Selecting for antibody scFv fragments with improved stability using phage display with denaturation under reducing conditions

Journal of Immunological Methods 296 (2005) 159 – 170 www.elsevier.com/locate/jim

Research paper

Selecting for antibody scFv fragments with improved stability using phage display with denaturation under reducing conditions Eeva-Christine Brockmann*, Matthew Cooper1, Nelli Strfmsten2, Markus Vehni7inen, Petri Saviranta3 Department of Biotechnology, University of Turku, Tykisto¨katu 6A 6th floor, FIN-20520 Turku, Finland Received 9 June 2004; received in revised form 27 October 2004; accepted 11 November 2004 Available online 2 December 2004

Abstract Stability of single-chain Fvs (scFvs) can be improved by mutagenesis followed by phage display selection where the unstable variants are first inactivated by, for example, denaturing treatment. Here we describe a modified strategy for the selection of stabilized antibody fragments by phage display, based on denaturation under reducing conditions. This strategy was applied to an anti-thyroid-stimulating hormone (TSH) scFv fragment which refolded remarkably during the selection if denaturation was carried out in conventionally used non-reducing conditions. Refolding was, however, efficiently prevented by combining denaturation with reduction of the intra-domain disulfide bridges, which created favourable conditions for selection of clones with improved stability. Using this strategy, scFv mutants with 8 – 9 8C improved thermal stability and 0.8–0.9 M improved stability for guanidinium chloride were found after 4 – 5 enrichment cycles. The most stable mutants selected contained either LysH66Arg or AsnH52aSer mutations, which are known to stabilize other scFvs. Periplasmic expression level of the mutants was also improved. D 2004 Elsevier B.V. All rights reserved. Keywords: Antibody engineering; Disulfide bond; Phage display; Protein denaturation; Protein stability

Abbreviations: scFv, single-chain Fv antibody fragment; VH, heavy chain of an antibody Fv fragment; VL, light chain of an antibody Fv fragment; GdmCl, guanidinium chloride; TSH, thyroid-stimulating hormone; DTT, dithiotreitol; GSH, reduced glutathione; GSSG, oxidized glutathione; MPB, N-(3-maleimidopropionyl)biocytin; BSA, bovine serum albumin; IPTG, isopropyl-h-d-galactopyranoside; EDTA, ethylenediaminetetraacetic acid. * Corresponding author. Tel.: +358 2 3338062; fax: +358 2 3338050. E-mail address: [email protected] (E.-C. Brockmann). 1 Current address: School of Biomedical and Molecular Sciences, University of Surrey, Guildford, GU2 7XH, UK. 2 Current address: Department of Environmental and Biosciences and Institute of Biotechnology, Biocenter 2, P.O. Box 56, Viikinkaari 5, FIN-00014 University of Helsinki, Finland. 3 Current address: VTT Technical Research Centre of Finland, Medical Biotechnology Group, P.O. Box 106, FIN-20521 Turku, Finland. 0022-1759/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jim.2004.11.008

160

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

1. Introduction Stability is an essential property for proteins of biotechnological interest. Among antibodies, singlechain Fv (scFv) fragments (Raag and Whitlow, 1995) may have limited stability and therefore scFvs have been targets of stability engineering (Wo¨rn and Plu¨ckthun, 2001). The strategies described to improve scFv stability include both rational and evolutionary means. By rational design, stability can be improved by introducing known stabilizing mutations or mutations predicted by canonical immunoglobulin sequence (Steipe et al., 1994) or structure (Ewert et al., 2003). Grafting complementarity-determining region (CDR) loops to a stable framework (Jung and Plu¨ckthun, 1997) and introducing a disulfide bridge between the VH and VL domains (Young et al., 1995) also increase stability. As designing stabilizing mutations requires special know-how and testing different mutations and their combinations is labour-intensive, evolutionary methods are often preferred. Compared to rational design, these methods combine mutagenesis with cell-based screening (Kolmar et al., 1995; Auf der Maur et al., 2001; Philipps et al., 2003) or selection by some display technique, such as phage display (Hoogenboom et al., 1998) or ribosome display (Schaffitzel et al., 1999) and are extremely powerful. Random mutagenesis overcomes the need for rational design, although mutations known to have stabilizing effects can also be introduced into a library. Using display techniques, a single clone with improved stability can be enriched among millions of others by repeating selection cycles. If the selection threshold is correctly set, enrichment can be obtained within a few cycles. In conjunction with phage display, stability selection can be based on irreversible inactivation of unstable proteins by heat or guanidinium chloride (GdmCl) or by protease digestion. Stable variants are then collected by antigen binding activity or phage infectivity (Sieber et al., 1998; Kristensen and Winter, 1998; Jung et al., 1999; Martin et al., 2001). If selection is based on GdmCl denaturation, the phage-displayed protein may refold during the collection step. Because of refolding, denaturation by GdmCl has been estimated to select less efficiently for stability than denaturation by temperature (Jung et al., 1999). Denaturation of scFv can,

however, be made irreversible by reduction and the loss of disulfide bridges inside VL and VH domains which contribute to domain stability (Glockshuber et al., 1992). Only intrinsically stable scFvs or single domains are able to fold in reducing conditions (Frisch et al., 1996; Wo¨rn and Plu¨ckthun, 1998a,b; Ohage and Steipe, 1999) and, accordingly, reduction has been used with ribosome display to improve scFv stability based on ability to fold under reducing conditions (Jermutus et al., 2001). Here we have used an immunoassay to study the efficiency of denaturation of an anti-thyroid-stimulating hormone (TSH) scFv fragment under different conditions and found that the selection efficiency could be improved by carrying out denaturation under reducing conditions. We have also successfully enriched stabilized variants of the scFv by phage display using denaturation in a reducing environment as a means to inactivate unstable scFv variants, thereby demonstrating the applicability of the strategy as a means to select for stability by phage display.

