| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Protein Engineering vol. 16 no. 5 pp. 381-386, 2003
© 2003 Oxford University Press
VL position 34 is a key determinant for the engineering of stable antibodies with fast dissociation rates
1UMR 7100, ‘Biotechnologie des Interactions Macromoléculaires’, Ecole Supérieure de Biotechnologie de Strasbourg (ESBS), Parc d’Innovation, boulevard Sébastien Brant, BP 10413, F-67412 Illkirch Cedex, 2Laboratoire d’Ingénierie des Macromolécules, Institut de Biologie Structurale Jean-Pierre Ebel (CEA/CNRS/UJF), 41 avenue Jules Horowitz, F-38027 Grenoble Cedex 1, France and 3Center for Surface Biotechnology, Uppsala Biomedical Center, Box 577, S-751 23 Uppsala, Sweden N.Hugo and M.Weidenhaupt contributed equally to this work.
4 To whom correspondence should be addressed. e-mail: daniele.altschuh{at}esbs.u-strasbg.fr
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Predictive engineering of antibodies exhibiting fast kinetic properties could provide reagents for biotechnological applications such as continuous monitoring of compounds or affinity chromatography. Based on covariance analysis of murine germline antibody variable domains, we selected position L34 (Kabat numbering) for mutational studies. This position is located at the VL/VH interface, at the base of the paratope but with limited antigen contacts, thus making it an attractive position for mild alterations of antigen binding properties. We introduced a serine at position L34 in two different antibodies: Fab (fragment antigen binding) 57P (Asn34Ser) and scFv (single chain fragment variable) 1F4 (Gln34Ser), that recognize peptides derived from the coat protein of tobacco mosaic virus and the oncoprotein E6, respectively. Both mutated antibodies exhibited similar properties: (i) expression levels of active fragments in Escherichia coli were markedly improved; (ii) thermostability was enhanced; and (iii) dissociation rate parameters (koff) were increased by 2- and at least 57-fold for scFv1F4 and Fab57P, respectively, while their association rate parameters (kon) remained unchanged. The L34 Ala and Thr mutants of both antibody fragments did not possess these properties. This first demontration of similar effects observed in two antibodies with different specificities may open the way to the predictive design of molecules with enhanced stability and fast dissociation rates.
Keywords: antibody engineering/binding kinetics/covariance/expression level/thermostability
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Stable antibody fragments with fast dissociation kinetics are valuable tools for in vitro applications such as affinity chromatography or monitoring of specific substances in continuous processes. The dissociation rate of antigen– antibody interactions may be increased by modifying the antigen binding site (paratope), but paratope residues play different roles in different antigen–antibody interactions. A mutational analysis of each new antibody is required to identify positions that influence antigen binding, without being essential. In addition, protein mutations affect other properties such as expression level and thermodynamic stability in an unpredictable manner (Rubingh, 1997
). The predictive engineering of recombinant antibody fragments therefore remains a challenge. Predictive engineering may be achievable by targeting positions playing a similar role in antibodies of different specificities, for instance because they belong to the common structural framework. In this study we have characterized the effect on antigen binding, expression level and thermostability of mutating position L34, which was selected (i) because of its non-random pattern of amino acid replacements in mouse germline sequence alignments, and (ii) because of its location at the variable light chain/variable heavy chain (VL/VH) interface and at the base of the paratope.
A covariance search on sequence alignments of mouse germline antibody variable domains (Choulier et al., 2000) indicated a non-random variability for position L34 (unpublished work). Two positions A and B are covariant if the nature of amino acids at position A is not independent of the nature of amino acids at position B. Their simultaneous replacement suggests that they are linked by functional constraints, in particular if the two positions are close together in the 3D structure of the protein (Altschuh et al., 1987). A mutational study of three pairs of spatially close covariant positions has shown that the distribution of surface charges influences the half-life of antibody fragments (Hugo et al., 2002). Residues that were found to covary with L34 are spatially distant in the 3D structure of antibody variable domains, which may indicate either a fortuitous correlation or a complex relationship between residues that cannot be unraveled by automated procedures (Mirny and Shakhnovich, 1999). The hypothesis of a functional role for residue L34 is, however, supported by the non-random distribution of amino acids. The method for covariance identification developed by Choulier et al. (Choulier et al., 2000) divides the sequences of a protein alignment into two groups, each group containing distinctive sets of amino acids at the covariant positions. In a mouse germline sequence alignment (Almagro et al., 1997, 1998), L34 displayed amino acids with small side chains (Ala, Ser, with Ser occurring only once) in one group and large polar side chains (His, Gln, Asp, Tyr, Glu) in the other, while apolar residues were not found. Thus, amino acids in the two groups possess different physico-chemical properties, whereas those within one group share similar properties, suggesting that allowed amino acids are restricted by functional constraints.
