, thetriple mutant was B60-fold more efficient than wild-type OleD.
Variant enzymes identified via directed evolution are often found tobe promiscuous toward substrates well beyond those originallyscreened against18. To assess the impact of directed evolution onsugar nucleotide donor promiscuity, wild-type OleD and the P67T/S132F/A242V mutant were probed using a simple end-point assaywith a library of 20 potential ‘unnatural’ UDP donors (6–26) as well asUDP-Glc and dTDP-Glc in the presence of 4 as the acceptor (Fig. 3a).This library included both commercially available sugar nucleotides(2, 6 and 10) and unnatural sugar nucleotides generated via che-moenzymatic synthesis, cumulatively representing alterations of thesugar at C¢1, C¢2, C¢3, C¢4 or C¢6. The product identities wereconfirmed by LC-MS and coelution with commercial standardswhere available (Supplementary Table 1 online). Of the 22 sugarnucleotides tested, only UDP-Glc (2), UDP-6-deoxyglucose (8) andUDP-6-thioglucose (9) led to detectable product with wild-type OleD,ranging from 1% to 14% conversion in 3 h (Fig. 3b and Supplemen-tary Table 1). Thus, with the exception of very limited tolerance to C¢6modification, wild-type OleD showed stringent sugar donor specifi-city. In contrast, the evolved triple mutant P67T/S132F/A242Vaccepted 15 of 22 sugar nucleotide donors examined, 12 of whichwere not detectable substrates of wild-type OleD, with syntheticimprovements ranging from 7- to 300-fold. Notably, tolerance tothe sugar C¢6 modification was most enhanced to allow new
WT
7B9
2C38B3
P67T
S132F
A242VG340W
I112T
P67T/S132F/A242V
0 I II III
2
4
6
8
40
0
Generation
Spe
cific
act
ivity
(nm
ols
prod
uct m
in–1
mg–
1 )
b
0 20 40 60 80 1002,000
3,000
4,000
5,000
6,000
7,000
Act
ivity
(∆
fluor
esce
nce
units
at 3
h)
a
WT OleD
pET28a
7B9
2C3
8B3
Figure 2 Outcome of OleD directed evolution. (a) Representative screeningdata for the glycosylation of fluorescent 4 illustrating B100 random
members from a mutant OleD library and the positive hits 2C3, 7B9 and
8B3. (b) Progression of GT activity toward 4. Clone 2C3 had a single amino
acid change (A242V), whereas clones 7B9 and 8B3 each had two mutations
(S132F/G340W and P67T/I112T, respectively). WT, wild-type.
HO OHO
HOUDP
7
a
b
HO OHO
HO
SH
UDP
9 10
HO OHO
HOUDP
8
OHO
HO
OH
UDP
12
H2N OHO
HO
OH
UDP
13
HO OHO
HO
N3
UDP
11
OHO
HOUDP
14
H2N OHO
HOUDP
15
HO OH2N
HOUDP
18
HO O
HO
OH
UDP
16
HO OH2N
HO
OH
UDP
17
HO OHO
H2N
OH
UDP
19
HO OHO
AcHN
OH
UDP
20
HO ON3
HO
OH
UDP
21
N3 OHO
HO
OH
UDP
22
HO
OHO
HO
OH
UDP
HO
OHO
HOOH
UDP
HO OHO
HOOH
UDP
24 25
HO OAcHN
HO
OH
UDP
23 26
HO O
HOHO
U(T)DP
OH
2/6
C′6-modified
C′4-modified C′3-modified
C′2-modified
UD
P-G
lc (
2)
TD
P-G
lc (
6)
UD
P-x
ylos
e (7
)
UD
P-6
-deo
xy-G
lc (
8)
UD
P-6
-thi
o-G
lc (
9)
UD
P-6
-azi
do-G
lc (
11)
UD
P-4
-deo
xy-G
lc (
12)
UD
P-4
-am
ino-
Glc
(13
)
UD
P-4
,6-d
ideo
xy-G
lc (
14)
UD
P-4
-am
ino-
6-de
oxy-
Glc
(15
)
UD
P-3
-deo
xy-G
lc (
16)
UD
P-3
-am
ino-
Glc
(17
)
UD
P-3
-am
ino-
6-de
oxy-
Glc
(18
)
UD
P-2
-am
ino-
Glc
(19
)
UD
P-2
-ace
tam
ido-
Glc
(20
)
0
10
20
30
40
50
60
70
80
90
100 WT
P67T/S132F/A242V
Per
cent
age
conv
ersi
on (
3 h)
>293
>103
14
157>137
>73
>43
>150 >157
>27>13 >13
>80>70
HO OHO
HO
COOH
UDP
Figure 3 Activity of wild-type OleD and variant P67T/A242V/S132F toward
a set of NDP-sugar donors. (a) Structures of the donors tested. NDP-sugar
donors that were not detectable substrates for either enzyme are shown in
red. (b) Conversion rates of wild-type OleD and P67T/A242V/S132F towardeach donor with 4 as acceptor. Donors that did not show activity with either
enzyme were omitted for clarity (10, 21–26). Numbers above bars indicate
fold improvement from wild-type OleD, using an estimated minimal detec-
tion limit of 0.3% conversion for 4. See Supplementary Methods for assay
conditions and detection method.
L E T T ERS
NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 10 OCTOBER 2007 6 5 9
C¢6-modified glycosides, including analogs (UDP-6-thioglucose (9)and UDP-6-azidoglucose (11)) that present the potential for furtherdownstream chemoselective diversification5.
