Single domain antibody libraries with maximized antibody developability properties

Abstract

Described herein are VHH antibody libraries with heavy chain variable domain framework scaffolds that have complementarity determining regions (CDRs) found in naturally occurring human antibodies, and methods for making such antibody libraries. The antibody libraries do not include members that contain one or more liabilities that affect one or more properties of the members.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of the filing date of U.S. Provisional Application No. 63 / 311,438, filed February 17, 2022, the entire contents of which are incorporated herein by reference.

[0002] Reference to sequence listing This application contains a Sequence Listing that has been submitted electronically in XML format and is incorporated herein by reference in its entirety. Said XML copy, created on Feb. 17, 2023, is named 112310-0034-70003WO00_SEQ.XML and is 98,332 bytes in size.

[0003] The present invention relates to a VHH antibody library comprising heavy chain variable domain framework scaffolds with complementarity determining regions (CDRs) found in naturally occurring human antibodies. [Background technology]

[0004] Currently, there are three recombinant antibody platforms that are primarily used to generate human antibodies for human therapy: (1) "humanization" of mouse monoclonal antibodies, (2) immunization of transgenic mice containing human antibody genes, and (3) in vitro selection from vast human antibody libraries. The immunization approach relies on the generation of an appropriate in vivo immune response and may not yield antibodies with the desired properties. In contrast, in vitro selection has the advantage that antibodies with specific properties can be directly selected and, once selected, can be easily improved in terms of affinity or specificity.

[0005] Generally, there are two types of antibody libraries: synthetic antibody libraries and natural antibody libraries. Synthetic antibody libraries can be constructed by introducing randomized complementarity determining region (CDR) sequences into antibody framework scaffolds, i.e., between framework region (FR) sequences. Such antibody libraries can have vast genetic diversity potential and improved expression through the selection of well-characterized framework scaffolds. However, synthetic antibody libraries also contain many non-functional antibody members, and much natural diversity is excluded due to the routine methods used to generate diversity within a limited set of framework scaffolds.

[0006] Antibody libraries created from natural sources, known as natural antibody libraries, have the advantage that the rearranged variable (V) genes undergo quality control in B cells, resulting in a much higher percentage of V genes that are biologically functional, even if the potential diversity is lower. Disadvantages include the challenge of obtaining large numbers of B cells to increase diversity, and the poor expression and poor biophysical properties of some antibodies recombinantly expressed in E. coli, yeast or mammalian cells. Single domain antibodies (sdAb, VHH, nanobody) consist of a single monomeric variable antibody domain. Like a whole antibody, it can selectively bind to a specific antigen. The low molecular weight makes sdAbs highly soluble and stable, in addition to improving their penetration into tissues, compared to F(ab) and F(ab')2 fragments. Furthermore, due to the lack of an Fc region, sdAbs do not exhibit complement-induced cytotoxicity. Therefore, sdAb libraries are expected to have various diagnostic and therapeutic applications. Summary of the Invention

[0007] Provided herein is an antibody library comprising a single domain heavy chain (VHH), which comprises a heavy chain framework region and a complementarity determining region (CDR), the CDR being derived from a naturally occurring antibody. Thus, one aspect of the disclosure features a VHH antibody library comprising a population of antibody VHH domains encoded by a plurality of nucleic acids, the population of antibody VHH domains comprising a population of CDR1, a population of CDR2, and a population of CDR3, which are located in the CDR1, CDR2, and CDR3 regions of the VHH genes, respectively. The amino acid sequences of CDR1, CDR2, and CDR3 may be derived from naturally occurring antibodies of a mammalian species, e.g., human or camelid.

[0008] In some embodiments, the heavy chain framework (scaffold) regions (framework 1, framework 2, framework 3 and framework 4) are derived from a single VHH domain from a therapeutic antibody, such as caplacizumab, embafolimab, gontivimab, isecalosumab, ozoralizumab, soneloximab or bovalilizumab, the coding nucleotide sequences of the exemplary parent scaffolds shown in SEQ ID NOs: 1, 2, 3 and 4, respectively (or the nucleotide sequences encoding the corresponding amino acid sequences shown in SEQ ID NOs: 53, 54, 55 and 56 due to codon degeneracy), the nucleotide sequences of deleted CDR1 versions of these parent scaffolds shown in SEQ ID NOs: 5, 7, 9 and 11, respectively, and the nucleotide sequences of deleted CDR2 versions of these parent scaffolds shown in SEQ ID NOs: 6, 8, 10 and 12, respectively. The scaffolds may also be based on modified versions of these therapeutic antibodies, such modifications being introduced to make the scaffold closer to the germline or other therapeutic VHH, to eliminate sequence liability or to enable Protein A binding. For example, isecarosumab can be mutated at two framework positions (V11L and L89V, Kabat numbering) to be closer to human germline VH3-23. The coding nucleotide sequence of an exemplary parent scaffold (or a nucleotide sequence encoding the corresponding amino acid sequence shown in SEQ ID NO:58 due to codon degeneracy) is shown in SEQ ID NO:57, the nucleotide sequence of a deleted CDR1 version of this parent scaffold is shown in SEQ ID NO:59, and the nucleotide sequence of a deleted CDR2 version of these parent scaffolds is shown in SEQ ID NO:60.

[0009] At least 90% of the population of heavy chain CDR1 and at least 90% of the population of heavy chain CDR2 are completely free of members with one or more liabilities that affect one or more properties of the antibody with the liabilities. Such VHH libraries containing CDRs derived from natural antibodies, such as human antibodies, will have a large number of functional members and will reflect the natural diversity of human antibodies. Removal of members with one or more liabilities as described herein will increase the proportion of members with desired properties, such as high yield when produced by recombinant techniques, high stability, reduced aggregation potential, reduced liabilities as described below, etc. Thus, the VHH libraries described herein will maximize the developability properties of antibodies.

[0010] The population of heavy chain CDR1's, the population of heavy chain CDR2's, and / or the population of heavy chain CDR3's are free (e.g., substantially free) of members that comprise one or more of the following liabilities: (i) a glycosylation site, (ii) a deamidation site, (iii) an isomerization site; (iv) any cysteine; (v) a net charge greater than 1; (vi) a tripeptide motif comprising at least two residues having aromatic side chains; (vii) a multispecificity site; (viii) protease-sensitive or hydrolytically susceptible sites; (ix) an integrin binding site, (x) lysine glycation site, (xi) a metal-catalyzed fragmentation site; (xii) multispecific aggregation sites; (xiii) a streptavidin-binding motif, (xiv) one or more arginines, (xv) a hydrophobic CDR sequence, (xvi) CDR mutations that reduce binding to Protein A

[0011] In this embodiment, the (i) glycosylation site comprises the motif NXS, NXT, or NXC, where X represents any naturally occurring amino acid residue except proline, the (ii) deamidation site comprises the motif NG, NS, NT, NN, NA, NH, ND, NQ, NF, NW, or NY, the (iii) isomerization site comprises the motif DT, DH, DS, DG, DN, DR, DY, or DD, the (vi) aromatic residue of the tripeptide motif comprises F, H, W, or Y, the (vii) multispecific site comprises the motif GG, GGG, RR, VG, W, WV, WW, WWW, YY, or WXW, where X represents any amino acid residue, and the (viii) protease sensitive or hydrolysis prone site comprises the motif DX, where X is P, G, S, V, Y, F, Q, K, L, or D, (ix) the integrin binding site comprises RGD, RYD, LDV, or KGD, (x) the lysine glycosylation site comprises KE, EK, or ED, (xi) the metal catalyzed fragmentation site comprises an HS, SH, KT, HXS, or SXH motif, where X represents any amino acid residue, (xii) the multispecific aggregation site comprises an X1X2X3 motif, where each of X1, X2, and X3 is independently selected from the group consisting of F, I, L, V, W and Y, and / or (xiii) the streptavidin binding motif comprises the motif HPQ, EPDW (SEQ ID NO: 49), PWXWL (SEQ ID NO: 50) (wherein X represents any amino acid residue), GDWVFI (SEQ ID NO: 51), or PWPWLG (SEQ ID NO: 52), and (xv) the hydrophobic CDR sequence is selected from the group consisting of Parkinson's disease, ... and (xvi) CDR mutations that reduce binding to Protein A include any mutation in the last amino acid of CDR2 to A, G, C, D, E, F, G, H, I, L, M, N, P, Q, S, V, W or Y, as defined by IMGT. Furthermore, at least 95% of the population of CDR1, at least 95% of the population of CDR2, and at least 95% of the population of CDR3 are completely free of non-functional members.

[0012] In some embodiments, the population of antibody heavy chain CDR1 fragments, the population of antibody heavy chain CDR2 fragments, and / or the population of antibody heavy chain CDR3 fragments does not include members that comprise at least two of (i)-(xvi). In some examples, at least the population of antibody heavy chain CDR1 fragments, the population of antibody heavy chain CDR2 fragments, and / or the population of antibody heavy chain CDR3 fragments does not include members that comprise (i)-(ix), and optionally does not include members that comprise one or more of (x)-(xvi).

[0013] In some embodiments, the antibody libraries described herein comprise a population of heavy chain CDR1s and / or a population of heavy chain CDR2s that do not include members that comprise one or more of (i)-(xvi), and the population of heavy chain CDR3s is derived from naturally occurring human antibodies (without removing one or more of the liabilities (i)-(xvi) disclosed herein), e.g., derived from human B lymphocytes or precursor cells thereof.

[0014] In some cases, members of the antibody libraries described herein comprise a heavy chain CDR1, a heavy chain CDR2, and / or a heavy chain CDR3 that are not derived from the same naturally occurring antibody, for example, at least 50% of the members of the antibody library do not comprise a heavy chain CDR1, a heavy chain CDR2, and / or a heavy chain CDR3 that are derived from the same naturally occurring antibody.

[0015] In another aspect, the disclosure features a method for generating a single domain heavy chain (VHH) antibody library, the method including: (a) providing a first plurality of nucleic acids encoding a population of naturally occurring CDR1 fragments; (b) providing a second plurality of nucleic acids encoding a population of naturally occurring CDR2 fragments; (c) providing a third plurality of nucleic acids encoding a population of naturally occurring CDR3 fragments; and (d) providing a nucleic acid gene encoding a common VHH domain including heavy chain framework regions framework 1, framework 2, framework 3, and framework 4; and inserting the first plurality of nucleic acids, the second plurality of nucleic acids, and the third plurality of nucleic acids into the CDR1 region, CDR2 region, and CDR3 region, respectively, of the common VHH domain, thereby generating a population of nucleic acids encoding a VHH domain library.

[0016] The heavy chain CDR1 fragment, the heavy chain CDR2 fragment, and the heavy chain CDR3 fragment may be derived from a naturally occurring antibody of a mammalian species disclosed herein. In some embodiments, the heavy chain CDR1 fragment, the heavy chain CDR2 fragment, and the heavy chain CDR3 fragment may be derived from the same mammalian species. Optionally, the common heavy chain variable region gene may also be derived from the same mammalian species.

[0017] In some embodiments, the antibody library includes members in which each of the heavy chain CDR1, heavy chain CDR2, and / or heavy chain CDR3 is not derived from the same naturally occurring antibody, for example, at least 50% of the members of the antibody library do not include a heavy chain CDR1, heavy chain CDR2, and / or heavy chain CDR3 derived from the same naturally occurring antibody.

