GB2197322A

GB2197322A - Bispecific antibodies

Info

Publication number: GB2197322A
Application number: GB08725811A
Authority: GB
Inventors: Michael Ronald Clark; Herman Waldmann
Original assignee: National Research Development Corp UK
Current assignee: National Research Development Corp UK
Priority date: 1986-11-05
Filing date: 1987-11-04
Publication date: 1988-05-18
Anticipated expiration: 2007-11-04
Also published as: GB8725811D0; WO1988003565A1; GB2197322B; JPH01501200A; AU616871B2; EP0289546A1; GB8626412D0; AU8156987A

Abstract

Bi-specific antibody molecules having a first binding affinity for a human T-cell receptor capable of activating killing and a second binding affinity for target cells are characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule. Such cytotoxic molecules and fragments thereof retaining the binding affinities of the whole molecule are of value in treating disease, particularly neoplastic, viral and parasitic diseases.

Description

SPECIFICATION Antibodies This invention relates to novel forms of antibody and their use as targeted cytotoxic agents, particularly in the treatment of neoplastic disease.

The treatment of neoplastic disease still remains an intractable problem despite the fact that a very wide range of cytotoxic agents has now been developed for use in the treatment of the disease. One recent approach utilises the phenomenon of effector cell retargeting (ECR) to destroy tumour cells. In this approach a bi-specific antibody is constructed which has both an anti-T cell and an anti-tumour antigen activity.In the first application of this technique (Staerz et al, Nature, T985, 314, 628 and Perez et al, Nature, 1985, 316, 354) a bi-specific antibody conjugate was constructed by chemical means but in a further application of the technique (Staerz and Bevan, Proceedings of the National Academy of Sciences of the USA, 1986, 83, 1453 and immunology Today, 1986, 7, 241) hybridoma technology has been employed to produce a bi-specific antibody molecule. The bi-specific antibody exerts its effect by binding both to a tumour cell or other form of target cell, such as a virally infected cell, and to a T-cell thereby effecting destruction of the former by the action of the latter.

The use of effector cell retargeting as described in the art does however disregard one important disadvantage of such an approach. Thus, it is possible for a bi-specific antibody to induce killing of the T-cells to which it binds through one of the natural Fc receptor mediated cell killing mechanisms of which the body is capable, for example via antibody-dependent cellmediated cytotoxicity (ADCC) involving K cells, neutrophils and macrophages, via phagocytosis by macrophages and cells of the reticuloendothelial system, or via complement activation. Such killing of T-cells will at best reduce the level of killing of the tumour or other target cells which is achieved and at worst will result in a complete lack of any target cell toxicity. It is an object of the present invention to overcome this previously quite unidentified problem.

Accordingly the present invention comprises a bi-specific antibody molecule having a first binding affinity for a human T-cell receptor capable of activating killing and a second binding affinity for target cells characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule, or a fragment thereof retaining the binding affinities of the whole molecule.

It will be appreciated that the bi-specific antibody molecules of the present invention differ from those bi-specific antibodies of the 1985 Staerz et al and Perez et al papers in as far as, although being bi-specific, the present antibodies have the normal form of an antibody molecule in which two light chains and two heavy chains are present. By way of contrast the bi-specific antibodies described in 1985 were conjugates produced by the chemical cross linking of two normal antibody molecules and contained four light chains and four heavy chains. Although the bi-specific antibodies described in the 1986 Staerz and Bevan papers consist of antibody molecules rather than conjugates, those authors failed to appreciate the problems inherent in the use of these molecules.The present invention is based on an appreciation that the two heavy chains must be selected to mitigate killing of human T-cells by the bi-specific antibody molecule.

Moreover, as described hereinafter, we have discovered that the close interaction between the two heavy chains in these molecules, but not in the conjugates, is such that even when the heavy chain having an affinity for a human T-cell receptor is of a type which would normally produce killing of the T-cells by natural Fc receptor mediated cell killing mechanisms, it is possible to counteract this killing through an appropriate selection of the heavy chain having an affinity for target cells.

The bi-specific antibody molecules of the present invention function by binding to T-cells in order to direct their toxicity and the T-cell binding affinity of the antibody molecule may be specific for any T-cell receptor which will cause killing. Thus the receptor may be associated with the ability of the T-cells to kill either directly or indirectly through the assistance of killing by other cell types or any other agent. Thus, the receptor may be one which is associated with the ability of cytotoxic T-cells to kill directly or with the ability of helper T-cells to assist killing by B-cells, or with any other indirect mode of killing (natural or artificial). Although receptors capable of activating direct killing are of particular interest, it will be appreciated that many receptors are associated with both direct and indirect modes of killing.

In particular, the binding affinity may be directed against one or both of the a and fi chains which comprise the T-cell antigen specific receptor termed Ti, and which are present on the vast majority of T-cells, or it may be directed against the receptor-associated CD3 unit (previously identified as T3) as a whole or against one of the individual chains thereof. Thus, studies with human cells presently indicate that the T-cell receptor exists as a complex of two chains identified as a (Mr about 50,000) and ss (Mr about 40,000) which are coded for by genes which are somatically rearranged in an analogous fashion to immunoglobulin genes. Each T-cell therefore possesses a unique rearrangement of genes coding for these tvvo chains.These two chains are found in association with at least three other chains which comprise the CD3 complex and are identified as y (Mr about 25,000), a (M, about 20,000) and e (Mr about 20,000).

Antibodies having a T-cell binding affinity, either for a known or a novel receptor, may be identified by an assay procedure which we have developed. Most hybridomas as well as secreting antibody have small amounts of cell surface antibody. Thus, for example, a mouse hybridoma making antibody against rat lgG2b is capable of trapping a rat lgG2b antibody by virtue of the small amount of antibody against rat lgG2b on its surface. If the hybridoma cells are labelled with a radioactive label, for example 5'Cr, and then incubated with a mixture of Tcells and monoclonal rat lgG2b antibodies against T-cells, the hybridoma cells will bind to the antibody which will in turn bind to the T-cells, thereby leading to killing of the hybridoma and consequent release of the radioactivity.Such a procedure therefore provides a means of detecting rat lgG2b antibodies with an appropriate T-cell binding capacity (i.e. those capable of inducing killing). By selection of an appropriate form of hybridoma such a screen can be conducted among antibodies of any species and class or subclass. The hybridomas producing antibodies detected in such an assay procedure may be used in preparing the bi-specific antibodies of the present invention by techniques described hereinafter.

Anti-human T-cell/anti-non human target cell bi-specific antibodies are of interest, particularly in a research context, for example as a model system in the investigation of the important requirements for cell killing with animal tumours such as those of the rat and mouse. However, the major area of interest of the present invention is in human medicine and bi-specific antibodies having a first binding affinity directed against human T-cells (and a second binding affinity directed against target cells present in the human body) are therefore of particular interest.Using assay procedures such as that described above it is possible to select from among hybridomas producing monoclonal antibodies having the ability to bind with human T-cells those which are capable of triggering killing by T-cells, one example of such an antibody being the rat IgG2b antibody produced by the hybridoma YTH 12.5.14.2 (and all related subclones) which is described in Example 3 hereinafter. This antibody binds to the T-cell antigen CD3 but other antibodies of different specificities may also be similarly selected. Other examples include the anti-human CD2 rat IgG2b antibodies produced by the hybridomas YTH 655(5)6 and YTH 616.7.10 (H.P. Tighe, Ph.D.Thesis entitled "Monoclonal antibodies against cell surface antigens involved in leukocyte function", University of Cambridge, 1987) and the anti-human CD3 mouse IgG2a antibodies produced by the hybridoma OKT3 (US Patent 4361549), as well as the group of anti-CD3 mouse lg antibodies listed by Kurrle et al, Leukocyte Typing II, Volume 1, Human T Lymphocytes, edited by Reinherz et al, Springer Verlag, 1985, page 137, and various of the antibodies listed for the Third International Workshop and Conference on Human Leukocyte Differentiation Antigens, Leukocyte Typing Ill, White Cell Differentiation Antigens, edited by McMichael, Oxford University Press, 1987, in particular those antibodies against the CD3 unit and also against the Ti receptor, as well as certain of the anti-CD2 antibodies.It will be appreciated that hybridomas producing all of these antibodies may of course be used in derivative forms (as reclones or subclones) or that analogous hybridomas of similar specificity may be employed.

The present invention thus further comprises a bi-specific antibody molecule having a first binding affinity for human T-cells and a second binding affinity for target cells present in the human body characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule.

