503 Phage display of combinatorial antibody libraries Christoph Rader and Carlos F Barbas Ill* Abbreviations CDR complementarity-determining region HA hemagglutinin peptide scFr single-chain variable region fragment encapsulated genotype via the phage surface forms the basis of the technique (reviewed in [3-S]). Typically, the antibody fragments are displayed on the surface of phage as either Fab fragments [6], single-chain variable region fragments (scFvs), or dimeric scFvs, also known as diabodies, which differ from scFvs in the reduced length of the linker peptide used and their preference to associate as dimers. Library construction is facilitated by the ready availability of phagemid vectors, which allow for construction and display of libraries of all of these types of antibody fragments using a single rare cutting restriction enzyme, S’Z. Selection of interesting members from the library based on the displayed antibodies’ binding specificity and affinity is generally performed over several rounds of selection and amplification in a process known as panning. Antibodies have been sleeted from immune humans [7-16,17*,18,19*,‘20**,21~], macaques (the variable region genes of macaques one highly homologous to those of humans, making cloning from immune macaques an attractive approach to the development of human antibodies [Z]), mice [23**,24,25’,26’,27’*,28*,29”], rabbits [30], and chickens [31*], as well as from large naive libraries [32-34,3Soo]. This review focuses on the most interesting studies using this technique published in 1996. Introduction Selection Monoclonal antibodies were isolated from the first combinatorial antibody library expressed by phage in 1989 [l]. Driven by the success of this initial report, incorporation of the filamentous phage display method of Smith into general practice was rapid [Z]. It should be noted that at the beginning there was much skepticism concerning both the likelihood of finding high affinity antibodies within a random combinatorial library of antibody fragments and the ability to display a 50 kDa heterodimeric Fab protein on the surface of filamentous phage. Some of the most skeptical voices of this time are now among the technique’s most prolific practitioners. There are now more than 300 published examples in the literature for this technique, which forms the basis of an annual practical course at the Cold Spring Harbor Laboratory. While in the past the selection of phage displayed antibody libraries was mainly restricted to panning on purified antigens immobilized on artificial surfaces, an increasing number of antibodies have now been selected by panning on cell surface expressed antigens [9,11,18,23**]. Cell panning has been used when the antigen was not available in purified form or when it has yet to be characterized at all, in other words in the search for new cell markers. For example, the majority of antigens expressed on the surface of human red blood cells are defined only serologically and are not available in purified form. Cell panning of combinatorial antibody libraries derived from auto- and alloimunized individuals has been used to select for these antigens [18]. Moreover, cell panning has become important for the selection of antibodies with therapeutic potential. In contrast to panning on purified immobilized antigen, cell panning selects for antibodies that are more likely to bind to epitopes accessible in &XX In addition, as a result of homophilic or heterophilic interactions of the antigen on the cell surface, selection is focused on biologically relevant epitopes. This has been suggested for the HIV-l envelope glycoprotein gp120. Oligomeric gp120, as presented on the HIV-l virion or an HIV-l infected cell, was found to select HIV-l neutralizing antibodies from a phage library much more efficiently than monomeric gpl20 [36*]. In fact, Parren et al. [36*] suggested that the ability of antigen preparations to select The selection of antibodies displayed an important methodology diagnostic, from combinatorial on the surface of filamentous for the generation and therapeutic natural immune responses. molecules libraries phage has become of reagent, and for the study of Using this technique, genes have been cloned from multiple species directly from large man-made repertoires genes. Recent studies demonstrate for the in vitro evolution of antibodies antibody or expressed of antibody-encoding that the technique allows to create molecules whose affinity for antigen exceeds that observed in nature. Addresses The Skaggs institute for Chemical Biology and the Department of Molecular Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, California 