[0018] The population of heavy chain CDR1's, the population of heavy chain CDR2's, and / or the population of heavy chain CDR3's are free (e.g., substantially free) of members that comprise one or more of the following liabilities: (i) a glycosylation site, (ii) a deamidation site, (iii) an isomerization site; (iv) any cysteine; (v) a net charge greater than 1; (vi) a tripeptide motif comprising at least two residues having aromatic side chains; (vii) a multispecificity site; (viii) protease-sensitive or hydrolytically susceptible sites; (ix) an integrin binding site, (x) lysine glycation site, (xi) a metal-catalyzed fragmentation site; (xii) multispecific aggregation sites; (xiii) a streptavidin-binding motif, (xiv) one or more arginines, (xv) a hydrophobic CDR sequence, (xvi) CDR mutations that reduce binding to Protein A

[0019] In this embodiment, the (i) glycosylation site comprises the motif NXS, NXT, or NXC, where X represents any naturally occurring amino acid residue except proline, the (ii) deamidation site comprises the motif NG, NS, NT, NN, NA, NH, ND, NQ, NF, NW, or NY, the (iii) isomerization site comprises the motif DT, DH, DS, DG, DN, DR, DY, or DD, the (vi) aromatic residue of the tripeptide motif comprises F, H, W, or Y, the (vii) multispecific site comprises the motif GG, GGG, RR, VG, W, WV, WW, WWW, YY, or WXW, where X represents any amino acid residue, and the (viii) protease sensitive or hydrolysis prone site comprises the motif DX, where X is P, G, S, V, Y, F, Q, K, L, or D, (ix) the integrin binding site comprises RGD, RYD, LDV, or KGD, (x) the lysine glycosylation site comprises KE, EK, or ED, (xi) the metal catalyzed fragmentation site comprises an HS, SH, KT, HXS, or SXH motif, where X represents any amino acid residue, (xii) the multispecific aggregation site comprises an X1X2X3 motif, where each of X1, X2, and X3 is independently selected from the group consisting of F, I, L, V, W and Y, and / or (xiii) the streptavidin binding motif comprises the motif HPQ, EPDW (SEQ ID NO: 49), PWXWL (SEQ ID NO: 50) (wherein X represents any amino acid residue), GDWVFI (SEQ ID NO: 51), or PWPWLG (SEQ ID NO: 52), and (xv) the hydrophobic CDR sequence is selected from the group consisting of Parkinson's disease, ... and (xvi) CDR mutations that reduce binding to Protein A include any mutation in the last amino acid of CDR2 to A, G, C, D, E, F, G, H, I, L, M, N, P, Q, S, V, W or Y, as defined by IMGT. Furthermore, at least 95% of the population of CDR1, at least 95% of the population of CDR2, and at least 95% of the population of CDR3 are completely free of non-functional members.

[0020] In some embodiments, the population of antibody heavy chain CDR1 fragments, the population of antibody heavy chain CDR2 fragments, and / or the population of antibody heavy chain CDR3 fragments does not include members that comprise at least two of (i)-(xvi). In some examples, at least the population of antibody heavy chain CDR1 fragments, the population of antibody heavy chain CDR2 fragments, and / or the population of antibody heavy chain CDR3 fragments does not include members that comprise (i)-(ix), and optionally does not include members that comprise one or more of (x)-(xvi).

[0021] In some embodiments, the antibody libraries described herein comprise a population of heavy chain CDR1s and / or a population of heavy chain CDR2s that do not include members that comprise one or more of (i)-(xvi), and the population of heavy chain CDR3s is derived from naturally occurring human antibodies (without removing one or more of the liabilities (i)-(xvi) disclosed herein), e.g., derived from human B lymphocytes or precursor cells thereof.

[0022] In some embodiments, the first plurality of nucleic acids, the second plurality of nucleic acids, and / or the third plurality of nucleic acids are generated by a process comprising: (a) obtaining amino acid sequences of the heavy chain CDR1 region, the heavy chain CDR2 region, and the heavy chain CDR3 region of a population of naturally occurring antibodies; (b) excluding amino acid sequences including any or all of (i) to (xvi) from the heavy chain CDR1 amino acid sequence, the heavy chain CDR2 amino acid sequence, and the heavy chain CDR3 amino acid sequence of (a) to obtain non-liability heavy chain CDR1 sequence, non-liability heavy chain CDR2 sequence, and non-liability heavy chain CDR3 sequence; and (c) synthesizing a first plurality of nucleic acids encoding the non-liability heavy chain CDR1 region, a second plurality of nucleic acids encoding the non-liability heavy chain CDR2 region, and a third plurality of nucleic acids encoding the non-liability heavy chain CDR2 region.

[0023] In some cases, the third plurality of nucleic acids is generated by a process comprising: (a) amplifying heavy chain CDR3 regions from a population of B cells; and (b) combining the third plurality of nucleic acids encoding the heavy chain CDR3 regions obtained in (a) with the remaining CDRs.

[0024] In some cases, the above process may further include isolating a functional member from the non-liability heavy chain CDR1 and CDR2 region and / or from the CDR3 region. For example, a functional member of a non-liability heavy chain CDR1 and CDR2 region, or a functional member of a CDR3 region, is isolated by expressing an antibody comprising a non-liability heavy chain CDR1 and CDR2 region, and / or a CDR3 region in a host cell such that the antibody is displayed on the surface of the host cell, isolating the antibody displayed on the host cell, and identifying a CDR1, CDR2, and / or CDR3 region within the displayed antibody that is a functional member of the CDR1, CDR2, and / or CDR3 region. Alternatively, functional members of the non-liability heavy chain CDR1 and CDR2 regions, and / or CDR3 regions are isolated by expressing an antibody comprising the non-liability heavy chain CDR1 and CDR2 regions, and / or CDR3 regions, optionally in fusion with a folding reporter that is β-lactamase or green fluorescent protein, or a fragment thereof, to obtain members with improved folding.

[0025] The details of one or more embodiments of the invention are set forth in the description below. Other features and advantages of the invention will be apparent from the following drawings and detailed description of several embodiments, and from the claims. [Brief description of the drawings]

[0026] [Figure 1] 1 shows sequence analysis of the indicated antibodies to predict Protein A binding sites (SEQ ID NOs: 61-71, from top to bottom). [Diagram 2]Comparative sequence analysis of selected VHH domains with the human VH3-23 germline sequence is shown (SEQ ID NOs: 72-77, from top to bottom). [Diagram 3] Schematic diagram of the assembly of the final VHH library from the parental VHH framework scaffolds, the deleted CDR (ΔCDR) versions used in the construction of the single-CDR libraries, and the filtered CDR libraries using either the duplicated native HCDR3 or the B-cell derived HCDR3. [Figure 4] FIG. 1 is a Western blot showing the expression of four VHH framework scaffolds in phage. [Diagram 5] Binding of phage expressing the four VHH framework scaffolds to their respective targets by ELISA is shown. [Figure 6] A-F show flow cytometry analysis of VHH expression on yeast cell surface using APC-conjugated protein A. A shows binding of protein A to yeast expressing caplacizumab framework scaffold. B shows binding of protein A to yeast expressing isecalosumab framework scaffold. C shows binding of protein A to yeast expressing soneloximab framework scaffold. D shows binding of protein A to yeast expressing bovalilizumab framework scaffold. E shows binding of protein A to yeast expressing positive control VH3-7 / Vk2-28 scFv. F shows no binding of protein A to yeast expressing negative control VH4-30-4 / Vk3-11 scFv. [Figure 7] Shown is a representation of the opposite orientation of the BsaI and SfiI sites that replace the CDR in one of the constructs (SEQ ID NOs: 78-80, top to bottom). [Figure 8] AB are schematic diagrams of sequence logos of CDRs selected for insertion into the library. A is a schematic diagram of CDR1 as the sequence logo selected for insertion into the library. B is a schematic diagram of CDR2 as the sequence logo selected for insertion into the library. [Figure 9]1 shows the Levenshtein distance distribution between an exemplary data set of 10,000 naturally occurring HCDR3 sequences. [Figure 10] 1 shows the cluster size distribution of exemplary 1,000 clusters generated. [Figure 11] A-B show exemplary clusters of HCDR3 from which a single sequence can be selected to represent a given cluster. Each line shows a separate HCDR3 sequence, and amino acids that do not match the cluster consensus are highlighted in black. A shows an exemplary cluster of HCDR3 of cluster A (SEQ ID NOs: 81-89, from top to bottom). B shows an exemplary cluster of HCDR3 of cluster B (SEQ ID NOs: 90-102, from top to bottom). [Figure 12] Flow cytometry analysis of library1-CDR1 and library1-CDR2 libraries before and after sorting for Protein A binding. [Figure 13] Flow cytometry analysis of Library 3-CDR1 and Library 3-CDR2 libraries before and after sorting for Protein A binding. [Figure 14] A-B show Western blots of libraries. A shows a Western blot of a library constructed with four framework scaffolds, caplacizumab, isecalosumab (1), soneloximab (3), and bovalilizumab (1). B shows a Western blot of an scFv library phage displaying four VHH framework scaffolds, caplacizumab, isecalosumab (1), soneloximab (3), and bovalilizumab (1). [Figure 15] Flow cytometry data for Library 1, Library 2, Library 3 and Library 4 selected against human interferon alpha are shown. [Figure 16] The association and dissociation rates of clones from all four libraries to interferon alpha as measured by SPR are shown. [Figure 17]Flow cytometry data for Library 1, Library 2, Library 3 and Library 4 selected against human B7-H4 are shown. [Figure 18] Flow cytometry data for library 1 and library 3 selected against human B7-H4 are shown. [Figure 19] The number of unique CDR3 clusters identified by next generation sequencing for human B7-H4 is shown. [Figure 20] 1 shows a histogram distribution of Levenshtein distances between CDR3 sequences (maximum 100) identified for B7-H4. [Figure 21] 1 shows a histogram distribution of the composite Levenshtein distances between the CDR1, CDR2, and CDR3 sequences (maximum 100) identified for B7-H4. [Figure 22] 1 shows the number of unique CDR3 clusters identified by next generation sequencing for human interferon alpha. [Diagram 23] 1 shows a histogram distribution of Levenshtein distances between CDR3 sequences (maximum 100) identified for interferon alpha. [Figure 24] 1 shows a histogram distribution of the composite Levenshtein distances between CDR1, CDR2, and CDR3 sequences (maximum 100) identified for interferon alpha. [Diagram 25] Expression levels assessed by Protein A capture by surface plasmon resonance are shown. [Figure 26] Flow cytometry analysis of library 2A-CDR1 and library 2A-CDR2 libraries before and after sorting for Protein A binding. The libraries were generated using a modified isecalosumab (1)(a) scaffold. [Figure 27] Flow cytometry data for library 2A selected against human interferon alpha is shown. [Figure 28]Flow cytometry data for library 2A selected against the receptor binding domain of Sars-CoV-2. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0027] It is generally accepted that the larger or more diverse an antibody library is, as measured in terms of the number of different antibodies, the better the antibodies that can be selected from it. Griffiths et al., EMBOJ 13(14):3245-3260, 1994 and Perelson et al., J Theor Biol., 81(4):645-70, 1979. The diversity of most antibody libraries is estimated by counting the number of transformants, assuming that each colony represents a different antibody.

[0028] In general, the diversity of VH genes is assumed to be equal to the number of colonies obtained, e.g., about 10 8 However, next-generation sequencing (NGS) revealed that the diversity of VH clones (unique HCDR3 amino acid sequences) was actually about 30-fold lower (3 × 10 6) D'Angelo et al., MAbs., 6(1): 160-72, 2014. Despite this apparent low diversity, many antibodies were selected from this library. For example, Sblattero et al., Nat Biotechnol.,18(1):75-80,2000, Glanville et al.,Curr Opin Struct Biol.,33:146-60,2015, Lou et al.,Journal of immunological methods;253(1-2):233-42,2001, Kehoe et al.,Mol Cell Proteomics,5(12):2350-63,2006,Ayriss et al.,J Proteome Res.6(3):1072-82,2007,Velappan et al.,Journal of immunological methods,321(1-2):60-9,2007,Lillo et al.,PLoS One,6(12):e27756,2011,Ferrara et al.,PLoS One, 7(11):e49535, 2012; Close et al., BMC Microbiol. 13:270, 2013; and Ferrara et al., MAbs, 7(1):32-41, 2015. NGS sequencing of another natural antibody library measured a lower VH diversity (2x10), despite the extremely high number of donors used (654). 5 ) and the estimated number of colonies was 3x10 10 Glanville et al., Proceedings of the National Academy of Sciences of the United States of America, 106(48):20216-21, 2009. Furthermore, Fantini et al. PLoS One. 12(5):e0177574, 2017 reported that the maximum diversity (number of colonies) was 6-16x10 6 , and estimated NGS diversity 3–9x10 6 Three libraries are described, each having:

[0029] The diversity found in CDR1-2 can be covered relatively easily by array-based oligonucleotide synthesis, but this is because the original diversity is 10 8 This may not be the case for CDR3s, which can easily exceed 10 different CDR3s. Even after excluding liabilities and CDRs that were seen less than 4 times, NovaSeq (3x10 9 When the diversity was evaluated using the 10 7 While genetic diversity is essential, effective functional diversity is even more important. High genetic diversity is not useful if the encoded antibody is not functional and cannot fold properly. Indeed, a single amino acid change in an antibody can lead to dramatic changes in expression levels and stability. Several publications have shown the superiority of natural antibody libraries over synthetic antibody libraries: Hugo et al., Protein Eng., 16(5):381-6, 2003; Wang et al., Proteins, 82(10):2620-30, 2014; and Chan et al., Journal of immunological methods, 373(1-2):79-88, 2011. Natural diversity has the advantage of being prescreened for functionality by the immune system. However, this has the disadvantage that some antibodies express and fold poorly in in vitro display systems, and diversity can be dominated by a few clones.