The second binding affinity of the bi-specific antibody molecules according to the present invention may be directed against any antigen present on the surface of a target cell. The target cell may be any cell which may be beneficially removed from the body. Examples include virally infected cells (viruses themselves not normally being attacked by T-cells), the virus being of various types including the influenza and rabies viruses, and both parasitized cells and parasites themselves including those responsible for malaria, leprosy, trypanosomiasis and schistomiasis, as well as tapeworms and other parasitic worms such as helminths. The preferred target cells are however tumour cells, the second binding affinity being against any tumour-associated antigen.The ideal situation would be for the affinity to be for a tumour-specific antigen, i.e. an antigen found on tumour cells only and not on normal cells. However, the B-cell Ig idiotype which has been employed in the treatment of B-cell malignancies such as BCLL is one of the very few examples of such antigens which exist, and in practice the tumour-associated antigen will usually also exist on normal cells. Preferably, the antigen is anomalously expressed at higher levels or in an appropriate way on tumour cells thereby allowing an enhanced level of antibodyantigen reaction with the tumour cells but even this is not completely necessary. Thus, for example, even in the extreme case where there is no differentiation between tumour cells and normal haemopoietic cells, the toxic effects of the treatment on the normal cells can be countered by the use of marrow transplants or of removal of the patient's own bone marrow prior to the treatment and its return thereafter, conveniently following separate treatment of the bone marrow in vitro for the removal of tumour cells.Examples of anti-tumour antibodies of particular interest are antibodies against antigens which define clusters of differentiation (CD) of the haemopoetic system recognised by groups of monoclonal antibodies standardised and characterised by International Workshops on Human Leukocyte Differentiation Antigens (Paris 1980, Boston 1983, Oxford 1986), and antibodies to the common acute lymphoblastic leukaemiaassociated antigen CALLA, the carcinoembryonic antigen as expressed on human colon carcinoma, and the human melanoma-assoeiated ganglioside GD3. A specific example is the human B-cell differentiation antigen CD19 which is expressed on normal B-cells and many malignant Bcells and B-cell lines.

The bi-specific antibody molecules described herein may be produced by the chemical linkage of the halves of two antibodies (which may be produced by classical techniques or by hybridoma technology), one of the first binding activity and another of the second binding activity.

However; particularly preferred bi-specific antibody molecules according to the present invention are those produced directly by the techniques of hybridoma technology modified as necessary for the production of bi- rather than mono-specific antibodies. Such techniques for the preparation of bi-specific antibodies, in general, are described in European Patent Application 0068763 and PCT Application WO 83/03679 which relate broadly to bi-specific antibody molecules produced by hybridoma technology. Examples of myeloma starting materials are the Y3-Ag 1.2.3 myeloma of European Patent No. 001459 (C.N.C.M. No. 1-078), the YB2/3.0.Ag.20 myeloma of European Patent No. 0043718, the myeloma P3-X63-Ag8 (A.T.C.C. No. CRL 1597) the myeloma NS1/1-Ag4-1 (A.T.C.C. No.TIB 18), and the myeloma P3 (Staerz et al, Journal of Immunology, 1985, 134, 3994-4000), as well as the various human myelomas referred to in the literature such as those of European Patent Applications 0062409 and 0148644, U.K. Patent 2086937 and U.S. Patent 4529694. It is preferred that each hybridoma is either derived from a myeloma which does not express a light chain or is a myeloma light chain loss variant, i.e. being HL rather than HLK. The techniques used in the preparation of the bi-specific antibodies of the present invention closely parallel those described in U.K. Patent Application 2,144,147A which may also be applied in the case of the present invention, although the binding affinities of the fusion partners are of course different. We have, however, developed a variant of these techniques which has proved to be of especial value.As in previous techniques described for the preparation of bi-specific antibody molecules our procedure involves the fusion of two hybridomas, but in the present instance one of these produces antibodies directed against T-cells and the other produces antibodies directed against target cells. The two hybridomas may be fused by the conventional procedures, as illustrated in the Examples, to produce a hybridoma secreting bi-specific antibody molecules having each of the specificities of the hybridomas from which they derive. It will be appreciated, however, that one of the most difficult stages of any hybridomaproducing fusion is the selection from the fusion mixture of the desired type of bi-specific hybridoma and the procedure we have developed involves features directed particularly to this end.

In our procedure, the first hybridoma used has a drug selectable marker which may conveniently be an enzyme deficiency, for example of thymidine kinase (TK) or particularly of hypoxanthine-guanine phosphorobosyl transferase (HPRT). Such hybridomas are obtained by selecting cells from a growth medium containing 5-bromouracyl deoxyribose or 2-aminopurine (for TK) or 8-azaguanine or particularly 6-thioguanine (for HPRT). Cells selected for growth on a medium containing the appropriate drug lack the enzyme in question and, in the case of both TK and HPRT, will therefore be incapable of growth in a medium containing hypoxanthine, aminopterin and thymidine (HAT) since the aminopterin blocks the main pathway for purine and pyrimidine synthesis and the lack of HPRT and TK removes the ability possessed by normal cells of utilising the hypoxanthine to make purines and the thymidine to make pyridimines.

The second hybridoma employed in our procedure is poisoned with a lethal dose of an irreversible biochemical inhibitor, for example diethylpyrocarbonate and particularly iodoacetamide.

Such an inhibitor poisons the cells but does not damage their DNA, which codes for immunoglobulin expression and also for the HPRT enzyme. Following the treatment of these cells they are washed to remove any excess of the inhibitor and are then ready for use, the overall structure of the cells remaining intact for several hours after the treatment, the fusion typically being carried out within 0.5 to 1 hour.

Fusion of the two hybridomas will produce a fused cell system which possesses the DNA from both hybridomas and in which. any short-term loss of vital enzyme function from the poisoned cells will be complemented by the enzymes derived from the other hybridoma. Following the fusion of the two hybridomas the fusion mixture is cultured in a medium free from inhibitors when cells of the unfused hybridoma which has been poisoned will gradually die whilst cells of the other unfused hybridoma and the fused cells will survive.Selection is then commenced with, for example, a HAT-containing medium when cells of the unfused hybridoma lacking TK or HPRT will die but the fused cells will survive since the enzyme deficiency is met by the DNA from the other hybridoma. lodoacetamide has been found to cause cell death within 1 to 24 hours but it has been found that a greater level of hybridisation is generally obtained if selection is delayed for 2 to 3 days, possibly because HPRT expression derived from the poisoned cell requires a little time to occur fully. Moreover, best results have been obtained using an equal or higher proportion of untreated cells to iodoacetamide-treated cells, for example from 1:1 to 10:1.

The mixture of cells from which both types of unfused hybridoma cell have been eliminated is then treated further in the usual way to isolate therefrom hybridomas secreting monoclonal antibodies having the desired bi-specificity. Such hybridomas may be cultured either in vitro or in vivo by conventional techniques in order to produce supplies of the monoclonal antibodies.

The mitigation, and desirably the substantial avoidance, of the killing of T-cells by the bispecific antibodies of the present invention is based on an appreciation that this problem can be overcome by the use of an appropriate combination of heavy chains in the bi-specific antibody molecules of the present invention. The overall structure of an immunoglobulin is determined by the interactions of the various globular domains of the individual chains with each other. These interactions consist of covalent bonds involving intra-chain and inter-chain disulphide bridges as well as non-covalent interactions involving both protein and carbohydrate groups.In order to activate the killing mechanisms, complement components and Fc receptors must bind to structures present in the different antibodies but different species, isotypes and allotypes of antibody have differences in portions of their protein sequences although they may have many similarities in other portions of these sequences. Thus different immunoglobulins will interact differently with complement components and Fc receptors and, in addition, when hybrid antibody molecules are made the two heavy chains may differ in sequence at crucial points for their interaction with each other and this may influence the properties of different combinations.

The immunoglobulins which constitute antibodies may be divided into several classes, the major ones of these being identified as IgG, IgA, lgM, IgD and IgE, of which some, in particular IgA and especially IgG, may be further divided into subclasses. The bi-specific antibody molecules according to the present invention preferably contain two heavy chains of the same class but may be of different subclasses within that class and may be of the same or a different subclass but relate to different species, the human, rat and mouse being of most interest, although immunoglobulins of other species, for example the rabbit, may be employed. Combinations of rat and mouse and rat and rat immunoglobulins are preferred to the mouse and mouse combination in terms of the higher level of stability of the corresponding hybridomas.

As regards the Fc receptor mediated killing mechanisms, particularly ADCC, we have found that although the nature of the heavy chain providing the T-cell binding affinity is of prime importance, the nature of the heavy chain providing the target cell binding affinity is also of importance. Thus, certain subclasses of immunoglobulin of a first species will not interact with the cell mediated effector mechanisms of a second species and with those subclasses which will interact when two heavy chains of the same species and subclass are present in the immunoglobulin it Is possible to interfere with this interaction by replacing one of the two heavy chains by an appropriate selection from another species or subclass.With the complement activation mechanism, the nature of the heavy chain providing the T-cell binding activity may be a more dominant factor in determining the level of complement activation which occurs than it is with ADCC, but the nature of the heavy chain providing the target cell binding activity can still exert an important effect as discussed hereinafter.

In selecting an appropriate combination of heavy chains for use in a bi-specific antibody molecule according to the present invention it must be appreciated that the species to which the two heavy chains relate and the species to which the antibody is to be administered are all of importance, Thus, a particular heavy chain species/subclass combination which does not produce killing of T-cells in the mouse or rat, for example, may well do so in the human, which is the species of choice as regards the T-cell binding affinity of the bi-specific antibodies. As indicated hereinbefore, human heavy chains may conveniently be used for at least one of the heavy chains in the bi-specific antibody molecules of the present invention. Indeed, a preferred choice is the use of two human heavy chains of the same class but of a different subclass.However, the relative lack of availability of human myelomas as compared with mouse and rat myelomas can pose a problem and in practice these species may often therefore represent the mammalian species of choice for the heavy chains.

