Engineering therapeutic antibodies for improved function.

The binding sites on human IgG1 for human Fc gamma receptor (Fc gamma R) I, Fc gamma RIIa, Fc gamma RIIb, Fc gamma RIIIa and neonatal FcR have been mapped. A common set of IgG1 residues is involved in binding to all Fc gamma Rs, while Fc gamma RII and Fc gamma RIII utilize distinct sites outside this common set. In addition to residues which abrogated binding to the Fc gamma R, several positions were found which improved binding only to specific Fc gamma Rs or simultaneously improved binding to one type of Fc gamma R and reduced binding to another type. Selected IgG1 variants with improved binding to Fc gamma RIIIa were then tested in an in vitro antibody-dependent cellular cytotoxicity (ADCC) assay and showed an enhancement in ADCC when either peripheral blood mononuclear cells or natural killer cells were used.

Generation of an LFA-1 antagonist by the transfer of the ICAM-1 immunoregulatory epitope to a small molecule.

The protein-protein interaction between leukocyte functional antigen-1 (LFA-1) and intercellular adhesion molecule-1 (ICAM-1) is critical to lymphocyte and immune system function. Here, we report on the transfer of the contiguous, nonlinear epitope of ICAM-1, responsible for its association with LFA-1, to a small-molecule framework. These LFA-1 antagonists bound LFA-1, blocked binding of ICAM-1, and inhibited a mixed lymphocyte reaction (MLR) with potency significantly greater than that of cyclosporine A. Furthermore, in comparison to an antibody to LFA-1, they exhibited significant anti-inflammatory effects in vivo. These results demonstrate the utility of small-molecule mimics of nonlinear protein epitopes and the protein epitopes themselves as leads in the identification of novel pharmaceutical agents.

Characterization and humanization of a monoclonal antibody that neutralizes human leukocyte interferon: a candidate therapeutic for IDDM and SLE.

We have developed a panel of murine monoclonal antibodies that recognize human interferon alpha. One of these mononclonal antibodies binds and neutralizes, with high affinity, all of seven tested recombinant human interferon alphas. This mononclonal antibody also neutralizes the interferon activity present in two independent pools of interferon alphas prepared following stimulation of human peripheral blood leukocytes. The complementary determining regions from this murine mononclonal antibody were transferred to a human IgG2 heavy chain and to a human kappa1 light chain. In addition, six (heavy chain) and two (light chain) amino acids were transferred from the framework regions. This generated a humanized mononclonal antibody that retained the specificity of the mouse parent. The humanized anti-interferon alpha antibody is a candidate therapeutic for those diseases, such as insulin-dependent diabetes, systemic lupus erythematosis, psoriasis and Crohn's disease, which are all characterized by pathological expression of interferon alpha.

Engineered antibodies with increased activity to recruit complement.

This manuscript describes two sites in a human IgG1 that, when mutated individually or in combination, result in a dramatic increase in C1q binding and complement-dependent cytotoxicity activity. These two residues, K326 and E333, are located at the extreme ends of the C1q binding epicenter in the C(H)2 domain of a human IgG. A mutation to tryptophan at K326 debilitates Ab-dependent cell-mediated cytotoxicity activity. In addition, substitutions of the residues E333 with serine and of K326 with tryptophan in a human IgG2 confer biological activity in the complement-dependent cytotoxicity assay in which the wild-type IgG2 is inactive. This study reveals that the residues K326 and E333 play a significant role in the control of the biological activity of an IgG molecule and can rescue the activity of an inactive IgG isotype.

High resolution mapping of the binding site on human IgG1 for Fc gamma RI, Fc gamma RII, Fc gamma RIII, and FcRn and design of IgG1 variants with improved binding to the Fc gamma R.

