EP0988321A2

EP0988321A2 - B7-binding molecules for treating immune diseases

Info

Publication number: EP0988321A2
Application number: EP98939523A
Authority: EP
Inventors: Katrien Lorre; Erwin Sablon; Marie-Ange Buyse; Alfons Bosman
Original assignee: Innogenetics NV SA
Current assignee: Fujirebio Europe NV SA
Priority date: 1997-06-20
Filing date: 1998-06-22
Publication date: 2000-03-29
Also published as: AU8800598A; CA2292415A1; WO1998058965A3; WO1998058965A2

Abstract

The present invention relates to molecules, such as diabodies, tri- and tetravalent antibodies and small antigen binding peptides, which can cross-link or cross-react with the costimulatory molecules B7.1 and B7.2 expressed on professional antigen presenting cells (APCs) leading to the inhibition of antigen-specific T cell activation. The present invention also concerns methods to produce these molecules and use of these molecules to treat diseases, such as transplant rejection, graft versus host disease (GVHD), allergy and autoimmune diseases, where antigen-specific T cell activation is pathogenic.

Description

B7-BINDING MOLECULES FOR TREATING IMMUNE DISEASES

FIELD OF THE INVENTION

The present invention relates to molecules which can cross-link: or cross-react with the costimulatory molecules B7.1 and B7.2 expressed on professional antigen presenting cells (APCs) leading to the inhibition of antigen-specific T cell activation. The present invention also concerns methods to produce these molecules and use of these molecules to treat diseases, such as transplant rejection, graft versus host disease (GVHD), allergy and autoimmune diseases, where antigen- specific T cell activation is pathogenic.

BACKGROUND OF THE INVENTION

^* Molecules which cross-link or cross-react with B7.1 and B7.2

It is known that the use of monoclonal antibodies (mAb's) for diagnostic or therapeutic purposes in vivo is limited because of their nature (the majority are murine mAb's), their suboptimal stability and affinity and their large molecular size. In order to solve these problems several modified antibodies, antibody constructs and peptide and nonpeptide antigen binding fragments have been developed by bioengineering or chemical methods. Murine mAb's were made less antigenic to humans by CDR grafting (Winter and Harris, 1993). mAb's were made more effective by conjugating chemotherapeutic drugs and other toxins to the antibodies (Ghetie and Vitetta, 1994). Another method to produce more effective antibodies is the development of bispecific antibody constructs capable of simultaneously binding two different epitopes on the same- or different antigens. These bispecific antibodies have been produced using a variety of methods: a) antibodies of different specificities or univalent fragments of pepsin-treated antibodies of different specificities have been chemically linked (Fanger et al., 1992); b) two hybridomas secreting antibodies of different specificity have been fused and the resulting bispecific antibodies from the mixture of antibodies were subsequently isolated; c) geniticcdly engineered single chain antibodies have been used to produce non-covalently linked bispecific antibodies (e.g. diabodies (Holliger et al., 1993), minibodies (Kostelny et al., 1992; Pack et al., 1993) and tetravalent antibodies (Pack et al; 1995; WO 96/13583 to Pack) or covalently-linked bispecific antibodies (e.g. chelating recombinant antibodies (Kranz et al., 1995), single chain antibodies fused to protein A or Streptavidin (Ito and Kurosawa, 1993; Kipriyanov et al., 1996) and bispecific tetravalent antibodies (EP 0517024 to Bosslet and Deeman; Morrison, 1993). Recently, also trivalent antibody constructs, named triabodies, have been described (Kortt et al., 1997; Iliades et al., 1997). These trivalent constructs may have a higher avidity in comparison to bivalent constructs and may be useful for diagnostic or therapeutic purposes in vivo. Phage display of Ab combinatorial libraries resulting in the production of high-affinity antibodies and screening of random DNA sequence phage display libraries for small antigen-binding peptides that mimic the antigen specific binding activity of an antibody are other approaches to produce more effective antigen-binding molecules

(for review see Hayden et al., 1997 and Hoogenboom, 1997). Also the production of low-molecular-weight nonpeptide molecules, which mimic the chemical structure and biological activity of their peptide counterpart, but have longer biological half-lives and better oral bioavailability, is an attractive alternative to obtain effective antigen- binding molecules (Wiley and Rich, 1993; Wendolowski et al., 1993 and Lybrand, 1995).

Peptidic or non-peptidic molecules which are derived from 'High Throughput Screening' (HTS) of chemical or natural libraries and inhibit protein-protein interactions (Sarubbi et al., 1996) can also be used as effective antigen-binding molecules. Also the generation of hybridomas, derived from immunized transgenic mice, containing large sections of the human immunoglobulin (Ig) gene loci in the germ line, integrated by the yeast artificial chromosomal (YAC) technology, is a method to obtain effective blocking antibodies (Mendez et al, 1997). Although antibodies against B7 antigens have been described (as for example the macaque antibodies to human B7. and B7.2 described in WO96/40S78 to Anderson et al.), no prior art exists regarding the production and existence of molecules which cross-link or cross-react with B7.1 and B7.2 and which do not comprise a variable domain of a monkey antibody or the extracellular domain of the B7-bindιng molecules CTLA4 or CD28

^* T cell activation

Activation of T cells is the result of ligand-receptor interactions. The TcR/CD3 complex has two functions in antigen-induced activation: a recognition function in which a specific antigen is recognized in the context of the appropriate MHC molecule, and a signaling function in which the recognition event is transmitted across the plasma membrane (Imboden et al., 1985; Weiss and Imboden, 1987). However, T cells require a second signal to induce proliferation and maturation into effector cells. This costimulatory signal is provided by the cell surface of APCs (Springer et al., 1987). Intercellular signaling after TcR MHC-peptide interaction in the absence of the costimulatory signal results in T cell inactivation in the form of clonal anergy (Mueller et al., 1989). The interaction of a number of accessory molecules present on the cell surface of T cells with known ligands on the APC have been implicated in providing the costimulatory signal in T cell activation: CD2 with its ligand CD58 (LFA-3), CD 11 a/CD 18 (LEA- 1 ) with CD54 (ICAM- 1 ), CD28 with B7, and CD29/CD49d 5VLA-4) with VCAM-1 (Selvaro] et al., 1987; Springer, 1990; Marlin and Springer, 1987, Van Noesel et al., 1988; Iinsley et al., 1990; Damle et al., 1991).

To date, the best candidate costimulatory signal leading to full T cell activation, or to T cell anergy, is generated by interaction of CD28 on the T cells with the B7 costimulatory molecules on APCs. In viiro studies have demonstrated that signaling via the CD28 costimulatory pathway can prevent the induction of anergy Harding and coworkers ( 1992) have demonstrated that activation of mouse T cell clones with anti-

CD3 Mab in the absence of APC results in anergy However, under these conditions cross-linking of the CD28 molecule using antι-CD28 Mab could prevent anergy induction. In contrast, when the same T cell clones were stimulated with antι-CD3 Mab and competent APC, addition of Fab fragments of anti-CD28 Mabs caused T cell anergy In addition, it has been demonstrated that mouse fibroblasts co-transfected with M HC-DR-7 and human B7 1 , but not with ICAM- 1 , could prevent anergy induction of DR-7-specific αlloreαctive human T cells (Boussiotis et al., 1993a).

^* The B7 costimulatory molecules

To date, two members of the B7 family have been molecularly cloned and functionally characterized: B7.1 (CD80), originally named B7/BB1, and B7.2 (CD86), originally named B70.

B7.1 is a monomeric transmembrane glycoprotein with an apparent molecular mass of 45-65 kDa and is a member of the immunoglobulin superfamily (Freeman et al., 1989). It was initially reported that the expression of the B7.1 molecule was restricted to activated B cells (Freeman et al., 1989) and monocytes stimulated with IFN-γ (Freedman et al., 1991). More recently, B7.1 expression has also been found on cultured peripheral blood dendritic cells (Young et al. 1992) and on m vitro activated T cells (Azuma et al., 1993a). The expression of the B7.1 molecule in a number of normal and pathological tissues has been examined by iinmunohistochemistry using the anti-B7.1 Mab B7-24 (Vandenberghe et al., 1993). In addition to the staining of activated B cells, the B7.1 molecule was shown to be constitutively expressed in vivo on dendritic cells in both lymphoid and non-lymphoid tissue. Monocytes/macrophages were only found to be positive under inflaminatory conditions and endothelial cells were always negative. Interestingly, the number of B7.1 positive cells in skin lesions of patients with acute GVHD was strongly increased compared to normal skin.

B7.2 is also a transmembrane glycoprotein with an apparent molecular mass of approximately 70 kDa and is also a member of the immunoglobulin superfamily (Freeman et al., 1993; Azuma et al., 1993b). The B7.2 molecule seems to have a very similar histo-distribution pattern to B7.1 , with the exception that the induction of cell- surface expression seems to be faster (Freeman et al., 1993) and that it is present on freshly isolated monocytes (Azuma et al., 1993b). Also the expression of the B7.2 molecule was found to overlap to a large extent with B7.1, with minor differences in some B cell subsets in and around germinal centers, which most likely reflects different activation status of the B cells (de Boer, unpublished results).

In addition to the two well-characterized B7 molecules at least two other B7-like molecules have been described One shares structural homology with B7 1 since it is recognized by the BBl anti-Mab This molecule has been identified on activated _B cells and activated keratinocytes (Boussiotis et al 1993b) This molecule can bind to CD28 but not CTLA-4 (see below) However this molecule does not seem to be able to stimulate T cells via CD28 A second B7-lιke molecule has been described on EBV- transformed B cell lines and functionally interacts with CTLA-4 on T cells resulting m apoptosis of previously activated T cells (Gribben et al 1995)

^* CD28 and CTLA-4 B7-bmdιna molecules

CD28 a homodimenc transmembrane glycoprotein with an apparent molecular mass of 44 kDa and a memoer of the immunoglobulin superfamily is expressed on approximately 95% of the CD4-posιtιve T cells and 50% of the CD8 positive T cells (June et al 1990) CD28 regulates a signal transduction pathway distinct from that induced by the TcR/CD3 complex (Vandenberghe et al , 1992) Many studies indicate that costimulation of T cells by cross-linking the CD28 molecule with Mab results in a greatly enhanced activation, which is accompanied by the production of large amounts of ιnterleukm-2 (IL 2) (Thompson et al 1989, June et al 1989) and other cytokmes Furthermore anti CD28 Mabs can be replaced by B7 the natural ligand for CD28 (Lmsley et al 1991 Gimmi et al 1991 de Boer et al 1992) The B7-CD28 interaction can result in a strong proluerative (de Boer et al 1992) as well as cytolytic T cell response (Van Gool et al 1993) Ligation of CD28 can generate two signal transduction pathways a calcium independent signal which is probably the most important and a calcium-dependent signal Cyclosporm A (CsA) which is currently the most important lmmunosuppressive drug can only completely prevent T cell activation when this activation is totally dependent upon a calcium calcmeuπum mediated intracellular event (Bierer et al 1993) Consequently the activation of human T cells through the TcR m tne presence of costimulation ^ιa B7-CD28 is resistant to inhibition by CsA (June et al 1987) The CTLA-4 molecule is closely related to CD28 The amino acid sequences of both CTLA-4 and CD28 predicted from their corresponding genes v^τaιch are located on the same chromosome, show a high degree of homology, especially in the transmembrane domain and some defined regions of the extracellular domain (Harper et al., 1991). Furthermore, both CTLA-4 and CD28 bind the B7 molecules (linsley et al., 1991; Freeman et al., 1993; Azuma et al., 1993b). However, CTLA-4 has, compared to CD28, a 10 to 20-fold higher affinity for B7 (linsley et al., 1991), and, whereas CD28 is constitutively expressed at relatively high levels on T cells, CTLA-4 is only expressed at low levels on activated T cells (Linsley et al., 1992). mRNA for CTLA-4 can be detected shortly after activation of T cells, cell surface expression is only found after 2-4 days in culture (Linsley et al., 1992). The exact function of CTLA-4 has been, until recently, controversial (summarized in Science Perspective, Allison and Krummel, 1995 and Refs therein). Antibodies to CTLA-4, when co-administered with suboptimal concentrations of anti-CD28 Mab, enhance T cell proliferation. However, when cross-linked, antibodies to CTLA-4 profoundly inhibit the proliferation of naive T cells. Moreover, CTLA-4 cannot replace CD28 to provide the costimulatory signal in CD28-deficient mice. Taken together, these data suggest that blockade of

CTLA-4 binding to its ligand(s) removes inhibitory signals, whereas aggregation (i.e. cross-linking) of CTLA-4 induces inhibitory signals which down-regulate T cell responses. This hypothesis was confirmed by the phenotype of CTLA-4 deficient mice which exhibit a lymphadenopathy of extreme magnitude (Waterhouse et al., 1995). The peripheral organs of these mice contain 5-10 times the normal number of lymphocytes, the vast majority of which are activated T cells as indicated by the expression of various activation markers. The latter findings clearly support the hypothesis that the function of CTLA-4 is to switch-off T cell responses which has, as we will discuss below, important implications for the design of optimal immunosuppressive strategies based on blocking the B7-CD28 interaction.

^* Transplant rejection

Recognition of alloantigen on transplanted tissue by CD4- helper T cells leads to the induction of a variety of lymphokines. These lymphokines subsequently drive the maturation of precursor to effector cells of different immune functions. Without immune suppression this cascade of events will result m repction of the grafted tissue The current strategies for prevention of graft rejection are based on the use of broad acting lmmunosuppressive drugs such as cyclosporm and corticosteroids These drugs not only suppress the immune reaction to the alloantigen on the transplanted tissue but also suppress the immune response to bacterial or viral pathogens In addition m order for the graft to survive these drugs must often be taken over long periods of time and therefore increase the nsk of senous infections nephrotoxicity and cancer Thus the immunological system of the patient after organ transplantation is placed in a dangerously unstable balance a low immunosuppression results m transplant rejection, whereas a high immunosuppression increases the risk of fatal microbial infections Due to a better understanding of the mechanism of action of the currently used lmmunosuppressive drugs and the fine tuning of dosmg regimes, acute rejection is not a major proDlem anymore The success rate (one year survival) of for instance kidney allografts is more than 95% However despite the progress m the management of acute rejection, allograft loss after a few years due to what is called chronic vascular rejection remains a large problem for which no effective treatments are available Consequently it has become crucial to develop and evaluate alternative approaches Since donor-antigen-specific lymphocyte hyporeactivity has been demonstrated in patients without chronic rejection of a transplanted lung, in contrast to those in whom chrome rejection did develop (Remsmoen et al 1994) the optimal therapy to prevent graft rejection should be based on the mauction of specific T cell tolerance to the donor tissue Thus an ideal drug treatment should induce clonal un-responsiveness of donor-reactive T cells (anergy) without the need for long-term non specific immunosuppression and without the occurrence of enrome vascular rejection

Furthermore several m vivo models have snown that induction of Drolonged graft acceptance is possible by interruption of the B7 1/B7 2 CD28 patnwer' Treatment with CTLA-4Ig (50 μg/d) given ever/ other day for 14 davs immediately after xenogeneic human pancreatic islet transplantation in mice resulted m long term graft survival (Lenschow et al 1992) However 90% of fully mismatched rat cardiac allografts were rejected in rats treated lntrapeπtoneally (IP) with 500 ug/a CTLA-4Ig during 7 days (Turkα et αl., 1992). CTIA-4Ig (50 μg/d), given intravenously (IV) at time of transplantation and then IP every other day on days 2 through 12, prolonged cardiac allograft survival in mice, but failed to prolong the survival of primary skin grafts (Pearson et al., 1994). The discrepancies between rat and mouse cardiac transplantation models can be explained by differences in immunogenicity of allografts, the differential impact of costimulatory adhesion molecules other than B7.1/B7.2, altered pharmacokinetics, and/or changed immunogemcity and affinity of the CTLA-4Ig administered. Blockage of the CD28-pathway wih CTLA-4Ig (500 μg/d TV or IP on days 0, 1,2,3,4,6, and 8) resulted in significant prolongation of small bowel transplant survival in rats compared to controls, although all grafts were rejected after

15 days (Pescovitz et al., 1994). Finally, treatment with CTLA-4 Ig ( 100 or 250 μg/d IP for 4 weeks) could reduce lethal murine GVHD in recipients of fully allogeneic bone marrow and significantly prolonged survival rates with up to 63% of mice surviving greater than 3 months post-transplantation (Blazar et al., 1994). The failure of CTLA-4Ig alone to induce anergy m vitro and m vivo, can most likely be explained by persistent

IL-2 production, induced by TCR triggering in combination with signaling from other accessory molecules on APC.

Taken together, an ideal drug treatment to prevent transplant rejection should induce donor-specific T cell anergy and reduce the need for long term non-specific immunosuppression. In this regard, no prior art exists demonstrating that administration of a molecule which cross-links, or cross-reacts with, B7.1 and B7.2 and which does not comprise the extracellular domain of CTLA4 or CD28, can, possibly in combination with a reduced amount of lmmunosuppressive agents, prevent allograft rejection.

Graft-versus-host disease

Graft-versus-host disease (GVHD), which is a major complication during allogeneic bone marrow transplantation, is initiated by immunocompetent donor T cells which recognize alloantigens of an immunocompromised host GVHD can be classified into acute- and chronic GVHD on the basis of histologic characteristics Acute GVHD involves necrosis of the epithelium of the skm, liver and gastrointestinal tract and is clinically characterized by skin rash, jaundice and diarrhea Chronic GVHD involves fibrosis and atrophy of the same organs as for acute GVHD and may result rn complete dysfunction of these affected organs Treatment -with both CsA and methotrexate have been shown to reduce the risk of acute GVHD Treatment with prednisone, alone or in combination with azathioprrne, has been shown to effectively reverse chrome GVHD in 50-75% of patients. Furthermore, incubating the donor marrow rn vitro with anti-T-cell mAb's plus complement or anti-T-cell immunotoxms has succesfully depleted the marrow of T cells and lowered the incidence of GVHD. However, patients receivmg T cell-depleted marrow did have an increased risk of rejection, higher relapse rate of leukemia and increased risk of fungal infection which indicated that donor T cells also play a protective role Furthermore, neither lmmunosuppressive therapy nor T cell depletion specifically prevent the recognition of host alloantigens by donor T cells An approach which would specifically target this recognition would preserve the protective donor T cells and eliminate the pathogenic

T cells Recently, Gribben et al (1996) demonstrated that host alloantigen-specific anergy in human donor T cells could be induced ex vivo by using anti-B7 1 plus anti- B7 2 mAb's or a CTLA4-Ig fusion protein However, no prior art exists demonstrating that administration of a molecule which cross-links, or cross-reacts with, B7 1 and B7 2 and which does not comprise the extracellular domain of CTLA4 or CD28, can, possibly m combination with a reduced amount of lmmunosuppressive agents, more efficiently prevent the alloantigen specific activation of donor T cells and thus GVHD

Autoimmune diseases

A number of studies indicate that costimulation through CD28 ligation might be the initiating event in autoimmumty The potential of both a primary signal via the TcR and B7 1 as a costimulatory signal for the generation of autoimmune diabetes has clearly been proven with transgenic mice (Guerder et al , 1994 Harlan et al , 1994) In these studies it is hypothesized that tolerance to peripheral antigens is induced by triggering the TcR m the absence of essential costimulatory signals Mice expressmg both B7.1 and a high level of primary antigens (MHC molecules or viral glycoproterns) on pancreatic beta cells developed autoimmune diabetes. The critical role of the absence of B7.1-mediated costimulation in the induction and maintenance of tolerance to peripheral antigens, and of the B7.1 -mediated signaling in the breakdown of T cell non-responsiveness, causing autoimmunity, was obvious.

The role of the B7.1/B7.2-CD28 interaction in the chronic activation state of T cells, which have been implicated in autoimmune diseases, has been strongly suggested in various studies. Using immunohistochemical techniques, strong B7.1 expression has been found in lesions of autoimmune diseases, such as rheumatoid arthritis and psoriasis. Furthermore, it has been demonstrated that blocking

B7.1/B7.2-CD28 interaction could block auto-antibody production and prolongation of life in a murine model of autoimmune disease that closely resembles systemic lupus erythematosus in humans (Finck et al., 1994). However, no prior art exists demonstrating that administration of a molecule which cross-links, or cross-reacts with, B7.1 and B7.2 and which does not comprise the extracellular domain of CTLA4 or CD28, can, possibly in combination with a reduced amount of immunosuppressive agents, more efficiently prevent autoimmune disorders.

Allergy

Initial expression to allergen in atopic individuals results in IgE production by B cells. This IgE will first sensitize local mast cells, and 'spill-over' IgE enters the circulation and binds on receptors on circulating basophils and tissue mast cells throughout the body. The IgE production is controlled by T helper 2 (Th2) cells, as IL-4 promotes the IgE synthesis by B cells.

The role of the B7.1/B7.2-CD28 interaction in the development of an allergic reaction has been strongly suggested in various studies. It was shown by Keane et al ( 1997) that the B7-CD28/CTLA-4 costimulatory pathway is required for the development of Th2 mediated allergic airway responses to inhaled antigens. In this study, sensitized A/J mice develop significant increases in airway responsiveness, bronchioalveolar lavage eosinophils, serum IgE levels and Th2-associated cytokine production following aspiration challenge with OVA Administration of CTI-A-4 Ig m this murine model either before Ag sensitization or before pulmonary Ag challenge abolished antigen induced hyperresponsiveness and pulmonary eosrnophils The level of Th2 cytokine IL-4 and of antigen specific Ab lsotypes IgGl and IgE was significantly decreased Furthermore Knnzman et al ( 1996) showed that blockade of costimulation with CTLA-

4-Ig inhibits airway hyperresponsiveness, inflammatory infiltration expansion of thonc lymphocytes, and allergen-specific responsiveness of thorac T cells in a munne model of allergic asthma However, no prior art exists demonstrating that administration of a molecule which cross-links, or cross-reacts with, B7 1 and B7 2 and which does not comprise the extracellular domain of CTLA-4 or CD28, can, possibly in combination with a reduced amount of lmmunosuppressive agents more efficiently prevent allergic reactions

AIMS OF THE INVENTION

It is clear from the literature that blocking B7 1 alone results in partial inhibition of T cell activation Indeed, activation of T cells by alloantigen-expressing monocytes appears predominantly dependent on B7 2 costimulation Furthermore we have previously demonstrated that, durmg a mixed lymphocyte culture (MLC) with monocytes as stimulator cells anti-B7 2 mAb alone could strongly but not completely inhibit the proliferative response Only CTLA4-Ig which binds both B7 1 and B7 2 d e cross-reacts with B7 1 and B7 2) or a combination of anti B7 1 plus anu-B7 2 mAb s gave maximal inhibition

Moreover, and as already indicated above, several recent in vivo models have shown that CTLA4-Ig alone fails to induce anergy The latter finding can most likely be explained by a persistent IL-2 production mduced by TCR triggering in combination with signalling from other accessory molecules such as LFA 1 (CD 1 la/CD 18) LFA-3 (CD58) ICAlvl's (CD54 CD102 CD50) or others on -APC s Indeed signalling through the IL-2 receptor gamma chain shortly after TCR triggering is able to prevent induction of anergy (Boussiotis et al 1994 Van Gool et al 1994) This pnenomenon was illustrated m an experiment in which purified T cells were cultured for six days with αlloαntigen-expressing EBV-trαnsformed B cells. After 6 days in the presence or absence of blocking agents, cells were harvested, cultured for 2 days in plain medium without any additions and restimulated with the same EBV cell line. Addition of CTLA4- Ig or the combination of anti-B7.1 plus anti-B7.2 did induce hyporesponsiveness. However, only the addition of CsA to the B7.1/B7.2 mAb's resulted in alloantigen- specific anergy. In another experiment, the combination of anti-B7.1 plus anti-B7.2 and CsA, but not each of them separately, prevented subsequent CTL generation in a secondary MLR performed with the same EBV B cell line but in the absence of the blocking agents. These experiments clearly demonstrate the superiority of the combined B7.1/B7.2 antibodies versus CTLA4-Ig. Therefore, the present invention aims at providing a molecule which cross-links, or cross-reacts with, B7.1 and B7.2, and which does not comprise the extracellular domain of CTLA-4 or CD28.

There are several other important reasons why a molecule which can crosslink, or cross-react with, B7.1 and B7.2, and which does not comprise the extracellular domain of CTLA-4 or CD28 is a preferred therapeutic to block accessory molecule function:

1) Such a molecule can be produced in a bacterial expression system. This ensures low production and quality control costs compared to CTLA4-Ig that requires mammalian fermentation. 2) An optimal blocking agent for B7.1 and B7.2 should have a high affinity for both B7 molecules. CTLA4-Ig is a high affinity ligand for both B7 molecules, however this high affinity is due to a very high on-rate. CTLA4-Ig has also an extremely fast off-rate (Linsley et al., 1995). Since both B7 molecules are not modulated from the cell surface after ligation with CTLA4-Ig or with specific mAb's, a molecule with a very slow off-rate is the ideal blocking agent for application in vivo. In this respect, molecules which can cross-link, or cross-react with, B7.1 and B7.2, and which do not comprise the extracellular domain of CTLA-4 or CD28 prove to be good therapeutic candidate molecules, because we have means to improve the binding characteristics of mAb's via the mutagenesis of the CDR regions. In contrast, improving the binding characteristics of CTLA4-Ig is extremely difficult given the nature of the molecule.

3) There are at least 3 CTLA4-Ig counter receptors. If any of these CTLA4-Ig counter receptors can interact with CTLA-4 but not with CD28 then the usage of CTLA4-Ig may mediate deleterious effects m vivo Because CTLA4 on activated T cells functions as a terminator of T cell activation one would prefer not to block a ligand that can interact with CTLA4 but not with CD28 Such a CTLA4 ligand was demonstrated recently (Gribben et al 1995) This as yet uncharacterised CTLA4 ligand induced antigen- specific apoptosis of previously activated T cells Neither B7 1 nor B7 2 mediate apoptosis via CTLA4 Activated T cells were rechallenged with an alloantigen-bearing EBV -positive B cell line m the presence of blocking agents The simultaneous addition of mAb's to B7 1 and B7 2 but not each of them separately strongly inhibited T cell proliferation and IL-2 production and most interestingly, strongly induced apoptosis of more than 90% of the cells Therefore the present invention corns at providing a molecule which cross links or cross-reacts with B7 1 and B7 2 and which does not comprise the extracellular domain of CTLA-4 or CD28

Primatized antibodies derived from macaque monoclonal antibodies which bind B7 1 and possibly B7 2 are described in WO 96/40878 to Anderson et al

However because of the complexity and ethical constrains of working with monkeys, the present invention cams at providing in a less-complex, more-ethically acceptable and more elegant manner a molecule which cross-links or cross-reacts with B7 1 and B7 2 and which does not comprise a variable domain of a monkey antibody In this regard Zhang and Johnson (1997) have also snown that macaque B7 1 and B7 2 molecules are highly homologues (an overall am o acid homology of greater that 90%) to their human counterparts and are specifically recognized by neutralizing murine anti-human B7 1 and B7 2 monoclonal antibodies These irndings strongly suggest that macaques will develop an auto-immune response upon mjecting/immunizmg them with human B7 1 and B7 2 as described m WO 96/40878 to

Anderson et al which will result m immunosuppression In other words the anti human B7 1 and B7 2 macaque antibodies will bmd to the mon e/ s own B7 1 and B7 2 molecules which will result m tne neutralization of the monkey s B7-medιated T cell activation The present invention aims at overcoming the latter complex and ethical constrains

More particularly the present invention aims at providing a molecule which cross-links, or cross-reacts with, B7.1 and B7.2, as described above and which comprises at least one first domain which binds B7.1 or B7.2 or cross-reacts with B7.1 and B7.2, at least one second domain which binds B7.1 or B7.2 or cross-reacts with B7.1 and B7.2, and optionally a third domain which couples the first and the second domcdn(s).

More specifically, the present invention aims at providing a molecule as described above wherein said first and second domain is a low-molecular weight nonpeptide molecule or a polypeptide such as an antibody or a humanized antibody, a single chain fragment or another fragment thereof which has largely retained the specificity of scdd antibody, or a small antigen-binding peptide which is neither an antibody nor derived from an antibody, and wherein said third domain is a polypeptide, any chemical coupling agent or any oligomerization domain.

More particularly, the present invention aims at providing bispecific antibodies such as miniantibodies, diabodies and bispecific tetravalent antibodies, trivalent antibodies, bispecific small antigen-binding peptides and bispecific low molecular weight nonpeptide molecules.

Furthermore, the invention aims at providing a method to produce said molecule which cross-links, or cross-reacts with, B7.1 and B7.2, as described above.

The invention also aims at providing a composition comprising said molecule which cross-links, or cross-reacts with, B7.1 and B7.2, and which does not comprise the extracellular domain of CTLA-4 or CD28, or a variable domain of a monkey antibody, in a pharmaceutically acceptable excipient which can be used as a medicament.

