combinatorial chemistry, part a

Preface
Combinatorial chemistry is a field that did not exist five years ago but
is so vibrant today, especially in medicinal chemistry, that almost every
major pharmaceutical company has a group working in this area and many
start-up companies have been formed with combinatorial chemistry as their
raison d'etre.
Like many other fast-breaking developments, this field had
its main origins in work done in academic research laboratories, and many
of the techniques were developed to solve specific problems in basic re-
search. The common feature of all combinatorial approaches is the genera-
tion of a complex mixture of molecules coupled to screens or selections
which can identify out of that mixture a single molecule with desired proper-
ties, e.g., as the ligand or inhibitor of an enzyme or as a macromolecule
with novel or enhanced properties. At the start most combinatorial libraries
were of biological molecules, mostly peptides or nucleic acids, but because
these molecules only rarely exhibit good pharmacological properties, in-
creasingly the libraries of interest to medicinal chemists are of small mole-
cules with a range of pharmacologically attractive properties.
Because of the rapid progress in this field, a follow-up to this volume
would not be possible in a single volume of
Methods in Enzymology,
but
at the time of the organization of this volume one could identify the main
themes that constitute this field and present the key technologies in a
single volume.
One of the earliest techniques for the generation and screening of a
diverse library of peptides was the display of random sequences in the coat
protein of single-strand DNA phages. The diversity of these libraries is
limited to the titers of phage one can obtain, typically >10 H particles/ml.
The phage coat protein can also accommodate entire proteins such as DNA
or RNA binding proteins that have been partially randomized so that

proteins with novel binding properties can be selected.
The techniques for the generation and screening of small molecule
libraries originated with peptides, and this volume contains a number of
early and still very useful techniques in this area. These libraries can be
screened by a number of very clever methods, including deconvolution of
different pools and the elegant and potentially very powerful encoded li-
braries.
The exploration of sequence space is most striking in the case of nucleic
acid libraries. Here, due to the power of the polymerase chain reaction,
libraries with diversities as high
as 1016
different molecules have been
explored. This is an extremely exciting area in which we are continually
xiii
xiv PREFACE
being surprised by the diversity of form and function possible within the
confines of the polynucleotide backbone. One can select RNA molecules
which can bind specifically to virtually any protein or small molecule and
also which can catalyze a diverse set of chemical reactions. And not only
can one explore sequence space in large libraries but as Tsang and Joyce
show in article [23] in this volume, one can expand that sequence space by
judicious mutagenesis during amplification between rounds of selection as
must have occurred during biological evolution.
It is clear, however, that much of the creative energy these days in this
field is being directed at inventing sophisticated methods for the generation
and screening of diverse kinds of small molecules, such as the pioneering
work by Ellman and colleagues on benzodiazepine libraries described in
this volume. Interestingly, the need to generate large diversity in these
libraries is not the key factor, and, instead, the ingenuity in the selection
of scaffolds and functional groups in generating the libraries will probably

be most important in generating interesting new pharmacological leads. In
this regard, one can expect interactions between computational chemistry
and combinatorial chemistry in which libraries are generated and screened
by computer methods in a search to find the most appropriate library for
a particular target. It is perhaps in that area that we should think now of
organizing a new volume in order to have something interesting for the
new millenium.
JOHN N. ABELSON
Contributors to Volume 267
Article numbers are in parentheses following the names of contributors,
Affiliations listed are current.
STEVEN C. BANVILLE (25), Chiron Corpora-
tion, Emeryville, California 94608
JOEL G. BELASCO (9), Department of Microbi-
ology and Molecular Genetics, Harvard
Medical School, Boston, Massachusetts
02115
SYLV1E E. BLONDELLE (13), Torrey Pines In-
stitute for Molecular Studies, San Diego,
California 92121
BARRY A. BUNIN (26), Department of Chem-
istry, University of California, Berkeley,
Berkeley, California 94720
CHARLIE L. CHEN (12), Hoechst Marion
Roussel, Tucson, Arizona 85737
JERZEY CIESIOLKA (19), Department of Mo-
lecular, Cellular, and Developmental Biol-
ogy, University of Colorado, Boulder, Colo-
rado 80309
RICHARD C. CONRAD (20), Department of

Chemistry, Indiana University, Blooming-
ton, Indiana 47405
RICCARDO CORTESE (6, 7), IRBM P. Angel-
etti, 00040 Pomezia, Rome, Italy
CHARLES CRAIK (3), Departments of Pharma-
ceutical Chemistry, Pharmacology, and
Biochemistry and Biophysics, University of
California, San Francisco, San Francisco,
California 94143
MILLARD G. CULL (10), Enzyco, Inc., Denver,
Colorado 80206
JEFFREY P. DAVIS (18), NeXstar Pharmaceuti-
cals, Inc., Boulder, Colorado 80301
JENNIFER M. DIAS (11), Affymax Research
Institute, PaiD Alto, California 94304
BARBARA DORNER (13), Torrey Pines Insti-
tute for Molecular Studies, San Diego, Cali-
fornia 92121
WILLIAM J. DOWER (11), Affymax Research
Institute, Palo Alto, California 94304
ANDREW D. ELLINGTON (20), Department of
Chemistry, Indiana University, Blooming-
ton, Indiana 47405
JONATHAN A. ELLMAN (26), Department of
Chemistry, University of California, Berke-
ley, Berkeley, California 94720
FRANCO FELICI (6, 7), IRBM P. Angeletti,
00040 Pomezia, Rome, Italy
GIANINE M. FIGLIOZZI (25), Chiton Corpora-
tion, EmeryviUe, California 94608

TIM FITZWATER (17), NeXstar Pharmaceuti-
cals, Inc., Boulder, Colorado 80301
GIOVANNI GALFR~ (6, 7), IRBM P. Angeletti,
00040 Pomezia, Rome, Italy
MARK GALLOP (16), Affymax Research Insti-
tute, Palo Alto, California 94304
CHRISTIAN M. GATES (10), Affymax Research
Institute, Palo Alto, California 94304
LORI GIVER (20), Division of Chemistry and
Chemical Engineering, Californm Institute
of Technology, Pasadena, California 91125
RICHARD GOLDSMITH (25), Chiron Corpora-
tion, Emveryville, California 94608
HARVEY A. GREISMAN (8), Department of Bi-
ology, Massachusetts Institute of Technol-
ogy, Cambridge, Massachusetts 02139
HYUNSOO HAN (14), Departments of Molecu-
lar Biology and Chemistry, The Scripps Re-
search Institute, La Jolla, California 92037
JACQUELINE L. HARRISON (5), United States
Biochemicals Pharma Ltd. (Europe), War-
ford WD1 8YH, United Kingdom
CHRISTOPHER P. HOLMES (16), Affymax Re-
search Institute, Palo Alto, California 94304
RICHARD A. HOUGHTEN (13), Torrey Pines
Institute for Molecular Studies, San Diego,
California 92121
MALI ILLANGASEKARE (19), Department of"
Molecular, Cellular, and Developmental Bi-
ology, University of Colorado, Boulder,

