STRUCTURAL AND MECHANISTIC DETERMINANTS OF ......P1: RHS/PLB P2: SDA/PLB QC: SDA/PBS/AGR T1: PBS...

P1: RHS/PLB P2: SDA/PLB QC: SDA/PBS/AGR T1: PBS

April 5, 1997 11:59 Annual Reviews AR031-02 AR31-02

Annu. Rev. Biophys. Biomol. Struct. 1997. 26:27–45Copyright c© 1997 by Annual Reviews Inc. All rights reserved

STRUCTURAL AND MECHANISTICDETERMINANTS OF AFFINITYAND SPECIFICITY OF LIGANDSDISCOVERED OR ENGINEEREDBY PHAGE DISPLAY

Bradley A. KatzArris Pharmaceutical Corporation, 385 Oyster Point Boulevard, Suite 3,South San Francisco, California 94080; email: [email protected]

KEY WORDS: selectivity, mechanism, peptide-receptor structures

ABSTRACTThe scope and utility of phage display is reviewed with emphasis on medical ap-plications and structure-based ligand and drug design, from literature mostly after1994. General principles by which phage-displayed peptides achieve affinity andselectivity for targets are described, along with selected structural or mechanisticstudies of the binding of peptides or proteins discovered or engineered by phagedisplay. Such engineered proteins whose wild-type or mutant crystal or 2D-NMRstructures yield insight about the basis for enhanced affinity or altered specificityinclude antibodies, zinc fingers, human growth hormone, protein A, and atrial na-triuretic peptide. Structures of complexes of de novo phage-discovered peptideligands with targets such as the Src SH3 domain, streptavidin, and erythropoi-etin receptor reveal the structural basis for receptor-peptide recognition in thesesystems.

CONTENTS

PERSPECTIVES AND OVERVIEW. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28General Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28Selected Medical Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

GENERAL PRINCIPLES BY WHICH PHAGE-DISPLAYED PEPTIDES ACHIEVE AFFIN-ITY AND SELECTIVITY FOR TARGETS. . . . . . . . . . . . . . . . . . . . . . . . . . . 31

271056-8700/97/0610-0027$08.00



28 KATZ

Core Residues of Small Peptides Determine Affinity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Appropriate Conformational Constraints Increase Affinity. . . . . . . . . . . . . . . . . . . . . . . . 32

MOLECULAR BASIS FOR AFFINITY AND SELECTIVITY OF PHAGE-DISPLAYEDPROTEINS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

Visualization of Receptor-Ligand Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33Affinity-Matured Human Growth Hormone. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Minimization of the Size of Natriuretic Peptide and of a Binding Domain from Protein A34Structural Plasticity in Phage-Displayed Fabs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Dimerization Improves Single-Chain Antibodies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35DNA Recognition by Phage-Displayed Zinc Fingers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

MOLECULAR BASIS FOR AFFINITY AND SELECTIVITY OF DE NOVO PEPTIDESDISPLAYED ON PHAGE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

L- andD-Peptide Ligands to the Src SH3 Domain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36Binding of Diverse Phage-Displayed Sequences to HLA-DR1. . . . . . . . . . . . . . . . . . . . . . 37Combining Phage Display with Structure-Based Design in the Streptavidin Model System38Small Peptide Agonist Dimerizes EPO Receptor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

PERSPECTIVES AND OVERVIEW

IntroductionPhage display has rapidly matured as a widespread technology for harnessingthe chemical and structural diversity of peptide libraries. Since its inception in1985 (82), phage display (reviewed in 10, 11, 14, 15, 17, 60, 67, 93) has rev-olutionized the discovery of novel peptide ligands that bind to diverse targets.Development of alternate biological systems for displaying combinatorial li-braries (64, 85) promises to expand the technology. The power of phage displayfor screening libraries is twofold. First, the expressed fusion peptide or proteinrecognition elements are directly linked to the genes encoding their replication,allowing facile identification and amplification of binding sequences. Second,phage-displayed moieties usually exhibit the same or similar functional char-acteristics as their native counterparts in solution.

Because proteins and peptides are the natural substrates or ligands in ubiqui-tous biological events, phage display is an ideal tool for studying macromolecu-lar interactions, central in both normal biological processes as well as in diseasestates. One active thrust in the pharmaceutical sector involves the discovery ofpeptide ligands as initial leads for subsequent drug development. The versatil-ity of phage display has become apparent through discovery of peptide epitopesthat bind to targets whose natural ligands or substrates are not only proteins orpeptides, but also nonpeptides (34, 80).

General ApplicationsPhage display of entire proteins, receptor domains (42), or lectin chains or do-mains (27) enables the engineering of properties of natural proteins (reviewedin 67) without detailed knowledge of structure-function relationships. Such



PHAGE PEPTIDE—RECEPTOR RECOGNITION 29

protein engineering is exemplified by the directed in vitro evolution of speci-ficity of enzymes (58, 91), of enzyme inhibitors (22, 61, 62, 85, 87, 88), ofDNA binding proteins (43, 58, 96), and of enhanced affinities of cytokines fortheir receptors (77, 86).

Substrate specificity mapping through presentation of peptides on phageas potential substrates (24, 83) has accelerated identification of highly activeand selective substrates without any a priori assumptions about the specificityor physiological substrates. Finally, phage display affords a powerful toolfor localization of epitopes of monoclonal (2, 5, 7, 84) and polyclonal (28)antibodies, and of many other binding proteins (18, 35, 38, 74–76, 94).

Selected Medical ApplicationsSURROGATE PEPTIDE LIGANDS FOR RECEPTORSA small cyclic peptideidentified by phage-display binds to five somatostatin receptor subtypes withnanomolar affinities (94). Small peptides bind specifically to the cell surfaceimmunoglobulin receptor of a human B-cell lymphoma cell line, and when con-jugated into dimers or tetramers, kill these cells in vitro with 50% inhibitoryconcentrations (IC50s) between 40 and 200 nM, providing a model for peptide-based therapy (75). Potent specific ligands for the human urokinase receptor, akey mediator of tumor cell invasion (34), compete with urokinase for receptorbinding with IC50s between 10 nM and 10µM, providing leads in the discoveryof drugs that inhibit cell-surface proteolysis.

TUMOR SUPPRESSOR PROTEIN (P53) AND INTEGRINSMutations that alter thefunction of p53 comprise the most common genetic alteration in human can-cers (18, 76). The honed specificity of a phage-displayed p53-binding peptideallowed discrimination between wild-type and correctly folded single-site mu-tants of p53 (18). Targeting subpopulations of a heterogeneous p53 pool can beviewed as a key step in developing new diagnostics and therapeutics for treatingcancers expressing mutant p53s.

Integrins are cell surface receptors that enable communication between theexterior and the interior of the cell necessary for anchorage and for redundantregulation of cell growth, differentiation, and migration (reviewed in 41). Theintegrinαvβ3 has been a target for discovery of small ligands by phage display(39, 54), as it is involved not only in physiological processes such as angiogen-esis and tissue repair, but also in pathological conditions such as osteoporosis,tumor cell metastasis, and adenoviral infection. Similarly, theα5β1 andαIIbβ3integrins have been the targets for peptide inhibition of tumor invasion andplatelet aggregation, respectively (39, 54).

