AU620568B2 - The expression of antibody fragments in protease deficient bacterial host cells - Google Patents

Description

7~ 1 i~ c r i I d AU-AI-23813/88 PCT WORLD INTELLECTUAL PROPERTY ORGANIZATION PA L IteIrnation Bure=u INTERNATIONAL APPLICATION PUBLISH U R E 5E O 8RATION TREATY (PCT) (51) International Patent Classification 4 (11) International Publication Number: WO 89/ 02465 C12N 15/00, C12P 21/00 A2 (43) International Publication Date: 23 March 1989 (23.03.89) (21) International Application Number: PCT/GB88/00747 (74) Agent: MERCER, Christopher, Paul; Carpmaels Ransford, 43 Bloomsbury Square, London WCIA (22) International Filing Date: 9 September 1988 (09.09.88) 2RA (GB).

(31) Priority Application Numbers: 8721230 (81) Designated States: AT (European patent), AU, BE (Eu- 8727431 ropean patent), CH (European patent), DE (European patent), FR (European patent), GB (European (32) Priority Dates: 9 September 1987 (09.09.87) patent), IT (European patent), JP, LU (European pa- 23 November 1987 (23.11.87) tent), NL (European patent), SE (European patent),

US.

(33) Priority Country: GB Published (71) Applicant (for all designated States except US): CELL- Without international search report and to be repu- TECH LIMITED [GB/GB]; 216 Bath Road, Slough, blished upon receipt of that report.

Berkshire SL1 4EN (GB).

(72) Inventors; and Inventors/Applicants (for US only) FIELD, Helen [GB/ GB]; 68 Banbury Road, Oxford, Oxfordshire OX2 .D J P.2 5 MAY 1989 6JP REES, Anthony, Robert [GB/GB]; 17 Narrow Hill, Witney, Oxfordshire YARRANTON, Geoffrey, Thomas [GB/GB]; 8 Sherwood Road, Win- AUSTRALIAN nersh, Nr. Wokingham, Berkshire RGl1 7NJ 17 APR k99 P A RENI OfF39 PATENT OFFK(E (54) Title: S"The expression of antibody fragments in protease deficient bacterial host cells".

(57) Abstr A process for the production of a polypeptide comprising direct expression of the polypeptide in bacterial host cells characterised in that an inducible expression system is used in combination with a protease deficient bacterial host system.

The process is particularly useful for the production of small polypeptides especially immunologically functional antibody fragments and in particular heavy and light chain polypeptides comprising complementary variable domains. Further aspects of the invention relate to suitable expression vectors and bacterial host systems.

V

i WO 89/02465 PCT/GB88/00747 "The expression of antibody fragments in protease deficient bacterial host cells".

Field of the invention This invention relates to recombinant DNA technology and in particular to processes for the production of polypeptides, especially immunologically functional immunoglobulin fragments such as immunoglobulin Fv fragments and derivatives thereof.

Background to the Invention Over recent years recombinant DNA technology has been applied to the production of a wide range of useful protein and polypeptide products. Thus the techniques of recombinant DNA genetic manipulation allow for the transformation of a host organism with a foreign gene.

The transformed host organism cells may then produce the protein or polypeptide for which the foreign gene codes.

Such modified, or transformed, host cells provide a reproducible culture source for the large scale production of polypeptides or protein using industrial fermentation techniques.

Many of the products to which recombinant DNA techniques are applied are polypeptides such as human or animal hormones. It has been found that when the polypeptide is of relatively small size, e.g. of less than about 150 amino acids in length, only low concentrations of the polypeptide product accumulate in the host cells when the product is produced as a direct expression product, e.g. as a methionine polypeptide product. Such low accumulations of product, coupled with the necessary time consuming and expensive processes of purification, render commercial working uneconomic. It appears that this low accumulation of product may be due, at least in part, to proteolytic turnover of the foreign product by the host cells.

In order to increase the yield of foreign products, such products have been produced as fusion proteins.in 'SUBSTITUTE

SHEET

S WO 89/02465 PCT/GB88/00747 -2which the polypeptide product is fused to a larger protein which is known to accumulate in the host organism. This may be achieved by ligating a gene coding for a protein known to be produced abundantly in the selected host organism, with a gene coding for the desired product, in the correct reading frame and without intervening stop codons. Examples of such abundantly produced proteins are anthranylate synthetase (Trp gene product), P-galactosidase (lacZ gene product), and chloramphenicol acetyltransferase (CAT). Such fusion proteins must be cleaved to recover the desired polypeptide. In general, the procedures used for recovery and purification of such fusion proteins and recovery of the desired products from the fusion proteins are complex and costly. Furthermore, the proteins referred to above, which are used to form a fusion protein with desired polypeptides, are large in comparison with a typical polypeptide product. The desired polypeptide therefore represents only a small percentage of the fusion protein produced by the transformed host cells, and thus the efficiency of the production process is low.

The limitations to the use of recombinant DNA technology, with regard to the production of relatively small size polypeptides, has been particularly felt in the application of recombinant DNA technology to the production of immunologically functional immunoglobulin fragments and in particular immunoglobulin Fv fragments.

For the purpose of the present description the term "immunologically functional immunoglobulin fragment" means an immunoglobulin fragment which has specific binding affinity for a corresponding antigenic determinant.

In recent years recombinant DNA technology has been WO 89/02465 PCT/GB8//00747 3 applied to the preparation of immunoglobulin (Ig) products, i.e. antibodies and parts thereof. Thus, DNA sequences coding for complete heavy and light Ig chains have been expressed in bacteria and the complete heavy and light chains reconstituted in vitro to give immunologically functional molecules (Boss et al., Nucleic Acids Research (1984), volume 12 No. 9, pages 3791-3806 and Cabilly et al., Proc. Natl. Acad. Sci. USA (1984), volume 81 pages 3273-3277).

The application of recombinant DNA technology to Ig production has opened up the possibility of producing engineered Igs or Ig fragments. For example, "chimeric antibodies" have been prepared (published European Patent Application EP-A-0171496 Research Development Corporation Japan; published European Patent Application EP-A- 0173494 Stanford University, and published International Patent Application WO 86/01533 Celltech) comprising mouse variable region and human constant region amino acid sequences. Also "altered antibodies" have been produced (International Patent Application WO 86/01533 Celltech) in which constant region domains have been replaced with a polypeptide of different function, such as an enzyme. Furthermore, "humanised" antibodies have been prepared (published European Patent Application EP-A-0239400 Winter) in which the hypervariable/complementarity determining regions of a mouse monoclonal antibody have been grafted into the variable framework regions of a human antibody.

Furthermore, it has been proposed to prepare immunologically functional Ig fragments including Fab', F(ab') 2 (published European Patent Application EP-A-0125023 Genentech) and Fv (published European Patent Application EP-A-0008994 Schering) fragments by recombinant DNA technology. Such antibody fragments have WO 89102465 PCT/GB88/00747 -4potential therapeutic uses: they lack effector functions, are less immunogenic, more soluble and potentially more susceptible to transfer across cell membranes than whole antibody molecules. Also they may be used to prepare altered antibody molecules in which the fragments are prepared, fused to other functional polypeptides, e.g.

enzymes, toxins etc., for instance, at the gene level by expression of fused DNA sequences.

