JP2011115175A - Super humanized antibody - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for humanizing antibodies based on selection of variable region framework sequences derived from human antibody gene. <P>SOLUTION: The selection is based on comparing the canonical CDR (complementarity determining region) structure type of CDR in variable region of nonhuman antibodies with the canonical CDR structure type of corresponding CDR derived from the library of the human antibody sequence (preferably germ line antibody gene segment). The human antibody variable region having the canonical CDR structure type resembling to the nonhuman CDR forms a subset of member human antibody sequences for selecting human framework sequences. The members of these subsets can be further ranked by amino acid similarity between the human CDR sequence and the nonhuman CDR sequence. The human sequences with higher rank are selected to provide framework sequences for constructing chimera antibodies. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

(Cross-reference to related applications)
This application claims priority to US Provisional Application No. 60 / 305,111 filed July 12, 2001.

(Statement of government interest)
The development of the present invention was supported in part by a grant from the National Institutes of Health (Grant No. CA-18029).

(Technical field)
The present invention relates to methods for humanizing antibodies, and more particularly to humanizing antibodies by generating chimeric antibodies comprising CDR sequences derived from non-humanized antibodies and framework sequences of human antibodies, and more particularly A method for selecting an appropriate human antibody framework sequence to perform its humanization, and more particularly as a basis for selecting an appropriate human framework sequence for a humanized antibody. It relates to the use of canonical CDR structural types of non-human antibodies compared to germline canonical CDR structural types.

(Background of the Invention)
Antibodies are natural proteins that the vertebrate immune system forms in response to foreign substances (antigens), primarily for protection against infection. Over a century, antibodies have been induced in animals under artificial conditions and collected for use in the treatment or diagnosis of disease states, or for biological studies. Each individual antibody-producing cell produces a single type of antibody having a chemically defined composition, whereas antibodies obtained directly from animal serum in response to antigen inoculation are actually individual It includes a collection of non-identical molecules generated from a collection of antibody producing cells (ie, a polyclonal antibody).

  Hybridoma technology provides a method for growing single antibody-producing cells for an infinite number of generations, using screening methods to identify clones of cells that produce antibodies that react with specific antigens did. Development of this technology has made it possible to produce infinite quantities of structurally identical antibodies with essentially any desired antigen specificity. Such antibodies are generally referred to as monoclonal antibodies, and most are derived from rodents. Sequencing a monoclonal antibody gene has made it possible to define the primary amino acid structure of the antibody.

  Advances in recombinant DNA technology have enabled the structural manipulation of antibody genes and the generation of modified antibody molecules with properties not obtainable by hybridoma technology. In the therapeutic field, one purpose of this method was to reduce the immunogenicity of rodent monoclonal antibodies in humans by altering the primary amino acid structure of rodent monoclonal antibodies. Reduction of immunogenicity of therapeutic antibodies is desirable. This is because induction of an immune response can cause a series of adverse effects in the patient, which are ultimately lethal from the accelerated elimination of therapeutic antibodies and the resulting loss of efficacy. Ranges up to anaphylaxis.

One strategy to reduce the immunogenicity of a foreign monoclonal antibody is to replace the light and heavy chain constant domains of that monoclonal antibody with a similar domain of human origin to intact the variable region domain of the foreign antibody. It was to leave. The light and heavy chain variable region domains are responsible for the interaction between the antibody and the antigen. The binding domain that binds the variable domain to the constant domain is located in a region away from the antigen binding site, and thus the binding domain between the variable domain and the constant domain generally does not interfere with antigen binding. A chimeric antibody molecule having a mouse variable domain attached to a human constant domain usually binds antigen with the same affinity constant as the mouse antibody from which the chimera was derived. Such chimeric antibodies are less immunogenic in humans than their full mouse counterparts. Nevertheless, antibodies that preserve the entire murine variable domain tend to elicit an immune response in a significant proportion of patients. For example, INFIXIMAB ^™ (a widely prescribed chimeric antibody considered safe) induced a human anti-chimeric antibody response in 7 of 47 patients with Crohn's disease (Rutgeerts, P. et al. (1999). ) Efficiency and safety of retraction with anti-tumor necrosis factor antibody (INFLIXIMB) to maine remission in Chron's disease, Gastroe1 76.

  It was predictable that humans elicit an immune response against all mouse variable domains. Thus, efforts to obtain variable domains with more human characteristics began even before clinical trials of such standard chimeric antibodies were reported. One category of methods often referred to as “humanization” involves the conversion of murine monoclonal antibody variable domains into a more human form by recombinantly constructing antibody variable domains with both murine and human characteristics. Objective. The humanization strategy is based on some consensus understanding of antibody structure data. First, the variable domain comprises a series of peptide sequence regions that are conserved in one species but differ between evolutionarily distantly related species (eg, mouse and human). Second, other contiguous regions are not conserved in certain species, but differ even between antibody producing cells in the same individual. Third, contact between antibodies and antigens occurs primarily through non-conserved regions of variable domains. Fourth, the molecular structure of antibody variable domains is sufficiently similar between species so that the corresponding amino acid residue positions between species can be identified based on the position alone without using experimental data.

  Humanization strategies share the premise that replacing amino acid residues characteristic of mouse sequences with residues found at corresponding positions in human antibodies reduces the immunogenicity of the resulting antibody in humans. However, sequence substitution between species usually results in a decrease in antibody binding to that antigen. Thus, humanization techniques address the need to replace original mouse sequences to reduce immunogenicity and the need for humanized molecules to retain sufficient antigen binding to be therapeutically useful. There is to match. This balance has been solved using two approaches.

  One approach (U.S. Pat. No. 5,869,619 and for Studnicka Padlan (1991) A possible procedure for reducing the immunogenicity of antibody variable domains while preserving their ligand binding properties, Molecular Immunology 28: exemplified by 489 to 498) Characteristically, human residues are determined or presumed to (i) play no significant chemical role in the interaction with the antigen, and (ii) are arranged with the side chain protruding into the solvent In place of murine variable domain residues that are determined or predicted to be Thus, external residues that are remote from the antigen binding site are humanized, while residues that form an interface between internal residues, antigen binding residues, and variable domains remain mouse. One advantage of this approach is to determine whether the residue does not play any significant chemical role in antigen binding, or whether the residue is located in a solvent in a particular three-dimensional antibody structure. Somewhat extensive experimental data is needed to determine this.

  Another more general approach (U.S. Pat. No. 5,225,539 to Winter and Jones et al. (1986) Replacing the complementarity determining regions in a human antibody with this from the mouse, Examples 52-52, 52-52. ), A series of murine variable domain peptide sequence regions that are believed to be conserved are replaced with corresponding regions from a human antibody. In this more general approach, all variable domain residues are humanized except for non-conserved regions involved in antigen binding. To determine the appropriate contiguous region for replacement, Winter and Jones et al. (1986) utilized a classification of antibody variable domain sequences previously developed by Wu and Kabat (1970).

  Wu and Kabat pioneered the alignment of antibody peptide sequences. And their contribution in this respect was several times. First. Through the study of sequence similarity between variable domains, they adopt a similar three-dimensional structure, play a similar functional role, interact as well as neighboring residues, and exist in similar chemical environments As far as possible, corresponding residues were identified that were more or less homologous across all antibodies in all vertebrate species. Second, they devised a peptide sequence numbering system in which homologous immunoglobulin residues were assigned the same position number. One skilled in the art can unambiguously assign what is now commonly referred to as Kabat numbering, to any variable domain sequence, based on no experimental data beyond the sequence itself. Third, for each Kabat numbered sequence position, Kabat and Wu calculated variability. By variability is meant the discovery of a few or many amino acids possible when the variable domain sequences are aligned. They identified three continuous regions with high variability embedded in four less variable continuous regions. Other researchers have previously noted variability in these regions (hypervariable regions) and hypothesized that this highly variable region represents the amino acid residue used for antigen binding. Kabat and Wu formally defined the residues that make up these variable regions and named these “complementarity determining regions” (CDRs). This refers to chemical complementarity between the antibody and the antigen. The role of this variable domain in three-dimensional folding (but not the role in antigen recognition) was attributed to the remaining less variable region (which is here the “framework region”). Fourth, Kabat and Wu established a public database of antibody peptide and nucleic acid sequences. This continues to be maintained and is well known to those skilled in the art.

