WO1991003493A1

WO1991003493A1 - Bi-or trispecific (fab)3 or (fab)4 conjugates

Info

Publication number: WO1991003493A1
Application number: PCT/GB1990/001335
Authority: WO
Inventors: Martin John Glennie
Original assignee: The University Of Southampton
Priority date: 1989-08-29
Filing date: 1990-08-29
Publication date: 1991-03-21
Also published as: AU6290090A

Abstract

Novel trimeric and tetrameric antibodies are disclosed, including bispecific and trispecific F(ab)3 and F(ab)4 antibodies. A simple and efficient method is described for the production of pure F(ab'η)3 antibodies, in which the individual antibody Fab' fragments are joined via stable thioether linkages. Hybrid molecules were constructed from mouse monoclonal antibodies with specificities for targetting cytotoxic effectors (human peripheral blood T cells) against 51Cr-labelled chicken red blood cells. Fab' fragments from two of the chosen antibodies were first coupled via their hinge-region SH groups using o-phenylenedimaleimide (oPDM), this bispecific fragment was then linked, again via the hinge region using oPDM, to a third Fab' fragment.

Description

Bi-or trispecific (Fab)₃ or (Fab)₄ conjugates

Antibodies which are bispecific with respect to the antigen they recognize have been used successfully in a number of applications. In immunochemistry they have been used to cross-link cellular antigen and detecting agent such as ferritin or horseradish peroxidase, doing away with the antibody conjugates used in more conventional methods. Similarly they have been used as

heterobifunctional protein cross-linkers for the immobilization of enzymes in a number of assays. Perhaps their greatest potential lies in their therapeutic use for the targeting of unwanted cells or pathogens by cytotoxic effector cells or pharmacologic agents such as drugs or toxins.

It has been demonstrated that bispecific F(ab'ɣ)₂ antibodies, in which one Fab'ɣ aim is directed at a lymphoma cell and the other binds to a riboscme-inactivating protein, such as ricin A chain or saporin, can target a toxic agent to tumour cells both in vitro and in vivo and prevent further growth. For targeting effector

T cells and polymorphonuclear leukocytes, bispecific antibodies have usually have been employed which cross-link the T cell receptor complex or the Fc receptor, respectively, onto the target cell and thereby mediate high levels of specific lysis. By using the appropriate derivatives in this way, it has been possible to show that both normal circulating T cells and single clones of cytotoxic T lymphocytes can be "armed" to destroy almost any specified target cell, and that lysis is independent of the major histocompatibility complex status of the cells involved. Furthermore, the bispecific antibodies do not simply serve to "glue" the two cell populations together, but in linking the effector and target cell actually trigger the lytic process.

In addition to therapeutic uses, bispecific antibodies have also been useful as tools for understanding some of the molecular interactions which occur when resting T cells are activated for

cytotoxicity and proliferation. At this point the relationship

between activation of effectors for lysis and the activation for proliferation is not clear. As a general rule effectors, such as

T cells, are not cytotoxic when resting but become cytotoxic when profilterating. However, certain effector populations can be

activated to become cytotoxic without being driven into

proliferation. While monoclonal antibodies directed at the antigen receptor complex (TCR-CD3) on T cells can, to a limited extent at least, πύmic antigen in triggering activation for cytotoxicity and proliferation, it has been shown that bispecific antibodies with dual specificity for the TCR-CD3 and one of a group of accessory

T cell molecules, such as CD2 CD4 or CD8, are more potent in this respect.

According to a first aspect of the present invention there is

provided a trimeric or tetraireric antibody, preferably bispecific or trispecific. By antibody is meant a moiety capable of binding to one or more specific sites on one or more specific antigens.

Trimeric antibodies consist of three structurally similar arms, such as three Fab arms, linked together. Tetrameric antibodies consist of four such arms.

Bispecific antibodies of the invention consist of two arms having a first antigen specificity, the third, and fourth if present, having a

second antigen specificity. Trispecific antibodies of the invention consists of three arms having, respectively, first, second and third antigen specificities, if present, the fourth arm has the same

antigen specificity as one of the first three arms. Preferred

antibodies of the invention are F(ab)₃ or F(ab)₄ antibodies, such as a bispecific or trispecific F(ab')₃ or F(ab')₄ antibody and

particularly preferably a bispecific or trispecific F(ab'ɣ)₃ or

F(ab'ɣ)₄ antibody. By Fab' antibody is meant an Fab antibody

fragment which has been generated by pepsin cleavage of an antibody.

Mtibody specificity in antibodies according to the invention may, however, be provided by antibody fragments from any source, including genetically engineered Fv fragments, which may have engineered on multiple residues suitable for forming links between fragments. Such fragments will form Fv₃ and Fv₄ antibodies.

Antibodies of the invention may have specificities for any antigens against which antibodies can be raised or engineered. They find particular application in therapy, especially against tumour cells, but also have applications in assay techniques.

