TY - JOUR
AU - McCoy, John M.
AB - Abstract The `FLITRX' random peptide library, consisting of dodecamer loop peptides displayed on a thioredoxin-flagellin scaffold on Escherichia coli, was used to select peptide sequences with affinity for a monoclonal antibody. These peptides were further screened for pH- and metal-sensitive antibody binding. Several zinc-sensitive peptides were identified, termed `switch epitopes'. A soluble, monomeric thioredoxin loop (`Trxloop') insertion analog of a FLITRX switch epitope was constructed and its antibody binding properties were characterized by Western blots. Zinc-dependent antibody recognition was maintained in the Trxloop protein although the apparent antibody affinity was lower. This Trxloop protein bound to an immobilized metal affinity chromatography matrix, similar to a `histidine-patch' thioredoxin variant, and was reversibly precipitated by 1 mM Zn2+ or Cu2+ ions. Residues important for zinc and antibody binding were determined by site-directed mutagenesis. The Trxloop antibody affinity was increased by saturation mutagenesis. Biotinylated Trxloop (`Biotrxloop') variants of the original and improved affinity Trxloop proteins were constructed and characterized by surface plasmon resonance measurements. Increased antibody affinity was partially due to a slower antibody desorption rate, although the relative adsorption rates were dependent on the amount of immobilized Biotrxloop protein, indicating an influence of avidity on the apparent affinity. Introduction Application of nature's rules and mechanisms in the rational design of novel proteins with desired functions is an important goal of protein engineering. Many cellular biological processes are regulated by recognition-specific interactions of proteins with other biomolecules, including other proteins, cell membrane components, nucleic acid sequences and enzymatic substrates. These protein–biomolecule interactions are often modulated by chemical modification or by changes in the cellular solvent conditions, including pH, temperature and metal ion concentration. Two well-known examples of proteins with binding affinity for other biomolecules controlled by metal ions are the calcium-modulated `EF hand' domain found in calmodulin and related proteins (Nakayama et al., 1992) and the zinc-modulated `zinc finger' domain found in numerous DNA-binding proteins (Kriwacki et al., 1992; Becker et al., 1995). Engineered proteins with controlled binding to other biomolecules have many potential applications, including therapeutic and industrial enzymes and inhibitors, signal transduction proteins, biosensors, biomolecular machines for information processing and storage (`bioorganic transistors') and in affinity purification of proteins. Researchers have used various methods to engineer binding switches into proteins. Some researchers (Miwa et al., 1991; Tawfik et al., 1994; Stayton et al., 1995) used chemical modification or conjugation with other compounds to yield proteins with pH-, temperature- and calcium-sensitive binding or enzymatic activity switches. Others (Gallagher et al., 1993; Walker et al., 1994; Neri et al., 1995; Smith et al., 1995; Briand et al., 1997) used molecular biology techniques to add or remove calcium-, copper- and zinc-sensitive peptide switches in various proteins. A number of fusion peptide affinity tags with binding switches have also been developed for use in protein purification, including the pH-sensitive Strep-tag peptide (Schmidt and Skerra, 1993) and the calcium-sensitive FLAG peptide (Hopp et al., 1988). The His-6 tag (Hochuli et al., 1988) can also be considered a pH and metal ion binding switch. The development of random peptide libraries displayed on phage and bacterial surfaces has resulted in a new set of tools for investigating specific protein interactions with other molecules. Two research groups (Bianchi et al., 1995; Ivanenkov et al., 1995) engineered naturally occurring calcium- and zinc-binding peptide motifs into phage display libraries to create conformationally constrained peptides, which can be screened for binding against any target ligand in a metal-dependent manner. The FLITRX Escherichia coli thioredoxin (Trx) scaffold random peptide library (Lu et al., 1995) has previously been used to identify conformationally constrained dodecamer peptide sequences specifically recognized by monoclonal antibodies (mAbs). The FLITRX peptide library has an estimated diversity of 1.77×108 different random dodecamer loop sequences and does not contain any predefined structural motif. Thus, novel peptide sequences might be identified in this library that mimic the functions of protein domains found in nature. We wanted to identify FLITRX peptides with pH- or metal ion-sensitive mAb recognition, termed `switch epitopes'. This work summarizes our investigation of this concept and represents an alternative approach to protein engineering; one in which a desired molecular function is identified through random selection and screening, rather than rational design. We demonstrate that zinc-sensitive switch epitope peptides recognized by a mAb can be identified using the FLITRX library and that these peptides can be grafted on to soluble, monomeric thioredoxin proteins and retain their switch epitope properties. We also show that the peptide–mAb affinity can be further improved by saturation mutagenesis while retaining the switch epitope property. Materials and methods Materials All chemicals and reagents were of reagent grade or better and were purchased from Sigma Chemical (St. Louis, MO) unless indicated otherwise. Restriction enzymes were from New England Biolabs (Beverly, MA), unless indicated otherwise. Oligonucleotides were provided by the Genetics Institute DNA Synthesis Group. DNA sequencing reagents (Sequenase 2.0 kit) were obtained from United States Biochemical (Cleveland, OH). Mouse anti-human interleukin-8 (IL-8) mAb, clone HIL8-NR7, was purchased from Devaron (Dayton, NJ). Anti-thioredoxin mAbs, GI TD1/9.5.2, GI TD1/7.2.27 and GI TD1/1.5.12 were provided by the Immunoassay Department at the Genetics Institute. Rabbit anti-mouse IgG polyclonal sera was purchased from Zymed Immunochemicals (South San Francisco, CA) or from Pharmacia Biosensor (Piscataway, NJ). Purified protein-A, radiolabeled with 125I, was purchased from DuPont New England Nuclear (Boston, MA). Polyclonal sheep anti-mouse IgG–horseradish peroxidase (HRP) conjugate and electrochemiluminescent (ECL) detection reagents were purchased from Amersham Life Science (Arlington Heights, IL). Avidin–alkaline phosphatase conjugate, alkaline phosphatase buffer, 5-bromo-4-chloro-3′-indolylphosphate p-toluidine (BCIP) and nitro-blue tetrazolium chloride (NBT) were purchased from Pierce Chemical (Rockford, IL). Chromatography media were purchased from Amersham Pharmacia Biotech (Piscataway, NJ), unless indicated otherwise. Growth media and E.coli strains 10× M9 salts were prepared by dissolving 60 g of Na2HPO4, 30 g of KH2PO4, 5 g of NaCl and 10 g of NH4Cl in 800 ml of deionized water, adjusting the pH to 7.4 with NaOH, followed by addition of deionized water to a volume of 1 liter. IMC growth medium was prepared by addition of 10% casamino acids, 10% 10× M9 salts, 0.5% glucose, 1 mM MgCl2 and 0.1 mM CaCl2 and 75% deionized water. All media and stock solutions were autoclaved or sterile filtered prior to use. Ampicillin (Amp) was added to make final concentration of 100 μg/ml. Soluble Trxloop and Biotrxloop proteins were expressed in E.coli strain GI-934. FLITRX library peptide selection Polystyrene tissue culture plates, 60 mm diameter (Nunc, Naperville, IL), were coated with 20 μg of purified murine anti-human IL-8 mAb diluted in 2.0 ml of deionized (DI) water for 2 h, with gentle shaking. The plates were then blocked with 10 ml of 50 mM tris(hydroxymethyl)aminomethane (Tris), pH 7.4, 150 mM NaCl (TS) buffer containing 1% non-fat dry milk and 1% methyl α-d-mannopyranoside. Shear elution of bound E.coli was performed by vortexing the plates for 30 s and washing with IMC growth medium. Zinc elution was performed on a pool of previously 3× shear selected FLITRX bacteria by adding 2.5 mM ZnCl2 to the wash solution during the final rinse step, after five normal wash steps of the mAb-coated plates. The bacteria were cultured overnight to saturation conditions in casamino acid (CAA)–Amp liquid medium and then plated on 150 mm agar CAA–Amp plates. Colonies were transferred to nitrocellulose membranes and grown on CAA–Amp–tryptophan (Trp) plates for 6 h at 30°C to induce flagellin protein expression. The original colonies were regrown at 30°C for 4 h. The membranes containing duplicate colonies were lysed overnight in a TS buffer containing 1% non-fat dry milk, 40 μg/ml lysozyme and 1 μg/ml DNase I. Colonies of interest were grown to saturation conditions in 5 ml test-tube cultures, mixed with sterile 50% glycerol and frozen at –80°C in cryovials for long-term storage. DNA sequencing of novel plasmids DNA sequencing of plasmids was performed manually using the USB Sequenase 2.0 kit according to the manufacturer's directions. The synthesized oligonucleotides 5′-GACAGTTTTG CACGGATGT-3′ and 5′-TCAGCGATTT ATCCAGAAT-3′ were used as primers for the top and bottom strands, respectively, with α32dATP used as the radioisotope label. SDS-PAGE A 2 ml volume of E.coli cell culture containing overexpressed protein was centrifuged and resuspended in 200 μl of 20 mM Tris buffer, pH 7.5 and diluted with 200 μl of 2× SDS buffer with 2% β-mercaptoethanol (BME) and heated to 90°C for 20 min to lyse the cells. A 20 μl volume of of each type of E.coli cell lysate was then loaded into wells of 10% Tricine Mini Gels (Novex) and run for 2 h at a constant current of 30 mA per gel. The gels were then stained with Coomassie Brilliant Blue R-250 dye. Dot Western blot screening procedures