In addition to its non-random variability, the structural location of L34 makes it a potentially interesting target for the engineering of binding properties. L34 is part of the complementarity determining region (CDR) L1 (L24–L34) as defined by Kabat et al. (Kabat et al., 1991) on the basis of sequence variability, but not as defined by Chothia and co-workers (Chothia et al., 1989; Al-Lazikani et al., 1997) on the basis of structural variability. In the study by Chothia et al. (Chothia et al., 1998), L34 was defined as a VL/VH interface residue belonging to the common structural core and conserving a hydrophilic or neutral side chain, with a mean accessible surface area of 8 Å2. In an analysis of the contacts between antibody and antigen in 26 complex structures, L34 presented on average five contacts, as compared with 23 contacts for the position most involved in antigen contacts (MacCallum et al., 1996). In summary, the location of L34 at the VL/VH interface and therefore central to the paratope, but with only limited contacts with the antigen, suggested a possibly general, but not essential, contribution to paratope geometry and consequently to binding affinity, making it a possible target for modulating dissociation rates. The location and limited surface accessibility of L34 in antibody D2.3 (Charbonnier et al., 1997) (PDB code 1YEC) harboring Asn at this position are illustrated in Figure 1.
|
We analyzed the functional effects of mutating position L34 in two different recombinant antibody fragments. Fab57P (Chatellier et al., 1996
) was elicited against the coat protein of tobacco mosaic virus (TMVP) and recognizes peptides corresponding to amino acids 134–151 of the protein. SvFv1F4 was elicited against protein E6 of human papillomavirus 16 linked to glutathione S-transferase (GST) and recognizes a peptide corresponding to the N-terminal sequence of E6 (Giovane et al., 1999). The V
regions of both antibodies belong to the mouse kappa II subgroup. Fab57P and scFv1F4 possess Asn and Gln, respectively, at position L34 and therefore belong to the group of sequences with large, polar side chains in the covariance analysis. These amino acids were replaced by Ser, Ala and Thr in both antibody fragments. Ala and Ser L34 occur in the second group of germline sequences and are thus compatible with antibody function in their original sequence context. Thr was chosen because of its larger volume compared with Ala and Ser, making the replacement more conservative. In addition, five multiple mutants of scFv1F4 were constructed, containing Ser L34 in combination with one or two other changes, in order to investigate the effect of the Ser L34 in slightly different contexts. The functions that were measured are bacterial expression level, antigen binding kinetics and specificity and thermostability. The L34 Ser mutants were found to possess biochemical and binding properties that improve their potential usefulness as biotechnological tools. |
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Synthetic peptides
Peptides were obtained from Synthem (Montpellier, France). They were synthesized with a C-terminal cysteine for the E6 peptides (scFv1F4 system) and an N-terminal cysteine for the TMVP peptides (Fab57P system), to allow oriented chemical coupling on sensor chips for Biacore analysis. In the Fab57P system, the wild-type (WT) peptide antigen corresponds to residues 134–151 of the TMVP and is denoted C-WTTMVP. In the scFv1F4 system, the WT peptide antigen corresponds to residues 7–15 of the E6 protein and is denoted WTE6-C. Two substituted peptides derived from this WT sequence were used: Q13AE6-C and R15AE6-C.
Site-directed mutagenesis
Site-directed mutagenesis was performed using the PCR-based QuikChange site-directed mutagenesis kit as described by the manufacturer (Stratagene, La Jolla, CA, USA). Mutagenized plasmids were systematically verified by sequencing the coding region of the antibody fragments on both strands.
