combinatorial chemistry, part a

[ 1] AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 3 [11 Affinity Maturation of Phage-Displayed If the sequences are capable of heritable mutation--phage display and random R N A and D N A

P r e f a c e

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 R N A 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

C o n t r i b u t o r s to V o l u m e 2 6 7

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

If the sequences are capable of heritable mutation phage display and random R N A and D N A 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

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 M o t Biol 149, 313 (1981)

7 j K Scott and G P Smith, Science 249, 386 (1990)

[ 1] AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 5

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 D N A 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

[ Mutantlibrary 1 1

,~ l-step selection with I lag receptor

~ mutagenesis

~ l-step selection with I lag receptor

ng ~4¢ 2-step

lO ng

,-step ~ 1-step ~ 1-step ~ 2-step ~ 2-step % 2-step

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

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 ( - 5 0 0 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

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 b a c k g r o u n d are p r o b a b l y t o o stringent to be useful I n the e x e m p l a r

e x p e r i m e n t , f o r instance, six d i f f e r e n t c o n d i t i o n s f o r r o u n d s 4 a n d 5 w e r e tried, yielding final eluates 5 A - 5 F (Fig 2)

Capture via Biotinylated Receptor

I f r e c e p t o r p r o t e i n is available in relatively p u r e form, it is c o n v e n i e n t

to b i o t i n y l a t e it at accessible e - a m i n o g r o u p s This allows it to be rapidly

a n d irreversibly c a p t u r e d o n s t r e p t a v i d i n - c o a t e d petri dishes u n d e r n o n d e -

n a t u r i n g c o n d i t i o n s a n d also facilitates E L I S A (see B i n d i n g Studies) (Nu-

m e r o u s alternative i m m o b i l i z a t i o n m e t h o d s are available, b u t will n o t be discussed here.) I n a typical p r o t o c o l , 1 0 - 4 0 / z g p r o t e i n is r e a c t e d with

5 0 - 4 0 0 / x M s u l f o s u c c i n i m i d y l - 6 - ( b i o t i n a m i d o ) h e x a n o a t e ( N H S - L C - b i o t i n ; Pierce C h e m i c a l Co., R o c k f o r d , I L ) in 4 4 / x l o f 0.1 M N a H C O 3 ; residual

[ 1] AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 9

Solution or

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

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

Details in Smith and Scott?

b Details in Sambrook et al 15

l i b r a r y , o r a 100-/.d p o r t i o n o f t h e e l u a t e f r o m t h e p r e v i o u s r o u n d o f affinity

10 PHAGE DISPLAY LIBRARIES [ 1 ]

TABLE II STANDARD PROCEDURES

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 A c i d s 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

T A B L E III

Escherichia coli STRAINS

MC1061" F hsdR rncrB A(araABC- Uninfectableb; streptomy-

1eu)6779 araD139 Alac174 cin resistant

galU galK strA thi thi

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

m k h 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 T F D B A (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 P H A G E 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 T T D B A (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 5 A - 5 F 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 M A T U R A T I O N OF P H A G E - B O R N E 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

1 4 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 D N A 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 D N A 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 D N A (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 ( 1 0 - 1 - 1 0 -4) o n 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 D N A 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

D N A 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

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, N R A W S E F L W Q H L A P V ; Clan 2,

R N W D L F A V S H M A A V ; Clan 4, R W W V S I D G L S F A R A V ; Clan 6, W H R Y Q V W R F P -

D F V V L ; Clan 7, W R R W F Y Q F P T P L A V A ; Clan 8, C F A N F S W G S S D C V L In Clan 2, one

of the r a n d o m codons is deleted so the corresponding r a n d o m peptide has only 14 amino acids The 6-mer motif F N F E ( V / I ) ( V / I / L / M ) is discussed in the text Each " u n i q u e " 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 A F F I N I T Y M A T U R A T I O N OF P H A G E - B O R N E L I G A N D S 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 T T D B A (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

T T D B A 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

2 0 PHAGE DISPLAY LIBRARIES [ 1 ]

T A B L E IV NUCLEOTIDE AND CORRESPONDING AMINO ACID SEQUENCES FOR SELECTED

1

[ 1] AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 21

T A B L E IV (continued)

