Antibody Library and Method

Abstract

This disclosure relates to methods of generating antibody libraries, antibody libraries produced using such methods, and variant antibodies. Presently, methods of improving antibody binding (affinity maturation assays) require the screening of vast libraries of antibody variants (often >10.sup.10) to identify a small fraction of variants with improved characteristics. The present invention involves taking the nucleotide sequence of the framework and complementarity determining region of a target antibody and identifying motifs which would be recognised by deamination somatic hypermutation enzymes. A small library of variants is then created which incorporate one or more of these mutations. It was found that a relatively high proportion of the variants have an increased affinity. The technique of the present invention was demonstrated on the trastuzumab and Cathepsin S antibodies, and the variants produced are also claimed.

Claims

1. A variant of a reference antibody, wherein said variant has (i) at least two amino acid changes in the light chain sequence when compared to the light chain amino acid sequence of the reference antibody, (ii) at least two amino acid changes in the heavy chain sequence when compared to the heavy chain amino acid sequence of the reference antibody, or (iii) at least one amino acid change in the light chain sequence when compared to the light chain amino acid sequence of the reference antibody and at least one amino acid change in the heavy chain sequence when compared to the heavy chain amino acid sequence of the reference antibody; wherein each of said amino acid changes are at amino acid residues independently encoded from a DNA segment of a variant DNA sequence, wherein said DNA segment of the variant differs from that of the corresponding DNA sequence encoding the reference antibody by a point mutation in a DNA motif susceptible to deamination by a somatic hypermutation inducing enzyme.

2. The variant antibody according to claim 1 wherein said somatic hypermutation inducing enzyme is Activation-Induced Deaminase (AID) and said DNA motif is DGYW or WRCH, where D is adenine, guanine or thymine, R is adenine or guanine, G is guanine, C is cytosine, H is adenine or cytosine or thymine, W is adenine or thymine, and Y is any cytosine or thymine.

3. The variant antibody according to claim 1, wherein said reference antibody is trastuzumab or a Cathepsin S antibody, wherein the Cathepsin S antibody is Fsn503h.

4. The variant antibody according to claim 3, wherein when said reference antibody is trastuzumab, said amino acid changes are selected from the group consisting of 1c9N, 1c9T, 1c9I, 1c9R, 1c9K, 1c25G, 1c25V, 1c25D, 1c31N, 1c31S, 1c31I, 1c32D, 1c32G, 1c32V, 1c32T, 1c32N, 1c32S, 1c321, 1c32P, 1c32L, 1c32F, 1c33L, 1c331, 1c34G, 1c34V, 1c34D, 1c38E, 1c38K, 1c40A, 1c40S, 1c40T, 1c43G, 1c43V, 1c43T, 1c43N, 1c43S, 1c43I, 1c43P, 1c43L, 1c43F, 1c46V, 1c46I, 1c47V, 1c51S, 1c51P, 1c51T, 1c76R, 1c76N, 1c76T, 1c76K, 1c76I, 1c79K, 1c79E, 1c80T, 1c80S, 1c80A, 1c85S, 1c85N, 1c85I, 1c89H, 1c90E, 1c90A, 1c91N, 1c91D, 1c91Y, 1c93S, 1c93N, 1c93I, 1c94S, 1c94N, 1c941, 1c101D, 1c102S, 1c102N, hc2L, hc2I, hc3H, hc4M, hc4V, hc13K, hc13E, hc14A, hc14T, hc14S, hc16A, hc16V, hc16D, hc23E, hc23G, hc23V, hc23T, hc23K, hc23R, hc23I, hc23P, hc23L, hc23S, hc24D, hc24G, hc24V, hc24T, hc24N, hc24S, hc24I, hc24P, hc24L, hc24F, hc26A, hc26V, hc26D, hc28K, hc35N, hc35D, hc35Y, hc48L, hc48I, hc49G, hc49S, hc56A, hc56V, hc56D, hc58S, hc58N, hc58I, hc61G, hc61V, hc61D, hc79G, hc79V, hc79D, hc82E, hc82K, hc85R, hc88D, hc88T, hc88S, hc88P, hc88G, hc92G, hc92V, hc92D, hc103A, hc103V, hc103D, hc106D, hc106G, hc106V, hc106T, hc106N, hc106S, hc106I, hc106P, hc106L, hc106F, hc114S, hc114N, and hc114I, and when said reference antibody is Fsn503h, said amino acid changes are selected from the group consisting of 1c12A, 1c12S, 1c12T, 1c19V, 1c28R, 1c32T, 1c32I, 1c45A, 1c45S, 1c45T, 1c5OH, 1c51V, 1c51F, 1c51I, 1c56L, 1c56F, 1c56I, 1c58K, 1c66S, 1c69A, 1c69V, 1c81T, 1c81I, 1c81N, 1c85P, 1c85S, 1c85T, 1c90L, 1c90F, 1c96I, 1c96S, 1c96I, 1c96N, 1c108N, hc3H, hc4V, hc4M, hc10A, hc10V, hc14A, hc14S, hc24G, hc24V, hc30T, hc30I, hc31R, hc31T, hc37L, hc37F, hc40P, hc40S, hc52S, hc52I, hc53S, hc53I, hc84T, hc84I, hc92G, and hc92V.

5. (canceled)

6. The variant antibody according to claim 4, wherein the variant light chain and heavy chain sequences do not differ from those of the reference antibody at any residue other than the recited amino acid residues.

7. The variant antibody according to claim 3, wherein when said reference antibody is trastuzumab, said variant antibody molecule has the combination of amino acid mutations as shown for any one of the variants in FIG. 11a, FIG. 11b, FIG. 11c, FIG. 11d, FIG. 11e, FIG. 11f, and FIG. 12, and when said reference antibody is Fsn503h, said variant antibody molecule has the combination of amino acid mutations as shown for any one of the variants in FIG. 3.