Moreover, ease of production of antibody fragments permits various scientific investigations: for example, a range of mutant antibody fragments may be prepared to study and analyse the relationship between primary structure and binding site specificity. Also antibody fragments may be used for structural studies of antibodies::" Fvs may be easier to crystallise than whole antibody molecules and thus permit structure determination by X-ray crystallography. Also NMR assignments are simpler with small antibody fragments.

Published European Patent Application EP-A-0008994 (Schering Corporation) proposes the use of recombinant DNA technology for production of immunologically functional Fvs. This patent application describes the preparation of a double stranded DNA sequence which codes for a variable region of a light or heavy chain of an Ig specific for a predetermined ligand but lacking nucleotides coding for amino acid residues superfluous to said variable region.

The expression of such DNA sequences in E. coli host cell is described and it is suggested that Fv heavy and light chain products are isolated from the supernatant obtained after centrifugation of lysed host cells. However, no results of product yield or reconstitution of immunologically functional Fvs are given Hitherto, there have been no reports in the i7 7T ~~SU T:;:Z3 5 literature describing the successful production of immunologically functional Fvs using recombinant DNA techniques involving expression of light and heavy chain variable polypeptides in bacteria. It appears therefore that, to date, a satisfactory solution to the problem of producing immunologically functional Fvs in bacteria, has not been forthcoming.

It has now been found, unexpectedly, that it is possible to accumulate satisfactory levels of relatively small polypeptide products, such as immunologically functional Fvs, by direct expression of genes coding for the polypeptide products in bacterial host cells.

SUMMARY OF THE INVENTION Accordingly, in a first aspect, the present invention provides a process for the production of an immunologically functional antibody fragment, wherein heavy and light chain polypeptides having complementary variable domains are directly expressed in a protease deficient bacterial host cell using an expression system in which vector copy number is inducible.

The process of the invention may be applied to production of polypeptides in general, though is particularly advantageous for production of relatively small polypeptides. Such relatively small polypeptides are characteristically of length less than about 140 amino I acids, usually of length less than 120 amino acids and Sespecially of length from about 20 to about 60 amino acids.

I IExamples of relatively small polypeptides which may be produced by the process of the invention include calcitonin mwspe/3543 91 7 17 r i mr -1 5 a human calcitonin 33 amino acids), calcitonin gene related peptide hCGRP 37 amino acids) and epidermal growth factor hEGF 55 amino acids). Other relatively small polypeptides may comprise polypeptide fragments of larger polypeptides and proteins, o ooooo 1

I

:-fl mwspe/3543 91 7 17 WO 89/02465 PCT/GB88/00747 6 including polypeptide fragments of cell surface receptors. In particular the invention may be applied to the preparation of antibody fragments, especially immunologically functional antibody fragments.

Accordingly, in a second aspect, the-present invention provides a process for the production of an immunologically functional antibody fragment comprising expression and preferably reconstitution of heavy and light chain polypeptides comprising complementary variable domains characterised in that an inducible expression system is used in combination with a protease deficient bacterial host strain.

The process of the invention may be applied to production of antibody fragments in general, provided the fragment has binding specificity for a corresponding antigenic determinant. It will be appreciated, also, that the heavy and light chain polypeptides are preferably of sufficient heterophilicity to permit association of heterodimer molecules. Thus the process may be used for production of (Fab') 2 and Fab' antibody fragments.

Preferably, the antibody fragment is an Fv fragment.

The antibody fragments may comprise natural antibody fragments, chimeric antibody fragments, variable domains derived from one species and class of antibody and remaining Ig sequences derived from another species or class of Ig), altered antibody fragments variable Ig domains plus an additional polypeptide sequence having different, non Ig function e.g. an enzyme or toxin), "humanised" antibody fragments wherein the Ig sequence comprises hypervariable/complementarity Sdetermining regions from one species, e.g. a mouse, *i7 1 -i I 1 1 1 b -j 1 i; 1 l 1 1 u 1 1 1 1 v 1 WO 89/02465 PCT/GB88/00747 7 grafted within framework regions of another species e.g.

human), and "engineered" antibody fragments wherein the Ig amino acid sequence has been altered from the natural sequence, e.g. by site directed techniques, with a view to altering a characteristic of the molecule e.g.

antigen binding affinity or specificity for instance, along the lines described by Roberts et al., Nature (1987). The antibody fragments may comprise suitable combinations of the above types of antibody fragment.

Typically, the antibody fragments comprise substantially complete heavy and light chain variable domains which may be derived from a single or more than one species.

The antibody fragments may have any desired antigen specificity. For example, the antibody fragments may have specificity for a cell specific antigen e.g. a tumour marker, T cell marker, etc. Alternatively, the fragment may have specificity for compounds of f-herapeutic or analytical interest, e.g. fibrin, or hormones etc. In particularly preferred embodiments, the antibody fragments comprise Fv/enzyme fragments, e.g. anti-fibrin Fv/ fibrinolytic enzyme fragments, or anti tumour Fv/enzyme fragments. The specificity of the antibody fragment may be the specificity of a natural antibody, humanised antibody or engineered antibody.

The inducible expression systems used in the process of the invention are systems in which vector copy number, and preferably also expression of DNA coding for the desired polypeptide or polypeptides, is inducible by external influences, such as temperature or metabolite.

For example, the expression system may comprise a runaway replication vector of the general type described in British Patent Specification GB-B-1557774 (Alfred Benzon

A/S).

t 1 i, WO 89/02465 PCT/GB88/00747 8 Preferably, however, the expression system comprises a dual origin vector, of the type described in British Patent Specification No. GB-A-2136814. A dual origin vector is a vector comprising two replication systems; a first origin of replication resulting in a .low copy number and stable inheritance of the vector, and a second, high copy number, origin of replication at which replication is directly controllable as the result of replacement or alteration by DNA manipulation of the natural vector sequence(s) which control replication at said second origin.

Accordingly, in a third aspect, the invention provides an inducible bacterial expression vector containing a DNA sequence coding for a relatively small polypeptide or relatively small polypeptides wherein the DNA sequence is located within the vector such that it is expressed as a direct expression product.

Preferably the vector comprises DNA sequences coding for either or both a heavy and a light chain polypeptide each comprising a substantially complete variable domain.

Most preferably the heavy and light chain polypeptides are variable domain polypeptides, i.e. v H and VL polypeptides, whether or not these are expressed from the same vector.

The DNA coding for the heavy and light chain polypeptides may be in separate expression systems. Preferably DNA coding for both the heavy and light chain polypeptides are contained in the same expression system and are co-expressed. Characteristically, the polypeptides are expressed directly, i.e. as methionine polypeptides.

