Development of a High-Throughput Screen for Protein Catalysts: Application to the Directed Evolution of Antibody Aldolases

Article abstract of DOI:10.1002/anie.200352117

A semiautomated high-throughput system has been developed to express and purify proteins and assay their catalytic activity. The screen can be used to evolve activity, selectivity, and expression levels of proteins directly or in combination with selections. To illustrate its potential, the system was applied to the directed evolution of catalytic antibodies with aldolase activity.

Full text of DOI:10.1002/anie.200352117

Angewandte
Chemie
Screening for Protein Catalysts
extensive manipulation of plates such as centrifugation and
sonication were avoided.
Development of a High-Throughput Screen for
Protein Catalysts: Application to the Directed
Evolution of Antibody Aldolases**
After extensive optimization, we developed a semiauto-
mated system capable of routinely screening 6000 clones/run
in a parallel 96-well format. Yields of protein range from 0.5–
3
9
mg/well (1 mL culture/well) and purities are greater than
0%. In addition, a wide range of antibodies have been
Jeff Gildersleeve, Alex Varvak, Shane Atwell,
Doug Evans, and Peter G. Schultz*
successfully expressed and purified using the system without
[
5]
modifying the protocol. A typical screen starts by trans-
forming bacteria (Top10) with a library of plasmids encoding
mutant antibodies (mouse–human chimeric Fabs) fused to
High-throughput screens can be powerful methods for
[
1–3]
[6]
identifying and evolving biological catalysts.
To date,
His6 tags. The expression vector is a modified pBAD
nearly all high-throughput screens (e.g. blue/white colony
screens, plate lifts, cell sorting, etc.) assay catalytic activity in
live cells or crude cell lysates. As a result, there are significant
limitations on the substrates, reactions, and conditions that
can be used. In addition, detection of products in the presence
of many proteins, lipids, nucleic acids, and other cellular
components can be problematic. To circumvent these limi-
tations, we have developed a high-throughput system to
express, purify, and assay the catalytic activity of proteins. By
carrying out assays on purified proteins, a much wider range
of substrates and reactions can be explored, including those
involving cell-impermeable substrates, endogenous back-
ground activities, and nonchromogenic products. To illustrate
the potential of this system, we have applied it to the directed
evolution of catalytic antibodies with aldolase activity.
plasmid with expression controlled by the arabinose pro-
moter. Single colonies are picked into individual wells of
shallow, 96-well plates containing 2xYT/amp using a robotic
colony picker. These starter cultures are grown overnight and
then 10 mL is transferred from each well to a corresponding
[
7]
1 mL expression culture in a 2-mL, deep, 96-well plate. To
expedite the process, the 96 liquid transfers required for each
plate are done in parallel using a robotic liquid handler fitted
with a 96-needle head. The starter cultures are grown to
saturation to ensure that each expression culture is inoculated
at about the same density. In addition, the starter cultures
serve as a source of plasmid DNA for hits identified during
the screen (starter cultures are stored at 48C until completion
of the screen). The cultures are then grown for 7 h at 378C and
250 rpm, and protein production is induced by parallel
addition of arabinose (4%, 50 mL/well). Cultures are shaken
at 258C and 250 rpm for an additional 12 h, and then 100 mL/
The first step involved the development of a general
system for expressing and purifying large numbers of anti-
[
4]
À1
bodies in parallel. The system should be amenable to
optimization of expression levels and catalytic activity as both
are important properties for antibody development. Our
approach was to grow high-density 1-mL cultures of E. coli in
well of 10x PBS that contains 10units DNasemL is added.
Cells are lysed by addition of 100 mL/well of deoxycholic acid
À1
(10 mgmL ), and His6-tagged proteins are captured by
addition of 50 mL/well of Ni-NTA resin. Each of these
additions is carried out with the robotic liquid handler.
After shaking for an additional 4 h, the plates are allowed to
sit for 5 min to allow the Ni-NTA resin to settle to the bottom
of the wells. The needles of the liquid handler are inserted
into the wells just above the resin, and the lysate is removed.
The resin is then washed once with 500 mL/well of PBS that
contains 5 mm imidazole and then three times with 500 mL/
well of PBS to afford the immobilized, purified antibodies.
To test the performance of the system, we conducted a
series of model studies using catalytic antibody 38C2, an
9
6-well plates, lyse the cells, and then capture the His-tagged
antibodies with Ni-NTA beads. To maximize the utility of the
system, a single set of expression and purification conditions
was required that would accommodate many different anti-
bodies. In addition, sufficient quantities of purified antibodies
must be produced to obtain good signal-to-noise ratios in
catalytic assays. The expression levels and purity of protein
must also be reproducible so that mutants with incremental
changes in activity and expression levels can be detected with
confidence. Finally, protocols should be compatible with
automation and high-throughput formats; steps requiring
[
8]
aldolase developed by Wagner, Lerner, and Barbas. This
antibody efficiently catalyzes aldol reactions with a broad
range of substrates; it has also been extensively characterized
[
9,10]
both mechanistically and structurally.
We found that 38C2
[
*] Prof. P. G. Schultz, Dr. J. Gildersleeve, Dr. A. Varvak, Dr. S. Atwell
Department of Chemistry, The Scripps Research Institute
La Jolla, CA 92037 (USA)
retains activity while immobilized on Ni-NTA resin (~ 50%
relative to the antibody in solution) and can be assayed
directly without elution. In a representative assay, 38C2 was
expressed and purified in 80 wells distributed over 20 plates
and then assayed for activity by addition of 200 mL/well of
Fax: (+1)858-784-9440
E-mail: schultz@scripps.edu
D. Evans
The Genomics Institute of the Novartis Research Foundation
[11]
substrate 1 (20 mm in PBS). The product of the retro-aldol
3115 Merryfield Row, SanDiego, CA 92121 (USA)
reaction, 2 (Scheme 1), can be detected easily using a
fluorescent plate reader (l = 330 nm, l = 440 nm). After
[**] This research was supported by the NIH (Grant No. GM56528). J.G.
ex
em
gratefully acknowledges a postdoctoral fellowship from the NIH
incubating the plates for 1 h at 258C and 250 rpm, the reaction
was transferred to white 96-well plates, and fluorescent signals
were measured. The average signal was 1100 with a CV of
30% (signal variation from well to well and plate to plate).
(
Grant No. 5 F32 AI10419). This is manuscript number 15847-CH of
The Scripps Research Institute.
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
Angew. Chem. Int. Ed. 2003, 42, 5971 –5973
DOI: 10.1002/anie.200352117
ꢀ 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5971

