In vitro selection for catalytic activity with ribosome display

Article abstract of DOI:10.1021/ja025870q

We report what is, to our knowledge, the first in vitro selection for catalytic activity based on catalytic turnover by using ribosome display, a method which does not involve living cells at any step. RTEM-β-lactamase was functionally displayed on riboso

Full text of DOI:10.1021/ja025870q

Published on Web 07/17/2002
In Vitro Selection for Catalytic Activity with Ribosome Display
Patrick Amstutz,† Joelle N. Pelletier,†,‡ Armin Guggisberg,§ Lutz Jermutus,†,
Sandro Cesaro-Tadic,† Christian Zahnd,† and Andreas Plu¨ckthun*,†
Biochemisches Institut, UniVersita¨t Zu¨rich, Winterthurerstrasse 190,
CH-8057 Zu¨rich, Switzerland, and Organisch-Chemisches Institut der
UniVersita¨t Zu¨rich, Winterthurerstrasse 190, CH-8057 Zu¨rich, Switzerland
Received February 8, 2002
Abstract: We report what is, to our knowledge, the first in vitro selection for catalytic activity based on
catalytic turnover by using ribosome display, a method which does not involve living cells at any step.
RTEM-â-lactamase was functionally displayed on ribosomes as a complex with its encoding mRNA. We
designed and synthesized a mechanism-based inhibitor of â-lactamase, biotinylated ampicillin sulfone,
appropriate for selection of catalytic activity of the ribosome-displayed â-lactamase. This derivative of
ampicillin inactivated â-lactamase in a specific and irreversible manner. Under appropriate selection
conditions, active RTEM-â-lactamase was enriched relative to an inactive point mutant over 100-fold per
ribosome display selection cycle. Selection for binding, carried out with â-lactamase inhibitory protein (BLIP),
gave results similar to selection with the suicide inhibitor, indicating that ribosome display is similarly efficient
in catalytic activity and affinity selections. In the future, the capacity to select directly for enzymatic activity
using an entirely in vitro process may allow for a significant increase in the explorable sequence space
relative to existing strategies.
Naturally occurring enzymes catalyze a wide variety of
chemical reactions and are increasingly used in pharmaceutical,
industrial, and environmental applications as a result of their
high reactivities and specificities. However, the direct improve-
ment of biocatalysts remains challenging, and the yet more
ambitious goal of developing enzymes with new catalytic
functions still seems almost elusive. Although our knowledge
of structure-function relationships of enzymes has significantly
increased, rational protein design is still a difficult task,
especially for improved catalysis. The recently developed
strategy of directed evolution can be used as a complement to
rational design. In directed evolution, a protein function of
interest is evolved in the laboratory by mimicking Darwinian
evolution in multiple successive rounds of diversification (library
generation) with subsequent selection or screening (reviewed
in refs 1-3).
selection methods sample the entire library in a single experi-
mental step. Such experiments require a direct coupling of the
phenotype, which is to be selected for, and its encoding genetic
information, the genotype.
For the selection of enzymatic activities, three general
approaches can be distinguished: methods performed entirely
in vivo, those working “partially” in vitro, and those performed
completely in vitro.1 In the in vivo approach, a genetic library
encoding enzyme variants is transformed into cells where the
variants are expressed and selection takes place. Because
selection protocols are generally based on a growth advantage,
e.g. complementation of an auxotrophy or resistance to a
cytotoxic compound,5-8 in vivo selection of catalysis is limited
to those activities giving rise to a growth advantage. Moreover,
microbial genomes have evolved to deal with environmental
selection pressure. The expression host can therefore be surpris-
ingly “creative” in escaping selection pressure, such as by higher
expression of a poor catalyst or the use of alternative metabolic
pathways, completely by-passing the enzymatic activity that is
the actual target of selection.
Screening, which involves the analysis of single protein
variants, can be automated for high-throughput protocols but
remains laborious and time-consuming, therefore limiting the
size of libraries which can be handled.4 As opposed to screening,
Partially in vitro methods can offer an alternative to in vivo
methods. In these methods, the library is also introduced into
cells, resulting in display of the protein of interest, usually on
the surface of phage,9 bacteria, or yeast.10 Selection then occurs
* Corresponding author. Tel. (+41-1) 635 55 70. Fax: (+41-1) 635 57
12. E-mail: plueckthun@biocfebs.unizh.ch.
† Biochemisches Institut, Universita¨t Zu¨rich.
§ Organisch-Chemisches Institut der Universita¨t Zu¨rich.
‡ Present address: De´partement de Chimie, Universite´ de Montre´al, C.
P. 6128, Succursale Centre-ville, Montre´al, PQ, Canada.
Present address: Cambridge Antibody Technology, The Science Park,
Melbourn, Cambridgeshire SG8 6JJ, U.K.
(5) Orencia, M. C.; Yoon, J. S.; Ness, J. E.; Stemmer, W. P.; Stevens, R. C.
Nat. Struct. Biol. 2001, 8, 238-242.
(6) Altamirano, M. M.; Blackburn, J. M.; Aguayo, C.; Fersht, A. R. Nature
2000, 403, 617-622. Retraction, Nature 2002, 417, 468.
(7) Crameri, A.; Raillard, S. A.; Bermudez, E.; Stemmer, W. P. Nature 1998,
391, 288-291.
(1) Griffiths, A. D.; Tawfik, D. S. Curr. Opin. Biotechnol. 2000, 11, 338-
353.
(2) Olsen, M.; Iverson, B.; Georgiou, G. Curr. Opin. Biotechnol. 2000, 11,
331-337.
(8) MacBeath, G.; Kast, P.; Hilvert, D. Science 1998, 279, 1958-1961.
(9) Dunn, I. S. Curr. Opin. Biotechnol. 1996, 7, 547-553.
(10) Mendelsohn, A. R.; Brent, R. Science 1999, 284, 1948-1950.
(3) Soumillion, P.; Fastrez, F. Curr. Opin. Biotechnol. 2001, 12, 387-394.
(4) Joo, H.; Lin, Z.; Arnold, F. H. Nature 1999, 399, 670-673.
9
9396
J. AM. CHEM. SOC. 2002, 124, 9396-9403
10.1021/ja025870q CCC: $22.00 © 2002 American Chemical Society

In Vitro Selection of a Displayed Enzyme
A R T I C L E S
in vitro, that is, outside the cell, allowing fine-tuning of the
selection pressure and selection conditions.
Both entirely in vivo or partially in vitro techniques require
a transformation step, limiting the applicable library size to
cellular transformation efficiencies. Typically, transformation
efficiencies of 107-108 cells/µg DNA for yeast and 109-1010
cells/µg DNA for Escherichia coli (E. coli)11,12 are achievable.
A considerable amount of work arises from the need of ligating
and transforming such a library after each round of randomiza-
tion. Furthermore, cytotoxic proteins cannot be displayed at all.
Technologies working completely in vitro can overcome these
limitations, because no living cells are involved at any step.