The invention also aims at providing an alternative and better method to inhibit antigen-specific T cell activation and/or to treat diseases of the immune system such as allograft rejection, GVHD, allergy and autoimmune diseases by using a molecule which cross-links, or cross-reacts with, B7.1 and B7.2 as described above. Moreover, the present inventors have been able to prove that, surprisingly, cross-linking of B7.1 and B7.2 using a molecule as described above leads to the inhibition of immunoactivators such as IL- 12. This characteristic is a valuable asset in order to suppress T cell mediated immune responses. Furthermore and as we indicated above m one of our experiments, the combination of antι-B7 1 plus anti-B7 2 and CsA but not each of them separately prevented subsequent CTL generation m a secondary MLR performed with an EBV- transformed cell tine in the absence of the blocking agents Most importantly, even when the addition of the antibodies and CsA was delayed for 24 hours the generation of CTL activity after restimulation was almost completely blocked Taken together these experiments clearly demonstrate, in addition to prevent costimulation via B7 1/2 signalling, the need to interfere with the TCR signal using CsA Therefore, the present invention aims at providing a molecule which cross-links, or cross-reacts with, B7 1 and B7 2, and which does not comprise the extracellular domain of CTLA-4 or CD28 or a variable domain of a monkey antibody which possibly can be administered in conjunction with CsA or other lmmunosuppressive chemicals in order to inhibit antigen-specific T cell activation and/or to treat diseases of the immune system such as allograft rejection, GVHD, allergy and autoimmune diseases All the corns of the present invention are considered to have been met by the embodiments as set out below

BRIEF DESCRIPTION OF FIGURES

Figure 1 shows the Map of baculo transfer vector pVL-Fc The vector contains

^* a E coli origin of replication and an ampiciline resistance prokaryotic expression unit for propagation of the plasmid DNA m E coli ^*baculovrral expression casette baculovrral polydπn promoter (Ppolh) followed by a BamHI cloning site allowing upstream fusion of a cDNA encoded ORF (e g hB7- ED) to a cDNA sequence encoding the Fc region (Hιnge-CH2 CH3) of a human IgGγ l

^* baculoviral polyhedrin locus genome sequences flanking the baculovirus expression casette Figure 2 shows the protein sequence of the hB7. lFc soluble fusion protein Amino Acids 1- 34 : potential eukaryotic secretory signal peptide Amino Acids 35-241 : extracellular domain of hB7.1 protein (mature protein) Amino Acids 242- 248 : introduced by PCR cloning strategy -Amino Acids 249-480 : human IGgl-Fc (Hinge-CH2-CH3)

Figure 3 shows the protein sequence of the hB7.2Fc soluble fusion protein Amino Acids 1-16 potential eukaryotic secretory signal sequence (Azuma et al., Nature, 1993) Amino Acids 17-239 : extracellular domain of hB7.2 protein (mature protein)

Amino Acids 240-245 : introduced by PCR cloning strategy Amino Acids 246-477 : human IGgl-Fc (Hinge-CH2-CH3)

Figure 4 shows the protein sequence of the hB7.1glu-glu soluble fusion protein Amino Acids 1-34 : potential eukaryotic secretory signal peptide

Amino Acids 35-242 .-extracellular domain of hB7.1 protein (mature protein) Amino Acids 243-251 : glu-glu detection/purification tag

Figure 5 shows the protein sequence of the hB7.2his soluble fusion protein Amino Acids 1-23 : potential eukaryotic secretory signal peptide (Azuma et al., Nature, 1993)

Amino Acids 24-238 : extracellular domain of hB7.2 protein (mature protein) Amino Acids 239-244 : histidine detection/purification tag

Figure 6 shows gelfiltrσtion profile of sB7.1 gluglu (b) and MW markers (a) 1 = 670 kD,

2 = 158 kD, 3 = 44 kD, 4 = 17 kD, 5 = 1.3 kD

Figure 7 shows gelfiltration profile of sB7.2 Hisδ (b) and MW markers (a) 1 = 670 kD, 2 = 158 kD, 3 = 44 kD, 4 = 17 kD, 5 = 1.3 kD

Figure 8 shows the neutralizing activity of scFv B7-24compared to parent B7-24 mAb Figure 9 shows the PhαgemidpCES 1 antibody genes V_L-C_L variable (V) and constant (C) region of the light chain V_H-C_H1, variable and first constant region of the heavy chain PlacZ promoter rbs, nbosome binding site, S signal sequence H6 six histidines stretch for IMAC purification, tag, c-myc-deπved tag amber amber codon that allows production of soluble Fab fragments in non-suppressor strains, gill, gene encoding one of the minor coat protems of filamentous phage Restriction sites used for cloning are indicated

Figure 10 shows sequences of hybrid immunoglobulin light chcan/CTLA-4 molecule

Figure 1 1 shows the experimental strategy for 'random DNA shuffling'

Figure 12 shows sequences of the CTLA-4CDR3/Vk light chain spiking oligonucleotides

Figure 13 shows sequences of CTLA-4 CDR3/ Vλ light chain spiking oligonucleotides

Figure 14 shows a BiTAb molecule

Figure 15 shows the DNA sequence of the BιTAbB7-24- 1 G 10H6 molecule

Nucleotides 1-72 pelB signal sequence Nucleotides 73- 415 VH region anti B7 1 Mab (B7-24) Nucleotides 416 - 460 (G₄S)₃ flexible linker Nucleotides 461 -787 VL region anti B7 1 Mab (B7-24) Nucleotides 788 - 820 Human IgG3 Lunge region

Nucleotides 821 - 925 Helix-Turn Helix Dimeπsation Domain Nucleotides 926 958 Human IgG3 Lunge domain Nucleotides 959 1325 VH region anti B7 2 Mab ( 1G10) Nucleotides 1326 - 1369 (G₄S0 flexible linker Nucleotides 1370- 1708 VL region anti B7 2 Mab ( 1 G 10)

Nucleotides 1709 - 1726 Hιs6 Tag Figure 16 shows the Protein Sequence of the BiTAbB7-24-lG10H6 molecule Amino Acids 1-24: pelB signal sequence Amino Acids 25- 138: VH region anti B7.1 Mab Amino Acids 139- 153: (G₄S)₃ flexible linker Amino Acids 154 - 262: VL region anti B7.1 Mab

Amino Acids 263- 273: Human IgG3 hinge region Amino Acids 274-308: Helix-Turn-Helix Dimerisation Domain Amino Acids 309-319: Human IgG3 hinge domain Amino Acids 320-446 : VH region anti B7.2 Mab Amino Acids 447- 461: (G₄S)₃ flexible linker

Amino Acids 462- 574: VL region anti B7.2 Mab Amino Acids 575- 580: His6 Tag

Figure 17 shows DNA sequence the BiTAblG10-B7-24H6 molecule Nucleotides 1 - 360 : VH region anti B7.2 Mab ( 1G10)

Nucleotides 361 - 405 : (G₄S)₃ flexible linker Nucleotides 406 - 744 : VL region anti B7.2 Mab (1G10) Nucleotides 745 - 777 : Human IgG3 hinge region Nucleotides 778 - 882 : Helix-Turn-Helix Dimerisation Domain Nucleotides 883 - 915 : Human IgG3 hinge domain

Nucleotides 916 - 1278 : VH region anti B7.1 Mab (B7-24) Nucleotides 1279 - 1314 : (G₄S)₃ flexible linker Nucleotides 1314 - 1650 : VL region anti B7.1 Mab (B7-24) Nucleotides 1651 - 1670 : His6 Tag

Figure 18 shows Protem Sequence of the BiTAb lG10-B7-24H6 molecule Amino Acids 1 - 120 : VH region anti B7.2 Mab Amino Acids 120 - 135 : (G₄S), flexible linker Amino Acids 136 - 248 : VL region anti B7.2 Mab Amino Acids 249 - 259 : Human IgG3 hinge region

Amino Acids 260 - 285 : Helix-Turn-Helix Dimerisation Domain Amino Acids 286 - 305 : Human IgG3 hinge domain Amino Acids 306 - 426 : VH region anti B7.1 Mab Amino Acids 427 - 441 : (G₄S)₃ flexible linker Amino Acids 442 - 550 : VL region anti B7.1 Mab Amino Acids 551 - 556 : His6 Tag

Figure 19 show^s the DNA sequence of the dimerisation domain HDH Nucleotides 1 - 33 : Human IgG3 hinge region Nucleotides 34 - 82 : helix-domcrin Nucleotides 83 - 90 : turn

Nucleotides 91 - 139 :helix-domain

Nucleotides 140 - 171 :Human IgG3 hinge region

Figure 20 shows the Protein sequence of the dimerisation domain HDH Amino Acids 1 - 1 1 : Human IgG3 hinge region

Amino Acids 12 - 27 : helix-domain

Amino Acids 28 - 31 : turn

Amino Acids 32 - 46 : helix - domain

Amino Acid 47 - 57 : Human IgG3 hinge region

Figure 21 shows the DNA sequence of the dimerisation domain JEM- 1

Nucleotides 1- 99 : JEIvl-1 dimerisation domain

Figure 22 shows the Protein seσuence of the dimerisation domain JEM-1 Amino Acids 1- 33 : JEM-1 dimerisation domain

Figure 23 shows DNA sequence of monospecific Diabody B7-24: VH-B7-24/5/VL-B7- 24/H6

Nucleotides 1-72 : pelB signal sequence Nucleotides 73 - 414 : VH region anti B7.1 Mab (B7-24)

Nucleotides 415 -429 : G₄S flexible linker Nucleotides 430 - 756 : VL region αnti B7.1 Mob (B7-24) Nucleotides 757 - 773 : His6 Tag

Figure 24 shows Protein Sequence of monospecific Diabody B7-24: VH-B7-24/5/VL- B7-24/H6

Amino Acids 1- 24: pelB signal sequence Amino Acids 25 - 138 : VH region anti B7.1 Mab Amino Acids 139 - 143 : G₄S flexible linker Amino Acids 144 - 252 : VL region anti B7.1 Mab Amino Acids 253 - 259: His6 Tag

Figure 25 shows DNA sequence of monospecific Diabody 1G10: VH-1G10/5/VL- 1G10/H6

Nucleotides 1-72 : pelB signal seσuence Nucleotides 73 - 433 : VH region anti B7.2 Mab (1G10)

Nucleotides 434- 447 : G₄S flexible linker Nucleotides 448 - 786 : VL region anti B7.2 Mab (1G10) Nucleotides 787 - 804 : His6 Tag

Figure 26 shows Protein Seσuence of monospecific Diabody 1G10: VH- 1 G 10/5/VL-

1G10/H6

Amino Acids 1 - 24 : pelB signal sequence Amino Acids 25 -144 : VH region anti B7.2 Mab Amino Acids 145 - 149 : G₄S flexible linker Amino Acids 150 - 262 : VL region anti B7.2 Mab

Amino Acids 263 - 268 : His6 Tag

Figure 27 shows DNA seσuence of bispecific Diabody I : VH-1G 10/5 VL-B7-24/H6 Nucleotides 1-1 17 : g3p - signal sequence Nucleotides 1 18 - 483 : VH region anti B7.2 Mab ( 1G10)

Nucleotides 484 - 498 : G,S flexible linker Nucleotides 499 - 825 VL region αnti B7 1 Mob (B7 24) Nucleotides 826 - 843 Hιs6 Tag

Figure 28 shows the Protem sequence of bispecific Diabody I VH 1 G 10/5/VL-B7- 24/H6

Ammo Acids 1- 39 g3p - signal sequence Ammo Acids 40 - 161 VH region anti B7 2 Mab Ammo Acids 162 - 166 G₄S flexible linker Ammo Acids 167 - 275 VL region anti B7 1 Mab Ammo Acids 276 - 281 Hιs6 Tag

Figure 29 shows the DNA sequence of bispecific Diabody II VH B7-24/5/VL IGl 0 Nucleotides 1 1 17 g3p signal seσuence Nucleotides 1 18 - 465 VH region anti B7 1 Mab (B7-24) Nucleotides 466 - 480 G₄S flexible linker

Nucleotides 481 - 819 VL region anti B7 2 Mab (IGl 0)

Figure 30 shows the Protem seσuence of bispecific Diabody II VH-B7-24/5/VL-1G10 Amino Acids 1 - 39 g3p - signal sequence Amnno Acids 40 - 155 VH region anti B7 1 Mab

.Ammo Acids 156 160 G₄S flexible linker Ammo Acids 161 273 VL region anti B7 2 Mab

Figure 31 shows tne DNA seσuence of monospecific Tπabody B7 24 VH-B7- 24/0/VL B7-24H6

Nucleotides 1-72 pelB signal seσuence Nucleotides 73 - 414 VH region anti B7 1 Mab (B7-24) Nucleotides 415 - 741 VL region anti B7 1 Mab (B7 24) Nucleotides 742 - 759 Hιs6 Tag

Figure 32 shows the Protein sequence of monospecific Tπabody B7-24 VH-B7 24/0NL-B7-24H6

Amino Acids 1- 24 : pelB signal seσuence Amino Acids 25 - 138 : VH region anti B7.1 Mab Amino Acids 139 - 247 : VL region anti B7.1 Mab Amino Acids 248 - 253 : His6 Tag

Figure 33 shows the DNA sequence of monospecific Triabody 1G10 : VH-1G10/0/VL- 1G10H6

Nucleotides 1-72 : pelB signal sequence Nucleotides 72 - 433 : VH region anti B7.2 Mab (1G10)

Nucleotides 434 - 771 : VL region anti B7.2 Mab ( 1G10) Nucleotides 772 - 789 : His6 Tag

Figure 34 shows the Protein sequence of monospecific Triabody 1G10 : VH- 1G10/0/VL-1G10H6

Amino Acids 1 - 24 : pelB signal seσuence

Amino Acids 25 - 144 : VH region anti B7.2 Mab

Amino Acids 145 - 257 : VL region anti B7.2 Mab

Amino Acids 258 - 263 : His6 Tag

Figure 35 shows the DNA sequence of bispecific Triabody I : VH- 1 G 10/0/VL-B7-

24/H6

Nucleotides 1-1 17 : g3p - signal sequence

Nucleotides 1 18 - 483 : VH region anti B7.2 Mab (1G10) Nucleotides 484 - 810 : VL region anti B7.1 Mab (B7-24)

Nucleotides 81 1 - 828 : Hiso Tag

Figure 36 shows the Protein sequence of bispecific Triabody I : VH- 1G10/0/VL-B7- 24/H6 Amino Acids 1- 39 : g3p - signal seσuence

Amino Acids 40 - 161: VH region anti B7.2 Mab Ammo Acids 162 - 270 VL region αnti B7 1 Mob Ammo Acids 271 - 276 Hιs6 Tag

Figure 37 shows the DNA sequence of bispecific Triabody II VH B7-24/0A^L-lG10 Nucleotides 1-1 17 g3p - signal sequence

Nucleotides 1 18 - 465 VH region anti B7 1 Mab (B7-24) Nucleotides 466 - 804 VL region anti B7 2 Mab (1G10)

Figure 38 shows the Protem sequence of bispecϋicTπabody II VH-B7-24/0/VL- 1 G 10 Ammo Acids 1- 39 g3p - signal sequence

Ammo Acids 40 - 155 VH region anti B7 1 Mab Ammo Acids 156 - 268 VL region anti B7 2 Mab

Figure 39 shows the geliiltration profile of scFv B7-24 LO (a) scFV B7-24 L5 (b) and scFv B7-24 L15 (c)

Figure 40 shows the bmdmg of unpurified scFv B7-24, B7-24 diabodies (scFv B7-24 L5)- and B7-24 tnabodies (scFV B7-24 L0) on B7 1 ED fusion protein

Figure 41 shows the bmdmg of unpurified scFv 1 G 10 1 G 10 diabodies (scFv 1 G 10

L5) and 1G10 tnabodies (scFv 1G10L0) on B7 2ED fusion proteins

Figure 42 shows bmdmg of unpurified scFv B7-24 B7-24 diabodies (scFv B7-24 L5)- and B7-24 tnabodies (scFV B7-24 L0) on different 3T6 cells

Figure 43 shows binding of semi-puπfied scFv B7 24 B7 24 diabodies (scFv B7 24 L5)- and B7-24 tnabodies (scFV B7-24 L0) on B7 1 ED fusion protein

Figure 44 shows Joindrng of semi-purified scFv B7-24 B7-24 diabodies (scF¹" B7-24 L5)- and B7-24 tnabodies (scFV B7-24 L0) on RPMI8866 cells Figure 45 shows the neutralizing activity of scFv B7-24 and B7-24 diabodies (scFv

Figure 46 shows the neutralizing activity of scFv B7-24 and B7-24 tnabodies (scFV B7-24 LO) a MLR

Figure 47 shows bmdmg of B7-24 diabodies (scFv B7-24 L5)- and B7-24 tnabodies (scFV B7-24 LO) gelfiltration fractions on B7 1ED fusion protem

Figure 48 shows the gelfiltration of B7-24/1G 10 crosslinked monoclonal antibodies

Figure 49 shows the bmdmg of the crosshnked monoclonal antibodies on 3T6 cells

Figure 50 shows the Biacore results of the crosshnked b7-24/lG10 monoclonal antibodies

DETAILED DESCRIPTION OF THE INVENTION

The invention described herein draws on previously published work and pending patent applications By way of example such work consists of scientific papers patents or pending patent applications All of these publications and applications cited previously or below are hereby incorporated by reference

The present invention is based on the finding that molecules which crosslink, or cross-react with, B7 1 and B7 2, and which do not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28 efficiently inhibit antigen specific T cell activation Accordingly these molecules can be used to prevent or treat (terms used interchangeably) diseases such as transplant rejection GVHD allergy and autoimmune diseases where antigen- specific T cell activation is pathogenic More particularb the present invention relates to a molecule which crosslinks or cross reacts with B7 1 and B7 2 and which does not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28 resulting in the inhibition of antigen-specific T cell activation

The expression "molecule which cross-reacts with B7 1 and B7 2 and which does not comprise the extracellular domain of CTLA-4 or CD28" refers to any molecule known m the art which simultaneously binds (not necessarily to a common epitope of B 1 and B7 2) or binds to a common epitope of B7 1 (decπbed by Freeman et al , 1989) and B7 2 (described by Freeman et al , 1993 and Azuma et al , 1993) and which does not comprise the extracellular domain of CTLA-4 or CD28 Molecules compnsmg the extracellular domains of CTLA-4 or CD28, such as CTLA-4Ig or CD28Ig respectively, are known to be able to cross-link, or crossreact with B7 1 and B7 2 (described m US patent N^c 5 434, 131 and 5 521,288 to linsley et al respectively) However we already indicated above why it is preferable not to use CTLA-4Ig as a therapeutic Also CD28Ig was proven to be a very poor inhibitor of T-cell activation Furthermore, the term ' a molecule which cross-finks B7 1 and B7 2" indicates that both B7 1 and B7 2 are physically bridged or connected by scad molecule It should also be clear that both B7 molecules which are connected by scad molecule can be expressed on the same cell or on different cells In this regard molecules with a rigid structure such as diabodies, tnabodies, small antigen bmdmg peptides and low-molecular weight nonpeptide molecules will crosslink B7 molecules on different cells and may not be able to crosslink B7 molecules expressed on the same cell whereas molecules with a flexible structure such as tetravalent antibodies will crosslink B7 molecules expressed on both the same and different cells Furthermore the expression ' crosshnking' may also imply that the B7 molecules on the B7-expressιng cells are not only physically bridged but that the cross-linking of both B7 molecules results m a biological effect on the B7-expressrng cells The latter biological efiect may comprise the inhibition of synthesis of lmmunoacuvating soluble mediators such as lnterleukin- 12 (IL-12) ιnterleukιn-6 (IL-6) ιnterleukιn-1 (IL- 1 ) and tumor necrosis factor-alpha (TNF- ) by the B7-expressrng cell or the activation of synthesis of lmmunosuppressive mediators such as mterleukin 10 (IL- 10) tumor growth factor-beta (TGF-β) and prostaglandms by the B7-expressmg cell or any other known biological effect which, preferably, favours the inhibition of T cell activation. The terms "a variable domain of a monkey antibody" specifically refer to the variable domains of the macaque monoclonal antibodies 7B6, 16C10, 7C10 and 20C9 described in WO 96/40878 (PCT/US96/10053) to Anderson et al. The latter reference is incorporated by reference in its entirety herein.

More specifically, the invention relates to a molecule which comprises at least one first domain which binds B7.1 or B7.2 or cross-reacts with B7.1 and B7.2, at least one second domain which binds B7.1 or B7.2 or cross-reacts with B7.1 and B7.2, and optionally a third domain which couples the first and the second domcrin(s). The term domain refers to any antibody-like polypeptide, any nonantibody-like peptide or any nonpeptide organic molecule. Examples of such molecules are diabodies which bind in a monovalent fashion to B7.1 and B7.2 or which bind in a bivalent fashion to a common epitope of B7.1 and B7.2 (hereafter termed B7.12), triabodies which bind simultaneously in a monovalent fashion to B7.1 and bivalently to B7.2 or in a monovalent fashion to B7.2 and bivalently to B7.1 or trivalently to B7.12, tetravalent antibodies which bind bivalently to B7.1 and B7.2 or tetravalently to B7.12, and small antigen binding peptides or low molecular weight nonpeptide molecules which bind mono-or multivalently to B7.1, B7.2 and/or B7.12. It should be clear that any other possible combination, for example a tetravalent antibody which binds bivalently to B7.12 and bivalently to B7.2, is also part of the present invention.

As used herein, the term "any chemical couphng agent or any oligomerization domain" refers to any molecule known in the art which is capable of coupling the said first and second domains to each other. Examples of such domains are the known leucine zipper of c-fos and c-jun (Kostelny et al, 1992; de

Kruii & Logtenberg, 1996), the polyglutamic acid-polylysine domains as described in US Patent N⁼ 5,582,996 to Curtis, the helix-turn-helix motif described by Pack et al. ( 1993) and the max -interacting proteins and related molecules as described in US Patent N^c 5,512473 to Brent and Zervos. The term "optionally" as used in the terms "optionally a third domain which couples the first and the second domcrin(s)" indicates that a third domain can, but does not have to, be part of a molecule of the present invention as described above

As used here , the term 'antibody' refers to polyclonal or monoclonal antibodies The term "monoclonal antibody' refers to an antibody composition having a homogeneous antibody population The term is not hmited regarding the species or source of the antibody nor is it intended to be limited by the manner m which it is made

As used here the term "humanized antibody" means that at least a portion of the framework regions of an lmmunoglobuhn are derived from human immunoglobulin sequences As used herem the term 'single chain antibody" refers to antibodies prepared by determining the bmdmg domains (both heavy and tight chains) of a bmdmg antiboav and supplymg a hnkmg moiety which permits preservation of the binding iunction This forms in essence a radically abbreviated antibody having only that part of the variable domain necessary for bmdmg the antigen Determination and construction of single chain antibodies are described in U S

Patent No 4 946 778 to Ladner et al

As used herem, the term "fragments (of antibodies)" refers to F_ab F_(ab)2, F and other fragments which retain the antigen bmdmg function and specificity of the parent antibody Antibodies to human B7 1 and human B7 2 are known in the art The present invention contemplates a new use lor such antibodies as detailed above

Monoclonal antibody B7 24 was prepared as described m the international application WO 94/01547

Monoclonal antibodies 5B5 and 1G10 were prepared essentially as described m U S Patent No 5 397 703 or international application WO 94/01547 and can be obtained at Innogeneticε N V Industπepark Zv jnaarde 7 box 4 B

9052 Ghent Belgium fax ^32 9 241 07 99

Monoclonal antibodies 3H 10 5F3 7B8 9D8 1 1B9 13B9 13D3 and Fl were prepared as described in example 1 (see further) and can be obtained at Innogeneϋcs N V Industπepark Zwijnaarde 7 box 4 B 9052 Ghent Belgium fax

-32 9 241 07 99 The present invention more specifically relates to miniantibodies, diabodies, triabodies, tetravalent antibodies, antigen-binding peptides and low molecular weight nonpeptide molecules which cross-link, or cross-react with, B7.1 and B7.2, and do not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28, and which are produced by the following methods:

1 ) chemical linkage of anti-B7.1 and anti-B7.2 antibodies or univalent fragments thereof foU owing a method as described by Fanger et al. (1992). See also the Examples section further in the present application.

2) geneticcdly engineering of non-covalently-linked miniantibodies as described by Pack et al. (1993), diabodies as described by Holhger et al. ( 1993) and tetravalent antibodies as described by Pack et al. (1995). See also the Examples section further in the present application.

3) genetically engineering of covalently-linked chelating recombinant antibodies as described by Kranz et al. (1995), single chain antibodies fused to protein A or Streptavidin as described by Ito and Kurosawa (1993) and Kipriyanov et al. (1996) and bispecific tetravalent antibodies as described in EP 0 517 024 to Bosslet and Seeman, and Coloma and Morrison (1997). See also the Examples section further in the present application.

4) genetically engineering of triabodies as described by Kortt et al ( 1997) and as given in the Examples section of the present application.

5) phage display of Ab combinatorial libraries resulting in the production of high- affinity antibodies and screening of random DNA sequence phage display libraries for small antigen-binding peptides as described in US patent numbers 5,403,484 and 5,571,698 and 5,223,409 to Ladner et al., Schultz and Schultz ( 1996), Parsons et al. (1996), McGuinness et al. (1996) and Hoogenboom (1997) and Georgiou et al.

( 1997). See also the Examples section further in the present application.

6) generation of hybridomas, derived from immunized transgenic mice, containing large sections of the human immunoglobulin (Ig) gene loci in the germ line, integrated by the yeast artificial chromosomal (YAC) technology, resulting in effective blocking antibodies as described by Mendez et al ( 1997).

7) rational drug design resulting in the production of low-molecular weight nonpeptide molecules as described by Wiley and Rich ( 1993), Wendolowski et al. ( 1993) and Lybrand (1995).

8) 'High Throughput Screening' (HTS) of chemical or natural libraries, resulting in the production of peptides or non-peptides as described by Sarubbi et al. ( 1996). The latter references are incorporated in their entirety herein.

As used herein, the term "small antigen-binding peptides or fragments" refers to any peptide (i.e. a polymer composed of at least two amino acids) which cross-links, or cross-reacts with, B7.1 and B7.2, and which does not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28. The term "low-molecular weight nonpeptide molecules" refers to any molecule which is not a peptide and which cross-links, or cross-reacts with, B7.1 and B7.2, and which does not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28.

The present invention further relates to a composition comprising a molecule which cross-links, or cross-reacts with, B7.1 and B7.2, and which does not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28 in a pharmaceutically acceptable excipient, possibly in combination with immunosuppressive agents such as cyclosporin A (Sandimmune, Neoral; Sandoz-Sangstat), FK 506 (Tacrolimus, Prograf; Fujisawa), rapamycin (Sirolimus; American Home Products), OKT-3 (anti-CD3 mAb; Johnson & Johnson),

OKT-4 (anti-CD4 mAb;Johnson & Johnson), SB-210396 (anti-CD4 mAb; Smimkline Beecham), T10B9 (anti-TcR antibody; Medlmmune), BTI 322 (anti-CD2 mAb; Biotransplant), Mycophenolate mofetil (Cellcept; Roche), anti-thymocyte immunoglobulin (Thymoglobulin (rabbit); Pasteur Merieux), anti-lymphocyte immunoglobulin (Lymphoglobulin (equine); Pasteur Merieux), anti-lymphocyte immunoglobulin (ATG Fresenius (rabbit); Hoechst Marion Roussel), azathioprine (Imuran; Glaxo Wellcome), leflunomide (Hoechst Marion Roussel), triple therapy combining cyclosporin A (Sandimmune, Neoral; Sandoz-Sangstat) with azathioprine (Imuran; Glaxo Wellcome) and glucocorticosteroϊds, adenosin deaminase inhibitor (Pentostatin; Warner-Lambert), purine nucleoside phosphorylase (PNP) inhibitor (Peldistine; Biocryst Pharmaceuticals), MHC-peptide (AUotrαp-2702 Sαngstαt) and IL-2 receptor mAb (Pasteur-Meneux Lederle/Amencan Home products Protem Design Labs, for use as a medicament to prevent allograft rejection, GVHD allergy and autoimmune diseases

As used herem, the term "composition" refers to any composition compnsmg as an active mgredient a molecule which cross-hnks, or cross-reacts with B7 1 and

B7 2, and which does not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28 according to the present invention possibly m the presence of suitable excipients known to the skilled man The molecule which cross-hnks, or cross-reacts with, B7 1 and B7 2, and which does not comprise a variable domain of a monkey antibody or the extracellular domain of

CTLA-4 or CD28 of the invention may thus be administered m the form of any suitable composition as detailed below by any suitable method of administration within the knowledge of a skilled man The preferred route of administration is parenterally In parenteral administration, the compositions of this invention will be formulated in a unit dosage rnjectable form such as a solution suspension or emulsion m association with a pharmaceutically acceptable excipient Such excipients are inherently nontoxic and nontherapeutic Examples of such excipients are saline, Ringer's solution, dextrose solution and Hank's solution Nonaqueous excipients such as fixed oils and ethyl oleate may also be used A preferred excipient is 5% dextrose rn saline The excipient may contain minor amounts of additives such as substances that enhance lsotomcity and chemical stability including buffers and preservatives

The molecule which cross-hnks, or cross-reacts with, B7 1 and B7 2 and which does not comprise a variable domain of a monkey anubody or the extracellular domain of CTLA-4 or CD28 of the invention are administered at a concentration that is therapeutically effective to prevent allograft rejection GVHD allergy and autoimmune diseases The dosage and mode of administration will depend on the individual Generally the compositions are administered so that the molecule cross-lmks or cross-reacts with B7 1 and B7 2 and which does not comprise a variable domain of a monkey antibody or the extracellular domain of

CTLA-4 or CD28 is given at a aose between 1 μg/kg and 10 mg/kg more preferably between 10 μg/kg and 5 mg/kg most preferably between 0 1 and 2 mg/kg Preferably, it is given as a bolus dose Contmuous short time infusion (durmg 30 minutes) may also be used If so, the molecule which cross-links or cross-reacts with, B7 1 and B7 2 and which does not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28 constructs or compositions compnsmg the same may be infused at a dose between 5 and 20 μg/kg/minute, more preferably between 7 and 15 μg/kg/mmute

According to the specific case, the "therapeutic ally effective amount" of a molecule which cross-hnks or cross-reacts with, B7 1 and B7 2, and which does not comprise a variable domain of a monkey antibody or the extracellular domain of

CTLA-4 or CD28 needed should be determined as bemg the amount sufficient to cure the patient in need of treatment or at least to partially arrest the disease and its complications Amounts effective for such use will depend on the seventy of the disease and the general state of the patient's health Single or multiple administrations may be required depending on the dosage and frequency as required and tolerated by the patient

With regard to the use of a molecule which cross-hnks or cross-reacts with, B7 1 and B7 2, and which does not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA-4 or CD28 to prevent allograft rejection it should be stressed that molecules of the present invention or the compositions compnsmg the same may be administerea before during or after the organ transplantation as is desired from case to case In case the molecules or the compositions comprising the same are administered directly to the host treatment will preferably start at the time of the transplantation and continue afterwards m order to prevent the activation and differentiation of host T cells against the MHC on the allograft In case the donor organ is ex vivo peπused with molecules or the compositions comprising the same treatment of the donor organ ex vivo will start before the time of the transplantation of the donor organ m order to prevent the activation and differentiation of host T cells agarnst the MHC on the allograft The present invention will now be illustrated py reference to the following examples which set forth particularly advantageous embodiments However it should be noted that these embodiments are illustrative and are not to be construed as restricting the invention in any way.