For identification of initial FLITRX hit clones, the membranes with lysed bacterial colonies were probed for positive mAb binding by binding the primary mAb in TS buffer for 2 h with gentle mixing at a concentration of 1 μg/ml in a total volume of 50 ml, followed by three 10 min rinses in TS buffer. Rabbit anti-mouse IgG polyclonal antibody, preincubated against strain GI-808 E.coli flagellin containing cell lysate, was then added to the membrane for 2 h, followed by three 10 min rinses with TS buffer. A 20 μl volume of of 125I-labeled protein-A was added at a 1:1000 dilution to 20 ml of TS buffer per blot and allowed to bind overnight. The blots were then rinsed three times in TS buffer, dried and exposed to X-ray film (X-OMAT-AR, Eastman Kodak, Rochester, NY) from 1 to 12 h, depending on the radioactivity of the blots. FLITRX hits were screened for switch epitope mAb binding by a variation of the previous immunoblot procedure. Single colonies of each selected mutant were picked with toothpicks and used to inoculate 200 μl of sterile IMC medium in the wells of polystyrene tissue culture plate. These liquid cultures were grown overnight to saturation and a colony replication tool was used to inoculate the surface of nitrocellulose membranes placed on top of 150 mm CAA–Amp agar plates with a duplicate culture pattern of colonies corresponding to those in the wells. The membrane inoccula were grown for 6 h at 30°C. Then the membranes were transferred to CAA–Amp–Trp plates and grown for 6 h at 30°C to induce protein expression. FLITRX dot-blot colonies were lysed by addition of lysozyme as described previously. Trxloop colonies were lysed by performing three freeze–thaw cycles of –80°C for 10 min, followed by 30°C for 10 min on the nitrocellulose membranes. All membranes were blocked overnight with 1% non-fat milk in water and then probed with the IL-8 mAb, using a variant of the Western blot procedure used to identify the original positive colonies. In some immunoblotting experiments, the metal salts CaCl2, ZnCl2, CuCl2, NiCl2 or chelators (ethylenediaminetetraacetic acid, EDTA) were added to TS buffer or the buffers 2-N-morpholinoethanesulfonic acid (MES), 2-(N-cyclohexylamino)ethanesulfonic acid (CHES) or acetate were used instead of Tris to obtain different solution pH values. The membranes were probed for positive mAb binding by binding the primary mAb for 2 h with gentle mixing at a concentration of 1 μg/ml in 50 ml buffer, with 10 μM to 2.5 mM CaCl2, ZnCl2, CuCl2, NiCl2 or 2.5 mM EDTA in pH 7.5 TS buffer or 50 mM pH 6.0 MES, 150 mM NaCl buffer or 50 mM pH 9.0 CHES, 150 mM NaCl buffer, followed by three 10 min rinses in the same buffer. Bound IL-8 mAb was then detected in the same manner as described above (incubation with polyclonal rabbit anti-mouse antibodies and 125I-labeled protein-A). Initial Western blot experiments with 2.5 mM ZnCl2 in the buffer resulted in non-specific precipitation of the IL-8 mAb; 0.1 mM ZnSO4 was found to yield good discrimination between zinc inhibited and non-zinc inhibited mAb epitopes. For Western blots of proteins isolated by SDS-PAGE, the proteins were transferred from tricine gels to nitrocellulose membranes in a Hoefer TE series Transfor unit in Western transfer buffer for 2 h at 60 mA. For soluble protein immunoblots, 50 μl of each soluble fraction of cell lysate were loaded on to a nitrocellulose or PVDF membrane and allowed to bind at 25°C for 1 h. Membranes were blocked and incubated with IL-8 mAb, as described above. Bound IL-8 mAb was labeled by 2 h incubation with 50 μg of alkaline phosphatase conjugated polyclonal goat anti-mouse antibody or 50 μg of polyclonal sheep anti-mouse IgG–horseradish peroxidase conjugate. The amount of bound mAb was then detected by addition of BCIP/NBT or ECL reagents and exposure of the membranes to X-ray film for 10–60 s in a darkroom. Construction of thioredoxin loop mutants For construction of the H6A mutant, pAL(H1,6)trxA was cut at two unique sites with AflII and NdeI restriction enzymes and the large fragment was gel purified on agarose gel. The two synthetic oligonucleotides 5′-TATGTCTGAT AAAATTATTG CTCTGACTGA TGATTCTTTT GATACTGATG TAC-3′ and 5′-TTAAGTACAT CAGTATCAAA AGAATCATCA GTCAGAGCAA TAATTTTATC AGACA-3′ were phosphorylated, annealed and ligated into the AflII and NdeI sites with T4 ligase, resulting in the vector pAL(H6A)trxA containing the two mutations H1S (wild-type reversion) and H6A. For creation of active site loop mutants, pALtrxA, pALBio(81)trxA-BirA (Smith et al., 1998) or the modified pAL(H6A)trxA plasmids were cut at a unique RsrII/CspI site using CspI restriction enzyme (Stratagene, La Jolla, CA). Oligonucleotides encoding top and bottom strands for each desired loop peptide insertion were designed with respect to optimal codon usage in E.coli, synthesized, phosphorylated, annealed and ligated using standard protocols (Sambrook et al., 1989). E.coli strain GI-934 were transformed with the ligated plasmids by electroporation and plated on to CAA–Amp plates. All mutations were confirmed by DNA sequencing. Construction of Z9 Trxloop peptide saturation mutagenesis library Two synthetic semi-random oligonucleotides were synthesized: 5′-GACTGACTG*G TCCAC1 AGGTA CATCCAAAAC ACTTCGGTCA CGCTCCAATC G*GTCCTCAGT CAGTCAG-3′ and 5′-CTGACTGACTG AGGACC-3′. Each base at an underlined region was synthesized with ~86% of the desired specific base and with 14% random N (25% A/25% C/25% G/25% T), with the exception of the first C1 in bold, which was synthesized with 14% V (33% A/33% G/33% T), to avoid introduction of a stop codon insertion (CAG→TAG). Statistical probability calculations predicted that the resultant oligonucleotide population would contain 7% with no mutations, 20% with single mutations, 29% with two mutations and 44% with three or more mutations. The * symbols denote AvaII restriction cleavage sites on either side of the semi-random region. Preparation of the final double-stranded semi-random oligonucleotides and E.coli library was accomplished by the procedure described previously for the FLITRX oligonucleotide library construction (Lu et al., 1995). The library of E.coli clones had an estimated diversity of 3.83×104 different clones, based on extrapolation of numbers of transformants plated on CAA–Amp plates. The library was estimated to contain ~75% single oligonucleotide inserts, based on SDS-PAGE of induced cell lysates of 24 random colonies. Expression and purification of Trxloop proteins E.coli strain GI-934 containing pALtrxA or pAL(81)Biotrx-BirA containing loop insertion peptides of interest were plated on CAA–Amp plates and incubated overnight at 30°C to obtain visible colonies. Single colonies were used to inoculate plasmid media in shaker tubes and grown overnight at 30°C to saturation. This culture was used to inoculate IMC media to an initial optical density at 550 nm wavelength (OD550) of 0.05 and grown at 30°C to 0.5 OD550. Trxloop protein expression was then initiated by addition of l-tryptophan to a final concentration of 100 μg/ml and the cultures were grown further for 6 h at 37°C. Cells were then centrifuged, washed and resuspended in 50 ml of 25 mM Tris buffer, pH 7.5, 2 mM EDTA. 1 mM phenylmethanesulfonyl fluoride (PMSF) and 1 mM p-aminobenzamidine (PABA) were added to Trxloop protein cell lysates, while one tablet of Complete protease inhibitor cocktail (Boehringer Mannheim/Roche Molecular Biochemicals, Indianapolis, IN) was added to Biotrxloop lysates. The cells were then lysed by two passes through a French press at a pressure of 1000 psi. Cell lysates were centrifuged at 100 000 g for 30 min, filtered through a 0.22 μm filter and loaded on to a Q Sepharose Fast Flow anion-exchange column. Trxloop and Biotrxloop proteins were eluted by applying a gradient of 0.0–0.5 M NaCl and relevant fractions were identified by SDS-PAGE, pooled and concentrated in an Amicon ultrafiltration cell (Amicon, Beverly, MA) using a YM-10 membrane. Trxloop proteins were loaded on to a HiPrep Sephacryl S-100 size-exclusion column pre-equilibrated with 150 mM NaCl, 50 mM Tris, pH 8.0, 2 mM EDTA buffer. Elution fractions containing pure Trxloop protein were concentrated with Amicon Centriprep-10 concentrators and stored at 4°C. Biotrxloop proteins were loaded on to a HiPrep Sephacryl S-200 size-exclusion column pre-equilibrated with 300 mM NaCl, 50 mM Tris, pH 8.0, 2 mM EDTA buffer. Fractions containing Biotrxloop proteins were concentrated on a YM-10 membrane and loaded on to a Toyopearl Phenyl 650-S column (TosoHaas, Montgomeryville, PA) in 2.0 M NaCl, 25 mM Tris, pH 8.0 buffer and eluted with a decreasing gradient of NaCl (2.0–0.0 M NaCl), desalted into 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), pH 7.5, 150 mM NaCl buffer and stored at 4°C. Optical densiometric measurement of Trxloop variant relative IL-8 mAb affinities The Western dot-blot autoradiogram x-ray film of IL-8 mAb binding to immobilized Trxloop proteins on a membrane was scanned with a light-transmitting optical densiometer and saved as a TIFF format computer image. All image intensity measurements were performed on a Macintosh computer using NIH Image version 1.62 software, with an intensity scale of 0–255, where 0 = white (no mAb binding) and 255 = black (intense mAb binding). The mean pixel intensity of each immobilized Trxloop variant dot region was determined by averaging the pixel values obtained by sampling relatively homogeneous dot-blot regions with a 10-pixel diameter circle. Care was taken to avoid non-homogeneous local dark spots resulting from radiolabeled protein-A aggregates. The background image intensity was an average of five measurements taken at different non-Trxloop positions on the immunoblot. The average background intensity was then subtracted from each of the Trxloop variant mean intensity values. IMAC retention of Trxloop proteins