PHAGE-ANTIBODIES Optimization of high-affinity antibodies by phage displayof combinatorial antibody libraries is a robust mimic of immune selection for



30 KATZ

natural antibody diversity, circumventing hybridoma and immunization tech-nologies. Phage-antibodies have found broad application both in medical sci-ence (reviewed in 92) and in plant biology (reviewed in 16). Evolution ofexquisite specificity of antibodies from large phage display repertoires is ex-emplified by a single-chain alloantibody developed as a diagnostic and ther-apeutic reagent capable of discriminating between two polymorphic forms oftheαIIb/β3 integrin differing by a single amino acid (36). Human Fabs, single-chain antibodies (19, 21, 98), or disulfide-stabilized Fv’s (9) can be isolatedwith specificities against virtually any targeted antigen, either foreign or self(26, 37), hapten (81), carbohydrate (21, 98), protein, DNA (3), or RNA (72).

ANTITUMOR AND ANTIVIRAL ANTIBODIES Phage display provides access to avast untapped pool of human monoclonal antitumor antibodies with clinicaland research potential (13, 78, 97). When combined with toxins or radioiso-topes, single-chain antibodies can be advantageous for cancer therapy becausetheir smaller size allows greater tumor penetration and faster clearance rates.Disulfide-stabilized Fv’s (whose applications to cancer targeting are reviewedin 73) have the further advantage over single-chain counterparts of being inca-pable of oligomerization.

Fabs isolated against a host of viral pathogens (reviewed in 12), includingthe human immunodeficiency virus type 1 (HIV-1; reviewed in 26), provideresearch tools, diagnostics, and potential pharmaceutical reagents for the pro-phylaxis and treatment of viral diseases. Physiologically relevant doses of aFab isolated from a phage display library prepared from bone-marrow of a long-term, asymptomatic, HIV-1 seropositive donor provided complete protectionagainst HIV-1 infection in a mouse model (68).

TAILORING SPECIFICITY THROUGH EPITOPE MASKING A recently developedphage display approach utilizes specific immobilized monoclonal antibodiesto capture native viral proteins of extracts of infected cells (25). The methodcircumvents the use of purified glycoprotein antigens whose conformationalintegrity may be compromised during purification, and permits faster estab-lishment of Fabs directed to multiple glycoproteins. The capture antibody,by masking an epitope on the captured antigen, restricts the specificity of thecloned Fabs to unmasked eptiopes. Sequential use of different capture antibod-ies allows tailoring the affinity of selected Fabs to specific regions and epitopeswithin the target viral proteins.

VACCINE DEVELOPMENT Random phage peptides selected against an anti-mucin core antibody defined the antibody’s specificity and uncovered epitopesfor cancer vaccine candidates (56). Using a highly potent and broadly reactiveHIV-1 neutralizing antibody as the probe, Keller et al identified hundreds of




reactive phage clones whose epitopes mimic the HIV-1 gp120 V3 loop, andwhich provide HIV vaccine candidate peptides (52).

The scope of phage display has expanded from applications involving tar-gets of well-defined homogeneous molecules, to whole cells (13, 20, 29, 33),to sera from clinically well-defined patients (15, 32, 66), the latter to identifydisease-specific epitopes without the need to identify the pathological antigen.Mimotope peptides that recognize antibodies of patients can be identified ascandidates for vaccine development in the absence of purified monoclonal an-tibodies (15, 32, 66). The phage itself is a suitable vector for immunogenicmimotopes (66), and multifunctional phage displaying both B- and T-cell epi-topes are superb immunogens (23, 66).

ORGAN TARGETING IN VIVO An organ-selective targeting approach based onin vivo screening of random peptide sequences identified peptides capable ofmediating selective localization of phage to brain and kidney blood vessels(70). The peptide sequences represent the first step towards identifying selectiveendothelial markers, with potential use in targeting cells, genes, and drugs intoselected tissues. Cancer therapy was envisioned with drug conjugates directedat tumor vasculatures containing specific targetable markers.

GENERAL PRINCIPLES BY WHICH PHAGE-DISPLAYEDPEPTIDES ACHIEVE AFFINITY AND SELECTIVITYFOR TARGETS

Core Residues of Small Peptides Determine AffinityTo completely sample all possible mutants that can be surveyed simultaneously(∼108), total randomization is limited to 4–6 residues due to codon degeneracy(14). Thus many searches for ligands are undertaken with random hexapeptidelibraries. The binding sequences obtained from such libraries often exhibit aconserved subset or core of residues constituting a binding motif. For example,phage-displayed hexapeptides selected for binding to integrins (39, 53, 54)share the RGD sequence, and those selected for binding to concanavalin A (80)share the YPY sequence. In general, the role of residues flanking the conservedcore sequences may be to provide peripheral, less specific interactions with thetarget, to exert a favorable influence on the conformation of the core motif, orto constrain its conformation through disulfides.

CONTINUOUS EPITOPES Sequences discovered by screening phage librariesare often identical to those of the recognition sequences of the correspondingnatural ligands. The core consensus sequences of phage-displayed epitopesrecognized by many monoclonal antibodies are similar to those of the epitopes



32 KATZ

of the proteins against which the antibodies were originally raised. Selectionof semisynthetic antibodies by phage display has also been shown to producesequences similar to those obtained from the natural selection process operatingin the immune system. Thus, phage display can faithfully reproduce, reveal,and even improve on fundamental recognition elements in biology.

MIMOTOPES From a viewpoint of drug design, one property of ligands discov-ered by phage display that can be improved over corresponding natural proteinligands is their size. Because of the practical limitation in the length of peptideto be completely randomized (noted above), high affinity for targets of smallphage-displayed peptides is often achieved through mimotopes whose shortersequences differ from those of the natural ligands, yet mimic the specificity.

Appropriate Conformational Constraints Increase AffinityDISULFIDE CONSTRAINTS A generally exploited principle to increase ligandaffinity is to constrain the conformations of the unbound ligand to be the sameas or similar to that of the protein-bound ligand (reviewed in 30). Increasedaffinity of constrained ligands compared with more flexible counterparts is thusachieved, owing to a decrease in the conformational entropy in the unboundstate. Such constrained peptides have further key advantages in structure-baseddrug design because they are amenable to structural determination in solution.Thus, disulfide constraints have been used in the discovery of high-affinitycyclic peptide ligands (34, 39, 53, 54, 69, 79, 94).

A phage-displayed hexapeptide library againstα5β1 integrin (54) yielded apotent disulfide-bridged RGD-containing peptide. Cyclic ligands, some con-strained by two disulfides, selective for four different integrins, were discoveredin cysteine-flanked penta-, hexa-, and heptapeptide libraries (39). A cyclic pep-tide with two disulfides was 20-fold more potent in inhibiting the attachmentof αvβ3 or αvβ5 integrin-expressing cells to vitronectin than a correspondingpeptide with one disulfide, and 200-fold more potent than linear counterparts.By selecting ligands against RGD-containing fibronectin fragments (insteadof against the receptor, as above), the same group discovered a disulfide-bonded octapeptide that bound specifically to fibronectin and RGD-containingfibronectin fragments (69).