Colorado 80309
X CONTRIBUTORS TO VOLUME
267
KATHRYN M. IVANETICH (15), Biomolecular
Resource Center, University of California,
San Francisco, San Francisco, California
94143
KIM D. JANDA (14), Departments of Molecu-
lar Biology and Chemistry, The Scripps Re-
search Institute, La Jolla, California 92037
NEBOJ~A JANJI¢ (18), NeXstar Pharmaceuti-
cals, Inc., Boulder, Colorado 80301
GERALD F. JOYCE (23), Departments of
Chemistry and Molecular Biology, The
Scripps Research Institute, LaJolla, Califor-
nia 92037
JACK D. KEENE (21), Department of Microbi-
ology, Duke University Medical Center,
Durham, North Carolina 27710
ROBERT C. LADNER (2, 4), Protein Engi-
neering Corporation, Cambridge, Massa-
chusetts 02138
ITE A. LA1RD-OFFRINGA (9), Departments of
Surgery and Biochemistry and Molecular
Biology, University of Southern California
Medical School, Los Angeles, California
90033
KIT S. LAM (12), Departments of Medicine,
Microbiology, and Immunology, Arizona
Cancer Center, University of Arizona, Col-

lege of Medicine, Tucson, Arizona 85724
MICHAE LEBL (12), Hoechst Marion Roassel,
Tucson, Arizona 85737
ALLESANDRA LUZZAGO (6, 7), IRBM P. An-
geletti, 00040 Pomezia, Rome, Italy
DEREK MACLEAN (16), A ffymax Research In-
stitute, Palo Alto, California 94304
IRENE MAJERFELD (19), Department of Mo-
lecular, Cellular, and Developmental Biol-
ogy, University of Colorado, Boulder, Colo-
rado 80309
WILLIAM MARKLAND (2, 4), Vertex Pharma-
ceuticals, Inc., Cambridge, Massachusetts
02139
EDITH L. MARTIN (10), Affymax Research In-
stitute, Palo Alto, California 94304
LARRY C. MATFHEAKIS (11), Affymax Re-
search Institute, PaiD Alto, California 94304
PAOLO MONACI (6, 7), IRBM P. Angeletti,
00040 Pomezia, Rome, Italy
SIMON C. NG (25), Chiron Corporation, Em-
eryville, California 94608
ZHI-JIE NI (16), Affymax Research Institute,
PaiD Alto, California 94304
TIM NICKLES (19), Department of Molecular,
Cellular, and Developmental Biology, Uni-
versity of Colorado, Boulder, Colorado
80309
ALFREDO NICOSIA (6, 7), IRBM P. Angeletti,
00040 Pomezia, Rome, Italy

PETER E. NmLSEN (24), Department of Medi-
cal Biochemistry and Genetics, Center for
Biomolecular Recognition, The Panum In-
stitute, DK-2200 N Copenhagen, Denmark
AHUVA NISSIM (5), The Institute of Hematol-
ogy, The Chaim Sheba Medical Centre,
Sachler School of Medicine, Tel Hashomer
52621, Israel
JOHN M. OSTRESH (13), Torrey Pines Institute
for Molecular Studies, San Diego, Califor-
nia 92121
CARL O. PABO (8), Department of Biology,
Howard Hughes Medical Institute, Massa-
chusetts Insitute of Technology, Cambridge,
Massachusetts 02139
MATrHEW J. PLUNKETr (26), Department of
Chemistry, University of California, Berke-
ley, Berkeley, California 94720
BARRY POLISKY (17), NeXstar Pharmaceuti-
cals, Inc., Boulder, Colorado 80301
EDWARD J. REBAR (8), Department of Biol-
ogy, Massachusetts Institute of Technology,
Cambridge, Massachusetts 02139
BRUCE L. ROBERTS (2, 4), Genzyme Corpora-
tion, Framingham, Massachusetts O1701
MARGARET E. SAKS (22), Division of Biology,
California Institute of Technology, Pasa-
dena, California 91125
JEFFREY R. SAMPSON (22), Division of Biol-
ogy, California Institute of Technology,

Pasadena, California 91125
DANIEL V. SANTI (15), Department of Phar-
maceutical Chemistry, University of Cali-
fornia, San Francisco, San Francisco, Cali-
fornia 94143
PETER J. SCHATZ (10), Affymax Research In-
stitute, Palo Alto, California 94304
CONTRIBUTORS TO VOLUME
267 xi
GEORGE P. SMITH (1), Division of Biological
Sciences, University of Missouri, Columbia,
Missouri 65211
PETER STROP (12), Hoechst Marion Roussel,
Tucson, Arizona 85737
Yu TIAN (20), Department of Chemistry, Indi-
ana University, Bloomington, Indiana
47405
JOYCE TSANG (23), Departments of Chemistry
and Molecular Biology, The Scripps Re-
search Institute, La Jolla, California 92037
CHENG-I WANG (3), Department of Pharma-
ceutical Chemistry, University of California,
San Francisco, San Francisco, California
94143
MARK WELCH (19), Department of Molecular,
Cellular, and Developmental Biology, Uni-
versity of Colorado, Boulder, Colorado
80309
SAMUEL C. WILLIAMS (5), Medical Research
Council Centre for Protein Engineering,

Cambridge CB2 2QH, United Kingdom
GREG WINTER (5), Medical Research Council
Centre for Protein Engineering, and Labo-
ratory of Molecular Biology, Cambridge
CB2 2QH, United Kingdom
QING YANG (3), Department of Pharmaceuti-
cal Chemistry, University of California, San
Francisco, San Francisco, California 94143
MICHAEL YARUS (19), Department of Molec-
ular, Cellular, and Developmental Biology,
University of Colorado, Boulder, Colo-
rado 80309
JINAN YU (1), Department of Pharmacology,
School of Medicine, University of Pitts'-
burgh, Pittsburgh, Pennsylvania 15261
DOMINIC A. Z1CHI (18), NeXstar Pharmaceu-
ticals, Inc., Boulder, Colorado 80301
SHAWN ZXr~NEr~ (19), Department of Molecu-
lar, Cellular, and Developmental Biology,
University of Colorado, Boulder, Colo-
rado 80309
RONALD N. ZUCKERMAN (25), Drug Design
and Development, Chiton Corporation,
Emeryville, California 94608
[ 1]
AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 3
[11 Affinity Maturation of Phage-Displayed
Peptide Ligands
By JINAN YU and GEORGE P. SMITH
Introduction

Many experiments in this volume start with large libraries of random
amino acid or nucleotide sequences of a certain length from which a tiny
subset is selected according to some criterion of "fitness" most often,
affinity for a chosen target receptor. In most cases the library represents
sequences of the same length exceedingly sparsely. Even the very best
(fittest) sequence in a sparse initial library may be much inferior to the
globally best sequence of the same length.
If the sequences are capable of heritable mutation phage display and
random RNA and DNA libraries fall into this category the problem of
sparseness might be addressed by encouraging fitter sequences to "evolve"
from parent sequences in the initial library. 1'2 This sort of artificial evolution
is exemplified by the "greedy" strategy: Step A, from the initial library
select the very best sequence; call this the "initial champion." Step B,
mutagenize the initial champion randomly, producing a "clan" of closely
related mutants. Step C, from that clan select the mutant with the very
best fitness. Step D, repeat Steps B and C as needed until an optimal ligand
is found. Each round of selection thus selects "greedily" for the very best
sequence available in the current population.
A drawback of the greedy strategy is that it can only explore close
relatives of the initial champion a tiny parish in the vast "space" of
possible sequences. Yet, for all we know, the best sequence in that neighbor-
hood may be far inferior to sequences lying totally elsewhere in sequence
space. Might it not then be worthwhile to explore the neighborhood of the
second-best sequence in the initial library? of the third best? of every
sequence with fitness above a certain threshold? In order thus to broaden
the search for fitter sequences, the stringency (fitness threshold) can be
reduced in the early rounds of selection, so as to include sequences some-
what inferior to the initial champion: Step A', from the initial library select
a mixture of sequences with diverse fitnesses (ideally, above a certain
threshold). Step B', mutagenize the entire population of selected sequences