8. (canceled)

9. The variant antibody according to claim 3, wherein the variant antibody is a variant of trastuzumab comprising,. relative to trastuzumab, the following amino acid changes: 1c9K, 1c43F, and hc106V.

10-14. (canceled)

15. The variant antibody according to claim 1, wherein said variant light chain and heavy chain sequences comprise in total at least three amino acid changes compared to the amino acid sequence of the reference antibody.

16. The variant antibody according to claim 1 wherein one or more of said amino acid changes are in framework regions of said variant antibody or are in CDRs of said variant antibody, or wherein all of said amino acid changes are in framework regions of said variant antibody.

17-18. (canceled)

19. The variant antibody according to claim 1, wherein the change of affinity of said variant antibody molecule relative to the reference antibody is greater than −2 and the change of stability of said variant antibody molecule relative to reference antibody is greater than −2.

20. A library of antibody molecules, wherein each antibody molecule is a variant of a reference antibody, wherein the amino acid sequence of each antibody molecule differs from the amino acid sequence of the reference antibody at one or more amino acid residues, wherein each of said amino acid residues are independently encoded from a DNA segment of a variant DNA sequence, wherein said DNA segment of the variant differs from that of the corresponding DNA sequence encoding the reference antibody by a point mutation in a DNA motif susceptible to deamination by a somatic hypermutation inducing enzyme.

21. The library according to claim 20, wherein said somatic hypermutation inducing enzyme is Activation-Induced Deaminase (AID) and said DNA motif is DGYW or WRCH, where D is adenine, guanine or thymine, R is adenine or guanine, G is guanine, C is cytosine, H is adenine or cytosine or thymine, W is adenine or thymine, and Y is any cytosine or thymine.

22. (canceled)

23. The library according to claim 20, wherein the nucleotide sequence encoding each of said antibody molecules does not differ from that of the nucleotide sequence encoding said reference antibody at any nucleotide residue other than nucleotide residues of said DNA motif

24. The library according to claim 20 wherein one or more of said DNA motifs are in DNA sequences which encode framework regions of said antibody molecule or are in DNA sequences which encode CDRs of said antibody molecule, or wherein all of said DNA motifs are in DNA sequences which encode framework regions of said antibody molecule.

25-26. (canceled)

27. The library according to claim 20 wherein the DNA sequence of each variant does not comprise or encode a deamination site, isomerisation site, N-linked glycosylation site or oxidation site which originates from said point mutation in said DNA motif.

28. A library of nucleotide sequences, wherein each member of the library encodes an antibody molecule of the library of antibody molecules according to claim 20.

29. A method of generating/producing a library of variant antibody molecules, wherein said variant antibody molecules are variants of a reference antibody, said method comprising the steps: a) providing a nucleotide sequence encoding the reference antibody, b) in said nucleotide sequence, identifying one or more DNA motifs susceptible to deamination by a somatic hypermutation inducing enzyme; c) for one or more of said DNA motifs, selecting at least one variant nucleotide residue to substitute for a residue of said DNA motif, wherein said substitution will result in a variant nucleotide sequence which encodes a variant antibody molecule having, relative to the reference antibody, a change in the amino acid sequence being encoded by said DNA motif; and d) repeating steps (b) and (c); such that a library containing a plurality of variants of said reference antibody is generated.

30. The method according to claim 29, wherein said method further comprises screening said library of variants to determine binding to an epitope to which the reference antibody binds, and using those variants determined to bind to said epitope with at least one of an affinity or a stability greater than a predetermined value relative to the reference antibody to generate an optimised library of variant antibody molecules.

31. (canceled)

32. The method according to claim 29, wherein said method is a computer implemented method.

33. The method according to claim 29, further comprising synthesising the variant antibody molecules.

34. (canceled)

35. The library according to claim 20, wherein the reference antibody is trastuzumab or the anti Cathepsin S antibody Fsn0503h.

36. The method according to claim 29, wherein the reference antibody is trastuzumab or the anti Cathepsin S antibody Fsn0503h.

Description

BRIEF DESCRIPTION OF THE FIGURES

[0104] Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying figures in which:

[0105] FIG. 1 shows amino acid sequences of the Fsn0503h antibody light chain and heavy chain showing potential mutations above each chain (SEQ ID Nos 1 & 2). Doubly circled amino acids were not permitted in the library due to their uncommon usage at that position. Doubly circled asterisks (*)represent STOP codons that also were not permitted. Singly circled amino acid regions are CDRs.

[0106] FIG. 2 is a schematic model showing predicted protein-protein interaction between Fsn0503h and Cathepsin S.

[0107] FIG. 3 shows Table 1 which lists affinity and stability predictions of amino acid mutations in a combinatorial fashion for 1 to 6 mutations. Groups of mutations highlighted in blue represent variants with both improved stability and affinity where the change is greater than −2.

[0108] FIG. 4 illustrates an ELISA comparison of each of the expressed variants with the parental Fsn0503h antibody.

[0109] FIG. 5 shows affinity (KD) measurements of each variant alongside the parental Fsn0503h antibody, as measured by BLI with an Octet instrument (Pall).

[0110] FIG. 6 illustrates an affinity on-off rate map of the variants of Fsn0503h. The parental antibody is shown as the hollow dot. The 4.45 nM line represents equal affinity to the parental antibody, and any variants with an increased affinity (<4.45 nM) are within the area to the left of the 4.45 nM line.

[0111] FIG. 7 shows amino acid sequences of the trastuzumab antibody light chain (SEQ ID NO: 3) showing in bold the amino acid residue at each position of the wild type antibody light chain and, where applicable, in the boxes to the right of each residue, potential mutations. Amino acids shown with two lines were not permitted in the library due to their uncommon usage at that position. Asterisks (*) in two lines represent STOP codons that also were not permitted. Complementarity determining regions CDRs and framework regions (FR) are indicated.