EXAMPLES

1. Making monoclonal Antibodies to B7.1 and B7.2

1.1 Making monoclonal Antibodies to B7.1

One female BALB/c mice was immunized (injected intraperitonec ly) four times (i.e., at days 0, 13, 26 and 213) with 5x10^C Sf9 insect cells that were infected with a recombinant baculovirus containing a human B7.1 cDNA. Three days after the last injection, spleen cells were retrieved from the immunized mice and used for cell fusion mainly according the procedure as described by Kόhler and Milstein ( 1975). To this end, dissociated splenocytes from the immunized mice were fused with murine myeloma cells SP2/0-Agl4 (ATCC, CRL-1581) at a ratio of 10:3 using a polyethylene glycol/DMSO solution. The fused cells were mixed up and resuspended in DMEM medium supplemented with hypoxanthine, thymidine, sodium pyruvate, glutamine, a non-essential amino acid solution, 10% inactivated fetal calf serum, 10% inactivated horse serum and 10% BM-Condimed. The cells were then distributed to 960 wells on tissue culture plates to which aminopterin was added 24 hours after the cell fusion. Each well contained between 1 to 5 growing hybridoma clones at the average. After eight days supernatants of the 960 wells were combined in groups of 10 for primary screening. The 96 pools were screened for the presence of specific antibody by FACS analysis using the B7.1 positive

Epstein-Barr virus (ΞBV)-transformed human B cell fine, ARC . For this, cells (5xl0⁵ cells/sample) were incubated for 30' at 4^~C with the different (100 ml undiluted) supernatants. .After washing of the cells with PBS supplemented with 2% inactivated FCS ands 0.02% azide, the cells were incubated another 30' at 4⁼C with goat anti-mouse antibodies conjugated to fluoresceine isothiocyanate (FTTC). Cells were washed twice in PBS supplemented with 2% inactivated FCS ands 0.02% αzide, fixed using 0.5% pαrαformαldehyde and analyzed with a FACScan flow cytometer (Becton Dickinson). The specific binding of the monoclonal antibodies is expressed as the mean fluorescence intensity in arbitrary units. The first screening yielded 3 positive pools. The thirty wells corresponding to the three positive pools were subjected to a second screening with the FACS screening technique described above using ARC cells (2.5x10⁵ cells/sample). This second screening provided one individual positive well containing antibodies reactive with the B7.1- expressing ARC cells This positive well was subcloned and one stable hybridoma clone named 5B5 was obtained. This hybridoma clone secrete mouse antibodies of the IgG3 isotype.

The antibody secreted by the hybridoma clone 5B5 was tested for specific binding to the human B7.1 molecule in a FACS experiment using the B7.1 -expressing ARC cells or the mouse fibroblasts, 3T6 cells, transfected with cDNA encoding human B7.1 molecule (3T6-B7.1 cellsXDe boer et al., 1992). As control non-transfected 3T6 cells were used. Cells (2.5x10⁵ cells/sample) were incubated with the supernatant of the hybridoma clone 5B5 for 30 min. at 4°C. Thereafter, the cells were washed (PBS supplemented with 2% inactivated FCS ands 0.02% azide) three times and incubated with FTTC-labeled goat anti-mouse antiserum (i.e. GAM-FTTC). The cells were also incubated with the GAM-FTTC alone. After another 3 washes, the cells were analysed for fluorescent staining using a FACScan instument. Results showed that the antibodies secreted by the 5B5 hybridoma clones specifically bound to the B7.1 -expressing ARC and 3T6/B7.1 cells whereas 3T6 cells that do not express B7.1, did not exhibit any binding significantly greater than that of the control monoclonal antibody or GAM-FTTC. Moreover, specific binding of the antibodies secreted by 5B5 hybridoma clones was further analyzed in a competition experiment. To this end, ARC cells (5x10⁵ ceMs/stcdning) were incubated with the supernatant of the 5B5 hybridoma clones together with 1 mg of biotinylated anti-B7.1 monoclonal antibody (B7-24) for 30' at 4^"C. Thereafter, cells were separated from the supernatant, washed and incubated for 30' at 4°C with FTTC-labeled streptavidin. Results showed that complete inhibition of binding of the anti-B7.1 monoclonal antibody (B7-24) could be obtained with 5B5 hybridoma clones.

Moreover, the antibodies, secreted by the 5B5 hybridoma clone were tested for their capacity to inhibit the proliferation of human peripheral blood T lymphocytes, activated with anti-CD3 (OKT-3) in the presence of the mouse 3T6 cells transfected with the cDNA encoding human B7.1 molecule and human CD32 (3T6/CD32/B7.1 )

(deBoer et al., 1992). Human peripheral blood T lymphocytes were isolated from buffy coat by density centriiugation.T cells were further purified by cold aggregation of the monocytes (4X 10⁶ PBMC/ml in RPMI bic + 10 % iFCS, 30' rotation at 4 °C; 15' on ice to separate the monocyte aggregates; supernatant contains the enriched T cells). T cells were further enriched by depletion of monocytes, B cells and NK cells using Lympho-Kwik T (One Lambda, Los Angeles, CA) according to the manufacturers protocol. Mitomycin C-treated 3T6/CD32/B7.1 ceUs(at 200μl Mitomycin C at 250μg/ml to 800 μl RPMI bic + 10 % iFCS during 45' at 37³C, washed twice in RPMI bic + 10 % iFCS) at 10⁴ cells /well were incubated with anti-CD3 mAb (OKT3, 0.5 mg/ml) for 1 h at 37°C fohowed by a 1 h incubation at 37°C with decreasing concentrations of the antibodies secreted by the 5B5 hybridoma clone. Subsequently purified T cells (5xl0⁴ cells/well) were added and incubated for 5 days. After 5 days of culture, the cells were pulsed for 6 to 8 h with 1 μCi [³H]- Thymidine, after which the cells were harvested using an automated cell harvester. [³H] -Thymidine incorporation was determined with a liquid scintillation counter.

Proliferation of T ceUs was performed in triplicate wells. Results showed that the proliferation of the T cells was inhibited by the antibodies secreted by the 5B5 hybridoma clones. Thus, the antibodies secreted by the 5B5 hybridoma clone are neutralizing antibodies.

J .2. Making monoclonal Antibodies to B7.2

1.2.1 Fusion 1

Two female BALB/c mice were immunized (injected intraperitoneally) four times (i.e., at days 0, 28, 56 and 208) with Sf9 insect cells that were infected with a recombinant baculovirus containing a human B7.2 cDNA. Three days after the last injection, spleen cells were retrieved from the immunized mice and used for cell fusion mainly according to the procedure as described by Kohler and Milstein ( 1975). Dissociated splenocytes from the immunized mice were fused with SP2/0 murine myeloma cells at a ratio of 10:3 using a polyethylene glycol/DMSO solution. The fused cells, derived from the two mice, were mixed up and resuspended in

DMEM medium supplemented with hypoxanthine, thymidine, sodium pyruvate, glutamine, a non-essential amino acid solution, 10% inactivated fetal calf serum and 10% inactivated horse serum. The cells were then distributed to 1440 wells on tissue culture plates to which arriinopterin was added 24 hours after the cell fusion. Each well contained between 1 to 5 growing hybridoma clones at the average.

After ten days, supernatants from the 1440 primary wells were combined in groups of five to form 288 pools for primary screening. The 288 pools were screened for the presence of specific antibody by FACS analysis using the B7.2 positive Epstein- Barr virus (EBV) transformed human B cell line, RPMI 8866.For this, cells (0.5- lx 10⁵ cells/sample) were incubated for 15' at 4°C with the different supernanants ( 100 μl undiluted). After washing twice in RPMI 1640 supplemented with 10% FCS, the cells were incubated for another 15' at 4°C with goat anti-mouse antibodies conjugated to fluorescein isothiocyanate (FTTC). The cells were washed twice in RPMI 1640 supplemented with 10% FCS and finally suspended in PBS supplemented with 1% BSA and 0.1 % NaN₃ and analyzed with a FACScan flow- cytometer (Becton Dickinson).

The specific binding of the monoclonal antibodies is expressed as the mean fluorescent intensity in arbitrary units. This first screening yielded eighteen positive pools. The ninety wells corresponding to the eighteen positive pools were subjected to a second screening, with the FACS screening technique described above using the B7.2-expressing human EBV-transformed B cell line RPMI 6688.

This second screening provided eleven individual positive wells containing antibodies reactive with the 37.2-expressing EBV-transformed human B cell fine RPMI 8866. These positive wells were subcloned and eight stable hybridoma clones named 3H10, 5F3, 7B8, 9D8, 1 1B9, 13B9, 13D3, 14F1 were obtained. These hybridoma clones secrete mouse antibodies of different IgG isotypes: IgGl :9D8;

IgG2a: 5F3, 7B8, 14F1; IgG2b: 3H10, 1 1B9, 13B9, 13D3. The antibodies secreted by hybridoma clones 3H 10 5F3, 7B8 9D8 1 1B9 13B9, 13D3, 14F1 were tested for specific bmdmg to the human B7 2 molecule B7 2- expressing EBV-transformed human B cells (RPMI 8866), freshly isolated peripheral blood human T cells and monocytes were incubated with the supernatant of the different hybridoma clones 3H 10, 5F3, 7B8, 9D8 11B9, 13B9 13D3 14Fl or an isotype matched control monoclonal antibody (control mAb) for 30 mm at 4°C Thereafter, the cehs were separated from the supernatant, washed three times and incubated with FTTC-labeled goat anti-mouse IgG antiserum d e GAM-FTTC) The cehs were also incubated with the GAM-FTTC alone After another 3 washes, the cehs were analysed for fluorescent staining usmg a FACScan mstument Results showed that the antibodies secreted by the 3H10, 5F3, 7B8, 9D8 1 1B9 13B9, 13D3, 14F1 hybridoma clones specifically bound to the B7 2-expressmg RPMI 8866 cehs and the B7 2 expressmg human monocytes whereas freshly isolated human peripheral blood T cehs that do not express B7 2, did not exhibit any bmdmg significantly greater than that of the control monoclonal antibody or GAM-FTTC In another experiment, mouse 3T6 cehs transfected with the cDNA encoding human B7 1 molecule (3T6-B7 1) or encoding the human B7 2 molecule (3T6-B7.2) (de Boer et al , 1992) were used to demonstrate that the antibodies secreted by hybridoma clones 3H10, 5F3, 7B8, 9D8, 1 1B9, 13B9, 13D3, 14F1 are specific for B7 2 Results showed that the antibodies secreted by the hybridoma clones 3H10 5F3 7B8, 9D8

1 1B9 13B9 13D3 14F1 specifically bound to the B7 2-expressmg 3T6 cehs (3T6- B7 2) whereas 3T6 cehs expressmg the B7 1 (3T6-B7 1) molecule did not exhibit any bmdmg significantly greater than that of the GAM FITC Moreover the antibodies secreted by the different hybridoma clones 3H 10, 5F3 7B8 9D8 1 1B9 13B9 13D3, 14F1, were tested for their capacity to inhibit the proliferation of human peripheral blood T lymphocytes activated vnth anu-CD3 (OKT-3) the presence of the B7 2 expressmg EBV transformed B cell line RPMI 8866 Human peripheral blood T lymphocytes were isolated irom buffi, coat by density centrifugation (Ficoll Pacque density 1 077) T cells were further purified by cold aggragation of the monocytes (4X 10^l PBMC/mhn RPMI bic - 10 % iFCS 30' rotation at 4 C 15 on ice to separate the monocvte aggregates supernatant contains the enriched T cells) T cehs were further enriched by depletion of monocytes, B cehs and NK cells usmg Lympho- Kwik T (One Lambda, Los Angeles, CA) according to the manufacturers protocol The purified T cehs proliferated in the presence of anti-CD3 mAb (OKT-3, 0 5μg/ml) and Mitomycme C treated RPMI8866 cehs (at 200μl Mytomicine C at 250μg/ml to 800 μl RPMI bic -r 10 iFCS durmg 45' at 37°C, wash twice in RPMI bic -^>- 10 % iFCS) Decreasing concentrations of the antibodies, secreted by the 3H10 5F3, 7B8, 9D8, 1 1B9, 13B9, 13D3, 14F1 hybridoma clones were added to the T ceh cultures. After 5 days of culture, the cehs were pulsed for 6 to 8 h with 1 μCi [³H] -Thymidine, after which the cehs were harvested usmg an automated ceh harvester [³HJ- Thymidine incorporation was determined with a liquid scintillation counter.

Proliferation of T cehs was performed m quadruplicate wehs Results showed that the proliferation of the T cehs was inhibitted by the antibodies, secreted by the different hybridoma clones 3H10, 5F3, 7B8, 9D8, 1 1B9 13B9, 13D3, 14F1 Thus, the antibodies secreted by the different hybridoma clones 3H10, 5F3, 7B8, 9D8, 1 1B9, 13B9, 13D3, 14F1 are neutralizing antibodies

1.2.2 Fusion 2

One female BALB/c mice was immunized (injected rntrapentoneally) four times d e , at days 0, 27 56 and 97) with 5x10^b Sf9 insect cehs that were infected with a recombinant baculovirus containing a human B7 2 cDNA Three days after the last injection, spleen cehs were retrieved from the immunized mice and used for cell fusion mamly according the procedure as described by Kohler and Mhstem (1975) Dissociated splenocytes from the immunized mice were fused w th SP2/0 munne myeloma cehs at a ratio of 10 3 usmg a polyethylene glycol/DMSO solution

The fused cehs were mixed up and resuspended DME medium supplemented with hypoxanthme, thymidine, sodium pyruvate glutamine a non-essential ammo acid solution 10% inactivated fetal calf serum, 10% inactivated horse serum and 10% BM-Conώmed The cehs were then distributed to 960 wehs on tissue culture plates to which σminopteπn was added 24 hours after the cell fusion Each well contained between 1 to 5 growing hybridoma clones at the average Aiter eight days supernatants of the 960 wehs were combined in groups of 5 for primary screening. The 192 pools were screened for the presence of specific antibody by FACS analysis using freshly isolated monocytes. Monocytes were isolated from buffy coat on Ficoh-Paque(density gradient 1.077; Pharmacia) gradients. After three washings, monocytes were removed by cold agglutination (Mentzer et al., 1986). A

15 ml tube containing 50x10⁶ PBMC in 10 ml of culture medium was slowly rotated for 30 minutes at 4°C. Monocyte aggregates were allowed to sediment over a 15 minute period, and the non aggregated cehs were carefully aspirated. Monocytes (5xl0⁵ cells/sample) were incubated for 30' at 4°C in PBS supplemented with 2% FCS, 0.02% sodium azide and 10% normal rabbit serum with the different ( 100 ml undiluted) supernatants. Thereafter the cehs were washed two times with PBS supplemented with 2% inactivated FCS ands 0.02% sodium azide and incubated with goat anti-mouse (IgG and IgM) antibodies conjugated to fluorescein isothiocyanate (FTTC) for 30' at 4°C After another two washes, cehs were fixed using 0.5% paraformaldehyde and analyzed with a FACScan flow cytometer

(Becton Dickinson). The specific binding of the monoclonal antibodies is expressed as the mean fluorescent intensity in arbitrary units. The first screening yielded five positive pools. The twenty-five wehs corresponding to the five positive pools were subjected to a second screening with the FACS screening technique described above using 5xl0⁵ freshly isolated monocytes for each staining. This second screening provided one individual positive well containing antibodies reactive with the monocytes. This positive weh was subcloned and the subclones were subjected to a FACS screening as described above using the B7.2 expressing EBV-transformed human B ceh hne RMPI8866, (5x10⁵ cehs/stcdning). Out of this subcloning one stable hybridoma clone named 1G10 was obtained. This hybridoma clone secrete mouse antibodies of the IgG2a isotype. The antibodies secreted by the hybridoma clone 1G10 were tested for specific binding to the human B7.2 molecule. To this end, FACS staining on mouse 3T6 cehs, tranfected with the cDNA encoding human B7.2 molecules (3T6/B7.2) and on B7.2-negative 3T6 cells was performed. The 3T6 and 3T6/B7.2 cells were firstly incubated for 20' at 4°C in PBS supplemented with 5% inactivated FCS and 0.02% sodium αzide and 10% normal rabbit serum Subsequently the cells were incubated with the supernatant of the hybridoma clone 1G10 for 30 mm at 4°C Thereafter the cehs were separated from the supernatant washed three times and incubated with fluorescein lsothyocyanate-labeled goat anti-mouse antiserum d e GAM-FTTC) The cehs were also incubated with the GAM-FTTC alone After another

3 washes the cehs were analysed for fluorescence staining using a FACScan flow cytometer (Becton Dickinson) Results showed that the antibodies secreted by the 1G10 hybridoma clones specifically bound to the B7 2-expressmg 3T6/B7 2 cehs whereas 3T6 cehs that do not express B7 2, did not exhibit any bmdmg significantly greater than that of the control monoclonal antibody or GAM-FTTC

Moreover, the antibodies secreted by the 1G10 hybridoma clone were tested for their capacity to inhibit the proliferation of human peripheral blood T lymphocytes activated with anti-CD3 (OKT-3) m the presence of the mouse 3T6 cehs transfected with the cDNA encoding human B7 2 molecule and human CD32 (3T6/CD32/B7 2) Human peripheral blood T lymphocytes were isolated from buffy coat by density centrifugation T cehs were further punned by cold aggregation of the monocytes (4X 10⁶ PBMC/mhn RPMI bic + 10 % iFCS, 30' rotation at 4 °C 15' on ice, supernatant contains the enriched T cehs) T cehs were further enriched by depletion of monocytes, B cehs and NK cehs usmg Lympho Kwik T (One Lambda, Los Angeles CA) according to the manufacturers protocol Mitomycin C-treated

3T6/CD32/B7 2 cehs(at 200μl Mitomycin C at 250μg/ml to 800 μl RPMI bic - 10 iFCS durmg 45' at 37°C wash twice m RPMI bic ^ 10 % iFCS) at 10' cehs well were incubated with anti-CD3 mAb (OKT3 0 5 mg/ml) for 1 h at 37^"C fohowed by a 1 h incubation at 37^~C with decreasing concentrations of the antibodies secreted by the 1G10 hybridoma clone Subsequently purified T cells (5x10' cells/well) were added and incubated for 5 days After 5 aavs of culture the cehs ere pulsed for 6 to 8 h with 1 μCi [Η] Thymidine after which the cehs were harvested usmg an automated cell harvester 0HJ-Thymiarne incorporation was determined w th a liquid scintillation counter Proliferation of T cells was peπormed m triplicate wehs Results showed that the proliferation oi the T cells was inhibited by the antibodies secreted by thelGlO hvondoma clones Thus the antibodies secreted b\ the 1G10 hybridoma clone are neutralizing antibodies.

2.Isolation of recombinant human antibodies that bind to human B7.1 or B7.2

A large naive human phage displayed Fab repertoire (Target Quest, Maastricht,

The Netherlands), which contains 4.1 x 10¹⁰ antibody molecules was screened for binding to human B7.1 or B7.2 molecules. The direct selection was performed on biotinylated B7.1 ED or B7.2ED fusion proteins (B7. lED-Bio or B7.2ED-Bio) and specific phage antibodies captured by streptavidin paramagnetic particles and a magnet. A..S both, B7.1 and B7.2 are presented as immunoglobulin fusion proteins, we performed the selection in the presence of a 10 fold molar excess of human IgG to compete out the antibodies specific for the Fc moiety. Different (4) rounds of selection were performed. In the first round of selection the human Fab library was tested for binding 500 nlvl B7.1 ED-Bio, in the second and third round of selection the previous selected Fob's were tested for binding 1 OOnM B7. lED-Bio and in the fourth round of selection the previous selected Fob's were tested for binding 10 nM B7.1 ED-Bio. In this approach the selection is for antibodies specific for the native conformation of the antigen. Specific clones are screened in a phage ELISA for binding to B7.1ED or B7.2ED fusion protein. The direct selection approach for hB7.1 binders resulted in 16 out of 40 clones that gave strong binding to B7.1 ED fusion protein and not to B7.2ED fusion protein or human IgG using phage ELISA. In this phage ELISA, antigen (B7.1ED or B7.2 ED fusion protein) was coated at 5ug/ml in l OOmM sodium hydrogen carbonate pH 9.6 in an ELISA plate and incubated overnight at 4°C. Wehs were washed 2 times with both PBS and PBS tween-20 (0.1%) and then blocked for at least 30 minutes at room temperature with 2%

Ivlarval PBS. Plates were washed a further 3 times with PBS and PBS plus tween. To each well 50ul of 4% Marval PBS was added and 50ul of phage supernatants. These were mixed by pipetting up and down and left for 1.5 hrs. Plates were washed 3 times with PBS and PBS plus tween. A secondary antibody of goat anti immunoglobulin was added to each well and incubated for 1 hour at room temp.

Antibody was discarded and washed 3 times with both PBS and PBS tween. Positive phαge antibodies were then detected using standard TMB staining and detection at OD 450. The reaction was stopped by adding lOOul of 2M H₂S0₄. Results showed that there was a high diversity of binders to B7-lIg in the 3^rd round of selection at lOOnM. Nine different patterns were obtained . After the 4^th round of selection on B7.1-Ig (lOnM) one binder was recovered. This clone may have been selected for higher affinity by the 10 fold reduction in antigen concentration from round 3 to 4, however an increased phage display efficiency cannot be excluded as a contributing factor to the preferential selection of this clone. The ten clones pCES-lFabl to 10 (1A, 1E,2A,2G,3E,3G,3H,4G,5C,5H), obtained by screening on B7.1 ED fusion protein using the Phage Display Technology were transformed to JM83 expressionstrcrin. An overnight culture was 20x diluted in LB+ lOOμg/ml amp+ 1% glucose and incubated at 28°C until an OD600 of 0.5 was reached. After removing the glucose of the medium, the culture was induced with 0. ImM IPTG and further incubated at 28°C during ± 18h. To isolate soluble periplasmic proteins , the method described by Neu and Heppel (1965) was used.

Briefly, cehs were harvested by centrifugation and resuspended in ice cold shockbuffer (lOOmM Tris-HCl pH 7.4; 20% sucrose, ImM EDTA, pH8). After incubation on ice during 10' with occasional stirring, the mixture was centrifuged at lO.OOOrpm during 1,5'. The supernatans was removed and the pellet was immediately resuspended in ice cold distihed water. After incubation on ice during

10' with occasional stirring, the mixture was centrifuged at 14.000rpm and the obtained supernatans was the soluble periplasmic fraction. A considerable amount was secreted into the periplasm and a small amount leaked in the supernatans. The Fob's ( 1A, 1E,2A,2G,3E,3G,3H,4G,5C,5H) were tested in a ELISA experiment. In this ELISA.., B7.1ED fusion protein ( 0.5 μg/ml) was directly coated on the plate (37°C, 2 hours) fohowed by blocking in PBS 0.1% casein for 1 hour at 37⁼C. After washing 5 times_^ wehs were subsequently incubated with soluble Fab ( lh, 37°C), washed 5 times and incubated ( lh, 37^DC) with HRP labeled mouse anti-human mAb (cone) followed by the addition of TMB as substrate. Plates were read at 450-595 nm on a microtiter plate reader. Results showed that the anti-B7.1 Fab molecules bind and thus recognize to the B7.1ED fusion molecule. Human B7.2 binding FAb molecules are selected in a similar way using the B7.2 ED fusion proteins in the different selection rounds.

It is necessary to ensure the removal of antibodies specific for the Fc moiety. Therefore the selection is done in a similar way as previous described, using B7. lglu-glu or B7.2 his molecules; which do not have the Fc portion.

3. Expression ofB7.1 extracellular domain and B7.2 extracellular domain fusion proteins and expression of soluble B7.1 glu-glu and soluble B7.2 his molecules

Soluble proteins are generated which comprise the extracellular domains (ED) of respectively human B7.1 and human B7.2 proteins carboxyterminahy fused to the Fc-domcrin of a human IgGγ 1 (further referred as ED fusion protein), the peptide EEEEYMPME (glu-glu epitope) or a six histidine peptide (his) (further referred as ED-glu-glu or ED-his proteins).

Ah B7 proteins are expressed in insect cehs using the recombinant baculovirus expression system (BENS). Alternative expression systems are used. The soluble B7.1ED and B7.2ED fusion proteins, B7.1ED glu-glu and B7.2ED his proteins are needed for the further selection and optimization of higher affinity antibodies.

3.1. Generation of recombinant baculovmons expressmg soluble h37.1ED or h B7.2 ED fusion proteins or hB7. lglu-glu or hB7.2 his protems.