Expression of the polypeptides may be optimised by suitable vector adjustment. The ribosome binding site length and sequence around the AUG start codon, may be optimised for,polypeptide expression. Also the Shine Dalgarno-ATG (SD-ATG) distance may be adjusted to optimise expression levels. For example, we have found that a Xr:J WO 89/02465 -PCT/GB8/00747 9 relatively long ribosome binding site sequence (nine nucleotides) gave enhanced levels of expression of heavy chain variable polypeptide. Also we have found that SD-ATG di.stances of greater than 5, preferably 10-14, nucleotides in length are desirable.

The protease deficient bacterial host strains used in the process of the invention are typically constitutively protease deficient and in particular include constitutively protease deficient strains of E.

coli. We have.found that E. coli B host cells are not satisfactory for use in the process of the invention.

With E. coli B, unsatisfactory levels of polypeptide product accumulate in the host cells. It was a surprising finding that other protease deficient strains permit accumulation of satisfactory levels of polypeptide product.

It may be noted that E. coli B strains are "naturally" lon- and as such they do not appear to be suitable as host strains for polypeptide-production.

Recent investigations have indicated that the lon gene is present in E. coli B strains and may be expressed under certain conditions such as heat stress. Thus E. coli B strains may not be truly protease deficient but retain some protease activity.

The protease deficient strain used in this invention may comprise a lon- strain or preferably an htpR- strain, e.g. E. coli CAG 456.' More preferably a multiply protease deficient strain is used, e.g. a doubly deficient strain, including especially strains which are both htpR- and lon-, such as E. coli CAG 629 or similar strains.

Accordingly, in a fourth aspect, the invention provides protease deficient bacterial host cells transformed with a vector according to the third aspect of i; the invention.

Following expression, the polypeptide product may be recovered and further treated as desired. Following j expression, the heavy and light chain fragments may be

I

i j sU 7 :7 E WO 89/02465 PCT/GB88/00747 1 10 recovered and reconstituted in vitro using methods similar to or the same as those known in the art Boss et al., loc. cit; Cabilly et al., loc. cit). We have found that the polypeptide products typically accumulate within the host cells in the form of insoluble inclusion bodies.

The polypeptide products may be recovered as part of the insoluble fraction after cell lysis. Any suitable method including Guanidine HC1 or lysozyme treatment or French Press may be used to lyse the cells. After recovery, the insoluble polypeptide products may be solubilised, denatured and"renatured using techniques well known in the art including treatment with chaotropic agents followed by dialysis and/or treatment with alkali followed by neutralisation, e.g. as described in British Patent Specification GB-B-2138004.

In-preferred embodiments it is envisaged that co-expressed heavy and light chain fragments i.e.

expressed from the same vector, may reconstitute in vivb In accordance with the present invention we have found that the combination of inducible expression system and protease deficient bacterial host cells leads to accumulation of satisfactory levels of relatively small polypeptide products.

Preferably, a protease deficient host cell is transformed with a vector comprising DNA sequences coding for either or both a heavy and a light chain polypeptide each comprising a substantially complete variable domain.

Most preferably the heavy and light chain polypeptides are variable domain polypeptides, i.e. VH and VL polypeptides, whether or not these are expressed from the same vector.

In accordance with the present invention we have Sfound that the combination of inducible expression system 1 and protease deficient bacterial host cells leads to accumulation of sufficient levels of light and heavy chain polypeptides to permit reconstitution of immunolgically functional Ig fragments.

SCT;' 2 C, WO 89/02465 "PCT/GB88/00747, 11 The invention is further described by way of illustration only in the following examples, which refer to the accompanying diagrams. The examples describe the production of recombinant heavy and light chain immunoglobulin variable region polypeptide fragments. It will be appreciated, however, that the invention is widely applicable to the production of relatively small polypeptides in general.

Brief Description of the Diagrams This invention is further described, by way of illustration only, in the following examples, which refer to the diagrams, in which, Figure 1 illustrates construction of rFv genes from full length Gloop2 clones; mutations are shown by black boxes in the double-stranded DNA. Hatched areas represent the signal sequences. vhg2: the heavy-chain was mutated at the 3' end of the variable region making 2+2 base changes using an oligonucleotide 31 bases long and creating a TGA stop codon and a 3' EcoRl site. The EcoRl fragment including vhg2 was excised and subcloned into pAT153. The existing Bgl2-Pvu2 fragment was replaced with a linker restoring the coding sequence and restriction sites, and containing an ATG start codon, encoding formyl-Met in E. coli, vlg2: double SDM using a 35mer and a 33mer at the 5' and 3' ends of the variable coding sequence, respectively made 8 and 3+4 base changes. The growth hormone GH, was excised from pMG939 via its 5' Bgl2 and 3' EcoRl sites, and replaced by either the vhg2 or vlg2 Bgl2-EcoRl fragments, to give pMG.HF2 or pMG.LF9, respectively. pPW.HF2 contains the 9 bp SD but is otherwise identical to pMG.HF2; Figure 2 illustrates expression of the rFv genes in E. Goli; detection of V H and VL by biosynthetic labelling in cells without protease-deficiences (1B373).

Autoradiographed SDS-PAGE of either uninduced or 'i a 1 U WO 89/02465 PCT/GB88/P0747 -12induced cells labelled for 2 minute-. The degradiation of V L is apparent even after 2 minutes; Figure 3 illustrates augmenting the yield of V H with a 9-bp Shine-Dalgarno sequence 9SD); Autoradiogram of SDS-PAGE of in vitro translation products from pMG939 9GHO, pMG.HF2 with 4 bp SD and pPW.HF2 with 9 bp sd no VH products was produced in 30 minutes from pMG.HF2 in vivo, Western blotting analysis of whole cells electrophoresed on 15% SDS-PA gels: pMG.HF2 and pPW.HF2 produced in CAG529 4.5 hours after induction.

Coomassie-blue stained 15% SDS-PAGE OF 1 A 600 n m samples 6 hours after induction of pMG.HF2 and pPW.HF2 in CAG629. Coomassie-blue stained 15% SDS-PAGE of 1 A 500 nm cell extracts transformed with pMG.LF9: U time 0 after induction; at time 6 hours after induction: 629 in host strain CAG629; 456 in host strain CAG456; P htpR parent strain, SC122. VL accumulated in CAG456 to half the level found in CAG629; Figure 4 illustrates demonstration of a heterodimeric rFv complex; the A 280 nm elution profiles (below) from Superose 12 of the rFv-containing reconstitutions (upper) and hose cell reconstitutions (lower) are compared with the Pep-l binding activities obtained for each fraction (top), for rFv and controls:

V

H or VL reconstituted separately, growth hormone (GH) and host cells Protein concentrations are within a factor of 2, with respect to V H and VL. Compare the elution positions of markers: bovine serum albuin (BSA) 66 kDa, growth hormone (GH) 22.5 kDa, soybean trypsin inhibitor (SBTI) 20.1 kDa and cytochrome c (CYT C) 12.4 kDa. Most of the monomer activity resided in reconstituted VH; Figure 5 illustrates the final step in the purification of rFv (Gloop2); silver-stained

