Useful Links:  IMGT (International ImMunoGeneTics Information system; useful information on germline gene repertoire)

Because mice are convenient for immunization and recognize most human anitgens as foreign, mAbs against human targets with therapeutic potential have typically been of murine origin. However, non-human mAbs contain substantial stretches of amino acid sequences that will be immunogenic when injected into a human patient. It is well known that after injection of a foreign antibody, such as a murine antibody, a patient can have a strong human anti-mouse antibody response that essentially eliminates the antibdy’s therapeutic utility after the initial treatment as well as the utility of any other subsequently administered murine antibody. The human anti-mouse antibody (HAMA) responselimits the adminsitration of murine antibodies. Humanization techniques are well known for producing mAbs which exhibit reduced immunogenicity in humans while retaining the binding affinity of the original non-human parental mAb.

Jone (Nature 321: 522-525, 1986) shown that CDRs from the HV region of mouse antibody B1-8 could replace the corresponding CDRs of a human myeloma protein so as to result in a new antibody with affinity of the B1-8 antibody. 

In addiiton to the immunogenicity of rodent mAbs when administered to humans, further limitations arise form weak recruitment of effector funciton and rapid clearance from serum. However, simple grafting of CDR sequences often yields humanized antibodies that bind antigen much more weakly than the parent murine mAb, and decreases in affinity of up to several hundrefold have been reported. As a means of simplifying antibody humanization, a number of different approaches have been developed. A common method is to select the human framework most homologous in sequence to that of the murine antibody of interest. In this way, the number of mismatches between the humanized and parent murine fraemwork is minimized, and it becomes less likely that a key framework residue will need to be replaced. However, even a single incorrect framework residue can have a dramatic effect on antigen binding. (Wells “Antibody humanization using monovalent phage display” 212(16), 1996). 

In some cases, transplanting hypervariable loops form rodent antibodies into human frameworks is sufficient to trasnfer high antigen binding affinity, whereas in other cases it has been necessary to also replace one or several fraemwork region residues. For a given antibody, a small number of FR resideues are anticipated to be important for antigen binding. Firs,t there are a few FR residues that directly contact antigen in crystal structures of anitobdy-antigen complexes. Second, a number of FR reisdues have been proposed as critically affecting the conformaiton of particular CDRs and thus their contribution to antigen bidning. (Shepard, “Humanization of an anti-p185HER2 antibody for human cancer therapy” Proc. Natl. Acad. Sci. USA, 89, pp. 4285-4289, May 1992). 

Once the non-human antibody regions targeted for grafting onto the human FR have been defined, the second step in a humanizaiton process is to identify the source of human FR domors. Regardless of the method chosesn to select the human FRs, two sources of human sequences have been uilized; mature and germline gene sequences. Mature sequences, which are products of immuen resposnes, carry somatic mutations generated by random processes and are not under the species selection, resulting in potential immunoglenic residues. Thus, to avoid immunogenic residues, human germline genes have increasinly been used. The physical map of the human H and L chains loci and the functional gemrline gene repertoire they encode has also been throughly characterized. A second potential advantage of using germline genes as FR donors is that comparison of germline gene and mature antiboy x-ray cyrstallography stuctures in bound and free ligand stages have shown that the fomrer molecuels are more flexible. A third step in a typical CDR grafting protocl is to define mutations that will resotre or prevent affinity losses. A substantial body of evidene indicates that the number of back mutations and the type of amino acid resplacements to introduce into the humanized antibody in order to restore binding depend on each particular case. Thus, back mutations should be carefully designed based on the sturcture or a model of the humanized antibody and tested experimentally. (Fransson “Humanization of antibodies” Frontiers in Bioscience 13, 1619-1633, 2008)

Definitions

Humanized antibodies refers to an antibody having at least one chain comprising variable region framework residues substantially from a human antibody chain (referred to as the acceptor immunoglobulin or antibody) and at least one complementrity determining region substantially form a mouse antibody (referred to as the donor immunoglobulin or antibody).

Chimeric antibodies should be distinguished from humanized antibodies in that chimeric antibodies combine murine variable regions with the human constant region (i.e., the framework regions are also murine).

De-immunization: identifies potential linear T cell epitopes present in the antibody (or in any protein) using bioinformatics. If these are detected, the sequence can be altered. In addition to diminished IgG2 and IgG4 subclass effector function, engineered human IgG with debilitated effector functions are avilable. It is known, for example, that absence of the carbohydrate attached to Asn297 of the Fc reusluts in reduced effector functions. When generating algycosyl mAbs either in mammalian cells (e.g., CHO) or transgenic plants (where plant/mammalina glycosylation differences might be a problem), the IgG Asn297 can be altered to anotehr amino acid such as Gln or Ala. (presta, Adavnced Drug Delivery Reviews 58, 2006, 640-656). 