Preferably, however, at least one arm of the antibody specific for a marker on a target, which may be a target cell such as tumour cell, and at least one arm is specific for a marker on an effector, which may be an effector cell such as a T cell, lymphocyte or macrophage, or it may be another cell toxin such as a ribosome-inactivating protein, for example saporin, ricin A chain or intact ricin, or another therapeutic agent to which antibodies can be raised or engineered, such as daunomycin or adriamycin.

In the case of a trispecific antibody, the trispecificity allows it to at once bind to an effector cell and to activate it. The third arm binds to the target cell. It is preferred that two arms of the trispecific antibody are specific to T cells, one of the CD3 molecule and the other to an accessory surface molecule such as the CD2, CD4 or CD8. Alternatively, both arms may be specific for CD2. In that event, the two arms are specific for different epitopes on CD2 such as T11₂ and T11₃ or T11₂ and T11₃.

Also in accordance with the first aspect of the invention there is provided a process for the preparation of a bispecific F(ab)₃ antibody comprising:

(i) dissociating a first F(ab)₂ antibody fragment having a first specificity into its two component Fab arms;

(ii) dissociating a second F(ab)₂ antibody fragment having a second specificity into its two component Fab arms; (iii) linking the first Fab arm from step (i) to the second Fab arm from step (ii), the molar ratio of first Fab arm: a second Fab arm being 2:1, to construct a bispecific F(ab)₃ antibody.

Also in accordance with the invention there is provided a process for the preparation of a trispecific F(ab)₃ antibody as described above comprising:

(ii) dissociating a second F(ab)₂ antibody fragment having a second specificity into its two component Fab arms;

(iii) linking the first Fab arm from step (i) to the Fab arm from step (ii) to construct a bispecific F(ab)₂ antibody;

(iv) dissociating a third F(ab)₂ antibody fragment having a third specificity into its two component Fab arms; and

(v) linking the bispecific F(ab)₂ antibody from step (iii) to the Fab arm from step (iv) to give specific F(ab)₃.

Preferably, the Fab fragments are generated by treating the antibodies providing the fragments with a proteolytic enzyme such as pepsin to give a monospecific F(ab)₂ fragment. This is split by reaction with, for example, 2-merc-aptcethanol to give Fab_SH fragments, in which the -S-S- links between the Fd chains of the F(ab)₂ fragments in the original antibody have been broken and reduced to -SH groups. In the preparation Fab specific antibody, Fab_SH fragments are prepared from two antibodies to give FabAg_SH and FabB_SH fragments. The linkage of the Fabg_jj fragments is effected by treating FabA_SH fragments with o-phenylenediamine (oPDM) to give FabA_mal fragments. These are combined with untreated FabB_SH fragments, in a 2:1 weight ratio, under cross-linking conditions to give bispecific FabAAB. The product was, reduced and alkylated with 20mM 2-mercaptoethanol and 25mM iodoacetamide respectively to remove any minor products which may have formed by oxidation or disulphide exchange, before fractionating an Ultrogel AcA44. In the

preparation of trispecific antibody Fab_SH fragments are prepared from three antibodies to give FabA_SH FabB_SH and FabC_SH fragments. The linkage of the Fab_SH fragments is effected by treating FabA_SH fragments with oPDM to give FabA_mal fragments. These are combined with untreated FabB_SH fragments under cross-linking conditions to give bispecific F(ab)2AB, as shown in Fig. 1. FabC_SH fragments are similarly treated with oPDM, to yield FabC_malfragments, which are combined with the F(ab)2AB by means of an -SH group on the F(ab)2AB to give trispecific F(ab)₃ABC antibody (see Fig. 2).

According to a second aspect of the invention there is provided a conjugate comprising an antibody according to the first aspect of the invention and an effector for which at least one of the arms of the antibody is specific. The invention also contemplates a process for the preparation of such a conjugate, in which the antibody is mixed with the effector.

The invention further contemplates a pack coirprising an

antibody according to the first aspect of the invention and, separately, an effector for which at least one of the arms of the antibody is specific.

The invention will be further described with reference to the example and to the figures, in which:

Figure 1 shows the postulated reaction between two Fab fragments to produce a bispecific F(ab)₂ antibody;

Figure 2 shows the proposed structure of a trispecific antibody according to the invention;

Figure 3 shows typical chromatography profiles obtained during the preparation of bispecific F(ab'ɣ)₃ and trispecific F(ab'ɣ)₃; Figures 4 (a) and (b) are graphs showing redirected cellular cytotoxicity of ⁵¹Cr-labelled chicken red blood cells (CRBC) mediated by normal peripheral blood lymphocyte (PBL) and bispecific F(ab'ɣ)₂ antibody;

Figure 5 is a graph showing redirected cellular cytotoxicity of ⁵¹Cr-labelled CRBC by PBL and one or two bispecific

F(ab'ɣ)₂ antibodies;

Figure 6 is a graph showing redirected cellular cytotoxicity of ⁵¹Cr-labelled CRBC by bispecific F(ab'ɣ)₂ and F(ab'ɣ)₃ derivatives;

Figure 7 shows the blocking of redirected cellular

cytotoxicity mediated by bispecific F(ab'ɣ)₂ and F(ab'ɣ)₃ derivatives.

