The invention relates to both a pharmaceutical composition having an endoproteolytic activity and a process for the (micro)biological production of a protein and for the in vitro cleavage of a protein, in particular a precursor protein by processing the protein with an endoproteolytically active enzyme.

The invention described herein is the result of a further study into the possible physiological significance of furin, a human protein described in European patent application EP-A- 0 246 709, which is the expression product of the fur gene located in the genome upstream of the human fes/fps proto- oncogene. The patent application referred to and other publications by the same research group (Roebroek et al. , Molec.Biol.Rep. H, 1986, 117-125; Roebroek et al. , EMBO J. __, 1986, 2197-2202; and Schalken et al. , J.Clin. Invest. _r 1987, 1545-1549) show that on the basis of the limited DNA data then available, it was impossible to determine the function of the product of the fur gene. What could be determined was that the furin is probably a membrane- associated protein which has a function in which certain recognition structures play a role. It was also observed at the time that the fur gene is expressed as a 4.5 kb mRNA in liver, kidney, spleen, thymus and brain, whereas the expression in lung tissue is very slight; in non-small-cell lung carcinomas, on the other hand, a highly increased expression was found to occur, on the ground of which the fur gene was suggested to have a utility as a tumor marker.

Within the framework of the above research, the complete nucleotide sequence of a genomic DNA fragment of about 21 kbp containing the fur gene has meanwhile been determined (Van den Ouweland et al. , Nucl.Acids Res. H, 1989, 7101-7102), while the nucleotide sequence of the corresponding fur cDNA has also

been determined (Van den Ouweland et al. , Nucl. Acids Res .18, 1990, 664) . On the basis thereof it is now possible for the fur gene to be completely characterized, at both the level of genomic organization structure and the level of the encoding sequences. From these encoding nucleotides sequences, the amino acid sequence of the furin can also be derived.

A computer analysis of this amino acid sequence has now surprisingly revealed that furin is highly similar to subtilisin-like proteases as encoded in yeast by the KEXl gene of Kluvveromyces lactis and the KEX2 gene of Saccharomyces cerevisiae _r and that furin is evidently the higher-eukaryotic form (found in Man- and in animals, such as monkey, cat, rat, mouse, chicken and Drosophila) of these endoproteases . More specifically it has been found that the furin exhibits a certain degree of homology with the catalytic domain of the hitherto described bacterial subtilisins (about 20 enzymes), such as thermitase of Thermoactinomyces vulσaris and subtilisin BPN' of Bacillus amyloliquefaciens. and exhibits a striking high homology with subtilisin-like proteases, such as the expression product of the KEXl gene of the yeast

Kluvveromyces lactis and the expression product of the KEX2 gene of the yeast Saccharomyces cerevisiae . The furin, which contains 794 amino acids, exhibits in the domain of the amino acids 97 to 577 an overall homology of about 80.0% with the amino acids 123-584 of the expression product of said KEXl gene (i.e., 41.6% identical amino acids and 38.3% conservative substitutes) and an overall homology of about 78.9% with the amino acids 134-597 of the expression product of said KEX2 gene (i.e., 39.4% identical amino acids and 39.5% conservative substitutes) . These amino acid regions of the yeast proteases comprise the subtilisin-like catalytic domains . The subtilisin-like domain of furin is situated in an amino- terminal furin fragment comprising the amino acids 108-464. With regard to the subtilisin-like proteases, reference is made to the following publications: Tanguy-Rougeau et al . , FEBS Letters 234, 1988, 464-470; Mizuno et al. , Biochem.

Biophys. Res Commun. 156, 1988, 246-254; Meloun et al. FEBS Letters 183, 1985, 195-200; Marklan et al . , J.Biol.Chem. 242. 1967, 5198-5211; Mizuno et al. , Biochem. Biophys. Res.Commun. 159. 1989, 305-311; Bathurst et al. , Science 23_5_, 1987, 348- 350; Thomas et al., Science 241, 1988, 226-230; Foster et al., Biochemistry ___., 1990 347-354; Fuller et al., PNAS USA §___, 1989, 1434-1438; Julius et al. , Cell 3_7, 1984, 1075-1089; Bourbonnais et al. , J.Biol.Chem. 263. 1988, 15342-15347; • Cosman et al . , Dev.Biol. Stand. ____, 1988, 9-13; Schubert 0 Wright et ^' al. , Nature 221, 1969, 235-242; Cunningham et al. , Yeast 5_, 1989, 25-33; Davidson et al., Nature 333, 1988, 93- 96.

