WO2013115425A1 - Methods for screening an antibody by one-cycle panning - Google Patents

Abstract

The present invention relates to a method for screening an antibody specifically binding to a target antigen and a method for producing a polypeptide specifically binding to a target polypeptide. This method could be applied to select an antibody or polypeptide from a large collection of homologous or heterogeneous molecules that are contained in combinatorial libraries.

Description

METHODS FOR SCREENING AN ANTIBODY BY ONE-CYCLE PANNING

The present invention relates to a method for screening an antibody specifically binding to a target antigen and a method for producing a polypeptide specifically binding with a target polypeptide.

An important goal of modern molecular biology is to link the genotypic information that comes from nucleic acid sequencing to the phenotype of normal and diseased organisms. Here, the decoded nucleic acid information is directly linked to phenotype, because, in evolutionary terms, this is an endpoint analysis, and one assumes a high degree of fitness for any deduced protein. In other words, the information obtained from nucleic acid sequencing already relates to a highly selected phenotype of a functioning organism. Although such studies are proceeding rapidly for many genomes, deep sequencing has not been systematically used for the analyses of very large libraries of molecules that are related by a sequence, such as antibodies. This task is very different than the study of the genome of an organism, because, for a given phenotype such as binding energy, one has a very large collection of related molecules that represent a spectrum of affinities. Furthermore, it is this very property of relatedness that makes the process of selective recovery of rare sequences from the larger collective so difficult (Edelman G, Gally JA (2001) Degeneracy and Complexity in Biological Systems. Proc Natl Acad Sci USA 98:13763-13768). These problems become more acute when one wishes to select antibodies against targets that are components of complex mixtures such as cell surfaces.

Although phage systems have the potential to link phenotype to genotype, their analysis is based on the process of selection which, when successful, necessarily results in the loss of information. For example, if molecules that bind less tightly but are otherwise related are discarded, one loses information concerning the evolutionary trajectory toward fitness. By contrast, nucleic acid sequencing of libraries only gives information that without selection is, in the main, unrelated to phenotype.

In other words, in the classical screening of combinatorial antibody libraries in phage, multiple rounds of selection are used such that members of the library with increased fitness for binding are increased at each round with the concomitant loss of less-fit molecules. Even though the trajectory toward fitness is not transparent, the process has been highly successful for purified antigens and many important experimental and therapeutic antibodies have been obtained. Although these methods are suitable in many cases, in certain circumstances the molecules that are lost may be more important than those that are selected. For example, if one wants an antibody that neutralizes a virus, those that have the highest affinity may be irrelevant because they bind to a region of the virus that is not a target for neutralization. Furthermore, in modern immunochemistry there are many ways to improve the binding energy of molecules that initially have lower affinities. Because binding energy comes from a concert of interactions, distant amino acid changes that were discarded along the evolutionary trajectory in favor of those at other positions that gave superior binding may be reinstated to give additive effects. Finally, because conventional procedures require many cycles of phage replication, irrelevant factors such as growth advantages may bias the selection.

In this regard, the present inventors have made an effort to develop an highly efficient and simple method for screening an antibody, and were able to select an antibody to a preselected epitope in the presence of a large number of irrelevant proteins on the E. coli surface. Namely, one begins with a large library of different proteins expressed in phage where only a very minor fraction of the members of the library exhibit a desired phenotype such as the binding to another protein or the surface of a cell. After an initial selection, the both noise and signal are existed. Next, the information content of the phenotypic pool is established by deep sequencing of its members. The nucleic acids encoding the molecules considered to be signal molecules are recovered by hybridization and converted back to phenotype by transformation of Escherichia coli, thereby completing the phenotype-information-phenotype-cycle.

It is an object of the present invention to provide a method for screening an antibody specifically binding to a target antigen.

It is another object of the present invention to provide a method for producing a polypeptide specifically binding to a target polypeptide.

The method of the present invention can be generalized for selection of antibodies against targets that are present as minor components of complex systems.

Fig. 1 shows the phenotype-information-phenotype cycle. The cycle starts with selections from a combinatorial antibody library displayed on the surface of phage. The phage library is selected against two populations of bacteria that either display the antigen or serve as control cells (phage specific for antigen are in the minority and are colored grey and other phage are indicated in black). After selection, bound phage are eluted and their phagemids (circles) are analyzed by deep sequencing. The sequencing results from paired samples were compared by a bioinformatics analysis and DNA representing sequences overrepresented in phage that bind to the antigen presenting E. coli are extracted. A biotinylated probe whose design is based on the VH CDR3 of interest is synthesized and hybridized in solution with singlestranded circular DNA isolated from phage particles. The ssDNA selected by hybridization is captured on magnetic streptavidin beads and released by heating. The ssDNA is converted to dsDNA before transformation of suitable host E. coli cells.

Fig. 2 shows antigen display on bacterial surfaces. Structural models (A, Left) the Lpp leader sequence and first nine amino acids of the E. coli major outer membrane lipoprotein (OmpA) were used to attach a variety of proteins to the E. coli surface. In this system, the posttranslational tripalmitoyl-S-glycerylcysteine component of the lipoprotein anchors the complex by inserting into the E. coli surface membrane. IL-6 was fused to the C terminus of OmpA (amino acids 46-159) and displayed on the surface of E. coli cells (A, Right). Alternatively, IL-6 was fused to the C terminus of the MBP and the fusion product was anchored to the cell wall by linking it to the C terminus of OmpA. The view was generated using PyMOL. Protein Data Bank codes are 1ALU for IL-6, 3PGF for MBP, and 1QJP for OmpA. The IL-6 is represented as a black ribbon diagram; MBP is rendered in magenta; and OmpA is colored light grey. The grey and black slabs represent the exoplasmic and cytoplasmic surface, respectively. (B) Expression of Lpp-OmpA-cytokine fusions. Cytokines were fused directly to the C terminus of OmpA. After induction at 22 ℃ for 3 h, the FLAG tag at the C terminus of fusion protein was detected by HRP-conjugated anti-FLAG antibody (black) and compared to a control antibody of the same isotype (grey). (C) Comparison of cytokine display level in the presence or absence of the MBP. IL-1, chemokine (C-C motif) ligand 28 (CCL28), and gp41 antigens were displayed in duplicate, either fused directly to OmpA or fused to MBP-OmpA. The cytokines on the cell surface were detected by cell-ELISA with HRP conjugated to anti-FLAG antibody (blak and white) and compared to a control antibody of the same isotype (light grey and dark grey). (D) Detection of displayed cytokines by flow cytometry After induction, cells were stained with phycoerythrin (PE)-labeled cytokine specific antibodies for 1h, washed, and 10,000 E. coli events were analyzed with a Becton Dickinson flow cytometer. Each histogram is represented in pairs representing the specific versus control antibodies as indicated in the legend in the figure.