Communications
improved antibodies can be reliably identified using the
high-throughput screen.
To further improve the activity and expression levels of
the aldolase antibodies, 26 single amino acid positions in close
proximity to the active site of 3D6 were randomized. These
residues included 31–36 (CDR1) and 89–97 (CDR3) on the
light chain and 33–37 (CDR1), and 94, 95, and 99–103
Scheme 1. 38C2-Catalyzed conversion of substrate 1 into 2.
(
CDR3) on the heavy chain. The 26 single amino acid libraries
[
12]
were combined and then screened (with 10 mm 1).
Two
With this level of variation, a two- to threefold change in
activity is detectable. For comparison, the initial signal for
substrate alone was about 220, the background signal after
incubation of substrate for 1 h in wells containing a catalyti-
cally inactive antibody was about 250, and the fluorescent
signal of cell cultures and crude cell lysates lacking substrate
was about 3000. The products were also analyzed by mass
spectrometry using an LC–TOF instrument equipped with a
unique mutants, Thr92 LAsn and Met34HVal, were identified
that resulted in yields threefold higher than those for 3D6 in
the high-throughput screen. The double mutant (1A4) was
generated, expressed, and characterized. Antibody 1A4 was
found to have a sixfold improvement in catalytic efficiency
(k /K ) and a 25-fold improvement in expression level over
38C2, corresponding to an overall 150-fold increase in
product yield. As with 38C2, 1A4 was found to be highly
selective for the S enantiomer of substrate 1. The generation
of antibody 1A4 from high-throughput screens of two
iterative focused libraries demonstrates that this is a reliable
and effective tool for directed evolution.
In conclusion, we have developed a high-throughput
screen capable of rapidly analyzing thousands of protein
mutants. The screen can be used to evolve activity, selectivity,
and expression levels of proteins directly or in combination
with selections. In addition to identifying mutants with
improved properties, one can use the screen to rapidly
compare protein library design strategies, evaluate the quality
of protein libraries, and determine the effectiveness of rounds
of selections.
cat
M
MUX eight-channel unit, albeit with lower overall throughput
À1
(
~ 500 samplesday ). Moreover, because the assays are
performed on resin-bound catalyst, up to three catalytic
assays can be conducted on the same sample of protein
without significantly affecting the signal-to-noise ratio and
variability by simply washing the resin twice with PBS
between assays. Furthermore, the antibodies can then be
eluted from the resin and analyzed by SDS-PAGE (sodium
dodecyl sulfate polyacrylamide gel electrophoresis) or
ELISA (enzyme-linked immunosorbent assay) to probe
expression levels and hapten binding. By conducting multiple
assays, we can simultaneously screen for mutants with
improved activity, selectivity, and/or expression levels.
To test the high-throughput screen as a tool for directed
evolution, we constructed two antibody aldolase libraries in
which key amino acid residues in the active site were
randomized to all 20 possible amino acid. One library had
residues 96–98 on the heavy chain randomized and the other
had residues 27B-D on the light chain randomized. From
these libraries, 17 unique mutants were identified that
produced signals two- to threefold higher than the 38C2
controls. The four heavy-chain mutants and 13 light-chain
mutants were then crossed and rescreened (with 20 mm 1) in
an effort to find combinations of mutations with additive
effects. Two clones, 2G7 and 3D6, resulted in yields 15-fold
higher than those for 38C2 in the screen. These mutants were
expressed in 1-L cultures, purified, and characterized in detail
Received: June 11, 2003 [Z52117]
Keywords: aldol reaction· a nt ibodies · directed evolution·
.
high-throughput screening · proteins
[1] N. Cohen, S. Abramov, Y. Dror, A. Freeman, Trends Biotechnol.
2001, 19, 507.
[
2] M. T. Reetz, Angew. Chem. 2002, 114, 1391; Angew. Chem. Int.
Ed. 2002, 41, 1335.
[
[
3] P. Soumillion, J. Fastrez, Curr. Opin. Biotechnol. 2001, 12, 387.
4] For other approaches, see: a) S. Yokoyama, Curr. Opin. Chem.
Biol. 2003, 7, 39, and references therein; b) A. Heyhurst, G.
Georgiou, Curr. Opin. Chem. Biol. 2001, 5, 683, and references
therein.
(
see Table 1). Both mutants have a 2.4-fold higher kcat value
and express at levels five- to sixfold higher levels than those
for the wild-type antibody, 38C2, demonstrating that
[
5] In addition to expressing and purifying thousands of mutants of
38C2, the following antibodies have also been successfully
expressed: a) 43C9: K. D. Janda, D. Schloeder, S. J. Benkovic,
R. A. Lerner, Science 1988, 241, 1188; b) 136.1: J. R. Jacobsen,
J. R. Prudent, L. Kochersperger, S. Yonkovich, P. G. Schultz,
Science 1992, 256, 365; c) 28B4: L. C. Hsieh, J. C. Stephans, P. G.
Schultz, J. Am. Chem. Soc. 1994, 116, 2167; d) 7G12: A. G.
Cochran, P. G. Schultz, Science 1990, 249, 781; e) AZ28: A. C.
Braisted, P. G. Schultz, J. Am. Chem. Soc. 1994, 116, 2211.
6] Detailed procedures for a typical high-throughput screen, the
construction of the expression vector, the construction of the
antibody libraries, and characterization of the antibodies can be
found in the Supporting Information.
Table 1: Amino acid sequences, catalytic properties, and expression
levels for selected antibodies.
[
a,b]
[a,c]
[d]
Antibody
Residues
kcat
KM
Expression
À1
À1
L27B-D H96-98 L92 H34 [min
]
[mm] [mgL
]
3
2
3
1
8C2
G7
D6
A4
L-H-T
D-V-L
L-V-S
L-V-S
F-Y-S
T
T
T
N
M
M
M
V
0.63
1.55
1.51
2.35
15
30
21
9
0.25
1.2
1.5
6.0
[
G-Q-K
G-Q-K
G-Q-K
[
[
a] kcat and
b] Æ0.04 min . [c] Æ1 mm. [d] Æ0.05 mgL
K
were measured using racemic substrate 1.
[7] A control antibody (typically 38C2) is expressed in 2–4 wells/
plate to serve as an internal control.
M
À1
À1
.
5
972
ꢀ 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
Angew. Chem. Int. Ed. 2003, 42, 5971 –5973

