REVIEW I M M U N O L O G Y TO D AY

Natural and designer binding sites made by phage display technology Hennie R. Hoogenboom and Patrick Chames In the past decade, the drive to develop completely human

T

large collections of antibody variable dohe creation of binding sites antibodies for human therapy has mains was needed; this was first described to order has been at the heart led to the development of phage in 1989, when partially degenerate oligoof many immunological and display technology. This technology nucleotides priming to the 59 and 39 end of biotechnological procedures. variable-region genes and the polymerase By immortalizing antigen-specific B cells, is able to deliver the ultimate in chain reaction (PCR) were used to amplify classical hybridoma technology harvests antibody engineering, that is, highhybridoma6,7 or large collections of variable binding sites that have been built, selected affinity fully human antibodies to genes8,9. Second, as whole antibodies could and optimized for binding to antigen by the not be functionally expressed in bacteria, a immune system. Monoclonal antibodies any antigen of choice. Here, this crucial discovery was that antibody frag(mAb) are thus made by immunizing aniapplication of phage display ments (Fab or single-chain Fv; Fig. 1) were mals and allowing in vivo processes, such functionally expressed in Escherichia coli as immune tolerance and somatic hypertechnology is reviewed, and the when they were secreted into the periplasm mutation, to shape the antigen combining many other antibody-engineering of the bacteria, which simulated the natusite. These mAb, with binding sites created avenues this technology offers are rally oxidizing environment of the endoplasin vivo, can be used for many research and mic reticulum10,11. By providing restriction diagnostic applications1. Monoclonal antihighlighted. bodies might be made less immunogenic, sites in the oligonucleotides used for PCR and therefore often more effective for human amplification, antibody libraries could thus therapy, by reformatting the binding site into chimaeric2 or comple- be cloned for expression in E. coli. Initially, such antibody libraries mentary determining region (CDR)-grafted antibodies3. Alterna- were expressed from phage lambda vectors9; a plaque-screening tively, the two variable domains of the binding site can be cloned assay with labelled antigen was then used to identify antigenand arranged into a large array of possible molecular formats and specific binding sites. Such time-consuming procedures were sizes, and expressed in a variety of hosts, ranging from bacteria, rapidly replaced by the third seminal development: the provision of lower eukaryotes such as yeast and fungi, to the higher eukaryotes, a link between phenotype and genotype, using phages. In 1990, including mammalian cells, transgenic animals and plants (reviewed McCafferty et al. showed that antibody fragments could be disin Ref. 4, see also Little et al., this issue). played on the surface of filamentous phage particles by fusion of the Over the past decade, molecular display technologies have been antibody variable genes to one of the phage coat proteins12. Antigendeveloped that allow us to create fully human antibodies for therapy specific phage antibodies could subsequently be enriched by mulwith ease. This article addresses how these technologies are now tiple rounds of affinity selection, because the phage particle carried revolutionizing the way in which we can build high-affinity binding the gene encoding the displayed antibody. This was originally sites from scratch, from any species (including humans), possibly reported for single-chain Fv fragments12, and later for Fab fragin an automated fashion, without the constraints imposed by ments13–15 and other antibody derivatives such as diabodies16, as the immune system, and with in-built features for downstream well as extended to various display systems (Fig. 1). With these findings in place, it became possible to make phage antibody applications. libraries by PCR cloning of large collections of variable-region genes, expressing each of the binding sites on the surface of a different Key discoveries in the development of antibody phage phage particle and harvesting the antigen-specific binding sites by in display vitro selection of the phage mixture on a chosen antigen. In the early Making and selecting antibody libraries 1990s, Winter et al. showed for the first time that phage display techThe concept of molecular display technology is the provision of a nology could be used to select antigen-specific antibodies from physical linkage between genotype (the antibody variable region libraries made from the spleen B cells of immunized mice17, thereby genes) and phenotype (antigen-binding) to allow simultaneous se- bypassing the requirement to immortalize the antigen-specific B lection of the genes that encode a protein with the desired binding cells, as in the hybridoma technology. Similarly, libraries were made function. This concept was successfully applied to small peptides by from human B cells taken from individuals immunized with antiSmith5, but the display of functional antibody repertoires on phages gen18, exposed to infectious agents19, with autoimmune diseases20 or required several additional discoveries. First, a procedure for accessing with cancer21. PII: S0167-5699(00)01667-4

0167-5699/00/$ – see front matter © 2000 Elsevier Science Ltd. All rights reserved.

