Phage display and selection of a site-directed randomized single-chain antibody Fv fragment for its affinity improvement

Molecular engineering of antibodies for therapeutic and diagnostic purposes

During the past ten years, monoclonal antibodies (mAbs) have taken center stage in the field of targeted therapy and diagnosis. This increased interest in mAbs is due to their binding accuracy (affinity and specificity) together with the original molecular and structural rules that govern interactions with their cognate antigen. In addition, the effector properties of antibodies constitute a second major advantage associated with their clinical use. The development of molecular and structural engineering and more recently of in vitro evolution of antibodies has opened up new perspectives in the de novo design of antibodies more adapted to clinical and diagnostic use. Thus, efforts are regularly made by researchers to improve or modulate antibody recognition properties, to adapt their pharmacokinetics, engineer their stability, and control their immunogenicity. This review presents the latest molecular engineering results on mAbs with therapeutic and diagnostic applications.

Generation of recombinant antibodies and means for increasing their affinity

Highly specific interaction with foreign molecules is a unique feature of antibodies. Since 1975, when Keller and Milstein proposed the method of hybridoma technology and prepared mouse monoclonal antibodies, many antibodies specific to various antigens have been obtained. Recent development of methods for preparation of recombinant DNA libraries and in silico bioinformatics approaches for protein structure analysis makes possible antibody preparation using gene engineering approaches. The development of gene engineering methods allowed creating recombinant antibodies and improving characteristics of existing antibodies; this significantly extends the applicability of antibodies. By modifying biochemical and immunochemical properties of antibodies by changing their amino acid sequences it is possible to create antibodies with properties optimal for certain tasks. For example, application of recombinant technologies resulted in antibody preparation of high affinity significantly exceeding the initial affinity of natural antibodies. In this review we summarize information about the structure, modes of preparation, and application of recombinant antibodies and their fragments and also consider the main approaches used to increase antibody affinity.

A novel Arnt-interacting protein Ainp2 enhances the aryl hydrocarbon receptor signalling

In an effort to better understand the Ah receptor nuclear translocator (Arnt)-dependent signaling mechanisms, we employed a phage display system to identify Arnt-interacting peptides. Human liver cDNA library was utilized to screen for Arnt-interacting peptides using an Arnt construct fused to thioredoxin (TH-ArntCDelta418). Two clones, namely Ainp1 and Ainp2 (Arnt-interacting peptide), were identified and subsequently Ainp2 was further characterized. Ainp2 interacts with TH-ArntCDelta418 in the GST pull-down and mammalian two-hybrid assays. Northern blot results revealed that Ainp2 is predominantly expressed in human liver. The putative full-length Ainp2 cDNA sequence was subsequently cloned using RACE PCR. Endogenous expression of Ainp2 was found in Jurkat cells at the mRNA and protein levels. Results from the transient transfection studies using a DRE-driven reporter plasmid and the real-time QPCR experiments examining the endogenous CYP1A1 expression showed that Ainp2 enhances the 3-methylchloranthrene-induced activity in HepG2 cells, suggesting that Ainp2 plays a role in the Arnt-dependent function

Construction and diversification of yeast cell surface displayed libraries by yeast mating: application to the affinity maturation of Fab antibody fragments

Yeast display is a powerful technology for the affinity maturation of human antibody fragments. However, the technology thus far has been limited by the size of antibody libraries that can be generated, as using current transformation protocols libraries of only between 10(6) and 10(7) are typically possible. We have recently shown that Fab antibodies can be displayed on the cell surface of Saccharomyces cerevisiae [van den Beucken, T., Pieters, H., Steukers, M., van der Vaart, M., Ladner, R.C., Hoogenboom, H.R., Hufton, S.E., 2003. Affinity maturation of Fab antibody fragments by fluorescent-activated cell sorting of yeast-displayed libraries. FEBS Lett. 546, 288-294]. This discovery and the knowledge that Fab antibodies are heterodimeric suggest that independent repertoires of heavy chain (HC) and light chain (LC) can be constructed in haploid yeast strains of opposite mating type. These separate repertoires can then be combined by highly efficient yeast mating. Using this approach, we have rapidly generated a naive human Fab yeast display library of over 10(9) clones. In addition, utilizing error-prone polymerase chain reaction, we have diversified Fab sequences and generated combinatorial and hierarchical chain shuffled libraries with complexities of up to 5 x 10(9) clones. These libraries have been selected for higher affinity using a repeating process of mating-driven chain shuffling and flow cytometric sorting.

