PEDS Advance Access published online on September 30, 2008
Protein Engineering Design and Selection, doi:10.1093/protein/gzn053
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Antibody library selection by the β-lactamase protein fragment complementation assay
1Department of Medical Sciences, and Interdisciplinary Research Centre on Autoimmune Diseases (IRCAD), University of Eastern Piedmont, Via Solaroli 17, 28100 Novara 2Department of Biology, University of Trieste, Via L. Giorgieri 10, 34127 Trieste, Italy 3B Division, MS-M888, Los Alamos National Laboratory, Los Alamos, NM 87545, USA
4 To whom correspondence should be addressed. E-mail: daniele.sblattero{at}med.unipmn.it
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionFundingReferences |
|---|
Protein fragment complementation assay (PCA) is based on the interaction between two protein partners (e.g. target antigen and antibody), which are genetically fused to the two halves of a dissected marker protein. Binding of the two partners reassembles the marker protein and hence reconstitutes its activity. In this work we have developed the first application of β-lactamase-based PCA for the isolation of single chain Fv fragments (scFvs) binding to the human receptor RON from a naïve library. Specific scFvs with the ability to immunoprecipitate could be isolated after a single round of PCA selection from an scFv repertoire previously pre-selected by phage display. Furthermore, the PCA was used to successfully map the epitopes recognized by the selected scFvs by screening them against a small library of random RON fragments.
Keywords: β-lactamase/PCA/phage display/scFv
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingReferences |
|---|
Several hundred genome sequencing projects continue to identify thousands of genes encoding proteins. The functional annotation of the gene products is mainly based on sequence homology to already characterized proteins. However, a substantial portion of predicted gene products (30–50%) lack functional assignment because they have no or too little homology to known proteins. An alternative strategy to define the functions of new genes is to study how their products (proteins) interact with other known molecules involved in biochemical processes. Thus, significant efforts in modern biological research have been focused on developing approaches to identify and study these interactions, preferably in a high throughput format. One crucial requirement in the study of any new gene product is the availability of an antibody, or other affinity reagent. Antibodies allow the rapid determination of tissue, cellular and subcellular distribution, they facilitate the identification of interaction partners by immunoprecipitation, and they can also provide important functional information.
Although antibodies have been traditionally obtained by immunization, a new class of technologies, based on selection from large diverse libraries, and epitomized by phage display, has been recently developed. This promises to overtake the use of animals, and allow the generation of specific binders by selection on small amounts of purified protein. Although this field is advancing rapidly, with a few groups producing small amounts of proteins on a genomic scale, on the whole, sufficient quantities of selectors are not available for most proteome projects. One way around this is to avoid the use of physical selectors altogether and to develop genetic selection methods that use DNA encoding either whole, or a part, of the gene of interest. The selection of binders from libraries is essentially a protein–protein interaction problem, with the yeast two-hybrid system (Fields and Song, 1989
) being the most widely used genetic selection method to identify such interactions. Although this has been adapted to the selection of antibodies (Visintin et al., 1999; Visintin et al., 2002), the poor transfection efficiency of yeast, and the intrinsic instability of antibodies within intracellular compartments, make it unsuitable for high throughput genomic approaches. Bacterial two-hybrid systems (Karimova et al., 1998; Hays et al., 2000; Joung, 2001) are also available, and in addition to interaction systems relying on transcriptional activation, the so-called protein fragment complementation assays (PCA) have also been described (Pelletier et al., 1998; Remy and Michnick, 1999; Michnick et al., 2000; Michnick, 2001; Tafelmeyer et al., 2004; Cabantous and Waldo, 2006; Kerppola, 2006a,b; O’Hare et al., 2008). These rely on protein–protein interactions to reassemble a protein with a selectable phenotype from constituent fragments that are otherwise unable to assemble. In general, a reporter enzyme is split into two fragments that are fused to test protein pairs. If these test proteins bind to each other, the enzyme fragments are brought sufficiently close together that correct folding is initiated and the reconstituted protein can be detected or selected for by the reconstitution of enzyme activity. To date, several enzymes or fluorescent proteins have been used for PCA in prokaryote or eukaryote systems, including dihydrofolate reductase (DHFR) (Pelletier et al., 1998), adenylate cyclase (Fransen et al., 2002), green fluorescent protein (GFP) (Remy and Michnick, 2001; Hu and Kerppola, 2003; Cabantous et al., 2005; Magliery et al., 2005), luciferases (Paulmurugan et al., 2002; Paulmurugan and Gambhir, 2003; Remy and Michnick, 2006), TEV protease (Wehr et al., 2006), β-galactosidase (Rossi et al., 1997), ubiquitin (Johnsson and Varshavsky, 1994) Trp1p (Tafelmeyer et al., 2004) and TEM β-lactamase (Galarneau et al., 2002, Wehrman et al., 2002). Depending on reporter and assay format, the interactions can be monitored via cell survival on selective media or via colorimetric, luminescent or fluorescent read-outs. The PCA has been used to analyze biochemical (Michnick, 2003) and secretory (Nyfeler et al., 2005) pathways, drug targets (MacDonald et al., 2006) as well as protein–protein interactions in general (Michnick et al., 2000; Galarneau et al., 2002; Kerppola, 2006a,b). In addition to these analytical uses, PCA has also been used to select proteins or peptides with specific properties (Pelletier et al., 1999; Remy and Michnick, 1999; Arndt et al., 2000; Mossner et al., 2001; Amstutz et al., 2006; Koch et al., 2006). Murine DHFR is one of the most commonly used PCA scaffolds for selections, and is based on the finding that the murine DHFR is not inhibited by the antibiotic trimethoprim, in contrast to the bacterial enzyme, which allows interactions to be detected and selected for by colony formation (Pelletier et al., 1998). It was initially utilized to demonstrate that antigen–antibody interactions can be selected in model experiments using defined antigen–antibody pairs known to function intracellularly (Mossner et al., 2001). This was further developed by the selection of two antibody fragments (Koch et al., 2006), and subsequently ankyrin-based binders (Amstutz et al., 2006) from large libraries. In this format, the two halves of mDHFR are genetically fused to test proteins (antigen or specific binder) which are co-expressed in the cytoplasm of Escherichia coli (E.coli) cells. Although this study showed that the PCA was a suitable strategy for complex library screening, the efficiency in this specific example appeared to be hampered by inherent limits of the format. The reducing environment of the cytoplasm restricts the formation of disulfide bonds and interferes with the correct folding and stability of the expressed proteins, particularly antibody fragments. As a consequence, this format shows undesired biases such as the selection of low-affinity interactions and a high number of false positive colonies due to incorrectly folded proteins that accumulate as inclusion bodies and that are engaged in spurious and non-specific interactions. The success of this PCA application was significantly improved when a pre-selection using ribosome display was carried out (Amstutz et al., 2006). With the aim to develop a strategy for high throughput screening of complex libraries, we derived a PCA based on the ampicillin resistance protein, TEM-1 β-lactamase (Bla). Bla is a small monomeric protein (29 kDa) that hydrolyzes the β-lactam ring of most penicillins. It can be split into two inactive peptide fragments, namely the 
peptide (19 kDa) and the 
peptide (10 kDa), that are unable to refold spontaneously. Bla PCA has been so far used to select an fragment harboring a tripeptide sequence at the C-terminal that greatly improves the complementation system (Wehrman et al., 2002) or to set up model systems for fluorescence-based monitoring of Bla activity in living bacterial or mammalian cells (Wehrman et al., 2002; Nord et al., 2005).