As shown by the above publications, it is especially the expression product of the KEX2 gene of the yeast species 5 Saccharomyces cerevisiae which has been well studied and characterized. It is a membrane-associated, calcium ions dependent endopeptidase with an enzyme specificity for paired basic amino acid residues; substrate proteins are cleaved at the carboxyl site of pairs of basic amino acids containing C arginine by this enzyme, which is to be defined here as a

"restriction" endopeptidase (by analogy to the nomenclature in restriction endonucleases in which a given nucleotide sequence is determinative of the cleavage of the DNA) . The location of the enzyme is probably in a structure of the Golgi complex. 5 The subtilisin-like domain and the Ca ²⁺ activation sequences are in the aminoterminal part of the protein. In the yeast Saccharomyces cerevisiae, the endopeptidase is involved in the proteolytic processing of precursors of killer toxin and pairing pheromone alpha factor, i.e., of pro-killer toxin and 0 pro-alpha factor. Furthermore, the endopeptidase is found to be capable of correctly cleaving the mouse neuroendocine- peptide precursor prepro-opiomelanocortin after introduction into certain mutant mammalian cell lines with disturbed proteolytic processing, and to ^' be capable of processing 5 proalbumin to mature albumin and to be capable of processing the precursor of the plasma C protein.

On the ground of the established similarities between the above known endopeptidases and the furin, it is postulated that the furin is a restriction endopeptidase which can be used for the processing of proteins, more specifically the processing of precursor proteins of polypeptide hormones, growth factors, toxins, enzymes, or other types of biologically relevant proteins . In this connection, in vitro applications are conceivable on the one hand, and in vivo applications on the other, including an application within the framework of a therapeutic treatment. For such applications, the human furin may be more suitable than the above known endopeptidases of non-human origin, or more generally an animal "furin" may be more suitable than an endopeptidase from lower organisms. The same applies to analogues or relatives of ^" furin not yet isolated, referred to herein as furin-like enzymes, belonging to a larger family of restriction- endoproteolytic enzymes of which furin is the first-found representative. The various members of this family will exhibit a high structural resemblance, although the sequence homology may be quite low, possibly as low as below 50% homology. Within this family, it will be possible to distinguish several enzyme classes _? such as a group of furin- like enzymes involved in the processing of constitutively secreted proteins and a group of furin-l±ke enzymes involved in the processing of proteins whose secretion is regulated (secretion through secretory granula) . It is possible that each of these furin-like enzymes is characteristically expressed in a limited number- of cell types in which the enzyir. is active as a processing enzym. A limited degree of overlap between the cell and tissue distribution of these enzymes is also conceivable and could very well be responsible for the known phenomenon of cell type-dependent differential processing of precursors.

The pituitary proteins PCI and PC2, described recently by Seidah et al. , DNA and Cell Biol.- __., 1990, 415-424, constitute examples of such furin-like .enzymes .

Through recombinant DNA techniques, it is possible to obtain large quantities of the protein furin. In prokaryotes, the fur gene can be expressed as a fusion protein with beta- galactosidase (pUR vector system) or the anthranilate synthetase (pATH vector system) . Another possibility is the synthesis of the fusion protein glutathion-S-transferase-furin (pGEX) . The advantage of this approach is that the furin can be split off by means of thrombin. The furin can also be synthesized as such in prokaryotes by placing the cDNA in the correct manner behind a suitable promotor. The pUR and pATH vector systems have been described in the European patent application referred to hereinbefore. pGEX is commercially available. Using the strong SV40 promotor, the fur cDNA can be expressed in suitable eukaryotic cells. In connection with glycosylation of the protein, this approach is preferred for certain purposes .