Expression of a recombinant protein using BEVS is realised by integration of a foreign gene expression casette, consisting of the protein-encoding cDΝA seσuence under transcriptional control of a strong baculoviral promoter (e.g. polyhedrin promoter), in the genome of the Autogxapha calif ornica Nuclear Polyhedrosis Virus (AcNPV), the prototype of the Baculoviridae, at a genome locus non-essential for in vitro replication of the virus (e.g. polyhedrin locus). As the baculovirus genome is too large, easy msertion of this recombinant expression casette (cDNA and viral promoter) by simple cloning techniques is not possible Therefore generation of the recombinant baculovirus genome was performed usmg a technology based on in vivo homologous recombination between the baculovirus w ld type genome and a transfer plasmid containing the foreign expression casette flanked by baculoviral genome sequences, both introduced msect cehs by classical cotransfection technology The resulting recombinant genome encodes a recombinant virus that is subsequently purified and amplified providing a high titre recombinant virus stock that can be used for large scale infection of msect cehs thereby producing large amounts of the foreign protem

3.1.1. Construcuon ofbaculo transfer plasmids for hB7.1ED Fc and human B7.2ED Fc recombinant baculovirus generation

3.1.1.1. Design of universal baculo transfer vector pVL-Fc

Fusion protems of both the hB7 1 and hB7.2 ED with the Fc part (Hιnge-CH2-CH3) of a humanlgGγ 1 immunoglobulin were expressed as recombinant baculovirus protems The cloning of the human IgGγ 1 -Fc sequence was performed by means of RT-PCR on mRNA isolated from stimulated human B cehs ( 16h coculture with mitomycin C treated mouse fibroblast (3T6) transfected w th cDNA encoding human CD40Lιgand, 3T6/hCD40L, m the presence of 20U/ml rhIL2), usmg specuic PCR-primers designed on the hlgGγ 1 seσuence pubhshed by Ellison et 01.( 1982, NAR 4071) The PCR amplified DNA fragment consisted of a 6 ammoacid encoding linker sequence mcludmg a BamHI site fohowed by the cDNA sequence codmg for the complete Hmge-CH2-CH3 domain of the IgGγl, and a stop codon The amplifiea sequence was inserted m the cloning vector pUC18 to allow insert seσuence analysis Subsequently the Fc fragment was recloned in the baculotransfei vector pVL1393 (Pharmmgen) as a BamHI-EcoRI (pUCrestπction site) DMA. fragment PVL1393 is a polyhedrin promoter controlled transfer vector designed for insertion of foreign expression casettes in the polyheaπn locus of a baculovirus genome The resulting transfer vector named pVL-Fc fusion vector (ICCG3048) (Figure 1) allowed in frame insertion of both the ED-cDNA of hB7 1 and hB7 2 usmg the BamHI clonmg site

3.1.1.2 Clonmg of full-size hB7 1 and hB72 cDNA sequences

The fuh size hB7 1 encoding cDNA sequence was originally generated by

RT-PCR performed on total RNA isolated from EBV-transformed human spleen cehs, usmg a hB7 1 specific pnmerset (MR67/MR68) designed based on the hB7.1 seσuence, pubhshed by Freeman et al., 1989, as described in de Boer.M et al , 1992 The amplified DNA was cloned m several eukaryotic expression vectors mcludmg pcDNAIneo (Invitrogen) for mammalian expression The resulting plasmid was named pcDNAIneohB7 1 (ICCG1713)

The fuh size hB7 2 cDNA sequence was originally cloned by RT-PCR on human peripheral blood mononuclear cehs usmg hB7 2 specific pnmerset IG2834/IG2833 The amplified fragment was mserted blunt end the EcoRV openened clonmg vector pBSK(+) for sequence confirmation, and later remserted in the mammalian expression vector pcDNA3 as an EcoRV-Xbal DNA fragment, resulting m pcDNA3hB7 2 (ICCG2307)

3 1 1 3 Insertion of the hB7 1ED and hB7.2 ED m pVL-Fc

The cDNA sequences codmg for the ED of hB7 1 and hB7 2 were then isolated by PCR on the above described plasmids pcDNAIneohB7 1 and pcDNA3hB7 2 containing respectively LιB7 1 and hB⁷ 2 fuh lenght cDNA, and usmg hB7 1ED or hB7 2ED-specιfic pπmersets The resulting amplified DNA fragments were originally mserted in the clonmg vector pUCl 8 for seσuence confirmation of the inserts and later remserted m the pVL-Fc (ICCG3048) by BamHI-Bghl clonmg resulting in in frame fusion of the B7 1ED and B7 2 ED to the IgGγ 1-Fc cDNA sequence already present in the pVL-Fc transfer vector resulting respectively the pVLshB7 1-Fc (ICCG3005) and pVLshB7 2-Fc (ICCG3004) baculotransfer plasmids (figure 2 3) 3.1.2. Construction ofbaculo transfer plasmid for hB7. lglu-glu recombinant baculovirus generation

Soluble hB7. lglu-glu was expressed as a fusion protein consisting of the extracellular domain (AA1-242) of the hB7.1 protein carboxyterminahy fused to the peptide EEEEYMPME , also named glu-glu epitope, for which monoclonal antibodies are available. The hB7. lglu-glu encoding fusion cDNA sequence was generated by RT-PCR performed on total RNA isolated from an EBV-transformed human spleen ceh line, using a hB7.1ED specific primerset (MR67/ MR145) based on the hB7.1 sequence pubhshed by Freeman et al., 1989 , as described in de

Boer,M. et al. , 1992.

The sense primer matched with the codons for caninoacids 1 and 2 of the hB7.1 and contained a Pstl cloning site upstream this ED-coding sequence. The antisense primer matches amino acid-codons 203-242 of the hB7.1 sequence fohowed by the sequence encoding the glu-glu detection/purification epitope, and a Kpnl cloning site (figure 4).

MR67sense primer : Pstl 5' gcg ctσcag cσtctgaagcc atg ggc c 3'

Met Gly ( l -2 of hB7.1 preprotein)

MR 145 antisense primer : Kpnl 5' qcσc σatacc tta etc cat ggg cat gta ttc etc ttc ctg gtt ate agg aaa atg ctg ttg 3'

Glu Met Pro Met Tyr Glu Glu Glu Glu Asn Asp Pro Phe His Gin Gin

Glu-glu epitope ~hB7.1 A_sn₂₄₂-Gln₂₀₂

The resulting 779bp PCR fragment was inserted in the baculo transfer vector pAcC8, a classical pAcYMl derived transfer vector (Matsuura et al.. 1987), as a Pstl- Kpnl fragment. The resulting pAcC8hB7.1 glu-glu transfer plasmid was used for insertion of the polyhedrin promoter controlled hB7. lglu-glu expression unit in the baculovirus polyhedrin locus by in vivo homologous recombination as described herunder.

3.1.3. Construction ofbaculo transfer plasmid for hB7.2his recombinant baculovirus generation

Soluble hB7.2his was expressed as a fusion protein of the extracellular domain (AA1-238) of B7.2 carboxyterminahy fused to a six histidine peptide. The hB7.2his tagged fusion cDNA seσuence was generated by PCR amplification performed on the plasmid pcDNAhB7.2(ICCG2307), containing the fuh size hB7.2 cDNA sequence as a template, using primerset IG8568 and IG8569. The PCR fragment, encoding AA l_Met -239_His of the hB7.2 preprotein fohowed by 5 extra histidine residues (figure 5), was inserted in the baculovirus transfer vector pAcSG2 (Pharmingen, San Diego) as a Xhol-Bghl fragment. The resulting pAcSG2hB7.2his transfer plasmid (ICCG3519) was used for insertion of a polyhedrin promoter controlled hB7.2his expression unit in the baculovirus polyhedrin locus by in vivo homologous recombination.

Sense primer IG8568 :

Xhol 5' ccg ctcσσσ ccttgcact atg gga ctg 3'

Met Gly Leu (AA 1-3 of hB7-2)

Antisense primer IG8569 :

Bglll 3' ggg ggt ctg gtg gta gtg gtg gta gtg att cct tctσσa agg 5'

Pro Pro Asp His His His His His His (AA 236-239 of LιB7.2 ÷ 5 His)

3.2. Generation recombinant baculovirions by in vivo homologous recombination Recombinant baculovirions containing the information for the expression of the different hB7 1 or hB7 2 soluble protems were generated following the 'baculogold transfection method' as described by Pharmrngen In the baculotransfer plasmids described above pVLhB7 lFc, pVLhB7.2Fc, pAcC8hB7 lglu-glu and pAcSG2hB7 2hιs, the fusion cDNA's were placed under transcriptional control of the strong late baculoviral poiyhedrm promoter The resulting expression casette (promoter and cDNA) was flanked by baculoviral genome sequences derived from the poiyhedrm genelocus, a locus non-essential for m vitro virus replication

The recombinant transfer plasmids were introduced by classical transfection methods in msect cehs (Spodoptera frugiperda , Sf9) together with the linearized genomic DNA of a modified version of the wild type AcNPV baculovirus genome, the Baculogold™ In this Baculogold™ baculovirus genome the poiyhedrm gene is replaced by the E coh lacZ gene and contains three Bsu36I restriction sites side respectively the lacZ gene and the baculoviral genes ORF603 and ORF1629, originally flanking the polyhedrin locus. Bsu36I linearization of this genome results m deletion of the lacZ insert and part of the baculovirus gene ORF1629 which is essential for m vitro replication of the virus In vivo homologous recombination between transfer plasmid and Baculogold™' genome at the poiyhedrm locus flanking sequences, resulted in the insertion of the fusionprotem expression casette in the baculoviral genome poiyhedrm locus) and m the rescue of the lethal deletion in ORF1629 Usmg this approach a recombination efticienty of > 99% was obtained, as wild tipe Baculogold¹¹ DNA did not encode viable virus

Individual recombinant baculovirions were subsequently isolated rrom the supernatant of transfected cells harvested 5 days post-transfection by a olaque- punfication assay Routinely 3 candidate recombinant baculovirions v/ere isolated for detailed protem expression pattern analysis 3.3. Analysis of protein expression pattern of recombinant baculovirus infected cells.

The isolated recombinants were characterized by kinetic infection studies in

Sf9 cehs (infection at MOI3, harvest ceh CM at 24, 48, 72 h post-infection) and analysis of the secreted recombinant fusion proteins by ( 1) Western blot using monoclonal antibodies specific for the Fc-tah (anti-Fc mAb) or specific for the glu- glu tag (anti-glu-giu mAb) or for the his-tag (anti-his mAb), and (2) in binding- inhibition studies using anti-B7.1,-and anti-B7.2 mAbs on B7.1 and B7.2 expressing ceh lines.

Western blot analysis ^•

The hB7.1ED and hB7.2ED fusion proteins were secreted as a mature glycosylated monomer protein with M.W. ± 62kDa (12.5% SDS-Laemmh, reducing conditions , Western blot analysis with anti-Fc mAb). In non-reducing conditions the proteins showed a high degree of ohgomerisation. Theoretical M.W. calculated for the non- glycosylated proteins is 53827 Da for hB7.2ED fusion protein and 54298 Da for hB7.1ED fusion protein. Optimal harvest time was determined at 72h post-infection for both proteins.

The hB7.1ED glu-glu protein was secreted as a mixture of mature differentially glycosylated monomer proteins with M.W. _r 33kDa ( 12.5% SDS-Laemmh, reducing conditions , Western blot analysis with anti-glu-glu mAb). Theoretical M.W. calculated for the non-glycosvlated protein is 28768 Da. Optimal harvest time was determined at 72h post-infection. Total deglycosylation of tine protein was obtained with N-glycosydaseF. The hB7.2hιs protem v as secreted as a mixture of mature differentially glycosylated monomer proteins with M.W. ± 33kDa ( 12.5% SDS-Laemmh, reducing conditions , Western blot analysis with anti-his mAb). Theoretical M.W. calculated for the non-glycosylated protein is 27799 Da. Optimal harvest time was determined at 72h post-infection.

Binding-inhibition assay on RPMI8866 cehs:

Both B7.1ED and B7.2ED fusion proteins as weh as the BF.1ED glu-glu and the B7.2his proteins were analysed in a binding-competition test on an EBV transformed human B-ceh line RPMI8866 expressing hB7.1 and hB7.2. Dilutionseries of the crude infected ceh conditioned medium (CM) was incubated with a fixed, suboptimal concentration of respectively anti-B7.1 (B7-24)(20ng/ceh pehet) or anti-B7.2(lG10)(30 ng/ceh pehet) mAbs. The antigen/antibody mixture was subsequently incubated with the RPMI8866 cehs and anti-B7.1 or anti-B7.2 antibody binding on the cehs was then determined by Fluorochrome activated ceh scan (FACs) using fluorescein isothyocyanate (FΙTC)-conjugated goat anti-mouse (Fab) antibodies .Inhibition of the anti-B7 mAb binding on the cehs by the B7ED fusion proteins or the B7. lEDglu-glu or B7.2EDhis proteins present in the tested sample was considered as a relative measure for the B7ED-protein concentration and functionality. Results are indicated as the highest possible sample dilution giving maximum inhibition of B7-specific mAb binding.

3.4. Production and purification of recombinant baculovirus expressed soluble h£7.1ED or hB7.2ED fusion protems , B7.1ED glu-glu or B7.2ED his proteins. High titre recombinant baculovirusstocks, prepared by classical virus amplification procedures, were used for production of the recombinant proteins following an in house optimized production procedure on plastic roller bottles. Insect cehs (e.g. Sf9) exponentially growing in spinner cultures in TCI 00/10% FCS medium (Gibco) were cohected by low speed centrifugation, resuspended at 2 10⁶ c/ml in TCI 00 and seeded in glass roller bottles at a maximum volume of 600ml/R1260 cm² surface, allowing optimum aeration of the cehs. A high titre virusstock was added at a multiplicity of infection (MOD of 3. Cehs and virus were incubated on a roher apparatus at 27°C for 48 h. Conditioned medium (CM) was harvested and cleared from cells and celldebris by low speed centrifugation. Alternative host ceh hnes (e.g. Sf21, High Five) , media and production procedures (e.g. spinner , bioreactor) are used. Quality control on the CM was performed by ( 1) Western blot analysis, using Fc specific mAb or anti-glu-glu mAb or anti-his mAb, and (2) hB7.1 and hB7.2 binding/competition assays by FACScan on hB7.1 and hB7.2 expressing recombinant ceh lines , as described previous. CM was stored at -70°C untih purification.

Purification 1XB7-1ED and hB7-2ED fusion proteins was done on a Prot A column, at neutral pH and at low salt concentrations. To avoid the problem of the presence of aggregates and the formation of aggregates during the purification procedure (due to the high amount of Cys present in the whole protein (1 1) and due to the fact that the Cys in the Fc region has the tendency to dimerise) DTT was added in the purification procedure. DTT was added in a concentration of 2mM at the starting material.

Purification of the hB7.1 gluglu, expressed from Baculo virus infected cehs, occurred by immunoaffinity onto a M24-coupled divinylsulf one activated matrix in expanded bed mode (UpFront, Denmark). The bound material which was eluted by pH 3.0 behaved on SDS-PAGE Western blotting (detection with anti gluglu) and coomassie staining under reducing and non-reducing conditions as a monomer

(MW : 35-40 kD). The sB7.1 monomeric band had a triplet nature with a smear in between which is thought to be caused by different gradations of glycosylation On gelfiltration chromatography performed with a Superdex 200 PC3 2/30 ( Smart Pharmacia) (Figure 6 gelfiltration profile of a MW markers 1 = 670 kD 2 = 158 kD, 3 = 44 kD, 4 = 17 kD, 5 = 1 3 kD, b sB7 1ED gluglu) a smgle peak at the elution place for a 44 kDa protem could be detected indicating that the sB7 1 gluglu behaves as a monomer

Purification of the hB72Hιs6, expressed from Baculo virus infected cehs, occurred by capturing the glycosylated protems on Lentil lectin chromatography Elution of the Lentil lectin column was performed by applying buffer containing 500 mM Methylmannopyranoside The only remaining contaminating protem with a

Molecular weight of -/- 70 kD was removed by Nι²"-IDA chromatography The 70 kD protem was recovered the flow through while the sB7 2hιs was recovered m the 200 mM imidazole elution SDS-PAGE Coomassie staining revealed that the resulting material was of high purity > 90% SDS-PAGE Western blotting (detection with anti His) run under reducing conditions showed a monomeπc band which had a triplet nature and smear m between (MW 35-40 kD) SDS-PAGE Western Blotting run under non-reducing conditions showed also as major amount the same triplet with between a smear but supplementary a band at the place for dimeπc material and some higher aggregation bands but with minor intensity The sB7 2hιs behaved heterogeneously on gelfiltration chromatography performed with a Superdex 200 PC 3 2/30 (Smart Pnarmacia) (Figure 7 gelfiltration profile of a MW markers 1 = 670 kD 2 = 158 kD 3 = 44 kD 4 = 17 kD 5 = 1 3 kD b sB7 2ED Hιs6) The major peak was around 44 kD corresponding to the monomeπc sB7 2 Another peal: about 5 % comparea to the mean peak eluted at the retention time ior dim eric material (-¹-/ 90 kD) This peak v/as preceded by an augmentation m aosorbency which means tnat also nigner sB7 2 oligomers were present The tnird determined peak (also about 5% compared to the major peak) eluted at 2 kD It is clear that this sB7 2Hιs6 protein has a higher tendency to iorm oligomers than the sB7 1 gluglu 3.5. Characterization of the B7 1ED and B7.2ED fusion proteins and the B7 lED-glu- glu and B7.2 his protein

Purified B7 1ED and B7.2 ED fusion molecules were tested in binding inhibition studies, using anti-B7 1 mAb (B7-B7-24) and antι-B7.2 mAb (1G10), on

RPMI8866 cehs expressmg both B7 1 and B7.2 molecules. A suboptimal concentration of anti-B7 1 mAb (B7-B7-24) (20ng/ceh pehet) or of anti-B7.2 mAb ( 1G10) (30 ng/ceh pehet) was preincubated for 20' at 4°C with different concentrations of B7.1ED or B7.2ED fusion molecule (serial dilution ranging from 1 μg/ml to 2ng/ml) Cehs (0.5 x 10⁶ cells/sample) were incubated for 20' at 4°C with the B7 1ED fusion molecule/B7-24 complexes or with the B7 1ED fusion molecule/lGlO complexes or with the B7.2ED fusion molecule/B7-24 complexes or with the B7.2ED fusion molecule/lGl O complexes JY cehs ( lxlO⁵ cells/sample) The results show that a low concentration of B7 1ED fusion molecule (37 5 ng/ceh pehet) specific inhibits the binding of anti-B7 1 mAb (B7-24 mAb) (20ng/pehet) and not the binding of anti-B7.2 mAb ( 1G10) (30ng/ ceh pehet) on the RPMI8866 cehs.Otherwise, the results show that a low concentration of B7.2ED fusion molecule (37.5 ng/ceh pehet) specific inhibits the binding of anti-B7.2 mAb ( 1G10) (30ng/ cell pehet) and not the binding of anti-B7.1 mAb (B7-24 mAb) (20ng/pellet) on the RPMI8866 cehs Purified B7 1ED glu-glu molecule and unpurified B7.2ED his molecule

(baculopvirus supernatant) v ere tested in bmdmg inhibition studies as described above, usmg antι-B7 1 mAb (B7-24) and anti-B7.2 mAb ( 1G10), on RPMI8866 cehs expressmg both B7 1 and B7 2 molecules. Cehs (0.5 x 10^f' cells/sample) were incubated for 20 at 4^"C with the B7 1ED glu-glu/B7-24 complexes or with the B7.2ED his/lGlO complexes The experiment was further performed as described above

The results snow that low concentrations of B7 1 giu-glu molecule (7 5 ng/pellet) are able to inhibit the bmdmg of antι-B7 1 mAb (B7-24) (20ng/cell pellet; to the RPMI8866 cehs and that unpurified B7 2 his (baculovirus supernatant) is able to mhibit the binding of antι-B7 2 mAb ( 1G10) (30 ng/ceh pellet) on the RPMI cehs Purified 37 1ED and B7.2ED fusion molecules were tested m ELISA , using antι-B7 1 mAb (B7-24) and antι-B7.2 mAb ( 1G10) Purified B7 1ED or purified B7.2ED fusion molecule were coated (4°C, overnight) on Nunc Maxisorb strips in carbonate buffer, pH 9.6 at different concentrations (500-50-10-0 ng/ml). After washing 5 times with PBS, wells were subsequently blocked with PBS 0.1% caseind hour, 37°C) and incubated with anti-B7.1 mAb (B7-24) (concentration range, 1/2 serial dilution, 1000 to 1 ng/ml) or with anti-B7.2 mAb (1G10) ( 1000 to lng/ml) respectively. After washing 5 times, wehs were incubated ( 1 hour, 37°C) with HRP conjugated Goat anti-mouse IgG Fcγ specific monoclonal antibody.The ELISA was developped with the substrate TMB. Plates were measured in a ELISA reader at 450 nm. Results show that B7.1ED and B7.2ED fusion molecules can bind anti-B7.1 monoclonal antibody (B7-24) and anti-B7.2 monoclonal anitbody (1G10) respectively at a coating concentration of 100 to 500 ng/ml.

Purified B7.1ED glu-glu molecule was tested in ELISA , using anti-B7.1 mAb (B7- 24). Purified B7.1ED glu-glu molecule (0.5 μg/ml) was coated (400 overnight) on Nunc Maxisorb strips in carbonate buffer, pH 9.6. The experiment was further performed as described above. Results show that B7.1ED glu-glu molecule can bind anti-B7.1 monoclonal antibody (B7-24) at a coating concentration of 0.5 μg/ml.

4. Generation of scFv's to B7.1 andB7.2

The anti-B7.1 mAb B7-24 and the anti-B7.2 mAb 1G10 are used for the generation of B7.1 and B7.2 scFv^'s respectively. Both the B7.1 and the B7.2 scFv's are used as positive controls in screening procedure for the generation of the B7.12 Mab. Moreover, they serve as the B7.1-, B7.2-binding component in the bispecific diabody, the bispecific BiTAb and the trispecific triabody (see further). Finally, these scFv's are used as competitor reagents in the selection for high affinity variants of the anti-B7.12 scFv.

4.1. Cloning VH and VL regions by PCR in pCANTAB5E vector

Anti-B7.1 (B7-24) Mab VH and VL regions available in a baculo expression vector are previously described in WO 94/01547. Starting from this vector, both variable regions wee PCR-amphfied using the degenerate primers available from the Pharmacia RPAS (Recombinant Phage Antibody System) Mouse scFv Module. In a second PCR, both VH and VL were linked using a short synthetic linker. After cleavage with the Sfil and NotI restriction sites, de B7-24 scFv was ligated in pCANTAB5E. To select for functional scFv's, phages were generated after ligation in pCANTAB5E. TGI cells were transformed with the ligation mixture and were infected with M13K07 helper phage to induce phage production. The produced phages were panned upon a B7.1 positive B cell line (JY cehs), washed and subsequently eluted from the JY cehs. Binding phages were screened for their binding capacity in a FACS analysis with this B7.1 positive B ceh line. Coπect binders were selected for further DNA sequence analysis. The sequence of B7-24 scFv clone 6 was found to be correct.

For the anti-B7.2 scFv's, the VH and VL regions were cloned by RT-PCR with a set of degenerate forward primers located at the 5' end of the VH or VL region and a backward primer located in the constant region of the heavy or tight chain. The resulting cDNA was cloned in pUC18 and consensus VH and VL sequences were determined.

Starting from correct clones in pUCl 8, the VL and VH regions were cloned in the pCANTAB5E vector using degenerated primers obtained from Pharmacia, transformed in TG 1 cehs, infected with M 13K07 helper phage, fohowed by panning upon a B7.2 positive B ceh line (JY cells), as described for the anti-B7.1 regions. Correct binders were selected for further DNA sequence analysis. The sequence of I G I OSCFV clone 31 was found to be correct.

4.2. Test expression of scFv's with phage FACS and ELISA

The recombinant phages expressing B7-24 scFV( clone 6) and 1G10 scFv (clone 21) were tested in FACS experiments using FY cehs, expressing B7.1 and B7.2 molecules (de Boer et al., 1992). JY cells ( lx 10⁵ cells/sample) were incubated for 20' at 4^CC with B7-24 scFv or 1G10 scFV phage diluted serial from undiluted to zero.

Aiter washing twice in RPMI 1640 supplemented with 10% FCS, the cells were subsequently incubated for another 20' at 4°C with a sheep anti-M13 monoclonal antibody, washed twice in RPMI 1640 supplemented with 10% FCS and incubated (20' at 4°C) with anti-sheep antibody conjugated to fluorescein isothiocyanate (FTTC). JY cehs were washed twice in RPMI 1640 supplemented with 10% FCS and finally suspended in PBS supplemented with 1% BSA and 0.1 % NaN₃ and analysed in a FAC-Scan flow cytometer (Becton Dickinson). The specific binding of the scFv phages is expressed as the mean fluorescence intensity in arbitrary units. The results showed that the B7-24 scFv and the lGlOscFv phages binds to the JY cehs. The mean fluorescence intensity (MFi) diminished when the phage is titrated. Evidence showing that the obtained scFv phages are specific for B7.1 or B7.2 was obtained in a competition FACS experiment on JY cehs, expressing B7.1 and B7.2 molecules. In this competition experiment the JY cehs were preincubated with the anti-B7.1 mAb (B7-24) or the anti-B7.2 mAb (1G10), fohowed by incubation with the scFv phages. jΥ cehs (lx 10⁵ cells/sample) were incubated for 20' at 4°C with 2μg anti-B7.1 mAb (B7-24) or with Tμg anti-B7.2 mAb ( 1G10). After washing twice in

RPMI 1640 supplemented with 10% FCS, the cehs were incubated for 20' at 4°C with B7-24 scFv or 1G10 scFv phage diluted serial from undiluted to zero. After washing twice in RPMI 1640 supplemented with 10% FCS, the cehs were subsequently incubated for another 20' at 4°C with a sheep anti-M13 monoclonal antibody, washed twice in RPMI1640 supplemented with 10% FCS and incubated for 20' at

4^CC with anti-sheep antibody conjugated to fluorescein isothiocyanate (FTTC). JY cehs were washed twice in RPMI 1640 supplemented with 10% FCS and finally suspended in PBS supplemented with 1% BSA and 0.1 % NaN₃ and analysed in a FAC-Scan flow cytometer (Becton Dickinson). The specific binding of the scFv phages is expressed as the mean fluorescence intensity in arbitrary' units. The results showed that the B7-24 scFv phage is not capable to bind the JY cehs preincubated with the parent monoclonal antibody, B7-24. No inhibition of the binding of the B7-24 scFV could be detected when JY cehs were preincubated with anti-B7.2 monoclonal antibody, 1G10. Similar results were obtained for the 1G10 scFV phage. The 1 G 10 scFv phage is not capable to bind the jΥ cehs preincubated with the parent monoclonal antibody, 1G10. No inhibition of the binding of the 1G10 scFV could be detected when JY cehs were preincubated with anti-B7.1 monoclonal antibody, B7-24.

Functional binding of the recombinant phages to their respective antigens was also tested in ELISA using the B7.1ED and B7.2ED fusion proteins. Plates were coated overnight at 4°C with anti-human IgG. After washing withPBS Tween, plates were incubated with B7.1ED or B7.2ED fusion proteins .After washing (PBS Tween), the plates were incubated with the phages (B7-24 phage or IGIO phage) in serial dilution from undiluted to zero or with the parent monoclonal antibodies (B7-24 or IGIO) serial diluted from lOOng/weh to zero. Binding was detected with a labelled anti-M13 or anti-mouse Ig antibody. As substrate TMB was used. Plates were measured at 450 nm. Ah incubations were done for 1 hour at 37°C Resultsshowed that B7-24 scFv and 1G10 scFv bind to theB7.1ED or B7.2ED fusion molecule respectively. The binding of the scFv phages is much higher than the binding of the parent antibodies. This is not due to a difference in binding affinity, but is the result of the different detection sytems used..

In case antibody VH and VL regions do not directly fold properly as scFv's, a small PM library in the pCantab5E vector (see section 6) is generated and specific mutations in VH-VL pairing domains are introduced in order to optimize scFv folding. Screening for variants with improved folding is done using immobilized ED fusion proteins.

4.3. Milligram scale production and purification of scFv's in E. Coli

To produce milligram amounts of purified ScFv, cDNA encoding the scFv's from the pCantab5E vector are transferred to an E. coli expression vector. The genes encoding the scFvB7-24H6 and scFvlGlOHδ were isolated by PCR with primer 6999 and 7002 for scFvB7-24 and primer 7001 and 7002 for scFvlGlO, v/hereby specific restrictionsites were created.

list of used primers:

6999: 5' -CGCGGTATACAGGAGTCTGGGGGAGGCTTAG- 3' 7000: 5' -CGCGCTCGAGCTTGGTCCCCTGACCGAAC- 3' 7001: 5' -CGCGGTATACAGCAGTCTGGACCTGAGCTG- 3' 7002: 5' -CGCGAGCGCTTGGTACCTCCACCGAACG- 3'

The PCR fragments were cloned in the pGEM-T-vector (Promega) and were subjected to DNA sequence analysis. After cleavage with Xhol - Bsil 107 and Eco47III - Bstl 107(Biolabs) of _PGEM-TscFvB7-24H6 and pGEM-TscFvlG10H6 respectively, the genes encoding scFvB7-24H6 and scFvlG10H6 were cloned in an E.coli expressionvector under control of the IPTG inducible lac-promotor .They are preceded by the pelB signal seσuence to obtain secretion into the periplasm. A C- terminal histidine-tag for purification purposes was included. The expression plasmids containing the scFv constructs were transformed in the E.coli expressionstrain JM83. (Knappik and Plϋckthun 1995). An overnight culture was 20x diluted in LB+ lOOμg/ml amp+ 1% glucose and incubated at 28°C until an OD600 of 0.5 was reached. After removing the glucose of the medium, the culture was induced with 0. ImM IPTG and further incubated at 28°C during ± 18h. To isolate soluble periplasmic proteins , the method described by Neu and Heppel (1965) was used. Briefly, cehs were harvested by centrifugation and resuspended in ice cold shockbuffer (lOOmM Tris-HCl pH 7.4; 20% sucrose, ImM EDTA, pH8). After incubation on ice during 10' with occasional stirring, the mixture was centrifuged at l O.OOOrpm during 1,5'. The supernatans was removed and the pellet was immediately resuspended in ice cold distihed water. After incubation on ice during 10' with occasional stirring, the mixture was centrifuged at 14.000rpm and the obtained supernatans was the soluble periplasmic fraction. Only 10-30 % of the expressed scFv 's was found in the periplasm as a soluble and correctly folded functional molecule and a small amount leaked in the culture supernatans. This small periplasmic fraction was purified using metal affinity chromatography. ScFv B7-24 from the soluble periplasmic fraction was purified up to >90% purity by applying it onto a Zn²"-IDA. column with subsequent wash and elution with 40 and 150 mM Imidazole. The same procedure is apphed for the purification of ScFv 1G10 from the soluble periplasmic fraction.. The major part of the scFv's formed inclusion bodies in the periplasm but were processed correctly.This was confirmed by NH₂-terminal sequence analysis.

4.4. Characterization andBIAcore analysis of scFv's

Binding specificity of the expressed scFv's is checked by ELISA experiments using the ED fusion proteins, and FACS analysis using EBV-transformed cehs

(RPMI 8866), that are positive for expression of both B7.1 and B7.2 molecules.

Binding of scFv's to their target molecules is detected by means of a his-tag present on these scFv's.