  Humanization methods using the Kabat classification disclosed by Winter and Jones include CDRs from one antibody and framework regions from another antibody that differ in species origin, specificity, subclass, or other characteristics Generate a chimeric antibody. However, no particular sequence or property can be attributed to the framework region, and in fact, Winter taught that any framework set could be combined with any CDR set. Since then, framework sequences have been recognized as important for conferring the three-dimensional structure of antibody variable regions necessary to maintain good antigen binding. Thus, the general humanization method described by Winter and Jones has the disadvantage of often resulting in inactive antibodies. Because these references reasonably select from many possible human framework sequences the sequences most likely to support the antigen binding required by a particular CDR region from a non-human antibody. This is because the necessary information is not provided. Subsequent developments in this area are improvements in the range of Winter to handle the loss of avidity for antigens observed with several humanized antibodies compared to the avidity of the corresponding murine antibody ( Avidity is a quantitative measure of antibody partitioning in the presence of antigen under conditions that approximate the chemical equilibrium between the free and antigen-bound forms. For reactions, avidity is the same as affinity, which is a biochemical equilibrium constant).

  US Pat. No. 5,693,761 to Queen et al. Discloses certain improvements based on Winter for humanizing antibodies. This is avidity to problems with structural motifs in the humanized framework that prevent CDRs from folding into bindable conformations found in mouse antibodies due to steric or other chemical incompatibilities. Based on the assumption of loss. To address this issue, Queen teaches the use of human framework sequences that are closely homologous in the linear peptide sequence to the framework sequence of the mouse antibody to be humanized. Thus, the Queen method focuses on comparing framework sequences between species. Typically, all available human variable domain sequences are compared to a specific mouse sequence and the percent identity between corresponding framework residues is calculated. The human variable domain with the highest percentage is selected to provide the framework sequence for the humanization project. Queen also teaches that it is important in humanized frameworks to retain certain amino acid residues from the mouse framework that are important to support CDRs in a bindable conformation. Potential importance is assessed from molecular models. Candidate residues for retention are typically those residues that are adjacent to a CDR in a linear sequence, or are physically within 6 cm of any CDR residue.

  In other approaches, the importance of particular framework amino acid residues can be determined once by returning single residues to the mouse sequence and assaying antigen binding, as described by Riechmann et al. (1988). Once a low avidity humanized construct is obtained, it is determined experimentally. Another exemplary approach for identifying the importance of amino acids in framework sequences is disclosed by US Pat. No. 5,821,337 to Carter et al. And US Pat. No. 5,859,205 to Adair et al. The These references disclose specific Kabat residue positions in the framework, which may require substitution in the humanized antibody with the corresponding mouse amino acid to preserve avidity. One of the disadvantages of the improvements by Queen and the Riechmann approach, Carter approach and Adair approach is that a large number of human framework sequences are needed for comparison and / or important amino acid residues. The guidelines for preservation are not completely sufficient to estimate functionality. Thus, constructed frameworks that are partly human and partly murine still frequently exhibit human immunogenicity or reduced antigen binding, thereby being suitable for therapeutic use. In order to get a good framework, it takes many iterations in building the framework.

A second type of improvement to Winter is described by Padlan et al. (1995) Identification of specificity-determining residings in antibodies, FASEB J. 9: 133-139; and Tamura et al. (2000) Structural correlations of an anti-carimanti enti enti enti enti enti edi enti enti edi enti edi enti enti edi enti enti edi enti edi enti edi enti edi enti edi ent i edi enti ent ri ed enti
only, J.M. Immunol. 164: 1432-1441. These references share the premise that increasing the proportion of human sequences characteristically in a humanized antibody decreases the immunogenicity of the antibody, thus disclosing methods for transplanting partial CDR sequences. To do. Determination of the three-dimensional structure of the antibody-antigen complex indicated that many residue positions assigned to the CDRs defined by Kabat and Wu were directly involved in antigen binding. These references have shown that transplanting a subset of CDR residues fully transfers antigen binding to humanized antibodies. However, humanized framework sequences are still required. These references do not teach methods for selecting sufficient human framework sequences for use with a given set of mouse CDRs.

  Thus, in order to support non-human CDR regions and to provide humanized antibodies that retain low immunogenicity and retain high antigen binding in humans, the critical Identify the appropriate human framework sequences reliably without the need to determine important amino acid residues and without having to repeat the construction multiple times to obtain a humanized antibody with appropriate therapeutic properties; There is a need in the art for methods for humanizing antibodies.

(Summary of the Invention)
The present invention provides this need by providing a method for generating high affinity and low immunogenic humanized antibodies without the need to compare framework sequences between non-human and human antibodies. Meet. The invention also provides humanized antibodies produced thereby. Rather than being based on human framework sequences as an analysis point, the methods provided herein canonical canonical CDR structural types of non-human antibodies (especially encoded by human germline sequences) Based on identifying candidate human antibody sequences to obtain appropriate human framework sequences as compared to CDR structural types.

  More particularly, a method of generating a humanized antibody is provided, the method comprising obtaining a peptide sequence of a subject variable region encoded by a non-human mature antibody gene, and at least 2 within the non-human antibody variable region. Includes operations to identify a first set of canonical CDR structure types for one CDR. Thereafter, a library of peptide sequences of human antibody variable regions of human antibodies is also obtained. In a preferred embodiment, the library comprises human germline variable region sequences encoded by germline nucleic acid segments. However, in other embodiments, the library can include mature human antibody sequences. In any case, the method includes identifying a canonical CDR structure type (ie, a second set of canonical CDR structure types) for at least two CDRs for each sequence in the library of human variable region sequences. Is included. From this library, a first set of canonical CDR structure types is compared with a second set of canonical CDR structure types (ie, mouse canonical CDR structure types are And the second set of canonical CDR structures for the CDR sequences at corresponding positions in the non-human variable region and the human variable region, respectively, the first set of canonical CDR structures. By selecting a human sequence that is the same as the type, a subset of candidate sequences is selected. The method includes these candidate human variable region sequences, including at least two of the CDR sequences derived from non-human variable regions (eg, of a mouse CDR) together with a framework region derived from the candidate human variable region sequence. Used as a basis for constructing chimeric molecules. The result of this construction is that the chimeric antibody contains each non-human CDR sequence in place of each of the corresponding human CDR sequences in the variable region, so that the framework sequence in the chimeric antibody is a candidate human It differs from a framework sequence by 10 amino acid residues or less. In certain embodiments, the framework sequence of the chimeric antibody differs from the human framework sequence by no more than 5 amino acid residues. In other embodiments, the framework sequence of the chimeric antibody differs from the human framework sequence by no more than 2 amino acid residues. In most embodiments, the operation of constructing a chimeric antibody molecule comprises constructing a nucleic acid sequence that encodes the chimeric antibody sequence.

  In an exemplary embodiment, the method comprises a subset of candidate human sequences by comparing the similarity at each amino acid residue position of a non-human CDR sequence to the corresponding human CDR sequence according to a ranking criterion. The method further includes the step of ranking the members. In certain implementations, human sequence candidates include only the top 25% of the ranked members. In some embodiments, the ranking criteria are a non-human CDR sequence and a human CDR at corresponding residue positions of at least one CDR or at least two CDRs (or most typically each corresponding CDR). Contains a score of amino acid identity with the sequence. In other embodiments, the ranking criteria is a score of amino acid homology between a non-human CDR and a human CDR at a corresponding position in at least one, at least two, or each of the corresponding CDRs. Include. In still other embodiments, the ranking criteria include both an amino acid identity score and an amino acid homology score for at least one, at least two, or each of the corresponding CDRs. This method can be implemented using CDRs defined by different systems. For example, in certain embodiments, the CDR is a CDR defined by Kabat, and in other embodiments, the CDR is a CDR loop defined by Chothia.

  This method is not limited to strict use of the exact CDR sequences of non-human sources or the exact sequences of the human framework from member sets. In certain embodiments, the method can also include substituting at least one amino acid residue of the non-human CDR sequence with a different amino acid, provided that no more than 4 residues are non-human light chain CDR1. , Non-human light chain CDR2, non-human light chain CDR3, non-human heavy chain CDR1, or non-human heavy chain CDR3, and 10 amino acids or less are replaced in non-human heavy chain CDR2. In other embodiments, the method can also include substituting at least one amino acid residue of the human framework sequence but no more than 10 amino acid residues with a different amino acid residue.