Two approaches that are carried out completely in vitro can
be distinguished. In one, a compartmentalization of individual
variants is achieved in water-in-oil emulsions.13 Since the in
situ detection of fluorescent products and direct optical sorting
of such droplets, harboring the catalytic proteins, have not yet
been reported, the physical link between genotype and pheno-
type is still required, and the emulsions have to be broken up
before an affinity selection or sorting can be performed. The
other approach makes use of the concomitant presence of mRNA
and nascent protein at the ribosome during an in vitro translation
for coupling of genotype and phenotype.14 Here, the most
prominent techniques are ribosome display;15,16 mRNA display,
also termed “in vitro virus”17 or “mRNA-protein fusion”;18 and
“ribosome-inactivation display system”.19 Below, we will
describe the first application of ribosome display to the selection
of catalytic activity in an effort to increase the library sizes
accessible beyond those in previously published accounts with
in vivo or partially in vitro methods.
To select catalysts in vitro, catalysis must be correlated or
coupled to binding. Although reversible binding of enzymes to
transition state analogues has been used frequently in the
selection of novel catalysts,20,21 a more direct selection for
enzymatic turnover can be achieved by the use of mechanism-
based (or “suicide”) inhibitors.22 These compounds are substrate
analogues that, upon turnover, are converted into a reactive
species that binds with the enzyme in a covalent manner, thus
causing irreversible inhibition. Mechanism-based inhibitors can
thus be used for labeling active enzyme molecules. The selection
of â-lactamase displayed on phage with a biotinylated mech-
anism-based inhibitor has been reported.23-26 In these reports,
Figure 1. Principle of ribosome display selection for catalytic activity.
DNA encoding the protein of interest is transcribed and translated in vitro.
Because the mRNA carries no stop codon and translation is stopped with
high Mg2+ concentrations, stable ternary complexes of mRNA (black),
ribosome (blue) and tethered nascent protein (red) are formed. These can
be used directly for selection with a biotinylated suicide inhibitor. Excess
inhibitor is removed by gel filtration, and the labeled complexes are captured
with avidin coated magnetic beads. After washing to remove any untrapped
ribosomal complexes, the selected complexes are destroyed to elute the
mRNA. Finally, reverse transcription (RT) and PCR are used to amplify
the genetic information of the selected clones.
active enzyme was enriched relative to less active mutants and
relative to penicillin binding protein. The elution of covalently
trapped clones from the matrix to which they had been linked
is generally achieved by cleaving within a linker region, either
chemically between the suicide inhibitor and the affinity tag or
enzymatically between the displayed protein and the phage.
Even though entirely in vitro display techniques have great
potential for selection of catalytic activity, this has not previously
been reported. We believe that ribosome display is particularly
well suited to applications based on mechanism-based inhibition,
because ribosome display offers an elegant solution for elution
of the selected genetic information. Because genotype and
phenotype are coupled in a noncovalent complex at the ribosome
(Figure 1), dissociation of these complexes and therefore elution
of the genetic information is easily achieved.
In the present study, we use ribosome display for selection
of catalytic activity with a suicide inhibitor. We have chosen
â-lactamase as a model enzyme, and we displayed it on the
ribosome in complex with its encoding mRNA. We have devised
a synthetic scheme for the preparation of a biotinylated
mechanism-based inhibitor of â-lactamase. We then applied the
inhibitor to the selective enrichment of active, displayed enzyme.
We show that selection of enzymes is possible entirely in vitro,
combining the advantages of both the in vitro technology of
ribosome display and direct selection for catalytic turnover with
a suicide substrate to select for protein catalysts.
(11) Inoue, H.; Nojima, H.; Okayama, H. Gene 1990, 96, 23-28.
(12) Sidhu, S. S.; Lowman, H. B.; Cunningham, B. C.; Wells, J. A. Methods
Enzymol. 2000, 328, 333-363.
(13) Tawfik, D. S.; Griffiths, A. D. Nat. Biotechnol. 1998, 16, 652-656.
(14) Amstutz, P.; Forrer, P.; Zahnd, C.; Plu¨ckthun, A. Curr. Opin. Biotechnol.
2001, 12, 400-405.
(15) Hanes, J.; Plu¨ckthun, A. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 4937-
4942.
(16) He, M.; Taussig, M. J. Nucleic Acids Res. 1997, 25, 5132-5134.
(17) Nemoto, N.; Miyamoto-Sato, E.; Husimi, Y.; Yanagawa, H. FEBS Lett.
1997, 414, 405-408.
(18) Roberts, R. W.; Szostak, J. W. Proc. Natl. Acad. Sci. U.S.A. 1997, 94,
12297-12302.
Results and Discussion
(19) Zhou, J. M.; Fujita, S.; Warashina, M.; Baba, T.; Taira, K. J. Am. Chem.
Soc. 2002, 124, 538-543.
(20) Schultz, P. G.; Lerner, R. A. Science 1995, 269, 1835-1842.
(21) Xu, J.; Deng, Q.; Chen, J.; Houk, K. N.; Bartek, J.; Hilvert, D.; Wilson, I.
A. Science 1999, 286, 2345-2348.
In this study we assessed the potential of ribosome display
for enzyme selection based on catalytic activity and on binding
specificity. In such an in vitro selection technology, the
theoretical library size would be limited only by the number of
active ribosomes in the reaction. As a model system, we
displayed RTEM â-lactamase in a ribosomal complex with its
encoding mRNA. We describe a rapid synthesis of a biotinylated
mechanism-based â-lactamase inhibitor for activity selection.
(22) Hilvert, D. Annu. ReV. Biochem. 2000, 69, 751-793.
(23) Soumillion, P.; Jespers, L.; Bouchet, M.; Marchand-Brynaert, J.; Winter,
G.; Fastrez, J. J. Mol. Biol. 1994, 237, 415-422.
(24) Vanwetswinkel, S.; Marchand-Brynaert, J.; Fastrez, J. Bioorg. Med. Chem.
Lett. 1997, 7, 239.
(25) Vanwetswinkel, S.; Avalle, B.; Fastrez, J. J. Mol. Biol. 2000, 295, 527-
540.
(26) Avalle, B.; Vanwetswinkel, S.; Fastrez, J. Bioorg. Med. Chem. Lett. 1997,
7, 479-484.
9
J. AM. CHEM. SOC. VOL. 124, NO. 32, 2002 9397

A R T I C L E S
Amstutz et al.
Scheme 1. Synthesis of Biotinylated Ampicillin Sulfone, a
Mechanism-Based Inhibitor for Selection of Active Site Serine
â-Lactamases.a
a (I) H2O, CH2Cl2 (4:1); NaHCO3; MOZ-ON. (II) H2O, KMnO4. (III)
CH2Cl2, TFA (10:1). (IV) H2O, EZ-Link Sulfo-NHS-LC-LC-Biotin (Pierce)
zyloxycarbonyloxyimino-2-phenyl-acetonitrile (MOZ-ON), but
not with benzylchloroformate or paranitrobenzylchloroformate,
yielded intact ampicillin sulfone upon deprotecting. In a fourth
step, the biotin moiety was added to yield the biotinylated
suicide inhibitor (Scheme 1). Our synthesis is simpler than that
of a similar penicillanic acid sulfone, because we do not need
a disulfide bond in the linker region between the penicillin and
the biotin moiety.31 This disulfide bridge was required in the
context of phage display selection of â-lactamase to release the
covalently trapped phages by cleaving with DTT. In contrast,
ribosome display does not require elution of the covalently
trapped and displayed protein, because the genetic information
coding for the displayed protein can easily be released by
chelating the Mg2+ ions so as to disrupt the ribosomal complex.