PROTEIN SCAFFOLDS The four-helix bundle framework of cytochrome b562,in which two loops were randomized (55), afforded mutants that bound a bovineserum albumin-hapten with aKd of ∼1µM, which is comparable to that for amonoclonal antibody directed against the conjugate (0.29µM). Tendamistat, aβ-sheet protein inhibitor ofα-amylase, was used as a scaffold for constrainingpeptides displayed on the loop regions to isolate mutants that bound specifi-cally to a monoclonal antibody (65). A potent and specific interleukin-6 an-tagonist was developed from a novel 61-residueβ-sheet “minibody” scaffold




corresponding to the variable domain of immunoglobulins, bearing two hyper-variable loops optimized by phage display (63). Optimization of hypervari-able loops on phage-displayed on Fabs or single-chain antibodies also yieldsconformationally constrained libraries; Figure 1 (See color insert) shows thestructure of a trisaccharide–single-chain antibody complex in which the sixcomplementarity-determining loops were randomized to increase affinity forthe carbohydrate antigen (21).

CONFORMATIONALLY HOMOGENEOUS LIBRARIES Libraries of peptides withpredetermined secondary structure can provide a conformational model of thepeptide pharmacophore directly from screening, thus allowing the design ofa suitable nonpeptidic scaffold to replace the peptide backbone in a processtermed selection-driven peptidomimetic design (5). A conformationally homo-geneous peptide library was produced from a structure-inducing template byrandomizing five positions occupying the solvent-exposed region of an eight-residueα-helical portion of a zinc finger motif (Figure 2, see color insert). Cou-pling zinc coordination with formation of a defined zinc finger motif structureprovided a built-in filter against selection of undefined or nonnative structuresamong the zinc-dependent ligand binding sequences in this library. Screen-ing against a monoclonal antibody yielded a well-defined consensus sequenceexhibiting strong, zinc-dependent binding to the antibody. The secondary struc-ture of a selected binding variant was shown to be equivalent to that of wildtype. Conformationally homogeneous libraries were considered advantageouswhen no structural information is available at the outset of a drug discoveryproject.

MOLECULAR BASIS FOR AFFINITY AND SELECTIVITYOF PHAGE-DISPLAYED PROTEINS

Visualization of Receptor-Ligand InteractionsAlthough the number of systems studied by phage display has burgeoned, manyfewer studies address the molecular basis by which either de novo peptidesor randomly mutagenized proteins develop affinity or specificity. The mostpowerful and direct method for visualizing the interactions responsible forbinding is determination of the structure of the ligand-target complex, by X rayor neutron crystallography, or by 2D NMR. When the structure of the target–natural ligand complex is known, modeling how alterations in the sequence ofthe natural ligand affect its interactions with the target may provide insight intothe affinity changes effected by phage display. Finally, in the absence of thestructure of the target, determination or modeling of the structure of the bindingpeptide or protein alone may provide insight into the determinants of affinityand selectivity.



34 KATZ

Affinity-Matured Human Growth HormoneThe affinity of human growth hormone (hGH) for its receptor has been improvedby phage display (86). Five selected regions, each comprising four residuesknown to contribute to binding from alanine scanning mutagenesis, were sep-arately mutagenized and mutants of higher affinity identified. Combination of15 mutations yielded a hormone with a 400-fold cumulative increase in affinity,which also exhibited an improved physicochemical property: better diffractionof crystals of the mutant (86).

Comparison of the crystal structures of the wild-type and mutant hormones(Figure 3, see color insert) showed that most of the secondary structural ele-ments are unchanged. A 3.0-Å shift in the main chain between residue 54 andminihelix 1 was attributed to a change of buried Phe54 to Pro. Conversion ofPhe10, buried in the four-helix core of wild type, to Ala causes a rigid bodyrotation of∼12◦ of helix 1 to bring its N terminus closer to helix 4. In addi-tion, the C terminus of helix 4 moves by∼1.1 Å toward helix 1. The spaceproduced from these shifts is filled by a water molecule, completely buried inthe hydrophobic core (hydrogen bonded to two carbonyl oxygens). Becausemost of the binding determinants for the high-affinity receptor occur on thesolvent accessible surface of helices 1 and 4, the Phe10→Ala substitution wasidentified as a potentially critical determinant of the enhanced affinity. Fourother substitutions at residues whose side chains are largely buried in wild typedo not cause any conformational differences but exert indirect effects on thebinding affinity, possibly by altering flexibility. Nine substituted residues werefound at the receptor-hormone interface in the wild-type complex.

Minimization of the Size of Natriuretic Peptideand of a Binding Domain from Protein AA combination of phage display, modeling, and alanine scanning mutagenesiswas used to reduce the size of natriuretic peptide (ANP), a hormone regulatingblood pressure and salt balance, from a 28- to 15-residue peptide retaining highbiopotency (57). Alanine scanning mutagenesis of ANP displayed on phageidentified seven residues most critical for receptor binding. Guided by the2D-NMR structure that showed that five of the critical residues form a clusterdefining a face of the molecule, Li et al replaced the native disulfide of ANP witha disulfide that placed the cluster of critical residues onto a smaller disulfide-bonded cyclic-peptide portion. The 100-fold decrease in binding affinity dueto the nonnative disulfide was restored by optimizing noncritical residues byphage display. The nine carboxyl residues were then deleted, and the peptidewas again reoptimized by phage display. Finally, the four N-terminal residueswere deleted to yield a 15-residue peptide whose affinity for the receptor (5.9nM) was about sevenfold weaker than that of the wild type (0.8 nM). The

P1: H

December 23, 1997 19:51 Annual Reviews AR034-00 AR34-FrontisP

Figure 1 Structure of single chain antibody [pdb 1 mfa (98)] showing 6 CDR’s opti-mized to increase affinity for carbohydrate antigen (21). The locations of the VH sub-stitutions Met34→Leu and Gly109→Ser, responsible for increasing affinity by pro-moting dimerization are indicated. Note that the peptide stretch linking L111 to H1is not visible due to disorder.

Figure 2 Structure of zinc-finger scaffold for achieving a conformationally homo-geneous library (5), from which high affinity antibody binders were selected, adaptedfrom Figure 1aof (5). The Zn+2 ion, coordinating residues (three letter code), and muta-genized wild-type and mutant residues (one letter codes) are indicated.

P1: H


Figure 3 Structure of affinity matured growth hormone [pdb 1 huw (86)], adapt-ed from Figure 3 of (86). The mutagenized and corresponding wild-type residues areindicated; those of the core are in italic.

P1: H


Figure 4 Structure of Zif268-DNA complex [pdb 1 zaa (71)], showing the wild-typeprotein residues that were mutagenized, adapted from Figure 1 of (43).

Figure 5 2D-NMR structure of SH3 complexed with class I ligand, RALPPLPRY,[pdb 1 rlp (31)] adapted from Figure 1a of (31). Shown are protein binding residues,Tyr92 (orange), Trp118 (blue), and Tyr136 (red), and ligand binding residues Leu3(green) and Pro4 (yellow).

