Furin: the prototype mammalian subtilisin-like proprotein-processing enzyme. Endoproteolytic cleavage at paired basic residues of proproteins of the eukaryotic secretory pathway.

Furin, the translational product of the recently discovered fur gene, appears to be the first known mammalian member of the subtilisin family of serine proteases and the first known mammalian proprotein-processing enzyme with cleavage selectivity for paired basic amino acid residues. Structurally and functionally, it resembles the prohormone-processing enzyme, kexin (EC 3.4.21.61), which is encoded by the KEX2 gene of yeast Saccharomyces cerevisiae. Most likely, furin is primarily involved in the processing of precursors of proteins that are secreted via the constitutive secretory pathway. Here, we review the discovery of the fur gene and describe the isolation of cDNA clones corresponding to human and mouse fur and to two fur-like genes of Drosophila melanogaster, Dfur1 and Dfur2. We also compare the structural organization of the various deduced furin proteins to that of yeast kexin, and of other members of the subtilisin family of serine proteases. Furthermore, the biosynthesis of biologically active human and mouse furin is evaluated. Finally, the cleavage specificity for paired basic amino acid residues of human and mouse furin is demonstrated by the correct processing of the precursor for von Willebrand factor.     

Molecular characterization of KEX1, a kexin-like protease in mouse Pneumocystis carinii

Expression screening of a Pneumocystis carinii-infected mouse lung cDNA library with specific monoclonal antibodies (mAbs) led to the identification of a P. carinii cDNA with extensive homology to subtilisin-like proteases, particularly fungal kexins and mammalian prohormone convertases. The 3.1 kb cDNA contains a single open reading frame encoding 1011 amino acids. Structural similarities to fungal kexins in the deduced primary amino acid sequence include a putative proenzyme domain delineated by a consensus autocatalytic cleavage site (Arg-Glu-Lys-Arg), conserved Asp, His, Asn and Ser residues in the putative catalytic domain, a hydrophobic transmembrane spanning domain, and a carboxy-terminal cytoplasmic domain with a conserved tyrosine motif thought to be important for localization of the protease in the endoplasmic reticulum and/or Golgi apparatus. Based on these structural similarities and the classification of P. carinii as a fungus, the protease was named KEX1. Southern blotting of mouse P. carinii chromosomes localized kex1 to a single chromosome of approximately 610 kb. Southern blotting of restriction enzyme digests of genomic DNA from P. carinii-infected mouse lung demonstrated that kex1 is a single copy gene. The function of kexins in other fungi suggests that KEX1 may be involved in the post-translational processing and maturation of other P. carinii proteins.     

Yeast KEX2 genes encodes an endopeptidase homologous to subtilisin-like serine proteases

Yeast Saccharomyces cerevisiae KEX2 gene previously isolated, was characterized as the gene encoding a calcium-dependent endopeptidase required for processing of precursors of alpha-factor and killer toxin. In this study, we report the amino acid sequence of the KEX2 gene product deduced from nucleotide sequencing. Our results indicate that the KEX2 gene contains a 2,442-bp open reading frame encoding a polypeptide of 814 amino acids. The deduced amino acid sequence contains a region extensively homologous to the members of subtilisin-like serine protease family near the N-terminus. A putative membrane-spanning domain near the C-terminus was also detected. These facts indicate that the KEX2-encoded protein may function as a membrane-bound, subtilisin-like serine protease.     

Cloning and functional expression of Dfurin2, a subtilisin-like proprotein processing enzyme of Drosophila melanogaster with multiple repeats of a cysteine motif.

Production of a variety of regulatory eukaryotic proteins, such as growth factors and polypeptide hormones, often involves endoproteolytic processing of proproteins at cleavage sites consisting of paired basic residues. The first known mammalian proprotein processing enzyme with such specificity is the human fur gene product furin. Structurally and functionally, furin is related to the subtilisin-like serine endoprotease kexin (EC 3.4.21.61) of yeast Saccharomyces cerevisiae; unlike kexin, it contains a cysteine-rich region with an unknown function. Here, we describe cloning and sequencing of a 5.8-kbp cDNA of the Dfur2 gene, a fur-like gene of Drosophila melanogaster, which we found expressed during various stages of development. This Dfur2 cDNA has an open reading frame for a 1680-residue protein, called Dfurin2. Dfurin2 contains similar protein domains as mammalian furin, however, it has an extended amino-terminal region and its cysteine-rich region is much larger than that of mammalian furin. Because of this latter phenomenon, we were able to identify a particular cysteine motif that was repeated multiple times in Dfurin2 but present only twice in mammalian furin. Furthermore, we show that Dfur2 encodes an endoproteolytic enzyme with specificity for paired basic amino acid residues as, in cotransfection experiments, correct cleavage was demonstrated of the precursor of the von Willebrand factor but not of a cleavage mutant. Finally, Dfur2 was mapped to region 14C of the X chromosome of D. melanogaster.     