Angewandte
Chemie
[
8] J. Wagner, R. A. Lerner, C. F. Barbas III, Science 1995, 270,
797.
1
[
9] T. Hoffmann, G. Zhong, B. List, D. Shabat, J. Anderson, S.
Gramatikova, R. A. Lerner, C. F. Barbas III, J. Am. Chem. Soc.
1998, 120, 2768.
[
10] C. F. Barbas III, A. Heine, G. Zhong, T. Hoffmann, S. Grama-
tikova, R. Bjornestedt, B. List, J. Anderson, E. A. Stura, I. A.
Wilson, R. A. Lerner, Science 1997, 278, 2085.
[
11] B. List, C. F. Barbas III, R. A. Lerner, Proc. Natl. Acad. Sci. USA
1998, 95, 15351.
[
12] The 26 libraries were constructed in four groups, based on the
location of the randomized residue: light CDR1, light CDR3,
heavy CDR1, and heavy CDR3 (see Supporting Information).
The four groups were transformed separately and screened as
described in the supporting information with a unique clone in
each well. Plasmid DNA from each hit identified in the screen
was obtained by isolating the DNA from the corresponding
starter culture well.
Angew. Chem. Int. Ed. 2003, 42, 5971 –5973
www.angewandte.org
ꢀ 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
5973