WO1999017801A2

WO1999017801A2 - Materials and methods relating to the stimulation of cells

Info

Publication number: WO1999017801A2
Application number: PCT/GB1998/002978
Authority: WO
Inventors: Farzin Farzaneh; David Darling
Original assignee: Cancer Research Campaign Technology Limited
Priority date: 1997-10-03
Filing date: 1998-10-05
Publication date: 1999-04-15
Also published as: GB9721075D0; AU9276398A; WO1999017801A3

Abstract

The application provides a method of producing a complex for activating an immune cell against an antigen. The method comprises the steps of: applying a linker to the surface of an activating cell or to the surface of a component of an activating cell, the linker attaching in a non-immunologically specific manner to said surface; and connecting a co-stimulator to the surface via the linker such that the co-stimulator can function in the activation of the immune cell.

Description

MATERIALS AMD METHODS RELATING TO THE STIMULATION OF CELLS

Field of the invention

The present invention relates to materials and methods involved in the stimulation of cells. Particularly, but not exclusively, it relates to cells or components thereof capable of stimulating an immune response against an antigen. Background to the Invention

T-cells play a crucial role in several events of the immune system (Julius et al . Immunol . Today. 1993 _/ 14 : 177 - 183) . Not only are they involved in the recognition of foreign antigens and destruction of virally infected cells, but they also provide help to B-cells for the production of antibodies involved in the neutralization and destruction of foreign antigens. Before T-cells can perform their immune function, they need to recognise their target antigen presented to them by professional antigen- presenting cells (APCs) , such as dendritic cells, macrophages, Langerhans cells, and B cells. These cells use immune molecules otherwise known as major histocompatibility complex (MHC) in the mouse, or human leucocyte antigen (HLA) in humans, to form complexes with the antigen in a form that is recognised by the T-cell

( Germain et al . Cell . 1994 ; 76 : 287-299) . There are two classes of MHC molecules, both of which are highly polymorphic. MHC class I is expressed normally on all nucleated cells and is mainly involved in the presentation of endogenous protein fragments derived either from self antigens, which are not recognised as foreign, or from viral gene products, which are recognised as foreign.

{ Howard et al . Curr Opin Immunol . 1995; 7 : 69 - 76) . MHC class II molecules are more restrictive and are expressed normally on the professional APCs (eg, B-cells, dendritic cells, macrophages, and Langerhans cells) . MHC class II molecules normally present fragments of proteins that have entered the cell by phagocytosis or endocytosis, mainly derived from infectious organisms such as bacteria and parasites {Harding et al . Trends Cell Biol . 1995_/ 5 : 105- 109) .

In order to recognise the MHC-antigen complex on the APCs, each T-cell uses its unique T-cell receptor (TcR) or CD3 complex, which, together with the CD8 or CD4 coreceptors, binds to MHC class I or II, respectively {Davi s et al . Nature . 1988 ; 334 : 395 -402) . However, TcR occupancy alone is not sufficient to activate T-cells to respond; in addition, these cells need to receive antigen- independent signals (co-stimulatory signals) obtained by engagement of some of their other cell surface molecules, with complementary ligands expressed on the surface of the APC { Van Seventer et al . Curr Opin Immunol . 1991 _/ 3 : 294 - 303 ) . Depending on their nature, these molecules serve two purposes: they stabilise the interaction of the T-cell with APCs and transduce co-stimulatory signals leading to cytokine secretion and/or proliferation and/or induction of their effector functions. Co-stimulatory signals are provided most efficiently by dendritic cells, followed (in order of descending potency) by macrophages, activated B- cells, and resting B-ceils. Although several cell surface molecules, such as LFA-3 and ICAM-1, have been implicated as potential co- stimulatory molecules, a large number of recent studies show that an important co-stimulatory activity required for interleukin-2 (IL-2) -driven proliferation of T-cells is provided by members of the CD28 family on the T-lymphocyte surface {Linsley et al . Annu Rev Immunol . 1993 / 11 : 191 -212) . The ligands for these molecules are members of the B7 family, for example, B7 .1 , and B7.2 {Boussiotis et al . Proc Natl Acad Sci USA . 1993 / 90 : 11059 - 11063 . )

Once the T-cell is activated to respond to a particular antigen, it is then capable of recognition and/or destruction of target cells displaying the same antigen. While the activation of T-cells requires the co- stimulatory function, the effector phase (ie, destruction of the target cell) does not usually require interaction with co-stimulatory molecules on the target cells. Many cells, other than APCs display antigens on their surfaces via MHC molecules. For example, tumour cells may display antigens specific to the tumour cell ("tumour antigens") via MHC class I molecules or in some cases, class II molecules. However, these cells differ from APCs in that they do not possess the cell surface molecules capable of transducing co-stimulatory signals to the T- cell.

The use of cancer cells expressing cytokines as vaccines for immune gene therapy of cancer is already a reality, and subject to a number of clinical trials {Marcel

T, et al . Hum . Gene Therapy 1996_/ 7 : 2025-2046) . B7.1 and B7.2 in an allogeneic setting offer exciting possibilities and have entered the clinic on a trial basis {Marcel T, et al . Hum. Gene Therapy 1996; 7 : 2025-2046) . B7.1 treatment in autologous protocols has yet to reach the trial stage. Reasons for this are multifarious, but include access to sufficient tumour cells and inefficiency of the currently available gene transfer and expression procedures.

It has become clear that a number of laboratories are approaching the problem from the different and exciting perspective of modifying cells ex vivo by non-genetic means such that the cells display the products of the genes of interest without the requirement for genetic modification {McHugh RS et al . Proc . Natl . Acad . Sci . USA 1995_/ 92 : 8059-8063 / Brunschwig EB et al , J. Immunol . 1995_/ 155: 5498 -5505_/ Brunschwig EB et al , FASEB . J. 1996_/ 10 : (6) pA1348 Abstr 2012_/ Brunschwig EB et al , FASEB J. 1995; 9: (4 pt2) pA811 Abstr 4700) . The B7.1 gene has been engineered as a fusion protein, with either the CD16B transmembrane domain {McHugh RS et al . Proc . Natl . Acad . Sci . USA 1995; 92 : 8059- 8063) or sequences from decay- accelerating factor (DAF) {Brunschwig EB et al , J. Immunol . 1995; 155: 5498 -5505; Brunschwig EB et al , FASEB . J. 1996; 10 : (6) pA1348 Abstr 2012; Brunschwig EB et al , FASEB J. 1995; 9 : (4 pt2 ) pA811 Abstr 4700) , which are subsequently glycosyl-phosphatidylinositol (GPI) -modified within the cells to acquire a "GPI anchor" . Purification of this protein/GPI species allows the complex as a whole to be applied to cells in such a way that the GPI portion of the molecule incorporates into the cell membrane and anchors B7.1 to the cell surface {McHugh RS et al . Proc . Natl . Acad . Sci . USA 1995; 92 : 8059-8063 ; Brunschwig EB et al , J. Immunol . 1995; 155: 5498 -5505; Brunschwig EB et al , FASEB . J. 1996; 10 : (6) pA1348 Abstr 2012) . So far the number of cells required (10 grams of CHO cells for 10-20 tests) to purify sufficient B7.1-GPI coupled with problematic batch quality {McHugh RS et al . Proc . Natl . Acad . Sci . USA 1995; 92 : 8059-8063 ) make this a difficult system to use. Summary of the Invention

The present applicants have appreciated that there is a need for a complex which has the ability to mediate a immune response by mimicking an APC, but that can be produced in an efficient and reliable way by avoiding many nucleic acid based procedures.

With this in mind, the applicants have investigated strategies for associating co-stimulators with target cells. In relation to the present invention, the term "co- stimulator" is used herein to define any stimulatory factor which can function in the activation of an immune cell. This definition includes classically defined costimulatory molecules, (ie. membrane bound or secreted products of APCs that activate signal transduction events in addition to those induced by MHC/TcR interactions) as well as adhesion molecules and their receptors and cytokines and their receptors. The applicants have realised that by associating a co-stimulator to the target cell by external conjugation, the more time consuming and more difficult nucleic acid based procedures which involve the introduction of genetic material into cells can be avoided. However, it is important that the co-stimulators are associated with the target cell such that they remain closely associated with it but also retain their co-stimulatory function.

Therefore, the present invention provides cells which mimic APCs in that they are capable of activating an immune response through one or more surface displayed co- stimulator (s) . The one or more surface displayed co- stimulator (s) may activate the specific immune response by acting in combination with a surface displayed antigen. The inventors were pleasantly surprised that molecular interferences between the linked molecules (see below) such as steric hindrance did not prevent the co-stimulator from functioning in the activation of an immune response.

The invention provides a method of producing a complex for activating an immune cell against an antigen which comprises the steps of applying a linker to the surface of an 'activating' cell or to the surface of a component of an activating cell, the linker attaching in a non- immunologically specific manner to the surface; and connecting a co-stimulator to the surface via the linker such that the co-stimulator can function in the activation of the immune cell.

In the above the term 'activating cell' is used. The term 'target cell' is also used. Both terms denote the same cell in that they both may be of any cell type which is capable of carrying on its surface, via a linker moiety, one or more co-stimulator (s) . The presence of the co- stimulator (s) allows the cell ('activating' or 'target') to play a role in the activation of an immune cell.

One or more co-stimulators may be connected to the surface via one or more linkers. If two, three or more co- stimulators are used they may function in combination to activate an immune response.

A co-stimulator may comprise part or all of a binding domain of an antibody. The co-stimulator may comprise part or all of a binding domain of an antibody which is connected to the linker via protein A or protein G or a fusion protein of protein A or protein G.

A co-stimulator may be capable of binding a cluster differentiation molecule. A cluster differentiation molecule may for example be selected from the CD2 , CD3 ,

CD5, CD16, CD28 or CD43 cluster differentiation families.

The linker may be biotin. The co-stimulator may be connected to a biotin linker via avidin or streptavidin. A co-stimulator may be a fusion protein comprising avidin or streptavidin.

The immune cell may be a splenocyte.

The immune cell may be a T cell. It may be a cytotoxic cell (e.g. a T_c cell, a natural killer (NK) cell, or a lymphokine-activated killer (LAK) cell) , or it may be an antigen presenting cell (APC) , a B cell, or a macrophage .

The activating cell may be a tumour cell.

The complex may be produced in vivo . The present invention also provides a method of activating an immune cell against an antigen by producing a complex as stated above and contacting the complex with an immune cell .

The immune cell can be activated in vitro or in vivo. Therapeutic methods are provided e.g. for the purposes of vaccination or treatment. In such a method, a complex obtainable as stated above is contacted with immune cells of a subject so as to activate them. This may be achieved by forming the complex in vivo or by forming the complex in vitro and then administering it to the subject.

The present invention also provides pharmaceutical products which comprise a complex obtainable as stated above. Also provided are kits for producing a medicament, the kit comprising a linker (e.g. biotin succinimide ester) and a co-stimulator and instructions for preparing a complex as stated above. The instructions may state a time frame for administration of the medicament to a mammalian patient .

The present invention also provides a method for evaluating the capability of a molecule or combination of molecules to activate an immune cell against an antigen. In the examples provided herein it is demonstrated that an 'activating cell' carrying a plurality of co-stimulators have therapeutic utility. The beneficial effect may be obtained by use of just one or two co-stimulators or by use of combinations of co-stimulators different to those specifically exemplified. Different individual therapeutic co-stimulators and effective combinations thereof may be readily identified using the techniques provided herein.