1 D. J. Kenan, D. E. Tsai, and J. D. Keene,
Trends Biochem. Sci.
19, 57 (1994).
2 j. W. Szostak,
Trends Biochern. Sci.
17, 89 (1992).
Copyright © 1996 by Academic Press~ Inc.
METHODS IN ENZYMOLOGY. VOL. 267 All rights of reproduction in any form reserved.
4
PHAGE DISPLAY LIBRARIES [ 1]
to produce many clans of mutants. Step C', from those clans select a mixture
of sequences with diverse fitnesses (ideally, above a slightly higher threshold
than in Step A'). Step D', repeat steps B' and C' as often as desired,
possibly increasing the stringency of selection with succeeding rounds. Step
E', after the final round of mutagenesis, stringently select the very best
sequence in the current population.
Alternating nonstringent selection with mutagenesis in this way makes
it possible to discover "dark horses": sequences in the initial library that
are inferior to the initial champion, yet can be mutated to even higher
fitness than can that champion. A dark horse will usually lie in a different
neighborhood than the initial champion, since in most cases two sequences
in the same small neighborhood will be able to mutate to the same local
optimum. Even a well-implemented experiment may fail to reveal dark
horses in any particular case (see Discussion), most obviously because there
are none to reveal. Still, dark horses may appear sufficiently frequently to
make this an attractive alternative to the greedy strategy.
When the fitness being selected for is affinity for a target receptor
molecule, the foregoing program is called "affinity maturation," the term
coined by immunologists for the interspersed rounds of selective stimulation
by antigen and somatic mutation of antibody genes that is thought to

give rise to antibodies with increasing affinity in the course of an immune
response. 3 This chapter covers affinity maturation from random peptide
libraries displayed on phage. The procedures and underlying principles will
be discussed in the context of a specific exemplar experiment in which
ligands for a model receptor were selected from a library of random 15-
mers. 4 The model receptor was S-protein, a 104-residue fragment of bovine
ribonuclease prepared by partial digestion with subtilisin; the other frag-
ment, S-peptide, corresponds to the N-terminal 20 amino acids. 5 Neither
fragment alone is enzymatically active, but when they are mixed, S-peptide
binds strongly to S-protein, restoring enzyme activity. 6
Vector, Initial Library, and Overall Plan
The procedures in this article are tailored for libraries in fUSE5 7 and
related vectors, which have a tetracycline (Tc) resistance determinant in
3 Eisen, H. N.,
in
"Molecular Evolution on Rugged Landscapes: Proteins, RNA and the
Immune System" (A. S. Perelson and S. A. Kauffman, eds.), p. 75. Addison-Wesley, New
York, 1991.
4 T. Nishi, H. Tsurui, and H. Saya,
Exp. Med.
11, 1759 (1993).
s F. M. Richards and P. I. Vithayathil,
I. Biol. Chem. 234,
1459 (1959).
6 H. C. Taylor, D. C. Richarson, I. S. Richardson, A. Wlodawer, A. Komoriya, and I. M.
Chaiken,
J. Mot Biol.
149, 313 (1981).
7 j. K. Scott and G. P. Smith,
Science

249, 386 (1990).
[ 1]
AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 5
BglI BgII
CTATTCTCACTCC-GCCGACGIGGGCT(NNK) 15GGGC~CGCT~ GGGCCGAAAcTGTTGAA
Forward primer ~ + Reverse primer
A D G A X15 G A A G A E T V E
FIG. 1. Nucleotide sequence near the beginning of the pill gene in the random 15-mer
library. 4 Only the plus strand the strand that is packaged in virions and that is anticomplemen-
tary to mRNA is shown. In the initial library, before selection, positions designated N had
(theoretically) an equal mixture of all four nucleotides, K an equal mixture of G and T. The
corresponding amino acid sequence at the N terminus of mature pill is shown in the one-
letter code; X~5 stands for the random 15-mer encoded by the degenerate codons. The PCR
priming sites used in construction of mutant libraries (see Mutagenesis) are underlined.
Cleavage of the PCR product at the flanking
BglI
sites releases a degenerate 60-bp fragment
that can be spliced to the
Sill-cleaved
fUSE5 vector.
the minus-strand origin 8 (changes required for other vectors 9 are obvious
and do not materially affect the discussion). Although the resulting defect
in minus-strand replication reduces plaque size to near invisibility, the
phage can be cloned and propagated as plasmids by infecting a Tc-sensi-
tive host and growing in medium containing Tc (filamentous phage do not
kill the host cell); phage are titered as transducing units (TU) by counting
Tc-resistant colonies. Only cells bearing F-pili can be infected, but the
pilus is not required for phage production by transfected cells. Expression
of Tc resistance by newly infected or transfected cells is induced by
culturing them ~30 min in a subinhibitory concentration of Tc (0.2

/zg/ml).
Phage libraries, including the receptor-specific mutant libraries created
in the course of affinity maturation (see Mutagenesis), are constructed by
splicing foreign DNA inserts into the gene for coat protein plIl (five copies
at one tip of the virus) or pVIII (thousands of copies forming the tube
surrounding the DNA). The peptide encoded by the insert is displayed
on the virion surface fused to the coat protein and is available to bind
macromolecular target receptors for which it has affinity.
The fUSE5 vector has two SfiI cloning sites near the beginning of the
plII gene, 7 between which a synthetic BglI fragment with 15 degenerate
codons was inserted to create the initial library for the exemplar experiment 4
(Fig. 1). Each clone has a particular sequence of 15 codons and displays
the corresponding 15-residue peptide. There are 3.3 × 1019 possible 15-
mers altogether, but only ~2 × 108 clones in the initial library a sparse
library indeed.
8 G. P.
Smith,
Virology
167,
156 (1988).
9 G. P. Smith and J. K. Scott,
Methods Enzymol.
217, 228.
6
PHAGE DISPLAY LIBRARIES [1]
Affinity maturation begins with alternating rounds of affinity selection
and mutagenesis, the stringency of selection being kept low (see Introduc-
tion). The phage population resulting from these alternating rounds
hopefully greatly enriched for receptor-binding clones is then subjected
to additional rounds of stringent selection without mutagenesis in order to

identify the highest-affinity clones, which are analyzed by sequencing and
binding studies. Figure 2 outlines the sequential arrangement of selection
steps (producing Eluates 1-3, 4A-4F, and 5A-5F) and mutagenesis steps
(producing Mutant Libraries 1 and 2) in the exemplar experiment; also
Conventional
selection
"~ 10 lag
[ Eluate 2' ]
~ 2-step
100 ng
I Eluate 3' I
] Initial librar~ [
~ l-step selection
with 10 lag receptor
. 1 ]
~ mutagenesis
[ Mutantlibrary 1 1
,~ l-step selection
with I lag receptor
I Eluate 2 I
~ mutagenesis
~ l-step selection
with I lag receptor
I Eluate 3 I
ng ~4¢ 2-step
lO ng
[ Eluate 4A[ I Eluate 4B [ [ Eluate 4C I ] Eluate 4D I LEluate 4E [ [ Eluate 4F I
,-step ~ 1-step ~ 1-step ~ 2-step ~ 2-step % 2-step
I lag 100 ng 10 ng I lag 100 ng 10 ng
"Eluate5A] ]Eluate5BI ~ ]Eluate5D I [Eluate5E] IEiuate5F]