[0112] FIG. 8 shows amino acid sequences of the trastuzumab antibody heavy chain (SEQ ID NO: 4) showing in bold the amino acid residue at each position of the wild type antibody antibody chain and, where applicable, in the boxes to the right of each residue, potential mutations. Amino acids shown with two lines were not permitted in the library due to their uncommon usage at that position. Asterisks (*) in two lines represent STOP codons that also were not permitted. Complementarity determining regions CDRs and framework regions (FR) are indicated.

[0113] FIG. 9 schematically illustrates how somatic hypermutations may be used to cause alterations to the DNA sequence of antibodies which results in amino acid changes in the protein structure of the antibody leading to enhanced affinity.

[0114] FIG. 10 illustrates the fully resolved crystal structure of trastuzumab in complex with human HER2 which was used to perform residue scanning. Domain IV of the extracellular domain of HER2 was found to bind to trastuzumab.

[0115] FIG. 11a shows Table 2 which lists trastuzumab variants having one mutation relative to the reference trastuzumab antibody. In column 2, L indicates light chain, while H indicates heavy chain. A Affinity is predicted free energy change (ΔΔG) in kcal/mol relative to trastuzumab affinity. A Stability is predicted free energy change (ΔΔG) relative to trastuzumab stability. Negative A Affinity and negative Δ Stability values are considered to represent improved affinity and improved stability relative to trastuzumab.

[0116] FIG. 11b shows Table 3 which lists trastuzumab variants having two mutations relative to the reference trastuzumab antibody. In column 2, L indicates light chain, while H indicates heavy chain.

[0117] FIG. 11c shows Table 4 which lists trastuzumab variants having three mutations relative to the reference trastuzumab antibody. In column 2, L indicates light chain, while H indicates heavy chain.

[0118] FIG. 11d shows Table 5 which lists trastuzumab variants having four mutations relative to the reference trastuzumab antibody. In column 2, L indicates light chain, while H indicates heavy chain.

[0119] FIG. 11e shows Table 6 which lists trastuzumab variants having five mutations relative to the reference trastuzumab antibody. In column 2, L indicates light chain, while H indicates heavy chain.

[0120] FIG. 11f shows Table 7 which lists trastuzumab variants having six mutations relative to the reference trastuzumab antibody. In column 2, L indicates light chain, while H indicates heavy chain.

[0121] FIG. 12 shows Table 8 which lists the 100 trastuzumab variants from tables 1 to 7 having both improved affinity and improved stability, ranked according to improved affinity.

[0122] FIG. 13 schematically illustrates the top 5 trastuzumab variants scored based on affinity using the predicted free energy change (MG) in kcal/mol. In the first and third ranked variants there are a total of five mutations, while rank 2 and 4 have four mutations and finally rank 5 contains three amino acid mutations. Domain IV of HER2 where trastuzumab binds is coloured cyan, while the scFv is coloured green and CDRs coloured red. Mutations are highlighted in yellow.

[0123] FIG. 14 illustrates sensorgrams from 20 samples showing report points used for ranking. Assay performed at 25° C.

[0124] FIG. 15 shows a graph illustrating off-rate ranking of Trastuzumab variants binding to HER2 and shows stability_early vs stability_late for identification of stable binder (best binders circled in blue). In total, 89 samples were analyzed and ranked with respect to binding stability, however for clarity only 29 are shown. Trastuzumab is commercially available antibody material. Wild-Type is antibody having sequence of trastuzumab and transiently expressed in parallel with all Trastuzumab variants/mutants. Mutant is a Trastuzumab variant. Mutants 5, 6 and 19 are trastuzumab variants listed as numbers 5, 6 and 19 in Table 8.Assay performed at 25° C.

[0125] FIG. 16. shows a graph illustrating off-rate ranking of Trastuzumab variants binding to HER2 and shows Stability_early vs stability_late for identification of stable binders from a small sub-set of the top binders shown in FIG. 15 (those binders closest the 100% left line). In total, 14 Trastuzumab variants were analyzed and ranked with respect to binding stability. Trastuzumab is commercially available antibody. Wild-Type is material transiently expressed in parallel with all Trastuzumab variants/mutants. Mutant; is a Trastuzumab variant. Mutants 5, 6 and 19 are trastuzumab variants listed as numbers 5, 6 and 19 in Table 8. Assay performed at 37° C.

[0126] FIG. 17 shows reference corrected BLI binding curves (black), monitored on a surface of non-covalently immobilised Trastuzumab (a), wild-type (b), mutant 5 (c) and (d) mutant 19 antibodies, for various HER2 concentrations (serial diluted two-fold; highest concentration shown on curve) at 37° C. in running buffer. The apparent dissociation rate constant (kd) and association rate (k.sub.a) constants were determined by globally fitting (red) a simple 1:1 interaction model, A+B=AB, to the sensorgrams using the software supplied with the instrument. Global fitting results are summarised in table 9.

[0127] FIG. 18 illustrates schematically the correlation between affinity prediction and measurement for trastuzumab variants.

[0128] FIG. 19 shows blank corrected BLI binding curves (black), monitored on a surface of non-covalently immobilised HER2 at a range of concentrations of wild-type (a), mutant 5 (b) and (c) mutant 19 antibodies, at 37° C. in running buffer. The apparent dissociation rate constant (kd) and association rate (k.sub.a) constants were determined by globally fitting (red) a simple 1:1 interaction model, A+B=AB, to the sensorgrams using the software supplied with the instrument. Global fitting results are summarised in table 10.