Thus the invention provides that one may evaluate the ability of a molecule or combination of molecules to activate an immune cell against an antigen which involves making a test complex by applying one or more linkers to the surface of an activating cell or to the surface of a component of an activating cell, the linker or linkers attaching in a non-immunologically specific manner to the surface and connecting the molecule or molecules being evaluated to the surface via the one or more linkers and testing the complex for ability to activate the immune cell. Activation of an immune cell can be ascertained as particularly disclosed herein and by other generally known methods e.g. by measuring release of IL-2.

The antigen may be of a tumour cell. The antigen may be substantially restricted to tumour cells or more prevalent on tumour cells as compared to non-tumour cells.

Thus, the present invention provides a complex for activating a specific immune response against an antigen, comprising a biotinylated target cell or component thereof displaying or capable of displaying said antigen; and a co-stimulator externally conjugated or otherwise connected to the biotinyated target cell via its biotinylated surface such that the co-stimulator and the antigen can act in combination to activate the immune response.

The present invention further provides a method of producing a complex for activating a specific immune response against an antigen, comprising the steps of obtaining a target cell or component thereof displaying the antigen or capable of displaying the antigen; biotinylating the surface of said target cell or component thereof ; and externally conjugating or otherwise connecting a co- stimulator to the biotinyated target cell or component thereof via its biotinylated surface such that the antigen and the co-stimulator can act in combination to activate the immune response .

The present invention also provides a method of producing a specific immune response against an antigen comprising obtaining a complex according to the first aspect and contacting said complex with immunological substances capable of eliciting a specific immune response against said antigen.

The present invention also provides a method of producing a specific immune response against an antigen comprising the steps of obtaining a target cell or component thereof displaying the antigen or capable of displaying the antigen; biotinylating the surface of said target cell or component thereof; externally conjugating or otherwise connecting a co- stimulator to the target cell or component thereof via its biotinylated surface to provide a complex that displays the antigen and the co-stimulator such that they can act in combination; contacting said complex with immunological substances capable of eliciting a specific immune response against said antigen.

As discussed herein, the complexes and methods of the present invention are of use for immunotherapy both for treatment and prophylaxis. They may also be used in evaluating the capability of molecules for activating immune cells (eg. T cells) .

By biotinylating the cell surface, the conjugation or connection of the co-stimulator is not dependent on particular cell surface molecules present on the cell surface. For this reason, the same method of conjugating or connecting the co-stimulator may be used for any cell type. This means that no knowledge is required of the cell in order for the co-stimulator to be conjugated.

In all of the above aspects, the target cell is preferably a cell that displays and re-displays antigen on its surface via an MHC protein. The target cell is most preferably a tumour cell. However, it will be appreciated that the target cell merely has to serve to display antigen. Therefore, it may be an inactivated cell or even a component of a cell which retains the ability to display the antigen but has been deprived of part or all of its components necessary for normal cellular function. This may be particularly useful with regard to a tumour cell where it may be desirable to deprive it of part or all of its tumour associated functions. It is usual for nearly all cells to have class I MHC proteins present. These proteins bind peptides and display them on the surface of the cell. The peptides are usually derived from self proteins. In addition, in some cells a virus particle may be digested by the cell and the resulting peptides present on the surface of the cell by its MHC proteins. These presented peptides act as antigens with regard to other immunological substances.

The present invention is particularly well suited to tumour cells. These cells present peptides (hereinafter termed "tumour antigens") on their surfaces by their MHC proteins. By activating a specific immune response against a tumour antigen in accordance with the present invention, all other tumour cells displaying that antigen will become the target of that response even though they do not display the co-stimulator.

The following description will be directed to tumour cells and tumour antigens. However it will be appreciated by the skilled person that the invention may be applied to other cells, for example those presenting foreign antigens or viral antigens. By externally conjugating or connecting the co- stimulator to a biotinylated cell, nucleic acid based procedures are avoided. The co-stimulator may be prepared with a conjugating or connecting assembly prior to attachment to the cell. This avoids expression of the co- stimulator from within the cell or having to provide the co-stimulator with a transmembrane binding member. Suitable conjugative assemblies include any binding members that link the co-stimulator and the biotinylated cell surface. For example, streptavidin or avidin with or without biotinylated protein G or protein A.

Preferred conjugating assemblies are described below. As mentioned above, all of these conjugative assemblies allow the co-stimulator to be conjugated to any cell type as long as its surface has been biotinylated. In other words, the conjugative assembly is not reliant on a particular surface molecule on the target cell and therefore no knowledge of the target cell surface is required beforehand.

The co-stimulator may be any member that is capable of binding to an immunological substance such that when acting in combination with an antigen a specific immune response is elicited against said antigen. The co-stimulator preferably binds to a surface molecule on the immunological substance and causes the transduction of co-stimulatory signals leading to responses such as cytokine secretion, proliferation and induction of other immunological responses .

Preferred co-stimulators include ligands, antibody binding domains or fragments thereof that can bind and activate surface molecules of immunological substances such as T-cells, antigen presenting cells (APC) , macrophages, or natural killer (NK) cells. Many co-stimulators will be apparent to the skilled person but preferred examples include members that will activate the CD28 receptor family found on T-cells or the CD16 family found on NK cells. Examples of such co-stimulators include those of the B7 family. Further co-stimulators include members that will activate the CD2 , CD3 , CD5 and CD43 receptors such as antibodies directed against said receptors or ligands such as LFA3 and ICAM I. Other co-stimulators will be apparent to the skilled person.

Particularly preferred co-stimulators are antibody binding domains that bind to and thereby activate downstream events via molecules or counter receptors of the immunological substances. Although other immunological substances will be apparent to the skilled person, for convenience the following description will focus on T- cells .

The co-stimulator may be externally conjugated or connected to the surface of the cell by any of the conjugating assemblies as described above that allows it to be displayed such that it can interact with the immunological substances. The co-stimulator may itself be biotinylated and conjugated to the target cell surface via an avidin or strepavidin bridge. Alternatively, the co- stimulator may be linked to a streptavidin or avidin molecule via some other chemical linking means, or they may be produced together as a fusion protein. In this way the strepavidin or avidin molecule will bind with the biotinylated tumour cell thereby conjugating the co- stimulator with the tumour cell.

As a further aspect of the present invention there is provided a method for anchoring a molecule to a tumour cell surface comprising the steps of biotinylating the tumour cell surface; coupling said molecule with avidin or streptavidin; and linking the molecule to the biotinylated cell surface via the avidin or streptavidin such that the molecule is anchored to the cell surface. The molecule may be biotinylated such that it binds to the streptavidin or avidin. In this way the streptavidin or avidin forms a bridge between the biotinylated cell surface and the biotinylated molecule. Alternatively, the molecule may be produced with the streptavidin or avidin as a fusion protein using standard known nucleic acid based procedures .

Co-stimulators may be antibody binding domains that bind to and thereby activate the surface molecules of the immunological substances . An antibody binding domain may comprise a whole antibody or may just be a fragment thereof. A particularly preferred antibody is anti-human CD28 monoclonal antibody.

As antibodies can be modified in a number of ways, the term "antibody" should be construed as covering any specific binding substance having a binding domain with the required specificity. Thus, this term covers antibody fragments, derivatives, functional equivalents and homologues of antibodies, including any polypeptide comprising an immunoglobulin like domain, whether natural or synthetic. Chimaeric molecules comprising an immunoglobulin binding domain, or equivalent, fused to another polypeptide are therefore included. Cloning and expression of chimaeric antibodies are described in EP-A- 0120694 and EP-A-0125023. It may be desirable to "humanise" non-human (eg murine) antibodies to provide antibodies having the antigen binding properties of the non-human antibody, while minimising the immunogenic response to the antibodies, eg when they are used in human therapy. Thus, humanised antibodies comprise framework regions derived from human immunoglobulins (acceptor antibody) in which residues from one or more complementary determining regions (CDR's) are replaced by residues from CDR's of a non-human species (donor antibody) such as mouse, rat or rabbit antibody having the desired properties, eg specificity, affinity or capacity. Some of the framework residues of the human antibody may also be replaced by corresponding non-human residues, or by residues not present in either donor or acceptor antibodies. These modifications are made to further refine and optimise the properties of the antibody. It has been shown that fragments of a whole antibody can perform the function of binding antigens . Examples of binding fragments are (i) the Fab fragment consisting of VL, VH, CL and CHI domains; (ii) the Fd fragment consisting of the VH and CHI domains; (iii) the Fv fragment consisting of the VL and VH domains of a single antibody; (iv) the dAb fragment (Ward, E.S. et al . , Nature 341, 544-546 (1989)) which consists of a VH domain; (v) isolated CDR regions;

(vi) F(ab')2 fragments, a bivalent fragment comprising two linked Fab fragments (vii) single chain Fv molecules (scFv) , wherein a VH domain and a VL domain are linked by a peptide linker which allows the two domains to associate to form an antigen binding site (Bird et al , Science, 242, 423-426, 1988; Huston et al , PNAS USA, 85, 5879-5883, 1988) ; (viii) bispecific single chain Fv dimers (PCT/US92/09965) and (ix) "diabodies", multivalent or multispecific fragments constructed by gene fusion (WO94/13804; P. Holliger et al Proc. Natl. Acad. Sci. USA 90 6444-6448, 1993) .

Where the co-stimulator is an antibody, the antibody itself may be biotinylated or a further linker molecule which is capable of binding to the antibody may be biotinylated. For example, protein G or protein A are capable of binding the Fc portion of an antibody. In accordance with the present invention these proteins may be biotinylated to form for example a protein G-biotin which binds to both the streptavidin/avidin bridge and the Fc portion of the antibody thereby conjugating the antibody co-stimulator to the surface of the tumour cell. Likewise, streptavidin/avidin may be coupled directly to these proteins to form, for example, a chimeric streptavidin- protein A molecule. This can be achieved as a fusion protein such that the molecule is expressible using standard procedures (Sano T, Biotech. N.Y. 1991; 9:1378- 1381) . Commercially available reagents may be used to biotinylate cell surface proteins. Once the cell has been biotinylated, it may be desirable to allow time for the cellular displayed antigen to be redisplayed just in case the antigen itself has also become biotinylated. In a preferred embodiment, an avidin bridge (Conrad MK et al. Met . Enzymol .1990 ; 184: 641-653) in conjunction with protein G-biotin (Goward CR et al , Biochem . J. 1990; 267: 171-177) may effectively couple an antibody such as mouse anti-human CD28 in an orientation specific manner. This then allows the antibody to remain associated with the cell long enough to allow the activation of the immune cells (e.g. T cells, APCs or NK cells) and provide the necessary co-stimulatory signal for the functional activation of the reactive T-cells. Aspects and embodiments of the present invention will now be illustrated, by way of example only, with reference to the accompanying figures. Further aspects will be apparent to those skilled in the art. All documents mentioned in this text are incorporated herein by reference .

Brief description of the drawings

Figure 1 Cartoon of Biotin/Avidin bridge. A representation of the cell membrane shows a putative accessible NH₂ group. This is then shown as biotinylated, and the biotin detected by streptavidin-FITC: biotinylated and avidin treated and the avidin detected by biotin-FITC: biotinylated and avidin treated and incubated with protein G' -biotin and mouse anti human CD28: and finally the above complex detected with FITC-conjugated F(ab')₂ fragment of rabbit anti-mouse immunoglobulins .