A

WO 89/02465 PCT/GB88/00747 -13- SDS-PAGE of fractions eluting from DEAE at pH 6.34 Tris), shoeing rFv eluting with high purity in the void volume (VOID), and more slowly during subsequent washing (WASH). VH appeared light yellow on silver staining, whereas VL was dark. Fractions were eluted with a 0-0.2M sodium chloride gradient and pooled (represented in lanes 1,2 and Coomassie-blue stained 15% SDS-PAGE of purified rFv fractions (lanes 1,2, where 2 is a mixture of rFv from washes and fractions eluted from DEAE at NaCl in 15mM Tris pH Lanes 3 and 4 show 2 jg and pg whole Gloop2, separated into its constituent heavy and light chains. rFv was likewise separated into its constituents, V H and VL migrating with apparent Mr of 11.5 and 9 kDa, respectively; Figure 6 illustrates purified rFv and its activity compared with Gloop2 and Fab' (Gloop2); each point is a mean of 8 data points, comprising 2 quadruplet determinations. Bars indicate the highest and lowest values obtained. Kds are obtained by a computer program using parametric and non-parametric analyses. The Kd values shown are a mean of these figures. Murine IgG (Sigma) controls showed no Pepl binding: values were a 100% at all concentrations (not shown); Figure 7 illustrates the fractionated Fv (Gloop2) reconstitutions; fractions 14 and.15 from Superose 12 fractionated Fv(Gloop2) reconstitutions, which fractions exhibit Pepl binding activity, were electrophoresed on a Phast gel system (Pharmacia). Fractions 14 and 15 from Fv prepared with (V x and without (V 3 lysozyme were compared with control fractions, 14 and 15, from fractionated reconstitutions of cells containing growth hormone (GH) or cells, using both native and reducing SDS-PAGE. A band representing Gloop2 Fv was clearly present on native PAGE. It migrated at high molecular weight, consistent with it being a dimeric complex with a size close to but

I~

WO 89/02465 PCT/GB88/00747 -14smallert-h-anrr.that of growth hormone (at 22.5kDa).

Reducing SDS-PAGE indicated the presence of VH and VL in these Pepl binding fractions, as a characteristic doublet migrating at a position consistent with a molecular weight of around 14 kDa; Figure 8 illustrates production of both heavy and light polypeptide chains; Coomassie-blue stained SDS-PAGE" of' whole cells, sampled at various times after i-nduction (at. time Two identical constructs for B72:3.VH, and: f-or VL (pMG.ROH4, pMG.ROH9, pMG.ROL5 and pMG.ROL'O) were- transformed into host strain E. coli CAG629? a:n-d induced for protein synthesis. Times of sampling were, in order, 0, 2, 4, 6 and 15 hours (for the last time course the 4 hour sample was omitted). V H and VL both accumulated to a maximum level after 6 hours.

Approximately equimolar amounts were synthesised, intermediate between levels obtained from Gloop2 VH and

VL;

Figure 9 illustrates the appearance of the VH and VL polypeptides in the insoluble fraction of host cells; insoluble fractions of cells were washed stringently with 2% Triton X100 (buffered) and solubilised in 7.6M guanidine buffer (see Cabilly et al., 1984). These cDntain the.polypeptides VH and VL, as assessed by C.oomassie-b.lue- stained 15% SDS-PAGE of different loadings of B 7 2 .13VH,, V. and Gloop2 VH and VL, in order. Lysozyme migrated at: a marginally higher position than B72.3 CH in the B72.3 examples; and Figure 10 shows Coomassie-blue stained SDS-PAGE showi-ng recombinant Gloop2 VH and VL expressed as insoluble material in Escherichia coli; c A 6 0 0 cells, i S= insoluble material, s soluble fraction, w supernatants from pellet washes. H VH and L VL.

i WO 89/02465 PCT/GB88/00747 15 Detailed Description of the Embodiments Example 1 The Fv fragment of an antibody having specificity for a characterised epitope of hen egg lysozyme (HEL) was produced by expression of heavy variable (VH) and light variable (VL) polypeptides in E.coli and reconstitution in vitro.

A panel of five murine monoclonals, Gloop 1 have been raised to a well characterised epitope, the loop of hen egg lysozyme (HEL), Darsley and A.R. Rees, (1985a)). Genes coding for the heavy and light chains of Gloop 2 (IgG2b, k) have been cloned and the nature of this antibody's interaction with HEL has been analysed by the combined strategy of site-directed mutagenesis (SDM) and a computer model derived from complementarity determining region (CDR) homology in known crystal structures de la Paz, B.J Sutton, M.J. Darsley and A.R. Rees, (1986); S. Roberts and A.R. Rees, (1986); and S. Roberts et al, (1987).

In the manipulation of the cloned Gloop2 heavy and light chain genes two amino-acid codons from the 5' end of 62b were included in the heavy-chain variable-region gene (vhg2); vlg2 consisted of light chain variable region sequences alone. Transcription signals were present in the parent dual-origin vector chosen for expression (pMG939). The parent vector contained the gene for a mammalian growth hormone and was used as a control for temperature-inducible expression and reconstitution of Pep-l binding activity. The vector was derived from pMG411 (Wright et al., 1986) Site-directed mutagenesis of the Gloop2 genes was according to the method of Carter et al. (1985), except that for the double mutagenesis (vlg2) the two mutagenic and the selection oligonucleotides were included in equimolar amounts to a total concentration equal to that described.

WO 89/02465 PCT/GB88/00747 -16- Oligonucleotides were made on an Applied Biosystems International plc Model 381A DNA synthesiser, and purified as described by Roberts and Rees (1986).

The entire vlg2 sequence was ascertined using the method of Sanger et al., (1977) and that of vhg2 by both dideoxy and Maxam and Gilbert sequencing.

Linker insertion at the 5' end of vhg2 (in pAT153) used a 17mer and a 13mer oligonucleotide, phosphorylating at only the blunt site. A 100 times molar ratio of oligonucleotides to vector fragment was used. Ligation was as described by Maniatis et al., (1982). The presence of the linker was screened for using radiolabelled 17mer as described by Carter et al., (1984), performing the prehybridisation in sealed bags at ambient temperature overnight. Colonies showing very strong signals were sequenced by the method of Maxam and Gilbert (1977).

Both rFv genes were excised with Bgl2 and EcoRl and after purification from low melt agarose by the method of Maniatis et al., (1982), ligated into the similarly purified Bgl2-EcoRl vector fragment from pMG939. ligation was',assessed by restriction analysis.

Gloop2 VH and VL polypeptides were directly expressed in severely protease-deficient strains of E.

coli, htpR", lon". These formyl-Methionine derivatives consisted of 116 and 109 amino acids for V

H

and VL respectively. The expression of the Gloop2 polypeptides was analysed by biosynthetic labelling in vivo (Figure 2).

In vigorously growing E. coli strain 1B373 both polypeptides were produced, but only VL was present at high levels after 2 minutes. Neither polypeptide accumulated in this strain or in the naturally occuring lon" strains B and B/r, as assessed by Coomassie-stained SDS-PAGE (not shown). Accumulation was detected only in host strains of E. coli deficient in the heat shock WO 89/02465 PCT/GB88/00747 17 response, htpR- strains (Baker et al., 1984) (Figure 3C,D). A lon" mutation in this background (CAG629) allowed even greater accumulation (Figure 3D), and this host was therefore used as the production strain.

