Patent Application
20080152586
This invention relates to target-binding polypeptides, especially polypeptides of high avidity and multiple specificity. In particular the invention relates to protein complexes which are polyvalent and/or polyspecific, and in which the specificity is preferably provided by the use of immunoglobulin-like domains. In one particularly preferred embodiment the protein complex is trivalent and/or trispecific.
Reagents having the ability to bind specifically to a predetermined chemical entity are widely used as diagnostic agents or for targeting of chemotherapeutic agents. Because of their exquisite specificity, antibodies, especially monoclonal antibodies, have been very widely used as the source of the chemical binding specificity.
Monoclonal antibodies are derived from an isolated cell line such as hybridoma cells; however, the hybridoma technology is expensive, time-consuming to maintain and limited in scope. It is not possible to produce monoclonal antibodies, much less monoclonal antibodies of the appropriate affinity, to a complete range of target antigens.
Antibody genes or fragments thereof can be cloned and expressed in E. coli in a biologically functional form. Antibodies and antibody fragments can also be produced by recombinant DNA technology using either bacterial or mammalian cells. The hapten- or antigen-binding site of an antibody, referred to herein as the target-binding region (TBR), is composed of amino acid residues provided by up to six variable surface loops at the extremity of the molecule.
These loops in the outer domain (Fv) are termed complementarity-determining regions (CDRs), and provide the specificity of binding of the antibody to its antigenic target. This binding function is localised to the variable domains of the antibody molecule, which are located at the amino-terminal end of both the heavy and light chains. This is illustrated in
Recombinant Fv fragments are prone to dissociation, and therefore some workers have chosen to covalently link the two domains to form a construct designated scFv, in which two peptides with binding domains (usually antibody heavy and light variable regions) are joined by a linker peptide connecting the C-terminus of one domain to the N-terminus of the other, so that the relative positions of the antigen binding domains are consistent with those found in the original antibody (see
Methods of manufacture of covalently linked Fv fragments are disclosed in U.S. Pat. No. 4,946,778 and U.S. Pat. No. 5,132,405. Further heterogeneity can be achieved by the production of bifunctional and multifunctional agents (Huston et al U.S. Pat. No. 5,091,513, and Ladner et al U.S. Pat. No. 4,816,397).
The construction of scFv libraries is disclosed for example in European Patent Application No. 239400 and U.S. Pat. No. 4,946,778. However, single-chain Fv libraries are limited in size because of problems inherent in the cloning of a single DNA molecule encoding the scFv. Non-scFv libraries, such as VH or Fab libraries, are also known (Ladner and Guterman WO 90/02809), and may be used with a phage system for surface expression (Ladner et al WO 88/06630 and Bonnert et al WO 92/01047).
For use in antibody therapy, monoclonal antibodies, which are usually of mouse origin, have limited use unless they are first “humanised”, because they elicit an antigenic response on administration to humans. The variable domains of an antibody consist of a β-sheet framework with six hypervariable regions (CDRs) which fashion the antigen-binding site. Humanisation consists of substituting mouse sequences that provide the binding affinity, particularly the CDR loop sequences, into a human variable domain structure. The murine CDR loop regions can therefore provide the binding affinities for the required antigen. Recombinant antibody “humanisation” by grafting of CDRs is disclosed by Winter et al (EP-239400).
The expression of diverse recombinant human antibodies by the use of expression/combinatorial systems has been described (Marks et al, 1991). Recent developments in methods for the expression of peptides and proteins on the surface of filamentous phage (McCafferty et al, 1991; Clackson et al, 1991) offer the potential for the selection, improvement and development of these reagents as diagnostics and therapeutics. The use of modified bacteriophage genomes for the expression, presentation and pairing of cloned heavy and light chain genes of both mouse and human origins has been described (Hoogenboom et al, 1991; Marks et al, 1991 op.cit. and Bonnert et al, WPI Acc. No. 92-056862/07)
Receptor molecules, whose expression is the result of the receptor-coding gene library in the expressing organism, may also be displayed in the same way (Lerner and Sorge, WO 90/14430). The cell surface expression of single chain antibody domains fused to a cell surface protein is disclosed by Ladner et al, WO 88/06630.
Affinity maturation is a process whereby the binding specificity, affinity or avidity of an antibody can be modified. A number of laboratory techniques have been devised whereby amino acid sequence diversity is created by the application of various mutation strategies, either on the entire antibody fragment or on selected regions such as the CDRs. Mutation to change enzyme specific activity has also been reported. The person skilled in the art will be aware of a variety of methods for achieving random or site-directed mutagenesis, and for selecting molecules with a desired modification. Mechanisms to increase diversity and to select specific antibodies by the so called “chain shuffling” technique, i.e. the reassortment of a library of one chain type e.g. heavy chain, with a fixed complementary chain, such as light chain, have also been described (Kang et al, 1991; Hoogenboom et al, 1991; Marks et al, 1992).
Our earlier International Patent Application No. PCT/AU93/00491 described recombinant constructs encoding target polypeptides having a stable core polypeptide region and at least one target-binding region, in which the target binding region(s) is/are covalently attached to the stable core polypeptide region, and has optionally been subjected to a maturation step to modify the specificity, affinity or avidity of binding to the target. The polypeptides may self-associate to form stable dimers, aggregates or arrays. The entire disclosure of PCT/AU93/00491 is incorporated herein by this cross-reference.
This specification did not predict that scFv-0 constructs in which the C-terminus of one V domain is ligated to the N-terminus of another domain, and therefore lack a foreign linker polypeptide, would form trimers. In contrast, it was suggested that, like constructs incorporating a linker, they would form dimers. A trimeric Fab′ fragment formed by chemical means using a tri-maleimide cross-linking agent, referred to as tri-Fab, has been described (Schott et al, 1993 and Antoniw et al, 1996). These tri-Fab molecules, also termed TMF, have been labelled with 90Y as potential agents for radioimmunotherapy of colon carcinoma, and have been shown to have superior therapeutic effects and fewer side-effects compared to the corresponding IgG. This was thought to result from more rapid penetration into the tumour and more rapid blood clearance, possibly resulting from the nature of the cross-linked antibody fragment rather than merely the lower molecular weight (Antoniw et al, 1996). However, these authors did not examine the affinity or avidity of either the IgG or the TMF construct.
Recombinant single chain variable fragments (scFvs) of antibodies, in which the two variable domains VH and VL are covalently joined via a flexible peptide linker, have been shown to fold in the same conformation as the parent Fab (Kortt et al, 1994; Zdanov et al, 1994; see
Various protein linking strategies have been used to produce bivalent or bispecific scFvs as well as bifunctional scFv fusions, and these reagents have numerous applications in clinical diagnosis and therapy (see
In the design of scFvs, peptide linkers have been engineered to bridge the 35 Å distance between the carboxy terminus of one domain and the amino terminus of the other domain without affecting the ability of the domains to fold and form an intact binding site (Bird et al, 1988; Huston et al, 1988). The length and composition of various linkers have been investigated (Huston et al, 1991) and linkers of 14-25 residues have been routinely used in over 30 different scFv constructions, (WO 31789/93, Bird et al, 1988; Huston et al, 1988; Whitlow and Filpula, 1991; PCT/AU93/00491; Whitlow et al, 1993, 1994). The most frequently used linker is that of 15 residues (Gly4Ser), introduced by Huston et al (1988), with the serine residue enhancing the hydrophilicity of the peptide backbone to allow hydrogen bonding to solvent molecules, and the glycyl residues to provide the linker with flexibility to adopt a range of conformations (Argos, 1990). These properties also prevent interaction of the linker peptide with the hydrophobic interface of the individual domains. Whitlow et al (1993) have suggested that scFvs with linkers longer than 15 residues show higher affinities. In addition, linkers based on natural linker peptides, such as the 28 residue interdomain peptide of Trichoderma reesi cellobiohydrolase I, have been used to link the VH and VL domains (Takkinen et al, 1991).
A scFv fragment of antibody NC10 which recognises a dominant epitope of N9 neuraminidase, a surface glycoprotein of influenza virus, has been constructed and expressed in E. coli (PCT/AU93/00491; Malby et al, 1993). In this scFv, the VH and VL domains were linked with a classical 15 residue linker, (Gly4 Ser)3, and the construct contained a hydrophilic octapeptide (FLAG™) attached to the C-terminus of the VL chain as a label for identification and affinity purification (Hopp et al, 1988). The scFv-15 was isolated as a monomer which formed relatively stable dimers and higher molecular mass multimers on freezing at high protein concentrations. The dimers were active, shown to be bivalent (Kortt et al, 1994), and reacted with soluble N9 neuraminidase tetramers to yield a complex with an Mr of ˜600 kDa, consistent with 4 scFvs dimers cross-linking two neuraminidase molecules. Crystallographic studies on the NC10 scFv-15 monomer-neuraminidase complex showed that there were two scFv-neuraminidase complexes in the asymmetric unit and that the C-terminal ends of two VH domains of the scFv molecules were in close contact (Kortt et al, 1994). This packing indicated that VH and VL domains could be joined with shorter linkers to form stable dimeric structures with domains pairing from different molecules and thus provide a mechanism for the construction of bispecific molecules (WO 94/13804, PCT/AU93/00491; Hudson et al, 1994, 1995).