To estimate the impact of each individual mutation on donorspecificity, a subset of six donors was used to probe the specificity ofthe single mutants P67T, S132F and A242V (Supplementary Table 2online) with 4 as the acceptor in an identical end-point assay. Thissubset consisted of the natural donor UDP-Glc (2), UDP-xylose (7),and the 6-deoxy (8), 6-azido (11), 4-deoxy (12) and 2-amino (19)derivatives. Notably, all single mutants led to r2% conversion withdonor analogs 7, 11, 12 and 19, and they all led to poor conversion(r12%) with 8. Thus, the improved donor range of the triple mutantlikely derives from the synergistic combination of P67T, S132F andA242V. The same donor subset using 1 as acceptor and triple mutantP67T/S132F/A242V as the catalyst led to the desired product in nearlyquantitative yield in every case (Supplementary Fig. 3 online). Giventhe role of differential glycosylation in modulating the biologicaleffects of macrolides—including bacterial ribosome 50S ribosomeinhibition19, immunomodulation20, and inhibition of Golgi trans-port21—this preliminary study is an advance toward the synthesis ofsuch analogs.
Given the ability of wild-type OleD to glycosylate a range of smallphenolics15, we examined the acceptor specificity of the wild-type andmutant OleDs by measuring the rate of glucoside formation using apanel of six additional acceptors (27–32; Table 1), including thenatural macrolide substrate 1. In each case, the appearance of amajor new product peak was monitored over a suitable time intervalduring which the rate of formation was linear. The products wereidentified by LC-MS, comparison to commercial standards whereavailable, and fluorescence measurements (Supplementary Figs. 4 and5 online). Compared with the natural substrate 1, wild-type OleDshowed a 50- to 500-fold lower activity with screening target 4,flavonoid 30 and isoflavones 31 and 32, whereas the charged coumar-ins 27 and 28 were essentially nondetectable substrates for wild-typeOleD. However, aminocoumarin 29 was accepted by the wild-typeenzyme, albeit at a rate B3,000-fold less than with 1. Once again instark contrast, the mutant P67T/S132F/A242V activity toward the
entire panel of small phenolics (4 and 27–32) was improved comparedwith wild-type OleD. Surprisingly, the largest improvement in activitywas not observed with the screening target 4; instead, it was observedwith 7-hydroxycoumarin-4-acetic acid (27). In addition to their well-known antioxidant activities, glycosylated coumarins and flavonoidshave diverse biological activities including anticancer22, antiangiogen-esis23, anti-HIV24 and anti-inflammatory25 effects. The modestimprovement toward aminocoumarin 29 is also notable, particularlyin the context of the observed donor promiscuity of P67T/S132F/A242V, as the antibiotic novobiocin (33) (a glycosylated version of 29)is an established inhibitor of DNA gyrase and induces degradation ofHsp90-dependent client proteins26. Concomitant with the enhance-ment of activity toward small aromatic acceptors, the evolved triplemutant glucosylated the natural acceptor 1 at a rate B54-fold less thanthat of the wild-type enzyme. However, the single mutant S132F wasB2-fold more active toward 1 than wild-type OleD. Apart from thisdifference, the single mutants P67T, S132F and A242V were less activethan the triple mutant toward all the acceptors tested; the P67Tmutation was clearly the most advantageous single functional mutation.
The recently determined structure of OleD allows structural inter-pretation for the functional effects of the mutations discovered in thisstudy13. The most advantageous single functional mutation at Pro67 ispresent in a loop region (amino acids 60–76, loop N3) that containsseveral prolines and follows b-sheet 3 in the N-terminal domain(Fig. 4). This loop is hypervariable in other GT-B fold GTs andconstitutes part of the acceptor binding site27. For example, the loopimmediately following Nb3 in GtfA (residues 57–72, Fig. 4b) forms abroad binding surface containing two prolines at positions 68 and 69(ref. 28). Coincidentally, this loop has been interrogated by mutationin at least one other reported example aimed at identifying theresidues responsible for donor selectivity in the urdamycin GTs. Inthe previous UrdGT mutagenesis study, sequence alignments wereused to guide the construction of a series of defined hybrid andresidue-exchanged UrdGT1b and UrdGT1c enzymes, the activitiesof which were assayed using a low-throughput HPLC assay29.Consequently, a 31-amino-acid region (residues 52–82) was deter-mined to be responsible for donor specificity, and mutation at a singleposition equivalent to a proline (Pro56) within this region was foundto be sufficient to alter UrdGT specificity30. In conjunction with thepresent OleD directed evolution study, this previous GT mutagenesisstudy highlights the importance of key prolines within the GT-B loopN3, which may (in part) define GT-B acceptor specificity.
Ser132 is also closely associated with the active site of OleD, whichis located at the N terminus of b-strand 5 in the N-terminal domain ofOleD (Fig. 4a). This residue is partially conserved as serine, threonineor alanine in related natural-product GTs. Comparison to thesequence and structure of the plant flavonoid GT-B fold enzyme
OleD 55 ARPVLYHSTLPGPDADP-----EAWGSTLLDNVEPVvGT1 55 DSMHTMQCNIKSYDISDGVPEGYVFAGRPQEDIELGtfA 48 VPMVPVGRAVRAGAREP--------GELPPGAAEV
OleD 109 LHDITSYPARVLARRWGVPAVSL PNLVAWKGVvGT1 117 VADAFIWFAADMAAEMGVAWLPFWT
SAGPNSLSTHV
GtfA 102 LP--AAVAVRSMAEKLG---IPYRYTVLSPDHLPS
OleD 236 LVSLGSAFTKQPAFYRECVRAFGNLPG---WHLVLVvGT1 275 YISFGTVTTPPPAEVVALSEALEASRVPFIWSLRDGtfA 225 YVGFGSSSRPATADAAKMAIKAVRASG-----RRI
8
9
12
3
4
5
CN
aLoop N3
Nβ5 Nα5
Cβ8 Cα7 Cβ9
Nβ6
b Loop N3Nβ3 Nα3
67
---
Figure 4 Location of functional amino acid mutations. (a) Crystal structure
of OleD (PDB ID 21YF) highlighting the locations of functional amino acid
mutations as colored spheres: Pro67, red; Ser132, magenta; Ala242, green.