Their small size, and their hydrophobicity, makes the V H and VL formyl-methionine derivatives vulnerable to heat-shock proteases, perhaps because they mimic the intracellular signals generated by stress that activate the heat shock system.

La, encoded by lon, is well-known for its scavenging properties particularly in the presence of small or abnormal polypeptides (Waxman and Goldberg, 1982); it is a heat-shock protein under htpR regulation (Goff et al., 1984). htpR" host strains have allowed product accumulation for other small proteins such as the 8.5 kDa complement factor C5a (Mandecki et al., 1986) and the amino-acid somatomedin-C (Buell et al, 1985). The htpR gene encodes a positive regulator of the heat-shock response, a 32-kDa c-factor for core RNA polymerase (Grossman et al., 1984) conferring specificity for heat shock promoters. The continuous presence of this and other factors (Zimarino and Wu, 1987) ensures the rapid response to stimulus, whether temperature or abnormal polypeptides (Burdon, 1986), observed when V H and VL were produced following heat-shock induction. The compensating activity of protease Ti (Hwang et al., 1987) may explain the lack of V accumulation in lon- strains.

|1 VH expression was improved by a longer Shine-Dalgarno sequence. Production of VH was augmented 2-fold by insertion of a 9 base-pair (bp) Shine-Dalgarno sequence (5'-AAGGAGGTA) into the trp ribosome binding site (vector pPW.HF2). Analysis by in vitro synthetic labelling showed improved expression (Figure 3A) which f ~resulted in increased accumulation of V H in CAG629 (Figure 3B,C). No significant changes in production levels were

SA

WO 89/02465 PCT/GB88/00747 18 noted for VL, although if anything, expression was decreased (Field, 1988) The longer Shine-Delgarno sequence had no effect on production of the more stable growth hormone, but may prove useful for the production of some smaller polypeptides 14.5 kDa). Alternative SD and leader sequences have previously been used to enhance expression (Greenfield et al., 1986; Sung et al., 1986).

Computer analysis of the ability of the message to base-pair with the 16S ribosomal RNA indicated that the longer Shine-Dalgarno may establish more stable binding to the appropriate ribosomal RNA sequence than did the original trp promoter. This is in agreement with other studies demonstrating that the affinity of the ribosome for the messenger RNA is important for efficient translation (Coleman et al., 1985; Jacob et al., 1987).

The influence of secondary structure in the ribosome binding site will depend on the gene sequence itself, thus the lack of improvement noted for V L expression (in vitro expression was decreased) may be explained by the 'introduction of unfavourable secondary structure into this region.

The influence of the post-initiation AUG sequences on gene expression is becoming well-documented (Looman et al., 1987; Kiltunow et al., 1987). Manipulation of the Shine-Dalgarno to AUG start codon distance is also known to affect synthesis (Fujisawa et al., 1985). However, reduction of the number of base-pairs from 14 to 12, 11 or had no pronounced effect on VH production from pMG.HF2, although once this distance was reduced to 5 bp, expression was greatly diminished (Field, 1988). Computer studies also indicated the VL was expressed at high levels due to enhanced messenger stability: a region of internal homology at the 3' end of the messenger RNA was postulated to form stable secondary structure deflecting i WO 89/02465 PCT/GB88/00747 -19acting exoribonucleases (Glass, 1982). This internal homology was absent in vhg2 (Field, 1988).

The microvax programs BESTFIT and COMPARE were used for analysis of homology. FOLD was used to assess the secondary structure of the messenger RNA. The sequence for the E. coli 16S ribosomal rRNA was taken from file ecotgix.seq, in the UIC database (University of Wisconsin).

Transformed cell lines were grown at 30 0 C overnight, in L-broth containing 100 ,g.ml 1 ampicillin and, for CAG629, 10 g.ml-1 tetracyclin also. 50mls L-broth containing drugs as appropriate were innoculated at 0.08 A600nm and grown at 30 0 C, shaking at ca. 200 rpm, to an

A

600 of 0.4, or 1.0 for htpR~ strains, transferred to a 45-50 0 C water bath and flash-induced to 42 0 C, shaking frequently. Cultures were then grown for up to 6 hours at 37 0 C, shaking, to allow product accumulation. Other volumes of cultures, 1 litre or 10mls, were grown, innculating and inducing at the same absorbance Both V H and VL products were found in the insoluble fraction of host cells, the inclusion bodies (IBs) (Field et al., 1988). The M r of VH and VL were similar, and just less than 14 kDa. Excessive hydrophobicity may account for this apparently diminished Mr, as noted for the Fv polypeptides produced by Hochman et al., (1973).

Fractionation of cell extracts into soluble and insoluble (inclusion body, IB) material was performed using the method of Schoemaker et al., (1985) with the following alterations: cells were lysed in a buffer containing 100mM Tris.HCl pH 7.5, 5mM EDTA and 270 Ag.

ml 1 lysozyme, and incubated with 0.2 mM PMSF for minutes on ice. Sodium deoxycholate was added as described and incubated on ice; the final incubation, on ice for 5 minutes, was with 0.04mg.ml-1 DNase I and MgCl 2 -u 1 1 :i WO 89/02465 PCT/GB88/00747 Reconstitution and partial purification of the rFv and controls was accomplished either following the protocol' of Cabilly et al., (1984) with reconstitution from:cell lysates, or the IBs were extracted by cell lysis with: lysozyme as described above, or else IBs were extracted.following French pressure cell lysis (sonication in the case of the growth hormone and host cell controls). At each stage of the manipulation (preceding solubilisation in guanidine), 50 ul 100mM phenylmethylsulphonyl fluoride (PMSF) and 35 of a cocktail of protease inhibitors (ImM EDTA/10/g.ml-1 y-g.ml l pepstatin/200 /g.ml-1 bacitracin/lmM benzamadine-HCl/0.1mM PMSF/20/g.ml-1 soybean trypsin inhibitor) were added. Pellets were washed twice with 2% Triton-Xl00 and once with water, dissolved in 7.6M guanidine-HCl, 50mM Tris-HCl pH8, ImM EDTA, 0.1M B-mercaptoethanol, and incubated for one hour at 37 0

C.

The extracts were diluted to give a potential rFv content of 50, g.ml 1 in 8M guanidine-HCL/50mM Tris-HCl pH 8/lmM.EDTA. VH and VL were also reconstituted separately and mixed! at 50 /g.ml-1, while growth hormone and host cell. control extracts were added to approximately the same host-cell protein content. These were each, separately reconstituted according to the method of Cabilly et al., (1984) except that 1 ml protease inhibitors were added to the final phosphate buffered saline dialysis medium. All reconstitutions were concentrated using (Amicon), and fractionated by gel filtration using Fast Protein Liquid Chromatography (FPLC) through Superose 12 (Pharmacia).

The process was also scaled-up and purification to homogeneity carried out. 2 litres VH-culture and 0.6 litres3VL-culture were lysed in 30mls and 10mls lysosyme lysi-s buffer containing protease inhibitors (above).