In the human, the IgG class is divided into the subclasses IgG1, lgG2, lgG3 and lgG4 (the IgA class being divided into the subclasses IgAl and IgA2) whilst in the mouse and rat only the IgG class is divided into subclasses, these being IgG1, IgG2a, IgG2b and lgG3 in the mouse and IgG1, IgG2a, lgG2b and IgG2c in the rat (the similarly named subclasses not necessarily having similar properties in the mouse and the rat). In practice each of the heavy chains is most likely to be of mouse or rat IgA or IgE, particularly lgM and especially IgG, the commonest situation being that they are each of an IgG subclass and especially mouse IgG1, IgG2a or lgG2b, and less commonly lgG3, or rat IgG1, IgG2a or IgG2b, and less commonly IgG2c.

In selecting a heavy chain having T-cell binding activity, the simplest course is to use a class or subclass, for example of the rat or mouse, which does not lead to the killing of T-cells in the species in question via an Fc receptor mediated mechanism. (The word isotype is used in the art to designate a particular class and/or subclass so that in the rat, for example, IgM, IgG1, IgG2a, lgG2b and lgG2c each constitute a separate isotype.) A convenient choice for the avoidance of ADCC killing of human T-cells is a heavy chain of the IgM class but such heavy chains are often particularly effective at producing killing through complement activation.In terms of activity and of common availability a heavy chain of the IgG subclass is therefore an usual choice and for the avoidance of killing through complement activation the order of preference in the mouse is lgG1 > lgG3 > lgG2a=lgG2b and in the rat is lgG1 > lgG2cIgG2a > lgG2b. As regards the avoidance of killing through other Fc receptor mediated killing mechanisms, particularly via ADCC involving K-cells, the preference in the mouse is IgG2b, lgG3 and IgG1 > lgG2a and in the rat IgG1 and especially IgG2a and lgG2c > lgG2b. As mentioned hereinbefore, however, it is possible in the case of such a mechanism to avoid killing, even though the heavy chain of T-cell binding activity is one such as mouse IgG2a or rat IgG2b which will promote killing, through the selection of a suitable form of heavy chain having target cell binding activity, in particular one of a different species or of a different isotype or allotype. Among the other rat IgG subclasses the preferences for inactivating the rat IgG2b heavy chain are a rat lgG2a or especially a rat lgG2c heavy chain. With such active heavy chains as rat lgG2b or mouse lgG2a a combination with an alternative form of heavy chain, either by species (for example rat IgG2b/mouse IgG1) or by subclass, is indicated since a rat IgG2b/rat IgG2b or mouse IgG2a/mouse IgG2a combination can generally be presumed to be effective in promoting killing through the ADCC mechanism and possibly also through the complement activation mechanism.The use of heavy chains of a different allotype or particularly a different species or isotype can also be of value when the anti T-cell heavy chain may exhibit only an insubstantial level of effectiveness in causing killing (i.e.

being substantially ineffective), possibly acting at a low level through only one of these two mechanisms as may be the case for example with some mouse IgG isotypes. Although similar combinations of isotype may be used in such instances, the use of a difference of species or isotype will mitigate even the insubstantial level of effectiveness and provides a clear indication of suitability for the heavy chain combination. In general the achievement of the mitigation of the killing of human T-cells may be assessed by a comparison with such bi-specific antibody molecules in which the heavy chain having the affinity for target cells is the same (species, isotype and conveniently allotype) as that having the affinity for human T-cells.

The present invention thus also includes a bi-specific antibody molecule having a first affinity for a human T-cell receptor capable of activating killing and a second binding affinity for target cells, for example target cells present in the human body, characterised in that the two heavy chains in the molecule are selected from different species, isotypes or allotypes to mitigate the killing of human T-cells by the molecule.

In addition to the use of the information given hereinbefore, simple test procedures may be used to determine appropriate combinations of heavy chains for use in the bi-specific antibody molecules of the present invention. Thus, if it is proposed to use in a bi-specific antibody molecule the heavy chain/light chain combination of a particular monoclonal antibody against Tcells which is itself capable of inducing the killing of T-cells, then the hybridoma producing this monoclonal antibody can be fused with a hybridoma producing a specified immunoglobulin type of monoclonal antibody of any irrelevant binding affinity. The resulting bi-specific antibody can then be tested to see if the T-cell killing ability of the first heavy chain/light chain combination is negated by combination with a heavy chain/light chain combination of the immunoglobulin type in question.Thus, for example, by fusing a hybridoma producing a rat IgG2b antibody against Tcells with a hybridoma producing any irrelevant rat IgG2a or lgG2c antibody it can be tested whether the ability of the IgG2b heavy chain to induce killing by both the ADCC mechanism and through the complement activation mechanism is retained or not. An alternative approach is to transfect cloned immunoglobulin genes into a hybridoma so that the cloned gene is expressed in the hybrid cell and mixed immunoglobulin molecules produced which can be assayed for activity in T-cell killing by both the ADCC and complement route.

Although such procedures may be used with advantage to identify heavy chain combinations of particular value, it will be appreciated that in general any difference of species or difference of isotype (i.e. either of class or subclass) is sufficient to mitigate the killing of T-cells by the bispecific molecule and that many differences of allotype will also achieve this result. Other criteria relating to the selection of combinations of heavy chains are discussed hereinafter.

Even though an appropriate combination of heavy chains not leading to the killing of T-cells is present in the bi-specific antibody, however, the procedures available for the preparation of that antibody, particularly when using hybridoma technology, can lead to a complex mixture of antibodies, some of which may show T-cell toxicity even though the bi-specific antibody does not. The various types of different antibody which may be present in the mixture obtained on preparing a bi-specific antibody molecule from two hybridomas not expressing a myeloma derived light chain are illustrated in the Fig. 1 which appears at the end of the specification.In theory, every possible combination of the two different light chains and two different heavy chains (shown in the Figure as black for those having T-cell binding activity and white for those having target cell binding activity) can occur and in practice some antibodies of each type may be presumed to be obtained although not necessarily in equal proportions. It is only in the case where the bi-specific antibody molecule heavy chain having a T-cell binding affinity will not induce the killing of T-cells by an Fc receptor mediated killing mechanism, such as ADCC or complement activation, that none of the types of antibody will be capable of destroying T-cells.

In a case where this heavy chain will cause the killing of T-cells via an Fc receptor mediated killing mechanism, types 2 and 4 will be toxic via this route irrespective of the nature of the target cell-binding heavy chain, types 1 and 5 only being prevented from also being toxic by the presence of an appropriate target cell-binding heavy chain which is such as to negate the activity of the other heavy chain. Accordingly even though the bi-specific type 1 monoclonal antibody molecules do not cause the killing of T-cells owing to the selection of an appropriate combination of heavy chains for the anti-T cell and anti-target cell halves of the molecule, types 2 and 4 will always possess the undesirable ability to kill T-cells unless the heavy chain they contain is inherently incapable of inducing Fc receptor mediated killing.It should be appreciated, however, that where their target cell-binding heavy chain is of a type which will induce Fc receptor mediated killing mechanisms the antibody molecules of types 3 and 7 (type 6 usually being inactivated by the anti-T cell heavy chain) will contribute an added mode of target cell killing to the T-cell mediated toxicity of the type 1 bi-specific antibody molecules (the antibody molecules of types 8, 9 and 10 are inactive as regards both T-cell and target cell binding in view of their mismatched light and heavy chains). In the preferred case where types 3 and 7 contribute added target cell toxicity they may therefore conveniently be retained.

In a further aspect of the present invention, therefore, the mixture of antibodies containing the bi-specific antibody molecules of the invention which is produced by a hybridoma system derived from two fused hybridomas or two other fusion partners may be fractionated to enhance the proportion of the bi-specific antibody molecules therein and preferably substantially to separate the bi-specific antibody from or at least to reduce the proportion of other species which are undesirable, i.e. any types of molecule which act as a diluent to the type 1 molecules and/or compete therewith for binding to T-cells and particularly any species which are toxic to T-cells by an Fc mediated receptor mechanism, but not necessarily other species which are toxic to target cells.

The techniques used for such purification are broadly analogous to those described in UK Patent Application 2,144,147A where the product it is desired to purify is one of type 4, being a monovalent monoclonal antibody. Such procedures include the use of affinity chromatography, ion exchange chromatography and chromatofocussing, particularly using fast protein liquid chromatography (FPLC-Pharmarcia Trade Mark) and high pressure liquid chromatography (HPLC) techniques.

Affinity chromatography can be exploited using either an antigen-containing column (for example an immunoglobulin) which will select for those species of molecule having the correct combination of heavy and light chains for specificity or, alternatively, an anti-isotype or protein A column can be used to separate on the basis of isotype. Ion exchange chromatography relies on the fact that at different pHs the charge on a protein varies as different side chains ionize so that the binding of protein to a charged column can be affected by ionic strength. A powerful application of ion exchange chromatography involves the separation of fractions on a first column at a first pH followed by the use of a second column at a second pH, the columns usually being of opposite charge so that cation exchange chromatography is followed by anion exchange chromatography, or vice versa.Chromatofocussing relies on the fact that at a particular selected pH, the protein has no net charge and will not bind to a charged column so that similar mixed proteins are separated on a basis of their pl. FPLC and HPLC offer different advantages and may be used in combination.

