Copyright c(cid:2) 2000 by Annual Reviews. All rights reserved
`
`ANTIBODY ENGINEERING
`
`?Annu. Rev. Biomed. Eng. 2000. 02:339–76
`
`CONTENTS
`INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
`ANTIBODY STRUCTURE AND THE RECOGNITION OF ANTIGENS . . . . . . . . 340
`ANTIBODY ENGINEERING TECHNOLOGIES . . . . . . . . . . . . . . . . . . . . . . . . . 341
`Antibody Cloning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
`Measuring Antibody Affinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
`Antibody Library Screening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
`Bypassing Immunization: Isolation of Antibodies from Large Libraries . . . . . . . . 349
`Affinity Maturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
`Humanization of Murine Antibodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
`Human Antibodies from Transgenic Mice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
`Engineering Antibody Avidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
`PRODUCTION OF RECOMBINANT ANTIBODIES . . . . . . . . . . . . . . . . . . . . . . 357
`ANTIBODY APPLICATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
`Neutralizing Antibodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
`Intracellular Antibodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
`Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
`Cancer Therapies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
`CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
`
`Jennifer Maynard1 and George Georgiou1,2
`1Department of Chemical Engineering and 2Institute for Cellular and Molecular Biology,
`University of Texas at Austin, Austin, Texas 78712; e-mail: ggmjd@mail.utexas.edu
`
`Key Words phage display, library screening, scFv, antigen, combinatorial libraries
`■ Abstract Antibodies are unique in their high affinity and specificity for a binding
`partner, a quality that has made them one of the most useful molecules for biotech-
`nology and biomedical applications. The field of antibody engineering has changed
`rapidly in the past 10 years, fueled by novel technologies for the in vitro isolation
`of antibodies from combinatorial libraries and their functional expression in bacteria.
`This review presents an overview of the methods available for the de novo generation
`of human antibodies, for engineering antibodies with increased antigen affinity, and for
`the production of antibody fragments. Select applications of recombinant antibodies
`are also presented.
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`1523-9829/00/0825-0339$14.00
`
`339
`
`Lassen - Exhibit 1058, p. 1
`
`

`

`MAYNARD (cid:2) GEORGIOU
`
`INTRODUCTION
`
`The significance of antibodies as diagnostic and analytical reagents has been known
`and exploited for almost a century. In recent years, antibodies have become in-
`creasingly accepted as therapeutic reagents, particularly for cancer but also for
`numerous other disorders. An indication of the emerging significance of antibody-
`based therapeutics is that over a third of the proteins currently undergoing clin-
`ical testing in the United States are antibodies. Until the late 1980s, antibody
`technology relied primarily on animal immunization or the expression of engi-
`neered antibodies in a eukaryotic host. However, the development of methods for
`the expression of antibody fragments in bacteria, together with the emergence
`of powerful techniques for screening combinatorial libraries and an expanding
`structure-function data base has opened unlimited opportunities for the engineer-
`ing of antibodies with tailor-made properties for specific applications. Antibodies
`of low immunogenicity, suitable for human therapeutic or diagnostic purposes,
`can now be engineered with relative ease. Such reagents can greatly enhance and
`complement other biomedical engineering technologies. This chapter presents an
`overview of the current methodologies for antibody isolation and functional opti-
`mization. Select applications of possible relevance to biomedical engineering are
`also discussed. However, we apologize in advance to the reader because, due to
`space limitations, it was not possible to cover numerous current or emerging areas
`of antibody technology.
`
`?340
`
`ANTIBODY STRUCTURE AND THE RECOGNITION
`OF ANTIGENS
`
`There are five classes of immunoglobulins: IgM, IgG, IgA, IgD, and IgE (3). From
`a biotechnology perspective, by far the most important class of antibodies is IgG
`and to a lesser extent IgM and IgA. IgMs are pentamers, and their large size results
`in rather poor pharmacokinetic properties, whereas their low specificity renders
`them less desirable than IgG antibodies for diagnostic applications. Secretory IgAs
`can potentially be very important as a means of passive immunization against
`genital, gastrointestinal, and oral pathogens. However, until recently production
`of useful amounts of monoclonal secretory IgAs has been problematic (4, 5).
`Antibodies belonging to the IgG class are homodimers of two identical polypep-
`tide chains of 450 amino acids (heavy chains) and two identical chains of 250 amino
`acids (light chains). The structure of each of the four heavy and two light chain
`domains has the characteristic immunoglobulin fold consisting of two antiparal-
`lel β-sheets with an intramolecular disulfide bond. The N-terminal domains of
`each chain are unique in that the three loops connecting the β-sheets are highly
`variable in length and sequence. The six hypervariable loops, or complementarity-
`determining regions (CDRs) form a unique surface that specifically recognizes
`and binds an antigen (see Figure 1).
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 2
`
`