Mechanism-Based Enzyme Inhibition. The ability of ampi-
cillin sulfone to irreversibly inhibit â-lactamase was tested by
steady-state kinetic assays. The inhibition pattern of the RTEM
â-lactamase from E. coli, obtained directly from in vitro
translation, was compared to that of the â-lactamase from
Enterobacter cloacae. Both enzymes behaved identically. The
enzyme inhibition profile as a function of time showed a
dramatic fall of activity within the first 20 min, followed by a
slower activity decay (Figure 3). These biphasic inhibition
kinetics are typical for mechanism-based inhibitors of â-lacta-
mases, such as clavulanic acid, 6-(methoxymethylene)penicil-
lanic acid and 6-â-((carboxy)methylsulfonamido)penicillanic
acid sulfone.27,32 The kinetics have been interpreted as being
due to a first inhibition phase where the enzyme is transiently
inhibited, while in a slower second phase, the compound inhibits
the enzyme irreversibly (Figure 2). Since not every turnover
results in enzyme deactivation (Figure 2), the inhibitor-to-
enzyme ratio needed for complete inhibition is a measure for
inhibitor potency. From various inhibition experiments, we
determined that this ratio is 8 × 104 for ampicillin sulfone, which
is about an order of magnitude higher than that for the previously
characterized â-lactamase suicide inhibitor sulbactam.28,33 This
reduced inhibition efficiency is most likely due to the fact that
Figure 2. Proposed interaction of ampicillin sulfone with the active site
serine â-lactamase.27,28 The active site serine attacks the carbonyl group of
the lactam ring, and before a water molecule can attack, the sulfone in
position 1 acts as a leaving group from carbon 5 and evokes the formation
of an acyl-enzyme intermediate (a). This acyl-enzyme intermediate is in
an equilibrium with a more stable tautomer (b), resulting in transient
inhibition. A normal deacylation, triggered by water, can also occur, and
no inhibition is seen (c). Finally, transamination with a lysine in the active
site can occur (d), which results in an irreversibly inactivated enzyme, bound
covalently to the inhibitor and the biotin moiety.
We demonstrate that the enzyme folds on the ribosome to its
correct three-dimensional structure and is active there. Further-
more, we demonstrate selection based on activity as well as
selection for affinity with this system.
Rapid Synthesis of a Bifunctional Mechanism-Based
Inhibitor: Biotinylated Ampicillin Sulfone. Penicillin ana-
logues with an electron withdrawing group in position 1 (Figure
2) are known to lead to opening of the thiazolidine ring after
nucleophilic attack of the â-lactam ring by the active-site serine
of type-A â-lactamase, forming an acyl-enzyme intermedi-
ate.27,28 The attack of a second active-site nucleophilic side chain
on this reactive intermediate results in a covalently inhibited
enzyme (Figure 2). Clavulanic acid and penicillanic acid sulfone
are well-characterized mechanism-based inhibitors of type-A
â-lactamase that follow this mechanism.29,30 By tethering biotin
to such an inhibitor, an affinity handle is generated that links
biotin exclusively to active enzyme molecules as a consequence
of enzymatic turnover. The resulting irreversible trapping allows
for the possibility to select directly for catalytic activity.23
We developed a four-step synthesis for biotinylated ampicillin
sulfone, a â-lactamase suicide inhibitor, with 64% overall yield.
The synthesis of ampicillin sulfone involved only protection of
the amino group of ampicillin, oxidation of the sulfur in the
thiazolidine ring to the sulfone, followed by deprotection
(Scheme 1). Since the compound is highly sensitive to hydrolysis
at the lactam ring, the choice of the protecting group proved to
be of utmost importance. Only protection with 4-methoxyben-
(27) Fisher, J.; Charnas, R. L.; Knowles, J. R. Biochemistry 1978, 17, 2180-
2184.
(31) Vanwetswinkel, S.; Fastrez, J.; Marchand-Brynaert, J. J. Antibiot. 1994,
47, 1041-1051.
(32) Brenner, D. G.; Knowles, J. R. Biochemistry 1981, 20, 3680-3687.
(33) Imtiaz, U.; Billings, E. M.; Knox, J. R.; Mobashery, S. Biochemistry 1994,
33, 5728-5738.
(28) Fisher, J.; Charnas, R. L.; Bradley, S. M.; Knowles, J. R. Biochemistry
1981, 20, 2726-2731.
(29) Brenner, D. G.; Knowles, J. R. Biochemistry 1984, 23, 5833-5839.
(30) Charnas, R. L.; Knowles, J. R. Biochemistry 1981, 20, 3214-3219.
9
9398 J. AM. CHEM. SOC. VOL. 124, NO. 32, 2002

In Vitro Selection of a Displayed Enzyme
A R T I C L E S
Figure 4. Analysis of nonspecific labeling of proteins by ampicillin sulfone.
A protein mix, here E. coli S30 extract, was incubated with biotinylated
ampicillin sulfone (0.1 mM). The proteins were separated from excess
inhibitor by gel filtration, followed by SDS-PAGE and Western blotting
for detection of biotin with avidin-alkaline phosphatase conjugate. The
arrow indicates the signal obtained from pyruvate carboxylase, a biotinylated
E. coli protein in the extract, used as internal standard. Lanes 1-5 and 7
show different labeling conditions and incubation times. In lane 6, the protein
mix was preincubated with 1 mM ampicillin sulfone at room temperature
for 1 h. The nonspecific labeling is most likely caused by the attack of
nucleophilic surface residues on the lactam ring and has been reported also
for ampicillin.35
Figure 3. Time course of RTEM-â-lactamase inhibition by ampicillin
sulfone. â-lactamase (10-9-10-10 M, from in vitro translation) was
incubated with ampicillin sulfone (1 mM) for different time periods at room
temperature, and the remaining activity was measured in nitrocefin assays.
The data, given as activity relative to that at time ) 0, were fitted as double
exponential decay. Similar inhibition curves have been reported for other
â-lactamase suicide inhibitors.27,31,34 The data for this plot were obtained
from three independent experiments.
ampicillin sulfone is derived from ampicillin, a very good
substrate of â-lactamase that is rapidly hydrolyzed.28 The
covalent nature of the inhibition was confirmed by dialysis of
the inhibited enzyme. Even upon extensive dialysis, no recovery
of activity could be detected, whereas the noninhibited enzyme
remained fully active (data not shown). The biotinylated form
of the inhibitor gave results comparable to ampicillin sulfone
itself in all assays (data not shown). These results confirm that
the biotinylated form of ampicillin sulfone acts as a bona fide
mechanism-based inhibitor of â-lactamase and could be used
for selection of â-lactamase activity.