A U G U S T Vo l . 2 1

No.8

371

REVIEW I M M U N O L O G Y T O D AY

Immunized

Disease related

Naïve

Cloned

A C C T G T A G

V-gene source

and

mRNA is isolated from B cells and used for cDNA synthesis Natural

CDR origin

Gene is synthesized

Mixed

Synthetic

mRNA V-gene

V-genes are amplified and re-assembled in functional antibody format scFv Fragment format

VH

Fab VL

VL

CL

VH

CH1

scFv or Fab genes are cloned into a display vector Phagemid (p8)

Phage (p3)

Phagemid (p3)

Yeast or E. coli

Ribosome

Display format

Immunology Today

Fig. 1. Choices when constructing a binding-site library. Before building a repertoire of binding sites, it is necessary to make several choices that are guided by the future application of the repertoire. The source of the variable domain genes and of the type of CDRs included will determine the specificity and frequency of antigen-specific clones in a binding-site library, as well as the quality (affinity, frequency of nucleotide mutations) of selected clones. High affinity antibody binding sites to immunogenic or disease-related antigens may be retrieved more easily from immune or disease-related libraries, respectively (left), while naïve or synthetic libraries (right) are a reliable source of antibodies to many different antigens. The antibody genes are amplified from their natural and synthetic sources by PCR (left) or assembled using synthetic oligonucleotides (right), or the natural and synthetic diversities are combined (middle). The format of the fragment displayed can also influence the outcome of the selection, and each format, scFv (right) or Fab (left), has pros and cons (in terms of multimerization, which is pronounced for the former, ease of cloning of the VH and VL genes, and production, size and stability, which may have their effect on the display level). Finally, the display format, which can be chosen to be either phage display using a major coat protein present in many copies (p8) or a minor coat protein with few copies (p3), displayed on phage (for p3 display) or phagemid particles (for p3 and p8 display), will have its impact on the selection outcome. With multivalent display system such as that achieved using phage p8, it will be more difficult to select higher affinity clones owing to possible avidity effects. The most frequently used systems are based on the phagemid systems using p3 display (middle), because these allow a high frequency of monovalent display, as well as an easy conversion from phage-displayed antibody to a soluble, secreted antibody. With cellular display systems, cell sorting might be used to enrich affinity variants, while ribosome display offers a complete cell-free molecular evolution system for antibodies (right).

Phage antibodies from nonimmune libraries Thus, phage display technology in the early 1990s had already shown the potential to replace the hybridoma technology by rescuing V-genes from immune B cells. Within a period of two years, however, two other seminal findings were reported that would bypass the use of immunization and animals altogether. First, it was shown that antibodies against many different antigens could be selected from a ‘naïve’ binding-site library, made from the light-chain and heavy-chain IgM–V-gene pools of B cells of a nonimmunized,

A U G U S T 372

Vo l . 2 1

No.8

healthy individual22 (Fig. 1). Second, libraries of synthetic antibody genes, with variable genes not harvested from immune sources but consisting of germline segments artificially provided with diversity by oligonucleotide cloning13,23 (Fig. 1), were shown to behave in a similar way to ‘naïve’ antibody libraries. It thus became possible to use primary (‘naïve’ or ‘synthetic’) antibody libraries, with huge collections of binding sites with different specificities, to select in vitro binding sites against most antigens, including nonimmunogenic molecules, toxic substances and targets conserved between species

REVIEW I M M U N O L O G Y TO D AY

(reviewed in Ref. 24). Since these key discoveries, there have been numerous reports on applications of phage antibody libraries25,26, ranging from basic research to drug development. In addition, many novel, related molecular display methods for antibodies have been described, including display systems on ribosomes27, bacteria28 and yeast cells29 (Fig. 1). These technologies follow similar concepts for in vitro selection and improvement of binding sites, as will be discussed in detail for phage display technology.

9

1

2

8

4 7 3

Selecting the binding site Selection for binding using purified antigen

6 5

The selection step is a key feature of any molecular display technology. A major adImmunology Today vantage of this in vitro procedure is that we can choose and manipulate the procedure and therefore the outcome of the selection Fig. 2. Principle of selecting and improving binding sites. A diverse repertoire containing from (reviewed in Ref. 30). Selections with puri- several millions to billions of different binding sites is incubated with the target (1), immobilized on fied antigen are essentially made in two a solid surface in this example. Several washing steps remove the irrelevant phage antibodies (2), and ways. The antigen can be attached to a solid the binders are then separated from the antigen, typically using a pH shock (3). Binders are then support (to plastic, by adsorption, or to amplified by infection and multiplication in E. coli (4) to generate a new repertoire highly enriched beads or a column matrix), the phage library in binding sites specific for the antigen. This cycle is repeated two to four times until binders domirun over the support, and antigen-bound nate the population and can be characterized as monoclonal antibodies (5). To affinity-mature these phages retrieved after rinsing the support by selected antibodies, their genes are isolated from the display vector (6) and diversity is introduced, for elution with acid or alkaline solutions. Alter- example, by shuffling one of the domains, or by using site-directed mutagenesis (or other methods natively, the binding event can take place described in the text) (7). This new repertoire is cloned (8) and phage particles of the new repertoire with antigen in solution, for example, using produced (9), to be used in selection procedure (1–4) with increasing stringency, to identify higher affinity biotinylated antigen or unlabelled antigen. variants of the binding site. Antigen-bound phages might be retrieved by incubation with streptavidin-coated magnetic beads or other lig- to the antigen, or by depletion or subtraction procedures using ands that capture the antigen (e.g. antibodies). The choice of antigen an excess of the irrelevant antigens or epitopes, for example, by concentration and binding–washing times can drive selection to- fluorescence-activated cell sorting or magnetic-activated cell sorting. wards the highest affinity antibodies in the population. After a few Alternatively, natural ligands of known specificity might be used as rounds of selection, individual monoclonal phage antibody clones a ‘pathfinder’ to direct the selection towards phage particles binding can be screened for antigen binding in ELISA or other assays. This in the vicinity of the ligand34. Despite the availability of these elegant selection methods, it has process is depicted in Fig. 2. By choosing the right design of the selection procedure (e.g. the use of competing antibodies or ligands, been difficult to isolate high-affinity antibodies to some antigens, indepletion of irrelevant phage antibodies, and so on), antibodies with cluding carbohydrates and seven-transmembrane receptors. In some exquisite binding features can be selected, for example, antibodies instances, the antibody repertoire might have a structural limitation to unique epitopes on highly related glycoproteins31, antibodies to in what antigen types it can recognize; in others, the phage antibodies epitopes exposed upon activation of the protein32 or antibodies to might not be able to access the antigen35. unique major histocompatibility complex–peptide complexes33.