The retention of wild-type Trx and Z9 Trxloop proteins on nitrilotriacetic acid (NTA) immobilized metal ion affinity chromatography (IMAC) resin charged with Zn2+ was performed as described by Lu et al. (1996), using NTA resin (Qiagen, Valencia, CA) charged with 0.5 M ZnSO4 and equilibrated with 25 mM Tris–HCl, pH 7.5, 100 mM NaCl, 1 mM imidazole buffer (buffer A). A 50 μl volume of purified wild-type thioredoxin or purified Trxloop Z9 protein was injected on to the column at a flow rate of 3 ml/min. The column was washed with two bed volumes of buffer A, followed by development of a 10 bed volume linear gradient of buffer B (buffer A with 100 mM imidazole). Eluting protein was detected by monitoring the absorbance at 280 nm. Assay of thioredoxin thiol oxidation state Ellman's reagent (Ellman, 1959), 5,5′-dithiobis(2-nitrobenzoic acid) (DTNB) (Pierce Chemical, Rockford, IL), was dissolved in 0.1 M sodium phosphate buffer, pH 8.0, to give a concentration of 4 mg/ml. Wild-type recombinant E.coli thioredoxin (Calbiochem, La Jolla, CA) was dissolved in 0.1 M phosphate buffer, pH 8.0 and stored with atmospheric oxygen for 5 days to allow intrachain disulfide bond oxidation. Reduced wild-type Trx was prepared by reacting 5 μl of BME with 250 μl of 0.5 mg/ml wild-type Trx for 30 min, followed by removal of BME on P-6 Bio-Spin desalting chromatography columns (Bio-Rad Laboratories, Hercules, CA). Freshly reduced, desalted wild-type Trx and Trxloop Z9 protein were mixed with 50 μl of DTNB solution and 2.5 ml of phosphate buffer and allowed to react for 15 min at 25°C. The absorbance of each solution was measured at 412 nm. A molar absorbance value of 14 150 l/mol.cm was used to determine free thiol concentration (Riddles et al., 1983). Protein concentrations were determined by using molar absorbance values measured at 280 nm of 13 700 l/mol.cm for wild-type Trx and the Trxloop Z9 variant (Holmgren and Reichard, 1967). Surface plasmon resonance affinity measurements All surface plasmon resonance (SPR) experiments were performed using chips and chemical coupling reagents purchased from Pharmacia Biosensor on a BIAcore 2000 instrument. Sterile filtered, degassed 10 mM HEPES, pH 7.4, 150 mM NaCl, 0.005% Tween-20 buffer [HBS(T)] also containing either 3.4 mM EDTA or various concentrations of ZnSO4 was used as the buffer for all SPR experiments. Purified Trxloop proteins, IL-8 mAb and rabbit anti-mouse IgG polyclonal antibodies were chemically immobilized on Pharmacia Biosensor Research Grade CM5 carboxymethyldextran chips by injection of N-hydroxysuccinimide (NHS) and N-ethyl-N′-(dimethylaminopropyl)carbodiimide (EDC), followed by neutralization of unreacted N-hydroxysuccinimide esters with ethanolamine. A number of SPR immobilization formats were tested using purified Trxloop proteins and the IL-8 mAb. An initial strategy of covalently immobilizing the IL-8 mAb or the Trxloop protein by amine coupling chemistry resulted in very poor sensitivity in the case of the IL-8 mAb and no detectable binding in the case of amine coupled Trxloop proteins. The IL-8 mAb epitope region of the Z9 Trxloop proteins contains a lysine and one or more histidine residues with amines that may react in the NHS/EDC coupling chemistry. Thus, Trxloop mAb recognition may have been eliminated or sterically blocked upon amine coupling of basic epitope residues. The Fab epitope recognition region of the IL-8 mAb may have also been chemically inactivated during amine coupling procedures. A second assay format involving chemical immobilization of rabbit anti-mouse polyclonal antibodies by amine coupling chemistry, followed by affinity capture of the IL-8 mAb also resulted in very poor sensitivity to Trxloop protein binding. Steric effects, low mAb surface density and the low molecular mass of Trxloop proteins may have contributed to the low sensitivity of this assay. A third SPR immobilization strategy involved capturing Trxloop proteins with amine-coupled monoclonal anti-Trx mAbs GI TD1/9.5.2, GI TD1/7.2.27 and GI TD1/1.5.12, previously demonstrated to bind Trxloop proteins strongly in Western blot assays (unpublished data). This strategy yielded a more sensitive response, with a total binding capacity of 100–400 response units (RU) of Trxloop protein and discrimination between strong and weak IL-8 mAb binders. However, Trxloop proteins captured by these amine-coupled mAbs rapidly desorbed within seconds to minutes, indicating that none of these anti-Trx mAbs were suitable capture ligands for Trxloop proteins. A fourth superior immobilization approach was developed involving the use of a biotinylated N-terminal peptide fused to Trxloop proteins, resulting in soluble `Biotrxloop' proteins (Smith et al., 1998). Biotrxloop proteins can be irreversibly captured on immobilized streptavidin surfaces via the biotinylated N-terminal fusion peptide. Up to 4000 RU of Biotrxloop protein can be captured per 4000 RU of immobilized streptavidin, corresponding to a 4:1 ratio of 15 kDa Biotrxloop monomer to 60 kDa streptavidin tetramer (1000 RU is equivalent to a surface concentration of 1 ng protein/mm2). Purified Biotrxloop proteins were captured on Pharmacia Biosensor SA immobilized streptavidin chips by diluting to concentrations of 1–100 μg/ml in HBS(T) buffer and injecting on to the chip at a flow rate of 20 μl/min. For measurement of mAb-Biotrxloop kinetic parameters ka and kd (mAb on-rate and off-rate) used in calculating the equilibrium binding constant Kd = kd/ka, ~100 response units (RU) of Biotrxloop proteins Z9, RS2-20 and Z9-H37A were individually captured on separate lanes of a chip, with a fourth lane left blank. The captured protein quantity of 100 RU corresponds to a Biotrxloop to streptavidin monomer molar ratio of 1:10, a condition chosen to prevent mass transfer-limited mAb adsorption and minimize multivalent capture and display of Biotrxloop proteins. A 100 μl volume of IL-8 mAb solution in HBS(T) containing 3.4 mM EDTA, 10 or 100 μM ZnSO4 buffer was sequentially injected on to all four lanes of the chip at a flow rate of 40 μl/min, at mAb bulk concentrations (Cb) of 15 nM, 30 nM, 150 nM, 300 nM, 1.5 μM and 3.0 μM. A 100 μl volume of 0.5 mg/ml Z9-H2F Trxloop protein was co-injected at the end of the adsorption step, to bind up desorbing IL-8 mAb and prevent re-adsorption on the captured Biotrxloop proteins. The Z9-H2F Trxloop variant possesses a high affinity for the IL-8 mAb and does not adsorb on the carboxymethyldextran or streptavidin on the SPR chips to any measurable extent (unpublished data). The adsorption rate constant kd was determined from a linear fit of ks versus Cb, with ks estimated from fits of 30 s of adsorption data at the various Cb values, using the standard BIAevaluation kinetic modeling software with a simple 1:1 interaction between the mAb and the Biotrxloop protein. The mAb desorption rate constant kd was determined by fitting the simple 1:1 interaction model to the initial 5 s of desorption data. Results Successful identification of zinc- and pH-sensitive switch epitopes in FLITRX library The metal-coordinating residues histidine, aspartate and glutamate were previously identified in various FLITRX loop dodecamer peptides selected for an IL-8 mAb (Lu et al., 1995). The small linear epitope length (4–6 amino acids) and variation of epitope location within the loop indicate that only a portion of the dodecamer loop peptide is needed for IL-8 mAb recognition, while other segments may remain solvent accessible. These observations suggested that a subset of FLITRX clones selected for mAb binding might also contain epitope and non-epitope metal-coordinating motifs and thereby function as metal-sensitive switch epitopes. A panel of 100 `hit' sequences that bind to an IL-8 mAb was selected by `biopanning' with a non-specific shear elution procedure. The FLITRX epitope sequences were similar to, but not identical with, those previously selected by Lu et al. (Lu et al., 1995) against the same IL-8 mAb, which recognizes the linear epitope sequence His18–Pro–Lys–Phe21 on the surface of wild-type human IL-8. The location of the epitope region in the loop peptide was highly variable, ranging from the N-terminal region to the C-terminal region of the loop sequence (unpublished data). The epitope sequence compositions for 55 of the highest affinity FLITRX clones are summarized in Table I. The first N-terminal epitope residue was usually a histidine residue, with occasional alanine and serine substitutions (Table I). The second epitope residue was highly variable in terms of size, charge and hydrophobicity, although proline and arginine were the two most common residues. The third epitope residue was nearly always a lysine, with several arginine substitutions. The fourth C-terminal epitope residue always had an aromatic side chain, with phenylalanine as the most common residue, although several histidine and tyrosine substitutions were also observed at this position. The panel of selected IL-8 mAb-binding FLITRX clones was screened for possible metal ion and pH switch epitopes by Western blot assays with the IL-8 mAb as a probe against FLITRX fusion peptides. A variety of different buffer conditions were employed during the IL-8 mAb binding step. Calcium, pH 9.0 CHES buffer and EDTA/EGTA had little effect on the observed binding of the IL-8 mAb to the panel of FLITRX variants (unpublished data). Many of the FLITRX fusion peptides exhibited lower affinity for the IL-8 mAb in pH 6.0 MES buffer, suggesting a broad influence of low pH on the mAb–epitope interaction (unpublished data). The S4, S5 and S19 FLITRX peptides exhibited zinc- and copper-sensitive mAb binding, indicative of switch epitopes (Table II). An additional zinc elution step was tried as a method for enriching the pool of shear elution IL-8 mAb-binding FLITRX mutants for zinc-sensitive variants. A Western