Tailoring in vitro selection for a picomolar affinity human antibody directed against vascular endothelial growth factor receptor 2 for enhanced neutralizing activity

Vascular endothelial growth factor (VEGF) and its receptors have been implicated in promoting solid tumor growth and metastasis via stimulating tumor-associated angiogenesis. We previously identified several fully human neutralizing anti-VEGF receptor 2 (or kinase inserting domain-containing receptor (KDR)) antibodies from a large antibody phage display library. These antibodies bind specifically to KDR, block VEGF/KDR interaction, and inhibit VEGF-induced proliferation of human endothelial cells and migration of KDR+ leukemia cells. Three of these antibodies, interestingly, share an identical heavy chain variable (VH) sequence. In this report, we constructed a new library comprising the single VH paired with the variable light chain (VL) repertoire obtained from the original naïve human library. Initial in vitro selection revealed that the single VH could pair with a number of different VL while retaining its specificity for KDR. However, a consensus VH/VL pair, clone 1121, was identified after three or four rounds of selection by tailoring the stringency of the panning conditions. Clone 1121 showed a >30-fold higher binding affinity to KDR (Kd, 100 pm) because of improvement on both association and dissociation constants and blocked VEGF/KDR interaction with an IC50 of approximately 1 nm, compared with that of 3-4 nm for the parent Fab fragments. Further, clone 1121 was more potent in inhibiting VEGF-stimulated KDR phosphorylation in endothelial cells. A binding epitope mapping study on clone 1121 and one of the parent clones, 2C6, demonstrated that both antibodies interacted with the third immunoglobulin domain within the extracellular region of KDR. Several peptide phage display libraries were utilized to further examine the fine binding specificities of the two antibodies. All of the 2C6-binding peptides are cysteine-constrained, whereas clone 1121 binds to both cysteine-constrained and linear peptides. It is noteworthy that most of the 2C6-binding peptides also cross-react with clone 1121, but none of the clone 1121-specific peptides binds to 2C6, indicating that clone 1121 retained part of the original binding epitope(s) of 2C6 while gaining new binding specificity. Taken together, our observation suggests that clone 1121 may have great clinical potential in anti-angiogenesis therapy. It further underscores the efforts to identify antibodies of high affinity for enhanced biological activities.

Phage randomization in a charybdotoxin scaffold leads to CD4-mimetic recognition motifs that bind HIV-1 envelope through non-aromatic sequences

Binding of HIV-1 gp120 to T-cell receptor CD4 initiates conformational changes in the viral envelope that trigger viral entry into host cells. Phage epitope randomization of a beta-turn loop of a charybdotoxin-based miniprotein scaffold was used to identify peptides that can bind gp120 and block the gp120-CD4 interaction. We describe here the display of the charybdotoxin scaffold on the filamentous phage fUSE5, its use to construct a beta-turn library, and miniprotein sequences identified through library panning with immobilized Env gp120. Competition enzyme-linked immunosorbent assay (ELISA) identified high-frequency phage selectants for which specific gp120 binding was competed by sCD4. Several of these selectants contain hydrophobic residues in place of the Phe that occurs in the gp120-binding beta-turns of both CD4 and previously identified scorpion toxin CD4 mimetics. One of these selectants, denoted TXM[24GQTL27], contains GQTL in place of the CD4 beta-turn sequence 40QGSF43. TXM[24GQTL27] peptide was prepared using solid-phase chemical synthesis, its binding to gp120 demonstrated by optical biosensor kinetics analysis and its affinity for the CD4 binding site of gp120 confirmed by competition ELISA. The results demonstrate that aromatic-less loop-containing CD4 recognition mimetics can be formed with detectable envelope protein binding within a beta-turn of the charybdotoxin miniprotein scaffold. The results of this work establish a methodology for phage display of a charybdotoxin miniprotein scaffold and point to the potential value of phage-based epitope randomization of this miniprotein for identifying novel CD4 mimetics. The latter are potentially useful in deconvoluting structural determinants of CD4-HIV envelope recognition and possibly in designing antagonists of viral entry.