Prevention of Nonspecific Protein Labeling by Ampicillin
Sulfone. Besides its reactivity, we also investigated the specific-
ity of the suicide inhibitor. Although hydrolysis of the lactam
ring, leading to formation of the reactive species, will be greatly
accelerated at the active site of the enzyme, the labile nature of
the lactam ring may allow for labeling outside of the active
site. In addition, ampicillin is known to react nonspecifically
with ꢀ-amino groups of proteins in a covalent manner, a cause
for ampicillin allergies.35 Nonspecific labeling could lead to
significant background in selection rounds by the suicide
inhibitor reacting with inactive enzyme displayed on the
ribosome or with ribosomal proteins. We thus investigated
possible nonspecific binding in order to minimize it. Aliquots
of E. coli S30 extract were incubated with biotinylated ampicillin
sulfone under various conditions and for different incubation
times. The proteins were then separated from excess suicide
inhibitor by gel filtration. The resulting samples were analyzed
by Western blotting. The results showed significant labeling of
proteins other than â-lactamase (Figure 4). Nonspecific labeling
also occurred when incubating the biotinylated suicide inhibitor
with purified control proteins, such as citrate synthase, GFP
(green fluorescent protein), anti-GCN4-single-chain Fv, and
M13 helper phage (data not shown). The signal intensity of
nonspecific labeling increased with time (lanes 1-4), temper-
ature (lane 5), and higher pH (lane 7) and could be competitively
inhibited by preincubation with ampicillin sulfone (lane 6)
(Figure 4). These observations are in agreement with the
published results on â-lactamase selections with phage display
using similar mechanism-based inhibitors.23-26 In all the phage
display experiments, the labeling time was short (<30 min),
the pH was kept below 6, and there was at least 1% BSA
present, presumably to reduce nonspecific labeling. Under those
conditions, the authors reported the selection of highly active
â-lactamase over less active mutants or penicillin binding
protein, which binds but does not hydrolyze the â-lactam ring.
The broadly based nonspecific labeling of proteins that are
not linked to their genotype, as illustrated above, is not likely
to be detrimental to the selection of displayed enzyme. However,
it suggests that nonspecific labeling of inactive displayed enzyme
and of ribosomal proteins may also occur under certain
conditions. To favor specific labeling in our in vitro system,
the selection for enzymatic activity of â-lactamase was carried
out at pH 6.0 and at 4 °C for short times (10 min to 1 h) (10
min in Figure 4, lane 1). These conditions are also compatible
with those required for ribosome display.
Ribosome-Displayed â-Lactamase is Active. In ribosome
display, selection is performed with a ternary complex of
mRNA, ribosome and displayed protein (Figure 1). For this
strategy to be successful, the protein must fold to its native three-
dimensional structure before the polypeptide is released from
the ribosome. Furthermore, the ribosome-bound protein must
be active. To allow the displayed protein of interest to exit the
ribosomal peptide channel and fold into its active conformation,
the encoding gene was fused to a 171-amino acid-long C-
terminal fusion partner TolA of E. coli serving as a tether. To
demonstrate functionality of â-lactamase in the ribosomal
complex, the complexes were separated by gel filtration from
enzyme released during translation, under conditions securing
complex stability (see Experimental Procedures). Usually such
protein-ribosome complexes are purified by sucrose gradient
(34) Brenner, D. G.; Knowles, J. R. Biochemistry 1984, 23, 5839-5846.
(35) Schneider, C. H.; De Weck, A. L. Nature 1965, 208, 57-59.
9
J. AM. CHEM. SOC. VOL. 124, NO. 32, 2002 9399

A R T I C L E S
Amstutz et al.
Table 1. Activity of Ribosome-Displayed â-Lactamase
sum of free and
ribosome-displayed
â-lactamasea
ribosome-displayed
ribosome-bound
â-lactamaseb
â-lactamase
background activity
in complex fractiond
c
after release
rel enzyme activitye
100 ( 4.8
50 ( 5.1
54 ( 7.0
17 ( 5.8
a Total activity measured after in vitro translation. b After gel filtration (activity of ribosomal complex fraction). c Complexes destroyed by RNAse A
treatment after gel filtration (activity of ribosomal complex fraction). d Complexes destroyed by RNAse A treatment prior to gel filtration (activity of ribosomal
complex fraction). e Data from three independent experiment measured in triplicates.
centrifugation.36-38 The use of small gel filtration columns is a
rapid and gentle alternative and allows the parallel processing
of a large number of samples. Thus, we were able to quantify
the amount of active â-lactamase in ribosomal complexes (Table
1). An aliquot of this fraction was treated with RNase A,
destroying the ribosomal complex, thus releasing the enzyme.
No significant increase in activity was measured, indicating that
ribosome-bound â-lactamase is fully active, and no additional
active enzyme molecules are obtained by releasing them from
Figure 5. â-Lactamase selection for activity and affinity. The DNA yield
after one round of ribosome display by affinity with gpHDBLIP (BLIP),
the ribosome (Table 1). When the RNase A treatment was
carried out prior to the gel filtration procedure, however, some
background activity was detected in the complex fraction (Table
1), either from insufficient complex destruction or incomplete
separation by gel filtration. These results confirm that the
tethered â-lactamase can fold to its correct three-dimensional
structure and can catalyze substrate turnover while still bound
to the ribosome. Comparing the total activity after in vitro
translation of the ribosome display construct prior to gel filtration
to the amount of ribosome-displayed activity, we estimated the
percentage of â-lactamase in ribosomal complexes to be ∼50%
of total â-lactamase produced. We presume that the other half
is released from the ribosome by hydrolysis of RNA or
proteolysis of the tether. The possibility that a fraction of the
produced protein is bound to the ribosome in an inactive state
and does not refold upon complex destruction remains, but that
seems unlikely, because â-lactamase is known to fold ef-
ficiently.39 Although co-translational folding and activity on the
ribosome have been demonstrated for firefly luciferase in
different experiments,36-38 we show here that â-lactamase also
folds and acquires activity on the ribosome, thereby fulfilling
the major requirements for ribosome display selections.
Affinity Selection and Selection for Activity Are Equally
Efficient. We extended our analysis of â-lactamase functionality
in the ribosome display format by performing selection rounds
based on affinity and catalytic activity. The â-lactamase
inhibitory protein (BLIP) binds near the active site of â-lacta-
mase with high affinity (Kd ) 0.6 nM)40 and had been used in
affinity selection with phage display.41,42 In contrast to His-
tagged BLIP, gpHDBLIP, a C-terminal fusion of BLIP to the
His-tagged protein D (gpHD),43 could be expressed at high
levels in inclusion bodies and could be refolded efficiently. The
fusion partner, gpHD, not only improved the expression of BLIP
by biotinylated ampicillin sulfone (suicide inhibitor) or with beads alone
(background) were compared. Only active â-lactamase was used, resulting
in PCR amplification of a band at 834 bp. Lane M indicates DNA molecular
weight marker. Background binding to the beads is very weak, and the
efficiency of selection for affinity and activity are comparable.
very significantly but also provides a His tag, which we used
for purification of the fusion protein and its immobilization for
affinity selection. We performed selection rounds with active
ribosome-displayed â-lactamase on either immobilized gpHD-
BLIP or the immobilized biotinylated suicide inhibitor. The
experiments were carried out as described44 with some modi-
fications to adapt the system for activity selection, notably lower
pH (pH 6.0) and the presence of 1% BSA, for reasons discussed
above. After in vitro translation and incubation on ice, aliquots
of ribosomal complexes displaying â-lactamase were incubated
for 1 h with gpHDBLIP or biotinylated ampicillin sulfone.