2. Materials and methods 2.1. Primers, buffers, culture media and reagents HeavyFor (5V-GGA ATT CGG CCC CCG AGG CCG CAG AGA CAG TGA CCA GAG T-3V), HeavyRev (5V-GGC GGC GGC GGC TCC GGT GGT GGT GGA TCC SAG GTG CAG CTG CAG GAG-3V), scFOR (5V-GGA ATT CGG CCC CCG AGG CC-3V) and scREV (5V-TTA CTC GCG GCC CAG CCG GCC ATG GCG-3V) were from TAG Copenhagen (Copenhagen, Denmark); pAKfor (5VTGAAATACCTATTGCCTACG-3V) and pAKrev (5VCGCCATTTTTCACTTCACAG-3V) were from Alpha DNA (Montreal, Canada); WO267 (5V-CTA GAC TAG TAC AAT CCC TGG GCA CAA TTT TC-3V), MK24 (5V-GAT GGC AAA CGC TAA TAA G-3V), WO1236 (5V-GCC CAG CCG GCC ATG GCG CVW ATT GTK CTM ACY CAG TC-3V) and WO1244 (5VGGA GCC GCC GCC GCC AGA ACC ACC ACC ACC AGA ACC ACC ACC ACC ACG TTT CAG CTC CAG CTT GG-3V) were synthesized in-house using standard phosphoramidite chemistry. SfiI restriction sites are underlined.

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

SB medium, LB agar plates, TBS and PBS buffers were prepared as described (Sambrook et al., 1989). In TBS/bovine serum albumin (BSA), 1% (w/v) BSA was included. TBT-0.05 and TBT-0.5 contained 0.05% (v/v) and 0.5% Tween20 in TBS/BSA. The concentrations of antibiotics on agar plates and culture media were 25 Ag/ml chloramphenicol and 10 Ag/ml tetracycline. DELFIA products (assay buffer, wash buffer and enhancement solution) were from PerkinElmer Life Sciences (Turku, Finland). Monoclonal anti-M13-phage antibody 9E7 (produced in-house) and TSH (Scripps Laboratories, San Diego, CA) were labelled with Eu-N1-isothiocyanate (Mukkala et al., 1989) according to the instructions in the DELFIA Eu-labelling kit (Perkin Elmer Life Sciences). TSH was biotinylated at NH2-groups with biotin isothiocyanate (Mukkala et al., 1993). 2.2. Cloning 5404 antibody in scFv format RNA was isolated from hybridoma TSH cells washed with PBS, using the Quick Prep mRNA purification kit (Amersham Biosciences, Uppsala, Sweden). cDNA was synthesized with a RNA PCR Gene Amp kit (Part. No. N808-0017, Perkin Elmer, Norwalk, CT). 5404 antibody VH and VL genes were amplified form the cDNA using Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA). VH gene was amplified as a VH–CH1 fragment using heavy primer 1 provided in an RPAS mouse scFv module kit (Amersham Biosciences) and IgG1 CH antisense primer WO267. VL gene was amplified using the light primer mix of the kit. The PCR products were cloned using pGEM-T vector system (Promega, Madison, WI) and clones were checked by sequencing. For cloning in scFv format, 5404 VH was then amplified with HeavyFor and HeavyRev primers and VL with WO1236 and WO1244 primers. The VH and VL PCR products were agarose gel purified and PCR assembled to VL–VH scFv with (Gly4Ser)4 linker between the VL and VH. ScFv was cloned at SfiI restriction sites into vectors of pAK series (Krebber et al., 1997). 2.3. ScFv libraries To produce an scFv library, cloned 5404 scFv was mutated by error-prone PCR (Cadwell and Joyce,

161

1992; Fromant et al., 1995). The library (4.0106 clones with on average 2.5 random amino acid point mutations per scFv) was cloned into phagemid vector pAK100 (Krebber et al., 1997; a gift from Andreas Plqckthun, Department of Biochemistry, University of Zqrich, Switzerland) and desalted ligation mixture was transformed by electroporation (Dower et al., 1988) into Escherichia coli XL1-Blue strain [recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac [F’ proAB lacI q ZDM15 Tn10 (Tetr)] (Stratagene). Phage stock was produced as described earlier (Korpima¨ki et al., 2002) by infecting the transformed library with VCS M13 helper phage (Kanr, Stratagene). Phages that displayed TSH binding activity were collected with biotinylated TSH, using the protocols described in Section 2.6. E. coli culture was infected with the collected phage, and after overnight culturing, phagemid DNA was isolated using a Quantum Prep Plasmid miniprep Kit (Bio-Rad Laboratories, Hercules, CA). ScFv genes were amplified from the phagemid DNA using primers pAKfor and MK24 and recombined using a DNA shuffling technique (Stemmer, 1994). The resultant library was cloned into vector pAK100 and phage stock was produced as with the first library. This library was used later in selections and had a diversity of 1.2107 clones (background from self-ligated vector was 2.1%) and on average there were three amino acid mutations in each scFv. 2.4. ScFv production, purification and biotinylation ScFv was produced periplasmically in E. coli strain XL1-Blue, using vector pAK3FC (Korpima¨ki et al., 2004). The vector introduces a Cys residue into the C-terminus of scFv for site-specific biotinylation and a hexahistidine tag for affinity purification. One liter cultures were grown at 30 8C, 240 rpm in SB medium supplemented with 0.2– 0.5% glucose, 25 Ag/ml chloramphenicol and 10 Ag/ ml tetracycline, induced at OD600=0.8 by addition of 0.2 mM isopropyl-h-d-thiogalactopyranoside (IPTG, Fermentas, Hanover, MD) and continued overnight at 26 8C, 220 rpm. Cells were harvested by centrifugation (10 min, 8000g, 4 8C), resuspended in 40 ml of 10 mM Na-phosphate buffer (pH 7.5), 2 mM ethylenediaminetetraacetic acid (EDTA) and disrupted using a French press. Eight percent

