ANTIGEN BINDING PROTEINS WHICH BIND TO THE pMHC HLA-DQ2.5:DQ2.5 PRESENTING A GLIADIN PEPTIDE

20210147552 · 2021-05-20

    Inventors

    Cpc classification

    International classification

    Abstract

    The present invention relates generally to the field of antigen binding proteins such as antibodies, in particular those which bind to HLA-DQ2.5:DQ2.5-glia-α1a, or which bind to HLA-DQ2.5:DQ2.5-glia-α2. The invention further relates to compositions and immunoconjugates comprising such antibodies and to methods of producing such antibodies. The invention also relates to methods and uses which employ such antibodies, for example in the treatment of celiac disease.

    Claims

    1. An antigen binding protein which binds to HLA-DQ2.5:DQ2.5 presenting a gliadin peptide, said antigen binding protein comprising at least one light chain variable domain and at least one heavy chain variable domain, each domain comprising three complementarity determining regions (CDRs), wherein (a) said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α2 and comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:425 (GGTX.sub.4X.sub.5X.sub.6X.sub.7X.sub.8X.sub.9X.sub.10X.sub.11) wherein X.sub.4 is F or S or N or V or G or Y; X.sub.5 is S or T or Q or L or R or N; X.sub.6 is S or G or M or E or P; X.sub.7 is Y or F or G or R; X.sub.8 is A or G or I or Y or V or M; X.sub.9 is no amino acid or any amino acid, preferably no amino acid or G or Y or H or S; X.sub.10 is no amino acid or any amino acid, preferably no amino acid or A or G; and X.sub.11 is no amino acid or any amino acid, preferably no amino acid or A; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:427 (IIPIFGTX.sub.8) wherein X.sub.8 can be any amino acid, preferably A or V; a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:429 (ARX.sub.3X.sub.4X.sub.5X.sub.6X.sub.7X.sub.8X.sub.9X.sub.10X.sub.11X.sub.12X.sub.13X.sub.14X.sub.15X.sub.16X.sub.17X.sub.18X.sub.19X.sub.20) wherein X.sub.3 is D or V or G; X.sub.4 is V or A or Y or R or Q; X.sub.5 is Q or I or Y or N or P or V or G; X.sub.6 is R or G or Y or T or S or I or P or L; X.sub.7 is M or G or V or D or Y or P or L or I; X.sub.8 is G or For S or C or Y or P or W; X.sub.9 is M or F or S or O or T; X.sub.10 is D or G or A or S or Y or R; X.sub.11 is V or Y or L or G or E; X.sub.12 is no amino acid or any amino acid, preferably no amino acid or F or D or S or W or Y or L; X.sub.13 is no amino acid or any amino acid, preferably no amino acid or D or Y or C or F or G or V; X.sub.14 is no amino acid or any amino acid, preferably no amino acid or Y or M; X.sub.15 is no amino acid or any amino acid, preferably no amino acid or S or Y or D or F; X.sub.16 is no amino acid or any amino acid, preferably no amino acid or P or F or V or Q; X.sub.17 is no amino acid or any amino acid, preferably no amino acid or H or D; X.sub.18 is no amino acid or any amino acid, preferably no amino acid F or Y; X.sub.19 is no amino acid or any amino acid, preferably no amino acid or D; and X.sub.20 is no amino acid or any amino acid, preferably no amino acid or Y; a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:419 (X.sub.1X.sub.2X.sub.3X.sub.4X.sub.5X.sub.6X.sub.7X.sub.8X.sub.9X.sub.10X.sub.11X.sub.12) wherein X.sub.1 is Q or G; X.sub.2 is D or S or T or N; X.sub.3 is I or V or S; X.sub.4 is S or L or N or I; X.sub.5 is N or S or Y or D or T or K; X.sub.6 is W or N or S or V or Y; X.sub.7 is no amino acid or any amino acid, preferably no amino acid or S or G; X.sub.8 is no amino acid or any amino acid, preferably no amino acid or N or G; X.sub.9 is no amino acid or any amino acid, preferably no amino acid or N or Y; X.sub.10 is no amino acid or any amino acid, preferably no amino acid or K or G; X.sub.11 is no amino acid or any amino acid, preferably no amino acid or N or Y; and X.sub.12 is no amino acid or any amino acid, preferably no amino acid or Y; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:421 (X.sub.1X.sub.2S) wherein X.sub.1 can be any amino acid, preferably D or G or W; X.sub.2 can be any amino acid, preferably S or A or V; a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:423 (X.sub.1X.sub.2X.sub.3X.sub.4X.sub.5X.sub.6X.sub.7X.sub.8X.sub.9X.sub.10X.sub.11) wherein X.sub.1 is Q or S; X.sub.2 is Q or S or H; X.sub.3 is F or Y; X.sub.4 is N or Y or D or T; X.sub.5 is S or N or O or W; X.sub.6 is Y or W or T or S or L; X.sub.7 is P or G; X.sub.8 is L or T or P or R; X.sub.9 is no amino acid or any amino acid, preferably no amino acid or T or V or R; X.sub.10 is no amino acid or any amino acid, preferably no amino acid or L or F; and X.sub.11 is no amino acid or any amino acid, preferably no amino acid or T; or wherein (b) said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α1a and comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:5 (GDSVSSNSAA), or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:5; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:6 (TYYRSKWYN), or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:6; a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:417 (ARDX.sub.4X.sub.5X.sub.6GWX.sub.9X.sub.10YGMDV) wherein X.sub.4 can be any amino acid, preferably S or R; X.sub.5 can be any amino acid, preferably S or T; X.sub.6 can be any amino acid, preferably S or T; X.sub.9 can be any amino acid, preferably H or N or G; and X.sub.10 can be any amino acid, preferably P or A; a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:8 (HDISSY), or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:8; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:9 (AAS) or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:9; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:415 (QX.sub.2LNSYPLX.sub.9X.sub.10) wherein X.sub.2 can be any amino acid, preferably Q or D; X.sub.9 can be any amino acid or no amino acid, preferably no amino acid or L; and X.sub.10 can be any amino acid or no amino acid, preferably no amino acid or T.

    2. The antigen binding protein of claim 1, wherein (a) said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α2 and comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:426; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:428; a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:430; a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:420; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:422; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:424; or wherein (b) said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α1a and comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:5, or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:5; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:6, or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:6; a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:418; a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:8 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:8; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:9 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:9; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:520 or SEQ ID NO:416.

    3. The antigen binding protein of claim 1 or claim 2, wherein said antigen binding protein comprises a light chain variable domain that comprises a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:435; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:437; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:439.

    4. The antigen binding protein of any one of claims 1 to 3, wherein said antigen binding protein comprises a light chain variable domain that comprises a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:436; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:438; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:521 or SEQ ID NO:440.

    5. The antigen binding protein of any one of claims 1 to 4, wherein said heavy chain variable region comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:522 or SEQ ID NO:441 or; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:42 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:42, or SEQ ID NO:168 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:168; and a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:443.

    6. The antigen binding protein of any one of claims 1 to 5, wherein said heavy chain variable region comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:523 or SEQ ID NO:442; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:42 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:42, or SEQ ID NO:168 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:168; and a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:443.

    7. The antigen binding protein of any one of claims 1 to 6, wherein said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α2, said antigen binding protein comprising at least one heavy chain variable region that comprises three CDRs and at least one light chain variable region that comprises three CDRs, wherein said light chain variable region comprises: (a) a variable light (VL) CDR1 that comprises the amino acid sequence of SEQ ID NO:368 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:368, (b) a VL CDR2 that comprises the amino acid sequence of SEQ ID NO:369 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:369, and (c) a VL CDR3 that comprises the amino acid sequence of SEQ ID NO:370 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:370; and wherein said heavy chain variable region comprises: (d) a variable heavy (VH) CDR1 that comprises the amino acid sequence of SEQ ID NO:365 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:365, (e) a VH CDR2 that comprises the amino acid sequence of SEQ ID NO:366 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:366, and (f) a VH CDR3 that comprises the amino acid sequence of SEQ ID NO:367 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:367.

    8. The antigen binding protein of any one of claims 1 to 6, wherein said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α1a and comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:41, or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:41; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:42, or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:42; a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:431; a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:44 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:44; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:45 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:45; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:46 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:46.

    9. The antigen binding protein of any one of claims 1 to 6 or claim 8, wherein said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α1a and comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:41, or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:41; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:42, or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:42; a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:432; a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:44 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:44; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:45 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:45; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:46 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:46.

    10. The antigen binding protein of any one of claims 1 to 7, wherein said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α2 and comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:433; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:168 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:168; a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:169 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:169; a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:170 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:170; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:171 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:171; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:172 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:172.

    11. The antigen binding protein of any one of claims 1 to 7 or claim 10, wherein said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α2 and comprises a variable heavy (VH) CDR1 comprising the amino acid sequence of SEQ ID NO:434; a variable heavy (VH) CDR2 comprising the amino acid sequence of SEQ ID NO:168 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:168; a variable heavy (VH) CDR3 comprising the amino acid sequence of SEQ ID NO:169 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:169; a variable light (VL) CDR1 comprising the amino acid sequence of SEQ ID NO:170 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:170; a variable light (VL) CDR2 comprising the amino acid sequence of SEQ ID NO:171 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:171; and a variable light (VL) CDR3 comprising the amino acid sequence of SEQ ID NO:172 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:172.

    12. The antigen binding protein of any one of claims 1 to 6 or claims 8 or 9, wherein said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α1a, said antigen binding protein comprising at least one heavy chain variable region that comprises three CDRs and at least one light chain variable region that comprises three CDRs, wherein said light chain variable region comprises: (a) a variable light (VL) CDR1 that comprises the amino acid sequence of SEQ ID NO:116 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:116, (b) a VL CDR2 that comprises the amino acid sequence of SEQ ID NO:117 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:117, and (c) a VL CDR3 that comprises the amino acid sequence of SEQ ID NO:118 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:118; and wherein said heavy chain variable region comprises: (d) a variable heavy (VH) CDR1 that comprises the amino acid sequence of SEQ ID NO:113 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:113, (e) a VH CDR2 that comprises the amino acid sequence of SEQ ID NO:114 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:114, and (f) a VH CDR3 that comprises the amino acid sequence of SEQ ID NO:115 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:115.

    13. The antigen binding protein of any one of claims 1 to 6 or claims 8 or 9, wherein said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α1a, said antigen binding protein comprising at least one heavy chain variable region that comprises three CDRs and at least one light chain variable region that comprises three CDRs, wherein said light chain variable region comprises: (a) a variable light (VL) CDR1 that comprises the amino acid sequence of SEQ ID NO:503 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:503, (b) a VL CDR2 that comprises the amino acid sequence of SEQ ID NO:504 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:504, and (c) a VL CDR3 that comprises the amino acid sequence of SEQ ID NO:505 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:505; and wherein said heavy chain variable region comprises: (d) a variable heavy (VH) CDR1 that comprises the amino acid sequence of SEQ ID NO:500 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:500, (e) a VH CDR2 that comprises the amino acid sequence of SEQ ID NO:501 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:501, and (f) a VH CDR3 that comprises the amino acid sequence of SEQ ID NO:502 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:502.

    14. The antigen binding protein of any one of claims 1 to 6, or claims 10 or 11, wherein said antigen binding protein binds to HLA-DQ2.5:DQ2.5-glia-α2, said antigen binding protein comprising at least one heavy chain variable region that comprises three CDRs and at least one light chain variable region that comprises three CDRs, wherein said light chain variable region comprises: (a) a variable light (VL) CDR1 that comprises the amino acid sequence of SEQ ID NO:314 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:314, (b) a VL CDR2 that comprises the amino acid sequence of SEQ ID NO:315 or a sequence containing 1 amino acid substitution, addition or deletion relative to SEQ ID NO:315, and (c) a VL CDR3 that comprises the amino acid sequence of SEQ ID NO:316 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:316; and wherein said heavy chain variable region comprises: (d) a variable heavy (VH) CDR1 that comprises the amino acid sequence of SEQ ID NO:311 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:311, (e) a VH CDR2 that comprises the amino acid sequence of SEQ ID NO:312 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:312, and (f) a VH CDR3 that comprises the amino acid sequence of SEQ ID NO:313 or a sequence containing 1, 2 or 3 amino acid substitutions, additions or deletions relative to SEQ ID NO:313.

    15. The antigen binding protein of any one of claims 1 to 14, wherein said antigen binding protein is an antibody.

    16. A composition comprising an antigen binding protein of any one of claims 1 to 15 and a diluent, carrier or excipient.

    17. An immunoconjugate comprising an antigen binding protein of any one of claims 1 to 15 attached to a therapeutic or diagnostic agent.

    18. A nucleic acid comprising a nucleotide sequence that encodes an antigen binding protein of any one of claims 1 to 15.

    19. An antigen binding protein of any one of claims 1 to 15 for use in therapy or diagnosis.

    20. An antigen binding protein of any one of claims 1 to 15 for use in the treatment or diagnosis of celiac disease.

    21. A method of treating celiac disease which method comprises administering to a patient in need thereof a therapeutically effective amount of an antigen binding protein of any one of claims 1 to 15.

    22. Use of an antigen binding protein of any one of claims 1 to 15 in the manufacture of a medicament for use in the treatment of celiac disease.

    23. A method of diagnosing celiac disease in a mammal comprising the steps of: (a) contacting a test sample taken from said mammal with one or more of the antigen binding proteins of any one of claims 1 to 15; (b) measuring the presence and/or amount and/or location of antigen binding protein-antigen complex in the test sample; and, optionally (c) comparing the presence and/or amount of antigen binding protein-antigen complex in the test sample to a control.

    24. A method of detecting HLA-DQ2.5:DQ2.5-glia-α1a or HLA-DQ2.5:DQ2.5-glia-α2, comprising contacting a composition suspected of containing HLA-DQ2.5:DQ2.5-glia-α1a or HLA-DQ2.5:DQ2.5-glia-α2 with an antigen binding protein of any one of claims 1 to 15 under conditions effective to allow the formation of HLA-DQ2.5:DQ2.5-glia-α1a or HLA-DQ2.5:DQ2.5-glia-α2/antigen binding protein complexes and detecting the complexes so formed.

    25. A method of producing an antigen binding protein of any one of claims 1 to 15, comprising the steps of (i) culturing a host cell comprising one or more of the nucleic acids of claim 18 under conditions suitable for the expression of the encoded antibody or protein, and (ii) isolating or obtaining the antigen binding protein from the host cell or from the growth medium/supernatant.

    26. The method of claim 25, further comprising a step of purification of the antigen binding protein product and/or formulating the antigen binding protein into a composition including at least one additional component, such as a pharmaceutically acceptable carrier or excipient.

    Description

    [0771] FIG. 1. Screening of HLA-DQ2.5:DQ2.5-glia-ala-specific binders and affinity measurements. (A) Representative ELISA (n=2) of normalized, rescued phage outputs from R2 and R3 analyzed for binding to HLA-DQ2.5 with either DQ2.5-glia-α1a or CLIP2 peptides. Phage displaying an irrelevant specificity (scFv anti-NIP) was included as control. mAb 2.12.E11 specific for the DQ2.5 β-chain was used to assess capture-level of pMHC. Error bars illustrate mean±SD of duplicates. (B) ScFvs selected in R2, R3, and R4 were batch-cloned into a vector for soluble expression and random clones expressed and analyzed for target reactivity by ELISA. Clones preferentially binding HLA-DQ2.5:DQ2.5-glia-α1a compared to HLA-DQ2.5:CLIP2 were chosen and sequenced. IGHV and IGKV gene segment usage was identified from the IMGT database. The pie charts show the gene segments used by the 11 unique clones. (C) The unique clones were expressed and purified and binding affinity to HLA-DQ2.5:DQ2.5-glia-α1a was determined by single cycle kinetics using a 3-fold concentration series ranging from 2 μM-0.025 μM scFv. Representative sensograms of clones R2A1-8E, R3A2-9F, and R4A1-3A which bound specifically and are shown as indicated (n=2-3). KDs were derived by fitting the data to a 1:1 Langmuir model. Steady state affinity evaluations are shown as inset figures. NA=not available.