Reduction of the linker length to shorter than 12 residues prevents the monomeric configuration and forces two scFv molecules into a dimeric conformation, termed diabodies (Holliger et al, 1993, 1996; Hudson et al, 1995; Atwell et al, 1996;
ScFv-0 VH-VL molecules have been designed without a linker polypeptide, by direct ligation of the C-terminal residue of VH to the N-terminal residue of VL (Holliger et al, 1993, McGuiness et al, 1996). These scFv-0 structures have previously been thought to be dimers.
We have now discovered that NC10 scFv molecules with VH and VL domains either joined directly together or joined with one or two residues in the linker polypeptide can be directed to form polyvalent molecules larger than dimers and in one aspect of the invention with a preference to form trimers. We have discovered that the trimers are trivalent, with 3 active antigen-combining sites (TBRs; target-binding regions). We have also discovered that NC10 scFv molecules with VL domains directly linked to VH domains can form tetramers that are tetravalent, with 4 active antigen-combining sites (TBRs).
We initially thought that these trimeric and tetrameric conformations might result from steric clashes between residues which were unique to the NC10scFv, and prevented the dimeric association. However, we have discovered that a second scFv with directly linked VH-VL domains, constructed from the monoclonal anti-idiotype antibody 11-1G10, is also a trimer and is trivalent, with 3 active TBRs. The parent antibody, murine 11-1G10, competes for binding to the murine NC41 antibody with the original target antigen, influenza virus N9 neuraminidase (NA) (Metzger and Webster, 1990). We have also discovered that another scFv with directly linked VH-VL domains (C215 specific for C215 antigen) is also a trimer.
We now propose that the propensity to form polyvalent molecules and particularly trimers is a general property of scFvs with VH and VL domains either joined directly together or joined with one or two residues in the linker polypeptide, perhaps due to the constraints imposed upon V-domain contacts for dimer formation. It will be appreciated by those skilled in the art that the polyvalent molecules can be readily separated and purified as trimers, tetramers and higher multimers.
Due to polyvalent binding to multiple antigens, trimers, tetramers and higher multimers exhibit a gain in functional affinity over the corresponding monomeric or dimeric molecules. This improved avidity makes the polymeric scFvs particularly attractive as therapeutic and diagnostic reagents. Furthermore the ability to utilise polycistronic expression vectors for recombinant production of these molecules enables polyspecific proteins to be produced.
The invention generally provides polyvalent or polyspecific protein complexes, in which three or more polypeptides associate to form three or more functional target-binding regions (TBRs). A protein complex is defined as a stable association of several polypeptides via non-covalent interactions, and may include aligned V-domain surfaces typical of the Fv module of an antibody (
The individual polypeptides which form the polyvalent complex may be the same or different, and preferably each comprise at least two immunoglobulin-like domains of any member of the immunoglobulin superfamily, including but not limited to antibodies, T-cell receptor fragments, CD4, CD8, CD80, CD86, CD28 or CTLA4.
It will be clearly understood that the length of the linker joining the immunoglobulin-like domains on each individual polypeptide molecule is chosen so as to prevent the two domains from associating together to form a functional target-binding region (TBR) analogous to Fv, TCR or CD8 molecules. The length of the linker is also chosen to prevent the formation of diabodies. Instead, three or more separate polypeptide molecules associate together to form a polyvalent complex with three or more functional target-binding regions.
In a first aspect the invention provides a trimeric protein comprising three identical polypeptides, each of which comprises immunoglobulin VH and VL domains which are covalently joined preferably without a polypeptide linker, in which the peptides associate to form a trimer with three active TBRS, each of which is specific for the same target molecule.
In a second aspect the invention provides a trimeric protein comprising three different polypeptides, each of which comprises antibody VH and VL domains or other immunoglobulin domains, which are covalently joined preferably without a polypeptide linker, in which the polypeptides associate to form a trimer with three active TBRs directed against three different targets.
In one preferred embodiment of the second aspect the trimer comprises one TBR directed to a cancer cell-surface molecule and one or two TBRs directed to T-cell surface molecules.
In a second preferred embodiment the trimer comprises one TBR directed against a cancer cell surface molecule (a tumour antigen), and a second TBR directed against a different cell surface molecule on the same cancer cell.
In a third preferred embodiment the trimer comprises two TBRs directed against the same cancer cell-surface molecule and one TBR directed to a T-cell surface molecule.
In one preferred embodiment of the second aspect, one TBR of the trimer can be directed to a costimulatory T-cell surface molecule, such as CTLA4, CD28, CD80 or CD86.
Particularly preferred trivalent or trispecific reagents according to the invention include the following:
1) Three identical VH-VL molecules (scFv×3) which are inactive as monomers but which form active trimers with 3 (identical) antigen combining sites (TBRs).
2) Three different VH-VL molecules (scFv×3) which are inactive as monomers but which form active trimers with 3 different antigen combining sites (TBRs).
In a third aspect the invention provides a tetrameric protein comprising four identical polypeptides, each of which comprises immunoglobulin VH and VL domains which are covalently joined preferably without a polypeptide linker, in which the peptides associate to form a tetramer with four active TBRs each with specificity to the same target molecule.
In a fourth aspect the invention provides a tetrameric protein comprising four different polypeptides each of which comprises antibody VH and VL domains or other immunoglobulin domains, which are covalently joined preferably without a polypeptide linker, in which the polypeptides associate to form a tetramer with four active TBRs directed against four different targets.
In one preferred embodiment of the fourth aspect the tetramer comprises one or more TBRs directed to a cancer cell-surface molecule and one or more TBRs directed to T-cell surface molecules.
In a second preferred embodiment the tetramer comprises one or more TBRs directed against a cancer cell surface molecule (a tumour antigen), and one or more TBRs directed against a different cell surface molecule on the same cancer cell.
In one preferred embodiment of the fourth aspect, one TBR of the tetramer is directed to a costimulatory T-cell surface molecule, such as CTLA4, CD28, CD80 or CD86.
It will be clearly understood that the molecules which form the polyvalent or polyspecific proteins of the invention may comprise modifications introduced by any suitable method. For example one or more of the polypeptides may be linked to a biologically-active substance, chemical agent, peptide, drug or protein, or may be modified by site-directed or random mutagenesis, in order to modulate the binding properties, stability, biological activity or pharmacokinetic properties of the molecule or of the construct as a whole. The linking may be effected by any suitable chemical means alternatively, where the protein of the invention is to be conjugated to another protein or to a peptide, this may be achieved by recombinant means to express a suitable fusion protein. It will also be appreciated that chemical modifications and disulphide bonds to effect interdomain cross-links may be introduced in order to improve stability. Selection strategies may be used to identify desirable variants generated using such methods of modification. For example, phage display methods and affinity selection are very well known, and are widely used in the art.
Mechanisms to increase diversity and to select specific antibodies by the so-called “chain shuffling” technique, i.e. the reassortment of a library of one chain type e.g. heavy chain, with a fixed complementary chain, such as light chain, have also been described (Kang et al, 1991; Hoogenboom et al, 1991; Marks et al, 1992; Figini et al, 1994).
In order to avoid the generation of an immune response in the subject to which the polyvalent reagent of the invention is administered, and to ensure that repeat treatment is possible, it is preferred that the molecules comprising the polyvalent reagent are of homologous origin to the subject to be treated, or have been modified to remove as far as possible any xenoantigens. Thus if the recipient is a human, the molecules will be of human origin or will be humanised (CDR-grafted) versions of such molecules. “Humanisation” of recombinant antibody by grafting of CDRs is disclosed by Winter et al, EP-239400, and other appropriate methods, e.g. epitope imprinted selection (Figini et al, 1994), are also widely known in the art.
Where the immunoglobulin-like domains are derived from an antibody, the TBR may be directed to a chemical entity of any type. For example it may be directed to a small molecule such as a pesticide or a drug, a hormone such as a steroid, an amino acid, a peptide or a polypeptide; an antigen, such as a bacterial, viral or cell surface antigen; another antibody or another member of the immunoglobulin superfamily; a tumour marker, a growth factor etc. The person skilled in the art will readily be able to select the most suitable antigen or epitope for the desired purpose.
According to a fifth aspect, the invention provides a pharmaceutical composition comprising a polyvalent or polyspecific reagent according to the invention together with a pharmaceutically-acceptable carrier.