Investigation of a tetracycline-regulated phage display system

A new vector (pGZ1) was developed for bacterial phage display of antibody fragments using a transcriptional regulation element with tight control. The tet(p/) degrees -based phasmid exhibits fully suppressed scFv background synthesis in the absence of inducer and is independent of glucose as a catabolite repressor. The vector is shown to be a useful alternative to commonly used lac(p/) degrees -regulated systems.

Fine tuning of an anti-testosterone antibody binding site by stepwise optimisation of the CDRs

BACKGROUND

We have previously reported specificity improvement of an anti-testosterone monoclonal antibody (3-C4F5) by random mutagenesis of the third complementarity determining regions (CDR3s) and by phage display selection.

OBJECTIVES

Here we extend the mutagenesis strategy to the other CDRs and select the mutant libraries using two different approaches in order to further fine-tune the binding properties of this recombinant Fab fragment.

STUDY DESIGN

To improve the affinity the new mutant libraries were selected by using limiting, decreasing concentrations of biotinylated testosterone (TES) in solution and capturing the binders on streptavidin-coated microtiter plate. The specificity was improved by preincubating the mutant libraries in solution with a high concentration of the most problematic cross-reacting steroid, dehydroepiandrosterone sulfate (DHEAS).

RESULTS

In two different light chain CDR1 mutant clones isolated from the affinity pannings, the relative TES affinity was increased over 10-fold while the cross-reactivities to related steroids were preserved at the same level as in the parental combined CDR3 mutant clone. New heavy chain CDR1 and light chain CDR2 mutants showing slightly decreased cross-reactivities were isolated from specificity selections. By combining compatible mutant CDRs together we were able to create a Fab fragment with over 12-fold higher relative TES affinity and significantly lower cross-reactivity to DHEAS when compared to the original monoclonal antibody 3-C4F5.

CONCLUSIONS

Our results demonstrate that a high-affinity and selective recombinant Fab fragment working over a wide TES concentration range with clinical samples could be generated by CDR mutagenesis and phage display selection.

A simple luciferase assay for signal transduction activity detection of epidermal growth factor displayed on phage

Studies on receptor-ligand interactions are important for the design of agonists or antagonists of natural ligands. We developed a luciferase reporter assay to screen epidermal growth factor receptor (EGFR) binding molecules rapidly for their ability to stimulate or inhibit signal transduction. Human EGF displayed on fd filamentous phage presented an activity similar to soluble EGF when tested for binding to the EGFR, for induction of cell cycle progression or in the luciferase assay. Two libraries of human EGF variants displayed on phage were constructed in which the aspartic acid residue at position 46 or the arginine residue at position 41 were randomised. EGF mutants displayed on phage were screened in parallel for binding to the EGFR using an ELISA assay and for transducing activity using the luciferase assay. Regarding the 46 position, most of the mutants retained the ability to bind the EGFR and their transducing activity corresponded perfectly with their binding. For the more crucial 41 position, only the wild-type EGF was able to bind the EGFR. Our approach allowed a simple determination of crucial positions and paved the way for identification of agonists with altered transduction activity.

Determination of the binding affinity of an anti-CD34 single-chain antibody using a novel, flow cytometry based assay

A single chain antibody (scFv) was constructed from a hybridoma expressing the anti-CD34 monoclonal antibody My10. The scFv was expressed in the mouse fibroblast cell line NIH 3T3, and purified from culture supernatant via an epitope tag fused to the C-terminus of the protein. The scFv equilibrium dissociation constant (KD) was determined to be 2.4 X 10(-7) M using a fluorescence based flow cytometry assay involving recognition of the epitope tag, and bound approximately 24-fold less avidly to CD34 expressing KG-1a cells than the native antibody My10. This novel and previously unreported method for determining antibody binding affinity offers several advantages over alternative methods. It is rapid and simple, and unlike methods that directly label the antibody, it involves no covalent modifications of antibody variable domain residues that could potentially interfere with antigen binding. The KD for the anti-CD33 antibody HuG1 (Caron et al. (1992) The biological and immunological features of humanized M195 (anti-CD33) monoclonal antibodies. Cancer Res. 52, 6761-6767) was determined as well. The close agreement of this value and the previously reported value, determined by a radioligand competition method, validates the use of this assay for antibody affinity determination. We discuss various potential applications for this anti-CD34 scFv.

