NO179210B - Method for producing a LIF polypeptide, recombinant DNA molecule encoding the LIF polypeptide and host cell to express this - Google Patents

Description

The present invention relates to a method for producing a LIF (leukemia inhibitory factor) polypeptide, a recombinant DNA molecule that codes for a LIF polypeptide and a host cell to express this.

These and other features of the invention appear in the patent claims.

Mature blood cells are produced by clonal proliferation and simultaneous differentiation of immature precursor cells (Metcalf, D.

(1984) The Hemopoietic Colony Stimulating Factors, Elsevier, Amsterdam). For normal hemopoietic cells, the two processes of proliferation and differentiation are closely linked so that the mature cells with a short lifespan are continuously replenished. At least four biochemically distinct growth factors, the colony-stimulating factors (CSFs), have been shown in vitro to stimulate the proliferation and differentiation of progenitor cells of granulocytes and macrophages: G-CSF, M-CSF, GM-CSF and multi-CSF (IL-3) ( see Metcalf, D.

(1987), Proe. R. Soc. B. 230, 389-423, for further explanation).

Unlike normal cells, leukemic myeloid cells are characterized by uncoupled proliferation and differentiation so that immature stem cells accumulate, retain their proliferative capacity and fail to differentiate. There has been considerable controversy as to the type of biological factors capable of inducing the differentiation of myeloid leukemia cells in vitro and whether certain factors are capable of inducing differentiation in the absence of proliferation. It has thus been proposed that proliferation and differentiation are necessarily mediated by different factors (Sachs, L. (1982), J. Cell. Physiol. Suppl., 1, 151-164). On the one hand, two groups have described and purified an activity (MGI-2 or D factor) capable of inducing differentiation of the murine myeloid leukemia cell line M1, which does not stimulate proliferation of normal stem cells (Lipton, J.H. and Sachs, L., 1981), Biochem. Biophys. Acta. 673, 552-569; Tomida, M., Yamamoto-Yamaguchi, Y., and Hozumi, M. (1984) J. Biol. Chem. 259, 10978-10982). On the other hand, the present inventors have recently shown that one of the CSFs, namely G-CSF, is a strong differentiation-inducing stimulus for the murine myeloid leukemia cell line, WEHI-3B D<+>, as well as being a proliferative and differentiating telling stimulus for normal cells (Nicola, N.A., Metcalf, D., Matsumoto, M. and Johnson, G.R. (1983), J. Biol. Chem. 258, 9017-9023). Moreover, Tomida, et.al. (Tomida, M., Yamamoto-Yamaguchi, Y., Hozumi, M., Okabe, T. and Takaka, F.

(1986), FEBS Lett., 207, 271-275) showed that recombinant G-CSF is also able to induce macrophage differentiation of M1 cells. The connection between these factors has been a hotly debated issue. The situation was further complicated by the finding that tumor necrosis factor α (TNFα) is able to stimulate the differentiation of the human myeloblast cell line ML-1 (Takeda, K., Iwamoto, S., Sugimoto, H., Takuma, T. , Kawatani, N., Noda, M., Masaki, A., Morise, H., Arimura, H., and Konno, K. (1986) Nature 323, 338-340).

In an attempt to resolve the discrepancies between the data from these groups, the medium treated with Krebs II ascites tumor cells used in a number of previous studies has now been biochemically fractionated and shown to contain not only authentic G-CSF and GM-CSF, active and normal stem cells as well as WEHI-3B D<+> cells, but also two biochemically different, but functionally similar, factors capable of inducing the differentiation of Ml cells. These latter factors have been named leukemia inhibitory factor (LIF)-A and (LIF)-B because their ability to suppress the proliferation of Ml leukemia cells in vitro and it has been shown that they do not induce the differentiation of WEHI-3B D<+> cells and do not stimulate the proliferation of normal granulocyte/macrophage stem cells.

LIF prepared in accordance with the present invention is defined as a molecule with the following two properties: (1) it has the ability to suppress the proliferation of myeloid leukemia cells such as M1 cells, with associated differentiation of the leukemia cells; and (2) it will compete with a molecule having the herein defined sequence of murine LIF or human LIF for binding to specific cellular receptors on M1 cells or murine or human macrophages.

LIF has potential for use as a therapeutic non-proliferative agent to suppress some forms of myeloid leukemia, as well as use as a response component to modify macrophage function and other responses to infections, and for clinical testing and research related thereto.

The expression "polypeptide with LIF activity" as used herein, denotes a polypeptide or glycopolypeptide with the properties of LIF as stated above, comprising, inter alia, polypeptides having an amino acid sequence that is wholly or partially homologous to the amino acid sequence of either murine or human LIF as set forth herein. This term also includes polypeptides that are wholly or partially homologous to only part of the amino acid sequence of murine or human LIF provided that the polypeptide has the properties of LIF as indicated above. The term further includes polypeptides or glycopolypeptides produced by expression in a host cell, which are inactive when expressed but which are processed by the host cell to yield active molecules, as well as polypeptides or glycopolypeptides which are inactive when expressed but which are selectively cleavable in vitro or in vivo to yield active molecules.

Murine LIF is a molecule with the following biological properties: (a) induction of macrophage differentiation in cells from the murine myeloid leukemia cell line M1 with loss of proliferative capacity and death of the clonogenic leukemia cells, an effect that is enhanced by G-CSF; (b) selective binding to high-affinity receptors on M1 cells and on normal murine monocytes and macrophages from the peritoneal cavity, spleen and bone marrow, the number of receptors increasing with macrophage maturation or functional activation, but not to granulocytic, erythroid or lymphoid

cells from these tissues;

(c) specific binding of 125I-LIF to high-affinity receptors is not subject to competition from G-CSF, GM-CSF, multi-CSF (interleukin-3), M-CSF, interleukins 1,2, 4 or 6, endotoxin or a variety of other growth factors, but is

exposed to competition from unbranded LIF,

(d) increase by bacterial endotoxin in the serum of normal or athymic mice, but not in endotoxin-resistant C3H/HeJ mice; (e) production using various tissues including lung, salivary gland, peritoneal cells and bone marrow; (f) reduction of the survival time in vitro of normal granulocyte-macrophage progenitor cells when grown in the absence of CSF; (g) an inability to suppress proliferation or induce differentiation of WEHI-3B D<+> murine myeloid leukemia cells, murine myeloid cells transformed to leukemiagenicity by infection with retroviruses expressing GM-CSF or multi-CSF, cf. murine leukemia cell lines WEH2 65 and WR19; (h) an inability to stimulate the proliferation of normal progenitor cells of granulocyte, macrophage, eosinophil, megakaryocyte, erythroid and mast cell lineage, and an inability to suppress clonal proliferation or alter the quantitative response to stimulation by CSFs in vitro, of progenitor cells of normal granulocytes, macrophages, megakaryocytes, eosinophils or natural cytotoxic cells, or the proliferation of cells of the continuous cell lines 32CD1.13 and FDCP-1; (i) an inability to compete with iodinated derivatives of granulocyte colony-stimulating factor (G-CSF) for binding to specific cellular receptors despite

the ability of G-CSF to induce differentiation in M1 and WEHI-3B D<+> murine myeloid leukemia cells and to

enhance the effect of LIF on Ml cells;

(j) the action of LIF has an inability to be inhibited by antisera produced specifically

against tumor necrosis factor (TNF);

(k) transcripts for murine LIF are substantially present in the cytoplasm of LB3 and E9.D4 T cells, they are not induced by the lectin concanavalin A in these cells and are present in a variety of other cell types.

Murine LIF is also determined to have the following properties:

(1) a single subunit glycoprotein with a molecular weight of 58,000±5,000 as determined by electrophoresis in 8-25% gradient polyacrylamide gels containing sodium dodecyl sulfate with or without a reducing agent (2-mercaptoethanol or dithiothreitol), and a molecular weight of 23,000±5,000 after treatment with the endoglycosidase , N-glycanase; (m) an isoelectric point between 8.6 and 9.3; (n) primary amino acid sequence as detailed in Figures 5, 6 and 10 herein; (o) is encoded by the nucleotide sequence detailed in Figures 5, 6 and 10 herein;

(p) a specific activity of 1-2 x 10<8> units/mg (where 50

units are defined as the amount of LIF that in 1 ml induces a 50% reduction in clone formation in murine M1 myeloid

leukemia cells);

(q) is encoded by a unique gene on murine chromosome 11 (as determined by analysis of a group of mouse/Chinese hamster ovary somatic cell hybrids) delimited by restriction endonuclease cleavage sites as shown in FIG. 15;

(r) is homologous to a human gene sequence delimited, in chromosomal DNA, by restriction endonuclease cleavage sites as shown in FIG. 23.

Human LIF is a molecule with the following biological properties in the native and/or yeast-derived recombinant product:

(a) induction of macrophage differentiation in cells of the murine myeloid leukemia cell line M1, with loss of proliferative capacity and death of the clonogenic leukemia cells; (b) no ability to stimulate the proliferation of normal human progenitor cells of granulocyte, macrophage, eosinophil and erythroid lineage; (c) an ability, in combination with G-CSF, to partially suppress the proliferation of cells of the human leukemia cell line U937 and, in combination with GM-CSF, the proliferation of the human leukemia cell lines U937 and HL60; (d) binds specifically to murine LIF receptors on M1 cells and murine macrophages and completely competes with murine <125>I-LIF for binding to such cells; (e) binds to specific cellular receptors on the human hepatoma cell line Hep-2G.

Human LIF is also determined to have the following properties:

(f) is identical to murine LIF with approximately 80% of the amino acid residues in the mature protein; (g) has the primary amino acid sequence detailed in Figures 21, 22 and 24; (h) is encoded by a unique gene bound in the chromosomal DNA at restriction endonuclease cleavage sites as shown in fig. 23; (i) has, as part of the sequence of its gene, the nucleotide sequence given in detail in fig. 21.

In a first aspect of the present invention, a method for the production of a LIF (leukemia inhibitory factor) polypeptide has been provided which is characterized in that it comprises: (a) providing host cells containing a recombinant DNA molecule which codes for the LIF polypeptide, in combination with transcriptional and translational regulatory sequences, as well as a selection marker, (b) culturing said cells under conditions leading to expression of the polypeptide, and

(c) recovering the polypeptide;

wherein the polypeptide comprises a mature polypeptide which is at least 78% homologous to an amino acid sequence selected from those shown in Figures 10, 22 and 24.

The invention provides a method for producing a substantially pure LIF. Substantially pure LIF is produced using host cells such as yeast cells, mammalian cells and E.coli, containing recombinant DNA molecules which code for the amino acid sequence of LIF, or some of it.

The term "substantially pure form" denotes a form of polypeptide or glycopolypeptide in which at least 90% of the polypeptide or glycopolypeptide appears as a single band by electrophoresis on a suitable sodium dodecyl sulfate-polyacrylamide gel, said band being consistent with LIF activity.

In another aspect, the present invention relates to a recombinant DNA molecule which codes for the LIF polypeptide and which is characterized by the fact that it comprises a nucleotide sequence as indicated in figures 5, 10 (bases 1-63), 21 or 24 (bases 1-543) , said polypeptide competing with a polypeptide with an amino acid sequence as indicated in Figures 10, 22 or 24 for binding to specific cellular receptors on M1 cells or murine or human macrophages. Said recombinant DNA molecule comprising a nucleotide sequence which, upon expression in a suitable host cell, codes for the LIF polypeptide is also characterized in that said nucleotide sequence hybridizes under conditions of 6 x SSC, 65°C or more stringent conditions, to a sequence of a LIF -coding insert selected from the LIF-coding inserts of clones pLIF7.2b, pLIFNKl, pLIFNK3 and pHGLIFBaml as defined in Figures 5, 6, 10 or 21. The invention further relates to a host cell which is characterized in that it has been transformed with a recombinant DNA molecule and has the ability to express the LIF polypeptide.

In connection with investigations concerning LIF, particularly with a view to providing alternative sources of this glycoprotein of both murine and human origin to enable use both clinically and otherwise, murine LIF has also been purified from Krebs II ascites tumor cells using of an efficient purification procedure (as described in more detail in NO patent application no. 885339) and has been subjected to partial amino acid sequence analysis as a first step towards chemical or biosynthetic production of this factor.