`

`?ANTIBODY ENGINEERING
`
`The recognition of antigens by high affinity antibodies is generally described
`as a “lock and key” fit, in that the conformation of the antibody is generally not
`greatly perturbed upon antigen binding. However, it should be kept in mind that
`not all antibody:antigen complexes fit this description, and large conformational
`changes in antibody binding pockets are not uncommon. MacCallum et al (6) have
`classified antibody binding sites into four classes: concave and moderately con-
`cave (mostly small molecule binders), ridged (mostly peptide binders), and planar
`(mostly protein binders). Because antigen binding is thermodynamically favor-
`able, it is accompanied by a decrease in free energy, and numerous biochemical
`and crystallographic studies have explored the energetics of protein-ligand binding
`(7). In general, antigen recognition is enthalpically driven by van der Waals inter-
`actions, salt bridges, and hydrogen bonds, whereas the entropic cost of forming an
`antibody:antigen complex is partially compensated by desolvation effects. Water
`molecules play a critical role in stabilizing the interaction, as they fill cavities
`where the geometric complementarity is imperfect and ensure that all hydrogen
`bond donors and acceptors are compensated (8). Within the binding pocket, “hot
`spots,” a small number of residues within the binding site that account for a large
`fraction of the interaction energy, can often be found (9) and are a common theme
`in biological recognition (10).
`
`ANTIBODY ENGINEERING TECHNOLOGIES
`
`Antibody Cloning
`Antibody engineering became possible with the development of hybridoma tech-
`nology in 1975, relying on the fusion of a myeloma cell line with B-cells from
`an immunized animal (11). Monoclonal antibody technology remains one of the
`core technologies of biotechnology, and thousands of medically and diagnostically
`relevant hybridomas have been developed.
`
`341
`
`Figure 1 Ribbon diagram
`from the crystal structure of
`the antidigoxin Fab, shown in
`complex with the small mole-
`cule steroid, digoxin. Frame-
`work variable
`regions
`are
`shown in white, CDR regions
`in light gray, digoxin in white.
`Note that only the CDR re-
`gions mediate antigen contact.
`[From PDB file 1IGJ (1).]
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 3
`
`