P1: H


Figure 6 (a) Structure of streptavidin-cyclo-Ac-[CHPQGPPC]-NH2 [pdb 1 sle (46)].Streptavidin residues are labeled inyellow, peptide residues inwhite. H-bonds mediatingpeptide binding are shown inyellow; the H-bond between the peptide Gln Nε2H and theπ system of Trp108 is incyan. (b) Structure of streptavidin-cyclo-Ac-[CHPQFC]-NH2[pdb 1 sld (46)].




successful reduction of the size of the ANP hormone demonstrates the powerof phage display when guided by a 2D-NMR structure of the hormone alone inthe absence of a structure of the receptor-ligand complex.

A similar approach used a protein A–immunoglobulin-Fc crystal structureand iterations of structure-based design and phage display. The Z-domain ofprotein A was minimized from a 3-helix bundle of 59 residues that binds to theFc fragment with aKd of 10 nM, to a 2-helix bundle of 33 residues that bindswith a Kd of 43 nM (8).

Structural Plasticity in Phage-Displayed FabsA variety of CDR1 sequences were isolated (81), which imparted affinities fordigoxin similar to that of the wild-type monoclonal antibody, including twoof higher affinity by 4.1-fold (Kd= 4.6× 10−11 M) and 1.8-fold (Kd= 1.0×10−11 M). The crystal structure of the wild-type Fab–digoxin complex showsthat antigen recognition occurs exclusively by shape complementarity (45).From this structure, it was concluded that significant local main-chain rear-rangement occurs in some of the mutants and that the CDR1 is more plasticthan anticipated.

Dimerization Improves Single-Chain AntibodiesDramatic increases in affinities (fromKds of∼µM to ∼nM) for antigens ofphage-displayed single-chain antibodies can occur from antibody dimeriza-tion and resulting increased avidity (21, 78), owing to the repetitive nature ofpolysaccharide (21) or c-erbb2 (78) antigens. Single-chain antibody dimersare formed through interaction of the variable light chain (VL) of one single-chain Fv with the variable heavy chain (VH) of a second molecule. The VHsubstitutions responsible for increasing affinity do not contact the antigen; theyoccur near the VL − VH interface (Figure 1) and have a profound effect on thedimer/monomer equilibrium. Microcalorimetry showed that enthalpy changesdominated the binding thermodynamics of the trisaccharide to both wild-typeand mutant single-chain antibodies (21).

DNA Recognition by Phage-Displayed Zinc FingersZinc finger mutants whose specificities were tuned either to wild-type or alteredoperator DNA sequences were engineered by phage display (43). Four residuesmediating DNA binding in the finger of helix 1 of the murine transcription factor,Zif268, were identified for random mutagenesis from the crystal structure of theZif268-DNA complex (71) (Figure 4, see color insert). The altered sequenceswere related to interactions in the crystal structure of the wild-type protein-DNAcomplex and to the predicted structural changes incurred by alterations in thesequences of selector DNA operators. Zinc fingers capable of distinguishingoperator sequences differing by a single base change were isolated.



36 KATZ

Observation of Arg at position 6, in all isolates selected with operators con-taining a 5′ guanine (G) in the 3′ triplet [either guanine-thymine-guanine [GTG]or guanine-cytosine-guanine (GCG)], was attributed to the requirement for acharged hydrogen bond donor to interact with the 5′ guanine, as seen in theX-ray structure of the native protein-DNA complex (Figure 4). Alteration ofthe 5′ G to T in this target triplet resulted in a complete consensus for Lys atposition 6. Complementary charged pairs or charged-neutral pairs were alwaysfound at position−1 and 2 of the protein, which suggests the importance ofmaintaining intramolecular interactions such as the salt bridge between Arg andAsp at these positions in the wild-type crystal structure (Figure 4a). Selectionof clones directed to the GTG 3′ triplet compared with GCG led to isolation ofclones in which side chains at position 3 larger than Ser were excluded by themethyl group of thymine. Finally, DNA sequence changes that were nearby,but not in direct contact with the protein, affected the amino acid preferencesin the zinc-finger mutants.

The specificity of each of the three fingers of Zif268 was also altered byWu et al (96), who randomly mutagenized six residues in each of the threefingers and then measured and compared affinities and kinetics of binding ofselected proteins by surface plasmon resonance. Changes in the specificity ofthe fingers were generally governed by changes in the stability of the protein-DNA complex as reflected in the the rate of dissociation of the protein fromDNA (koff) values. The sequence of the finger 1 mutant with the highest affinityand selectivity for wild-type DNA differed radically from that of the wild-typeprotein. Also, because many of the selected mutants contained a Pro in regionsthat areα-helical in wild type, it was concluded that the secondary structure ofthis region may be altered in these proteins, and thus there is no single generalcoding relationship between zinc finger protein and target DNA sequence.

Modularity of the zinc finger domain was considered from the viewpoint offusing individually selected domains to produce molecules potentially capableof recognizing any DNA sequence; a unique site in the human genome couldbe targeted by a six-finger complex. The functioning of such zinc fingers aspositive or negative genetic switches was envisioned in the context of developingspecific antiviral or antitumor agents (96).

MOLECULAR BASIS FOR AFFINITY AND SELECTIVITYOF DE NOVO PEPTIDES DISPLAYED ON PHAGE

L- andD-Peptide Ligands to the Src SH3 DomainThe Src homology 3 (SH3) domain, a 50-residue domain present in many cellu-lar proteins, mediates intracellular signal transduction that is important for cell




growth, migration, differentiation, and responses to the external environment.Pathogens also exploit interactions mediated by SH3 domains by mechanismsrecently reviewed (6). Targeting the SH3 domain with pharmaceuticals thatmimic the natural ligand-binding motif is a means of blocking specific cellularevents to modulate both physiologic and pathological processes. SH3 bindingmotifs and determinants of binding specificity for Src, Fyn, Lyn, Yes, Abl, andphosphatidylinositol 3-kinase were elucidated by phage display (75). Cyclic,disulfide-bridgedD-peptide ligands to the SH3 domain were also identified by“mirror-image phage display,” which involves chemical synthesis of the SH3domain in theD-amino acid configuration, followed by selection from a phagedisplay library expressing randomL-amino acid peptides (79).

Two types of SH3 domainL-peptide ligands (class I and class II) have beenidentified. Class I ligands, identified both by phage display and by a biasedpeptide library, have the consensus sequence RXLPPLP. Class II ligands, iden-tified by a biased peptide library, have the consensus sequence9PPLPXR (9 isa hydrophobic residue). 2D-NMR structures (31) of complexes of SH3 domainwith class I (Figure 5, see color insert) and class II ligands show that the boundligands adopt a left-handed polyproline type II helix. However, the direction-ality of the helix of bound class I ligands is opposite to that of bound classII ligands. The peptide orientation is determined by a salt bridge formed bythe terminal Arg of the ligands and the conserved Asp 99 of the SH3 domain.Residues at positions 3, 4, 6, and 7 of both peptide classes intercalate into theligand binding site. In contrast to theL-peptide ligands for the SH3 domain, thepositively charged residues in theD-peptide ligands are located in the middle ofa stretch of conserved residues, suggesting that the mode of ligand binding isdifferent (79). Heteronuclear 2D-NMR studies on binding of aD-peptide ligandindicate that its binding site partially overlaps that for the physiological ligands.