Immunoglobulin G (IgG) Fc receptors play a critical role in linking IgG antibody-mediated immune responses with cellular effector functions. A high resolution map of the binding site on human IgG1 for human Fc gamma RI, Fc gamma RIIA, Fc gamma RIIB, Fc gamma RIIIA, and FcRn receptors has been determined. A common set of IgG1 residues is involved in binding to all Fc gamma R; Fc gamma RII and Fc gamma RIII also utilize residues outside this common set. In addition to residues which, when altered, abrogated binding to one or more of the receptors, several residues were found that improved binding only to specific receptors or simultaneously improved binding to one type of receptor and reduced binding to another type. Select IgG1 variants with improved binding to Fc gamma RIIIA exhibited up to 100% enhancement in antibody-dependent cell cytotoxicity using human effector cells; these variants included changes at residues not found at the binding interface in the IgG/Fc gamma RIIIA co-crystal structure (Sondermann, P., Huber, R., Oosthuizen, V., and Jacob, U. (2000) Nature 406, 267-273). These engineered antibodies may have important implications for improving antibody therapeutic efficacy.

TrkA amino acids controlling specificity for nerve growth factor.

Neurotrophins are important for the development and maintenance of the vertebrate nervous system, mediating their signal into the cell by specific interaction with tyrosine kinase receptors of the Trk family. The extracellular portion of the Trk receptors has been previously proposed to consist of a cysteine-rich motif, a leucine-rich motif, a second cysteine-rich motif followed by two immunoglobulin-like domains. Earlier studies have shown that a major neurotrophin-binding site in the Trk receptors resides in the second immunoglobulin-like domain. Although the individual amino acids in TrkA involved in binding to nerve growth factor (NGF) and those in TrkC involved in binding to neurotrophin-3 have been mapped in this domain, the Trk amino acids that provide specificity remained unclear. In this study, a minimum set of residues in the human TrkC second immunoglobulin-like domain, which does not bind nerve growth factor (NGF), were substituted with those from human TrkA. The resulting Trk variant recruited binding of NGF equivalent to TrkA, maintained neurotrophin-3 binding equivalent to TrkC, and also bound brain-derived neurotrophin, although with lower affinity compared with TrkB. This implies that the amino acids in the second immunoglobulin-like domain that determine Trk specificity are distinct for each Trk.

Inhibitory Fc receptors modulate in vivo cytotoxicity against tumor targets.

Inhibitory receptors have been proposed to modulate the in vivo cytotoxic response against tumor targets for both spontaneous and antibody-dependent pathways. Using a variety of syngenic and xenograft models, we demonstrate here that the inhibitory FcgammaRIIB molecule is a potent regulator of antibody-dependent cell-mediated cytotoxicity in vivo, modulating the activity of FcgammaRIII on effector cells. Although many mechanisms have been proposed to account for the anti-tumor activities of therapeutic antibodies, including extended half-life, blockade of signaling pathways, activation of apoptosis and effector-cell-mediated cytotoxicity, we show here that engagement of Fcgamma receptors on effector cells is a dominant component of the in vivo activity of antibodies against tumors. Mouse monoclonal antibodies, as well as the humanized, clinically effective therapeutic agents trastuzumab (Herceptin(R)) and rituximab (Rituxan(R)), engaged both activation (FcgammaRIII) and inhibitory (FcgammaRIIB) antibody receptors on myeloid cells, thus modulating their cytotoxic potential. Mice deficient in FcgammaRIIB showed much more antibody-dependent cell-mediated cytotoxicity; in contrast, mice deficient in activating Fc receptors as well as antibodies engineered to disrupt Fc binding to those receptors were unable to arrest tumor growth in vivo. These results demonstrate that Fc-receptor-dependent mechanisms contribute substantially to the action of cytotoxic antibodies against tumors and indicate that an optimal antibody against tumors would bind preferentially to activation Fc receptors and minimally to the inhibitory partner FcgammaRIIB.

Mapping of the C1q binding site on rituxan, a chimeric antibody with a human IgG1 Fc.