Selection for function Selection for binding using nonpurified antigen Many other sources of antigen have been used for selection, ranging from whole cells or cell lysates, tissues and protein blots, to bacteria that display recombinant antigen. When using such complex sources, selection of phage antibodies to irrelevant antigens or epitopes has to be reduced, for example, by specific elution using a known ligand

The in vitro procedure permits selection for function as well as binding capabilities. Selections can be performed under conditions that mediate the selection of phage antibodies with a particular characteristic, for example, under reducing conditions to retrieve disulfidefree yet stable antibodies36, or in the presence of proteases to select for well-folded molecules37. Antibodies might also be selected with

A U G U S T Vo l . 2 1

No.8

373

REVIEW I M M U N O L O G Y T O D AY

or for a particular functional activity, for example, for receptor crosslinking, signalling, gene transfer or catalysis. The availability of the cloned antibody genes further offers the possibility of combining phage selections (for binding) with other selections, such as a neutralizing activity of the antibody when expressed intracellularly38 (reviewed in Ref. 25).

Selection based on phage infectivity Besides these very successful in vitro selection methods, several methods have been described based on the restoration of phage infectivity using the antibody–antigen interaction. The idea is simple: only phage antibodies that bind to the antigen would provide the phage with the necessary elements (e.g. a functional minor coat protein) to restore infectivity of the phage particle. The farthest developed system, selectively infective phage, however, still needs a very high affinity between antibody and antigen to be effective, and is thus not yet suited for applications with most primary phage libraries (reviewed in Ref. 39). With an increase in the understanding of the phage infection process, in the future, other more generally applicable infectivity-mediated selection procedures based on bacterial antigen display and phage antibody libraries, might well be developed40.

Subtractive selection The availability of very large repertoires of different binding sites, in combination with in vitro selection procedures, have opened up applications in the field of target discovery. By subtractive selection, for example on tumour cell populations with subtraction on normal cells, antibodies to overexpressed and possibly novel cellsurface or tissue antigens can be isolated21. Many attempts to identify novel target antigens have been described, with one of the best illustrated applications the phage antibody-assisted cloning of CD55 as an overexpressed cell-surface target on lung carcinoma cells41. This application of large phage antibody libraries, their use as a large collection of molecular probes for target identification, is expected to yield valuable targets for both therapeutic and diagnostic applications.

High-throughput selection and screening As phage display selection is amenable to automation, highthroughput screening of selected populations has become possible. With appropriate equipment, antibodies to hundreds of different antigens, all appropriately tagged for detection, immobilization or purification, can be selected simultaneously. This has clear applications in functional genomics and proteomics research, not in the least to obtain binding sites specific for the many gene and protein products that need to be characterized in this field42. Furthermore, the availability of large collections of recombinant antibodies will aid the development of antibody-based chips and arrays43, opening up many new applications for display-derived binding-site libraries.

A U G U S T 374

Vo l . 2 1

No.8

Exploring the natural binding site Binding sites evolved by the immune system might be rescued from immune hosts and cloned using phage-display technology, as has been done for mice, rats, rabbits, camels and humans. In humans, phage antibody libraries made from donors who naturally mount an immune response (with viral infections, bearing tumours or with autoimmune disease) have been used to generate both human antibodies with high affinity and specificity for particular antigens, and to investigate the humoral response in disease. In these libraries, antigen exposure has increased the frequency of antigen-specific B cells, and somatic hypermutation might have affinity maturated the binding site, features that will increase the likelihood of success with the in vitro phage-selection technique. ‘Naïve’ antibody libraries have been more generally used. These are built from diverse lymphoid sources, including peripheral blood, bone marrow, spleen or tonsils, and thus harvest naturally assembled binding sites22. To minimize biases of the libraries, owing to the inclusion of V-genes of unrelated immune responses in the donor, the B cells can be enriched for IgM–IgD surface expression44, or the variable heavy-chain genes can be amplified from IgM mRNA (Ref. 22). Provided the repertoire constructed is diverse, larger libraries give rise both to many more different antibodies against a given antigen, and also to higher affinity antibodies sites (reviewed in Ref. 24). As a consequence, there have been several efforts to establish very large primary libraries, by brute-force cloning sites31,45,46 or by site-specific recombination procedures47,48, to yield collections of 109 to 1011 potentially different binding sites. From the best libraries, antibodies in the 1–200 nM range are routinely selected. These library sizes are close to the total number of clones that can be effectively sampled using in vitro selection, and larger libraries are not expected to yield much better antibodies. Instead, the binding sites can be fine-tuned in vitro for binding to their antigen in a process that resembles the in vivo affinity maturation process.