Fig. 3 shows schematic illustration of the three approaches used for scFv recovery based on the VH CDR3 DNA sequences. (A) Overlap PCR. The specific scFv were recovered by overlap PCR using primers (grey and white arrows) complementary to the VH CDR3 sequence as well as to vector sequences using phagemid pools as templates. The recovered scFv was ligated into the phagemid vector. After transformation into XL1-blue cells, the antibody genes present in randomly picked colonies were characterized by Sanger sequencing. (B) Rolling circle. Complementary primers based on selected VH CDR3 sequences were annealed to denatured phagemids followed by rolling circle amplification. Methylated and hemimethylated template DNA was removed by digestion with DpnI, leaving only the newly synthesized molecules which were then used to transform XL1-blue cells. The antibody genes present in the recovered colonies were characterized by Sanger sequencing. (C) Hybridization. The phage single-stranded phage DNA was recovered by hybridization to a probe selected for its specificity against the scFv antibody genes of interest. The ssDNA recovered by an affinity step using a biotinylated probe was converted to dsDNA prior to transformation of E. coli cells.

Fig. 4 shows selection of IL-1RA antibodies from a spiked-in library. Phage encoding the IL-1RA binding scFv (H9) were spiked into a ScFv naive combinatorial antibody library containing 3.0 x 10⁹ members at a ratio of one H9 encoding phage to 10⁹ irrelevant phage. Two rounds of selection were carried out using a subtractive panning format in which at each round phage were first incubated with control bacteria and those that did not bind were next selected on bacteria displaying the IL-1RA antigen. The PCR-amplified VH repertoires from phage that bound to the paired samples at each round were sequenced using Roche 454 pyrosequencing. The VH CDR3s were identified and the frequency of each unique VH CDR3 was determined for the paired samples from the same round of selection (control versus experimental) to obtain a ratio of frequencies. In addition to the spiked-in H9 clone, we noted selective appearance of another clone (H1). (A) All CDR3s that had greater than 10 reads in at least one sample (ca. 0.3%) were ordered by similarity. The phylogenetic tree was constructed based on multiple sequence alignments via MAFFT and ClustalW2 tree generation methods (Left). The log2-fold change in frequency of the first round (black) and the second round (grey), and Z score of the first round (white) and the second round (light grey) are shown (Right). Arrows indicate clones H1 and H9 (Right). (B) The sequences with highest differential ratio were recovered with overlap PCR and corresponding phages were prepared. The degree of binding to IL-1RA was tested by phage-ELISA using different concentrations of three different phage clones (H1, H9, and irrelevant). Phage were added to IL-1RA or BSA-coated wells. After washing, the bound phage were detected with HRP-conjugated phage antibody. (C) The scFvs were converted to full-length IgG that was produced in HEK293F cells and purified by protein G chromatography. Different concentrations of IgG were added to IL-1RA using BSA as a control protein. The bound antibody was detected with an HRP-conjugated antibody to the human IgG.

Fig. 5 shows epitope-directed antibody discovery. (a) Structure of IL-6. The structural representation was generated by PyMol. The Protein Data Bank code is 1ALU. The structure on the left represents IL-6 with amino acids 1-27 truncated (Δ1-27). The cartoon on the right representsWT IL-6. The N-terminal 1-27 amino acids are highlighted in black and shown as ribbon diagrams. (b) Detection of wild-type and truncated IL-6 displayed on bacterial surfaces by flow cytometry. Wild-type and truncated IL-6 (Δ1-27) were displayed on the bacterial cell surface by fusing them to the C terminus of OmpA (Fig. 2). The induced bacteria were stained separately with two phycoerythrin (PE)-labeled mAbs. Monoclonal antibody 1 (R&D, catalog number IC206P) recognizes both the full-length and truncated version of IL-6, whereas mAb 2 (Biolegend, catalog number 501106) only recognizes the full-length version of the molecule. Irrelevant antibodies of the same isotype were used for control studies. After 1 h incubation at 24 ℃, bacteria were washed and analyzed by flow cytometry. (c) Phylogenetic tree of CDR3 sequences from pyrosequencing. The PCR-amplified VH encoding DNA from phage bound to bacteria displaying WT or truncated IL-6 (Δ1-27) from the second round of selection was subjected to pyrosequencing. After a bioinformatics analysis, all CDR3 sequences observed greater than two times in at least one sample (ca. 0.1% frequency) were ordered by similarity. The phylogenetic tree on the left was constructed based on multiple sequence alignment via MAFFT and ClustalW2 tree generation methods. The log2-fold change in frequency (black) and Z score (grey) are shown on the right. (d) Binding of scFv to full-length and truncated IL-6 in phage-ELISA. Different concentrations of scFv displayed on phage were added to plates coated either with full-length or truncated human IL-6 (hIL-6) or BSA. The truncated IL-6 was prepared from HEK293F cells transfected with the truncated IL-6 overexpression vector and was purified on an anti-FLAGtag column. After 1 h incubation at 37 ℃, the wells were washed five times with PBS, after which HRP-conjugated antiphage antibody was added. Incubation was for 1 h at 37 ℃. After five washes with PBS, 2,2-azinobis(3-ethylbenzthiazoline-6-sulfonic acid) was added and the OD at 405 nm was measured. (e) Competitive ELISA to validate the epitope to which the selected antibody bound. ScFv protein N27-1 was purified fromthe bacterial culture supernatant and periplasmic space with a anti-FLAG antibody column. Increasing concentrations of the commercial IL-6 binding antibody (R&D, mAb2061), which we determined bound to the N terminus of the protein (Fig. 5b), weremixedwith a constant concentration (1 ㎍/mL) of the selected scFv N27-1 and themixtureswere added to wells coated with recombinant hIL-6 (black line). An irrelevant mAb against IL-1RA was used as control (grey line). HRP-conjugated anti-FLAG antibody was added to detect the scFv. ScFv binding to BSA is shown as a light grey dashed line. The curves represent the concentration-dependent competitive inhibition of the signal generated by scFv binding by either the commercial mAb that bound to the N terminus of IL-6 (black) or an irrelevant antibody (grey).