Figure 2 Conjugation of mouse T-cells with biotinamidocaproate N-hydroxysuccinimide ester. 2x10 7 TIB232 cells in 400μl of ice cold PBS pH8.0 on ice were biotinylated with 3μl of biotinamidocaproate N- hydroxysuccinimide ester as described in the materials and methods. The cells were then washed twice in RPMI+10%FCS, and adjusted to 6xl0⁵/ml with or without the addition of 0.05% sodium azide. At which point 5xl0⁵ untreated (A) or biotinylated (B) cells were stained with streptavidin-FITC

(B) . This was repeated at 18 (C-E) , 48 (F-H) and 72 (I-K) hours, with either untreated (C,F & I) biotinylated (D,G &

J) or 0.05% sodium azide incubated biotinylated cells (E,H & K) . At each time the cells were analysed by FACS as described.

Figure 3 Biotinylation of cell surface proteins changes the immunological profile of the cells. 4xl0⁷ TIB232 B7 clone 2 cells in two 400μl aliquots on ice, were either incubated with 3μl of DMSO or 3μl of 366mM biotinamidocaproate N- hydroxysuccinimide ester in DMSO as described in materials and methods. After 30 minutes incubation cells were washed twice and placed in culture in RPMI+10%FCS at 5xl0⁵/mιl. Immediately afterwards 5xl0⁵ cell aliquots of either untreated (DMSO alone) (A, C, E,G, I , K,M & 0) or biotinylated

(B,D,F,H, J,L,N & P) cells were analysed for immunofluoresence as described (A-J) or were analysed after

24 hours in culture (K-P) . Cells were either unstained (A,B,K & L) , streptavidin-FITC incubated (C,D,M & N) , stained with ICAM-1 followed by rabbit anti rat FITC or secondary alone (*) (E and F) , FITC conjugated anti-mouse CD80 (B7-1) antibody (G,H,0 & P) or anti-MHC H-2 Class 1 (_H_₂ ^b,k,s, _P followed by FITC-conjugated F(ab')₂ fragment of rabbit anti mouse immunoglobulins (I & J) or secondary alone (*) as described and analysed by FACS as described.

Figure 4 Biotinylated cells can bind streptavidin. 2xl0⁷ TIB232 cells in two 400μl aliquots on ice, were either incubated with 3μl of DMSO or 3μl of biotinamidocaproate N- hydroxysuccinimide ester as described in materials and methods. After 30 minutes incubation cells were washed twice and resuspended in 2ml HBSS+1%FCS and further divided into five 400μl aliquots to which was added lmg/ml streptavidin to a final concentration of 0 (B & D) , 7 (E & F) , 40 (G & H) and 160μg/ml (I-K) . After 30 minutes on ice cells were washed twice in 10ml HBSS+1%FCS and 5xl0⁵ cells stained either with biotin-FITC (A,B,E,G,I & J) or streptavidin-FITC (C,D,F,H,K & L) and analysed by FACS as described.

Figure 5 Cells labelled with avidin can bind biotinylated antibody. 2xl0⁷ TIB232 cells 400μl on ice were incubated with 3μl of biotinamidocaproate N-hydroxysuccinimide ester as described in materials and methods. After 30 minutes cells were washed twice in RPMI+10%FCS, adjusted to 8xl0⁵/ml and cultured overnight. After 18 hours 2xl0⁷ biotinylated and untreated cells were avidin (or mock) treated, washed as described in materials and methods and resuspended in HBSS+1%FCS to a concentration of 2xl0⁷/ml. lOOμl was taken and biotinylated antibody bound as described in materials and methods. Cells were stained with either biotin-FITC (C & D) or FITC-conjugated F(ab')₂ fragment of rabbit anti mouse immunoglobulins (A,B,E-H) and analysed by FACS as described.

Figure 6 Cells with bridged avidin can be used to assemble a complex with protein G' -biotin and antibody. 8xl0⁷ TIB232 cells in 1.6ml on ice were incubated with 6μl of biotinamidocaproate N-hydroxysuccinimide ester as described, washed twice in RPMI+10%FCS and cultured at 37°C overnight. After 18 hours 2.4xl0⁷ biotinylated cells

(B,D,E,G,H, I,K,M,N, P,Q & R) and untreated equivalents (A,C,F,J,L & 0) were avidin treated as described, after which cells were washed twice in HBSS+1%FCS, resuspended in 300μl of HBSS+1%FCS, lOOμl aliquots (8xl0⁶each) were then added either to carrier alone (A- I) or a preformed complex of 42μg mouse anti-human CD28 + 8.5μg protein G' -biotin (J- R) . After a further 30 minutes cells were washed 2x in DMEM+10%FCS (DMEM is deficient in biotin) , and a sample taken immediately for staining and FACS analysis as described (T=0 time point, A,B,J & K) , the remaining cells were divided into 2 and cultured either at 37°C for 3.5 hours (C,E,L & N) and 6 hours (F,H,0 & Q) or 4°C for 3.5 hours (D & M) or 6 hours (G & P) harvested and stained for FACS analysis. I & R represent duplicates of B & K that were returned to culture after staining and incubated for a further 6 hours at 37°C .Before processing the biotinylation efficiency was checked and found to have worked as normal (data not shown) . All samples stained as indicated on figure and analysed by FACS as described.

Figure 7 Mouse anti -human CD28 antibody complexed transiently with cells is functional as a co-stimulator. 8xl0⁷ TIB232 cells were biotinylated in 1.6ml PBS pH8.0 essentially as previously described and cultured overnight. Biotinylated and untreated cells were then taken and processed as previously described with lmg/ml avidin for every 7xl0⁷ cells in 900μl HBSS+1%FCS and then 20μg protein G' -biotin in 200μl for every 3.5xl0⁷ cells and lOOμg anti- CD28 in 200μl for every 3.5 xlO⁷ cells, in the combinations described below. After complex assembly cells were washed four times and cell samples analysed by FACS to confirm efficiency of the complex formation. The remaining cells were kept in RPMI+10%FCS on ice. Cells were then added to the Jurkat assay as described. Conditions were either Jurkat cells alone, untreated (control) cells incubated with mouse anti-CD28 (C/CD28) , untreated (control) cells incubated with avidin+protein G' -biotin+anti-CD28

(C/AV/PG/CD28) , biotinylated cells incubated with avidin+protein G' -biotin+anti-CD28 (B/AV/PG/CD28) or biotinylated cells incubated with avidin and protein G' - biotin alone (B/AV/PG) . The human IL-2 ELISA detects Human IL-2 alone, and does not cross-react with recombinant murine IL-2 in our hands (data not shown) . Data points show the pg/ml of quadruplicate determinations of IL-2 (+ one standard deviation) . Figure 8 Mouse anti-human antibody complexed transiently with cells co-stimulates as a function of the antigenic target. 7xl0⁷ TIB232 cells were biotinylated in 1.6ml PBS pH8.0 essentially as previously described and cultured overnight. Biotinylated cells were then taken and either processed as previously described with 2mg/ml avidin for every 3.6xl0⁷ cells in 360μl HBSS+1%FCS, or mock avidin treated, cells were washed twice and resuspended in lOOμl HBSS+1%FCS for every 1.2xl0⁷ cells. To this was added a premix (30 minutes) of 6.5μg of protein G' -biotin and 34μg of either anti CD28, CD3 or CD14. After complex assembly cells were washed four times and cell samples analysed by FACS to confirm efficiency of the complex formation. The remaining cells were kept in RPMI+10%FCS on ice. Cells were then added to the Jurkat assay as described. Conditions were either Jurkat cells alone (No addition) , biotinylated cells, no avidin, with protein G' -biotin+anti- CD28 (B+PG/CD28) , biotinylated cells incubated with avidin+protein G' -biotin+anti-CD28 (B/AV+PG/CD28) , biotinylated cells incubated with avidin+protein G' - biotin+anti-CD3 (B/AV+PG/CD3) , and finally biotinylated cells incubated with avidin+protein G' -biotin+anti-CD14 (B/AV+PG/CD14) . The human IL-2 ELISA detects Human IL-2 alone, and does not cross-react with recombinant murine IL- 2 in our hands (data not shown) . Data points show the pg/ml of triplicate determinations of IL-2 (± one standard deviation) . Crossed bars represent those samples whose IL- 2 concentration exceeded the accurate range of the ELISA and were plotted as simply more than 2000pg/ml. Figure 9 illustrates the optimization of non specific T cell activation and IL-2 secretion in vitro. Splenocytes from naive DBA/2 female mice were plated at 5xl0⁴/well in round bottom 250μl microtire plates. Hamster antibodies directed against the murine CD28 and CD3e and Rat antibodies against murine CD5 were added to a final concentration of 0. lμg/ml each either alone or in combination. In addition, half the wells also included 5xl0⁴ P815 cells. The plate was incubated at 37°C for 48hrs, the supernatant harvested and assayed for the presence of murine IL-2 by ELISA. Results are expressed as pg/ml IL- 2/5X10⁴ splenocytes/48hrs, each point is the mean of three values. The value <15pg represents the limit of reliable detection from the IL-2 standard dose curve.

Figure 10 shows a comparison of two antibody complexing procedures for the induction of IL-2 secretion. Splenocytes from naive DBA/2 female mice were plated at 5xl0⁴/well in round bottom 250μl microtire plates. Antibodies (either biotinylated or otherwise) directed against the murine CD28, CD3e and CD5 were added to a final concentration of 0.1 or O.Olμg/ml each in the presence or absence of 5x10" non-modified P815 cells. Additionally P815 cells were biotinylated and avidin bridged and further incubated with:

1) P815 biotinylated and avidin brigdged (BAV)

2) Biotinylated anti CD3+CD5+CD28 (BAV-Bab's) 3) Biotinylated protein G' (BAV-PG)

4) Biotinylated protein G' and CD3+CD5+CD28 (BAV- PGab's) and lxlO⁵ of these cells cocultured with splenocytes. The plate was incubated at 37°C for 48hrs, the supernatant harvested and assayed for the presence of murine IL-2 by ELISA. Results are expressed as pg/ml IL-2/5X10⁴ splenocytes/48hrs, each point is the mean of three values. The value <15pg represents the limit of reliable detection from the IL-2 standard dose curve.

Figure 11 shows a comparison of the stability of two antibody complexing procedures . Splenocytes from naive DBA/2 female mice were plated at 5xl0⁴/well in round bottom 250μl microtire plates. P815 cells were biotinylated and avidin bridged and further incubated with: 1)P815 Biotinylated and avidin brdiged 2) Biotinylated anti CD3+CD5+CD28 (BAV-Bab's) 3) Biotinylated protein G' (BAV-PG)

4) Biotinylated protein G'and CD3+CD5+CD28 (BAV-PGab's) and 5xl0⁴ of these cells cocultured with splenocytes either immediately after complexing, or after 90 minutes in RPMI+10%FCS at 37°C followed by a washing step. The plate was incubated at 37°C for 48hrs, the supernatant harvested and assayed for the presence of murine IL-2 by ELISA. Results are expressed as pg/ml IL-2/5X10⁴ splenocytes/48hrs, each point is the mean of three values. The value <60pg represents the limit of reliable detection from the IL-2 standard dose curve. Figure 12 illustrates an assessment of antibody complexing procedure by FACS. After complexing, (T=0, A-H) Unstained, biotinylated (A) and control unmodified (B) cells were analysed, followed by control unmodified (C) and biotinylated cells (D) incubated with avidin-FITC to analyse biotinylation. Control cells and biotinylated and avidin bridged (BAV Az) cells were also incubated with biotin-FITC (E and F) respectively, detecting avidin bound to biotinylated cells. Finally, BAV Az (G) and BAV Bab's (H) incubated with Rabbit anti Rat F(ab')2 FITC and Goat anti Hamster IgG FITC determine the efficiency of biotin antibody complexing with BAV cells. The process described for (A-H) was repeated with samples incubated overnight in RPMI+10%FCS at 37°C, washed and stained (I-M) . (I) - Control unmodified, unstained. (J) - Control unmodified stained with biotin-FITC. (K) - BAV Az - stained with biotin-FITC. (L) - BAV Az - stained with Rabbit anti Rat FITC and Goat anti Hamster FITC. (M) - BAV Bab's - stained with Rabbit anti Rat FITC and Goat anti Hamster FITC.