[0030] The present disclosure provides, at least in part, by providing that library members are functionally pre-screened by the immune system to eliminate members containing potential liabilities that may be poorly expressed, and to screen for liabilities using common screening systems (e.g., yeast display, phage display, or folding reporters such as β-lactamase, see, e.g., Saunders et al., Nat. Che Biol., 12:94-101; 1988, and D'Angelo et al., BMC genomics 12, suppl. 1, S1-S5; 2011, or green fluorescent protein, see, e.g., Waldo, et al., Nat. Biotechnol., 17:691-5; 1999, Cabantous, et al., PLoS ONE., 3:e2387; 2008, and Cabantous, et al., J Struct Funct The goal of the present disclosure is to construct single domain antibody libraries that contain natural diversity, such that antibody members are aggregated and / or folded poorly in the presence of cleavage (see, e.g., J. Immunol. 1999, 6:113-9; 2005). Thus, in some embodiments, the disclosure features methods for generating highly diverse and highly functional antibody libraries by combining naturally occurring CDRs, including naturally occurring CDRs that contain in vivo generated somatic mutations, within an antibody scaffold such that the members of the antibody library are expected to be well expressed and / or folded and lack liability.

[0031] As used herein, the term "liability" refers to a motif in an antibody (e.g., located in the heavy chain region) that adversely affects one or more desirable properties of the antibody (e.g., stability, good expression in an expression or display system, proper folding, no or reduced aggregation, solubility, no or reduced integrin binding, no or reduced glycosylation, no or reduced deamidation, no or reduced isomerization, no unpaired cysteines, or no or reduced protease susceptibility, etc.). Because such antibody libraries are composed of highly functional members, they are expected to be functionally much larger than libraries of similar genetic size in which antibodies containing any of these liabilities exist. In other words, the antibody libraries disclosed herein have much greater effective diversity.

[0032] I. VHH antibody libraries and methods of construction Provided herein are single domain antibody (sdAb) libraries (VHH antibody libraries) comprising heavy chain elements as described herein, and methods of generating such antibody libraries, in which the heavy chain CDRs are inserted into a preselected heavy chain variable domain gene framework scaffold, also as described herein. The heavy chain CDR1, CDR2, and / or CDR3, as well as the preselected heavy chain variable domain framework scaffold, may be from a mammalian species, e.g., human, mouse, rat, rabbit, dog, pig, or camelid, such as camel or llama. In some cases, the heavy chain CDR1, CDR2, and CDR3 may be derived from antibodies of the same mammalian species (e.g., human or camelid). Optionally, the preselected heavy chain variable domain genes may be derived from the same mammalian species. Alternatively, the heavy chain CDR1, CDR2, and / or CDR3, as well as the optional preselected heavy chain variable domain genes, may be derived from naturally occurring antibodies of different mammalian species. In some embodiments, the heavy chain CDRs and the preselected variable domain genes are all from the same mammalian species, e.g., human.

[0033] VHH antibodies (used interchangeably in the plural) are single monomeric variable antibody domains capable of specifically binding to target antigens, such as carbohydrates, polynucleotides, lipids, polypeptides, etc., via an antigen recognition site. As used herein, the term "antibody" refers to nanobodies, single domain antibodies (also known as nanobodies, e.g., VHH antibodies, such as those found in camelids). H Heavy chain-only antibodies (HcAbs) are naturally produced by camelids and sharks. The antigen-binding portion of HcAbs is composed of VHH fragments. Vincke et al., Methods Mol Biol. 911:15-26 (2012).

[0034] VHH antibodies consist of heavy chain variable regions (VHs) that are subdivided into hypervariable regions, also called "complementarity determining regions" ("CDRs"), interspersed with more highly conserved regions known as "framework regions" ("FRs") or scaffolds. H ) included. Each V His typically composed of three CDRs and four FRs, arranged from amino terminus to carboxy terminus in the following order: FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4. The lengths of framework regions and CDRs can be precisely defined using methodologies known in the art, for example, by the Kabat definition, the Chothia definition, the AbM definition, and / or the Contact definition, all of which are well known in the art. See, e.g., Kabat, E. A., et al. (1991) Sequences of Proteins of Immunological Interest, Fifth Edition, USDepartment of Health and Human Services, NIH Publication No. 91-3242; Chothia et al., (1989) Nature 342:877; Chothia, C. et al. (1987) J. Mol. Biol. 196:901-917; Al-lazikani et al. (1997) J. Mol. Biol. 273:927-948; and Almagro, J. Mol. Recognit. 17:132-143 (2004). See also, hgmp.mrc.ac.uk and bioinf.org.uk / abs.

[0035] The antibody libraries disclosed herein include VHH antibody libraries. Exemplary steps for constructing the antibody libraries described herein may include: (1) identifying a suitable heavy chain variable domain gene scaffold; (2) generating a vector based on a scaffold containing a single insertion site for each CDR (the remaining CDRs can remain unchanged); (3) identifying naturally occurring CDRs by analysis of a database of naturally occurring antibody sequences (which may be obtained by sequencing members of a natural antibody library); (4) eliminating sequences from a database of naturally occurring CDR sequences that are likely to code for liabilities; (5) synthesizing the remaining set of CDRs as oligonucleotides; (6) Inserting the CDRs at appropriate sites within the pre-modified scaffold (each scaffold contains CDRs at only one site) (e.g., a set of identified HCDR1s are inserted into the HCDR1 sites of the modified scaffold).

[0036] In some embodiments, the CDRs (e.g., CDR1, CDR2, CDR3, or a combination thereof) identified as described herein can be experimentally screened or selected for good folding and / or expression, and can be screened or selected for liability such as poor folding, poor expression, polyreactivity, or aggregation. The selected CDRs can be inserted into a complete V domain within the scaffold. The resulting complete V domain can be further screened and selected for good folding and / or expression, and / or can be screened and selected for liability such as poor folding, poor expression, polyreactivity, or aggregation. The selected VH complete scaffold can be assembled and cloned into a suitable display vector (e.g., phage or yeast) for screening of antibodies with the desired binding specificity.

[0037] A. Selection of Heavy Chain Variable Domain Framework Scaffold In some embodiments, the heavy chain framework scaffolds used in the construction of the antibody libraries described herein may be derived from commercially available therapeutic antibodies (e.g., those for which marketing authorization has been granted by the U.S. Food and Drug Administration or / and the European Medicines Agency) or therapeutic antibodies currently undergoing clinical trials, e.g., Phase II or Phase III trials. As used herein, a therapeutic antibody refers to an antibody molecule that is an approved pharmaceutical product (e.g., in the US, in the EP, or in other jurisdictions such as CA or JP) or that has undergone / or is currently undergoing clinical trials in an appropriate jurisdiction, e.g., in the U.S. or Europe. The following is an exemplary list of factors to be considered when selecting a framework scaffold: (i) Currently approved for therapeutic use or in clinical trials; (ii) It has affinity for Staphylococcus aureus protein A, facilitating purification during production. VHH domains are similar to human VH3 domains, many of which are known to have affinity for protein A. (iii) lacking cysteine ​​residues, except for the canonical pair in frameworks 1 and 3 (Kabat positions 22 and 92); (iv) has an arginine residue in framework 2, as opposed to the canonical leucine residue (Kabat position 45) present at the VH / VL interface of human antibodies; (v) Has a canonical tryptophan residue as the first residue (Kabat position 103) in framework 4. In the final library, this region was sourced from a human donor and the majority of sequences have a tryptophan residue at this position.

[0038] The germline heavy chain variable domain genes used in such therapeutic antibodies can be examined for properties such as aggregation, hydrophobic interactions, multispecificity, monomericity, expression levels and purification characteristics. Those with the desired properties can be selected as framework scaffolds for library construction. The properties and selection criteria are shown in Table 1. This is shown as an illustrative example of therapeutic antibodies considered, some of which are multi-domain / multispecific (numbers in brackets indicate the domains analyzed). A sequence analysis of these therapeutic VHH antibodies is shown in Figure 1.

[0039] In some specific examples, the VH framework scaffolds used in the antibody libraries described herein are derived from caplacizumab, embafolimab, gontivimab, isecalosumab, ozoralizumab, soneloximab, or bovalilizumab. Exemplary antibody characteristics are shown in Table 2. As used herein, "derived from" refers to the use of the VH gene of any of these therapeutic antibodies with no modifications or with one or more mutations introduced into one or more of the framework regions, e.g., up to five amino acid substitutions (e.g., up to four, three, two, or one amino acid substitutions) in the VH gene (e.g., in one or more of the framework regions).

[0040] In some cases, the mutation introduced into the germline VH gene or the mutation introduced into the VH gene of the reference therapeutic antibody (e.g., those listed in Table 2) can be a conservative substitution. As used herein, "conservative amino acid substitution" refers to an amino acid substitution that does not change the relative charge or size characteristics of the protein in which the amino acid substitution is made. Variants can be prepared according to such methods known to those skilled in the art, such as those found in references that summarize methods for changing polypeptide sequences, such as Molecular Cloning: A Laboratory Manual, J. Sambrook, et al., eds., Second Edition, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, 1989, or Current Protocols in Molecular Biology, FMAusubel, et al., eds., John Wiley & Sons, Inc., New York. Conservative substitutions of amino acids include substitutions made within the following groups of amino acids: (a) M, I, L, V, (b) F, Y, W, (c) K, R, H, (d) A, G, (e) S, T, (f) Q, N, and (g) E, D.

[0041] The nucleotide sequences of exemplary VH framework scaffolds are shown in Table 3. The nucleotide sequences of deleted CDR1 and CDR2 versions of exemplary VH framework scaffolds are shown in Table 4.

[0042] Any of the VH framework scaffolds described herein can be used to construct cassettes that allow for cloning one or more of a plurality of nucleic acids, each encoding a diverse population of heavy chain CDRs, into the corresponding framework scaffold at the corresponding CDR positions.

[0043] In some cases, restriction enzyme recognition sites can be introduced into the heavy chain framework scaffold adjacent to the CDR1 or CDR2 regions to clone multiple nucleic acids encoding a diverse population of heavy chain CDR1 or heavy chain CDR2, respectively. In one example, restriction enzyme recognition sites can be introduced into the heavy chain framework scaffold adjacent to all of the CDR1 and CDR2 regions to clone diverse heavy chain CDR1 and CDR2 at the corresponding positions.

[0044] 2 and 3 show exemplary schemes for the construction of cassettes for introducing heavy chain CDR1, CDR2, or CDR3 diversity into a heavy chain framework scaffold. The resulting cassettes can be placed into suitable expression vectors for producing the encoded antibodies in suitable expression, display, or folding reporter systems.

[0045] B. Heavy Chain CDR Population The heavy chain CDR1, CDR2 and / or CDR3 populations contained in the antibody library may be derived from naturally occurring human antibodies. Such CDR sequences can be obtained by sequencing naturally occurring antibodies (e.g., human antibodies) in an existing natural antibody library and analyzing the heavy chain sequences thus obtained by conventional methods to identify the heavy chain CDR sequences.Alternatively, or in addition, the CDR sequences of naturally occurring antibodies can be found in a public database of naturally occurring antibody sequences (e.g., human antibody sequences or Camelid VHH antibody sequences), such as the NCBI database, the IMGT database, the sequences described in Jackson et al., J. Immunol. Methods, 324:26, 2007, and / or the sequences described in Lee et al., Immunogenetics, 57:917, 2006, the observed antibody space described in Kovaltsuk, A. et al. (antibodymap.org), the observed antibody space: A Resource for Data Mining Next-Generation Sequencing of Antibody Repertoires. Journal of Immunology, doi:10.4049 / jimmunol.1800708 (2018), and / or the iReceptor database described in Corrie, B.D. et al. iReceptor: A platform for querying and analyzing (ireceptor.irmacs.sfu.ca). It can be obtained by analyzing the sequences of such antibodies in the sequence databases described in antibody / B-cell and T-cell receptor repertoire data across federated repositories. Immunol Rev 284, 24-41, doi:10.1111 / imr.12666 (2018), and / or Briney, B. et al., Commonality despite exceptional diversity in the baseline human antibody repertoire. Nature, doi:10.1038 / s41586-019-0879-y (2019).

[0046] The lengths of framework regions and CDRs can be precisely defined using methodologies known in the art, such as by the Kabat definition, the IMGT definition, the Chothia definition, the AbM definition, and / or the Contact definition, all of which are well known in the art. For example, Kabat, EA, et al. (1991) Sequences of Proteins of Immunological Interest, Fifth Edition, USDepartment of Health and Human Services, NIH Publication No. 91-3242, Chothia et al., (1989) Nature 342:877, Chothia, C. et al. al.(1987)J.Mol.Biol.196:901-917, Al-lazikani et al(1997)J.Molec.Biol.273:927-948, Lefranc,MPet al.IMGT unique numbering for immunoglobulin and T cell receptor variable domains and Ig superfamily V-like domains.Dev Comp Immunol 27, 55-77 (2003) and Almagro, J. Mol. Recognit. 17:132-143 (2004). See also hgmp.mrc.ac.uk, IMGT.org and bioinf.org.uk / abs.