In accordance with one aspect, the present invention relates to a one-cycle method for screening an antibody specifically binding to a target antigen, comprising following steps: (a) providing an antibody library of different proteins displayed on the surface of phages; (b) contacting the antibody library with (i) a group of cells which displays the target antigen (test group) and (ii) another group of cells which does not displays target antigen or displays less amounts of target antigen (control group); (c) removing the unbound phages and sequencing nucleotide of the bound phages; (d) comparing nucleotide sequences of the test group and control group; and (e) selecting nucleotide sequences the test group which are not common with the control group and identifying these sequences as sequences encoding the antibody specifically binding to the target antigen.

Preferably, the method for screening an antibody specificially binding with a target antigen may further comprise (f) extracting single-stranded DNA from phage particles and hybridizing to a tagged probe; and (g) converting the single-stranded DNA to double stranded DNA.

Preferably, the step (c) may be accomplished by deep sequencing.

The term "deep sequencing" may refer to the "depth of coverage", meaning how many times a single base is read during the sequencing process. Deep sequencing implies that you are sequencing to a depth that allows each base to be read hundreds of times. This allows identification of very rare sequence variants (mutations). Namely, deep sequencing may refer to the coverage, or depth, of the process that is many times larger than the length of the sequence under study. The deep sequencing method may be pyrosequencing but it is not limited thereto.

Preferably, the cells may be selected from the group consisting of bacteria, yeast and eukaryotic cells but it is not limited thereto.

Preferably, the antibody library may comprise single chain variable region fragments (scFvs) or single domain antibodies (dAbs) but it is not limited thereto.

Preferably, the antibody library may comprise a library of human antibodies but it is not limited thereto.

Preferably, the target antigen of step (b) may be an arfificial polypeptide or natural polypeptide but it is not limited thereto.

The term "antibody" refers to a polypeptide chain(s) which exhibits a strong monovalent, bivalent or polyvalent binding to a given antigen, epitope or epitopes. Antibodies used in the invention can have sequences derived from any vertebrate, camelid, avian or pisces species. They can be generated using any suitable technology, e.g., hybridoma technology, ribosome display, phage display, gene shuffling libraries, semi-synthetic or fully synthetic libraries or combinations thereof. The term "antibody" as used in the present invention includes intact antibodies, antigen-binding polypeptide fragments and other designer antibodies that are described below or well known in the art (see, e.g., Serafini. J Nucl. Med. 34:533-6,1993).

Antibodies to be used in the invention also include antibody fragments or antigen-binding fragments which contain the antigen-binding portions of an intact antibody that retain capacity to bind the cognate antigen. Examples of such antibody fragments include (i) a Fab fragment, a monovalent fragment consisting of the V_L, V_H, C_L and C_H1 domains; (ii) a F(ab')₂ fragment, a bivalent fragment comprising two Fab fragments linked by a disulfide bridge at the hinge region: (iii) a Fd fragment consisting of the V_H and C_H1 domains; (iv) an Fv fragment consisting of the V_L and V_H domains of a single arm of an intact antibody; (v) disulfide stabilized Fvs (dsFvs) which have an interchain disulfide bond engineered between structurally conserved framework regions; (vi) a single domain antibody (dAb) which consists ofa V_H domain (see, e.g., Ward et aI., Nature 341:544-546. 1989); and (vii) an isolated complementarity determining region (CDR).

Antibodies suitable for practicing the present invention also encompass single chain antibodies. The term "single chain antibody" refers to a polypeptide comprising a V_H domain and a V_L domain in polypeptide linkage, generally linked via a spacer peptide, and which may comprise additional domains or amino acid sequences at the amino- and/or carboxyl-termini. For example, a single-chain antibody may comprise a tether segment for linking to the encoding polynucleotide. As an example, a single chain variable region fragment (scFv) is a single-chain antibody. Compared to the V_L and V_H domains of the Fv fragment which arc coded for by separate genes, a scFv has the two domains joined (e.g.. via recombinant methods) by a synthetic linker. This enables them to be made as a single protein chain in which the V_L and V_H regions pair to form monovalent molecules.

The various antibodies or antigen-binding fragments described herein can be produced by enzymatic or chemical modification of the intact antibodies or synthesized de novo using recombinant DNA methodologies or identified using phage display libraries. Methods for generating these antibodies or antigen-binding molecules are all well known in the art. For example, single chain antibodies can be generated using phage display libraries or ribosome display libraries, gene shuffled libraries. In particular, scFv antibodies can be obtained using methods described in, e.g., Bird et al., Science 242:423-426. 1988: and IIuston et al., Proc. Natl. Acad. Sci. USA 85:5879-5883, 1988. Fv antibody tragments can be generated as described in Skerra and Pliickthun, Science 240:1038-41, 1988. Disulfide- stabilized fv fragments (dsfvs) can be made using methods described in. e.g.. Reiter et al., Int. J. Cancer 67:113-23. 1996. Similarly, single domain antibodies (dAbs) can be produced by a variety of methods described in, e.g., Ward et al., Nature 341:544-546. 1989: and Cai and Garen, Proc. Natl. Acad. Sci. USA 93:6280-85. 1996.