Figure 13 illustrates the survival of female DBA/2 mice injected with modified P815 Tumour cells. 6-8 week old DBA/2 female mice were injected intrapertioneally with 2xl0⁵ cells/animal and culled on the basis of large ascites or obvious ill health. The modifications of the P815 are:

P815 control - P815 cells with no modifications; P815 Bab's - P815 cells with a mixture of biotinylated anti CD3 ,

CD28 and CD5 antibodies; P815 BAV Az - biotinylated P815 cells subsequentially incubated with avidin and the incubated in the presence of 0.075% Azide. P815 BAV Bab's biotinylated P815 cells subsequently incubated with avidin, then incubated with a mixture of biotinylated anti

CD3 , CD28 and CD5 antibodies in ther presence of 0.07%

Azide .

Figure 14 illustrates the disease status of the female DBA/2 mice injected with modified P815 Tumour cells prior to their death. Animals were examined every 2-3 days and thier abdomens palpated for signs of tumour lumps or ascites . Figure 15 illustrates a determination of the ability of modified P815 tumour cells to activate T-cells in vitro. Splenocytes (5xl0⁴) were plated in round bottom microtitre plates to which was added either antibody in solution, control P815 or modified cells, and the plates incubated at 37°C, additionally 4 and 24 hours later additional wells with splenocytes were then incubated with antibody or modified cells that had been returned to culture in RPMI+10%FCS at 37°C. And the plates returned to 37°C . The supernatant from these plates was harvested 72 hours after the initial culture and assayed by ELISA for the presence of murine IL-2 as a marker for non-specific activation of T cells.

Figure 16 Biotinylation of cell membranes is not limited to murine cell lines. Two primary AML samples (KR and TC) that do well in primary culture were taken for biotinylation 10 days after having been recovered from frozen. 5xl0⁶ cells in 500μl of PBS pH8.0 on ice were incubated with 3μl of biotinamidocaproate N- hydroxysuccinimide ester as described in materials and methods. After 30 minutes on ice cells were washed twice as described and stained with biotin-FITC as described in materials and methods. Control TC cells (A) do not stain with streptavidin-FITC whilst those that have been biotinylated (B) show a 2 log shift in fluorescence after streptavidin-FITC. The same is seen with the second sample KR cells after streptavidin-FITC when the control cells (C) are compared with those following biotinylation (D) , all samples analysed by FACS as described. Detailed description and exemplification of the invention Conjugation of mouse TIB232 cells wi th biotinamidocaproate N-hydroxysuccinimide ester

Exposed NH₂ groups on cell surface proteins (primarily side groups of lysine) are readily biotinylated by succinimide ester derivatives (Demant EJF et al . Biochim . Biophys . Acta 1991; 1118: 83-90). Incubation of, mouse AKR thymoma, TIB232 cells with 2.75mM biotinamidocaproate N- hydroxysuccinimide ester routinely resulted in a 2 log increase in FITC fluorescence, as measured by streptavidin- FITC immunofluoresence, compared to untreated cells. However the immunofluoresence on these cells rapidly declined and by 72 hours there was no more than a threefold difference [Figure 2 (A compared with B, I compared with J; see also Figure 8) ] . This decay was not seen when cells were in the presence of 0.05% sodium azide; approximately a greater than 1 log difference remaining after 72 hours (E compared with C and D, H with F and G, and K with I and J) . Thus biotinylated cells lose membrane bound biotin by an active process which can be blocked by the metabolic poison sodium azide at a concentration of 0.05%. At this concentration after 72 hours 50% of cells in sodium azide were dead; 0.1% azide was found to be even more toxic whilst 0.01% azide though less toxic was only marginally effective in preventing the disappearance of the detectable biotin. Sodium azide was therefore not use in subsequent studies unless otherwise specified.

Biotinylation of cell surface proteins changes the immunological profile of the cells Biotinylation has variable effects on the immunological profile of B7.1 transfected TIB232 cells (Figure 3). No difference was apparent in autofluorescence (A,B,K & L) , in the binding of secondary antibody (E*,F*,I* & J*), the immunodetection of adhesion protein ICAM-1 (E & F) or the detection of MHC Class I expression (I & J) . This is in contrast to the effect seen on the detection of a transfected and expressed mouse B7.1 (CD80) protein where detectable surface expression drops to half the normal level after biotinylation (G compared with H, Mean Fluorescence Index: MFI=72 & 31 respectively) . Levels return to near normal after 24 hours in culture (O & P, MFI=76 & 58 respectively) .

Biotinylated cells can bind streptavidin

Biotinylated cells can be incubated with high concentrations of streptavidin in such a way that the tetrameric structure of streptavidin allows the molecule to both stick to the cell surface and still retain active sites for the further binding of biotin-FITC. This is demonstrated in Figure 4: control, untreated, cells bind neither biotin-FITC (A) or streptavidin-FITC (C) and though biotinylated cells do not bind biotin-FITC (B) they bind large amounts of streptavidin-FITC (D) . After incubation with increasing amounts of streptavidin, biotinylated cells lose their affinity for streptavidin-FITC (D-L) . This is mirrored by their increasing affinity for biotin-FITC (B- J) . Surprisingly even at the highest concentration of streptavidin, little non-specific binding is seen, as determined by the inability of non-biotinylated cells incubated with 160μg/ml streptavidin to bind either biotin- FITC (I) or streptavidin-FITC (K) . Thus, cells can be treated in such a way that enables them to bind biotinylated proteins, though once the avidin is bound it is no longer detected after 24 hours in culture (DMEM+10% FCS, no biotin in DMEM) data not shown.

Cells labelled with avidin can bind biotinylated antibody The present applicants wished to determine whether cells once biotinylated and incubated with avidin can retain an anchor of active avidin in such a way that biotinylated antibody can then be bound to the cell surface via the avidin-biotin bridge. This was investigated using a biotinylated antibody (Mouse anti human HLA-DR) to bind to the avidin on the cell which we then detected with a FITC- conjugated F(ab')₂ fragment of rabbit anti mouse immunoglobulins. This is demonstrated in Figure 5 where as controls the present applicants show that Untreated (A) and biotinylated (B) cells bind little FITC-conjugated F(ab')₂ fragment of rabbit anti mouse immunoglobulins. Untreated but avidin incubated cells (C) bind little biotin-FITC compared with biotinylated cells which have been incubated with avidin (D) indicating both the attachment of avidin, via the biotin on the cell surface, and its ability to bind further biotin in the form of biotin-FITC. Whilst untreated avidin incubated cells (E) also bind little FITC- conjugated F(ab')₂ fragment of rabbit anti-mouse immunoglobulins, but biotinylated cells after incubation with avidin (F) were found to bind substantially more (4-5 fold increase in intensity) , indicating a non-specific binding of the antibody to avidin. However, untreated cells incubated with avidin followed by the biotinylated mouse anti human HLA-DR (G) bind the same amount of FITC- conjugated F(ab')₂ fragment of rabbit anti mouse immunoglobulins as (A) and (E) , whilst biotinylated/avidin incubated cells with biotinylated mouse anti human HLA-DR (H) bind 10-fold more FITC-conjugated F(ab')₂ fragment of rabbit anti mouse immunoglobulins than the same cells in the absence of biotinylated mouse anti-human HLA-DR (F) . Additional data not shown demonstrated that identical but non-biotinylated mouse anti human HLA-DR does not bind non- specifically to biotinylated/avidin treated cells, indicating that the binding of mouse anti-human HLA-DR to avidin coated cells was on the basis of the antibody-biotin conjugate. Thus biotinylated antibody can be bound to the cell surface by an avidin bridge, and detected on the basis of a species specific constant region indicating that steric hinderance as a result of tight binding to avidin does not totally abrogate further antibody binding.

Cells with bridged avidin can be used to assemble a complex with protein G' -biotin and antibody A more orientation specific antibody binding can be achieved with protein G' -biotin (Figure 6). Thus the antibody itself need not be biotin conjugated. Cells that were biotinylated and then cultured overnight were avidin treated as described and incubated with protein G' - biotin/mouse anti-human CD28 activating antibody (Baroja ML et al . Cell . Immunol . 1989; 120: 205-217) complexes. Figure 6 shows that after biotinylation, incubation with avidin allows biotin-FITC to detect avidin on the surface of only the biotinylated cells (B) and not untreated (non- biotinylated) cells (A) . The presence of avidin on the surface then serves as a target/anchor for protein G' - biotin/anti CD28 complexes, the presence of which are detected after incubation with rabbit anti-mouse immunoglobulins F (ab' ) ₂-FITC. Cells that were biotinylated, avidin incubated and then further incubated with the protein G' -biotin/CD28 complex (K) , show almost a 2 log shift increase in fluorescence over the staining with non- biotinylated cells (Fig 6 compare K with J) . These staining profiles indicate that the entire complex (biotin/avidin/protein G' -biotin/anti-CD28 can be assembled in a way predictable from the individual capabilities of the component parts. Non specific staining by rabbit anti- mouse immunoglobulins F(ab')₂-FITC of biotinylated cells treated with avidin and protein G' -biotin is also seen (see also Figure 5F) , but is less than 10% of that seen with additional CD28 (K) , data not shown.

Stability of the complex

When cells are returned to culture after avidin treatment, either at 37°C or 4°C, over a period of 3.5 and 6 hours the background staining of Untreated (non-biotinylated) cells treated with avidin and stained with biotin-FITC remains constant; compare Fig 6 (A, T=0) with C, 3.5 hours & F, 6 hours) , duplicate cells which were kept at 0°C were identical to those kept at 37°C and therefore not included.

However when biotinylated cells were avidin incubated and then immediately stained with biotin-FITC (B) the initial fluorescence decreased substantially when the cells were retested after 3.5 hours at 4°C (D) but little further reduction was seen at 6 hours (G) , this loss of detectable avidin was greater when cells were kept at 37°C with a continuous drop in detectable avidin over the 3.5 (E) and 6 hour (H) time points. Interestingly, cells stained at T=0 (B) and then returned to culture for 6 hours retained far more stain (I) . The temperature dependence of the change coupled with the retention of stain on biotin-FITC labelled cells in culture suggests that disappearance of the avidin is at least in part due to endocytosis of the cell membrane associated complex.

The entire avidin/protein G' -biotin/anti-CD28 complex on biotinylated cells (K) again shows deterioration at 4°C over the first 3.5 hours (M) with less change over the subsequent 2.5 hours (P) . Once again the background at 37°C (L, 3.5 hours and O, 6 hours) remains the same and identical to the background seen at 4°C (data not shown) . At 37°C (N, 3.5 hours & Q, 6 hours) the complex deteriorates over the entire time course, but remains above background even after 6 hours (comparing Q with O) : whilst stained cells (K) that were returned to culture for a further 6 hours (R) once again retained far more stain.

