Altered antibodies

[0001] This is a continuation-in-part of application Ser. No. 903776.

[0002] 1. Field of the Invention

[0003] The present invention relates to altered antibodies in which at least part of the complementarity determining regions (CDRs) in the light or heavy chain variable domains of the antibody have been replaced by analogous parts of CDRs from an antibody of different specificity. The present invention also relates to-methods for the production of such altered antibodies. The term altered antibody is used herein to mean an antibody in which at least one residue of the amino acid sequence has been varied as compared with the sequence of a naturally occuring antibody.

[0004] 2. Descripton of the Prior Art

[0005] Natural antibodies, or immunoglobulins, comprise two heavy chains linked together by disulphide bonds and two light chains, each light chain being linked to a respective heavy chain by disulphide bonds. The general structure of an antibody of class IgG (ie an immunoglobulin (Ig) of class gamma (G)) is shown schematically in FIG. 1 of the accompanying drawings.

[0006] Each heavy chain has at one end a variable domain followed by a number of constant domains. Each light chain has a variable domain at one end and a constant domain at its other end, the light chain variable domain being aligned with the variable domain of the heavy chain and the light chain constant domain being aligned with the first constant domain of the heavy chain. The constant domains in the light and heavy chains are not involved directly in binding the antibody to the antigen.

[0007] Each pair of light and heavy chains variable domains forms an antigen binding site. The variable domains of the light and heavy chains have the same general structure and each domain comprises four framework regions, whose sequences are relatively conserved, connected by three hypervariable or complementarity determining regions (CDRs) (see Kabat, E. A., Wu, T. T., Bilofsky, H., Reid-Miller, M. and Perry, H., in Sequences of Proteins of Immunological Interest, US Dept. Health and Human Services, 1983 and 1987). The four framework regions largely adopt a beta-sheet conformation and the CDRs form loops connecting, and in some cases forming part of, the beta-sheet structure. The CDRs-are held in close proximity by the framework regions and, with the CDRs from the other variable domain, contribute to the formation of the antigen binding site.

[0008] For a more detailed account of the structure of variable domains, reference may be made to: Poljak, R. J., Amzel, L. M., Avey, H. P., Chen, B. L., Phizackerly, R. P. and Saul, F., PNAS USA, 70, 3305-3310, 1973; Segal, D. M., Padlan. E. A., Cohen, G. H., Rudikoff, S., Potter, M. and Davies, D. R., PNAS USA, 71, 4298-4302, 1974; and Marquart, M., Deisenhofer, J., Huber, R. and Palm, W., J. Mol. Biol., 141, 369-391, 1980.

[0009] In recent years advances in molecular biology based on recombinant DNA techniques have provided processes for the production of a wide range of heterologous polypeptides by transformation of host cells with heterologous DNA sequences which code for the production of the desired products.

[0010] EP-A-0 088 994 (Schering Corporation) proposes the construction of recombinant DNA vectors comprising a ds DNA sequence which codes for a variable domain of a light or a heavy chain of an Ig specific for a predetermined ligand. The ds DNA sequence is provided with initiation and termination codons at its 5- and 3-termini respectively, but lacks any nucleotides coding for amino acids superfluous to the variable domain. The ds DNA sequence is used to transform bacterial cells. The application does not contemplate variations in the sequence of the variable domain.

[0011] EP-A-1 102 634 (Takeda Chemical Industries Limited) describes the cloning and expression in bacterial host organisms of genes coding for the whole or a part of human IgE heavy chain polypeptide, but does not contemplate variations in the sequence of the polypeptide.

[0012] EP-A-0 125 023 (Genentech Inc.) proposes the use of recombinant DNA techniques in bacterial cells to produce Igs which are analogous to those normally found in vertebrate systems and to take advantage of the gene modification techniques proposed therein to construct chimeric Igs, having amino acid sequence portions homologous to sequences from different Ig sources, or other modified forms of Ig.

[0013] The proposals set out in the above Genentech application did not lead to secretion of chimeric Igs, but these were produced as inclusion bodies and were assembled in vitro with a low yield of recovery of antigen binding activity.

[0014] The production of monoclonal antibodies was first disclosed by Kohler and Milstein (Kohler, G. and Milstein, C., Nature, 256, 495-497, 1975). Such monoclonal antibodies have found widespread use not only as diagnostic reagents (see, for example, Immunology for the 80s, Eds. Voller, A., Bartlett, A., and Bidwell, D., MTP Press, Lancaster, 1981) but also in therapy (see, for example, Ritz, J. and Schlossman, S. F., Blood, 59, 1-11, 1982).

[0015] The recent emergence of techniques allowing the stable introduction of Ig gene DNA into myeloma cells (see, for example, Oi, V. T., Morrison, S. L., Herzenberg, L. A. and Berg, P., PNAS USA, 80, 825-829, 1983; Neuberger, M. S., EMBO J., 2, 1373-1378, 1983; and Ochi, T., Hawley, R. G., Hawley, T., Schulman, M. J., Traunecker, A., Kohler, G. and Hozumi, N., PNAS USA, 80, 6351-6355, 1983), has opened up the possibility of using in vitro mutagenesis and DNA transfection to construct recombinant Igs possessing novel properties.

[0016] However, it is known that the function of an Ig molecule is dependent on its three dimensional structure, which in turn is dependent on its primary amino acid sequence. Thus, changing the amino acid sequence of an Ig may adversely affect its activity. Moreover, a change in the DNA sequence coding for the Ig may affect the ability of the cell containing the DNA sequence to express, secrete or assemble the Ig.

[0017] It is therefore not at all clear that it will be possible to produce functional altered antibodies by recombinant DNA techniques.

[0018] However, colleagues of the present Inventor have devised a process whereby chimeric antibodies in which both parts of the protein are functional can be secreted. The process, which is disclosed in International Patent Application No. PCT/GB85/00392 (WO86/01533) (Neuberger et al. and Celltech Limited), comprises:

[0019] a) preparing a replicable expression vector including a suitable promoter operably linked to a DNA sequence comprising a first part which encodes at least the variable domain of the heavy or light chain of an Ig molecule and a second part which encodes at least part of a second protein;

[0020] b) if necessary, preparing a replicable expression vector including a suitable promoter operably linked to a DNA sequence which encodes at least the variable domain of a complementary light or heavy chain respectively of an Ig molecule;

[0021] c) transforming an immortalised mammalian cell line with the or both prepared vectors; and

[0022] d) culturing said transformed cell line to produce a chimeric antibody.

[0023] The second part of the DNA sequence may encode:

[0024] i) at least part, for instance the constant domain of a heavy chain, of an Ig molecule of different species, class or subclass;

[0025] ii) at least the active portion or all of an enzyme;

[0026] iii) a protein having a known binding specificity;

[0027] iv) a protein expressed by a known gene but whose sequence, function or antigenicity is not known; or

[0028] v) a protein toxin, such a ricin.

[0029] The above Neuberger application only shows the production of chimeric antibodies in which complete variable domains are coded for by the first part of the DNA sequence. It does not show any chimeric antibodies-in which the sequence of the variable domain has been altered.

[0030] EP-A-O 173 494 (The Board of Trustees of the Leland Stanford Junior University) also concerns the production of chimeric antibodies having variable domains from one mammalian source and constant domains from another mammalian source. However, there is no disclosure or suggestion of production of a chimeric antibody in which the sequence of a variable domain has been altered: indeed, hitherto variable domains have been regarded as indivisible units.

[0031] The present invention, in a first aspect, provides an altered antibody in which at least part of a CDR in a light or heavy chain variable domain has been replaced by analogous part(s) of a CDR from an antibody of different specificity.

[0032] The determination as to what constitutes a CDR and what constitutes a framework region is made on the basis of the amino-acid sequences of a number of Igs. However, from the three dimensional structure of a number of Igs it is apparent that the antigen binding site of an Ig variable domain comprises three looped regions supported on sheet-like structures. The loop regions do not correspond exactly to the CDRs, although in general there is considerable overlap.

[0033] Moreover, not all of the amino-acid residues in the loop regions are solvent accessible and in at least one case it is known that an amino-acid residue in the framework region is involved in antigen binding. (Amit, A. G., Mariuzza, R. A., Phillips, S. E. V. and Poljak, R. J., Science, 233, 747-753, 1986).

[0034] It is also known that the variable regions of the two parts of an antigen binding site are held in the correct orientation by inter-chain non-covalent interactions. These may involve amino-acid residues within the CDRs.

[0035] Further, the three dimensional structure of CDRs, and therefore the ability to bid antigen, depends on the interaction with the framework regions: thus in some cases transplanting CDRs to a different framework might destroy antigen binding.

[0036] In order to transfer the antigen binding capacity of one variable domain to another, it may not be necessary in all cases to replace all of the CDRs with the complete CDRs from the donor variable region. It may, eg, be necessary to transfer only those residues which are accessible from the antigen binding site. In addition, in some cases it may also be necessary to alter one or more residues in the framework regions to retain antigen binding capacity: this is found to be the case with reshaped antibody to Campath 1, which is discussed below.

[0037] It may also be necessary to ensure that residues essential for inter-chain interactions are preserved in the acceptor variable domain.

[0038] Within a domain, the packing together and orientation of the two disulphide bonded beta-sheets (and therefore the ends of the CDR loops) are relatively conserved. However, small shifts in packing and orientation of these beta-sheets do occur (Lesk, A. M. and Chothia, C., J. Mol. Biol., 160, 325-342, 1982). However, the packing together and orientation of heavy and light chain variable domains is relatively conserved (Chothia, C., Novotny, J., Bruccoleri, R. and Karplus, M., J. Mol. Biol., 186, 651-653, 1985). These points will need to be borne in mind when constructing a new antigen binding site so as to ensure that packing and orientation are not altered to the deteriment of antigen binding capacity.

[0039] It is thus clear that merely by replacing at least part of one or more CDRs with complementary CDRs may not always result in a functional altered antibody. However, given the explanations set out above, it will be well within the competence of the man skilled in the art, either by carrying out routine experimentation or by trial and error testing to obtain a functional altered antibody.

[0040] Preferably, the variable domains in both the heavy and light chains have been altered by at least partial CDP replacement and, if necessary, by partial framework region replacement and sequence changing. Although the CDRs may be derived from an antibody of the same species class or even subclass as the antibody from which the framework regions are derived, it is envisaged that the CDRs will generally preferably be derived from an antibody of different species and/or from an antibody of different class or subclass.

[0041] Thus, it is envisaged, for instance, that the CDRs from a mouse antibody could be grafted onto the framework regions of a human antibody. This arrangement will be of particular use in the therapeutic use of monoclonal antibodies.