As Indicated above, by an appropriate selection of anti-human T-cell heavy chain which does not exhibit T-cell toxicity or by the combination with a toxic anti-human T-cell heavy chain of an anti-target cell heavy chain of a different species or isotype it is generally possible to achieve the desired aim of the present invention to provide a bi-specific antibody molecule which substantially avoids the killing of human T-cells. By the use of purification techniques as indicated above it is possible to remove other components toxic to human T-cells which would otherwise cause a composition containing the bi-specific antibody molecules to exhibit T-cell toxicity.

The position with regard to the desirable degrees of fractionation of the mixture of antibody molecules of types 1 to 10 may be considered, by way of added exemplification, in the case of those bi-specific antibodies containing different combinations of rat lgG2 heavy chains. As indicated previously, rat IgG2b heavy chains are effective at inducing the killing of human T-cells by an Fc receptor mediated killing mechanism whilst lgG2a and IgG2c heavy chains are not and, moreover, will act to negate the toxic effect of an lgG2b heavy chain when combined therewith in the same bi-specific antibody molecule.A rat IgG2b anti-T-cell/rat lgG2b anti-target cell bispecific antibody is thus itself expected to be toxic to human T-cells whilst a rat lgG2b anti-T cell/rat IgG2a (or 2c) anti-target cell bi-specific antibody is not itself expected to show such toxicity. However, this latter type of antibody will be obtained in admixture with type 2 and 4 antibodies which will show such toxicity and are therefore preferably removed. A rat lgG2a (or 2c) anti-T-cell/rat IgG2a (or IgG2c) anti-target cell bi-specific antibody should neither be toxic to human T-cells itself nor be obtained in admixture with type 2 and 4 antibodies which show such toxicity so that the advantage of fractionation is only in the removal of other antibodies acting as a diluent to the bi-specific antibody.The fourth type of bi-specific antibody, which is rat lgG2a (or 2c) anti-T-cell/rat IgG2b anti-target cell, has a particular advantage over the other types in that the bi-specific antibody should not be toxic to human T-cells itself, nor should the type 2 and 4 antibodies, but the type 3 and 7 antibodies should contribute an added target cell toxicity through Fc receptor mediated killing mechanisms (type 6 will be inactivated against target cells by the IgG2a (or 2c) anti-T cell heavy chain) and no fractionation is required unless this is to remove inactive diluent antibodies.

The order of value for such rat IgG2 combinations in human therapy is thus (for the anti-Tcell/anti-target cell combination): lgG2b/lgG2b < IgG2b/IgG2a (or 2c) < lgG2a (or 2c)/lgG2a (or 2c) < lgG2a (or 2c)/lgG2b. Other rat IgG combinations of particular interest for human therapy are lgG2b/lgG1, which should not be toxic to T-cells as the bi-specific antibody but will provide toxic type 2 and 4 molecules, and especially lgG1/lgG2b which should not be toxic to T-cells as types 1, 2, 4 and 5 but should provide molecules of types 3 and 7 which contribute target cell toxicity.Alternatively, the four possible different combinations of IgG1 with lgG2a or IgG2c may be used, which would all be expected to behave generally similarly to the IgG2a (or 2c)/lgG2a (or 2c) combinations described above. As regards the mouse, the combinations of similar isotopes IgG1/lgG1, lgG2a/lgG2a and lgG2b/lgG2b are generally somewhat less preferred (particularly lgG2a/lgG2a for the reasons given for rat IgG2b/rat IgG2b above) than all of the possible dissimilar isotype combinations of IgG1, lgG2a and IgG2b, which are of interest.Some of these dissimilar isotype combinations will of course be of greater interest than others on similar reasoning to that applied in the rat case so that, for example, combinations in which the anti-Tcell binding is provided by mouse IgG2a will be expected disadvantageously to provide toxic type 2 and 4 molecules whilst those in which the anti-target cell binding is provided by mouse IgG2a will be expected advantageously to provide toxic type 3 and 7 molecules. The whole range of rat IgG/mouse IgG and mouse IgG/rat IgG combinations are also of some interest.

As indicated previously, the bi-specific antibody may generally be used in the form of a fragment retaining the binding affinities of the whole molecule, particularly a F(ab')2 fragment.

Although the use of F(ab')2 portions of the antibody molecule can have certain advantages it also has disadvantages so that, being smaller, their half life in serum is shorter and, in particular, the antibody molecules of types 3 and 7 will not contribute the Fc-mediated toxicity discussed above in the F(ab')2 form.

The preparation of the F(ab')2 portion of the antibody molecule requires the use of an appropriate enzyme system to effect cleavage at a suitable point in the heavy chains to thereby remove the Fc region of the heavy chains (or alternatively at least that of the anti-T cell heavy chain) whilst retaining the remaining portion of the two heavy chains, each linked to its light chain and also to the other heavy chain via disulphide bridges.It will be appreciated that different species and isotypes have different amino acid sequences in the cleavage area of the immunoglobulin and that an appropriate enzyme specific for the amino acid sequence in question must be used; Furthermore, where the two heavy chains differ in species or isotype not all cysteine groups in one chain will necessarily be connected through a bridge with a cysteine group in the other chain and the possibility of certain modes of cleavage yielding Fab rather than F(ab')2 fragments must be considered.It will therefore be necessary to select an enzyme, for example pepsin, papain, V8 protease, etc., and conditions of pH, temperature and time for its use which are appropriate to the bi-specific antibody system in question in order to produce a F(ab')2 fragment (or other fragment lacking only one Fc region, in particular that of the T-cell binding heavy chain) retaining the two specificities of the whole antibody. Having selected the appropriate enzyme and conditions, as identified by the production of a product having the characteristics indicated above, the techniques for the production of the F(ab')2 fragments of the bi-specific antibodies according to the present invention are broadly similar to those described in the literature for the preparation of the F(ab')2 fragments of mono-specific antibodies.

The present invention thus includes (a) a F(ab')2 fragment of an antibody molecule having a first binding affinity for a human T-cell receptor capable of activating killing and a second binding affinity for target cells characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule, and also (b) a process for the preparation of such a fragment which comprises treating the antibody molecule with a suitable enzyme system to effect cleavage thereof to yield this fragment.

It will be appreciated that the fractionation of the different antibody molecules of types 1 to 10 as discussed hereinbefore is still of value where the F(ab')2 fragment is used in order to remove other species acting as a diluent and/or as competitors for binding to T-cells. The fractionation may be carried out either before, or preferably after, formation of the F(ab')2 fragment.

It will be appreciated that the present invention therefore further includes (a) an antibody molecule having a first binding affinity for a human T-cell receptor capable of activating killing and a second binding affinity for target cells characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule, or a fragment thereof having the binding affinities of the whole molecule, for use in surgery, therapy or diagnosis and also (b) the use of an antibody molecule having a first binding affinity for a human T-cell receptor capable of activating killing and a second binding affinity for target cells characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule, or a fragment thereof having the binding affinities of the whole molecule, for the manufacture of a medicament for use in the treatment of neoplastic or other disease.

The bi-specific antibody molecules and fragments thereof described herein may be formulated for use in various ways, which will however, usually involve the use of a physiologically acceptable diluent or carrier which will conveniently be sterile and preferably also pyrogen-free for certain uses. This may take various forms, for example phosphate buffered saline, saline, balanced salt solution and dextrose solution. However, phosphate buffered saline may be mentioned especially as often being suitable.

The composition may, if desired, be presented in unit dosage form, i.e. in the form of discrete portions containing a unit dose, or a multiple or sub-unit dose.

The bi-specific antibody molecules or fragment thereof may be administered in various ways, for example intravenously, intraperitoneally or possibly intracerebrally, the mode of administration being selected to be appropriate to the type and localisation of the tumour or other target cells and also for ease of administration by the clinician and for the safety of the patient. In general, however, parenteral administration, and particularly intravenous injection, will often be used. As regards dosages of the bi-specific antibody molecule or fragment thereof, the exact dosages will depend upon the potency of the reagents, the tumour or other disease burden of the patient and the patient's body weight/surface area ratio.It may, however, be stated by way of guidance that a dosage of between 1 to 25 mg of the antibody molecule or fragment, for example approximately the same amount of each, will often be suitable, conveniently used in a 7 to 10 day regimen involving 1 dose of each per day, i.e. a 10 day course of treatment involving the administration of a total dosage of 10 to 250 mg of the antibody or fragment to the patient. It will be appreciated however that dosages outside this range may be used where appropriate although a particular advantage of the present invention is the low dosages which may be used in many cases, i.e. towards the lower end of the range stated or even below it, thereby possibly even avoiding the setting up of an immune response to the bi-specific antibody molecules and thus allowing repeated usage.