Peptide N u m b e r Nucleotide s e q u e n c e s s e q u e n c e " of clones

and 72, along with the parent peptide, were chosen for analysis by inhibition ELISA (next subsection)

Inhibition E L I S A

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]

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 T T D B A (Table I), whereas ligand- coated wells receive 190/xl T T D B A 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

[ 1] AFFINITY M A T U R A T I O N OF P H A G E - B O R N E LIGANDS 23 determined in preliminary titration experiments After equilibration over- night at 4 °, wells are washed rapidly five times with TBS/Tween (Table I)

in a plate washer (total wash time - 2 5 sec) All wells, including those with the biotinylated BSA standards, are immediately reacted for 30 rain at room temperature with 65 txl of 5/zg/ml alkaline phosphatase-conjugated streptavidin (AP-SA; Table I) in AP-SA diluent (Table I); washed with TBS/Tween thoroughly 10 times on the plate washer; and filled with 100 /zl of NPP substrate solution (Table I) The difference between the optical density (OD) at 405 and 490 nm is read at 3-min intervals over a 60-min period on a kinetic plate reader in order to obtain a slope (mOD/min) for each well (for OD differences up to 1.5, time dependence is linear, with correlation coefficients exceeding 0.995)

The dependence of slope on input concentration for the biotinylated BSA standards is modeled by a cubic equation, which is used in turn to transform the slopes for the other wells to equivalent relative amounts of biotinylated protein captured, which we call Y The Y values are averaged for wells containing no inhibitor to give the maximum value, Ymax ; and percentage inhibition for the other wells is calculated as 100(Ymax - Y)/

Ymax •

Theoretical inhibition curves can be calculated assuming unbound re- ceptors are simultaneously in equilibrium both with receptors bound to inhibitor and with receptors bound to immobilized ligand The curves are governed not only by the parameter of interest the Ko of the inhibitor for the receptor but also by two nuisance parameters concerning the immobilized ligand that are not generally known: its effective concentration (the amount actually available for reaction with receptor divided by the reaction volume) and its Ko for the receptor Under a broad range of conditions, however, the inhibition curve is insensitive to the nuisance parameters and is sensitive to the inhibitor KD, and a value of the latter that brings the curve into accord with the data should be close to reality (assuming the underlying kinetic model is realistic) The effect of varying the parameters within their plausible ranges can be checked by computer Figure 6 shows results of an inhibition ELISA for the Clan-1 parent and Clan-1 mutant Clones 88 and 72; also included for comparison is S-peptide and a previously studied 17 hexapeptide ligand, YNFEVL, match- ing the 6-met motif FNFE(V/I)(V/I/L/M) (see Sequence Analysis; tyrosine

is substituted for phenylalanine at the first position in some clones with this motif) S-peptide was available only as free peptide, but the other four were tested both in the form of phage and as chemically synthesized free peptides Best-fitting theoretical curves are shown in solid lines, each labeled with the corresponding KD For the tightest-binding inhibitors Clone-88 phage and perhaps S-peptide the fit of the theoretical curve is insensitive

2 4 PHAGE DISPLAY LIBRARIES [1]

- 1 5 0 ligands/virion (5 × 101° virions in 50/~1 TBS per well for 3 hr at room temperature) The concentration of biotinylated S-protein 17 was 2 nM, and inhibitor peptides were present

at concentrations indicated on the abscissa Phage inhibitors (black squares) were assumed

to display five intact peptides/virion one fused to each plII coat-protein molecule Free peptide inhibitors (shaded squares) were either purchased (S-peptide; Sigma Chemical Co.)

or synthesized chemically The parent peptide and mutant peptides 88 and 72 have an acetylated