162

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

glycerol was added to minimize scFv aggregation and lysate was clarified by centrifugation (20 min, 11,000g 4 8C). ScFv was purified from the lysate first at pH 6 with SP Sepharose Fast Flow cation exhange chromatography (Amersham Biosciences) and eluted with 400 mM NaCl at pH 6.3. This was followed by purification with Ni-NTA immobilized metal affinity chromatography superflow (QIAGEN, Hilden, Germany). The C-terminal Cys-residue was biotinylated in a Ni-NTA column. For this, scFv was reduced at pH 8 with 5 mM 2-mercaptoethanol (Sigma-Aldrich, St. Louis, MO). 2-Mercaptoethanol was then removed by washing the column. 100 AM N-(3-maleimidopropionyl)biocytin (MPB, SigmaAldrich) was added and scFv was biotinylated by MPB during a 1.5h incubation at room temperature. Unreacted MPB was then washed away and scFv was eluted with 400 mM imidazol in phoshphate buffer. Unbiotinylated scFv was removed using a SoftLink avidin column (Promega). 2.5. Measuring the stability of soluble scFv and phage infectivity To measure stability, biotinylated scFv was incubated for 2 h at 4–60 8C (0.5 Ag/ml scFv) or 1 h in 0–5 M GdmCl (5 Ag/ml scFv) in TBS/BSA with and without 1 mM dithiotreitol (DTT) and then diluted 1/ 10 (thermally denatured samples) or 1/50 (GdmCl denatured samples) in DELFIA assay buffer. Biotinylated scFv was bound to a streptavidin plate (Innotrac Diagnostics Oy, Turku, Finland) following a 1-h incubation with slow shaking at room temperature. After four washes with DELFIA wash buffer, Eulabelled TSH was added, and after 1 h incubation at room temperature the plate was washed four times. Time-resolved fluorescence of Eu3+ was measured after 5 min development with DELFIA enhancement solution and the signal was compared to the signal obtained with untreated scFv. The effect of denaturing treatments on phage infectivity was studied by infecting cultured XL1Blue cells with aliquots of phage denatured as described above and plating them onto LB agar plates supplemented with tetracycline and chloramphenicol. Infectivity was determined as the percentage of transforming units (tfu) compared with untreated control.

2.6. Enrichment of the libraries Stabilized scFv mutants were enriched from the library by phage display selection. First, immunoreactive phage (21010–11012 tfu phage/200 Al TBT0.05) were collected by biotinylated TSH and Cellection biotin binding magnetic beads (4106 beads, Dynal Biotech, Oslo, Norway). After five washes with TBT-0.5, bound phage were eluted by digestion with DNase (100 U, Dynal Biotech). XL1Blue cells were infected with the eluate and new phage stock was produced as described earlier (Korpima¨ki et al., 2002). To select the most stable clones, the following panning rounds were preceded by phage denaturation with either elevated temperature or GdmCl. The conditions used in selection are shown in Table 1. Phage that were denatured with GdmCl were additionally allowed to refold for 1 or 2 h by 1/50 dilution in refolding buffer (TBT with 0.1% BSA, 1 mM reduced (GSH) and 0.2 mM oxidized (GSSG) glutathione). New phage stocks were produced. Overnight phage production was carried out at 30 8C to favour the expression of stable scFv variants. Individual phage clones were isolated from the enriched libraries by infecting 50 Al of fresh XL1Blue culture with 100–1000 tfu phage and plating the infected cells on LB agar plates containing tetracycline and chloramphenicol. Phage stocks were produced separately for individual clones. 2.7. Measuring phage antigen binding activity Phage (51012 tfu/ml) were denatured for 1 h either at 40 8C or in 1.5 M GdmCl (in TBS pH 8.0, 1 or 2 mM DTT) and diluted in DELFIA assay buffer or refolding buffer as described above with the measurement of soluble scFv. Residual antigen binding activity was measured by a two-site immunofluorometric assay. First, phage were bound to streptavidin microtitration wells coated with biotinylated TSH (12 ng/well). After 1 h incubation with slow shaking at room temperature, the plate was washed four times with DELFIA wash buffer. The detection antibody, Eu-labelled 9E7 anti-phage monoclonal antibody (50 ng), was then added and incubated for 1 h. Finally, the plate was washed four times and time-resolved fluorescence of Eu3+ was measured after 5 min development with DELFIA enhancement solution.