Furin can be purified by standard biochemical techniques in the presence of protease inhibitors. Furin is active in a relatively acidic medium with a pH of 5.5, as occurs in secretory granula, but the protein maintains its activity also at pH 7.5. By virtue of this, a 0.2 M sodium acetate buffer (pH 5.5) or Tris-HCl buffer (7.0) may be used in. vitro. The activity of the enzyme furin depends on the presence of C ² " ions . For the in. vitro enzyme activity, a calcium concentration of 2-5 mM has been found to be optimal. The presence of metal chelators such as EDTA will greatly inhibit the activity of furin. Furthermore, the presence of heavy metal ions such as Zn ²⁺ , Hg ²⁺ and Cu ²⁺ should be avoided. The substance o-phenanthrolin binds heavy metals except Ca ²⁺ and thus has no adverse effect on the enzymatic activity of furin. Low concentrations of phenyl methyl sulphonyl fluoride (PMSF) and diisopropyl fluorophosphate (DFP) up to 5 mM have no inhibitory effect. At higher concentrations of PMSF, the enzyme function is inhibited. An in, vitro incubation for two hours at 37°C is sufficient for the processing of the protein to be cleaved.

Furin can be used for the endoproteolytic processing of various proteins. This makes it possible, for example, for in vitro produced precursor proteins to be specifically cleaved to form biologically active compositions which may be used as additional agents for the treatment of diseases in which the precursors are not split or to ^" an insufficient degree. Generally speaking,- furin may be said to be suitable in the processing of biologically relevant proteins.

The protein furin may also find application as a medicament, so that patients deficient in an endoprotease may be treated by administering furin, so that an adequate processing of precursor proteins is yet possible. As a result, the cleavage products may perform their function, and it will be possible for any disturbingly high levels of precursor proteins to be reduced.

Furin is possibly also applicable for clearing depositions with substrate proteins in, for example, the blood circulation system, so that obstructions of vital organs may be remedied by the administration of furin. Furin is further also applicable in the commercial production of all sorts of biologically active substances (e.g., other enzymes) if processing is a production step therein.

The invention relates in the first place to a pharmaceutical composition comprising one or more pharmaceutically acceptable carriers, diluents or adjuvants, as well as an endoproteolytically active amount of furin or a furin-like enzyme, or a fragment or derivative of furin or furin-like enzyme having an endoproteolytic activity. The proteolytic activity is maintained when the carboxy- terminal region with the transmembrane domain therein has been split off. Instead of the complete furin or furin-like enzyme, therefore, according to the invention, use can be made of a fragment of the enzyme which still contains the part responsible for the proteolytic activity. One suitable

fragment is, for example, the furin fragment consisting of amino acids 108-464.

The activity of the furin or furin-like enzyme, or of an endoproteolytically active fragment thereof, can further be manipulated by introducing mutations. The invention accordingly also extends to derivatives of furin or furin-like enzyme still having endoproteolytic activity.

According to a preferred embodiment according to the invention of such a pharmaceutical composition, the furin or the furin-like enzyme, or the fragment or derivative of furin or furin-like enzyme having endoproteolytic activity, used in the composition, has been obtained from prokaryotic or eukaryotic cells which through genetic engineering with recombinant DNA or RNA have acquired the ability of expressing the furin, furin-like enzyme, fragment or derivative of furin or furin-like enzyme, whether or not in the form of a fusion protein, while in case the furin, furin-like enzyme, fragment or derivative of furin or furin-like enzyme is produced by the cells as a fusion protein, the fusion protein has been processed to split off the furin, furin-like enzyme, fragment or derivative of furin or furin-like enzyme from the fusion protein.

Another possibility, however, is for the source of the furin or furin-like enzyme used to be cells which by nature are capable of producing the furin or furin-like enzyme, for example, a suitable tumor cell line.

A particularly preferred embodiment of the invention concerns a pharmaceutical composition containing furin itself. An alternative particularly preferred embodiment of the invention concerns a pharmaceutical composition comprising an aminoterminal fragment of furin comprising at least the amino acids 108-464 of furin.

The invention further relates to a process for the in vitro cleavage of a protein by treating the protein w th an endoproteolytically active enzyme, in which, in accordance with the present invention, the protein is treated in the

presence of Ca ²⁺ ions with furin or a furin-like enzyme, or an endoproteolytically active fragment, derivative or fusion protein of furin or furin-like enzyme as the endoproteolytically active enzyme. 5 The treatment will commonly be carried out at physiologically occurring pH and temperature values, i.e., at a pH within the range of 4-9 and at a temperature of about 37°C. .