Periplasmic expression
Fab and scFv proteins were expressed in the periplasm of Escherichia coli using a protocol adapted from Rauffer-Bruyère et al. (Rauffer-Bruyère et al., 1997). Briefly, IPTG-induced expression was carried out at 30°C for 14 h. Harvested cells were equilibrated in spheroblast buffer (30 mM Tris, pH 8.0), 1 mM EDTA, 20% (w/v) saccharose, then osmotically shocked by diluting the cells twice in 0.5x periplasmic buffer supplemented with lysozyme (0.1 mg/ml). The volume of periplasmic extracts corresponded to one-tenth of the volume of E.coli cultures. Overnight dialysis at 4°C in HBS (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA) extemporally supplemented with 1 mM PMSF (phenylmethanesulfonyl fluoride) was followed by concentration using Centri-prep 10 devices (Amicon, Charlotte, NC, USA). Concentrated crude periplasmic extracts were kept at –20°C.
Biacore measurements
All experiments were performed at 25°C on a Biacore 2000 or 1000 instrument. All solutions and buffers were filtered through a 0.22 µm filter (Millipore) and running buffers were extensively degassed under vacuum. The running buffer was HBS containing 0.05% surfactant P20 (Biacore, Uppsala, Sweden). All surfaces were regenerated by injecting 10 µl of a 50 mM HCl solution.
Preparation of sensor surfaces for Biacore measurements
CM5 and B1 sensor chips were used for analysis of the Fab57P and scFv1F4 interactions, respectively. Small amounts of peptides were immobilized for kinetic analysis and large amounts were immobilized for measuring the active concentrations of antibody fragments. The thiol coupling chemistry (BIA Applications Handbook, Biacore) was used as described previously (Choulier et al., 1999; Ben Khalifa et al., 2000), using peptide solutions at 50–100 µg/ml in 10 mM formic acid, pH 3.2. The desired peptide level was achieved by varying the activation time of the matrix using a solution containing N-ethyl-N'-[3-(diethylamino)propyl]carbodiimide (EDC) and N-hydroxysuccinimide (NHS), and/or the peptide injection time. Both the activation and peptide injection times were between 5 and 10 min for preparing highly immobilized surfaces. The peptide levels were between 400 and 700 resonance units (RU). For surfaces with low concentrations of peptide, the activation times were 1 and 10 min on CM5 and B1 sensor chips, respectively, while peptide injection times were between 6 and 60 s (10 µl of peptide solution at a flow-rate of 100 and 10 µl/min, respectively). Between 6 and 20 RU of peptide were immobilized, in order to reach a maximum analyte response (RUmax) of 300 and 150 RU for Fab57P and scFv1F4, respectively, at saturating analyte concentrations. The TMVP (tobacco mosaic virus protein) was immobilized on a CM5 sensor chip by amine coupling chemistry, using TMVP obtained by acid degradation of the virus (Dubs et al., 1992), at a concentration of 100 µg/ml in 10 mM formic acid, pH 3.2. The TMVP surface was washed by repeated 1 min injections of 100 mM glycine–HCl, pH 2.0, until a stable baseline was reached. Blank surfaces were obtained in the same manner as kinetic surfaces except that no peptide was injected.
Determination of the concentrations of active antibody fragments
The active concentrations were determined as described previously for the Fab57P (Choulier et al., 1999) and scFv1F4 (Ben Khalifa et al., 2000) by using calibration curves. Dilutions of the periplasmic extracts containing the antibody fragments were injected at 10 µl/min on a surface where a high concentration of peptide had been immobilized (mass transport limitation) and the reaction rate was recorded for a 15 s interval at 17.5 (Fab57P) or 30 s (scFv1F4) after injection. Under these conditions, the reaction rate is directly proportional to the concentration of analyte (Karlsson et al., 1993).
Kinetic measurements
The antibody fragments in crude periplasmic extracts were diluted in running buffer and injected at five concentrations between 1 and 100 nM, at a flow-rate of 30 µl/min during 3 min, on a blank surface (flow cell 1) and two peptide surfaces (flow cells 2 and 3). Each analyte solution was also injected on a highly immobilized surface (flow cell 4) to measure its active concentration immediately before the kinetic run. Kinetic data were interpreted with the BIAevaluation 3.0 software (Biacore) using a Langmuir model of interaction, as described previously (Choulier et al., 1999; Ben Khalifa et al., 2000). Fab57P–NL34S was injected at concentrations up to 1000 nM.