In the ELISA experiment, mouse anti-his mAb (1/1000 cone) is coated on the plate in PBS for 2h at 37°C. After washing, B7-24 scFv is added in a twofold serial dilution starting from lμg/ml to 15 ng/ml and incubated for lh at 37°C . The fuh size mAb B7- 24 was added as positive control (lμg/ml). After washing, the B7.1 ED fusion protein (500ng/ml) is added for 30' at 37 °C and detected by sheep anti-human HRP (cone,

1 hour, 37 °C). TMB was added as substrate. Results were measured at 450nm.Results showed that purified B7-24 scFv (2 μg/ml) binds the B7.1 ED fusion protein in this ELISA.

Purified B7-24 scFv was also tested and compared with the fuh size mAb B7-24 for its binding capacity to the human B7.1 molecule present on the RPMI8866 cehs, a EBV transformed B ceh line positive for hB7.1 and hB7.2. RPMI 8866 cehs were incubated with B7-24 scFv or, as control, with anti-B7.1 mAb (B7-24). Cehs (0.5 xlOøsσmple) were incubated for 30' at 4°C with B7-24scFv or with anti-B7.1 mAb (B7-24).After washing twice in RPMI supplemented with 10% FCS, the cehs were incubated for another 30' at 4°C with a mouse anti-His mAb. After washing twice in

RPMI supplemented with 10% FCS, the cells were incubated for another 30' at 4°C with a biotinylated goat anti-mouse IgG(Fab)-antibody fohowed by an incubation (30' at 4°C) with streptavidine conjugated with phycoerithrine (PE). The cells were washed twice in RPMI 1640 supplemented with 10%FCS and finally suspended in PBS supplemented with 1% BSA and 0.1 % NaN₃ and analyzed with a Facs Scan flow cytometer (Becton Dickinson). The specific binding of the monoclonal antibodies is expressed as the mean fluorescence intensity in arbitrary units. The results show that the B7-24scFv binds to the RPMI 8866 cehs, expressing both B7.¹ and B7.2 molecules.

The purified scFvb7-24 were tested for their capacity to block the T ceh APC interaction in a mixed lymphocyte culture. T cehs were purified out of whole heparinized blood on Ficoh-Paque (density 1.077, Pharmacia Biotech) density gradients. The peripheral blood mononuclear cehs (PBMC) present in the interface were washed three times in 40 ml of RPMIbic supplemented with 10% inactivated FCS. Subsequently, monocytes were removed by cold aggregation (Mentzer et al., 1986). To this end, PBMC were resuspended in 40 ml of RPMIbic supplemented with 10% inactivated FCS and slowly rotated for 30' at 4°C. Monocyte aggregates were allowed to sediment over a 15' period incubation on ice, and the non- aggregated cehs containing enriched T cehs and B cehs were carefully aspirated and centrifuged for 10' at 1200 rpm. T cehs were further enriched using Lympho- Kwik-T (One lambda Inc, Los Angeles, CA). This reagents contains a mixture of anti-monocyte and anti-B ceh mAbs and complement. Lymphocytes were resuspended in 3 ml Lympho-Kwik-T and the mixture was incubated for 45' at 37°C. Subsequently, cehs were resuspended in 0.5 ml of PBS and centrifuged for 2' at 2000 rpm and washed twice. The mouse fibroblast ceh line, 3T6, tranfected with the cDNA encoding human B7.1 and human CD32 (3T6/CD32/B7.1 ) was mitomycin C treated. Ceh pehet of 1 subconfluent falcon was dissolved in 800 ml RPMIbic and 200 ml Mitomycin C (250 mg/ml) and incubated for 40' at 37°C fohowed by two washes. Subsequently, cehs are suspended in 30 ml RPMIbic and incubated for 15' at 37°C fohowed by one additional wash step. In the mixed lymphocyte culture (MLR), the mitomycin C treated 3T6/CD32/B7.1 cehs ( 10⁴ cells/well) were incubated with OKT3 ( 1 μg/ml) for 1 h at 37 C followed by a 1 Li incubation at 37°C with decreasing concentrations of mAbB7-24 or scFvB7.24. Subsequently, purified T cehs (5x10⁴ cell/well) were added and incubated for 5 days. After 5 days, cehs were incubated with 1 mCi (^JH)-thymidine for 6 h and harvested using an automated cell harvester. ( ³H)-thymidine incorporation was determined vith a liquid scintillation counter. Results showed that the recombinant scFvB7-24 antibodies displayed neutralizing activities, comparable to those of the parent B7-24 mAb, as measured in MLR.

The binding affinity of the scFv's is compared with that of Fab fragments of the parent antibody in order to evaluate whether the scFv fragments have the same affinity as the intrinsic affinity of the parent antibody, which is reflected in the binding characteristics of the Fab fragments. The ED fusion proteins are used as target molecules and the conditions for appropriate analysis of the different molecules are optimized for BIAcore analyses. These accumulated data give an indication of the extend with which the affinity is lowered by the construction of a scFv and set a goal for the increase in affinity obtained using Parsimonius mut agenesis (PM).

6. Generation of novel B7.12 Mabs or small antigen binding peptides/ fragm ents (microproteins)

This novel molecule is a single mAb that can bind to both B7.1 and B7.2. In spite of the fact that both B7.1-B7.2 ligands, CD28 and CTLA4, bind B7.1 and B7.2 with comparable affinities, it has not been possible to produce murine Mabs which react with both B7.1 and B7.2. This is probably due to the fact that the active B7.1- B7.2 epitopes, which are conserved between B7.1 and B7.2, are also sufficiently conserved between mouse and human to ensure that idiotypes against the human epitopes would be suppressed in the mouse as anti-self.

Mice are immunized with peptides derived from the homologous sites between B7.1 and B7.2.and, as an alternative route, human Mabs which react with high affinity' to both human B7 molecules are made via antibody engineering. Tv/o types of phage display libraries are used to obtain such antibodies by biopanning on immobilized B7.1ΞD and B7.2ED fusion proteins. As an alternative approach, a phαge-displαy conformαtionαl constrained library is screened which ends up in the development of small antigen binding peptides or fragments (also called microproteins) interacting with B7.1 and B7.2.

4.1. Phage display libraries

4.1.1. Generating 'rationally' and 'random' designed CTLA4-like libraries

In the first series of experiments ('rational' approach), CTLA4 elements are grafted onto a limited set of structurally homologous antibody scaffolds to define which elements within CTLA4 may be replaced without loosing B7.1-B7.2 binding. Otherwise, irnmunoglobulin sequences are grafted onto structurally and functionally important CTLA-4 sequences. CTLA4 is a member of the immunoglobulin superfamily, and as such its main extracellular region is buht up of a domain with a basic immunoglobulin fold. On the basis of a model of CTLA4 and a comparative analysis to known antibody variable domains, it is determined which CTLA4-elements are grafted (CDR1-3 Including the 'MYPPPY" sequence, possibly FR residues), onto which scaffold. The 'scaffolds' are derived from the structurally most homologous antibody V-genes. For example, FR2 and FR3 are most homologous to human lambda tight chains with respect to amino acid sequence, with some homology for human kappa regions. 'Structural' homology^', which in our case is the similarity between antibody structure and the CTLA-model, aided by predictions on similarity of CDR-canonical folds determine which 'scaffolds' are used. This part of the work requires detailed computer modelling and is performed by Dr. Johan Desmet, K.U.Leuven, Campus Kortrijk, Interdisciphnair Research

Centrum, Belgium (Desmet et al., 1992; Lasters and Desmet, 1993; Lasters et al., 1995 and Desmet et al., 1997).

Such chimaeric molecules are created and tested for binding to B7.1 and B7.2; this generates very useful information for the second stage of this approach. Large repertoires of selected antibodies (and their FR) are then made and are provided with a minimum of CTLA4-elements; these molecules are displayed on the surface of phαge. From these hybrid hbrαries molecules that bind to both B7.1 and B7.2 are selected.

The CTLA4 CDR regions are shuffled at the DNA level with the selected antibody FR sequences; oligonucleotides encoding the CTLA4 CDR regions fused to neighbouring antibody-derived framework residues are combined with 4 repertoires of frameworks (FR1-4) by PCR-assembly. Certain templates (VH, V- kappa or V-lambda) are chosen based on successful identification of closely homologous FR sequences.

The display of CTLA-4 on filamentous phage is a pre-requisite for the rational design approach using CTLA-4 as a scaffold. We have designed primers to clone human CTLA-4 into our phage display vector pCES 1. We have cloned hCTLA-4 extracellular domain into pCESl as a ApaLl/Notl fragment. This fuses CTLA-4 to the N-terminus of p3 for display on filamentous phage. We have decided to use a CTLA-4 fragment as described in (Metzler et al., 1997) except we wih not include the C-terminal cysteine residue (Cys 123) believing that this may present problems with the correct folding and presentation of the CTLA-4 p3 fusion on filamentous phage. In figure 9 the PhagemidpCES 1 is represented : antibody genes : V_L-C_L, variable (V) and constant (C) region of the tight chain; V_H-C_H1, variable and first constant region of the heavy chain; PlacZ, promoter; rbs, ribosome binding site; S, signal sequence; H6, six histidines stretch for IMAC purification; tag, c-myc-derived tag; amber, amber codon that allows production of soluble Fab fragments in non- suppressor strains, gill, gene encoding one of the minor coat proteins of filamentous phage. Restriction sites used for cloning are indicated. Clones with the correct sized insert were further tested by BsfNl fingerprint and shown to have the expected pattern. Phage were prepared and tested in ELISA for binding to B7-lIg and B7-2Ig. A strong signal was obtained showing binding to both B7.1-Ig and B7.2-Ig. and not to BSA or plastic . Soluble CTLA-4 also showed some binding to B7-1 and B7-2 in ELISA, however the signal was only twice background. It is interesting to see what the enrichment factor is of CTLA-4 phage for the ligands B7- 1 and B7-2, and to determine if it wih indeed be possible to select novel binding molecules. CTLA-4 binds to B7-1 and B7-2 with Kd of 0.4 and 2.2μM respectively with a very fast off rate (koff > 0.4 sec^_1)(Greene et al., 1996 ; Van der Merve et al., 1997). The enrichment factor of CTLA-4 phage on B7-lIg and B7-2Ig was 5100 and 2687 respectively. This enrichment factor is very high , of the order 10³ per round of selection and is presumeably due to the avidity effect of display on filamentous phage. This means that, provided chimeric CTLA-4/Vgene molecules are be expressed and displayed efficiently, that selection is possible even for lower affinity interactions than the wild-type CTLA-4/B7-12 interaction. In a second rational design approach using immunoglobulin as a scaffold, a rational protein design can identify a structurally homologous protein scaffold (ie a member of the immunoglobulin superfamily (e.g V_H, V_L, CD2, CD 122, CD28) upon which we can graft structurally and function ally relevant regions of CTLA-4. For example, the immunoglobulin hght chain A27(DPK22) is the most frequently used human hght chain in vivo and can be used as a scaffold. Light chains are generally more soluble than heavy chains, and light chains are more structurally compatible with CTLA-4 than heavy chains. We have designed a molecule that can be synthesised from oligonucleotides for display on filamentous phage. Features which have been incorporated are (Figure 10_:Hybrid light chcrin/CTLA-4 molecule): -CDR 1 of CTLΛ-4 introduced at a structurally compatible site of A27 as EY-CDR 1 -

VRVTV -CDR3 of CTLA-4 introduced at a structurally compatible site of A27 as KVEL-CDR3-

GIG We have introduced degeneracy at Ileu51 and phe 65 so as to generate constructs with both the naturally occurting hght chain residues at these sites and also to allow for the introduction of the non-Ig disulphide bridge which is present in CTLA-4 We have introduced degeneracy at Phe 74 so as to generate constructs with both the naturally occurting residue and also the vaiine (Strand E) of CTLA-4 which may be structurally important. Restriction sites have been introduced by silent mutagenesis so as to be able to replace CDR1, 2 and 3 and also further engineering of molecule based on structural, functional and solubility modelling predictions. Restriction sites of Apa l/Notl have been introduced for cloning into phage display vector for binding studies and further manipulation using either rational or random approaches.

The CK hght chain is appended to the C terminus as aKpnl/Notl fragment to solve the solubility problem, as expected with a unpaired tight chain molecule, . The generated molecules are manipulated in a rational way using modelling predictions to engineer changes to resolve any functional, structural or solubility difficulties. Furthermore this molecule is fed into the random approach.

In the random approach, 'random' framework pools, derived from naive or immune Ig pools are used to identify apparent compatible antibody-derived FR regions. We have used use DNA shuffling (Stemmer., 1994) to generate libraries of chimeric CTLA-4 immunoglobulin molecules. By shuffling human CTLA-4 with a pool of human V_H and V_L antibody chains and their subsequent display we select novel B7-

1/2 binding molecules. Molecules are fused to the N-terminus of the minor phage coat protein p3 and selected for binding to B7.1/2. Our experimental strategy is shown in figure 11 We have generated two repertoires of chimeric molecules which have reamplified with primers specific for the N terminus of V_L and also the C-terminal of CTLA-4. The rational for using a C-terminal primer specific for CTLA-4 was that this is functionally the most important region and should be retained if possible in any new B7-1/2 binding molecule. The tv/o repertoires that have been constructed in pCES 1 have been DNA shuffled at 50°C and also at a lower temperature 45°C

Hu-CTLA-4/VL hybrid

HU-VK Back ApaLl ( 17 primers) Hu-CTLA-4-For-No/ 1 ( 1 primer)

Type of library Temperature of shuffle Library size

pCES l (VL/CTLA-4)-l 45°C l . le6

PCES 1 (VL/CTLA-4)-2 50°C 8.4 e5

We have also generated two repertoires of chimeric CiT-A-4/__mmunoglobulin molecules displayed on phage where we have amplified with primers specific for both the N and C-termini of immunoglobulin hght chains.

Hu-CTLA-4/VL hybrid i 1

I I J

Hu-Vκ Back ApaLl (6 primers) Hu-Jκ-For-NoH (5 primers)

Type of Library Temperature of shuffle Library size pCESl(VL/CTLA-4)-3 45°C 1.3 e 6

PCES l(VL/CTLA-4)-4 50°C 7.8 e 5

To Introduce CTLA-4 CDR3 mto hght chain repertoire we have designed primers which can be used to introduce the functionally important CTLA-4 CDR3 sequence into a immunoglobulin light chain repertoire at a structurally compatible position. These oligonucleotides are also used to spike DNA shuffling experiments

C7LA-4 CDR3/hght chain oligoucleoude design- V_L-FR3— IvIYPPPY— V_LFR4

1mm unoglobulm light chain repertoire

Ctlα-4 cdr3/hght chain oligo

/ ^* Hujλ-For Not (3 primers)

Hu VK back ApaLl (6 primers) HUJKFOΓ Not (5 primers)

HuVλ back ApaLl (11 primers)

Sequences of the CTLA-4 CDR3/Vk light chain spiking oligonucleotides are designed so as to amplify all VK sequences and are presnted in figure 12 Sequences of CTLA-4 CDR3/ Vλ light chain spiking oligonucleotides are designed so as to amplify all Vλ sequences and are presented in figure 13

In all cases it is crucial to improve the selectivity of the selected molecules for B7.1, B7.2 or both. Chain shuffling is used for this purpose.

4.1.2 Genercding dual-binding Fab fragments by alternating selections bom a large naive antibody repertoire

This method selects antibodies binding to both B7.1 and B7.2 from a large ncrive Fab library, which contains 4.1 x 10¹⁶ antibody molecules (Target Quest) displayed on phage. Selections are alternated between (biotinylated) B7.1ED and B7.2ED fusion protein and with excess of non-labeled Ig to deplete Ig-binding antibodies. This selection mode yields antibodies that target the (partially) overlapping binding site of CTLA4 on B7.1 and B7.2. It is also necessary to ensure the removal of antibodies specific for the junction between B7.1/2 and the Fc moiety; therefor alternating selections are done with KB7. lglu-glu and B7.2 his proteins, without the Fc portion.Gene diversification is required to improve the selectivity of the antibody. For this, the library quality is optimized using Trinucleotide-directed Mutagenesis (TRIM) technology from Morphosys (Virnekάs et al. 1994).

4.1.3 Generating of small B7.1 and B7.2 binding peptides/ fragments (microproteins)

This metinod selects smαh antigen-binding peptides from random DNA sequence phage display libraries composed of cyclic peptides (Tn6, Tn8, Tn9, TnlO and HTS) and variants from human-origin proteins (Kunitz and endothelin libraries) ( described in US patent numbers 5,403,484 and 5,571,698 and 5,223,409 to Ladner et al.). These libraries allow to examine several hundred million compounds and useful 'hits' emerge. Both the structured peptide' hits' and the human origin 'hits' give rise to suitable clinical leads. To obtain ligands that bind both B7.1 and B7.2 the libraries are screened in three different ways:

1) Bind to B7.1 and elution of bound phage with soluble B7.2

2) Bind to B7.2 and elution of bound phage with soluble B7.1

3) Alternate binding to B7.1 (e.g. rounds 1 and 3) and B7.2 (e.g. rounds 2 and 4)

4.2. Classical immunization

In spite of the fact that we did not succeed in generating murine mAbs, reacting with both B7.1 and B7.2, upon immunization with B7.1 or B7.2 respectively, we generate crossreacting murine mAbs by immunizing homologous sites from B7.1 and B7.2. The sequences of human B7.1ED and B7.2ED were compaired by an alignement. Five regions from human B7.1ED and humanB7.2ED showed homology and peptides of these sites were synthesized.

Three peptides were derived from human B7.1 : B7peptide 1 : region 66-94 human B7.1

B7peptide2: region 137-148 human B7.1 B7peptide4: region 200-212 human B7.1 Two peptides were derived from human B7.2: B7peptide3: region 124-135 human B7.2 B7peptide5: region 183-194 human B7.2 The peptides were synthesized on Tentαgel S resm (Rαpp Polymere GmbH Germany) usmg a Rainin Symphony/Multiplex synthesizer with standard Fmoc-chemistry and HOBtiTBTU (TBTU 2-( lH-Benzotnazole-l -yl) -1, 1 3 3-tetramethyluromum tefrafluoroborate, HOBt N-Hydroxybenzotriazole) m situ activation Standard double couphngs were performed usmg a 4 fold excess of ammo acids the presence of equimolar amounts of HOBt and TBTU for 2 x 20 minutes The Fmoc protecting group was removed by mild base treatment usmg 2%pιperιdιne/2%DBU (l,8-Dιazabιcyclo[5 4 0]undec-7-ene) m DMF (Dimethylformamide) Biotinylation was performed by dissolving biotin m 30% dimethylsulfoxide /

70% dimethylformamide with m situ activation

After completion of the peptide-resm complex the peptide is cleaved of the resm by incubating for 2 5 h with 90% tπfluoroaceuc acid/ 5% thιoamsole/3% ethanedιthιol/2% amsole The peptide is precipitated from the cleaving mixture usmg t-burylmethylether After centnfugation the pellet is washed 3 times with t-butylmethylether and dried overnight under vacuum The purity of the crude peptide is checked on RP-HPLC

The following B7 peptides were synthesized B7peptidel (IGP1458) Acetvl WQKEKKMVLTGGXGGEYKNRTIFD CONH2

B7peptide2 (IGP1459) Acetyl LSVKADFPTPSI - CONH2 B7peptide3(IGP1460) Acetyl LSVLANFSQPEI CONH2

B7peptιde4 (IGP1461) Acetyl SQDPETELYAVS CONH2 B7peptide5 (IGP1462) Acetyl SQDNVTELYDVS - CONH2

B7Peptιdel B7peptιde2 B7peptιde3 and B7peptιde4 are biotiniiated peptides

B7peptιdel was synthesized as ajoranched peptide

These peptides are able to induce an immune response in the mice

For this purpose mice are immunized with these different peptides containing overlapping sites from B7 1 and B7 2 A classical immunization scheme of 4 injections at 4 wee S interval is fohowed Different weh known routes of r ection different carriers and different adjuvants are tested. Bleedings are obtained at day 0 (pre immuun serum), day 66 (bleeding 1) and at day 94 (bleeding 2). Pre and post immuun sera are tested in ELISA for reactivity with B7.1 and B7.2 using the B7.1ED and B7.2ED fusion proteins. Mice developping antibodies reacting with both B7.1 and B7.2 are boosted and sacrificed 3 days after the boost.

The fusion is performed using a standard protocol. Briefly, splenic cehs from the immune mouse are fused with SP2/0 myeloma cells in ration 10:3 using PEG/DMSO. Fused cehs are plated in 96-weh plates at density of 5 x 10³-10⁴ cehs per weh and the first screening is performed 7 to 10 days later by ELISA for reactivity with B7.1 and B7.2. Positive clones are subcloned by limiting dilution and tested for reactivity with B7.1 and B7.2. Those hybridomas reacting with both B7.1 and B7.2 are selected for further characterization.

The selected hybridomas are then tested for their affinity for B7.1 and B7.2 in BIAcore experiments. They are also tested in an in vitro system of humanT ceh-APC interaction, in order to show their efficiency as blocking agents for allogeneic T ceh activation, ability to induce anergy, and ability to exclude any direct stimulating effect on the APC. The efficacy of these mAbs as immunosuppressive agents is tested in in vitro models as outlined further.

5. Affinity maturation of anti-B7.12, anti-B7.1 and anti-B7.2 by

Parsimonious Mutagenesis

The generation of humanized versions of rodent Mabs has successfully been achieved by various methods. Humanized Mabs exhibit the same ligand- binding properties as the original rodent Mabs, but in general with severely decreased binding affinity. In addition, human Mabs (hulvlabs) derived from ncrive human libraries only have primary response-level affinities, i.e., sub-micromolar. However, such affinities are sthl 2-3 orders of magnitude below what is usually required for therapeutic efficacy, and affinity maturation will have to be completed m vitro. The affinity of the newly generated anti-B7.12 mAbs is maturated using m vitro mutagenesis techniques. Also the anti-B7.1 and anti-B7.2 scFv, necessary for the different constructs (see further) are affinity maturated. In vivo antibodies are affinity matured in a stepwise fashion, by gradually incorporating mutations that cause small incremental improvement of the affinity

(for review see Hoogenboom H., 1997). Mutations affect affinity indirectly by influencing the positioning of side chains contacting the antigen, by providing new contact residues (particularly when they are located in or near the centre of the antigen combining site), or by replacing 'repulsive' or low affinity contact residues with contact residues with more favourable energetics. The in vitro affinity maturation process essentially involves three steps, ( 1) the introduction of diversity in the antibody V-genes, creating a 'secondaryi library', (2) the selection of the higher affinity from the low affinity variants, and (3) the screening to allow discrimmation of antibody variants with differences in affinity or kinetics of binding. The selection is chosen to favour kinetic parameters such as off-rate or affinity; this hinges on the use of hmited and decreasing amounts of antigen, and on performing the selections in solution rather than by avidity-prone panning. For example antibodies of the highest affinity can be preferentially selected by using the antigen concentration at or below the desired dissociation constant. The areas that are best targeted for mutagenesis differ for each individual antibody. Diversity is introduced either more-or-less randomly, as with error-prone PCR, by using mutator strains, by V-gene chain shuffling, or by DNA shuffling. Chain shuffling is used successfully to obtain a 300-fold increase in the affinity of an anti-hapten antibody, originally isolated from a naive human scFv library. In this experiment first the hght chain, then the heavy chain segment (without CDR3) is replaced by a repertoire of human partner domains somatically mutated in vivo, and from these repertoires the best variants are selected on antigen A similar approach is applied to antibodies binding to protem antigens, with a more modest 6-fold affinity improvement for an anti-c-erbB2 antibody. Cha shuffling appears to be the method of choice when the starting dissociation constant is > 10 nlvl or the koff is faster than 10 ^J sec ', but is still useful even for high affinity antibodies to identify those residues in the antibody involved in contacting the antigen, which could in a second step by (partially) randomised. For higher affinity antibodies (Kd < 10 nM or koff < 10 ^~3 sec ^"1) the antibody is CDR regions is targeted, using oligonucleotide directed mutagenesis or spiked oligonucleotides and PCR. With over 100 CDR residues constituting the antigen combining site, a choice is made as to where to start. If the structure of the antibody-antigen complex is known, residues that contact the antigen or may influence other residues contacting the antigen are targeted. Such residues are also determined experimentally, by chain shuffling, by alcmine-scanning of the CDR-regions, by parsimonious mutagenesis (see below), or by modelling. Residues involved in maintaining the main chain conformation of the CDR are conserved; residues that modulate affinity are randomized, ideally 4-6 residues at a time to allow efficient sampling of the sequence space. Sequential targeting is preferred, as additive effects of mutants obtained by targeting CDRis in parallel and combining mutations prove frequently unpredictable. For the different antibodies that are considered for affinity maturation, the following schedules are fohowed.

In theCTLA-4 element containing antibodies the shuffle between CTLA-4 and antibody elements yields Fv-like molecules with an Ig-fold that bind to both B7.1 and B7.2. The affinity of these molecules is low, around 10 to 100 nM for the Kd. The maturation concerns antibody-like molecules, in which it remains unclear which areas of the molecule are involved in antigen binding. Therefore the first V-gene segment is exchanged via DNA-shuffling (basically similar to the first experiments, now with more defined subsets of V-gene derived segments), and in a parallel approach, the XLl-Red mutator strain. These methods guarantee the highest possible affinity gain without needing the 'educated guess' that is required to apply oligo-directed mutagenesis or Parsimonious Mutagenesis (see below). These experiments also indicate v/hich CDR- and FR-residues are involved in binding, information which aid in the construction of libraries using spiked oligonucleotides and P.M. and re-selection of the repertoire. The maturation of the human B7.1 /2 antibodies concerns genuine antibodies, with dual binding activity, which facilitate choice of mutagenesis method. The method that is applied here depends on the affinity of the starting antibody. If the off-rate if higher than 10^"3 s^'1, the choice wih be on chain shuffling, because this is a fast and straight forward procedure, with the highest chance to retrieve antibodies with improvements beyond this treshold. If the off-rate is slower, or has been made slower using this procedure, in the next step parsimonious mutagenesis is applied.

Both the heavy and hght chain CDR'ε are targeted, in a sequential way, to derive antibody variants, which are than selected using limited amounts of biotinylated antigens (retaining parallel dual selections to retain affinity for both target. Initially both H3 and L3 are targeted; however, other CDR's also have to be diversified, because the binding site is extraordinary broad due to the dual binding capacity.

The P.M. procedure is explained further in the text.

It is necessary to affinity mature the murine B7.1 and B7.2 antibodies beyond the values that are normally found in antibodies, made by the natural immune system, as these antibodie are used to build larger molecules with two adjacent binding sites based on the murine Fv fragments.

The method of choice depends on the starting affinity, and fohowε the previous schedule (chain shuffle mutator strain) and as an alternative making use of spiked oligonucleotides to mutate H3 and L3.