    [0772] FIG. 2. The HLA-DQ2.5:DQ2.5-glia-ala-specific mAbs are highly specific. The three hIgG1 mAbs and isotype control mAb were reformatted to hIgG1, expressed by transient transfection in human 293E cells and purified from supernatants before assessment of specificity. (A and B) Competition ELISAs where the mAbs were pre-incubated with titrated amounts of (A) soluble pMHCs, HLA-DQ2.5:DQ2.5-glia-α1a and HLA-DQ2.5:DQ2.5-glia-α2, or the corresponding free peptides, or (B) 33mer α-gliadin before assessment of ability to compete with binding to plate-bound HLA-DQ2.5:DQ2.5-glia-α1a (n=3). (C) Eight different HLA-DQ2.5:gluten peptide complexes and HLA-DQ2.5:CLIP2 were used in ELISA for specificity analysis (n=2). mAb 2.12.E11 was included to control pMHC capture levels. In each set of 9 bars, moving from left to right, bar 1 is HLA-DQ2.5-glia-α1a, bar 2 is CLIP2, bar 3 is HLA-DQ2.5-glia-γ1, bar 4 is HLA-DQ2.5-glia-γ2, bar 5 is HLA-DQ2.5-glia-γ3, bar 6 is HLA-DQ2.5-glia-γ4c, bar 7 is HLA-DQ2.5-glia-ω1, bar 8 is HLA-DQ2.5-glia-ω2, bar 9 is HLA-DQ2.5-glia-α2.

    [0773] FIG. 3. Mapping the fine-specificity and the structural basis for specificity. (A) Flow cytometric analysis of A20 B cells expressing HLA-DQ2.5 with covalently coupled DQ2.5-glia-α1a or CLIP2 peptides stained with the hIgG1 mAbs or hIgG1 isotype control mAb. Bound mAbs were detected using a biotinylated secondary anti-human IgG1 followed by streptavidin-RPE (n=2). (B) The hIgG1 mAbs were used to stain a panel of either HLA-DQ2.5:peptide or HLA-DQ2.2:peptide expressing A20 B cells and binding was analyzed by flow cytometry. Data are shown as the ratio median fluorescent intensity (MFI) of the hIgG1 mAbs compared to hIgG1 isotype control mAb (n=2). For each set of 3 bars, moving from left to right, bar 1 represents the mAb R2A1-8E, bar 2 represents the MAb R3A2-9F and bar 3 represents the MAb R4A1-3A. (C) Peptide alignment of DQ2.5-glia-α1a (forest green) and DQ2.5-glia-ω1 (cyan). Residues differing are underlined in the peptide sequences. Based on the crystal structure of HLA-DQ2.5:DQ2.5-glia-α1a (PDB ID 1S9V) [Kim, C. Y., et al., 2004]. (D) Overlay of the top three docking models of the Fvs onto HLA-DQ2.5:DQ2.5-glia-α1a [Kim, C. Y., et al., 2004]. Peptide residues that were mutated in the fine-specificity analysis and the HLA-DQ2.5/HLA-DQ2.2 polymorphisms are illustrated (aY22 and aS72). (E) In all three models V.sub.H CDR3 is placed close to p7. (F) CDR1 and CDR3 of V.sub.L both contain residues in close proximity to p9 in two of the Fv models. (G) In two of the Fv models V.sub.L CDR1 (D28 and S36) is placed in close proximity to aS72, with potential H-bond formation. (D-G) Coloring in molecular structures as follows: V.sub.H and V.sub.L, black; MHCa, grey; MHC, light orange; DQ2.5-glia-α1a, forest green; CDR1 and CDR2 of V.sub.H, deep purple; CDR3 of V.sub.H, red; CODR1 and CDR2 of V.sub.L, deep teal; CDR3 of V.sub.L, blue; aS72 and aY22, hot pink; p7 and p9, pale green; H-bonds, yellow dashes. (E-G) Residues within 5 Å of p7, p9 and aS72, respectively, are shown.

    [0774] FIG. 4. Specific detection of gluten peptide presentation in context of HLA-DQ2.5. (A) Monocytes isolated from human PMBCs were in vitro differentiated to monocyte-derived DCs, loaded with peptide and stained with hIgG1 mAb R3A2-9F or isotype control mAb before flow cytometric analysis (n=1). (B) Single-cell suspensions of intestinal biopsies obtained from patients undergoing gastroduodenoscopy were stained with a panel of antibodies to phenotypically characterize DQ2.5-glia-α1a presenting cells along with R3A2-9F mIgG2b. Bound R3A2-9F was detected using a FITC-conjugated secondary anti-mouse IgG2b Ab. Samples from three HLA-DQ2.5.sup.+ untreated CD (UCD) patients with Marsh 3B/C were run in parallel. The mean percent of mIgG2b mAb R3A2-9F positive cells compared to no primary antibody is shown ±SD. Mφ, macrophages.

    [0775] FIG. 5. PCs and B cells of gut biopsies present the DQ2.5-glia-α1a peptide. Detection of DQ2.5-glia-α1a presentation among PCs and B cells in single-cells suspension prepared from intestinal biopsies from either untreated CD (UCD) or treated CD (TCD) patients or healthy controls. mIgG2b mAb R3A2-9F or R4A1-3A were used for detection and percent positive cells was determined relative to use of secondary antibody alone. (A) Representative flow cytometric gating strategy to identify PCs and B cells from single-cell suspensions. (B) Percentage of specific HLA-DQ2.5:DQ2.5-glia-α1a detection among CD19.sup.+ PCs, CD45.sup.+ PCs, CD45.sup.− PCs, and B cells in HLA-DQ2.5.sup.+ UCD CD patients (n=18) compared to controls (grouped healthy and non-HLA-DQ2.5.sup.+CD patients, n=15). Two-tailed P-values are shown (unpaired t-test). (C) Stratification of the control patients among the CD19.sup.+ PCs from (A). Ctrl HLA-DQ2.5.sup.+ (n=5), Ctrl HLA-DQ2.5.sup.− (n=5), UCD HLA-DQ2.5.sup.+ (n=18), TCD HLA-DQ2.5.sup.+ (n=3), UCD HLA-DQ8.sup.+ (n=1), and UCD HLA-DQ2.2.sup.+ (n=1). (D) The HLA-DQ2.5.sup.+ UCD patients (n=18) were stratified according to Marsh score as indicated. (B-D) Non-CD ctrl patients did not have mucosal alterations. Each data point represents an individual subject. Red (i.e. horizontal) bars indicate mean percentage.

    [0776] FIG. 6. DQ2.5-glia-α1a presenting PCs express TG2-specific IgA and MHC class II. (A and B) PC subsets were sorted by flow cytometry from single-cells suspensions from HLA-DQ2.5.sup.+ UCD patients (n=3) all with positive serum anti-TG2 IgA titers and with marsh score 3B/C. (A) Representative micrographs of sorted PCs subsets as indicated. Two individual cells within each group are shown (B) Representative TG2-specific ELISPOT using the sorted PC subsets as indicated. TG2-specific IgA autoantibodies were captured onto TG2-coated plates and detected using AP-conjugated anti-IgA Ab. T cells were used as negative control. (C) Percentage MHC class expression among APC present in single-cell suspensions from HLA-DQ2.5.sup.+ UCD patients (n=4). Each data point represents an individual subject; Mo, monocytes; Mφ, macrophages; two-tailed p-values from unpaired t-test; *, P≤0.05; **, P≤0.01; ***, P≤0.001; ****, P≤0.0001; ns, not significant.

    [0777] FIG. 7. Validation of purified scFv clones and SPR binding analysis to control pMHCs. The 11 unique scFv clones were expressed and purified from E. coli periplasmic fractions. (A) SDS-PAGE gel analysis of the scFv clones which specifically bound DQ2.5:DQ2.5-glia-ala. Gels were run after purification by IMAC and size exclusion chromatography under non-reducing and reducing conditions. scFv size of approx. 30 KDa is indicated. (B) Analytical gel filtration profiles of the candidate scFv clones and HLA-DQ2.5:DQ2.5-glia-α1a as indicated after a freeze/thaw cycle as the samples were subjected to prior to SPR analysis. (C and D) Representative SPR sensograms of the candidate scFv clones for binding to (C) HLA-DQ2.5:CLIP2 (n=2) and (D) HLA-DQ2.5:DQ2.5-glia-α2 (n=2). (E and F) Overlays of the DQ2.5-glia-α1a peptide with (E) CLIP2 and (F) DQ2.5-glia-α2. Peptide sequences are indicated and the residues differing are underlined. Based on the crystal structures of HLA-DQ2.5:DQ2.5-glia-α1a (PDB: 1S9V) and HLA-DQ2.5:DQ2.5-glia-α2 (PDB: 4OZH) [Kim, C. Y., et al., 2004; Petersen, J., et al., 2014].

    [0778] FIG. 8. Reformatting of candidate clones to hIgG1 and SPR binding analysis. The three specific clones were reformatted to hIgG1, expressed by transient transfection of human 293E cells and purified from supernatants along with isotype control hIgG1 mAb (anti-NIP). (A) SDS-PAGE gel of purified hIgG1 mAbs run under non-reducing and reducing conditions. Appropriate bands at approx. 150 KDa for full-length hIgG1 and bands at 50 KDa and 23 KDa (reduced heavy and light chains, respectively) are indicated. (B) SPR sensograms of hIgG1 mAbs along with the corresponding scFv fragments were run over HLA-DQ2.5:DQ2.5-glia-α1a to validate gain in functional affinity after reformatting to full-length mAbs.

    [0779] FIG. 9. Fine-specificity assessment. (A) Flow cytometric analysis of A20 B cells transduced to express HLA-DQ2.5 with covalently coupled DQ2.5-glia-α1a or CLIP2 peptides stained with biotinylated mAb 2.12.E11 or isotype control mAb, followed by RPE-conjugated streptavidin (n=2). (B) Flow cytometric assessment of the pMHC expression level of the panel of A20 B cells transduced with either HLA-DQ2.5 or HLA-DQ2.2 with covalently coupled peptide. Q indicated native (glutamine) DQ2.5-glia-α1a epitope. Unless specified, all epitopes are in the deamidated forms. All cells were stained with biotinylated mAb 2.12.E11 followed by streptavidin-RPE (n=2). (C and D) Representative SPR sensograms showing binding to (C) HLA-DQ2.5:DQ2.5-glia-ω1 (n=2) and (D) HLA-DQ2.2:DQ2.5-glia-α1a (n=1) after capture of pHLA on sensor chips and injection of scFv clones as indicated. (E) SPR binding analysis of the DQ2 conformational-specific mAb SPV-L3 to evaluate the conformational integrity of HLA-DQ2.5:DQ2.5-glia-α1a, HLA-DQ2.5:DQ2.5-glia-ω1, HLA-DQ2.5:DQ2.5-glia-ω2, and HLA-DQ2.2:DQ2.5-glia-α1a as indicated after binding experiments.

    [0780] FIG. 10. Construction of mIgG2b mAbs and flow cytometric analysis of single-cell suspensions from CD patient biopsies. (A) SDS-PAGE gels of the mAbs R3A2-9F and R4A1-3A and isotype control mAb after reformatting to mIgG2b and purification from supernatants of transfected HEK293E cells. Full-length mIgG2b of approx. 150 KDa run under non-reducing conditions and separated heavy and light chains at approx. 50 KDa and 23 KDa run under reducing conditions are indicated. (B) Representative ELISA showing retained specificity of mIgG2b mAbs R3A2-9F and R4A1-3A after reformatting (n=2). mAb 2.12.E11 was included to control pMHC capture levels. (C) The figure is based on FIG. 3b showing detection of HLA-DQ2.5:DQ2.5-glia-α1a using mAb R3A2-9F with or without use of FcR block. Single-cell suspensions of intestinal biopsies from 3 patients all being HLA-DQ2.5.sup.+ with Marsh 3B/C were run in parallel.

    [0781] FIG. 11. Flow cytometric gating strategy and analysis of PCs and B cells presenting DQ2.5-glia-α1a peptide. Single-cell suspensions were prepared from intestinal biopsies, cells were stained with indicated antibodies and immediately analyzed by flow cytometry. (A) Representative gating strategy for detection of gluten peptide presentation is shown. FSC-A, FSC—H and SSC-W were used to gate out doublet cells. (B) Stratification of the control patients among the CD45.sup.+ PCs, CD45.sup.− PCs and B cells from FIG. 4b. Ctrl HLA-DQ2.5.sup.+ (n=5), Ctrl HLA-DQ2.5.sup.− (n=5), UCD HLA-DQ2.5.sup.+ (n=18), TCD HLA-DQ2.5.sup.+ (n=3), UCD HLA-DQ8.sup.+ (n=1), and UCD HLA-DQ2.2.sup.+ (n=1). mIgG2b mAb R3A2-9F or R4A1-3A were used for detection and percent positive cells was determined relative to use of secondary antibody alone. Each data point represents an individual subject; non-CD ctrl patients did not have mucosal alterations; red (i.e. horizontal) bars indicate mean percentage.

    [0782] FIG. 12. Gating strategy for detection of MHC class II on APC subsets. Single-cell suspensions prepared from intestinal biopsies were stained with indicated antibodies and immediately analyzed by flow cytometry. Representative gating strategy for detection of MHC class II on PCs, B cells and bulk DCs, monocytes and macrophages are shown. HLA staining (red) is overlaid isotype control staining (black). FSC-A and FSC-H were used to gate out doublet cells.

    [0783] FIG. 13. SPR experiments with candidate scFvs: scFvs were analyzed in SPR using a single cycle kinetics method. They were tested for binding to HLA-DQ2.5:DQ2.5-glia-α2. They were tested for cross-reactivity with HLA-DQ2.5:DQ2.5-glia-ala and HLA-DQ2.5:DQ2.5-w2. The scFv concentrations varied for different candidates but were the same for all three antigens.

    [0784] FIG. 14. Different HLA-DQ2.5:gluten peptide complexes and HLA-DQ2.5:CLIP2 were used in ELISA for specificity analysis of HLA-DQ2.5:DQ2.5-glia-α2 antibodies. The peptides are annotated according to Sollid, L. M., et al., (Immunogenetics, 2012, 64(6): 455-460) (A). An antibody specific for HLA-DQ2.5:DQ2.5-glia-α1a and an antibody with irrelevant specificity were used as positive and negative controls (B). mAb 2.12.E11 specific for the DQ2 β-chain was included to control pMHC capture levels (C). Error bars illustrate mean±SD of duplicates. The biotinylated pMHCs captured were: HLA-DQ2.5:DQ2.5-glia-α2 (native (P4Q)), HLA-DQ2.5:DQ2.5-glia-α2 (deamidated P4E), HLA-DQ2.5:DQ2.5-glia-ω2, HLA-DQ2.5:DQ2.5-α1a, HLA-DQ2.5:DQ2.5-hor3, HLA-DQ2.5:DQ2.5-glia-γ2, and HLA-DQ2.5:CLIP2. HSA=human serum albumin; BSA=bovine serum albumin.

    [0785] FIG. 15. ELISA against different pMHC complexes with affinity matured clones (scFv): Binding of the affinity matured antibodies to HLA-DQ2.5:CLIP, HLA-DQ2.5:DQ2.5-glia-ala, and HLA-DQ2.5:DQ2.5-glia-α2. 2.12.E11 was used to control for functionality of pMHC molecules and similar levels of pMHC immobilization. In each set of 3 bars, moving from left to right, bar 1 represents HLA-DQ2.5:DQ2.5-glia-aa, bar 2 represents HLA-DQ2.5:DQ2.5-glia-α2, bar 3 represents HLA-DQ2.5:CLIP.

    [0786] FIG. 16. SPR experiments with affinity matured clones (scFv): Binding kinetics of the affinity matured variants towards the two gliadin complexes (HLA-DQ2.5:DQ2.5-glia-α1a (over a1a) and HLA-DQ2.5:DQ2.5-glia-α2 (over a2)) was analysed in SPR. All curves are normalized to the mother clones. The affinity matured scFv bound their targets and showed different off-rates. All of them showed improved off-rates compared to the mother clone. None of them was cross-reactive to the other α-gliadin complex (only depicted for 12.F6, 3.F6, 15.A6, and 4.7C).