Purification of mutant and WT antibody fragments
Purification of Fab57P-NL34S and Fab57P-WT was achieved by ion-exchange chromatography and affinity chromatography as described elsewhere (Gu et al., 2002). Immunoaffinity purification of scFvs 1F4-QL34S and 1F4-WT was performed as described previously (Giovane et al., 1999). Purified fragments were dialyzed in HBS buffer supplemented with 1 mM PMSF at 4°C and kept at –20°C.
Thermostability of Fab and scFv fragments
Purified Fab and scFv fragments were diluted in HBS buffer supplemented with the protease inhibitor cocktail Complete (Roche Molecular Biochemicals, Basel, Switzerland) to a concentration of 100–150 nM active antibody fragments. Samples of 150 µl were heated for 10 min at a given temperature and immediately stored on ice. After centrifugation (112 g, 5 min, 4°C), the remaining active concentration of the samples was determined on a Biacore 1000 upgrade instrument. Samples of 55 µl were injected at 10 µl/min on a high-density surface immobilized with WTE6-C (500 RUs) or C-WTTMVP (500 RU for the WT and 1500 RU for the NL34S mutant) for scFv1F4 and Fab57P molecules, respectively. The residual activity of heated samples was determined by reporting the slope of the resulting sensorgram (30 s after injection for scFv1F4 fragments, 17.5 s after injection for Fab57P-WT and 8 s after injection for Fab57P-NL34S) on their respective calibration curves. Relative residual activities were calculated as the difference between a reference sample incubated on ice (100% activity) and the samples heated to different temperatures and T50 values were determined as the temperature at which the antibody activity has fallen by 50%.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Expression and purification of antibody fragments
The various antibody fragments were expressed in BMH 71-18 E.coli cells. Concentrations of active antibody fragments were measured by Biacore. Although expression yields vary in different experiments, we found that antibody fragments containing mutation QL34S (scFv1F4) or NL34S (Fab57P) were consistently expressed at higher active levels than the parent WT molecule (Table I). A substantial increase in expression level compared with the WT was also observed for double mutants scFv1F4-Q34S-D70T and scFv1F4-Q34S-G44R. Fragments containing Ala L34 or Thr L34 were expressed at active levels similar to the WT in the 1F4 system, whereas no activity could be detected in the 57P system (Table I). Highest absolute expression yields were obtained for fragments scFv1F4-QL34S (1472 µg/l E.coli culture) and scFv1F4-QL34S-DL70T (1439 µg/l E.coli culture) while the scFv1F4-WT was maximally expressed at 261 µg/l E.coli culture.
|
Kinetic measurements
The kinetic parameters for the interaction of all antibody fragments with the WT peptide antigen were measured. Kinetic data were interpreted by the global fit procedure. Results are summarized in Table II, where data for the Ser L34 mutants are highlighted in bold. In both systems, the association rate parameter (kon) was only marginally affected by the mutations, with a ratio kon(mut)/kon(WT) in the range 0.6–1.3. The presence of Ser L34 resulted in 2.3- and 83-fold increases in dissociation rate parameter (koff) relative to the WT in the scFv1F4 and Fab57P systems, respectively, for the unpurified molecules. Similar results were observed with molecules purified by affinity chromatography, i.e. 2.4- and 57-fold increases in koff for the BCCP-fused (Sibler et al., 1999) scFvs and for the Fabs, respectively (Table II).
|
We tested the possibility of further increasing koff in the scFv1F4 system by combining the Ser mutation with other mutations that affected dissociation. For example, mutations DL60S and DL70T in the scFv1F4 system increased koff by
50 and 20% relative to the WT, respectively (Weidenhaupt et al., 2002). However, koff values determined for the double and triple mutants were similar to those observed for the single mutant scFv1F4-QL34S, with a ratio relative to the WT between 1.9 and 2.3 (Table II). ScFvs 1F4-QL34A and QL34T showed increases in koff of 1.2- and 1.5-fold, respectively. Typical sensorgrams corresponding to the injection of six concentrations of Fab57P-WT (3–90 nM) on a peptide surface are shown in Figure 2A. The fast dissociation between Fab57P-NL34S and the peptide antigen (Figure 2B) was close to the limit of measurements achievable by Biacore. A control experiment on an irrelevant peptide surface demonstrated the absence of non-specific binding of the mutant Fab (Figure 2C). The steady-state affinity, evaluated by plotting the equilibrium RU level against the concentration of injected Fab57P-NL34S (60–960 nM), was (6.6 ± 0.4)x106 M–1, which is in agreement with the apparent equilibrium affinity constant K'A of 6.1x106 M–1 deduced from the ratio kon/koff.