N terminus and an amidated C terminus; S-peptide and the YNFEVL 6-mer 17 have free N and C termini S-peptide was quantified by amino acid analysis, the others spectrophotometri- cally [H Edelhoch, Biochemistry 6, 1948 (1967)] Solid lines are best-fitting theoretical inhibi- tion curves whose corresponding KD values are indicated Theoretical inhibition curves (solid lines) were calculated assuming that the Kt) for binding of receptor to the immobilized peptide (one of the nuisance parameters) was 100 nM, a value consistent with previous affinity measurements, and that the effective concentration of immobilized peptide (the other nuisance parameter) was 10 n M (in the 190-~1 reaction volume), - 1 5 % of the theoretical maximum

if all 2 ~g of phage used to coat a well was immobilized and available for reaction For all inhibitors but Clone-88 phage and (possibly) S-peptide, the curve's fit to the data is sensitive

to the inhibitor KD and insensitive to the two nuisance parameters

in affinity for the target receptor

to explain this result will help illuminate how affinity maturation works

In the early stages of affinity maturation, a dark horse clan and the clan

of the initial champion expand and mutate in a sort of race toward their respective local optima Although by definition the dark horse's optimum has higher affinity than the initial champion's, that by no means guarantees that the dark horse clan will win For the dark horse starts with a selective disadvantage: its clan expands more slowly in the population than the initial champion's at first, slowing the exploration for affinity-enhancing mutations The dark horse clan thus has limited opportunity to reach a mutant with sufficient selective advantage to reverse the clan's ultimate decline to obscurity in the population

Previous work on how S-protein binds S-peptide hints that dark horse iigands for this receptor may be hard to find Four residues of S-peptide F8, Q l l , H12, and M13 are deeply buried in its a-helical-bound form (Fig 7) They lie close together on one face of the helix, and the first three seem to be critical for binding in that they are strictly conserved in ribonucleases from 41 other species (Fig 7) The fourth buried position may not be quite as critical: the methionine found there in bovine fibonuclease is substituted with isoleucine and valine in other species, and synthetic S-peptide analogs with isoleucine, valine, or leucine at this position bind S-protein as strongly as does S-peptide itself (Fig 7) Still, other substitu- tions greatly weaken binding (Fig 7), pointing to its importance in the interaction Taken together, this work suggests FxxQH(M/L/I/V) as a criti- cal a-helical-binding motif This motif is preserved in the Clan-1 parent

26 PHAGE DISPLAY LIBRARIES [ 1]

to the parent The invariant NH2-AIa-Asp-GIy-Ala that precedes the 15-residue random peptide in all phage clones (see Fig 1) is shown in lowercase letters The four "buried"

residues of S-peptide have solvent-accessible surfaces areas of 10 ~z or less in the complex

with S~rotein, whereas the other S-peptide residues have an average accessible surface area

of 48 A 2 [E E Kim, R Varadarajan, H W Wyckoff, and F M Richards, Biochemistry 31,

12304 (1992); F M Richards, H W Wyckoff, J L Mouning, and J W Schilling, in "Atlas

of Molecular Structures in Biology" (D C Phillips and F M Richards, eds.), Vol 1 Oxford Univ Press (Clarendon), Oxford, 1973] 20 Substitutions in ribonucleases from 41 other species are shown [J J Beintema, W M Fitch, and A Carsana, Mol BioL Evol 3, 262 (1986); J J Beintema, Life Chem Rep 4, 333 (1987)], 1° as are substitutions in synthetic S-peptide analogs that bind S-protein well (KD < 333 nM at 25 ° and pH 6) or poorly (KD > 100/~M) [P R Connelly, R Varadarajan, J M Sturtevant, and F M Richards, Biochemistry 29, 6108 (1990)]

selected from the random peptide library, as well as in mutant Clones 72 and 88 (Fig 7), with one exception: at the last motif position, Clone 72 has histidine, which is not one of the substitutions that have been studied in synthetic S-peptide analogs For the sake of concreteness in what follows, the critical binding motif will be assumed to be FxxQH(M/L/I/V/H); amendments compelled by future data will make no substantial difference

to the argument At all but one of the nonmotif positions (including the two designated "xx" shown earlier), the Clan-1 peptides differ from S-peptide (Fig 7); apparently many different combinations of amino acids

at these positions are compatible with strong binding a finding with im- portant implications, as we shall see