Improving the binding site The antibodies selected from many of the immune and the large ‘single pot’ repertoires are often directly useful for the researcher, for example for ELISA, western blot or immunofluorescence. In some instances, their affinity is not sufficient, such as for many diagnostic applications or some applications in immunotherapy. Affinity maturation might be bypassed by the construction of multivalent molecules, as reviewed in Refs 49,50; however, there will be occasions in which in vitro affinity maturation of the selected antibodies is required. With phage display, as in the in vivo immune system, antibodies can be affinity-matured in a stepwise fashion, by incorporating mutations and selecting variant cells with decreasing amounts of antigen (reviewed in Ref. 30; Fig. 2). Various procedures for introducing diversity in the antibody genes have been described, ranging from more-or-less random strategies (e.g. V-gene chain shuffling, error-prone PCR, mutator strains or DNA shuffling), to strategies targeting the CDR regions of the antibody for mutagenesis (e.g. oligonucleotide-directed mutagenesis, PCR). PCR might be guided by information from the natural somatic hypermutation mechanism,

REVIEW I M M U N O L O G Y TO D AY

Table 1. Major differences between the three types of binding-site repertoires Type of repertoire

Advantages

Disadvantages

Main applications

Immune

High frequency of binders In vivo affinity maturation

Library has to be rebuilt for each antigen Bias against self-antibodies Antigen has to be immunogenic

Study of humoral response in disease (viral infections, cancer and autoimmunity) High-affinity diagnostic antibodies

Naïve

Can be used against any antigen Has to be very large to obtain good affinities Scaffold and somatic mutations Bias towards certain sequences and selected by evolution against self

Synthetic

Expand the structural scope Has to be very large to obtain good More suitable for use in automated of binding-site repertoires affinities selection and in microarray field Special ‘built-in’ design possible A high percentage of ‘non-viable’ sequences (genomics or proteomics) for specific applications Immunogenicity?

for example, by targeting natural hot spot (positions that are frequently mutated in vivo, and are more likely than others to generate improved affinity), or by mimicking loop deletions and insertions51. Affinity variants are selected under conditions that favour clones with a higher affinity, or an improved on- or off-rate. Can this process work more effectively in vitro than in vivo? During in vivo B-cell selection in the germinal centre, there is an affinity ceiling estimated to be around 1010 M21, beyond which selective enrichment is not likely to occur52. By using in vitro selection procedures, phage antibodies with approximately tenfold higher affinities (1011 M21)53,54 have been made. Perhaps a combination of rational and empirical approaches, in which the antibody is diversified in a stepwise fashion and selected for higher affinity, could, in the future, bring us closer to the ultra-high affinities described for some natural protein–ligand pairs, for example, avidin–biotin (1015 M21). One possible disadvantage of this in vitro process is that the affinity improvement can be accompanied by the appearance of a modified fine-specificity or unwanted crossreactivity55. The natural immune system might act quickly to remove such potentially deleterious variants; extensive screening of the in vitro affinity-matured antibodies for changes in specificity is thus required.

Exploring the designer binding site Binding-site repertoires with naturally rearranged V-genes are potentially heavily biased towards certain sequences and clones, owing to evolutionary pressures and the immune history of the host. This will result in biases and redundancy in natural binding-site libraries (Table 1). The use of laboratory-assembled variable-region genes would provide complete control over the genetic makeup, allow engineering of the individual building blocks, and a choice of where and how the diversity could be introduced in these repertoires. It was thought that with these possibilities, antibody libraries could be built with a high level of relevant diversity that might outperform the ‘naïve’ binding-site melting pot.

Selection of high-affinity human antibody for imaging or therapy

Methods for generating designer binding sites Many different binding-site designs have been proposed. In the first, synthetic antibody library56,57, a large collection of human germline gene segments was provided using PCR with a complete random diversity in the CDR3 regions of the heavy-chain segment, and later47 with limited diversity in the CDR3 for the light-chain segments. This design reflects the limited natural germline diversity in the other CDR regions. From this47 and related libraries, antibodies to many different targets and up to nanomolar affinity have been isolated. Several other synthetic repertoire designs have been described, varying from using one rearranged VH and VL segment and introducing two randomized CDR3 regions23, to using segments with completely randomized CDRs (Refs 58,59). The design of the binding site might combine elements of, or extend diversity beyond what is generated by the in vivo immune system. Perhaps a larger fraction of functionally relevant antibodies might be obtained by targeting the natural sites of both primary and secondary diversity, for example, the CDR3 of the heavy chain as primary ‘signature’ for the antibody, and the natural hot spots for somatic hypermutation in the other CDRs (Ref. 60). Also, libraries with non-natural CDR regions (e.g. mixed between species or artificially combined61, or with longer or shorter CDR1/2 loops) could well expand the structural scope of the binding-site repertoire, or might be used to build up the antigen combining site around the antigen62. The first binding-site libraries with a propensity for binding to a chosen antigen or antigenic epitope have been constructed63. This is a large step towards modelling-based and experimentally assisted de novo binding-site design64.