92037, USA *e-mail: carlos@scripps.edu Current Opinion in Biotechnology 1997, 8:503-508 http://biomednet.com/elecref/0958186900800503 0 Current Biology Ltd ISSN 0958-l 669 The basic principles of the filamentous phage display technique are illustrated in Figure 1. A large library of phage is prepared following the cloning of genes encoding antibody heavy and light chain fragments; because heavy and light chains are recombined randomly, we talk about combinatorial antibody libraries. Each phage has the potential to display a unique antibody with a specific antigen binding site on its surface, which is the selectable phenotype of interest. Within the phage coat is the genotype that encodes the displayed molecule. This linkage of displayed antibody phenotype with strategies: cell panning and beyond 504 Protein engineering Figure 1 (a) V V Helper phage Male E. coli Xba I Xho I Sfi I Sfi I Gene Ill fragment trP Sfi I Fab or scFvsfi Vectors designed for the expression of antibody fragments I 0 1997 Current Opinion in Biotechnology on the phage surface or in soluble form. Typically, phagemid vectors have been utilized for the display of combinatorial antibody libraries on the surface of filamentous phage. Antibody fragments are fused to the carboxy-terminal domain of the minor coat protein of filamentous phage, the gene Ill protein, which is displayed in low copy number at one end of the phage. Whereas the amino-terminal domain of the gene III protein binds to the F’ pili, allowing for infection of male E. coli, the membrane-bound carboxy-terminal domain is important for phage promoter drives the synthesis of a dicistronic transcript. translation of two separate polypeptide chains encoded a heavy chain (VR-C&l) fused to the carboxy-terminal assembly. (a) In the Fab display version of pComb3H 161 a single IacZ Two ribosome binding sites (Shine and Dalgarno sequences; SD) give rise to the sequentially. One encodes a complete light chain (VL-CL), the other the Fd fragment domain of the gene Ill protein. The gene Ill fragment is followed of by the frp transcriptional terminator. To generate a combinatorial antibody library, light chain and Fd fragment encoding sequences are amplified independently and combined randomly in pComb3H. The Sacl and Xbal restriction enzyme sites are provided for cloning light chain encoding sequences. The Xhol and Spel sites are provided for cloning Fd fragment encoding sequences. Alternatively, the sequences encoding light chain and Fd fragment can be fused by overlap extension PCR and cloned directionally using two asymmetric sites of the rare eightbase cutter Sfi I. The leader peptides ompA and pelB target both polypeptides to the periplasm of E. coli, where the soluble light chain and the membrane-bound Fd fragment associate via a disulfide bond. The transformation of male E. coli with Fab encoding pComb3H and the subsequent addition of helper phage leads to the production of phage that, as their phenotype, display one copy of Fab fragment linked to the phage surface by the gene Ill protein fragment and which contain the corresponding single-stranded phagemid as their genotype. (b) In the scFv display version of pComb3H the variable immunoglobulin domains of light chain (Vt) and heavy chain (V,), are fused by a peptide linker (L). While a long, flexible linker allows for the monovalent scFv display, a shorter linker leads to the formation of scFv dimers (diabodies) with two antigen-binding sites. (C) Digestion of pComb3H, encoding a selected Fab or scFv, with Spel and Nhel and self ligation removes the gene Ill fragment and allows for the production of soluble Fab or scFv. Alternatively, Fab- or scfv-encoding sequences are generally subcloned using the Sfil sites for high-level expression into the compatible expression vector pPhoA-HGHA [6], which adds a histidine tag (H6) to the amino terminus and a hemagglutinin decapeptide (HA) to the carboxyl terminus of the expressed antibody fragment for detection and purification. Phage display of combinatorial antibody libraries neutralizing antibodies from phage libraries could be used as a measure of the quality of the antigen as a vaccine. Taking a great step forward, Pasqualini and Ruoslahti [37**] developed a methodology for the in vtio panning of phage displayed peptide libraries in order to select peptides that home to target organs. Phage were injected in the tail vein of mice, allowed to circulate for a few minutes, and were isolated and reamplified from brain and kidney tissue after the mice had been snap frozen in liquid nitrogen. After three rounds of in &JO panning, peptide sequences that selectively