  This method also recognizes that in certain circumstances, a non-human variable region can include a CDR sequence having a canonical form that is not present in the human variable region. If each of the three non-human CDRs is a light chain CDR, select a human sequence if one of the three non-human CDR sequences is a canonical structure type not present in the library of human variable region sequences The action includes selecting a human variable region sequence having a canonical CDR that is different from the non-existing non-human CDR type at the corresponding position, provided that the different canonical human CDR type is a non-human CDR type. Have a length that is less than or greater than two amino acid residues than the length of this non-existing canonical CDR structure type. Typically, if the non-existing CDR sequence is canonical type 1, this action includes selecting a human sequence having a canonical type 2 CDR at the corresponding position, or non-human. If the CDR sequence is canonical type 5, this operation involves selecting a human sequence having a canonical type 4 or 3 CDR at the corresponding position.

In most embodiments, the non-human variable region is a mouse variable region. Similarly, in most embodiments, a library of human variable region sequences is used as a human framework source as a human _Vk sequence, human _Vλ sequence, human _VH sequence, human _JH sequence, human _Jk sequence or Contains the human _Jλ sequence. In most embodiments, the method, the chimeric antibodies with both chimeric variable light chains and chimeric variable heavy chains, typically, comprising combining a human framework derived from the V _k sequence and V _H sequences . In an exemplary embodiment, the chimeric variable light chain and the chimeric variable heavy chain are formed to be Fab fragments, or (Fab) ₂ molecules, or single chain Fv molecules, or the chimeric variable light chain and The chimeric heavy chain is assembled with human antibody constant regions to form a complete antibody.

  This method is applicable to transforming a subject antibody sequence of any subject species into a less immunogenic form that is suitable for use in the object species, which is By generating a chimeric antibody comprising a species framework sequence in combination with a CDR from the principal species. In such cases, the methods described above perform the same operations, the variable region can be derived from any subject species, and the object variable region is derived from any object species for which the antibody is used. possible. Thus, for example, in various embodiments, the subject antibody is chimerized using a framework sequence from a bovine source, a porcine source, a mouse source, or a horse source, respectively, Ized antibodies, murine antibodies, or equine antibodies can be formed.

  In another aspect, the present invention provides a composition comprising a chimeric antibody molecule produced according to the disclosed method. Since this method uses a novel method of identifying appropriate object framework sequences for combination with the subject CDR sequences, the resulting chimeric antibody is also novel. Accordingly, provided herein is a humanized antibody comprising a chimeric antibody variable region comprising at least two non-human CDR sequences fused adjacent to a human variable region framework sequence. The human framework sequence is selected from a subset of framework sequences characterized by having 10 amino acid residues or less, wherein the 10 amino acid residues or less are between the variable region of the chimeric antibody and the variable region of the human antibody Different from framework sequences in human antibody variable regions having at least two human CDR sequences having the same canonical structure type as the non-human CDR sequences for at least two corresponding CDR positions between.

This non-human variable region CDR is typically derived from a mouse. This human variable region sequence is typically a V _k sequence, a V _λ sequence, a V _H sequence, a J _H sequence, a J _k sequence or a J _λ sequence. Most typically, a chimeric antibody comprises a chimeric antibody sequence for each of the variable light and variable heavy chains. In exemplary embodiments, the chimeric variable light chain and the chimeric variable heavy chain are formed into Fab fragments, or (Fab) ′ ₂ molecules, or single chain Fv molecules, or chimeric variable light chains and chimeric heavy chains. Are assembled with human antibody constant regions in the form of complete antibodies. Most typically, the human variable region sequence is a sequence derived from a human germline variable region fragment. In other embodiments, the human variable region sequence is a sequence derived from a human mature antibody.

In preferred embodiments, the humanized antibody has a dissociation constant for its antigen of at least 10 ⁶ M ⁻¹ , preferably at least 10 ⁷ M ⁻¹ , and more preferably at least 10 ⁸ M ⁻¹ . Typically, this humanized antibody does not elicit an immune response when administered to a human. Certain embodiments illustrating the invention bind to a humanized antibody that binds to a scorpion venom antigen, a humanized antibody that binds to the human CD28 receptor, a humanized antibody that binds to human lysozyme, or a human glutamate decarboxylase (GAD65). Humanized antibodies were included.

FIG. 1 shows a library of human germline _VH gene segments. FIG. 2 shows a library of human germline _Vk gene segments. FIG. 3 shows a part of mouse D1.3 (anti-chicken lysozyme) antibody variable light chain sequence and mouse DL. 3 shows a selected subset of human germline _Vk variable region sequences having the same type of canonical CDRs as the three light chain sequences. This subset is DL. 3 The highest ranked sequences are presented first, ranked by amino acid sequence similarity between the CDRs and the human CDRs. FIG. 4 shows part of the murine D1.3 antibody variable heavy chain sequence and its DL. 3 shows a selected subset of human germline _Vk variable region sequences with canonical CDRs of the same type as 3. FIG. This subset is ranked according to the similarity of the corresponding CDR amino acid sequences as in FIG. Figure 5 shows the amino acid sequence of V _H variable region of the chimeric V _k variable region amino acid sequence and humanized D1.3 antibodies, this illustrates one aspect of the present invention. FIG. 6 shows the nucleic acid sequence of a DNA construct that encodes (and expresses) the humanized chimeric D1.3 antibody of FIG. 5, which represents another aspect of the present invention. FIG. 6 shows the nucleic acid sequence of a DNA construct that encodes (and expresses) the humanized chimeric D1.3 antibody of FIG. 5, which represents another aspect of the present invention. FIG. 7 is a graph showing antigen binding by humanized D1.3 antibody. This has an affinity constant greater than 10 ⁸ M ⁻¹ and represents one embodiment of the present invention. FIG. 8 shows a portion of the mouse variable light chain sequence of an anti-human CD28 antibody named 9.3 and a human reproductive strain having the same type of human canonical CDR as the mouse 9.3 variable light chain sequence at the corresponding position. A selected subset of the series V _k variable region sequence is shown, which is ranked by amino acid sequence similarity as in FIG. FIG. 9 shows a selection of a portion of the murine variable heavy chain sequence of the 9.3 antibody and a human germline V _H variable region sequence having the same type of human canonical CDR as the murine variable heavy chain sequence at the corresponding position. This is also ranked by amino acid sequence similarity. FIG. 10 shows a humanized anti-human CD28 (Hu9.3) Fab fragment with a chimeric variable heavy chain and a variable light chain, which represents another embodiment of the present invention. FIG. 11 is a graph showing antigen binding by Hu9.3 fragment. This has an affinity constant greater than 10 ⁶ M ⁻¹ and represents one embodiment of the present invention. FIG. 12 shows a humanized anti-scorpion toxin Fab fragment having a chimeric variable heavy chain and a variable light chain, which represents another embodiment of the present invention. FIG. 13 shows a humanized anti-human glutamate decarboxylase (GAD65) Fab fragment with a chimeric variable heavy chain and a variable light chain, which represents another embodiment of the present invention. FIG. 14 is a graph showing antigen binding by humanized anti-GAD65 Fab fragment. This has an affinity constant greater than 10 ¹¹ M ⁻¹ and represents one embodiment of the present invention.

(Detailed description of the invention)
In the following description, reference is made to various references that can assist one of ordinary skill in the art in understanding and practicing the present invention to a full extent. Accordingly, each reference cited in the following description is incorporated herein by reference in its entirety. To better assist in understanding various embodiments of the present invention, it may be useful to explain the meaning of certain terms used herein.

  A “mature antibody gene” is a gene sequence encoding an immunoglobulin that has undergone a maturation process such that the particular immunoglobulin is expressed, eg, a lymphocyte (eg, in a hybridoma) , B cells) or expressed in antibody producing cells. The term encompasses mature genomic sequences, mature cDNA sequences, or other mature nucleic acid sequences that encode such mature genes, which are isolated and / or for expression in other cell types. Recombination has been performed. Mature antibody genes have undergone a variety of maturation and rearrangements that structurally distinguish between antibody genes encoded in all cells other than lymphocytes and these mature antibody genes. Mature antibody genes in humans, rodents, and many other mammals are obtained by fusion of V and J gene segments in the case of antibody light chains and V gene segments, D in the case of antibody heavy chains. It is formed by the fusion of a gene segment and a J gene segment. Many mature antibody genes require point mutations and subsequent fusions, some of which increase the affinity of the antibody protein for a particular antigen.