Binding of Diverse Phage-DisplayedSequences to HLA-DR1The major histocompatibility complex (MHC) class II molecules are highlypolymorphic membrane glycoproteins that bind peptide fragments of proteinsand display them for recognition by T lymphocytes to generate an immuneresponse (38, 44). Crystallographic analysis of endogenous peptides associatedwith human leukocyte antigen DR (HLA-DR1) show that binding is mediatedby 14 hydrogen bonds between the peptide main chain and MHC side chains.The hydrogen bonds, along with location of MHC pockets, induce the peptideto adopt a polyproline II conformation.

The crystal structure suggests two components to the mechanism by whichMHC class II molecules form tightly bound complexes with peptides of di-verse sequences. One component involves interactions with specific peptide



38 KATZ

side chains of the MHC pockets, many of which can accommodate a numberof different side chains, thereby reducing the restrictions placed on the bindingsequences. These pockets probably recognize the conserved and allele-specificanchor residues identified by phage display for three different (HLA-DR)molecules (38). The second component to peptide binding involves conservedMHC residue interactions with the peptide main chain, providing a set of inter-actions that are independent of the peptide sequence and MHC polymorphism.

Combining Phage Display with Structure-Based Designin the Streptavidin Model SystemAs a paradigm for understanding how de novo peptides discovered by phagedisplay achieve high affinity and selectivity, my colleagues and I have used strep-tavidin as a model receptor to identify high-affinity (Kd∼ 0.4 µM) disulfide-bridged cyclic peptide ligands (34), and to determine crystallographically themolecular basis by which the ligands are recognized (46). Streptavidin, atetrameric protein secreted byStreptomyces avidinii, functions as an antibi-otic by binding its natural nonpeptide ligand, biotin (vitamin H) with one of thehighest known affinities (Kd = 10−14 M) for a noncovalent interaction. Thechoice of streptavidin as a model receptor was also guided by an objective toprobe the structural relationship between peptide and nonpeptide ligands boundin a common site of a single target. Such information could provide insightinto the principles involved in converting initial peptide leads (either the naturalknown substrates or ligands of a target, or those discovered by phage display)into small molecule drugs.

The streptavidin-bound crystal structures of two disulfide-bridged cyclicpeptides (cyclo-Ac-[CHPQGPPC]-NH2, andcyclo-Ac-[CHPQFC]-NH2) (Fig-ure 6, see color insert) and of a linear peptide (FSHPQNT) showed that the boundligands share a common HPQ conformation and make common interactionswith streptavidin, although significant differences in structures and interactionsoccur for flanking residues (46). In all complexes, bound peptides adopt a typeI β-turn reflected by a hydrogen bond between the His main chain carbonyland the main-chain amide NH of thei +3 residue in the bound cyclic peptides.In boundcyclo-Ac-[CHPQFC]-NH2, there is an additional hydrogen bond, in-dicative of anα-helix, between the main-chain His carbonyl and the main-chainC-terminal Cys amide. Location of the disulfides of the bound cyclic peptideson the surface is consistent with their proposed entropic effect on binding (34);they do not interact with the protein and thus do not directly contribute enthalpi-cally to the binding. Binding interactions for both cyclic and linear peptidesinclude hydrophobic interactions, direct hydrogen bonds, and hydrogen bondsmediated by water molecules that play an integral role in the binding of thepeptide ligands. The directionalities of all hydrogen bonds mediating binding




Figure 7 Schematic showing H-bonds mediating binding ofcyclo-Ac-[CHPQGPPC]-NH2 tostreptavidin (48).

of both the linear and cyclic peptides could be unambiguously determined(Figure 7).

MECHANISM OF BINDING Plasmon resonance measurements showed that theaffinities of the HPQ-containing peptide ligands decrease significantly belowpH ∼ 6.3, but they remain relatively constant above pH∼ 6.3 (48). Thestructures of the streptavidin-bound peptides determined at pHs ranging from2.0–10.5 indicated that at all these pHs, the His of the peptide is uncharged;there is a hydrogen bond between Nδ1 and the Gln amide proton (Figures 6 and7). Thus, deprotonation of the peptide His is required for high-affinity binding.

STRUCTURE-BASED DESIGN The structures of the complexes enabled structure-based design of high-affinity streptavidin-binding ligands conformationallyconstrained by thioether crosslinks, which have some advantages over disul-fides (49). The relationship between the structure of streptavidin-biotin and ofstreptavidin-cyclic peptide ligands enabled design of a small organic ligand,glycoluril, with a Kd of 2.5µM and with structural features common to bothbiotin and the cyclic peptide ligands from which it was designed (50). Strep-tavidin also proved to be an ideal model system for applying phage displayto development of receptor dimerizing agents. Topochemical crystal lattice–



40 KATZ

mediated dimerization, through disulfide interchange, of adjacentcyclo-Ac-[CHPQGPPC]-NH2 ligands bound in either the I222 (51) or I4122 (47) latticesproduced ligand dimers that dimerized streptavidin.

Small Peptide Agonist Dimerizes EPO ReceptorDimerization triggers many important biological processes (4; reviewed in 1,40). The action of the hGH receptor relies on dimerization of the cytokine-binding domains to produce a 1:2 cytokine-cytokine receptor complex that hasapproximate C2 symmetry (involving a rotation) (reviewed in 90). Recep-tor dimerization is also induced by erythropoietin (EPO) (89), a protein thatregulates the proliferation and differentiation of immature erythroid cells. Re-cently, phage display has been applied to the development of an EPO-receptordimerizing agonist peptide, a 20-residue disulfide-bridged ligand, unrelated insequence to EPO. The peptide competes with EPO in receptor binding assays,induces cellular proliferation of EPO-responsive cell lines, and exhibits signif-icant erythropoeitic effects in mice (95).

Crystallography showed that the peptide forms a 2:2 complex with the re-ceptor (59). The peptide monomers are associated through a contact surfacearea of 320Å2 formed by twofold related hydrophobic cores and by hydrogenbonds between twofold relatedβ-sheet segments. A portion of each peptidemonomer interacts with both receptors. The peptide and receptor dimers bury840Å2 and 880Å2 of surface area, respectively, in the complex. The peptidedimer provides hydrophobic interactions, mostly from its hydrophobic coreand hydrogen bonds, primarily linking the two twofold-relatedβ-turns to thereceptors. Interactions provided by the receptor are from recognition loopsequivalent to those in the hGH-hGH receptor structure. However, the dimerassembly differs markedly from that of the hGH receptor complex, suggestingthat more than one mode of dimerization can induce signal transduction andcell proliferation. Through the powerful combination of phage display andcrystallography, discovery of a small EPO mimetic, followed by determinationof its mode of interaction with the EPO receptor, now provides a structuralfoundation for development of a nonpeptide, small-molecule drug for anemia.

CONCLUSIONS

Phage display stands out as an icon of modern molecular biology. In manysystems, it has been tightly coupled to drug discovery and has had a signifi-cant impact on medical science. Targets have been successfully extended fromhomogeneous molecules to whole cells, to sera, and finally to organs in vivo.The mirror-image phage display approach promises to further extend the tech-nology by allowing identification of metabolically stableD-peptide ligands tosmall synthesizable proteins. The structural basis for recognition of phage-




displayed peptides for targets has been probed in a number of systems. Thefuture will surely bring more applications and innovations in the technologyfor probing biological processes and identifying drug leads, as well as morestructural and mechanistic insights into the molecular basis whereby phage-displayed peptides and proteins achieve affinity and selectivity for their targets.