target brain and kidney were identified. Because the phage were allowed to circulate for but a few minutes, it is unlikely that they left the blood vessels; therefore, the peptides were assumed to bind cell surface molecules selectively expressed on the endothelial cells of brain and kidney [37”]. In principle, in vjvo panning should be feasible using phage displayed antibody libraries too. Selective targeting of endothelial markers by peptides or antibodies promises applications in gene and cancer therapy. Cell surface molecules selectively expressed on mature or angiogenic endothelial cells that form the blood vessels of tumors have become a promising target for cancer therapy [38]. Only a few molecular markers of tumor blood vessels are known, however, making in viva panning of phage displayed peptide and antibody libraries a promising too1 for their identification. Peptides and antibodies selected by tumor targeting will find many diagnostic and therapeutic applications. Antibodies to clinically relevant targets Since their first application to cloning antibody segments involved in the human immune response by the laboratories of Barbas, Burton, and Lerner, phage antibodies have become an important resource for the development of therapeutic antibodies and the study of the human antibody repertoire as it relates to disease. Antibodies have been selected to bind the classical therapeutic targets, tumor antigens [9,25’,26*,32] and viral antigens (reviewed in [7]; [8,15,17’,21*]). Furthermore, this methodology has facilitated studies of human immune responses and autoimmunity [10,11,13,14,16,19*]. The first human IgE antibodies with known specificity have now been isolated from phage displayed antibody libraries [ZO”]. IgE Fabs specific for a grass pollen allergen were selected by screening a phage displayed IgE Fab library derived from peripheral blood lymphocytes of an allergic individual. By interfering with the binding of IgE polyclonal antibodies to allergens, IgE Fab selected from phage displayed antibody libraries may become a therapeutic tool for the treatment of allergies [ZO**]. Another antibody directed to a clinically relevant target that became accessible only through phage display is one directed to the human platelet alloantigen 1 located on integrin aIIbB3 [12]. Human platelet alloantigen 1 can be encoded by two alleles that differ in one nucleotide. Due to this difference, 98% of the Caucasian Rader and Barbas 505 population have a leucine at position 33 of cxI&, while 2% have a proline at this position. Pro33 homozygotes are at risk of producing Leu33-specific antibodies after transfusion or during pregnancy. Monoclonal antibodies directed to native O(II& Leu33 would be an important tool for phenotyping and therapeutic applications, but have not been obtained by hybridoma technology. Griffin and Ouwehand [12] have now selected an anti-o&B3 Leu33 human scFv by screening a phage displayed scFv library derived from peripheral blood lymphocytes of an alloimmunized pregnant individual. As shown by flow cytometry, the isolated scFv selectively binds to platelets expressing ~I& LeU33. A particularly interesting report details the generation of an antibody that selectively binds to a peptide-MHC complex [23”]. For this, a phage displayed Fab library was generated from spleens of mice that had been immunized with a peptide-MHC complex consisting of an influenza virus derived hemagglutinin (HA) peptide and the mouse MHC class I molecule Kk. To eliminate unspecific binders efficiently the library was panned on cells that displayed HA-Kk, alternating with panning on HA-Kk-coated microspheres. One of the selected Fabs, Fab13.4.1, bound to HA-Kk with an affinity of about 50nM. Neither isolated HA, isolated Kk, nor a complex between Kk and a different peptide was bound by Fab13.4.1. Thus, Fab13.4.1 binds a particular peptide-MHC complex selectively, an activity reserved to T-cell receptors in viva. Antibodies specific for peptide-MHC complexes will be important diagnostic and therapeutic tools: the contents of cells containing intracellular parasites or aberrant proteins related to a cancer can now be seen with a soluble antibody and, if desired, the cell can be targeted for destruction by antibody effector functions or tethered toxins. Directed evolution of antibody affinity It is generally assumed that the ability to create antibodies of higher affinity will provide therapeutic and diagnostic benefits. Phage display provides a format for the directed molecular evolution of antibody affinity (reviewed in [7]). While low affinity antibodies are readily improved with respect to