WO1996001319A1 - Variants of leukemia inhibitory factor - Google Patents

Abstract

The invention provides proteins which are variants of human Leukemia Inhibitory Factor (hLIF) with altered binding affinities to the LIF receptor and to the gp130 receptor. Such proteins include hLIF which has been altered by substitution at, insertion into, or deletion within a portion of said sequence such that the affinity for the LIF receptor is lowered compared to LIF, or a fragment of said protein which contains said alteration. Such proteins also include hLIF which has been altered by substitution at, insertion into, or deletion within a portion of said sequence such that the affinity for the gp130 receptor is lowered compared to LIF, or a fragment of said protein which contains said alteration.

Description

VARIANTS OF LEUKEMIA INHIBITORY FACTOR

The present invention relates to artificial variants of Leukemia Inhibitory Factor (LIF) with agonist or antagonist activity, and hybrid proteins incorporating domains from LIF

Leukaemia inhibitory factor (LIF) is a secreted polyfunctional cytokine which elicits a diversity of biological effects on many cell types. LIF expression has been detected in a variety of cell lines and primary tissues, including primordial germ cells, neurons, embryonic stem cells, adipocytes, hepatocytes and osteoblasts (reviewed Metcalf, 1992 (Growth Factors 7, 169-173); Heath, 1992 (Nature 359, 17)), and stimulated T lymphocytes and monocytes, brain glial cells, liver fibroblasts, bone marrow stromal cells, thymic epithelial cells and uterine endometrial gland cells just prior to blastocyst implantation. Non-functional mutants have also revealed an essential role for LIF in the process of embryonic implantation in the mouse (Stewart et al., 1992, (Nature 359, 76-79); Escary et al., 1993 (Nature 363, 361-364)).

The biological activities attributed to LIF are widespread. LIF can inhibit proliferation and induce macrophage differentiation in the murine leukemic myeloid cell line M1. It can synergize with IL-3 to stimulate the production of primitive hematopoietic progenitor and megakaryocyte colonies in vitro and increase the number of megakaryocytes and platelets in vivo. LIF can enhance the synthesis of acute phase proteins in liver hepatocytes stimulate the proliferation of myoblasts in culture and inhibit the lipoprotein iipase that mediates the transport to, and accumulation in, adipocytes. In neuronal developments, LIF is a cholinergic neuronal differentiation factor that up-regulates neuropeptides and acetylcholine synthesis in sympathetic neurons in vitro and in vivo. In addition, LIF is a neurotrophic, as well as a survival, factor in sensory neuron development in vitro. In bone metabolism, LIF is a paracrine or autocrine regulator that acts directly on osteoblasts, and indirectly on osteoclasts to inhibit bone resorption and regulate bone formation. LIF inhibits the differentiation and maintains the pluripotent developmental potential of embryonic stem cells. In LIF "knockout" mice, females have a specific defect in uterine function that prevents implantation of the blastocyst. Overexpression of LIF in mouse embryos, however, will result in inhibition of differentiation of embryonic ectodermal cells.

Both human and mouse LIF cDNAs encode mature LIF polypeptides of 180 amino acid residues with a predicted molecular mass of approximately 20kDa. The mature form of the LIF protein is present as a monomer in solution. In eukaryotic cells LIF is subjected to extensive post-translational glycosylation. Native human and mouse LIF are highly glycosylated (Moreau et al., 1988, (Nature 336, 690-692); Smith et al., 1988 (Nature 336, 688-690)), single chain molecules varying in molecular masses from approximately 38-67 kDa. Both human and murine LIFs have multiple N- and O-linked glycosylation sites and six conserved cysteine residues that are involved in three intramolecular disulfide bridges. The non-glycosylated, E.coli-expressed, recombinant human LIF is indistinguishable from native LIF in its biological activities in vitro. Human and murine mature LIF exhibit a 78% sequence identity at the amino acid level, whereas human LIF is equally active on both human and mouse cells, murine LIF is approximately 1000 times less active on human cells.

Sequence comparisons, analysis of genomic organisation and structure predictions have led to the proposal (Bazan, 1991 Neuron 7, 197-208) that LIF is a member of a family of ligands, which are characterized by a four a-helix bundle topology. The prototype member of this family is growth hormone (GH) and subsequent structural studies have confirmed that several other proposed members of this set of molecules conform to the predicted four helix structure (reviewed by Sprang and Bazan, 1993, Curr. Op. Struct. Biol. 3, 815-827). Extensive structural and functional analyses of the interaction between GH and its receptor (de Vos et al., 1992, Science 255, 306-312; reviewed by Wells and de Vos, 1993, Annu. Rev. Biophys. Biomol. Struct. 22, 329-351) have revealed that the functional GH signalling complex is composed of a single molecule of ligand complexed to two molecules of receptor. Receptor subunit homodimerisation, and consequential signalling, is brought about by the sequential interaction of two physically distinct sites on the ligand with essentially similar binding sites on each receptor subunit.

The biological effects of LIF are mediated by interaction (k_d ~ 10^-9 M) between the ligand and a specific LIF receptor subunit (LIF-R: Gearing et al., 1991, EMBO J. 10, 2839-2848) which is a member of the 'cytokine-binding' family of receptor subunits (reviewed by Cosman et al., 1990, Trends Biochem. Sci. 15, 265-270; Cosman, 1993 Cytokine 5, 95-106) characterised by a conserved pattern of cysteine residues and a WSXWS amino acid motif. LIF-R also contains a cytoplasmic motif which has been found to be common to other signal transducing elements (Taga and Kishimoto, 1992 FASEB J. 6, 3387-3396). Formation of a high affinity signalling complex (K_d ~ 10^-11 M) requires the association of the LIF/LIF-R complex with another transmembrane signal transducing molecule gp130 (Gearing et al., 1992 Science 255, 1434-1437; Gearing et al., 1992, Ciba. Found. Symp. 167, 245-255) which itself exhibits features of the cytokine family of receptors (Hibi et al., 1990). Recent results indicate that intracellular signal transduction mediated by LIF-R/gp130 heterodimerisation involves activation of members of the JAK family of cytoplasmic tyrosine kinases (Lutticken et al., 1994 Science 263, 89-92; Stahl et al., 1994 Science 263, 92-95). Some of the abovementioned activities of LIF are shared with four other cytokines: oncostatin M (OSM: Malik et al., 1989 Mol. Cell Biol. 9, 2847-2853), interleukin 6 (IL-6: Van Snick et al., 1988 Eur. J. Immunol. 18, 193-197), ciliary neurotrophic factor (CNTF: Stockli et al., 1989 Nature 342, 920-923), and interleukin 11 (IL-11: Paul et al., 1990 Proc. Natl. Acad. Sci. USA 87, 7512-7516; Kawashima et al., 1991 FEBS Lett. 283, 199-202).

Interleukin 6 (IL-6) is a multifunctional protein produced by lymphoid and non-lymphoid cells as well as by normal and transformed cells. OSM is a pleiotropic cytokine that can affect the growth and differentiation of a variety of normal and tumor cells. IL-11 is a growth factor with multiple effects on both hematopoietic and nonhematopoietic cell populations.

The molecular basis of these common functions appears to be the shared use of gp130 in all cases, and LIF-R in some cases, in their mechanism of action. Direct or indirect evidence indicates that LIF, OSM and IL-11 all generate intracellular signals as a result of ligand-mediated heterodimerisation of a ligand-specific transmembrane receptor subunit and gp130 (Liu et al., 1992 Biol. Chem. 267, 16763- 16766; Gearing et al 1992, ibid; Yin et al., 1993 Immunol. 151, 2555-2561; Taga et al., 1992 Proc. Natl. Acad. Sci. USA 89, 10998-11001).

Besides its growth inhibitor activities on the human A375 melanoma and other solid tumor cells, OSM also has grown stimulatory activities on normal fibroblasts, AIDS-Kaposi's sarcoma derived cells, and a human erythroleukemia cell line, TF-1. Other OSM-mediated activities include: the induction of differentiation of the M1 murine myeloid leukemia cell line; the inhibition of differentiation of murine embryonic stem (ES) cells; the stimulation of plasminogen activator activity in cultured bovine aortic endothelial cells as well as in human synovial fibroblasts; the stimulation of acute phase protein production in primary hepatocytes; and LDL uptake and up-regulation of cell surface LDL receptors in HepG2 cells. OSM is a heat and acid stable, single chain glycoprotein. cDNA derived from U937 lymphoma cells predicts a 252 amino acid (aa) residue peptide containing a 25 amino acid hydrophobic N-terminal signal sequence peptide. Cleavage of the signal sequence results in a 227 amino acid pro-cytokine that undergoes C-terminal processing to form a mature peptide 195 to 196 amino acid in length. Two potential N-glycosylation sites and five cysteine residues are present in the mature protein.