blot screen of 12 clones from this enrichment step is shown in Figure 1. Enrichment of zinc-sensitive sequences over zinc-insensitive sequences was minor; only two out of 36 zinc elution clones, Z9 and Z27 exhibited zinc-sensitive mAb binding. The loop peptide sequences of all the zinc-sensitive FLITRX clones are listed in Table II. The more prevalent type of zinc switch epitope contained three histidine residues in the loop sequence, including one or two histidine residues in the mAb epitope region, as shown in the S4, S19, Z9 and Z27 sequences. Both of the highly zinc-sensitive S4 and Z9 peptides contained two mAb epitope histidine residues, with a histidine substituted for phenylalanine in the epitope region (HPKF → HPKH). Another type of zinc switch epitope, identified in the S5 clone (Table II), contained two tyrosine residues flanking the IL-8 mAb epitope region, in addition to one epitope histidine residue. The presence of other acidic or polar residues in the variable loop region was not sufficient to impart zinc switch epitope behavior, as observed for the zinc-insensitive S12 and Z32 sequences (Table II). The Z9 FLITRX loop peptide exhibited the highest relative mAb affinity of all the zinc-sensitive IL-8 mAb binding peptides, as determined by Western blot analysis of FLITRX protein cell lysates (unpublished data). This Z9 peptide was chosen for further characterization experiments because of its higher relative mAb affinity and strong zinc sensitivity. Properties of soluble monomeric thioredoxin loop (`Trxloop') switch epitope proteins The Western blot data clearly indicate that zinc switch epitopes can be identified in the FLITRX library using variations on normal immunoblot procedures. However, the FLITRX library uses multivalent display of thioredoxin fusion peptides on large, insoluble flagellin fibers that cannot be purified by normal protein liquid chromatography techniques and are prone to avidity effects during selection procedures. We therefore investigated the feasibility of grafting a FLITRX switch epitope peptide on to soluble, monomeric E.coli Trx protein, with the expectation that the zinc-dependent mAb binding properties would be maintained. A thioredoxin loop insert (`Trxloop') analog of the zinc switch epitope Z9 FLITRX protein was constructed by a cassette mutagenesis technique with the pALtrxA plasmid (LaVallie et al., 1993). The complete amino acid sequence of the Z9 Trxloop protein is given in Figure 2. This Z9 Trxloop protein was highly expressed in soluble form in E.coli, contributing >10% of the total soluble cellular protein, as estimated from SDS–PAGE (unpublished data), similar to other previously described Trxloop fusions (LaVallie et al., 1993). The Z9 Trxloop protein exhibited zinc-dependent IL-8 mAb binding in various Western blot assays, both as a cell lysate and as a purified wild-type protein (Figure 3), in a manner similar to that observed for the Z9 FLITRX protein. However, the monomeric Z9 Trxloop protein exhibited a much lower apparent affinity for the IL-8 mAb than the corresponding multimeric Z9 FLITRX protein, as autoradiographic detection of mAb binding with 125I-labeled protein-A required 3–6-fold longer exposure times to X-ray film (unpublished data). The presence of three histidine residues in the Z9 Trxloop peptide suggested that this protein might behave in a similar manner to a `histidine-patch' Trx mutant, described by Lu et al. (Lu et al., 1996). Chromatographic retention experiments with purified Z9 Trxloop protein on an NTA IMAC column loaded with zinc or copper showed specific binding of the protein, with 75 mM imidazole required for elution (Figure 4b). Wild-type Trx was not retained on the IMAC column (Figure 4a). Equilibrium dialysis of the Z9 Trxloop protein with 1 mM ZnCl2 or 1 mM CuCl2 in 20 mM pH 7.5 Tris buffer resulted in quantitative precipitation of the protein (unpublished data), whereas 1 mM NiCl2 only resulted in partial precipitation of the Z9 Trxloop protein. The zinc and copper protein precipitates could be quantitatively re-dissolved by dialysis with EDTA. Wild-type Trx did not precipitate after dialysis with any of the same metal ion buffer solutions. An assay with Ellman's reagent (DTNB) indicated that purified, oxidized wild-type Trx and Z9 Trxloop proteins did not have any reactive thiols, whereas freshly reduced wild-type Trx had 1.7 out of 2.0 thiols in reactive form. Site-directed mutagenesis of histidine residues in Z9 Trxloop protein The Western blot, IMAC retention and metal ion equilibrium dialysis experiments suggested that the Z9 Trxloop insertion peptide specifically coordinated one or more zinc ions through several amino acid side chains in the loop peptide region. Under normal oxidizing conditions, the two Trx/Trxloop cysteine residues probably form an intrachain disulfide bond, preventing metal binding by the Cys32 and wild-type Trx Cys35 or Trxloop Cys48 residues. No other residues with high affinity for transition metal ions, such as serine, threonine, aspartate, glutamate, tyrosine or methionine, are present in the Z9 loop peptide sequence. Thus, the three Trxloop Z9 variant loop histidine residues His37, His40 and His43 are the most likely candidates for zinc-binding ligands (Arnold, 1991); possibly in conjunction with a fourth His6 residue derived from the wild-type Trx structure. To test this prediction, each of the four Z9 Trxloop histidine residues was individually mutated to alanine or phenylalanine, residues that do not bind transition metals. The four Z9 Trxloop variants H6A, H37A, H40F and H43A were purified and tested for IL-8 mAb binding affinity and zinc switch behavior by performing Western dot-blots under non-denaturing conditions (Figure 3). The H6A variant clearly retained the zinc-dependent mAb recognition. The H37A variant had a very low mAb affinity, but no longer exhibited zinc-dependent mAb binding. The H40F variant exhibited higher apparent mAb affinity and did not exhibit zinc-dependent mAb binding. Western blot assays of the purified H43A mutant protein did not clearly indicate zinc-dependent mAb binding, although H43A crude cell lysate exhibited moderate switch epitope behavior (unpublished data). Other substitutions at this position also resulted in IL-8 mAb zinc switch epitopes (see below). These results indicated that zinc-sensitive mAb binding was conferred by residues His37 and His40 in the mAb epitope region and that residues His6 and His43 were not critical for zinc-dependent mAb binding. However, these results did not rule out the possibility that residue His43 also coordinated zinc. Increased Trxloop mAb affinity by saturation mutagenesis The multivalent display characteristics of peptides displayed in the FLITRX library cannot reliably distinguish between high intrinsic affinity and high avidity sequences, even when zinc was used specifically to elute clones during selection. The same avidity versus affinity problem in phage display has been overcome by development of monomeric phage display libraries (Lowman et al., 1991) and fusions of selected peptides to monomeric proteins, such as maltose binding protein (Choi et al., 1997). We suspected that selective saturation mutagenesis (Gayle et al., 1993) of a Trxloop peptide might further improve the IL-8 mAb affinity while retaining the switch epitope property. Thus, we performed saturation mutagenesis on nine residues in the Z9 Trxloop variable loop region to validate this hypothesis, which has the sequence G33PQVHPKHFGHAPIGP48. The design strategy was to allow mutations in the underlined non-epitope loop linker regions and the highly variable Pro38 epitope residue in the Z9 loop region sequence; the two mAb epitope His37 and His40 residues critical for zinc switch epitope properties and the epitope Lys39 residue were retained to preserve IL-8 mAb recognition. A library of Z9 Trxloop variants was constructed and screened for high IL-8 mAb affinity using Western blots of Trxloop E.coli colony lysates. Initial screening of this library yielded numerous high IL-8 mAb affinity variants, many of which had multiple loop peptide inserts as indicated by DNA sequencing and SDS-PAGE of expressed Trxloop proteins (unpublished data). Therefore, the panel of high-affinity Z9 Trxloop variants was further screened by SDS-PAGE of the expressed Trxloop proteins to verify molecular masses indicative of single loop inserts. This process yielded 17 Trxloop variants containing 1–5 mutations with 2–10 times higher apparent IL-8 mAb affinity than the original Z9 Trxloop protein and two variants with mAb affinities equivalent to the Z9 Trxloop protein, as summarized in Table III. The RS2-20 Trxloop variant had the highest apparent IL-8 mAb affinity of all the single epitope variants identified in the saturation mutagenesis screening procedure. These variants all exhibited zinc switch epitope behavior in Western blot assays (Figure 5). The S1-7, S1-19, S1-31, S1-50, P1-25, P2-12, P3-17 and P3-36 variants all contained His43 substitutions (Table III), yet exhibited zinc switch epitope behavior (Figure 5, unpublished data), confirming the non-essential role of His 43 in switch epitope function. Four F41C loop substitution variants were identified: S1-7, P1-25, P1-26 and P3-36. The P1-26 variant also had a shortened loop sequence. Non-reducing SDS-PAGE indicated that the purified P3-36 Trxloop variant exists as a mixture of several disulfide-bonded dimers and monomer that is reduced to a single monomer species after reduction with BME (unpublished data). These results suggest that the novel loop Cys41 residues readily formed intrachain disulfide bonds with each other or either of the wild-type Trx Cys32 and Cys49 residues. Measurement of Biotrxloop IL-8 mAb affinities by surface plasmon resonance A Western blot assay (Figure 5) suggested a 10-fold improvement in IL-8 mAb