Improving designer binding-site libraries Despite these engineering opportunities, to date none of the designer libraries has yielded higher affinity antibodies than those that can be isolated from well-sampled large ‘naïve’ antibody libraries. This might be due to design limitations and also to suboptimal

A U G U S T Vo l . 2 1

No.8

375

REVIEW I M M U N O L O G Y T O D AY

experimental library construction and selection conditions. The procedures to create a randomized region in the V-regions, however, all suffer from introducing nonsense mutations, frameshifts and nonviable sequences, which will lead to a decrease of correctly folded and displayed antibody variants (Table 1). Methods to increase functional diversity (the number of clones in a library that display a functional binding site on their surface), include the use of oligonucleotides that provide more representative distributions of amino acid at each randomized codon in the library, and the use of well-expressed V-genes in the library. Germline segments might indeed be engineered for high-level expression in E. coli (Ref. 65); alternatively, individual heavy- and light-chain repertoires are displayed on a phage and enriched for expression via a generic binding ligand (protein A/L binding; Tomlinson and Winter, unpublished). In both cases, the selected antibodies are relatively well expressed in E. coli, which will have its advantages for some applications, but this has not been translated into a better average affinity for the selected antibodies. It therefore still remains to be seen what the best design or experimental strategies will be for constructing designer binding-site libraries. The natural binding-site library, with its structural scope yet overall stability, its economic yet highly effective usage of germline diversity and its process to affinity-mature and select higher affinity variants, will inspire our designer binding-site libraries in the years to come. Such new, super antibody libraries, with many applicationdriven features, will open up many new fields of (nontherapeutic) applications for antibodies. The use of the various types of bindingsite libraries is summarized in Table 1.

Phage display-based human therapeutic antibodies Where antibodies have, however, a very strong competitive advantage, is in their use as reliable, relatively easy to produce, welltolerated and well-behaved immunotherapeutics73. In this regard, immunogenicity has proved to be a major drawback in clinical applications of antibodies, thus for most therapeutic antibodies one would prefer a fully human antibody sequence. Less than ten years after the description of the concepts, the first phage display technology-derived antibodies, made by either using guided selection or direct selection from ‘naïve’ libraries, and directed to tumour necrosis factor a (TNFa) and transforming growth factor b2 (TGFb2), respectively, are now in Phase II/III clinical trials. In guided selection, a rodent antibody was rebuilt by shuffling the antibody variable domains and using phage selection to generate a human version of the antibody with similar binding characteristics74. Indeed, using phages, a series of very high-affinity murine and human antibodies have been developed for medical applications, for example, in the field of tumour targeting75–79. Alternatively, human antibodies can be made by applying hybridoma technology to mice with human antibody genes, or can be derived from rodent antibodies by humanization (which might be faster using phages80). The elements that govern a possible anti-globulin response are similar for human or humanized antibodies, and relate to primary amino acid sequence, the presence of T-cell epitopes, and the nature of the target cell and of the target. This makes the immunogenicity of these antibodies equally difficult to predict. Data from clinical trials with phage-derived human antibodies are awaited, to compare their immunogenicity and specificity with those of the many humanized antibodies in the clinic.

Smaller Ig-fold-based binding sites Nature has evolved to use two protein domains with an immunoglobulin fold in nearly all of its natural antigen combining sites. Nevertheless, engineering of binding sites based on a single variable domain with immunoglobulin fold8 might yield ligands with specific applications that antibodies with two variable domains cannot offer. Several single-domain antibody formats have been explored with phage display, including the naturally occurring heavy-chain only antibodies from camels66, as well as engineered, murine or human, heavy- or light-chain only domains67,68 (H.R.H. Hufton et al., unpublished). In addition, other nonantibody proteins with a single immunoglobulin fold have been equipped with novel binding specificities, including tendamistat69, domains of fibronectin70 and the extracellular domain of CTLA-4 (Ref. 71). Potential advantages of such single-domain binding molecules might be envisaged in terms of their ease and yield of production, targeting of certain antigen types (e.g. enzyme active sites and ligand-binding pockets of receptors), and in their fast engineering into multimeric or multivalent reagents. Furthermore, an increasing list of other protein folds are being developed as scaffolds for creating novel binding sites (reviewed in Ref. 72). Such singleIg folds and other artificial binding sites might eventually become major competitors for antibodies in many of the present applications.

A U G U S T 376

Vo l . 2 1

No.8

Concluding remarks Our efforts to make antibodies without using animals and outside the immune system have now provided us with very powerful tools for in vitro selection and evolution of binding-site libraries. Significant advances in the past decade with phage and other display methodologies, binding-site library design, refined selection procedures and instrumentation for automation, are making display technologies increasingly popular for creating binding sites for use in all areas of research, and in medical and industrial applications. Binding sites generated in vitro are therefore set to replace hybridomas and animals eventually as the major source of binding sites. The greatest challenge for the future will be to translate our ability to create binding sites with tailored size, affinity, valency and sequence, into antibody molecules with improved clinical benefit. Studies to understand antibody efficacy and to assess the role of antibody affinity, effector function and pharmacokinetics in clinical efficacy, should drive novel developments in therapeutic binding and effector-site engineering.