Here we describe a procedure based on a Bla PCA for screening of antibody libraries in E.coli. Using as test proteins either a pair of characterized antigen/antibody molecules or complex libraries we exploit the system to demonstrate its suitability for simple, straightforward and high-throughput screening of protein–protein interactions. Although initially, we had hoped that it would be possible to carry out direct selection of specific antibodies from complex libraries using β-lactamase PCA, the results described later show that such direct selection suffers from problems not unlike those described earlier when DHFR is used in direct antibody selection. We were similarly able to overcome these problems using a phage display pre-selection step, rather than ribosome display. The results show that antibodies selected using this combined approach are more broadly useful than those selected by traditional phage display and ELISA screening, showing functionality in western blotting and immunoprecipitation, as well as ELISA.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingReferences |
|---|
Strains and cell lines
The E.coli strain DH5F' was used for DNA manipulations and for the PCA. The E.coli strain BL21-DE3 was used for the periplasmic expression of the single chain Fv fragments (scFvs) cloned in the pET20b(+) vector. NIHRon, a mouse embryonic fibroblast cell line stable transformed with the human Ron, used for the immunoprecipitation assays, was cultured in D-MEM medium supplemented with 10% fetal calf serum (Sigma), 2 mM glutamine and 100 U/ml penicillin, 100 U/ml streptomycin.
The vector pUC119 harboring the Bla coding sequence was used as template for a PCR generation of β-lactamse fragments. The -196 fragment corresponding to residues 196–286 was amplied with the primers Amp_Omega_sense AGCTGCTAGCGGCAAACCAATCCCAAACCCACTGCTGGGCCTGGATAGTACTCTACTTACTCTAGCTTCCCG and Amp_anti_EcoRI GCTAGAATTCTTACCAATGCTTAATCAGTG. The PCR fragment was cloned into the NheI and EcoRI restriction sites of the vector pPAO2 (Zacchi et al., 2003). The -195 fragment comprising residues 1–195 was amplified by three successive PCR reactions using a sense primer Alfa_amp_sense AGCTGCGCGCGCCAAATTCTATTTCAAGGAGACAGTCATAATGGGTATTCAACATTTCC and three different antisense primers called ALFA_AMP_Anti1 CCACCGCCACCGGAGCCACCTCCACCGACGTCGACAAATTCCCGCCCATTTTCGCCAGTTAATAGTTTGC, ALFA_AMP_Anti2 GGCCGCAAGCTTGTCGACGGAGCTCGAATTCGGATCCGCCACCGCCGCTGCCACCGCCACCGGAGCCACC, and ALFA_AMP_Anti3 GCGCGCTTAGCAGCCGGATCTCAGTGGTGGTGGTGGTGGTGCTCGAGTGCGGCCGCAAGCTTGTCGACGG. The resulting sequence was cloned as BssHII fragment into the vector pMPK1 (Mayer, 1995), a pBluescript II (Invitrogen) based vector where the ampicillin resistance encoding gene has been replaced by the gene encoding resistance to kanamycin. A point mutation was inserted at the position 182 M to T (Huang and Palzkill, 1997) in the fragment by site-directed mutagenesis using the QuickChange Site-Directed Mutagenesis Kit (Stratagene).
Cloning of Ron antigens domains
DNA manipulations were made according to the standard molecular biology techniques (Sambrook et al., 1989). All the Ron products were amplified from a pBSK-hRON, a plasmid carrying the full-length cDNA of the human macrophage stimulating 1 receptor (RON) (NM_002447
[GenBank]
.2 GI:153946392).
The Ron1 domain, the entire intracellular portion of human Ron, except for the 12 carboxyl-terminal arginine–proline rich residues (aas 984–1388), was PCR amplified with primer For1 (5'CGCGAATTCATGGGGAGGAAGCAGCTAG 3') and primer Back4 (5'GCAAGCTTGCGTACATTCCCTGGCATGG 3'). The Ron2 fragment (aas 984–1078) was PCR amplified with primer For1 and primer Back3 (5'GCAAGCTTCTCATGGGGAATCAGCACATC 3'). To obtain the Ron3 fragment (aas 1037–1388), PCR was done with primer For2 (5'CGCGAATTCGGAGCATCCTTCTCCGATAG 3') and primer Back4. The DNA fragments Ron1, Ron2 and Ron3 were cut with EcoRI and HindIII and inserted into the p vector in frame with a coding sequence for a six-histidine tag (6His-tag), to yield p-Ron1, p-Ron2 and p-Ron3 expression vectors. As negative Control, we used p-
G which expresses a domain of the V. Cholerae ZOT protein (GI:155314) named G (Di Pierro et al., 2001).
Cloning of scFv fragments in p vector
To obtain the p-sc7 plasmid, the nucleotide sequence of sc7 was PCR amplified from pHENsc7, a phagemid vector carrying the scFv, with the forward primer BssHII-sc7 (5'AGC GGC GCG CAT GCC CAG GTG CAG CTG CAG GAG TCG 3') and the reverse primer NheI-sc7 (5' GCC G CT AGC ACC TAG GAC GGT CAG CTT GGT 3'). The sc7 gene was cut with BssHII and NheI and ligated into p.
To obtain p-G2, the gene encoding for the previously isolated scFv G2 against G protein was cut from pDAN5 with BssHII and NheI and ligated into p.