Polypeptides are polymer chains comprised of amino acid residue monomers which are joined together through amide bonds (peptide bonds). The amino acids may be thc L-optical isomer or thc D-optical isomer. In general, polypeptides refer to long polymers of amino acid residues, e.g., those consisting of at least more than 10, 20. 50, lOO, 200, 500, or more amino acid residue monomers. However, unless otherwise noted, the term polypeptide as used herein also encompass short peptides which typically contain two or more amino acid monomers, but usually not more than 10, 15, or 20 amino acid monomers.

The term "target" refers to a molecule or biological cell of interest that is to be analyzed or detected, e.g., a nucleotide, an oligonucleotide, a polynucleotide, a polypeptide, a protein, or a blood cell. Preferably, the target antigen of step (b) is an artificial polypetide or natural polypeptide, but it is not limited thereto.

A cell has been "transformed" by an exogenous or heterologous polynucleotide when such polynueleotide has been introduced inside the cell. The transforming DNA may or may not be integrated (covalently linked) into the genome of the cell. In prokaryotes, yeast, and mammalian cells for exmnple, the transforming polynucleotide may be maintained on an episomal element such as a plasmid. With respect to eukaryotic cells, a stably transformed cell is one in which the transforming polynucleotide has become integrated into a chromosome so that it is inherited by daughter cells through chromosome replication. This stability is demonstrated by the ability of the eukaryotic cell to establish cell lines or clones comprised of a population of daughter cells containing the transfanning polynucleotide. A "clone" is a population of cells derived from a single cell or common ancestor by mitosis.

An exemplary scheme showing the strategy for the phenotype-information-phenotype cycle is illustrated in Fig. 1. The method of the present invention is designed for the selection of antibodies where the target molecule is a minor component of an otherwise complex mixture. Essentially, the method integrates phenotypic information derived from a binary selection format to a stringent affinity-based recovery of genotype. To generate a phenotype-information-phenotype cycle (Fig. 1), one begins with a large library of different proteins expressed in phages where only a very minor fraction of the members of the library exhibit a desired phenotype such as binding to another protein or the surface of a cell. After an initial selection, the both noise and signal are existed. Next, the information content of the phenotypic pool is established by deep sequencing of its members. In the examples, the information of interest concerns the frequency distribution of antibody molecules that bind to the target relative to the control. The central concept is that, because the starting library is very large, a given sequence should not be seen multiple times unless it has been selected. In addition to the frequency distribution of sequences, one could derive many other relevant informational parameters such as protein homologies and/or predicted secondary and tertiary structures. Importantly, unlike an analysis of immunized animals that informs as to what has happened, a study of very large naive or synthetic antibody libraries allows one to determine what can happen.

The nucleic acids encoding the molecules considered to be signal are recovered by hybridization and converted back to phenotype by transformation of Escherichia coli thereby completing the phenotype-information-phenotype-cycle. Although we generated the cycle for antibodies, it should work for any large collection of homologous or heterogeneous molecules that are genetically encoded or even for organic molecules contained in DNA-encoded combinatorial libraries.

The important advantage of the cycle described here lies in its potential go beyond selection against protein singularities to find rare antibodies in which the target is a component of an otherwise complex mixture. The two main situations where this problem arises is in the selection of cognate antigen-antibody pairs in the library against library format and the selection of antibodies that bind to proteins in which, in each case, the target molecule may be only a minor component of a cell surface. Classically, these selections have been much more difficult than those against purified proteins.

In this regard, the fact that we were able to select an antibody to a preselected epitope in the presence of a large number of irrelevant proteins on the E. coli surface is encouraging. Another advantage of selecting against proteins in situ is that they exist in a functional conformation that in many cases may be altered when they are removed from their natural environment.

Even though the antibody molecule lends itself to the cycle studied here because it contains variable regions, the technique can be generalized because, even when libraries of highly related molecules are constructed, individual members can be given uniqueness by incorporating artificial diversity elements ("bar codes") in the vector.

A much deeper informatics analysis than that applied herein can be used to study the information content of the nucleic acid pool generated from the phenotypic step of our process. The present inventors anticipate that, when more sophisticated algorithms are used to study the phenotypic collective, less transparent but otherwise important parameters will be deciphered. Finally, the information content will be a function of the number of reads from deep sequencing. Although the 3,000 reads used here were sufficient to establish the principle of the cycle, we expect that many other interesting antibodies will be revealed as the number of reads approaches a million or more.

In accordance with another aspect, the present invention relates to a method for producing a polypeptide specifically binding to a target polypeptide, comprising one cycle of the following steps: (a) providing a library of different proteins displayed on the surface of phages; (b) contacting the library with (i) a group of cells which displays the target polypeptide (test group) and (ii) another group of cells which does not display the target polypeptide or displays less amounts of the target polypeptide (control group); (c) removing the unbound phages and sequencing the nucleotides of the bound phages; (d) comparing the nucleotide sequences of the test group and control group; (e) selecting nucleotide sequences from the test group which are not common with the control group and identifying these sequences as sequences encoding polypeptide specifically binding with the target polypeptide; (f) extracting single-stranded DNA from phage particles and hybridizing to a tagged probe; (g) converting the single-stranded DNA to double stranded DNA; and (h) producing the polypeptide specifically binding with a target polypeptide by transforming it into host cells.

Preferably, the step (c) may be accomplished by deep sequencing.

Preferably, the cells may be selected from the group consisting of bacteria, yeast and eukaryotic cells but it is not limited thereto.

Preferably, the library may comprise single chain variable region fragments (scFvs) or single domain antibodies (dAbs) but it is not limited thereto.

Preferably, the library may comprise a library of human polypeptides but it is not limited thereto.

Preferably, the target polypeptide of step (b) may be an arfificial polypeptide or natural polypeptide but it is not limited thereto.