Mouse anti -human CD28 antibody complexed transiently wi th cells is functional as a costimulator in vi tro Having demonstrated co-localization of the mouse anti-human CD28 with target cells, the present applicants then tested the stimulatory ability of bound protein G' -biotin/antibody in a functional assay for CD28 mediated stimulation of T- cells. TIB232 cells that were complexed as described in the legend to Figure 7, with appropriate controls. Once assembled these cells were co-cultured with human Jurkat T- cells. These cells possess both the CD28 receptor and the CD3 complex and respond to stimulation via CD3 with PHA, or protein kinase C activation with PMA, by secreting small amounts of IL-2 into the medium (Nunes JA et al . J. Exp . Med. 1994; 180: 1067-1078) . Combinations of anti-CD28 with PHA or PMA synergistically increase the level of IL-2 secretion. Figure 7 shows that Jurkat cells in the absence of stimulation secrete little detectable IL-2 either in the absence of TIB232 cells or with any of the modifications except Biotin+AVidin+Protein G' -biotin+anti-CD28 (B/AV/PG/CD28) . Jurkat cells incubated with PMA alone show little IL-2 secretion, and other conditions show no significant change until, once again, the PMA treated cells are incubated with B/AV/PG/CD28. Under these conditions the stimuli appear synergistic. A surprising result is seen when the cells are stimulated with PHA here we see that both B/AV/PG/CD28 and C/AV/PG/CD28

(Control+AVidin+Protein G' -biotin+anti-CD28) both synergise with PHA treatment of Jurkat cells. This is a reproducible phenomenon, and odd since neither avidin or protein G' - biotin bind non-specifically to a sufficient degree to anchor anti-CD28 to these cells as determined by FACS analysis. The present applicants can only conclude that firstly the threshold of synergistic interaction with CD28 and PMA is higher (ie less sensitive) , than for PHA, and that the bioassay of the Jurkat cells detects lower levels of presented anti-CD28 than the FACS analysis. Finally the addition of anti-CD28 alone has little effect on the Jurkat cells, and the addition of more, with B/AV/PG/CD28 , is not effective. Incubation of cells with anti-CD28 alone (Control/anti-CD28, C/CD28) therefore does not allow this antibody to persist through the washing steps prior to the assay, neither does the presence of avidin and protein G' - biotin, (B/AV/PG) , without the anti-CD28 stimulate IL-2 production in these cells. Therefore these results strengthen the suggestion that the stimulation of IL-2 production by the T-cell line is the product of its CD28 activation by the anti-CD28 antibody bound to the TIB232 cells .

Figure 8 demonstrates that the presence of antibody per se on the TIB232 cells does not allow them to activate Jurkat cells. Thus whilst anti CD28 (B/AV/PG/CD28) co- stimulates with either PMA or PHA, anti CD3 (B/AV/PG/CD3) co-stimulates only with PMA, and an isotype matched but irrelevant antibody (CD14) that does not bind to Jurkat cells (unpublished results) (B/AV/PG/CD14) is unable to co- stimulate under any of the conditions tested. TIB232 cells that were biotinylated but not avidin treated failed to bind Protein G' -biotin and anti-CD28 (B+PG/CD28) and thus showed little co-stimulatory activity. Non specific T cell activation in vitro using combinations of antibodies

The murine mastocytoma cell line P815 forms tumours in syngeneic DBA/2 mice. It had been shown by other groups that these cells responded well to conventional gene therapy, ie cells transfected with murine B7.1 do not form tumours in these animals.

In this and subsequent experiments, except where otherwise indicated, P815 cells were biotinylated by the method described herein except that the biotinylation was conducted at room temperature. The 30 minute incubation on ice during the biotinylation was found to be toxic to the cells over a period of 24 hours. A subsequently discovered previous report indicated that cold shock induced DNA fragmentation (now called apoptosis) in P815 cells.

The use of room temperature worked so well that biotinylation was much better than before. However because of this increased efficiency it was found that adding avidin was a problem: there was so much biotin that the tetrameric avidin tended to be inactivated (ie all the biotin binding sites were used up) ; the avidin would then bind little of the biotinylated antibody. Thus it is preferable to wait more than 40 hours before attempting to add avidin, by which time the biotin on the cells has dropped to manageable levels. It will of course be possible to reduce the amount of the biotinylation reagent in order to shorten this time. Allowing this growth time may also allow the cells to re-express potentially immunologically important membrane proteins in an unmodified form.

T cell activation with combinations of antibodies was first analyzed using in vitro DBA/2 splenocyte activation assays. As noted above, it is well established that for full T cell activation the engagement of the T-cell receptor with the MHC/antigen complex requires a second signal (so called costimulation) , and that a very potent costimulator in this context is B7.1. Since the anti-CD28 antibody alone (clone 37.51, which mimics the action of B7.1 binding to its T cell counter receptor CD28) , did not appear to be effective in vivo, it was investigated whether combining it with a hamster antibody directed against the epsilon chain of CD3 (CD3e, T-cell receptor complex) would generate an effective T-cell activation in vitro. In addition it was investigated whether the murine system behaved like the human system in that an antibody directed against the CD5 on T cells (counter receptor for CD72 found on B-cells) can synergize with either CD3e or CD28 pathway of stimulation, and in that a triple (CD5, CD3e, CD28) combination activates T-cells more efficiently than either two alone .

Figure 9 illustrates the optimization of non specific T cell activation and IL-2 secretion in vitro. Splenocytes from naive DBA/2 female mice were plated at 5xl0⁴/well in round bottom 250μl microtire plates. Hamster antibodies directed against the murine CD28 and CD3e and Rat antibodies against murine CD5 were added to a final concentration of 0. lμg/ml each either alone or in combination. In addition, half the wells also included 5xl0⁴ P815 cells. The plate was incubated at 37°C for 48hrs, the supernatant harvested and assayed for the presence of murine IL-2 by ELISA. Results are expressed as pg/ml IL- 2/5X10⁴ splenocytes/48hrs, each point is the mean of three values. The value <15pg represents the limit of reliable detection from the IL-2 standard dose curve.

Figure 9 shows that a similar, though not identical, situation to humans is also the case in the murine system. Choosing a low dose of antibody in solution, none of the antibodies alone or in any combination induce IL-2 secretion from naive DBA/2 splenocytes. However when the splenocyte/antibody combinations are co-cultured with P815 tumour cells, we see a profound induction of IL-2 secretion for some combinations and very little effect in others. In the presence of P815, the splenocytes secrete no IL-2 in the presence of anti CD5 or CD28 alone or in a combination of anti CD5 and CD28. Antibody to the CD3e alone has a marginal effect on IL-2 secretion which is not changed in combination with CD5 , but is increased to 235pg/ml in combination with CD28. The most potent condition appears to be the triple combination of antibody to CD5,CD3e and CD28. We thus decided to use this combination in further experiments. The present applicants were encouraged by the increased IL-2 production seen in combination with P815 cells, which may be due to other immune costimulators on the surface of the tumour cells (ie ICAM-1 or others at present unknown) . It is believed that the IL-2 secretion is not at this stage tumour specific, but a non-specific reaction to the antibody combination.

In many ways the experiment detailed in Figure 9 mimics the effects that we would like to achieve in vivo. In vitro the tumour P815 cells and the antibody are colocalized and presented to the splenocytes merely because of the constraints imposed upon it by the incubation vessel. The cells and antibody are thus not at liberty to move away from each other as they undoubtedly would do in vivo. This is the problem that the antibody linking procedure (eg. with biotin and avidin) was designed to solve .

Comparison of two antibody completing procedures for the induction of IL-2 secretion in vi tro.

Figure 10 shows the results of experiments where different procedures are used to link the triple combination of antibodies to P815 cells and present them to splenocytes. In this experiment cells were biotinylated, avidin bridged, and then either incubated with a cocktail of biotinylated antibodies directed against CD3e, CD5 and CD28, or incubated with the equivalent amount of the identical non-biotinylated antibodies which had been preincubated with biotinylated recombinant Protein G' . Thus antibody was complexed to the P815 cells either directly using biotinylated antibody, or indirectly and in an orientation specific manner via Fc specific binding to biotinylated protein G' . Details of the experiments are as follows.

Splenocytes from naive DBA/2 female mice were plated at 5xl0⁴/well in round bottom 250μl microtire plates. Antibodies (either b otinylated or otherwise) directed against the murine CD28, CD3e and CD5 were added to a final concentration of 0.1 or O.Olμg/ml each in the presence or absence of 5xl0⁴ non-modified P815 cells. Additionally P815 cells were biotinylated and avidin bridged and further incubated with:

1) P815 biotinylated and avidin bridged (BAV) 2) Biotinylated anti CD3+CD5+CD28 (BAV-Bab's)

3) Biotinylated protein G' (BAV-PG)

4) Biotinylated protein G' and CD3+CD5+CD28 (BAV- PGab's) and 1x10^s of these cells cocultured with splenocytes. The plate was incubated at 37°C for 48hrs, the supernatant harvested and assayed for the presence of murine IL-2 by ELISA. Results are expressed as pg/ml IL-2/5X10⁴ splenocytes/48hrs, each point is the mean of three values. The value <15pg represents the limit of reliable detection from the IL-2 standard dose curve.

It will be observed from Figure 10 that BAV cells or BAV-PG cells in combination with splenocytes results in little or no IL-2 secretion, but BAV-Bab's and BAV-PGab's results in the secretion of 147 and 864 pg/ml IL-2/5xl0⁴/48 hours respectively. It appears that both complexing procedures result in antibody associated with the P815 cells that can activate splenocytes to secrete IL-2. Some but not all of the increased potency with the Protein G' methodology may be due to the higher potency of non biotinylated antibodies. This can be seen in the standard incubation with non modified antibody (ab's) or biotinylated antibody (Bab's) with non modified P815 coculture .

Thus we can stimulate IL-2 secretion with antibody complexed cells in vitro and protein-G methodology seems to be the more effective. However repeating this assay with antibody complexed cells that were then incubated at 37°C for 24 hours, washed and plated with splenocytes showed that this IL-2 secretion induction was no longer present under any of the complexing procedures . The antibody has either fallen off or been endocytosed.

An investigation of the relative stability of the two procedures is shown in Figure 11. Figure 11 shows a comparison of the stability of two antibody complexing procedures. Splenocytes from naive DBA/2 female mice were plated at 5xl0⁴/well in round bottom 250μl microtire plates. P815 cells were biotinylated and avidin bridged and further incubated with:

1)P815 Biotinylated and avidin bridged (BAV) 2) Biotinylated anti CD3+CD5+CD28 (BAV-Bab's) 3) Biotinylated protein G' (BAV-PG)

4) Biotinylated protein G'and CD3+CD5+CD28 (BAV-PGab's) and 5xl0⁴ of these cells cocultured with splenocytes either immediately after complexing, or after 90 minutes in RPMI+10%FCS at 37°C followed by a washing step. The plate was incubated at 37°C for 48hrs, the supernatant harvested and assayed for the presence of murine IL-2 by ELISA. Results are expressed as pg/ml IL- 2/5X10⁴ splenocytes/48hrs, each point is the mean of three values. The value <60pg represents the limit of reliable detection from the IL-2 standard dose curve.

At T=0 once again the protein G' methodology is the most potent for IL-2 secretion (735pg/ml compared to 210pg/ml) . However after 90 minutes (T=90) the protein G' procedure has dropped in potency by 36% to only 64% of the original T=0 value, this is compared to a maintenance of more than 90% in the biotinylated antibody procedure. Lower potency applied over a longer period may well be the preferred option for therapy.

Having tested the antibody combination and complexing procedure in vitro the combinations were tested in vivo.

Modification of P815 Tumour Cells and their Assessment in Vivo Modification of P815 cells .