`

`MAYNARD (cid:2) GEORGIOU
`
`?342
`
`A revolution in antibody technology began in the late 1980s when efficient sys-
`tems for the cloning and expression of antibody genes in bacteria were developed
`(12). Cloning the genes of monoclonal antibodies from murine hybridomas was
`greatly simplified by the discovery of the polymerase chain reaction (PCR), and
`bacterial expression systems allowed the rapid and facile production of functional
`recombinant antibody fragments for analysis (13).
`To clone antibodies, mRNA is isolated from hybridoma, spleen, or lymph
`cells, reverse transcribed into DNA, and antibody genes are amplified by PCR.
`This strategy requires oligonucleotide primers that can recognize any antibody
`gene. Numerous primer sets have been published, with 5(cid:3) primers based on the
`N-terminal sequence of purified antibodies (14), rapid amplification of cDNA ends
`(15), antibody leader sequences (16), and most popularly, primers based on known
`variable region framework amino acid sequences from the Kabat (17) and V-base
`databases (18–20).
`An artifact of the process used to generate hybridomas is that many myeloma cell
`lines express irrelevant heavy or light chains, in addition to the desired monoclonal
`antibodies (21, 22). The general cloning strategy outlined above will amplify all
`antibody genes present at the mRNA level, and multiple heavy and light chain genes
`in addition to the desired ones may thus be cloned (20). Although techniques have
`been developed to selectively remove aberrant mRNA (23), it is critical that cloned
`antibody genes are expressed in functional form to confirm their ability to bind
`antigen (20). In addition, mutations are often introduced by the degenerate primers
`themselves, even when the correct chain is cloned (24), further underscoring the
`need to screen cloned genes for function.
`Once the correct genes encoding VH and VL domains have been identified,
`they can be assembled in a number of forms suitable for expression or further
`manipulation. The smallest antibody-derived polypeptide that can bind antigen
`with reasonable affinity is a single VH chain. VH chains are very unstable and
`prone to aggregation in vivo because the largely hydrophobic area that normally
`forms the interface with the VL domain is exposed to the solvent. However, one
`class of VH chains derived from camelids, such as camels or llamas, are naturally
`found without a light chain, yet exhibit high antigen affinities (25). Sequence and
`structural analysis of the camelid VH chains (26, 27) has been used to guide the
`rational design of mutations in a human VH chain that rendered it stable in the
`absence of a VL chain (28–30). Alternatively, a unique murine VH chain found to
`be naturally stable without a VL chain has been engineered to recognize different
`antigens (31).
`In general, the presence of both the VH and VL chains is needed for high stability
`and antigen affinity. VH and VL chains can be expressed as separate polypeptides
`in bacteria where they assemble into Fv fragments (see Figure 2). However, in
`these dimeric proteins the two polypeptide chains are held together by noncovalent
`interactions and are therefore prone to dissociation and aggregation. The two chains
`can be covalently assembled by engineering an interchain disulfide bond to give
`a dsFv antibody. This design is more stable than an Fv fragment, but is difficult
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 4
`
`

`

`343
`
`Figure 2 Schematic showing various antibody fragments of biotechnological and clinical
`interest. Each block represents one antibody domain with a characteristic immunoglobulin
`fold. Black bars, inter-chain di-sulfide bonds (horizontal) or intra-domain linkages (verti-
`cal); longer curved lines, genetically engineered polypeptide linkers. [Adapted from (2).]
`
`?ANTIBODY ENGINEERING
`
`to produce by fermentation, and the disulfide bond can be reduced under mild
`conditions. Recombinant DNA techniques can introduce a short polypeptide linker
`to fuse the VH and VL chains together into a scFv antibody fragment (see Figure 2).
`scFvs are relatively small (26–27 kDa), generally quite stable, and are encoded by
`a single gene, which simplifies genetic manipulations. The most common linker
`is a flexible (Gly4Ser)3 decapentapeptide (32). The two variable domains can be
`connected either as VH-linker-VL or VL-linker-VH, with the former being more
`common. The order of the two domains can affect expression efficiency, stability,
`and the tendency to form dimers in solution (33, 34). If a scFv is found to have
`poor stability or low affinity compared with the parental antibody, engineering the
`linker sequence may improve function. A variety of linkers have been designed
`based on structural considerations, screening of combinatorial linker libraries, or
`natural linker sequences occurring in multi-domain polypeptides (35–40).
`In addition to scFvs, the other commonly used recombinant antibody fragments
`are Fabs (see Figure 2). Fabs consist of two polypeptide chains, one containing the
`light chain variable and constant domains, VL-Cκ or λ, the other a truncated heavy
`chain containing the variable domain and one constant domain, VH-CH1. Just as
`in intact IgG immunoglobulins, the two chains are linked together by a disulfide
`bond. The more extensive interface between the two chains and the presence of the
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 5
`
`