Rituxan (Rituximab) is a chimeric mAb with human IgG1 constant domains used in the therapy of non-Hodgkin's B cell lymphomas. This Ab targets B cells by binding to the cell-surface receptor, CD20. In our investigation of the mechanism of B cell depletion mediated by Rituximab, we first constructed mutants of Rituximab defective in complement activation but with all other effector functions intact. Our results demonstrate that the previously described C1q binding motif in murine IgG2b constituting residues E318, K320, and K322 is not applicable to a human IgG1 when challenged with either human, rabbit, or guinea pig complement. Alanine substitution at positions E318 and K320 in Rituximab had little or no effect on C1q binding and complement activation, whereas alanine substitution at positions D270, K322, P329, and P331 significantly reduced the ability of Rituximab to bind C1q and activate complement. We have also observed that concentrations of complement approaching physiological levels are able to rescue >60% of the activity of these mutant Abs with low affinity for C1q. These data localize the C1q binding epicenter on human IgG1 and suggest that there are species-specific differences in the C1q binding site of Igs.

Mapping the intercellular adhesion molecule-1 and -2 binding site on the inserted domain of leukocyte function-associated antigen-1.

By extensive mutagenic analysis of the inserted domain (I-domain) of the alpha-chain (CD11a) of the leukocyte function-associated antigen-1 (LFA-1), we have defined a putative binding surface for intercellular adhesion molecules 1 and 2 (ICAM-1 and -2). This analysis showed that individually mutating Leu-205 or Glu-241 to alanine completely abolished LFA-1 binding to ICAM-1 or -2 without affecting I-domain structure, as assayed by antibody binding. Mutating Thr-243 to alanine also had a profound effect on LFA-1 binding to ICAM-1 and -2, as seen by complete loss of binding to ICAM-1 and a significant reduction (70% decrease) in binding to ICAM-2. Mutating Glu-146 to alanine reduced LFA-1 binding to ICAM-1 or -2 by 70%, and mutating His-264 or Glu-293 to alanine reduced binding to ICAM-1 or -2 by about 30-40%. Mutating Thr-175 to alanine reduced binding to ICAM-1 by about 30% and binding to ICAM-2 by about 70%. Interestingly, mutating Lys-263 to alanine preferentially abolished LFA-1 binding to ICAM-2. Using these data, we have generated a model of the interface between the LFA-1 I-domain and residues in the first domain of ICAM-1 that have been shown to be critical for this interaction. In addition, this model, together with the ICAM-2 crystal structure, has been used to map residues that are likely to mediate LFA-1 I-domain binding to ICAM-2.

Humanization of an antibody against human protein C and calcium-dependence involving framework residues.

A murine monoclonal antibody to the zymogen form of human protein C that blocks protein C activation by thrombin-thrombomodulin both in vitro and in vivo has been humanized using the consensus and resurfacing methods. While the binding of the parent murine antibody to protein C is calcium-dependent (1.5-2-fold decrease in binding without calcium), the humanized antibody exhibited a significant increase in its calcium-dependence (5-fold decrease in binding without calcium). Two exposed human framework residues in the variable light domain of the humanized antibody, aspartic acid L1 and glutamine L3, are responsible for the increase in calcium-dependence.

An efficient route to human bispecific IgG.

Production of bispecific IgG (BsIgG) by coexpressing two different antibodies is inefficient due to unwanted pairings of the component heavy and light chains. To overcome this problem, heavy chains were remodeled for heterodimerization using engineered disulfide bonds in combination with previously identified "knobs-into-holes" mutations. One of the variants, S354C:T366W/Y349'C:T366'S:L368'A:Y407++ +'V, gave near quantitative (approximately 95%) heterodimerization. Light chain mispairing was circumvented by using an identical light chain for each arm of the BsIgG. Antibodies with identical light chains that bind to different antigens were identified from an scFv phage library with a very restricted light chain repertoire for the majority (50/55) of antigen pairs tested. A BsIgG capable of simultaneously binding to the human receptors HER3 and cMpI was prepared by coexpressing the common light chain and corresponding remodeled heavy chains followed by protein A chromatography. The engineered heavy chains retain their ability to support antibody-dependent cell-mediated cytotoxicity as demonstrated with an anti-HER2 antibody.