IL-6 and CNTF show a more complex involvement of gp130 in ligand mediated signalling. IL-6 interacts with the IL-6 receptor subunit (IL6-R) which is frequently found expressed in a soluble form (reviewed by Kishimoto et al., 1992, Science 258, 593-597; Taga and Kishimoto, 1992, ibid). The IL-6/IL6-R complex mediates cellular signalling by formation of gp 130 homodimers (Murakami et al., 1993, Science 260, 1808-1810). The effects of IL-6 on different cells are numerous and varied. The effect on B cells is to stimulate differentiation and antibody secretion. IL-6 also affects T cells, acting as a co-stimulant with sub- optimal concentrations of PHA or Con A to stimulate IL-2 production and IL-2 receptor expression. IL-6 exhibits growth factor activity for mature thymic or peripheral T cells and reportedly enhances the differentiation of cytotoxic T cells in the presence of IL-2 or IFN-γ.

The human IL-6 cDNA sequence predicts a precursor protein of 212 amino acids with two potential N-glycosylation sites. The hydrophobic N-terminal 28 amino acid signal peptide is cleaved to produce a mature protein of 184 amino acids with four cysteine residues and a predicted molecular weight of 21 kDA.

IL-11 is a pleiotropic cytokine that was originally detected in medium conditioned by an IL-1α- stimulated primate bone marrow stromal cell line, PU-34, by its ability to stimulate the proliferation of the IL-6-dependent murine plasmacytoma cell line T1165.85.2.1 in the presence of excess neutralizing anti-IL-6 antibodies. Similar to IL-1, IL-6, G-CSF, and c-kit ligand, IL-11 has blast cell growth factor activity and can synergize with IL-3 and IL-4 to shorten the G_o period of early hematopoietic progenitors. Although IL-11 alone will not support the growth of megakaryocyte colonies, it will synergize with IL-3 to increase the number, size and average ploidy value of megakaryocyte colonies formed from bone marrow cells.

The human IL-11 cDNA encodes a 199 amino acid precursor polypeptide with a 21 amino acid residue hydrophobic signal peptide that is processed proteolytically to generate the 178 amino acid residue mature form of IL-11. CNTF acts by heterodimerisation of gp130 with LIF-R which is arbitrated by a complex of CNTF and a specific CNTF receptor subunit (CNTF-R: Davis et al., 1991, Science 253, 59-63; Ip et al., 1992, Cell 69, 1121-1132; Davis et al., 1993, Science 260, 1805-1808) which can also exist in a soluble form. In both cases, a non-signalling, specificity-determining receptor subunit interacts with two transmembrane transducing components which include at least one molecule of gp130. LIF is most similar in sequence to the "LIF-R binding" cytokines OSM and CNTF (Bazan, 1991, Neuron 7, 197- 208), in particular human LIF exhibits about 20% amino acid identity with OSM (Rose and Bruce, 1991, Proc. Natl. Acad. Sci. USA 88, 8641-8645; Bruce et al., 1992, Prog. Growth Factor Res. 4, 157-170). Cardiotrophin (CT) is a cytokine which also binds through the gp130 receptor system. CT induces a hypertrophic (swelling) response in heart cells (Pennica et al, Proc. Natl. Acad. Sci. USA., 1995, 92, 1142-6). Taken together these findings suggest that ligand-mediated homodimerisation of gp130, or heterodimerisation of gp130 with another transmembrane receptor subunit, is essential for signal transduction in this group of ligands (reviewed by Stahl and Yancopoulos, 1993, Cell 74, 587-590). By analogy with the GH model it might therefore be expected that the 'gp130-dependent' set of cytokines should exhibit two functionally distinct sites of interaction; one which is involved in interaction with the ligand specific receptor subunit, and the second which interacts with the common gp130 transducer. The participation of a third specificity conferring receptor subunit in signalling, via the formation of a trimolecular receptor complex, may involve a third site on the ligand. Despite extensive functional analysis of gp130-mediated signalling (reviewed by; Kishimoto et al., 1992, ibid; Taga and Kishimoto, 1992, ibid), very little is known of the structural basis of ligand/receptor interactions in this system. Such information is essential both for understanding the molecular basis of signal transduction and the future design of ligand and family-specific response modifiers with therapeutic potential. Owczarek et al., 1993 (EMBO J. 12, 3487-3495), have published mutagenesis data relating to LIF.

It is probable that murine LIF, human LIF and OSM exhibit closely related three dimensional structures. It has been established, however, that murine and human LIF exhibit distinct differences in biological specificity. Whilst human LIF is able to bind to murine LIF-R and exert a biological effect on murine cells (Smith et al., 1988, ibid; Moreau et al., 1988, ibid) murine LIF exhibits relatively low affinity for human LIF-R (Owczarek et al., 1993, ibid) and, as described below, exhibits significantly reduced potency in bioassays dependant upon human LIF-R and human gp130. Moreover, OSM has been shown to bind LIF-R and gp130 albeit with lower affinity than LIF itself (Gearing and Bruce, 1992, New Biol. 4, 61-65; Baumann et al 1993, Biol. Chem. 268, 8414-8417).

In the present invention, the 3-dimensional structure of LIF has been established for the first time. This has enabled us to identify those regions of LIF which are crucial to the interaction between LIF and its receptors, i.e. gp130 and LIF-R, and to examine the regions of LIF which may account for the differences between LIF and related molecules mentioned above.

Brief Description of the Drawings.

Figure 1 A: Representation of the three dimensional structure of human LIF determined by crystallisation studies. The positions of residues of particular interest, described in the specification, are indicated. B: Representation of LIF from opposite orientation to (A). Figure 2: Sequence alignments for human and murine LIF (SEQ ID Nos. 1 & 2 respectively), human OSM (SEQ ID No. 3), human CNTF (SEQ ID No. 4) and human cardiotrophin (hCT, SEQ ID No. 5) based on structural considerations. Conserved residues are highlighted in dark shading and conservative substitutions are shown in light shading. Large open boxes represent the positions of the helices in murine LIF, and the residue numbers refer to the murine LIF sequence. The graph below the alignment indicates the solvent accessibility of each murine LIF residue in gradations of 20, 40, 60, 80 ϋ²

Figure 3. A) Composition of human/mouse LIF chimeras. For amino acid numbering refer to figure 2. B) Biological activity of LIF mutants in human Ba/F3 and murine Da-1a assays. Values represent the mean of triplicate samples. Standard deviation of replicate samples was less than 10% of the mean. C & D) Competitive inhibition of ¹²⁵I labelled human LIF binding to COS cells transfected with the human LIF-R by C) M, H, HM, MH D) H, 150-180, MH, OSM, 161-180, and V177A. Results are expressed as the number of counts bound to the cells at a particular concentration of unlabelled competitor (B) divided by the number of counts bound to the cells when no unlabelled competitor was present (B_o). Values represent the mean of three experiments.

Figure 4: Antagonism of hLIF by a site 2 variant of hLIF. Figure 5: Antagonism of OSM by a site 2 variant of hLIF

The present inventors have identified 3 sites which mediate the binding of LIF to its receptor. Site 3 is from residues 150 to 160, site 2 comprises residues 25-38 and 120-128, and site 1 comprises residues 161 to 180. Unless specified to the contrary, the numbering herein refers to that contained in the human sequence set out above. It has now been found that sites 1 and 3 are responsible for the interaction of LIF to LIF-R, whereas site 2 is responsible for the interaction of LIF to gρ130. This has enabled the provision of variants of LIF which have either a higher or lower affinity for the components of the LIF receptor than natural LIF. Such variants can be used as LIF agonists or antagonists depending upon their affinity. In particular variations in sites 1 and 3 provide agonists of LIF which enhanced binding to the receptor. Variations in site 2 provide LIF antagonists. The present invention thus provides a protein which comprises the sequence of human Leukemia Inhibitory Factor (hLIF (SEQ ID No. 1)) which has been altered by substitution at, insertion into, or deletion within a portion of said sequence such that the affinity for the LIF receptor is lowered compared to LIF, or a fragment of said protein which contains said alteration. Preferably the protein or fragment thereof has a binding affinity for gp130 which is substantially the same as hLIF.

Such proteins desirably contain alterations from the hLIF sequence at or within one or more of the residues of site 1 from amino acids 161 to 180 and/or of the residues of site 3 from 150 to 160. Desirably, the affinity for such proteins ("site 1 or 3 mutants") for LIF is lowered at least about 2, preferably 10 fold compared to LIF. Preferably, such proteins thereof will have an EC50 protein / EC50 hLIF binding ratio (as measured in the examples) of at least 40, and preferably greater than 100. Such proteins thereof will also preferably have a EC50 protein / EC50 gp130 binding ratio of between about from 0.01 to 10, e.g. from 0.1 to 5.0 or from 0.5 to 5.0.

The invention further provides a protein which comprises the sequence of human Leukemia Inhibitory Factor (hLIF (SEQ ID No. 1)) which has been altered by substitution at, insertion into, or deletion within a portion of said sequence such that the affinity for the gp 130 receptor is lowered compared to LIF, or a fragment of said protein which contains said alteration. Preferably such a protein or fragment thereof has a binding affinity for LIF receptor which is substantially the same as hLIF.