Validation of PCA between human Ron e sc7
To test PCA between sc7 and Ron, DH5F' cells carrying the vector p-sc7 were co-transformed with p-Ron1/2/3 or p-G. After incubation for 1 h at 37°C (without selective pressure), ampicillin at the concentration of 30 µg/ml was added and the cells were grown for further 45 min. The cells were plated either for titering onto 2xTY agar plates supplemented with 50 µg/ml kanamycin and 34 µg/ml chloramphenicol, or for selection onto 2xTY agar plates supplemented with increasing concentrations of ampicillin (30–50–70–100 µg/ml), 34 µg/ml chloramphenicol and 1 mM IPTG. The plates were incubated at 28°C for 48 h.
PCA with a human naïve scFv library and isolation of the selected clones
To obtain phage pre-selected mini-libraries, a naïve phage scFv library (Sblattero and Bradbury, 2000) was used for two rounds of selection on GST–Ron1 fusion protein using immunotubes (Marks et al., 1991). The resulting repertoire of scFv genes was cut with BssHII and NheI and subcloned into the p vector. Electrocompetent DH5F'p-Ron1 were electroporated with the p-scFv filtered libraries and plated onto agar plates with 50 or 100 µg/ml ampicillin, 34 µg/ml chloramphenicol and 1 mM IPTG for the complementation assay, and onto 2xTY supplemented with 50 µg/ml kanamycin and 34 µg/ml chloramphenicol for titrating the electroporation efficiency. The plates were grown at 28°C for 48 h.
Sequences of the VH (heavy variable) and VL (light variable) domains of PCA-selected p-scFv clones were determined using the BigDye Terminator sequencing kit on an Applied Biosystems 377XL automated DNA sequencer (Foster City, CA, USA). Genes were aligned using the IMGT/VQUEST database (Giudicelli et al., 2006).
Periplasmic expression of the scFvs-
The -Ron bacterial clones, the PCA-selected scFv clones, and the negative Control G2 were grown in liquid broth to OD600 0.4, and the production of the recombinant protein was induced by overnight growth at 28°C in 20 ml of 2xTY supplemented with either 50 µg/ml kanamycin or 34 µg/ml chloramphenicol and 1 mM IPTG. To obtain periplasmic fractions, bacterial pellets were first incubated with hyperosmotic buffer, 1/20 vol (200 mg/ml sucrose, 1 mM EDTA, 30 mM Tris–HCl pH 8.0) for 30 min in ice, and further incubated with hyposmotic buffer, 1/20 vol (5 mM Mg2SO4) for 10 min in ice. Supernatants were collected and frozen at –20°C.
Expression of - and - fusion proteins was analyzed by western blotting using monoclonal antibodies to 6His-tag and to SV5 respectively, and alkaline phosphatase staining (Roche).
Analysis of scFv/Ron interaction by western blotting
Total protein extracts of DH5F' expressing GST or GST–Ron fusion proteins were separated by SDS–PAGE and transferred onto nitrocellulose membranes (Schleicher & Schuell). The filters were incubated with PCA-selected scFvs or G2 periplasmic fractions diluted 1:5 in PBS, 5% non-fat dry milk, 0.1% Triton X-100 and incubated overnight at 4°C. The immunocomplexes were detected by alkaline phosphate staining using an anti-SV5 (Hanke et al., 1992) monoclonal antibody, and an AP-conjugated anti-mouse antibody.
Analysis of scFv/Ron interaction with ELISA
Purified native GST–Ron, GST and G proteins were diluted in PBS to a concentration of 2 µg/ml and used to coat wells (100 µl/well) of a 96 microtiter plate (Maxisorp, Nunc) for 1 h at 37°C. Wells were saturated with blocking solution (PBS, 5% non-fat dry milk, 0.1% Triton X-100, 200 µl/well) for 1 h at 37°C. Periplasmic fractions, diluted in blocking solution (100 µl/well), were added to the wells and incubated for 1 h at 37°C. After four washes in PBS plus 0.1% Triton X-100, binding was detected with an anti-SV5 monoclonal antibody diluted 1:2500 in blocking solution (100 µl/well) for 1 h at 37°C, and after four washes in PBS plus 0.1% Triton X-100, with a horseradish peroxidase-conjugated anti-mouse antibody (Dako Cytomation) diluted 1:5000 in blocking solution (100 µl/well), incubated for 1 h at 37°C. After four washes with PBS plus 0.1% Triton X-100 and one with PBS, immunocomplexes were revealed with TMB substrate (Sigma) and the optical density was measured at 450 nm.
Soluble expression of the scFvs
The pET20b(+) vector (Novagen) was modified to insert a BssHII site downstream the sequence coding for the pectate lyase (pelB) leader and a NheI site upstream the coding sequence for the histidine tag. The resulting vector was used to subclone the scFv genes from selected p-scFvs. The recombinant constructs (pET–scFvs) encode carboxyl-terminal histidine tagged scFvs fused to the pelB leader peptide. Periplasmic recombinant proteins were obtained as described earlier.
Binding affinity ranking of scFv by competitive ELISA
A ranking of the affinity constant of the selected scFvs was determined according to Friguet et al. (1985). Briefly, affinity purified scFvs were incubated with different concentrations of purified GST–Ron. At the equilibrium state, the free antibodies were evaluated in ELISA, in wells coated with 2 µg/ml of purified GST–Ron. The value of the antigen in solution at the half-maximal signal of optical density at 450 nm was considered as the dissociation constant.
Surface plasmon resonance measurements
Surface plasmon resonance (SPR) measurements were done using a Biacore 2000 instrument. Briefly, 200 RUs of purified GST–Ron antigen were immobilized on CM5 sensor chip (Biacore) by standard amine-coupling and following ethanolamine deactivation procedures. For measurement of the kinetic constants, the purified 2.6 and sc7 scFvs were injected at concentrations of 100, 50, 25, 12.5, 6.25, 3.12 and 1.56 nM, at a flow rate of 10 µl/min. Association and dissociation phases were monitored for 1 and 5 min, respectively, and the regeneration consisted of two 20 s injection of 10 mM Gly, pH 3, at a flow rate of 10 µl/min. The analysis was performed for each scFv in two flow cells in parallel, one of them representing a blank spot used for referencing. Sensorgrams were fit to a 1:1 interaction model using the BIAevaluation software.
NIHRon cells were harvested from six plates (10 cm in diameter, 80% of confluence) and lysed in 1.8 ml of ice-cold DIM buffer (10 mM PIPES, 100 nM NaCl, 5 mM MgCl2, 300 mM sucrose, 5 mM EGTA, 10%Triton X-100, supplemented with 1 µg/ml leupeptin, 1 mM PMSF) for 20 min in ice. The lysate was clarified by centrifugation at 13 000 r.p.m. for 10 min at 4°C and incubated with Ni-NTA resin (Invitrogen) for pre-clearing.