Hereinafter, the present invention will be described in more detail with reference to Examples. However, these Examples are for illustrative purposes only, and the invention is not intended to be limited by these Examples.

Method

Example 1: Bacteria Surface Display of Antigen

The signal peptide and first nine amino acids of the Escherichia coli major outer membrane lipoprotein and amino acids 46-159 of the outer membrane protein A (OmpA) was cloned into the pCGMT vector and the gene encoding the antigen was inserted directly behind OmpA. Alternatively, the gene encoding the maltose-binding protein (MBP) was inserted between the OmpA and antigen to display an MBP antigen fusion protein. To facilitate the detection of displayed antigen, a FLAG tag was added at the C terminus of the antigen. The plasmid was transformed into TOP10 F' bacteria. The bacteria were cultured overnight at 37 ℃ in super broth (SB) medium supplemented with 2% glucose and 50 ㎍/mL carbenicillin (CARB). The medium was changed to glucose-free SB medium and induced at 24 ℃ for 2 h without addition of IPTG.

Example 2: Whole-Cell ELISA

Freshly induced antigen displaying bacteria (2 x 10⁸) were incubated with HRP-conjugated anti-FLAG antibody (Sigma A8592) for 1 h at 37 ℃. Bacteria were centrifuged and washed five times with PBS and 2,2'-azinobis(3-ethylbenzthiazoline-6-sulfonic acid) (ABTS) developer [450 ㎍/mL ABTS, 0.01% H2O2, 100 mM citrate buffer (pH 4.0)] was added. After 15 min for color development, cells were centrifuged and 50 ㎕ of supernatant was transferred to an ELISA plate reader and the A₄₀₅ was measured.

Example 3: Flow Cytometry Analysis

Induced bacteria were incubated with 2 ng/mL phycoerythrin-labeled cytokine-specific antibodies at 22 ℃ for 1 h, washed twice with PBS and resuspended in PBS at 5 x 10⁷ cells per mL, 10,000 E. coli events were measured with a Becton Dickinson flow cytometer, and data were analyzed with Flowjo software. The antibodies were obtained from the following vendors: anti-human(h)TNF, Biolegend 502908; antihCX3CL1, R&D Systems IC365P; anti-hIL-6 R&D Systems IC206, Biolegend, 501106; and anti-hIL-4, Biolegend 500703.

Example 4: Subtractive Panning

Phage with a titer of 10¹¹ cfu were preincubated for 1 h with 5 x 10⁸ of control bacteria. After centrifugation, the supernatant was transferred to freshly induced antigendisplaying bacteria and incubated for 1 h. Unbound phage were washed away by pelleting bacteria, resuspending them in PBS/3% BSA for five times, and transferring the bacteria to a new tube. Bound phages were eluted with 100 ㎕ glycine elution buffer [200 mM glycine, 1 mgmL BSA, 0.05% Tween 20 (pH 2.2)]. The buffer containing phage was neutralized with 7㎕ 2 M Tris, after which XL1-blue cells infected with phage were plated and cultured at 30 ℃ overnight. Cells were scraped from the plate into 100 mL of SB with 2% glucose to OD₆₀₀ of 0.1. The culture was incubated at 37 ℃ until the OD₆₀₀ reached 0.8. Helper phage VCSM13 was added for 30 min at 37 ℃ without shaking, followed by shaking for 90 min at 37 ℃ at 300 rpm. Cells were centrifuged to remove the glucose-containing medium and resuspended in 100 mL of SB with 50 ㎍/mL CARB, 10 ㎍/mL tet, and 70 ㎍/mL kanamycin and incubated by shaking at 30 ℃ overnight. Phage were precipitated on ice with 4% PEG and 3% NaCl and resuspended in 1% BSA/PBS. Amplified phages that had a titer of 10¹¹ cfu were subjected to the next round of subtractive panning.

Example 5: Roche 454 Pyrosequencing of VH

Bar-coded primers were designed to amplify VH from phagemid pools of paired bacteria-bound phage. Phusion Hot Start High-Fidelity DNA Polymerase (New England Biolabs F-549) was used and 10 cycles of PCR were carried out to reduce bias incurred due to PCR amplification. The amplicons of VH were purified from 1% agarose gels and deep sequenced according to Roche 454 GS FLX instructions.

Example 6: Bioinformatics Analysis

Antibody sequences were analyzed by BLAST downloaded from the National Center for Biotechnology Information and compared to germ-line IGVH, IGDH, IGJH sequences obtained from the International Immunogenetics Information System database. Matching was done using the MEGABLASTalgorithm. Germ-line genes were assigned to each read using the MEGABLASTalgorithm. MEGABLAST was run requiring at least 80% sequence identity and using 11-bp word size setting to match IGVH genes. Reads were assigned to genes based upon a hierarchy of BLAST result parameters where the next parameter in the hierarchy was considered only if a read matched multiple genes with the previous parameter. The hierarchy was ordered to be more permissive to gaps in the BLAST alignments because of known homopolymer sequencing errors using 454 technology as follows: highest bit score, highest percent identity, longest alignment length, least number of mismatches, and, finally, least number of gaps. IGJH genes were assigned similarly, requiring 80% sequence identity, using a 7-bp word size, and requiring the read to span at least 23 bases of the J genes. CDR3 sequences were extracted by a more rigorous Smith-Waterman alignment of the identified IGVH and IGJH for each read in order to more precisely define the CDR3 boundary nucleotides.

Example 7: Recovery of Single-Chain Fv (scFv) with Overlapping PCR

The variable heavy-chain complementarity determining region 3 (VH CDR3) and scFv flanking vector sequence-specific primers were used to amplify the scFv fragments with phagemid pools from round two as template. The two resulting DNA fragments were assembled by overlapping PCR, digested with SfiI, ligated into SfiI digested pcGMT3 phagemid and transformed into XL1-blue competent cell. Five to 20 colonies for each scFv were inoculated and minipreps of the overnight culture were sequenced by Sanger sequencing.