Such proteins desirably contain alterations from the hLIF sequence at or within one or more of the residues of site 2 from amino acids 25 to 38 and/or from 120 to 128. Desirably, the affinity for such proteins ("site 2 mutants") for gp130 is lowered at least about 2, preferably 10 fold compared to LIF. Preferably, such proteins thereof will have an EC50 protein / EC50 gp130 binding ratio (as measured in the examples) of at least 10, and preferably greater than 10, e.g greater than 100. Such proteins thereof will also preferably have a EC50 protein / EC50 hLIF binding ratio of between about from 0.01 to 10, e.g. from 0.1 to 5.0 or from 0.5 to 5.0.

Preferred fragments of proteins of the invention will have the ability to bind to the LIF receptor or to gp130 in competition with LIF, or related proteins including murine LIF, oncostatin, ciliary neurotrophic factor or cardiotrophin. This can be measured by assaying the ability of LIF to stimulate the growth of cells in culture in the presence and absence of such fragments. Generally, fragments of LIF or its variants will be at least 10, preferably at least 15, for example 20, 25, 30, 40, 50, 60 or 100 amino acids in length. The fragments may be made by synthetic methods such as synthesis on a solid phase or by expression of a recombinant DNA in an expression vector in a host cell, wherein the vector comprises DNA encoding the peptide operably linked to a promoter compatible with the host cell. The vector may also contain transcription termination signals. Proteins of the invention may also be made by recombinant DNA technology, as described below. In addition, the present findings have enabled us to provide novel hybrid proteins which will require specific combinations of receptor molecules at the cell surface in order to initiate signalling via those receptors. This can provide novel proteins which can be targeted to a specific subset of cell types, in particular those which express a LIF receptor and a second receptor, for example and interleukin receptor.

Mammalian homologues or fragments thereof of LIF which have the alterations mentioned above in the corresponding residues of their sequences may also be used in the present invention. Such homologues can be obtained by routine cloning procedures, e.g. by using the hLIF cDNA sequence as a probe to obtain another mammalian LIF from a cDNA library made from cells of the mammal which express LIF. The human and mammalian LIF proteins may be altered using standard techniques of genetic engineering known per se (e.g. see Sambrook et al (Molecular Cloning: A Laboratory Manual, 1989) and which are further illustrated in the examples below.

The invention further provides a recombinant protein or fragment thereof which comprises a site (as defined above) from LIF or a variant thereof. For example, the recombinant protein may comprise a protein or fragment thereof of the invention which further comprises, at the N- or C- terminus, all or part of the sequence of a cytokine including a cytokine selected from the group consisting of murine LIF (SEQ ID No. 2), Oncostatin (SEQ ID No. 3), ciliary neurotrophic factor (SEQ ID No. 4) and cardiotrophin (SEQ ID No. 5).

When such recombinant proteins are based upon site 1 or 3 mutants of LIF it is desirable that their affinity for LIF is lowered at least about 2, preferably 10 fold compared to LIF. Preferably, such proteins thereof will have an EC50 protein / EC50 hLIF binding ratio (as measured m the examples) of at least 40, and preferably greater than 100. Such proteins thereof will also preferably have a EC50 protein / EC50 gp130 binding ratio of between about from 0.01 to 10, e.g. from 0.1 to 5.0 or from 0.5 to 5.0. Likewise, recombinant proteins based upon site 2 mutants preferably have an affinity for gp130 which is lowered at least about 2, preferably 10 fold compared to LIF. Preferably, such proteins thereof will have an EC50 protein / EC50 gp130 binding ratio (as measured in the examples) of at least 10, and preferably greater than 10, e.g greater than 100. Such proteins thereof will also preferably have a EC50 protein / EC50 hLIF binding ratio of between about from 0.01 to 10, e.g. from 0.1 to 5.0 or from 0.5 to 5.0. Such recombinant proteins based on LIF proteins of the invention may have the general structure:

X¹-A-X²

where X¹ is an N-terminal sequence of amino acids or hydrogen, A is amino acids 25-38 of LIF or a variant thereof, and X² is a C-terminal sequence or a carboxy group; provided that the protein is not naturally occurring mammalian LIF.

X¹ is preferably the sequence of human or murine LIF from the N-terminal to the amino acid 24. X² may comprise the sequence of human or murine LIF, or a chimera thereof, from amino acid 39 to the C-terminal of the human or murine LIF, or chimera thereof.

Preferably such a recombinant protein is of the formula:

X¹-A-Z-B-X²

where A, X¹ and X² are as defined above, B is amino acids 120-128 of LIF or variant of this region thereof and Z is a sequence of amino acids linking A and B.

In this embodiment of the invention, Z, X¹ and X² are preferably derived from the same protein, preferably another cytokine (e.g. OSM, IL-6, IL-11 or CNTF or cardiotrophin (CT)) which also interacts with gp130. Preferably, the regions A and B are grafted into the other cytokine in place of the regions of that cytokine which correspond to site 2 of LIF when that cytokine is aligned with LIF. In the case of OSM or CNTF this may be done by reference to Figure 2 of the examples.

In such a case, residues 3-16 and/or 96-104 of mature human CNTF (residues 19-32 and/or 112-120 of the SEQ ID No. 4) may be replaced by the residues 25-38 and/or 120-128 or variants thereof respectively of LIF. Similarly, residues 12-24 and/or 112-120 of mature human OSM (residues 41-53 and/or 141-149 of SEQ ID No. 3) may be replaced. Likewise, residues 29-42 and/or 127-135 (of SEQ ID No. 5) of human cardiotrophin may be replaced.

In a further embodiment of the invention, the site 1 and/or site 3 regions of LIF or variants thereof may be grafted into the corresponding regions of other cytokines (including those mentioned above). This will provide further recombinant proteins which interact with the LIF receptor but will also require another receptor at the cell surface in order to mediate their effect. This enables the hybrid molecule to be targeted to a specific subset of cells which express a particular combination of receptors.

Such recombinant proteins will have the general structure:

X¹-S1-X², or

X¹-S3-X², or

X¹-S3-Z-S1-X²,

where X¹, Z, and X², are as defined above, S1 is site 1 or a variant thereof, and S3 is site 3 or a variant thereof.

Thus, in one embodiment of the invention, residues 125-135 and/or 136-155 of mature human CNTF (146-156 and/or 157-176 of SEQ ID 4) may be replaced by the residues 150-160 and/or 161-180 or variants thereof respectively of LIF. Similarly, residues 150-160 and/or 161-180 of mature human OSM (179-189 and/or 190-209 of SEQ ID No. 3) may be replaced.

The other cytokine is preferably IL-6, IL-11 OSM, CNTF or CT. The cytokine may also be LIF of a different species, e.g. murine. The fusion proteins may be obtained by expression of recombinant DNA. cDNA sequences of the cytokines are available and can be manipulated to be spliced in-frame to the relevant portion of LIF or a variant thereof using techniques known per se in the art. For example, in the case of variants in which a site 1 and/or site 3 region or variant thereof is to be grafted into the framework of DNA encoding another cytokine, PCR primers directed to the regions X¹, SI, Z, S3 and X² may be made and these regions amplified separately. The primers can be designed to overlap or to contain restriction sites which may be used to splice the fragments together. In this sense, the process is analogous to preparing DNA encoding recombinant antibodies in which murine CDRs are spliced into a human framework region. Other variant and hybrid proteins according to the invention may be made in this manner. Reference may be made to Sambrook et al (Molecular Cloning: A Laboratory Manual, 1989) for details of such techniques.

Alterations to LIF which are amino acid substitutions are particularly preferred, although deletions or insertions or 1, 2, 3, 4, 5 or more amino acids are also possible. Preferably the variant comprises 1, 2, 3, 4, 5, 6, 7, 8, or 10 substitutions.

Preferred residues which may be substituted are:

In site 1: K170, V175 and V177;

In site 2: Q25, S28, Q32, S36, D120, 1121, G124 and S127; and

In site 3: T150, S151, K153, F156, K158 and K159. Preferred substitutions are in particular, in site 2, changes which result in an opposite charge where the residue being substituted has a charge. In sites 1 and 3 conserved changes are preferred. Examples of site 2 changes include D120 to K or R. Examples of site 1 or 3 changes include K170 to R170; V175 and/or V177 to G, A, L or I; T150 to S; S151 to T; K153 to R; F156 to Y; and K158 to R. In addition, any of the residues of sites 1, 2 or 3 may be changed to A.