[0129] FIG. 20 shows double-reference corrected BLI binding curves (black), monitored on a surface of non-covalently immobilised HER2 at coated at a range of concentrations, presented to a concentration of wild-type (a), mutant 5 (b) and (c) mutant 19 antibodies, at 37° C. in running buffer. The apparent dissociation rate constant (kd) and association rate (k.sub.a) constants were determined by globally fitting (red) a simple 1:1 interaction model, A+B=AB, to the sensorgrams using the software supplied with the instrument. Global fitting results are summarised in tables 11, 12 and 13.

[0130] FIG. 21 displays the Expression yield of Fsn0503 variants expressed as mg of purified IgG by Octet immunoassay obtained per ml of transfected CHO supernatant. Wild-Type shown in orange.

[0131] FIG. 22 displays the melting point determination values (TM1 and TM2) of Fsn0503 variants in degrees centrigrade. Wild-Type is in position 1.

[0132] FIG. 23 Shows the measurement of proportion of monodisperse (non-aggregated) molecule of Fsn0503 variants by size-exclusion chromatography. Wild-Type shown in orange.

EXAMPLES

[0133] Materials and Methods

[0134] Library Design

[0135] An antibody library was generated for each of (i) a humanised anti-Cathepsin S antibody, Fsn0503h (Fusion Antibodies, Belfast) (Kwok et al. Molecular Cancer 2011, 10:147) and (ii) trastuzumab (Roche) by mimicking the natural somatic hypermutation in humans, searching for specific DNA sequence motifs RGYW in the 3′ to 5′ strand and WRCY in the 5′ to 3′ strand, recognised by the AID enzyme, responsible for making mutations within antibodies in the human body.

[0136] The DNA sequence has a point mutation introduced at the guanine in position 2 of the DNA sequence motif RGYW or the cytosine in position 3 of the DNA sequence motif WRCY, where the nucleotide is mutated to be any other nucleotide. The AID enzyme in humans introduces natural mutations at these positions. This can cause a possible change to the amino acid at a single point. All these possible mutations were identified throughout the antibody DNA sequence in both directions in the 3′ to 5′ and 5′ to 3′ strands to generate all the amino acids that would be possible naturally throughout the antibody sequence to form the initial library.

[0137] Each library's size was then curated by removing any recognised sequence liabilities, such as deamidation sites, isomerization sites, n-linked glycosylation sites and oxidation sites from the newly generated amino acid sequences.

[0138] Molecular Docking

[0139] For each library, the antibody prediction tool in Maestro 11.7 (Schrodinger) was used to perform homology modelling, to generate an antibody model based on the amino acid sequences for the heavy and light variable chain regions. The antigen was imported from PDB. In the case of Fsn0503h, the inventors imported the crystal structure of human Cathepsin S (Cat S) with a C25S mutation with a bound drug (PDB Code: 3MPE). In the case of the library for trastuzumab variants, the extracellular domain of human epidermal growth factor receptor 2 (HER2) (PDB Code: 1N8Z) was used.

[0140] The protein preparation wizard (Bioluminate, Schrodinger) was used to assign bond orders (using Chemical Component Dictionary (CCD) database), add hydrogens, create zero-order bonds to metals, create disulphide bonds, convert selenomethionines to methionines, fill in missing side chains and loops using Prime, and generating het states using Epik, for both the antibody and antigen models. The structures were further refined using ProtAssign (Bioluminate, Schrodinger) to define hydroxyl, asparagine, glutamine and histidine states. Waters with less than three bonds to non-waters were removed. Finally, the structures were minimized using the OPLS3e forcefield. (Bioluminate, Schrodinger).

[0141] Cathepsin S was docked to the surface of the antibody model using the protein-protein docking tool, Prime. Only the CDR regions of the antibody were considered for molecular docking. Non-CDR regions were masked. Using in vitro information about the epitope of Cat S, a suitable docked pose was selected based on rank and from inspection of shape complementarity and surface interactions (using the protein interaction analysis tool).

[0142] Likewise, for the trastuzumab variants, the extracellular domain of human epidermal growth factor receptor 2 (HER2) (PDB Code: 1N8Z) was docked to the surface of the antibody model using the protein-protein docking tool, Prime. Only the CDR regions of the antibody were considered for molecular docking. Non-CDR regions were masked. A suitable docked pose was selected based on rank and from our inspection of shape complementarity and surface interactions (using the protein interaction analysis tool).

[0143] Combinatorial Mutation Analysis

[0144] Residue scanning was performed on the docked pose of the antibody-antigen complex. Informed mutations were made to the antibody, avoiding highly conserved residues, to increase the affinity of the antibody for the antigen and to enhance stability.

[0145] Residue scanning was first done by generating models with a single amino acid variation from the original structure, and repeated (with the same mutations) for up to 6 simultaneous variations from the wild type Ab. The residue mutation tool calculated the stability and affinity of the mutants relative to the original wild type antibody-antigen complex.

[0146] The variants were then sorted by difference in affinity and difference in stability relative to the wild type. Results that scored below a threshold of -2 for difference in stability and difference in affinity were selected and ranked based on the combination of the two scores (prioritising difference in affinity).

[0147] The best variants of the Fsn503h antibody and of the trastuzumab antibody were synthesised and analysed in vitro.