We gratefully acknowledge S. Hufton and P. Henderikx for critically reading the manuscript.

REVIEW I M M U N O L O G Y TO D AY

Hennie Hoogenboom ([email protected]) is at Target Quest BV, a subsidiary of Dyax Corp, PO Box 5800, 6202 AZ Maastricht, The Netherlands; both Patrick Chames and Hennie Hoogenboom are at the Dept of Pathology, Maastricht University and University Hospital Maastricht, 6202 AZ Maastricht, The Netherlands. References 1 Winter, G. and Milstein, C. (1991) Man-made antibodies. Nature 349, 293–299 2 Morrison, S.L. et al. (1984) Chimeric human antibody molecules: mouse antigen-binding domains with human constant region domains. Proc. Natl. Acad. Sci. U. S. A. 81, 6851–6855 3 Jones, P.T. et al. (1986) Replacing the complementarity-determining regions in a human antibody with those from a mouse. Nature 321, 522–525 4 Hudson, P.J. (1998) Recombinant antibody fragments. Curr. Opin. Biotechnol. 9, 395–402 5 Smith, G.P. (1985) Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface. Science 228, 1315–1317 6 Orlandi, R. et al. (1989) Cloning immunoglobulin variable domains for expression by the polymerase chain reaction. Proc. Natl. Acad. Sci. U. S. A. 86, 3833–3837 7 Chiang, Y.L. et al. (1989) Direct cDNA cloning of the rearranged immunoglobulin variable region. Biotechniques 7, 360–366 8 Ward, E.S. et al. (1989) Binding activities of a repertoire of single immunoglobulin variable domains secreted from Escherichia coli. Nature 341, 544–546 9 Huse, W.D. et al. (1989) Generation of a large combinatorial library of the immunoglobulin repertoire in phage lambda. Science 246, 1275–1281 10 Better, M. et al. (1988) Escherichia coli secretion of an active chimeric antibody fragment. Science 240, 1041–1043 11 Skerra, A. and Pluckthun, A. (1988) Assembly of a functional immunoglobulin Fv fragment in Escherichia coli. Science 240, 1038–1041 12 McCafferty, J. et al. (1990) Phage antibodies: filamentous phage displaying antibody variable domains. Nature 348, 552–554 13 Hoogenboom, H.R. et al. (1991) Multi-subunit proteins on the surface of filamentous phage: methodologies for displaying antibody (Fab) heavy and light chains. Nucleic Acids Res. 19, 4133–4137 14 Garrard, L.J. et al. (1991) Fab assembly and enrichment in a monovalent phage display system. Biotechnology 9, 1373–1377 15 Chang, C.N. et al. (1991) Expression of antibody Fab domains on bacteriophage surfaces. J. Immunology 147, 3610–3614 16 McGuinness, B.T. et al. (1996) Phage diabody repertoires for selection of large numbers of bispecific antibody fragments. Nat. Biotechnol 14, 1149–1154 17 Clackson, T. et al. (1991) Making antibody fragments using phage display libraries. Nature 352, 624–628 18 Persson, M.A. et al. (1991) Generation of diverse high-affinity human monoclonal antibodies by repertoire cloning. Proc. Natl. Acad. Sci. U. S. A. 88, 2432–2436 19 Burton, D.R. et al. (1991) A large array of human monoclonal antibodies to type 1 human immunodeficiency virus from combinatorial libraries of asymptomatic individuals. Proc. Natl. Acad. Sci. U. S.A. 88, 10134–10137 20 Graus, Y.F. et al. (1997) Human anti-nicotinic acetylcholine receptor recombinant Fab fragments isolated from thymus-derived phage display libraries from myasthenia gravis patients reflect predominant specificities in serum and block the action of pathogenic serum antibodies. J. Immunol. 158, 1919–1929 21 Cai, X. and Garen, A. (1995) Anti-melanoma antibodies from melanoma patients immunised with genetically modified autologous