Loop N3 is shown in yellow, UDP and non-natural acceptor erythromycin are
shown in cyan and orange, respectively, and b-strands, where visible, are
numbered sequentially. (b) Sequence alignment of OleD (Entrez Protein
accession code ABA42119), the plant GT VvGT1 (Entrez Protein accession
code AAB81683), and GtfA (Entrez Protein accession code AAB49292).
Secondary structure of OleD is shown above the OleD sequence. b-sheets,
blue; a-helices, red; loop N3, yellow. OleD Pro67 is highlighted red, OleD/
VvGT1 Ser132/Thr141 are highlighted magenta, OleD Ala242 is highlighted
green, and the conserved serine/threonine that likely interacts with
a-phosphate of the donor is highlighted black.
L E T TERS
66 0 VOLUME 3 NUMBER 10 OCTOBER 2007 NATURE CHEMICAL BIOLOGY
VvGT1 (ref. 31) reveals a likely role for Ser132 in binding the donorsugar. The VvGT1 equivalent to the OleD Ser132 (Thr141) forms ahydrogen bond with the C6 OH of UDP-2-deoxy-2-fluoroglucose.Thus, mutation of Ser132 in OleD may alter binding of thedonor. Similarly, Ala242 is partially conserved as alanine, serineor threonine in related GTs. In the OleD structure, Ala242 followsSer241, which forms a hydrogen bond to the a-phosphate of thedonor—an interaction that is also observed in the VvGT1 struc-ture. Mutation of Ala242 may therefore affect binding of the di-phosphate moiety of the UDP donor or in turn alter binding of thesugar moiety.
In an effort to overcome the limited substrate specificity of anatural-product GT, we have used directed evolution to improve thepromiscuity of the macrolide GT OleD. Aside from the considerableexpansion of both donor and acceptor specificity of the evolvedenzyme, a number of additional key elements of this study are ofimportance. First, this work illustrates the ability to substantially alterGT specificity and proficiency via a single (or a few combined)mutation and thereby provides promise for future GT engineeringefforts where assay design may limit throughput. Second, this studylends support to the observation that an increase in enzyme profi-ciency leads to an increase in promiscuity32, and it also reveals thepotential to create promiscuous variants of other natural-product GTssimply by screening for efficiency toward a single acceptor-donor pair.Third, this study exposes the GT-B fold loop N3 as a potential focalpoint for future GT engineering efforts and, in the context of OleD,presents a new scaffold for further rational redesign or directedevolution. Finally, given that many of the acceptors for the newlyevolved GT are therapeutically relevant (for example, macrolides,flavonoids, isoflavones, aminocoumarins and coumarins), this broadlypromiscuous ‘universal’ GT presents opportunities in drug discoveryand will also enhance ongoing efforts to develop bacterial hosts forin vivo glycorandomization33.
METHODSGeneral. Bacterial strain E. coli BL21(DE3)pLysS was purchased from Strata-
gene. NovaBlue was purchased from Novagen. Plasmid pET28/OleD was a gift
(see Acknowledgments), and pET28a was purchased from Novagen. All other
chemicals were reagent-grade and purchased from Fluka, New England Biolabs
or Sigma, unless otherwise stated. Primers were ordered from Integrated DNA
Technologies. Oleandomycin was purchased from MP Biomedicals Inc. Phe-
nolic substrates (27, 28, 30, 31, 32) were from Indofine Chemical Company
Inc. Novobiocic acid (29) was prepared as previously described from Novo-
biocin (33)8. Product standard 4-methylumbelliferyl-7-O-b-D-glucoside (5)
was purchased from Sigma, and daidzein-7-O-b-D-glucoside (34) standard
was purchased from Fluka. Analytical HPLC was performed on a Rainin
Dynamax SD-2/410 system connected to a Rainin Dynamax UV-DII absor-
bance detector. Mass spectra were obtained using electrospray ionization on an
Agilent 1100 HPLC-MSD SL quadrupole mass spectrometer connected to a
UV-Vis diode array detector. For LC-MS analysis, quenched reaction mixtures
were analyzed by analytical reverse-phase HPLC with a 250 mm � 4.6 mm
Gemini 5m C18 column (Phenomenex), using a gradient of 10–90% CH3CN in
0.1% formic acid/H2O for 20 min at 1 ml min–1, with detection at 254 nm.
Library preparation. The random mutant library was prepared via error-prone
PCR using the Stratagene GeneMorph II Random Mutagenesis Kit, as described
by the manufacturer with pET28/OleD as template. The primers used for
amplification of the OleD gene were T7 FOR (5¢-TAA TAC GAC TCA CTA TAG
GG-3¢) and T7 REV (5¢-GCT AGT TAT TGC TCA GCG G-3¢). Amplified
product was digested with NdeI and HindIII, purified by agarose gel electro-
phoresis (0.8% w/v agarose), extracted using the QIAquick Gel Extraction Kit
(QIAgen), and ligated into similarly treated pET28a. The ligation mixtures were
used to transform chemically competent NovaBlue cells and single colonies
used to prepare plasmids for DNA sequencing, which revealed that the library
had the desired mutation rate of 1 or 2 amino acid mutations per gene product.