Theoretically, the initial 2 × 108 clone library should contain some 1500 clones with the FxxQH(M/L/I/V/H) motif -even more if the same motif

in other registers is considered If nonmotif positions play only a supporting role, for which many different combinations of amino acids are equally suited, it is not surprising that some of these 1500 clones, including the

[ 1] AFFINITY MATURATION OF PHAGE-BORNE LIGANDS 2 7 Clan-1 parent, have good affinity for the receptor By the same token, the local optima near all or most of these clones are plausibly nearly equivalent

to one another in affinity In these circumstances, the local optimum in the neighborhood of a dark horse (if there is one) will typically have only a marginal selective advantage over the local optimum in the neighborhood

of the initial champion too marginal, perhaps, to overcome the initial selective disadvantage of the dark horse itself

If the requirements at the nonmotif positions were somewhat more stringent, the results of the exemplar experiment might have been altogether different In that case, perhaps none of the 1500 initial clones that (theoreti- cally) match the FxxQH(M/L/I/V/H) motif would have had as high affinity

as the best of the clones with the 6-mer motif FNFE(V/I)(V/I/L/M), which abounded in Eluate 1 from the 15-met library (Fig 4) and predominated

in selections from a random hexapeptide library, t°'17 Yet, provided the requirements at the nonmotif positions were not too stringent, one of those

1500 clones might have acquired one or two mutations in the course of affinity maturation that gave it a strong selective advantage over all FNFE(V/I)(V/I/L/M) clones Affinity maturation would have revealed a dark horse, and with it a superior binding motif entirely different from the motif in the initial champion Without affinity maturation, such discoveries would not be possible

No matter how good the ligand revealed by affinity maturation, there

is always the possibility that somewhere in the vast reaches of sequence space a much better ligand remains to be discovered Affinity maturation can only explore the close neighborhoods of receptor-binding clones that happen to be present in the initial library Still, given the utter impossibility

of a truly global search, those neighborhoods seem the most promising territory in which to concentrate our limited resources

Acknowledgments

Supported by U.S Army Research Office Grant DAAL03-92-G-0178, Department of Health and Human Services Grant GM41478, and the University of Missouri Molecular Biology Program Our collaborators David Schultz and John Ladbury did the conventional affinity selection experiment and the structural analysis summarized in Fig 7 We thank Robert Davis for excellent technical assistance

The generation of a display phage library involves several steps A

"display gene" is either cloned or constructed using synthetic oligonucleo- tides and is inserted between the DNA encoding a signal peptide and the structural gene of gene VIII or III of a filamentous bacteriophage The resulting parental display phage is then used as a framework for the propa- gation of a phage display library by utilizing variegated (mutated) synthetic DNA, which is inserted into a preselected (directed) portion of the display gene Each display variant in the library is physically linked, via fusion to the coat protein encapsidating the bacteriophage DNA, to its own encoding display gene For this reason, the selection of binding variants and the ready determination of the predicted amino acid sequence of the variants from D N A analysis enable rapid and effective screening of a library against

a number of targets Such screening can be performed in a cyclic and iterative fashion such that a small subpopulation of display variants with the highest affinity for the target are selected (test tube evolution) This chapter describes, in detail, the generation of a bacteriophage- display library based on a protease inhibitor and the selection from this library of a series of variants with high affinity and specificity for different

1 G P Smith and J K Scott, Methods Enzymol 217, 228 (1993)

2 M A McLafferty, R B Kent, R C Ladner, and W Markland, Gene 128, 29 (1993)

3 G Winter, A D Griffiths, R E Hawkins, and H R Hoogenboom, Annu Rev Immunol

12, 129 (1994),

4 D R Corey, A K Shiau, Q Yang, B A Janowski, and C S Craik, Gene 128, 129 (1993)

5 W Markland, B L Roberts, M J Saxena, S K Guterman, and R C Ladner, Gene 109,

13 (1991)

Copyright © 1996 by Academic Press, Inc

[2] SELECTION FOR PROTEASE INHIBITORS 2 9

human proteases of potential pharmaceutical value We demonstrate that such a process is timely and precise and that the iterative quality of the operation maximizes selection so as to derive the highest affinity molecules

