Taking immunogenicity assessment of therapeutic proteins to the next level.

Therapeutic proteins provide innovative and effective therapies for numerous diseases. However, some of these products are associated with unwanted immunogenicity that may lead to clinical consequences such as reduced or loss of efficacy, altered pharmacokinetics (PK), general immune and hypersensitivity reactions, and neutralisation of the natural counterpart (e.g. the physiological hormone). Regulatory guidance on immunogenicity assessment needs to take into consideration a great diversity of products, indications and patient populations as well as constantly advancing manufacturing technologies. Such guidance needs to be sufficiently specific while, at the same time, allowing interactive discussion and adjusted benefit-risk weighing of each product on a case-by-case basis, e.g. for a unique treatment of a life threatening disease acceptable treatment risks may differ considerably from the ones in case of less serious disease. This theme was the focus of the international conference "Taking immunogenicity assessment of therapeutic proteins to the next level", held at the Paul-Ehrlich-Institut in Langen, Germany, on the 10-11. June 2010. The objectives of the conference were to highlight how the field could move from that of a mere description of risk factors to a system of risk assessment and mitigation, as well as an understanding of the impact of unwanted immunogenicity on the overall benefit/risk consideration for a medicinal product. More than 150 experts from industry, academia and regulatory authorities worldwide discussed the phenomenon of undesired immunogenicity from different perspectives. The conference focussed on issues relevant to three areas: (1) new European guidelines that are currently the subject of discussion; (2) testing strategies for immunogenicity assessment; and (3) scientific progress on the product-related factors that may contribute to the development of pathogenesis of immunogenicity, in particular in the field of protein aggregation and post-translational modifications. This report provides an overview of issues, insights, and conclusions that were discussed and achieved during the meeting.

Complete amino acid sequences of variable regions of two human IgM rheumatoid factors, BOR and KAS of the Wa idiotypic family, reveal restricted use of heavy and light chain variable and joining region gene segments.

Evidence derived from the complete amino acid sequences of the variable regions of both the heavy and light chains of two members (BOR and KAS) of the Wa idiotypic family of human rheumatoid factors suggests that not only are the light chains of these molecules derived from possibly one variable region gene segment, but the heavy chain variable regions are all derived from the VHI subgroup of human V region genes. These molecules exhibit a surprising conservation in the size of D region, and all use the JH4 gene element. This restriction in use of VL, VH, D, and JH suggests all of these elements may play a crucial role in either antigen binding and/or expression of the crossreactive idiotype.

Comparison of the effector functions of human immunoglobulins using a matched set of chimeric antibodies.

Cell lines have been established that secrete a matched set of human chimeric IgM, IgG1, IgG2, IgG3, IgG4, IgE, and IgA2 antibodies that are directed against the hapten 4-hydroxy-3-nitrophenacetyl. These chimeric antibodies secreted from mouse plasmacytoma cells behave exactly like their authentic human counterparts in SDS-PAGE analysis, binding to protein A and in a wide range of serological assays. The antibodies have been compared in their ability to bind human C1q as well as in their efficacy in mediating lysis of human erythrocytes in the presence of human complement. A major conclusion to emerge is that whereas IgG3 bound C1q better than did IgG1, the chimeric IgG1 was much more effective than all the other IgG subclasses in complement-dependent hemolysis. The IgG1 antibody was also the most effective in mediating antibody-dependent cell-mediated cytotoxicity using both human effector and human target cells. These results suggest that IgG1 might be the favoured IgG subclass for therapeutic applications.

Selective expression of a VHIV subfamily of immunoglobulin genes in human CD5+ B lymphocytes from cord blood.