LIF that has been radioactively derivatized with I<125> has been used in a receptor competition assay to identify cells and tissues of murine and human origin that synthesize and produce LIF. As a result of this investigation, a recombinant DNA library of DNA copies of mRNA from one of these sources has been screened and murine LIF-encoding clones identified by hybridization with radiolabeled oligonucleotides encoding portions of the previously determined murine LIF amino acid sequence, and the above-mentioned murine LIF cDNA clones have been subjected to nucleotide sequence analysis. Moreover, the cloned murine LIF coding sequence has been used as a hybridization probe to identify a human gene encoding a homologue of murine LIF and hybridization conditions have been established under which said hybridization between species can be carried out in an efficient manner. As a result, recombinant DNA clones containing DNA sequences encoding the human homologue of murine LIF have been identified.

As a result of this construction of LIF-encoding sequences, recombinant DNA molecules have been constructed that use cloned murine or human LIF-encoding DNA sequences to direct the synthesis of clonally pure LIF, either in whole or in part, in e.g. cultured mammalian cells, in yeast or in bacteria.

In a particular embodiment, the invention relates to expression of cloned human and murine LIF genes in yeast cells, in fibroblasts and in E.coli, and modification of the cloned gene so that it can be expressed as such, as well as establishment of the biological and biochemical properties of recombinant human and murine LIF.

In this embodiment, the invention also relates, as mentioned, to recombinant DNA molecules containing nucleotide sequences that code for human or murine LIF, either completely or partially, in a form in which said nucleotide sequences are capable of controlling the synthesis and production of human or murine LIF in yeast cells, fibroblasts or E.coli, either in whole or in part. Cloning vectors (such as plasmids) and, as mentioned, host cells with such recombinant DNA molecules inserted into them are also provided. In one embodiment of the invention, the human and murine LIF gene sequences are modified and installed in a yeast expression vector, YEpsecl. Yeast cells have been transformed with the obtained recombinants and the medium conditioned with the thus transformed yeast cells has been shown to contain a factor with biological properties analogous to those of natural human and murine LIF.

Viewed against the background of the potential for the use of LIF in the treatment of patients with some forms of myeloid leukemia and patients with certain infections, pharmaceutical preparations have been obtained comprising LIF, in particular human LIF, either completely or partially and which are prepared by e.g. use of cloned LIF-coding DNA sequences or by chemical synthesis. The pharmaceutical preparations may also contain at least one other biological regulator of blood cells, such as G-CSF or GM-CSF.

Subexample 1

The following sub-example shows the steps used to obtain the N-terminal amino acid sequence and the amino acid sequences of various peptides formed by proteolytic cleavage with trypsin or Staphylococcus V8 protease.

LIF purified using the procedure as described in ex. 1 in NO patent application no. 885339 (15 Jig) was applied to a microtube

(2 mm) column packed with Brownlee RP 300 beads (C8) and eluted on an HPLC system with a linear gradient from 0-60% acetonitrile in 0.1% trifluoroacetic acid. The protein eluted as a single peak in a volume of 100 µl. This was applied to a fiberglass dish and dried, then placed in a sequencing chamber of an Applied Biosystem (Model 47OA) gas phase protein sequencer. The protein was sequenced by Edman chemistry and the phenylthiohydantoin derivatives from the individual amino acids were identified by HPLC on an Applied Biosystems 12OA PTH analyzer.

A separate sample of LIF (20 Hg) also purified using the above procedure was diluted to 2 ml in 6 m guanidine hydrochloride, 0.2 M tris-HCl buffer pH 8.5, 2 mM dithiothreitol and incubated at 37 °C for 2 hours. Then a 30-fold molar excess of iodoacetic acid was added and the solution was incubated at 37°C for a further 15 min. The solution was applied to the same microtube HPLC column and eluted as described above. The reduced and carboxymethylated derivative of LIF (RCM-LIF) eluted 3 min. later from the column than untreated LIF and RCM-LIF (200 µl) was diluted to 1 L with 0.1 M tris-HCL buffer pH 8.0 containing 0.02% Tween 20, 1 mM CaCl2 and 2 µg TPCK- trypsin treated. This incubation was allowed to continue for 18 hours at 37°C and the mixture was again applied to the microtube column and eluted in the same manner as described above. Individual peptides appeared as UV-absorbing peaks, and were collected individually. Some of these peptides were sequenced directly as described above while others were repurified on the same column but using a 0-60% acetonitrile gradient in 0.9% NaCl pH 6.0 prior to sequencing.

A second sample of RCM-LIF (200 µl = 10 µg) was diluted to

1 ml of 0.05 M sodium phosphate buffer pH 7.8 containing 0.002 M EDTA and 0.02% Tween 20, and was digested with 2 µg of Staphylococcus aureus V8 protease for 18 hours at 37°C. The reaction mixture was applied to the same microtubule column and eluted as described above.

The 2 6 amino acids defining LIF from the amino terminus were, in order, found to be: N-Pro-Leu-Pro-Ile-Thr-Pro-Val-X-Ala-Thr-X-Ala-Ile-Arg-His-Pro- Cys-His-Gly-Asn-Leu-Met-Asn-Gln-Ile-Lys

The amino acid sequences of several tryptic peptides were found to be:

1. His-Pro-Cys-His-Gly-Asn-Leu-Met-Asn-Gln-Ile-Lys

2. Met-Val-Ala-Tyr...

3. Gly-Leu-Leu-Ser-Asn-Val-Leu-Cys..

4. Leu-Gly-Cys-Gln-Leu-Leu-Gly-Thr-Tyr-Lys

5. Val-Leu-Asn-Pro-Thr-Ala-Val-Ser-Leu-Gln-Val-Lys

6. Asn-Gln-Leu-Ala-Gln-Leu-X-Gly-Ser-Ala-Asn-Ala-Leu-Phe-Leu-Val-Glu-Leu-Tyr ....

7. Leu-Val-Ile-Ser-Tyr...

8. Val-Gly-His-Val-Asp-Val-Pro-Val-Pro-Asp-His-Ser-Asp-Lys-Glu-Ala-Phe-Gln.

9. Leu-X-Ala-Thr-Ile-Asp-Va1-Met-Arg.

10. Gln-Val-Ile-Ser-Val-Val-X-Gln-Ala-Phe.

Amino acid sequence one for a peptide formed by cleaving LIF with V8 protease was: Ala-Phe-Gln-Arg-Lys-Lys-Leu-Gly-Cys-Gln-Leu-Leu^Gly-Thr-Tyr-Lys-Gln- Val-Ile-Ser-Val-Val-Val-Gln-Ala-Phe.

Subexample 2

The following sub-example shows the steps used to obtain radioactively derivatized LIF and to identify LIF sources using receptor competition assays.

Fig. 1 relates to subexample 2: Binding of radioiodinated pure LIF (approximately 2 x 10<5> cpm/ng) to M1 myeloid leukemia cells at 0°C. Ml cells (5 x 10<6>) were incubated with <125>I-LIF (4 x 105 cpm) with or without competing factors for 2 hours after which bound radioactivity was separated from unbound radioactivity by centrifugation through fetal calf serum.

A. Titration of unlabeled pure LIF and pure multi-CSF, GM-CSF, G-CSF or M-CSF as competing factors in indicated dilutions (10 µl of each dilution added to a final volume of 80 µl).

The highest concentrations of each were 10<4 >units/ml, respectively; 3.5 x 10" units/ml; 3 x IO<4> units/ml; 5 x 10<4 >units/ml; 4 x 10<4> units/ml.

B. The ability of pure LIF or lipopolysaccharide (LPS) or concentrated (4-8 times) medium treated with various cells or serum from endotoxin-injected mice to compete with <125>I-LIF for binding to M1 cells as above.

LIF purified as described in ex. 1 in NO patent application no. 8853 3 9 was radioactively labeled with 1<25>I on tyrosine residues as described previously (Nicola, N.A. and Metcalf, D., J. Cell. Physiol. 128:180-188, 1986) and gave <125>I-LIF with a specific radioactivity of about 2 x IO<5> cpm/ng. <125>I-LIF bound specifically to M1 myeloid leukemia cells as well as to murine "adult" bone marrow cells, spleen cells, thymus cells, and peritoneal exudate cells. Autoradiographic analysis of the cells indicated that <125>I-LIF only bound specifically to macrophages and their progenitor cells, with receptor numbers increasing with cell maturation and not detectable for any other types of hemopoietic cells. This suggests that LIF may be clinically useful for application in modulating macrophage function in a number of disease states. The specific binding of <125>I-LIF to Ml myeloid leukemia cells or to peritoneal macrophage populations was inhibited with unlabeled LIF, but not with a variety of other hematopoietic growth factors, including G-CSF, GM-CSF, M-CSF (CSF-1) , multi-CSF (interleukin 3), interleukins 1,2,4,6, endotoxin or other growth factors (see e.g. Fig. 6). The ability of cell-treated medium or cell extracts to inhibit the binding of <1>25I-LIF to such cell populations (radioactive receptor assay for LIF) (Fig. 6B) thus provided a specific measurement for the presence of LIF production or storage by various cells and tissues . Medium conditioned with lectin-activated LB3 T cells will e.g. strongly inhibit specific binding of <125>I-LIF to M1 cells and thus indicate that LB3 cells produce LIF.

EXAMPLE 1

The following example shows the steps used to obtain a cloned complementary DNA copy of the mRNA encoding murine LIF. Figures 2 to 5 deal with different steps in the method which is described in the following. Figure 2 deals with step 1 of the cloning of LIF cDNA clones: accumulation of mRNAs for hemopoiesis growth regulators in the cytoplasm of LB3 cells after stimulation with the lectin concanavalin A. Cytoplasmic polyadenylated RNA (5 µg) from WEHI-3B D" cells (lane 1), T cell clone LB3 cells cultured at 10<6 >cells/ml with 5 µg/ml concanavalin A for 5 hours (lane 2), and unstimulated LB3 cells (lane 3) were electrophoresed on formaldehyde-agarose gels, transferred to nitrocellulose and hybridized with a <32>P-GM-CSF probe (lane a) or a <32>P-multi-CSF probe (lane b) (for details see Kelso, A., Metcalf, D. and Gough, N.M., J. Immunol. 136:1718-1725, 1986). Figure 3 relates to step 3 of the cloning of LIF cDNA clones: synthetic oligonucleotide used for identification of LIF clones. A portion of the amino acid sequence of murine LIF near the N-terminal end (residues 15-26, see subex. 1) is shown in the top line, the possible combinations of mRNA sequences that could code for this peptide are shown in the middle and etc. the igonucleotide probe complementary to this mRNA sequence is shown at the bottom. Note that for all amino acid residues except Cys 17 and His 18, the oligonucleotide is only complementary to the codons most commonly used in mammalian coding regions (Grantham et.al.,

Nucleic Acids Res. 9:43-77, 1981). For Cys 17, the sequence is complementary to both codons. For His 18, the oligonucleotide is complementary to the less frequently used CAU codon. It was unlikely that the CAC codon would be used as it would form, in conjunction with the neighboring Gly codon, the unusual CpG dinucleotide (Swartz et.al, J. Biol. Chem. 238:1961-1967, 1962). Figure 4 relates to step 4 in the cloning of LIF cDNA clones: The identification of LIF cDNA clones by hybridization with the oligonucleotide as illustrated in fig. 3. Figure 5 relates to step 5 of the cloning of LIF cDNA clones: nucleotide sequence cDNA clone pLIF7.2b and LIF amino acid sequence. The sequence of the mRNA synonymous strand of the cDNA is listed from 5' to 3' with the predicted amino acid sequence of LIF given above. The numbers at the end of lines indicate the position of the final residue (amino acid or nucleotide) on that line. The amino acid sequence is numbered consecutively from the first residue encoded in this clone. Amino acid sequence regions determined by analysis of the protein are underlined. There are no discrepancies between the two. Pro 9 was the N-terminal residue determined by direct amino acid sequence analysis (subex. 1). Seven potential N-linked glycosylation sites (Neuberger, A. et.al. in Glycoproteins, Gottschalk, A., ed., Elsevier, Amsterdam, pages 450-490, 1972) are indicated by asterisks and four potential O-linked glycosylation sites (Takahashi , N. et.al., Proe. Nat. Acad. Sei. USA, 81: 2021-2025, 1984) with shading.