Subsequently, all the transformants from this library were pooled and cultured
overnight. Plasmids were prepared from this culture and used to transform
chemically competent E. coli BL21(DE3)pLysS, which was screened as
described below.
Site-directed mutagenesis. Site-specific OleD variants were constructed using
the Stratagene QuikChange II Site-Directed Mutagenesis Kit, as described by
the manufacturer. Constructs were confirmed to carry the correct mutations via
DNA sequencing.
Screening. Individual colonies were used to inoculate wells of a 96-deep-well
microtiter plate in which each well contained 1 ml of LB medium supple-
mented with 50 mg ml–1 kanamycin. Culture plates were tightly sealed with
AeraSea breathable film (Research Products International Corp.). After cell
growth at 37 1C for 18 h with shaking at 350 r.p.m., 100 ml of each culture was
transferred to a fresh deep-well plate containing 1 ml of LB medium
supplemented with 50 mg ml–1 kanamycin. The original plate was sealed and
stored at 4 1C, or a glycerol copy was made by mixing 100 ml of each culture
with 100 ml of 50% (v/v) glycerol and stored at –80 1C. The freshly inoculated
plate was incubated at 37 1C for 2–3 h with shaking at 350 r.p.m. Expression of
the N-terminal His6-tagged OleD was induced at an optical density at 600 nm
(OD600) of B0.7 via the addition of IPTG to a final concentration of 0.4 mM,
and the plate was incubated for 18 h at 18 1C. Cells were harvested by
centrifugation at 3,000g for 10 min at 4 1C, the cell pellets were thoroughly
resuspended in 50 mM Tris-HCl (pH 8.0) containing 10 mg ml–1 lysozyme
(Sigma) at 4 1C, and the plates were subjected to a single freeze-thaw cycle
to lyse the cells. Cell debris was then collected by centrifugation at 3,000g
for 20 min at 4 1C, and 50 ml of the cleared supernatant was used for
enzyme assay.
For the assay, cleared supernatant was mixed with an equal volume (50 ml) of
50 mM Tris-HCl (pH 8.0) containing 10 mM MgCl2, 0.2 mM 4 and 1.0 mM 2
using a Biomek FX Liquid Handling Workstation (Beckman Coulter). Upon
mixing, the initial fluorescence (lex ¼ 350 nm, lex ¼ 460 nm) was measured
using a FLUOstar Optima plate reader (BMG Labtechnologies). The reactions
were subsequently incubated for 3 h at 30 1C, at which time the fluorescence
measurement was repeated. Activity of the clones was expressed as the
difference in fluorescence intensity between 0 h and 3 h.
Other methods. For all other methods, including probing acceptor/donor
specificity, sugar nucleotide synthesis, and determination of kinetic parameters,
see Supplementary Methods.
Accession codes. Protein Data Bank: The OleD crystal structure can be
found as PDB ID 2IYF. Entrez Protein: the sequence alignment of OleD
can be found under accession code ABA42119, the sequence of plant GT
VvGT1 is under accession code AAB81683, and the sequence of GtfA is
under accession code AAB49292. All sequences were deposited as part of
previous studies.
Note: Supplementary information and chemical compound information is available onthe Nature Chemical Biology website.
ACKNOWLEDGMENTSWe are grateful to the School of Pharmacy Analytical Instrumentation Center foranalytical support, H.-W. Liu (University of Texas-Austin) for plasmid pET28/OleD and S. Singh for helpful discussions. This work was supported in part bythe US National Institutes of Health grants AI52218 and U19 CA113297. J.S.T. isa University of Wisconsin H.I. Romnes Fellow.
AUTHOR CONTRIBUTIONSG.J.W. contributed to the experimental design, experimental execution andmanuscript drafting; C.Z. contributed experimental reagents and consultation;and J.S.T. contributed to the experimental design and manuscript drafting.
COMPETING INTERESTS STATEMENTThe authors declare competing financial interests: details accompany the full-textHTML version of the paper at http://www.nature.com/naturechemicalbiology/.
L E T T ERS
NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 10 OCTOBER 2007 6 6 1
Published online at http://www.nature.com/naturechemicalbiology
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions
1. Weymouth-Wilson, A.C. The role of carbohydrates in biologically active natural pro-ducts. Nat. Prod. Rep. 14, 99–110 (1997).
2. Thorson, J.S. & Vogt, T. in Carbohydrate-Based Drug Discovery (ed. Wong, C.-H.)685–712 (Wiley-VCH, Weinheim, Germany, 2002).
3. Ahmed, A. et al. Colchicine glycorandomization influences cytotoxicity and mechanismof action. J. Am. Chem. Soc. 128, 14224–14225 (2006).
4. Griffith, B.R., Langenhan, J.M. & Thorson, J.S. ‘Sweetening’ natural products viaglycorandomization. Curr. Opin. Biotechnol. 16, 622–630 (2005).
5. Fu, X. et al. Antibiotic optimization via in vitro glycorandomization. Nat. Biotechnol.21, 1467–1469 (2003).
6. Zhang, C., Albermann, C., Fu, X. & Thorson, J.S. The in vitro characterization of theiterative avermectin glycosyltransferase AveBI reveals reaction reversibility and sugarnucleotide flexibility. J. Am. Chem. Soc. 128, 16420–16421 (2006).
7. Zhang, C. et al. Exploiting the reversibility of natural product glycosyltransferase-catalyzed reactions. Science 313, 1291–1294 (2006).
8. Albermann, C. et al. Substrate specificity of NovM: implications for novobiocinbiosynthesis and glycorandomization. Org. Lett. 5, 933–936 (2003).
9. Zhang, C., Fu, Q., Albermann, C., Li, L. & Thorson, J.S. The in vitro characterization ofthe erythronolide mycarosyltransferase EryBV and its utility in macrolide diversification.ChemBioChem 8, 385–390 (2007).