Humanization Techniques

Humanized antibodies can be produced using a variety of techniques including CDR-grafting (WO 91/09967 and US Pat Nos: 5,225,539, 5530101 and 5585089), veneering or resurfacing, and chain shuffling (US 5,565,332).

CDR-graft: The substitution of mouse CDRs into a human variable domain framework is most likely to result in retention of their correct spatial orientation if the human variable domain framework adopts the same or similar conformation to the mouse variable framework from which the CDRs originated. This is achieved by obtaining the human variable domains from human antibodies whose framework sequences exhibit a high degree of sequence identity with the murine variable framework domains from which the CDRs were dervied. The heavy and light chain variable fraemwork regions can be dervied from the same or different human antibody sequences. Having identified the complementarity determining regions of the murine donor immunoglobulin and appropriate human acceptor immunoglobulines, the next step is to determine which, if any, residues form these components should be substituted to optimize the properties of the resulting humanized antibody. In general, substitution of human amino acid residues with murine should be minimized, because introduction of murine reisdues increases the risk of the antibody eliciting a human anti mouse antibody response in humans. Certain amino acids from the human variable region framework residues are selected for substitution based on their possible influence on CDR conformation and/or binding to antigen. The unnatural juxtaposition of murine CDR regions with human variable framework region can result in unnatural conformational restraints, which, unless corrected by substitution of certin amino acid residues, lead to loss of binding affinity. The selection of amino acid residues for substitution is determined, in part, by computer modeling. The cahins to be modeled are compared for amino acid sequence similarity with chains or domains of solved 3 D structures. The selection of amino acid residues for substituion can also be determined in part by examinaiton of the characteristics of the amino acids at particular locations.

Veneering/resurfacings: The process of veneering involves selectively replacing FR residues (e.g., a murine heavy or light chain variable region, with human FR residues) in order to provide a xenogeneic molecule comprising an antigen binding portion which retains substantially all of the native FR protein folding structure. Veneering techniques are based on the understanding that the antigen binding characteristics fo an antigen binding portion are determined primarily by the structure and relative disposition of the heavy and light chain CDR sets within the antigen-assocation surface. Thus, antigen association specificity can be preserved in a humanized antibody only where the CDR structures, their interaction with each other and their interaction with the rest of the variable region domains are carefully maintained. By using veneering techniques, exterior (e.g., solvent-accessbile) FR residues which are readily encountered by the immune system are selectively replaced with human residues to provide a hybrid molecule that comrpises either a weakly immunogenic, or substantially non-immunogenic veneered surface. In other other cases, FW residues in the FW regions will be substituted with the corresponding residue from the CDR donor antibody to alter, preferably improve, antigen binding. These FW substitutions are identified by modeling of interactions of the CDR and FW residues to identify FW residues important for antigen binding and sequence comparison to identify unusual FW residues at particular positions (US 5,585,089; Riechmann, Nature 1988, 332).

Best-fit: The choice of human variable domains, both light and heavy, to be used in making the humanized antibodies is to reduce antigenicity. According to the so-called “best fit” method, the sequence of the variable domain of a rodent antibody is screened against the entire library of known human variable-domain sequences. The human sequences which are most closely related to that of the rodent are then screened for the presences of specific residues that may be critical for antigen binding, appropriate structural formation and/or stability of the intended humanized mAb. (Sims, J. Immunol., 151:2296 (1993)). The resulting FW sequences matching the desired criteria are then used as the human donor FW regions for the humanized antibody.

Consensus sequences: Another method uses a particular FW derived from the consensus sequence of all human antibodies of a particular subgroup of light or heavy chains. The same FW may be used for several different humanized anti-CD22 antibodies (Carter, Proc. Natl. Acad. Sci USA, 89: 4285 (1992).

Accordingly, an alternative approach is to humanize anitbodies using only a single human framework, regardless of the sequence of the parent murin anitbody. This mehtod has been successfuly used to humanize a number of murine antibodies using a framework derived form consensus sequences of the most abundant human subclasses, namely VLk subgroup I (VL, kI) and VH subgraoup III (VHIII). Use of the most common human VL and VH frameworks minimizes any potential immunogenicity of the humanized antibody and also eliminates possible idiosyncracies assocaited with any one particular frawework. yields of antibody when expressed recobminantly in either E. coli or eukaryotic expression systems, an important consideration for antibodies that are to go into large scale development. Based on the collective data, the framework residues that most often influences antigen binding are also consistently derived form a set of only 11 residues. In addition, this framework has been demonstrated to give good (Wells “Antibody humanization using monovalent phage display” 212(16), 19996).