High resolution mapping of the binding site of TrkA for nerve growth factor and TrkC for neurotrophin-3 on the second immunoglobulin-like domain of the Trk receptors.

Neurotrophic factors are important for survival and maintenance of neurons during developmental and adult stages of the vertebrate nervous system. The neurotrophins mediate their signal into the cell by specific interaction with tyrosine kinase receptors of the Trk family. The extracellular immunoglobulin-like domain of the Trk receptors adjacent to the membrane has previously been shown to be the dominant element for specific neurotrophin binding. Using computer graphics models of the human TrkA and TrkC immunoglobulin-like domains as a guide, the residues involved in binding to their respective neurotrophins were mapped by mutational analysis. TrkC primarily utilizes loop EF, between beta-strands E and F, for binding. In contrast, TrkA utilizes the EF loop as well as additional residues, the latter being prime candidates for determining the specificity of TrkA versus TrkC. When selected TrkC and TrkA mutants with reduced binding were expressed on NIH3T3 cells, neurotrophin-induced autophosphorylation was strongly reduced or absent.

Mapping the charged residues in the second immunoglobulin-like domain of the vascular endothelial growth factor/placenta growth factor receptor Flt-1 required for binding and structural stability.

Flt-1 is one of two receptor tyrosine kinases through which the angiogenic factor vascular endothelial growth factor (VEGF) functions. Placenta growth factor (PlGF) is an additional ligand for Flt-1. The second immunoglobulin-like domain in the extracellular domain of Flt-1 has previously been identified as the region containing the critical ligand-binding determinants. We analyzed the contribution of charged residues within the first three domains of Flt-1 to ligand binding by alanine-scanning mutagenesis. Domain 2 residues Arg159, Glu208 and His223-Arg224 (together) affect both VEGF and PlGF binding, while Glu137, Lys171, His223, and Arg224 affect PlGF but not VEGF. Several charged residues, especially Asp187, are important in maintaining the structural integrity of domain 2. In addition, some residues in domain 3 contribute to binding (Asp231) or provide for additional discrimination between ligands (Arg280-Asp283).

Humanization of an anti-vascular endothelial growth factor monoclonal antibody for the therapy of solid tumors and other disorders.

Vascular endothelial growth factor (VEGF) is a major mediator of angiogenesis associated with tumors and other pathological conditions, including proliferative diabetic retinopathy and age-related macular degeneration. The murine anti-human VEGF monoclonal antibody (muMAb VEGF) A.4.6.1 has been shown to potently suppress angiogenesis and growth in a variety of human tumor cells lines transplanted in nude mice and also to inhibit neovascularization in a primate model of ischemic retinal disease. In this report, we describe the humanization of muMAb VEGF A.4.6.1. by site-directed mutagenesis of a human framework. Not only the residues involved in the six complementarity-determining regions but also several framework residues were changed from human to murine. Humanized anti-VEGF F(ab) and IgG1 variants bind VEGF with affinity very similar to that of the original murine antibody. Furthermore, recombinant humanized MAb VEGF inhibits VEGF-induced proliferation of endothelial cells in vitro and tumor growth in vivo with potency and efficacy very similar to those of muMAb VEGF A.4.6.1. Therefore, recombinant humanized MAb VEGF is suitable to test the hypothesis that inhibition of VEGF-induced angiogenesis is a valid strategy for the treatment of solid tumors and other disorders in humans.

Mutational analysis of thrombopoietin for identification of receptor and neutralizing antibody sites.