[0042] At present, if a mouse monoclonal antibody is injected into a human, the human body's immune system recognises the antibody as foreign and produces an immune response thereto. Thus, on subsequent injections of the mouse antibody into the human, its effectiveness is considerably reduced by the action of the body's immune system against the foreign antibody. In the altered antibody of the present invention, only the CDRs of the antibody will be foreign to the body, and this should minimise side effects if used for human therapy. Although, for example, human and mouse framework regions have characteristic sequences, to a first approximation there seem to be no characteristic features which distinguish human from mouse CDRs. Thus, an antibody comprised of mouse CDRs in a human framework may well be no more foreign to the body than a genuine human antibody.

[0043] Even with the altered antibodies of the present invention, there is likely to be an anti-idiotypic response by the recipient of the altered antibody. This response is directed to the antibody binding region of the altered antibody. It is believed that at least some anti-idiotype antibodies are directed at sites bridging the CDRs and the framework regions. It would therefore be possible to provide a panel of antibodies having the same partial or complete CDR replacements but on a series of different framework regions. Thus, once a first altered antibody became therapeutically ineffective, due to an anti-idiotype response, a second altered antibody from the series could be used, and so on, to overcome the effect of the anti-idiotype response. Thus, the useful life of the antigen-binding capacity of the altered antibodies could be extended.

[0044] Preferably, the altered antibody has the structure of a natural antibody or a fragment thereof. Thus, the altered antibody may comprise a complete antibody, an (Fab)₂ fragment, an Fab fragment, a light chain dimer or an Fv fragment. Alternatively, the altered antibody may be a chimeric antibody of the type described in the Neuberger application referred to above. The production of such an altered chimeric antibody can be carried out using the methods described below used in conjunction with the methods described in the Neuberger application.

[0045] The present invention, in a second aspect, comprises a method for producing an altered antibody comprising:

[0046] a) preparing a first replicable expression vector including a suitable promoter operably linked to a DNA sequence which encodes at least a variable domain of an Ic framework regions consisting at least parts of framework regions from a first antibody and CDRs comprising at least part of the CDRs from a second antibody of different specificity;

[0047] b) if necessary, preparing a second replicable expression vector including a suitable promoter operably linked to a DNA sequence which encodes at least the variable domain of a complementary Ig light or heavy chain respectively;

[0048] c) transforming a cell line with the first or both prepared vectors; and

[0049] d) culturing said transformed cell line to produce said altered antibody.

[0050] Preferably, the cell line which is transformed to produce the altered antibody is an immortalised mammalian cell line, which is advantageously of lymphoid origin, such as a myeloma, hybridoma, trioma or quadroma cell line. The cell line may also comrpise a normal lymphoid cell, such as a B-cell, which has been immortalised by transformation with a virus, such as the Epstein-Barr virus. Most preferably, the immortalised cell line is a myeloma cell line or a derivative thereof.

[0051] Although the cell line used to produce the altered antibody is preferably a mammalian cell line, any other suitable cell line, such as a bacterial cell line or a yeast cell line, may alternatively be used. In particular, it is envisaged that E. Coli derived bacterial strains could be used.

[0052] It is known that some immortalised lymphoid cell lines, such as myeloma cell lines, in their normal state secrete isolated Ig light or heavy chains. If such a cell line is transformed with the vector prepared in step a) of the process of the invention, it will not be necessary to carry out step b) of the process, provided that the normally secreted chain is complementary to the variable domain of the Ig chain encoded by the vector prepared in step a).

[0053] In general the immortalised cell line will not secrete a complementary chain, and it will be necessary to carry out step b). This step may be carried out by further manipulating the vector produced in step a) so that this vector encodes not only the variable domain of an altered antibody light or heavy chain, but also the complementary variable domain.

[0054] Alternatively, step b) is carried out by preparing a second vector which is used to transform the immortalised cell line.

[0055] The techniques by which such vectors can be produced and used to transform the immortalised cell lines are well known in the art, and do not form any part of the invention.

[0056] In the case where the immortalised cell line secretes a complementary light or heavy chain, the transformed cell line may be produced for example by transforming a suitable bacterial cell with the vector and then fusing the bacterial cell with the immortalised cell line by spheroplast fusion. Alternatively, the DNA may be directly introduced into the immortalised cell line by electroporation. The DNA sequence encoding the altered variable domain may be prepared by oligonucleotide synthesis. This requires that at least the framework region sequence of the acceptor antibody and at least the CDRs sequences of the donor antibody are known or can be readily determined. Although determining these sequences, the synthesis of the DNA from oligonucleotides and the preparation of suitable vectors is to some extent laborious, it involves the use of known techniques which can readily be carried out by a person skilled in the art in light of the teaching given here.

[0057] If it was desired to repeat this strategy to insert a different antigen binding site, it would only require the synthesis of oligonucleotides encoding the CDRs, as the framework oligonucleotides can be re-used.

[0058] A convenient variant of this technique would involve making a symthetic gene lacking the CDRs in which the four framework regions are fused together with suitable restriction sites at the junctions. Double stranded synthetic CDR cassettes with sticky ends could then be ligated at the junctions of the framework regions. A protocol for achieving this variant is shown diagrammatically in FIG. 6 of the accompanying drawings.

[0059] Alternatively, the DNA sequence encoding the altered variable domain may be prepared by primer directed oligonucleotide site-directed mutagenesis. This technique in essence involves hybridising an oligonucleotide coding for a desired mutation with a single strand of DNA containing the region to be mutated and using the signle strand as a template for extension of the oligonucleotide to produce a strand containing the mutation. This technique, in various forms, is described by: Zoller, M. J. and Smith, M., Nuc. Acids Res., 10, 6487-6500, 1982; Norris, K., Norris, F., Christainsen, L. and Fiil, N., Nuc. Acids Res., 11, 5103-5112, 1983; Zoller, M. J. and Smith, M., DNA, 3, 479-488 (1984); Kramer, W., Schughart, K. and Fritz, W.-J., Nuc. Acids Res., 10, 6475-6485, 1982.

[0060] For various reasons, this technique in its simplest form does not always produce a high frequency of mutation. An improved technique for introducing both single and multiple mutations in an M13 based vector, has been described by Carter et al. (Carter, P., Bedouelle H. and Winter, G., Nuc. Acids Res., 13, 4431-4443, 1985).

[0061] Using a long oligonucleotide, it has proved possible to introduce many changes simultaneously (as in Carter et al., loc. cit.) and thus single oligonucleotides, each encoding a CDR, can be used to introduce the three CDRs from a second antibody into the framework regions of a first antibody. Not only is this technique less laborious than total gene synthesis, but is represents a particularly convenient way of expressing a variable domain of required specificity, as it can be simpler than tailoring an entire V_H domain for insertion into an expression plasmid.

[0062] The oligonucleotides used for site-directed mutagenesis may be prepared by oligonucleotide synthesis or may be isolated from DNA coding for the variable domain of the second antibody by use of suitable restriction enzymes. Such long oligonucleotides will generally be at least 30 bases long and may be up to or over 80 bases in length.

[0063] The techniques set out above may also be used, where necessary, to produce the vector of part (b) of the process.

[0064] The method of the present invention is envisaged as being of particular use in reshaping human monoclonal antibodies by introducing CDRs of desired specificity. Thus, for instance, a mouse monoclonal antibody against a particular human cancer cell may be produced by techniques well known in the art. The CDRs from the mouse monoclonal antibody may then be partially or totally grated into the framework regions of a human monoclonal antibody, which is then produced in quantity by a suitable cell line. The product is thus a specifically targetted, essentially human antibody which will recognise the cancer cells, but will not itself be recognised to any significant degree, by a human's immune system, until the anti-idiotype response eventually becomes apparent. Thus, the method and product of the present invention will be of particular use in the clinical environment.

[0065] The present invention is now described, by way of example only, with reference to the accompanying drawings.

[0066] In the drawings:

[0067] FIG. 1 is a schematic diagram showing the structure of an IgG molecule,

[0068] FIG. 2 shows the amino acid sequence of the V_H domain of NEWM in comparison with the V_H domain of the BI-8 antibody;

[0069] FIG. 3 shows the amino acid and nucleotide sequence of the HuV_NP gene.

[0070] FIG. 4 shows a comparison of the results for HuV_NP-IgE and MoV_NP-lgE in binding inhibition assays;

[0071] FIG. 5 shows the structure of three oligonucleotides used for site directed mutagenesis;

[0072] FIG. 6 shows a protocol for the construction of CDR replacements by insertion of CDR cassettes into a vector containing four framework regions fused together;

[0073] FIG. 7 shows the sequence of the variable domain of antibody D1.3 and the gene coding therefor;

[0074] FIG. 8 shows a protocol for the cloning of the D1.3 variable domain gene;

[0075] FIG. 9 illustrates nucleic acid and amino acid sequences of the variable domains of antibodies to Campath-1, with FIG. 9a representing the heavy chain and FIG. 9b representing the light chain;

[0076] FIG. 10 illustrates the sequence of the HuVLLYS gene and derived amino acid sequence;

[0077] FIG. 11 illustrates the sequences of the HuVLLYS gene and derived amino acid sequence, with asterisks marking the CDRs;

[0078] FIG. 12 illustrates a strategy for producing a reshaped human antibody having rat CDRs;

[0079] FIG. 13 illustrates loop Phe 27 to Tyr 35 in the heavy chain variable domain of the human myeloma protein KOL;

[0080] FIG. 14 illustrates the results of complement lysis and ADCC for various antibodies;

[0081] FIG. 15 illustrates the results of complement lysis and ADCC of various further antibodies;

[0082] FIG. 16A to D are 4 graphs of fluorescence emission spectra of mouse and humanised anti-lysozyme antibody in the presence of two equivalents of lysozyme;

[0083] FIG. 17 is a graph illustrating spectral change at fixed wavelength as a function of lysozyme concentration on titration of antibody samples;

[0084] FIG. 18 illustrates the plasmid for expression of the Fv fragment of a reshaped anti-lysozyme antibody;

[0085] FIG. 19 illustrates the results of SDS acrylamide (16%) gel analysis of the Fv fragments and other units;

[0086] FIG. 20 illustrates the results of native acrylamide (8%) gel analysis at pH 7.5 of the Fv fragments and other units; and

[0087] FIG. 21 illustrates the results of native acrylamide (8%) gel analysis at pH4 of the Fv fragments and other units.

[0088] This example shows the production of an altered antibody in which the variable domain of the heavy chains comprises the framework regions of a human heavy chain and the CDRs from a mouse heavy chain.

[0089] The framework regions were derived from the human myeloma heavy chain NEWM, the crystallographic structure of which is known (see Poljak et al., loc. cit. and Bruggemann, M., Radbruch, A., and Rajewsky, K., EMBO J., 1, 629-634, 1982.)