Step 1: Preparation of RNA containing LIF mRNA. Cells of the cloned murine T cell line LB3 (Kelso, A. and Metcalf, D. Exptl. Hematol. 13:7-15, 1985) were stimulated with leetin concanavalin A for 6 hours to increase the accumulation in the cytoplasm of mRNAs encoding various factors that regulate the growth and differentiation of hemopoiesis cells.

(Kelso, A., Metcalf, D. and Gough, N.M., J. Immunol. 136:1718-1725, 1986). (Fig. 2 shows, for example, the increased production of mRNA

which encodes the hemopoietic growth factors GM-CSF and multi-CSF in cells stimulated in this way.) Cytoplasmic RNA was prepared from 5 x IO<8> concanavalin A-stimulated LB3 cells by using

by a technique described previously (Gough, N.M., J. Mol. Biol. 165:683-699, 1983). Polyadenylated mRNA molecules

was separated from ribosomal RNA by two rounds of chromatography on oligo-dT cellulose using standard procedures (Maniatis, T. et.al., Molecular Cloning, Cold Spring Harbor,

New York, 1982).

Later, it was found by Northern blotting that unlike mRNAs for GM-CSF and multi-CSF, LIF mRNA is present substantially in LB3 cells and its abundance is not increased by concanavalin A stimulation (Gearing, D.P. et.al.,

EMBO J. 6: 3995-4002, 1987).

Step 2: Synthesis and cloning of a library of LB3 cDNA molecules.

Double-stranded DNA copies of LB3 mRNA prepared as described

above was synthesized by standard procedure (Maniatis, T.

et.al., Molecular Cloning, Cold Spring Harbor, New York, 1982

and Gough, N.M. et al., Nature 309:763-767 1984). Briefly, 10 µg of cytoplasmic polyadenylated LB3 mRNA was used as a template for the synthesis of single-stranded complementary DNA (cDNA) in a reaction catalyzed by avian myeloblastosis virus reverse transcriptase and initiated with oligo-dT. After completion of the reaction, the mRNA was degraded by incubation at 65 °C for 1 h in 0.3 M NaOH, 1 mM EDTA After neutralization of the base and extraction of cDNA by ethanol precipitation, the single-stranded cDNA was converted to double-stranded in a reaction catalyzed by the Klenow fragment of E.coli DNA polymerase I. The "hairpin loop" structure of the cDNA was then cut by treatment with single-stranded specific nuclease S1 and then extensions of deoxycytidine residues (approximately 20-30 residues in length) were attached to each end of the double-stranded cDNA using the enzyme terminal deoxynucleotidyl transferase as described (Michelson, A.M. and

Orkin, S. J. Biol. Chem. 257:14773-14782, 1982). cDNA extended with dC was fractionated by electrophoresis on a 1.5% agarose gel and molecules with a length greater than 500 bp were recovered and annealed to a plasmid DNA molecule (pJL3, Gough, N.M. et.al., EMBO J. 4 :645:653, 1984) which was cleaved with the restriction endonuclease Sac 1 and to which extensions of deoxyguanosine residues were attached (Michelson, A.M. and Orkin, J., J. Biol. Chem. 257:14773-14782, 1982). The extended cDNA and plasmid molecules were annealed as described (Gough, N.M. et.al., Biochemistry 19:2702-2710, 1980) and E.coli MC1061

(Casadaban, M. and Cohen, S., J. Mol. Biol. 138:179-207, 1980)

transformed with the base-paired cDNA/plasmid mixture and approximately 50,000 independently transformed bacterial colonies were selected by growth on agar plates containing 10 µg/ml ampicillin. The transformed bacterial colonies were removed from the agar plates by washing with liquid growth medium containing 50 µg/ml ampicillin and stored as 10 independent pools in 10% glycerol at -70°C.

Step 3: Design of Oligonucleotide Probes.

A sequence of 26 amino acids at the amino-terminal end of LIF and residues from 11 different peptides derived by cleavage of LIF with trypsin and V8 protease has been determined (see subex. 1). These amino acid sequences provided the basis for the design of oligonucleotides complementary to certain defined regions of mRNA encoding LIF for use as hybridization probes to identify cDNA clones corresponding to LIF mRNA.

Each naturally occurring amino acid is coded for in its corresponding mRNA by a specific combination of 3 ribonucleoside triphosphates (a codon) (eg, Watson, J. Molecular Biology of the Gene, 3rd ed., Benjamin/Cummings, Menlo Park, Calif., 1976). Certain amino acids are specified by only one codon, while others are specified by as many as 6 different codons (eg, Watson, J. Molecular biology of the Gene, supra). Therefore, since a number of different combinations of nucleotide sequences will actually be able to code for any particular amino acid sequence, a number of different degenerate oligonucleotides will therefore have to be constructed to accommodate every possible sequence that potentially codes for that peptide. However, there are certain technical difficulties when using highly degenerate oligonucleotides as hybridization probes. Since, however, for a given amino acid, not all codons are used with equivalent frequency (Grantham, R. et.al., Nucleic Acids Res. 9:43-73, 1981) and then, in the mammalian genome, the CpG dinucleotide is underrepresented, as it occur at only 20 to 25% of the frequency expected from the base composition (Swartz, M.N. et.al., J. Biol. Chem. 238:1961-1967, 1962), it is often possible to predict the likely nucleotide sequence that encodes for a given peptide and thus reduce the complexity of a given oligonucleotide probe. Based on the foregoing, a number of oligonucleotides corresponding to different LIF peptides were designed and synthesized using standard procedures.

Step 4: Screening of an LB3 cDNA library for LIF-encoding clones.

For screening colonies of bacteria by hybridization with oligonucleotide probes, 10,000-15,000 bacterial colonies from each pool of the above LB3 cDNA library were grown on agar plates (containing 50 µg/ml ampicillin), transferred to nitrocellulose filter dishes and plasmid DNA was amplified by incubation of the filters on agar plates containing 200 µg/ml chloramphenicol (Hanahan, D. and Meselson, M. Gene 10:63-67, 1980). After regrowth of colonies on the original plate, another nitrocellulose filter was prepared as above. The master plate was cultured again and then stored at 4°C. Plasmid DNA was released from bacterial colonies and fixed to the nitrocellulose filters as described previously (Maniatis, T. et.al., Molecular Cloning, Cold Spring Harbor, New York, 1982). Before hybridization, the filters were incubated for several hours at 37°C in 0.9 M NaCl, 0.09 M sodium citrate, 0.2% Ficoll., 0.2% polyvinylpyrrolidone, 0.2% bovine serum albumin, 50 µg/ml heat denatured salmon sperm DNA, 50 ug/ml E.coli tRNA, 0.1 M ATP and 2 mM sodium pyrophosphate. Hybridization was carried out in the same solution containing an additional 0.1% NP40, at 37°C for 18 hours. 500 ng of the synthetic oligonucleotide as shown in fig. 8, was radioactively labeled in a reaction catalyzed by polynucleotide kinase and containing 500 jiCi [γ-<32>P]ATP (specific activity 2000-3000 Ci/mmol). Unincorporated [γ-<32>P]ATP was then separated from the radiolabeled oligonucleotide by ion exchange chromatography on a NACS-PREPAC column (Bethesda Research Laboratories) according to the manufacturer's instructions. Radiolabeled oligonucleotide was included in the hybridization reaction at a concentration of approximately 20 ng/ml. After hybridization, the filters were extensively washed in 0.9 M NaCl, 0.09 M sodium citrate, 0.1% sodium dodecyl sulfate at various temperatures (as described below) and after each wash, the filters were autoradiographed.

The rationale for performing multiple successive washes was that at lower temperatures, all clones with even a small degree of homology to the oligonucleotide (as little as about 15 nucleotides) would show up as autoradiographic spots appearing on duplicate filters. When the temperature is increased, the oligonucleotide probe will disappear from clones with the lowest level of homology and the corresponding autoradiographic signals will thus be lost. Clones with the highest degree of homology (and thus clones that represent the best candidates for containing LIF cDNA sequences) will retain hybridization at the highest temperature. With this strategy, it is thus possible to focus directly on the strongest candidates. Fig. 4 shows the performance of a set of clones on a double filter pair containing 15,000 LB3 cDNA clones when the wash temperature was increased from 46°C to 66°C. Several clones only retained hybridization to the oligonucleotide at lower temperatures, while several retained hybridization even at 66°C (eg clones 1 and 2). The latter clones were thus selected for further analysis and the corresponding bacterial colonies were removed from the master plates, purified and plasmid DNA was prepared from each bacterial clone using standard procedures (Maniatis, T. et.al., Molecular Cloning, Cold Spring Harbor, 1982 ). Initial structural analyzes of these clones (by determining the size of the inserted cDNA, by mapping the location of cleavage sites for different restriction endonucleases and by testing for hybridization with different oligonucleotides corresponding to different LIF peptides) indicated that each of these clones were indeed identical; i.e., they represented different isolates of the same original clone. Thus, further detailed analyzes were carried out only on one clone, pLIF7.2b.

Step 5: Determination of the nucleotide sequence of pLIF7. 2b.

Nucleotide sequence analysis of the cDNA portion of clone pLIF7.2b was carried out by the dideoxy chain termination method (Sanger, F. et.al., Proe. Nati. Acad. Sei., USA 74:5463-5467, 1977), using alkaline or heat-denatured double-stranded plasmid DNA as template, with a series of oligonucleotides complementary both to the regions of the vector (pJL3) flanking the cDNA and to sequences in the cDNA insert as primers, and using both the Klenow fragment of E.coli DNA polymerase I and avian myeloblastosis virus reverse transcriptase as polymerases in the sequencing reactions. The sequence thus determined for the cDNA part of clone pLIF7.2b is shown in fig. 5. Such analyzes confirm that this clone indeed contains a DNA copy of an mRNA capable of encoding the LIF molecule, as all the amino acid sequences previously determined for different peptides of the LIF molecule can be found within the single translational reading frame spanning over the entire cDNA and is not interrupted by stop codons. However, this clone does not contain a complete copy of the LIF coding region as (a): it does not, at the 5' end, extend past a region encoding a putative hydrophobic leader sequence initiated at a methionine codon and (b): it includes not, at its 3' end, an "in-frame" translational stop codon. At the 5' end, however, it extends past the beginning of the region that codes for the mature protein as determined by comparison with the previously determined amino-terminal amino acid sequence (residue Pro 9 in Fig. 5).

EXAMPLE 2

The following example describes the steps used to construct a full-length copy of the murine LIF coding region, install this coding region into a yeast expression vector, and prepare murine LIF. Figures 6-9 relate to different steps in the method described below. Figure 6: concerns step 1 of the expression of murine LIF in yeast cells: the C-terminal amino acid sequence of LIF. The amino acid sequence at the C-terminal end of the pLIF7.2b reading frame is shown above the sequence of a Staphylococcal V8 protease and a tryptic peptide derived from LIF purified from Krebs ascites tumor cells. Note that these latter peptides extend past the pLIF7.2b LIF sequence by 9 amino acids and terminate at the same residue. The nucleotide and amino acid sequence in the corresponding region of clone pLIFNK1 is shown in the bottom line, which confirms the C terminus detected by direct amino acid sequencing. The numbering system is the same as in fig. 5. Figure 7: Regarding step 2 of the expression of murine LIF in yeast cells: reconstruction of pLIF7.2b cDNA for introduction into the yeast expression vector, YEpsecl. Partial nucleotide sequence of pLIF7.2b cDNA and its encoded amino acid sequence are shown in the top row. The numbering is in accordance with Figures 5 and 6. The second and third lines show the sequence of pLIFmut1 and pLIFmut2 respectively. Asterisks indicate stop codons. Restriction sites (Barn HI, Hind III and Eco RI) used for cloning in YEpsecl (Baldari, C. et.al., EMBO J. 6: 229-234, 1987) and pGEX-2T (Smith, D.B. and Johnson, K.G., Gene, 1988) is indicated. The bottom line shows the partial sequence of the K.lactis killer toxin signal sequence in YEpsecl. The sequence Gly-

Ser encoded by the Barn HI restriction site is efficiently recognized by signal peptidase (Baldari, C. et.al, EMBO J. 6: 229-234, 1987). Figure 8: concerns step 4 in the expression of murine LIF in yeast cells: the biological activity of yeast-derived recombinant LIF in cultures of Ml leukemia cells. Titration in M1 cultures of yeast-derived recombinant LIF (-•-) against purified native LIF-A (-o-j) shows equal concentration-dependent induction of differentiation in M1 colonies (panel A) and suppression of colony formation (panel B). Each point represents the mean value from duplicate cultures Figure 9: concerns step 5 of the expression of murine LIF in yeast cells: the ability as different dilutions of authentic native murine LIF-A (-o-), conditioned medium from yeast cells containing LIFmut1/YEpsec1 recombinant induced with galactose (-•- ), and conditioned medium from the same yeast cells, but not induced with galactose (-+-), tend to compete with native <125>I-LIF-A for binding to cellular receptors on murine peritoneal cells.