Excess ampicillin sulfone or unbound gpHDBLIP was removed
by gel filtration, and the complexes selected with the biotinylated
suicide inhibitor were captured with avidin-agarose, whereas
the ones binding to gpHDBLIP were immobilized via a tetra-
His antibody which was captured with protein G-PLUS agarose.
Extensive washing removed unbound complexes. The mRNA
was eluted, and after RT and PCR, the yields were quantified
according to band intensities on an agarose gel (Figure 5). We
observed identical results with affinity- and activity-based
selections, indicating that the ribosome-displayed â-lactamase
is active in terms of both binding and enzymatic activity and
that selection for activity and affinity in this system are equally
efficient.
Selection of Active â-Lactamase over an Inactive Mutant
Completely In Vitro. To assess the efficiency and specificity
of selection for activity with ribosome display, the enrichment
of active â-lactamase over an inactive point mutant was tested.
The serine residue at position 70 of the active enzyme, whose
hydroxyl group carries out the nucleophilic attack of the lactam
ring (Figure 2), was replaced by alanine, yielding an inactive
mutant enzyme with otherwise unimpaired structure.45-48 Ri-
(36) Kolb, V. A.; Makeyev, E. V.; Spirin, A. S. J. Biol. Chem. 2000, 275,
16597-16601.
(37) Makeyev, E. V.; Kolb, V. A.; Spirin, A. S. FEBS Lett. 1996, 378, 166-
170.
(38) Frydman, J.; Erdjument-Bromage, H.; Tempst, P.; Hartl, F. U. Nat. Struct.
Biol. 1999, 6, 697-705.
(39) Vanhove, M.; Guillaume, G.; Ledent, P.; Richards, J. H.; Pain, R. H.; Frere,
J. M. Biochem. J. 1997, 321, 413-417.
(44) Hanes, J.; Jermutus, L.; Plu¨ckthun, A. Methods Enzymol. 2000, 328, 404-
(40) Albeck, S.; Schreiber, G. Biochemistry 1999, 38, 11-21.
(41) Huang, W.; Petrosino, J.; Palzkill, T. Antimicrob. Agents Chemother. 1998,
42, 2893-2897.
430.
(45) Sigal, I. S.; DeGrado, W. F.; Thomas, B. J.; Petteway, S. R., Jr. J. Biol.
Chem. 1984, 259, 5327-5332.
(42) Huang, W.; Zhang, Z.; Palzkill, T. J. Biol. Chem. 2000, 275, 14964-14968.
(43) Forrer, P.; Jaussi, R. Gene 1998, 224, 45-52.
(46) Dalbadie-McFarland, G.; Neitzel, J. J.; Richards, J. H. Biochemistry 1986,
25, 332-338.
9
9400 J. AM. CHEM. SOC. VOL. 124, NO. 32, 2002

In Vitro Selection of a Displayed Enzyme
A R T I C L E S
sulfone. Here, no enrichment of active enzyme was detected
(Figure 6, lane C2). The possibility of different translation
efficiencies of mutant and active construct was ruled out by
performing radioactive translations and comparing protein yields
after SDS-PAGE and autoradiography. The band intensities
were identical for both constructs (data not shown). It follows
that active â-lactamase is efficiently enriched with ribosome
display in vitro due to its catalytic activity.
Figure 6. Restriction analysis for detection of enrichment of active
â-lactamase with ribosome display. The results of two independent
experiments are shown, where E1 represents a usual result and E2, the most
successful result obtained. In each experiment, an MscI digest of the DNA
pool was carried out after one round of enrichment for enzymatic activity
of â-lactamase with ribosome display. The round was performed with a
ratio of inactive mutant to active â-lactamase of 10:1. After RT, the
â-lactamase gene was amplified (834 bp) by PCR and digested with MscI,
resulting in cleavage of only the DNA encoding the inactive mutant (662
bp + 172 bp). Lane M indicates DNA molecular weight marker. Lane C1
shows the input mRNA directly reverse-transcribed and amplified by PCR
without undergoing a display cycle. C2 shows the background of a standard
enrichment round without biotinylated ampicillin sulfone present. E1 and
E2 represent two independent enrichment rounds performed with biotiny-
lated ampicillin sulfone. The enrichment factors were estimated to be 110
for E1 and 1400 for E2.
Conclusions
In conclusion, we have shown for the first time that ribosome
display may be used for selection of enzymes based on their
catalytic activity. It follows that this strategy is ideally suited
to performing directed evolution of enzymes. This novel
potential of ribosome display was demonstrated by selecting
â-lactamase based on its enzymatic activity. We show that the
enzyme â-lactamase folds correctly on the ribosome and is
active in the ternary complex, thus fulfilling the major require-
ments for ribosome display. To select for enzymatic activity of
â-lactamase with ribosome display, we synthesized ampicillin
sulfone, a mechanism-based inhibitor of â-lactamase, which can
be derivatized with biotin, yielding a bifunctional activity label.
This label was applied to ribosome display experiments, in
which active â-lactamase could be enriched over an inactive
mutant >100-fold/selection round. The efficiency of selection
for activity was comparable to that of affinity selection, which
was carried out with immobilized BLIP.
Taken together, we demonstrate that ribosome display, a
technique that combines the advantages of very rapid selection
rounds with the use of large libraries, known to be capable of
evolving high affinity binders, may also be applied to the
selection of catalytic proteins. Therefore, we believe this method
will have great potential for directed evolution of enzymes and
the selection of new catalytic proteins.
bosome display was performed with RNA encoding active and
mutant â-lactamase mixed in a ratio of 1:10. After translation,
biotinylated ampicillin sulfone was added, and after 30 min,
excess suicide inhibitor was removed by gel filtration. The
biotin-labeled complexes were captured with streptavidin mag-
netic beads, and the mRNA was eluted and amplified by RT
and PCR. Five completely independent selection experiments
were performed. Enrichment of the active enzyme was quanti-
fied by two independent methods: first, an MscI restriction site
had been introduced along with the S70A mutation, allowing
restriction fragment analysis to distinguish the PCR products
encoding active and inactive enzyme. Figure 6 illustrates a
representative enrichment (lane E1, factor of 110) as well as
the best enrichment obtained (lane E2, factor of 1400). To
independently determine the magnitude of the enrichment factor,
gene pools after one round of ribosome display were sequenced
for the 2 experiments illustrated in Figure 6, and the signal
intensities of the respective bases were analyzed, giving
enrichment factors of 110 and 250 (data not shown). The fact
that the enrichment factor varied within 1 order of magnitude
in different selection experiments is most likely due to the small
amount of inhibitor used, giving a low yield difficult to quantify,
and the inhibitor concentration was not increased to prevent
unspecific labeling.