A “germline antibody gene” or germline antibody gene fragment is an immunoglobulin sequence encoded by a non-lymphocyte cell that has not undergone a maturation process that results in gene rearrangement and mutation for expression of a particular immunoglobulin. It is. One of the advantages provided by the various embodiments of the present invention is that germline antibody genes may conserve essential amino acid sequence structures that are more characteristic of individuals in animal species than mature antibody genes. High, and therefore, comes from the recognition that when used therapeutically in that species, it is unlikely to be recognized as coming from a heterogeneous source. FIGS. 1 and 2 show the peptide sequences of human germline antibody genes encoding human variable heavy chain region (V _H ) and variable light chain region (V _k ) antibodies (ie, immunoglobulins). Each of these sequence listings illustrates a library of human antibody genes, particularly a library of human germline antibody genes.

  “CDR” is the complementarity determining region within antibody variable sequences. For each variable region, there are three CDRs (designated as CDR1, CDR2 and CDR3) in each of the variable heavy and variable light chain sequences. The exact boundaries of these CDRs are defined differently by different systems, but all have residues that overlap within what constitutes a so-called “hypervariable region” within the variable sequence. The system described by Kabat (CITE) not only provides a clear residue numbering system applicable to any variable region of an antibody, but also provides the exact residue boundaries that define these CDRs. These CDRs can be referred to as Kabat CDRs. Chothia and co-workers (CITE) have found that certain sub-parts within the Kabat CDR adopt nearly identical peptide backbone configurations, despite the great diversity at the amino acid sequence level. These sub portions are named L1, L2 and L3 or H1, H2, and H3, where “L” and “H” indicate the light and heavy chain regions, respectively. These regions may be referred to as Chothia CDRs, which have boundaries that overlap with Kabat CDRs. Table I shows the overlap of Chotia and Kabat CDRs by the Kabat residue numbering system.

  (Table I)

  Other boundaries defining CDRs that overlap with Kabat CDRs are described by Padlan (1995) or MacCallum (1996). Still other CDR boundary definitions may not strictly follow one of the above systems, but nevertheless they overlap with Kabat CDRs, but they are defined as specific residues or residues It may be shortened or lengthened in view of the prediction or experimental finding that the group or even the entire CDR does not significantly affect antigen binding. Although the methods used herein may utilize CDRs defined according to any of these systems, preferred embodiments use CDRs defined by Kabat or CDRs defined by Chonthia.

  The “framework” or “framework sequence” is the remaining sequence obtained by subtracting the CDR from the variable region. Since the exact definition of a CDR sequence can be determined by different systems, the meaning of the framework sequence is affected by different interpretations accordingly. For the purposes of clarity herein, framework sequences refer to sequences within the variable region of an antibody other than those defined as being CDR sequences, so that the framework The exact sequence depends only on how the CDR is defined. For example, the CDRs used in the methods provided herein are usually a subset of what is considered a Kabat CDR, but, for example, in the case of heavy chain CDR1, framework residues in the Kabat system. Residues classified as are also included.

  The “canonical CDR structure type” is a structure type named by Chothia (CITE). Chothia and co-workers have found that the critical position of the CDRs of many antibodies adopts nearly identical peptide backbone configurations, despite great diversity at the amino acid sequence level. Thus Chothia defined one or a few “canonical structures” for each CDR in each chain. Each canonical structure primarily identifies a set of peptide backbone tangles for successive segments of amino acid residues that form a loop. The canonical CDR structure types defined by Chothia are listed in Table II.

  (Table II)

  “Corresponding CDR” refers relatively to the CDR between two different variable sequences that correspond in position within two different variable sequences. Thus, for example, mouse light chain CDR1 corresponds to human light chain CDR1, and vice versa. Because each is located at a position defined in the Kabat numbering system, whether the actual boundary of its CDRs is defined by Kabat, by Chothia, or by some other system. Because it does. Similarly, a “corresponding” residue, “corresponding” sequence, or “corresponding” amino acid refers to the relative residue position between two different peptide sequences positioned by the Kabat numbering system.

  The purpose of the method provided herein, which can be referred to as CDR grafting, is to provide a formulation for reaching the appropriate human framework sequences for humanizing the subject non-human antibody. In all previous CDR grafting methods, the selection of humanized framework sequences was based on comparing the human framework to the subject (mouse) framework. In contrast, the basis of the methods described herein is based on the similarity between the CDRs of the subject antibody and the CDRs of the subject antibody, without comparing framework sequences between the two antibodies. Selecting human antibodies to provide work.

  Similarity of a candidate human antibody sequence to the subject CDR is assigned for each domain at two levels. Primarily, the same three-dimensional structure of the CDR peptide backbone is searched. The experimentally determined atomic coordinates of the subject CDR are rarely available. Thus, the three-dimensional similarity is approximated by determining the Chothia canonical structure type of the subject CDR and excluding candidates carrying different canonical structures from further consideration. Second, the residue-by-residue homology between the subject CDR and the remaining human candidate CDRs is considered and the candidate with the highest homology is selected.

  Choosing the highest homology is based on various criteria used to rank candidate human variable regions that have the same canonical structure as the subject non-human variable region. The criteria for ranking a selected set of members can be by amino acid sequence identity or amino acid sequence homology or both. Amino acid identity is a simple score of match for each amino acid residue position. Similarity due to amino acid homology is similarity at each position in a characteristic residue structure. Homology is described, for example, in Henikoff and Henikoff (1992) Amino acid substituting matrices from protein blocks, Proc. Natl. Acad. Sci 89: 10915-10919 can be scored according to the tables and procedures described by or by the BLOSUM series described by Henikoff and Henikoff (1996).

The steps of these methods are as follows:
The peptide sequences of the heavy and light chain variable domains of the main antibody are determined. These can be done in several ways (eg, DNA sequencing of individual genes after conventional cDNA cloning; DNA sequencing of reverse transcripts of subject hybridomas or cloning products amplified from DNA by polymerase chain reaction; or purification. The peptide sequence of the antibody protein produced).

  The Kabat numbering system (Kabat et al., 1991) is applied to the heavy and light chain sequences of the subject non-human antibody.

  The canonical structure type is determined for each of the CDRs of the subject non-human antibody. This determination takes into account the guidelines discussed in Chothia and Lesk (1987), Chothia et al. (1992), Tomlinson et al. (1995), Martin and Thornton (1996), and Al-Lazikani et al. (1997). Made from the test. The salient features of canonical structure determination for each of the CDRs are as follows.

  For the heavy chain CDR1, three canonical structure types are currently known. Assigning new sequences is easy. This is because each canonical structure type has a different number of residues. When Kabat numbering is assigned to this sequence, as described in Al-Lazikani et al. (1997), the numbering for residues 31-35 is as follows for each canonical structure: .

Canonical structure type 1: 31, 32, 33, 34, 35
Canonical structure type 2: 31, 32, 33, 34, 35, 35a
Canonical structure type 3: 31, 32, 33, 34, 35, 35a, 35b.

For the heavy chain CDR2, four canonical structure types are currently known. Some have a unique number of residues and their unique Kabat numbering at positions 52-56, ie canonical structure types 1:52, 53, 54, 55, 56
Canonical structure type 4: 52: 52a, 52b, 52c, 53, 54, 55, 56
Easily distinguished from

  Canonical structure types 2 and 3 for heavy chain CDR2 have an equal number of residues and therefore must be distinguished by a clue in their sequence, as discussed by Chothia et al. (1992). Don't be. The Kabat numbering of the segment containing these cues is 52, 52a, 53, 54, 55. The canonical structure type 2 has Pro or Ser at the position 52a, Gly or Ser at the position 55, and there are no restrictions on other positions. The canonical structure type 3 has Gly, Ser, Asn, or Asp at the position 54, and there are no restrictions on other positions. These criteria are sufficient to resolve the correct assignment in most cases. Further, framework residue 71 is commonly Ala, Val, Leu, Ile or The for canonical structure type 2 and Arg for canonical structure type 3 in common.

  Heavy chain CDR3 is the most diverse of all CDRs. This is generated by a genetic process unique to lymphocytes, some of which are random in nature. As a result, the canonical structure of CDR3 was difficult to estimate. In either case, the human germline V gene segment does not encode any part of CDR3. This V gene segment terminates at Kabat position 94, while positions 95-102 encode CDR3. For these reasons, the canonical structure of CDR3 is not considered for selecting candidate human sequences.

  For the light chain CDR1, six canonical structure types are currently known for CDR1 in the kappa chain. Each canonical structure type has a different number of residues, so the assignment of the canonical structure type to the new sequence is apparent from Kabat numbering at residue positions 27-31.