Preferably, the treatment is carried out at a pH of 5- 10 7.5, more preferably 5.5-7.0-.

Also, according to the invention, it is preferable for the treatment to be carried out at a temperature of 20-50°C, more preferably 30-40°C.

Furthermore, according to the invention, it is preferable 15 for the treatment to be carried out at a calcium concentration of 1-10 mM, more preferably 2-5 mM.

According to a particularly preferred embodiment of the invention, the treatment is carried out in the presence of o- phenanthrolin or an equivalent agent for binding heavy metals 20 other than calcium.

The process according to the invention comprises a treatment of a substrate to be processed as such with furin (or with a furin-like enzyme) as such, i.e., furin in an isolated or purified form, but also comprises a treatment with 25 or within cells, in particular genetically engineered mammalian cells in which furin is expressed. Preferably, these are carefully selected, genetically engineered mammalian cells (such as COS-1 cells, CHO cells and endothelial cells) with high levels of expression of both the fur gene and a gene

^• 3Λ encoding for the substrate to be processed. As well known to those skilled in the art, a greatly enhanced expression can be realized by gene amplification or by using strong proirtotors. The invention even extends to applications involving transgenic animals, and is therefore not limited to in vitro protein production and protein cleaving processes. The invention accordingly also comprises mammalian cells and

mammals comprising DNA originating from recombinant DNA encoding for furin or furin-like enzyme, and capable of expressing the furin or furin-like enzyme. Endothelial cells, for example, are particularly ideal as mammalian cells for transport of therapeutic genes and gene products through the body by reason of their distribution throughout the entire body (at the surface of blood vessels, lung tissue, and the like) and by virtue of their interaction with all sorts of components in the circulation of body fluids (the bloodstream and the like) . Thus endothelial cells of a patient suffering from some disease resulting from a disturbance in the processing of pro-proteins, could be genetically engineered after being isolated from the body to remedy the defect by introducing an active fur gene, and subsequently the genetically modified cells could be re-transplanted into the patient. Such a gene therapy, however, is not limited to endothelial cells .

A preferred embodiment of the invention consists in a process for the (micro)biological production of a protein by culturing genetically engineered cells expressing a pro-form of the protein as well as furin, and possibly isolating the protein formed. For this purpose both prokaryotic and eukaryotic cells can be used, but cells of higher eukaryotes are preferred. For example, yeast cells or still better plant cells can be used. It is, however, particularly preferable to use genetically engineered mammalian cells.

The expression "pro-form" means a form of the protein which should or may be converted into the desired protein by processing. It may be a natural pro-form or prepro-form of the protein, but also a synthetic pro-form which is the result of a recombinant DNA construct in which the gene coding for the desired protein is preceded by an added signal or leader sequence.

As regards the substrates to be processed, generally speaking proteins with paired basic amino ^' acid residues can serve as a substrate. The additional presence of a basic amino

acid residue in the -4 position relative to the cleavage site (i.e., .4 positions before the cleavage site) will lead to higher efficiency. The following examples are mentioned as possible substrates for processing by furin without completeness being pretended: precursors encoded by the transforming growth factor β (TGF-β) gene family of growth and differentiation factors (such as TGF-βl, TGF-β2, TGF-β3, TGF-β4, TGF-β5, activiv, inhibin, Xenopus laevis Vgl gene product, Mullerian Inhibiting Substance [MIS], decapenta- peptide gene complex of Drosophila embryos and bone morpho- genetic protein, see Sporn and Roberts, Anal. N.Y. Acad. Sci. 593, 1990, 1-6) , precursors of growth factors, such as β-Nerve Growth Factor (β-NGF) and insulin, precursors of clotting factors, such as von Willebrand Factor, Protein C, Factor IX ^■ and Factor X, hormones and neuropeptides, such as Proopio- melanocortin, Proe'nkephalin, Prodynorphin, Provasopressin, Prooxytocin, ProCRF (corticotropin releasing factor) , ProGRF (growth hormone releasing factor) , Prosomatostatin, Proglucagon, Procalcitonin, ProCGRP (calcitonin gene-related peptide) , ProVIP (vasoactive intestinal peptide) , Procaerulin and ProELH (egg laying hormone) , interleukins, interferons, and hematopoietic factors .