affinity, antibodies that already possess nanomolar binding of their antigen require more subtle refinements to improve their already high affinity [7]. Several strategies that have been recently explored are saturation mutagenesis of complementarity determining regions (CDR) (CDR walking mutagenesis) [6,39”,40,41], chain shuffling [41-44], and the use of mutator strains of Eschric/ria co/i [45*], which provide mutations in r&o akin to error-prone PCR in vim. There appears to be consensus that the CDR walking mutagenesis strategy is the most reliable of affinity maturation methodologies yet devised [6,39”]. This approach has been applied to human anti-HIV [6] and anti-c-erbB-2 [39*.] antibodies to create antibodies with monovalent picomolar affinities not found in nature. In this approach, saturation mutagenesis of a CDR is constrained to libraries that examine all 506 Protein engineering possible amino acids in the target CDR. Two strategies are evident, either sequential or parallel optimization of CDR. In the sequential approach, the library of antibodies with a single randomized CDR is screened by several rounds of panning against the antigen. The selected clone(s) are then used in the construction of a second library where a different single CDR is randomized. The resulting library is screened by additional rounds of panning, and the selected clone(s) are used in the construction of a third library and so on. Sequential optimization takes into account that optimal binding may result from the interdependence of CDRs. In the parallel approach, independent libraries are constructed where each library represents the randomization of a given CDR. Each library of antibodies is screened by several rounds of panning against the antigen. Then, the individually optimized CDRs are combined. If the free energy change of individually optimized CDRs combined is nearly equal to the sum of the free energy changes observed in the single optimized CDRs, the free energy changes are said to be additive. As additivity within the antibody binding site is virtually impossible to predict, sequential CDR optimization is preferred over parallel. To date the greatest gains in affinity have resulted from the optimization of heavy chain CDR3 and light chain CDR3 regions, but there are sure to be many exceptions. An advantage of this strategy over PCR or mutator strain induced mutagenesis is that modifications to the parental antibody are constrained to the hypervariable regions. It is expected that changes in the primary sequence in these regions are less likely to generate immunogenic antibodies than changes in the more sequence constrained framework regions. Acknowledgements Conclusions Methodologies for the creation and selection of combinatorial antibody libraries on the surface of phage continue to be improved and extended. Antibodies have been selected to molecules both large and small to study protein recognition, immunity, and problems in basic and applied sciences. While it was anticipated that in viva recombination of heavy and light chain sequences based on Cre/lox strategies [46] would allow for the creation of libraries that exceed 1012 molecules, in practice this approach has yet to demonstrate its potential because of severe problems with the genetic stability of the vectors. These problems are not, however, insurmountable. Other interesting strategies are evolving that enable the cloning of genes encoding both antibodies and their antigens using phage display. For example, Krebber et a/. [47] investigated a phage display system with the potential to examine library versus library interactions. It may be possible in the future to clone a cDNA together with a specific antibody which binds it. There is still hope that phage display of combinatorial antibody libraries “will be the component of a machine that, before the turn of the century, will evolve proteins on demand to meet new challenges in biotechnology” [4]. Christoph Rader was supported by the Swiss National Science Foundation and Carlos F Barbas III by grants from the National Institute of Health and an Investigator Award from the Cancer Research Institute. References and recommended reading Papers of particular interest. published within the annual period of review, have been highlighted as: . .. of special interest of outstanding interest 1. Huse WD. Sastrv L. lverson SA Kana AS. Altina-Mees M. Button Dd, Be&o& SJ, Lerner’ RA: &em&on of a large combinatorial library of the immunoglobulin repertoire in phage lambda. Science 1969, 246:1275-l 261. 2. Smith GP: Filamentous fusion phage: novel expression vectors that display cloned antigens on the surface of the virion. Science 1965, 226:1315-l 317. 