For immunoprecipitation analysis, 30 µl of Ni-NTA bound scFvs were incubated with 250 µl of the pre-cleared lysate overnight at 4°C in rotation. As positive or negative Controls, equal amounts of lysates were incubated with protein A-sepharose bound anti-Ron antibody (C-20 Santa Cruz) or with Ni-NTA bound G2. After four washes in DIM buffer, immunocomplexes were detected by western blot using anti-Ron C-20 and ECL chemiluminescence (Amersham).
Ron fragment mini-library and epitope mapping
Ten microgram of Ron1 DNA fragment were sonicated (20 min pulses, at 100% power output, in ice), and fragments ranging between 200 and 400 bp were gel purified, blunted with T4 polymerase and ligated into the StuI site of p. DH5F' cells were transformed to yield a Ron fragment library of 104 cfu. Plasmid DNA from the library was used to transform DH5F' cells harboring either p-sc7, p-sc1.1, p-sc1.3, p-sc1.5 or p-sc2.6. Transformants were plated in presence of 70 or 100 µg/ml ampicillin, 34 µg/ml chloramphenicol and 1 mM IPTG, for PCA selection. The Ron fragments of selected clones were sequenced.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingReferences |
|---|
Construction of p
and p expression vectors and their validation for PCA
To exploit the Bla PCA as a suitable genetic tool for screening large antibody libraries, we started by verifying whether such a method could be used to select antigen–antibody interacting pairs in the periplasmic compartment. Two plasmids were constructed, p encoding the Bla fragment (aa 1–195, N-terminal) and p, encoding the fragment (aa 196–286, C-terminal) (Fig. 1A and B). Both plasmids are derived from the pBluescript II (Mayer, 1995) vector whose Bla gene has been substituted by the kanamycin (Kan) or the chloramphenicol (Cat) resistance genes in p or in p, respectively. In the p plasmid, the expression of chimeric genes is driven by the inducible lac-promoter, fusion proteins contain the fragment with the original TEM-1 signal peptide for periplasmic export, a flexible (Gly4-Ser)3 linker joining the complementing fragment to the test protein and a 6HIS-tag at the C-terminal. The fragment carries at the carboxyl terminal the residue substitution M182T (Huang and Palzkill, 1997), and the tripeptide NGR shown to enhance the activity and stability of the enzyme (Wehrman et al., 2002). In the p plasmid the PelB leader drives secretion of the scFv fragment into the periplasmic space. The scFv is followed by a SV5 tag (Hanke et al., 1992) and by the fragment (aa 196–286) of the Bla.
|
To determine whether antigen–antibody interactions could occur in the periplasm of E.coli cells, we used a pair of previously characterized interacting proteins: the intracellular portion of the human tyrosine-kinase receptor RON (Gaudino et al., 1994
), and an scFv (sc7) recognizing it selected from a naïve antibody phage library and previously shown to bind the juxtamembrane domain of RON with a Kd of 50 nM (Secco et al., 2004). This pair of proteins was genetically fused to the β-lactamase fragments in p and p plasmids respectively.
Three p-Ron plasmids were constructed: p-Ron1 consisting of the entire intracellular portion of the receptor (aa residues from 984 to 1388 referred to the full-length protein lacking for the 12 C-terminal arginine–proline rich tail), p-Ron2 from aa 984 to 1078, and p-Ron3 from aa 1037 to 1388 (Fig. 1C). The expression of the -Ron and -sc7 fusion products and their export to the periplasm were checked by western blot using anti-tag antibodies (Fig. 1D). As shown, all the proteins were correctly secreted with the smallest fragment, -Ron2, better exported into the periplasm, but partially degraded compared with -Ron1 and -Ron3.
E.coli cells were co-transformed with each p-Ron construct and p-sc7 or the p-G2 Control plasmid (carrying a scFv specific for the V. cholerae toxin G) and plated onto agar either containing increasing amounts of ampicillin or chloramphenicol/kanamycin to confirm the presence of both vectors. As shown in Fig. 2, the formation of ampicillin-resistant colonies was observed only in cells co-transformed by p-sc7 and p–Ron1 or p–Ron2 consistent with the presence of the juxtamembrane domain recognized by the scFv sc7. Interestingly, we observed that the number and dimension of colonies were inversely correlated to the amount of ampicillin, indicating that the sensitivity to antibiotic toxicity could be a simple means to select clones expressing more stable recombinant proteins among genetically homogenous cells. Furthermore, despite the presence in both proteins of the epitope target for -sc7, we observed that the colonies carrying -Ron1 were more sensitive to ampicillin than those with -Ron2. It is likely that this feature can be ascribed to differences in expression levels between the two proteins; the shorter being expressed, folded and/or exported more efficiently to the periplasmic space (Fig. 1D). No colonies were detected from cells either co-transformed with p-Ron3 and p-sc7 or with p-G2, even at low ampicillin concentrations. Moreover, no colonies were obtained from cells co-transformed with p-G and p-sc7, confirming the specificity of sc7.
|
Screening of phage-filtered mini-libraries by Bla PCA
Since the results obtained with the test pairs showed that Bla PCA could be used to select antigen–antibody interactions, we decided to evaluate whether the method could be used to screen complex libraries representing repertoires of human antibodies. Some published results (Koch et al., 2006) and our preliminary experiments (data not shown) demonstrated that the selection of antibodies directly from large libraries by PCA is impaired by limits that appear to be intrinsic to most complementation systems: the selective advantage given to small easily expressed fragments which are able to interact non-specifically and hence reconstitute enzyme activity, and the inability of antibodies known to be present in the library to provide a sufficiently strong selective advantage over these small fragments to be identified after selection. With this in mind, we decided to investigate whether phage display could be a means to filter out stable and folded scFvs scaffolds prior to PCA. A large naïve phage display library (Sblattero and Bradbury, 2000) was used for two rounds of selection on a purified GST–Ron1 fusion protein with a classical selection protocol using immunotubes (Marks et al., 1991). Colonies obtained after infection with phages eluted from both the first and second rounds of selection were collected, DNA extracted and scFv coding sequences were subcloned into the p vector to construct filtered mini-libraries 1 and 2. The DNA from both mini-libraries was used to transform E.coli cells carrying the p-Ron1 plasmid, and double transformants were plated in the presence of ampicillin of increasing concentration (50 and 100 µg/ml). The total number of double transformed colonies and selected colonies are represented in Table I.