[0090] The CDRs were derived from the mouse monoclonal antibody B1-0.8 (see Reth et al., loc. cit.), which binds the hapten NP-cap (4-hydroxy-3-nitrophenyl acetyl-caproic acid: K_NP-CAP1.2 uM).

[0091] A gene encoding a variable domain HuV_NP comprising the B1-8 CDRs and the NEWM framework regions, was constructed by gene synthesis as follows.

[0092] The amino acid sequence of the V_H domain of NEWM is shown in FIG. 2, wherein it is compared to the amino acid sequence of the V_H domain of the B1-8 antibody. The sequence is divided into framework regions and CDRs according to Kabat et al. (loc. cit.). Conserved residues are marked with a line.

[0093] The amino acid and nucleotide sequence of the HuV_NP gene, in which the CDRs from the B1-8 antibody alternate with the framework regions of the NEWM antibody, is shown in FIG. 3. The HuV_NP gene was derived by replacing sections of the MOV_NP gene in the vector pSV-V_NP (see Neuberger, M. S., Williams, G. T., Mitchell, E. B., Jouhal, S., Flanagan, J. G. and Rabbitts, T. H., Nature, 314, 268-270, 1985) by a synthetic fragment encoding the HuV_NP domain. Thus the 5 and 3 non-encoding sequences, the leader sequence, the L-V intron, five N-terminal and four C-terminal amino acids are from the MoV_NP gene and the rest of the coding sequence is from the synthetic HuV_NP fragment.

[0094] The oligonucleotides from which the HuV_NP fragment was assembled are aligned below the corresponding portion of the HuV_NP gene. For convenience in cloning, the ends of oligonucleotides 25 and 26b form a Hind II site followed by a Hind III site, and the sequences of the 25/26b oligonucleotides therefore differ from the HuV_NP gene.

[0095] The HuV_NP synthetic fragment was built as a PstI-Hind III fragment. The nucleotide sequence was derived from the protein sequence using the computer programme ANALYSEQ (Staden, R., Nuc. Acids. Res., 12, 521-538, 1984) with optimal codon usage taken from the sequences of mouse constant domain genes. The oligonucleotides (1 to 26b, 28 in total) vary in size from 14 to 59 residues and were made on a Biosearch SAM or an Applied Biosystems machine, and purified on 8M-urea polyacrylamide gels (see Sanger, F. and Coulson, A., FEBS Lett., 107-110, 1978).

[0096] The oligonucleotides were assembled in eight single stranded blocks (A-D) containing oligonucleotides

[0097] 1,3,5,7 (Block A), 2,4,6,8 (block A), 9,11,13a,13b (Block B), 10a, 10b,12/14 (block B), 15,17 (block C), 16,18 (block C), 19, 21, 23, 25 (block D) and 20, 22/24, 26a, 26b (block D).

[0098] In a typical assembly, for example of block A, 50 pmole of oligonucleotides 1,3,5 and 7 were phosphorylated at the 5 end with T4 polynucleotide kinase and mixed together with 5 pmole of the terminal oligonucleotide 1 which had been phosphorylated with 5 uCi gamma-³²-P ATP (Amersham 3000 Ci/mmole). These oligonucleotides were annealed by heating to 80 C. and cooling over 30 minutes to room temperature, with unkinased oligonucleotides 2, 4 and 6 as splints, in 150 ul of 50 mM Tris.Cl, ph 7.5, 10 mM MgCl₂. For the ligation, ATP (1 mM) and DTT (10 mM) were added with 50 U T4 DNA ligase (Anglian Biotechnology Ltd.) and incubated for 30 minutes at room temperature. EDTA was added to 10 mM, the sample was extracted with phenol, precipitated from ethanol, dissolved in 20 ul water and boiled for 1 minute with an equal volume of formamide dyes. The sample was loaded onto and-run on a 0.3 mm 8M-urea 10% polyacrylamide gel. A band of the expected size was detected by autoradiography and eluted by soaking.

[0099] Two full length single strands were assembled from blocks A to D and A to D using splint oligonucleotides. Thus blocks A to D were annealed and ligated in 30 ul as set out in the previous paragraph using 100 pmole of olignucleotides 10a, 16 and 20 as splints. Blocks A to D were ligated using oligonucleotides 7, 13b and 17 as splints.

[0100] After phenol/ether extraction, block A-D was annealed with block A-D, small amounts were cloned in the vector M13mp18 (Yanish-Perron, C., Vieria, J. and Messing, J., Gene, 33, 103-119, 1985) cut with PstI and Hind III, and the gene sequenced by the dideoxy technique (Sanger, F., Nicklen, S. and Coulson, A. R., PNAS USA, 74, 5463-5467, 1977).

[0101] The MoV_NP gene was transferred as a Hind III-BamHI fragment from the vector pSV-V_NP (Neuberger et al., loc. cit.) to the vector M13 mp8 (Messing, J. and Vieria, J., Gene, 19, 269-276, 1982). To facilitate the replacement of MoV_NP coding sequences by the synthetic HuV_NP fragment, three Hind II sites were removed from the 5 non-coding sequence by site directed mutagenesis, and a new Hind II site was subsequently introduced near the end of the fourth framework region (FR4 in FIG. 2). By cutting the vector with PstI and Hind II, most of the V fragment can be inserted as a PstI-Hind II fragment. The sequence at the Hind II site was corrected to NEWM FR4 by site directed mutagenesis.

[0102] The Hind III-Bam HI fragment, now carrying the HuV_NP gene, was excised from M13 and cloned back into pSV-V_NP to replace the MoV_NP gene and produce a vector pSV-HuV_NP. Finally, the genes for the heavy chain constant domains of human Ig E (Flanagan, J. G. and Rabbitts, T. H., EMBO J., 1, 655-660, 1982) were introduced as a Bam HI fragment to give the vector pSV-HuV_NP.HE. This was transfected into the mouse myeloma line J558 L by spheroplast fusion.

[0103] The sequence of the HuV_NP gene in pSV-HuV_NP.HE was checked by recloning the Hind III-Bam HI fragment back into M13 mp8 (Messing et al., loc. cit.). J558L myeloma cells secrete lambda 1 light chains which have been shown to associate with heavy chains containing the MoV_NP variable domain to create a binding site for NP-cap or the related hapten NIP-Cap (3-iodo-4-hydroxy-5-nitrophenylacetyl-caproic acid) (Reth, M., Hammerling, G. J. and Rajewsky, K., Eur. J. Immunol., 8, 393-400, 1978).

[0104] As the plasmid pSV-HuV_NP.HE contains the gpt marker, stably transfected myeloma cells could be selected in medium containing mycophenolic acid. Transfectants secreted an antibody (Huv_NP-IgE) with heavy chains comprising a HuV_NP variable domain (ie a humanised mouse variable region) and human epsilon constant domains, and lambda 1 light chains from the J558L myeloma cells.

[0105] The culture supernatants of several gpt clones were assayed by radioimmunoassay and found to contain NIP-cap binding antibody. The antibody secreted by one such clone was purified from culture supernatant by affinity chromatography on NIP-cap Sepharose (Sepharose is a registered trade mark). A polyacrylamide-SDS gel indicated that the protein was indistinguishable from the chimeric antibody MoV_NP-IgE (Neuberger et al., loc. cit.).

[0106] The HuV_NP-IgE antibody competes effectively with the MoV_NP-IgE for binding to both anti-human-IgE and to NIP-cap coupled to bovine serum albumin.

[0107] Various concentrations of HuV_NP-IgE and MoV_NP-IgE were used to compete with the binding of radiolabelled MoV_NP-IgE to polyvinyl microtitre plates coated with (a) Sheep anti-human-IgE antiserum (Seward Laboratories); (b) NIP-cap-bovine serum albumin; (c) Ac38 anti-idiotypic antibody; (d) Ac 146 anti-idiotypic antibody; and (e) rabbit anti-MoV_NP antiserum. Binding was also carried out in the presence of MoV_NP-IgM antibody (Neuberger, M. S., Williams, G. T. and Fox, R. O., Nature, 312, 604-608, 1984) or of JW5/1/2 which is an IgM antibody differing from the MoV_NP-IgM antibody at 13 residues mainly located in the V_H CDR2 region.

[0108] The results of the binding assays are shown in FIG. 4, wherein black circles represent HuV_NP white circles MoV_NP black squares MoV_NP-IgM and white squares JW5/1/2. Binding is given relative to the binding in the absence of the inhibitor.

[0109] The affinities of HuV_NP-IgE for NP-cap and NIP-cap were then measured directly using the fluorescence quench technique and compared to those for MoVNP-IgE, using excitation at 295 nm and observing emission at 340 nm (Eisen, H. N., Methods Med. Res., 10, 115-121,1964).

[0110] Antibody solutions were diluted to 100 nM in phosphate buffered saline, filtered (0.45 um pore cellulose acetate) and titrated with NP-cap in the range 0.2 to 20 uM. As a control, mouse DI-3 antibody (Mariuzza, R. A., Jankovic, D. L., Bulot, G., Amit, A. G., Saludjian, P., Le Guern, A., Mazie, J. C. and Poljak, R. J., J. Mol. Biol., 170, 1055-1058, 1983), which does not bind hapten, was titrated in parallel.

[0111] Decrease in the ratio of the fluorescence of HuV_NP-IgE or Huv_NP-IgE to the fluorescence of the D1-3 antibody was taken to be proportional to NP-cap occupancy of the antigen binding sites. The maximum quench was about 40% for both antibodies, and hapten dissociation constants were determined from least-squares fits of triplicate data sets to a hyperbola.

[0112] For NIP-cap, hapten concentration varied from 10 to 300 nM, and about 50% quenching of fluorescence was observed at saturation. Since the antibody concentrations were comparable to the value of the dissociation constants, data were fitted by least squares to an equation describing tight binding inhibition (Segal, I. H., in Enzyme Kinetics, 73-74, Wiley, New York, 1975).

[0113] The binding constants obtained from these data for these antibodies are shown in Table 1 below.

	TABLE 1
		KNP-cap	KNIP-cap
	MoVNP-IgE	1.2 uM	0.02 uM
	HuVNP-IgE	1.9 uM	0.07 uM

[0114] These results show that the affinities of these antibodies are similar and that the change in affinity is less than would be expected for the loss of a hydrogen bond or a van der Waals contact point at the active site of an enzyme.

[0115] Thus, it has been shown that it is possible to produce an antibody specific for an artificial small hapten, comprising a variable domain having human framework regions and mouse CDRs, without any significant loss of antigen binding capacity.