FIG. 2. Outline of the exemplar affinity maturation of ribonuclease S-protein ligands.
Arrows labeled "l-step selection" and "2-step selection" correspond to rounds of affinity
selection by the one- and two-step methods described under Affinity Selection; the amount
of receptor (biotinylated S-protein) used in each round is shown. All eluates but 3' and 5A-5F
were amplified (see
Quantifying Yield and Amplifying Eluates
under Affinity Selection) before
being mutagenized or subjected to the next round of affinity selection. Arrows labeled "muta-
genesis" correspond to PCR mutagenesis and mutant library construction (see Mutagenesis).
Also shown is a conventional affinity selection experiment (without mutagenesis) that was
carried out in parallel with affinity maturation. 1° Thus, Eluate 2' was selected directly from
Eluate 1, and Eluate 3' from Eluate 2', without mutagenesis.
[ l I AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 7
shown is a conventional selection experiment without mutagenesis (Eluates
2' and 3') that was carried out in parallel for comparison.
TM
In the sections that follow, the principles and practice of affinity matura-
tion will be discussed in detail, with the exemplar experiment serving
throughout as an illustration. Table I gives the formulas or recipes for
solutions and preparations, Table II describes standard procedures, and
Table III lists Escherichia coli strains.
Affinity Selection
Each affinity selection step starts with a mixture of phage and seeks to
select from that mixture phage whose displayed peptide binds the target
receptor. These phage are specifically "captured" by immobilizing the re-
ceptor on a solid surface (e.g., a plastic petri dish); unbound phage are
washed away and captured phage are eluted (still in infective form), yielding
a selected subset of the original phage mixture that is called an "eluate."
Stringency
The stringency of affinity selection is controllable in some degree by

the choice of conditions, as will be detailed later. The logic of affinity
maturation calls for low stringency (thus high yield) in the early rounds of
selection (see Introduction). There is an additional argument even in
conventional selection without mutagenesis for choosing high yield in the
very first round of selection, whose input consists of all clones in the initial
library. Because the library has many clones, each clone is represented by
few particles (-500 TU/clone on average in the exemplar experiment);
consequently, if the yield for a binding clone is not high in the first round
(>0.2% in the exemplar experiment), that clone has a good chance of being
lost, and of course can never be recovered. In later rounds, especially after
the last round of mutagenesis, stringency can be increased in order to select
for the tightest binder.
There is a limit to stringency, however. The reason is that there is always
a background yield of nonspecifically bound phage; if stringency is set
too high, the yield of specifically captured phage will fall far below the
background of nonspecifically bound phage, and all power of discrimination
in favor of high affinity is lost.
In practice, because the relationship between selection conditions and
stringency is unknown in advance, it is advisable to explore a range of
conditions in the final rounds of selection; those whose yields are close
10 D. A. Schultz, J. E. Ladbury, G. P. Smith, and R. O. Fox, unpublished (1995).
8 PHAGE DISPLAY LIBRARIES [ 11
TABLE I
SOLUTIONS AND PREPARATIONS
Solution or
preparation Description
Acrylamide gel
AP-SA (500/zg/ml
stock)
AP-SA diluent

Biotin (10 mM stock)
Biotinylated BSA (2
mg/ml stock)
Blocking solution
BSA (50 mg/ml
stock)
Dialyzed BSA (50
mg/ml stock)
Elution buffer a
18.75 ml 38% (w/w) acrylamide/2% (w/w) bisacrylamide, 10 ml
5× TBE and 21.25 ml water are mixed and degassed; 20/~1
N,N,N',N'-tetramethylethylenediamine
and 375/~1 10% (w/w)
ammonium persulfate are added to initiate polymerization
Alkaline phosphatase-conjugated streptavidin; Jackson Immuno-
Research Laboratories (West Grove, PA); dissolved in 5 mM
Tris-HC1 (pH 8), 125 mM NaC1, 10 mM MgC12, 1 mM ZnC12,
50% (v/v) glycerol; stored at 4 °
1 mg/ml bovine serum albumin (BSA), 0.1% Tween 20, 1 mM
MgCI2, 0.1 mM ZnCI2 in TBS
1 N NaOH is added slowly to a stirred suspension until the solid
dissolves and the pH reaches 6-9; filter-sterilized; stored at
-20 °
Biotinamidocaproyl-labeled BSA, 8.9 biotin/molecule; Sigma
Chemical Co. (St. Louis, MO), A6043; dissolved at 2 mg/ml in
water; filter-sterilized; stored at 4 °
0.1 M NaHCO3, 5 mg/ml dialyzed BSA, 0.1/zg/ml streptavidin,
200/zg/ml NAN3; filter-sterilized; stored at 4°; reused until mi-
crobial contamination is evident
BSA, Fraction V; Sigma Chemical Co.; filter-sterilized; stored at

4 °
BSA, extensively dialyzed; Sigma Chemical Co. A6793; presumed
to be free of biotin; filter-sterilized; stored at 4 °
0.1 N HC1, 1 mg/ml BSA, pH adjusted to 2.2 with glycine; filter-
sterilized; stored at room temperature
to background are probably too stringent to be useful. In the exemplar
experiment, for instance, six different conditions for rounds 4 and 5 were
tried, yielding final eluates 5A-5F (Fig. 2).
Capture via Biotinylated Receptor
If receptor protein is available in relatively pure form, it is convenient
to biotinylate it at accessible e-amino groups. This allows it to be rapidly
and irreversibly captured on streptavidin-coated petri dishes under nonde-
naturing conditions and also facilitates ELISA (see Binding Studies). (Nu-
merous alternative immobilization methods are available, but will not be
discussed here.) In a typical protocol, 10-40/zg protein is reacted with
50-400/xM sulfosuccinimidyl-6-(biotinamido)hexanoate (NHS-LC-biotin;
Pierce Chemical Co., Rockford, IL) in 44/xl of 0.1 M NaHCO3; residual
[ 1] AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 9
TABLE I
(continued)
Solution or
preparation Description
NAP buffer 80 mM NaC1, 50 mM NH4H2PO4, pH adjusted to 7.0 with
NHnOH; autoclaved; stored in refrigerator or room temper-
ature
Just before use, 10/xl 1 M MgC12 and 100/zl of 50 mg/ml
p-nitrophenylphosphate (stored at -20 °) added to 10 ml 1 M
diethanolamine (pH adjusted to 9.8 with HC1)
16.6 mM (NH4)2SO4, 67 mM Tris-HCl (pH 8.8), 6.1 mM MgC12,
6.7 p,M NazEDTA (pH adjusted to 8.0 with NaOH), 0.17 rag/

ml BSA
14.5% (w/w) polyethylene glycol 8000, 16.9% (w/w) NaC1
Any bacterial culture medium, such as NZY °
0.2/zg/ml Tc in SOC medium b
0.5 M Tris, 0.5 M
H3BO3,
10 mM Na2EDTA
50 mM Tris-HCl (pH 7.5), 0.1 M NaCI; autoclaved; stored at
room temperature
0.5% (v/v) Tween 20 in TBS; autoclaved; stored at room temper-
ature
1 : 1 (v/v) mixture of filter-sterilized 40 mg/ml tetracycline (Tc)
and autoclaved glycerol (cool before mixing); stored at -20 °
Rich medium (e.g., NZY) with 20 p,g/ml tetracycline (Tc)"
Petri dishes with agar medium containing 40/xg/ml tetracycline
(Tc) ~
10 mM Tris-HC1 (pH 8), 1 mM Na2EDTA (pH adjusted to 8.0
with NaOH); autoclaved; stored at room temperature
1 mg/ml dialyzed BSA, 200 ~g/ml NaN3 in TBS/Tween
NPP substrate
PCR buffer
PEG/NaCI
Rich medium
SOC/Tc
TBE (5× stock)
TBS
TBS/Tween
Tc (20 mg/ml stock)
Tc medium
Tc plates