10. Hu, Y. & Walker, S. Remarkable structural similarities between diverse glycosyltrans-ferases. Chem. Biol. 9, 1287–1296 (2002).
11. Hancock, S.M., Vaughan, M.D. & Withers, S.G. Engineering of glycosidases andglycosyltransferases. Curr. Opin. Chem. Biol. 10, 509–519 (2006).
12. Aharoni, A. et al. High-throughput screening methodology for the directed evolution ofglycosyltransferases. Nat. Methods 3, 609–614 (2006).
13. Bolam, D.N. et al. The crystal structure of two macrolide glycosyltransferases providesa blueprint for host cell antibiotic immunity. Proc. Natl. Acad. Sci. USA 104,5336–5341 (2007).
14. Quiros, L.M., Aguirrezabalaga, I., Olano, C., Mendez, C. & Salas, J.A. Two glycosyl-transferases and a glycosidase are involved in oleandomycin modification during itsbiosynthesis by Streptomyces antibioticus. Mol. Microbiol. 28, 1177–1185(1998).
15. Yang, M. et al. Probing the breadth of macrolide glycosyltransferases: in vitroremodeling of a polyketide antibiotic creates active bacterial uptake and enhancespotency. J. Am. Chem. Soc. 127, 9336–9337 (2005).
16. Mayer, C. et al. Directed evolution of new glycosynthases from Agrobacterium beta-glucosidase: a general screen to detect enzymes for oligosaccharide synthesis. Chem.Biol. 8, 437–443 (2001).
17. Collier, A.C., Tingle, M.D., Keelan, J.A., Paxton, J.W. & Mitchell, M.D. A highlysensitive fluorescent microplate method for the determination of UDP-glucuronosyltransferase activity in tissues and placental cell lines. Drug Metab. Dispos. 28,1184–1186 (2000).
18. Carr, R. et al. Directed evolution of an amine oxidase possessing both broad substratespecificity and high enantioselectivity. Angew. Chem. Int. Edn Engl. 42, 4807–4810(2003).
19. Katz, L. & Ashley, G.W. Translation and protein synthesis: macrolides. Chem. Rev. 105,499–528 (2005).
20. Amsden, G.W. Anti-inflammatory effects of macrolides–an underappreciated benefit inthe treatment of community-acquired respiratory tract infections and chronic inflam-matory pulmonary conditions. J. Antimicrob. Chemother. 55, 10–21 (2005).
21. Bonay, P., Munro, S., Fresno, M. & Alarcon, B. Intra-Golgi transport inhibition bymegalomicin. J. Biol. Chem. 271, 3719–3726 (1996).
22. Lacy, A. & O’Kennedy, R. Studies on coumarins and coumarin-related compounds todetermine their therapeutic role in the treatment of cancer. Curr. Pharm. Des. 10,3797–3811 (2004).
23. Yuan, H.Q., Junker, B., Helquist, P. & Taylor, R.E. Synthesis of anti-angiogenicisocoumarins. Curr. Org. Synth. 1, 1–9 (2004).
24. Williams, C.A. & Grayer, R.J. Anthocyanins and other flavonoids. Nat. Prod. Rep. 21,539–573 (2004).
25. Fylaktakidou, K.C., Hadjipavlou-Litina, D.J., Litinas, K.E. & Nicolaides, D.N. Naturaland synthetic coumarin derivatives with anti-inflammatory/antioxidant activities. Curr.Pharm. Des. 10, 3813–3833 (2004).
26. Burlison, J.A., Neckers, L., Smith, A.B., Maxwell, A. & Blagg, B.S. Novobiocin:redesigning a DNA gyrase inhibitor for selective inhibition of Hsp90. J. Am. Chem.Soc. 128, 15529–15536 (2006).
27. Mulichak, A.M., Lu, W., Losey, H.C., Walsh, C.T. & Garavito, R.M. Crystal structureof vancosaminyltransferase GtfD from the vancomycin biosynthetic pathway: interac-tions with acceptor and nucleotide ligands. Biochemistry 43, 5170–5180 (2004).
28. Mulichak, A.M. et al. Structure of the TDP-epi-vancosaminyltransferase GtfA from thechloroeremomycin biosynthetic pathway. Proc. Natl. Acad. Sci. USA 100, 9238–9243(2003).
29. Hoffmeister, D., Ichinose, K. & Bechthold, A. Two sequence elements ofglycosyltransferases involved in urdamycin biosynthesis are responsible for substratespecificity and enzymatic activity. Chem. Biol. 8, 557–567 (2001).
30. Hoffmeister, D. et al. Engineered urdamycin glycosyltransferases are broadened andaltered in substrate specificity. Chem. Biol. 9, 287–295 (2002).
31. Offen, W. et al. Structure of a flavonoid glucosyltransferase reveals the basis for plantnatural product modification. EMBO J. 25, 1396–1405 (2006).
32. Oberth+ur, M. et al. A systematic investigation of the synthetic utility ofglycopeptide glycosyltransferases. J. Am. Chem. Soc. 127, 10747–10752 (2005).
33. Hui, J.P., Yang, J., Thorson, J.S. & Soo, E.C. Selective detection of sugar phosphates bycapillary electrophoresis/mass spectrometry and its application to an engineered E. colihost. ChemBioChem 8, 1180–1188 (2007).