[0047] The heavy chain CDR1, CDR2, and / or CDR3 sequences thus obtained may be further analyzed to remove those containing liabilities, e.g., those listed in Table 6. In some cases, the heavy chain CDR1, CDR2, and / or CDR3 sequences containing one of the liabilities listed in Table 6 (e.g., glycosylation sites, deamidation sites, isomerization sites, unpaired cysteines, net charges greater than one (e.g., in HC CDR1-2), tripeptide motifs containing at least two aromatic residues (which may affect viscosity), motifs that promote aggregation, (viii) multispecific sites such as those containing motifs GG, GGG, RR, VG, W, WV, WW, WWW, YY, or WXW (wherein X represents any amino acid residue), protease-sensitive sites (fragmentation-sensitive sites), or integrin-binding sites). Figures 8A-8B show logo representations of the final CDR1 and CDR2 sequences. Additional or different sequence liabilities in any combination can be similarly removed from the database of CDRs, thereby providing a list of CDRs with no sequence liabilities.

[0048] Alternatively or in addition, potential glycosylation sites, such as lysine glycosylation sites, can be removed. Glycosylation sites refer to sites in a protein molecule that can be linked to a sugar molecule through a non-enzymatic process. Exemplary glycosylation sites include, but are not limited to, KE, EK, and ED. Additional liabilities include metal-catalyzed fragmentation sites (e.g., HS, SH, KT, HXS, or SXH, where X represents any amino acid residue), multispecific aggregation sites (e.g., having a motif of X1X2X3, where each of X1, X2, and X3 is independently F, I, L, V, W, and Y), and streptavidin binding motifs (e.g., HPQ EPDW (SEQ ID NO: 49), PWXWL (SEQ ID NO: 50), where X represents any amino acid residue, GDWVFI (SEQ ID NO: 51), and PWPWLG (SEQ ID NO: 52)).

[0049] Substantially free means that the number of heavy chain CDRs containing the liability is less than 20% in the library, e.g., less than 15% or less than 10%. In some examples, heavy chain CDR1, CDR2, and / or CDR3 sequences containing two or more (e.g., 3, 4, 5, 6, 7, or more) of the above-mentioned liabilities can be removed such that the resulting library does not contain (substantially free or completely free) members containing the excluded liabilities. In one example, heavy chain CDR1, CDR2, and / or CDR3 sequences containing all of the liabilities listed in Table 6 can be removed such that the resulting library does not contain (substantially free or completely free) members containing any of the liabilities. Alternatively, or in addition, heavy chain CDR1, CDR2, and / or CDR3 sequences containing all of the liabilities listed in Table 6 can be removed such that the resulting library does not contain (substantially free or completely free) members containing any of the liabilities, as shown in Figures 8A-8B. In one particular example, heavy chain CDR1, CDR2, and / or CDR3 sequences that contain all of the liabilities disclosed herein can be removed such that the resulting library is free of (substantially free of or completely free of) members that contain any of the liabilities.

[0050] In some instances, heavy chain CDR1 and CDR2 sequences that contain one or more of the liabilities, such as those listed in Table 6, can be removed, but heavy chain CDR3 sequences can be derived from naturally occurring human antibodies without removing members with liabilities. Alternatively, heavy chain CDR3 sequences that contain one or more liabilities can also be removed.

[0051] In some instances, heavy chain CDR1, CDR2, and / or CDR3 sequences with abnormal lengths can also be excluded.

[0052] In some examples, heavy chain CDR1 and CDR2 members that contain deamidation sites (e.g., NG, NS, NT, NN, NA, NH, ND, NQ, NF, NW or NY), isomerization sites (e.g., DT, DH, DS, DG, DN, DR, DY or DD), aggregation sites (FHW), motifs that affect viscosity (e.g., HYF and HWH), motifs that indicate low developability (e.g., net charge >+1 in HCDR1-2), unpaired cysteines, multispecificity sites (e.g., GG, GGG, RR, VG, W, WV, WW, WWW, YY, or WXW, where X refers to any amino acid residue), and glycosylation sites (e.g., NXS, NXT, or NXC, where X is any amino acid residue except proline) can be excluded. In some examples, one or more of the following liabilities in the heavy chain CDR1 and heavy chain CDR2 members can also be excluded: additional glycosylation sites (e.g., NXC, where X is any amino acid residue except proline), additional deamidation sites (e.g., NA, NH, and / or ND), additional isomerization sites (e.g., DT and / or DH), lysine glycosylation sites (e.g., KE, EK, and ED), integrin binding sites (e.g., RGD, RYD, LDV, and KGD), protease sensitive sites (fragmentation sites) (e.g., DP, DG, DS, DV, DY, DF, DQ, DK, DL, and DD), metal catalyzed fragmentation sites (e.g., HS, SH, KT, HXS, and SXH, where X represents any amino acid residue), multispecific aggregation sites (e.g., having a motif of X1X2X3, where each of X1, X2, and X3 is independently F, I, L, V, W, or Y), and / or streptavidin binding sites (e.g., HPQ EPDW (SEQ ID NO: 49), PWXWL (SEQ ID NO: 50), where X represents any amino acid residue, GDWVFI (SEQ ID NO: 51), and PWPWLG (SEQ ID NO: 52)).

[0053] In some instances, heavy chain CDR3 members having one or more of the liabilities described herein may also be excluded. Alternatively, heavy chain CDR3 members may be included that are derived directly from a naturally occurring antibody without removing one or more of the liabilities described herein.

[0054] Heavy chain CDR1, CDR2, and / or CDR3 sequences from a naturally occurring antibody, either excluding sequences containing one or more liabilities or maintaining all sequences, can be used as templates to synthesize nucleic acids encoding and replicating the CDR sequences. Such nucleic acids can be inserted into the corresponding CDR positions of the VH scaffolds disclosed herein and are referred to as "replicated native CDRs."

[0055] If desired, the expression vector having a VH scaffold with one or more heavy chain CDRs inserted can be introduced into a suitable expression / display system to isolate functional members. Functional members include those having one or more superior properties, such as better expression and display in a suitable display system, improved folding, reduced aggregation or polyreactivity, and / or higher Tm. Such functional members can be identified by harvesting host cells that display antibodies produced from the expression vector and sequencing the corresponding heavy chain CDR sequences encoded by the expression vector in the harvested host cells.

[0056] For example, the initial antibody library may also be sorted by staining yeast-displayed antibodies with a conformational probe that detects correct antibody folding. Traxlmayr et al., Arch Biochem Biophys. 526(2):174-80, 2012. Examples of such conformational probes include Protein A, which can bind to VH3 and VHH domains (Hillson et al., The Journal of experimental medicine. 178(1):331-6, 1993; Akerstrom et al., 1994; J. Imm Methods, 177(1-2):151-63, 1994; and Roben et al., J. Immunology 154(12):6437-45, 1995), as well as indole-3-butyric acid (Alves et al., Langmuir, 28(25):9640-8, 2012; Alves et al., Anal 154(12):6437-45, 1995), which binds to the "nucleotide binding site" found in all antibodies (Rajagopalan et al., Proceedings of the National Academy of Sciences of the United States of America, 93(12):6019-24, 1993). Chem., 84(18):7721-8, 2012; Alves et al., Bioconjug Chem., 25(7):1198-202, 2014; and Mustafaoglu et al., Biotechnol Bioeng., 112(7):1327-34, 2015).

[0057] Previous use of conformational probes has been shown to predict high expression and thermostability in yeast display (Traxlmayr et al., 2012; Shusta et al., J Mol Biol. 292(5):949-56, 1999; Traxlmayr et al., Biochim Biophys Acta., 1824(4):542-9, 2012; Traxlmayr et al., Protein Eng Des Sel., 26(4):255-65, 2013; and Hasenhindl et al., Protein Eng Des Sel., 26(10):675-82, 2013). In this approach, antibody fragments that are well expressed and well folded are selected. Rather than positive selection for good display, individual CDR libraries can be depleted of CDRs containing liability. For example, to perform yeast display sorting, we performed adaptation screens used for antibody screening (Yang et al., MAbs., 5(5): 787-94, 2013; Kelly et al., MAbs., 7(4): 770-7, 2015; Kohli et al., MAbs. 7(4): 752-8, 2015; Obrezanova et al., MAbs., 7(2): 352-63, 2015; Wu et al., Protein Eng Des Sel., 28(10): 403-14, 2015; Yang et al., MAbs., 9(4): 646-53, 2017; Xu et al., Protein Eng Des Sel., 26(10): 663-70, 2013, and Kelly et al., MAbs., 9(7):1036-40, 2017), isolating yeast-displayed antibodies that correspond to more “developable” phenotypes allows for the selection of appropriate CDRs, which can then be combined to create high-functioning libraries.

[0058] Examples of such selection include multispecific reagents, heparin, or chaperones, retaining only antibodies that do not bind such substances. Stability can be further increased by applying a heat shock step (Traxlmayr et al., 2012; Shusta et al., J Mol Biol. 292(5):949-56, 1999; Traxlmayr et al., Biochim Biophys Acta., 1824(4):542-9, 2012; Traxlmayr et al., Protein Eng Des Sel., 26(4):255-65, 2013, and Hasenhindl et al., Protein Eng Des Sel., 26(10):675-82, 2013).

[0059] Sequences encoding functional members of heavy chain CDR1, CDR2, and / or CDR3 may be used as templates to synthesize nucleic acids encoding such functional members or may be used directly. The resulting nucleic acids can then be inserted into VHH scaffolds as described herein to generate antibody libraries, also as described herein. In some embodiments, the antibody libraries disclosed herein are substantially free of non-functional members, e.g., contain less than 10% (e.g., less than 8%, less than 5%, less than 3%, less than 1%, or less) non-functional members.

[0060] C. Antibody Library The antibody libraries described herein may include a plurality of nucleic acids encoding a population of antibody VHH domains that collectively comprise a population of CDR1s, a population of CDR2s, and a population of CDR3s located in the CDR1, CDR2, and CDR3 regions of VHH genes and inserted into the corresponding CDR positions. Furthermore, the amino acid sequences of CDR1, CDR2, and CDR3 are derived from naturally occurring antibodies, and at least 90% of the population of heavy chain CDR1s and at least 90% of the population of heavy chain CDR2s are completely free of liability-containing members.

[0061] In some embodiments, the antibody libraries described herein are heavy chain or VHH libraries that contain a plurality of nucleic acids encoding a plurality of antibody heavy chain variable domains. In some examples, the heavy chain or VHH library contains at least 10 2 Diverse heavy chain CDR1 (at least 10 2 unique heavy chain CDR1 sequences), e.g., at least 10 3 , 10 4 , 10 5 Diversity or at least 10 6 Alternatively, or in addition, the heavy chain or VHH library may comprise at least 10 2 Diverse heavy chain CDR2 (at least 10 2 unique heavy chain CDR2 sequences), e.g., at least 10 3 , 10 4 , 10 5 Diversity or at least 10 6 In another example, the heavy chain or VHH library may comprise at least 10 2 Diverse heavy chain CDR3 (at least 10 2 unique heavy chain CDR3 sequences), e.g., at least 10 3 , 10 4 , 10 5 Diversity, at least 10 6 Diversity, at least 10 7 Diversity, or at least 10 8 It may include diversity.

[0062] In some examples, a heavy chain or VHH library may contain diversity only in the heavy chain CDR1 or heavy chain CDR2. In one particular example, a heavy chain or VHH library contains diversity in all of the heavy chain CDR1, CDR2, and CDR3 regions.

[0063] In some embodiments, the heavy chain or VHH library is a secondary library generated for affinity maturation of a preselected antibody (parent antibody) with binding activity against a target antigen. Such a secondary library may contain diversity in one or two of the heavy chain CDR regions while retaining the other CDR sequence(s) of the parent antibody. For example, the secondary library may contain the same heavy chain CDR1 and CDR2 sequences as the parent antibody, as well as a diverse population of heavy chain CDR3 sequences. Alternatively, the secondary library may contain the same heavy chain CDR3 sequence as the parent antibody, as well as a diverse population of heavy chain CDR1 and / or CDR2 sequences.