Affinity maturation in vitro is typically attempted by chain shuffling, or by random mutagenesis of the antibody CDRs and screening or selection for higher affinity variants. Either complete randomization or error-prone PCR is usually used as the mutational operator, but these are extremely inefficient probes for searching protein sequence space. To search protein sequence space as efficiently as possible we have developed a proprietary computer-assisted method for otigonucleotide-directed scanning mutagenesis, called Parsimonious Mutagenesis (PM; Balint and Larrick, 1993; Schier et al., 1996). The efticiency of PM is based on a rational reduction in the size of the seσuence space which must be searched to one that can be completely encompassed in manageable libraries, and thoroughly searched with available screening techniques. This rational reduction of the sequence space is based on several observations. From the available X-ray crystαhogrαphic data, between 15 and 22 residues of a protein antigen are typically in contact with a similar number of residues in the Ab combining site (Novotny et al., 1989; Laver et al., 1990; Novotny, 1991). However, energy calculations and mutational studies indicate that a subset of only —30% of the contact residues, or 5-8 on average, contribute most of the binding energy (Novotny et al., 1989; Novotny, 1991; Mylvaganam et al., 1991; Tulip et al., 1992; Nuss et al., 1993). The remaining contact surface, therefore, presents multiple opportunities to develop additional high-affinity contacts, needing only a means to identify them. Consistent with the observation that most of the binding energy is concentrated in a few contacts, other studies have shown that affinities can often be profoundly altered by subtle changes in structure. For example, incremental contributions to binding energy of typical amino acid side chain groups have been measured as the comparative affinities of tigands which differ only by the presence or absence of the group. Such studies show that loss of a methyl group in a high-energy contact by replacement of Ala with Gly, Thr with Ser, or he with Val, could cost up to nearly three logs in binding energy. The same is true for the loss of a key carboxylate, as by replacement of Asp with Ala. Loss of a key hydroxyl by replacement of a Ser with Ala, or Tyr with Phe, could cost nearly six logs. In a hydrogen bond, gain or loss of 1 A between proton donor and acceptor could be worth — 1 kcal or a factor of ~ 5 in affinity. These observations suggest that significant gains in antibody affinity may be achieved by a few key substitutions. To define a fully-accessible sequence space which is maximally enriched for sequences which contain new high-affinity contacts and retain ah of the parental high-affinity contacts, we make use of two additional observations: (i) most high-affinity contacts are concentrated in the CDR3s, comprising 20-30 sites and

(ii) the genetic code has evolved to maximize the frequency of adaptive substitutions among the 5-7 substitutions for each amino acid which may result from single base changes. Based on the foregoing we define the relevant sequence space for at least the first round of antibody affinity optimization as including all permutations of 1-5 mutations in up to 30 sites of the heavy and light chain CDR3s, and including ah possible combinations of one-base amino acid substitutions at each site. This comprises a sequence space of ~ 10⁹ different sequences. Using PM, it is possible to construct a library which would contain the entire sequence space in — 10⁹ clones, and using phage display it is possible to pan the entire library against immobilized antigen. To assist with the construction of PM libraries, we have developed a computer program, called PMCAD, for the design of mutagenic oligonucleotides. The program computes optimum nucleotide (nt) mixtures for each position to be mutagenized in the oligonucleotides, based on the parent amino acid (aa) sequence, the user-selected mutation frequency distribution, and the desired sets of alternative amino acids. In this case we wish to maximize the abundance of 5-hit mutants so that ah permutations with ah possible combinations of up to 6 substitute aa at each site would be represented in a library of ~ 10⁹. Such a library automatically contains ah possible 6-aa combinations for ah mutants with 1-4 hits also. A mutation frequency of 16.7% at each of 30 aa positions produces a binomial distribution in which 5-hit mutants comprise 19% of the library and mutants with 1-5 hits comprise 61% of the library. Substitute amino acid sets are controlled by using degenerate (doping) codons which may be selected to encode mixtures of more and less conservative analogs of the parent amino acid. NNT and NNG codons are particularly useful in this case because for most aa, they encode six different equally frequent 1-base codon changes to more and less conservative analogs. As is typical of doping codons in PM, the frequencies of 2- and 3-base codon changes is so low that the 1-base frequencies are only —5% below what they would be if they were the only substitutions. We have previously shown that PM is particularly adept at identifying positions which can be improved most easily (Schier et al., 1996). Positions which were repeatedly altered in selected higher-affinity variants could be further improved by saturation mutagenesis. At such sites the parent residue may make a negative contribution to affinity which could be relieved by multiple alternatives. However, more intensive mutagenesis may be requfred to reveal the best alternative at each site. This is consistent with the contention of Novotny ( 1989, Biochem. 28, 4735; 1991,

Molec. Immunol. 28, 201) that from crystal structures even the highest affinity antibodies can be seen to make repulsive contacts with the antigen High affinity Mab fragments obtained by PM and expressed on the phage ceh surface only constitute a very small fraction of the total ceh population We therefore use various enrichment procedures These enrichment procedures are ah based on functional recognition of the specific antigen by the Mab fragment expressed on the surface of the phage

Selected putative high affinity Mab fragments are analyzed by BIAcore usmg B7.1ED and B7 2ED fusion protems Alternatively, additional PM primers are designed for the generation of new libraries, fohowed by a new round of selection for high affinity Mabs

6 Generation ofBiTAb, Diabody and Triabody molecules

The overall com of the present invention is to generate a pharmaceutical composition that can simultaneously block the B7 1-CD28 and the B7 2-CD28 costimulatory pathways

7 1 Generation of BiTAb (anti-B7 l/anti-B7 2, anti-B7 12/anti-B7 12) molecules

As an alternative to a diabody and a triabody molecule, a BiTAb molecule can be constructed A BiTAb molecule has a less rigid structure compared to a diabody and a triabody, and cross-lmks and/or cross-reacts with the costimulatory molecules B7 1 and B7 2 expressed on the membrane of professional antigen- presenting cells, leading to the inhibition of antigen-specific T cell acuvation A BiTAb molecule b ds m a bivalent fashion to B7 1 or B7 12 on the one hand and to B7 2 or to B7 12 on the other hand Tne Dindmg affinity of the individual scFv s will be maturated using Parsimonious Mutagenesis (see anove) Together with the divalent bmdmg properties of this molecule we end up with a molecule with a very slow off-rate, bemg an ideal blocking agent for the application m vivo The intrinsic affinity of the different binding domains of the antibody constructs for their respective antigens is of mmor importance as both antigens normally appear as membrane bound components thus promoting multivalent binding of the construct if the antigens are present on the same ceh surface. A BiTAb is a Bispecific Tetravalent Antibody molecule. The molecule consists of 4 scFv's; two anti B7.1 scFv's and two anti B7.2 scFv's (tetravalency). One single BiTAb is a homodimer of two identical molecules, each containing both an anti B7.1 and anti B7.2 scFv (bispecificity). An anti B7.1 and an anti B7.2 scFv are linked together using a dimerisation domain, which drives the homodimerisation of the molecule (see figure 14). The VH and VL regions of the anti B7.1 and anti B7.2 Mab's are cloned from their respective hybridoma' s by RT-PCR with a set of degenerate primers, for example the ones used in the Pharmacia RPAS Mouse scFv module. In a second PCR, both VH and VL are linked using a short synthetic linker. After cleavage with the appropriate restriction sites, the scFv coding sequences are ligated in a scFv expressionvector such as pCANTAB5E (Pharmacia). To select for functional scFv's, phages wih be generated after ligation in pCANTAB5E. These phages are panned against B7.1 or B7.2 positive ceh lines. Binding phages are screened for their binding capacity in a FACS analysis with the respective B7.1 or B7.2 positive cell-lines. Correct binders are selected for further DNA sequence analysis. Anti B7.1 (B7-24) and B7.2 (IGIO) scFv's are used as building blocks to generate the B7.1/B7.2 BiTAbs using standard recombinant DNA techniques. A single BiTAb subunit starts with an anti-

B7.1 scFv or an anti-B7.2 scFv fohowed by a dimerisation domain flanked by flexible linkers. The dimerisation domain in its turn links C-terminahy to the anti-B7.2 scFv or the anti-B7. IscFv. Finally a detection and purification tag is added at the extreme C-terminus of the molecule. The sequence coding for the dimerisation domain and the flanking linkers is made synthetically using the metinod described by Stemmer ei al. (1995). Herefor we considered the optimal codon usage for E.Coii-expression. This synthetic domain is subsequently linked to both the anti B7.1 and anti B7.2 (with tag) scFv's. As tinkers betv/een the dimerisation domain and the scFv's we use either the widely used (G₄S)₃ sequence (Hoogenboom er al, 1991) or the flexible and proteolysis-resistant truncated human IgG3 upper hinge region (Pack & Plύckthun, 1992). As dimerisation domain we used a leucine zipper type of domain (Kostelny et al, 1992; de Kruif & Logtenberg, 1996) or the helix-turn- helix motif described by Pack et al. (1993). An alternative dimerisation domain whereby a possible immunogenic reaction is reduced/avoided, is the JEM-1- peptide. This human peptide shows a 'leucine zipper' dimerisation motif with limited homology to Fos Jun proteins (Duprez et al. 1997). A Cystein could be included to enhance stability. As the C-terminal detection and purification tag we use a hexahistidine sequence. For immunogenic reasons we decided to make BiTAb molecules both with or without a Hisδ tag. The sequences are assembled in such a way that functional domains are easily replaceable using unique restriction sites present in the molecule.

The BiTAb coding sequence is inserted in anE.coϋ expressionvector with an appropriate secretion signal such as pelB or ompA signal sequence and is placed under control of an inducible promotor such as P_lac, ?,_._. or P_L. The BiTAb expressionplasmid is introduced into suitable E.coli expressionstrctins such as JM83 or HB2151. Expression levels in the E.coli periplasm and extent of dimerisation is analysed using small-scale expression experiments. Coπectly dimerised BiTAbs are purified from the E.coli periplasm using IMAC chromatography. The DNA sequence and the protein sequence of the described BiTAbB7-24- 1G10H6 are represented in figure 15 and 16, respectively and the sequences of

BiTAb IGl 0-B7-24H6 in figure 17 and 18, respectively.

7.1.1. Practical information

7.1.1.1. Single-step assembly of the dimerisation domain consisting of a helix-turn- helix domain flanked by the truncated human IgG3 hinge region as a flexible linker: HDH-domain (figure 19 and 20).

Equal amounts ( l Opmoles) from each of the 10 ohgo's (primer 7624 to 7633) were combined and the mixture was subsequently diluted 100-fold in 50μl PCR mix containing lOmM Tris-HCl pH9/ 2.2mM MgClø 50mM KCl/ 0.2mM each dNTP / 0.1% TritonX- 100/ 1 U of Tαq polymerase/ 0.1 U of Pfu polymerase. The PCR program consisted of 35 cycles at 50°C for 30s. The obtained gene was amplified in a second PCR reaction. The gene assembly reaction mixture was dhuted 40-fold in 100 μl PCR mix containing lOmM Tris-HCl pH9/ 2.2mM MgCl₂/ 50mM KCl/ 0.2mM each dNTP / 0.1% TritonX- 100/ 5U of Taq polymerase/ 0.1U of Pfu polymerase/ 2 outside primers (primers 7622 and 7623) at a concentration of 1 μM. The PCR program consisted of 23 cycles at 48°C for 30s. The gene-amplified DNA was purified and ligated in the pGEM-T-vector. The assembled DNA sequence was subjected to DNA sequence analysis.

List of used primers:

7622: 5' -CGCGCTCGAGATCAAACGGACC- 3'

7623: 5' -CGCGGAATTCGCGTTCGCGACTAG - 3'

7624: 5' -CGCGCTCGAGATCAAACGGACCCCGCTGGGTGATACCACTC- 3' 7625: 5' -CAGTTCACCTCCGGAGGTATGAGTGGTATCACCCAGCGGG- 3'

7626: 5' -ATACCTCCGGAGGTGAACTGGAAGAGCTGTTGAAACATCT- 3' 7627: 5' -GACCTT CAGCAGTTCTTTCAGATGTTTCAACAGCTCTTC - 3' 7628: 5' -GAAAGAACTGCTGAAAGGTCCGCGGAAAGGTGAACTGGAG - 3' 7629: 5' -TTCAGGTGC TCAGCAATTCCrCCAGTTCACCTTTCCGCG - 3' 7630: 5' -GAATTGCTGAAGCACCTGAAAGAGCTGTTGAAAGGTACCC -3'

7631: 5' -ATGGGTAGTATCACCCAGGGGGGTACCTTTCAACAGCTCT - 3' 7632: 5' -CCCTAGGTGATACTACCCATACCAGCGGTCAGGTGCAACT -3' 7633: 5' -CGCGGAATTCGCGTTCGCGACTAGTTGCACCTGACCGCTGGT-

7.1.1.2. Introduction of an appropriate restrictionsite just before the start of the VH- domatin of the gene encoding the scFvB7-24 or scFv 1G10.

To obtain a perfect in-phase fusion of the gene encoding the C-terminal scFvB7-24 or scFvlGlO to the gene encoding the HDH-domain in the gene encoding BiTAbB7- 24-1G10H6 or BiTAb 1G10-B7-24H6 respectively, α Spel restrictionsite was introduced just before the start of the gene encoding the VH-domain of the gene encoding scFvB7-24 and scFvlGlO. This is done by PCR with the appropriate primers (primers 8067 and 8075). The PCR fragment was ligated in the pGEM-T vector and subjected to DNA sequence analysis.

List of used primers:

8067: 5' -GGACTAGTTCAGGTGCAGCTACAGCAGTCTG -3' 8075: 5' -GCCAGTGAATTCTATTAGTGGTGATG -3'

7.1.1.3. Construction of BiTAbB7-24-lG10H6 subunit

For the construction of BTTABB7-24-1G10H6 subunit, three fragments were ligated in-phase, namely the gene encoding the N-terminal scFvB7-24, the gene encoding the HDH-domctin and the gene encoding the C-terminal scFvl Gl 0H6. In that way the gene encoding BiTAbB7-24-lG10H6 subunit, preceded by the pelB secretion signal, was cloned in an E.coli expressionvector under control of the lac-promotor.

Fragment 1: pscFvB7-24H6 cleaved with XhoI-EcoRI

Fragment 2: pGEM-THDH cleaved with Xhol-Spel Fragment 3: pGEM-TscFv 1 G 1 Os-e cleaved with Spel-EcoRI

This expressionplasmid pBiTAbB7-24-lG10H6 was transformed in the expression strain JM83.

7.1.1.4. Expression of BiTAbB7-24- IGl 0H6 in JM83.

. n overnight culture of pBiTAbB7-24-lG10H6 in JM83 was 20x diluted in LB÷ lOOμg/ml amp+ 1% glucose and incubated at 28°C until an OD600 of 0.5-0.6 was reached. After removing the glucose of the medium, the culture was induced with 0. ImM IPTG and furtiner incubated at 28°C during ± 18h. Periplasmic fractions were prepared as explained before. Only a very small amount was secreted into the periplasm and allmost ah the expressed protein formed insoluble cytoplasmic inclusion bodies. This was confirmed by NH₂-terminal sequence analysis whereby the pelB secretion signal was still observed. This means that BiTAbB7-24-lG10H6 couldn't be weh secreted into the periplasm. Therefor, we cloned the gene encoding BiTAbB7-24-lG10H6 and the gene in the opposite sense, encoding BiTAb 1 G 10-B7-24H6 in the pIGFti2-expressionvector (Innogenetics) without the pelB- secretion signal. This means that we are purifying BiTAb-molecules from cytoplasmic inclusion bodies .

7.1.1.5. Construction of the gene encoding BiTAbB7-24- 1 G 10H6 without a secretion signal.

By PCR with specific primers 8299 and 8301, the gene encoding BiTAbB7-24- 1G10H6 was isolated in such a way that the pelB secretion signal was removed and that specific sites were created for ligation in the E.coli expressionvector pIGRI-2. The obtained PCR fragment was ligated in the pBSK(+) and subjected to

DNA sequence analysis. For the construction of pIGRI-2BiTAbB7-24-lG10H6, three fragments were isolated and fused tin-phase. 1. pIGRI-2: blunted Ncol ÷ Sail .2. _PBSKBiTAbB7-24-lG10H6: blunted Sapl + BspEI 3. PBSKBiTABB7-24- 1 G 10H6 : BspEI ÷ Sail

By ligation of the three fragments, the gene encoding BiTAbB7-24-lG10H6 without a secretion signal was cloned under control of the pL-promotor and the obtained plasmid was transformed to the E.coh expressionstrain MC1061(pAcI), SG4044(pAcI), UT5600(pAcI). After induction, cytoplasmic inclusion bodies will be formed and purification from inclusion bodies wih be performed.

List of used primers:

8299: 5' -GGCCGCTCTTCGCAGCTACAGGAGTCTGG -3' 8301: 5' -CGACGTCGACTATTAGTGGTGATGGTG -3' 7.1.1.6. Single-step assembly of JEM-1 peptide using the method of Stemmer et al. (Figure 21 and 22)

This human peptide shows a 'leucine zipper' dimerisation motif with hmited homology to Fos/Jun proteins.

Equal amounts (lOpmoles) from each of the ohgo's (primers 8528 to 8534) were combined and the mixture was subsequently diluted 100-fold in 50μl PCR mix containing 1 OmM Tris-HCl pH9/ 2.2mM MgCl₂/ 50mM KCl/ 0.2mM each dNTP / 0.1 % TritonX- 100/ 1 U of Taq polymerase/ 0.1 U of Pfu polymerase. The PCR program consisted of 30 cycles at 52°C for 30s. The obtained gene was amplified in a second

PCR reaction. The gene assembly reaction mixture was diluted 40-fold in 100 μl PCR mix containing 1 OmM Tris-HCl pH9/ 2.2mM MgCl₂/ 50mM KCl/ 0.2mM each dNTP / 0.1% TritonX-100/ 5U of Taq polymerase/ 0.1U of Pfu polymerase/ 2 outside primers (primers 8526 and 8527) at a concentration of 1 μM. The PCR program consisted of 30 cycles at 50°C for 30s. The gene-amplified DNA was purified and ligated in the pGEM-T-vector. The assembled DNA sequence was subjected to DNA sequence analysis.

The dimerisation domain HDH in the BiTAb molecule wih be replaced by this human dimerisation domain to reduce immugenecity.

List of used primers:

8526: 5' -CGCGTCCGGAGACCTGCAGTAC -3' 8527: 5' -CGCGCCTAGGGGGGTCTGTTC -3' 8528: 5' -CGCGTCCGGAGACCTGCAGTACCACTCGAACGTC -3' 8229: 5' -CGCGCCTAGGGGGGTCTGTTCAGACAGCTGCG -3'

8230: 5' -TGGCGCGTGAAAAAAACCAGCTGATCCTGGAAAACGAAGC -3' 8231: 5' -GCTGGGTCGTAACACCGCGCAGCTGTCTGAACAGACCCCC -3' 8232: 5' -CGGTGTTACGACCCAGCGCTTCGTTTTCCAGGATCAGCTG -3' 8233: 5' -GTTTTTTTCACGCGCCAGACGTTCGAAGTGGTACTGCAGG -3' 8234: 5' -CGCCGCCCCTAGGGGGGTCTGTTCAGACAGCTGCG -3' The functional affinity of the BiTAb constructs is measured using surface plasmon resonance (SPR) analysis measured with an optical biosensor (BIAcore from Pharmacia). As it wih be difficult to mimic the density, the proportion, the distribution and the membrane diffusion capacity of the 2 antigens during the Biacore measurements, a qualitative analysis of the binding of these constructs is performed on the 3 antigens separately and on combinations of fixed concentrations of the 2 molecules. The binding curves are compared with the binding curves obtained with the monospecific antibody constructs (anti-B7.2 antibody, anti-B7.1 antibody or anti-B7.12 scFv). These results indicate whether the BiTAb constructs have an advantage over monospecific holoantibodies with respect to their potential membrane binding. The final evaluation is done on the basis of the immunophysiological properties of these molecules.

7.2. Generation of diabody molecules (anti-B7. l/anti-B7.2 and anti-B7.12/anti- B7.12)

Diabodies are dimeric antibody fragments. In each polypeptide, a heavy- chain variable domain (V_H) is linked to a light-chain variable domain (V_L) but unlike scFv's, each antigen-binding site is formed by pairing of one V_H and one V_L domain from the two different polypeptides. This is achieved by shortening the linker between the V_H and V_L domains in each molecule (Holtiger et al., 1993). Since diabodies have two antigen-binding sites they can be bispecific. Mono- or bi-specific bivalent molecules are generated by shortening the flexible linker seσuence between the VH and VL of the anti-B7.1 scFv B7-24, the anti-B7.2 scFv IGIO and the scFv molecule with dual specificity for B7.1 and B7.2 (B7.12) to between five and ten residues (Gly4Ser to (Gly4Ser)2) and for the bi-specific molecules by cross-pairing the variable heavy and tight chain domains from the tv/o scFvs with different antigen recognition (B7.1/B7.2 and B7.12/B7.12). As a first trial we started the construction of monospecific diabodies B7-24 and 1G10 (figure 23,24 and 25, 26). As an example for the different steps involved in such a construction, v/e have documented the construction of a anti-B7.1 diabody B7-24. By PCR with specific primers the linker in scFvB7-24 was reduced to 5 or 10 aminoacids. Herefor, the VH domain of scFvB7-24 was amplified with primers 8218 and 8070( 5aa) ,and the VL domain of scFvB7-24 with primers 8073 and 8075 . In this process a Sapl restriction site was introduced at the N-terminus of the VH-domctin and a EcoRI restriction site was introduced at the C-terminus of the VL domain.

Both PCR fragments were annealed and amplified by making a mix of equal amounts of both fragments, primers 8218 and 8075, 200μM dNTP, 1U Pfu- polymerase and 1 Ox Pfu-buffer. The annealed fragment of 804 bp was ligated in the pBSK(+) and subjected to DNA sequence analysis. The gene scFvB7-24L5H6, preceded by the pelB-secretion signal, was cloned in the pTrc99A expressionvector under control of the IPTG inducible pTrc-promotor and transformed to the JM83 E.coli expressionstrain. This was done by cleavage of pBSKB7-24L5H6 with Sapl, fohowed by T4polymerase and EcoRI . The vector was cleaved with Ncol , fohowed by T4polymerase and EcoRI. Both fragments were ligated to obtain the expressionvector pTrcB7-24L5H6.

The same procedure was used for the construction of the three other expressionplasmids using the appropriate primers, namely pTrcB7-24L10H6 (primers 8218, 8071, 8074 and 8075), pTrclG10L5H6 (primers 8218, 8070, 8073 and 8075) and pTrclG10L10H6(primers 8218, 8071, 8074 and 8075). A Cystein could be included to enhance stability.

Ah 4 expressionplasmids were transformed to JM83. An overnight culture was 20x diluted in LB+ lOOμg/ml amp+ 1% glucose and incubated at 28°C until an OD600 of 0.5 was reached. After removing the glucose of the medium, the culture was induced with 0. ImM IPTG and further incubated at 28°C during ± 18h. Periplasmic fractions were prepared as explained before. Only a small amount (10-30%) was secreted into the periplasm

List of used primers:

8218:

5' -GGCCGCTCTTCGAAATACCTATTGCCTACGGCAG -3' 8070: 5'CTGAGTGAGCTCGATGTCCGATCCGCCACCGCCTGAGGAGACGGTGACCGT GGTC -3'

8071: 5'CTGAGTGAGCTCGATGTCCGATCCGCCACCGCCAGAGCCACCTCCGCCTGA

GGAGACGGTGACCGTGGTC -3'

8073: S'GTCACCGTCTCCTCAGGCGGAGGTGGCTCTGACATCGAGCTCACTCAGTCTC C -3' 8074:

S'GTCACCGTCTCCTCAGGCGGAGGTGGCTCTGGCGGTGGCGGATCGGACAT CGAGCTCACTCAGTCTCC -3'

8075: 5' -GCCAGTGAATTCTATTAGTGGTGATG -3'.

The produced bivalent monospecific scFv B7.24 diabody molecules are purified from the periplasmic extract .

The bispecific diabodies (figure 27,28,29,30) are constructed otin a simhlar way as described for the monospecific diabodies.

7.3. Generation of triabody molecules (B7.1/B7.1/B7.2 or B7.1/B7.2/B7.2 or B7.12/B7.12/B7.12)

The diabody construct can target only two molecules being B7.1 and B7.2 or

B7.12 and B7.12 on the APC. A construct, which could target three molecules, e.g two B7.2 and one B7.1 molecule involved in this costimulatory pathway, namely a triabody, is aimed in this invention. A triabody is a mono-, a bi- or a trispecitic molecule recognizing simultaneously e.g two B7.2 and one B7.1 molecules. Similar to the diabody, the triabody is a molecule with a rigid structure that prevents simultaneously binding to the three targets and so prevents the activation of the APC. Each antigen-binding site is formed by pairing of one V_H and one V_L domain from the same or from two different polypeptides. This is achieved by shortening the linker between the V_H and V_L domains in each molecule to between - 1 to 5 aa, preferable to 0 or -1 (refering numbering of Kabat). Additional association domains or disulfide bridges are incorporated in order to drive triabody formation as well as to maintain stability. As this construct can only bind monovalently to the target molecules, it means that, as for the diabody construct, the binding of the ScFv components in the triabody construct must be of high affinity. It is therefore important to increase the binding affinity of the individual ScFv's (see also diabody).

As a first trial we started the construction of monospecific triabodies B7-24 and IGIO (figure 31,32 and 33,34). As an example for the different steps tinvolved in such a construction, we have documented the construction of a anti-B7.1 triabody B7- 24. By PCR with specific primers the linker in scFvB7-24 was removed. Herefor, the VH domain of scFvB7-24 was amplified with primers 8218 and 8069, and the VL domain of scFvB7-24 with primers 8072 and 8075. In this process a Sapl restriction site was introduced at the N-terminus of the VH-domain and a EcoRti restriction site was introduced at the C-terminus of the VL domain. Both PCR fragments were annealed and amplified by making a mix of equal amounts of both fragments, primers and , 200μM dNTP, 1U Pfu-polymerase and lOx Pfu-buffer. The annealed fragment of bp was ligated in the pBSK(-τ-)-vector and subjected to DNA sequence analysis.

The gene scFvB7-24L0H6, preceded by the pelB-secretion signal, was cloned in the pTrc99A expressionvector under control of the IPTG inducible pTrc-promoior. This was done by cleavage of pBSKB7-24L0H6 with Sapl, fohowed by T4polymerase and EcoRI . The vector was cleaved with Ncol , followed by T4polymerase and EcoRI. Both fragments were ligated to obtain the expressionvector pTrcB7-24L0H6. The same procedure was used for the construction of the gene encoding the subunit of the monospecific triabody 1G10 using the appropriate primers 8218, 8069, 8072 and 8075.

Both expressionplasmids, pTrcB7-24L0H6 and pTrclG10L0H6 were transformed to JM83. An overnight culture was 20x diluted in LB+ lOOμg/ml amp+ 1% glucose and incubated at 28°C until an OD600 of 0.5 was reached. After removing the glucose of the medium, the culture was induced with 0. ImM IPTG and further incubated at 28⁼C during ± 18h. Periplasmic fractions were prepared as explained before. Only 5 a small amount ( 10-30%) was secreted into the periplasm

List of used primers:

8218: 5' -GGCCGCTCTTCGAAATACCTATTGCCTACGGCAG -3' o 8069: 5' -CTGAGTGAGCTCGATGTCTGAGGAGACGGTGACCGTGGTC -3'

8072: 5' -GTCACCGTCTCCTCAGACATCGAGCTCACTCAGTCTCC -3' 8075: 5' -GCCAGTGAATTCTATTAGTGGTGATG -3'

Most of the expressed proteins formed insoluble, periplasmic inclusion bodies. 5 The bispecific triabodies are constructed in a similar way as described for the monospecvific triabodies (figure 35,36 and37,38).

The multimeric behaviour of the purified and unpurified molecules is analyzed. The ability of the molecules to bind the different antigens is tested in ELISA, FACS, T 0 ceh proliferation assay and BIACORE as described under section 7.5.

7.4 Purification of diabodie and tήabodi molecules

Purification of diabodies and triabodies from the ScFv B7-24 molecule out of the 5 soluble periplasmic fraction occurred under native as weh as denaturating conditions. The soluble periplasmic fraction corresponding with 20 to 30% of the expressed product was purified on Ni^2_r-IDA under native conditions. The bound material was eluted by a 40 mM Imidazole wash step and a 200 mM Imidazole elution. The 200 mM Imidazole elution was divided (based upon SDS-PAGE CBB 0 staining) into 3 pools (PI, P2 and P3) for the ScFv M24 L5 and into 2 pools for the

ScFvM24 L0 (PI and P2). PI from both constructs ( ScFvM24 L5 and L0) contained the peαkfrαctions with the highest protein concentration but also with the highest amount impurities. P2 (ScFvM24 LO) and P3 (ScFv M24 L5) contained the tailing fractions of the elution peak, low in protein concentration but with a higher ScFv purity. P2 from ScFvM24 L5 contained the fractions in between PI and P3 with an 5 intermediate purity. Elution fractions were not as pure compared to the earlier native purification performed with ScFvM24. This can be caused due to the application of Ni^2÷ instead of Zn^2÷ or due to the fact that more E.coli host proteins are co-induced and co-expressed in the culture of the dia- and triabodies compared to the ScFV M24 culture. P2 and P3 from ScFvM24 L5 and P2 from l o ScFvM24 LO were analysed in FACS and ELISA and MLR.

P3 (ScFvM24 L5) and P2 (ScFvM24 LO) being the fractions with the highest purity , based on SDS-PAGE CBB staining, were concentrated and injected onto a Superdex 200 PC 3.2/30 gelfiltration column (Pharmacia) and compared with the gelfiltration profile of ScFvM24 LI 5. On the gelfiltration proftie of the ScFv M24 L5

15 (diabody) two important peaks could be detected: the one with highest absorbence eluting at the same place were ScFvM24 LI 5 eluted, being the elution place for a 27 kD protein, indication for monomeric material, the other one eluting at the elution place for a 54 kD protein, indicating that diabodies are formed. On the gelfiltration profile of the ScFvM24L0 also two important peaks could be detected

20 but with a much lower resolution. Both peaks had approximately the same absorbence, the first eluted at the place for a 54 kD protein (diabody) and the second one eluted a fraction earlier than the elution place for a 27 kD protein. (Figure 39: gelfiltration profile of a : ScFvM24 LO, b : ScFvM24 L5 and c : ScFvM24 LI 5). No clear peak at the supposed elution place for a triabody ÷l- 80 kD could be 25 detected at this applied concentration. In a second SEC run a lOfold more concentrated sample of ScFvM24 L5 and ScFvM24 LO has been applied on the same column under identical conditions. A decrease in resolution v/as observed but also a peak corresponding with the expected retention volume time for a triabody (MW 80 kD) appeared. For ah three chromatograms (ScFvM24 LI 5, L5

30 and L0) gelfiltration fractions 5 up to 17 were analysed in ELISA.

Most of the expressed proteins formed insoluble, periplasmic inclusion bodies. The multimeric behaviour of the purified molecules is analyzed. The ability of the purified molecules to bind the different antigens is tested in BIAcore and ELISA experiments as described under section 6.4.

7.5. Characterization of the diabody and triabody molecules

The ability of the purified and unpurified diabody and triabody molecules to bind their specific antigen is tested in ELISA using the B7.1ED and B7.2ED fusion proteins. The cell-binding properties of these molecules are determined in FACS experiments. The molecules are further analyzed for their capacity to block T ceh-

APC interactions.