|
Forty eight random colonies were then picked from each round of selection including the phage display selected colonies and from both plates with different ampicillin concentration. The specificity of the clones was tested by ELISA using periplasmic extracts probed against GST–Ron1 fusion protein, GST alone, and the protein
G (Di Pierro et al., 2001) as negative control. As can be seen in Table I, no positive colonies were selected from either the phage display or the PCA-derived clones obtained after a single round of selection. After the second round of selection, however, 24 of the 48 phage display clones examined were positive comprising two different scFvs, whereas after selection by PCA using the second round output, 40 of 48 clones were positive after selection on 100 µg/ml ampicillin and three different clones were identified. At lower ampicillin concentrations (50 µg/ml), the percentage of positive clones was slightly lower (34 out of 48) but the number of different clones (7), was significantly higher and their reactivity is reported in Fig. 3A. Interestingly these seven clones included all those selected on higher ampicillin concentrations, but not those isolated by phage display.
|
The specificity of the seven isolated scFvs clones was also tested in vivo. The p
-scFv vectors were separated from the p-antigen and transformed back in E.coli containing p-Ron1 or p-G as negative control. Bacteria were grown onto plates with kanamycin/chloramphenicol to confirm double transformation efficiency, and onto plates with ampicillin 50 µg/ml for the PCA assay. As can be seen in Fig. 3B bacterial growth was observed only as a result of the interaction between the scFvs and their cognate antigens.
The seven positive scFvs were sequenced to determine the V gene belonging. Sequence analysis and comparison with the human Ig gene sets [IMGT/VQUEST database (Giudicelli et al., 2006)], showed a restricted V gene use for both VL and VH families (Table II and supplementary File 1; Supplementary data are available at PEDS online) with a predominant usage of VL1 and VH3 gene families. It is noteworthy that the phage antibody sc7 previously selected on Ron, also carries VL1 and VH3 gene fragments (Secco et al., 2004). This finding suggests that this set of fragments might constitute a stable framework appropriate for specific Ron–antibody binding in the periplasmic compartment. In contrast, the D and J sequences, as well as the CDR3 sequences, did not show any significant bias.
|
Phage display filtering yields stable scFv scaffolds suitable for PCA selection
Since the scFvs used to perform the ELISA described earlier were indeed products of fusion with the β-lactamase fragment, it could be argued that the stability and solubility of the scFv proteins is due to fusion to the enzyme moiety rather than the phage display filtering. To examine this possibility, we subcloned some of the positive scFvs into a modified version of the pET20b(+) vector and expressed them as 6His-tagged scFvs in the periplasmic space of transformed E.coli cells, strain BL21-DE3. All scFvs were soluble and could be purified by affinity chromatography using Ni-NTA resin; the yield from a 1 l shake flask culture was between 2 and 5 mg, which is significantly higher than the yield usually obtained for scFvs selected from this library by phage display (0.1–2 mg). Furthermore, the scFvs selected by PCA appeared to show significantly less degradation than phage-selected scFvs (data not shown).
The biophysical and binding properties of these scFvs were further analyzed by using them as primary antibodies in ELISA tests where they showed similar positivity to those obtained as β-lactamase fusions, or in western blots on bacteria lysates containing GST–Ron1 or GST alone (Fig. 4). To investigate the ability of these recombinant antibodies to bind native human Ron, purified scFvs were incubated with cell lysates from NIH-Ron, a mouse fibroblast cell line that expresses the human receptor, and the immunocomplexes were analyzed by western blot (Fig. 5). All the scFvs were able to immunoprecipitate both the 170 kDa precursor and the 150 kDa mature form of Ron.
|
|
A ranking of the binding affinity of the selected clones was obtained by competitive ELISA, according to Friguet et al. (1985)
. As expected, all the binders showed affinities which could be estimated to be in the same medium to low nanomolar KD range. In particular, scFvs 2.6 and sc7, showing the highest and the lowest affinity respectively, were chosen for analysis by SPR. The kinetic constants and binding affinity, obtained by fit of a simple 1:1 interaction model, were determined to be: Ka (1/Ms) 4.3 x 105, Kd (1/s) 2.49 x 10–3, KD 5.78 nM for 2.6; and Ka 1.07 x 105, Kd 6.33 x 10–3, KD 59.1 nM for sc7. Consequently, according to the SPR analysis and competitive ELISA, the affinity of all the scFv clones ranged from 5 to 60 nM (supplementary Table I; Supplementary data are available at PEDS online). Our results corroborate previously reported data focusing on the importance that particular antibody scaffolds have in performing successful PCA. Here we demonstrated that the enrichment of such scaffolds can be accomplished by a single cycle of selection of a pre-selected phage antibody library.
As outlined earlier, the combined phage display/PCA-based approach presented here facilitates the isolation of antibodies suited for a wide range of immunological applications. Given that after such selection, scFvs are cloned within the p vector, we also explored the possibility that such selected scFvs could be directly used for epitope mapping by screening against a library of fragments derived from the original target protein gene cloned into p. To test this hypothesis, we amplified the DNA encoding the intracellular portion of human Ron by PCR and fragmented it by sonication. DNA fragments ranging 200–400 bp were repaired, gel purified and inserted into p to generate a Ron fragment mini-library of 104 individual clones. To check the library diversity, 25 clones were randomly picked and the DNA sequence of inserts was determined. DNA fragments covering the entire protein encoding sequence inserted in both orientations were identified (Fig. 6). The Ron fragment mini-library was used to transform E.coli cells carrying five different p-scFvs: 1.1, 1.3, 1.5 and 2.6 as well as the clone p-sc7 used as a positive control. Double transformants were plated onto agar plates containing kanamycin/chloramphenicol or 70 or 100 µg/ml of ampicillin. The ratio between the ampicillin and kanamycin/chloramphenicol-resistant clones was in the range of 10–5, indicating active selection. Several ampicillin-resistant clones were picked from each transformation and the p DNA inserts were sequenced. All selected p clones contained fragments that overlap the juxtamembrane region of Ron (Fig. 6). This finding is consistent with the data obtained from PCA selection of the scFvs, since all positive scFvs were selected by antigen containing the juxtamembrane domain (Ron1). Sequence comparison of the selected Ron fragments showed that two major epitopes exist in this region: one comprised between residues 983 and 1047 of the Ron membrane – proximal domain and recognized by the sc7, 1.1, 1.3 and 1.5 scFvs and a second between residues 1032 and 1107 of the same domain recognized by the 2.6 antibody.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingReferences |
|---|
Bacterial PCA is a procedure used to select interacting proteins that offers some advantages over more established methods. Compared with the yeast two-hybrid system, E.coli provides short duplication times and the ability to use highly complex cDNA libraries as a result of the high transformation efficiencies attainable. With respect to phage display the co-expression of antigens with library members during selection, and the theoretical possibility of having to carry out only a single cycle of selection to retrieve interactors, has the potential to significantly shorten the general time required for a selection procedure, as well as potentially avoid the need to express and purify proteins.