WO 89/02465 PCT/GB88/00747 21 Lysed cells were centrifuged in 30ml Corex tubes at 1200g for 10 minutes, then washed three times with 2% 1'riton-Xl00 in 100mM Tris-HCl pH 7.5/5mM EDTA, adding protease inhibitors each time. Pellets were resuspended in 20mls and 10mls of 6M guanidine buffer (as above).

Extracts,were fractionated on a 300ml Sephacryl S-200 (Pharmacia) column equilibrated in 6M guanidine-HC1/50mM Tris-HCl pH 8.0/5mM EDTA/lmM A-mercaptoethanol. V

H

and VL- containing fractions were pooled separately. A potential rFv concentration of 200 yg.ml 1 was reconstituted as described above, using per dialysis bag per 625mls urea buffer. Following concentration by ultrafiltration using a YM5 membrane (Amicon) the reconstitutions were further fractionated on the same Sephacryl S-200 column equilibrated in PBS. The rFv fractions were pooled and dialysed into 20mM Tris pH 6.34 and loaded onto a 13ml (20cm) DEAE (DE-52, Whatman) column equilibrated in the same buffer. rFv eluted in the void. For refolding by dilution, guanidine lysates were diluted from 5M to 2M guanidine with PBS pH 7.5, mixed and then further diluted to 0.6M guanidine. Pepl binding of a control prepared by dialysis as above, was compared.

In order to demonstrate the reconstitution of rFv heteroduplex in partially purified matiial washed IBs or frozen host cell pellets were solubilised and reconstituted. Binding to Pep-l, a proteolytic fragment of HEL containing the loop antigen (and to which the Gloops were raised), was tested by the competition assay described below, and found only in reconstitutions containing putative rFv, and not in controls consisting of similarly prepared host cells, host cells carrying the parent vector (expressing recombinant growth hormone), or VH and VL alone (not shown). The presence of lysozyme in the lysis buffer was demonstrated not to alter Pepl binding.

WO 89/02465 PCT/GB88/00747 22 These reconstitutions were partially purified by gel filtration through Superose-12 (Figure A Pep-1 binding complex in VH plus VL reconstitutions eluted with a retention time consistent with the apparent Mr of a dimer 21,000). A peak on the A280nm elution profile at this position was unique to putative rFv-containing fractions, and eluted close to the position of growth hormone (Mr' 22,500). Pep-l binding was also obtained from reconstitutions of VH or VL alone, eluting at a retention time consistent with monomers, at Mr' -12,000 (in fraction 17). VH and VL therefore did not form homodimers, and rFv must be heterodimeric.

The binding assay comprised binding 50 per well of 10/g.ml 1 affinity purified goat anti-mouse IgG in PBS/0.02% sodium azide or coating buffer (0.05M sodium carbonate/bicarbonate, pH 9.6/0.02% sodium azide) to microtitre plates at 4 0 C overnight, or 4 hours at ambient. After washing for 3x 1 minute with PBS/0.05% Tween 20/0.02% sodium azide, the wells were blocked with casein in PBS/azide or coating buffer, for 1-2 hours at 37 0 C. After washing, Gloop2 at /g.ml 1 was added as a second antibody layer in washing buffer, incubating for 4 hours at ambient or overnight at 4 0

C.

The wells were washed and test samples were serially diluted in triplicate (for the partially purified rFv/controls) or quadruplet for pure rFv assays (25 1 Superose 12 fractions were tested in duplicate), into .i PBS/azide. 125 i-labelled Pep 1 in washing buffer was added to a final volume of 50 100% values were obtained by adding only labelled Pep 1 to the well; non-specific binding was obtained by adding 100 g.ml-1 Gloop2 with the hot Pep 1. After equilibration overnight at 23 0 C plates were washed and individual wells cut out and counted on an LKB Clinigamma counter.

T

WO 89/02465 PCT/GB88/00747 -23- Computer analysis of the binding data was using the program SANCOL (R Ryan, 1988).

Protein concentrations were determined using the BCA assay (Pierce) and purity estimated by BCA and weighing densitometry traces made using a Joyce Mk III C8 double beam recording microdensitometer. Amino-acid analysis was performed using the Pico Tag (TM) system (Waters Associates Inc.) Following the final step of purification, shown in Figure 5, Gloop2 rFv was subjected to amino-acid analysis (Table Correct amino acid composition and a purity of 98% was demonstrated. Purity was confirmed by silver-stained SDS-PAGE.

Lf,,j (1 i -1 i 2 WO 89/02465 PCT/GB88/00747 -24- Table 1 Amino acid composition of purified rFv Amino acid Theoretical composition Observed composition (moles amino acid/mole rFv) Ala Arg Asx *Cys Glx Gly His Ile Leu +Lys Met Phe Pro *Ser *Thr Trp +Tyr Val Total 14.9 12.8 1.6 22.8 17.5 0.2 9.4 23.0 18.2 3.2 10.4 5.7 27.6 20.3 15.2 7.1 4 13 7 225 No.

Table 1 The amino acid composition of purified rFv was obtained by Pico Tag analysis, with amounts obtained in rmoles. Values were normalised to Asx. *Errors are known to occur in Cys, Ser, Thr because of oxidation (underestimated); +Tyr and Lys may be masked by Tris peaks (overestimated). Errors in the range of 2% for the more reliable estimates.

WO 89/02465 PCT/GB88/00747 The antigen (Pep-l) binding of rFv was compared with that of Fab'(Gloop2) and the whole MAB (Figure Within the limits of this type of experiment, the Kds were in good agreement, at 21 nM for whole Gloop2, 17nM for Fab'(Gloop2) and 9nM for rFv(Gloop2).

The extinction coefficient,£ was experimentally determined to be 38300 M- 1 .cm 1

(A

280 S 26,AM), compared with a theoretical value of 35620, obtained by addition of Trp and Tyr extinction values.

Purified rFv(Gloop2) retaining the binding activity observed in the intact monoclonal antibody may offer a solution to some of the problems encountered in studies of antigen-antibody binding. Studies are hampered by a low rate of production of protein antigen-antibody (Fab') co-crystals. The availability of antigen binding fragments in large amounts will allow physical studies.

Table 2 Purification step Fv/VH/VL (mg) Yield Purity Cells VH (21itres) 35.5 100 2 VL (1/3.5 40.0 100 7 Litres) Inclusion bodies >26-30 >90 7-12 S-200/guanidine 25-40 75-100 13-30 Reconstitution 4.8* 6 31 S-200/PBS 3.8 5 42 DEAE 1.6 1.9 >98 Table 2: yields of rFv or its components obtained throughout the purification protocol described. During reconstitution, BSA was added as a carrier, *This figure represents loss of protein drring dialysis: 80% losses were recorded at this stage Reconstitution of VH and VL into active Fv is in itself an efficient process.