In summary, all experiments clearly showed enrichment of
active enzyme over inactive mutant. We concluded that after
only one round of mechanism-based selection with ribosome
display, an enrichment >100-fold was achieved. To demonstrate
that the enrichment was, indeed, due to specific labeling of
active â-lactamases, several control experiments were per-
formed. To ascertain that enrichment was not an artifact of RT
and PCR, input mRNA was directly applied to this procedure,
showing no enrichment of the active species (Figure 6, lane
C1). To rule out mRNA stability during translation as enrich-
ment source, a second control was performed. A normal
ribosome display round was performed using the same ratio of
input mRNA but in the absence of biotinylated ampicillin
Experimental Section
Synthesis of Biotinylated Ampicillin Sulfone. Ampicillin was
obtained from Roche Diagnostics, and 4-methoxybenzyloxycarbony-
loxyimino-2-phenyl-acetonitrile (MOZ-ON) and other chemicals were
from Fluka. IR spectra were measured on a Perkin-Elmer 297
spectrometer; NMR spectra, on a Bruker ARX-300 (1H) and a Bruker
ARX-75 (13C) spectrometer; and mass spectra, on a Perkin-Elmer Sciex
API III+ ESI-MS.
Ampicillin (1.5 g, 4.02 mmol) was dissolved in 4 mL of H2O. One
equiv of MOZ-ON (1.25 g, 4.02 mmol) in 4 mL dioxane was added.49
The pH of the reaction mixture was adjusted to 7.8. After stirring for
6 h, the mixture was treated with 8 mL of H2O (pH 7.8) and extracted
with EtOAc (3×, 100 mL). The pH was adjusted to 3.0 with 10% H3-
PO4, and the aqueous phase was extracted with EtOAc (5×, 100 mL).
The combined EtOAc extractions were dried with NaSO4 and evapo-
rated in vacuo. MOZ-ampicillin (1.58 g, 3.08 mmol, 76.5% yield) was
collected as an egg white powder: IR (CHCl3 [cm-1]) 3405, 2960,
1785, 1725, 1695, 1612, 1515, 1495; 1H NMR (300 MHz, DMSO): δ
1.42, 1.55, 3.74 (3× 3H, 3s), 4.23 (1H, s), 4.98-4.99 (2H, m), 5.38-
5.41 (1H, m), 5.50-5.54 (2H, m), 6.91 (2H, d like m, J ) 6.9), 7.24-
7.35 (5H, m), 7.44-7.53 (2H, m), 7.88 (1H, br d, J ) 7.9), 8.99 (1H,
br d, J ) 7.7); 13C NMR (DMSO): δ 26.4, 30.2, 54.9 (3q), 57.4, 58.1
(2d), 63.5 (s), 65.4 (t), 67.1, 70.2 (2d), 113.6 (2C, d), 125.4 (d), 127.1,
128.0 (2× 2C, 2d), 128.6 (1C, s), 129.5 (d), 137.7, 155.6, 158.9, 168.9,
170.4, 173.2 (s); MS: 514.2 m/z.
(47) Jacob, F.; Joris, B.; Frere, J. M. Biochem. J. 1991, 277, 647-652.
(48) Mazzella, L. J.; Pazhanisamy, S.; Pratt, R. F. Biochem. J. 1991, 274, 855-
859.
(49) Chen, S. T.; Wang, K. T. Synthesis 1989, 36-37.
9
J. AM. CHEM. SOC. VOL. 124, NO. 32, 2002 9401

A R T I C L E S
Amstutz et al.
MOZ-ampicillin (513 mg, 1.00 mmol) was dissolved in 3.6 mL of
cold H2O, and the pH was adjusted to 7.0. KMnO4 (190 mg, 1.2 mmol)
was dissolved in 5 mL H2O and treated with 65 µL 10% H3PO4. The
KMnO4 solution was added dropwise to the first solution on ice, while
maintaining the pH between 6.8 and 7.3 with 10% H3PO4 and 10%
NaOH. After 40 min, the reaction mixture was brought to pH 8.0 and
filtered over Celite, and the filtrate was acidified to pH 3.0 with 10%
H3PO4.50 The product was extracted with EtOAc (5×, 100 mL, dried
with NaSO4, and concentrated in vacuo to afford 495 mg (91% yield)
of MOZ-ampicillin sulfone as an egg white powder: IR (CHCl3 [cm-1])
3410, 2960, 1810, 1730, 1700, 1612, 1515, 1495, 1328, 1118; 1H NMR
(DMSO): δ 1.33, 1.46, 3.74 (3 × 3H, 3s), 4.36 (1H, s), 4.97 (2H, s),
5.31 (1H, d, J ) 4.5), 5.51 (1H, d, J ) 8.3), 5.90 (1H, dd, J ) 4.4,
8.8), 6.90 (2H, d, J ) 8.6), 7.28-7.31 (5H, m), 7.35-7.50 (2H, m),
7.97 (1H, br d, J ) 8.5), 8.65 (1H, br d, J ) 8.8); 13C NMR (DMSO):
δ 17.4, 19.8, 55.1 (3q), 56.8, 59.1, 63.7 (3d), 64.6 (s), 65.4 (d), 67.3
(t), 113.8, 127.0 (2× 2C, 2d), 128.4 (s), 128.7 (d), 129.1, 129.9 (2×
2C, 2d), 136.0, 156.0, 159.5, 168.9, 170.6, 173.6 (6s); MS: 546.2 m/z.
A solution of MOZ-ampicillin sulfone (220 mg, 0.40 mmol) in 2
mL of CH2Cl2 and 0.2 mL of TFA was stirred at room temperature for
30 min.49 The solvent was removed in vacuo. The product was triturated
with ether that was removed as the supernatant after centrifugation.
This procedure was repeated twice, and the product was dried in vacuo.
Ampicillin sulfone (150 mg, 0.39 mmol, 97% yield) was isolated as a
pale yellow powder: IR (KBr [cm-1]): 3390, 2980, 1802, 1674, 1525,
sulfone on ice. Aliquots were taken after various time points and excess
inhibitor was removed by gel filtration (NAP-500 column, Pharmacia,
according to the manufacturer’s instructions). Alternatively, one aliquot
was incubated at 25 °C for 2 h while another was kept on ice at a pH
8 for 2 h. For inhibition studies, one aliquot was preincubated with 1
mM ampicillin sulfone for 1 h at room temperature prior to incubation
with biotinylated ampicillin sulfone (as described above). The proteins
were separated by SDS-PAGE (12%) and blotted onto a membrane
(Immunoblot, Millipore). The membrane was blocked with milk, and
the presence of the biotinylated suicide inhibitor was detected with an
avidin-alkaline phosphatase conjugate (Pierce).