Canonical structure type 1:27, 29, 30, 31
Canonical structure type 2: 27, 28, 29, 30, 31
Canonical structure type 3: 27, 27a, 27b, 27c, 27d, 27e, 27f, 28, 29, 30, 31
Canonical structure type 4: 27, 27a, 27b, 27c, 27d, 27e, 28, 29, 30, 31
Canonical structure type 5: 27, 27a, 27b, 27c, 27d, 28, 29, 30, 31
Canonical structure type 6:27, 27a, 28, 29, 30, 31.

  For light chain CDR2, only a single canonical structure type is known for CDR2 in kappa light chains. Thus, the assignment is automatic, except for exceptional subject antibody sequences.

For light chain CDR3, up to six canonical structure types have been described for CDR3 in the kappa chain, but three of these are rare. The three common types can be distinguished by their length as reflected in Kabat numbering at residue positions 91-97:
Canonical structure type 1:91, 92, 93, 94, 95, 96, 97 (also accompanied by an essential Pro at position 95 and Gln, Asn or His at position 90)
Canonical structure type 3: 91, 92, 93, 94, 95, 97
Canonical structure type 5: 91, 92, 93, 94, 95, 96, 96a, 97.

  After identifying the canonical CDR structure type of the subject non-human antibody, human genes of the same chain type (heavy chain or light chain) having the same canonical structure type combination as the subject antibody form a candidate set of human sequences Is identified as follows. In a preferred embodiment, only the peptide sequences of human germline immunoglobulin VH and Vk gene fragments are considered for comparison. Most of these gene fragments have been discovered and have already been assigned to canonical structure types (Chothia et al., 1992, Tomlinson et al., 1995). Additional V gene fragments not disclosed by these references appear to exist between the sequences provided herein and listed in FIGS. For the heavy chain, CDR1 and CDR2 matches to the mouse canonical structure type are evaluated, and mismatched genes are eliminated. For the light chain, the integrity of CDR1 and CDR2 of each human sequence to the canonical structure type of the main antibody is first evaluated. The possibility that residues 89-95 of the candidate Vk gene form CDR3 of the same canonical structure type as the main antibody is evaluated, this assessment assumes fusion of the gene with the J region, and CDR3 canonical By applying criteria for CDR structural typing to the fusion sequence. Inconsistent sequences are excluded.

  In another embodiment suitable when the variable domain of the subject antibody is a canonical structure type that is not available in the human genome, a human germline V having a canonical structure type that is three-dimensionally similar but not identical. Genes are considered for comparison. Such a situation often occurs with kappa chain CDR1 in mouse antibodies and includes the following two examples. All six possible canonical structure types are observed in this CDR in mouse antibodies, while the human genome encodes only canonical types 2, 3, 4, and 6. In these situations, canonical CDR structural types with amino acid residue lengths that fall in two of the amino acid residue lengths of the subject non-human sequence can be selected for comparison. For example, if a type 1 canonical structure is found in the subject antibody, a human Vk sequence with canonical structure type 2 should be used for comparison. If a type 5 canonical structure is found in a murine antibody, a human Vk sequence having either canonical structure type 3 or 4 should be used for comparison.

  In another embodiment, reconstructed mature human antibody sequences can be considered for sequence comparison. Such a comparison can be assured under various circumstances, where it is known that the mature human sequence is (1) very close to the germline; (2) not immunogenic in humans Or (3) includes, but is not limited to, a canonical structure type identical to that of the main antibody but not found in the human germline.

In a preferred embodiment, for each candidate V gene having a matching canonical structure type, residue-by-residue sequence identity and / or sequence homology with the subject sequence is also evaluated to rank the candidate human sequences. Is done. In certain embodiments, the evaluated residues are as follows:
Chain CDR residue position κ 1 26-32
κ 2 50-52
κ 3 91-96
Heavy 1 31-35
Heavy 2 50-60
In a preferred embodiment, residue-by-residue homology is first scored by the number of identical amino acid residues between the subject sequence and the candidate human sequence. The human sequence used for the subsequent construction of the converted antibody is selected from 25% of the candidates with the highest score. In other embodiments suitable where several candidate sequences have similar identity scores, similarity between non-identical amino acid residues can be further considered. Aliphatic and aliphatic, aromatic and aliphatic, or polar and polar fits between the main and object residues are added to the score. In another embodiment, quantitative assessment of sequence homology can be performed using amino acid substitution matrices such as the BLOSUM62 matrix of Henikoff and Henikoff (1992).

  The object sequence for the framework region on the C-terminal side of the CDR3 sequence is selected from a set of known human germline J segments. Preferred J peptide sequences are the homology for each J segment residue for sequence positions where CDR3 and J overlap using the scoring criteria specified for evaluation of candidate V genes as described above. Is selected by evaluating. The J gene segment peptide sequence used for subsequent construction of the converted antibody is selected from 25% of the candidates with the highest scores.

  In one embodiment, the chimeric variable chain comprises at least two CDRs derived from the subject non-human sequence and a framework sequence derived from the candidate human sequence. In other embodiments, the chimeric light chain comprises three CDRs from the subject non-human sequence and a framework sequence from the candidate human sequence. In other embodiments, the chimeric heavy chain comprises at least two CDRs of the main body weight chain and a framework sequence of a candidate human heavy chain. In another embodiment, the chimeric heavy chain comprises each CDR from the main body weight chain and the framework sequence of the candidate human heavy chain. In yet another embodiment, the chimeric antibody heavy chain comprises CDR1 and CDR2 from the subject non-human sequence, residues 50-60 for CDR3 from the candidate human heavy chain and residues 61-65 for CDR4, Includes with candidate human sequence framework sequences. In another embodiment, the chimeric heavy chain sequence comprises each CDR derived from the subject non-human sequence, framework sequences 27-30 derived from the subject sequence, and framework sequences derived from the candidate sequence. However, in all cases, the chimeric antibody molecule comprises no more than 10 amino acid residues in the framework sequence that differ from the amino acid residues in the framework sequence of the candidate human variable region.

  In another embodiment suitable where increased affinity of the humanized antibody is desired, residues in the CDRs of the converted antibody may be further substituted with other amino acids. Typically, 4 amino acid residues or less in a CDR are changed, and most typically 2 residues or less in a CDR are changed (except for heavy chain CDR2), but as many as 10 The residues of can be changed. Similarly, in certain embodiments, some of the amino acids in the framework sequence can be changed. In all embodiments, no more than 10 amino acid residues are changed.

The humanized antibody sequence is then physically assembled by gene synthesis methods and recombinant protein expression methods known by those skilled in the art. The final form of the humanized sequence with the chimeric variable chain produced by the methods disclosed herein can take many forms. Most typically, chimeric antibodies are produced by the construction of a nucleic acid sequence encoding a chimeric variable chain that is recombinantly expressed in an appropriate cell type. One of the most typical forms of the chimeric antibody is a Fab antibody fragment. Other suitable forms of chimeric antibodies include (Fab) ′ ₂ molecules, or single chain Fv molecules. Still other forms may include additional fusions to the constant domains of human antibodies that form complete antibodies. In preferred embodiments, both the variable light chain and the variable heavy chain are humanized. However, in other embodiments, the variable light chain and variable heavy chain can be mixed (ie, one is a fully murine variable chain (heavy or light chain) and the other is a humanized variable chain). ).

In most embodiments, the method includes screening candidate chimeric antibodies to select antibodies having a certain dissociation constant for the antigen appropriate for the intended use. In most embodiments, humanized antibodies produced according to these methods have a dissociation constant of at least 10 ⁶ M ⁻¹ , at least 10 ⁷ M ⁻¹ , or at least 10 ⁸ M ⁻¹ . A dissociation constant of at least about 10 ⁸ M ⁻¹ is preferred for most therapeutic applications.

  The following examples illustrate the invention by showing specific embodiments for humanized antibodies that bind to various types of antigens for illustrative purposes. Those skilled in the art will appreciate that many other specific embodiments can be generated using the methods disclosed herein, and that the present invention is not limited by the specific embodiments. .

Example 1
(Humanized anti-chicken lysozyme)
Mouse antibody D1.3 binds to chicken lysozyme antigen. The peptide sequence of the variable domain of D1.3 was obtained from Protein Data Bank accession number 1 VFA. The light chains were numbered according to Kabat. Mouse CDRs were assigned canonical structure types as follows:
The light chain CDR1, numbered according to Kabat, has the sequence:

Consists of. CDR1 has canonical structure type 2 since there are no insertions or deletions between residues 27 and 31.

  The light chain CDR2, numbered according to Kabat, has the sequence:

Consists of. This is not an exceptional array and its canonical structure type is type 1.