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

163

Table 1 Conditions used in phage display selection and enrichment of stability in the course of panning Panning round

T1 T2 T3 T4 T5 G1 G2 G3 G4 G5

Selection Denaturation

Refolding

Selection stringency (%)

40 8C, overnight 35 8C, 1 mM DTT, 2 h 35 8C, 1 mM DTT, 2 h 40 8C, 1 mM DTT, 2 h 40 8C, 1 mM DTT, 2 h 1.0 M GdmCl, 1 mM DTT, 1.0 M GdmCl, 2 mM DTT, 1.0 M GdmCl, 2 mM DTT, 1.5 M GdmCl, 2 mM DTT, 1.5 M GdmCl, 2 mM DTT,

– – – – – 2 2 1 1 1

85 10 56 10 n.a. 2 39 24 19 19

1 1 1 1 1

h h h h h

h, h, h, h, h,

1/50 1/50 1/50 1/50 1/50

dilution dilution dilution dilution dilution

b

Stability of enriched phage (%)a 1 4 13 50 62 3 6 9 22 42

a Stability of the enriched phage as residual antigen binding activity to biotinylated TSH after 1 h denaturation at 40 8C, 1 mM DTT (rounds T1–T5) or in 1.5 M GdmCl, 2 mM DTT (rounds G1–G5). The residual antigen binding activity of wild type scFv in these conditions is 2% and 1%, respectively. b Selection stringency as residual antigen binding activity of phage used for selection after the treatments.

The residual antigen binding activity was defined as the ratio of the signal obtained with denatured phage relative to that of untreated phage.

and washed four times. Time-resolved fluorescence of Eu3+ from bound Eu-TSH was measured after 5 min development in DELFIA enhancement solution. ScFv (ng/ml) was quantified using the calibration curve.

2.8. Analysis of the expression level of scFv ScFv mutants were cloned at SfiI sites into vector pAK300 (Krebber et al., 1997) and transformed into XL1-Blue strain. To analyse the expression level, scFv variants were expressed in 50 ml flask cultures with SB medium, chloramphenicol, tetracycline and 0.2% glucose. Expression was initiated at OD600=0.6– 0.8 with addition of 0.4 mM IPTG and continued for 18 h at 26 8C, 220 rpm or 37 8C, 300 rpm, after which the amount of soluble scFv was measured from culture media and periplasmic extracts. Periplasmic extracts were prepared suspending cells collected by centrifugation (10 min, 5000g, 4 8C) in extraction buffer (70 mM Tris–Cl pH 9.0, 2 mM EDTA, 1 mg/ml lysozyme). Proteins were released from the cells during a 10-min incubation at 30 8C after which cell debris was removed by centrifugation. ScFv concentration in the samples was measured by competitive immunoassay. First, 50 ng of biotinylated scFv was diluted in DELFIA assay buffer and bound for 1 h on to streptavidin microtiter wells. Then after 4 washes with DELFIA wash buffer, an equilibrated mixture of Eu-labelled TSH and sample scFv or calibrator (purified scFv) dilutions was added, incubated for 1 h at room temperature with slow shaking

2.9. Measuring the thermodynamic stability of soluble scFv ScFvs were expressed from vector pAK300 with no free Cys residue. They were purified as described with vector pAK3FC (Section 2.4), but the biotinylation step was omitted. After Ni-NTA, scFvs were purified to homogeneity with a Resource cation exchange column (Amersham Biosciences) using A¨ KTA Explorer system (Amersham Biosciences). To measure the thermodynamic stability, scFvs were equilibrated (at 10 Ag/ml) overnight at 10 8C in TBS containing 0–5 M GdmCl. Fluorescence emission at 300–400 nm was recorded using a Cary Eclipse spectrofluorometer (Varian Instruments, Walnut Creek, CA) with excitation at 280 nm. The averaged maximum wavelength of five parallel measurements was recorded.

3. Results 3.1. Efficient denaturation under reducing conditions In this study we aimed to improve the stability of a cloned 5404 anti-TSH scFv and overcome problems

164

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

possibly related to its low stability, such as high tendency to inactivate and aggregate during purification and storage. To improve the stability we used random mutagenesis followed by phage display selection. Denaturation was used to inactivate the unstable variants in the library and the surviving variants were then collected based on binding to TSH. We first tried selection by thermal denaturation because this had proven to be efficient (Jung et al., 1999). However, using this strategy the wild type scFv could not be inactivated efficiently enough to get high stringency for selection. Therefore we looked for conditions where the wild scFv could be completely inactivated and other strategies that could be used for selection. We tried reduction because we thought this was theoretically reasonable and it had been shown to work with ribosome display (Jermutus et al., 2001). To find optimal stringency for selection, we studied the efficiency of inactivation of the wild type scFv by measuring the residual antigen binding activity of biotinylated scFv to europium-labelled TSH after denaturing treatments. To minimize the effect of GdmCl on TSH binding, all samples were diluted 1/ 50 to contain no more than 0.1M GdmCl during the assay. However, this establishes favourable conditions for the refolding of denatured scFv. As scFv was active even after denaturation in 6 M GdmCl and an increase in residual antigen binding activity was observed after denaturation in non-reducing conditions at GdmCl concentrations higher than 1 M with the wild type scFv (Fig. 1b), it is obvious that scFv refolds during the immunoassay. Inefficient inactivation of scFv by heat (Fig. 1a) suggests that refolding occurs also after thermal denaturation. Efficient inactivation of the scFv could only be obtained when denaturation was carried out in a reducing environment, i.e. in the presence of dithiotreitol (Fig. 1a and b). Selection of stable clones by phage display follows a similar procedure with binding to TSH after denaturation in either GdmCl or by heat, and therefore reduction is also again essential to prevent refolding. 3.2. Effect of reducing conditions on phage infectivity Loss of phage infectivity in the selection procedure can limit the use of some denaturing conditions, so the effect of denaturing treatments on M13 phage infectivity was studied. Denaturation under reducing

Fig. 1. Antigen binding activity of the anti-TSH scFv fragments after denaturation in reducing (1 mM DTT) and in non-reducing (no DTT) conditions. (a) Denaturation of the wild type scFv with increased temperature and (b) in GdmCl. (c) Denaturation of stabilized T5.10 scFv in GdmCl.