Other conserved substitutions may be made according to the following table indicates conservative substitutions, where amino acids in the same group within the second column and preferably within the same line of the third column may be substituted for each other:

The present invention also provides a nucleic acid, e.g. a DNA, encoding a protein or fragment thereof of the invention, and a vector comprising such nucleic acid. The vector may be an expression vector, wherein said nucleic acid is operably linked to a promoter compatible with a host cell. The invention thus also provides a host cell which contains an expression vector of the invention. The host cell may be bacterial (e.g. E.coli), insect, yeast or mammalian (e.g. hamster or human).

Host cells of the invention may be used in a method of making a protein or fragment thereof of the invention which comprises culturing the host cell under conditions in which said protein or fragment thereof is expressed, and recovering the protein or fragment thereof in substantially isolated form. The protein or fragment thereof may be expressed as a fusion protein.

The invention further provides pharmaceutical formulations. Such formulations comprise a protein or fragment thereof of the invention together with a pharmaceutically acceptable carrier or diluent. Pharmaceutically acceptable carriers or diluents include those used in formulations suitable for oral or parenteral (e.g. intramuscular or intravenous) administration. The formulations may conveniently be presented in unit dosage form and may be prepared by any of the methods well known in the art of pharmacy. Such methods include the step of bringing into association the active ingredient with the carrier which constitutes one or more accessory ingredients. In general the formulations are prepared by uniformly and intimately bringing into association the active ingredient with liquid carriers or finely divided solid carriers or both, and then, if necessary, shaping the product.

For example, formulations suitable for parenteral administration include aqueous and non-aqueous sterile injection solutions which may contain anti-oxidants, buffers, bacteriostats and solutes which render the formulation isotonic with the blood of the intended recipient; and aqueous and non-aqueous sterile suspensions which may include suspending agents and thickening agents, and liposomes or other microparticulate systems which are designed to target the polypeptide to blood components or one or more organs.

Proteins of the invention may be used to target specific groups of cells in the body, in order to either stimulate or inhibit growth. Particular conditions which may be treated include neuronal disorders, including degenerative diseases of the nervous system such as Parkinson's Disease, conditions which require nerve regeneration following trauma, disorders of the blood system including leukemias, bone osteoporosis and weight loss. The proteins may also be useful in assisting embryonic implantation in IVF procedures.

Treatment of a patient with a protein or fragment thereof of the invention will comprise administering to a patient in need of treatment an effective amount of the protein or fragment thereof (or composition containing the protein). The proteins according to the invention may be administered to human patients or other mammals by any route appropriate to the condition to be treated, suitable routes including oral or parenteral (including intramuscular or intravenous), intradermal). It will be appreciated that the preferred route may vary with, for example, the condition of the recipient. For each of the above-indicated utilities and indications the amount required of the protein or fragment thereof will depend upon a number of factors including the severity and nature of the condition to be treated and the identity of the recipient and will ultimately be at the discretion of the attendant physician. In general, however, for each of these utilities and indications, a suitable, effective dose will be in the range 0.1 to 100 μg per kilogram body weight of recipient per day, preferably in the range 1 to 10 μg per kilogram body weight per day. The desired dose may if desired be presented as two, three, four or more sub-doses administered at appropriate intervals throughout the day. These sub-doses may be administered in unit dosage forms. Proteins and fragments thereof of the invention may also be used in in vitro screening methods to identify antagonists or agonists of LIF. Accordingly, the present invention provides a method of screening candidate agonist or antagonist substances of LIF which comprises bringing a candidate substance into contact with a cell responsive to LIF in the presence of LIF;

measuring the response of the cell;

comparing said response with the response when said candidate and a protein or fragment thereof of the invention is brought into contact with said cell; and

selecting those candidate substances which show agonist or antagonist activity. By using proteins or fragments thereof of the invention with the ability to bind either LIF-R or gp130 in the above manner, it is possible to select candidate substances which show true agonist or antagonist activity against LIF. This is because stimulation or inhibition of cell growth in the presence of LIF which is blocked or activated by the candidate substance can be compared to the activity of the candidate substance in the presence of a LIF protein variant or fragment thereof of the invention. This allows a distinction to be made between candidate substances which have a general effect on cells and those which interact with or through the LIF/gp130 receptor system.

Suitable candidate substances include peptides (e.g. of from 5 to 20 amino acids) based on part of the sequence of LIF or other cytokines, synthetic or naturally occurring pharmaceutical drugs or plant extracts.

The proteins and fragments thereof of the invention may also be used in in vitro screening methods to identify or characterize new cytokines which also bind to the gp130 receptor and/or LIF receptor. For example, proteins or fragments thereof of the invention can be used to antagonise the binding of cytokines to the gp130 receptor, providing information about the binding profile and mode of action of said cytokine. Screening programmes may also include the use of LIF variants of the invention to bind to soluble forms of the LIF receptor and/or gp 130 in solution.

The present invention also provides antibodies capable of binding to proteins and fragments thereof of the invention. Such antibodies desirably bind to the protein or fragments thereof of the invention with an affinity which is at least 10 fold, e.g. 100 fold or 1000 fold higher than their affinity to human and/or murine LIF. In the case of proteins of fragments thereof of the invention which also comprises sequences of another cytokine, the antibodies will also have an affinity for that cytdleine which is at least 10 fold, e.g. 100 fold or 1000 fold lower than their affinity to the protein or fragment thereof of the invention. The affinity of antibodies of the invention to a LIF and/or other cytokines may be determined by routine techniques known in the art per se. An antibody of the invention may be monoclonal or polyclonal. For the purposes of this invention, the term "antibody", unless specified to the contrary, includes fragments of whole antibodies which retain their binding activity for a protein or fragment thereof of the invention. Such fragments include Fv, F(ab') and F(ab')₂ fragments, as well as single chain antibodies. In addition, monoclonal antibodies according to the invention may be analyzed (eg. by DNA sequence analysis of the genes expressing such antibodies) and humanized antibody with complementarity determining regions of an antibody according to the invention may be made, for example in accordance with the methods disclosed in EP-A-0239400. Monoclonal antibodies may be prepared by conventional hybridoma technology using the proteins or peptide fragments thereof, as an immunogen or, in the case of modified antibodies or fragments, by recombinant DNA technology, eg by the expression in a suitable host vector of a DNA construct encoding the modified antibody or fragment operably linked to a promoter. Suitable host cells include bacterial (eg. E.coli), yeast, insect and mammalian. Polyclonal antibodies may also be prepared by conventional means which comprise inoculating a host animal, for example a rat or a rabbit, with a protein or fragment thereof of the invention and recovering immune serum.

The following examples provide an analysis of LIF function by "homolog-scanning" mutagenesis, and reveal two regions of the LIF molecule involved in receptor interaction and biological function. The first, located within the D-helix comprising residues 161-180, and the second, located between residues 150-160 at the C-terminus of the CD loop - two surface regions that are separated by the AB loop. There are significant differences between these findings and the study of Owczarek et al. (1993, ibid) who also analyzed a series of human/mouse LIF chimeras for interaction with human LIF-R in a similar COS cell transfection system. In particular, whilst Owczarek et al observed an increase in competition with substitution mutants containing human residues 130-160 (which would include the region 150-160 identified in this study) they did not detect an increase in competition for binding to human LIF-R with substitution mutants containing the region 161-180. Although this might reflect detailed differences in the methods employed it is noteworthy that restoration of activity by substitution of residues 161-180 in the present studies employed a molecular framework in which residues 1-98 were also of human origin whereas Owczarek et al employed a framework of murine origin. This implies that the interaction of LIF with LIF-R may also be influenced by species differences between residues in the N-terminal half of the molecule. In addition, Owczarek et al identified a secondary effect for binding to LIF-R within the region of LIF residues 103-130. Five of the six solvent exposed amino acid differences for residues 103-130 between human and mouse LIF cluster around the BC loop, a region that is spatially close to the turn into the D helix (residues 150-160) implicated in both studies. Example 1 - Crystallization Studies.

Murine LIF was expressed as a fusion protein with glutathione-S-transferase in E. coli strain JM109. Murine LIF cDNA encoding the mature form of the polypeptide was cloned into the bacterial expression vector pGEX-2T, as described by Mereau et al., 1993 Cell Biol. 122, 713-719. For large- scale protein inductions (30 L) cultures were grown in LB + ampicillin (100 mg/ml) at 37°C, 300 rpm until they reached mid-log phase (A₆₀₀=0.6-0.8). IPTG was then added to the culture to a final concentration of 0.1 mM. Cultures were incubated at 37°C for a further 3 hours. Intracellular fusion protein was recovered from cell extracts by affinity binding to a slurry of glutathione sepharose (glutathione sepharose 4B; Pharmacia; 100 ml solution of 50%) in MTPBS (150 mM NaCl, 16 mM Na₂HPO₄, 4 mM NaH₂PO₄, pH 7.3) for 2 hours at 4°C. This was followed by washing once with 5 bead volumes of 0.5% octyl-b-glucopyranoside in MTPBS, then one wash each, with 50 mM Tris.HCl, pH 8.5, 150 mM NaCl and 50 mM Tris.HCl, pH 8.5, 150 mM NaCl, 2.5 mM CaCl₂. Isolation of recombinant LIF was achieved by cleavage of the fusion protein with human thrombin (T3010; Sigma) whilst attached to the matrix in 50 mM Tris.HCl, pH 8.5, 150 mM NaCl, 2.5 mM CaCl₂. Thrombin was added to a final enzyme:protein ratio of 1:100 and digestion was carried out at room temperature for 6 h. After thrombin digestion, the supernatant was separated and combined with five washes of the gel matrix. The supernatant from this reaction was dialysed against two changes of 20 mM MES, pH 6.0, 12 h at 4°C.