Example 8: Recovery of scFv with the Hybridization Method

Single-stranded DNA was purified from the phage particles selected in the second round with QIAprep Spin M13 kit (Catalog number 27704). A biotinylated probe corresponding to the most variable region of VH CDR3 was used to capture the ssDNA from the pool. Four micrograms of the total ssDNA library were mixed with 50 ng of the biotinylated probe in hybridization solution [5 x standard saline phosphate/EDTA (0.18 M NaCl/10 mM phosphate, pH 7.4/1 mM EDTA (SSPE), 0.1% Tween 20] and hybridized at 45 ℃ for 12 h. The hybrid was captured with streptavidin-coated paramagnetic beads (Dynabeads M-280 Streptavidin 112.05D) by incubation at 22 ℃ for 1 h in binding buffer (1 x SSPE, 0.1% Tween 20). The beads were collected with a magnet and washed in wash buffer (2 x SSPE, 0.1% Tween 20) six times. The beads were resuspended in water and heated at 90 ℃ for 1 min to release the captured DNA. The released ssDNA was annealed at 55 ℃ with a primer having the same sequence as the probe and the entire circle of ssDNA was replicated at 72 ℃ to form dsDNA and transformed into XL1-blue competent cell. The transformants were plated on CARB plates and 20 colonies were characterized with Sanger sequencing.

Example 9: Phage ELISA

ELISA plates were coated overnight at 4 ℃ with 50 ㎕ of PBS containing 200-ng coating protein [IL-1 receptor antagonist (IL-1RA), IL-6, or truncated IL-6). Wells were washed twice with PBST and blocked with 50 ㎕ of PBS/3% BSA for 1h at 37 ℃. Different concentrations of phage in PBS Tween 20 (PBST)/3%BSA were added and incubated for 1 h at 37 ℃. Wells were washed five times with PBST and then 50 ㎕ of HRP-conjugated antiphage antibody in PBST/3% BSA was added and incubated for 1 h at 37 ℃. Wells were again washed five times with PBST, and 50 ㎕ of ABTS developer was added. After 15 min at 22 ℃, the A₄₀₅ was measured.

Result

Example 10: Initiating the Cycle Using a Phage Against E. coli Binary Selection System

Most combinatorial antibody libraries are selected against purified proteins. To establish the first part of the cycle, an E. coli format to express the antigen was utilized so as to establish phenotype by a method that both might be generally useful and is representative of complex systems such as cell surfaces where a given antigen is but one component of an otherwise complex mixture. In this system, the interaction between antigens that are expressed on the surface of E. coli and antibodies expressed on the surface of phage are studied. Although the antigens were expressed on the surface of E. coli, other systems using yeast or eukaryotic cells could be implemented. In the phage system, the antibody protein is expressed on the surface of the phage and the DNA encoding its heavy and light chains is packaged in the interior of the phage, thereby linking genotype to phenotype in a way that will ultimately yield the information content of the system. This format allows for a stringent test of our method because the nonspecific binding of phage to the surface of E. coli and/or the binding to irrelevant molecules can be very high and, in our hands, is often difficult to circumvent. In this regard, the binary selection format is critical because the proper control is not in doubt. Thus, one can compare phage that bind to the surface of E. coli or other cells that express antigen to those that do not.

Two different expression systems were tested. Ten different proteins were expressed on the surface of E. coli by linking them to either the cell surface outer membrane protein A (OmpA) or a maltose-binding protein (MBP)-OmpA fusion (Fig. 2) . In each case, the signal peptide and transmembrane helix was removed from the target protein (Fig. 2A). The success of the expression was determined by ELISA and flow cytometry analyses using antibodies that bound to a FLAG tag on the expressed proteins (Fig. 2 B and C) as well as those that bound to the native conformation of the target molecule by flow cytometry (Fig. 2 D). The best expression system was somewhat dependent on the particular protein studied. For example, chemokine (C-C motif) ligand 28 (CCL28) and IL-1β expressed best when fused to the MBP-OmpA fusion, whereas TNF-α and IL-6 could be efficiently displayed when directly fused to OmpA as determined by ELISA (Fig. 2 B and C) and flow cytometry (Fig. 2D). Importantly, in these systems, the necessity of purifying the target antigen is avoided and molecules that historically have been difficult to purify such as IL-6 are easily displayed.

To initially determine whether the bacterial display format could be used to select binding proteins from combinatorial antibody libraries, the ability to select phage against a bacterially expressed 12 amino acid long peptide epitope from the retrovirus glycoprotein 41 (gp41) or the full-length IL-1 receptor antagonist (IL-1RA) was studied. Phage containing either antigp41 or anti-IL-1RA antibodies expressed on their gene III protein were enriched 20- to 50-fold when E. coli expressing the cognate antigen as opposed to control E. coli were used for enrichment. Phage that expressed an irrelevant antibody were not selected (Table 1).

Table 1

Example 11: Linking Information to Phenotype

To first approximation, the information of interest relates to frequency with which given nucleic acid sequences appear in the phenotypic pool and the ratio of their abundance in control versus experimental selections (frequency ratio). This information is obtained by pyrosequencing of the DNA contained in the selected phage populations followed by a bioinformatic analysis of the sequences. To return to phenotype, the DNA sequences of interest must be selectively recovered. Three different methods were tested for recovery of nucleic acid sequences considered to be signal (Fig 3). The diversity of the variable heavy-chain complementarity determining region 3 (VH CDR3) is generated by rearrangement of a limited number of VH, diverse heavy (DH), and joining heavy (JH) gene segments and is significantly increased by the addition and deletion of nucleotides in the formation of the junctions between gene segments. The added nucleotides are known as P nucleotides and N nucleotides, which represent the most variable region of each antibody sequence. Therefore, in all three methods, a probe (or primer) complementary to the P and N nucleotides and the D region of the gene encoding the VH CDR3 region of the antibody molecule was used (Table 2).