LB: 10 g tryptone, 5 g yeast extract, 10 g NaC1 per liter

SOC: 20 g tryptone, 5 g yeast extract, 1.8 g glucose, 10 mM NaCI, 2.5

mM KCI, 10 mM MgC12, 10 mM MgSO4 per liter

YND: 6.7 g yeast nitrogen base, 5 g ammonium sulfate, 5 g casamino acids, 182 g sorbitol per liter

All components obtained from Difco Labs (Detroit, MI)

Plasmin substrates: N-p-Tosyl-Gly-Pro-Lys-p-nitroanilide (Sigma,

St Louis, MO) and N-t-Boc-Val-Leu-Lys-7-amido-4-methylcouma- rin (Sigma)

3 0 PHAGE DISPLAY LIBRARIES [2]

T h r o m b i n reaction buffer: 50 m M Tris (7.5), 150 m M NaCI, 0.1% (w/v) P E G , 0.05% Triton X-100

T h r o m b i n substrates: Benzoyl-Phe-Val-Arg p-nitroanilide (Calbio- chem, La Jolla, CA) and N-t-Boc-Val-Pro-Arg-7-amido-4-methyl- coumarin (Sigma)

Kallikrein reaction buffer: 20 m M Tris (7.5), 150 m M NaC1, 0.1% P E G ,

1 m M E D T A

Kallikrein substrates: N-Benzoyl-Pro-Phe-Arg-p-nitroanilide (Sigma) and Pro-Phe-Arg-7-amido-4-methylcoumarin (Sigma)

All substrates are stored as frozen aqueous stocks (10 mM)

P C R Reagents and Primers

P C R (polymerase chain reactions) are u n d e r t a k e n using reagent kits commercially available (Perkin-Elmer Cetus) and the G e n e A m p P C R System 9600 T h e P C R primers are located at the 5' and 3' ends of the gene insertion site within the III gene of M13 and are as follows:

3 P C R U P : 5' C G G C G C A A C T A T C G G T A T C A A G C T G 3'

3PCRDN: 5' C A T G T A C C G T A A C A C T G A G T T T C G T C 3'

M e t h o d s

Parental Gene: Design and Construction

In our previous phage-display work with a Kunitz molecule, 5-7 we have used bovine pancreatic trypsin inhibitor (BPTI) F o r further work in this field we have decided to choose a h u m a n Kunitz molecule While there is

no known directly homologous molecule to B P T I in humans, there are several molecules which contain Kunitz homology domains, e.g., inter- a-trypsin inhibitor (ITI), the amyloid protein, and lipoprotein-associated coagulation inhibitor (LACI)

L A C I , also known as tissue factor pathway inhibitor (TFPI) or extrinsic pathway inhibitor (EPI), is involved in the extrinsic pathway of coagulation

T h e predicted primary sequence of L A C I s indicates three t a n d e m domains with homology to Kunitz-type protease inhibitors, s It has been p r o p o s e d that Kunitz domain 2 is required for efficient binding to factor Xa and that

6 B L Roberts, W Markland, A C Ley, R B Kent, D W White, S K Guterman, and

R C Ladner, Pro¢ Natl Acad Sci U.S.A 89, 2429 (1992)

7 B L Roberts, W Markland, K Siranosian, M J Saxena, S K Guterman, and R C Ladner,

Gene 121, 9 (1992)

8 T J Girard, L A Warren, W F Novotny, K M Likert, S G Brown, J P Miletich, and

G J Broze, Jr., Nature (London) 338, 518 (1989)

[2] SELECTION FOR PROTEASE INHIBITORS 31

W, tryptophan; and Y, tyrosine

domains 1 and 2 are required for inhibition of factor VIIa/tissue factor activity, while the function of Kunitz domain 3 is uncertain 9

The LACI domain 1 (LACI-D1) gene is designed (Fig 1) (based on the published c D N A sequence) to contain unique restriction enzyme sites