L E T TERS
66 2 VOLUME 3 NUMBER 10 OCTOBER 2007 NATURE CHEMICAL BIOLOGY
A simple selection strategy forevolving highly efficient enzymes
Martin Neuenschwander, Maren Butz,Caroline Heintz, Peter Kast & Donald Hilvert
Combining tunable transcription with an enzyme-degradation
tag affords an effective means to reduce intracellular enzyme
concentrations from high to very low levels. Such fine-tuned
control allows selection pressure to be systematically
increased in directed-evolution experiments. This facilitates
identification of mutants with wild-type activity, as shown
here for an engineered chorismate mutase. Numerous selection
formats and cell-based screening methodologies may benefit
from the large dynamic range afforded by this easily
implemented strategy.
Genetic selection can greatly facilitate the search for rare catalysts invery large protein libraries1. Auxotrophic strains, which grow only ifprovided with a protein that functionally replaces a missing cellularenzyme, are frequently used for this purpose. However, growthrepresents an indirect (and imperfect) readout of catalyst activity.Even mediocre catalysts may provide sufficient activity for cells togrow at wild-type levels, making it difficult to distinguish the mostactive variants from their less effective counterparts. As a result, it canbe difficult or impossible to optimize relatively inefficient enzymesthrough multiple rounds of mutagenesis and selection.
We faced this problem when we tried to improve an engineeredchorismate mutase. The dimeric helical bundle chorismate mutasefrom Escherichia coli (EcCM) was successfully converted into a func-tional hexamer (hEcCM) by inserting a five-amino-acid hinge loop intothe middle of the long H1 helix spanning the parent dimer2, but thetopological change was accompanied by a 2- to 3-order-of-magnitudedecrease in activity (Table 1). Activity was partially recovered whenhEcCM was subjected to two rounds of random mutagenesis andselection in a chorismate mutase–deficient E. coli strain (KA12/pKIMP-UAUC3; Fig. 1a)4. Nevertheless, the best variant, tEcCM,which contained three mutations and possessed a trimeric quaternarystructure, still had a 14-fold lower kcat value than the parent EcCMdimer (Table 1), and further improvements were not possible becausetEcCM already conferred wild-type levels of growth to its host4.
In theory, selection pressure in a complementation assay can beincreased by decreasing the intracellular catalyst concentration.This might be accomplished by switching to a weakly active promoter,low gene dose or inefficient ribosomal binding sites for catalystproduction. Such strategies have been profitably exploited for the
directed evolution of aspartate transaminases5, for example. As reclon-ing of a library is laborious and prone to loss of diversity, and the effecton selection pressure only qualitatively predictable, inducible systemsthat provide a tightly controlled and graded transcriptional response toan external inducer represent potentially attractive alternatives.
Regulable promoters that use arabinose6 or tetracycline7,8 as indu-cer compounds have been well characterized. Both allow homo-geneous gene expression over a broad dynamic range. To constructselection plasmids that combine inducer dose-dependent gene expres-sion with the convenience of high-copy plasmids, we opted for thetetracycline-inducible Ptet system7, which does not require specificallyengineered host strains. We replaced the weak, constitutive blapromoter on our selection plasmid pKECMB2 (Fig. 1a) with amodified Ptet promoter cassette (Fig. 1b), which includes a down-stream T7 promoter, to simplify protein overproduction of selectedvariants. As expected, auxotrophic KA12/pKIMP-UAUC cells harbor-ing the gene for the very weakly active hEcCM under the control ofthis promoter system (on plasmid pKT) did not grow under selectiveconditions in the absence of inducer, but they regained prototrophy athigh tetracycline concentrations (Table 2). In contrast, cells containingthe more active tEcCM and EcCM variants grew even in the absence ofinducer. Thus, background transcription affords sufficient amounts ofthese more active catalysts to fully satisfy the metabolic needs of thecell. Clearly, in this case, tight transcriptional control alone is notsufficient to reduce protein concentrations to a level low enough toallow discrimination between a moderately active catalyst (tEcCM)and the parent enzyme (EcCM).
To reduce protein concentration further, we fused an 11-amino-acidSsrA degradation signal9 to the C terminus of the catalyst (Fig. 1c).The SsrA tag targets the catalyst for rapid degradation by theintracellular ClpXP protease. The efficacy of this strategy was estab-lished using green fluorescent protein as a reporter (SupplementaryFig. 1 and Supplementary Methods online). The presence of thedegradation tag in plasmid pKTS also increased the dynamic range ofthe chorismate mutase selection system, as demonstrated by
Table 1 Catalytic parameters of the evolved hinge-loop variants
Varianta kcat (s–1) Km (mM) kcat/Km (M–1s–1)
EcCMb 14 350 41,000
hEcCMb 0.15 2,100 70
tEcCMb 1.0 34 30,000
EcCM-200/4c 12 ± 1 270 ± 20 45,000
aThe sequences of the variants are provided in Supplementary Figure 4 online. bRef. 4,pH 6.5. The catalytic parameters of EcCM are similar at neutral and acidic pH. cThis work.
Protein production and characterization is described in Supplementary Methods.Kinetic measurements were performed at 20 1C in PBS (10 mM Na2HPO4, 2 mMKH2PO4, 137 mM NaCl, 2.7 mM KCl, pH 7.4) supplemented with 0.1 mg/ml bovineserum albumin.
Received 21 March; accepted 24 August; published online 16 September 2007; doi:10.1038/nbt1341
Laboratory of Organic Chemistry, ETH Zurich, Honggerberg HCI F 339, CH-8093 Zurich, Switzerland. Correspondence should be addressed to D.H.([email protected]).