TE
T'FDBA
Details in Smith and Scott?
b Details in Sambrook
et al. 15
reagent is quenched with ethanolamine, nonbiotinylated carrier protein is
added, and quenched reagent is removed by ultrafiltration as detailed else-
where .9
The biotinylated receptor can be used in two ways (details in the next two
subsections): in "one-step" selection, phage are captured by a biotinylated
receptor that has been preimmobilized on the surface of a streptavidin-
coated petri dish; whereas in "two-step" selection, phage are reacted with
biotinylated receptor in solution, then subsequently captured on a streptavi-
din-coated dish. In either case, the input phage number is typically about
101~ TU a 10-/zl portion of the initial library, a 100-/xl portion of a mutant
library, or a 100-/.d portion of the eluate from the previous round of affinity
11D. W. Leung, E. Chen, and D. V. Goeddel,
Technique
1, 11 (1989).
10 PHAGE DISPLAY LIBRARIES [ 1
]
TABLE II
STANDARD PROCEDURES
Culture supernatant
DNA extraction a
Electrocompetent
cells b
PEG precipitation
Propagation and
processing of

phage c
Removal of super-
natant
Starved cells c
Titering TU c
Vector DNA c
Clear cells from a grown culture by two successive centrifugations
(4000-12,000 g, 4 °, 10 min), saving supernatant each time.
Extract with phenol and chloroform, precipitate with ethanol, and
dissolve in buffer or water to desired concentration.
Wash log-phase MC1061 cells (Table III) twice with ice-cold 1 mM
4(-2-hydroxyethyl)-l-piperazineethanesulfonic acid (pH adjusted
to 7.0 with NaOH) and once with ice-cold 10% (v/v) glycerol by
centrifugation (3500 g, 4 °, 10-15 min) and gentle resuspension;
gently resuspend final cell pellet in 1/800 culture volume ice-cold
10% glycerol; use immediately without freezing.
To phage in 1 volume medium or other solution add 0.15 vol
PEG/NaCI (Table I) and incubate for at least 4 hr at 4-25°; cen-
trifuge or microfuge (at least 7500 g, 10-30 rain, 4-25 °) to pellet
phage; remove all supernatant (see below); dissolve pellet in de-
sired buffer (up to -5 x 1013 virons/ml); centrifuge or microfuge
briefly to pellet insoluble matter, transferring cleared superna-
tant to new vessel
Large scale: Propagate phage in l-liter cultures, precipitate with
PEG, purify by CsC1 density equilibrium centrifugation. Small
scale: Propagate phage in 1.5-ml cultures, precipitate with PEG.
Aspirate or decant supernatant from centrifuged pellet, recentri-
fuge (maintaining centrifugal orientation) to drive residual super-
natant to bottom, aspirate residual supernatant.
Pellet log-phase K91, K91Kan, or K91BlueKan cells (Table III) by

centrifugation; resuspend gently in 1 culture volume 80 mM
NaCI; shake gently at 37 ° for 45 rain; pellet by centrifugation; re-
suspend in 1/20 culture volume cold NAP buffer (Table I); store
at 4 ° for up to 1 week.
Infect 10/xl starved cells with 10-/zl phage dilutions for 10-30 min
at room temperature in a 17 × 100-mm tube; dilute with i ml
rich medium containing 0.2/~g/ml Tc (Table I); shake for 30-60
min at 37°; spread 200/zl on Tc plate (Table I).
Cleave at cloning sites, isopropanol precipitate to remove "stuffer"
between sites.
"Details in Sambrook
et aL 15
b W. J. Dower, J. F. Miller, and C. W. Ragsdale,
Nucleic Acids Res.
16, 6127 (1988).
c Details in Smith and Scott. 9
selection. The procedure and the amount of receptor (biotinylated S-pro-
tein) used in each selection step in Fig. 2 are indicated.
One-Step Selection
A 35-mm petri dish is coated with 400/zl of 10/zg/ml streptavidin in
0.1 M NaHCO3 for at least 1 hr at room temperature, then blocked with
[ 1]
AFFINITY MATURATION OF PHAGE-BORNE L1GANDS
1 l
TABLE III
Escherichia coli
STRAINS
Strain Sex Chromosomal genotype Characteristics
MC1061" F
hsdR rncrB A(araABC-

Uninfectableb; streptomy-
1eu)6779 araD139 Alac174
cin resistant
galU galK strA thi
thi
lacZ::mkh e thi
K91' Hfr Cavalli
K91Kan d Hfr Cavalli
K91BlueKan Hfr Cavalli
lacZAM15 lacY::mkh e lacl °
thi
Infectable
Infectable; kanamycin re-
sistant
Infectable; kanamycin resis-
tant; co-donor for Lac c~
complementation
P. S. Meissner, W. P. Sisk, and M. L. Berman,
Proc. Nat. Acad. Sci. U.S.A.
84, 4171 (1987).
h F- cells cannot be infected, but can support intracellular replication and virus production.
A A- derivative of K38 [L. B. Lyons and N. D. Zinder,
Virology
49, 45 (1972)].
d Details in Smith and Scott. 9
mkh
is the "mini-kan hopper" transposon [J. C. Way, M. A. Davis, D. Morisato, D. E.
Roberts, and N. Kleckner,
Gene
32, 369 (1984)], which confers kanamycin resistance.

blocking solution (Table I) for 2 hr; after washing five times with TBS/
Tween (Table I) from a squirt bottle (slapping the dish face down on a
clean paper towel each time), the desired amount of biotinylated receptor
(0.01-10/zg S-protein in the exemplar experiment, Fig. 2) is added in 400
/zl TFDBA (Table I); the dish is allowed to react at least 2 hr at 4 °, washed
five times with TBS/Tween to remove unbound receptor, and filled with
400/xl of TTDBA. In order to block unoccupied biotin-binding sites on
the streptavidin, 4/xl of 10 mM biotin (Table I) is added to the dish, which
is rocked at room temperature for 10 min before adding input phage (there
is no need to remove excess free biotin). The dish is rocked (usually at 4 °,
but sometimes at other temperatures) for 4 hr and is washed 10 times with
TBS/Tween as described earlier. Bound phage are eluted from the dish
with 400/zl of elution buffer (Table I) for 10 rain, transferred to a microtube,
and neutralized by mixing with 75/zl of 1 M Tris-HC1 (pH 9.1).
When the amount of biotinylated receptor is enough to saturate the
immobilized streptavidin (1-10/xg per 35-ram dish), this procedure gives
the maximum achievable yield, which can reach 20% of the input phage.
When, as in the exemplar experiment, each phage particle displays multiple
copies of the random peptide, this high yield is plausibly attributed to
attachment of a single virion to two or more neighboring receptor molecules;
a particle captured multivalently in this fashion may dissociate from the
solid surface exceedingly slowly, even if the underlying monovalent affinity
is only modest. As the density of immobilized receptor is decreased, this
12
PHAGE DISPLAY LIBRARIES [ 1]
"avidity effect" is reduced, possibly to the point where the yield from
monovalent attachment comes to dominate the output conditions that
should strongly favor high affinity.
Two-Step Selection
Input phage are equilibrated overnight at 4 ° with the desired amount of