According to a sixth aspect the invention provides a method of treatment of a pathological condition, comprising the step of administering an effective amount of a polyspecific reagent according to the invention to a subject in need of such treatment, wherein one TBR of the reagent is directed to a marker which is:
a) characteristic of an organism which causes the pathological condition, or
b) characteristic of a cell of the subject which manifests the pathological condition,
and a second TBR of the reagent binds specifically to a therapeutic agent suitable for treatment of the pathological condition.
Preferably two different TBRs of the reagent are directed against markers of the pathological condition, and the third to the therapeutic agent, or alternatively one TBR of the reagent is directed to a marker for the pathological condition or its causative organism, and the two remaining TBRs of the reagent are directed to two different therapeutic agents. It is contemplated that the method of the invention is particularly suitable for treatment of tumours, in which case suitable therapeutic agents include but are not limited to cytotoxic agents, toxins and radioisotopes.
According to a seventh aspect the invention provides a method of diagnosis of a pathological condition, comprising the steps of administering a polyvalent or polyspecific reagent according to the invention to a subject suspected of suffering from said pathological condition, and identifying a site of localisation of the polyvalent or polyspecific reagent using a suitable detection method.
This diagnostic method of the invention may be applied to a variety of techniques, including radio imaging and dye marker techniques, and is suitable for detection and localisation of cancers, blood clots etc.
In another preferred embodiment of this aspect of the invention there is provided an imaging reagent comprising:
a) a trimer of the invention in which all three components (TBRs) of the trimer are directed to a molecular marker specific for a pathological condition and in which the trimer is either labelled with radioisotopes or is conjugated to a suitable imaging reagent.
b) a trimer of the invention in which either two TBRs of the trimer are directed to two different markers specific for a pathological condition or site, and the third is directed to a suitable imaging reagent;
c) one TBR of the trimer is directed to a marker characteristic of a pathological condition, such as a tumour marker, a second TBR is directed to a marker specific for a tissue site where the pathological condition is suspected to exist, and the third is directed to a suitable imaging agent, or
d) one TBR of the trimer is directed to a marker characteristic of the pathological condition and the remaining two TBRS are directed to two different imaging agents.
In one preferred embodiment of the invention, one component of the polyspecific molecule is a non-antibody immunoglobulin-like molecule. These Ig-like molecules are useful for binding to cell surfaces and for recruitment of antigen presenting cells, T-cells, macrophages or NK cells. The range of Ig-like molecules for these applications includes:
a) The Ig-like extracellular domain of CTLA4 and derivatives (Linsley et al, 1995). CTLA4 binds to its cognate receptors B7-1 and B7-2 on antigen presenting cells, either as a monomer (a single V-like domain) or as a dimer or as a single chain derivative of a dimer.
b) The Ig-like extracellular domains of B7-1 and B7-2 (CD80, CD86 respectively; Peach et al, 1995, Linsley et al, 1994) which have homology to Ig variable and constant domains.
In a preferred embodiment, the Ig-like domains described above are affinity-matured analogues of the natural mammalian sequence which have been selected to possess higher binding affinity to their cognate receptor. Techniques for affinity maturation are well known in the field, and include mutagenesis of CDR-like loops, framework or surface regions and random mutagenesis strategies (Irving et al, 1996). Phage display can be used to screen a large number of mutants (Irving et al, 1996). CTLA4 and CD80/86 derivatives with enhanced binding activity (through increases in functional affinity) have application in preventing transplant rejection and intervening in autoimmune diseases. These molecules interfere with T-cell communication to antigen presenting cells, and can either activate T-cells leading to proliferation with application as an anti-cancer reagent, or decrease T-cell activation, leading to tolerance, with application in the treatment of autoimmune disease and transplantation (Linsley et al, 1994, 1995). These molecules can also be used to activate NK cells and macrophages once recruited to a target site or cell population.
In a further preferred embodiment, trispecific reagents comprise dimeric versions of CTLA4 or CD80/86 or one molecule of each species, which is expected to result in further affinity enhancement and with similar therapeutic applications as described above.
In a further preferred embodiment, one component of the trispecific reagents may comprise a non Ig-like domains, such as CD40, to manipulate the activity of T and NK cells.
a) Gel stained with Coomassie brilliant blue G-250;
b) Western blot analysis of the same fractions using anti-FLAG™ M2 antibody.
a) the ‘sandwich’ complex between two tetrameric neuraminidases and four scFv dimers based on crystallographic data of the neuraminidase-Fab complex (Tulip et al, 1992; Malby et al, 1994) and scFv-15 monomer complex (Kortt et al, 1994),
b) the complex between scFv-5 dimer and anti-idiotype 3-2G12 Fab′,
c) the scFv-0 trimer (c.f.
d) the scFv-0 binding three anti-idiotype Fab′ fragments to form a complex of Mr 212 kDa.
a) boomerangs; NC10 VH-VL scFv-5 diabody/3-2G12 Fab complex,
b) Y-shaped tripods; NC10 VH-VL scFv-0 triabody/3-2G12 Fab complex,
c) V-shaped projections; NC10 VH-VL scFv-0 triabody/3-2G12 Fab complex, and
d) X-shaped tetramers; NC10 VL-VH scFv-0 tetramer/3-2G12 Fab complex.
Magnification bar 50 nm.
A: An scFv comprising VH-L-VL where L is a linker polypeptide as described by Whitlow et al and WO 93/31789; by Ladner et al, U.S. Pat. No. 4,946,778 and WO 88/06630; and by McCafferty et al (1991) and by McCartney et al. (1995).
B: A single polypeptide VH-L1-VL-L2-VH-L3-VL which forms two scFv modules joined by linker polypeptide L2, and in which the VH and VL domains of each scFv module are joined by polypeptides L1 and L3 respectively. The design is described by Chang, AU-640863.
C: Two scFv molecules each comprising VH-L1-VL-L2 (a,b), in which the VH and VL domains are joined by linker polypeptide L1 and the two scFv domains are joined together by a C-terminal adhesive linkers L2a and L2b. The design is described by Pack et al, PI-93-258685.
D: The design of PCT/AU93/00491, clearly different to A, B and C above. A single scFv molecule VH-L-VL comprises a shortened linker polypeptide L which specifically prevents formation of scFvs of the type A, B or C, and instead forces self-association of two scFvs into a bivalent scFv dimer with two antigen combining sites (target-binding regions; TBR-A). The association of two different scFv molecules will form a bispecific diabody (TBRs-A,B).
The invention will be described in detail by reference only to the following non-limiting examples and to the figures.
N9 neuraminidase was isolated from avian (tern) influenza virus following treatment of the virus with pronase and purified by gel filtration as described previously (McKimm-Breschkin et al, 1991).
Monoclonal anti-idiotype antibodies 3-2G12 and 11-1G10 were prepared in CAF1 mice against NC10 and NC41 anti-neuraminidase BALB/c monoclonal antibodies (Metzger and Webster, 1990). Anti-neuraminidase antibody NC41 and the anti-idiotype antibodies 3-2G12 and 11-1G10 were isolated from ascites fluid by protein A-Sepharose chromatography (Pharmacia Biotech). Purified antibodies were dialysed against 0.05 M Tris-HCl, 3 mM EDTA, pH 7.0 and digested with papain to yield F(ab′)2 as described (Gruen et al, 1993). The F(ab′)2 fragment from each antibody was separated from Fc and undigested IgG by chromatography on protein A-Sepharose, and pure F(ab′)2 was reduced with 0.01 M mercaptoethylamine for 1 h at 37° C. and the reaction quenched with iodoacetic acid. The Fab′ was separated from the reagents and unreduced F(ab′)2 by gel filtration on a Superdex 75 column (HR 10/30) in PBS, 7.4.
The molecular size and aggregation state of affinity purified scFvs were assessed by size exclusion FPLC on Superose 6 or 12, or Superdex 75 HR 10/30 (Pharmacia) columns in PBS, pH 7.4. The ability of the scFv-0, scFv-5 and scFv-10 to bind to antigen and anti-idiotype Fab′ fragments, and the size of the complexes formed, was also assessed by size exclusion FPLC on Superose 6 in PBS, pH 7.4. The columns were equilibrated with a set of standard proteins as described previously (Kortt et al, 1994).
The molecular mass of scFv-0, scFv-5 and scFv-10, and that of the complexes formed with antigen and anti-idiotype antibody Fab′ fragments with each scFv, was determined in 0.05 M phosphate-0.15 M NaCl, pH 7.4 by sedimentation equilibrium in a Beckman model XLA ultracentrifuge.