Reliable cloning of functional antibody variable domains from hybridomas and spleen cell repertoires employing a reengineered phage display system

A prerequisite for the use of recombinant antibody technologies starting from hybridomas or immune repertoires is the reliable cloning of functional immunoglobulin genes. For this purpose, a standard phage display system was optimized for robustness, vector stability, tight control of scFv-delta geneIII expression, primer usage for PCR amplification of variable region genes, scFv assembly strategy and subsequent directional cloning using a single rare cutting restriction enzyme. This integrated cloning, screening and selection system allowed us to rapidly obtain antigen binding scFvs derived from spleen-cell repertoires of mice immunized with ampicillin as well as from all hybridoma cell lines tested to date. As representative examples, cloning of monoclonal antibodies against a his tag, leucine zippers, the tumor marker EGP-2 and the insecticide DDT is presented. Several hybridomas whose genes could not be cloned in previous experimental setups, but were successfully obtained with the present system, expressed high amounts of aberrant heavy and light chain mRNAs, which were amplified by PCR and greatly exceeded the amount of binding antibody sequences. These contaminating variable region genes were successfully eliminated by employing the optimized phage display system, thus avoiding time consuming sequencing of non-binding scFv genes. To maximize soluble expression of functional scFvs subsequent to cloning, a compatible vector series to simplify modification, detection, multimerization and rapid purification of recombinant antibody fragments was constructed.

Affinity improvement of single antibody VH domains: residues in all three hypervariable regions affect antigen binding

BACKGROUND

Through antibody engineering, immunoglobulins can be tailored for their particular application. In this respect, small recognition units are desired for the targeting of antigens in obstructed locations like solid tumors.

OBJECTIVES

To design efficient, minimum size recognition units, heavy chain variable regions (VH) had previously been modified for the use as antigen specific, single domain antibody fragments. To develop a rational approach to improve affinity, antigen binding is investigated here by analysing the effect of randomisations of CDR1 and 2 residues in VH domains specific for hapten and protein ligands.

STUDY DESIGN

Randomised repertoires were displayed on phage and affinity selected to improve and analyse antigen binding. Affinities of newly selected VH domains were determined in their soluble format to assess the role of modified residues in binding.

RESULTS

In four of five randomisation experiments, a new VH with an improved antigen affinity compared to the primary VH was selected. Dissociation constants decreased from 160 nM to 25 nM or 47 nM (CDR1 or CDR2 randomisation of an anti-Ox VH), from 300 nM to 31 nM (CDR2 randomisation of an anti-NIP VH) and from 3.1 microM to 1.6 microM (CDR2 randomisation of an anti-lysozyme VH).

CONCLUSIONS

Thus the affinity of VH domains can be improved after site specific, secondary randomisations in CDR1 and CDR2, phage display and antigen selection. As differences in the CDR3 sequences had formed the only difference between the primary VH domains used in this study, the effect of CDR1 and CDR2 mutations of affinity is consistent with a participation of all three CDRs in antigen binding by single VH domains.

Affinity maturation of recombinant antibodies using E. coli mutator cells

BACKGROUND

Phage libraries can display repertoires of antibodies which are greater in number than the mammalian immune response. However, the selected antibodies often have low binding affinity to their target antigen or hapten (KD below 10(-6) M), which is characteristic of the primary immune repertoire. There is a need for procedures to mimic somatic hypermutation through antigen driven affinity maturation, thereby increasing the affinity of selected immunoglobulins.

OBJECTIVE

To investigate the effectiveness of mutation and affinity selection of recombinant antibody genes with mutator E. coli cells, incorporating phage-display strategies.

STUDY DESIGN

Unique human scFvs were selected from a naive Fd-phage library. These genes were mutated by propagation in mutD5 mutator E. coli cells (mutD5-FIT) which were competent for Fd (M13) based phagemid transfections and generated point mutations (transversions and transitions) in the scFv genes. Individual phage-displayed scFvs were affinity selected from the mutation library and were assayed as soluble scFvs by ELISA and BIAcore for binding to antigen.