biotinylated receptor in TTDBA (Table I; typically ~100/zl). The reaction
solution is then added to a dish that has been previously coated with
streptavidin as in one-step selection and filled with 400/zl of TTDBA. After
rocking for 10 rain at room temperature to permit capture by immobilized
streptavidin, the dish is washed and eluted as in one-step selection.
During the equilibrium step, receptors (assuming they are monovalent)
bind phage reversibly according to solution-phage equilibrium kinetics. If
there is little dissociation and reassociation during the subsequent 10-min
capture step, the situation at the beginning of the capture step will largely
determine the relative yields of different clones. If, at the other extreme,
receptors dissociate and reassociate very rapidly during the capture step,
two-step selection is really equivalent to an abbreviated one-step selection.
If desired, reassociation can be suppressed during the capture step by
adding a competitive ligand for the receptor at high concentration (such a
competitor S-peptide was available in the exemplar experiment, but
was not in fact used). In practice, two-step selection gives considerably
lower yields than one-step selection, even when reassociation is not sup-
pressed (next subsection).
Quantifying Yield and Amplifying Eluates
Eluates that are to serve as input for mutagenesis or further rounds of
affinity selection (e.g., all eluates but 3' and 5A-5F in Fig. 2) are amplified
by propagating the phage in fresh host cells. An eluate from the initial
library or from a mutant library (e.g., Eluates 1-3 in Fig. 2) is first concen-
trated and washed once with TBS on a Centricon 30-kDa ultrafilter (Ami-
con, Danvers, MA) to give a final volume of 100/zlg; this allows the entire
eluate to be amplified, reducing the chance that a binding clone will be
lost (see
Stringency).
Eluates from subsequent rounds, in which every clone
is represented by many thousands or millions of phage particles, are used

without concentration.
A 100-/zl portion of starved cells (Table II) is infected with eluate (the
entirety of a concentrated eluate or a 100-/~1 portion of an unconcentrated
one) for 10-30 min at room temperature. The infected cells are inoculated
into 20 ml rich medium (Table I) containing 0.2/zg/ml Tc (Table I) and
are shaken for 30-60 min at 37 °. After adding additional Tc to a final
[ 1]
AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 13
concentration of 20/zg/ml, 200-/zl portions of appropriate dilutions of the
culture are spread on Tc plates (Table I) to quantify the output of the
affinity selection. At the same time, input phage are titered in the ordinary
way (Table II); the yield of each affinity selection can be calculated by
dividing the output by the input.
Meanwhile, the main 20-ml culture is shaken overnight at 37 °. Phage
are partially purified from the culture supernatant (Table II) by two PEG
precipitations (Table II), ending up in 200/zl TBS (optionally containing
0.02% NaN3 as preservative). The physical particle concentration in this
"amplified eluate" is ~5 x 10 t3 virions/ml, regardless of the titer in the
unamplified eluate; the titer is -0.5-5 × 1012 TU/ml.
Figure 3 shows yields from successive rounds of affinity selection in the
exemplar experiment. Results for Eluates 1-3, 4A, and 5A (Fig. 2) can
be directly compared, as they were obtained under essentially the same
conditions: one-step selection with saturating levels of receptor. The yield
from the first round is close to background (-3 × 10-5%), reflecting the
rarity of receptor-binding clones in the initial library. Even if binding clones
are enriched a millionfold over nonbinding ones, the output of this round
may still be dominated by phage that have been captured nonspecifically
(see Sequence Analysis). The yield increases to a maximum by the third
round, however, as binding clones come to dominate. Yields in the fourth-
and fifth-round eluates other than 4A and 5A reflect the stringency of the

selection conditions: as the amount of receptor decreases, so does yield,
and at a given level of receptor, one-step selection gives higher yields than
o~
1
0.1
0.01
0.~1
0.0~1
0.00001
* l-step, 1 lag
¢~ •
,t l-step, 100 ng
• 1-step, 10 ng
e 2-step, 1 lag
a-'- 2-step, 100 ng
', , e 2-step, 10ng
• 1-step, 1-10 lag
I I I I
2 3 4 5
Round of selection
FiG. 3. Yields from affinity selections in Fig. 2. The six branches from the third round
(Eluate 3 in Fig. 2) correspond to the six parallel fourth and fifth rounds of selection, leading
to final Eluates 5A-5F (Fig. 2); the procedure (one- or two-step) and amount of receptor
used in each branch are shown.
14 PHAGE DISPLAY LIBRARIES
[ 1]
two-step. Yields in Eluates 4F and 5F (two-step, 10 ng) are not far above
background, alerting us that selection with that amount of receptor may
actually be less stringent than with larger amounts (see
Stringency).

Mutagenesis
The essence of affinity maturation is to mutagenize many clones with
a range of affinities, not just the single best clone (see Introduction). The
mutagenesis method must therefore be able to accommodate many clones
simultaneously (e.g., 150,000 clones were represented in Eluate i of Fig.
2), ruling out methods based on degenerate synthetic oligonucleotides.
Error-prone polymerase chain reaction 11-13 (PCR) is particularly suitable
for this purpose because it focuses mutations on the codons for the displayed
peptide. The PCR template is viral DNA from the previous amplified eluate
(see
Quantifying Yield and Amplifying Eluates). The
product, carrying
abundant base substitutions, is cloned back into the original vector to make
a "mutant library" (e.g., Mutant Libraries 1 and 2 in Fig. 2) in which each
clone from the eluate is represented by a large "clan" of mutants.
In the exemplar experiment, we used a PCR procedure (details in the
next paragraph) in which inosine 5'-triphosphate (ITP) was added to the
reaction mixture. In fact, ITP was mistakenly used for deoxyinosine 5'-
triphosphate,
TM
which might be expected to increase base substitutions by
being incorporated promiscuously by DNA polymerase. The error went
undiscovered because preliminary experiments indicated that ITP pro-
moted all six kinds of base pair substitutions relatively uniformly, although
the reason for this effect is obscure. In any case, as the results of the
exemplar experiment show (next section), ITP-supplemented PCR indeed
introduced abundant substitutions in the phage-displayed peptides.
PCR template is prepared by extracting phage DNA (Table 11) from
80/xl of amplified eluate (-4 x
1012

virions = 20/zg DNA) and dissolving
it in 80/zl water; a typical yield is ~6/zg as estimated by gel electrophoresis.
The reaction mixture contains 5 ng of this DNA and 2/zg each of forward
and reverse primers (Fig. 1) in 500 tzl PCR buffer (Table I) supplemented
with 0.5 mM MnCI2, 0.2 mM each deoxyribonucleoside triphosphate, 0.2
mM ITP, and 50 units/ml
Taq
polymerase (Promega, Madison, WI). The
solution is divided equally into five tubes, overlaid with mineral oil, and
subjected to 20 temperature cycles (1 min at 94 °, 1 min at 50 °, 4 min at
12 R. C. Cadwell and G. F. Joyce, PCR Methods AppL 2, 28 (1992).
a3 R. C. Cadwell and G. F. Joyce, PCR Methods Appl. 3, $136 (1994).
14 H. Gram, L. A. Marconi, C. F. Barbas, III, T. A. Collet, R. A. Lerner, and A. S. Kang,
Proc. Natl. Acad. Sci. U.S.A. 89, 3576 (1992).
[ 1 ] AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 15
70°). The DNA is extracted (Table II) and dissolved in 300/zl TE (Table
I); a typical yield is 5/zg as estimated by gel electrophoresis.
The bulk of the PCR product is digested in 600/xl with 1200 units
BglI,
which cleaves in the invariant flanking sequences and releases a degenerate
insert fragment with overhanging 3' ends that are compatible with the
fUSE5 vector (Fig. 1; other enzymes would be used for vectors with different
cloning sites). The digested fragments are electrophoresed in a 16-mm-
wide well in a 140 × 160 × 1.5-mm 15% acrylamide gel (Table I), which
is stained briefly with 0.5/zg/ml ethidium bromide and illuminated with a
long-wave (366-nm) transilluminator to reveal the digestion products. A
gel piece containing insert fragments of the correct size (60 bp in the
exemplar experiment; see Fig. 1) is excised, and the DNA is electroeluted, ~5
extracted (Table II), and dissolved in 200/.d water; a typical yield is 800 ng.
The entire degenerate insert preparation is ligated as described 9 to 67