The BIAcore™ biosensor (Pharmacia Biosensor AB, Uppsala Sweden), which uses surface plasmon resonance detection and permits real-time interaction analysis of two interacting species (Karlsson et al, 1991; Jonsson et al, 1993), was used to measure the binding kinetics of the different NC10 scFvs. Samples for binding analyses were prepared for each experiment by gel filtration on Superdex 75 or Superose 12 to remove any cleavage products or higher molecular mass aggregates which may have formed on storage. The kinetic constants, ka and kd, were evaluated using the BIAevaluation 2.1 software supplied by the manufacturer, for binding data where the experimental design correlated with the simple 1:1 interaction model used for the analysis of BIAcore™ binding data (Karlsson et al, 1994).
Solutions of the two complexes; NC10 scFv-5 diabody/Fab, NC10 scFv-0 triabody/Fab, and also a mixture of NC10 scFv-0 triabody/Fab with free 3-G12 anti-idiotype Fab were examined by electron microscopy. In each case, proteins were diluted in phosphate-buffered saline (PBS) to concentrations of the order of 0.01-0.03 mg/ml. Prior to dilution, 10% glutaraldehyde (Fluka) was added to the PBS to achieve a final concentration of 1% glutaraldehyde. Droplets of ˜3 μl of this solution were applied to thin carbon film on 700-mesh gold grids which had been glow-discharged in nitrogen for 30 s. After 1 min the excess protein solution was drawn off, followed by application and withdrawal of 4-5 droplets of negative stain (2% potassium phosphotungstate adjusted to pH 6.0 with KOH). The grids were air-dried and then examined at 60 kV in a JEOL 100B transmission electron microscope at a magnification of 100,000×. Electron micrographs were recorded on Kodak SO-163 film and developed in undiluted Kodak D19 developer. The electron-optical magnification was calibrated under identical imaging conditions by recording single-molecule images of the NC10 antibody (Fab) complex with its antigen, influenza virus neuraminidase heads.
Measurements of particle dimensions were made on digitised micrographs using the interactive facilities of the SPIDER image processing suite to record the coordinates of particle vertices. Particle arm lengths and inter-arm angles were calculated from the coordinates for 229 diabodies and 114 triabodies.
The NC10 scFv antibody gene construct with a 15 residue linker (Malby et al, 1993) was used for the shorter linker constructions. The NC10 scFv-15 gene was digested successively with BstEII (New England Labs) and SacI (Pharmacia) and the polypeptide linker sequence released. The remaining plasmid which contained NC10 scFv DNA fragments was purified on an agarose gel and the DNA concentrated by precipitation with ethanol. Synthetic oligonucleotides (Table 1) were phosphorylated at the 5′ termini by incubation at 37° C. for 30 min with 0.5 units of T4 polynucleotide kinase (Pharmacia) and 1 mM ATP in One-Phor-All buffer (Pharmacia). Pairs of complementary phosphorylated oligonucleotide primers (Table 1) were premixed in equimolar ratios to form DNA duplexes which encoded single chain linkers of altered lengths.
These duplexes were ligated into BstEII-SacI restricted pPOW NC10 scFv plasmid using an Amersham ligation kit. The ligation mixture was purified by phenol/chloroform extraction, precipitated with ethanol in the usual manner, and transformed into E. coli DH5α (supE44, hsdR17, recA1, endA1, gyrA96, thi-1, relA1) and LE392 (supE44, supF58, hsdR4, lacY1, galK2, galT22, metB1, trpR55). Recombinant clones were identified by PCR screening with oligonucleotides directed to the pelB leader and FLAG sequences of the pPOW vector. The DNA sequences of the shortened linker regions were verified by sequencing double-stranded DNA using Sequenase 2.0 (USB).
The new NC10 scFv gene constructs, in which the VH and VL domains were linked with linkers of 10 ((Gly4Ser)2), 5 (Gly4Ser) and zero residues, are shown in
The pPOW NC10 scFv constructs, with 0, 5 and 10 residues linkers as described in Example 1, were expressed as described by Malby et al, (1993) for the parent scFv-15. The protein was located in the periplasm as insoluble protein aggregates associated with the bacterial membrane fraction, as found for the NC10 scFv-15 (Kortt et al, 1994). Expressed NC10 scFvs with the shorter linkers were solubilised in 6M guanidine hydrochloride/0.1 M Tris/HCl, pH 8.0, dialysed against PBS, pH 7.4 and the insoluble material was removed by centrifugation. The soluble fraction was concentrated approximately 10-fold by ultrafiltration (Amicon stirred cell, YM10 membrane) as described previously (Kortt et al, 1994) and the concentrate was applied to a Sephadex G-100 column (60×2.5 cm) equilibrated with PBS, pH 7.4; fractions which contained protein were analysed by SDS-PAGE and the scFv was located by Western blot analysis using anti-FLAG™ M2 antibody (Eastman Kodak, New Haven, Conn.). The scFv containing fractions were pooled, concentrated and purified to homogeneity by affinity chromatography using an anti-FLAG™ M2 antibody affinity resin (Brizzard et al, 1994). The affinity resin was equilibrated in PBS pH 7.4 and bound protein was eluted with 0.1 M glycine buffer, pH 3.0 and immediately neutralised with 1M Tris solution. Purified scFvs were concentrated to ˜1-2 mg/ml, dialysed against PBS, pH 7.4 which contained 0.02% (w/v) sodium azide and stored frozen.
The purity of the scFvs was monitored by SDS-PAGE and Western blot analysis as described previously (Kortt et al, 1994). The concentrations of the scFv fragments were determined spectrophotometrically using the values for the extinction coefficient (ε0.1%) at 280 nm of 1.69 for scFv-15, 1.71 for scFv-10, 1.73 for scFv-5 and 1.75 for scFv-0 calculated from the protein sequence as described by Gill and von Hippel (1989).
For N-terminal sequence analysis of the intact scFv-0 and scFv-5 and the two lower molecular mass cleavage products, the protein bands obtained on SDS-PAGE were blotted on to a Selex 20 membrane (Schleicher and Schuell GmbH, Germany), excised and sequenced using an Applied Biosystems Model 470A gas-phase sequencer.
Soluble NC10 scFv-10, scFv-5 and scFv-0 fragments were each purified using a two step procedure involving gel filtration and affinity chromatography after extraction of the E. coli membrane fraction with 6 M guanidine hydrochloride, and dialysis to remove denaturant. The solubilised protein obtained was first chromatographed on Sephadex G-100 gel filtration to resolve three peaks (peaks 1-3, as shown in
Affinity chromatography of the Sephadex G-100 peak 1 from
Gel filtration on a calibrated Superdex 75 column of affinity purified scFvs showed that the NC10 scFv-10 (
Gel filtration of affinity-purified NC10 scFv-0 yielded a single major symmetrical peak with an apparent molecular mass of approximately 70 kDa (
A partial specific volume of 0.71 ml/g was calculated for scFv-5 and scFv-0 from their amino acid compositions, and a partial specific volume of 0.7 ml/g was calculated for the scFv-neuraminidase complexes, from the amino acid compositions of scFvs and the amino acid and carbohydrate compositions of neuraminidase (Ward et al, 1983). A partial specific volume of 0.73 ml/g was assumed for the scFv-anti-idiotype 3-2G12 Fab′ complex. The complexes for ultracentrifugation were prepared by size exclusion FPLC on Superose 6. The results are summarized in Table 2.
The molecular masses of 54 and 52.4 kDa, respectively, for scFv-10 and scFv-5 confirmed that they were dimers. The molecular mass of 69 kDa determined for the NC10 scFv-0 suggested that it was a trimer composed of three scFv-0 chains, but this molecular mass is lower than expected for such a trimer (calculated Mr of 78 kDa). Analysis of the sedimentation data gave linear in c versus r2 plots (Van Holde, 1975), indicating that under the conditions of the experiment scFv-5 dimer and scFv-0 trimer showed no dissociation. Furthermore, the sedimentation equilibrium results did not indicate a rapid equilibrium between dimer and trimer species to account for this apparently low molecular mass for NC10 scFv-0 trimer.
Purified NC10 scFv-5 and scFv-10 dimers at concentrations of ˜1 mg/ml showed no propensity to form higher molecular mass aggregates at 4° C., but on freezing and thawing higher-molecular mass multimers were formed (data not shown). These multimers were dissociated readily in the presence of 60% ethylene glycol, which suppresses hydrophobic interactions. In contrast the NC10 scFv-0 showed no propensity to aggregate on freezing and thawing, even at relatively high protein concentrations.