Cleaved protein was further purified on a Mono S cationic exchange column. Elution was carried out with a linear gradient of 0-1 M NaCl in 20 mM MES, pH 6.0. Positive fractions from a single peak were then pooled and concentrated by ultrafiltration (Amicon membrane; molecular weight cut-off of 3,000 Da) to 10 mg/ml for use in crystallization trials. Time of flight mass spectrometry was performed on purified LIF after HPLC purification, using a Finnigan Lasermat (matrix assisted laser desorption, nitrogen laser at 337 nm). LIF samples ( ~ 50 pmol) were analyzed using a sinapinic acid (11 mg/ml) matrix at a sample to matrix molar ratio of 1:5000. Amino acid sequencing and mass spectroscopic analysis confirmed that the resultant protein comprised the mature LIF sequence with a single additional N-terminal glycine residue arising from the thrombin cleavage site. Bioassays demonstrated that this modified, non glycosylated form of LIF retained full biological activity. A broad survey of crystallization conditions yielded several microcrystalline forms. Optimisation of conditions produced a bouquet of narrow, needle-like LIF crystals from 40% PEG 8000, 5 mM MES pH 6.0, 15 mg/ml LIF. lOmM MES was also suitable. Macroseeding of individual "needles" reduced nucleation and reproducibly produced crystals of a sufficient size for diffraction studies.

Diffraction measurements, data processing and reduction, phasing, map interpretation, refinement and analysis of the crystals was performed as described in UK patent application 9413316.2 filed 1st July 1994 and in Robinson et al, 1994, Cell 77, 1101-1116. The structure of LIF was analyzed as reported in Robinson et al, ibid. LIF is a compact molecule with overall dimensions of approximately 22 ü x 28 ϋ x 46 ϋ. As had been predicted (Bazan, 1991, ibid), the LIF structure conforms to the up-up- down-down four helix bundle topology common to the hematopoietic growth factors (Figure 1). The structure thus comprises 4 main a-helices conventionally labelled A, B, C and D, linked by two long loops (AB and CD) and one short loop BC. This topological motif may be considered in terms of two pairs of antiparallel a-helices B:C and A:D. The B and C helices (29 and 27 residues respectively), are relatively straight and pack in a classic antiparallel manner, tilted to cross approximately half way down their length. The A helix (27 residues) and the shorter D helix (23 residues with an additional 3 C-terminal residues which depart from the classic hydrogen bonding pattern) exhibit pronounced kinks (most notably in helix A) which serve to maximise the region of close packed, solvent inaccessible core formed on addition of this antiparallel a-helix pair to the B:C pair to form the central four helix bundle. These kinks require breaks in the normal a-helix hydrogen bonding pattern, substitute hydrogen bonds are made to the polar sidechains of serines (Ser-36 in helix A and Ser-174 in helix D) and tightly bound water molecules (Figure 1). The compact core is predominantly composed of hydrophobic residues contributed by the four a-helices. However, prior to helix A, from Asn-9 to Leu-22, the N-terminal region is wrapped around the molecule; the long loops AB (the first part of which contains a fifth short a-helix A' ) and CD are similarly tightly packed against the four helix bundle. Thus these three regions also contribute to the molecular core. The N-terminal region is pinned to the four helix bundle at the bottom of helix C by two disulphide bridges (Cys-12 to Cys- 134 and Cys-18 to Cys-131). Similarly the third disulphide bridge (Cys-60 to Cys-163) tethers the first part of the AB loop to the top of helix D. This disulphide bonding pattern confirms that reported by Nicola et al., (1993, Biochem. Biophys. Res. Commun. 190, 20-26). The compact nature of the molecule is further emphasized on inspection of crystallographic temperature factors. These indicate no regions of high conformational mobility apart from the N-terminal region prior to Asn-21.

With the possible exception of the number of exposed valine residues the surface of murine LIF is relatively characteristic of that expected for a small globular protein and shows no pronounced clustering of positive or negative charge. Two proline residues are in the cis conformation; Pro-17 before a disulphide bridge in the N-terminal region and Pro-51 at the start of the AB loop. The helical cytokine structures have been classified (Boulay and Paul, 1993, Curr. Biol. 3, 573-581; Sprang and Bazan, 1993, ibid) in terms of two subgroups; short-chain (Sc) and long-chain (Lc) cytokines. Nine examples of structures based on the cytokine four helix bundle topology are already known of which three are dimeric molecules (macrophage colony stimulating factor, MCSF: Pandit et al., 1992, Science 258, 1358-1362; interleukin5, IL-5: Milburn et al., 1993, Nature 336, 172-176; interferon γ, IFNγ: Ealick et al., 1991, Science 252, 698-702; Samudzi et al., 1991, Biol. Chem. 266, 21791-21797) and one is a five helix bundle (interferon β, INF/3: Senda et al., 1992, EMBO J. 11, 3193-3201). Structural superpositions of the LIF structure with the monomeric Sc, cytokines granulocyte macrophage colony stimulating factor (GMCSF: Diederichs et al., 1991, Science 254, 1779-1782; Walter et al., 1992, Mol. Biol. 224, 1075-1085) interleukin 2 (IL-2: Bazan and McKay, 1992, Science 257, 410-413) and interleukin 4 (IL-4: Walter et al., 1992, Biol. Chem. 267, 20371- 20376; Wlodawer et al., 1992, FEBS Lett. 309, 59-64; Smith et al., 1992, Mol. Biol. 224, 899-904; Powers et al., 1993, Biochemistry 32, 6744-6762), reveal limited structural similarity to LIF (number of Ca equivalences ranging from 83 to 91 with rms distances 2.00-2.30 ϋ). The LIF structure clearly falls within the Lc cytokine subgroup with greatest structural similarity to GH (de Vos et al., 1992, ibid; number of Ca equivalences = 111, rms distance = 2.48 u) and, in particular, to GCSF (Hill et al., 1993; number of Ca equivalences = 117, rms distance = 1.92 ϋ). As has been previously noted for the helical cytokines (Sprang and Bazan, 1993, ibid) the structural homology between LIF, GCSF and GH is not echoed in any strict conservation of key residues or disulphide bonds.

Structural superpositions of LIF onto GH and GCSF were examined. The lengths of the four helices and their relative positions are, to a first approximation, conserved between these three molecules; as are the conformations and positions of major portions of the long AB and CD loops. However, detailed comparisons of the structures indicate several basic points of variation. The B and C helices and short BC loop superimpose well between structures apart from the helical conformation of the BC loop and the related kink in helix C which distinguish the GH structure. Helix D is a mm shorter in LIF but again matches the other structures well, particularly GCSF which has a similarly curved fourth helix. Indeed the major variation in the four helices of LIF compared to GH and GCSF occurs in helix A which exhibits a distinctive kink in LIF which is thus far unique to this helical cytokine. The first part of the AB loop in all three structures contains a short helical region and the C-terminal half of this loop superimposes well between LIF and GCSF, however, this loop is markedly shorter in LIF and this is manifested in the acute angle at which the first half of the loop crosses in front of the D helix. Thus the AB loop in LIF overlays the surface of the D helix at a point approximately one third of the way down its length rather than at its N-terminus as in GH. The conformation of the CD loop is similar in all three molecules, but in both GCSF and GH this long loop contains highly flexible regions. In contrast the CD loop in LIF has a well defined single mainchain conformation throughout its length, this rigidity appears to be inherent to the molecule since this region is not involved in lattice contacts within the crystal. Finally the N-terminal region in LIF, prior to helix A, follows a unique path wrapping around the base of the four helix bundle. Thus the major distinctive features of the LIF structure within the Lc cytokine family are the N-terminal region, the kink in helix A and the position of the AB loop on crossing helix D.

Example 2 - Sequence Comparisons

Figure 2 includes a sequence alignment between human and murine LIF. The positions of the a-helices and the degree of residue solvent accessibility are indicated based on the murine LIF structure. The two sequences differ for 39 residues with no insertions or deletions. Of these differences none seem likely to perturb the structure greatly. The majority of the changes are at solvent exposed residues which are distributed evenly over the surface of the molecule and cannot, taken in isolation, indicate the structural basis of the species specificity which is observed for binding to human LIF-R (Owczarek, et al., 1993, ibid and below).