affinity for the RS2-20 Trxloop variant relative to the Z9 variant. We wanted to verify this increase in affinity by measuring accurate equilibrium binding constants for the IL-8 mAb–Trxloop interaction and examine the kinetics and reversibility of mAb–Trxloop binding as a function of zinc concentration with a more sensitive SPR technique. Biotin-labeled Biotrxloop variants with loop sequences corresponding to the original Z9, high mAb affinity RS2-20 and low mAb affinity Z9-H37A Trxloop variants were prepared by cassette mutagenesis, purified and analyzed with respect to IL-8 mAb binding and zinc concentration by SPR measurements. The IL-8 mAb specifically bound to the immobilized Z9 Biotrxloop protein in the absence of zinc (Figure 6A) and did not bind to the unmodified carboxymethyldextran chip matrix (unpublished data). Some of the bound IL-8 mAb molecules rapidly desorbed in the absence of zinc, whereas a significant fraction of the bound IL-8 mAb molecules was very slow to desorb under these conditions, suggestive of two populations of low- and high-affinity mAb binding sites. The majority of the remaining tightly bound IL-8 mAb was rapidly eluted by addition of 100 μM ZnSO4, although a minor accumulation of irreversibly bound protein (~20 RU) was apparent after each mAb adsorption/zinc elution/EDTA wash cycle. Further SPR experiments indicated that 100 μM ZnSO4 almost completely inhibited IL-8 mAb binding to the immobilized Z9 Biotrxloop protein (Figure 6B). The presence of 10 μM ZnSO4 decreased the amount of bound mAb by 25–33% relative to zinc-free buffer and 1 μM ZnSO4 had little effect on IL-8 mAb binding (Figure 6B). Similar results were observed for the RS2-20 Biotrxloop protein (unpublished data). A relatively small amount of IL-8 mAb bound to the Z9-H37A Biotrxloop variant, compared with the Z9 and RS2-20 variants, in agreement with previous Western blot studies of the corresponding Trxloop variants (Figure 3). These results indicate that the IL-8 mAb rapidly binds to immobilized Z9 and RS2-20 Biotrxloop proteins in an epitope-specific manner, that a significant fraction of mAb molecules remain tightly bound and slowly desorb in the absence of zinc and that the majority of bound IL-8 mAb molecules can be rapidly eluted in a largely reversible manner by addition of 100 μM zinc. A more quantitative SPR kinetic analysis of IL-8 mAb–Biotrxloop interactions was performed using experimental conditions designed to minimize mass transfer and avidity effects. Adsorption rates, desorption rates and equilibrium Kd values for mAb binding to Z9 and RS2-20 Biotrxloop proteins in the absence of zinc, under conditions of monovalent display (100 RU Biotrxloop:4000 RU streptavidin) are listed in Table IV. An ~2-fold decrease in IL-8 mAb Kd value was measured for the RS2-20 mutant versus the original Z9 variant, compared with the 10-fold higher apparent affinity observed in Western blots (Figure 5). The IL-8 mAb exhibited a slower adsorption rate (`on-rate') and slower desorption rate (`off-rate') with the RS2-20 Biotrxloop protein than the Z9 Biotrxloop protein under these conditions (Figure 6C). Approximately 600 RU of mAb was bound to the Z9 Biotrxloop protein at a time of 2.5 min, wheras only 440 RU of mAb was bound to the RS2-20 protein within the same interval. Thus, more IL-8 mAb bound to the Z9 peptide than the RS2-20 peptide after several minutes, but it also desorbed more quickly. An additional gain in RS2-20 IL-8 mAb affinity relative to the Z9 Biotrxloop variant was apparent under conditions of multivalent Biotrxloop protein display on streptavidin using a ratio of 2000 RU Biotrxloop:4000 RU streptavidin (Figure 6D). The IL-8 mAb exhibited a faster adsorption rate to the RS2-20 Biotrxloop variant than the Z9 Biotrxloop variant after 1 min, resulting in binding of 50% more mAb (800 more RU) to the RS2-20 protein than the Z9 protein at a time of 2.5 min. This mAb kinetic binding behavior was opposite to that observed under conditions of monovalent Biotrxloop display, where 160 RU more IL-8 mAb was bound to the Z9 Biotrxloop protein than the RS2-20 variant at a time of 2.5 min (Figure 6C). Discussion FLITRX library contains zinc- and pH-sensitive IL-8 mAb switch epitopes The variation in the FLITRX loop epitope sequence, location of the linear epitope sequence and non-epitope sequence composition indicated that many different amino acid sequences are recognized by the same IL-8 mAb. We successfully exploited this allowed epitope diversity to identify a small number of FLITRX zinc switch epitopes, which occurred at low frequency (~1–5%) in the pool of selected IL-8 mAb binders. The failure of the FLITRX zinc-elution selection procedure to significantly enrich the library for zinc switch epitopes may have resulted from the extremely high avidity of the interaction between `20 000-mer' FLITRX fibers on E.coli and the monolayer of IL-8 mAb adsorbed on the polystyrene plate. Specific zinc-induced elution of E.coli with high-affinity FLITRX flagella is probably a very slow kinetic dissociation event. The tendency of zinc switch epitopes to have three loop histidine residues was not surprising, as histidine is one of the dominant residues in protein interactions with transition metal ions such as Cu2+ and Zn2+ (Yip et al., 1989; Arnold, 1991; Porath, 1992). There are numerous examples of proteins with structure and binding properties modulated by the coordination of divalent Zn2+ ions to either three or four histidine residues (Berg and Shi, 1996). However, multiple histidines are not strictly necessary for zinc switch behavior, as demonstrated by the tyrosine-rich S5 epitope sequence (YHGKFY). The choice of IL-8 mAb with its corresponding conserved histidine epitope residue may have resulted in a higher than typical probability for zinc switch epitopes and a lower probability of calcium-sensitive switch epitopes. Ionization of this conserved epitope histidine may also account for the general decrease in IL-8 mAb affinity observed at pH 6.0 versus pH 7.5 for many of the IL-8 mAb FLITRX clones. The FLITRX flagellin protein export pathway appears to discriminate against single cysteine residues in the variable loop region, as none were identified in any of the selected FLITRX library clones. In contrast, four single cysteine substitutions were identified in Trxloop proteins. The presence of reduced, unpaired reactive thiols probably precludes proper export and assembly of FLITRX proteins into functional flagella fibers. Rapid disulfide bond formation between the two adjacent, wild-type Trx cysteine residues probably accounts for their tolerance in the FLITRX protein folding and assembly pathway. Trxloop variant retains zinc-dependent mAb recognition of FLITRX protein The Z9 loop peptide exhibited the same IL-8 mAb recognition and zinc switch epitope properties whether it was displayed as a FLITRX, Trxloop or Biotrxloop fusion protein. This demonstrated that the core Trx–peptide fusion maintained function, conformation and solvent accessibility independent of any N- and C-terminal fusions to the Trx scaffold protein. Trxloop and Biotrxloop fusion proteins were found to be superior to FLITRX proteins for precise affinity characterization and affinity maturation of loop peptides, because of their high expression, solubility, monomeric association state and ease of purification. However, Trxloop proteins frequently exhibit lower apparent affinities for target molecules than their corresponding parent FLITRX proteins, as observed in this and previous work (Lu et al., 1995). The lowered effective IL-8 mAb affinity probably results from decreased avidity of loop peptides displayed by monomeric Trxloop molecules versus highly multivalent FLITRX flagellin fibers. Dimeric Trxloop fusions partially overcome this problem (Lu et al., 1995), consistent with this explanation. Zinc switch epitope and metal binding properties of Z9 Trxloop variant The Z9 Trxloop variant has a much higher affinity for zinc than wild-type Trx, as indicated by the strong zinc IMAC retention and quantitative zinc-induced precipitation observed for the Z9 Trxloop variant versus the lack of zinc-induced precipitation and IMAC retention observed for wild-type Trx. Strong retention on a zinc IMAC column was previously observed for a `histidine-patch' Trx variant (Lu et al., 1996); the wild-type Trx His6 residue failed to bind the zinc IMAC column in these studies. Zinc-induced precipitation has also been observed for a histidine-tagged fusion protein (Lilius et al., 1991). These results indicate that the high zinc affinity of the Z9 Trxloop variant relative to wild-type Trx primarily results from loop histidine residues and not wild-type Trx His6, Cys32 or Cys35/Cys49. All three Z9 loop histidine residues (His37, His40 and His43) probably bind zinc ions, although only His37 and His40 are necessary for IL-8 mAb zinc switch epitope behavior. The higher apparent affinity of the Z9 Trxloop protein for Zn2+ and Cu2+ ions and lower apparent affinity for Ni2+ ions are consistent with those predicted by the Irving–Williams series (Irving and Williams, 1953). The higher affinity for zinc versus copper may result from the lack of ligand field stabilization energy exhibited by the Zn2+ ion (Berg and Shi, 1996) compared with other transition metals. The Z9 FLITRX/Trxloop His–Pro–X–His metal binding motif is different from a previously described His–X–X–X–His α-helical metal binding motif (Arnold, 1991). The Z9 loop peptide conformation may partially mimic the extended loop/3–10 helix structure of the wild-type IL-8 protein epitope (Clore et al., 1990). Saturation mutagenesis of Z9 Trxloop variant and avidity effects on mAb binding The Z9 Trxloop IL-8 mAb affinity was successfully increased by saturation mutagenesis of variable epitope and non-epitope regions. Common trends in the loop sequences