Human B lymphocytes expressing the CD5 surface antigen (CD5+ B cells) constitute a subset capable of producing polyspecific antibodies recognizing a variety of self antigens. The repertoire of antibodies produced by CD5+ and CD5- B cells is different. However, it is not yet established whether this distribution is reflected in different immunoglobulin variable region gene (IgV) use. Rearrangement of heavy chain IgV (IgVH) genes represents one of the first identifiable stages in the maturation of B cells, and occurs in a developmentally ordered fashion. The repertoire of IgVH gene expression is highly restricted during fetal life but diversifies progressively after birth. A high frequency of VH gene use from the relatively small VHIV gene family has previously been demonstrated in human fetal liver B cells. In the present study, 102 B cell lines established by Epstein-Barr Virus-transformation of separated CD5+ and CD5- cord blood B cells, were examined for the frequency of IgV expression using monoclonal antibodies to cross-reactive idiotypes (CRI). The results demonstrate a relatively high frequency of VHIV gene use (30%) in B cells from cord blood. Furthermore, two mutually exclusive CRI associated with distinct subgroups of the VHIV family are segregated in their association with either subset of B cells. One CRI is exclusively expressed in lines established from CD5+ B cells while the other is associated with lines established from CD5- B cells.

Differential glycosylation of polyclonal IgG, IgG-Fc and IgG-Fab isolated from the sera of patients with ANCA-associated systemic vasculitis.

Post-translational modifications (PTMs) of proteins produced in vivo may be tissue, developmentally and/or disease specific. PTMs impact on the stability and function of proteins and offer a challenge to the commercial production of protein biotherapeutics. We have previously reported a marked deficit in galactosylation of oligosaccharides released from polyclonal IgG isolated from sera of patients with the anti-neutrophil cytoplasmic antibodies (ANCA) associated vasculitides; Wegener's granulomatosis (WG) and microscopic polyangiitis (MPA). Whilst normal polyclonal IgG molecules are glycosylated within the IgG-Fc region, approximately 20% of molecules also bear oligosaccharides attached to the variable regions of the light or heavy chain IgG-Fab. It is of interest, therefore to compare profiles of oligosaccharides released from the IgG-Fc and IgG-Fab of normal IgG with that isolated from the sera of patients with WG or MPA. This study shows that whilst the oligosaccharides released from ANCA IgG-Fc are hypogalactosylated those released from IgG-Fab are galactosylated and sialylated. These results show that hypogalactosylation of IgG-Fc is not due to a defect in the glycosylation or processing machinery. It rather suggests a subtle change in IgG-Fc conformation that influences the addition of galactose. Remarkably, this influence is exerted on all plasma cells. Interestingly, a licensed monoclonal antibody therapeutic, produced in Sp2/0 cells, is also shown to be hypogalactosylated in its IgG-Fc but fully galactosylated in its IgG-Fab.

Idiotypic and subgroup analysis of human monoclonal rheumatoid factors. Implications for structural and genetic basis of autoantibodies in humans.

Rheumatoid factors (RFs) in humans have been studied intensively because of their association with autoimmune and lymphoproliferative diseases. Many human IgM-RFs express cross-reactive idiotypes (CRIs) and have homologous light chains, some of which are encoded by a single V kappa gene, termed V kappa 325. However, although antibody activity generally requires the interaction between heavy and light chain variable regions, much less is known about structural relationships among RF heavy chains. To delineate further the structural and genetic basis of RF autoantibody synthesis, we generated "sequence-dependent" reagents specific for the human heavy and kappa light chain subgroups, and used them to analyze a panel of 27 monoclonal RFs. In addition, these proteins were tested for the expression of a heavy chain-associated CRI (G6), and a light chain-associated CRI (17.109). The results showed that most 17.109-reactive RFs contain heavy chains of the VHI subgroup, which bear the G6 idiotypic marker. However, among the 14 17.109-reactive RFs, two have heavy chains of the VHII subgroup, and another two contain heavy chains of the VHIII subgroup. Previously, we have shown that 17.109 is a phenotypic marker of the human V kappa 325 gene. Accordingly, these results demonstrate that the same human V kappa gene can combine with several VH genes from different VH gene subgroups to generate RF activity.

Immunogenicity and antigenicity of immunoglobulins: detection of human immunoglobulin light-chain carbohydrate, using concanavalin A.