Figure 8 is a graph showing redirected cellular cytotoxicity of ⁵¹Cr-labelled CRBC by trispecific F(ab'ɣ)₃.

Figures 9(a) and (b) are graphs showing the case of blocking of trispecific antibodies;

Figure 10 is a graph showing proliferation (activation) responses of PBL to different Fab'ɣ derivatives;

Figure 11 shows proliferation responses of PBL to different

Fab'ɣ derivatives in the presence and absence of target cells;

Figure 12 shows the redirected cellular cytotoxocity of human tumour cells (Namalwa) with trispecific antibody; and Figure 13 shows the redirected cellular cytotoxocity of human tumour cells with a trispecific antibody triggering through CD2.

EXAMPLES

Materials:

All cell culture was performed in supplemented DMEM

[Dulbecco's Minimum Essential Medium containing glutamine

(200 mM), pyruvate (100 mM), penicillin and streptomycin

(100 IU/ml), fungizone (2 μg/ml) and 10% FCS (myoclone)

(Giboco Ltd, Paisley, Scotland)], or in supplemented REMI [RPMI 1640 Medium containing the same

supplements as the DMEM, but with the FCS replaced by 10% normal human serum which had been incubated at 56°C for 30 min to destroy any complement-mediated cytotoxic activity].

Antibodies:

A mouse IgGl monoclonal antibody, E₁₁C₁₂, reacting with chicken red blood cells (CRBC) was raised using conventional hybridoma

technology. BALB/c mice were immunized in a protocol which delivered CRBC (approx 10⁹) s.c. in CFA and IFA (Difco, Detroit, MI) on days 0 and 14 respectively, and i.p. in DMEM on day 24. Four days later splenic mononuclear cells were fused with the NS-1 (P3/NS-1/1-Ag4.1) mouse myeloma line at a ratio of 2:1 by using a standard somatic fusion protocol with polyethylene glycol 4000 (E.Merck, Darmstadt, Germany). Hybridoma cells secreting anti-CRBC antibody were identified by immunofIuorescence staining and flow cytometry as described previously and cloned by limiting dilution.

Additional hybridoma cell lines producing the antibodies OKTI (CD5), OKT3 (CD3) and OKT11 (CD2) were obtained from the American Type Culture Collection (ATCC, Rockville, Maryland), and the hybridoma 3G8 (CD16) was a gift from Dr. D. Segal, NIH, Bethesda, Maryland. All hybridoma cells were expanded as ascitic tumors in pristaneprimed (BALB/c × CBA) Fl mice. The 7S IgG fraction of monoclonal ascites was isolated as described by precipitation in 2 M amronium sulfate, followed by ion exchange chromatography on Trisacryl-M-DEAE (LKB-Produkter AB, Bromma, Sweden) .

Antibody F(ab'(ɣ )₂ fragments from IgG were prepared by limited proteolysis with pepsin at pH 4.1-4.2 in 0.1 M sodium acetate. The reaction being monitored at regular intervals by rapid fractionation of 100μgg samples on a GF 250 HPLC column (Zorbac), and then, when less than 10% of the IgG remained, the digestion was stopped by adjusting the pH to 8.0 with saturated Tris-base and the products fractionated on Ultrogel AcA44 (LKB). Example 1: Preparation of bispecific F(ab'ɣ)₃ antibodies:

F(ab'ɣ')₂ from the two chosen mouse antibodies at 10 mg/ml in 0.2 M Tris-HCI buffer, pH 8.0, containing 10 mM EDTA was reduced by addition of 20 mM 2-mercaptoethanol for 30 min at 30°C. At this point both reduced Fab'ɣ (Fab'ɣ_SH) samples were chilled to 4°C, a temperature which was maintained throughout the remainder of the preparation including the chromatography stages, before running through Sephadex G-25 equilibrated in a buffer of 50 irM sodium acetate, pH 5.3, containing 0.5 mM EDTA. A half volume of 12 irM o-phenylenedimaleimide (oPEM) dissolved in chilled dimethylformamide was then added to one of the two Fab'&g_jj samples. After 30 min the maleimidated Fab'o' (Fab'fø^) was separated from other solutes in the reaction mixture by passage through Sephadex G-25 equilibrated in the 50 mM sodium acetate, pH 5.3, buffer containing 0.5 mM EDTA. It was then added immediately to the Fab'ɣ_SH antibody component of the heterotrimer in a 2:1 molar ratio and concentrated to

approximately 5 mg/ml by ultrafiltration under nitrogen using a Diaflo membrane in a chilled Amicon chamber. After incubation for 18 hours, the pH of the reaction mixture was adjusted to 8 using 1 M Tris-HCl, pH 8.0. , before reducing with 2-mercaptoethanol at a final concentration of 20 mM for 30 min at 30°C. Finally the bispecific F(ab'ɣ)₃ was separated from other products and residual reactants by passage through Ultrogel AcA44 equilibrated in 0.2 M Tris-HCl. 10 mM EDTA, pH 8.0.