From sequence alignment LIF was predicted to belong to the helical cytokine family and assigned to a subgroup which also comprises OSM and CNTF (Bazan, 1991), molecules that have subsequently been shown to bind LIF-R and hence are referred to here as the "LIF-R binding" subgroup. The LIF structure is the first to be determined for a member of this subgroup. A revised sequence alignment for LIF, OSM, CNTF and cardiotrophin (CT) based on conservation of major structoral features, is presented in Figure 2. Clearly for this level of sequence identity structural variations may occur between the molecules at several points. However, the alignment does show strong conservation of key structural residues along the lengths of all four helices, most notably in helix D, which supports the assumption of structural equivalence for residues in these regions. Given the location of the four main helices in the sequences gross comparisons may also be made with respect to the rest of the LIF structure. Thus relative to LIF the N-terminal region is truncated and the C-terminal region is extended for both OSM and CNTF. The AB loop in LIF and OSM is tethered by a disulphide bridge to the equivalent point on helix D. The BC loop is lengthened in OSM and, to a lesser extent, in CNTF.

The majority of the residues which are identical or conserved in nature (e.g. retaining apolar character) within the "LIF-R binding" subgroup contribute to the stability of the molecular core and are thus inaccessible to solvent. There are, however, a number of conserved residues which have over 20 ϋ² of accessible area in the LIF structure and thus contribute to the surface characteristics of the molecule. Of these five may be considered to be highly exposed to solvent (40 ϋ²of accessible area). In the LIF structure these correspond to Ser-36 (or Asp) which mediates the hydrogen bonding at the kink in helix A, Ala-117 (apolar) on helix C, Phe-156 (totally conserved), Arg-158 (or Lys) and Lys- 159 (totally conserved) which are at the start of helix D. These residues may have a functional role and as such are prime candidates for mutagenesis.

Example 3 - Expression vectors for human and chimeric LIF Human LIF was cloned into the pGEX-2T expression plasmid in an identical manner to murine LIF as set out in Example 1. Chimeric (human/murine) LIF proteins were constructed by taking advantage of a unique Sma I site located at analogous positions in both species of cDNA (human and murine) and unique Sma I and Eco RI sites located in the parental pGEX-2T vector at the 3' end of the LIF inserts. The HM and MH chimeras were constructed with Sma I restriction fragments and the chimeras 161-180 and 150-180 were constructed with Sma I/Eco RI fragments assembled by SOE (splicing by overlap extension; Higiuchi et al., 1988, Nucleic Acids Res. 16, 7351-7367). Subcloning resulted in the production of H-MLIF and M-HLIF plasmids pGEX-2T expression vector plasmids. The fragment and junctions were sequenced by the dideoxy chain termination method using Sequenase (US Biochemical) and Circumvent kits (NE Biolabs). HLIF and species chimeric LIF proteins were produced under similar conditions to those for murine LIF, with the exceptions that protein inductions were carried out at 22°C instead of 37°C, and a reverse phase step replaced the final Mono S column. Recombinant human OSM was purchased from Preprotech Inc. Example 4 - Site Directed H-M LIF proteins

Generation of site-directed mutants in the C-terminus of the human-murine LIF chimeric protein species was achieved by incorporation of mutagenic and selection primers directly into the pGEX-H- MLIF plasmid using the Transformer Site-Directed Mutagenesis kit (Clontech Laboratories Inc.). Mutagenic primers consisted of sequences of murine LIF containing the desired point mutation and selection primers contained sequences of pGEX-2T which converted a unique Eco RV site to an Eco RI site within the expression vector. Point mutations were made at amino acid positions: 152(D to G), 168(T to K), 174(S to A) and 177(V to A). All mutagenised H-MLIF fragments were sequenced by the dideoxy chain termination method using the Sequenase kit.

Example 5 - Functional Analysis of LIF mutants

The LIF mutants and chimeras of the Examples 3 and 4 were tested for their biological activity in two bioassays. The first was a murine bioassay based on the LIF dependant growth of the murine Da- la cell line (Moreau et al., 1988, ibid; Godard et al., 1988, Blood 6, 1618-1623) and the second, a human bioassay, based upon the ability of experimental molecules to support the multiplication of murine Ba/F3 cells co-transfected with human LIF-R and human gp130. Da-1a cells were maintained in RPMI 1640 (ICN Flow Laboratories) supplemented with glutamine (2 mM), penicillin (50 IU/ml), streptomycin (50 mg/ml), 10% FCS (selected batches) and 10 ng/ml recombinant mouse LIF. Ba/F3 [gp130 + LIF-R] cells were maintained in RPMI 1640 medium supplemented with 8% FCS, glutamine (2 mM), penicillin (50 IU/ml), streptomycin (50 mg/ml) and recombinant human LIF (25 ng/ml).

To make the transfected Ba/F3 cells, human gp130 (Professor T. Kishimoto, Osaka, Japan) was subcloned into the pRcNeo (InVitrogen) eukaryotic expression vector. Parent Ba/F3 cells (Palacios et al., 1985) were electroporated (960 mF, 360 V) in 0.8 ml PBS containing 20 mg of pRcNeogp130, and then subjected to G418 (Gibco-BRL) selection (1 mg/ml) in RPMI medium supplemented with 8% fetal calf serum (FCS: ICN Flow Laboratories) and murine IL3. WEHI-3 conditioned medium was added at 10% to the RPMI culture medium and used as a source of the murine IL3 (Lee et al., 1982, Immunol. 128, 2393-2398). The pKCSRa eukaryotic expression vector containing the cDNA encoding the human LIF-R (kindly provided by Dr M. M. Hallet) was then electroporated into the Ba/F3 [gp130] cells and positive clones were selected in culture medium supplemented with 240 IU/ml of human LIF the source of which was obtained from a transfectant CHO cell line expressing the human LIF protein (Moreau et al., 1988, ibid). After 15 days of culture, the Ba/F3 [LIF-R + gp130] LIF- dependent cell line was isolated. Biological assays were performed using the recombinant factors as previously described (Moreau et al., 1988, ibid; Godard et al., 1988, ibid). In brief, cells were washed 3 times with a large volume of LIF-free medium before being cultured at a density of 10⁵ cells/ml in RPMI medium supplemented with glutamine, penicillin, streptomycin and 10% FCS, in the presence of twofold successive dilutions of the factors to be tested. Assays were performed in triplicate. Cellular proliferation was assessed by MTT (Sigma) staining (Mosmann, 1983, Immunol. Methods 65, 55) after 72 h. Biological activity was determined with reference to a standardized sample of human LIF using the arbitrary definition of 1 unit of biological activity = 1 ng of human LIF

The results of these studies are shown in Figure 3. Using the murine Da- la assay, murine LIF, human LIF, the human/murine LIF chimeras and point mutants exhibited similar biological activities, confirming that the mutant molecules had retained structoral integrity. OSM was also biologically active in this assay but exhibited an approximately 3 fold lower specific activity than human LIF. The biological properties of these molecules did, however, exhibit significant differences when analyzed by their ability to stimulate Ba/F3 cell proliferation by means of association with the human LIF-R and gp130. In particular, murine LIF exhibited 30-fold lower activity than human LIF. This observation confirms that the observed species difference in activity between murine and human LIF arises from a difference in affinity for LIF-R and/or gp130. Two chimeras were tested which comprised either the N-terminal 'half of murine LIF (residues 0-98) and the C-terminus of human LIF (residues 99-180), or the reciprocal substitution of the N-terminal 'half' of human LIF and the C-terminal region of murine LIF (HM-LIF). In the Ba/F3 assay MH-LIF was similar to human LIF in activity whereas HM-LIF was equivalent to murine LIF in activity. These observations suggest that species-specific activity of LIF resides in the C-terminal region of the molecule (residues 99-180). HM-LIF was then subjected to further mutations in which increasing regions of murine C-terminal sequence were substituted for human sequence. In addition a series of point substitotions were made at individual amino acids which differed between human and murine LIF in sequence. The strategy behind this approach was to detect regions of murine LIF sequence in the HM chimera which could be 'activated', by substitotion mutagenesis, into functionality in the human bioassay thereby reconstructing regions of the molecule required for activity mediated by human LIF-R and gp130. Given the potentially discontinuous nature of the protein binding sites (Cunningham and Wells, 1993, Mol. Biol. 234, 554-563), HM-LIF was chosen as a template molecule for substitotion mutagenesis to permit the identification of sequences in the region 99-180 whose species-specific activity depended upon contributions from residues 1-98.

The results of these experiments are also shown in Figure 3. Substitution of residues 161-180 of HM- LIF by the equivalent human residues resulted in a 10-fold increase in specific activity, revealing that the sequence differences within the D helix contribute to the species difference between murine and human LIF. Point mutation of solvent exposed residue 177(Val to Ala) resulted in a 7-fold increase in the activity of the HM chimera. No increase in activity was observed by substitotion of buried residue 168(Thr to Lys). These observations reveal that residues 161-180, which comprise the C- terminal end of the D helix play an important, but partial, role in the species specific activity of LIF.