It will be appreciated that the present invention includes the use of an antibody molecule having a first binding affinity for a human T-cell receptor capable of activating killing and a second binding affinity for target cells characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule, or a fragment thereof having the binding affinities of the whole molecule, in the treatment of neoplastic or other disease.In particular, it includes a method for aiding the regression and palliation of neoplastic or other disease which comprises administering to a patient in need thereof an amount therapeutically effective in achieving such regression and palliation of an antibody molecule having a first binding affinity for human T-cells and a second binding affinity for tumour or other target cells characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule, or a fragment thereof having the binding affinities of the whole molecule. Alternatively, however, the bi-specific antibodies may be used for the removal of neoplastic cells from bone marrow in vitro, thereby allowing autologous bone marrow transplantation to be used in the treatment of malignancy.

The present invention is illustrated by the following Examples.

Example 1 Preparation of hybridoma-producing monoclonal antibodies specific for the human CD3 antigen and for the mouse Thy-l antigen (1) Anti-CD3 component The Lou rat myeloma cell line Y3-Ag 1.2.3 (CNCM, 1-078) is fused with spleen cells from a DA rat immunised with human lymphocytes according to the procedure described by Clark and Waldmann, Methods in Hematology, 1986, 13, 1-20, and the fusion mixture worked up as described therein selecting for hybridomas producing monoclonal antibodies having specificity for the human CD3 antigen by reactivity with all human peripheral T-cells, cross inhibition with mouse monoclonal antibodies such as UCHT-1 (Burnset et al, Journal of Immunology, 1982, 129, 1451) and OKT-3 (US Patent 4361549), and immunoprecipitation.The selected hybridoma producing an antibody of such specificity is hypoxanthine-guanine phosphoribosyl transferase positive (HPRT+) and expresses one spleen cell-derived light chain and a second myelomaderived light chain of the kappa 1 a allotype.

Selection is made for myeloma light chain loss variants by cell cloning on semi-solid agar (Clark and Waldmann, ibid) and then assaying for the loss of rat kappa-la allotype expression using a sensitive red cell haemagglutination assay (Clark, Methods in Enzymology, 1986, 121, 548-556). The variant selected is cloned on semi-solid agar and then maintained in culture in Iscoves modification of Dulbecco's medium (IMDM--Gibco Europe) supplemented with 1 to 5% v/v foetal calf serum (FCS) and buffered with bicarbonate using 5% CO2 in air. For short term handling of the cells the bicarbonate is replaced by extra HEPES buffer and NaCI to maintain the ionic strength.

(2) Anti-Thy-l component The Lou rat myeloma cell line Y3-Ag 1.2.3 (CNCM, 1-078) is fused with spleen cells from a DA rat immunised with the mouse Thy-1 antigen according to the procedure described by Cobbold et al, Molecular Biology and Medicine, 1983, 1, 285-304, selecting for hybridomas producing monoclonal antibodies having specificity for the Thy-1 antigen as described therein using an assay based directly upon binding to this antigen.The selected hybridoma producing a monoclonal antibody having specificity for the mouse Thy-1 antigen is then selected further for a variant of this hybridoma which is HPRT negative by effecting culture in increasing concentrations of medium containing the selective drug 6-thioguanine (Clark and Waldmann, ibid). The selected HPRT- variant is cloned and cultured as for the hybridoma (1).* * In a first variant of this procedure the hybridoma (1) is 6-thio-guanine resistant and the hybridoma (2) is poisoned with iodoacetamide in step (3) (see Example 5), in a second variant 8azaguanine is used instead of 6-thioguanine in either the original or first variant procedures, and in a third variant the hybridoma (2) is a myeloma light chain loss variant as well as or instead of the hybridoma (1), or one or both hybridomas are derived from a myeloma which does not express a light chain.

(3) Hybridoma-Hybridoma Cell Fusion Prior to fusion the cells of the HPRT+ hybridoma (1) (5x 106) are washed into phosphate buffered saline (PBS) by centrifugation at 200xg and are resuspended in 10 ml PBS containing 5 mM iodoacetamide. The cells are incubated on ice for 30 minutes and are then washed into HEPES buffered IMDM. These iodoacetamide-poisoned cells are mixed with cells of the hybridoma (2) (5x 107), the cell mixture is washed once with HEPES buffered IMDM and the mixed cells are then pelleted at 200xg. Cell fusion is induced by treating the cell pellet for 2 minutes with 1 ml of a 50% w/v solution of polyethylene glycol 1500 in PBS whilst stirring (Clark and Waldmann, ibid).The cells are washed once with HEPES buffered IMDM and are resuspended in bicarbonate buffered IMDM containing 5% v/v foetal calf serum, then being plated out into 48 x 2 ml culture wells and cultured at 37"C under 5% CO2. On the following day a control containing iodoacetamide-treated, but unfused, cells is typically totally non-viable, whilst in the cultures of the fused cells the majority are observed to be viable.

After culture of 48 hours selection is commenced with hypoxanthine, aminopterine and thymidine (HAT selective medium Clark and Waldmann, ibid), the HPRT- hybridoma (2) cells being non-viable in this medium. Culture in the HAT selective medium is continued for 2 or more weeks and the following assays are then employed to screen for hybrid cell lines producing bispecific antibody.Firstly, a series of mouse monoclonal antibodies specific for rat immunoglobulin chains is employed as isotyping reagents in a rapid red-cell linked haemagglutinin assay (Clark, ibid). Examples of such antibodies are MAR 18.5 (Lanier et al, Hybridoma, 1982, 1, 125-131), which is specific for all rat kappa light chains, RG11/15.5 (Springer etna!, Hybridoma, 1982, 1, 257-273). which is specific for rat kappa 1 b allotype light chains, NORIG 1. 1.6 and NORIG7.16.2 (Hale et al, Journal of Immunology, 1985, 134, 3056), which are specific for the rat IgG2b isotype, and NORIG31.12.14 (Hale et al, ibid), which is specific for the rat lgG2c isotype. The same reagents are also used in sandwich enzyme-linked immunoassays as follows.

A relevant anti-isotype monoclonal antibody is used to coat plastic microtitration plates by incubating a serum free culture supernatant at 100 ,ul per well overnight at 4"C. Unbound antibody is then washed off with PBS and the wells are filled with blocking buffer (PBS plus 1% w/v bovine serum albumen and 0.1%-w/v sodium azide) and the coated plates are stored at 4"C until needed. 100 Al of the supernatants to be tested are incubated in the wells for 1 hour at room temperature and unbound antibody is washed away with PBS containing 0.1% w/v bovine serum albumen. Bound antibody is then detected using a second anti-isotype monoclonal labelled with biotinyl succinimide ester followed by subsequent layers of streptavidin-peroxidase (Amersham) and finally the enzyme substrate ortho-phenylenediamine, the colour change of the reaction being determined at 492 nm. Cells from wells which are positive in both assay procedures are cloned twice in semi-solid agar and a suitable clone is selected, this being maintained in culture as for the mono-specific hybridoma (1)." * In a variant of this procedure the anti-Thy-1 hybridoma (2) is replaced by a hybridoma specific for a target cell present in the human body, particularly a tumour cell.In a further variant, which may conveniently be combined with the first, one or both of the hybridomas (1) and (2) are replaced by a hybridoma derived from a human myeloma, for example a myeloma as described in European Patent Applications 0062409 and 0148644, in UK Patent 2086937 and in US Patent 4529694.

Example 2 Production and fractionation of monoclonal antibodies from bi-specific hybridoma The bi-specific hybridoma produced in Example 1 is maintained in low serum culture (IMDM containing 1% v/v FCS) for 48 hours at 37"C using 5% CO2 in air and the culture supernatant is then concentrated by precipitation with ammonium sulphate added to a level of 50% w/v. The precipitate is redissolved in the minimum volume of water and desalted into 50 mM malonate buffer at pH 5.5 on Sephadex G25. This preparation is then filtered through a 0.2 micron filter and is subjected to FPLC (Pharmacia) ion-exchange chromatography on a Mono S column (Pharmacia).Bound protein is eluted using a linear salt gradient of O to 1 M sodium chloride, antibody-containing peaks being determined using immunoglobulin isotype specific assays of the type described in Example 1 and the pooled fractions of breakthrough and bound/eluted material are each desalted into PBS by gel filtration through Sephadex G25 and are stored at 4"C until required.

Example 3 Preparation of rat IgG2b anti-human CD3/rat /gG2c anti-mouse Thy-l bi-specific hybridoma SHN 20. 12 The procedure of Example 1(3) was followed using as hybridoma (1) the hybridoma YTH 12.5.14 (this is identical to the sister clone YTH 12.5.22 described by Cobbold and Waldmann, Nature, 1984, 308, 460-462) and as hybridoma (2) the hybridoma YBM 29.2.1 (Cobbold et al, ibid). YTH 12.5.14 produces a rat IgG2b monoclonal antibody and expresses a myeloma-derived kappa light chain of the la allotype. The myeloma chain loss variant of YTH 12.5.14, YTH 12.5.14.2, secretes a light chain unreactive with all tested anti-rat kappa reagents and may therefore be presumed to express a light chain of the lambda class.YBM 29.2.1 is an HL hybridoma which is a myeloma chain loss variant of YBM 29.2. It produces a rat lgG2c monoclonal antibody and expresses a kappa-lb allotype light chain. The 6-thioguanine resistant variant was YBM 29.2.1TG6. The fusion mixture was screened using the appropriate isotyping reagents as listed in Example 1. Growth was observed in only one of the total of 48 wells, this well containing an antibody having the immunoglobulin chains, and also the specificities for antigens, to be expected from a fusion of the two parental hybridomas. Moreover it was possible to show that the YBM 29.2.1-derived kappa light chain and lgG2c heavy chain were as associated in mixed antibody molecules with the YTH 12.5.14-derived IgG2b heavy chain.The culture was cloned on semi-solid agar and all growing clones (12/12) were found to be positive when assayed again, no chain loss variants being detected. Clone SHN 20.12 was selected for use.