Example 2: Prepartion of trispecific F(ab'ɣ )₃ antibodies:

Firstly, bispecific F(ab'ɣ )₂ antibodies were made by a similar method to that of Example 1, except that the Fab' fragments of Example 1 were mixed in a 1:1 molar ratio, to give bispecific F(ab'ɣ)₂.

Bispecific F(ab'ɣ)₂ was conjugated with a Fab'ɣ_mal from a third antibody. This latter reaction relies on having free -SH groups available in the bispecific F(ab'ɣ)₂ derivative. The procedure and reaction conditions for linking the Fab'ɣ_mal to the F(ab'ɣ)_2SH were similar to those used Example 1: bispecific F(ab'δ)₂ and Fab'ɣ₂ fragments from the third antibody at 5 mg/ml in 0.2 M Tris-HCl buffer, pH 8.0, containing 10 mM EDTA were reduced by addition of 20 mM 2-mercapoethanol for 30 min at 30°C. The samples were chilled to 4°C and run through Sephadex G-25 which had been equilibrated in a buffer of 50 mM sodium acetate, pH 5.3, containing 0.5 mM EDTA. The third Fab'ɣ_SH antibody species was then maleimidated using o-PDM as in the bispecific antibody preparation and finally the bispecific F(ab'ɣ)_2SH and Fab'ɣ_malwere mixed together at a weight ratio of 1:4 for 18 hours at 4°C. Following chormatography on AcA44, 150 kDa-sized material (i.e. F(ab'ɣ)₃) was harvested and concentrated.

Trispecific F(ab'ɣ)₄ was also generated during the trispecific F(ab'ɣ)₃ preparation. It emerged from the AcA44 column at a position which corresponded to that of a protein with a molecular weight of approximately 200 kDa. This size is consistent with the joining of four Fab'ɣ fragments during the reaction. Apparently, the bispecific F(ab'ɣ)₂ has conjugated with two Fab'ɣ_mal fragments from the third antibody.

The final products in the reaction mixtures were reduced and alkylated with 20 mM 2-mercaptoethanol and 25 mM iodoacetamide respectively to remove any minor products which may have formed by oxidation or disulphide exchange, before fractionating on Ultrogel AcA44.

ese methods, the following antibodies can be prepared:

Bispecific F(ab)₃ antibodies

For targeting cytotoxic T cells to unwanted cells:

CD2 × CD2 × target*

CD3 × CD3 × target

CD5 × CD5 × target

CD3 × target × target

Trispecific F(ab)₃ antibodies

For targeting cytotoxic T cells to unwanted cells:

CD2 × CD3 × target

CD2 × CD5 × target

CD3 × CD5 × target

CD3 × LFA3 × target

CD3 × CD4 × target

CD3 × CD8 × target

CD2(T11₁) × CD2' (T11₃) × target

* target = CRBC or a tumor specific antigen For targeting cytotoxic agents to unwanted cells: toxin ** × Ig × CD19 toxin × Ig × CD22 toxin × Ig × CD37

** toxin = ribosome inactivating proteins such as saporin or ricin

The proposed structure of a trispecific F(ab'ɣ)₃ antibody of the invention specific for antigens A, B and C is shown in Fig. 2.

Although not shown, some of the ɣ-L chain disulfide bonds will be reduced during the preparation. Such reduction is known not to compromise antigen binding activity in Fab fragments. Before the final product was alkylated with iodoacetamide, one hinge-region sulphydryl (-SH) group remained, offering the potential for linking at least one more Fab'ɣ_mal fragment yielding F(ab'ɣ)₄, as mentioned above. The groups joined to the cysteinyl sulfur are: Q is carboxyamidomethyl 2, a blocking group; and R is

o-phenylenedisuccimimidyl.

Typical chromatography profiles obtained during the preparation of (a) bispecific F(ab'ɣ)₃ and (b) trispecific F(ab'ɣ)₃ derivatives are shown in Fig. 3 (a) and (b) respectively

(a) A reaction mixture containing Fab' _mal and Fab' _SH at a weight ratio of 2:1 was reduced and alkylated and then fractionated on Ultrogel AcA44 in 0.2 M Tris HCl, pH 8.0. The unreacted Fab' fragments and the bispecific F(ab'ɣ)₂ and F(ab'ɣ)₃ are indicated.