N-terminal analysis of the two bands from the Fv fragment produced during the isolation of the NC10 scFv-0 (
Influenza virus neuraminidase, a surface glycoprotein, is a tetrameric protein composed of four identical subunits attached via a polypeptide stalk to a lipid and matrix protein shell on the viral surface (Colman, 1989). Intact and active neuraminidase heads (Mr 190 kDa) are released from the viral surface by proteolytic cleavage in the stalk region (Layer, 1978). The four subunits in the neuraminidase tetramer are arranged such that the enzyme active site and the epitope recognised by NC10 antibody are all located on the upper surface of the molecule (distal from the viral surface). This structural topology permits the binding in the same plane of four NC10 scFv-15 monomers or four Fab fragments (Colman et al, 1987; Tulip et al, 1992) such that the tetrameric complex resembles a flattened box or inverted table with the neuraminidase as the top and the four Fab fragments projecting as the legs from the plane at an angle of 45°. This suggests that a bivalent molecule may be able to cross-link two neuraminidase tetramers to form a ‘sandwich’ type complex (
Size-exclusion FPLC on a calibrated Superose 6 column showed that both the NC10 scFv-10 (
Gel filtration of the isolated 600 kDa NC10 scFv-10-neuraminidase complex showed that it was extremely stable to dilution, with only a small amount of free neuraminidase and NC10 scFv-10 appearing when complex at a concentration of 2 nM was run on the Superose 6 column. The linearity of the in c versus r2 plots (Van Holde, 1975) of the sedimentation data, demonstrated in Example 3, showed that both complexes were homogeneous with respect to molecular mass and indicated that discrete and stoichiometric complexes were formed. Complex formation with different molecular ratios of scFv to neuraminidase (from 1:4 to 8:1) yielded only the 600 kDa complex. Interestingly, complexes with 4 scFv dimers binding to 1 neuraminidase (˜400 kDa) or aggregated complexes in which more than two neuraminidases were cross-linked were not observed.
Size exclusion FPLC on Superose 6 showed that anti-idiotype 3-2G12 Fab′ formed stable complexes with NC10 scFv-15 monomer, NC10 scFv-5 and NC10 scFv-0, Sedimentation equilibrium analyses of the isolated complexes gave molecular masses consistent with the scFv-15 binding one Fab′, NC10 scFv-5 binding two Fab's and the NC10 scFv-0 binding three Fab′ molecules, as shown in Table 2 and
It will be appreciated that
Electron micrographs of the NC10 scFv-5 diabodies complexed with two anti-idiotype 3-2G12 Fab molecules (Mr ˜156 kDa) showed boomerang-shaped projections with the angle between the two arms ranging from about 60°-180°, as shown in
In micrographs of NC10 scFv-0 triabodies complexed with three 3-2G12 Fab molecules (Mr ˜212 kDa), most fields showed a mixture of predominantly Y-shaped and V-shaped projections (
The Y-shaped projections were unlikely to be planar, as invariably one of the Fab legs appeared foreshortened. The V-shaped projections were interpreted as tripods (triabody complexes) lying on their sides on the carbon film, with two Fab legs forming the V and the third Fab leg extending upward and out of the stain, which would account for the increase in density sometimes observed at the junction of the V.
The V-shaped structures were clearly different to the boomerang-shaped diabody complexes, both in the angle between Fab arms and in the projected density in the centre of the molecules, consistent with the expected models (
Triabodies are obviously flexible molecules, with observed angles between Fab arms in the NC10 triabody/Fab complexes distributed around two angles of mean 136° and one of mean 80°, and are not rigid molecules as shown schematically in
a) Binding of NC10 scFvs to Anti-Idiotype 3-2G12 Fab′
In a series of experiments anti-idiotype 3-2G12 Fab′ and the NC10 scFv-15 monomer, scFv-10, scFv-5 and scFv-0 were also immobilised at pH 4.0 via their amine groups. Binding analyses were performed in HBS buffer (10 mM HEPES, 0.15 M NaCl, 3.4 mM EDTA, 0.005% surfactant P20, pH 7.4) at a constant flow rate of 5 μl/min.
Immobilised 3-2G12 Fab′ could be regenerated with 10 μl 0.01 M sodium acetate buffer, pH 3.0 without loss of binding activity. A comparison of the binding of the NC10 scFv-15 monomer, scFv-10 and scFv-5 dimers, and scFv-0 trimer showed that the monomer dissociated rapidly, and non-linear least squares analysis of the dissociation and association phase, using the single exponential form of the rate equation, gave a good fit to the experimental data. These results are shown in
This table shows the apparent kinetic constants for the binding of NC10 scFv-15 monomer to immobilised tern N9 neuraminidase and anti-idiotype 3-2-G12 Fab′ fragment determined in the BIAcore™ The kinetic constants were evaluated from the association and dissociation phase using non-linear fitting procedures described in BIAevaluation 2.1. The binding experiments were performed in 10 mM HEPES, 0.15 NaCl, 3.4 mM EDTA, 0.005% surfactant P20, pH 7.4 at a flow rate of 5 μl/min. Tern N9 neuraminidase (1360 RU) and 3-2-G12 Fab′ (1000 RU) were immobilised via amine groups using the standard NHS/EDC coupling procedure.
The NC10 scFv-10 and scFv-5 dimers and scFv-0 trimer/anti-idiotype complexes showed apparently slower dissociation, as illustrated in
b) Binding of Anti-Idiotype 3-2G12 Fab′ to Immobilised NC10 scFv-15 Monomer and scFv-10, scFv-5 and scFv-0
i) NC10 scFv-15 Monomer
Although the scFv-15 monomer was readily immobilised (˜2000 Response Units; RU), less than 10% of the protein was active, as indicated by the total amount of anti-idiotype Fab′ that could be bound to the surface as calculated from the RU increase. Logarithmic transformation of the dissociation phase data showed significant deviation from linearity which permitted only approximate values of the binding constants to be estimated (Table 4).
ii) scFv-10, scFv-5 and scFv-0
In contrast, the three NC10 scFvs with the shorter linkers were not readily immobilised via their amine groups, since only 200-550 RU of protein could be immobilised after several injections of protein at a flow rate of 2 μl/min. Binding experiments with anti-idiotype 3-2G12 Fab′ indicated that approximately 30-50% of the immobilised scFv-10, scFv-5 and scFv-0 were active, as calculated from the total bound RU response. The results are shown in Table 4. As for immobilised NC10 scFv-15 monomer, analysis of the data showed deviation from linearity on logarithmic transformation of dissociation data and poor fits when the data was analysed by non-linear regression. These non-ideal effects associated with BIAcore™ binding data may arise either from the rate of molecular diffusion to the surface contributing to the kinetic constants (mass transfer effect) (Glaser, 1993; Karlsson et al, 1994) or from the binding heterogeneity of the immobilised molecules resulting from the non-specific immobilisation procedure used, or both. These effects contribute to lowering the measured rate constants and affect the estimated binding constants. A comparison of the rate constants for the binding of 3-2G12 Fab to each of the four immobilised NC10 scFvs shows that the apparent affinity for the interaction of 3-2G12 Fab with each scFv is similar, as shown in Table 4. Increases in affinity that are shown in
The starting template for construction of the short Tinkered scFvs was the zero-linked NC10 scFv-0 gene construct in the vector pPOW as described in Example 1, in which the 5′ end of the VL sequence is linked directly to the 3′ end of the VH sequence. The constructions were designed to add nucleotides coding for one, two, three or four glycine residues between the 3′ end of the VH and the 5′ end of the VL sequence.
Four sets of complementary oligonucleotide primers were synthesised as shown in Table 5 to add the extra codons between the VH and VL sequences, using the QuikChange™ Site-Directed Mutagenesis procedure (Stratagene Cloning Systems, La Jolla, Calif.).
15 ng NC10 scFv-0 DNA was subjected to PCR in a 50 μl reaction volume containing 5 μl reaction buffer supplied with the kit, 20 pmoles of the complementary oligonucleotide primers, 2.5 nmoles of each dNTP, and 2.5 units Pfu DNA polymerase. Thermal cycling conditions were: (95° C., 30 secs) 1 cycle; (95° C., 30 sec; 55° C., 1 min; 68° C., 12 min) 18 cycles. 1 μl Dpn I restriction enzyme (10 U/μl) was added to each sample and incubated at 37° C. for 90 min to digest dam methylated, non-mutated parental DNA. 2 μl of each reaction mixture was used to transform electrocompetent XL1-Blue cells (recA endA 1 gyrA96 thi-1 hsdR17 supE44 relA1 lac [f′ proAB lacIqZΔM15 Th10 (tetr)]) (1×109 cfu/μg), aliquots of which were incubated overnight on YT-amp100 plates at 30° C.
Mutants containing the correct nucleotide insertions were selected by DNA sequencing of plasmid DNA from a number of individual colonies across the region targeted for mutation, using Sequenase ver 2.0 (US Biochemicals) and the oligonucleotide primer TACATGCAGCTCAGCAGCCTGAC (SEQ ID NO. 17). Clones having the correct mutations were subjected to small scale expression in 5 ml 2YT/amp200 as described in Malby et al (1993) to confirm that the construct could produce a full length, in-frame product. Culture samples were analysed by SDS-PAGE and Western Blot with anti-FLAG® M2 antibody. The selection criterion was a positive reaction at the correct migration position. One positive clone was selected from this screen for each of the four constructions.