[0116] As shown in FIG. 4(d), the HuV IgE antibody has lost the MoV_NP idiotypic determinant recognised by the antibody Acl46. Furthermore, HuV_NP-IgE also binds the Ac38 antibody less well (FIG. 4(c)), and it is therefore not surprising that HuV_NP-IgE has lost many of the determinants recognised by the polyclonal rabbit anti-idiotypic antiserum (FIG. 4(e)).

[0117] It can thus be seen that, although the HuV_NP-IgE antibody has acquired substantially all the antigen binding capacity of the mouse CDRs, it has not acquired any substantial proportion of the mouse antibody's antigenicity.

[0118] The results of FIGS. 4(d) and 4(e) carry a further practical implication. The mouse (or human) CDRs could be transferred from one set of human frameworks (antibody 1) to another (antibody 2). In therapy, anti-idiotypic antibodies generated in response to antibody 1 might well bind poorly to antibody 2. Thus, as the anti-idiotyic response starts to neutralise antibody 1 treatment could be continued with antibody 2, and the CDRs of a desired specificity used more than once.

[0119] For instance, the oligonucleotides encoding the CDRs may be used again, but with a set of oligonucleotides encoding a different set of framework regions.

[0120] The above work has shown that antigen binding characteristics can be transferred from one framework to another without loss of activity, so long as the original antibody is specific for a small hapten.

[0121] It is known that small haptens generally fit into an antigen binding cleft. However, this may not be true for natural antigens, for instance antigens comprising an epitopic site on a protein or polysaccharide. For such antigens, the antibody may lack a cleft (it may only have a shallow concavity), and surface amino acid residues may play a significant role in antigen binding. It is therefore not readily apparent that the work on artificial antigens shows conclusively that CDR replacement could be used to transfer natural antigen binding properties.

[0122] Therefore work was carried out to see if CDR replacement could be used for this purpose. This work also involved using primer-directed, oligonucleotide site-directed mutagenesis using three synthetic oligonculeotides coding for each of the mouse CDRs and the flanking parts of framwork regions to produce a variable domain gene similar to the HuV_NP gene.

[0123] The three dimensional structure of a complex of lysozyme and the antilysozyme antibody D1.3 (Amit et al., loc. cit.) was solved by X-ray crystallography. There is a large surface of interaction between the antibody and antigen. The antibody has two heavy chains of the mouse IgG1 class (H) and two Kappa light chains (K), and is denoted below as H₂K₂.

[0124] The DNA sequence of the heavy chain variable region was determined by making cDNA from the mRNA of the D1.3 hybridoma cells, and cloning into plasmid and M13 vectors. The sequence is shown in FIG. 7, in which the boxed residues comprise the three CDRs and the asterisks mark residues which contact lysozyme.

[0125] Three synthetic oligonucleotides were then designed to introduce the D1.3 V_HCDRs in place of the V_HCDRs of the HUV_NP gene. The Hu_NP gene has been cloned into M13 mp8 as a BamHI-Hind III fragment, as described above. Each oligonucleotide has 12 nucleotides at the 5 end and 12 nucleotides at the 3 end which are complementary to the appropriate HuV_NP framework regions. The central portion of each oligonucleotide encodes either CDR1, CDR3, or CDR3 of the D1.3 antibody, as shown in FIG. 5, to which reference is now made. It can be seen from this Figure that these oligonucleotides are 39, 72 and 48 nucleotides long respectively.

[0126] 10 pmole of D1.3 CDR1 primer phosphorylated at the 5 end and annealed to lug of the M13-HuV template and extended with the Klenow fragment of DNA polymerase in the presence of T4 DNA ligase. After an oligonucleotide extension at 15 C., the sample was used to transfect E. Coli strain BHM71/18 mutL and plaques gridded and grown up as infected colonies.

[0127] After transfer to nitrocellulose filters, the colonies were probed at room temperature with 10 pmole of D1.3 CDR1 primer labelled at the 5 end with 30 Ci ³²P-ATP. After a 3 wash at 60 C., autoradiography revealed about 20% of the colonies had hybrdidised well to the probe. All these techniques are fully described in Oligonucleotide site-directed mutagenesis in M13 an experimental manual by P. Carter, H. Bedouelle, M. M. Y. Waye and G. Winter 1985 and published by Anglian Biotechnology Limited, Hawkins Road, Colchester, Essex CO2 8JX. Several clones were sequenced, and the replacement of HuV_NP CDR1 by D 1.3 CDR1 was confirmed. This M13 template was used in a second round of mutagenesis with D1.3 CDR2 primer; finally template with both CDRs 1&2 replaced was used in a third round of mutagenesis with D1.3 CDR3 primer. In this case, three rounds of mutganesis were used.

[0128] The variable domain containing the D1.3 CDRs was then attached to sequences encoding the heavy chain constant regions of human IgG2 so as to produce a vector encoding a heavy chain Hu*. The vector was transfected into J558L cells as above. The antibody Hu*₂L₂ is secreted.

[0129] For comparative purposes, the variable region gene for the D1.3 antibody was inserted into a suitable vector and attached to a gene encoding the constant regions of mouse IgG1 to produce a gene encoding a heavy chain H* with the same sequence as H. The protocol for achieving this is shown in FIG. 8.

[0130] As shown in FIG. 8, the gene encoding the D1.3 heavy chain V and C_H1 domains and part of the hinge region are cloned into the M13 mp9 vector.

[0131] The vector (vector A) is then cut with NcoI, blunted with Klenow polymerase and cut with PstI. The PStI-NcoI fragment is purified and cloned into PstI-HindII cut MV_NP to replace most of the MV_NP coding sequences. The ^MVNP vector comprises the mouse variable domain gene with its promoter, 5 leader, and 5 and 3 introns cloned into M13 mp9. This product is shown as vector B in FIG. 8.

[0132] Using site directed mutagenesis on the single stranded template of vector B with two primers, the sequence encoding the N-terminal portion of the C_H1 domain and the PstI site near the N-terminus of the V domain are removed. Thus the V domain of D1.3 now replaces that of V_NP to produce vector C of FIG. 8.

[0133] Vector C is then cut with HindIII and BamHI and the fragment formed thereby is inserted into HindIII/BamHI cut M13 mp9. The product is cut with Hind III and SacI and the fragment is inserted into PSV-V_NP cut with Hind III/SacI so as to replace the V_NP variable domain with the D1.3 variable domain. Mouse IgG1 constant domains are cloned into the vector as a SacI fragment to produce vector D of FIG. 8.

[0134] Vector D of FIG. 8 is transfected into J558L cells and the heavy chain H* is secreted in association with the lambda light chain L as an antibody H*₂L₂.

[0135] Separated K or L light chains can be produced by treating an appropriate antibody (for instance D1.3 antibody to produce K light chains) with 2-mercaptoethanol in guanidine hydrochloride, blocking the free interchain sulphydryls with iodoacetamide and separating the dissociated heavy and light chains by HPLC in guanidine hydrochloride.

[0136] Different heavy and light chains can be reassociated to produce functional antibodies by mixing the separated heavy and light chains, and dialysing into a non-denaturing buffer to promote re-association and refolding. Properly reassociated and folded antibody molecules can be purified on protein A-sepharose columns. Using appropriate combinations of the above procedures, the following antibodies were prepared.

	H2K2	(D1.3 antibody)
	H*2L2	(D1.3 heavy chain, lambda light chain)
	H*2K2	(recombinant equivalent of D1.3)
	Hu*2L2	(humanised D1.3 heavy chain, lambda
		light chain)
	Hu*2K2	(humanised D1.3)

[0137] The antibodies containing the lambda light chains were not tested for antigen binding capacity. The other antibodies were, and the results are shown in Table 2.

	TABLE 2
		Dissociation constant
	Antibody	for lysozyme (nM)
	D1.3 (H2K2)	14.4
	D1.3 (H2K2)	15.9, 11.4
	(reassociated)
	recombinant D1.3 (H*2K2)	9.2
	(reassociated)
	humanised D1.3 (Hu*2K2)	3.5, 3.7
	(reassociated)

[0138] The affinity of the antibodies for lysozyme was determined by fluroresecent quenching, with excitation at 290 nm and emission observed at 340 nm. Antibody solutions were diluted to 15-30 ug/mg in phosphate buffered saline, filtered (0.45 um-cellulose acetate) and titrated with hen eggwhite lysozyme. There is quenching of fluoresence on adding the lysozyme to the antibody (greater than 100% quench) and data were fitted by least squares to an equation describing tight binding inhibition (I. H. Segal in Enzyme Kinetics, p73-74, Wiley, New York 1975). This data suggests that the binding of the humanised antibody to lysozyme is tighter than in the original D1.3 antibody. Subsequent results suggest that the affinities of the humanised and mouse antibodies are both less than 5 nM with 2 mol of lysozyme molecules binding 1 mol of antibody: see Verhoeyen, M., Milstein, C. and Winter, G., Science, 239, 1534-1536 (1988). Although the work described in Verhoeyen et al. suggests that the reshaped antibody may have a weaker affinity for lysozyme than the original mouse antibody it is clear that the humanised antibody binds lysozyme effectively and with a comparable affinity to D1.3. (within a factor of 10).

[0139] Further work on fully humanised antibody to lysozyme is discussed below, in Example 4.

[0140] Further work has been carried out with an antibody to the antigen Campath-1, which is potentially of great therapeutic use, in which both light and heavy chain variable domains were reshaped. In this case, transfer of the CDRs only resulted in production of a reshaped antibody which bound poorly to the antigen as compared with the original antibody. A single mutation in the framework produced greatly enhanced binding affinity.

[0141] The Campath-1 antigen is strongly expressed on virtually all human lymphocytes and monocytes, but is absent from other blood cells including the hemopoietic stem cells (Hale, G., Bright, S., Chumbley, G., Hoang, T., Metcalf, D., Munro, A. J. & Waldmann, H. Blood 62,873-882 (1983)). A series of antibodies to Campath-1 have been produced, including rat monoclonal antibodies of IgM, IgG2a, and IgG2c isotypes (Hale, G., Hoang, T., Prospero, T., Watts, S. M. '& Waldmann, H. Mol. Biol. Med. 1,305-319 (1983)) and more recently IgG1 and IgG2b isotypes have been isolated as class switch variants from the IgG2a secreting cell line YTH 34.5HL (Hale, G., Cobbold, S. P., Waldmann, H., Easter, G., Matejtschuk, P. '& Coombs, R. R. A. J. Immunol. Meth. 103, 59-67 (1987)). All of these antibodies with the exception of the rat Ig.G2c isotype are able to lyse efficiently human lymphocytes with human complement.