  The light chain CDR3, numbered according to Kabat, has the sequence:

Consists of. Due to the length and Pro at position 95, this sequence is consistent with canonical structure type 1.

  In the compilation in FIG. 2 and Tomlinson et al. (1995), 21 non-overlapping G germline Vk genes are (1) CDR1 with canonical structure type 2, (2) CDR2 with canonical structure type 1, and ( 3) encodes a sequence having the ability to form canonical structure type 1 in CDR3. These are listed under the D1.3Vk sequence in FIG. Their sequences at residue positions including Chothia canonical structure types are also shown, stratifying the human Vk gene in FIG. 3 according to the number of identities per residue in these sequences. L23 has seven identities, while the next three items on this list have six. In addition, L23 has conserved residue positions 91 and 92 in CDR3, which is superior to the next three candidates. Therefore, L23 is selected for the humanized construct.

  None of the human Jk segments in FIG. 3 match Arg in D1.3 at position 96, and all are identical at the next three positions. Jk4 overlapping the GGG motif at positions 99 to 101 of D1.3 is the best fit for the J segment and is used for humanized construction.

  The heavy chain variable domain of D1.3 is numbered according to Kabat and is also shown in FIG. CDRs were assigned canonical structure types as follows.

  The sequence in the region of heavy chain CDR1 is

It is. This sequence lacks inserted residues and is therefore assigned to canonical structure type 1.

  Kabat CDR2 of D1.3 is the sequence

Have Since there is no insertion between residues 52 and 56, canonical structure type 1 is assigned to CDR2. The human germline _VH gene was assumed to have canonical structure type 1 in CDR1. Canonical structure type 1 in CDR2 was obtained from Chothia et al. (1992) and FIG. This is listed in FIG.

  Segments selected for homology evaluation are 27-35 (corresponding to additional residues forming Kabat CDR1 + Cothia canonical structure) and 50-60 (corresponding to Kabat CDR2 lacking residues 61-65) there were. They rarely participate directly in antigen binding. The first two items have 8 identities in these segments when compared to the mouse sequence, and the next 5 have 7 identities. Therefore, the top 25% of items in the similarity ranking are any of two genes having eight identities and those having seven. Any of these seven genes are suitable candidates for humanized construction, but some are preferred. It is due to conservation at non-identical residues. Three with Glu or Arg replacing Met at residue 50 are eliminated. This is because an embedded three-dimensional structure may be obtained by burying the charged side chain in the center of the hydrophobic segment. Therefore, V71-4 was selected for the remaining four.

  JH4 is clearly the best fit for the C-terminus of CDR3.

  A chimeric humanized antibody was designed by combining the Kabat CDR of D1.3 with the Kabat framework encoded by V71-4, JH4, L23 and Jk4. The sequence of the heavy and light chain variable domains of this antibody is shown in FIG.

  Synthetic variable domain genes encoding humanized Vk and VH were prepared from synthetic oligonucleotides by the method of Ye et al. (1992) (incorporated herein by reference). These genes were then transferred to the Fab expression vector pAK19 described by Carter et al. (1992) (incorporated herein by reference). The DNA sequence of the synthetic gene and the DNA sequence of the Fab expression cassette of pAK19 are shown in FIG. Recombinant Fab was obtained from E. coli. expressed in E. coli, released from the periplasm by osmotic shock, and purified by chromatography in lysozyme-Sepharose.

  The affinity of SHuD1.3 for lysozyme was determined by the fluorescence quench method described by Foote and Winter (1992). This method relies on changes in the intrinsic tryptophan fluorescence of antibodies and antigens upon complex formation. In the experiment of FIG. 7, 200 nM humanized D1.3 Fab was titrated with a small amount of concentrated lysozyme solution. Fluorescence data was fit to the titration equation by the least squares method to give a value for the dissociation constant and a standard error of 23 ± 5 nM. By comparison, the Kd of D1.3 IgG is known to have a Kd of 4 nM (Foote and Winter, 1992). Therefore, the humanized antibody in Example 1 has the same antigen specificity as the main mouse antibody, and binds to the antigen with an affinity reduced to less than 1/6 compared to the main antibody.

(Example 2)
(Humanized anti-human CD28)
A mouse anti-human CD28 antibody named 9.3 was used as the non-human subject antibody. A mouse 9.3 hybridoma strain was isolated. This is described by Hansen et al. (1980).

  The 9.3 heavy and light chain variable regions were cloned by reverse transcription and polymerase chain reaction. Started with messenger RNA, isolated by the guanidinium isothiocyanate procedure (Chomczynski and Sacchi, 1987) followed by chromatography on an oligo-dT column. Amplification was primed using an oligonucleotide complementary to the constant region and an oligonucleotide corresponding to the signal peptide region or region of the N-terminal framework sequence.

  The light chains were numbered according to Kabat and the CDRs were assigned canonical structure types as follows with reference to FIG.

  The light chain CDR1 numbered by Kabat has the sequence

Consists of. Due to the insertion residue between 27 and 31, CDR1 has canonical structure type 5.

  The light chain CDR2, numbered by Kabat, has the sequence

Consists of. This is not an exceptional array. Its canonical structure type is 1.

  The light chain CDR3 numbered by Kabat has the sequence

Consists of. Due to its length and Pro at position 95, this sequence matches canonical structure type 1.

  The Vk sequence with canonical structure type 5 in CDR1 is not shown in the human germline, but structures 3 and 4 are similar to canonical structure type 5. These were further considered.

  In the compilation in FIG. 2, eight non-overlapping human germline Vk genes are represented in CDR1, having canonical structure type 3 or 4, (2) CDR2 having canonical structure type 1, and (3) canonical structure in CDR3. It encodes a sequence having the ability to form type 1. These are listed under the 9.3 Vk sequence in FIG. Their sequences in the Kabat CDR are also shown. The human Vk genes in FIG. 3 are ranked according to the number of identities per residue at the residue positions that form the Chothia canonical structure. The B3 gene has seven identities at these positions, while the next three on this list have five. Therefore, B3 was selected for humanization construction. Since this scoring was based on Kabat CDR positions rather than Chothia, B3 is still the first candidate. The Tyr residue encoded in 5 'of human Jk2 matched exactly to the corresponding position of 9.3. This germline fragment was therefore used.

  The 9.3 heavy chain variable domains were numbered by Kabat. This is also shown in FIG. CDRs were assigned canonical structure types as follows.

  The sequence in the region of heavy chain CDR1 is

It is. This sequence lacks any inserted residues. Therefore, canonical structure type 1 is assigned.

  The 9.3 Kabat CDR2 has the sequence

Have Since there is no insertion between residues 52 and 56, canonical structure type 1 is assigned to CDR2.

  The human germline VH gene was assumed to have canonical structure type 1 in CDR1. Canonical structure type 1 in CDR2 was obtained from Chothia et al. (1992) and FIG. This is listed in FIG.

  Segments selected for homology evaluation are 27-35 (corresponding to additional residues forming Kabat CDR1 + Cothia canonical structure) and 50-60 (corresponding to Kabat CDR2 lacking residues 61-65) there were. They rarely participate directly in antigen binding. The sequence is scored for the number of identical residues when compared to 9.3 and ranked by the score in FIG. Gene DP-45 has the highest number of identities 10 in a residue-by-residue comparison with 9.3, and the next six items have all nine. DP-45 was selected for humanization construction.

  Of the human JH segments, JH4 had the closest homology to the CDR C-terminus in 9.3 and was used in this construction.

  A chimeric humanized antibody variable domain was designed by combining the sequences as follows. The light chain variable domain is the Kabat CDR sequence of the 9.3 light chain (excluding residue 34, which was not considered important for antigen recognition, and thus became the same Ala as the residue in B3 at that position) , Up to residue 88, identical to B3 and identical to Jk2 at positions 98-108 (except for residues 70 and 72. These combine with residue 71 to form the glycosylation motif that these residues form. It was kept the same as 9.3 to preserve). The heavy chain variable domain is the 9.3 heavy chain Kabat CDR sequence (excluding residues 60-65. These were thought to be unimportant for antigen recognition, so they were identical to the DP-45 sequence at those positions. And Kabat framework sequence that is identical to DP-45 up to residue 94 and identical to JH4 at residues 103-113.

  The sequence of the heavy chain variable domain and the light chain variable domain of this antibody is shown in FIG. Recombinant Fab fragments containing the variable domains of these sequences were prepared as described in Example 1, except that affinity chromatography on protein-G Sepharose was used for purification. As a control, a Fab fragment composed of a mouse variable domain and a human constant domain was prepared by the same method. Similarly, a hybrid Fab fragment composed of a human constant domain, a mouse 9.3 heavy chain variable domain, and a humanized light chain variable domain was also prepared.