Table 2

Initial studies showed that when the frequency of a sequence in the phenotypic pool was high (above approximately 5%), standard overlap PCR amplification using the VH CDR3 specific primer together with a vector specific primer allowed its selective recovery. However, when the frequency of the target sequence was 1% or less of the pool, this PCR-based recovery process was too promiscuous to be useful, probably because of off-target binding of the primer. Likewise, when rolling circle amplification was attempted, the background was too high, perhaps because of the very limited DpnI activity on the hemimethylated template. Thus, these two standard methods did not allow utilization of the full power that derives from the information content of the phenotypic pool. In the third approach, instead of using the phagemids from bacteria, single-stranded DNA with minus polarity was extracted directly from phage particles and hybridized to a biotinylated version of the probe. Unlike the other two approaches, this method can be made more selective because it uses affinity purification in a process in which a wide range of conditions can be used to control the stringency of the nucleic acid recovery step. This method also avoids artifacts caused by selective amplification of certain sequences, such that the recovered copy number is directly related to phenotype. Further, because the heavy and light chains are present in the same phage ssDNA molecule, its isolation maintains this linkage, which encodes the phenotypically successful chain pairing. Finally, there is no need to synthesize DNA and construct new vectors, which can be very time consuming and expensive.

To determine the ideal conditions for obtaining selectivity using this affinity approach, a single-stranded phage vector encoding the scFv portion of an antibody that binds to IL-1RA as well as spectinomycin (SPEC) antibiotic resistance was added to a complex pool of vectors that encoded irrelevant antibody sequences and carbenicillin (CARB) antibiotic resistance (Fig 3 and Table 3). Selectivity was determined by measuring the ratio of SPEC- to CARB-resistant colonies after transformation of E. coli. After hybridization of the probe at either 45 or 65 ℃, the hybrid complex was captured on streptavidin beads and the single-stranded phage genome was detached by melting at 90 ℃ for 1 min. E. coli was transformed with either the single-stranded phage DNA or double-stranded DNA that was generated by T4 or Taq DNA polymerase (Table 3). The best results were obtained when hybridization was at 45 ℃ and transformation was carried out using Taq DNA polymerase to generate doublestranded DNA (Table 3). Remarkably, under these conditions, selective recovery of the desired sequences approached 100%.

Table 3

To optimize the parameters for scFv recovery by hybridization (Fig. 3C), a model system was constructed. ssDNA was purified from phage particles with QIAprep Spin M13 kit (catalog no. 27704). The ssDNA encoding scFv H9 and spectinomycin resistance (SPEC^r) was spiked into a pool of ssDNA encoding irrelevant scFVs and carbenicillin resistance (CARB^r) at a ratio of 1.100. A biotinylated probe corresponding to H9 VH CDR3 was used to capture the H9 encoding ssDNA from the pool. Four micrograms of the total ssDNA library was mixed with 50 ng of the biotinylated probe in hybridization solution (5 SSPE, 0.1% Tween 20) and hybridized at either 45 or 65 ℃ for 12 h. The hybrid was captured with streptavidin-coated paramagnetic beads (Dynabeads M-280 Streptavidin 112.05D) by incubation at 22 ℃ for 1 h in binding buffer (1 x SSPE, 0.1% Tween 20). The beads were collected with a magnet and washed in 2 x SSPE, 0.1% Tween 20 six times. The beads were resuspended in water and heated at 90 ℃ for 1 min to release the captured DNA. The released DNA was used to transform XL1-blue cells directly or converted to dsDNA before transformation. Two methods to convert ssDNA to dsDNA were compared. ssDNA was annealed at 55 ℃ with a primer having the same sequence as the probe that was phosphorylated at the 5 position and the entire circle of ssDNA was replicated at 72 ℃ to form a dsDNA with a nick in the newly synthesized strand. Alternatively, ssDNA was annealed with the same primer and converted to closed dsDNA with T4 DNA polymerase plus T4 DNA ligase at 22 ℃. The transformants were plated on CARB plates and SPEC plates and the total colony number determined (CARB plate colony number plus SPEC plate colony number). The selectivity was determined by measuring the number of SPEC-resistant colonies as a function of the total number of colonies (SPEC plate colony numberthe total colony number).

Example 12: Selective Power of the Cycle

To initially validate the cycle and test its resolving power, "spiked-in" libraries were constructed in which phage-expressing antibodies known to bind to the IL-1RA were added to a generic naive library such that they were present only once in a library of 1.0 x 10⁹ members. Two rounds of selection were carried out. After each round of selection, the information content of the pool was determined by pyrosequencing using over 3,000 reads (Fig. 4A). Even after one round, obvious information about the differential frequency of selected clones is present. Interestingly, in addition to the spiked-in H9 clone, we noted selective appearance of another clone (H1). Based on this information, the two clones (H1 and H9) were considered enriched relative to binding to control E. coli, even though there was still approximately 50-fold more noise than signal. The two clones thought to be signal were carried through the rest of the phenotype-informationphenotype cycle. Antibodies selected by this cycle were shown to bind specifically to the IL-1RA target antigen (Fig. 4B and C).

This experiment also illustrates the difficulty that may be encountered by conventional panning against complex systems where the selection for unwanted proteins may be favored (Table 4). In toto, these experiments suggest that the resolving power of the method is at least one part in 1.0 x 10⁹.

Table 4

Example 13: Selection from Authentic Libraries

Although the spiked-in libraries described above suggested the resolving power of the cycle was robust, one might argue that a previously selected phage could already have replicative advantage.

Therefore, the cycle starting with an authentic naive combinatorial antibody library was tested. The method to find antibodies that bound to a predetermined epitope on the IL-6 protein was challenged. In this case, even interactions with the unwanted portion of the IL-6 molecule are noise. A naive combinatorial antibody library containing 3.0 x 10⁹ members was selected against E. coli displaying either full-length IL-6 or a truncated version where 27 N-terminal amino acids were deleted (Fig. 5a and b). After two rounds of selection, three sequences representing two classes of molecules were determined to be uniquely present in phage reactive with the full-length IL-6 molecule by comparing the number of times the sequence appears when it is selected against experimental versus control cells (frequency ratio) (Fig. 5c). The genes encoding the selected antibodies are present at frequencies ranging between 0.9% and 2.8% of the pool and would likely have been missed in the conventional iterative selection of combinatorial antibody libraries (Table 5, 6 and 7).