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

conditions had an effect on M13 phage infectivity. Under reducing conditions phage lost their infectivity at 50 8C, i.e. about 10 8C lower than in a non-reducing environment (Fig. 2a). In non-reducing conditions M13 phage tolerate even 6 M GdmCl (Jung et al., 1999), but under reducing conditions, phage infectivity is reduced to b10% in 2 M GdmCl. Phage infectivity is, however, largely regained by refolding (Fig. 2b), so the loss of phage infectivity in a reducing environment does not limit the use of phage display in stability selection. To avoid the effect of refolding on the selection process, refolding should be done after elution of the selected phage.

Fig. 2. Effect of denaturing treatment on M13 phage infectivity. (a) Infectivity after incubation at different temperatures in reducing (1 mM DTT) and in non-reducing (no DTT) conditions. (b) Infectivity after GdmCl denaturation in reducing conditions (2 mM DTT) and followed by 2 h refolding in buffer with a redox shuffling system.

165

3.3. Enrichment of stable scFv fragments by phage display Two different selection strategies were used to enrich variants of 5404 scFv. The first strategy was based on thermal denaturation under reducing conditions and aimed at efficient enrichment of the most stable scFv mutants. The other strategy was based on GdmCl denaturation under reducing conditions, but denaturation was combined with a short refolding step to see whether mutants with improved folding properties and reduced tendency to aggregate during folding could be enriched at the same time. Refolding was initiated by diluting the denatured phage in a buffer containing GSH and GSSG. The redox-shuffling system created by GSH and GSSG should enable correct disulfide bond formation (Lilie et al., 1998) and thereby ease the folding of scFv into active forms. Exact conditions used in the two different selection procedures are shown in Table 1. Enrichment of stability was followed by measuring the residual antigen binding activity after denaturation under reducing conditions for each enriched phage library (Table 1). During the first panning rounds with thermal denaturation, mild selection was used to ensure that clones with possibly only a slightly improved stability were not lost. When enrichment was evident, the stringency was increased to allow more efficient selection of the most stable variants. During the 4th and 5th panning rounds, denaturation conditions were used that almost completely inactivated the wild type scFv, thus allowing very efficient selection of stable variants. Under non-reducing conditions such an efficient inactivation of the wild type scFv would not have been possible, based on the results shown in Fig. 1 and the limits of phage stability (Fig. 2a). After five panning rounds, 62% of the antigen binding activity of the thermally selected library was retained during 2 h denaturation at 40 8C under reducing conditions (the selection used in the last panning round), compared with 2% residual activity of the phage carrying wild type scFv. When the library was selected with chemical denaturation and combined refolding, the residual antigen binding activity of the enriched phage library after the treatments used in the last panning round (denaturation in 1.5 M GdmCl followed by refolding) was 27%. Using the same

166

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

conditions, the wild type phage had 1% residual antigen binding activity after denaturation and a total of 11% activity after denaturation and refolding. After five panning rounds, individual clones were isolated for analysis from both libraries. Clones were randomly taken from the 4th round (15 clones from both selection schemes) and the 5th round (5 clones from both selection schemes). The properties of these clones are shown in Fig. 3, as screened in phage format. All the mutants selected by thermal denaturation (later called T4.xx and T5.xx) were more stable than the wild type scFv and 12 of the 20 thermally selected clones preserved more than 70% of their antigen binding activity at 40 8C (under reducing conditions). Stability of these mutants to GdmCl was also improved and correlated with thermal stability (r=0.73), but denaturation in 1.5 M GdmCl differentiated the clones less well because of higher levels of inactivation. The mutants selected by combined GdmCl denaturation and refolding (later called G4.xx

and G5.xx) were altogether less stable to both GdmCl and thermal denaturation than the mutants selected by thermal denaturation (Fig. 3a and b). Nine of the 20 clones selected were less stable to heating than the least stable thermally selected clone, and four of these clones exhibited stability similar to the wild type scFv. However, some stable clones were enriched and five clones had N70% residual antigen binding activity after denaturation at 40 8C under reducing conditions, where the wild type is completely inactivated. Both stability and refolding contributed to the selection with GdmCl and, indeed, refolding of some of the clones was improved compared to the wild type scFv phage (Fig. 3b). 3.4. Characterization of the selected clones Five clones were selected from each enriched library for more detailed analysis. Seven of these 10 clones were unique by sequence (Table 2). Two

Fig. 3. Screening of enriched phage clones. (a and b) Stability distribution of enriched clones, measured as residual antigen binding activity to TSH after thermal denaturation (2 h at 40 8C, 1 mM DTT) and GdmCl denaturation (1 h in 1.5 M GdmCl, 2 mM DTT). (b) Effect of refolding in selection by GdmCl, shown as residual antigen binding activity to TSH after GdmCl denaturation (1 h in 1.5 M GdmCl, 2 mM DTT) and refolding (1/50 dilution in refolding buffer). Properties of the wild type 5404 scFv are shown for comparison.