The invention can also be applied to proteins which do not by themselves require endoproteolytic processing. Examples are gene constructs in which, for reasons of good processing (glycosylation) or ready purification (secretion) a sequence encoding for the desired protein is coupled to a suitable signal sequence, as has been proposed earlier, for example, for the production of erythropoietin in yeast cells by Elliott et al., Gene 2 , 1989, 167-180. In that publication, a gene construct is described from the leader region of prepro-alpha factor placed before the erythropietin sequence. The processing of the resulting synthetic precursor is effected in the yeast cells by the KEX2 gene product present therein. A further illustration of the invention will be given with reference to the accompanying drawings, in which

Fig. 1 shows the amino acid sequence (in the one-letter code) of the purin consisting of 794 amino acids;

Fig. 2 shows diagrammatically the furin gene, cDNA and protein. a. Genomic organization of part of the fur gene. Exon 1 (about 120 bp) is located at 7.2 kb upstream of exon 2. The asterisk above exon 2 indicates the position of the initiation codon, and the arrow head above exon 16 the stop codon. Non- coding sequences are represented by black boxes. B=BamHI; E=E£o_RI; K=KpnI; S=SalI; P=PstI; X=X___.I . b. Schematic distribution of exons in the cDNA of fur. c. The putative localization of the various protein domains in furin. The largest ^' exon (exon 16) encodes nearly the entire Cys-rich domain, the transmembrane domain and the cytoplasmic domain. The exons 2-12 encode the presumptive prepro and catalytic domains, with codons for the active site residues Asp46 (D) , His87 (H) and Ser261 (S) in exons 5, 7 and 10, respectively. This intron/exon distribution is of the same degree of complexity as observed in the trypsin family of serine proteases. Vertical arrows indicate pairs of basic residues (Arg-Arg, Lys-Arg) which are potential autoprocessing sites; the pairs of basic amino acid residues Arg310-Lys311 and Arg341-Lys342 are possibly involved in proteolytic cleavage. The N-terminus of the mature protein is assumed to begin at amino acid residue 108 directly behind the triplet of potential cleavage sites (Lys-Arg-Arg-Thr-Lys-Arg) because an arginine residue (Argl04) at the -4 position relative to the proposed cleavage site has been found to enhance cleavage efficiency. The regions in which the amino acid sequences of furin, Kexl (Kluyveromyces lactis) and Kex2 (Saccharomyces cerevisiae) exhibit similarity include parts of the prepro domain, the entire catalytic domain (47% identity in 322 residues) and the entire middle domain (26-31% identity in 138 residues) . There is no significant similarity in the

transmembrane and cytoplasmic domains, while the Cys-rich domain is not present in the two yeast proteins .

Fig. 3 contains a comparison of the amino acid sequences of (hfur) human furin, (kexl) Kexl protease, (kex2) Kex2 protease, (ther) thermitase, (subC) subtilisin Carlsberg, and (subB) subtilisin BPN' ..

On the right-hand side, the numbering of the amino acid residues from the putative N-terminus of the mature enzymes is given, and for furin also along the top. Probably, furin has a prepro segment of 107 residues terminating with the sequence Lys-Arg-Arg-Thr-Lys-Arg, which has three potential cleavage sites for auto-activation.

(V) identical residues and ( . ) conservative substitutions in all six sequences; ( : ) identical residues in at least four sequences. Pairs of basic residues in furin, Kexl and Kex2 are indicated in lower case letters . The sequence alignment is taken from a multiple alignment of more than 20 members of the subtilisin family of serine proteases and a superposition of the three-dimensional structures of thermitase, subtilisin Carlsberg and subtilisin BPN' determined by X-ray crystallography. This superposition of three-dimensional structures leads to an extended consensus core, as shown by solid bars, with distances" between topologically equivalent Cα atoms of less than 1.5 A. Secondary structural elements common to all three proteins are indicated as (α) α-helix, (β) β-sheet,

(t) β-turn and (s) bend.