/zg SfiI-cleaved vector DNA (Table II); the ligation product is extracted
(Table II) and dissolved in 200/xl TE. The DNA is mixed in 15-txl portions
(nominally containing 5 /zg vector DNA) with 200 /zl electrocompetent
cells (Table II) in an ice-cold 2-mm cuvette (Bio-Rad Laboratories, Rich-
mond, CA); the mixture is shocked by charging a 25-tzF capacitor to 2.5
kV and discharging it through a 400-1) resistor in parallel with the cuvette.
The shocked cells are immediately suspended in 4 ml SOC/Tc (Table I),
shaken at 37 ° for 1 hr, and pooled with cells from two other electroporations
in 1 liter Tc medium (Table I) in a 2.8-liter baffled Fernbach flask. After
spreading 200-/xl portions of appropriate dilutions
(10-1-10 -4) on
Tc plates
(Table I) to determine the number of independent transfectant clones
(4 × l0 s and 7 ×
10 9
in Mutant Libraries 1 and 2 in Fig. 2), the main
cultures are shaken vigorously overnight at 37 ° . All the cultures are pooled,
and phage are partially purified from 1 liter of pooled culture supernatant
(Table II) by two successive PEG precipitations (Table II), with the final
buffer being 10 ml of TBS (Table I); the physical particle concentration is
-5 × 1013 virions/ml and the titer is 0.5-5 × 1012 TU/ml.
Sequence Analysis
A rapid microplate-based sequencing procedure 16 serves as the prelimi-
nary screen for choosing clones to characterize further. When clones are
available in the form of plaques or colonies of infected cells, they are
propagated and virions are PEG precipitated in 96-well microplates using
~s j. Sambrook, E. F. Fritsch, and T. Maniatis, "Molecular Cloning: A Laboratory Manual,"
2nd ed. Cold Spring Harbor Lab., Cold Spring Harbor, NY, 1989.
~6 S. J. Haas and G. P. Smith,
BioTechniques

15, 422 (1993).
16
PHAGE DISPLAY LIBRARIES [ 1]
multichannel pipetters and a rotor that allows whole microplates to be
centrifuged. Viral DNA is extracted simply by adding alkali solution con-
taining end-labeled primer to the phage pellet; the alkali dissolves and
disassembles virions, and after neutralization with acid, the released viral
DNA anneals with primer to form primed template. Alternatively, when
phage are available in solution because they have already been propagated
and processed (Table II), portions are dispensed to wells, and an equal
volume of twofold concentrated alkali/primer solution is added before
neutralization. In either case, aliquots of each primed template are dis-
pensed into a number of wells, to which are added termination mixes
containing T7 DNA polymerase, deoxyribonucleoside triphosphates, and
dideoxy terminators (most often in overlapping combinations to provide
redundant sequence information); after polymerization, sequencing reac-
tions are loaded and electrophoresed side by side in a sequencing gel.
Even larger numbers of clones can be analyzed in a preliminary fashion
by a rapid "one-lane" variant of the microplate method. Here, instead of
dispensing primed template into multiple wells, a single termination mix
containing two dideoxy terminators is added directly to the entire primed
template preparation, and each reaction is electrophoresed in a single lane
of a sequencing gel. The sequence information thus obtained is incomplete
but is more than sufficient to classify clones into clans, each of which clearly
derives from a single parent clone in the initial library. In some cases, one-
lane sequencing also allows dominant mutations within a clan to be recog-
nized.
Ordinarily, only clones from the final round of selection need be ana-
lyzed, but in the exemplar experiment, clones from earlier rounds and from
the parallel conventional selection experiment were also analyzed in order

to document the progress of selection (see Fig. 2). Clones from Eluates 2
and 5A-5F (65-109 each) were classified into clans by one-lane sequencing,
whereas clones from Eluates 1, 2', and 3' (15-42 each) were sequenced
completely. Eight clans were identified altogether, as well as a group of
clones whose displayed peptide could be aligned in various registers with
a 6-mer S-protein-binding motif, FNFE(V/I)(V/I/L/M), that had already
been identified from a library of random hexamers1°'17; clones classified as
"unique" were found only once in a single eluate.
The progression of clan membership during successive rounds of selec-
tion is shown in Fig. 4, and in key respects reflects the expectations of an
in vitro
evolutionary process. Unique clones are prominent in very early
rounds but essentially disappear in later rounds, as expected if they repre-
sent predominantly the background of nonspecifically captured phage. Clan
17 G. P. Smith, D. A. Schultz, and J. E. Ladbury,
Gene
128, 37 (1993).
[ 11
AFFINITY MATURATION OF PHAGE-BORNE LIGANDS
17
_=
.=.
_o
tJ
IClan 2
Clan 3
~ _
'- _ _
Clan
4" -

m
Clan 5
Clan
6" " -
Elan
7- " -
-
- ,,. . :
Clan ~ - - =
~5-mer motif - " = $
1
~ ___
Eluate
FIc. 4. Clan membership in selected eluates from Fig. 2. Each section shows the proportion
of analyzed clones from the various eluates belonging to the indicated clan; the ordinate runs
from 0 to 100%. Also shown in the top section is the frequency of Clan 1 clones with the
L8Q mutation. The parent peptide sequences, when known (Clans 3 and 5 were identified
only by one-lane sequencing), are as follows: Clan 1, NRAWSEFLWQHLAPV; Clan 2,
RNWDLFAVSHMAAV; Clan 4, RWWVSIDGLSFARAV; Clan 6, WHRYQVWRFP-
DFVVL; Clan 7, WRRWFYQFPTPLAVA; Clan 8, CFANFSWGSSDCVL. In Clan 2, one
of the random codons is deleted so the corresponding random peptide has only 14 amino
acids. The 6-mer motif FNFE(V/I)(V/I/L/M) is discussed in the text. Each "unique" clone
appears only once in a single eluate.
4 and the 6-mer motif FNFE(V/I)(V/I/L/M) rise and then fall again; per-
haps these phage have a numerical advantage in the initial library, but
eventually lose the competition to higher-affinity clones. Clan 1 increases
from initial obscurity to prominence in the third round of the conventional
selection experiment (Eluate 3') and overwhelming dominance by the fifth
round of the affinity maturation experiment (Eluates 5A-5F). In all likeli-
hood, the parent of this clan is the tightest-binding clone in Eluate 1, and

possibly the champion of the entire initial library.
18 PHAGE DISPLAY LIBRARIES [ 1]
Clans 6 and 7 seemed at first to be good candidates for dark horses
(see Introduction). They are absent from Eluates 1-3, 2', and 3', but rise
to 31 and 14%, respectively, in Eluate 5C arguably the most stringently
selected phage population (Fig. 3; a single Clan-6 clone also appears in
Eluate 5F). In fact, however, phage capture assays (next section) show that
they do not bind the S-protein receptor specifically (data not shown). A
possible explanation for their prominence in Eluate 5C is that the low
density of immobilized S-protein in the one-step selection of that eluate
and its antecedent, Eluate 4C, permitted these phage to bind unimpeded
to other immobilized species, such as streptavidin or the bovine serum
albumin (BSA) used to block the dishes.
Among the Clan-1 clones, one particular mutation, which could be
readily detected even by one-lane sequencing, was very prominent in the
fifth-round eluates, accounting for 68% of Clan 1 in Eluate 5B and 73% in
Eluate 5E (Fig. 4). This mutation will be called L8Q because it causes a
leucine 0 glutamine substitution at position 8 of the displayed 15-mer.
The rise of this mutation to prominence suggests that it confers an advantage
during affinity selection most likely (although not necessarily) because it
improves affinity for the target receptor.
Sixty-eight clones from Clan 1 (L8Q mutants were deliberately under-
represented, but otherwise clones were chosen essentially randomly from
Eluates 5A-5F) were propagated and processed on the 1.5-ml scale (Table
II), and the coding sequence for the displayed random 15-mer peptide was
determined completely as described earlier. The nucleotide and corre-
sponding peptide sequences are reported in Table IV. The parent sequences
are shown at the top; where a mutant clone differs from the parent, the
mutant residue is shown; at other positions a dot is shown to indicate
identity with the parent. In the nucleotide sequences, silent mutations (i.e.,