Additional activation, and almost complete recovery of activity in the human bioassay, occurred upon further substitotion of human for murine sequences in the region corresponding to the solvent exposed 'turn' at the C-terminal end of the CD loop (residues 150-160). No activation was observed by point substitotion of solvent exposed residue 152(Asp to Gly) suggesting that this residue does not contribute to species specificity. Additional substitution of CD loop sequences including substitotions of the 'proline rich' motif (residues 145-150) resulted in substantial loss of biological activity in both the human and mouse bioassays, suggesting that substitutions in this region may result in conformational perturbation. These observations, therefore, reveal a second region of the LIF molecule conferring species-specific activity which is located between residues 150-160, corresponding to the solvent- exposed turn at the C-terminal end of the CD loop and the top of the D helix. The activity of OSM in the BA/F3 assay was of particular interest due to the relatively low degree of sequence identity between human OSM and murine LIF and human LIF. OSM was able to stimulate the multiplication of the transfected BA/F3 cells with a specific activity approximately 2-fold lower than human LIF. OSM was, in this respect, more similar to human LIF than murine LIF. Hence residues which are common between OSM and human LIF but differ from murine LIF may be important for biological activity. According to the revised structure based sequence alignment of human LIF, murine LIF and OSM presented in figure 2 there are four residues conforming to this criterion, of which three have reasonable solvent accessibilities in the murine LIF structure: Val-56 (Leu in human LIF and OSM); His- 150 (Thr in human LIF and OSM) and Glu-154 (Asp in human LIF and OSM). His-150 and Glu-154 are located within the region of the CD loop demonstrated to be involved in the species specific activity of LIF, whilst Val-56 is in the small A' helix of the AB loop which crosses the D helix, separating the two functionally implicated regions.

Example 6 - LIF-R interaction

The above studies do not indicate whether the observed biological differences are due to the interaction of LIF with LIF-R, gp 130 or a combination of both. This was investigated by testing the ability of LIF mutants to compete for binding with ¹²⁵I labelled human LIF to human LIF-R expressed by transfection in COS cells.

Human LIF-R was transiently expressed in COS-7 cells by electroporating aliquots of 0.8 X 10⁷ cells at 330 mV and 500 μF in the presence of 30 μg of human LIF-R cDNA subcloned into the expression vector PXMT2. The surviving cells were plated at a density of 10⁵ cells. Seventy two hours later the cells were used in binding assays conducted as described previously (Mereau et al., 1993, ibid) with two modifications. The dilutions of unlabelled, LIF (0.01 nM to 300 nM) for competition were added prior to the 0.2 nM human ¹²⁵I-LIF and the cells were solubilized in 1 M NaOH in place of a detergent lysis buffer. Recombinant human LIF was radiolabelled with Iodobeads (Pierce) and was used at a specific activity of 375 cpm/fmole. These results of the binding experiments show that the affinity of most mutants for human LIF-R is comparable to their activity in the human BA/F3 bioassay (Figure 3b). Thus the MH mutant exhibited significantly greater ability (ED50= 13.5 ±2.9 nM) to compete with human ¹²⁵I-LIF binding to human LIF-R than the HM counterpart (ED50 > 100 nM). These findings are essentially similar to those of Owczarek et al. (1993, ibid). Further substitotion of residues 161-180 in the HM chimera resulted in a significant increase in binding to human LIF-R (ED50=81.7±6.9 nM) although, as observed in the biological investigations, this did not lead to complete restoration of function. The additional substitution of human residues 150-160 into the HM chimera (ED50=2.9 ± 0.4 nM) resulted in a mutant with equivalent activity to the MH chimera. This finding indicates that prominent sites of interaction with human LIF receptor are located in two regions, between residues 150-160 and 161- 180 respectively. The point mutant of HM-LIF 177(Val to Ala) provided an interesting exception to the parallel behaviour of mutants in the human bioassay and competition for binding to human LIF-R. Whilst this mutant exhibited an activation of biological function in the human bioassay it did not exhibit significant restoration of competition for binding to human LIF-R (ED50 > 100 nM). The basis for this discrepancy is not clear although it is worth noting that the requirements for the formation of a ligand-mediated high affinity LIF-R/gp130 complex over a period of time sufficient to elicit a biological response may involve structural features of the ligand additional to those required for simple interaction with LIF-R. Finally, in agreement with the findings of Gearing and Bruce (1992), OSM was also able to compete for human ^I25I-LIF binding to human LIF-R with a reduced apparent affinity (ED50=51.3 ±6.9 nM) compared to human LIF (ED50=0.96 ±0.05 nM).

The crystal structure of the GH ligand/receptor complex (De Vos et al., 1992, ibid) provides a paradigm for receptor binding by the helical cytokines. There are two receptor binding sites on GH. The higher affinity site I involves the AB loop and the C-terminal half of helix D; site II is formed by residues from helix A and helix C. The work of Cunningham and Wells, (1993, ibid) indicates that a small number of key residues within the receptor binding sites are critical in stabilizing the ligand/receptor complex. Both sites I and II in GH are binding sites for signal-transducing receptor components; the resultant subunit homodimerization initiates signal transduction. Data from site- directed mutagenesis studies of IL-2 (Zurawaski et al., 1993, EMBO J. 12, 5113-5119) and IL-4 (Kruse et al., 1993, EMBO J. 12, 5121-5129) provide good evidence that the equivalent sites I and II on these helical cytokines also serve as the binding sites of the signal-transducing components. It is therefore instructive to examine the current structure/function data for LIF in the context of the GH based model for receptor binding.

The mutagenesis data implicate residues in the region 161-180 of murine LIF in receptor interaction. This region would correspond in location to Site I of the GH model. The difference in activity between mutants in which the N terminal half is derived from either mouse (Owczarek et al., 1993, ibid) or human sequences also indicates that the activity of this site either includes, or is influenced by, additional sequences in the region 1-98. This could involve the region of the AB loop which crosses the D helix of LIF or regions of helix A which are able to interact with helix D by virtue of physical proximity. This assignment of site I therefore resembles the GH prototype. Homolog-scanning, however, has not revealed the binding site for gp130. A number of distinctive surface features of the LIF structure, such as the kink in helix A and Ser-36 (conserved Ser or Asp) and Ala-117 (conserved apolar) on helix C, cluster within site II of the GH receptor binding model. By analogy site II may be tentatively assigned as a gp130 binding site in the LIF sub-family of helical cytokines, which, in the absence of any species difference in activity in human/mouse LIF chimeras may be comprised of functionally conserved residues. However this classical two site model fails to account for all of the data obtained in the above examples. The data for human-murine LIF chimeras strongly implicate the C-terminus of the CD loop in species specific binding to LIF-R. Also, surface residues whose natures are conserved between LIF, OSM and CNTF cluster in this region. This part of the LIF structure is more exposed than the equivalent area in GH because of the different position of the AB loop. The additional area implicated in receptor binding could be viewed as an extension of the standard site I, however the receptor binding mode observed for GH cannot be trivially adjusted to accommodate extra interactions with this region.

Thus it is proposed that a single LIF-R molecule binds LIF at two sites, the classic GH site I and a new site designated site III (at the top of helix D). This model may have broader applicability since it also suggests a mode of binding for the specificity-determining components which are additionally required for formation of some cytokine ligand/receptor complexes (reviewed by Stahl and Yancopoulos, 1993, ibid). This 'three binding site' model extends the conceptual framework within which to consider further studies of receptor binding modes for the helical cytokine family.

Example 7 - LIF mutants

A series of plasmids encoding LIF mutants were made using the pGeX-hLif expression vector system described in Example 1. Mutant DNA sequences were created by PCR overlap (Ho et al., 1989) using pGeX-hLif and oligonucleotides containing an appropriate coding change for each individual mutant. The LIF mutants were expressed as glutathione-S-transferase fusion proteins in E. coli JM109 as described in Example 1.

Oncostatin-M, amino acids 1-196 was a kind gift from Dr. D Staunton, was also expressed as a glutathione-S-transferase fusion protein and purified as above.

The LIF mutants were tested for their ability to bind to LIF-R and to gp130, and for their ability to stimulate the Ba/F3 cell line, which expresses both hLIF-R and hGP130, as described in Example 5. To obtain LIF-R and gp130 receptor protein, expression plasmids were made. PCR was used to amplify the region coding for amino acids 2-538 of the human LIF-R. This fragment was cloned into the pIG plasmid (Simmons, 1992, Cloning Cell Surface Molecules by Transient Expression in Mammalian Cells in Cellular Interaction in Development. IRL Press) using a Hind III site 5' to the first ATG codon and a Bam HI site 3' to the codon for amino acids 528, both sites were introduced by the PCR primers. The PCR primer matching the N-terminus sequence also introduced an optimised Kosak sequence, while the other primer created a splice donor sequence between codon 528 and the Bam HI site. An identical approach was used to clone the human gp130 into the pIG plasmid, except the region amplified coded for residues 1-328. The nucleotide sequence of both constructs was confirmed by DNA sequencing with Sequenase (USB).