|
In order to verify whether the faster dissociation of Fab57P-NL34S compared with Fab57P-WT is also observed for the protein antigen, five concentrations of Fab57P-WT (60–3 nM) and Fab57P-NL34S (400–25 nM) were injected on a surface immobilized with 1000 RU of TMVP. The koff values calculated from the post-injection phase using a Langmuir model were (2.9 ± 0.4)x10–4 and (81.1 ± 1.0)x10–4 s–1 for the two Fabs, respectively, indicating a 28-fold faster dissociation of Fab57P-NL34S from the protein antigen compared with Fab57P-WT. These measurements also confirm that the Fabs dissociate faster from the peptide than from the protein (Choulier et al., 1999
), with a koff value 5–10 times larger. Binding specificity could not be evaluated for Fab57P-NL34S because its fast dissociation rate precludes precise kinetic measurements, but was evaluated for BCCP-scFvs 1F4-QL34S, QL34A and QL34T, using two peptide variants, Q13AE6-C and R15AE6-C, which are known to be recognized more weakly by scFv1F4-WT compared with peptide WTE6-C. The interaction of these scFvs with the peptide variants was too weak to be quantified, indicating a conserved ability to discriminate between related antigens, as illustrated in Figure 3 for the interaction of scFvs 1F4-WT, 1F4-QL34S and 1F4-QL34T with peptides WTE6-C and R15AE6-C.
|
Thermostability
Thermostability measurements were carried out with purified scFv1F4 and Fab57P fragments. The determination of a T50 value (Table III), corresponding to the temperature at which 50% of the binding activity of the antibody is lost, allowed us to show that the Ser mutation at position L34 increased the resistance to heat denaturation in both the Fab57P and scFv1F4 systems.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Our data show that the replacement of large polar side chains at position L34 by Ser in two different antibodies has three effects. Compared with the parent antibody fragment, the mutants showed increased expression levels of active molecules in E.coli, increased thermostability and increased dissociation rate parameters for the interaction with their cognate antigens, with an essentially unaffected association rate parameter. Such properties were not observed for the Ala L34 or Thr L34 mutants in either antibody. Although numerous papers have reported the influence of antibody mutations on binding activity, protein stability or expression level, only a few studies have compared the effects of substitutions on several different properties in more than one antibody.
Structure–function relationships
A structural explanation for the observed effects would require detailed investigations by X-ray crystallography or NMR and modeling. Position L34 is likely to play a complex role, by influencing the VL/VH packing, determining the conformation of CDR loops and in some cases directly contacting the antigen. The L34 Ser mutations have affected koff indicating a non-destructive alteration of antigen–antibody contacts, as expected, whereas kon, which mainly depends on diffusion and relative molecular charges, was similar for all interactions. The koff increase was much weaker in the 1F4 than the 57P system, but other factors such as buffer composition and temperature may be varied to accelerate dissociation further (Hugo et al., 2002) and achieve the desired kinetic rates. The effect on antigen binding of mutating position L34 has been reported in one case: the Gly L34 to Ser change in antibody 48G7 resulted in a 5-fold lower affinity for its hapten. From an examination of the crystal structure of the WT complex, this was attributed to the repulsive steric interaction between the Ser side chain and the hapten (Wedemayer et al., 1997).