Homology modeling can also be used to identify residues unique to the murine antibody sequences that are predicted to be critical to the structure of the antibody combining site (e.g., the CDRs). Homology modeling is a computational method whereby approximate 3 dimensional coordinates are generated for a protein. The source of initial coordinates and guidance for their further refinement is a second protein, the reference protein, for which the 3 dimensional coordinates are known and the sequence of which is related to the sequence of the first protein. The relationship between the sequences of the 2 proteins is used to generate a correspondence between the reference protein and the protein for which coordinates are desired, the target protein. The primary sequences of the reference and target proteins are aligned with coordinates of identical portions of the two proteins transferred directly form the reference protein to the target protein. Coordinates for mismatched portions of the two proteins (e.g., from residue mutations, insertions, or deletions) are constructed from generic structural templates and energy refined to insure consistency with the already transferred model coordinates. This computations protein structure may be further refined or employed direclty in modeling studies. The quality of the model structure is determiend by the accruacy of the contention that the reference and target proteins are related and the precision with which the sequence alignment is constructed.

Veneered antibodies are non-human antiobdy fragments in which the exposed residues in the framework regions are replaced to match the human residues at those positions.

Retention of Specificity-Determining Residues (SDRs): Not all CDR residues are critical in the complementarity of Ag-Ab surfaces. In fact, analysis of the known structures of Ag-Ab complexes suggests that only 20-33% of CFR residues are involved in the Ag contact. These residues are designated as “SDRs” and usually located at positions that display high variability. (Tamaura, J. Immunology, 2000, 164: 1432-1441).

Phage Display: Guided selection, Framework shuffling:

In a frameowrk library approach a collection of residue variants are introduced at specific positions in the FR followed by panning of the library to select the FR that best supports the grafted CDR. An alternative to this FR library approach, a buided selection approach has been used asin the case of Humira in which guided selection allows for the isolation of completely human antibodies. VH and VL domain of a given non-human anitobdy specific for a particular epitope with a human VH or VL librane and specific human V domains are selected agaisnt the antigen of interest. In teh case of Humara, a rodent antibody agasint human TNF-alpha was humanized by first combining the rodent VH with a library of human L chains. The library was then selected agasint NF-alpha by pahge display and the selected VL was cloned into a birary of human VH chains and selected agasint tNF-lpaha. As a resutl, a fully human antibody with similar affinity as the rodent antibody was isoalted. (Fransson “Humanization of antibodies” Frontiers in Bioscience 13, 1619-1633, 2008)

Wu (“Antibody humanizaiton by framework shuffling” Methods 36 (2005) 43 60) disclose humanization of a mouse mAb using framework shuffling. First mAb was raised against the human receptor tyrosine kinsease EphA2 which is upregulated in many cancer cell lines. The 6 CDRs were fused in frame to pools of correspoinding individual human frameworks which encompased all known H and L *k) chain human germline genes. The resulting Fab combinatorial libraries were then screened for binding to the antigen. 

Wells (“Antibody humanization using monovalent phage display” 212(16), 1996) discloses randomizing a small set of critical fraemwork residues and by monovalent display of the resultant library of antibody molecules on the surface of filamentous pahge, optimal fraemwork sequences are identified via affinity based section.

Use of transgenic animals which contain human immunoglobuilin genes: In addition to issues of safety, half-life and effector function, the process of humanizing murine antibodies often reduces their affinity considerably, which adversely impacts efficacy. To address these concerns, mice which bear human heavy and light chain immunoglobuilin gene loci introduced as minichromosomes or transgenes have been developed. One such mouse is the Hu-MAb-Mouse (GenPharm International), which contains one of three human heavy chain transgenes desiganted HC2, HCo7 and HCo12 and the human light chain transgene Kco5. The heavy chain transgene constructs HC2, HCo7 and HCo12 are comprised of human immunoglobuilin heavy chain variable (VH), diversity (D) and joining (JH) segments alsong with the u. y1, and/or y3 constant (C) region exons, the associated swithc regions, the JH intronic enhancer and the rate 3′ heavy chain enhancer. The light chain trangene construct KCo5 is comprised of human immunoglobuilin light chain variable (Vk), joining Jk and constant Ck reegion segments.  Such transgenic mice can be immunized with pathogens such as toxin for bacterial disease which results in the production of hybridomas which can then be selected for antibodie against such toxins. (Tzipori, Clinical Microbiology Reviews, Oct. 2004, p. 926-941).