Prohormone processing by yeast proteases.

Investigations of the precursors of alpha-pheromone and killer toxin in the yeast Saccharomyces cerevisiae have defined the genes coding (KEX1 and KEX2) for the proteases which are responsible for their processing. In addition to processing at pairs of basic residues it is evident that yeast can also process at monobasic sites. We present data on the Kex1p and Kex2p enzymes, their cellular localization, and their post-translational modification. In addition initial characterisation of the monobasic specific protease and the isolation of mutants defective in this activity are presented. The use of the yeast system as a model for the processing of mammalian prohormones is discussed.     

Human fur gene encodes a yeast KEX2-like endoprotease that cleaves pro- beta-NGF in vivo

Extracts from BSC-40 cells infected with vaccinia recombinants expressing either the yeast KEX2 prohormone endoprotease or a human structural homologue (fur gene product) contained an elevated level of a membrane-associated endoproteolytic activity that could cleave at pairs of basic amino acids (-LysArg- and -ArgArg-). The fur-directed activity (furin) shared many properties with Kex2p including activity at pH 7.3 and a requirement for calcium. By using antifurin antibodies, immunoblot analysis detected two furin translation products (90 and 96 kD), while immunofluorescence indicated localization to the Golgi apparatus. Coexpression of either Kex2p or furin with the mouse beta- nerve growth factor precursor (pro-beta-NGF) resulted in greatly enhanced conversion of the precursor to mature nerve growth factor. Thus, the sequence homology shared by furin and the yeast KEX2 prohormone processing enzyme is reflected by significant functional homology both in vitro and in vivo.     

The gene encoding DRAP (BACE2), a glycosylated transmembrane protein of the aspartic protease family, maps to the down critical region

We applied cDNA selection methods to a genomic clone (YAC 761B5) from chromosome 21 located in the so-called 'Down critical region' in 21q22.3. Starting from human fetal heart and brain mRNAs we obtained and sequenced several cDNA clones. One of these clones (Down region aspartic protease (DRAP), named also BACE2 according to the gene nomenclature) revealed a striking nucleotide and amino acid sequence identity with several motifs present in members of the aspartic protease family. In particular the amino acid sequences comprising the two catalytic sites found in all mammalian aspartic proteases are perfectly conserved. Interestingly, the predicted protein shows a typical membrane spanning region; this is at variance with most other known aspartic proteases, which are soluble molecules. We present preliminary evidence, on the basis of in vitro translation studies and cell transfection, that this gene encodes a glycosylated protein which localizes mainly intracellularly but to some extent also to the plasma membrane. Furthermore DRAP/BACE2 shares a high homology with a newly described beta-secretase enzyme (BACE-1) which is a transmembrane aspartic protease. The implications of this finding for Down syndrome are discussed.     

Reduced proteolysis of secreted gelatin and Yps1-mediated alpha-factor leader processing in a Pichia pastoris kex2 disruptant

Heterologous proteins secreted by yeast and fungal expression hosts are occasionally degraded at basic amino acids. We cloned Pichia pastoris homologs of the Saccharomyces cerevisiae basic residue-specific endoproteases Kex2 and Yps1 to evaluate their involvement in the degradation of a secreted mammalian gelatin. Disruption of the P. pastoris KEX2 gene prevented proteolysis of the foreign protein at specific monoarginylic sites. The S. cerevisiae alpha-factor preproleader used to direct high-level gelatin secretion was correctly processed at its dibasic site in the absence of the prototypical proprotein convertase Kex2. Disruption of the YPS1 gene had no effect on gelatin degradation or processing of the alpha-factor propeptide. When both the KEX2 and YPS1 genes were disrupted, correct precursor maturation no longer occurred. The different substrate specificities of both proteases and their mutual redundancy for propeptide processing indicate that P. pastoris kex2 and yps1 single-gene disruptants can be used for the alpha-factor leader-directed secretion of heterologous proteins otherwise degraded at basic residues.     

Characterization of the yeast KEX1 gene product: a carboxypeptidase involved in processing secreted precursor proteins.