    [0787] FIG. 17. Biophysical characterization of antibodies. A+B: Fab fragments were ranked based on off-rates in SPR, with the clone 4.7C marked with a single asterisk (*) and the clone 3.C11 marked with a double asterisk (**) (A: binding to HLA-DQ2.5:DQ2.5-glia-α1a, B: binding to HLA-DQ2.5:DQ2.5-glia-α2). C+D: Melting temperatures of the mother clones and the affinity matured Fab fragments with the clone 4.7C marked with a single asterisk (*) and the 3.C11 clone marked by a double asterisk (**) (C: HLA-DQ2.5:DQ2.5-glia-α1a, D: HLA-DQ2.5:DQ2.5-glia-α2). E-G: Representative sensorgrams of 4.7C (E), and RF117 (F) binding to HLA-DQ2.5:DQ2.5-glia-α1a, and 3.C11 binding to HLA-DQ2.5:DQ2.5-glia-α2 (G) (n≥2). RF117 denotes a mutant combining the sequence of 5.6A with the CDR H3 of 4.7C. H: The leads were reformatted to full-length hIgG1 and analyzed in ELISA against a panel of related soluble peptide:HLA-DQ2.5 complexes. In each set of 6 bars, from left to right, bar 1 is (a), bar 2 is (c), bar 3 is (b), bar 4 is (d), bar 5 is (e) and bar 6 is (f).

    [0788] FIG. 18. Antibodies stain engineered A20 mouse B cells. A20 cells were engineered to express HLA-DQ2.5 with covalently linked peptide. They were stained with the mother clones and the high affinity variants (n=2). A: Representative histograms are shown for the high affinity variants. B: Median fluorescence intensities are shown for all antibodies.

    [0789] FIG. 19. Antibodies stain plasma cells from celiac disease small intestinal biopsies. Single cell suspensions were prepared from either untreated HLA-DQ2.5.sup.+ celiac disease patients (i.e. on a gluten-containing diet, n=8) or controls with a healthy mucosal histology (n=3). (A and B) Cells were gated as live, large lymphocytes, CD3−, CD11c−, CD14−, CD38+, CD27+, CD19+, CD45+ plasma cells in (A) celiac patients and (B) controls. Bound mIgG2b antibodies were detected with an Alexa-546-conjugated secondary antibody and the frequency of positive cells was calculated and compared to use of an isotype control antibody (isotype). The secondary antibody only (FMO) is also shown. Each celiac disease patient is shown in a unique symbol.

    [0790] FIG. 20. Determination of TCR-reconstructed SKW3 T cell peptide sensitivity and assessment of anti-pMHC mAb inhibition capacity. (A-C) Representative dose-response curves (n=2) of T cell activation measured as % CD69-postive cells following co-cultivation of peptide-pulsed B cells (dump gated on CD19 expression) and the SKW3 T cells 380 (A), S16 (B) and 364 (C). Error bars illustrate mean±SD of duplicates from one of two independent assays, and estimated EC50, quality of fit (R.sup.2) values and the peptide used in pulsing are given within each graph. Based on the dose-response curves for each individual T cell, peptide concentrations of the indicated peptides resulting in about 60% activation were chosen baseline for an Ab inhibition assay (D-F), where the SKW3-380 (D), SKW3-S16 (F) and SKW3-364 (F) were treated as indicated with either 1 μM of the anti-pMHC mAbs, or 0.1 μM pan anti-HLA mAbs. The presented data are given as the % activation normalized to the absolute T cell activation in the absence of Ab, which was set to 100%. Error bars illustrate mean±SD of duplicates.

    EXAMPLES

    Example 1

    [0791] Identification of Antibodies that Specifically Bind to HLA-DQ2.5:DQ2.5-Glia-α1a

    Results

    [0792] Phage Selection of Recombinant Antibodies to HLA-DQ2.5 with Bound DQ2.5-Glia-α1a

    [0793] To isolate HLA-DQ2.5:DQ2.5-glia-α1a specific binders, we performed phage selections using a naïve, fully-human scFv-phage library [Loset, G. A., et al., 2005]. We performed four rounds (R1-R4) of selection using recombinant soluble, biotinylated pMHC (peptide MHC complex) with covalently linked peptide. After R3, the polyclonal library outputs showed preferential HLA-DQ2.5:DQ2.5-glia-ala reactivity compared to HLA-DQ2.5:CLIP2, with a large increase in antigen reactivity from R2 to R3 (FIG. 1A). To remove low affinity clones, we performed a stringent R4 before single-clone scFv binding analysis of R2, R3 and R4 outputs. A total of 75 independent clones reacted preferentially with HLA-DQ2.5:DQ2.5-glia-α1a (Table S1), representing 11 unique clones, with a preferential usage of IGHV6-1, as well as IGKV1-9 and IGKV1-39 (FIG. 1B).

    [0794] We next expressed and purified all 11 unique scFv clones in E. coli (FIGS. 7A and B, data not shown) and performed SPR to analyze binding affinity and specificity using HLA-DQ2.5 with DQ2.5-glia-α1a or the control peptides, DQ2.5-glia-α2 and CLIP2. Several scFvs bound pMHC, and three (R2A1-8E, R3A2-9F (also referred to as 106) and R4A1-3A (also referred to as 107)) bound specifically to HLA-DQ2.5:DQ2.5-glia-α1a with a monomeric affinity ranging in the nanomolar range (FIG. 1C, FIG. 7C-E, Table S2). R3A2-9F was highly enriched, constituting 60 out of 75 binding clones in the soluble scFv screen. Although R3A2-9F and R4A1-3A were found to differ by only one amino acid, R4A1-3A appeared only once among the screened clones (Table S1).

    [0795] The Selected Antibodies are Highly Specific Towards HLA-DQ2.5:DQ2.5-Glia-α1a

    [0796] To increase the functional affinity of the interaction, we reformatted and expressed the three binders as human IgG1 (hIgG1) mAbs (FIG. 8A). SPR confirmed pMHC-specific binding and a substantial gain in avidity, resulting in approximately 160-fold increase in half-life (FIG. 8B). To confirm a requirement for DQ2.5-glia-α1a peptide recognition strictly in the context of HLA-DQ2.5, we performed competition ELISA using soluble pMHC and free peptide. Indeed, only soluble HLA-DQ2.5:DQ2.5-glia-α1a, and not peptide alone, competed with the plate-bound complex for binding to the mAbs (FIG. 2A). Of note, DQ2.5-glia-ala provided as part of a 33mer peptide fragment which binds efficiently to HLA-DQ2.5 [Shan, L., et al., 2002], was not able to inhibit mAb binding to pMHC (FIG. 2B).

    [0797] Next, we extended the specificity analysis with 7 HLA-DQ2.5-gluten-peptide complexes in ELISA. This panel included common epitopes from γ- and ω-gliadin to which CD patients mount T-cell responses. None of the mAbs bound any of the complexes other than HLA-DQ2.5:DQ2.5-glia-α1a (FIG. 2C), not even the highly similar DQ2.5-glia-ω1, which differs from DQ2.5-glia-α1a in p7 and p9 only. Taken together, these results show that the mAbs exclusively recognize DQ2.5-glia-α1a bound to HLA-DQ2.5 and are not cross-reactive with HLA-DQ2.5 in complex with the other gluten peptides tested.

    [0798] Mapping Fine-Specificity of the Candidate mAbs

    [0799] To validate mAb binding to pMHC on cells, we utilized murine A20 B cells transduced with HLA-DQ2.5 with covalently linked DQ2.5-glia-α1a or CLIP2 peptides. When assessed for binding, all mAbs bound specifically to cells displaying the DQ2.5-glia-α1a epitope, while none bound CLIP2 (FIG. 3A and FIG. 9A).

    [0800] To further map fine-specificity, we screened for binding against a panel HLA-DQ2.5:peptide or HLA-DQ2.2:peptide expressing A20 B cells (FIG. 9B). Covalent attachment of the peptides to MHC largely eliminates effects of differences in peptide off-rates, enabling comparative assessment of binding. None of the mAbs bound the highly similar DQ2.5-glia-ω1 epitope or HLA-DQ2.2 with DQ2.5-glia-α1a (FIG. 3B). We also corroborated this finding with SPR using the mAbs and soluble, recombinant pMHCs (FIG. 9C-E). DQ2.5-glia-α1a and DQ2.5-glia-ω1 differ in p7 and p9 only (FIG. 3C). Thus, we constructed pL7Q and pY9F variants of DQ2.5-glia-α1a to resemble DQ2.5-glia-ω1 in these positions. All three mAbs bound the pL7Q variant, albeit not as strongly as DQ2.5-glia-α1a (FIG. 3B). However, while mAb R2A1-8E bound the pY9F variant, mAbs R3A2-9F and R4A1-3A did not (FIG. 3B). Of the polymorphic residues that differ between HLA-DQ2.5 and HLA-DQ2.2, the α72 residue is the only one in position for direct interactions (FIG. 3D). To map a potential effect of the HLA-DQ2.5 residue, we constructed the HLA-DQ2.5:DQ2.5-glia-α1a αS721 mutant (S in HLA-DQ2.5 and I in HLA-DQ2.2). The mAb R2A1-8E was the only one to bind the αS721 variant. As before, we did not observe binding to CLIP2 or DQ2.5-glia-α2 (FIG. 3B). Furthermore, the native, non-deamidated DQ2.5-glia-α1a (DQ2.5-glia-ala-Q) was not recognized.

    [0801] To understand the molecular basis for the observed specificity of the mAbs, we built Fv homology models using the V region sequence of mAb R4A1-3A. These models represent the highly similar mAbs R3A2-9F and R4A1-3A, but not mAb R2A1-8E, which differs in sequence and thus cannot be rationalized based on the models. We then docked the models to the available crystal structure of HLA-DQ2.5:DQ2.5-glia-aa [Kim, C. Y., et al., 2004]. The top three lowest-energy models were highly similar and positioned the scFv in a diagonal manner across the pMHC (FIG. 3D). In all three models, the CDR-H3 was positioned close to p7, with residues W111.1 and H112.1 within 5 Å of the L in p7 (FIG. 3E). Although no direct interactions are indicated in the models, the pL7Q substitution could indirectly be sensed causing the small reduction in MFI as observed (FIG. 3B). Similarly, both CDR-L1 and CDR-L3 are in close proximity to p9 (FIG. 3F). Three residues are close enough to interact with the Y, and one of these residues, D28, potentially forms a H-bond with αS72 of the MHC, giving a molecular explanation to the lost binding of the highly similar mAbs R2A3-9F and R4A1-3A (FIG. 3G). Taken together, fine-specificity analysis using single mutants revealed that the mAbs utilized distinct binding modes, and that mAbs R3A2-9F and R4A1-3A exhibited a greater specificity for DQ2.5-glia-α1a compared to mAb R2A1-8E.

    [0802] Detection of Cell Surface HLA-DQ2.5:DQ2.5-Glia-α1a Complexes

    [0803] As all efforts to characterize specificity and affinity of the antibodies were conducted using recombinant HLA-DQ2.5 with covalently coupled peptide, either soluble or cell-bound, we next examined if the mAbs could bind HLA-DQ2.5.sup.+ cells loaded with soluble peptide. For this purpose, we isolated monocytes using PBMCs from a healthy HLA-DQ2.5.sup.+ donor and in vitro differentiated to monocyte-derived DCs and loaded the cells with peptide. Using mAb R3A2-9F, we specifically detected cells presenting DQ2.5-glia-α1a (FIG. 4A).

    [0804] B Cells and CD19.sup.+ PCs are the Major DQ2.5-Glia-α1a Presenting Subsets in the Intestinal Mucosa of CD Patients

    [0805] Encouraged by the ability of the mAb R3A2-9F to specifically stain cells exogenously loaded with DQ2.5-glia-α1a peptide, we generated single-cell suspensions of intestinal biopsies from HLA-DQ2.5.sup.+ untreated CD patients, and co-stained the freshly isolated cells with mouse IgG2b (mIgG2b) versions of mAb R3A2-9F together with antibodies specific for other APC surface markers (FIG. 4B and FIGS. 10A and B). Unexpectedly, we observed binding of mAb R3A2-9F almost exclusively to PCs (large, viable, CD19.sup.+/−-CD27.sup.+CD38.sup.+) and B cells (smaller, viable, CD19.sup.+CD38.sup.−), whereas very few CD11c.sup.+CD14.sup.− DCs and CD11c.sup.+CD14.sup.+ or CD11c.sup.−CD14.sup.+ macrophages stained positive (FIG. 4B). Analysis of three patients analyzed in parallel showed an average of 27.4% and 35.4% mAb R3A2-9F positive PCs and B cells, respectively. Importantly, pre-blocking of FcγRs did not affect staining (FIG. 10C).

    [0806] We then compared the level of peptide presentation by B cells and PCs as detected by mAb R3A2-9F or R4A1-3A staining of both untreated CD and treated CD patients (i.e. on gluten-free diet) and matched with non-CD healthy controls. Notably, these mAbs stained cells similarly. Small intestinal PCs can be separated into 3 major subsets with distinguished longevity based on CD19 and CD45 expression; CD19.sup.+CD45.sup.+ PCs which are dynamically exchanged, and CD19-CD45+ PCs and CD19.sup.−CD45.sup.− PCs which are long-lived subsets and exhibits little and no replacement, respectively (herein these subsets are referred to as CD19.sup.+, CD45.sup.+, and CD45.sup.− PCs, respectively; FIG. 11A). Of these PC subsets, we found the CD19.sup.+ PCs to display most HLA-DQ2.5:DQ2.5-glia-α1a complexes, followed by the CD45.sup.+ and the CD45.sup.− PCs, with an average of 19%, 11% and 7% positive cells among HLA-DQ2.5.sup.+ untreated CD patients, respectively (FIGS. 5A and B). An average of 16% of the B cells were positive (FIG. 5B). Further analysis of patients revealed that all PC subsets and the B cells of treated CD patients stained negative, comparable to both HLA-DQ2.5.sup.+ and HLA-DQ2.5.sup.− healthy controls (FIG. 5C and FIG. 11B). Additionally, both HLA-DQ8.sup.+ and HLA-DQ2.2.sup.+CD patients with active disease were negative. Possibly, there were a higher number of positive PCs from patients with high Marsh scores (FIG. 5D). In summary, we found PCs and B cells to be the main cell types presenting DQ2.5-glia-α1a on HLA-DQ2.5, with the highest level of staining in the CD19.sup.+ PC subset.

    [0807] DQ2.5-Glia-Aa Presenting PCs Express TG2-Specific IgA

    [0808] To further validate our observations, we sorted four populations of PCs by use of mAb R4A1-3A and TG2-antigen multimers; bulk PCs, bulk TG2.sup.+ PCs, R4A1-3A.sup.+ TG2.sup.+ PCs and R4A1-3A.sup.+ TG2.sup.− PCs. All groups of sorted cells were microscopically confirmed to be PCs, with the typical PC morphology characterized by large nuclei and little cytoplasm (FIG. 6A). This further strengthens our observations from flow cytometry, largely excluding unspecific mAb binding by cells such as macrophages and DCs. Moreover, culturing and subsequent TG2-specific ELISPOT using the sorted PCs verified that the cells secrete IgA antibodies specific for TG2 (FIG. 6B). Importantly, within the TG2.sup.+ PC and the R4A1-3A.sup.+ TG2.sup.+ PC populations spots were clearly visible, while none were found in the R4A1-3A.sup.+ TG2.sup.− sorted PC population, nor in the T-cell negative control. Among bulk PCs spots were only visible when using many cells, in line with the approximately 10% TG2-specific PCs in the inflamed mucosa of CD patients.

    [0809] Intestinal PCs Express MHC Class II

    [0810] Despite the fact that intestinal PCs appear to have a functional BCR, and thus Ag capturing capacity, they are thought to lack APC properties by virtue of transcriptional silencing of the MHC class loci. The specific detection of gluten peptide presentation on HLA-DQ2.5 requires MHC class II expression in human intestinal PCs in CD patients. To experimentally verify this, we performed flow cytometric staining showing that the CD19.sup.+, CD45.sup.+ and CD45.sup.− PCs all indeed express MHC class II, albeit to a lower level as compared to DCs, monocytes and macrophages or B cells (FIG. 6C and FIG. 12).