Table 5

Table 6

Table 7

In fact, when Sanger sequencing of the DNA from 150 phagemids from the selected pool was carried out, only the most frequent (2.8%) of the three sequences deduced from deep sequencing was present and it appeared only once. Furthermore, if one uses a standardized format, the information from previous searches becomes part of the accumulated database. For example, sequences N27-5 and N27-6 were selected in multiple screens and likely encode antibodies against bacterial proteins (Table 4 and 5).

The present inventors tested the significance of the different frequency ratios by replicating six phage clones that had frequency ratios ranging between 1 and 57.7 (Table 6). Their binding to the two different IL-6 molecules expressed on the E. coli surface was determined by measuring the titer of bound phage. When the frequency ratio is above 10 (N27-1 and N27-2), the isolated phage clones show a strong binding preference for the cognate E. coli as determined by the titer of bound phage. However, no specific binding to the expressed target is observed when the ratio approximates one, even if the frequency is very high (N27-5). Antibodies that bind with a high frequency but do not discriminate could either bind to an irrelevant E. coli surface protein or a region of IL-6 other than the N terminus. When the absolute frequency is too low, the credibility of the ratio is compromised, because it is not statistically significant (N27-3 and N27-4). Thus, neither the absolute frequency nor the selectivity ratio by themselves gives sufficient information and the two parameters should be combined when selecting candidate antibodies. Therefore, a Z score derived from the two-proportion Z test was introduced. Given the fraction of times an individual CDR3 is observed in each sample and the total number of CDR3 sequences per sample, the standard formulation of the two-proportion Z test was utilized to estimate the number of standard deviations, derived from the expected distribution of differences between two normally distributed proportions, that separated our observed two proportions. This measure accounts for both the ratio between two proportions and the absolute frequency difference of two proportions. For example, the top two ranks for selection of IL-6 N-terminal binding antibody chosen based on the Z score are proven correct. In general, the Z score was used to combine the highest differential ratio with the absolute frequency when selecting candidate antibodies.

To complete the cycle, the clones encoding the antibodies were recovered by the hybridization method (Fig 3C). Sequence analysis of 20 randomly picked clones revealed that 19 out of 20 had the correct sequence. The single incorrect sequence represented the most abundant member and was present in both the experimental and control pools.

Initially, the selected sequences were studied by phage ELISA for their ability to bind to purified WT IL-6. Interestingly, only one of the two selected antibodies (N27-1) bound to purified IL-6 when it was adsorbed to plastic surfaces for the ELISA, thereby underscoring the importance of controlling the conformation of target molecules during the selection process (Table 6). Purified antibody N27-1 reacted with the full-length but not the truncated IL-6 protein by ELISA and competed with a commercial antibody known to neutralize the function of IL-6 (Fig. 5d and e). Although the commercial antibody was known to block the binding of IL-6 to its receptor, the exact portion of the molecule to which it bound was unknown. These experiments show that the antibody that blocks the binding of IL-6 to its receptor interacts with the 27 N-terminal amino acids (Fig. 5b and e).

Claims (13)

1. A one-cycle method for screening an antibody specifically binding to a target antigen, comprising following steps:

   (a) providing an antibody library of different proteins displayed on the surface of phages;

   (b) contacting the antibody library with (i) a group of cells which displays the target antigen (test group) and (ii) another group of cells which does not displays target antigen or displays less amounts of target antigen (control group);

   (c) removing the unbound phages and sequencing nucleotide of the bound phages;

   (d) comparing nucleotide sequences of the test group and control group; and

   (e) selecting nucleotide sequences the test group which are not common with the control group and identifying these sequences as sequences encoding the antibody specifically binding to the target antigen.
2. The method of claim 1, further comprising (f) extracting single-stranded DNA from phage particles and hybridizing to a tagged probe; and (g) converting the single-stranded DNA to double-stranded DNA.
3. The method of claim 1, wherein step (c) is accomplished by deep sequencing.
4. The method of claim 1, wherein the cells are selected from the group consisting of bacteria, yeast and eukaryotic cells.
5. The method of claim 1, wherein the antibody library comprises single chain variable region fragments (scFvs) or single domain antibodies (dAbs).
6. The method of claim 1, wherein the antibody library comprises a library of human antibodies.
7. The method of claim 1, wherein the target antigen of step (b) is an arfificial polypeptide or natural polypeptide.
8. A method for producing a polypeptide specifically binding to a target polypeptide, comprising one cycle of the following steps:

   (a) providing a library of different proteins displayed on the surface of phages;

   (b) contacting the library with (i) a group of cells which displays the target polypeptide (test group) and (ii) another group of cells which does not display the target polypeptide or displays less amounts of the target polypeptide (control group);

   (c) removing the unbound phages and sequencing the nucleotides of the bound phages;

   (d) comparing the nucleotide sequences of the test group and control group;

   (e) selecting nucleotide sequences from the test group which are not common with the control group and identifying these sequences as sequences encoding polypeptide specifically binding with the target polypeptide;

   (f) extracting single-stranded DNA from phage particles and hybridizing to a tagged probe;

   (g) converting the single-stranded DNA to double stranded DNA; and

   (h) producing the polypeptide specifically binding with a target polypeptide by transforming it into host cells.
9. The method of claim 8, wherein step (c) is accomplished by deep sequencing.
10. The method of claim 8, wherein the cells are selected from the group consisting of bacteria, yeast and eukaryotic cells.
11. The method of claim 8, wherein the library comprises single chain variable region fragments (scFvs) or single domain antibodies (dAbs).
12. The method of claim 8, wherein the library comprises a library of human polypeptides.
13. The method of claim 8, wherein the target polypeptide of step (b) is an arfificial polypeptide or natural polypeptide.

Images