We have identified and partially characterized the Saccharomyces cerevisiae KEX1 gene product, Kex1p, to assess its role in processing secreted protein precursors. Anti-Kex1p antibodies identified a 113-kilodalton protein that was absent in cells in which the KEX1 gene has been disrupted and that was more abundant in cells overexpressing the KEX1 gene. Kex1p was found to be a membrane-associated glycoprotein with N-linked carbohydrate. The N-linked oligosaccharide(s) was modified in a progressive manner after synthesis, causing the glycoprotein to slowly increase in mass to 115 kilodaltons. After a Kex2p-mediated cleavage event at specific pairs of basic amino acids, alpha-factor and K1 killer toxin precursors have COOH-terminal dibasic residue extensions and require a carboxypeptidase B-like enzyme to process the precursors to maturity. A carboxypeptidase activity, with apparent specificity for basic amino acids, was detected in KEX1 cells. Disruption of the KEX1 gene abolished this activity, while overexpression of KEX1 increased it. Our results provide biochemical evidence consistent with earlier genetic work, that KEX1 encodes a serine carboxypeptidase involved in the processing of precursors to secreted mature proteins.     

Nucleotide sequence of cloned cDNA for human pancreatic kallikrein

Cloned cDNA sequences for human pancreatic kallikrein have been isolated and determined by molecular cloning and sequence analysis. The identity between human pancreatic and urinary kallikreins is indicated by the complete coincidence between the amino acid sequence deduced from the cloned cDNA sequence and that reported partially for urinary kallikrein. The active enzyme form of the human pancreatic kallikrein consists of 238 amino acids and is preceded by a signal peptide and a profragment of 24 amino acids. A sequence comparison of this with other mammalian kallikreins indicates that key amino acid residues required for both serine protease activity and kallikrein-like cleavage specificity are retained in the human sequence, and residues corresponding to some external loops of the kallikrein diverge from other kallikreins. Analyses by RNA blot hybridization, primer extension, and S1 nuclease mapping indicate that the pancreatic kallikrein mRNA is also expressed in the kidney and sublingual gland, suggesting the active synthesis of urinary kallikrein in these tissues. Furthermore, the tissue-specific regulation of the expression of the members of the human kallikrein gene family has been discussed.     

Secretion of biologically active porcine prophospholipase A2 by Saccharomyces cerevisiae. Use of the prepro sequence of the alpha-mating factor

The cDNA coding for porcine pancreatic prophospholipase A2 (proPLA) has been cloned and expressed in Saccharomyces cerevisiae. Expression and secretion of proPLA could only be obtained after fusing the proPLA to the prepro sequence of the yeast alpha-mating factor. Upon secretion, the fusion protein was cleaved by the KEX2 protease yielding a 140-amino-acid zymogen-like form of the phospholipase A2. This protein was purified in high yield by ion-exchange chromatography. Limited proteolysis with trypsin cleaved the 'zymogen' to yield active phospholipase A2, which was indistinguishable from the authentic porcine pancreatic enzyme. These results show that a protein with a disulphide bridge content as high as 7 per 124 amino acid residues can be correctly processed by the yeast secretory apparatus.     

Molecular cloning and expression of a G25K cDNA, the human homolog of the yeast cell cycle gene CDC42.

G25K is a low-molecular-mass GTP-binding protein with a broad distribution in mammalian tissues. A cDNA clone was isolated by using oligonucleotides corresponding to the partial amino acid sequence of purified human G25K. The cDNA encodes an 191-amino-acid polypeptide containing GTP-binding consensus sequences and a putative farnesylation site at the C terminus. The sequence exhibits 50 and 70% identities to the mammalian rho and rac proteins, respectively, and an 80% identity to the Saccharomyces cerevisiae CDC42 gene product. Insect Sf9 cells infected with recombinant baculovirus vectors expressing the G25K cDNA produced a 25-kDa protein that bound GTP and was recognized by antibodies specifically reactive to G25K. G25K appears to be the human homolog of the CDC42 gene product, since expression of the G25K cDNA in S. cerevisiae suppressed both cdc42-1 and cdc24-4 temperature-sensitive lethal mutations.     