Binding specificities of the unpurified monospecific scFvB7-24, B7-24 diabody (scFvB7-24 L5) and B7-24 triabody (scFvB7-24 LO) molecules and of the unpurified monospecific scFvlGl O, 1G10 diabody (scFvlGlO L5) and triabody (scFvlGlO LO) molecules were analysed in ELISA experiments using the B7.1ED and B7.2ED fusion proteins. As the periplasmic preparations of scFvlGl O and IGI O triabody (scFvlGlO LO) contained very low amounts of scFv as demonstrated on Western blot, these two extracts were firstly concentrated 3.5 times before analysis in ELISA. For the ELISA, rabbit anti-human IgG Fc fragment specific antibodies were coated on MaxiSorb plates at a concentration of 1 mg/ml in PBS buffer for 2h at 37°C or overnight at 4³C followed by blocking vith PBS 0.1% casein for lh at 37^DC. Subsequently, 100 ml of B7.1ED fusion protein (500ng/ml) orB7.2ED fusion protein (500ng/ml) in PBS 0.1% casein was added and incubated for lh at 3700 After three washes with PBS 0.05% Tween20, the different antibodies were added in a fourfold serial dilution in PBS 0.1% casein starting from 1/1 dilution for the unpurified recombinant antibody preparations and from 1 mg/ml for purified scFvB7-24 and for the B7-24 (positive control) and 1G10 mAbs (positive control). Crude periplasmic preparation of nontransformed JM83 cehs was included as a negative control. After three washes with PBS 0.05% Tween20, wells with the recombinant antibodies were incubated vith mouse anti-his monoclonal antibody at 1 mg/ml for lh at 37^rC fohowed by three wash steps with PBS 0.05% Tween20. Subsequently, the plates were incubated for lh at 37°C with biotinylated goat anti-mouse (IgG +IgM) antibodies at 1.5 mg/ml fohowed by five washes with PBS 0.05% Tween20 and incubation with peroxidase-conjugated streptavidin at 0.1 mg/ml for lh at 37°C. Color was generated by incubation with TMB for 30' at RT. The reaction was stopped by addition of 50 ml/well of 2N H₂S0₄ and the absorbance at 450 and 595 nm was measured in an ELISA reader. For detection of B7-24 and 1G10 binding, incubation with mouse anti-his monoclonal antibody was omitted whereas ah other incubation steps using antibodies and peroxidase-conjugated streptavidin were identical. OD values hsted out in the figures are corrected for the blank OD value (OD value obtained if no recombinant antibody, B7-24 or 1G10 mAb was included in the ELISA). Results (figure 40) show that the unpurified scFvB7-24, B7-24 diabodies (scFvB7-24 L5) and B7-24 triabodies (scFvB7-24 L0) were able to btind the indirectly coated B7.1ED fusion protein. Binding was specific to the B7.1ED fusion protein since no significant binding to B7.2ED fusion protein was detected . Unpurified scFvlGlO, 1G10 diabody (scFvlGl O L5) and 1G10 triabody (scFvlGlO

L0) were able to bind to indirectly coated B7.2ED fusion protein (tigure 41) but not to the B7.1ED fusion protein , demonstrating that these scFvlGlO recombinant antibodies bind specifically to the B7.2ED fusion protein.

The unpurified monospecific scFvB7-24, B7-24 diabody and B7-24 triabody molecules were further analysed for their binding capacity to the B7-1 molecules on the membrane of mouse 3T6-CD32-B7.1 cehs, which are transfected with the cDNA encoding human B7.1. The 3T6-CD32 cehs and 3T6-CD32 cehs transfected with the cDNA, of human B7.2 (3T6-CD32-B7.2) were included as a negative control. In addition, binding of these molecules on B7.1 expressing EBV transformed human B cehs, RPMI 8866 was evaluated. To this end, 3T6-CD32, 3T6-CD32-B7.1, 3T6-

CD32-B7.2 and RPMI 8866 cehs (5x10^f' cells/staining) were preincubated for 20' at 4^;C in 50 ml FACS buffer (PBS supplemented with 5% inactivated FCS and 0.02% sodium azide) supplemented with 10% normal rabbit serum. Tne different antibodies were added in a twofold serial dilution starting from 1/1 dilution for the periplasmic preparations of scFvB7-24, B7-24 diabody (scFvB7-24 L5) and B7-24 triabody (scFvB7-24 L0) and from 1 mg/ml for purified scFvB7-24 (positive control) and incubated for 20' at 4°C. Periplasmic preparation of nontransformed JM83 cehs at the same dilutions as described above was included as a negative control. After three washes in FACS medium, cehs were incubated with mouse anti-his monoclonal antibody at 2 mg/stctintiig for 20' at 4°C fohowed by 3 washes with FACS buffer. Subsequently, cehs were incubated with 1.4 mg/staining of fluorescein (FTTC)-conjugated goat anti-mouse IgG Fc fragment specific antibodies for 20' at 4°C, followed by three wash steps and analysis of the cehs for fluorescent stctinting using a FACScan instrument. Cehs only incubated with mouse anti-his antibody fohowed by incubation with FTTC-conjugated goat anti-mouse IgG Fc fragment specific antibodies were included as a negative control for staining.

FACS results using the periplasmic preparations at a twofold dilution and the purified scFvB24 at 1 mg/ml are tisted out in figure 42a-b-c. The thin line represents the negative control statining whereas bold lines represent staining with the different crude periplasmic preparations and with purified scFvB7-24. These results show that unpurified scFvB7-24, B7-24 diabodies (scFvB7-24 L5) and B7-24 triabodies (scFvB7-24 LO) bound to the two B7.1 expressing ceh lines:3T6-CD32-B7.1 cehs and RMPI8866 cells. Binding of these recombinant antibodies was found to be specific for B7.1 as they did not bind to 3T6-CD32 and to 3T6-CD32-B7.2 cehs, which do not express B7.1 (data not shown). Binding specificities of the semi-purified monospecific B7-24 diabody

(scFvB7-24 L5 fraction p2 and p3) and triabody (scFvB7-24 L0 fraction p2) w^*ere checked with two different ELISA settings using the B7.1ED and B7.2ED fusion proteins. In a first ELISA, rabbit anti-human Fc antibody was coated at a concentration of lμg/ml in PBS buffer for 2h at 37°C followed by blocking with PBS 0.1% casein for lh at 3TC. Subsequently 100 μl of B7.1ED fusion protein (500ng/ml) was added and incubated for lh at 37^DC. Purtiied scFvB7-24 and semi-purified B7- 24 diabody (scFvB7-24 L5 fraction p3) or B7-24 triabody (scFvB7-24 L0 fraction p2) molecules were added in a twofold serial dilution starting from 500 ng/ml to 7.5 ng/ml and incubated for lh at 37°C. After three washes with PBS 0.05% Tween20, wells were incubated for lh at 37°C with mouse anti-his monoclonal antibody fohowed by three washes and incubation for lh at 37°C with biotinylated goat anti- mouse (IgG +IgM) antibodies at 1.5 mg/ml. After five washes with PBS 0.05% Tween20, incubation with peroxidase-conjugated streptavidin at 0.1 mg/ml for lh at 37°C was performed. Color was generated by incubation with TMB for 30' at RT. The reaction was stopped by addition of 50 ml/well of 2N H₂S0₄ and the absorbance at 450 and 595 nm was measured in an ELISA reader.

Binding specificities of the semi-purified monospecific B7-24 diabody (fraction p2 and p3) and triabody (fraction p2) molecules were also evaluated in a second ELISA. In this ELISA, wehs were coated with mouse anti-his monoclonal antibodies ( 1 mg/ml) in PBS for 2h at 37°C or overnight at 4°C fohowed by blocking with PBS 0.1% casein. Purified scFvB7-24 and semi-purified monospecific B7-24 diabody

(scFvB7-24 L5 fraction p2 en p3) and B7-24 triabody (scFvB7-24 L0 fraction p2) molecules were added in a twofold serial dilution starting from 1 μg/ml to 15 ng/ml and incubated for lh at 37°C Subsequently 100 ml of B7.1ED fusion protein (500 ng/ml) or B7.2ED fusion protein (500ng/ml) (negative control) were added (lh at 37°C), fohowed by incubation with peroxidase-conjugated sheep anti-human Ig

(whole antibody) at 1/1000 for lh at 37°C. As substrate TMB was used. Plates were measured at 450 and 595 nm.

Results of the first ELISA setting (figure 43) show that the semi-purified fractions of B7-24 diabody (scFvB7-24 L5 p3) and of scFvB7-24 triabody (scFvB7-24 L0 p2) displayed binding to indirectly coated B7.1ED fusion protein, comparable with the binding of purified scFvB7-24. Similar results were obtained in the second ELISA setting. Here also, B7-24 diabody and B7-24 triabody display binding to B7.1ED fusion protein. The observed binding was specific as no significant binding was observed to the B7.2ED fusion protein. These semi-purified monospecific B7-24 diabody (scFvB7-24 L5 fraction p2 and p3) and triabody (scFvB7-24 L0 fraction p2) molecules are further analysed for their binding capacity to B7-1 molecules on the membrane of B7.1 expressing EBV transformed human B cells, RPMI8866. Cehs (5x10^J cells/staining) were preincubated for 20' at 4^CC in 50 ml FACS buffer (PBS supplemented with 5% inactivated FCS and 0.02% sodium azide) supplemented with 10% normal rabbit serum. Subsequently, the cehs were incubated with fourfold serial dilutions of purified scFvB7-24 semi-purified B7-24 diαbody (scFvB7-24 L5 fraction p2 and p3) or semi-purified B7 24 triabody (scFvB7-24 LO fraction p2), starting from 1 mg/ml for 20' at 4°C After three washes with FACS medium the cehs were incubated with mouse anti-his monoclonal antibody at 2 mg/stammg for 20' at 4°C followed by 3 washes Subsequently cells were incubated with biotinylated goat anti-mouse

(IgG+IgM) antibodies at 1 5 mg/stammg and lmg/stammg of R-phycoerythrrn (PE)- conjugated streptavidin for 20' at 4°C Cehs were analysed for fluorescent staining usmg a FACScan instrument Between each incubation step cehs were washed three times with FACS buffer Cehs only incubated with mouse anti-his antibody fohowed by incubation with biotinylated goat anti-mouse antibody and R-PE conjugated streptavidin were included as a negative control for staining FACS results usmg Img/ml of punned scFvB7-24 semi-purified B7 24 diabody (scFvB7-24 L5 fraction p2 and p3) and semi purified B7-24 triabody (scFvB7-24 LO fraction p2) are tisted out in figure 44 a-b The tinm tine represents the negative control staining whereas bold tines represent staining with the different recombinant antibodies

Results show that the semi-puruied B7-24 diabodies and B7-24 tnabodies bmd to the human B7 1 expressmg RPMI8866 cehs, in a comparable way as scFvB7-24

The semi purified constructs are further analyzed for their capacity to block T ceh-APC mteractions T cells were purified out of whole hepar nized blood on Ficoh Paque (density 1 077 Pharmacia Biotech) density gradients The peripheral blood mononuclear cehs (PBMC) present m the t erf ace were washed three times m 40 ml of RPMIbic supplemented with 10% inactivated FCS Subsequently monocytes were removed by cold aggregation (Mentzer et al 1986) To this end PBMC '-^rere resuspended in 40 ml oi RPLIIbic supplemented with 10% inactivated FCS ana slowly rotated lor 30 at 4 C Monocyte aggregates were allowed to sediment over a 15 period incubation on ice ana the non aggregated cells containing enncned T cehs and B cehs were carefully aspirated and centrifuged for lO at 1200 rpm T cells v/ere iurther enriched using Lympho-Kwik-T (One lambda Inc Los Angeles C ) This reagents contains a mixture of ami monocyte and anti B ceh mAbs and complement Lymphocytes were resuspended in 3 ml

Lympho Kwik T and tne mixture was incubated for 45' at 37~C Subsequently cehs were resuspended in 0 5 ml of PBS and centrifuged for 2' at 2000 rpm and washed twice

The mouse fibroblast cell tine, 3T6, tranfected with the cDNA encoding human B7_ 1 and human CD32 (3T6-CD32-B7 1) was mitomycin C treated Cell pellet of 1 subconfluent falcon was dissolved in 800 ml RPMIbic and 200 ml Mitomycm C (250 mg/ml) and incubated for 40' at 37°C followed by two washes Subsequently, cells are suspended m 30 ml RPMIbic and incubated for 15' at 37^DC fohowed by one additional wash step In the mixed lymphocyte culture (MLR), the mitomycm C treated 3T6-CD32-B7 1 cehs (10⁴ cehs/weh) were incubated with OKT3 (1 mg/ml) for 1 h at 37°C fohowed by a 1 h incubation at 37°C with threefold serial dilutions of B7-

24 mAb recombinant purified scFvB7-24 or semi-purified B7-24 diabody (scFvB7-24 L5 fractions p2 and p3) or B7-24 triabody (scFvB7-24 L0 fraction p2) starting from 1 μg/ml Subsequently purified T cehs (5x10⁴ cell/well) were added and incubated for 5 days After 5 days cehs were mcubated with 1 mCi ( H]-thymιdιne for 6 h and harvested usmg an automated ceh harvester ( ³H)-thymιdιne incorporation was determmed with a liquid scintillation counter.

Results showed that the recombinant scFvB7-24 antibodies, the B7-24 diabody (tigure 45) and B7-24 triabody (figure 46) molecules displayed neutralizing activities, comparable to those of the parent B7-24 mAb, as measured in MLR Gelfiltration fractions 5 to 17 of the B7-24 diabody (scFvB7-24 L5 p3) and B7-

24 triabody (scFv37-24 L0 p2) chromatography and fractions 10 to 17 of the scFvB7 24 chromatography were evaluated for specific bmdmg in ELISA To this end, mouse anti-his antibody ( 1 mg/ml) was coated in PBS for 2h at 37^"C fohowed by blocking with PBS 0 1% casern for lh at 37^"C The different gelfiltration fractions of scFvB7-24 B7 24 diabodies (scFvB7-24 L5 p3) and B7 24 tnabodies (scFvB7-24 L0 p2) were added twofold serial dilutions and indicated for lh at 37 C fohowed by incubation with 100 ml of B7 1ED fusion protem (500 ng/ml) for lh at 37°C Subsequently the wehs were mcubated with peroxidase-conjugated sheep anti- human Ig (whole antibody) at dtiuuon 1/1000 for lh at 37^')C Color was generated by incubation for 30 at RT with TMB Plates were read at 450 and 595 nm OD values tisted out m the figures are corrected for the blank OD value (OD value obtained if no recombinant antibody was included m the ELISA setting) These results confirm that B7-24 triabody (ScFvB7-24 LO) (figure 47, lower panel) displays reactivity the triabody region as weh as in the diabody region and the monomeπc region, while B7-24 diabody (scFvB7-24 L5) (tigure 47 middle panel) has reactivity m the diabody region and monomeπc region In contrast, scFvB7-24 (tigure 12 upper panel) has only reactivity m the monomeπc region This confirms that scFvB7-24 LO mdeed has the capacity of forming trimers (tnabodies) and d ners (diabodies), while ScFvB7-24 L5 has only the capacity of forming dimers (diabodies) ScFvB7-24 LI 5 remains monomenc

8 Evaluation of the B7 blockmg agents

8 J Chemically cross-linked anτι-B7 antir>odιes suppress the proliferation of human T cells and induce IL- 12 production m peripheral blood mononuclear cells (PBMC)

As a proof of concept we wanted to test the lmmunosuppressive effect of chemical cross-linked anti-B7 1 mAb and anti-B7 2 mAb (further refeπed as anti- B7 1/B7 2 mAb) We followed the protocol described 'Current protocols of Immunology for 'Chemically cross linking antibodies with SPDP'(Current Protocols of Immunology, Supplement 14 Unit 2 13) Briefly, anti-B7 1 mAb (B7-24) and anti- B7 2 mAb ( IGIO) are cross-linked usmg the heterobrfunctional compound N- succιnιmιdvl-3-(2-pyrιdyldιthιol) propionate (SPDP) This reagent binds randomly to e-amino groups on lysme residues and forms reducible disultide bonds between antibodies The use of a bifunctional cross tinker is very important because the mtention here is to create a oifuncuonai antibody (recognizing two different antigens) and therefore only a cross linking between antι-B7 1 mAb (B7-24) and anti-B7 2 mAb ( IGIO) wih be performed The formation of IGIO IGI O or B7-24-B7-24 is not desirable The resulting bispecinc molecules consists of aggregates of antibodies of varying size linked together at random sites The bifunctional antibody v/as obtained by performing a size exclusion chromatography (Superdex G 200) (SEC), which separates the dimeπc antibody (300 kDa) from the monomenc antibodies (150 kDa) The gelfiltration resulted in 4 mam peaks (figure 48) Peak 2 and 3 appeared at the suspected place for dimenc (B7-24/1G10 crosstinked mAbs) and monomenc mAbs (B7-24 mAb) Peak 1 eluted at the place for ohgomers As we 5 have already seen that the purified 1G10 is retained much longer onto gelfiltration column we can suppose the peak 3 and 4 are dimeπc and monomenc 1G10 Gefftitrationfractions 19, 20 and 21 (peak 2), containing dimenc B7-24/1G10 crosstinked mAbs were tested and compared with a mixture of antι-B7 1 mAb (B7- 24)- and anti-B7 2 mAb ( IGl 0) for their bmdmg capacity to the human B7 1 and B7.2 o molecules Mouse 3T6 cells tranfected with the cDNA encoding human CD32 and the human B7 1 molecule (3T6-CD32-B7 1) or encoding the human B7 2 molecule (3T6-CD32B7 2) (De Boer et al , 1992) were mcubated with fractions 19 (3 μg/cell pehet) and 20 (2μg/ceh pehet) of the cross-linked anti-B7 1/B7 2 mAb or as control with a mixture of anti-B7 1 mAb (B7-24) (0 5 μg/ceh pehet) or anti-B7 2 mAb 5 ( IGl 0)(0 5 μg/ceh pehet) Cehs (0 5-lx 10⁵ cells/sample) were mcubated for 15' at

4³C with the different mAbs After washing twice m RPMI1640 supplemented with 10% FCS, the cehs were mcubated for another 15' at 4°C with goat anti-mouse antibodies conjugated to fluorescein isothiocyanate (FTTC) The cells were washed twice m RPMI 1640 supplemented with 10% FCS and finally suspended m PBS 0 supplemented with 1% BSA and 0 1% NaN₃ and analyzed with a FACScan flow cytometer (Becton Dickinson) The specuic bmdmg ot the monoclonal antibodies is expressed as the mean fluorescence intensity in arbitrary units The results (tigure 49 a-b) showed that the chemical cross-linked antι-B7 1/B7 2 mAb binds both the human B7 1 molecule (3T6-CD32-B7 1 cells) and the human B7 2 molecule (3T6- 5 CD32-B7 2 cehs) indicating that the chemical cross-linked mAb exists of a B7 1 and a B7 2 recognizing part

Gelftitrationfractions 19 20 and 21 (peak 2) containing dimenc B7-240G1 O crosshnked mAbs were tested and compared with a mixture oi anti-B⁷ 1 mAb (B7 24)- and anti-B7 2 mAb ( 1G10) for their bmdmg capacity to the human B7 1 and B7 2 0 molecules in a coating ELISA with hB7 1ED and hB7 2ED fusion protems 500 ng of hB7 1ED and hB7 2ED fusion protems were coated in PBS for 2n at 37 C fohowed by blocking m PBS 0 1% cαse for lh at 37°C Coating without B7ED fusion protems was mcluded as a negative control Subsequently two different dilutions (1/1000 and 1/10000) of above mentioned fractions were incubated for 1 h at 37^"C In this way, the amount of material added to the ELISA correlates approx with 250 ng/ml and 25 ng/ml for the dimer fraction 20 Detection was done by incubation with goat anti-mouse IgG (Fcg) HRP labeled antibody for 1 h at 37°C fohowed by the addition of TMB as substrate Plates are read at 450-595 nm on a microtiter plate reader Results shoxed that the B7-24/1G10 crosstinked monoclonal antibodies are able to bmd both hB7 1 and hB7 2 No signal was detected on the bianco coating conditions

The simultaneous bmdmg of B7 1 and B7 2 to the cross-linked antι-B7 1/B7 2 mAb was also evaluated usmg SPR-analysis with the BIAcore instrument (Pharmacia Biosensor AB Uppsala Sweden) In this experiment hB7 2ED fusion protem was immobilized directly onto a CM5 sensor chip usmg an amine coupting according to the manuiacturer s procedure Briefly, the sensor chip surface was initially activated with N-hydroxysuccinimide and N-ethyl-N'-(3- ethylamιnopropyl)carbodιιmιde A continuous flow of 5 μg/ml soluble hB7 2ED fusion protem was mjected over the activated surface at pH 4 8 Residual unreacted ester groups were blocked with ethcmolamine This coupting procedure resulted in an immobilization level of about 3450 RU (resonance units) Onto this coupled B7 2Fc the cross tint ed B7 24/1G10 construct (pool 20) was mjected and a specific interaction of about 1500 RU could be monitored (Figure 50) Subsequent injection of hB7 1ED fusic n protein resulted m an increase of about 460 RU These results clearly indicate that B7 24/1G10 preparation contains cross linked constructs that recognize both B7 1 and B7 2 As the bmdmg of the second B7 molecule (B7 1ED fusion protem) is iairly hign we can concluae that most 1G10 containing molecules are B7-24/1G10 cross links

The m virro proliferation of CDS - activated human T lymphocytes in the presence of B7 1 and B7 2 molecules expressed on the membrane of different antigen presenting cells (APC) is largely blocked bv the combination of anti B7 1 mAb and antι-B7 2 mAb In our experiments human peripheral blood mononuclear cells (PBMC) were isolated from buffy coat by density centrifugation and cultured with anti-CD3 mAb (OKT3 0 3 μg/ml) in the presence of a mixture of antι-B7 1 mAb (B7- 24 0 5 μg/ml) and anti-B7 2 mAb (1G10 0 5μg/ml) or m the presence of the chemical crosstinked anti-B7 1/B7 2 mAbs ( 1 μg/ml) Our results show that chemical crosstinked anti-B7 1/B7 2 mAbs are able to stronger inhibit this proliferation of activated human T lymphocytes compared to the combination of the two mAbs Thus the chemical crosshnked anti-B7 1/B7 2 mAbs have a stronger immunosuppression potency than the combination an anti-B7 1 mAb and an anti- B7 2 mAb Activation of human PBMC isolated from buffy coat by density centrifugation, by anti-CD3 mAb (OKT3 0 3 μg/ml) induces the production of IL-12 by the monocytes present in these cultures This IL 12 production by the human PBMC cultures is suppressed by tne combination of anti-B7 1 mAb (B7 24) and anti-B7 2 mAbs ( 1G10) Chemical crosstinked anti-B7 1/B7 2 mAbs are able to stronger inhibit this IL- 12 production compared to the combination of the two mAbs As IL- 12 activates

T lymphocytes and induces IFN-γ release by Th-1 lymphocytes this stronger inhibition by the crosstinked anti-B7 1/B7 2 mAbs, compared to the inhibition by the combination of the two mAbs alone, means indirectly a stronger inhibition of the T lymphocyte activation Thus, again these experiments demonstrate a stronger immunosuppression potency of the chemical crosstinked anti-B7 1/B7 2 mAbs compared to the combination of an anti-B7 1 mAb and an antι-B7 2 mAb In a second proliferation experiments, the crosstinked monoclonal antibodies were further evaluated to block T cell-APC mteractions in a MLR wherein T cells are cultured vith RPMI 8877 cells a EBV tranformed B ceh hne expressmg human B7 1 and human B7 2 molecules T cehs were purified out of whole hepaπnized blood on

Ficoll Paque (aensity 1 077 Pharmacia Biotecn) density gradients Tne peripheral blood mononuclear cehs (PBMC) present m the mterface v/ere washed three times m 40 ml oi RPMIbic supplemented with 10% inactivated FCS Subseσuentiv monocvtes were removed by cold aggregation (Mentzer et al 1986) To this end PBMC were resuspended m 40 ml of RPMIbic supplemented with 10% inactivated

FCS and slowly rotated for 30 at 4"C Monocyte aggregates were allowed to sediment over α 15' period incubation on ice, and the non-aggregated cells containing enriched T cehs and B cehs were carefully aspirated and centrifuged for 10' at 1200 rpm. T cells were further enriched using Lympho-Kwik-T (One lambda Inc, Los Angeles, CA). This reagents contains a mixture of anti-monocyte and anti-B ceh mAbs and complement. Lymphocytes were resuspended in 3 ml

Lympho-Kwik-T and the mixture was incubated for 45' at 37°C. Subsequently, cells were resuspended in 0.5 ml of PBS and centrifuged for 2 at 2000 rpm and washed twice.

The EBV tranformed B cell tine, RPMI8877, was mitomycin C treated. Ceh pellet of 1 subconfluent falcon was dissolved in 800 ml RPMIbic and 200 ml Mitomycin C (250 mg/ml) and incubated for 40' at 37°C fohowed by two washes. Subsequently, cells are suspended in 30 ml RPMIbic and incubated for 15' at 37^SC followed by one additional wash step. In the mixed lymphocyte culture (MLR), the mitomycin C treated RPMI8877 cells ( 10⁴ cehs/weh) were incubated with OKT3 ( 1 mg/ml) for 1 h at 37°C followed by a 1 h incubation at 37°C with serial dilutions of crosshnked

B7.24/1G10 mAbs (fraction 19 or 20) or with the combination of B7.24 ( 1 μg/ml) and 1G10 (lμg/ml). Subsequently, purified T cehs (5x10⁴ cell/well) were added and incubated for 5 days. After 5 days, cehs were incubated with 1 mCi (³H0thymidine for 6 h and harvested using an automated ceh harvester. ( ³H)-thymidine incorporation was determined with a liquid scintillation counter. Results showed

(figure) that the crosshnked b7.24/lG10 monoclonal antibodies displayed neutralizing activities, comparable to those of the parent B7-24 mAb, as measured in MLR. Our results show that chemical crosstinked anti-B7.1/B7.2 mAbs are able to stronger or comparable inhibit this proliferation of activated human T lymphocytes, compared to the combination of the two mAbs. Thus, the chemical crosstinked anti-B7.1/B7.2 mAbs have a stronger or comparable immunosuppression potency than the combination an anti-B7.1 mAb and an σnti- B7.2 mAb.

8.2 Induction of donor specific tolerance in a rhesus monkey skin transplant model The immunosuppressive effect of the combined treatment with Cyclosporin A (CsA) and a monoclonal antibody directed against the B7.1 molecule (B7-24 mAb) is excπnined in a rhesus monkey skin transplant model. Animals receive the experimental medication B7.24 mAb(0.5 mg/kg) and CsA (5 mg/kg) for 10 days starting one day before the day of transplantation. Transplant rejection is monitored by scoring the skin grafts. At several time points during the expertinent, blood samples are taken to determine the blood level of CsA, serum levels of the mAb anti-CD80, the rhesus monkey anti-mouse anti-body (RAMA) response, anti- donor antigen antibody reponse. Combined treatment of the anti-CD80 mAb and CsA in this rhesus monkey skin transplantation model results in increased skin graft survival time (control 10 days, treatment 14 days) but fails to induce donor specific tolerance. Thus, short term immunosuppression, resulting in prolonged skin graft survival can be obtained using CsA and B7-24 mAb as prophylactic treatment in rhesus monkey skin transplant model. Data has been accumulating that blocking both CD80 and CD86 molecules using or CTLA-4-Ig or a combination of anti-CD80 mAb and anti-CD86 mAb can induce often indefinite prolongation of allograft survival in rat and mouse cardiac transplantation models (Bashuda et al., 1996; Lenschow et al., 1992, 1995; lin et al, 1993; Pearson et al., 1994). Therefore , it can be suggested that the combination of anti-B7.1 mAb and CsA in this rhesus monkey skin transplantation model only results in a prolongation of the graft survival time, but fails to induce donor-specific non-responsiveness, since the B7.2- CD28 pathway is not blocked.