7i c- -I WO 89102465 PCT/GB88/00747 -26- Expression levels of VH and VL were estimated at 2% and 7% of total bacterial protein, respectively. Using Cabilly's protocol for reconstitution, and other fractionation methods as described, 1 litre of shake-flask culture expressing VH and 0.3 litre VL culture yielded 0.7 mgs 98% pure rFv, representing 1.9% of the starting material (see Table This low yield was due mainly to protein losses during reconstitution, where more than of V polypeptides were lost. This problem was later alleviated by reconstitution involving a simple dilution of V polypeptides from guanidine into phosphate-buffered saline pH Hochman et al., (1973) have demonstrated the reconstitution of variable domains to be a very efficient process, with 87% recovery of the original binding activity. This compared favourably with the 17% recovery of Fab' activity from fully denatured polypeptides (Haber, 1964). The presence of the second domain militated against efficient refolding due to interdomain interactions (Tsunenaga et al,, 1987).

In addition, cultures may be industrially fermented to up to 40 times the cell density obtainable at bench-level (harvesting OD 600 nm was 1.5) allowing us to predict that tens of milligrams of purified rFv may be produced from 1 litre each of cells expressing VH and VL.

In vitro translation employing the method supplied with the prokaryotic translation kit from Amersham International was followed and the 125 I-Met labelled products analysed by SDS-PAGE and autoradiography.

Molecular weight protein markers were either 14 C-labelled (Amersham) or prestained (Gibco-BRL).

In Western blotting experiments proteins were transferred from 10-15% PA gels at 60V for 3 hours or at overnight, and the filter probed using the method of WO 89/02465 PCT/GB88/00747 -27- Burnette (1981), with the following alterations; the blot sandwich was made from 2 pieces of whatman 3MM enclosing the gel and nitrocellulose (Schleicher and Schuell) of 0 .1 /m pore size. The electrode solution contained Tris base, 192mM glycine and 20% methanol. Blocking was with 0.9% saline, 20mM Tris pH 7.6, 0.05% Triton X-100 and casein hammarsten (BDH), with four changes over 1 hour at ambient. The anti-VH or -VL sera (below) were applied by spreading in Iml blocking buffer, and incubated in a humidity chamber for 1 hour at ambient. Washing was with four changes of buffer without casein, and 2-4 ~Ci 125 I-labelled Protein A added in the same manner as the antisera. After incubating at ambient for 1 hour, the filter was washed in a high salt wash buffer containing 1M NaC1, 20mM Tris pH 7.6 and 0.4% N-lauroyl sarcosine four cimes at ambient, air dried and autoradiographed.

To prepare rabbit antisera chloramphenicol acetyl transferase (CAT) fused to VH or VL at its C-terminus were produced in E. coli. The vector pCT202 was used to construct the vectors direct in expression of CAT-VH and CAT-VL. Insoluble material was prepared for both recombinants, which were electroeluted following preparative SDS-PAGE (with or without Coomassie staining) using a Biotrap system (Schleicher and Schuell), After dialysis into PBS the protein concentration was estimated and injected using the following schedule: 60 'g in Sa complete Freunds adjuvant in lml intradermally into the back and scruff; after 16 days, 130, g in incomplete Freunds in 0.8ml intramuscularly into the haunch; after a further 38 days, 50 yg in PBS was injected into haunch and scruff, and the next day 200frg in PBS intravenously into the ear vein. Bleeding was also from the ear vein, one week later.

WO 89/02465 PCT/GB88/00747 -28- The route for direct polypeptide expression in E.

coli, using an inducible vector in conjunction with heat-shock protease deficient host strains was useful for production of another, unrelated rFv.

Example 2 The Fv fragment of an antibody having specificity for a carbohydrate epitope of a breast carcinoma antigen (Colcher et al., 1981) was produced by expression of heavy variable (VH) and light variable polypeptides in E.

coli and reconstituted in vitro in the same manner as for the Gloop2 Fv.

This murine monoclonal antibody will have potential application in tumour therapy. To reduce its immunogenicity in the human system, a chimera was constructed (Whittle et al., 1987). The variable region was placed onto human constant regions by genetic engineering. The retension of binding activity demonstrated the independence of the binding of the variable region (or Fv).

VH and VL genes were constructed from the previously cloned complete (but unrearranged see Whittle et al., op. cit.) heavy and light chain genes by insertion of an ATG (encoding formyl-Methionine) start and a stop codon.

A 3' Bgl2 and a 5' EcoRI site were also engineered, such that the entire Fv sequence was contained in two genes and bounded by restriction sites compatible with the expression vector used.

end manipulation was by insertion of linkers, as described for Gloop2 VH gene construction. The linker region sequences were verified by dideoxy sequencing of the double-stranded plasmids, after insertion into the Svector, pMG939.

The growth hormone gene of the vector described in the previous example, pMG939, was removed, and the VH or WO 89/02465 PCT/GB88/00747 -29- VL genes for B72.3 expression inserted, forming the plasmids pMG.ROH4/9 and pMG.ROL5/10.

The dual origin vector above was used to produce both heavy and light polypeptide chains (120 and 109 amino acids, respectively) in the protease deficient strain E.

coli CAG629 (htpR- and lon") (Figure Both polypeptides appeared in the insoluble fraction of host cells (Figure 9).

Concluding Remarks The small size of rFvs should simplify studies of antibody combining sites by physical methods. For X-ray crystallography, the switch site angle (between variable and constant domains) need not be determined, reducing computation times, and the similarity of tertiary structure between rFvs will-allow rapid assessment of data from new rFv crystals. Furthermore, not all Fabs may be induced to crystallise (Davies and Metzger, 1983), and it is possible that rFvs, lacking the flexible elbow bend, will crystallise more easily. The ability to produce VH and VL separately will allow their complete assignment by NMR spectroscopy, and assessment of the interdomain interaction.

E. coli will not glycosylate its products. Although the variable domains of immunoglobulins are non-stoichiometrically glycosylated in vivo, the carbohydrate is not usually involved in antigen-binding (Taniguchi et al., 1985) and specifically, absence of carbohydrate is found not to effect IgG2b function (Nose and Wigzell, 1983).

The conserved nature of variable domain pairing (Chothia et al., 1985) suggests that the methods described here for reconstitution of rFv binding from chaotropic buffer will also apply to any rFv. The methods described here will be employed in antigen-binding studies. A i 1 i WO 89/02465 WO 8902465PCT/GB88/00747 30 potentially high yield, cytoplasmnic route for V polypeptide expression may encourage the development of immunotoxin constructs, or chimeric therapeutic agents with low inununogenciity.

WO 89/02465 WO 8902465PCT/GB88/00747 References Baker, Grossman, A.D. and Gross, C.A. (1984) Proc. Natl.

Acad. Sci. USA, 81, 6779-6783.

Boss, M.A. Kenten, C.R. Enitage, S.J. (1984) Nucleic Ac.

Res. 12, 3791-3806.

Buell, Schulz, M-F Selzer, Chollet, Movva, N.R., Semon, Escanez, Kawashima, E. (1985) Nucleic Ac. Res.

13, 1923-1938. Burdon, R.H. (1981) Biochem. J. 240, 313-324.

Burnette, W.N. (1981) Anal. Biochem. 112, 195-203.

Cabilly, S, Riggs, Pande, Shively, Holmes, W.E., Rey, Perry, Wetzel R. Heyneker, H.L. (1984) Proc.