Large-scale expression and purification of NC10 scFv-1, scFv-2, scFv-3 and scFv-4 were performed as described in Example 2, but with the chromatography step on Sephadex G-100 omitted. SDS PAGE and Western Blot of the bound fraction from affinity chromatography on immobilised anti FLAG revealed that they contained predominantly NC10 scFv.
Estimation of Molecular Mass of NC10 scFv-1, scFv-2, scFv-3 and scFv-4
Aliquots of affinity purified NC10 scFv-1, scFv-2, scFv-3, scFv-4 were individually analysed by FPLC on a calibrated Superose 12 column. Elution profiles are shown in
Activity of TBRs—Formation of Complexes with 3-2G12 Fab′ and EM Imaging
Complexes were formed between 3-2-G12 Fab′ and affinity purified NC10 scFv-2 and scFv-3, as described for scFv-0 and scFv-5 (Example 4), isolated by FPLC on Superose 6 and used for EM imaging, also as described for scFv-0 and scFv-5.
The absence of any free scFv peak in the FPLC profile after the formation of complexes in the presence of excess Fab′ indicated that both scFv-2 and scFv-3 were completely active. The elution time for the scFv-2/Fab complex was identical to that found previously for the scFv-0/Fab complex, and is consistent with scFv-2 being a trimer. The scFv-3/Fab complex had an identical elution time to that found previously for the scFv-5/Fab complex, and is consistent with the scFv-3 being a dimer.
EM images of scFv-2/Fab and scFv-3/Fab complexes showed results which were consistent with our previous observations that the NC10 scFv-2 was a stable trimer similar to scFv-0 and scFv-3 was a stable dimer similar to scFv-5. These images appear identical to either scFv-5 dimer complexes or scFv-0 trimer complexes shown in
The VH and VL genes were amplified by PCR from the parent 11-1G10 hybridoma, and joined into an scFv-0 gene by ligation between codons for C-terminal VH-Ser113 and N-terminal VL-Gln1 by PCR overlap-extension. For 11-1G10 the zero-linkered scFv is defined as the direct linkage of VH-Ser113 to VL-Gln1. The scFv-0 gene was cloned into the Sfi1-Not1 sites of the expression vector pGC which provides an N-terminal pelB leader sequence and C-terminal FLAG octapeptide tag tail (Coia et al, 1996). The entire DNA sequence of the cloned scFv-0 insert was determined using DNA purified by alkaline lysis and sequencing reactions performed using the PRISM Cycle Sequencing Kit (ABI). This confirmed that the 11-1G10 scFv-0 gene comprised a direct ligation between codons for the C-terminal VH-Ser113 and N-terminal VL-Gln1.
HB101 E. coli containing the scFv-0 gene in pGC were grown in 2×YT supplemented with 100 μg/ml ampicillin and 1% glucose at 37° C. overnight and then subcultured in the absence of glucose at an A600 of 0.1, and grown at 21° C. until A600 was 1.0. Expression was induced by addition of IPTG to 1 mM and cells cultured for 16 hours at 21° C. under conditions which release the contents of the periplasmic space into the culture supernatant, presumably by cell lysis, to yield soluble and biologically active scFv (Coia et al, 1996). Cells and culture supernatant were separated by centrifugation, and samples of cell pellet and supernatant were analysed on a 15% SDS-PAGE gel, followed by Western blot analysis using M2 anti-FLAG antibody (Kortt et al, 1994) and goat anti-mouse IgG (H+L)HRP (BioRad) as the second antibody to visualise the expressed product.
The expressed scFv-0 was purified from supernatant by precipitation with ammonium sulphate to 70% saturation at 21° C. followed by centrifugation at 10000 g for 15 minutes. The aqueous phase was discarded, and the pellet resuspended and dialysed in PBS at 4° C. overnight. Insoluble material was removed by centrifugation at 70,000 g and the supernatant was filtered through a 0.22 μm membrane and affinity purified on either an M2 anti-FLAG antibody affinity column (Brizzard et al, 1994) or an NC41 Fab Sepharose 4B affinity column. The affinity resin was equilibrated in TBS (0.025M Tris-buffered saline, pH 7.4) and bound protein was eluted with gentle elution buffer (Pierce). The scFv-0 was concentrated to about 1 mg/ml, dialysed against TBS and stored at 4° C. SDS-PAGE analysis of the affinity purified scFv-0 revealed a single protein band of 27 kDa which on Western analysis reacted with the anti-FLAG M2 antibody (
The affinity-purified 11-1G10 scFv-0 was as described in Example 5. For the other proteins described in this example, the 11-1G10 scFv-15 (comprising a 15 residue linker in the orientation VH-(Gly4Ser)3-VL) was synthesised under similar conditions to the scFv-0 described in Example 5 above. The 11-1G10 scFv-15 was isolated by gel filtration as a 27 kDa monomer and shown to be stable at 4° C. for several weeks, similar to previous studies with different scFv-15 fragments. NC41 and 11-1G10 Fab fragments were prepared by proteolysis from the parent hybridoma IgG as described previously in this specification. 11-1G10 scFv-0 and scFv-15 were fractionated by size exclusion FPLC on either a Superdex 75 HR10/30 column or a Superose 12 HR10/30 column (Pharmacia) in PBS to determine the molecular size and aggregation state.
The complexes formed between 11-1G10 scFv and NC41 Fab were analysed and isolated by size exclusion FPLC on a Superose 12 column in PBS (flow rate 0.5 ml/min). The FPLC columns were calibrated with standard proteins as described (Kortt et al, 1994). The molecular mass of each isolated complex was determined by sedimentation equilibrium on a Beckman model XLA centrifuge as described previously (Kortt et al, 1994) using partial specific volumes calculated from amino acid compositions. An upgraded Pharmacia BIAcore™ 1000 was used for analysis of the binding of monomeric 11-1G10 scFv-15 and trimeric 11-1G10 scFv-0 to immobilised NC41 Fab as described (Kortt et al, 1994). The resulting binding curves were analysed with BIAevaluation 2.1 software (Pharmacia Biosensor), to obtain values for the apparent dissociation rate constants.
Gel filtration of affinity purified scFv-0 by FPLC on either a Superdex 75 column (
The complexes of NC41 Fab with either scFv-15 monomer or scFv-0 trimer were isolated by size exclusion FPLC chromatography and analysed by sedimentation equilibrium in a Beckman Model XLA ultracentrifuge. The molecular mass was determined experimentally by the method described by Kortt et al, 1994 at 20° C. The calculated MW of NC41 Fab is 47273 Da, scFv-15 is 28427 Da and scFv-0 is 26466 Da.
In comparison, the scFv-15 fragment of 11-1G10 (comprising a 15 residue linker in the orientation VH-(Gly4Ser)3-VL) was also synthesised using the pGC vector in HB2151 E. coli cells, and then purified as a stable monomer with a Mr ˜27 kDa determined by gel filtration and sedimentation equilibrium (
Gel filtration analysis by FPLC on a Superose 12 column showed that all the scFv-0 interacted with NC41 Fab to form a stable complex of Mr ˜245 kDa (
The NC10 scFv-0 (VL-VH) gene encoded the pelB leader immediately followed by the N-terminal residues of DIEL for the VL gene. The C-terminus of the VL gene encoded residues KLEIR107 (where R is unusual for VL). The N terminus of the VH (residues QVQL) immediately followed to form a linkerless construct. The C-terminus of the VH terminated with residues VTS112, and was immediately followed by a C-terminal FLAG™ sequence for affinity purification. The NC10 scFv-0 VL-VH gene was, then subcloned and expressed in the heat inducible expression vector pPOW using methods described in Kortt et al, 1994 and Examples 1-4 above. The isolation of NC10 scFv-0 (VL-VH) from the E. coli cell pellet required extraction and solubilisation with 6M GuHCl, preliminary purification using a Sephadex G-100 column, and affinity purification using an anti-FLAG M2 affinity column, using methods described in Kortt et al, 1994.