In this work we describe the use of PCA based on the Bla for library screening. The structural and functional properties of β-lactamase make it a good potential PCA reporter: it is monomeric, relatively small and its functional localization in the periplasmic space via the leader sequence is advantageous for proteins, such as antibody fragments, that require oxidizing environments for correct folding. The interactions are easily detected by bacterial growth of cells co-transformed with plasmids expressing interacting pairs. We have used p and p vectors that respectively allow the expression of an antigen fused to the fragment (aa 1–195) and of recombinant antibody fragments to the fragment (aa 196–286) of Bla. We first demonstrated the applicability of the system by analyzing the interaction between a previously characterized antigen–antibody pair (Secco et al., 2004): sc7, a scFv isolated from a phage display library, and the human receptor Ron. Using these proteins we confirmed that only bacteria containing both plasmids were able to grow on ampicillin plates and no growth was observed with non-interacting scFvs. We also noticed that the number and dimensions of surviving colonies reflects both the ampicillin concentration used for selection and the expression levels of the target antigen.
One common feature to the successful use of genetic selection methods, including PCA, to select high affinity binders from naïve libraries, is the use of a pre-selection step. In our case we used phage display, while ribosome display was used for the successful isolation of ankyrins (Rossi et al., 1997; Koch et al., 2006). Such pre-selection steps are likely to be effective for a number of reasons: (i) they reduce the total diversity while increasing the concentration of specific binders; (ii) they improve the expression and folding of pre-selected clones; (iii) they eliminate low affinity non-specific binders as a result of the stringency of the pre-selection; and (iv) as they involve recloning, small DNA fragments encoding polypeptides with polyreactive potential, will be eliminated. By carrying out this phage-based pre-selection step, we were able to increase the diversity of selected antibodies when compared with selection carried out by phage display alone. Selected scFvs were subcloned and expressed as recombinant proteins. All clones were soluble when expressed independently of the fragment showing that the fragment does not interfere with folding and that the system allows the isolation of functional recombinant antibodies. These were used as immunological reagents and all showed good binding characteristics against human Ron assayed in western blotting, ELISA, SPR analysis and immunoprecipitation. This last point is significant, since it is rare that antibodies isolated by phage display, unless selected directly on cell surfaces (Goenaga et al., 2007; Kurosawa et al., 2008) are used to recognize selecting antigens in immunoprecipitation (Marks et al., 1991; Vaughan et al., 1996; Sblattero and Bradbury, 2000; Schofield et al., 2007). The success we have observed is likely to be due to the application of two selection parameters (phage display and PCA) rather than a single one (phage display alone). Antibodies surviving both selection gates are more likely to recognize the native form of the target, as required for immunoprecipitation, rather than the denatured forms often found when antigens are adsorbed to plastic (Butler et al., 1997).
Although purified protein has not been required for selection, we and others have nevertheless used purified protein to subsequently confirm the specificity of selected antibodies (Fig. 3A). Given that far more antigen is consumed during screening than selection, this is an important qualification. However, the experiments illustrated in Fig. 3B show that protein production and purification can probably be abolished completely if a panel of p vectors expressing a number of different target proteins is prepared. By carrying out the PCA of selected antibodies against such a panel, it should be possible to rapidly identify those antibodies which bind specifically and those which are cross-reactive without the need to prepare protein. The only caveat of this approach is the possibility that such fusion proteins may not be correctly folded, although this is also true for proteins expressed and purified from E.coli.
The binding affinity of scFvs for purified RON was determined by SPR analysis and ELISA and compared with that of the parent sc7. The affinities of the selected scFvs ranged from 5 to 60 nM, which is generally somewhat better than the affinities routinely selected from naïve phage antibody libraries (Vaughan et al., 1996; de Haard et al., 1999, Sblattero and Bradbury, 2000). Sequence analysis shows that, although they are clearly different antibodies, the VH and VL families of the selected antibodies are all derived from those VH/VL combinations that have been previously shown to give single melting point transitions, and are consequently the most stable VH/VL combinations within the context of scFv structures (Ewert et al., 2003), suggesting that the combination of phage display pre-selection followed by PCA was able to select the most stable scFvs.
In the final set of experiments, we used PCA to identify the epitopes recognized by the selected scFvs. In general, methods used to identify epitopes fall into four broad classes: gene fragmentation and the selection of coding elements recognized by the antibody or other binding molecules by phage or yeast display (Di Niro et al., 2005; Kawamura et al., 2006; Levy et al., 2007); the selection of recognized epitopes from random peptide phage libraries (Smith, 1991; Balass et al., 1993; Kay et al., 1993); random mutagenesis to identify clones that have lost a specific epitope within phage or yeast display libraries (Stoop et al., 2000; Chao et al., 2004; Oliphant et al., 2005); and finally the recognition of linear peptide epitopes using overlapping synthetic peptides (Frank and Overwin, 1996; Frank, 2002). Gene fragmentation and random mutagenesis methods are the most effective, as they are able to identify both linear and conformational epitopes. We adapted the gene fragmentation method from display to PCA by expressing a library of Ron fragments in combination with each of the individual scFvs, and sequencing the Ron fragments of the surviving clones. Sc7, 1.3 and 1.5 all recognize the same epitope restricted to the first 30 amino acids of the membrane proximal region of Ron while 2.6 interacts with a different epitope found slightly downstream. This represents a significant new use for PCA, providing single step selection of clones displaying recognized epitopes which can be rapidly identified by sequencing, once the epitope library is created. This is especially true when the scFvs themselves are derived by PCA selection, and can be carried out without expression, purification or labeling of either antibody or recognized fragments. Although we created a random fragment library, targeted domain specific libraries can be just as easily made.