Both the human LIF-R and the gp130 receptor were expressed as fusion proteins with the Fc region (hinge-CH2-CH3) of human IgG-1. The pIG expression plasmids were transfected by the calcium- phosphate technique into a human epithelial kidney cell line 293T, which expresses the large T antigen of SV40 (Dubridge et al., 1987, Molecular and Cellular Biology 7, 379-387). After transfection of 293T cells the media was changed to a serum free media (Ultra-Cho, Biowhittaker) and the Fc-fiision proteins were left to accumulate in the media for six days. Receptor-Fc proteins were purified from clarified supernatants by chromatography on protein-A sepharose (Pharmacia). Elution of the receptor was achieved with 0.1 M citric acid, pH 3.0 and subsequent neutralisation with Tris base. Purity of the receptor-Fc proteins was assessed as >90%.

Recombinant human LIF produced as described above, was iodinated using iodobeads (Pierce) and ¹²⁵I-Iodine (Amersham) as recommended by the manufactorers. Iodinated LIF had a specific activity of 1-4 x 10³ cpm/fmol and was equally active in the bioassay as wild type hLIF. Recombinant Oncostatin-M was biotinylated with Biotinamidocaproate N-hydroxysuccinimide ester (Sigma) following the procedure of (Harlow and Lane, 1988, Antibodies: A laboratory manual. New York, Cold Spring Harbor Laboratory).

All binding assays were performed in Nunc maxisorp 96 well plates. For LIF-R binding, plates were first coated with 100 ml of protein A (Sigma) at 1 mg/ml, blocked with PBS-BSA (1%) and then incubated with a 100 ml of Lif-R also at lmg/ml. Initial experiments indicated that the supernatant from 293T cells transfected with LIF-R worked equally well as purified LIF-R, therefore all subsequent experiments used LIF-R containing supernatant. After LIF-R binding the plates were washed with PBS 0.05% Tween 20 and then used for the binding assay. Competition binding assays between ¹²⁵I-hLif and wildtype hLif or mutant hLif were performed in a volume of 60 ml in a buffer of RPMI (Life Technologies), 20 mM Hepes pH 7.2, 25 mg/ml BSA, 2 mg/ml azide, 1 mM PMSF and 1 mM EDTA overnight at 4°C. After washing twice with 200 ml PBS 0.05% Tween 20, bound ¹²⁵I-hLif was released by incubation in 80 ml of 1 M NaOH and counted in a gamma counter (LKB). All Lif-R binding studies were performed in triplicate for at least two independent experiments. Binding stodies to gp130 was performed in a similar manner to that for LIF-R. These assays differed in that for gp130 competition binding, biotinylated Oncostatin-M was used instead of ¹²⁵I-hLIF and the bound Oncotatin-M was detected by incubation with a streptavidin-horseradish peroxidase conjugate (Amersham). Specifically, after washing plates with PBS, 0.05% Tween 20 the wells were rinsed with PBS and then incubated with 100 ml of streptavidin-horseradish peroxidase (1/1000 dilution) in PBS-1 % BSA. After again washing with PBS the horseradish peroxidase was detected by incubation with the chomagen OPD (Orthophenylenediamine, Dako) according to manufactorers instructions. Absorbance was read at 492 nM in a 96 well plate reader (Anthos). All gp130 binding stodies were performed in duplicate for at least two independent experiments. Results from the binding experiments were first expressed as the concentration of mutant protein required to give 50% inhibition of binding, this figure was then expressed as a ratio with the concentration of hLif required to give 50% inhibition of binding.

Example 7a - Site 1 mutants.

In a first series of experiments, a number of single amino acid substitotions were made to the sequence of human LIF. The binding and stimulation results are shown in Table 1.

The results of from Table 1 show that mutations in site 3 of LIF (K153, F156 and K159) cause LIF to loose affinity to LIF-R, which in turn renders the mutant proteins biologically less active. However the site 3 mutations do not affect the ability of the protein to bind to gp130. -

Although not all site 3 mutants tested showed the same amount of decrease in activity, no further LIF mutants tested with changes outside site 3 (residues 150-160) showed any significant decrease in LIF-R binding ability.

Example 7b - Site 2 mutants.

In a further series of experiments, a number of multiple mutations were made in site 2 of LIF, to provide 6 different LIF mutants, 01-06. Mutants O1 and O2 contained amino acid variants in the C helix region of LIF. Mutants 03 and 04 contained changes in the A helix region within site 2 (residues 25-38). The mutations of O1 and 03 were combined to provide 05. Likewise, 06 contained the mutations of 02 and 04. The results are shown in Table 2.

The results of Table 2 show that binding to the LIF receptor is maintained by these mutants, even though biological activity is virtually eliminated because of the lack of affinity of the mutants to gp130. Thus these mutants will behave as LIF antagonists, competing with LIF for the LIF receptor.

Example 7c - antagonism of OSM and hLIF.

The 04 mutant was assayed for its ability to antagonise the action of OSM and hLIF in the Baf- LIFR/gp130 system described above. Antagonism of hLIF is shown in Figure 4, and of OSM in Figure 5. In both cases, increasing concentrations of the 04 LIF antagonist, particularly above 1, e.g 10 ng/ml, provided significant antagonism of hLIF and OSM.

Claims

1. A protein which comprises the sequence of human Leukemia Inhibitory Factor (hLIF (SEQ ID No. 1)) which has been altered by substitotion at, insertion into, or deletion within a portion of said sequence such that the affinity for the LIF receptor is lowered compared to LIF, or a fragment of said protein which contains said alteration.

2. A protein or fragment thereof according to claim 1 which has been altered at or within or more of the residues from 150 to 160 of SEQ ID No. 1.

3. A protein or fragment thereof according to claim 2 wherein said alterations comprise

substitotions at the positions selected from one or more of the group consisting of 153, 156 and 159.

4. A protein or fragment thereof according to any one of claims 1 to 3 wherein the protein or fragment thereof further comprises, at the N- or C- terminus, all or part of the sequence of a growth factor selected from the group consisting of murine LIF (SEQ ID No. 2), Oncostatin (SEQ ID No. 3), ciliary neurotrophic factor (SEQ ID No. 4) and cardiotrophin (SEQ ID No. 5).

5. A protein or fragment thereof according to any one of claims 1 to 4 which has a binding affinity for gp130 which is substantially the same as hLIF.

6. A protein which comprises the sequence of human Leukemia Inhibitory Factor (hLIF (SEQ ID No. 1)) which has been altered by substitution at, insertion into, or deletion within a portion of said sequence such that the affinity for the gp130 receptor is lowered compared to LIF, or a fragment of said protein which contains said alteration.

7. A protein or fragment thereof according to claim 6 which has been altered at or within or more of the residues from 25 to 38, or from 120 to 128 of SEQ ID No. 1.

8. A protein or fragment thereof according to claim 7 wherein said alterations comprise

substitutions at the positions selected from one or more of the group consisting of 25, 28, 32, 36, 120, 121, 124 and 127.

9. A protein or fragment thereof according to any one of claims 6 to 8 wherein the protein or fragment thereof further comprises, at the N- or C- terminus, all or part of the sequence of a growth factor selected from the group consisting of murine LIF (SEQ ID No. 2), Oncostatin (SEQ ID No. 3), ciliary neurotrophic factor (SEQ ID No. 4) and cardiotrophin (SEQ ID No. 5).

10. A protein or fragment thereof according to any one of claims 6 to 9 which has a binding affinity for LIF receptor which is substantially the same as hLIF.

11. A nucleic acid encoding a protein or fragment thereof according to any one of claims 1 to 10.

12. A vector comprising a nucleic acid according to claim 11.

13. A vector according to claim 12 which is an expression vector, wherein said nucleic acid is operably linked to a promoter compatible with a host cell.

14. A host cell which contains an expression vector according to claim 13.

15. A method of making a protein or fragment thereof as defined in any one of claims 1 to 10 which comprises cultoring the host cell of claim 14 under conditions in which said protein or fragment thereof is expressed, and recovering the protein or fragment thereof in substantially isolated form.

16. A composition comprising a protein or fragment thereof as defined in any one of claims 1 to 10 and a pharmaceutically acceptable carrier or diluent.

17. A protein or fragment thereof according to any one of claims 1 to 10 for use in a method of treatment of the human or animal body by therapy.

18. A method of screening candidate agonist or antagonist substances of LIF which comprises bringing a candidate substance into contact with a cell responsive to LIF in the presence of LIF;

measuring the response of the cell;

comparing said response with the response when said candidate and a protein or fragment thereof according to any one of claims 1 to 10 is brought into contact with said cell; and

selecting those candidate substances which show agonist or antagonist activity.

19. An antibody capable of binding a protein or fragment thereof of as defined in any one of claims 1 to 10.