[0064] II. Antibody library screening Any of the antibody libraries described herein can be used to screen for antibodies with binding specificity to an antigen of interest. The antibodies encoded by the nucleic acids in the library can be expressed and displayed using a suitable expression / display system, for example, a cell-free display system (e.g., ribosome display), a phage display system, a prokaryotic cell-based display system (e.g., bacterial display), or a eukaryotic cell-based display system (e.g., yeast display or mammalian cell display). In certain embodiments, the antibody library is expressed and displayed on yeast cells. In other embodiments, the antibody library is expressed and displayed on phage particles (phage display). In other embodiments, more than one display system is used, for example, phage display followed by yeast display.

[0065] The library of antibodies can be expressed / displayed in any format in a suitable system, such as those described herein.

[0066] Phage display is a protein display format that uses bacteriophages (e.g., phages f1, fd, and M13). In this system, at least one antibody chain (e.g., a heavy chain) is usually covalently linked to a bacteriophage coat protein, such as the gene III protein, gene VIII protein, or major coat protein (see, e.g., WO 00 / 71694). Phage display has been described, for example, in US 5,223,409, Smith (1985) Science 228:1315-1317; WO 92 / 18619; WO 91 / 17271; WO 92 / 20791; WO 92 / 15679; WO 93 / 01288; WO 92 / 01047; WO 92 / 09690; WO 90 / 02809, de Haard et al. (1999) J. Biol. Chem 274:18218-30, Hoogenboom et al. (1998) Immunotechnology 4:1-20, Hoogenboom et al. (2000) Immunol Today 2:371-8, Fuchs et al. (1991) Bio / Technology 9:1370-1372, Hay et al. (1992) Hum Antibod Hybridomas 3:81-85, Huse et al. (1989) Science 246:1275-1281, Griffiths et al. (1993) EMBO J 12:725-734, Hawkins et al. (1992) J Mol Biol 226:889-896, Clackson et al. (1991) Nature 352:624-628, Gram et al. (1992) PNAS 89:3576-3580, Garrard et al. (1991) Bio / Technology 9:1373-1377, and Hoogenboom et al. (1991) Nuc Acid Res 19:4133-4137.

[0067] Bacteriophage displaying the protein component can be grown and recovered using standard phage preparation methods, such as PEG precipitation from the growth medium. After selection of individual display phages, the nucleic acid encoding the selected protein component can be isolated from cells infected with the selected phage after amplification, or from the phage itself. Individual colonies or plaques can be selected and the nucleic acid isolated and sequenced.

[0068] In other embodiments, eukaryotic expression / display systems, such as yeast cells or mammalian cells, can be used to express and display the libraries of VHHs described herein. Yeast display is a protein display format in which a protein component (e.g., an antibody component) is directly or indirectly linked to a yeast cell wall protein (e.g., Aga1p or Aga2p). In some cases, the VHHs can be covalently fused to a yeast cell wall protein for direct display. In other examples, the association between the VHHs and the yeast cell wall component can be mediated by an intermediate agent. Yeast display is described, for example, in Cho et al., J. Immunol. Methods, 220(1-2):179-188, 1998; Boder et al., Methods Enzymol. 192(2):243-248, 2000; van den Beucken et al., FEBS Lett 546(2-3):288-294, 2003; and Boder et al., Arch Biochem Biophys 526(2):99-106, 2012.

[0069] To screen a VHH library as described herein to isolate VHHs capable of binding to a target antigen, the library of VHHs can be contacted with the target antigen under suitable conditions that allow antibody-antigen binding. Phage particles or host cells that show VHH binding to the target antigen can be isolated, for example, by retention or by a support material on which the target antigen is immobilized, and optionally amplified, and the nucleic acid encoding the displayed VHH can be determined. The screening process can be repeated multiple times and display systems can be used in combination. Different antigens can be used as needed to select VHH members with the desired binding specificity or for negative selection to eliminate VHH members with binding activity for non-target antigens.

[0070] Screening of VHHs derived from the libraries described herein can be performed by any suitable means. For example, binding activity can be assessed by standard immunoassays and / or affinity chromatography. Determination of the ability of candidate VHHs to bind to therapeutic targets can be assayed in vitro, for example, using a BIACORE™ instrument that measures the binding rate of VHHs to a given target antigen based on surface plasmon resonance. In vitro assays can be performed using any of a number of animal models and then tested in humans, if desired. Cell-based biological assays are also contemplated.

[0071] Lead VHHs identified in the VHH library screening may be subjected to affinity maturation as described herein. The secondary library resulting from affinity maturation may be screened for binders with desired properties, e.g., high binding affinity and / or binding specificity, according to routine practices and / or the disclosure provided herein.

[0072] general technique The practice of the present disclosure will employ, unless otherwise indicated, conventional techniques of molecular biology (including recombinant techniques), microbiology, cell biology, biochemistry, and immunology, which are within the skill of one in the art. Such techniques are fully described, for example, in the following publications: Molecular Cloning: A Laboratory Manual, second edition (Sambrook, et al., 1989) Cold Spring Harbor Press; Oligonucleotide Synthesis (M. J. Gait, ed. 1984); Methods in Molecular Biology, Humana Press; Cell Biology: A Laboratory Notebook (J. E. Elis, ed., 1989) Academic Press; Animal Cell Culture (R. I. Freshney, ed. 1987); Introduction to Cell and Tissue Culture (J. P. Mather and P. E. Oberts, 1998) Plenum Press; Cell and Tissue Culture: Laboratory Procedures (A. Doyle, J. B. Griffiths, and D. G. Newell, eds. 1993-8) J. Wiley and Sons; Methods in Enzymology (Academic Press, Inc.); Handbook of Experimental Immunology (D. M. Weir and CCBlackwell, eds.): Gene Transfer Vectors for Mammalian Cells (JMMiller and MP Calos, eds., 1987); Current Protocols in Molecular Biology (FMAusubel, et al. eds. 1987); PCR: The Polymerase Chain Reaction, (Mullis, et al., eds. 1994); Current Protocols in Immunology (JEColigan et al., eds., 1991); Short Protocols in Molecular Biology (Wiley and Sons, 1999); Immunobiology (C.A. Janeway and P. Travers, 1997); Antibodies (P. Finch, 1997); Antibodies: a practice approach (D. Catty., ed., IRL Press, 1988 - 1989); Monoclonal antibodies: a practical approach (P. Shepherd and C. Dean, eds., Oxford University Press, 2000); Using antibodies: a laboratory manual (E. Harlow and D. Lane (Cold Spring Harbor Laboratory Press, 1999); The Antibodies (M. Zanetti and J.D. Capra, eds. Harwood Academic Publishers, 1995); DNA Cloning: A practical Approach, Volumes I and II (D.N. Glover ed. 1985); Nucleic Acid Hybridization (B.D. Hames & S.J. Higgins eds. (1985》; Transcription and Translation (B.D. Hames & S.J. Higgins, eds. (1984》; Animal Cell Culture (R.I. Freshney, ed. (1986》; Immobilized Cells and Enzymes (lRL Press, (1986》; and B. Perbal, A practical Guide To Molecular Cloning (1984); F.M. Ausubel et al. (eds.).

[0073] It should be noted that there seem to be some encoding or formatting issues in the original text, such as "《" which might cause problems in proper display and further processing. The translation is done as accurately as possible based on the provided text.Without further elaboration, it is believed that one skilled in the art can utilize the present invention to its fullest extent based on the above description.The following specific embodiments are therefore to be construed as merely illustrative, and in no way limiting to the remainder of the disclosure.All publications cited herein are incorporated by reference for the purposes or subject matter referred to herein. EXAMPLES

[0074] Example 1: Identification of suitable VHH molecules for use as framework scaffolds. This example illustrates that selecting a suitable VHH framework scaffold for CDR insertion is essential to generate a highly diverse and highly functional single domain antibody (sdAb) library. The general rationale for selecting a framework scaffold is as follows: (i) the framework scaffold is selected from known antibodies; (ii) the framework scaffold has been widely used by others; (iii) The framework scaffold contains germline sequences, which may or may not be consensus sequences.

[0075] Herein, an alternative approach is provided to identify suitable VHHs for use as framework scaffolds for antibody library construction aimed at the development of therapeutic single domain antibodies. Five factors were considered: (i) Currently approved for therapeutic use or in clinical trials; (ii) It has affinity for Staphylococcus aureus protein A, facilitating purification during production. VHH domains are similar to human VH3 domains, many of which are known to have affinity for protein A. (iii) lacking cysteine ​​residues, except for the canonical pair in frameworks 1 and 3 (Kabat positions 22 and 92); (iv) has an arginine residue in framework 2, as opposed to the canonical leucine residue (Kabat position 45) present at the VH / VL interface of human antibodies; (v) Has a canonical tryptophan residue as the first residue (Kabat position 103) in framework 4. In the final library, this region was sourced from a human donor and the majority of sequences have a tryptophan residue at this position.

[0076] Nine therapeutic VHH domains belonging to seven different therapeutic molecules were analyzed (Table 1), some of which were multidomain / multispecific (numbers indicate domains analyzed). Furthermore, ozoralizumab, isecalosumab, soneloximab, and bovalilizumab all share one domain targeting human serum albumin (HSA), which was analyzed only once. Protein A binding was predicted based on amino acid sequence. 1,2 (Figure 1). Four therapeutic VHH domains, namely caplacizumab, isecalosumab (1), soneloximab (3), and bovalilizumab (1), were selected as the final framework scaffold options since they fulfilled the above five requirements (Table 2 and Figure 2).

[0077] One of these (isecalosumab) was also synthesized with two framework mutations (V11L and L89V) making them identical to the other selected frameworks and the VH3-23 germline.

[0078] For each of the therapeutic antibodies listed in Table 2, three vectors were designed and synthesized as described in Example 4. A diagrammatic representation of the vectors encoding the original framework scaffolds, CDR1 derived from the four original framework scaffolds listed in Table 2, and the CDR2 framework scaffolds is shown in Figure 3. Naturally occurring replicated CDRs can be inserted into these selected exemplary antibody framework scaffolds. [Table 1]

[0079] Example 2: Expression of selected framework scaffolds on the surface of filamentous bacteriophage. The phage display technique consists of expressing foreign molecules on the surface of bacteriophages, more commonly filamentous bacteriophage (M13). This technique is widely used for antibody development. To evaluate the suitability of the four selected framework scaffolds used with the above mentioned technique, four polynucleotides (Table 3) encoding VHHs corresponding to each framework scaffold (Table 3) were synthesized and inserted into a phage display vector (phagemid pDAN5). This vector contains a cloning site upstream of g3 of the filamentous phage. Restriction enzyme sites for BssHII and NheI were inserted into the four polynucleotides with an SV5 tag sequence downstream of the NheI site to aid in the detection of recombinant protein expression. [Table 2] [Table 3-1] [Table 3-2] [Table 3-3]

[0080] The four constructs (VHH therapeutic framework scaffolds inserted into pDAN5) were transformed into E. coli (strain Omnimax2) and phage particles were produced with the assistance of helper phage M13KO7. Display of VHH by the phage was assessed by SDS-PAGE+Western blot using an antibody specifically recognizing the SV5 expression tag (Figure 4) and confirmed by the observation of a pIII-VHH band.

[0081] Phage particles were also used in ELISA assays to detect framework scaffolds binding to their respective targets. Targets vWF, ADAMTSL5, IL-17A / IL-17F, or IL6R were individually immobilized in wells of a 96-well microtiter plate. Wells were blocked using PBST / 5% milk, washed, and incubated with phage particles. After washing away unbound phages, phage binding was determined by incubation with horseradish peroxidase anti-M13 phage antibody followed by visualization with TMB (3,3',5,5'-tetramethyl-benzidine, absorbance at 450 nm). Specific binding activity was observed for each framework scaffold against its intended target, but not against other targets (Figure 5).

[0082] Example 3: Expression of framework scaffolds on the surface of the yeasts Saccharomyces Cerevisiae and Staphylococcus Aureus - Protein A binding assessment by flow cytometry. Yeast display technology consists of expressing foreign molecules on the surface of yeast cells, typically the budding yeast Saccharomyces cerevisiae. This technology is widely used for antibody development. To evaluate the suitability of the four selected parent framework scaffolds used with the above mentioned technology, four polynucleotides (Table 3) encoding VHHs corresponding to each framework scaffold were inserted into a yeast display vector (phagemid pSYD). This vector contains a cloning site upstream of the AGA2 gene and consists of restriction enzyme sites BssHII and NheI (inserted in the four polynucleotides) and an SV5 tag sequence downstream of the NheI site, useful for detecting the expression of recombinant proteins.