[0148] Antibody Synthesis

[0149] Transient Transfection:

[0150] Suspension adapted ExpiCHO cells were routinely cultivated at 4−6×10.sup.6 cells/ml at 130 rpm, 37° C., 8% CO.sub.2, in ExpiCHO Expression Medium in 500 ml vented Erlenmeyer flasks. For each of the Fsn0503h variants, 1 μg/ml of DNA was diluted in 4% (v/v) OptiPRO SFM in a centrifuge tube. In a separate tube, 0.32% (v/v) ExpiFectamine was diluted in 3.7% OptiPRO SFM. The ExpiFectamine/OptiPRO mix was then added to the DNA/OptiPRO mix and incubated at room temperature for 3 minutes before adding to 25 ml ExpiCHO cells at a final density of 6×10.sup.6 cells/ml in 125 ml vented Erlenmeyer flasks. Each transfected culture was cultivated at 37° C., 8% CO.sub.2 and 130 rpm overnight. Twenty hours post transfection, cells were supplemented with 0.6% (v/v) ExpiCHO enhancer and 24% (v/v) ExpiCHO feed. Cultures were then transferred to incubators at 32° C., 5% CO.sub.2 and 130 rpm. Cultures were harvested by centrifugation at 4000 rpm for 40 minutes at 18° C.

[0151] Purification:

[0152] Two-step Fsn0503h WT and variant antibody purifications were performed using a Tricorn 5/50 column (GE) packed with 1 ml of MabSelect™ PrismA (GE) followed by a 10 ml (2×5 ml) Hitrap Desalting (Desalt) column (GE). The MabSelect™ PrismA affinity medium was chosen for its high mAb binding and specificity properties and its alkali tolerance for efficient Cleaning-in-place (CIP). All steps were performed at room temperature, using a flow-rate of 4 ml/min, unless otherwise stated. After loading (performed using the AKTA sample pump), the protein A column was washed (in reverse flow mode) with 10 column volumes (CV) of PBS followed by a one-step elution (in reverse flow mode) with 100 mM glycine, pH 3.0. The protein A eluate was collected in a 2 ml loop when the absorbance was above 120 mAU at 280 nm (AKTA equipped with a 10 mm flow cell) and injected immediately onto the pre-equilibrated Desalt column. The Desalt peak elution was collected in a 96-well-2 ml block at 2-8° C. when the eluate had an absorbance above 100 mAU. To avoid any cross-contamination, the automated process also included a CIP of both the affinity and desalt columns. CIP was performed between each sample, for all contact pathways, using 0.2M NaOH (reverse flow mode was used for column cleaning).

[0153] The level of expression was determined as the total yield of material, following purification per ml of culture media.

[0154] Trastuzumab variants were synthesised using similar techniques.

[0155] Affinity Ranking

[0156] Enzyme-Linked Immunosorbent Assay:

[0157] MaxiSorp 96 well plates were coated with the sixty six 0503 variants at 1 g/ml in PBS for 24 hrs at 4° C. To obtain the EC50 results, the variants were serially diluted from 1000 ng/ml to T1 ng/ml with PBS and coated in duplicate. The standard curve was prepared with the parental 0503 antibody coated at 1 μg/ml in PBS for 24 hrs at 4° C. After 24 hrs, the MaxiSorp plates were washed three times with PBS-T. 200 μl of SuperBlock was added to each well, removed and replaced three times. 100 μl of Cat S antigen at 200 ng/ml was added to each well and left shaking at RT, 150 rpm for 1 hr 30 mins. The plates were washed with PBS-T three times and dried. 100 μl of anti-his-HRP at 5 μg/ml was added to each well and left for 1 hr 30 mins shaking at 150 rpm, RT. The plates were washed three time PBS-T, once with PBS and dried. 100 μl of TMB was added to each well and incubated at 37° C. for 10 mins, which was followed by 50 μl of 1M HCL and the absorbance of the plates was measured at 450 nm.

[0158] Affinity Ranking Using the Octet RED96 System (Fsn503h Variants):

[0159] Affinity ranking assays were performed by first capturing IgG using anti-human Octet biosensors (ForteBio part no. 18-5060) followed by a baseline step of 2 minutes in HBS-EBT buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 1 mg/ml BSA, and 0.05% Tween-20, pH 7.4). The mAb capture biosensors were then submerged in wells containing 200 ng/ml of recombinant Cathepsin S antigen for 10 minutes (association step), followed by a 10 min dissociation step in running buffer. To allow for double reference correction, IgG-captured sensors were dipped into wells containing only buffer and blank sensors were also dipped into wells containing the antigen. This referencing provided a means of compensating for both the natural dissociation of the capture IgG and also non-specific binding of the antigen to the sensor surface. All steps were performed at 25° C. in HBS-EBT buffer at a constant flow-rate of 1000 rpm. New sensors were used for each sample. Dissociation rate constants (koff) were calculated using the ForteBio Data Analysis software. All consumables used were those recommended by ForteBio.

[0160] Antibody Quantification Using Biolayer Interferometry (the Octet RED96 System):

[0161] To measure IgG content, 200 μL aliquots of antibody standards (spanning 0.06 to 512 μg/ml) and IgG containing cell supernatants (diluted within the measurable range of calibration curve) were prepared in duplicate using 1× HBS-EBT buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 1 mg/ml BSA, and 0.05% Tween-20, pH 7.4) and placed in the wells of a 96-well black microtiter plate (Greiner Bio-One part no. 655209). All samples and standards were measured in duplicate using protein A Biosensors (Fortebio PN 18-5010). The plate was placed in the Octet and allowed to equilibrate to 25° C. in the thermostatted chamber. The run was initiated by placing the sensors in the wells and measuring the change in layer thick-ness (in nanometers, nm) with time, all under computer control. Data were taken for each set of eight samples at a time (one plate column is measured simultaneously) for 180-600 sec at a flow rate of 400-1000 rpm (orbital flow). Data were processed automatically using the Octet User Software version 3.1. The measurement time and flow rate were altered according to the sensitivity required.