3. Barbas CF Ill: Recent advances Biotechnol 1993, 4:526-530. 4. Burton DR, Barbas CF Ill: Human antibodies libraries. Adv lmmunol 1994, 57:191-280. 5. Hoogenboom HR: Designing and optimizing library selection strategies for generating high-affinity antibodies. 7iends Biotechnoll997, 15:62-70. 6. Yang WP, Green K, Pinz-Sweeney S, Briones AT, Burton DR, Barbas CF Ill: CDR walkin mutagenesis for the affinity maturation of a potent human anti-HIV-1 antibody into the picomolar range. J Mel Bioll995, 264:392-403. 7. Barbas CF Ill, Burton DR: Selection and evolution of highaffinity human anti-viral antibodies. Trends Biotechnol 1996, 141230-234. 8. Binley JM, Ditzel HJ, Barbas CF Ill, Sullivan N, Sodroski 1, Parren PWHI, Burton DR: Human antibody responses to HIV-1 ~~41 cloned in phage display libraries suggest three major epitopes are recognized and give evidence for conserved antibody motifs in antigen binding. AIDS Res Hum Retroviruses 1996, 12:911-924. 9. Cai X. Garen A: A melanoma-soecific VH antibodv cloned from a fusion phage library of a vaccinated melanoma- patient Proc Natl Acad Sci USA 1996, 93:6260-6285. 10. Duenas M, Chin LT, Malmborg AC, Casalvilla R, Ohlin M, Borrebaeck CAK: In vitro immunization of naive human B cells yields high affinity immunoglobulin G antibodies as illustrated by phage display. hmunology 1996, 89:1-7. 11. Eggena M, Targan SR, lwanczyk L, Vidrich A, Gordon LK, Braun J: Phage display cloning and characterization of an immunoaenetic marker (oerinuclear anti-neutrophil cytoplasmicantibody) in ulcerative colitis. J lmmunbl 1996, 156:4005-4011. 12. Griffin HM, Ouwehand WH: A human monoclonal antibody specific for the leucine-33 (PI Al, HPA-la) form of platelet glycoprotein llla from a V gene phage display library. Blood 1995, 86:4430-4436. 13. McIntosh RS, Asghar MS, Watson PF, Kemp EH, Weetman AP: Cloning and analysis of IQGK and IgGh anti-thyroglobulin autoantibodies from a patient with Hashimoto’s thyroiditis: evidence for in viva antigen-driven repertoire selection. J lmmunol 1996, 157:927-935. 14. Ohlin M, Borrebaeck CAK: Characteristics of human antibody repertoires following active immune responses in viva. MO/ lmmunoll996, 33:563-592. 15. Pilkington GR, Duan L, Zhu M, Keil W, Pomerantz RI: Recombinant human Fab antibody fragments to HIV-1 Rev and Tat regulatory proteins: direct selection from a combinatorial phage display library. Mel Immunol 1996, 33:439-460. 16. Roben P, Barbas SM, Sandoval L, Lecerf JM, Stellar BD, Solomon A, Silverman GJ: Repertoire cloning of lupus anti-DNA autoantibodies. J C/in Invest 1996, 98:2827-2637. 1z . Samuelsson A, Yari F, Hinkula J, Ersoy 0, Norrby E, Persson MAA: Human antibodies from phage libraries: neutralizing activity against human immunodeficiency virus type 1 equally in phage display. Curr Opin from combinatorial Phage display of combinatorial antibody libraries Rader and Barbas improved after expression as Fab and IgG in mammalian cells. Eur J lmmunoll996, 26:3029-3034. This report suggests that Fab expressed in E. co/i is not identical, in tens of binding and virus neutralization, to antibody expressed in eukaryotic cells. Antibody derived from expression in a eukaryotic host may be superior in terms of the fraction of properly folded protein that is produced as compared with that expressed in prokaryotes. This observation may be antibody dependent. 16. Siegel DL: Isolation of human anti-red blood cell antibodies by repertoire cloning. Ann NY Acad Sci 1995, 764:547-556. Silverman GJ, Pires R, Bouvet JP: An endogenous sialoprotein and e bacterial B cell superantigen compete in their VH familyspecific binding interactions with human Igs. J lmmunol 1996, 167:4496-4502. An interesting report on the investigation of superantigen interactions as probed with libraries. Williamson RA, Peretz D, Smorodinsky N, Bastidas R, Serban H, Mehlhom I, DeArmond SJ, Prusiner SB, Burton DR: Circumventing tolerance to generate autologous monoclonal antibodies to the prion protein. Proc Nat/ Acad Sci USA 1996, 93:7279-7262. Construction of combinatorial libraries from prion knockout mice and immunization with prion protein enables the first selection of high-affinity antibodies to this important protein. This strategy may be an important one for the generation of antibodies to highly conserved proteins. 29. .. 30. 19. . Steinberger P, Kraft D, Valenta R: Construction of e combinatorial IgE library from an allergic patient J Biol Chem 1996, 271 :10967-l 0972. Due to their low frequency within the expressed antibody repertoire, production of monoclonal IgE antibodies of known specificity had not been achieved previously. This study may lead to advances in the treatment of allergy. 