NATURE BIOTECHNOLOGY VOLUME 25 NUMBER 10 OCTOBER 2007 1 14 5
BR I E F COMMUN I CAT I ONS©
2007
Nat
ure
Pub
lishi
ng G
roup
ht
tp://
ww
w.n
atur
e.co
m/n
atur
ebio
tech
nolo
gy
complete suppression of cell growth in the absence of inducer for alltested chorismate mutase variants (Table 2). Restoration of growthupon addition of specific tetracycline concentrations on solid mediaroughly correlated with specific activity, distinguishing the mostweakly active variant, hEcCM, from the more active tEcCM andEcCM variants. The correlation between growth rate, tetracyclineconcentration and specific activity was even more apparent in liquidculture (Fig. 2a). At the highest tetracycline concentrations tested,wild-type growth rates were achieved with the tEcCM and EcCMvariants, but not with the weakly active hEcCM. Moreover, atintermediate tetracycline concentrations, cells harboring tEcCMgrew more slowly than those with wild-type EcCM. The ability tocontrol selection stringency simply by adjusting tetracycline concen-tration raises the possibility of evolving topologically novel catalyststhat are more active than tEcCM.
To test the utility of this system in a directed evolution experiment,we inserted library fragments that encode the first 93 residues ofhEcCM, diversified by error-prone PCR and DNA shuffling4, into thepKTS acceptor vector in place of a stuffer fragment, in-frame with thelast seven residues of hEcCM fused to the SsrA tag. After transforma-tion of the KA12/pKIMP-UAUC selection strain (1.5 � 107 transfor-mants), library clones were picked randomly and sequenced to checklibrary quality. With the exception of the R44C substitution, whichoccurred in 40% of the sequences because of an apparent DNAshuffling artifact, mutations were evenly distributed over the entirehEcCM gene. Aliquots of the library were plated onto selective M9cplates containing varying amounts of tetracycline to identify activecatalysts based on their ability to complement the chorismate mutasedeficiency3. The number of complementing clones decreased withdecreasing tetracycline concentration (Fig. 2b), consistent with thehypothesis that reducing intracellular enzyme concentration increasesselection pressure.
Sequence analysis of 96 clones revealed a mutational bias in activevariants that correlates roughly with selection stringency. Theemergence and disappearance of specific mutations upon increasingselection pressure is illustrated in Figure 2c (see also SupplementaryFig. 2 online). At a complementation frequency of 14 ± 5%
(400 ng/ml tetracycline), an H66R mutation occurred frequentlyand was further enriched at a complementation frequency of 2 ± 1%(200 ng/ml tetracycline). An S42L mutation was also enriched underthe latter conditions, whereas the R44C mutation, which was prevalentbefore selection, occurred less frequently, presumably because itprovides no catalytic benefit. As selection pressure further increased(r100 ng/ml tetracycline), the complementation frequency droppedto o0.1% and only false positives were observed. The latter had lostthe degradation tag mainly through frameshift mutations, and weretherefore presumably produced at elevated concentrations. The factthat the most stringent conditions only yielded false positives illus-trates the importance of fine-tuning the selection pressure in theseexperiments to maximize the yield of highly active variants.
The genes of six clones selected at a complementation frequency of2 ± 1% were retransformed, and the transformants grew faster than sixout of seven variants selected at a complementation frequency of 14 ±5% (Supplementary Fig. 3 online). For six fast-growing variants, thedegradation tag was replaced with a (His)6 tag for protein purification,and the enzymatic activity of the four variants that could be producedin soluble form was determined in vitro. They showed uniformly highVmax values comparable to EcCM (Supplementary Table 1 online).The highest catalytic efficiency was exhibited by variant EcCM-200/4,which contained four mutations relative to hEcCM (A9V, S42L, H66R,T87I; see also Supplementary Fig. 4 online). It eluted from a gelfiltration column as a trimer and catalyzed the rearrangement ofchorismate to prephenate with a kcat of 12 s–1 and a kcat/Km of 45,000M–1s–1 (Table 1). The turnover number, which represents a 75-foldimprovement over hEcCM and a tenfold improvement relative to thebest previously characterized variant tEcCM, is similar to that ofwild-type EcCM (Table 1). This result is notable as high kcat values are
tetR
pKTS
cm ssrAPtetA
CM
SsrA ClpXP
TetR tc TetRtc
pKECMB
Pbla
Chorismate
TyrPhe
Prephenate
cm CMCM
CMCM
CM
tetR
pKT
cmPtetA
TetR tc TetRtc
CM
CM
a
b
c
KA12/pKIMP-UAUC/pKECMB
KA12/pKIMP-UAUC/pKT
KA12/pKIMP-UAUC/pKTS
Table 2 Benchmark complementation assays with chorismate
mutases having different activities
pKT derivative encoding pKTS derivative encoding
[tc] (ng/ml) EcCM tEcCM hEcCM EcCM tEcCM hEcCM
0 + + 0 0 0 0
10 + + 0 0 0 0
50 + + + + + 0
300 + + + + + +
+, cell growth; 0, no cell growth. Streak-outs of KA12/pKIMP-UAUC cells containing theindicated selection plasmids were evaluated after 2 d of incubation at 30 1C on M9cmedium plates14. [tc] is the tetracycline concentration in M9c medium. pKT places thegene under control of the tetracycline-inducible Ptet system, pKTS additionally encodesa degradation tag fused to the catalyst.
Figure 1 Selection plasmid design. (a) The catalyst gene (cm) on
plasmid pKECMB2 is constitutively expressed under control of the bla
promoter. E. coli strain KA12 is deficient in chorismate mutase activity
and also requires plasmid pKIMP-UAUC, which encodes two prephenate-
processing enzymes, for growth on minimal medium3. (b) Selection
plasmid pKT provides graded and homogeneous transcriptional control
of catalyst production from promoter PtetA. The tetR gene and its promoter
region are located upstream of cm, so that the TetR repressor simultaneouslyregulates transcription of catalyst and TetR repressor genes13. The
tetracycline-resistance determinant of the Tn10 transposon (tetA) is
integrated in the KA12 chromosome3. (c) Selection plasmid pKTS permits
graded transcriptional control and limited enzyme half-life. The ssrA
sequence is incorporated as a downstream genetic fusion to the catalyst
gene. The resulting enzyme carries the degradation tag at its C terminus,
and is directed to the ClpXP protease, where it is degraded. Plasmid
construction details are provided in Supplementary Methods online.