(b) A reaction mixture containing bispecific F(ab'ɣ) _{2 SH} and

Fab' _mal at a weight ratio of 1:4 was reduced and alkylated and then fractionated as for (a). Fab'ɣ, F(ab'ɣ)₂, F(ab'ɣ)₃ and putative F(ab'ɣ)₄ are indicated. Target cells:

Chicken red blood cells (CRBC) and Namalwa lymphoma cells were used as targets throughout the study. As regards the CRBC, fresh blood was collected from Rhode Island Red

Chickens into preservative-free heparin and washed in DMEM before storing in supplemented DMEM at 4ºC. For

radiolabelling, 50μl of CRBC (5×10⁸) or 200μl of Namalwa (2 × 10⁷) were first incubated in 250 μl Na₂ ⁵¹CrO₄ (Amersham International UK) for 40 min at 37ºC and then washed four times in DMEM before resuspending at 2×10⁵/ml in

supplemented DMEM.

Effector Cells:

Donors for PBL were healthy laboratory personnel in the age range of 21 to 55 years. Blood was collected into

preservative-free heparin and separated by flotation on

Ficoll-Hypaque (Lymphoprep, Nyeguard, Oslo, Norway). Cells collected at the interface were washed in phosphate-buffered saline (PBS) and resuspended in supplemented DMEM for use in cytotoxicity or proliferation assays.

Redirected cellular cytotoxicity (RCC) assay:

Cytotoxicity was measured by a standard ⁵¹Cr-release assay in 96-well, U-bottomed, microculture plates (Gibco). Each well received 50 μl of antibody diluted in supplemented DMEM, followed by 10⁴ 5¹-Cr-labeled CRBC (50 μl) or 10^{4 51}Cr-labeled Namalwa cells (50 μl) and 2×10⁵ or 5×10⁵ PBL

effectors (100 μl) respectively in supplemented DMEM. The cell mixtures were then sedimented by centrifugation (230xG for five min) before incubating at 37 ° C in a humidified atmosphere of 5% CO₂ in air for 4, 8 or 21 hours to allow lysis. Finally the cells were sedimented at 420×G for five min and 100-μl aliquots of supernatant removed to assess the release of ⁵¹Cr from target cells. Percentages of specific ⁵¹Cr release were calculated by the usual method using detergent lysis with 1% Nonidet P40 to give maximum ⁵¹Cr release.

Graphs showing redirected antibody dependent cellular cytotoxicity of ⁵¹Cr-labeled CRBC mediated by normal PBL and bispecific F(ab'ɣ)₂ antibody are shown in Fig. 4: (a) shows lytic activity of the CD3 × CRBC antibody (100 ng/ml) in 4 and 8 hour assays using fresh PBL from six healthy donors.

(b) shows lysis of CRBC using PBL from one healthy donor and various concentrations (as indicated) of CD3 × CRBC and CD16 × CRBC in 4, 8 or 21 hour assays. The assay time is indicated on each of the CD3 × CRBC titration curves, but is emitted from those of the CD16 × CRBC derivative due to their proximity.

Note that while there is considerable variation between donors, all have shown appreciably higher levels of lysis in the longer assay. With two of the donors it was only in this longer 8 hour assay that significant release of ⁵¹Cr could be measured at all. This result is confirmed and extended for one donor in Figure 4b, which shows that near maximal lytic activity was approached in eight hours with an antibody concentration of 100 ng/ml.

A graph showing redirected antibody dependent cellular cytotoxicity of ⁵¹Cr-labeled CRBC by PBL and one or more bispecific F(ab'ɣ)₂ antibodies is shown in Fig. 5. Lysis was measured in an 8 hour assay using fresh PBL from one donor and the derivatives indicated.

Results indicated that while the CD3 × CRBC F(ab'ɣ)₂ lysed CRBC at concentrations as low as 4 ng/ml, almost 1000 times more CD2 × CRBC

F(ab' )₂ was necessary to achieve appreciable levels of

cytotoxicity. Mixing the CD2 bispecific reagents

with the CD3 × CRBC antibody resulted in no additive effects, giving similar levels of killing to those seen with the CD3 derivative alone.

Fig. 6 shows that the redirected cellular cytotoxicity against ⁵¹Cr-labelled CRBC is considerably higher for bispecific F(ab'ɣ)₃ derivatives than for bispecific F(ab'#)₂ derivatives. For example, the CD3 × CD3 × CRBC trimsr was found to be up to 125 times more potent than the equivalent dimer, CD3 x CRBC, giving significant activity at concentrations below 0.1ng/ml. The efficiency of these bispecific trimers is not dependant upon two

Fab'ɣ arms being bound to the effector cells, since the derivative

CD3 × CRBC × CRBC gives a similar cytotoxicity result. The

increased efficiency probably results from the increased b-mding

avidity of the bispecific trimers. Blocking studies with excess

Fab'ɣ (Fig. 7) showed that the redirected cellular cytotoxicity

mediated by a trispecific trimer is far more readily inhibited by

blocking with the appropriate Fab'ɣ antibody than is that of a

bispecific trimer.

Fig. 7a shows a comparison of the cytotoxicity of CD2 × CRBC antibody with that of CD2 × CD2 × CRBC antibody, blocked with CD2 antibody.