`

`Ab + Ag
`
`Ab − Ag
`
`kon−→←−
`koff
`
`MAYNARD (cid:2) GEORGIOU
`
`disulfide bond confer increased stability to denaturation. Although the expression
`of Fab requires the association of two chains, it often occurs quite efficiently in
`bacteria (41, 42). On the other hand, the presence of two chains somewhat compli-
`cates genetic manipulations, and the larger size of these proteins may limit their
`bioavailability for certain therapeutic applications.
`
`Measuring Antibody Affinity
`The most important performance criterion for immunoglobulins or recombinant
`antibody fragments is the equilibrium binding constant for the formation of an
`antibody:antigen complex:
`
`?344
`
`where, KD = koff/kon.
`Accurate determination of the equilibrium constant, KD, is not trivial, as it re-
`quires the quantitation of the concentration of both the complex and either free
`antibody or antigen in solution under conditions that do not perturb the equilib-
`rium (43, 44). One of the most reliable and relatively easy equilibrium techniques
`is competition ELISA (45). Techniques requiring greater sophistication and spe-
`cialized equipment include (a) fluorescence quenching, whereby the binding of
`antibodies to fluorescent antigens or to fluorescent conjugates quenches the flu-
`orescence, providing a convenient way to determine equilibrium concentrations;
`(b) isothermal titration microcalorimetry used to determine the enthalpy and equi-
`librium constant from which the entropy of binding can calculated (46); (c) flow
`cytometry, a convenient way to measure the dissociation rate constant koff for re-
`combinant antibodies displayed on the surface of microbial cells (47–49); and (d)
`surface plasmon resonance (SPR).
`In recent years, SPR has emerged as the method of choice for measuring the
`kinetics of protein-ligand interactions (50, 51) and can be used to analyze antibod-
`ies having KD values from 10−5 to 10−11 M (52). Although not without pitfalls,
`SPR is a relatively simple and fast technique. Briefly, the antigen is immobilized
`on a derivatized dextran matrix on a sensor chip, and the amount of antibody that
`binds under laminar flow conditions is detected as a change in resonance angle of
`incident light, which is reported as “response units” (see Figure 3). The kinetics
`of the bimolecular association reaction are calculated from experiments with dif-
`ferent antibody concentrations. The dissociation rate is measured as a decrease in
`the amount of bound protein when the solution flowing over the chip surface con-
`tains free antigen (as a competitor) and no antibody. Usually, the key parameter
`that determines the KD is the rate of dissociation of the antibody:antigen com-
`plex, which can be determined readily by SPR without needing a purified protein
`preparation. koff can be determined using antibodies displayed on the surface of
`bacteriophages, in crude bacterial lysates, and even with antibodies synthesized
`by in vitro transcription-translation (53, 54; G Georgiou, unpublished results).
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 6
`
`