22 23

24 25 26 27

28

29 30

31

32

33

34 35

36 37 38

39 40 41

42 43 44

45

tumor cells: selection of specific antibodies from single-chain Fv fusion phage libraries. Proc. Natl. Acad. Sci. U. S. A. 92, 6537–6541 Marks, J.D. et al. (1991) By-passing immunization: human antibodies from V-gene libraries displayed on phage. J. Mol. Biol. 221, 581–597 Barbas, C.F. et al. (1992) Semisynthetic combinatorial libraries: a chemical solution to the diversity problem. Proc. Natl. Acad. Sci. U. S. A. 89, 4457–4461 Winter, G. et al. (1994) Making antibody by phage display technology. Annu. Rev. Immunol. 12, 433–455 Hoogenboom, H.R. et al. (1998) Antibody phage display technology and its applications. Immunotechnology 4, 1–20 Dall’Acqua, W. and Carter, P. (1998) Antibody engineering. Curr. Opin. Struct. Biol. 8, 443–450 Hanes, J. and Pluckthun, A. (1997) In vitro selection and evolution of functional proteins by using ribosome display. Proc. Natl. Acad. Sci. U. S. A. 94, 4937–4942 Georgiou, G. et al. (1997) Display of heterologous proteins on the surface of microorganisms: from the screening of combinatorial libraries to live recombinant vaccines. Nat. Biotechnol. 15, 29–34 Boder, E.T. and Wittrup, K.D. (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nat. Biotechnol. 15, 553–557 Hoogenboom, H.R. (1997) Designing and optimizing library selection strategies for generating high-affinity antibodies. Trends Biotechnol. 15, 62–70 de Haard, H.J. et al. (1999) A large non-immunized human Fab fragment phage library that permits rapid isolation and kinetic analysis of high affinity antibodies. J. Biol. Chem. 274, 18218–18230 Horn, I.R. et al. (1999) Selection of phage-displayed Fab antibodies on the active conformation of Ras yields a high affinity conformationspecific antibody preventing the binding of c-Raf kinase to Ras. FEBS Lett. 463, 115–120 Chames, P. et al. Direct selection of a human antibody fragment binding to the tumor T cell epitope HLA-A1/MAGE-A1 from a non immunized phage–Fab library. Proc. Natl. Acad. Sci. U. S. A. (in press) Osbourn, J.K. et al. (1998) Pathfinder selection: in situ isolation of novel antibodies. Immunotechnology 3, 293–302 Hoogenboom, H.R. et al. (1999) Selection-dominant and nonaccessible epitopes on cell-surface receptors revealed by cell-panning with a large phage antibody library. Eur. J. Biochem. 260, 774–784 Proba, K. et al. (1998) Antibody scFv fragments without disulfide bonds made by molecular evolution. J. Mol. Biol. 275, 245–253 Kristensen, P. and Winter, G. (1998) Proteolytic selection for protein folding using filamentous bacteriophages. Fold. Des. 3, 321–328 Gargano, N. and Cattaneo, A. (1997) Rescue of a neutralizing anti-viral antibody fragment from an intracellular polyclonal repertoire expressed in mammalian cells. FEBS Lett. 414, 537–540 Jung, S. et al. (1999) Selectively infective phage (SIP) technology: scope and limitations. J. Immunol. Methods 231, 93–104 Borrebaeck, C.A. (1998) Tapping the potential of molecular libraries in functional genomics. Immunol. Today 19, 524–527 Ridgway, J.B. et al. (1999) Identification of a human anti-CD55 singlechain Fv by subtractive panning of a phage library using tumor and nontumor cell lines. Cancer Res. 59, 2718–2723 Lueking, A. et al. (1999) Protein microarrays for gene expression and antibody screening. Anal. Biochem. 270, 103–111 Abbott, A. (1999) A post-genomics challenge: learning to read patterns of protein expression. Nature 402, 715–720 Hawkins, R.E. and Winter, G. (1992) Cell selection strategies for making antibodies from variable gene libraries: trapping the memory pool. Eur. J. Immunol. 22, 867–870 Sheets, M.D. et al. (1998) Efficient construction of a large nonimmune phage antibody library: the production of high-affinity human singlechain antibodies to protein antigens. Proc. Natl. Acad. Sci. U. S. A. 95, 6157–6162