9 G J Broze, Jr., T J Girard, and W F Novotny, Biochemistry 29, 7539 (1990)

32 PIJAOE DlSeLAV LmRARIZS 121

at positions which fulfill the following requirements: (a) enable the initial cloning of the synthetic gene into the display phage vector (MAEX.IIIT), (b) enable the replacement of segments of the LACI gene with variegated (mutating) synthetic DNA, and (c) allow for the presence of diagnostic restriction enzyme sites, of value in the verification of constructs and during library generation and screening The DNA sequence is designed to max- imize codon usage in E coli The LACI-D1 gene is assembled from four synthetic oligonucleotides to give a final synthetic gene containing a 5'

EagI overhang and a 3' KasI overhang This is ligated into the prepared display vector

The display phage vector (MAEX.III 7) is a III-display system of the polyvalent type in which the fusion-III protein is constitutively expressed

It is based on an M13 vector into which an ampicillin resistance gene has been cloned at the intergenic region The vector is cleaved in two stages with KasI and EagI Following complete cleavage, the restricted vector is precipitated using standard methods, and 100-/A volume ligations are set

up using T4 ligase and ligase buffer such that the prepared vector and gel- purified assembled synthetic oligonucleotides are at ratios of 1 : 0, 1 : 5, 1 : 10, and 1:20 The ligation mixtures are electroporated into E coli cells (XLI- Blue; Stratagene) and plated for plaques A number of plaques are picked and restreaked for temporary storage Plaques from these plates are ana- lyzed using PCR techniques for the presence of an insert of the correct size, containing the exact number and type of restriction enzyme sites expected from the gene design D N A sequence analysis of the clones showed them to be correct

Functional Assay of Displayed LACI-D1

The presence of a novel molecule fused to the III gene product (espe- cially in a polyvalent system) can be frequently intimated by a change in plaque morphology; usually the plaque size is smaller At its most extreme, such a fusion product causes the complete loss of phage infectivity, in which case a gene VIII product bacteriophage display system or a monovalent (phagemid) display system should be considered as an alternative The possession of polyclonal or monoclonal antibodies specific for the displayed molecule or domain can be used to immunoprecipitate the display phage or affect the infectivity of the display phage Such experiments, with suitable controls, will indicate the display of the appropriate immunoreac- tive species A true functional assay is the ideal assessment of the correct display and accessibility of the molecule of interest The display of protease inhibitors naturally lends itself to a functional assay related to the binding

of the displayed inhibitors to their cognate proteases LACI-D1 should

[2] SELECTION FOR PROTEASE INHIBITORS 33 bind to a trypsin-like protease Such binding (to trypsin-coated wells or agarose-immobilized trypsin) should be significant when compared to a nondisplay phage (e.g., M13) or a Kunitz-display phage with a different specificity (e.g., the human neutrophil elastase binder EPI16) Such an experiment with trypsin agarose beads (Pierce, Rockford, IL) (see general phage-binding section) compared MAEX.III (nondisplay phage) with LACI-DI.III and BPTI.III (bovine pancreatic trypsin inhibitor phage, a known trypsin binder) The assay demonstrates that the LACI-DI.III dis- play phage has a significant binding capability relative to the nondisplay phage and a binding to trypsin comparable to that of the BPTI.III dis- play phage

Phage Display Library: Design and Construction

Protein variants are generated by the introduction of variegated syn- thetic oligonucleotide duplexes into a suitably prepared parental gene vec- tor, in this instance LACI-DI.III The library was designed to be made in two stages: phase I and phase II The phase I region encompasses the DNA encoding the P1 region of LACI-D1, the part of the molecule known to interact directly with the target protease (variegation scheme shown in Fig 2a) The phase II region is the DNA encoding the loop of amino acids underlying the main P1 loop of the inhibitor, considered to have an indirect affect on the binding affinities of mutant inhibitors (variegation scheme shown in Fig 2b) The residues chosen to be variegated or fixed were selected by means of sequence homologies contained within the family of Kunitz molecules; it is also known that, in the case of BPTI, residues 10

to 20 and 31, 32, and 34 to 40 are in direct contact with residues contained within the active site of trypsin

The degree of variation at any one position was chosen such that the level of predicted amino acid diversity is maximized relative to the level

of codon redundancy The presence of stop codons should be prevented where possible; when this is not possible, the use of suppressor strains should be considered