Visit the Annual Reviews home pageathttp://www.annurev.org.

Literature Cited

1. Austin DJ, Crabtree RG, Schreiber SL.1994. Proximity versus allostery: the roleof regulated protein dimerization in biol-ogy.Chem. Biol.1:125–29

2. Balass M, Heldman Y, Cabilly S, GivolD, Katchalski-Katzir E, et al. 1993. Iden-tification of a hexapeptide that mimicsa conformation-dependent binding site ofacetylcholine receptor by use of a phage-epitope library.Proc. Natl. Acad. Sci. USA90:10638–42

3. Barbas SM, Ditzel HJ, Salonen EM, YangWP, Silverman GJ, et al. 1995. Human au-toantibody recognition of DNA.Proc. Natl.Acad. Sci. USA92:2529–33

4. Belshaw PJ, Ho SN, Crabtree GR,Schreiber SL. 1996. Controlling proteinassociation and subcellular localizationwith a synthetic ligand that induces het-erodimerization of proteins.Proc. Natl.Acad. Sci. USA93:4604–7

5. Bianchi E, Folgori A, Wallace A, NicotraM, Acali A, et al. 1995. A conformation-ally homogeneous combinatorial peptidelibrary. J. Mol. Biol.247:154–60

6. Bliska J. 1996. How pathogens exploitinteractions mediated by SH3 domains.Chem. Biol.3:7–11

7. Böttger V, Böttger A, Lane EB, Spruce BA.1995. Comprehensive epitope analysis ofmonoclonal anti-proenkephalin antibodiesusing phage display libraries and syntheticpeptides: revelation of antibody fine speci-ficities caused by somatic mutations in thevariable genes.J. Mol. Biol.247:932–46

8. Braisted AC, Wells JA. 1996. Minimizing abinding domain from protein A.Proc. Natl.Acad. Sci. USA93:5688–92

9. Brinkmann U, Chowdhury PS, RoscoeDM, Pastan I. 1995. Phage display ofdisulfide-stabilized Fv fragments.J. Im-munol. Methods182:41–50

10. Burritt JB, Bond CW, Doss KW, Je-

saitis AJ. 1996. Filamentous phage displayof oligopeptide libraries.Anal. Biochem.238:1–13

11. Burton DR. 1995. Phage display.Im-munotechnology1:87–94

12. Burton DR. 1995. Human antibodies to vi-ral pathogens from phage display libraries.Vaccines 95: Mol. Approaches Control In-fect. Dis. Annu. Meet., 12th,ed. RM Cha-nock, pp. 1–11. Cold Spring Harbor, NY:Cold Spring Harbor Lab. Press

13. Cai X, Garen A. 1995. Anti-melanomaantibodies from melanoma patients im-munized with genetically modified autol-ogous tumor cells: selection of specificantibodies from single-chain Fv fusionphage libraries.Proc. Natl. Acad. Sci. USA92:6537–41

14. Clackson T, Wells JA. 1994. In vitro se-lection from protein and peptide libraries.Trends Biotechnol.12:173–84

15. Cortese R, Monaci P, Nicosia A, LuzzagoA, Felici F, et al. 1995. Identification ofbiologically active peptides using randomlibraries displayed on phage.Curr. Opin.Biotechnol.6:73–80

16. Crosby WL, Schorr P. 1995. Principlesand applications of recombinant antibodyphage display technology to plant biology.Methods Cell Biol.50:85–99

17. Daniels DA, Dion A, Lane DP. 1995.Phage peptide libraries for the selection ofhighly specific ligands.Exp. Opin. Ther.Pat.5:901–12

18. Daniels DA, Lane DP. 1994. The charac-terization of p53 binding phage isolatedfrom phage peptide display libraries.J.Mol. Biol. 243:639–52

19. de Kruif J, Boel E, Logtenberg T. 1995.Selection and application of human sin-gle chain Fv antibody fragments from asemi-synthetic phage antibody display li-brary with designed CDR3 regions.J. Mol.



42 KATZ

Biol. 248:97–10520. de Kruif J, Terstappen L, Boel E, Logten-

berg T. 1995. Rapid selection of cellsubpopulation-specific human monoclonalantibodies from a synthetic phage anti-body library.Proc. Natl. Acad. Sci. USA92:3938–42

21. Deng SJ, MacKenzie CR, Hirama T,Brousseau R, Lowary TL, et al. 1995. Ba-sis for selection of improved carbohydrate-binding single-chain antibodies from syn-thetic gene libraries.Proc. Natl. Acad. Sci.USA92:4992–96

22. Dennis MS, Herzka A, Lazarus RA. 1995.Potent and selective Kunitz domain in-hibitors of plasma kallikrein designed byphage display.J. Biol. Chem.270:25411–17

23. Di Marzo Veronese F, Willis AE, Boyer-Thompson C, Appela E, Perham RN. 1994.Structural mimicry and enhanced immuno-genicity of peptide epitopes displayed onfilamentous bacteriophage. The V3 loop ofHIV-1 gp120.J. Mol. Biol.243:167–72

24. Ding L, Coombs GS, Strandberg L, NavreM, Corey DR, et al. 1995. Origins of thespecificity of tissue-type plasminogen acti-vator.Proc. Natl. Acad. Sci. USA92:7627–31

25. Ditzel HJ, Binley JM, Moore JP, SodroskiJ, Sullivan N, et al. 1995. Neutralizing re-combinant antibodies to a conformationalV2- and CD4-binding site–sensitive epi-tope of HIV-1 gp120 isolated by using anepitope-masking procedure.J. Immunol.154:893–906

26. Ditzel HJ, Burton DR. 1995. Antibodyrepertoire to self and the virus in HIV-1 in-fection probed by phage display libraries.Vaccines 95: Mol. Approaches ControlInfect. Dis. Annu. Meet., 12th,ed. RMChanock, pp. 19–26. Cold Spring Harbor,NY: Cold Spring Harbor Lab. Press

27. Djojonegoro BM, Benedik MJ, WilsonRC. 1994. Bacteriophage surface displayof an immunoglobulin-binding domain ofStaphylococcus aureusprotein A. Bio-Technology12:169–72

28. Dybwad A, Bogen B, Natvig JB, FoerreOE, Sioud M. 1995. Peptide phage librariescan be an efficient tool for identifying anti-body ligands for polyclonal antisera.Clin.Exp. Immunol.102:438–42

29. Dziegiel M, Nielsen LK, Andersen PS,Blancher A, Dickmeiss E, et al. 1995.Phage display used for gene cloning of hu-man recombinant antibody against the ery-throcyte surface antigen, rhesus D.J. Im-munol. Methods182:7–19

30. Fairlie DP, Abbenante G, March DR. 1995.Macrocyclic peptidomimetics—forcing

peptides into bioactive conformations.Curr. Med. Chem.2:654–86

31. Feng S, Chen JK, Yu H, Simon JA,Schreiber SL. 1994. Two binding orienta-tions for peptides to the Src SH3 domain:development of a general model for SH3-ligand interactions.Science266:1241–47