SDS-PAGE and Western blot analysis of purified NC10 scFv-0 (VL-VH) gave a major protein band at ˜30 kDa. FPLC analysis of purified scFv-(VL-VH) on a Superose 12 HR10/30 column (Pharmacia) run at a flow rate of 0.5 ml/min gave a major protein peak eluting at 22.01 minutes with a distinct shoulder on the trailing edge of the peak (
Purified NC10 scFv-0 (VL-VH) tetramer and NC10 scFv-0 (VL-VH) trimer reacted with anti-idiotype 3-2G12v Fab to yield complexes of 4 Fab/tetramer and 3 Fab/trimer, demonstrating the tetravalent and trivalent nature of the two NC10 scFv-0 (VL-VH) molecules. EM analysis of complexes of the isolated NC10 scFv-0 VL-VH trimer and tetramer complexed with 3-2G12 anti-idiotype Fab showed images of tripods and crosses consistent with the trimers having 3 functional TBRs and the tetramers having 4 active TBRs, as shown in
The strategy for construction of the zero-linked C215 scFv antibody gene construct was as described in Example 7 in which the 5′ end of the VL sequence (Glu1) is linked directly to the 3′ end of the VH sequence (Ser113). The VH and VL genes of C215 (Forsberg et al, 1997) were amplified by PCR from the parent Fab coding region, and joined into an scFv-0 gene by PCR overlap-extension. The scFv-0 gene was cloned into the Sfi1-Not1 sites of the expression vector pGC, which provides an N-terminal pelB leader sequence and C-terminal FLAG octapeptide tag tail (Coia et al, 1996). The C-terminus of the VL terminated with residues ELK107, and was immediately followed by the C-terminal FLAG™ sequence for affinity purification. The scFv-0-linker gene was also cloned into the NdeI-EcoRI sites of the expression vector pRSET™, which is a cytoplasmic expression vector. The oligonucleotides used to amplify the C215 with the correct restriction sites for cloning into pRSET are:
The entire DNA sequences of the cloned scFv-0 inserts were determined using DNA purified by alkaline lysis and sequencing reactions were performed using the PRISM Cycle Sequencing Kit (ABI). This confirmed that the C215 scFv-0 gene comprised a direct ligation between codons for the C-terminal VH-Ser121 and N-terminal VL-Glu1.
HB101 E. coli expression of the C215 scFv-0 was performed as detailed in Example 7 The C215 scFv-0 was concentrated to about 1 mg/ml, dialysed against TBS and stored at 4° C. SDS-PAGE analysis of the affinity purified scFv-0 revealed a single protein band of Mr ˜28 kDa which on Western analysis reacted with the anti-FLAG M2 antibody. N-terminal sequence analysis of the Mr ˜28 kDa, protein gave the expected sequence for the N-terminus of the C215 VH domain, and confirmed that the pelB leader sequence had been correctly cleaved.
The affinity-purified C215 scFv-0 was as described in Example 10.
Gel filtration of affinity-purified C215 scFv-0 by FPLC on a calibrated Superose 12 column (HR10/30) revealed a major peak of Mr ˜85 kDa, (an apparent trimer) with a retention time of 20.20 mins. as shown in
Construction of Three Discrete Bispecific Ig-Like V Domains which are Designed to Assemble into Trimers with Three Different binding specificities: CTLA-4-0 Linked to CD86, CTLA-4-0 Linked to UV-3 VL and UV-3 VH-0 Linked to CD86.
The Ig-like V domains were separately amplified by PCR from the parent coding region with appropriate oligonucleotides pairs which are listed in table 6: #4474/#4475 (UV-3 VH), #4480/4481 (UV-3 VL), #4470/#4471 (human CTLA-4) (Dariavach 1988), #4472/#4473 (CD86 V domain) respectively.
Human CTLA-4 and CD86 (Aruffo and Seed 1987) were joined into a 0-linker gene construct by a linking PCR with oligonucleotides #4470 & #4473. Human CTLA-4 and UV-3 VL were joined into 0-linker gene construct by a linking PCR with oligonucleotides #4478 & # 4471 and UV-3 VH and human CD86 were joined into 0-linker gene construct by a linking PCR with oligonucleotides #4474 & #4477. This produced ligation between codons for C-terminal UV-3 VH-Ala114 and N-terminal CD86-Ala1 by PCR overlap-extension. The Ig-like V domain 0-linker gene constructs were cloned into the Sfi1-Not1 sites of the expression vector pGC, which provides an N-terminal pelB leader sequence and C-terminal FLAG octapeptide tag tail (Coia et al, 1996). Ligation between codons for C-terminal CTLA-4-Ala112 and N-terminal CD86-Ala1 by PCR overlap-extension produced Ig-like V domain 0-linker gene constructs which were cloned into the Sfi1-Not1 sites of the expression vector pGC. Ligation between codons for C-terminal CTLA-4-Ala112 and N-terminal UV-3-VL-Glu1 by PCR overlap-extension was used to produce the Ig-like V domain 0-linker gene construct, which was cloned into the Sfi1-Not1 sites of the expression vector pGC. The C-terminus of the VL was immediately followed by the FLAG™ sequence for affinity purification.
The entire DNA sequence of the cloned Ig-like V domains with 0-linkers was determined, using DNA purified by alkaline lysis and sequencing reactions performed using the PRISM Cycle Sequencing Kit (ABI). This confirmed that the Ig-like V domain 0-linker gene constructs comprised direct ligation between codons for each of the domains. Expression was as described in Example 5. Gel filtration of affinity-purified CTLA-4-0-CD86, CTLA-4-0-UV-3 VL or UV-3 VH-0-CD86 by FPLC on a calibrated Superose 12 column revealed major peaks at ˜20.00 mins for each construct (data not shown), consistent with the retention time of trimer. 8M urea or other disaggregating reagents are used to dissociate and prevent the formation of homotrimers. Mixing the purified CTLA-4-0-CD86, CTLA-4-0-UV-3 VL and UV-3 VH-0-CD86 Ig-like V domains and removing the disaggregating reagent by gel filtration or dialysis forms the trispecific trimer.
Design of scFv-0 Molecules Lacking a Foreign Flexible Linker Polypeptide
The design of VH-VL and VL-VH ligations was initially based on the precise distances between N- and C-terminal residues from the crystal structure of NC10 scFv-15 (Kortt et al, 1994). Previous studies have investigated the design of flexible linker peptides to join VH and VL domains to produce scFvs (Huston et al, 1991; Ragg and Whitlow, 1995), and the effect of the linker structure on the solution properties of scFvs (Holliger et al, 1993; Desplancq et al, 1994; Whitlow et al, 1994; Alfthan et al, 1995; Solar and Gershoni, 1995). ScFvs with the classical 15-residue linker, (Gly4 Ser)3 described by Huston et al, (1989, 1991) can exist as a monomers, dimers and higher molecular mass multimers (Holliger et al, 1993; Whitlow et al, 1994; Kortt et al, 1994). This propensity of scFvs to dimerise was exploited further by Whitlow et al, (1994) to make bispecific dimers by linking VH and VL domains of two different antibodies (4-4-20 and CC49) to form a mixed scFv and then forming an active heterodimer by refolding a mixture of the two scFv in the presence of 20% ethanol, 0.5 M guanidine hydrochloride. The main disadvantage of this approach with 15 residue or longer linkers is that different VH and VL pairings show different dimerization and dissociation rates. A variety of scFv-type constructs is illustrated in
A: An scFv comprising VH-L-VL where L is a linker polypeptide as described by Whitlow et al and WO 93/31789; by Ladner et al, U.S. Pat. No. 4,946,778 and WO 88/06630; and by McCafferty et al and by McCartney et al.
B: A single polypeptide VH-L1-VL-L2-VH-L3-VL which forms two scFv modules joined by linker polypeptide L2, and in which the VH and VL domains of each scFv module are joined by polypeptides L1 and L3 respectively. The design is described by Chang, AU-640863 and by George et al.
C: Two scFv molecules each comprising VH-L1-VL-L2 (a,b), in which the VH and VL domains are joined by linker polypeptide L1 and the two scFv domains are joined together by a C-terminal adhesive linkers L2a and L2b. The design is described by Pack et al, PI-93-258685.
D: This design of PCT/AU93/00491, which is clearly different to A, B and C above. A single scFv molecule VH-L-VL comprises a shortened linker polypeptide L which specifically prevents formation of scFvs of the type A, B or C, and instead forces self-association of two scFvs into a bivalent scFv dimer with two antigen combining sites (target-binding regions; TBR-A). The association of two different scFv molecules will form a bispecific diabody (TBRs-A,B).