[0142] In addition, the IgG2b antibody YTH 34.5HL-G2b, but not the other isotypes, is effective in antibody dependent cell mediated cytotoxicity (ADCC) with human effector cells (Hale et al, 1987, loc. cit.). These rat monoclonal antibodies have found important application in the context of immunosuppression, for control of graft-versus-host disease in bone marrow transplantation (Hale et al, 1983, loc. cit.); the management of organ rejection (Hale, G., Waldmann, H., Friend, P. '& Calne, R. Transportation 42,308-311 (1986)); the prevention of marrow rejection and in the treatment of various lymphoid malignancies (Hale, G., Swirsky, D. M., Hayhoe, F. G. J. & Waldmann, H. Mol. Biol. Med. 1,321-334 (1983)). For in-vivo use, the IgG2b antibody YTH 34.5HL-G2b seems to be the most effective at depleting lymphocytes, but the use of any of the antibodies in this group is limited by the antiglobulin response which can occur within two weeks of the initiation of treatment (Hale, Swirsky et al, 1983, loc. cit.).

[0143] The sequences of the heavy and light chain variable domains of rat IgG2a Campath-1 antibody YTH 34.5HL were determined by cloning the cDNA (FIG. 9), and the hypervariable regions were identified according to Kabat et al, loc. cit. Sequence information is given in the lower lines of FIG. 9, with the CDRs identified in boxes.

[0144] In the heavy chain variable domain there is an unusual feature in the framework region. In most known heavy chain sequences Pro(41) and Leu(45) are highly conserved: Pro(41) helps turn a loop distant from the antigen binding site and Leu(45) is in the beta bulge which forms part of the conserved packing between heavy and light chain variable domains (Chothia, C., Novotny, J., Bruccoleri, R. '& Karplus, M. J. Mol. Biol. 186, 651-663 (1985)). In YTH 34.5HL these residues are replaced by Ala(41) and Pro(45), and presumably this could have some effect on the packing of the heavy and light chain variable domains.

[0145] Working at the level of the gene and using three large mutagenic oligonucleotides for each variable domain, in a single step the hypervariable regions of YTH 34.5HL were mounted on human heavy or light chain framework regions taken from the crystallographically solved proteins NEW for the heavy chain (Saul, F. A., Amzel, M. '& Poljak, R. J. J. Biol. Chem. 253,585-597 (1978)) and from a protein based closely on the human myeloma protein REI for the light chain (Epp, 0., Colman, P., Fehlhammer, H., Bode, W., Schiffer, M. '& Huber, R. Eur. J. Biochem. 45,513-524 (1974)). The NEW light chain was not used because there is a deletion at the beginning of the third framework region of the NEW light chain. The resulting reshaped heavy chain variable domain HuVHCAMP is based on the HuVHNP gene (Kabat et al, loc. cit. and Jones, P. T., Dear, P. H., Foote, J., Neuberger, M. S. '& Winter, G. Nature 321, 522-525 (1986)) with the framework regions of human NEW alternating with the hypervariable regions of rat YTH 34.5HL. The reshaped light chain variable domain HuVLCAMP is a similar construct, except with essentially the framework regions of the human myeloma protein REI, with the C-terminal and the 3 non-coding sequence taken from a human J_k-region sequence (Hieter, P. A., Max, E. E., Seidmann, J. G., Maizel, J. V. Jr '& Leder, P. Cell 22,197-207 (1980)). Sequence information for the variable domain of the reshaped antibody is given in the upper lines in FIG. 9. The sequences of oligonucleotide primers are given and their locations on the genes are also marked in FIG. 9.

[0146] Considering the above in further detail, mRNA was purified (Kaartinen, M., Griffiths, G. M., Hamlyn, P. H., Markham, A. F., Karjalainen, K., Pelkonen J. L. T., Makela, O. '& Milstein, C. J. Immunol. 130,320-324 (1983)) from the hybridoma clone YTH 34.5HL (gamma 2a, k^b), and first strand cDNA made by priming with oligonucleotides complementary to the 5 end of the CH1 (oligonucleotide I) and the Ck exons (oligonucleotide II). cDNA was cloned and sequenced as described in Gubler, U. '& Hoffman, B. J. Gene 25, 263-269 (1983) and Sanger, F., Nicklen, S. A. '& Coulson, A. R. Proc. natl. Acad. Sci USA 74, 5463-5467 (1977).

[0147] For expression of the rat heavy chain variable domain RaVHCAMP, two restriction sites (XbaI and SalI) were introduced at each end of the cDNA clone in M13 using mutagenic oligonucleotides III and V respectively, and the XbaI-SalI fragment excised. Simultaneously, the corresponding sites were introduced into the M13-HuVHNP gene using oligonucleotides IV and VI, and the region between the sites exchanged. The sequence at the junctions was corrected with oligonucleotides VII and VIII, and an internal BamHI site removed using the oligonucleotide IX, to create the M13-RaVHCAMP gene. The encoded sequence of the mature domain is thus identical to that of YTH 34.5HL.

[0148] The reshaped heavy chain variable domain (HuVHCAMP) was constructed in an M13 vector by priming with three long oligonucleotides simultaneously on the single strand containing the M13-HuVHNP gene (see Kabat et al, loc. cit and Jones et al, loc. cit).). The mutagenesis techniques used were similar to those described in Carter et al loc. cit, using the host 71-18 mutL and without imposing strand selection. Each oligonucleotide (X, XI and XII) was designed to replace each of the hypervariable regions with the corresponding region from the heavy chain of the YTH 34.5HL antibody.

[0149] Colony blots were probed initially with the oligonucleotide X and hybridisation positives were sequenced: the overall yield of the triple mutant was 5%. Ser27 to Phe and Ser27 to Phe, Ser30 to Thr mutants (to be described below) of M13 mpB-HuVHCAMP were made with the mixed oligonucleotide XIII.

[0150] The reshaped light chain variable domain (HuVLCAMP) was constructed in an M13 vector from a gene with framework regions based on human REI. As above, three long oligonucleotides (XIV, XV, and XVI) were used to introduce the hypervariable regions of the YTH 34.5HL light chain.

[0151] Construction of the humanised light chain variable domain is described in greater detail in the following seven paragraphs.

[0152] (1) The humanised light chain variable domain (HuVLCAMP) was constructed in three stages, utilising a humanised light chain variable domain (HuVLLYS) which had been constructed for other purposes.

[0153] (a) The first stage involved the gene synthesis of a humanised light chain variable domain gene (HuVLLYS). The HuVLLYS gene incorporates human framework regions identical to the most common residue in each position in the Kabat alignment of the human kappa subgroup I, except for residues 97-108, which were identical to those in the human J1 fragment described in Heiter, P., Maizel, J '& Leder, P. J. Biol. Chem. 257, 1516-1522 (1982). The sequences of the framework regions are very similar to the crystallographically solved light chain structure REI. The CDRs in HuVLLYS were identical to those in the mouse antilysozyme antibody (D1.3) light chain (unpublished). A 30 bp sequence, identical to the sequence following the genomic JI segment, was introduced to the 3 side of residue 108. BamH1 and EcoRI restriction sites were introduced at the 3 end of the synthetic gene, and a PstI site at th 5 end. The gene synthesis of HuVLLYS is described in paragraphs (2) to (5) below, and the sequence of the gene and the derived amino acid sequence is given in FIG. 10.

[0154] (b) The second stage involved the introduction of the HuVLLYS gene in place of the heavy chain variable domain in the vector M13-MOVHNP and this is described in paragraphs 6 and 7 below. Thus the light chain variable domain utilises the promoter and signal sequence of a heavy chain variable domain: at the 3 end of the gene the sequence is derived from the human light chain J1 segment as described in paragraph (1a). The sequence of the HuVLLYS gene and the derived amino acid sequence is given in FIG. 11.

[0155] (c) The third stage involved the conversion of HuVLLYS to a humanised light chain variable domain with the CDRs of Campath-1 specifity.

[0156] 2. For the synthesis of the HuVLLYS gene, three blocks of oligonucleotides (PK1-5, KK1-5 and KE1-8 in the table in paragraph 3 below were cloned separately into M13 vectors, and sequenced. Each cloned block was excised and ligated together into M13 mp19 to create the HuVLLYS gene.

[0157] 3. Oligonucleotides listed below were produced on an Applied Biosystems 380B synthesizer. Each oligonucleotide was size-purified, 10 nmol being subjected to electrophoresis on a 2040 cm 12% polyacrylamide, 7M urea gel, eluted from the gel by dialysis against water, and lyophilized. For gene synthesis or mutagenesis, a 50 pmol aliquot of each purified oligonucleotide was phosphorylated in a 20 ul reaction mixture with 50 mM Tris-Cl (pH 8.0), 10 mM MgCl₁ 5 mM dithiothreitol, 1 mM ATP, and 5 units of polynucleotide kinase, incubated at 37 for 30 minutes. When used as hybridization probes, gel-purified oligonucleotides were phosphorylated in a similar fashion, except on a 15 pmol scale with an excess of ³²P labeled ATP.

name	sequence (5-3)
PK1	GACATCCAGATGACCCAGAGCCCAAGCAGCCTGAGCGCCAGCGTG
	GGT
PK2	GACAGAGTGACCATCACCTGTAGAGCCAGCGGTAACATCCACAAC
	TACCTGGCTTGGTAC
PK3	CAAGCCAGGTAGTTGTGGATGTTACCGCTGGCTCTACAGGTGAT
PK4	GGTCACTCTGTCACCCACGCTGGCGCTCAGGCT
PK5	GCTTGGGCTCTGGGTCATCTGGATGTCTGCA
KK1	CAGCAGAAGCCAGGTAAGGCTCCAAAGCTGCTGATCTACTACACC
	ACCA
KK2	CCCTGGCTGACGGTGTGCCAAGCAGATTCAGCGGTAGCGGTAGCG
	GTAC
KK3	CGCTACCGCTACCGCTGAATCTGCT
KK4	TGGCACACCGTCAGCCAGGGTGGTGGTGTAGTAGATCAGC
KK5	AGCTTTGGAGCCTTACCTGGCTTCTGCTGGTAC
KE1	CGACTTCACCTTCACCATCAGCAGCCTCCAGCCAGAGGACATCGC
CACCTACTACTGCC
KE2	AGCACTTCTGGAGCACCCCAAGGACGTTCGGCCAAGGGACCAAGG
TGGA
KE3	AATCAAACGTGAGTAGAATTTAAACTTTGCTTCCTCAGTTGGATC
CTAG
KE4	AATTCTAGGATCCAACTGAGGAAGCAAAGTTTAAA
KE5	TTCTACTCACGTTTGATTTCCACCTTGGTCCCTT
KE6	GGCCGAACGTCCTTGGGGTGCTCCAGAAGTGCTGGCAGTAGTAG
KE7	GTGGCGATGTCCTCTGGCTGGAGGCT
KE8	GCTGATGGTGAAGGTGAAGTCGGTAC
PK0	TCATCTGGATGTCGGAGTGGACACCT

[0158] 4. The construction of individual blocks is described for the PK1-5 block, but KK1-5 and KE1-8 blocks were cloned as KpnI-KpnI and KpnI-EcoRI fragments respectively in a similar way. 4 ul portions of each oligonucleotide PK1, PK2, PK3, PK4 and PK5, kinased as in paragraph 3, were combined and annealed at 80 C. for 5 minutes, 67 C. for 30 minutes, and allowed to cool to room temperature over the span of 30 minutes, 0.1 ul of this annealing mix was ligated with 20 ng of PstI/KpnI digested M13-mp19, in 10 ul 50 mM Tris-Cl (pH7.5), 10 mM MgCl₂, 10 mM dithiothreitol, 1 mM ATP, 120 units T4 DNA ligase (Biolabs), and incubated 12 hours at 15 C. The ligation mix was used to transfect competent E. coli strain BMH 71-18, plated with BCIG and IPTG, and a clone containing the complete PstI-KpnI insert was identified.