    TABLE-US-00027 TABLE S1 Table S1. Overview of selection and candidate clones. Selection Input Output Enrichment # positive # unique # specific round/strategy.sup.a (cfu) (cfu) factor.sup.b clones.sup.c clones.sup.d clones.sup.e R1 1.32 × 10.sup.11 2.73 × 10.sup.6 ND ND ND ND R2A1 6.50 × 10.sup.11 5.11 × 10.sup.5 0.38 5/188 1 R2A1-8E R2A2 6.50 × 10.sup.11 1.04 × 10.sup.6 7.73 3/188 3 0 R3A1 7.35 × 10.sup.11 1.78 × 10.sup.6 3.08 41/188  1 * R3A2 8.00 × 10.sup.10 4.95 × 10.sup.5 3.87 9/188 2 R3A2-9F R4A1 3.83 × 10.sup.11 3.09 × 10.sup.5 0.33 6/188 2 R4A1-3A/* R4A2 4.83 × 10.sup.10 1.71 × 10.sup.5 0.57 11/188  5 0/* Total — — — 75/1128 11 3 .sup.aSelection was performed using two parallel strategies from R2-R4; A1 without soluble competitor and A2 with soluble competitor. In both cases the selection stringency was increased for each round. .sup.bEnrichment factor was determined using the ratio from the current selection round divided by the ratio from the previous round. The ratio was obtained by dividing the phage output (cfu.sup.ampR) on the phage input (cfu.sup.ampR) in a selection round. .sup.cDetermined by ELISA after scoring clones with a signal/background ratio above background level (set by empty E. coli XL1-Blue) as positive. .sup.dDetermined by sequencing of single clones. Some clones are not included due to out-of-frame mutations or unattainable sequence. .sup.eDetermined by SPR binding analysis. *Same clone as R3A2-9F ND, not determined Note; V.sub.H of clone R4A1-3A contains the recognition sequence of one of the enzymes used to sub-clone the scFv cassette from the phagemid to the vector for soluble expression. As the cloning step was performed prior to screening of the libraries, clones with this particular sequence may have been lost. Thus, its frequency may be under-estimated in the selection output.

    TABLE-US-00028 TABLE S2 Table S2. Kinetics of the scFv-HLA-DQ2.5:DQ2.5-glia-α1a interaction. Single cycle kinetics.sup.a Steady state Clone k.sub.on (M.sup.−1s.sup.−1) k.sub.off (s.sup.−1) K.sub.D (M) SE K.sub.D (M) K.sub.D (M) SE K.sub.D (M) R2A1-8E NA NA NA NA 2.03 × 10.sup.−7 1.20 × 10.sup.−8 R3A2-9F 1.29 × 10.sup.5 0.01262 9.79 × 10.sup.−8 5.63 × 10.sup.−8 1.42 × 10.sup.−7 2.20 × 10.sup.−8 R4A1-3A 2.89 × 10.sup.5 0.02151 7.43 × 10.sup.−8 7.14 × 10.sup.−8 6.70 × 10.sup.−8 1.30 × 10.sup.−8 .sup.aKinetics were determined by fitting data to a 1:1 Langmuir binding model. .sup.bSteady state K.sub.D was derived from the single cycle kinetics runs. NA, not available.

    DISCUSSION

    [0811] In this study, we report the generation of mAbs highly specific for HLA-DQ2.5 in complex with one of the immunodominant T-cell epitopes in CD, DQ2.5-glia-α1a. By utilizing these mAbs, we identify B cells and PCs as the main APCs presenting gluten peptides in the inflamed intestine of CD patients.

    [0812] TCR and TCR-like mAb binding to the same pMHC complex have been compared before, usually revealing distinct fine-specificities with mAb binding modes ranging from highly tilted to TCR-like docking. For the HLA-DQ2.5:DQ2.5-glia-α1a complex, binding experiments using 11 T-cell clones showed a striking dependence on p7, as a pL7A mutation completely abrogated binding for all clones, whereas alternations of all other positions resulted in clone-dependent effects. In the case of a model TCR clone, a pL7Q mutation reduced binding, while a pY9F mutation increased binding, presumably translating into HLA-DQ2.5:DQ2.5-glia-ω1 cross-reactivity. Among the three HLA-DQ2.5:DQ2.5-glia-α1a-specific mAbs described in this paper, R3A2-9F and R4A1-3A behaved similarly, while mAb R2A1-8E showed a distinct recognition pattern. All mAbs bound irrespective of p7 mutation, while p9 mutation abrogated binding for mAbs R3A2-9F and R4A1-3A, explaining the lack of binding to the DQ2.5-glia-ω1 epitope. R3A2-9F and R4A1-3A differ by one amino acid in the framework of V.sub.L, while mAb R2A1-8E shares V.sub.H with the two former mAbs, but utilizes a different V.sub.L. While p7 appears solvent exposed and the pL7Q mutation reduced mAb binding, p9 is generally considered a buried anchor residue. However, crystallographic analysis of HLA-DQ2.5 in complex with DQ2.5-glia-α1a has shown that the p9 residue can either act as an anchor or be positioned outside of the pocket. Docked models of mAb R4A1-3A and HLA-DQ2.5:DQ2.5-glia-α showed that the CDR-L1 and CDR-L3 loops are in close proximity to p9 and αS72. Although there are no direct interactions with p9, D28 of CDR-L1 interacts with αS72, an interaction that would be disrupted after mutation to the αI72 found in HLA-DQ2.2, possibly explaining the lack of binding to HLA-DQ2.2 with DQ2.5-glia-α1a. The molecular basis for the lack of mAb binding to the DQ2.5-glia-ala epitope with the corresponding native sequence remains unclear. The p6E is generated by TG2 mediated deamidation and acts as an anchor residue involved in an extensive H-bond network with both the peptide and MHC. It is conceivable that this H-bond network will rearrange in the presence of the native Q, which may then directly or indirectly influence the ability of the mAb to bind.

    [0813] HLA-DQ2.5 is largely resistant to HLA-DM-mediated peptide exchange, resulting in an extraordinarily high level of CLIP peptides presented by HLA-DQ2.5.sup.+ APCs. As a consequence, the relative proportion of HLA-DQ2.5 loaded with other peptides is assumed to be low. Cross-reactivity to CLIP2 would limit the utility of the mAbs as specific reagents. As seen in both binding experiments using recombinant molecules in ELISA and SPR, as well as in flow cytometry after staining of DQ2.5+ cells with covalently coupled CLIP2 peptide, CLIP2 is not detected. Each naturally processed antigenic pMHC complex has been estimated to occur in numbers of 10-1000 per cell. This contrasts the high density of TCR on T cells, estimated to about 50.000 molecules per cell, thus, in comparison with tetramer detection of a TCR, the use of TCR-like mAbs is challenging, particularly for detection of inefficiently loaded antigens. Still, we were able to detect both in vitro loaded cells and pMHC complexes generated in vivo after gluten consumption by CD patients.

    [0814] The intestinal compartment of conventional APC is dominated by macrophages, with smaller populations of DCs, naïve and memory B cells. DCs have been suggested to participate in priming of T cells in CD. We detect very few DCs, which might in part be explained by their low density in the intestine, and by migration to the mesenteric lymph node after antigen uptake. Although the affinities of mAbs R3A2-9F and R4A1-3A are in line with those reported for other TCR-like mAbs isolated from naïve libraries, it might be too low to detect scarce cell populations.

    [0815] B cells and PCs expressing Ig specific for gliadin and TG2 are characteristics of CD. The dominant B-cell lineage in the lamina propria is PCs, which are found at high densities, constituting 25-35% of the total mononuclear cell population, whereas there is only a minor population of memory B cells and very few naïve B cells. In CD, 1% and 10% of the PCs are specific for gliadin and TG2, respectively. The role of these cells and the antibodies they produce in disease development and maintenance has been unclear. Our findings indicate their involvement as APCs for gluten-reactive CD4.sup.+ T cells. Comparing the 3 major subsets of small intestinal PCs, we found the CD19.sup.+ subset to present DQ2.5-glia-α1a most efficiently. This subset is highly dynamic and undergoes constant renewal, whereas the CD45.sup.+ and CD45.sup.− PCs are long-lived and more static, in particular the CD45.sup.− PCs where we detected the lowest level of peptide presentation. The observed lack of correlation between DQ2.5-glia-α1a presentation and serum anti-TG2 IgA titer is in line with the previous observation that the frequency of TG2-specific PCs does not correlate with serum Ab titers.

    [0816] Although the conventional view is that B cells are not efficient activators of naïve T cells, B cells have been shown to be efficient APCs when they recognize the same antigen as the responding T cell. In a murine model of systemic lupus erythematosus (SLE), activation of naïve self-reactive T cells was shown to depend on B cells. In CD, a hapten-carrier model has been suggested for efficient presentation of gluten peptides by TG2-specific B cells, whereby BCR-bound TG2 is itself associated with the gluten peptide, or has catalyzed the coupling of the peptide to neighboring molecules. The presence of DQ2.5-glia-α1a-presenting B cells builds on these observations and strengthens the hapten-carrier hypothesis in activation of T cells. The phenotype of the CD19.sup.+CD38.sup.− B cells we identified has been thoroughly investigated. This population was found to constitute mostly memory B cells (CD27.sup.+IgD=.sup.−-IgA.sup.+) with a minor population of naïve-mature B cells (CD27.sup.− IgD.sup.+IgM.sup.+), most likely representing a variable contribution from isolated lymphoid follicles.

    [0817] The ability of PCs to act as APCs is controversial, and conflicting results have been reported from human and murine studies. Murine PCs have been shown to process antigen and activate naïve T cells. This has proven much more difficult to verify for human PCs, despite the observation that IgA and IgM PCs in the bone marrow and lamina propria have functional BCR that is able to transmit intracellular signals and internalize antigen. Nevertheless, MHC expression and an ability to activate T cells have been demonstrated for human myelomas. Additionally, human bone marrow and splenic PCs have been shown to express MHC class II. Expression from the MHC class II locus is believed to be controlled by CIITA. Upon PC maturation, CIITA expression is lost, leading to silencing of MHC class II expression. However, epigenetic mechanisms as well as carcinogenesis have been shown to induce class II expression in PCs, both by reactivation of CIITA and in its absence. We have shown that the PCs present gluten peptides on HLA-DQ2.5, which indicates that PCs function as APCs.

    [0818] In summary, we have isolated highly specific HLA-DQ2.5:DQ2.5-glia-α1a-specific mAbs, and we found PCs and B cells to be the main cell types presenting DQ2.5-glia-α1a in the intestinal lesion of CD patients. The mAbs are highly specific, detecting DQ2.5-glia-α1a solely in the context on HLA-DQ2.5. The lack of detection in HLA-DQ2.2.sup.+ and HLA-DQ8.sup.+ untreated CD patients strongly implies that our clear staining of HLA-DQ2.5.sup.+ untreated CD patients is not an artifact caused by a highly inflamed tissue. The treatment for CD is to completely abstain from gluten. However, for a fraction of CD patients, this is not curative and this group is in need of novel therapeutic intervention. Up to 50% of the gluten-reactive CD4.sup.+ T.sub.H cells in the active CD lesion may be focused on either of the immunodominant DQ2.5-glia-α1a and DQ2.5-glia-α2 epitopes. Importantly, selective blocking of dominating epitopes in HLA-driven diseases has been shown to ameliorate disease. The previously unappreciated ability of PCs to act as APCs, and the observed importance of B cells in gluten peptide presentation may also offer instructive clues for understanding of other T-cell driven autoimmune diseases.

    [0819] Materials and Methods

    [0820] Human and Animal Material

    [0821] Duodenal biopsy material was obtained according to approved protocols (Regional Ethics Committee of South-Eastern Norway approval 2010/2720 S-97201), and all subjects gave informed written consent. CD diagnosis was given according to the British Society for Gastroenterology guidelines including clinical history, anti-TG2 serological testing, HLA typing and histological analysis of small intestinal biopsies obtained by esophagogastroduodenoscopy and forceps sampling from the duodenum.

    [0822] Recombinant pMHC Expression and Purification

    [0823] Recombinant HLA-DQ2.5 or HLA-DQ2.2 with covalently coupled gluten-derived peptides containing the T-cell epitopes DQ2.5-glia-α1a (QLQPFPQPELPY, underlined 9mer core sequence), DQ2.5-glia-α2 (PQPELPYPQPE), DQ2.5-glia-ω1 (QQPFPQPEQPFP), DQ2.5-glia-ω2 (FPQPEQPFPWQP), DQ2.5-glia-γ1 (PEQPQQSFPEQERP), DQ2.5-glia-γ2 (QGIIQPEQPAQL), DQ2.5-glia-γ3 (TEQPEQPYPQP), DQ2.5-glia-γ4c (TEQPEQPFPQP) and CLIP2 (MATPLLMQALPMGAL) were generated as previously described [Fallang, L. E., et al., 2008, Quarsten, H., et al., 2001]. Briefly, insect cell produced soluble, recombinant pMHC was affinity purified using mAb 2.12.E11 [Viken, H. D., et al., 1995] specific for DQ2 and occasionally by size exclusion using Superdex 200, followed by site-specific biotinylation using BirA (Avidity). Recombinant pMHC used for SPR was further purified by size exclusion using Superdex 200 after biotinylation.

    [0824] Selection and Rescue of scFv Phage Libraries

    [0825] HLA-DQ2.5:DQ2.5-glia-α1a-specific binders were isolated from a naïve human scFv library (described in Loset, G. A., et al., 2005). Dynabeads MyOne Streptavidin T1 beads (Invitrogen) and phages (1.32×10.sup.11 cfu.sup.ampR in R1) were blocked 1 h using either 4% non-fat skim milk powder or 2% BSA (essentially fatty acid free) in PBS, alternating the blocking reagent for each selection round. 1 ml pre-blocked phage samples were incubated 1 h with 80 nM biotinylated HLA-DQ2.5:CLIP2 for negative selection (R1, R2, R3), before transfer to tubes containing beads and further incubated for 30 min. Beads containing captured HLA-DQ2.5:CLIP2 with bound phage were absorbed on a magnet and supernatant containing unbound phage was transferred to new tubes and incubated 1 h with 80 nM biotinylated HLA-DQ2.5:DQ2.5-glia-α1a for positive selection, before transfer to beads as before. After 5 washes with PBS with 0.05% Tween-20 (PBST) and 5 washes with PBS (a brief vortex between each wash), bound phages were eluted by 30 min incubation with 0.5 ml trypsin/EDTA. In subsequent rounds, all samples were selected using two strategies; alterative 1 as in R1, and alternative 2, by addition of 16.6 nM non-biotinylated HLA-DQ2.5:DQ2.5-glia-α1a competitor in solution [Zahnd, C., et al., 2010]. All incubations were performed using rotation at room temperature (RT). The selection stringency was increased for each round; washing 10+10 in R2, 20+20 in R3 and R4, decreasing antigen amount 10 times for each round, and 100 times for R4. The eluted output was used to infect 9.5 ml E. coli XL1-Blue (Stratagene) at OD.sub.600 nm 0.6 in 2× YT-TG (30 μg/ml tetracycline and 0.1 M glucose). Additional 0.05 M glucose was added to the cultures immediately before infection to ensure complete shutdown of the lac promoter. Infection was allowed for 30 min/80 rpm/37° C., followed by 30 min/220 rpm/37° C. incubation. Cultures were harvested by centrifugation, plated onto Bio-Assay dishes (NUNC) containing 2×-YT-TAG (30 μg/ml tetracycline, 100 μg/ml ampicillin, and 0.1 M glucose) and incubated overnight at 30° C. Cells were scraped and re-inoculated to OD.sub.600 nm 0.05 in 50 ml 2×YT-TAG. M13K07 (GE Healthcare) at MOI 20 was added at OD.sub.600 nm 0.1-0.2 and allowed to infect as before, followed by medium replacement to 2×YT-AK (100 μg/ml ampicillin and 50 μg/ml kanamycin), and incubated overnight at 30° C. Cultures were centrifuged and supernatants were filtrated using 0.22 μm filters. Virions were purified and concentrated by PEG precipitation as described [Marks, J. D., et al., 1991]. Phage was spot-titrated onto nitrocellulose filters essentially as before [Koch, J. et al., 2000]. Antigen-specific clones identified after selection were sequenced by GATC Biotech.