Thrombopoietin (TPO) is a hematopoietin important for megakaryocyte proliferation and production of blood platelets. We sought to characterize how TPO binds and activates its receptor, myeloproliferative leukemia virus receptor. The erythropoietin-like domain of TPO (TPO1-153) has been fused to the gIII coat protein of M13 bacteriophage. Forty residues were chosen for mutation to alanine using the criteria that they were charged residues or predicted to be solvent-exposed, based on a homology model. Phage enzyme-linked immunosorbent assay was used to determine affinities for binding to both the TPO receptor and five anti-TPO1-153 monoclonal antibodies. Mutations at mostly positively charged residues (Asp8, Lys14, Lys52, Lys59, Lys136, Lys138, Arg140) caused the greatest reduction in receptor-binding affinity. Most of these residues mapped to helices-1 and -4 and a loop region between helix-1 and helix-2. Two of the monoclonal antibodies that blocked TPO binding and bioactivity had determinants in helix-4. In contrast, the other three monoclonal antibodies, which were effective at blocking TPO activity but did not block initial binding of TPO to its receptor, had epitopes predominantly on helix or 3. These results suggest that TPO has two distinct receptor-binding sites that function to dimerize TPO receptors in a sequential fashion.

Antibody humanization using monovalent phage display.

Antibody humanization often requires the replacement of key residues in the framework regions with corresponding residues from the parent non-human antibody. These changes are in addition to grafting of the antigen-binding loops. Although guided by molecular modeling, assessment of which framework changes are beneficial to antigen binding usually requires the analysis of many different antibody mutants. Here we describe a phage display method for optimizing the framework of humanized antibodies by random mutagenesis of important framework residues. We have applied this method to humanization of the anti-vascular endothelial growth factor murine monoclonal antibody A4.6.1. Affinity panning of a library of humanized A4.6.1 antibody mutants led to the selection of one variant with greater than 125-fold enhanced affinity for antigen relative to the initial humanized antibody with no framework changes. A single additional mutation gave a further 6-fold improvement in binding. The affinity of this variant, 9.3 nM, was only 6-fold weaker than that of a murine/human chimera of A4.6.1. This method provides a general means of rapidly selecting framework mutations that improve the binding of humanized antibodies to their cognate antigens and may prove an attractive alternative to current methods of framework optimization based on cycles of site-directed mutagenesis.

Remodeling domain interfaces to enhance heterodimer formation.

An anti-p185HER2/anti-CD3 humanized bispecific diabody was previously constructed from two cross-over single-chain Fv in which YH and VL domains of the parent antibodies are present on different polypeptides. Here this diabody is used to evaluate domain interface engineering strategies for enhancing the formation of functional heterodimers over inactive homodimers. A disulfide-stabilized diabody was obtained by introducing two cysteine mutations, VL L46C and VH D101C, at the anti-p185HER2.VL/VH interface. The fraction of recovered diabody that was functional following expression in Escherichia coli was improved for the disulfide-stabilized compared to the parent diabody (> 96% versus 72%), whereas the overall yield was > 60-fold lower. Eleven "knob-into-hole" diabodies were designed by molecular modeling of sterically complementary mutations at the two VL/VH interfaces. Replacements at either interface are sufficient to improve the fraction of functional heterodimer, while maintaining overall recoverable yields and affinity for both antigens close to that of the parent diabody. For example, diabody variant v5 containing the mutations VL Y87A:F98M and VH V37F:L45W at the anti-p185HER2 VL/VH interface was recovered as 92% functional heterodimer while maintaining overall recovered yield within twofold of the parent diabody. The binding affinity of v5 for p185HER2 extracellular domain and T cells is eightfold weaker and twofold stronger than for the parent diabody, respectively. Domain interface remodeling based upon either sterically complementary mutations or interchain disulfide bonds can facilitate the production of a functional diabody heterodimer. This study expands the scope of domain interface engineering by demonstrating the enhanced assembly of proteins interacting via two domain interfaces.

Specificity determinants in neurotrophin-3 and design of nerve growth factor-based trkC agonists by changing central beta-strand bundle residues to their neurotrophin-3 analogs.