Fig. 7b shows a comparison of the cytotoxicity of CD3 × CRBC

antibody with that of CD3 × CD3 × CRBC antibody, blocked with CD3

antibody.

These graphs show that redirected cellular cytotoxicity mediated by bispecific dimers is completely inhibited by the addition of 25μg/ml of the appropriate Fab' ɣ, whilst under similar conditions

bispecific trimers still showed cytotoxic activity with blocking

Fab'ɣ at 500μg/ml.

A graph showing redirected antibody dependent cellular cytotoxicity of ⁵¹Cr-labelled CRBC by trispecific F(ab'ɣ)₃ is shown in Fig. 8.

Lysis was measured in an 8 hour cytotoxicity assay using fresh PBL from one donor and the derivatives indicated.

The results show that a trispecific F(ab'ɣ)₃ derivative containing anti-CD2 and -CD3 antibody specificities was at least 100 times more potent than the best bispecific F(ab'ɣ)₂ derivative and could

promote significant levels of lysis at concentrations well below 0.1 ng/ml.

From these results, it appears that trispecific F(ab)₃ antibodies

according to the invention when used therapeutically will be

effective at doses as low as 1% of those required for similar

bispecific F(ab)₂ antibodies. This has clear advantages both in cost of treatment and directly to the patient who will take a lower dosage of antibody.

To confirm that both the arms of the trispecific antibody specific for the effector were required for full activity we performed RCC assays in the presence of blocking Fab' ɣantibody. Figure 9(a), shows the cytotoxicity of the CD2 × CD3 × CRBC antibody in the presence of, respectively, CD2 antibody, CD3 antibody and a mixture of CD2 antibody and CD3 antibody as blocker. The graph shows that when using CD2 antibody CD3 antibody at 500 μg/ml, a concentration known to be sufficient to block bispecific derivatives (Fig. 7) no reduction in the activity of this trispecific reagent was observed. It was only when both these blocking antibodies were included in the assay that any reduction in activity occurred. With Fab' from both CD2 antibody and CD3 antibody each 500μg/ml, specific ⁵¹Cr release was reduced from 65% to 15%. In a similar experiment the results of which appear in Fig. 9(b), the trispecific F(ab'ɣ)₃ antibody CD3 × CD5 × CRBC also demonstrated redirected cellular cytotoxicity which was much more resistant to blocking than a bispecific F(ab' ɣ)₂ (See Fig. 7). In this example however, the blocking was slightly more effective than that with the CD2 × CD3 × CRBC reagent. While this increased sensitivity to blocking could reflect a reduced avidity by the CD3 × CD5 × CRBC antibody, it is probably also affected by the relatively poor cytotoxic potency of this reagent.

Incorporation of [³H]thymidine:

Proliferation of normal T cells in response to mitogenic antibody derivatives was assessed in vitro. Peripheral blood lymphocytes (PBL) isolated from Ficoll-Hypaque were cultured at 37°C in 96-well, U-bottomed, microculture plates (10⁵/well) in supplemented RPMI containing the various antibody derivatives, together with or without CRBC (200 μl /well). After 48 hours each well was pulsed for 16 hours at 37°C with 1 μCi [%]thymidine (Amersham) and the incorporated radioactivity harvested onto glass microfibre filters and assessed as described previously. All experimental points were determined in triplicate. A graph showing proliferation (activation) responses of PBL to different Fab' ɣ derivatives is shown in Figs. 10 and 11. Fresh PBL from one donor were cultured for 48 hours with the Fab' 2fderivative or mixture of derivatives indicated.

The results show that in the absence of CRBC only the trispecific antibody, CD2 × CD3 × CRBC, and to a lesser extent CD3 × CD5 × CRBC, gave a significant proliferation signal. In contrast when CRBC were included in the cultures a number of antibodies, particularly

CD2 × CD2 × CRBC and a mixture of CD2 × CRBC + CD3 × CRBC were highly mitogenic.

Bispecific or trispecific F(ab)₃ antibodies with the appropriate specificities can be constructed which will first activate cellular effectors, such as T cell or monocytes, and then target them to destroy any unwanted cell. In the case of trispecific antibodies, efficient activation would appear to benefit from two antibody specificities reacting with molecules on the surface of the effector cell. The third Fab specificity of the trispecific F(ab)₃ is then available to target against any unwanted cell. Target cells in this system could include neoplastic cells, virally infected host cells (including HIV), autoreactive host cells (B or T lymphocytes) or invading pathogens, including bacteria and viruses.