Concanavalin A binding to glycoprotein bands on nitrocellulose blots was used to detect mannose, sorbose, N-acetylgalactosamine and/or glucose residues on 100% (31/31) of human Bence Jones protein light chains, following sodium dodecyl sulphate-polyacrylamide gel electrophoresis. All (20/20) light chains form IgG myeloma proteins and light chains from a preparation of normal polyclonal human IgG were also bound by concanavalin A. The specificity of concanavalin A for glycoproteins was demonstrated by its binding to human Fc fragments and a human monoclonal anti-Rhesus D antibody (REG-A), but not to human albumin pFc' fragments and aglycosylated REG-A derived from cells grown in the presence of the glycosylation inhibitor tunicamycin. These results suggest that all Bence Jones proteins and light chains from myeloma IgG proteins contain mono- or oligosaccharides linked O-glycosidically to serine or threonine residues.

Structural analysis of human IgG-Fc glycoforms reveals a correlation between glycosylation and structural integrity.

Antibodies may be viewed as adaptor molecules that provide a link between humoral and cellular defence mechanisms. Thus, when antigen-specific IgG antibodies form antigen/antibody immune complexes the effectively aggregated IgG can activate a wide range of effector systems. Multiple effector mechanisms result from cellular activation mediated through a family of IgG-Fc receptors differentially expressed on leucocytes. It is established that glycosylation of IgG-Fc is essential for recognition and activation of these ligands. IgG antibodies predominate in human serum and most therapeutic antibodies are of the IgG class. The IgG-Fc is a homodimer of N-linked glycopeptide chains comprised of two immunoglobulin domains (Cgamma2, Cgamma3) that dimerise via inter-heavy chain disulphide bridges at the N-terminal region and non-covalent interactions between the C-terminal Cgamma3 domains. The overall shape of the IgG-Fc is similar to that of a "horseshoe" with a majority of the internal space filled by the oligosaccharide chains, only attached through asparagine residues 297.To investigate the influence of individual sugar (monosaccharide) residues of the oligosaccharide on the structure and function of IgG-Fc we have compared the structure of "wild-type" glycosylated IgG1-Fc with that of four glycoforms bearing consecutively truncated oligosaccharides. Removal of terminal N-acetylglucosamine as well as mannose sugar residues resulted in the largest conformational changes in both the oligosaccharide and in the polypeptide loop containing the N-glycosylation site. The observed conformational changes in the Cgamma2 domain affect the interface between IgG-Fc fragments and FcgammaRs. Furthermore, we observed that the removal of sugar residues permits the mutual approach of Cgamma2 domains resulting in the generation of a "closed" conformation; in contrast to the "open" conformation which was observed for the fully galactosylated IgG-Fc, which may be optimal for FcgammaR binding. These data provide a structural rationale for the previously observed modulation of effector activities reported for this series of proteins.

Molecular definition of interaction sites on human IgG for Fc receptors (huFc gamma R).

Evidence from several experimental approaches allows us to conclude that the primary amino acid sequence of the lower hinge region (residues 234-237) of human IgG molecules determines recognition by human Fc gamma RI, Fc gamma RII and Fc gamma RIII. Glycosylation of the CH2 domain is also essential, although the carbohydrate is not accessible for direct interaction with ligands. The role of the carbohydrate moiety may be to maintain a protein conformation that allows accessibility to amino acid side chains essential for ligand recognition and binding. It appears logical that the evolutionarily-related Fc gamma R molecules should interact with overlapping non-identical sites on the IgG molecule.

Aglycosylated chimaeric human IgG3 can trigger the human phagocyte respiratory burst.

This study investigates the capacity of a complexed aglycosylated chimaeric human IgG3 antibody to induce the respiratory burst in human monocyte-like U937 cells. It demonstrates that the aglycosylated antibody, prepared by cell culture in tunicamycin, retains significant capacity to trigger this effector function which was assayed as superoxide generation. Erythrocytes sensitized with near maximal levels of aglycosylated IgG3 were able to trigger > 80% of the superoxide generation triggered by the glycosylated antibody from U937 cells induced to differentiate by interferon gamma and the aglycosylated IgG3 gave half maximal responses at sensitization levels only 72% higher than those required by the glycosylated form. Aglycosylated IgG3 was, however, much less effective in triggering superoxide generation by interferon gamma treated U937 cells at low sensitization levels as threshold responses required only 60 glycosylated IgG3 molecules per erythrocyte compared with 16,000 aglycosylated molecules. In addition, these studies indicate significant differences between the target cell to effector cell ratios which permit IgG sensitized erythrocytes to stimulate the respiratory burst and those which stimulate ADCC in the same effector cell type.