[0083] The four constructs (VHH framework scaffolds inserted in pSYD) were transformed into S. cerevisiae (strain EBY100) and the expression of VHH was induced using galactose. The expression of VHH presentation on yeast was measured by flow cytometry using allophycocyanin (APC)-conjugated Protein A. As controls, two different single-chain Fv (scFv) constructs were also used, one with a human VH3 heavy chain (positive control for Protein A binding) and one with a human VH4 heavy chain (negative control for Protein A binding). All four VHH framework scaffolds showed binding to Protein A as indicated by an increase in fluorescence (shift of the population to the right on the x-axis). None of the uninduced cell populations showed binding to Protein A, indicating that the observed signal originates from the inducible expression of VHH at the surface of the yeast cells (Figure 6).

[0084] Example 4: Construction of vectors for experimental CDR screening. For each of the four libraries created using the four framework scaffolds shown in Table 2, three polynucleotides were synthesized encoding the VHH corresponding to each framework scaffold. One of the three synthesized polynucleotides encoded the unmodified VHH (parent) (Example 3), and the remaining two polynucleotides were modified to replace one of the original CDRs with a combination of restriction enzyme recognition sites including two inverted BsaI sites (a type II enzyme that cuts outside its recognition sequence), an additional SfiI site to ensure cleavage of the vector and to serve as a spacer between the BsaI sites, and an ochre stop codon to prevent frameshifting and expression of background sequences (Figures 3 and 7). To prevent potential contamination of the final library with camelid CDR1 and CDR2, or CDR1 and CDR2 containing sequence liabilities, these CDRs were replaced with close human homologous sequences without the sequence liabilities of the two polynucleotides containing type II restriction enzyme recognition sites.

[0085] Each of these modified polynucleotides encoding the framework scaffolds was cloned into a yeast display vector (pSYD plasmid), and the presence of a stop codon within this sequence prevented expression of the framework scaffold on the yeast surface until the missing CDR was replaced with a functional CDR. The nucleotide and amino acid sequences of exemplary framework scaffolds with missing CDRs are shown in Table 4. Due to codon degeneracy, alternative nucleotide sequences encoding the amino acid sequences shown in Table 4 can instead be used in the present invention. [Table 4-1] [Table 4-2] [Table 4-3]

[0086] Example 5: Generation of a database of CDR sequences and informatic elimination of CDRs exhibiting potential liability. To generate a database of naturally occurring CDRs (CDRs found in naturally occurring antibodies such as human antibodies), we performed next generation sequencing (NGS) of variable genes derived from a total of 40 human donors and grouped into four pools (pools A-D) containing 10 donors each. B cells bearing CD19+ on their surface were purified from Leukopak using magnetic beads. Total RNA was extracted from the cells, after which polyA+RNA was isolated and reverse transcribed using primers specific for the CH1 region of human IgG and IgM. Heavy chain variable regions were amplified by PCR, gel purified, and sequenced using a Novaseq SP 2x250 nucleotide, paired-end sequencing. After merging the paired-end reads and applying quality filters, over 227 million reads were obtained (Table 5). Analysis of the variable gene sequences allowed the identification of a number of CDRs, as shown in Table 6. [Table 5] [Table 6]

[0087] Since VHH antibodies are similar to human VH3 antibodies, we chose to only include CDR1 and CDR2 sequences derived from the human VH3 germline. A total of 128,935 and 389,908 CDR1 and CDR2 sequences were identified, respectively. The IMGT CDR definitions were used.

[0088] Example 6: Informatics elimination of CDRs resulting from sequencing errors and showing potential sequence liability. Following the creation of the CDR database as disclosed in Example 5 above, CDRs that may have arisen as a result of sequencing errors were also computationally eliminated. In general, the presence of a particular CDR in more pools indicates that it is more likely to be genuine and not the result of a sequencing error. Alternatively, the identified CDRs can also be compared to publicly available datasets to verify their presence. It is clear that different threshold numbers can be selected depending on the number of total reads, the number of different pools, and the number of total unique CDRs identified. In this case, we chose to keep CDR sequences found in at least three of our sequenced donor pools, or found in two of our sequenced donor pools plus at least one publicly available dataset, or found in one of our sequenced donor pools plus at least two publicly available datasets. After eliminating CDR sequences that do not meet the aforementioned criteria, a total of 70,093 and 149,962 CDR1 and CDR2 sequences remained, respectively.

[0089] Sequence liabilities are short amino acid sequences that have the potential to create biophysical liabilities in proteins that contain them. A list of exemplary liabilities is provided in Table 6, but it is clear that additional sequence-based liabilities may be identified. The list of unique CDRs previously identified was examined for the occurrence of the listed liabilities, and all CDRs containing liabilities were computationally removed from the list of unique CDRs. In addition, CDRs with unusual lengths (different from 8 amino acids) were also removed. Finally, a total of 22,062 and 16,704 CDR1 and CDR2 sequences, respectively, remained. Logo representations of the final CDR1 and CDR2 sequences are shown in Figures 8A-8B. Additional or different sequence liabilities in any combination can be similarly removed from the database of CDRs, resulting in a list of CDRs without sequence liabilities.

[0090] Example 7: Synthesis and amplification of final CDR corresponding oligonucleotides. The sequences encoding the amino acid sequences found in naturally occurring CDRs are called naturally duplicated CDRs. Oligonucleotides encoding the naturally duplicated CDRs corresponding to those identified for CDR1 and CDR2 after the elimination step described in the above examples were synthesized (Agilent Technologies, Inc., Santa Clara, CA). The CDR-encoding sequences in these oligonucleotides were flanked by 5' and 3' sequences homologous to the framework vector into which the CDR-encoding sequences were cloned. The homologous sequences were used both to amplify the CDR-encoding oligonucleotides and to insert the amplified oligonucleotides into the yeast display vector.

[0091] The pool of duplicated native CDRs amplified using primer pairs specific for each library framework scaffold and CDR position and cloned by homologous recombination into the yeast display vector described in Example 4 resulted in eight different single CDR loop libraries (two libraries per framework scaffold - CDR1-2). The pool of oligonucleotides was subjected to amplification using the following primers: For CDR1: Primers specified in HCDR1-F and HCDR1-R in Table 7 For CDR2: Primers specified in HCDR2-F and HCDR2-R in Table 7

[0092] Exemplary amplification and assembly primer sequences are shown in Tables 7 and 8, respectively. [Table 7-1] [Table 7-2] [Table 8]

[0093] The diversity found in CDR1-2 can be covered relatively easily by array-based oligonucleotide synthesis, but this is because the original diversity is 10 8 This may not be the case for CDR3s, which can easily exceed 10 different CDR3s. Even after excluding liabilities and CDRs seen less than 4 times, NovaSeq (3x10 9 When the diversity was evaluated using the 10 7 This represents a potential 10x increase in synthetic HCDR3 diversity that is addressable by array-based synthesis. 6 This can be addressed by either limiting the diversity of CDR1-2 to fewer sequences or by combining synthetic CDR1-2 diversity with naturally diverse HCDR3 amplified from donor lymphocytes.

[0094] A total of 10 9RNA from Leuko Pak B lymphocytes from 10 donors containing more than 1000 B cells was isolated using the Miltenyi STRAIGHTFROM LEUKOPAK CD19 kit. cDNA was prepared using primers annealing within the IgM constant region. HCDR3 was amplified from cDNA using forward primers (F-L1-HCDR3, F-L3-HCDR3, F-L4-HCDR3) and reverse primer (JH-universal) listed in Table 7. The generated amplicon was then re-amplified with framework scaffold-specific forward primer (HCDR3-ASS-F) and JH-universal reverse primer listed in Table 8. This amplification adds a sequence homologous to the framework vector at the 5' end into which HCDR3 is cloned.

[0095] Example 8: Generation of non-redundant CDR3 duplicated diversity for insertion into libraries. To insert naturally duplicated HCDR3s into VHH CDR3 sites, a subset of sequenced HCDR3s is identified up to the synthesis limit of array (or any other) based synthesis. Currently, such limit is about 1,000,000 sequences. Such HCDR3s are best selected from a sequence database of HCDR3s by their mutual differences, in order to select the widest possible diversity, taking into account the limited diversity accessible using naturally duplicated diversity. Machine learning or artificial intelligence is used to identify a subset of HCDR3s that are as different from each other as possible.

[0096] Many methods known to those skilled in the art are available for identifying the desired subset of HCDR3s, including unsupervised learning clustering algorithms such as K-means, DBSCAN, OPTICS, BIRCH, hierarchical clustering, and the like.

[0097] As exemplified here, 10,000 distinct HCDR3 amino acid sequences derived from next generation sequencing of CD19+ B cells were first assigned a distance score relative to all other sequences in the dataset. The distance metric used is the Levenshtein distance, but one skilled in the art can easily identify other suitable metrics of similarity (e.g., Needleman-Wanscher alignment score) or dissimilarity (e.g., Hamming distance) to serve the same purpose. For a given dataset, the distribution of calculated Levenshtein distances is shown in Figure 9. The scores were then used to feed a modified version of the K-means clustering algorithm suitable for large datasets, setting the number of desired clusters to 1,000, i.e., less than one-tenth the number of input sequences ("Web Scale K-Means Clustering" D. Sculley, Proceeds of the 19th International conferon on World wide web (2010)). Depending on how much one wishes to compress the dataset, a larger or smaller number of clusters can be specified. The size distribution (number of HCDR3s per cluster) is shown in Figure 10. Nearly 500 clusters contain only one sequence, meaning that they have little similarity to other clusters in the data set. Larger clusters are also formed, the largest of which has 52 different HCDR3 sequences. Two exemplary HCDR3 clusters are shown in Figures 11A and 11B.

[0098] A single sequence can be selected from each cluster to be included in the library.Many forms of representative selection can be used by those skilled in the art, for example, sequences that are closest to the cluster consensus and do not have any sequence liability, such as those defined in Example 6.The clustering process can be repeated with larger or similarly sized data sets until a practical synthesis upper limit is reached.The amplification of such sequences can be performed as described for CDR1 and CDR2 in Example 7.

[0099] Example 9: Cloning of CDRs into single-site CDR vectors and selection of functional CDRs. Naturally duplicated CDRs are synthesized based on criteria that should ensure their functionality (e.g., in-frame and removal of liability), but oligonucleotide synthesis may not be 100% accurate. In addition to the problem of inaccurate sequence, other unidentified liability that causes low expression or polyreactivity may be encoded by synthesized oligonucleotide. Selecting functional CDRs may be one of the options to address this issue.

[0100] Each of the CDRs remaining from the above elimination steps was cloned into an appropriate yeast display framework scaffold vector. The coding sequences of exemplary framework scaffolds are shown in Table 4.

[0101] Using CDR1 of library 1 as an example, the CDR1 framework scaffold vector was digested with BsaI and Sfil, leaving a gap at the site of CDR1. The collection of cleaved vectors and amplified CDR1 polynucleotides were then co-transformed together into yeast as described in Example 7. Inside the yeast cells, homologous recombination between the cleaved vector and the CDR1 polynucleotide results in the insertion of the CDR1 polynucleotide into the CDR1 framework scaffold vector. The entire population of CDR1 yeast framework scaffold vectors carrying each of the CDR1 polynucleotides identified above constitutes the CDR1 yeast display library (Figure 3). In the case of the CDR1 framework scaffold vectors, all parts of the VHH domain are constant, except for the CDR1, which is evaluated. Selection of functional CDR1 was performed by sorting for Protein A binding. After the VHH display is induced, the yeast cells are incubated with magnetic beads covalently linked to Protein A. The positive population (VHH binding Protein A) is purified using MACS (magnetic activated cell sorting). The higher the level of VHH that binds to Protein A, the higher the probability that cells will bind to the nanoparticles (especially when competition is used by having the total number of cells far exceed the binding capacity of the nanoparticles), resulting in enrichment of strong binding sequences.

[0102] By sorting a single CDR library by affinity for Protein A, in addition to eliminating CDRs that may adversely affect the affinity of the VHH for Protein A, CDRs that contain stop codons, frameshifts, or are poorly expressed can be eliminated as well. Effective display on the yeast surface has previously been correlated with improved stability and folding in a variety of proteins (Cherf, GM and JR Chochran (2015). "Applications of Yeast Surface Display for Protein Engineering." Methods Mol Biol 1319: 155-175; Pavoor, TV, et al., (2012) "An enhanced approach for engineering thermally stable proteins using yeast display." Protein engineering, design & selection: PEDS 25(10): 625-630; Pepper, LR, et al., (2008). "A decade of yeast surface display technology: where are we now?" Comb Chem High Throughput Screen 11(2): 127-134). After sorting each CDR library, DNA encoding well-expressed CDRs is obtained by isolating DNA from yeast cells expressing well-folded CDRs. Unfiltered and filtered libraries for libraries 1 and 3 are shown in Figures 12 and 13. Populations were analyzed by flow cytometry and Protein A binding levels (x-axis) are represented as histograms. This analysis shows a clear improvement after enrichment with libraries that showed binding levels similar to or exceeding the original parent molecule from which the framework scaffold was generated.