Yeast KEX1 protease cleaves a prohormone processing intermediate in mammalian cells

A vaccinia virus vector was used to express the yeast KEX1 gene, which encodes a prohormone carboxypeptidase specific for the removal of basic amino acids from prohormone processing intermediates, in mammalian cells. When produced in BSC-40 cells, Kex1p was localized to the perinuclear region and conferred a large increase in enzymatic activity characteristic of this carboxypeptidase. Expression of the KEX1 gene together with the yeast KEX2 gene, which encodes a prohormone endopeptidase specific for cleavage at pairs of basic amino acids, and the mouse proopiomelanocortin (mPOMC) cDNA in BSC-40 cells resulted in the full conversion of mPOMC to mature peptides including gamma-lipotropin. This in vivo processing of mPOMC to mature peptides by the KEX2/KEX1 gene products demonstrates a significant functional homology of the basic prohormone processing machinery in yeast and neuroendocrine cells.     

Amino acid sequence of endothiapepsin. Complete primary structure of the aspartic protease from Endothia parasitica

The amino acid sequence of endothiapepsin, the aspartic protease from Endothia parasitica has been determined. The enzyme consists of 330 residues. The sequence determination was performed exclusively at the protein level. The homology of this fungal milk-clotting enzyme with aspartic proteases is demonstrated by alignment with pepsin, chymosin, gastricsin, renin, and cathepsin D from various vertebrates and proteinase A from Saccharomyces cerevisiae showing 25-30% identity. The identity with mucor rennin from Mucor pucillus was 21% and with penicillopepsin from Penicillium janthinellum 53%, the fungal enzymes thus representing the lowest as well as the highest degree of homology.     

Determination of the carboxyl termini of the alpha and beta subunits of yeast K1 killer toxin. Requirement of a carboxypeptidase B-like activity for maturation

The carboxyl-terminal sequences of the two polypeptide chains of the Saccharomyces cerevisiae K1 killer toxin were determined by protein sequencing and amino acid analysis of peptide fragments generated from the mature, secreted toxin. The COOH-terminal amino acid of the beta chain is histidine 316, the final residue encoded by the precursor gene. The COOH terminus of the alpha chain is at alanine 147 of the preprotoxin. Amino acid composition data for the purified toxin are consistent with that predicted from the gene sequence of the preprotoxin where the alpha and beta subunits consist of amino acid residues 45-147 and 234-316, respectively. The molecular weight of the mature alpha beta dimer is about 20,658. The COOH-terminal sequence determination completes the location of the toxin subunits in the precursor, and its configuration may be represented as prepropeptide-Pro-Arg-alpha-Arg-Arg-gamma-Lys-Arg-beta, where gamma represents the interstitial glycosylated peptide. The COOH terminal side of the paired basic residues (Arg-148 Arg-149 and Lys-232 Arg-233 of preprotoxin) are endoproteolytic processing sites for the product of the KEX2 gene (Julius, D., Brake, A., Blair, L., Kunisawa, R., and Thorner, J. (1984) Cell 37, 1075-1089), and thus maturation of the alpha subunit of killer toxin apparently requires a carboxypeptidase B-like activity. A possible candidate for this activity is the product of the KEX1 gene (Dmochowska, A., Dignard, D., Henning, D., Thomas, D.Y., and Bussey, H. (1987) Cell, in press).     

Recombinant HIV1 protease secreted by Saccharomyces cerevisiae correctly processes myristylated gag polyprotein

The protease of the human immunodeficiency virus type I (HIV1) was expressed both intracellularly and extracellularly in Saccharomyces cerevisiae. Intracellular expression of the protease was achieved by fusing a 179 amino acid precursor form of the protease to human superoxide dismutase (hSOD). Self-processing of the viral enzyme from the hybrid precursor was demonstrated to occur within the yeast host. Secretion of the protease was achieved by fusing the leader sequence of yeast alpha-factor to the precursor form of the protease or to the 99 amino acid mature form of the protease. Authentic and active forms of the retroviral enzyme were detected in yeast supernatants of cells expressing the precursor or the mature form of the protease. A D25E active site variant of the retroviral enzyme exhibited diminished autocatalytic activity when expressed intracellularly or secreted from yeast. The wild-type protease was active in an in vitro assay on the natural substrate, myristylated gag precursor, Pr53gag. Correct processing of Pr53gag at the Tyr 138-Pro 139 junction was confirmed by amino terminal sequence analysis of the resulting capsid protein (CA, p24). The secreted protease was purified to homogeneity from yeast media using preparative isoelectric focusing and reverse-phase HPLC. Amino terminal sequence analysis showed a sequence beginning at amino acid 1 of the mature enzyme (Pro) and another sequence beginning at amino acid 6 (Trp). This shorter sequence may represent a natural autolytic product of the protease.