Human Fc gamma RI triggering of the mononuclear phagocyte respiratory burst.

This study investigates the role played by Fc gamma RI and Fc gamma RII in triggering the respiratory burst induced in human monocyte-like U937 cells by monoclonal chimaeric anti-NIP antibodies expressing the human IgG subclass and mouse IgG2b heavy chains. Respiratory burst activity was measured as superoxide generation. Four separate lines of evidence indicate a predominant role for Fc gamma RI in triggering superoxide generation induced by erythrocytes sensitized with up to the maximum of 100,000 IgG molecules per cell. Firstly, erythrocytes sensitized with mouse IgG2b anti-NIP antibodies which are not recognized by human Fc gamma RI, did not induce a response but when residue Glu-235 was replaced by Leu to give the lower hinge sequence of mouse IgG2a which is recognized by Fc gamma RI, the mutant bound to Fc gamma RI and induced a response equal to 80% of that given by chimaeric human IgG3. Chimaeric human IgG3 antibodies with amino acid substitutions in the lower hinge showed reduced activity and the greatest reductions (< 32% of wild type antibody activity) were associated with changes at Leu-235 which is critical for recognition by Fc gamma RI. Secondly, chimaeric human IgG4 antibodies which are not recognized by Fc gamma RII, were able to induce superoxide generation. The rank order of abilities of chimaeric human IgG subclass antibodies to induce responses was IgG3 > IgG1 > IgG4 > > IgG2. Thirdly, responses induced by chimaeric human IgG were inhibited by concns of monomeric human IgG3 in the nM range. Finally, chimaeric human IgG3 induced responses were inhibited by anti-Fc gamma RI, but not anti-Fc gamma RII monoclonal antibodies. Consistent with a major role for Fc gamma RI in triggering the responses of U937 cells, erythrocytes sensitized with chimaeric human IgG3 did not induce superoxide generation by neutrophils which express Fc gamma RII and Fc gamma RIII, or eosinophils which express Fc gamma RII, but neither of which expresses Fc gamma RI.

Localisation of the monocyte-binding region on human immunoglobulin G.

Earlier studies, which provided indirect evidence for the involvement of the C gamma 2 domain of human immunoglobulin G (IgG) in human immunoglobulin G (IgG) in human monocyte binding, have been extended to further localise the site of interaction on human IgG. A number of IgGs from several different species and fragments of human IgGs were assayed for ability to inhibit the interaction of radio-labelled human IgG and the human monocyte. By comparison of the amino-acid sequences of those IgGs found to exhibit relatively tight, intermediate or weak binding to human monocyte Fc receptors we are able to postulate a possible monocyte-binding site on human IgG. In addition, the results have implications for the applicability of monoclonal antibodies and antisera when used in the presence of human monocytes and possibly macrophages.

The use of anti-IgG monoclonal antibodies in mapping the monocyte receptor site on IgG.

Monoclonal antibodies (MAbs) directed against epitopes on the C gamma 1, C gamma 2, C gamma 3 and C gamma 2-C gamma 3 interface regions of human IgG were used to attempt to localize the monocyte Fc receptor (FcR) binding site. The MAbs have been assayed for their capacity to inhibit the interaction between 125I-labelled IgG (125I-IgG) and human monocytes or human histiocytic lymphoma U937 cells. Two MAbs specific for epitopes on the N-terminal region of the C gamma 2 domain, and one MAb recognizing an epitope in the C gamma 2-C gamma 3 inter-domain region inhibited binding of 125I-IgG to monocyte FcRs. The remaining MAbs, against a C-terminal C gamma 3 domain epitope, another C gamma 2/C gamma 3 region epitope and the G1m(f) allotope on the C gamma 1 domain did not inhibit the interaction. The capacity of the MAbs to bind to their respective epitopes on cell surface FcR-bound IgG was also studied, using indirect radiobinding and immunofluorescence assays. All of the MAbs, except those with C gamma 2 domain specificities, were able to detect FcR-bound IgG under these conditions. The results confirm the role of the C gamma 2 domain in the interaction of IgG with monocytes and demonstrate that epitopes in the C gamma 3 and C gamma 2-C gamma 3 regions are not involved in the binding site.