Example 4 Preparation and fractionation of SHN 20. 12 monoclonal antibodies Low serum culture supernatants from the hybridoma SHN 20.12 of Example 3 were produced as described in Example 2 and then subjected to FPLC ion exchange chromatography under the conditions described in Example 2, these conditions providing a separation of the parental antibody types. Thus at pH 5.5 in 50 mM malonate buffer using a Pharmacia Mono S column with a linear salt gradient of O to 1 M sodium chloride it was found that the YTH 12.5.14.2 antibody binds strongly whereas the other parental antibody, YBM 29.2.1, occurs in the breakthrough fractions. When the specific antibody preparation SHN 20.12 was subjected to chromatography under these conditions it was found that the antibody mixture produced by the hybridoma did not give rise to any well resolved unique peaks.Instead, both preparations showed some broadening of the peak corresponding to the elution position of YTH 12.5.14.2.

The fractions of the breakthrough and of the bound and eluted material were therefore pooled separately and dialysed into PBS for use in the effector cell retargeting assay procedure.

Example 5 Preparation of rat IgG2b anti-human CD3/mouse IgG1 anti-human CD 19 bi-specific hybridoma SHR 1.6.1 A procedure similar to that of Example 1(3) was followed using as hybridoma (1) the hybridoma YTH 12.5.14.2TG101 and as hybridoma (2) the anti-human CD19 hybridoma 8EB1BU12.

In this case, however, it is the hybridoma (1) which is HPRT- and the hybridoma (2) which is poisoned with iodoacetamide. Thus YTH 12.5.14.2TG101 is a 6-thioguanine resistant variant of the myeloma chain loss variant YTH 12.5.14.2 which is described in Example 3. SEB1BU12 produces a mouse IgG1 antibody which, rather than being specific for the mouse This1 antigen as is the hybridoma (2) used in the main procedure of Example 1, is instead specific for the human CD19 antigen, this being a human B cell differentiation antigen expressed on both normal B cells and many malignant B-cells (the source of the 8EB1BU12 hybridoma was Ling and McLennan of Birmingham University, England).The wells showing growth were selected by the detection of CD3 binding to human T-cells and of CD19 binding to human B-cells using fluorescein isothiocyanate (FITC) labelled anti-rat mouse immunoglobulin (Hudson and Hay, "Practical Immunology", Blackwell Scientific Publications, 1976, page 11). On the basis of this selection, well number 1 was cloned and then re-cloned on semi-solid agar and the clone SHR 1.6.1 selected for use.

Example 6 Preparation and fractionation of SHR 1.6. 1 monoclonal antibodies Low serum culture supernatants from the hybridoma SHR 1.6.1 of Example 5 were produced and processed as described in Example 2 but, instead of using a Mono-S column for purification, the filtered preparation was subjected to ion exchange chromatography on a TSK-SPW(LKB) column under HPLC with a pH 5.8, 50 mM malonate buffer and a O to 1 M sodium chloride gradient.The various peaks were dialysed into phosphate buffered saline (PBS) to yield the following concentrations of material after dialysis: Fraction Peak Concentration (mg/ml) Tubes 2,3,4,5 and 6 pooled (breakthrough) 4.1 Tube 9 1 0.57 Tubes 10,11 and 12 pooled 2 1.43 Tube 13 3 1.07 Tube 14 4 0.36 Tube 15 5 0.36 Tubes 16,17 and 18 pooled 6 0.43 Fluorescence assays against a B-cell line and a T-cell line indicated that the breakthrough fraction was anti-CD19 active, the tube 13 and 14 fractions were each anti-CD3 active, and the tube 9 and pooled tube 10, 11 and 12 fractions were both anti-CD19 and anti-CD3 active.

Example 7 Assay of rat IgG2b/rat IgG2c bi-specific SHN 20. 12 monoclonal antibodies in effector cell retargeting (1) Preparation of human cytotoxic effector cells Venous blood was collected from healthy donors and was defibrinated using glass beads.

Mononuclear cells were isolated from the interface following density gradient centrifugation on Ficoll-Hypaque and were then washed into bicarbonate buffered IMDM containing 5% v/v FCS.

These cells were plated out at S x 106 cells per ml in the same culture medium containing 100 ,ug/ml of a mitogenic rat lgG2c monoclonal antibody such as YTH 361.1. After 3 to 5 days in culture the cell blasts were washed and used in assays for cytotoxicity. The potency of these effectors in mediating cytotoxicity could be demonstrated by using the anti-CD3 secreting hybridoma YTH 12.5.14.2 described in Example 3 as a target cell line in 5'Cr release assays. The blast cells produced in this way also contain 10-15% Fc receptor positive cells (as detected by Fe rosetting) which are abie to mediate ADCC. For certain experiments the ADCC effectors were inhibited by preincubating the washed effector cells with 1 ,ug/ml of the anti-Fc receptor antibody CLB-Fcr gran I (Tetteroo et a/, Leucocyte Typing II, 1985, Volume 3, page 27-Springer Verlanger, New York) for 15 minutes at room temperature. The antibody was not washed away prior to plating the cells out in the assay.

(2) Effector cell retargeting assay The mouse This1 positive thymona cell line EL-4 (A.T.C.C. reference number TIB40) was used as the target cell line. Cells were maintained in exponential phase in IMDM containing 2% v/v FCS until required for assay. Approximately 5x 106 cells were then spun down at 200xg and the pellet resuspended in 200 ,ul IMDM containing 1 50 ,uci S'Cr-sodium chromate. The cells were incubated for 37"C for 45 minutes and were washed into HEPES buffered IMDM. Suitable dilutions of the antibodies to be tested for effector cell retargeting were prepared in 100 jli volumes in round bottomed micro-titre wells.The radiolabelled EL-4 targets were then added at a level of 104 cells per well in a volume of 50 all of HEPES buffered IMDM containing 5% v/v heat inactivated FCS and washed cytotoxic effector cell blasts prepared as under (1) were added at a suitable effector to target ratio in a volume of 50 ul of HEPES buffered IMDM containing 5% v/v heat inactivated FCS. The cell mixture was incubated at 37"C for 4 hours and 100 'jI of supernatant was harvested to determine the released radioactivity (all of the assays being carried out in replicates of 3 or 4) by measurement of gamma radiation using a Phillips gamma counter model PW 4800.

(3) Comparison of activity of SHN 20. 12 antibodies with that of YTH 12.5. 14.2 and YBM 29.2. 1 parental antibodies and mixtures thereof Supernatants from cells in the stationary phase of the cultures prepared as described in Example 2 were utilised for the study of the YTH 12.5.14.2 and YBM 29.2.1 antibodies, and 1:1 mixtures thereof, whilst the bound/eluted preparation of Example 4 was utilised for the SHN 20.12 antibodies. The starting concentration of antibody was estimated from the antibody isotyping assays to lie in the range of 0.01 to 0.1 mg/ml. In each case the antibodies were assayed for their ability to effect the killing of EL-4 targets through human blasts as described in (2) above (the antibody YTH 361.1 having been used in the procedure of (1) above).The results of titrating antibody over the range 1/10 to 1/106 at a constant effector to target ratio (15:1) are shown in Fig. 2 where (a) is supernatant from a culture of Y3-Ag 1.2.3 myeloma cells used as a control, (b) is YTH 12.5.14.2, (c) is YBM 29.2.1, (d) is YTH 12.5.14.2+YBM 29.2.1, and (e) is SHN 20.12. It will be seen that the YBM 29.2.1 antibody does not mediate the killing of target cells in the presence of the effector cells but that the YTH 12.5.14.2 antibody does show a small but significant level of mediation which is probably due to a bystander effect of T-cell activation by this antibody (rat IgG2b isotype antibodies such as YTH 12.5.14.2 are very effective in ADCC and the human blast cells contain 10-15% Fc receptor positive cells).The action of the YTH 12.5.14.2/YBM 29.2.1 mixture will be seen to be little different from that of YTH 12.5.14.2 alone. Compared with the killing obtained with this mixture of the parental antibodies the bi-specific antibody SHN 20.12 derived therefrom shows very efficient killing of the targets. The hybrid supernatant therefore shows a much improved level of killing to the mixture of the parental antibodies despite the fact that the mixture of antibodies which is present together with the bi-specific antibodies includes combinations which would be expected to compete for binding to target or effector cells.

The effect of titrating effector to target ratio at a constant antibody concentration (1:100 v/v dilution of supernatant) is shown in Fig. 3 (the designations being similar to Fig. 1). Once again the potency of the SHN 20.12 antibodies at directing target cell killing can be seen to be very high, effective killing being achieved at relatively low effecto to target ratios and at very low antibody concentrations.