`

`345
`
`Figure 3 Affinity measurement by surface plas-
`mon resonance. (A) Physical configuration: anti-
`gen is covalently bound to a sensor chip; the anti-
`body flows over the surface. Binding is detected by
`a decrease in light intensity for a given incident an-
`gle. (B) A typical sensorgram. During the associa-
`tion phase, antibody flows over the chip and binds
`immobilized antigen; response units (RUs) in-
`crease. During dissociation, buffer with free anti-
`gen flows over the chip, and antibody dissociates
`from antigen, resulting in exponential decay of the
`RU signal with a characteristic rate constant koff.
`
`?ANTIBODY ENGINEERING
`
`Antibody Library Screening
`The in vitro isolation and engineering of antibodies hinges on the ability to screen
`large combinatorial libraries of antibody genes to isolate clones that exhibit spe-
`cific binding to a desired target molecule. Efficient high throughput screening
`technologies had to be developed because library sizes between 105–1011 clones
`have to be screened for most immunotechnology applications. During the last 10
`years, a variety of techniques of increasing sophistication and sensitivity have
`been developed to aid in the isolation of high affinity antibodies from such large
`libraries. These have been reviewed recently (56–58), and therefore the topic is
`covered here only briefly.
`The simplest library screening strategy is to clone an antibody gene repertoire
`into a λ bacteriophage expression vector (59). Upon infection of a lawn of bacterial
`cells growing on a plate, zones of clearance (plaques) containing antibodies re-
`leased from lysed cells, as well as bacteriophage progeny, are formed. The plate is
`overlayed with a membrane filter that has a high binding capacity for proteins. The
`filter is then reacted with a radiolabeled antigen, and spots corresponding to the
`position of plaques that produce specific antibodies are visualized by autoradiogra-
`phy. The bacteriophage encoding the specific antibodies are then isolated from the
`plaques and used for further analysis. This technique has been used successfully
`for the de novo isolation of antibodies and for affinity improvement applications
`(59, 60). However, it is more cumbersome than other library screening techniques,
`particularly when large libraries are being used.
`The screening of large libraries of recombinant proteins can be greatly simplified
`by establishing a direct physical link between a gene, the protein it encodes, and
`the molecule recognized by the protein. Such a link can be established using a
`variety of display technologies that are exploited for the screening of libraries in
`an efficient, high throughput manner (56, 58). Both in vivo and in vitro library
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 7
`
`

`

`MAYNARD (cid:2) GEORGIOU
`
`Figure 4 The three major antibody display sys-
`tems. (A) Antibody display on filamentous bacterio-
`phage as a gpIII fusion. Usually one fusion protein
`is displayed per phage. (B) Cell surface display. In
`this case, the gene is maintained on a plasmid. (C)
`Ribosome display. During an in vitro protein syn-
`thesis reaction, a complex is formed between the
`ribosomes, mRNA, and a properly folded antibody.
`
`?346
`
`display systems have been developed. For in vivo methods, the protein is encoded
`by a recombinant gene and displayed on the surface of a biological particle such as
`a virus or a whole cell (see Figure 4). Currently, the most widely used technique for
`library screening is based on the display of antibodies on the surface of filamentous
`bacteriophages (57). Antibody genes are fused in-frame to phage genes encoding
`surface-exposed proteins, most commonly pIII. The gene fusions are translated into
`chimeric proteins in which the two domains fold independently. Phage displaying
`an antibody specific for a ligand can be readily enriched by selective adsorption
`onto immobilized ligand, a process known as panning (see Figure 5). The bound
`phage is desorbed from the surface, usually by acid elution, and amplified through
`infection of Escherichia coli cells. Usually, 5–8 rounds of panning, desorption, and
`amplification are sufficient to select for phage displaying specific antibodies, even
`from very large libraries (up to 1011 clones). Once specific phage have been isolated,
`they can be used directly to obtain an estimate of antibody ligand affinity by ELISA
`or SPR. Also, phage infection of certain E. coli host strains can be employed to
`produce the antibody protein alone without being tethered to the pIII protein.
`A wide variety of techniques for addressing specific technical issues in the
`screening of libraries have been developed. These include (a) systems for gen-
`erating phage that predominantly display only one copy of antibody per phage
`particle, thus eliminating avidity effects during selection (57); (b) methods for
`guaranteeing the release of bound phage from the solid support by disrupting the
`chemical linkage that is used for ligand immobilization onto the matrix; (c) phage
`enrichment by binding to soluble rather than immobilized ligands, to allow better
`discrimination of high affinity clones (61); (d) selectively infective phage (SIP), a
`technique for obtaining very efficient enrichment of the desired clones by requir-
`ing that only phage bound to ligand are able to infect bacteria for amplification
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 8
`
`