Mapping and comparison of the interaction sites on the Fc region of IgG responsible for triggering antibody dependent cellular cytotoxicity (ADCC) through different types of human Fc gamma receptor.

In the present study 3-iodo-4-hydroxy-5-nitrophenacetyl (NIP)-specific antibodies were compared for induction of antibody dependent lysis of NIP-derivatised red blood cells effected by pre-stimulated U937 or HL-60 cells and by K cells. The chimaeric antibodies have heavy chains corresponding to human IgG subclasses 1-4, and include site-directed mutants of IgG3 as well as the aglycosylated form of IgG3; a mouse IgG2b antibody and a site-directed mutant IgG2b were also examined. rIFN stimulated U937 or HL-60 cells express increased levels of Fc gamma R1 compared to unstimulated cells; PMA stimulated HL-60 and U937 cells express an increased level of Fc gamma R11 compared to unstimulated cells; K cells express Fc gamma R111. Using these effector cell populations and the target cells mentioned above, we have compared anti-NIP antibodies with different heavy chain constant domains for their ability to induce ADCC through human Fc gamma R1, Fc gamma R11 and Fc gamma R111. The results suggest that all three human Fc gamma receptors appear to recognise a binding site on IgG within the lower hinge (residues 234-237) and trigger ADCC via this site, but that each receptor sees this common site in a different way. The possibility that other amino acid residues also participate in the binding/triggering site(s) cannot be excluded.

Molecular recognition of antibody (IgG) by cellular Fc receptor (FcRI).

Earlier studies from this and other laboratories have provided indirect evidence for the involvement of the C gamma 2 domain of human IgG in the binding of IgG to the high affinity monocyte Fc receptor (FcRI). Two approaches have been used to extend these studies and to further localize the site of interaction on human IgG. Firstly, monoclonal antibodies (MAbs) directed against different epitopes on IgG were assayed for their capacity to inhibit the binding of radiolabelled IgG to human monocytes or U937 cells. The capacity of the MAbs to interact with their respective epitopes on FcR-bound IgG was also studied using indirect radiobinding and immunofluorescence assays. Secondly, a number of IgGs from several different species and fragments of human IgGs were assayed for their ability to inhibit the binding of radiolabelled IgG to human monocytes. The amino acid sequences of those IgGs exhibiting relatively tight, intermediate or weak binding to monocyte FcRs were compared. On the basis of these studies a possible monocyte FcR-binding site on human IgG is postulated, involving the lower hinge region of IgG (residues Leu 234-Ser 239) with possible involvement of the nearby N-proximal bend and two beta-strands (Gly 316-Lys 338).

Immunogenic and antigenic epitopes of immunoglobulins--V. Reactivity of a panel of monoclonal antibodies with sub-fragments of human Fc gamma and abnormal paraproteins having deletions.

The specificity of a panel of 27 monoclonal antibodies reactive with human Fc gamma has been further defined through reactivity profiles with sub-fragments of Fc gamma and IgG paraproteins having deletions within C gamma 2 or C gamma 3. Antibodies are identified that are directed to discontinuous epitopes requiring native Fc gamma for expression. Other antibodies are reactive with isolated C gamma 2 or C gamma 3 domains. Sub-class specific and sub-class restricted antibodies allow correlation of epitope expression with primary structure.

The solution conformations of the subclasses of human IgG deduced from sedimentation and small angle X-ray scattering studies.

The solution conformations of human immunoglobulin G subclass molecules have been investigated by sedimentation and small-angle X-ray scattering techniques. Both methods qualitatively indicate IgG3 to be an extended molecule relative to IgG1. Sedimentation data have been collected for a number of paraproteins of all four subclasses and the hinge-deleted IgG1Dob protein. The known crystal structure of Dob allows the use of this protein as a basis for the proposal of models of the average conformations of IgG subclasses which are consistent with experimental s(0)20,w values. IgG1 is suggested to have a hinge length of 0-15 A and non-coplanar Fab arms; IgG2 to be effectively hingeless with folded-back Fab arms; IgG3 to have an extended hinge of the order of 100 A and IgG4 to be effectively hingeless and T-shaped. The possible correlation of these conformations with subclass function is discussed.