RESULTS

The in vivo mutation of phage-displayed scFvs in E. coli mutD5-FIT, combined with affinity selection against antigen, produced scFv molecules with improved binding activity. The point mutations which resulted in single amino acid substitutions frequently produced ten fold increases in apparent binding affinity. Structural comparisons revealed that these point mutations were in framework regions (adjacent to the CDRs) and within the CDRs. In one case the apparent affinity of an anti-glycophorin scFv after mutation in the VL framework region close to CDR3 increased by 10(3). However, this increase in apparent affinity was accompanied by an increased propensity to dimerise and form aggregates.

CONCLUSIONS

A strategy for the rapid affinity maturation of scFv and Fab antibody fragments has been developed which utilises mutator strains of E. coli and incorporates phage display of antibody repertoires (libraries).

Contribution of antibody heavy chain CDR1 to digoxin binding analyzed by random mutagenesis of phage-displayed Fab 26-10

We constructed a bacteriophage-displayed library containing randomized mutations at H chain residues 30-35 of the anti-digoxin antibody 26-10 Fab to investigate sequence constraints necessary for high affinity binding in an antibody of known crystal structure. Phage were selected by panning against digoxin and three C-16-substituted analogues. All antigen-positive mutants selected using other analogues also bound digoxin. Among 73 antigen-positive clones, 26 different nucleotide sequences were found. The majority of Fabs had high affinity for digoxin (Ka 3.4 x 10(9) M-1) despite wide sequence diversity. Two mutants displayed affinities 2- and 4-fold higher than the parental antibody. Analysis of the statistical distribution of sequences showed that highest affinity binding occurred with a restricted set of amino acid substitutions at positions H33-35. All clones save two retained the parental Asn-H35, which contacts hapten and hydrogen bonds to other binding site residues in the parental structure. Positions H30-32 display remarkable diversity, with 10-14 different substitutions for each residue, consistent with high affinity binding. Thus complementarity can be retained and even improved despite diversity in the conformation of the N-terminal portion of the H-CDR1 loop.

Basis for selection of improved carbohydrate-binding single-chain antibodies from synthetic gene libraries

A technique is described for the simultaneous and controlled random mutation of all three heavy or light chain complementarity-determining regions (CDRs) in a single-chain Fv specific for the O polysaccharide of Salmonella serogroup B. Sense oligonucleotides were synthesized such that the central bases encoding a CDR were randomized by equimolar spiking with A, G, C, and T at a level of 10% while the antisense strands contained inosine in the spiked regions. Phage display of libraries assembled from the spiked oligonucleotides by a synthetic ligase chain reaction demonstrated a bias for selection of mutants that formed dimers and higher oligomers. Kinetic analyses showed that oligomerization increased association rates in addition to slowing dissociation rates. In combination with some contribution from reduced steric clashes with residues in heavy-chain CDR2, oligomerization resulted in functional affinities that were much higher than that of the monomeric form of the wild-type single-chain Fv.

Antibody VH domains as small recognition units

To develop immunoglobulin based recognition units of minimum size, a human heavy chain variable domain (VH) was designed for selection of phage displayed VH. Non-specific binding of the VH through its interface for the light chain variable domain (VL) was prevented through three mutations (G44E, L45R and W47G) in this interface. These mutations were introduced to mimic camelid antibody heavy chains naturally devoid of light chain partners. The third hypervariable loop of the modified VH was then randomised to yield a repertoire of 2 x 10(8) independent clones, which was displayed on phage and selected through antigen binding. VH clones specific for hapten and protein antigens were isolated. Soluble VH was expressed with an isoleucine residue at position 47 to improve expression and stability compared to VH containing a glycine residue at this position, which however was preferable for phage selection. Affinities of soluble VH for hapten were between 100 nM and 400 nM. The VH domains were highly specific, stable and well expressed in Escherichia coli. These positive biophysical properties and their small size make them attractive for biotechnological applications.

Identification of residues in the monoclonal antitumor antibody L6 important for binding to its tumor antigen