8.3 Induction of donor specific tolerance in a rhesus monkey kidney transplant model

A combination of anti-B7.1 mAb, anti-B7.2 mAb and CsA seems to be an optimal treatment to induce donor-specific tolerance during allo-transplantσtion. Rhesus monkeys receive a kidney transplantation from an allogeneic donor mismatched for at least one MHC class I and one class II antigen. Graft survival time is indicative for the immunosuppressive potency of the therapy combining anti-B7.1 mAb, αnti-B7 2 mAb and CsA The mAbs (0 5 mg/kg at day - 1 0 25 mg/kg on day 0 till 12) are given daily for 14 days starting at day -1 The CsA ( 10 mg/kg) is given daily for 35 days starting at day 1 At several time pomts durmg the experiment blood samples are taken to determine the blood level of CsA serum levels of the mAbs anti-B7 1 and anti-B7 2, the rhesus monkey anti-mouse anti-body (RAMA) response anti-donor antibody reponse

In a pharmacokinetic -toxicology (PK-Tox) study, with the same design as the transplantation study, no signs of acute cytotoxicity of the anti-B7 1 mAb and anti- B7 2 mAb m combination with CsA are observed In the skm transplantation study (combination of anti-B7 1 mAb and CsA), the anti-B7 1 concentrations decreases to undetectable values after 9 days due to a RAMA response Usmg the combination of anti-B7 1 mAb anti-B7 2 mAb and CsA, total mouse mAb serum concentrations remained high and are even detectable on day 20 No RAMA response can be detected in the monkeys in this PK-Tox study Thus combined treatment with anti- B7 1 mAb anti-B7 2 mAb and CsA is able to prevent an lmmunological response against mouse mAbs, indicating the high immunosuppressive potency of this combination

Rhesus monkeys, not receivmg any treatment after a kidney graft transplantation, reject the graft 6 days after transplantation Combined treatment of the animals with antι-B7 1 mAb and antι-B7 2 mAb strongly prolonges the kidney graft survival time (22 days and 35 days) Using this combmed treatment with antι-B7 1 mAb and antι-B7 2mAb the mouse mAb serum concentrations remains high and both anti- B7 2 mAb and anti-B7 1 mAb are still detectable on day 15 As long as the mAbs are m circulation no or only a weak RAMA response can be measured As the concentration of circulating mAbs become low the F AMA responses become high and the immunosuppressive potency of this comornation is lost Longer treatment with the combination oi antι-CD80 mAo and antι-CD86 mAb can enhance the immunosuppressive capacity of this combination Also combined treatment of the animals after kidney transplantation with anti-B7 1 mAb and the antι-B7 2 mAb together with CsA can enhance the immunosuppressive capacity REFERENCES

Allison, J.P., and M.F. Krummel. 1995. The yting and yang of T ceh costimulation. Science 270(5238):932.

Armitage, R.J. 1994. Tumor necrosis factor receptor superfamily members and thefr ligands. [Review]. Current Opinion in Immunology 6:407.

Azuma, M., D. Ito, H. Yagita, K. Okumura, J.H. Phillips, LL. Lanier, and C. Somoza. 1993b. B70 antigen is a second ligand for CTLA-4 and CD28. Nature 366:76.

Azuma, M., H. Yssel, J.H. Phillips, H. Spits, and L.L. Lanier. 1993a. Functional expression of B7/BB1 on activated T lymphocytes. Journal of Experimental Medicine 177:845.

Baltint, R., and J. Larrick. 1993. Antibody engineering by parsimonious mutagenesis. Gene 137:109.

Bashuda, H., K. Setino, and M. Kano. 1996. Specific acceptance of cardiac allografts after treatment with antibodies to CD80 and CD86 in mice. Transplant

Proc. 28: 1039.

Bierer, B.E., G. Hollander, D. Fruman, and S.J. Burakoff. 1993. Cyclosporin A and FK506: molecular mechanisms of immunosuppression and probes for transplantation biology. Curr. Opin. Immunol. 5:763.

Blazar, B.R., P.A. Taylor, P.S. Linsley, and D.A. Vallera. 1994. In vivo blockade of CD28/CTLA4 : B7/BB1 interaction with CTLA4-Ig reduces lethal murine graft-versus- host disease across the major histocompatibility complex barrier in mice. Blood 83:3815. Boussiotis, V.A., G.J. Freeman, G. Gray, J. Gribben, and L.M. Nadler. 1993. B7 but not ICAM- 1 costimulation prevents the induction of human ahoantigen specific tolerance. Journal of Experimental Medicine 178: 1753.

Boussiotis, V.A., G.J. Freeman, J.G. Gribben, J. Daley, G. Gray, and L.M. Nadler.

1993a. Activated human B lymphocytes express three CTLA-4 counterreceptors that costimulate T ceh activation. Proc. Natl. Acad. Sci. USA 90(23): 1 1059.

Boussiotis, V.A., G.J. Freeman, J.D. Griffin, G.S. Gray, J.G. Gribben, and L.M. Nadler. 1994. CD2 is tinvolved in maintenance and reversal of human ahoantigen- specific clonal anergy. J. Exp. Med. 180:1665.

Caux, C, C. Massacrier, B. Vanbervliet, B. Dubois, C. Van Kooten, I. Durand, and J. Banchereau. 1994. Activation of human dendritic cehs through CD40 cross-linking. Journal of Experimental Medicine 180: 1263.

Coloma, M., and S. Morrison. 1997. Design and production of novel tetravalent bispecific antibodies. Nature Biotech. 15: 159.

Damle, N.K., P.S. Linsley, and J.A. Ledbetter. 1991. Direct helper T ceh induced B ceh differentiation involves interaction between T cell antigen CD28 and B cell activation antigen B7. Eur. J. Immunol. 21: 1277.

de Boer, M., P. Parren, J. Dove, F. Ossendorp, G. van der Horst, and J. Reeder. 1992.

Functional characterization of a novel antι-B7 monoclonal antibody. Eur. J. Immunol. 22(12):3071

de Boer, M., P. Parren, J Dove, P. Ossendorp, G. van der Horst, and J. Reeder. 1992. Functional characterization of a novel anti-B7 monoclonal antibody. Eur. J.

Immunol. 22:3077. de Kruif J and T Logtenberg 1996 Leucme zipper dimeπsed bivalent and bispecific scFv antibodies from a semi synthetic antibody phage display library /

5 Desmet J M De Maeyer B Hazes and I Lasters 1992 The dead-end elimination theorem and its use in protem side-chain positioning Nature 356 539

Desmet J I Wilson, M Joniau M De Maeyer and I Lasters 1997 Computation of the bmdmg of fully flexible peptides to protems with flexible side chains FASEB J 0 1 1 164

Ellison J W B J Berson and L E Hood 1982 The nucleotide sequence of a human immunoglobulin C gamma 1 gene Nucleic Acids Res 10(13) 4071

5 Fanger M W , P M Morganelh and P M Guyre 1992 Bispecific antibodies Cut

Rev Immunol 12(3-4) 101

Fmck, B K P S Linsley and D Wofsy 1994 Treatment of munne lupus with CTLA4Ig Science 265(5176) 1225

20

Freedman A S G J Freeman K Rhynhart and L M Nadler 1991 Selective induction of B7/BB 1 on mterferon-gamma stimulated monocytes a potential mechanism for amplification of T cell activation through the CD28 pathway Cellular Imm unologγ 137 429 o ς

Freeman G J A S Freedman J M Segti G Lee J F Whitman and L M Nadler 1989 B7 a new member ot the Ig supeπamtiy with unique expression on activated and neopiastic B cells Journal oi Immunology 143 27 P

30 Freeman G J J G Gribben V A Boussiotis J W Ng V A Restivo Jr L A Lombard

G S Gray and L M Nadler 1993 Clonmg of B7 2 a CTLA-4 counter receptor that costimulαtes human T cell proliferation Science 262 909

Georgiou, G , C Stathopoulos P Dougherty A Nayak, B Iverson and R Curtiss III 1997 Display of heterologous protems on the surface of microorganisms from the screening of combinatorial libraries to live recombinant vaccmes Nature

Bioiecn 15 29

Ghetie, M A , and E S Vitetta 1994 Recent developments in mnmunotoxm therapy Curr Opm Immunol 6(5) 707

Gimmi, C D , G J Freeman, J G Gribben K Sugita A S Freedman, C Morrmoto, and L M Nadler 1991 B-ceh suriace antigen B7 provides a costimulatory signal that mduces T cells to proliferate and secrete mterleukm 2 Proc Natl Acad Sci USA 88 6575

Gribben J G , G J Freeman, V A Boussiotis, P Rennert, C L Jeltis, E Greenfeld, M Barber, V A Restivo Jr , X Ke, and G S Gray 1995 CTLA4 mediates antigen- spectiic apoptosis of human T cehs Proc Natl Acad Sci USA 92(3) 81 1

Gribben J G E C Guinan V A Boussiotis, X Y Ke L Linsley C Sieff G S Gray

G J Freeman and L M Nadler 1996 Complete Plockade of B7 family-mediated costimulation is necessary to mduce human alloantigen-specific anergy a method to ameliorate graft-versus-host disease and extend the donor pool Blood 87 4887

Guerder S D E Picareha P S Linsley, and R A Flavell 1994 Costimulator B7- 1 confers antigen-presenting-cell tunction to parencnymal tissue ana m conjunction with tumor necrosis factor alpha leads to autoimmunity in trans gemc mice Proc Natl Acaa Sci USA 91( 1 1) 5138

Harding F A J G McArthur J A Gross D H Raulet and J P Allison 1992 CD28- mediated signally costimualtes murine T cehs and prevents mduction of anergy m T-ceh clones. Nature 356:607.

Hαrlαn, D.M., H. Hengαrtner, M.L. Huang, Y.H. Kang, R. Abe, R.W. Moreadith, H. Pircher, G.S. Gray, P.S. Ohashi, and G.J. Freeman. 1994. Mice expressing both B7- 1 and viral glycoprotein on pancreatic beta cells along with glycoprotein-specific transgenic T cehs develop diabetes due to a breakdown of T-lymphocyte unresponsiveness. Proc. Natl. Acad. Sci. USA 91 (8):3137.

Harper, K., C. Bolzano, E. Rouvier, M.G. Mattei, M.F. Luciani, and P. Golstein. 1991. CTLA-4 and CD28 activated lymphocyte molecules are closely related in both mouse and human as to sequence, message expression, gene structure, and chromosomal location. /. Immunol. 147: 1037.

Hayden, M. S., L.K. Gihhand, and J.A. Ledbetter. Antibody engineering. Curr. Opin. Immunol. 9:202-212.

Holtiger, P., T. Prospero, and G. Winter. 1993. "Diabodies": small bivalent and bispecific antibody fragments. Proc. Natl. Acad. Sci. USA 90(14):6444.

Hoogenboom, H. 1997. Designing and optimising library selection strategies for generating high-affinity antibodies. Trends Biotech. 15:62.

Hoogenboom, H., A. Griffiths, K. Johnson, D. Chisweh, P. Hudson, and G. Winter. 1991. Ivlulti-subunit proteins on the surface of filamentous phage: methodologies for displaying antibody (Fab) heavy and light chatins. Nucl. Acids Res. 15:4133.

Hoogenboom, H.R. 1997. Designing and optimizing library selection strategies for generating high-affinity antibodies. Trends Bioiechnol. 15(2):62.

Iliades, P., Kortt A.. A., and P. Hudson 1997. Triabodies: single chain Fv fragments without linker form trivalent trimers. FEES Letters 409:437-441. Imboden, J.B., and J.D. Stobo. 1985. Transmembrane signalling by the T ceh antigen receptor. Perturbation of the T3-antigen receptor generates inositol phosphates and releases calcium ion from intracehular stores. /. Exp. Med. 161 :446

Ito, W., and Y. Kurosawa. 1993. Development of an artificial antibody system with multiple valency using an Fv fragment fused to a fragment of protein A. /. Biol. Chem. 268:20668.

Ito, W., and Y. Kurosawa. 1993. Development of an artificial antibody system with multiple valency using an Fv fragment fused to a fragment of protein A. /. Biol.

Chem. 268(27):20668.

Jenkins, M.K., and J.G. Johnson. 1993 . Molecules tinvolved in T cell costimulation. Curr. Opin. Immunol. 5:361.

June, C.H., J.A. Ledbetter, M.M. Gihespie, T. lindsten, and C.B. Thompson. 1987. T ceh proliferation involving the CD28 pathway is associated with cyclosporine- resistant interleukin 2 gene expression. Mol. Cell. Biol. 7:4472.

June, C.H., J.A. Ledbetter, T. Lindsten, and C.B. Thompson. 1989. Evidence for the involvement of three distinct signals in the induction of IL-2 gene expression in human T lymphocytes. J. Immunol. 143: 153.

June, C.H., J.A. Ledbetter, P.S. Unsley, and C.B. Thompson. 1990. Role of the CD28 receptor in T-ceh activation. Immunology Today 1 1:21 1.

Kipriyanov, S.M., M. Little, H. Kropshofer, F. Breitling, S. Gotter, and S. Dubel. 1996. Affinity enhancement of a recombinant antibody: formation of complexes vith multiple valency by a single-chain Fv fragment-core streptavidin fusion. Protein Ξng. 9(2):203. Kipriyαnov, S., M. Little, H. Kropshofer, F. Breithng, S. Gotter, and S. Dubel. 1996. Affinity enhancement of a recombinant antibody: formation of complexes with multiple valency by a single-chain Fv fragment-core streptavidin fusion. Prot. Eng: 9:203.

Knappik, A., and A. Pluckthun. 1995. Engineered turns of a recombinant antibody improve its in vivo folding. Protein Eng. 8( 1 ):81.

Kortt, A., M. Lah, G. Oddie, C. Gruen, J. Burns, L. Pearce, J. Atwell, A. McCoy, G. Howlett, D. Metzger, R. Webster, and P. Hudson. 1997. Single-chain Fv fragments of amti-neuraminidase antibody NC10 containing five- and ten -residue linkers form dtiners and with zero-residue linker a trimer. Prot. Eng. 10:423.

Kostelny, S., M. Cole, and Y. Yun Tso. 1992. Formation of a bispecific antibody by the use of leucine zippers. /. Immunol. 148:1547.

Kranz, D.M., T.A. Patrick, K.E. Brigle, M.J. Spineha, and E.J. Roy. 1995. Conjugates of folate and anti-T-ceh-receptor antibodies specifically target folate-receptor-positive tumor cells for lysis. Proc. Natl. Acad. Sci. USA 92(20):9057.

Kranz D., M. Gruber, and E. Wilson. 1995. Properties of bispecific single chain antibodies expressed in E.coli. J. Hematother. 4:403.

Kreane-Myers, A., W.C. Gause, P.S. Linsley, S.J. Chen and M. Wills-Karp. 1997. B7- CD28/CTLA-4 costimulatory pathways are requfred for the development of T helper ceh 2-mediated allergic airway responses to inhaled antigens. J. Immunol. 158: 2042

Krinzman, S.J., G.T. De Sanctis, M. Cernadas, D. Mark, Y. Wang, J. Listman, L. Kobzik, C. Donovan, K. Nassr, I. Katona, D.C. Christiani, D.L. Perkins and P.W. Finn.

1996. Inhibition of T cell costimulation abrogates airway hyperresponsiveness in a murine model. J. Cltin. Invest. 98:2693.

Lαrsen, C.P., E.T. EHwood, D.Z. Alexander, S.C. Ritchie, R. Hendrix, C. Tuckerburden, H.R. Cho, A. Aruffo, D. Hollenbaugh, and P.S. Linsley. 1996. Long-term acceptance of skin and cardiac allografts after blocking CD40 and CD28 pathways. Nature 381:434.

Lasters, I., and J. Desmet. 1993. The fuzzy-end elimination theorem: correctly implementing the side chain placement algorithm based on the dead-end elimination theorem. Prot. Eng. 6:717.

Lasters, I., M. De Maeyer, and J. Desmet. 1995. Enhanced dead-end elimination in the search for the global minimum energy conformation of a cohection of protein side chains. Pror. Eng. 8:815.

Lover, W.G., G.M. Air, R.G. Webster, and S.J. Smith-Gill. 1990. Epitopes on protein antigens: misconceptions and realities. Cell 61(4):553.

Lenschow, D.J., Y. Zeng, and R.J. Thislethwcrite. 1992. Long-term survival of xenogeneic pancreatic islets grafts induced by CTLA-4 Ig. Science 257:789.

Lenschow, D.J., Y. Zeng, K.S. Hathcock, L.A. Zuckerman, G. Freeman, J.R. Thislethwaite, G.S. Gray, R.J. Hodes, and J.A. Bluestone. 1995. Inhibition of transplant rejection following treatment with anti-B7-2 and anti-B7-l antibodies. Transplantation 60( 10): 1 171.

Lin, H., S.F. Bollin, and P.S. Linsley. 1993. Long-term acceptance of major histocompatibility complex mismatched cardiac allografts induced by CTLA-4Ig plus donor-specific trc fusion. /. Exp. Med. 178: 1801.

Linsley, P.S., JΛ.L. Greene, T.P. Bradshaw, J.A. Ledbetter, C. Anasetti, and Ν.K.

Damle. 1992. Coexpression and functional cooperation of CTLA-4 and CD28 on activated T lymphocytes. /. Exp. Med. 176: 1595.

Linsley, P.S., J. Ledbetter, R. Peack, and J. Bajorath. 1995. CD28/CTLA-4 receptor structure, binding stoichiometry and agopregation during T cell activation. Res. Immunol. 146(3): 130.

Linsley, P.S., E.A. Clark, and J.A. Ledbetter. 1990. T-ceh antigen CD28 mediates adhesion with B cehs by interacting with activation antigen B7 BB-1. Proc. Natl. Acad. Sci. USA 87:5031.

Linsley, P.S., W. Brady, L. Grosmaire, A. Aruffo, N.K. Damle, and J.A. Ledbetter. 1991. Btinding of the B ceti antigen B7 to CD28 costimulates T ceh proliferation and interleukin 2 mRNA accumulation. Journal of Experimental Medicine 173:721.

Lybrand, T.P. 1995. Ligand protein docking and rational drug design. Curr. Opin.

Struct. Biol. 5(2):224.

Martin, S.D., and T.A. Springer. 1987. Purified intercellular adhesion molecule- 1 (ICAM-1) is a ligand for lymphocyte function-associated antigen (LFA-1). Cell 51(5):813.

McGuinness, B., G. Walter, K. FitzGerald, P. Schuler, W. Mahoney, A,. Duncan, and H. Hoogenboom. 1996. Phage diabody repertoires for selection of large numbers of bispecific antibody fragments Nature Biotech. 14: 1 149.

iviendez, M.J., L.L. Green, J.R.F. Corvalan, X-C Jia, C.E. Maynard-Currie, X-D. Yang, M.L. Gaho, D.M. Louie, D.V. Lee, K.L. Erickson, J. Luna, C.M.N. Roy, H. Abderrahim, F. Kirschenbaum, M. Noguchi, D.H. Smith, A. Fukushima, J.F. Hales, M.H. Finer, C.G. Davis, K.M. Zsebo, and A. Jakobovits. 1997. Functional transplant of megabase human immunoglobulin loci recapitulates human antibody response in mice. Naiure genetics 15:146. Mueller, D.L., M.K. Jenkins, and R.H. Schwartz. 1989. Clonal expansion versus functional clonal inactivation: a costimulatory signalling pathway determines the outcome of T cell antigen receptor occupancy. Annu. Rev. Immunol. 7:445.

Mylvaganam, S.E., Y. Paterson, K. Kctiser, K. Bowdish, and E.D. Getzoff. 1991.

Biochemical implications from the variable gene sequences of an anti-cytochrome c antibody and crystahographic characterization of its antigen-binding fragment in free and antigen-complexed forms. /. Mol. Biol. 221(2):455.

Novotny, J., R.E. Bruccoleri, and F.A. Saul. 1989. On the attribution of btinding energy in antigen- antibody complexes McPC 603, D1.3, and HyHEL-5. Biochemistry 28(11):4735.

Novotny, J. 1991. Protein antigenecity: a thermodynamic approach. Mol. Immunol. 28(3):201.

Nuss, J.M., P.B. Whitaker, and G.M. Air. 1993. Identification of critical contact residues in the NC41 epitope of a subtype N9 influenza virus neuraminidase. Proteins 15(2): 121.

Pack, P. 1995. Tetravalent miniantibodies with high avidity assembling in Escherichia coli. /. Mol. Biol. 246(1):28.

Pack, P., and A. Plύckthun. 1992. Miniantibodies: use of amphipathic helices to produce functional, flexibly linked dimeric Fv fragments with high avidity in

Escherichia coli. Biochemistry 31 : 1579.

Pack, P., K. Muher, R. Zahn, and A. Pluckthun. 1995. Tetravalent miniantibodies with high avidity assembling in Escherichia coli. /. Mol. Biol. 246:28.

Pack, P., M. Kujau, V. Schroekh, U. Knϋpfer, R. Wenderoth, D. Riesenberg, and A. Pluckthun 1993 Improved bivalent miniantibodies with identical avidity as whole antibodies, produced by high ceh density fermentation of Escherichia coli Bio/Technology 11 1271

Parsons H , J Earnshaw J Wilton K Johnson P Schueler W Mahoney, and J

McCafferty 1996 Directing phage selections towards specific epitopes Prot Eng 9 1043

Pearson, T C , D Z Alexander, K J Wrnn, P S Lmsley, R P Lowry and C P Larsen 1994 Transplantation tolerance duced by CTLA4-Ig Transplantation 57 1701

Pescovitz M D R Yang 0 L u M H Collins and J L Grosfeld 1994 CD28- pathwσy blockade with CTLA-4Ig leads to prolongation ol small bowel transplant survival in rats Transplantation Proceeώngs 26 1618

Remsmoen, N L , R M Bol an, K Savik, K Butters A J Matas and M I Hertz 1994 Improved long-term graft outcome m lung transplant recipients who have donor antigen-specific hyporeactivity Journal of Heart & Lung Transplantation 13 30

Sarubbi E S D Yanofsky R W Barrett and M Denaro 1996 A cell-free nonisotopic high-throughput assay for inhibitors of type- 1 mterleukm 1 receptor Analytical Biochemistry 237 70

Schier R R Bahnt A McCall G Apeh J Larπck and J Marks 1996 Identification of functional and structural ammo acid residues by parsimonious mutagenesis Gene 169 147

Scnultz J 1996 Tne combinatorial library a multifunctional resource Bioiecnnol Prog 12 729

Selvaraj P M L Plunkett M Dustin M E Sanders S Shaw and T A, Springer

ill 1987. The T lymphocyte glycoprotein CD2 binds the cell surface ligand LFA-3. Nature 326:400.

Springer, T.A., M.L. Dusttin, T.K. Kishimoto, and S.D. Martin. 1987. The lymphocyte function-associated LFA-1, CD2, and LFA-3 molecules: ceh adhesion receptors of the immune system. Ann. Rev. Immunol. 5:223.

Springer, T.A. 1990. Adhesion receptors of the immune system. Nature 346:425.

Thompson, C.B., T. Lindsten, J.A. Ledbetter, S.L. Kunkel, H.A. Young, S.G. Emerson,

J.M. Leiden, and CH. June. 1989. CD28 activation pathway regulates the production of multiple T-ceh derived lymphokines/cytokines. Proceedings of the National Academy of Sciences of the United States of America 86: 1333.

Tulip, W.R., J.N. Varghese, R.G. Webster, W.G. Lover, and P.M. Colman. 1992.

Crystal structures of two mutant neuraminidase-antibody complexes with amino acid substitutions in the interface. /. Mol. Biol. 227( 1): 149.

Turka, L.A., P.S. Linsley, H. Lin, W. Brady, J.M. Leiden, R.Q. Wei, M.L. Gibson, X.G. Zheng, S. Myrdal, D. Gordon, T. Bailey, S.F. Boiling, and C.B. Thompson. 1992. T- ceh activation by the CD28 ligand B7 is required for cardiac allograft rejection in "vivo. Proceedings of the National Academy of Sciences of the United States of America 89: 1 1 102.

Van Gool, S.W., M. de Boer, and J.L. Ceuppens. 1994. Atioantigen-specific anergy is induced by anti-B7 and cyclosporin A /. Exp. Med. 179:715

Van Gool, S.W., M. de Boer, and J.L Ceuppens. 1993. CD28 ligation by mAb or B7/BB1 provides an accessory signal for the CsA-resistant generation of cytotoxic T ceh activity. /. Immunol. 150:3254. Van Noesel, C, F. Miedema, M. Brouwer, M.A. De Rie, L.A. Aarden, and R.A. Van Lier. 1988. Regulatory properties of LFA-1 alpha and beta chains in human T- lymphocyte activation. Nature 333:850.

Vandenberghe, P., G.J. Freeman, L.M. Nadler, M.C. Fletcher, M. Kamoun, L.A.

Turka, J.A. Ledbetter, C.B. Thompson, and CH. June. 1992. Antibody and B7/BB1- mediated ligation of the CD28 receptor induces tyrosine phosphorylation in human T cehs. Journal of Experimental Medicine 175:951.

Vandenberghe, P., J. Delabie, M. de Boer, C. De Wolf-Peeters, and J.L. Ceuppens.

1993. In situ expression of B7/BB1 on antigen presenting cehs and activated B cehs: an immunohistochemical study. Int. Immunol. 5:317.

Vimekas, B., L. Ge, A. Pluckthun, K.C. Schneider, G. We nhofer, and S.E. Moroney. 1994. Trinucleotide phosphoramidites: ideal reagents for the synthesis of mixed oligonucleotides for random mutagenesis. Nucleic Acids Research 22(25):5600.

Waterhouse, P., J.M. Penninger, E. Timms, A. Wakeham, A. Shahinian, K.P. Lee, C.B. Thompson, H. Griesser, and T.W. Mak. 1995. Lymphoproliferative disorders with early lethality in mice deficient in Ctla-4. Science 270(5238):985.

Weiss, A., and J.B. Imboden. 1987. Ceh surface molecules and early events tinvolved in human T lymphocyte activation. Adv. 1mm unol . 41 : 1.

Wendoloski, J.J., J. Shen, M.T. Olivia, and P.C. Weber. 1993. Biophysical tools for structure-based drug design. Pharmac. Tner. 60: 169.

Wiley, R.A., and D.H. Rich. 1993. Peptidomimetics derived from natural products. Medicinal Research Reviews 13(3):327.

Winter, G., and W.J. Harris. 1993. Humanized antibodies. Immunol. Today 14(6):243. Young, J.W., L. Koulovα, S.A. Soergel, E.A. Clark, R.M. Steinman, and B. Dupont. 1992. The B7/BB1 antigen provides one of several costimulatory signals for the activation of CD4 + T lymphocytes by human blood dendritic cells in vitro [pubhshed erratum appears in J Clin Invest 1993 Apr;91(4): 1853]. Journal of Clinical

Investigation 90:229.

Zhang, D., and R.P. Johnson. 1997. Molecular cloning and comparative analysis of the rhesus macaque costimulatory molecules CD80 (B7.1) and CD86 (B7.2). Ceil Immunol. '97.

Claims

1. A molecule which cross-links, or cross-reacts with, B7.1 and B7.2, and which does not comprise a variable domain of a monkey antibody or the extracellular domain of CTLA4 or CD28.

2. A molecule according to claim 1, wherein said molecule comprises at least one first domain which binds B7.1 or B7.2 or cross-reacts with B7.1 and B7.2, at least one second domain which binds B7.1 or B7.2 or cross-reacts with B7.1 and B7.2, and optionally a third domain which couples the first and the second domcdn(s).

3. A molecttie according to claim 2, wherein said first and second domain is a polypeptide or a low-molecular-weight nonpeptide molecule.

4. A molecule according to claim 2, wherein said first and second domain is an anti-B7.1 antibody or an anti-B7.2 antibody or an antibody which cross-reacts with both B7.1 and B7.2, or, a humanized antibody, a single-chain fragment or another fragment thereof which has largely retained the specificity of said antibodies.

5. A molecttie according to claim 2 wherein said third domain is a (poly)peptide, any chemical coupting agent or any oligomerization domain.

6. A molecule according to claim 4, wherein said anti-B7.1 and anti-B7.2 antibody are monoclonal antibody B7-24 or-5B5 and monoclonal antibody IGIO, 3H10, 5F3, 7B8, 9D8, 1 1B9, 13B9, 13D3 or 14F1, respectively.

7. A molecule according to claims 1 to 6 selected from the group consisting of miniantibodies, diabodies, trivalent antibodies, tetravalent antibodies, small antigen -binding peptides and low-molecular -weight nonpeptide molecules.

8. A molecule according to any of claims 1 to 7, wherein said cross-linking of B7.1 and B7 2 results the inhibition of lmmuno-activating soluble mediators and/or the activation of lmmuno-inhibiting soluble mediators

9 A molecule according to claim 8, wherem scad rmmuno-activatmg soluble mediators are selected from the group consisting of IL-1 IL-6 IL-12 and TNF-a, and wherem said lmrnuno-inhibiting soluble mediators are selected from the group consisting of IL- 10, TGF-╬▓ and prostaglandms

10 A method for producmg a molecule according to any of claims 1 to 9

1 1 A pharmaceutical composition compnsmg a molecule according to any of claims 1 to 9 m a pharmaceutically acceptable excipient

12 A molecule according to any of claims 1 to 9 or a composition according to claim 11 for use as a medicament

13 A molecule according to any of claims 1 to 9 or a composition according to claim 11 possibly in combination with immunosuppressive agents for use m inhibiting antigen-specific ceh activation

1 A molecule according to any of claims 1 to 9 or a composition according to claim 11 possibly m combination with immunosuppressive agents for use m preventing or treating diseases of the immune system in particular for preventing or treating graft rejection graft versus host disease allergy and autoimmune diseases

15 molecule according to claims 13 or 14 wherem scad immunosuppressive agent is cnosen irom the group consisting of cyclosporm A FK 506 rapamycin OI T-3 OKT-4 SB 210396 T10B9 BTI 322 Mycophenolate mofetil anti-thymocyte globulin anti-lymphocyte immunoglobulin the combination of cyclosporm A azathioprine and glucocorticosteroids azathioprine leflunomide adenosrn de╬▒min╬▒se inhibitor; purtine nucleoside phosphoryl╬▒se inhibitor; MHC-peptide and IL-2 receptor mAb.