20. .. Tsui P, Tometta MA, Ames RS, Bankosky BC, Griego S, Silverman C, Porter T, Moore G, Sweet RW: Isolation of a neutralizing human RSV antibody from e dominant nonneutralizing immune repertoire by epitope-blocked panning. J lmmunoll996, 167:772-760. Often selection from libraries can lead to the isolation of antibodies directed against a dominant epitope. By blocking a dominant epitope with a previously isolated antibody and repanning, antibodies against rare epitopes may be recovered. This study confirms the efficacy of the epitope masking procedure described earlier in studies of anti-HIV-l antibody selection. Samuelsson A, Chiodi F, Ohman P, Putkonen P, Norrby E, Persson MAA: Chimeric macaque/human Feb molecules neutralize simian immunodeficiency virus. Virology 1995, 207:495-502. Andersen PS, Stryhn A, Hansen BE, Fugger L, Engberg J, Buus S: A recombinant antibody with the antigen-specific. major histocompatibility complex-restricted specificity of T cells. Proc Nat/ Acad Sci USA 1996, 93:1620-l 624. A mouse Fab that binds selectively to a mouse peptide-MHC complex was selected by panning a Fab library derived from immune mice. Antibodies specific for peptide-MHC complexes will be important diagnostic and therapeutic tools. 23. .. 24. Bespalov IA, Purmal AA, Glackin MP, Wallace SS, Melamede RJ: Recombinant Phabs reactive with 7, 6-dihydro-6-oxoguanine, a major oxidative DNA lesion. Biochemistry 1996, 35:2067-2076. Dinh Q, Weng NP, Kiso M, lshida H, Hasegawa A, Marcus DM: High affinity antibodies against Lex and sialyl Lex from a phage display library. J lmmunol 1996, 167:732-736. High-affinity anticarbohydrate antibodies have been thought to be rare, if not impossible, to produce. This report demonstrates that selection allows for the rapid isolation of this interesting class of antibodies. Yamanaka HI, lnoue T, Ikeda-Tanaka 0: Chicken monoclonal antibody isolated by a phage display system. J lmmunoll996. 157:1156-l 162. Cloning antibodies from immune chickens may allow for isolation of specificities not generated in other animals because of conservation. Chicken libraries are by far the easiest to prepare since chickens possess a single functional VH and VL gene which is diversified via gene conversion. Primers are available for cloning chicken antibodies in the pComb3H phagemid vector. 32. Carnemolla B, Neri D, Castellani P, Leprini A, Neri G, Pini A, Winter G, Zardi L: Phege antibodies with pan-species recognition of the oncofetal angiogenesis marker fibronectin ED-B domain. Int J Cancer 1996, 66:397-405. 33. Fan J, Griffiths AD, Lockhart A, Cross RA, Amos LA: Microtubule minus ends can be labelled with a phage display antibody specific to alpha-tubulin. J MO/ Biol 1996, 269:325-330. 34. Stausbol-Gron B, Wind T, Kjaer S, Kahns L, Hansen NJV, Kristensen P, Clark BFC: A model phage display subtraction method with potential for analysis of differential gene expression. FEBS Lett 1996,391:71-75. Vaughan TJ, Williams AJ, Pritchard K, Osboum JK, Pope AR, Eamshaw JC, McCaffetty J, Hodits RA, Wilton J, Johnson KS: Human antibodies with sub-nanomolar affinities isolated from a large non-immunized phage display library. Nat Biotechnol 1996, 14:309-314. Antibodies of good affinity may be directly selected from large repertoires of V genes. This report involves the selection of scFv fragments which often aggregate, leading to an overestimate of their monovalent binding affinity. A general rule in combinatorial antibody selection is if your application demands a SCFV, select a scFv. lf your application requires a Fab or a whole antibody, select from a Fab library. Antibodies selected as scFv do not generally bind their antigen well when they are converted for expression in the Fab or whole antibody format. 35. .. 36. . Lorimer IA, Keppler-Hafkemeyer A, Beers RA, Pegram CN, Bigner DD, Pastan I: Recombinant immunotoxins specific for a mutant epidermal growth factor receptor: targeting with a single chain antibody variable domain isolated by phage display. Proc Nat/ Acad Sci USA 1996, 93:14615-l 4620. A mouse scFv that binds to a mutant human epidermal growth factor receptor was selected by panning an scFv library derived from mice immunized with a synthetic peptide. An immunotoxin produced by fusing the selected scFv to Pseudomonas exotoxin A revealed good affinity, cytotoxicity, and stability. 