Natl. Acad. Sci. USA 81, 3273-3277.

Carter, Bedouelle, H. Winter, G. (1985) Nucleic Ac. Res.

13, 4431-4442.

Carter, Winter, Wilkinson, A.J. Fersht. (1984) Cell 38, 835-840.

Chothia, Novotny, Bruccoleri, R. Karplus, M. (1985) J.

Mol. Biol. 186, 651-663.

Colcher, Hand, Nuti, Schlon,, J. (1981) Proc. Natl.

Acad. Sci. USA 78, 3199-3203.

Coleman, Inouye, M. Nakamura, K. (1985) J. Mol. Biol. 181, 139-143.

Darsley, M.J. Rees, A.R. (1985a) EMBO J. 4, 383-392.

Darsley, M.J. Rees, A.R. (1985b) EMBO J. 4, 393-398.

Davies, D.R. Metzger, H. (1983) Ann. Rev. Immunol. 1, 87-lib.

A

~Ti ii i 'I WO 89/02465 PCT/GB88/00747 -32 de la paz, Sutton, Darsley, M.J. Rees, A.R. (1986) EMBO J. 5, 415-425.

Field, H (1988) D.Phil thesis, Oxford University, U.K.

Field, Yarranton, G.T. Rees, A.R. (1988) In "Vaccines Cold Spring Harbor Publications, Pages 29-34.

Fujisawa, Ito, Ikeyama, S. Kikuchi, M. (1985) gene 23-29.

Glass, R.E. (1982) gene Function. E. coli and its heritable elements. Croom helm, London, pp. 58-70.

Goff, Casson, L.P. Goldberg, A.L. (1984) proc. Natl.

Acad. Sci. USA 81, 6647-6651.

Greenfield, Dovey, Lawyer, F.C. Gelfand, D.H. (1986) Bio/Technology 4, 1006-1011.

Grossman, Erickson, J.W. Gross, C.A (1984) Cell 38, 383-390.

Haber, E. (1964) Proc. Natl. Acad. Sci. USA 52, 1099-1106 Hendershot, Bole, D. and Kearney, J.F. (1987) Immunology Today 8, 111-114.

Hochman, Inbar, D. Givol, D. (1973) Biochemistry 12, 1130-1135 Hwang, Park, Chung, C.H. Goldberg, A.L. (1987) Proc. Natl. Acad. Sci. USA 84, 5550-5554.

Jacob, Santer, M. Dahlberg, A.E. (1987) Proc. Natl.

Acad. Sci. USA 84, 4757-4761.

4) t WO 89/02465 PCT/GB88/00747 -33 Kenten Helm, Ishizaka, Cattini, P. Gould, H.

(1984) Proc. Natl. Acad. Sci. USA 81, 2955-2959.

Klausner, A. (1986) Bio/Technology 4, 1041-1043.

Koltuno, A.M. Gregg, K. Rogers, G.E. (.1987) Nucleic Ac. Res.

7795-7808.

Looman, Bodlaender, Cornstock, Eaton, D., Jhurani, de Boer, H.A. van Knippenberg, P.H. (1987) EMBO J. 6, 2489-2492.

Mandecki, Powell, Mollison, Carter, G.W. Fox, J.L. (1986) Gene 43, 131-138.

Maniatis, Fritsch, E.F. 'Sambrook, J. (1982) Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, New York.

Maxam, A.M. Gilbert, W. (1987) Proc. Natl. Acad. Sci USA 74, 560-654.

Nose, M. Wigzell, h. (1983) Proc. Natl. Acad. Sci. USA 6632-6636.

Roberts, Cheetham, J.C. Rees, A.R. (1987) Nature (Lond.) 328, 731-734.

Roberts, S. Rees, A.R. (1986) Protein Engineering 1, 59-65.

Ryan, R. (1988) D. Phil thesis, Oxford University, U.K.

Sanger, Nicklen, S. Coulson, A.R. (1987) Proc. Natl. Acad.

Sci. USA 74, 5463-5467.

WO 89/02465 PCT/GB88/00747 Schoemaker, Brasnett, A.H. Marston, F.A.O. (1985) EMBO J. 4 775-780.

Sen, J. and Beychok, S. (1986) Proteins: structure, function& genetics 1, 256-262.

Sharon, J. Givol, D. (1976) Biochemistry 15, 1591-1594.

Sung, Yao, Zahab, D.M. Narang, S.A. (1986) Proc.

Natl. Acad. Sci. USA 83, 561-565.

Taniguchi, Mizuochi, Beale, Dwek, Rademacher, T.W. and Kobata, A. (1985) Biochemistry 24, 5551-5557.

Tsunenaga, Goto, Kawata, Hamag-uchi, K. (1987) Biochemistry 26, 6044-6051.

Verhoeyen, M. and Riechmann, L. (1988) Bioessays 8, 74-78 Waxman, L. Goldberg, A.L. (1982) Proc. Natl. Acad. Sci USA 79, 4883-4887.

Whittle, Adair, Lloyd, Jenkins, Devine, J., Schlom, Raubitschek, Coicher, Bodmer, M4 (1987) Protein Engineering 1, 499-505.

Wright, Humnphreys, G.O. Yarranton G.T. (1986) gene 49, 311-321. Zimarino, V. Wu, C. (1987) Nature (Lond.) 327, 727-730.

Claims (11)

1. A process, for the production of an immunologically functional antibody fragment, wherein heavy and light chain polypeptides having complementary variable domains are directly expressed in a protease deficient bacterial host cell using an expression system in which vector copy number is inducible.

2. The process of claim 1, wherein the antibody fragment is an Fv fragment.

3. The process of claim 1 or claim 2, wherein the protease deficient bacterial host cell is an E. coli strain.

4. The process of claim 3, wherein the protease deficient bacterial host strain is a lon- strain.

5. The process of claim 3, wherein the protease deficient bacterial host strain is a hptR~ strain.

6. The process of any one of claims 1 to 5, wherein the protease deficient bacterial host strain is multiply protease deficient.

7. A copy number inducible bacterial expression vector containing a DNA sequence coding for a heavy or light chain immunoglobulin polypeptide having a substantially complete variable domain wherein the DNA sequence is located within the vector such that it is expressed as a direct expression product.

8. A copy number inducible bacterial expression vector containing DNA sequences coding for heavy and light chain immunoglobulin polypeptides having substantially complete variable domains wherein the DNA sequences are located within the vector such that they are expressed as direct expression products. mwspe#3543 91 11 27

9. A protease deficient bacterial host cell transformed with the expression vector of Claim 7 or Claim 8. A process according to any one of Claims 1 to 9 substantially as hereinbefore described.

11. A vector according to any one of Claims 7 or 8 substantially as hereinbefore described.

12. A host cell according to Claim 9 substantially as hereinbefore described. DATED this 17 July 1991 ::10 CARTER SMITH BEADLE Fellows Institute of Patent Attorneys of Australia Patent Attorneys for the Applicant: S" CELLTECH LIMITED *s mwspe/3543 91 7 17