[0159] 5. The three cloned blocks were excised from 10 ug double-stranded replicative form of the thee M13 vectors, by digestion with PstI/KpnI (block PK1-5), KpnI (block KKI-5) and KpnI/EcoRI (block KE1-8). The inserts were separated from the vector by electrophoresis on a 2020 cm 12% polyacrylamide gel, eluted from the gel slices with 0.5 M NH₄OAc, 10 mM Mg (OAc)₂₁ 0.1 mM EDTA, 0.1% SDS, and purified by phenol extraction and ethanol precipitation. All three fragments were ligated to PstI/EcoRI cut M13-mp19. 200 white plaques were transferred by toothpick to a fresh 2TY plate, and grown as a grid of infected colonies. The plate was blotted with nitrocellulose filters, which were then treated with 0.5 M NaOH, followed by 1M Tris-Cl (pH7.5), and baked 1 hr at 80 C. under vacuum. The filters were washed at 67 C. in 3 Denhardt's solution, 2SSC, 0.07% SDS, followed by 6SSC at room temperature. Filters were then probed with the radiolabeled oligonucleotides KK3 or kK4 in 3 ml of 6SSC at 37. Following hybridization with both olignucleotides, positive colonies were picked for DNA sequencing. A phage clone containing correctly assembled blocks was designated M13-HuVLLYS.

[0160] 6. To introduce the-HuVLLYS gene in place of the heavy chain variable domain in the vector M13-MOVHNP (described in Jones et al, loc. cit) as MV with HindIII site at the 3 end of the reading frame), double-stranded replicative form DNA of phages M13-HuVLLYS and M13-MOVHNP were prepared and digested with PstI and BamHI. The insert of M13-HuVLLYS was isolated on a polyacrylamide gel, and the vector portion of M13-MOVHNP was isolated on an agarose gel. The purified fragments were ligated and transfected into E. coli strain BMH71-18, and the resulting plaques probed with oligonucleotide KK3 to identify the insert. The clone was designated M13-HuVLLYS*.

[0161] 7. In the M13-HuVLLYS* gene, to join the signal sequence of MOVHNP correctly to the 5 end of the HuVLLYS gene (at the PstI site), single stranded DNA was prepared and altered by oligonucleotide directed mutagenesis with the primer PKOsee paragraph (3) for sequence. The mutant clone was designated M13-HuVLLYS.

[0162] As previously mentioned the Campath-1 light chain variable domain was derived from the HuVLLYS domain, and the reshaped human heavy (HuVHCAMP) and light (HuVLCAMP) chain variable domains were then assembled with constant domains in three stages as illustrated in FIG. 12. In FIG. 12 sequences of rat origin are marked in black, and those of human origin in white. The recombinant heavy and light chains are also marked using a systematic nomenclature.

[0163] The illustrated procedure permits a step-wise check on the reshaping of the heavy chain variable domain (stage 1), the selection of the human isotype (stage 2), and the reshaping of the light chain variable domain and assembly of human antibody (stage 3). The vector constructions were genomic, with the variable domains excised from the M13 vectors and cloned as HindIII-BamHI fragments and the constant domains as BamHI-BamHI fragments in either pSVgpt (heavy chain) (Mulligan, R. C. '& Berg, P. Proc. natl. Acad. Sci USA 78,2072-2076 (1981)) or pSVneo (light chain) (Southern, P. J. '& Berg, P. J. Mol. Appl. Genetics 1,327-341 (1981)) vectors. The heavy chain enhancer was included to the 5 side of the variable domain, and expression of both light and heavy chains was driven from heavy chain promoter and the heavy chain signal sequence.

[0164] The human gamma 1 (Takahashi, N., Ueda, N. S., Obata, M., Nikaido, T. & Honjo, T. Cell 29,671-679 (1982)), gamma 2 (Flanagan, J. G. & Rabbits, T. H. Nature 300,709-713 (1982)), gamma 3 (Huck, S., Fort, P., Crawford, D. H., Lefranc, M.-P. & Lefranc, G. Nucl. Acid Res. 14,1779-1789 (1986), gamma 4 (Clark, M. & Waldmann, H. J. N. C. I. (in press) and K (Heiter et al, loc. cit) constant domains, and the rat gamma 2b (Bruggemann, M., Free, J., Diamond, A., Howard, J., Cobbold, S. & Waldmann, H. Proc. natl. Acad. Sci. USA 83,6075-6079 (1986)) constant domains were introduced as BamHI-BamHI fragments. The following plasmids were constructed and transfected into lymphoid cell lines by electroporation (Potter, H., Weir, L. & Leder, P. Proc. natl. Acad. Sci. USA 81,7161-7163 (1984)). In stage 1, the pSVgpt vectors HuVHCAMP-RaIgG2B, and also two mutants for reasons to be explained below, HuVHCAMP(Ser27 to Phe)-RaIgG2B, HuVHCAMP(Ser27 to Phe, Ser30 to Thr)-RaIgG2B) were introduced into the heavy chain loss variant of YTH34.5HL. In stage 2, the pSVgpt vectors RaVHCAMP-RaIgG2B, RaVHCAMP-HuIgG1, RaVHCAMP-HuIgG2, RaVHCAMP-HuIgG3, RaVHCAMP-HuIgG4 were transfected as described above. In stage 3, the pSV-gpt vector Hu(Ser27-Phe, Ser30-Thr)VHCAMP-HuIgG1 was cotransfected with the pSV-neo vector HuVLCAMP-HuIgK into the rat myeloma cell line Y0 (Y B2/3.0 Ag 20) (Galfre, G. & Milstein, C. Meth. Enzymol. 73,1-46 (1981)). In each of the three stages, clones resistant to mycophenolic acid were selected and screened for antibody production by ELISA assays. Clones secreting antibody were subcloned by limiting dilution (for Y0) or the soft agar method (for the loss variant) and assayed again before 1 litre growth in roller bottles.

[0165] Heavy Chain Variable Domain

[0166] In stage 1, the reshaped heavy chain variable domain (HuVHCAMP) was attached to constant domains of the rat isotype IgG2b and transfected into a heavy chain loss variant of the YTH34.5 hybridoma. The loss variant carries two light chains, one derived from the Y3 fusion partner (Galfre et al., loc. cit). The cloned rat heavy chain variable domain (RaVHCAMP) was also expressed as above.

[0167] Antibodies were harvested at stationary phase and concentrated by precipitation with ammonium sulphate, followed by ion exchange chromatography on a Pharmacia MonoQ column. The yields of antibody were measured by an ELISA assay directed against the rat IgG2b isotype, and each adjusted to the same concentration (Clark and Waldmann loc. cit).

[0168] The HuVHCAMP and RaVHCAMP antibodiesall of the rat IgG2b isotypewere compared in a direct binding assay to the Campath-1 antigen (obtained from a glycolipid extract from human spleen), and also in complement lysis of human lymphocytes. For measuring the binding to antigen, the partially purified Campath-1 antigen was coated onto microtitre wells. Bound antibody was detected via a biotin labelled anti-rat IgG2b monoclonal antibody (Clark & Waldmann loc. cit), developed with a streptavidin-peroxidase conjugate (Amersham plc). Complement lysis of human lymphocytes with human serum as the complement source was as described in Hale, Hoang et al (1983) loc. cit. For both binding and complement assays, the titres for the antibodies were determined by fitting the data to a sigmoid curve by a least squares iterative procedure (Hale, Hoang et al (1983) loc. cit), and the concentration of antibody giving 50% maximal binding or lysis was noted.

[0169] The results are given in Table 3.

TABLE 3
Reshaping the heavy chain variable domain
	Concentration of
	antibody in ug/ml at
	50% binding or lysis
heavy chain variable	antigen	complement
domain	binding	lysis
RaVHCAMP	0.7	2.1
HuVHCAMP	27.3	(*)
HuVHCAMP (Ser27 to Phe)	1.8	16.3
HuVHCAMP (Ser27 to Phe, Ser30 to Thr)	2.0	17.6
(*) Complement lysis with the HuVHCAMP variable domain was too weak for the estimation of lysis titre.

[0170] Compared with the original rat antibody, or the engineered equivalent, the antibody with the reshaped heavy chain domain HuVHCAMP bound poorly to the Campath-1 antigen and was weakly lytic. This suggested an error in the design of the reshaped domain.

[0171] There are several assumptions underlying the transfer of hypervariable loops from one antibody to another, and in particular that the antigen binds mainly to the hypervariable regions. These are defined as regions of sequence (Kabat et al, loc. cit) or structural (Chothia, C. & Lesk, A. J. Mol. Biol. 196,901-917 (1987)) hypervariability, and the locations of hypervariable regions are similar by either criterion, except for the first hypervariable loop of the heavy chain. By sequence the first hypervariable loop extends from residues 31 to 35 (Kabat et al, loc. cit) and by structure from residues 26 to 32 (Chothia et al, (1987) loc. cit). Residues 29 and 30 form part of the surface loop, and residue 27 which is phenylalanine or tyrosine in most sequences including YTH34.5HL, helps pack against residues 32 and 34.

[0172] By way of illustration, see FIG. 13 which illustrates loop Phe27 to Tyr35 in the heavy chain variable domain of the human myeloma protein KOL which is crystallographically solved (Marquardt, M., Deisenhofer, J., Huber, R. & Palm, W. J. Mol. Biol. 141,368-391 (1980)). The backbone of the hypervariable region according to Kabat et al, (loc. cit.) is highlighted, and a 200% van der Waal surface is thrown around Phe27 to show the interactions with Tyr32 and Met34 of the Kabat hypervariable region. In the rat YTH34.5HL heavy chain, these three side chains are conserved, but in HuVHCAMP, Phe27 is replaced by Ser: this is because, unlike most human heavy chains, in NEW the phenylalanine is replaced by serine, which would be unable to pack in the same way as phenylalanine. To restore the packing of the loop, a Ser(27) to Phe mutation was made in HuVHCAMP, and also a double mutation Ser(27) to Phe, Ser(30) to Thr (as mentioned above).