Step 1: determination of the amino acid sequence of the C-terminus of murine LIF.

cDNA clone pLIF7.2b in ex. 1 contains an incomplete copy of murine LIF mRNA. At the 5' end, the cDNA encodes 8 residues N-terminal to the first residue determined by N-terminal amino acid sequence analysis (Pro 9 in Fig. 5). At the 3' end, however, pLIF7.2b is incomplete as it does not contain an "in frame" translational stop codon. Inspection of the amino acid sequence of two LIF peptides determined by direct amino acid sequencing (subex. 1) suggests that pLIF7.2b was missing only 27 nucleotides of the coding region at the 3' end. As shown in fig. 6, a V8-derived peptide started at residue Ala 162 in the cDNA sequence and extended past the amino acid sequence deduced from the cDNA clone by 9 residues. A tryptic peptide contained in the V8 peptide terminated at the same residue, suggesting that Phe 187

is the C-terminal residue of the protein. To confirm this conclusion, a LIF cDNA clone overlapping with pLIF7.2b and extending towards 3' was isolated and subjected to nucleotide sequence analysis.

In order to construct and map a suitable cDNA library from which to isolate such a clone, a number of different mRNA samples were screened using Northern blot hybridization to identify the RNA samples with the highest concentration of LIF mRNA molecules. Cytoplasmic polyadenylated RNA, prepared essentially as described previously (Gough, N.M., J. Mol. Biol. 165: 683-699, 1983), was fractionated on 1% agarose gels containing 20 mM morpholinopropane sulfonic acid (MOPS), 5 mM sodium acetate , 1 mM EDTA (pH 7.0) plus 6% (w/w) formaldehyde, filters containing RNA were immersed in 2 x SSC, containing 0.2% Ficoll, 0.2% polyvinylpyrrolidone, 0.2% bovine serum albumin, 2 mM sodium pyrophosphate, 1 mM ATP, 50 µg/ml denatured salmon sperm DNA and 50 µg/ml E.coli tRNA, at 67°C for several hours. Hybridization was in the same buffer plus 0.1% SDS at 67°C. The probe used to detect LIF transcripts consisted of an approximately 750 bp Eco RI - Hind III fragment containing the cDNA insert of pLIF7.2b subcloned into pSP65. This fragment contains not only the cDNA sequence, but also G-C extensions and approximately 150 bp of the pJL3 vector sequence. Riboprobes of approximately 2 x 10<9> cpm/µg were derived by transcription of this SP6 sub-clone using reagents from BRESA (Adelaide). The probe was included in hybridization at approximately 2 x 10<7 >cpm/ml. The filters were washed extensively in 2 x SSC, 2 mM EDTA, 0.1% SDS at 67°C and finally in 0.2 x SSC at 67°C before autoradiography.

Such analyzes showed that LIF transcripts were present in small amounts in a number of haemopoiesis cell lines and that there was considerable variation in the amount of LIF mRNA in different batches of Krebs RNA. Two batches of Krebs ascites tumor cell RNA were selected for synthesis of two cDNA libraries. cDNA library was constructed using reagents from Amersham

(Product no. RPN.1256 and RPN.1257) and using the supplier's instructions; the cloning vector being XGT10. Approximately 4 x 10<5> recombinant clones were obtained and screened by hybridization with an oligonucleotide corresponding to a sequence of 36 residues at the 3'-end of the pLIF7.2b sequence: nucleotides 500-535 (inclusive) in fig. 5.

Phage plaques representing the Krebs cDNA library were grown at a density of approximately 50,000 plaques per 10 cm petri dish, transferred in duplicate to nitrocellulose and processed using standard techniques (Maniatis, T. et.al., Molecular Cloning, Cold Spring Harbor, 1982). Before hybridization, the filters were incubated for several hours at 37°C in 6 x SSC (SSC = 0.15 M NaCl, 0.015 M sodium citrate), 0.2% Ficoll,

0.2% polyvinylpyrrolidone, 0.2% bovine serum albumin, 2 mM sodium pyrophosphate, 1 mM ATP, 50 mg/ml denatured salmon sperm DNA and 50 u/ml E.coli tRNA. Hybridization is carried out in the same solution containing 0.1% NP40, at 37°C for 16-18 hours. The above oligonucleotide probe, which was radiolabeled using [γ-<32>P] ATP and polynucleotide kinase to a specific activity of about 10<9> cpm/ug and separated from unincorporated label by ion exchange chromatography on a NACS column (Bethesda Research Laboratories ), was included in the hybridization at a concentration of approximately 20 ng/ml. After hybridization, the filters were washed extensively in 6 x SSC, 0.1% sodium dodecyl sulfate at 60°C. A plaque representing clone A.LIFNK1, positive on duplicate filters, was picked and screened again at a lower density, as before.

The cDNA insert of approximately 950 bp in X.LIFNK1, which hybridized with the above oligonucleotide, was subcloned into a plasmid vector (pEMBL8+, Dente, L. et.al, Nucl. Acids Res. 11: 1645-1655, 1983) to to form clone pLIFNK1. Nucleotide sequence analysis of the cDNA inserted into pLIFNK1 was performed by the dideoxy chain termination method (Sanger, F. et.al., Proe. Nat. Acad. Sei., USA 74: 5463-5467, 1977) using alkali-denatured double-stranded plasmid DNA as template (Chen, E.Y. and Seeburg, P.H., DNA 4: 165-170, 1985), a series of oligonucleo-

times that were complementary both to regions of the vector surrounding the cDNA and to the sequences inside the cDNA insert as primers, and using both the Klenow fragment of E.coli DNA polymerase and AMV reverse transcriptase as polymerase in the sequencing reactions. The nucleotide sequence of part of the cDNA inserted into pLIFNKl is shown in Figure 6. The amino acid sequence specified by pLIFNKl is identical at the C-terminal end to the 9 amino acids that were predicted by direct amino acid sequencing to constitute the C-terminal end of LIF, and confirms that Phe 187 is the C-terminal residue of LIF, then, in the pLIFNK1 cDNA sequence, the codon for this residue is immediately followed by an "in frame" translational stop codon.

Step 2: Construction of a LIF codon region in a yeast expression vector.

Initial production of the protein encoded by the LIF cDNA clones was achieved in a eukaryotic system (yeast), so that the expressed product will be glycosylated, secreted and correctly folded. The expression vector used, YEpsecl (Baldari, C. et.al., EMBI J. 6: 229-234, 1987), provides an N-terminal leader sequence derived from the killer toxin gene of Kluyveromyces lactis, which has previously been shown to direct the efficient secretion of interleukin 1 (Baldari, C. et.al., EMBO J. 6: 229-234, 1987), transcribed from a galactose-inducible hybrid GAL-CYC promoter.

To express the protein encoded by pLIF7.2b in this vector, it was necessary to modify the cDNA in several ways (see Fig. 7). At the 5' end it was necessary not only to remove the few nucleotides specifying the partial mammalian leader sequence, but also to include an appropriate restriction endonuclease cleavage site (Barn HI) to allow insertion "in-frame" with the K.lactis leader and retaining an appropriate signal peptidase cleavage site (Gly-Ser). At the 3' end, two versions of the coding region for LIF were constructed. A version (LIFmut1) was constructed such that a stop codon followed immediately after the last codon of pLIF7.2b (Gin 178). The second version (LIFmut2) was constructed by adding a sequence encoding the 9 amino acid residues known to be absent from pLIF7.2b (see above), followed by an "in-frame" translational stop codon. An appropriate restriction endonuclease cleavage site (Hind III) complemented both constructed versions. All these modifications were achieved by means of oligonucleotide-mediated mutagenesis: the approximately 750 bp Eco RI - Hind III fragment carrying the 535 bp cDNA insert of pLIF7.2b, ligated by means of G-C extensions and part of the pJL3 vector was subcloned into plasmid pEMBL8+ (Dente, L. et.al., Nucl. Acids Res. 11: 1645-1655, 1983) and single-stranded DNA was prepared as described (Cesarini, G. and Murray, J.A.H. in Setlow, J.K. and Hollaender, A. ( eds), Genetic Engineering: Principles and Methods, Plenum Press, New York, vol. 8, 1987). In vitro mutagenesis was carried out as described (Nisbet, I.T. and Beilharz, M.W., Gene Anal. Techn. 2: 23-29, 1985), using oligonucleotides of 35, 51 and 61 bases, respectively, to modify the 5' end and the 3' end of the cDNA as indicated above. Barn HI - Hind III fragments containing the modified LIF cDNA sequences were ligated into plasmid YEpsecl and the nucleotide sequence of the inserts in the obtained recombinants was determined.

Step 3: Introduction of YEpsecl/LIF recombinants into yeast cells.

S.cerevisiae strain GY41 (leu2 ura3 adel his4 met2 trp5 gall

cir+; x 4003-5b from the Yeast Genetic Stock Centre, Berkeley) was transformed by the polyethylene glycol method (Klebe, R.J. et.al., Gene 25: 333-341, 1983). Transformants were selected and maintained on a synthetic minimal medium (2% carbon source, 0.67% yeast nitrogen base (Difco) supplemented with 50 µg/ml of the required amino acids) under uracil deprivation. Expression of insert sequences in plasmid YEpsecl was achieved by growing transformants in either a non-selective complete medium (1% yeast extract, 2% peptone) or in a synthetic minimal medium, each containing 2% galactose.

Step 4: Determination of the biological properties of the yeast-derived LIF.

Experiments concerning differentiation-inducing activity and leukemia-suppressive activity of yeast-conditioned medium were carried out in 1 ml cultures containing 300 ml cells (from Dr. M. Hozumi, Saitama Cancer Research Centre, Japan) in Dulbecco's modified Eagle's medium with a final concentration of 20% fetal calf serum and 0.3% agar. Material to be analyzed was added in serially diluted 0.1 ml volumes to the culture dish before adding the cell suspension in agar medium. The cultures were incubated for 7 days in a fully humidified atmosphere at 10% CO 2 . The cultures were scored using a dissecting microscope at 35x magnification, a score that differentiated all colonies with a rosette of dispersed cells or that consisted entirely of dispersed cells. Morphological examination of the colonies was carried out by fixing the whole culture with 1 ml of 2.5% glutaraldehyde, after which dry cultures were stained on microscope slides using acetylcholinesterase/Luxol Fast Blue/Haematoxylin.

Measurements for colony stimulating activity were carried out using 75,000 C57BL bone marrow cells as described previously (Metcalf, D., The Hemopoietic Colony Stimulating Factors Elsevier, Amsterdam, 1984). Assays regarding differentiation-inducing activity on WEHI-3B D<+> cells were carried out as described previously (Nicola, N. et.al., J. Biol. Chem. 258: 9017-9023, 1983).

Medium from yeast cultures containing the full-length coding region (LIFmut2), but not from cultures of non-transformed yeast, yeast containing the vector YEpsecl alone, or yeast containing the incomplete coding region (LIFmut1), is able to induce typical macrophage differentiation in cultures of M1 colonies (Fig. 8A). As is the case with purified native LIF, the yeast-derived material, with increasing concentrations, will also progressively reduce the number and size of M1 colonies that develop (Fig. 8B). Comparison with purified native LIF indicates that the yeast-conditioned medium contained up to 16,000 units/ml (approximately 130 ng/ml) of LIF. Yeast-derived LIF, which purified native LIF-A, fails to stimulate colony formation with normal granulocyte-macrophage progenitor cells or to induce differentiation in, or suppress proliferation of, WEHI-3B D<+> leukemia colonies.