  The ability of these three Fabs to bind to CD28 was tested by ELISA. CD28Ig coated plates were incubated with Fab solutions at concentrations ranging from 1 pM to 10 mM. Binding was then assayed using anti-human kappa immunoconjugate. The generated binding isotherm was processed to obtain an equivalent concentration for the maximum half-value binding (EC50) of that antibody to CD28Ig as described in Jin et al. (1992) (incorporated herein by reference). It was determined. This analysis, shown in FIG. 11, showed that the murine Fab has an EC50 of 20 nM, the EC50 of Hu9.3 is 630 nM, and the EC50 of the hybrid Fab is 30 nM. The similarity of this hybrid Fab and mouse Fab avidity suggests that the reduced binding by humanization 9.3 can be attributed to interactions involving weakened heavy chains, and thus light chain-only humanization is minimal. Caused a loss of avidity.

(Example 3)
(Humanized anti-scorpion toxin)
A mouse anti-scorpion toxin antibody named BCF2 was used as the non-human main antibody of the humanized anti-scorpion toxin. A mouse BCF2 hybridoma line has been described, and the efficacy of BCF2 in a mouse model was demonstrated by Licea et al. (1996). The sequence of the variable domain of BCF2 was disclosed by Selisako et al. (1999). This is shown in FIG.

  The canonical structure type of the light chain was determined as previously described. They were type 5 for CDR1, type 1 for CDR2, and type 1 for CDR3. The canonical structure type of the heavy chain CDR is type 1 for CDR1 and type 2 for CDR2. A humanized form of BCF2 was designed using the considerations discussed above for the selection of human germline V and J gene sequences.

  The light chain variable domain consisted of the Kabat CDR sequence of the BCF2 light chain and a framework sequence that is identical to human gene A2 / DPK12 up to residue 88 and identical to Jk4 at positions 98-108. The heavy chain variable domain is the BCF2 heavy chain Kabat CDR sequence (excluding residues 62-65. These were considered not important for antigen recognition. Therefore, they are identical to the 1-f / DP3 sequence at those positions. And Kabat framework residues which are identical to 1-f / DP3 up to residue 94 and are identical to JH6 at residues 103-113.

The sequence of the heavy chain variable domain and light chain variable domain of the humanized BCF2 antibody is shown in FIG. Recombinant Fab fragments containing variable domains with these sequences were prepared as described in Example 2. As a control, the (Fab) ′ ₂ fragment was prepared by pepsin digestion of mouse BCF2 IgG obtained from hybridoma cells.

The ability of the two Fabs to bind to CD28 was tested using toxins immobilized on the sensor chip surface and antibodies in the supernatant using a BIAcore biosensor instrument. This method is described by Joensson et al. (1991), incorporated herein by reference. Subsequently, Fab solutions with concentrations varying over at least a 10-fold range were passed through the chip to observe the dissociation phase. The sensorgram was continued with buffer alone in the liquid phase to observe dissociation. The affinity as the dissociation equilibrium constant Kd was determined from the ratio of the kinetic rate constants kon / koff. Individual affinities were 10 nM for mouse (Fab) ′ ₂ and 140 nM for the humanized form.

Example 4
(Humanized anti-human GAD65)
NGAD65, an antibody against the human glutamate decarboxylase 65 kDa isoform.

  The sequence of the mouse NGAD65 hybridoma strain and its antibody variable region was described by Hampe et al. (2001). Their sequence is shown in FIG. The first two residues of the light chain are removed. Because those residues were derived from the oligonucleotides used for cloning.

  The canonical structure type of the light chain CDRs was determined to be type 4 for CDR1, type 1 for CDR2, and type 1 for CDR3. The canonical structure type of the heavy chain CDRs was determined to be type 1 for CDR1 and type 2 for CDR2.

  A humanized form of NGAD65 was designed using the considerations discussed above for the selection of human germline V and J gene sequences. The light chain variable domain consisted of a Kabat CDR sequence of the NGAD65 light chain and a framework sequence identical to the human Vk gene A17 / DPK18 up to residue 88 and identical to Jk3 at positions 98-108. The heavy chain variable domain is identical to the 1-v sequence at those positions since the Kabat CDR sequence of BCF2 heavy chain (excluding residues 61-65. This was not considered important for antigen recognition. ) And a Kabat framework sequence that is identical to 1-f / DP3 up to residue 94 and identical to JH4 at residues 103-113.

  The heavy chain variable domain and light chain variable domain sequences of the humanized NGAD65 antibody are shown in FIG. Recombinant Fab fragments containing these sequences were prepared as described in Example 2. As a control, a Fab fragment composed of a variable domain and a constant domain of mouse NGAD65 was prepared by the same method.

  The ability of these two Fabs to bind antigen was tested by immunoprecipitation assay. Radioactive human glutamate decarboxylase was prepared by in vitro translation using 35S-methionine. The labeled antigen was incubated overnight with either of these two Fab fragments at various concentrations. Protein G-Sepharose beads were then added to sequester the Fab and any bound antigen. Radioactivity was determined by scintillation counting and EC50 was determined visually from the midpoint of the plot of bound radioactivity versus concentration of Fab fragment. EC50 values (0.36 pM for mouse Fab and 9 pM for humanized Fab) were obtained. Even considering the loss of human antibody affinity to mouse antibody by a factor of 25, this humanized antibody still binds antigen sufficiently to be used in humans in therapy and has an affinity of 25 There is no need to mutate this sequence further to compensate for the loss to a fraction.

  The methods provided herein use a mouse mature antibody gene as a source for a first Chothia canonical CDR and a human antibody gene as a source for a second Chothia canonical CDR. Illustrated by. These examples are particularly suitable for the construction of humanized antibodies for use in human therapeutic applications. Such humanized antibodies not only contain sufficient mouse amino acid sequences to retain the three-dimensional structure necessary for strong antigen binding, but are sufficient human antibodies to prevent undesirable immunogenicity in humans Also includes sequences. However, one of ordinary skill in the art will recognize that the methods disclosed herein are equally applicable for preparing transformed antibodies comprising chimeric hypervariable regions from any two different vertebrate species. To do.

  In a more general sense, a first antibody sequence, which is originally selected by its binding to an antigen, can be referred to as a “subject” antibody sequence. Typically, the subject antibody sequence is of mouse or rat origin. The second antibody sequence (which is selected from the antibody sequences of the target animal) is referred to as the “object” antibody sequence. This antibody sequence of interest is typically derived from a human or domestic animal that is the subject of the therapeutic treatment. An antibody composition comprising a chimeric hypervariable region according to the methods of the present invention results in a third antibody sequence that may be commonly designated as “converted” antibody sequence. This transformed antibody sequence differs in specific defined structural features from each of the subject antibody sequence and the object antibody sequence, and other specific defined features from each of the subject antibody sequence and the object antibody sequence. Identical in structural features.

  (References)

Claims (46)