[0173] The two mutants showed a significant increase in binding to CAMPATH-1 antigen and lysed human lymphocytes with human complement. See the results given in Table 3. Thus the affinity of the reshaped antibody could be restored by altering the packing between the hypervariable regions and the framework by a single Ser(27) to Phe mutation. This suggests that alterations in the Kabat framework region can enhance the affinity of the affinity of the antibody, and extends previous work in which an engineered change in the hypervariable region yielded an antibody with increased affinity (Roberts, S., Cheetham, J. C. & Rees, A. R. Nature 328,731-734 (1987)).

[0174] Heavy Chain Constant Domains

[0175] In stage 2 (FIG. 12), the rat heavy chain variable domain was attached to constant domains of the human isotypes IgG1, 2, 3, and 4, and transfected into the heavy chain loss variant of the YTH34.5 hybridoma.

[0176] Antibody was harvested from cells in stationary phase, concentrated by precipitation with ammonium sulphate and desalted into phosphate buffered saline (PBS). Antibodies bound to the Campath-1 antigen coated on microtitre plates, were assayed in ELISA directed against the rat k light chain (Clark & Waldmann loc cit), and adjusted to the same concentration. The antibodies were assayed in complement lysis (as described above) and ADCC with activated human peripheral blood mononuclear cells (Clark & Waldmann loc. cit and Hale, G. Clark, M. & Waldmann, H. J. Immunol. 134,3056-3061 (1985)). Briefly, 510⁴ human peripheral blood cells were labelled with ⁵¹Cr and incubated for 30 minutes at room temperature with different concentrations of antibody. Excess antibody was removed and a 20 fold excess of activated cells added as effectors. After 4 hours at 37 C. death was estimated by ⁵¹Cr release.

[0177] The results are shown in FIG. 14, in which the results for rat heavy chain variable domain attached to different human isotypes are represented as follows:

	IgG1	empty squares
	IgG2	empty circles
	IgG3	solid squares
	IgG4	empty triangles

[0178] Results of lysis with the antibody YTH34.5HL are represented by solid circles.

[0179] In complement lysis (FIG. 14a), the human IgG1 isotype proved similar to the YTH34.5HL-G2b, with the human IgG3 isotype less effective. The human IgG2 isotype was only weakly lytic and the IgG4 isotype non-lytic. In ADCC (FIG. 14b) the human IgG1 was more lytic than the YTH34.5HL-G2b antibody. The decrease in lysis at higher concentration of the rat IgG2b and the human IgG1 antibody is due to an excess of antibody, which causes the lysis of effector cells. The human IgG3 antibody was weakly lytic, and the IgG2 and IgG4 isotypes were non-lytic.

[0180] The human IgG1 isotype was therefore suitable for a reshaped antibody for therapeutic use. Other recent work also suggests the IgG1 isotype as favoured for therapeutic application. When the effector functions of human isotypes were compared using a set of chimaeric antibodies with an anti-hapten variable domain, the IgG1 isotype appeared superior to the IgG3 in both complement and cell mediated lysis (Bruggemann, M., Williams, G. T., Bindon, C., Clark, M. R., Walker, M. R., Jefferis, R., Waldmann, H. & Neuberger, M. S. J. Exp. Med. (in press). Furthermore, of two mouse chimaeric antibodies directed against cell surface antigens as tumour cell markers, with human IgG1 or IgG3 isotypes, only the IgG1 isotype mediated complement lysis (Liu, A. Y., Robinson, R. R., Hellstrom, K. E., Murray, E. D. Jr., Cheng, C. P. & Hellstrom, I. Proc. natl. Acad. Sci. USA 84,3439-3443 (1987) and Shaw, D. R., Khasaeli, M. B, Sun, L. K., Ghraeyeb, J., Daddona, P. E., McKinney, S. & Lopuglio, A. F. J, Immunol. 138,4534-4538 (1987)).

[0181] Light Chain

[0182] In stage 3 (FIG. 12), the reshaped heavy chain was completed, by attaching the reshaped HuVHCAMP (Ser27 to Phe, Ser30 to Thr) domain to the human IgG1 isotype. The reshaped light chain domain HuVHCAMP was attached to the human Ck domain. The two vectors were cotransfected into the non-secreting rat Y0 myeloma line.

[0183] Antibody HuVHCAMP (Ser27 to Phe, Thr30 to Ser)-HuIGG1,) HuVLCAMP-HuIGK was purified from supernatants of cells in stationary phase by affinity chromatography on protein A Sepharose. The antibody was at least 95% (by wt) pure. The yield (about 10 mg/l) was measured spectrophotometrically. Complement and ADCC assays were performed as described in connection with FIG. 14.

[0184] The results are shown in FIG. 15, in which the results for reshaped human antibodies are represented by squares and those for rat YTH34.5HL antibodies are represented by solid circles.

[0185] The purified antibody proved almost identical to the YTH34.5HL-G2b antibody in complement lysis (FIG. 15a). In cell mediated lysis the reshaped human antibody was more reactive than the rat antibody (FIG. 15b). Similar results to the ones in FIG. 15b were obtained with three different donors of target and effector cells (data not shown). Furthermore the antibody was as effective as YTH34.5HL-G2b in killing leukaemic cells from three patients with B Cell lymphocytic leukaemia by complement mediated lysis with human serum.

[0186] The rat antibody and fully humanised antibody were compared in a direct binding assay to Campath-1 antigen. Antibody concentrations were determined as described in FIGS. 14 and 15. The amount of rat antibody bound to partially purified Campath-1 antigen was determined as described in connection with Table 3. The amount of human antibody bound was determined by an ELISA assay using a biotinylated sheep anti-human IgG antibody (Amersham).

TABLE 4
Reshaping the heavy and light chain variable domains
simultaneously
	Concentration of antibody
	in ug/ml at 50% binding
	antigen
antibody	binding
RaVHCAMP RalGG2B
RaVHCAMP RaKappa	1.01
HuVHCAMP (Ser 27 to Phe, Ser30 to Thr)
HulGGl
HuVLCAMP HuKappa	1.11

[0187] Thus by transplanting the hypervariable regions from a rodent to a human antibody of the IgG1 subtype, the antibody can be reshaped for therapeutic application.

[0188] The strategy illustrated in FIG. 12 is stepwise assembly to allow any problems to be detected at each stage (reshaping of heavy chain variable domain, selection of constant domain and reshaping of light chain variable domain). It is quite possible to build the reshaped antibody in a single step assembly, i.e. constructing the two reshaped variable domains, attaching to appropriate constant domains and cotransfecting into e.g. YO.

[0189] Following the work described in Example 2, a fully humanised anti-lysozyme antibody with reshaped heavy and light chain variable domains was constructed.

[0190] The heavy chain variable region was constructed as described in Example 2 above, and the light chain variable region was constructed as described in Example 3 above.

[0191] Heavy and light chain constructs were prepared from 1 L of bacterial culture by CsCl density gradient ultracentrifugation. 20 ug of each plasmid was digested with Pvul and co-transfected into 10⁷ NSO cells by electroporation. Transformants were selected by growth in medium containing mycophenolic acid, in a 24-well tissue culture plate. After two weeks growth, aliquots of cells were removed from each well, incubated overnight with ³⁵S-methionine, an the supernatant medium affinity adsorbed with Protein A-Sepharose beads (Pharmacia). Absorbed proteins were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), followed by autoflurography. Clones were isolated by limiting dilution from the wells which had yielded both heavy and light chain bands on the autofluorogram. The radioincorporation screening method was again employed to identify those clones secreting a complete antibody. Of these, one cell line was chosen and propagated for storage and further analysis.

[0192] A 2L culture of the cell line was grown to saturation in Dulbecco's modifed Eagle medium supplemented with 10% fetal calf serum. Antibody was concetrated from the culture medium by ammonium sulfate precipitation. The precipitate was redissolved in phosphate-buffered saline, pH 7.4(PBS), dialyzed, and chromatographed on a column of lysozyme-Sepharose (prepared by reaction of 20 mg lysozyme per ml of CNBr-activated Sepharose CL-4B). The column was washed with 0.5 M NaCl, 0.1 M Tris chloride, pH 8.5, and subsequently with 50 mM Et₂NH. Immunoglobulin-containing fractions eluting with the latter wash were identified by SDS-PAGE followed by Coomassie Blue staining; these were pooled and dialyzed against PBS. The dialyzed material was applied to a column of protein A-Sepharose. The column was washed with PBS, followed by 0.1 M citrate buffers in the order pH 6, 5, 4, 3. A peak eluting at pH 4 (the pH expected for elution of a human immunoglobulin of the gamma 2 isotype) was identified as homogeneous immunoglobulin by SDS-PAGE. This was dialyzed vs PBS for storage. Its concentration was determined spectrophotometrically using an extinction coefficient at 280 nm of 1.4 cm¹ (mg/ml)¹.

[0193] The fluorescence emission spectra of mouse and humanised antilysozyme in the presence of two equivalents of lysozyme show a loss of intensity and a hypsochromic shift relative to the calculated sum of the spectra of free antibody and free lysozyme. This quenching effect is indicative of an interaction between lysozyme and each antibody. Sets of spectra are shown in FIGS. 16A-D. Solution conditions prevailing during the measurement of these spectra were 200 nM immunoglobulin and 400 nM lysozyme (separately or in combination), in PBS at a temperature of 20 C. Spectroscopic conditions employed consisted of an excitation wavelength of 280 nm with a 5 nm bandwidth, and an emission bandwidth of 2.5 nm. Data acquisition was with a Perkin-Elmer LS-5B spectrofluorimeter interfaced to a Macintosh microcomputer, which in turn was used for data manipulation and display.

[0194] The spectral change at fixed wavelength was measured as a function of lysozyme concentration. Antibody samples were titrated in the spectroflurimeter with small aliquots of a concentrated lysozyme solution, in parallel with a control antibody, which did not interact with lysozyme, at an identical concentration. The fluorescence was determined after each addition. Titration data are shown in FIG. 17 (filled squares, humanized; open squares, mouse). The spectral change is expressed as a percent of the maximum change observed at saturation, and titrant amounts are put on a ratio scale to facilitate comparison of the two sets of data. Actual conditions for the measurements were for the humanized antibody: 200 nM, 10, 290 nm excitation, 390 nm emission; for the mouse antibody: 50 nM, 25 280 excitation, 340 emission. The titration showed an equivalence point of 1.9 binding sites per mole for the humanized antibody, and 1.8 for the mouse, extremely close to the 2 antigen binding sites expected for an immunoglobulin G. The data do not allow deduction on exact binding constant for the interaction of lysozyme and humanized antibody. However it appears to be in the range 5-50 nM.