In order to confirm that the killing of EL-4 target cells by effectors in the presence of the SHN 20.12 bi-specific antibodies was not Fc receptor dependent, experiments were performed using effectors treated with an anti-Fc receptor monoclonal antibody CLB-FcR gran I. It was found that the observed killing mediated by SHN 20.12 supernatant was not altered in any way.

Example 8 Comparison of activity of rat IgG2b/rat lgG2c bi-specific SHN 20. 12 monoclonal antibodies with rat IgG2b/rat IgG2b bi-specific SHM 15.3 monoclonal antibodies (The rat lgG2b/lgG2b antibodies are used for comparative purposes only and, not showing mitigation of T-cell toxicity, do not fall within the scope of the invention.) (1) Preparation of rat lgG2b anti-human CD3/rat IgG2b anti-mouse Thy-1 SHM 15.3 monoclonal antibodies The procedure of Example 1(3) was followed using as hybridoma (1) the hybridoma YTH 12.5.14 described under Example 3 and as hybridoma (2) the hybridoma YTS 154.7.7 (Cobbold et al, ibid). YTH 12.5.14 provides a myeloma chain loss variant YTH 12.5.14.2 which is again described under Example 3.YTS 154.7.7 is a myeloma light chain loss variant of the parent hybridoma YTS 154.7 which produces a rat IgG2b monoclonal antibody and, like the hybridoma YBM 29.2.1 of Example 3, expresses a kappa-1b allotype light chain. The 6-thioguanine resistant variant of YTS 154.7.7 was YTS 154.7.7.TG9.40. The fusion mixture was screened using the appropriate isotyping reagents as listed in Example 1. Growth was observed in twenty four of the forty eight wells and all of the growing wells were found to have the immunoglobulin chains and specificities for antigens expected from the parental antibody types. The cultures giving strongest binding in the assays were selected for cloning on semi-solid agar and all were found to be positive, no chain loss variants being detected. Clone SHM 15.3 was selected for use.

Low serum culture supernatants from the hybridoma SHM 15.3 were produced as described in Example 2 and purified to give products consisting of the pooled fractions of breakthrough and of bound and eluted material, as described for the hybridoma SHN 20.12 in Example 4.

(2) Effector cell retargeting assay (a) The assay procedure described in Example 7 utilising the mouse This1 positive thymoma cell line EL-4 was applied to the SHM 15.3 monoclonal antibodies prepared as described in (1) above and to the parental antibodies used singly and in 1:1 admixture (all antibodies used in form of supernatants). The assay showed that the bi-specific antibodies exhibited more efficient killing of target cells than a 1:1 mixture of the parenteral antibodies YTH 12.5.14.2 and YTS 154.7, the bi-specific antibodies possessing a similar titre to the monospecific YTS 154.7 antibodies alone. When the effectors were treated with the anti-Fc receptor monoclonal antibody CLB-FcR gran I the observed mediated killing by SHM 15.3 antibodies was unaffected.This is similar to the observation with SHN 20.12 monoclonals but, by way of contrast, the killing produced by YTS 154.7 antibodies was eliminated by the use of CLB-FcR gran I antibodies.

(b) The assay procedure of Example 7 was employed but replacing the EL-4 cell line with the mouse This1 positive BW5 147 cell line (CRL 1588, PHLS European Collection of Animal Cell Cultures, Porton Down, England). The procedure was carried out with supernatants from cells in the stationary phase of the cultures for the three parental antibodies and with each of the pooled breakthrough and bound and eluted fractions for the two bi-specific antibodies.It was found that the parental CD3 antibody YTH 12.5 and the parental IgG2c anti-Thy-l antibody YBM 29.2 did not elicit any killing of the target cells but that the parental IgG2b anti-Thy-l antibody YTS 154.7 did induce substantial killing of the target cells, as did both groups of pooled fractions of both bi-specific antibodies. However, there were differences in the relative titres among the bi-specific antibodies as will be seen from Fig. 4 where the percentage specific target cell lysis is shown for titrations in the range of 10-1 to 10-7 mg/ml. In Fig. 4, (a) is YTS 154.7, (b) is YBM 29.2.1, (c) is YTH 12.5.14.2, (d) is bound/eluted SHN 20.12, (e) is breakthrough SHN 20.12, (f) is breakthrough SHM 15.3, and (g) is bound/eluted SHM 15.3.It will be seen that both the breakthrough and the bound/eluted fractions of SHN 20.12 gave a similar titre to YTS 154.7 (compare at 10-5 mg/ml) but SHM 15.3 showed a 10 fold lower titre for the breakthrough fraction (10-4 mg/ml similar to 10-5 mg/ml for YTS 154.7) and a 100 fold lower titre for the bound/eluted fraction (10-3 mg/ml similar to 10-5 mg/ml for YTS 154.7).

It will be seen therefore that Fig. 4 indicates that the fractions of the lgG2c/lgG2b SHN 20.12 are substantially better than those of the lgG2b/lgG2b SHM 15.3 in effector cell retargeting.

(3) Assay of Fc receptor mediated killing Freshly obtained resting peripheral blood mononuclear cells from a good K-cell donor were used as effectors in place of the T-cell blasts in the assay procedure of Example 5. It was found that target cell killing was only observed with the IgG2b YTS 154.7 and IgG2b/lgG2b SHM 15.3 antibodies which substantiates the results reported in 2(a) above using the CLB-FcR gran I antibody and suggests that the IgG2b/lgG2c SHN 20.12 bi-specific antibodies were unable to solicit any significant Fc receptor mediated killing of the target cells. Thus the killing of target cells obtained with the SHN 20.12 antibodies in the effector cell retargeting assays is attributable to the action of the T-cell blasts as effectors and apparently does not involve any significant contribution from ADCC.

(4) Assay of ability of bi-specific antibodies to kill effector cells (a) Complement mediated lysis Human effector cells were prepared as described in Example 7(1) and were labelled with 5'Cr as described for the mouse target cells in Example 7(2). Dilutions of the purified antibody fractions into HEPES buffered IMDM were made in volumes of 100 Al in micro-titre wells.

Labelled cells were added at 104 cells per well in a volume of 50 Cil and finally 50 Al of autologous human serum (from the original donor of the mononuclear effector cells) was added as a complement source. The plates were incubated at 370C for 1 hour and 100 ,ul of supernatant was harvested to determine released radioactivity.

This procedure was applied to similar antibody preparations as identified as (c) to (g) under 2(b) above, titrating these preparations in the range of antibody concentration from 10- to 5x 10-5 mg/ml. In addition the procedure was applied to a preparation of antibody which is an ion-exchange purified form of the monovalent species which is produced, together with bivalent and inactive species by the hybridoma SHL 45.6.1.SHL 45.6.1 is an anti-CD3 monoclonal antibody-producing hybridoma similar to the hybridoma YTH 12.5.22 (see Cobbold and Waldmann, ibid). Comparing the breakthrough and bound/eluted fractions of SHN 20.12 and SHM 15.3 with the controls provided by (c) and its equivalent monovalent preparation it was found that the breakthrough fractions of both bi-specific antibodies gave levels of killing similar to that of YTH 12.5 whilst the bound and eluted fractions showed a greater level of killing intermediate between that of YTH 12.5.14 and its monovalent equivalent. However, neither fraction of either SHN 20.12 or SHM 15.3 showed any detectable complement lysis at concentrations below 4x10-4 mg/ml at which level effective levels of both ADCC and effector cell retargeting activity are observed.

(b) Effector cell killing by effectors A variation of the effector cell retargeting assay of Example 7 was performed in which some (20%) of the human T-cell blast population was labeled with radioactive chromium as described for the mouse target cells of Example 7(2), and these cells were then added to dilutions of the antibodies to be assayed. Finally, unlabelled human blast effector cells were added so that the final cell numbers per well were as for the effector cell retargeting assays described above and the cultures were incubated for 4 hours at 37"C before harvesting the supernatant to determine released radioactivity (the effector cells were not treated with CLB-FcR gran I in this procedure).

This assay procedure was applied to similar antibody preparations as identified as (a) to (g) under 2(b) above, the results being shown in Fig. 5a and b; for reasons of clarity in showing the plots of preparations (g), (f) and (a) the plot of preparations (c) is not shown in Fig. 5b but only in Fig. 5a. It will be seen that the antibodies caused lysis of the human effector cells at concentrations in the same range as those which caused mouse target cell lysis (compare with Fig. 4).However, it should be noted that for the same antibody concentrations the YTH 12.5.14.2 parental antibody shows a titre (3x 10-7 mg/ml, estimated as the concentration giving 15% specific cytotoxicity) which is about 30 fold greater than the breakthrough and bound/eluted fractions of the bi-valent antibody SHM 15.3 (compare with 10-5 mg/ml value) and about 10 fold greater than the breakthrough fraction and 100 fold greater than the bound/eluted fraction of the bi-valent antibody SHN 20.12 (compare with 3x 40-6 and 2x 10-5 mg/ml values, respectively). In addition the antibody YTH 12.5 shows a marked prozone over the range tested.