Residues known to be involved in substrate or inhibitor binding in thermitase, subtilisin Carlsberg and subtilisin

BPN' through main-chain or side-chain interactions are marked with asterisk. Essential residues of the active site (D, H and

S) and of the oxyanion hole (N) are underlined. The loops corresponding to the strongest Ca ion binding sites in thermitase are indicated by <==Ca==>.

Boundaries of exons encoding sequences of the presumptive catalytic domain of furin are located behind residues 17, 60,

86, 115, 173, 244, 278 311 and 352.

Fig. 4 shows a schematic model of the catalytic domain of furin. The model is based on a ribbon drawing of subtilisin. The active site, consisting of the residues Asp46, His87 and Ser261, is situated at the top centre. The C-terminal extension (dashed) containing additional domains begins at the opposite side of the catalytic domain.

Predicted positions of 8 short inserts (solid black) , including an extended N terminus, relative to subtilisin, are seen to be situated in surface loops and in connections between conserved α-helix and β-sheet secondary structural elements.

Predicted positions of two stabilizing calcium ions, Cal and Ca2 as in thermitase, are indicated by hatched spheres in the external loops 98-105 and 68-77, respectively. All of the side chain carboxyl groups required for the coordination of these two calcium ions as in thermitase are also present in furin in topologically equivalent residues in these loops; in addition, Asp8 and Asp55 are present to coordinate Cal.as in thermitase and subtilisin, where in topologically equivalent positions either Gin or Asp are the ligands .

Predicted disulfide bridges Cysl04-Cys253 and Cysl96- Cys226 (or Cysl98-Cys226) are shown in dotted lines.

Negatively charged side chain groups on the substrate binding site (top) of the furin molecule are shown as forked stalks and correspond to residues 46, 47, 84, 121, 123, 126 150, 151, 152, 192, 194, 199, 241, 248 and 255. Most of these charges are not present in equivalent positions in subtilisins and thermitase. Many of these negatively charged residues could interact directly with paired basic residues in the substrate, as they are probably located in or near the PI and P2 binding pockets for lysine and/or arginine.

The model described for the catalytic domain of furin also applies to the Kexl and Kex2 proteases since essentially all of the important elements described above are present in all three proteins.

Fig. 5 shows diagrammatically the structural organization of prepro-vWF of wild type von Willebrand Factor (top part) and prepro-vWFgly763 of the mutant vWFgly763 (lower part) . Internal homologous domains are indicated by open boxes; Al, A2 and A3 represent a triplicated domain; B embodies the homologous domains Bl, B2 and B3; Cl and C2 represent a duplicated domain; Dl, D2, D3 and D4 represent four repeated domains and D ¹ represents a partly duplicated domain. The solid line indicates the remaining amino acid sequences . The aminoterminal part contains a signal peptide of 22 amino acid residues. The cleavage site after arginine residue at position 763, which consists, of a pair of basic amino acid residues, is marked with an arrow. The nucleotide sequence of the DNA region around the cleavage site is given, and the deduced amino acid sequence presented in one-letter notation. The point mutation in pro-vWFgly 763 is marked with an asterisk.

There will now be given an example of a process according to the invention in which the endoproteolytic activity of furin is used in the processing of the precursor of the von Willebrand Factor (pro-vWF) as a substrate. With regard to the structure of prepro, pro and mature vWF, reference is made to Verweij et al. , EMBO J. J5, 1986, 1839-1847, and Verweij et al., J.Biol.Chem. 263, 1988, 7921-7924. Pro-vWF consists of a pro-polypeptide (741 amino acid residues) and, at the C terminus, mature vWF (2050 amino acid residues) . As explained in the above publications, mature vWF is formed from pro-vWF by proteolytic processing next to the paired basic amino acids Lys762-Arg763, and COS-1 cells are a suitable host for the synthesis of constitutively secreted vWF after transfection of full-length prepro-vWF cDNA. The activity of furin in endo¬ proteolytic processing was tested for both pro-vWF and the mutant pro-vWFgly763 described by Voorberg et al., EMBO J. j9, 1990, 797-803. The DNA coding for this mutant pro-vWFgly763 contains a guanosine instead of an adenosine in the 2407 position of full-length prepro-vWF cDNA. As a result of this mutation, the cleavage site Lys762-Arg763 of the

propolypeptide is replaced by Lys762-Gly763 in the pro-vWF precursor protein mutant .