    [0826] Reformatting from Phage to Soluble Expression

    [0827] scFv cassettes encoding selected clones were retrieved either by batch-cloning a midi-prepped library glycerol stock (for R3 screening) or by PCR amplification directly from the phage stocks (for R2 and R4 screening). Briefly, the scFv cassette was cloned as NcoI/NotI fragment from the phagemid pSEX81 pL-NBLκ into pFKPEN [Gunnarsen, K. S., et al., 2010], placing the scFv in-frame with c-myc and his-tags, and transformed into E. coli XL1-Blue. Alternatively, the scFv cassette was retrieved directly from the PEG precipitated phages stocks by PCR using Phusion HotStart DNA polymerase (Thermo Scientific). 1 μl phage stocks were used with 0.5 μM primers scTCR_fw 5′-CTCAGCCGGCCATGGCC-3′ (SEQ ID NO: 516) and scTCR_rv 5′-TTTGGATCCAGCGGCCGC-3′ (SEQ ID NO:517), 0.2 mM dNTPs, annealing temperature 60° C. The PCR was purified from agarose gel and the scFv cassette cloned as NcoI/NotI fragment into pFKPEN.

    [0828] Soluble Prokaryotic Protein Expression and Purification

    [0829] Soluble prokaryotic expression both for library screening and large-scale single-clone production was performed essentially as before [Gunnarsen, K. S., et al., 2010]. For single clone screening in 96-deep well plates, single colonies were picked and inoculated into 400 μl LB-AG, sealed with AirPore Tape Sheet (QIAGEN) and incubated with shaking at 750 rpm/37° C. overnight using Titramax (Heidolph). 50 μl of the cultures were transferred to plates containing fresh LB-AG and incubated with shaking at 600 rpm/4 h/37° C. before medium replacement to 450 μl LB-A supplemented with 0.1 mM IPTG. The plates were incubated with shaking at 600 rpm/30° C. overnight. For large-scale scFv expression, the cells were inoculated into 1 L 2×YT-AG in baffled shaker flasks, and incubated at 37° C./220 rpm overnight. The cultures were then re-inoculated to OD.sub.600 nm 0.025 using 1 L 2×YT-AG. Medium was replaced to 2×-YT-A when the cultures reached OD.sub.600 nm 0.6 and incubation continued overnight at 30° C./250 rpm. Periplasmic fractions containing expressed scFv were harvested by resuspension of cell pellets in ice-cold periplasmic extraction solution (50 mM Tris-HCl, 20% sucrose, 1 mM EDTA, pH 8) supplemented with 1 mg/ml lysozyme and 0.1 mg/ml RNase A, using 300 μl for 96-well cultures and 80 ml for 1 L cultures, and incubate for 1 h/rotation/4° C. Periplasmic fraction were harvested by centrifugation and protein either used directly in screening or filtered (0.22 μm filters) and purified by IMAC (HiTrap, GE Healthcare) followed by size exclusion using HiLoad Superdex 200 (GE Healthcare) run in PBS supplemented with 150 mM NaCl and pH adjusted according to the pl of the proteins. Superdex 200 3.2/300 was used for analytical size exclusion.

    [0830] SDS-PAGE

    [0831] To visualize purified protein, 2 μg was mixed with BOLT™ LDS sample buffer, heated 5 min at 95° C. before separation on 12% NUPAGE BT gels in BOLT™ MES SDS running buffer (reagents from Novex) at 165 V/35 min along with Spectra prestained multicolor broad-range ladder (Thermo Scientific). Gels were stained with coomassie gel stain. Samples were in some cases reduced using DTT.

    [0832] ELISA

    [0833] 96-well MaxiSorp microtiter plates (Nunc) were coated overnight at 4° C. with NeutrAvidin (Avidity, 10 ug/ml in PBS), before blocking with 4% skim milk powder in PBS (w/v). Biotinylated pMHC was captured onto the NeutrAvidin. Due to variations in biotinylation levels, different pMHCs were normalized to give the same signal as 63 ng/well of HLA-DQ2.5:DQ2.5-glia-α1a in ELISA with mAb 2.12.E.11 detection. Phage, periplasmic fractions containing scFvs, or 0.5 μg/ml hIgG1 diluted in PBS with 0.05% Tween-20 (PBST) were added to the wells, and detected with either anti-M13-HRP (Amersham Biosciences, 1:5000), anti-His-tag-HRP (AbD Serotech, 1:5000), or polyclonal anti-human IgG Fc-ALP (Sigma, 1:2000) in PBST, respectively. 0.2 μg/ml mAb 2.12.E11 [Viken, H. D., et al., 1995] was detected using polyclonal anti-mouse IgG Fc-ALP (Sigma, 1:2000). HRP ELISAs were developed by addition of TMB solution (Calbiochem), while ALP ELISAs were developed with 1 mg/ml phosphatase substrate in diethanolamine buffer before absorbance reading at 620 nm (450 nm in the case of HCl addition) or 405 nm, respectively. Assays were performed at RT with duplicate wells, except for single clone screenings with only one well per sample. Between each layer, the plates were washed 3× with PBST. In competition ELISAs, 0.1 μg/ml hIgG1 was pre-incubated 30 min with non-biotinylated pMHC or peptides 2-fold diluted from 1 μM. Deamidated gliadin peptides used were 12mer DQ2.5-glia-α1a (QLQPFPQPELPY, 12mer DQ2.5-glia-α2 (PQPELPYPQPQL) (SEQ ID NO: 524) and 33mer (LQLQPFPQPELPYPQPELPYPQPELPYPQPQPF).

    [0834] Binding Analysis by SPR

    [0835] Binding specificity, affinity, and kinetics was determined on Biacore T100 (T200 sensitivity enhanced) (GE Healthcare). 1000 RU of NeutrAvidin (Avidity) diluted in acetate buffer pH 4.5 was immobilized on CM3 sensor chips (GE Healthcare) by amine coupling, before capture of 150-300 RU biotinylated pMHC. Samples were 3-fold diluted from 2 μM in PBS supplemented with 150 mM NaCl and 0.05% surfactant P20 for estimation of scFv affinity and binding kinetics. Data were acquired at 30 ul/min using the single cycle kinetics method (data collection rate 10 Hz), with an association time 120 sec and a final dissociation of 600 sec. Alternatively, a kinetic/affinity method was employed using the same conditions as above, with association times as indicated in the figure panels. For half-life comparisons of scFv and hIgG1 variants, 0.25 μM was used of each protein, with association time 120 sec and dissociation of at least 600 sec. All experiments were performed at 25° C. The surface was regenerated by a 10 sec injection of Glycine-HCl pH 2.2 at 10 ul/min. Presence of functionally folded pMHC was verified after running samples by injection of 0.2 μM mAb SPV-L3, binding only correctly folded DQ2. NeutrAvidin reference flow cell values were subtracted before data analysis using Biacore T200 Evaluation Software, version 1.0 and RI set to constant. A 1:1 Langmuir binding model was used for determination of K.sub.D. Figures were prepared using GraphPad Prism 7.

    [0836] IgG Cloning and Eukaryotic Protein Expression and Purification

    [0837] Synthetic gene fragments encoding V.sub.H and V.sub.L from the selected scFvs together with intronic splice donor sites (Genscript) were cloned as Bsml-BsiWi fragment into the IgG genomic expression vectors pLNOH2.sub.NIP and pLNOk.sub.NIP [Norderhaug, L., et al., 1997], encoding constant human gamma1 with N297G mutation and constant human kappa domains, respectively, exchanging the existing specificity for the hapten NIP [Neuberger, M. S., 1983]. Alternatively, synthetic gene fragments encoding V.sub.H and V.sub.L were ordered together with codon optimized mouse gamma2b or mouse kappa cDNA, respectively, and cloned as Bsml-BamHI fragments into the vectors described above. E. coli Top10F (Life Technologies) were transformed with the resulting plasmids before preparation of DNA. HEK293E cells (ATCC) were co-transfected with the expression vectors encoding Ig H and L chains using Lipofectamine 2000 (Invitrogen) and grown in DMEM supplemented with 10% fetal calf serum, 2 mM L-glutamine, 50 U/ml Streptomycin and 50 U/ml Penicillin. Medium was collected and replaced every day/every second day for two weeks, followed by filtration (0.22 μm) and purification on either HiTrap protein L (GE Healthcare) or a NIP-coupled column (in-house prepared) using 0.2 M glycine-HCl pH 3.0 for elution followed by rapid neutralization using 1 Tris-HCl pH 8. Protein containing fractions were further purified by size exclusion on HiLoad Superdex 200 (GE Healthcare) run in PBS supplemented with 150 mM NaCl and pH adjusted according to the pl of the proteins.

    [0838] Retroviral Transduction of A20 Murine B Cells and Flow Cytometry

    [0839] A20 B cells expressing HLA-DQ2.5 with the covalently attached DQ2.5-glia-ala (deamidated=E, native=Q, pL7Q and pY9F variants), DQ2.5-glia-ω1, DQ2.5-glia-α2 as well as HLA-DQ2.5:CLIP2 have been described. The construct encoding HLA-DQ2.5:DQ2.5-glia-α1a-Sα72I was generated by cloning a BgIII/BamHI codon-optimized synthetic DNA fragment (Genscript) encoding the HLA-DQ2.5 α-chain (DQA1*05:01) with the Sα72I mutation into the pMIG-II-eGFP retroviral plasmid (Holst J, Vignali K M, Burton A R, Vignali D A., Nat Methods. 2006; 3(3):191-197) already encoding HLA-DQ2.5:DQ2.5-glia-α1a. HLA-DQ2.2:DQ2.5-glia-α1a was generated by exchange of the HLA-DQ2.5 α-chain (DQA1*05:01) with a BgIII/BamHI codon-optimized synthetic DNA fragment encoding the HLA-DQ2.2 α-chain (DQA1*02:01). The constructs were made to have identical ectodomains as in the soluble, recombinant pMHCs. The resulting pMIG-II-eGFP-rDQ2.5:peptide plasmids and the pAmpho plasmid were then co-transfected (Lipofectamine 2000, Invitrogen) into GP2-293 packaging cells (Clontech), before virus-containing supernatants were collected 48 and 72 h after transduction and cell debris removed by centrifugation and filtration (0.45 μm). Fifty thousand A20 cells were incubated with 1.3 ml of virus-containing supernatant supplemented with 10 μg/ml polybrene and centrifuged 3000 g at 32° C. for 90 min. The supernatant was then discarded and the cells cultured in RPMI with 10% FCS for 5 days, before cells were sorted using FACSAria II (BD Biosciences) based on eGFP expression level. The A20 B cells were cultivated in RPMI-10% FCS. Data was analyzed using FlowJo software V10.

    [0840] Differentiation and Flow Cytometric Detection of Peptide-Loaded Monocyte-Derived DCs

    [0841] Monocyte-derived DCs were prepared from PBMCs from DR3/DQ2-positive blood donors [Qiao, S. W., et al., 2004]. Briefly, monocytes were positively selected from PBMCs using anti-CD14 MicroBeads (Miltenyi Biotec) and cultured in RPMI 1640 with 10% FCS containing 1000 U/ml GM-CSF and 500 U/ml IL-4 (both from R&D Systems). On day 6, DCs were matured using 150 ng/ml LPS for 48 h, supplemented with 40 μM deamidated DQ2.5-glia-α1a peptide (QLQPFPQPELPY after 24 h. For flow cytometry, cells were washed twice with flow buffer buffer (PBS with 2% FCS) in V-bottomed 96-well plates, incubated for 45 min on ice with 10 μg/ml of mAb R3A2-9F or isotype control mAb. After washing, the cells were incubated as before with 2 μg/ml F(ab′)2 anti-human IgG-biotin (Southern Biotech), followed by 30 min incubation with streptavidin-RPE (Invitrogen). After staining, cells were washed once with flow buffer and immediately analyzed on FACSCalibur (BD Biosciences), and data were analyzed with FlowJo 10.0.7 software (Tree Star). All antibodies were diluted in flow buffer.

    [0842] Isolation of Single-Cell Suspensions from Duodenal Biopsies and Flow Cytometry

    [0843] To obtain single-cell suspensions from duodenal biopsies, epithelial cells were removed by three washing steps with PBS containing 2 mM EDTA and 1% FCS for 15 min at 37° C. The remaining lamina propria was minced and digested in RPMI containing 2.5 mg/ml liberase and 20 U/ml DNase I (both from Roche) at 37° C. for 1 h. Cells were passed through 100 μM cell strainers (Falcon) and washed three times in PBS. Single-cell suspensions from lamina propria were stained in V-bottomed 96-well plates with either 10 μg/ml mIgG2b mAb R3A2-9F and R4A1-3A or isotype control mAb (mIgG2b/κ, Sigma) for 30 min, followed by staining with 1 μg/ml secondary antibody goat anti-mouse IgG2b conjugated to either Alexa-488 or FITC (Southern Biotech). Cells were subsequently stained with the following fluorochrome-conjugated mAbs targeting human cell-surface markers for 30 min on ice; CD14-APC-Cy7 (clone HCD14), CD45-v510 (clone H130), HLA-DR-bv605 (clone L243) (all Biolegend) and CD11c-v450 (clone B-Ly6); or CD3-APC (clone OKT3, eBioscience), CD11c-APC (clone S-HCl-3), CD14-APC (clone HCD14), CD45-BV510 (clone H130), HLA-DR-BV605 (clone L243), CD19-PE-Cy7 (clone HIB19), CD38-APC-Cy7 (clone HIT2) and CD27-BV421 ((clone 0323) all from Biolegend). For panHLA detection single-cell suspensions were stained with 10 μg/ml anti-human DR/DQ/DP/Dx (clone CR3/43, Santa Cruz Biotechnology) or isotype control mAb (mouse IgG1/κ, clone MOPC-21,BD Pharmingen; or clone AD1.1.10, AbD Serotec) followed by 1 μg/ml secondary mAb anti-mouse IgG1-PE (clone A85-1, BD Biosceinces) or Ab anti-mouse-PE (BD Pharmingen), before staining with the following fluorochrome-conjugated mAbs targeting human cell-surface markers; CD3-FITC (clone OKT3, Biolegend), CD11c-BV450 (clone Bly-6, BD Horizon), CD14-Pacific Blue (clone M5E2, BD Pharmingen), CD45-BV510, CD19-PE-Cy7, CD38-APC-Cy7. Additionally, cells were occasionally stained with fluorescent TG2 multimers where prepared by preincubation of biotinylated TG2 with either streptavidin-RPE (Life Technologies) or Strep-tactin-APC (iba solutions for life sciences). Propidium iodide was used to exclude dead cells and FcRs were blocked using human FcR Blocking Reagent (Miltenyi Biotec). All antibodies and reagents were diluted in PBS containing 5% FCS and 0.1% sodium azide and incubations were performed on ice. Cells were washed between each staining layer. After staining, cells were washed once and immediately acquired on LSR Fortessa cytometer (BD Biosciences), and data were analyzed with FlowJo 10.0.7 software (Tree Star).

    [0844] Sorting of PCs, Validation of Morphology by Light Microscopy and TG2 ELISPOT

    [0845] Single-cells suspensions from duodenal biopsies were stained with Alexa-488-conjugated hIaG1 R3A2-9F or hIaG1 anti-RSV as isotype controls mAbs (both in-house conjugated) together with the following antibodies against cell-surface markers; CD4, CD8 and CD14-Pacific Blue (Biolegend), CD11c-BV450 (BD Biosciences), CD27-PE-Cy7 (eBioscience), IgA-PE (Southern Biotech), and multimerised TG2 (Strep-tactin-APC, iba solutions for life sciences). PCs were sorted using FACSAriall (BD Biosciences) directly into RPM1640 without phenol red using a 100 μM nozzle. Sorted cells were spun down, resuspended in fresh medium and kept in culture at 37° C., 5% CO.sub.2 over night before imaging live cells in culture using a 40×NA 0.5 objective on an inverted Leica DM IL microscope equipped with a Axiocam MRc camera (Zeiss).