Size fractionation on a Sephacryl S-200 column indicated that yeast-derived LIF co-eluted with proteins with apparent molecular weights from 67,000 to 150,000 Daltons, compared to 58,000 Daltons for LIF derived from Krebs cells.

Absence of LIF activity in medium conditioned with yeast containing an incomplete coding region (LIFmut1) suggests that the nine hydrophobic C-terminal residues missing in this engineered unit may be required for LIF function. While these residues may interact with the receptor, they may also form part of a hydrophobic core in the protein, and thus their absence may prevent correct folding. Alternatively, they may in some respects be required for efficient secretion of LIF.

Step 5: Receptor binding specificity of yeast-derived murine LIF and purified native LIF-A from Krebs II ascites cell treated medium.

Purified murine LIF-A (see ex. 1 in NO patent application no. 885339) was iodinated as indicated previously (subex. 2). Peritoneal cells were harvested by lavage from mice in which a large amount of macrophages had been induced in the peritoneal cavity with thioglycolate. These cells were washed and resuspended to 2.5 x 10<6>/50 µl in Hepes-buffered RPMI medium containing 10% fetal calf serum. Cells in 50 µl samples were incubated with 200,000 cpm <125>I-LIF-A (10 µl, in the same medium) and 10 µl control medium or serial dilutions of unlabeled pure murine LIF-A or yeast-derived murine LIF, diluted two times. At appropriate concentrations, supernatant from galactose-induced yeast containing the LIFmut2 construct (Fig. 7) competed with 12<5>I-LIF-A for binding to its receptors on peritoneal cells, whereas uninduced yeast supernatant did not (Fig. 9). The degree of competition was similar to that of authentic native LIF-A, indicating that the yeast-derived LIF-A contained all the information necessary for binding to LIF-A cellular receptors.

EXAMPLE 3

The following example shows the steps used to express recombinant murine LIF in mammalian cells.

Accompanying figure 10 relates to step 1 of the method described under: The nucleotide sequence of the cDNA part of clone PLIFNK3. The nucleotide sequence of the mRNA synonymous strand is present in the orientation 5' to 3' with the deduced amino acid sequence of LIF given above. The N-terminal amino acid residue determined previously is indicated as +1.

The Eco RI site at the 5' end of the cDNA and the Xba I site spanning the stop codon (used when inserting the LIF coding region into pMP-Zen in step 2) are indicated.

Step 1: Determination of the amino acid sequence of the N-terminal leader sequence of murine LIF.

The cDNA clones pLIF7.2b and pLIFNKl (see above) contain incomplete copies of LIF mRNA. Although together they contain the complete coding region for the mature part of the LIF protein, they do not contain the complete hydrophobic leader sequence required for secretion of LIF from mammalian cells.

To isolate a cDNA clone containing the region coding for the hydrophobic leader, additional IO<6> cDNA clones constructed as above (Ex. 2) were screened (as above), using the same batch of Krebs mRNA as a template, using two nucleotides that probe corresponding to a sequence of 35 residues at the 5'-end and 36 residues at the 3'-end of the pLIF7.2b sequence (nucleotides 67-102 and 500-535 in Fig. 5).

Two plaques representing clones A,LIFNK2 and Å,LIFNK3 that were positive on duplicate filters were picked and rescreened at lower density, as before. When Å.LIFNK3 was shown to hybridize with each of the two oligonucleotides used, this was chosen for further analyses.

The cDNA insert of approximately 1400 bp in A.LIFNK3 was subcloned into the plasmid vector pEMBL 8" (Dente, L., et.al., Nucleic Acids Res. 11:1645-1655, 1983) to form clone pLIFNK3. Nucleotide sequence analysis of the cDNA the insertion of pLIFNK3 was carried out as for pLIF7.2b and pLIFNK1 (above).

The nucleotide sequence of the cDNA insert in pLIFNK3 is shown in fig. 10, and indicates that pLIFNK3 contains a complete LIF coding region: there is an initiation codon (AUG) in the 23-25 position of the pLIFNK3 sequence, which precedes a sequence that codes for a typical hydrophobic leader sequence of 24 amino acid residues. The coding region extends to the same translational stop codon as defined earlier by pLIFNK1 (above).

Step 2: Introduction of a LIF coding region into a mammalian expression vector.

The mammalian expression vector chosen initially was the retroviral expression vector pMP-Zen. The vector is derived from the Moloney murine leukemia virus-based vector pZIPNeo SV(X)

(Cepko, C.L. et.al., Cell 37:1053-1062, 1984) by omitting the neo<R> gene along with nearby SV40 and plasmid sequences, retaining an Xho I expression site. The 3' region of the vector is also modified to incorporate the long terminal repeat (LTR) enhancer of the myeloproliferative sarcoma virus (MPSV) (Bowtel, D.D.L. et.al., Mol. Biol. Med. 4:229-250, 1987 ). This vector was primarily chosen since the LIF/PMP-Zen recombinant can be "packaged" into helper-free infectious retrovirus particles by passage through \|/2 cells

(Mann, R., et al., Cell, 33:153-159, 1983). Such virus particles can then be used to effectively introduce (by infection) the LIF/pMP-Zen recombinant into a number of murine cell types (Mann, R. et.al., Cell 33:153-159, 1983). Second, this particular vector, which uses MPSV LTRs for expression of the promoter coding region, has been shown to direct the efficient expression of other specific hematopoietic growth and differentiation factors including GM-CSF.

The segment of pLIFNK3 is selected for insertion into pMP-Zen extending from position 1 (an Eco Ri site) to position 630 (an Xba 1 site spanning the stop codon) - see fig. 10. This segment was chosen as it contains little more than the region coding for pre-LIF and excludes the entire 3' untranslated region, which may contain sequences that confer mRNA instability (eg Shaw, G. and Kamen, R., Cell 46:659-667, 1986; Verma, I.M. and Sassone-Corsi, P., Cell 51: 513-514, 1987). To insert this segment into pMP-Zen, it was first inserted between the Eco RI and Xba I sites of pIC20H (Marsh, J.L. et.al., Gene 32: 481-485, 1984) using standard techniques (Maniatis et. al., 1982, supra). The cDNA insert was then recovered from the polylinker of the pIC20H plasmid by Sal I plus Xho I cleavage, thus forming a LIF cDNA fragment with cohesive ends suitable for inserting into the Xho I cloning site of pMP-Zen. The insertion of the LIF cDNA fragment into pMP-Zen was accomplished by standard techniques (Maniatis et al., 1982, supra).

Step 3: Introduction of the pMP-Zen/LIF recombinant into murine fibroblasts.

pMP-Zen/LIF DNA was introduced into vj/2 fibroblasts by electroporation (Potter et al., PNAS 81:7161-7165, 1984).

30 µg of pMP-Zen/LIF DNA plus 3 µg of pSV2Neo DNA (Southern, P.J. and Berg, P., J. Mol. App. Genet 1:327-341, 1982) were mixed with 1 x 10<6>\jf2 fibroblasts in 1 ml DME/10% FCS (Dulbecco's modified Eagle's medium containing 10% fetal calf serum) and subjected to a pulse of 500v at a capacitance of 25 uF (using a BioRad Gene-Pulser model No. 1652078). Transfectants were initially selected on the basis of resistance to the antibiotic G418 conferred by pSV2Neo DNA. G418 resistant Xf2 cells were selected in 400 µg/ml G418 using standard procedures (Mann, R. et.al, Cell 33: 153-159, 1983). Of the 19 G418-resistant clones examined, 2 also contained the pMP-Zen/LIF construct as determined by LIF activity detectable in γ2 conditioned medium.

Step 4: Determination of the biological properties of vj/2 — derived LIF.

Assays for differentiation-inducing activity and leukemia-suppressive activity of conditioned medium from pMP-Zen/LIF containing \}/2 cells were carried out as described previously. Medium from cultures of IO<6> \\ f2 cells in 3 ml DME/10% FCS was measured for differentiation-inducing activity. The results for two positive cultures are shown in the following table:

Thus, significant levels of biologically active recombinant murine LIF can be produced in this expression system.

Step 5: Transmission of pMP-Zen/LIF retrovirus \|/2 cells to hemopoietic cells.

Infectious pMP-Zen/LIF retrovirus could be transferred from the "parental \j/2 cell lines to hemopoietic cells of the line FDC-P1 (Dexter, T.M. et.al., J. Exp. Med. 152:1036-1047, 1980) by co-cultivation. 10<6>\|/2 cells were mixed with 10<6> FDC-P1 cells in 10 ml DME/10% FCS containing the optimal concentration of WEHI-3BD" conditioned medium for the growth of FDC-P1 cells . After 2 days' incubation, the non-adherent FDC-P1 cells were removed, washed free of all adhered γ2 cells and cultured for 16 hours in 3 ml of the same medium. Treated medium was harvested and measured for differentiation-inducing and leukemia-suppressive

activity as described earlier. The results are shown in the following table:

Thus, \y2 clones, \j/2-lla and \|/2-lld, are capable of transmitting biologically active pMP-Zen/LIF retroviruses to hemopoietic cells. These clones will therefore be able to be used to infect normal murine hemopoiesis progenitor cells which can be used to reconstitute the hemopoiesis system of radiation-treated mice to study the effects of high-level expression of native LIF on normal murine hemopoiesis, in a manner analogous to that used for Zen/GM-CSF virus infection.

EXAMPLE 4

The following example shows the steps used to express recombinant murine LIF in E.coli cells.

Accompanying fig. 11 relates to step 1 of the method described under: The nucleotide sequence in the glutathione-S-transferase/thrombin cleavage site/LIF link in plasmid pGEX-2T/LIF, and the sequence in the link of the coded fusion protein. The C-terminal end of glutathione-S-transferase, the thrombin cleavage site and the N-terminal end of the murine LIF parts of the tripartite fusion protein together with the nucleotide sequence encoding this amino acid sequence are shown. The expected site of cleavage by thrombin is indicated by an arrow.

Step 1: Introduction of a LIF coding region into an E.coli expression vector.

The expression vector used to express LIF in E.coli was pGEX-2T (Smith, D.B. and Johnson, K.S., Gene, 1988), which directs the synthesis of foreign polypeptides as fusions to the C-terminus of Sj26, a 26kD glutathione-S-transferase. (E.C. 2.5.1.18) encoded by the parasitic helminth Schistosoma japonicum. In most cases, fusion proteins have been shown to be soluble in aqueous solutions and can be purified from impure bacterial lysates under non-denaturing conditions by affinity chromatography on immobilized glutathione. The special vector pGEX-2T has been constructed so that the glutathione-S-transferase carrier can be cleaved from the fusion protein by cleavage with the site-specific protease thrombin.

The complete coding region of murine LIF, derived from plasmid pLIFmut2 (see Ex. 2 and Fig. 7 above) was inserted as a Barn HI - EcoRI fragment into the multiple cloning site of pGEX-2T (Smith D.B. and Johnson, K.S., supra) , thus placing the LIF coding region 3' and in the same translational reading frame as that of the glutathione S-transferase and thrombin cleavage site. Thus, the LIF protein will be located C-terminal to these elements in a tripartite glutathione-S-transferase/thrombin cleavage site/LIF fusion protein (see Fig. 11). It should be noted that the position of the thrombin cleavage site is such that two amino acid residues (Gly-Ser) will be attached to the N-terminal end of the LIF protein after thrombin cleavage. The construction of the above plasmid, pGEX-2T/LIF, was achieved by standard techniques and the plasmid was introduced into E.coli NM 522 (Gough, J.A. and Murray, N.M., J. Mol. Biol. 166: 1-19, 1983) using standard techniques.

Step 2: Expression and purification of glutathione-S-transferase/LIF fusion protein.