[0195] Reshaped Fv fragments of the anti-lysozyme antibody D1.3 (Verhoeyen et al, loc. cit) were constructed. The heavy chain variable region was reshaped by combining human framework (FR) sequences from the myeloma protein NEW (Saul F. A., Amzel, M., Poljak R. J., J. Biol. Chem. 253.585 (1978)) with the mouse D1.3 CDRs which provide the antigen specifically (Verhoeyen et al, loc. cit). The reshaped light chain contains human FRs from human kappa consensus sequence (Kabat et al, loc. cit) similar to the sequence of the Bence Jones protein REI (Epp, O., et al, Eur. J. Biochem. 45, 513 (1974)) combined with the D1.3 light chain CDRs.

[0196] FIG. 18 illustrates the plasmid for the expression of the Fv-fragment of the reshaped version of the antilysozyme antibody D1.3. The plasmid was transfected by electroporation (Potter, H., Weir, L., Leder, P. Proc. Natl. Acad. Sci. USA 81,7161 (1984)) into the non-producer myeloma cell line NSO (Galfre, G., Milstein, C., Meth. Enzymol 73, 1 (1981)). Transfectants were selected with mycophenolic acid (Mulligan, R. C., Berg, P., Proc. Natl. Acad. Sci. USA 78,2072-2076).

[0197] The genes (HuVHLYS and HuVLLYS) for the VH and VL domains were produced as HindIII-BamHI fragments in M13 for the expression of the whole antibody (see M. Verhoeyen et al. Science loc. cit. for sequence of VH, see Riechmann, I., Clark, M., Waldmann, H., Winter, G., Nature in press for VL-framework sequences and see Verhoeyen, M., Berek, C., Winter, G., Nucleic Acid. Res. submitted for the VL CDRs). At the 3end of their coding sequence two stops codons followed by a SacI-site were introduced by priming with oligonucleotides I and II on the corresponding single strands. Between the RNA start site and the translation start of the leader sequence in both genes a HindIII site was introduced using oligonucleotide III. The resulting HindIII-BamHI fragments were cloned into a pSVgpt vector (Riechmann et al, Nature loc cit). The vector contains a EcoR-HindIII fragment of an Ig-heavy chain enhancer (IcH enh) as a linker. The 3 SacI-BamHI fragment of both genes was then exchanged with a SacI-BamHI fragment of the human kappa constant region (3end C_k) (Hieter, P. A. et al. Cell 22, 197 (1980)) to provide a polyadenylation signal. Into the HindIII site of both vectors a HindIII-HindIII fragment of the HCMV immediate-early gene Stenberg, R. M. et al. J. Virol 49, 190(1984), Boshart, M. et al., Cell 41, 521 (1985)) containing its enhancer, promotor and the first non-translated exon (HCMV enh-pro) were cloned. The complete VL-gene (containing Ig-enhancer, HCMV-promoter, VL-coding region and polyadenylation signal) was then subcloned as an EcoRI-fragment into pBGS18 (Spratt, B., et al., Gene 41,337 (1986)) and the resulting vector pBGS-HuVLLYS was cloned into the pSVgpt-HuVHLYS vector as a BamHI fragment as shown in FIG. 18.

[0198] The final plasmid PLRI further contained the resistance genes for the drugs ampicillin (amp^R), kanamycin (kan^R) and mycophenolic acid (Eco gpt) two col EI origins of replication (col EI ori) and the SV40 enhancer (SV40 enh pro). The BamHI (B), HindIII (H), EcoRI (E) and SacI (S) restriction sites used for cloning steps are indicated. The diagram is not to scale. Oligonucleotides I5-GAG AGG TTG GAG CTC TTA TTA TGA GGA GAC-3, II5-AAG TTT AAA GAG CTC TAC TAT TTG ATT TC-3, III5-CTC AGT AAG CTT AGA GAG A-3

[0199] Both heavy and light chain variable domains were combined in a single plasmid to facilitate the selection of transfectants using the gpt selection system (Mulligan, R. C., Berg. P., Proc. Natl. Acad. Sci. USA 78,2072,2076). Pools of transfected cell clones were analysed on SDS-acrylamide gels after ³⁵S methionine incorporation and affinity purification of culture supernatants with lysozyme Sepharose. The cloned cell line used for the preparation of Fv-fragments secreted about 8 mg/L when grown in roller bottles. Thus it is possible to produce Fv fragments in myeloma cells with yields similar to recombinant versions of intact antibodies (Neuberger, M. S., Wiliams, G. T., Fox, R. O., Nature 312,604 (1984), Riechmann, I. et al, Nature, loc. cit).

[0200] The Fv fragment contains two chains of about 12 KD (calculated values 12,749 for VH and 11,875 for VL) when analysed on SDS gels. See results in FIG. 19, in which lysozyme was run in lane 1, Fv-fragment plus lysozyme in lane 2, affinity purified Fv-fragment in lane 3, isolated VL-domain in lane 4, isolated VH-domain in lane 5) and size markers in lanes 6). The Fv-fragment and the lysozyme/Fv-fragment complex were eluted from the bands in the native gel in FIG. 20 (lanes 2,3). All samples were applied in buffer containing beta mercaptoethanol. The Fv-fragment is secreted in a functional form, as it can readily be purified from the culture supernatant with lysozyme Sepharose (Fv-fragments from cell culture supernatants were prepared by filtering through two layers of Whatmann 3 MM paper, adsorption to lysozyme coupled to CnBr-Sepharose (Pharmacia), extensive washing with phosphate buffered saline and elution with 50 mM diethylamine. Eluates were immediately adjusted to pH 7.5)

[0201] When the purified Fv-fragment was investigated on an HPLC sizing column (Biozorbax GF250) in phosphate buffered saline, only a single peak was observed and its retention time did not change between concentrations of 70 and 0.3 mg/L.

[0202] The Fv-fragment was also analysed on native acrylamide (8%) gels. See results in FIG. 20, in which lysozyme was run in lane 1, lysozyme/Fv fragment complex plus free lysozyme in lane 2, affinity purified Fv-fragment in lane 3, isolated VL-domain in lane 4 and isolated VH-domain in lane 5. Gel and running buffer contained 40 mM Tris, 8.3 mM sodium acetate, 0.4 mM Na₂EDTA and was adjusted to pH 7.6 with acetic acid. No stacking gel was used, the gel was run with reversed polarity. Here the Fv-fragment runs as a single band, that contains both the VH and the VL domain when analysed on SDS gels (compare lane 3 in FIGS. 19 and 20). This band can be shifted on the native gel, when the antigen lysozyme is added. The shifted band contains lysozyme, VH and VL domain in similar amounts when analysed on SDS-gels (compare lane 2 in FIGS. 19 and 20). Further, the isolated VL domain runs as a diffused band with a mobility different to the Fv-fragment on the native gel (lane 4, FIG. 20). The isolated VH does not run into the gel because of its net charge at pH 7.5.

[0203] (The VL and VH-domains were separated on a Mono-S column (Pharmacia) in 50 mM acetic acid, 6 M urea (adjusted to pH 4.8 with NaOH) using 0 to 0.3 M NaCl gradient over 6 minutes. The VH was sufficiently pure according to SDS gel analysis. The VL was further purified after desalting into phosphate buffered saline on a Biozorbax GF250 (DuPont) sizing column to get rid of residual VH-VL heterodimer) These results strongly suggest that the predominant form of the Fv-fragment at pH 7.5 is an associated VH-VL heterodimer. Also its apparent molecular weight in ultracentrifuge sedimentation analysis was about 23.5 kD. The same was observed with Fv-fragments obtained by proteolytic digestion (Inbar, D., Hochmann, J., Givol, D., Proc. Natl. Acad. Sci USA 69,2659 (1972), Kakimoto, K., Onoue, K., J. Immunol 112,1373 (1974), Sharon, J., Givol, D., Biochemistry 15,1591 (1976)).

[0204] The formation of VH-VL heterodimers was further established, when Fv fragments were incubated at a concentration of 0.5 mg/ml in phosphate buffered saline with 3.7% formaldehyde overnight. Crosslinked VH-VL heterodimers of about 25 kD were formed (Purified, biosynthetically ³⁵S-methionine labelled VH domain was incubated in 3.7% formaldehyde/PBS overnight in the absence or presence of excessive unlabelled VH-VL heterodimer. When analysed on SDS gels crosslinked, labelled VH VL heterodimers (molecular weight of about 25 kD) are formed from isolated labelled VH only in the prescence of unlabelled Fv-fragment. No formation of dimers could be detected in the absence of unlabelled Fv-fragment). Lysozyme-Sepharose purification of the crosslinked material showed that the crosslinked VH-VL heterodimer is still active. Overloading of SDS gels with crosslinked material also made visible a small fraction (less than 5%) of slightly lower molecular weight material suggesting the formation of crosslinked VL homodimers. No higher molecular weight band for possible VH homodimers was observed.

[0205] Nevertheless dissociation was observed when the Fv-fragment was analysed on native acrylamide gels at pH4.5. Under these conditions the VH and the VL formed each a single band see results in FIG. 21, in which lysozyme was run in lane 1, lysozyme plus Fv-fragment in lane 2, affinity purified Fv-fragment in lane 3, isolated VL-domain in lane 4 and isolated VH-domain in lane 5. Incubation of antibodies at low pH has been used historically to facilitate their proteolytic digestion (Connell, G. E., Porter, R. R, Biochem. J. 124,53P (1971)) probably reflecting the same underlying structural change.

[0206] Although the Fv-fragment is predominantly associated at neutral pH, it is in a dynamic equilibrimun; the purified biosynthetically labelled VH domain exchanges with the unlabelled VH domain when incubated with an excess of unlabelled VH-VL heterodimer, because labelled VH-VL heterodimers can be trapped by crosslinking with formaldehyde.

[0207] However, the dissociation of Fv-fragments should not cause problems in diagnostic or therapeutic applications. For structural studies, for which high protein concentrations are used Fv-fragments will certainly be of considerable advantage without further treatment. They should especially simplify the assignment of signals in NMR-spectra, if the same beta-sheet frameworks are used for Fv-fragments with different specificities.

[0208] It will of course be understood that the present invention has been described above purely by way of example, and modifications of detail can be made within the scope of the invention as defined in the appended claims.