DNA Constructs for the Model System. The â-lactamase gene
encoding the double cysteine to alanine mutant was PCR amplified
51
from the plasmid pAP5•2C•2A
with the primers SDA-BLA
5′-(AGACCACAACGGTTTCCCTCTAGAAATAATTTTGTTTAA-
CTTTAAGAAGGAGATA TATCCATGGACTACAAAACCAGCCA-
GAAACGCTGGTGAAAGT), which codes for the Shine Dalgarno box
and the beginning of the â-lactamase gene, and BLArev 5′-(ACA-
CAGGCCCCCGAGGCCCAATGCTTAATCAGTGA), annealing at
the end of the â-lactamase gene and introducing a unique SfiI restriction
site, while removing the stop codon. To create an inactive â-lactamase
to use as a negative control for activity, the active site serine was
mutated to alanine (S70A) by PCR mutagenesis. A unique MscI
restriction site was added at the site of the mutation to allow for
discrimination of the genes encoding active and inactive enzyme. In
parallel, a 171 amino acid portion of the tolA gene (residues 131 to
302) was amplified by PCR from chromosomal E. coli DNA in order
to serve as a C-terminal tether for the displayed â-lactamase. The
primers used were tolAforSFI 5′-(TATATGGCCTCGGGGGCCGAAT-
TCCAGAAGCAAGCTGAAG), introducing an SfiI-site, and tolAmrev
5′-(CCGCACACCAGTAAGGTGTGCGGTTAGCTCACCGAAAA-
TATCATC), which introduced a RNA-stabilizing 3′-loop. The inactive
S70A mutant and the active â-lactamase were fused to the tolA spacer
by SfiI digestion-ligation. The resulting constructs were further
amplified by PCR with the primers tolAmrev and T7B 5′-(ATAC-
GAAATTAATACGACTCACTATAGGGAGACCACAACGG), which
introduced an RNA stabilizing 5′-loop and the T7 promoter for in vitro
transcription. Both constructs were verified by DNA sequencing.
1
1458, 1433, 1376, 1323, 1116; H NMR (DMSO): δ 1.32, 1.42 (2×
3H, 2s); 4.33 (1 H, s); 5.31-5.29 (2H, m); 5.95 (1 H, dd, J ) 8.4,
4.5,); 7.30-7.54 (5H, m); 9.12 (1H, d like m, J ) 8.5); 13C NMR
(DMSO): δ 17.0, 19.1 (2q); 54.7, 55.6, 63.1 (3d); 63.7 (s), 64.6 (d),
127.4, 128.7 (2× 2C, 2d), 129.1 (d), 132.9 167.7, 168.0, 173.1 (4s),
MS: 382.2 m/z.
EZ-LinkSulfo-NHS-LC-LC-Biotin (7.6 mg, 1.14 × 10-2 mmol,
Pierce) was dissolved in 7 mL borate buffer pH 7.8. Ampicillin sulfone
(7.1 mg, 2.3 × 10-2 mmol) was added. This solution was sonicated
and stirred for 1 h at room temperature. Tris(hydroxymethyl)ami-
nomethane base (2 mg, 1.7 × 10-2 mmol) was added, stirred for an
additional 10 min, and lyophilized to afford 71.5 mg of a white powder
(as a complex with borate) from which 30 mg was dissolved in MeOH
and applied to a silica gel column (1 g, in a Pasteur pipet). The product
was eluted with 10 mL of MeOH, which was removed in vacuo. The
product was obtained quantitatively as a white powder from ampicillin
sulfone. MS: 834.4 m/z.
Inhibition Studies. â-Lactamase (from E. cloacae (Fluka) or E. coli
R-TEM â-lactamase from in vitro translation) in concentrations from
10-9 to 10-10 M (as determined by activity measurements,51 and
assuming that the specific activity of the in vitro translated â-lactamase
is similar to that of the bacterially expressed enzyme) was incubated
at room temperature in 50 mM potassium phosphate buffer (pH 6.0,
5% DMSO), with or without 1 mM ampicillin sulfone. DMSO is
required to dissolve ampicillin sulfone. Aliquots of 500 µL were taken
after 10, 20, 30, 45, and 90 min and were mixed with 500 µL of 0.4
mM nitrocefin (Becton Dickinson) in 50 mM potassium phosphate
buffer (pH 6.0, 1% DMSO). Reaction kinetics were followed at OD486
during 2 min at room temperature with a Perkin-Elmer Lambda 20
spectrophotometer.51 To show the covalent nature of the inhibition, a
sample (500 µL) of inhibited and, as a control, not inhibited enzyme
were dialyzed against 500 mL potassium phosphate buffer (50 mM,
pH 7) in “Slide-A-Lycer 10K” devices (molecular weight cut-off, 10
kD, 0.5-3 mL sample volume) (Pierce) for 2 h before activity was
measured.
Separation of Ternary Complexes from Free â-Lactamase by
Gel Filtration. For ribosome display, the in vitro transcription and in
vitro translation of the constructs were carried out essentially as
previously described.44 Following a 10-min translation in 110 µL, the
reaction was stopped by 4-fold dilution in ice-cold wash buffer (WB;
50 mM potassium phosphate buffer, pH 6.0; 150 mM NaCl; 50 mM
MgCl2). A 250-µL aliquot was treated with 2.5 µL RNase A (Qiagen,
10 mg/mL in WB) for 180 min at 4 °C to release ribosome-bound
â-lactamase, while another aliquot was treated with WB only. An
aliquot (200 µL) of each solution was applied to a 1-mL gel filtration
column (CL-4B, Pharmacia) that had been equilibrated with 10-20
mL of ice-cold WB. Fractions of 100 µL were collected, and the enzyme
activity was assayed in a nitrocefin assay (200 µM nitrocefin in WB
buffer, pH 6.0, 1% DMSO). In addition, an aliquot of diluted translation
mix was treated with RNaseA for 1 h at room temperature without
subsequent gel filtration. The â-lactamase activity of this sample was
determined in order to quantify the total activity, as free enzyme,
resulting from translation.
Enrichment for Catalytic Activity with Ribosome Display. RNA
coding for active as well as inactive S70A mutant â-lactamase was
mixed in a ratio of 1 to 10 (10 µg total of RNA in 10 µL water) and
translated in 110 µL for 10 min.44 Translation was stopped by 4-fold
dilution into ice-cold WB with 1% BSA (bovine serum albumin,
Sigma). Biotinylated ampicillin sulfone (stock solution, 0.01 M in
DMSO) was added to a final concentration of 0.01 mM. A control
sample contained the same final concentration of DMSO but no
biotinylated ampicillin sulfone. After 30 min at 4 °C, the reaction was
stopped by removal of the excess biotinylated ampicillin sulfone by
Western Blot Analysis. E. coli S30-extract (as used for ribosome
display 44) was diluted 2-fold in 50 mM potassium phosphate buffer
(pH 6.0, 1% DMSO) and incubated with 0.1 mM biotinylated ampicillin
(50) Johnson, D. A.; Panetta, C. A.; Cooper, D. E. J. Antibiot. 1963, 28, 1927-
1928.
(51) Laminet, A. A.; Plu¨ckthun, A. EMBO J. 1989, 8, 1469-1477.