A brief overview of the library construction will be given Phase I varie- gation is achieved by ligating synthetic oligonucleotide duplexes with NsiI-

and MluI-compatible ends (Fig 2a) into the cleaved replicating form (Rf) DNA of the parental vector (LACI-DI.III) The resultant ligated material

is electroporated into E coli and plated onto LB.Amp plates to obtain phage- producing ampicillin-resistant colonies from which the display library was recovered To incorporate phase II variegation, synthetic oligonucleotide du- plexes with MluI- and BstEII-compatible ends (Fig 2b) are ligated into Rf DNA derived from one of the following: (i) the parental construct; (ii) the

34 PHAGE DISPLAY LIBRARIES [2]

phase I library, making a combinatorial library, i.e., the summation of the two partial libraries; or (iii) the display phage selected from the phase I screening of the target protease The third possibility will be described here The variegation scheme for phase I allows for 6.6 × 105 D N A sequences (3.1 × 105 protein sequences) and for phase II 4096 D N A sequences (1600 protein sequences), giving a combined total of 2.7 × 108 D N A sequences

[2] SELECTION FOR PROTEASE INHIBITORS 35 (5.0 × 10 v protein sequences) In the project described here, since previously selected display phage were used as the origin of Rf D N A for the generation

of the Phase II library, the final level of target accessible variegation is probably in the range of 105 to 10 6 variants

Variegated Heteroduplex Generation

Since large sections of the synthetic oligonucleotide used in the genera- tion of the phage display library will be variegated (mutated), it is not possible to form a simple homoduplex As an alternative, synthetic oligonu- cleotide heteroduplexes can be formed in one of several ways In the case

of the phase I library described here, a variegated partial heteroduplex was formed by annealing two "bottom" primers located at the 5' and 3' termini (see Fig 2a) to the complete "top strand" synthetic oligonucleotide, such that the correct 5' NsiI and 3' MluI overhangs were formed Note that no

"fill-in" reactions are undertaken, hence the final product, which is ligated into the prepared vector, is a partial heteroduplex

Preparation o f L A CI-DI.III Vector

The restriction enzyme cleavage of the parental vector is performed using standard molecular biological techniques Twenty micrograms of LACI-DI.III Rf D N A is cleaved with NsiI in a 100-~1 reaction at 37 ° for

2 hr Complete cleavage of the vector is determined MluI is added to the

same tube (Note: this enzyme cuts well in the NEB NsiI buffer.) Complete

digestion is again determined Dephosphorylation of the cut vector is per- formed to ensure the lowest background possible for ligation, i.e., removal

of the wild-type LACI-D1 display phage which has been only singly cut and which has, after all, the highest ligation efficiency This is important when generating a phage display library since the background will be all wild-type LACI-Dl-display phage while the remainder of the library con- tains the large diversity of the library Too large a background results in the considerable outnumbering of any one variant by the parental molecule which may be significant in instances with a particular target

Trial Library Ligation

Ligation, on a small scale, using the components derived in the previous sections, is performed to optimize the vector to insert ratio in the ligation reaction The cut/dephosphorylated vector (0.5 /~g) is added to insert at molar ratios of 1 : 0, 1 : 1, 1 : 5, or 1 : 10 in a 25-/A volume containing 1 /~1

of T4 ligase (NEB; 200,000 units per ml) and is incubated at 16 ° overnight Dilutions of the ligation mixtures, together with suitable controls, are elec-

36 P H A G E D I S P L A Y L I B R A R I E S [21 troporated into an E coli strain (XL1-Blue) and plated for ampicillin- resistant phage-producing colonies or alternatively as plaques in a lawn of XL1-Blue cells PCR analysis of the colonies and plaques is performed; namely restriction enzyme analysis of the PCR products The parental LACI-D1 gene contains unique BspHI and RsrlI recognition sites whereas the library variant genes lost both sites If the library background level is too high (>5 to 10%, depending on the library size), a postligation restriction enzyme cleavage of the ligation mixture with the just-described two enzymes should reduce it to more acceptable levels