27. .. McGuinness BT, Walter G, FiizGerald K, Schuler P, Mahoney W, Duncan AR, Hoogenboom HR: Phage diabody repertoires for selection of large numbers of bispecific antibody fragments. Nat Biotechnoll996, 141149-l 154. Study of a model system for the selection of dimeric scFvs that possess binding sites with two unique specificities. This technique holds potential for the development of antibodies that cross-link molecules on the cell surface or act as molecular adhesives. 26. . Tang Y, Jiang N, Parakh C, Hilvert D: Selection of linkers for a catalytic single-chain antibody using phage display technology. J Biol Chem 1996, 271 :15662-l 5666. Expression of antibodies in the scFv format is often an unsuccessful undertaking. This report demonstrates that selection of the linker sequence connecting the variable region fragments affects both affinity for antigen and the level of scFv expression. Parren PWHI, Fisicaro P, Labrijn AF, Binley JM, Yang WP, Ditzel HJ, Barbas CF Ill, Burton DR: In vitro antigen challenge of human antibody libraries for vaccine evaluation: the human immunodeficiency virus type 1 envelope. J Viral 1996, 70:9046- 25. . 26. . Lang IM, Barbas CF Ill, Schleef RR: Recombinant rabbit Fab with binding activity to type-l plasminogen activator inhibitor derived from a phage-display library against human alphagranules. Gene 1996, 172:295-296. 31. . 21. . 22. 507 9050. Based on the HIV-l -neutralizing ability of human Fab that were selected by panning a Fab library derived from a HIV-1 positive individual on different preparations of the HIV-l envelope glycoprotein, the authors suggest that the ability of antigen preparations to select neutralizing antibodies from combinatorial antibody libraries could be used as a measure of the quality of the antigen as a vaccine. 37. Pasqualini R, Ruoslahti E: Organ targeting in viva using phage .. display peptide libraries. Nature 1996, 360:364-366. Peptides that home to target organs were selected by in viva panning of phage displayed peptide libraries, a technique that should also be feasible with phage displayed antibody libraries. Selective in viva targeting by peptides or antibodies promises applications in gene and cancer therapy. 36. Barinaga M: Designing therapies that target tumor blood vessels. Science 1997, 2751462-464. 39. .. Schier R. McCall A, Adams GP, Marshall KW, Merritt H, Yii M, Crawford RS, Weiner LM, Marks C, Marks JD: Isolation of picomolar affinity anti-c-erbB-2 single-chein Fv by molecular evolution of the complementarity determining regions in the center of the antibody binding site. J MO/ Biol 1996, 263:551567. CDR walking mutagenesis focused on LCDRB and HCDR3 leads to large gains in affinity for this therapeutically interesting s&v. See also [6] for a comprehensive investigation of the CDR walking mutagenesis strategy as it relates to additivity within antibody combining sites. Currently, this strategy has led to the greatest success in improving the affinity of already high affinity antibodies. 40. Short MK, Jeffrey PD, Kwong RF, Margolies MN: Contribution of antibody heavy chain CDRl to digoxin binding analyxed by 508 protein engineering light chains to restriction of antibody recognition of the HIV-1 principal neutralization determinant J lmmunoll996, 156:18761603. random mutagenesis of phage-displayed Fab 26-10. J Biol Chem 1995,270:29541-29550. 41. ThompsonJ, Pope T, Tung JS, Chan C, Hdlis G, Mark G, JohnsonKS: Affinity maturation of a high-affinity human monoclonal antibody against the third hypervariable loop of human immunodeficiency virus: use of phage display to improve affinity and broaden strain reactivity. J MO/ Biol 1996, 256~77-00. 42. Ohlin M, Owrnan H, Mach M, Borrebaeck CAK: Light chain shuffling of a high affinity antibody results in a drift in epitope recognition. Mel Immunoll995, X%47-56. 43. Schier R, Bye J, Apell G, McCall A, Adams GP, MalmqvistM, Weiner LM, Marks JD: Isolation of high-affinity monomeric human anti-c-erbB-2 single chain Fv using affinity-driven selection. J MO/ Biol 1996, 255:26-43. _ 44. Watkins B4 Davis AE, Fiorentini S, di MarzoVeronaseF, Reitz MS Jr: Evidence for distinct contributions of heavy and 45. . Low NM, Holliger P, Winter G: Mimicking somatic hypermutatlon: affinity maturation of antibodies dlsplayed on bacteriophage using a bacterial mutator strain. J MO/ Biol 1996, 260:359-366. By multiple rounds of growth in the mutator strain E. co/i mutD5 followed by panning,the affinityof a human scFv to the hapten 2-phenyl-5-oxazolone was improved. 46. TsurushitaN. Fu H. Warren C: Phase disolaY vectors for in viva recombinatibn of immunoglobulihhea~ and light chain genes to make large combinatorial libraries. Gene 1996, 172:59-63. 47. Krebber C, Spada S, Desplancq D, PluckthunA: Co-selection of cognate antibody-antigen pairs by selectively-infective phages. FEBS Leti 1995, 377~227-231.