L6 is a monoclonal antitumor antibody which recognizes an epitope located in a 42-residue extracellular domain of a tumor-associated approximately 22 kDa glycoprotein antigen. The L6 mAb localizes to solid tumors in vivo and triggers complement activation and antibody-dependent cellular cytotoxicity. It has been the subject of phase I clinical trials. Previously, we had reported the derivation and analysis of a three-dimensional model of the L6 Fv. The model suggests that L6 displays a generally aromatic CDR surface. We aim at improving the affinity for tumor antigen of L6 by in vitro mutagenesis. As the first step toward this end, we have attempted to identify residues critical for the binding of L6 to tumor antigen. On the basis of the model, seven residues were selected which we thought may be critical for L6 antigen binding. Criteria for the selection of these residues were their accessibility and central position on the CDR surface and the residue character. Large polar or charged residues such as arginine, asparagine, and tyrosine were preferred. Nine site-specific single and double mutants were generated using oligonucleotide-directed mutagenesis in an M13 expression vector encoding the L6 Fab. The binding of these mutant Fabs to the L6 tumor antigen and a set of three anti-idiotypic antibodies was quantified in an ELISA. In eight out of nine mutants, binding to L6 tumor antigen was either abolished or substantially reduced. In contrast, the binding of the mutants to the anti-idiotypic antibodies was largely unaffected, suggesting that no significant structural perturbations were introduced as a consequence of these mutations.

Genetically engineered antibodies for diagnostic pathology

Antibody genes can be cloned, genetically manipulated, and expressed in both homologous and heterologous expression systems to produce viable antigen-binding proteins complete with natural effector functions. Manipulation of antibody genes permits the expression of fusion proteins or truncated proteins that retain antigen-binding activity. The new antibody technologies are becoming increasingly sophisticated, permitting the alteration of antigen-binding responses, the transfer of antigen specificity between antibodies, and the expression of minimal-size antigen-binding protein domains. These new molecules have been made mostly for studies on function or to provide molecules suited for in vivo diagnosis and therapy; very few have been specifically designed for, or used for, diagnostic histopathology. We describe here the adaptation of small antibody derivatives for use in immunohistochemistry. Molecules suitable for this purpose need only to possess specific antigen-binding ability and some means of detection of antigen-bound material. Detection could be by recognition of a genetically fused flag or tag epitope, by the fusion of an enzyme whose activity can be assayed, or by fusion with a protein that can interact with pre-existing histopathological reagents.

Antibody design: beyond the natural limits

Dissection of antibody-antigen interactions requires a knowledge of antibody structure, the ability to model accurately the conformation of antibody-combining sites, and an understanding of the energetic factors governing the interactions. When this understanding has reached the point where the molecular shape and chemical character of a combining site necessary to define a particular specificity and binding requirement can be designed, the antibody repertoire will have been extended 'beyond the natural limits'.

Catalytic antibodies: perusing combinatorial libraries

Combinatorial libraries are a promising alternative for isolating catalytic antibodies produced by the immune system in response to the transition-state analog of a given reaction. Large, diverse panels of antibodies with high affinity for the transition-state analog can be isolated using screening or selection approaches. Furthermore, we have estimated that nucleotide sequences that bear close similarity to the sequence for a known catalytic antibody occur in combination at frequencies sufficient for their detection in such libraries.

'Camelising' human antibody fragments: NMR studies on VH domains

A human heavy chain variable domain (VH) was expressed in bacteria for structural analysis by NMR spectroscopy. NMR analysis was initially impossible due to the short transverse proton relaxation time of the VH, probably caused by aggregation through the exposed interface naturally in contact with the light chain. The relaxation time was improved to normal values when this interface was mutated to mimic heavy chains of camel antibodies naturally devoid of light chains and through the use of the detergent CHAPS. Assignment of NMR signals will now be possible after isotopic labeling. Implications for the design of VH domains as minimum size immunoreagents are outlined.

Surface display of antibodies

To screen antibody libraries that contain many millions of different clones, a selection system is required with an efficiency comparable to that of the immune system. This can be achieved by displaying antibodies on the surface of microorganisms containing the antibody's gene, analogous to the expression of the IgM antigen receptor on the surface of unactivated B-lymphocytes. Specific clones can then be selected using immobilized antigens. The minor coat protein of filamentous phages, pIII, which initiates the infection of E.coli by binding to their F-pili, and the major coat protein, pVIII, have been used as carriers for displaying antibodies on the phage surface. Recombinant antibodies have also been targeted to the cell surface of bacteria by fusing them with outer membrane components derived from lipoproteins, OmpA and an IgA protease. However, only the pIII system has been routinely used for screening antibody libraries. Here we describe the various antibody surface display systems and the screening of antibody libraries generated from the gene repertoire of lymphocytes and by gene synthesis. Finally, we have made a short comparison of the bacterial production of Fabs versus single chain antibodies (scFv).