`

`347
`
`Figure 5 Antibody affinity maturation and library screening cycle. This cycle is similar
`for all antibody display systems discussed in the text, but monovalent phage display is
`shown here as an example. The steps of the cycle are (a) a library is created and displayed
`as a phage fusion protein; (b) the phage are allowed to bind the immobilized antigen;
`(c) nonspecific and low affinity phage are removed by stringent washing; (d) bound phage
`are eluted and allowed to infect E. coli cells to amplify the gene by phage production;
`and (e) the panning cycle is repeated 5–8 times to identify a small number of high affinity
`antibody genes.
`
`?ANTIBODY ENGINEERING
`
`(62, 63); and (e) display on lytic phages that assemble in the bacterial cytoplasm, a
`strategy especially suited to the production of antibodies that can fold in the cytosol
`and are thus useful for gene therapy applications (64, 65). A number of excellent
`reviews describe these and other advances in the applications of the phage display
`technology for antibody library screening (56, 57, 63, 66).
`Examples of some of the most successful applications of the phage display
`technology in antibody engineering include (a) the de novo isolation of high affin-
`ity human antibodies from repertoire libraries; (b) isolation of genes encoding
`antigen-specific antibodies from immunized animals and from hybridomas; (c)
`affinity maturation: antibodies with extremely low KD binding constants, in the
`low picomolar range, have been generated; (d) isolation of antibodies with in-
`creased resistance to chemical or thermal denaturation; and (e) antibodies specific
`to cell surface markers or receptors that cannot be produced in purified form have
`been isolated by panning directly against whole cells and tissue samples.
`Antibodies can also be displayed on the surface of microbial cells such as E. coli
`and Saccharomyces cerevisiae (47, 49, 67–70). For screening purposes, a library of
`cells, each displaying multiple copies of a different antibody variant, is incubated
`with a fluorescently tagged ligand in buffer. Cells displaying antibodies that bind
`the ligand become fluorescently labeled and are isolated by fluorescence-activated
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 9
`
`

`

`MAYNARD (cid:2) GEORGIOU
`
`?348
`
`cell sorting (FACS). With flow cytometry, the binding of each clone in the library
`to a particular ligand is determined quantitatively. Parameters such as ligand con-
`centration or time for the dissociation of antibody:ligand complexes can be easily
`optimized to ensure the isolation of only the highest affinity antibodies. These
`features are particularly significant for antibody affinity maturation, and in fact,
`the limited data reported so far indicate that cell surface display may be a superior
`screening technology for that purpose (47, 48, 71). In addition, the screening of
`cell-displayed libraries always results in the isolation of antibodies that can be
`expressed well in the host cell. This is because, in order for a cell to be detected
`and sorted by the flow cytometer, it must have a minimum threshold fluorescence
`(roughly equivalent to the binding of at least 1000 ligand molecules conjugated to
`a small-molecule fluorescent dye). This means that the cell must be able to dis-
`play at least 1000 correctly folded antibody molecules. If an antibody is expressed
`poorly, it will not be represented in a sufficiently high number of copies to allow the
`isolation of the corresponding cell. In contrast, the screening of phage-displayed
`libraries by panning poses no requirement for good expression, as one copy is dis-
`played per particle. As a result, antibodies isolated from phage-displayed libraries
`often have very poor expression characteristics.
`At present, cell surface display and FACS cannot be used for the de novo
`isolation of antibodies from natural or synthetic repertoire libraries because the
`maximum library diversity that can be realistically screened by FACS is about
`5×108. This is about two orders of magnitude smaller than the library size needed
`for the isolation of high-affinity antibodies from repertoire libraries (66). However,
`repertoire libraries can be screened by 1–2 rounds of panning to reduce the library
`diversity followed by cell display technologies that guarantee the isolation of
`high-affinity, well expressed antibodies. The isolation of antihapten antibodies
`from repertoire libraries by such a tandem phage-cell surface display technology
`has been demonstrated (G Chen, A Hayhurst, J Thomas, B Iverson, Georgiou, in
`preparation).
`In vitro protein synthesis is an excellent means for the facile analysis of large
`numbers of site-specific protein mutants (73, 74; JA Maynard, B Iverson, G
`Georgiou, in preparation). In addition, several exciting in vitro library screening
`techniques have been developed that completely circumvent the need to use a host
`cell for protein synthesis. In vitro screening technologies employ RNA polymerase
`for mRNA synthesis from a DNA template. The mRNA is then translated into pro-
`tein using ribosomal extracts, amino acids, and various cofactors. The advantage
`of in vitro library screening methods is that the library size that can be screened is
`vast, potentially 1015 different sequences. In theory at least, screening such highly
`diverse libraries could result in the discovery of entirely new protein functions that
`cannot be found using cell or bacteriophage libraries. However, this point has not
`yet been demonstrated experimentally, perhaps because in vitro library screening
`techniques have been developed only within the last few years (76).
`Ribosome display, originally described by Mattheakis et al (77), and adopted
`to the screening of antibody libraries by Pluckthun (78, 79) and Taussig (80),
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 10
`
`