mutations that do not change the encoded peptide) are indicated with
lowercase letters. The nucleotide sequences demonstrate that PCR muta-
genesis succeeded in introducing abundant mutations. Most of these, includ-
ing all the silent ones, are scattered more or less randomly among the
clones, suggesting that they confer little or no selective advantage. Two
mutations, however, show evidence of selection in that they are found
repeatedly in many clones: the L8Q mutation that has already been dis-
cussed and the L12H mutation (Table IV). Clones with these mutations,
along with several other clones, were chosen for binding studies (next
section).
Binding Studies
The aim of affinity maturation is to identify high-affinity receptor li-
gands, but it is conceivable that in a close competition phage clones might
[ 11
AFFINITY MATURATION OF PHAGE-BORNE LIGANDS
19
be selected on the basis of slight growth advantages or other subtle traits
unrelated to affinity. For this reason, it is important to evaluate indepen-
dently the affinity of the peptides displayed by the winning phage. Numerous
methods for estimating affinity are available, but this chapter focuses on
two that allow peptides to be studied in the form of whole virions, without
having to synthesize them chemically.
Phage Capture Assay
The phage capture assay (also called "micropanning" 9) is similar in
principle to affinity selection, but is a microplate-based, analytical method
for assessing the binding strength of individual clones rather than a prepara-
tive method for affinity-selecting binding clones from complex phage mix-
tures. In general, high yield in this assay is expected to correlate with high
affinity between receptor and the phage-borne ligand, although the exact
relationship is unknown and undoubtedly complex. The results provide a

basis for choosing a smaller number of clones for more definitive analyses
like inhibition ELISA.
The procedure described here is for a 24-well culture dish (e.g., Falcon
3047, Becton Dickinson, Lincoln Park, N J), but is readily adapted to 96-
well microplates. Wells are coated with 2 ~l of 200/~g/ml streptavidin in
200 ~l TBS (Table I) overnight at 4 ° and are blocked with blocking solution
(Table I) for 2 hr at room temperature. After washing four times with
TBS/Tween (Table I) from a squirt bottle, biotinylated receptor (100 ng
S-protein in the exemplar experiment) in 200/~l of TTDBA (Table I) is
pipetted into the wells. The dish is incubated overnight at 4 ° in a humidified
plastic box and is then washed six times with TBS to remove unbound
receptor. Phage clones ( 5 x 109 virions = 0.5-5 × 108 TU) in 100/zl
TTDBA are added to a set of wells; if desired, a known receptor ligand
can be added to a second set of wells before adding phage to see if it
competitively inhibits phage capture. After a 4-hr incubation at 4 °, wells
are emptied by aspiration and washed 10 times with TBS/Tween. Elution
buffer (100/xl; Table I) is pipetted into each well, the dish is incubated for
10 rain at room temperature to elute phage, and the eluates are transferred
to microtubes containing 19/~1 of 1 M Tris-HC1 (pH 9.1) to neutralize the
acid in the elution buffer. These eluates (output phage) are titered (Table
II) along with dilutions of input phage in order to quantify yield.
Figure 5 shows the phage capture results for 11 clones from Table
IV. Clone 88, which has the predominant L8Q mutation, seems to give a
significantly higher yield than the other clones; in all cases, capture was
strongly inhibited by 1/~M S-peptide, as expected. It is possible that discrim-
ination among clones would have been improved if the surface density of
receptor molecules in the microplate wells had been reduced. Clones 88
20 PHAGE DISPLAY LIBRARIES [ 1 ]
TABLE IV
NUCLEOTIDE AND CORRESPONDING AMINO ACID SEQUENCES FOR SELECTED

CLAN-1 CLONES
Peptide Number
Nucleotide sequence a sequence" of clones
AATCGGGCTTGGTCTGAGTTTCTGTGGCAGCATCTTGCGCCTGTT NRAWSEFLWQHLAPV
a
a ,
c
C
a
c
a
C
t c c a
c c C
G. .c D
G GA C E
A c Q
.GC g S
a G G
A C T R
A R
a a c.A c H
A H
.C A T H
A C H
a cAA A K H
c A H
t AC H
G a A G H
a.C A

13(Parent)
4
1
1
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
1
1
1
1
1
l(Clone 72)
1
[ 1]
AFFINITY MATURATION OF PHAGE-BORNE LIGANDS
21

TABLE IV
(continued)
Peptide Number
Nucleotide sequence s sequence" of clones
C h
A C Q A
A G Q.R
A G a Q.R
.G A S Q
.G c A S Q
t A Q
A c Q
h Q
g A
a h
h a
G h
G A a c
C G A
a.T A A Ca
c.T a V
T V
G A T D.T V
Q
Q
Q
.C Q
.A Q
.A Q
V.Q, ! A

1
2
2
1
1
1
1
1
4
1
1
1
1
1
1
l(Clone 88)
1
1
1
a Periods signify nucleotides or amino acids that are the same as at the corresponding
position of the parent sequence Lowercase letters in the nucleotide sequences
are silent mutations, which do not change the encoded peptide
and 72, along with the parent peptide, were chosen for analysis by inhibition
ELISA (next subsection)
Inhibition ELISA
Inhibition ELISA determines the affinity of a test peptide for receptor
by measuring the ability of various concentrations of the peptide in solution
to competitively inhibit binding of receptor to an immobilized ligand. If
affinity is high enough (dissociation equilibrium constant KD < 1 /xM),
peptides can be analyzed in the form of whole virions; otherwise (or in

22
PHAGE DISPLAY LIBRARIES [ 1]
0.25~
0.15-~
i
o.oo~ i
<
c~
Z
C~
E
iiii ,
H
c~ c~ c~ ~.
no S-pepfide
S-pepfide
Fic. 5. Phage capture assay of 11 clones from Table IV.
addition), peptides can be synthesized chemically and analyzed free in so-
lution.
A known ligand for the receptor most often, 5 x 101° virions of one
of the affinity-selected clones in 50/zl TBS (Table I) is adsorbed to wells
of a 96-well microplate for 3 hr at room temperature. At the same time,
10 wells in the same microplate are coated with 50-/zl portions of biotiny-
lated BSA standards (0-40.5 ng/ml in TBS containing 100 tzg/ml nonbiotin-
ylated BSA; Table I). Because all the standard wells receive the same total
concentration of BSA, the fraction of the input biotinylated BSA that
becomes irreversibly immobilized to the plastic surface should be constant
from well to well. Therefore, the amount of biotinylated BSA immobilized
in each well is presumably directly proportional to the input concentration
of biotinylated BSA. All wells are blocked at least 1 hr at room temperature

with 5% nonfat dry milk in TBS and washed five times with TBS/Tween
(Table I) on a plate washer to remove unbound material. The biotinylated
BSA standard wells receive 190 /zl TTDBA (Table I), whereas ligand-
coated wells receive 190/xl TTDBA containing a constant concentration
of biotinylated receptor and graded concentrations of inhibitor peptides,
either free or in the form of virions (phage are propagated on a l-liter
scale and purified by CsCI density gradient centrifugation; see Table II).
The receptor concentration is fixed at a level just sufficient to give a work-
able ELISA signal (15-30 mOD/min; see below) without inhibitor, as