The increased thermostability of the Ser mutants may be explained by a tighter VL/VH packing. Their increased level of expression may be related to increased stability, although the expression level is not always related to stability for periplasmic proteins (Wörn and Plückthun, 2001) (review). Other factors could influence the level of expression of active fragments, such as a decreased propensity of the scFvs to aggregate or to form wrong domain pairing. Stabilization of scFv fragments by VL/VH interface mutations has been achieved before by either increasing the strength of interface contacts (Zhu et al., 1997; Chowdhury et al., 1998; Tan et al., 1998) or adding inter-domain disulfide bridges (Glockshuber et al., 1990; Reiter et al., 1996; Rajagopal et al., 1997). In two of these cases (Chowdhury et al, 1998; Tan et al., 1998), the mutations have been shown both to decrease binding affinity and to increase expression level and/or stability, indicating that several different VL/VH positions can be engineered to improve such properties. However, in all cases, only one antibody fragment was analyzed.
Possible generality of the effects
General effects have rarely been reported in antibody mutational studies. Two examples of stabilizing framework mutations observed in two different antibodies were mentioned by Wörn and Plückthun (Wörn and Plückthun, 2001): the replacement of Lys by Arg at position H66 and the introduction of Pro at position L8 (normally Glu or Gln).
Our study suggests a possible generality for three properties following the introduction of Ser at position L34: increased thermostability, dissociation rate and expression level. The magnitude of the effects varied for the two antibody fragments, but the tendency was the same in both systems. We propose that these mutational effects may be general for antibodies possessing large polar side chains at position L34, which represent the majority of murine antibodies. The frequency of occurrence of such amino acids (Asn, Asp, Glu, Gln, Lys, Arg, His, Tyr), calculated using the Kabat Database of Sequences of Proteins of Immunological Interest (http://immuno.bme.nwu.edu/) (Johnson and Wu, 2001), is 53% for all kappa light chains (4569 entries), 72% for mouse kappa light chains (2612 entries) and 30% for human kappa light chains (1555 entries). The frequencies of occurrence of Ala are 37, 20 and 63%, respectively. Apolar side chains are rare, while the frequencies of occurrence of Ser and Thr are 7, 6 and 3% in the three groups of sequences, respectively.
A high level of expression, improved stability and increased dissociation rate are essential properties for antibodies to be used in affinity chromatography or for monitoring continuous processes. Our study may open the way to the predictive optimization of antibody fragments for various biotechnological applications.
|
Acknowledgements |
|---|
We thank Dr Annie-Paule Sibler for the gift of purified BCCP fused scFvs and Mr Christophe Guerin for the production of Fabs. This work was supported by grants from the Biotechnology Program of the European Community (contract number BIO4-CT98-0502), from the Association pour la Recherche sur le Cancer (contract number 5173) and from the Ministry of Defense (contracts 99 34 043/DSP/STTC and 20216 DSP/SREA/F).
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Al-Lazikani,B., Lesk,A.M. and Chothia,C. (1997) J. Mol. Biol., 273, 927–948.[CrossRef][Web of Science][Medline]
Almagro,J.C., Hernandez,I., Ramirez,M. and Vargas-Madrazo,E. (1997) Mol. Immunol., 34, 1199–1214.[CrossRef][Medline]
Almagro,J.C., Hernandez,I., Ramirez,M.C. and Vargas-Madrazo,E. (1998) Immunogenetics, 47, 355–363.[CrossRef][Medline]
Altschuh,D., Lesk,A.M., Bloomer,A.C. and Klug,A. (1987) J. Mol. Biol., 193, 693–707.[CrossRef][Web of Science][Medline]
Ben Khalifa,M., Weidenhaupt,M., Choulier,L., Chatellier,J., Rauffer-Bruyère,N., Altschuh,D. and Vernet,T. (2000) J. Mol. Recognit., 13, 127–139.[CrossRef][Medline]
Charbonnier,J.B., Golinelli-Pimpaneau,B., Gigant,B., Tawfik,D.S., Chap,R., Schindler,D.G., Kim,S.H., Green,B.S., Eshhar,Z. and Knossow,M. (1997) Science, 275, 1140–1142.