Neurotrophic factors mediate their signal by binding to specific cell surface receptors of the trk family. The binding sites of neurotrophin-3 (NT-3) and nerve growth factor (NGF) to their preferred receptors trkC and trkA, respectively, were previously determined by mutational analyses. These and other studies showed that trkA can discriminate between NGF and NT-3 primarily by recognition of their N-terminal residues. The mechanism of trkC discrimination, however, remained unclear especially since the most important residue in NT-3 involved in binding to trkC, R103, is conserved in all neurotrophins. In this study residues that are part of the central beta-strand bundle of NT-3 and are not conserved among the neurotrophins were grafted onto NGF and tested for recruitment of trkC affinity. Exchange of NGF residues at positions 18, 20, 23, 29, 84, and 86 by their NT-3 counterparts resulted in NGF variants that bound to trkC, while maintaining their affinity to trkA, and were able to induce autophosphorylation and differentiation of PC12 cells expressing trkC. These variants show that the amino acid at position 23 (glycine in NGF, threonine in NT-3) is critical for trkC recognition while other residues fine tune the specificity of NT-3 for trkC. The results demonstrate the importance of nonconserved residues of the central beta-strand bundle region for the interaction of NT-3 with trkC and emphasize the different mechanism of specificity determination that is employed in the NT-3/trkC and NGF/trkA ligand/receptor pairs.

Identification of the binding site in intercellular adhesion molecule 1 for its receptor, leukocyte function-associated antigen 1.

Intercellular adhesion molecule 1 (ICAM-1, CD54) is a member of the Ig superfamily and is a counterreceptor for the beta 2 integrins: lymphocyte function-associated antigen 1 (LFA-1, CD11a/CD18), complement receptor 1 (MAC-1, CD11b/CD18), and p150,95 (CD11c/CD18). Binding of ICAM-1 to these receptors mediates leukocyte-adhesive functions in immune and inflammatory responses. In this report, we describe a cell-free assay using purified recombinant extracellular domains of LFA-1 and a dimeric immunoadhesin of ICAM-1. The binding of recombinant secreted LFA-1 to ICAM-1 is divalent cation dependent (Mg2+ and Mn2+ promote binding) and sensitive to inhibition by antibodies that block LFA-1-mediated cell adhesion, indicating that its conformation mimics that of LFA-1 on activated lymphocytes. We describe six novel anti-ICAM-1 monoclonal antibodies, two of which are function blocking. Thirty-five point mutants of the ICAM-1 immunoadhesin were generated and residues important for binding of monoclonal antibodies and purified LFA-1 were identified. Nineteen of these mutants bind recombinant LFA-1 equivalently to wild type. Sixteen mutants show a 66-2500-fold decrease in LFA-1 binding yet, with few exceptions, retain binding to the monoclonal antibodies. These mutants, along with modeling studies, define the LFA-1 binding site on ICAM-1 as residues E34, K39, M64, Y66, N68, and Q73, that are predicted to lie on the CDFG beta-sheet of the Ig fold. The mutant G32A also abrogates binding to LFA-1 while retaining binding to all of the antibodies, possibly indicating a direct interaction of this residue with LFA-1. These data have allowed the generation of a highly refined model of the LFA-1 binding site of ICAM-1.

Human anti-IgE monoclonal antibody blocks passive sensitization of human and rhesus monkey bladder.

IgE antibodies may play an important role in noninfectious urinary inflammation, including interstitial cystitis (IC). In this study, urinary bladder strips were passively sensitized for 20 hours with serum from a ragweed-sensitive patient in the absence or presence of an anti-human IgE monoclonal antibody (MaE11) at 0, 1-, or 5-fold IgE concentration. The urinary bladder strips then were suspended in a superfusion apparatus for measurement of contraction and histamine release in response to antigen E (AgE) challenge. Non-sensitized tissues did not react to AgE challenge, whereas AgE challenge of passively-sensitized tissues resulted in time-dependent bladder contraction and histamine release. MaE11 abolished AgE-induced contraction and histamine release in a concentration-dependent manner. The safety of MaE11 was confirmed by its failure to contract or release histamine from passively-sensitized bladder tissues. The results of this study suggest that MaE11 may have immunotherapeutic benefit for amelioration of IgE-mediated diseases of the urinary system.