EXAMPLES

Identification of translation products encoded by full length fur cDNA transfected into COS-1 cells

To further characterize the fur gene product furin, experiments were performed to synthesize this protein in eukaryotic cells under the control of the SV40 late promotor and to use this material in an approach to elucidate its function. To identify the translation products of the fur gene, an immunological approach was selected. A polyclonal antiserum raised in rabbits to a recombinant furin hybrid protein as described in "Materials and methods" was used. In Western blot analysis of proteins in total lysates of bacteria transformed with pMJ109, pMJ119 or pEWl DNA, the polyclonal antiserum recognized β-gal-Δfurinl, 336trpE-AS-Δfurinl and GST- Δfurin2, respectively. In control experiments, the antiserum did not react with the trpE-encoded polypeptide chain of anthralinate synthetase or glutathion-S-transferase. Using this antiserum, Western blot analysis was performed to detect fur gene-encoded proteins in COS-1 cells transfected with pSVLfur. Based upon nucleotide sequence data of fur cDNA, synthesis of a primary translation product with a calculated molecular weight of 87 kDa may be expected. Two proteins with apparent molecular weights of about 90 kDa and 100 kDa, respectively, were detected in transfected COS-1 cells as compared to non-transfected COS-1 cells (results not shown) . The presence of two forms of furin in COS-1 cells transfected with pSVLfur DNA indicates that furin is subject to post-translational modification. It is possible, that the 100 kDa protein is a glycosylat ^' ed form of the primary product of about 90 kDa. However, it is also tempting to speculate that the 100 kDa polypeptide would represent the pro-form,

while the 90 kDa polypeptide represents mature furin, generated by proteolytic (auto)processing of the peptide bond between residues 107 and 108.

It is noted that non-transfected COS-1 cells also contain small amounts of immunoreactive material with apparent molecular weights of 90, 60 and 40 kDa. Although the identity of these proteins remains to be established, it is conceivable that the 90 kDa protein represents a low amount of endogenous furin. The data indicate that the transfected fur genetic sequences are indeed transcribed and translated making it possible to test the biological function of the fur products .

Proprotein processing activity of furin

In transfection experiments with 10μg pSVLvWF DNA, the 360 kDa pro-vWF precursor protein and the 260 kDa mature vWF protein were found in virtually equal proportions in the conditioned medium. The formation of mature vWF in the cells obtained by transfection is attributed to a processing by endogenous furin, which is expressed in COS-1 cells as shown by Northern blot analysis of mRNA i-solated from COS-1 cells and by immuno-precipitation analysis with a polyclonal rabbit anti-furin serum. In similar transfection experiments of COS-1 cells with lOμg pSVLvWFgly763 DNA, it was found that pro-vWFgly763 was formed and secreted constitutively in the culture medium as a 360 kDa protein. There was no endoproteolytic processing to mature vWF (260 kDa) . The same result was found when a cotransfection of 5μg pSVLvWFgly763 DNA and 5μg pSVLfur DNA was carried out. No processing of pro-vWFgly763 to mature vWF was observed.

On the other hand, when COS-1 cells were cotransfected with 5μg pSVLvWF DNA and 5μg pSVLfur DNA, a complete processing of pro-vWF to mature vWF was found.