Chatellier,J., Rauffer,N., Van Regenmortel,M.H.V., Altschuh,D. and Weiss,E. (1996) J. Mol. Recognit., 9, 39–51.[CrossRef][Web of Science][Medline]
Chothia,C. et al. (1989) Nature, 342, 877–883.[CrossRef][Medline]
Chothia,C., Gelfand,I. and Kister,A. (1998) J. Mol. Biol., 278, 457–479.[CrossRef][Web of Science][Medline]
Choulier,L., Rauffer-Bruyère,N., Ben Khalifa,M., Martin,F., Vernet,T. and Altschuh,D. (1999) Biochemistry, 38, 3530–3537.[CrossRef][Medline]
Choulier,L., Lafont,V., Hugo,N. and Altschuh,D. (2000) Proteins, 241, 475–484.
Chowdhury,P.S., Vasmatzis,G., Beers,R., Lee,B. and Pastan,I. (1998) J. Mol. Biol., 281, 917–928.[CrossRef][Web of Science][Medline]
Dubs,M.C., Altschuh,D. and Van Regenmortel,M.H.V. (1992) J. Chromatogr., 597, 391–396.[CrossRef][Medline]
Giovane,C., Travé,G., Briones,A., Wasylyk,B. and Weiss,E. (1999) J. Mol. Recognit., 12, 141–152.[CrossRef][Web of Science][Medline]
Glockshuber,R., Malia,M., Pfitzinger,I. and Plückthun,A. (1990) Biochemistry, 29, 1362–1367.[CrossRef][Medline]
Gu,Z., Weidenhaupt,M., Ivanova,N., Pavlov,M., Xu,B., Su,Z.G. and Janson,J.C. (2002) Protein Express. Purif., 25, 174–179.[Medline]
Hugo,N., Lafont,V., Beukes,M. and Altschuh,D. (2002) Protein Sci., 11, 2697–2705.[CrossRef][Medline]
Johnson,G. and Wu,T.T. (2001) Nucleic Acids Res., 29, 205–206.
Kabat,E.A., Wu,T.T., Reid-Miller,M., Perry,H.M., Gottesman,K.S. and Foeller,C. (1991) Sequences of Proteins of Immunological Interest. 5th edn. US Department of Health and Human Services, National Institutes of Health, Washington, DC.
Karlsson,R., Fägerstam,L., Nilshans,H. and Persson,B. (1993) J. Immunol. Methods, 166, 75–84.[CrossRef][Medline]
MacCallum,R.M., Martin,A.C. and Thornton,J.M. (1996) J. Mol. Biol., 262, 732–745.[CrossRef][Web of Science][Medline]
Mirny,L.A. and Shakhnovich,E.I. (1999) J. Mol. Biol., 291, 177–196.[CrossRef][Web of Science][Medline]
Rajagopal,V., Pastan,I. and Kreitman,R.J. (1997) Protein Eng., 10, 1453–1459.
Rauffer-Bruyère,N., Chatellier,J., Weiss,E., Van Regenmortel,M.H.V. and Altschuh,D. (1997) Mol. Immunol., 34, 165–173.[CrossRef][Web of Science][Medline]
Reiter,Y., Brinkmann,U., Lee,B. and Pastan,I. (1996) Nature Biotechnol., 14, 1239–1245.[CrossRef][Web of Science][Medline]
Rubingh,D.N. (1997) Curr. Opin. Biotechnol., 8, 417–422.[CrossRef][Medline]
Sibler,A.P., Kempf,E., Glacet,A., Orfanoudakis,G., Bourel,D. and Weiss,E. (1999) J. Immunol. Methods, 224, 129–140.[Medline]
Tan,P.H., Sandmeier,B.M. and Stayton,P.S. (1998) Biophys. J., 75, 1473–1482.[Web of Science][Medline]
Wedemayer,G.J., Patten,.A., Wang,L.H., Schultz,P.G. and Stevens,R.C. (1997) Science, 276, 1665–1669.
Weidenhaupt,M., Ben Khalifa,M.B., Hugo,N., Choulier,L., Altschuh,D. and Vernet,T. (2002) J. Mol. Recognit., 15, 94–103.[Medline]
Wörn,A. and Plückthun,A. (2001) J. Mol. Biol., 305, 989–1010.[CrossRef][Web of Science][Medline]
Zhu,Z., Presta,L.G., Zapata,G. and Carter,P. (1997) Protein Sci., 6, 781–788.[Web of Science][Medline]
Received October 3, 2002; revised March 7, 2003; accepted April 4, 2003.
CiteULike 
Connotea 
Del.icio.us What's this?