A Q I – – – – – – – – – – – – – – – – – T T R T CDR1 S – – – L – – – R – – – – – – K L – – P P P – –

The residues are numbered according to Kabat et al. (1991). a Total expression level including periplasmic protein and scFv leaked into growth medium. b The denaturation midpoints are denaturing conditions in the presence of 1 mM DTT, where the antigen binding activity of C-terminally biotinylated scFv to TSH is reduced to 50%.

T (8C)

36 44 38 45 n.a. 37 40 37 0.4 1.2 0.8 1. 3 n.a. 0.8 0.9 0.8

GdmCl (M) 37 8C

0.2 7.8 13 25 n.a 0.2 0.2 0.2 2.5 19 16 14 n.a 6.0 3.9 3.5 Q L – – – – – – – – – – – – H P FR4 S – P – – – – – R – G – – – – – S N – – – – – – K M – – – – R – R – – – – – – I FR3

66 82 82a 83 84 105 108 26 8C 52a

N S – – – – – – CDR2 Q – – – – – L – FR1 L V V – – V – – FR4 T – A – – A – –

91 94 97 106 3

S I – V – – T – T – – – – – – – CDR3 I – – T T T T – FR3

72

A S – – – – T – T T – – – – – FR1 Wild type T4.2 T4.39/G4.31 T5.10/T5.11/G5.13 T4.16 G4.7 G4.38 G5.14 Region

P A – – – – – – FR2

9 10 15 18 22 25 26 27c 27d 43 Residue

T – R – – – – –

VL-position scFv clone

Table 2 Isolated scFv mutants: Mutations, expression and stability characteristics

VH-position

Expression levela (mg/l) Denaturation midpointb

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

167

sequences (T5.10/T5.11/G5.13, later T5.10 and T4.39/G4.31, later T4.39) had been enriched by both selection schemes. The proteins were purified and stability was characterized in an immunoassay that measured the residual antigen binding activity of soluble biotinylated scFv after denaturation under reducing conditions. T5.10 together with T4.2 had a markedly improved stability with denaturation midpoints of 45 and 44 8C or 1.3 and 1.2 M GdmCl, respectively, compared to the wild type scFv with denaturation midpoints of 36 8C or 0.4 M GdmCl. The stability of these clones was confirmed by thermodynamic analysis (Fig. 4). Also, the rest of the sequenced mutants had improved stability (Table 2). Periplasmic expression levels of the sequenced clones was also improved (Table 2), except for G4.38 and G5.14, the clones that were taken for characterization based on their ability to refold. T4.2, T4.39 and T5.10 had the highest expression levels with a 5–8 fold increased expression at 26 8C and a 40–125 fold improved expression at 37 8C (Table 2). The highest expression level was achieved with T5.10 at 37 8C (25 mg/l), and the T5.10 scFv was expressed at higher levels at 37 8C than at 26 8C (14 mg/l), even though at 37 8C most of the protein leaked into the growth medium during overnight expression.

Fig. 4. GdmCl-induced equilibrium unfolding of T5.10 and T4.2 scFv fragments detected as a change in intrinsic protein fluorescence.

168

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

4. Discussion In this study, we have shown that scFv fragments with improved stability can be selected with phage display using denaturation in a reducing environment as a means to inactivate unstable variants. Reduction prevents refolding of the denatured structures, resulting in an efficient and irreversible inactivation. As a result, selection is based more purely on stability rather than on competence to refold, and the selection stringency is increased. Reducing conditions have previously been used with ribosome display to select for improved stability (Jermutus et al., 2001). However, selection was based on the ability to fold without the formation of disulfide bonds, rather than on the ability to withstand denaturing conditions, as was the case here. Denaturation with concomitant reduction of disulfide bridges also mimics the idea of an initial destabilization of scFv structure by removal of disulfide bridges, which can be used to increase selection stringency (discussed by Wo¨rn and Plu¨ckthun, 1998a), but as no reverse engineering of the destabilized residues is needed, stability engineering is more convenient here. To the best of our knowledge, this is the first description of the use of reduction in phage display selection for stability. In antibody phage display, the capture step usually requires non-denaturing conditions, which may favour the refolding of some of the denatured antibody fragments. GdmCl denaturation is often at least partially reversible unlike thermal denaturation, and more efficient selection of stability has therefore been obtained with thermal selection than by chemical denaturation with GdmCl (Jung et al., 1999). The wild type 5404 scFv antibody, however, also refolds after thermal denaturation (Fig. 1a). As refolding can clearly compromise the efficiency of stability selection, it is desirable to prevent this process as completely as possible. Using immunoassay simulations we could conveniently predict denaturing conditions that, combined with the reducing environment, led to irreversible inactivation of the wild type scFv but allowed even slightly more stable variants to be enriched. The results show that stability is efficiently selected with thermal denaturation under reducing conditions. As thermal stability of the selected scFv clones correlated under reducing conditions with chemical stability, it should be possible to select for