32. Folgori A, Tafi R, Meola A, Felici F, Galfr´eG, et al. 1994. A general strategy to identifymimotopes of pathological antigens usingonly random peptide libraries and humansera.EMBO J.13:2236–43

33. Fong S, Doyle LV, Devlin JJ, Doyle MV.1994. Scanning whole cells with phage-display libraries: identification of peptideligands that modulate cell functions.DrugDev. Res.33:64–70

34. Giebel LB, Cass R, Milligan DL, Young D,Arze R, et al. 1995. Screening of cyclic pep-tide phage libraries identifies ligands thatbind streptavidin with high affinity.Bio-chemistry34:15430–35

35. Goodson RJ, Doyle MV, Kaufman SE,Rosenberg S. 1994. High-affinity uroki-nase receptor antagonists identified withbacteriophage peptide display.Proc. Natl.Acad. Sci. USA91:7129–33

36. Griffin HM, Ouwehand WH. 1995. A hu-man monoclonal antibody specific for theleucine-33 (P1A1, HPA-1a) form of plateletglycoprotein IIIa from a V gene phage dis-play library.Blood86:4430–36

37. Griffiths AD, Malmqvist M, Marks JD, ByeJM, Embleton MJ, et al. 1993. Human anti-self antibodies with high specificity fromphage display libraries.EMBO J.12:725–34

38. Hammer J, Valsasnin P, Tolba K, Bolin D,Higelin J, et al. 1993. Promiscuous and al-lele specific anchors in HLA-DR bindingpeptides.Cell 74:197–203

39. Healy JM, Murayama O, Maeda T, YoshinoK, Sekiguchi K, et al. 1995. Peptide lig-ands for integrinα5β3 selected from ran-dom phage display libraries.Biochemistry34:3948–55

40. Heldin CH. 1995. Dimerization of cell sur-face receptors in signal transduction.Cell80:213–23

41. Hynes RO. 1992. Integrins: versatility,modulation, and signalling in cell adhe-sion.Cell 69:11–25

42. Jacobsson K, Frykberg L. 1995. Cloningof ligand-binding domains of bacterial re-ceptors by phage display.BioTechniques18:878–85

43. Jamieson AC, Kim S-H, Wells JA. 1994.In vitro selection of zinc fingers with al-tered DNA-binding specificity.Biochem-istry 33:5689–95




44. Jardetzky TS, Brown JH, Gorga JC, SternLJ, Urban RG, et al. 1996. Crystallo-graphic analysis of endogenous peptidesassociated with HLA-DR1 suggests a com-mon, polyproline II-like conformation forbound peptides.Proc. Natl. Acad. Sci. USA93:734–38

45. Jeffrey PD, Strong RK, Sieker LC, ChangCYY, Campbell RL, et al. 1993. 26-10 Fab-digoxin complex: affinity and specificitydue to surface complementarity.Proc. Natl.Acad. Sci. USA.90:10310–14

46. Katz BA. 1995. Binding to protein tar-gets of peptidic leads discovered by phagedisplay: crystal structures of streptavidin-bound linear and cyclic peptide ligandscontaining the HPQ sequence.Biochem-istry 34:15421–29

47. Katz BA. 1996. Preparation of a proteindimerizing ligand by topochemistry andstructure-based design.J. Am. Chem. Soc.118:2355–56

48. Katz BA, Cass RT. 1996. In streptavidin-bound peptide ligands containing the HPQsequence the pKa of the histidine is lessthan 2.5.J. Biol. Chem.In press

49. Katz BA, Johnson CR, Cass RT. 1995.Structure-based design of high affinitystreptavidin binding cyclic peptides con-taining thioether cross-links.J. Am. Chem.Soc.117:8541–47

50. Katz BA, Liu B, Cass RT. 1996. Structure-based design tools: structural and ther-modynamic comparison with biotin of asmall molecule that binds to streptavidinwith micromolar affinity.J. Am. Chem. Soc.118:7914–7920

51. Katz BA, Stroud RM, Collins N, Liu B,Arze R. 1995. Topochemistry for prepar-ing ligands that dimerize receptors.Chem.Biol. 2:591–600

52. Keller PM, Arnold BA, Shaw AR, TolmanRL, Van Middlesworth F, et al. 1993. Iden-tification of HIV vaccine candidate pep-tides by screening random phage epitopelibraries.Virology193:709–16

53. Koivunen E, Gay DA, Ruoslahti E. 1993.Selection of peptides binding to theα5β1integrin from a phage display library.J.Biol. Chem.268:20205–10

54. Koivunen E, Wang B, Ruoslahti E. 1995.Phage libraries displaying cyclic peptideswith different ring sizes: ligand specifici-ties of the RGD-directed integrins.Bio-Technology13:265–70

55. Ku J, Schultz PG. 1995. Alternate pro-tein frameworks for molecular recognition.Proc. Natl. Acad. Sci. USA92:6552–56

56. Laing P, Tighe P, Kwiatkowski E, MilliganJ, Price M, et al. 1995. Selection of pep-tide ligands for the antimucin core antibody

C595 using phage display technology: def-inition of candidate epitopes for a cancervaccine.Clin. Mol. Pathol.48:M136–41

57. Li B, Tom JYK, Oare D, Yen R, FairbrotherWJ, et al. 1995. Minimization of a polypep-tide hormone.Science270:1657–60

58. Light J, Lerner RA. 1995. Random mutage-nesis of staphylococcal nuclease and phagedisplay selection.Bioorg. Med. Chem.3:955–67

59. Livnah O, Stura EA, Johnson DL, Middle-ton SA, Mulcahy LS, et al. 1996. Func-tional mimicry of a protein hormone by apeptide agonist: the EPO receptor complexat 2.8Å. Science273:464–71

60. Lowman HB. 1997. Bacteriophage displayand discovery of peptide leads.Annu. Rev.Biophys. Biomol. Struct.26:XXX–XX

61. Markland W, Ley AC, Ladner RC. 1996.Iterative optimization of high-affinity pro-tease inhibitors using phage display. 2.Pasma kallikrein and thrombin.Biochem-istry 35:8058–67

62. Markland W, Ley AC, Lee SW, LadnerRC. 1996. Iterative optimization of high-affinity protease inhibitors using phage dis-play. 1. Pasmin.Biochemistry35:8045–88057

63. Martin F, Toniatti C, Salvati AL, CilibertoG, Cortese R, et al. 1996. Coupling pro-tein design and in vitro selection strategies.Improving specificity and affinity of a de-signedβ-protein IL-6 antagonist.J. Mol.Biol. 255:86–97

64. Mattheakis LC, Bhatt RR, Dower WJ.1994. An in vitro polysome display systemfor identifying ligands from very large pep-tide libraries.Proc. Natl. Acad. Sci. USA91:9022–26

65. McConnell SJ, Hoess RH. 1995. Tendami-stat as a scaffold for conformationally con-strained phage peptide libraries.J. Mol.Biol. 250:460–70

66. Meola A, Delmastro P, Monaci P, LuzzagoA, Nicosia A, et al. 1995. Derivation of vac-cines from mimotopes. Immunologic prop-erties of human hepatitis B virus surfaceantigen mimotopes displayed on filamen-tous phage.J. Immunol.154:3161–72