[0103] In the examples provided herein, we sorted for improved Protein A binding levels. However, a similar approach can be taken using any selective method that distinguishes yeast-displayed antibodies with desirable properties (e.g., high expression, low polyreactivity, good developability) from yeast-displayed antibodies with undesirable properties (e.g., low expression, high polyreactivity, low developability). For example, to select for antibodies with reduced polyreactivity, clones displaying VHH antibodies that do not bind to multispecific reagents are selected. Examples of such multispecific reagents include those previously described (Hotzel, I. et al. A strategy for risk mitigation of antibodies with fast clearance. MAbs 4, 753-760, doi:10.4161 / mabs.22189(2012); Xu, Y. et al. Addressing polyspecificity of antibodies selected from an in vitro yeast presentation system: a FACS-based, high-throughput selection and analytical tool. Protein Eng Des Sel 26, 663-670, doi:10.1093 / protein / gzt047(2013); Kelly, RL et al. Chaperone proteins as single component reagents to assess antibody nonspecificity. MAbs 9, 1036-1040, doi:10.1080 / 19420862.2017.1356529(2017).).

[0104] Example 10: Assembly of a full-length VHH library. Once the individual CDR libraries were screened for Protein A binding (CDR1 and CDR2) or amplified from donor B cells (CDR3), the CDRs were assembled into VHH domains containing diversity in all three CDRs. Individual CDRs and adjacent framework regions were amplified using the primers listed in Table 8. As shown in Figure 3, three individual fragments corresponding to CDR1, CDR2, and CDR3 (and their adjacent frameworks) of each library were assembled using overlap PCR to form the VHH chains.

[0105] Example 11: Cloning into a phage display vector (pDan5). Once the VHH libraries were assembled, they were ligated into a phage display vector such as pDAN5 to investigate their functionality. This vector contains a cloning site upstream of g3 of the filamentous phage, consisting of restriction enzyme sites for BssHII and NheI. The VHHs generated in Example 10 contained a BssHII restriction enzyme recognition site upstream of the VHH and NheI downstream of the VHH. The PCR products were then digested with the same enzymes to generate sticky ends. The pDAN5 plasmid was grown in E. coli, extracted by alkaline lysis, and purified on a cesium chloride / ethidium bromide gradient. The plasmid was digested with the same enzymes and the backbone was purified by agarose gel electrophoresis extraction followed by chromatography to remove contaminants. The backbone was ligated to the VHH library overnight at 16°C using T4 DNA ligase. The ligation was purified and electrotransformed into electrocompetent E. coli TG1 cells. The transformed cells were plated on agar plates containing carbenicillin, glucose, and sucrose to select for bacteria that had received the plasmid. A total of 2.3 × 10 10 Transformants were obtained (Table 9). [Table 9]

[0106] Example 12: Generation of bacteriophage particles containing Western blots. Transformed bacteria were grown in shake flasks containing liquid 2xYT medium + carbenicillin + glucose (glucose to inhibit VHH expression) at 37°C until an OD600nm of 0.5 was reached. Bacteria were superinfected with M13KO7 helper phage (multiplicity of infection of 5) for 30 min at 37°C without shaking and for 30 min at 37°C with shaking. Bacteria were centrifuged, the medium removed and replaced with 2xYT medium + carbenicillin + kanamycin, and grown for 16 h at 25°C in a shaker incubator.

[0107] To recover the phage particles, the culture was centrifuged to separate the bacteria from the supernatant in which the phage were found. The supernatant was mixed with 20% PEG 8000 + 2.5M NaCl solution in a 5:1 ratio, which precipitates the phages, allowing them to be recovered by centrifugation. The supernatant was discarded and the phage pellet was resuspended in PBS solution. The display of VHHs by the phages was assessed by SDS-PAGE + Western blot using an antibody that specifically recognizes the expression tag (SV5), as shown in Figure 14B.

[0108] Example 13: Selection of antibodies against interferon-alpha-2b by combined phage and yeast display using libraries. After construction and production of phage particles, the library was screened against the target of interest. Selection can be performed using phage display alone, but it is preferable to combine phage display technology with yeast display technology. 12 phage particles were used in two rounds of selection against biotinylated recombinant interferon-alpha using the Kingfisher magnetic bead system. 2x10 phage particles coated with biotinylated protein (100-400 nM) were used in two rounds of selection against biotinylated recombinant interferon-alpha using the Kingfisher magnetic bead system. 7Streptavidin-conjugated magnetic beads (Dynabeads M-280) were washed, coated with antigen, incubated with phage particles, and washed again to remove non-binders. Phage particles were then eluted by lowering the pH and infecting F'-fimbriated bacteria (Ominmax-2T1, Thermo Fisher Scientific). Phages were grown and the selection cycles were repeated. After two rounds of phage enrichment, VHHs were PCR amplified and transferred by homologous recombination into a C-terminal yeast display system (pDNL6 yeast display vector). Here, VHHs are displayed fused to the C-terminus of Aga-2. The transformed yeast was then induced for VHH display by adding galactose to the medium. The induced yeast minilibrary was then used for three more rounds of enrichment against biotinylated recombinant human antigens by fluorescence-activated cell sorting. The target concentration in the first sorting round was 100 nM and reduced to 5 nM in rounds 2 and 3. Two of the libraries underwent additional rounds of negative selection to remove streptavidin-binding clones. After these multiple rounds of phage and yeast sorting enrichment, the recovered populations were analyzed by flow cytometry to test for binding to antigen in decreasing concentrations of antigen and in the absence of antigen, and to check for non-specific binding to the secondary reagent (Figure 15). The results show that this library can successfully yield high affinity binders to the antigens tested.

[0109] Example 14: Determination of the affinity of selected antibodies. Affinity determination of antibodies selected from a naive library using the phage + yeast display protocol described in Example 13 was performed according to the approach described herein. The binding affinity of the VHH antibody variants thus obtained to interferon-alpha was examined using a Carterra LSA instrument. Briefly, the VHH population was subcloned into a pDAN5 vector and transformed into E. coli Top10F' cells to produce soluble VHH molecules. Supernatants from bacteria expressing VHH proteins were immobilized on a Carterra LSA HC30M chip coupled to an anti-SV5 antibody. The chip was activated with 1:1:1 100 mM MES pH 5.5, 100 mM S-NHS, 400 mM EDC (all reconstituted in MES 5.5) and 100 μL of each was mixed in a vial immediately before running the assay. Monoclonal goat anti-SV5 IgG was immobilized at 50 μg / mL and subsequently deactivated with 1 M ethanolamine pH 8.5.

[0110] Supernatants were diluted in HBSTE buffer and circulated across the anti-SV5 surface for 5 minutes. Antigens were tested in a 4-fold dilution series starting at 150 nM. Antigen samples were tested from the lowest to the highest concentration. Data were processed using Carterra LSA software. Results were double referenced, trimmed, and fitted using a 1:1 kinetic model with fixed Rmax and floating T0 parameters. The highest antigen concentration injection was excluded to improve the fit.

[0111] As shown in FIG. 16, the affinities of antibodies selected directly from the libraries constructed as described in Examples 1-12 were shown to be extremely strong, with the majority having affinities below 10 nM.

[0112] Example 15: Selection of antibodies against B7-H4 protein by combinatorial phage and yeast display using libraries. To further validate the library construction and concept, it was selected against another target of interest (B7-H4 protein). A similar approach was taken as shown in Example 12, in this case phage particles were subjected to selection against biotinylated antigen using a similar protocol. After two rounds of phage enrichment, VHHs were PCR amplified and transferred by homologous recombination into a C-terminal yeast display system (pDNL6 yeast display vector), where VHHs are displayed fused to the C-terminus of Aga-2. The transformed yeast was then induced for VHH display by adding galactose to the medium. The induced yeast minilibrary was then used for another round of positive selection against biotinylated antigen at 100 nM concentration. This round was followed by a negative selection to select only those clones that did not show binding to the secondary reagents (anti-SV5 monoclonal antibody labeled with a PE fluorophore and streptavidin labeled with an AlexaFluor633 fluorophore). All four sub-libraries were subjected to another round of positive selection, here at an antigen concentration equal to 20 nM, followed by libraries 1 and 3 at an antigen concentration equal to 5 nM. After these rounds of phage and yeast selection enrichment, the recovered populations were analyzed by flow cytometry to test for binding to the antigen in decreasing concentrations of antigen and in the absence of antigen, and to check for non-specific binding to the secondary reagent. The populations tested were those obtained after two positive and one negative rounds of yeast display selection (Figure 17) and those obtained after three positive and one negative rounds of yeast display selection (Figure 18). The results show that this library can successfully yield high affinity binders to the antigens tested.

[0113] Example 16: Next generation sequencing of selected yeast populations for the antigens interferon-alpha and B7-H4. To identify the sequences present in a given selected population, low-throughput DNA sequencing techniques such as the Sanger method, or alternatively, next-generation sequencing techniques such as those in the Illumina platform, can be applied. The latter can provide a much larger number of sequences, allowing for better profiling of the selected population. Selected populations from interferon-alpha and B7-H4 selection campaigns, such as those illustrated in Table 10, were subjected to next-generation sequencing using Illumina Miseq. The resulting sequences were annotated to identify scaffold and CDR identity. Also, to identify more distinct clones, clustering techniques such as those described in Example 8 can be applied, or in this example, clustering was performed by recoding the CDR3 amino acid sequence to reflect their physicochemical properties. They were then compared with each other to quantify the Levenshtein distance, and finally clustered using the OPTICS method.

[0114] Of the four VHH libraries, a total of 215 CDR3 clusters were identified for B7H4 (Figure 19), and 248 CDR3 clusters were identified for interferon-alpha for these four libraries (Figure 22). For each population, up to 100 unique CDR3 sequences were compared to each other using the Levenshtein distance, and alternatively, up to 100 of the top representative CDR3s of each cluster were selected for comparison using the Levenshtein distance. The results for B7-H4 are shown in Figure 20, and the results for interferon-alpha are shown in Figure 23. It is also possible to compare all CDRs between the identified clones. Up to 100 unique clones were compared for all three CDRs using the Levenshtein distance, and alternatively, up to 100 representative clones of each cluster were selected for comparison using the Levenshtein distance. The results for B7-H4 are shown in Figure 21, and the results for interferon-alpha are shown in Figure 24. As can be seen in both figures, the representatives of the identified clusters and the top unique CDR3s are significantly different from each other. [Table 10]

[0115] Example 16: Antibody Maturation. The following approach can be used to select optimized VHHs and assemble CDRs into mature antibodies. First, two libraries are created: one in which CDR1 is replaced with the CDR from a naive library made using the corresponding framework scaffold, and CDR2 and 3 are kept constant; the other in which CDR2 is replaced with the CDR from a naive library made using the corresponding framework scaffold, and CDR1 and 3 are kept constant. The libraries are selected for high affinity binding to the antigen. Then, the two selected libraries are combined and further selected for higher affinity to the target to obtain mature antibodies.

[0116] Example 17: Construction of libraries with modified frameworks. The selected scaffold can be modified to achieve a desired goal, such as making it more similar to the native germline. The isecarosumab (1) scaffold was mutated at two separate positions: V11L and L89V (Kabat numbering) to generate a novel scaffold that is closer to both human germline (IGHV3-23) and alpaca germline (IGHV3-3). This scaffold was subjected to repeated yeast display CDR filtering for the Protein A binding process (Figure 26). The filtered CDRs were amplified, combined with the amplified CDR3 from the donor, cloned into the pDAN5 phage display vector, and transformed into E. coli TG1 cells by electroporation to produce 4.8x10 9 Transformants (library 2A) were generated.