stability also by denaturation with GdmCl, if refolding is prevented by reduction. Refolding does not completely prevent the selection of stable clones, as seen in the experiment with GdmCl denaturation followed by refolding (Fig. 3b), but selection of stable clones would be much more efficient without refolding. Since refolding is easily prevented by reduction of the denatured structures, denaturation under reducing conditions could be useful in stability selection of scFvs other than 5404. In cases where refolding occurs, the residual antigen binding activity, as measured by immunoassay, is not proportional to the actual thermodynamic stability. Comparison of the GdmCl unfolding curve of, for example, T5.10 scFv (Fig. 4) with the measured antigen binding activity after denaturation in either reducing or non-reducing conditions (Fig. 1c) shows that the immunoassay data corresponds to the thermodynamic stability only when the denaturation is made irreversible. Indeed, the immunoassay measurement of residual antigen binding activity after denaturation under reducing conditions can be useful as an indicative measure of scFv stability in cases where the determination of the actual thermodynamic stability is not possible or practical. Based on the immunoassays wild type scFv refolds poorly after partial denaturation in low GdmCl concentrations in a non-reducing environment (Fig. 1b) and the scFv is possibly trapped in a conformation that does not refold easily. This kind of situation may arise, e.g., if one of the domains is less stable than the other so that it is unfolded, while the other domain remains folded (Ramm et al., 1999). This may result in aggregation. We were interested in whether the tendency to aggregate could be reduced by selecting for refolding of partially unfolded structures. However, improved refolding from a partially denatured state was achieved also by selecting for stability as seen with T5.10 (Fig. 1c), and it may not be necessary to overcome the trap by selecting for refolding. Another hypothesis was that refolding might favour selection of variants that have good in vivo folding properties. However, this idea was not supported since the expression level was low with the clones that were characterized by good refolding on phage but possessed only a moderate stability (G4.38 and G5.14; Table 2). Thus selection for stability by irreversible denaturation is likely to produce more favourable

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

properties than selection by refolding from a partially denatured state. In folded VL and VH domains the intra-domain disulfide bridges are buried and protected from reducing agents (Kikuchi et al., 1986). Unfolding is therefore a prerequisite for reduction and reducing conditions as such are not likely to cause scFvs to denature. However, in intact antibody CL domains the disulfide-bridge can be reduced without any denaturing agent, since CL undergoes global unfolding even under physiological conditions, but high concentrations of reductant are needed for reduction (Kikuchi et al., 1986). Compared to intact domains, scFvs are stabilized through inter-domain interactions, which make them less prone for unfolding (Ja¨ger and Plu¨ckthun, 1999) and in the mild reducing conditions used in this study, the disulfide bridges of scFv are therefore probably not reduced if there is no denaturing agent. The phage minor coat protein does not unfold solely by reduction, since the infectivity of M13 phage is not affected by reducing agents themselves (Rakonjac and Model, 1994). The most stable scFv variants T5.10 and T4.2 contained the VH domain mutations LysH66Arg and AsnH52Ser, respectively, which are known to stabilize other scFv fragments (Wo¨rn and Plu¨ckthun, 1998a). These mutations were first found through stability selection with an engineered, totally disulfide-free variant of an scFv naturally lacking the disulfide bond in VH. (Proba et al., 1998). The destabilizing effect caused by the removal of the remaining disulfide bond from the VL domain could be overcome by the LysH66Arg mutation alone (Proba et al., 1998). The fact that the same LysH66Arg mutation was also found in our selection, based on denaturation combined with reduction, suggests that it may not be necessary to remove the disulfide bridges genetically in order to create an extremely powerful stability selection.

Acknowledgements We thank Medix Biochemica (Kauniainen, Finland) for providing genetic material for the 5404 antibody. De´borah Braun is thanked for help with panning. The study was financially supported by the National Technology Agency of Finland (TEKES).

169

References Auf der Maur, A., Escher, D., Barberis, A., 2001. Antigen independent selection of stable intracellular single-chain antibodies. FEBS Lett. 508, 407. Cadwell, R., Joyce, G., 1992. Randomization of genes by PCR mutagenesis. PCR Methods Appl. 2, 28. Dower, W.J., Miller, J.F., Ragsdale, C.W., 1988. High efficiency transformation of E. coli by high voltage electroporation. Nucleic Acids Res. 16, 6127. Ewert, S., Honegger, A., Plqckthun, A., 2003. Structure-based improvement of the biophysical properties of immunoglobulin VH domains with a generalizable approach. Biochemistry 42, 1517. Frisch, C., Kolmar, H., Schmidt, A., Kleemann, G., Reinhardt, A., Pohl, E., Uso´ n, I., Schneider, T.R., Fritz, H.-J., 1996. Contribution of the intramolecular disulfide bridge to the folding stability of REIV, the variable domain of a human immunoglobulin n light chain. Fold. Des. 1, 431. Fromant, M., Blanquet, S., Plateau, P., 1995. Direct random mutagenesis of gene sized DNA fragments using polymerase chain reaction. Anal. Biochem. 224, 347. Glockshuber, R., Schmidt, T., Plqckthun, A., 1992. The disulfide bonds in antibody variable domains: effect on stability, folding in vitro and functional expression in Escherichia coli. Biochemistry 31, 1270. Hoogenboom, H.R., de Bruine, A.P., Hufton, S.E., Hoet, R.M., Arends, J.W., Roovers, R.C., 1998. Antibody phage display technology and its applications. Immunotechnology 4, 1. J7ger, M., Plqckthun, A., 1999. Domain interactions in antibody Fv and scFv fragments: effects on unfolding kinetics and equilibria. FEBS Lett. 462, 307. Jermutus, L., Honegger, A., Schwesinger, F., Hanes, J., Plqckthun, A., 2001. Tailoring in vitro evolution for protein affinity or stability. Proc. Natl. Acad. Sci. U. S. A. 98, 75. Jung, S., Plqckthun, A., 1997. Improving in vivo folding and stability of a single-chain Fv antibody fragment by loop grafting. Protein Eng. 10, 959. Jung, S., Honegger, A., Plqckthun, A., 1999. Selection for improved protein stability by phage display. J. Mol. Biol. 294, 163. Kabat, E., Wu, T., Perry, H., Gottesman, K., Foeller, C., 1991. Sequences of Proteins of Immunological Interest, 5th ed. NIH Publication 91-3242US Department of Health and Human Services, Washington, DC. Kikuchi, H., Goto, Y., Hamaguchi, K., 1986. Reduction of the buried intrachain disulfide bond of the constant fragment of the immunoglobulin light chain: global unfolding under physiological conditions. Biochemistry 25, 2009. Kolmar, H., Frisch, C., Gotze, K., Fritz, H.J., 1995. Immunoglobulin mutant library genetically screened for folding stability exploiting bacterial signal transduction. J. Mol. Biol. 251, 471. Korpim7ki, T., Rosenberg, J., Virtanen, P., Karskela, T., Lamminm7ki, U., Tuomola, M., Vehni7inen, M., Saviranta, P., 2002. Improving broad specificity hapten recognition with protein engineering. J. Agric. Food Chem. 50, 4194. Korpim7ki, T., Hagren, V., Brockmann, E.-C., Tuomola, M., 2004. Generic lanthanide fluoroimmunoassay for the simultaneous