67. O’Neil K, Hoess RH. 1995. Phage display:protein engineering by directed evolution.Curr. Opin. Struct. Biol.5:443–49

68. Parren PW, Ditzel HJ, Gulizia RJ, BinleyJM, Barbas III CF, et al. 1995. Protectionagainst HIV-1 infection in hu-PBL-SCIDmice by passive immunization with a neu-tralizing monoclonal antibody against thegp120 CD4-binding site.AIDS9:F1–6

69. Pasqualini R, Koivunen E, Ruoslahti E.1995. A peptide isolated from phage dis-



44 KATZ

play libraries is a structural and functionalmimic of an RGD-binding site on integrins.J. Cell Biol.130:1189–96

70. Pasqualini R, Ruoslahti E. 1996. Organ tar-geting in vivo using phage display peptidelibraries.Nature380:364–66

71. Pavletich N, Pabo CO. 1991. Zinc finger-DNA recognition: crystal structure of aZif268-DNA complex at 2.1Å. Science252:809–17

72. Powers JE, Marchbank MT, DeutscherSL. 1995. The isolation of U1 RNA-binding antibody fragments from autoim-mune human-derived bacteriophage dis-play libraries.Symp. RNA Biology, I. RNA-Protein Interactions, Nucleic Acids Symp.Ser.33, pp. 240–43

73. Reiter Y, Pastan I. 1996. Antibody engi-neering of recombinant Fv immunotoxinsfor improved targeting of cancer: disulfide-stabilized Fv immunotoxins.Clin. CancerRes.2:245–52

74. Renschler MF, Bhatt RR, Dower WJ,Levy R. 1994. Synthetic peptide ligandsof the antigen-binding receptor induce pro-grammed cell death in a human B-celllymphoma. Proc. Natl. Acad. Sci. USA91:3623–27

75. Rickles RJ, Botfield MC, Zhou X-M,Henry PA, Brugge JS, et al. 1995. Phagedisplay selection of ligand residues im-portant for Src homology domain bind-ing specificity.Proc. Natl. Acad. Sci. USA92:10909–13

76. Roth J, Dittmer D, Rea D, Tartaglia J, Pao-letti E, et al. 1996. p53 as a target forcancer vaccines: recombinant canarypoxvirus vectors expressing p53 protect miceagainst lethal tumor cell challenge.Proc.Natl. Acad. Sci. USA93:4781–86

77. Saggio I, Gloaguen I, Poiana G, Laufer R.1995. CNTF variants with increased bio-logical potency and receptor selectivity de-fine a functional site of receptor interaction.EMBO J.14:3045–54

78. Schier R, Bye J, Apell G, McCall A,Adams GP, et al. 1996. Isolation of high-affinity monomeric human anti-c-erbb2single chain Fv using affinity-driven selec-tion. J. Mol. Biol.255:28–43

79. Schumacher TNM, Mayr LM, Minor DLJr, Milhollen MA, Burgess MW, et al.1996. Identification ofD-peptide ligandsthrough mirror-image phage display.Sci-ence271:1854–57

80. Scott JK, Loganathan D, Easley RB,Gong X, Goldstein IJ. 1992. A familyof concanavalin A binding peptides froma hexapeptide epitope library.Proc. Natl.Acad. Sci. USA89:5398–402

81. Short MK, Jeffrey PD, Kwong R-F, Mar-golies MN. 1995. Contribution of anti-body heavy chain CDR1 to digoxin bind-ing analyzed by random mutagenesis ofphage-displayed Fab 26-10.J. Biol. Chem.270:28541–50

82. Smith GP. 1985. Filamentous fusion phage:novel expression vectors that displaycloned antigens on the virion surface.Sci-ence228:1315–17

83. Smith MM, Shi L, Navre M. 1995. Rapididentification of highly active and selectivesubstrates for stromelysin and matrilysinusing bacteriophage peptide display li-braries.J. Biol. Chem.270:6440–49

84. Stephen CW, Helminen P, Lane DP. 1995.Characterization of epitopes on human p53using phage-displayed peptide libraries:insights into antibody-peptide interactions.J. Mol. Biol.248:58–78

85. Sternberg N, Hoess RH. 1995. Display ofpeptides and proteins on the surface of bac-teriophageλ. Proc. Natl. Acad. Sci. USA92:1609–13

86. Ultsch MH, Somers W, Kossiakoff AA, deVos AM. 1994. The crystal structure ofaffinity-matured human growth hormone at2 Å resolution.J. Mol. Biol.236:286–99

87. Van Meijer M, Roelofs Y, Neels J, Hor-revoets AJG, Van Zonneveld A-J, et al.1996. Selective screening of a large phagedisplay library of plasminogen activator in-hibitor 1 mutants to localize interactionsites with either thrombin or the variableregion 1 of tissue-type plasminogen acti-vator.J. Biol. Chem.271:7423–28

88. Wang C-I, Yang Q, Craik CS. 1995.Isolation of a high affinity inhibitor ofurokinase-type plasminogen activator byphage display of ecotin.J. Biol. Chem.270:12250–56

89. Watowich SS, Hilton DJ, Lodish HF. 1994.Activation and inhibition of erythropoeitinreceptor function: role of receptor dimer-ization.Mol. Cell. Biol.14:3535–49

90. Wells JA. 1996. Binding in the growth hor-mone receptor complex.Proc. Natl. Acad.Sci. USA.93:1–6

91. Widersten M, Mannervik B. 1995. Glu-tathione transferases with novel active sitesisolated by phage display from a library ofrandom mutants.J. Mol. Biol.250:115–22

92. Winter G, Griffiths AD, Hawkins RE,Hoogenboom HR. 1994. Making antibod-ies by phage display technology.Annu. Rev.Immunol.12:433–55

93. Winter J. 1994. Bacteriophage display:peptide libraries and drug discovery.DrugDev. Res.33:71–89

94. Wright RM, Gram H, Vattay A, Byrne S,




Lake P, et al. 1995. Binding epitope ofsomatostatin defined by phage-displayedpeptide libraries.Bio-Technology13:165–69

95. Wrighton NC, Farrell FX, Chang R,Kashyap AK, Barbone FP, et al. 1996.Small peptides as potent mimetics of theprotein hormone erythropoietin.Science273:458–63

96. Wu H, Yang WP, Barbas III CF. 1995.Building zinc fingers by selection: towarda therapeutic application.Proc. Natl. Acad.

Sci. USA92:344–4897. Yelton DE, Rosok MJ, Cruz G, Cosand

WL, Bajorath J, et al. 1995. Affinity mat-uration of the BR96 anti-carcinoma anti-body by codon-based mutagenesis.J. Im-munol.155:1994–2004

98. Zdanov A, Li Y, Bundle DR, Deng S-J,MacKenzie R, et al. 1994. Structure of asingle-chain antibody variable domain (Fv)fragment complexes with a carbohydrateantigen at 1.7Å resolution. Proc. Natl.Acad. Sci. USA91:6423–27

STRUCTURAL AND MECHANISTIC DETERMINANTS OF ......P1: RHS/PLB P2: SDA/PLB QC: SDA/PBS/AGR T1: PBS...

Documents

Transcript of STRUCTURAL AND MECHANISTIC DETERMINANTS OF ......P1: RHS/PLB P2: SDA/PLB QC: SDA/PBS/AGR T1: PBS...