A comparison of the just-described analyses for plaques and colonies derived from the same ligation can give useful indications of infectivity problems associated with members of the phage display library A skewing

in favor of the wild-type display phage relative to the variant display phage

or a disparity in sequences associated with either colony or plaque-derived DNA is likely to indicate an infectivity problem

Library Ligation and Large-Scale Transformation

Based on the results obtained from the previous section, a large-scale ligation was set up using 5-10/zg of cut and dephosphorylated vector to which an insert is added at the predetermined molar ratio in a ligation volume of 250/zl containing 5/zl T4 ligase (NEB; 200,000 units per ml)

A postligation cut with BspHI and RsrlI was performed

The potential for differential infectivity of the components of a phage display library, leading to a skewing of the library composition, can be reduced by plating the library as ampicillin-resistant, phage-producing colo- nies instead of as plaques in a lawn of bacterial cells, which require rounds

of infectivity or by phage production in a liquid culture, where such skewing could be even more pronounced

The number of potential independent transformants can be estimated from the trial ligation or by transforming a small aliquot of the large-scale library ligation The large bioassay dishes (245 x 245 mm), available from Nunc, Inc (Naperville, IL), are ideal for plating out the transformed cells from a library ligation Such dishes require 250 ml of LB agar (plus 50/zg per ml of ampicillin) and are large enough to plate approximately 2 million transformants/transductants per plate The number of plates can be esti- mated from previous determinations and can be made in advance

An F- strain of E coli (XL1-BlueMR, Stratagene) was used for the large- scale transformation to prevent the likelihood of spurious bacteriophage infection during cell growth and preparation or in the posttransformation grow out It should also prevent potential infectivity skewing of the library during the biological amplification of the transformants

[2] SELECTION FOR PROTEASE INHIBITORS 37 The total processed library ligation is split into two halves Each portion

is electroporated on separate days using freshly prepared electrocompetent cells as a safeguard against poor electrocompetent cell preparation or con- tamination Positive and negative controls are included Half the library ligation is added to 400/zl of fresh electrocompetent XL1-BlueMR cells and left on ice for 10 min This is split into 12 separate prechilled cuvettes (approximately 35 /.d per cuvette) ready for electroporation (Bio-Rad, Richmond, CA, Gene Pulser; 1.7 kV, 25/zF) The time constant should be between 18 and 19 msec when the pulse controller is set at 800 ohm Each pulsed sample is taken up in 1 ml SOC and the individual electroporations are combined (12 ml total) Aliquots of the combined electroporations, together with controls, are taken, diluted, and titered for plaque-forming units (pfu) on LB agar plates in a lawn of XL1-Blue cells This is a measure

of the number of independent transformations achieved and should be at least twice the designed size of the library The remainder are grown in a shaker/incubator at 37 ° for 30 min Aliquots are taken and titered for colony-forming units (cfu) by plating on LB agar ampicillin (50/xg per ml) plates A measure of the transduction efficiency of the operation can be determined by comparing the number of cfu relative to the known indepen- dent transformants measured as pfu The remainder of the library transfor- mation grow out is plated on the large Nunc LB agar Amp (50/xg per ml) plates Two such plates, 6 ml of grow out per plate, are used All platings are incubated at 37 ° overnight

Plaque and colony counts are performed the next day A suitable num- ber of colonies and/or plaques is picked for PCR analysis and sequencing,

as previously described

Processing of Transduced Colonies to Generate a Phage Stock

Retrieving the display phage from the ampicillin-resistant colonies is a simple matter of suspending the bacterial cell/bacteriophage mixture in a suitable medium, followed by removal of the cells and precipitation of the display phage LB is added to the colony-containing plates (50 ml of LB per Nunc plate), and the colonies are scraped with a glass rod spreader This suspension is placed on a shaker at room temperature for 30 min followed by centrifugation and the addition of one-quarter volume of phage precipitation solution The bacteriophage are pelleted by centrifugation, redissolved in 20 ml of phage high salt resuspension buffer, the bacteria and debris spun out, and the bacteriophage reprecipitated The final pellet

is resuspended in LB (5 to 10 ml) and azide is added to a final concentration

of 0.05% The display library phage stock is titered and stored at 4 ° Titration data from the bacterial transformations and the final library