It will be seen that of the four SHN 20.12 and SHM 15.3 preparations the lowest level of lysis is observed for the SHN 20.12 bound/eluted fraction which is also the fraction giving the highest cytotoxic titre against the mouse target cells. (It is believed that the cytotoxicity observed in both (a) and (b) for the breakthrough and bound/eluted fractions of SHN 20.12 is attributable to contaminants as illustrated in Fig. 1, rather than to the bi-specific antibody moiecules.) Example 9 Activity of rat IgG2b/mouse lgG 1 bi-specific SHR 1.6. 1 monoclonal antibodies against human target cells and human T-cells (1) Effector cell retargeting assay The assay procedure was based on that described in Example 7(2) but replacing the mouse Thy-1 positive thymoma cell line EL-4 by a human CD19, HLA-class II positive cell line HWLCL produced by the transformation of human B-cells with Epstein-Barr virus. The human effector cells were used in a ratio to the HWLCL target cells of 10:1 and the antibodies for test were used in a range of dilutions from 1:10 to 1:106 for culture supernatants and 1:102 to 1:107 for column fractions.The following antibody preparations were assayed: supernatants from cells in the stationary phase of cultures prepared as described in Example 2 from (a) the parental hybridomas YTH 12.5.14.2 and 8EB1BU12 (a 1:1 v/v mixture of the supernatants), and (b) the hybridoma SHR 1.6.1 of Example 5; and column fractions produced by the purification of SHR 1.6.1 as described in Example 6 corresponding to (c) tubes 2-6 (breakthrough), (d) tube 9, (e) tubes 10-12, (f) tube 13, and (g) tube 14.

The results obtained are shown in Fig. 6 from which it will be seen that the SHR 1.6.1 antibodies mediate the killing of human C19-positive cells by effector T-cells but the parental antibodies substantially do not and that fractions 9 and 10-12 (peaks 1 and 2) contain most of the bi-specific antibody activity which mediates the ECR of B-cells by T-cells.

(2) Assay of effector cell killing The assay procedure was based on that described in Example 8(4)(a) using the column fractions (d), (e) and (f) as described in (1) above, (h) supernatant from cells in the stationary phase prepared as described in Example 2 for the hybridoma SHL 45.6.1 referred to in Example 8(4)(a) and (i) a monovalent preparation from the hybridoma SHL 45.6.1 also referred to in Example 8(4)(a). Each of the preparations was diluted on a 1:3 v/v basis six times commencing with the following concentrations of antibody in the first wells which are diluted 1:3 v/v in the first dilution for the first determination: (d) 10.0 ,ug, (e) 23.8 ug, (f) 17.8 ijg, (h) 16.7 ,ug and (i) 13.2 ug.

The results obtained are shown in Fig. 7 from which it will be seen that the lowest level of complement mediated lysis results from preparation (d) (fraction 9). It is noteworthy that prepa ration (d) also gives the highest level of ECR among the fractions. Since the rat IgG2b heavy chain is one which is toxic to T-cells, the results obtained indicate that the mouse IgG1 heavy chain negates the toxicity of the rat IgG2b heavy chain. Monospecific anti-CD3 rat IgG2b/rat IgG2b antibodies will however be markedly toxic and it is likely that preparation (d) (fraction 9) contains substantially pure bi-specific antibody but in preparation (e) (fractions 10-12) the bispecific antibody is contaminated with monospecific anti-CD3 antibody thereby explaining the substantially increased T-cell toxicity of this preparation.

Claims

1. A bi-specific antibody molecule having a first binding affinity for a human T-cell receptor capable of activating killing and a second binding affinity for target cells characterised in that the two heavy chains in the molecule are selected to mitigate the killing of human T-cells by the molecule, or a fragment thereof retaining the binding affinities of the whole molecule.

2. An antibody molecule or fragment thereof according to Claim 1, in which the target cells are tumour cells.

3. An antibody molecule or fragment thereof according to Claim 1 or 2, in which the two heavy chains each separately are derived from the human, rat or mouse and are of the IgG class of immunoglobulins.

4. An antibody molecule or fragment thereof according to Claim 1, 2 or 3, in which T-cell killing is mitigated by the use of a first heavy chain having a T-cell binding affinity which is substantially ineffective at promoting killing through Fc receptor mediated killing mechanisms.

5. An antibody molecule or fragment thereof according to Claim 4, in which said heavy chain is derived from the rat and is of the lgG2a, IgG2c or IgG1 sub-class.

6. An antibody molecule or fragment thereof according to Claim 5, in which the two heavy chains are of the isotypes, for the anti-T-cell/anti-target cell specificities respectively, of lg- G2a/lgG2a, IgG2a/lgG2c, IgG2c/lgG2a, IgG2c/lgG2c, IgG2a/lgG2b or IgG2c/lgG2b.

7. An antibody molecule or fragment thereof according to Claim 4, in which said heavy chain is derived from the mouse and is of the IgG2b, IgG1 or lgG3 subclass.

8. An antibody molecule or fragment thereof according to Claim 1, 2 or 3, in which the first heavy chain having a T-cell binding affinity is effective at promoting killing through Fc receptor mediated killing mechanisms but T-cell killing is mitigated by the presence of a second heavy chain having target cell binding affinity which renders said first heavy chain less effective.

9. An antibody molecule or fragment thereof according to Claim 8, in which the first and second heavy chains are of a different species, isotype or allotype.

10. An antibody molecule or fragment thereof according to Claim 9, in which the first heavy chain is rat lgG2b or mouse IgG2a.

11. An antibody molecule according to Claim 9 or 10, in which the first and second heavy chains are of the same class.

12. An antibody molecule or fragment thereof according to Claim 11, in which the heavy chain is derived from the rat and is of the isotypes, for the first chain/second chain respectively, IgG2b/lgG2a or lgG2b/lgG2c or is derived from the mouse and is of the isotypes, for the first chain/second chain respectively, lgG2a/lgG1 or IgG2a/lgG2b.

13. A bi-specific antibody molecule having a first binding affinity for a human T-ceil receptor capable of activating killing and a second binding affinity for target cells characterised in that the two heavy chains in the molecule are of a different species or isotype, or a fragment thereof retaining the binding affinities of the whole molecule.

14. An antibody molecule or fragment thereof according to Claim 13, in which the target is a human tumour cell.

15. An antibody molecule or fragment thereof according to Claim 13 or 14, in which the two heavy chains are derived from different species among the human, mouse and rat.

16. An antibody molecule or fragment thereof according to Claim 13 or 14, in which the two heavy chains are either derived one from the rat and the other from the mouse or are of different rat or of different mouse isotypes.

17. An antibody molecule according to any of Claims 13 to 16, in which the two heavy chains are of the same class.

18. An antibody molecule or fragment thereof according to Claim 16, in which the heavy chains are both derived from the rat and are of the isotypes, for the anti T-cell/anti target cell specificities respectively, lgG2a/IgG2c, IgG2c/lgG2a, IgG2a/lgG2b or IgG2c/lgG2b or are both derived from the mouse and are of the isotypes, for the anti-T-cell and anti-target cell specificities respectively, of lgG1/lgG2b, lgG2b/lgG1, IgG1/lgG2a or lgG2b/lg2a.

19. An antibody molecule or fragment thereof according to any of Claims 8 to 12, which is substantially free from antibody molecules which have two of said first heavy chains and in which at least one of these heavy chains is in combination with its corresponding light chain, and from fragments of said antibody molecules retaining the binding affinities of the whole molecule.

20. An antibody molecule or fragment thereof according to any of Claims 13 to 18, which is substantially free from antibody molecules which have two heavy chains with said first binding affinity and in which at least one of these heavy chains is in combination with its corresponding light chain, and from fragments of said antibody molecules retaining the binding of the whole molecule.

21. An antibody molecule or fragment thereof according to any of the preceding claims, in which the heavy chain having a binding affinity for target cells is effective at promoting killing through Fc receptor mediated killing mechanisms, the molecule or fragment thereof being in admixture with antibody molecules having two such heavy chains and in which at least one of these heavy chains is in combination with its corresponding light chain, or with fragments thereof retaining the binding affinities of the whole molecule.

22. A fragment of an antibody molecule according to any of the preceding claims which is either the F(ab')2 fragment of the whole molecule or a fragment of the whole molecule lacking only that Fc region which is that of the heavy chain having a T-cell binding affinity.

23. A composition comprising an antibody molecule or fragment thereof according to any of Claims 1 to 22, together with a physiologically acceptable diluent or carrier.

24. An antibody molecule or fragment thereof according to any of Claims 1 to 22, for use in surgery, therapy or diagnosis.

25. The use of an antibody molecule or fragment thereof according to any of Claims 1 to 22 for the manufacture of a medicament for use in the treatment of a neoplastic, viral or parasitic disease.

26. A method for aiding the regression and palliation of a neoplastic, viral or parasitic disease which comprises administering to a patient in need thereof an amount therapeutically effective in achieving such regression and palliation of an antibody molecule or fragment thereof according to any of Claims 1 to 22.

27. A process for the preparation of a bi-specific antibody molecule or fragment thereof according to any of Claims 1 to 22, which comprises culturing a hybridoma which expresses said bi-specific antibody molecule in order to produce said molecule and thereafter, where appropriate, treating this to produce a fragment thereof.

28. A bi-specific antibody molecule whenever produced according to the process of Claim 27, or an obvious equivalent thereof.