Immunogenic and antigenic epitopes of immunoglobulin--XX. Denaturation of human IgG3 by free radicals.

It has previously been demonstrated that exposure of polyclonal IgG to free radicals results in denaturation evidenced by aggregation, auto-fluorescence and destruction of cysteine, proline and aromatic amino acids. In the present study we have used a panel of monoclonal antibodies (McAb) to epitopes expressed on the IgG3 heavy chain to detect changes in antigenicity. When IgG3 was exposed to u.v. irradiation, as a source of free radicals, subclass specific epitopes were rapidly lost whilst other epitopes were unaffected. Prolonged exposure resulted in further denaturation and a progressive loss of expression of further epitopes. The IgG3 subclass specific McAb are specific to epitopes localized to the hinge region of IgG3. Thus, this exposed cysteine and proline rich region is shown to be particularly vulnerable to free radical attack; however, prolonged exposure results in structural alterations throughout the heavy chain.

Restricted immunoglobulin variable region gene usage by hybridoma rheumatoid factors from patients with systemic lupus erythematosus and rheumatoid arthritis.

Rheumatoid factors (RFs) are autoantibodies that are produced by approximately 75% of patients with rheumatoid arthritis (RA). Their role in pathogenesis is not well understood. In this study of 81 human hybridoma IgM antibodies derived from unstimulated peripheral blood B-cells of patients with RA and systemic lupus erythematosus (SLE), we have demonstrated that idiotypes associated with RFs derived from patients with mixed cryoglobulinemia were expressed by approximately 60% of RFs and 6% of IgM antibodies lacking RF activity. The specificity of the RFs for the Fc portion of IgG only (monospecificity) or for Fc and additional self antigens (polyreactivity) was found to correlate with the expression of specific heavy chain associated idiotypes. The VH3 associated RF idiotypes, D12 and B6, were expressed by 0/16 (0%) of monospecific RFs compared with 6/22 (27%) of polyreactive RFs. The predominant use of VH3 was verified by analysis of the expressed Ig with VH family specific anti-peptide antibodies. The light chains expressed by both populations of IgM RFs were found to be predominantly VKIII, both by detection of specific epitopes/idiotypes and V family analysis. This non-random gene usage of both the heavy and light chains suggests that there is a selective expression of V regions in the RF producing B-cells in patients with RA and SLE. We suggest that different antigen-driven, clonal selection events may occur which result in either monospecific RFs or polyreactive RFs.

Interaction of human IgG chimeric antibodies with the human FcRI and FcRII receptors: requirements for antibody-mediated host cell-target cell interaction.

Chimeric monoclonal antibodies (McAb), specific for the hapten 5-iodo-4-hydroxy-3-nitrophenacetyl (NIP), expressing human IgG1, IgG2, IgG3 and IgG4 subclass constant domains, have been examined for their ability to interact with the human FcRII receptor. Human red blood cells (RBC) sensitized by each of these McAbs have been assayed for their ability to form rosettes with the human histiocytic lymphoma U937 cell line, human B cell line Daudi and erythroblastoid K562 cell line. IgG1 and IgG3 sensitized RBC formed significant rosettes with the FcR- and FcRII+ Daudi and K562 cell lines, the percentage of cells forming rosettes being directly proportional to the degree of sensitization of the RBC. Bromelin treating Daudi cells did not alter this pattern of reactivity, whereas bromelin treated FcRI+ and FcRII+ U937 cells formed significant resettes with IgG1, IgG3 and IgG4 sensitized RBC, demonstrating a difference in the IgG subclass specificity between human FcRI and FcRII. Murine IgG2b anti-NIP sensitized RBC did not form rosettes with any cell line tested; however, RBC sensitized by some members of a panel of murine IgG1 McAb, specific for the glycophorin A molecule, were able to form rosettes with Daudi, U937 and K562 cells. This interaction was enhanced by bromelin treating the Daudi or U937 cells and can be correlated to the disposition of the epitopes recognized, relative to the target cell membrane, those McAbs recognizing epitopes furthest from the RBC surface being most effective in interacting with FcRII. The data are interpreted in terms of a simple model for antibody-mediated cell--cell interaction.