To induce the expression of the glutathione S-transferase/LIF fusion protein, 10 ml cultures of E.coli NM 522 cells containing pGEX-2T/LIF were grown to logarithmic phase in Luria extract, and isopropyl~β-D-thiogalactopyranoside was added to a concentration of 0.1 mM. After a further 4 hours of growth, during which the glutathione-S-transferase/LIF gene is expressed, the cells were harvested by centrifugation and resuspended in 1 ml of mouse tonicity phosphate-buffered saline (MTPBS: 150 mM NaCl, 16 mM Na2HP04, 4 mM NaH2PO4, pH 7, 3). Cells were lysed on ice by gentle sonication and after addition of Triton X-100 to 1%, degraded cell material was removed by centrifugation (10,000 g, 5 min. 4°C). The clarified supernatant was mixed at room temperature in a 50 ml polypropylene tube on a rotating plate with 200 µl 50% glutathione agarose beads (sulfur bond, Sigma). (Before use, the beads were pre-swollen in MTPBS, washed twice in the same buffer and stored in 1 MTPBS at 4°C as a 50% solution, v/v. ) After absorption (5 min.), the beads were collected by centrifugation (500 g, 1 min.) and washed three times in MTPBS/Triton X-100. The glutathione-S-transferase/LIF fusion protein was then eluted by competition with free glutathione: the beads were incubated with 100 µl of 50 mM Tris. HCl, 5 mM reduced glutathione (pH 7.5) for 2 min. at room temperature and the spheres were removed by centrifugation. The elution step was carried out twice, and the two samples of 100 µl of the eluate were pooled.

100 µl of the glutathione-S-transferase/LIF fusion protein was treated with thrombin as described (Smith, D.B. and Johnson, K.S., supra).

1 µl samples of uncleaved and thrombin-cleaved glutathione-S-transferase LIF fusion proteins were electrophoresed on a Pharmacia Phast Gel (8-25% polyacrylamide gradient gel). After staining with Coomassie Blue, a single protein species with a relative molecular weight of -4 6 kDa appeared in the non-cleaved preparation, and in the thrombin-cleaved preparation two major bands of ~2 6 kDa and -20 kDa appeared (corresponding to the expected sizes of the fusion protein (46 kDa), glutathione-S-transferase (26 kDa) and LIF proteins (20 kDa) respectively). By estimating the mass of the Coomassie-stained bands on this gel, the yield of glutathione-S-transferase/LIF protein from the original 10 ml E.coli culture was estimated to be -1.5 µg.

Assays for differentiation-inducing activity and leukemia-suppressive activity of both the glutathione-S-transferase/LIF fusion protein and the thrombin-cleaved LIF preparation were performed using murine M1 cells (as described previously). Both preparations of LIF were determined to be biologically active, with specific activities in excess of 7x10<6 >units/mg.

EXAMPLE 5

The following example shows the steps used to determine whether the murine genome contains any other genes closely related to the gene encoding the sequence present in clone pLIF7.2b (Ex. 1), and deriving a map of the location of restriction endonuclease cleavage sites around the LIF gene.

The accompanying figures relate to different steps in the method described below: Figure 13: relates to step 1: hybridization of mouse DNA with a pLIF7.2b-derived probe under more and less stringent conditions. BALB/c liver DNA cleaved with the indicated restriction enzymes was examined with respect to LIF gene sequences as described in ex. 5 step 1. In the left lane, hybridization was performed at 65°C in 2 x SSC, and final washing was performed at 65°C in 0.2 x SSC. In the right lane, hybridization and washing were both carried out in 6 x SSC at 65°C. In the left lane, linear pLIF7.2b plasmid DNA (5.6 kb) was included in an amount equivalent to 10 and 1 copy per haploid mouse genome (400 pg and 40 pg, respectively) with an estimated molecular weight for the haploid mouse genome of 3 x 10<9> bp (Laird, C.D., Chromosoma, 32: 378-406, 1971). The size of the hybridization genome fragments is indicated. Figure 14: relates to step 2: Hybridization of mouse DNA cleaved with different restriction endonucleases, either alone or in paired combinations, with a mouse LIF probe. The cleaved DNA was electrophoresed on 0.8% agarose gels and hybridized with the pLIF7.2b cDNA fragment as described in ex. 5, step 1, under very stringent conditions. Figure 15: relates to step 2: Restriction endonuclease cleavage map of the murine LIF gene. The region of chromosomal DNA containing sequences corresponding to the cDNA clone pLIF7.2b is indicated by a square. The restriction seats provided above (Eco R5) and below (Xba I, Barn HI, Pst I and Stu I) the main line, have not been briefed regarding the main line. The relative location of seats within the complex below the line is as indicated. Figure 16: relates to step 3: Chromosomal determination of the murine LIF gene. In field 1 BALB/c mouse embryo DNA was electrophoresed, in field 2 DNA from Chinese hamster ovary cells and in fields 3-8 DNA from the hybrid clones I-18A HAT, I-18A-2a-8AG; EBS 58; EBS 11; EBS 4; and I-13A-la-8AG (Cory, S. et al, EMBO J. 2: 213-216, 1983). The murine chromosome content of these hybrids is given in Cory S. et.al., (EMBO J. 2: 213-216, 1983). All DNA was digested with Barn HI. The position of the 3 kbp Barn HI fragment carrying the murine LIF gene is indicated.

Step 1: Determination of the number of LIF-related genes in the murine genome.

On the basis of the data (see ex. 1 in NO patent application no. 885339) that medium conditioned with Krebs II cells contains two biochemically separable but functionally similar factors which are capable of inducing the differentiation of M1 cells (LIF-A and LIF -B), one wanted to determine how many genes there are in the murine genome related to the species that have been purified and cloned. Southern blotting of murine genome DNA cleaved with different restriction endonucleases was therefore hybridized with a probe derived from pLIF7.2b under both highly stringent and less stringent conditions.

20 (lg samples of high molecular weight genomic DNA from BALB/c mouse livers were digested completely with various restriction endonucleases, fractionated by electrophoresis in 0.8% agarose gels and transferred to nitrocellulose. Before hybridization, the filters were incubated for several hours at 65°C in either 6 x SSC (low stringency conditions) or 2 x SSC (high stringency conditions) (SSC = 0.15 M NaCl, 0.015 M sodium citrate), 0.2% Ficoll, 0.2% polyvinylpyrrolidone, 0.2% bovine serum albumin . were then washed in either 6 x SSC, 0.1% SDS at 65°C (low stringency conditions) or in 2 x SSC, 0.1% SDS at 65°C, followed by 0.2 x SSC, at 65° C (very stringent conditions) as given in detail in the text of Fig. 13. The hybridization probe used was as previously given, the approximate 750 bp Eco RI - Hind III fragment spanning the cDNA insert in pLIF7.2b radiolabeled to a specific activity of about 4 x 10<8> cpm/ug by nick translation and included in the hybridization at a concentration of about 2 x 10< 7 >cpm/ml.

In mouse liver DNA (Fig. 13) as well as in DNA from Krebs II cells, LB3 T cells and WEHI265 monocytic cells (not shown), the LIF probe detected unique Eco RI Barn HI and Hind III fragments of approximately 11.3 and 13 kbp, respectively . The same hybridization pattern was evident at both highly (0.2 x SSC, 65°C) and less stringent conditions (6 x SSC, 65°C) (Fig.

13). When hybridizing under less stringent conditions and washing conditions, no further bands appeared. Likewise, unique hybridization fragments were also detected in Pst I, Stu I, Sac 1, Eco R5 and Bgl II cleaved genomic DNA (Fig. 13). Moreover, in the experiment shown in Fig. 11, pLIF7.2b plasmid DNA was included at a concentration equivalent to 10 and 1 copy per haploid mouse genome (lanes 1 and 2). The hybridization intensity of the genomic LIF sequences was not very different from the intensity of the unique gene standard (lane 2).

Taken together, the foregoing data indicate that the LIF gene is unique in the murine genome, with no close relatives. Thus, it is unlikely that the two species of LIF are products of different genes, but rather that they represent post-transcriptional or post-translational variants of the same gene product.

Step 2: Derivation of a restriction endonuclease cleavage map of the murine LIF gene.

In order to determine the location of restriction endonuclease cleavage sites in and around the murine LIF gene, and thus provide a molecular fingerprint of this gene, DNA from mouse livers or from Krebs ascites tumor cells was digested with different restriction endonucleases and subjected to Southern blotting as described in step 1 (under very stringent hybridization and washing conditions). Examples of such experiments are shown in fig. 14. Analyzes of the size of the cleavage products make it possible to create a map of the location of each cleavage site around the LIF gene (fig. 15).

Step 3: Chromosomal determination of the murine LIF gene.

To determine the chromosome on which the mouse LIF gene is located, DNA from ovarian somatic cell hybrid cell lines from six mouse/Chinese hamsters was examined and which contain different mouse chromosomes (Cory, S. et.al., EMBO J. 2: 213-216, 1983 ; Francke, U. et.al., Cytogenet. Cell. Genet. 19: 57-84, 1977). Southern blot analyzes carried out as described in step 1 (using very stringent hybridization and washing conditions) indicated that the 3 kbp Barn HI fragment containing the murine LIF gene was absent in all the hybrids (fig. 16). As chromosome 11 is the only mouse chromosome not contained in any of these lines (Cory, S. et.al., EMBO J. 2: 213-216, 1983), a characterization of mouse/Chinese hamster hybrids (Francke, U . et.al., Cytogenet. Cell. Genet. 19: 57-84, 1977), it is likely that the LIF gene is on this chromosome. For the same reason, it also has the murine GM-CSF gene (Barlow, D.P. et.al-, EMBO J. 6: 617-623, 1987) and the Multi-CSF gene (Iheo, J.N. and Kozak, C.A., National Cancer Institute, Frederick Cancer Research Facility, Annual Report, 1984) has been assigned to chromosome 11, which has been confirmed by genetic linkage studies (Barlow, D.P. et.al., EMBI J. 6: 617-623, 1987).

EXAMPLE 6

The following example shows the steps used to establish conditions under which the cloned murine LIF cDNA can be used to identify, by hybridization, a human gene or mRNA or recombinant DNA clones containing human LIF coding sequences.

The accompanying figure (Fig. 17) relates to step 2 of the method described under: hybridization of <32>P-labeled fragment of cDNA from clone pLIF7.2b under different conditions to genomic DNA of both mouse and human origin. In each case, lane 1 contains 15 µg murine (LB3) DNA and lanes 2 and 3 contain 15 µg human DNA (from Raji or Ramos cell lines, respectively). The hybridization and washing conditions used for each filter are described in step 2. The approximate 10 kbp Eco RI fragment containing the murine LIF gene, and the approximate 9 kbp Eco RI fragment containing the human homolog, are shown with arrows. Molecular weight standards are given to the left. Two different auto-radiographic images (16 hours and 62 hours) are shown.

Step 1: Preparation and radioactive labeling of a fragment of cDNA from pLIF7.2b.

pLIF7.2b plasmid DNA was amplified by growth in E.coli MC1061 (Casadaban, M. and Cohen, S., J. Mol. Biol. 138:179-207, 1980), extracted from the E.coli cells using standard procedures (Maniatis, J. et.al., Molecular Cloning, Cold Spring Harbor,

New York, 1982) and purified by CsCl gradient centrifugation. pLIF7.2b plasmid DNA thus purified was digested with restriction endonucleases Eco RI and Hind III to release a cDNA-containing fragment of -770 bp from the vector (pJL3), which was resolved from vector sequences by electrophoresis through a 1.5% agarose gel .

200 ng pLIF7.2 cDNA fragment thus purified was radiolabeled to a specific activity of approximately 3 x 10<8> cpm/ug by nick translation (Rigby, P.W.J., Dieckmann, M., Rhodes, C. and Berg, P., J. Mol. Biol. 113: 237-251, 1977) in a reaction containing 100 µCi [α<32>P]-dATP. The radioactively labeled cDNA fragment was purified from unincorporated label by precipitation in the presence of 1 M sodium perchlorate and 33% isopropanol.

Step 2: Hybridization of Southern blot of mouse and human genome DNA with a pLIF7.2b cDNA probe.

High molecular weight genomic DNA (15 µg) from the murine T-cell line LB3 (lane 1) and from two human B-cell lines (Raji and Ramos; lanes 2 and 3) digested with the restriction endonuclease Eco RI was electrophoresed through a 0.8% agarose gel and transferred to nitrocellulose using standard techniques (Southern, E.M., J. Mol. Biol. 98:503-517, 1975). Five identical Southern blots were prepared containing each of these three DNAs. Before hybridization, the filters were incubated for several hours at 55°C in either 0.9 M NaCl, 0.09 M sodium citrate, 0.2% Ficoll, 0.2% polyvinylpyrrolidone, 0.2% bovine serum albumin, 50 µg/ml heat denatured salmon sperm DNA, 50 ug/ml E.coli tRNA, 0.1 mM ATP and 2 mM sodium pyrophosphate (filters a, b, c, d) or in 0.3 M NaCl, 0.03 M sodium citrate, with the same added ingredients ( filter e). Hybridization with <32>P-labeled pLIF7.2b cDNA prepared in step 1 was carried out in the same solutions as the prehybridization, containing an additional 0.1% sodium dodecyl sulfate, at 55°C (filters a, b, c), or 65°C (filters d, e) for 16 hours. <32>P-labeled cDNA was denatured by boiling before hybridization and was included in the hybridization reaction at -10<7> cpm/ml.