[0117] Example 18: Generation of polyclonal binding populations from libraries with modified frameworks. The modified library was used for phage particle production and screened against the target of interest. Selection can be performed using phage display alone, but it is preferable to combine phage display and yeast display techniques. 12 phage particles were used in two rounds of selection against biotinylated recombinant interferon-alpha (IFN) and receptor binding domain (RBD) derived from Sars-CoV-2 using the Kingfisher magnetic bead system. 2x10 phage particles coated with biotinylated proteins (100-400 nM) were used in two rounds of selection against biotinylated recombinant interferon-alpha (IFN) and receptor binding domain (RBD) derived from Sars-CoV-2 using the Kingfisher magnetic bead system. 7 Streptavidin-conjugated magnetic beads (Dynabeads M-280) were washed, coated with antigen, incubated with phage particles, and washed again to remove non-binders. Phage particles were then eluted by lowering the pH and infecting F'-fimbriated bacteria (Ominmax-2T1, Thermo Fisher Scientific), and the output titers are shown in Table 11. Phages were propagated and the selection cycle was repeated. After two rounds of phage enrichment, VHHs were PCR amplified and transferred by homologous recombination into a C-terminal yeast display system (pDNL6 yeast display vector). Here, VHHs are displayed fused to the C-terminus of Aga-2. The transformed yeast was then induced for VHH display by adding galactose to the medium. The induced yeast minilibrary was then used for two more rounds of enrichment against biotinylated recombinant antigen by fluorescence-activated cell sorting. In the case of RBD, a third round was performed using negative sorting to remove streptavidin binders. In both positive enrichment rounds, the target concentration was kept at 100 nM. After phage and yeast sorting enrichment, the recovered populations were analyzed by flow cytometry to test for binding to antigen and checked for non-specific binding to secondary reagents in the absence of antigen (IFN: FIG. 27; RBD: FIG. 28). The results show that the library successfully yielded binders to the antigens tested. [Table 11]

[0118] Other embodiments All features disclosed herein may be combined in any combination. Each feature disclosed herein may be replaced by an alternative feature serving the same, equivalent, or similar purpose. Thus, unless expressly indicated otherwise, each feature disclosed is only one example of a generic series of equivalent or similar features.

[0119] From the above description, those skilled in the art can easily ascertain the essential features of the present invention, and can make various changes and modifications to the present invention to adapt it to various applications and situations without departing from the spirit and scope of the present invention. Accordingly, other embodiments are within the scope of the following claims.

[0120] Equivalent While several inventive embodiments have been described and illustrated herein, those of ordinary skill in the art will readily envision various other means and / or structures for performing the functions and / or obtaining one or more of the results and / or advantages described herein. Each such change and / or modification is deemed to be within the scope of the inventive embodiments described herein. More generally, those of ordinary skill in the art will readily appreciate that all parameters, dimensions, materials, and configurations described herein are intended to be exemplary, and that the actual parameters, dimensions, materials, and / or configurations will depend on the particular application or applications for which the teachings of the present invention are used. Those of ordinary skill in the art will recognize or be able to ascertain, using no more than routine experimentation, many equivalents to the specific inventive embodiments described herein. Thus, the foregoing embodiments are presented by way of example only, and it will be understood that, within the scope of the appended claims and their equivalents, the inventive embodiments may be practiced other than as specifically described and claimed. Inventive embodiments of the present disclosure are directed to each individual feature, system, article, material, kit, and / or method described herein. Furthermore, any combination of two or more of such features, systems, articles, materials, kits, and / or methods is included within the inventive scope of the present disclosure, unless such features, systems, articles, materials, kits, and / or methods are mutually inconsistent.

Claims

1. A VHH antibody library, comprising: Nucleic acids encoding framework region 1, framework region 2, framework region 3 and framework region 4; and A plurality of nucleic acids encoding a population of VHH genes comprising a population of CDR1s, a population of CDR2s, and a population of CDR3s located in the CDR1, CDR2, and CDR3 regions of said VHH genes, wherein the amino acid sequences of said CDR1s, said CDR2s, and said CDR3s are naturally replicated from naturally occurring antibodies, and at least 90% of said population of CDR1s and at least 90% of said population of CDR2s are completely free of members comprising one or more of the following: (i) a glycosylation site comprising the motif NXS, NXT, or NXC, where X represents any naturally occurring amino acid residue except proline; (ii) a deamidation site comprising an NG, NS, NT, NN, NA, NH, ND, NQ, NF, NW, or NY motif; (iii) an isomerization site comprising a DT, DH, DS, DG, DN, DR, DY or DD motif; (iv) any cysteine; (v) a net charge greater than 1; (vi) a tripeptide motif containing at least two residues with aromatic side chains, including F, H, W, or Y; (vii) a multispecific site comprising the motifs GG, GGG, RR, VG, W, WV, WW, WWW, YY, or WXW, where X represents any amino acid residue; (viii) a protease-sensitive or hydrolytically susceptible site comprising a DX motif, where X is P, G, S, V, Y, F, Q, K, L, or D; (ix) an integrin binding site comprising RGD, RYD, LDV, or KGD; (x) a lysine glycosylation site including KE, EK, or ED; (xi) a metal-catalyzed fragmentation site comprising an HS, SH, KT, HXS, or SXH motif, where X represents any amino acid residue; (xii) X １ X ２ X ３ A multispecific aggregation site comprising the motif １ , X ２ , and X ３ each of which is independently selected from the group consisting of F, I, L, V, W and Y), (xiii) a streptavidin binding motif comprising the motif HPQ, EPDW (SEQ ID NO: 49), PWXWL (SEQ ID NO: 50) (wherein X represents any amino acid residue), GDWVFI (SEQ ID NO: 51), or PWPWLG (SEQ ID NO: 52); (xiv) one or more arginines; (xv) hydrophobic CDR sequences that are summed using reference numbers from Parker JM et al. Biochemistry. 1986 Sep 23;25(19):5425-32, resulting in a value less than zero; (xvi) CDR mutations that reduce binding to Protein A, including any mutation in the last amino acid of said CDR2 to A, G, C, D, E, F, G, H, I, L, M, N, P, Q, S, V, W or Y according to the IMGT definition; The VHH antibody library, comprising: wherein at least 95% of the population of CDR1, at least 95% of the population of CDR2, and at least 95% of the population of CDR3 are completely free of non-functional members; framework region 1, framework region 2, framework region 3 and framework region 4 are derived from a single therapeutic antibody VHH; and The VHH antibody library, wherein each framework region can contain up to five amino acid substitutions.

2. 2. The VHH antibody library of claim 1, wherein the single VHH domain is selected from the therapeutic antibody selected from the group consisting of caplacizumab, embafolimab, gontivimab, isecalosumab, ozoralizumab, soneloximab, and bovalilizumab.

3. 3. The VHH antibody library of claim 1 or claim 2, wherein the naturally replicated nucleic acid sequences of the CDR1, the CDR2, and the CDR3 are synthetic.

4. 3. The VHH antibody library of claim 1 or claim 2, wherein the nucleic acid sequences of the CDR3s are naturally replicated by obtaining sequences derived from heavy chain CDR3s from donor lymphocytes.

5. 1. A method for generating an antibody library, comprising: (a) providing a first plurality of nucleic acids encoding a population of naturally replicated CDR1 fragments; (b) providing a second plurality of nucleic acids encoding a population of naturally replicated CDR2 fragments; (c) providing a third plurality of nucleic acids encoding a population of naturally replicated CDR3 fragments; (d) providing nucleic acid genes encoding a common VHH domain comprising heavy chain framework region 1, heavy chain framework region 2, heavy chain framework region 3, and heavy chain framework region 4; and inserting the first plurality of nucleic acids, the second plurality of nucleic acids, and the third plurality of nucleic acids into the CDR1 region, the CDR2 region, and the CDR3 region of the common VHH domain, respectively, thereby generating a population of nucleic acids encoding a VHH domain library, wherein at least 90% of the population of heavy chain CDR1s and at least 90% of the population of heavy chain CDR2s are completely free of members comprising one or more of the following: (i) a glycosylation site comprising the motif NXS, NXT, or NXC, where X represents any naturally occurring amino acid residue except proline; (ii) a deamidation site comprising an NG, NS, NT, NN, NA, NH, ND, NQ, NF, NW, or NY motif; (iii) an isomerization site comprising a DT, DH, DS, DG, DN, DR, DY or DD motif; (iv) any cysteine; (v) a net charge greater than 1; (vi) a tripeptide motif containing at least two residues with aromatic side chains, including F, H, W, or Y; (vii) a multispecific site comprising the motifs GG, GGG, RR, VG, W, WV, WW, WWW, YY, or WXW, where X represents any amino acid residue; (viii) a protease-sensitive or hydrolytically susceptible site comprising a DX motif, where X is P, G, S, V, Y, F, Q, K, L, or D; (ix) an integrin binding site comprising RGD, RYD, LDV, or KGD; (x) a lysine glycosylation site including KE, EK, or ED; (xi) a metal-catalyzed fragmentation site comprising an HS, SH, KT, HXS, or SXH motif, where X represents any amino acid residue; (xii) X １ X ２ X ３ A multispecific aggregation site comprising the motif １ , X ２ , and X ３ each of which is independently selected from the group consisting of F, I, L, V, W and Y), (xiii) a streptavidin binding motif comprising the motif HPQ, EPDW (SEQ ID NO: 49), PWXWL (SEQ ID NO: 50) (wherein X represents any amino acid residue), GDWVFI (SEQ ID NO: 51), or PWPWLG (SEQ ID NO: 52); (xiv) one or more arginines; (xv) hydrophobic CDR sequences that are summed using reference numbers from Parker JM et al. Biochemistry. 1986 Sep 23;25(19):5425-32, resulting in a value less than zero; (xvi) CDR mutations that reduce binding to Protein A, including any mutation in the last amino acid of said CDR2 to A, G, C, D, E, F, G, H, I, L, M, N, P, Q, S, V, W or Y according to the IMGT definition; wherein at least 95% of the population of heavy chain CDR1s, the population of heavy chain CDR2s, and the population of heavy chain CDR3s are completely free of non-functional members; framework 1, framework 2, framework 3 and framework 4 are derived from a single VHH domain derived from a therapeutic antibody consisting of one or more VHH domains; and The method, wherein each framework region can contain up to five amino acid substitutions.

6. 6. The method of claim 5, wherein the therapeutic antibody consisting of one or more VHH domains is selected from the group consisting of caplacizumab, embafolimab, gontivimab, isecalosumab, ozoralizumab, soneloximab and bovalilizumab.

7. the first plurality of nucleic acids, the second plurality of nucleic acids, and the third plurality of nucleic acids, (a) obtaining the amino acid sequences of the heavy chain CDR1 region, the heavy chain CDR2 region, and the heavy chain CDR3 region of a population of naturally occurring antibodies; (b) excluding amino acid sequences containing any or all of (i) to (xvi) from the heavy chain CDR1 amino acid sequence, the heavy chain CDR2 amino acid sequence, and the heavy chain CDR3 amino acid sequence of (a), thereby obtaining a non-liability heavy chain CDR1 sequence, a non-liability heavy chain CDR2 sequence, and a non-liability heavy chain CDR3 sequence; (c) synthesizing the first plurality of nucleic acids encoding the non-liability heavy chain CDR1 region, the second plurality of nucleic acids encoding the non-liability heavy chain CDR2 region, and the third plurality of nucleic acids encoding the non-liability heavy chain CDR2 region; 7. The method of claim 5 or 6, produced by a process comprising:

8. the third plurality of nucleic acids comprising: (a) amplifying the heavy chain CDR3 region from a population of B cells; (b) combining the third plurality of nucleic acids encoding the heavy chain CDR3 regions obtained in (a) with the remaining CDRs; The method of claim 7 produced by a process comprising:

9. the processes for generating the first plurality of nucleic acids, the second plurality of nucleic acids, and the third plurality of nucleic acids further comprise isolating functional members from the non-liability heavy chain CDR1 and CDR2 regions and / or from the CDR3 region; (i) the functional member of the non-liability heavy chain CDR1 and CDR2 region, or the functional member of the CDR3 region, is isolated by expressing the antibody comprising the non-liability heavy chain CDR1 and CDR2 region, and / or CDR3 region in a host cell such that the antibody is displayed on the surface of the host cell, isolating the antibody displayed on the host cell, and identifying within the displayed antibody the CDR1, CDR2, and / or CDR3 region that is a functional member of the CDR1, CDR2, and / or CDR3 region; or (ii) said functional members of said non-liability heavy chain CDR1 and CDR2 regions, and / or CDR3 regions are isolated by expressing an antibody comprising said non-liability heavy chain CDR1 and CDR2 regions, and / or CDR3 regions, optionally in fusion with a folding reporter, which is β-lactamase or green fluorescent protein, or a fragment thereof, to obtain members with improved folding; The method of claim 7.

10. The antibody library of claim 1 , wherein the CDR1 comprises a human homolog sequence.

11. The antibody library of claim 10 , wherein the CDR2 comprises a human homolog sequence.

12. The antibody library of claim 11, wherein the framework regions are derived from a single VHH domain derived from a therapeutic antibody consisting of one or more VHH domains.

13. 6. The method of claim 5, wherein the heavy chain CDR1, CDR2 and CDR3 fragments and the heavy chain variable domain gene are derived from a naturally occurring antibody of a mammalian species.

14. 14. The method of claim 13, wherein the mammalian species is human or camelid.

Images