[0162] Affinity Ranking Using Biolayer Interferometry (the Octet RED96 System) (Trastuzumab Variants):

[0163] Affinity ranking assays were performed by first capturing *IgG using anti-human Octet biosensors (ForteBio part no. 18-5060) followed by a baseline step of 2 minutes in HBS-P+ buffer (10 mM HEPES, 150 mM NaCl, 1 mg/ml BSA, and 0.05% Tween-20, pH 7.4). The mAb capture biosensors were then submerged in wells containing 5 nM of recombinant HER2 (Acro Biosystems; P/N. H5225) antigen for 15 minutes (association step), followed by a 20 min dissociation step in running buffer. To allow for double reference correction, IgG-captured sensors were dipped into wells containing only buffer and blank sensors were also dipped into wells containing the antigen. This referencing provided a means of compensating for both the natural dissociation of the capture IgG and also non-specific binding of the antigen to the sensor surface. Steps were performed at either 25° C. or 37° C. in HBS-EBT buffer at a constant flow-rate of 1000 rpm. New sensors were used for each sample. Dissociation rate constants (koff) were calculated using the ForteBio Data Analysis software. All consumables used were those recommended by ForteBio.

[0164] *To allow for similar loading levels the quantified (as described above) IgG containing cell supernatants were diluted to an identical concentration.

[0165] K.sub.D Determination Using Biolayer Interferometry (the Octet RED96 System):

[0166] Kinetic assays were performed by first capturing IgG using anti-human Fc Octet biosensors followed by two baseline steps of 2 minutes each in in HBS-P+ buffer running buffer. The mAb capture biosensors were then submerged in wells containing various concentrations of HER2 for 15 mins followed by 20 mins of dissociation time in running buffer. To allow for double reference correction, IgG-captured sensors were dipped into wells containing only buffer and blank sensors were also dipped into wells containing the antigen concentration series. This referencing providing a means of compensating for both the natural dissociation of the capture IgG and also non-specific binding of the antigen to the sensor surface. All steps were performed at 37° C. in kinetics buffer at a constant flow-rate of 1000 rpm.

[0167] Melting Point Determination

[0168] Antibodies commonly display 2 measurable melting points in analysis, designated TM1 and TM2 as a result of thermal denaturing of different parts of the assembled molecule. These values were determined by Thermal Shift Assay. Solutions containing 5 μl of Sypro Orange (diluted 1/200 in PBS, pH 7.4); Molecular Probes) and 45 μl of 0.3 mg ml.sup.−1 antibody were added to low profile PCR tubes (Bio-Rad; TLS0851). Tubes were sealed with optical ultra-clear sealing caps (Bio-Rad; TCS0803) and heated in an i-Cycler iQ5 real-time PCR detection system (Bio-Rad) from 20 to 90° C. in increments of 1° C. Fluorescence changes in the wells of the plate were monitored simultaneously with a charge-coupled (CCD) camera. The wavelengths for excitation and emission were 485 and 575 nm, respectively. The temperature midpoint for the protein unfolding transition, T.sub.m, was calculated using the Bio-Rad iQ5 software.

[0169] Determination of Monodispersity

[0170] The level of monodispersity, defined as free individual molecules of immunoglobulin within an antibody preparation, of each variant was also shown to differ from the wild type molecule. This measurement is commonly used as an indicator of aggregation propensity for antibody molecules. Aggregation is the tendency of protein molecules to associate into multimeric complexes, diminishing the solubility and activity of an antibody preparation over time and is a key attribute contributing to the stability of an antibody molecule in solution.

[0171] This was determined on the purified antibody solutions using size-exclusion chromatography. All samples were diluted to a final concentration of 0.1 mg/ml using phosphate buffered saline. Highly purified samples of antibodies were loaded independently onto a Superdex 200 increase 10/300 GL gel-filtration column. 50 μl samples were injected and the column flow rate was maintained at 0.75 μl min.sup.−1. Separations and equilibration steps were performed in phosphate buffered saline at 22° C. Protein peaks were monitored using absorbance at 280 & 214 nm and spectra were analysed using the Unicorn emulation software package (GE healthcare). Results are reported in V.sub.r (ml) and relative peak area (%) of each peak.

[0172] Results

Example 1 Cathepsin S antibody Variants

[0173] Library Design

[0174] The DNA sequences of the variable domains of the Fsn0503h antibody light and heavy light chains were analysed for motifs susceptible to mutation during somatic hypermutation, and the corresponding potential amino acid results of these mutation plotted above the parental sequence, as shown in FIG. 1. Any undesirable amino acids or STOP codons generated as a result of the DNA mutations were identified.

[0175] It was a surprising discovery that there are more mutations within the framework regions than the CDRs, and in particular that CDR-H3 has no functional mutations.

[0176] Molecular Docking

[0177] The parental Fsn0503h antibody variable domains (Kwok et al. Molecular Cancer 2011, 10:147) were docked with the Cathepsin S protein using Schrödinger molecular docking software as described in the Methods. The outcome of the docking procedure is shown in FIG. 2.

[0178] Combinatorial Mutagenesis

[0179] Within the docked structures amino acid residue mutations were introduced and the relative difference in both affinity and stability predicted. The mutations were introduced in increasing numbers until no further benefits to stability or affinity were predicted. The results of the combinatorial mutagenesis are shown in FIG. 3. The inventors identified 66 variants that had a predicted improvement in both affinity and stability. The DNA coding for these variants was then synthesised and the antibodies expressed and purified as described in the Methods.

[0180] ELISA

[0181] Each of the purified variants was analysed by ELISA for binding to recombinant Cathepsin S protein. The results (FIG. 4) show that 12 of the variants have a higher relative OD than the parental Fsn0503h, potentially indicating a higher affinity for the Cathepsin S target.

[0182] Affinity Ranking

[0183] In order to determine an accurate affinity comparison of each of the variants with the parental Fsn0503h antibody the interactions with Cathepsin S were measured by BLI using an Octet instrument as described in the Methods.

[0184] The results show that approximately 50% of the variants have improved affinity when measured against the average reading for the Fsn0503h antibody (FIG. 6). The association (Kon) and dissociation (Kdis) rates are given in FIG. 5.