4xl0⁷ P815 cells in RPMI+10%FCS were pelleted (room temperature, 5 minutes, 200g) and resuspended in 1.6ml Ca/Mg free PBS pH=8.0. This solution was divided into 2x0.8ml in fresh polypropylene tubes and 6μl of 366mM Biotinamidocaproate N-hydroxysuccinimide ester (BSuE) was added to the dry portion of one tube (B) . The tube was flicked to rapidly dilute the BSuE, and the tube incubated at room temperature for 30 minutes. After incubation, cells (with biotinylation [Biotin] and without biotinylation [Control] ) were diluted into 20 mis of RPMI+10%FCS and allowed to stand for 10 minutes at room temperature. Cells were then pelleted (RT, 10 minutes, 200g) , resuspended to 5xl0⁵/ml (viability 94%) and placed at 37°C. Cells were maintained below lxl0⁶/ml for 48 hours.

After 48 hours cells were counted and 2.5xl0⁷ cells pelleted (RT, 10 minutes, 200g) and resuspended in 0.75ml of HBSS+ 1%FCS.

To tube [Biotin] was added 200μl of lOmg/ml Avidin in PBS pH=8.0 (Now labelled BAV) This results in a final concentration of 2.1 mg/ml . To tube [Control] was added 200μl of PBS pH=8.0 alone (still labelled C)

After 30 minutes incubation at room temperature cells were made up to 10 ml with HBSS+1%FCS, pelleted (RT, 10 minutes, 200g) resuspended in 5ml HBSS+1%FCS, pelleted again as above, and finally resuspended in 0.5ml

HBSS+1%FCS.

At this point BAV was divided into 2 portions of 0.25ml each (1.25xl0⁷ cells each). One portion was then added to 0.75ml of biotinylated antibody solution comprising:

125μg Rat anti Mouse CD5(ly-l) , IgG2a, Pharmingen clone 53.7-3; and

125μg Armenian Hamster anti Mouse CD3e, IgG, Pharmingen clone 145-2C11; and 125μg Syrian Hamster anti Mouse CD28 , IgG, Pharmingen clone 37.51.

These antibodies were stored in 0.1% Azide, so the final concentration of azide in this portion [BAV Bab's] was 0.075%. The other portion was adjusted to contain the equivalent amount of sodium azide [BAV Az] and have a volume of lmL.

The [control], [BAV Az] and [BAV Bab's] tubes were then incubated at room temperature for a further 30 minutes to allow for avidin/biotin interaction and binding. After this the cells were resuspended in 10ml HBSS+1%FCS, pelleted (RT, 5min, 200g) , resuspended in 5 ml HBSS+1%FCS, pelleted (RT, 5min, 200g) , resuspended in 5ml HBSS with NO FCS, counted, and adjusted to the required concentration. At this point there were 3 cell populations:

P815 Control - mock treated all the way through including the incubations and the washes, but with no modification. P815 BAV Az - Biotinylated cells, subsequently incubated with avidin, and then incubated in the presence of 0.075% Azide.

P815 BAV Bab's - Biotinylated cells, subsequently incubated with avidin and then incubated with a mixture of Biotinylated anti CD3 , CD28 and CD5 in the presence of 0.075% Azide.

In addition there was the remains of the original stock of biotinylated cells Biotin. These four cell populations were then used for three assays done in the following order: 1) Immunofluorecence and FACS analysis (Figure 12)

2) Injection into test female DBA/2 mice (Figures 13 and 14)

3) In vitro determination of the ability to activate T- cells (Figure 15)

1 - Jmmuπojfluorescence and FACS analysis .

In order to test that the full modifications have proceeded as normal before injection into the animal was performed, the cells were analyzed by FACS analysis.

5xl0⁵ cells were washed in HBSS + 1%FCS and the cells resuspended in either: lOOμl of HBSS+1%FCS (Unstained) ; or lOOμl of HBSS+1%FCS containing lμg Avidin-FITC (AV-FITC) ; or lOOμl of HBSS+1%FCS containing 0.5μg/ml Biotin-FITC (B- FITC) ; or lOOμl of HBSS+1%FCS containing a mixture of lOμg Polyclonal Rabbit anti Rat F(ab')2 FITC (Serotec ST-AR 49) and 2μg Polyclonal Goat anti Hamster IgG FITC (Serotec AA/17F) .

The anti Rat FITC detects antibody to CD5 and the anti Hamster FITC detects antibody to CD3e and CD28.

After 30 minutes at room temperature, cells were washed twice in HBSS+1%FCS and analyzed by FACS. FACS profiles (Figure 12) show an analysis of green fluorescence (FL1) on a logarithmic X axis, with the y axis as the event counter indicating the number of cells with each level of fluorescence .

Figure 12 illustrates an assessment of antibody complexing procedure by FACS. After complexing, (T=0, A-H) Unstained, biotinylated (A) and control unmodified (B) cells were analysed, followed by control unmodified (C) and biotinylated cells (D) incubated with avidin-FITC to analyse biotinylation. Control cells and biotinylated and avidin bridged (BAV Az) cells were also incubated with biotin-FITC (E and F) respectively, detecting avidin bound to biotinylated cells. Finally, BAV Az (G) and BAV Bab's

(H) incubated with Rabbit anti Rat F(ab')2 FITC and Goat anti Hamster IgG FITC determine the efficiency of biotin antibody complexing with BAV cells. The process described for (A-H) was repeated with samples incubated overnight in RPMI+10%FCS at 37°C, washed and stained (I-M) . (I) - Control unmodified, unstained. (J) - Control unmodified stained with biotin-FITC. (K) - BAV Az stained with biotin-FITC. (L) - BAV Az stained with Rabbit anti Rat FITC and Goat anti Hamster FITC. (M) - BAV Bab's stained with Rabbit anti Rat FITC and Goat anti Hamster FITC.

Thus Figure 12 shows that the presence of biotin alone on cells does not increase their fluorescence (compare A and B) , and the presence of biotin is confirmed in (D) compared with control cells in (C) . This biotin can be used to assemble the first part of the avidin bridge as seen in (F) compared with control cells in (E) . Finally the avidin detected in (F) can be used to complex with CD5 , CD28 and CD3e detected in (H) compared with (G) . After 18 hours in culture, the avidin detectable on BAV Az (F) has mostly disappeared when assayed in (K) , and is little different from background (J) . P815 BAV Bab's still retain a small amount of detectable antibody, compare (M) with (L) , though this is much reduced from that staining seen at T=0 (H) .

Thus most of the antibody complexed with the avidin bridged cells disappears in 18 hours.

2) - Injection into test female DBA/ 2 mice. During the immunofluorescence and FACS analysis of stained samples, the remaining cells for injection were kept at room temperature in HBSS alone at lxl0⁶/ml.

The control cells were divided into two and to one tube the antibody mixture was added to a final concentration of 10μg of each antibody for every 2xl0⁶ cells. This represents the amount of antibody bound to the BAV Bab's cells if 50% of the available had bound. Thus there were now 4 samples. P815 control (8 animals) control for washing procedure P815 Bab's (8 animals) control for antibodies alone but in solution

P815 BAV Az (7 animals) control for modification, azide, but no antibody

P815 BAV Bab's (10 animals) experimental group, with antibody complexed with the cells.

Each 6-8 week old DBA/2 female was injected I. P. with 200μl of cells (lxl0⁶/ml in HBSS alone, 2_*10^sr/animal) ■ Animals were then monitored 2-3 times a week and scored on the presence of either tumour lumps or ascites; they were culled on the basis of large ascites or obvious ill health.

The results of this experiment are detailed in Figure 13 showing % survival with time and Figure 14 showing % disease free with time.

Figure 13 shows that animals injected with P815 control cells were all dead by day 54, and that those animals injected with P815 Bab's (ie control cells with antibodies in solution) were all dead by day 43. This indicates that the antibodies were not protecting the animals from the challenge. In contrast 80% of the animals injected with P815 BAV Bab's were still alive on day 54, compared to those animals injected with modified cells alone (P815 BAV Az) , of whom only 30% were still alive at that time. The survival of animals with antibodies complexed to their cell surface is greater than any of the controls, despite the fact that most of this antibody (detected by FACS) had gone from these cells.

Figure 14 illustrates the disease status of these same animals prior to their demise. Animals were examined every 2-3 days and their abdomens palpated for signs of tumour lumps or ascites. All animals developed signs of disease by day 9 (% Disease free =0) , in virtually all cases characterized by a small lump 2 -3mm diameter on the abdomen. This did not seem to be associated with the lymph nodes, but may have been the result of inflammation. However, many of the lumps remained whilst subsequent ascites developed, which at later time points became the most serious indicator of disease. Further investigation in subsequent experiments will need to determine the nature of the lumps. Disease in the majority of these animals is progressive from day 9 onwards. In the P815 BAV Bab's group, however, 50% of animals show no sign of disease, or disease that shows only transient manifestation (see day 23, 28 and 49), from day 16 onwards.

3) In vi tro determination of the ability of modified P815 tumour- cells to activate T-cells .

After the injection of the animals, the remaining modified cells were used in an in vitro T cell activation assay. A spleen was removed from a DBA/2 animal of the same batch as the in vivo experiment. The spleen was disrupted and the white cells counted; no attempt was made to purify T cells.

Figure 15 illustrates a determination of the ability of modified P815 tumour cells to activate T-cells in vitro. Splenocytes (5xl0⁴) were plated in round bottom microtire plates to which was added either antibody in solution, control P815 or modified cells, and the plates incubated at 37°C, additionally 4 and 24 hours later additional wells with splenocytes were then incubated with antibody or modified cells that had been returned to culture in RPMI+10%FCS at 37°C. And the plates returned to 37°C. The supernatant from these plates was harvested 72 hours after the initial culture and assayed by ELISA for the presence of murine IL-2 as a marker for non-specific activation of T cells. Figure 15 shows that cell samples taken from those prepared for injection were capable of inducing IL-2 secretion in vitro. This activity remains for at least 4 hours at 37°C but is gone by T=18. The amount of IL-2 secretion also appears dependent on the number of P815 BAV Bab's cells plated. Thus it can be confirmed that the cells injected were capable of inducing IL-2 in vitro. It has also been determined that the P815 BAV Bab's cells themselves secreted no IL-2 in the same experiment (data not shown) .

Biotinylation of cell membranes is not limi ted to murine cell lines

In order for this to be in any way applicable to human treatment, primary human cells must be capable of responding to biotinylation in the same way as the cell lines so far tested. Figure 16 demonstrates that two primary AML patient samples (KR and TC) can both be kept in culture and biotinylated. In fact all cells tested so far are easily biotinylated, including murine myeloid 32Dp210 cells as well as human myeloid HL-60, U937, NB4 and K562 cells .

DISCUSSION

The present invention offers tremendous potential for introducing co-stimulators onto the surface of target cells making them able to provide stimulation of immunological cells such as T-cells. The strategy allows the use of readily available reagents and thus, in theory, no costly or complicated reagents need to be manufactured for this strategy to be used in therapy. The orientation of the bound antibody is such that its antigen binding variable region is available for further interactions. The presented data shows that cells can be readily biotinylated to a high level under conditions that are not toxic to the cells. The biotin on the cell surface is progressively lost, but this can be slowed down with the use of metabolic inhibitors. Since the initial efficiency of biotinylation is high, the rate of reduction even in the absence of metabolic inhibitors can be tolerated as even after 72 hours in culture the presence of biotin on the cell surface can be readily detected.