of the nineteen increased mAb affinity Z9 Trxloop variants included a conserved Pro38 epitope residue, substitution of more flexible residues for other loop proline residues, conserved or improved hydrophobic interactions and increased positive charge resulting from cationic substitutions. Acidic substitutions were rarely observed at any of the variable loop residue positions. The highest IL-8 mAb affinity RS2-20 variant had F41W, P45L and I46F substitutions, two of which involved aromatic substitutions suggestive of increased hydrophobic interactions and one which increased loop flexibility. The second highest IL-8 mAb affinity S1-19 variant had G42D, H43R and P45L substitutions that may enhance ionic interactions and increase loop flexibility. Thus, more than one route to improved affinity is possible. The Western blot screening process for high IL-8 mAb affinity Z9 Trxloop variants in this work was partially confounded by several avidity effects. First, the bivalent IL-8 mAb bound more tightly to multiple loop epitope peptides than single epitope peptides; use of monovalent IL-8 mAb Fab fragments would minimize this avidity effect during screening. Avidity effects in the form of enhanced IL-8 mAb affinity have previously been observed with covalent Trxloop dimer fusion proteins compared with monomeric Trxloop proteins (Lu et al., 1995). Second, the Z9 Trxloop plasmid library construction method did not prevent occasional creation of larger loop peptides containing multiple epitopes; alternative plasmid library construction approaches would avoid this problem. Third, overexpressed Trxloop proteins may be adsorbed on nitrocellulose or PVDF immunoblot membranes at near-monolayer surface concentrations. This could cause avidity effects during Western blot screening resulting from single mAb molecules binding to adjacent membrane-bound Trxloop protein molecules. Monovalent phage display selection of Trxloop proteins might be more efficient at identifying high mAb-affinity variants than the Western blot screening approach (Lowman et al., 1991; Lowman and Wells, 1993). Finally, unlike the FLITRX display system, Trxloop proteins overexpressed in E.coli can tolerate single cysteine substitutions in the loop region, leading to the formation of disulfide-bond linked dimeric Trxloop proteins. This problem could be avoided by excluding possible cysteine codons in the synthetic semi-random oligonucleotides used to construct the plasmid library variable loop region. SPR results indicate higher avidity, slower off-rate for RS2-20 vs Z9 Biotrxloop proteins The SPR experiments demonstrated that the Trxloop IL-8 mAb affinity could be increased without loss of the zinc switch epitope function. The increased IL-8 mAb affinity (smaller Kd value) of the RS2-20 variant was entirely due to a 2-fold slower off-rate, as the on-rate was actually slightly slower than that of the parent Z9 variant under monovalent display conditions. Avidity effects may explain the discrepancy between the 10-fold increase in apparent IL-8 mAb affinity observed in immunoblot assays and the 2-fold decrease in Kd value. SPR experiments with Biotrxloop proteins also demonstrated that the zinc switch effect on IL-8 mAb binding was rapid and largely reversible following removal of bound zinc ions with EDTA, consistent with prior immunoblot results. The slightly irreversible adsorption of IL-8 mAb after zinc elution and EDTA washing steps may have been due to adsorption and physical entrapment of denatured IL-8 mAb aggregates in the carboxymethyldextran matrix. The observed biphasic IL-8 mAb desorption behavior may have resulted from the presence of both spatially isolated, monomeric and spatially adjacent, multimeric Biotrxloop epitopes immobilized on streptavidin within the carboxymethyldextran matrix. Larger amounts of immobilized Biotrxloop protein (500–2000 RU), as used in the initial qualitative assays result in partially multivalent display of IL-8 mAb epitopes. Multivalent display of Biotrxloop proteins in mAb-affinity SPR assays may more closely approximate the Trxloop monolayer surface display conditions that probably occur in immunoblot assays. The SPR studies suggested that the Z9 and RS2-20 Biotrxloop proteins have an approximate zinc equilibrium dissociation constant (Kd value) in the range 10–100 μM. However, the actual Zn2+ Kd value could be much smaller (0.1–10 μM), as 10 μM Zn2+ was sufficient to inhibit completely IL-8 mAb binding to the Z9 Trxloop protein in Western blot assays. The effective solution concentration of zinc in SPR assays may have been decreased owing to competitive binding of Zn2+ ions by the anionic carboxymethyldextran matrix. Conclusion We successfully demonstrated that the FLITRX random peptide library can be used to identify peptide loops on thioredoxin with affinity properties modulated by transition metal ions and pH. Thus, random peptide libraries, such as the FLITRX library, can be screened to identify peptide loop `modules' with interesting physico-chemical properties that are potentially useful as `molecular switch' control elements. These loop modules can then be grafted on to soluble wild-type Trx and Biotrx proteins to facilitate further characterization. The resultant Trxloop proteins also have potential as fusion partners with other proteins, e.g. as subunits in a novel fusion protein scaffold used to detect loop peptide interactions with other molecules in a yeast two-hybrid interaction trap (Colas et al., 1996). One theoretical implication of this work is that novel therapeutic enzymes and signal transduction proteins with superior pharmacological properties could be designed by insertion of surface loop peptides that possess suitable switch epitope properties. Table I. Summary of FLITRX loop region amino acid epitope sequence composition for anti-interleukin 8 monoclonal antibody obtained by shear elutiona Interleukin-8 epitopeb  Observed FLITRX peptide IL-8 mAb epitope residues (frequency of occurrence)  a55 FLITRX clones selected for binding to IL-8 monoclonal antibody using a shear elution procedure loop. The loop region peptide sequences were identified by sequencing DNA of corresponding plasmids, as described in Materials and methods.  bWild-type IL-8 epitope sequence: S14KPFHPKFIKEL25.  His18  His (49/55), Ala (4/55), Ser (2/55)  Pro19  Pro (21/55), Arg (13/55), {Ala, Glu, Gly, His, Ile, Lys, Leu, Asn, Gln, Ser, Val} (21/55)  Lys20  Lys (52/55), Arg (3/55)  Phe21  Phe (50/55), His (4/55), Tyr (1/55)  Interleukin-8 epitopeb  Observed FLITRX peptide IL-8 mAb epitope residues (frequency of occurrence)  a55 FLITRX clones selected for binding to IL-8 monoclonal antibody using a shear elution procedure loop. The loop region peptide sequences were identified by sequencing DNA of corresponding plasmids, as described in Materials and methods.  bWild-type IL-8 epitope sequence: S14KPFHPKFIKEL25.  His18  His (49/55), Ala (4/55), Ser (2/55)  Pro19  Pro (21/55), Arg (13/55), {Ala, Glu, Gly, His, Ile, Lys, Leu, Asn, Gln, Ser, Val} (21/55)  Lys20  Lys (52/55), Arg (3/55)  Phe21  Phe (50/55), His (4/55), Tyr (1/55)  View Large Table II. Selected FLITRX loop region amino acid sequences with affinities for anti-IL-8 mAb obtained by shear and zinc elution, showing histidines and variation of linear antibody epitope location (underlined) FLITRX clonea  Variable loop region amino acid sequenced  IL-8 mAb FLITRX binding sensitive to zinc?  aFLITRX clones were identified by shear elution or a combination of shear and zinc elution as described in Materials and methods. The apparent relative IL-8 mAb affinities of the FLITRX clones were determined by immunoblotting the immobilized FLITRX proteins in the presence of 2.5 mM EDTA or 100 μM ZnCl2 as described in Materials and methods.  bFLITRX clones denoted by S prefix were selected by a non-specific shear elution procedure.  cFLITRX clones denoted by Z prefix were selected by performing an additional zinc elution procedure after three rounds of shear elution.  dThe putative IL-8 mAb epitope regions in each FLITRX loop sequence are underlined.  S4b  RQSGAHGHPKHN  Yes  S5  GKPAGYHGKFYG  Yes  S12  SSARGTLHRKFG  No  S19  PHKKFPKPHHER  Yes  Z9c  QVHPKHFGHAPI  Yes  Z27  HDGRAVHSKFGH  Moderate  Z32  DRDTRQEHRKFG  No  FLITRX clonea  Variable loop region amino acid sequenced  IL-8 mAb FLITRX binding sensitive to zinc?  aFLITRX clones were identified by shear elution or a combination of shear and zinc elution as described in Materials and methods. The apparent relative IL-8 mAb affinities of the FLITRX clones were determined by immunoblotting the immobilized FLITRX proteins in the presence of 2.5 mM EDTA or 100 μM ZnCl2 as described in Materials and methods.  bFLITRX clones denoted by S prefix were selected by a non-specific shear elution procedure.  cFLITRX clones denoted by Z prefix were selected by performing an additional zinc elution procedure after three rounds of shear elution.  dThe putative IL-8 mAb epitope regions in each FLITRX loop sequence are underlined.  S4b  RQSGAHGHPKHN  Yes  S5  GKPAGYHGKFYG  Yes  S12  SSARGTLHRKFG  No  S19  PHKKFPKPHHER  Yes  Z9c  QVHPKHFGHAPI  Yes  Z27  HDGRAVHSKFGH  Moderate  Z32  DRDTRQEHRKFG  No  View Large Table III. Z9 Trxloop saturation mutagenesis variant loop region amino acid sequences and apparent IL-8 mAb affinities Trxloop variant  Loop region amino acid sequencec  Relative affinity for IL-8 mAbd  Ratio of relative IL-8 mAb affinity to original Z9 clonee  aOriginal (parent) Z9 Trxloop variant loop peptide sequence identified during screen of IL-8 mAb binding FLITRX clones for zinc switch epitopes. The complete Z9 Trxloop sequence is given in Figure 2.  