[0185] Influence on Expression

[0186] The 66 variants expressed in CHO as described above also demonstrated a range of level of expression relative to the wild type. IgG level was determined by a quantitative human IgG immunoassay on the BLI Octet instrument following purification and expressed the total amount relative to the supernatant volume purified.

[0187] Modulation of Melting Point

[0188] The 66 variants also demonstrated a range of variation in stability characteristics associated with antibody molecules. This included changes in melting temperature profile at 2 melting points commonly observable in immunoglobulin molecules. The results of melting point determination are shown in FIG. 22. It is noteable that some variants such as Mut 6 and Mut 49 appear to have lost their distinct 2-phase melting pattern, with only a single melting temperature being observable.

[0189] Modulation of Monodispersity

[0190] This characteristic was analysed by analytical size exclusion chromatography of the 66 variants and show a range of values amongst the analysed variants as presented in FIG. 23.

Example 2—Trastuzumab Variants

[0191] Library Design

[0192] A library of antibody variants was produced for trastuzumab based on naturally occurring somatic hypermutation in humans. Using the method described above somatic hypermutation was replicated, producing naturally occurring mutations in the DNA sequence of trastuzumab and subsequently translated to their respective amino acid mutations, as illustrated in FIG. 9. These mutations were compiled into a library of potential mutations at specific sites in both the complementary determining regions (CDRs) as well as in the framework regions (FIGS. 7 and 8), in which potential amino acid results of the method are detailed next to the parental sequence. Any undesirable amino acids or STOP codons generated as a result of the DNA mutations were identified.

[0193] The fully resolved crystal structure of trastuzumab in complex with the extracellular domain of human epidermal growth factor receptor 2 (HER2) (PDB Code: 1N8Z) is illustrated in FIG. 10. The library of trastuzumab mutations illustrated in FIGS. 7 and 8 was used for mutant scanning. Mutants were scanned consecutively whereby a single amino acid mutation was first analysed, followed by two mutations up until six total mutations in the heavy and light chains and subsequently variants were ranked based on affinity and stability. As with the Fsn503h library, the Schrödinger molecular docking software was used as described in the Methods.

[0194] Within the docked structures amino acid residue mutations were introduced and the relative difference in both affinity and stability predicted. The mutations were introduced in increasing numbers until no further benefits to stability or affinity were predicted. The results of the combinatorial mutagenesis are detailed in the tables of FIGS. 11 and 12. A total of 558 potential variants for trastuzumab were found to have both improved affinity and improved stability. The top five ranked affinities for variants are illustrated in FIG. 13. Among these variants, a total of 5 mutations across the heavy and light chains were found to be the most common. The variants were scored relative to the parental antibody affinity using the predicted free energy change (MG).

[0195] Off-Rate Ranking

[0196] A panel of 89 Trastuzuamb variants were screened for binding to the HER2 antigen using biolayer interferometry (see materials and methods for details). Sensorgrams from 20 samples with the report points indicated as colour bars are shown in FIG. 14. FIG. 15 shows a scatterplot of the report points stability_early plotted against stability_late. The best binders (14 in total), with a high binding stability and slow dissociation, are highlighted with a blue circle. The off-rate ranking experiments were repeated for these mutants at 37° C., to provide more biologically relevant data (FIG. 16). Mutant 19 appeared to have enhanced stability compared to that of WT Trastuzumab and was taken forward for kinetic analysis.

[0197] KD Determination

[0198] All samples were diluted in freshly prepared running buffer. Either Trastuzumab or a variant (Mutant) were immobilised onto the surface of a series of biosensors using the capture methods described (see materials and methods). HER2 was passed over the surface to generate a binding response. Binding data for the HER2 interactions were collected at 37° C. on the biosensors. A dilution series of the HER2 antigen (5 nM-0.078 nM) was used in the association step, in order to fit results globally and get the best values for k.sub.a, k.sub.d, and K.sub.D. The response data for the binding of antigen to the surface immobilised IgG were fitted to a 1:1 binding model, yielding the data traces (red—see FIG. 17). The experiments were performed in duplicate and the average kinetic parameters are summarised in Table 9. For example, the data shows that Mutant 19 exhibits an apparent ˜2-fold affinity increase compared with that of the wild-type control and commercial Trastuzumab. The increase in affinity is primarily due to the slower Kd, 6.04E-05 compared to 1.20E-04 for the wild-type antibody. FIG. 18 schematically shows a comparison of actual versus predicted affinity for tratuzumab variants selected by in silico predictions of improved affinity and stability. Actual Octet affinity ranking correlates reasonably well for the top 20 predictions

TABLE-US-00001 TABLE 9 Kinetic parameters (fit to 1:1 interaction model) for Trastuzumab variants and HER2 interaction K.sub.D k.sub.a k.sub.d Mean Name (pM) (m.sup.−1s.sup.−1) (s.sup.−1) R.sup.2 X.sup.2 Rmax Trastuzumab  192 ± 21 1.10E+06 ± 4.50E+04 2.12E−04 ± 3.15E−05 0.9960 0.2185 0.2112 HER2 WT 126 ± 6 1.01E+06 ± 1.85E+04 1.28E−04 ± 7.50E−06 0.9946 0.1283 0.1693 MUT 5 139 ± 5 1.05E+06 ± 5.00E+04 1.46E−04 ± 1.50E−06 0.9977 0.1395 0.2553 MUT 19  52 ± 16 1.12E+06 ± 1.30E+05 6.04E−05 ± 2.47E−05 0.9958 0.0984 0.1405

[0199] R.sup.2 values indicate how well the fit and experimental data correlate and above 0.95 are considered a good fit; X.sup.2 is the sum of the squared deviation should be generally below 3 X.sup.2 is the measure of error between the experimental data and the fitted line. The smaller the X.sup.2 indicates a better fit.