1 14 6 VOLUME 25 NUMBER 10 OCTOBER 2007 NATURE BIOTECHNOLOGY
B R I E F C O M M U N I C AT I O N S©
2007
Nat
ure
Pub
lishi
ng G
roup
ht
tp://
ww
w.n
atur
e.co
m/n
atur
ebio
tech
nolo
gy
important for industrial biocatalysis, where high conversion of sub-strate to product is desired10.
In conclusion, pairing a tunable promoter with a degradation tagcan provide very low but adjustable catalyst concentrations withincells. By providing systematic control over selection stringency, thisstrategy facilitates the evolution of substantially more active enzymesthan is possible with systems reliant on weak constitutive geneexpression4. Extension of this approach for regulating stringencyshould benefit any cell-based selection1,11 or screening11,12 system inwhich the setting of a threshold for minimum activity allows the bestvariants to be distinguished from less interesting ones.
Note: Supplementary information is available on the Nature Biotechnology website.
ACKNOWLEDGMENTSThis work was generously supported by the Schweizerischer Nationalfonds andthe ETH Zurich.
AUTHOR CONTRIBUTIONSM.N., P.K. and D.H. designed research; M.N., M.B. and C.H. performed theexperiments; M.N., M.B., P.K. and D.H. analyzed data; M.N., P.K. and D.H.wrote the paper.
10 100 1,0000.00
0.04
0.08
0.12
0.16
0.20
400 200 100 50 25 00.001
0.01
0.1
1
10
Com
pl. f
requ
ency
(%
)
Gro
wth
rat
e (h
–1)
Tetracycline (ng/ml) Tetracycline (ng/ml)0.01 0.1 1 10 1000
20
40
60
80
100
Compl. frequency (%)
Occ
urre
nce
(%)
EcCMtEcCMhEcCM
R44CS42LH66RArtifacts
a b c
Figure 2 Tetracycline-dependent growth in selective M9c medium and influence of tetracycline concentration on the selection process. (a) KA12/pKIMP-
UAUC cells were transformed with the pKTS selection plasmid encoding wild-type EcCM, tEcCM or hEcCM. Growth curves were determined for each
transformant. Error bars indicate the s.d. of the curve fit in each growth experiment. (b) Complementation frequency among gene library members on
selective M9c plates as a function of tetracycline concentration. (c) Mutation bias as a function of complementation frequency, determined using 24
sequences originating from three independent selection experiments for each selection regime. The occurrence of false positives lacking the degradation tag
and the relatively frequent mutations H66R, S42L and R44C is plotted for different selection regimes. Clones grown on nonselective rich medium plates
(100% complementation) were examined to assess library size, quality and sequence diversity before selection. See Supplementary Figure 2 online for
alignments of all 96 sequences used for this analysis and the Supplementary Methods for detailed experimental protocols for the liquid growth tests, library
construction and selection experiments.
Published online at http://www.nature.com/naturebiotechnology
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions
1. Taylor, S.V., Kast, P. & Hilvert, D. Angew. Chem. Int. Ed. Engl. 40, 3310–3335(2001).
2. MacBeath, G., Kast, P. & Hilvert, D. Protein Sci. 7, 325–335 (1998).3. Kast, P., Asif-Ullah, M., Jiang, N. & Hilvert, D. Proc. Natl. Acad. Sci. USA 93,
5043–5048 (1996).4. Vamvaca, K., Butz, M., Walter, K.U., Taylor, S.V. & Hilvert, D. Protein Sci. 14,
2103–2114 (2005).5. Yano, T., Oue, S. & Kagamiyama, H. Proc. Natl. Acad. Sci. USA 95, 5511–5515
(1998).6. Khlebnikov, A., Risa, O., Skaug, T., Carrier, T.A. & Keasling, J.D. J. Bacteriol. 182,
7029–7034 (2000).7. Hansen, L.H., Ferrari, B., Sorensen, A.H., Veal, D. & Sorensen, S.J. Appl. Environ.
Microbiol. 67, 239–244 (2001).8. Lutz, R. & Bujard, H. Nucleic Acids Res. 25, 1203–1210 (1997).9. Karzai, A.W., Roche, E.D. & Sauer, R.T. Nat. Struct. Biol. 7, 449–455 (2000).10. Burton, S.G., Cowan, D.A. & Woodley, J.M. Nat. Biotechnol. 20, 37–45 (2002).11. Turner, N.J. in Enzyme Assays (ed. J.L. Reymond) 139–161, (Wiley-VCH, Weinheim,
2006).12. Jestin, J.L. & Kaminski, P.A. J. Biotechnol. 113, 85–103 (2004).13. Hillen, W. & Berens, C. Annu. Rev. Microbiol. 48, 345–369 (1994).14. Gamper, M., Hilvert, D. & Kast, P. Biochemistry 39, 14087–14094 (2000).
NATURE BIOTECHNOLOGY VOLUME 25 NUMBER 10 OCTOBER 2007 1 14 7
B R I E F C OMMUN I C AT I O N S©
2007
Nat
ure
Pub
lishi
ng G
roup
ht
tp://
ww
w.n
atur
e.co
m/n
atur
ebio
tech
nolo
gy