bThis Trxloop variant and all following variants were obtained by screening a saturation mutagenesis library of the original Z9 loop sequence, Q35VHPKHFGHAPI46, in which the underlined residues were randomly varied by using degenerate oligonucleotides in a cassette mutagenesis procedure, as described in Materials and methods. Trxloop variants with high IL-8 mAb affinity were identified by an immunoblot procedure and screened for single loop inserts by SDS–PAGE as described in Materials and methods.  cThe amino acid numbering for the first residue of each peptide sequence is number 35. Amino acid sequences of Trxloop variable loop regions were obtained by sequencing DNA of corresponding plasmids isolated from E.coli clones, as described in Materials and methods.  dRelative intensity of Trxloop variant IL-8 mAb binding in Western immunoblot assay, corrected for background intensity, as described in Materials and methods.  eRatio of relative IL-8 mAb binding intensity for each Trxloop variant to the parent Z9 Trxloop variant.  Z9a (original)  QVHPKHFGHAPI  7.8  (1.0)  S1-7b  QVHPKHCGRARI  59.1  7.6  S1-19  QVHPKHFDRALI  79.7  10.3  S1-31  QVHPKHFGRASF  18.3  2.4  S1-49  RVHPKHFGHASI  19.7  2.5  S1-50  HVHPKHFGRAPF  7.6  1.0  P1-25  PVHPKHCGYVQG  35.0  4.5  P1-26  QVHPKHCGH  27.1  3.5  P2-12  QVHPKHFGRALI  22.9  2.9  P2-48  RVHPKHFAHAHI  25.0  3.2  P2-50  QVHPKHFGHALF  16.6  2.1  P3-3  QVHPKHFGHAII  7.8  1.0  P3-12  QVHPKHFGHVAI  17.8  2.3  P3-14  QVHPKHFRHAPI  28.1  3.6  P3-17  QVHPKHFGTARI  16.8  2.2  P3-36  QVHPKHCGDSRV  54.9  7.1  P3-71  RVHPKHFRHAPI  22.7  2.9  P3-76  QVHPKHFVHATI  13.1  1.7  RS2-9  RAHPKHFGHRPF  14.9  1.9  RS2-20  QVHPKHWGHALF  86.0  11.1  Trxloop variant  Loop region amino acid sequencec  Relative affinity for IL-8 mAbd  Ratio of relative IL-8 mAb affinity to original Z9 clonee  aOriginal (parent) Z9 Trxloop variant loop peptide sequence identified during screen of IL-8 mAb binding FLITRX clones for zinc switch epitopes. The complete Z9 Trxloop sequence is given in Figure 2.  bThis Trxloop variant and all following variants were obtained by screening a saturation mutagenesis library of the original Z9 loop sequence, Q35VHPKHFGHAPI46, in which the underlined residues were randomly varied by using degenerate oligonucleotides in a cassette mutagenesis procedure, as described in Materials and methods. Trxloop variants with high IL-8 mAb affinity were identified by an immunoblot procedure and screened for single loop inserts by SDS–PAGE as described in Materials and methods.  cThe amino acid numbering for the first residue of each peptide sequence is number 35. Amino acid sequences of Trxloop variable loop regions were obtained by sequencing DNA of corresponding plasmids isolated from E.coli clones, as described in Materials and methods.  dRelative intensity of Trxloop variant IL-8 mAb binding in Western immunoblot assay, corrected for background intensity, as described in Materials and methods.  eRatio of relative IL-8 mAb binding intensity for each Trxloop variant to the parent Z9 Trxloop variant.  Z9a (original)  QVHPKHFGHAPI  7.8  (1.0)  S1-7b  QVHPKHCGRARI  59.1  7.6  S1-19  QVHPKHFDRALI  79.7  10.3  S1-31  QVHPKHFGRASF  18.3  2.4  S1-49  RVHPKHFGHASI  19.7  2.5  S1-50  HVHPKHFGRAPF  7.6  1.0  P1-25  PVHPKHCGYVQG  35.0  4.5  P1-26  QVHPKHCGH  27.1  3.5  P2-12  QVHPKHFGRALI  22.9  2.9  P2-48  RVHPKHFAHAHI  25.0  3.2  P2-50  QVHPKHFGHALF  16.6  2.1  P3-3  QVHPKHFGHAII  7.8  1.0  P3-12  QVHPKHFGHVAI  17.8  2.3  P3-14  QVHPKHFRHAPI  28.1  3.6  P3-17  QVHPKHFGTARI  16.8  2.2  P3-36  QVHPKHCGDSRV  54.9  7.1  P3-71  RVHPKHFRHAPI  22.7  2.9  P3-76  QVHPKHFVHATI  13.1  1.7  RS2-9  RAHPKHFGHRPF  14.9  1.9  RS2-20  QVHPKHWGHALF  86.0  11.1  View Large Table IV. IL-8 mAb-Biotrxloop kinetic and equilibrium binding constants for original Z9 and improved affinity RS2-20 variantsa Biotrxloop variant  kab (s–1 M–1)  kdc (s–1)  Kdd (M)  aSurface plasmon resonance measurements were performed using metal-free HBS(T) buffer which contains 3.4 mM EDTA, using buffer flow-rates and monomeric display immobilization conditions described in the Materials and methods section.  bDetermined from linear fit of Cb versus ks for Cb = 150 nM, 300 nM, 1.5 μM, 3.0 μM mAb concentration, where ks = –(kaCb + kd), fitted to standard 1:1 interaction model using manufacturer's software.  cAverage of kd values from Cb = 150 nM, 300 nM, 1.5 μM and 3.0 μM mAb desorption data fitted to standard 1:1 interaction model using manufacturer's software.  dEquilibrium dissociation constant calculated from kd/ka ratio.  Z9  3.4×103  0.54  162×10–6  RS2-20  2.7×103  0.17  63×10–6  Biotrxloop variant  kab (s–1 M–1)  kdc (s–1)  Kdd (M)  aSurface plasmon resonance measurements were performed using metal-free HBS(T) buffer which contains 3.4 mM EDTA, using buffer flow-rates and monomeric display immobilization conditions described in the Materials and methods section.  bDetermined from linear fit of Cb versus ks for Cb = 150 nM, 300 nM, 1.5 μM, 3.0 μM mAb concentration, where ks = –(kaCb + kd), fitted to standard 1:1 interaction model using manufacturer's software.  cAverage of kd values from Cb = 150 nM, 300 nM, 1.5 μM and 3.0 μM mAb desorption data fitted to standard 1:1 interaction model using manufacturer's software.  dEquilibrium dissociation constant calculated from kd/ka ratio.  Z9  3.4×103  0.54  162×10–6  RS2-20  2.7×103  0.17  63×10–6  View Large Fig. 1. View largeDownload slide Western immunoblot screen of selected FLITRX clones for differential IL-8 mAb binding in the presence (A) and absence (B) of zinc. (A) shows significant IL-8 mAb binding to the circled Z9 FLITRX clone and other clones in zinc-free 1 mM EDTA buffer. (B) shows a failure of the IL-8 mAb to bind the circled Z9 FLITRX clone in the presence of 100 μM zinc buffer, while the other FLITRX clones still exhibit significant IL-8 mAb binding, indicating that the Z9 clone behaves as a zinc switch epitope. FLITRX E.coli cell lysates were immobilized on nitrocellulose membranes and probed with the IL-8 mAb in the absence and presence of 100 μM zinc in 50 mM Tris, pH 7.5, 150 mM NaCl, 1% α-methylmannoside buffer solution, as described in Materials and methods. Fig. 1. View largeDownload slide Western immunoblot screen of selected FLITRX clones for differential IL-8 mAb binding in the presence (A) and absence (B) of zinc. (A) shows significant IL-8 mAb binding to the circled Z9 FLITRX clone and other clones in zinc-free 1 mM EDTA buffer. (B) shows a failure of the IL-8 mAb to bind the circled Z9 FLITRX clone in the presence of 100 μM zinc buffer, while the other FLITRX clones still exhibit significant IL-8 mAb binding, indicating that the Z9 clone behaves as a zinc switch epitope. FLITRX E.coli cell lysates were immobilized on nitrocellulose membranes and probed with the IL-8 mAb in the absence and presence of 100 μM zinc in 50 mM Tris, pH 7.5, 150 mM NaCl, 1% α-methylmannoside buffer solution, as described in Materials and methods. Fig. 2. View largeDownload slide Complete amino acid sequence of Z9 Trxloop protein. The italicized region denotes the loop insert peptide that is not found in wild-type E.coli thioredoxin. The four residues in bold denote potential zinc-coordinating histidine residues which were altered by site-directed mutagenesis to alanine or phenylalanine. Fig. 2. View largeDownload slide Complete amino acid sequence of Z9 Trxloop protein. The italicized region denotes the loop insert peptide that is not found in wild-type E.coli thioredoxin. The four residues in bold denote potential zinc-coordinating histidine residues which were altered by site-directed mutagenesis to alanine or phenylalanine. Fig. 3. View largeDownload slide Western dot-blot of IL-8 mAb binding to wild-type IL-8 protein and Z9 Trxloop histidine substitution variants in the absence (A) of zinc (1 mM EDTA) and presence (B) of zinc (10 μM ZnCl2). 1, IL-8-Trx fusion protein control; 2, H6A Z9 Trxloop variant; 3, parent Z9 Trxloop protein; 4, H37A Z9 Trxloop variant; 5, H40F Z9 Trxloop variant; 6, H43A Z9 Trxloop variant. Residues His37 and His40 in the mAb epitope region are important for the zinc switch epitope property, whereas the non-loop His6 residue is clearly not necessary for zinc switch epitope behavior. Fig. 3. View largeDownload slide Western dot-blot of IL-8 mAb binding to wild-type IL-8 protein and Z9 Trxloop histidine substitution variants in the absence (A) of zinc (1 mM EDTA) and presence (B) of zinc (10 μM ZnCl2). 1, IL-8-Trx fusion protein control; 2, H6A Z9 Trxloop variant; 3, parent Z9 Trxloop protein; 4, H37A Z9 Trxloop variant; 5, H40F Z9 Trxloop variant; 6, H43A Z9 Trxloop variant. Residues His37 and His40 in the mAb epitope region are important for the zinc switch epitope property, whereas the non-loop His6 residue is clearly not necessary for zinc switch epitope behavior. Fig. 4. View largeDownload slide Chromatogram of wild-type Trx (A) and Z9 Trxloop variant (B) retention on zinc–NTA IMAC column. The Z9 Trxloop protein bound to the zinc matrix of the column and required 75 mM imidazole for elution. Wild-type thioredoxin was not retained on the column. Experimental conditions are described in Materials and methods. Fig. 4. View largeDownload slide Chromatogram of wild-type Trx (A) and Z9 Trxloop variant (B) retention on zinc–NTA IMAC column. The Z9 Trxloop protein bound to the zinc matrix of the column and required 75 mM imidazole for elution. Wild-type thioredoxin was not retained on the column. Experimental conditions are described in Materials and methods. Fig. 5. View largeDownload slide IL-8 mAb Western blots of example high-affinity Z9 Trxloop variants identified by screening a saturation mutagenesis library of the Z9 Trxloop sequence. The loop insert region of the Z9 Trxloop sequence was randomly mutated at low frequency at the underlined regions of the loop sequence: Q35VHPKHFGHAPI46. (A) IL-8 mAb immunoblot performed with 1 mM EDTA, 50 mM MOPS, pH 7.5, 150 mM NaCl buffer, conditions as described in Materials and methods. The two S1-19 and RS2-20 Trxloop variants exhibited approximately 10-fold