`

`349
`
`relies on the formation of a ternary complex between ribosomes, mRNA, and
`the polypeptide (see Figure 4). Complexes containing folded antibodies with a
`desired specificity are enriched by panning against immobilized ligand. The mRNA
`of the ribosome-mRNA-protein complexes is reverse transcribed to produce the
`DNA encoding the antibodies responsible for the binding of the complexes to
`the immobilized ligand. The DNA is then transcribed by RNA polymerase to
`begin another cycle of ternary complex formation and selection. Alternatively, a
`covalent link may be established directly between mRNA and the protein it encodes
`(81). Covalent mRNA-protein complexes are formed via the reactive amino acid
`analogue puromycin. The advantage of this technology is that covalent mRNA-
`protein is more stable and can be subjected to harsh screening conditions to enrich
`antibodies or other proteins with increased stability.
`The screening of combinatorial libraries of antibody genes is one of the most
`important tools in antibody engineering. The ability to screen large libraries has
`enabled the development of technologies that mimic the mammalian immune re-
`sponse, allowing the isolation of antibodies without animal immunization, and the
`engineering of antibodies with very high affinity, increased stability and improved
`effector functions.
`
`?ANTIBODY ENGINEERING
`
`Bypassing Immunization: Isolation of Antibodies
`from Large Libraries
`In the mammalian immune system, antibodies with high affinities are created
`in two stages. First, combinatorial recombination and other diversity-generating
`mechanisms result in the formation of a large collection of assembled, germline
`antibody genes. Second, B-lymphocytes expressing antibodies of low affinity but
`high specificity for a target antigen undergo a process of somatic hypermutation
`in which a high number of mutations accumulate in antibody genes, preferentially
`within the CDRs (82, 83). B-lymphocytes expressing affinity-matured antibodies
`exhibiting slower antigen dissociation kinetics are then selected and expanded (3).
`Antibody technology can successfully mimic the process by which antibodies
`are selected by the immune system. This is accomplished by cloning large libraries
`of antibody genes and selecting for binding to a desired target. If the library is
`large enough (many contain 109 different clones, similar to the number of distinct
`rearranged human antibodies) an antibody with specificity for almost any antigen
`should be represented and can be selected.
`Immune libraries were first developed as an alternative to hybridoma technol-
`ogy. Immunized rodents or humans exposed to a desired antigen through vacci-
`nation or disease have high levels of circulating antibodies to the antigen, and
`the corresponding mRNAs are highly represented among total antibody mRNAs.
`Therefore, even small libraries (∼105) from immunized donors give rise to specific
`antibodies, and because the antibody genes have experienced affinity maturation
`in vivo, antibodies can potentially be isolated that do not require further affinity
`improvement. Screening of libraries from immunized animals has been used to
`
`Annu. Rev. Biomed. Eng. 2000.2:339-376. Downloaded from www.annualreviews.org
`
` Access provided by Reprints Desk, Inc. on 07/20/19. For personal use only.
`
`Lassen - Exhibit 1058, p. 11
`
`

`

`MAYNARD (cid:2) GEOR