A U G U S T Vo l . 2 1

No.8

377

REVIEW I M M U N O L O G Y T O D AY

46 Vaughan, T.P. et al. (1996) Human antibody with sub-nanomolar affinities isolated from a large non-immunized phage display library. Nat. Biotechnol. 14, 309–314 47 Griffiths, A.D. et al. (1994) Isolation of high affinity human antibodies directly from large synthetic repertoires. EMBO J. 13, 3245–3260 48 Sblattero, D. and Bradbury, A. (2000) Exploiting recombination in single bacteria to make large phage antibody libraries. Nat. Biotechnol. 18, 75–80 49 Holliger, P. and Winter, G. (1997) Diabodies: small bispecific antibody fragments. Cancer Immunol. Immunother. 45, 128–130 50 Hudson, P.J. and Kortt, A.A. (1999) High avidity scFv multimers; diabodies and triabodies. J. Immunol. Methods 231, 177–189 51 de Wildt, R.M. et al. (1999) Somatic insertions and deletions shape the human antibody repertoire. J. Mol. Biol. 294, 701–710 52 Foote, J. and Eisen, H.N. (1995) Kinetic and affinity limits on antibodies produced during immune responses. Proc. Natl. Acad. Sci. U. S. A. 92, 1254–1256 53 Schier, R. et al. (1996) Isolation of picomolar affinity anti-c-erbB-2 single-chain Fv by molecular evolution of the complementarity determining regions in the center of the antibody binding site. J. Mol. Biol. 263, 551–67 54 Yang, W.P. et al. (1995) CDR walking mutagenesis for the affinity maturation of a potent human anti-HIV-1 antibody into the picomolar range. J. Mol. Biol. 254, 392–403 55 Ohlin, M. et al. (1996) Light chain shuffling of a high affinity antibody results in a drift in epitope recognition. Mol. Immunol. 33, 47–56 56 Hoogenboom, H.R. and Winter, G. (1992) By-passing immunisation. Human antibodies from synthetic repertoires of germline VH gene segments rearranged in vitro. J. Mol. Biol. 227, 381–388 57 Nissim, A. et al. (1994) Antibody fragments from a ‘single pot’ phage display library as immunochemical reagents. EMBO J. 13, 692–698 58 Garrard, L.J. and Henner, D.J. (1993) Selection of an anti-IGF-1 Fab from a Fab phage library created by mutagenesis of multiple CDR loops. Gene 128, 103–109 59 Soderlind, E. et al. (1995) Domain libraries: synthetic diversity for de novo design of antibody V regions. Gene 160, 269–272 60 Pini, A. et al. (1998) Design and use of a phage display library. Human antibodies with subnanomolar affinity against a marker of angiogenesis eluted from a two-dimensional gel. J. Biol. Chem. 273, 21769–21776 61 Jirholt, P. et al. (1998) Exploiting sequence space: shuffling in vivo formed complementarity determining regions into a master framework. Gene 215, 471–476 62 Lamminmaki, U. et al. (1999) Expanding the conformational diversity by random insertions to CDRH2 results in improved anti-estradiol antibodies. J. Mol. Biol. 291, 589–602 63 Kirkham, P.M. et al. (1999) Towards the design of an antibody that recognises a given protein epitope. J. Mol. Biol. 285, 909–915 64 Winter, G. (1990) Gene technologies for antibody engineering. Behring. Inst. Mitt. 87, 10–20 65 Knappik, A. et al. (2000) Fully synthetic human combinatorial antibody libraries (HuCAL) based on modular consensus frameworks and CDRs randomized with trinucleotides. J. Mol. Biol. 296, 57–86 66 Arbabi Ghahroudi, M. et al. (1997) Selection and identification of single domain antibody fragments from camel heavy-chain antibodies. FEBS Lett. 414, 521–526 67 Davies, J. and Riechmann, L. (1996) Single antibody domains as small recognition units: design and in vitro antigen selection of camelized, human VH domains with improved protein stability. Protein Eng. 9, 531–537 68 Reiter, Y. et al. (1999) An antibody single-domain phage display library of a native heavy chain variable region: isolation of functional singledomain VH molecules with a unique interface. J. Mol. Biol. 290, 685–698

A U G U S T 378

Vo l . 2 1

No.8

69 McConnell, S.J. and Hoess, R.H. (1995) Tendamistat as a scaffold for conformationally constrained phage peptide libraries. J. Mol. Biol. 250, 460–470 70 Koide, A. et al. (1998) The fibronectin type III domain as a scaffold for novel binding proteins. J. Mol. Biol. 284, 1141–1151 71 Hufton, S.E. et al. (2000) Development and application of cytotoxic T lymphocyte–associated antigen 4 as a protein scaffold for the generation of novel binding ligands. FEBS Lett. 475, 225–231 72 Nygren, P.A. and Uhlen, M. (1997) Scaffolds for engineering novel binding sites in proteins. Curr. Opin. Struct. Biol. 7, 463–469 73 Breedveld, F.C. (2000) Therapeutic monoclonal antibodies. Lancet 355, 735–740 74 Jespers, L.S. et al. (1994) Guiding the selection of human antibodies from phage display repertoires to a single epitope of an antigen. Biotechnology 12, 899–903 75 Begent, R.H. et al. (1996) Clinical evidence of efficient tumor targeting based on single-chain Fv antibody selected from a combinatorial library. Nat. Med. 2, 979–984 76 Schier, R. et al. (1996) Isolation of picomolar affinity anti-c-erbB-2 single-chain Fv by molecular evolution of the complementarity determining regions in the center of the antibody binding site. J. Mol. Biol. 263, 551–567 77 Henderikx, P. et al. (1998) Human single-chain Fv antibodies to MUC1 core peptide selected from phage display libraries recognize unique epitopes and predominantly bind adenocarcinoma. Cancer Res. 58, 4324–4332 78 Birchler, M. et al. (1999) Selective targeting and photocoagulation of ocular angiogenesis mediated by a phage-derived human antibody fragment. Nat. Biotechnol. 17, 984–988 79 Beiboer, S.H. et al. (2000) Guided selection of a pan carcinoma specific antibody reveals similar binding characteristics yet structural divergence between the original murine antibody and its human equivalent. J. Mol. Biol. 296, 833–849 80 Baca, M. et al. (1997) Antibody humanization using monovalent phage display. J. Biol. Chem. 272, 10678–10684

Keeping up with the Trends… Immunology in other Trends journals The role of nonhuman primate models in AIDS vaccine development M.I. Johnston: Molecular Medicine Today 6, 255–293 Potential role of TNF-α in the pathogenesis of insulin resistance and type 2 diabetes D.E. Moller: Trends in Endocrinology and Metabolism 11, 212–217 Driving cells into atherosclerotic lesions – a deleterious role for viral chemokine receptors? H. Hengel and C. Weber: Trends in Microbiology 8, 294–296