The present applicants have demonstrated that potentially important surface proteins are targets for biotinylation, but not universally, so whilst the immunological profile of B7.1 on cells is altered this is not the case with CD54 (ICAM-1) or the MHC Class 1 on TIB232 cells. Additionally, after 24 hours in culture the profile of detectable B7.1 returns to near normal levels. Putative tumour antigens may be targets for biotinylation and this may then destroy their immunogenicity, but it would be unlikely that biotinylation would interfere with all of them, and that fresh product would not be re- expressed with additional culture. This has been shown to be the case with a surface marker found to be interfered with in these studies (ie B7.1). Once the cells have been labelled with biotin, either avidin or streptavidin can be used to provide a bridge to other reagents. One initial worry was that the stoichiometry of the avidin treatment, despite the tetrameric nature of the protein, would result in the inactivation of further biotin binding activity of the bridge. The present applicants have found that as long as the streptavidin /avidin concentration is sufficiently high this is not a problem.

Biotinylated antibody binds with avidin treated biotinylated cells in a way that was predictable from the nature of biotinylated antibody. The conjugation reaction to biotinylate antibody was performed using its succinimide ester derivative, and does not target constant regions, but reacts primarily with lysine NH₂ side groups (Demant EJF et al. Biochim . Biophys . Acta 1991; 1118: 83-90). Since biotin labelling procedures for antibody aim to avoid interfering with antigen recognition (Johnstone A et al Blackwell Science, Oxford, 1996, p277-289) some of the bound antibody should be correctly orientated. An alternative strategy was also utilized where the orientation of the bound antibody was predictable. This alternative strategy provides added flexibility to the potential user.

Since they were not actually cross-linking in this study the present applicants were somewhat surprised that the avidin protein G' -biotin complex once assembled on the cell surface disappears quite rapidly. At least one major component of this deterioration involved the internalization of the complex into the cell, presumably by endocytosis. This process appears to be stimulated by the addition of more components to the cell membrane, given that the decrease in biotin level on the cells takes place over a period of days, whilst avidin complexed to the cells deteriorates in hours. Even with the rapid deterioration however, a detectable amount remains after six hours in culture, and sufficient remains attached, even after extensive washing of the cells, for anti-CD28 to be used in the Jurkat co-stimulation assay.

Other labelling strategies have been utilized to simplify the system as much as possible and reduce the size of the complex on the cell surface. One of the first, that proved ultimately fruitless, was to avoid the biotinylation completely and avidinylate the cells directly with the commercially available maleimide and hydrazide derivatives of avidin. The problem with these was that the pre- treatment required to optimize the presence of the correct targets (SH groups for the maleimide (Christiaansen JE et al . J. Immunol . Meth . 1984; 74: 229-239) and modified cell surface glycoprotein for the hydrazide (Spiegel S et al . J^". Immunol . 1981; 127: 572-575 , Kahne T et al . J. Immunol . Meth . 1994; 168: 209-218.)), were quite cytotoxic and the labelling not very efficient. The high molecular weight of avidin compared with biotin also precluded the application of sufficiently high molar concentrations of derivatised avidin. Other reagents such as the N- (biotinoyl) dipalmitoyl-L-Q!-phosphatidylethanolamine (biotin-DPPE Muzykantov VR et al . J. Immunol . Meth. 1993; 158: 183-190.) conjugate was capable of only minimal association with the membrane and was very unstable. Synthesis of biotinylated cationic lipidophiles such as 1, 1' -didodecyl- , 3 , 3 ' , 3 ' - tetramethylindocarbocyanine perchlorate (Dil) of the indocarbocyanine family, which are often used as membrane probes (De Clerk LS et al . J. Immunol . Meth . 1994; 172: 115- 124) and may ultimately be the reagent of choice, are apparently difficult to synthesize. Any antibody binding domain (either biotinylated or not) and any protein (of those that can be biotinylated) can be anchored to the cell surface and this will allow the rapid evaluation of potential immune co-stimulators or combinations thereof either in vitro (eg. through splenocyte IL-2 assays) or within _.in vivo mouse models (eg. by injecting complexes of tumour cells with the potential immune co-stimulator) . Thus cells lacking adhesion molecules can be labelled transiently with the adhesion molecule to allow the costimulation. A slow release capability for cytokine may be feasible by the careful choice of linker groups in the complex introducing protease sensitive sites. By way of example, questions such as the relative efficacy of an anti-CD28 antibody and the natural counter receptor for CD28 (B7.1 which can provide an inhibitory signal to T-cells expressing CTLA4 ( Walunas et al . Immuni ty 1994; 1: 405-413, Krummel MF et al . J. Exp . Med. 1995; 182: 459-465.)), can be rapidly anserwed; combinations such as anti-CD28 with either anti-CD3 or anti-CD2 (Li YW et al . J. Exp. Med. 1996; 183: 639-644) can be rapidly evaluated. The present invention thus provides novel methods of in vitro or in vivo screening for, or evaluation of, potential immune costimulators or combinations thereof, e.g. molecules capable of activating T cells.

The present applicants have realised that the ability to modify large numbers of tumour cells ex-vivo in a very short time is more advantageous than more difficult nucleic acid based procedures .

The strategy may ultimately be further simplified using a chimeric streptavidin-protein G module or chimeric streptavidin-protein A molecule (Sano T et al . Biotech . N. Y. 1991; 9: 1378-1381). This could allow the whole process to be simplified to a two step reaction: 1) biotinylation followed by: 2) application of a commercially available reagent comprising the antibody (or antibody- avidin conjugate) of choice complexed with streptavidin/avidin-protein A/G at the optimal molar ratio. Finally it has been shown that some recombinant single chain Fv species whilst retaining affinity have reduced avidity for antigen due to their monomeric structure. This can be reversed in some cases by a single chain Fv- core/streptavidin fusion protein (Kipriyanov SM et al . Protein Eng. 1996; 9: 203-211) . The, streptavidin mediated, tetrameric structure of the molecule can enhance the avidity of the Fv. Recombinant single chain Fv's of anti CD28 (9.3) may not initially look like a good choice as a Fab fragment of anti CD28 (9.3) does not demonstrate co- stimulatory activity whilst a F(ab')₂ fragment of the same antibody is as active as the whole molecule (Dalme NK et al. J. Immunol . 1988; 140: 1753-1761) . This may not be due to lack of avidity as this Fab fragment can compete quite effectively with native B7.1 in blockade of CD28 pathway dependent proliferation of ICAM-1 primed T cells

(Dalme NK et al . J. Immunol . 1994; 152: 2686-2697). It is of interest to determine whether Fv (9.3) fusions with streptavidin would reactivate costimulatory activity. Such a fusion protein would, if active, combine the ability to express large amounts of a recombinant Fv that retains avidity and fortuitously possesses the ability to bind biotin. Such a reagent, combining many advantages in one molecule, would provide further possibilities of in vivo non-gene, immunotherapy. Injection into a tumour site with biotin-succinimide ester followed by the application of recombinant single chain Fv anti CD28 fused to streptavidin may indeed be a possibility. Using a combination of three antibodies complexed with tumour cells the present applicants demonstrated T-cell activation in vitro as detected by IL-2 secretion and a growth inhibitory effect of these antibodies on tumour cells in vivo. The applicants then demonstrated that the complexing of these antibodies onto tumour cells also has an effect when such cells are introduced into an in vivo mouse model. The present applicants are currently generating data which shows that mice which have rejected tumour cells modified by this procedure are immune to a subsequent challenge with unmodified cells (at present 6 out of 7 mice are tumour free, compared to zero out of 5 age matched naive controls) . This data indicates a strong suggestion of vaccination potential .

The instability of the complex on the modified tumour cells (see Figure 15) suggests that a transient presence of the activating antibodies co-localised with the tumour cells is sufficient to induce the rejection of these tumour cells. This is consistent with the observed initial development of disease (Figure 14) followed by its subsequent disappearance. This also implies that the modified tumour cells can be used as a tumour cell vaccine to induce the remission of an already established disease.

In the in vivo situation, the initial effect may well be non specific, resulting in essentially mitogenic activation of T cells. However, it may result in a cytokine rich environment that provides a good platform for the rare tumour cell specific T-cells to be activated, proliferate and become efficient effector cells. Any non- tumour specific T cells activated by interaction with the complexed antibodies (or other complexed costimulators) are unlikely to proliferate since the local environment will not contain cells presenting the antigens for which they are specific.

Evidence of auto-immune disease in the animals inoculated with the modified tumour cells was not seen and so there is no evidence of induction of auto-immune disease. This again suggests that the initial antibody mediated T cell activation achieved by inoculation of the mice with the modified cells results in anti-tumour activity but not wide-spread and non-specific anti-host responses. The present applicants have not yet checked whether the animals which rejected their tumours have developed a specific immunity only to the tumour cells used in the study (P815) , but they expect this to be the case.

Further potentially therapeutic antibodies and combinations thereof have not yet been tried, though the methodology described herein can clearly be used to screen for such antibodies or combinations thereof. The treatment of animals with established disease using this method is also to be attempted. The complexes and methods of the present invention have considerable potential therapeutic utility. Firstly, tumour cells could be extracted from a patient and complexed ex vivo to an immune costimulator or a combination of such costimulators. The immune costimulators may have been identified by a screening method provided by the present disclosure; they may for example be cell surface receptors, cytokines, antibodies, or domains or fragments of antibodies. The costimulators could be complexed to the tumour cells using the methods exemplified herein, for example using the biotin-avidin system with or without the use of protein G or protein A. Alternative known methods of directly linking the costimulators to the tumour cell surface without specialist knowledge of the tumour cell surface could also be used.

The complexed cells would then be reinjected into the patient. This method would be particularly suitable for the treatment of leukaemias and lymphomas where there are sufficiently large numbers of easily accessible tumour cells in the blood. By way of example, one application may be the treatment of Minimal Residual Disease since it is possible to freeze Acute Myeloid Leukaemia (AML) cells prior to biotinylation and use. The method could also be used for the treatment of solid tumours if sufficient numbers of tumour cells could be obtained, e.g. by biopsy, surgery or aspiration techniques. It is further possible that the complexing of the costimulators with tumour cells could take place in vivo, thereby facilitating the treatment of solid tumours. By way of example, biotin-succinimide ester could be injected into a tumour site and followed by injection of the immune costimulator (s) .

MATERIALS AND METHODS

The following sections provide further information on the materials and methods used in the preceding experiments . Monoclonal antibodies

Mouse antibody 9.3 (Baroja ML et al . Cell . Immunol . 1989; 120: 205-217). (IgG2a) recognizing human CD28 (lmg/ml in PBS) , was a generous gift from Dr Karl Hellstrδm at Bristol-Myers Squibb, Seattle. Mouse antibody (IgG2a, clone HIT3a) recognizing human CD3 , 0.5mg/ml in 150mM NaCl (Pharmingen 30111A) . Mouse antibody (IgG2a, clone M5E2) recognizing human CD14 , 0.5mg/ml in 150mM NaCl (Pharmingen 30541A) . Biotin conjugated mouse (IgG2a) anti human HLA-DR, and identical non-biotinylated antibody, both 25μg/ml (Becton Dickinson 347361 and 347360 respectively)

FITC-conjugated hamster anti-mouse CD80 (B7-1) , IgG 0.5mg/ml used at 1:20 dilution (Pharmingen 09604D) . Rat (IgG2a) anti-mouse CD54 (ICAM-1) , lmg/ml used at 1:10 dilution (Serotec MCA818)

Mouse (IgG2) anti mouse MHC H-2 class 1 (H-2^b'^k'^s'^p, lmg/ml used at 1:10 dilution (Biodesign M4918M)

Cells

Mouse AKR1.G.1.OUAR.1.26, ( Hyman R et al . Immunogenetics 1980; 10: 261-271) AKR strain specific thymoma cell line was obtained from the ATCC Tumour Immunology Bank, hereafter referred to as TIB232 and grown routinely in RPMI+10%FCS, 2mM L-glutamine, lOOμg/ml streptomycin, lOOU/ml penicillin between 5xl0⁴/ml and lxl0⁶ml. Human Jurkat JH6.2 ( Nunes JA et al . J. Exp . Med. 1994; 180: 1067-1078) cells were a kind gift from Dr D Cantrell, Lymphocyte Activation Laboratory, Imperial Cancer Research Fund, London. They were recloned and colonies screened by FACS analysis for CD3 expression. Clone 5 expressed the highest levels of CD3 and was used subsequently in the study. Cells were grown in RPMI+10%FCS, 2mM L-glutamine, lOOμg/ml streptomycin, lOOU/ml penicillin, regularly re-checked for CD3 expression and discarded after more than 1 month in culture.