After hybridization, the filters were washed in either 0.9 M NaCl, 0.09 M sodium citrate, 0.1% sodium dodecyl sulfate at 55°C (filter a) 60°C (filter b) or 65°C (filters c, d) or in 0.3M NaCl, 0.03M sodium citrate, 0.1% sodium dodecyl sulfate at 65°C (filter e). After thorough washing, the filters were autoradiographed at -70°C using Kodak XAR-5 film and two intensifying screens. Fig. 17 shows the results of such an experiment, in which two of the hybridization/washing systems described (filters d and e) allowed both the murine LIF gene (present on an approximately 10 kbp Eco RI fragment) and the human LIF gene (present on an approximately 9 kbp Eco RI fragment) was detected over non-specific residual background hybridization. All other conditions tested gave an unacceptably high level of background hybridization (filters a, b, c).

EXAMPLE 7

The following example shows the steps used to obtain a cloned human gene homologous to murine LIF.

The accompanying figures relate to various steps in the method described below. Figure 18: relates to step 1 in the cloning of the human LIF gene: detection of the LIF gene by means of Southern blot hybridization using a mouse LIF cDNA probe. Genomic DNA from the human cell line RAMOS was cleaved with the indicated restriction endonucleases and hybridized under conditions described in ex. 6, step 1, with a mouse cDNA fragment derived from pLIF7.2b. Figure 19: relates to step 1 in the cloning of the human LIF gene: detection of the LIF gene by means of Southern blot hybridization using a mouse LIF cDNA probe under a variety of hybridization conditions. Genomic DNA from the human cell line RAMOS (H) or mouse liver DNA (M) cleaved with restriction endonuclease Barn HI was hybridized with a murine LIF cDNA probe (pLIF7.2b) under a series of hybridization and washing conditions. Temperatures and concentrations of SSC used for hybridization are indicated in the top line, and wash conditions in the second line (see Ex. 7, step 1). Figure 20: relates to step 2 in the cloning of the human LIF gene: restriction endonuclease digestion of three possible LIF gene clones. 1 µg DNA from A. phage of clones A.HGLIF1, XHGLIF2 and X.HGLIF3 was digested with Sal I, Barn HI or Pst I, electrophoresed on a 0.8% agarose gel (left panel), transferred to nitrocellulose and hybridized (as given previously) with PLIF7.2 cDNA. The approximate sizes of the hybridization fragments are indicated. Figure 21: relates to step 3 in the cloning of the human LIF gene: the nucleotide sequence of the mRNA-synonymous strand of a 1.3 kbp segment of XHGLI1 spanning the human LIF gene (H) is given in a 5' to 3' briefing. Corresponding nucleotide sequence of the murine LIF mRNA (M) derived from cDNA clones pLIF7.2b, pLIFNK1 and pLIFNK3 is listed below the human gene and is given in lowercase letters. Identities between mouse sequences and the human sequences are indicated by asterisks. The putative N-terminal residue of mature human LIF, by analogy with mouse LIF, is given as +1. Figure 22: relates to step 3 in the cloning of the human LIF gene: amino acid sequence of human LIF and comparison with mouse LIF. The amino acid sequence of mature murine LIF (M) as determined by direct amino acid sequencing, and analyzes of the cDNA clones pFLIF7.2b, pLIFNKl and pLIFNK3 is given in the top row, with the corresponding sequence of human LIF (H), as deduced from sequence-

one of Å.HGLIF1, given below. Identities are indicated by asterisks.

Figure 23: relates to step 4 in the cloning of the human LIF gene: restriction endonuclease cleavage map of the human LIF gene. Exons of the human gene homologous to pLIF7.2b (Fig. 21) are indicated as squares. The direction of transcription of the gene is indicated by the arrow below the line.

Step 1: Detection of human LIF gene with a mouse probe.

A method has been described previously for the use of a radioactively labeled fragment of mouse LIF cDNA cloned pLIF7.2b as a hybridization probe to detect the human LIF gene.

Figure 18 shows that this method allows the human LIF gene to be detected in human genome DNA cleaved with a number of restriction endonucleases. Analyzes such as this, and other gels not shown, revealed the sizes of DNA fragments formed by a variety of restriction endonucleases on which the human LIF gene is located. Such data are important in the subsequent steps of this example, not only to establish conditions under which the mouse probe can be used as a hybridization probe to detect the human, but also to provide diagnostic restriction map data to aid in the identification of human genome LIF clones.

A greater degree of homology between mouse sequences and the human LIF sequences was evident when mouse and human genome DNA cleaved with Barn HI was hybridized with a mouse LIF cDNA probe under a variety of hybridization and washing conditions, (fig. 19). When the hybridization and washing conditions were made more stringent so that background staining was reduced, a unique ∼3 kbp fragment hybridizing with the mouse probe appeared. The human gene clearly retained significant hybridization even at 65°C in 0.2 x SSC.

Step 2: Screening of a human genome library and isolation of a clone containing the LIF gene.

A library of human genomic DNA, partially digested with Sau 3A and ligated into the X phage cloning vector EMBL3A, was screened for LIF gene-containing clones by hybridization with mouse cDNA as a probe. The fragment of LIF cDNA was radioactively labeled and the hybridization conditions were as given above (ex. 1). Briefly, phage plaques representing the genome library were grown at a density of -50,000 plaques per 10 cm Petri dish and transferred to microcellulose as described (Maniatis, T. et.al., Molecular

Cloning, Cold Spring Harbor, 1982). Before hybridization, filters were incubated for several hours at 65°C in 6 x SSC (SSC =

0.15 M NaCl, 0.015 M sodium citrate), 0.2% Ficoll, 0.2% polyvinylpyrrolidone, 0.2% bovine serum albumin, 2 mM sodium pyrophosphate, 1 mM ATP, 50 ug/ml denatured salmon sperm DNA and 50 ug/ml E. coli tRNA. Hybridization as in the same solution containing 0.1% SDS, at 65°C for 16-18 hours. The LIF cDNA fragment, radiolabeled by nick-translation using [a<32>P] dATP to a specific activity of -2 x 10<8> cpm/ug was included in the hybridization at a concentration of -2 x 10 <6> cpm/ml. After hybridization, the filters were washed thoroughly in 6 x SSC, 0.1% sodium dodecyl sulfate at 65°C and then autoradiographed. Plaques that were positive on duplicate filters were picked and screened again at a lower density, as before.

Three clones were thus identified and purified: XHGLIF1, 2 and 3. To determine the relationship between these clones and to determine which, if any, contain the human LIF gene, DNA was prepared from each clone and digested with these restriction endonucleases: Sal I which releases the entire segment of cloned genomic DNA and Barn HI and Pst I which cleave in and around the LIF gene to form characteristic fragments of -3 kbp and 1.8 kbp and 0.6 kbp respectively (as determined above). After cleavage of recombinant phage DNAs and reresolution by electrophoresis on agarose gels (Fig. 20, left panel), the DNA was transferred to nitrocellulose and hybridized with the mouse LIF cDNA probe (under the conditions given above) to release fragments containing LIF the gene (Fig. 20, right panel). This analysis showed that (a) all three clones turned out to be identical; (b) all contain a -9 kbp segment of genomic DNA containing a region homologous to the mouse LIF probe;

(c) the region homologous to mouse cDNA is present on a

3 kbp Barn HI fragment and 1.8 and 0.6 kbp Pst I fragments, characteristics of the human LIF gene (see above). Thus, it was concluded that all three clones contain a segment of chromosomal DNA that includes the human LIF gene.

Step 3: Determination of the nucleotide sequence of the human LIF gene.

The ~3 kbp Barn HI fragment of ?iHGLIFl shown above to hybridize with the mouse LIF cDNA probe was recloned into the plasmid vector pEMBL8<+> leading to clone pHGLIFBaml and subjected to nucleotide sequence analysis. Nucleotide sequencing was carried out by the dideoxy chain termination method (Sanger, F. et.al., Pro. Nati. Acad. Sei., USA. 74: 5463-5467, 1977) using alkaline denatured double-stranded plasmid DNA as template, a series of oligonucleotides that are complementary to the sequences in the gene as primers and using both the Klenow fragment of E.coli DNA polymerase I and avian myeloblastosis virus reverse transcriptase as polymerases in the sequencing reactions.

The entire nucleotide sequence of the Barn HI fragment spanning the LIF gene (2840bp) was determined, of which 12 97bp are shown in fig. 21. Alignment of this sequence with the murine LIF mRNA sequence showed that the sequences coding for the mature human LIF protein are present on two exons separated by an intron of 693bp (fig. 21). For the region that codes for the mature protein, there is a high degree of homology between the two species both at the nucleotide and amino acid sequence level. Within exon 1 there is 88% nucleic acid sequence homology (114/129 residues compared) and 91% amino acid sequence homology (39/43) for the region coding for the mature protein (position 58-186). Exon 2 is somewhat less homologous, 77% at the nucleotide level (318/411) and 74% at the amino acid level (101/136) in the coding region. When looking at the mature protein as a whole, mouse and human LIF sequences determined here are identical in 140 out of 179 positions (78%) without insertions or deletions (Fig. 22). Also, many of the differences are very conservative substitutions (Lys:Arg, Glu:Asp and Leu:Val:Ala).

At the 5' end of the codon for the N-terminal proline residue of mature LIF, (position 58 in Fig. 21), the human gene is homologous to the murine LIF mRNA through a region encoding most of the hydrophobic leader. However, the entire leader is not encoded on this exon, as the mRNA and gene sequences diverge in a typical RNA splice site (TCCCCAG) (Mount, S.M., Nucleic Acids Res. 10:459-472, 1982). The exon specifying the 5' untranslated region and the first residues of the leader are not present within 1097 bp 5' of this splice site. In the mouse LIF gene, the exon specifying the first 6 amino acid residues of the hydrophobic leader is located -1.5 kbp 5' of the analogous splice site.

Step 4: Derivation of a restriction endonuclease cleavage map of the human LIF gene.

In order to determine the location of the restriction endonuclease cleavage sites in and around the human LIF gene, and thus provide a molecular fingerprint of this gene, human genome DNA (from the RAMOS cell line) was cleaved with different restriction endonucleases alone and in pairwise combinations and subjected to Southern blot analyzes as described in ex. 5, except that the probe used was the 3 kbp Barn HI fragment derived from pHGLIFBaml described in step 3 above and radioactively labeled by means of nick translation. Analyzes of the thus derived data (not shown) as well as that shown in Fig. 18 and derived from analyzes of A.HGLIF1 and pHGLIFBaml, yielded the restriction endonuclease cleavage map shown in FIG. 23.

EXAMPLE 8

The following example details modifications made to the cloned human LIF gene to allow expression in yeast cells and determination of the biological and biochemical properties of the recombinant human LIF thus derived.

Figure 12: relates to step 1: oligonucleotides used to modify the human LIF gene for incorporation into YEpsecl. Oligonucleotide (a) corresponds to the 5' end of the coding region (residues 31 to 69 in Fig. 21). Oligonucleotide (b) corresponds to the middle of the coding region (residues 163 to 186 and 880 to 903 in Fig. 21).