The bispecific F(ab'ɣ)₃ antibodies approximately 100 times more potent than the equivalent bispecific F(ab'ɣ)₂ antibodies. For example, CD3 × CD3 × CRBC antibody was still fully active at below lng/ml. This surprising increase in performance does not appear to arise from more efficient triggering of cytotoxic T cells, at least as judged by triggering of mitosis. An F(ab'ɣ)₃ antibody containing two CRBC specific Fab'ɣ arms and one T cell specific Fab'ɣ arm also demonstrated a similar improvement in performance. It seems likely that bispecific F(ab'ɣ)₃ antibodies, because they bind to one cell surface through two Fab'o arms, couple target and effector cells together with an increased avidity over equivalent bispecific F(ab'ɣ)₂ antibody. Blocking studies were consistent with this interpretation, showing that the two arms of a bispecific F(ab'ɣ)₃ required at least 20 times more free Fab'o from the appropriate antibody to inhibit cytotoxicity than did equivalent F(ab'ɣ)₂ reagent.

In the case of trispecific antibodies, it may be advantageous to have two antibody Fab arms reacting with the target cell and one Fab arm recruiting either a cellular effector or a pharmacological agent such as a toxin. As with the effector T cell, other cells, including B cells and monocytes, can be activated when bound by two antibodies reacting with the appropriate surface molecules.

Activiated cells which show a high proliferative rate are also often more susceptible to destruction by pharmacological agents such as cytotoxic drugs and toxins; the more rapidly growing tumours, such as childhood acute lymphoblastic leukaemia, are often the most sensitive to conventional chemotherapy. Thus, a trispecific antibody, with two Fab arms directed at the target cell and one at a pharmacological agent, could first activate the target cell and then deliver a poison while it remained in a hypersensitive state.

In addition to the advantages of being able to activate effector or target cells, trispecific antibodies, because they have two Fab arms binding to one surface, also display an increased avidity for that surface. A trispecific F(ab)₃ derivative will cross-link two cell surfaces together significantly more strongly than a mixture of two bispecific F(ab)₂ antibodies. This advantage may also be applied to immunoassays including enzyme-linked immunosorbent assays or radio-immunoassays. In this situation a trispecific antibody with two different binding sites for a single antigen, such as an enzyme, protein or peptide, and a third Fab arm for a second protein, can be used as a single step cross-linker increased avidity.

Other immunoassays in which a trispecific F(ab)₃ derivatives may be useful include situations where it is necessary to capture three different immunogenic antigens, such as enzymes, proteins or peptides, into a tight immune complex. An enzyme and its substrate could be captured directly from solution onto a solid surface in this wayy Redirected cellular cytotoxicity (RCC) of human tumour cells (Namalwa) with trispecific antibody.

Fresh human blood mononuclear cells (PBL) and ⁵¹Cr-labelled Namalwa cells were mixed in a standard 20 h

cytotoxicity assay with derivatives of the shown

specificities. The differentiation antigen CD37 was used as a tumour marker for delivering effectors to the Namalwa cells. The results are shown in Figure 12 of four RCC assays, using PBL from different donors (donor 1-4). In each case the trispecific antibody [anti-CD2 × anti-CD3 × anti-CD37] is far more active than any bispecific antibody. Note that this enhance lysis is specific in that the

derivative [anti-CD2 × anti-CD3 × anti-CRBC], which is mitogenic to T cells and was highly potent in RCC against CRBC (Fig. 8), showed no cytotoxic activity (Donor 2). Also a mixture of two bispecific antibodies, [anti-CD3 x anti- CD37] + [anti-CD2 × anti-CD37], was no more active than

[anti-CD3 × anti-CD37] alone (Donor 3). These results confirm the CRBC results by showing that a derivative with a pair of antibody arms (anti-CD2 × anti-CD3) which activate cytotoxic T cells, become highly potent reagents for killing unwanted targets when converted into a trispecific antibody by the addition of an anti-target cell arm.

RCC of human tumour cells with a trispecific antibody triggering through CD2.

The experimental conditions were the same as those used in Figure 12, but using effectors from two different donors and derivatives with the specificities shown in Figure 13. The results in Figure 13 show that, for both donors a trispecific antibody derivative containing a mitogenic pair of anti-CD2 antibodies, anti-CD2 (T11₁) and anti-CD2'

(T11₃), is as, or more, active than the original trispecific antibody [anti-CD2 × anti-CD3 × anti-CD37]. Note that three other trimeric antibodies, [anti-CD3 × anti-CD3 anti-CD37], [anti-CD3 × anti-CD4 × anti-CD37] and [anti-CD3 × anti-CD5 × anti-CD37], which are not mitogenic to T cells do not show the enhance activity of the [CD2 × CD3 × CD37] and [CD2 × CD2' ×CD37] derivatives, but are probably more active than a bispecific F(ab')₂ [anti-CD3 × anti-CD37],

Claims

1. A trimeric or tetrameric antibody.

2. A bispecific or trispecific antibody according to claim 1.

3. An antibody according to any preceding claim cαrprising at least one arm specific for a marker on a first moiety and at least one arm specific for a marker on a second moiety.

4. A trispecific antibody according to any preceding claim comprising one arm specific for a marker on a first moiety and two arms specific, respectively, for different markers on a second moiety.