170

E.-C. Brockmann et al. / Journal of Immunological Methods 296 (2005) 159–170

screening of 18 sulfonamides using an engineered antibody. Anal. Chem. 76, 3091. Krebber, A., Bornhauser, S., Burmester, J., Honegger, A., Willuda, J., Bosshard, H.R., Plqckthun, A., 1997. Reliable cloning of functional antibody variable domains from hybridomas and spleen repertoires employing a reengineered phage display system. J. Immunol. Methods 201, 35. Kristensen, P., Winter, G., 1998. Proteolytic selection for protein folding using filamentous bacteriophages. Fold. Des. 3, 321. Lilie, H., Schwarz, E., Rudolph, R., 1998. Advances in refolding of proteins produced in E. coli. Curr. Opin. Biotechnol. 9, 497. Martin, A., Sieber, V., Schmid, F.X., 2001. In-vitro selection of highly stabilized protein variants with optimised surface. J. Mol. Biol. 309, 717. Mukkala, V.-M., Mikola, H., Hemmil7, I., 1989. The synthesis and use of activated N-benzyl derivatives of diethylenetriaminetetraacetic acids: alternative reagents for labelling of antibodies with metal ions. Anal. Biochem. 176, 319. Mukkala, V.-M., H7nninen, E., Hemmil7, I., 1993. Synthesis of two biotin isotiocyanate derivatives. Proceedings of the 8th European Symposium on Organic Chemistry. P422, Sitges, Spain. Ohage, E., Steipe, B., 1999. Intrabody construction and expression: I. The critical role of VL domain stability. J. Mol. Biol. 291, 1119. Philipps, B., Hennecke, J., Glockshuber, R., 2003. FRET-based in vivo screening for protein folding and increased protein stability. J. Mol. Biol. 327, 239. Proba, K., Wfrn, A., Honegger, A., Plqckthun, A., 1998. Antibody scFv fragments without disulfide bonds made by molecular evolution. J. Mol. Biol. 275, 245. Raag, R., Whitlow, M., 1995. Single-chain Fvs. FASEB J. 9, 73. Rakonjac, J., Model, P., 1994. The influence of the reducing agent dithiotreitol in vitro on infectivity of f1 particles. Gene 142, 153.

Ramm, K., Gehrig, P., Plqckthun, A., 1999. Removal of the conserved disulfide bridges from the scFv fragment of an antibody: effects on folding, kinetics and aggregation. J. Mol. Biol. 290, 535. Sambrook, J., Fritsch, E., Maniatis, T., 1989. Molecular Cloning: A Laboratory Manual, 2nd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Schaffitzel, C., Hanes, J., Jermutus, L., Plqckthun, A., 1999. Ribosome display: an in vitro method for selection and evolution of antibodies from libraries. J. Immunol. Methods 231, 119. Sieber, V., Plqckthun, A., Schmid, F., 1998. Selecting proteins with improved stability by a phage-based method. Nat. Biotechnol. 16, 955. Steipe, B., Schiller, B., Plqckthun, A., Steinbacher, S., 1994. Sequence statistics reliably predict stabilizing mutations in a protein domain. J. Mol. Biol. 240, 188. Stemmer, W.P.C., 1994. DNA shuffling by random fragmentation and reassembly: in vitro recombination for molecular evolution. Proc. Natl. Acad. Sci. U. S. A. 91, 10747. Wfrn, A., Plqckthun, A., 1998a. Mutual stabilization of VL and VH in single-chain antibody fragments, investigated with mutants engineered for stability. Biochemistry 37, 13120. Wfrn, A., Plqckthun, A., 1998b. An intrinsically stable antibody scFv fragment can tolerate the loss of both disulfide bonds and fold correctly. FEBS Lett. 427, 357. Wfrn, A., Plqckthun, A., 2001. Stability engineering of antibody single-chain Fv fragments. J. Mol. Biol. 305, 989. Young, N.M., MacKenzie, C.R., Narang, S.A., Oomen, R.P., Baenziger, J.E., 1995. Thermal stabilization of a single-chain Fv antibody fragment by introduction of a disulfide bond. FEBS Lett. 377, 135.