9
9402 J. AM. CHEM. SOC. VOL. 124, NO. 32, 2002

In Vitro Selection of a Displayed Enzyme
A R T I C L E S
gel filtration (250 µL ribosomal complexes, 1 mL CL-4B column, as
described above). Avidin-coated agarose beads (100 µL) (Roche
Diagnostics), equilibrated in ice-cold WB containing 6% biotin-depleted
sterilized milk,44 were added to the flow-through (500 µL) and gently
shaken at 4 °C for 30 min to capture the labeled complexes. The beads
were then washed five times with 500 µL of ice-cold WB. During this
procedure, the beads were transferred to a new prechilled tube. To elute
the mRNA, 250 µL EB20 buffer (50 mM Tris-HCl, pH 7.5, 20 mM
EDTA) was added, and the solution was gently shaken for 10 min at
4 °C before centrifugation (2 min, 2000g). The supernatant (200 µL)
was used for mRNA purification (High Pure RNA Isolation Kit, Roche
Diagnostics). Reverse transcription (RT) and PCR was carried out as
described,44 except that the primers used were SD-BLA and BLArev.
Monitoring the Enrichment of Active â-Lactamase Relative to
the Inactive S70A Mutant. The product from RT-PCR was purified
with a spin column (QIAquick, Qiagen) according to the manufacturer’s
protocol. The DNA was subsequently analyzed by restriction analysis
with MscI or by sequencing. After digestion with MscI, the ratio of
active (uncleaved) to mutant (cleaved) â-lactamase was estimated from
the relative intensity of bands on a 2.5% Metaphor agarose gel (FMC)
using the program NIH image. DNA sequencing was carried out with
10 ng PCR product according to the manufacturer’s protocol (SequiTh-
erm EXCEL II, Epicenter Technologies). The enrichment factor was
estimated by determining the relative signal intensities of the 2
nucleotides that were mutated in the inactive S70A enzyme, with NIH
image.
and concentrated to 2.5 mg/mL. One liter of E. coli culture yielded 60
mg of gpHDBLIP, which was frozen in liquid nitrogen and stored at
-80 °C. The protein gave a single band in SDS-PAGE upon staining
with Coomassie-Blue. The â-lactamase inhibitory activity of gpHDBLIP
was quantified in â-lactamase inhibition experiments as described for
ampicillin sulfone (data not shown).
Comparison of Affinity Selection and Selection for Catalytic
Activity. Ribosome display was carried out as described above, but
only with active â-lactamase. After stopping the translation, the stopped
mix was incubated for 4 h on ice. A 200-µL aliquot was treated for 1
h on ice with biotinylated ampicillin sulfone (final concentration, 0.01
mM), while another aliquot was treated with gpHDBLIP (final
concentration, 0.3 µM). The selection procedure for the aliquot treated
with biotinylated ampicillin sulfone was as described above. A gel
filtration step was performed, as described for the fraction with the
suicide inhibitor. The ribosomal complexes binding to gpHDBLIP were
captured via the His6 tag by incubating the flow-through with 10 µL
of Tetra-His antibody (Qiagen, 0.1 µg/µL, in WB) and 100 µL of protein
G PLUS-Agarose (Santa Cruz) equilibrated in WB in the presence of
4% sterilized milk powder. The agarose was washed as described for
the magnetic beads. Rescuing of the mRNA, RT, and PCR was carried
as out described above.
Abbreviations: RT, reverse transcription; PCR, polymerase chain
reaction; GSH and GSSG, the reduced and oxidized forms of glu-
tathione, respectively; GdnHCl, guanidine hydrochloride; bp, base pairs;
BLIP, â-lactamase inhibitory protein; E. coli, Escherichia coli; GFP,
green fluorescent protein; gpHD, His-tagged gene product D from phage
lambda; IPTG, isopropyl-â-D-thiogalactoside; PAGE polyacrylamide
gel electrophoresis; PBS, phosphate-buffered saline; MOZ-ON, 4-meth-
oxybenzyloxycarbonyloxyimino-2-phenyl-acetonitrile.
â-Lactamase Inhibitory Protein (BLIP) Expression as pD Fusion.
We constructed a fusion protein of BLIP with protein D, carrying an
N-terminal His6 tag (gpHD). The BLIP (Swiss-Prot P35804) gene was
43
amplified by PCR and cloned into the vector pAT39 between the
BamHI and HindIII sites to yield gpHDBLIP, consisting of gpHD, a
Gly-Ser linker (GSGGGSGGGG), and BLIP. The fusion protein was
amplified from pAT39 by PCR and ligated into the vector pET40-b
(Novagen), carrying a kanamycin resistance gene. Plasmids were
verified by sequencing and subsequently used for expression in the E.
coli strain BL21 (DE3) [pLysS] (Stratagene) in 1 L of LB medium
containing 100 µg/mL kanamycin. Expression was induced with IPTG
(1 mM) at OD600 ) 0.8. The cells were grown for 3 h at 37 °C and
harvested by centrifugation. The cell pellet was resuspended in PBS
(10 mM sodium phosphate buffer, pH 7.4, 140 mM NaCl, 15 mM KCl),
and the cells were lysed by sonication. The resulting crude extract was
centrifuged (5 min, 6000 g, 4 °C), and the pelleted inclusion bodies
were washed once with PBS and resuspended with a solution containing
6 M GdnHCl, 0.1 M Tris, pH 8.0, and 10 mM imidazole. The
solubilized inclusion bodies were applied to a Ni-NTA column (15
mL) (Qiagen) and washed with GdnHCl buffer. The protein was
refolded on the column by replacing the GdnHCl-buffer with refolding
buffer (20 mM Tris pH 8, 500 mM NaCl, 20% glycerol, GSH (0.2
mM)/GSSG (1 mM)) in a linear gradient and eluted with 250 mM
imidazole in refolding buffer.52 The protein was dialyzed against PBS
Note Added in Proof: After this paper was submitted, it
was reported (Takahashi, F.; Ebihara, T.; Mie, M.; Yanagida,
Y.; Endo, Y.; Kobatake, E.; Aizawa, M. FEBS Lett. 2002, 514,
106-110) that 4 random point mutants of dihydrofolate
reductase, which had been obtained in a ribosome display
selection for binding to the inhibitor methotrexate, were active.
This suggests that enrichment of active enzyme variants may
also be possible simply by binding to an inhibitor under some
circumstances, not requiring the turnover-based selection re-
ported here, even though inactive mutants that still bind to an
inhibitor are certainly conceivable.
Acknowledgment. We thank Prof. Dr. Manfred Hesse for
the possibility to perform the organic synthesis in his laboratory
facilities. We also thank Patrik Forrer, Michael Stumpp, Kaspar
Binz, and Stephen Marino for fruitful discussions and technical
advice. This work was supported by Schweizerischer Nation-
alfonds Grant 3100-046624.96/1. Joelle Pelletier was a recipient
of a fellowship from le Conseil de Recherche en Sciences
Naturelles et en Ge´nie du Canada.
(52) Holzinger, A.; Phillips, K. S.; Weaver, T. E. BioTechniques 1996, 20, 804-
806, 808.
JA025870Q
9
J. AM. CHEM. SOC. VOL. 124, NO. 32, 2002 9403