    [0846] MultiScreenHTS IP Filter Plate (0.45 μm) ELISPOT plates (Millipore) were activate with a 1 min incubation with 20 μl 35% ethanol, solution discarded and wells washed 3× with 200 μl PBS before coating with 100 μl 5 μg/ml TG2 (Phadia) in PBS or PBS only to negative control wells and incubated overnight at 4° C. Wells were 3× washed with 200 μl PBS and block with 200 μl 1% w/v BSA in PBS for 2 h at RT. Sorted cells were added in a total volume of 100 μl and plates incubated at 37° C., 5% C02 for 3 days. Cell were aspired and wells washed 3× with 200 μl PBS and 6× with 200 μl PBST before incubation with AP-conjugated goat anti-human IgA (Sigma, 1:2000) in 100 μl 1% BSA in PBST for 1.5 h at RT. Wells were wash 6× with 200 μl PBST, 3× with 200 μl PBS, 2× with 200 μl dH.sub.2O, followed by detection of spots by addition of 100 μl BCIP/NBT solution (AP Conjugate Substrate Kit, Bio-Rad). The reaction was stopped with extensively washing under running water. Plates were dried and read by ImmunoSpot Analyzer. Spots were counted manually. The BW58α.sup.−β.sup.−.human CD4 T cell hybridoma transduced with TCR364 was used as negative control.

    [0847] Antibody Modeling

    [0848] Antibody homology models were generated essentially as described in [Weitzner, B. D., et al., 2017]. The sequences of the light and heavy variable regions were saved in fasta format, aligned to homologs with known structure, and grafted together into models using Rosetta Antibody. Multiple templates of the V.sub.L-V.sub.H orientation [Marze, N. A., et al., 2016] were used, resulting in 10 grafted models. The grafted models were further refined by de novo CDR H3 loop modeling and the V.sub.L—V.sub.H docking. During modeling, the CDR H3 was constrained to the kinked conformation with a harmonic potential [Weitzner, B.D. and J.J. Gray, 2017]. We obtained a total of 2,800 Fv models. Models for docking were selected based on low Rosetta energy and good V.sub.L-V.sub.H orientation. To keep some of the diversity generated by the multi-template grafting, we considered models from at least three different V.sub.L-V.sub.H orientation templates. In the end, 10 Fv models were docked to the pMHC complex, using SnugDock [Sircar, A. and J.J. Gray, 2010].

    [0849] Antibody Docking to pMHC

    [0850] A crystal structure of unliganded HLA-DQ2.5:DQ2.5-glia-α1a is available in the PDB (1S9V, [Kim, C. Y., et al., 2004]) and was used as the docking partner for the antibody models. To alleviate pre-existing steric clashes, the structure was “relaxed” in the Rosetta energy function [Conway et al., 2014]. The top 10 Fv models and the relaxed pMHC structure were prepared for docking by running the ensemble prepack protocol as described in [Weitzner, B. D., et al., 2017]. The initial orientation was chosen based on the solved TCR:pMHC interaction (4OZI, [Petersen, J., et al., 2014]). Docking consisted of an initial spin around the Ab-Ag center-of-mass axis and uniformly sampled from 0 to 360°, and additional random perturbations consisting of small translations and rotations, with values sampled from Gaussian distributions centered at 3 Å and 8°, respectively. During docking, the flexible CDR H2 and H3 loops were refined by kinematic loop closure and the V.sub.L-V.sub.H orientation was refined by V.sub.L—V.sub.H docking. A total of 1,000 models were generated using SnugDock. The final models were picked based on low Rosetta energy, reasonable orientation relative to the pMHC, and agreement with experimentally observed specificities.

    REFERENCES

    [0851] Shan, L., et al., Structural basis for gluten intolerance in celiac sprue. Science, 2002.297(5590): p. 2275-9. [0852] Loset, G. A., et al., Construction, evaluation and refinement of a large human antibody phage library based on the IgD and IgM variable gene repertoire. J Immunol Methods, 2005. 299(1-2): p. 47-62. [0853] Kim, C. Y., et al., Structural basis for HLA-DQ2-mediated presentation of gluten epitopes in celiac disease. Proc Natl Acad Sci USA, 2004. 101(12): p. 4175-9. [0854] Petersen, J., et al., T-cell receptor recognition of HLA-DQ2-gliadin complexes associated with celiac disease. Nat Struct Mol Biol, 2014. 21(5): p. 480-8. [0855] Quarsten, H., et al., Staining of celiac disease-relevant T cells by peptide-DQ2 multimers. J Immunol, 2001. 167(9): p. 4861-8. [0856] Viken, H. D., et al., Characterization of an HLA-DQ2-specific monoclonal antibody. Influence of amino acid substitutions in DQ beta 1*0202. Hum Immunol, 1995. 42(4): p. 319-27. [0857] Zahnd, C., C.A. Sarkar, and A. Pluckthun, Computational analysis of off-rate selection experiments to optimize affinity maturation by directed evolution. Protein Engineering Design and Selection, 2010. 23(4): p. 175-184. [0858] Marks, J. D., et al., By-passing immunization. Human antibodies from V-gene libraries displayed on phage. J Mol Biol, 1991. 222(3): p. 581-97. [0859] Koch, J., F. Breitling, and S. Dubel, Rapid titration ofmultiple samples of filamentous bacteriophage (M13) on nitrocellulose filters. Biotechniques, 2000. 29(6): p. 1196-8, 2002. [0860] Gunnarsen, K. S., et al. Periplasmic expression of soluble single chain T cell receptors is rescued by the chaperone FkpA. BMC Biotechnol, 2010. 10, 8 DOI: 10.1186/1472-6750-10-8. [0861] Norderhaug, L., et al., Versatile vectors for transient and stable expression of recombinant antibody molecules in mammalian cells. J Immunol Methods, 1997. 204(1): p. 77-87. [0862] Neuberger, M.S., Expression and regulation of immunoglobulin heavy chain gene transfected into lymphoid cells. EMBO J, 1983. 2(8): p. 1373-8. [0863] Mach, N., et al., Differences in dendritic cells stimulated in vivo by tumors engineered to secrete granulocyte-macrophage colony-stimulating factor or Flt3-ligand. Cancer Res, 2000. 60(12): p. 3239-46. [0864] Qiao, S. W., et al., Dependence of antibody-mediated presentation of antigen on FcRn. Proc Natl Acad Sci USA, 2008. 105(27): p. 9337-42. [0865] Qiao, S. W., et al., Antigen presentation to celiac lesion-derived T cells of a 33-mer gliadin peptide naturally formed by gastrointestinal digestion. J Immunol, 2004. 173(3): p. 1757-62. [0866] Weitzner, B. D., et al., Modeling and docking of antibody structures with Rosetta. Nat Protoc, 2017. 12(2): p. 401-416. [0867] Marze, N.A., S. Lyskov, and J.J. Gray, Improved prediction of antibody VL-VH orientation. Protein Eng Des Sel, 2016. 29(10): p. 409-18. [0868] Weitzner, B.D. and J.J. Gray, Accurate Structure Prediction of CDR H3 Loops Enabled by a Novel Structure-Based C-Terminal Constraint. J Immunol, 2017. 198(1): p. 505-515. [0869] Sircar, A. and J.J. Gray, SnugDock: paratope structural optimization during antibody-antigen docking compensates for errors in antibody homology models. PLoS Comput Biol, 2010. 6(1): p. e1000644. [0870] Conway, et al. Protein Sci. 2014 January; 23(1):47-55.

    Example 2

    [0871] Identification of Antibodies that Specifically Bind to HLA-DQ2.5:DQ2.5-Glia-α2

    [0872] Antibodies with specificity for HLA-DQ2.5:DQ2.5-glia-α2 were isolated by use of phage display. Three rounds of selection against the target complex were performed using a large, naïve human single chain fragment variable (scFv) library and different display formats (high valence vs. low valence display, pIII fusions vs. pIX fusions). Single clones were isolated from the selection output and screened for target binding and cross-reactivity to related pMHC complexes. All the scFv identified that bind specifically to HLA-DQ2.5:DQ2.5-glia-α2 use a VH1 gene segment.

    [0873] The sequences of identified clones 6 (also referred to as 206), 17 (also referred to herein as 217), 18 (also referred to herein as 218), 20 (also referred to herein as 220), 21 (also referred to herein as 221), 23 (also referred to herein as 223), 26 (also referred to herein as 226 or 261) and 28 (also referred to herein as 228) are set forth elsewhere herein.

    [0874] These clones in the scFv format, were tested for their specificity and affinity for HLA-DQ2.5:DQ2.5-glia-α2 using surface plasmon resonance (SPR) experiments.

    [0875] First the target antigen was immobilized on a NeutrAvidin coated chip and binding of candidates (scFv clones) was tested briefly. The above mentioned clones demonstrated binding to the target antigen.

    [0876] The clones were analyzed in single cycle kinetics experiments regarding their binding to HLA-DQ2.5:DQ2.5-glia-α2, HLA-DQ2.5:DQ2.5-glia-α1a, HLA-DQ2.5:DQ2.5-glia-ω2 (FIG. 13). Binding to HLA-DQ2.5:DQ2.5-glia-α2 was observed for the above clones and dissociation constants Kd were calculated from two independent experiments with different ligand concentrations. The same single cycle kinetics method was used to analyze binding to HLA-DQ2.5:DQ2.5-glia-α1a and HLA-DQ2.5:DQ2.5-glia-ω2.

    [0877] The clones demonstrated specificity for the HLA-DQ2.5:DQ2.5-glia-α2 (FIG. 13) In particular, none of the clones, demonstrated significant cross-reactivity to HLA-DQ2.5:DQ2.5-glia-α1a.

    [0878] The affinities of the scFvs for HLA-DQ2.5:DQ2.5-glia-α2 were in the range of approximately 10 nM to 5 μM (Table Y).

    TABLE-US-00029 TABLE Y (dissociation constants (K.sub.D)) KD* (affinity **) KD (kinetics***) scFv 6  (20 ± 10) nM  (12.6 ± 0.9) nM   scFv 17 (4,900 ± 200) nM    (5,080 ± 20) nM  scFv 18 (2,700 ± 800) nM   (2,400 ± 600) nM  scFv20 (900 ± 200) nM  (730 ± 70) nM scFv 21 (700 ± 100) nM  (590 ± 70) nM scFv23 (2,300 ± 300) nM   (2,200 ± 300) nM  scFv 26 (300 ± 100) nM  (170 ± 50) nM scFv 28  (240 ± 80) nM  (110 ± 70) nM *Dissociation constants KD were determined as mean values from two separate single cycle kinetics experiments with different scFv concentrations. Errors were calculated as standard deviations. ** KD was determined by using a fit to the responses for different concentrations. ***KD was also determined by estimating the on- and off-rates. For all candidates the dissociation constants estimated with the two methods have overlapping errors.

    [0879] The ELISA was performed as essentially as per the method in Example 1. Briefly, 96-well MaxiSorp microtiter plates (Nunc) were coated overnight at 4° C. with NeutrAvidin (Avidity, 10 ug/ml in PBS), before blocking with 4% skim milk (SM) powder in PBS (w/v). Biotinylated pMHC was captured onto the NeutrAvidin.

    [0880] The biotinylated pMHCs captured were:

    [0881] HLA-DQ2.5:DQ2.5-glia-α2 (native (P4Q))

    [0882] HLA-DQ2.5:DQ2.5-glia-α2 (deamidated P4E)

    [0883] HLA-DQ2.5:DQ2.5-glia-ω2

    [0884] HLA-DQ2.5:DQ2.5-α1a

    [0885] HLA-DQ2.5:DQ2.5-hor3

    [0886] HLA-DQ2.5:DQ2.5-glia-γ2

    [0887] HLA-DQ2.5:CLIP2

    [0888] Due to variations in biotinylation levels, the different pMHCs were normalized to give the same signal as 63 ng/well of HLA-DQ2.5:DQ2.5-glia-α1a in ELISA with mAb 2.12.E.11 detection (2.12.E.11 is a monoclonal antibody specific for the DQ2 β-chain). 0.5 μg/ml hIgG1 (antibodies 206, 220 and 228) was diluted in PBS with 0.05% (v/v) Tween-20 (PBST) were added to the wells, and detected with polyclonal anti-human IgG Fc-AP (Sigma, 1:2000) in PBST. mAb 2.12.E11 (0.2 μg/ml) was detected using polyclonal anti-mouse IgG Fc-AP (Sigma, 1:2000). AP ELISAs were developed with 1 mg/ml phosphatase substrate in diethanolamine buffer before absorbance reading at 405 nm. Assays were performed at RT with duplicate wells. Between each layer, the plates were washed 3× with PBST.

    [0889] The ELISA results (FIG. 14A) indicate that the antibodies show preferential binding to the HLA-DQ2.5:DQ2.5-glia-α2 molecules as compared to the other tested pMHCs.

    [0890] FIG. 14B is a control experiment which demonstrates that the 107 antibody (hIgG1), which specifically binds to HLA-DQ2.5:DQ2.5-glia-α1a, does not bind to any of the other tested pMHCs, and also that an isotype control does not bind to any of the tested pMHCs. FIG. 14C is a control experiment which shows consistent pMHC immobilization levels.

    Example 3

    [0891] Affinity Matured Antibodies

    [0892] Affinity matured antibodies that specifically bind to HLA-DQ2.5:DQ2.5-glia-α1a or that specifically bind to HLA-DQ2.5:DQ2.5-glia-α2 were generated.

    [0893] Starting from the 107 (R4A1-3A) “mother clone” that specifically binds HLA-DQ2.5:DQ2.5-glia-α1a, affinity matured second generation clones were generated. The sequences of six such clones are set forth elsewhere herein, 4.5D (or 107-4.5D), 4.6D (or 107-4.6D), 4.6C (or 107-4.6C), 4.7C (or 107-4.7C), 5.6A (or 107-5.6A) and 15.6A (107-15.6A).

    [0894] Starting from the 206 “mother clone” that specifically binds HLA-DQ2.5:DQ2.5-glia-α2, affinity matured second generation clones were generated. The sequences of six such clones are set forth elsewhere herein, 2.B11 (or 206-2B11), 3D.8 (or 206-3D.8), 3.C7 (or 206-3.C7), 3.C11 (or 206-3.C11), 3.F6 (or 206-3.F6) and 12.F6 (206-12.F6).

    [0895] ELISA

    [0896] To assess the specificity of the affinity matured clones ELISA experiments were performed. ELISA wells were coated with NeutrAvidin (10 μg/mL), blocked with PBS supplemented with 0.05% tween and 5% skim milk powder, biotinylated forms of HLA-DQ2.5:CLIP, HLA-DQ2.5:DQ2.5-glia-α1a, and HLA-DQ2.5:DQ2.5-glia-α2 molecules were immobilised and the scFv were added (10 μg/mL in PBS). The scFv were detected with a mouse anti-myc antibody and a secondary anti-mouse antibody coupled to horseradish peroxidase (HRP). The reaction was stopped by adding 1M HCl to the wells. Absorbance was read at 450 nm.

    [0897] As shown in this ELISA experiment (FIG. 15), all tested affinity matured scFvs are specific to the relevant HLA-DQ2.5:DQ2.5-peptide antigen (pMHC) and do not cross-react to HLA-DQ2.5 with the other α-gliadin or CLIP bound.

    [0898] Surface Plasmon Resonance (SPR)

    [0899] To assess the affinity improvement, we performed Surface Plasmon Resonance (SPR) experiments (FIG. 16). We immobilized Neutravidin on a CM3 sensor chip and captured biotinylated pMHC (HLA-DQ2.5:DQ2.5-glia-α1a or HLA-DQ2.5:DQ2.5-glia-α2). The different scFv were run over the pMHC molecules. We confirmed that there is no cross-reactivity between the two targets.

    [0900] All of the affinity matured antibodies showed improved affinity relative to their respective mother clone. The affinity matured scFv bound their targets and showed different off-rates. All of them showed improved off-rates compared to the mother clone. None of them was cross-reactive to the other α-gliadin pMHC complex (only depicted in FIG. 16 for 12.F6, 3.F6, 15.A6, and 4.7C).

    [0901] Based on the rate of the decaying signal, we ranked the antibodies and chose 4.7C and 2.B11 as lead candidates for high affinity binding to HLA-DQ2.5:DQ2.5-glia-α1a and HLA-DQ2.5:DQ2.5-glia-α2, respectively. This also matches the results obtained in ELISA experiments where these candidates showed the highest signals.

    Example 4

    [0902] Antibody Modelling/Docking

    [0903] To assess the “footprint” (or “recognition motif” or “codon”) to which antibodies of the present invention may bind, antibody modelling was done using the 107 antibody that specifically binds to HLA-DQ2.5:DQ2.5-glia-α1a and the 206 antibody that specifically binds to HLA-DQ2.5:DQ2.5-glia-α2.