5. An antibody according to claim 3 or 4 in which the first and second moieties are, respectively, a target cell and an effector.

6. An antibody according to claims 3 or 4 in which the first and second moieties are, respectively, an effector and a target cell.

7. A antibody according to claim 5 or 6 in which the effector is an effector cell.

8. An antibody according to claims 5 or 6 in which the effector is a therapeutic agent effective to destroy the target cell.

9. An antibody according to claim 8, in which the therapeutic agent is a conventional chemotherapeutic compound to which

antibodies can be raised, such as daunomycin or adriamycin.

10. An antibody according to any of claims 5 to 9 in which the target cell is a tumour cell.

11. A trispecific antibody according to any of claims 1 to 5, in which two arms are specific for, respectively, TCR-CD3 and an accessory molecule on effector T cells.

12. A trispecific antibody according to any of claims 1 to 5 comprising two arms specific for, respectively, different sites on a Type 1 or Type 2 ribosome-inactivating protein and at least one arm specific for a marker on a target cell.

13. A trispecific antibody according to any of claims 1 to 4, 6 and 8 to 10 comprising two arms specific for, respectively, two different markers on a target cell and at least one arm specific for a Type 1 or Type 2 ribosome-inactivating protein.

14. A trispecific antibody according to claim 12 or 13, in which the ribosome-inactivating protein is saporin, ricin A chain or intact ricin.

15. A bispecific antibody according to any of claims 1 to 3, 5 to 7 and 10 comprising two arms specific for TCR-CD3.

16. A bispecific antibody according to any of claims 1 to 3, and 5 comprising two arms specific for a target cell.

17. An ant-ibody according to any preceding claim in which the ironomer arms are linked by a -S-(o-phenylenedisuccinimidyl)-S- group.

18. A F(ab)₃ or F(ab)₄ antibody according to any preceding claim.

19. An F(abɣ)₃ or F(abɣ)₄ antibody according to any preceding claim.

20. An F(ab')₃ or F(ab')₄ antibody according to any preceding claim.

21. An F(ab' ɣ) ₃ or F(ab'ɣ)'₄ antibody according to any preceding claim.

22. A Fv₃ or Fv₄ antibody according to any of claims 1 to 17.

23. A process for the preparation of a bispecific F(ab)₃ antibody comprising:

(i) dissociating an F(ab)₂ antibody fragment having a first specificty into its two component Fab arms;

(ii) dissociating a second F(ab)₂ antibody fragment having a second specificity into its two component Fab arms; and

(iii) linking the first Fab arm from step (i) to two component Fab arms from step (ii) to give bispecific F(ab)₃.

24. A process for the preparation of a trispecific F(ab)₃ antibody comprising:

(i) dissociating an F(ab)₂ antibody fragment having a first specificity into its two component Fab arms;

(v) linking the bispecific F(ab)₂ antibody from step (iii) to the Fab arm from step (iv) to give trispecific F(ab)₃.

25. A process according to claim 23 or 24 in which the F(ab)₂ antibody fragments used in steps (i), (ii) and (iv) are obtained by treating antibodies having the said specificities with pepsin or other proteolytic enzyme.

26. A process according to any of claims 23 to 25 in which one or more of steps (i), (ii) and (iv) are carried out by treatment of the F(ab)₂ antibody fragment with a reducing agent such as

2-mercaptoethanol.

27. A process according to any of claims 23 to 26 in which the linking of step (iii) is effected by treating the dissociated Fab fragment produced in one of steps (i) and (ii) with

o-phenyleneάimaleimide and combining the thus treated Fab fragment with the untreated Fab fragment produced in the other of steps (i) and (ii) under crrjss-linki-ng conditions to give the bispecific F(ab)₂ fragment.

28. A process according to any of claims 24 to 27 for the reporting of a trispecific F(ab)₃ antibody in which the linking of step (v) is effected by treating the Fab fragment produced in step (iv) with o-phenyleled-imaleimide and combining the thus treated Fab fragment with the bispecific F(ab)₂ fragment produced in step (iii) under cross linking conditions to give F(ab)₃.

29. A conjugate comprising a bispecific or trispecific antibody according to any of claims 1 to 22 or produced by a process according to any of claims 23 to 28 ccmplexed with an effector to which at least one arm of the antibody is specific.

30. A process for the preparation of a conjugate according to claim 29, comprising mixing the antibody with the effector.

31. A pharmaceutical composition comprising a pharmaceutically acceptable carrier or diluent and, as active ingredient, either: a bispecific or trispecific antibody according to any of claims 1 to 22; a bispecific or trispecific antibody produced by a process according to any of claims 23 to 28; or a conjugate

according to claim 29, or a conjugate produced by a process according to claim 30.

32. A pack comprising: a bispecific or trispecific antibody according to any of claim 1 to 22 or produced by a process according to any one of claims 23 to 28 and, separately therefrom, an effector for which at least one arm of the antibody is specific.