Although much work remains to be carried out to be able to use PCA for routine direct selection from naïve libraries, the work reported here and elsewhere (Mossner et al., 2001; Amstutz et al., 2006, Koch et al., 2006) indicate the power of this technique to select specific interacting partners when selection is carried out from relatively small libraries enriched in highly functional molecules. In the experiments described here, this is illustrated by the ease with which specific scFvs and the targets they recognize were isolated when either the scFv library was pre-selected by phage display, or a small library of protein fragments was used. In the experiments reported here, and in general, transformation is used to introduce plasmids encoding potential interacting partners into E.coli. This is relatively inefficient, requiring considerable work to produce cells sufficiently competent to assess the interaction between large libraries and potential targets. This is especially so, if a large library is to be assessed against a number of different targets, and could be considerably facilitated if a more efficient method could be used to introduce plasmid DNA into bacteria. We recently developed a method to generate pure populations of phagemid particles using bacterial packaging cell lines from any plasmid containing a filamentous phage packaging signal (Chasteen et al., 2006). Once a library has been transformed into these cells, unlimited quantities of phagemid particles can be easily and rapidly prepared. These provide a powerful alternative to transformation, reaching transduction efficiencies of 100% by infection, and should allow the more rapid and efficient application of this method.
The p and p vectors we used here carried different antibiotic resistance genes, but the same pUC origin of replication. This was based on our earlier work which showed that plasmids carrying identical origins of replication could co-exist for extended periods, even in the absence of any selective pressure (Velappan et al., 2007). By obviating the need to create plasmids with different origins of replication, the copy numbers of the plasmids encoding potential interacting targets are likely to be far more similar than when different origins are used. Given that the same promoter is also used for both plasmids, this will tend to result in more balanced expression levels of the interacting species, and hopefully more efficient selection, than when different origins and promoters are used, although as shown in Fig. 1D this does not overcome expression differences caused by the intrinsic differences between the genes being expressed.
In conclusion we have succeeded in selecting specific high affinity scFvs from large naïve libraries using a combination of phage display pre-selection and the PCA β-lactamase strategy. We have also demonstrated how this strategy can be rapidly used to identify epitopes recognized by selected scFvs, in a way which does not require any protein expression or purification.
|
Funding |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingReferences |
|---|
This work was supported with grant from Compagnia San Paolo (Torino) to C.S. Ricerca Sanitaria Applicata-CIPE Project to U.D. and EC Marie Curie Research Training Network, contract n. MRTN-CT-2006-036032 to RM.
|
Footnotes |
|---|
Edited by Daniel Tawfik
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionFundingReferences |
|---|
Amstutz P., Koch H., Binz H.K., Deuber S.A., Pluckthun A. Protein Eng. Des. Sel. (2006) 19:219–229.
Arndt K.M., Pelletier J.N., Muller K.M., Alber T., Michnick S.W., Pluckthun A. J. Mol. Biol. (2000) 295:627–639.[CrossRef][Web of Science][Medline]
Balass M., Heldman Y., Cabilly S., Givol D., Katchalski-Katzir E., Fuchs S. Proc. Natl Acad. Sci. USA (1993) 90:10638–10642.
Butler J.E., Navarro P., Sun J. Immunol. Invest. (1997) 26:39–54.[Web of Science][Medline]
Cabantous S., Terwilliger T.C., Waldo G.S. Nat. Biotechnol. (2005) 23:102–107.[CrossRef][Web of Science][Medline]
Cabantous S., Waldo G.S. Nat. Methods (2006) 3:845–854.[CrossRef][Web of Science][Medline]
Chao G., Cochran J.R., Wittrup K.D. J. Mol. Biol. (2004) 342:539–550.[CrossRef][Web of Science][Medline]
Chasteen L., Ayriss J., Pavlik P., Bradbury A.R. Nucleic Acids Res. (2006) 34:e145.
de Haard H.J., van Neer N., Reurs A., Hufton S.E., Roovers R.C., Henderikx P., de Bruine A.P., Arends J.W., Hoogenboom H.R. J. Biol. Chem. (1999) 274:18218–18230.
Di Niro R., Ferrara F., Not T., Bradbury A.R., Chirdo F., Marzari R., Sblattero D. Biochem. J. (2005) 388:889–894.[CrossRef][Web of Science][Medline]
Di Pierro M., Lu R., Uzzau S., Wang W., Margaretten K., Pazzani C., Maimone F., Fasano A. J. Biol. Chem. (2001) 276:19160–19165.
Ewert S., Huber T., Honegger A., Pluckthun A. J. Mol. Biol. (2003) 325:531–553.[CrossRef][Web of Science][Medline]
Fields S., Song O. Nature (1989) 340:245–246.[CrossRef][Web of Science][Medline]
Frank R. J. Immunol. Methods (2002) 267:13–26.[CrossRef][Web of Science][Medline]
Frank R., Overwin H. Methods Mol. Biol. (1996) 66:149–169.[Medline]
Fransen M., Brees C., Ghys K., Amery L., Mannaerts G.P., Ladant D., Van Veldhoven P.P. Mol. Cell. Proteomics (2002) 1:243–252.
Friguet B., Chaffotte A.F., Djavadi-Ohaniance L., Goldberg M.E. J. Immunol. Methods (1985) 77:305–319.[CrossRef][Web of Science][Medline]
Galarneau A., Primeau M., Trudeau L.E., Michnick S.W. Nat. Biotechnol. (2002) 20:619–622.[CrossRef][Web of Science][Medline]
Gaudino G., Follenzi A., Naldini L., Collesi C., Santoro M., Gallo K.A., Godowski P.J., Comoglio P.M. EMBO J. (1994) 13:3524–3532.[Web of Science][Medline]
Giudicelli V., Duroux P., Ginestoux C., Folch G., Jabado-Michaloud J., Chaume D., Lefranc M.P. Nucleic Acids Res. (2006) 34:D781–D784.
Goenaga A.L., et al. Mol. Immunol. (2007) 44:3777–3788.[CrossRef][Web of Science][Medline]
Hanke T., Szawlowski P., Randall R.E. J. Gen. Virol. (1992) 73:653–660.
Hays L.B., Chen Y.S., Hu J.C. Biotechniques (2000) 29:288–290. 292, 294 passim.[Web of Science][Medline]
Hu C.D., Kerppola T.K. Nat. Biotechnol. (2003) 21:539–545.[CrossRef][Web of Science][Medline]
Huang W., Palzkill T. Proceedings of the National Academy of Sciences of the United States of America (1997) 94:8801–8806.
Johnsson N., Varshavsky A. Proc. Natl Acad. Sci. USA (1994) 91:10340–10344.
Joung J.K. J. Cell Biochem. Suppl. (2001) Suppl. 37:53–57.[Medline]
Karimova G., Pidoux J., Ullmann A., Ladant D. Proc. Natl Acad. Sci. USA (1998) 95:5752–5756.