    [0904] Methods

    [0905] We chose to use the RosettaAntibody and SnugDock applications (software) to generate models of the docked complexes of the antibodies with their pMHC targets. Both applications belong to the Rosetta software suite for macromolecular structure prediction and design.

    [0906] RosettaAntibody's performance was tested in the antibody modeling assessment II (AMA-II) [B. D. Weitzner, et al. 2014]. It predicted all framework regions and 76% of non-H3 CDR loops at sub-Ångström accuracy. RosettaAntibody further produced the best H3 models for 4 out of 11 targets compared to competitors. It can be regarded as among the best available computational methods for prediction of antibody structures.

    [0907] SnugDock's ability to correctly predict antibody:antigen complexes was benchmarked on 15 solved structures when the software was first published [A. Sircar and J. J. Gray, 2010]. When analyzing the top 10 lowest energy models produced by SnugDock in combination with a method called EsembleDock (Chaudhury and Gray, 2008), a model of at least acceptable quality was found in 14 out ofthe 15 candidates.

    [0908] Antibody homology models were obtained essentially as described in [B. D. Weitzner, 2016a]. The amino acid sequences of the light and heavy variable regions in fasta format were aligned to homologs with solved crystal structure, and grafted together into models using Rosetta Antibody. To improve the accuracy we used multiple templates of the V.sub.L-V.sub.H orientation [N. A. Marze and J. J. Gray, 2016], resulting in 10 grafted models. The grafted models were refined by de novo CDR H3 loop modeling and V.sub.L-V.sub.H docking. The CDR3 loop of the heavy chain was constrained to the kinked conformation with a harmonic potential [B. D. Weitzner and J. J. Gray, 2016b]. We obtained 2,800 Fv models. We selected models for docking based on low Rosetta energy and V.sub.L—V.sub.H orientations within the ranges that are observed in solved antibody structures. In order to maintain structural diversity generated by the multi-template grafting, we considered models from at least three different V.sub.L-V.sub.H orientation templates. 10 Fv models were docked to the pMHC complex, using SnugDock [A. Sircar and J. J. Gray, 2010].

    [0909] We used the crystal structure of the binary complex of HLA-DQ2.5:DQ2.5-glia-α1a (PDB ID 1S9V [C.-Y. Kim, 2004]) and the crystal structure of HLA-DQ2.5:DQ2.5-glia-α2 in complex with T cell receptor JR5.1 (PDB ID 4OZF [Petersen et al., 2014]) as docking partners.

    [0910] The pMHC structures were first “relaxed” in the Rosetta energy function [Conway et al., 2014] to remove pre-existing steric clashes. The top 10 Fv models and the relaxed pMHC structure were prepared for docking by running the ensemble prepack protocol as described in [Weitzner et al, 2016a]. The initial orientation was selected based on the solved TCR:pMHC interaction (4OZI and 4OZF [J. Petersen, 2014]). Docking consisted of an initial spin around the Ab-Ag center-of-mass axis uniformly sampled from 0 to 360°, and additional random perturbations consisting of small translations and rotations, with values sampled from Gaussian distributions centered at 3 Å and 8°, respectively. The flexible CDR2 and CDR3 loops of the heavy chain were refined by kinematic loop closure and the V.sub.L-V.sub.H orientation was refined by V.sub.L—V.sub.H docking. For each antibody, 1,000 models were generated using SnugDock. The final models were picked based on low Rosetta energy, reasonable TCR-like orientation relative to the pMHC, as well as agreement with experimentally observed specificities.

    [0911] The recognition motif was identified by visual inspection of the three and four lead docking models of 107 (0063, 0158, 0195) and 206 (0064, 0083, 0107, 0265), respectively.

    [0912] Results

    [0913] A diagonal binding mode of the antibodies across the pMHC groove, similar to the one observed for TCRs. This has been observed for TCR like antibodies before and is supported by our docking models.

    [0914] The docking models predict the variable light chain of 107 and 206 to be positioned mostly over MHC and the C terminal end of the peptide. The variable heavy chain is positioned over both MHC and peptide.

    [0915] Based on the strong enrichment of different IGVH gene families during the phage display selections (IGVH6-1 for the α1a selection and IGVH1-69 for the α2 selection) it is likely that the heavy chains contribute strongly to peptide specificity.

    [0916] The models of 107 and 206 in complex with pMHC predict CDR3 loop of the light chain to be oriented to the DQB1*02 chain of MHC. CDR1 loop of the light chain is oriented towards to DQA1*05 chain.

    [0917] Because of the sequence conservation and the closely related IGKV-1 genes used by 107 and 206, we think that IGKV-1 drives binding to the MHC, HLA-DQ2.5.

    [0918] This is exemplified by IGKV1-9 and IGKV1-12 gene usage for the 107 and 206 antibodies, respectively.

    [0919] The models show VH6-1 (for antibody 107) and VH1-69 (for antibody 206) to be supporting binding to DQB 02:05 (MHC contacts) and additionally harboring peptide specificity.

    [0920] Based on the antibody modelling, the recognition motif in the conserved stretches of the antibody light chain is predicted to be comprised of a set of residues including but not limited to an N at position 92 of the VL chain of the 107 or 206 antibodies (which is in CDR3), an S at position 93 of the light chain of the 107 or 206 antibodies (which is in CDR3), and a Y at position 94 of the light chain of the 107 and 206 antibodies (which is in CDR3), a D at position 28 of the of the VL chain of the 107 or 206 antibodies (which is in CDR1), and an S at position 30 of the VL chain of the 107 and 206 antibodies (which is in CDR1).

    [0921] These light chain residues are predicted to interact with a set of MHC residues including a Y at position 60 of the MHC beta chain, a Q at position 64 of the MHC beta chain, a D at position 66 of the MHC beta chain, an R at position 70 of the MHC beta chain, an H at position 68 in the MHC alpha chain, an S at position 72 of the MHC alpha chain, and an R at position 76 of the MHC alpha chain.

    [0922] Thus, this antibody modeling study indicates that, surprisingly, the 107 antibody that specifically binds to HLA-DQ2.5:DQ2.5-glia-α1a and the 206 antibody that specifically binds to HLA-DQ2.5:DQ2.5-glia-α2, despite having different antigen specificities, share a common recognition motif which recognises the HLA-DQ2.5 MHC molecule. This recognition motif involves residues found in the variable light chain of the antibodies, as discussed above.

    REFERENCES

    [0923] B. D. Weitzner, D. Kuroda, N. Marze, J. Xu, and J. J. Gray, “Blind prediction performance of RosettaAntibody 3.0: Grafting, relaxation, kinematic loop modeling, and full CDR optimization,” Proteins Struct. Funct. Bioinforma., vol. 82, no. 8, pp. 1611-1623, 2014. [0924] A. Sircar and J. J. Gray, “SnugDock: Paratope Structural Optimization during Antibody-Antigen Docking Compensates for Errors in Antibody Homology Models,” PLoS Comput. Biol., vol. 6, no. 1, p. e1000644, January 2010. [0925] B. D. Weitzner, J. R. Jeliazkov, S. Lyskov, N. Marze, D. Kuroda, R. Frick, N. Biswas, and J. J. Gray, “Modeling and docking antibody structures with Rosetta,” Nat. Publ. Gr., vol. 12, no. D pp. 1-28, 2016a. [0926] N. A. Marze and J. J. Gray, “Improved prediction of antibody VL-VH orientation,” Protein Eng. Des. Sel., no. 2011, pp. 1-9, 2016. [0927] B. D. Weitzner and J. J. Gray, “Accurate structure prediction of CDR H3 loops enabled by a novel structure-based C-terminal ‘kink’ constraint,” J. Immunol., vol. 2017, 2016b. [0928] C.-Y. Kim, H. Quarsten, E. Bergseng, C. Khosla, and L. M. Sollid, “Structural basis for HLA-DQ2-mediated presentation of gluten epitopes in celiac disease.,” Proc. Natl. Acad. Sci. U.S.A, vol. 101, no. 12, pp. 4175-4179, 2004. [0929] J. Petersen, V. Montserrat, J. R. Mujico, K. L. Loh, D. X. Beringer, M. van Lummel, A. Thompson, M. L. Mearin, J. Schweizer, Y. Kooy-Winkelaar, J. van Bergen, J. W. Drijfhout, W.-T. Kan, N. L. La Gruta, R. P. Anderson, H. H. Reid, F. Koning, and J. Rossjohn, “T-cell receptor recognition of HLA-DQ2-gliadin complexes associated with celiac disease.,” Nat. Struct. Mol. Biol., vol. 21, no. 5, pp. 480-8, May 2014. [0930] P. Conway, M. D. Tyka, F. DiMaio, D. E. Konerding, and D. Baker, “Relaxation of backbone bond geometry improves protein energy landscape modeling,” Protein Sci., vol. 23, no. 1, pp. 47-55, 2014. [0931] Chaudhury and Gray, “Conformer Selection and Induced Fit in Flexible Backbone Protein-Protein Docking Using Computation and NMR Ensembles”, J Mol Biol. 2008 Sep. 12; 381(4):1068-87.

    Example 5

    [0932] Biophysical Characterization of Affinity Matured pMHC-Specific Antibodies

    [0933] In order to assess improvements in binding strength of the 107 and 206 derived binders and to choose lead clones, we performed SPR (surface plasmon resonance) and ranked the antibodies based on off-rates (FIG. 17A+B and Table Z). Strongly improved off-rates were observed for all clones tested. As expected, the 107-derived clones 5.6A and 15.6A, both from the random, error-prone library (i.e. clones derived from a random library made by error prone PCR across the entire scFv), had less pronounced improvements in off-rate compared to the CDR3 mutants (i.e. mutants derived from a library of CDR3 mutated clones). Based on these results, we chose 4.7C and 3.C11 as leads for binding to HLA-DQ2.5:DQ2.5-glia-α1a and HLA-DQ2.5:DQ2.5-glia-α2, respectively.

    [0934] We next assessed the thermostability of all Fab fragments by determining their melting temperatures by nanoDSF (nano differential scanning fluorimetry) (FIG. 17C+D). Whereas most HLA-DQ2.5:DQ2.5-glia-α1a binders had improved thermostability compared to 107, the HLA-DQ2.5:DQ2.5-glia-α2 binders surprisingly had slightly lower melting temperatures than the 206 mother clone. The lead clones 4.7C and 3.C11 had the highest thermostabilities among the selected clones from the targeted libraries (i.e. from the library of CDR3 mutated clones). In line with the rational for generating the random mutagenesis libraries, the mutants 5.6A and 15.6A had the highest improvements in thermostability, with 5.6A being the highest with a melting temperature 3.3° C. higher than 107.

    [0935] To harness both the improved off-rate and improved thermostability of the HLA-DQ2.5:DQ2.5-glia-α1a binders, a combination mutant, RF117, was generated, combining the lowest off-rate CDR3 loop (4.7C) with the most stable clone (5.6A). Affinities of the lead antibodies were determined in SPR (FIG. 17E-G). In concordance with the improved (lower) off-rates, all candidates had a strong improvement in affinity, with 4.7C, and 3C11 having K.sub.Ds of 170±40 pM and 88±11 pM, respectively (Table Z). This is a 400-fold improvement for 4.7C and a 2,600-fold improvement for 3.C11. The combination mutant RF117 had a K.sub.D of 20±17 μM, an approximately 3,600-fold improvement. To our knowledge, RF117 has the highest reported monomeric affinity of a pMHC-specific antibody in any human system.

    [0936] The 2.sup.nd generation antibodies were then expressed as full-length hIgG1 and tested for specific binding in ELISA (FIG. 17H). In agreement with previous results, both 4.7C and RF117 bound exclusively to the target complex HLA-DQ2.5:DQ2.5-glia-α1a and 3.C11 was specific for HLA-DQ2.5:DQ2.5-glia-α2.

    TABLE-US-00030 TABLE Z Kinetics and affinity of affinity matured variants. Kinetics and affinity Steady state.sup.c Clone k.sub.on (M.sup.−1s.sup.−1) k.sub.off (s.sup.−1) K.sub.D (M) SE K.sub.D (M) K.sub.D (M) SE K.sub.D (M) HLA-DQ2.5:DQ2.5-glia-α1a binders scFv 107.sup.a  2.89 × 10.sup.5 0.02151 .sup. 7.43 × 10.sup.−8 7.44 × 10.sup.−8 6.70 × 10.sup.−8 1.30 × 10.sup.−8 Fab 107.sup.a  2.24 × 10.sup.5 0.01601 .sup. 7.14 × 10.sup.−8 5.54 × 10.sup.−8 7.41 × 10.sup.−8 1.20 × 10.sup.−9 Fab 107.sup.b 2.377 × 10.sup.5 0.01698 7.145 × 10.sup.−8  NA NA Fab 4.7C.sup.b 3.776 × 10.sup.5 9.819 × 10.sup.−5 2.600 × 10.sup.−10 NA NA Fab 4.5D.sup.b 3.830 × 10.sup.5 1.657 × 10.sup.−3 4.327 × 10.sup.−9  NA NA Fab 4.6C.sup.b 6.972 × 10.sup.5 2.617 × 10.sup.−4 3.754 × 10.sup.−10 NA NA Fab 4.6D.sup.b 4.638 × 10.sup.5 5.413 × 10.sup.−4 1.167 × 10.sup.−9  NA NA Fab 5.6A.sup.b 1.002 × 10.sup.6 3.442 × 10.sup.−3 3.436 × 10.sup.−9  NA NA Fab 15.6A.sup.b 5.883 × 10.sup.5 2.209 × 10.sup.−3 3.755 × 10.sup.−9  NA NA Fab 4.7C 9.007 × 10.sup.5 1.770 × 10.sup.−4 1.966 × 10.sup.−10 5.40 × 10.sup.−7 NA NA Fab 4.7C 6.697 × 10.sup.5 9.685 × 10.sup.−5 1.446 × 10.sup.−10 5.60 × 10.sup.−7 NA NA Fab RF117 1.829 × 10.sup.6 5.861 × 10.sup.−5 3.024 × 10.sup.−11 3.70 × 10.sup.−7 NA NA Fab RF117 2.048 × 10.sup.6 1.550 × 10.sup.−5 7.569 × 10.sup.−12 5.60 × 10.sup.−7 NA NA HLA-DQ2.5:DQ2.5-glia-α2 binders Fab 206.sup.a  1.02 × 10.sup.6 0.2291  .sup. 2.24 × 10.sup.−7 2.25 × 10.sup.−7 2.54 × 10.sup.−7 2.70 × 10.sup.−9 Fab 206.sup.b 1.077 × 10.sup.6 0.2462  2.280 × 10.sup.−7  NA NA Fab 2.B11.sup.b 2.208 × 10.sup.6 3.285 × 10.sup.−4 1.488 × 10.sup.−10 NA NA Fab 3.C11.sup.b 1.669 × 10.sup.6 6.967 × 10.sup.−6 4.174 × 10.sup.−11 NA NA Fab 3.C7.sup.b 2.006 × 10.sup.6 7.111 × 10.sup.−4 3.544 × 10.sup.−10 NA NA Fab 3.D8.sup.b 1.545 × 10.sup.6 1.729 × 10.sup.−3 1.119 × 10.sup.−9  NA NA Fab 3.F6.sup.b 8.096 × 10.sup.5 1.225 × 10.sup.−3 1.513 × 10.sup.−9  NA NA Fab 12.F6.sup.b 8.039 × 10.sup.5 7.456 × 10.sup.−4 9.275 × 10.sup.−10 NA NA Fab 3.C11 1.190 × 10.sup.6 1.126 × 10.sup.−4 9.462 × 10.sup.−11 3.10 × 10.sup.−7 NA NA Fab 3.C11 1.387 × 10.sup.6 1.118 × 10.sup.−4 8.064 × 10.sup.−11 1.30 × 10.sup.−7 NA NA Fab 3.C11 1.335 × 10.sup.6 1.206 × 10.sup.−4 9.039 × 10.sup.−11 1.20 × 10.sup.−7 NA NA Kinetics were determined by fitting data to a 1:1 Langmuir binding model. .sup.aDetermined from single cycle kinetics. .sup.bDetermined from one concentration of protein in off-rate screening. .sup.cSteady state K.sub.D was derived from the single cycle kinetics runs. NA, not available.