Primary human acute myeloid leukaemia (AML) cells were collected by leucapheresis, and mononuclear cells obtained by density gradient separation (Ficoll-Hypaque, Pharmacia Biotech Ltd) and cryopreserved in medium with 30% autologous serum and 10% DMSO. Cells for culture were washed in medium after thawing and resuspended to between 5xl0⁶ and lxl0 /ml in Iscove's Modified Eagle's Medium

(IMDM) with lOOμg/ml streptomycin, lOOU/ml penicillin, 10%

FCS (all from Sigma) , lOng/ml recombinant human GM-CSF

(kind gift from Schering-Plough Ltd, Bury St. Edmunds,

UK) , 20ng/ml recombinant human Stem Cell Factor (SCF) (225- SC, R&D systems, Oxford) and lOng/ml recombinant human IL-3 (kind gift from Sandoz Pharmaceuticals, Camberley, UK) . Cells were pelleted each day and resuspended to the above cell concentration. Under these conditions the two cells types KR and TC maintained 95% viability over 10 days (data not shown) . .Reagents

Biotinamidocaproate N-hydroxysuccinimide ester (Sigma B2643) reconstituted to 166.6mg/ml (366mM) in DMSO and stored at -20°C, can be frozen and thawed numerous times whilst still retaining activity. Other reagents were prepared as stock solutions, aliquoted and stored at -20°C Streptavidin (Sigma S8276) (lmg/ml in sterile water) Avidin (Sigma A9390) (5mg/ml in sterile PBS pH8.0) Streptavidin-FITC (Sigma S3762) (0.5mg/ml in sterile PBS pH8.0), used at 1:100 dilution: Biotin-FITC (Sigma B8889) (lmg/ml in PBS pH8.0), used at 1:100 dilution: recombinant Protein G' -biotin (Sigma P8045) (lmg/ml in PBS pH7.0): recombinant protein G' (Sigma P4689) (lmg/ml in PBS pH7.0) : Phorbol Myristate Acetate (PMA) (Sigma P8139) (lmg/ml in DMSO) : PHA (Phytohemagglutinin P, Pharmacia Biotech 27- 3707) (lOmg/ml in water) . Biotinylation and avidin bridging

For biotinylation: cells in exponential phase were washed once with HBSS+1%FCS at room temperature, and once with PBS pH8.0 in the absence of serum, and resuspended in 400-800μl of ice-cold PBS pH8.0 to the indicated cell concentration (see Figure Legends) , transferred to a fresh polypropylene tube and placed on ice. To a dry portion of the side of the tube 3-6μl of 366mM Biotinamidocaproate N- hydroxysuccinimide ester in DMSO was added and the tube flicked to dilute quickly (reversing the order of the addition resulted in crystallization of the reagent), the cells were then immediately returned to ice and incubated for 30 minutes with periodic gentle agitation. Cells were then washed twice in RPMI+10%FCS and returned to culture in RPMI+10%FCS overnight at 37°C, or immediately assessed for biotinylation efficiency as described under immunofluoresence .

For streptavidin or avidin treatment: biotinylated cells were cultured overnight counted, and l-8xl0⁷ cells were washed twice with HBSS+1%FCS and resuspended in HBSS+1%FCS in the presence or absence of avidin (usually at lmg/ml, unless otherwise stated) , incubated for 30 minutes on ice, and washed twice in ice-cold HBSS+1% FCS and either analyzed by FACS as described under immunofluoresence, or further processed.

For binding of biotinylated antibody to modified cells: cells were biotinylated and avidin treated (as described above) and incubated in lOOμl of HBSS+l%FCS/lxl0⁶ cells in the presence of 0.5μg of biotin conjugated mouse IgG2a anti human HLA-DR, or identical non-biotinylated antibody, incubated on ice for 30 minutes and washed twice in HBSS+1%FCS. Any bound antibody was detected as described under immunofluoresence .

For protein G' -biotin/anti human CD28, CD3 or CD14 complexing: a 5:1 μg ratio (Sigma: binding capacity 5mg human IgG per mg protein G'approx) of mouse anti human CD28 antibody and protein G^; -biotin was incubated on ice for 30 minutes. This complex was then added to cells and incubated on ice for a further 30 minutes after which they were washed twice with HBSS+1%FCS and binding of the complex was detected as described under immunofluoresence. Cells not for binding determination were cultured in DMEM+10%FCS. See Figure 1 for cartoon representation of above treatments . Immunofl uoresence

Detection of cell surface conjugated biotin: cells

(5xl0⁵) were incubated in lOOμl of HBSS/1% FCS containing streptavidin-FITC at 1:100 dilution, left for 30min on ice, washed twice, resuspended in 0.5ml of buffer and analysed by flow cytometry.

Detection of cell surface avidin bound biotin: cells (5xl0⁵) that bound avidin were assayed by incubating the cells in lOOμl HBSS/1% FCS containing biotin-FITC at 1:100 dilution, left for 30min on ice, washed twice, resuspended in 0.5ml of buffer and analysed by flow cytometry.

Immunodetection of cell surface molecules and of bound MAbs: cell surface detection of ICAM-1, MHC class I or B7.1 was performed by incubating 5xl0⁵ cells in lOOμl HBSS/l% FCS containing the appropriate antibodies, left on ice for 30min and washed twice in same buffer. For the immunodetection of bound MAbs, cells were incubated in lOOul of HBSS/1% FCS containing FITC-conjugated F(ab')₂ fragment of rabbit anti-mouse immunoglobulins (Dakopatts, Denmark) for the detection of mouse MAbs or FITC-conjugated F(ab')₂ fragment of purified rabbit IgG anti-rat IgG (Serotec, Oxford, UK) for the detection of rat MAbs. Cells were left to incubate for 30min on ice, washed twice and resuspended in 0.5ml of HBSS/l% FCS. All cells were analysed on a Becton-Dickinson FACScan flow cytometer equipped with a 15-mW argon laser (Becton-Dickinson & Co., Mountain View, CA) . Two thousand events were collected per sample using a logarithmic data mode setting for fluorescence intensity over 4 log decades (a total of 1024 channels, i.e. 256 channels/decade). Jurkat Co- stimulation assay

Jurkat cells were plated out at 50μl aliquots of 2xl0⁶/ml (lxlO⁵ cells/ round bottom well of a 96 well plate) with lOng/ml PMA, 2μg/ml PHA or lμg/ml mouse anti-human CD28

(Nunes JA et al . J. Exp . Med. 1994; 180: 1067-1078) in the presence or absence of 4xl0⁵ modified TIB232 cells in a final volume of 200μl. After 24 hours in culture lOOμl of supernatant was harvested and assayed for the presence of human IL-2.

Determination of human IL-2 by ELISA

Human IL-2 duoset (Genzyme) performed essentially according to manufacturers instructions.

Claims

1. A method of producing a complex for activating an immune cell against an antigen, which method comprises the steps of: applying a linker to the surface of an activating cell or to the surface of a component of an activating cell, said linker attaching in a non-immunologically specific manner to said surface; and connecting a co-stimulator to the surface via the linker such that the co-stimulator can function in the activation of the immune cell.

2. A method according to claim 1 wherein 2, 3 or more costimulators are connected to said surface via one or more linkers.

3. A method according to claim 1 or claim 2 wherein a said costimulator comprises part or all of a binding domain of an antibody.

4 . A method according to claim 1 or claim 2 wherein a said costimulator comprises part or all of a cell surface receptor.

5. A method according to any preceding claim wherein the immune cell is a splenocyte.

6. A method according to any preceding claim wherein the immune cell is a T cell.

7. A method according to any preceding claim wherein a said co-simulator is capable of binding a cluster differentiation molecule.

8. A method according to claim 7 wherein said a cluster differentiation module is selected from the CD3, CD5 or CD28 cluster differentiation families.

9. A method according to any preceding claim wherein a said linker is biotin.

10. A method according to claim 9 wherein a said costimulator is connected to the biotin linker via avidin or streptavidin.

11. A method according to anyone of claims 1 to 3 and 5 to 10, wherein said costimulator comprises part or all of a binding domain of an antibody is connected to the linker via protein A or protein G or a fusion protein of protein A or protein G.

12. A method according to claim 9 wherein a said costimulator is a fusion protein comprising avidin or streptavidin .

13. A method according to any preceding claim wherein said activating cell is a tumour cell.

14. A method according to any preceding claim wherein said complex is producing in vivo.

15. A method of activating an immune cell against an antigen, which method comprises the steps of producing a complex according to any one of claims 1 to 14 and contacting the complex with said immune cell.

16. A method according to claim 15 wherein said immune cell is activated in vitro.

17. A method according to claim 15, wherein said immune cell is activated in vivo.

18. A method of therapy for the purposes of vaccination or treatment which comprises contacting complex obtainable according to any one of claim 1 to 14 with immune cells of a subject so as to activate said immune cells .

19. A method according to claim 18 which comprises administering the complex to the subject.

20. A pharmaceutical product which comprises a complex obtainable according to any one of claims 1 to 13, 34 and 35 and a pharmaceutically acceptable excipient.

21. A kit for producing a medicament which kit comprises a linker and a costimulator and instructions for carrying out a method according to any one of claims 1 to 14, 34 and 35.

22. A kit according to claim 21 wherein the instructions state a time frame for administration of the medicament to a mammalian object.

23. A method of evaluating the capability of a molecule or combinations of molecules to activate an immune cell against an antigen which comprises: making a test complex by (i) applying one or more linkers to the surface of an activating cell or to the surface of a component of an activating cell; said linker or linkers attaching in a non- immunologically specific manner to said surface, and (╧ï) connecting the molecule or molecules being evaluated to the surface via the one or more linkers; and testing the complex for ability to activate the immune cell.

24. A method according to claim 23 wherein the immune cell is a splenocyte.

25. A method according to claim 23 or claim 24 wherein the immune cell is a T cell.

26. A method according to any one of claims 23 to 25 wherein the molecule being evaluated is a said co- simulator is capable of binding a cluster differentiation molecule .

27. A method according to claim 26 wherein said a cluster differentiation module is selected from the CD3, CD5 or CD28 cluster differentiation families.

28. A method according to anyone of claims 23 to 27 wherein a said linker is biotin.

29. A method according to claim 28 wherein the molecule being evaluated is connected to the biotin linker via avidin or streptavidin.

30. A method according to anyone of claims 23 to 29 wherein the molecule being evaluated comprises part or all of a binding domain of an antibody is connected to the linker via protein A or protein G or a fusion protein of protein A or protein G.

31. A method according to claim 28 wherein the molecule being evaluated is a fusion protein comprising avidin or streptavidin.

32. A method according to any one of claims 23 to 31 wherein the antigen is of a tumour cell.

33. A method according to claim 3^~2 wherein the antigen is specific to a tumour cell.

34. A method according to any one of the preceding method claims wherein the said surface displays or is capable of displaying the antigen.

35. A method according to any one of the preceding method claims wherein the immune cell is a cytotoxic cell.