Kawamura M., et al. Biochem. Biophys. Res. Commun. (2006) 346:198–204.[CrossRef][Web of Science][Medline]
Kay B.K., Adey N.B., He Y.S., Manfredi J.P., Mataragnon A.H., Fowlkes D.M. Gene (1993) 128:59–65.[CrossRef][Web of Science][Medline]
Kerppola T.K. Nat. Methods (2006) a 3:969–971.[CrossRef][Web of Science][Medline]
Kerppola T.K. Nat. Rev. Mol. Cell Biol. (2006) b 7:449–456.[CrossRef][Web of Science][Medline]
Koch H., Grafe N., Schiess R., Pluckthun A. J. Mol. Biol. (2006) 357:427–441.[CrossRef][Web of Science][Medline]
Kurosawa G., et al. Proc. Natl Acad. Sci. USA (2008) 105:7287–7292.
Levy R., Forsyth C.M., LaPorte S.L., Geren I.N., Smith L.A., Marks J.D. J. Mol. Biol. (2007) 365:196–210.[CrossRef][Web of Science][Medline]
MacDonald M.L., et al. Nat. Chem. Biol. (2006) 2:329–337.[CrossRef][Web of Science][Medline]
Magliery T.J., Wilson C.G., Pan W., Mishler D., Ghosh I., Hamilton A.D., Regan L. J. Am. Chem. Soc. (2005) 127:146–157.[CrossRef][Web of Science][Medline]
Marks J.D., Hoogenboom H.R., Bonnert T.P., McCafferty J., Griffiths A.D., Winter G. J. Mol. Biol. (1991) 222:581–597.[CrossRef][Web of Science][Medline]
Mayer M.P. Gene (1995) 163:41–46.[CrossRef][Web of Science][Medline]
Michnick S.W. Curr. Opin. Struct. Biol. (2001) 11:472–477.[CrossRef][Web of Science][Medline]
Michnick S.W. Curr. Opin. Biotechnol. (2003) 14:610–617.[CrossRef][Medline]
Michnick S.W., Remy I., Campbell-Valois F.X., Vallee-Belisle A., Pelletier J.N. Methods Enzymol. (2000) 328:208–230.[Web of Science][Medline]
Mossner E., Koch H., Pluckthun A. J. Mol. Biol. (2001) 308:115–122.[CrossRef][Web of Science][Medline]
Nord O., Gustrin A., Nygren P.A. FEMS Microbiol. Lett. (2005) 242:73–79.[CrossRef][Web of Science][Medline]
Nyfeler B., Michnick S.W., Hauri H.P. Proc. Natl Acad. Sci. USA (2005) 102:6350–6355.
O’Hare H., Juillerat A., Dianiskova P., Johnsson K. J. Microbiol. Methods (2008) 73:79–84.[CrossRef][Web of Science][Medline]
Oliphant T., et al. Nat. Med. (2005) 11:522–530.[CrossRef][Web of Science][Medline]
Paulmurugan R., Gambhir S.S. Anal. Chem. (2003) 75:1584–1589.[Medline]
Paulmurugan R., Umezawa Y., Gambhir S.S. Proc. Natl Acad. Sci. USA (2002) 99:15608–15613.
Pelletier J.N., Campbell-Valois F.X., Michnick S.W. Proc. Natl Acad. Sci. USA (1998) 95:12141–12146.
Pelletier J.N., Arndt K.M., Pluckthun A., Michnick S.W. Nat. Biotechnol. (1999) 17:683–690.[CrossRef][Web of Science][Medline]
Remy I., Michnick S.W. Proc. Natl Acad. Sci. USA (1999) 96:5394–5399.
Remy I., Michnick S.W. Proc. Natl Acad. Sci. USA (2001) 98:7678–7683.
Remy I., Michnick S.W. Nat. Methods (2006) 3:977–979.[CrossRef][Web of Science][Medline]
Rossi F., Charlton C.A., Blau H.M. Proc. Natl Acad. Sci. USA (1997) 94:8405–8410.
Sambrook J., Fritsch E.F., Maniatis T. Molecular Cloning: A Laboratory Manual (1989) 2nd edn. Cold Spring Harbour Laboratory Press.
Sblattero D., Bradbury A. Nat. Biotechnol. (2000) 18:75–80.[CrossRef][Web of Science][Medline]
Schofield D.J., et al. Genome Biol. (2007) 8:R254.[CrossRef][Medline]
Secco P., Ferretti M., Gioia D., Cesaro P., Bozzo C., Marks J.D., Santoro C. J. Immunol. Methods (2004) 285:99–109.[CrossRef][Web of Science][Medline]
Smith G.P. Curr. Opin. Biotechnol. (1991) 2:668–673.[CrossRef][Medline]
Stoop A.A., Jespers L., Lasters I., Eldering E., Pannekoek H. J. Mol. Biol. (2000) 301:1135–1147.[CrossRef][Web of Science][Medline]
Tafelmeyer P., Johnsson N., Johnsson K. Chem. Biol. (2004) 11:681–689.[Web of Science][Medline]
Vaughan T.J., Williams A.J., Pritchard K., Osbourn J.K., Pope A.R., Earnshaw J.C., McCafferty J., Hodits R.A., Wilton J., Johnson K.S. Nat. Biotechnol. (1996) 14:309–314.[CrossRef][Web of Science][Medline]
Velappan N., Sblattero D., Chasteen L., Pavlik P., Bradbury A.R. Protein Eng. Des. Sel. (2007) 20:309–313.
Visintin M., Tse E., Axelson H., Rabbitts T.H., Cattaneo A. Proc. Natl Acad. Sci. USA (1999) 96:11723–11728.
Visintin M., Settanni G., Maritan A., Graziosi S., Marks J.D., Cattaneo A. J. Mol. Biol. (2002) 317:73–83.[CrossRef][Web of Science][Medline]
Wehr M.C., Laage R., Bolz U., Fischer T.M., Grunewald S., Scheek S., Bach A., Nave K.A., Rossner M.J. Nat. Methods (2006) 3:985–993.[CrossRef][Web of Science][Medline]
Wehrman T., Kleaveland B., Her J.H., Balint R.F., Blau H.M. Proc. Natl Acad. Sci. USA (2002) 99:3469–3474.
Zacchi P., Sblattero D., Florian F., Marzari R., Bradbury A.R. Genome Res. (2003) 13:980–990.
Received August 29, 2008; revised August 29, 2008; accepted September 5, 2008.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  D. R. Bowley, T. M. Jones, D. R. Burton, and R. A. Lerner Libraries against libraries for combinatorial selection of replicating antigen-antibody pairs PNAS, February 3, 2009; 106(5): 1380 - 1385. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||