    Materials and Methods

    ELISA

    [0937] EIA/RIA plates were coated with 10 μg/ml NeutrAvidin in PBS (100 μl/well) and incubate overnight at 4° C. Plates were blocked with 5% skim milk powder in PBS-T (300 μl/well) for 1 h/RT with agitation. Biotinylated pMHCs were prepared as per in Example 1 herein. Equal amounts of biotinylated pMHC (normalized to 300 ng/ml) were captured for 1 h/RT and followed by addition of 0.5 μg/ml purified pMHC-specific antibodies. hIgG1 were detected with anti-hIgG-ALP (Sigma Aldrich, 1:3,000). All antibodies were diluted in PBS-T. Plates were developed with TMB solution (Calbiochem) and the enzymatic reaction stopped by addition of 1 M HCl and read at 450 nm or developed with 1 mg/ml phosphatase substrate (Sigma Aldrich) in dietanolamine buffer and read at 405 nm using a microplate reader (Tecan sunrise).

    SPR

    [0938] Kinetics of antibody binding to pMHC were determined using a Biacore T200 (GE Healthcare). Briefly, NeutrAvidin (10 μg/ml in 10 mM sodium acetate, pH 4.5) was coupled onto a CM3 sensor chip to 1000 response units (RU) by amine coupling. Biotinylated pMHCs were prepared as per in Example 1 herein. Soluble, recombinant, biotinylated pMHC (1 μg/ml) was then captured at approximately 80-90 RU by passing over the flow cells at 10 μl/min. Antibody samples (scFv or Fab fragments) in PBS supplemented with 150 mM NaCl and 0.05% (v/v) surfactant P20 were run over the surface at various concentrations using either single cycle kinetics or a multi cycle method. For off-rate ranking, all samples were used at 0.5 μM. Binding experiments were performed at 25° C. with a flow-rate of 30 μl/min. The surface was regenerated using either Glycine-HCl pH 2.2 or Diethylamine pH 11 when necessary. Binding data were buffer subtracted and NeutrAvidin-reference-cell subtracted using the T200 Evaluation Software v1.0. Kinetic constants were determined by fitting the data to a 1:1 Langmuir binding model.

    Example 6

    Structural Basis for Improved Affinity

    [0939] In an effort to visualize the interaction interfaces, docking models of the two leads, 4.7C and 3.C11, were generated. Interestingly, the mutations responsible for the increased affinity of 4.7C, are not seen to be directly involved in binding to the pMHC, but rather to stabilize the CDR H3 loop with additional hydrogen bonds. The overall docking geometry does not appear to be changed. In contrast, 3.C11 is predicted to form several new interactions with pMHC via the CDR H1 loop as a result of the increased loop length (2 amino acids). In the model, the mutated CDR H1 loop is positioned where the CDR H3 loop was in the mother clone. This suggests that it takes over the function of the CDR H3 loop, which is displaced to the periphery of the interface. The interfaces were further analyzed using Rosetta's InterfaceAnalyzer. The solvent accessible surface area (SASA) of the interface increased from the mother clones to the high affinity variants in both cases, and Rosetta further estimated lower binding energies for the improved variants. The binding energies are also improved when normalized to the interface SASA, meaning that the gained affinity is likely both due to a larger interface surface and to improved interactions across the interface.

    Materials and Methods

    Antibody Modeling

    [0940] Structural models of Fv fragments were generated as described (Weitzner et al. 2017 and supra). CDR loops and framework regions were separately aligned to homologs with known structures and grafted together using RosettaAntibody. We used 10 templates for the V.sub.L/V.sub.H orientation (Marze and Gray 2016, supra) resulting in 10 grafted models. The grafted relaxed models were further improved by de novo CDR H3 modeling and VLVH docking. The CDR H3 was constrained to a kinked conformation (Weitzner and Gray 2016, supra) and a total of 2,800 Fv models were generated. The 10 final models were selected based on low Rosetta energy and V.sub.L/V.sub.H orientations within the natural distribution. Models were taken from at least three initial grafted templates to maintain diversity.

    Antibody Docking to pMHC

    [0941] Crystal structures of HLA-DQ2.5:DQ2.5-glia-α1a (1S9V) (Kim et al. 2004) and HLA-DQ2.5:DQ2.5:glia-α2 (4OZF) (Petersen et al. 2014, supra) were retrieved from the PDB and “relaxed” into the Rosetta energy function. We used the cocrystal structure with the T-cell receptor (40Z) (Petersen et al. 2014, supra) as a template for an initial orientation of the Fv models relative to the pMHC. We used SnugDock+EnsembleDock starting with 10 antibody Fv models (Sircar and Gray 2010, supra) to generate 1,000 docking models for each antibody. Random moves during docking consisted of a spin around the antibody:antigen center-of-mass axis sampled from 0-360°, and additional random translations and rotations, sampled from Gaussian distributions centered a 3 Å and 8°, respectively. CDRs H2 and H3 were refined by kinematic loop closure and V.sub.L/V.sub.H orientations were improved by V.sub.L/V.sub.H docking. The final models were ranked and selected based on low Rosetta energy, occurrence of “energy funnels”, and an orientation relative to the pMHC that agrees with the observed specificities. Rosetta's InterfaceAnalyzer application was used to obtain information about binding energies and interfaces.

    Example 7

    [0942] Detection of Cell-Surface pMHC

    [0943] Having demonstrated specificity and improved affinity to soluble, recombinant, biotinylated pMHC molecules, we tested whether the antibodies would stain pMHC complexes on the surface of cells. To this end, A20 mouse B cells were engineered to express variants of covalently linked pMHC complexes. Functional cell-surface pMHC expression was verified and the A20 cells were stained with the engineered hIaG1 variants (FIG. 18A+B). 107 and its offspring 4.7C and RF117 bound the deamidated target specifically. The high affinity clone 3.C11 bound to both native and deamidated HLA-DQ2.5:DQ2.5-glia-α2.

    Materials and Methods

    B Cell Lines

    [0944] The murine A20 B cell lymphoma had previously been engineered to express HLA-DQ2.5 or HLA-DQ2.2 with different peptide variants covalently linked to the MHC β-chain (Kristin Støen Gunnarsen et al. 2017, JCI Insight 2 (17) doi:10.1172/jci.insight95193). Notably, the ectodomains are identical as in the soluble pMHC molecules used for selection, screening and characterization of antibody binding by SPR and ELISA. All cells were cultured under standard conditions in RPMI 1650 supplemented with 10% FCS, 0.1 mg/ml Streptomycin and 100 U/ml Penicillin.

    Flow Cytometry

    [0945] A20 murine B cells were stained for flow cytometry experiments using the pMHC-specific antibodies. For staining A20 B cells, pMHC-specific hIgG1 antibodies were used at 5 μg/mL together with rat anti-mouse CD16/CD32 block (BD, 1:200). Bound hIgG1 were detected with biotinylated goat F(ab′)2 anti-human IgG (Southern Biotech, 2 μg/ml) followed by streptavidin R-PE (Invitrogen, 2 μg/ml). All stainings were performed on ice using V-bottom shaped 96-well plates and an equal number of cells were used in each staining (at least 100,000).all PBS supplemented with 2% FCS was used to wash cells and for dilution of antibodies and streptavidin.

    [0946] Data was acquired using an Attune N×T flow cytometer and analyzed using FlowJo software v10.4.1.

    Example 8

    Staining Human Small Intestinal Biopsy Material

    [0947] We generated fresh single-cell suspensions of intestinal biopsies from either untreated HLA-DQ2.5.sup.+ celiac disease patients or control patients and stained them with the pMHC specific mIgG2b antibodies as well as antibodies against different APC surface markers (FIG. 19). We have previously detected the highest levels of DQ2.5-glia-α1a presentation on CD19+CD45+ plasma cells (see Example 1 herein). These cells have been characterized in the small intestinal mucosa and were found to be dynamically exchanged (Landsverk et al. 2017, Journal of Experimental Medicine, 214(2):309-317). The high affinity variants also stain CD19+CD45+ plasma cells and confirm that these cells present gliadin peptides in celiac disease patients. 4.7C stains a similar percentage of cells as the 107 mother clone, while 3.C11 appears to stain a slightly higher number of cells. Two out of six samples are consistently negative for pMHC using all three antibodies. Only one sample (#5) is negative for staining with the mother clone but has a positive population when stained with 4.7C and especially 3.C11. The two control subjects confirm that there is little background staining with all antibodies used.

    Materials and Methods

    Human Material

    [0948] Duodenal biopsy material was obtained according to approved protocols (Regional Ethics Committee of South-Eastern Norway approval 2010/2720 S-97201), and all subjects gave informed written consent. CD (celiac disease) diagnosis was given according to the British Society for Gastroenterology guidelines including clinical history, anti-TG2 serological testing, HLA typing and histological analysis of small intestinal biopsies obtained by esophagogastroduodenoscopy and forceps sampling from the duodenum (Ludvigsson et al. 2014, Gut, 63 (8):1210-28). Small intestinal resections (duodenum-proximal jejunum tissue) were obtained from nonpathological small intestine during Whipple procedure (pancreatoduodenectomy) of pancreatic cancer patients who gave informed written consent (approval 2010/2720 S-97201). Only material with confirmed normal histology was included.

    Isolation of Single-Cell Suspensions from Duodenal Biopsies and Small Intestinal Resections and Flow Cytometry

    [0949] Single-cell suspensions from duodenal biopsies or from small intestinal resection were prepared as described (Landsverk et al. 2017, supra) and analyzed by flow cytometry as detailed in the Table below.

    Antibodies used for Staining of pMHC on APCs in human biopsy material and anlysis by flow cytometry:

    TABLE-US-00031 Antigen Conjugate Clone Supplier Dilution mlgG2b — 107, 4.7C, 3.C11 In-house 10 μg/ml 107/4.7C/3.C11 Isotype control — OMV In-house 10 μg/ml mlgG Alexa-546 Polyclonal Invitrogen  1 μg/ml CD3 FITC OKT3 Biolegend 1:20 CD11c APC S-HCl-3 BD Biosciences 1:20 CD14 APC HCD14 Biolegend 1:20 CD14 APC-Cy7 HCD14 Biolegend 1:20 HLA-DR BV605 L243 Biolegend 1:20 CD45 BV510 H130 Biolegend 1:20 CD19 PE-Cy7 HIB19 Biolegend 1:20 CD38 APC-Cy7 HIT2 Biolegend 1:20 CD27 BV421 O323 Biolegend 1:20

    Example 9

    [0950] Inhibition of T cell activation using affinity maturated mAbs 4.7C and 3.C11

    Results

    [0951] The Anti-pMHC Specific mAbs 4.7C (Also Referred to Herein as 107-4.7C) and 3.C11 (Also Referred to Herein as 206-3.C11) Exhibit Strong HLA and Peptide Dependent In Vitro Inhibitory Capacity of T Cell Activation

    To assess whether or not the affinity maturated lead candidate mAb clones 4.7C and 3.C11 have relevant T cell inhibitory capacity, and thus might function as disease (e.g. celiac disease) modifying agents, we generated T cell receptor (TCR) reconstructed SKW3 T cells clones expressing three different human TCRs derived from celiac patients. Two of these TCRs are specific for DQ2.5:DQ2.5-glia-α2 (clone S16 and 364), whereas the last TCR is specific for DQ2.5:DQ2.5-glia-α1a (clone 380). We then characterized the peptide dose-response of T cell activation using a human HLA-DQ2.5 positive B cells line (Raji) as antigen presenting cells (APCs) loaded with exogenous peptide. For the subsequent T cell inhibition, we fixed the amount of exogenous specific peptide to the concentration resulting in about 60% of full T cell activation (FIG. 20, A-C). After having loaded the APC with peptide, we then added the pMHC-specific and control Abs followed by adding T cells and continue incubation ON (overnight). When measuring the resulting T cell activation, indeed we observed an interference of T cell activation directly conferred by the Abs (FIG. 20, D-F). In the case of the mAb 4.7C, there was an about 20% reduction in T cell activation, but only where the APC was loaded with the peptide containing the correct epitope (FIG. 20, D). Correspondingly, the pan anti-DQ Ab (clone SPVL3) exhibited a close to complete inhibition, whereas the pan anti-DR Ab (clone L243) had no apparent effect (FIG. 20, D). Importantly, the mAb 3.C11 had no inhibitory effect on this SKW3-380 T cell activation underscoring the peptide specificities of these anti-pMHC mAbs. Conversely, the situation was opposite when the mAb 3.C11 was used to inhibit activation of the SKW3-S16 and 364 T cells (FIGS. 20, E and F). Here, a complete inhibition on par with the pan anti-DQ Ab was seen in the case of both T cells, whereas the 4.7C had a negligible effect. Thus, we conclude that the observed T cell inhibitory capacity seen with mAb 4.7C and 3.C11 is both peptide and DQ dependent underscoring the high specificity and strong target binding capacity of these mAbs.

    [0952] Materials and Methods

    [0953] IgG Protein Expression and Purification

    [0954] Purified full-length human IgG1 protein harboring the VH and VL domains of the identified affinity maturated clones 4.7C and 3.C11 were custom produced in HEK293 cells by Genscript based on the provided VH and VL domain amino acid sequences.

    [0955] Retroviral Transduction of Human SKW3 T Cells and Flow Cytometry

    [0956] The human T cell line SKW3 and the retroviral vector pMSCV were purchased from CLS Cell Lines Service GmbH and Clontech Laboratories, Inc, respectively. Based on the published T cell receptor (TCR) sequences (PMID: 24777060-Petersen et al., 2014, Nat. Struct. Mol. Biol. 21(5):480-8, 28878121—Gunnarsen et al., 2017, JCI Insight, 2 (17), and 29649333—Gunnarsen et al., 2018, PLoS One, 13(4)e0195868), TCRs 380, 364 and s16 were reconstructed by gene synthesis as human/mouse chimeric TCRs as described (PMID: 28878121, supra), and cloned into pMSCV (performed by Genscript). Retroviral transduction of the SKW3 cells was performed using the Retro-X Universal Packaging System (Clonetech) according to the manufacturer's instructions. Stable, homogenous TCR-redirected SKW3 T cells were obtained by standard cell expansion and FACS sorting using a FACSAria II cytometer (BD Biosciences) based on their TCR expression levels assessed by H57-Alexa647 (Thermo Fisher Scientific) antibody staining. The TCRs transduced SKW3 cells were validated for peptide-specific activation using a panel of known agonistic and antagonistic peptides, essentially as described (PMID: 28878121, supra), using CD69 up-regulation as activation marker assessed by anti-hCD69-APC (BD Biosciences) antibody staining. Data was acquired on a BD Accuri C6 cytometer (BD Biosciences) and analyzed using FlowJo software V10 (Tree Star).

    [0957] T Cell Activation and Inhibition Assays

    [0958] For T cell activation assays 50,000 Raji cells, which natively express HLA-DQ2.5 (PMID: 19845894—Bentley et al., 2009, Tissue Antigens, 74(5):393-403), were incubated in RPMI/10% FCS at 37° C./ON with titrated amounts of peptide (as indicated in the figures), followed by washing to remove remaining free peptide and addition of 40,000 SKW3 T cells and growth 37° C./ON (overnight) before being analysed in flow. The following peptides were used (epitopes are underlined): DQ2.5-glia-α1a (QLQPFPQPELPY) and DQ2.5-glia-α2 (PQPELPYPQPE). As a control, Cell Stimulation Cocktail containing PMA and ionomycin (eBioscience, 1:500) was added to wells containing SKW T cells only. Data analysis (EC50 determination) and figures were prepared using GraphPad Prism 7. Based on the established dose-response in T cell activation, a peptide concentration estimated to result in about 60% T cell activation (measured as CD69 upregulation) was chosen for the inhibitory assays. Following ON (overnight) incubation with peptide as above and washing, either 1 μM (final concentration) mAb 4.7C or 3.C11 were added to the Raji cells, before T cells were added and incubation continued ON. The resulting T cell activation was measured as above. As control Abs, either 0.1 μM (final concentration) of pan-anti-DR (clone L243: Thermo Scientific) of pan-anti-DQ (clone SPVL3: BD Biosciences) was added on parallel. Relative inhibitory capacity was estimated by normalizing the data according to the T cell activation in the absence of mAb, which was set to 100% activation (gray left bar in FIG. 20, D-F).