A method for producing a humanized antibody comprising:
Obtaining a peptide sequence of the main variable region encoded by the non-human mature antibody gene;
Identifying a first set of canonical CDR structural types for at least two CDRs within the non-human antibody variable region;
Obtaining a peptide sequence library of human antibody variable regions for a human antibody selected from the group consisting of a human antibody segment encoded by a human germline gene and a mature antibody;
Identifying a second set of canonical CDR structural types for at least two CDRs for each peptide sequence in the library of human variable region sequences;
Comparing the first set of canonical CDR structure types to the second set of canonical CDR structure types, and the second set of canonical CDR structures in the non-human variable region and in the human variable region, respectively. Selecting a subset of member peptide sequences from the library by selecting human peptide sequences that are the same as the first set of canonical CDR structural types for CDR sequences at corresponding positions; and non-human variable Constructing a chimeric molecule comprising at least two of the region-derived CDR sequences and a framework region from at least one member of a subset of the selected human variable region peptide sequences, comprising: The chimeric antibody comprises each of the non-human CDR sequences in place of at least two of the human CDR sequences at corresponding positions in the variable region. The framework sequence of the chimeric antibody differs from the human framework sequence of the selected member peptide sequence by no more than 10 amino acid residues;
Including the method.   2. The method of claim 1, wherein the framework sequence of the chimeric antibody differs from the human framework sequence of the selected member peptide sequence by no more than 5 amino acid residues.   2. The method of claim 1, wherein the framework sequence of the chimeric antibody differs from the human framework sequence of the selected member peptide sequence by no more than 2 amino acid residues.   2. The method of claim 1, wherein the selected subsets by comparing the similarity of the amino acid residues of the non-human CDR sequence by position to the corresponding human CDR sequence according to a ranking criterion. The method further comprises ranking the members.   The method of claim 4, wherein the selected subset includes only the top 25% of the ranked members.   6. The method of claim 5, wherein the ranking criteria comprises a score of amino acid identity between the non-human CDR sequence and the human CDR sequence at corresponding residue positions of at least one CDR. Method.   6. The method of claim 5, wherein the ranking criteria comprises a score of amino acid identity between the non-human CDR sequence and the human CDR sequence at corresponding residue positions of at least two CDRs. Method.   6. The method of claim 5, wherein the ranking criteria includes a score of amino acid identity between the non-human CDR sequence and the human CDR sequence at the corresponding residue position of each CDR.   7. The method of claim 6, wherein the ranking criteria further comprises an amino acid homology score between the non-human CDR and the human CDR at a corresponding residue position of at least one CDR. .   7. The method of claim 6, wherein the ranking criteria further comprises an amino acid homology score between the non-human CDR and the human CDR at corresponding residue positions of at least two CDRs. .   8. The method of claim 7, wherein the ranking criteria further comprises a score of amino acid homology between the non-human CDR and the human CDR at the corresponding residue position of each CDR.   2. The method of claim 1, wherein constructing the chimeric antibody sequence comprises constructing a nucleic acid sequence encoding the chimeric antibody sequence.   The method of claim 1, wherein the CDR is a CDR defined by Kabat.   2. The method of claim 1, wherein the CDR is a CDR loop defined by Chothia.   2. The method of claim 1, wherein the operation of constructing the chimeric antibody sequence further comprises substituting at least one amino acid residue of the non-human CDR sequence with a different amino acid, provided that 4 residues The following are substituted in any of the non-human light chain CDR1, non-human light chain CDR2, non-human light chain CDR3, non-human heavy chain CDR1, or non-human heavy chain CDR3, and 10 amino acids or less are non-human heavy chain CDR2 Replaced by in a method.   The method according to claim 1, wherein the operation of constructing the chimeric antibody sequence comprises replacing at least one amino acid residue of the human framework sequence but not more than 10 amino acid residues with a different amino acid residue. The method further comprising the step of:   2. The method of claim 1, wherein each of the three non-human CDRs is a light chain CDR and one of the three non-human CDR sequences is not present in the library of human variable region sequences. If it is quasistructured, the selecting action further comprises selecting a human variable region sequence having a canonical structure type CDR that is different from the non-existing non-human CDR type at the corresponding position. Wherein the different human canonical structure type has a length that is less than or greater than 2 amino acid residues by the length of the non-canonical non-canonical structure type.   18. The method of claim 17, wherein if the non-existent CDR sequence is canonical type 1, the action comprises selecting a human sequence having a canonical type 2 CDR at the corresponding position. Or when the non-human CDR sequence is canonical type 5, the operation comprises selecting a human sequence having canonical type 4 or canonical type 3 CDRs at the corresponding position. The method of inclusion.   2. The method of claim 1, wherein the non-human variable region is a mouse variable region. The method of claim 1, said library of human variable region sequence, V _k sequence, V _lambda sequences, V _H sequence, J _H sequences, the group consisting of a library of J _k sequence and J _lambda sequences More selected, the method.   2. The method of claim 1, wherein the operation of constructing the chimeric antibody comprises constructing a chimeric antibody sequence for each of the variable light and variable heavy chains. 24. The method of claim 21, wherein the chimeric antibody sequence comprises a framework derived from a human _Vk sequence and a framework derived from a human _VH sequence. 23. The method of claim 22, wherein the chimeric variable light chain and the chimeric heavy chain form a molecule selected from the group consisting of a Fab fragment, a (Fab) ' ₂ molecule, and a single chain Fv molecule. Aggregated, method.   24. The method of claim 22, wherein the chimeric variable light chain and the chimeric heavy chain are further assembled with a human antibody constant region domain to form a complete antibody.   2. The method of claim 1, wherein the human variable region sequence is a germline variable region fragment sequence.   2. The method of claim 1, wherein the human variable region sequence is a sequence derived from a mature human antibody. 2. The method of claim 1, wherein the dissociation constant of the humanized antibody for its antigen is determined and a dissociation constant of at least 10 ^{< 6} > M ^<-1> , at least 10 ^{< 7} > M ^<-1> , or at least 10 < ⁸ > M ^<-1 >. A method further comprising selecting an antibody having A method for producing a converted antibody comprising:
Obtaining a peptide sequence of a subject variable region encoded by a subject species mature antibody gene;
Identifying a first set of canonical CDR structure types for at least two CDRs in the subject variable region;
Obtaining a peptide sequence library of an object antibody variable region for an object species antibody selected from the group consisting of antibodies encoded by germline antibody genes and mature antibody genes;
Identifying a second set of canonical CDR structural types for at least two CDRs for each peptide sequence in the object variable region sequence library;
Comparing the first set of canonical CDR structure types to the second set of canonical CDR structure types, and corresponding positions in the subject variable region and the object variable region, respectively, of the second set of CDR structures. Selecting a subset of member peptide sequences from the library by selecting object peptide sequences that are the same as the first set of canonical CDR structure types for the CDR sequences in
Constructing a chimeric molecule comprising at least two of the CDR sequences from the subject variable region and a framework region from at least one member of a subset of the selected object variable region peptide sequences comprising: As a result, the chimeric antibody comprises each of the subject CDR sequences in place of each of the subject CDR sequences at corresponding positions in the variable region, wherein the framework sequence of the chimeric antibody comprises the selected member peptide The sequence differs from the object framework sequence by no more than 10 amino acid residues;
Including the method. A humanized antibody,
A chimeric antibody variable region comprising at least two non-human CDR sequences fused adjacent to a human variable framework sequence, the human framework sequence comprising: a variable region of the chimeric antibody variable region; and a variable region of the human antibody. Having no more than 10 amino acid residues from framework sequences in human antibody variable regions having at least two human CDR sequences having the same canonical structure type as the non-human CDR sequences at least two corresponding CDR positions between A humanized antibody selected from a subset of framework sequences characterized by:   30. The humanized antibody of claim 29, wherein the non-human variable region is a mouse variable region. A humanized antibody of claim 29, wherein the human variable region sequence, V _k sequence, V _lambda sequences, V _H sequence, J _H sequences is selected from the group consisting of J _k sequence and J _lambda sequences , Humanized antibodies.   30. The humanized antibody of claim 29, wherein the chimeric antibody comprises a chimeric antibody sequence for each of the variable light and variable heavy chains. 30. The humanized antibody of claim 29, wherein the chimeric variable light chain and the chimeric heavy chain form a molecule selected from the group consisting of a Fab fragment, a (Fab) ' ₂ molecule, and a single chain Fv molecule. Humanized antibodies assembled together.   30. The humanized antibody of claim 29, wherein the chimeric molecule further comprises a human antibody constant region domain so as to form a complete antibody.   30. The humanized antibody of claim 29, wherein the human variable region sequence is a sequence derived from a human germline variable region fragment.   30. The humanized antibody of claim 29, wherein the human variable region sequence is a sequence derived from a human mature antibody. 30. The humanized antibody of claim 29, wherein the humanized antibody is at least 10 ^{< 6} > M ^<-1> , at least 10 ^{< 7} > M ^<-1> , or at least 10 ^{< 8} > M ^<-1 > against the antigen of the humanized antibody. A humanized antibody having a dissociation constant of   30. The humanized antibody of claim 29, wherein the humanized antibody does not elicit an immune response when administered to a human.   30. The humanized antibody of claim 29, wherein the humanized antibody binds to a scorpion venom antigen.   40. The humanized antibody of claim 39, wherein the humanized antibody consists of SEQ ID NO :.   41. The humanized antibody of claim 40, wherein the humanized antibody binds to a lysozyme antigen.   42. The humanized antibody of claim 41, wherein the humanized antibody is composed of SEQ ID NO :.   30. The humanized antibody of claim 29, wherein the humanized antibody binds to human CD28 antigen.   44. The humanized antibody of claim 43, wherein the humanized antibody consists of SEQ ID NO :.   30. The humanized antibody of claim 29, wherein the humanized antibody binds to a human glutamate decarboxylase antigen.   46. The humanized antibody of claim 45, wherein the humanized antibody consists of SEQ ID NO :.

Images