Materials and methods

Molecular cloning

pSVLfur contains a 4.1 kb full-length fur cDNA fragment, starting 117 nucleotides upstream of the ATG start codon and ending 21 nucleotides downstream of the poly-A addition site cloned into the EcoRI site of pSVL (Wells et al . , Nucl. Acids Res. 11, 1983, 7911-7925) . In pSVLfur, expression of the fur cDNA sequences is under control of the SV40 late promotor. pMJ109 consists of a 2.2 kb Smal/Smal human fur cDNA fragment molecularly cloned into the Hindlll site of plasmid pUR291; the 2.2 kb fur cDNA encompasses the carboxyterminal region of furin, which in pMJ109 is fused in phase to the β-galactosidase (β-gal) encoding sequences using the polylinker region constructed just in front of the stop codon in lacZ . pMJ119 consists of the same 2.2 kb fur cDNA fragment but here molecularly cloned into the Sm l site of plasmid pATHl, which results in the in phase fusion of the furin sequences to the first 336 amino acid residues of the trpE-encoded portion of anthranilate synthetase (336trpE-AS) . Finally, in case of pEWl, a 3.5 kb Bglll/EcoRI fur cDNA fragment is cloned into pGEX-3X and fused in phase to glutathion-S-transferase (GST) . Upon proper induction of protein synthesis in bacteria transformed with pMJ109, pMJ119 or pEWl, production of relatively large quantities of β-gal-Δfurinl (MW 170 kDa) , 336trpE-AS-Δfurinl (MW 90 kDa) and GST-Δfurin2 (MW 100 kDa) , respectively, was observed.

Preparation of polyclonal anti-furin antibodies and im unoblotting

Polyclonal anti-furin antibodies were raised in rabbits to the β-gal-Δfurinl hybrid protein synthesized in bacteria transformed by pMJ109. For immunizations, partially purified hybrid protein preparations were used. Upon size fractionation

of the bacterial proteins by SDS-PAGE, the gel region containing the hybrid protein was excised and the protein content removed electrophoretically. Western blotting experiments with extracts of transfected COS-1 cells were performed as follows. COS-1 cells transfected with 10 μg of pSVLfur DNA were maintained 48 hr post-transfeetion in serum- free medium. At this time, the cells were washed twice with 10 mM sodium phosphate (pH 7.4), 0.14 M NaCl and then lyzed in "immunoprecipitation buffer (IPB) " consisting of 10 mM Tris- HCl (pH- 7.8), 150 mM NaCl, 5 mM EDTA, 1% (v/v) Nonidet P-40, 10 mM benzamidine, 5 mM N-ethylmaleimide and 1 mM phenylmethylsulfonyl fluoride (PMSF) . An aliquot of the cell extract was run under reducing conditions on a 8% (w/v) SDS- polyacrylamide gel and-, subsequently, proteins were transferred to nitrocellulose (Schleicher and Schuell) .

Detection of furin was performed by incubating the blot with the rabbit anti-furin serum described above.

DNA transfection, radiolabeling of cells and immuno- precipitation analysis

Monkey kidney COS-1 cells were propagated in Iscove's modified minimal medium, supplemented with fetal calf serum (10% v/v) and antibiotics [penicillin (100 U/ml) and streptomycin (100 μg/ml) ] . Twenty four hours upon seeding, the semi-confluent cells were transfected with 20 μg of DNA in 2 ml of Iscove's modified minimal medium, supplemented with 200 μg/ml DEAE-dextran. The transfection procedure used included a chloroquine shock (Luthman and Magnusson, Nucl. Acids Res. __1 , 1983, 1295-1308) . After transfection, cells were maintained in the medium described above for 48 hours. Prior to radiolabelling, medium was removed and the cells incubated for 1 h in RPMI medium, lacking methionine . Subsequently, cells were labelled for 4 hours in the presence of [ ³⁵ S]methionine (50 μCi/ml, specific activity > 800

Ci/mmol) , followed by a chase of 14 hours with unlabelled

methionine (final concentration 1 mM) . After centrifugation for 5 min at 13,000 x g, the labelled culture media were adjusted to 1 x IPB. Preclearance of the media was performed by incubating twice with gelatin-Sepharose and, subsequently, with preformed complexes of rabbit pre-immune serum with

Protein A-Sepharose. Immunoprecipitation of radiolabelled vWF- related proteins was carried out by preformed complexes of an IgG preparation, derived from rabbit anti-vWF (Dakopatts, Glostrup, Denmark) with Protein A-Sepharose. Immunoprecipitates were extensively washed with IPB and pelleted through a discontinuous 10-20% (w/v) sucrose gradient dissolved in IPB supplemented with 0.5% desoxycholate and 10 mM Tris-HCl (pH 7.8), respectively. Immunoprecipitates were analysed under reducing conditions on a 5% SDS-polyacrylamide gel.