[0200] Monomeric KD Determination

[0201] In order to further verify the affinity of the top variants and to understand the contribution of avidity to the measurement of affinity within the top 2 variants (MUT 5 and MUT 19), these molecules were prepared by enzymatic cleavage as Fragment Antigen Binding (Fab) fragments, consisting of the variable domain and CH1 constant domain of the heavy and light chain, in a monomeric form, with a single antigen binding domain per molecule.

[0202] An initial measurement was made after coating the probes with 2.5 μg/ml biotinylated Her2-recombinant Fc fusion protein. A range of concentrations of each of the 3 monomeric antibodies were assessed for binding to the Her2 surface. In this case double-referencing was not applied. The sensorgram data is shown in FIG. 19 and resulting kinetic calculations (on a 1:1 model) are listed in Table 10.

TABLE-US-00002 TABLE 10 Kinetic parameters (fit to 1:1 interaction model) for Trastuzumab variants as monomeric Fab and HER2 interaction at a HER2 coating concentration of 2.5 μg/ml K.sub.D Capture level k.sub.a k.sub.d K.sub.D K.sub.D Steady Mean Antibody (nm) (m.sup.−1s.sup.−1) (s.sup.−1) (M) (pM) State R.sup.2 X.sup.2 R.sub.max WT (4) 0.719 2.05E+06 6.17E−04 3.01E−10 300 6.40E−10 0.9513 0.141 0.059 Mut 5 (5) 0.853 1.14E+06 2.90E−04 2.56E−10 256 1.20E−10 0.9856 0.107 0.106 Mut 19 (3) 0.676 2.17E+06 1.13E−05 5.20E−12 5.2 1.00E−10 0.9848 0.0289 0.061

[0203] Sensors were then prepared by coating with 0.625, 1.25 or 2.5 μg/ml of biotinylated Her2-recombinant Fc fusion protein. The purified Fab fragments were then tested for their binding to these probes. The sensorgram traces are shown in FIG. 20 for these measurements, with details shown in Tables 11, 12 and 13. This data was double-referenced to exclude any drift due to non-specific interactions at the probe surface, unlike the previous experiment.

TABLE-US-00003 TABLE 11 Kinetic parameters (fit to 1:1 interaction model) for Trastuzumab variants as monomeric Fab and HER2 interaction at a HER2 coating concentration of 2.5 μg/ml (double-referenced) Sample K.sub.D K.sub.D Error K.sub.ON K.sub.ON Error K.sub.OFF K.sub.OFF Error R.sup.2 X.sup.2 R.sub.MAX WT 2.86E−10 <1.0E−12 1.18E 06 5.34E−65 3.37E−04 <1.0E−07 0.9938 0.0012 0.0921 MUT 19 1.88E−10 2.74E+56 1.19E+06 3.24E+68 2.22E−04 3.24E+62 0.9959 0.0014 0.1157 MUT 5 3.55E−10 <1.0E−12 1.29E+06 2.94E−65 4.59E−04 <1.0E−07 0.9872 0.0035 0.1138

TABLE-US-00004 TABLE 12 Kinetic parameters (fit to 1:1 interaction model) for Trastuzumab variants as monomeric Fab and HER2 interaction at a HER2 coating concentration of 1.25 μg/ml (double-referenced) Sample K.sub.D K.sub.D Error K.sub.ON K.sub.ON Error K.sub.OFF K.sub.OFF Error R.sup.2 X.sup.2 R.sub.MAX WT 3.12E−10 <1.0E−12 1.10E+06 1.73E−65 3.43E−04 <1.0E−07 0.9908 0.0012 0.0717 MUT 19 1.68E−10 3.79E+56 1.03E+06 3.91E+68 1.73E−04 3.91E+62 0.9914 0.0021 0.0917 MUT 5 4.50E−10 <1.0E−12 1.26E+06 2.24E−65 5.67E−04 <1.0E−07 0.9883 0.0021 0.0904

TABLE-US-00005 TABLE 13 Kinetic parameters (fit to 1:1 interaction model) for Trastuzumab variants as monomeric Fab and HER2 interaction at a HER2 coating concentration of 0.625 μg/ml (double-referenced) Sample K.sub.D K.sub.D Error K.sub.ON K.sub.ON Error K.sub.OFF K.sub.OFF Error R.sup.2 X.sup.2 R.sub.MAX WT 3.15E−10 <1.0E−12 1.00E+06 1.79E−65 3.16E−04 <1.0E−07 0.9831 0.0013 0.0552 MUT 19 <1.0E−12 <1.0E−12 1.08E+06 1.94E−62 <1.0E−07 N/A 0.9894 0.0017 0.0698 MUT 5 3.40E−10 <1.0E−12 1.34E+06 2.10E−65 4.57E−04 <1.0E−07 0.9787 0.0018 0.0642

[0204] The affinity value appears to vary as a result of coating concentration so it is difficult to assign a definitive value for the increase in affinity for MUT 5 and MUT 19 compared to the WT Trastuzemab molecule from these monomeric analysis. Without wishing to be bound by theory, this is thought to be due to the problem of working at the limits of sensitivity of the FortéBio Octet Biosensor instrument.

[0205] Conclusion

[0206] Using the methods of the present invention, the inventors have successfully demonstrated that the affinity of an antibody for its target can be improved without the need to generate a very large physical library of antibody variants and subsequent selection/screening process. In addition, a pool of variants are generated which, upon expression, display a range of attributes of interest including affinity, expression level and physiochemical characteristics, which are all of interest to the potential developability of an antibody molecule.

[0207] Although the invention has been particularly shown and described with reference to particular examples, it will be understood by those skilled in the art that various changes in the form and details may be made therein without departing from the scope of the present invention.