higher apparent IL-8 mAb affinities than the parent Z9 Trxloop protein. (B) IL-8 mAb immunoblot performed with 100 μM ZnCl2, 50 mM MOPS, pH 7.5, 150 mM NaCl buffer, conditions as described in Materials and methods. No IL-8 mAb binding is observed for any of the variants in the presence of 100 μM ZnCl2. Fig. 5. View largeDownload slide IL-8 mAb Western blots of example high-affinity Z9 Trxloop variants identified by screening a saturation mutagenesis library of the Z9 Trxloop sequence. The loop insert region of the Z9 Trxloop sequence was randomly mutated at low frequency at the underlined regions of the loop sequence: Q35VHPKHFGHAPI46. (A) IL-8 mAb immunoblot performed with 1 mM EDTA, 50 mM MOPS, pH 7.5, 150 mM NaCl buffer, conditions as described in Materials and methods. The two S1-19 and RS2-20 Trxloop variants exhibited approximately 10-fold higher apparent IL-8 mAb affinities than the parent Z9 Trxloop protein. (B) IL-8 mAb immunoblot performed with 100 μM ZnCl2, 50 mM MOPS, pH 7.5, 150 mM NaCl buffer, conditions as described in Materials and methods. No IL-8 mAb binding is observed for any of the variants in the presence of 100 μM ZnCl2. Fig. 6. View large Download slide View large Download slide Surface plasmon resonance measurement of Biotrxloop mAb binding kinetics and affinities. (A) `Zinc-switch' effect of IL-8 mAb affinity for Z9 Biotrxloop protein. The left arrow `a' indicates injection of IL-8 mAb and subsequent binding of the mAb to the immobilized Z9 Biotrxloop using HBS(T) buffer containing 3.4 mM EDTA. The middle arrow `b' shows injection of IL-8 mAb in HBS(T) buffer containing 100 μM ZnSO4, with no binding of the IL-8 mAb to the Z9 Biotrxloop protein. The right arrow `c' shows another sequential injection of IL-8 mAb in HBS(T) buffer containing 3.4 mM EDTA, again followed by IL-8 mAb binding to the Z9 Biotrxloop protein. (B) Effect of zinc concentration on IL-8 mAb binding to Z9 Biotrxloop protein as monitored by SPR. The left arrow `a' indicates start of IL-8 mAb injection, the right arrow `b' indicates the end of the IL-8 mAb injection. IL-8 mAb binding to captured Z9 Biotrxloop protein was highest in HBS(T) buffer with 3.4 mM EDTA (▪), with 30% inhibition of mAb binding in HBS(T) buffer with 10 μM ZnSO4 (○) and nearly complete inhibition of mAb binding in HBS(T) buffer with 100 μM ZnSO4 (▵). (C) SPR comparison of IL-8 mAb binding to captured Z9 (•) and RS2-20 (□) Biotrxloop proteins under conditions of monovalent Biotrxloop display (100 RU captured Biotrxloop: 4000 RU immobilized streptavidin). The left arrow `a' indicates start of IL-8 mAb injection, the right arrow `b' indicates the end of the IL-8 mAb injection. (D) SPR comparison of IL-8 mAb binding to captured Z9 (•) and RS2-20 (□) Biotrxloop proteins under conditions of multivalent Biotrxloop display (2000 RU captured Biotrxloop: 4000 RU immobilized streptavidin). The left arrow `a' indicates start of IL-8 mAb injection and the right arrow `b' indicates the end of the IL-8 mAb injection. Fig. 6. View large Download slide View large Download slide Surface plasmon resonance measurement of Biotrxloop mAb binding kinetics and affinities. (A) `Zinc-switch' effect of IL-8 mAb affinity for Z9 Biotrxloop protein. The left arrow `a' indicates injection of IL-8 mAb and subsequent binding of the mAb to the immobilized Z9 Biotrxloop using HBS(T) buffer containing 3.4 mM EDTA. The middle arrow `b' shows injection of IL-8 mAb in HBS(T) buffer containing 100 μM ZnSO4, with no binding of the IL-8 mAb to the Z9 Biotrxloop protein. The right arrow `c' shows another sequential injection of IL-8 mAb in HBS(T) buffer containing 3.4 mM EDTA, again followed by IL-8 mAb binding to the Z9 Biotrxloop protein. (B) Effect of zinc concentration on IL-8 mAb binding to Z9 Biotrxloop protein as monitored by SPR. The left arrow `a' indicates start of IL-8 mAb injection, the right arrow `b' indicates the end of the IL-8 mAb injection. IL-8 mAb binding to captured Z9 Biotrxloop protein was highest in HBS(T) buffer with 3.4 mM EDTA (▪), with 30% inhibition of mAb binding in HBS(T) buffer with 10 μM ZnSO4 (○) and nearly complete inhibition of mAb binding in HBS(T) buffer with 100 μM ZnSO4 (▵). (C) SPR comparison of IL-8 mAb binding to captured Z9 (•) and RS2-20 (□) Biotrxloop proteins under conditions of monovalent Biotrxloop display (100 RU captured Biotrxloop: 4000 RU immobilized streptavidin). The left arrow `a' indicates start of IL-8 mAb injection, the right arrow `b' indicates the end of the IL-8 mAb injection. (D) SPR comparison of IL-8 mAb binding to captured Z9 (•) and RS2-20 (□) Biotrxloop proteins under conditions of multivalent Biotrxloop display (2000 RU captured Biotrxloop: 4000 RU immobilized streptavidin). The left arrow `a' indicates start of IL-8 mAb injection and the right arrow `b' indicates the end of the IL-8 mAb injection. 1 To whom correspondence should be addressed. Present address: Biochemistry and Molecular Biology Department, The Pennsylvania State University, University Park, PA 16802, USA. E-mail: bct2@psu.edu 2 Present address: Biogen, Inc., 14 Cambridge Center, Cambridge, MA 02142, USA Many thanks are due to Bonnie Rup, Kristin Murray, Stan Rosenberg and Lori Fitz for advice and training on the BIAcore instrument, Edward LaVallie, Kristin Murray, Jeffrey Feldman, Vincent Ling, Lisa Racie-Collins, Elizabeth DiBlasio-Smith, Brian Kelly, Jeff Deetz, the Process Development Department, the Formulation Department and the Analytical Chemistry Department at Genetics Institute for assistance and valuable discussions and the DNA synthesis group for the synthesis of oligonucleotides. References Arnold,F.H. ( 1991) Bio/Technology , 9, 151–156. Google Scholar Becker,K.G., Nagle,J.W., Canning,R.D., Biddison,W.E., Ozato,K. and Drew,P.D. ( 1995) Hum. Mol. Genet. , 4, 685–691. Google Scholar Berg,J.M. and Shi,Y. ( 1996) Science , 271, 1081–1085. Google Scholar Bianchi,E. et al. ( 1995) J. Mol. Biol. , 247, 154–160. Google Scholar Briand,L., Chobert,J.M., Tauzin,J., Declerck,N., Leonil,J., Molle,D., Tran,V. and Haertle,T. ( 1997) Protein Eng. , 10, 551–560. Google Scholar Choi,S.J., Ahn,M., Lee,J.S. and Jung,W.J. ( 1997) Mol. Cells , 7, 575–581. Google Scholar Clore,G.M., Appella,E., Yamada,M., Matsushima,K. and Gronenborn,A.M. ( 1990) Biochemistry , 29, 1689–1696. Google Scholar Colas,P., Cohen,B., Jessen,T., Grishina,I., McCoy,J. and Brent,R. ( 1996) Nature , 380, 548–550. Google Scholar Ellman,G.L. ( 1959) Arch. Biochem. Biophys. , 82, 70–77. Google Scholar Gallagher,T., Bryan,P. and Gilliland,G.L. ( 1993) Proteins , 16, 205–213. Google Scholar Gayle,R.B. et al. ( 1993) J. Biol. Chem. , 268, 22105–22111. Google Scholar Hochuli,E., Bannwarth,W., Dobeli,H., Gentz,R. and Stuber,D. ( 1988) Bio/Technology , 6, 1321–1325. Google Scholar Holmgren,A. and Reichard,P. ( 1967) Eur. J. Biochem. , 2, 187–196. Google Scholar Hopp,T.P., Pricket,K.S., Price,V.L., Libby,R.T., March,C.J., Cerretti,D.P., Urdal,D.L. and Conlon,P.J. ( 1988) Bio/Technology , 6, 1204–1210. Google Scholar Irving,H. and Williams,R.J.P. ( 1953) J. Chem. Soc., Part III , 3192–3210. Google Scholar Ivanenkov,V.V., Jamieson,G.A., Gruenstein,E. and Dimlich,R.V. ( 1995) J. Biol. Chem. , 270, 14651–14658. Google Scholar Kriwacki,R.W., Schultz,S.C., Steitz,T.A. and Caradonna,J.P. ( 1992) Proc. Natl Acad. Sci. USA , 89, 9759–9763. Google Scholar LaVallie,E.R., DiBlasio,E.A., Kovacic,S., Grant,K.L., Schendel,P.F. and McCoy,J.M. ( 1993) Bio/Technology , 11, 187–193. Google Scholar Lilius,G., Persson,M., Bulow,L. and Mosbach,K. ( 1991) Eur. J. Biochem. , 198, 499–504. Google Scholar Lowman,H.B. and Wells,J.A. ( 1993) J. Mol. Biol. , 234, 564–578. Google Scholar Lowman,H.B., Bass,S.H., Simpson,N. and Wells,J.A. ( 1991) Biochemistry , 30, 10832–10838. Google Scholar Lu,Z., Murray,K.S., Van Cleave,V., LaVallie,E.R., Stahl,M.L. and McCoy,J.M. ( 1995) Bio/Technology , 13, 366–372. Google Scholar Lu,Z., DiBlasio-Smith,E.A., Grant,K.L., Warne,N.W., LaVallie,E.R., Collins-Racie,L.A., Follettie,M.T., Williamson,M.J. and McCoy,J.M. ( 1996) J. Biol. Chem. , 271, 5059–5065. Google Scholar Miwa,T., Damrongchai,N., Shinohara,H., Ikariyama,Y. and Aizawa,M. ( 1991) J. Biotechnol. , 20, 141–150. Google Scholar Nakayama,S., Moncrief,N.D. and Kretsinger,R.H. ( 1992) J. Mol. Evol. , 34, 416–448. Google Scholar Neri,D., de Lalla,C., Petrul,H., Neri,P. and Winter,G. ( 1995) Bio/Technology , 13, 373–377. Google Scholar Porath,J. ( 1992) Protein Express. Purif. , 3, 263–281. Google Scholar Riddles,P.W., Blakeley,R.L. and Zerner,B. ( 1983) Methods Enzymol. , 91, 49–60. Google Scholar Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning. A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Google Scholar Schmidt,T.G. and Skerra,A. ( 1993) Protein Eng. , 6, 109–122. Google Scholar Smith,J.W., Tachias,K. and Madison,E.L. ( 1995) J. Biol. Chem. , 270, 30486–30490. Google Scholar Smith,P.A., Tripp,B.C., DiBlasio-Smith,E.A., Lu,Z., LaVallie,E.R. and McCoy,J.M. ( 1998) Nucleic Acids Res. , 26, 1414–1420. Google Scholar Stayton,P.S., Shimoboji,T., Long,C., Chilkoti,A., Chen,G., Harris,J.M. and Hoffman,A.S. ( 1995) Nature , 378, 472–474. Google Scholar Tawfik,D.S., Chap,R., Eshhar,Z. and Green,B.S. ( 1994) Protein Eng. , 7, 431–434. Google Scholar Walker,B., Kasianowicz,J., Krishnasastry,M. and Bayley,H. ( 1994) Protein Eng. , 7, 655–662. Google Scholar Yip,T.T., Nakagawa,Y. and Porath,J. ( 1989) Anal. Biochem. , 183, 159–171. Google Scholar © Oxford University Press
TI - Investigation of the `switch-epitope' concept with random peptide libraries displayed as thioredoxin loop fusions
JF - Protein Engineering, Design and Selection
DO - 10.1093/protein/14.5.367
DA - 2001-05-01
UR - https://www.deepdyve.com/lp/oxford-university-press/investigation-of-the-switch-epitope-concept-with-random-peptide-MKtB5wd3y8
SP - 367
EP - 377
VL - 14
IS - 5
DP - DeepDyve
ER -