Linkers of less than 12 residues are too short to permit pairing between VH and VL domains on the same chain, and have been used to force an intermolecular pairing of domains into dimers, termed diabodies (Holliger et al, 1993, 1996; Zhu et al, 1996; PCT/AU93/00491; WO 94/13804; WO 95/08577). Holliger et al, 1993, 1996, WO 94/13804 and WO 95/08577 described a model of scFv diabodies with VH domains joined back-to-back, and suggested that these structures required a linker of at least one or two residues. This model was confirmed in a crystal structure of a 5-residue diabody (Perisic et al, 1994), but it was noted that scFv-0 could not be fitted to this conformation, even with severe rotations of the VH domains. Desplancq et al, (1994) described a series of scFvs with linkers of 10, 5 and zero residues, and concluded on the basis of FPLC analyses that these scFvs were predominantly dimers with minor amounts of monomer, Alfthan et al (1995) also reported that scFvs with small linkers, down to 2 residues in length, formed dimers. McGuinness et al (1996) claimed that bispecific scFv-0 molecules were diabodies and could be displayed and selected from bacteriophage libraries. However, none of these studies performed precise molecular size determination on the expressed soluble products to confirm whether dimers were actually formed.
scFv Trimers
We have now discovered that the NC10 scFv-0 yielded a molecular mass on FPLC and sedimentation equilibrium analysis of 70 kDa, significantly higher than expected for a dimer (52 kDa), and less than that for a trimer (78.5 kDa) (Table 2). Binding experiments with anti-idiotype 3-2G12 Fab′ showed that the scFv-0 formed a complex of Mr of 212 kDa, consistent with three Fab′ fragments binding per scFv-0. This result confirmed that the 70 kDa NC10 scFv-0 was a trimer, and that three pairs of VH and VL domains interact to form three active antigen-combining sites (TBRs). This scFv-0 structure showed no propensity to form higher molecular mass multimers. The NC10 scFv-0 trimer also bound to neuraminidase, but the arrangement of the antigen combining sites is such that a second antigen binding site on NC10 scFv-0 could not cross-link the neuraminidase tetramers into ‘sandwiches’, as shown for the scFv-10 and scFv-5 dimers in
A computer graphic model, shown in
In these models of NC10 scFv-0 trimers (
In contrast, dimeric structures have been proposed for scFv-0 in which only VH domains are in contact between Fv modules (Perisic et al, 1994). These dimeric structures impose severe steric constraints when the linker is less than 3 residues in length. Our data show that trimers are exclusively favoured over dimers for both NC10 scFv-0 and 11-1G10 scFv-0, Steric constraints probably prevent the dimer formation and result in the trimeric configuration as the generally preferred conformation for scFv-0 molecules.
Binding Affinities of scFvs
Binding studies using the BIAcore™ biosensor showed that all the scFvs tested bound to immobilised anti-idiotype 3-2G12 Fab′. In the case where the dimers and trimer were used as analyte, the kinetic constants were not evaluated because multivalent binding resulted in an avidity effect and invalidated the kinetic interaction model. Experiments with the immobilised NC10 scFv-0 showed that the affinity of each antigen combining site (TBR) for anti-idiotype 3-2G12 Fab′ was essentially identical (Table 4), and that the increases in affinity shown in
The gain in affinity through multivalent binding (avidity) makes these multimeric scFvs attractive as therapeutic and diagnostic reagents. Furthermore, the construction of tricistronic expression vectors enables the production of trispecific scFv-0 reagents.
In conclusion, this specification shows that linkers of 10 or 5 residues joining the NC10 VH and VL domains result in the exclusive formation of bivalent dimers. The pairing of VH and VL domains from different molecules results in non-covalently crossed diabodies. For the scFv-5 and scFv-10 constructs monomers do not form, and any observed monomeric species are proteolytically-produced Fv fragments. The direct linkage of NC10 VH and VL domains as scFv-0 produced a trimer, with three antigen combining sites (TBRs) capable of binding antigen. Previous scFv-0 constructs have been reported to be dimers, which suggests that C-terminus and N-terminus residues in those constructs have some flexibility and may act as a short linker (Holliger et al, 1993). Indeed, the allowed flexibility between Fv modules of a 5-residue linked diabody has recently been modelled (Holliger et al, 1996), and presumably linkers of less than 5 residues would severely restrict this flexibility.
We initially thought that the trimeric conformation was unique to NC10 scFv-0, perhaps due to steric clashes between V-domains which prevented the dimeric association. However, we show in this specification that NC10 scFv molecules linked with up to 2 flexible residues between the V-domains also form trimers. We also show that the reverse orientation, for NC10 scFv-0 VL-VH is a trimer, but can also be a tetramer. Furthermore, we show that a second scFv-0 in VH-VL orientation, constructed from the anti-idiotype 11-1G10 antibody, can be a trimer, and possess trivalent specificity. We also show that a third scFv-0 in VH-VL orientation, constructed from the C215 antibody, can also form a trimer.
This specification describes methods of producing trimeric scFv-0 molecules constructed by direct ligation of two immunoglobulin-like domains, including but not limited to scFv-0 molecules in VH-VL and VL-VH orientations, and teaches the design of polyspecific reagents.
Ig-like V domains of non-antibody origin have also been joined without a linker in a construct equivalent to the scFv-0 to form trimers, and we have shown here the joining of CD86 (Ig-like V domain) to CTLA-4 (Ig-like V domain), as well as joining each of these to UV-3 VH and UV-3 VL respectively. The trimer formation by each of these constructs teaches that polyspecific and in this case trispecifc trimers can form as shown in
In the design of the trimeric NC10 scFv-0 residues Ser112 and Asp1 were ligated as a direct fusion of domains and, presumably, the absence of a flexible linker prevents the dimeric configuration. The C-terminal residue Ser112 was chosen from precise structural data, obtained by crystallographic analysis (Malby et al, 1994), as being immediately adjacent to the last residue constrained by hydrogen bonding to the VH domain framework before the start of the flexible hinge region. Similarly, Asp1 of VL was known to be hydrogen-bonded to the V-domain framework and was close to the antigen-binding site, but was not involved in antigen binding. Using a similar rationale, the NC10 scFv-0 VL-VH molecules were synthesised as a direct ligation of the C-terminal VL residue Arg107 to the N-terminal VH residue Gln1 (residues taken from Malby et al, 1994), and shown to associate into a stable trimer by FPLC analysis (
Since there are no structural data for 11-1G10, we assumed from structural homology that direct ligation of VH-Ser113 to VL-Gln1 would similarly prevent the formation of a flexible linker, unless there is unfolding of the terminal β-strands from the V-domain framework. The 11-1G10 scFv-0 exclusively formed trimers (
Taken together, our examples of scFv-0 molecules demonstrate that directly ligated VH-VL or VL-VH domains form trimeric scFv-0 molecules and in some cases, form a tetramer. The residues chosen for ligation of VH-VL or VL-VH should be close to the V-domain framework, and can either be determined experimentally, or can be predicted by homology to known Fv structures (Malby et al, 1994). Presumably, additional residues that form a more flexible linker will allow the formation of diabodies (Holliger et al, 1993; PCT/AU93/00491; WO 94/13804; WO 95/08577).
ScFv-0 molecules can be easily modelled into a symmetric trimeric conformation without interdomain steric constraints (
It will be appreciated by those skilled in the art that the effect of V-domain orientation and the requirement up to two residues in the flexible linker may be different for other scFv molecules, but that the preferred linker length and V-domain orientation can be easily determined using the designed iterative alterations described in this specification.
This specification predicts that the polymeric configuration, and particularly trimers and tetramers, will be the preferred stable conformation in many other scFv-0 molecules. The increased tumour to blood ratio reported for bivalent scFv dimers over monomers (Wu et al, 1996), presumably resulting from higher avidity and reduced clearance rates, offers advantages for imaging, diagnosis and therapy. The further gain in affinity through avidity makes trimeric and tetrameric scFvs attractive for in vivo imaging and tumour penetration as an alternative reagent to diabodies (Wu et al, 1996) and multivalent chemical conjugates (Antoniuw et al, 1996, Casey et al, 1996; Adams et al, 1993; McCartney et al, 1995).
The design of bivalent diabodies directly led to the design of bispecific diabodies using dicistronic vectors to express two different scFv molecules in situ, VHA-linker-VLB and VHB-linker-VLA, which associate to form TBRs with the specificities of the parent antibodies A and B from which the V-genes were isolated (Holliger et al, 1993, 1996; WO 94/13804; WO 95/08577). The linker sequence chosen for these bispecific diabodies, Gly4Ser, provided a flexible and hydrophilic hinge.
In a similar process, and using the inventive steps described in this specification, tricistronic vectors can be designed to express three different scFv-0 molecules in situ, VHA-VLB, VHB-VLC, and VHC-VLA which will associate to form a trispecific trimer with TBRs equivalent to the parent antibodies A,B,C from which the V-genes have been obtained. The three VH-VL scFv-0 molecules can associate into a trispecific trimer in a schematic configuration similar to that shown in
Similarly, tetramers with four active TBRs can be formed by association of four scFv identical molecules, and tetraspecific tetrabodies can be formed by association of four different scFv molecules, preferably expressed simultaneously from tetracistronic vectors.
It will be apparent to the person skilled in the art that while the invention has been described in some detail for the purposes of clarity and understanding, various modifications and alterations to the embodiments and methods described herein may be made without departing from the scope of the inventive concept disclosed in this specification.
Reference cited herein are listed on the following pages, and are incorporated herein by this reference.
| Number | Date | Country | Kind |
|---|---|---|---|
| PL 4973 | Sep 1992 | AU | national |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 10367956 | Feb 2003 | US |
| Child | 11692643 | US | |
| Parent | 09147142 | Mar 1999 | US |
| Child | 10367956 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 08403853 | May 1995 | US |
| Child | 09147142 | US |