The part of this oligonucleotide which is complementary to exon 1 is underlined with dashes, and that which is complementary to exon 2 with dots. Oligonucleotide (c) corresponds to the 3' end of the coding region (from position 1279 in Fig. 21). Oligonucleotides (a) and (b) introduce the indicated restriction endonuclease cleavage sites. Figure 24: relates to step 1: nucleotide sequence of, and amino acid sequence coded for, the synthetic human LIF cDNA derived by mutagenesis of the cloned human LIF gene. Barn HI and Hind III cleavage sites introduced by oligonucleotides (a) and (c) (Fig. 12) are indicated. The putative N-terminal amino acid of mature LIF (by analogy with the mouse) is indicated as +1. Figure 25: relates to tin 3: Induction of differentiation in colonies of M1 leukemia cells by dilutions of purified native murine LIF (o o) and treated medium from yeast cells containing the YEpsecl/HLIF recombinant induced with galactose (• •). Medium from non-induced yeast cultures containing the YEpsecl/HLIF recombinant (o o) was inactive. Average data from repeated 7-day cultures are given. Figure 26: relates to step 4: competition between yeast-derived HLIF and native murine 125I-LIF for binding to specific receptors on murine M1 cells. Dilutions of authentic native murine LIF-A (o o) , recombinant murine LIF (• •) and treated medium from yeast cells containing the YEpsecl/HLIF construct that were uninduced (■ ■) and induced with galactose (□ □) were tested for the ability to compete with native <125>I-LIF-A for binding to cellular receptors on murine M1 cells at 37°C, as previously reported.

Step 1: Modification of the human LIF gene for expression in yeast.

The isolation of a recombinant DNA clone containing the human LIF gene and the nucleotide sequence of the above-mentioned gene has been given previously (ex. 7). The nucleotide sequence of 1297 bp of DNA spanning the two exons that code for the mature human LIF protein is shown in fig. 21.

Murine recombinant LIF has previously been produced in yeast cells using the yeast expression vector, YEpsecl (Ex. 2). This vector provides an N-terminal leader sequence derived from the killer toxin gene of Kluyveromyces lactis, transcribed from a galactose-inducible hybrid GAL-CYC promoter.

In order to express the protein encoded by the human LIF gene in this vector, it was necessary to modify the gene in several ways.

At the 5' end of the region coding for the mature protein, a restriction endonuclease Barn HI cleavage site was introduced to allow insertion in frame with the K. lactis leader and to retain an appropriate signal peptidase cleavage site (Gly-Ser). The same modification as previously applied to mouse cDNA (pLIF7.2b) was carried out here. In the middle, the 693 bp intervening sequence was removed, joining two exons in the same translational reading frame. At the 3' end, a second translational stop codon was introduced immediately 3' of the natural stop codon, followed by a Hind III site for insertion into YEpsecl. All modifications were achieved by oligonucleotide-mediated mutagenesis: the ~3 kbp BAM HI fragment spanning the LIF gene was subcloned into the plasmid pEMBL8+, (Dente, L. et.al., Nucleic Acids Res. 11:1645-1655, 1983) and single-stranded DNA was prepared by Fl superinfection. In vitro mutagenesis was carried out as described (Nisbet, I.T. and Beilharz, M.W., Gene Anal. Techn. 2:23 29, 1985), using oligonucleotides of 39, 48 and 39 bases, respectively, to modify the 5' end, the middle and the 3' end of the gene as given above (Fig. 12). The nucleotide sequence of the modified human LIF coding region is given in fig. 24.

Step 2: Introduction of the YEpsecl/HLIF recombinant into yeast cells.

S.cerevisiae strain GY1<+>(leu2 ura3 ade2 trpl cir+; from G. Cesareni, EMBL Heidelberg) was transformed by the polyethylene glycol method (Klebe, R.J. et.al., Gene 25: 333-341, 1983). Transformants were selected and maintained on a synthetic minimal medium (2% carbon source, 0.67% yeast nitrogen base (Difco) supplemented with 50 µg/ml of the required amino acids) under uracil deprivation. Recombinant HLIF was prepared by either of the two methods. Either (1), Ura<+> transformants were grown to stationary phase in a non-selective medium containing 2% galactose, and the medium was assayed for LIF activity or (2), Ura<+> transformants were grown to stationary phase in a selective minimal medium containing 2% glucose. The cells were then washed and resuspended in the same volume of the selective minimal medium containing 2% ethanol and cultured for 8 h to overcome glucose repression. Transcription of the HILF insert was then induced by diluting the cells (1:10) in a synthetic minimal medium containing 2% galactose. Samples of the culture supernatant were removed at various times after induction, filtered through Millipore filters (0.2 µm) and analyzed directly for LIF activity.

Step 3: Determination of the biological properties of the yeast-derived HLIF.

Given the high degree of sequence similarity between mouse and human LIF, the activity of yeast-derived human LIF on murine M1 cells was assessed. Assays of differentiation-inducing activity and leukemia-suppressive activity of yeast-conditioned medium were carried out in 1 ml cultures containing 300 murine Ml cells (from Dr M. Hozumi, Saitama Cancer Research Centre, Japan) in Dulbecco's modified Eagle's medium with a final concentration of 20% fetal calf serum and 0.3% agar. Materials to be analyzed were added in serially diluted 0.1 ml volumes to the culture dishes before the addition of the cell suspension in agar medium. The cultures were incubated for 7 days in an atmosphere saturated with humidity of 10% CO 2 in air. The cultures were scored using a dissecting microscope at 35x magnification, a score that differentiated colonies with a rosette of dispersed cells or that were composed entirely of dispersed cells. Morphological examination of the colonies was carried out by fixing the whole culture with 1 ml of 2.5% glutaraldehyde after which the dry cultures were stained on microscope slides using acetylcholinesterase/Luxol Fast Blue/Haema-toxylin.

Medium from galactose-induced cultures of yeast containing the human coding region of YEpsecl, but not from uninduced cultures of the same yeast cells, from cultures of non-transformed yeast or yeast containing the vector YEpsecl alone, is able to induce typical macrophage differentiation in cultures of Ml colonies (Fig. 25). As with murine LIF, the yeast-derived human material also progressively reduced the number and size of M1 colonies that developed with increasing concentrations. Comparison with purified native murine LIF indicated that yeast-treated medium contained up to 50,000 units/ml of human LIF.

Step 4: Yeast-derived human receptor binding specificity

LIFE.

Purified murine LIF-A was iodinated as indicated previously. Ml cells were washed and resuspended to 2.5 x 10<6>/50 µl in Hepes-buffered RPMI medium containing 10% fetal calf serum. Cells in 50 µl samples were incubated with 200,000 cpm of 125 I-LIF-A (10 µl in the same medium) and 10 µl control medium or serial dilutions (diluted twofold) of unlabeled pure murine LIF-A or treated medium from galactose-induced or uninduced cultures of yeast transformants containing the YEpsecl/HLIF construct.

Medium treated with galactose-induced yeast cells containing the YEpsecl/HLIF recombinant is able to compete with native murine <125>I-LIF-A for binding to specific cellular receptors on murine M1 cells to the same extent as native and recombinant murine LIF-A, at 37°C. (Fig. 26) and at 0°C (not shown). Medium from uninduced yeast cells and from yeast cells containing the vector YEpsecl alone did not compete. Thus, there appears to be a strong conservation of the receptor-binding region in murine and human LIF, which is consistent with the high degree of primary amino acid sequence similarity.

Step 5: Purification, sequencing and iodination of yeast-derived human LIF.

Human LIF in medium treated with galactose-induced yeast cells containing the YEpsecl/HLIF recombinant was purified using steps 2 and 4 in Ex. 1 in NO patent application no. 885339 with the exception that the LIF activity that binds to the DEAE-Sepharosé CL-6E column and is eluted with the salt gradient was collected in step 2. Purified yeast-derived human LIF was labeled with radioactive iodine by incubating 1 (lg human LIF in 50 µl 0.2 M sodium phosphate buffer, pH 7.2, with 1 mCi Na 125I

(2.7 µl) and 5 µl of 0.2 mM ICI in 2 M NaCl for 60 sec. <125>I-LIF was separated from unincorporated 125I by passing the reaction mixture through a column of Sephadex G-25 M (Pharmacia) equilibrated in phosphate-buffered (20 mM, pH 7.4) saline (0.15 M) containing 0.02% Tween 20 and 0.02% sodium azide. Human <125>I-LIF was electrophoresed on an 8-25% gradient sodium dodecyl sulfate-polyacrylamide gel as a single broad band with an apparent molecular weight of 170,000. Human <125>I-LIF bound specifically to murine Ml cells and bone marrow cells and

confirmed the ability of human LIF to bind to murine LIF receptors. Purified yeast-derived human LIF (approximately 10 µg)

was subjected to amino-terminal amino acid sequencing as described in subex. 1 and yielded a single sequence:

Ile-Thr-Pro-Val-X-Ala

This is identical to the predicted amino acid sequence of human

LIF (Fig. 22) except that the sequence begins at it

fourth amino acid compared to the start of the murine sequence (see subex. 1) which is:

Pro-Leu-Pro-Ile-Thr-Pro-Val-Asn-Ala

This indicates that the purified yeast-derived human LIF is lacking

the corresponding first three amino acids to the mouse sequence, but is still biologically active and still able to bind to the murine LIF receptor. The first three amino acids thus prove to be indispensable for the biological activity of

LIFE.

Claims (21)

1. Method for producing a LIF (leukemia inhibitory factor) polypeptide, characterized in that it comprises: (a) providing host cells containing a recombinant DNA molecule that codes for the LIF polypeptide, in combination with transcriptional and translational regulatory sequences, as well as a selection marker, (b) culturing said cells under conditions leading to expression of the polypeptide, and (c) recovery of the polypeptide, wherein the polypeptide comprises a mature polypeptide which is at least 78% homologous to an amino acid sequence selected from those shown in Figures 10, 22 and 24. 2. Method as stated in claim 1, characterized in that the mature polypeptide has complete homology with the amino acid sequence as shown in figure 24. 3. Method as stated in claim 1, characterized in that the polypeptide is glycosylated with the help of the host cell. 4. Method as stated in claim 1, characterized in that the promoter is a galactose-inducible hybrid GAL-CYC promoter. 5. Method as stated in claim 1, characterized in that the signal sequence which codes for expression of a leader sequence which controls the secretion of the polypeptide is derived from the signal sequence in the killer toxin gene of Kluyveromyces lactis. 6. Method as stated in claim 1, characterized in that the vector is derived from Moloney murine leukemia virus. 7. Method as stated in claim 6, characterized in that the vector further comprises the LTR enhancer of myeloproliferative sarcoma virus. 8. Method as stated in claim 7, characterized in that the vector lacks the 3' untranslated region in native LIF mRNA. 9. Method as stated in claim 1, characterized in that the vector is one that directs the expression of a glutathione-S-transferase/LIF fusion protein. 10. Method as stated in claim 9, characterized in that the vector is one that directs the expression of a glutathione S-transferase/thrombin cleavage site/LIF fusion protein. 11. Recombinant DNA molecule that codes for a LIF polypeptide, characterized in that it comprises a nucleotide sequence as indicated in figures 5, 10 (bases 1-63), 21 or 24 (bases 1-543), said polypeptide competing with a polypeptide with an amino acid sequence as indicated in figures 10, 22 or 24 for binding to specific cellular receptors on M1 cells or murine or human macrophages. 12. Recombinant DNA molecule as set forth in claim 11 comprising a nucleotide sequence which, upon expression in a suitable host cell, codes for the LIF polypeptide, characterized in that said nucleotide sequence hybridizes under conditions of 6 x SSC, 65°C or stricter conditions, to a sequence of a LIF-coding insert selected from the LIF-coding inserts in clones pLIF7.2b, pLIFNKl, pLIFNK3 and pHGLIFBaml as defined in Figures 5, 6, 10 or 21. 13. Molecule as stated in claim 12, characterized in that the finished polypeptide has complete homology with the amino acid sequence as shown in figure 24. 14. Molecule as stated in claims 11 - 13, characterized in that it further comprises a replication starting point whereby said molecule is made suitable for use as a cloning vector. 15. Molecule as stated in claim 14, characterized in that it further comprises a promoter sequence which is operably linked to the said nucleotide sequence whereby the said molecule is made suitable for use as an expression vector. 16. Host cell, characterized in that it is transformed with a recombinant DNA molecule as stated in claim 11 and has the ability to express the LIF polypeptide. 17. Host cell as stated in claim 16, characterized in that the host cell is a yeast cell. 18. Host cell as stated in claim 17, characterized in that the host cell is of the species Saccharomyces cerevisiae. 19. Host cell as stated in claim 16, characterized in that the host cell is a mammalian cell. 20. Method as stated in claim 19, characterized in that the host cell is haemopoietic cells. 21. Method as stated in claim 16, characterized in that the host cell is E. coli cells.